Food Flavour Technology

P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth November 25, 2009 15:44 Printer Name: Yet to Come P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

Second Edition

Edited by Andrew J. Taylor and Robert S.T. Linforth Division of Food Sciences, University of Nottingham, UK

A John Wiley & Sons, Ltd., Publication P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

This edition ﬁrst published 2010 C 2010 Blackwell Publishing Ltd

Blackwell Publishing was acquired by John Wiley & Sons in February 2007. Blackwell’s publishing programme has been merged with Wiley’s global Scientiﬁc, Technical, and Medical business to form Wiley-Blackwell.

Registered ofﬁce John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom

Editorial ofﬁces 9600 Garsington Road, Oxford, OX4 2DQ, United Kingdom 2121 State Avenue, Ames, Iowa 50014-8300, USA

For details of our global editorial ofﬁces, for customer services and for information about how to apply for permission to reuse the copyright material in this book please see our website at www.wiley.com/wiley-blackwell.

The right of the author to be identiﬁed as the author of this work has been asserted in accordance with the Copyright, Designs and Patents Act 1988.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books.

Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought.

Library of Congress Cataloging-in-Publication Data

Food ﬂavour technology / edited by Andrew J. Taylor and Robert S.T. Linforth. Ð 2nd ed. p. cm. Includes bibliographical references and index. ISBN 978-1-4051-8543-1 (hardback : alk. paper) 1. Flavour. 2. Flavouring essences. 3. FlavourÐAnalysis. I. Taylor, A. J. (Andrew John), 1951- II. Linforth, Robert S. T. TP418.F65 2010 664.07Ðdc22 2009028000

A catalogue record for this book is available from the British Library.

Set in 10/12 pt Times by Aptara R Inc., New Delhi, India Printed in Singapore

1 2010 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

Contents

List of contributors xi Preface xiii

1 Creating and formulating flavours 1 John Wright 1.1 Introduction 1 1.1.1 A little history 1 1.2 Interpreting analyses 2 1.3 Flavour characteristics 3 1.3.1 Primary characters 3 1.3.2 Secondary characteristics 4 1.3.3 Taste effects 5 1.3.4 Complexity 6 1.3.5 Flavour balance 6 1.3.6 Unfinished work 7 1.4 Applications 8 1.4.1 Ingredient factors 8 1.4.2 Processing factors 10 1.4.3 Storage factors 10 1.4.4 Consumption factors 11 1.5 Flavour forms 11 1.5.1 Water-soluble liquid flavours 11 1.5.2 Clear water-soluble liquid flavours 12 1.5.3 Oil-soluble liquid flavours 13 1.5.4 Emulsion-based flavours 13 1.5.5 Dispersed flavours 13 1.5.6 Spray-dried flavours 14 1.6 Production issues 15 1.7 Regulatory affairs 16 1.8 A typical flavour 16 1.9 Commercial considerations 19 1.9.1 International tastes 19 1.9.2 Abstract flavours 20 1.9.3 Matching 21 1.9.4 Customers 22 1.10 Summary 22 References 23 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

iv Contents

2 Flavour legislation 24 Jack Knights 2.1 Introduction 24 2.2 Methods of legislation 24 2.3 Legislation in the United States 26 2.4 International situation: JECFA 27 2.5 Council of Europe 28 2.6 European community 30 2.6.1 Background Ð national to EU legislation 30 2.6.2 The 1988 Council Directive 31 2.6.3 Smoke ﬂavourings 2003 Directive 40 2.6.4 Developments 2008 onwards 41 2.7 Current EU Situation and the future 47 References 48

3 Basic chemistry and process conditions for reaction flavours with particular focus on Maillard-type reactions 51 Josef Kerler, Chris Winkel, Tomas Davidek and Imre Blank 3.1 Introduction 51 3.2 General aspects of the Maillard reaction cascade 51 3.2.1 Intermediates as flavour precursors 54 3.2.2 Carbohydrate fragmentation 58 3.2.3 Strecker degradation 61 3.2.4 Interactions with lipids 62 3.3 Important aroma compounds derived from Maillard reaction in food and process flavours 65 3.3.1 Character-impact compounds of thermally treated foods 65 3.3.2 Character-impact compounds of process flavours 70 3.4 Preparation of process flavours 74 3.4.1 General aspects 74 3.4.2 Factors influencing flavour formation 74 3.4.3 Savoury process flavours 78 3.4.4 Sweet process flavours 80 3.5 Outlook 80 References 81

4 Biotechnological ﬂavour generation 89 Ralf G. Berger, Ulrich Krings and Holger Zorn 4.1 Introduction 89 4.2 Natural ﬂavours: market situation and driving forces 89 4.3 Advantages of biocatalysis 90 4.4 Micro-organisms 91 4.4.1 Biotransformation and bioconversion of monoterpenes 91 4.4.2 Bioconversion of C13-norisoprenoids and sesquiterpenes 95 4.4.3 Generation of oxygen heterocycles 96 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

Contents v

4.4.4 Generation of vanillin, benzaldehyde and benzoic compounds 97 4.4.5 Generation of miscellaneous compounds 99 4.5 Enzyme technology 101 4.5.1 Liberation of volatiles from bound precursors 101 4.5.2 Biotransformations 101 4.5.3 Kinetic resolution of racemates 103 4.6 Plant catalysts 104 4.6.1 Plant cell, tissue and organ cultures 104 4.6.2 Callus and suspension cultures 105 4.6.3 Organ cultures 105 4.6.4 Plant cell biotransformations 107 4.7 Flavours through genetic engineering 107 4.7.1 Genetically modiﬁed micro-organisms 108 4.7.2 Isolated enzymes from genetically modiﬁed micro-organisms 109 4.7.3 Plant rDNA techniques 110 4.8 Advances in bioprocessing 112 4.8.1 Process developments in microbial and enzyme systems 112 4.8.2 Process developments of plant catalysts 114 4.9 Conclusion 114 References 115

5 Natural sources of flavours 127 Peter S.J. Cheetham 5.1 Introduction 127 5.2 Properties of flavour molecules 129 5.2.1 Flavour perception 129 5.2.2 Differences in sensory character and intensity between isomers 141 5.2.3 Extraction of flavours from plant materials 142 5.2.4 Commercial aspects 146 5.2.5 Economic aspects 147 5.2.6 Safety aspects 147 5.3 Dairy flavours 147 5.3.1 Background 147 5.3.2 Cream and butter 148 5.3.3 Cheese 149 5.4 Fermented products 151 5.4.1 Hydrolysed vegetable proteins 151 5.4.2 Chocolate 152 5.4.3 Tea 153 5.4.4 Coffee 154 5.4.5 Beer 154 5.4.6 Wine 156 5.4.7 Sweeteners 158 5.5 Cereal products 158 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

vi Contents

5.6 Vegetable sources of flavour 159 5.6.1 Spice flavours 159 5.6.2 Mushroom 161 5.6.3 Garlic, onion and related flavours 161 5.6.4 Brassica flavours, including mustard and horseradish 163 5.6.5 ‘Fresh/green/grassy’ 164 5.6.6 Nuts 164 5.6.7 Other vegetables 165 5.6.8 Fermented vegetables 165 5.7 Fruit 165 5.7.1 Apples 166 5.7.2 Pears 167 5.7.3 Grapefruit 167 5.7.4 Blackcurrant 167 5.7.5 Raspberry 168 5.7.6 Strawberry 168 5.7.7 Apricot and peach 169 5.7.8 Tomato 169 5.7.9 Cherry 169 5.7.10 Tropical fruit flavours 170 5.7.11 Vanilla 170 5.7.12 Other fruits 171 5.7.13 Citrus 171 5.7.14 Citrus processing 172 5.8 Other flavour characteristics 174 5.9 Fragrance uses 174 5.10 Conclusion 175 References 175

6 Useful principles to predict the performance of polymeric ﬂavour delivery systems 178 Daniel Benczedi« 6.1 Overview 178 6.2 Introduction 178 6.3 Compatibility and cohesion 179 6.4 Sorption and swelling 182 6.5 Diffusion and release 184 References 187

7 Delivery of flavours from food matrices 190 Saskia M. van Ruth and Jacques P. Roozen 7.1 Introduction 190 7.2 Flavour properties 191 7.3 Thermodynamic aspects of flavour delivery 191 7.3.1 Definition of gas/product partition coefficients and activity coefficients 191 7.3.2 Types of binding 193 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

Contents vii

7.3.3 Lipid–flavour interactions 194 7.3.4 Carbohydrate–flavour interactions 195 7.3.5 Protein–flavour interactions 196 7.4 Kinetic aspects of flavour delivery 197 7.4.1 Principles of interfacial mass transfer 198 7.4.2 Liquid food products 200 7.4.3 Semi-solid food products 200 7.4.4 Solid food products 201 7.5 Delivery systems: food technology applications 202 7.6 Conclusions 203 References 203

8 Modelling flavour release 207 Robert S. T. Linforth 8.1 Introduction 207 8.2 Equilibrium partition models 208 8.2.1 The air/water partition coefficient 208 8.2.2 Estimation of Kaw using QSPR 209 8.2.3 Effect of lipid on volatile partitioning 211 8.2.4 QSPR estimation of the air/emulsion partition coefficient 212 8.2.5 Internet models and databases 213 8.3 Dynamic systems 214 8.3.1 Modelling flavour release from a retronasal aroma simulator 214 8.3.2 Non-equilibrium partition modelling of volatile loss from matrices 215 8.3.3 Modelling the gas-phase dilution of equilibrium headspace 216 8.3.4 Modelling the gas-phase dilution of equilibrium headspace above emulsions 218 8.3.5 Modelling the rate of volatile equilibration in the headspace above emulsions 219 8.4 In vivo consumption 220 8.4.1 Modelling release from emulsions during consumption 222 8.4.2 Effect of gas flow on volatile equilibration above emulsions 222 8.4.3 Modelling volatile transfer through the upper airway 223 8.4.4 Non-equilibrium partition model for in vivo release 223 8.4.5 Modelling flavour release using timeÐintensity data 224 8.4.6 QSPR of in vivo volatile release from gels 224 8.5 Conclusion 226 References 227

9 Instrumental methods of analysis 229 Gary Reineccius 9.1 Analytical challenges 229 9.2 Aroma isolation 231 9.2.1 Aroma isolation methods based on volatility 231 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

viii Contents

9.2.2 Aroma isolation methods using solvent extraction 237 9.2.3 Solid-phase micro-extraction 238 9.2.4 General considerations in preparing aroma isolates 241 9.2.5 Aroma isolation summary 241 9.3 Selection of aroma isolation method 242 9.3.1 ‘Complete’ aroma profile 242 9.3.2 Key components contributing to sensory properties 243 9.3.3 Off-notes in a food product 243 9.3.4 Monitoring aroma changes in foods 244 9.3.5 Using aroma compound profiles to predict sensory response 244 9.3.6 Summary comments on isolation methods 245 9.4 Aroma isolate fractionation prior to analysis 245 9.4.1 Fractionation of concentrates prior to analysis 245 9.5 Flavour analysis by gas chromatography 249 9.5.1 High-resolution gas chromatography 249 9.5.2 Gas chromatographyÐolfactometry 250 9.5.3 Specific gas chromatographic detectors 254 9.6 Flavour analysis by HPLC 254 9.7 Identification of volatile flavours 255 9.7.1 Gas chromatography 255 9.7.2 Infrared spectroscopy 256 9.7.3 Mass spectrometry 257 9.8 Electronic ‘noses’ 261 9.9 Summary 262 References 262

10 On-line monitoring of flavour processes 266 Andrew J. Taylor and Robert S.T. Linforth 10.1 Introduction 266 10.2 Issues associated with in vivo monitoring of flavour release 268 10.2.1 Speed of analysis 268 10.2.2 Analysis of different chemical classes 268 10.2.3 Sensitivity 269 10.2.4 Identification of analysed compounds 270 10.2.5 Interfering factors 270 10.2.6 Non-volatile tastants 270 10.3 Pioneers and development of on-line flavour analysis 271 10.4 On-line aroma analysis using chemical ionisation techniques 272 10.4.1 Analysis via atmospheric pressure chemical ionisation 272 10.4.2 Analysis via PTR 275 10.4.3 Analysis via selected ion flow tube 275 10.4.4 Calibration 276 10.4.5 Suppression 277 10.4.6 Assigning ions to compounds for unequivocal identification 277 10.4.7 Summary 279 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

Contents ix

10.5 Analysis of tastants using direct mass spectrometry 279 10.6 Applications 280 10.6.1 Breath-by-breath analysis 281 10.6.2 Flavour reformulation in reduced fat foods 283 10.6.3 Flavour release in viscous foods 285 10.6.4 Measuring aroma release in ethanolic beverages 285 10.6.5 Monitoring ﬂavour generation on-line 286 10.6.6 Rapid headspace proﬁling of fruits and vegetables 289 10.7 Future 290 References 290

11 Sensory methods of ﬂavour analysis 296 Ann C. Noble and Isabelle Lesschaeve 11.1 Introduction 296 11.2 Analytical tests 296 11.2.1 Discrimination tests 296 11.2.2 Intensity rating tests 298 11.2.3 TimeÐintensity rating 299 11.2.4 TasteÐsmell interactions 301 11.2.5 Descriptive analysis 301 11.2.6 Quality control tests 303 11.3 Consumer tests 304 11.3.1 Purpose of consumer tests 304 11.3.2 Methods 304 11.4 Sensory testing administration 304 11.4.1 Facilities 304 11.4.2 Test administration 305 11.4.3 Experimental design 306 11.5 Selection and training of judges 308 11.5.1 Human subject consent forms and regulations 308 11.5.2 Judges 308 11.6 Statistical analysis of data 309 11.6.1 Analytical tests 309 11.6.2 Consumer tests 311 11.7 Relating sensory and instrumental ﬂavour data 312 11.8 Summary 314 References 314

12 Brain imaging 319 Luca Marciani, Sally Eldeghaidy, Robin C. Spiller, Penny A. Gowland and Susan T. Francis 12.1 Introduction 319 12.2 Cortical pathways of taste, aroma and oral somatosensation 320 12.2.1 Basic brain anatomy and function 320 12.2.2 Central gustatory pathways 321 12.2.3 Central olfactory pathways 323 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

x Contents

12.2.4 Central oral somatosensory pathways 325 12.2.5 Interaction and association of stimuli 325 12.3 Imaging of brain function 327 12.3.1 Methodologies to image brain function 327 12.3.2 Functional magnetic resonance imaging 328 12.3.3 fMRI design for ﬂavour processing 338 12.3.4 Behavioural data and subject choice 341 12.3.5 Measurement limitations 341 12.4 Brain imaging of ﬂavour 343 12.4.1 Brain imaging of taste 343 12.4.2 Brain imaging of aroma 343 12.4.3 Imaging cortical associations 344 12.4.4 Texture and the ‘taste of fat’ 345 12.4.5 The issue of the ‘super-tasters’ 345 12.5 Future trends 345 References 346

Index 351 P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

List of Contributors

Daniel Benczedi« Jack Knights Firmenich SA, Corporate Research and Duston, Northampton, Development, Switzerland UK

Ralf G. Berger Ulrich Krings Institut fur¬ Lebensmittelchemie Institut fur¬ Lebensmittelchemie Gottfried Wilhelm Leibniz Universitat¬ Gottfried Wilhelm Leibniz Universitat¬ Hannover, Germany Hannover, Germany

Imre Blank Isabelle Lesschaeve Nestle« Product Technology Centre, Orbe, Wine Aroma Wheels, Davis, CA, USA Switzerland Robert S.T. Linforth Samworth Flavour Laboratory, Division of Peter S.J. Cheetham Food Sciences, University of Nottingham, Hatton Park, Warwick, Warwickshire, UK Loughborough, Leics, UK

Tomas Davidek Luca Marciani Nestle« Product Technology Centre, Orbe, Nottingham Digestive Diseases Centre Switzerland NIHR Biomedical Research Unit, Nottingham University Hospitals, Sally Eldeghaidy University of Nottingham, Nottingham, UK Sir Peter Mansﬁeld Magnetic Resonance Centre, School of Physics and Astronomy, Ann C. Noble University of Nottingham, Nottingham, UK Wine Aroma Wheels, Davis, CA, USA

Susan T. Francis Gary Reineccius Sir Peter Mansﬁeld Magnetic Resonance University of Minnesota, Food Science and Centre, School of Physics and Astronomy, Nutrition, St Paul, MN, USA University of Nottingham, Nottingham, UK Jacques P. Roozen Penny A. Gowland Institute of Food Safety, Wageningen Sir Peter Mansﬁeld Magnetic Resonance University and Research Center Centre, School of Physics and Astronomy, Wageningen, The Netherlands University of Nottingham, Nottingham, UK Robin C. Spiller Josef Kerler Nottingham Digestive Diseases Centre Nestle« Product Technology Centre, Orbe, Biomedical Research Unit, Nottingham Switzerland University Hospitals, University of Nottingham, Nottingham, UK P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

xii List of Contributors

Andrew J. Taylor John Wright Division of Food Sciences, University of Princeton, NJ Nottingham, Sutton Bonington, USA Loughborough, UK Holger Zorn Saskia M. van Ruth Institut fur¬ Lebensmittelchemie und Institute of Food Safety, Wageningen Lebensmittelbiotechnologie, University and Research Center Justus-Liebig-Universitat,¬ Gie§en, Germany Wageningen, The Netherlands

Chris Winkel Givaudan UK Ltd, Ashford, Kent, UK P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

Preface

Food Flavour Technology was originally designed as a textbook to give a broad introduction to the formulation, origins, analysis and performance of flavours. Since 2002, when the book was first published, there have been developments in several areas, which necessitated a review of the book’s content. Specifically, there have been developments in the science and technology available for the study of flavour, changes in European regulatory processes and changes in consumer attitudes to food flavour. The original chapter headings have been retained, as all of them are still relevant, but the chapters have been revised by the authors to include new material that has appeared since 2002. Some new chapters have also been added. The aim of the book is to provide coverage of flavour technology topics that are relevant to scientists who are beginning to specialise in the area. Information on flavour research can be found in research papers published in scientific journals, but flavour researchers also like to present results at conferences and there is a wealth of information available in conference proceedings such as those from the Weurman, Wartburg and American Chemical Society symposium series. The chapter authors have tried to incorporate all this information into the chapters, so as to give a good overview of the science available on a particular topic. The creation of flavourings is the starting point for the book as this outlines the methodology and constraints faced by flavourists. This is followed by a second set of constraints that are the result of the new European flavour legislation. This is a very new area where there is still much discussion as to how the regulations will be, and should be, interpreted, and there are the usual inconsistencies and omissions that will be discovered and debated over the next few years. The origins of flavours are described in three chapters covering thermal generation, biogeneration and natural sources. The current consumer trend is to demand ‘natural’ ingredients in foods, and flavour manufacturers have adjusted their raw materials and processes to comply with this need as well as complying with the cost issues. Delivery of flavours using encapsulation or through an understanding of the properties of the food matrix is described in the next two chapters, and this section is followed by chapters describing the different ways to analyse flavours using instrumental, modelling and sensory techniques. Two new chapters have been added to introduce experimental techniques that are useful to the study of flavour. On-line flavour monitoring has been established for over 10 years and has been used to study a range of flavour processes. Measuring aroma release during eating and probing the link between the flavour profiles produced in vivo and the resulting sensory perception of the flavour has been one aspect that has received considerable attention. The effect of reactants and process conditions has been studied in thermally generated flavours, and on-line analysis also provides a high-throughput technique. In situations where there is a need to analyse hundreds or thousands of samples (e.g. individual fruits to study the link between fruit flavour and plant genetics), on-line analysis can gather large quantities of data to understand the complexities of plant breeding. The other new chapter describes P1: SFK/UKS P2: SFK FM BLBK221-Taylor/Linforth December 2, 2009 17:29 Printer Name: Yet to Come

xiv Preface

the techniques available to image the signals in the brain during food consumption and how the data can be used to study the perceptual process. Brain imaging is still relatively new in flavour studies, and the challenge is to carry out the experiments so as to obtain high-quality data and then to interpret the data to understand how the measured brain activity relates to perception. While the book describes the availability of science and technology to help the flavour industry, consumer attitudes in some parts of the world are limiting the uptake of these ideas. It is difficult to generalise these attitudes, but there seems to be a fear that food is no longer wholesome and that some of our current disease states (especially obesity) are the fault of the food and flavour industries. In this atmosphere, innovation needs to show some direct benefit for the consumer as well as the manufacturer, but the mood may change if the predicted changes in energy availability and climate take place and food becomes limiting. Against this rather negative mood, there are some interesting new aspects that may help us develop flavours in a more positive way. The discovery of both taste and odour receptors in the gut, followed by evidence that the sweet taste receptor is actively involved in glucose uptake, offers new potential to link flavour, not just with food intake but also with food uptake. Already there are patents covering the use of antisweet compounds such as lactisole to decrease glucose uptake in the gut, and the notion that flavours could be designed to influence nutrient intake through intake and uptake is interesting and one that merits intensive study. As ever, the only certainty in flavour research is that there will be changes and that work will be needed to apply these changes so as to produce acceptable flavours. The book editors and chapter authors hope that this book will assist future generations in this goal. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

1 Creating and formulating ﬂavours John Wright

1.1 INTRODUCTION

There are many different approaches to flavour creation and no one approach has a monopoly on the truth. Any successful technique must simply recognise the fundamental structures of flavours and then proceed logically to the goal. Some flavourists rely totally on blotters (strips of filter paper that are used to assess the odour of a mixture by sniffing). Some never touch them and make everything up to taste. Some flavourists throw most of the ingredients in at the start and some prefer to build up the composition step by step. Arguments about the logic, or lack of logic, inherent in some of these creative approaches miss the point. I have known good flavourists who use techniques that seem to me to be impossibly complicated and impractical. What all successful flavourists have in common is the ability to imagine the interactions between a very complex blend of raw materials and to use intuition and creative originality to fashion a work of art. Many successful flavourists are trained as scientists, but some had no scientific training whatever. Scientific method alone, without the spark of creativity, would mean that a single flavour would be a lifetime’s work.

1.1.1 A little history The flavour industry originated in the latter half of the nineteenth century with essential oil distillation and botanical extraction as the main sources of raw materials, often with a strong link to the pharmaceutical industry. Simple chemicals were available by the turn of the century, and during the first half of the twentieth century the fledgling flavour industry was increasingly driven by chemical research. For the flavourist of those times (who was often a pharmacist or chemist-turned flavourist), the task of making flavours was purely creative. Very little was known of nature, other than the major components of essential oils and a very limited number of chemicals that had been isolated from food and successfully identified. Most new chemicals that were synthesised had no possible value in flavours. The few that proved useful became the starting point for the synthesis of every possible related compound. Thus, the available raw materials were concentrated in a few obvious areas. Flavours created in this era were often not very close to the character of the real food, but some of them displayed real creativity and became accepted standards in their own right. The advent of gas chromatography and mass spectroscopy marked a real turning point for the industry. For the first time it was possible to see, in some detail, the chemicals used by nature to flavour food. The advance was, understandably, treated with some caution. What had been a purely creative and artistic profession could possibly be reduced to analytical P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

2 Food Flavour Technology

routine. The early analyses quickly dispelled all concerns. On reconstitution it was never possible to recognise anything more than a passing resemblance to the original target. Relieved flavourists quickly settled back to the old routine, but the more astute among them recognised a few diamonds in the mud. Among the first useful results from the new analytical techniques were pyrazines and unsaturated aliphatic alcohols. Chemicals such as trimethylpyrazine gave a true-to-nature roasted note to nut and chocolate flavours. Earlier flavours had been forced to rely on old- fashioned phenolic compounds such as dimethyl resorcinol. Dimethyl resorcinol provided a hint of roasted character but, at the same time, drowned the flavour in an uncharacteristic rubbery phenolic soup. cis-3-Hexenol gave an authentic green note to a multitude of fruit flavours, which previously had to depend on methylheptine carbonate to achieve a modicum of freshness (although tinged with melons and violets). Many of the failings of the early analytical techniques have now been overcome. Analyses are still not easy to interpret and different techniques can give very contradictory results, but they should form the starting point for the work of a good flavourist.

1.2 INTERPRETING ANALYSES

For virtually all flavours the nucleus is nature. We may or may not aim to reproduce nature accurately, but fully understanding nature is essential even for a caricature. Analysis is therefore the first step. Usually, several different types of analyses will be available (see Chapters 9 and 10 for details of the different flavour analyses available). Headspace analyses emphasise the more volatile components and are relatively true to the character of the food being analysed. The quantification of headspace analyses can usually be improved by applying vapour pressure correction factors. Early headspace analyses lacked detail and failed to capture less volatile components, but these shortcomings have now been largely overcome. Extract analyses are less accurate and contain more artefacts. They are often representative of a rather cooked character, but they do emphasise the less volatile components. Stir bar sorptive extraction is a good, nonintrusive, analytical technique and offers a wide-ranging analysis of liquids. Specialised analyses are often carried out to investigate the high-boiling components and also the sulfur and nitrogen compounds. The flavour of food will often vary depending on the plant variety as well as the growing or cooking conditions, and many analyses will quantify these differences. In consequence, the flavourist will often first have to correlate a wealth of information about the target food. The correlated list can be daunting, often running into many hundreds of different chemicals. The quantification used by the flavourist should be derived from the best of the headspace and stir bar results, corrected for vapour pressure, with extract results pressed into service for the less volatile chemicals Ð an impossibly complex problem on the face of it. The ‘trick’ of being a successful flavourist hinges on the ability to imagine the smell of complex mixtures, but a mixture of several hundred ingredients is far too complex to imagine. The first priority is to simplify the problem. Simplification can be carried out in three stages. The first stage is relatively easy. Many of the chemicals that have been found will be present well below their threshold levels and it might seem safe to ignore them all. Some caution is needed because synergistic and additive effects are common. The best approach at this stage is to build in a comfortable margin of error and retain any questionable chemicals. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 3

The second stage of simplification is to eliminate those chemicals that are likely to be artefacts. Artefacts can be present in the original food, produced during the separation process prior to analysis or produced during the actual analysis. Again, in cases of doubt, retain rather than discard. The final stage of simplification is to reject those notes in the target food that are genuinely present but are not desirable. Examples would be the trace by-products of fermentation and enzymatic browning in fresh fruits. Even the simplified analysis will usually be of daunting complexity. At this stage it is beneficial to try to reconstitute the analysis by mixing the flavour components in the proportions identified by the various analyses and then smelling and/or tasting the mixture. The result is certain to be disappointing, but it will serve to clarify the key aroma characteristics of the target food. Sometimes it is feasible to recreate the conditions of the original analysis using the reconstitution. Reanalysis will highlight the odd errors of identification, but it will invariably give a much improved quantitative base to start from. A second reconstitution may now give a recognisable product, but not one that anybody would be remotely happy to buy. It is time to abandon the strictly scientific approach and move on to the more abstract creative approach.

1.3 FLAVOUR CHARACTERISTICS

Smelling and tasting the target food will give the ﬂavourist a good idea of which aroma characteristics are important. Reconstituting the analysis will clarify this assessment even further and may well add a few unexpected notes. The aroma characteristics can be divided into two broad categories, primary and secondary characters.

1.3.1 Primary characters Primary characters are essential to the recognition of the target food. They constitute the basic skeleton of the flavour. Good examples are ‘violet’ (␣-ionone) in raspberries and ‘clove’ (eugenol) in bananas. It is impossible to create a realistic flavour without some contribution from these notes. Secondary characters are not essential for recognition but contribute an optional descriptive characteristic. Good examples are ‘leaf green’ (cis-3-hexenol) in strawberries and ‘dried’ (2-methylbutyric acid) in apricots. In both cases it is perfectly possible to make good, authentic flavours without these notes. Their effect is simply to vary the type of flavour to green strawberries and dried apricots, respectively. Strictly speaking, the primary characteristics can also be regarded, in some circumstances, as having secondary characteristics as well. A raspberry flavour with unnaturally emphasised ␣-ionone will smell distinctly violet. This is not a problem because the object of this exercise is, once again, simplification. It allows the flavourist to balance the primary characteristics in isolation and leave the secondary characteristics for later. Flavours vary greatly in the complexity of their primary recognition characteristics. The simplest example, at first sight, is probably vanilla. On its own, the chemical vanillin smells recognisably of vanilla. For many vanilla flavours in common use worldwide, this is all the primary character needed. Where consumers are accustomed to a more complex flavour, such P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

4 Food Flavour Technology

as the character of real vanilla beans, vanillin alone will not suffice to build a recognisable skeleton. Strawberry is a more complex flavour, and a more complex mixture of notes is required to achieve a recognisable flavour. In this example, ‘peach’ (␥-decalactone), ‘fruity’ (ethyl R butyrate), ‘guava’ (methyl cinnamate) and ‘candy’ (Furaneol ), blended in the correct proportions, would be the primary characters for the strawberry flavour skeleton. Some of the primary characteristics will be simple and will be represented by just one chemical in the analysis. Others may be more complex and may be represented by several chemicals. An example is the ‘peach’ note in fruit flavours. Major contributors to this note in many fruit products are ␥-decalactone and ␥-dodecalactone. Both chemicals have a similar ‘peach’ odour, but the taste characteristics intensify (and the odour strength decreases) with increasing molecular weight. When several chemicals contribute, the balance between the different components may need to be adjusted from that indicated by the analysis. Fortunately, that task can often be deferred until the basic skeleton of the flavour has been devised. It is easy to introduce unnecessary complication at this stage. In our peach example, we will find numerous additional lactones of similar structure in an analysis and it is tempting to think of them as part of a very complex primary characteristic. In reality, the additional lactones are not essential for peach recognition and are secondary notes. The flavourist is now ready to begin the real creative work. The objective is to achieve the best possible combination of what is now a reasonably limited number of chemicals to obtain a recognisable flavour skeleton. The analysis can be taken as a starting point, but it is no more than that. Even if the analysis is entirely quantitatively accurate, which is unlikely, it is still probably a long way from the optimum blend. It represents, at best, a specific example of the target food rather than one with every characteristic optimised Ð something that never quite occurs in nature. Ultimately, individual notes should be emphasised or reduced to make the flavour more attractive than the specific example of nature that has been analysed. It is possible, at this point, to try to take a relatively scientific approach and blend the two most important components first. The next step would be to determine the best level for the third component, and so on. The problem with this approach is that the presence of the third component alters the ideal balance between the first two components. The scientific approach rapidly becomes unimaginably complex and impossibly time-consuming. The best approach is to plunge in, taking the analysis as a starting point, and experiment with blends to understand the role of each of the primary characteristics. Speed is normally vital for commercial reasons, but it is also vital if the flavourist is to remain fresh and able to smell accurately. For that reason it is best just to use blotters at this stage and to experiment with large rather than cautious changes. If an addition is overdone, it can be blended back very quickly. If it is underdone, it is a slow process to carry on adding small quantities and there is a very real risk that the nose will fatigue to the chemical being added.

1.3.2 Secondary characteristics Once the basic skeleton has been built, the ﬂavourist has to concentrate on the more complex secondary characteristics. These can generally be worked on in groups. Green notes, for example, usually contain several subcategories and many different chemicals. Our strawberry ﬂavour would almost certainly contain the common ‘leaf green’ character cis-3-hexenol, but it could also contain lesser quantities of ‘fruity green’ (cis-3-hexenyl acetate), ‘apple green’ (trans-2-hexenal), ‘melon green’ (melonal; 2,6-dimethyl-5-heptenal), ‘unripe green’ P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 5

(hexanal) and ‘tropical green’ (cis-3-hexenyl butyrate). Once again the empirical approach is used to optimise this blend. Working through all the secondary characteristics will probably take some time. It is still best to use a blotter at this stage and to experiment with large rather than small changes. By now, the first stirrings of pride should be evident. It is time to taste the flavour. Tasting solutions should be simple and appropriate. If, for example, the target is a fruit and contains sugar and acid, then the taster should contain sugar and acid for the flavour to be appreciated accurately. Forget the end application at this stage. Two problems are apparent. The first, and most obvious, is that the balance between the components will seem a little different in aqueous solution from the way it appeared on the blotter. This is something flavourists learn to allow for when using blotters and is usually only a problem for trainees. Blotters offer three great advantages to the flavourist. They allow a very quick evaluation of each flavour. They also allow the simple comparison of many variants. Blotters uniquely offer a panorama of different aspects of your flavour as they air off and the more volatile components evaporate. This is a big advantage because it allows you to smell ‘through’ the flavour as it evaporates. The odour approximates that experienced in a simple taster for a relatively short time, usually about 5Ð10 minutes after dipping. The second problem is that some of the real taste (as opposed to odour) characteristics may be partially or even totally missing. For some flavours, such as roast beef, the taste element is obviously vital. Even when it is not so obviously important, for example in bananas, it is still surprisingly vital to the realism of the flavour. Correcting the taste imbalance is the next step in the flavour creation process.

1.3.3 Taste effects Taste effects are normally confined to individual flavouring ingredients that are highly water soluble or have a high molecular weight. Research on taste has lagged far behind that on odour, so natural extracts are still widely used to confer subtle taste effects. Maltol is a good example of a water-soluble taste effect ingredient. Maltol has a pleasant candyfloss odour, and a lingering sweet aftertaste, and is claimed to have flavour-enhancing properties (Labbe et al., 2007). It forms an important part of the aroma of a number of flavours, but the use of maltol as a taste ingredient dwarfs its use as an odour ingredient. Ethyl maltol is stronger than maltol, has similar taste and odour characters, but is not found in nature. Furaneol is even stronger than ethyl maltol and is found widely in nature. The only drawback to the use of Furaneol is that it can be easily oxidised. Vanillin is another water-soluble ingredient frequently used for its sweet taste effect and vanilla odour. Vanillin is widely found in nature and can be integrated into many flavour types. The taste effect of high-molecular-weight ingredients can be illustrated by the lactones in dairy flavours. The two most important lactones in all dairy flavours are ␦-decalactone and ␦-dodecalactone. ␦-Decalactone provides an excellent creamy odour in dairy flavours. ␦-Dodecalactone has a similar odour but has only about 10% of the odour strength of ␦-decalactone. The two ingredients have similar costs, and if odour were the only consideration, it would not make any sense to use ␦-dodecalactone. The higher molecular weight of ␦-dodecalactone gives it a noticeable creamy, oily taste. If cost were no object, the best combined taste and odour results would be achieved by a mixture of ten parts of ␦-dodecalactone and one part of ␦-decalactone. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

6 Food Flavour Technology

Many high-boiling, nature-identical chemicals have been little used in flavours because of the historical emphasis on odour rather than taste. They can often play a very useful role in enhancing taste characteristics even though they have little or no effect on the odour of the flavour. A wide range of natural botanical extracts have useful taste characteristics. Kola nut extract has a good astringent character, ginger extract has a hot character, Saint John’s bread extract has an attractive fruity sweetness and gentian extract has a lingering bitterness. All of these extracts also possess noticeable odours, and care must be taken to blend in their odour when they are added to a flavour for their taste effect.

1.3.4 Complexity Flavour formulations vary radically in complexity. The simplest flavour can be based on just one component. Many flavours, just like nature, contain hundreds of ingredients. Which is best? Very simple flavours have been popular since the earliest days of the flavour industry. Vanillin, isoamyl acetate and benzaldehyde have been the most popular single-component examples. Verysimple flavours may represent an attractive caricature, but they never taste like a real food. At the other extreme, very complex flavours often lack impact and can taste flat and characterless. Complex flavours can be deliberate (the result of slavishly following every detail of an analysis) or accidental (the result of lazy blending of flavours and intermediates). If a natural character is desired, then the optimum level of complexity is often the minimum number of components required to prevent the taster from perceiving the individual characters. This level of complexity can vary from perhaps as few as 15 components in simple fruit flavours to up to 100 in the most complex flavour of cooked food. There are, however, some important exceptions to this rule. The key problem with complex flavours is that a mixture of two chemicals usually smells weaker than the sum of its parts. The perceived intensity of flavour chemicals has a logarithmic rather than a linear relationship with concentration. At low concentrations, near the threshold, the logarithmic relationship does not hold because the chemical is not perceived at all until it reaches the threshold level. At high concentrations the relationship also does not hold because the nose fatigues to the stimulus. The lower extremes of the concentration scale explain synergistic effects, which otherwise appear to contradict the rule that a mixture smells weaker than the sum of its parts (see Keller and Vosshall, 2004, for more information on measuring odour psychophysics). Traces of components that, tasted individually, would be well below their threshold level can thus have significant positive effects in mixtures. At the other extreme, it is unwise to use so much of any single ingredient that the taster will quickly become fatigued. A mixture of two or more chemicals with complementary odours can often give better results.

1.3.5 Flavour balance Evaluating the flavour in tasters may also involve quite a number of modifications to improve the overall balance of the flavour. Once you have something you are basically happy with, it is a good idea to try out variations of concentration of the flavour in the taster. This is a little known, but extremely critical, way of evaluating a new flavour. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 7

Most flavours in nature are not particularly sensitive to changes in concentration. If you add twice as many apricots to a yoghurt, apart from the added acidity and sweetness, the yoghurt just tastes twice as strongly of apricots. The flavour does not become unbalanced. Most flavours created by humans do not fare nearly so well. It is possible to draw an analogy with jigsaw puzzles. An ingredient that does not have a counterpart in nature, in the flavour being created, can be seen as a large misshapen piece in the jigsaw puzzle. Not only is that specific piece out of place, but it also forces many of the other components out of balance. It may be possible, with enough effort, to get this flavour to taste right in a specific application and at a specific dose rate. The flaws will immediately become obvious if the application or the dose rate is varied because the apparent strengths of the different components will not change in unison. A prime example of an ‘alien’ unbalancing ingredient is ethyl methyl phenyl glycidate (strawberry aldehyde). This chemical is seductively attractive to flavourists because it smells more like strawberries than any other ingredient they have. It is very hard to turn your back on something that seems likely to give your flavour such a great start. It is not found anywhere in nature and it is certainly not found in strawberries. As we saw earlier, the natural character recognition skeleton of strawberry is a combination of ‘peach’, ‘fruit’, ‘guava’ and ‘candy’ primary characteristics. Ethyl methyl phenyl glycidate has a very complex odour, with a little of each of these notes. ‘Peach’ and ‘guava’ dominate and the chemical also has a strong ‘jammy’ character. It follows that if ethyl methyl phenyl glycidate is used in a strawberry flavour, it is impossible to build up the rest of the character recognition skeleton in the correct balance and it is also impossible to avoid some degree of ‘jammy’ character. This phenomenon is a powerful argument for using only those ingredients that are found in nature in the target flavour. This is undoubtedly the ideal, but as long as the odour character of a potential raw material is close to that of a naturally occurring ingredient, it is often possible to use it effectively. This sort of substitution would be very desirable if the naturally occurring ingredient were prohibitively expensive, impossible to make or very unstable. Tasting the flavour at double the optimum dose rate will make unbalanced components horribly obvious. Once those problems are corrected, the flavour should, at last, be something that is ready to show to other people. As with everything else involved in flavour creation, opinions vary radically about when and how to solicit opinions from other flavourists, non- flavourists and sensory panels. One thing is certain Ð I do not know of a single instance of a really successful flavourist who works in complete isolation.

1.3.6 Unfinished work An old saying cautions that you should ‘never show fools or children unfinished work’. Like many old sayings, it has an uncomfortable kernel of truth. It certainly highlights a real dilemma for the aspiring flavourist. Successful flavourists must be able to memorise and recognise a formidable range of raw materials. They must also have the ability to imagine the effect of complex mixtures and the creative spark to use these talents to make original flavours. A further essential requirement for this formidable being is an abundant helping of self-confidence. By self-confidence I certainly do not mean arrogance. Input from others is vital and it should never be treated with contempt. Self-confidence is essential to keep the flavourist sane in the face of well-meaning, but often contradictory, suggestions and criticism. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

8 Food Flavour Technology

Help for a trainee during the early stages of the creation of a new flavour is really the preserve of a mentor who is deeply involved in the project. There are always many different possible approaches to any problem, and it may not be obvious to other flavourists in which direction the trainee is trying to go. Their advice in the early stages of a project is likely to be wildly contradictory. Advice once the flavour has taken shape can be sought from a wide variety of sources. Other flavourists can be very helpful in a number of ways. They can give quick, and often accurate, assessments based solely on blotters. They will often have original ideas of raw materials to try out. Some will work and some will not, but the extra source of ideas is invaluable. Other flavourists can often pick out mistakes that the originator has missed or, more frequently, has become too saturated with the flavour to notice. It is important to recognise that the advising flavourist is often making an impromptu suggestion based on a quick evaluation. However good the flavourists, their suggestions are not necessarily gold dust. Sensory panels, especially expert panels, can be a valuable source of guidance on matches, hedonic ratings of new flavours and profiling. Panels are especially helpful in matching work. No flavourist is ever completely satisfied with a match of another flavour and a panel provides a reality check. Preference mapping, linked to profiling, can provide real insight into the best way to optimise a flavour for a specific consumer group. Simple sensory panels should be avoided as they all tend to lead in the direction of a bland, uninteresting flavour that offends nobody but, equally, excites nobody (see Chapter 11 for more detail on sensory testing methodologies). Other, noncreative, staff can also be a useful source of criticism. It is often helpful to involve applications and sales staff. It is, after all, very difficult for sales staff to sell something that has not first been sold to them. Sensory panels and noncreative staff can rarely comment on blotters or simple tasters. The flavour must first be applied to a realistic end product.

1.4 APPLICATIONS

All flavours are used in end products that impose some requirements on the finished flavour because of interactions with the finished food. These interactions can be broadly grouped into four categories Ð ingredients, processing, storage and consumption.

1.4.1 Ingredient factors The most important factor is the fat content of the finished product. Flavour chemicals vary in polarity and consequently in fat solubility. Taste thresholds in fat are much higher than in water. The partition of different components in a flavour may vary, and this can alter the balance of the perceived aroma. It is often possible to adjust the formulation, and the methods described in Chapter 10 have been used to measure aroma release and then rebalance flavours in foods with different fat contents (Shojaei et al., 2006). However, an alternative approach is to avoid drastic differences in the polarities of the flavour components. In all foods containing fat, added flavour will slowly partition between the fat and the aqueous phases on storage. This effect can be partly avoided by adding separate flavours to the fat and the aqueous P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 9

phases, but this is a laborious approach and will rarely be sufficiently accurate to avoid subsequent partition effects. Care must be taken in application trials to store the finished food sufficiently long before tasting to allow the partition of the flavour to be substantially completed. The lipophilic gum base in chewing gum has an effect similar to that of fat, but the problems are aggravated because the flavour is gradually extracted by chewing. If there are differences in the polarities of the flavour components, the chewing gum will appear to taste mainly of the most polar components at the start of chewing. Eventually only the nonpolar chemicals will be extracted. A completely fat-soluble flavour may be necessary for some applications. At the other extreme, entirely water-soluble flavours are essential for clear soft drinks. In both cases it is difficult to produce a balanced profile within a restricted range of polarity. In these examples it is sometimes helpful to depart from the essentially nature- based approach we have used so far. All flavourists should keep a reference record of the characteristics of all the raw materials they have encountered. This database can usually be used to find a chemical with a similar odour character to a problem raw material but with different physical or chemical properties. Natural extracts and oils often contain chemicals with widely differing polarities. They can be processed by distillation, solvent extraction and chromatography to reduce these differences. The most common example of this type of process is the deterpenisation of lemon oil. Lemon oil contains about 90% of terpene hydrocarbons, which are nonpolar, low boiling and susceptible to oxidation, and contribute little to the overall flavour character. The oil also contains about 6% of oxygenated chemicals, which are polar, relatively high boiling and less susceptible to oxidation, and provide most of the flavour character. The level of lemon oil that would be required to impart an acceptable flavour level to lemonade would result in a level of terpene hydrocarbons in the drink well in excess of their limit of solubility. The oxygenated chemicals would be readily soluble at this level, so a clear drink could be obtained by removing the hydrocarbons from the oil. Chromatography and solvent extraction are obvious possibilities. Distillation also works because of the difference between the boiling points of the terpene hydrocarbons and most of the oxygenated chemicals. Some loss of the true lemon character is inevitable owing to processing and the small, but significant, flavour contribution made by the hydrocarbons. This is more than justified by the gain in stability to oxidation. Solvent extraction generally gives better results than distillation because this method retains the most volatile aliphatic chemicals, which are responsible for the fresh, juicy character of many citrus oils, especially orange oil. Solvent extraction is discussed in more detail later in this chapter. Distillation, if it is used to produce a terpeneless oil, unfortunately also removes the high-boiling antioxidants that are present in cold-pressed citrus oils. Major components of a flavour may themselves cause problems in a finished food. These problems are often changes in texture or in the stability of emulsions. The solvents are the most likely culprits, and in many instances a change of solvent will provide a cure. Where flavour dose rates are very high, particularly in chewing gum, individual flavour chemicals may also be responsible. When this happens, the flavour can often be modified, but sometimes the only possible solution is to modify the formulation of the application. This may also be an issue if the flavour necessarily contains large quantities of a food additive, for example, an acid. This could happen, for instance, in a natural flavour containing significant quantities of concentrated fruit juices. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

10 Food Flavour Technology

Carbohydrates in a finished food may have a binding effect under certain conditions, but this is not a frequent problem. The most obvious example is the loss of flavour in bread on storage because of flavour binding to the helices in starch molecules. Flavour binding by proteins is a more serious problem. Most protein molecules are folded in such a way that the nonpolar amino acid side chains are on the inside and the polar groups are on the outside. Flavour chemicals can interact with the nonpolar interior regions of the protein and cause it to unfold. They can be bound into the protein by absorption at the protein surface or inclusion in the nonpolar interior. Protein flavour binding is most evident in processes involving heat and is most pronounced with carbonyl flavour chemicals. Chemical interactions and partitioning effects make it essential to wait at least 24 hours before tasting some applications.

1.4.2 Processing factors Minor effects from processing include those from filtration, aeration and freezing, but by far the most important factor is heat. This may cause chemical changes in the flavour, but the main problem is the loss of volatiles. This may have the effect of reducing the fresh top note of a flavour. If the key recognition chemicals have widely different boiling points, heat could render the flavour unrecognisable. The choice of solvent can reduce this problem. In some instances, volatile chemicals may be replaced by higher boiling analogues. It is usually possible to change the balance of the flavour to allow for differential losses, but this solution gives a flavour that is suitable for only a limited range of applications. In processes involving considerable heat, such as bakery and extrusion, the best solution is multiple encapsulation. In this process a spray-dried flavour is coated with a high-melting-point fat. This process protects the flavour until the fat melts. Chapter 6 introduces the key concepts required to encapsulate flavours effectively.

1.4.3 Storage factors Some wines and cheeses improve with age, but they are the exceptions rather than the rule. Flavour stability in an application should ideally at least match the shelf-life of the food itself. When a flavour is added to food, some chemical changes, such as the hydrolysis of acetals, occur quite quickly. There are frequently subtle differences in flavour character after only one day. In the longer term, oxidation is responsible for most of the changes in flavour during storage. When most flavour chemicals oxidise, the effect is simply perceived as loss of flavour because the flavour chemical has a much stronger odour than its oxidation products. When some incidental components, such as the hydrocarbons in a lemon flavour, oxidise, the effect is often perceived as an off-note. Substitution of flavour components with more stable alternatives and the use of antioxidants usually reduce the problem to manageable proportions. Migration of flavour chemicals into or through food packaging materials can sometimes occur. It may lead to a detectable loss of flavour or cross-contamination problems. A change in packaging material is the best cure, but, if this is not possible, it may be practical to reformulate the flavour without the problem of chemicals. This may change the profile of the flavour. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 11

Tea bags present a very specific packaging problem because of the size of the holes in the tea bags. Liquid flavour can be spread directly onto tea leaves, but this leaves the flavour very prone to oxidation and evaporative losses. Spray-dried flavours need to be agglomerated to prevent them from falling through the holes in the tea bags. Tea dust can be included in the agglomerated flavour to give it a similar appearance to the tea in the bags. Tea used in tea bags is generally of small particle size, and care needs to be taken to ensure that the size of the agglomerated flavour particles matches that of the tea.

1.4.4 Consumption factors Many of the processing problems can resurface when the food is consumed. This is particularly common when a powder flavour is used in a dry convenience food. The final factor that may influence the formulation of the flavour is the temperature at which the finished food is consumed. At temperatures below room temperature, as in the case of ice cream, the intensity of the whole flavour is reduced. The intensity of the most volatile chemicals is reduced relatively more than the rest of the flavour, and they may need to be increased. Caution should be exercised because the food can warm up in the mouth. Foods that are consumed hot are more difficult to flavour. The high temperature increases the intensity of the flavour, particularly the more volatile chemicals. At the same time it may cause a relatively greater loss of the same components.

1.5 FLAVOUR FORMS

Liquid and powder flavours can be split into a number of major types, each of which poses some specific problems for the flavourist.

1.5.1 Water-soluble liquid flavours These are by far the most common types of flavours. The flavour chemicals and natural components are dissolved in a simple solvent, most commonly propylene glycol, triacetin or ethanol, with the possible addition of water. If the flavour contains significant amounts of solids, such as vanillin or maltol, then the quantities added must remain well within their limit of solubility. Storage conditions can be much harsher in real life than in a laboratory, and a large safety margin should be built in. The same consideration should be applied to the nonsolid components of the flavour, but problems are not as common in this area. Propylene glycol is usually the solvent of choice. It is stable, virtually characterless in use, and confers some stability in applications involving heat processes. The drawbacks of propylene glycol are that it is not a strong solvent, that it is not natural, that the levels of use are restricted in some countries, and that it forms acetals and ketals quite readily with carbonyl flavour chemicals. Acetal and ketal formation can be inhibited by the addition of water to the flavour, but this makes an already weak solvent even weaker. Acetals and ketals can actually be useful in some applications because they may protect the parent carbonyl from oxidation during storage of the flavour. They will later break down to release the parent carbonyl in many applications in the presence of water. The most serious problem deriving from acetal and ketal formation is that many acetals and ketals are only poorly soluble in P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

12 Food Flavour Technology

propylene glycol. This can result in the puzzling phenomenon of an initially clear flavour, gradually phase separating and forming two layers. Ethanol is also widely used, especially where there are no duty handicaps (many countries impose high taxes on ethanol). It is relatively stable, has a mild but pleasant character in use, is readily available in a natural form, and can be diluted significantly with water. The drawbacks are relative instability in applications involving heat processes, religious restrictions in some countries, flammability and the formation of acetals and ketals. The last factor is less important because ethanol is a strong solvent and can still remain effective if sufficient water has been added to inhibit acetal and ketal formation. Triacetin is not a solvent of choice for most applications. It is not very water soluble, and when it does dissolve, it decomposes to glycerol and acetic acid. Triacetin has a slight bitter taste and acetic acid has a noticeable ‘vinegar’ odour. Triacetin can be the solvent of choice when water solubility is not critical (as in many confectionery applications), when propylene glycol is restricted, when the components of the flavour will not dissolve readily in ethanol or propylene glycol and, most importantly, when propylene glycol has an undesirable effect on the texture of the finished food. Chewing gum is the most important example. Propylene glycol hardens chewing gum, but triacetin acts as a plasticiser. Water is not added to triacetin-based flavours because it would hydrolyse the triacetin. Other solvents may be useful in specific cases. Triethyl citrate is similar in many respects to triacetin. It is poorly water soluble, but is odourless and confers heat stability. The major difficulty with triethyl citrate is the bitter aftertaste, which severely restricts the level of use. Diacetin is also similar to triacetin but is generally less effective. Glycerol is a very weak solvent but can be used effectively in conjunction with ethanol in natural extracts to confer some heat stability. Lactic acid is not generally a very effective solvent but can be useful, in mixtures, for some problematic raw materials, especially maltol. Benzyl alcohol has a faint floral character and is a good solvent but is prone to oxidation to benzaldehyde. Benzyl benzoate is stable and relatively odourless. It can be used in solvent mixtures, especially for oil-soluble flavours, but has an unpleasant flavour at high levels. Many of these lesser solvents are not universally recognised as solvents. They may be permitted as flavouring ingredients, but care must be taken of the level of use.

1.5.2 Clear water-soluble liquid flavours This category is very similar except for the requirement that the end product, usually a beverage, should be crystal clear. Most flavour raw materials are entirely water soluble at their normal level of use. The exceptions are limited to a few chemicals that need to be used at relatively high levels (usually esters), chemicals that can form insoluble polymers on storage and terpene hydrocarbons. Terpene hydrocarbons are found in many natural essential oils. They have limited use as flavouring ingredients (there are exceptions, such as myrcene), and they are prone to oxidation. The hydrocarbons can be removed from essential oils by distillation, chromatography or solvent extraction. The most effective method is solvent extraction (often called ‘washing’) because it causes least change in the character of the original oil. The most effective solvent is a mixture of ethanol and water, but propylene glycol can also be used. The extraction is carried out by dissolving the oil (for example, orange oil) in ethanol, adding water to throw out the hydrocarbons (commonly called ‘terpenes’), chilling the mixture and allowing it to stand for 2 days. The terpenes float to the top of the mixture, which can then be drawn off P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 13

and filtered. A little extra alcohol is added as the final stage to prevent the flavour becoming cloudy if it is stored in the cold. The process can be speeded up by the use of a coalescer, a metal mesh that coalesces the oil droplets. Propylene glycol ‘washings’ are difficult to make because of the viscosity of the solvent and the small amount of water that can be added. The use of a coalescer is virtually essential to make propylene glycol-based ‘washings’. A surprising, but effective, alternative way to produce clear beverages is through the use of low-payload, small particle size emulsions. The flavour must contain only very limited amounts of terpene hydrocarbons for the process to work. This method is widely used for cola flavours.

1.5.3 Oil-soluble liquid flavours Oil-soluble flavours are needed where the end product is an oil or a fat. They are also used where the end product cannot tolerate water. Both ethanol and propylene glycol contain small amounts of water, so these solvents cannot be used in water-sensitive products such as chocolate. Natural or synthetic (medium-chain triglyceride) vegetable oils can be used as solvents. The problems are susceptibility to oxidation (for the natural oils) and poor solvent power. Many of the chemicals that are important for taste effects are highly polar and poorly soluble in oils. Some of the minor solvents discussed earlier, such as benzyl benzoate and triethyl citrate, can be particularly effective in oil-soluble flavours. They all have some drawbacks and may be more effective when used as mixtures. If this does not work, one possible solution is to dispense with traditional solvents altogether and use the major components of the flavour to dissolve the solids. This is not always possible without adding excess quantities of weak- tasting esters such as ethyl acetate. When it can be done, the resulting flavour may be highly concentrated and very difficult to dose accurately in an industrial environment. Essential oils can be effective ‘solvents’ for some oil-based flavours. This is especially true of citrus flavours. The natural oil gives a realistic, complex background and added flavour ingredients give powerful specific character.

1.5.4 Emulsion-based flavours Emulsions, based for example on orange oil, are often used to give cloud to a beverage, but they can also be a cheap and effective way of delivering a flavour where cloud is not an issue. The water-soluble components, such as vanillin, can be dissolved in the gum solution (typically gum arabic or modified starch is used as emulsifiers), and the remaining components can be mixed together to form an oil phase, which is then emulsified. Potential problems include the clumping or separation of the oil phase, the hydrolysis of susceptible flavour ingredients and the microbiological stability of the emulsion over an extended period, especially once the container has been opened. Ideally, for these applications, the oil phase should constitute around 5% and certainly not more than 10% of the emulsion.

1.5.5 Dispersed ﬂavours Dispersions are a similar, cheap and cheerful, way of delivering ﬂavours in powder form. If all the ingredients are solids, they may be mixed together and diluted with a carrier such P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

14 Food Flavour Technology

as lactose. If some of the ingredients are liquids, they are mixed together and spread on the carrier before the solids are mixed in. This method works if all the ingredients are relatively high boiling and not susceptible to oxidation. Even so, it produces flavours with a relatively short shelf-life and it is difficult to mix the flavours so that they are entirely homogeneous.

1.5.6 Spray-dried flavours Spray-drying is the method of choice for powder flavours (see Chapter 6 for the mechanisms of flavour encapsulation). The flavour is typically emulsified in an aqueous gum solution, and then dried by spraying into a hot chamber. This method is preferable to dry mixing because the resulting flavour is stronger and much more stable to evaporation and oxidation. Spray-drying works so well because the sprayed droplets form a semipermeable shell very quickly, long before most of the water has evaporated. The semipermeable shell allows water to pass, continuing the drying process, but is relatively impermeable to most flavour components. This is true of even the smallest flavour molecules, such as ethyl acetate and acetaldehyde. Only a small proportion of the ethyl acetate or acetaldehyde added to a flavour survives spray-drying, but, without the effect of the semipermeable shell, logic would dictate that the loss on drying would be virtually 100%. Flavourists are not expected to be experts in the area of spray-drying and the many variants of this technique, but they should know enough to get the best out of the process. The first issue is the way the flavour is added to the emulsion. The criteria are much the same as those for liquid emulsions, except that there is no need for the emulsion to be stable in the long term. The ingredients of the flavour should be split into water-soluble and oil-soluble keys. The water-soluble components should be dissolved in the gum solution. The oil-soluble components should be emulsified in the resulting mixture. This emulsion does not need to be stable for longer than it takes to dry the batch, but it should be emulsified to a reasonably small and uniform particle size. Poor emulsification will result in more surface oil, flavour loss and susceptibility to oxidation. Some solvents should not be used in the formulation of the keys. Ethanol will increase the flavour loss, and propylene glycol (in more than trace amounts) will make the powdered flavour hygroscopic. Triacetin works well in most instances. The maximum loading of the oil phase is around 30% of the dry weight, but drying losses increase steeply after 20%, as does the amount of surface oil. For cost-effectiveness, 20% is a good maximum to aim for. The second issue is the composition of the gum solution. Gum acacia is the most widely used material, although some modified starches can give equally good results. Gum acacia varies widely in quality and care should be taken to buy 100% pure gum from a reputable source. The cheap gum that has been cut back with filler is always of poor value. The flavourist should be free to control the proportion of pure gum used. It is a waste of money to use 100% gum acacia as the carrier. In spray-dried flavours, 30% is the absolute maximum quantity of gum needed for even the most challenging applications. In many cases as little as 10% is all that is needed to form the semipermeable film during drying. The filler, usually maltodextrin, is important because a high dextrose equivalence is needed to make the shell of the spray-dried particle less permeable to oxygen. One unintentional advantage of using reduced levels of gum acacia is that the viscosity of the emulsion is lower. This allows the solid content of the emulsion to be increased, while still keeping to a viscosity level that can be handled readily. Higher solids mean more throughput and less energy costs. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 15

The third issue is the processing conditions. They should ideally be set for each flavour. The emulsion should not be warm because this will damage the flavour. The inlet temperature must be adjusted so that particles hitting the sides of the drier do not stick. In general, the best results are obtained with the highest inlet temperatures and the highest throughput. The outlet temperature should be reduced as much as possible, but not so much that the spray-dried particles contain significant moisture when they leave the drier.

1.6 PRODUCTION ISSUES

One of the most difficult challenges facing flavour companies is the link (or lack of it) between the creative flavourists and the production staff. It is possible to sidestep the issue by introducing a complete department to sort out problems, but it is obviously much better not to have them in the first place. More thought should be given during the training of flavourists to the possibility of a spell in production. QC training is often included, but nothing beats getting your hands dirty and learning about the practical issues first hand. The first issue is the total number of raw materials available for use. A sensible number can be reached by adding to the number of GRAS (generally recognised as safe; see Chapter 2 for information on the legal status of flavours) and European raw materials (around 3500), the number of sensible variations of natural products (around 500) and the legal variations (natural, organic, kosher, etc.) (around 1500). A sensible maximum is 6000. Not many companies can boast such a small list, but the cost, quality and service problems associated with large raw material lists are formidable. The total number of flavour formulations is also often quoted by production as a key problem metric, but it is only a problem if the operations function is so inefficient that it is necessary to keep stock of finished flavours. Very few flavours are now sold off-the-shelf and the tailor-made flavour is becoming the standard. It makes much more practical sense, and it is much more cost-effective, to concentrate on controlling the number of raw materials. The second real issue is the number of raw materials in any individual formulation. It is simply not possible to justify more than 100 ingredients in a flavour. Depending on the type of flavour, the optimum number of ingredients can vary between 15 and 100 but in most cases the best effect is obtained using between 15 and 50 ingredients. The cost of compounding and the service problems associated with very complex flavours are both serious issues. Accurate compounding in a production environment is very different from the situation in a laboratory. The use of solutions should be tailored to production needs and kept to a minimum necessary for accurate weighing. Old solutions should be discarded. The use of a single key ingredient may be helpful in some instances to separate out all the very low- volume items. Outside this restricted context, the use of keys and the blending of flavours in general are real headaches for production. They are also, frankly, indicative of lazy work on the part of the flavourist concerned. The correct compounding order may be obvious to the flavourist, but it must be specified in a formulation to assist production. Other important notes are the need to filter (which can often be avoided by better selection of raw materials or more careful formulation) and full details of any processes. The originating flavourist should always be involved in the quality control testing of the first batch in case there are problems scaling up the flavour. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

16 Food Flavour Technology

1.7 REGULATORY AFFAIRS

Flavourists should receive extensive training on regulatory issues, not simply the widely varied global flavour regulations but also the implications for finished foods and labelling (see Chapter 2 for further information on the safety and legislation of flavours). With the current time scales for projects, it is not practical to expect that a final regulatory check should be anything more than a safety net. A generally conservative approach should be taken, and wherever possible GRAS ingredients should be used. The regulations in Europe and the US are increasingly well harmonised, so this restriction is usually practical. For any country, the IOFI (International Organisation of the Flavour Industry) guidelines represent the minimum standard, irrespective of the lack of local regulations. Natural certification of raw materials should not be accepted without critical evaluation. Natural standards vary by country and common sense should be applied.

1.8 A TYPICAL FLAVOUR

Raspberry flavour is a good learning tool. It is relatively simple, but not so simple that it does not contain a multitude of useful lessons. To illustrate the process of flavour creation, we will work on an imaginary, but typical, customer project. The task at hand is to create a nature-identical flavour, with a profile, that the customer has described as true to nature, fresh and red. The end use is hard candy. Let us imagine that the flavourist has two analyses to work from. One derived from the analysis of an extract from the fruit and the other derived from the analysis of the headspace over the fruit. In these two, hypothetical analyses, 362 different chemicals have been identified, 271 in the extract analysis and 203 in the headspace analysis. In both cases the quantification is expressed as a percentage of the total volatiles. The headspace analysis is also quantified with an added vapour pressure correction. Table 1.1 gives the quantification of those chemicals from the analyses that we will consider using to create a simple flavour. The first step is to decide which chemicals in either analysis represent primary characteristics. In the case of raspberries, the violet note is clearly essential. The analyses contain both ␣-ionone and ␤-ionone. ␣-Ionone has a clean ‘violet’ note and ␤-ionone has a ‘violet’ note in addition to a strong ‘cedar’ note. To keep things simple there is an obvious temptation

Table 1.1 Flavour components identiﬁed in analyses of raspberry ﬂavour.

Cost (in order Extract Headspace Vapour pressure of appearance) (%) (%) adjusted (%) ␣-Ionone 4.00 0.70 8.000 ␤-Ionone 1.80 0.50 9.500 4-Hydroxyphenylbutan-2-one 0.50 — — Damascenone 0.05 0.02 0.150 Dimethyl sulﬁde 0.02 1.50 0.001 Acetyl methyl carbinol 0.50 0.20 0.002 Ethyl acetate 5.00 9.80 0.040 cis-3-Hexenol 8.00 0.60 0.030 cis-3-Hexenyl acetate 0.02 0.04 0.010 ␦-Decalactone 0.60 — — P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 17

to ignore the complications of ␤-ionone and work with ␣-ionone alone. This simplification might work, but it is probably unwise. The ‘cedar’ note of ␤-ionone generates a ‘pippy’ or ‘seedy’ effect in raspberry flavours. This note is hardly a primary characteristic, but it is normally attractive. If ␤-ionone is ignored at this stage, then a later correction to add a ‘seedy’ note will necessitate a rebalancing of the ‘violet’ character. The next step is to establish an estimate of the correct concentration of the 70/30 mixture (these proportions are derived from the extract analysis) of ␣-ionone and ␤-ionone using a simple taster. A good starting level would be 0.25 ppm (part per million or milligram/ kilogram). The most common dilution of flavours in beverages is 0.05% rtd (ready to drink). This is equivalent to 0.035% ␣-ionone and 0.015% of ␤-ionone in the flavour. The other primary characteristic is not quite so easy to identify. The flavour of ␣-ionone alone is simply ‘floral, violet’. The missing character should confer a specifically ‘berry’ note. The only feasible candidate in the analyses is 4-hydroxyphenylbutan-2-one. This chemical has a distinct ‘berry’ aroma, even a specific hint of raspberries. Again trial and error can be used to establish a good balance between these two chemicals. A good starting level would be around 2% in the flavour, but later in the process this will prove to be too high (once other ingredients with somewhat similar characteristics have been added) and the final level is 1%. At this stage we already have a recognisable raspberry skeleton. We can move on to the optional, secondary components. The customer wants a true-to-nature character, but also describes the target as ‘red’ and ‘fresh’. Neither of these descriptors is very specific, so the flavourist has to try to guess the customer’s wishes. This dilemma is very common and illustrates the need to work with customers to establish specific descriptors. ‘Red’ can reasonably be taken to mean red raspberries rather than black (so no musk character), ripe rather than unripe (so ripe, ‘fruit’ notes and restricted ‘green’ notes). ‘Fresh’ can be taken to mean an absence of ‘jammy’ or ‘cooked’ notes. It might also indicate high ‘green’ notes, which certainly confer freshness. A more moderate level of ‘green’ notes is probably a good idea because ‘red’ was also specified. High levels of ‘green’ notes, especially ‘raw, green’ notes, give an unripe effect. A good choice for the red ‘fruit’ note is the ‘damson’ character of damascenone. This chemical is found widely in nature and is, justifiably, a favourite with flavourists. The ‘damson’ character of damascenone adds richness and a deep ‘fruity’ character to our fledgling raspberry flavour. Taking the extract analysis as a guide (damascenone is fairly high boiling), the levels of the ionones used in the flavour indicate a level of 0.0004% of damascenone in the flavour. This seems very low indeed. To obtain the correct character we must increase the level in the flavour to 0.04%. Dimethyl sulfide is also an excellent ripe ‘fruit’ note in dilution, although at high levels it has a ‘cabbage’ character and can make the flavour seem cooked and ‘jammy’. The addition of dimethyl sulfide also improves our flavour dramatically, but 0.01% is the most we can add before the character becomes slightly ‘jammy’. This is, however, far more than the level indicated by the vapour pressure-corrected headspace analysis. The ‘buttery’ note of acetyl methyl carbinol will also, surprisingly, add to the ‘ripe’, ‘red’ character. The concentration that would be required in the flavour, on the basis of the amount found in the extract analysis, relative to ␣-ionone, is around 0.005%. This level works well and provides the required note. The final ‘fruity’, ‘red’ note is ethyl acetate. The headspace analysis indicates a very low level, but the extract analysis (which would be expected to give a low result) indicates a level broadly similar to that of ␣-ionone. Increasing that level a little to 0.10% in the flavour gives an attractive result. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

18 Food Flavour Technology

The most obvious green note is cis-3-hexenol, but this chemical has a ‘leaf, green’ character, similar to fresh-cut grass. Like damascenone, cis-3-hexenol is found very widely in nature and is often the first choice when a ‘fresh’ character is desired. We could add a low level of cis-3-hexenol, but we would run the risk of introducing an ‘unripe’ note. A much better choice for this flavour would be cis-3-hexenyl acetate, which has a softer ‘fruity, green’ character and very little unripe note. The analyses would indicate a low level of cis-3- hexenyl acetate, but a higher level is necessary because we are not adding any cis-3-hexenol. In practice, the ideal level is 0.02% in the flavour. This flavour will smell reasonable but taste very thin. Only two components of the flavour so far have a significant taste effect Ð damascenone and 4-hydroxyphenylbutan-2-one. The flavour has a degree of ‘berry’ depth of taste, but needs added ‘sweet’ character. The addition of 2% of maltol (not found in the analysis) will help to solve this problem, but is obviously not the ideal solution. Maltol has a ‘candyfloss’ aroma and imparts a lingering sweet aftertaste. It adds depth, but is too a simple character. One other addition that will help to add depth is ␦-decalactone. This chemical has a ‘creamy’ character and trial, and error establishes an ideal concentration in the flavour of 0.05%. This level is higher than that indicated by the extract analysis. The flavour we have developed thus far is much too simple and will taste like an obvious mixture of separate notes. The addition of a small amount of jasmine absolute (0.02%) will add complexity and traces of desirable ‘berry’ (from benzyl acetate), ‘lavender’ (from linalool), ‘animalic’ (from indole) and ‘jasmine’ (from methyl jasmonate) notes. The final stage in the creation of our very simple raspberry flavour is to make allowances for the processes involved in making hard candy. The only factor is heat, so the most volatile components must be increased to allow for the losses in processing. Ethyl acetate should be increased to 0.20%, dimethyl sulfide to 0.02%, acetyl methyl carbinol to 0.009% and cis-3-hexenyl acetate to 0.03%. In real life, the process of developing this flavour would be much more complicated, but this simple example serves to illustrate the principles involved. The composition of the final flavour is compared to the two analyses in Table 1.2. The analyses help, but they are a long way from the quantification of the finished flavour. This example illustrates the high level of creative input, even when analyses are taken as the starting point.

Table 1.2 Comparison of raspberry ﬂavour analysis (from Table 1.1) with formulation of a raspberry ﬂavour suitable for hard candies.

Cost (in order Extract VP adjusted Flavour of appearance) (%) (%) (%) ␣-Ionone 4.00 8.000 0.035 ␤-Ionone 1.80 9.500 0.015 4-Hydroxyphenylbutan-2-one 0.50 — 1.000 Damascenone 0.05 0.150 0.040 Dimethyl sulﬁde 0.02 0.001 0.020 Acetyl methyl carbinol 0.50 0.002 0.009 Ethyl acetate 5.00 0.040 0.200 cis-3-Hexenol 8.00 0.030 — cis-3-Hexenyl acetate 0.02 0.010 0.030 Maltol — — 2.000 ␦-Decalactone 0.60 — 0.050 Jasmine absolute — — 0.020 P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 19

1.9 COMMERCIAL CONSIDERATIONS

1.9.1 International tastes We are all accustomed to rapidly increasing globalisation, and with it the assumption that one product can be sold in all markets. There are cases where, with sufficient advertising, this is manifestly true. In most cases, however, the assumption does not hold. Regional tastes for most of the key flavour types still override global stereotypes. The regional tastes are often derived from historical familiarity and may fade in time, but, for now, they are very important. The main regional preferences for the most important flavour categories are summarised as follows:

Beef : Roast beef is the preferred profile in the UK. Grilled beef reigns supreme in the US, and in much of Asia boiled beef is the main profile. Cheese: Cheddar is far and away the most important type of cheese in terms of flavour sales. Blue cheese and Parmesan are very small categories compared with Cheddar. Cheddar can be broken down into two main types by region: sharp and mild. Sharp Cheddar is best defined (ironically) by aged US or Canadian Cheddar cheese and represents the target profile in Europe. The taste preferences of US consumers are very different. In this region a mild, creamy, buttery character is preferred. Cherry: In Europe the hawthorn note of Morello cherries is preferred, but in the US benzaldehyde is the prominent character. Chocolate: Milk chocolate predominates in most markets outside Europe and the milk component often has a cooked character. Some popular milk chocolates also have an added signature note such as cinnamon or almond. Dark chocolate is popular in Europe and can have pronounced burnt and bitter characteristics in this market. Lemon: In the UK especially, but also in much of Continental Europe, a high citral level is liked. The European taste also likes an exaggerated level of jasmine character. In the US, lemon is milder and more floral, and in much of the rest of the world citral is the defining note, sadly, sometimes accompanied by the waxy character of oxidised oil. Mango: As with most other tropical fruits the situation is the exact reverse of berry flavours. The genuine character, with its strong terpene, skin note is preferred in Asia and Latin America. In Europe and the US a pale imitation flavour is preferred, with much reduced skin and sulfur notes and an emphasis on melon and peach. Milk: Fresh milk and dairy flavours are optimum in the US and Europe. In Asia a boiled, condensed milk note is preferred, and in Latin America the even more caramelised ‘dulche de leche’ is ideal. Orange: In Europe there is the strange contradiction that the flavour of processed juice is liked, presumably for nostalgic reasons, together with the pungent, fresh note of acetaldehyde. An exaggerated hint of violet is also liked in many orange flavours for confectionery. Fresh juice character is popular in the US and, for the rest of the world, cold-pressed orange peel oil is the most popular character. Raspberry: In the US a strong violet character is preferred, but in Europe this note is muted and balanced by fresh and green characters. In Asia real raspberries are a rarity and an old-fashioned candy character is preferred. Strawberry: At first sight, strawberry would seem to be easily standardised. Not so. It is one of the most difficult flavours to fit into its many regional variations. In the US, strawberry is generally sweet and slightly jammy. Green notes are not liked. In most of Europe the preferred character is fresh and distinctly green. Within Europe, the French taste is P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

20 Food Flavour Technology

for a pronounced jasmine note and the Spanish taste is for strawberry jam. In Asia the preferences are more abstract and old-fashioned because of relative unfamiliarity with the real fruit. Vanilla: Alcoholic genuine vanilla extract, with rum and fruity overtones, deﬁnes the US taste. In France the taste veers towards a creamy hawthorn note, similar to the rare Tahitian natural vanilla extract. In Germany a hint of balsam is appreciated, and in much of northern Europe, a simple vanillin taste is preferred. The UK preference is for a distinctly buttery note.

1.9.2 Abstract flavours Most flavours have an obvious natural target. The flavour may be realistic or have a degree of abstraction, especially for children’s products. Many flavours, such as lemon-lime, are blends of recognisable natural targets. Very few flavours are basically abstract. The most important examples are cola and tutti-frutti. Cola flavours are all quite complex blends. The main characters are distilled lime oil, cassia oil, nutmeg oil and vanilla extract. In some instances, the caramel colour also contributes a characteristic flavour. The flavour ages very quickly in the bottling syrup because of the high level of phosphoric acid. Matching cola flavours is especially difficult because of the noticeable change in flavour in the syrup and the need to age the syrup before evaluation. Tutti-frutti flavours, in contrast, are quite widely varied. The main characters are banana, orange, pineapple, vanilla and berries. They can be classified into two broad types Ð those based on banana and those based on berries. The banana family is usually built around isoamyl acetate and the berry family is usually built around ␣-ionone. There are other interesting, but less commercially important, abstract flavours. Root beer was originally based on sassafras oil, vanilla and methyl salicylate, as well as a host of minor ingredients. Sassafras oil has not been used for many years (because it contains safrole over which there are safety concerns) and the substitutes vary in effectiveness. Sarsaparilla and dandelion and burdock are similar products. Cream soda flavours were similarly modified in line with regulatory requirements and now consist of vanillin with lactone-based hay and cream notes. Cachou flavours are intensely perfumed and are often based on combinations of violet, rose and musk. Another, accidental, category of abstract flavours is the group of the best of the flavours from the early years of the industry. They were often not exactly recognisable, but were triumphs of artistry over paucity of raw materials and became standards in their own right. Blackcurrant and cherry flavours are good examples. Early blackcurrant flavours were based on one simple raw material, buchu leaf oil. This raw material reproduced nothing more than the ‘catty’ character of blackcurrants and it was unpleasantly harsh and minty. Creativity improved this simple base by adding a mixture of ingredients, most importantly vanillin and ␣-ionone. The harshness was covered and the pleasant abstract confection is still the basis of many blackcurrant-flavoured foods today. The character is not close to blackcurrants, but it is an instantly recognisable standard. Early cherry flavours relied on benzaldehyde in a similar way and were not particularly attractive. Creativity gradually improved on this ingredient and evolved a complex flavour with anisaldehyde and para-methoxy acetophenone as the main additions. This flavour type is exceptionally attractive. It is not close to any known variety of cherry, but it is still immensely popular. P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 21

1.9.3 Matching No project is less welcome to the typical flavourist than one that requires matching. The very idea of matching someone else’s work is profoundly unattractive. Improving on someone else’s work is quite another matter and represents a real challenge, but simple matching is boring. Behind the flavourist’s manifest hostility is not just the simple lack of challenge and novelty; there is also the commercial fact that matching work represents by far the least rewarding use of creative time. Most matching projects are obtained by novice sales staff as a way of gaining entry to the account. Most successful matches are simply used to pressure the existing supplier to reduce prices. Very few product managers will risk changing the flavour of a successful consumer product to save a few cents, especially when it must involve expensive consumer trials. Some very successful flavour companies will generally not accept matching projects, and it is not evident that their customer standing has suffered as a result. Despite all the objections, there are occasionally good reasons to carry out matching work. The customers may have become genuinely hostile to their current supplier and wish to change at any price. The need to carry out matching may also derive from a reduction in the number of suppliers in a core supplier programme. The customer may wish to duplicate a competitor’s existing consumer product, although this type of project is more often directed towards beating the existing product. Matching generally starts with an analysis. Normally, it will be a direct analysis of the existing flavour, if the customer is serious. In some cases it will be a consumer product. Once the analysis has been completed, it should be reconstituted and reanalysed (in consumer product if necessary) to pick up errors in identification and quantification. This corrected analysis represents the starting point for the flavourist. The target flavour will often contain natural extracts and essential oils, so the first job is to allocate all the components derived from natural sources correctly. This is usually a matter of experience and the analyst should also be able to provide some guidance. The trickiest part is usually trying to determine which processes (solvent extraction, concentration, etc.) have been applied to the natural raw materials. Trial and error, as well as quite a few reanalyses, overcome this hurdle. The major chemical components should be identified and quantified from the analysis. It is often helpful to carry out liquid chromatography of the main chemicals to improve the accuracy of the quantification. One issue that is often forgotten is the need to get the solvents in liquid flavours identified and quantified correctly. Flavourists sometimes assume that they will not be important in the end product. That is not always true, but incorrect solvent balance always hinders rapid evaluation of matches on blotters. All that remains are the trace components. Unfortunately, this is often where most of the problems lie. Traces of sulfur chemicals, for example, can be very difficult to pick up on analysis, but can be a vital part of the flavour. Matches are often carried out under severe time pressure and it is easy to become stale and run out of ideas. Involving other flavourists is a must and can be especially helpful in generating ideas about the identity of missing trace components. It is also useful to have available a bank of analyses of flavours from the same competitor. In many companies the same ideas are trotted out with surprising regularity. Sensory panel work is essential to validate the accuracy of the final match. It may also help persuade the customer to accept the change. Caution should be exercised about blindly P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

22 Food Flavour Technology

accepting routine panel results. If a very close match is required, an expert panel may be needed.

1.9.4 Customers A flavour is, sadly, nothing, if it is not sold. Part of the tremendous ‘buzz’ of being a successful flavourist is the feeling of having created something really good. The other part of the ‘buzz’ is seeing your product on the supermarket shelves, enjoyed by thousands, or perhaps millions, of people. The best flavourists do not divorce themselves from customer involvement and the art of selling. Most successful flavourists reach the inescapable conclusion that they are the best judge of the market and the best flavour for a specific project. This conclusion has some basis in fact, but the sad truth is that customer involvement is usually the key to success. Customers usually do know their own market best and especially their own brands. They should be encouraged to guide the creative process and take a genuinely active part in the overall profile of the finished flavour. One additional advantage of this approach is that the customers buy into the process and regard the resultant flavour as ‘theirs’, not without some justification. The only barrier to this approach is the communication problem. If the project has passed through intermediaries (sales, marketing, etc.) then communication is very difficult, however thoroughly the project information has been gathered. The use of intermediaries also wastes a considerable amount of time that could otherwise be put to good creative use. The only practical solution is direct contact between the flavourist and the application specialists in the customer’s laboratories. Descriptive terms need to be defined and understood and simple examples (such as cis-3-hexenol for ‘leaf green’) can help a great deal. Knowledge of the customer’s application processes is also needed, and the involvement of an application specialist from the flavour company is often vital. Problems are often caused by the interactions between the flavour and the application ingredients and processes. In many cases a small change to the customer’s formulation or process can save the day. Sensory evaluation can play a useful part if it is used in the right context. This is particularly true when expert panels are used to divine the precise flavour profiles preferred by a target market segment. This is usually a specialised area where the flavour company probably has more depth of knowledge than the customer. Most projects go through a number of iterations before they are concluded successfully, and it is vital in this process to remember that ‘the customer is always right’.

1.10 SUMMARY

Flavour creation is still more of an art than a science. Science provides a vital understanding of nature and a broad palette of raw materials. Science may also provide insights into the preferences of a target group of consumers and some understanding of the mechanisms of taste and smell, but science cannot yet replace the intuitive, creative skills of a good flavourist. Readers requiring more information on the art of flavour creation can find more information in Flavor Creation (Wright, 2004). P1: SFK/UKS P2: SFK c01 BLBK221-Taylor/Linforth November 17, 2009 16:27 Printer Name: Yet to Come

Creating and formulating ﬂavours 23

REFERENCES

Labbe, D., Rytz, A., Morgenegg, C., Ali, S. and Martin, N. (2007) Subthresold olfactory stimulation can enhance sweetness. Chem. Senses 32, 205Ð214. Keller, A. and Vosshall, L.B. (2004) Human olfactory psychophysics. Curr. Biol. 14, R875ÐR878. Shojaei, Z.A., Linforth, R.S.T., Hort, J., Hollowood, T.A. and Taylor, A.J. (2006) Measurement and manipulation of aroma delivery allows control of perceived fruit ﬂavour in low and regular fat milks. Int. J. Food Sci. Technol. 41, 1192Ð1196. Wright, J. (2004) Flavor Creation. Allured, Carol Stream, Illinois. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

2 Flavour legislation Jack Knights

2.1 INTRODUCTION

Flavourings represent a class of food additives that has had comparatively little legislation. This is probably due to the relatively large number of ingredients used in flavourings, the very small quantities involved and the difficulty of regulating added substances (which are also naturally present in foodstuffs) by analytical methods. It was estimated about 50 years ago that only 50 defined flavouring substances were used to the extent of over 50 kg/year in flavourings in the UK (a corresponding figure for the USA appears to be about 200 substances). These figures represent an average consumption of about 1 mg/person per year, at which level even many of the natural toxic materials, such as arsenicals or cyanides, are harmless. One of the characteristics of food additives in general, and flavouring substances in particular, is the low degree of risk they pose to the consumer in spite of the public perception of their undesirability. Around 1965, national regulators in several countries began to become interested in controlling food flavourings, although they had no evidence of untoward risks from their use. However, they felt that, without detailed knowledge regarding the composition of flavourings, they were laying themselves open to criticism by food activists. Initially, it was believed that all compound flavourings would have to be registered and approved by the appropriate authority, but the flavour industry managed to convince the legislative authorities that this was unnecessary and impractical. Currently, flavour legislation is carried out at national and international levels. The background history to flavour legislation is important as it explains why there are different systems in place and why they differ in their approach. The following sections will explain the situation in the USA, in the international arena and in Europe. Since the US system has been fairly stable for some years, in terms of its approach and procedures, the main emphasis is on the European Union (EU) situation where the formation of a large trading body over the last 40 years has led to attempts to harmonise legislation in many areas. EU harmonisation is a slow process and despite a directive in 1988, the legislation proposed in 2008 will not be fully in force until 2011, so it is necessary to consider existing legislation (Section 2.6.2) and future legislation (Section 2.6.4).

2.2 METHODS OF LEGISLATION

There are fundamentally two basic methods of legislation: P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 25

(1) Positive: Only those materials that are listed are permitted to be used, to the exclusion of all others. This type of legislation has the disadvantage that research into possible new materials is frustrated by the requirement for publication before use, unless the time required for commercialisation is built into the regulation. However, regulatory authorities generally prefer positive listing because they believe that it offers the consumer maximum protection and is relatively easy to police. (2) Negative: All materials are allowed except those that are listed. In the case of food legislation, this needs to be supplemented by a statement that none of the materials used are injurious to health. This firmly puts the requirement for safety of the materials onto the manufacturer, but is strongly objected to by most national legislative bodies on the basis that they do not have control over what is used. The system encourages the research and development of novel materials to provide more authentic flavourings, to the overall benefit of the consumer.

The above-mentioned systems may be combined so that some groups of materials are controlled by positive lists while other groups are covered by negative lists. This is still the system in use in the EU, although it is gradually being superseded by total positive lists. The relative effects on research of positive and negative legislation systems are illustrated in Fig. 2.1. The figure shows the number of new flavour compounds discovered year by year on the basis of published data (patents and journal papers) and an estimate of the postulated numbers of new compounds, i.e. those additional compounds discovered by flavour houses but not published. The number of compounds added to the USA GRAS list is also shown. The difference in the number of compounds discovered and those accepted on the GRAS (positive legislation) list suggests that the GRAS process does not encourage the development of new flavouring compounds. In contrast, many of the new compounds found use as flavours in Europe, where the legislative system was more flexible.

Fig. 2.1 Total volatile compounds in food, 1965–1990, derived from Volatile Compounds in Food (TNO Biotechnology and Chemistry Institute, Utrechtseweg 48, 3700 AJ Zeist, The Netherlands), and from FEMA/GRAS listing, 1965–1990. ‘Postulated’ is estimated as 10% non-published ﬂavour industry research. , postulated; , published; , GRAS listed. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

26 Food Flavour Technology

Since 2000, most of the research into new flavouring substances by the flavour industry has been abandoned due to the impending provision of published positive lists, the loss of confidentiality and the exorbitant costs of toxicological clearance of very minor food-related substances.

2.3 LEGISLATION IN THE UNITED STATES

The system that has been adopted in the USA since 1965 is the positive list. The initial work was undertaken by the USA Food and Drug Administration (FDA) as part of Title 21 of the Code of Federal Regulations. Some 27 flavouring substances were evaluated and classified as generally recognised as safe (GRAS), presumably on the basis of long usage without untoward effect. At that time the Flavor and Extract Manufacturing Association (FEMA) proposed to assist the FDA by using independent experts to evaluate the other flavouring ingredients that were known to be in use at the time, classifying most of them as GRAS and allocating the numbers 2001-3124 to them (Hall and Oser, 1965, 1970). FEMA later set up a panel of independent experts to evaluate flavouring materials on behalf of the FDA (Hallagan and Hall, 1995). This group was later named FEXPAN and consists of independent international toxicologists who undertake the safety evaluation of novel flavouring ingredients. This procedure has continued to the present and there have been further 16 reports allocating GRAS status to a further 1542 flavouring ingredients (Oser and Hall, 1972, 1973, 1974, 1975, 1976, 1977, 1978, 1979; Oser et al., 1984, 1985; Burdock et al., 1990; Smith and Ford, 1993; Smith et al., 1996, 1997, 2001, 2003, 2005, 2009; Newberne et al., 1998, 1999, 2000; Waddell et al., 2007). It should be noted that GRAS listing applies to all ingredients for use in flavourings, including non-flavouring materials such as solvents, carriers, emulsifiers and antioxidants. The GRAS number makes no distinction between natural and artificial ingredients in the list; a given substance has the same entry irrespective of its status. However, foods (e.g. butter) may be used in the formulation of flavourings without being GRAS listed. An entry in the GRAS list is accompanied by average usual and average maximum levels in ppm (parts per million equivalent to milligram/kilogram) that may be used in 34 categories of foodstuffs. These levels are provided for novel flavour materials by the submitter and agreed by FEXPAN. In the USA, the only alternative designations are ‘natural flavor’ and ‘artificial flavor’, and the latter has to be displayed directly with the name of the food as well as in the ingredients list. It is therefore a highly undesirable designation for a food and is avoided wherever possible. The term ‘natural flavor’ or ‘natural flavoring’ is defined in Title 21 of the Code of Federal Regulations, Chapter 1 §101.22(a)(3) as the essential oil, oleoresin, essence or extractive, protein hydrolysate, distillate, or any product of roasting, heating or enzymolysis, which contains the flavouring constituents derived from spices, fruits or fruit juice, vegetable or vegetable juice, edible yeast, herbs, barks, buds, roots, leaves or similar plant materials, meat, seafood, poultry, eggs, dairy products or fermentation products thereof, whose significant function in food is flavouring rather than nutritional. Natural flavours include the natural essence or extractives obtained from plants listed in §182.10, 182.20, 182.40, 182.50 and Part 184 of this chapter and the substances listed in §172.510 of this chapter (Code of Federal Regulations, 1990, Title 21). P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 27

It has become normal in the USA to designate various categories of natural flavourings as follows: r FTNF (from the named fruit): As the name suggests, these consist solely of extracts or distillates derived from the named fruit. For instance, strawberry FTNF could consist of concentrated strawberry juice with added strawberry distillate. It may not contain material from any other natural source. Flavourings of this category tend to be very expensive in r use (they are usually very weak) and not very stable. WONF (with other natural flavourings): These must contain more than 51% derived from the named source but may contain other natural flavouring ingredients. For instance, strawberry WONF could consist of 51% concentrated strawberry juice fortified with other fruit juices or natural chemicals. These flavourings are still expensive in use because of r the price of the named ingredient. Natural flavour: These must contain only natural ingredients, but the type or source is not defined. For instance, natural strawberry flavour may contain ingredients from any source so long as they are classified as natural.

The solvent or carrier has to be on the GRAS list (or it can be a food), but this does not affect the natural or artificial status of the flavouring. In recent years, a great deal of research effort has been directed to the preparation of natural versions of many of the significant, defined chemicals that are important in flavours. This has included such techniques as heat-induced reactions often at elevated temperatures and high pressures (see Chapter 3) and direct esterification using enzymes as catalysts (see Chapter 4), on the assumption that if the reactants are natural, then the finished product may be so designated. It should be noted that process flavours and smoke extracts are also regarded as natural under US law, whereas in Europe both these categories are separately designated and specifically non-natural.

2.4 INTERNATIONAL SITUATION: JECFA

JECFA is the Joint Food and Agriculture Organisation of the United Nations (FAO)/World Health Organisation (WHO) Expert Committee on Food Additives. It was set up in 1956 to evaluate the safety of food additives, residues of veterinary drugs in food, and naturally occurring toxicants and contaminants in food. JECFA serves as a scientific advisory body to FAO, WHO, their member states and the Codex Alimentarius Commission, primarily through the Codex Committee on Food Additives and Contaminants (CCFAC), regarding the safety of food additives including flavouring substances. The committee establishes acceptable intakes on the basis of toxicological data and related information on the substance being evaluated. It also develops general principles for assessing the safety of chemicals in food. The requirement to keep up-to-date with scientific developments in toxicology and related disciplines necessitates a constant review of evaluation procedures (Munro et al., 1999). JECFA has to date evaluated some 600 flavouring substances and, more importantly, indicated that many flavouring agents are used in such small quantity that a full evaluation may not be appropriate. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

28 Food Flavour Technology

JECFA evaluations and procedures are of international significance since they are not influenced by any particular group. JECFA evaluations are taken into account by both FEXPAN and the European Food Safety Authority (EFSA) in their evaluation of the safety of flavourings for the USA and Europe, respectively.

2.5 COUNCIL OF EUROPE

The Council of Europe ad hoc Working Party on Natural and Artificial Flavouring Substances was set up in 1965 as a subsidiary body to the Sub-Committee on the Health Control of Foodstuffs. The work carried out by this group had no legal standing but has been important in shaping the EU legislation described in Section 2.6. The initial aims of the Council of Europe were r to draw up a list of natural and artificial flavourings that could be used in foodstuffs r without hazard to public health; to draw attention to those flavourings that presented hazard to public health.

It is important in the context of the Council of Europe lists to understand the deﬁnitions they have used in classifying ﬂavourings in their initial publication (Council of Europe, 1974).

(i) A flavouring is a substance that has predominantly odour-producing properties and that possibly affects the taste. (ii) A natural flavouring is a substance obtained from vegetable and sometimes animal sources, exclusively through the appropriate physical processes. Those biological processes that occur spontaneously, and roasting, are assimilated to physical processes. (iii) An artificial flavouring is a substance that has flavouring properties and that has been obtained by a chemical process. This term includes (a) substances that exist in natural products; (b) substances not present, or as yet undiscovered in natural products.

These definitions are unsatisfactory in a number of ways and were modified in the 1981 publication (Council of Europe, 1981) as follows. These definitions refer to materials and flavouring substances considered acceptable as flavourings and do not necessarily apply to flavourings found in commerce:

(i) Flavouring properties are those that are predominantly odour-producing and that possibly affect the taste. (ii) A flavouring substance is a chemically defined compound that has flavouring properties. It is obtained either by isolation from a natural source or by synthesis. (iii) Natural sources of flavourings are products of plant or animal origin from which flavourings may be obtained exclusively through appropriate physical processes or by biological processes that occur spontaneously (e.g. fermentation). (iv) Natural flavourings may be defined as complex mixtures derived from natural sources that have flavouring properties. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 29

These definitions were carried on more or less unchanged in the 1992 edition (Council of Europe, 1992). In the first two editions of the list, the natural source materials were classified into the following groups:

N1: Fruits and vegetables or parts thereof consumed as food. No restriction on the parts used under the usual conditions of consumption is proposed. N2: Plants and parts thereof, including herbs, spices and seasonings commonly added to foodstuffs in small quantities, the use of which is considered acceptable with a possible limitation of an active principle in the final product. N3: Plants and parts thereof that, in view of their long history of use without evidence of acute untoward effects, are temporarily acceptable for continued use, in the traditionally accepted manner, in certain beverages and other foodstuffs. The Committee of Experts, however, stresses that insufficient information is available for an adequate assessment of their potential long-term toxicity. N4: Plants and parts thereof that are used for flavouring purposes at present but cannot be classified owing to insufficient information.

A number of N2 and N3 were listed with active principles such as coumarin or safrole, and these data have been the basis of Annex II in the EU Flavourings Directive. The 1992 edition does not have a section on natural source materials. In the 1974 and 1981 editions, the artificial flavouring substances were classified into lists as follows:

6. Flavouring substances that may be added to foodstuffs without hazard to public health. 7. Flavouring substances that may be temporarily added to foodstuffs without hazard to public health. 8. Flavouring substances not fully evaluated. This category was dropped from the 1981 edition despite containing a number of commonly used ﬂavouring substances, several of which were on the FEMA/GRAS list.

The 1994 edition took a totally different approach to the listing of artificial flavouring substances. Here, they were classified into chemical groups and for each substance a summary consisting of name, category, structure, Council of Europe, FEMA and CAS numbers, upper levels of use, natural occurrence and toxicity data was provided. A further series of reports on natural source materials has been published (Council of Europe, 2000, 2007, 2008). Guidelines on the production of flavouring preparations by enzymatic and microbiological processes (Council of Europe, 1998b) and tissue culture (Council of Europe, 1998a) also have been published. The Council of Europe Expert Committee has also produced guidelines on the production of thermal process flavourings, their ingredients and processing conditions (Council of Europe, 1995) as follows:

(1) Ingredients added prior to processing (a) A protein nitrogen source (b) A carbohydrate source P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

30 Food Flavour Technology

(c) A fat or fatty acid source (d) Other ingredients; herbs and spices and their extracts, water, thiamin, ascorbic, citric, lactic, fumaric, succinic and tartaric acids, guanylic and inosinic acids, inositol, lecithin, pH regulators and siloxanes as antifoaming agents (2) Ingredients added after processing (a) Flavourings (b) Authorised food additives (3) Processing conditions (a) The temperature of the product should not exceed 180◦C. (b) The duration of thermal processing should not exceed 15 minutes at 180◦C with correspondingly longer times at lower temperatures; heating up and cooling down time should be as short as practical (c) The pH during processing should not exceed the value of 8.0 (d) Flavourings and food additives should only be added after processing is complete; ﬂavour enhancers may be added before processing but only in minimum amounts necessary for ﬂavour generation (4) Purity criteria (a) Heavy metals as in EU directive (b) Benzo[a]pyrene not more than 1 µg/kg (c) Benzo[a]anthracene not more than 2 µg/kg (d) Amino-imidazo-azaarenes not detectable by the most sensitive routine method available (5) Safety evaluation. As a minimum requirement, the following toxicological studies should be performed: (a) A gene mutagenicity test (b) A test for chromosome damage in vivo or in vitro (c) A 90-day feeding study in animals

As stated at the beginning of this section, the Council of Europe lists have no legal standing but the work of the Council of Europe Committee of Experts has provided much of the basis for the current EU legislation on flavourings. Many members of the committee are also members of the Flavourings Sub-Committee of the EU Scientific Committee for Food, responsible for the evaluation of flavouring substances for the current EU positive list.

2.6 EUROPEAN COMMUNITY

2.6.1 Background – national to EU legislation Several countries in Europe have had flavouring regulations in some form going back many years. In the former West Germany, the initial regulation governing the use of artificial flavouring substances was enacted in 1959 (Essenzen VO, 1959). In this sense, the term ‘artificial’ refers to substances that are not chemically identical to materials present in natural products. This regulation listed only five artificial flavouring substances that were permitted and a short list of processed foods in which they were allowed. This regulation was amended in 1970 (Essenzen VO, 1970) to allow several more artificial substances in the same restricted list of foods. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 31

In the UK in 1965, the Ministry of Agriculture, Fisheries and Food (MAFF) published a report on flavouring agents (MAFF, 1965) and a further report in 1976 (MAFF, 1976). Both these reports recommended that flavouring ingredients should be controlled by positive listing, but no action was taken to implement this. However, the latter report acknowledged that substances chemically identical to natural substances (‘nature identical’) were different from artificial flavouring substances. It also drew attention to the work of the Council of Europe in the flavouring field. In 1988 the first attempt was made to harmonise the flavour regulations of the member states of the European Community. This took the form of a Council Directive on the approximation of laws of the Member States relating to flavourings and to source materials for their production (EC, 1988a). The primary reason for the Directive was the protection of human health, but within these limits, to also take account of economic and technical needs.

2.6.2 The 1988 Council Directive The 1988 Directive was intended to lay down a framework for ﬂavour legislation. It speciﬁed the procedures and principles that future legislation should consider and these included the following:

(a) General purity criteria (b) Definitions (c) Labelling (d) Appropriate provisions for the inventories created by Decision 88/389/EEC (EC, 1988b) (e) Specific purity and microbiological criteria (f) Limitation of certain components of vegetable or animal raw materials (g) Drawing up of lists of additives, solvents and diluents for flavourings

The following sections consider only those articles (EC, 1988a) that deal with nonprocedural matters and elaborate the principles and the potential effects of legislation on the food and ﬂavour industry.

2.6.2.1 Article 1 (definitions) Article 1 of the 1988 Council Directive attempted to define some terms that would help set the scope (Section 2.6.2.2) of the future legislation. The text below lists the key definitions that were proposed with some commentary to explain the reasons for writing specific definitions in this manner and how theses definitions (and omissions) led to a working framework:

1. This Directive shall apply to ‘flavourings’ used or intended for use in or on foodstuffs to impart odour and/or taste, and to source materials used for the production of flavourings. 2. For the purposes of this directive: (a) ‘Flavouring’ means flavouring substances, flavouring preparations, smoke flavourings, process flavourings or mixtures thereof.

Unlike most earlier definitions and normal commercial practice, this definition does not include the solvent or carrier but only the active part of the finished flavouring. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

32 Food Flavour Technology

(b) ‘Flavouring substance’ means a deﬁned chemical substance with ﬂavouring properties which is obtained: (i) By appropriate physical processes (including distillation and solvent extraction) or enzymatic or microbiological processes from material of vegetable origin either in the raw state or after processing for human consumption by traditional food- preparation processes (including drying, torrefaction and fermentation), (ii) By chemical synthesis or isolated by chemical processes and which is chemically identical to a substance naturally present in material of vegetable or animal origin as described in (i), (iii) By chemical synthesis but which is not chemically identical to a substance naturally present in material of vegetable or animal origin as described in (i).

These three definitions refer to flavouring substances that, in practice, are called natural, nature identical and artificial, respectively. The initial part of the definition refers to a defined chemical substance but is understood to include a defined mixture of defined substances, e.g. citral, which is a mixture of cis- and trans-isomers. The natural substance definition includes non-foods, e.g. cedarwood, and thus the nature-identical definition includes non-food-identical substances. There is no definition of ‘a traditional food preparation process’ or ‘appropriate physical process’, which leaves it open to individual interpretation:

(c) ‘Flavouring preparation’ means a product, other than the substances deﬁned in (i) above, whether concentrated or not, with ﬂavouring properties, which is obtained by appropriate physical processes (including distillation and solvent extraction) or by enzymatic or microbiological processes from material of vegetable or animal origin, either in the raw state or after processing for human consumption by traditional food-preparation processes (including drying, torrefaction and fermentation).

This definition covers such products as essential oils, concentrated essential oils, essential oil terpenes and isolates, oleoresins, resinoids, absolutes, extracts and tinctures of natural source materials (including the extraction solvent if this is a flavour carrier), distillates of natural source materials and fruit juices either concentrated or not used for their flavouring properties. It suffers from the same imprecision as in the foregoing paragraph:

(d) ‘Process ﬂavouring’ means a product which is obtained according to good manufacturing practices by heating to a temperature not exceeding 180◦C for a period not exceeding 15 minutes a mixture of ingredients, not necessarily themselves having ﬂavouring properties, of which at least one contains nitrogen (amino) and another is a reducing sugar.

This definition is defective in that it does not apply to the majority of commercial process flavours, which are heated for 1Ð4 hours, even though at lower temperatures. By negotiation with the commission, it has been unofficially agreed that longer times at lower temperatures are appropriate and a doubling of the time for each 10◦C decrease in temperature is acceptable. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 33

The deﬁnition does not specify what other ﬂavouring ingredients may be added or whether sources that provide reducing sugars are covered.

(e) ‘Smoke ﬂavouring’ means a smoke extract used in traditional foodstuffs smoking processes.

This is an impossible deﬁnition since traditional smoking processes do not use smoke extract. It is assumed that it refers to smoke generated by a method similar to that used in traditional smoking processes that is then condensed or absorbed in a carrier to form a smoke extract. This is the meaning that was used by the European ﬂavour industry but has now been amended by the 2003 Smoke Extract legislation (see Section 2.6.3).

3. Flavourings may also contain foodstuffs as well as additives necessary for the storage, use, dissolution or dilution, or processing aids where these are covered by other Community provisions.

These materials are not within the ambit of paragraph 2(a) above and thus do not form part of the ﬂavouring but are additives to it.

2.6.2.2 Article 2 (scope) Having set the deﬁnitions in the previous section, the scope of the directive was then described as follows:

The Directive shall not apply to: r Edible substances and products intended to be consumed as such, with or without reconsti- r tution. r Substances which have exclusively a sweet, sour or salt taste. Material of vegetable or animal origin, having inherently ﬂavouring properties, where they are not used as ﬂavouring sources.

The first bullet point excludes such products as fruit juices and the third bullet herbs and spices. The position of flavour enhancers such as monosodium glutamate and other amino acids is not clear. They are not really in accordance with the second bullet point since they certainly have some taste apart from being salty. The latest inventory of flavouring substances includes some amino acids and they are also included in the proposed raw materials for the preparation of process flavourings.

2.6.2.3 Article 4 (restricted substances) As described in Section 2.2, the directive included a general statement that ﬂavourings should be safe for human consumption and that certain categories of compounds should be avoided because of their toxicity. Article 4 describes these limitations as follows: P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

34 Food Flavour Technology

Table 2.1 Maximum limits for certain undesirable substances present in foodstuffs as consumed as a result of the use of ﬂavourings (EC, 1998; Annex I).

Substance Foodstuffs (µg/kg) Beverages (µg/kg) Benzo[a]pyrene 0.03 0.03 Benzo[a]anthracene 0.06 0.06

Member States shall take all measures necessary to ensure that:

(a) Flavouringsr do not contain any element or substance in a toxicologically dangerous quantity Subject to any exceptions provided for in the specific criteria of purity referred to in Article 6(2) third indent, they do not contain more than 3 mg/kg of arsenic, 10 mg/kg of lead, 1 mg/kg of cadmium and 1 mg/kg of mercury. (b) The use of flavourings does not result in the presence in foodstuffs as consumed of undesirable substances listed in Annex I [see Table 2.1] in quantities greater than those specified therein.

These limits are extremely low when compared with the levels of the same compounds in the surface layers of smoked or roasted meats (around 1 ppm).

(c) The use of flavourings and of other food ingredients with flavouring properties does not result in the presence of substances listed in Annex II [see Table 2.2] in quantities greater than those specified therein.

Annex II represented the position as at March 2001. Subsequently, capsaicin, estragole and methyleugenol have been added to the list.

2.6.2.4 Article 5 (inventories) This article is concerned with making appropriate decisions concerning the proposed inventories of ﬂavour compounds created by Council Decision 88/389/EEC (EC, 1988b) as follows:

The Commission shall, within 24 months of the adoption of this Decision and after consultation of the Member States, establish an inventory of:

(a) Flavouring sources composed of foodstuffs and of herbs and spices normally considered as foods (b) Flavouring sources composed of vegetable or animal raw materials not normally considered as food (c) Natural flavouring substances (d) Synthetic flavouring substances chemically identical to substances in foodstuffs, herbs and spices (e) Synthetic flavouring substances chemically identical to substances in vegetable or animal raw materials not normally considered as foodstuffs, herbs and spices, artificial flavouring substances (f) Source materials used in the production of smoke and process flavourings and the reaction conditions under which they are prepared P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 35

Table 2.2 Maximum limits for certain substances obtained from flavourings and other food ingredients with flavouring properties present in foodstuffs as consumed in which flavourings have been used (EC, 1998; Annex II).

Foodstuffs Beverages Exceptions and/or special Substance (mg/kg) (mg/kg) restrictions Agaric acid 20 20 100 mg/kg in alcoholic beverages and foodstuffs containing mushrooms Aloin 0.1 0.1 50 mg/kg in alcoholic beverages ␤-Asarone 0.1 0.1 1 mg/kg in alcoholic beverages and seasonings used in snack foods Berberine 0.1 0.1 10 mg/kg in alcoholic beverages Coumarin 2 2 10 mg/kg in certain types of caramel confectionery 50mg/kginchewinggum 10 mg/kg in alcoholic beverages Hydrocyanic acid 1 1 50 mg/kg in nougat, marzipan or its substitutes or similar products 1 mg/% alcohol by volume/kilogram in alcoholic beverages 5 mg/kg in canned stone fruit Hypericine 0.1 0.1 10 mg/kg in alcoholic beverages 1 mg/kg in confectionery Pulegone 25 100 250 mg/kg in mint- or peppermint-ﬂavoured beverages 350 mg/kg in mint confectionery Quassine 5 5 10 mg/kg in confectionery in pastille form 50 mg/kg in alcoholic beverages Safrole and 1 1 2 mg/kg in alcoholic beverages with less isosafrole than 25% alcohol by volume 5 mg/kg in alcoholic beverages with more than 25% alcohol by volume 15 mg/kg in foodstuffs containing mace and nutmeg Santonin 0.1 0.1 1 mg/kg in alcoholic beverages with more than 25% alcohol by volume Thujone (␣ and ␤) 0.5 0.5 5 mg/kg in alcoholic beverages with less than 25% alcohol by volume 10 mg/kg in alcoholic beverages with more than 25% alcohol by volume 25 mg/kg in foodstuffs containing preparations based on sage 35mg/kginbitters

None of the above-listed substances may be added as such to flavourings or to foodstuffs. They may be present in a foodstuff either naturally or following the addition of flavourings prepared from natural raw materials. Groups (c)Ð(e) have been combined into a single inventory under Commission Regulation 2232/96 (EC, 1996). This regulation required member states to notify the commission of those flavouring substances allowed in their territory. In practice the local flavour industry, under the guidance of the European Flavour and Fragrance Association (EFFA), supplied a list of the substances that were being used in their territory to the local regulatory organisation for that P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

36 Food Flavour Technology

body to examine and, if appropriate, pass on to the commission by 23 November 1996. The Commission was required to consolidate the submissions into a single inventory representing the flavouring substances that were currently being used in the EU (EC, 1999). It was permitted to disclose substances under code to maintain confidentiality of their identity (EC, 1998), but this route has been taken only for about 12 substances. The Standing Committee for Foodstuffs was required to examine and, if necessary, amend the inventory by 23 September 1999. The final inventory contained approximately 2500 substances and contained almost 500 substances that were not on the US GRAS lists. The final permitted list would be dependent on safety evaluation of each individual substance by the Scientific Committee for Food. This was delegated to the flavouring subcommittee, which examined the substances by chemical grouping under the Scientific Co-operation Procedure (SCOOP) (EC, 1994). The data required for evaluation were provided by EFFA on the form illustrated in Fig. 2.2. The substances contained in the inventory were meant to be permitted in each member state of the EU until a final permitted list was established in about 2005. However, both Germany and Italy maintained their restricted lists of artificial flavouring substances that were in force before 1988. Some preliminary progress has been made on group (f). The European Commission has not yet published any guidelines on process flavourings, but it is likely that they will follow the Council of Europe recommendations. The flavour industry stance, as agreed by International Organisation of the Flavour Industry (IOFI, 1989), was somewhat similar to those guidelines except for item ‘1 (d) Other ingredients’ where ‘herbs and spices and their extracts’ was expanded to include ‘flavouring substances identified therein’. There can be no argument that this significantly increased the risk, since the presence of these materials is already permitted owing to their occurrence in herbs and spices. An initial draft document concerning smoke flavourings (EC, 2001) set out the types of wood that could be used to produce the smoke used in the preparation of the smoke extract, the conditions under which the smoke could be generated, the general specification for the product and the toxicological data necessary for its approval. The way the document was written suggested that specific smoke flavourings from specified suppliers were the only ones that would be approved. This is currently the position in Sweden, where there is a positive list of named smoke flavourings to the exclusion of all others. This is not in accord with the Community anticompetition legislation.

2.6.2.5 Article 6 (additives for flavourings) As mentioned previously, legislation should consider not only the active flavouring compounds but also the other materials used to preserve, stabilise and deliver the flavourings in foods. This article requires that the following lists of authorised substances be agreed: r r Additives necessary for the storage and use of flavourings r Products used for dissolving and diluting flavourings Processing aids where these are not covered by other community provisions

The ﬁrst two bullet points above have proved extremely difﬁcult in obtaining agreement with the member states. Discussions have been taking place with the commission since 1990 to try to thrash out a system that will satisfy all parties involved. These have consisted of permitting P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 37

Flavouring Substance Butyl but-2-enoate

EFFA No 0301 CoE No - CAS No 591-63-9

FEMA No - EINECS No 2-966X JECFA No -

FL No 09.324

Chemical name on register Butyl but-2-enoate

IUPAC name 2-Butenoic acid, (E)-, butyl ester

Synonyms Crotonic acid, butyl ester, (E)-

Physical form Liquid

Boiling point ( °C, 76 Torr) 80(42 T) Melting point (°C)

Ref. index lower value (20°C) 1.425 Ref. index upper value 1.431

Density lower value (25°C) Density upper value

Minimum assay value 95 Identity test B1’721’863

Solubility Insoluble in water Solubility in ethanol 1 mL in 1 mL 95%EtOH

Chemical group 1 Origin (Codex, CAC) nat./nat.ident.

Food category Beverages, excluding dairy products Normal dosage in ppm 5

Maximum dosage in ppm 25

Volume of use by EFFA kg/a 14

Sensory Colourless liquid with fruity banana odour description

Food source Pawpaw (Asimina Triloba Dunal.) in ppm 0.024

Interpretive study 2000-3

Recent studies

Butyl but-2-enoate

Volume of use by EFFA: kg/year 14

Application

Fig. 2.2 Example of initial data submission required by SCOOP for ﬂavouring substances, and volume of use by EFFA. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

38 Food Flavour Technology

Food category Normal dosage Maximum dosage (ppm) (ppm)

Dairy products (excluding those listed 735 elsewhere)

Fats and oils and fat emulsions 5 25

Edible ices including sherbets and water 10 50 ices

Processed fruits and vegetables 7 35

Sugar and chocolate confectionery 10 50

Cereals and cereal products 5 25

Bakery wares 10 50

Meat and meat products 2 10

Fish and fish products 2 10

Soups, sauces and seasonings 5 25

Foodstuffs intended for particular nutritional 10 50 purposes

Beverages excluding dairy products 5 25

Ready-to-eat savouries 20 100

Composite foods (e.g. casseroles) 5 25

Fig. 2.2 (Continued)

automatically all the additives that are permitted for use in food or those that are quantum satis and the others by positive list. Neither of these alternatives is regarded as satisfactory, mainly on the grounds that they would allow additives that are not permitted in a particular foodstuff to be added without declaration through the ﬂavour composition. The discussion is ongoing and agreement is of considerable importance since individual member state rules are being used to frustrate free trade in ﬂavourings. Article 6 also deals with methods of sampling and analysis and microbiological criteria, none of which has been implemented to date.

2.6.2.6 Article 9 (labelling) One of the biggest consumer issues with food is understanding its composition and safety. Consumer pressure groups are demanding more information on packs of food so that consumers can ‘make a choice’ although the history of food labelling in the EU shows how P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 39

these good intentions can backfire. The introduction of the E number system in the EU was designed to give consumers confidence that additives in their food have been fully tested and are safe for consumption. Instead, the popular press convinced consumers that E numbers meant ‘chemicals’ in their food and that E numbers should be avoided. This has led to food manufacturers in several countries either using trivial names in the ingredient list, e.g. ascorbic acid (which is perceived as good by the consumer) rather than E 300, or simply by removing additives with E numbers. It is unlikely that these moves by the food industry have had any significant benefit for the consumer. Applying this background to flavourings that contain many different chemicals creates serious problems both for regulation and for consumer understanding. Article 9 therefore attempted a solution, which provided consumer safety without entailing a huge long list of compounds on packs of finished food.

(1) Flavourings not intended for sale to the final consumer may not be marketed unless their packaging or containers bear the following information, which should be easily visible, clearly legible and indelible: (a) The name or business name and address of the manufacturer or packer, or of a seller established in the community. (b) Either the word ‘flavouring’ or a more specific name or description of the flavouring.

This enables descriptions such as ‘lemon ﬂavouring’ or ‘natural ﬂavouring’ if appropriate.

(c) Either the statement ‘for foodstuffs’ or a more specific reference to the foodstuff for which the flavouring is intended. (d) A list in descending order of weight of the categories of flavouring ingredients presentr as follows: r Natural flavouring substances r Nature-identical flavouring substances r Artificial flavouring substances r Flavouring preparations r Process flavourings Smoke flavourings (e) In the case of mixtures of flavourings with other substances referred to in Article 6, ar list in descending order of weight in the mixture of: r The categories of flavourings classified as in (d) above The names of each of the other substances or materials or their E numbers

It is not clear whether this means a single list or two lists. In general, the ﬂavour industry uses the latter.

(f) An indication of the maximum quantity of each component contained in Annexes I and II or sufficient information to enable the food producer to comply with the limits for the finished food. (g) An indication identifying the consignment. (h) The nominal quantity in units of mass or volume. (2) The word ‘natural’, or any other word having substantially the same meaning, may only be used for flavourings in which the flavouring component consists of exclusively flavouring preparations and/or natural flavouring substances. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

40 Food Flavour Technology

Paragraph 2 means that flavourings containing process flavourings or smoke flavourings cannot be labelled as natural, in contrast to the US position. The above requirement is complied with in all EU member states except Italy, where flavourings containing nature identical flavouring substances are also designated as natural. If the sales description of the flavouring contains a reference to a foodstuff or flavouring source, the word ‘natural’, or any word having substantially the same meaning may not be used unless the flavouring components have been isolated solely or almost solely from the flavouring source concerned. This means that in ‘natural lemon flavouring’ the flavouring components have to be natural and derived solely or almost solely from lemon. The term ‘almost solely’ is not defined but, by agreement with the 1988 Commissioner, a level of greater than 90% from the named source would be acceptable to both the commission and industry. This leaves the question of how to designate natural flavourings that are not derived solely or almost solely from the named source. This is of course a matter for the individual Member States. In the UK, wordings such as ‘natural lemon flavour flavouring’ or ‘natural flavouring lemon type’ have been used, but the matter has never been tested in the courts to decide whether these are acceptable.

(3) By derogation from paragraph 1, the information required by paragraphs 1(d), (e) and (f) may appear merely on the trade documents relating to the consignment supplied prior to the delivery, provided the phrase ‘intended for the manufacture of foodstuffs and not for retail’ appears in a conspicuous part of the packaging or container of the products in question.

It is not clear whether this phrase should replace that in paragraph 1(c) or be in addition to it. It is normal for the former to be used.

(4) Member States shall refrain from laying down requirements more detailed than those contained in this article, concerning the manner in which the particulars provided for are to be shown. These particulars shall be given in terms easily understood by the purchaser. This shall not prevent them being given in various languages.

This provision has been ignored by at least one member state by invoking the Reserved Dairy Descriptions Directive to prevent the use of milk, cream, butter or cheese in the description of ﬂavourings. After complaints by other member states this was abandoned, as was the insistence that solely their own language be used on label of ﬂavourings for their country.

2.6.3 Smoke flavourings 2003 Directive In 2003, the European Parliament and the Council produced a new document (EC, 2003), covering smoke flavourings as defined in Directive 88/388/EEC. It was felt that the protection of human health with respect to smoke flavouring was inadequate and these should be regulated separately. The intention of the document was to authorise and regulate certain smoke preparations to the exclusion of all others. The regulation defined ‘primary smoke condensates’ and ‘primary tar fractions’ and referred to derived products therefrom. The conditions of generation of the primary products were defined as controlled burning, dry distillation, or treatment with superheated steam in a controlled oxygen environment with a maximum temperature of 600◦C. The smoke is condensed and may be physically treated P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 41

to achieve phase separation, isolation and/or purification. The primary products must not contain more than 10 ␮g/kg benzo[a]pyrene and 20 ␮g/kg benz[a]anthracene. In order to obtain authorisation for a primary product, an application must be made to the competent authority in a member state giving information on the type of wood used, the detailed production process, the quantitative and qualitative composition including its variability and analytical methodology. Also required is information on the intended use levels in, or on, specific food or food categories and detailed toxicological data. The data provided will be evaluated by EFSA that will provide an opinion as to the safety of the primary product and any conditions or restrictions to its use. The European Commission will take note of the EFSA opinion and, if acceptable, will authorise the use of the primary product, giving it a unique identifying code. This code has to be used in all transactions involving the primary product including product derived therefrom. The only stage at which this code does not have to be displayed is on food products intended for direct sale to the final consumer. If smoke flavourings are used as ingredients of other flavourings and impart a smoky flavour, they must be declared separately in the ingredients list. This must include their identifying code except in the ingredients list of food intended for direct sale to the final consumer.

2.6.4 Developments 2008 onwards In mid-2008, the Council of the European Union produced drafts of four regulations, commonly termed the ‘Food Improvement Agents Package’. These draft regulations mark a shift in the approach to flavour legislation which is now contained in an overall scheme that includes food additives and enzymes. The first of these documents, the Common Authorisation Procedure (EC, 2008a), describes the establishment of a common authorisation procedure for food additives, food enzymes and food flavourings after adoption of a Regulation of the European Parliament and of the Council. The second document, ‘Chapter I General Prin- ciples’ outlines what the regulation is about and introduces a common assessment and authorisation procedure (hereinafter referred to as the ‘common procedure’). The common procedures will apply to food enzymes (EC, 2008b), food additives (EC, 2008c) and food flavourings including food ingredients with flavouring properties (EC, 2008d) and specifically exclude smoke flavourings falling within the scope of Regulation (EC) 2065/2003 (as discussed in Section 2.6.4). The third document ‘Chapter II Common Procedure’ lays down the procedural arrangements for updating the lists of authorised substances and the criteria, according to which, substances can be included on the lists. The common procedure may be started on the ini- tiative of the commission or following an application, either by a member state or by an interested party. The commission shall seek the opinion of the EFSA, but this may not be required if the update does not have an effect on human health. EFSA is required to give its opinion within 6 months of a valid application and forward it to the commission, the member states and where applicable, the applicant. This period may be extended where EFSA requests additional information by agreement with the applicant. Within 9 months of EFSA giving its opinion, the commission shall submit a draft regulation updating the community list. There is a provision for confidentiality with the submission but since this may not include the name and clear description of the substance, it is of little value particularly in the cases of flavouring substances. The fourth document ‘Flavourings Regulation’ is specific for flavourings (EC, 2008d) and was published on 31 December 2008 but is intended to apply from 20 January 2011. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

42 Food Flavour Technology

Certain aspects apply from 20 January 2009 and compliance with the community list shall apply from 18 months after its date of application. In some ways this regulation is similar to Directive 88/388 (EC, 1988a) but with signiﬁcant differences, and a comparison of the previous sections with those below will highlight the changes.

2.6.4.1 Article 2 (scope) The scope has been modified compared to that described in Section 2.6.2.2 by including food ingredients with flavouring properties. The regulation also applies to foodstuffs containing flavourings and/or food ingredients with flavouring properties as well as to their source materials. It does not apply to smoke flavourings or substances having exclusively a sweet, sour or salty taste, raw foods, non-compounded foods or mixtures of spices and/or herbs, mixtures of tea and mixtures for infusion as such as long as they have not been used as food ingredients. In the new regulation, ‘flavourings’ may be made or consist of flavouring substances, flavouring preparations, thermal process flavourings, smoke flavourings (even though the regulation does not apply to them) and two new categories Ð ‘flavour precursors’ and ‘other flavourings’ Ð and mixtures of the above-mentioned categories.

2.6.4.2 Article 3 (deﬁnitions) The following are the key differences between the situation explained in Section 2.6.2.1 and the 2008 proposals:

‘Flavouring substance’ is defined as before but no distinction is made between nature identical and artificial flavouring substances. ‘Natural flavouring substance’ is defined as before but with the proviso that if it is prepared for human consumption, it must be by using one of the processes defined in Annex II (see Table 2.3). ‘Flavouring preparation’ is defined as before but with the above Annex II proviso. These do not require evaluation and approval if they are derived from food.

Table 2.3 List of traditional food preparation processes (ANNEX II; EC 2008a).

Chopping Coating Heating, cooking, baking, frying (up to Cooling ◦ 240 C at atmospheric pressure) and ◦ pressure cooking (up to 120 C) Cutting Distillation/rectiﬁcation Drying Emulsiﬁcation Evaporation Extraction, incl. solvent extraction in accordance with Directive 88/344/EEC Fermentation Filtration Grinding Infusion Maceration Microbiological processes Mixing Peeling Percolation Pressing Refrigeration/Freezing Roasting/Grilling Squeezing Steeping P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 43

Table 2.4 Conditions for the production of thermal process ﬂavourings and maximum levels for certain substances in thermal process ﬂavourings (ANNEX V; EC, 2008a).

Part A: Conditions for the production ◦ (a) The temperature of the products during processing shall not exceed 180 C ◦ (b) The duration of the thermal processing shall not exceed 15 minutes at 180 C with correspondingly longer times at lower temperatures, i.e. a doubling of the heating time for each decrease of temperature ◦ by 10 C, up to a maximum of 12 hours (c) The pH during processing should not exceed the value of 8.0 Part B: Maximum levels for certain substances Substance Maximum levels (µg/kg) 2-Amino-3,4,8-trimethylimidazo[4,5-f ] quinoxaline (4,8-DiMeIQx) 50 2-Amino-1-methyl-6-phenylimidazol [4,5-b] pyridine (PhIP) 50

‘Thermal process flavouring’ is defined as before but is subdivided by its ingredients being food and/or source materials other than food. If the ingredients are food and the finished flavouring complies with Annex V (Table 2.4), these do not require evaluation and approval. All other thermal process flavourings do require evaluation and approval. ‘Smoke flavouring’ is defined as in Regulation (EC) 2065/2003 (Section 2.6.3). ‘Flavour precursor’ shall mean a product, not necessarily having flavouring properties itself, intentionally added to food for the sole purpose of producing flavour by breaking down or reacting with other components during food processing. It may be obtained from food or source materials other than food. If obtained from food, these do not require evaluation and approval. ‘Other flavouring’ shall mean a flavouring added or intended to be added to food in order to impart odour and/or taste not covered by the other definitions. ‘Food ingredient with flavouring properties’ shall mean a food ingredient other than flavourings, which may be added to food for the main purpose of adding flavour to it or modifying its flavour and which contribute significantly to the presence in food of certain naturally occurring undesirable substances. These do not require prior evaluation and approval. ‘Appropriate physical process’ shall mean a physical process that does not intentionally modify the chemical nature of the components of the flavouring, without prejudice to the listing of traditional food preparation processes in Annex II, and does not involve, inter alia, the use of singlet oxygen, inorganic catalysts, metal catalysts, organometallic reagents and/or UV radiation. The processes are listed in Table 2.3.

Flavourings may contain food additives as permitted by Regulation (EC) 1333/2008 Annex III Part 4 (EC, 2008c), and/or other food ingredients incorporated for technological purposes.

2.6.4.3 Articles 8 and 9 (evaluation and approval) All categories of flavourings listed in the foregoing section require prior evaluation and approval unless specifically defined out.

2.6.4.4 Article 6 (presence of certain substances) In a similar way to previously (Section 2.6.2.3, Table 2.2), there is a list of substances that may not be added to food as such (Table 2.5; Annex III EC, 2008d). There is also a list of certain P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

44 Food Flavour Technology

Table 2.5 Substances and limits in ﬂavourings (Annex III; EC, 2008d).

Part A: Substances that shall not be added as such to food Agaric acid Aloin Capsaicin 1,2-Benzopyrone, coumarin Hypericine ␤-Asarone 1-Allyl-4-methoxybenzene, estragole Hydrocyanic acid Menthofuran 4-Allyl-1,2-dimethoxybenzene, methyl eugenol Pulegone Quassin 1-Allyl-3,4-methylene dioxy benzene, safrole Teucrin A Thujone (alpha and beta)

Part B: Maximum levels of certain substances, naturally present in flavourings and food ingredients with flavouring properties, in certain compound food as consumed to which flavourings and/or food ingredients with flavouring properties have been added Maximum Name of the Compound food in which the presence of level substance the substance is restricted (mg/kg) ␤-Asarone Alcoholic beverages 1.0 1-Allyl-4-methoxybenzene Dairy products 50 Estragole Processed fruits, vegetables (including mushrooms, 50 fungi, roots, tubers, pulses and legumes), nuts and seeds Fish products 50 Non-alcoholic beverages 10 Hydrocyanic acid Nougat, marzipan or its substitutes or similar 50 products Canned stone fruits 5 Alcoholic beverages 35 Menthofuran Mint/peppermint-containing confectionery, except 500 micro breath-freshening confectionery Micro breath-freshening confectionery 3000 Chewing gum 1000 Mint/peppermint-containing alcoholic beverages 200 4-Allyl-1,2- Dairy products 20 dimethoxybenzene Methyleugenol Meat preparations and meat products including 15 poultry and game Fish preparations and fish products 10 Soups and sauces 60 Ready-to-eat savouries 20 Non-alcoholic beverages 1 Pulegone Mint/peppermint-containing confectionery, except 250 micro breath-freshening confectionery Micro breath-freshening confectionery 2000 Chewing gum 350 Mint/peppermint-containing non-alcoholic 20 beverages Mint/peppermint-containing alcoholic beverages 100 P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 45

Table 2.5 (Continued)

Quassin Non-alcoholic beverages 0.5 Bakery wares 1 Alcoholic beverages 1.5 1-Allyl-3,4- Meat preparations and meat products including 15 methylenedioxy benzene, poultry and game safrole Fish preparations and ﬁsh products 15 Soups and sauces 25 Non-alcoholic beverages 1 Teucrin A Bitter-tasting spirit drinks or bitter 5 Liqueurs with a bitter taste 5 Other alcoholic beverages 2 Thujone (alpha and beta) Alcoholic beverages except those produced from 10 Artemesia species Alcoholic beverages produced from Artemesia 35 species Non-alcoholic beverages produced from 0.5 Artemesia species Coumarin Traditional and/or seasonal bakery ware 50 containing a reference to cinnamon in the labelling Breakfast cereals including muesli 20 Fine bakery ware with the exception of traditional 15 and/or seasonal bakery ware containing a reference to cinnamon in the labelling Desserts 5

substances, naturally present in flavourings and food ingredients with flavouring properties, which are limited in certain compound foods to which they have been added (Annex III Part B). The maximum levels do not apply to compound foods that are not listed or where no flavourings have been added, and the only food ingredients with flavouring properties that have been added are fresh, dried or frozen herbs and spices.

2.6.4.5 Article 7 (the use of certain source materials) There is also a list of source materials that may not be used for the production of flavourings (Table 2.6, Annex IV Part A) and flavourings derived from certain source materials for which there are specified conditions of use (Annex IV Part B).

2.6.4.6 Article 10 (community list of flavouring and source materials) As stated earlier, all flavouring substances, whether natural or not, have to be evaluated by EFSA and approved by the European Commission as laid down in the common authorisation procedure (Regulation (EC) 1331/2008). The approved substances, including conditions of use if appropriate, will form the community list. Only flavouring substances that are contained in the community list may be used in flavourings. The community list of flavouring substances is likely to apply from the middle of 2012. The regulation refers to all the classes of flavouring ingredients requiring evaluation and approval, but to date no attempt has been made to undertake the evaluation of other than flavouring substances. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

46 Food Flavour Technology

Table 2.6 List of source materials to which restrictions apply for their use in the production of ﬂavourings and food ingredients with ﬂavouring properties (Annex IV; EC, 2008d).

Part A: Source materials that shall not be used for the production of flavourings and food ingredients with flavouring properties Source material Latin name Common name Tetraploid form of Acorus Tetraploid form of calamus L. calamus Part B: Conditions of use for flavourings and food ingredients with flavouring properties produced from certain source materials Source material Latin name Common name Conditions of use Quassia amara L. and Quassia Flavourings and food products Picrasma excelsa (Sw) with flavouring properties produced from the source material may only be used for the production of beverages and bakery wares Laricifomes officinales (Vill.: White agaric Flavourings and food products Fr) Kotl. Et Pouz or Fomes mushroom with flavouring properties officinalis produced from the source material may only be used for the production of alcoholic beverages Hypericum perforatum L. St John’s wort Teucrium chamaedrys L. Wall germander

2.6.4.7 Articles 14 and 15 (labelling of ﬂavourings not intended for sale to the ﬁnal consumer) The general labelling requirements are similar to the present legislation (Section 2.6.2.6) with the addition of a date of minimum durability or use by date.

2.6.4.8 Article 16 (specific requirements for the use of the term ‘natural’) This differs significantly from the present flavouring legislation and the new definition of natural is given below. The term ‘natural’ may only be used if the flavouring component comprises only flavouring preparations and/or natural flavouring substances. This is the same as previously. The term ‘natural flavouring substance(s)’ may only be used for flavourings in which the flavouring component contains exclusively natural flavouring substances. The term ‘natural’ may only be used in combination with a reference to a food, food category or a vegetable or animal flavouring source if the flavouring component has been obtained exclusively or by at least 95% w/w from the source material referred to. The present legislation requires 90% from the specified source material. The label description shall read ‘natural “food(s) or food category or source(s)” flavouring’. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 47

The term ‘natural “food(s) or food category or source(s)” flavouring with other natural flavourings’ may only be used if the flavouring component is partially derived from the source material referred to, the flavour of which can easily be recognised. The term ‘natural flavouring’ may only be used if the flavouring component is derived from different source materials and where a reference to the source materials would not reflect their flavour or taste.

2.6.4.9 Article 29 (designation of ﬂavourings in the list of ingredients on ﬁnished foods) Directive 2000/13/EC Annex III (Section 2.6.2.6) shall be replaced by the following:

(1) Flavourings shall be designated by the terms Ð ‘flavourings’ or a more specific name or description of the flavouring; Ð ‘smoke flavouring(s)’ if the flavouring component contains smoke flavourings and imparts a smoky flavour to the food.

This means that, if the flavouring contains a smoke flavouring as defined in Regulation (EC) 2065/2003 and imparts a smoky flavour to the food, both ‘flavourings’ and ‘smoke flavouring(s)’ must appear in the ingredients list on the food.

(2) The term ‘natural’ for the description of ﬂavourings shall be in accordance with Article 16 of Regulation (EC) 1334/2008.

2.7 CURRENT EU SITUATION AND THE FUTURE

The introduction of the 2008 draft Regulations means that EU flavour legislation is currently between the old and the new systems with new regulations being phased in up to 2012. This makes a complex situation for food and flavour manufacturers. The new regulations have cleared up some of the omissions in the earlier regulations and provided clearer definitions in some areas, but it is very difficult to write legislation that covers the huge range of flavour compounds found in foods and all the potential usages that may arise. Inevitably, there will be further amendments when such issues are recognised. There appears to be a further problem, namely, whether flavourings that comply with the new Regulation and not the old Directive are legal before the date of implementation of the new Regulation. This has yet to be resolved by the European Commission. It is widely recognised both among regulators and the international flavour industry that the lack of harmonisation of flavouring legislation across the world represents a significant barrier to trade. The acceptance in one country of a flavouring material banned in another cannot possibly be on the basis of safety; it is usually on the basis of commercial interest. The International Organisation of the Flavour Industry (IOFI) has been pressing legislators to accept flavouring materials on the basis of safety evaluations performed for other groups, e.g. the acceptance by European Scientific Committee for Food of evaluations done by FEXPAN. The international nature of the flavour industry is providing data to FEXPAN to enable the evaluation of the 500 or so additional flavouring substances that are on the European inventory so that they may be granted GRAS status for use in the USA. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

48 Food Flavour Technology

REFERENCES

Burdock, G.A., Wagner, B.M., Smith, R.L., Munro, I.C. and Newbeme, P.M. (1990) Recent progress in the consideration of flavoring ingredients under the Food Additive Amendment. FEMA GRAS Substances 15. Food Technol. 44(2), 82. Code of Federal Regulations Title 21 (1990) Food and drugs. Chapter 1, §101.22 (a) (3). Council of Europe (1974) Natural Flavouring Substances, Their Sources, and Added Artificial Flavouring Substances, Council of Europe, Maisonneuve, Paris. Council of Europe (1981) Flavouring Substances and Natural Sources of Flavourings, 3rd edn, Council of Europe, Maisonneuve, Paris. Council of Europe (1992) Flavouring Substances and Natural Sources of Flavourings, 4th edn, Council of Europe, Maisonneuve, Paris. Council of Europe (1995) Guidelines on the Production of Thermal Process Flavourings, Council of Europe Publishing, Paris, ISBN 92-871-2811-1. Council of Europe (1998a) Guidelines for Flavouring Preparations Produced by Plant Tissue Culture, Council of Europe Publishing, Paris, ISBN 978-92-871-3738-8. Council of Europe (1998b) Guidelines for Flavouring Preparations Produced by Enzymatic or Microbiolog- ical Processes, Council of Europe Publishing, Paris, ISBN 978-92-71-2586-6. Council of Europe (2000) Natural Sources of Flavourings Ð Report No. 1, Council of Europe Publishing, Paris, ISBN 978-92-871-4324Ð2. Council of Europe (2007) Natural Sources of Flavourings Ð Report No. 2, Council of Europe Publishing, Paris, ISBN 978-92-871-6156-7. Council of Europe (2008) Natural Sources of Flavourings Ð Report No. 3, Council of Europe Publishing, Paris, ISBN 978-92-871-6422-3. EC (1988a) Council directive of 22 June 1988 on the approximation of the laws of the member states relating of flavourings and to source materials for their production (88/388/EEC). Official Journal of the European Communities, No. L 184/61(15/7/88). See also UK implementation: The Flavouring in Foods Regulations S.I. 1992 No. 1971, The Flavouring of Food (Amendment) Regulations S.I. 1994 No. 1486. EC (1988b) Council decision of 22 June 1988 on the establishment, by the commission, of an inventory of the source materials and substances used in the preparation of flavourings (88/389/EEC). Official Journal of the European Communities, No. L 184/67 (15/7/88). EC (1994) European Commission decision of 20 September 1994 establishing the inventory and distribution of tasks to be undertaken within the framework of co-operation by member states in the scientific examination of questions relating to food (1994/652/EC). Official Journal of the European Communities, No. L 253/29 (29/9/94). EC (1996) European Parliament and Council Regulation (EC) of 28 October 1996 laying down a community procedure for flavouring substances used or intended for use in Foodstuffs (2232/96). EC (1998) European Commission Recommendation 98/282/EC. EC (1999) European Commission decision of 23 February 1999 adopting a register of flavouring substances used in or on foodstuffs (1999/217/EC). Official Journal of the European Communities, No. L 084/1 (27/3/99) amended by Decision 2000/489/EC, Official Journal of the European Communities,No.L 197/53 (3/8/00). EC (2001) European Commission Working Document WGF/007/01. Draft European Parliament and Coun- cil Regulation for smoke flavourings used or intended to be used in or on foodstuffs (22 February 2001). EC (2003) Regulation No 2065/2003 of the European Parliament and of the Council of 10 November 2003 on smoke flavourings used or intended for use in or on foods. EC (2008a) Regulation No 1331/2008 of the European Parliament and of the Council of 16 Decem- ber 2008 establishing a common authorisation procedure for food additives, food enzymes and food flavourings. EC (2008b) Regulation No 1332/2008 of the European Parliament and of the Council of 16 December 2008 on food enzymes and amending Council Directive 83/417/EEC, Council Regulation (EC) No 1493/1999, Directive 2000/13/EC, Council Directive 2001/112/EC and Regulation(EC) No 258/97. EC (2008c) Regulation No 1333/2008 of the European Parliament and of the Council of 16 December 2008 on food additives. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

Flavour legislation 49

EC (2008d) Regulation No 1334/2008 of the European Parliament and of the Council of 16 December 2008 on flavourings and certain food ingredients with flavouring properties for use in and on foods and amending Council Regulation (EEC) No 1601/91. Regulations (EC) No 2232/96 and (EC) No 110/2008 and Directive 2000/13/EC Essenzen VO (1959) Verordnung uber Essenzen und Grundstoffe (Essenzen VO) vom 19 Dezember 1959 (BGB1.IS.747). Essenzen VO (1970) Verordnung uber Essenzen und Grundstoffe (Essenzen VO). Der Neufassung vom 9 Oktober 1970 (BGBL I S. 1389). Hall, R.L. and Oser, B.L. (1965) Recent progress in the consideration of flavoring ingredients under the Food Additive Amendment. FEMA GRAS Substances 3. Food Technol. 19(2), Part 2, 151; FEMA GRAS Substances 4. Food Technol. 24(5) 25. Hall, R.L. and Oser, B.L. (1970) Recent progress in the consideration of flavoring ingredients under the Food Additive Amendment. FEMA GRAS Substances 3. Food Technol. 19(2), Part 2, 151; FEMA GRAS Substances 4. Food Technol. 24(5) 25. Hallagan, J.B. and Hall, R.L. (1995) FEMA GRAS Ð a GRAS assessment program for flavor ingredients. Regul. Toxicol. Pharmacol. 21, 422Ð430. IOFI, International Organisation of the Flavour Industry (1989) Code of Practice Section III IOFI guidelines for the production and labelling of process flavourings (October 1989). MAFF (1965) Food Standards Committee Report on Flavouring Agents, HMSO, London. MAFF (1976) Food Additives and Contaminants Committee Report on the Review of Flavourings in Food, FAC/REP/22, HMSO, London. Munro, I.C., Kennepohl, E. and Kroes, R. (1999) A procedure for the safety evaluation of flavouring substances. Food Chem. Toxicol. 37, 207Ð232. Newberne, P., Smith, R.L., Doull, J., Goodman, J.I., Munro, I.C., Portoghese, P.S., Wagner, B.M., Weil, C.S., Woods, L.A., Adams, T.B. and Hallagan, J.B. (1998) GRAS flavoring substances 18. Food Technol. 52(9), 65. Newberne, P., Smith, R.L., Doull, J., Goodman, J.I., Munro, I.C., Portoghese, P.S., Wagner, B.M., Weil, C.S., Woods, L.A., Adams, T.B., Hallagan, J.B. and Ford, R.A. (1999). Correction to GRAS flavoring substances 18. Food Technol. 53(3), 104. Newberne, P., Smith, R.L., Doull, J., Feron, V.J., Goodman, J.I., Munro, I.C., Portoghese, P.S., Waddell, W.J., Wagner, B.M., Weil, C.S., Adams, T.B. and Hallagan, J.B. (2000) GRAS flavoring substances 19. Food Technol. 54(6), 66, 68Ð70, 72Ð74, 76Ð84. Oser, B.L., Ford, R.A. and Bernard, B.K. (1984) Recent progress in the consideration of flavoring ingredients under the Food Additive Amendment. FEMA GRAS Substances 13. Food Technol. 38(10), 65. Oser, B.L., Ford, R.A. and Bernard, B.K. (1985) FEMA GRAS Substances 14. Food Technol. 39(11), 108. Oser, B.L. and Hall, R.L. (1972) Recent progress in the consideration of flavoring ingredients under the Food Additive Amendment. FEMA GRAS Substances 5. Food Technol. 26(5), 35. Oser, B.L. and Hall, R.L. (1973) FEMA GRAS Substances 6. Food Technol. 27(1), 64. Oser, B.L. and Hall, R.L. (1974) FEMA GRAS Substances 7. Food Technol. 27(11), 56. Oser, B.L. and Hall, R.L. (1975) FEMA GRAS Substances 8. Food Technol. 28(9), 76. Oser, B.L. and Hall, R.L. (1976) FEMA GRAS Substances 9. Food Technol. 29(9), 70. Oser, B.L. and Hall, R.L. (1977) FEMA GRAS Substances 10. Food Technol. 31(1), 65. Oser, B.L. and Hall, R.L. (1978) FEMA GRAS Substances 11. Food Technol. 32(2), 60. Oser, B.L. and Hall, R.L. (1979) FEMA GRAS Substances 12. Food Technol. 33(7), 65. Smith, R.L. and Ford, R.A. (1993) Recent progress in the consideration of flavoring ingredients under the Food Additive Amendment. FEMA GRAS Substances 16. Food Technol. 47(6), 104. Smith, R.L., Newberne, P., Adams, T.B., Ford, R.A., Hallagan, J.B. and the FEMA Expert Panel (1996) GRAS flavoring substances 17. Food Technol. 50(10), 72. Smith, R.L., Newberne, P., Adams, T.B., Ford, R.A., Hallagan, J.B. and the FEMA Expert Panel (1997) Correction to GRAS flavoring substances 17. Food Technol. 51(2), 32. Smith, R.L., Doull, J., Feron, V.J., Goodman, J.I., Munro, I.C., Newberne, P.M., Portoghese, P.S., Waddell, W.J., Wagner, B.M., Adams, T.B. and McGowen, M.M. (2001) GRAS flavoring substances 20. Food Technol. 55(12), 34Ð36, 38, 40, 42, 44Ð55. Smith, R.L, Cohen, S.M., Doull, J., Feron, V.J., Goodman, J.I., Marnett, I.J., Portoghese, P.S., Waddell, W.J., Wagner, B.M. and Adams, T.B. (2003) GRAS flavoring substances 21. Food Technol. 57(5), 46Ð48, 50, 52Ð54, 56Ð59. P1: SFK/UKS P2: SFK c02 BLBK221-Taylor/Linforth November 17, 2009 16:31 Printer Name: Yet to Come

50 Food Flavour Technology

Smith, R.L., Cohen, S.M., Doull, J., Feron, V.J., Goodman, J.I., Marnett, I.J., Portoghese, P.S., Waddell, W.J., Wagner, B.M. and Adams, T.B. (2005) GRAS flavoring substances 22. Food Technol. 59(8), 24Ð28, 31Ð32, 34, 36Ð62. Smith, R.L., Waddell, W.J., Cohen, S.M., Feron, V.J., Marnett, L.J., Portoghese, P.S., Rietjens, I.M.C.M., Adams, T.B., and Taylor, S. (2009) GRAS flavoring substances 24. Food Technol. 63(6), 46Ð105. Waddell, W.J., Cohen, S.M., Feron, V.J., Goodman, J.I., Marnett, L.J., Portoghese, P.S., Rietjens, I.M.C.M., Smith, R.L., Adams, T.B., Lucas Gavin, C., McGowen, M.M. and Williams, M.C. (2007) GRAS flavoring substances 23. Food Technol. 61(8), 22Ð24, 26Ð28, 30Ð49. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

3 Basic chemistry and process conditions for reaction ﬂavours with particular focus on Maillard-type reactions Josef Kerler, Chris Winkel, Tomas Davidek and Imre Blank

3.1 INTRODUCTION

Maillard reaction technology is used by the flavour and food industry for the production of process/reaction flavours or generating flavour upon food processing (in-process flavour generation). Process flavours are complex building blocks that provide similar aroma and taste properties to those found in thermally treated foodstuffs such as meat, chocolate, coffee, caramel, popcorn and bread. The Maillard reaction between a reducing sugar and a food-grade nitrogen source is the principal underlying reaction, which is responsible for flavour and colour development. However, Maillard-type reactions may also give rise to undesirable molecules that need to be limited using mitigation concepts. This review provides a summary of general aspects of the Maillard reaction in flavour formation in view of reinforcing distinct desirable flavour notes. An overview of important aroma compounds of thermally treated foodstuffs and process flavours is given. In addition, the patent literature and other publications relating to reaction flavour production and their process conditions are discussed. This chapter focuses on Maillard-type reactions and only partially deals with other reactions occurring during process flavour preparation (e.g. lipid oxidation).

3.2 GENERAL ASPECTS OF THE MAILLARD REACTION CASCADE

The ‘Maillard reaction’ is of great importance for flavour and colour formation of thermally treated foodstuffs. It is a complex cascade of many different types of reactions rather than one single reaction type, even though it is initiated by an amino-carbonyl reaction step. The thermal generation of flavours in foods, process flavours and model systems has accordingly been the subject of many symposia and reviews (e.g. Parliment et al., 1989, 1994; Weenen et al., 1997; Reineccius, 1998; Tressl and Rewicki, 1999; Cerny, 2007; Yeretzian et al., 2007). A first milestone in the history of Maillard chemistry was the publication of the well- known Hodge scheme (Hodge, 1953). Although Hodge’s studies focused only on Maillard browning, this scheme provided a framework that also covered important reaction routes for the formation of aroma compounds. The work of Hodge triggered a large number of studies on the elucidation of important intermediates and pathways of the Maillard reaction (see reviews by Ledl and Schleicher, 1990; Tressl and Rewicki, 1999). Isotopic labelling of sugars and/or amino acids in conjunction with GC-MS (gas chromatographyÐmass spectrometry) analysis (Tressl et al., 1993; Gi and Baltes, 1995; Keyhani and Yaylayan, 1996) and trapping of P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

52 Food Flavour Technology

reactive intermediates (Nedvidek et al., 1992; Hofmann, 1999) are key techniques for the improved understanding of Maillard reaction pathways. Improvements and refinements of the Hodge scheme were presented by Tressl et al. (1995). Their scheme provides an excellent overview of Maillard reaction pathways leading to the formation of volatile compounds. For the optimisation of reaction flavours, however, a strong emphasis is required on those routes that are involved in the generation of key aroma compounds. This can be achieved by first evaluating the character-impact compounds of a process flavour or model system using a combination of sensorial and instrumental analysis on the basis of the odour activity value concept (Grosch, 1994). The mechanistic studies can then be focused on the key substances only. The work of Hofmann (1995) is an excellent example of such an approach. Figure 3.1 gives an overview of the pathways that are involved in the formation of important aroma compounds during Maillard reaction. There are three main routes involved in flavour generation. All three routes start with imine formation between a reducing sugar and an amino acid. The Amadori (derived from aldoses) or Heyns (derived from ketoses) rearrangement products are important intermediates of the early phase of the Maillard reaction. Route A (see also Section 3.2.1) leads to the formation of 1- and 3-deoxyosones, which on cyclisation, reduction, dehydration and/or reaction with hydrogen sulfide result in heterocyclic aroma compounds. Route B (see also Section 3.2.2) is characterised by fragmentation of the sugar chain through retro-aldolisation or ␣-/␤-cleavage. By aldol-condensation of two sugar fragments or a sugar fragment and an amino acid fragment, heterocyclic aroma compounds are generated on cyclisation, dehydration and/or oxidation reactions. Alterna- tively, the fragments can react with hydrogen sulfide and form very potent alicyclic flavour substances. Route C (see also Section 3.3) involves the so-called Strecker degradation of amino acids, which is catalysed by dicarbonyl or hydroxycarbonyl compounds. The reaction

Amino acid + Sugar

Amadori/Heyns Fragmentation products B rearrangement compounds

Retro-aldol or H23 S, NH , amino - amino acid α (β)-cleavage acid fragments

1- and 3-Deoxyosones Condensation products - H S A 2 H S, NH reduction Amino acid 23

Cyclisation products Cyclisation products Strecker degradation

- H2 S reduction C dehydration

Flavour substances

Fig. 3.1 Major pathways for the formation of ﬂavour substances during Maillard reaction. A, B and C denote the three key pathways. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 53

Fig. 3.2 Reaction of glucose (I) and glycine leading to the Amadori compound N-(1-deoxy-D-fructos-1- yl)glycine (V, open-chain form) and related degradation reactions (adapted from Davidek et al., 2002).

is a ‘decarboxylating transamination’ and the resulting Strecker aldehydes are potent ﬂavour compounds. Strecker aldehydes can also be formed directly from Amadori rearrangement products (ARPs) or Heyns rearrangement products (HRPs). A more detailed, but still simpliﬁed scheme is depicted in Fig. 3.2 (Davidek et al., 2002) showing the formation of Amadori compounds, N-substituted 1-amino-1-deoxy-ketoses (V) representing an important class of Maillard intermediates (Ledl and Schleicher, 1990). They are formed in the initial phase of the Maillard reaction by Amadori rearrangement of the corresponding N-glycosylamines (II), the latter obtained by condensation of amino acids and aldoses such as glucose (I) as shown in pathway A. The importance of Amadori compounds stems from the fact that their formation as well as decomposition can be initiated under mild conditions. Thus, the formation of Amadori compounds represents a low-energy pathway of sugar degradation. The chemistry of Amadori compounds has recently been reviewed (Yay- layan and Huyghues-Despointes, 1994). Degradation of the Amadori compound V by 1,2- enolisation (pathway B) and 2,3-enolisation (pathway D) leads to the formation of 3-deoxy- 2-hexosulose (VII) and 1-deoxy-2,3-hexodiulose (XII), respectively, as already suggested by P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

54 Food Flavour Technology

Hodge (1953). In parallel to these pathways, other ␣-dicarbonyls can be formed by enolisation. For example, transition metal-catalysed oxidation of 1,2-enaminol IV can lead via pathway C to osones such as glucosone (IX). Pathway E gives rise to the 1-amino-1,4-dideoxy-2,3- diulose (XIII) by elimination of the C4-OH group of the 2,3-endiol (VI). If the iminoketone (VIII) formed by oxidation of 1,2-enaminol (IV) is not hydrolysed, then Strecker aldehydes can be formed in the course of pathway C by direct oxidative degradation of the Amadori compound and decarboxylation of X, as proposed by Hofmann and Schieberle (2000a). The so-called carbon module labelling (CAMOLA) technique (Schieberle et al., 2003; Schieberle, 2005) has been introduced as an advanced tool to quantify the relative contribution of carbohydrate fragments (e.g. C1ÐC4 fragments derived from Route B in Fig. 3.1) in comparison with transient intermediates with intact carbon chain conﬁguration (e.g. deoxyosones; formed via Route A in Fig. 3.1) in the generation of Maillard-derived aroma compounds. The authors used a model system containing 1:1 mixtures of unlabelled and 13 C6-labelled glucose as well as unlabelled proline and then measured the labelling pattern 12 13 13 ( C6, C3 and C6) of the caramel-like odourant 4-hydroxy-2,5-dimethyl-3(2H)-furanone (FuraneolR ) by GC-MS. The study revealed that, under dry heating conditions, Furaneol was generated entirely via the intact sugar skeleton, whereas in aqueous solution, 63% of the furanone was derived from the recombination of two C3-fragments. This work can be considered as another milestone in the history of research on Maillard chemistry. The CAMOLA technique has recently also been applied for studying the formation of furan in both model systems and foodstuffs (Limacher et al., 2008).

3.2.1 Intermediates as flavour precursors ARPs and HRPs are relatively stable intermediates and have been detected in various heat- processed foods (Eichner et al., 1994). Since ARPs and HRPs can easily be synthesised (Van den Ouweland and Peer, 1970; Yaylayan and Sporns, 1987), their potential as flavour precursors has been evaluated in several studies. Doornbos et al. (1981), for example, found that the ARP derived from rhamnose and proline is a useful precursor for the generation of the potent caramel-like odourant 4-hydroxy-2,5-dimethyl-3(2H)-furanone. ARPs and HRPs have also been reported to be good precursors for Strecker aldehydes and, in the absence of oxygen, also for 1- and 3-deoxyosones (Hofmann and Schieberle, 2000a). At higher pH values, ARPs and HRPs easily undergo cleavage of the carbohydrate chain, yielding fission products such as 2,3-butanedione and pyruvaldehyde (Weenen and Apeldoorn, 1996). When cysteine is heated with reducing sugars, thiazolidine carboxylic acids (TCAs) are formed instead of ARPs or HRPs (de Roos, 1992). TCAs are relatively stable in anionic form, which is probably the main reason for the inhibitory effect of cysteine in Maillard reactions, especially at higher pH. This can also explain the more efficient formation of important meat sulfur compounds at low pH (Hofmann and Schieberle, 1998a). A research disclosure (Anonymous, 1979), however, describes the use of TCAs as precursors for meat flavours. TCAs of glyceraldehyde, fructose or xylose were reacted as such or in conjunction with organic acids (e.g. succinic acid, malic acid and citric acid) or fatty acids (e.g. oleic acid and linoleic acid). Using a pH of 6Ð7 and temperatures between 50 and 100◦C, the TCA of glyceraldehyde and cysteine was reported to result in a beef-like aroma, whereas TCAs of fructose or xylose and cysteine yielded ‘meaty/savoury’ flavours. Other important Maillard reaction intermediates are the deoxyosones. In general, 1-deoxyosones are more important flavour precursors than 3-deoxyosones, whose formation is favoured under neutral/slightly basic and acidic pH, respectively. Although 1-deoxyglucosone has been synthesised by Ishizu et al. (1967), 1-deoxyosones are too P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 55

unstable to be used as precursors. 3-Deoxyosones, however, are more stable and are easily obtainable from compounds such as difructoseglycine (Anet, 1960). The structure and reactivity of various 3-deoxyosones have been extensively studied by Weenen and Tjan (1992, 1994) and Weenen et al. (1998). By using various 1H NMR and 13C NMR techniques, the authors showed that 3-deoxypentosone and 3-deoxyglucosone consist almost exclusively of monocyclic and bicyclic (hemi)acetal/(hemi)ketal structures. 3-Deoxypentosones and 3- deoxyhexosones are good precursors for furfural and (5-hydroxymethyl)furfural under acidic conditions. Under basic conditions, they undergo cleavage of the carbohydrate chain and can form pyrazines in the presence of an N-source. Hofmann and Schieberle (2000b) showed that acetylformoin, which is formed from 1- deoxyhexosone, is an effective precursor for 4-hydroxy-2,5-dimethyl-3(2H)-furanone (Fu- raneol). The amounts of Furaneol obtained from acetylformoin were significantly enhanced in the presence of reductones such as ascorbic acid or methylene reductinic acid as well as the Strecker-active amino acid proline. The reaction between acetylformoin and proline also resulted in high amounts of the cracker-like odourant 6-acetyltetrahydropyridine (ACTP). A number of articles and patents report the production of meat-like aromas by reacting 4-hydroxy-5-methyl-3(2H)-furanone (norfuraneol) with hydrogen sulfide or cysteine. Van den Ouweland and Peer (1968) were first to file a patent on the use of this precursor system to prepare 3-mercaptomethylfurans, which exhibit meat-like character. Later, several studies identified sulfur-containing aroma compounds derived from the reaction of norfuraneol and hydrogen sulfide (e.g. Van den Ouweland and Peer, 1975; Whitfield and Mottram, 1999). The latter authors showed that this precursor system is capable of producing compounds such as 2-methyl-3-furanthiol (MFT) and 2/3-mercapto-3/2-pentanone in relatively high amounts. These thiols are also key aroma compounds of heated meat (Kerscher and Grosch, 1998). Shu and Ho (1989) and Zheng et al. (1997) studied the reaction of the methyl homologue 4- hydroxy-2,5-dimethyl-3(2H)-furanone (Furaneol) with hydrogen sulfide or cysteine, which also gave rise to meat-like aromas. Among the identified sulfur-containing volatiles, however, the methyl homologue of MFT (2,5-dimethyl-3-furanthiol) was not found in the reaction mixtures. Unilever patented processes for the preparation of savoury flavours using the precursor systems and reaction conditions shown in Fig. 3.3 (Turksma, 1993; Rosing and Turksma, 1997). When 2,5-dimethyl-2-(2-hydroxy-3-oxo-2-butyl)-3(2H)-furanone (diacetyloligomer; R and R = CH3; process A in Fig. 3.3), which can be obtained by heating 2,3-butanedione under acidic conditions (Doornbos et al., 1991), is reacted with cysteine and hydrogen sulfide, high amounts of 2,5-dimethyl-3-furanthiol (DMFT) are generated (Turksma, 1993). For example, almost 40% yield of DMFT was obtained after 1 hour at 120◦C using a polar organic solvent, acidic conditions and super-atmospheric pressure (100Ð2500 kPa). DMFT, which was found to have a ‘meaty taste and roasted meat aroma’ was also formed from 2,5-dimethyl-3(2H)-furanone. Using similar conditions as for the diacetyloligomer, Rosing and Turksma (1997) reacted 4-hydroxy-2,5-dimethyl-(2-hydroxy-3-oxo-2-butyl)-3(2H)- 1 3 4 furanone (Fig. 3.3; R − R = CH3,R = acetyl; process B) with cysteine and hydrogen sulfide. This process resulted in a flavouring with sweet, onion-like, meaty aroma, odours attributed to high amounts of 2,5-dimethyl-4-mercapto-3(2H)-furanone and 2,5-dimethyl- 4-mercapto-3(2H)-thiophenone. However, these two sulfur compounds were not found to significantly contribute to the flavour of heated meat. Mottram et al. (1998) filed a patent on flavouring agents that serve as precursors for generating cooked (e.g. cooked meat) flavours in foodstuffs in situ. The authors claim P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

56 Food Flavour Technology

rosrucerP niaM tnarodo

O SH R' Cysteine + H2S A O R R O Glycerol O OH 90–120°C 1–3 hours, 100–2500 kPa R, R’ = CH3 or H

OH O OH SH Cysteine + H S R2 2 B OH R1 R2 R1 O Glycerol X R3 R4 90–120°C 1-3 hours, 100–2500 kPa R1 = CH3, C2H5 or H R2 = alkyl, C1-4 or H R3 = alkyl, C1-5 R4 = organic radical, C1-6, H or O-atoms X = O or S

Fig. 3.3 Precursors, reaction conditions and main aroma compounds of two savoury process ﬂavours (according to Turksma, 1993 (A), and Rosing and Turksma, 1997 (B)).

a long list of precursor substances that are capable of developing flavour during microwave cooking or conventional oven cooking with reduced cooking times. This list comprises several sulfur compounds such as hydrogen/ammonium/sodium sulfides, cysteine, thiamine, onion and garlic as well as ‘non-sulfur-containing post-rearrangement Mail- lard products such as furanones (e.g. 4-hydroxy-5-methyl-3(2H)-furanone, 4-hydroxy-2,5- dimethyl-3(2H)-furanone, 2-methyl-4,5-dihydro-3(2H)-furanone), pyranones (e.g. maltol, 5-hydroxy-5,6-dihydromaltol), 3-deoxyglucosone, ketones (e.g. cyclotene) and aldehydes. The precursor mixtures were encapsulated or spray-dried and applied to the foodstuffs through dusting or inclusion prior to the heat treatment. In two other studies, a similar precursor system consisting of unsaturated aldehydes and hydrogen sulfide resulted in aroma blocks with deep-fried notes (Van den Ouweland, 1989; Zhang and Ho, 1989). As shown in this chapter, meat-like flavours can be obtained from the reaction of intermediate precursor systems such as 4-hydroxy-5-methyl-3(2H)-furanone (norfuraneol) and cysteine. Cerny and Davidek (2003) investigated the efficiency of this precursor system in the generation of the meat-like odourants 2-methyl-3-furanthiol (MFT) and 3-mercapto- 13 2-pentanone (MP) relative to their formation from ribose/cysteine. Reacting C5-labelled ribose together with unlabelled norfuraneol and cysteine under aqueous conditions (pH 5, ◦ 13 95 C for 4 hours), the authors showed that mainly C5-labelled MFT was formed, suggesting that norfuraneol is less important as intermediate of MFT. In contrast, MP was found unlabelled and hence originated from norfuraneol. On the basis of these results, as well as 13 additional mechanistic studies using the CAMOLA technique (heating of ribose and [ C5]- ribose (1 + 1) with cysteine under above-mentioned conditions), a new reaction pathway for the formation of MFT and MP from ribose via 1,4-dideoxyosone was proposed (Fig. 3.4). P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 57

Ribose

O OH O OH

OH OH OH O O O O 1,4-Dideoxyosone

+ H2S + H2S

O SH O OH OH OH OH SH O O

- H2O - H2O + H2S

O SH OH SH OH OH SH O O - H2O SH - H2O

+ Cys - H2O O

2-Methyl-3-furanthiol O SH

HO N O

OH SH SH 3-Mercapto-2-pentanone - CO2 - H2O

Red. SH O N NH2 NH + H2O + H2O

- NH3 SH SH SH SH SH - O

Fig. 3.4 Proposed formation pathway for 3-mercapto-2-pentanone and 2-methyl-3-furanthiol from ribose and cysteine via the 1,4-dideoxyosone route (adapted from Cerny and Davidek, 2003). P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

58 Food Flavour Technology

3.2.2 Carbohydrate fragmentation Carbohydrate fragments have been found to originate from deoxyosones, ARPs or HRPs, as well as from the sugar directly (Ledl and Schleicher, 1990; Weenen, 1998) (Fig. 3.1). The extent of the sugar cleavage reactions depends on the pH and on the reaction medium, with fragmentation favoured at higher pH values (pH ≥7) and in aqueous systems. Using isotopic labelling of the sugar molecule in conjunction with GC-MS analysis, C5/C1,C4/C2 and C3/C3 fission reactions were established (review by Tressl et al., 1995). The proposed cleavage routes involve retro-aldolisation, vinylogous retro-aldolisation, ␣- and ␤-dicarbonyl cleavage (reviewed by Weenen, 1998; Tressl and Rewicki, 1999). Retro-aldolisation is by far the most accepted fragmentation route. Studying the generation of acetic acid (which is a major carbohydrate fragmentation product and also a good marker for the 2,3-enolisation pathway, as exclusively formed from 1-deoxy-2,3-diuloses) during Maillard reaction of glucose and glycine under aqueous conditions (90Ð120◦C, pH 6Ð8), Davidek et al. (2006a) revealed that the organic acid is mainly formed through a hydrolytic ␤-dicarbonyl cleavage pathway (Fig. 3.5). The authors also evidenced that ␤-dicarbonyl cleavage, which can be seen as an acyloin cleavage or a reverse Claisen-type reaction, represents a general cleavage route for diacylcarbinol intermediates of the Maillard reaction under aqueous conditions and that the frequently reported hydrolytic ␣-dicarbonyl cleavage can be ruled out as a sugar fragmentation mechanism. Furthermore, a new sugar fragmentation pathway has been suggested to occur under oxidative conditions (Davidek et al., 2006b) by oxidative ␣-dicarbonyl cleavage with a BaeyerÐVilliger-type rearrangement as key steps. The sugar degradation pathway will mainly depend on the reaction conditions (Fig. 3.6). Carbohydrate cleavage products have been analysed by several authors (Nedvidek et al., 1992; Weenen and Apeldoorn, 1996; Hofmann, 1999). Since these intermediates are reactive dicarbonyl and hydroxycarbonyl compounds, trapping agents such as 1,2-diaminobenzene and ethoxamine hydrochloride were used to transform them into stable quinoxaline and O-ethyloxime derivatives, respectively. Weenen and Apeldoorn (1996) studied the formation of glyoxal, methylglyoxal, 2,3-butanedione and 2,3-pentanedione in both Maillard and caramelisation reactions. The results are shown in Table 3.1. The study revealed that sugar fragmentation is highest in the presence of a Strecker-inactive amine functionality (cyclohexylamine), followed by a Strecker-active amino acid (alanine). Without amine (caramelisation reaction), the extent of fragmentation was even lower and no detectable amounts of 2,3-butanedione and 2,3-pentanedione were observed. The yields of the pentanedione were relatively high in alanine model systems, indicating that the Strecker aldehyde, acetaldehyde, is involved in its formation. In addition, the ARP of glucose and alanine was found to be an efficient ␣-dicarbonyl precursor, whereas the 3-deoxyglucosone/alanine reaction mixture yielded only low concentrations of fission products (Table 3.1). The latter result is in good agreement with the finding that 3-deoxyglucosone is also a poor pyrazine precursor (Weenen and Tjan, 1994). Hofmann (1999) studied the time course of the formation of carbohydrate degradation products in thermally treated solutions of either xylose or glucose with alanine. The author showed that during the first 10 minutes of the Maillard reaction, glyoxal is the most abundant fragmentation product from both xylose and glucose. Its formation can be explained by retro- aldol cleavage of 2-xylosulose or 2-glucosulose. After 20 minutes of the Maillard reaction, the main fission products were found to be methyl glyoxal and hydroxy-2-propanone, with xylose yielding higher amounts than glucose. One year later, Yaylayan and Keyhani (2000) P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 59 O OH 3 OH CH CHO C C + 4 1 C C COOH H + 4 3 H H -3-oxobutanal -3-oxobutanal HOH OC OH O O 3 3 CH 1 OH C C C CH CH 1 6 H Acetylformoin OH OH 2 OH OH O 3 COOH -dicarbonyl cleavage mechanism, indicating 3 + ␤ CHO CH C C C CH 4 4 1 CH 6 H + + H H OH O O O OH O 3 3 3 OH C C C C CH CH CHO C C CH COOH C C + 3 1 6 3 3 6 H H H HOH CH Acetylformoin 6 OH OH BC 2-Hydroxy-3-oxobutanal 2-Hydroxy 2-Hydroxy-3-oxobutanal O 2 as the key intermediate by the hydrolytic H OH COOH OH OH O 2 3 3 , 2006a). + C CH CHO C C CH CH 3 1 H

et al. + H OH O OH OH O 2 3 OH OH OH 2 2 Davidek C C C CH CH 1 CHO CH CHO CH C C 6 6 3 H HC 6 Erythrose H HOH OH OH O O 3 2 1-Deoxy-2,4-hexodiulose 1-Deoxy-2,4-hexodiulose CH C C CH 3 6 A OH 1-Deoxy-2,3-tetrodiulose OH OH 2 2 C C CH C=O C CH 6 3 6 H HOH Tetrulose HOH + + H O OH OH OH OH 3 O 2 3 CH C C CHO C C CO C CH CH 3 6 1 6 COOH COOH H H 3 H 3 OH OH OH 2 + CH CH C CH C C OH 1 3 2-Hydroxy-3-oxobutanal 2-Hydroxy-3-oxobutanal 1-Deoxy-2,3-hexodiulose 1-Deoxy-2,3-hexodiulose H HOH Formation of acetic acid from glucose via 1-deoxy-2,3-hexodiulose Fig. 3.5 the various possible degradation pathways (adapted from P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

60 Food Flavour Technology

CH3 C O COOH CO A1 H C OH H C OH + HOHC

H C OH CH2OH CH CH2OH 3 Erythronic acid 1-Deoxy-2,3-hexodiulose COOH Acetic acid CH2OH C O + HOHC B1 CH3 CH2OH C O Tetrulose H C OH C O HCOH CH3 CH OH C O 2 B2 + CH OH 1-Deoxy-2,4-hexodiulose 2 COOH Acetol HOHC

CH2OH CH 3 Glyceric acid H C OH CH OC A2 3 H C OH C O + COOH HCOH Lactic acid CH2OH 1-Deoxy-3,4-hexodiulose

Fig. 3.6 Scheme summarising possible pathways of sugar fragmentation via 1-deoxyhexodiuloses. A1 and A2, oxidative ␣-dicarbonyl cleavage; B1 and B2, hydrolytic ␤-dicarbonyl cleavage by nucleophilic − attack of OH at the C-2 and C-4 carbonyl group, respectively (adapted from Davidek et al., 2006b).

performed a mechanistic study to determine the origin of Maillard intermediates such as glycolaldehyde, methylglyoxal, hydroxy-2-propanone and 3-hydroxy-2-butanone. 2,3-Butanedione (diacetyl) and 2,3-pentanedione are aroma-active carbohydrate cleavage products, contributing to a sweet-caramel odour of coffee (Grosch, 2001), and, for example, in the presence of hydrogen sulﬁde or cysteine, they are precursors of important sulfur aroma compounds such as 2-mercapto-3-butanone and 2/3-mercapto-3/2-pentanone (Hofmann, 1995). Yaylayan and Keyhani (1999) investigated the origin of these two dicarbonyl compounds in glucose/alanine Maillard model systems under pyrolytic conditions. Using labelled glucose or alanine, the authors showed that 90% of the formed pentanedione requires the participation of the C2/C3 atoms of alanine, whereas diacetyl was derived from the sugar chain only. This result is supported by the ﬁnding of Hofmann (1995) that P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 61

Table 3.1 Formation of ␣-dicarbonyl products (according to Weenen and Apeldoorn, 1996).

α-Dicarbonyl products (µg) Methyl 2,3-Buta- 2,3-Penta- Amine Carbohydrate Glyoxal glyoxal nedione nedione No Glucose 26 11 – – No Fructose 28 15 – – No Xylose 62 17 – – No 3-Deoxyglucosone 23 57 – – No Fru-Alaa 103 101 98 18 Alanine Glucose 58 43 41 38 Alanine Fructose 45 28 22 25 Alanine Xylose 27 81 28 42 Alanine 3-Deoxyglucosone 16 56 11 21 Alanine Fru-Alaa 81 67 81 22 Cyclohexylamine Glucose 618 865 227 39 Cyclohexylamine Fructose 691 1104 265 89 Cyclohexylamine Xylose 591 925 614 101 Cyclohexylamine 3-Deoxyglucosone 317 583 146 25 Cyclohexylamine Fru-Alaa 509 454 232 39

1 N-1-(deoxy-d-fructosyl)-l-alanine (Amadori rearrangement product of glucose and alanine).

2,3-pentanedione is formed by reacting acetaldehyde and hydroxy-2-propanone. Schieberle 13 et al. (2003) applied the CAMOLA technique to a Maillard model system of glucose ( C6 + 12 C6 = 1 + 1) and proline and showed that 2,3-butanedione was formed by 87 and 13% through the recombination of C3 + C1 and C2 + C2 fragments (Route B in Fig. 3.1), respectively. As a result, no diacetyl is generated from the intact carbon chain (Route A in Fig. 3.1).

3.2.3 Strecker degradation In the presence of ␣- or vinylogous dicarbonyl compounds, ␣-amino acids can undergo a ‘decarboxylating transamination’, which results in the formation of aldehydes with one carbon atom less than the amino acid (Strecker aldehydes) (Schonberg¬ and Moubacher, 1952). The Strecker degradation of amino acids is a key reaction in the generation of potent aroma compounds during Maillard-type processes (Ledl and Schleicher, 1990) (see also Fig. 3.1). Certain amino acids (leucine, valine, methionine or phenylalanine) are known to produce Strecker aldehydes with significant odour strength such as 3-methylbutanal, methylpropanal, methional or phenylacetaldehyde. These aldehydes have been confirmed as key contributors to many thermally processed foods (Hofmann et al., 2000). Besides aldehyde formation, Strecker degradation also contributes to flavour formation during Maillard reaction by reducing dicarbonyls to hydroxycarbonyls (e.g. formation of 1,4-dideoxyosone from 1- deoxyosone) (Nedvidek et al., 1992) or by generating ␣-aminocarbonyl compounds, which are pyrazine precursors (Weenen and Tjan, 1994). Weenen and van der Ven (1999) studied the formation of phenylacetaldehyde in Maillard model systems, including reactions of phenylalanine with various sugars, ␣-dicarbonyl and hydroxycarbonyl compounds as well as ARPs. The authors found that methyl glyoxal was the most efficient dicarbonyl compound for the formation of phenylacetaldehyde, followed by 3-deoxyerythrosone, glyoxal, 3-deoxyxylosone and 3-deoxyglucosone. Hydroxycarbonyl compounds such as dihydroxyacetone and glyceraldehyde also yielded high amounts of the P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

62 Food Flavour Technology

HO HO OH COOH 2+ HO COOH O2,Me OH H+ OH OH COOH OH HO N N H OH HO H O + N + Me 2+ HO H Me ,H2O2 O O O O

OH OH HO OH HO OH H2O N N O O O OH CO2,H2O COOH OH H HO OH

NH O 2

Fig. 3.7 Formation of phenylacetaldehyde via an oxidative degradation of N-(1-deoxy-D-fructosyl)-L- phenylalanine (Fru-Phe) (adapted from Hofmann and Schieberle, 2000a).

Strecker aldehyde. Sugars were less reactive, with the reactivity decreasing in the order erythrose, xylose, fructose and glucose. In addition, the authors showed that the Amadori compound Fru-Phe is a superior phenylacetaldehyde precursor to the corresponding sugar/amino acid mixture. This finding was also confirmed by Hofmann and Schieberle (2000a). In addition, Hofmann and Schieberle (2000a) revealed that the yields of phenylacetaldehyde formed from the Amadori compound Fru-Phe were significantly increased in the presence of oxygen and copper (II) ions. On the basis of the observation that 1,2-hexodiulose is also generated in high amounts under these conditions, they proposed a mechanism for the formation of phenylacetaldehyde from Fru-Phe (Fig. 3.7). Hofmann et al. (2000) additionally found that in the reaction of phenylalanine and glucose, considerable amounts of phenylacetic acid were generated. While the formation of phenylacetaldehyde showed an optimum pH of 5 and was not influenced by oxygen, the acid was most abundant at pH 9 in the presence of oxygen and copper (II) ions. The authors proposed a mechanism for the generation of phenylacetic acid that involves a similar oxidation step to that shown in Fig. 3.7. Cremer and Eichner (2000) studied the influence of the pH on the formation of 3-methylbutanal during Maillard reaction of glucose and leucine. They showed that the formation rate of the aldehyde was higher at pH 7 than at pH 5 or 3. This was consistent with the high degradation rate of the Amadori compound Fru-Leu at pH 7, suggesting that the ARP was a good precursor for 3-methylbutanal. However, Chan and Reineccius (1994) showed that the optimal reaction conditions of different Strecker aldehydes vary, owing to differences in stability of the aldehydes.

3.2.4 Interactions with lipids In addition to the Maillard reaction, lipid oxidation is another major reaction occurring in process flavour production and food systems. Both reaction cascades include a whole network of different reactions in which extraordinary complex mixtures of compounds are obtained, triggering important changes in food flavour, colour, texture and nutritional value, with desirable and undesirable consequences. In addition, both reactions are intimately interrelated as shown in Fig. 3.8 (Zamora and Hidalgo, 2005) and the products of each reaction influence the other. Furthermore, there are common intermediates and products in both pathways. The existing data suggest that both Maillard reaction and lipid peroxidation are so closely interrelated that they should be considered simultaneously to understand the P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 63

Lipid Reducing Amino N-Substituted sugar compound glycosylamine Oxidation Amadori rearrangement Oxidation Lipid hydroperoxides

1-Amino-1deoxy-2-ketose Hydroperoxide decomposition

Sugar dehydration Lipid peroxyl and hydroxyl radicals Sugar fragmentation Sugar enolisation Strecker degradation Formation of stable compounds Cleavage Polymerisation Glyoxal Methylglyoxal Volatile and non-volatile monomers Aldehydes Others Ketones Ketones • Schiff base of HMF or furfural Alcohols Cyclic peroxides Alcohols • Reductones Epoxides Hydroperoxy Epoxides • Fission products Dimers compounds Hydrocarbons Hydroxyalkylpyrroles • Aldehydes Polymers Acids Cleavage • Others

Pyrrole Carbonyl-amine Aldol condensation polymerisation polymerisation Aldol condensation Carbonyl-amine polymerisation Carbonyl-amine polymerisation

Aroma Taste Antioxidants Coloured compounds Neo-formed toxicants

Fig. 3.8 Known interactions between Maillard reaction and lipid oxidation pathways in nonenzymatic browning development (adapted from Zamora and Hidalgo, 2005).

reaction mechanisms, kinetics, and products in the complex mixtures of carbohydrates, lipids and proteins occurring in food systems and process flavours. In these systems, lipids and carbohydrates are competing in the chemical modification of amino compounds (e.g. proteins and phospholipids). Therefore, although there are significant differences between Maillard reaction and lipid peroxidation, many aspects of both reactions can be better understood if they are included in only one general carbonyl pathway that can be initiated by both lipids and carbohydrates (Hidalgo and Zamora, 2005). As an example, 2-pentylpyridine has been reported (Henderson and Nawar, 1981) as an interaction product of linoleic acid and valine, with 2,4-decadienal and ammonia being the key intermediates (Fig. 3.9). Its formation was studied by Kim et al. (1996) in model systems

Lipid COOH O Oxidation Linoleic acid (E,E)-2,4-Decadienal

NH2 Maillard NH3 R COOH reaction

Amino acid Ammonia (asparagine, glutamine)

[O] NH N N H

Fig. 3.9 Formation of 2-pentylpyridine from 2,4-decadienal and an amino acid (adapted from Henderson and Nawar, 1981). P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

64 Food Flavour Technology

by reacting 2,4-decadienal with amino acids (glycine, aspartic acid, asparagine, glutamic acid and glutamine) at 180◦C for 1 hour (pH 7.5). The relative yields of alkylpyridine formation from the reactions were asparagine Ͼ glutamine Ͼ aspartic acid Ͼ glutamic acid Ͼ glycine. When amide-15N-labelled glutamine and asparagine were heated with 2,4- decadienal, the relative contribution of amide nitrogens to the formation of alkylpyridine was determined. Approximately half the nitrogen atoms in 2-pentylpyridine formed from asparagine, originated from the amide nitrogens of asparagine, whereas when glutamine was the reactant, almost all the nitrogen atoms came from the amide nitrogens in glutamine. The above-mentioned results may indicate that both free ammonia and ␣-amino groups bound in amino acids can contribute to the formation of alkylpyridines, but free ammonia does so more effectively. The formation of Strecker aldehydes in the presence of lipid oxidation products is another example, illustrating the interactions between Maillard and lipid intermediates. Strecker degradation of amino acids is one of the most important reactions leading to ﬁ- nal aroma compounds in the Maillard reaction. Hidalgo and Zamora (2004) have studied the reaction of 4,5-epoxy-2-alkenals with phenylalanine. In addition to N-substituted 2- (1-hydroxyalkyl)pyrroles and N-substituted pyrroles, which are major products of the reaction, the formation of both the Strecker aldehyde, phenylacetaldehyde, and 2-alkylpyridines was also observed. The aldehyde was only produced from the free amino acid (not esteri- ﬁed), suggested to be produced through imine formation, which is then decarboxylated and hydrolysed (Fig. 3.10). This reaction also produces a hydroxyl amino derivative, which is the origin of the 2-alkylpyridines. These data indicate that Strecker-type degradation of amino acids occurs at low temperature by some lipid oxidation products. This is a proof of the interrelations between lipid oxidation and Maillard reaction, which are able to produce common products by analogous mechanisms. However, recent results suggest that, analogously to carbohydrates, certain lipid oxidation products may also degrade certain amino acids to

H H N 2 − 1 O H2O R N R1 + O O O OH O O 56 H − CO2 O

1 H + H O R1 N R NH2 + 2 OH OH 13 14

R1 N 15

Fig. 3.10 Strecker-type degradation of phenylalanine produced by 4,5-epoxy-2-alkenals (adapted from Hidalgo and Zamora, 2004). P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 65

undesirable compounds, e.g. vinylogous derivatives such as styrene (Hidalgo and Zamora, 2007). Simulating food flavours by the process flavour approach requires precursors and recipes that are close to food or which well represent the composition of food. Therefore, lipids are key ingredients in the reaction flavour system to obtain boiled chicken notes. Similarly, polyphenols play an important role in generating cocoa flavour. The role of these specific components (lipids, polyphenols, vitamins) may be (i) generating new specific flavour compounds (Whitfield, 1992) as compared to the pure Maillard system (e.g. 2-pentylpyridine) or (ii) intervening in the Maillard reaction cascade and thus alerting the overall flavour composition. The latter is probably of higher relevance.

3.3 IMPORTANT AROMA COMPOUNDS DERIVED FROM MAILLARD REACTION IN FOOD AND PROCESS FLAVOURS

In the flavour industry, process flavours are often developed by empirical means, i.e. the reaction conditions are optimised using organoleptic evaluation. However, this approach should be accompanied by analytical evaluation of the products. Knowledge of the important aroma compounds derived from the Maillard reaction in food and process flavours is essential in order to focus investigations into reaction mechanisms enhancing the key aroma compounds. In addition, the knowledge of their formation pathways and key intermediates allows the development of multi-step approaches (e.g. two-step reactions), which aim at optimising reaction conditions of each of the flavour generation stages (e.g. from sugar/amino acid mixture to deoxyosones, from deoxyosones to target flavour compounds). One of these steps (often the first step) can also be a biotransformation that yields an intermediate that, on thermal treatment, releases the key aroma compound. An example for such an approach is the biogeneration of 2-(1-hydroxyethyl)-4,5-dihydrothiazole through yeast fermentation, which releases 2-acetylthiazoline on microwave heating (Bel Rhild et al., 2002). It is recommended that the evaluation of character-impact aroma compounds should be based on a combination of instrumental and sensorial analysis. A well-established approach involves the determination of the odour activity values (ratio of concentration and odour threshold values) of aroma compounds (Grosch, 1994). This requires quantitative analysis of a selected number of aroma compounds, the importance of which has been screened by GC- olfactometry (GC-O). Suitable screening techniques such as aroma extract dilution analysis or headspace dilution analysis are available (Ullrich and Grosch, 1987; Guth and Grosch, 1993a, b). As these methods do not consider interactions of different aroma compounds, they should be combined with organoleptic evaluations of reconstituted model mixtures. Our selection of important Maillard-derived aroma compounds (see Sections 3.3.1 and 3.3.2) is primarily based on results of GC-O techniques or other sensory evaluations or on quantitative data.

3.3.1 Character-impact compounds of thermally treated foods Character-impact compounds of thermally treated foods that are formed during Maillard reaction are summarised in Table 3.2. Their identiﬁcation in foodstuffs such as meat, bread, P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

66 Food Flavour Technology ) ), 27 ), ), coffee ( 7 ), beer ( 21 ), tea ( 29 ), yeast ( 40 ) b ) ) 17 ), chocolate ( 17 ), meat ( 41 ) 7 26 ), chocolate ( 17 ), cocoa ( 17 ) ( 2–6 ), yeast ( 20,), 40 coffee ( 7, 8 ) 3–6, 9 ), yeast40 ), ( coffee ( 7, 8 ), sesame 20 ) 20 ) 20 ) 15 ) ( ( ( ( 12 ) 3, 9, 19 ), chicken ( 5 ), coffee ( 7 ), French fries 3, 9, 19 ), chicken ( 4, 5 ), pork ( 6 3, 9 ), chicken ( 4, 5 ), sesame 10 ), ( ﬁsh ( 22 ) 12 ) 12 ) 3, 4, 13 ), chicken3, ( 14 ), yeast ( 40 ) a t t t t o s s a a c a a e e a o e e Beef ( Yeast ( Beef ( Yeast ( Coffee ( Beef ( ( 21 ), tea ( 29 ), chocolate ( 17 ), cocoa ( 17 ), yeast ( 40 Beef ( French fries ( 21 ), potato chipbread ( 18 ), ( 24 ), ﬁsh yeast ( 22, ( 40 ) 23 ), Beef ( Detected in Coffee ( popcorn ( 27 ), bread ( 24 ), chocolate( 28 ), ( 17 ), French sesame fries ( Beef ( 3, 9, ), 19 chicken ( 6 ), coffee ( 7 Chicken ( 4 ( 10 ), popcorn ( 11 ) Beef ( Beef ( b g/kg) )M ) ␮ )M 1 1 16 )C ) ) ) 26 ) 1 15 )Y 25 ) 1 0.004 ( — 1.15 ( 0.05 ( — — 0.05 ( 20 )Y 0.3 ( 5 ) 0.2 ( 18 ) water ( 7.5 ( 26 ) 0.01 ( 3.0 ( Odour threshold in 0.7 ( 1 reaction in various thermally treated foods. , roast beef Roasty, brothy Meaty, sweet, sulfury 0.007 ( Meaty, thiamine-like Meaty, onion-like, coffee-like — Cooked meat-like Cooked meat-like Caramel-like, sweet Caramel-like, fruity Seasoning-like Cooked potato-like Roasty, popcorn-like, burnt 1.0 ( Odour description Seasoning-like Roasty, sulfury Meaty, sweet, sulfury 0.018 ( Sulfury, catty Sulfury, catty )- )-

H H )- )-furanone Caramel-like, strawberry-like 10)-furanone ( 26

H H H Important aroma compounds derived from the Maillard furanone thiophenone (sotolon) furanone (abhexon) 10 2-Methyl-3-furylthioacetate 1112 1-(2-Methyl-3-furylthio)-ethanethiol 4-Hydroxy-2,5-dimethyl-3(2 Meaty 17 2-Ethyl-4-hydroxy-5-methyl-3(2 13 Methional 14 2-Acetyl-2-thiazoline No. Compound Table 3.2 1 2-Methyl-3-furanthiol 15 4-Hydroxy-2,5-dimethyl-3(2 9 2-Furfurylmethyl disulﬁde 18 3-Hydroxy-4-methyl-5-ethyl-2(5 16 3-Hydroxy-4,5-dimethyl-2(5 2 2-Furfurylthiol 45 3-Mercapto-2-pentanone 2,5-Dimethyl-3-furanthiol 3 3-Mercapto-2-butanone 6 2-Methyl-3-(methylthio)furan 8 2-Methyl-3-(methyltrithio)furan A: Compounds containing sulfur 7 2-Methyl-3-(methyldithio)furan B: Compounds containing oxygen P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction flavours 67 ), , 7 38 , 25 , Cerny and 19 , Kerler and 13 ,Guthand ), fish ( 22, 23 ), (1972); ), French fries ( 21 ), fish

et al. ), bread ( 36 ) 31 , Milo and Grosch (1995); , Gasser and Grosch (1991); 6 34 ), popcorn ( 11 ), sesame ( 10 ), 37 , Schieberle and Grosch (1994); 18 , Guadagni 24 , Rychlik and Grosch (1996); ) 28 ), popcorn ( 39 ), French fries ( 21 ), bread ( 24 ), tea ( 29 ), ), French fries ( 20 ), tea ( 29 ) ( 7 7 7 24 ), rice ( 24 ), popcorn ( 11 ) 24 ), French fries ( 21 ), cocoa ( 17 ), chocolate ), chicken ( 5 ), coffee ( 8 12 , Mottram and Madruga (1994); 3, 5 ), yeast ( 40 ), chocolate ( 17 ), cocoa ( 17 ), e ( 2, 3 ), French fries ( 21 ), yeast ( 40 ) 2, 19 ), cocoa ( 17 ), chocolate ( 17 ), bread ( 24 ), 22, 23 ), coffee ( 7 ), bread ( 24 ) m t ( 30 , Schieberle (1991b); a a h s e s e i Bread ( beef ( Bread ( coffee ( Beef ( 19, 25 ), chicken ( 14 ), coffee ( 7 ), sesame ( 10 bread ( 24 ), French fries ( 21 ),( 17 ), cocoa popcorn ( 17 ), ( 36 ) chocolate Beef ( 19 ), chicken ( 5 ), breadcoffee ( 24 ), ( chocolate ( 17 ), beer ( 30 ) Bread ( ( 17 ), popcorn ( 36 ) Beef ( 19, 25 ), chicken( 21 ), ( 14 ), cocoa coffee ( 17 ), ( 7 ), chocolate French ( 17 ), fries popcorn sesame ( 36 ), ( 10 ), bread ( 37 ), yeast40 ) ( Coffee ( 7 ), French fries ( 21 ) ( 31 ) Beef ( 3, 9 ), bread ( 24 ), coffee ( 7 Beef ( French fries ( 21 ), coffee ( 7 ), teaBeef ( 29 ), ( 9 yeast ( 40 ) Chocolate ( 17 ), cocoa ( 17 ), yeast ( 40 ) Coffee ( 5 , Kerler and Grosch (1997); 23 , Milo and Grosch (1996); 11 , Schieberle (1991a); 36 , Schieberle and Grosch (1987); ) 33 ) 23 ) 17 , Schnermann and Schieberle (1997); 32 ) 18 ) 29 , Guth and Grosch (1993b); 0.35 ( 5 )M 1.6 ( 35 ) 0.16 ( 7 ) 15 ( 0.4 ( 18 ) 0.09 ( 7 ) — 62 ( 38 )S 25 ( 9 30 ( 32 )F — (2000); , Gasser and Grosch (1990); 4

et al. 41 , Madruga and Mottram (1995). 10 , Schieberle (1993a); 22 , Milo and Grosch (1993); 35 , Buttery and Ling (1995); 28 , Schieberle (1996); 16 , Schieberle (1982);

et al. thy, roasty Malty, cocoa-like Roasty, popcorn, bread-like 0.1 ( Earthy, roasty Roasty, cracker-like Earthy, roasty Earthy, roasty Buttery Malty, fruity, pungent 0.7 ( Buttery, green Earthy, roasty Roasty, sweet, nutty Honey-like, sweet, ﬂowerySolvent-like 4 ( , Kerscher and Grosch (1997); , Guth and Grosch (1994); ¨ unch and Schieberle (1998); 3 9 was assessed by quantitative data or by GC-O techniques. 34 , Buttery 27 , Schieberle (1993b); 40 ,M 21 , Wagner and Grosch (1997); (1983); (1995);

15 , MacLeod and Ames (1986); et al. (1991);

et al.

et al. 33 , Buttery , Gasser and Grosch (1988); 39 , Schieberle (1995); 2 , Semmelroch and Grosch (1995); 14 , Kerler (1996); (1992); 8 26 , Semmelroch 20 , Werkhoff (1975);

et al.

et al. , Hofmann (1995); The sensory signiﬁcance of the1 aroma compounds a b Grosch (1993a); Teranishi Grosch (2001); 32 ,Blank Grosch (1996); Grosch (1993); 19 2/3-Methylbutanal 25 2-Acetyl-1-pyrroline 26 6-Acetyltetrahydropyridine 27 2-Ethyl-3,5-dimethylpyrazine 28 2-Ethyl-3,6-dimethylpyrazine 20 Methylpropanal 29 2,3-Diethyl-5-methylpyrazine 30 2-Ethenyl-3,5-dimethylpyrazine 32 2-Acetylpyrazine 31 2-Ethenyl-3-ethyl-5-methyl-pyrazine Ear 23 2,3-Butanedione 21 Phenylacetaldehyde 22 Acetaldehyde 24C: Compounds 2,3-Pentanedione containing nitrogen P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

68 Food Flavour Technology

coffee, cocoa, chocolate, sesame, popcorn, French fries, tea, fish and yeast, as well as odour description and odour thresholds of the aroma substances, is covered. In the following discussion of Table 3.2, emphasis is given to important aroma compounds of meat, yeast, coffee and bread. The meaty character of boiled beef, pork or chicken is mainly due to sulfur compounds such as MFT, 2-furfurylthiol, 3-mercapto-2-butanone, 2/3-mercapto-3/2-pentanone, DMFT, methanethiol, hydrogen sulfide and methional (Gasser and Grosch, 1988, 1990; Mottram and Madruga, 1994; Kerscher, 2000). The precursors of these aroma substances in meat are known to be free or bound C5-sugars such as ribose, ribose phosphate and inosine monophosphate as well as sulfur-containing compounds such as thiamine, cysteine, glutathione and methionine. After the publication of the patent of Morton et al. (1960), in which meat aroma formation established from the reaction of cysteine and ribose was described, a great number of patents and publication on meat-like process flavours followed (see Section 3.4.3). It is well known that species-specific differences in the aroma of cooked meats such as beef and chicken are mainly due to concentration and composition differences in lipid- derived flavour substances. Kerscher and Grosch (1998) and Kerscher (2000) confirmed these findings and showed that significant differences also exist for Maillard-derived aroma compounds. Cooked beef, for example, was found to contain higher amounts of the sulfur compounds MFT and 2-furfurylthiol as well as the caramel-like 4-hydroxy-2,5-dimethyl- 3(2H)-furanone, whereas MP and methional were more important in cooked chicken. The character-impact compounds of yeast extracts were found to be very similar to those of cooked meat, which is due to similar pools of Maillard precursors. Munch¬ and Schieberle (1998) reported high odour activity values for the sulfur compounds MFT, 2-furfurylthiol, MP and methional, the Strecker aldehydes 3-methylbutanal and phenylacetaldehyde, as well as the furanones 4-hydroxy-2,5-dimethyl-3(2H)-furanone (Furaneol) and 3-hydroxy-4,5- dimethyl-2(5H)-furanone (sotolon). Werkhoff et al. (1991) additionally identified 2-methyl- 3-(methylthio)furan, 2-methyl-3-furylthioacetate, 1-(2-methyl-3-furylthio)ethanethiol and 4- hydroxy-2,5-dimethyl-3(2H)-thiophenone (thiofuraneol) in yeast. The authors claimed that these sulfur compounds contribute significantly to the meaty character of yeast. Besides yeast, onion extracts are also used in process flavourings. Widder et al. (2000) have identified 3-mercapto-2-methylpentan-1-ol as new powerful aroma compounds in both process flavourings (containing onion extract) and raw onions. This odourant that exhibits a pleasant meat broth, sweetish, onion and leek-like character (at a low concentration of 0.5 ppb in water), is suggested to be formed by aldol condensation of two molecules of propanal, followed by hydrogen sulfide addition at the double bond to yield 3-mercapto-2-methylpentanal. The aldehydes are finally reduced enzymatically to the corresponding alcohol. The Maillard reaction is also a key reaction in flavour formation during roasting of coffee. The precursor pool in green coffee comprises a complex mixture of various soluble sugars such as glucose, fructose, galactose and sucrose. In addition, the amount of polymeric arabinose and rhamnose was found to decrease during roasting, which indicates that these sugars are also involved in caramelisation and Maillard processes (Tressl, 1989). The total amino acid content drops by about 30% during roasting. Especially, the amino acids Ð lysine, serine, threonine, arginine, histidine, methionine and cystine Ð are degraded to a high extent during the roasting process (Belitz et al., 2008). Semmelroch and Grosch (1995) reported the simulation of the aroma of Arabica and Robusta coffee brews using reconstituted mixtures of 23 aroma compounds. The authors showed that the Maillard products Ð 2- furfurylthiol, Furaneol, sotolon, methanethiol, 2,3-butanedione, 2,3-pentanedione, 2-ethyl- 3,5-dimethylpyrazine (EDMP), 2,3-diethyl-5-methylpyrazine (DEMP), methylpropanal and P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 69

OH O O OH + N N Alapyridaine O OH O H COO + O N 5-MPC COOH OH OH O N COOH N H O O OH OH 1-Oxo-2,3-dihydro-1H- OH Dru-Glu indolizinium-6-olate 5-MPF

Fig. 3.11 Chemical structures of some recently identiﬁed taste molecules formed by Maillard-type reactions. Fru-Glu, N-(1-deoxy-D-fructos-1-yl)-L-glutamic acid; 5-MPC, 5-methyl-2-(1-pyrrolidinyl)-2-cyclopenten-1- one; MPF, and 5-methyl-4-(1-pyrrolidinyl)-3(2H)-furanone.

3-methylbutanal Ð contribute to the flavour of coffee brews. They also investigated the aroma differences between Arabica and Robusta coffee. The more earthy/roasty and less caramel character of Robusta was found to be due to higher concentrations of the pyrazines EDPM and DEMP as well as to lower amounts of Furaneol and sotolon, respectively. The flavour of cereal products, especially of bread, has been studied extensively, and the results have been reviewed (Grosch and Schieberle, 1997). 2-Acetyl-1-pyrroline (ACP) and 6-acetyltetrahydropyridine (ACTP) are responsible for the pleasant roasty character of wheat bread crust and popcorn. However, these compounds are not important in wheat bread crumb and rye bread. Both ACP and ACTP are generated by a reaction of proline with reducing sugars or sugar breakdown products (Schieberle, 1990). When ornithine instead of proline was reacted, only ACP was formed. The fact that yeast contains relatively high amounts of ornithine explains why ACP concentrations in bread are strongly dependent on the amount of yeast used in the baking process (Grosch and Schieberle, 1997). Other important Maillard- derived aroma compounds of wheat and rye bread are Furaneol and the Strecker aldehydes methional, 3-methylbutanal and methylpropanal. They contribute to the caramel-like and malty aroma of bread. Another breakthrough has been the identification of new taste-active molecules and taste modifiers as result of Maillard-type reactions (Fig. 3.11), which contribute to the overall flavour perception (Hofmann, 2005). As an example, Alapyridaine has been identified in Maillard model systems containing hexose sugars and alanine as well as in beef broth as an essential compound enhancing the sweet taste and umami character in beef broth (Ottinger and Hofmann, 2003; Soldo et al., 2003). Glycoconjugates of glutamic acid, namely the N-glycoside dipotassium N-(d-glucos-1-yl)-l-glutamate and the corresponding Amadori compound N-(1-deoxy-d-fructos-1-yl)-l-glutamic acid (Fru-Glu), were found by systematic sensory studies to exhibit a pronounced umami-like taste, with recognition taste thresholds of 1Ð2 mmol/L, close to that of monosodium glutamate (MSG) (Beksan et al., 2003). Contrary to MSG, they do not show the sweetish and slightly soapy by-note, but evoke an intense umami, seasoning, and bouillon-like taste. Added to a bouillon base, which did not contain any taste enhancers, both glycoconjugates imparted a distinct umami character similar to the control sample containing the same amount of MSG on a molar basis (Schlichtherle-Cerny et al., 2002). The Amadori compound Fru-Glu has been reported in dried tomatoes as an example (Eichner et al., 1994). Furthermore, thermal treatment of aqueous solutions of xylose, P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

70 Food Flavour Technology

rhamnose and l-alanine led to a rapid development of a bitter taste of the reaction mixture (Frank et al., 2003). Liquid chromatography/mass spectrometry (LC/MS) and nuclear magnetic resonance (NMR) spectroscopy revealed 1-oxo-2,3-dihydro-1H-indolizinium-6- olates as the key compounds and most signiﬁcant contributors to the intense bitter taste of this process ﬂavour mixture. Thermally treated glucose/L-proline mixtures that contain ‘cooling’ compounds were recently reported. These Maillard systems generate 5- methyl-2-(1-pyrrolidinyl)-2-cyclopenten-1-one (5-MPC) and 5-methyl-4-(1-pyrrolidinyl)- 3(2H)-furanone (MPF) as key compounds contributing to the cooling sensation without imparting aroma notes (Hofmann et al., 2001; Ottinger et al., 2001a). They have also been found in dark malt (10Ð100 ␮g/kg) (Ottinger et al., 2001b).

3.3.2 Character-impact compounds of process flavours Meat-like process flavours are often prepared by reacting cysteine and/or thiamine with sugars, although pentoses such as xylose and ribose are preferably used. There is a range of additional precursor sources such as pectin hydrolysates (C-source), hydrolysed vegetable proteins and wheat protein hydrolysates (N-sources) as well as hydrogen sulfide, inorganic sulfides, onion, garlic and cabbage (S-sources). The products serve as building blocks in the creation of meat flavours (see also Section 3.4.3). They can also be described as middle notes that are combined with base notes (mainly taste compounds and taste enhancers) and top notes (mainly compounded flavourings) (savoury flavour pyramid according to Yeretzian et al., 2007). The important sulfur-containing compounds in process flavourings derived from either cysteine/ribose or thiamine reaction systems are quite similar (Table 3.3). The aroma of both cysteine- and thiamine-based process flavours is determined by MFT, 3-mercapto-2-butanone, 2/3-mercapto-3/2-pentanone, 2-methyl-3-thiophenethiol and 2-thenylthiol (Guntert¬ et al., 1992, 1996; Hofmann and Schieberle, 1995, 1997). The same compounds are also responsible for the meaty character of process flavours that are based on 5-inosine monophosphate (5-IMP) and cysteine (Zhang and Ho, 1991; Madruga and Mottram, 1998). Although 5-IMP is more abundant than ribose in raw meat, it is a much poorer precursor for these sulfur compounds than ribose, when reacted with cysteine (Mottram and Nobrega, 1998). Other character-impact compounds that are formed primarily in thiamine or thiamine/cysteine reaction systems are 2-methyl-4,5-dihydro-3-furanthiol, 1- (methylthio)ethanethiol, mercaptoacetaldehyde and 2-methyl-1,3-dithiolane (Guntert¬ et al., 1996). Many of the thiols mentioned above are also important aroma substances in glucose/cysteine or rhamnose/cysteine process flavourings (Hofmann and Schieberle, 1997). However, both reaction systems contain other characteristic aroma compounds. For example, 2-(1-mercaptoethyl)furan and its thiophene derivative are only formed from glucose and cysteine, whereas 4-hydroxy-2,5-dimethyl-3(2H)-furanone (Furaneol) and 3-hydroxy-6-methyl- 2(2H)-pyranone belong to key odourants of the rhamnose system. The latter two substances are responsible for the strong caramel and seasoning-like character of rhamnose/cysteine process blocks, which render them very suitable for application in beef flavours. Another compound having seasoning-like character is 3-hydroxy-4,5-dimethyl-2(5H)-furanone (sotolon). Sotolon was found to contribute to the aroma of various cysteine-derived reaction flavours (Hofmann and Schieberle, 1995, 1997). In contrast to the other compounds mentioned above, the formation of sotolon is less influenced by the type of sugar. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 71 ) ), ), ), ), ), ), ), ), 3 3 3 ) ) 6 6 ), )

Continued 6 ( ), glutathione/ribose ( ), glutathione/ribose ( ), thiamine/cysteine ( 5 ), ), glutathione/ribose ( ) ), thiamine/methionine ( 5 ) ) ) 5 ) 5 ), cysteine/IMP ( 6, 7 ), ) ), cysteine/ribose-5-P ( 7 b ), cysteine/IMP ( 6, 7 ), ), thiamine ( 5 ), cysteine/IMP6 ), ( Cysteine/IMP ( Thiamine/cysteine ( Detected in Cysteine/ribose ( 2, 3, 6 ), cysteine/glucose4 ( cysteine/rhamnose ( 4 thiamine ( 3, 5 ), thiamine/cysteinecysteine/IMP ( 5 ), ( 6, 7 ), cysteine/ribose-5-P ( Cysteine/ribose ( 2, 3, 6 ), cysteine/glucose4 ( cysteine/rhamnose ( 4 thiamine ( 3 cysteine/ribose-5-P ( 6 cysteine/ribose-5-P ( 6 Thiamine ( 3, 5 ), thiamine/cysteine ( 5 ) Cysteine/ribose ( 2, 3, ), 6 ( 3 glutathione/ribose Cysteine/ribose ( 2, 6 ), cysteine/glucosecysteine/rhamnose ( 4 ( 4 cysteine/IMP ( 6, 7 ), cysteine/ribose-5-P ( Cysteine/ribose ( 2, 3, ), 6 cysteine/rhamnose cysteine/glucose ( 4 ( 4 thiamine ( 3, 5 ), thiamine/cysteinecysteine/IMP ( 5 ), ( 6 Methionine/ribose ( 9 ), thiamine/methionine ( 5 ) Cysteine/ribose ( 2, 6 ), cysteine/IMP ( 6 cysteine/ribose-5-P ( 6 Cysteine/ribose ( 2, 6 ), cysteine/glucose ( 4 thiamine/cysteine ( cysteine/ribose-5-P ( 6 b ) ) ) ) ) 1 1 1 g/kg) in water ␮ 0.00002 ( 10 ) 0.007 ( 1 ) — 3.0 ( — 0.02 ( Odour threshold ( 0.042 ( 1 — 0.7 ( 0.2 ( 8 ) a Meaty, sulfury Sulfury, roasty, coffee-like 0.01 ( 1 Odour description Meaty, sulfury, sweet Sulfury, catty Sulfury, putrid Cabbage-like Meaty, sulfury Meaty, sulfury Sulfury, catty Cooked potato-like Sulfury, roasty Character-impact compounds of process ﬂavourings. No. Compound Table 3.3 1 2-Methyl-3-furanthiol 2 2-Furfurylthiol 4 Mercapto-2-propanone 3 Mercaptoacetaldehyde 910 2-Methyl-4,5-dihydro-3-furanthiol Bis(2-methyl-3-furyl) disulﬁde 5 3-Mercapto-2-butanone 6 3/2-Mercapto-2/3-pentanone A: Compounds containing sulfur 78 Methional 2-Methyl-3-thiophenethiol 11 2-Thenylthiol P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

72 Food Flavour Technology ), ), ), ), ), 4 4 4 ) ) ) ) ) ) ) ) 4 4 6 5 ), thiamine/methionine ( 5 ) 5 ) ), cysteine/glucose ( 4 2 b Detected in Cysteine/rhamnose ( Cysteine/glucose ( 4) Cysteine/rhamnose ( Thiamine/cysteine ( cysteine/rhamnose ( 4 Cysteine/ribose ( 2 ), cysteine/glucose ( cysteine/rhamnose ( 4 Cysteine/glucose ( 4 ) Cysteine/ribose ( 2 ), cysteine/glucose ( 4 Cysteine/glucose ( 4 ) Cysteine/ribose ( 2, 3 ), thiaminethiamine/cysteine ( 5 ), ( 5 ), cysteine/IMP ( 6,cysteine/ribose-5-P 7 ), ( Thiamine/cysteine ( Cysteine/ribose ( 2 ), cysteine/glucose ( cysteine/rhamnose ( 4 Cysteine/ribose ( 2 ), cysteine/glucose ( cysteine/rhamnose ( 4 Cysteine/ribose ( cysteine/rhamnose ( 4 b ) ) ) ) 1 1 12 ) 11 ) g/kg) in water ␮ 0.048 ( 1 ) 0.049 ( 1 ( Odour threshold 0.022 ( 0.038 ( 1 1.25 ( 1 ) — 10 ( 0.2 ( 1.0 ( 13 ) — — — Odour description Sulfury, roasty Caramel-like, sweet, meaty 24.0 ( Sulfury, roasty Sulfury, burnt Roasty, popcorn-like Sulfury Sulfury, burnt Sulfury, egg-like Roasty, popcorn-like Sulfury, putrid Sulfury, burnt Thiamine-like, meaty Sulfury, putrid )-

H )

Continued ( thiophenone No. Compound Table 3.3 12 5-Methyl-2-furfurylthiol 13 5-Methyl-2-thenylthiol 14 2-(1-Mercaptoethyl)furane 23 5-Acetyl-2,3-dihydro-1,4-thiazine 24 4-Hydroxy-2,5-methyl-3(2 15 2-(1-Mercaptoethyl)thiophene 18 2-Methyl-1,3-dithiolane 16 2-Methyltetrahydrothiophen-3-one 20 Methanethiol 21 Ethanethiol 22 2-Acetyl-2-thiazoline 19 Hydrogen sulﬁde 17 1-(Methylthio)ethanethiol P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 73 20 , ) ), ), , Zhang 4 18 ) 7 13 , Cerny and ) ) ) ), proline/glucose ( 15 ), 4 4 ), cysteine/rhamnose ( 4 4 19 , Kerler and Grosch (1997); 18 ), proline/lactose ( , Mottram and Nobrega (1998); 6 12 , Guth and Grosch (1994); (1996); Serine/maltose ( Cysteine/rhamnose ( Cysteine/rhamnose ( Cysteine/ribose ( 2 ), cysteine/glucose ( Cysteine/ribose ( 2 ), cysteine/glucose4 ( cysteine/rhamnose ( 4 Cysteine/ribose ( 2 ), proline/xylose ( 16 ) Cysteine/glucose ( cysteine/rhamnose ( 4 proline/rhamnose ( 16 ), lysine/rhamnose ( 16 ) Proline/glucose ( 15 ) Proline/glucose ( 15 ) Proline/glucose ( 15 ) 18 , Fickert (1999);

et al. (1970); ¨ untert

et al. 5 ,G ) ) 1 14 ) 17 , Pittet 21 ) 22 ) 14 ) 23 ) 11 , Pippen and Mecchi (1969); 35000 ( 17 ) 15.0 ( 1.15 ( 19 ( 0.3 ( 19 ) 0.09 ( 20 ) 0.1 ( (1990); (1984);

et al. (1975).

et al.

et al. , Hofmann and Schieberle (1997); 16 , Decnop 10 , Buttery 4 23 , Teranishi Caramel-like, burnt chicory 8500 ( 1 Caramel-like Caramel-like, sweet Earthy, roasty Caramel, strawberry-like 10 ( Seasoning-like Roasty, popcorn-like Roasty, caramel-like Roasty, burnt, caramel-like 1.6 ( , Gasser (1990); 3 was assessed by quantitative data or by GC-O techniques. 15 , Roberts and Acree (1994); )- , Meynier and Mottram (1995);

H 9 )-furanone )-furanone 22 , Buttery and Ling (1995);

H H (1995); )-furanone )-pyranone )-pyranone Seasoning-like

H H H (1972); (1983); et al.

et al.

et al. , Hofmann and Schieberle (1995); 2 , Guadagni 8 21 , Buttery 14 , Semmelroch (norfuraneol) (maltol) furanone (homofuraneol) (Furaneol) (sotolon) , Hofmann (1995); The sensory signiﬁcance of the1 aroma compounds a b and Ho (1991); Grosch (2001); Grosch (1993); 25 4-Hydroxy-2,5-dimethyl-3(2 31 2,3-Diethyl-5-methylpyrazine 30 3-Hydroxy-6-methyl-2(2 29 3-Hydroxy-4,5-dimethyl-2(5 27 2-Ethyl-4-hydroxy-5-methyl-3(2 28 3-Hydroxy-2-methyl-4(4 26 4-Hydroxy-5-methyl-3(2 32 2-Acetyl-1-pyrroline B: Compounds containing oxygen C: Compounds containing nitrogen 3334 6-Acetyltetrahydropyridine 2-Acetylpyridine P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

74 Food Flavour Technology

Furaneol, which is also a key ingredient of caramel-like process flavours, can be efficiently prepared by reacting rhamnose or other 6-deoxyhexoses with lysine, proline or hydroxyproline (Decnop et al., 1990). Another important caramel-like aroma compound, 3-hydroxy-2-methyl-4(4H)-pyranone (maltol), is a key substance of process flavours that are prepared from disaccharides (maltose or lactose) and proline or serine (Fickert, 1999). In thermally treated solutions of proline and glucose or fructose, ACP and ACTP have been evaluated as important aroma compounds, contributing a roasty, popcorn-like odour to these process flavours (Roberts and Acree, 1994; Schieberle, 1995). ACP and ACTP have also been reported to be character-impact compounds of various thermally treated cereal products (see also Section 3.3.1). In addition, Roberts and Acree (1994) showed that Furaneol and 2-acetylpyridine contribute to the flavour of the proline/glucose model system.

3.4 PREPARATION OF PROCESS FLAVOURS

3.4.1 General aspects In Europe, flavourings that are obtained by thermal treatment of a reducing sugar and a food- grade nitrogen source such as amino acids, peptides, food proteins, hydrolysed vegetable proteins (HVPs) and yeasts are referred to as thermal process flavourings (official terminology revised by EU in 2008; see Chapter 2 for more details). These products have been designated as a separate class of flavours and are identified as complex mixtures that have been converted to flavours by heat processing. In the US, the term ‘process flavours’ does not exist in regulatory terms. Maillard reaction flavours are considered natural or artificial flavours, depending on whether the starting materials and process are considered natural. The International Organisation of the Flavour Industry (IOFI) has established a guideline for manufacturers of process flavours. This guideline is part of the IOFI’s Code of Practice and defines the types of raw materials and general reaction conditions (for instance, a maximum temperature/time treatment of 180◦C/15 minutes, pH ≤8) (reviewed by Manley, 1995). The US Department of Agriculture, however, did not set a guideline for manufacture but established labelling criteria for materials used to produce a process flavour (Lin, 1995). Although processed flavours that are prepared according to the IOFI guidelines were considered GRAS (generally recognised as safe) in the US in 1995, the regulatory status of Maillard reaction flavours still lacks clarity with respect to the GRAS specification. This is due to lack of information on whether processed flavours contain heterocyclic amines in amounts sufficient to affect their safe use in foods. Maillard reaction technology is commonly used by the flavour industry to produce complex building blocks that provide similar aroma and taste properties to thermally treated foodstuffs such as meat, chocolate, coffee, caramel, popcorn or bread. Although flavour formation during the Maillard reaction is quantitatively a minor pathway, process flavours are very important to the flavour industry. This is because these complex blocks exhibit unique flavour qualities, are difficult to copy and are relatively cheap, being based on low production costs and high flavour potency of the aroma compounds formed.

3.4.2 Factors influencing flavour formation The factors that influence flavour formation and, thus, the sensory properties of process flavours, are the type of sugar and amino acid, pH, reaction media, water activity as well as P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 75

Table 3.4 Flavour types of processed sugar–amino acid model mixtures.

Amino Temperature Flavour Sugar acid (◦C) description Referencesa Glucose Cysteine 100–140 Meaty, beefy 1, 6 Ribose Cysteine 100 Meaty, roast beef 1, 2, 6 Ascorbic acid Threonine 140 Beef extract, meaty 1 Ascorbic acid Cysteine 140 Chicken 1 Glucose Serine or glutamine or 100–220 Chocolate 1 tyrosine Glucose Leucine 100 Chocolate 3 Glucose Threonine 100 Chocolate 3 Glucose Phenylalanine 100–140 Floral, chocolate 1 Ribose or Threonine 140 Almond, marzipan 1 xylose Glucose Proline 100–140 Nutty 1 Glucose Proline or 180 Bread, baked 3, 6 hydroxyproline Glucose Alanine 100–220 Caramel 1, 6 Glucose Lysine 110–120 Caramel 4 Xylose Lysine 100 Caramel, buttery 5 Ribose Lysine 140 Toast 1 Glucose Valine 100 Rye bread 3 Glucose Arginine 100 Popcorn 3 Glucose Methionine 100–140 Cooked potatoes 1, 3 Glucose Isoleucine 100 Celery 1 Glucose Glutamine or —b Nutty 6 asparagine

a 1, Lane and Nursten (1983); 2, Morton et al. (1960); 3, Herz and Schallenberger (1960); 4, McKenna (1988); 5, Apriyantono and Ames (1990); 6,Yaylayanet al. (1994). b Microwave heating (640 W for 2–4 minutes).

temperature and time (see reviews by Shibamoto, 1983; Reineccius, 1990). In general, the sensory quality of a process flavour is less influenced by the type of sugar than by the amino acid. Several authors (Herz and Schallenberger, 1960; Lane and Nursten, 1983; Yaylayan et al., 1994) studied the variety of odours produced in Maillard model systems, comprising two or more components in reaction systems containing various sugars (or ascorbic acid) with each of the protein-derived amino acids. The flavour characters of some of these processed sugarÐamino acid model mixtures are summarised in Table 3.4. Cysteine is the favoured amino acid to produce meat-like flavours, both on heating with reducing sugars or alone (Lane and Nursten, 1983). These authors also obtained chicken and beef aromas by reacting, respectively, cysteine and threonine with ascorbic acid. Chocolate flavours can be prepared by heating glucose with amino acids such as serine, glutamine, tyrosine, leucine, threonine or phenylalanine (Herz and Schallenberger, 1960; Lane and Nursten, 1983). Phenylalanine also gives rise to a floral aroma (in reaction with glucose or alone), whereas threonine yields nutty aromas when reacted with ribose or xylose. Proline is the favoured amino acid for the production of bread-like and baked flavours (Lane and Nursten, 1983; Yaylayan et al., 1994). However, Schieberle (1992a) showed that the yeast- derived amino acids ornithine and citrulline are even more effective precursors for ACP, which is a key aroma compound of bread crust. Herz and Schallenberger (1960) reported the generation of rye bread and popcorn aromas by heating valine or arginine with glucose. In P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

76 Food Flavour Technology

addition, alanine and lysine were found to give caramel flavours, and glutamine and arginine, nutty flavours. Besides the type of sugar and amino acid, pH is another important factor determining aroma of process flavours. It is well known to the flavour industry that meat flavours are preferably prepared at low pH (4Ð5.5), whereas roast and caramel flavours are obtained under neutral or slightly basic conditions. Madruga and Mottram (1995) as well as Hofmann and Schieberle (1998a) showed that important sulfur-containing compounds in meat, such as MFT, 2-furfurylthiol and 2-methyl-3-(methyldithio)furan, are preferably formed at a pH of 3Ð4. Sensorial evaluations of thermally treated model mixtures of ribose or 5-IMP and cysteine revealed that the highest scores for boiled meat character were obtained when the reactions were carried out at pH 4.5 (ribose) and 3 (5-IMP) (Madruga and Mottram, 1998). Reaction media and water activity of the Maillard reaction systems are additional factors that influence aroma generation. Besides buffered aqueous solutions, solvents such as propylene glycol, glycerol, triacetin or fats and oils, as well as their emulsions or mixtures with water, are used. Vauthey et al. (1998), for example, filed a patent on the generation of roast chicken aroma using a cubic phase system. This system was prepared by introducing a melted monoglyceride (saturated in C16 and C18) into an aqueous phosphate buffer solution. Compared to the same reaction in phosphate buffer, the flavour formed in the cubic system was more intense, corresponding to higher amounts of sulfur compounds such as MFT. Shu and Ho (1989) investigated the reaction of cysteine and 4-hydroxy-2,5-dimethyl- 3(2H)-furanone in varying proportions of water and glycerol. They found that a superior roasted/meaty character was obtained in the aqueous system. The influence of the water activity on pyrazine formation during Maillard reaction was studied by Leathy and Reineccius (1989a). The authors observed that pyrazine formation was optimal at an Aw of about 0.75. Schieberle and Hofmann (1998) compared the character-impact compounds of cysteine- based process flavours formed in aqueous solution and under dry heating conditions. In a cysteine/ribose model system, dry heating yielded higher amounts of key odourants with roasty notes such as 2-furfurylthiol, 2-acetyl-2-thiazoline and 2-propionyl-2-thiazoline as well as 2-ethyl- and 2-ethenyl-3,5-dimethylpyrazine, whereas the meat-like sulfur compounds MFT and MP were found in comparable or lower concentrations, respectively. Their study also revealed that the amounts of the 3-deoxyosone-derived compounds 2-furfural and 5-methylfuran-2-aldehyde were significantly higher in the dry-heated model systems, whereas the formation of the 1-deoxyosone-derived 4-hydroxy-2,5-dimethyl-3(2H)-furanone (Furaneol) was enhanced in aqueous solution. An explanation for this finding could be that, under dry heating conditions, caramelisation processes are favoured relative to Maillard reaction. In caramelisation processes, 1,2-enolisation of the sugar molecule is preferred over 2,3-enolisation, leading to the formation of high amounts of 3-deoxyosones (Kroh, 1994). Apart from the precursor composition (e.g. availability of amino acids as reactants), pH and the reaction medium, other reaction parameters such as the presence of catalysts (e.g. phosphates) as well as temperature and time have a major effect on the sensory properties of process flavourings. Many of these parameters have extensively been studied for the generation of the caramel-like smelling Furaneol from 6-deoxyhexoses (Schieberle, 1992b; Havela-Toledo et al., 1997, 1999; Hofmann and Schieberle, 1997, 1998b; Schieberle and Hofmann, 2002). For example, the yield of Furaneol from rhamnose (pH 7, 150◦C, 45 minutes) was increased about 40 times when the malonate buffer was replaced with phosphate buffer (Schieberle, 1992b). Even a higher increase of its yield (70-fold) was observed when the pH of rhamnose/cysteine system (145◦C, 20 minutes) was increased from 3 to 7 P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 77

(Schieberle and Hofmann, 2002). Recently, Illmann et al. (2009) used a fractional factorial design to identify critical reaction parameters affecting kinetics of Furaneol generation from rhamnose under cooking conditions (120◦C). The importance of the reaction parameters was found to decrease in the following order: phosphate concentration > concentration of precursors Ͼ pH Ͼ rhamnose to lysine ratio. The experimental design approach achieved very high yields of Furaneol (about 40 mol%). The type of amino acid was also shown to affect the yield of Furaneol from rhamnose. Lysine was most efficient in generating the caramel-like odourant, followed by alanine, serine, glycine and threonine. On the other hand, much lower yields were obtained in the presence of proline and especially cysteine (Davidek et al., 2009). The generation of 3-hydroxy-2-methyl-4(4H)-pyranone (maltol), another important caramel-like aroma compound, was shown to strongly depend on the reaction media. Rather low yields of maltol were obtained when lactose was heated with proline in aqueous systems (13.5 mmol/mol lactose) or when the precursors were dry heated (7.2 mmol/mol lactose). However, the yield of maltol increased significantly when water was replaced by propylene glycol (70 mmol/mol lactose; Cerny, 2003). In addition, the knowledge of the reaction kinetics of aroma compounds helps to explain the influence of temperature and time on their formation (Reineccius, 1990). Although flavour formation is a multi-step reaction sequence, Arrhenius kinetics have been found to describe flavour formation well in model and real food systems. Stahl and Parliment (1994) used an ingenious device to obtain clear time/temperature conditions for model systems and determined the activation energies of flavour compounds. Leathy and Reineccius (1989b) showed that the sensory quality of a product is less influenced by temperature/time in a model system that is designed to result in similar key aroma compounds (e.g. pyrazines). This can be explained by the fact that these compounds have similar activation energies. However, their study focused on dialkylpyrazines and did not consider the more potent trialkylpyrazines, which were suggested to have different formation pathways (Amrani-Hemaimi et al., 1995; Schieberle and Hofmann, 1998). As a result, flavour generation during Maillard reaction is in most cases strongly influenced by temperature and time, also because Maillard reaction flavours are complex mixtures of different classes of key aroma compounds. Lee (1995) proposed a different approach, on the basis of differential equations, to describe Maillard kinetics. The model produced was able to simulate the Maillard reaction not just as a function of time and temperature but also as a function of reactant concentrations and pH. From this simulation, the relative amounts of reactants could be plotted against time. Examples given include the concentrations of aldose and Amadori compounds over a 10-hour reaction and the time profiles of enolisation compounds from both the 1,2- and 2,3-pathways under defined pH conditions. The use of stable intermediates as flavour precursors and/or the application of multi- step reactions are another way of optimising processing conditions on the basis of the required application. Blank et al. (2003a,b) compared the generation of odourants in Maillard reaction systems containing glucose and proline (Glc/Pro) or the corresponding Amadori compound fructosyl-proline (Fru-Pro). The major odourants found in both systems were similar and included ACP,4-hydroxy-2,5-dimethyl-3(2H)furanone (Furaneol), acetic acid, 3- hydroxy-4,5-dimethyl-2(5H)furanone (sotolon), 2,3-butanedione and ACTP. However, their concentrations as well as their contributions to the final flavour differed. For example, the formation of Furaneol was favoured from Fru-Pro, namely at pH 6 and 7. On the other hand, the reaction system Glc/Pro gave rise to relative high yields of ACP and ACTP. Although both roasted-smelling popcorn odourants showed high sensory relevance in both reaction systems, P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

78 Food Flavour Technology

they dominated especially the aroma of the Glc/Pro process flavouring. These results also indicate that there is no benefit in using Amadori compound for the formation ACTP and ACP. The following sections give a survey of the patent literature in the area of savoury and sweet process flavours. Emphasis is given to meat-like process flavours.

3.4.3 Savoury process flavours The great majority of patents based on Maillard reaction technology have been directed to the production of meat-like process flavours. Most of these reaction flavours indicate cysteine and thiamine as the essential sulfur-containing precursor compounds. In 1960, the basic concept of Maillard flavour technology was beginning to emerge with the Unilever patent of Morton et al. (1960). The authors disclosed Maillard processes for the production of cooked beef and pork flavours by reacting ribose or mixtures of ribose and glucose with cysteine and either additional amino acids or deflavoured protein hydrolysates from cod fish flesh, casein, groundnut or soya. The group of additional amino acids was consisted of ␤-alanine, glutamic acid, glycine, ␣-alanine, threonine, histidine, lysine, leucine, serine and valine. All flavours were prepared in water at a pH between 3 and 6 and a temperature around 130◦C. Using similar reaction conditions, May and Morton (1960) and May (1961) also prepared meat flavours through reacting cysteine with glyceraldehyde or furfural in combination with deflavoured cod fish hydrolysates or amino acid mixtures. Jaeggi (1973) patented processes for boiled and roasted beef flavours, which were also based on the reaction of ribose and cysteine. However, the inventor used methionine and proline as additional amino acids and carried out the reaction in glycerol or groundnut oil instead of water. Methionine as the sole sulfur source was also reported to result in beef flavourings when reacted with xylose and cysteine-free hydrolysed plant protein (Van Pottelsberghe de la Potterie, 1973). Tandy (1985) prepared process flavourings with white chicken meat character using leucine and cysteine in combination with the reducing sugars arabinose and glucose. In addition, he found that the use of rhamnose instead of glucose (as well as the addition of serine) provided a more aromatic, characteristic white meat chicken flavour. International Flavors and Fragrances (IFF) filed a number of patents on the preparation of chicken, beef and pork flavours using cysteine in combination with thiamine, often in carbohydrate-free systems. This demonstrates that thiamine is capable of replacing carbohydrates by providing similar intermediates and aroma compounds as found in Maillard systems. Intense beef flavours, for example, were obtained by refluxing cysteine, thiamine and carbohydrate-free vegetable protein hydrolysate (HVP) in water or waterÐethanol mixtures (IFF, 1965). The addition of beef tallow was found to result in a beef flavour with a ‘pan-dripping’ character. In addition, chicken flavourings were prepared by heating cysteine, thiamine and HVP in combination with other ingredients such as ␤-alanine, glycine and ascorbic acid, whereas pork flavours required the addition of methionine and lard. Similar processes for chicken, beef and pork flavours were disclosed in other IFF patents (IFF, 1967; Giacino, 1968a, 1969; Katz and Evers, 1973). Chicken aroma was found to be improved, for example, by adding diacetyl and hexanal (Giacino, 1968a) or mercaptoalkanones (Katz and Evers, 1973) to the processed flavours. Dihydroxyacetone, pyruvic acid or pyruvic aldehyde in combination with thiamine and HVPs were claimed to result in beef flavourings with improved cooked note (Giacino, 1968b). Kerscher (2000) investigated the analytical assessment of the species-specific character of beef, chicken and pork. His results (see Section 3.3.1) cannot explain the choice of the P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 79

ingredients for the preparation of chicken, beef and pork flavours that are disclosed in the IFF patents mentioned above, but are in agreement with the findings of Chen and Tandy (1988). The authors developed species-specific beef and chicken flavourings through oxidation of oleic and linoleic acids, respectively. Their processes involved heat treatment of oleic or linoleic acid in the presence of air at high temperatures of about 300◦C as well as trapping of the resulting aroma fraction in cold traps. The authors also claimed flavour blocks that resembled roast, grilled, bloody and braised beef in different fractions of the distillate of oxidised oleic acid. In terms of alternative sulfur sources to cysteine and thiamine, Giacino (1970) found that process flavours with similar characters were obtained when cysteine was replaced by taurine. Patents of the Corn Products Company also describe the production of Maillard reaction flavours using taurine in combination with HVPs and xylose (Corn Products Company, 1969; Hack and Konigsdorf, 1969). From a scientific point of view, however, the finding that cysteine can be replaced by taurine has to be questioned, because there are no studies that report the generation of important meat aroma compounds from taurine. In addition, Tai and Ho (1997) could detect only trace amounts of volatile sulfur compounds in a Maillard model system containing cysteinesulfinic acid and glucose. Broderick and Linteris (1960) used derivatives of mercaptoacetaldehyde such as 2,5-dihydroxy-1,4-dithiane, diethyl- or dithio- acetals and hemimercaptals as precursors to impart meat-like flavour to canned simulated meat and vegetable products on sterilisation. The patent of Heyland (1977) involved the use of hydrolysed onion, garlic and cabbage in combination with HVP, ribose and beef fat for the preparation of beef flavours. Other flavour companies developed meat flavourings with sulfur sources such as hydrogen, sodium or ammonium sulfides (Godman and Osborne, 1972; Gunther, 1972), methionine (Van Pottelsberghe de la Potterie, 1972) or egg white (Theron et al., 1975). Yeast extracts or yeast hydrolysates have traditionally been used either as precursors for the thermal generation of meat flavourings or as taste-enhancing ingredients, in blends with process flavours. The advantage of yeast extracts is that they are a relatively cheap, natural source of amino acids and thiamine. In addition, their high content of glutamate and 5-ribonucleotides, particularly inosine 5-monophosphate and guanosine 5-monophosphate, provides complexity, body and flavour enhancement. Nestle,« for example, filed several patents in which the preparation of beef and chicken process flavours using yeast extracts was disclosed (Nestle,« 1966; Rolli et al., 1988; Cerny, 1995). Cerny (1995) also developed bouillon flavours using yeast cream, which is enriched in hydrogen sulfide. Such a yeast cream was obtained by incubating baker’s yeast with elemental sulfur. In order to obtain complete meat flavouring products, process flavours are blended with several other ingredients, which provide aroma (e.g. compounded aroma blocks referred to as top notes or topnote flavours), taste, taste enhancement, mouth feel and body (referred to as base notes). Besides topnote flavours, yeast extracts, hydrolysed vegetable proteins and the monosodium salts of glutamate, inosinate and guanylate, ingredients such as onion, garlic, celery and/or caramel powder, animal or vegetable fats, gelatine and spices are often used in meat flavour compositions. The use of some yeast extracts, however, is limited owing to their undesirable ‘yeasty’ character. Therefore, research groups have developed processes for manufacturing yeast hydrolysates with improved meat-like taste, in which the yeasty notes are absent. De Rooij and Hakkaart (1992), for example, improved the meaty character of yeast hydrolysates prepared from several yeast species by combining the enzymatic degradation of yeast cells with an additional fermentation step, which was carried out using lactic acid-producing micro-organisms or additional yeasts. Hyoky¬ et al. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

80 Food Flavour Technology

(1996) developed a method for the production of yeast extracts in which undesirable bitter and yeasty ﬂavour notes were removed. The authors evaluated several non-ionic and slightly basic macroporous polymeric adsorbents as well as activated carbon for their ability to bind bitter and other undesirable ﬂavouring substances of yeast hydrolysates, without binding yeast peptides, amino acids or nucleotides. The best results were obtained with Amberlite XAD-16 and Amberlite XAD-765, which are a non-ionic styrene/divinylbenzene copolymer and a weakly basic phenolformaldehyde polymer, respectively.

3.4.4 Sweet process flavours A few patents and articles on the generation of chocolate and caramel flavours are discussed here. Rusoff (1958) prepared artificial chocolate flavours by heating partially hydrolysed proteins with sugars. The Maillard reactions were performed between protein hydrolysates derived from casein, soy, wheat gluten or gelatine and mixtures of pentoses and hexoses. The reaction medium contained 30% water and the reaction temperature ranged between 130 and 150◦C. These ‘base chocolate flavours’ were rounded off by adding ingredients such as caffeine, theobromine and tannins before or after the heating process. The need for hydrolysed proteins to generate chocolate flavours through Maillard reaction wasstressedbyRodel¬ et al. (1988). The authors based their investigations on precursor studies of Mohr et al. (1971, 1976), who found that only mixtures of peptide and free amino acid fractions isolated from fermented raw cocoa beans developed cocoa aroma on thermal treatment. The study of Rodel¬ et al. (1988) covered a complete range of parameters such as source of protein, rate of hydrolysis, source of enzyme, amount of sugar, water content as well as temperature and time. The Maillard reaction flavours obtained were evaluated organoleptically and analytically. The authors revealed that gelatine that is enzymatically hydrolysed by more than 20% is an appropriate protein source for producing cocoa flavours through Maillard reaction. The quality of the cocoa aroma was further affected by water and sugar contents, whereas the source of enzyme had no influence. A water content of at least 5% and a sugar content of 20 g per 100 g hydrolysed protein gave a positive effect on flavour. In addition, temperature and time, which were the most sensitive parameters, were optimised at 144◦C and 21 minutes. Pittet and Seitz (1974) disclosed processes for the preparation of various flavours resem- bling chocolate, sweet corn, popcorn, bread, cracker and caramel toffee. The flavours were prepared by heating cyclic enolones such as Furaneol, maltol or cyclotene with amino acids in propylene glycol or glycerol. The temperatures ranged between 120 and 205◦C. Choco- late flavours, for example, were obtained when valine or leucine was reacted with maltol or Furaneol, whereas proline (in combination with Furaneol or maltol) yielded cracker-like, popcorn, sweet corn and bread aromas. In addition, caramel toffee and burnt sugar flavours were established from proline and cyclotene or ethyl cyclotene. Gilmore (1988) also developed a caramel butterscotch flavour by heating a mixture of sugar syrup and butter in the presence of ammonia at a pH of 7 and a temperature of about 100◦C.

3.5 OUTLOOK

Flavour formation is a minor but important pathway within the complex cascade of chemical reactions occurring during Maillard processes. A strong focus on the key ﬂavour compounds (aroma and taste-active molecules), their precursors and reaction routes is required for the P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 81

optimisation of process flavours. In addition, the formation of undesirable molecules needs to be taken into account when it comes to the optimisation of processing parameters. The better understanding of the influence of various process parameters on the formation of both the key aroma compounds and their precursors is still a challenge for future research. Experimental design and kinetic parameter estimations are good tools for limiting the amount of experiment required. In addition, the evaluation of important taste compounds derived from the Maillard reaction, the elucidation of their formation pathways as well as the understanding of how the interaction of aroma and taste compounds affects the sensorial quality of the flavours are certainly research areas that are worth investigating.

REFERENCES

Amrani-Hemaimi, M., Cerny, C. and Fay, L.B. (1995) Mechanisms of formation of alkylpyrazines in the Maillard reaction. J. Agric. Food Chem. 43, 2818Ð2822. Anet, E.F.L.J. (1960) 3-Deoxyhexosones. J. Am. Chem. Soc. 82, 1502Ð1504. Anonymous (1979) Foodstuff with flavor precursor. Res. Disclosure 177–188, 17918. Apriyantono, A. and Ames, J.M. (1990) Volatile compounds produced on heating lysine with xylose. In: Flavour Science and Technology (eds Y. Bessiere` and A.F. Thomas), Wiley, Chichester, UK, pp. 117Ð 120. Beksan, E., Schieberle, P., Robert, F., Blank, I., Fay, L.B., Schlichtherle-Cerny, H. and Hofmann, T. (2003) Synthesis and sensory characterization of novel Umami-tasting glutamate glycoconjugates. J. Agric. Food Chem. 51, 5428Ð5436. Belitz, H.-D., Grosch, W. and Schieberle, P. (2008) Food Chemistry, 4th edn, Springer, Berlin. Bel Rhild, B., Fleury, Y., Devaud, S., Fay, L.B., Blank, I. and Juillerat, M.A. (2002) Biogeneration of roasted notes based on 2-acetyl-2-thiazoline and its precursor 2-(1-hydroxyethyl)-4,5-dihydrothiazole. In: Heteroatomic Aroma Compounds, ACS Symposium Series 826 (eds G.A. Reineccius and T.A. Reineccius), American Chemical Society, Washington, DC, pp. 179Ð190. Blank, I., Devaud, S., Matthey-Doret, M., Pollien, P.., Robert, F. and Yeretzian, C. (2003b) Formation of odour-active compounds in Maillard model systems based on proline. In: Flavour Research at the Dawn of the Twenty-First Century, Proceedings of the 10th Weurman Flavour Research Symposium (eds J.-L. Le Quer« eandP.X.Eti« evant),« Lavoisier, Intercept, London, pp. 458Ð463. Blank, I., Devaud, S., Matthey-Doret, M. and Robert, F. (2003a) Formation of odorants in Maillard model systems based on L-proline as affected by pH. J. Agric. Food Chem. 51, 3643Ð3650. Blank, I., Sen, A. and Grosch, W. (1992) Aroma impact compounds of Arabica and Robusta coffee. Colloq. Sci. Int. Cafe 14, 117Ð129. Broderick, J.J. and Linteris, L.L. (1960) Flavoring agents and process for imparting a meat-like flavor to an edible composition. US Patent 2955 041. Buttery, R.G., Haddon, W.F., Seifert, R.M. and Turnbaugh, J.G. (1984) Thiamin odor and bis(2-methyl-3- furyl) disulfide. J. Agric. Food Chem. 32, 674Ð676. Buttery, R.G. and Ling, L.C. (1995) Volatile flavor components of corn tortillas and related products. J. Agric. Food Chem. 43, 1878Ð1882. Buttery, R.G., Ling, L.C. and Juliano, B.O. (1982) 2-Acetyl-1-pyrroline: an important aroma compound of cooked rice. Chem. Ind. (London) 23, 958Ð959. Buttery, R.G., Ling, L.C., Juliano, B.O. and Turnbaugh, J.G. (1983) Cooked rice aroma and 2-acetyl-1- pyrroline. J. Agric. Food Chem. 31, 823Ð826. Cerny, C. (1995) Proced« e« de transformation du soufre el« ementaire« en sulfure. European Patent 778350. Cerny, C. (2003) Maltol generation from lactose in dry systems. In: Flavour Research at the Dawn of the Twenty-First Century, Proceedings of the 10th Weurman Flavour Research Symposium (eds J.L. Le Quer« e« and P.X. Eti« evant),« Lavoisier, Intercept, London, pp. 544Ð547. Cerny, C. (2007) Process flavourings. In: Flavourings. Production, Composition, Applications, Regulations (ed. H. Ziegler), Wiley-VCH, Weinheim, Germany, pp. 274Ð297. Cerny, C. and Davidek, T. (2003) Formation of aroma compounds from ribose and cysteine during Maillard reaction. J. Agric. Food Chem. 51, 2714Ð2721. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

82 Food Flavour Technology

Cerny, C. and Grosch, W. (1993) Quantification of character-impact odor compounds of roasted beef. Z. Lebensm.-Unter. Forsch. 196, 417Ð422. Chan, F. and Reineccius, G.A. (1994) The reaction kinetics for the formation of isovaleraldehyde, 2-acetyl- 1-pyrroline, di(H)di(OH)-6-methylpyranone, phenylacet-aldehyde, 5-methyl-2-phenyl-2-hexenal, and 2- acetylfuran in model systems. In: Maillard Reactions in Chemistry, Food and Health (eds T.P. Labuza, G.A. Reineccius, V. Monnier, J. O’Brien and J. Baynes), The Royal Society of Chemistry, Cambridge, pp. 131Ð139. Chen, T.-K. and Tandy, J.S. (1988) Production of flavour. European Patent 295509. Corn Products Company (1969) Process for the production of seasonings with a flavor similar to meat extract. GB Patent 1148449. Cremer, D.R. and Eichner, K. (2000) The influence of the pH value on the formation of Strecker aldehydes in low moisture model systems and in plant powders. Eur. Food Res. Technol. 211, 247Ð251. Davidek, T., Clety, N., Aubin, S. and Blank, I. (2002) Degradation of the Amadori compound N-(1-deoxy- D-fructos-1-yl)glycine in aqueous model systems. J. Agric. Food Chem. 50, 5472Ð5479. Davidek, T., Devaud, S., Robert, F. and Blank, I. (2006a) Sugar fragmentation in the Maillard reaction cascade: formation of acetic acid by a hydrolytic ␤-dicarbonyl cleavage mechanism. J. Agric. Food Chem. 54, 6667Ð6676. Davidek, T., Illmann, S., Gouezec,« E., Rytz, A., Schuchmann, H.P. and Blank, I. (2009) Effect of reaction conditions on generation of 4-hydroxy-2,5-dimethyl-3(2H)-furanone from rhamnose. In: 12th Weurman Flavor Research Symposium, Interlaken, Switzerland, July 1Ð4, 2008. Davidek, T., Robert, F., Devaud, S., Arce Vera, F. and Blank, I. (2006b) Sugar fragmentation in the Maillard reaction cascade: formation of short-chain carboxylic acids by a new oxidative ␣-dicarbonyl cleavage pathway. J. Agric. Food Chem. 54 6677Ð6684. Decnop, C., van Dort, J.M. and de Hey, J.T. (1990) Hydroxyfuranone preparation. European Patent 398417. De Rooij, J.F.M. and Hakkaart, M.J.J. (1992) A process for the preparation of food flavours. European Patent 191513. de Roos, K.B. (1992) Meat flavor generation from cysteine and sugars. In: Flavor Precursors: Thermal and Enzymatic Conversions (eds R. Teranishi, G.R. Takeoka and M. Guntert),¬ American Chemical Society, Washington, DC, pp. 203Ð216. Doornbos, T., Haring, P.G.M. and Van Der Heijden, A. (1991) Use of diketone in food products. European Patent 293957. Doornbos, T., Van Den Ouweland, G.A.M. and Tjan, S.B. (1981) Amadori compounds, derived from 6-deoxy sugars, as flavour precursor. In: Progress in Food and Nutrition Science, Maillard Reactions in Food (ed. C. Eriksson), Pergamon Press, Oxford, pp. 57Ð63. Eichner, K., Reutter, M. and Wittmann, R. (1994) Detection of Amadori compounds in heated foods. In: Thermally Generated Flavors: Maillard, Microwave, and Extrusion Processes (eds T.H. Parliament, M.J. Morello and R.J. McGorrin), American Chemical Society, Washington, DC, pp. 42Ð54. Fickert, B. (1999) Untersuchungen zur Bildung von Aromastoffen bei der Malzung¬ von Getreide. PhD thesis, Technical University of Munich, Garching, Germany. Frank, O., Jezussek, M., Hofmann, T. (2003) Sensory activity, chemical structure, and synthesis of Mail- lard generated bitter-tasting 1-oxo-2,3-dihydro-1H-indolizinium-6-olates. J. Agric. Food Chem. 51, 2693Ð2699. Gasser, U. (1990) Fluchtige¬ Verbindungen aus gekochtem Rind- und Huhnerfleisch,¬ Fleischaromen und Fleischextrakt. PhD thesis, Technical University of Munich, Garching, Germany. Gasser, U. and Grosch, W. (1988) Identification of volatile flavor compounds with high aroma values from cooked beef. Z. Lebensm.-Unters. Forsch. 186, 489Ð494. Gasser, U. and Grosch, W. (1990) Primary odorants of chicken broth. A comparative study with meat broths from ox and cow. Z. Lebensm.-Unters. Forsch. 190,3Ð8. Gasser, U. and Grosch, W. (1991) Aroma von gekochtem Schweinefleisch. Lebensmittelchemie 45, 15Ð16. Gi, U.-S. and Baltes, W. (1995) Model reactions on roast aroma formation, 15. Investigations on the formation of pyrido[3,4-d]imidazoles during the Maillard reaction. J. Agric. Food Chem. 43, 2226Ð2230. Giacino, C. (1968a) Poultry flavor composition and process. US Patent 3394017. Giacino, C. (1968b) Product and process of reacting a proteinaceous substance with a sulfur-containing compound to provide a meat-like flavor. US Patent 3394015. Giacino, C. (1969) Meat flavor compositions. GB Patent 1146337. Giacino, C. (1970) Meat flavor compositions. US Patent 3519437. Gilmore, L.T. (1988) Process for preparing a caramel butterscotch flavor syrup. US Patent 4753814. Godman, J.L. and Osborne, D.R.D. (1972) Edible products. GB Patent 1285568. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 83

Grosch, W. (1994) Determination of potent odourants in foods by aroma extract dilution analysis (AEDA) and calculation of odour activity values (OAVs). Flavour Frag. J. 9, 147Ð158. Grosch, W. (2001) Flavour of chemistry III: Coffee and volatile compounds. In: Coffee, Recent Developments (eds O. Vitzthum and R.J. Clarke), Blackwell Science, Oxford, pp. 68Ð89. Grosch, W. and Schieberle, P. (1997) Flavor of cereal products Ð a review. Cereal Chem. 74(2), 91Ð97. Guadagni, D.G., Buttery, R.G. and Turnbaugh, J.G. (1972) Odor thresholds and similarity ratings of some potato chip components. J. Sci. Food Agric. 23, 1435Ð1444. Guntert,¬ M., Bertram, H.-J., Hopp, R., Silberzahn, W., Sommer, H. and Werkhoff, P. (1996) Thermal generation of flavor compounds from thiamin and various amino acids. In: Recent Developments in Flavor and Fragrance Chemistry (eds R. Hopp and K. Mori), VCH-Verlagsgesellschaft, Weinheim, Germany, pp. 215Ð240. Guntert,¬ M., Bruing,¬ J., Emberger, R., Hopp, R., Kopsel,¬ M., Surburg, H. and Werkhoff, P. (1992) Thermally degraded thiamin: a potent source of interesting flavor compounds. In: Flavor Precursors: Thermal and Enzymatic Conversions (eds R. Teranishi, G.R. Takeoka and M. Guntert),¬ American Chemical Society, Washington, DC, pp. 140Ð163. Gunther, R. (1972) Flavoring compositions produced by reacting hydrogen sulfide with a pentose. US Patent 3642497. Guth, H. and Grosch, W. (1993a) 12-Methyltridecanal, a species-specific odorant of stewed beef. Lebensm. Wissensch. Technol. 26, 171Ð177. Guth, H. and Grosch, W. (1993b) Identification of potent odorants in static headspace samples of green and black tea powders on the basis of aroma extract dilution analysis (AEDA). Flavour Frag. J. 8, 173Ð178. Guth, H. and Grosch, W. (1994) Identification of the character impact odorants of stewed beef juice by instrumental analysis and sensory studies. J. Agric. Food Chem. 42, 2862Ð2866. Hack, A.W. and Konigsdorf, W. (1969) Production of seasonings with a flavor similar to meat extract. US Patent 3480447. Havela-Toledo, E., Naim, M., Zehavi, U. and Rouseff, R.L. (1997) 4-Hydroxy-2,5-dimethyl-3(2H)-furanone formation in buffers and model solutions of citrus juice. J. Agric. Food Chem. 45, 1314Ð1319. Havela-Toledo, E., Naim, M., Zehavi, U. and Rouseff, R.L. (1999) Effect of L-cysteine and n-acetyl- L-cysteine on 4-hydroxy-2,5-dimethyl-3(2H)-furanone (Furaneol), 5-hydroxymethylfurfural and 5- methylfurfural formation and browning in buffer solutions containing rhamnose or glucose and arginine. J. Agric. Food Chem. 47, 4140Ð4145. Henderson, S.K. and Nawar, W.W. (1981) Thermal interaction of linoleic acid and its esters with valine. J. Am. Oil Chem. Soc. 58, 632Ð635. Herz, W.J. and Schallenberger, R.S. (1960) Some aromas produced by simple amino acid-sugar reaction. Food Res. 25, 491Ð494. Heyland, S. (1977) Verfahren zur Herstellung eines Aromatisierungsmittels. German Patent 2546035. Hidalgo, F.J., Zamora, R. (2004) Strecker-type degradation produced by the lipid oxidation products 4,5- epoxy-2-alkenals. J. Agric. Food Chem. 52, 7126Ð7131. Hidalgo, F.J. and Zamora, R. (2005) Interplay between the Maillard reaction and lipid peroxidation in biochemical systems. Ann. N. Y. Acad. Sci. 1043, 319Ð326. Hidalgo, F. J. and Zamora, R. (2007) Conversion of phenylalanine into styrene by 2,4-decadienal in model systems. J. Agric. Food Chem. 55, 4902Ð4906. Hodge, J.E. (1953) Chemistry of browning reactions in model systems. J. Agric. Food Chem. 1, 928Ð943. Hofmann, T. (1995) Charakterisierung intensiver Geruchsstoffe in Kohlenhydrat/Cystein-Modellreaktionen und Klarung¬ von Bildungswegen. PhD thesis, Technical University of Munich, Garching, Germany. Hofmann, T. (1999) Quantitative studies on the role of browning precursors in the Maillard reaction of pentoses and hexoses with L-alanine. Eur. Food Res. Technol. 209, 113Ð121. Hofmann, T. (2005) Taste-active Maillard reaction products: the ‘tasty’ world of nonvolatile Maillard reaction products. Ann. N. Y. Acad. Sci. 1043, 20Ð29. Hofmann, T., Munch,¬ P.and Schieberle, P.(2000) Quantitative model studies on the formation of aroma-active aldehydes and acids by Strecker-type reactions. J. Agric. Food Chem. 48, 434Ð440. Hofmann, T., Ottinger, H., Frank, O., Soldo, T., Cerny, C., Robert, F. and Blank I. (2001) Use of alpha-keto enamine derivatives as cooling ingredients. Eur. Pat. Appl., EP 1157617 A2. Hofmann, T. and Schieberle, P. (1995) Evaluation of the key odorants in a thermally-treated solution of ribose and cysteine by aroma extract dilution techniques. J. Agric. Food Chem. 43, 2187Ð2194. Hofmann, T. and Schieberle, P.(1997) Identification of potent aroma compounds in thermally-treated mixtures of glucose/cysteine and rhamnose/cysteine using aroma extract dilution techniques. J. Agric. Food Chem. 45, 898Ð906. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

84 Food Flavour Technology

Hofmann, T. and Schieberle, P. (1998a) Quantitative model studies on the effectiveness of different precursor systems in the formation of the intense food odorants 2-furfurylthiol and 2-methyl-3-furanthiol. J. Agric. Food Chem. 46, 235Ð241. Hofmann, T. and Schieberle, P. (1998b) Identification of key aroma compounds generated from cysteine and carbohydrates under roasting conditions. Z. Lebensm.-Unters. Forsch. 207, 229Ð236. Hofmann, T. and Schieberle, P. (2000a) Formation of aroma-active Strecker-aldehydes by a direct oxidative degradation of Amadori compounds. J. Agric. Food Chem. 48, 4301Ð4305. Hofmann, T. and Schieberle, P. (2000b) Acetylformoin Ð an important progenitor of 4-hydroxy-2,5-dimethyl- 3(2H)-furanone and 2-acetyltetrahydropyridine during thermal food processing. In: Flavour 2000 – Perception, Release, Evaluation, Formation, Acceptance, Nutrition/Health (ed. M. Rothe), Eigenverlag, Bergholz-Rehbrucke,¬ Germany, pp. 311Ð322. Hyoky,¬ G., Sarkki, M.-L. and Tylli, M. (1996) Method of producing a yeast extract usable in foodstuffs, where undesirable flavouring agents in the extract have been removed. World Patent 96/38057. IFF, International Flavors & Fragrances, Inc. (1965) Composition and process for preparation thereof. GB Patent 1099711. IFF, International Flavors & Fragrances, Inc. (1967) Flavoring compositions. GB Patent 1069104. Illmann, S., Davidek, T., Gouezec,« E., Rytz, A., Schuchmann, H.P. and Blank, I. (2009) Generation of 4- hydroxy-2,5-dimethyl-3(2H)-furanone from rhamnose as affected by reaction parameters Ð experimental design approach. J. Agric. Food Chem. 57, 2889Ð2895. Ishizu, A., Lindberg, B. and Theander, O. (1967) 1-Deoxy-D-erythro-2,3-hexodiulose, an intermediate in the formation of D-glucosaccharinic acid. Carbohydr. Res. 5, 329Ð334. Jaeggi, K. (1973) Process for the manufacture of meat flavors. US Patent 3761287. Katz, I. and Evers, W.J. (1973) Flavoring compositions and processes utilizing alpha-ketothiols. US Patent 3773524. Kerler, J. (1996) Objektivierung eines Aromafehlers (Warmed-over Flavour) bei aufgewarmtem¬ Huhner-und¬ Rindfleisch. PhD thesis, Technical University of Munich, Garching, Germany. Kerler, J. and Grosch, W. (1996) Odorants contributing to warmed-over flavor (WOF) of refrigerated cooked beef. J. Food Sci. 61(6), 1271Ð1274, 1284. Kerler, J. and Grosch, W. (1997) Character impact odorants of boiled chicken: changes during refrigerated storage and reheating. Z. Lebensm.-Unters. Forsch. 205, 232Ð238. Kerscher, R. (2000) Objektivierung tierartspezifischer Aromaunterschiede bei erhitztem Fleisch. PhD thesis, Technical University of Munich, Garching, Germany. Kerscher, R. and Grosch, W. (1997) Comparative evaluation of potent odorants of boiled beef by aroma extract dilution and concentration analysis. Z. Lebensm.-Unters. Forsch. 204,3Ð6. Kerscher, R. and Grosch, W. (1998) Quantification of 2-methyl-3-furanthiol, 2-furfurylthiol, 3-mercapto-2- pentanone, and 2-mercapto-3-pentanone in heated meat. J. Agric. Food Chem. 46, 1954Ð1958. Keyhani, A. and Yaylayan, V.A. (1996) Elucidation of the mechanism of pyrazidone formation in glycine model systems using labeled sugars and amino acids. J. Agric. Food Chem. 44, 2511Ð2516. Kim, Y.-S., Hartman, T.G., Ho, C.-T. (1996) Formation of 2-pentylpyridine from the thermal interaction of amino acids and 2,4-decadienal. J. Agric. Food Chem. 44, 3906Ð3908. Kroh, L.W. (1994) Caramelisation in food and beverages. Food Chem. 51, 373Ð379. Lane, M.J. and Nursten, H.E. (1983) The variety of odors produced in Maillard model systems and how they are influenced by reaction conditions. In: Maillard Reaction in Food and Nutrition (eds G.R. Waller and M.S. Feather), American Chemical Society, Washington, DC, pp. 141Ð158. Leathy, M.M. and Reineccius, G.A. (1989a) Kinetics on the formation of alkylpyrazines: effect of pH and water activity. In: Thermal Generation of Aromas (eds T.H. Parliament, R.J. McGorrin and C.-T. Ho), American Chemical Society, Washington, DC, pp. 196Ð208. Leathy, M.M. and Reineccius, G.A. (1989b) Kinetics on the formation of alkylpyrazines: effect of amino acid type and type of sugar. In: Flavor Chemistry (eds R. Teranishi, R.G. Buttery and F. Shahidi), American Chemical Society, Washington, DC, pp. 76Ð91. Ledl, F. and Schleicher, E. (1990) New aspects of the Maillard reaction in foods and in the human body. Angew. Chem. Int. Ed. Engl. 29, 565Ð706. Lee, P.S. (1995) Modeling the Maillard reaction Ð a computer-simulation and a discussion of its application to Maillard reaction analysis and design. In: Flavor Technology: Physical Chemistry, Modification, and Process (eds C.-T. Ho, C.T. Tan and C.H. Tong), American Chemical Society, Washington, DC, pp. 74Ð97. Limacher, A., Kerler, J., Davidek, T., Schalzried, F. and Blank, I. (2008) Formation of furan and methylfuran by Maillard-type reactions in model systems and food. J. Agric. Food Chem. 56, 3639Ð3647. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 85

Lin, L.J. (1995) Regulatory status of Maillard reaction flavors. In: Savory Flavors (ed. T.W. Nagodawithana), Esteekay Associates, Wisconsin, pp. 7Ð15. MacLeod, G. and Ames, J. (1986) 2-Methyl-3-(methylthio)furan: a meaty character impact compound from cooked beef. Chem. Ind. 3, 175Ð177. Madruga, M.S. and Mottram, D.S. (1995) The effect of pH on the formation of Maillard-derived aroma volatiles using a cooked meat system. J. Sci. Food Agric. 68, 305Ð310. Madruga, M.S. and Mottram, D.S. (1998) Sensory analysis of a model system using 5-IMP and cysteine at different pH. Cienc.ˆ Tecnol. Aliment. 18(4), 397Ð404. Manley, C.H. (1995) Process flavors and precursor systems. In: Savory Flavors (ed. T.W. Nagodawithana), Esteekay Associates, Wisconsin, pp. 16Ð25. May, C.G. (1961) Flavouring substances and their preparation. GB Patent 858333. May, C.G. and Morton, I.D. (1960) Process for the preparation of a meat flavor. US Patent 2934436. McKenna, A.B. (1988) Influence of heating on the formation of caramel flavour, brown colour and hydroxymethylfurfural in reconstituted concentrated skim milk. New Zeal. J. Dairy Sci. Technol. 23, 363Ð372. Meynier, A. and Mottram, D.S. (1995) The effect of pH on the formation of volatile compounds in meat- related model systems. Food Chem. 52, 361Ð366. Milo, C. and Grosch, W. (1993) Changes in the odorants of boiled trout (Salmo fario) as affected by the storage of the raw material. J. Agric. Food Chem. 41, 2076Ð2081. Milo, C. and Grosch, W. (1995) Detection of odor defects in boiled cod and trout by gas chromato- graphyÐolfactometry of headspace samples. J. Agric. Food Chem. 43, 2076Ð2081. Milo, C. and Grosch, W. (1996) Changes in the odorants of boiled salmon and cod as affected by the storage of the raw material. J. Agric. Food Chem. 44, 459Ð462. Mohr, W., Landschreiber, E. and Severin, T. (1976) Zur Spezifitat¬ des Kakaoaromas. Fett. Wiss. Technol. 78(2), 88Ð95. Mohr, W., Rohrle,¬ M. and Severin, T. (1971) Uber¬ die Bildung des Kakaoaromas aus seinen Vorstufen. Fett. Wiss. Technol. 73(8), 515Ð521. Morton, I.D., Akroyd, P. and May, C.G. (1960) Flavouring substances and their preparation. GB Patent 836694. Mottram, D.S., Ames, J.M., Mlotkicwiez, J.A., Copsey, J.P. and Anderson, A. (1998) Flavouring agents. World Patent 98/42208. Mottram, D.S. and Madruga, M.S. (1994) Important sulfur-containing aroma volatiles in meat. In: Sulfur Compounds in Foods (eds C.J. Mussinan and M.E. Keelan), American Chemical Society, Washington, DC, pp. 180Ð187. Mottram, D.S. and Nobrega, C.C. (1998) Formation of volatile sulfur compounds in reaction mixtures containing cysteine and three different ribose compounds. In: Food Flavors: Formation, Analysis and Packaging Influences (eds E.T. Contis, C.-T. Ho and C.J. Mussinan), Elsevier Science, Amsterdam, pp. 483Ð492. Munch,¬ P. and Schieberle, P. (1998) Quantitative studies on the formation of key odorants in thermally-treated yeast extracts using stable isotope dilution assays. J. Agric. Food Chem. 46, 4695Ð4701. Nedvidek, W., Ledl, F. and Fischer, P. (1992) Detection of 5-hydroxymethyl-2-methyl-3(2H)-furanone and of ␣-dicarbonyl compounds in reaction mixtures of hexoses and pentoses with different amines. Z. Lebens.-Unters. Forsch. 194, 222Ð228. Nestle« (1966) Flavouring agent. GB Patent 1032334. Ottinger, H., Bareth, A. and Hofmann, T. (2001b) Characterization of natural ‘cooling’ compounds formed from glucose and L-proline in dark malt by application of taste dilution analysis. J. Agric. Food Chem. 49, 1336Ð1344. Ottinger, H., Hofmann, T. (2003) Identification of the taste enhancer alapyridaine in beef broth and evaluation of its sensory impact by taste reconstitution experiments. J. Agric. Food Chem. 51, 6791Ð6796. Ottinger, H., Soldo, T., Hofmann, T. (2001a) Systematic studies on structure and physiological activity of cyclic ␣-keto enamines, a novel class of ‘cooling’ compounds. J. Agric. Food Chem. 49, 5383Ð5390. Parliment, T.H., McGorrin, R.J. and Ho, C.-T. (eds) (1989) Thermal Generation of Aromas, American Chemical Society, Washington, DC. Parliment, T.H., Morello, M.J. and McGorrin, R.J. (eds) (1994) Thermally Generated Flavors: Maillard, Microwave, and Extrusion Processes, American Chemical Society, Washington, DC. Pippen, E.L. and Mecchi, E.P. (1969) Hydrogen sulfide, a direct and potentially indirect contributor to cooked chicken aroma. J. Food Sci. 34, 443Ð446. Pittet, A.O., Rittersbacher, P. and Muralidhara, R. (1970) Flavor properties of compounds related to maltol and isomaltol. J Agric. Food Chem. 18, 929Ð932. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

86 Food Flavour Technology

Pittet, A.O. and Seitz, E.W. (1974) Flavoring compositions and processes. US Patent 3782973. Reineccius, G.A. (1990) The influence of Maillard reactions on the sensory properties of foods. In: The Mail- lard Reaction in Food Processing, Human Nutrition and Physiology (eds P.A. Finot, H.U. Aeschbacher, R.F. Hurrell and R. Liardon), Birkhauser¬ Verlag, Basel, Switzerland, pp. 157Ð170. Reineccius, G.A. (1998) Kinetics of flavor formation during Maillard browning. In: Flavor Chemistry: Thirty Years of Progress (eds R. Teranishi, E.L. Wick and I. Hornstein), Kluwer Academic/Plenum Publishers, New York, pp. 345Ð352. Roberts, D.D. and Acree, T.E. (1994) Gas chromatography-olfactometry of glucose-proline Maillard reaction products. In: Thermally Generated Flavors: Maillard, Microwave, and Extrusion Processes (eds. T.H. Parliment, M.J. Morello and R.J. McGorrin), American Chemical Society, Washington, DC, pp. 71Ð79. Rodel,¬ W., Habisch, D. and Ruttloff, H. (1988) Formation of cocoa flavour by Maillard reaction. Charact. Prod. Appl. Food Flavours 2, 301Ð309. Rolli, K., Roschli,¬ D. and Sihver, J.J. (1988) Process for preparing a flavouring ingredient. European Patent 286838. Rosing, E.A.E. and Turksma, H. (1997) Process for the preparation of a savoury flavour. European Patent 784936. Rusoff, I.I. (1958) Flavor. US Patent 2835592. Rychlik, M. and Grosch, W. (1996) Identification and quantification of potent odorants formed by toasting of wheat bread. Food Sci. Technol. (London) 29, 515Ð525. Schieberle, P. (1990) The role of free amino acids present in yeast as precursors of the odorants 2-acetyl- 1-pyrroline and 2-acetyltetrahydropyrridine in wheat bread crust. Z. Lebensm.-Unters. Forsch. 191, 206Ð209. Schieberle, P. (1991a) Primary odorants in popcorn. J. Agric. Food Chem. 39, 1141Ð1144. Schieberle, P. (1991b) Primary odorants of pale lager beer. Differences to other beers and changes during storage. Z. Lebensm.-Unters. Forsch. 193, 558Ð565. Schieberle, P. (1992a) Bildung wichtiger Rostaromastoffe¬ in Lebensmitteln aus Getreide. Getreide, Mehl und Brot 46, 338Ð342. Schieberle, P. (1992b) Formation of Furaneol in heat-processed foods. In: Flavour Precursors – Thermal and Enzymatic Conversion (eds R. Teranishi, G.R. Takeoka and M. Guntert),¬ ACS Symposium Series 490, Washington, DC, pp. 164Ð174. Schieberle, P. (1993a) Studies on the flavour of roasted white sesame seeds. In: Progress in Flavour Pre- cursor Studies – Analysis, Generation and Biotechnology (eds P. Schreier and P. Winterhalter), Allured Publishing, Carol Stream, IL, pp. 343Ð360. Schieberle, P. (1993b) Untersuchungen zum Aromabeitrag und zur Bildung von 4-Hydroxy-2,5-dimethyl- 3(2H)-furanon in thermisch behandelten Lebensmittel. Lebensmittelchemie 47, 15Ð16. Schieberle, P. (1995) Quantification of important roast-smelling odorants in popcorn by stable isotope dilution assays and model studies on flavor formation during popping. J. Agric. Food Chem. 43, 2442Ð 2448. Schieberle, P. (1996) Odor-active compounds in moderately roasted sesame. Food Chem. 55, 145Ð152. Schieberle, P. (2005) The carbon module labeling (CAMOLA) technique: a useful tool for identifying transient intermediates in the formation of Maillard-type target molecules. Ann. N. Y. Acad. Sci. 1043, 236Ð248. Schieberle, P., Fischer, R. and Hofmann T. (2003) The carbohydrate module labeling technique: a useful tool to clarify formation pathways of aroma compounds formed in Maillard-type reactions. In: Flavour Research at the Dawn of the Twenty-First Century, Proceedings of the 10th Weurman Flavour Research Symposium (eds. J.-L. Le Quer« eandP.X.Eti« evant),« Lavoisier, Intercept, London, 447Ð452. Schieberle, P. and Grosch, W. (1987) Evaluation of the flavor of wheat and rye bread crusts by aroma extract dilution analysis. Z. Lebensm.-Unters. Forsch. 185, 111Ð113. Schieberle, P. and Grosch, W. (1994) Potent odorants of rye bread crust. Differences from the crumb and wheat bread crust. Z. Lebensm.-Unters. Forsch. 198, 292Ð296. Schieberle, P. and Hofmann, T. (1998) Characterization of key odorants in dry-heated cysteine-carbohydrate mixtures: comparison with aqueous reaction systems. In: Flavor Analysis (eds C.J. Mussinan and M.J. Morello), American Chemical Society, Washington, DC, pp. 320Ð330. Schieberle, P. and Hofmann, T. (2002) New results on the formation of important Maillard aroma compounds. In: Advances in Flavours and Fragrances: From the Sensation to the Synthesis (ed. K.A. Swift), Royal Society of Chemistry, Cambridge, pp. 163Ð177. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

Basic chemistry and process conditions for reaction ﬂavours 87

Schieberle, P., Hofmann, T. and Munch, P. (2000) Studies on potent aroma compounds generated in Maillard- type reactions using the odor-activity-value concept. ACS Symp. Ser. 756, 133Ð150. Schlichtherle-Cerny, H., Affolter, M., Blank, I., Cerny, C., Robert, F., Beksan, E., Hofmann, T. and Schieberle, P. (2002) Amadori and Heyns rearrangement products as flavoring compounds for imparting umami taste to food products. Eur. Pat. Appl. EP 1252825 A1. Schnermann, P. and Schieberle, P. (1997) Evaluation of key odorants in milk chocolate and cocoa mass by aroma extract dilution analysis. J. Agric. Food Chem. 45, 867Ð872. Schonberg,¬ A. and Moubacher, R. (1952) The Strecker degradation of a-amino acids. Chem. Rev. 50, 261Ð277. Semmelroch, P. and Grosch, W. (1995) Analysis of roasted coffee powders and brews by gas chromatographyÐolfactometry of headspace samples. Food Sci. Technol. (London) 28, 310Ð313. Semmelroch, P., Laskawy, G., Blank, I. and Grosch, W. (1995) Determination of potent odorants in roasted coffee by stable isotope dilution assays. Flavour Frag. J. 10,1Ð7. Shibamoto, T. (1983) Heterocyclic compounds in browning and browning/nitrite model systems. In: Instru- mental Analysis of Foods,Vol.1 (eds G. Charalambous and G. Inglett), Academic Press, New York, pp. 229Ð277. Shu, C.-K. and Ho, C.-T. (1989) Parameter effects on the thermal reaction of cystine and 2,5-dimethyl-4- hydroxy-3(2H)-furanone. In: Thermal Generation of Aromas (eds T.H. Parliment, R.J. McGorrin and C.-T. Ho), American Chemical Society, Washington, DC, pp. 229Ð241. Soldo, T., Blank, I., Hofmann, T. (2003) (+)-(S)-Alapyridaine Ð a general taste enhancer? Chem. Senses 28, 371Ð379. Stahl, H.D. and Parliment, T.H. (1994) Formation of Maillard products in the proline-glucose model system Ð high-temperature short-time kinetics. In: Thermally Generated Flavors, Maillard, Microwave and Ex- trusion Processes (eds T.H. Parliment, M.J. Morello and R.J. Mc Gorrin), American Chemical Society, Washington, DC, pp. 251Ð262. Tai, C.-Y. and Ho, C.-T. (1997) Influence of cysteine oxidation on thermal formation of Maillard aromas. J. Agric. Food Chem. 45, 3586Ð3589. Tandy, J.S. (1985) Chicken flavorants and the processes for preparing them. GB Patent 2157538. Teranishi, R., Buttery, R.G. and Guadagni, D.G. (1975) In: Geruchs- und Geschmacks-stoffe (ed. F. Drawert), Verlag Hans Carl, Nurnberg,¬ Germany, pp. 177Ð186. Theron, P.P.A., von Fintel, R.G.H., Saisselin, A.L. and Vissers, A.M. (1975) Flavouring agents. GB Patent 1382335. Tressl, R. (1989) Formation of flavor compounds in roasted coffee. In: Thermal Generation of Aromas (eds T.H. Parliment, R.J. McGorrin and C.T. Ho), American Chemical Society, Washington, DC, pp. 285Ð301. Tressl, R., Helak, B., Kersten, E. and Rewicki, D. (1993) Formation of flavor compounds by Maillard reaction. In: Recent Developments in Flavor and Fragrance Chemistry (eds R. Hopp and K. Mori), Verlag Chemie, Weinheim, Germany, pp. 167Ð181. Tressl, R., Nittka, C. and Kersten, E. (1995) Formation of isoleucine-specific Maillard products from [1- 13C]-d-glucose and [1-13C]-d-fructose. J. Agric. Food Chem. 43, 1163Ð1169. Tressl, R. and Rewicki, D. (1999) Heat generated flavors and precursors. In: Flavor Chemistry: Thirty Years of Progress (eds R. Teranishi, E.L. Wick and I. Hornstein), Kluwer Academic/Plenum Publishers, New York, pp. 305Ð325. Turksma, H. (1993) Process for the preparation of savoury flavours. European Patent 571031. Ullrich, F. and Grosch, W. (1987) Identification of the most intense volatile flavor compounds formed during autoxidation of linoleic acid. Z. Lebensm.-Unters. Forsch. 184, 277Ð282. Van den Ouweland, G.A.M., Demole, E.P. and Enggist, P. (1989) Process meat flavor development and the Maillard reaction. In: Thermal Generation of Aromas (eds T.H. Parliment, M.J. Morello, R.J. McGorrin and C.T. Ho), American Chemical Society, Washington, DC, pp. 433Ð441. Van den Ouweland, G.A.M. and Peer, H.G. (1968) Mercapto furane and mercapto thiophene derivatives. GB Patent 1283912. Van den Ouweland, G.A.M. and Peer, H.G. (1970) Synthesis of 3,5-dihydroxy-2-methyl-5,6-dihydropyran- 4-one from aldohexoses and secondary amine salts. Recl. Trav. Chim. Pay. B. 89, 750Ð754. Van den Ouweland, G.A.M. and Peer, H.G. (1975) Components contributing to beef flavor. Volatile compounds produced by the reaction of 4-hydroxy-5-methyl-3(2H)-furanone and its thio analog with hydrogen sulfide. J. Agric. Food Chem. 23, 501Ð505. Van Pottelsberghe de la Potterie, P.J. (1972) Verfahren zur Herstellung von Geschmacksstoffen mit Rinder- bratenaroma. German Patent 2149682. Van Pottelsberghe de la Potterie, P.J. (1973) Beef flavor. US Patent 3716380. P1: SFK/UKS P2: SFK c03 BLBK221-Taylor/Linforth November 17, 2009 14:4 Printer Name: Yet to Come

88 Food Flavour Technology

Vauthey, S., Leser, M. and Milo, C. (1998) An aroma product comprising saturated C16 and C18 monoglyc- erides. European Patent 1008305. Wagner, R. and Grosch, W. (1997) Evaluation of potent odorants of French fries. Food Sci. Technol. (London) 30, 164Ð169. Weenen, H. (1998) Reactive intermediates and carbohydrate fragmentation in Maillard chemistry. Food Chem. 62, 393Ð401. Weenen, H. and Apeldoorn, W. (1996) Carbohydrate cleavage in the Maillard reaction. In: Flavour Science: Recent Developments (eds A.J. Taylor and D.S. Mottram), The Royal Society of Chemistry, Cambridge, pp. 211Ð216. Weenen, H., Kerler, J. and van der Ven, J. (1997) The Maillard reaction in flavour formation. In: Flavours and Fragrances (ed. K.A.D. Swift), The Royal Society of Chemistry, Cambridge, pp. 153Ð171. Weenen, H. and Tjan, S.B. (1992) Analysis, structure, and reactivity of 3-deoxyglucosone. In: Flavor Precursors: Thermal and Enzymatic Conversions (eds R. Teranishi, G.R. Takeoka and M. Guntert),¬ American Chemical Society, Washington, DC, pp. 217Ð231. Weenen, H. and Tjan, S.B. (1994) 3-Deoxyglucosone as flavour precursor. In: Trends in Flavour Research (eds H. Maarse and D.G. van der Heij), Elsevier Science, Amsterdam, pp. 327Ð337. Weenen, H. and van der Ven, J.G.M. (1999) Formation of Strecker aldehydes. In: Book of Abstracts, 218th ASC National Meeting, New Orleans, American Chemical Society, Washington, D.C. Weenen, H., van der Ven,J.G.M., van der Linde, L.M., van Duynhoven, J. and Groenewegen, A. (1998) C4, C5, and C6 3-deoxyosones: structures and reactivity. In: The Maillard Reaction in Foods and Medicine (eds J. O’Brien, H.E. Nursten, M.J.C. Crabbe and J.M. Ames), The Royal Society of Chemistry, Cambridge, pp. 57Ð64. Werkhoff, P., Bretschneider, W., Emberger, R., Guntert,¬ M., Hopp, R. and Kopsel,¬ M. (1991) Recent developments in the sulfur flavour chemistry of yeast extracts. Chem. Mikrobiol. Technol. Lebensm. 13, 30Ð57. Whitfield, F.B. (1992) Volatiles from interactions of Maillard reactions and lipids. Crit. Rev. Food Sci. Nutr. 31, 1Ð58. Whitfield, F.B. and Mottram, D.S. (1999) Investigation of the reaction between 4-hydroxy-5-methyl-3(2H)- furanone and cysteine or hydrogen sulfide at pH 4.5. J. Agric. Food Chem. 47, 1626Ð1634. Widder, S., Sabater Luntzel,¬ C., Dittner, T. and Pickenhagen, W. (2000) 3-Mercapto-2-methylpentan-1-ol, a new powerful aroma compound. J. Agric. Food Chem. 48, 418Ð423. Yaylayan,V.A.,Forage, N.G. and Mandeville, S. (1994) Microwave and thermally induced Maillard reactions. In: Thermally Generated Flavors: Maillard, Microwave, and Extrusion Processes (eds T.H. Parliament, M.J. Morello and R.J. McGorrin), American Chemical Society, Washington, DC, pp. 449Ð456. Yaylayan, V.A. and Huyghues-Despointes, A. (1994) Chemistry of Amadori rearrangement products: analysis, synthesis, kinetics, reactions and spectroscopic properties. Crit. Rev. Food Sci. Nutr. 34, 321Ð369. Yaylayan, V.A. and Keyhani, A. (1999) Origin of 2,3-pentanedione and 2,3-butanedione in d-glucose/l- alanine Maillard model systems. J. Agric. Food Chem. 47, 3280Ð3284. Yaylayan, V.A. and Keyhani, A. (2000) Origin of carbohydrate degradation products in l-alanine/d- [13C]glucose model systems. J. Agric. Food Chem. 48, 2415Ð2419. Yaylayan, V. and Sporns, P. (1987) Novel mechanisms for the decomposition of 1-(amino acid)-1-deoxy-d- fructoses (Amadori compounds): a mass spectrometric approach. Food Chem. 26, 283Ð305. Yeretzian, C., Blank, I. and Palzer, S. (2007) Process flavourings. In: Flavourings. Production, Composition, Applications, Regulations (ed. H. Ziegler), Wiley-VCH, Weinheim, Germany, pp. 549Ð572. Zamora, R. and Hidalgo, F.J. (2005) Coordinate contribution of lipid oxidation and Maillard reaction to the nonenzymatic food browning. Crit. Rev. Food Sci. Nutr. 45, 49Ð59. Zhang, Y. and Ho, C.-T. (1989) Volatile compounds formed from thermal interaction of 2,4-decadienal with cysteine and glutathione. J. Agric. Food Chem. 37, 1016Ð1020. Zhang, Y. and Ho, C.-T. (1991) Formation of meatlike aroma compounds from thermal reaction of inosine 5-monophosphate with cysteine and glutathione. J. Agric. Food Chem. 39, 1145Ð1148. Zheng, Y., Brown, S., Ledig, W.O., Mussinan, C. and Ho, C.-T. (1997) Formation of sulfur-containing flavor compounds from reactions of Furaneol and cysteine, glutathione, hydrogen sulfide, and alanine/hydrogen sulfide. J. Agric. Food Chem. 45, 894Ð897. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

4 Biotechnological ﬂavour generation Ralf G. Berger, Ulrich Krings and Holger Zorn

4.1 INTRODUCTION

Aromas and flavours possess antimicrobial, medicinal and signalling properties, as well as acting as food preservatives. They were even used to preserve human corpses in ancient Egypt. Above all, it is their alluring sensory properties that promised financial profits high enough to prompt the Phoenicians, Arabs, and, later, the Portuguese, Dutch, Spanish and Venetiansto discover and conquer entire countries. Today’s captains are scientists, their ocean is the metabolic flow, their ships and weapons are gene shuttles and biochemical knowledge, and their targets are no longer leaves, fruits and seeds, but cells, enzymes and genes. This chapter describes the following: r How micro-organisms transform flavour precursors directly into flavour molecules (biotransformation) or produce flavours along multi-step processes (bioconversion and de r novo synthesis). r How enzymes catalyse hydrolytic or other chemical reactions leading to flavours. r How single plant cell in sterile culture may replace field-grown flavour producers. r How the dissemination of methods of genetic engineering fertilises bioflavour research. How laboratory developments have been successfully transferred into industrial applications.

No mention is made of flavours derived from traditional and genetically engineered starter cultures, because informative reviews exist, e.g. on fermented meat flavours (Tjener and Stahnke, 2007), on dairy flavours (Smit et al., 2004), on bread flavours (Hansen and Schieberle, 2005) and on wine flavours (Swiegers et al., 2008).

4.2 NATURAL FLAVOURS: MARKET SITUATION AND DRIVING FORCES

Beverages are the largest market segment for added flavours, accounting for about one-third of the worldwide flavour sales. Around 90% of the flavours added to beverages in the EC are natural, and about 80% in the US; savoury products contain about 80% naturals in both the EC and the US, while the percentages for flavoured dairy products are 50% in the EC and 75% in the US (the remaining proportions are nature-identical and artificial flavours). These numbers clearly reflect a strong consumer preference for naturalness. This scientifically P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

90 Food Flavour Technology

unfounded, vague chemophobia has also been extended to furniture, clothing and cosmetics. As a result, the attribute ‘natural’ is an excellent marketing point, and the use of ‘natural’ flavour sources is the last bastion for food manufacturers to suggest the naturalness of their product explicitly. According to the definitions of the Code of Federal Regulations in the US (CFR, 1993) and to the mandatory guidelines of the Council of the European Communities (88/388/EWG of 22 June 1988; 91/71/EWG and 91/72/EWG of 16 January 1991), aromas generated by biotechnology are classified as natural if the starting materials used were ‘natural’. The revised flavour regulation, scheduled for coming into effect in autumn 2010, will no longer distinguish between nature-identical and artificial flavours; they will be termed ‘flavouring substances’ (EC 1334/2008 of 16 December 2008). Article 3.2(c) defines a ‘natural flavouring substance’ as ‘a flavouring substance obtained by appropriate physical, enzymatic or microbiological processes from material of vegetable, animal or microbiological origin either in the raw state or after processing for human consumption by one or more of the traditional food preparation processes listed in Annex II. Natural flavouring substances correspond to the substances that are naturally present and have been identified in nature’ (www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//TEXT+TA+P6-TA-2008- 0331+0+DOC+XML+V0//EN). The flavour industry, accounting for some legal uncertainties, has adopted a product- oriented point of view on the basis of the GRAS (generally recognised as safe; Waddell et al., 2007) procedure and has set its own standards. The guidelines of the International Organisation of the Flavour Industry demand that the bioprocessing of flavours is in accordance with good manufacturing practice and with the general principles of hygiene of the Codex Alimentarius. Bioflavours shall comply with national legislation, and safety must be ‘adequately established’. Doubts about contamination with microbial toxins or the like have been resolved as the volatile nature of aroma compounds facilitates efficient downstream processing and exclusion of recombinant nucleic acids or other undesired non-volatiles by distillation or lipophilic extraction. This is thought to sufficiently provide protection of the consumer’s health. The ‘all-natural’ mania of the consumer and the favourable legal situation are, however, only two of the forces driving the development of bioflavours. A bundle of driving forces that may be subdivided into ‘business pull’ and ‘technical push’ can be listed. Among the business factors listed are the decline of availability of some traditional raw materials, an increasing market size, modern processes and ingredient formulations, and the growing importance of functional, ethnic and exotic products (Bauer, 2000; Cheetham, 2004). Tech- nical factors include again and again improved analytical methods (Steinhart et al., 2000), improved bioprocessing techniques, genetic engineering, and an improving understanding of structureÐactivity relationships of flavours and fragrances (for example, ‘olfactophore models’; Kraft et al., 2000).

4.3 ADVANTAGES OF BIOCATALYSIS

If a chemosynthetic route results in a mixture of products or isomers, a subsequent separation may be more expensive than a comparable but more selective bioprocess. Trace impurities of a chemosynthetic compound may adulterate the sensory character of a product, and sensory activity is often dependent on an exact stereochemical structure, as the sense of smell is chiral. Biocatalysts not only provide high regiospeciﬁcity and stereospeciﬁcity but show P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 91

high reaction velocity even at low molar fraction. Although biocatalysts are not free from inherent drawbacks, such as operational instability (isolated enzymes), processing cost and sometimes occurrence of side reactions, the advantages of biocatalysts, such as ecological compatibility or multi-step synthesis, cannot be matched by a chiral chemical catalyst. While the closing of mass cycles and sustainable production have gained high priority in the chemical industries, nature itself sets the example. Biocatalysis will not provide immediate solutions for all ﬁne chemicals, but highly prized ﬂavours and fragrances appear to present a particularly suitable playground for biotechnologists. Many authors have discussed one class or various groups of aroma compounds under quite different aspects. For an introduction to the literature until 1997 and a description of some processes already operating, the reader is referred to Berger (1995), Eti« evant« and Schreier (1995) and Berger (1997). More recent biotechnological advances, which will broaden and facilitate the use of biocatalysts for industrial production of volatile chemicals, are described in this chapter.

4.4 MICRO-ORGANISMS

Intact microbial cells possess active transport systems, enzyme arrangements optimised by evolution, and they regenerate their biocatalytic molecules together with the co-factors required; costs arise for the equipment needed to maintain a suitable biochemical environment, but costs for isolation and stabilisation of the biocatalyst are inapplicable. Bacterial pathways to volatile flavours are often based on strong hydrolytic properties. Incomplete substrate oxidation and Strecker degradation of amino acids yield a spectrum of products (Rizzi, 2008). Yeasts, as unicellular non-filamentous fungi, are more complex eukaryotic cells that show a much more diverse biochemistry including a broad range of volatile metabolites, such as esters, lactones, aldehydes and phenolics (Debourg, 2000). The most developed fungal species, the class of basidiomycetes, show a complicated sexual cycle, pseudo-tissue formation, and the distinct ability to degrade native cellulose or lignin aerobically. Volatile flavours from all chemical classes were found in basidiomycete fruit bodies and cell cultures, with a particular emphasis on volatile phenylpropanoic and phenolic compounds (Berger and Zorn, 2004; Wu et al., 2007).

4.4.1 Biotransformation and bioconversion of monoterpenes Many essential oils are dominated by monoterpene hydrocarbons. Because of their low sensory activity, low water solubility and tendency to autoxidise and polymerise, they are usually rectiﬁed from the oil and regarded as processing waste. These properties in conjunction with their role as physiological precursors of high-valued oxyfunctionalised terpenoids turn terpene hydrocarbons, such as limonenes, pinenes and terpinenes, into ideal starting materials for microbial transformations (Berger et al., 1999b; de Carvalho and da Fonseca, 2006). The amount of R-(+)-limonene separated from cold-pressed citrus peel oil was estimated at 36 000 tons per year, and the steam distillation of pine oils delivered 160 000 tons of ␣-pinene and 26 000 tons of ␤-pinene (Ohloff, 1994; Nonino, 1997). The ﬁrst systematic studies on terpene hydrocarbon transformation date back to the early 1960s (Bhattacharyya et al., 1960; Prema and Bhattacharyya, 1962), and since then a vast number of publications on the subject has appeared. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

92 Food Flavour Technology

Fig. 4.1 Products of allylic oxidations: (I) carveol; (II) perillyl alcohol; (III) p-mentha-1,8-diene-4-ol; (IV) p-mentha-2,8-diene-1-ol; (V) isopiperitenol; (VI) verbenol; (VII) myrtenol.

4.4.1.1 Allylic hydroxylation From a biotechnological point of view, the allylic hydroxylation is one of the most important transformation reactions leading to compounds with direct or indirect economic importance (carveol, borneol, verbenol, nootkatol). Concurrent rearrangements of double bonds are explained either by radical transition states, as they occur at the end of the cytochrome P450 mono-oxygenase cycle, or by radical-cationic intermediates, typical of peroxidase-mediated reactions. Transformation organisms range from bacteria, such as Pseudomonas through deuteromycetes, such as Penicillium or Aspergillus to higher fungi, such as the basidiomycetes. A number of frequently occurring transformation products of limonene and ␣-pinene are compiled in Fig. 4.1. High regioselectivities have been observed. While some strains preferentially attacked ring carbons, other strains hydroxylated exocyclic allylic positions. Because of its abundance, limonene has been the most popular transformation substrate in recent years. Its regiospecific hydroxylation by a strain of Pseudomonas putida yielded up to 3 g of perillic acid per litre within 5 days (Speelmans et al., 1998). The success of this transformation was based on the selection of a limonene-tolerant micro-organism, biphasic operation, and careful optimisation of reaction conditions including the co-substrate glycerol. The same strategy of pre-screening microbial strains, using substrate-enriched nutrient media, succeeded in the case of Pseudomonas alcaligenes, which transformed (+)- limonene into the hydration product ␣-terpineol (Teunissen and de Bont, 1995). However, when 120 Gram-positive bacterial strains were selected by their growth on carvone as the sole source of carbon, none of them transformed limonene efficiently to carvone (Van der Werf and De Bont, 1998). Other regiospecific transformations of (+)-limonene were observed for the basidiomycete Pleurotus sapidus, which produced mainly carveols and carvone (Onken and Berger, 1999a), and for a non-conventional Hormonema black yeast to yield P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 93

trans-isopiperitenol (Van Rensburg et al., 1997). In the latter case, a varying morphological appearance was associated with unstable product yields. The enantiospecific transformation of racemic limonene to (+)-␣-terpineol by Penicillium digitatum (Tan et al., 1998) encouraged these researchers to immobilise the cells in calcium alginate and to perform an airlift reactor study (Tan and Day, 1998). While good molar conversion was achieved, the absolute yields remained in the lower milligram per litre range. Marostica and Pastore (2007) have covered recent work on limonene biotransformation. Starting with a report on the microbial hydroxylation of ␣-pinene (Bhattacharyya et al., 1960), the history of pinene transformation now spans more than four decades. Work on the transformation of the pinenes, as well as other bicyclic monoterpenes, such as cineoles, camphor and carene, is summarised by Trudgill (1994). Recently, classical (ultra-violet) UV- mutagenesis preceded the selection of strains of Aspergillus, which converted ␣-pinene either to verbenol (Agrawal et al., 1999) or to verbenone (Agrawal and Joseph, 2000). This group has recently studied (+)-␣-pinene transformation by a Pseudomonas strain and found high yields of several volatiles, among them a novel flavour, dihydrocarveol acetate (Divyashree et al., 2006). As long as the public concern about genetic engineering applications in the food sector persists, random mutagenesis will continue to claim scientific interest. The transformation of monoterpenols, such as citronellol, geraniol and nerol, using Botry- tis cinerea followed the same general routes (Bock et al., 1988). Cystoderma carcharias,a basidiomycete, transformed citronellol to 3,7-dimethyl-1,6,7-octanetriol and several minor products, among them the flavour-impact compounds E/Z-rose oxide (Onken and Berger, 1999b). Nerol or citral was transformed by isolates of Penicillium and Aspergillus to a number of other monoterpenes and to the fruity smelling C2-shortened 6-methyl-5-hepten-2-one (Demyttenaere and De Pooter, 1998; Demyttenaere and De Kimpe, 2000). A microculti- vation method in conjunction with solid-phase micro-extraction was used to speed up the process monitoring on a laboratory scale. Another study confirmed the minor role of yeasts in the formation of terpenes of wine and beer flavour (King and Dickinson, 2000). Neither Saccharomyces cerevisiae nor Kluyveromyces lactis or Torulospora delbrueckii accumulated significant amounts of the transformation products Ð linalool, ␣-terpineol and terpin hydrate. Low yields and the lack of stereoselectivity for linalool and ␣-terpineol suggest that chemical transformation overlapped with the biocatalysed routes.

4.4.1.2 Oxidation of non-activated carbons This reaction may be called the Holy Grail of biotechnology. However, there will be a low chance of incidence with oligo-isoprenoid compounds, as most of them carry carbonÐcarbon double bonds, where attack will be favoured. Aliphatic bicyclic compounds, such as fenchol, borneol and 1,8-cineole, were shown by the group of Kieslich to be selectively converted to mono- and diols by a Zygomycete, Diplodia bisporus and by Bacillus cereus (Abraham et al., 1988). A chemically related reaction is the terminal hydroxylation of saturated fatty acids catalysed by Torulopsis yeasts, a reaction permitting access to macrocyclic musks (Williams, 1999).

4.4.1.3 Epoxidation Epoxidations are typical cytochrome P450-mediated reactions. The reaction products were either isolated as stable products or considered to be intermediates, on the basis of the P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

94 Food Flavour Technology

Fig. 4.2 Degradation pathway of ␣-pinene in Pseudomonas ﬂuorescens.

vicinal diols identified. Obviously, the pH of the incubation medium governs the extent of hydrolysis. Asymmetric dihydroxylation of alkene moieties was obtained when strains of P. putida were exposed to isoprene or to related dienes (Boyd et al., 2000). Further, diol oxidation was successfully controlled by addition of propylene glycol as an inhibitor. Substrates with high ring tension, such as ␣-pinene epoxide, were often transformed by fungal strains to a number of products, among them campholenic aldehyde, trans-sobrerol, and E/Z- carveols. In such cases, biocatalysis appears at first glance not to have any advantage over a chemical transformation. The same epoxide substrate, however, was efficiently degraded by Pseudomonas fluorescens and Nocardia strains to a small set of products belonging to one specific pathway (Best et al., 1987; Griffiths et al. 1987; Fig. 4.2). An unusual (4R,8R)-limonene-8,9-epoxide was the only transformation product when a Xanthobacter isolate selected on cyclohexane as the sole carbon source was exposed to (4R)-limonene (Van der Werf et al., 2000). Many other transformation pathways of limonene were compiled by the same authors (Van der Werf et al., 1999). The conversion of ␤-myrcene by the edible fungus Pleurotus ostreatus was investigated using trideutero-labelled ␤-myrcene or presumed intermediates as the substrates. Myrcene diols were formed from the cleavage of several myrcene epoxides identified as the immediate reaction products (Krings et al., 2008). The diverse substrates and reactions catalysed by cytochrome P450 isoforms were summarised in a recent review (Bernhardt, 2006).

4.4.1.4 Oxidation of alcohols Interest in the widespread oxidation of primary and secondary terpenols rests on the fact that many terpene aldehydes and ketones are more volatile and more powerful (in ﬂavour terms) than the corresponding alcohols. Yeasts and some higher fungi may deliver the target compounds in yields close to 100%. Further oxidation to carboxylic acids is rarely observed (Fig. 4.2 shows one of the few exceptions), because the free acyl moieties are quickly transformed to activated conjugates for further metabolism through ␤-oxidation and related pathways. Another pathway along which terpene acids are formed is the BaeyerÐVilliger oxidation: P. putida strains were able to insert an oxygen atom next to a carbonyl function, and the subsequent ring opening yielded a degradable acid intermediate (Abraham et al., 1988).

4.4.1.5 Hydration Hydration of a double bond or of a ring bridge may proceed as an acid- or enzyme-catalysed reaction. As always, a cell-free control experiment is indispensable to assess the extent of chemical side reactions. As a rule of thumb, troublesome chemical hydration occurs at ambient temperature only at a pH Ͻ 4. Useful information on this aspect is sometimes hidden in reports on off-ﬂavour formation in acidic food matrices, such as citrus juices (Haleva-Toleo P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 95

et al., 1999). The required biocatalytic activity appears to be located mainly in bacteria, such as Escherichia coli, Bacillus stearothermophilus and Pseudomonos gladioli, but also in P. digitatum. Candida tropicalis transformed ␣-pinene to ␣-terpineol in a previously un- matched yield of 77% (Chatterjee et al., 1999a).

4.4.2 Bioconversion of C13-norisoprenoids and sesquiterpenes The same fundamental transformation steps as for monoterpenes apply for larger oligo- isoprenoids and related compounds. The increased structural complexity, however, allows more substructures of the molecules to fit into the active sites of enzymes, resulting in a mixture of products with a broad range of concentrations. C13-Norisoprenoids of the ionone/damascone group have received particular attention, because they belong to the small group of highly potent character-impact components with pleasant floral fruity flavours. In tomatoes and maize, a carotenoid cleavage dioxygenase was found to asymmetrically cleave numerous carotenoids to yield, for example, 6-methyl-5-hepten-2-one from lycopene (Vogel et al., 2008). Similar activities were reported from higher fungi. A concerted biogeneration of ␤-ionone and some related compounds has become known using submerged grown fungal mycelium (Zorn et al., 2003a). Several of these fungal enzymes, all so far belonging to the large peroxidase family, were isolated from the source basidiomycete, sequenced, and cloned, and one of them is now commercial (de Boer et al., 2005). A summary on this subject is available (Rodriguez-Bustamante and Sanchez, 2007). A broad screening of Streptomyces strains showed that some strains transformed ␣-ionone regioselectively to 3-hydroxyionone without many side products (Lutz-Wahl et al., 1998). Racemic ␣-ionone was selectively transformed to trans-(3R,6R) and (3S,6S) alcohols. In contrast, there was limited transformation of ␤-ionone to the 4-hydroxy product with no formation of the 3-hydroxy product. Strains of Aspergillus niger metabolised ␤-ionone much better and yielded almost 100% conversion to 3- and 4-hydroxy products in a fed- batch process (Larroche et al., 1995). An advanced mathematical model of the mass transfer rates in this two-phase system was presented (Grivel et al., 1999). Because of the observation that the growth inhibition of B. cinerea by patchoulol faded with time, the metabolic fate of the fungicide was followed (Aleu et al., 1999). Numer- ous hydroxy compounds were found, indicating first steps of a detoxification pathway. As these hydroxy compounds are substrates for further degradation, only transient accumulation of volatile oxidation products can be expected. The group of Miyazawa has issued a series of papers looking, for example, at the conversion of substrates, such as (+)-cedrol, (+)-aromadendrene, (−)-␣-bisabolol, ␤-selinene, (−)-globulol, ␥-gurjunene and farnesol isomers (Miyazawa et al., 1997, 1998; Nankai et al., 1998, and references therein). A plant pathogenic fungus, Glomerella cingulata, was the strain of choice, and the problem of low substrate solubility was partially circumvented by using monoalcohols instead of the hydrocarbon compounds. As a general outline, exomethylene carbons and isopropyl substituents were oxidised non-stereoselectively, while the oxidation at ring positions proceeded stereoselectively, indicating a more restricted conformational situation at the active site of the enzyme. Some evidence is accumulating that ␤-ionone and other terpenes (De-Oliveira et al., 1999) and sesquiterpenes (Sime, 2000) are quite efficient inhibitors of cytochrome enzymes. This would mean that low substrate solubility was responsible for the not only low product P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

96 Food Flavour Technology

yields reported in the past but also the inhibition of the ﬁrst step of substrate detoxiﬁcation, particularly if the substrate was administered to the cells all at once.

4.4.3 Generation of oxygen heterocycles A number of oxygen-containing heterocycles, such as the well-known flavour compounds maltol and Furaneol (and some alkanolides), have received much attention because, besides possessing flavour themselves, they may also affect the smell, taste and umami impressions of other flavour compounds. Furaneol (2,5-dimethyl-4-hydroxy-2H-furan-3-one) was obtained by converting 6-deoxyhexoses along a Maillard-imitating soft chemistry route (Whitehead, 1998). This approach suffers, however, from the lack of economic sources of the precursor 6-deoxy sugars. Furaneol and the related 2- (or 5-) ethyl-5- (or 2-) methyl-4-hydroxy- 2H-furan-3-one have been frequently found in oriental food fermented in the presence of certain yeasts (Hayashida et al., 1998). Concentrations of the two compounds (in the milligram per litre range) accumulated after supplementing Zygosaccharomyces rouxii in simple fermentation media with an autoclaved mixture of a single amino acid and reducing sugars (Hayashida et al., 1999). Hexoses favoured the formation of Furaneol, while pentoses, such as ribose, stimulated formation of the ethyl-substituted furanone with a nonlinear correlation. Glutamate was the most efficient source of nitrogen. The authors, well aware of the competing chemical pathway of formation, described the formation of furanones in aging miso and simple fermentation media as a yeast-dependent reaction and suggested that ‘many amino acids form compounds able to act as Furaneol precursors’. The same experiments did not result in the formation of significant amounts of furanones, when the heating of the precursor combination was replaced by a filter sterilisation (0.22 ␮m) protocol. Even higher concentrations of 2- (or 5-) ethyl-5- (or 2-) methyl-4-hydroxy-2H-furan-3-one (74.3 mg/L in the presence of l-alanine) were reported subsequently (Sugawara and Sakurai, 1999). The occurrence of volatile lactones, such as 4,5-dimethyl-3-hydroxy-5H-furan-2-one (sotolon), in higher fungi (Lizarraga-Guerra« et al., 1997) underscores the potential of fungal cells as catalysts in bioprocesses. The most common pathway is ␤-oxidative degradation of hydroxy fatty acids followed by intramolecular esterification to the respective 4- and 5-alkanolides with fruity and fatty odour notes, but a number of different pathways have also been reviewed (Gatfield, 1997). 6-Pentyl-␣-pyrone is a lactone with coconut-like flavour, and its formation by Trichoderma harzianum, Trichoderma viride and other fungi is well documented. Various surface and submerged fermentations (Kalyani et al., 2000) and solid-state fermentation on sugarcane bagasse (Sarhy-Bagnon et al., 2000) have yielded up to almost 1 g of volatile product per litre. The lactone inhibits the growth of some phytopathogenic fungi and may be formed as a part of the chemical self-defence system of the Trichoderma fungi. T. harzianum was also able to convert castor oil into 4-decanolide (Serrano-Carreon« et al., 1997). Using a 14-L bioreactor, rheological studies on the shear-sensitive mycelial cells were performed. A lactone yield of 5.3 g/kg of castor oil was calculated according to the principle of extractive fermentation. Indeed, the age of industrial bioflavours started with a patented process to convert ricinoleic acid (12-hydroxyoleic acid) to 4-decanolide using Candida yeast (Farbood and Willis, 1983). Initially, the market price was US$20 000/kg, which created a lot of enthusiasm among biotechnologists. Process development was facilitated by access to an inexpensive precursor substrate, by the widespread occurrence of the pathway in fungi and by relatively simple P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 97

bioprocessing conditions. An alternative access was opened by the transformation of 3-decen- 4-olide to the reduced lactone using baker’s yeast (Gatfield and Sommer, 1997). Among the more recently described producers were Sporidiobolus yeasts (Haffner and Tressl, 1998; Dufosse et al., 2000). Today, 4-decanolide commands a market price 40-fold lower than initially. A survey of the pathways to 4-decanolide was presented (Krings and Berger, 1998). Less work has been devoted to sulfur-containing heterocycles. A patent (Bel Rhlid et al., 1997a,b) claims a bioprocess on the basis of the baker’s yeast that converted cysteamine or cysteine and hydroxy or oxopropanoic acid (or derivatives) to a flavour mixture containing 2- methyl-3-furanthiol, mercaptopentanone and 2-acetyl-2-thiazoline. The composition is used to intensify meat flavour of food and pet food.

4.4.4 Generation of vanillin, benzaldehyde and benzoic compounds Vanillin is the world’s number one flavour chemical. Vested with a unique flavour characteristic, the compound also exhibits antioxidant, flavour-enhancing and bitterness-masking properties. Three orders of magnitude of difference of the market prices of chemosynthetic (around US$12/kg) and Vanilla pod-derived vanillin (up to US$16 000/kg) have again triggered a lot of research. Vanilla planifolia is an orchid and fixes CO2 according to the rather uncommon CAM (Crassulacean acid metabolism) pathway. Multi-component stable isotope analysis of vanillin will therefore be able to discriminate various ‘natural’ vanillins according to the isotope patterns of the vanillin precursor substrates (Hener et al., 1998). Among the vanillin precursors listed (Rabenhorst, 2000; Xu et al., 2007) are ferulic acid, eugenol and isoeugenol, vanillylamine, methoxytyrosine, coniferyl aldehyde, coumaric acid, and others. More than two dozen microbial strains, mainly soil bacteria and higher fungi, and at least two different enzymes, converted these substrates to vanillin. The most efficient process (Ͼ11 g/L in 30 hours) is based on an actinomycete of the Amycolatopsis family and a batch-fed dosage of ferulic acid (Rabenhorst, 2000). This retro- Claisen-type cleavage also resulted in metabolite overflow, subsequent secretion, and accumulation of amounts in gram of vanillin by Streptomyces setonii (Muheim and Lerch, 1999). A P. putida strain was cultivated by the same authors. Although ferulic acid catabolism was fast, vanillin did not accumulate, as it was oxidised faster than ferulic acid. Accumulation of vanillin also failed in growing cultures of a Nocardia strain (Li and Rosazza, 2000). A purified carboxylic acid reductase from the same strain, however, quantitatively reduced vanillic acid to vanillin. Isoeugenol from various essential oils is another inexpensive precursor substrate. Owing to its considerable cytotoxicity, the development of a high-yielding bioprocess towards vanillin was prevented for a long time. With the aid of enrichment techniques, isoeugenol-tolerant strains have now been selected. Cell-free extracts of a Bacillus strain yielded 0.9 g vanillin per litre (Shimoni et al., 2000), and Rhodococcus rhodochrous tolerated 15 g of isoeugenol per litre, which was converted with a molar yield of about 60% to vanillin (Chatterjee et al., 1999b). The phenylpropanoid pool provides precursors for benzoic volatiles in flavour extracts from both cell cultures (Venkateshwarlu et al., 2000) and fruiting bodies (Rosecke¬ and Konig,¬ 2000). Benzaldehyde, mandelates, phenylacetates and their derivatives occur regularly. Mul- tiple pathways of the degradation of phenylpropanoid compounds appear to exist. Although less likely to occur than with terpenes, mere chemical conversion cannot be ruled out: a cell-free extract from Lactobacillus plantarum deaminated phenylalanine to phenylpyruvic P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

98 Food Flavour Technology

acid, which was then chemically oxidised to benzaldehyde and other compounds at pH 8 if certain metal cations were present (Nierop Groot and de Bont, 1998). A P. fluorescens strain, able to grow on ferulic acid as the sole carbon source, contained high levels of an inducible feruloyl-CoA ligase and a vanillin dehydrogenase (Narbad and Gasson, 1998). Acetyl-CoA was verified as a cleavage product by 13C NMR, but the mechanism of the cleavage steps remained open. The mandelate pathway was demonstrated in the related P. putida by feeding benzoylformate (Simmonds and Robinson, 1998), and also in the higher fungus Gloeophyl- lum odoratum by analysing volatile and non-volatile (trimethylsilylated) constituents of a hot-water extract from the fruiting bodies (Rosecke¬ and Konig,¬ 2000). Flavour chemists and natural products chemists often stick to their selective non-polar and medium-polar extraction solvents, respectively, and are thus prone to overlook the respective non-soluble intermediates of a given pathway, which might impede biogenetic understanding. Originating from the Strecker degradation of phenylalanine, the rose-like compound 2- phenylethanol is a common volatile in yeast fermentation flavours (Fabre et al., 1998). Metabolites of labelled l-phenylalanine in the higher fungus Bjerkandera adusta were the non-volatile (E)-cinnamic, phenylpyruvic, phenylacetic, mandelic and benzoylformic acids, and volatiles such as benzaldehyde and benzyl alcohol (Lapadatescu et al., 2000). The direct ␤-oxidation of cinnamic acid to benzoic acid was concluded from the occurrence of acetophenone, the product of the spontaneous decarboxylation of the supposed intermediate ␤-oxophenylpropanoic acid. Pycnoporus cinnabarinus, another white-rot fungus, was the subject of optimisation studies resulting in a high-density culture yielding >1.5 g of vanillin per litre (Oddou et al., 1999; Stentelaire et al., 2000). Vanillin formation was favoured by reduced concentration of dissolved oxygen, high carbon dioxide, gentle agitation, low specific growth rate and the application of a non-selective adsorbent. On the basis of this well-explored model, 5-2H-labelled ferulic acid was fed to the fungus (Krings et al., 2001; Fig. 4.3). The major labelled phenolic compounds identified were four lignans: the methyl esters of ferulic and vanillic acid, (E)-coniferyl aldehyde and alcohol, vanillic acid, vanillin, and vanillyl alcohol. Labelled 4-hydroxy-3-methoxyacetophenone occurred, suggesting the decarboxylation of free 4-hydroxy-3-methoxybenzoylacetic acid. Detailed mass spectrometric examination revealed traces of 4-hydroxy-3-methoxybenzoylacetic acid methyl ester and 3-hydroxy-(4-hydroxy-3-methoxyphenyl) propanoic acid methyl ester in the culture medium. Hence, the fungal degradation of the phenylpropenoic side chain, a principal key step of lignin decomposition, should proceed by analogy to the oxidation of fatty acids. Raspberry ketone, 4-(4-hydroxyphenyl)butan-2-one, is one of the character-impact components of raspberry flavour. If the compound were isolated from raspberry fruit, natural raspberry ketone would cost 6 million dollars per kg (cost of fruit only), compared to $US8Ð10/kg for the synthetic product. An enzymatic pathway was proposed involving the ␤-glucosidase-catalysed hydrolysis of the naturally occurring betuloside from the European white birch (Betula alba) to release betuligenol, which is transformed by an Acetobacter alcohol dehydrogenase (ADH) into the ketone. Plant cell culture or biomimetic acyl anion transfer reactions using aldolases were also suggested, but none of the known routes appear to have met industrial processing requirements (Whitehead, 1998). The basidiomycete Nidula niveo-tomentosa synthesised traces of the ketone and its corresponding alcohol de novo starting from simple nutrients, such as glucose and amino acids. A systematic attempt was made to improve the productivity of this fungus in submerged culture (Boker¬ et al., 2001). Variation of the composition of the nutrient medium supported by a factorial experimental design yielded a 50-fold increase in metabolite concentrations. This allowed for a P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 99

Fig. 4.3 ␤-Oxidation-like degradation of ferulic acid by Pycnoporus cinnabarinus.

follow-up labelling study using phenylpropanoid precursor substrates, and it turned out that l-phenylalanine was degraded to a benzylic intermediate and then side chain elongated in two subsequent steps (Zorn et al., 2003b). Exposure to UV light stimulated the growth of the basidiomycete and the synthesis of raspberry alcohol and ketone. As this is one of the very rare examples of a UV light-induced formation of a fungal volatile, differentially expressed proteins were identiﬁed by means of 2D-electrophoresis and ab initio sequenced by ESI- MS/MS spectrometry. The encoding nucleotide sequences were cloned. Several stress and growth-related enzymes were up-regulated as a response to irradiation, but clear evidence for the involvement of polyketide synthases is still missing (Taupp et al., 2008).

4.4.5 Generation of miscellaneous compounds Short-chain methyl-branched fatty acids are not only important to cheese and other fermentation flavours, but are also added, for example, to strawberry flavours. Their generation from fusel oil constituents by strongly oxidising bacteria is feasible. Gluconobacter species preferentially produced (S)-2-methylbutanoic acid from 2-methylbutanols of known enantiomeric composition (Schumacher et al., 1998). Mechanistic aspects were studied in detail (Gatfield et al., 2000). Esterification of such fatty acids with ethanol using Geotrichum P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

100 Food Flavour Technology

yeast removes the necessity to substitute inactive lipase during reverse-hydrolytic enzyme- based processes (Daigle et al., 1999). Work on the production of 1-octen-3-ol, one of the mushroom flavour-impact compounds generated through the lipoxygenase/hydroperoxide lyase pathway, is continuing (Assaf et al., 1997). Enantioselective reduction of 2-octanone by Lactobacillus fermentum and other organisms produced (S)-2-octanol with high enantiomeric excess (Molinari et al., 1997). An extended screening of food-grade yeast species showed that some of them accumulated elevated concentrations of ␣-hydroxy ketones (acyloins). Zygosaccharomyces bisporus cells, for example, accepted a wide variety of aldehyde substrates and linked them to pyruvate to form ␣-hydroxy ketones (Neuser et al., 2000a). The enantiopurity of the products depended strongly on the substrate structure. Odour qualities and threshold values of 34 acyloins were evaluated using a GCÐolfactometry dilution technique, and 23 of them possessed novel and pronounced flavour properties (Neuser et al., 2000b). This catalytic property is known to rely on a pyruvate decarboxylase, and the respective enzyme was isolated from Z. bisporus and compared with crude preparations from S. cerevisiae, K. lactis and Kluyveromyces marxianus. Conversion rates of more than 50% showed that the potential of this type of enzyme to catalyse the formation of aliphatic acyloins has been underestimated. An 1856 bp cDNA coding for the monomeric unit of the enzyme was amplified and sequenced (Neuser et al., 2000c). Aliphatics with sulfur functions often possess low-odour detection threshold values. Of interest for cheese flavour, for example, is methanethiol. Species commonly present in ageing cheese, such as Lactococcus, Lactobacillus, and Brevibacterium, degraded l-methionine through methionine aminotransferase or ␥-lyase activities to methanethiol (Dias and Weimer, 1998). The enzyme activities were quantitatively assessed, but no attempts at further purification were made. Particularly active formers of sulfur compounds are strains of Geotrichum candidum, which develop early in cheese ripening (Berger et al., 1999a; Demarigny et al., 2000). Methanethiol, as well as di- and trisulfides, and several thio fatty acids with straight and branched chains were identified. From results obtained with l-[(S)-methyl-2H]methionine, two different pathways for the generation of the sulfur compounds were suggested. An extended range of microbial strains from surface-ripened cheese is currently under study (Deetae et al., 2007). Whether an odorous molecule is perceived by humans as a positive or negative (off- flavour) sensation depends on the actual concentration, the composition of the matrix and the presence of other complementing volatiles. Hence, off-flavours attract the same industrial attention as pleasant volatile flavours. How the production of volatiles by yeast can be affected by bioprocess conditions has become evident during the development of continuous brewing processes (Van Iersel et al., 2000). Immobilised yeast showed slower growth and decreased amino acid metabolism, and, thus, released increased concentrations of oxo acids into the fermentation medium. High levels of undesirable aldehydes may occur due to the activity of pyruvate decarboxylase, a problem present in the dealcoholisation of beer. As these oxo acids are the precursors of the above-mentioned acyloins, this problem may be turned to advantage if the generation of acyloins is aimed at. Strains of Penicillium, Trichoderma, Aspergillus, Mucor, Monilia and one of Streptomycetes were isolated from cork (Caldentey et al., 1998). Growth on cork resulted in the formation of off-flavours, but growth on malt extract medium did not. Many aliphatic alcohols and carbonyls, sesquiterpenes and some halogenated aromatics were identified, but no structural assignments of the off-notes were reported. The phenolic off-flavour of fermented soy products is now clearly attributable to 4-ethyl- and 4-vinylguaiacol (Suezawa et al., 1998; Karmakar et al., 2000). Bacteria, P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 101

mainly strains of Bacillus, Pseudomonas and Staphylococcus that degrade phenylpropanoid precursors, are responsible for this problem.

4.5 ENZYME TECHNOLOGY

If a single chemical step is to be biocatalysed, it appears to be a waste to maintain all the thousands of concurrent metabolic steps of a living cell that are not wanted. A single enzyme, whether in free form or included in gel or microcapsules or bound onto solid supports, could perform the desired reaction just as well. Some problems associated with the use of enzymes may include location of the activity needed, isolation and puriﬁcation, maintenance of activity during preparation and use, and, for kinases, methyl transferases and oxidoreductases, inevitable co-factor requirements. A review on enzymes in ﬂavour biotechnology by Menzel and Schreier (2007) shows how some of these disadvantages can be overcome. Most enzymes currently used in the food industry stem from only about 25 micro-organisms. Many possible sources of enzymes, for example marine organisms (Chandrasekaran, 1997), have remained almost unexplored so far. With the rapid progress in basic enzymological and genetic knowledge even NAD+-dependent redox enzymes gain interest, particularly if a synthetic electron sink, such as dichlorophenol indophenol, is accepted and high stereoselectivity is observed (Van der Werf et al., 1999).

4.5.1 Liberation of volatiles from bound precursors The application of glycosidases and, less frequently, of lyases, for the liberation of preformed flavours from non-volatile precursors (Winterhalter and Skouroumounis, 1997) should not be confused with the maceration of recalcitrant plant materials, when enzymes serve merely as processing aids (for example, Sakho et al., 1998). The majority of reports deal with wine flavour (Table 4.1). ␤-d-Glucopyranosidase, ␣-l-arabinofuranosidase, ␤-apiosidase, ␣-l- rhamnosidase and a cysteine conjugate ␤-lyase were shown to enhance the level of volatiles in musts, wines and fruit juices. While strains of Aspergillus and yeasts are typical sources of these enzymes, the lyase was from Eubacterium limosum (Tominaga et al., 1998). Flavour enhancement in common fruits and in vanilla bean (Pu et al., 1998) was often performed using fungal ␤-glucosidases. Tropical fruits and black tea leaves, however, also contain less abundant glycosides, such as primeverosides, vicianosides and rutinosides, which may not be sufficiently hydrolysed by side activities of the standard glycosidases. The long history of enzyme-modified cheese is discussed in reviews (Kilcawley et al., 1998; Klein and Lortal, 1999). The proteolysis of minced fish tissue to produce a ‘seafood flavour’ (Imm and Lee, 1999) was likewise adopted from traditional food biotechnology. The reverse reaction may also be useful: lyophilised cells of Xanthomonas campestris and of Stenotrophomonas maltophilia synthesised l-menthyl ␣-d-glucopyranoside anomer-selectively from l-menthol (Nakagawa et al., 2000). An impressive molar yield of more than 99% in 48 hours was reported. The glucoside obtained is slowly hydrolysed in the oral cavity, thereby acting as a flavour depot.

4.5.2 Biotransformations The interfacial enzymology of lipid hydrolases is useful in ﬂavour generation, and there are, as with all carbonyl reactions, two options for shifting the reaction equilibrium: ﬁrst, towards P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

102 Food Flavour Technology

Table 4.1 Hydrolases of oenological relevance.

Flavour precursor Enzymes References Glycoconjugated aroma compounds Endo-, exogenous glycosidases Winterhalter and Skouroumounis (1997) Mono- and diglycosides of terpenols ␣-L-Arabinofuranosidase, Spagna et al. (1998) (linalool, ␣-terpineol, citronellol, nerol, ␤-D-glucopyranosidase geraniol) Monoglycosides of terpenols ␤-D-Glucosidase Yanai and Sato (1999) Apiofuranosylglucosides of geraniol ␤-Apiosidase Guo et al. (1999) and linalool ␣-L-Rhamnopyranoside ␣-L-Rhamnosidase Orejas et al. (1999) Glycosides of nerol, geraniol, linalool, ␤-Glucosidase Gueguen et al. (1997) ␥-terpinene, 2-phenylethanol, benzyl alcohol S-Cysteine conjugates of volatile thiols Cysteine conjugate ␤-lyase Tominaga et al. (1998) (4-mercapto-4-methylpentane-2-one, 4-mercapto-4-methylpentan-2-ol, 3-mercaptohexan-1-ol) Glycosides of fruits, particularly Vitis Exogenous and endogenous Sarry and Gunata (2004) vinifera glycosidases Terpenyl glycosides Bacterial, yeast, plant Maicas and Mateo glycosidases (2005) Glycosides of juices Glycosidases Pogorzelski and Wilkowska (2007)

the liberation of odorous fatty acids from triacylglycerols and related esters, and second, towards the synthesis of a variety of volatile esters from different acyl and alkyl precursor moieties in a microaqueous environment. A broad range of reaction conditions including organic media is tolerated and can be varied to achieve the desired selectivities (Saxena et al., 1999). The lipase from Candida antarctica, which is a stable, versatile enzyme with broad substrate specificity, has been characterised in detail and become a classical catalyst. Some of the more than 170 lipases have been analysed by X-ray crystallography down to the 1.5 Å level of resolution; thus, the position of the helical lid that buries the active triad and conformational changes exerted on this by bipolar chemicals have been well established (http://www.rcsb.org/pdb/). The acetates of common acyclic monoterpenols are used as flavours and fragrances. These volatiles can be produced either by direct esterification or by transesterification using the C. antarctica lipase SP435 in microaqueous n-hexane as reaction solvent (Claon and Akoh, 1994a,b). Whole cells of Hansenula saturnus or Pichia species in an interface reactor and n-decane (Oda et al., 1995; Oda and Ohta, 1997) and dry mycelium of Rhizopus delemar in n-heptane (Molinari et al., 1998a,b) were successfully used for the same purpose. The reverse hydrolytic process runs on an industrial scale and constitutes one of the fundamentals of modern flavour biotechnology. Lipases of C. antarctica and of Rhizomucor miehei formed esters in high yields under solvolytic conditions and vacuum, where the substrate itself acted as a solvent (Chatterjee and Bhattacharyya, 1998a,b). The reaction solvent also affects the enantioselectivity. The synthesis of (−)-citronellyl oleate from racemic citronellol was achieved using a Candida cylindracea lipase in supercritical carbon dioxide, but only in P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 103

a narrow window of pressure and temperature (Ikushima et al., 1996). Some contradictory results suggest, however, that a generalisation of these findings should be treated with caution (Michor et al., 1996). The synthesis of (3Z)-hexenyl butanoate in n-hexane or in solvent-free medium was likewise achieved using the Mucor or Candida lipases (Bourg-Garros et al., 1997; Kim- Jungbae et al., 1998). Aliphatic alcohols from fusel oil, a side-product of the distillation of spirits, were reacted with dodecanoic acid (De Castro et al., 1999). Yields dropped with decreasing chain length of the aliphatic acyl moiety. Recent work also modified the structure of the alkyl moiety. Phenylethanol was transformed to its acetate, a Koji-style wine flavour impact, using supercritical carbon dioxide and lipase PS (Wen et al., 1999), and thioethyl, thiobutyl and thiohexyl propanoate, butanoate and valerate were produced using different immobilised lipases (Cavaille-Lefebvre and Combes, 1997). An example of a successful redox process was the oxidation of vanillylamine to vanillin by an amine oxidase from A. niger, or by a monoamine oxidase from E. coli (Yoshida et al., 1997).

4.5.3 Kinetic resolution of racemates The preferred cleavage of one enantiomer of a racemic (usually ester) mixture is one possible route to enantiopure products, an approach patented for the generation of the bulk flavour chemical l-menthol in the early 1970s. Many subsequent papers and patents confirm the competitiveness and industrial usefulness of this mature bioprocess. This classical approach was followed to resolve (R)- and (S)-karahanaenol produced from ring enlargement of limonene epoxide (Roy, 1999). The racemic monoterpenol acetate was resolved using a Pseudomonas cepacia lipase, and the 4R-isomer alcohol was preferentially obtained through alcoholysis (Fig. 4.4). Kinetic resolution in the course of the esterification of menthol was a novel aspect arising from work with lipase from Candida rugosa, originally focusing on the formation of methyl esters as a flavour depot (Shimada et al., 1999). When a menthol racemate was esterified with oleate in an emulsion containing 30% water, 96% esterification and an enantiomeric excess of 88% of the l-enantiomer menthyl ester were found. The operational stability of some lipases is remarkable. In a study undertaken to separate racemates of ibuprofen and 1-phenylethanol, the commercial enzyme remained partly active even after 14 hours at 140◦C and 15 MPa (Overmeyer et al., 1999). These examples together with Table 4.2 show the broad range of substrates amenable to hydrolysis with various lipases. The theoretical modelling of the energy differences between the transition states of the diastere- omeric enzymeÐsubstrate complexes and of substrate inhibition will aid in further refinement of kinetic resolution processes (Berendsen et al., 2006). There remains the drawback that the

Fig. 4.4 Enantiospeciﬁc alcoholysis of karahanaenol acetate by Pseudomonas cepacia lipase: (I) terpinolene oxide; (II) (R)-karahanaenol; (III) (S)-karahanaenol acetate (adapted from Roy, 1999). P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

104 Food Flavour Technology

Table 4.2 Kinetic resolution of racemates: a mature technique in ﬂavour enzymology.

Flavour compounds Enzymes References Lactones and epoxides Reductase Danchet et al. (1998) 1-Phenylethanol Immobilised lipase (Mucor miehei)Fringset al. (1999) 1-Phenylethanol Lipase (Pseudomonas sp.) Ceynowa and Koter (1997) 2-Phenyl-1-propanol Lipases Goto et al. (2000) Ibuprofen, 1-phenylethanol Commercial lipase (Novozym R ) Overmeyer et al. (1999) ␥-, ␦-Lactones Lipase (Pseudomonas sp.) Enzelberger et al. (1997) 2-Methylbutanoic acid methyl ester Lipase (R. miehei) Kwon et al. (2000)

resolved substrates are usually derived from chemosynthesis; hence, the produced flavours are not naturals in a legal sense. Tailor-made enzymes are now at hand. With the invention of directed evolution in combination with high-throughput screening systems, literally any enzyme can be modified according to changes of substrate or reaction specificity. Micro-organisms that cannot be cultured in vitro deliver their enzymes through gene partial sequences found in soils, deep- sea sources or geysers. Progress in sequence-based biocatalyst discovery allows to explore the bewildering catalytic diversity of nature. Techniques for the stabilisation and immobilisation of enzymes have reached a high degree of perfection. Further impetus comes from the spreading idea of white biotechnology, an industrial approach focusing on renewable substrates and environmentally friendly production processes.

4.6 PLANT CATALYSTS

Most of the natural flavours currently processed by the flavour and fragrance industry are obtained by extraction or distillation of parts of field-grown plants. Their huge biosynthetic potential represents an attractive starting point for the in vitro culture of plant cells for aroma biotechnology (Fu, 1999). At the same time, the high degree of structural and biochemical organisation of a plant cell impedes the biotechnological approach: specialised plant cells tend to dedifferentiate in vitro, they demand complex nutrient media and low-shear bioreactors, and usually do not accumulate and excrete elevated levels of volatile flavours. The enzymes of a phototrophic metabolism are not easily inducible by carbon substrates. This explains why plant cells in sterile culture are not as responsive to precursor substrates as are most micro-organisms. Numerous attempts at the production of flavour and aroma compounds by plant cell, tissue and organ cultures have been described (Berger, 1995; Dorneburg¬ and Knorr, 1996; Kumar et al., 1998). A collection of promising recent results obtained on the laboratory scale is presented in the following sections.

4.6.1 Plant cell, tissue and organ cultures Plant tissues, excised from differentiated and surface-sterilised materials and placed on phytoeffector-containing agar medium, start to divide again and form non-differentiated cell clusters (‘callus’). After transfer to liquid medium, a ﬁne suspension of single cells and smaller cell aggregates develops. This propagation of plant cells is based entirely on mitotic events; the full genetic potential to form ﬂavours is retained in each cultured cell. However, P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 105

these cells dedifferentiate during subculturing and do not accumulate ﬂavours, even if isolated from a ﬂavour-bearing source tissue.

4.6.2 Callus and suspension cultures Exceptionally high amounts of odorous mono- and sesquiterpenes (0.34% total oil content in cells plus medium) were recovered from callus cultures of the Brazilian snapdragon (Otacanthus coeruleus), an ornamental pot plant from east Brazil (Ronse et al., 1998). The amount of essential oil extracted from the nutrient media was higher than the intracellular amount. High sucrose treatments (Ͼ40 g/L) increased the quantities of oil found in the medium due to the high osmotic stress. In cell cultures of two genotypes of rosemary, the concentrations of calcium ions, sucrose and plant growth regulators significantly affected the yields of camphene, 1,8-cineole, linalool, camphor, borneol and bornyl acetate (Tawfik et al., 1998). Photomixotrophic callus cells of grapefruit (Citrus paradisi Macf.), lemon (Citrus limon (L.) Burm.) and lime (Citrus aurantiifolia (Christm. et Panz.) Swingle) generated monoterpenes, although no cytodifferentiation was found by electron microscopy (Reil and Berger, 1996a). Chlorophyll content and the formation of oligo-isoprenoid volatiles were positively correlated with high light intensities. An optimisation of the growth medium and the light regime resulted in the identification of more than 40 mono- and sesquiterpenes and aliphatic aldehydes in grapefruit callus. The maximum yield was 186 mg of volatiles per kg callus accumulated in 4 weeks, representing about 5% of the volatiles accumulated in flavedo tissue in about 13 months. A similar correlation existed for white diosma (Coleonema album Thunb.) photomixotrophic cell cultures (Reil and Berger, 1997). An extended photoperiod and certain concentrations of phytoeffectors were the conditions for the formation of volatiles including limonene and phellandrenes. The light conditions also affected the production of compounds by vanilla (V. planifolia) cultures, particularly of 4-hydroxy-3-methoxybenzyl alcohol (vanillyl alcohol) (Havkin-Frenkel et al., 1996). Microbial infections, whether caused by an endogenous infection of the source tissue or by a secondary infection, present a serious experimental problem because most micro-organisms simply overgrow the slower plant cells. Otherwise, microbial infections are often accompanied by the formation of elicitors, components involved in plant chemical communication and defence. As volatile flavours may be part of a plant’s response to a microbial attack, flavour biogenesis can be elicited. A persistent contamination with Pseudomonas mallei was reported for cell cultures of rosemary (Rosmarinus officinalis L.), producing cineole and ␣-pinene (Shervington et al., 1997). An (unintended) elicitation may explain this observation. A combination of photomixotrophy and elicitation led to the excretion of volatiles by cell cultures of parsley (Petroselinum crispum (Mill.) Nym.). Treatment with autoclaved homogenate of the wood-destroying basidiomycetes Polyporus umbellatus or Tyromyces sambuceus elicited a spicy odour, imparted by elemicin (5-allyl-1,2,3-trimethoxybenzene), 3-n-butylphthalide, (Z)- and (E)-butylidenephthalide, sedanenolide and (Z)-ligustilide (Reil and Berger, 1996b) (Fig. 4.5).

4.6.3 Organ cultures There is abundant experimental evidence for the correlation of cytodifferentiation and the formation of secondary plant products. Visible differential gene activity and biochemical P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

106 Food Flavour Technology

n Z Z

Fig. 4.5 Flavours elicited in cell cultures of parsley.

specialisation occur with morphological differentiation. Non-embryogenic cell lines derived from immature juice vesicles of sweet orange (Citrus sinensis (L.) Osbeck) failed to form the characteristic flavour constituents, but organised embryogenic cells emitted a fruity aroma and produced 420 mg of volatiles per kg tissue (Niedz et al., 1997). Some hairy root cultures, derived from the transformation of aseptic plantlets with Agrobacterium rhizogenes,were shown to be capable of forming volatile flavours. The essential oils of hairy root cultures of anise (Pimpinella anisum L.) differed significantly from those of the fruits (Santos et al., 1998; Andarwulan and Shetty, 1999). While the major components of the essential oil from the hairy root cultures were the anethole precursor (E)-epoxypseudoisoeugenyl 2-methylbutanoate, zingiberene, ␤-bisabolene, geijerene and pregeijerene, the terpene spectrum of the fruits was dominated by (E)-anethole (Fig. 4.6). The maintenance of morphological stability with no dedifferentiation or greening remains an experimental challenge (Santos et al., 1999). Matsuda et al. (2000) generated mutant hairy roots of musk melon (Cucumis melo L.) by means of t-DNA insertion. Of more than 6500 clones, 5 fragrant hairy root clones were selected. The volatile compounds were identified as (Z)-3-hexenol, (E)-2-hexenal, 1-nonanol and (Z)-6-nonenol, which also determine the flavour of the fruits. Aroma production was stable for more than 3 years. The yield of aroma compounds was about 6.5-fold higher than in ripe melon fruit.

Fig. 4.6 Secondary metabolites produced by hairy root cultures of anise. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 107

4.6.4 Plant cell biotransformations Biotransformations of monoterpenoid alcohols, aldehydes, ketones and oxides by plant and microbial cell cultures have been reviewed by Shin (1995), and the conversion of monoterpenes, steroids and indole alkaloids using cell cultures was summarised by Hamada and Furuya (1999). Particular attention was dedicated to the regio- and stereospecifity of the reactions and to immobilisation techniques. Immobilised and free cells of yams (Dioscorea deltoidea) and of kangaroo apple (Solanum aviculare) oxidised (−)-limonene to (Z)- and (E)-carveol to carvone (Section 4.4.1.1). The preferred formation of either carvone or (Z)- and (E)-carveol depended on the immobilisation medium (Vanek et al., 1999). The same allylic oxidation of a terpene hydrocarbon was performed by grapefruit suspension cells (Reil and Berger, 1996a): exogenous valencene was converted to nootkatone via the 2-hydroxy derivative. The synthesis of menthol by peppermint (Mentha piperita L.) cell suspension cultures has been investigated thoroughly (Park et al., 1997). Distinct hydroxylation activity towards terpenes was found. Application of (−)-(4R)- and (+)-(4S)-isopiperitones, for example, yielded the corresponding 7-hydroxyisopiperitones followed by conversion to the respective glucopyranosides. The common pattern of slow flavour synthesis was observed with various Allium tissue cultures (onion, garlic and chive) (Mellouki et al., 1996). On the basis of the good knowledge of the biosynthetic pathway in Allium species, the low yields were attributed to the low concentrations of the flavour precursors, the (+)-S-alk(en)yl-l-cysteine sulfoxides rather than to a lack of C-S lyase activity (Prince et al., 1997). Addition of cysteine, glutathione or methionine increased the yields of methyl- and propenylcysteine sulfoxides. Chilli pepper (Capsicum frutescens) suspension cells accumulated vanillin, vanillic acid and ferulic acid after supplementation with isoeugenol (Ramachandra Rao and Ravishankar, 1999). The biotransformation rate was improved by the simultaneous addition of ␣-cyclodextrin and conversion substrate, by immobilising cells with sodium alginate, and by application of fungal elicitors. Attempts to further optimise the yields were reported (Ramachandra Rao and Ravishankar, 2000), but it will be difficult to compete with the very advanced microbial processes described above. Another transformation of commercial interest is the demethylation of methyl N-methylanthranilate by peroxidases from various plant sources to produce the Concord grape and citrus flavour-impact methyl anthranilate (Van Haandel et al., 2000). The tremendous recent rise of the price of crude oil has revived the interest in plant cells as factories for fine chemicals. There is, however, a big gap between our knowledge of the structure and properties of volatile flavours and our understanding of the pathways of formation and their regulation (Schwab et al., 2008). Only recently, work is being focused on enzymes and genes involved in their biosynthesis. For example, the genome of rice (Oryza sativa L.), not famous as an essential oil plant, is supposed to contain about 50 genes encoding putative terpene synthases (Cheng et al., 2007). Modification of flavour of plants by genetic engineering bears an entire set of opportunities, but depends on the knowledge of genes encoding for the relevant reactions.

4.7 FLAVOURS THROUGH GENETIC ENGINEERING

Genetic engineering provides the tools to turn biochemical knowledge into organisms, which, from an anthropocentric point of view, are regarded as improved. The tools comprise simple gene deletions or ampliﬁcations, DNA rearrangements in a species, trans-species gene P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

108 Food Flavour Technology

transfer and bioanalytical screening and monitoring techniques. The food industry is prepared for the new possibilities, as soon as the public concern has been overcome (Pridmore et al., 2000). Another obstacle was presented in recent years by the uncertain legal situation. This has been remedied in Europe by the guideline No. 50/2000 of the European Commis- sion, which deals with the labelling of food and food additives containing ingredients and flavours from genetically modified organisms (GMOs). The guideline, in Article 3, demands the labelling of GMO-derived flavours if a measurable content of proteins or DNA is present as the result of genetic engineering. The International Organisation of the Flavour Industry set pragmatic standards some years ago to ensure the safety of the consumer. A breakthrough of novel and cheaper flavours is predicted, once the first successful products are available and accepted in the marketplace (Muheim et al., 1998).

4.7.1 Genetically modified micro-organisms A lot of the recent work on genetically modified micro-organisms has focused on phenylpropanoic acid metabolism. A strain of P. fluorescens grew on ferulic acid as the sole carbon source and contained a gene coding for an enoyl-SCoA hydratase/lyase enzyme for the conversion of feruloyl-SCoA to vanillin (Barghini et al., 2007) (Fig. 4.7). Heterologous expression of the gene in E. coli proved its function. The results suggest a degradation route different from the ␤-oxidation-like chain-shortening reaction mentioned above for P. cinnabarinus. A transposon mutant of the same species grew on p-coumaric acid as the

Fig. 4.7 Conversion of ferulic acid to vanillin; micro-organisms versus vanilla (modiﬁed after Krings et al., 2001; Barghini et al., 2007). P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 109

sole carbon source and produced vanillin, also in the upper gram per litre range (Civolani et al., 2000). The group of Steinbuchel¬ has put much effort into elaborating the same degradation process and confirmed the involvement of genes of a Pseudomonas sp. by gene disruption and characterisation of the resulting mutants (Overhage et al., 1999a). The inactivation of genes coding for vanillin-degrading enzymes by insertion of omega elements was pursued to achieve accumulation of the intermediate vanillin (Overhage et al., 1999b). On the basis of the evaluation of genetic and enzymatic properties, a similar pathway was identified in the above-cited Amycolatopsis sp. (Achterholt et al., 2000). Patent protection for a series of enzymes and genes was acquired (Steinbuchel¬ et al., 1997). Mutants of Yarrowia lipolytica were created by disrupting one or several acyl-CoA oxidase coding genes (Wache« et al., 2000). The oxidase isoenzymes were believed to be involved in the biotransformation of methyl ricinoleate to 4-decanolide, as they catalyse the initial acyl-CoA dehydrogenation to (2E)-enoyl-CoA acids. A clear correlation of gene disruption and lactone formation was demonstrated. The same methodical approach was applied to an S. cerevisiae strain to accumulate ethyl hexanoate in sake (Asano et al., 2000). A fatty acid activation gene, coding for a synthase acting on exogenous fatty acids, was disrupted. As a consequence of the derepression of the endogenous pathway, an increased de novo production of fatty acids and of the desired ester compound was found. In view of the commercial importance of fermented food, the genetics of traditional starter cultures and of yeasts is increasing an interest. For example, the gene encoding for the enzyme that converts l-methionine to methanethiol was cloned from Brevibacterium linens, and a knockout technique showed that this gene was essential for flavour development in cheese (Yvon et al., 2006). Similarly, enzymes of Lactobacillus casei responsible for the formation of volatile sulfur compounds from non-volatile precursors were found by homology studies, and the resulting two recombinant proteins showed the expected lyase activity (Irmler et al., 2008). The transfer of such procaryotic genes into eukaryotic hosts has also become a routine procedure. As a result, the manipulation of Saccharomyces using genes from E. coli has been reported; the formation of sulfur compounds was among the traits engineered (Swiegers et al., 2008).

4.7.2 Isolated enzymes from genetically modified micro-organisms The well-understood enzymology of fermented food is a solid starting point for attempts to improve flavour generation. Smit et al. (2000) dealt with proteolysis and further degradation of amino acids in cheese, and Rijnen et al. (1999) characterised a lactococcal aminotransferase. A heterologous glutamate dehydrogenase was transferred into Lactococcus lactis to evaluate the impact of this enzyme in a cheese model using radiolabelled amino acids (Rijnen et al., 2000). Increased production of ␣-ketoglutarate and a general improvement of flavour development were found. A wine yeast was supplemented with an Aspergillus nidulans endoxylanase gene to increase fruity flavour notes (Ganga et al., 1999). In the reverse of this process, A. nidulans received an endoglucanase gene from Trichoderma longibrachiatum to release flavours from grape macerates (Villanueva et al., 2000). An endogenous S. cerevisiae alcohol acetyltransferase was overexpressed to increase acetate ester formation in wine and distilled beverages (Lilly et al., 2000). Constitutive expression at high levels was observed by Northern blot hybridisation in yeast transformants. While ethyl octanoate and decanoate levels remained unaffected, a profound increase of ethyl acetate, isopentyl acetate and P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

110 Food Flavour Technology

2-phenylethyl acetate was reported. In some cases, substrate transport to the site of enzymatic conversion may be rate limiting in ﬂavour formation. The citrate uptake of L. lactis, for example, determined the production of acetoin and diacetyl (Drider et al., 1998).

4.7.3 Plant rDNA techniques New expectations for high-yielding cell cultures have been raised by the penetration of recombinant DNA (rDNA) techniques into the plant sciences (Van Berge, 1998). The basic understanding of plant enzymes and regulatory mechanisms has benefited from stable-isotope precursor techniques and the concerted modification of enzymes (Mosandl et al., 2000). Leahy and Roderick (1999) and Takeoka (1999) have compiled information on fruits and vegetables of specific principles of flavour genesis from non-volatile precursors. Gene coding for these enzymes are well accessible from cell cultures and can either be used to specifically modify food plants or be transferred into a suitable microbial expression system. Since the introduction of the FlavrSavrTM tomato in 1994, rDNA techniques have developed rapidly, but most efforts were directed towards improved pest or pesticide resistance of field crops. Flavour quality was neglected, although sensory improvement was obviously a primrose path to gain better consumer acceptance of genetically modified food.

4.7.3.1 Flavours from genetically engineered food plants InasingleAgrobacterium-mediated transformation, three genes from daffodil (Narcissus pseudonarcissus) and from Erwinia uredovora were introduced to the genome of rice to express the entire provitamin A pathway (Ye et al., 2000). The resulting ‘golden rice’ forms ␣,␣-carotene from endogenous geranylgeranyl diphosphate. Intended to fight off vitamin A deficiency in Third World countries, the example shows what genetic engineering could do in the flavour area: carotenes are precursors not only of vitamins but also of flavours, such as the C13-norisoprenoid ionones (Winterhalter, 1996; Winterhalter and Skouroumounis, 1997; Section 4.4.2). Changes of the fatty acid profile were monitored after overexpressing a yeast ⌬ -9- desaturase gene in tomato fruits (Lycopersicon esculentum Mill.) (Wang et al., 1996). A concomitant increase of the concentrations of (Z)-3-hexenal and (Z)-3-hexen-1-ol, derived from linolenic acid peroxidation and degradation, was found in aroma extracts from the transgenic fruit. Speirs et al. (1998) transformed tomatoes using constructs containing a tomato alcohol dehydrogenase (ADH) cDNA in a sense orientation relative to the tomato polygalac- turonase promoter to allow fruit-ripening specific expression of the cDNA. The transformed fruits displayed enhanced ADH activities, which affected the redox balance between some of the aldehydes and the corresponding alcohols associated with tomato flavour (Prestage et al., 1999). The levels of lipoxygenase (LOX) mRNA and of lipoxygenase activity of tomatoes were decreased by an antisense construct (Griffiths et al., 1999). This rDNA contained a fruit-specific promoter, a highly conserved 1.2 kb antisense fragment of the cDNA of tomato lipoxygenase a and no terminator. In contrast to expectations, the transgenic fruit did not show a significantly reduced formation of the lipoxygenase-derived volatile flavours. The authors suggested either that the amount of LOX in the fruit tissue was not rate-limiting in flavour formation or that some key isozymes of LOX remained unaffected by the antisense approach. The results show that, in this field of flavour research, the command of powerful P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 111

molecular biological methods has advanced faster than our knowledge of the enzymatic details. The group of Croteau has studied enzymes of the plant terpene biosynthesis for years. Papers report the isolation of oil gland cDNAs from spearmint and peppermint coding for regiospecific limonene hydroxylases (Lupien et al., 1999; Haudenschild et al., 2000). The cDNAs were overexpressed in E. coli and S. cerevisiae to further characterise the enzymes. The cytochrome P450 enzyme of spearmint introduced the hydroxy group in 6- position to yield (−)-trans-carveol, while the enzyme from peppermint led to the 3-allylic hydroxylation product, (−)-trans-isopiperitenol. Both enzymes showed a high degree of sequence homology. Such data constitute valuable models for studying structureÐactivity relationships in cytochrome P450-catalysed reactions. (S)-linalool synthase from Clarkia breweri (Onagraceae), an annual plant native to California, was expressed in appropriate host plants, resulting in enhanced accumulation of fragrant volatiles (Dudareva and Pichersky, 2000). The detailed paper gives a good account of the subcellular sites of formation of volatile monoterpenes, sesquiterpenes and phenylpropanoids. Ripening of climacteric fruits is controlled by ethene (ethylene) produced from methionine by the key enzyme 1-aminocyclopropane-1-carboxylic acid oxidase (ACO). Cantaloupe Charentais melon (C. melo var. cantalupensis, Naud. cv Vedrandais)« was transformed with an ACO antisense gene (Bauchot et al., 2000). The reduction of ethene synthesis led to delayed ripening and suppressed the biosynthesis of flavours. Application of exogenous ethylene restored an aroma profile similar to that of the control fruit without antisense ACO.

4.7.3.2 Genetically engineered plant enzymes Enzymes involved in the biosynthesis of terpenoids in plants, such as acetoacetyl-CoA thiolase (EC 2.3.1.9), isopentenyl-diphosphate isomerase (EC 5.3.3.2), mevalonate kinase (EC 2.7.1.36) and 3-hydroxy-3-methylglutaryl-CoA synthase (EC 4.1.3.5), are well known (Van der Heijden et al., 1998). Synthases catalysing the cyclisation of the intermediate linear isoprenoid diphosphates into mono- and polycyclic hydrocarbons are of particular interest, because these are the metabolic starting points from which a large variety of oxyfunctionalised, commercial derivatives branch off. Both monoterpene and sesquiterpene synthases were cloned and heterologously expressed to elucidate the substrate specificity and mechanistic details of the cyclisation reaction (Landmann et al., 2007). Plants possessing specific cyclase genes can thus be engineered by a back-transfer of these genes into a target plant host (Beale and Phillips, 1999). The formation of volatile aliphatic and aromatic esters in strawberry fruit (Fragaria sp.) depends on specific enzymes, such as aminotransferase, pyruvate decarboxylase, thiolase, ADH or acyl transferase (Schwab et al., 2006). DNA sequences that code for these activities were isolated from strawberry fruits, characterised and cloned. Likewise, acyltransferases and esterases isolated from apple (Malus domestica Borkh.), mango (Magnifera indica L.) and banana (Musa sp.) were described. Microarray cDNA assays support rapid studies of the expression profile of large subsets of genes in given tissues (Lemieux et al., 1998; Schwab et al., 2006). Another approach to isolate ripening-related genes utilised the differential screening of a high-quality cDNA library (Manning, 1998). Twenty-six ripening-related cDNAs were identified from strawberry fruit. Cultured plant cells constitute a convenient, inexhaustible pool of homogeneous nucleic acids. A kinetin-treated cell suspension culture of vanilla (V. planifolia Andr.) served as the P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

112 Food Flavour Technology

mRNA source for constructing a cDNA library to isolate clones coding for 4-coumarate-CoA ligase (4CL) (EC 6.2.1.12) and caffeic acid O-methyltransferase (EC 2.1.1.6), key enzymes of the phenylpropanoid metabolism (Xue and Brodelius, 1998). Down-regulation of 4CL by an antisense technique was supposed to result in a redirection of phenylpropanoid precursors from lignin biosynthesis to vanillin and related compounds. The activity of hydroperoxide (HPO)-lyase is thought to be rate limiting for the degradation of polyunsaturated fatty acids along the lipoxygenase pathway. The gene coding for this enzyme was cloned from banana and heterologously expressed in yeast cells to generate a continuous source of active lyase (Muheim et al., 1997). Another plant HPO- lyase was purified 300-fold from tomatoes (Suurmeijer et al., 2000). The tomato enzyme cleaved only 13-hydroperoxides from linoleic acid and ␣-linolenic acid, whereas HPO-lyase from alfalfa (Medico sativa L.) also accepted 9-hydroperoxides as substrates to form the respective volatile C9-aldehydes (Noordermeer et al., 1999). Classical tools of molecular biology, such as His-tagging of proteins and immobilised metal affinity chromatography of the tagged species, support the search for flavour enzymes (Santiago-Gomez et al., 2007). The short-term aim is to express both active lipoxygenase and HPO-lyase in a recombinant micro-organism. Mint oil or other traditional plant sources of leaf alcohol and other C6 and C9 flavours would then be replaced by microbial sources. Enzymatic release of flavour from non-volatile glycosides, another form of flavour precursors, not only enhances the concentration of perceivable flavour in foods but turns waste materials, such as peelings, skins and stems, into a renewable source of natural flavours. The practical application, however, is limited owing to low activity at neutral pH and strong inhibition by glucose. The construction of chimeric genes and their overexpression in different hosts is the subject of ongoing research (Winterhalter, 1996; Winterhalter and Skouroumou- nis, 1997 (and references therein)).

4.8 ADVANCES IN BIOPROCESSING

Many laboratory-scale bioprocesses for volatile flavours show productivities high enough for study of the conversion of a substrate and analysis of inducible enzymes and volatile products. To operate the same process successfully on an industrial scale requires an appreciable increase of productivity. Mass and energy transfer in the bioreactor need to be improved, inhibitory effects need to be eliminated and product recovery should be integrated. A sufficient level of oxygen, for example, may not be easily achieved by increasing the inlet pressure because substrate and product may be prone to gas stripping. As their metabolic products, most flavour precursors are medium polar to lipophilic, which entails problems of low substrate solubility and high volatility and cytotoxicity. Lipophilic substrates dissolve into the membranes of cells or interact with the active conformation of enzymes. Solvent- tolerant recombinant bacteria hold much promise for alleviating these persistent problems of bioprocessing of lipophilic chemicals (Verhoef et al., 2007).

4.8.1 Process developments in microbial and enzyme systems It is always a good idea to start with a cheaply available renewable precursor substrate. Apple pomace, soybean products, sugar beet pulp, cassava bagasse and all kinds of processing P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 113

wastes of the food and agricultural industries may contain enough precursor substrate to be used as such (Lesage-Meessen et al., 1999; Pandey et al., 2000a,b; Berger, 2007). Many different effects of immobilisation may lead to a net gain of productivity, as in the case of benzaldehyde synthesis by basidiomycetes (Lapadatescu et al., 1997). Organic/aqueous biphasic systems are a bioengineering response to the different solubilities and sensitivities of substrate, product and catalyst. Microemulsions present an intriguing extension of this approach because of the much larger area of mass exchange (Orlich and Schomacker,¬ 1999). It has been clearly demonstrated that some enzymes work well in supercritical fluids (Section 4.5.2) and in ionic liquids. The latter are composed of, for example, an imidazolium cation and a hexafluoro phosphate, but both ions can be changed and tuned to the respective application. Enzymes not only work well in these ‘green solvents’ (Harjani et al., 2007) but sometimes show improved stability and re-usability, although the reasons are not yet fully understood (Feher et al., 2007). Flavour development in blue cheese depends on the action of P. roqueforti metabolism. Fatty acids liberated from triacylglycerols can undergo decarboxylation as a result of an over- flowing ␤-oxidation. Decarboxylation of the intermediate free 2-oxoacids leads to methyl ketones, such as 2-heptanone, 2-nonanone and 2-undecanone. A combination of immobilisation and biphasic reaction supported the conversion of short-chain alkyl esters of hexanoic, octanoic and dodecanoic acid to 2-alkanones by spores of P. roqueforti (Park et al., 2000). The problem of a lacking metabolic trait can be solved by the consecutive action of two biocatalysts. A comprehensive patent describes (mainly prokaryotic) micro-organisms for the degradation of the propenoic side chain of ferulic acid, and a second set of (mainly eukaryotic) micro-organisms for converting vanillic acid into vanillin (Cheetham et al., 1999). The same two-step process was successful using A. niger for the first step and P. cinnabarinus for the second one (Lesage-Meessen et al., 1999). One step further, the principle of co-cultivation of different species depends on an internal regulation of the bioprocess, but has been applied with few problems to cheese and other traditional fermented foods (Martin et al., 1999; Midje et al., 2000). Novel developments in enzyme technology are surface-coated lipases (Huang et al., 1998) and lipases imprinted by pre-incubation with an ester and a surfactant before freezing the conformation by lyophilisation (Gonzalez-Navarro« and Braco, 1998). Enzyme catalysis is also possible in the gas phase, if the enzymes’ hydration shell is maintained, but there is currently no known application for the synthesis of volatile flavours. Hydrolysis, reverse hydrolysis, carboligation and even co-factor-dependent redox reactions in the gas phase were demonstrated (Trivedi et al., 2006; Mikolajek et al., 2007). No specific preference for a certain type of bioreactor for flavour biotechnology can be recognised. Many authors rely on classical stirred tanks or variants thereof, and examples of a direct comparison of the same reaction proceeding in two different reactor systems are quite rare. Solid-state fermentation was applied to bacteria (Besson et al., 1997) as well as to many fungi (Pandey et al., 2000b). Using a fluidised bed reactor, more than 100 g of terpenyl ester per day was produced in a 2-L bioreactor (Laboret and Perraud, 1999). Not much attention was paid to the development of suitable techniques of in situ recovery of flavours. Standard extraction solvents denature or kill most of the biocatalysts; distillation at ambient pressure or under vacuum will produce the same result. A simple and sometimes efficient means is the direct addition of inert resins to the cell suspension; in case of a highly volatile flavour mounting, an adsorbent tube to the waste air outlet of the bioreactor is an efficient variant (Krings and Berger, 2008). Lipophilic, macroporous polystyrenes offer good mechanical stability and a large adsorptive surface of several hundreds of square metre per gram. Zeolites share with polystyrenes the disadvantage of being non-selective (Treffenfeldt P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

114 Food Flavour Technology

et al., 1999), but are chemically perfectly inert and easily reused. First attempts to create more selective polystyrenes have been undertaken (Gehrke et al., 2000). In view of so many parameters that might eventually affect a bioprocess, more emphasis should be placed on factorially designed experiments and statistical evaluation of results. One of the few examples refers to ester formation by K. marxianus (Medeiros et al., 2000), another one to raspberry compound formation by Nidula niveo-tomentosa (Boker¬ et al., 2001).

4.8.2 Process developments of plant catalysts Strategies towards improved productivity are key to any commercialisation of plant cell cultures (Goldstein, 1999). Progress was achieved by the selection of high-yielding cell lines, variation of the chemical and physical environment, supplementation of precursors and elicitors, in situ product removal and immobilisation techniques. Simultaneous application of several different strategies may result in a synergistic response and a pronounced secondary metabolite formation (Berger, 1995; Pedersen et al., 1999; Shuler, 1999; Verpoorte et al., 1999). Current proposals to improve bioreactor designs for suspension cultures and for organised cultures were presented by Scragg (2007). Oxygen transfer rates in large-scale root culture reactors were discussed by Tescione et al. (1999). The immobilisation of cells or enzymes will inevitably decrease oxygen (and nutrient) transfer, but this disadvantage appears to be overcompensated by improved longevity of the catalyst, simplified recovery of biocatalyst and product and sometimes increased productivity (Fu, 1999). Tomato flavour enzymes were harnessed as a crude enzyme preparation in a commercial ultrafiltration unit, operated as a hollow-fibre reactor to produce hexanal from linoleic acid (Cass et al., 2000). The reactor proved to be stable over an operation period of 5 days, indicating that flavour production with immobilised membrane-associated enzymes in a hollow-fibre reactor is a promising technique. It allowed retention of the enzyme system and substrate with concomitant formation and removal of the product, as monitored by solid- phase micro-extraction. Water-soluble crude protein preparations from green pea, soybean or buckwheat were immobilised in calcium alginate gels (Nagaoka and Kayahara, 2000). This simple catalyst produced optically pure 1-(4-substituted phenyl) ethanol by selective oxidation of one enantiomer of a racemic mixture. After three consecutive reactions, no decrease of yield or optical purity was observed.

4.9 CONCLUSION

Based on data of 2002 and assuming an annual growth rate of 4%, the global flavour and fragrance market is estimated to exceed US$20 billion in 2009 (Short, 2002), with a share of 10% or more for biotechnology-derived compounds (Shamel and Udis-Kessler, 2000). In the relatively mature flavour industry, biotechnology is a vital alternative to generate natural compounds (Sime, 2000), and all leading companies in the flavour market are operating industrial-scale bioprocesses. More than 100 flavours from biotechnology are estimated to be on the market, either as constituents of composed in-house flavours or as pure compounds. The continuing scientific progress is also documented by numerous chapters in recent flavour books authored by scientists from both academia and industry (Table 4.3). Only few cost considerations have been published (Krings and Berger, 1998). As a rule of thumb, a concentration of 1 g of flavour per litre should be achieved to render a process P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 115

Table 4.3 Recent ﬂavour books containing chapters on biotechnology.

Years Editors Title/source 1997 K.A.D. Swift Flavours and Fragrances/The Proceedings of the RCS/SCI International Conference on Flavours and Fragrances, Warwick, UK 1997 R.G. Berger Biotechnology of Aroma Compounds/Advances in Biochemical Engineering Biotechnology, 55; T. Scheper (managing editor) 1999 K.A.D. Swift Current Topics in Flavours and Fragrances: Towards a New Millennium of Discovery, Kluwer Academic, Dordrecht 1999 R. Teranishi, E.L. Flavor Chemistry: Thirty Years of Progress/Proceedings of an ACS Wick, I. Hornstein Symposium, Boston, MA 2000 S.R. Risch, C.-T. Ho Flavor Chemistry: Industrial and Academic Research/ACS Symposium Series 756, ACS, Washington, DC 2000 P. Schieberle, Frontiers of Flavour Science/The Proceedings of the Ninth Weurman K.-H. Engel Flavour Research Symposium, Freising, Germany 2004 T. Hofmann, State-of-the-Art in Flavour Chemistry and Biology, Proceedings of the M. Rothe, 7th Wartburg Symposium on Flavour Chemistry and Biology, P. Schieberle Eisenach, DFA, Garching, Germany 2007 A. Taylor, J. Hort Modifying Flavour in Food, Woodhead, Cambridge, UK 2007 H. Ziegler Flavourings, 2nd ed., Wiley-VCh, Weinheim, Germany 2007 T. Hofmann, Recent Highlights in Flavor Chemistry and Biology, Proceedings of W. Meyerhof, the 8th Wartburg Symposium on Flavor Chemistry and Biology, P. Schieberle Eisenach, DFA, Garching, Germany 2007 R.G. Berger Flavours and Fragrances – Chemistry, Bioprocessing and Sustainability, Springer, Berlin, Germany 2008 D. Havkin-Frenkel, Biotechnology in Flavor Production, Wiley, Chichester, UK F.C. Belanger

economically interesting. Evidently, more progress in the ﬁeld is required before biocatalyti- cal systems will seriously compete on a broad industrial scale with the conventional sources.

REFERENCES

Abraham, W.R., Arfmann, H.-A., Stumpf, B., Washausen, P. and Kieslich, K. (1988) Microbial transformations of some terpenoids and natural compounds. In: Bioﬂavour ’87 (ed. P. Schreier), Walter de Gruyter, Berlin, pp. 399Ð414. Achterholt, S., Priefert, H. and Steinbuchel,¬ A. (2000) Identiﬁcation of Amycolatopsis sp. strain HR 167 genes, involved in the bioconversion of ferulic acid to vanillin. Appl. Microbiol. Biotechnol. 54, 799Ð 807. Agrawal, R., Deepika, N.U.A. and Joseph, R. (1999) Strain improvement of Aspergillus sp. and Penicillium sp. by induced mutation for biotransformation of ␣-pinene to verbenol. Biotech. Bioeng. 63, 249Ð252. Agrawal, R. and Joseph, R. (2000) Bioconversion of ␣-pinene to verbenone by resting cells of Aspergillus niger. Appl. Microbiol. Biotechnol. 53, 335Ð337. Aleu, J., Galan, J.R., Galan, R.H. and Collado, I.G. (1999) Biotransformation of the fungistatic sesquiterpenoid patchuol by Botrytis cinerea. J. Nat. Prod. 62, 437Ð440. Andarwulan, N. and Shetty, K. (1999) Phenolic content in differentiated tissue cultures of untransformed and Agrobacterium-transformed roots of anise (Pimpinella anisum L.). J. Agric. Food Chem. 47, 1776Ð1780. Asano, T., Kawadu, M., Kurose, N., Tarumi, S. and Kawakita, S. (2000) Effect of the FAA1 gene disruption of sake yeast on the accumulation of ethyl caproate in sake mash. J. Biosci. Bioeng. 89, 609Ð611. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

116 Food Flavour Technology

Assaf, S., Hadar, Y. and Dosoretz, C.G. (1997) 1-Octen-3-ol and 13-hydroperoxylinoleate are products of distinct pathways in the oxidative breakdown of linoleic acid by Pleurotus pulmonarius. Enzyme Microbiol. Technol. 21, 484Ð490. Barghini, P., Di Gioia, D., Fava, F. and Ruzzi, M. (2007) Vanillin production using metabolically engineered Escherichia coli under non-growing conditions. Microb. Cell Fact. 6, 13; doi: 10.1186/1475-2859- 6-14. Bauchot, A.D., Mottram, D.S. and John, P. (2000) Aroma formation in Cantaloupe Charentais melon. In: Frontiers of Flavour Science (eds P. Schieberle and K.-H. Engel), Deutsche Forschungsanstalt fur¬ Lebensmittelchemie, Garching, Germany, pp. 463Ð468. Bauer, K. (2000) Tropical fruit flavors: a flavorist’s perspective. Cereal Foods World 45, 204Ð207. Beale, M.H. and Phillips, A.L. (1999) The terpenoid pathway Ð closing the loop. Biochem. Soc. Trans. 27, A17. Bel Rhlid, R., Blank, I. and Gunilla, C.G. (1997a) Process for the preparation of flavouring compositions and use of these compositions in bakery products. European Patent 0951841. Bel Rhlid, R., Blank, I. and Cerny, Ch. (1997b) Process for the preparation of flavouring compositions and use of these compositions in foodstuffs. European Patent 0951840. Berendsen, W.R., Gendrot, G., Freund, A. and Reuss, M. (2006) A kinetic study of lipase-catalyzed reversible kinetic resolution involving verification at miniplant-scale. Biotechnol. Bioeng. 95, 883Ð892. Berger, C., Khan, J.A., Molimard, P., Martin, N. and Spinnler, H.E. (1999a) Production of sulfur flavors by ten strains of Geotrichum candidum. Appl. Environ. Microbiol. 65, 5510Ð5514. Berger, R.G. (1995) Aroma Biotechnology, Springer, Berlin. Berger, R.G. (ed.) (1997) Biotechnology of Aroma Compounds, Springer, Berlin. Berger, R.G. (ed.) (2007) Flavours and Fragrances, Springer, Berlin, Germany. Berger, R.G., De Bont, J.A.M., Eggink, G., Gehrke, M., Da Fonseca, M.M., Gros, J.-B., Van Keulen, F., Krings, U., Larroche, C., Leak, D.J. and Van Der Werf, M. (1999b) Biotransformations in the flavour industry. In: Current Topics in Flavours and Fragrances (ed. K.A.D. Swift), Kluwer Academic, Dordrecht, The Netherlands, pp. 139Ð170. Berger, R.G. and Zorn, H. (2004) Flavors and fragrances. In: Advances in Fungal Biotechnology for Industry, Agriculture, and Medicine (eds J.S. Tkacz and L. Lange), Kluwer Academic/Plenum Publishers, New York, pp. 341Ð358. Bernhardt, R. (2006) Cytochromes P 450 as versatile biocatalysts. J. Biotechnol. 124, 128Ð145. Besson, I., Creuly, C., Gros, J.B. and Larroche, C. (1997) Pyrazine production by Bacillus subtilis in solid- state fermentation on soybeans. Appl. Microbiol. Biotechnol. 47, 489Ð495. Best, D.J., Floyd, N.C., Magalhaes, A., Burfield, A. and Rhodes, P.M. (1987) Initial enzymatic steps in the degradation of ␣-pinene by Pseudomonas fluorescens NCIMB 11671. Biocatalysis 1, 147Ð 159. Bhattacharyya, P.K., Prema, B.R., Kulkarni, B.D. and Pradhan, S.K. (1960) Microbiological transformation of terpenes: hydroxylation of ␣-pinene. Nature 187, 689Ð690. Bock, G., Benda, I. and Schreier, P. (1988) Microbial transformation of geraniol and nerol by Botrytis cinerea. Appl. Microbiol. Biotechnol. 27, 351Ð357. Boker,¬ A., Fischer, M. and Berger, R.G. (2001) Raspberry ketone from submerged cultured cells of the basidiomycete Nidula niveo-tomentosa. Biotechnol. Prog. 17, 568Ð572. Bourg-Garros, S., Razafindramboa, N. and Pavia, A.A. (1997) Synthesis of (Z)-3-hexen-1-yl butyrate in hexane and solvent-free medium using Mucor miehei and Candida antarctia lipases. J. Am. Oil Chem. Soc. 74, 1471Ð1475. Boyd, D.R., Clarke, D., Cleij, M.C., Hamilton, J.T.G. and Sheldrake, G.N. (2000) Bacterial biotransformation of isoprene and related dienes. Monatsh. Chem. 131, 673Ð685. Caldentey, P., Fumi, M.D., Mazzoleni, V. and Careri, M. (1998) Volatile compounds produced by microorganisms isolated from cork. Perfum. Flavor. 13, 185Ð188. Cass, B.J., Schade, F., Robinson, C.W., Thompson, J.E. and Legge, R.L. (2000) Production of tomato flavor from a crude enzyme preparation using a hollow-fiber reactor. Biotechnol. Bioeng. 67, 372Ð377. Cavaille-Lefebvre, D. and Combes, D. (1997) Lipase synthesis of short-chain flavour thioesters in solvent-free medium. Biocatal. Biotransform. 15, 265Ð279. Ceynowa, J. and Koter, I. (1997) Lipase-catalyzed kinetic resolution of (R, S)-1-phenylethyl propionate in an enzyme membrane reactor. Acta Biotechnol. 17, 253Ð264. CFR (1993) US Government Printing Office, Office of Electronic Information Dissemination Services 732 N. Capital Street, Washington, DC 20402, USA. http://www.gpoaccess.gov/ cfr/index.html. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 117

Chandrasekaran, M. (1997) Industrial enzymes from marine microorganisms: the Indian scenario. J. Marine Biotechnol. 5, 86Ð89. Chatterjee, T. and Bhattacharyya, D.K. (1998a) Synthesis of terpene esters by alcoholysis reaction with Mucor miehei lipase in a solvent-free system. J. Am. Oil Chem. Soc. 75, 651Ð655. Chatterjee, T. and Bhattacharyya, D.K. (1998b) Synthesis of terpene esters by an immobilized lipase in a solvent-free system. Biotechnol. Lett. 20, 865Ð868. Chatterjee, T., De, B.K. and Bhattacharyya, D.K. (1999a) Microbial oxidation of ␣-pinene to (+)-␣-terpineol by Candida tropicalis. Indian J. Chem., Sect. B: Org. Chem. Incl. Med. Chem. 38B, 515Ð517. Chatterjee, T., De, B.K. and Bhattacharyya, D.K. (1999b) Microbial conversion of isoeugenol to vanillin by Rhodococcus rhodochrous. Indian J. Chem., Sect. B: Org. Chem. Incl. Med. Chem. 38B, 538Ð541. Cheetham, P.S.J. (2004) Bioprocesses for the manufacture of ingredients for foods and cosmetics. Adv. Biochem. Eng./Biotechnol. 86, 83Ð158. Cheetham, P.S.J., Gradley, M.L. and Sime, J.T. (1999) Flavour/aroma materials and their preparation. World Patent 2000050622. Cheng, A.-X., Xiang, C.-Y., Li, J.-X., Yang, C.-Q., Hu, W.-L., Wang, L.-J., Lou, Y.-G. and Chen, X.- Ya. (2007) The rice (E)-␤-caryophyllene synthase (OsTPS3) accounts for the major inducible volatile sesquiterpenes. Phytochemistry 68, 1632Ð1641. Civolani, C., Roncetti, P., Barghini, A.R., Ruzzi, M. and Schiesser, A. (2000) Bioconversion of ferulic acid into vanillin acid by means of a vanillate-negative mutant of Pseudomonas fluorescens strain BF14. Appl. Environ. Microbiol. 66, 2311Ð2317. Claon, P.A. and Akoh, C.C. (1994a) Enzymatic synthesis of geranyl acetate in n-hexane with Candida antarctica lipases. J. Am. Oil Chem. Soc. 71, 575Ð578. Claon, P.A. and Akoh, C.C. (1994b) Effect of reaction parameters on SP435 lipase-catalyzed synthesis of citronellyl acetate in organic solvent. Enzyme Microbiol. Technol. 16, 835Ð838. Daigle, P., Gelinas, P., Leblanc, D. and Morin, A. (1999) Production of aroma compounds by Geotrichum candidum on waste bread crumb. Food Microbiol. 16, 517Ð522. Danchet, S., Buisson, D. and Azerad, R. (1998) Microbial reduction of varying size cyclic ␤-keto esters. Stereoselective synthesis of chiral lactones and epoxides. J. Mol. Catal. B: Enzym. 5, 255Ð259. de Boer, L., Meima, R.B., Zorn, H., Scheibner, M., Hulsdau,¬ B. and Berger, R.G. (2005) Novel enzymes for use in enzymatic bleaching of food products, PTC/EPA2005/053329, PCT/EP2006064132. de Carvalho, C.C.R. and da Fonseca, M.M.R. (2006) Biotransformation of terpenes. Biotechnol. Adv. 24, 134Ð142. De Castro, H.F., Moriya, R.Y., Oliveira, P.C. and Soares, C.M.F. (1999) Fusel oil as precursor for aroma generation by biotransformation using lipase: scientific note. Appl. Biochem. Biotechnol. 77–79, 817Ð826. Debourg, A. (2000) Yeast flavor metabolites. Eur. Brew. Conv. Monogr. 28, 60Ð73. Deetae, P., Bonnarme, P., Spinnler, H.E. and Helinck, S. (2007) Production of volatile aroma compounds by bacterial strains isolated from different surface-ripened French cheeses. Appl. Microbiol. Biotechnol. 76, 1161Ð1171. Demarigny, Y., Berger, C., Desmasures, N., Gueguen, M. and Spinnler, H.E. (2000) Flavor sulfides are produced from methionine by two different pathways by Geotrichum candidum. J. Dairy Res. 67, 371Ð 380. Demyttenaere, J.C.R. and de Kimpe, N. (2000) Biotransformation of geraniol and nerol by sporulated surface cultures of Aspergillus niger.In:Frontiers of Flavour Science (eds P. Schieberle and K.-H. Engel), Deutsche Forschungsanstalt fur¬ Lebensmittelchemie, Garching, Germany, pp. 400Ð404. Demyttenaere, J.C.R. and de Pooter, H.L. (1998) Biotransformation of citral and nerol by spores of Penicillium digitatum. Flavour Frag. J. 13, 173Ð178. De-Oliveira, A.C.A.X., Fidalgo-Neto, A.A. and Paumgartten, F.J.R. (1999) In vitro inhibition of liver monooxygenases by ␤-ionone, 1,8-cineole, (−)-menthol and terpineol. Toxicology 135, 33Ð41. Dias, B. and Weimer, B. (1998) Conversion of methionine to thiols by lactococci, lactobacilli,andbrevibac- teria. Appl. Environ. Microbiol. 64, 3320Ð3326. Divyashree, M.S., George, J. and Agrawal, R. (2006) Biotransformation of terpenic substrates by resting cells of Aspergillus niger and Pseudomonas putida isolates. J. Food Sci. Technol. 43, 73Ð76. Dorneburg,¬ H. and Knorr, D. (1996) Generation of colors and fragrancies in plant cell and tissue cultures. Crit. Rev. Plant Sci. 15, 141Ð168. Drider, D., Santos, J.M., Arraiano, C.M. and Lopez, P. (1998) RNA processing is involved in the posttran- scriptional control of the citQRP operon from Lactococcus lactis biovar diacetylactis. Mol. Gen. Genet. 258, 9Ð15. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

118 Food Flavour Technology

Dudareva, N. and Pichersky, E. (2000) Biochemical and molecular genetic aspects of floral scents. Plant Physiol. 122, 627Ð634. Dufosse, L., Feron, G., Perrin, C., Djian, A. and Spinnler, H.E. (2000) Influence of bioconversion of ricinoleic acid into ␥-decalactone: optimisation of the production using the yeasts Sporidiobolus salmonicolor and Sporidiobolus ruinenii.In:Frontiers of Flavour Science (eds P. Schieberle and K.-H. Engel), Deutsche Forschungsanstalt fur¬ Lebensmittelchemie, Garching, Germany, pp. 389Ð394. Enzelberger, M.M., Bornscheuer, U.T., Gatfield, I. and Schmid, R.D. (1997) Lipase catalysed resolution of ␥-and␦-lactones. J. Biotechnol. 56, 129Ð134. Eti« evant,« P. and Schreier, P. (eds) (1995) Bioflavour 95, INRA, Paris. Fabre, C.E., Blanc, P.J. and Goma, G. (1998) 2-Phenylethyl alcohol: an aroma profile. Perfum. Flavor. 23, 43Ð45. Farbood, M.I. and Willis, B.J. (1983) ␥-Decalactone. World Patent 8301072. Feher, E., Major, B., Belafi-bako, K. and Gubicza, L. (2007) On the background of enhanced stability and reusability of enzymes in ionic liquids. Biochem. Soc. Trans. 35, 1624Ð1627. Frings, K., Koch, M. and Hartmeier, W. (1999) Kinetic resolution of 1-phenylethanol with high enantioselectivity with native and immobilized lipase in organic solvents. Enzyme Microbiol. Technol. 25, 303Ð 309. Fu, T.-J. (1999) Plant cell and tissue culture for food ingredient production: safety considerations. In: Plant Cell and Tissue Culture for the Production of Food Ingredients (eds T.-J. Fu et al.), Kluwer Academic/Plenum, New York, pp. 237Ð250. Ganga, A.M., Pinaga, F., Valles, S., Ramon, D. and Querol, A. (1999) Aroma improving in microvinification processes by the use of a recombinant wine yeast strain expressing the Aspergillus nidulans xlnA gene. Int. J. Food Microbiol. 47, 171Ð178. Gatfield, I.L. (1997) Biotechnological production of flavour-active lactones. In: Biotechnology of Aroma Compounds (ed. R.G. Berger.), Springer, Berlin, pp. 221Ð238. Gatfield, I.L., Hilmer, J.M., Werkhoff, P. and Bertram, H.J. (2000) Mechanistic aspects of the formation of natural-2-methylbutyric acid via fermentative oxidation of the corresponding alcohols. Adv. Food Sci. 22, 100Ð106. Gatfield, I.L. and Sommer, H. (1997) Facile conversion of 3-decen-4-olide to (+)-R-4-decanolide using baker’s yeast. Adv. Food Sci. 19, 12Ð14. Gehrke, M., Krings, U. and Berger, R.G. (2000) Selective recovery of flavour compounds using reversed- phase polystyrene adsorbents. Flavour Frag J. 15, 108Ð114. Goldstein, W.E. (1999) Economic considerations for food ingredients produced by plant cell and tissue culture. In: Plant Cell and Tissue Culture for the Production of Food Ingredients (eds T.-J. Fu et al.), Kluwer Academic/Plenum, New York, pp. 195Ð214. Gonzalez-Navarro,« H. and Braco, L. (1998) Lipase-enhanced activity in flavor ester reactions by trapping enzyme conformers in the presence of interfaces. Biotechnol. Bioeng. 59, 122Ð127. Goto, M., Kawasaki, M. and Kometani, T. (2000) Enzymatic resolution of 2-phenyl-1-propanol by enantioselective hydrolysis of its ester having a bulky group in an acyl moiety. J. Mol. Catal. B: Enzym. 9, 245Ð250. Griffiths, A., Prestage, S., Linforth, R.S.T., Zhang, J.L., Taylor, A.J. and Grierson, D. (1999) Fruit-specific lipoxygenase suppression in antisense-transgenic tomatoes. Postharv. Biol. Technol. 17, 163Ð174. Griffiths, E.T., Bociek, M., Harries, P.C., Jeffcoat, R., Sissons, D.J. and Trudgill, P.W. (1987) Bacterial metabolism of ␣-pinene: pathway from ␣-pinene oxid to acyclic metabolites in Norcardia sp. strain P18.4. J. Bacteriol. 169, 4972Ð4979. Grivel, F., Larroche, C. and Gros, J.B. (1999) Determination of the reaction yield during biotransformation of the volatile and chemically unstable compound ␤-ionone by Aspergillus niger. Biotechnol. Prog. 15, 4980Ð4984. Gueguen, Y., Chemardin, P., Pien, S. and Arnaud Alain, G.P. (1997) Enhancement of aromatic quality of muscat wine by the use of immobilized ␤-glucosidase. J. Biotechnol. 55, 151Ð156. Guo, W., Salmon, J.-M., Baumes, R., Tapiero, C. and Gunata,¬ Z. (1999) Purification and some properties of an Aspergillus niger ␤-apiosidase from an enzyme preparation hydrolyzing aroma precursors. J. Agric. Food Chem. 47, 2589Ð2594. Haffner, T. and Tressl, R. (1998) Stereospecific metabolism of isomeric epoxyoctadecanoic acids in the lactone-producing yeast Sporidiobolus salmonicolor. Lipids 33, 47Ð58. Haleva-Toledo, E., Naim, M., Zehavi, U. and Rouseff, R.L. (1999) Formation of ␣-terpineol in citrus juices, model and buffer solutions. J. Food Sci. 64, 838Ð841. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 119

Hamada, H. and Furuya, T. (1999) The selectivity by plant biotransformation. In: Plant Cell and Tissue Culture for the Production of Food Ingredients (eds T.-J. Fu et al.), Kluwer Academic/Plenum, New York, pp. 113Ð120. Hansen, A. and Schieberle, P. (2005) Generation of aroma compounds during sourdough fermentation: applied and fundamental aspects. Trends Food Sci. Technol. 16, 85Ð94. Harjani, J.R., Naik, P.U., Nara, S.J. and Salunkhe, M.M. (2007) Enzyme-mediated reactions in ionic liquids. Curr. Org. Synth. 4, 354Ð369. Haudenschild, C., Schalk, M., Karp, F. and Croteau, R. (2000) Functional expression of regiospecific cytochrome P450 limonene hydroxylases from mint (Mentha ssp.) in Escherichia coli and Saccharomyces cerevisiae. Arch. Biochem. Biophys. 379, 127Ð136. Havkin-Frenkel, D., Podstolski, A. and Knorr, D. (1996) Effect of light on vanillin precursors formation by in vitro cultures of Vanilla planifolia. Plant Cell, Tissue Organ Cult. 45, 133Ð136. Hayashida, Y.,Nishimura, K. and Colin Slaughter, J. (1998) Formation mechanisms of flavor-active furanones, 4-hydroxy-2,5-dimethyl-3(2H)-furanone (HDMF) and 4-hydroxy-2(or 5)-ethyl-5(or 2)-methyl-3(2H)- furanone (HEMF), in mugi-miso, fermented soy-bean paste with barley-koji. Nippon Jozo Kyokaishi 93, 730Ð738. Hayashida, Y., Kuriyama, H., Nishimura, K. and Slaughter, J.C. (1999) Production of 4-hydroxyfuranones in simple media by fermentation. Biotechnol. Lett. 21, 505Ð509. Hener, U., Brand, W.A., Hilkert, A.W., Juchelka, D., Mosandl, A. and Podebrad, F. (1998) Simultaneous online analysis of 18O/16Oand13C/12C ratios of organic compounds using GC-pyrolysis-IRMS. Z. Lebensm.-Unters. Forsch. A 206, 230Ð232. Huang, S.Y., Chang, H.L. and Goto, M. (1998) Preparation of surfactant-coated lipase for the esterification of geraniol and acetic acid in organic solvents. Enzyme Microbiol. Technol. 22, 552Ð557. Ikushima, Y., Saito, N., Hatakeda, K. and Sato, O. (1996) Promotion of a lipase-catalyzed esterification in supercritical carbon dioxide in the near-critical region. Chem. Eng. Sci. 51, 2817Ð2822. Imm, J.Y. and Lee, C.M. (1999) Production of seafood flavor from redhake (Urophycis chuss) by enzymatic hydrolysis. J. Agric. Food Chem. 47, 2360Ð2366. Irmler, S., Raboud, S., Beisert, B., Rauhut, D. and Berthoud, H. (2008) Cloning and characterization of two Lactobacillus casei genes encoding a cystathionine lyase. Appl. Environ. Microbiol. 74, 99Ð106. Kalyani, A., Prapulla, S.G. and Karanth, N.G. (2000) Study on the production of 6-pentyl-␣-pyrone using two methods of fermentation. Appl. Microbiol. Biotechnol. 53, 610Ð612. Karmakar, B., Vohar,R.M., Nandanwar, H., Sharma, P.,Gupta, K.G. and Sobti, R.C. (2000) Rapid degradation of ferulic acid via 4-vinylguaiacol and vanillin by a newly isolated strain of Bacillus coagulans. J. Biotechnol. 80, 195Ð202. Kilcawley, K.N., Wilkinson, M.G. and Fox, P.F. (1998) Enzyme-modified cheese. Int. Dairy J. 8, 1Ð10. Kim-Jungbae, Altreuter, D.H., Clark, D.S. and Dordick-Jonathan, S. (1998) Rapid synthesis of fatty acid esters for use as potential food flavours. J. Am. Oil Chem. Soc. 75, 1109Ð1114. King, A. and Dickinson, J.R. (2000) Biotransformation of monoterpene alcohols by Saccharomyces cerevisiae, Torulaspora delbrueckii and Kluyveromyces lactis. Yeast 16, 499Ð506. Klein, N. and Lortal, S. (1999) Attenuated starters: an efficient means to influence cheese ripening Ð a review. Int. Dairy J. 9, 751Ð762. Kraft, P., Bajgrowicz, J.A., Denis, C. and Frater,« G. (2000) Odds and trends: recent developments in the chemistry of odorants. Angew.Chem.Int.Ed.39, 2980Ð3010. Krings, U. and Berger, R.G. (1998) Biotechnological production of flavours and fragrances. Appl. Microbiol. Biotechnol. 49,1Ð8. Krings, U. and Berger, R.G. (2008) In situ recovery of the aroma compound perillene from stirred-tank cultured Pleurotus ostreatus using gas stripping and adsorption on polystyrene. Biotechnol Lett. 30, 1347Ð1351. Krings, U., Hapetta, D. and Berger, R.G. (2008) A labeling study on the formation of perillene by submerged cultured oyster mushroom, Pleurotus ostreatus. Appl. Microbiol. Biotechnol. 78, 533Ð541. Krings, U., Pilawa, S., Theobald, C. and Berger, R.G. (2001) Phenyl propenoic side chain degradation of ferulic acid by Pycnoporus cinnabarinus Ð elucidation of metabolic pathways using [5-2H]-ferulic acid. J. Biotechnol. 85, 305Ð314. Kumar, C.G., Sharma, A. and Kanawjia, S.K. (1998) Prospects for bioproduction of food flavours. Indian Food Ind 17, 98Ð111. Kwon, D.Y., Hong, Y.J. and Yoon, S.H. (2000) Enantiomeric synthesis of (S)-2-methylbutanoic acid methyl ester, apple flavor, using lipases in organic solvent. J. Agric. Food Chem. 48, 524Ð530. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

120 Food Flavour Technology

Laboret, F. and Perraud, R. (1999) Lipase-catalyzed production of short-chain acids terpenyl esters to the food industry. Appl. Biochem. Biotechnol. 82, 185Ð198. Landmann, Ch., Fink, B., Festner, M., Dregus, M., Engel, K.-H. and Schwab, W. (2007) Cloning and functional characterization of three terpene synthases from lavender (Lavandula angustifolia). Arch. Biochem. Biophys. 465, 417Ð429. Lapadatescu, C., Feron, G., Vergoignan, C., Djian, A., Durand, A. and Bonmarme, P. (1997) Influence of cell immobilisation on the production of benzaldehyde and benzyl alcohol by the white-rot fungi Bjerkandera adusta, Ischnoderma benzoinum and Dichomitus squalens. Appl. Microbiol. Biotchnol. 47, 708Ð714. Lapadatescu, C., Ginies, C., Le Quere, J.-L. and Bonnarme, P. (2000) Novel scheme for biosynthesis of aryl metabolites from l-phenylalanine in the fungus Bjerkandera adusta. Appl. Environ. Microbiol. 66, 1517Ð1522. Larroche, C., Creuly, C. and Gros, J.B. (1995) Fed-batch biotransformation of ␤-ionone by Aspergillus niger. Appl. Microbiol. Biotechnol. 43, 222Ð227. Leahy, M.M. and Roderick, R.G. (1999) Fruit flavor biogenesis. In: Flavor Chemistry (eds R. Teranishi, E.L. Wick and I. Hornstein.), Kluwer Academic/Plenum, New York, pp. 275Ð286. Lemieux, B., Aharoni, A. and Schena, M. (1998) Overview of DNA chip technology. Mol. Breed. 4, 277Ð289. Lesage-Meessen, L., Stentelaire, C., Lomascolo, A., Couteau, D., Asther, M., Moukha, S., Record, E., Sigoillot, J.-C. and Asther, M. (1999) Fungal transformation of ferulic acid from sugar beet pulp to natural vanillin. J. Sci. Food Agric. 79, 487Ð490. Li, T. and Rosazza, J.P.N. (2000) Biocatalytic synthesis of vanillin. Appl. Environ. Microbiol. 66, 684Ð687. Lilly, M., Lambrecht, M.G. and Pretorius, I.S. (2000) Effect of increased yeast alcohol acetyltransferase activity on flavor profiles of wine and distillates. Appl. Environ. Microbiol. 66, 744Ð754. Lizarraga-Guerra,« R., Guth, H. and Lopez,« M.G. (1997) Identification of the most potent odorants in huitla- coche (Ustilago maydis) and austern pilzen (Pleurotus sp.) by aroma extract dilution analysis and static head-space samples. J. Agric. Food Chem. 45, 1329Ð1332. Lupien, S., Karp, F., Wildung, M. and Croteau, R. (1999) Regiospecific cytochrome P450 limonene hydroxylases from mint species: cDNA isolation, characterization, and functional expression of (−)-4S- limonene-3-hydroxylase and (−)-4S-limonene-6-hydroxylase. Arch. Biochem. Biophys. 368, 181Ð192. Lutz-Wahl, S., Fischer, P., Schmidt-Dannert, C., Wohlleben, W., Hauer, B. and Schmid, R.D. (1998) Stereo- and regioselective hydroxylation of ␣-ionone by Streptomyces strains. Appl. Environ. Microbiol. 64, 3878Ð3881. Maicas, S. and Mateo, J.J. (2005) Hydrolysis of terpenyl glycosides in grape juice and other fruit juices: a review. Appl. Microbiol. Biotechnol. 67, 322Ð335. Manning, K. (1998) Isolation of a set of ripening-related genes from strawberry: their identification and possible relationship to fruit quality traits. Planta 205, 622Ð631. Marostica, J.M.R. and Pastore, G.M. (2007) Biotransformation of limonene: a review of the main metabolic pathways. Quim. Nova 30, 382Ð387. Martin, N., Savonitto, S., Molimard, P., Berger, C., Brousse, M. and Spinnler, H.E. (1999) Flavor generation in cheese curd by coculturing with selected yeast, mold, and bacteria. J. Dairy Sci. 82, 1072Ð1080. Matsuda, Y., Toyoda, H., Sawabe, A., Maeda, K., Shimizu, N., Fujita, N., Fujita, T., Nonomura, T. and Ouchi, S. (2000) A hairy root culture of melon produces aroma compounds. J. Agric. Food Chem. 48, 1417Ð1420. Medeiros, A.B.P., Pandey, A., Christen, R.J.S. and Soccol, C.R. (2000) Optimization of the production of aroma compounds by Kluyveromyces marxianus in solid-state fermentation using factorial design and response surface technologies. Biochem. Eng. J. 6, 33Ð40. Mellouki, F., Vannereau, A. and Cosson, L. (1996) Les Precurseurs d’arome dans des cultures cellulaires d’allium. Acta Bot. Gallica 143, 131Ð136. Menzel, M. and Schreier, P. (2007) Enzymes and flavour biotechnology. In: Flavours and Fragrances (ed. R.G. Berger.), Springer, Berlin, Germany, pp. 489Ð505. Michor, H., Marr, R., Gamse, T., Schilling, T., Klingsbichel, E. and Schwab, H. (1996) Enzymatic catalysis in supercritical carbon dioxide: comparison of different lipases and a novel esterase. Biotechnol. Lett. 18, 79Ð84. Midje, D.L., Bastian, E.D., Morris, H.A., Martin, F.B., Bridgeman, T. and Vickers, Z.M. (2000) Flavor enhancement of reduced fat cheddar cheese using an integrated culturing system. J. Agric. Food Chem. 48, 1630Ð1636. Mikolajek, R., Spiess, A.C. and Buechs, J. (2007) Feasibility of gas/solid carboligation: conversion of benzaldehyde to benzoin using thiamine diphosphate-dependent enzymes. J. Biotechnol. 129, 723Ð725. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 121

Miyazawa, M., Honjo, Y. and Kameoka, H. (1997) Biotransformation of acyclic terpenoid (2E,6E)-farnesol by plant pathogenic fungus Glomerella cingulata. Phytochemistry 43, 105Ð109. Miyazawa, M., Honjo, Y.and Kameoka, H. (1998) Biotransformation of the sesquiterpenoid (+)-␥-gurjunene using a plant pathogenic fungus Glomerella cingulata, as a biocatalyst. Phytochemistry 49, 1283Ð1285. Molinari, F., Bertolini, C. and Aragozzini, F. (1997) Microbial biotransformations to obtain (R)- and (S)-2- octanol. Ann. Microbiol. Enzymol. 47, 131Ð137. Molinari, F., Villa, R. and Aragozzini, F. (1998a) Selective esterifications catalyzed by dry microbial cells in organic solvent. Med. Fac. Landbouw. Univ. Gent 63, 1263Ð1270. Molinari, F., Villa, R. and Aragozzini, F. (1998b) Production of geranyl acetate and other acetates by direct esterification catalyzed by mycelium of Rhizopus delemar in organic solvent. Biotechnol. Lett. 20, 41Ð44. Mosandl, A., Fuchs, S., Wust,¬ M. and Beck, T. (2000) Biosynthetic studies on monoterpenoids using stable isotope labelling and enantioselective analysis. In: Frontiers of Flavour Science (eds P. Schieberle and K.-H. Engel.), Deutsche Forschungsanstalt fur¬ Lebensmittelchemie, Garching, Germany, pp. 433Ð439. Muheim, A., Hausler,¬ A., Schilling, B. and Lerch, K. (1997) The impact of recombinant DNA-technology on the flavour and fragrance industry. In: Flavours and Fragrances (ed. K.A.D. Swift.), The Royal Society of Chemistry, Cambridge, pp. 11Ð20. Muheim, A., Hausler,¬ A., Schilling, B. and Lerch, K. (1998) The impact of recombinant DNA technology on the flavor and fragrance industry. Perfum. Flavor. 23, 21Ð27. Muheim, A. and Lerch, K. (1999) Towards a high yield bioconversion of ferulic acid to vanillin. Appl. Microbiol. Biotechnol. 51, 456Ð461. Nagaoka, H. and Kayahara, H. (2000) Resolution and synthesis of optically active alcohols with immobilized water-soluble proteins from green pea, soybean and buckwheat as new bio-catalysts. Biosci. Biotechnol. Biochem. 64, 781Ð784. Nakagawa, H., Dobashi, Y.,Sato, T., Yoshida,K., Tsugane, T., Shimura, S., Kirimura, K., Kino, K. and Usami, S. (2000) ␣-Anomer-selective glucosylation of menthol with high yield through a crystal accumulation reaction using lyophilized cells of Xanthmonas campestris WU-9701. J. Biosci. Bioeng. 89, 138Ð144. Nankai, H., Miyazawa, M. and Kameoka, H. (1998) Biotransformation of the sesquiterpenoid (2Z,6Z)- farnesol by the plant pathogenic fungus Glomerella cingulata. Phytochemistry 47, 1025Ð1028. Narbad, A.N. and Gasson, M.J. (1998) Metabolism of ferulic acid via vanillin using a novel CoA-dependent pathway in a newly-isolated strain of Pseudomonas fluorescens. Microbiology 144, 1397Ð1405. Neuser, F., Zorn, H. and Berger, R.G. (2000a) Formation of aliphatic and aromatic ␣-hydroxy ketones by Zygosaccharomyces bisporus. Z. Naturforsch. 55c, 560Ð568. Neuser, F., Zorn, H. and Berger, R.G. (2000b) Generation of odorous acyloins by yeast pyruvate decarboxy- lases and their occurrence in sherry and soy sauce. J. Agric. Food Chem. 48, 6191Ð6195. Neuser, F., Zorn, H., Richter, U. and Berger, R.G. (2000c) Purification, characterisation and cDNA sequencing of pyruvate decarboxylase from Zygosaccharomyces bisporus. Biol. Chem. 381, 349Ð354. Niedz, R.P., Moshonas, M.G., Peterson, B., Shapiro, J.P. and Shaw, P.E. (1997) Analysis of sweet orange (Citrus sinensis L. Osbeck) callus cultures for volatile compounds by gas chromatography with mass selective detector. Plant Cell, Tissue Organ Cult. 51, 181Ð185. Nierop Groot, M.N. and de Bont, J.A.M. (1998) Conversion of phenylalanine to benzaldehyde initiated by an aminotransferase in Lactobacillus plantarum. Appl. Environ. Microbiol. 64, 3009Ð3014. Nonino, E.A. (1997) Where is the citrus industry going? Perfum. Flavor. 22, 53Ð58. Noordermeer, M.A., Veldink, G.A. and Vliegenhart, J.F.G. (1999) Alfalfa contains substantial 9- hydroperoxide lyase activity and a 3Z:2E-enal isomerase. FEBS Lett. 443, 201Ð204. Oda, S., Inada, Y., Kato, A., Matsudomi, N. and Ohta, H. (1995) Production of (S)-citronellic acid and (R)-citronellol with an interface bioreactor. J. Ferment. Bioeng. 80, 559Ð564. Oda, S. and Ohta, H. (1997) Double coupling of acetyl coenzyme A production and microbial esterification with alcohol acetyltransferase in an interface bioreactor. J. Ferment. Bioeng. 83, 423Ð428. Oddou, J., Stentelaire, C., Lesage-Meessens, L., Asther, M. and Ceccaldi, B.C. (1999) Improvement of ferulic acid bioconversions into vanillin by use of high-density cultures of Pycnoporous cinnabarinus. Appl. Microbiol. Biotechnol. 53,1Ð6. Ohloff, G. (1994) Scent and Fragrances, the Fascination of Odors and Their Chemical Perspectives, Springer, Berlin. Onken, J. and Berger, R.G. (1999a) Effects of R-(+)-limonene on submerged cultures of the terpene transforming basidiomycete Pleurotus sapidus. J. Biotechnol. 69, 163Ð168. Onken, J. and Berger, R.G. (1999b) Biotransformation of citronellol by the basidiomycete Cystoderma carcharias in an aerated membrane bioreactor. Appl. Microbiol. Biotechnol. 51, 158Ð164. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

122 Food Flavour Technology

Orejas, M., Ibanez, E. and Ramon, D. (1999) The filamentous fungus Aspergillus nidulans produces an ␣-l-rhamnosidase of potential oenological interest. Lett. Appl. Microbiol. 28, 383Ð388. Orlich, B. and Schomacker,¬ R. (1999) Enzymatic reduction of a less water-soluble ketone in reverse micelles with NADH regeneration. Biotechnol. Bioeng. 65, 357Ð361. Overhage, J., Priefert, H. and Rabenhorst, J. (1999a) Biotransformation of eugenol to vanillin by a mutant of Pseudomonas sp. strain HR199 contructed by disruption of the vanillin dehydrogenase (vdh) gene. Appl. Microbiol. Biotechnol. 52, 820Ð828. Overhage, J., Priefert, H. and Steinbuchel,¬ A. (1999b) Biochemical and genetic analyses of ferulic acid catabolism in Pseudomonas sp. strain HR199. Appl. Environ. Microbiol. 65, 4837Ð4847. Overmeyer, A., Schrader Lippelt, S., Kasche, V. and Brunner, G. (1999) Lipase-catalyzed kinetic resolution ◦ ◦ of racemates at temperatures from 40 C to 160 C in supercritical CO2. Biotechnol. Lett. 21, 65Ð69. Pandey, A., Soccol, C.R., Nigam, P., Soccol, V.T., Vandenberghe, L.P.S. and Mohan, R. (2000a) Biotechno- logical potential of agro-industrial residues. II: cassava bagasse. Biores. Technol. 74, 81Ð87. Pandey, A., Soccol, C.R. and Mitchell, D. (2000b) New developments in solid state fermentation: I. Biopro- cesses and products. Process Biochem. 35, 1153Ð1169. Park, O.J., Holland, H.L., Khan, J.A. and Vulfson, E.N. (2000) Production of flavour ketones in aqueous- organic two-phase systems by using free and microencapsulated fungal spores as biocatalysts. Enzyme Microbiol. Technol. 26, 235Ð242. Park, S.-H., Kim, K.-S., Suzuki, Y. and Kim, S.-U. (1997) Metabolism of isopiperitones in cell suspension culture of Mentha piperita. Phytochemistry 44, 623Ð626. Pedersen, H., Chin, C.-K. and Dutta, A. (1999) Yield improvement in plant cell cultures by in situ extraction. In: Plant Cell and Tissue Culture for the Production of Food Ingredients (eds T.-J. Fu et al.), Kluwer Academic/Plenum, New York, pp. 129Ð138. Pogorzelski, E. and Wilkowska, A. (2007) Flavour enhancement through the enzymatic hydrolysis of glycosidic aroma precursors in juices and wine beverages: a review. Flavour Frag. J. 22, 251Ð254. Prema, B.R. and Bhattacharyya, P.K. (1962) Microbiological transformation of terpenes II. Transformations of ␣-pinene. Appl. Microbiol. 10, 524Ð528. Prestage, S., Linforth, T., Rob, S., Taylor, A.J., Lee, E., Speirs, J. and Schuch, W. (1999) Volatile production in tomato fruit with modified alcohol dehydrogenase activity. J. Sci. Food Agric. 79, 131Ð136. Pridmore, R.D., Crouzillat, D., Walker, C., Foley, S., Zink, R., Zwahlen, M.C., Brussow, H., Petiard, V. and Mollet, B. (2000) Genomics, molecular genetics and the food industry. J. Biotechnol. 78, 251Ð258. Prince, C.L., Shuler, M.L. and Yamada, Y. (1997) Altering flavor profiles in onion (Allium cepa L.) root cultures through directed biosynthesis. Biotechnol. Prog. 13, 506Ð510. Pu, F., Jiang, M., Zhang, Z.J., Zhang, J.S. and Hao, F.H. (1998) Study on Vanilla curing enzyme treatment method. Acta Bot. Yunnanica 20, 355Ð361. Rabenhorst, J. (2000) Biotechnological production of natural aroma chemicals by fermentation processes. In: Biotechnology: Biotransformations II, Vol. 8b (ed. D.R. Kelly.), Wiley-VCH, Weinheim, pp. 333Ð350. Ramachandra Rao, S. and Ravishankar, G.A. (1999) Biotransformation of isoeugenol to Vanilla flavour metabolites and capsaicin in suspended and immobilized cell cultures of Capsicum frutescens: study of the influence of ␤-cyclodextrin and fungal elicitor. Process Biochem. 35, 341Ð348. Ramachandra Rao, S. and Ravishankar, G.A. (2000) Vanilla flavour: production by conventional and biotechnological routes. J. Sci. Food Agric. 80, 289Ð304. Reil, G. and Berger, R.G. (1996a) Accumulation of chlorophyll and essential oils in photomixotrophic cell cultures of Citrus sp. Z. Naturforsch. 51c, 657Ð666. Reil, G. and Berger, R.G. (1996b) Elicitation of volatile compounds in photomixotrophic cell culture of Petroselinum crispum. Plant Cell, Tissue Organ Cult. 46, 131Ð136. Reil, G. and Berger, R.G. (1997) Variation of chlorophyll and essential oils in photomixotrophic cell cultures of Coleonema album (Thunb.). J. Plant Physiol. 150, 160Ð166. Rijnen, L., Bonneau, S. and Yvon, M. (1999) Genetic characterization of the mayor lactococcal aromatic aminotransferase and its involvement in conversion of amino acids to aroma compounds. Appl. Environ. Microbiol. 65, 4873Ð4880. Rijnen, L., Courtin, P., Gripon, J.C. and Yvon, M. (2000) Expression of a heterologous glutamate dehydrogenase gene in Lactococcus lactis highly improves the conversion of amino acids to aroma compounds. Appl. Environ. Microbiol. 66, 1354Ð1359. Risch, S.J. and Ho, C.-T.. (eds) (2000) Flavor Chemistry, ACS, Washington, DC. Rizzi, G.P. (2008) The Strecker degradation of amino acids: newer avenues for flavor formation. Food Rev. Int. 24, 416Ð435. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 123

Rodriguez-Bustamante, E. and Sanchez, S. (2007) Microbial production of C13-norisoprenoids and other aroma compounds via carotenoid cleavage. Crit. Rev. Microbiol. 33, 211Ð230. Ronse, A.C., De Pooter, H., Van De Vyver, A. and De Proft, M.P. (1998) Otacanthus species: in vitro culture, plant propagation, and the production of essential oil. In: Biotechnology in Agriculture and Forestry (ed. Y.P.S. Bajaj.), Springer, Berlin, pp. 305Ð319. Rosecke,¬ J. and Konig,¬ W.A. (2000) Odorous compounds from the fungus Gleophyllum odoratum. Flavour Frag. J. 15, 315Ð319. Roy, A. (1999) Chemoenzymatic synthesis of homochiral (R)- and (S)-karahanaenol from (R)-limonene. J. Agric. Food Chem. 47, 5209Ð5210. Sakho, M., Chassagne, D., Jaus, A., Chiarazzo, E. and Crouzet, J. (1998) Enzymatic maceration: effects on volatile components of mango pulp. J. Food Sci. 63, 975Ð978. Santiago-Gomez, M.P., Vergeley, C., Policar, C., Nicaud, J.-M., Belin, J.-M., Rochette, L. and Husson, F. (2007) Characterization of purified green bell pepper hydroperoxide lyase expressed by Yarrowia lipolytica: radicals detection during catalysis. Enz. Microb. Technol. 41, 13Ð18. Santos, P.M., Figueiredo, A.C., Oliveira, M.M., Barosso, J.G., Pedro, L.G., Deans, S.G., Younus, A.K.M. and Scheffer, J.J.C. (1998) Essential oils from hairy root cultures and from fruits and roots of Pimpinella anisum. Phytochemistry 48, 455Ð460. Santos, P.M., Figueiredo, A.C., Oliveira, M.M., Barosso, J.G., Pedro, L.G., Deans, S.G., Younus, A.K.M. and Scheffer, J.J.C. (1999) Morphological stability of Pimpinella anisum hairy root cultures and time-course study of their essential oils. Biotechnol. Lett. 21, 859Ð864. Sarhy-Bagnon, V., Lozano, P., Saucedo-Castaneda,˜ G. and Roussos, S. (2000) Production of 6-pentyl- ␣-pyrone by Trichoderma harzianum in liquid and solid state cultures. Process Biochem. 36, 103Ð 109. Sarry, J.-E. and Gunata, Z. (2004) Plant and microbial glycoside hydrolases: volatile release from glycosidic aroma precursors. Food Chem. 87, 509Ð521. Saxena, R.K., Ghosh, P.K., Gupta, R., Davidson, W.S., Bradoo, S. and Ruchi, R. (1999) Microbial lipases: potential biocatalysts for the future industry. Curr. Sci. 77, 101Ð115. Schieberle, P. and Engel, K.-H.. (eds) (2000) Frontiers of Flavour Science, Deutsche Forschungsanstalt fur¬ Lebensmittelchemie, Garching, Germany. Schumacher, K., Asche, S., Heil, M., Mosandl, A., Engel-Kristen, K. and Rauhut, D. (1998) Mircobiological oxidation of 2-methylbutanol of differing enantiomeric ratios by Gluconobacter species. Z. Lebensm.- Unters. Forsch. A 207, 74Ð76. Schwab, W., Davidovich-Rikanati, R. and Lewinsohn, E. (2008) Biosynthesis of plant-derived flavor compounds. Plant J. 54, 712Ð732. Schwab, W., Lunkenbein, S., Salentijn, E.M.J. and Aharoni, A. (2006) Genetic engineering of strawberry flavour. Dev. Food Sci. 43, 39Ð44. Scragg, A.H. (2007) The production of flavours by plant cell cultures. In: Flavours and Fragrances (ed. R.G. Berger.), Springer, Berlin, pp. 599Ð614. Serrano-Carreon,« L., Flores, C. and Galindo, E. (1997) ␥-Decalactone production by Trichoderma harzi- nanum in stirred bioreactors. Biotechnol. Prog. 13, 205Ð208. Shamel, R. and Udis-Kessler, A. (2000) The sweet taste of success. Chem. Ind. 10, 27Ð28. Shervington, A., Darwish, R. and Afifi, F. (1997) Volatile oil produced by microbial infected Rosmarinus officinalis callus and suspension cultures. Alexandria J. Pharm. Sci. 11, 81Ð84. Shimada, Y., Hirota, Y., Baba, T., Kato, S., Sugihara, A., Moriyama, S., Tominaga, Y. and Terai, T. (1999) Enzymatic synthesis of l-menthyl ester in organic solvent-free system. J.Am.OilChem.Soc.76, 1139Ð 1142. Shimoni, E., Ravid, U. and Shoham, Y. (2000) Isolation of a Bacillus sp. capable of transforming isoeugenol to vanillin. J. Biotechnol. 78, 1Ð9. Shin, S.-W. (1995) Biotransformation of exogenous monoterpenoids by plant cell culture. Saengyak Hakhoechi (Korean J. Pharmacol.) 26, 227Ð238. Short, P. (2002) Reshaping flavors and fragrances. C&EN 80, 32Ð37. Shuler, M.L. (1999) Overview of yield improvement strategies for secondary metabolite production in plant cell culture. In: Plant Cell and Tissue Culture for the Production of Food Ingredients (eds T.-J. Fu et al.), Kluwer Academic/Plenum, New York, pp. 75Ð84. Sime, J. (2000) Biotransformation versus extraction technology. Int. Food Ingredients 1, 31Ð32. Simmonds, J. and Robinson, G.K. (1998) Formation of benzaldehyde by Pseudomonas putida ATCC 12634. Appl. Microbiol. Biotechnol. 50, 353Ð358. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

124 Food Flavour Technology

Smit, G., Verheul, A., Van Kranenburg, R., Ayad, E., Siezen, R. and Engels, W. (2000) Cheese flavour development by enzymatic conversions of peptides and amino acids. Food Res. Int. 33, 153Ð160. Smit, G., Wouters, J.T.M. and Meijer, W.C. (2004) Quality and flavor of fermented products. Handbook of Food and Beverage Fermentation Technology. Marcel Dekker, New York, pp. 89Ð111. Spagna, G., Romagnoli, D., Angelo, M. and Bianchi, G.P.P.G. (1998) A simple method for purifying glycosidases: ␣-l-arabinofuranosidase and ␤-d-glucopyranosidase from Aspergillus niger to increase the aroma of wine. Part I. Enzyme Microbiol. Technol. 22, 298Ð304. Speelmans, G., Bijlsma, A. and Eggink, G. (1998) Limonene bioconversion to high concentrations of a single and stable product, perillic acid, by a solvent resistant Pseudomonas putida strain. Appl. Microbiol. Biotechnol. 50, 538Ð544. Speirs, J., Lee, E., Holt, K., Yong-Duk, K., Scott, N.S., Loveys, B. and Schuch, W. (1998) Genetic manipulation of alcohol dehydrogenase levels in ripening tomato fruit affects the balance of some flavor aldehydes and alcohols. Plant Physiol. 117, 1047Ð1058. Steinbuchel,¬ A., Priefert, H. and Rabenhorst, J. (1997) Syntheseenzyme fur¬ die Herstellung von Conifer- ylalkohol, Coniferylaldehyd, Ferulasaure,¬ Vanillin und Vanillinsaure¬ und deren Verwendung. European Patent 0845532. Steinhart, H., Stephan, A. and Bucking,¬ M. (2000) Advances in flavor research. J. High Resolut. Chromatogr. 23, 489Ð496. Stentelaire, C., Lesage-Meessen, L., Oddou, J., Bernard, O., Bastin, G., Ceccaldi, B.C. and Asther, M. (2000) Design of a fungal bioprocess for vanillin production from vanillic acid at scalable level by Pycnoporus cinnabarinus. J. Biosci. Bioeng. 89, 223Ð230. Suezawa, Y., Yoshioka, N. and Mori, H. (1998) Bioconversions of ferulic acid and p-coumaric acid to volatile phenols by Aspergillus spp. and bacteria found in soy sauce koji and mashes. Nippon Nogei Kagaku Kaishi 72, 43Ð49. Sugawara, E. and Sakurai, Y. (1999) Effect of media constituents on the formation by halophilic yeast of the 2 (or 5)-ethyl-5 (or 2)-methyl-4-hydroxy-3(2H)-furanone aroma component specific to miso. Biosci. Biotechnol. Biochem. 63, 749Ð752. Suurmeijer, C.N.S.P., Perez-Gilabert, M., Van Unen, D.-J., Van Der Hijden, H.T.W.M., Veldink, G.A. and Vliegenthart, J.F.G. (2000) Purification, stabilization and characterization of tomato fatty acid hydroperoxide lyase. Phytochemistry 53, 177Ð185. Swiegers, J.H., Saerens, S.M.G. and Pretorius, I.S. (2008) The development of yeast strains as tools for adjust- ing the flavor for fermented beverages to market specifications. In: Biotechnology in Flavor Production (eds D. Havkin-Frenkel and F.C. Belanger.), Blackwell Publishing., Oxford, UK, pp. 1Ð55. Swift, K.A.D.. (ed.) (1999) Current Topics in Flavours and Fragrances, Kluwer Academic, Dordrecht, The Netherlands. Takeoka, G. (1999) Flavor chemistry of vegetables. In: Flavor Chemistry (eds R. Teranishi, E.L. Wick and I. Hornstein.), Kluwer Academic/Plenum, New York, pp. 287Ð304. Tan, Q. and Day, D.F. (1998) Bioconversion of (R)-(+)-limonene by immobilized Penicillium digitatum (NRRL 1202). Appl. Microbiol. Biotechnol. 49, 96Ð101. Tan, Q., Day, D.F. and Cadwallader, K.R. (1998) Bioconversion of (R)-(+)-limonene by P. digitatum (NRRL 1202). Proc. Biochem. 33, 29Ð37. Taupp, D.E., Nimtz, M., Berger, R.G. and Zorn, H. (2008) Proteome response of Nidula niveo-tomentosa to UV light, Mycologia 100, 529Ð538. Tawfik, A.A., Read, P.E. and Cuppett, S.L. (1998) Rosmarinus officinalis L. (Rosemary): in vitro culture, regeneration of plants, and the level of essential oil and monoterpenoid constituents. In: Biotechnology in Agriculture and Forestry (ed. Y.P.S. Bajaj.), Springer, Berlin, pp. 349Ð365. Teranishi, R., Wick, E.L. and Hornsteim, I. (eds) (1999) Flavor Chemistry, Kluwer Academic/Plenum, New York. Tescione, L., Asplund, P. and Curtis, W.R. (1999) Reactor design for root culture: oxygen mass transfer limitations. In: Plant Cell and Tissue Culture for the Production of Food Ingredients (eds T.-J. Fu et al.), Kluwer Academic/Plenum, New York, pp. 139Ð156. Teunissen, M.J. and De Bont, J.A.M. (1995) Will terpenes be of any significance in future biotechnology. In: Bioflavor ’95 (eds P. Eti« evant« and P. Schreier.), INRA, Versailles, Cedex, pp. 329Ð330. Tominaga, T., Des Peyot, G.C. and Dubourdieu, D. (1998) A new type of flavor precursor in Vitis vinifera L. cv. Sauvignon blanc: S-cysteine conjugates. J. Agric. Food Chem. 46, 5215Ð5219. Treffenfeldt, W., Vollmer, H., Preuss, A., Berger, R.G., Krings, U. and Latza, E. (1999) Procedure for the purification and separation of flavor-effective or fragrance molecules by adsorption. German Patent 19835542. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

Biotechnological ﬂavour generation 125

Trivedi, A.H., Spiess, A.C., Daussmann, T. and Buechs, J. (2006) Effect of additives on gasÐphase catalysis with immobilised Thermoanaerobacter species alcohol dehydrogenase (ADH T). Appl. Microbiol. Biotechnol. 71, 407Ð414. Trudgill, P.W. (1994) Microbial metabolism and transformation of selected monoterpenes. In: Biochemistry of Microbial Degradation (ed. C. Ratledge.), Kluwer Academic, London, pp. 31Ð61. Tjener, K. and Stahnke, L.H. (2007) Flavor. In: Handbook of Fermented Meat and Poultry (ed. F. Toldra), Blackwell Publishing, Ames, Iowa, pp. 227Ð239. Van Berge, P. (1998) Flavors into the 21st century. Perfum. Flavor. 23, 1Ð12. Van Der Heijden, R., Schulte, A.E., Ramos Valdivia, A.C. and Verpoorte, R. (1998) Characterization of some isoprenoid-biosynthetic enzymes from plant cell cultures. In: New Frontiers in Screening for Microbial Biocatalysts (eds K. Kieslich, C.P. Van der Beek, J.A.M. De Bont and W.J.J. Van den Tweel.), Elsevier Science, Amsterdam, pp. 177Ð184. Van Der Werf, M.J. and De Bont, J.A.M. (1998) Screening for microorganisms converting limonene into carvone. In: Frontiers in Screening for Microbial Biocatalysis (eds K. Kieslich, C.P. Van der Beek, J.A.M. De Bont and W.J.J. Van. den Tweel.), Elsevier Science, Amsterdam, pp. 231Ð234. Van Der Werf, M.J., Keijzer, P.M. and Van der Schaft, P.H. (2000) Xanthomonas sp. C 20 contains a novel bioconversion pathway for limonene. J. Biotechnol. 84, 133Ð144. Van Der Werf, M.J., Van der Ven, C., Barbirato, F., Eppink, M.E.M., De Bont, J.A.M. and Van Berkel, W.J.H. (1999) Stereoselective carveol dehydrogenase from Rhodococcus erythropolis DCL14. J. Biol. Chem. 274, 26296Ð26304. Van Haandel, M.J.H., Saraber, F.C.E., Boersma, M.G., Laane, C., Flemming, Y., Weenen, H. and Rietjens, I.M.C.M. (2000) Characterization of different commercial soybean peroxidase preparations and use of the enzyme for N-demethylation of methyl-N-methylanthranilate to produce the food flavor methylanthranilate. J. Agric. Food Chem. 48, 1949Ð1954. Van Iersel, M.F.M., Brouwer Post, E., Rombouts, F.M. and Abee, T. (2000) Influence of yeast immobilization on fermentation and aldehyde reduction during the production of alcohol-free beer. Enzyme Microbiol. Technol. 26, 602Ð607. Van Rensburg, E., Moleleki, N., Van der Walt, J.P. and Van Dyk, M.S. (1997) Biotransformation of (+)- limonene and (−)-piperitone by yeasts and yeast-like fungi. Biotechnol. Lett. 19, 779Ð782. Vanek, T., Valterova, I. Vankova, R., Vaisar, T. (1999) Biotransformation of (-)-limonene using Solanum aviculare and Dioscorea deltoidea immobilized plant cells. Biotechnol. Lett. 21, 625Ð628. Venkateshwarlu, G., Chandravadana, M.V., Pandey, M., Tewari, R.P. and Selvaraj, Y. (2000) Volatile flavor compounds from oyster mushroom (Pleurotus florida) in submerged culture. Flavour Frag. J. 15, 320Ð 322. Verhoef, S., Ruijssenaars, H.J., de Bont, J.A.M. and Wery, J. (2007) Bioproduction of p-hydroxybenzoate from renewable feedstock by solvent-tolerant Pseudomonas putida S12. J. Biotechnol. 132, 49Ð56. Verpoorte, R., Van Der Heijden, R., Ten Hoopen, H.J.G. and Memelink, J. (1999) Novel approaches to improve plant secondary metabolite production. In: Plant Cell and Tissue Culture for the Production of Food Ingredients (eds T.-J. Fu et al.), Kluwer Academic/Plenum, New York, pp. 85Ð100. Villanueva, A., Ramon, D., Salvador, V., Lluch, M.A. and MacCabe, A. (2000) Heterologous expression in Aspergillus nidulans of a Trichoderma longibrachiatum endonuclease of enological relevance. J. Agric. Food Chem. 48, 951Ð957. Vogel, J.T., Tan, B.-C., McCarty, D.R. and Klee, H.J. (2008) The carotenoid cleavage dioxygenase 1 enzyme has broad substrate specificity, cleaving multiple carotenoids at two different bond positions. J. Biol. Chem. 283, 11364Ð11374. Wache,« Y., Laroche, C., Bergmark, K., M¿ller-Andersen, C., Aguedo, M., Le Dall, M.-T., Wang, H., Nicaud, J.-M. and Belin, J.-M. (2000) Involvement of acyl coenzyme A oxidase isozymes in biotransformation of methyl ricinoleate into ␥-decalactone by Yarrowia lipolytica. Appl. Environ. Microbiol. 66, 1233Ð 1236. Waddell, W.J., Cohen, S.M., Feron, V.J., Goodman, J.I., Marnett, L.J., Portoghese, P.S., Rietjens, I.M.C.M., Smith, R.L., Adams, T.B., Gavin, C.L., McGowen, M.M. and Williams, M.C. (2007) GRAS flavoring substances 24. Food Technol. (Chicago) 61, 22Ð24, 26Ð28, 30Ð49. Wang, C., Chin, C.-K., Ho, C.-T., Hwang, C.-F., Polashock, J.J. and Martin, C.E. (1996) Changes of fatty acids and fatty acid-derived flavor compounds by expressing the yeast ⌬ -9 desaturase gene in tomato. J. Agric. Food Chem. 44, 3399Ð3402. Wen, C.Y., Jen, L.Y. and Sum, H.L. (1999) Effect of water on continuous enzymatic synthesis of phenylethyl acetate in supercritical carbon dioxide. J. Chin. Agric. Chem. Soc. 37, 87Ð94. Whitehead, I.M. (1998) Challenges to biocatalysis from flavor chemistry. Food Technol. 52, 40Ð46. P1: SFK/UKS P2: SFK c04 BLBK221-Taylor/Linforth November 17, 2009 16:34 Printer Name: Yet to Come

126 Food Flavour Technology

Williams, A.S. (1999) The synthesis of macrocyclic musks. Synthesis 10, 1707Ð1724. Winterhalter, P. (1996) Carotenoid-derived aroma compounds: biogenetic and biotechnological aspects. In: Biotechnology for Improved Foods and Flavors (eds G.R. Takeoka, R. Teranishi, P.J. Williams and A. Kobayashi), ACS, Washington, DC, pp. 295Ð308. Winterhalter, P. and Skouroumounis, G.K. (1997) Glycoconjugated aroma compounds: occurrence, role and biotechnological transformation. Adv. Biochem. Eng. Biotechnol. 55, 73Ð105. Wu, S., Zorn, U., Krings, U. and Berger, R.G. (2007) Volatilesfrom submerged and surface cultured beefsteak fungus Fistulina hepatica. Flavour Frag. J. 22, 53Ð60. Xu, P.,Hua, D. and Ma, C. (2007) Microbial transformation of propenylbenzenes for natural flavor production. Trends Biotechnol. 25, 571Ð576. Xue, Z.-T. and Brodelius, P.E. (1998) Kinetin-induced caffeic acid O-methyltransferases in cell suspension cultures of Vanilla planifolia Andr. and isolation of caffeic acid O-methyltransferase cDNAs. Plant Physiol. Biochem. 36, 779Ð788. Yanai, T. and Sato, M. (1999) Isolation and properties of ␤-glucosidase produced by Debaryomyces hansenii and its application in winemaking. Am.J.Enol.Viticult.50, 231Ð235. Ye, X., Al-Babili, S., Kloti,¬ A., Zhang, J., Lucca, P.B.P. and Potrykus, I. (2000) Engineering the provitamin A(␤-carotene) biosynthetic pathway into (carotenoid-free) rice endosperm. Science 287, 303Ð305. Yoshida, A., Takenaka, Y., Tamaki, H., Frebort, I., Adachi, O. and Kumagai, H. (1997) Vanillin formation by microbial amine oxidases from vanillylamine. J. Ferment. Bioeng. 84, 603Ð605. Yvon, M., Amarita, F., Nardi, M., Chambellon, E., Delettre, J. and Bonnarme, P. (2006) Identification of the gene responsible for the synthesis of volatile sulfur compounds in Brevibacterium linens. Dev. Food Sci. 43, 49Ð52. Zorn, H., Fischer-Zorn, M. and Berger, R.G. (2003b) A labeling study to elucidate the biosynthesis of 4-(4- hydroxyphenyl)-butan-2-one (raspberry ketone) by Nidula niveo-tomentosa. Appl. Environm. Microbiol. 69, 367Ð372. Zorn, H., Langhoff, S., Scheibner, M. and Berger, R.G. (2003a) Cleavage of ␤,␤-carotene to flavor compounds by fungi. Appl. Microbiol. Biotechnol. 62, 331Ð336. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

5 Natural sources of ﬂavours Peter S.J. Cheetham

5.1 INTRODUCTION

This chapter attempts to provide examples of the biochemical basis for the wide variety of flavours obtained from plant sources. These include some of the more important sources of the flavour materials themselves and their precursors, which can be converted into flavours by subsequent cooking, enzymatic or microbial steps. There are many excellent articles on particular aspects of natural flavours Ð such as essential oils, citrus or fruit flavours Ð but this chapter tries to show the relationships between the source materials and processing methods, and the chemical compositions of the resulting flavours. The flavour characteristics of the resulting flavour extracts and the way they dictate the end uses and social and economic values of the flavours are discussed. To illustrate these effects, examples are included from a wide range of sources, from the wide variety of flavour chemicals contained in these sources and from the very varied organoleptic characteristics produced when they are consumed (for further reading see Fruit and Vegetable Flavour: Recent Advances and Future Prospects, Bruckner¬ and Wyllie, 2008). It is hoped that this chapter will be of interest to those involved in formulating natural flavour chemicals and natural extracts into complete flavours, to scientists interested in understanding the elusive relationship between the structures of flavour chemicals and their organoleptic sensations, and to those involved in developing new methods of making natural flavour chemicals or extracts to replace the traditional sources. The important principles and concepts are presented, along with some representative examples Ð such as for extraction technology (citrus), commercial background (vanilla) and isomeric effects on flavour character and thresholds of detection (menthols). Some of the key problem areas are described and some promising opportunities for the future are identified. In preparing this chapter, I have to acknowledge many sources of useful and interesting information, all of which are cited in the references. In introducing the subject of plant-derived flavour, it is important to consider what the benefits of flavours are to the end users and therefore what the challenges are for flavour suppliers if significant improvements are to be made (Alston, 1992). If improvements in one or more of the following benefits can be achieved then the flavour offered for sale will most probably be differentiated from similar products and a competitive advantage will be gained that could lead to its widespread use. These benefits include the following: r r Improved flavour quality, intensity, complexity, identity with fresh products, etc. r Enhanced product taste stability/shelf-life extension Extended ranges of use such as improved heat/bake stability P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

128 Food Flavour Technology r r Off-taste masking r Replacement of manufacturing/food processing and cooking procedures r Reduction/replacement of undesirables, e.g. peanut and monosodium glutamate (MSG) r Premium positioning such as natural, kosher, halal or organic certification And of course, cost reductions, for instance, that could result from greater flavour intensity so that the flavour can be used at a lower concentration in the food or beverage, so that the user benefits from a lower cost in use.

In addition, a rule of innovation for consumer products and services is that they evolve to providing something at a price that is affordable to ordinary people that was previously available only to a social elite, historically, the nobility, and now entertainment and sports stars and other celebrities. It is also worth considering what the key challenges are in trying to develop new processes to make natural flavours. These include the availability of natural raw materials at reasonable costs, quantities and reproducible qualities. If the required raw materials are not readily available then expensive and time-consuming backwards integration into raw material production has to be justified, such as for hydroxy acids production for lactone manufacture, for instance for ␥-hydroxycaprylic acid. Natural flavours have to have high aroma/flavour quality and so have to be sold in highly purified forms, especially to eliminate ‘off-notes’. Therefore, intensive and expensive product extraction and isolation processes (downstream processing), including for key intermediates on many occasions, and for the removal of materials introduced earlier in the process such as solvent residues, are invariably necessary. Thus, downstream processing adds high capital and operating costs, especially as promising new technologies such as in situ product recovery have not proved to be generally applicable. Natural flavours are required in only relatively small quantities and often only on an ad hoc basis. Thus, economies of scale are not gained. Another consequence is that multi-purpose/multi-product batch-wise manufacturing facilities are necessary to make a range of products using the same equipment, on only a semi-regular basis, which means that specialist equipment needed to make just a single new natural flavour chemical is difficult to justify. Also, continuous processes that can be more cost-effective cannot be done, genetic engineering is not currently acceptable, and proof of identity has to be obtained to obtain kosher/halal/organic certification, which may constrain the choice of raw materials, processing aids and process conditions. The growth of suitable species of plants as sources of natural flavours is still a very competitive technology for a number of reasons. For instance, agriculture is a relatively low-cost technology and because of the relatively high value of flavours their production can compete with most other crops for arable land despite the low gravimetric yields of flavours. However, the promise of plant cell culture and plant cell tissue culture as sources of natural flavour chemicals has not been fulfilled, due to, for instance, slow growth rates and susceptibility to microbial contamination. Plant sources of flavours are also competitive because bioprocessing is still a relatively limited technology as only a limited range of commercially available enzymes and whole cell biocatalysts is available so far. For instance, introducing oxygen into molecules and hydroxylation in particular is very important for making many natural flavours, but there are limited successful examples. Thus, across the whole of bioprocessing technology there is only quite a short list of successful hydroxylation processes, for example hydroxyproline, carnitine, 6-hydroxynicotinic acid, d-␤-hydroxy isobutyric acid and steroid hydroxylations (Cheetham, 2004). Also, despite the very wide-ranging metabolic capabilities of wild-type P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 129

micro-organisms, few are sufﬁciently high yielding to be the basis for cost-effective bioprocesses due to, for instance, precursor toxicity problems and poor precursor uptake by micro-organisms. Furthermore, there are still only poor screening capabilities for microbial strains with new biocatalytic capabilities, including a declining skill base in microbial metabolism and physiology, skills that are important in enabling effective screening.

5.2 PROPERTIES OF FLAVOUR MOLECULES

The chemical nature of a given flavour molecule determines its properties and how it behaves in a wide range of environments. Flavour perception is one example (Section 5.2.1), as is the effect of isomerism on perceived flavour quality (Section 5.2.2). Extraction of compounds from biological sources (Section 5.2.3) also utilises the properties of the molecules for the design of suitable extraction processes. Chemical properties determine the safety in use of flavour chemicals (Sections 5.2.4) as well as the ease with which they can be prepared and shipped, thus affecting the price and the economics of the whole industry (Section 5.2.4). Plants are important sources of not just flavours but also of flavour precursors, and flavour can be not just desirable but undesirable, creating off-tastes that can ruin the value of foods and beverages.

5.2.1 Flavour perception Flavour is a combination of sensory inputs derived from taste receptors in the tongue that detect basic sensations such as sweet, bitter, sour and probably the savoury umami sensation as well; together with specialised receptors such as for hotness, and with these sensations from the mouth combined with odour sensations from the nose, which are orders of magnitude more sensitive than taste and able to discriminate much more widely as regards the particular types of molecule that are encountered. In addition, the vomeronasal organ in the nose that in animals is responsible for sensing pheromones may make some contribution, and also flavour perception is influenced by numerous other factors such as the colour, temperature and texture of the food or beverage, together with psychological factors such as expectations and appetite. As is well known, it is difficult to properly taste foods if one is unable to smell them as well. In real life, as opposed to the laboratory, what is perceived is a complex mixture of flavour chemicals that are sensed over different time scales so that the overall flavour quality and intensity will vary with time. The conditions of cooking also influence the flavour composition of foods, for instance cooking dehydrates foods and liberates fats that can act as solvents, and, especially in combination with high cooking temperatures can allow flavour-generating reactions to occur that would never take place at low temperatures and in aqueous conditions.

5.2.1.1 Receptors and flavour sensing Tastants are molecules that interact with receptors in the taste buds of the mouth to produce sensations that are broadly classified as sweet, bitter, salt and sour, although now umami flavour (due specifically to MSG) is recognised as a fifth taste sensation. Other sensations in the mouth (and to some extent in the nose) are derived from interaction of chemicals with the trigeminal receptors, epitomised by the hot sensation of chilli. Aromas are volatile molecules P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

130 Food Flavour Technology

that are perceived by receptors in the nose so that their effects are dependent on their vapour pressures, temperature, etc. Taste and smell have probably evolved as a way to assess the quality and safety of foods. Unlike sight, hearing and touch, the senses of taste and smell have not yet been fully explained. Aroma and taste receptors have been isolated and analysed. They have common structural features, belonging to the transmembrane G-protein-coupled receptor family, that detect molecules and amplify the signal generated, via second messenger molecules such as cyclic adenosine monophosphate (AMP) and stimulating the olfactory neurone. The signalling process involves activation of the G-protein by the receptorÐodourant complex, which involves the exchange of bound guanosine monophosphate (GMP) for guanosine triphosphate (GTP), release of a subunit of the G-protein that in turn activates adenyl cyclase to produce cyclic AMP or some other similar secondary messenger molecules. Cyclic AMP activates a protein kinase, the kinase phosphorylates gated ion channels that allows ion movements across the membrane, that thus becomes depolarised and causes the release of neurotransmitter at the synapse with a sensory axon and the generation of an action potential in the axon. The action potential transmits the sensory information to glomerulus cells that integrate all the signals from similar receptor cell, and then the integrated signals are transmitted via mitral cells to the central nervous system. The transmembrane domains, especially domains 3 and 4, appear to be the site of interaction of the receptors with the flavour and aroma molecules that initiate the taste-signalling process. A number of types of olfactory receptor have been identified so far and any one receptor appears to be able to bind a small range of related molecules. These receptors interact with well-defined properties of the molecules eliciting the response. For example, sweet taste is due to the arrangement of certain functional groups on the molecule that form the glycophore, which requires closely associated proton donning, proton accepting and hydrophobic groups, in defined spatial positions relative to each other, that interact with complementary groups arranged in the sweetness receptor to initiate the perception of sweetness (the AH/B X system) (Shallenberger, 1996). Malnic et al. (1999) and Ngai et al. (1993) provided experimental evidence to support a combinatorial mechanism for aroma receptors, whereby each type of receptor can detect a range of related odourant molecules, probably to differing extents. Similarly, any one aroma molecule can bind to more than one type of receptor, probably with different affinities. Individual olfactory neurons appear to only express one or a few receptor genes so that the perception of taste depends only on which combination of olfactory neurons has been stimulated. A single flavour chemical will stimulate a small and well-defined number of receptors. A formulated flavour containing a mixture of flavour chemicals will stimulate a greater number of receptors, depending on how chemically complex the flavour is. Most food and beverages will contain a large number of quite different flavour molecules that will bind with a wide range of different receptors and hence stimulate even more complex combinations of olfactory neurons. Even then these signals undergo much processing on the way to the brain to give us our taste perception of whatever we are eating or drinking. This could explain the ways in which smells and tastes can be learned and offers explanations of how molecules as dissimilar as hydrogen cyanide and benzaldehyde can both be perceived as having an almond-like character, whereas iso- propylbenzaldehyde smells of cumin, quite unlike almonds. Human taste receptors for sweet, bitter, umami taste have been identified and characterised by Nelson et al. (2001), Li et al. (2002) and Ozeck et al. (2004). A notable recent advance has been made by Triballeau et al. (2008). They used a com- putational (virtual) high-throughput screening method to identify molecules with higher P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 131

affinities for an amino acid-binding olfactory GCP (G protein-coupled) receptor than the naturally occurring agonists for that receptor, such as lysine and arginine. The screening process involved attempting to dock the structures of the molecules with a computer model of the receptor. The screen began with a database of 1.6 million commercially available chemical structures, which was reduced in three stages to the 46 most suitable molecules that were then put through the receptor-docking procedure. The most active ‘new odourants’ found included diaminopimelic acid, which is a metabolite in the biosynthesis of lysine, l-canavaline, (S)-4-oxalysine and l-glutamic acid ␥-para-nitroanilide that was most active. The activities of the most active molecules identified were confirmed experimentally by electrophysiology using goldfish olfactory epithelium since excitatory neuronal responses were generated by all four of the above-mentioned molecules at concentrations down to ca 10−8 M. But using this in vitro test method the diaminopimelic acid was the most active molecule. In addition to identifying new odourants this study identified potential antagonists of the G protein-coupled receptor (GPCR) that bind to the receptor, but do not allow it to initiate a signal. This approach potentially gives a new method to enable superior flavour molecule structures to be identified. Another new method is the creation of transgenic animals that express just a single type of odourant receptor.

5.2.1.2 Thresholds Usually, the log of the intensity of a flavour chemical is directly proportional to the log of its concentration, although the character of the flavour that is perceived may change considerably with its concentration and in a whole food or beverage situation the concentration of flavour chemicals can be reduced through binding to proteins or polysaccharides. To be active as a tastant or as an aroma, the concentration of the molecule has to be above its taste or aroma threshold, which varies greatly from molecule to molecule (Table 5.1). The only exception is when mixtures of high-threshold materials are present, but in such cases the overall intensity is less than expected from the intensities of the individual materials. Odour thresholds can vary widely depending on comparatively minor changes in chemical structure. Thus, whereas vanillin has a good vanilla-like flavour, isovanillin is almost tasteless. Similarly, 2-methyl- 3-ethylypyrazine and 3-butyl-3-methoxypyrazine have thresholds of 130 and 0.001 ␮g/L, respectively, a 130 000-fold difference. Likewise, the intensity of a flavour is roughly related to its concentration in the food or beverage divided by its threshold for perception. Up to 10 000 different flavour molecules have been found in foods and beverages, of which about 10% contain sulfur. A particular food or beverage will usually contain a very large number of flavour and aroma chemicals. This may approach 1000 compounds, especially for materials such as coffee, chocolate or bread in which a range of molecules have been generated by fermentation, and then that number has been amplified by their degradation and recombination during thermal processing. Because the taste and aroma thresholds for some molecules are so low, even very minor chemical constituents of foods and beverages can make significant contributions to the flavour quality and character as the flavour impact of a chemical is the product of its concentration and the reciprocal of its threshold. For instance, total flavour molecules are commonly present at about 20 g/t of food, i.e. at 20 mg/kg. Assuming that there is a minimum of 20 individual flavour chemicals present, then any one will be present, on average, at a concentration of 1 mg/kg. Hence, it is easy to understand why comparatively few flavour chemicals are produced even on a tonne scale. If it is also assumed that any individual flavour molecule P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

132 Food Flavour Technology

Table 5.1 Odour and taste threshold values of terpenoid enantiomers in water in parts per billion (␮g/L).

Odour Taste Compound threshold threshold (S)-(+)-␣-Phellandrene (dill) 200 200 (R)-(−)-␣-Phellandrene (terpeny, medicinal) 500 (R)-(+)-Fenchone (fennel) 510 (S)-(-−)-Fenchone (fennel) 350 (R)-(+)-Carvone (caraway) 85–130 (S)-(−)-Carvone (spearmint) 2 (R)-(+)-Carvone 600 (S)-(−)-Carvone 43 (R)-(+)-(E)-␣-Ionone (fruity, raspberry-like) 0.5–5 (S)-(−)-(E)-␣-Ionone (woody, cedar-like) 20–40 (R)-(+)-(E)-␣-Damascone (woody, slightly fruity) 100 1 (S)-(−)-(E)-␣-Damascone (fruity, slightly woody) 1.5 1 (+)-cis-2-Methyl-4-propyl-1,3-oxathiane (sulfury, rubbery, tropical) 2 (−)-cis-2-Methyl-4-propyl-1,3-oxathiane (ﬂat, estery, camphory-like) 4 (+)-Nootkatone (grapefruit) 800 (−)-Nootkatone (woody) 600 000 − R-(+)-1-p-Menthene-8-thiol (grapefruit) 2 × 10 5 − S-(−)-1-p-Menthene-8-thiol (grapefruit) 8 × 10 5

From Boelens et al. (1993) with permission.

needs to be present at about ten times its threshold of perception to have a significant flavouring effect, then the flavour molecule has to have a threshold of perception of less than 100 ␮g/L. Flavours also vary in character and intensity depending on which isomer of the flavour chemical is used. Common examples are d- and l-carvones, which taste and smell of caraway and spearmint, respectively. Again, this effect suggests that interaction with receptors with three-dimensional binding sites must be involved (see Sell, 2001, for the views on receptor theories). In several cases, the quality and character of a flavour changes with its concentration, owing to different molecules having different affinities for the flavour receptors. This contrasts with many other biological effects, in which activity varies linearly depending on concentration. However, it must always be remembered that actual perception of flavour quality, character and intensity is very subjective and very dependent on prior consumer experience, so that large regional, cultural and age differences exist, even for some quite basic flavour sensations. Flavours with quite different taste sensations can (surprisingly) complement each other Ð for example gammon and pineapples, strawberries and cream. Also, molecules with quite different chemical structures can have very similar taste sensations, such as the sesquiterpene (R)-nootkatone and the thioterpenoid 1-p-menthene-8-thiol, which both taste of grapefruit. This phenomenon was first noticed for benzaldehyde and hydrogen cyanide that both give the impression of almonds.

5.2.1.3 Range of flavour compounds Some 2000Ð3000 different flavour chemicals are in use commercially. They are conventionally divided into sweet and savoury flavours. Plants are probably the major sources of sweet flavours and also some types of savoury flavour such as vegetable and nut flavours, but P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 133

Table 5.2 Odourants of cocoa mass: results of aroma extraction dilution analysis.

Compound Odour quality Flavour dilution factor 3-Methylbutanal Malty 1024 Ethyl-2-methylbutanoate Fruity 1024 Hexanal Green 512 Unknown Fruity, waxy 512 2-Methoxy-3-isopropylpyrazine Peasy, earthy 512 (E)-2-Octenal Fatty, waxy 512 Unknown Tallowy 512 2-Methyl-3-(methyldithio)furan Cooked meat-like 512 2-Ethyl-3,5-dimethylpyrazine Earthy, roasty 256 2,3-Diethyl-5-methylpyrazine Earthy, roasty 256 (E)-2-Nonenal Tallowy, green 256 Unknown Pungent, grassy 128 Unknown Sweet, waxy 128 Phenylacetaldehyde Honey-like 64 (Z)-4-Heptenal Biscuit-like 64 ␦-Octenolactone Sweet, coconut-like 64 ␦-Decalactone Sweet, peach-like 64

From Belitz and Grosch (1999) with permission.

generally plants are not a good source of roasted and meat-like flavour chemicals. The minor exceptions illustrate the types of flavour material that are generally unavailable from plant sources, such as 2-methyl-3-(methyldithio) furan that occurs in cocoa mass (see Table 5.2) and its disulfide which is present in black tea (Table 5.3), methyl(2-methyl-3-furyl) disulfide which is present in milk chocolate (Table 5.4) and methanethiol that has been identified in coffees (Table 5.5). Vegetable and cereal materials do of course generate savoury flavours when cooked at high temperatures especially in the presence of oil or fat, for instance fried and roasted potatoes and parsnip, fried rice and grilled peppers and tomatoes. In addition, plants are relatively poor sources of the precursors of meat-like flavour chemicals such as cysteine and methionine, although the key meat flavour precursor thiamine is present in reasonable quantities in some plant materials such as brown (unpolished rice). Typically, complete, well-rounded commercial flavours, made to flavour foods and beverages, contain

Table 5.3 Odourants of black tea: results of aroma extraction dilution analysis.

Compound Odour quality Flavour dilution factor (E)-␤-Damascenone Boiled apple 512 Linalool Floral 512 3-Hydroxy-4,5-dimethyl-2(5H)-furanone Seasoning-like 512 4-Hydroxy-2,5-dimethyl-3(2H)-furanone Caramel 512 3-Methyl-3,4-nonanedione Strawy, hay-like 256 Bis(2-methyl-3-furyl)disulﬁde Boiled meat-like 256 cis-4-Heptenal Biscuit-like 256 1-Octen-3-one Mushroom-like 128 Vanillin Vanilla 128 trans-2-trans-4-Decadienal Deep fried 64 2-Phenylethanol Floral 64 trans-2-Nonenal Fatty green 64

From Belitz and Grosch (1999) with permission. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

134 Food Flavour Technology

Table 5.4 Aroma extract analysis on milk chocolate.

Compound Odour, character Isovaleraldehyde Malty, sharp 2-Ethyl-3,5-dimethylpyrazine Potato chip 2-Methylbutyric acid Sweaty Isovaleric acid Sweaty 5-Methylhept-2-en-3-one Hazelnut 1-Octen-3-one Mushroom 2-Ethyl-3,6-dimethylpyrazine Nutty, earthy 2,3-Diethyl-5-methylpyrazine Potato chip trans-2-Nonenal Green, tallow trans-2-trans-4-Decadienal Fatty, waxy trans-2-trans-4-Nonadienal Fatty ␥-Decalactone Sweet, peach Methyl(2-methyl-3-furyl)disulﬁde Sulfurous, meaty

From Rowe and Tangel (1999) with permission.

at least 20Ð50 different flavour chemicals. Many more will often be present if plant extracts are used, as when essential oils are included in the formulation. Many flavour chemicals, including most fruit flavours and many vegetable flavours, are manufactured as pure single chemicals, indeed often as single isomers. Others, especially savoury flavour chemicals, such as reaction flavours, are made as mixtures, often referred to as flavour ‘blocks’. Flavours can

Table 5.5 Concentrations and odour activity values of potent odourants of brews prepared from Arabica and Robusta coffees.

Concentration (mg/kg) Odour activity valuea Odourant Arabica Robusta Arabica Robusta

2-Furfurylthiol 1.7 1.7 1.7 × 105 1.7 × 105 2-Ethyl-3,5-dimethylpyrazine 0.33 0.94 2.1 × 103 5.1 × 103 2,3-Diethyl-5-methylpyrazine 0.095 0.31 1.1 × 103 3.4 × 103 (E)-␤-damascenone 0.195 0.205 2.6 × 105 2.7 × 105 Methional 5.7 2.8 29 14 3-Mercapto-3-methylbutyl formate 5.5 4.3 1570 1230 Guaiacol 170 1230 68 490 4-Vinylguaiacol 1640 5380 82 270 4-Ethylguaiacol 51 635 1 13 Vanillin 220 740 9 30 4-Hydroxy-2,5-dimethyl-3(2H)-furanone 4510 2480 450 250 3-Hydroxy-4,5-dimethyl-2(5H)-furanone 77 31 257 103 5-Ethyl-4-hydroxy-5-methyl-3(2H)-furanone 8.7 4.4 1 Ͻ1 2-Ethyl-4-hydroxy-5-methyl-3(2H)-furanone 840 670 42 29 2,3-Butanedione 2750 2400 183 160 2,3-Pentanedione 1570 750 52 25 2-Isobutyl-3-methoxypyrazine 1.0 0.17 200 34 Propanal 435 435 44 44 Methylpropanal 800 1380 1140 1970 2-Methylbutanal 650 1300 500 1000 3-Methylbutanal 550 925 1570 2640 Methanethiol 210 600 1050 3000

From Semmelroch and Grosch (1996), with permission from the American Chemical Society. aThe odour activity values were calculated by dividing the concentration by the odour threshold value in water. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 135

also vary a great deal in price, from cheap yeast extracts to scarce and expensive specialities such as truffles. It is the flavourist’s role to formulate the various flavour chemicals, extracts, blocks, etc., to produce compounded flavours that reproduce the authentic flavour of the traditional products (see Chapter 1). Many flavour chemicals rely on one or more functional groups for their activity. However, the presence of a functional group is not essential, as hydrocarbons such as limonene have flavour. In many cases, particular chemicals are essential flavour constituents and, without them, a good flavour of the particular fruit or vegetable cannot be achieved. In some foods, the flavour quality is predominantly given by a single flavour chemical, the so-called ‘character- impact’ compounds, such as benzaldehyde for cherry flavours and vanillin for vanilla flavours. When used alone, these give the characteristic flavour of the product, although invariably the complete flavour as extracted from the source material Ð as cherry fruit or vanilla beans in these cases Ð is chemically and organoleptically more complex. Other flavour chemicals are used in combination with the character-impact compounds to create an overall impression. Flavour chemicals such as the C6 alcohols and aldehydes are fairly ubiquitous, contributing a ‘green’ fresh character to many fruit and vegetable flavours. In many situations, particular flavour chemicals can be used effectively to create flavours in which they do not naturally occur. An example is 4-hydroxy-2,5-dimethyl-3(2H)-furanone, which is widely used in both fruit and savoury flavours but is only found in a restricted number of sources such as strawberries, pineapples, beer, bread and coffee. These differences also hold for flavour and aroma extracts such as essential oils. Mint oil is chiefly used as a source of l-menthol, which has a prized combination of mint taste and cooling sensation, but other plant extracts are used only to supplement flavours, as in the use of buchu oil to supplement blackcurrant. This is analogous to the traditional use of herbs and spices to enhance and/or modify the taste of another food or beverage. It is also illustrative of the general approaches to making and using natural flavouring materials (Fig. 5.1). Work still continues to increase the range of natural flavours, and a key aspect of this is developing new improved analytical methods. For example, Ishikawa et al. (2004) en- closed plants in a container under simulated natural conditions, pumped humidified air in and analysed the exit air. This new method proved to be more effective than conventional

Fig. 5.1 General approach and speciﬁc example for production of ﬂavoured products. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

136 Food Flavour Technology

headspace analysis or solvent extraction methods, especially for the analysis of oxygenated components.

5.2.1.4 Character-impact flavours In many, but by no means all cases, just a single flavour chemical gives an impression that determines the perceived identity of any product containing it. An obvious example is vanillin that gives a vanilla flavour. Others include p-1-menthen-8-thiol (grapefruit), benzaldehyde (almonds), 1-p-hydroxyphenyl-3-butanone (raspberry) and 2-methoxy-4- methyl-4-butanethiol (blackcurrant) in fruits; 2-sec-butyl-3-methoxy pyrazine (carrots), 1- octen-3-ol (mushroom) and 2-trans-6-cis-nonadienal (cucumber) and 1,2-dithiocyclopentene (asparagus) in vegetables; 2-acetyl-1-pyrrolidone (roasted), 4-hydroxy-2,5-dimethyl-3(2H)- furanone (caramel) and trans-5-methyl-2-hepten-4-one (nutty) for more savoury flavours, and with other examples mentioned later in this chapter. A character-impact flavour chemical does not necessarily have to have a low threshold of perception; for instance, (S)- (+)-␣-phellandrene is the character-impact chemical of dill despite its having a rather high threshold of perception, because it is present in high concentrations and because no other flavour chemicals are present in high enough concentrations to compete with it. Next, there are many flavour chemicals that, although not absolutely distinctive of a particular product, are essential if a good flavour is to be formulated. Thus, in many cases the same flavour chemical can occur in quite different tasting foods. Watercress, paprika and coffee Ð all contain 2-isobutyl-3-methoxy pyrazine; octen-1-ol occurs not just in mushrooms but also in Camembert cheese and lenthionine is present in shiitake mushrooms and contributes to the characteristic taste of lamb meat, 2-methyl-3-methyldithio furan occurs in cooked meat and in chocolate, 3-thiohexanol and 3-thiohexyl acetate both occur in passion fruit and in Sauvignon wine, and 3-methoxy and 3-isopropyl pyrazines both occur in bell peppers and Sauvignon wine. In addition, some other constituents not normally thought of as flavours can contribute to creating a distinctive taste, such as citric acid present in lemons and oranges, tartaric acid that is present in grapes and isocitric acid that is present in blackberries. Then, as a third tier, there are a host of flavour chemicals that are useful to improve the quality and value of flavour and to impart more subtle variations in character.

5.2.1.5 Consumption indices The proportion of any flavour consumed as present in a natural source divided by that present from deliberate addition to the food or beverage is termed the consumption ratio. Thus, vanillin, which is mostly consumed as chemically synthesised material, rather than as present in vanilla bean extracts, has a consumption ratio of 0.02, whereas methyl 2-pyrrolyl ketone, which is almost exclusively consumed as a flavour molecule in roast beef, has a consumption ratio of 2807 and so is called a food predominant-flavour chemical. The various isomers of methoxy isopropylpyrazine are consumed fairly evenly in the forms of tomatoes and added flavour, and so has a consumption ratio of 3.5 (Stofberg, 1983).

5.2.1.6 Flavour quality Flavour quality is comparatively easy to assess, although the detection and identiﬁcation of individual ﬂavour molecules present at very low concentrations can be challenging (see also P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 137

Table 5.6 Comparison of the perception threshold values of C6 ﬂavour aldehydes in different media.

Air Water Vegetable Mineral oil Water Milk Aldehyde (mg/m3) (ppm) oil (ppm) (ppm) (ppm) (ppm) Hexanal 0.053 0.0045 0.12 0.32 0.076 0.05 trans-2-Hexenal 0.220 0.017 — 10.0 0.45 0.067 trans-2-trans-4- 0.0018 0.01 — 0.27 —— Hexadienal trans-3-Hexenal — — — 1.2 —— cis-3-Hexenal 0.0022 — — 0.11 —— trans-4-Hexenal 0.0026 — — ——

Chapter 9). In most cases, the concentration of any single flavour chemical in the end food or beverage product rarely exceeds 10Ð50 ppm (that is 0.01Ð0.05 g/L). The flavour may be supplied diluted in a food-acceptable solvent such as propylene glycol or ethanol; with a bulking agent, such as maltodextrins; or even microencapsulated, such as in cyclodextrins. Flavour molecules can also have some other functional effects, contributing desirable colour or antimicrobial, or antioxidant activities such as possessed by rosemary extracts, as well as some uses as fragrances. The perceived quality and intensity of a flavour depends on the environment in which it is presented, for instance its temperature or pH. Table 5.6 illustrates the way in which the aroma and taste thresholds for a particular molecule vary, depending on whether it is presented to the nose or mouth dissolved in water, or in vegetable or mineral oil, or in air. These data show how the perception thresholds of various C6 flavour molecules differ depending on the matrix in which they are presented. The effect of protein solutions on flavour release is illustrated by Fig. 5.2, which shows that as the protein concentration increases,

Fig. 5.2 The inﬂuence of native proteins on the headspacer concentrations of allyl isothiocyanate (solid curves) and diacetyl (dashed lines) in water. , egg white; , bovine serum albumin; , casein. From Land (1994) with permission. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

138 Food Flavour Technology

Table 5.7 Comparison of high- and low-fat ice cream vanilla formulations.

Ingredient Reformulation factor Log P Vanillin 0.75 0.40 Phenol 0.50 0.89 p-Cresol 0.30 1.28 4-Ethylguaiacol 0.13 1.74 Eugenol 0.08 1.94 Ethyl benzoate 0.03 2.64 Methyl cinnamate 0.02 2.79 Anethole 0.01 3.33

the headspace concentrations of both allyl isothiocyanate and diacetyl decrease, and that different proteins have different effects, presumably because they bind the flavour molecules to different extents. A particularly big influence comes from the fat content of the food, because of the ability of flavours to partition between the water and fat phases to different extents depending on their lipophilicity/hydrophilicity, requiring rebalancing of the flavour formulations to give the taste required by consumers. Table 5.7 shows this effect for vanilla ice cream flavour, with the flavour reformulation factor being that required to produce the same initial flavour intensity in 0% fat ice cream as in 15% fat ice cream. In addition to their flavour characteristics, some essential oils and flavour chemicals possess some antimicrobial or antioxidant properties. However, these are generally not powerful enough at the low concentrations flavours are used at in products to make a very useful contribution to ensuring the stability of food and beverage products.

5.2.1.7 Changes on storage Changes in the chemical composition (and thus in taste and smell quality) frequently occur on storage. Sometimes these changes are desirable, as in the maturation of fine wines and spirits, but they can often be negative (Sinki et al., 1997). Changes can be due to interactions between components, such as esterification, aldol condensation, acetal formation, oxidation or reduction. For instance, a frequent occurrence is the lipoxygenase-mediated breakdown of unsaturated fatty acid into aldehyde and alcohol fragments with undesirable flavours. Oxidation of aldehydes to acids and of thiols to disulfides and the reaction of aldehydes to form acetals are other common mechanisms of flavour degradation. Changes can also be due to the processing conditions or to irradiation, microbial contamination or air oxidation. Protection is sometimes possible and cost-effective by encapsulating the flavour, for instance in cyclodextrins or gum arabic.

5.2.1.8 Off-flavours Off-flavours can greatly diminish the edibility and value of foods and beverages. Human beings probably find some flavours so objectionable because they are associated with the presence of toxins or contaminants and so have arisen during evolution as a food safety warning system. Positive or negative perceptions are however dependent on the particular food or beverage being tasted, as some flavour chemicals are perceived as positive in some products but negative in others. Well-known examples include the ‘beany’ off-taste of some P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 139

soy products that is due to the presence of aldehydes and the off-taste in beers, referred to as ‘sunstruck-beer’ caused by the formation of prenyl mercaptan (3-methyl-2-butene-1-thiol), but which is actually a desirable contributor to the flavour of coffee. Similarly, 3-mercapto- 3-methylbutyl formate is desirable in coffee but is undesirable when it is produced in beer. trans-2-trans-4-Decadienal is a major contributor to the flavour of chickens, but of course is considered an off-flavour when formed in potatoes. Methional is a desirable flavour component of many meat, fried and savour products and also of potatoes, but can be formed as an off-tasting flavour orange juice, or in milk where it is formed by a photo-oxidation process with riboflavin acting as the photosensitiser. Other problems in orange juices arise from the formation of 2-methoxy-4-vinylphenol (vinyl guaiacol) or of 2-methyl-3-furan thiol, which are desirable flavours when present in coffee and in beef products, respectively. Similarly, d-limonene is a well-accepted contributor to the flavour of citrus products, but can be oxidised to carvone in orange juices creating an off-flavour. As expected methods have been developed for preventing the formation of off-flavours or of removing them once they have formed. For example, immobilised lactase purified from Trametes hirsuta significantly reduced the amounts of the off-flavours guaiacol and 2,6-dibromophenol produced in apple juice by contaminating micro-organisms such as Acti- nomycetes, Alicyclobacillus and Clostridium (Schroeder et al., 2008).

5.2.1.9 Flavour precursors The natural chemicals that have absolutely no flavour can be very valuable as flavour precursors, with the flavour being created either in a defined and controlled manufacturing operation or taking place during the preparation or cooking of food. Unsaturated fatty acids from oils and triglycerides can be oxidised to form aldehydes and corresponding alcohols, amino acids from proteins can form aldehydes by Strecker degradation, and sugars from polysaccharides can undergo Maillard reactions to give flavours. Thus, rhamnose derived by enzyme treatment of naringin and hesperidin present in citrus materials is a very important raw material for the manufacture of natural Furaneol. Carotenoids form ionones, damascones and damascenones and cysteine, methionine and other biochemicals are sources of sulfur for incorporation into flavours. When the coenzyme thiamine is subject to cooking processes, it forms 2-methyl-3-furanthiol and its disulfide that are very powerful meat-beef tasting flavour chemicals. This is an important process despite the concentrations of thiamine being very low in most living tissues, because the 2-methyl-3-furanthiol and its disulfide have such low flavour thresholds of 7 × 10−3 and 2 × 10−8 ␮g/L, respectively. In fact, most vegetable flavours do not exist in the living vegetable plant material, but are formed from precursors by the action of endogenous enzymes liberated during the cutting up or mastication of the vegetables. One precursor can also give rise to flavour chemicals with quite different taste characteristics as a result of metabolism in different ways in different plants. Thus, a common route of metabolising linoleic acid is by the formation of its coenzyme A ester and then three cycles of ␤-oxidation. In pears further metabolism is by isomerisation, ␤-oxidation, desaturation and then esterification with ethanol to form 2-trans-4-cis-decadienoic acid ethyl ester, whereas in dairy products further metabolism is by hydration and lactonisation to form cis-6-dodecen-␥-lactone. Then, again in some fruits and vegetables such as apples and celery further metabolism is by lipoxygenase, non-enzymic oxidation and then decarboxylation to form 1,3-trans-5-cis-undecatriene, which is incidentally a good example of a hydrocarbon, P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

140 Food Flavour Technology

with no functional groups, that has ﬂavour properties. Finally, sucrose not just is an important sweetener but is also the raw material in the manufacture of the high-intensity sweetener SucraloseTM (4,1,6-trichlorogalactosucrose), which is 650 times as sweet as sucrose.

5.2.1.10 Flavour enhancers Flavour enhancers boost the intensity of other ﬂavours. The most common is the savour ﬂavour enhancer MSG. Most of the food industries usage is produced as a pure crystalline form by fermentation, but MSG is also present in some plant protein hydrolysates.

5.2.1.11 Plants as sources of enzymes that produce flavours Whereas most commercial enzymes are from microbial sources, plants are the source of some enzymes that are important in producing some flavours, including from plant-derived raw materials. Examples are the use of the soy bean lipoxygenase to make the 13-hydroperoxide intermediate of hexenal flavour molecules, lipoxygenase and hydroperoxide lyase enzymes that are present in mushroom stalks to produce 1-octen-3-ol from linoleic acid, and almond meal as a source of the ␤-glucosidase and mandelonitrile lyase enzymes required to make benzaldehyde from the mandelonitrile glucoside that is present in some plant materials, such as cherry stone kernels.

5.2.1.12 Historical development of food flavourings Originally, the only flavourings available were essential oils and other natural extracts, which were quite variable depending on their source and processing variations. Supply of these commodities was obviously limited in terms of both availability and quality, especially as the size of markets increased in the nineteenth century with the growth of modern consumerism. Then, as a result of advances in chemical analysis and synthetic organic chemistry, came an increasing number of nature-identical and synthetic flavour chemicals, which allowed improved fidelity to the original materials and greater flavour intensity, stability and reproducibility. This is reflected in a comparison of modern and traditional flavour formulations for cherry (Tyrell, 1995) (Table 5.8). Around 1985, a range of natural flavour chemicals produced by enzymic, microbial or mild chemical processes became available, so that good-quality flavour formulations can be made whose compositions closely resemble the analysis of natural extracts of the fruit. A big advantage of this approach is that a range of new raw materials can be utilised that may be cheaper and more available than the traditional raw materials used to manufacture flavours, in the same way that corn starch has become a cheap and large-scale raw material for glucose and fructose sweeteners. The main challenge in this approach remains the development of high-yielding and cost-effective processes. This is because, despite the variety of microbially produced flavours that occur in foods and beverages, these generally only form quite low concentrations of product, otherwise the product would be inedible. However, these flavour-producing strains are only rarely productive enough for use in the manufacturing of concentrated flavours. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 141

Table 5.8 Comparison of typical components of traditional and modern natural cherry ﬂavours.

Traditional cherry-type WONF fortiﬁer Almond bitter oil Ethyl alcohol Cinnamon bark oil Geranium oil bourbon Cognac green oil Neroli oil Coriander oil Rose oil Clove bud oil Natural black cherry WONF Acetaldehyde Coffee essence Acetic acid Davana oil Isoamyl acetate Ethyl acetate Benzaldehyde Ethyl alcohol Benzyl acetate Ethyl benzoate Benzyl alcohol Ethyl butyrate Butyric acid cis-3-Hexenal Caproic acid cis-3-Hexenol Capric acid Maltol Cherry essence concentrate

From Tyrell (1995) with permission. WONF, with other natural ﬂavourings.

5.2.2 Differences in sensory character and intensity between isomers In many cases particular isomers of flavour chemicals are formed in very pure forms. Thus, both R(−)-1-octen-3-ol and R(+)-trans-␣-ionone occur with enantiomeric excesses of over 90%. Marked differences in aroma/flavour characteristics and intensity (as assessed by threshold values) are common for isomers (Koppenhoefer et al., 1994). For instance, l- and d-carvones have spearmint and caraway flavours, respectively, and out of the eight menthol isomers (due to its having three chiral centres) and analogues, such as menthones and menthyl acetates, only one has the prized combination of good mint flavour plus cooling sensations (Table 5.9). Clark (1998) has reviewed menthol flavour, while Benn (1998) has given details of the sniff-GC and aroma extract dilution analysis of mint oils. That isomers really do have quite different flavour characteristics was irrefutably demonstrated by interconverting the (R)-(+) and (S)-(−) isomers of limonene and showing that the characteristic flavours of each isomer were the same, irrespective of whether they had been isolated or synthesised de novo. Boelens et al. (1993) thoroughly reviewed the information and concluded that three main categories of enantiomeric difference can be identified (see Table 5.1): r When sensory properties of the two enantiomers differ only slightly in intensity or in r quality (e.g. terpenoid hydrocarbons) When the enantiomers have the same main character but differ in secondary notes and r intensity (such as for aliphatic and monoterpene alcohols) When the odours of the enantiomers differ in both quality and intensity (e.g. carvones and nootkatones)

The effects of enantiomer type on odour threshold and odour character are shown in Table 5.1. These differences are quite common. For instance, the (R)-(+) and (S)-(−)-enantiomers P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

142 Food Flavour Technology

Table 5.9 Flavour, cooling and bitter threshold values of enantiomers of isomers of menthol, methone and menthyl acetates in ppm (␮g/L) in 5% sucrose solution.

Flavour threshold Cooling threshold Bitter threshold

L-Menthol 0.4 0.8 10–20 D-Menthol 0.3 3 20–30 D,L-Menthol 0.4 1–2 10–20 L-Isomenthol 0.6 30 50 D-Isomenthol 0.7 7 20 D,L-Isomenthol 0.6 10 20–30 L-Neomenthol 0.65 20–30 30 D-Neomenthol 0.5 3 20–30 D,L-Neomenthol 0.4 6 20–30 L-Neoisomenthol 1.0 6 Ͼ25 D-Neoisomenthol 0.2 Ͼ25 Ͼ25 D,L-Neoisomenthol 0.4 Ͼ25 Ͼ25 L-Menthyl acetate 0.4 3 10–20 D-Menthyl acetate 0.8 20–30 20–30 L-Menthone 0.01 0.1–0.2 50 D-Menthone 0.015 2–3 20–30 L-Isomenthone 0.035 0.8 4 D-Isomenthone 0.15 1–2 5

From Boelens et al. (1993) with permission.

of carvone show the well-known caraway and spearmint character, while the (R)-(+)-(E) and (S)-(−)-(E) enantiomers of ␣-damascone exhibit both fruity and woody notes, with the former enantiomer more fruity than woody while the latter shows the reverse (Table 5.1).

5.2.3 Extraction of flavours from plant materials Nearly all the anatomical parts of plants can be consumed as foods and can be the source of flavours. Just for vegetables these include stem tubers (potatoes), swollen tap roots (carrots), swollen hypocotyls (beetroots), bulbs (onions), axillary buds (brussel sprouts), flower buds (artichoke), swollen inflorescences (broccoli), stem sprouts (asparagus), main buds (lettuce), petioles (celery), leaf blades (spinach) and swollen leaf bases (leek). This variety helps to explain the range of different flavour chemicals present in our diet, such as methional in potatoes, carrotol in carrots and geosmin in beetroot, that give them their individual flavours and that can be obtained by the processing of agricultural products (Naf and Velluz, 2000). Figures 5.3 and 5.4 show the general approaches adopted. For plant sources, the tissue must usually be disrupted by mechanical, thermal or enzymic methods to allow the extraction of flavour and aroma materials in good yields. In some cases the required molecules are evenly distributed throughout the plant material, while in others they are present in specialised structures, such as the oil sacs that hold mint oils, present on the underside of the mint leaves. Plant cell walls are in three layers, the middle lamella and the primary and secondary cell walls. The middle lamella binds between cells and is mostly composed of pectin. The primary cell wall consists of cellulose fibres, together with pectins, hemicelluloses and proteins. The secondary cell wall contains lignin and pectin. It is these structures that must be disrupted to release flavour and aroma chemicals contained with the cells. Typical overall compositions of the cell walls are shown in Table 5.10. Enzymes commonly used in the manufacture of P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 143

Fig. 5.3 General approaches to production of plant extract ingredients and products.

plant extracts include polygalacturonidase, pectin and pectate lyases and esterases, cellulases and hemicellulases. There are four main ways of producing ﬂavour materials:

1. By direct extraction from the natural source, very often by steam distillation. Other methods are preferred for some materials; for instance, cold expression is the method of choice for obtaining citrus oils. Depending on the method used, various terms are used for the extract, including essential oil, absolute, extract, resinoid and oleoresins. Examples include coffee, cocoa and hop extracts, whereas benzoin and chicle gum are resinoids; oleoresins include many spices, such as pepper, capsicum, ginger or paprika. These spice oleoresins are particularly useful because they contain many of the non-volatile components not

Plant materials

Non-polar

Fig. 5.4 Traditional methods for the production of various aroma/ﬂavour products (underlined) by the extraction of plant materials. 1Absorption into fat. 2Concretes contain the volatile aroma chemicals, but also waxes. Concretes are usually used as intermediates. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

144 Food Flavour Technology

Table 5.10 Major components of the cell walls of some important fruit sources of ﬂavour.

Fresh weight of cell (%) Pectin Hemicellulose Cellulose Glycoprotein Cherries 0.51 0.06 0.17 0.32 Pineapples 0.21 0.35 0.27 0.12 Mangoes 1.02 0.23 0.59 0.32 Apples 0.54 0.34 0.70 0.15 Pears 0.42 0.22 0.40 0.12

present in the corresponding essential oils. Classification of a flavouring material is not always straightforward; for instance, vanilla extract is, technically speaking, really an oleoresin. Specific flavour chemicals may then be isolated in a pure form on a laboratory scale using the LikensÐNickerson apparatus that enables simultaneous extraction and distillation (see also Chapter 9), or on a larger scale once extracted as above, by fractional distillation. For example, Naf et al. (1990) used the LikensÐNickerson apparatus to isolate a precursor of ␤-damascenone from the leaves of Lycium halimifolium Mil (Solanaceae). 2. Compounded flavours are usually complex mixtures of chemically or naturally synthesised flavour molecules, often also incorporating extracts as produced above, and with the chemically or biochemically synthesised chemicals usually based on naturally occurring flavour chemicals discovered by careful chemical analysis. There is also a category of semi-compounded flavours in which flavour chemicals are formulated with fruit juices and/or sugar syrups for beverages or dairy products, or with spices, HVP (hydrolysed vegetable protein), MSG (monosodium glutamate) or salt for savoury foods. 3. Reaction flavours are produced by compounding appropriate precursor molecules, usually various sugars, amino acid sources and sulfur-containing compounds, and then heating or cooking to accelerate chemical reactions, such as the Maillard reaction (see also Chapter 3), that form the required mixture of flavour materials. Coffee and tea are examples. The strategy used in the past seems to have been to select materials with complex flavours together with another functional benefit, such as the stimulant properties of caffeine and theobromines in coffee and tea, respectively. The same strategy also applies to beer and wine, and to dairy products such as cheeses and yoghurts that in addition to their complex flavours are in one sense long storage life forms of milk, that only remains palatable for a very short period by comparison. 4. Enzyme and/or fermentation reactions can be used to produce flavours. These could be traditional, such as the conversion of wine or other sources of ethanol into vinegar, which is principally acetic acid. Modern processes tend to focus on the formulation of individual flavour chemicals such as decalactones and vanillin, but can also produce flavour blocks such as enzyme-modified cheese flavours. These can be extremely expensive when first introduced into the market, but prices usually soon fall as competing suppliers emerge (Fig. 5.5).

Irrespective of the methods used, the prime goal is taste quality, followed by considerations of cost, health, nutrition, naturalness and convenience, such as for microwaved foods, and considerations of novelty, as in flavours for exotic foods. New sources of aroma and flavour compounds are consistently being sought. One highly innovative approach is to search for them in flowers in the canopy of rainforests (McGee and Purzycki, 1999). P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 145

Fig. 5.5 Changes in the market price of natural ␥-decalactone. From Hausler¨ and Munch¨ (1998).

Quite recently, a new factor affecting the ease of extraction of flavours from plant sources has been discovered, and its exploitation has allowed very good improvements in yields. Many flavour molecules are present in their plant material of origin as glycosides. For instance, damascenone has been shown to be present as its ␤-glucoside in Lycium halimolium Mil. (Solanaceae) (Naf et al., 1990). Similarly, tea flavour chemicals have been shown to be present in fresh tea leaf in glycosidically bound forms (Wang et al., 2000); these include hexenol, benzyl alcohol, 2-phenylethanol, methylsalicylate, geraniol, linalool and four isomers of linalool oxide. These tend to be released by the action of endogenous glycosidases during tea processing, as well as by enzymes breaking up the plant matrix to make the glycosides more accessible and to facilitate subsequent extraction of the liberated flavour molecules. A comprehensive review of the occurrence of glycosidically bound plant C13 norisoprenoids as flavour precursors has been made by Winterhalter and Schreier (1994). In particular, the use of enzymes to release glycosidically bound flavours of passion fruit has been described by Chassagne et al. (1995). A more recently discovered method that plants use for binding flavour molecules is as cysteine-sulfur conjugates, such as the passion fruit flavour chemicals identified by Tominaga and Dubourdieu (2000). In this study 3-mercaptohexan-1-ol and 3-mercapto- 3-methylbutan-1-ol were produced from non-volatile fruit extracts following treatment with a cell-free extract of Eubactrium limosum as a source of ␤-carbon-sulfur lyase enzyme activity. Secondly, the S-(3-hexan-1-ol)-l-cysteine precursor was identified in the fruit by gas liquid chromatography/mass spectroscopy (GC/MS) analysis. Subsequently, a high- pressure/performance chromatography/mass spectroscopy (HPLC/MS) assay based on stable isotope dilution and one-step sample preparation has been developed to improve the detection and quantification of cysteine-sulfur conjugates involving the synthesis of deuter- ated S-3-hexan-1-ol-cysteine, which is hydrolysed by yeast ␤-carbon-sulfur lyase to release the volatile thiol 3-mercaptohexan-1-ol (Lusier et al., 2008). The changes that take place in a plant material following harvesting also affect extraction methods and efficiency, owing to the biochemical processes initiated by the trauma of harvesting. Either increases or decreases in the desired materials can result. Table 5.11 shows these effects for ␥-decalactone in strawberries and peaches. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

146 Food Flavour Technology

Table 5.11 Comparative analysis of ␥-decalactone in the headspace volatiles of living and picked fruits.

Living fruit, Picked fruit, Picked as Fruit ␥-decalactone (%) ␥-decalactone (%) percentage of living Ripe New Jersey strawberry 9.5 0.3 3 Ripe New Jersey cling peach 2.5 39.2 1570

Adapted from Mookherjee and Wilson (1990) with permission.

Whereas most fruit flavours occur preformed in the plant tissue, most vegetable flavour molecules are formed by enzymic action once the plant tissue has been disrupted, and in addition many flavours are formed by decomposition of precursors, especially by heating. For instance, heating sugars causes isomerisation and dehydration reactions to form flavour molecules such as hydroxymethylfurfural, acrolein (propenal) and 4-hydroxy-2,5-

dimethylfuran-3-one. Polymerisation of such sugar degradation products forms the brown colours characteristic of caramel, beer, bread and toast.

5.2.4 Commercial aspects The worldwide mercantile sales of flavours are substantial, estimated at about US$4.5 billion in 1994 (Hartmann, 1995), which excludes the considerable internal production and captive use of flavours by many companies. Sales were growing at 6Ð7% per annum, with 90% of sales in the big developed areas and 40% of sales in Europe. Sales of flavours are greatest for use in beverages, followed by flavours for savoury products (35 and 26% of total sales, respectively). The major consumer trends include convenience, freshness and ethnic and slimming foods or beverages. The largest producers of pure or semi-pure flavour chemicals are the specialised flavour and fragrance companies such as IFF, Givaudan, Firmenich, Symrise and Takasago. Their competitive advantages lie in their proprietary know-how in manufacturing and formulating flavour materials to produce more complex, value-added flavours. Over the last 20 years there has been a strong business trend towards globalisation, which has resulted in the consolidation of the flavour and fragrance industry, with the acquisition of many companies and the broadening of the range of products each large company can provide. Globalisation and consolidation further down the supply chain have been considerable among food manufacturers, who are the customers for flavours and aroma chemicals. To take just one example, the Pepsico/Frito-Lay group (which includes, for instance, Walkers Crisps in the UK) now has a 40% share of the world sales of savoury snack products. Next is Procter and Gamble (including the Pringles brand), which holds just 5% of the world market. This makes Pepsi/Frito-Lay the only truly global company in this market, and relegates all the others to somewhat regional and/or niche market positions. The biggest producers of more complex flavour extracts such as essential oils are usually traditional companies, with the biggest concentration present in the Grasse region of Southern France and many others located in tropical countries such as India and Indonesia. Their competitive advantages are in specialised cultivation, extraction and blending techniques. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 147

5.2.5 Economic aspects As regards the economics of the production of flavours from plant sources there is a paradox as in principle plants are easy and cheap to grow and there is an enormous diversity of plants many of which contain distinctive flavour materials, especially as well over 10 000 secondary metabolites have been found in plants. However, less than 20% of plant secondary metabolites have been chemically characterised, only a small proportion of the plants known to humans are easily intensively cultivated and only a proportion of these give economic yields of flavours, taking into consideration the low concentrations of flavour compounds present in plant materials and the substantial costs of extracting and purifying them to an acceptable quality, as well as other factors required such as resistance to pest damage and competition from weeds, post-harvest stability and ease of plant breeding to improve yields. So, in practice, the vast majority of plant-derived flavour materials are sourced from a surprisingly small proportion of the earth’s biodiversity.

5.2.6 Safety aspects Safety is an important aspect of flavour manufacture. In one sense, flavours offer a comparatively low risk because they are used at such low concentrations in most cases, and, if too high a concentration were to be used, the resulting food or beverage would be inedibly strongly flavoured. Thus, the use of flavours is, in practice, self-limiting. One new approach to ensuring the safety of foods is through the control and monitoring of the production processes using hazard analysis by critical control point (Ropkins and Beck, 2000). In addition, there are the solvents, bulking agents and carriers that usually do not contribute to flavour but serve to stabilise the flavour or make it easier to use. The legal framework surrounding flavour usage is discussed in Chapter 2 and the article by Somogyi (1996) contains further, relevant information.

5.3 DAIRY FLAVOURS

5.3.1 Background Dairy flavours are particularly interesting because of the wide range of products in which they are found, such as beverages, yoghurts, butters, spreads and cheeses, and because of the wide range of flavour chemicals involved. Raw milk has a very mild but highly complex flavour due to the presence of hundreds of different flavour chemicals, as well as its characteristic mouthfeel due to fat globules suspended as a colloidal suspension. The chemicals that make the biggest contribution to the flavour of raw and gently pasteurised milk are dimethyl sulfide formed from methionine, diacetyl made from citric acid, 2-methylbutanol, 4-cis-heptenal, 3- butenyl isothiocyanate, 2-trans-nonenal, ethanol, 2- and 3-methylbutanals, 4-pentenenitrile, 2-hexanone, hexanal, ethyl butyrate, 2,4-dithiapentane, heptanal, benzonitrile, 1-octen-3-one, 1-octen-3-ol, nonanal, p-cresol, 2-trans,4-trans-nonadienal and ␦-decalactones and dode- calactones, in rough order of contribution to the flavour. Sweetness is conferred by lactose, acidity by organic acids such as citrate; milk salts also have an effect. Low-temperature pas- teurisation has little effect on milk flavour. The use of higher temperatures introduces some sulfur flavours, such as hydrogen sulfide produced by the decomposition of milk proteins, and also some methylketones produced by thermal decarboxylation of ␤-keto acids and P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

148 Food Flavour Technology

␥- and ␦-lactones (formed from hydroxy fatty acids originally present as glycerides in the raw milk and released by heating). The effects of thermal processing are very marked in ultra-high temperature (UHT) milk, which has a distinct cooked flavour. This is due to disulfide exchange reactions between the milk proteins. Attempts have been made to reverse this reaction using the enzyme disulfide isomerase. Other off-flavours can be due to the photo-oxidation of methionine to methional, and by the action of microbial contaminants, such as the conversion of phenylalanine into phenylacetaldehyde and 2-phenylethanol by Streptomyces lactis var. maltigenes.Themost important contributors to flavour are 2-heptanone, 2-nonanone, dimethyl sulfide, diacetyl, 4-cis-heptenal, ␦-dodecalactone, hydrogen sulfide, 3-methylbutanal, dimethyl disulfide, hexanal, 2,3,4-trithiopentane, 2-trans-nonenal, 2-undecanone, ␦-decalactone and ␥-decalactone. The most extreme example of heating on milk flavour is in sterilised milk and condensed milk products, where temperatures have been high enough for Maillard flavours to be formed, such as by reaction between lactose and amino acids, to produce maltol.

5.3.2 Cream and butter Milk-derived products also offer a wide range of flavours. Cream and butter flavour depends greatly on free fatty acids and ␦-lactones. For instance, cis-6-dodecene-␥-lactone is an important flavour component of butter. It is formed from linolenic acid by ␤-oxidation followed by hydration, hydrolysis of the coenzyme A ester and subsequent lactonisation. Sour milk products such as yoghurts depend on metabolites of lactic acid bacteria, such as diacetyl and butanediol, lactic acid and ethanol, that are formed by the metabolism of the citric acid originally present in raw milk. The composition of flavour chemicals in a sweet cream butter is given in Table 5.12.

Table 5.12 Formulation of sweet cream butter aroma.

Concentration (mg/kg Compound of sweet cream butter) Diacetyl 4 3-Methylbutanal 0.01 4-cis-Heptenal 0.006 2-Phenylethanal 0.002 Acetic acid 57 Valeric acid 0.15 Phenol 0.01 p-Cresol 0.005 Guaiacol 0.002 Ethyl butyrate 0.002 ␥-Decalactone 3 ␦-Decalactone 3.6 ␦-Dodecalactone 6.9 Hydrogen sulﬁde 0.01 Methylthiol 0.01 Dimethyl sulﬁde 0.06 Indole 0.006 Skatole 0.67 Sodium glutamate 1.0

Data from Boelens and van Gemert (1993) with permission. pH 4.6 adjusted by lactic acid. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 149

Dairy products are often used as ingredients in other foods to modify the flavour or texture of a food or beverage. One example that has been studied is the addition of cream to heated raspberries, which resulted in a marked decrease in the flavour impact of the most intense flavour chemicals present in the raspberries, ␤-damascenone, vanillin, sotolone and 1-nonen-3-one and hydroxyphenylbutanone (raspberry ketone).

5.3.3 Cheese Although derived from milk, cheese flavour is principally determined by complex and individual processing methods (Eaton, 1994). Although hundreds of different cheeses are sold, it has been estimated that there are only 18 basic cheese types. The most important factor is the flavour production by micro-organisms present in and/or on the cheese. A good example is the formation of the ␦-lactone flavours, shown below.

Free fatty acids make a great contribution to cheese flavours, especially in Parmesan and Romano cheeses. Whereas long-chain fatty acids make little contribution, short-chain fatty acids (such as valeric acid) are important despite their high flavour thresholds. In particular, unsaturated fatty acids, such as trans-2-butenoic and hexenoic acids, give a sharp tangy flavour, and thioesters, such as methyl thiobutyrate and methyl (2-methyl)thiobutyrate, give aripeflavour. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

150 Food Flavour Technology

Intermediate-sized fatty acids have a soap-like flavour and so, in cheese preparation, only lipases that have a substrate specificity for glycerides containing short-chain fatty acids are used (e.g. pancreatic lipase). This use of enzymes as food additives has been employed to its greatest extent in the manufacture of enzyme-modified cheese flavours. This uses carefully selected esterases, lipases and proteases to develop cheese flavours, some 20- to 30-fold stronger than those of conventional cheeses. A particularly useful processing strategy is to use the enzymes sequentially in a ‘cascade’ to maximise their effectiveness. In addition, some more specialist approaches have been tested, such as the use of methioninase to form methanethiol from methionine (Lindsay and Rippe, 1986). As well as lipids, carbohydrates are very important for cheese flavours. Initially, milk lactose is metabolised into lactic acid by lactic acid bacteria. Lactic acid is the only product formed by streptococci and some lactobacilli, whereas Leuconostoc strains produce lactic and acetic acids and carbon dioxide in equimolar amounts. This results in a more open structure to the cheese, forming gas holes in some cases, or allowing aeration that can lead to the formation of blue-veined cheese varieties. In addition to these general processes, cheese flavour involves a great variety of micro-organisms producing a range of chemically quite different flavour chemicals (Bakker and Law, 1994). For instance, Streptomyces diaactylactus forms diacetyl, which is the basis of cottage cheese flavour, and Propionibacterium shermanii converts lactic acid into propionic acid, which is characteristic of Swiss cheeses. Other examples include pyrazines that occur in cheese and are produced by strains such as Pseudomonas perolens and Pseudomonas taetrolens and Corynebacterium glutamicum. Diacetyl (2,3-butanedione) contributes to milk and other dairy flavours and is produced by Lactobacillus lactis.Itis formed from citric acid via acetolactate, and its formation is stimulated by the addition of citric acid and by acidic pH, required by the citrate transport system. Methanethiol and its esters are also important contributors to some cheese flavours, such as Limburger, and are made by bacteria such as Brevibacterium linens and Penicillium freudenreichii. Methylthioacetate is found in cheeses, such as Gruyere` and Emmentaler, that are produced by propionic acid bacterial fermentations. Other important cheese flavour chemicals include butyric acid, propionic acid, various ␥- and ␦-lactones and also the methyl ketones produced by fungal Penicillium species, which are the characteristic flavour chemicals of blue cheeses such as Roquefort. The initial cheese fermentation is carried out by lactobacilli such as Lactobacillus casei. The subsequent ripening of the cheeses is much more idiosyncratic. Additional surface ripening by bacteria or fungi occurs in many cases, adding much of the flavour characteristics of particular cheeses. Quite often a mixed population of strains is present with successive growths of different microbial strains. For instance, with Roquefort, Camembert, Brie and others, the initial growth on the surface is of yeasts such as Kluyveromyces lactis or Kluyveromyces fragilis, Saccharomyces cerevisiae or Debaryomyces hansenii, which deaminate amino acids and oxidise lactic acid, causing an increase in pH. This increased pH encourages the growth of Penicillium roqueforti that gives the cheese its blue colour, especially when holes are bored into the cheese to allow the penicillium to colonise the interior of the cheese. P. roqueforti produces lipases that liberate free fatty acids that not only contribute to the flavour of the cheese but are also the substrates for the production of methyl ketones and also of ␦-lactones. This lipase activity rate-limits methyl ketone formation, rather than the conversion of the free fatty acids by ␤-oxidation. 2-Heptanone is the predominant methylketone of blue cheese, and 2-nonanone that of soft cheese. In addition, P. roqueforti has a good proteolytic activity, so that extensive hydrolysis of the cheese proteins can occur, producing large amounts of complex mixtures of flavour peptides (and amino acids), P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 151

which also contribute to flavour and can be further converted into other flavour molecules by decarboxylation, deamination and transamination. A most important factor has been to select proteolytic enzymes that minimise the bitterness of the hydrolysates they make. This involves using enzymes with a substrate specificity such that they tend not to produce peptides containing hydrophobic amino acids at the end of the peptide chains. These cheese proteins also have important influences on cheese texture, which is as important as flavour for consumer appreciation. Similar processes occur with different cheese varieties. Camembert is surface-ripened by Penicillium camemberti, and Brie is ripened by a mixture of P. camemberti and the bacterium B. linens. B. linens acting alone is responsible for the surface ripening of Limberger and similar cheese varieties. It acts only after salt-tolerant yeasts have raised the pH to about 6. In contrast to P. roqueforti, B. linens does not produce lipases but does make proteases, metabolising the resulting amino acids into a range of flavour chemicals such as 3-methylbutanol, 2-phenylethanol, methanethiol and many others. Flavour defects can also occur; for instance, reduction of diacetyl to acetoin by diacetyl reductase, which is produced by some cheese starter culture strains, is deleterious because the acetoin is much less highly flavoured. The extent of reduction is determined by the redox potential of the cheese. The variety of cheese flavours can be illustrated by Camembert cheese, which is characterised by a mushroom note given by 1-octen-3-ol. Camembert’s flowery note is due to 2-phenylethanol and its acetate; the hazelnut note is given by 1,3-dimethoxybenzene and methyl cinnamate; and the garlic note by molecules such as 2,4-dithiapentane, 2,4,5- triathiahexane and 3-methylthiol-2,4-dithiapentane.

5.4 FERMENTED PRODUCTS

5.4.1 Hydrolysed vegetable proteins Hydrolysed vegetable proteins (HVPs) have been used since before scientific times. A good example is soy sauce, which relies on the activities of naturally occurring ‘contaminant’ strains of Aspergillus, Lactobacillus and Saccharomyces. HVPs are remarkable as they enable meat-like savoury tastes to be produced from entirely vegetable raw materials that can be used to improve the palatability of quite cheap and abundant cereal foods much more cost-effectively than if meat-derived flavours were to be used. This effect is achieved not only by the presence of flavours generated during processing but also by flavour enhancers such as MSG, 5-IMP and 5-GMP. HVPs are produced from wheat gluten, defatted soy, peanut or cotton seed flours and other such materials, especially from wheat gluten, which has a high glutamic acid content and so yields high concentrations of MSG in the final product. Hydrolysis is at low pH and around 90◦C for several hours, requiring acid-resistant pressure vessels. Under these conditions, proteins are broken down into amino acids, and these can undergo Maillard reactions with sugars, also produced from the vegetable starting materials. Then, the hydrolysate is filtered, decolorised and concentrated or spray-dried (Swaine, 1993). The resulting HVP has a high salt content, derived from the acid used to carry out the hydrolysis, but unfortunately also contains low concentration of mono- and dichloropropanols. Because these are known carcinogens, they have to be maintained at low levels, for instance below 1 and 50 ppb, respectively. HVP bouillon-type flavours are complex and diffuse meat flavours that cannot easily be described in terms of a few flavour chemicals and that have a ‘warm’, salty and spicy P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

152 Food Flavour Technology

ﬂavour character. 5-Methyl-4-thiotetrahydrofuran-3-one and 2-methyltetrahydrothiophene- 3-thiol are top notes. Other ﬂavour chemicals include 5-ethyl-4-methyl-3-hydroxy-2(5H)- furanone, methional, sulfurol and 3-hydroxy-4,5-dimethyl-2(5H)-furanone, which give a meaty, savoury character.

The role of sulfurol is particularly interesting, as it has a relatively high ﬂavour threshold of 10 mg/L, but the sulfurol note often increases in intensity with storage. The related molecule 2-methyltetrahydrofuran-3-thiol may contribute to this, as it has a much lower ﬂavour threshold, and may be formed from sulfurol during cooking.

By comparison, the key ﬂavour-contributing chemicals in soy sauce are furanones especially sotolone, Furaneol and 2(5)-ethyl-4-hydroxy-5(2)-methyl-3-(2H)-furanone. Zygosac- caromyces rouxii is important in the development of the ﬂavour of soy sauce.

5.4.2 Chocolate Cocoa mass contains a complex mixture of flavour chemicals with both sweet and savoury characters. Further complexity is then added to the flavour as the first chocolate is made, with additional flavour chemicals with dairy characteristics introduced from the milk solids added, and then often additional ingredients such as nuts, raisins or coconut introduce yet more flavour chemicals. Some of the flavour compounds characteristic of chocolate and cocoa, especially savoury flavour chemicals, are formed during the fermentation process (Baigrie, 1994) that the beans undergo after harvesting (Table 5.2). During the fermentation, first sugars are metabolised by yeasts to produce ethanol, then lactic acid bacteria convert citric acid from the bean extracts into ethanol, and then acetic acid bacteria metabolise this ethanol into acetic acid (Rombaults, 1952). Some other flavours are produced by the fermentation, such as ethyl- 2-methylbutanoate, tetramethylpyrazine and some other pyrazines. The bitter taste notes are provided by theobromine and caffeine carried over from the original beans, together with diketopiperazines formed from the thermal decomposition of proteins during the roasting step. Other amino acids released during the fermentation are the precursors for other flavour chemicals such as 3-methylbutanol, phenylacetaldehyde, 2-methyl-3-(methyldithio)furan, 2-ethyl-3,5-dimethyl- and 2,3-diethyl-5-methylpyrazine. 2-Acetyl-1-pyrroline is formed by Bacillus cereus acting in the later stages of the fermentation, and also during the drying of the pulp, and so at least some of the acetyl-1-pyrroline of cocoa appears to be formed microbially, and not just by thermally induced reaction during the later stages of cocoa processing (roasting etc.). This carryover of flavours from fermentation appears to be a P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 153

general phenomenon, as tetramethylpyrazine has been shown to be made by Bacillus subtilis during the processing of the Japanese fermented soybean product called natto, which has an odour very characteristic of pyrazines. Following fermentation of the bean extract, it is dried, during which polyphenol oxidase activity continues, giving rise to new flavour molecules. Overall, the cocoa flavour is very dependent on the precise conditions and duration of harvesting, fermentation, drying and roasting stages. On further processing to chocolate, the sugar added is the main addition to the flavour, together with cocoa butter and some added flavouring materials, plus special ingredients such as nuts and coffee paste (Table 5.4).

5.4.3 Tea Tea flavour and aroma are complex and depend greatly on the particular type of tea leaf used and the processing conditions. For instance, fully fermented black tea has a very different character from that of partially fermented oolong tea. Initially, the leaves of Camellia sinensis are allowed to wither, allowing a loss of water, and then the leaves are macerated so that phenolics and other phytochemicals especially flavanols, such as epigallocatechin gallate, are allowed to react with endogenous enzymes. Then, the macerated leaves are allowed to ‘ferment’ at an elevated temperature so that extensive enzyme reactions take place, followed by ‘firing’ at higher temperatures, which cuts short the enzyme activity but allows chemical reactions, often involving the products of the enzyme reactions, to proceed at a high rate. Some of the important black tea aroma and flavour chemicals are shown in Table 5.3 (Belitz and Grosch, 1999). Others include diacetyl, methylpropanal and 2- and 3-methylbutanals. Many of these chemicals are produced by degradation of carotenoids and unsaturated fatty acids originally present in the tea leaf by the action of endogenous enzymes. Subsequently, further chemicals are produced during the firing of the tea, for instance by Strecker reactions. Chemicals produced by fatty acid degradation, such as cis-1,5-octadien- 3-one, cis-3-hexenal and 3-methyl-2,4-nonanedione, are especially present in green tea and give it its green fresh flavour character. In addition, teas, and especially black teas, have a very marked astringent character. This is due to the presence of thearubigins, which are complex phenols formed by the dimerisation of flavonols by endogenous polyphenol oxidases to form theaflavins, which give tea its red colour, and then further polymerisation to produce the thearubigins. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

154 Food Flavour Technology

5.4.4 Coffee Hundreds of flavour and aroma chemicals are present in roasted coffee. Many result from the thermal decomposition of carbohydrates and phenols, especially chlorogenic acids, that are present at significant concentrations in green beans during roasting. In addition, there are marked differences in flavour character and flavour chemical compositions between different coffees. This is due to the different varieties of coffee plants, different mixtures of beans and, of course, different ways of roasting. Only a minority of the compounds present actually contribute to coffee aroma. Furfurylthiol can be detected at 0.005 ␮g/L and smells of roast coffee at 0.01Ð0.5 ␮g/L, and 5-methylfurfurylthiol is detected at 0.05 ␮g/L and gives a stale coffee aroma when present at concentrations of 1Ð10 ␮g/L (Tressl and Silwar, 1981). It is also reported that 2-furfurylthio, 3-methyl-2-butenethio and 3-thio- 3-methylbutylformate give the roasted coffee-sulfur flavour note. The phenol-smoky note is given by guaiacol and 4-vinylguaiacol, and the earthy roast and caramel odours are given by pyrazines and furanones, respectively, with volatile acids such as formic and acetic acids also modifying the overall flavour. 3-Methyl-2-butene-1-thiol and 3-thio-3-methylbutanol and its formate have been found in roasted coffee (Blank, 1992; Holscher, 1992). Table 5.5 shows the concentrations and aroma values of many of the more important aroma and flavour chemicals of Arabica and Robusta coffees, which show significant differences in composition, thus accounting for the quite different tastes and aromas of different varieties of coffee, even when roasted and brewed in the same ways. In addition to the molecules described in Table 5.5, Semmelroch and Grosch (1996) include the following chemicals as contributing to coffee flavour and aroma Ð acetaldehyde, propanal, methylpropanal, 2- and 3-methylbutanals, 2-methyl-3-furanthiol, methanethiol, dimethyl trisulfide and 2-ethenyl-3,5-dimethyl- and 2- ethenyl-3-ethyl-5-methylpyrazines Ð which go some way to explaining the complexity and individual variations of coffee flavours. The other major factor affecting coffee quality is the length of storage, as volatile aroma compounds are lost, especially from ground coffee, with methanethiol and 2,3-pentanedione being used as indicators of coffee freshness, because they are rapidly lost on storage, especially the 2,3-pentanedione (Holscher, 1992).

5.4.5 Beer Beer flavour is made up of flavour chemicals contributed by the original ingredients, including added materials such as hops, together with yeast metabolites, and also as a result of changes that take place during maturation. For a review, see Verhagen (1994). One important variable is the type of cereal used and the degree of roasting. For instance, the more extensive roasting used prior to the brewing of dark beers creates methyldihydroxyfuranone, which gives these beers their caramel note. Many of the flavour chemicals characteristics of beer are produced by yeast enzymes acting on amino acids present in the malt. For example, leucine can undergo transamination and decarboxylation to produce 3-methylbutanal.

Other flavour compounds include 2-phenylethanol produced from phenylalanine, which contributes a floral note, dimethyl sulfide and 3-methylbutylformate, 4-vinylguaiacol P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 155

produced from ferulic acid, which gives a smoky/woody flavour note, and esters such as ethyl hexanoate and ethyl butyrate, as well as a number of organic acids such as acetic, lactic, citric, malic and propionic acids, plus amino acids derived from the malt. The individual yeast strains used to brew some speciality beers produce different flavour compounds. For instance, concentrations of vinylguaiacol of over 3 mg/L are found in German top-fermented wheat beer, well above its flavour threshold of 1 mg/L, whereas vinylguaiacol is well known as an off-flavour when present in many bottom-fermented beers. One important change that occurs in the maturation of lagers is a reduction in the concentration of diacetyl, which gives an undesirable buttery flavour note. Diacetyl is metabolised very slowly by yeast during the long lautering period, but now the enzyme diacetyl reductase can be added to accelerate this process and reduce maturation costs. 3-Methyl-3-butene- 1-thiol has been identified as the off-flavour caused in beer by exposure to UV light, the so-called ‘sunstruck’ off-note.

It is produced from the hop component humulone, and its flavour threshold has been estimated at 0.2Ð0.3 ng/L (Holscher, 1992). Similarly, trans-2-nonenal is a major source of stale flavours of beer. The use of hops is very important in flavouring most beers. Hops are an extract of Humulus lupulis that contain humulones (␣-acids) such as humulone, cohumulon and adhumalone, and lupulones (␤-acids) such as lupulone, colupulone and adlupulone, as well as terpene flavour chemicals such as myrcene, humulene and caryophyllene. The concentrations, relative amounts and types of the various bitterness conferring acids are all important contributors to beer flavour quality. During boiling of the wort, the humulones are isomerised into isohumulones, which are more soluble and more bitter tasting, and then can be further converted into humulinic acids,

which are less bitter. Lupulones are converted by boiling into hulupones and luputriones, which are less bitter than the lupulones. Since the isohumulones are soluble and much more bitter than the hulupones and luputriones, it is the humulones of the hops that make the greatest contribution to flavour, and so the humulone content of the different hop varieties is important both as regards quality control and as a breeding trait. Hops are a good example of a multifunctional ingredient. This is because, in addition to their flavouring properties, hops act as a beer clarifier by precipitating proteins, they have a preservative effect, and the pectins present help to stabilise beer foam. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

156 Food Flavour Technology

5.4.6 Wine The flavour chemicals responsible for the flavour and aroma of wines are either derived from the original grapes, such as ethyl-3-thiol propionate, as metabolites of the yeast fermentation, or develop during the maturation of the wine. Thousands of cultivars of grapevines (Vitis vinifera) have been developed worldwide, and many variations on the wine-producing process, so that enormous variations in the aroma, taste and chemical compositions of different wines occur. Fermentation has relatively little effect on wine flavour. This is mostly determined by the flavour molecules extracted from the grape, so that the time that the grape skins remain during the fermentation is important, whereas less commonly the wine flavour is modified by the practice of leaving the wine in contact with the yeast lees for an extended period after fermentation has ceased. Wine flavour is especially reliant on ethyl esters (ethyl acetate, propanoate, pentanoate, hexanoate, octanoate, decanoate) as well as the hexyl, 2-phenylethyl, 3-methylbutyl and ethyl acetates, plus a number of type-specific flavour chemicals. ␤-Damascenone is specific to Chardonnay and Reisling wines, ethyl cinnamate, ␤-ionone and linalool to Muscatel, and methyl anthranilate is the character-impact compound of Concord and Lambrusco wines. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 157

Ethyl-3-thio-propanoate is also a characteristic flavour chemical of Concord grapes, having a flavour threshold of 2 × 103 ␮g/L. It has a flavour of fruit and grape at low concentrations, but gives an animalic note when present at higher concentrations (Kolor, 1983). Perhaps the most striking examples of these grape variety-specific flavour chemicals are the character-impact compounds of Sauvignon varieties, such as 4-thio-4-methyl-2-pentanone (cat ketone), which is a key flavour component of Sauvignon grapes and wine, with a flavour threshold of 3 mg/L. 3-Methoxy-3-isobutylpyrazine also contributes to the aroma of Sauvignon wine and also to the taste of bell peppers. Furthermore, 3-thiohexanol and its acetate occur both in Sauvignon wine and in passion fruit, explaining why some wine tasters describe Sauvignon wines, especially Sauvignon blanc, as having a passion fruit character.

Many of the flavour compounds present in wine occur as glycoside precursors in the grape juice and are released during the wine-making process. Muscat grapes have been shown to contain concentrations of free linalool, geraniol, nerol and ␣-terpineol of 100, Ͻ5, Ͻ5 and Ͻ5 ␮g/kg fruit, respectively, so that only the linalool contributes to taste, as linalool has a flavour threshold of only 6 ␮g/L. But bound concentrations of these molecules of 390, 330, 170 and 40 ␮g/kg are present, so that release of the bound fractions by treatment with the appropriate glycosidases increases the concentrations of the other three flavour molecules closer to their thresholds. Changes in flavour during wine maturation can be desirable or adverse. For instance, ‘musty’ off-flavours can be due to the formation of geosmin or 1-octen-3-one, often caused by unwanted microbial or enzyme action. Desirable improvements in quality are also common during the maturation of wines. For instance, during the maturation of sherry, the amounts of acetals, esters and sotolone increase relative to the concentrations of ethanol and volatile acids. Botrytis is common microbial contaminant that infects grapes while still on the vine. It can produce 1-octen-3-ol (mushroom flavour) and sotolone, which has a honey-like flavour that can be perceived as either an off-note or a desirable flavour note as in the case of Tokai and Sauterne wines. A secondary fermentation that is often prized is the malolactate fermentation carried out by certain Lactobacillus or Leuconostoc strains. Country wines, as well as fine vintage wines, also have their characteristic flavour chemicals, which determine their individual flavours. For instance, Jorgenson et al. (2000) analysed elderflower wine by the GC-sniffing technique. They found that cis-rose oxide, nerol oxide, hotrienol and nonanal contributed to the special elderflower aroma, whereas linalool, ␣-terpineol, 4-methyl-3-penten-2-one and cis-␤-ocimene contributed floral notes.

Fruity odours were due to pentanal, heptanal and ␤-damascenone, and fresh notes to hexenal, hexanol and cis-3-hexenol. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

158 Food Flavour Technology

5.4.7 Sweeteners Although not strictly considered as flavours, there are several reasons for mentioning sweeteners in the context of plant-derived flavours. Firstly, the bulk sweeteners sucrose, glucose and high-fructose corn syrups are produced from plant sources, by extraction, enzyme hydrolysis of starch, and by the isomerisation of glucose using immobilised glucose isomerase, respectively. Secondly, high-intensity sweeteners such as AspartameR and Sucralose (SplendaR) are used in rather similar ways to flavours, and one of them, the intensely sweet protein thaumatin (TalinR), is obtained by extraction from the African plant Thaumatococcus daniellii. In addition, thaumatin has taste-modifying properties and has found a number of food and beverage uses that exploit this property. Quite a number of other sweet proteins have been discovered in plants including brazzein, which is 2000 times as sweet as sucrose, monellin, stevioside, glycyrrhizin, phyllodulcin, osladin and curculin. In addition, two other plant proteins, miraculin and Gymnema silvestre extract, have been found to possess taste-modifying properties.

5.5 CEREAL PRODUCTS

Background information about cereal flavour was published by Eriksson (1994). The flavour and aroma of bread crust are due to a number of compounds either arising from the yeast fermentation or formed owing to the higher temperatures and lower water environment the crust experiences during baking as compared to the bread crumb. 4-Hydroxy-2, 5-dimethyl- 3-furanone and 2- and 3-methylbutanals are responsible for the roasted, malty, caramel impression. Methylfurfuryl disulfide has been reported to give the ‘golden brown crust aroma’ (Mulders, 1976). Other aroma chemicals present in bread include 6-acetyltetrahydropyridine, 2-methyl-3-ethylpyrazine, 5-methyl-5H-cyclopentapyrazine, 2-methylfurfuryl disulfide, 2- acetylthiazole, 2- and 3-methylbutanals, methylpropanal, cis- and trans-2-nonenal and 2- acetylpyrazine, as well as 2-acetyl-1-pyrroline that is especially produced from the yeast metabolites ornithine, proline and 2-ketopropanal in the bread crust.

2-Acetyl-1-pyrroline has a very low aroma threshold, but concentrations decrease rapidly as the bread ages, due to air oxidation. This marked instability makes 2-acetyl-1-pyrroline a rather difficult flavour chemical to use by adding to processed foods. Other common bread flavour molecules present in the crumb include methylpropanal, 2-decanal, 2-nonenal, diacetyl, methional and 1-octen-3-one, plus 2-phenylethanol if the bread has been fermented for a long time. Fatty acid decomposition, followed by the condensation of the aldehyde products to form molecules such as 2-butyl-2-octenal and 2-butyl-2-heptenal, are responsible for rancid off-notes in cereals. Similarly, 2-methylisoborneol and geosmin have been found to be responsible for musty off-odours in cereals and are thought to be microbially produced. Analysis of 36 aroma chemicals in six different flavoured rice types showed that when taking into account the thresholds and concentrations of the various aroma chemicals detected, that over 97% of the aroma value of all the six types of rice was due to combinations of just six of the aroma chemicals. These were 2-acetyl pyrroline and the aldehydes hexanal, P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 159

(E)-2-nonenal, octanal, heptanal and nonanal, but with the relative proportions of each varying from rice type to type. It was also found that these six aroma chemicals together with another seven more minor aroma chemicals were responsible for the differences in ﬂavour between the six types of rice (Yang et al., 2008).

5.6 VEGETABLE SOURCES OF FLAVOUR

5.6.1 Spice ﬂavours Spices encompass a huge range of taste and odour sensations, achieved using an equally wide range of chemicals. These include capsaicin and dihydrocapsaicin (peppers), sotolone (fenugreek Ð curry), zingerone (ginger), trans-2-dodecenal (coriander), eugenol (cloves and cinnamon), 1,8-cymene (rosemary, cardamon, allspice and sage), methyl chavicol (basil), fenchone and anethole (fennel), anthranilic acid ester (mandarin), l-menthol and l-menthone (peppermint), d-carvone (caraway), l-carvone (spearmint), thymol (thyme), safranol (saf- fron) and ␣-phellandrene (dill). P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

160 Food Flavour Technology

Flavour molecules that have a pungent ﬂavour are found in ginger and pepper. They show a particular structural similarity. Peppers contain capsaicinoid chemicals such as capsaisin and related molecules that stimulate the pain receptors in the mouth. Piperine is the active pungent constituent of pepper. By contrast, bell peppers are not hot, and they, together with chilli and paprika peppers, contain 2-isobutyl-3-methoxypyrazine, which is believed to be biosynthesised from leucine. Pyrazines can also be produced by micro-organisms, but most of the pyrazines present in foods are generated as reaction ﬂavours by non-enzymic processes during the cooking of foods.

Ginger contains gingerols and shogoals, which vary in proportions depending on the type of ginger used and storage conditions. As can be seen, gingerols are simply a hydrated form of the corresponding shogoals. Ginger ﬂavour is particularly interesting as changes can occur on storage, with gingerol dehydrating into shogoal, and in turn shogoal can be cleaved into zingerone and hexanal, which adds to taste complexity.

The perceived degree of hotness is measured in Scoville units, and capsaicin is so potent that even when diluted by 105-fold it will still blister the tongue. It is 70 times as hot as piperine and 1000 times hotter than zingerone. The great variety and potency of flavour chemicals present in plants and especially the low concentrations they occur in makes it tempting to speculate what developments may be possible using genetic engineering techniques. Terpenes are a good target for improvement because despite being good sources of flavour and fragrance materials they are usually only present in quite low concentrations in their plant sources so that they tend to be expensive or even uneconomic to exploit. An exception is mint oils that are produced in comparatively high concentrations because they are stored in specialised structures on the underside of the plants leaves. An important step forward has been taken by co-expressing the enzymes that make the isoprenoid precurors of terpenes or sesquiterpenes, together with terpene synthases or sesquiterpene synthases, and both in the same subcellular location, either in the cytosol or in plastids, both of which are different locations to where these enzymes are usually located, so as to ensure more efficient supplies of precursors for terpene synthesis (Wu, 2006). Thus, for terpene synthesis, tobacco limonene synthase was co-expressed with geranyl phosphate synthase to provide it with precursor. For sesquiterpene synthesis, patchoulol synthase was co-expressed with farnesyl diphosphate synthase to provide it with precursor. The most striking result was a great enhancement of sesquiterpene production if the sesquiterpene P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 161

synthase is expressed in the plastids combined with a plastid farnesyl diphosphate synthase, which enhanced the production of the sesquiterpene amorpha-4, 11-diene precursor of the antimalarial drug artemisinin by a factor of 40 000-fold. Many spices are processed into essential oils and then used as such. Some of the larger volume essential oils are produced from cassia, cinnamon leaf, clove bud, coriander, dill, rosemary, sassafras and star anise. By contrast, vegetable oils have very different ﬂavour compositions. For instance, extra virgin olive oil is reported to contain acetic acid, cis-3-hexenol and cis-2-nonenol, as well as the ethyl esters of isobutyric, 2-methylbutyric and cyclohexanoic acids, and on occasions 4-methoxy-2-methyl-2-butanethiol that also provides blackcurrants with their characteristic ﬂavour.

5.6.2 Mushroom The most well-known mushroom flavour chemical is 1-octen-3-ol, which has a taste threshold of 0.01 ␮g/L. In addition, over one hundred other flavour chemicals have been found in mushrooms. They add to the complexity of the mushroom taste and help explain the differences in taste between the various types of mushrooms. This range of flavour chemicals includes 1-octen-3-one, 1- and 3-octanols, 3-octanone, trans-2-octen-1-ol and trans- 2-octenal. Whereas easily the best known, widest used and commercially most important mushroom species is Agaricus bisporus, some flavour chemicals are characteristic of particular species of mushrooms. trans-Nerolidol and fokienol are the main volatile components of Lentinellus cochleatus, and lenthionine; a cyclic sulfur-containing flavour molecule (1, 2, 3, 5, 6-pentathiepane) is characteristic of shiitake mushrooms (L. edodes). The oyster mushroom (Pleurotus florida) was grown in submerged culture, and the most abundant flavour chemicals were found to be anisaldehyde, 3-methyl-1-butanol, 3-methyl-1-propanol, benzaldehyde and 1-octen-3-one (Venkateshwarlu et al., 2000).

5.6.3 Garlic, onion and related flavours Garlic flavours are based on (di)allyl disulfide (di-(2-propenyl) disulfide), which constitutes 90% of the active flavour of garlic oil. It is produced from the precursor alliin (s-allyl-l- cysteine sulfoxide), via the intermediate allicin, by alliinase enzyme, which is only released when the vegetable tissue is crushed. 1-Propenyl disulfide is also present in garlic oils, but so far no manufacturing process has been developed to make it more readily available for formulation into flavours.

As well as the allyl disulfide, the trisulfide, mixed disulfide, mercaptan and other products are produced from allyl sulfide by disproportionation and rearrangement reactions. These increase the complexity of the flavour formed. In particular, the savoury flavour note increases with the sulfur content of the flavour molecule. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

162 Food Flavour Technology

Flavour complexity is also increased because some 5-methyl- and 5-propylcysteine sulfoxide precursors are also present in garlic and give rise to the corresponding flavour chemicals. In addition, chemical interactions with other food components occur, and the overall flavour can be modified by the presence of other flavour chemicals. For instance, in the presence of 3-hydroxy-2,5-dimethyl-4(H)-furanone a sweet, roasted character is produced. Onion flavours are based on propyl derivatives, such as propyl methane and propane thiosulfonates, rather than the allyl derivatives found in garlic, and so the flavour chemicals formed tend to be saturated rather than unsaturated. 5-Methyl and 5-propyl precursors also contribute flavour chemicals. Alkyl alkanethiosulfonates, especially propyl methanethiosul- fonate and propyl propanethiosulfonates and alkyl thiosulfonates, are characteristic of the flavour of raw onion, and have odour thresholds of 1.7 and 1.5 ␮g/L, respectively (Boelens et al., 1993).

Cooked onion flavour is given by dipropyl disulfide, and cis- and trans-2-propenylpropyl disulfides, which have odour thresholds of 3.2 and 2.0 ␮g/L, respectively, and also by various other sulfides such as trisulfides.

In fried onions the characteristic ﬂavour chemicals formed are 2-(propyl- dithio)dimethylthiophenes, which have odour thresholds of 0.01Ð0.05 ␮g/L (Kuo and Ho, 1992).

It has been found that the concentration of the potent onion flavour 3-mercapto-2- methylpentan-1-ol, which has a taste threshold of 0.03Ð4 ␮g/L and is also present in chives, scallions and leeks, but not in garlic, is present in four- to eightfold higher concentrations in sliced and stored, and in cooked onions than in raw onions (Granvogle et al., 2004). This is presumably because it is generated in situ by enzymes released when the onion tissue is disrupted that act on a yet-to-be-identified precursor molecule. The corresponding aldehyde is also present in onion. Other vegetables of the allium group contain similar flavour chemicals; for instance, leeks contain methylpropyl di- and trisulfides and dipropyl trisulfide. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 163

Table 5.13 Natural sources of isothiocyanates.

Isothiocyanate Source Methyl Cleome spinosa Ethyl Lepidivum menzieli t-Butyl Puntranjiva roburghii t-Butenyl-4 Rapeseed Isophenyl Horseradish Benzyl Garden cress, nasturtium p-Hydroxybenzyl White mustard Phenylethyl Watercress 3-Indoylmethyl Brassica species 2-Hydroxy-3-butenyl Brassica species n-Methoxy-3-indomethyl Brassica species

From Clark (1992) with permission.

5.6.4 Brassica flavours, including mustard and horseradish Isothiocyanates are the characteristic pungent flavouring components in Brassica plants, including mustard, horseradish, broccoli, cabbage and cress. The particular form of isothiocyanate present varies from species to species (Table 5.13), with each species being characterised by the possession of a particular isothiocyanate, such as p-hydroxybenzyl isothiocyanate in white mustard, phenylethyl isothiocyanate in watercress, 2-propenyl, 3-butenyl as well as 2-phenylethyl isothiocyanates in cabbage as compared to isophenyl isothiocyanate in horseradish. 3-Methylthiopropylisothiocyanate and 4-methylthiolbutylisothiocyanates are characteristic of cauliflower and broccoli, respectively. The isothiocyanates are formed from glucosinolate precursors, via unstable thiohydroxamate-O-sulfate intermediates, by the action of myrosinase (thioglucosidase) that is liberated when the plant tissue is disrupted, such as by mastication or maceration during processing. The main precursors are allyl isothiocyanates, but methyl, ethyl and isopropyl isothiocyanates are also present, and products of their reaction with myrosinase also contribute to the flavour. In addition to the formation of isothiocyanates, lesser amounts of nitrile and thiocyanate products are also formed from the unstable intermediates produced by myrosinase. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

164 Food Flavour Technology

Allyl isothiocyanate has a rather particular use, namely as a denaturant to give a repellent bitter taste to many household cleaning products, so as to deter children from accidentally poisoning themselves by drinking these often attractively labelled products.

5.6.5 ‘Fresh/green/grassy’

This ﬂavour note is conferred by C6 alcohols and aldehydes, typically trans-2-hexenal (leaf aldehyde) and cis-3-hexenol (leaf alcohol), as well as others such as trans-2-hexenol and cis-3-hexenyl butyrate (for reviews see Hatanaka, 1993; Fabre and Goma, 1999). The former compound is produced by isomerisation of the unstable cis-3-hexenal, which is the original product formed by the action of lipoxygenase, and then by lyase enzyme, on linoleic acid. cis-3-Hexenal isomerisation is catalysed by both acid and base, so it can only be used effectively at its optimal pH stability point. This is a considerable limitation as cis-3-hexenal has a very low aroma threshold of 0.25 ␮g/L, whereas trans-2-hexenal and cis-3-hexenal are less intense, having thresholds of 17 and 70 ␮g/L, respectively (that is about 64-fold and 280-fold higher).

The nuances of flavour between the various C6 flavour molecules have been compared by Whitehead et al. (1995). Their descriptions were hexanal, fatty, green and fruity; trans-2-hexenal, leafy green and fruity; cis-3-hexenal (unstable); hexan-1-ol, oily green; trans-2-hexen-1-ol, sharp, green, fruity; cis-3-hexen-1-ol, grassy green, fresh. These C6 flavours are also responsible for the off-flavours that develop in some foods on prolonged storage owing to the action of lipoxygenase on unsaturated fatty acids. This has been com- bated by blanching to inactivate the enzymes, and irradiation can also stabilise vegetables against the formation of off-flavours. The fresh green flavour sensation extends from C6 molecules, through C7 and C8 to C9 nonadienols, which have green vegetable-cucumber-melon characters, especially 3,6- nonadien-1-ol (violet leaf alcohol or cucumber alcohol), which is characteristic of cucumbers. This was the first nonadienol to be isolated and studied. Nonadienols are also formed by lipoxygenase action on linoleic acid and occur in melon and some grapes as well as in cucumbers. Nonadienals give a good illustration of the relationships between structure, flavour character and flavour price (Table 5.14). A further flavour variation is the formation of C11 hydrocarbons, such as 1,3-trans-5-cis-undecatriene and 1,3-trans-5,8-cis-undecatetraene found in various fruits and vegetables, by ␤-oxidation of linoleic acid and then lipoxygenase activity, non-enzymic oxidation and decarboxylation.

5.6.6 Nuts The distinctive tastes of nuts are contributed by ﬁlbertone (5-methylhept-2-en-3-one) and various pyrazines that give a roasted taste character. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 165

Table 5.14 The organoleptic impression of the C9 series.

Material Price (US$/kg) Impression Nonanal 20 Fatty, ﬂoral Nonan-1-ol 28 Oily, ﬂoral cis-2-Nonen-1-ol 650 Waxy melon trans-2-Nonenal 200 Fatty, waxy-orris trans-2-Nonen-1-ol 400 Waxy green cis-3-Nonen-1-ol 330 Waxy green melon cis-6-Nonenal 1500 Fresh melon cis-6-Nonen-1-ol 800 Powerful melon trans-2-trans-4-Nonadienal 600 Powerful fatty green trans-2-trans-4-Nonadien-1-ol 600 Mild-fatty trans-2-cis-6-Nonadienal 1800 Green vegetable, cucumber trans-2-cis-6-Nonadien-ol-1 2300 Green cucumber trans-2-trans-6-Nonadienal 2300 Green citrus trans-3-cis-6-Nonadien-1-ol 2300 Melon, cucumber cis-3-cis-6-Nonadien-1-ol 3500 Melon, cucumber

From Clark (1999) with permission.

5.6.7 Other vegetables Carrots contain 2-sec-butyl-3-methoxy pyrazine; celery, phthalides and dihydroph- thalides; beetroot, geosmin; potato, 2-isopropyl-3-methoxy pyrazine, 2,5-dimethyl pyrazine and 3-(methylthiol) propanal (methional); trufﬂes, dimethyl sulﬁde and bis-methylthio methane; peas, a variety of pyrazines such as 3-isopropyl, 3-sec-butyl and 3-isobutyl-2- methoxypyrazines; and asparagus contains 1,2-dithiocyclopentene, which is produced from 1,2-dithiolane-4-carboxylic acid during cooking.

5.6.8 Fermented vegetables A range of fermented vegetable products have been developed, often in prescientiﬁc times. Naturally occurring microbial strains used such as Lactobaccilli, Leuconostoc and Pediococ- cus strains are most often involved producing acids that have a preservative effect, improving the digestibility of the vegetables and giving them new ﬂavours. Common fermented vegetable products include pickled cucumbers and cabbage, such as sauerkraut.

5.7 FRUIT

Development of the flavour of a fruit can occur either when still on the parent plant or after harvesting. Ripening after harvesting is associated with a significant increase in the metabolic activity and an increase in CO2 production by the fruit and is termed climacteric ripening. Climacteric fruits include apples and pears, and plums, bananas, peaches and tomatoes, whereas non-climacteric fruits that ripen before harvesting include oranges and lemons, strawberries, cherries, melons and grapes. As will be seen below, a great variety of different types of flavour chemicals contribute to the flavours of fruits. One type of flavour chemical that is particularly characteristic of fruit is cyclic terpenes such as ␤-damascenone and ␤-ionone. They are both formed by the P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

166 Food Flavour Technology

breakdown of carotenoids by the action of lipoxygenase and are reported to have ﬂavour thresholds of around 0.002 and 0.007 ␮g/L, respectively.

5.7.1 Apples Esters are the main contributors. For instance, ethyl 2-methylbutyrate is a contributor to apple ﬂavour, with an odour threshold of 0.1 ␮g/L, but does not confer a distinct apple ﬂavour by itself. This requires a range of other acids, aldehydes, alcohols and esters such as cis-3-hexenol, trans-2-hexenal, ␤-damascenone, ethyl butyrate, cis-3-hexenylbutyrate and hexyl-2-methylbutyrate.

Flavour chemical synthesis in apples has been shown to take place just under the skin by experiments in which precursors such as butanol were supplied to the fruit, and measuring the distribution of the esters formed from the precursor over the cross section of the apple (Berger, 1990). In some cases ester synthesis was stimulated to such an extent that the apples tasted too strong to eat. An interesting study of the supply of exogenous n-butanol vapour to apples as a substrate for flavour ester formation shows that, whereas butanol penetrates deeply into the apple, flavour esters predominate close to the surface of the apple, strongly indicating that the enzymes responsible for flavour formation are located in, or close to, the skin of the apple. The great increase in ester concentrations formed from the supplied precursor Ð up to 200 times the levels normally present Ð proves that effective concentrations of the enzymes are present. Probably these esters are formed via coenzyme A ester intermediates. The characteristic flavours of particular varieties of fruit can often be accounted for by the presence of one variety-specific flavour chemical or a small number of such chemicals. For instance, the characteristic aniseed flavour note of the apple variety ‘Ellison’s Orange’ has been identified as being due to the presence of 1-methoxy-4-propenylbenzene (estragole). Much more basic flavour characteristics are variety dependent, as with acid apple varieties (Cox) as compared with sweeter varieties of lower acidity (Jonagold). The actual situation is more subtle as acidity tends to reduce during storage, unless the fruit is stored under a controlled low-oxygen atmosphere, and can result in depletion of important flavour compounds such as butyl and hexyl acetates. This behaviour has led to selective breeding studies to produce new hybrid varieties with improved flavour storage characteristics. The 2-methylbutyl esters predominate not only in apples but also in strawberries, papaya, blackcurrant, pineapples and other fruits. But in cider, which is also produced from apple juice, considerable amounts of 3-methylbutyl ester flavour chemicals occur. Other fruit esters frequently present, but without conferring a flavour character are ethyl-2-methylbutyrate, ethyl butyrate and ethyl hexanoate, which have odour thresholds of 1, 0.1 and 1Ð2 ␮g/L, respectively. The complexity of the various flavours possible is clearly illustrated by sensory mapping of apple drinks, which shows clear variations in perception resulting from the differences in types of apple, processing methods, and so on. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 167

5.7.2 Pears Pear flavour is chiefly given by esters of unsaturated fatty acids, such as ethyl 2,4- decadienoate, ethyl-2-octenoate, hexyl acetate, ethyl-4-decenoate, butyl acetate and ethyl butyrate. A particularly important flavour component of pears is 2-trans-4-cis-decadienoic acid ethyl ester. It is believed to be produced in the fruit from linoleic acid by ␤-oxidation, isomerisation of a double bond and further ␤-oxidation, followed by desaturation and esterification with ethanol.

5.7.3 Grapefruit A very good example of character-impact compounds is provided by grapefruit. For some time it has been recognised that the sesquiterpene (R)-nootkatone has a potent grapefruit ﬂavour character with a quite low odour threshold of 1 ␮g/L. More recently, it has been discovered that a quite different chemical, (R)-(+)-p-1-menthene-8-thiol (grapefruit mercaptan), also gives grapefruit character and has a remarkably low threshold of 0.00002 ␮g/L. Indeed, it is so potent that it only exhibits the grapefruit character when diluted to below 10 ␮g/L; at higher concentrations it only has a nondescript rubbery odour. By contrast, the (S)-(−)-p-1-menthene-8-thiol isomer is reported to have an unpleasant sulfur ﬂavour, and a quite different threshold of detection.

5.7.4 Blackcurrant Various sulfur-containing flavour chemicals give the characteristic flavour of blackcurrant. Examples are cat ketone (4-thio-4-methylpentan-2-one), blackcurrant mercaptan (4- methoxy-2-methyl-2-butanethiol) and particularly 8-thio-p-menthan-3-one. Blackcurrant mercaptan is naturally present in blackcurrants, and it is also found in olive oil and has a flavour threshold of 0.2 ␮g/L. 8-Thio-p-menthan-3-one is found in buchu leaf oil and is used to boost the flavour of blackcurrant extracts as is the cat ketone. Only the 1S,4R-isomer out of the four possible isomers was described as blackcurrant-like (Kopke and Mosandl, 1992). 1-Methoxy-3-methyl-3-butanethiol also has a blackcurrant flavour and a threshold of 0.08Ð 3 ␮g/L, and has been detected in virgin olive oils (Guth and Grosch, 1991). The composition of blackcurrant leaf oil from different cultivators extracted by steam distillation and low-temperature solvent extraction has been reported by Marriott (1988). P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

168 Food Flavour Technology

5.7.5 Raspberry 1-p-Hydroxyphenylbutan-3-one (raspberry ketone), which has a ﬂavour threshold of 5 ␮g/kg, gives raspberries its characteristic ﬂavour.

Other ﬂavour chemicals that make a big contribution to raspberry ﬂavour include ␤-damascenone, vanillin, 4-hydroxy-2,5-dimethyl-3-furanone, sotolone, 1-nonen-3-one, cis-3-hexenal, ␣- and ␤-ionones and the ethyl esters of 5-hydroxyoctanoic acid and 5-hydroxydecanoic acid, with these esters capable of hydrolysing and cyclising to form lactones during cooking.

5.7.6 Strawberry 4-Hydroxy-2,5-dimethylfuran-3-one and cis-3-hexenol are important in determining strawberry ﬂavour. Also contributing are methyl butanoate, ethyl 2-methylbutanoate, methyl-2- methylbutanoate, acetic acid, 2,3-butanedione and methyl and ethyl cinnamates. Signiﬁcant changes occur on cooking or freezing. On freezing, concentrations of hydroxy-dimethylfura- none increase 6-fold and cis-3-hexenol concentrations fall 60-fold. On heating, cis-3-hexenol concentrations also fall and the concentrations of hydroxydimethylfuranone, ␤-damascenone, 2,4-decadienol and guaiacol increase markedly. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 169

For comparison, it is interesting to note that early strawberry ﬂavours used the synthetic compounds ethylphenylglycidate and ethyl methylphenylglycidate, the merits of which are discussed in Chapter 1.

5.7.7 Apricot and peach Various flavour chemicals contribute to apricot flavour. These include myrcene, limonene, p- cymene, terpinolene, ␣-terpineol, geraniol and geranial, linalool, acetic and 2-methylbutyric acids, trans-2-hexenol, and the lactones ␥-caprolactone, ␥-octalactone, ␥-dodecalactone, ␦-octalactone and ␦-decalactone. In peaches, the main flavour character is provided by lactones, especially ␥-decalactones, and also other C6 to C12 ␥-lactones and C10 and C12 ␦-lactones. Other contributors include benzyl aldehyde, benzyl alcohol, ethyl cinnamate, isopentyl acetate, linalool, ␣-terpineol, ␣- and ␤-ionones, 6-pentyl-␣-pyrone, and hexanal, cis-3-hexenal and trans-2-hexenal.

5.7.8 Tomato Tomato cultivars have been analysed for their volatile flavour composition using aroma extract dilution analysis followed by GC-MS. Some flavour chemicals such as 1-pentene-3- one, trans, trans and trans, cis-2,4-decadienals and 4-hydroxy-2,5-dimethyl-3(2H) furaneol were present in higher concentrations in the preferred cultivars, whereas others such as methional, phenylacetaldehyde, 2-phenylethanol or 2-isobutylthiazole were present in the less preferred cultivars, which may give a chemical explanation for the popularity of the different tomato cultivars (Mayer et al., 2008). The flavour of tomatoes has been enhanced by expressing the geraniol synthase gene of lemon basil (Ocimum basilicum) under control of the tomato ripening-specific polygalactur- onase promoter. The enhanced flavour of the transgenic fruit was due to increased formation of geraniol as the tomato ripens, which then also leads to the increased formation of not only geraniol-derived flavour compounds such as geranial, neral and nerol but also citronellol and thus citronellol-derived flavour compounds such as citronellal. The enhanced geraniol formation is, however, at the expense of the accumulation of lycopene. This is because the substrate for geraniol synthesis is geranyl diphosphate, which is also the substrate for lycopene synthesis (Davidovitch-Rikanati et al. (2007).

5.7.9 Cherry The main ﬂavour character-determining chemical is benzaldehyde. It is generated from a cyanogenic precursor glycoside by endogenous glycosidase and lyase enzymes. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

170 Food Flavour Technology

5.7.10 Tropical fruit flavours Tropical fruit flavours such as passion fruit are becoming increasingly popular with western consumers. A typical flavour chemical that contributes to the distinctive taste of yellow passion fruit is tropathiane (2S,4R 2-methyl-4-propyl-1,3-oxathiane). It is present in passion fruit along with other sulfur-containing flavour chemicals such as 3-mercapto-1-hexanol and 3-(methylthiol) hexanol and only the 2S,4R-isomer has the passion fruit flavour character. By comparison, crimson passion fruit contains terpenoid megastigmatrienes. Melons and pineapples are well-appreciated tropical fruits. The flavour of melon is contributed to by thio esters such as methyl and ethyl (methylthiol) acetates, 2-(methylthiol) ethyl acetate, methyl and ethyl 3-(methylthiol) ethyl acetates and propionates and with 3-(methyl thiol) propyl acetate and methyl and ethyl-2-(methylthio)propionate present in pineapples. However, one Asian fruit that has too distinctive a taste for most western palates is durian. It is also characterised by sulfur-containing flavour chemicals such as 2-isopropyl-4-methyl thiazole and methyl thiohexanoate.

5.7.11 Vanilla Vanilla ﬂavour has always been one of the most popular of ﬂavours. Vanilla beans are the fruit of the epiphytic orchids Vanilla planifolia and to a lesser extent Vanilla tahitiensis. Some 2000 t/year of vanillin beans are produced, with Madagascar being the largest producer by far. North America is the largest market, using about 1400 t/year, and Europe and Japan use 450 and 70 t/year, respectively (Todd, 1998).

Vanilla extracts are generally made by extracting the beans with ethanolÐwater. The vanillin and other flavours are formed during the curing process by glycosidase action on glycoside precursors present in the green beans. The concentration of vanillin and other flavour chemicals in the extract depends on the proportions of bean and extracting solvent used: this is commonly referred to as the ‘fold’ of the extract. The flavour obtained varies considerably with source, depending especially on the variety of vanilla grown, climate, the soil type and cultivation conditions, and the methods of harvesting and processing, especially the method of curing. The vanilla flavour is composed of a wide range of aldehydes, ketones, alcohols, esters, ethers, hydrocarbons, oils, waxes and resins (Adedeji et al., 1993). Of these, vanillin constitutes the ‘heart’ of the flavour. In a detailed analysis of 10 different vanilla extracts, only 19 of a total 194 organic compounds detected were present in all 10 extracts. These extracts were Bourbon (two types), Tahitian, Bali (two types), Java, Mexican, Tonga, Costa Rican and Jamaican. Some of the aroma chemicals are definitive for particular types of vanilla extract. For instance, anise acid and anisaldehyde are only found in concentrations sufficient to influence the flavour in Tahitian vanilla extract. On the other hand, some of the organic compounds detected will contribute little or nothing to the flavour and aroma, because they have only a low flavour impact or/and because they are only present in low concentrations. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 171

Most vanilla beans are sold to North America. This is because Americans prefer sweetness in foods and beverages, so that relatively high concentrations of vanilla extract are formulated, and because Americans consume a lot of ice cream, which is the single biggest use for vanilla extracts. Vanilla continues to be the biggest selling sweet flavour, and the identity of vanilla flavour is strictly regulated in the USA, which defines three categories of vanilla ice cream: all natural, natural supplemented with artificial, and artificial (categories 1, 2, and 3, respectively). Traditionally, Americans preferred the relatively harsh, phenolic flavour notes of Indonesian vanilla, which has a quite smoky aroma. Indonesian vanilla extracts have been cheaper than Madagascan vanilla, and so are cost-effective for use in type 2 vanilla flavour ice cream supplemented with artificial vanillin. Much vanillin is produced by chemical processes for supplementation of vanilla extract, but now genuine naturally produced vanillin is beginning to be available from microbiological processes.

5.7.12 Other fruits Isopentyl acetate is characteristic of bananas; important ﬂavour constituents of plums are benzaldehyde, ␥-decalactone, linalool and methyl cinnamate, and pineapples contain 4-hydroxy-2,5-dimethyl-3-(2H)-furanone and (S)-ethyl-2-methylbutyric acid.

5.7.13 Citrus

Orange ﬂavour is contributed by minor constituents of the orange oil such as C8ÐC11 saturated aldehydes (octanal, decanal), undecanal terpenes such as ␤-sinensal, esters such as ethyl butanoate, ethyl-2-methylbutanoate and ethyl isobutyrate, and unsaturated aldehydes such as cis-3-hexenal, and especially trans-2-decenal. By contrast, mandarin ﬂavour contains ␣-sinensal.

Citral, which is a mixture of the two stereoisomers, geranial and neral, is a characteristic flavour of lemons. Linalool, myrcene and limonene, which are the major components of citrus oils, also make a contribution to flavour. Traces of hydrogen sulfide and dimethyl sulfide are also present in all citrus fruits and contribute to the overall flavour. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

172 Food Flavour Technology

Unfortunately, the excellent flavour and aroma of fresh, cold-pressed citrus oils is unstable. This is because the unsaturated hydrocarbon terpenes such as limonene, which comprise 80Ð95% of the oil but have very little aroma, are rapidly oxidised to produce molecules that cause strong off-tastes. Therefore, various techniques have been developed to reduce or remove the hydrocarbon constituents of oils before they can become oxidised. Unlike with some sources of flavours, considerable efforts have been made in the plant breeding of citrus for improved flavour characteristics.

5.7.14 Citrus processing Citrus is a good example of large-scale agroprocessing that uses a refinery approach and has flavour materials as just one of its products. A range of citrus crops are processed to produce essential oils for use both as flavours and fragrance ingredients. These include bergamot, grapefruit, lemon, lime, mandarin, sweet and bitter orange. The orange processing industry worldwide has a turnover of over US$2 billion per year. Over 400Ð500 million trees are under cultivation, mainly in Brazil and Florida, which have about 150 and 60 million trees, respectively, and some 75 000 t/year of d-limonene and orange oil is produced from the fruit and also special essential oils from particular orange varieties such as petitgrain and neroli oils. These figures give an impressive indication of the size of the industry (Bovill, 1996), especially when it is considered that the yield of D-limonene from the fruit is only 0.27%, compared with a yield of 53% for the juice. Oranges are squeezed to extract the juice, which is concentrated by evaporation about 6·5-fold before being shipped. The remaining orange peel is then subject to pressing or rasping to extract the so-called cold-pressed oil, which is the source of many useful flavour and fragrance molecules. These mainly oxygenated terpenes are a minority of the oil, which is mostly composed of limonene and some other ‘hydrocarbon’ terpenes. Since oxidation of the terpenes can cause off-flavour, a major objective of citrus oil processing is to reduce its hydrocarbon content. The traditional method is by fractional distillation followed by washing, or by ‘folding’. This involves taking a many-fold concentrate and directly dissolving it in a waterÐethanol solution. Although distillation is cheap, some volatile flavour materials are lost, some thermal degradation occurs, and the sesquiterpene hydrocarbons are not removed and so are still available for oxidation. Therefore, a molecular still (thin-film evaporator) has to be used to reduce residence times and to minimise thermal degradation. An alternative method is to wash the citrus oil with ethanolic solutions, as the valuable oxygenated terpenes are soluble in ethanol whereas the hydrocarbon terpenes are insoluble. The hydrocarbon extract is called washed citrus oil, and the alcohol solution containing concentrated oxygenated terpenes is called ‘washed extracts’. Although effective, this counter-current process is difficult to operate because of the difficulty in efficiently contacting the two phases without emulsifying them so finely that their subsequent separation is difficult. Counter-current extraction with liquid carbon dioxide, which has a polarity as an extracting solvent close to that of hexane, has been used. However, this method suffers from high capital costs. Absorption column chromatography using silica gel as the absorbent and ethyl acetate or hexane as the eliciting solvent has been developed (Moyler and Stephens, 1992). The newest method for removing the hydrocarbon terpenes is the so-called poroplast extraction method. This involves the aqueous phase passing through a column containing a low-polarity stationary phase on to which the hydrocarbons absorb, with the oxygenated hydrocarbons being carried through and out of the column by a flow of aqueous ethanol. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 173

Table 5.15 A comparison of the compositions of conventionally prepared and poroplast-extracted hydrocarbon-free orange oils.

Conventional Poroplast-extracted Linalool 28.0 29.9 Decanal 17.0 12.72 Octanal 11.6 14.17 Aldehyde acetals 8.0 — Geranial 6.5 4.21 Neral 4.0 2.81 Dodecanal 3.3 1.93 Citronellal 3.2 2.87 ␣-Terpineol 3.0 2.90 Octanol 2.5 2.25 Nonanol 2.3 2.57 Nerol 1.2 0.25 Perillaldehyde 1.1 1.89 ␤-Sinensal 0.9 0.31 ␣-Sinensal 0.6 0.24 Nootkatone 0.5 0.73 Terpinen-4-ol 0.4 — Octyl acetate 0.4 — Geraniol 0.2 0.43 D-Limonene 0.2 — trans-Limonene oxide 0.2 0.63 Piperitenone 0.2 — Neryl acetate 0.2 — trans-Nerolidol 0.2 0.12 cis-Limonene oxide — 0.23 Elemol — 0.48 trans-2-Nonenal — 0.22 trans-Carveol — 0.38 2,4-Decadienol — 0.29

From Fleisher (1994) and Moyler and Stephens (1992) with permission.

Once extraction is complete, a fresh charge of oil is applied to the column and the separation procedure is repeated (Fleisher, 1994). The compositions of conventional and poroplast- extracted hydrocarbon orange oils are compared in Table 5.15. New solvents have been tested for extraction of essential oils; for instance, 1,1,2-2- tetrafluoroethane (hydrocarbon-134a) (Wilde and McClory, 1994). Although this is a poor solvent, the product is made in the form of a clear, mobile oil with only a very low-solvent residue, and with no degradation of the solutes occurring. In addition, the solvent can be reused without further processing, and the product can be used without additional processing, in contrast to conventional extraction processes that produce a concrete that must be further refined by treatment with ethanol. A relatively new technique for flavour processing is pervaporation. An example of the utility of pervaporation is its use to process bilberry juice using capillary membranes, which successfully separated and concentrated trans-2-hexen-1-ol, which is one of the major flavour- impact components of bilberry (Garcia et al., 2008). One critical factor, whichever extraction method is used, is the influence of harvesting time on the chemical composition of the extract produced (Chalchat et al., 1997), particularly if an enantiomeric analysis is required (Dugo et al., 2001). This is one of the factors that P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

174 Food Flavour Technology

makes the standardisation of the chemical composition of extracts difficult and meeting tight product specifications challenging (Moyler and Moss, 1998). Another important factor is the presence of any pesticide residues (Dugo et al., 1997). The methods developed for some of the more traditional flavour products have also provided a basis for recovery and purification processes for new bioprocesses; for instance, see the review of Fabre et al. (1996) on the processing of 2-phenylethanol. Other essential oils used as food flavours include onion, garlic and leek, mint (from the species Mentha arvaensis, piperita, citrata as well as spicata (spearmint) and pulegium (pennyroyal)), sage, basil, almond, buchu, caraway, celery fennel, parsley, ginger and clove.

5.8 OTHER FLAVOUR CHARACTERISTICS

Sweetness is very important in ﬂavour perception, and sucrose is a plant extract produced by hot water extraction of sugar cane or sugar beet. Similarly, glucose and fructose sweeteners are produced by the action of ␣-amylase and glucoamylase acting on starch extracted from cereals followed by glucose isomerase to produce fructose. In addition, there are the more recently introduced sweeteners, such as AspartameTM and Sucralose. Acidity and sourness in foods and beverages is achieved using organic acids such as citric, malic and tartaric acids. Originally, these were made by extraction from citrus fruits, apples and grapes, respectively. Now, production of citric and malic acids is mostly done by fermentation, although tartaric acid is still extracted from grapes and is an important component of wines. The type of acid used affects the intensity and quality of the ﬂavour. Some less common acids are also consumed; for instance, isocitric acid is the major organic acid in blackberries. Bittering materials can also be derived from plant extracts; in particular hops, used to bitter beer, contain humulone, cohumulone and adhumulone. Other bittering materials include quinine used to bitter soft drinks, and naringin and limonene that make oranges and grapefruit bitter.

5.9 FRAGRANCE USES

Many plant-derived ﬂavour chemicals are also used as aroma chemicals to fragrance products, especially to give ﬂoral, citrus and other notes that are preferred in cosmetics as well as personal care and household products such as soaps and air fresheners, respectively. These include geraniol, linalool, citronellol, jasmine, damascones and rose oxide, as well as strawberry, grape, melon. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 175

5.10 CONCLUSION

This chapter has explored the range of flavour chemicals available from natural resources and the variety of organoleptic effects that they produce, together with an indication of how they can be used to flavour foods and beverages and their commercial value. It illustrates principles involved, some of what is already known, and current trends. In particular, it also indicates how much still needs to be discovered and the potential of some emerging new technologies. An indication of the great variety and range of natural flavour chemicals has been presented. Besides chemical diversity, there are also enormous variations in the intensities of aroma/flavour chemicals, as measured by their detection thresholds, which span a range of 16 orders of magnitude. Even within a single ‘class’ of flavour chemicals, there is a wide variation in flavour/aroma characters. Despite this, some good structureÐfunction relationships have been identified for both taste and aroma molecules. Flavour creation is still something of an art, and much research is still required to put flavours on a truly rational and predictive basis.

REFERENCES

Adedeji, J., Hartman, T.G. and Ho, C.T. (1993) Perfum. Flavour. 18, 24Ð33. Alston, F.H. (1992) Bioformation of Flavours (eds R.L.S. Patterson et al.), Royal Society of Chemistry, Cambridge, pp. 33Ð41. Baigrie, B.D. (1994) Understanding Natural Flavours (eds J.R. Piggott and A. Paterson), Chapman and Hall, London, pp. 268Ð282. Bakker, J. and Law, B.A. (1994) Understanding Natural Flavours (eds J.R. Piggott and A. Paterson), Chapman and Hall, London, pp. 283Ð295. Belitz, H.D. and Grosch, W. (1999) Food Chemistry, 2nd edn, Springer-Verlag, Berlin. Benn, S. (1998) Perfum. Flavour. 23, 5Ð18. Berger, R.G. (1990) Perfum. Flavour. 15, 33Ð39. Blank, I. (1992) Z. Lebensm.-Unters. Forsch. 195, 239Ð245. Boelens, M.H., Boelens, H. and Van Gemert, L.J. (1993) Perfum. Flavour. 18, 1Ð18. Boelens, M.H. and van Gemert, L.J. (1993) Perfum. Flavour. 18, 29Ð39. Bovill, H. (1996) Perfum. Flavour. 21,9. Bruckner,¬ B. and Wyllie, S.G. (eds) (2008) Fruit and Vegetable Flavour: Recent Advances and Future Prospects, CRC Press, Boca Raton, FL. Chalchat, J.C., Michet, A. and Pasquier, B. (1997) Perfum. Flavour. 22, 15. Chassagne, D., Bayonove, C., Crouzet, J. and Baumes, R. (1995) Bioﬂavour 95, Dijon, INRA, Paris (Les Colloques No. 75), pp. 217Ð222. Cheetham, P.S.J. (2004) New Trends and Developments in Biochemical Engineering,Vol.86 (ed. T. Scheper), Springer, Berlin-Heidelberg-New York, pp. 83Ð158. Clark, G.S. (1992) Perfum. Flavour. 17, 107Ð109. Clark, G.S. (1998) Perfum. Flavour. 23, 33Ð48. Clark, G.S. (1999) Perfum. Flavour. 24, 25Ð30. Davidovitch-Rikanati, R., Sitrit, Y., Tadmor, Y., Iijima, Y., Bilenko, N., Bar, E., Carmona, B., Fallik, E., Dudai, N., Simon, J.E., Pichersky, E. and Lewinsohn, E. (2007) Nat. Biotechnol. 25, 899Ð904. Dugo, G., Mondello, L., Cotroneo, A., Bonaccorsi, I. and Lamonica, G. (2001) Perfum. Flavour. 26, 20Ð36. Dugo, G., Saitta, M., DiBella, G. and Dugo, P. (1997) Perfum. Flavour. 22, 33Ð44. Eaton, D.C. (1994) Bioprocess Production of Flavour, Fragrance and Color Ingredients (ed. A. Gabelman), Wiley Interscience, New York, pp. 169Ð204. Eriksson, C. (1994) Understanding Natural Flavours (eds J.R. Piggott and A. Paterson), Chapman and Hall, London, pp. 283Ð297. Fabre, C. and Goma, G. (1999) Perfum. Flavour. 24,1Ð8. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

176 Food Flavour Technology

Fabre, C.E., Blanc, P.J., Marty, A., Goma, G., Sauchon, I. and Voilley, G.A. (1996) Perfum. Flavour. 21, 27Ð39. Fleisher, A. (1994) Perfum. Flavour. 19, 11Ð15. Garcia, V., Diban, N., Gorri, D., Keiski, R., Urtiaga, A. and Ortiz, I. (2008) J. Chem. Technol. Biotechnol. 83, 973Ð982 Granvogle, M., Christlbauer, M. and Schieberle, P. (2004) J. Agric. Food Chem. 52, 2797Ð2802. Guth, H. and Grosch, W. (1991) Fat. Sci. Technol. 93, 335Ð339. Hartmann, H. (1995) Perfum. Flavour. 20, 35Ð42. Hatanaka, A. (1993) Phytochemistry 34, 1201Ð1218. Hausler,¬ A. and Munch,¬ T. (1998) ASM News 63, 551Ð559. Holscher, W. (1992) J. Agric. Food Chem. 40, 655Ð658. Ishikawa, M., Honda, T., Fujita, A., Kurobayashi, Y. and Kitahara, T. (2004) Biosci. Biotechnol. Biochem. 68, 454Ð457. Jorgenson, U., Hansen, M., Christensen, L.P, Jensen, K. and Kaack, K. (2000) J. Agric. Food. Chem. 48, 2376Ð2383. Kolor, M.G. (1983) J. Agric. Food Chem. 31, 1127Ð1129. Kopke, T. and Mosandl, A. (1992) Z. Lebensm.-Unters. Forsch. 194, 327Ð376. Koppenhoefer, B., Behnisch, R., Epperlein, U. and Holzschuh, H. (1994) Perfum. Flavour. 19, 1Ð10. Kuo, M.C. and Ho, C.T. (1992) J. Agric Food Chem. 40, 111Ð117, 1906Ð1910. Land, D.G. (1994) Understanding Natural Flavours (eds J.R. Piggott and A. Paterson), Chapman and Hall, London, pp. 298Ð306. Li, X., Staszewiski, L., Xu, H., Durick, K., Zoller, M. and Adler, E. (2002) Proc. Natl. Acad. Sci. USA 99, 4692Ð4696. Lindsay, R.C. and Rippe, J.K. (1986) In: Biogeneration of Aromas (eds R. Croteau and T. Parliament), Symposium Series 317, American Chemical Society, Washington, DC, pp. 286Ð308. Lusier, J-L., Buettner, H., Volker, S., Rausis, T. and Frey, U. (2008) J. Agric. Food Chem. 56, 2883Ð2887. McGee, T. and Purzycki, K. (1999) Perfum. Flavour. 24, 1Ð10. Malnic, B., Hirono, J., Sato, T. and Buck, L.B. (1999) Cell 96, 713Ð716. Marriott, R.J. (1988) Flavours and Fragrances: A World Perspective, Proc. 10th Int. Congress of Essential Oils (eds B.M. Lawrence et al.), Elsevier Science, Amsterdam, pp. 387Ð392. Mayer, F., Takeoka, G.R, Buttery, R.G., Whitehand, L.C., Naim, M. and Rabinowithch, H.D. (2008) J. Agric. Food Chem. 56, 3749Ð3757. Mookherjee, B. and Wilson, A. (1990) Perfum. Flavour. 15, 27Ð49. Moyler, D. and Moss, N. (1998) Perfum. Flavour. 23, 37Ð39. Moyler, D.A. and Stephens, M.A. (1992) Perfum. Flavour. 23, 37Ð39. Mulders, E.J. (1976) Chem. Ind. (London) 613Ð614. Naf, R. and Velluz, A. (2000) Flavour Frag. J. 15, 329Ð334. Naf, R., Vellez, A. and Thommen, W. (1990) Tetrahedron Lett. 31, 6521Ð6522. Nelson, G, Hoon, M.A., Chandrasakar, J., Zhang, Y., Ryba, N.J. and Zuker, C.S. (2001) Cell 106, 381Ð390. Ngai, J., Dowling, M.M., Buck, L., Axel, R. and Chess, A. (1993) Cell 72, 657Ð666. Ozeck, M., Brust, P., Xu, H. and Servant, G. (2004) Eur. J. Pharmacol. 489, 139Ð149. Rombaults, J.E. (1952) Proc. Soc. Appl. Bacteriol. 15, 103Ð111. Ropkins, K. and Beck, A.J. (2000) Trends Food Sci. Technol. 11, 10Ð21. Rowe, D.J. and Tangel, B. (1999) Perfum. Flavour. 24, 36Ð44. Schroeder, M., Pollinger-Zierler, B., Aichernig, N., Siegmund, B. and Guebitz, G.M. (2008) J. Agric. Food Chem. 56, 2485Ð2489. Sell, C. (2001) Perfum. Flavour. 26,2Ð7. Semmelroch, P. and Grosch, W. (1996) J. Agric. Food Chem. 44, 537Ð543. Shallenberger, R.S. (1996) Food Chem. 56, 209Ð214. Sinki, G., Assaf, R. and Lombardo, J. (1997) Perfum. Flavour. 22, 23Ð31. Somogyi, L. (1996) Chem. Ind. March 170Ð173. Stofberg, J. (1983) Perfum. Flavour. 8, 61Ð64. Swaine, R.L. (1993) Perfum. Flavour. 18, 35Ð38. Todd, H. (1998) Perfum. Flavour. 23, 23Ð25. Tominaga, W. and Dubourdieu, D. (2000) Identiﬁcation of cysteinylated aroma precursors of certain volatile thiols in passion fruit juice. J. Agric. Food chem. 48, 2874Ð2876. Tressl, R. and Silwar, R. (1981) J. Agric Food Chem. 29, 1078Ð1082. P1: SFK/UKS P2: SFK c05 BLBK221-Taylor/Linforth November 23, 2009 13:38 Printer Name: Yet to Come

Natural sources of ﬂavours 177

Triballeau, N., Van Name, E., Laslier, G., Cai, D., Pollard, G., Soremsoen, P.W., Hoffmann, R., Bertrand, H-O., Ngai, J. and Acher, F.C (2008) Neuron 60, 767Ð774. Tyrell, M. (1995) Perfum. Flavour. 20, 13Ð22. Venkateshwarlu, G., Chandravandana, M.V., Pandey, M., Tewari, R.P. and Selvaraj, Y. (2000) Flavour Frag. J. 15, 320Ð322. Verhagen, L.C. (1994) Understanding Natural Flavours (eds J.R. Piggott and A. Paterson), Chapman and Hall, London, pp. 283Ð297. Wang, D., Yoshimura, T., Kubota, K. and Kobayashi, A. (2000) J. Agric. Food Chem. 48, 5411Ð5418. Whitehead, I.M., Muller, B.L. and Dean, C. (1995) Cereal Food. World 40, 193Ð197. Wilde, P.F. and McClory, P.G. (1994) Perfum. Flavour. 19, 25Ð26. Winterhalter, P. and Schreier, P. (1994) Flavour Frag. 9, 281Ð297. Wu, S. (2006) Nat. Biotechnol. 24, 1441Ð1447. Yang, D.S., Shewfelt, R.L., Lee, K-S. and Kays, S.J. (2008) J. Agric. Food Chem. 56, 2780Ð2787. P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

6 Useful principles to predict the performance of polymeric ﬂavour delivery systems Daniel Benczedi´

6.1 OVERVIEW

In this chapter, we review semi-empirical principles of chemical engineering that are particularly useful to understand the phase behaviour of glassy polymers applied to encapsulate flavours and fragrances. Hildebrand’s solubility parameter is used to outline how the polarity of a polymer increases its cohesive energy density, its density, its glass transition temperature and how it reduces its permeability towards apolar molecules such as gaseous oxygen and most flavouring ingredients. Flory’s model of polymer solutions is coupled to a generalised Freundlich adsorption model to highlight the limiting phase behaviour of polar polymer glasses at low partial pressures of water and of hydrogels at the saturation vapour pressure of water. This approach suggests that sigmoidal vapour sorption phenomenology is the macroscopic fingerprint of glassy materials. Below the glass transition point, water sorption is therefore assumed to take place on the pre-existing sites formed by an excess free volume, which characterises glassy polymers. This excess free volume needs typically to be minimised in controlled release applications to minimise the diffusion of oxygen and flavour molecules. Mass transport is further described with Fick’s laws of diffusion to highlight the effects of polymer cross-linking or plasticisation, as induced by temperature or partial pressure of water. Diffusion-controlled kinetics, scaling with the square root of time, is distinguished from zero-order kinetics observed when mass transport is polymer relaxation-controlled.

6.2 INTRODUCTION

The encapsulation of flavours was initially motivated by the need to protect volatile food components from evaporation and oxidation by locking them in a solid carrier able to dissolve rapidly in water (Schultz et al., 1956). Nowadays, more emphasis is laid on slowing aqueous dissolution down and on triggered release systems activated typically by heat and mechanical stress. Controlled flavour release thus provides new solutions to the food industry and enables the development of innovative consumer goods. In this context, the primary objective of flavour delivery systems is to maximise flavour perception by generating impact and persistence when needed. Flavours are traditionally encapsulated by spray-drying an aqueous polymer solution, and the resulting glassy powders have a fine granulometry, favouring quick aqueous dissolution. The selective retention of flavour molecules as water evaporation proceeds during P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

Useful principles to predict the performance of polymeric ﬂavour delivery systems 179

spray-drying is reviewed elsewhere (Rosenberg et al., 1990; Coumans et al., 1994; King, 1995; Liu et al., 2000). A coarser granulometry can be achieved by other encapsulation processes in order to influence the kinetics of aqueous dissolution or of flavour oxidation by varying the surface to volume ratio of flavour delivery systems, as reviewed elsewhere (Kondo, 1979; Bakan, 1986; Jackson and Lee, 1991; Shahidi and Han, 1993; Risch and Reineccius, 1995; Benczedi,« 1999; Gibbs et al., 1999; Qui and Xu, 1999; Ubbink and Schoonman, 2004; Gouin, 2004; Chen et al., 2006). The formulation ingredients commonly used to design edible flavour delivery systems include polysaccharides, proteins, fats and waxes used sometimes in combination to reduce the migration of oxygen, flavours or moisture (Kester and Fennema, 1986; Miller and Krochta, 1997). As we shall see, flavours are generally well retained and protected from oxygen by hydrophilic polymers whereas hydrophobic ones are used to protect delivery systems from moisture. To be commercially successful, the selected encapsulation process has to be cost- effective and adapted to the phase behaviour of the selected barrier material. Starch hydrolysates are excellent oxygen and flavour barriers used frequently in combination with octenyl succinylated starches or gum acacia to provide some lipophilic character to the carrier (Buffo and Reineccius, 2000). To improve oxidation stability of orange oil, starch hydrolysates are used in combination with lower molecular weight plasticisers such as amorphous sucrose or corn starch syrups (Anandaraman and Reineccius, 1986). This may at first appear as counter-intuitive because, above the glass transition temperature, Tg, plasticisation is generally coupled to a density reduction followed by an increase in permeability. The opposite seems to take place below Tg when the addition of a plasticiser results in a densification of the polymer, as observed at low water concentrations in starch (Guo, 1994; Benczedi,« 2001). The cross-linking of polymers is useful to slow down or prevent aqueous dissolution and to provide systems that release flavour on exposure to heat, to chemical or mechanical stress. For instance, gelatin and starch build thermally reversible gels, while alginates and pectins can be cross-linked by calcium ions and hydrophobic effects (Harris, 1990). Chemical cross- linking typically leads to novel food products, which need to be registered as such, as in the example of gelatin and gum acacia cross-linked by glutaraldehyde. The semi-empirical principles of chemical engineering presented in the following are particularly useful to predict the performance and limitations of controlled release systems (Prausnitz, 1999). Hildebrand’s solubility parameter is used to estimate the polarity or cohesive energy density of flavours and polymers to assess their mutual thermodynamic compatibility (Hildebrand et al., 1970; Barton, 1991). Flory’s model of polymer solutions is coupled to a generalised Freundlich adsorption to highlight the limiting phase behaviour of glassy polymers at low partial pressures of water, and Fick’s laws of diffusion are used to interpret mass transport in polymers (Flory, 1953; Atkins, 1994).

6.3 COMPATIBILITY AND COHESION

The second law of thermodynamics states that the mixing of molecules is spontaneous when the contribution of the heat of mixing, ⌬ H, and of the entropy of mixing, ⌬ S, leads to a negative Gibbs free energy of mixing, ⌬ G.

⌬ G = ⌬ H − T ⌬ S Ͻ 0 (6.1) P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

180 Food Flavour Technology

In the absence of any heat of mixing, the sole driving force is the conﬁgurational entropy gain that is proportional to the number of molecules involved and is thus less when the mixture contains a polymeric component. The sign and absolute value of ⌬ H is thus critical for spontaneous mixing to occur, more particularly when polymers are involved. Following Hildebrand, the heat of mixing is approximated as follows:

2 ⌬ H = (␦1 − ␦2) × V1␾2 (6.2)

where V1 is the molar volume of the solvent, ␾2 is the volume fraction of polymer and ␦ is the solubility parameter deﬁned as the square root of the cohesive energy density, c.e.d. (V is the molar volume, R is the universal gas constant, T is the temperature and ⌬ HV is the enthalpy of vaporisation).

␦ = (c.e.d.)1/2 = (⌬ H V − RT)1/2/V 1/2 (6.3)

For two chemical species to be miscible, the square of the difference between their solubility parameters should be small. If this is not the case the difference can be minimised with a co-solvent of intermediate ␦ value. Originally formulated for regular solutions, Hilde- brand’s approach has been extended to include associative attractions (hydrogen bonding) in addition to non-polar and polar attractions. As a consequence of the specific nature of such forces, materials with weak cohesive forces cannot make the associations needed to dissolve materials with strong attractive forces (Barton, 1991). In the polarity scale presented in Table 6.1, heptane represents a molecule held together by apolar attractions, and water one held together by apolar, polar and associative attractions. A value estimated for gaseous diatomic oxygen is added to emphasise its apolar character. It can be seen that the range of hydrophobicity characteristics of essential oils spans from limonene R to eugenol, although some flavour ingredients such as Furaneol show a hydrophilicity closer to that of ethanol and, of sucrose, a widespread carbohydrate flavour carrier ingredient. The proximity of solubility parameters of sucrose and starch (Table 6.2) suggests that the former is a good plasticiser for the latter. If the vaporisation enthalpy is not available experimentally,

Table 6.1 Molecular weight, M, boiling point, T b, and solubility parameters, ◦ δ, of selected chemicals at 25 C.

◦ 1/2 M (Da) Tb ( C) δ (MPa) ◦ Diatomic oxygen 32 Gaseous at 25 C11.7 n-Heptane 100 98 15.1 D-Limonenea 136 176 16.6 D-Carvonea 150 230 19.0 n-Octanol 130 194 21.1 Acetaldehyde 44 21 21.1 Eugenola 164 255 23.3 Ethanol 46 79 26.4 ◦ Furaneolb 128 Crystalline at 25 C 26.8 ◦ Sucrose 342 Crystalline at 25 C 45.1 Water 18 100 48.0

δ calculated with group contributions compiled by Beerbower and listed by Barton (1991). a Barton (1991) and Peppas and Am Ende (1997). b Firmenich trade name of 4-hydroxy-2,5-dimethyl-3(2H)-furanone. P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

Useful principles to predict the performance of polymeric ﬂavour delivery systems 181

◦ Table 6.2 Solubility parameter δ,density␳, difference between experimental temperature (25 C) and − glass transition temperature (T T g), and rounded oxygen P(O2) and water permeability P(H2O) of polymers.

1/2 3 − ◦ a a δ (MPa) ρ (g/cm ) T Tg ( C) P(O2) P(H2O) − Polyethylene 16 1.0 Ͼ0101 102 − Poly(vinyl acetate) 19 1.2 ≈0102 104 − Poly(vinyl alcohol) 30 1.3 Ͻ0106 Dissolves Starch 35b 1.5b 0b — Dissolves

Miller and Krochta (1997), Benczedi´ et al. (1998) and Brandrup et al. (1999). a Permeability of polymers is expressed in cm3 µm/(m2 day kPa).

as in the case of polymers, indirect methods are used to estimate ␦ experimentally and group contribution methods are available to calculate ␦ for a molecule of known chemical structure (van Krevelen, 1976; Barton, 1991; Hu et al., 1997; Tse et al., 1999). Hildebrand’s approach is thus useful to predict polymerÐsolvent compatibility, to rank polymers according to their hydrophilicity and, more generally, to predict the impact of formulation ingredients on the morphology and performance of delivery systems (Michaels et al., 1975; Rowe, 1988; Sakellariou and Rowe, 1991; Archer, 1992; Moldenhauer and Nairn, 1992; Peppas and Am Ende, 1997; Rodriguez et al., 2000). Table 6.2 shows the solubility parameter of selected polymers in which polyethylene and starch represent the lipophilic and hydrophilic end of the polarity scale, respectively. The increase of polarity corresponds to an increase of density and glass transition temperature. It is also apparent from Table 6.2 that the polarity of a polymer increases its water permeability coefficients while its oxygen permeability coefficient decreases. The glass transition temperature, Tg, of polymers is therefore expected to increase not only with their molecular weight but also with their polarity and stiffness. The addition of a low-molecular-weight diluent decreases the Tg of a polymer by so- called plasticisation. Most flavour molecules are poor plasticisers for the polar polymers typically used to encapsulate them. On the other hand, water is a powerful plasticiser for polar polymeric carriers, as suggested by the magnitude of their solubility parameters, and the Tg of hydrophilic polymers is therefore dependent on the partial vapour pressure of water (Benczedi« et al., 1998). The surprising densification of polymers by plasticiser molecules observed below their glass transition point is reported as polymer antiplasticisation in the literature because it affects the mechanical and barrier properties of polymer glasses in ways opposite to that expected of plasticisation (Fischer et al., 1985). Its occurrence can thus not be predicted by Hildebrand’s polarity scale. Antiplasticisation is usually described with free-volume models, assuming that the specific volume of a molecular liquid always contains a fraction of space left unoccupied by the molecules. The permanent redistribution of this free volume by random thermal agitation causes density fluctuations by which molecular transport is assumed to take place in liquids. In polymer glasses there is an excess free-volume component, which is not redistributed by thermal agitation. As we shall see, the occurrence of an excess free volume can be detected by the particular sigmoidal vapour and gas sorption phenomenology of glassy polymers. The latter suggests that an exothermal sorption takes place at fixed pre-existing sites composing the excess free volume of glassy polymeric materials (Meares, 1954; Cohen and Turnbull, 1959; Berens, 1975; Fan and Singh, 1989; Debenedetti, 1995; Benczedi,« 2001). P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

182 Food Flavour Technology

6.4 SORPTION AND SWELLING

The solubility of volatile components is characterised in sorption experiments by measuring their equilibrium volume fraction of solvent (indexed with 1), ␾1, sorbed by polymer (indexed 0 with 2) at any given solvent activity, a1 = p1/p1 . The deviation from ideal solution property is deﬁned by the activity coefﬁcient, ␥ 1 = a1/␾1, an excess thermodynamic function relating the actual free energy change on mixing to the free energy change observed in the ideal case, ⌬ GE = ⌬ Gactual − ⌬ Gideal. ⌬ GE ⌬ H E ⌬ SE ln ␥ = = − (6.4) 1 RT RT R

In the ideal case, ln ␥ 1 = 0 and ␥ 1 = 1 over the entire concentration range and the distribution of molecules is random, as predicted by Raoult’s law for non-interacting and equal-sized molecules. Positive and negative deviations of ␥ 1 from 1 correspond to en- dothermal and exothermal deviations from ideal mixing, as predicted by the quasi-chemical approximation (Guggenheim, 1952). If ␥ 1 is invariant with concentration at inﬁnite dilution ∞ of the solvent in the polymer, the mole fraction-based limiting activity coefﬁcient ␥ 1x is given by

␥ ∞ = p1 1x (6.5) Hx1

where H is the Henry’s law constant (Deshpande et al., 1974). Because of the low conﬁgurational entropy gain characterising polymerÐsolvent mixtures, ␥ 1 is always larger than predicted by Raoult’s law, as expressed semi-empirically by Flory and Huggins with a concentration-independent interaction parameter, ␹ (de Gennes, 1979).

2 ln ␥1 = (␾2 + ␹␾2) (6.6)

∞ For polymerÐsolvent mixtures characterised by ␹ = 0, Equation 6.6 yields ln ␥ 1 = 1or ∞ equivalently ␥ 1 ≈ 2.7. Rubbery polymerÐsolvent mixtures (T Ͼ Tg) are typically characterised by 0 Ͻ␹ Ͻ1/2, and polymer phase separation is predicted at larger ␹ values. Solvent uptake by rubbery polymers is thus usually endothermic over the entire concentration range, and ␹ shows typically only weak composition dependence (Guggenheim, 1952; Flory, 1953). An extension of the model to more than two components is useful to describe polymer phase separation, as occurring during simple coacervation (Tompa, 1956; Hsu and Prausnitz, 1973). In contrast, the water sorption isotherm of polar polymers is typically sigmoidal (type IV sorption) at ambient temperatures because these polymers are usually frozen in a glassy state at low solvent concentrations (Benson and Seehof, 1951; McLaren and Rowen, 1951). This exothermal deviation from Flory’s approximation observed with glassy polymers corresponds to the case where ln ␥ ∞ Ͻ 1. At higher plasticiser concentrations and/or temperatures, these solids are brought above their glass transition point and the Flory’s phenomenology is recovered at (T − Tg) Ͼ 0. As a further consequence, the activity coefﬁcient is no longer independent of concentration, as postulated in Flory’s approximation. Sigmoidal sorption phenomenology can be reproduced by coupling Flory’s approximation, with a generalised Freundlich model accounting for solvent adsorption at the low solvent concentrations, where T Ͻ Tg (Sips, 1948, 1952; Benczedi« et al., 1998). This P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

Useful principles to predict the performance of polymeric ﬂavour delivery systems 183

1 c T < Tg χ = 0 T > T g χ = 1/2

p φ 1 γ < 1 γ >1

χ = 1

0 0 1 a1 ␾ Fig. 6.1 Sorption isotherms showing the volume fraction of solvent in a polymer 1 as a function of the ␥ = solvent activity, a1. The dashed diagonal line represents Raoult’s law, and the activity coefﬁcient, 1 ␾ a1/ 1, is thus smaller than 1 above the line (exothermal mixing) and greater than 1 below it (endothermic mixing). The phase behaviour predicted with the Flory’s approximation is shown by setting ␹ = 0 (upper dashed line), ␹ = 1/2 (intermediate dashed line) and ␹ = 1 (lower dashed line); the latter illustrates ﬁnite polymer swelling in the phase domain characterised by solvent clustering, i.e. ␹Ͼ1/2. The sigmoidal (plain line) is typical for water sorption by a hydrophilic polymer frozen in a glassy state at low water activity (McLaren and Rowen, 1951; Benczedi´ et al., 1998). The corresponding gas adsorption phenomenology in polymer glasses is shown in the insert, where c is the ratio of gas to polymer volume at standard temperature, T, plotted as a function of the gas pressure, p (Paul, 1985; Ganesh et al., 1992).

approach suggests that the exothermal solvent uptake characterised by ln ␥ 1 Ͻ 1atlow solvent concentrations proceeds by adsorption into pre-existing sites provided by the excess free volume of polymers at T Ͻ Tg. The same phenomenology is reported for gas adsorption by glassy polymers with a sorption following Henry’s law above Tg, whereas below Tg, the gas sorption first follows Langmuir adsorption phenomenology prior to recovery of the Henry mode at higher partial pressures (Paul, 1985; Ganesh et al., 1992). At the higher solvent concentrations corresponding to (T − Tg) Ͼ 0, polymerÐsolvent mixtures recover positive ␹ values. Swollen hydrogels characterised by ␹ > 1/2 can be observed at high solvent concentrations if a complete solubilisation of the polymeric network is hindered by residual polymerÐpolymer interactions (Zimm and Lundberg, 1955; Berens, 1975; Stannett et al., 1980; Benczedi« et al., 1998). This particular case of finite polymer swelling without complete polymer solubilisation is illustrated in Fig. 6.1 by setting ␹ >1/2 in Flory’s approximation. At equilibrium, the osmotic pressure vanishes as the activity of the solvent inside the gel becomes identical to that outside the gel. The Flory treatment assumes that the total osmotic pressure of a non-ionic gel is the result of a swelling force, driven by polymerÐsolvent affinity, and opposed by a contracting (shrinking) force due to the rubber-like elasticity of the three-dimensional polymeric network. The swelling ratio, Q, defined as volume ratio of swollen to dry gel, decreases when the density of cross-links increases, or equivalently, as the polymer chain length between cross-links is reduced (Flory, 1953). It is also worth noting that ␹ decreases typically with temperature reflecting an enhanced solubilisation unless the polymer phase behaviour is characterised by a lower critical solution P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

184 Food Flavour Technology

temperature, as in the case, for instance, of some cellulose ethers in water (Doelker, 1993). Last but not least, electrostatic interactions need to be considered in systems containing acidic (electrophilic) functional groups such as carboxyl groups ionised at pH Ͼ pKa and basic (nucleophilic) functional groups such as amino groups ionised at pH Ͻ pKb. The swelling is then driven by the repulsion between the ﬁxed ionised functional groups of polymers until they are neutralised by hydration following Donnan equilibrium, i.e. electroneutrality in each phase and phase equilibrium for each mobile ion (Khare and Peppas, 1995; Prausnitz, 1995). Electrosteric stabilisation and complex coacervation are examples of industry applications in which polyelectrolytes play a major role.

6.5 DIFFUSION AND RELEASE

The transport of gases or liquids through amorphous polymeric barriers is characterised by a permeability coefficient, P. The permeation of molecules typically takes place in the amorphous domains of polymers, while the crystalline domains are impermeable to transport of molecules (Paul, 1985; Avranitoyannis et al., 1994). The thermodynamic component of the permeability coefficient, P, describing how many solvent molecules are soluble in a given rubbery polymer, is the solubility coefficient, S, the reciprocal of the Henry’s law constant, H = S−1. The kinetic component of P is the diffusion coefficient, D, describing how fast molecules translate in a given polymer. At infinite dilution of the permeating molecule, the permeability coefficient is the product of solubility and diffusion coefficients.

P = SD (6.7)

Permeation data are then interpreted either using an activated process approach or using the free-volume approach. In an activated process, the energy for diffusion is postulated to arise from the need to separate polymer segments sufficiently to allow the permeating molecule to make a unit diffusion jump. In the free-volume approach, the permeating molecule is postulated to move from one place to another only when local free volume around this molecule exceeds a certain critical value (Paul, 1985). Under stationary conditions (steady-state equilibrium), Fick’s first law of diffusion defines the rate of mass transfer across an isotropic material as proportional to the concentration gradient measured perpendicular to the section. Beyond stationary conditions, Fick’s second law relates the rate of change of concentration at a point to the spatial variation of concentration at that point (Atkins, 1994). The following power law is used to illustrate how Fick’s laws of diffusion can be used to analyse the release of functional molecules (flavour, fragrance, drug, etc.) from a polymer or the uptake of solvent (water) by a polymer:

Mt n = knt (6.8) M∞

where Mt is the amount of molecules released at time t, M∞ is the equilibrium value at large t (sum of active component released or of solvent absorbed), k is a rate constant and n is a scaling exponent. If a linear release curve is observed on plotting Mt/M∞ as a function of t1/2, the release or uptake is diffusion-controlled. A scaling exponent n of 1/2 is thus the ﬁngerprint of the Brownian motion of molecules diffusing in a stationary concentration gradient as predicted by Fick’s diffusion law (case I transport) (Fan and Singh, 1989). P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

Useful principles to predict the performance of polymeric ﬂavour delivery systems 185

T > T << g De 1 De << 1 n = 0.5 n = 0.5 D = D0 D = D(a ) T = Tg 1

De >> 1 n = 0.5 D = D0

De >> 1 0.1< De <10 n = 0.5 n > 0.5 D = D(a ) T < Tg 1 D = D(a1 ; t)

0 1 a1 − Fig. 6.2 Schematic temperature–solvent activity (T a1) diagram of mass transport in polymer–solvent systems (adapted from Hopfenberg and Frisch, 1969). The solid line is the glass transition temperature, T g, of the system. The dashed lines delimit domains of mass transport characterised by a Deborah number, De, a scaling exponent in Equation (6.9) and a diffusion coefﬁcient, which is either independent of 0 concentration and time, D , dependent on concentration but independent of time, D(a1), or dependent on both concentration and time, D (a1; t).

If a linear sorption curve is obtained by plotting Mt/M∞ as a function of t, the amount released is constant in time (n = 1) and the release kinetics are controlled by the relaxation of the polymerÐsolvent system (case II transport: zero-order kinetics). For values of the exponent n between 1/2 and 1, the release reﬂects both diffusion and relaxation of the polymerÐsolvent system (anomalous transport). The scaling exponent n is thus larger than 1/2 whenever it reﬂects the time dependence of polymer relaxation around the glass transition point of the polymerÐsolvent system (Fujita, 1961; Crank and Park, 1968). This is illustrated in Fig. 6.2 with a dimensionless Deborah number, De, used to distinguish the limiting cases of mass transport in polymers by the ratio of two characteristic times, a characteristic relaxation time for the polymerÐsolvent system, ␭, and a characteristic diffusion time, ␪ (Vrentas and Duda, 1975).

␭ = De ␪ (6.9)

In the liquid state characterised by De 1, diffusion occurs within a viscous medium, while in the solid state characterised by De 1, diffusion occurs within an elastic medium. As shown in Fig. 6.2, the diffusion coefﬁcient is only time dependent for 0.1 De 10 in the viscoelastic regime around the glass transition points where anomalous mass transport is observed because ␭ and ␪ are of the same order of magnitude (Ferry, 1980; Frisch, 1980; Fan and Singh, 1989). As shown in Table 6.2, hydrophilic polymers are at the same time good oxygen (␦ = 11.7 1/2 1/2 MPa ) barriers but poor water (␦ = 48 MPa ) barriers and oxygen permeability, P(O2), in polyethylene (␦ = 16 MPa1/2) is ﬁve orders of magnitude larger than in polyvinyl alcohol (␦ = 30 MPa1/2) (Salame and Steingiser, 1977; Miller and Krochta, 1997). As the size of P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

186 Food Flavour Technology

the permeating molecule increases, P is expected to decrease because a linear increase in S following an elevation of the boiling point is more than compensated by an exponential drop in D (Naylor et al., 1989). Diffusion is related to the mobility of polymer chains and thus to the temperature of the system relative to its glass transition temperature. As the temperature is lowered and approaches Tg, the free volume available for diffusion decreases. Diffusion coefficients exhibit either a continuous or a discontinuous change at Tg. The change is continuous if the size of the diffusing molecule is smaller than the average void size and diffusion occurs by localised, activated jumps from one pre-existing cavity to another. As the size of the diffusing molecule grows, the number of polymer segment rearrangements involved in an activated jump increases and the process becomes dependent on the excess (sub-Tg) free volume of the system, as reviewed elsewhere (Fan and Singh, 1989). An increase of the size of the permeating molecules decreases D by only two orders 1/2 of magnitude in natural rubber (T Ͼ Tg, ␦ = 17 MPa ) prior to reaching an asymptotic minimum value, while D decreases by ten orders of magnitude in glassy poly(vinyl chloride) 1/2 (T Ͻ Tg, ␦ = 20 MPa ). D is reported to increase by up to three orders of magnitude when the permeating molecule is elongated rather than spherical (Naylor et al., 1989). The effect of temperature on D is of the same order when a packaging material is brought to retort temperature or when a hydrophilic polymer is plasticised by water following exposure to high relative humidity (DeLassus et al., 1988; DeLassus, 1994). The plasticisation of the polymer makes the diffusion coefficient time dependent as reflected by n Ͼ 1/2 (Yapel et al., 1994; Beck and Tomka, 1997). A diffusion-controlled permeation (n = 1/2) of gas or vapour in rubbery polymers occurs when plasticisation or swelling is negligible. It can be thus observed either when the chemical affinity is insufficient ((␦1 − ␦2) 0) or at very low vapour pressures (infinite dilution). As a consequence, the diffusion coefficient is invariant with time (D(a1) in Fig. 6.2) or invariant with time and composition (D0 in Fig. 6.2), respectively. In packaging applications, the partial pressure of volatile compounds is often less than 0.2 and insufficient for plasticisation to affect the permeability as illustrated by the Fickian diffusion of d-limonene in rubbery polyethylene (Hernandez, 1986; DeLassus, 1994). In controlled release applications, the partial pressure of flavours is much higher and polar polymers of appropriate polarity and molecular weight are needed to prevent a flavour-induced plasticisation and to minimise oxygen permeability (Anandaraman and Reineccius, 1986). Diffusion coefficients can also be adjusted by cross-linking, which reduces polymer segment mobility and is able to change the release kinetics from diffusion-controlled (n = 1/2) at low cross-link density, to polymer relaxation-controlled (n = 1) at high cross- link density, as shown for eugenol release in ethanol from gels of poly(2-hydroxyehtyl methacrylate) (Peppas and Am Ende, 1997). Finally, a variation of the nature and degree of substitution of cellulose ethers is also able to affect permeation and release. When keeping the same active compound and the same surrounding solvent, methyl cellulose exhibits diffusion-controlled release (n ≈ 1/2), whereas release from hydroxyethyl-, hydroxypropyl- and hydroxypropyl methylcellulose is characterised by 1/2 Ͻ n Ͻ 1 (Rodriguez et al., 2000). These observations suggest that the same strategy can be applied to provide sustained flavour release in high moisture conditions if the polymers are selected to limit carrier dissolution above 70Ð80% relative humidity and in bulk water. The choice of the right delivery system is ultimately a balance between performance (e.g. oxygen and moisture stability) and cost. P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

Useful principles to predict the performance of polymeric ﬂavour delivery systems 187

REFERENCES

Anandaraman, S. and Reineccius, G. (1986) Stability of encapsulated orange peel oil. Food Technol. 40(11), 88Ð93. Archer, W.L. (1992) Hansen solubility parameter for selected cellulose ether derivatives and their use in pharmaceutical industry. Drug Dev. Ind. Pharm. 18(5), 599Ð616. Atkins, P.W. (1994) Physical Chemistry, 5th edn, Oxford University Press, Oxford. Avranitoyannis, I., Kalichevsky, M., Blanshard, J.M.V. and Psomiadou, E. (1994) Study of diffusion and permeation of gases in undrawn and uniaxially drawn films made from potato and rice starch conditioned at different relative humidities. Carbohydr. Polym. 24, 1Ð15. Bakan, J.A. (1986) Microencapsulation. In: The Theory and Practice of Industrial Pharmacy, 3rd edn (eds L. Lachman, H.A. Lieberman and J.L. Kanig), Lea & Febiger, Philadelphia, PA, pp. 412Ð429. Barton, A.F.M. (1991) Handbook of Solubility Parameters and Other Cohesion Parameters, 2nd edn, CRC Press, Boca Raton, FL. Beck, M.I. and Tomka, I. (1997) Effects of the degree of substitution in ethylcellulose on the clustering of sorbed water. J. Macromol. Sci. Phys. B36(1), 19Ð39. Benczedi,« D. (1999) Estimation of the free volume of starch-water barriers. Trends Food Sci. Technol. 10, 21Ð24. Benczedi,« D. (2001) Plasticization and antiplasticization of starch glasses by water. In: ISOPOW 8: Water Science for Food Health, Agriculture and Environment (eds Z. Berk, R.B. Leslie, P.J. Lillford and S. Mizrahi), Technomic Publishing, Lancaster, PA, pp. 273Ð282. Benczedi,« D., Tomka, I. and Escher, F. (1998) Thermodynamics of amorphous starch-water systems. Macro- molecules 31(9), 3055Ð3074. Benson, S.W. and Seehof, J.M. (1951) The surface of proteins: IV. sorption of polar gases. J. Am. Chem. Soc. 73, 5053Ð5058. Berens, A.R. (1975) Solubility of vinyl chloride in poly(vinyl chloride). Angew. Makromol. Chem. 47, 97Ð110. Brandrup, J., Immergut, E.H. and Grulke, E.A. (eds) (1999) Handbook of Polymers, 4th edn, John Wiley & Sons, New York. Buffo, R. and Reineccius, G.A. (2000) Optimization of gum acacia-modified starch-maltodextrin blends for the spray drying of flavours. Perfum. Flavour. 25, 37Ð51. Chen, L., Remondetto, E. and Subirade, M. (2006) Food protein-based materials as neutraceutical delivery systems. Trends Food Sci. Technol. 17, 272Ð283. Cohen, M.H. and Turnbull, D. (1959) Molecular transport in liquids and glasses. J. Chem. Phys. 31(5), 1164Ð1169. Coumans, W.J., Kerkhof, P.J.A.M. and Bruin, S. (1994) Theoretical and practical aspects of aroma retention in spray-drying and freeze-drying. Dry. Technol. 12(1Ð2), 99Ð149. Crank, J. and Park, G.S. (eds) (1968) DiffusioninPolymers, Academic, New-York. de Gennes, P.G. (ed.) (1979) Scaling Concepts in Polymer Physics. Cornell University Press, Ithaca, NY. Debenedetti, P.G.(1995) Metastable Liquids: Concepts and Principles. Princeton University Press, Princeton, NJ. DeLassus, P.T. (1994) Permeation of flavours and aroma through glassy polymers. Tappi J. 77(1), 109Ð 113. DeLassus, P.T., Strandburg, G. and Howell, B.A. (1988) Flavour and aroma permeation in barrier film: the effects of high temperature and high humidity. Tappi J. 71(11), 177Ð181. Deshpande, D.D., Patterson, D., Schreiber, H.P. and Su, C.S. (1974) Thermodynamic interactions in polymer systems by gas-liquid chromatography. Macromolecules 7(4), 530Ð535. Doelker, E. (1993) Cellulose derivatives. In: Advances in Polymer Science,Vol.107, Springer-Verlag, Berlin, pp. 199Ð265. Fan, L.T. and Singh, S.K. (1989) Controlled release: a quantitative treatment. In: Polymer Properties and Applications,Vol.13, Springer-Verlag, Berlin. Ferry, J.D. (1980) Viscoelastic Properties of Polymers, 3rd edn, John Wiley & Sons, New York. Fischer, E.W., Hellmann, G.P., Spiess, H.W., Horth,¬ F.J., Ecarius, U. and Wehrle, M. (1985) Mechanical properties, molecular motion and density fluctuations in polymer-additive mixtures. Makromol. Chem. Suppl. 12, 189Ð214. Flory, P.J. (1953) Principles of Polymer Chemistry. Cornell University Press, Ithaca, NY. Frisch, H.L. (1980) Sorption and transport in glassy polymers: a review. Polym. Eng. Sci. 20(1), 2Ð13. P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

188 Food Flavour Technology

Fujita, H. (1961) Diffusion in polymer-diluent systems. Fortschr. Hochpolym.-Forsch. 3, 1Ð47. Ganesh, K., Nagarajan, R. and Duda, J.L. (1992) Rate of gas transport in glassy polymers: a free volume based predictive model. Ind. Eng. Chem. Res. 31, 746Ð755. Gibbs, B.F., Kermasha, S., Alli, I. and Mulligan, C.N. (1999) Encapsulation in the food industry: a review. Int. J. Food. Sci. Nutr. 50, 213Ð224. Gouin, S. (2004) Microencapsulation: industrial appraisal of existing technologies and trends. Trends Food Sci. Technol. 15, 330Ð347. Guggenheim, E.A. (1952) Mixtures, Clarendon, Oxford. Guo, J.-H. (1994) A theoretical and experimental study of additive effects of physical aging and antiplasticization on the water permeability of polymer film coatings. J. Pharm. Sci. 83(3), 447Ð 449. Harris, P. (ed.) (1990) Food Gels, Elsevier Applied Food Science Series, London. Hernandez, R.J. (1986) Evaluation of the aroma barrier properties of polymer films. J. Plastic Film Sheeting 2(3), 181Ð211. Hildebrand, J.H., Prausnitz, J.M. and Scott, R.L. (1970) Regular and Related Solutions, van Nostrand- Reinhold, Princeton, NJ. Hopfenberg, H.B. and Frisch, H.L. (1969) Transport of organic micromolecules in amorphous polymers. J. Polym. Sci. Polym. Lett. B7, 405Ð409. Hsu, C.C. and Prausnitz, J.M. (1973) Thermodynamics of polymer compatibility in ternary systems. Macro- molecules 7(3), 320Ð324. Hu, Y. , Zhou, H., Liu, H., Wu, D.T. and Prausnitz, J.M. (1997) Estimation of vapor-liquid equilibria for polymer solutions by a group-contribution method. Fluid Phase Equilibr. 134, 43Ð54. Jackson, L.S. and Lee, K. (1991) Microencapsulation and the food industry. Lebensm. Wiss. Technol. 24, 289Ð297. Kester, J.J. and Fennema, O.R. (1986) Edible coatings: a review. Food Technol. 40(12), 47Ð59. Khare, A.R. and Peppas, N.A. (1995) Swelling/deswelling of anionic copolymer gels. Biomaterials 16(7), 559Ð567. King, C.J. (1995) Spray drying: retention of volatile compounds revisited. Drying Technol. 13(5Ð7), 1221Ð1240. Kondo, A. (1979) Microcapsule Processing and Technology (ed. J.W. Van Valkenburg), Dekker, New York. van Krevelen, D.W. (1976) Properties of Polymers: Estimation and Correlation with Chemical Structure, 2nd edn, Elsevier, Amsterdam. Liu, X.-D., Furuta, T., Yoshii, H. and Linko, P. (2000) Retention of emulsified flavour in a single droplet during drying. Food Sci. Technol. Res. 6(4), 335Ð339. McLaren, A.D. and Rowen, J.W. (1951). Sorption of water vapor by proteins and polymers: a review. J. Polym. Sci. 7(2Ð3), 289Ð324. Meares, P. (1954) The diffusion of gases through polyvinylacetate. J. Am. Chem. Soc. 76, 3415Ð3422. Michaels, A.S., Wong, P.S.L., Prather, R. and Gale, R.M. (1975) A thermodynamic method of predicting the transport of steroids in polymeric matrices. AIChE J. 21, 1073Ð1080. Miller, K.S. and Krochta, J.M. (1997) Oxygen and aroma barrier properties of edible films: a review. Trends Food Sci. Technol. 8, 228Ð237. Moldenhauer, M.G. and Nairn, J.G. (1992) The control of ethylcellulose microencapsulation using solubility parameters. J. Control. Release 22, 205Ð218. Naylor, T.D.V. (1989) Permeation properties. In: Comprehensive Polymer Science,Vol.2 (eds C. Booth and C. Price), Pergamon Press, Oxford, pp. 643Ð668. Paul, D.R. (1985) Transport property of polymers. In: Applied Polymer Science,Vol.285, American Chemical Society Symposium Series, Washington, D.C., pp. 253Ð274. Peppas, N. and Am Ende, D.J. (1997) Controlled release of perfumes from polymers. II. Incorporation and release of essential oils from glassy polymers. J. Appl. Polym. Sci. 66, 509Ð513. Prausnitz, J.M. (1995) Some new frontiers in chemical engineering thermodynamics. Fluid Phase Equilibr. 104, 1Ð20. Prausnitz, J.M. (1999) Thermodynamics and the other chemical engineering sciences: old models for new chemical products and processes. Fluid Phase Equilibr. 158–160, 95Ð111. Qui, Z.H. and Xu, A. (1999) Starch-based ingredients for flavour encapsulation. Cereal Foods World 44(7), 460Ð465. Risch, S.J. and Reineccius, G.A. (eds) (1995) Encapsulation and Controlled Release of Food Ingredients, ACS Symposium Series 590, Washington, D.C. P1: SFK/UKS P2: SFK c06 BLBK221-Taylor/Linforth November 17, 2009 13:13 Printer Name: Yet to Come

Useful principles to predict the performance of polymeric ﬂavour delivery systems 189

Roberts, D.D. and Taylor, A.J. (eds) (2000) Flavour Release, ACS Symposium Series, 763, Washington, D.C. Rodriguez, C.F., Bruneau, N., Barra, J., Alfonso, D. and Doelker, E. (2000) Hydrophilic cellulose derivatives as drug delivery carriers. Handbook of Pharmaceutical Controlled Release Technology (ed. D.L. Wise), Dekker, New York, pp. 1Ð30. Rosenberg, M., Kopelman, I.J. and Talmon, J. (1990) Factors affecting retention in spray-drying microencapsulation of volatile materials. J. Agric. Food Sci. 38, 1288Ð1294. Rowe, R.C. (1988) Binder-substrate interactions in tablets: a theoretical approach based on solubility parameters. Acta Pharm. Technol. 34, 144Ð146. Sakellariou, P. and Rowe, R.C. (1991) Phase separation and morphology in ethylcellulose/cellulose acetate phthalate blends. J. Appl. Polym. Sci. 43, 845Ð855. Salame, M. and Steingiser, S. (1977) Barrier polymers. Polym. Plast. Technol. Eng. 8(2), 155Ð175. Schultz, T.H., Dimick, K.P. and Mackower, B. (1956) Incorporation of natural fruit ﬂavours into fruit juice powders: I. locking of citrus oil in sucrose and dextrose. Food Technol. 10, 57Ð60. Shahidi, F. and Han, X.-Q. (1993) Encapsulation of food ingredients. Crit. Rev. Food Sci. Nutr. 33(6), 501Ð547. Sips, R. (1948) On the structure of a catalyst surface. J. Chem. Phys. 16(5), 490Ð495. Sips, R. (1952) On the structure of a catalyst surface. J. Chem. Phys. 18(8) 1024Ð1026. Stannett, V., Haider, M., Koros, W.J. and Hopfenberg, H.B. (1980) Sorption and transport of water vapor in glassy poly(acrylonitrile). Polym. Eng. Sci. 20(4), 300Ð304. Tompa, H. (1956) Polymer Solutions, Butterworth Scientiﬁc Publications, London. Tse, G., Blankstein, D., Shefer, A. and Shefer, S. (1999). Thermodynamic prediction of active ingredient loading in polymeric microparticles. J. Control. Release 60, 77Ð100. Ubbink, J. and Schoonman, A. (2004) Flavour Delivery Systems, in Kirk-Othmer Encyclopedia of Chemical Technology, John Wiley & Sons, Hoboken, NJ. Vrentas, J.S. and Duda, J.L. (1975) Diffusion in polymer-solvent systems: III construction of Deborah number diagrams. J. Polym. Sci. 15, 441Ð453. Yapel, R.A., Duda, J.L., Lin, X. and von Meerwall, E.D. (1994) Mutual and self-diffusion of water in gelatin: experimental measurement and predictive test of free volume theory. Polymer 35(11), 2411Ð2416. Zimm, B.H. and Lundberg, J.L. (1955) Sorption of vapors by high polymers. J. Phys. Chem. 60, 425Ð428. P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

7 Delivery of ﬂavours from food matrices Saskia M. van Ruth and Jacques P. Roozen

7.1 INTRODUCTION

Human perception of food flavour and texture during consumption is a complicated process in which taste, mouth feel, vision, olfaction, the trigeminal system and auditory signals contribute to the total appreciation of a food product (Sheperd, 1995; Meiselman, 1996; Visschers et al., 2006). Non-oral characteristics, such as olfactory and visual cues, have been used to evaluate the edibility of foods since ancient times (de Wijk et al., 2004). Visual factors have been shown to affect the amount of food ingested (Prinz and Heath, 2000). Anticipatory food sensations and related chemosensation and neural encoding are important for understanding how individual differences may contribute to overeating and the current obesity epidemic (Beaver et al., 2006; Small et al., 2008). It is generally accepted that aroma, taste, texture and mouth feel account for the major stimuli that contribute to the perception of flavour. Stimulation occurs when compounds from the food come into contact with receptor cells in the mucous membranes of nose (odour/aroma) and mouth (taste) or when food structures such as emulsions or rigid cell walls affect the chewing process (texture) or interact with the mouth lining (mouth feel) (Taylor, 1996). Mouth feel responses are concerned with the heat sensation of spices and the cooling sensation of menthol. Taste is concerned with the sensations of sweet, sour, salty, bitter and umami, which are associated with receptors on the tongue. Aroma is a much broader sensation and encompasses an estimated 10 000 or more different odours (Reineccius, 1993). The flavour perceived during eating not simply is an addition of the four basic stimuli but is a complex pattern that has different characteristics for particular foods. When food is eaten, flavour molecules are released from the food into the mouth and the volatile flavour compounds pass back up through the nasopharynx into the nose. A sufficiently high concentration of flavour molecules has to be released from the food to stimulate the olfactory system and elicit a response. Flavour release and delivery depend on the nature and concentration of volatile compounds present in the food, as well on as their availability for perception as a result of interactions between the major components and the aroma compounds in the food (Bakker et al., 1995). Food composition factors and eating behaviour determine the extent of flavour release, delivery and perception (Bakker et al., 1996). With the growing range of new foods available, many with lower fat or lower sugar formulations than the traditional foods, it is becoming increasingly important to understand the factors that affect the perception of flavour, including how flavour is released from food matrices, in order to deliver an acceptable flavour from these foods. Knowledge of the binding P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 191

behaviour of flavour compounds in relation to various food components and their rates of partitioning between different phases is of great practical importance for the flavouring of foods, in determining the relative retention of flavours during processing or the selective release of specific compounds during processing, storage and mastication (Kinsella, 1988). Thermodynamic and kinetic factors control the release of flavour from food products and thus its delivery. The influence of properties of the flavour compounds as such and the thermodynamic aspects and kinetic aspects of flavour release as a function of food composition and oral manipulation (salivation, mastication) are discussed in the following sections. Finally, a few flavour delivery systems that are of practical importance from a food technology perspective are summarized in Section 7.5.

7.2 FLAVOUR PROPERTIES

Flavour delivery depends on the availability of the flavour compounds in the gas phase and, therefore, on the affinity of the flavour compounds for the food matrix. Various properties of the flavour compounds determine the interactions with food components, e.g. molecular size, functional groups, shape and volatility (Kinsella, 1988). Properties such as molecular weight, vapour pressure, boiling point, octanolÐwater partition coefficient (log P) have been used to predict the volatility of the compounds under static conditions (Roberts and Acree, 1996; Linforth et al., 2000; van Ruth et al., 2000).

7.3 THERMODYNAMIC ASPECTS OF FLAVOUR DELIVERY

Interactions between flavour substances and major food components are of two types: attractive and repulsive interactions. Attractive interactions involve fixation of flavour compounds on food components, whereas repulsive interactions concern the release of aroma compounds. The nature of these interactions depends on the physicochemical properties of the compounds and the food matrix (Le Thanh et al., 1992).

7.3.1 Definition of gas/product partition coefficients and activity coefficients The discussion of thermodynamic aspects of flavour release, such as phase partitioning, requires a definition of gas/product partition coefficients and activity coefficients. Flavour release will only take place if the gas/product phase equilibria are disturbed. In other words, non-equilibrium is the driving force for mass transfer. Equilibrium between the gas phase and the product phase exists only if there is no effective transfer at the productÐgas interface (de Roos, 2000). The equilibrium partition coefficient can be expressed as

Ci K i = g i (7.1) Cp P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

192 Food Flavour Technology

and

pi H i = i (7.2) Cg

i i i where K is the partition coefficient, H is the Henry’s law constant, Cp is the food product i i concentration (mole fraction), Cg is the gas-phase concentration (mole fraction) and p is the partial pressure in the gas phase of the flavour compound i, all at equilibrium. Partition coefficients can also be expressed using concentrations in the gas and product phases. Gas/product partition coefficients are temperature dependent: the log-transformed gas/liquid partition coefficient is linearly related to the temperature (Kolb et al., 1992). Henry’s law explains the behaviour of the flavour compound and holds for a restricted range of conditions (Taylor, 1998). The concentration of the flavour compounds should be such that they can be considered infinitely dilute. Furthermore, molecules must remain unimolecular. Compounds that dissociate, such as organic acids (de Roos and Sarelse, 1996), or that associate, e.g. those that form micelles (Piggott et al., 1996), do not show ideal behaviour. Another important property is the relative volatility of the volatile component with respect to water, since it determines the relative proportions in which the aroma compound and water come off during an equilibrium vaporisation. This relative volatility ␣iw is defined by the i ratio of the partition coefficient of flavour compound i (Kp ) to the partition coefficient of w w w w water (K ). K in turn is defined as Cg /Cp .

K i Ci Cw ␣iw = = g p w i w K Cp Cg

Activity coefﬁcients in the liquid phase, ␥ i, are deﬁned by reference to the concept of an ideal solution:

i = ␥ i i i0 P Cp P (7.3)

where Pi0 is the vapour pressure of the pure component i at the temperature in question. The activity coefficient is a measure of the degree of compatibility of i with the liquid phase, the tendency for intermolecular forces to develop between i and the major constituents of the food, in comparison to the strength of intermolecular forces among the major components of the food itself. A similar definition may be made for the activity coefficient of water. Activity coefficients may also be defined for the gas phase, but these tend to be important only at pressures higher than those normally encountered under food conditions. Putting the equations for compound i and water together and recognising that by Dalton’s law

Pi ␥ i = (7.4) P where P is the total pressure, gives

␥ i Pi0 K i = (7.5) P P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 193

and

␥ i Pi0 ␣iw = (7.6) ␥ w Pw0

For many important flavour compounds, the vapour pressure of the pure substance is not greatly different from that of water. On the other hand, values of ␥ i tend to be very large, with values of the order of 1000 being common. These very large activity coefficients stem from the fact that common volatile flavour compounds tend to be relatively non-polar, and therefore relatively incompatible with a highly polar, aqueous solution in terms of intermolecular forces. Conversely, values of ␥ w tend to be around 1, since water is a major component of many food systems (King, 1983).

7.3.2 Types of binding 7.3.2.1 Absorption and adsorption Binding of flavour compounds in food systems is synonymous with ‘sorption’ in its broad sense, including adsorption, absorption, physicochemical binding and chemical binding. Adsorption and absorption are types of binding specific for low-moisture food systems. Dry foodstuffs consist of particles of variable size with an outer surface and, usually, an inner surface made up of fine pores and channels. Volatile compounds can therefore be sorbed onto both the outer and the inner surfaces: this process is called adsorption. The aroma compound may also ‘dissolve’ in the material of the particle: this process is called absorption. In crystalline nutrients of low molecular mass, the process is mainly adsorption to the outer surface. Pores play no great role. Sorption of sugars and salts is normally physical, according to molar heat of sorption at low vapour pressures of volatiles. In special cases, even crystalline substances can bind very large amounts of aroma compounds, irreversibly under some conditions. An example is the binding of volatile acids and amines to amino acids (Maier, 1975). Physical and chemical binding of various food components, which are the same for low- and high-moisture food systems, is discussed in the following section.

7.3.2.2 Physicochemical and chemical binding Flavour binding/complex formation in food systems is the result of specific physicochemical and chemical interactions between major food components and the flavour compounds. It is important to discriminate here between dissolved, bound and total flavour concentration. Only the free dissolved flavour molecules exert a vapour pressure (de Roos, 2000). Fixation of aroma substances in food results from a number of binding processes:

(a) Chemicalr binding Covalent bonds, which are irreversible and involve the transfer of electrons between two atoms (b) Physicochemicalr binding r van der Waals forces r Hydrogen bonds r Hydrophobic interactions Ionic bonds (Solms et al., 1973; Kim and Min, 1988; Voilley et al., 1990) P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

194 Food Flavour Technology

The composition of the food matrix determines the extent and type of aroma binding. Apart from the aqueous phase present in high-moisture foods, the major food components with respect to binding are lipids, carbohydrates and proteins.

7.3.3 Lipid–flavour interactions Most lipids are hydrophobic, non-polar materials that exist naturally as liquids (oils) or solids (fats). Lipids may be regarded as material of biological origin consisting of one or more of the following classes: free fatty acids; mono-, di- and triglycerides; phospholipids; sterols; plasmalogens; and lipoproteins (Forss, 1969). Of all food components, lipids probably have the strongest impact on gas/product partitioning. In lipid-containing food systems, lipophilic flavour compounds are bound to the lipid molecules by weak, reversible van der Waals forces and unspecific hydrophobic interactions (Plug and Haring, 1993). Lipids act as solvent for lipid-soluble, hydrophobic flavour compounds. Table 7.1 illustrates the effect of the interactions of flavour compounds and the oil phase by comparing the gas/liquid partition coefficients of flavour compounds in sunflower oil, in water and in a mixture of oil and water. The generally hydrophobic nature of flavour compounds results in considerable differences in headspace composition if the lipid phase is removed, as is the case in fat-free foods. In the absence of fat, the food matrix retains lipophilic flavours poorly and the resulting headspace concentrations are high, as indicated in Table 7.1 (Plug and Haring, 1993). Binding to the water phase tends to reduce the volatility of polar compounds in much the same way that oils bind non-polar flavour compounds (Forss, 1969).

Table 7.1 Gas/liquid partition coefficients (K × 1000) of 18 flavour compounds in sunflower oil, water and a 3:2 mixture thereof.

Oil Oil–water mix Water Alcohols 1-Propanol 3.8 1.0 0.6 1-Butanol 1.3 0.8 0.8 3-Methyl-1-butanol 0.6 0.6 1.2 2-Pentanol 0.9 0.8 1.7 1-Hexanol 0.5 0.6 2.4 2-Nonanol 0.3 0.3 5.6 Ketones 2-Butanone 4.8 3.9 4.3 2,3-Butanedione 4.9 2.6 1.9 2-Heptanone 0.5 0.6 15.6 2-Octanone 0.3 0.4 21.8 2-Decanone 0.4 0.4 25.5 Aldehydes Hexanal 0.6 0.9 23.6 Heptanal 0.3 0.4 35.8 Octanal 0.2 0.3 44.1 Esters Ethyl acetate 5.3 5.8 13.0 Propyl acetate 2.0 2.7 21.8 Butyl acetate 0.5 0.7 22.8 Ethyl butyrate 1.1 1.3 25.0 P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 195

Buttery et al. (1971, 1973) found that gasÐoil partition coefficients of aliphatic aldehydes and ketones decreased with increasing chain length of the flavour molecule. Later on, other authors (Gijs et al., 2000; Haahr et al., 2000) have reported a similar relationship. The effect of chain length can be explained by the lipophilicity of the flavour compounds, which is an important factor with respect to the affinity of aldehydes, ketones, esters, thioesters, sulfides and disulfides for the lipid phase (Piraprez et al., 1998; Gijs et al., 2000). Fat concentration (Schirle-Keller et al., 1994) and composition (Druaux et al., 1998), pH (van Ruth et al., 1999) and temperature (Hall and Andersson, 1983) determine the extent of interactions between lipids and small molecules. The occurrence of compoundÐliquid interactions can also be expressed as activity coefficients ␥ i. Interactions cause the value of ␥ i to differ from 1. Compounds such as 2,5- dimethylpyrazine in oil have ␥ i values that are lower than 1. This is the result of the size of the flavour molecule, which differs significantly from that of the solvent molecules and results in repulsive forces. For other compounds in oil, ␥ i values larger than 1 indicate attractive forces between flavour compounds and oil (Druaux et al., 1998).

7.3.4 Carbohydrate–flavour interactions The retention of flavour compounds in systems rich in carbohydrates is more complex than the retention caused by lipids. Simple sugars (e.g. glucose and maltose) produce an increase in vapour pressure for a number of components at low concentrations and a marked decrease for others (Buttery et al., 1971; Nawar, 1971). However, higher concentrations of simple sugars generally result in increased gas/liquid partition coefficients (Nahon et al., 2000; Hansson et al., 2001). A sort of ‘salting-out’ effect is likely to be the reason for this phenomenon, whereby the sugar interacts with water, increasing the concentration of flavour compounds in the remaining volume of free water (Voilley et al., 1977). This hypothesis was confirmed by Kieckbusch and King (1979), who calculated partition coefficients of some acetates in sucrose solutions on a free-water basis. Although initially a sharp increase in partition coefficients was found with increasing sucrose concentrations, the corrected partition coefficients remained nearly constant. Whereas some studies have used just a few, similar flavour molecules to investigate the effect of sugars on partition, Friel et al. (2000) used 40 different compounds and sugar concentrations up to 60%. The partition behaviour was expressed as a quantitative structureÐproperty relationship model and the factors describing behaviour were log P and some topological (molecular shape) factors. Generally speaking, significant changes in headspace were seen only for sugar concentrations above 20%. Binding of flavour compounds to simple sugars is not likely as it can occur only through loose hydrogen bonds, in which case the flavour compounds have to compete with the water molecules. Polysaccharides, such as dextrins and gums, are known to interact with flavour compounds, and are used to stabilise flavours in food preparations (Versic, 1988). Dextrins can reduce the activity coefficients of flavour compounds in water and, accordingly, gas/liquid partition coefficients (Lebert and Richon, 1984). The binding is of hydrogen bond type (Solms et al., 1973), which results in competition of flavour compounds for the binding sites (Goubet et al., 2000). Among enzyme-modified starch derivatives, cyclodextrins are known to entrap flavour compounds of specific geometry and polarity (Szente and Szejtli, 1988) and are used for flavour encapsulation (Hedges et al., 1995). Gums, such as xanthan and guar gum, are generally used as thickeners and also exhibit interactions with flavour compounds. The type P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

196 Food Flavour Technology

of compound affects the extent of binding. As competition between flavour compounds with respect to binding to these gums has been observed, the binding mechanism is likely to be of a more general hydrogen bond nature (Roberts et al., 1996; Yven et al., 1998). The interactions of flavour substances with starch are of special importance, since starch is one of the most commonly found components in food systems. Interactions involve the formation of so-called starch inclusion complexes. Inclusion compounds are not the result of chemical reactions but have been defined as addition compounds in which one entity fits into and is surrounded by the lattice of the other. Starch combines with a variety of substances to form inclusion compounds that are insoluble at room temperature. The inclusion compounds can be formed by the addition of particular substances to a molecularly dispersed solution of starch. Usually, the complex is formed from a hot solution by slow cooling in presence of an excess of guest molecules. The formation of a helical arrangement of amylose molecules has been recognised to be responsible for inclusion complex formation. It is also conceivable that single helices are induced; in this case the flavour compound is located in the free space between the helices (Osman-Ismail and Solms, 1973; Escher et al., 2000). Alcohols, aldehydes, ketones, terpenes and fatty acids have been reported to form inclusion complexes with starch (Osman-Ismail and Solms, 1973; Solms et al., 1973; Nussli,¬ 1998). Starch affects flavour retention not only at molecular level by the complexation of flavour compounds with amylose and amylopectin; it also has an effect at supramolecular level through crystallisation of inclusion complexes, and at colloidal level through formation of dispersions in which aggregation, phase separation and network formation of starch and amylose complexes occur (Escher et al., 2000).

7.3.5 Protein–flavour interactions Two types of interactions can occur between flavour compounds and proteins: (a) reversible physical adsorption via non-covalent interaction and (b) chemical reaction via covalent linkages. In the first case, the heat released by the reaction is less than 20 kJ/mol. In the second case, the heat released is at least 40 kJ/mol and includes formation of salts, amides and esters formation, and aldehyde condensation with NH2 and SH groups. Flavour compounds, especially aldehydes, can react either with free amino acids or with free amino groups of proteins and reversibly form Schiff bases. However, the process by which flavour compounds bind to proteins through covalent linkages is irreversible, as is seen in the case of interaction between formaldehyde and proteins. Formaldehyde reacts not only with primary amino groups in proteins but also with sulfhydryl groups. Flavour compounds bind to protein only when binding sites are available; that is, if the sites are not engaged in proteinÐprotein or other interactions. Loops projecting into the aqueous phase are especially favourable sites for interaction with flavour compounds (Kim and Min, 1988). The reversible and non-covalent binding of a flavour compound obeys the Scatchard equation:

V bound = K (n − V ) at equilibrium (7.7) V bound

where Vbound is the number of moles of ﬂavour compound bound per mole of protein, V is the molar concentration of the free volatile compound, K is the association constant and n is the total number of binding sites per mole of protein. As binding proceeds, the protein can undergo conformational changes and more binding sites can become available. Non-polar P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 197

flavour compounds can diffuse into the hydrophobic core of the protein, replace intra- or intermolecular proteinÐprotein hydrophobic interactions and result in a change in the protein solubility. The amount of flavour adsorbed to proteins increases with the hydrophobicity of the proteins. The amount of irreversibly bound acetone and ethanol, however, increases with the polarity of the proteins (Kim and Min, 1988). As was found for non-specific binding to carbohydrates, some flavour compounds exhibit competition for binding into the hydrophobic pocket of proteins (Muresan and Leguijt, 1998; Jouenne and Crouzet, 2000). Infrared spectroscopy and fluorescence quenching have indicated that flavour compounds seem to have different binding sites (Dufour and Haertle,« 1990; Guichard and Langourieux, 2000). Lactoglobulin, albumin and soy proteins are the most intensively studied proteins with respect to flavourÐprotein interactions.

7.4 KINETIC ASPECTS OF FLAVOUR DELIVERY

Flavour release is determined by thermodynamic and kinetic factors; this is illustrated by the differences in ﬂavour release under static and dynamic conditions in Fig. 7.1. The factors inﬂuencing the equilibrium concentrations have been discussed in preceding paragraphs. Kinetic factors determine the rate at which equilibrium is achieved.

Fig. 7.1 Flavour release under (a) equilibrium and (b) dynamic conditions from water and oil. , water; , oil. P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

198 Food Flavour Technology

Once a food is ingested, it is rapidly coated by a thin film of saliva. It can therefore be assumed that flavour, on release from a solid food, must first pass through the saliva phase before partitioning into the headspace of the oral cavity. In the case of liquid and semi-solid foods, flavour is already in the liquid phase and therefore can be released into the headspace directly. Thus, passage of flavours from the food to the headspace is a three-phase arrangement involving the food, saliva and gas phases. It is unlikely that simple diffusion of flavour molecules in the bulk phases of the food or saliva can determine the rates of release in the mouth, as mastication disturbs diffusion gradients and generates fresh interfaces (Harrison, 2000).

7.4.1 Principles of interfacial mass transfer Under the non-equilibrium conditions that exist during eating, the driving force for transfer of flavour compounds across the interface is the difference in flavour concentration between product and gas phases (see also Chapter 8). The rate of the unidirectional diffusion from the product to the gas phases is determined by the concentration gradients as well as the mass transfer coefficients of the flavour compounds in each of the phases (Fick’s law). The rate of mass transfer in product (denoted by subscript p) and gas phase (subscript g) can be described as

dM p = k [Cint − C ] (7.8) dt p p p dM g = k [C − Cint] (7.9) dt g g g

where M is the total mass of flavour compound diffusing in the two phases and kp and kg are int the mass transfer coefficients. Cg, Cp and C are the concentrations of the flavour compound in the product phase, the gas phase and at the interface, respectively. Figure 7.2 illustrates the

Fig. 7.2 Flavour concentrations in gas and product phases under non-equilibrium conditions (redrawn int from de Roos and Graf, 1995), where Cg and Cp are the concentrations in the gas and product phases, C is the concentration at the interface and V g and V p are the gas- and product-phase volumes, respectively. P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 199

concentrations at the interface and in the gas and product phases. Flavour compound diffusion is based on two mechanisms: molecular and eddy diffusion. Molecular or static diffusion is the random movement of the molecules in the stagnant fluid. Typical molecular diffusivities are 10−5 and 10−9 m2/second in gas and liquid aqueous phases, respectively. The rate of molecular diffusion varies only slightly between flavour compounds. The second mechanism is eddy or convective diffusion, which transports element or eddies of the fluid from one location to another, carrying with them the dissolved flavour compounds. The rate of eddy diffusion is usually much higher than the rate of molecular diffusion and is independent of flavour type (de Roos, 2000). In general, it is assumed that diffusion of flavour compounds in the gas phase is extremely rapid and as a result the concentration gradient in the gas phase is neglected (Harrison et al., 1997; Harrison and Hills, 1997a). Consequently, the concentration of the flavour compound at the product side of the interface determines the i i concentration in the gas phase (Cp = Cg/K ), which allows reformulation of the previous equations as dM C p = k g − C (7.10) dt p K i p

The concentration gradient depends on the depletion of the flavour compound at the interface. Depletion is favoured by a high gas pressure and a low mass transfer coefficient of the flavour i compound in the product phase. If compounds are completely depleted at the interface (Cp ∼ 0), the release of these kinetically controlled conditions is similar for all flavour compounds as kp varies only slightly (molecular diffusion) or not at all with the type of flavour compound (eddy diffusion) (de Roos, 2000). Three mathematical models (see also Chapter 8), which differ in the mechanisms of mass transport, have been derived for predicting flavour release under dynamic conditions:

(a) Stagnant-film theory: The stagnant-film model assumes that the boundary layers at the interface are stagnant and that mass is transported through these layers as a result of molecular diffusion. The mass transport coefficient k varies with the first power of the diffusion coefficient D and the reciprocal of the effective thickness of the stagnant layer (Hills and Harrison, 1995). (b) Penetration theory: The penetration theory takes into account that the boundary layers are often not completely stagnant and that there is also mass transport by eddy diffusion. It is assumed that a volume element of liquid from the bulk comes into contact with the interface layers, and is exposed to the second phase for a definite interval. During this time, equilibrium is attained by surface layers through a process of unsteady-state molecular diffusion of flavour into the gas phase, before the volume element is remixed with the bulk liquid. In the penetration model, k varies with the square root of the diffusion coefficient (Harrison et al., 1997). (c) Non-equilibrium partition model: The non-equilibrium partition model assumes that mass transfer takes place only by eddy diffusion. The independence of the diffusion constant allows a multiple extraction model. It is assumed that flavour compounds are extracted from the product with infinitesimal volumes of gas. During successive extraction, full equilibrium is achieved only at the gasÐproduct interface in the infinitesimal volumes of product and gas phases. After each extraction, the initial flavour concentrations at the surface of the product are restored by diffusion and turbulence before the next extraction takes place (McNulty and Karel, 1973; de Roos and Graf, 1995). P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

200 Food Flavour Technology

7.4.2 Liquid food products Release rates of flavour compounds from liquid foods, both directly after consumption and after swallowing will depend on how the flavour compounds interact with other components of the food during dilution. Firstly, the concentrations of free flavour compounds are affected by the thermodynamic interactions described in Section 7.3. The majority of flavour compounds are hydrophobic and therefore preferentially partition into the lipid phase and not into the aqueous or gas phases. Dilution with saliva of a lipid- containing liquid food will therefore shift the aroma partitioning and change the release kinetics (Harrison, 1998; van Ruth et al., 2001). Diffusion of flavour compounds between lipid and aqueous phases is extremely rapid in liquid foods (Harrison et al., 1997). The mass transfer coefficient is influenced by the viscosity of the phases, and thus by the lipid fraction and droplet size. Experimental studies confirm the importance of these factors for flavour release (Charles et al., 2000). Flavour release from lipid-containing liquid foods, such as emulsions, has been described by the penetration theory (Harrison et al., 1997; Harrison and Hills, 1997a). Furthermore, macromolecules, such as proteins and polysaccharides, can bind both reversibly and irreversibly to flavour compounds, as described in Section 7.3, thus reducing the free flavour available for release. Diluting the macromolecular concentration with saliva will also shift the binding equilibrium and change the flavour release kinetics. In addition, the presence of these food components will affect the overall viscosity of the saliva, further influencing the rates of release. Dissociation between the bound and unbound states of a flavourÐmacromolecule complex is extremely fast and therefore will not affect the rate of flavour release. At first sight, this may seem surprising, but it is a consequence of the extremely small amount of flavour released in the headspace compared with the amount of free flavour retained in the aqueous phase. Reversible binding of flavour compound and macromolecule can be described with first-order chemical kinetics. The penetration theory allows modelling of flavour release across the gasÐliquid interface of solutions containing macromolecules (Harrison and Hills, 1997b). Interactions between methyl ketones and ␤-lactoglobulin (Andriot et al., 2000) and flavour compounds and liquid gelatine resulted in effects on flavour release that are in agreement with the penetration theory (Bakker et al., 1998). Apart from the direct effect of viscosity on flavour release kinetics, the viscosity of saliva will also influence the residual thickness coating the inside of the oral cavity long after the majority of the food has been swallowed, thus potentially influencing the aftertaste. Breath-by-breath mass spectrometry has shown that some flavour compounds persist in the nose-space long after the food has been swallowed (Taylor, 1996). Viscosity is an important parameter of flavour release in liquid food systems. It determines the diffusion coefficient and, as a consequence, the mass transfer coefficient. Food components determining the viscosity of the liquid foodÐsaliva mixture have an effect on flavour release kinetics. A practical example is liquids with high-sucrose concentrations, flavour release from which shows penetration model behaviour (Nahon et al., 2000).

7.4.3 Semi-solid food products For semi-solid foods such as gels, which possess melting points below the mouth temperature, the driving force for ﬂavour release is the rate at which heat can diffuse into the gel matrix and initiate melting. For harder gels with melting points above mouth temperature, the diffusion of sucrose from the surface of the gel into the adjacent saliva phase is the rate-limiting P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 201

step for ﬂavour release, because it lowers the melting temperature of the surface. In vitro release from gelatine gels containing sucrose shows good agreement with the stagnant layer model. The model can also be applied to other foods when ﬂavour release is accompanied by a melting transition. Ice creams and chocolates are expected to behave like soft gels with low-sucrose concentrations, the rate-limiting step of which is thermal diffusion (Harrison and Hills, 1996).

7.4.4 Solid food products For simple solid foods, where dissolution of the sugar matrix determines flavour release, e.g. boiled sweets, the driving force for release across the interface is the sucrose gradient between the food product and saliva. Flavour release from this matrix shows stagnant layer behaviour. As the matrix dissolves, all flavours are simultaneously released into the surrounding saliva, from where they partition into the headspace of the oral cavity. The mass transfer coefficient of this type of food is determined by oral manipulation. Higher rates of manipulation will reduce the thickness of the stagnant layer, will increase mass transfer rates and, therefore, will increase flavour release (Hills and Harrison, 1995). Flavour release and delivery from non-disintegrating solid foods are more complex. Trans- fer of flavour compounds to the gaseous phase is a three-phase problem for foods of this type, with saliva as the intermediate medium. In principle, the relationship for mass transfer from saliva to the gas phase is similar to that for liquid foods (Harrison et al., 1998; de Roos, 2000), described earlier in this section. The rate-limiting step for release from solid foods is the transport of flavour compounds across the foodÐsaliva interface (Harrison et al., 1998; de Roos, 2000; Lian, 2000). Both the particle and the surrounding fluid affect the transfer of the compounds (Lian, 2000). Flavour compounds are transferred from the food into the saliva at different rates determined by the productÐsaliva partition coefficients. The transfer can be described by the stagnant-layer theory. The physical processes occurring during eating, such as saliva flow, mastication and swallowing, are important factors determining the particle size distribution and thus food product interface. Initial rates of flavour release from solid foods are less sensitive to chewing frequency and saliva flow rate, but are extremely dependent on the fracture mechanics of the food and the mass transfer coefficient. This implies that the structure and composition of a food determine the initial rate of flavour release. With longer eating times, eating behaviour begins to play a role in flavour release rates. Increasing chewing frequency and/or the efficiency of particle selection increases the rate at which new surfaces are created and therefore the rate of flavour release into the saliva phase. Once the food has been swallowed, the flavour-enriched saliva will continue to release flavour compounds into the oral cavity (Harrison et al., 1998). The basic characteristics of hard candy allowed them to be used as model food system to investigate flavour release theories. The delivery of flavour compounds from hard candy to saliva for subsequent release/perception is a function of the product dissolution. Therefore, the ratio of non-volatile to volatile compounds remains constant throughout the consumption period. Hills and Harrison (1995) showed on the basis of the direct measurement of dye from hard candy during dissolution (by spectrophotometry) that the two-layer stagnant film theory best describes the mechanism of flavour release from hard candy. With the same model system, Schober and Peterson (2004) studied flavour–flavour interactions for l-menthol and 1,8-cineole. The release rates of both l-menthol and 1,8-cineole in breath analysis were more rapid and at a higher concentration when the compounds P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

202 Food Flavour Technology

were added to hard candy separate from one another in comparison to their addition as a mixture. Flavour delivery has been described from a theoretical perspective in the previous sections for liquid food, semi-solid food, very simple solid food and non-disintegrating solid food. For many other foods, however, mechanisms such as dissolution, melting and hydration play a role. Breaking of the food structure might increase viscosity. Furthermore, released flavour compounds may bind to macromolecules and partition into lipids and therefore reduce the free flavour present in the saliva. In some foods, the microenvironment may determine flavour release and delivery rather than the food’s gross composition. Those factors further complicate the elucidation of flavour delivery from complex solid foods (Harrison et al., 1998) and are a challenge for future model development research.

7.5 DELIVERY SYSTEMS: FOOD TECHNOLOGY APPLICATIONS

Controlled release at the right place at the right time is a key functionality. A timely and targeted release improves the effectiveness of food flavours, broadens the application range and ensures optimal dosage, thereby improving cost-effectiveness for the food manufacturer. Volatile, sensitive and reactive food compounds can be turned into stable ingredients through encapsulation (Gouin, 2004). Polymer-based delivery systems have been developed extensively for the biomedical and pharmaceutical sectors to protect and transport bioactive compounds to target functions (de Wolf and Brett, 2000; Langer and Peppas, 2003). In spite of successful elaboration of many synthetic polymers as delivery systems, these cannot be used for food applications as many of them are not food grade (Chen et al., 2006). Never- theless, micro- or nano-encapsulation techniques have also been used for a variety of food applications: (a) taste and odour masking, (b) prolonging organoleptic effects of flavour or other sensory markers, (c) delivery of other types of functional ingredients, (d) protection of food ingredients that are chemically unstable under storage or cooking conditions and (e) release of flavouring ingredients under specific circumstances (Yeo et al., 2005). Various encapsulation technologies such as spray-drying/cooling/chilling, fluidised bed, coacervation, alginate beads, liposomes, spinning disk, centrifugal co-extrusion, extrusion and supercritical fluid-based techniques have been employed for encapsulation of food and flavour ingredients (Gouin, 2004; Yeo et al., 2005). The main lipid-based encapsulation systems that can be used in foods are liposomes, cochleates and archaeosomes (Mozafari et al., 2006). Liposomes are composed of one or more lipid and/or phospholipids bilayers and can contain other molecules. Unilamellar vesicles are liposomes that contain a single bilayer membrane only (Taylor et al., 2005). These liposomes can incorporate and release two materials with different solubilities simultaneously: these systems can accommodate water-soluble compounds together with lipid-soluble agents. Cochleates are stable lipid-based carriers, comprising mainly a negatively charged lipid and a divalent cation. They have a cigar-shaped multilayered structure, consisting of a continuous, solid, lipid bilayer sheet rolled up in a spiral fashion with little or no aqueous space. In this way hydrophobic, amphiphilic, negatively or positively charged molecules can be delivered (Mozafari et al., 2006). Archaeosomes are liposomes made from one or more of the polar ether lipids extracted from the archaeobacteria (Mozafari et al., 2006). Compared with liposomes, archaeosomes are relatively more thermostable and more resistant to oxidation and chemical and enzymatic hydrolysis (Patel et al., 2000). P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 203

Proteins possess unique functional properties including their ability to form gels and emulsions, which allow them to be an interesting material for the encapsulation of bioactive compounds. Protein-based delivery systems are particularly interesting because various modifications allow them to form complexes with polysaccharides, lipids or other biopolymers and a wide variety of compounds can be incorporated (Chen et al., 2006). The delivery systems can also conjugate with nutrients via either primary amino groups or sulfhydryl groups (Weijermanna et al., 2005). Various classes of ligand-binding proteins have been reviewed by de Wolf and Brett (2000). In the carbohydrates group, chitosan is one of the most valued polysaccharides for drug delivery in biomedical sciences (Des Rieux et al., 2006). Cyclodextrins are popular for flavour applications. The main advantage of cyclodextrin is the unique release characteristics and the thermal and chemical stability imparted to the flavour compounds while entrapped within the cyclodextrin. Other delivery candidates in the carbohydrates group are starches, starch hydrolysates, maltodextrins and gums (Bangs and Reineccius, 1988).

7.6 CONCLUSIONS

Flavour release and delivery are complex owing to the rapidly changing conditions during eating. Many factors affecting flavour delivery have been determined. Still, only sporadic attempts have been made to study several factors simultaneously in order to determine the contribution of the individual factors to flavour release and delivery, and validate theoretical models, because of the large-scale experiments required. In the present nano era, delivery systems have attracted renewed scientific interest. For the food industry the understanding of the factors that determine a timely and targeted flavour delivery remains of key importance to successful product development.

REFERENCES

Andriot, I., Harrison, M., Fournier, N. and Guichard, E. (2000) Interactions between methyl ketones and ␤-lactoglobulin: sensory analysis, headspace analysis, and mathematical modelling. J. Agric. Food Chem. 48, 4246Ð4251. Bakker, J., Boudaud, N. and Harrison, M. (1998) Dynamic release of diacetyl from liquid gelatin in the headspace. J. Agric. Food Chem. 46, 2714Ð2720. Bakker, J., Brown, W., Hills, B., Boudard, N., Wilson, C. and Harrison, M. (1996) Effect of the food matrix on flavour release and perception. In: Flavour Science, Recent Developments (eds A.J. Taylor and D.S. Mottram), Royal Society of Chemistry, Cambridge, pp. 369Ð374. Bakker, J., Langley, K.R., Elmore, J.S., Brown, W., Martin, A. and Salvador, D. (1995) Release of flavour from food matrices and effect on flavour perception. In: Aroma: Perception, Formation, Evaluation (eds M. Rothe and H.-P. Kruse), Deutsche Institut fur¬ Ernahrungsforschung,¬ Potsdam-Rehbrucke,¬ Germany, pp. 191Ð207. Bangs, W.E. and Reineccius, G.A. (1988) Corn starch derivatives. In: Flavor Encapsulation, ACS Symposium Series 370 (eds S.J. Risch and G.A. Reineccius), American Chemical Society, Washington, DC, pp. 12Ð28. Beaver, J.D., Lawrence, A.D., van Ditzhuijzen, J., Davis, M.H., Woods, A. and Calder, A.J. (2006) Individual differences in reward drive predict neural responses to images of food. J. Neurosci. 26, 5160Ð5166. Buttery, R.G., Bomben, J.L., Guadagni, D.G. and Ling, L.C. (1971) Some considerations of the volatilities of organic flavor compounds in foods. J. Agric. Food Chem. 19, 1045Ð1048. Buttery, R.G., Guadagni, D.G. and Ling, L.C. (1973) Flavor compounds: volatilities in vegetable oil and oilÐwater mixtures. Estimation of odor thresholds. J. Agric. Food Chem. 21, 198Ð201. Charles, M., Rosselin, V., Beck, L., Sauvageot, F. and Guichard, E. (2000) Flavor release from salad dressings: sensory and physicochemical approaches in relation with the structure. J. Agric. Food Chem. 48, 1810Ð1816. P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

204 Food Flavour Technology

Chen, L., Remondetto, G.E. and Subirade, M. (2006) Food protein-based materials as nutraceutical delivery systems. Trends Food Sci. Technol. 17, 272Ð283. de Roos, K.B. (2000) Physiochemical models of flavor release from foods. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 126Ð141. de Roos, K.B. and Graf, E. (1995) Non-equilibrium partition model for predicting flavor retention in microwave and convection heated foods. J. Agric. Food Chem. 43, 2204Ð2211. de Roos, K.B. and Sarelse, J.A. (1996) Volatile acids and nitrogen compounds in prawn powder. In: Flavour Science, Recent Developments (eds A.J. Taylor and D.S. Mottram), Royal Society of Chemistry, Cam- bridge, pp. 13Ð18. de Wijk, A., Polet, I.A., Engelen, L., van Doorn, R.M. and Prinz, J.F. (2004) Amount of ingested custard dessert as affected by its color, odor, and texture. Phys. Behav. 82, 397Ð403. de Wolf, F.A. and Brett, G.M. (2000) Ligand-binding proteins: their potential for application in systems for controlled delivery and uptake of ligands. Pharmacol. Rev. 52, 207Ð236. Des Rieux, A., Fievez, V., Garinot, M., Schneider, Y.-J. and Preat, V. (2006) Nanoparticles as potential oral delivery systems of proteins and vaccines: a mechanistic approach. J. Control. Release 116, 1Ð27. Druaux, C., Le Thanh, M., Seuvre, A.-M. and Voilley, A. (1998) Application of headspace analysis to the study of aroma compoundsÐlipids interactions. J. Am. Oil Chem. Soc. 75, 127Ð130. Dufour, E. and Haertle,« T. (1990) Binding affinities of ␤-ionone and related flavor compounds to ␤-lactoglobulin: effects of chemical modifications. J. Agric. Food Chem. 38, 1691Ð1695. Escher, F.E., Nussli,¬ J. and Conde-Petit,« B. (2000) Interactions of flavor compounds with starch in food processing. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 230Ð245. Forss, D.A. (1969) Role of lipids in flavors. J. Agric. Food Chem. 17, 681Ð685. Friel, E.N., Linforth, R.S.T. and Taylor, A.J. (2000) An empirical model to predict the headspace concentration of volatile compounds above solutions containing sucrose. Food Chem. 71, 309Ð317. Gijs, L., Piraprez, G., Perpete,` P., Spinnler, E. and Collin, S. (2000) Retention of sulfur flavours by food matrix and determination of sensorial data independent of the medium composition. Food Chem. 69, 319Ð330. Goubet, I., Le Quere,« Semon,« E. and Seuvre, A. (2000) Competition between aroma compounds for the binding on ␤-cyclodextrins: study on the nature of interactions. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 246Ð259. Gouin, S. (2004) Micro-encapsulation: industrial appraisal of existing technologies and trends. Trends Food Sci. Technol. 15, 330Ð347. Guichard, E. and Langourieux, S. (2000) Interactions between ␤-lactoglobulin and flavour compounds. Food Chem. 71, 301Ð308. Haahr, A.-M., Bredie, W.L.P., Stahnke, L.H., Jensen, B. and Refsgaard, H.H.F. (2000) Flavour release of aldehydes and diacetyl in oil/water systems. Food Chem. 71, 355Ð362. Hall, G. and Andersson, J. (1983) Volatile fat oxidation products II. Influence of temperature on volatility of saturated, mono- and di-unsaturated aldehydes in liquid media. Lebensm. Wiss. Technol. 16, 326Ð 366. Hansson, A., Andersson, J. and Leufven,« A. (2001) The effect of sugars and pectin on flavour release from a soft drink-related model system. Food Chem. 72, 363Ð368. Harrison, M. (1998) Effect of breathing and saliva flow on flavor release from liquid foods. J. Agric. Food Chem. 46, 2727Ð2735. Harrison, M. (2000) Mathematical models of release and transport of flavours from foods in the mouth to the olfactory epithelium. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 179Ð191. Harrison, M., Campbell, S. and Hills, B.P. (1998) Computer simulation of flavor release from solid foods in the mouth. J. Agric. Food Chem. 46, 2736Ð2743. Harrison, M. and Hills, B.P. (1996) A mathematical model to describe flavour release from gelatine gels. Int. J. Food Sci. Technol. 31, 167Ð176. Harrison, M. and Hills, B.P. (1997a) Effects of air flow-rate on flavour release from liquid emulsions in the mouth. Int. J. Food Sci. Technol. 32,1Ð9. Harrison, M. and Hills, B.P. (1997b) Mathematical model of flavor release from liquids containing aroma- binding macromolecules. J. Agric. Food Chem. 45, 1883Ð1890. Harrison, M., Hills, B.P., Bakker, J. and Clothier, T. (1997) Mathematical models of flavor release from liquid emulsions. J. Food Sci. 62, 653Ð658, 664. P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

Delivery of ﬂavours from food matrices 205

Hedges, A.R., Shieh, W.J. and Sikorski, C.T. (1995) Use of cyclodextrins for encapsulation in the use and treatment of food-products. In: Encapsulation and Controlled Release of Food Ingredients,Vol.590 (eds S.J. Risch and G.A. Reineccius), Washington, DC, American Chemical Society, pp. 60Ð71. Hills, B.P. and Harrison, M. (1995) Two-film theory of flavour release from solids. Int. J. Food Sci. Technol. 30, 425Ð436. Jouenne, E. and Crouzet, J. (2000) Determination of apparent binding constants for aroma compounds with ␤-lactoglobulin by dynamic coupled column liquid chromatography. J. Agric. FoodChem. 48, 5396Ð5400. Kieckbusch, T.G. and King, C.J. (1979) Partition coefficients for acetates in food systems. J. Agric. Food Chem. 27, 504Ð507. Kim, H. and Min, D.B. (1988) Interaction of flavor compounds with protein. In: Flavor Chemistry of Lipid Foods (eds D.B. Min and T.H. Smouse), American Chemical Society, Washington, DC, pp. 404Ð420. King, J. (1983) Physical and chemical properties governing volatilisation of flavor and aroma components. In: Physical Properties of Food (eds M. Peleg and E.B. Bagley), AVI, Westport, CT, pp. 399Ð421. Kinsella, J.E. (1988) Flavour perception and binding to food components. In: Flavor Chemistry of Lipid Foods (eds D.B. Min and T.H. Smouse), American Chemical Society, Washington, DC, pp. 376Ð403. Kolb, B., Welter, C. and Bichler, C. (1992) Determination of partition coefficients by automatic equilibrium headspace gas chromatography by vapor phase calibration. Chromatographia 34, 235Ð240. Langer, R. and Peppas, N.A. (2003) Advances in biomaterials, drug delivery, and bionanotechnology. Am. Inst. Chem. Eng. J. 49, 2990Ð3006. Le Thanh, M., Thibeaudeau, P., Thibaut, M.A. and Voilley, A. (1992) Interactions between volatile and non-volatile compounds in the presence of water. Food Chem. 43, 129Ð135. Lebert, A. and Richon, D. (1984) Infinite dilution activity coefficients of n-alcohols as a function of dextrin concentration in water-dextrin systems. J. Agric. Food Chem. 32, 1156Ð1161. Lian, G. (2000) Modeling flavor release from oil-containing gel particles. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 201Ð211. Linforth, R.S.T., Friel, E.N. and Taylor, A.J. (2000) Modelling aroma release from food using physicochemical parameters. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 166Ð178. Maier, H.G. (1975) Binding of volatile aroma substances to nutrients and foodstuffs. In: Proceedings of the International Symposium on Aroma Research (eds H. Maarse and P.J. Groenen), Pudoc, Wageningen, pp. 143Ð157. McNulty, P.B. and Karel, M. (1973) Factors affecting flavour release and uptake in O/W emulsions. I. Release and uptake models. J. Food Technol. 8, 309Ð318. Meiselman, H.L. (1996) The contextual basis for food acceptance, food choice and food intake: the food, the situation and the individual. In: Food Choice, Acceptance and Consumption (eds H.L. Meiselmand and H.J.H. MacFie), Blackie Academic & Professional, London, pp. 239Ð263. Mozafari, M.R., Flanagan, J., Matia-Merino, L., Awati, A., Omri, A., Suntres, Z.E. and Singh, H. (2006) Review: recent trends in the lipid-based nanoencapsulation of antioxidants and their role in foods. J. Sci. Food Agric. 86, 2038Ð2045. Muresan, S. and Leguijt, T. (1998) Fluorimetric analysis of interactions between ␤-lactoglobulin and small ligands. In: COST Action 96. Proceedings of the Meeting in Goteborg¬ (ed. A. Zacharoff), European Commission, Luxembourg, pp. 45Ð52. Nahon, D.F., Harrison, M. and Roozen, J.P. (2000) Modelling flavor release from aqueous sucrose solutions, using mass transfer and partition coefficients. J. Agric. Food Chem. 48, 1278Ð1284. Nawar, W.W. (1971) Some variables affecting composition of headspace aroma. J. Agric. Food Chem. 19, 1057Ð1059. Nussli,¬ J. (1998) Complexation behavior of amylose with small ligands in aqueous starch systems. PhD thesis No. 12518, Swiss Federal Institute of Technology (ETH), Zurich,¬ Switzerland. Osman-Ismail, F. and Solms, J. (1973) The formation of inclusion compounds of starches with flavor substances. Lebensm. Wiss. Technol. 6, 147Ð150. Patel, G.B., Agnew, B.J., Deschatelets, L., Fleming, L.P. and Sprott, G.D. (2000) In vitro assessment of archaesome stability for developing oral delivery systems. Int. J. Pharm. 194, 39Ð49. Piggott, J.R., Gonzalez« Vinas,˜ M.A., Conner, J.S., Withers, S.J. and Paterson, A. (1996) Effect of chill filtration on whisky composition and headspace. In: Flavour Science, Recent Developments (eds A.J. Taylor and D.S. Mottram), Royal Society of Chemistry, Cambridge, pp. 319Ð324. Piraprez, G., Herent,« M.-F. and Collin, S. (1998) Determination of the lipophilicity of aroma compounds by RP-HPLC. Flavour Frag. J. 13, 400Ð408. P1: IFM/UKS P2: SFK c07 BLBK221-Taylor/Linforth November 21, 2009 7:53 Printer Name: Yet to Come

206 Food Flavour Technology

Plug, H. and Haring, P. (1993) The influence of flavourÐingredient interactions on flavour perception. Food Qual. Prefer. 5, 95Ð102. Prinz, J.F. and Heath, R.M. (2000) Bolus dimensions in normal chewing. J. Oral Rehabil. 2, 765Ð768. Reineccius, G. (1993) Biases in analytical flavor profiles introduced by isolation method. In: Flavor Mea- surement (eds C.T. Ho and C.H. Manley), Marcel Dekker, New York, pp. 61Ð76. Roberts, D.D. and Acree, T.E. (1996) Retronasal flavor release in oil and water model systems with an evaluation of volatility predictors. In: FlavorÐFood Interactions (eds R.J. McGorrin and J.V. Leland), American Chemical Society, Washington, DC, pp. 179Ð187. Roberts, D.D., Elmore, J.S., Langley, K.R. and Bakker, J. (1996) Effect of sucrose, guar gum, and car- boxymethylcellulose on the release of volatile flavor compounds under dynamic conditions. J. Agric. Food Chem. 44, 1321Ð1326. Schirle-Keller, J.-P., Reineccius, G.A. and Hatchwell, L.C. (1994) Flavor interaction with fat replacers: effect of oil level. J. Food Sci. 59, 813Ð815, 875. Schober, A.L. and Peterson, D.G. (2004) Flavor release and perception in hard candy: influence of flavor compound-compound interactions. J. Agric. Food Chem. 52, 2623Ð2627. Sheperd, R. (1995) Psychological aspects of food choice. Food Sci. Technol. Today 9, 178Ð182. Small, D.M., Veldhuizen, M.G., Felsted, J., Mak, Y.E. and McGlone, F. (2008) Separable substrates for anticipatory and consummatory food chemosensation. Neuron 57, 786Ð797. Solms, J., Osman-Ismail, F. and Beyeler, M. (1973) The interaction of volatiles with food components. Can. Inst. Food Sci. Technol. J. 6, A10ÐA16. Szente, L. and Szejtli, J. (1988) Stabilisation of flavours by cyclodextrins. In: Flavor Encapsulation (eds S.J. Risch and G.A. Reineccius), American Chemical Society, Washington, DC, pp. 148Ð157. Taylor, A.J. (1996) Volatile flavor release from foods during eating. Crit. Rev. Food Sci. Nutr. 36, 765Ð784. Taylor, A.J. (1998) Physical chemistry of flavour. Int. J. Food Sci. Technol. 33, 53Ð62. Taylor, T.M., Davidson, P.M., Bruce, B.D. and Weiss, J. (2005) Liposomal nanocapsules in food science and agriculture. Crit. Rev. Food Sci. 45, 587Ð605. Taylor, T.M., Gaysinsky, S., Davidson, P.M., Bruce, B.D. and Weiss, J. (2007) Characterization of antimicrobial-bearing liposomes by G-potential, vesicle size, and encapsulation efficiency. Food Bio- phys. 2,1Ð9. van Ruth, S.M., Grossmann, I., Geary, M. and Delahunty, C.M. (2001) Interactions between artificial saliva and twenty aroma compounds in water and oil model systems. J. Agric. Food Chem. 49(5), 2409Ð2413. van Ruth, S.M., O’Connor, C.H. and Delahunty, C.M. (2000) Relationships between temporal release of aroma compounds in a model mouth system and their physico-chemical characteristics. Food Chem. 71, 393Ð399. van Ruth, S.M., Roozen, J.P., Posthumus, M.A. and Jansen, F.J.H.M. (1999) Volatile composition of sunflower oil-in-water emulsions during initial lipid oxidation: influence of pH. J. Agric. Food Chem. 47, 4365Ð4369. Versic, R.J. (1988) Flavor encapsulation: an overview. In: Flavor Encapsulation (eds S.J. Risch and G.A. Reineccius), American Chemical Society, Washington, DC, pp. 1Ð11. Visschers, R.W., Jacobs, M.A., Frasnelli, J., Hummel, T., Burgering, M. and Boelrijk, A.E.M. (2006) Cross- modality of texture and aroma perception is independent of orthonasal or retronasal stimulation. J. Agric. Food Chem. 54, 5509Ð5515. Voilley, A., Lamer, C., Dubois, P. and Feuillat, M. (1990) Influence of macromolecules and treatments on the behavior of aroma compounds in a model wine. J. Agric. Food Chem. 38, 248Ð251. Voilley, A., Simatos, D. and Loncin, M. (1977) Gas phase concentrations of volatiles in equilibrium with a liquid aqueous phase. Lebensm. Wisse. Technol. 10, 45Ð49. Weijermanna, J., Lochmanna, D., Georgensa, C. and Zimmer, A. (2005) Albumin-protamine-oligonucleotide- nanoparticles as a new anti-sense delivery system. Part 2. Cellular uptake and effect. Eur. J. Pharm. Biopharm. 59, 431Ð438. Yeo, Y., Bellas, E., Firestone, W., Langer, R. and Kohane, D.S. (2005) Complex coacervates for thermally sensitive controlled release of flavour compounds. J. Agric. Food Chem. 53, 7518Ð7525. Yven, C., Guichard, E., Giboreau, A. and Roberts, D.D. (1998) Assessment of interactions between hydrocolloids and flavor compounds by sensory, headspace, and binding methodologies. J. Agric. Food Chem. 46, 1510Ð1514. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

8 Modelling ﬂavour release Robert S. T. Linforth

8.1 INTRODUCTION

Why model flavour release? The main driver behind model production is to describe the behaviour of a system using a mathematical equation, or a series of equations. These can then be used to predict how a given system should function, or how the performance of the system will be affected by any changes made to it. In terms of flavour release (or delivery), the ultimate objective is to understand and estimate the intensity and timing of aroma delivery to the consumer. There are two main approaches to the construction of models. The first is essentially a theoretical approach, using the principles of physics and chemistry to describe how a system is likely to behave. Equations are constructed that contain parameters, describing the attributes of the system (the flavour molecules, matrix and the phases that surround them) and the way these components interact to influence flavour release. Because this is a purely theoretical approach, these models are not initially dependent on experimental data being available; however, experimental data are necessary for model validation. This is one of the major advantages of this approach, since it allows the production of models for systems, even if analytical methods are too insensitive to study them. The second approach is data driven. Data (such as the headspace volatile concentrations) are collected for a range of flavour compounds, or a range of food matrices. A model is then constructed with sufficient components to describe the variation in the data. These models may contain some of the same parameters used in the theoretical approach described above, or they may contain parameters that just numericise the differences that occur in the ranges of the system studied (e.g. differences in molecular weight). The principal difference between the two approaches is the amount of data required. A theoretical model may only require five to ten experimental values for it to be tested and validated, whereas empirical models require much larger data sets to increase their accuracy. However, there are limits to the size of data sets that should be generated, because a model that includes every compound of interest would have nothing left for it to predict. Models can therefore exhibit many different forms and cover different aspects of a particular system. Key questions that can be asked of any model are the following:

r r What information is needed to make a prediction from the model? r Which are the most important factors in the model? How reliable are the predictions obtained? P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

208 Food Flavour Technology

The last point is perhaps the most important question that can be asked about any model, namely how well does it work? Model validation in its simplest form should be a relatively easy task (assuming data can actually be obtained). Use the model to make predictions about how a system should perform, obtain data on the behaviour of the system under the conditions used for the prediction, and then compare the data obtained experimentally with the predicted values. The quality of the correlation (determined statistically) between the two sets of values is a good indicator of the validity of the model. The system modelled can be considered as one of three basic types. The first type is the simplest models that describe the partitioning of volatile compounds from a matrix into the gas phase under equilibrium conditions: typically, these have been developed for aqueous solutions or solutions containing volatile and non-volatile solutes or, in some cases, a second liquid phase such as lipid. Similar studies with low-water-content matrices are less common due to ‘complicating’ factors such as the heterogeneity of the system and the difficulty of defining factors such as mass transfer and surface area in such systems. The second type of model is more complex and attempts to model dynamic partitioning where the gas phase above a matrix is disturbed or diluted to simulate the processes that happen in real life. These models now include a temporal (time) dimension. In addition, the exact conditions used (flow rate, volumes, etc.) can result in different degrees of volatile depletion in the headspace, and mixing within the matrix, complicating the comparison of different systems. In contrast, measurements of volatile partitioning in static systems at equilibrium give the same ‘release’, providing temperature and pressure are constant. The third type of model attempts to describe flavour release during eating, where a number of factors (dilution with saliva, temperature changes, bolus breakdown or dissolution, phase inversion, swallowing and breathing) can act in concert, making these models the most complex of all (the temporal dimension is almost unavoidable). They have also been difficult to validate, principally because of the lack of experimental data available for comparison with their predictions.

8.2 EQUILIBRIUM PARTITION MODELS

8.2.1 The air/water partition coefficient The static partition of a volatile compound between the food and gas phase (at equilibrium) has the potential to describe the maximum volatile concentration that may occur above a sample, under defined conditions of temperature and pressure. One of the simplest static partition situations that can be modelled is the equilibrium partitioning of a volatile compound, between water and air in a closed vessel (Equation 8.1). The concentration of a volatile in the air (Ca) above a solution is directly proportional to the concentration in the liquid phase, in this case water (Cw), when in an ideal solution. This can be expressed as a single value, the air/water partition coefficient (Kaw):

Ca Kaw = (8.1) Cw

This ratio does not, however, have the power of prediction; it simply describes the state of equilibrium. However, given the air/water partition coefﬁcient, it is possible to estimate the equilibrium concentration of the volatile in the headspace, for any concentration of the volatile in an ideal solution. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 209

Estimates of Kaw can be obtained from fundamental values such as the vapour pressure of a compound, the molar volumes of the gas and liquid phases and the compound’s activity coefficient (Voilley et al., 1977). These values are however not known for all compounds. Measuring them experimentally is not an ideal solution Ð it would be just as easy to measure Kaw itself. The activity coefficient for a compound can be estimated using methods such as analytical solution of groups (ASOG) or universal functional group activity coefficients (UNIFAC). These methods predict the physical and topological properties of compounds using algorithms that calculate the sum of each functional group’s contributions (Reid et al., 1987). Since the values for the group contributions have themselves been derived by reduction of experimental data, they too are semi-empirical values. The vapour pressure of a compound can be estimated from tables listing critical properties of compounds, or these too can be estimated using group contribution methods. These methods for estimating Kaw can yield estimates close to those obtained experimentally (Marin et al., 1999). It is important to note that there can be substantial differences in the values obtained by different methods of parameter estimation (Sorrentino et al., 1986), which would in turn influence the accuracy of the Kaw estimate.

8.2.2 Estimation of K aw using QSPR

In addition to the methods described above, Kaw estimates can be obtained using quantitative structureÐproperty relationships (QSPR) as outlined in Fig. 8.1. The QSPR method is effectively an extension and development of the group contribution approach. Group counting is based on the principle that certain aspects of the compound (in this case the number and

Measure property Select (statistically) Calculate of volatile parameters that parameters to compounds correlate with describe compounds properties

Model equation Build model by property = regression analysis a*Para 1 + b*Para 2 + c of parameters and Model failed properties

No Predict properties of Correlate with Is correlation good further compounds actual properties enough? using model equation of compounds Yes

Model validated

Fig. 8.1 QSPR modelling of the behaviour of volatiles. The parameters would be molecular descriptors for the compounds, and a and b would be regression coefﬁcients for individual parameters (Para 1 and Para 2, respectively), with intercept c. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

210 Food Flavour Technology

type of subunits present) influence its behaviour, and a value is ascribed to each one empirically. Such methods are simple and can be performed without too much difficulty. With computer technology, we have the potential to rapidly perform large numbers of complex calculations, so as to determine a wide range of parameters to describe aroma compounds (these would be virtually impossible manually). The parameters can be calculated for any compound; first, the compound is drawn using the software package, and then the computer calculates the desired parameters on the basis of this structure. Calculations can be carried out on the structure under a range of conditions, e.g. as a pure compound or, in aqueous solution, substantially increasing the number of individual parameters that can be estimated. Parameters can include simple values, such as group counts and the molecular weight, or more complex terms (describing the whole molecule), such as the solvent-accessible surface area and molecular energies. Having obtained a wide range of estimates for parameters that describe each compound, these can then be statistically compared against a data set, such as tables of known values for Kaw. Once the key descriptors of compounds have been found that describe their behaviour, these can be formulated into an equation. This equation can then be used to estimate the Kaw of further compounds based on the values for the key descriptors for that molecule. The quality of the model is determined statistically by the correlation coefficient (R2) and 2 2 the cross-validated correlation coefficient (Rcv ). R values approaching 1.0 indicate a high correlation between the values predicted by the model and those of the data set, and if the 2 2 value of Rcv is close to that of R , the model should have good predictive power. QSPR models have been developed to estimate the solubility of compounds in water and their vapour pressures as separate functions, which can be combined to estimate the air/water partition coefficient (Katritzky et al., 1998). QSPR models have also been produced to directly estimate the Kaw of volatile compounds. Katritzky et al. (1996) used a data set of 406 structurally diverse compounds to develop a model (Equation 8.2) with a final R2 of 2 0.9407 and an Rcv of 0.9386, indicative of good predictive power. Key terms in the model were HDCA(2), related to the hydrogen-bonding ability of the compound; O + 2*N, the number of oxygen and nitrogen atoms present; EHOMOÐELUMO, related to the dispersion energy of a polar solute in solution; PCWTE, which is the most negative partial charge weighted topographical electronic index; and finally Nrings, the number of rings present in the compound. All these terms can be calculated using chemical modelling software and require no experimental data before further predictions can be made. The model for Kaw shown in Equation (8.2) was generated, not with the food industry in mind, but other situations such as the fate of environmental pollutants. Consequently, the ‘structurally diverse compounds’ also included halogenated organic compounds and hydrocarbons, which are not of major importance to the food industry, except perhaps as contaminants. The opportunity certainly exists for the generation of QSPR models based on data sets of organoleptically significant compounds: ∗ Log Kaw = 1/(42 × HDCA(2) + 0.71 × (O + 2 N)

− 0.17 × (EHOMO − ELUMO) E + 0.13 × PCWT + 0.79 × Nrings + 2.82) (8.2)

Therefore, just as the group counting approach estimates the effect of a ÐCH2OH group on a property, QSPR determines a coefﬁcient expressing the inﬂuence of a parameter (which may be associated with part or all the molecules) on the behaviour of a compound. This has P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 211

the advantage that QSPR can readily take into account not only the presence but also the position of functional groups, and can incorporate the overall shape of the molecule. For both approaches, a data set is required to determine the inﬂuence of a group or a QSPR parameter. Further models in addition to those for Kaw can effectively be obtained by either approach to describe how volatile compounds behave in food.

8.2.3 Effect of lipid on volatile partitioning Food systems are rarely simple solutions of volatile compounds in water, and other solutes or phases may influence partitioning substantially. One of the most significant influences on volatile behaviour is the concentration of lipid in the system. The partitioning effect of emulsions can be described using the equation of Buttery et al. (1973), where the air/oil (Kao) and air/water (Kaw) partition coefficients are combined with their respective oil and water volume fractions (Fo and Fw) to produce an overall air/emulsion (Kae) partition coefficient:

1 Kae = (8.3) (Fo/Kao + Fw/Kaw)

This model functions adequately in some cases (Doyen et al., 2001), but not others (Voilley et al., 2000). This may be due to interactions with the emulsifier rather than just the lipid phase itself. Equation (8.3) was originally generated from studies of oil and water layers without any added emulsifiers. Consequently, it does not include any terms associated with the potential chemical interaction with emulsifiers, or their presence as a bulk component of the system. The behaviour of compounds that dissociate in aqueous environments, such as the organic acids, will be poorly described by Equation (8.3) because partition will depend on the degree of dissociation of the acid. Further, models have been developed to describe the partitioning of these compounds in lipid-containing systems, including factors such as the pH of the system and the pKa of the acid (de Roos and Sarelse, 1996), shown in a simplified form (assuming equal volumes of gas and liquid phases) in Equation (8.4). On the basis of this equation, a decrease of the pH of the system relative to the pKa of the acid would increase pKa/pH, increasing the significance of the term Fw/Kaw and therefore decreasing Kae.This would reflect the association of the acid at pH values below its pKa, enhancing its potential to partition into the lipid.

1 Kae = (8.4) Fo/Kao + (Fw/Kaw)(1 + pKa/pH)

The key parameters for both of these models (Equations 8.3 and 8.4) are the air/oil and air/water partition coefﬁcients. Kaw can be estimated for compounds using either thermodynamic parameters or the QSPR approach. We do not have the same options with Kao; therefore, unless a published value for Kao exists, it would be necessary to determine (i.e. measure) Kao before an estimate of Kae could be made. This is the major limitation of many models that experimentation has to be performed (or previously published values obtained) before an estimate can be made. However, once the partition coefﬁcients have been determined, it is possible to estimate Kae for any oil fraction. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

212 Food Flavour Technology

8.2.4 QSPR estimation of the air/emulsion partition coefﬁcient A QSPR model (Carey et al., 2002) described the concentration of volatile compounds in the headspace above low-lipid-concentration emulsion systems relative to that of water, and the ratio was used to express the ‘lipid effect’:

Concentration of volatile in headspace above emulsion × 100 Lipid effect (%) = (8.5) Concentration of volatile in headspace above water

The key components of the model were (log P)2 (an estimate of the octanol/water partition coefﬁcient squared), log sol (a QSPR estimate of solubility in water; Liang and Gallagher, 2 1997), a dipole vector term squared (DV ) and the oil fraction (Fo). The model was based 2 2 on 92 observations and had an R value of 0.83 and an Rcv of 0.80. The model was tested with an external test set of 25 compounds (i.e. ones not used in the development of the original model), which showed a correlation coefﬁcient of R2 = 0.82 between the observed and predicted values. The effect of (log P)2 and the oil fraction is shown in Fig. 8.2 (for a constant value of log sol and DV 2). The contour lines on the plot link points, which would

16.0

100

12.0 40 2 60

Log P 8.0

4.0

100

0.0 0.00 0.0005 0.001 0.0015 0.002 Oil fraction

Fig. 8.2 The effect of oil fraction and (log P)2 on the partitioning behaviour of compounds (for log sol = −2.0, and DV 2 = 3). P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 213

be expected to show the same change in their headspace volatile concentration based on their (log P)2 and the oil fraction. In Fig. 8.2, a compound with a (log P)2 of 6.1, in an emulsion with an oil fraction of 0.0005, would decrease in headspace concentration to the same extent (i.e. 20% decrease, lipid effect = 80%) as a compound with a log P2 of 1.8 in an emulsion with an oil fraction of 0.0015. Therefore, using the contour plot in Fig. 8.2 we can see how the volatile headspace concentration would be affected as changes in (log P)2 and Fo occurred. It is interesting to note that the contour lines in Fig. 8.2 are in fact curved, due to the use of quadratic terms in the QSPR function (Equation 8.6) and the interactive term between (log P)2 and oil fraction. This reﬂects that as (log P)2 increases, there is a general decrease in the relative headspace concentration of volatiles up to a (log P)2 of around 10, after which the headspace concentration increases again (presumably due to steric effects restricting partitioning into the emulsion). An estimate of the change in volatile headspace concentration relative to that of water can be made for compounds simply by calculating their log P, log sol and dipole vector (at a given oil fraction) and substituting these values into Equation (8.6):

Lipid effect = 107 − 6.3 × (log P)2 − 3.2 × log sol + 0.28 × DV2 4 2 +1.0E + 6 × Fo + 0.39 × (log P) − 2.0E + 5 × (log P) 2 ×Fo + 7.6E + 5 × log sol × Fo − 9.3E + 4 × DV × Fo (8.6)

The model does not estimate the absolute headspace concentration of aroma compounds, as the lipid concentration increases, it speciﬁcally models the relative change in headspace concentration. However, from this and Equation (8.3) it is possible to estimate the relative difference between Kao and Kaw, and since we can estimate Kaw by either the thermodynamic properties of a compound or using QSPR, this model also provides a means by which Kao itself could be estimated for a compound. It is important to remember that models based on data such as the emulsion model of Carey et al. (2002) are based on data collected using a speciﬁc range of experimental factors. In this case the oil fraction used was from 0.0 to 0.002 (0.0Ð0.2% lipid), and the quality of estimates for higher oil fractions would be less reliable. Equally with any QSPR model, it is important to know the range of compounds and conditions used in the development of the model. For example, it might be unwise to try and estimate the behaviour of acetaldehyde (molecular weight 44 Da) using a model based on compounds with molecular weights more than 100 Da.

8.2.5 Internet models and databases The rise of the internet has allowed the potential to develop and share databases and models for the estimation of the basic parameters governing volatile behaviour. EpisuiteTM from the United States Environmental Protection Agency was developed to provide data and models to help predict the behaviour of volatile compounds in lakes, rivers and sediments. The factors affecting volatile behaviour in these environments are effectively the same as those affecting flavour behaviour in food. Episuite can be used to estimate the octanol/water partition coefficient, Henry’s law constant (and hence the air/water partition coefficient), vapour pressure, aqueous solubility of compounds and numerous other parameters. The models used by Episuite are based on group or bond contribution methods similar to those P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

214 Food Flavour Technology

described in Section 8.1. The main advantages of using Episuite are its ease of use, the fact that it also accesses and displays a large database of experimental values, plus the fact that it is maintained and updated and that it is free to use. Consequently, it is now possible for researchers to use a common source of values rather than having many diverse models varying from research group to research group.

8.3 DYNAMIC SYSTEMS

Models for dynamic systems can take many forms due to the potential variation in the systems themselves. Systems may be unstirred, in which case stagnant layers may form and the movement of volatiles is then dependent on processes such as diffusion. Alternatively, as agitation increases, factors such as eddy diffusion exert a more significant influence on volatile release (see Chapter 7 for discussion of flavour release mechanisms).

8.3.1 Modelling flavour release from a retronasal aroma simulator An empirical model of flavour release from systems containing oil or thickeners at different temperatures was developed by Roberts and Acree (1996). The data set they used was obtained using their retronasal aroma simulator (RAS), which is a headspace sampling system designed to mimic eating conditions in mouth. In this instance, the sampling method results in one single measurement of volatile release, and the system is dynamic by virtue of the sampling apparatus through which a stream of air passes. The model developed was similar to that of Carey et al. (2002) in that it focused, not on absolute concentrations, but on the relative differences between a range of matrices and water. The equation generated to describe the air/product partition coefficient Kap (Equation 8.7) has three main components. The first relates to the partition between air and water for a compound and its affinity for lipids (the authors would have used the oil/water partition coefficient instead of log P had data been available). This is effectively an alternative form of Equation (8.3) in which the numerator and the denominator are divided by Kaw. The second describes the influence of the thickeners on flavour release due to changes in viscosity of the food phase. The third and final term describes the change in volatile behaviour as temperature varies. Va is the volume of air (Va has little impact in this equation, assuming similar volumes of headspace were analysed since the model is relative to water), ␩ is the viscosity of the matrix (cps), T is temperature in degrees Kelvin and T0 and ␩0 are standard conditions 298 K and 1 cps, respectively.

0.1 Kaw × Va ␩0 T − T0 Kap = × × exp (8.7) log P × Fo + Fw ␩ 21.5

The model described the behaviour of aroma compounds used in the development of the model well (R2 = 0.90), whereas single factors, such as log P, vapour pressure and boiling point, showed little correlation with the data. Data from the RAS have been compared with actual volatile concentration in the breath during the consumption of a range of foods, which has shown that, for some systems, it generates similar volatile proﬁles, although at higher P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 215

concentrations than observed in vivo (Deibler et al., 2001). By extrapolation, Equation (8.7) may be related to matrix-driven changes in volatile concentration in vivo. It is interesting to note (from Equation 8.7) that, if identical volumes of headspace are sampled, from equiviscous systems at the same temperature, then it is only the ﬁrst component of Equation (8.7) that has any effect. Since this is very similar in form to Equation (8.3), we have a situation where a dynamic headspace system is effectively modelled by a model that describes a static equilibrium.

8.3.2 Non-equilibrium partition modelling of volatile loss from matrices de Roos and Wolswinkel (1994) have also developed equations to describe the partitioning of volatile compounds in a range of matrices (relative to water itself) including thickened solutions, with and without the addition of lipid. Equation (8.8) models the amount of volatile retained in the product phase (Xn), after a series of n extraction steps, relative to the initial volatile concentration (X0). Their approach is typical of the theoretical approach to model development (Fig. 8.3): ∗ ∗ n V K 1 − V Xn = p pa + p + ∗/ ∗ (8.8) X0 Vp (Kpa Va Vp ) Vp

where Va and Vp are the volumes of the air and product phases, respectively, and it is assumed that only a small proportion of these volumes (Va* and Vp*) are actually in equilibrium at the air product interface (Kpa being the product/air partition coefﬁcient). The ratio between

Generate model Collect data from Use part of data set to equation based on model system theoretical estimate values for considerations constants in the model

Predict behaviour of remainder of data using model equation, constants and key variables

Does model describe No Model failed behaviour with Yes Model validated sufficient accuracy?

Fig. 8.3 Theoretical modelling of the behaviour of volatiles. The constants would be components of the equation that should be the same for all compounds, and the key variables are typically compound-speciﬁc factors, such as partition coefﬁcients. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

216 Food Flavour Technology

Table 8.1 Values used for the parameters in Equation (8.8) to model the behaviour of volatiles in water and 1% CMC with 1% oil added (from de Roos and Wolswinkel, 1994).

Parameter Water 1% CMC 1% CMC + 1% oil

V p*/V p 0.001 0.000135 0.000068 V a*/V p* 30 140 250 N 14 000 14 000 14 000

the volumes in equilibrium with each other (Va*/Vp*) is affected by gas flows, mixing in the solution and diffusion, and thereby represents differences in the mass transfer. The relative values of this ratio to Kpa determine whether it is the equilibrium or mass transfer that is the dominant factor, affecting volatile losses into the gas phase. Values of Vp*/Vp, Va*/Vp* and n can be determined experimentally from the study of volatile retention in different systems (given that Kpa is known). In practice, n can be set to a high number and the two volume ratios each determined as single values, following the solution of simultaneous equations. These values will be similar for different compounds (the differences between compounds are expressed by the values for Kpa) if the model is correct. When the model was used to describe the retention of volatile compounds in water, Va*/Vp* was found to be much smaller than Kpa, suggesting that the losses of volatile were mainly influenced by equilibrium conditions. For solutions of carboxy-methylcellulose (CMC) and CMC + oil, however, Va*/Vp* was larger, reflecting changes in mass transport. The values obtained for the components of the equation (Table 8.1) show a decrease in the proportion of the product in equilibrium with the gas phase (Vp*/Vp) as resistance to mass transfer increases (Va*/Vp*). The change in Vp*/Vp was independent of the value for Va*/Vp*, which de Roos and Wolswinkel attributed to changes in mass transport, not only in the liquid phase but also in the gas phase as viscosity increased. For the CMC solutions, the changes in Va*/Vp* were not accompanied by changes in Kpa itself; hence, the new values for Va*/Vp* had a direct effect on volatile retention. The addition of lipid did alter the equilibrium partition coefficient (the Kpa for methyl benzoate was about 400 for water and 900 for 1% oil), but the change was not as great as that for Va*/Vp* such that this system was also more dependent on restrictions to mass transfer than the equilibrium partition coefficient. The estimates of the amount of volatile retained, using the values in Table 8.1 and Equation (8.8), showed a good correlation with the experimental values (Fig. 8.4); hence the model has been validated. Once values have been found for the parameters present in Equation (8.8) for a given matrix, it should be possible to predict the proportion of any volatile retained in solution, based on values for Kpa.

8.3.3 Modelling the gas-phase dilution of equilibrium headspace It is also possible to model the changes in volatile headspace concentration with time, using online analytical techniques. Dilution of the gas phase is easily achieved by introducing a gas flow through the headspace and monitoring volatile concentration in the outflow. The apparatus used by Marin et al. (1999) consisted of a flask containing an aqueous sample with a headspace volume of about 25 mL. There were two ports for gas flow, one in and one out. After a period of equilibration, the out port was connected to a mass spectrometer P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 217

100 Water CMC 75 CMC + oil

50 Calculated

0 0 25 50 75 100 Observed Fig. 8.4 Observed versus calculated values (based on Equation 8.8 and the values in Table 8.1) for volatile retention in water, and solutions containing CMC or CMC + oil (from de Roos and Wolswinkel, 1994).

(designed to measure the concentration of volatiles in real time). This drew headspace out of the flask (about 70 mL/minute) causing air to flow in via the second port, thereby diluting the headspace. The first signal detected by the mass spectrometer effectively corresponded to undiluted equilibrium headspace, which served as a reference point. Thereafter, compounds were found to vary in the rate at which their headspace concentration decreased with time. The key element driving this was described by Equation (8.9), where ko is the overall mass transfer coefficient, and kg and kl are the mass transfer coefficients in the gas and liquid phases, respectively.

1 1 K = + aw (8.9) ko kg kl

Given that kg and kl were similar for different volatile compounds, the main factor responsible for differences in ko (which resulted in the different rates of headspace depletion for the compounds studied) was the air/water partition coefficient. Compounds with high values −2 for Kaw (10 ) were readily depleted from the headspace, whereas compounds with low −5 Kaw values (10 ) were not (Fig. 8.5). When compounds have a low Kaw,thevalueof Kaw/kl decreases in significance (trends towards zero) and the main driving factor is then the mass transfer coefficient in the gas phase. Similarly, when Kaw values are high, Kaw/kl increases in significance, and the mass transfer in the liquid phase becomes important such that surface depletion of volatiles from the solution decreases their maximum potential headspace concentration. This model describes the behaviour of compounds in situations where equilibrium headspace is disturbed by an external gas flow, such as sniffing a glass of wine. The headspace system, which the model describes, is very specific, but other parameters can be built into the model to describe how other factors such as the airflow rate and surface area of the solution affect the overall mass transfer coefficient (Marin et al., 2000). The two parameters, kg and kl, were not determined by solving simultaneous equations, but by an iterative process of parameter fitting using Matlab. The initial values were set to those found in the literature; these were gradually modified until the theoretical change P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

218 Food Flavour Technology

−5 Low Kaw 10 Intensity

−2 High Kaw 10

Time

Fig. 8.5 Schematic representation of the differences in the rate of headspace depletion for compound with different air/water partition coefﬁcients as a function of time.

in headspace volatile concentration matched the experimental data. The fact that kg and kl were similar for different compounds was indicative of the validity of the model. If substantial differences in kg and kl were needed (for each compound) to ﬁt theoretical curves to experimental data, they would be little more than variable ﬁtting factors.

8.3.4 Modelling the gas-phase dilution of equilibrium headspace above emulsions In addition to describing the behaviour of purely aqueous systems, this model has been applied to the study of emulsion systems. An emulsion can result in a decrease in the volatile headspace concentration, and the new equilibrium partition coefficient Kae will be lower than Kaw. If the differences in the two partition coefficients are large enough, they may result in differences in behaviour during headspace dilution similar to those shown schematically in Fig. 8.5. This, however, will depend on the behaviour of the volatile in the aqueous phase. If the volatile can readily partition between the lipid and the water, then it can partition between the liquid and gaseous phases. If, however, the rate of partitioning between the lipid and the water is slow, the volatiles may be effectively trapped in the lipid phase. Substantial differences were observed in the rate of gas-phase dilution (relative to the initial gas-phase concentrations) of ethyl octanoate in the presence of emulsified lipid (Fig. 8.6). It was found (Doyen et al., 2001) that the ethyl octanoate partitioned readily between the lipid and the water phases such that the emulsion appeared to behave as one homogenous phase. These differences were related to the change in the partition coefficient, which altered overall mass transfer as described by Equation (8.9). Kaw/kl would have decreased in significance and the relative headspace concentration would depend more and more on the gas-phase mass transfer coefficient as the lipid fraction increased. ko only showed significant change as Kae varied, with the two mass transfer coefficients kg and kl having similar values to those found for water in this system. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 219

100 0.0200 0.0100 0.0050 0.0025 0.0000

Rel HS concentration (%) 25

0 0 01 02 Time (minute)

Fig. 8.6 Behaviour of ethyl octanoate in emulsions (oil fractions from 0.0 to 0.02). Relative changes in headspace concentration over time during headspace dilution. Reprinted with permission from Doyen et al. (2001), copyright 2001 American Chemical Society.

8.3.5 Modelling the rate of volatile equilibration in the headspace above emulsions Harrison et al. (1997) also developed models to describe the behaviour of emulsion systems. The theoretical system they modelled was the inverse of that modelled by Marin et al. (1999). They modelled the way in which an emulsion would release aroma into the gas phase, from initial conditions in which there were no molecules of the aroma compound in the gas phase, up to the point of equilibrium. Their model was based on penetration theory where bulk elements of the emulsion phase come into contact with the interface, so that partitioning can take place, rather than a stagnant layer situation. The equation they produced describes the concentration of the volatile in the gas phase over time cg(t): Kaece(0) Va hD Aaet cg(t) = × 1 − exp − 1 + × (8.10) KaeVa/Ve+1 KaeVa Va

On ﬁrst sight, this equation looks rather complex. However, Va and Ve are the volumes of air and emulsion phases, respectively, which for simplicity we could assume to be identical and equal to 1. Equation (8.10) could then be written as follows: Kaece(0) 1 cg(t) = × 1 − exp − 1 + × hD Aaet (8.11) Kae+1 Kae

−2 Kae for aroma compounds is typically less than 10 ; therefore, Kae + 1 is effectively 1 and likewise 1 + 1/Kae can be approximated by 1/Kae to give Equation (8.12). hD Aaet cg(t) = Kaece(0) × 1 − exp − (8.12) Kae

This allows us to see more clearly the key elements of the equation. The ﬁrst term in Equation (8.12), Kaece(0), is effectively the partition coefﬁcient multiplied by the initial P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

220 Food Flavour Technology

concentration of the compound in the emulsion: this product gives the headspace concentration at equilibrium (when time (t) trends towards infinity). The second part of the equation, 1 − exp(−(hDAaet/Kae)), governs the rate at which the headspace moves from its initial concentration (i.e. containing no volatile) towards equilibrium. The rate is influenced by the mass transfer coefficient hD, the area of the interface (Aae) and the partition coefficient. The most interesting part of this section is the mass transfer coefficient (hD(␾,d)), defined for an emulsion with a specific droplet size (d in ␮m) and oil fraction (␾) which Harrison and co-workers described with the equation: 1 + s ␾ h (␾, d) = h (0) × exp(−b(ln 10) × (8.13) D D 2 d

In Equation (8.13), hD(0) is the mass transfer coefficient of the compound in water and is modified by the oil fraction, droplet size and the exponent s, related to the contact time of the bulk elements of the emulsion and the surface and a constant b. Using these equations and assuming constant values for all but one parameter, it is possible to determine the extent to which a given parameter may influence volatile behaviour within a range of values (e.g. an oil fraction range from 0.0 to 1.0). The theoretical relationship between oil fraction and droplet size showed that, at a given oil fraction, the mass transfer coefficient should increase as the droplet size increases. Equally, they showed that, for an emulsion with a specific droplet size, the mass transfer coefficient should increase as oil fraction decreases. The model can also be used to determine the actual value of parameters such as the exponents, from experimental data for the rate of equilibration, where the characteristics of the system (␾, d, etc.) are known. This is possible despite there are many terms in Equations (8.12) and (8.13) because most of them have fixed values that are easily measured, leaving very few to determine mathematically via the solution of equations. It is interesting to note the different methods used in the estimation of the overall mass transfer coefficients (Equations 8.9 and 8.13). Both are dependent on a modification of values for liquid-phase mass transfer coefficients. Equation (8.9) (with Kae substituted for Kaw) does not have any components related to droplet size, but does express the influence of oil fraction via Kae. Such factors were not necessary in this estimation of overall mass transfer, since in the system studied, the particle size was small and uniform such that it did not restrict or influence volatile partitioning. Extrapolating this approach to mass transfer estimation to systems with a higher oil fraction (where droplet coalescence is likely to occur) may result in error. This highlights one of the important aspects of models; it is important to remember how they were derived, and in the case of models based on data, the limits of the dataset.

8.4 IN VIVO CONSUMPTION

There are fewer models that describe volatile release during the eating process, mainly due to the number of variables that need to be considered and the limited amount of experimental data available. The more variables present in an equation, the harder it is to solve mathematically. Deﬁning the value of constants in the equation is also an issue, a situation only made worse by limited datasets. Many of the dynamic headspace models that have been produced were generated in an attempt to describe volatile behaviour under non-equilibrium conditions, such as might be found in vivo. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 221

25 Breath/headspace (%)

0 −1 −2 −3 −4 −5

Log Kaw

Fig. 8.7 In vivo breath volatile concentration relative to the static equilibrium headspace concentration (%) as a function of the air/water partition coefﬁcient, observed during the consumption of aqueous solutions of aroma compounds.

The volatile concentrations in the breath during the consumption of standard foods (cheese, biscuits, etc.) can be substantially lower (about 100-fold) relative to those in the headspace (Deibler et al., 2001). Even simple systems such as aqueous solutions of volatiles, which have minimal restrictions on mass transfer (as described by de Roos and Wolswinkel, 1994), can release far lower amounts of aroma than expected simply based on the air/water partition coefﬁcient (Linforth et al., 2002). The relationship between volatile release into the breath and headspace is dependent on the air/water partition coefﬁcient itself (Fig. 8.7). Compounds with high Kaw values −2 −5 (10 ) show poor release relative to those with lower Kaw values (10 ) compared with their thermodynamic equilibrium, similar to their headspace concentration stability during dilution (Section 8.3.3). The reason for the similarity of the two processes in both cases is due to the relationship with Kaw. In the case of the dynamic headspace dilution, the compounds with high Kaw values have to transfer a high proportion of molecules into the gas phase to try and maintain equilibrium, and the surface layer is rapidly depleted of volatile. In vivo the limited surface area available for volatile transfer into the gas phase and inaccessibility of the bulk phase also results in surface depletion of volatiles with high Kaw values as they try and equilibrate with a proportionately larger gas phase. This shows the limited extent of mixing between surface and bulk phases that can occur in vivo, which should be considered in model development. The relationship shown in Fig. 8.7 can be described by an equation (Equation 8.14) which has such simplicity of form that it barely looks like a model. However, the equation describes the data and can be used for prediction of release from aqueous phases including beverages and more importantly saliva, which is a phase that many volatiles pass through as they release from foods.

Breath (%) =−19 × K − 23 (8.14) Headspace aw P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

222 Food Flavour Technology

8.4.1 Modelling release from emulsions during consumption Many models of food flavour behaviour were formulated before in vivo data were available and often assume equilibrium is achieved. One such model is that of McNulty (1987), which considered the fate of emulsions during consumption. He argued that, during consumption, saliva dilutes the sample, and since the air/emulsion partition coefficient of an emulsion is dependent on the oil and water fractions (Equation 8.3), the partition coefficient will change during the eating process, thereby affecting volatile release. In order for this to happen, the aroma compounds need to partition readily between the oil and water phases (otherwise the volatiles would effectively be trapped in the lipid), a phenomenon that has been observed experimentally (Fig. 8.6). The greater the dilution, the more significant this factor will become in determining the breath volatile concentration. In contrast to emulsions, the fate of volatiles in an aqueous solution is purely dependent on the extent of salivary dilution, more dilution resulting in a lower aqueous phase concentration, and hence lower breath volatile concentration. This would make release from the two phases more similar in vivo and much less than observed in equilibrium headspace studies. Do such processes affect volatile release in vivo? Is there sufficient dilution by saliva to affect volatile delivery? McNulty (1987) presented sensory data that were consistent with a difference in emulsion behaviour compared to purely aqueous systems, a difference also observed by de Roos and Wolswinkel (1994). Direct measurement of breath volatile concentrations also agrees with this hypothesis (Doyen et al., 2001). However, it is estimated that a 10- to 30-fold dilution of a sample with saliva would be necessary to account for the in vivo release observed (Linforth et al., 2002), which seems highly unlikely. An alternative explanation is that the lipophilic compounds in water typically have high Kaw values and deliver very inefficiently (Fig. 8.7), whereas those in emulsions have much lower Kaw values, yet deliver more efficiently. The net result is that release from the two systems is more similar than expected from equilibrium data and is principally caused by differences in partition- dependent mass transfer and not salivary dilution. The further implication of this is that lipids reduce the odour of foods (via the orthonasal route), whilst still allowing effective retronasal delivery, thus making the flavour experience more balanced relative to low-fat equivalents.

8.4.2 Effect of gas flow on volatile equilibration above emulsions One model that predicted changes in mass transfer in vivo, but did not take into account dilution of the sample by saliva, was proposed by Harrison and Hills (1997) in an extension to their initial model. The initial model (Equation 8.10) described gas-phase equilibration under static headspace conditions, and was then expanded to incorporate the effects of gas flow rate, a key factor when considering the dynamics of release in mouth. As with the model that studied equilibration rates under static conditions, one of the key factors in the equation was the mass transfer coefficient. Changes in the mass transfer coefficient as the oil fraction increased resulted in predictions of higher steady-state gas-phase volatile concentrations than would be expected on the basis of their oil/water partition coefficients. On the basis of the oil/water partition coefficients, the equilibrium headspace concentration for heptanone at oil fractions of 0.1 and 0.5 would be approximately 20 and 2.5% of that observed for water. The predicted steady-state gas-phase concentrations (using the model), however, were 60 and 20% relative to that of water. This again shows the potential for mass transfer-dependent reductions in flavour delivery differences without the need for salivary dilution. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 223

8.4.3 Modelling volatile transfer through the upper airway Harrison (2000) has attempted to model, not only the in-mouth partition of volatiles, but also their transfer through the upper airways. The potential complexity of the model was reduced by selecting chewing gum as the food system, since it shows minimal changes in surface area during the eating process. Volatile release from the chewing gum itself was described by Equation (8.15), where the mass of volatile transferred across the interface with time (dm/dt) was dependent on the surface area of the gum (A), mass transfer (hD), the saliva/gum partition Ksg and the concentration of the volatile in the gum and saliva (cg and cs, respectively). dm = Ah (K c (t) − c (t)) (8.15) dt D sg g s

Higher values for Ksg (0.2) would (according to the model) result in higher initial volatile release followed by a decrease over several minutes, whereas lower values (0.05) would result in a more constant release and a lower overall saliva volatile concentration. Having modelled the movement of volatiles from the gum into the saliva (based on the assumption that all volatile release takes place via saliva), the effect of salivary dilution and the air/saliva partition coefficient was estimated to give the in-mouth gas-phase volatile concentration. Thereafter, the transfer of volatiles from the throat through the upper airway was modelled by an equation containing nine parameters, describing factors such as the dimensions of the upper airway, breath velocity, mass transfer and partition coefficients. This latter equation effectively modelled volatile movement through a tube where the volatiles partitioned in and out of the liquid surface. It was predicted that a single pulse of hydrophilic compounds entering the gas flow of the tube would be retarded and smeared (as they partitioned in and out of the mucous membranes), relative to hydrophobic compounds that would show much less retention and peak broadening. Unfortunately, no experimental evidence was available to confirm the predictions of the models. Experimental data might have also helped to define the useful operating range of some parameters, and allowed others to be reduced to constants (e.g. those related to diffusion in air, breath velocity and the length of the throat). This emphasises that purely theoretical models need experimental data at some stage to allow further development of the model and, ultimately, to test the validity of the model.

8.4.4 Non-equilibrium partition model for in vivo release de Roos and Wolswinkel (1994) also considered in vivo release of volatiles from chewing gum using their non-equilibrium partition approach, and produced a model describing in vivo release during consumption. The initial model (based on the partition of volatile from the gum to the saliva) contained only three parameters, and did not correlate closely with experimental data, so further physicochemical parameters were introduced into the equation. The ﬁnal equation generated (Equation 8.16) included both the gum/water (Kgw) and air/water partition coefﬁcients, since both release into saliva and the gas phase appeared to be important factors (in contrast to the model proposed by Harrison, 2000). The saliva/gum volume ratio (Vw/Vg) was one of the key parameters incorporating the effect of volatile partitioning into saliva, whereas the volume ratio between the gum and gas phase Va/Vg affected the extent of partitioning into the gas phase.

K Xn = gw n (8.16) X0 (Kgw + Vw/Vg + (Va/Vg)Kaw) P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

224 Food Flavour Technology

As with their model for the dynamic equilibrium (Section 8.3.2), in vitro values for Vw/Vg and Va/Vg can be determined from experimental results. In this instance, there was ‘natural variation’ due to differences between people, and estimates for these parameters were based on the average panellist. Once these values had been determined, the amount of volatile retained in the gum phase was estimated, based on Kgw and Kaw for each compound. In addition, it was possible to consider the effect of short or long eating periods by using different values for n. Thereafter, it may be assumed that the changes in breath volatile concentration will be proportional to that of the chewing gum volatile concentration over time. The one limitation with this approach to empirical modelling is the range of foods to which it can be applied. The dataset (residual volatile in bolus) was produced by the extraction of volatiles from chewing gum (after it had been chewed for different periods) followed by quantitative analysis. In the case of chewing gum where the bolus remains intact and unswallowed, this is a relatively easy task. For foods where the bolus disintegrates, dissolves, or is swallowed, the collection of suitable data presents more of a problem. In this instance alternative experimental approaches are necessary.

8.4.5 Modelling flavour release using time–intensity data Data sets for empirical modelling can also be gathered from individuals eating foods and recording the change in perceived flavour with time (timeÐintensity (TI) methodology). Here, the data set takes into account not only the processes of mass transfer, diffusion, etc., in mouth but the whole process of volatile transport to the olfactory epithelium and the perceptual mechanism. Moore et al. (2000) prepared a series of protein gels that contained fat droplets and a volatile. Samples of the gels were eaten by their panel, TI curves were recorded, and the parameters were extracted from the TI curves (maximum intensity etc.) compared with those estimated from the model. The model took into account factors such as particle size reduction due to chewing, swallowing of volatile-laden saliva, retronasal airflow, the area of the salivaÐair interface and mass transfer between the saliva and air. In addition, measurement of gel texture and fat droplet size was made to see if these had a significant impact on perception that could be described by the model. The best correlation between experimental values and those obtained from the model was for the area under the curve up to the point of maximum intensity and the time to maximum intensity. Other values did not show such strong correlations, which may have been because at 30s, the panellists were allowed to swallow, substantially altering the amount of bolus retained in mouth. No correlations were found between the fat droplet size or the hardness of the gels and volatile release. This may have been due to the limited differences in these parameters, which had a range of ×10 and ×5, respectively. This lack of variation in the samples (and subsequently the TI data set) is one of the key problems encountered in this type of work because, without sufficient variation in the data set, it is virtually impossible to develop and evaluate empirically based models.

8.4.6 QSPR of in vivo volatile release from gels Flavour release data sets can be generated by direct measurement of the volatile composition of the breath during eating (Linforth et al., 1996). This does not take into account the P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 225

perception of the compounds, but it does allow the study of a wide range of different compounds with different properties. Release of the different compounds can be readily compared in absolute terms (e.g. mg/m3), whereas it is difﬁcult to express the perception of very diverse compounds on one unifying sensory scale. Linforth et al. (2000) used this approach to the study of ﬂavour release from gelatine/sugar gels. They found that the maximum breath volatile concentration (Imax) observed varied by a factor of 10 000 depending on the compound studied (all compounds were present in the gels at the same concentration), providing a good range of variation in the data set, which was suitable for modelling. Since the gel matrix and the panellists were constant, and the key differences were due to the compounds themselves, QSPR was the obvious choice for model generation. Physicochemical parameters were calculated using chemical modelling software to describe the properties of each compound numerically. These were then analysed statistically to determine which parameters best described the variation in the data set. Three parameters were selected and an equation was produced (Equation 8.17), which could be 2 2 used to estimate Imax.TheR of the model was 0.88 and the Rcv was 0.82, indicating reasonable predictive power. The parameters of Equation (8.17) describe the hydrophobicityÐhydrophilicity of a compound (log P), vapour pressure (log ␳ L; Liang and Gallagher, 1998) and the compound’s size and shape (Hartree energy). All these values were calculated using chemical modelling software. From the model, it is possible to predict the behaviour of any other compound by calculating the three parameters from the molecular structure and substituting them into Equation (8.17):

=− . + . × + . × − . × − . × 2 Log Imax 1 3 0 9 log P 0 7 log pL 0 06 Energy 0 15 (log P) 2 2 3 + 0.13 × (log pL) + 1.4E − 3 × Energy − 7.3E − 6 × Energy (8.17)

The model can be represented as a two-dimensional contour plot (Fig. 8.8), where the lines join regions of the plot where the values of Imax are the same. Log Imax for a hydrophilic compound with a high vapour pressure (e.g. ethanol) would be found in the top left of Fig. 8.8. In contrast, the Imax for a hydrophobic compounds with lower vapour pressure (e.g. decanol) would be found towards the opposite corner. The contour lines clearly run from the top left towards the lower right-hand corner indicating that the Imax values would be the same for these two diverse compounds, and this is indeed the case, with ethanol and decanol having virtually identical release characteristics in this system. The contour plot also shows the range over which parameters exert the greatest influence. For example, log P has a major effect on a compound’s behaviour up to a value of 1.40. Thereafter, increasing log P has virtually no effect on Imax. The QSPR models describing volatile release may be simple linear relationships, where one factor directly influences another. However, such relationships rarely continue indefi- nitely and in many instances quadratic (or cubic) functions may best represent the relationship between a modelling parameter and the response being modelled, hence the power terms in Equation (8.17) and the curved contours in Fig. 8.8. In addition to modelling the differences in Imax, it was also possible to describe the variations in the temporal dimension using the same parameters. Consequently, the entire flavour release curve could be estimated for any given compound (Fig. 8.9). P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

226 Food Flavour Technology

2.77

1.46 1 0.5 L p

0.15 Log -0.5 0 -1 -1.5

-2 -1.16 -2.5

-2.47 -1.32 0.04 1.40 2.75 4.11 Log P

3 Fig. 8.8 Contour plot of log Imax (mg/m ) as a function of log P and log pL for a ﬁxed Hartree energy value.

8.5 CONCLUSION

There are clearly many different approaches to modelling aroma release. However, it is evident that there is one common theme that there are major differences between compounds, which need to be taken into account in any model. This can be achieved using partition

1.5

Observed 3 Predicted 1

0.5 Concentration (mg/m )

0 0 0.5 1 Time (minute)

Fig. 8.9 Observed and predicted release curves for hexanol present at 100 ppm in a 6% gelatine gel. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

Modelling ﬂavour release 227

coefficients, mass transfer coefficients or physicochemical parameters, one or more of which can be found in every equation describing volatile behaviour. Models can describe a range of systems including the static partition, dynamic partition or in vivo behaviour of volatile compounds and their interactions with different matrices. They can be generated by a theoretical consideration of the system or empirically from a data set. If models are generated via a data set, the model should (ideally) be validated using an experimental test set of samples not used in the development of the original model: the quality of the model is then determined by the accuracy with which the model predicts the values for the test set. This guards against models based on random correlations between parameters and data, which do not truly describe the variation in the data set. Likewise, theoretical models should be validated, and equations should be generated (using real data) and solved to find the values of parameters that are constant in a given system (e.g. gas flow rate or volume). Thereafter, the model should predict all available experimental data keeping the constants constant, and using values for key variables in the equation (e.g. partition coefficients). It is important to remember the range of the data (volatiles and matrices) used to develop and test a model, since extrapolation beyond these limits has little (if any) validity.

REFERENCES

Buttery, R.G., Guadagni, D.G. and Ling, L.C. (1973) Flavour compounds: volatilities in vegetable oil and oil-water mixtures. Estimation of odour thresholds. J. Agric. Food Chem. 21, 198Ð201. Carey, M.E., Asquith, T., Linforth, R.S.T. and Taylor, A.J. (2002) The effect of a cloud emulsion matrix on volatile partitioning. J. Agric. Food Chem. 50, 1985Ð1990. de Roos, K.B. and Sarelse, J.A. (1996) Volatile acids and nitrogen compounds in prawn powder. In: Flavour Science: Recent Developments (eds A.J. Taylor and D.S. Mottram), Royal Society of Chemistry, Cambridge, pp. 13Ð18. de Roos, K.B. and Wolswinkel, C. (1994) Non-equilibrium partition model for predicting flavour release in the mouth. In: Trends in Flavour Research (eds H. Maarse and D.G. Van der Heij), Elsevier Science B.V., Amsterdam, pp. 15Ð32. Deibler, K.D., Acree, T.E., Lavin, E.H., Taylor, A.J. and Linforth, R.S.T. (2001). Flavor release measurements with retronasal aroma simulator. In: Flavour 2000 (ed. M. Rothe), Eigenverlag, Bergholz Rehbrucke, Germany, pp. 41Ð48. Doyen, K., Carey, M., Linforth, R.S.T., Marin, M. and Taylor, A.J. (2001) Volatile release from an emulsion: headspace and in-mouth studies. J. Agric. Food Chem. 49, 804Ð810. Harrison, M. (2000) Mathematical models of release and transport of flavors from foods in the mouth to the olfactory epithelium. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 179Ð191. Harrison, M. and Hills, B.P. (1997) Effects of air flow-rate on flavour release from liquid emulsions in the mouth. Int. J. Food Sci. Technol. 32,1Ð9. Harrison, M., Hills, B.P., Bakker, J. and Clothier, T. (1997) Mathematical models of flavor release from liquid emulsions. J. Food Sci. 62, 653Ð664. Katritzky, A.R., Mu, L. and Karelson, M. (1996) A QSPR study of the solubility of gases and vapors in water. J. Chem. Inf. Comput. Sci. 36, 1162Ð1168. Katritzky, A.R., Wang, Y.L., Sild, S., Tamm, T. and Karelson, M. (1998) QSPR studies on vapor pressure, aqueous solubility, and the prediction of water-air partition coefficients. J. Chem. Inf. Comput. Sci. 38, 720Ð725. Liang, C. and Gallagher, D. A. (1997) Prediction of physical and chemical properties by quantitative structure- property relationships. Am. Lab. March, 34Ð40. Liang, C.K. and Gallagher, D.A. (1998) QSPR prediction of vapor pressure from solely theoretically-derived descriptors. J. Chem. Inf. Comput. Sci. 38, 321Ð324. P1: SFK/UKS P2: SFK c08 BLBK221-Taylor/Linforth November 23, 2009 13:46 Printer Name: Yet to Come

228 Food Flavour Technology

Linforth, R.S.T., Friel, E.N. and Taylor, A.J. (2000) Modeling aroma release from foods using physicochemical parameters. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 166Ð178. Linforth, R.S.T., Ingham, K.E. and Taylor, A.J. (1996) Time course profiling of volatile release from foods during the eating process. In: Flavor Science: Recent Developments (eds A.J. Taylor and D.S. Mottram), Royal Society of Chemistry, London, pp. 361Ð368. Linforth, R., Martin, F., Carey, M., Davidson, J. and Taylor, A. J. (2002) Retronasal transport of aroma compounds. J. Agric. Food Chem. 50, 1111Ð1117. Marin, M., Baek, I. and Taylor, A. J. (1999) Volatilerelease from aqueous solutions under dynamic headspace dilution conditions. J. Agric. Food Chem. 47, 4750Ð4755. Marin, M., Baek, I. and Taylor, A. J. (2000) Flavor release as a unit operation: mass transfer approach based on a dynamic headspace dilution method. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 153Ð165. McNulty, P.B. (1987) Flavour release Ð elusive and dynamic. In: Food Structure and Behavior (eds J.M. Blanshard and P. Lillford), Academic Press, London, pp. 245Ð258. Moore, I.P.T., Dodds, T.M., Turnbull, R.P. and Crawford, R.A. (2000) Flavor release from composite dairy gels: a comparison between model predictions and time-intensity experimental studies. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 381Ð394. Reid, R.C., Prausnitz, J.M. and Poling, P.E. (1987) Properties of Gases and Liquids, McGraw-Hill, New York. Roberts, D.D. and Acree, T. (1996) Model development for flavour release from homogenous phases. In: Flavour Science: Recent Developments (eds A.J. Taylor and D.S. Mottram), Royal Society of Chemistry, Cambridge, pp. 399Ð404. Sorrentino, F., Voilley, A. and Richon, D. (1986) Activity coefficients of aroma compounds in model food systems. J. Am. Inst. Chem. Eng. 32, 1988Ð1993. Voilley, A., Espinosa Diaz, M.A., Draux, C. and Landy, P (2000) Flavour release from emulsions and complex media. In: Flavor Release (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 142Ð152. Voilley, A., Simatos, D. and Loncin, M. (1977) Gas phase concentration of volatiles in equilibrium with a liquid aqueous phase. Lebensmitt. Wiss. Technol. 10, 45Ð49. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

9 Instrumental methods of analysis Gary Reineccius

This chapter provides an insight into how one approaches the instrumental analysis of flavour. Only the aroma component of flavour is discussed; this excludes both taste and chemesthetic effects that are also universally considered to be components of flavour. The aim is to provide an understanding of the unique challenges faced in this analysis and how and why particular analytical approaches are taken to solve a particular flavour problem. Issues of sample preparation, aroma isolation, compound selection (when an objective), quantification and identification are discussed. Since this topic is very broad, one cannot provide much detail, thus reviews (e.g. Reineccius and Anandaraman, 1984; Widner, 1990; Marsili, 1997, 2002a Parliment, 1997; Reineccius, 2006) or original research papers are suggested for greater depth.

9.1 ANALYTICAL CHALLENGES

The aroma component of flavour is due to a complex mixture of volatile organic chemicals. A ‘simple’ flavour may have 100Ð300 volatile constituents (e.g. strawberry or grape). Foods that are more ‘complex’ in flavour, for example, those resulting from the Maillard reaction (e.g. coffee, meat or chocolate), may contain 900 or more volatile constituents (Nijssen et al., 1996). Of these volatiles, a limited number may be adequate to characterise the aroma of a food. For example, benzaldehyde is recognised by most individuals as cherry and citral is recognised as lemon. While in some cases one chemical may provide a recognisable odour character, most food aromas require 10Ð30 additional volatile compounds to provide a more rounded aroma that is characteristic of the food (see also Chapter 1). Thus, while the analytical chemists may wish to focus their efforts on a given chemical or a limited number of chemicals, they must attempt to deal with hundreds of chemicals simultaneously, many of which make little or no contribution to aroma. The complexity of the analytical task is further complicated by the large number of volatile compounds that make up the total ‘pool’ of aroma chemicals. The task would be sufficiently formidable if the total pool of aroma constituents equalled that of the most complex flavour, for example meat, and all other flavours were a result of a different balance of these components. Unfortunately, the total pool of aroma constituents identified to date now exceeds 7000. Thus, in aroma analysis, the potential number of aroma compounds one could conceivably encounter is immense. The wide chemical diversity of aroma chemicals found in foods further complicates these studies. Aroma compounds comprise a large number of different chemical structures P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

230 Food Flavour Technology

and, therefore, of physical and chemical properties. It is difficult to match an analytical method to a physical property, such as volatility or molecular size, when there is such a broad range in these properties. The only attribute that all aroma chemicals have in common is volatility, but the wide range in volatility one encounters renders this commonality of limited value. For example, hydrogen sulfide has a boiling point of −60◦C and vanillin one of 284◦C: a method based on volatility designed to isolate hydrogen sulfide is not likely to be appropriate for the isolation of vanillin. It is equally difficult to design an isolation or analytical technique on a chemical property such as reaction with a derivatising reagent (e.g. 2,4-dinitrophenylhydrazine for carbonyls) when aroma constituents comprise so many different chemical structures. There is often little in common among chemicals known to be odour active, other than they contribute to aroma. It is also problematic that aroma compounds may be sensorially significant when present in extremely low concentrations. The analytical chemist may have to work with femtogram or attogram (10−18) quantities of flavour constituents (Acree, 1993). This problem is ac- centuated by the fact that these traces of aroma chemicals are generally in a complex food system that contains thermally labile constituents (e.g. sugars, proteins, lipids and vitamins), good emulsifiers (e.g. lecithins and proteins), aqueous and fat-soluble components and other volatile components (water). A seemingly impossible task (complete aroma isolation) becomes even more complicated as one considers it. One can readily appreciate that the isolation and analysis of aroma from a food product is indeed challenging (Teranishi, 1998). No single method yields an accurate picture of the aroma constituents in a food. Every method produces some picture of the aroma profile, but the profile is strongly determined by the biases introduced by the methodology itself. The aroma profile obtained from using five different isolation techniques on the same standard mixture of volatiles is illustrated in Fig. 9.1. Ideally, the bars would all be the same height if the methods were 100% efficient in the extraction of all volatiles. It is quite clear that each method gives its own unique profile. Thus, in method selection, the analyst must be

Fig. 9.1 Inﬂuence of aroma isolation method on the recovery of selected aroma compounds from an aqueous solution (equal concentrations by weight). 1, ethanol; 2, propanol; 3, butanol; 4, octane; 5, decane; 6, ethyl propanoate; 7, ethyl butanoate; 8, ethyl pentanoate; 9, 2-heptanone; 10, acetophenone; 11, benzyl acetate; 12, methyl salicylate; 13, L-carvone; 14, ␤-ionone; 15, methyl anthranilate; 16, ethylmethylphenyl glycidate; 17, isoeugenol. (Reprinted with permission from Leahy and Reineccius, 1984; copyright 1984 American Chemical Society.) P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 231

knowledgeable in the ﬁeld and have a clear understanding of the study objectives. There is no universal approach to the instrumental analysis of aroma; each analysis must have a uniquely designed protocol. With this introduction, approaches for aroma isolation are discussed.

9.2 AROMA ISOLATION

Numerous methods can be used to isolate aroma constituents from other food components (proteins, carbohydrates, water, fats, minerals, vitamins, etc.). The primary principles used to isolate aroma constituents from the major food components are volatility and/or solubility. It is problematic that water is, with few exceptions, the most abundant volatile constituent in a food. This creates a problem in aroma isolation since isolation methods based on volatility also include water from the food. The analyst does not obtain just aroma constituents but rather a dilute ‘solution’ of aroma constituents in water. The aroma components must then be isolated from water to permit concentration and further analysis. This additional step introduces further errors to the method (such as aroma losses and artefact contamination) as well as adding time. Most (but not all) aroma compounds have greater solubility in organic solvents than in water, while the bulk of the major food constituents are the opposite. Unfortunately, this rule is complicated by the fact that food lipids are also soluble in organic solvents. Thus, a solvent extract of a food yields not only the aroma constituents but also triglycerides, mono- and diglycerides, phospholipids, vitamins, chlorophyll, carotenoids, and so on. A useful aroma isolate cannot be prepared via solvent extraction if lipids are present in the food unless further steps are taken to separate the aroma from the lipids. Again, this additional handling adds the potential for more error (loss of volatiles and potential for artefact formation) and adds time to the analysis. The ability to utilise either volatility or solubility as a basis for aroma isolation is further complicated by the range in volatility and/or solubility displayed by aroma constituents. Earlier, it was noted that hydrogen sulfide has a boiling point of −60◦C while vanillin boils at 284◦C (with some degradation). Compounds such as hydrogen sulfide have little solubility in organic solvents, while terpenes are very soluble in non-polar solvents (e.g. diethyl ether). It is evident that no physical or chemical properties are uniquely held by aroma constituents to permit their isolation from foods, nor are the ‘somewhat unique’ properties (e.g. boiling points) similar enough to permit an efficient total aroma isolation. The following section focuses on the impact of methodology but present little detail on the methodology itself. The numerous in-depth reviews referenced early in this chapter and selected text references are recommended for greater detail.

9.2.1 Aroma isolation methods based on volatility One property that an aroma constituent must inherently possess is volatility. It must exhibit sufﬁcient vapour pressure to be present in the gas phase at a concentration detectable by the olfactory system. Thus, it is understandable that numerous aroma isolation techniques are based on volatility, such as static headspace, dynamic headspace, molecular distillation, steam distillation and direct injection techniques (where food is placed in the apparatus itself and heated to volatilise aroma constituents). Some initial generalisations are made about these approaches, and method-speciﬁc comments follow. All the methods based on volatility will be strongly biased towards those aroma P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

232 Food Flavour Technology

constituents that are most volatile in the food system to be studied. This qualiﬁcation about ‘the system to be studied’ must be included since volatility is dependent on the food composition. Efﬁciency of recovery by these methods does not follow in order of volatility (vapour pressure) of the pure compounds but, rather, of its vapour pressure over the food system. Generally, compounds with the greatest vapour pressure in the pure states are low-molecular- weight substances that often have some water solubility. Dissolution in water reduces vapour pressure according to Raoult’s law, and thus, these very volatile constituents may not be easily recovered from aqueous-based foods. The medium-volatility compounds, which are still very volatile but have lower water solubility, often have the highest vapour pressure in aqueous-based food systems. High-molecular-weight compounds have little vapour pressure in the pure form or in aqueous systems and, thus, are not readily recovered. One can apply similar reasoning to volatiles that are lipophilic when placed in lipid-containing food systems. Thus, methods based on volatility are very biased towards aroma constituents that have the greatest vapour pressure over a given food.

9.2.1.1 Static headspace methods Direct analysis of the equilibrium headspace above a food product would appear to be an ideal method for aroma studies. This method analyses exactly what the olfactory receptors receive. Also, the method is very simple and gentle Ð one simply draws a few millilitres of vapour above a food into a gas-tight syringe and makes a direct injection into a gas chromatograph (GC). Schaefer (1981) has illustrated the primary limitation of static headspace methodology Ð inadequate sensitivity (see Table 9.1). Since direct headspace injections into a GC are generally limited to 10 mL or less, one can see that only volatiles present at concentrations exceeding 10−7 g/L (headspace) will be detected by gas chromatography (ﬂame ionisation detection), and only those at concentrations exceeding 10−5 g/L will be adequate for mass spectrometry (MS). Since the concentration of volatiles above a food product generally ranges from about 10−4 to 10−10 g/L (or less) (Weurman, 1974), only the most abundant volatiles will be detected by direct headspace sampling. Trace component analysis will require some method of headspace concentration that permits sampling of large volumes of headspace (100Ð1000 L), thereby compensating for low headspace concentrations. Sensitivity of the method may be enhanced to some extent via the use of very sensitive GC detectors and/or headspace enrichment. For example, the use of a GC equipped with a

Table 9.1 Minimum concentrations of a substance in a given volume of air required for gas chromatographic (GC-ﬂame ionisation detector) analysis or identiﬁcation by mass spectrometry (Schaefer, 1981).

Air volumea GC (g/L) MS (g/L) − − − − 1mL 10 5–10 6 10 3–10 4 − − − − 10 mL 10 6–10 7 10 4–10 5 − − − − 100 mL 10 7–10 8 10 5–10 6 − − − − 1L 10 8–10 9 10 6–10 7 − − − − 10 L 10 9–10 10 10 7–10 8 − − − − 100 L 10 10–10 11 10 8–10 9 − − − − 1000 L (1 m3)1011–10 12 10 9–10 10

aSample volume put into a GC or MS. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 233

pulsed-flame photometric detector has been shown to provide detection limits ranging from 0.1to2␮g/L for volatile sulfur compounds in beer (Li et al., 2008). Chemiluminescence GC is commonly used to measure hydrogen sulfide in human breath (morning ‘bad’ breath) in the very low ppb range (Paetznick et al., 2009). Thus, sensitivity is less problematic if a very sensitive detector can be used in analysis. If the analytes of interest are not amenable to detection using very sensitive GC detectors, the sample may be enriched in volatiles by preparing a distillate of a food product and analysing the distillate by headspace methods. Enrichment of the headspace may also be accomplished through the addition of soluble salts to the aqueous food product. The salts tend to drive the organic volatiles from solution into the vapour phase (Jennings and Filsoof, 1977). It is of interest that the enhancing effect is not similar for all volatiles (Roberts and Pollien, 2000). The use of sodium chloride to enrich headspace volatiles may quantitatively distort the headspace profile. A second disadvantage of headspace methods is that it is difficult to do quantitative studies using them. The analytical data one receives are on the amount of an aroma constituent in the headspace. The relationship between concentrations in the headspace versus those in the food can be very complex and must be determined experimentally. The issue of quantification is discussed later in this chapter. The advantages and shortcomings of static headspace sampling dictate its applications (Wampler, 1997). It is often used in quality control situations where only major components need to be measured. Although the components measured might not actually be responsible for the flavour attributes being monitored, if there is a good correlation between flavour quality and the components measured, the goal has been accomplished. For example, Buttery and Teranishi (1963) used headspace analysis of 2-methylpropanal and 2- and 3-methylbutanal as indicators of non-enzymatic browning in potato granules. Sullivan et al. (1974) have used this technique to do additional work on the flavour quality of dehydrated potatoes. It is commonly used to indicate the oxidative quality of edible oils (by monitoring hexanal formation).

9.2.1.2 Methods based on purging and trapping (headspace concentration) Headspace trapping methods are commonly called dynamic headspace or purge-and-trap methods. In these methods, the sample is purged with an inert gas, such as nitrogen or helium, which strips aroma constituents from the sample (Fig. 9.2). The volatiles in the purge gas must then be trapped (somehow removed) from the gas stream. The aroma constituents may be trapped via a cryogenic, Tenax (or alternative polymer), charcoal or other suitable trapping system. This approach favours the isolation of constituents with the highest vapour pressure, as has been discussed. Additional distortion of the aroma profile results from the aroma-trapping technique. A cryogenic trap is the least selective of the traps. It will remove and contain virtually any aroma constituent if properly designed and operated. The primary problem with a cryogenic trap is that it will also trap water Ð the most abundant volatile in nearly all foods. Thus, one obtains an aqueous distillate of the product, which must then be solvent extracted to recover the aroma fraction. Solvent extraction will again alter the true aroma profile. A Tenax trap is very widely used for aroma trapping. Despite its wide usage, it has a low surface area and, therefore, a low adsorption capacity. In addition, Tenax has a low affinity for polar compounds (hence it does not retain much water) and high affinity for non-polar compounds. An aroma compound such as hydrogen sulfide would not be retained at all on this material (Reineccius and Liardon, 1985). P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

234 Food Flavour Technology

Fig. 9.2 Schematic examples of apparatus for the isolation of volatiles via dynamic headspace techniques. The system in (a) uses a cryotrap (vacuum operation); that in (b) uses a Tenax trap (ambient pressure operation). (Reprinted with permission from Guntert et al., 1998; copyright 1998 American Chemical Society.)

Buckholz et al. (1980) demonstrated the biases associated with Tenax traps during a study on peanut aroma. They found a ‘breakthrough’ of peanut aroma (through two traps in series) after only 15 minutes of purging at 40 mL/minute. In an evaluation of the sensory properties of the material collected on the Tenax trap, they found that a representative peanut aroma had been collected by the trap after 4 hours of purging. Shorter or longer purge times did not produce an aroma characteristic of peanuts. In fact, the majority of purging conditions did not yield an aroma isolate characteristic of the sample. The isolate was biased both by the volatiles preferentially going into the purge gas and by the Tenax trapping method. The work of Guntert et al. (1998) found that the Tenax trapping method (ambient pressure) did not yield as true an aroma profile as vacuum distillation (cryotrapping) (apparatus presented in Fig. 9.2). This difference in performance may have been due in part to the operating conditions used for the Tenax system. As Buckholz et al. (1980) noted, operating conditions have a strong influence on the composition of the aroma isolate. Sucan et al. (1998) used response surface methodology to optimise their purge and trap method for a study of dry dog food aroma. This approach, once optimised, yielded a very good quality aroma isolate. Activated carbon traps have a strong affinity and large capacity for most aroma constituents. As little as 1Ð10 mg of carbon will trap the volatiles from 10Ð100 L of purge gas (Schaefer, 1981). The primary concern with charcoal traps is that they may not give up their aromatic components without artefact production. It has been suggested that this problem can be minimised through the use of a good-quality coconut charcoal. Despite the concerns noted about this approach to aroma analysis, it is commonly used in the field today. Several manufacturers offer completely automated systems that simplify the task and add considerable precision to the data. Wampler (1997) has provided a review of this technique. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 235

9.2.1.3 Distillation methods Distillation can be deﬁned broadly to include high-vacuum molecular distillation (the vacuum headspace method apparatus shown in Fig. 9.2 would be more appropriately termed high- vacuum distillation), steam distillation, or simple heating of the food and sweeping of the ‘distilled’ aroma constituents into a GC. High-vacuum distillation may be applied to pure fats or oils, solvent extracts of fat-containing foods or aqueous-based foods (e.g. fruit). Since fats and oils are essentially anhydrous, additional extractions (or sample manipulations) would not be required to remove any co-distilled water. The use of high-vacuum distillation for the isolation of volatiles from solvent extracts has frequently been used to provide good-quality extracts for aroma extraction dilution assays (discussed later). This distillation process is likely to require an additional solvent extraction step since diethyl ether is commonly used as the extracting solvent and this will extract some water as well. The high-vacuum distillation of fresh food products uses product moisture to co-distil volatiles. Water is always the major part of the distillate and a secondary extraction is mandatory. Thus, an extraction is commonly a part of this type of aroma isolation process. The primary sources of aroma proﬁle distortion come from the distillation process and subsequent solvent extraction. Steam distillation may be accomplished in several ways. The product may simply be put in a rotary evaporator (if liquid, or initially slurried in water if solid) and a distillate collected. This distillate would be solvent extracted to yield an aroma isolate suitable for GC analysis. The most common steam distillation method employs simultaneous distillation/solvent extraction (LikensÐNickerson). This is one of the oldest methods for obtaining aroma isolates. Chaintreau (2001) has provided a very good review of this method and its evolution. Figure 9.3 shows an atmospheric pressure system (a) and a vacuum system (b).

Sealable water bath

Fig. 9.3 Simultaneous steam distillation/solvent extraction apparatus: (a) atmospheric pressure; (b) vacuum operation (Chaintreau, 2001; John Wiley & Sons Ltd, reproduced with permission). P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

236 Food Flavour Technology

The primary difference is that the vacuum system has to have joints that are airtight and all parts of the apparatus have to be under tight temperature control. In either approach, the aroma profile ultimately obtained is influenced by volatility of the aroma compounds (initial isolation), solubility during solvent extraction of the distillate and, finally, volatility again during the concentration of the solvent extract. The aroma isolate prepared by simultaneous distillation/extraction (atmospheric or reduced-pressure operation) contains nearly all the volatiles in a food, but their proportions may only poorly represent the true profile in a food (Fig. 9.1). This method is efficient at recovering medium- to high-boiling point compounds and a liquid isolate is obtained. This liquid isolate is quite concentrated, which facilitates mass spectrometric work or repeated injections for further studies. Distillation, as defined here, also includes direct thermal analysis techniques. These techniques involve the heating of a food sample in an in-line desorber (i.e. in the carrier gas flow of the GC). Generally, aroma compounds are thermally desorbed from the food and then cryofocused on the GC column to enhance chromatographic resolution. This technique has been used for a number of years for the analysis of lipids and was later modified to include aqueous samples (Dupuy et al., 1971; Legendre et al., 1979). Aqueous samples were accommodated by including a water trap after the desorption cell. This general approach has been incorporated into the short-path thermal desorption apparatus discussed by Hartman et al. (1993) and Grimm et al. (1997). In this apparatus, shown schematically in Fig. 9.4, a sample of food is placed in the desorption tube and quickly heated. The volatiles are distilled into the gas flow, which carries them into the cooled injection system, where they are cryofocused prior to injection into the analytical column.

Fig. 9.4 Short-path thermal desorption apparatus (Grimm et al., 1997). P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 237

The issue of water in the sample often limits sample size even when it is a minor component of the food. Since most of these methods require cryofocusing prior to gas chromatography, small amounts of water will tend to freeze in the cryotrap (often the GC column), blocking the carrier gas ﬂow. Thus, the sample size (and therefore, sensitivity) is often limited by the moisture content of the sample (the method is applied to samples Ͻ5% moisture; Rothaupt, 1998). The primary bias inherent in distillation approaches is again relative volatility of the aroma constituents. Additional concerns involve the technique used to remove water from the sample (if the sample is aqueous-based) and the potential for artefact production due to heating of the sample. There is a substantial body of information demonstrating aroma formation (in this case, artefact formation) due to heating. Some reactions proceed rapidly at temperatures as low as 60◦C (see also Chapter 3); therefore, the aroma proﬁle can be greatly altered via the formation of artefacts due to heating of the sample during isolation.

9.2.2 Aroma isolation methods using solvent extraction One of the simplest and most efficient approaches for aroma isolation is direct solvent extraction. The major limitation of this method is that it is most useful on foods that do not contain any lipids. If the food contains lipids, they will also be extracted along with the aroma constituents, and the lipids must be separated from the aroma-containing solvent extract prior to further analysis. Aroma constituents can be separated from lipid-containing solvent extracts using techniques such as molecular distillation, steam distillation, purge- and-trap or dialysis Ð all of which further complicate the isolation process and introduce additional biases in isolation. A second consideration in solvent extraction is solvent purity. Solvents must be of the highest quality, which often necessitates in-house distillation prior to use. One must be mind- ful that various qualities of solvents can be purchased and GC grade is highly recommended (not HPLC or other quality). Furthermore, a reagent (solvent) blank must always be run to monitor solvent artefacts irrespective of the quality of the solvent. Solvent extraction can be as simple as putting the food sample (e.g. apple juice) into a separatory funnel, adding a solvent (e.g. dichloromethane) and shaking. The dichloromethane is collected from the separatory funnel, dried with an anhydrous salt and then concentrated for GC analysis. Alternatively, the process may be much more costly and complicated, involving, for example, a pressure chamber and supercritical CO2 (Jennings and Filsoof, 1977). Supercritical CO2 has the advantage that it has a very low boiling point (and so is efficiently separated from extracted volatiles), it leaves no ‘residue’ to interfere with any subsequent sensory analysis, it penetrates food matrices and its solvent properties can be altered through temperature and pressure changes or the use of chemical modifiers (e.g. methanol). Negative aspects of this solvent include its high cost due to pressure requirements, small sample sizes (most commercial extractors) and its highly non-polar nature (without modifiers). The use of modifiers such as methanol negates some of the advantages noted earlier. Morello (1994) has provided a good example of its use and thoughtful discussion of the technique. The biases imposed on the aroma profile by solvent extraction relate to the relative solubility of various aroma constituents in the organic/aqueous phases. A graphic comparison of the recovery of aroma compounds from a model flavour system (in water) using pentane versus dichloromethane as solvent is presented in Fig. 9.5. It is obvious that neither solvent gave 100% recovery of all aroma constituents and that dichloromethane extraction gave quite a different aroma isolate from that using pentane. Cobb and Bursey (1978) have made P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

238 Food Flavour Technology

Fig. 9.5 Batch solvent extraction of a model system showing the inﬂuence of solvent ((a) pentane and (b) dichloromethane) on recovery of volatiles. 1, ethanol; 2, propanol; 3, butanol; 4, octane; 5, decane; 6, ethyl propanoate; 7, ethyl butanoate; 8, ethyl pentanoate; 9, 2-heptanone; 10, acetophenone; 11, benzyl acetate; 12, methyl salicylate; 13, L-carvone; 14, ␤-ionone; 15, methyl anthranilate; 16, ethylmethylphenyl glycidate; 17, isoeugenol. (Reprinted with permission from Leahy and Reineccius, 1984; copyright 1984 American Chemical Society.)

a similar comparison of solvent effect on recovery of a model aroma system from 12% (v/v) ethanol in water (Table 9.2). Recoveries of aroma constituents were low and variable, depending on the solvent chosen and aromatic component being extracted. While it is obvious that even a simple solvent extraction introduces substantial bias into an aroma proﬁle, combining solvent extraction with another technique (e.g. to separate aroma components from extracts containing lipids) adds more bias, for example, applying a distillation technique to a solvent extract that contains lipids selects for the most volatile components (now from an oil phase).

9.2.3 Solid-phase micro-extraction Solid-phase micro-extraction (SPME) is a relatively new technique for the isolation of food aromas. Pawliszyn’s group (1997) was the ﬁrst to develop this method and apply it in

Table 9.2 Recovery of model compounds from an alcohol–water (12% v/v) system (Cobb and Bursey, 1978).

Recovery (%)a Compounds extracted Freon II Dichloromethane Ether Isopentane Ethyl butanoate 66 43 – 16 2-Methyl-1-propanol 34 55 22 32 3-Methyl-1-butanol 63 66 50 48 1-Hexanol 85 67 23 38 Benzaldehyde 83 54 18 20 Acetophenone 53 41 34 20 Benzyl formate 75 56 21 25 2-Phenethyl butanoate 46 48 25 17 Methyl anthranilate 62 59 57 27

aBatch separating funnel extraction; 757 mL of model system extracted 6 × 50 mL solvent. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 239

environmental analysis. Since then, it has become the most widely used technique for the analysis of volatiles in foods. Nongonierma et al. (2006) have published an in-depth critical review of this technique that this author feels is required reading before using this technique. For SPME, an inert fibre is coated with an adsorbent (of which there are several choices). The adsorbent-coated fibre is placed in the sample headspace or immersed in the sample itself and allowed to adsorb volatiles. The ‘loaded’ fibre is then thermally desorbed into a GC carrier gas flow and the released volatiles are analysed. A schematic of the device used to adsorb volatiles is presented in Fig. 9.6 and the overall process is shown in Fig. 9.7. The coated fibre is a modified syringe in which the needle is retractable and is coated with adsorbent. The feature of retractability affords protection to the fibre against physical damage and contamination. SPME is an equilibrium technique and, therefore, the volatiles profile one obtains is strongly dependent on sample composition, sampling parameters and the absorption properties (selectivity and capacity) of the fibre coating (Marsili, 2002b; Roberts et al., 2000). A particular weakness inherent to the method is the very low quantity of absorbent available for volatile absorption. A typical SPME fibre, e.g. a 100 ␮m PDMS fibre, has only about 0.5 ␮L of phase (David and Sandra, 2003). This results in the method, giving poor recoveries of analytes having an oil/water partition coefficient (Kow) less than 10 000 (this includes a large P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

240 Food Flavour Technology

Fig. 9.7 Schematic showing the steps involved in the use of an SPME device: (a) extraction procedure; (b) desorption procedure. (Courtesy of Supelco Inc.)

number of aroma compounds). Also, a low-phase ratio gives variable extraction for non-polar volatiles due to competitive reactions between the SPME fibre, the aqueous phase, the glass walls of the sample vessel, and the stir bar used to stir samples. These difficulties can be minimised by increasing the phase ratio. A stir bar method has been developed to address this issue (David and Sandra, 2003). In this method, an adsorbent phase is coated on an inert stir bar (glass). The coating acts as an extracting solvent as opposed to an adsorbent; thus, phase volume as opposed to surface area is important. A stir bar may have a 25Ð250 ␮m phase coating over a larger surface area (i.e. a stirring bar) to yield much better extraction efficiencies than SPME. The theoretical extraction efficiency approaches 100% for analytes having kow of more than 500. In most applications a stir bar is immersed in the product to be analysed, allowed to come to equilibrium with the liquid being analysed (30Ð240 minutes), rinsed with water, dried (wiped dry) and then either thermally desorbed into a GC or solvent extracted. It has been found that food samples containing fat levels below 2Ð3% or alcohol levels below 10% can be extracted with this technique. The availability of larger extraction phase volumes results in better quantitative data as well as greatly improved sensitivities. The speed and simplicity of this method and refinements over SPME make it quite attractive. Similarly to all the other methods described thus far, SPME and stir bar methods afford a certain view of the volatile composition of the food. This view is determined by factors common to headspace or extraction techniques as well as the unique effects contributed by the adsorption process. To use the method effectively, one has to be very familiar with the factors that influence volatile recovery. These factors have been discussed in detail in the review article cited above. If the method provides an isolate that has the components one wishes to measure and it is adequately reproducible, the method is quite attractive. There are no solvents for contamination, it is simple (has been automated) and it is sensitive and rapid. There are several other approaches for the isolation of food aromas; those that have been mentioned are the primary methods in use today. As noted earlier, the reader is encouraged to go to more in-depth reviews for detail. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 241

9.2.4 General considerations in preparing aroma isolates While the isolation methods themselves are selective in their recovery of aroma constituents from foods, and therefore give an aroma proﬁle unrepresentative of the food, the aroma proﬁle may be further biased by other means. Some of these considerations are outlined below.

9.2.4.1 Sample preparation Many food products contain active enzyme systems, for example in unprocessed plant or animal tissue (see also Chapter 5). These enzyme systems may become active if the food is chopped, ground or macerated to facilitate aroma isolation. If the enzyme systems are not inactivated (e.g. by heating, or addition of alcohol or sodium chloride) prior to starting analysis, the aroma proﬁle may change greatly. We may want this change to occur if this is how a food is normally consumed, or we may want to inhibit enzymatic changes. In either event, enzymatic action may introduce changes in the aroma proﬁle and we must take this effect into consideration.

9.2.4.2 Contamination by artefacts The analyst is typically working in the ppm (milligram/kilogram) or lower concentration ranges. There are numerous ways in which volatile constituents may contaminate the aroma isolate at such low levels. One must be extremely careful of water quality if the sample is mixed with any water (or steam). Organic solvents are seldom sufﬁciently pure to be used in aroma isolation without additional cleanup (typically by distillation). Any polymer-based materials (containers or tubing) are common sources of contamination. Antifoam additives may contribute as many components to an aroma isolate as does the food itself. Stopcock or vacuum greases are known sources of contamination. Bottle closures must be Teﬂon-coated rather than rubber to prevent the closure from both absorbing some aroma components and contributing others. Blanks must always be run to determine the magnitude of contamination contributed by the system. Every effort must be made to minimise contamination from all sources.

9.2.4.3 Thermally induced artefacts Foods are very good reaction systems (see also Chapter 3). At least 3000 volatile compounds have been identiﬁed to date arising from the heating of foods. Many techniques for aroma isolation involve heating of the food sample. This is done in order to more effectively remove volatiles from the food matrix. Extreme care must be exercised in heating food samples to ensure that the aroma isolate collected truly represents that of the food and is not contaminated by thermally induced artefacts. We have a rule that we do not heat raw (not thermally processed) foods above 60◦C in aroma isolation. We may go above 60◦Cfor thermally processed foods, but we always choose to use minimal heat exposure.

9.2.5 Aroma isolation summary Since every method preferentially selects those aroma constituents that meet certain physical or chemical criteria (e.g. solubility, volatility or afﬁnity), one must ‘make do’ and compensate P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

242 Food Flavour Technology

for having a very biased analytical view of the aroma constituents in a food product. That this view is biased does not mean that it is useless or even of lesser value than a truly accurate picture. We need to choose our methodology wisely so that we measure the aroma components we need to monitor to solve our problem, that is, so that they are contained in the aroma ‘view’ we consciously select. Furthermore, one must recognise that the most commonly used approaches in the literature may not be the best or even suitable for a given task. A particular task requires a unique method. The frequency of a method appearing in the literature is more often linked to the size of the research group than to any other factor. A particular research group may be large and doing similar work, and thus, their particular methodology appears frequently. Also, individuals have certain biases Ð no two researchers will approach the same problem in the same manner. With that said, this author’s view (including biases) of method selection is discussed in the following section.

9.3 SELECTION OF AROMA ISOLATION METHOD

One cannot choose a method for aroma isolation without ﬁrst deﬁning the objectives of the study. For example, one may wish

(1) to obtain a ‘complete’ aroma isolate to accurately identify and quantify every aroma constituent in a food; (2) to identify only key components of an aroma proﬁle, that is, those components that are responsible for the characteristic sensory properties; (3) to identify an off-note in a food product; (4) to monitor aroma changes with time; or (5) to predict sensory attributes.

Each of these tasks imposes different requirements on the methodology.

9.3.1 ‘Complete’ aroma profile Obtaining a complete aroma profile is one of the most difficult tasks to accomplish. It is given that no individual isolation technique will yield an accurate analytical profile. Thus, one must use several isolation techniques in combination. A good combination would be a static headspace method to obtain a profile of the most volatile and most abundant volatiles. One can follow this with a purge-and-trap method to obtain data on the less volatile and less abundant constituents (no solvent to ‘cover’ early-eluting components of interest). The task would finish with a solvent extraction (if there are no lipids in the food) or simultaneous distillation/extraction method to obtain a profile of the least volatile aroma components. This combination of techniques should yield a reasonably complete view of the aroma profile (for example, see Qian, 2000). Each of the profiles will be biased (as discussed), so that quantitative data will have to be obtained using the food system plus some quantification approach. This might include the use of multiple internal standards (Siek and Lindsay, 1968), a method of standard addition (adding known quantities of each individual pure component and determining increases in GC peak areas versus amount added and relating this back to original peak area; Qian, 2000) or, ideally, use of isotopically labelled pure standard compounds (adding a known quantity of P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 243

stable isotope-labelled compound and monitoring aroma isolate via selected ion-monitoring MS to obtain a ratio between labelled and unlabelled compounds of interest; Milo and Blank, 1998). Traditionally, ﬂavour chemists have been less than rigorous in this respect. It has long been a habit to simply use GC percentage area in tabulations of aroma compounds and their quantities found in foods. The data obtained and reported in this manner are typically grossly in error due to the many biases in aroma isolation and GC analysis. Occasionally, a researcher will add an internal standard and report quantitative data in terms of the internal standard. This approach offers little or no improvement over GC peak area. Obtaining accurate quantitative data is a very formidable task that many researchers choose to short cut.

9.3.2 Key components contributing to sensory properties Since the task of identifying these, in general, is discussed later in the chapter, the immediate discussion focuses only on the methods one might use in preparing an aroma isolate for this purpose. While much of the early research done for this purpose used only one isolation method (most commonly high-vacuum distillation/sublimation), it was recognised that this method did not provide a satisfactory isolate of the very volatile aroma compounds. These volatiles were lost during sublimation, extraction and/or concentration of the aroma isolate, or with the solvent front during chromatography. Thus, the methods employed underestimated the importance of the more volatile aroma compounds. The work done today generally combines two or more isolation methods to provide a more complete view of the food aroma. Typically, a rigorous technique such as high-vacuum distillation (or solvent-assisted ﬂavour evaporation; Engel et al., 1999) is combined with a static headspace technique. Qian (2000), in fact, chose to use a solvent extraction method to determine free fatty acids, a static headspace method to see the more volatile aroma compounds, a purge-and-trap method for the intermediate volatility aroma compounds and a high-vacuum distillation/sublimation method for the least volatile compounds (in Parmesan cheese).

9.3.3 Off-notes in a food product The requirements imposed on isolation methodology for identification of off-notes are much less stringent than those imposed by the first two tasks. Virtually any isolation method can be used that yields an aroma isolate containing the off-note. Selection of the isolation method can initially be guided by experience. For example, off-notes that have a less characteristic aroma but are more generically solvent-like (e.g. contamination by food packaging or printing inks) can initially be approached using an absorption or even perhaps a static headspace method. Off-notes that are very characteristic and heavier in sensory character (e.g. earthy, musty, cooked or burnt) may require more rigorous methods such as steam distillation or high- vacuum distillation. Irrespective of method, it is essential that the aroma isolate of a control (no off-note) and the off-flavoured product be produced and then sensorially evaluated to ensure that the chemical constituents responsible for the off-notes have actually been isolated from the food. Method validation can be done in different ways. For example, if the isolate is a liquid, a blotter can be dipped in it, the solvent evaporated and the blotter smelled for the off-note. Alternatively, the isolate can be applied to a GC and then the effluent smelt for the off-note. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

244 Food Flavour Technology

The analytical approach to finding the odourants causing an off-flavour then involves analysing the isolates from the good and bad samples by gas chromatography while smelling the GC column effluent to locate the chemical constituent (GC peak) corresponding to the characteristic off-note. One must recognise that any food isolate may contain odourants that are ‘unpleasant’. However, the presence of unpleasant odourants in the GC effluent may simply be due to their concentration in the effluent. They may not be a source of off-flavour in the food itself. Thus, it is essential that odourants (GC peaks) be selected by sniffing that are characteristic of the off-flavour in the food. Ideally, the aroma isolation procedure should yield an isolate sufficiently concentrated for GC-MS identification of the tainting compound. If there is no GC peak in the chromatogram where the offending aroma is smelled (or too little compound is present for a good mass spectrum), then the aroma isolation procedure needs to be changed so that larger quantities of the tainting chemical are obtained, or identification may be attempted on gas chromatographic retention data (retention indices) and sensory description.

9.3.4 Monitoring aroma changes in foods There are many situations in which the flavour chemist wishes to monitor changes in food aroma over time. For example, one may wish to analytically monitor flavour losses from a food product during storage (e.g. from coffee), the formation of desirable flavours (e.g. in wine or cheese ageing) or the appearance of off-flavours (e.g. through lipid oxidation). For these types of problems, one has to consider at least four factors:

(1) Does the aroma isolation method provide data on the aroma compounds of interest? (2) Is the method sufﬁciently robust to be stable over time? (3) Is there adequate precision in the method to see the anticipated variation? (4) Is the method rapid enough to be used in the study?

The importance of each factor depends on the task at hand. For example, a storage study with 10Ð20 samples to be analysed each day (or even each week) will preclude any method that is very time-consuming (distillation methods may be problematic). One would like to use an automated headspace method (static or dynamic). Precision is typically obtained through the use of automated methods and/or analytical standards. Automated headspace methods may have a coefficient of variation (CV = standard deviation/mean) of only 2Ð3% (compound dependent), while a distillation method may have a CV ranging from 10 to 50%. Many of the analytical methods in flavour research suffer from poor reproducibility. The range of problems and priorities one may encounter in this type of study makes any further discussion difficult.

9.3.5 Using aroma compound profiles to predict sensory response This task generally does not require a complete or an accurate volatile profile. The goal is to produce an aroma profile that is reproducible (precision aids prediction reliability) and contains some components related to the sensory attributes one desires to predict (see also Chapter 11). The aroma constituents used in prediction are generally not causally but only statistically correlated. Most often this task is used in a production or quality control setting, so time, ease and reliability are additional factors influencing method choice. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 245

Static headspace and direct analysis isolation techniques are well suited for this task. Both approaches are routinely used in the quality control of fats and oils for the analysis of hexanal as an indicator of rancidity. In some cases, static headspace may be replaced by an adsorption method (purge and trap or SPME) to yield more information for obtaining valid statistical correlations. There is a great deal of literature available demonstrating correlations between GC aroma proﬁles and some sensory attributes (e.g. coffee attributes) or other information of interest (e.g. geographical origin of olive oils or wines). Absorption methods have been automated to greatly reduce operator time required and, to a limited extent, operator expertise.

9.3.6 Summary comments on isolation methods While aroma isolation from foods is generally based on either volatility or solubility of the aroma compounds, numerous methods have been developed to apply these principles to this task. Some methods depend solely on volatility (e.g. headspace, direct analysis and molecular distillation) or solubility (solvent extraction and absorption methods). Other methods depend on combinations of volatility and solubility for isolation (e.g. simultaneous steam distillation/solvent extraction). One must be conscious that none of these approaches yield an aroma isolate for further analysis that either qualitatively or quantitatively truly represents the aroma of the food. Each method provides a unique view of the aroma profile that may be more or less suited to addressing the problem faced by the flavour chemist. In method selection, the chemist must define the problem and consider which isolation method is best suited to his or her particular task. Most problems in the flavour area can be addressed with existing technology, but the chemist must choose wisely among the technologies to yield a solution.

9.4 AROMA ISOLATE FRACTIONATION PRIORTOANALYSIS

It may prove advantageous to pre-treat the aroma isolate before GC analysis (Parliment, 1997). Even though high-resolution GC columns and specialised detectors are available to the analytical flavour chemist, the complex nature of the flavour isolates often makes complete resolution or adequate concentration impossible. Some pre-fractionation of flavour isolates prior to GC simplifies analysis. The fractionation of flavour isolates into classes of chemicals with similar chemical properties also permits the use of specialised gas chromatographic columns and detectors. For example, columns that are particularly adapted to the separation of basic compounds may be selected and nitrogen-specific detectors utilised. The following section discusses some of the more common methods applied to the fractionation of flavour concentrates.

9.4.1 Fractionation of concentrates prior to analysis 9.4.1.1 Acid/base separations This method utilises the differential solubility of ionised and non-ionised species in aqueous and organic solvents. Ionisation of ionisable compounds is generally accomplished via pH control of the aqueous solvent. A typical acid/base separation scheme is presented in Fig. 9.8. Whereas this scheme assumes one is starting with a ﬂavour isolate in an organic solvent, the P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

246 Food Flavour Technology

Fig. 9.8 Fractionation scheme for the of volatiles into acidic, neutral and basic fractions (Reineccius and Anandaraman, 1984).

same technique is easily applied to acid/base fractionation of aqueous distillates. The distillate is initially adjusted in pH prior to solvent extraction and then a similar pH adjustment scheme is followed. Subfractionation of acidic compounds can be accomplished by extracting the organic phase with 5% sodium carbonate initially to extract the strong acids (e.g. carboxylic acids) and then 5% NaOH to extract weakly acidic compounds (e.g. phenols). In this case, there would be two aqueous phases on pH adjustment, one of which would yield an isolate of weak acids and the other of strong acids. Acid/base separations are relatively simple and rapid, and do not require sophisticated equipment or expensive reagents. This fractionation simplifies the following GC analysis by reducing the number of components to be separated, permits further concentration than might otherwise be possible and allows a more specific column choice and tailoring of GC operating conditions. The acid/base/neutral fractions can be subjected to sensory analysis, which provides information on what type of flavour is responsible for the flavour notes of interest. However, this technique has two major drawbacks. The separation between classes is often incomplete, and poor recoveries are common. Efficient extraction requires large volumes of solvents and multiple extractions. Multiple extractions with large volumes of P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 247

solvent can be tedious and result in dilute solutions of ﬂavour compounds in the organic phase. This would mean greater losses during the concentration step. Artefact formation is possible under either alkaline or acidic conditions: esters are susceptible to hydrolysis under acidic and alkaline conditions to yield the corresponding alcohols and acids; epoxides can yield diols in the presence of acid and water.

9.4.1.2 High-pressure liquid chromatography The fractionation of flavour concentrates using high-pressure liquid chromatography (HPLC) has been used to a limited extent. HPLC is an attractive method for fractionation of flavour concentrates since it utilises a different set of physical properties for component separation than does GC. The potential exists for flavour fractionation on molecular sieves, adsorption, and reversed-phase or normal phase chromatography. A logical sequence would be to use adsorption chromatography first since this technique has the greatest column capacity and is capable of handling the widest range of types of compounds (Teitelbaum, 1977). This would accomplish a fractionation based on adsorption affinity. These individual fractions could then be again fractionated on a normal or reversed-phase HPLC column. At this point, substantial separation of the rather complex flavour concentrate has been accomplished and would definitely simplify the subsequent GC analysis. The general use of chemically bonded reversed-phase columns for HPLC makes this method attractive (Parliment, 1981). Reversed-phase columns are very inert and will not bleed into the eluting solvent. Therefore, they present minimal concern for contribution to artefact formation. Evaporation of the HPLC solvent may not be necessary for subsequent gas chromatography. Jennings (1979a,b) has demonstrated the applicability of splitless GC for the analysis of dilute solutions. Also, today, large-volume GC injection systems are available that further facilitate this approach. This means that solvent extraction and concentration of the collected fractions from the HPLC might not be necessary, eliminating concerns about solvent extraction efficiency, contamination and losses during concentration of the extracting solvent. There are a few disadvantages to the fractionation of flavour concentrates via HPLC. One concern is for contamination of the flavour concentrates due to HPLC solvents and repeated transfers. If the HPLC eluent must be extracted and concentrated for gas chromatography, then one has also to be concerned with extraction efficiency, solvent impurities and losses during solvent evaporation. The method requires expensive equipment, columns and high- purity solvents.

9.4.1.3 Silicic acid

Silicic acid (H2SiO3) and the dehydrated derivative silica gel (SiO2) have been widely used in separating organic compounds via conventional column chromatography. Silica gel, depending on water content, will adsorb organic compounds with different affinities. Elution of these adsorbed compounds with an appropriate organic solvent or elution gradient results in a chromatographic separation. If suitable choices of silica gel activity (amount of water in the gel), solvents and other chromatographic parameters (adsorbent/adsorbate ratio, column height, column diameter, elution rate, and so on) are used, efficient functional group separations can be accomplished. The separations are primarily due to differences in polarity of compounds in the mixture. Utilising this chromatographic technique, flavour concentrates can be fractionated into a number of less complex mixtures. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

248 Food Flavour Technology

The primary advantages of silica gel fractionations are the efficiency and simplicity of the method. The method requires only simple chromatographic columns and inexpensive silica gel and solvents. The separation achieved via silica gel chromatography can be quite efficient, depending on the chromatographic conditions chosen. Solvent choice, activity of silica gel, column dimensions and elution rates are the primary factors determining fractionation achieved. Silica gel fractionation also has some inherent disadvantages. A good understanding of these disadvantages is essential for the proper use of this technique. Silica gel may not release the adsorbed compounds quantitatively. Recoveries from a silica gel column can vary considerably (65Ð95%), depending on the nature of the compounds. Recovery of polar compounds is generally less than that of the less polar ones (e.g. hydrocarbons). Recoveries of the same compound with the same eluting solvent can vary if the activity of the silica gel is not the same. This can lead to inaccurate quantitative data. The errors are cumulative if repeated silica gel fractionation is attempted prior to analysis. Compounds containing certain labile or reactive functional groups are known to undergo chemical transformation in contact with highly activated adsorbents. For example, active silica gel can promote isomerisation reactions. Epoxides formed from the oxidation of tri- and tetra-substituted double bonds are very susceptible to isomerisation by active silica even at room temperature. Double-bond migration is also known to occur in terpene hydrocarbons in the presence of silica gel. Artefact formation can be minimised or eliminated by carefully controlling the activity of the silica gel and keeping the contact time to a minimum. Faster elution rates can be used only at the expense of some fractionation efficiency. In the published literature, one may note many instances where the authors did not mention either the elution rate or the activity of the silica gel. Standard methods are available for the determination of activity, and procedures are available to prepare silica gels of lower activities from fully active gel (Hernandez et al., 1961). Another means of reducing contact time is to use shorter columns when resolution permits. An unnecessarily large adsorbent/adsorbate ratio should not be used. If such unfavourable ratios are employed, isomerisation reactions will be facilitated due to increased contact with active centres or adsorption sites.

9.4.1.4 Preparative gas chromatography Preparative GC as a fractionation technique offers the greatest separating power of the methods discussed. To provide sufficient column capacity (load), either 3 or 6 mm (outer diameter) packed columns are typically used for resolution of components. The liquid phase of the preparative column is generally chosen to be of opposite polarity to the column used for the analytical study. Column effluent is either split via an effluent splitter (10:1 exit/detector) or passed through a non-destructive detector. This effluent is usually trapped in a cold trap for rechromatography on an analytical column. The primary advantages to preparative GC fractionation are the extreme efficiency of separation and the fact that fractions collected are solvent-free. If a large number of chromatographic cuts are taken from a run, each fraction collected can contain only a few compounds for further analysis. If the preparative GC work is done on a column of opposite polarity to the analytical column, separation by the analytical column is simplified. Since the collected fractions are solvent-free, sensory analysis is easily done on individual fractions. This often aids in narrowing the flavour study to a relatively simple fraction or few fractions. The solvent-free aspect also means that trace components in a flavour isolate can be made the P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 249

major components in the fraction collected. One chooses the chromatographic cuts such that trace components are collected separately from the major components. This greatly enhances the separation and identification of trace components of food flavour. Preparative GC may be criticised as a method that subjects the flavour isolate to high temperatures and active surfaces. A survey of early-published literature in flavour fractionation reveals that invariably a packed metal column of over 3 m has been used in the preparative GC work. Large surface areas offered by the solid support and the inner metal surface of the column could prove extremely detrimental to labile compounds or could irreversibly adsorb other compounds. The residence time of high-boiling compounds may also promote artefact formation. Rechromatography and subfractionation increase the concerns. Recent efforts have recognised these potential errors and have used deactivated solid supports, all- glass injection ports, glass columns and glass-lined transfer lines. Preparative GC generally complicates quantification such that it is used only for qualitative studies.

9.5 FLAVOUR ANALYSIS BY GAS CHROMATOGRAPHY

Gas chromatography is the single most widely used technique in flavour studies. It is exceptionally uncommon to see any analytical flavour work that does not include this instrument. This is not unexpected, since compounds most commonly of interest are those that are volatile and therefore may contribute to aroma. Gas chromatography has tremendous separating power, sometimes in excess of 200 000 theoretical plates per column. This attribute is essential for the separation of complex flavour isolates. Detectors available for GC work offer excellent sensitivity, providing picogram detection levels, and yet may be general (flame ionisation) or very specific (flame photometry, electron capture and nitrogen/phosphorus). It is not unexpected, then, that flavour research advanced greatly in the mid-1960s when GC became readily available to the flavour chemist. A tabulation of flavour compounds identified in foods lists only 500 compounds found by 1963. Only 15 years later, this number had increased to over 3000 and, as noted earlier, today over 7000 have been identified. We will not go into a general discussion of GC. There are many excellent books on this topic (Jennings et al., 1997; Handley and Adlard, 2001; Mondello et al., 2001; Niessen, 2001). We will, however, go into some aspects of GC that are particularly relevant to GC studies in flavour research.

9.5.1 High-resolution gas chromatography The primary purpose of gas chromatography in flavour research is to separate aroma mixtures into individual components. Once they are separated, quantification and identification become primary concerns. Owing to the complex nature of most food flavours, resolving power is a critical need in GC. Therefore, capillary column GC (high-resolution gas chromatography, HRGC) is the standard today. The primary disadvantage of using capillary columns in flavour analysis is their limited capacity. Flavour studies may require the trapping of individual components for infrared, NMR, sensory evaluation, or other analyses. Column capacity is typically being enhanced through the use of thick-film coatings as opposed to wide-bore columns (smaller losses in efficiency). Whereas a typical fused-silica column is limited to less than 100 ng of each component, thick-film fused-silica columns will handle 500 ng of each component without overloading. While this is still not sufficient for some needs, the use P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

250 Food Flavour Technology

of automated injection systems and capillary fraction collectors can result in the collection of substantial amounts of material. When HRGC does not offer adequate resolving power for a particular application, resolving power can be enhanced through the use of cold trapping and rechromatography (off-line) or multi-dimensional GC (in-line). The former method employs a fraction collector that can take heart cuts out of a GC run; the operator will then manually re-inject each cut onto a second GC. The second GC run is typically equipped with a different GC column phase to enhance resolving power. The in-line version of this process automatically does the heart cutting and re-injection of the sample into a second GC column (and oven). This can be used to automatically do heart cutting of single-column fractions, or the trapping and rechromatography can be done repeatedly to encompass the entire GC run. This latter technique has various names including comprehensive GC, GC×GC, or multi-dimensional GC. In comprehensive GC, a cryotrap is used to collect the eluent of a first column in short periods (e.g. 3 seconds). After collecting 3 seconds of column effluent, the column effluent is then diverted to a second cryotrap while the just loaded trap is heated rapidly and the collected material is vaporized and then passed into a second column of different polarity. Two limitations of this instrument are that the separation occurring in the first column is compromised to some extent by collecting 3 seconds of effluent, and the second chromatographic run must be done in 3 seconds to be ready for the next trapped fraction. Thus, the separating abilities of this second column are very limited. This technique is finding the greatest application when interfaced with a time-of-flight mass spectrometer (TOF-MS) (discussed in a later section). Wright (1997) and Mondello et al. (2001) have presented a discussion and examples of the application of multi-dimensional chromatography in flavour analysis. In cases where one is accomplishing greater peak resolution than is necessary, the use of fast GC is possible (White, 2009). This technique takes advantage of very small diameter, short columns and high-temperature programme rates (ca. 30◦C/second). GC runs may be in the seconds instead of minutes. Fast GC is used primarily in very simple flavour analyses where quantitative analysis is desired.

9.5.2 Gas chromatography–olfactometry Gas chromatography of flavour extracts provides the opportunity to use a rather unique detector, the human nose. The usual GC detector provides very little information for the flavour chemist Ð simply a line on a paper (or monitor). A trained person can smell the GC effluent and tell the potency (sensory intensity) of a GC peak, character and often even the chemical identity of a peak. Compounds with extremely low sensory thresholds (e.g. sulfur compounds) may be detected by smell but not detected by the GC. The nose may tell us that we have to concentrate the sample further or use a different isolation technique to prepare the sample for GC. A chromatogram obtained via smelling of the GC effluent, and therefore containing sensory information, is called an ‘aromagram’. Research on gas chromatographyÐolfactometry (GC-O) techniques was very prevalent in the 1980s and 1990s (Blank, 1997; Leland et al., 2001) and finds very broad use today.

9.5.2.1 GC-O to obtain aromagrams Simultaneous GC detection and odour proﬁling are accomplished using an efﬂuent splitter, a non-destructive detector, or chromatography without and with a detector (two GC runs). P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 251

When an effluent splitter is used, the effluent is generally split such that a larger portion goes to the nose than to the GC detector (10:1). Whereas the nose is often more sensitive than the GC detector, dilution in air at the time of smelling and human response time necessitate a split in favour of the nose. The thermal conductivity (TC) detector is the most common choice for odour profiling when a non-destructive detector is desired. Whereas a TC detector may seem to be a better choice than effluent splitting since the entire sample is available for GC detection and smelling, a TC detector is not as sensitive as other GC detectors. Flame ionisation detectors (FIDs) have a 104Ð106 times lower detection limit than a TC detector. Therefore, even with a 10:1 split (noseÐdetector), the use of an FID is preferred to a TC non-destructive approach. The final approach, that of making one GC run with effluent smelling (GC column is not connected to the GC detector) followed by a second GC run with GC detection (no smelling) is also done. This offers the maximum sample quantity to both the GC detector and the nose. The disadvantage is the time required for two GC runs and ambiguity of what odour goes with what peak when resolution is difficult. There are numerous weaknesses in using GC-O methodologies (Friedrich and Acree, 2000). It is often criticised as being a subjective method, yielding inconsistent results. However, independent judgements of well-trained subjects can minimise this first concern. The duration of a routine GC run is often more than 30 minutes. Odour fatigue can set in well before the end of the analysis, leading to incorrect odour descriptions. Thus, one must be conscious of the time a subject is asked to perform this task (it may be limited to 20 minutes). Odour characteristics of some flavour compounds tend to vary as a function of concentration. Skatole (3-methylindole) has a characteristic faecal odour at high levels but becomes pleasant, sweet and warm at very low levels. Fortunately, there are not many aroma compounds exhibiting such a large concentration-dependent odour character. Furthermore, attempts to indicate the perceived intensity of a GC peak can be in error owing to masking in mixtures. Finally, in-line condensation of some compounds can result in persistent background odours. The splitter and the transfer lines should be well conditioned and adequately heated to render them odour-free. Despite these potential pitfalls, GC-O is an invaluable tool to the flavour chemist and has found broad application in this field (Leland et al., 2001).

9.5.2.2 GC-O to select key odourants in foods As mentioned earlier, the aroma, i.e. the volatile component, of a food is generally made up of a very complex mixture of volatiles. It is well accepted that not all these volatiles contribute to sensory perception. Many compounds are present in a food at concentrations below the level needed to elicit a sensory response. Research has generally shown that 10Ð30 volatiles are adequate to reproduce the aroma of any food studied thus far. The analytical challenge is to determine which of perhaps 800+ aroma compounds (such as meat aroma) are needed to properly reproduce the sensory character of a food. Several different approaches to this challenge have appeared in the literature. The earliest work in this area is now more than 50 years old (Patton and Josephson, 1957). They proposed estimating the importance of a flavour compound by the ratio of the compound to its threshold concentration. This ratio is known as the odour activity value (OAV) (also as odour value, odour unit, flavour unit or aroma value). This ratio indicates by how much the actual concentration of a compound exceeds its sensory threshold. They suggested that compounds present above their sensory threshold concentrations in a food are significant contributors to its aroma, whereas those occurring below threshold are not. Patton P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

252 Food Flavour Technology

Fig. 9.9 Schematic of the GC-O system used in obtaining Charm data (Acree, 1993).

and Josephson (1957) proposed this method as a guide ‘that may not hold in some instances’. The OAV concept was applied to mixtures by Guadagni et al. (1966), who suggested that if the perceived intensity of odourants in a mixture is additive, the relationship between OAV of single component in the mixture and the OAV of that mixture is

OAVi1 + OAVi2 +···+OAVin = OAVm (9.1)

where i1, ..., in represent compound 1, ..., n in mixture m. Thus, the relative contribution of a compound to a mixture could be described as the ratio of its OAV to the OAV of the mixture. Guadagni et al. (1966) noted that this implied nothing about the odour quality of the final mixture and nothing about the relationship between the stimulus concentration and sensation above threshold. Since the introduction of the OAV concept, GC-O and OAV approaches have been extensively used to screen for ‘significant’ odourants in food. Two major screening procedures for determining the key odourants in food are based on this concept. Grosch developed the aroma extract dilution analysis (AEDA; Ullrich and Grosch, 1987) and a recent variation, the aroma extract concentration analysis (AECA; Kerscher and Grosch, 2000); Charm analysis was developed by Acree et al. (1984) (Fig. 9.9). These two methods are evaluated by GC- O, a dilution (or concentration) series of the original aroma extract from a particular food and attempt to rank the key odourants in order of potency. The highest dilution at which a substance is smelled is defined as its dilution value. The dilution value is proportional to the OAV evaluated in air. Both AEDA and Charm methodologies originally proposed that the larger the dilution value, the greater the potential contribution of that compound to the overall aroma. With time, data interpretation has changed. Researchers now consider AEDA, OAV and Charm methodologies to be screening in nature. These methodologies are used to determine those aroma compounds most likely to make a contribution to the odour of a food, recognising that sensory work (e.g. recombination studies) needs to be done to determine which aroma compounds are truly contributory. Interpretation has changed owing to recognition that the methods violate certain sensory P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 253

Fig. 9.10 GC-O system used by McDaniel for obtaining OSME data (McDaniel et al., 1990).

rules or psychophysical laws (Frijters, 1978; Piggot, 1990; Abbot et al., 1993; Mistry et al., 1997; Audouin et al., 2001). Two other GC-O methods have also found application for this purpose. One is called OSME and the other NIF (nasal impact frequency) or SNIF (surface of nasal impact frequency). OSME was developed by McDaniel et al. (1990). In her method, a panellist evaluates the aromas eluting from a GC column and responds by moving a variable resistor as aroma intensity changes (Fig. 9.10). Thus, one is obtaining intensity and duration measurements of each GC peak. There are no dilutions made of the sample, which facilitates the use of a larger number of judges as opposed to a single judge doing dilutions. This adds further validity to the method. The importance of an odourant to the overall aroma is judged on the basis of relative sensory intensities during snifﬁng. This is a fundamental difference between the dilution methods (AEDA, OAV and Charm) and OSME. The dilution methods assume that compounds present at the greatest multiple of their threshold are most important to aroma. This violates a basic law of sensory science in that there is a power function relationship between concentration and sensory intensity and that relationship is different from one aroma compound to another. Thus, one cannot unequivocally rank compound intensity based on OAV, Charm or AEDA value. This weakness is recognised and these values are now considered as screening techniques as opposed to providing ‘hard’ numbers: compounds with the highest values are candidates for further study to evaluate their true contribution. The NIF (or SNIF) method was developed by Pollien et al. (1997). In this method, eight to ten untrained individuals sniff the GC efﬂuent (one at a time). They simply note when they smell an odour. The aroma isolate used is adjusted in strength such that about 30 odourants are perceivable to the sniffers. This adds an element of selection in that only the more intense aroma compounds will be evaluated. The number of sniffers detecting an odourant is tabulated and plotted. Those odourants (GC peaks) being detected by the greatest number of individuals are considered the most important. The method has its weaknesses. One problem is that, for two compounds, one may be barely over the sensory threshold of all sniffers while P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

254 Food Flavour Technology

another may be a great distance above its sensory threshold for all sniffers, and yet both these compounds would be viewed as being equal by this methodology. As in searching for the ideal method of aroma isolation, there is no perfect method for selecting key odourants in foods. Each method has weaknesses. The result is that we ﬁnd more sensory work being done to evaluate the analytical data. Very often, researchers are using recombination studies followed by sensory analysis to determine what is contributing to the aroma of a food and what is not. If we are to attempt any re-creation of a ﬂavour, we must have a list of odourants to study and have some reasonable basis for their selection and ranking. The GC-O methods that have been developed have served this purpose with varying levels of success.

9.5.3 Specific gas chromatographic detectors The nature of aroma compounds being studied (elemental composition) and the type of analysis desired may govern the choice of GC detector. For example, an alkali flame ionisation detector (AFID) may be employed to facilitate detection of nitrogen-containing compounds and a pulsed-flame photometric detector (or chemiluminescence detector) to analyse sulfur- containing compounds. If a non-destructive detector is desired, a micro-cross-section thermal conductivity detector is one of the choices. FIDs (a non-specific detector) are the most commonly used detectors in flavour analysis because of their high sensitivity, long life and relatively low cost. However, specific or selective detectors are often used to good advantage in flavour analysis. Selective detectors have several advantages. For example, the relative sensitivity of a specific detector can be many-fold higher with respect to a certain element as compared with an FID. An alkali FID can have 35 000Ð75 000-fold greater response for nitrogen and phosphorus when compared with an equal amount (weight basis) of carbon. This illustrates how a detector can aid analysis of a trace pyrazine or any nitrogen-containing compound in aroma analysis. Selective detection can also aid in identification of compounds. At times, mass spectral data cannot be interpreted with adequate certainty to make identification in GC-MS analysis of flavours. Specific detectors can provide information that facilitates interpretation of such spectra and subsequent identification. Employing specific detectors in aroma analysis cannot be to the analyst’s disadvantage. However, it should be mentioned that these detectors require precise calibration for quantitative analysis. One should also keep in mind that some specific detectors only enhance the response of selected elements. They may still give limited response to other elements. Therefore, the results could be misleading.

9.6 FLAVOUR ANALYSIS BY HPLC

Application of HPLC in aroma analysis has been quite limited. Major reasons for this are lack of sensitivity over a broad range of compounds and resolution. Although analysis of non-volatile derivatives of aroma compounds and fractionation of aroma isolates have been achieved via this approach, a complete flavour analysis has not been done using HPLC. Gas chromatography has been used extensively in flavour analysis, and advancements in column technology and detectors have made extremely high resolution and sensitivity possible. This, to some extent, has overshadowed the progress in the development of techniques for flavour analysis using HPLC. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 255

With that said, HPLCs (and UPLC Ð ultra performance liquid chromatography) have found application for some specific volatiles and are the method of analysis for non-volatiles (i.e. taste substances). For example, HPLC has been used to measure maltol, theobromine, ethyl maltol, catechin, vanillic acid, caffeine, vanillin, epicatechin and ethyl vanillin in chocolate (Risner and Kiser, 2008). Engel et al. (2002) used HPLC to quantify the glucosinolates in cauliflower, which were found to correlate with bitterness intensity. Hofmann’s group has published extensively on the use of HPLC for the separation and quantification of taste substances in foods (e.g. Hufnagel and Hofmann, 2008; Schmiech et al., 2008; Toelstede et al., 2009). With greater emphasis being placed on the importance of taste substances to flavour perception, HPLC methods are becoming more common in flavour research.

9.7 IDENTIFICATION OF VOLATILE FLAVOURS

In most flavour studies, it is desirable and most often necessary to identify the flavour components of interest (note the qualification ‘of interest’). The problem may be to determine the volatiles that are important to coffee flavour or those that produce a taint in green beans. A decision should be made prior to starting identifications as to what should be identified. Despite rather sensitive and sophisticated instrumentation to aid in identification, the task of identifying every volatile component in a flavour isolate should be restricted only to the young and/or foolish. Decisions about what needs to be identified should be based on sensory evaluation, i.e. GC-O data. One can then often narrow the task to the identification of a limited number of components. Although there are a multitude of methods available to the chemist for the identification of unknowns, only a few are useful to the flavour chemist. Unfortunately, the choices are very limited because of the extremely small quantities of material available to work with. Some flavour chemicals make a contribution to sensory perception when present in the food at parts per trillion (or less). This translates into the flavour chemist having a few nanograms, at best, to work with following isolation and concentration procedures. That is a long way from the milligram quantities necessary for many classical identification techniques. While quantities of components are the primary limiting factors determining the identification method, one should also recognise that most identification methods require pure compounds. This means that they will generally have to be trapped pure (one hopes) from a GC run. HRGC provides limited sample capacity, and owing to the limitations discussed, the flavour chemist depends most heavily on GC and GC-MS for identifications. Infrared and nuclear magnetic resonance may also be used but much less frequently.

9.7.1 Gas chromatography Gas chromatography is primarily used as a technique for the separation of mixtures. However, owing to its extreme sensitivity, substantial effort has been directed towards developing iden- tiﬁcation methods utilising GC. These methods are based primarily on retention time data.

9.7.1.1 Retention time/retention indices Retention time, in an absolute sense, is not a suitable parameter for compound identiﬁcation. Minor variations in operator technique, hardware or operational parameters produce retention P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

256 Food Flavour Technology

time shifts. Thus, relative retention times, based on the retention time of an unknown on co- injection of a homologous series of n-paraffins or ethyl esters, are used. The n-paraffin standard was developed by Kovats (more commonly called ‘retention indices’ today) and has become the most widely used retention index system. Compilations of retention indices of aroma compounds are available from several sources including Acree (2001), Kondjoyan and Berdague (1996) and the LRI and Odour Database (www.odour.org.uk). The flavour industry often uses ethyl ester standards. Substantial work has been done on polar GC columns (Carbowax 20M), and the non-polar n-paraffins do not provide as good retention references as the more polar ethyl esters on these columns. The use of retention indices alone for compound identification is considered weakly tentative even if the retention indices agree with published standards on two different column phases. A polar and a non-polar phase should be selected for evaluation. A necessary word of caution is that published retention indices may differ significantly from those run in another laboratory despite the use of a relative retention index. These differences arise due to operator technique, minor column differences and operational parameters. Therefore, retention indices are very accurate only when they are obtained on the same column and instrument. Published data are useful, however, in narrowing choices or giving ideas of possible identity. The analyst must have additional data, supporting an identification beyond GC data. This may be aroma character, MS or nuclear magnetic resonance data.

9.7.1.2 Selective GC detectors Information about elemental composition may be obtained by the use of specific GC detectors. The most commonly used GC detectors have been discussed earlier in this chapter, but several other detectors are also available. GC detectors available include flame ionisation, flame photometric, alkali flame ionisation, electrolytic conductivity, microcoulometric, chemiluminescence, microwave plasma (atomic emission), radioactivity and electron capture detectors. Information on these detectors can be found in the literature. The utility of selective detectors in aiding in the identification of an unknown is rather evident and will not be pursued in this chapter.

9.7.2 Infrared spectroscopy Infrared (IR) spectroscopy was at one time the second most commonly used method in flavour research (second only to GC). Compounds were separated by GC and trapped for IR to aid in identifications. Two factors have combined to move this technique now into a distant third position behind MS. The relatively large sample requirements of IR (1Ð10 ␮g) are not compatible with current HRGC. Also, the direct coupling of MS and gas chromatography and the development of low-cost, low-resolution MS instruments have made it an extremely desirable (and ubiquitous) tool for the flavour chemist. IR spectroscopy provides information on the functional groups present in an unknown. Basically, it measures the vibrational modes of a molecule, which are determined by structure and functional groups. The most common problem in applying IR to flavour research is obtaining enough pure sample to obtain a usable absorption spectrum. These requirements are, to some extent, mutually exclusive since HRGC is required for assuring purity but provides little capacity. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 257

The simplest approach has been to couple IR with GC. Innovations in IR detectors and gas cells, the use of interferometer mirrors as replacements for the monochromator and Fourier transform infrared spectroscopy have permitted the direct coupling of GC and IR. Unfortunately, gas-phase IR spectra of some compounds may differ greatly from their liquid- phase spectra. This may complicate IR interpretation and compound identification since most of the IR literature is based on liquid-phase IR. This approach also suffers from running in real time, the result being poor (inadequate) sensitivity. The alternative approach has been to use systems in which the GC effluent is frozen directly on a rotating mirrored surface. The mirrored surface is then scanned by IR ‘off-line’ to give a liquid spectrum and greater sensitivity. These instruments tend to be expensive and require considerable skill in use and, thus, have found limited application in our field.

9.7.3 Mass spectrometry 9.7.3.1 Compound identification or quantification Mass spectrometry has become the second most commonly used instrumental technique in flavour research. MS is exceptionally well adapted to flavour research since it is readily connected to GC, has excellent sensitivity (10Ð100 pg) and provides more structural information than any other spectroscopic method. There is little question that MS has been so successful in flavour research that other identification techniques have not been developed or fully applied in this area. The widespread use of GC-MS has resulted in large comprehensive spectral libraries with efficient, cheap, computerised spectrum-matching systems. This adds ease of interpretation to an already attractive method. Mass spectrometers may be classed as low-resolution (LR) or high-resolution (HR) instruments. The LR instruments provide mass measurements to the closest whole mass unit. Since many combinations of elements may give the same unit mass, LR may provide molecular mass but does not provide elemental composition. High-resolution instruments will provide sufficiently accurate mass measurements to permit determination of elemental composition. The majority of flavour work in the past has utilised LR instruments. This is primarily because LR instruments are cheaper to purchase and operate than high-resolution instruments. MS is generally used in the flavour area either to determine the identity of an unknown or to act as a mass-selective GC detector. As mentioned, MS as an identification tool is unequalled by other instruments. The systems have largely become turnkey systems that require little or no operator expertise. If the operator can do GC, he or she can do MS. Comprehensive MS libraries and efficient searching algorithms make identification simple. Herein lies a danger. MS will provide the best match (suggest the identity) for any unknown irrespective of the validity of the match. The neophyte often accepts the proposed identifications without question and obtains incorrect identifications. It is essential that any MS identifications be supported by other data, for example, GC retention data, IR, NMR or odour character. The use of a mass spectrometer as a selective GC detector can facilitate some quantification problems that would be difficult or impossible by other techniques. For this purpose, the mass spectrometer is operated in the selected ion or multiple-ion detection mode. In this mode, it continuously measures only selected ions at very short intervals throughout a GC run. This makes it possible to separately quantify two components that are co-eluting from the GC (this may be two different compounds or a compound and its stable isotope counterpart used in quantitative studies). The mass spectrometer is generally set up to measure two or three ions P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

258 Food Flavour Technology

unique to each component being quantified. It is wisest to use two or three ions to minimise the probability of an individual ion being contributed by another compound. The computer will then reconstruct mass chromatograms for each of the masses quantified. The extreme specificity and sensitivity of this technique make selected ion monitoring a very valuable technique. Selected ion monitoring can also be used to enhance GC detection sensitivity or do suspect screening runs. The MS detector is generally more sensitive than flame ionisation detection. Thus, a gain is made in instrument sensitivity. If one knows or suspects certain compounds (e.g. off-flavour taints) in a GC profile, selected ion monitoring is an excellent means of screening the flavour extract for these compounds. Poor GC resolution or limited sample may make interpretation of complete MS scans difficult, whereas selected ion monitoring is quite simple in these situations. Despite the use of high-resolution capillary columns, we often find poorly resolved peaks in the GC analysis of complex flavour isolates. Inadequate separation makes identification by MS problematic and quantification more tedious. Recent innovations, i.e. the development of fast scanning MS systems and sophisticated data analysis software, have greatly simplified both tasks. Fast MS scanning permits the software to identify MS ions that are changing together in a run (partially resolved components) and identify that collection of ions as separate, unique components. For example, the total ion chromatogram of the MS output shown in Fig. 9.11 would be one fairly broad peak. However, if fast scanning is used so that small differences in MS time profiles are available and the data are analysed by peak deconvolution software, the ion chromatogram shown in this figure emerges. One can readily see that this peak is composed of at least six compounds, six apexes and sets of ions are

Fig. 9.11 Example of peak deconvolution (Leco Corp.) P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 259

slightly off set in time. Entering each individual set of ions in the MS database resulted in the identification of three of the components making up what appeared to be a single peak. Obtaining a good identification of a mixed spectrum would have been very problematic. The resultant peaks may also be integrated for quantification purposes. A review discussing the benefits of this technique has been published by Holland and Gardner (2002).

9.7.3.2 Direct MS of food aromas to measure aroma release MS may be used for the direct measurement (no separation) of volatiles released from foods or in human breath during eating (Taylor, 1996). This has applications in understanding how food composition and texture influence flavour perception. The process has been evolutionary as one would expect. One of the earliest methods for measuring aroma release from foods by MS simply passed a carrier gas over a food in a vial that was shaken with glass beads while held at constant temperature. The gas eluting from the vial was fed directly into an electrospray ionisation mass spectrometer (Lee, 1986). This technique lacked sensitivity and measured total volatiles as opposed to individual aroma components. Oxygen and moisture in the system were limiting factors. Soeting and Heidema (1988) and Springett et al. (1999) chose to use a membrane MS inlet to remove the interfering components, but this adds elements of compound discrimination and reduces resolution, response time and sensitivity. The best approach to date was developed by Taylor et al. (2000b) and involves the use of an atmospheric pressure ionisation inlet coupled to MS. Taylor et al. (2000b) interfaced this inlet to a quadrupole mass spectrometer, while Grab and Gfeller (2000) used an ion trap mass spectrometer. These inlet systems are extremely robust and tolerate the host of non-aroma components, water and oxygen, prevalent in the human breath or on foods. The MS system and its performance can be best understood by the illustrations that follow. The inlet system designed by Taylor et al. (2000b) is presented in Fig. 9.12. In this system, the breath is drawn into the ionisation source by a Venturi effect created by high-nitrogen gas flows (source gas). The volatiles in the breath are ionised by the corona discharge pin and drawn into the MS analyser. The mass spectrometer then monitors individual ions, characteristic of the compounds of interest. The raw data look like that shown in Fig. 9.13 for chewing a mint gum. The top line indicates chewing times, the acetone line is indigenous to the breath and indicates breathing times, the next two lines (carvone and menthone) reflect their release from the gum during chewing and the bottom line is time/intensity data from a panellist. These data are tabulated and smoothed to result in the view of aroma release presented in Fig. 9.14.

Fig. 9.12 APCI inlet designed by Taylor et al. (2000b). P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

260 Food Flavour Technology

Fig. 9.13 Raw data obtained on aroma release during chewing on a mint gum (Taylor et al., 2000a).

Since Taylor’s seminal work in this area, many researchers have made modifications to this technique such as using other types of ionisation (e.g. proton transfer mass spectrometry), mass analyser (ion trap or time of flight), means of sampling and sample entry into the MS. In addition, substantial effort has been put into understanding the issues in sampling human breath or designing an apparatus to simulate the human eating process (Deibler and van Ruth, 2005). A quick search on SciFinder yields 136 references using the key words of ‘aroma release from foods’. This topic has been a major research focus in attempting to understand how flavours and foods interact to provide the stimuli that provide the basis of human perception. The abundance of literature and breadth of research make it impossible to adequately summarise the developments in this area for it can constitute a book in itself (see also Chapter 10). A great deal of knowledge has been gained on food and flavour interactions. Similarly, the ability to analytically measure the stimuli provided to a human subject on eating has provided substantial insight into the process of human flavour perception.

Time (minute) Fig.r 9.14 The release of aroma, sucrose and ﬂavour perception in chewing gums (Taylor et al., 2000a). , Sucrose release; , menthone release; , time–intensity. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 261

9.8 ELECTRONIC ‘NOSES’

The finding that some materials change in electrical properties when exposed to chemical vapours resulted in the emergence of a new field Ð the electronic nose. Persaud and Dodd (1982) were the first to publish on the concept of a model nose, which was an attempt to use a sensor-based machine to model human olfaction. The technique gained stride in the late 1990s and one can now find 1433 citations to electronic noses in the SciFinder database (published in the last 11 years). A comprehensive review of this instrument is provided by Roeck et al. (2008). Electronic noses function by analysing a sensor array response to a complete aroma: there is no separation of aroma components. The sensory array response to any given aroma is correlated (using pattern recognition software) to sensory panel data. Using neural network software and many ‘acceptable’ and ‘unacceptable’ samples, the system determines a sensor response pattern that is representative of a fresh milk versus a spoiled milk, for example. The technique is particularly attractive for quality control applications where categorisation is desired. The sensors are key components of this system (Hodgins, 1997). Currently, there are numerous types of sensors including semiconductor gas sensors (metal oxides), surface acoustic wave devices, optical, biosensors, conducting polymers and MS-based sensors (Roeck et al., 2008). In the current instruments, it is common to combine sensor types to gain a wider range in responses. At first glance, the technique appears to be ideal in that there is no need for separation of volatiles. This can result in very rapid analysis. Also, it seems to be based on a process similar to the human olfactory system in that both the electronic nose and human olfactory systems consist of a host of receptors (sensors) and yield a pattern of response to any given aroma. The brain, in the case of the human, and the computer, in the case of the electronic nose, make judgements based on a pattern recognition process as to the aroma and its quality. Thus, the speed is attractive and the theoretical foundation appears to be rational. However, human sensors are very different from machine sensors as is the process of perception, which is based on many sensory inputs, not just the volatile profile being presented. A further weakness of such instrumentation is that one has no clear idea of what the instrument is responding to in making a judgement. One chooses to evaluate some sensory parameters (e.g. staling or rancidity during storage) and then asks the instrument to develop a means of predicting that sensory parameter. In the end, the instrument uses some stimuli/response pattern to make a prediction, but one has no idea of what the instrument was measuring. For example, roast and ground coffee gives off CO2 during ageing. In a storage study where one is determining the sensory quality of coffee and obtaining electronic nose correlations, the instrument could be responding to CO2 as opposed to any oxidised flavour. As long as CO2 outgassing is correlated to lipid oxidation, the relationship is good. If it is not correlated in all situations, then the relationship is likely to be invalid in other systems or studies. The point is that the human brain uses causative input/patterns to make judgements, while the electronic nose uses patterns that are not necessarily causative and may be only casually or haphazardly related. Another potential concern about this type of instrumentation is that sensors may respond to water vapour or CO2, and these responses may dominate or unduly alter sensor patterns. The sensors also deteriorate with time (or can be ‘poisoned’), therefore changing response, which makes shelf-life studies problematic or frequent calibration necessary. These weaknesses P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

262 Food Flavour Technology

largely relegate the technique to quality control situations as opposed to research studies. However, one must recognise the inherent weaknesses of the technique so as not to misuse it.

9.9 SUMMARY

Quantitative isolation and analysis of volatile flavour compounds are fraught with practical difficulties, and an appreciation of the limitations this places on data interpretation is essential. This is clearly illustrated by the poor performance of flavours formulated from analytical data alone. Flavour analyses can be used successfully in comparing samples with the same matrix (e.g. a hydrocolloid solution), but extrapolating data beyond the limits of the analysis can lead to gross errors.

REFERENCES

Abbot, N., Etievant, P., Issanchou, S. and Langlois, D. (1993) Evaluation of the representativeness of the odor of beer extracts prior to analysis by GC eluate sniffing. J. Agric. Food Chem. 41, 1698Ð1703. Acree, T.E. (1993) Gas-chromatographyÐolfactometry in flavor analysis. In: Flavor Measurement (eds C.H. Manley and C.T. Ho), Marcel Dekker, New York, pp. 77Ð94. Acree, T.E. (2001) Flavornet web site. http://www.nysaes.cornell.edu/flavornet/ (accessed 27 August 2009). Acree, T., Barnard, J. and Cunningham, D.G. (1984) A procedure for the sensory analysis of gas chromatographic effluents. Food Chem. 14, 273Ð286. Audouin, V., Bonnet, F., Vickers, Z.M. and Reineccius, G.A. (2001) Limitations in the use of odor activity values to determine important odorants in foods. In: Gas Chromatography-Olfactometry, ACS Symposium Series 782 (eds J.V. Leland, P. Schieberle, A. Buettner and T. Acree), American Chemical Society, Washington, DC, pp. 156Ð171. Blank, I. (1997) Gas chromatographyÐolfactometry in food aroma analysis. In: Techniques for Analyzing Food Aroma (ed. R. Marsili), Marcel Dekker, New York, pp. 293Ð330. Buckholz, L.L., Withycombe, D.A. and Daun, H. (1980) Application and characteristics of polymer adsorption method used to analyze flavor volatiles from peanuts. J. Agric. Food Chem. 28, 760Ð765. Buttery, R.G. and Teranishi, R. (1963) Measurement of fat oxidation and browning aldehydes in food vapors by direct vapor injection gasÐliquid chromatography. J. Agric. Food Chem. 11, 504Ð507. Chaintreau, A. (2001) Simultaneous distillationÐextraction: from birth to maturity Ð review. Flavour Frag. J. 16, 136Ð148. Cobb, C.S. and Bursey, M.W. (1978) Comparison of extracting solvents for typical volatile components of eastern wines in model aqueous-alcoholic systems. J. Agric. Food Chem. 26, 197Ð199. David, F.T. and Sandra, P. (2003) Stir-bar sorptive extraction of trace organic compounds from aqueous matrices. LCGC 21(2), 109, 111Ð113, 116Ð117. Deibler, K.D. and van Ruth, S. (2005) Simulation of mouth conditions for flavor analysis. In: Handbook of Food Analytical Chemistry: Pigments, Colorants, Flavors, Texture, and Bioactive Food Components (ed. R.E. Wrolstad), John Wiley & Sons, Inc., Hoboken, NJ, pp. 313Ð327. Dupuy, H.P., Fore, S.P. and Goldblatt, L.A. (1971) Elution and analysis of volatiles in vegetable oils by gas chromatography. J.Am.OilChem.Soc.48, 876. Engel, W., Bahr, W., Schieberle, P. (1999) Solvent assisted flavour evaporation Ð a new and versatile technique for the careful and direct isolation of aroma from complex food matrices. Eur. Food Res. Technol. 209, 237Ð241. Engel, E., Baty, C., le Corre, D., Souchon, I. and Martin, N. (2002) Flavor-active compounds potentially implicated in cooked cauliflower acceptance. J. Agric. Food Chem. 50(22), 6459Ð6467. Friedrich, J.E. and Acree, T.E. (2000). Issues in gas chromatographyÐolfactometry methodologies. In: Flavor Chemistry, ACS Symposium Series 756 (eds S.J. Risch and C.-T. Ho), American Chemical Society, Washington, DC, pp. 124Ð132. Frijters, J.E. (1978) A critical analysis of the odor unit number and its use. Chem. Senses Flavour 3(2), 227Ð233. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 263

Grab, W. and Gfeller, H. (2000) Flavorspace Ð a new technology for the measurement of fast dynamic changes of flavour release during eating. In: Frontiers of Flavour Science (eds P. Schieberle and K.H. Engel), Deutsche Forschungsanstalt fur¬ Lebensmittel, Garching, Germany, pp. 261Ð270. Grimm, C.C., Lloyd, S.W., Miller, J.A. and Spanier, A.M. (1997) The analysis of food volatiles using direct thermal desorption. In: Techniques for Analyzing Food Aroma (ed. R. Marsili), Marcel Dekker, New York, pp. 59Ð79. Guadagni, D.G., Buttery, R.G. and Harris, J. (1966) Odour intensities of hop oil components. Sci. Food Agric. Technol. 17, 142Ð144. Guntert, M., Krammer, G., Sommer, H. and Werkhoff, P. (1998) The importance of vacuum headspace methods for the analysis of fruit flavors. In: Flavor Analysis, ACS Symposium Series 705 (eds C.J. Mussinan and M.J. Morello), American Chemical Society, Washington, DC, pp. 38Ð60. Handley, A.J. and Adlard, E.R. (eds) (2001) Gas Chromatographic Techniques and Applications.Sheffield Academic Press, Sheffield, London, p. 337. Hartman, T.G., Lech, J., Karmas, K., Rosen, R.T. and Ho, C.T. (1993) Flavor characterization using adsorption trapping Ð thermal desorption or direct thermal desorptionÐgas chromatography and gas chromatography/mass spectrometry. In: Flavor Measurement (eds C.H. Manley and C.T. Ho), Marcel Dekker, New York, pp. 37Ð60. Hernandez, R., Hernandez, R., Jr., and Axelrod, A. (1961) Standardization of silica acid for chromatography. Anal. Chem. 33, 370Ð373. Hodgins, D. (1997) The electronic nose: sensor array-based instruments that simulate the human nose. In: Techniques for Analyzing Food Aroma (ed. R. Marsili), Marcel Dekker, New York, pp. 331Ð371. Holland, J.F. and Gardner, B.D. (eds) (2002) The advantages of GC-TOFMS for flavor and fragrance analysis. In: Flavor, Fragrance and Odor Analysis (ed R. Marsili), Marcel Dekker, New York, pp. 107Ð138. Hufnagel, J.C. and Hofmann, T. (2008) Quantitative reconstruction of the nonvolatile sensometabolome of a red wine. J. Agric. Food Chem. 56(19), 9190Ð9199. Jennings, W.G. (1979a) Vapor-phase sampling. J. High Resout. Chromatogr. Chromatogr. Commun. 2, 221Ð224. Jennings, W.G. (1979b) The use of glass capillary columns for food and essential oil analysis. J. Chromatogr. Sci. 17, 636Ð639. Jennings, W.G. and Filsoof, M. (1977) Comparison of sample preparation techniques for gas chromatographic analysis. J. Agric. Food Chem. 25, 440Ð445. Jennings, W., Mittlefehldt, E. and Stremple, P. (1997) Analytical Gas Chromatography, 2nd edn, Academic Press, San Diego, CA. Kerscher, R. and Grosch, W. (2000) Comparison of the aromas of cooked beef, pork and chicken. In: Frontiers of Flavour Science (eds P. Schieberle and K.H. Engel), Deutsche Forschungsanstalt fur¬ Lebensmittel, Garching, Germany, pp. 17Ð20. Kondjoyan, N. and Berdague, J.L. (1996) A Compilation of Relative Retention Indices for the Analysis of Aromatic Compounds, Edition du Laboratoire Flaveur, Station de Recherche sur la Viande, Sainte Genes Champanelle, France. Leahy, M.M. and Reineccius, G.A. (1984) Comparison of methods for the analysis of volatile compounds from aqueous model systems. In: Analysis of Volatiles: New Methods and Their Application (ed. P. Schreier), DeGruyter, Berlin, pp. 19Ð47. Lee, W.E. (1986) A suggested instrumental technique for studying dynamic flavor release from food products. J. Food Sci. 51, 249Ð250. Legendre, M.G., Fisher, G.S., Fuller, W.H., Dupuy, H.P. and Rayner, E.T. (1979) Novel techniques for the analysis of volatiles in aqueous and nonaqueous systems. J. Am. Oil Chem. Soc. 56, 552Ð555. Leland, J.J., Schieberle, P., Buettner, A. and Acree, T.E. (2001) Gas Chromatography–Olfactometry: The State of the Art, ACS Symposium Series 782, American Chemical Society, Washington, DC. Li, H., Jia, S. and Zhang, W. (2008) Rapid determination of low-level sulfur compounds in beer by headspace gas chromatography with a pulsed flame photometric detector. J. Am. Soc. Brew. Chem. 66(3), 188Ð191. Marsili, R. (ed.) (1997) Techniques for Analyzing Food Aroma, Marcel Dekker, New York. Marsili R. (ed.) (2002a) Flavor, Fragrance, and Odor Analysis, Marcel Dekker, New York, p. 425. Marsili, R. (2002b) SPME comparison studies and what they reveal. In: Flavor,Fragrance, and Odor Analysis (ed. R. Marsili), Marcel Dekker, New York, pp. 205Ð227. McDaniel, M.R., Miranda-Lopez, R., Watson, B.T., Michaels, N.J. and Libbey, L.M. (1990) Pinot noir aroma: a sensory/gas chromatographic approach. In: Flavors and Off-Flavors (ed. G. Charalambous), Elsevier, Amsterdam, pp. 23Ð36. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

264 Food Flavour Technology

Milo, C. and Blank, I. (1998) Quantification of impact odorants in food by isotope dilution assay: strengths and limitations. In: Flavor Analysis, ACS Symposium Series 705 (eds C.J. Mussinan and M.J. Morello), American Chemical Society, Washington, DC, pp. 250Ð259. Mistry, B.S., Reineccius, T.A. and Olson, L.K. (1997) Gas chromatographyÐolfactometry for the determination of key odorants in foods. In: Techniques for Analyzing Food Aroma (ed. R. Marsili), Marcel Dekker, New York, pp. 265Ð292. Mondello, L., Bartle, K. and Lewis, A. (2001) Multidimensional Chromatography, John Wiley & Sons, Chichester, UK, p. 448. Morello, M.J. (1994) Isolation of aroma volatiles from an extruded oat ready-to-eat cereal. In: Thermally Generated Flavors, ACS Symposium Series 543 (eds T.H. Parliment, M.J. Morello and R.J. McGorrin), American Chemical Society, Washington, DC, pp. 95Ð101. Niessen, W. (ed.) (2001) Current Practice of Gas Chromatography–Mass Spectrometry, Marcel Dekker, New York, p. 507. Nijssen, L.M., Visscher, C.A., Maarse, H., Willemsens, L.C. and Boelens, M.H. (1996) Volatile Compounds Found in Food, 7th edn, TNO Nutrition and Food Research Institute, Zeist, The Netherlands. Nongonierma, A., Cayot, P., Le Quere, J.-L., Springett, M. and Voilley, A. (2006) Mechanisms of extraction of aroma compounds from foods, using adsorbents. Effect of various parameters. Food Rev. Int. 22, 51Ð94. Paetznick, D., Peppard, T. and Reineccius, G.A. (2009) 8th International Conference of the International Society for Breath Odor Research, April 26Ð28, Dortmund, Germany. Parliment, T.H. (1981) Concentration and fractionation of aromas on reverse-phase adsorbents. J. Agric. Food Chem. 29, 841Ð845. Parliment, T.H. (1997) Solvent extraction and distillation techniques. In: Techniques for Analyzing Food Aroma (ed. R. Marsili), Marcel Dekker, New York, pp. 1Ð26. Patton, S. and Josephson, D. (1957) A method for determining significance of volatile flavor compounds in foods. Food Res. 22, 316Ð318. Pawliszyn, J. (1997) Solid Phase Microextraction: Theory and Practice,VCH,NewYork. Persaud, K. and Dodd, G. (1982) Analysis of discrimination mechanisms in the mammalian olfactory system using a model nose. Nature 299, 352Ð355. Piggot, J.R. (1990) Relating the sensory and chemical data to understand flavor. J. Sens. Stud. 4, 261Ð272. Pollien, P., Ott, A., Montigon, F., Baumgartner, M., Munoz-Box,˜ R. and Chaintreau, A. (1997) Hyphen- ated headspace-gas chromatographyÐsniffing technique: screening of impact odorants and quantitative aromagram comparisons. J. Agric. Food Chem. 45(7), 2630Ð2637. Qian, M. (2000) Study of Parmesan Cheese Aroma, PhD dissertation, University of Minnesota, St Paul, MN. Reineccius, G.A. (2006) Flavor Chemistry and Technology. Taylor & Francis, Boca Raton, FL, p. 489. Reineccius, G.A. and Anandaraman, S. (1984) Analysis of volatile flavors in foods. In: Food Constituents and Food Residues: Their Chromatographic Determination (ed. J.E. Lawrence), Marcel Dekker, New York, pp. 195Ð293. Reineccius, G.A. and Liardon, R. (1985) The use of charcoal traps and microwave desorption for the analysis of headspace volatiles above heated thiamine solutions. In: Topics in Flavour Research (eds R.G. Berger, S. Nitz and P. Schreier), H. Eichorn, Marzling-Hangenham, Germany, pp. 125Ð138. Risner, C.H. and Kiser, M. (2008) High-performance liquid chromatography procedure for the determination of flavor enhancers in consumer chocolate products and artificial flavors. J. Sci. Food Agric. 88(8), 1423Ð1430. Roberts, D.D. and Pollien, P. (2000) Dependence of ‘salting out’ phenomenon on the physical chemical properties of the compounds. In: Frontiers of Flavour Science (eds P. Schieberle and K.H. Engel), Deutsche Forschungsanstalt fur¬ Lebensmittel, Garching, Germany, pp. 311Ð315. Roberts, D.D., Pollien, P.and Milo, C. (2000) Solid phase microextraction method development for headspace analysis of volatile flavor compounds. J. Agric. Food Chem. 48, 2430Ð2437. Roeck, F., Barsan, N. and Weimar, U. (2008) Electronic nose: current status and future trends. Chem. Rev. 108(2), 705Ð725. Rothaupt, M. (1998) Thermo desorption as sample preparation technique for food and flavor analysis by gas chromatography. In: Flavor Analysis (eds C.J. Mussinan and M.J. Morello), ACS Symposium Series 705, American Chemical Society, Washington, DC, pp. 116Ð122. Schaefer, J. (1981) Isolation and concentration from the vapor phase. In: Handbuch der Aroma Frischung (ed. H.B. Maarse), Academie-Verlag, Berlin. p. 44. Schmiech, L., Uemura, D. and Hofmann, T. (2008) Reinvestigation of the bitter compounds in carrots (Daucus carota L.) by using a molecular sensory science approach. J. Agric. Food Chem. 56(21), 10252Ð10260. P1: SFK/UKS P2: SFK c09 BLBK221-Taylor/Linforth November 21, 2009 7:59 Printer Name: Yet to Come

Instrumental methods of analysis 265

Siek, T.J. and Lindsay, R.C. (1968) Volatile components of milk fat steam distillates identified by gas chromatography. J. Dairy Sci. 57(12), 1887Ð1896. Soeting, W.J. and Heidema, J. (1988) A mass spectrometric method for measuring flavor concentration/time profiles in human breath. Chem. Senses 13, 607Ð617. Springett, M.B., Rozier, V. and Bakker, J. (1999) Use of fibre interface direct mass spectrometry for the determination of volatile flavour release from model systems. J. Agric. Food Chem. 47, 1125Ð1131. Sucan, M.K., Fritz-Jung, C. and Ballam, J. (1998) Evaluation of purge-and-trap parameters: optimization using a statistical design. In: Flavor Analysis, ACS Symposium Series 705 (eds C.J. Mussinan and M.J. Morello), American Chemical Society, Washington, DC, pp. 22Ð37. Sullivan, J.F., Konstance, R.P., Calhoun, M.J., Talley, F.B., Cording, J., Jr., and Panasiuk, O. (1974) Flavor and storage stability of explosion-puffed potatoes. Nonenzymatic browning. J. Food Sci. 39, 58Ð60. Taylor, A.J. (1996) Volatile release from foods during eating. Crit. Rev. Food Sci. Nutr. 36(8), 765Ð784. Taylor, A.J., Linforth, R.S.T., Baek, I., Marin, M. and Davidson, J.M. (2000a) Flavour analysis under dynamic conditions: measuring the true profile sensed by consumers. In: Frontiers of Flavour Science (eds P. Schieberle and K.H. Heinze), Deutsche Forschungsanstalt fur¬ Lebensmittel, Garching, Germany, pp. 255Ð260. Taylor, A.J., Linforth, R.S.T., Harvery, B.A. and Blake, A. (2000b) Atmospheric pressure chemical ionization mass spectrometry for in vivo analysis of volatile flavour release. Food Chem. 71, 327Ð338. Teitelbaum, C.L. (1977) A new strategy for the analysis of complex flavors. J. Agric. Food Chem. 25, 466Ð470. Teranishi, R. (1998) Challenges in flavor chemistry: an overview. In: Flavor Analysis, ACS Symposium Series 705 (eds C.J. Mussinan and M.J. Morello), American Chemical Society, Washington, DC. Toelstede, S., Dunkel, A. and Hofmann, T. (2009) A series of kokumi peptides impart the long-lasting mouthfulness of matured Gouda cheese. J. Agric. Food Chem. 57(4), 1440Ð1448. Ullrich, F. and Grosch, W.Z. (1987) Identification of the most intense odor compounds formed during the autoxidation of linoleic acid. Z. Lebensm.-Unters. Forsch. 184, 277Ð282. Wampler, T.P. (1997) Analysis of food volatiles using headspaceÐgas chromatographic techniques. In: Tech- niques for Analyzing Food Aroma (ed. R. Marsili), Marcel Dekker, New York, pp. 27Ð58. Weurman, C. (1974) Sampling in airborne odorant analysis. In: Human Responses to Environmental Odors (eds A.J. Turk, W. Johnston and D.G. Moulton), Academic Press, New York, pp. 263Ð328. White, R.L. (2009) Fast GC with a small volume column oven and low power heater. Chromatographia 69(1Ð2), 129Ð132. Widner, H.M. (1990) Recent developments in instrumental analysis. In: Flavour Science and Technology (eds Y. Bessiere and A.F. Thomas), Wiley, Chichester, UK, pp. 180Ð190. Wright, D.W. (1997) Application of multidimensional gas chromatography techniques to aroma analysis. In: Techniques for Analyzing Food Aroma (ed. R. Marsili), Marcel Dekker, New York, pp. 113Ð141. Zhang, Z., Yang, M.L. and Pawliszyn, J. (1994) Solid phase-microextraction: a solvent-free alternative for sample preparation. Anal. Chem. 66, 844AÐ857A. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

10 On-line monitoring of ﬂavour processes Andrew J. Taylor and Robert S.T. Linforth

The aim of this chapter is to introduce readers to the techniques used to monitor flavours on-line. The rationale for on-line monitoring is given, along with the analytical requirements for the real time, in vivo measurement of aroma release. This sets the specification for the analytical method as well as demonstrating some of the difficulties and limitations associated with on-line flavour analysis. A brief history of on-line analysis is given to explain how the current techniques evolved and this is followed by an overview of the techniques currently available. Finally, examples of applying on-line monitoring of flavours are given to illustrate the power of the technique, and some future needs and trends are discussed.

10.1 INTRODUCTION

Analysis of flavour compounds has been a major interest for analysts and for flavour technologists over many years. On the analytical side, the advent of gas chromatography (GC) and then high-performance liquid chromatography (HPLC) allowed the detailed composition of many flavour materials to be elucidated in terms of compound identification and quantification. Besides giving food technologists a powerful tool to study the composition of flavours, the data could also be used to link flavour composition with sensory flavour quality. In other words, the aim was to understand questions such as the following: r r Why do particular mixtures of compounds taste pleasant? Why do quite small changes in some flavour components cause significant changes in sensory scores?

Although understanding the link between composition and sensory quality is an attractive idea, there are very few published papers that compare the flavour composition of a product to its sensory attributes. One example is that of Togari et al. (1995) who used a chemometric approach to correlate flavour compositions of different tea samples with individual sensory attributes. The various sensory attributes of tea (e.g. the ‘fresh floral’ attribute) were expressed as equations based on the concentrations of those compounds that contributed positively and negatively to the sensory attribute, with each compound being given a weighting, calculated to give the best fit of the experimental data P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 267

(Equation 10.1).

Fresh ﬂoral =−0.482[methylpyrazine] + 0.931[linalool] + 0.329[(E)-2-hexenyl hexanoate] − 0.217[5, 6-epoxy-␤-ionone] (10.1) + 0.615[jasmine lactone] − 0.958

The linear nature of the equations is a little surprising as the well-known psychophysical laws (Fechner, 1860) suggest a non-linear relationship between the odour stimulus concentration and the perceived odour intensity. Steven’s ‘power’ law (Stevens, 1969) describes the situation for a single compound as follows:

P = aSb (10.2)

where P is the perceived intensity of a taste or smell and S is the stimulus concentration multiplied by a scaling factor a and an exponent b, whose values typically lie between 0.55 (coffee odour) and 1.3 (sucrose) (Coren et al., 1999). The apparent simplicity of the equation, the lack of citations or similar publications since 1995 and failed attempts in our laboratory to use this type of approach in a predictive way suggest that the relationships may hold for the samples tested but are not robust enough to apply to other samples. It also suggests that the concept of predicting sensory behaviour from the flavour compositions is flawed. One of the reasons for this failure is that the concept ignores the changes that take place during consumption of food when the multitude of flavour compounds in a food are subjected to processes such as chemical changes, hydration, partition, mass transfer and dilution, which change the flavour profiles delivered to the flavour receptors. The situation was well described by Darling et al. (1986) who commented that ‘it is difficult to define what fraction of the total added flavour in the food is available for perception’. Studies on other human senses, such as vision, have described a similar situation and defined the differences between the whole object and the signals received at the receptors in terms of the ‘distal’ and ‘proximal’ stimuli (Coren et al., 1999). For food flavour, the distal stimulus represents the flavour composition in the food while the proximal stimulus represents those elements of the distal stimulus that actually reach and activate the receptors, i.e. the flavour profile after oral processing. Measuring these changes is a significant analytical challenge. One approach is to mimic the mastication process in vitro using model mouths but successful mouth mimics have proved very difficult to develop as outlined in Chapter 7. A second approach is to mathematically model flavour release in vivo using physicochemical and engineering principles (Chapter 8). This produces theoretical models, which rely on approximations and generalisations about the masticatory behaviour of humans and which generally lack appropriate experimental validation. The third approach, experimentally measuring the aroma and taste profiles delivered to the flavour receptors in vivo, has therefore been proposed to take into account the factors involved in oral processing and overcome the difficulties in modelling the eating process. However, monitoring such changes in vivo also raises interesting analytical issues but the potential benefits are as follows: r Measuring the taste and aroma profiles close to the receptors (i.e. the proximal stimulus) r will show the effect of eating and mastication on the flavour profile of the food. These in vivo flavour profiles should also correlate better with perceived flavour. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

268 Food Flavour Technology

The desire to link ﬂavour composition with sensory perception and the problems with other approaches (modelling etc.) have therefore driven researchers to investigate methods for measuring ﬂavour release during eating.

10.2 ISSUES ASSOCIATED WITH IN VIVO MONITORING OF FLAVOUR RELEASE

To achieve successful monitoring of aroma and taste release in vivo, there needs to be a dialogue between analysts and physiologists to identify the key issues (Taylor, 2002) and to ensure that the protocols developed will serve the purpose. The issues are presented and discussed in the following sections.

10.2.1 Speed of analysis Odour and taste recognition are rapid processes. It takes around 700Ð800 milliseconds between administration of an odour to the nose and its recognition (Laing and Macleod, 1992) and similar times were observed for tastants (Kuznicki and Tumer, 1988). In addition, human breathing is a tidal event with duration of about 5 seconds (Sherwood, 2006). To obtain adequate resolution of these physiological processes, about 50Ð250 data points are needed per breath and this then sets the sampling time to 100Ð20 milliseconds/data point, respectively. Experience has shown that slower sampling times can miss some of the detail of the eating process (for instance, the individual chewing movements that pump air into the throat; Hodgson et al., 2003), and accurate determination of the maximum intensity of a compound in each individual breath can also suffer (see Section 10.6.1). Using sampling speeds of 20Ð100 milliseconds rules out chromatographic methods as no existing technique can provide such fast separation and multiple analyses would be too time-consuming. Instead, the concept that analytes in air could be monitored using direct mass spectrometry (where compounds are directly sampled into the mass spectrometer and differentiated only by their m/z values) has been proposed, based on previous work from the chemical warfare (Ketkar et al., 1991) and health ﬁelds (Benoit et al., 1983). In the following sections, the suitability of direct mass spectrometry to address the other analytical requirements is tested.

10.2.2 Analysis of different chemical classes The next analytical requirement is the need to analyse a wide variety of compounds at low concentrations. The number of compounds in flavours varies. Complex natural flavours such as roast coffee can contain several hundred different compounds (Flament, 2002), although a smaller number are sensorially significant. Using odour activity values, 14 compounds were proposed as being the key odourants in coffee (Semmelroch et al., 1995) whereas a similar study using Charm analysis yielded a list of 30 compounds (Deibler et al., 1998). Whatever the exact number, it seems that there is no need to monitor the release behaviour of every compound in a flavour; studying a smaller, selected subset of compounds may monitor the key odourants and indicate the general trends. This approach is the basis for QSPR modelling of flavour release as described in Chapter 8 where the physicochemical properties of a subset of compounds have been used to build a model from which the behaviour of other compounds can be predicted. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 269

10.2.3 Sensitivity Sensitivity of monitoring is a key issue as flavour compounds have very different sensory thresholds. The common salty and sweet taste compounds are generally sensorially active in the range of 1Ð200 g/L, whereas some bitter compounds have extremely low thresholds (Schmiech et al., 2008) as do the high-intensity sweeteners such as aspartame (13 mg/L). For aroma compounds, thresholds vary from a few parts per trillion by volume (pptv, pL of aroma per L air), through parts per billion (ppbv, nL/L) to parts per million (ppmv, ␮L/L). The compilations of taste and odour thresholds (Van Gemert and Nettenbreijer, 1977; Devos et al., 1990; Rychlik et al., 1998) show wide ranges of values, which have been attributed to the variation in the population and to the difficulty in delivering consistent aroma stimuli to panellists when determining the odour thresholds (Vuilleumier et al., 2002). The values are usually quoted as concentrations of odourants in water or in air, and the latter values are pertinent for the case of on-line monitoring. Thus, to monitor all compounds at their odour threshold values, a highly sensitive technique is needed (Taylor, 2002). There is also an argument that aroma should be measured below these levels as there is evidence that subthreshold levels of aromas can be perceived in the presence of tastants (Dalton et al., 2000). From this discussion, the analytical needs become clear. The first need is that on-line analysis to measure aroma release in the nose should cover a wide dynamic range, potentially across 6 orders of magnitude. The second is the requirement to analyse small amounts of analytes, given the low concentrations and the small volumes of air that will be sampled. The amount available can be calculated using data from the literature on breath volume and breathing rate and some simple assumptions. For example, take the case of an aroma delivered to the nose at 100 ppbv (100 nL aroma/L of air). A human breath has a typical volume of 500 mL and an exhalation occurs 12 times a minute (or every 5 seconds) (Purves et al., 2001), so the flow rate is around 100 mL/second. If the on-line mass spectrometer samples breath for 100 milliseconds, then the amount of aroma available for analysis in that period is calculated as follows (molecular weight of the aroma compound is set at 100 Da):

r Volume of breath sampled in 100 milliseconds = 100 × 0.1 = 10 mL air r × = r 10 mL of a 100 ppbv aroma contains 10/1000 100 nL 1nLaroma An aroma compound of molecular weight 100 would contain the MW (100 g) in 22.4 L r at standard temperature and pressure, so 1 L of aroma contains 4.46 g aroma Weight of aroma in 1 nL = 4.46 ng

The amounts estimated above for a 100 ppbv aroma concentration are in line with current mass spectrometer detection limits, but if we consider the pptv threshold levels of some aroma compounds, then the amount for analysis decreases by a factor of 103 and the amount available is 4.46 pg, even less if the aroma compound has an MW less than 100 Da. Detection limits of mass spectrometers are compound speciﬁc and depend on several factors such as the efﬁciency of ionisation, the stability and degree of fragmentation of the ions formed, as well as the type of mass detector used. This aspect is further discussed in Section 10.4. Therefore, direct mass spectrometry could potentially meet the sensitivity limits as well as the speed requirements for in vivo measurements of aroma release. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

270 Food Flavour Technology

10.2.4 Identification of analysed compounds Classical flavour analysis using gas chromatographyÐmass spectrometry (GC-MS) or liquid chromatographyÐmass spectrometry (LC-MS) provides information on the quantity and identity of compounds. Because there are many isomeric compounds present, and because those isomers can represent very different aromas, it is important to identify isomers, whether positional (e.g. 2-methylbutanal or 3-methylbutanal) or stereoisomers (e.g. l-menthol and d-menthol). For unequivocal identification, two or more separate analytical methods are customary. When using direct mass spectrometry, the information is limited to the presence of ions and their intensities. This limitation and ways to address it are discussed in more detail in Section 10.4.

10.2.5 Interfering factors Sampling aromas from the nose during eating will also introduce the constituent gases of air (nitrogen, oxygen, carbon dioxide) as well as water vapour into the analytical apparatus. The gases will be present in much higher amounts than the aroma compounds and, ideally, will not interfere with the analysis. For direct mass spectrometry of aromas, a key factor is choosing ionisation conditions that do not ionise air components. Water vapour can condense in parts of the sampling or analysis system and lead to aroma losses due to solubilisation, or water may interfere in the analytical process. Combating the former problem is relatively easily solved by using heated transfer lines to maintain temperatures above 100◦C and prevent condensation. The latter problem needs some understanding of the analytical process to be used. For instance, water vapour in electron-impact mass spectrometry adversely affects the process, so inlets are designed to minimise water ingress while letting aroma compounds pass through. Other ionisation methods are water tolerant (see Section 10.4).

10.2.6 Non-volatile tastants Initially, analysis of non-volatile tastants seems to present fewer problems than analysis of odourants. The ease of access to receptors on the tongue and the generally high concentrations of tastants compared to odourants are positive points, but against this, the difﬁculty of sampling the tastants from the mixture of masticated food and saliva in the mouth is a major limitation for many foods. Continuous sampling of tastants during eating is also an issue (Davidson et al., 2000), and sampling is generally achieved by swabbing the tongue at intervals of 5 seconds or so. Interference in the analysis of taste compounds is mainly associated with presence of salivary proteins and inorganic salts, which may interfere with the ionisation process in direct liquid mass spectrometry (Taylor and Lin- forth, 2003). Some tastants such as the high-intensity sweeteners or the very bitter natural compounds are present at much lower levels than the sugars and acids and do require a signiﬁcantly higher degree of sensitivity, but, usually, detection beyond the taste threshold can be achieved (Davidson et al., 2000). There is also the complication of sampling tastants delivered to the saliva phase and not the bolus (the mass of masticated food in the mouth), whereas, with aroma, the bolus and the aroma compounds delivered to the nose are spatially separated. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 271

10.3 PIONEERS AND DEVELOPMENT OF ON-LINE FLAVOUR ANALYSIS

It is always difficult to determine the exact sequence of research in the development of scientific ideas such as direct mass spectrometry. The version presented here is based partly on publication date and partly on the transfer of methods from other fields to the analysis of flavours. Our reading of the literature was initially influenced by the article from Benoit et al. (1983), which described an attempt to directly measure the volatile compounds in exhaled breath. The issues described in the article are mainly concerned with interfacing humans (who deliver one or two breaths) into the mass spectrometer system, which operates on a continuous basis, and ensuring the connection between the high-vacuum region of the mass detector and the atmospheric pressure of the human is controlled for the safety of the human subject and for the efficient operation of the mass spectrometer. The inlet system used valves and mixing chambers and was relatively complicated compared to more modern designs for in vivo aroma analyses. The instrument used by Benoit et al. also had a relatively low scanning time and was unable to provide a full scan in the exhalation time. The next notable publication came from the Unilever laboratories in Vlaardingen (Soeting and Heidema, 1988) and described a system based on electron-impact mass spectroscopy. Interfacing of breath to the reduced pressure in the ionisation chamber was achieved using a thin membrane that also excluded water vapour. The device gave the first breath-by- breath traces of odourant release (butanone) although with limited resolution and sensitivity. Other research groups explored the electron-impact system further. To improve the interfacing between atmospheric pressure and the vacuum in the electron impact (EI) source, a fibre interface (Springett et al., 1999) as well as different membranes and operating conditions were tested (Reid and Wragg, 1995). A commercial EI instrument (Balzers Ther- mostar) was produced in the late 1990s, which used a flow restrictor to sample air at a fixed rate and overcome the problems of membrane inlets. Response times were good, due to inlet design, but sensitivity was in the low ppmv range. The major drawback of electron-impact ionisation is that the fragmentation patterns, which identify individual compounds, make interpretation of the spectra extremely difficult when mixtures of odourants are present. At the time, methods for deconvoluting the data were not readily available although such software has now been developed (e.g. by the National Institute for Science and Technology, http://www.cstl.nist.gov/projects/fy05/iais05mallard.pdf) and is commercially available. The complexity of the spectra produced, and the difficulty in interpreting the data when mixtures of compounds were analysed, led other workers to consider softer ionisation techniques that produced ions that represented the molecular weight of the compound with just one or two fragments, thus making it easier to assign ions to compounds. Several soft ionisation methods using chemical ionisation were investigated for detection of pollutants and chemical warfare agents, and the extreme sensitivity (parts per trillion or pL/L) of one publication was especially attractive (Ketkar et al., 1991). Closer inspection of the information, however, was a reminder that sensitivity is compound-dependent and the sensitivity may not always translate into the flavour area. For the chemical warfare agents tested, sampling time was 15 seconds and the ions appeared in a relatively noise-free area of the spectrum, hence giving good signal-to-noise ratios and therefore good sensitivity. Their structures also encourage ion stability, which also improves sensitivity. However, when analysing odorous compounds, some of them perform well, as they meet these criteria; others P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

272 Food Flavour Technology

do not and sensitivity measurements really need to carry a caveat, which deﬁnes the sampling time and chemical species monitored as well as other factors such as background noise.

10.4 ON-LINE AROMA ANALYSIS USING CHEMICAL IONISATION TECHNIQUES

From the discussion in the previous section, the need to ionise flavour compounds to produce a simple mass spectrum and to assign compounds to ions has been established. Chemical ionisation is a much softer process than electron impact. Ionisation energies (IEs) are measured in electron volts (eV), and 70 eV is the standard energy used on commercial electron-impact mass spectrometers principally so that standard spectral libraries can be used to aid identification of unknown compounds. Chemical IEs are much lower (the equivalent of about 10 eV) and usually result in ionisation of the intact molecule with limited fragmentation. An early publication reported the use of a modified chemical ionisation mass spectrometer to monitor aroma release from an ethanolic solution (Elmore and Langley, 1996). Helium was used as the carrier gas, and a jet separator removed gas from the analyte ions. Dwell time per analyte was 100 milliseconds, but sensitivity was not quoted as release was measured in relative terms. For volatile organic compounds, ionisation using the proton transfer reaction (PTR) via + the hydronium ion (H3O ) as the ‘reagent’ is considered to be an almost universal mechanism and is represented by the following equation, where R represents an odourant molecule:

+ + H3O + R → H2O + RH (10.3)

Ionisation occurs when there is a suitable difference between the proton affinity of the hydronium ion and the analyte molecule. Proton affinity can be expressed in theoretical terms, but simply, it is a measure of how strongly the proton is associated with the molecule. Water has a proton affinity of 691 kJ/mol and will potentially transfer the proton to a molecule with a higher proton affinity. Table 10.1 shows the proton affinity of some common reagent ions, the major constituents of air and some aroma compounds. By using protonated water to donate a proton in the PTR, nitrogen, oxygen or carbon dioxide are not ionised as their proton affinity is less than that of water. However, most odourants are ionised. Unlike other physicochemical parameters, such as log P or vapour pressure, where estimated values can be obtained using group contribution software like EPI suite (US EPA, 2009 R Estimation Programs Interface SuiteTM for Microsoft Windows, v 4.00; United States En- vironmental Protection Agency, Washington, DC), estimating proton affinity values is highly complex and prohibits the calculation of estimates from chemical structure. Therefore, there is no ready database containing proton affinity values for odourants.

10.4.1 Analysis via atmospheric pressure chemical ionisation Although proton transfer using hydronium ions is the major mechanism to ionise odourants so that they can be detected and analysed by direct mass spectrometry, the process can be achieved in different ways. Atmospheric pressure chemical ionisation is the simplest process (Raffaelli, 1997; Byrdwell, 2001). A needle is charged to 3Ð4 kV, and a corona discharge is P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 273

Table 10.1 Proton afﬁnities of compounds used as reagent ions (italicised), some air constituents and some aroma compounds.

Proton affinity Compound (kJ/mol) Oxygen 421 Nitrogen 493.8 Nitric oxide 531 Carbon dioxide 540.5 Ethylene 680.5 Carbon disulfide 681.9 Cyclohexane 686.9 Water 691 Formic acid 742 Methyl sulfide 773.4 Ethanol 776.4 Benzyl alcohol 778.3 Hexanal* 795.7 Diacetyl 801.9 Acetone 812 Dimethyl disulfide 815.3 Benzaldehyde 834 Ethyl acetate 835.7 Limonene 875 Pyrazine 877.1 Thiazole 904 Indole 933.4

From Hunter and Lias (1998) and *(Blake et al., 2008).

produced at atmospheric pressure between the needle and the entrance to the high-vacuum region of the mass spectrometer. In the corona, electrical energy drives a sequence of reactions (for a full description see Byrdwell, 2001), which ultimately forms hydronium ions. This process takes place within a very small distance from the needle, and reaction between hydronium ions and the odourant molecules takes place just outside this region. Since the + ionisation process has limited control, water clusters (H2O)nH also form (Bruins, 1991) and the efficiency of proton transfer is known to decrease with increasing cluster size (Sunner et al., 1988). However, in the gas-phase atmospheric pressure chemical ionisation (APCI) sources built in our laboratory, the use of a relatively high flow rate of nitrogen make-up gas (several litres per minute), means the monomer and dimer are the major species detected with very little trimer present and no sign of higher clusters. Successful aroma analysis by APCI requires suitable interfaces. The APCI interfaces designed for liquid flows are not compatible with gas flows as the dead volumes and flow rates are not well matched to breath- by-breath analysis. One solution is the MS-Nose, originally manufactured by Micromass (now Waters) and based on a patent from our laboratory (Linforth and Taylor, 1998). A typical breath-by-breath trace obtained on the MS-Nose is shown in Fig. 10.1. A commercial soft confectionery gel was eaten and the release of four ethyl esters was monitored. The sharp spikes denote individual chewing events, while the groups of spikes represent a single exhalation and the sample was swallowed at 2.8 minutes. The traces show the fast time response of the MS-Nose, which is due to minimum dead volume and fast linear flow rates. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

274 Food Flavour Technology

Fruit sweet2 SIR of 4Channels ES+ 100 131.001.00Da 7.88e6

0 Fruit sweet2 SIR of 4Channels ES+ 100 117.001.00Da 1.23e6

0 Fruit sweet2 SIR of 4Channels ES+ 100 103.001.00Da 5.07e6

0 Fruit sweet2 SIR of 4Channels ES+ 100 89.001.00Da 1.67e6

0 Time 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 2.90 3.00

Fig. 10.1 Breath-by-breath monitoring showing release of four ethyl esters from a soft confectionery gel. Breath background was measured from 2 to 2.1 minutes and then the gel was placed in mouth and chewed, with swallowing of the sample at 2.8 minutes.

The Micromass and Waters machines possess a useful software feature that allows the cone voltage (the voltage applied to the sampling cone, which leads to the high-vacuum region) to be set for each individual ion. The cone voltage is a major factor in fragmentation, and experiments in our laboratory have deﬁned the optimal cone voltage to obtain ‘clean’ spectra (containing just the protonated molecular ion for maximum sensitivity and spectral clarity) or it can be set to induce some fragmentation to help resolve compounds with potentially the same m/z values (as discussed in Section 10.4.6). P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 275

10.4.2 Analysis via PTR PTR ionisation was developed for general volatile compound analysis and applied to flavour and other applications by Lindinger et al. (1998). In this technique, hydronium ions are formed separately (either by a radioactive source or by a hollow cathode) and introduced + to react with the analyte after a ‘clean-up’ process, which ensures only H3O ions are introduced, although water clusters may appear in the mass spectrum due to interaction of hydronium with water in the sample flow. Reaction takes place under reduced pressure in a drift tube and ions are then analysed either by a quadrupole or by a time-of-flight (TOF) mass detector (Jin et al., 2007). A key feature of PTR-MS is the ability to increase sensitivity by increasing the sampling time, whereas APCI sensitivity is less time dependent. For some applications this offers possibilities of measuring volatile compounds at very low levels, and parts per trillion (pL/L) sensitivity can be achieved for certain compounds. Although PTR + uses primarily H3O as the proton transfer agent, other reagents can also be used and can give the technique extra power. An adaptation of PTR-MS that uses multiple reagent ions has been named chemical ionisation reaction mass spectrometry (Blake et al., 2006). Because of the highly controlled environment, the amounts of compounds analysed by PTR can be calculated from theoretical principles although there are some reports that the presence of additional water vapour or carbon dioxide in the sample can affect this calculation (Keck et al., 2008) and this issue is discussed further in Section 10.4.4. Commercial PTR machines are available through Ionicon, Innsbruck, Austria, and the proceedings of their biennial conferences (Hansel and Mark,¬ 2007) report advances in the various analytical fields such as food (Biasioli, 2007).

10.4.3 Analysis via selected ion flow tube + Selected ion flow tube (SIFT) ionisation uses a microwave discharge to generate the H3O reagent ions. The process has been reviewed by Smith and Spanel (Smith and Spanel, 2005) who have been the main developers of the technique in the UK although a New Zealand research group has also been active (see for example Francis et al., 2007; Milligan et al., 2007) and this has led to the production of a commercial machine (Syft). The Syft machine + + is programmed to deliver not only hydronium ions but also NO and O2 , which can be used to introduce some selectivity into the analysis by targeted ionisation of certain chemical species, based on the different reaction mechanisms of the reagent ions. Although PTR and APCI can be set up to use other reagent ions, the Syft machine is designed to switch between the reagents during an analysis and thus extract as much information as possible from the sample. In the SIFT machines, hydronium ions are created and mass filtered to give a pure + H3O reagent ion, which leads to conventional proton transfer. However, water ion clusters and analyteÐwater clusters are sometime observed as the hydronium ions associate in the ion flow tube to some extent, with the formation of analyteÐwater clusters due to a process known as ligand switching (Equation 10.4).

+ H3O + (H2O)n + CH3OH → H3O (H2O)n−1CH3OH + H2O (10.4)

+ + With NO and O2 as reagent ions, the reaction mechanisms principally involve rapid charge transfer rather than the proton transfer/ligand switching seen with PTR and APCI. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

276 Food Flavour Technology

This mechanism produces the molecular ion (M+) as shown in Equations (10.5) and (10.6).

NO+ + M → NO + M+ (10.5) + + → + + O2 M O2 M (10.6)

+ + SIFT precursor ions, NO and O2 , are also able to ionise analyte molecules by the process of hydride ion transfer. This reaction occurs when the IEs of the precursor ion and analyte are similar. For example, when NO+ (IE = 9.26 eV) reacts with the molecule anisole (IE = 8.2 eV), ionisation always occurs via charge transfer, because the IE of anisole is sufﬁciently less than that of NO+ (Smith et al., 2003). However, in the reaction of NO+ with ethylene glycol (IE = 9.3 eV), there is insufﬁcient difference in IE and thus charge transfer is inhibited. The most common reaction mechanism in this case is hydride ion (H+) transfer (Equation 10.7).

+ + → + + NO C2H5OC2H5 C2H5OC2H4 HNO (10.7)

These reactions are, thus, characterised by the production of an (M − H)+ ion and an HNO molecule (Smith et al., 2003). The process will proceed via an exothermic pathway when the hydride affinity of the precursor ion is greater than that of the conjugate base of the analyte (Arnold et al., 1998). SIFT offers a good dynamic range, with the ability to monitor odourants from concentrations as low as pptv to ppmv. As mentioned previously, the limit of detection is compound- dependent and readers should not assume that all compounds can be analysed at pptv levels. In selected applications, SIFT is a powerful technique for monitoring specific volatile organic compounds, but for breath-by-breath analysis of mixtures it is not ideal due to the complex ion spectra formed from the different ionisation pathways and the difficulty of interpreting the data.

10.4.4 Calibration To calculate the actual amounts of volatile compounds monitored by the three techniques, two methods are used. For APCI, it is customary to introduce known concentrations of the test odourants, either as the equilibrium headspace obtained from solutions of the appropriate compounds or by injecting a few microlitres of solution of the test compound in cyclohexane into the make-up gas flow using a syringe pump (Taylor and Linforth, 2003). The latter concept requires solubility of the test compound in cyclohexane and full vaporisation of the solution, which is relatively easy to achieve in a heated, high-velocity gas flow. Cyclohexane has a proton affinity less than that of water, so ionisation by proton transfer is not favoured although a few ions associated with cyclohexane are observed in the spectrum, probably due to ionisation via different routes or due to trace impurities. From the calibration curves, the amount of the compounds can be calculated by proportion. In our laboratory, we run cali- brations daily and a diagnostic is also run using a standard solution of 2,5-dimethylpyrazine to check that machine performance is within acceptable limits. Most problems with APCI are due to accumulation of residues in the heated fused-silica transfer lines (which impact on temporal resolution) or deposits on the sample cone and the associated ion optics. Re- placement of the fused-silica or standard cleaning procedures on the mass spectrometer side usually solve the issues of poor performance. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 277

For PTR and SIFT, it is possible to calculate the amount of compound present from the ion ﬂux in the drift tube regions, using fundamental principles of ion physics. However, under certain conditions (Ammann et al., 2006), these principles may not hold and conventional calibration with known standards seems to be good laboratory practice (Keck et al., 2007).

10.4.5 Suppression The PTR is quantitative providing there is an excess of hydronium ions so that all compounds are charged. In cases where there is an excess of one volatile compound, the available charge may be consumed by the excess molecules and there may be insufficient charge to ionise other molecular species quantitatively. In limiting hydronium conditions, particularly where one of the compounds has a high proton affinity, ‘charge stealing’ may occur where the other compounds are left un-ionised and the analysis is no longer quantitative. One simple way to assess when these conditions occur is to introduce headspace from a compound with a relatively low proton affinity (3-methylbutanal) into the system and then bleed in 2,5-dimethylpyrazine in increasing quantities and observe when the 3-methylbutanal signal deviates from its original value (see Section 10.6.5 for an example). Suppression is an issue with all types of ionisation (although rarely mentioned), but APCI is particularly susceptible because the ionising power is quite low. However, by careful experimentation it is usually possible to find an operational window where quantitative data can be obtained. Experience has shown that analysis of volatile compounds from the Maillard reaction can be problematic as, initially, oxygenated compounds are formed, then nitrogenous compounds, and if the latter suppresses the former, the data can be badly misinterpreted unless operating windows have been established (Channell and Taylor, 2005). PTR-MS, by its nature, generates higher levels of hydronium ion, which provides a wider operating window, but in our experience, there seems to be a trade-off as higher hydronium ion concentrations lead to some clustering (which can be partly removed by controlling the electric field conditions in the drift tube) but the increased number of ions also seems to encourage ionÐion interaction, leading to some fragmentation (see for example Aprea et al., 2007).

10.4.6 Assigning ions to compounds for unequivocal identiﬁcation On-line mass spectrometry is best at monitoring the release of known compounds and, for many experimental studies, researchers have chosen aroma compounds that produce different ions so there is no doubt about the assignment of a particular ion(s) to a compound. However, in more complex systems, or in samples where the aroma composition is unknown, several compounds can produce the same ion and assignment then becomes more difﬁcult. Some resolution of the problem can be achieved either by introducing more selectivity into the ionisation process or by using complementary analyses. In APCI-MS, the cone voltage controls fragmentation of ions and some isobaric compounds; e.g. hexanal and hexenol (both MW 100) can be resolved by favouring dehydration + of hexenol to form an ion of the form [M − H2O + H] (m/z 83) whereas hexanal will produce the protonated molecular ions (m/z 101) when suitable cone voltages are applied. The ability to set the cone voltage for each ion monitored in the selected ion monitoring software allows cone voltage to be used to maximum effect. In PTR, the E/N parameter plays a similar role to cone voltage in APCI. E represents the voltage gradient in the drift tube P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

278 Food Flavour Technology

while N represents the gas density (or pressure), and the units of E/N are Townsends (Td). The first region of the drift tube is typically set to values between 80 and 120 Td, while the last region before entering the mass analyser is set between 200 and 300 Td to break up ion clusters. Altering the E/N settings can optimise the signal intensity for the molecular ion or induce fragmentation in the same way that cone voltage operates in APCI. However, unless individual E/N values can be set for each ion through software, a single, compromise value has to be set, depending on the nature of the compounds being analysed. Separation of isobaric compounds and the various ways of resolving them on APCI have been reported in some detail by Jublot et al. (2005). Besides cone voltage, the possibility of using MSÐMS in an ion trap detector was investigated. Some resolution of positional isomers was possible, but no resolution of stereoisomers was possible. Some isobaric compounds have been resolved using a two-stage PTR procedure (Blake et al., 2008; Inomata and Tanimoto, + 2008) where initial ionisation with H3O was followed by ionisation using reagents such as protonated acetone. In this way, ethyl acetate and 1,4-dioxane could be resolved, even though at ppm levels. A second approach is based on GC separation of the compounds found in the headspace of a sample followed by simultaneous detection with EI-MS and APCI detectors in parallel. The GC-APCI trace can then be examined at a particular peak and the ion masses present identified and compound identity established using the data from the parallel GC-EI trace. The APCI chromatogram is then searched for the same ion masses at different retention times. If the ion only appears at one retention time, then this indicates that a compound can be unequivocally assigned to that ion. If the ion occurs at in other peaks in the APCI chromatogram, then the possibility of using several ions (and the ratio between the ion intensities) to identify a compound in the APCI trace is examined. Sometimes, the masses will indicate the presence of isobaric compounds that cannot be further resolved and so are reported as isobaric groups, e.g. ‘methylbutanals’, i.e. a potential mixture of 2- and 3-methylbutanal. The technique was applied by Sivasundaram et al. (2003) to investigate the ions and products of the Maillard reaction and has also been applied to coffee (Lindinger et al., 2005), tea (Wright et al., 2007) and cheese crackers (Pozo-Bayon et al., 2008). In the tea example (Wright et al., 2007), six ions in the APCI spectrum could be unequivocally assigned to individual compounds, five ions were associated with isobaric compounds (e.g. 2- and 3-methylbutanal and pentanal) or stereoisomers (e.g., heptenals or heptadienals) and a further four ions were associated with known compounds but with contributions from some unknown impurities. The combination of the two techniques certainly helps in assigning ions to compounds as well as determining the degree of confidence that can be applied to interpretation of the data. The Maillard reaction is especially challenging for direct mass spectrometric monitoring of the reaction. When monitoring acrylamide on-line, an ion trap mass spectrometer was used to fragment the parent ion at m/z 72 to yield a sibling ion at m/z 55 (Cook et al., 2005; Channell et al., 2008). The ratio of the two ions was determined for an acrylamide standard and, if the same ratio was observed in Maillard reactions generating acrylamide, then it was assumed that these ions represented acrylamide alone, without interference from other compounds. A similar approach may be beneficial in other analyses where one major compound is to be monitored. MSÐMS can help resolve some compounds (Jublot et al., 2005) but, generally, the low molecular weight of volatile compounds tends to result in simple, common fragment ions, which are not always useful in compound identification. Similarly, accurate mass determination can help in resolving ions with same nominal mass (e.g. m/z 155 in coffee), but where accurate mass analysis shows the presence of two ions P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 279

at m/z 155.0594 and m/z 155.1436 (unpublished data from our laboratory). However, the number of compounds that can be resolved in this way is limited. To assist in identiﬁcation of intermediates and end products from the Maillard formation of acrylamide, Channell et al. (2008) adopted the labelling procedure of Granvogl and Schieberle (2006), which used 15N and 13C labels in the reducing sugar and asparagine reactants. Potential ions corresponding to key intermediates were monitored with and without labelled reactants, and the change in mass of the ion was checked against the published chemical pathways to determine whether the change in mass was in accordance with the changes expected. By this means, some ions were monitored in the knowledge that they represented speciﬁc intermediates.

10.4.7 Summary There are a range of techniques and equipment that can be used for on-line monitoring of volatile compounds. Speed, sensitivity, interference and identification are all issues that need to be taken into consideration depending on the application, as discussed in Section 10.6. Apart from these general issues, each technique has its strengths and weaknesses, and the choice of a particular method depends on the application and the availability of equipment. Some equipment is purpose built for the task (e.g. the Ionicon PTR-MS system), and the design incorporates the necessary adaptations to cope with the relatively low mass ions produced. There are few volatile compounds with a molecular weight above 250 Da, and therefore, ion transmission and detection can be optimised for a mass range of 0Ð500 amu, which is more than adequate for volatile analysis. Adapting mass spectrometers designed for liquid chromatography to gas-phase analysis requires modification to the interface region to reduce the dead volumes and provide a heated transfer line to avoid condensation. Since liquid chromatography deals with non-volatile compounds with molecular weights up to several thousand Da, both ion transmission and detection are suboptimal for lower molecular weight volatile compounds. The original Micromass MS-Nose machines replaced some of the ion optics to address the former problem, and by operating the mass detector between 40 and 250 amu, the mass range could be adapted for most volatile compounds. However, these potential factors need to be considered when adapting existing machines or designing new machines. Custom-built mass spectrometers with a variety of ion sources and mass detectors can be ordered from specialist companies such as Kore (Ely, UK), while other suppliers such as MKS (San Jose, California), Hiden (Warrington, UK), GSG (D-76646 Bruchsal, Germany) and KR Analytical (Sandbach, Cheshire, UK) can supply off-the-shelf gas analysers/mass spectrometers, or customise them to specific applications.

10.5 ANALYSIS OF TASTANTS USING DIRECT MASS SPECTROMETRY

As mentioned previously (Section 10.2.6), continuous sampling of tastants from the tongue cannot yet be accomplished, so samples are taken at speciﬁc time points and analysed off-line. However, the result is a large number of samples (50Ð200 depending on the experiment), and rapid analyses to quantify a limited number of known tastants are required. Direct liquid- phase mass spectrometry has been developed to provide the rapid analysis needed (Taylor and Linforth, 2003). Basically, the saliva sample is taken from the tongue using a cosmetic cotton P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

280 Food Flavour Technology

bud and solubilised in methanol/water. An aliquot of this solution (20 ␮L) is then injected through a piece of fused silica, directly into the source of a mass spectrometer operating either in the liquid APCI or in the electrospray configuration. The mass detector is set in the selected ion mode to monitor the ions of interest, and the amount present in saliva can be back calculated after calibration with authentic standards. Electrospray analysis suffers from sodium adducts, which are formed between target analytes and sodium ions from saliva. These adduct ions complicate the spectra as well as interfering with quantification. By using a short desalt- ing column, prior to ionisation, cleaner spectra were obtained and better quantification was achieved (Davidson et al., 2000). The release and persistence of a wide range of tastants have been successfully monitored using this technique. In beer, the starch oligomers can be seen up to a degree of polymerisation of around 30, while aspartame can be monitored well below its taste threshold. The common organic acids, bulk sweeteners and bitter compounds such as hop acids can also be monitored. The traces from this type of sampling and analysis give crude release curves with data points about every 5Ð10 seconds depending on the frequency and replication of sampling (Davidson et al., 2000). Although aroma compounds have been the focus for most flavour analyses, the recognition that flavour is a multi-modal perception has reinvigorated work to monitor taste as well as the other modalities (texture, viscosity, colour, etc.). Chapter 12 describes brain imaging to study the interaction of the taste and aroma modalities and takes the analysis to a higher level along the flavour perception pathway (see Section 12.4.3). To summarise this section, it is obvious that direct analysis of tastants meets the sensitivity, identification and interference criteria but not the speed criterion. Whether this is a crucial omission is not at all clear. Although reaction time to tastants is in the millisecond range, experiments where salt was pulsed into the mouth over cycles of a few seconds produced con- flicting reports of perceived saltiness. The study from our laboratory showed no sensory effect of pulsing, with perception well correlated with total amount delivered (Morris-Martinet et al., 2009), while Busch et al. (2009), using a very similar protocol, did find an effect of pulsing on saltiness perception. In mint-flavoured chewing gum, it is well documented that the presence of a sweetener over the prolonged time that gum is chewed is essential in enhancing the mint flavour (Davidson et al., 1999) and this seems to be the basis for some long-lasting chewing gums such as the Cadbury-Adams product, Stride. Further fundamental investigations are needed to determine whether timing of tastant delivery affects overall flavour perception. The current methods for monitoring tastants in vivo are rather crude, and more sophisticated methods need to be developed so that taste release from difficult matrices such as chocolate can be determined.

10.6 APPLICATIONS

The previous sections outlined the scientific background to flavour monitoring and the following sections give some published examples of their applications. The techniques are highly relevant to industry, but for obvious reasons of confidentiality, publications from flavour and food companies are relatively rare. However, the contribution of direct mass spectrometry to research at Nestle was reviewed at the 2007 PTR-MS meeting (Lindinger et al., 2007) and the Nestle team has published some other general reviews of mass spectrometry within the company (Fay et al., 2001; Lindinger et al., 2008). Although direct mass spectrometry was originally developed to measure aroma release during eating, it has found several other uses P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 281

since it became readily available in the late 1990s, and some applications are described in the following sections.

10.6.1 Breath-by-breath analysis One of the major attractions of breath-by-breath analysis of aroma release was to use it in the formulation of flavours and give flavourists another tool with which to create flavours that perform well in a wide variety of food matrices. Typically, flavourists create a flavour from various raw materials in a simple solution (Chapter 1, Section 1.8). However, applying that flavour to foods ranging from milk shakes to ice cream to biscuits requires reformulation for each application with the result that flavour houses carry up to a hundred different formulations for popular flavours such as strawberry, each formulation custom-designed to deliver excellent flavour in a specific food product. Formulating these flavours is a time- consuming and difficult task but is increasingly important as food manufacturers comply with the need to reduce the fat, sugar and salt content of their products, while still delivering food that tastes as good as the original product. The hypothesis was that if the direct mass spectrometry techniques could measure the profile of an excellent flavour in vivo, then this represented the ‘gold standard’ flavour profile that needed to be delivered from all other products to recreate the flavour excellence. New products could be formulated, consumed by a panel and the flavour profile measured in vivo; this could then be compared against the ‘gold standard’ profile. The comparison would show, not only how close the two profiles were, but indicate which specific compounds needed to be adjusted in the formulation to achieve ‘gold standard’ flavour delivery. As with all hypotheses, the concept is simple, but the practice is a little more difficult and the confounding factors are now addressed. From a theoretical point of view, aroma release from foods during eating is governed by sound physicalÐchemical principles, involving mass transfer of volatile and non-volatile flavours from the food to the flavour receptors. However, depending on the food matrix consumed, many other factors (mastication, breathing rate, swallow time) are involved as described in Chapter 7. Early work demonstrated that individuals eating food produced quite reproducible aroma release profiles but the difference between individuals was large (Hollowood et al., 2005). This cast doubt on the ability of direct mass spectrometry techniques to give the necessary information so that flavours could be reliably reformulated. A seminal experiment tackled this question by feeding regular and low-fat milk samples (containing the same amount of flavour) to about 90 people and carrying out simultaneous sensory and in-nose monitoring of a single volatile (Shojaei et al., 2006). There was a clear sensory and instrumental difference between the samples. The amount of volatile in the low-fat milk was then decreased based on the in-nose data, so that both the low-fat and regular fat milks should deliver the same amount of the test volatile compound when consumed. The experiment was then repeated with the regular fat milk sample containing the original volatile content and the low fat-milk sample, formulated to deliver the same amount of volatile to the nose during consumption. Overall, measurement of the in-nose delivery showed the same levels from both milks. The sensory testing showed the panellists could not discriminate the flavour of the two milk samples. The results of this experiment suggested that, although there was considerable variation in aroma release between people, reformulating by using the panel mean values delivered the expected sensory perception in this simple system. It is interesting to note that the intrasample differences were much smaller in vivo than observed by headspace (Shojaei P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

282 Food Flavour Technology

et al., 2006). If reformulation had been based on headspace analysis, the result would have been worse than making no correction. Information from commercial sources suggests that this approach is valid with complex commercial flavours and can help transfer a successful flavour from one matrix to another. While recalculating the amounts moves the reformulation process in the right direction, flavourists still need to rebalance the flavour as it is impossible to measure all flavour components directly and the inherent variation in the population means the data produced should be viewed as a guide, rather than as precise numbers, for reformulation. To address the difficulty of measuring all flavour components so as to provide detailed reformulation information, the idea of using quantitative structureÐproperty relationship to predict the release of any compound is an obvious solution (Linforth et al., 2000; Taylor and Linforth, 2001). The approach (see Chapter 8, Section 8.4.6) is to select about 20 compounds whose physicochemical properties are spread across the volatility and hydrophobicity range and which are amenable to direct mass spectrometry (i.e. give good signal-to-noise ratios and are in parts of the spectrum where they can easily be identified and quantified). The compounds are incorporated into the desired food matrix and the samples are consumed by a panel of people and the aroma release is measured. An empirical model is produced by correlating the release data with a range of physicochemical parameters (e.g. hydrophobicity, volatility, charge separation on the molecule and molecular weight) calculated from software such as EPI Suite. Reasonable fits can be achieved but, since the models are empirical, they cannot be transferred to other food matrices and a new model needs to be constructed for each food matrix. The other major issue with breath-by-breath analysis is how to handle and manage the data produced. Typically, one panellist consumes three replicate samples and generates three flavour profiles. The traces are noisy due to mastication and swallowing and can be smoothed using standard moving point algorithms as found in most spreadsheets. Depending on the purpose of the experiment, the whole curve can be analysed or key parameters can be extracted using the techniques adopted for analysis of sensory timeÐintensity curves such as maximum intensity, time to maximum intensity and area under the curve (see for example Palsgard and Dijksterhuis, 2000). Correlation of instrumental aroma release data with sensory data is a complex process and two types of analysis have been used, one involving correlation of individual aroma release and timeÐintensity curves, the other using the mean values of parameters extracted from all the curves. An example of the former approach (Hollowood et al., 2005) was to fit the instrumental and sensory curves from a large data set, using the power law relationship of Stevens (1969) coupled with the adaptation modification proposed by Overbosch (1986). A reasonable fit was achieved if an improved version of the Overbosch adaptation effect was included in the Steven’s law relationship. A recent example of the second approach involves predicting the sensory quality of coffee from headspace analysis of brewed espresso samples (Lindinger et al., 2008). Here, the mean maximum release intensity was correlated with sensory attributes of coffee using a multivariate statistical approach. Currently, the relationship between aroma release and sensory perception is not fully established. Early work on release from gels suggested that the rate of aroma release was important (Baek et al., 1999), but in the light of more recent experiments on aromaÐtaste interactions (see for example Hort and Hollowood, 2004), the differences observed may have been due to changes in tastant release. This is an area where more research is needed, but experiments need to be well designed to avoid confounding factors as described for the Baek example, and powerful data analysis tools are needed to take into account the potential P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 283

interactions between aromas in producing the flavour perception in the brain (Lindinger et al., 2008). Sampling exhaled air from the nose of panellists during eating can be accomplished using a simple tube in one nostril or by using a tube in each nostril. The Nestle system where panellists wear a set of adapted spectacles carrying the twin nasal tubes is a neat solution to the problem of sampling air consistently while not interfering too much with the ability of the panellists to consume the sample in a realistic way. Health and safety considerations need to be observed as a sealed connection between the panellist’s nasal airways, and the vacuum in the mass spectrometer must be avoided. Given the flow rates into the source (14Ð 200 mL/minute) and the relatively low pressures involved, this is not a practical problem but can raise potential issues with ethics boards. In Section 10.2.1, the theoretical sampling rate needed to observe the detail of aroma release is discussed. The effect can be clearly seen in the experimental data presented in Fig. 10.2. The traces show the release of ethyl butyrate, which produces a protonated molecular ion at m/z 117, during chewing of a fruit-flavoured chewing gum. The data were obtained using a Kore PTR-TOF-MS, which recorded each individual ion arrival time. Using the TOF software, the raw data were processed to construct release traces equivalent to 62.5, 250 and 1000 milliseconds sampling. Figure 10.2 clearly shows that fast sampling of data improves the detail of the aroma release traces. The process simply adds up the ion count for particular intervals, and it is this mechanism that changes the scale of the ion count on the three traces (Fig. 10.2). For examination of fast-moving processes, such as monitoring the airflow during swallowing (Hodgson et al., 2003), fast sampling is essential to obtain accurate and detailed data. For breath-by-breath analysis to measure aroma release in vivo, sampling rates between 10 milliseconds and 2 seconds have been reported but it is unclear what effect sampling rate has on issues such as reproducibility and quantification of aroma release.

10.6.2 Flavour reformulation in reduced fat foods The general principles of flavour reformulation are presented in Section 10.6.1. Since fat acts as a reservoir of aroma, it has a major effect on aroma release (and perception) when the levels are reduced past certain critical levels. Brauss et al. (1999) studied aroma release from regular and low-fat yoghurts with the same aroma content and showed how the aroma profiles changed in terms of their maximum intensity and shape of release. Decreasing fat content in yoghurt caused an increase in the maximum intensity of release for hydrophobic compounds as well as a more rapid decrease after maximum aroma intensity. Using the data from the aroma release curves, the aroma content could be adjusted to address the maximum release intensity, but restoring the shape of the release profile required an additional mechanism. Kant et al. (2004) hypothesised that an alternative flavour reservoir, such as ␤-cyclodextrin, could be a solution to both issues and demonstrated that this was the case, using aroma release measurements and sensory data. This experiment established the principle of flavour reservoirs, but ␤-cyclodextrin is not a suitable food ingredient and alternatives are needed. This basic approach has been applied to a wide range of foods as shown in Table 10.2 to help in flavour reformulation. The subject of retaining quality in low-fat foods has been reviewed recently (Hort and Cook, 2007). P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

Signal f or Mass 117

10000

80 00

60 00

40 00

20 00

0 10 20 30 40 50 1 second integration Time in seconds

Signal f or Mass 117

30 00

20 00

10 00

0 10 20 30 40 50 60 0.25 second integration Time in seconds

Signal f or Mass 117

1200

1000

800

600

400

200

0 10 20 30 40 50 60 0.0625 second integration Time in seconds

Fig. 10.2 Breath-by-breath traces for ethyl butyrate at different sampling frequencies showing how detail is lost as sampling time increases. The traces are reconstructed from a Kore PTR-MS ﬁle, which was obtained at maximum scan rate and then the data were processed to yield traces equivalent to 1000 milliseconds (top trace), 250 milliseconds (middle), and 62.5 milliseconds (bottom) sampling. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 285

Table 10.2 Publications describing the effect of fat on aroma release in foods.

Food product Reference Salad dressing Charles et al. (2000) Frankfurters Chevance and Farmer (1999) Ice cream Chung (2004) Cheese Delahunty et al. (1996a,b) Dairy dessert Gonzalez-Tomas et al. (2008) Emulsions Malone et al. (2003), Weel et al. (2004) Custard Martuscelli et al. (2008) Milk Roberts et al. (2003), Miettinen et al. (2004), Shojaei et al. (2006) General Shamil et al. (1992), Plug and Haring (1993), Taylor et al. (2008)

10.6.3 Flavour release in viscous foods Measurement of aroma release has also been applied to study the well-documented phenomenon that increasing the viscosity of a food (or a flavoured model system) decreases the perceived flavour intensity. Baines and Morris (1989) studied the effects as a function of thickener concentration and proposed that the thickener content could affect the aroma release, the tastant release or both. Measurement of aroma release when samples of different viscosity were consumed showed no significant differences in aroma release (Cook et al., 2002; Hollowood et al., 2002; Cook et al., 2003a,b). These findings were supported by data on gel systems that form different viscosities in mouth during eating (Weel et al., 2002). Again, there was no significant change in aroma release but marked changes in sensory perception of the flavour from the gels. The emphasis therefore turned to measuring the effect of viscosity on tastant release, but the difficulties in measuring this effect in vivo have hampered research and in vitro methods have been applied (Ferry et al., 2006). These experiments illustrate how on-line monitoring of aroma release has been used to understand the nature of tasteÐaromaÐviscosity interactions in model systems and in real foods (see for example Burseg et al., 2009).

10.6.4 Measuring aroma release in ethanolic beverages Ethanolic beverages such as beers, wines and spirits represent a serious analytical challenge because the ethanol content far exceeds the volatile content and is likely to interfere with ionisation of the aroma compounds. Initial experiments studying the reagent ions present in the headspace of ethanolÐwater mixtures showed that water cluster ions were dominant at ethanol concentrations up to 4%, but above this level, ethanol cluster ions became dominant. Monitoring aroma compounds reliably across the ethanol levels was not possible. To overcome this problem, Aznar et al. introduced ethanol into the make-up gas of the APCI system to maintain a stable ethanol concentration in the source and thus stabilise the ionisation process (Aznar et al., 2004). With this modiﬁcation in place, the release of aroma from ethanolic solutions was studied using the dynamic headspace dilution technique. This showed that, although ethanol tended to reduce the gas-phase aroma concentration at equilibrium due to a changed partition value, the concentration in the gas phase during dynamic headspace dilution was greater for some compounds than for the corresponding water control (Tsachaki P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

286 Food Flavour Technology

et al., 2005; Tsachaki et al., 2008). This behaviour can be attributed to the Marangoni effect, a sequence of events that starts with ethanol evaporation from the airÐliquid interface. This causes changes in interfacial tension, which sets up convection currents in the bulk phase and effectively stirs the bulk phase, therefore avoiding aroma concentration gradients in the liquid phase (Tsachaki et al., 2008). By developing this analysis, the dynamics of aroma release from ethanolic solutions were determined. However, this behaviour is related to the bulk-phase release of aroma from ethanolic beverages and therefore applies to the situation prior to drinking, for instance, the headspace above a glass of wine or spirits. It will not relate to aroma release in vivo where release occurs from thin films of liquid spread around the mouth lining and where the bulk phase does not really exist. When using ethanol as a proton transfer reagent, it is important to understand that the ethanol clusters have different proton transfer efficiencies and thus it is essential to keep the proportions of the clusters constant. Ethanol has also a higher proton affinity than water (Table 10.1) and thus some compounds will not be ionised by ethanol.

10.6.5 Monitoring flavour generation on-line The direct mass spectrometry techniques have also been applied to monitor thermal generation of flavours either from model systems or from real foods. Volatile products of the Maillard reaction were monitored using APCI (Turner et al., 2002a), and then the physical state of the solid reactant matrix was monitored using physical techniques such as reflectance Fourier transformed infrared (FTIR) (Turner et al., 2002b; Sivasundaram et al., 2003). A series of publications from the Nestle research laboratories documented the work on coffee flavour generation during roasting (Yeretzian et al., 2000, 2002; Dorfner et al., 2003). Monitoring specific chemical reactions in coffee using on-line direct mass spectrometry has also been reported (Muller et al., 2006). Acrylamide formation has also been studied using on-line techniques (Pollien et al., 2003; Cook and Taylor, 2005; Channell et al., 2008) as have the formation of food contaminants (furan and methyl furan; Mark et al., 2006) and aromas (Channell and Taylor, 2005). The analytical challenges in these applications relate to efficient sampling of the reaction products, the potential for suppression of some ions (as other ions with higher proton affinities are formed and steal charge) and the assignment of ions to compounds so that the data can be interpreted correctly. Sampling of products from Maillard reactions is achieved by flowing gas over or through the reactor and interfacing the gas flow to the ion source. Difficulties can be encountered as some components, particularly polar molecules such as maltol, have a tendency to stick to the transfer lines. The result is that the compound is not monitored at its time of formation and it may release slowly and interfere with subsequent runs if not fully removed. The use of higher temperatures in the transfer lines (150Ð180◦C) and a high gas flow rate seem to remove the effect. Suppression is an issue as the compounds produced from a Maillard reaction cannot be fully predicted and their relative amounts depend on the reactant proportions and the processing conditions. APCI has limited ionising ‘power’ and is susceptible to suppression unless it is operating in a suitable window (see Section 10.4.5). In our laboratory, we have used 3-methylbutanal and 2,5-dimethylpyrazine to assess when suppression occurs. The Kore PTR-TOF-MS, which was designed with a high primary ion count of 500 000 counts/second to minimise suppression, was tested for suppression using this system. A linear calibration curve for 2,5-dimethylpyrazine over the P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 287

range 0Ð1000 ppbv was obtained first, and then the calibration was repeated in the presence of different concentrations of 3-methylbutanal. No change in the calibration curves was seen until the 3-methylbutanal content was above 100 ppmv. Thus, 3-methylbutanal only affected 2,5-dimethylpyrazine analysis when present in a 100× excess. The experiment was repeated by obtaining a calibration curve for 3-methylbutanal (0Ð1000 ppbv) and then determining the effect of 2,5-dimethylpyrazine at concentrations ranging from 0 to 1000 ppmv. Again, only concentrations of 2,5-dimethylpyrazine above 100 ppmv (equivalent to a 100× excess) changed the calibration data (unpublished data). These simple experiments can provide some estimates of suppression and also indicate when data can be interpreted with confidence and when it should be treated with caution. Identification of the many products and intermediates in the Maillard reaction is a major task, even by GC-MS, but assigning ions to compounds is essential to use the results from on-line monitoring. The combined GC-MS technique and the stable isotope labelling experiments described in Section 10.4.6 can certainly assist in identification. On-line monitoring can also be used to study the effects of processing conditions on flavour generation in a time-effective way. Conventional studies heat the reaction mixture for a set time, extract and analyse it by GC-MS, and then repeat the procedure for the other time points. In contrast, on-line analysis monitors continuously, although its ability to identify all the compounds present is limited. For this reason, on-line monitoring is useful for screening a range of reactant compositions or processing conditions to determine where and when significant changes occur in the product profile. The samples of interest can then be further studied using GC-MS analysis to provide the full information about composition. We used this approach in a study on acrylamide formation where more than 30 different reactant compositions were studied over a 3-day period, allowing us to rapidly identify the effects of composition on acrylamide formation (Channell et al., 2008). On-line monitoring can also be used to study the effect of secondary reactants as a reaction is underway. Using the Strecker reaction as a model, the effect of humidity on a particular part of the reaction sequence was studied (Taylor, 2005; Cook et al., 2006). In a model system containing l-valine and glucose, three key Maillard intermediates can be identified from mechanistic literature. Figure 10.3 shows the reaction of valine with a glucose breakdown product, pyruvaldehyde, for clarity. The three intermediates were monitored as the reactor was heated from 130 to 170◦C, with the carrier gas stream pulsed to contain three sequential levels of water (1 → 0 → 5 → 1 ␮L/minute) delivered by total evaporation into a 20-mL/minute reactor gas flow. As anticipated, accelerated thermal generation occurred as the reaction film was heated from 130 to 170◦C whereas the effect of humidity was consistent with mechanistic projections. The dehydration reaction necessary to form Intermediate 1 was strongly facilitated by low moisture conditions and suppressed by elevated humidity, whilst the imine hydrolysis reaction involved in the formation of Intermediate 3 was suppressed by anhydrous conditions and enhanced at elevated water levels. Intermediate 2 produced by the decarboxylation of Intermediate 1 appeared largely unaffected by humidity level and was simply generated on heating. The role of water in the Strecker reaction was further investigated using leucine and glucose, which form 3-methylbutanal. The reaction was started with a constant water level of 1 ␮L/minute (Fig. 10.4), and 3-methylbutanal was formed and monitored by the MH+ ion at m/z 87. When 18O-labelled water was introduced, the ion at m/z 87 was replaced by anewionatm/z 89 and the increase of two mass units showed that water was a reactant at that level. A return to standard water led to the reappearance of the ion at m/z 87. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

288 Food Flavour Technology

H H Dehydration H CO H CO2 H O 2 O CO2H OH H2N N NH2 H N + H O O OH O Decarboxylation Hydrolysis 2-Methylpropanal Aminoacetone Valine Pyruvaldehyde Mol. Wt.: 189 Mol. Wt.: 171 H2O Mol. Wt.: 72 Mol. Wt.: 73 + H Mol. Wt.: 117 Mol. Wt.: 72 MH =172 MH+=74 Intermediate 1 N Intermediate 3 H OH

H2O H

H O

Mol. Wt.: 127 MH+=128 Hydrolysis O Intermediate 2 + H NH2 O 2-Methylpropan-1-amine Mol. Wt.: 73

Fig. 10.3 Strecker reaction pathway between valine and pyruvaldehyde showing the three intermediate compounds monitored by on-line APCI-MS.

In these examples, on-line monitoring can be used to probe the detailed reaction sequence and understand the effect of gaseous reactants such as water or hydrogen sulﬁde as well as the interaction between time, temperature and water content. When interpreting the data, it should be borne in mind that only compounds in the gas phase are measured whereas it is often the amounts remaining in the solid (food) phase that are of interest and deliver ﬂavour

4.0E+07

3.5E+07 m/z 87

89) 3.0E+07 m/z 89 m/z 2.5E+07 87and 2.0E+07 m/z

1.5E+07

1.0E+07 Ion intensity ( intensity Ion 5.0E+06

0.0E+00 00.511.522.533.544.55 Time (minute)

18 ◦ Fig. 10.4 Effect of a pulse of H2 O on the formation of 3-methylbutanal at 130 C from a system containing leucine and glucose. Water was introduced at 1 ␮L/minute and vapourised into the carrier gas 18 ﬂow up to 2.5 minutes, H2 O was introduced from 2.5 to 4 minutes, and then replaced by standard water. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 289

to the consumer. Separate experiments are needed to establish the connection between gas- phase and solid-phase contents. For acrylamide, it was found that gas-phase contents were directly related to amounts in the solid phase and a calibration curve could be constructed although it was temperature dependent (Cook and Taylor, 2005). On-line mass spectrometry for following chemical reactions is a relatively new technique although there are some published examples in the literature on non-food applications (see for example Brum and Dell’Orco, 1998; Dell’Orco et al., 1999; Brum et al., 2001; Mark et al., 2006). It is a powerful tool for studying real foods as well as the dynamics of individual reaction steps in the Maillard reaction. Improvements to identiﬁcation and to data analysis are needed to make the technique easier to apply and more user-friendly.

10.6.6 Rapid headspace profiling of fruits and vegetables The role of on-line monitoring for screening reactions is mentioned in Section 10.6.5, and screening is also useful in situations where the sheer number of samples and/or the rapidly changing metabolism of the samples rules out conventional GC-MS. Monitoring fruit quality via the volatile profiles in the headspace is one such situation and can be used to analyse individual fruits as well as batches of fruits. This is of particular relevance in genetic studies where the genetics of individual fruits needs to be correlated with individual volatile profiles. Biasioli et al. (2006) carried out a 3-year study of strawberries and were able to identify strawberry cultivars by their PTR-MS fingerprint obtained from strawberry headspace. The results were consistent over the 3 years, showing that it was a robust identification and not affected by the different environmental conditions experienced by the crops in the 3 different years. In this trial, 30 strawberries were analysed. Further work extended the study to other strawberry samples and developed both the hardware and software requirements to carry out such studies in fruits and vegetables (Carbone et al., 2006; Granitto et al., 2007). Tomato volatiles have also been studied by on-line mass spectrometry. Tomato flavour is partly formed during ripening and partly formed through the lipoxygenase pathway when the fruit is eaten. Therefore, headspace sampling of the intact fruit is inappropriate and a controlled maceration device was developed to analyse tomatoes (Boukobza et al., 2001). The analysis takes 3 minutes and gives information on both the volatiles and the enzyme activity of the lipoxygenase pathway. Up to 100 tomatoes have been analysed in 1 day in our laboratory, and the simplicity of the analysis means it can be used as an ‘open-access’ facility so that tomato fruits are analysed within a certain time of picking to maintain consistency in their analysis. When using introgression lines to identify quantitative trait loci, it is necessary to screen several hundred fruits that may ripen over a period of weeks and so daily analysis during the harvest period is needed. Apple quantitative trait loci have also been investigated using PTR-MS (Zini et al., 2005). Other research using on-line methods has studied the general lipoxygenase pathways to flavour (Dunphy et al., 2000), while Grab and Gfeller (2000) related the tomato aroma profile at different times with temporal sensory attributes. On-line monitoring as a high-throughput quality control procedure for fruits and vegetables is also of interest, and the general approach to successful fingerprinting for quality control has been explained (Granitto et al., 2008) and applied to a variety of berry fruits (strawberry, raspberry, blackberry as well as redcurrants; Boschetti et al., 1999). The concept of analysing individual, high-value fruits to predict time to ripeness is an attractive idea but requires suitable volatile markers, rapid analyses to achieve high throughput and a portable, rugged and cheap mass spectrometer to survive the environment in packing sheds. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

290 Food Flavour Technology

10.7 FUTURE

On-line mass spectrometry has been available since the mid-1990s and, for flavour analysis, is considered a mature technique. The number of machines in use for flavour analysis is difficult to gauge but must be between 20 and 100 units spread throughout the food and flavour industry as well as in universities and research institutes. On the wish list for the future, increased sensitivity is always uppermost in the mind of analysts. Pushing sensitivity to the next level will be difficult and will probably combine hardware developments with increasingly sophisticated software, e.g. to reduce background signal. TOF mass analysers are now available on direct mass spectrometry machines and give potential for further hyphenation (e.g. with ion traps) to make best use of their resolving power, irrespective of the number of ions. Currently, TOF analysers are working at about 10% duty cycle (i.e. they are only detecting ions for 10% of the time), and some research groups are trying to apply Hadamard transform procedures (Brock et al., 1998) to TOF analysers to improve the situation (50% duty cycle is apparently achievable). Miniaturisation of mass spectrometers to make them more portable is also on the wish list. Since the amu range needed for volatile analysis is low, smaller mass analysers can be used, which require smaller vacuum pumps. Some prototype devices running off 12 V power supplies are undergoing bench testing and will hopefully be commercially available in the near future. More powerful data analysis capability is also needed, first to automate preliminary analysis of the data and then to correlate the on-line instrumental data with sensory data. The Nestle approach described by Lindinger et al. (2008) shows the way to go with instrumentalÐsensory analysis as we are still unable to answer the questions posed in Section 10.1, namely: r r Why do particular mixtures of compounds taste pleasant? Why do quite small changes in some flavour components cause significant changes in sensory scores?

From the information presented in this chapter, the future direction for on-line monitoring to improve our understanding of flavour generation, flavour release and flavour perception requires cross-disciplinary teams. However, to fully understand flavour perception, there is also a need to study flavour perception at a higher level (i.e. postreceptor) as this too may help us identify the key factors driving flavour perception. When we come to try and relate flavour compositions to perception, we need more scientific disciplines, so the ideal team might contain capabilities in analysis, sensory, data processing, brain imaging, neural processing and psychology. It is perhaps not surprising that a ‘multi-modal’ research team is needed to study the multi-modal phenomenon that is flavour perception.

REFERENCES

Ammann, C., Brunner, A., Spirig, C. and Neftel, A. (2006) Technical note: water vapour concentration and ﬂux measurements with PTR-MS. Atmos. Chem. Phys. 6, 4643Ð4651. Aprea, E., Biasioli, F., Mark, T.D. and Gasperi, F. (2007) PTR-MS study of esters in water and water/ethanol solutions: fragmentation patterns and partition coefﬁcients. Int. J. Mass Spectrom. 262, 114Ð121. Arnold, S.T., Viggiano, A.A. and Morris, R.A. (1998) Rate constants and product branching fractions for the + + reactions of H3O and NO with C-2-C-12 alkanes. J. Phys. Chem. A 102, 8881Ð8887. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 291

Aznar, M., Tsachaki, M., Linforth, R.S.T., Ferreira, V. and Taylor, A.J. (2004) Headspace analysis of volatile organic compounds from ethanolic systems by direct APCI-MS. Int. J. Mass Spectrom. 239, 17Ð25. Baek, I., Linforth, R.S.T., Blake, A. and Taylor, A.J. (1999) Sensory perception is related to the rate of change of volatile concentration in-nose during eating of model gels. Chem. Senses 24, 155Ð160. Baines, Z.V. and Morris, E.R. (1989) Suppression of perceived flavour and taste by food hydrocolloids. In: Food Colloids,Vol.75 (eds R.D. Bee, P. Richmond and J. Mingens), Royal Society of Chemistry, Cambridge, UK, pp. 184Ð192. Benoit, F.M., Davidson, W.R., Lovett, A.M., Nacson, S. and Ngo, A. (1983) Breath analysis by atmospheric- pressure ionization mass-spectrometry. Anal. Chem. 55, 805Ð807. Biasioli, F. (2007) PTR-MS in Food Science and Technology: A Review, Innsbruck University Press, Obergurgl. Biasioli, F., Gasperi, F., Marini, F. and Mark, T.D. (2006) Characterization of strawberry genotypes by PTR-MS spectral fingerprinting: a three-year study. Proceedings of the Vth International Strawberry Symposium, 497Ð500. Blake, R.S., Patel, M., Monks, P.S., Ellis, A.M. Inomata, S. and Tanimoto, H. (2008) Aldehyde and ketone discrimination and quantification using two-stage proton transfer reaction mass spectrometry. Int. J. Mass Spectrom. 278, 15Ð19. Blake, R.S., Wyche, K.P.,Ellis, A.M. and Monks, P.S.(2006) Chemical ionization reaction time-of-flight mass spectrometry: multi-reagent analysis for determination of trace gas composition. Int. J. Mass Spectrom. 254, 85Ð93. Boschetti, A., Biasioli, F., van Opbergen, M., Warneke, C., Jordan, A., Holzinger, R., Prazeller, P., Karl, T., Hansel, A., Lindinger, W. and Iannotta, S. (1999) PTR-MS real time monitoring of the emission of volatile organic compounds during postharvest aging of berryfruit. Postharvest Biol. Technol. 17, 143Ð151. Boukobza, F., Dunphy, P. and Taylor, A.J. (2001) Measurement of lipid oxidation-derived volatiles in fresh tomatoes. Postharvest Biol. Technol. 23, 117Ð131. Brauss, M.S., Linforth, R.S.T., Cayeux, I., Harvey, B. and Taylor, A.J. (1999) Altering the fat content affects flavour release in a model yogurt system. J. Agric. Food Chem. 47, 2055Ð2059. Brock, A., Rodriguez, N. and Zare, R.N. (1998) Hadamard transform time of flight mass spectrometry. Anal. Chem. 70, 3735Ð3741. Bruins, A.P. (1991) Mass spectrometry with ion sources operating at atmospheric pressure. Mass Spectrom. Rev. 10, 53Ð77. Brum, J. and Dell’Orco, P. (1998) On-line mass spectrometry: real-time monitoring and kinetics analysis for the photolysis of idoxifene. Rapid Commun. Mass Spectrom. 12, 741Ð745. Brum, J., Dell’Orco, P., Lapka, S., Muske, K. and Sisko, J. (2001) Monitoring organic reactions with on- line atmospheric pressure ionization mass spectrometry: the hydrolysis of isatin. Rapid Commun. Mass Spectrom. 15, 1548Ð1553. Burseg, K.M.M., Linforth, R.S.T., Hort, J. and Taylor, A.J. (2009) Flavor perception in biscuits; correlating sensory properties with composition, aroma release, and texture. Chemosen. Percept. 2, 70Ð78. Busch, J.H.L.C., Tournier, C., Knoop, J.E., Kooyman, G. and Smit, G. (2009) Temporal contrast of salt delivery in mouth increases salt perception. Chem. Senses 34, 341Ð348. Byrdwell, W.C. (2001) Atmospheric pressure chemical ionization mass spectrometry for analysis of lipids. Lipids 36, 327Ð346. Carbone, F., Mourgues, F., Biasioli, F., Gasperi, F., Mark, T.D., Rosati, C. and Perrotta, G. (2006) Development of molecular and biochemical tools to investigate fruit quality traits in strawberry elite genotypes. Mol. Breeding 18, 127Ð142. Channell, G.A. and Taylor, A.J. (2005) On line monitoring of the Maillard reaction using a film reactor coupled to ion trap mass spectrometry. In: Process and Reaction Flavors,Vol.905 (eds D.K. Weerasinghe and M.K. Sucan), American Chemical Society, Washington, DC, pp. 181Ð191. Channell, G.A., Wulfert, F. and Taylor, A.J. (2008) Identification and monitoring of intermediates and products in the acrylamide pathway using on-line analysis. J. Agric. Food Chem. 56, 6097Ð6104. Charles, M., Rosselin, V., Beck, L., Sauvageot, F. and Guichard, E. (2000) Flavor release from salad dressings: sensory and physicochemical approaches in relation with the structure. J. Agric. Food Chem. 48, 1810Ð1816. Chevance, F.F.V. and Farmer, L.J. (1999) Release of volatile odor compounds from full-fat and reduced-fat frankfurters. J. Agric. Food Chem. 47, 5161Ð5168. Chung, S.J. (2004) Release of artificial cherry flavor from ice creams varying in fat and fat replacers. J. Sens. Stud. 19, 211Ð236. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

292 Food Flavour Technology

Cook, D.J., Channell, G.A., Abd Ghani, M. and Taylor, A.J. (2006) Thermal flavour generation: insights from mass spectrometric studies. In: Flavour Science: Recent Advances and Trends,Vol.43 (eds W.L.P. Bredie and M.A. Petersen), Elsevier, Amsterdam, pp. 569Ð572. Cook, D.J., Channell, G.A. and Taylor, A.J. (2005) On line monitoring of acrylamide formation. In: Chemistry and Safety of Acrylamide in Food,Vol.561 (eds M. Friedman and D.S. Mottram), Springer, New York, pp. 303Ð316. Cook, D.J., Hollowood, T.A., Linforth, R.S.T. and Taylor, A.J. (2002) Perception of taste intensity in solutions of random coil polysaccharides above and below c*. Food Qual. Prefer. 13, 473Ð480. Cook, D.J., Hollowood, T.A., Linforth, R.S.T. and Taylor, A.J. (2003a) Oral shear stress predicts flavour perception in viscous solutions. Chem. Senses 28, 11Ð23. Cook, D.J., Linforth, R.S.T. and Taylor, A.J. (2003b) Effects of hydrocolloid thickeners on the perception of savory flavors. J. Agric. Food Chem. 51, 3067Ð3072. Cook, D.J. and Taylor, A.J. (2005) On-line MS/MS monitoring of acrylamide generation in potato- and cereal-based systems. J. Agric. Food Chem. 53, 8926Ð8933. Coren, S., Ward, L.M. and Enns, J.T. (1999) Sensation and Perception, Harcourt Brace, Fort Worth, TX. Dalton, P., Doolittle, N., Nagata, H. and Breslin, P.A.S. (2000) The merging of the senses: integration of subthreshold taste and smell. Nat. Neurosci. 3, 431Ð432. Darling, D.F., Williams, D. and Yendle, P. (1986) Physico-chemical interactions involved in aroma transport processes from solution. In: Interactions of Food Components (eds G.G. Birch and M.G. Lindley), Elsevier Applied Science, London, pp. 165Ð188. Davidson, J.M., Hollowood, T.A., Linforth, R.S.T. and Taylor, A.J. (1999) The effect of sucrose on the perceived flavour intensity of chewing gum. J. Agric. Food Chem. 47, 4336Ð4340. Davidson, J.M., Linforth, R.S.T., Hollowood, T.A. and Taylor, A.J. (2000) Release of non-volatile flavor components in vivo. In: Flavor Release:Linking Experiments, Theory and Reality, Vol. 763 (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, D.C., pp. 99Ð111. Deibler, K.D., Acree, T.E. and Lavin, E.H. (1998) Aroma analysis of coffee brew by gas chromatography- olfactometry. Dev. Food Sci 40, 69Ð78. Delahunty, C.M., Piggott, J.R., Conner, J.M. and Paterson, A. (1996a) Comparative volatile release from traditional and reduced-fat cheddar cheeses upon mastication in the mouth. Ital. J. Food Sci. 8, 89Ð98. Delahunty, C.M., Piggott, J.R., Conner, J.M. and Paterson, A. (1996b) Comparison of dynamic flavour release from hard cheeses and analysis of headspace volatiles from the mouth with flavour perception during consumption. J. Sci. Food Agric. 71, 273Ð281. Dell’Orco, P., Brum, J., Matsuoka, R., Badlani, M. and Muske, K. (1999) Monitoring process-scale reactions using API mass spectrometry. Anal. Chem. 71, 5165Ð5170. Devos, M., Patte, F., Rouault, J., Laffort, P. and Van Gemert, L.J. (1990), Standardized Human Olfactory Thresholds, IRL Press, Oxford. Dorfner, R., Ferge, T., Kettrup, A., Zimmermann, R. and Yeretzian, C. (2003) Real-time monitoring of 4-vinylguaiacol, guaiacol, and phenol during coffee roasting by resonant laser ionization time-of-flight mass spectrometry. J. Agric. Food Chem. 51, 5768Ð5773. Dunphy, P., Boukobza, F., Chengappa, S., Lanot, A. and Wilkins, J. (2000) Wound response in plants: an orchestrated symphony. In: Flavor Release: Linking Experiments, Theory and Reality (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, D.C., pp. 44Ð57. Elmore, J.S. and Langley, K.R. (1996) Novel vessel for the measurement of dynamic flavor release in real time from liquid foods. J. Agric. Food Chem. 44, 3560Ð3563. Fay, L.B., Yeretzian, C. and Blank, I. (2001) Novel mass spectrometry methods in flavour analysis. Chimia 55, 429Ð434. Fechner, G.T. (1860) Elemente der Psychophysik, Breitkopf and Hartel, Leipzig, Germany. Ferry, A.L., Hort, J., Mitchell, J.R., Cook, D.J., Lagarrigue, S. and Pamies, B.V. (2006) Viscosity and flavour perception: why is starch different from hydrocolloids? Food Hydrocolloid. 20, 855Ð862. Flament, I. (2002) Coffee Flavor Chemistry, Wiley, Chichester, UK. Francis, G.J., Milligan, D.B. and McEwan, M.J. (2007) Gas-phase reactions and rearrangements of alkyl + + + esters with H3O ,NO ,andO2 : a selected ion flow tube study. J. Phys. Chem. A 111, 9670Ð9679. Gonzalez-Tomas, L., Bayarri, S., Taylor, A.J. and Costell, E. (2008) Rheology, flavour release and perception of low-fat dairy desserts. Int. Dairy J. 18, 858Ð866. Grab, W. and Gfeller, H. (2000) Flavorspace Ð a technology for the measurement of fast dynamic changes of flavor release during eating. In: Flavor Release: Linking Experiments, Theory and Reality (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 33Ð43. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 293

Granitto, P.M., Biasioli, F., Aprea, E., Mott, D., Furlanello, C., Mark, T.D. and Gasperi, F. (2007) Rapid and non-destructive identification of strawberry cultivars by direct PTR-MS headspace analysis and data mining techniques. Sens. Actuators, B 121, 379Ð385. Granitto, P.M., Biasioli, F., Furlanello, C. and Gasperi, F. (2008) Efficient feature selection for PTR-MS fingerprinting of agroindustrial products. ICANN 5164, 42Ð51. Granvogl, M. and Schieberle, P.(2006) Thermally generated 3-aminopropionamide as a transient intermediate in the formation of acrylamide. J. Agric. Food Chem. 54, 5933Ð5938. Hansel, A. and Mark,¬ T.D. (2007) 3rd International Conference on Proton Transfer Reaction Mass Spectrom- etry and Its Applications. Innsbruck University Press, Obergurgl. Hodgson, M., Linforth, R.S.T. and Taylor, A.J. (2003) Simultaneous real time measurements of mastication, swallowing, nasal airflow and aroma release. J. Agric. Food Chem. 51, 5052Ð5057. Hollowood, T.A., Chang, K.-M., Hort, J., Langridge, J.P., Fearn, T. and Taylor, A.J. (2005) Temporal flavour perception in vivo. In: State-of-the-Art in Flavour Chemistry and Biology (eds T. Hofmann, M. Rothe and P. Schieberle), Deutsche Forschungsanstalt fur¬ Lebensmittelchemie, Garching, Germany, pp. 160Ð168. Hollowood, T.A., Linforth, R.S.T. and Taylor, A.J. (2002) The effect of viscosity on the perception of flavour. Chem. Senses 27, 583Ð591. Hort, J. and Cook, D.J. (2007) Formulating low fat foods: the challenge of retaining flavour quality. In: Modifying Flavour in Food (eds A.J. Taylor and J. Hort), Woodhead, Cambridge, UK, pp. 131Ð145. Hort, J. and Hollowood, T.A. (2004) Controlled continuous flow delivery system for investigating taste-aroma interactions. J. Agric. Food Chem. 52, 4834Ð4843. Hunter, E.P.L. and Lias, S.G. (1998) Evaluated gas phase basicities and proton affinities of molecules: an update. J. Phys. Chem. Ref. Data 27, 413Ð656. Inomata, S. and Tanimoto, H. (2008) Differentiation of isomeric compounds by two-stage proton transfer reaction time-of-flight mass spectrometry. J. Am. Soc. Mass Spectrom. 19, 325Ð331. Jin, S.P., Li, J.Q., Han, H.Y., Wang, H.M., Chu, Y.N. and Zhou, S.K. (2007) Proton transfer reaction mass spectrometry for online detection of trace volatile organic compounds. Prog. Chem. 19, 996Ð1006. Jublot, L., Linforth, R.S.T. and Taylor, A.J. (2005) Direct atmospheric pressure chemical ionisation ion trap mass spectrometry for aroma analysis: speed, sensitivity and resolution of isobaric compounds. Int. J. Mass Spectrom. 243, 269Ð277. Kant, A., Linforth, R.S.T., Hort, J. and Taylor, A.J. (2004) Effect of beta cyclodextrin on aroma release and flavor perception. J. Agric. Food Chem. 52, 2028Ð2035. Keck, L., Hoeschen, C. and Oeh, U. (2008) Effects of carbon dioxide in breath gas on proton transfer reaction-mass spectrometry (PTR-MS) measurements. Int. J. Mass Spectrom. 270, 156Ð165. Keck, L., Oeh, U. and Hoeschen, C. (2007) Corrected equation for the concentrations in the drift tube of a proton transfer reaction-mass spectrometer (PTR-MS). Int. J. Mass Spectrom. 264, 92Ð95. Ketkar, S.N., Penn, S.M. and Fite, W.L. (1991) Real-time detection of parts-per-trillion levels of chemical warfare agents in ambient air using atmospheric-pressure ionization tandem quadrupole mass-spectrometry. Anal. Chem. 63, 457Ð459. Kuznicki, J.T. and Tumer, L. (1988) Temporal dissociation of taste mixture components. Chem. Senses 13, 45Ð62. Laing, D.G. and Macleod, P. (1992) Reaction time for the recognition of odor quality. Chem. Senses 17, 337Ð346. Lindinger, C., Labbe, D., Pollien, P., Rytz, A., Juillerat, M.A., Yeretzian, C. and Blank, I. (2008) When machine tastes coffee: instrumental approach to predict the sensory profile of Espresso coffee. Anal. Chem. 80(5), 1574Ð1581. Lindinger, C., Pollien, P., Ali, S., Spadone, J.-C. and Robert, F. (2007) General Overview: Applications of PTR-MS in Food Research at Nestle´. 3rd International Conference on Proton Transfer Reaction Mass Spectrometry and Its Applications, Innsbruck University press, Obergurgl. Lindinger, C., Pollien, P., Ali, S., Yeretzian, C., Blank, I. and Mark, T. (2005) Unambiguous identification of volatile organic compounds by proton-transfer reaction mass spectrometry coupled with GC/MS. Anal. Chem. 77, 4117Ð4124. Lindinger, W., Hansel, A. and Jordan, A. (1998) On-line monitoring of volatile organic compounds at pptv levels by means of proton-transfer-reaction mass spectrometry (PTR-MS) Ð medical applications, food control and environmental research. Int. J. Mass Spectrom. 173, 191Ð241. Linforth, R.S.T., Friel, E.N. and Taylor, A.J. (2000) Modeling flavor release from foods using physicochemical parameters. In: Flavor Release: Linking Experiments, Theory and Reality,Vol.763 (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 166Ð178. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

294 Food Flavour Technology

Linforth, R.S.T. and Taylor, A.J. (1998) Apparatus and methods for the analysis of trace constituents of gases. European Patent. EP 0819937A2. Malone, M.E., Appelqvist, I.A.M. and Norton, I.T. (2003) Oral behaviour of food hydrocolloids and emulsions. Part 2. Taste and aroma release. Food Hydrocolloid. 17, 775Ð784. Mark, J., Pollien, P., Lindinger, C., Blank, I. and Mark, T. (2006) Quantitation of furan and methylfuran formed in different precursor systems by proton transfer reaction mass spectrometry. J. Agric. Food Chem. 54, 2786Ð2793. Martuscelli, M., Savary, G., Pittia, P.and Cayot, N. (2008) Vapourpartition of aroma compounds in strawberry flavoured custard cream and effect of fat content. Food Chem. 108, 1200Ð1207. Miettinen, S.M., Hyvonen, L., Linforth, R.S.T., Taylor, A.J. and Tuorila, H. (2004) Temporal aroma delivery from milk systems containing 0Ð5% added fat, observed by free choice profiling, time intensity, and atmospheric pressure chemical ionization-mass spectrometry techniques. J. Agric. Food Chem. 52, 8111Ð8118. Milligan, D.B., Francis, G.J., Prince, B.J. and McEwan, M.J. (2007) Demonstration of selected ion flow tube MS detection in the parts per trillion range. Anal. Chem. 79, 2537Ð2540. Morris-Martinet, C., Koliandris-Gady, A.-L., Wolf, B., Hort, J. and Taylor, A.J. (2009) Effect of pulsed or continuous delivery of salt on sensory perception over short time intervals. Chemosen. Percept. 2,1Ð8. Muller, C., Lang, R. and Hofmann, T. (2006) Quantitative precursor studies on di- and trihydroxybenzene formation during coffee roasting using ‘in bean’ model experiments and stable isotope dilution analysis. J. Agric. Food Chem. 54, 10086Ð10091. Overbosch, P. (1986) A theoretical-model for perceived intensity in human taste and smell as a function of time. Chem. Senses 11, 315Ð329. Palsgard, E. and Dijksterhuis, G. (2000) The sensory perception of flavor release as a function of texture and time: a time intensity study using flavored gels. J. Sens. Stud. 15, 347Ð359. Plug, H. and Haring, P.G.M. (1993) The role of ingredient-flavour interactions in the development of fat-free foods. Trends Food Sci. Technol. 4, 150Ð152. Pollien, P., Lindinger, C., Yeretzian, C. and Blank, I. (2003) Proton transfer reaction mass spectrometry, a tool for on-line monitoring of acrylamide formation in the headspace of Maillard reaction systems and processed food. Anal. Chem. 75, 5488Ð5494. Pozo-Bayon, M.A., Schirle-Keller, J.P. and Reineccius, G.A. (2008) Determining specific food volatiles contributing to PTR-MS ion profiles using GC-EI-MS. J. Agric. Food Chem. 56, 5278Ð5284. Purves, W.K., Heller, C., Sadava, D. and Orians, G.H. (2001) Life, the Science of Biology, Sinauer Associates, Sunderland, MA. Raffaelli, A. (1997) Atmospheric pressure ionization (ISI and APCI). In: Selected Topics and Mass Spec- trometry in the Biomolecular Sciences,Vol.504 (ed. R.M. Caprioli), Kluwer Academic Publishers, Amsterdam, pp. 17Ð31. Reid, W.J. and Wragg, S. (1995) Flavour Release Measurement by in-Breath Mass Spectrometry. I. Devel- opment of Membrane Interface, Leatherhead, UK, pp. 1Ð34. Roberts, D.D., Pollien, P. and Watzke, B. (2003) Experimental and modeling studies showing the effect of lipid type and level on flavor release from milk-based liquid emulsions. J. Agric. Food Chem. 51, 189Ð195. Rychlik, M., Schieberle, P. and Grosch, W. (1998) Compilation of Odor Thresholds, Odor Qualities and Retention Indices of Key Food Odorants, Deutsche Forschungsanstalt fuer Lebensmittelchemie, Garching, Germany. Schmiech, L., Uemura, D. and Hofmann, T. (2008) Reinvestigation of the bitter compounds in carrots (Daucus carota L.) by using a molecular sensory science approach. J. Agric. Food Chem. 56, 10252Ð10260. Semmelroch, P., Laskawy, G., Blank, I. and Grosch, W. (1995) Determination of potent odorants in roasted coffee by stable isotope dilution assays. Flavour Frag. J. 10,1Ð7. Shamil, S., Wyeth, L.J. and Kilcast, D. (1992) Flavour release and perception in reduced fat foods. Food Qual. Prefer. 3, 51Ð60. Sherwood, L. (2006) Fundamentals of Physiology: A Human Perspective, Thomson Brooks/Cole, Belmont, CA. Shojaei, Z.A., Linforth, R.S.T., Hort, J., Hollowood, T.A. and Taylor, A.J. (2006) Measurement and manipulation of aroma delivery allows control of perceived fruit flavour in low and regular fat milks. Int. J. Food Sci. Technol. 41, 1192Ð1196. Sivasundaram, L.R., Farhat, I.A. and Taylor, A.J. (2003) On-line monitoring of the maillard reaction. In: Handbook of Flavor Characterization: Sensory Analysis, Chemistry and Physiology (eds K.D. Deibler and J. Delwiche), Marcel Dekker, New York, pp. 379Ð388. P1: SFK/UKS P2: SFK c10 BLBK221-Taylor/Linforth November 21, 2009 8:4 Printer Name: Yet to Come

On-line monitoring of ﬂavour processes 295

Smith, D. and Spanel, P. (2005) Selected ion flow tube mass spectrometry (SIFT-MS) for on-line trace gas analysis. Mass Spectrom. Rev. 24, 661Ð700. Smith, D., Wang, T.S. and Spanel, P. (2003) A SIFT study of the reactions of H2ONO+ ions with several types of organic molecules. Int. J. Mass Spectrom. 230,1Ð9. Soeting, W.J. and Heidema, J. (1988) A mass-spectrometric method for measuring flavor concentration- time profiles in human breath. Chem. Senses 13, 607Ð617. Springett, M.B., Rozier, V. and Bakker, J. (1999) Use of fiber interface direct mass spectrometry for the determination of volatile flavor release from model food systems. J. Agric. Food Chem. 47, 1125Ð1131. Stevens, S.S. (1969) Sensory scales of taste intensity. Percept. Psychophys. 6, 302Ð308. Sunner, J., Nicol, G. and Kebarle, P. (1988) Factors determining relative sensitivity of analytes in positive mode atmospheric-pressure ionization mass-spectrometry. Anal. Chem. 60, 1300Ð1307. Taylor, A.J. (2002) Release and transport of flavours in vivo: physico chemical, physiological and perceptual considerations. Comp. Rev. Food Sci. 1, 45Ð57. Taylor, A.J. (2005) Real-time analysis of flavour release and perception. Foods & Food Ingred. J. Jpn 210, 859Ð870. Taylor, A.J. and Linforth, R.S.T. (2001) Modelling flavour release through quantitative structure property relationships (QSPR). Chimia 55, 448Ð452. Taylor, A.J. and Linforth, R.S.T. (2003) Direct mass spectrometry of complex volatile and non-volatile flavour mixtures. Int. J. Mass Spectrom. 223–224, 179Ð191. Taylor, A.J., Shojaei, Z.A., Bayarri, S., Hollowood, T.A. and Hort, J. (2008) Balancing flavour attributes in reduced fat foods. In: Recent Highlights in Flavor Chemistry and Biology (eds T. Hofmann, W. Meyerhof and P. Schieberle), Deutsche Forschungsanstalt fur Lebensmittelchemie, Garching, Germany, pp. 53Ð58. Togari, N., Kobayashi, A. and Aishima, T. (1995) Relating sensory properties of tea aroma to gas chromatographic data by chemometric calibration methods. Food Res. Int. 28, 485Ð493. Tsachaki, M., Gady, A.-L., Kalopesas, M., Athes, V., Linforth, R.S.T., Marin, M. and Taylor, A.J. (2008) Effect of ethanol, temperature and gas flow rate on volatile release from aqueous solutions under dynamic headspace dilution conditions. J. Agric. Food Chem. 56, 5308Ð5315. Tsachaki, M., Linforth, R.S.T. and Taylor, A.J. (2005) Dynamic headspace analysis of the release of volatile organic compounds from ethanolic systems by direct APCI-MS. J. Agric. Food Chem. 53, 8328Ð8333. Turner, J.A., Linforth, R.S.T. and Taylor, A.J. (2002a) Real time monitoring of thermal flavour generation. J. Agric. Food Chem. 50, 5400Ð5405. Turner, J.A., Sivasundaram, L.R., Ottenhof, M.-A., Farhat, I.A., Linforth, R.S.T. and Taylor, A.J. (2002b) Monitoring chemical and physical changes during thermal flavor generation. J. Agric. Food Chem. 50, 5406Ð5411. Van Gemert, L.J. and Nettenbreijer, A.H. (1977) Compilation of Odor Threshold Values in Air and Water, National Institute for Water Supply, Voorburg, The Netherlands. Vuilleumier, C., Cayeux, I. and Velazco,M.I. (2002) Dose-response curves of odor and taste stimuli: influence of sweetening agents. In: Chemistry of Taste: Mechanisms, Behaviors, and Mimics, Vol. 825 (eds Given, P. and Paredes, D.), American Chemical Society, Washington, DC, pp. 140Ð157. Weel, K.G.C., Boelrijk, A.E.M., Alting, A.C., van Mil, P., Burger, J.J., Gruppen, H., Voragen, A.G.J. and Smit, G. (2002) Flavor release and perception of flavored whey protein gels: perception is determined by texture rather than by release. J. Agric. Food Chem. 50, 5149Ð5155. Weel, K.G.C., Boelrijk, A.E.M., Burger, J.J., Jacobs, M.A., Gruppen, H., Voragen,A.G.J. and Smit, G. (2004) Effect of emulsion properties on release of esters under static headspace, in vivo, and artificial throat conditions in relation to sensory intensity. J. Agric. Food Chem. 52, 6572Ð6577. Wright, J., Wulfert, F., Hort, J. and Taylor, A.J. (2007) Effect of preparation conditions on release of selected volatiles in tea headspace. J. Agric. Food Chem. 55, 1445Ð1453. Yeretzian, C., Jordan, A., Brevard, H. and Lindinger, W. (2000) On-line monitoring of coffee roasting by proton-transfer-reaction mass-spectrometry (PTR/MS). In: Flavor Release,Vol.763 (eds D.D. Roberts and A.J. Taylor), American Chemical Society, Washington, DC, pp. 112Ð123. Yeretzian, C., Jordan, A., Badoud, R. and Lindinger, W. (2002) From the green bean to the cup of coffee: investigating coffee roasting by on-line monitoring of volatiles. Eur. Food Res. Technol. 214, 92Ð104. Zini, E., Biasioli, F., Gasperi, F., Mott, D., Aprea, E., Mark, T.D., Patocchi, A., Gessler, C. and Komjanc, M. (2005) QTL mapping of volatile compounds in ripe apples detected by proton transfer reaction-mass spectrometry. Euphytica 145, 269Ð279. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

11 Sensory methods of ﬂavour analysis Ann C. Noble and Isabelle Lesschaeve

11.1 INTRODUCTION

Similarly to instrumental analysis of flavour, analytical sensory tests must be conducted under standardised and controlled conditions. However, many factors influence subjects psychologically and physiologically, and so protocols for sensory studies require control of additional variables. For example, it is important to remove all external clues that can bias perception. Tests must be conducted in facilities that prevent distraction. Studies must be designed to account for sequence effects that influence responses. Further, analytical sensory judges require training, unlike instruments, although both need to be calibrated and tested for reproducibility. The type of test and judge depends on the purpose of the test. Most analytical tests answer one of the following questions: ‘Is there a difference?’, ‘What is the difference?’ and ‘How large is it?’ These are done in sequential fashion and can provide very detailed information about the sensory properties of flavours or flavoured products. In contrast to these analytical evaluations, subjective preference (hedonic) tests are used to determine factors influencing consumer acceptance. The purpose of this chapter is to introduce the basic principles of sensory evaluation and the factors that must be considered in sensory analyses of flavour. Excellent references for standard sensory methods include Carpenter et al. (2000), Lawless and Heymann (1998), Stone and Sidel (1985) and Meilgaard et al. (2006).

11.2 ANALYTICAL TESTS

11.2.1 Discrimination tests 11.2.1.1 Difference tests Initially, informal or formal tests should be conducted to determine whether perceptible differences exist between products. When the products differ by differences that are too small to be described, the two most common difference tests are the duoÐtrio test and the triangle test (Lawless and Heymann, 1998). When the difference can be deﬁned, paired tests are used to ask which sample is higher in a speciﬁed attribute. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 297

11.2.1.2 Threshold tests Threshold tests are performed to determine the sensitivity of subjects to a specific compound or to estimate the compound’s contribution to flavour. The threshold value is the concentration of a compound at which a detectable difference in aroma or taste is found (detection threshold) or at which the characteristic odour or taste can be recognised (recognition threshold). Typically, these values are determined by difference tests, with the sets presented in increasing order of concentration. A compilation of threshold values is available in an American Society for Testing and Materials publication (Stahl, 1978). Determination of threshold values should be used in quality control (QC) tests of taints to select judges who are sensitive to specific compounds. Similarly, threshold determination is useful in studying factors that influence individual differences, such as age, gen- der or disease. In most flavour investigations, threshold values are of only limited use for several reasons. The value applies only to the tested product under specific testing conditions since the threshold level varies with temperature, sample composition and the method by which the values were determined. Guadagni et al. (1968) proposed the concept of odour units, whereby the concentration of a component is expressed as the concentration divided by the threshold value. However, as all flavourists and perfumers know, neither the intensity nor the nature of suprathreshold concentrations of an odourant can be predicted from the threshold value and the number of odour units present. Compounds increase in intensity at different rates and can demonstrate marked differences in quality at different concentrations. At best, only ‘guesstimates’ of the contribution of individual compounds to flavour can be made from the threshold value and volatiles composition (see Chapter 1 for a discussion of the conversion of analytical data to deliver sensory flavour goals).

11.2.1.3 Gas chromatography–olfactometry Analogously to threshold value determinations, gas chromatographyÐolfactometry (GC-O) methods have been used to identify compounds that are ‘impact’ (characteristic) volatiles or at least ‘odour-active’ compounds. Interpretation of GC-O results is subject to problems similar to those discussed above in threshold testing. Although specific compounds may be ‘odour-active’, their contribution to the flavour in the mixture may not correspond to GC-O predictions. Despite this caveat, the utility of GC-O has been demonstrated by Guth (1997) in some specific applications. Forty-four odour-active compounds identified in Gewurztraminer¬ wine were added to a model solution at concentrations equal to those in the wine. The solution was rated very similar in odour to the parent wine. When one of the most odour-active compounds, cis-rose oxide, was omitted, the solution was no longer perceived as similar to the wine, demonstrating the importance of this compound to the Gewurztraminer¬ aroma. Several GC-O methods, as discussed elsewhere in this text (see Chapter 9), have been employed to detect compounds that contribute to aroma: Charm (Acree et al., 1984), aroma extract dilution analysis (AEDA) (Ullrich and Grosch, 1987), OSME (McDaniel et al., 1990), olfactory global analysis (OGA) (Ott et al., 1997) and nasal impact frequency (NIF) (Pollien et al., 1999). The methods vary primarily in that Charm and AEDA are dilution techniques, whereas OSME, OGA and NIF rely on judges’ consensus in evaluation of one concentration. Sources of error that occur in GC-O applications include inappropriate preparation of the extract for GC analysis (Zellner et al., 2008), judges not being trained or becoming fatigued. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

298 Food Flavour Technology

Typically in GC-O studies, judges are required to sniff a run for a maximum of 20 minutes so as to minimise fatigue. Another source of error is the difﬁculty in perceiving peaks that elute very rapidly. Hanaoka et al. (2001) noted that subjects who had faster breathing rates showed a tendency for higher frequency of odour detection. In one experiment, when subjects were asked to breathe faster, the primary effect was an increase in intensity of the odour. Usually, the make-up air is humidiﬁed to minimise the discomfort of dry nasal passages experienced by judges. In contrast to observations made by others over the years and to observations made in our laboratory, Hanaoka et al. (2000) reported that omitting the moisture in the make-up gas did not affect subjects’ comfort. For AEDA (or Charm), extracts are presented to a few individuals in serial dilution until no further odour is perceived. The dilution factor (FD) is reported for each individual as an indication of the ‘strength’ of the odour in AEDA. For OGA (Van Ruth et al., 1995) and OSME, odour-active volatiles detected by consensus are reported. De- spite the different methodologies, Le Guen et al. (2000) found AEDA, Charm and OSME yielded similar results in evaluation of the impact compounds of mussels. The limitations of GC-O were addressed recently by Barbe et al. (2008). The direct impact of ␤-damascenone on wine aroma was overestimated by GC-O due to its very low detection threshold, its wide range of contents and dependence on the composition of the medium.

11.2.2 Intensity rating tests To measure the size and nature of differences in flavours or flavoured products, the intensity of a specific attribute or attributes is rated by trained judges. There are several different scaling procedures and types of scales. Category scales or unstructured line (graphic) scales are used most frequently. Another method is magnitude estimation (ratio scaling) in which the intensity of a sensation is rated relative to the intensity of a reference. For example, if the reference is defined as an intensity of 10, and the sample is twice as intense, it is rated 20. Magnitude matching (or cross-modal matching) is a variation in which the intensity of tastes, smells or mouth feel is rated relative to a standard sound or light (Marks et al., 1988). The labelled magnitude scale (LMS), which is a combination of the ratio and category scales, was developed by Green et al. (1993). In contrast to category and graphic scales, which are anchored at the ends by the terms ‘low’ and ‘high’, the LMS is anchored by ‘barely detectable’ and ‘strongest you can imagine’ (see Fig. 11.1). For most studies in which the sensory properties of flavoured systems are being measured, the unstructured line scales or category scales are ideal. For a normal range of intensities, the same results are obtained using ratio scaling as found with category or graphic scales, but the latter are simpler to use than magnitude estimation methods (Giovanni and Pangborn, 1983). To measure responses of individuals who perceive stimuli very differently because of inherent physiological differences, the LMS scale or cross-modal matching better reflects the strength of the sensation perceived by the individual (Prutkin et al., 2000). In many studies in which category scales were used to rate bitterness, no differences in bitterness intensity were found between subjects who were sensitive (tasters) to the bitter compound propylthiouracil (PROP) and those who were insensitive (non-tasters) (Hall et al., 1975; Mela, 1989; Cubero- Castillo, 1999). In contrast, tasters rated intensity of bitterness, saltiness and the burn of capsaicin more strongly than non-tasters using cross-modal matching (Bartoshuk et al., 1994) or LMS (Bartoshuk et al., 2000). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 299

Fig. 11.1 Examples of scales used for rating intensity: (a) labelled magnitude estimation; (b) unstructured (graphic) line scales; (c) category scale.

11.2.3 Time–intensity rating 11.2.3.1 Individual attributes (time–intensity) versus multiple attributes: temporal dominance of sensation Temporal procedures are used to characterise persistent sensations, such as astringency and bitterness, and to monitor perceived intensity over time as flavour is released during ingestion and mastication. As the concentration of a single compound is increased, maximum intensity and total duration increase, whereas only small differences in time to maximum intensity occur. When the structure or composition of a system is altered, the rates of release of tastants or odourants are affected, as well as the maximum intensities. Examining the rate of onset and rate of decay of intensity can be used to model perception of taste and mouth feel (Pfeiffer et al., 2000) and suggest explanations for mechanisms of perception (Linforth et al., 1999). Recently, sensory timeÐintensity (TI) has been coupled with instrumental TI to characterise flavours more dynamically. Subjects rate perceived intensity while the concentrations of volatiles in the nasal passage (breath-by-breath analysis) (Linforth and Taylor, 1993; Chapter 10) or compounds in the oral cavity (Davidson et al., 1998, 1999) are simultaneously being measured. In temporal dominance of sensation (TDS), terms are selected to describe flavours perceived post-ingestion of foods and judges trained in use of the method and use of the terms. After swallowing the sample, judges select the dominant sensation and rate its intensity for 5 minutes. Whenever a different sensation predominates, it is chosen and intensity is rated. The TDS score is calculated from the intensity and duration of each attribute (see Pineau et al., 2004). TDS data can also be represented by curves showing, for each product, the percentage of judges who selected the attribute as dominant at a specific time. In TDS of gels containing different levels of odourants (peach and mint), citric acid, cooling agent and xanthan gum, reliable information was obtained, which was close to the standard descriptive analysis (DA) of the same gels. In addition, TDS provided information on the dynamic of perception after product consumption that was not available by DA (Labbe et al., 2009). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

300 Food Flavour Technology

11.2.3.2 Analysis of TI data Rating sensory intensity continuously (TI) and choosing the dominant sensation over time (TDS) are difficult tasks that require more training than do simple scaling tests. The way in which judges move their tongues, the rate at which they chew and their salivary flow rates Ð influence their perception of intensity and persistence of the sensation. Even with extensive training in TI, individual judges have characteristic curves with idiosyncratic, yet reproducible, patterns. Despite the difference in temporal patterns, judges can be trained to give reproducible responses that are consistent across samples. In many studies, TI parameters (such as time to maximum intensity, maximum intensity and total duration) are extracted from the raw TI curves for each sample and each judge and then averaged. Alternatively, the mean intensity rating at each time is calculated and the results are expressed as an average TI curve. TDS curves are analysed as the frequency of selection of a specific sensation versus that of chance alone. TDS scores are analysed by analysis of variance (for details see Pineau et al., 2004; Labbe et al., 2009).

11.2.3.3 Effect of mode of sampling Although continuous rating of intensity of ‘flavour by mouth’ attributes before and after spitting or swallowing the stimulus has many applications, TI methods are inappropriate for analysis of aromas that are not continuously evaluated. When odourants are rated when sniffed (or inhaled), the sensation of aroma decreases almost immediately after the subject stops sniffing the samples, as illustrated in Fig. 11.2. When a solution of 150 ppm menthol was sniffed for 10 seconds, overall menthol intensity increased rapidly, and then equally quickly decreased when the subject stopped sniffing (curve 4 in Fig. 11.2). In contrast, the perception of intensity of menthol was more intense and lasted far longer when the same

Time (second)

Fig. 11.2 Average intensity ratings over time of 150 ppm menthol evaluated under four conditions: sipping menthol + 2 g/L caffeine (curve 1) or menthol alone (curve 2); snifﬁng menthol for 10 seconds while tasting either 2 g/L caffeine (curve 3) or water (curve 4); n = 20 judges × 2 replications. Modiﬁed from Opet (1989). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 301

solution was sipped and then spat at 10 seconds (curve 2) since the residual menthol in the mouth was still detected (Opet, 1989).

11.2.4 Taste–smell interactions In many TI studies, cognitive interactions between taste, trigeminal sensations and smell are found (Noble, 1996). For example, increasing the sourness or sweetness of orange-flavoured solutions, without altering the flavour concentration, produced an increase in the intensity and duration of fruitiness (Bonnans and Noble, 1993). The overall intensity of menthol, which has a bitter taste, minty smell and cooling trigeminal sensation, is similarly affected by the bitterness of caffeine (Opet, 1989). When subjects tasted solutions of 150 ppm menthol to which 2 g/L caffeine had been added (curve 1, Fig. 11.2), menthol intensity was rated more intense and lasted longer than when menthol alone was sipped (curve 2). In the same study, cognitive enhancement also occurred when aroma and taste were perceived independently using a dual-delivery system (Hornung and Enns, 1984). When menthol was sniffed and caffeine was tasted simultaneously (curve 3), the perception of menthol intensity was rated higher and lasted longer than when menthol was sniffed and water was tasted (curve 4) (Opet, 1989). A similar cognitive interaction was seen in a study in which both TI and ‘breath by breath’ analyses were conducted (Davidson et al., 1999). Subjects chewed menthone-flavoured gum and rated the intensity of minty flavour, while the concentration of menthone in the nasal passage was monitored. At the same time, the residual concentration of sucrose in the oral cavity was also monitored. As shown in Fig. 11.3, the perceived intensity of the menthone flavour in the gum matched the change in sugar concentration rather than the high concentration of menthone in the nose. Subjects probably became fatigued to the menthone, yet continued to rate ‘mintiness’ that was enhanced by or associated with the residual sweetness.

11.2.5 Descriptive analysis The technique of DA provides a quantitative analytical characterisation of aroma, taste and mouth feel as described in detail elsewhere (Stone et al., 1974; Hootman, 1992; Heymann Normalised data (%)

Time (minute)

Fig. 11.3 Sucrose release (), menthone release (|) and perceived overall mint ﬂavour intensity () from chewing gum. Reprinted with permission from Davidson et al. (1999). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

302 Food Flavour Technology

et al., 1993; Lawless and Heymann, 1998). The first requirement for a meaningful DA is development of a vocabulary that will describe the differences in flavour among the samples in specific concrete terms. For examples of these, see the lexicons for flavour of beer (Meilgaard et al., 1982), wine (Noble et al., 1987), whiskey (Shortread et al., 1979) and cheese (Berodier et al., 1997; Pagliarini et al., 1991). For each DA study, terms are derived that describe the differences in aromas and flavours of the experimental samples. Reference standards are provided to define each term and train judges to rate each attribute consistently, as detailed for cheese (Murray and Delahunty, 2000), beer (Meilgaard et al., 1982) and wine (Noble et al., 1987). When immediate feedback was given to panelists about their performance, training time was reduced by 50% in wine DA (Findlay et al., 2006). After the judges are trained, the intensity of each term is rated in each product. A variation of the DA method, known as free-choice profiling, in which judges each use their own terms, can also be used, but the results are very difficult to interpret (Williams and Langron, 1984; Williams and Arnold, 1985; Beal and Mottram, 1993). Plotting the ratings from the DA data reveals the flavour profiles of the samples as shown in the following two examples. Heating wines accelerates normal ageing reactions and produces changes in wine flavour. To characterise the effect of storage at an elevated temperature on wine flavor, Chardonnay wines were stored at 40◦C for 0, 15 and 30 days and then pro- filed by DA (de la Presa Owens and Noble, 1997). As shown in Fig. 11.4, the intensity of the floral and fruity notes (tropical fruit, green apple) was reduced by 20% after storage for 15 days, although only the decrease in green apple was statistically significant. After 30 days, the intensities of citrus, tropical fruity, green apple and floral aromas were significantly decreased by nearly 80%. At the same time, the intensity of tea, honey, rubber and oak aromas increased. Products heated in a microwave do not have the typical nutty, toasted or caramel aromas that are produced by Maillard browning reactions in conventional thermal cooking. Descriptive analysis was used to monitor the differences in aroma between thermally heated solutions of cysteine-HCl and glucose and those heated in a microwave (Song, 1990). In one trial, the pH of the microwaved samples was adjusted to 2, 7 and 9. In Fig. 11.5, the mean ratings for each term in each sample are plotted to illustrate the aroma profiles. The roasted,

◦ Fig. 11.4 Aroma profiles of Chardonnay wines heated for 0, 15 or 30 days at 40 C. Centre of the figure, low intensity; perimeter, higher intensity for each term; LSD, least significant difference. Wines differ significantly (p Ͻ 0.05) if their intensity means differ by more than the LSD; n = 15 judges × 3 replications. Modified from de la Presa Owens and Noble (1997). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 303

Fig. 11.5 Aroma profiles of aqueous solutions of 0.25 mol/L cysteine-HCl and 0.25 mol/L glucose that ◦ were heated conventionally for 40 hours at 100 C with refluxing or microwaved for 5 minutes. Centre of the figure, low intensity; perimeter, higher intensity for each term; n = 15 judges × 3 replications; , thermal pH 9; , microwave pH 9; , microwave pH 7; , microwave pH 2 (Song, 1990).

nutty, brown sugar and popcorn notes were higher in the pH 9 microwaved treatment than in all other samples, including the thermally heated one.

11.2.6 Quality control tests For QC tests, the difference tests described above are too sensitive. Many alternative methods are used in quality control or quality assurance programmes, as reviewed in detail by Munoz˜ et al. (1992) (see also Yantis, 1992; Carpenter et al., 2000). The QC manual developed for evaluating corks for off-odours describes the procedures and guidelines for making accept or reject decisions (Butzke and Suprenant, 1998). The ‘in/out’ method is a decision-making tool for evaluating daily production. On-line judgements are made as to whether a product is within or outside the product specifications. The disadvantage of this approach is the lack of information provided and the difficulty in defining the specification limits. ‘Difference from control’ tests (degree of difference tests) rate the size of the difference of a production sample from a ‘control’ or ‘standard’ one. Category scales labelled with terms describing the degree of difference (none, very slight, slight, moderate, large, extreme) or unstructured line scales anchored by ‘no difference’ and ‘extreme difference’ can be used. Although this method is simple, its disadvantage is that the overall rating for the size of difference from the standard product provides no information about the nature of the differences. Moreover, simple difference tests find differences that are smaller than consumers can usually perceive. Prescott et al. (2005) proposed determination of consumer rejection thresholds as a QC method for off-flavour monitoring, such as cork taint in wine. Some production sites use descriptive analyses for a more comprehensive approach to QC. Trained judges rate the attributes that have been found to be most important in describing differences among products. Pre-set specifications are determined to define the acceptable range of attribute ratings for making QC judgements. Many quality assurance programmes rely on consumer inputs to define product specifications in terms of sensory or flavor profiles (Munoz,˜ 1992). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

304 Food Flavour Technology

11.3 CONSUMER TESTS

11.3.1 Purpose of consumer tests Consumer tests or affective tests are conducted primarily to determine how much a product is liked by a targeted population of consumers, usually in blind tasting condition where the objective is to characterise sensory preferences independently from any marketing inputs. However, the latter do affect consumer food choice by creating expectations about the product sensory quality and other potential beneﬁts. Consumer tests are conducted for product improvement/optimisation, development of new products, assessment of market potential, product category review and support for advertising claims (Meilgaard et al., 2006).

11.3.2 Methods The methodologies can be categorised into qualitative and quantitative affective tests. Qual- itative methods are used in the exploratory phases of the new product development to test a new flavour concept with a small group of consumers. Focus groups can be useful to identify factors that influence the consumer’s preference (McNeill et al., 2000), although several authors have argued about their relevance in providing objective directions for product development in terms of sensory/flavour attributes (Lawless and Heymann, 1998; Husson et al., 2001; Lesschaeve et al., 2002). In many industries, decisions are made on the basis of the managers’ intuition while observing the focus group’s ‘consumer speak’. To add some objectivity in the elicitation process, the repertory grid methodology (Thomson and McEwan, 1988) can be used to determine the attributes (e.g. flavour and colour) or features (e.g. easy opening packaging) of products that seem important to consumers. Sensory preference for a product is most frequently determined with paired preference tests between two samples; for several samples, products are ranked in order of increasing preference. Liking is usually rated on hedonic category scales, anchored by terms such as ‘dislike extremely’ to ‘like extremely’. The 9-point hedonic scale developed by Pilgrim and Peryam in 1957 is still widely used by market researchers in North America, although sensory scientists tend to favour unstructured linear scales or labelled magnitude scales for their ratio properties. During a consumer test, it is tempting to ask respondents what they liked or disliked in the tested products. However, Lesschaeve (2006) showed that using attributes generated by consumers to guide product development can be misleading. The term ‘vanilla/oak’ tended to be used to describe liking for wine, whereas ‘smoky/oak’ tended to reflect unpleasant sensory experience. For more information on consumer testing, consult Meilgaard et al. (2006).

11.4 SENSORY TESTING ADMINISTRATION

11.4.1 Facilities 11.4.1.1 Analytical tests Analogously to instrumental measurements, judges need to be calibrated and experiments should be conducted under reproducible conditions. However, unlike the response of instruments, the perception of judges is inﬂuenced by psychological and physiological factors that can cause additional sources of ‘experimental noise’. Consequently, sensory experiments P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 305

are conducted under conditions designed to minimise or exclude bias and distraction. For example, the evaluation area is separated from the preparation area so that the judges are not given clues about the experiment and so that odours can be controlled. The environment in which sensory tests are conducted should be temperature-controlled, quiet and odour-free. The testing facility should have a clean source of odourless air that is maintained at positive air pressure to prevent outside odours from interfering. Ideally, booths with partitions between the judges eliminate distraction and prevent interaction between judges (see Eggert and Zook, 1986). Controlled lighting should be used either to mask differences in appearance or to stan- dardise the light source and conditions under which colour and appearance are tested. If the samples vary in colour or appearance when only ﬂavour is being evaluated, products should be served in black glasses or evaluated under masking red light to remove visual cues. Conditions for conducting sensory tests that are linked with instruments, such as GC-O, breath-by-breath analysis or MRI, should also be optimised to exclude distraction by the laboratory environment. The area must be protected from noise and odour. If this is not possible, other activities in the laboratory should be curtailed during sensory testing.

11.4.1.2 Consumer tests Typically, consumer tests are conducted at a central location or at home. Central location settings should simulate a natural context of consumption as much as possible, as recent research showed that liking scores can vary signiﬁcantly between a laboratory (clinical) environment and a real restaurant (King et al. 2007). With the increasing presence of computers at home and the use of the internet for domestic purposes, on-line surveys are becoming popular among consumer researchers. Most traditional sensory software now has web-based applications, allowing the use of sensory methods and respective experimental designs for on-line tests with or without product evaluation.

11.4.2 Test administration 11.4.2.1 Analytical tests Samples should be presented in identical containers coded with random numbers to prevent bias from extraneous clues such as brand or treatment. Standard tulip-shaped clear wine glasses are optimal for evaluation of aroma, while plastic cups or beakers can be used when only taste or mouth feel is being evaluated. Providing watch glasses or Petri dishes as sample lids during aroma evaluations increases aroma intensity and reduces unwanted odours in the tasting facility. Whether their participation involves simple or complex tasks, subjects should not be the experimenters (sensory administrators). In cases in which subjects are required to pour or distribute their own samples, error can be introduced by non-uniform sample sizes and by mixing up of samples. Further, judges may learn about the design of the experiment. Asking subjects to perform complex tasks on their own, such as initiating timing with a stopwatch, tasting a sample and turning a device on and off at speciﬁed times, distracts from their ability to concentrate on rating intensity and introduces error. Preferably, subjects should not be required to monitor time while performing a series of tasks. Software for acquisition of sensory data can be programmed to prompt subjects when P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

306 Food Flavour Technology

to sniff or sip, when to initiate rating and when to spit. Without this automation, a sensory analyst needs to work one-to-one with the subject to control and regulate all such details.

11.4.2.2 Consumer tests Guidelines similar to those outlined above for analytical tests can be followed when consumers assess the products without any information (blind conditions) in a central location or at home. However, hedonic or preference tests can be complemented with assessments made with information (image, brand or allegations) to determine the relative weight of non- sensory and sensory cues on consumers’ hedonic appreciation. Depending on the product category, non-sensory cues may affect the initial sensory hedonic evaluation of the product (Deliza and MacFie, 1996; Jaeger, 2006).

11.4.3 Experimental design The nature and number of the experimental samples determine the specific sensory test as well as the design of the experiment. When there are too many products to be evaluated in one session without fatiguing the judges, experimental designs should be used to randomly assign samples to different sessions. For experimental designs, see texts such as Cochran and Cox (1957). As with analytical tests, the number of samples that can be evaluated for preference in one session without fatigue is dependent on both the nature of the samples and the demands of the sensory test. Fewer samples can be rated in a session when strong, pungent or persistent flavours are being evaluated than when less intense flavours or those with shorter duration are being tested. Three to four samples may be the most that can be evaluated in one session for a DA in which 10Ð15 attributes are being rated. To determine whether fatigue has occurred in the training sessions, serve the same sample first and last. If results are not reproducible, try presenting fewer samples. To reduce fatigue, subjects should breathe fresh air or rinse with water between samples. The specific interstimulus protocol varies with each product. Warm water should be used for rinsing between oily or fatty samples, while rinses of gelatin alternated with water help reduce astringency. In almost all analytical sensory tests, the samples are not swallowed but spat to avoid introducing a new variable: judge satiation or intoxication in the case of alcoholic beverages. Small cardboard or plastic containers with lids are an inoffensive way to facilitate this. Both the context of presentation and the sequence in which samples are presented affect perception, as discussed in Lawless and Heymann (1998). Hence, samples should be served in randomised orders to eliminate sequence and carry-over effects. The order of presentation of difference tests (paired, duoÐtrio or triangle sets) should be randomly assigned to avoid fatigue. To minimise guessing, the position of the samples within each set should be randomised. Designs for difference tests are presented by Stone and Sidel (1985) and Meilgaard et al. (2006). For rating tests, the order in which samples are presented should be randomised to eliminate errors that arise from systematic contrasts. For example, an acid solution at pH 3.25 was rated lower in sourness when presented after a pH 3.0 sample than when presented after a solution with pH of 3.5 (Norris, 1982). Similarly, shifts in intensity were also observed in sweetness ratings of a fruit beverage. Mid-range juices were less intense in the context of stronger items and more intense in the context of weaker items (Diamond and Lawless, 2001). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 307

Samples presented in the first position are usually rated higher, and often significantly higher, than subsequent samples irrespective of sample identity. This occurs with analytical rating as well as in preference tests. In an interlaboratory study of preference, eight coffees were presented in a Latin square design. The preference score for the first position, regardless of coffee identity, was significantly higher than for the other positions (Network, 1996). To eliminate the effect of the first sample being rated differently, a warm-up can be presented first and its scores can be discarded (Stone and Sidel, 1985). In evaluation of some products, time-order or carry-over effects occur (Amerine et al., 1965). In the evaluation of astringent or bitter products, the second sample is almost always perceived as higher in intensity. Thus, the design must randomise the order of presentation as well as balance the number of times each specimen is presented in a specific sequence. This can be done using William’s Latin squares designs, which are balanced for first-order carry- over effects (MacFie and Bratchell, 1989; Schlich, 1993). In some cases, the carry-over effect is compound specific. In a study of bitterness perception, different compounds were presented in sequence. No matter what compound was presented first, the perceived bitterness of all bitterants in the study increased when presented after another bitter stimulus. Yet caffeine increased bitterness of compounds that followed it by the largest amount, although bitterness of caffeine was least affected by other compounds. In contrast, the increase in bitterness of quinine was higher than that of the other bitter compounds in the study (Cubero-Castillo and Noble, 2001). Several compounds that stimulate the trigeminal sense show both sensitisation (increased perception of intensity) and desensitisation, depending on the time during presentation of samples, as well as on the specific chemical (Green and Lawless, 1991; Karrer and Bartoshuk, 1991). Pungency and burn of compounds such as capsaicin initially cause a sensitisation in which each subsequent sample is perceived as more intense when samples are presented at 1- to 5-minute intervals. When the sample is presented again after a break of 15Ð20 minutes, desensitisation occurs. The intensity of burn is greatly reduced, as shown in Fig. 11.6 (Green, 1989). Thus, evaluation of ‘hot/spicy’ products must be made carefully for valid results.

Time (minute)

Fig. 11.6 Perceived burning intensity of capsaicin over time. Sensitisation increases the burning intensity as capsaicin is applied 25 times at 1-minute intervals. After a 15-minute break, the intensity rating is greatly decreased, showing that desensitisation has occurred. Reprinted with permission from Green (1989). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

308 Food Flavour Technology

11.5 SELECTION AND TRAINING OF JUDGES

11.5.1 Human subject consent forms and regulations Government regulations require that researchers respect and protect the rights and welfare of individuals participating as subjects in studies. In conducting research involving human subjects, the individual’s rights and well-being must be the primary concern. Judges may stop participating without penalty and cannot be coerced into participating against their will. Although the guidelines for approval of sensory protocols vary with country and company, all judges must sign a human consent form before participating in a study. In this form, subjects must be informed of risks and beneﬁts of the study, such as the presence of sulfur dioxide in case the subject is allergic to it. In this form, it must be explained in general terms what the judges will be evaluating without revealing the purpose of the experiment or the exact nature of the variables to be tested.

11.5.2 Judges 11.5.2.1 Analytical tests 11.5.2.1.1 Judge selection Initially, judges should be selected on the basis of availability and motivation. Recruit more judges than are needed. The specific requirements for selecting judges vary with the type of test. If a QC test involves detection of a specific off-odour, such as the mouldy or musty notes of geosmin in water or of trichloroanisole in wine corks, select judges who are very sensitive to these compounds. To determine judges’ sensitivity, a series of paired tests can be presented to the judge in which a blank is compared with low concentrations of the specific chemical, or a formal threshold test can be done. For more general studies, judges’ performance should be examined. Retain subjects who perform reproducibly and are consistent with other judges, as discussed below.

11.5.2.1.2 Training At the beginning of any new testing, introduce each judge to the scorecard, the rating scale and the procedures. Initially, present samples that are very different from each other. This is effective as a training tool, since the specific attributes of interest are more easily perceived, such as a ‘floral’ note in a set of Riesling wines. More importantly, it helps morale. For unspecified difference tests, once judges are familiar with the testing procedure, only one training session, if any, is required. For paired tests or scaling tests, training is needed to define the specific attribute being tested. For DA, 1Ð6 weeks of training is required. The first sessions are discussions, in which reference standards are presented and their appropriateness for rating the flavour of the experimental products is reviewed. Subsequent sessions focus on refining the standards and training judges to use the terms consistently. Complex tests, such as TI studies, require extensive training to train the judge to coordinate activities such as initiating the recording, sipping, spitting and possibly sipping subsequent samples, while focusing on rating the attribute. When TI is combined with breath-by-breath analysis, even more training is needed so that the judge can breathe with controlled cadence, while focusing on rating intensity. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 309

For paired tests, scaling, DA or TI tests, reference standards should be developed that deﬁne the attributes, such as those that have been used for cheese (Murray and Delahunty, 2000), beer (Meilgaard et al., 1982) and wine (Noble et al., 1987). Presenting ‘high’- and ‘low’-intensity standards for each attribute helps to train panellists in the use of the scale and align the concepts.

11.5.2.1.3 Evaluating judges’ performance Keeping records on judges’ performance will permit selection of judges who are consistent, sensitive and reproducible. For difference tests, the cumulative number of correct judgements can be tracked (Amerine et al., 1965). In DA or TI tests, practice sessions should be conducted that use the same protocol as that used in formal testing. Prior to each DA session, the judge should smell and/or taste each of the references, and then rate each attribute. Inspect the data immediately to detect judges who are using a term inconsistently with the other subjects. This judge should be presented with samples or reference standards that illustrate ‘low’ and ‘high’ intensities of the terms that are not being used correctly. After familiarising himself or herself with these, the judge should rate more coded samples, and the results should be reviewed again. The data from the training sessions should be analysed before initiation of the formal data collection, as described below. This saves time and money. Rating 40 attributes and then ﬁnding that only 4 varied signiﬁcantly across the samples is a waste of time. For any type of test, give rewards (cookies, money, gifts) and provide positive feedback. Motivated judges are more focused and have better performance.

11.5.2.2 Consumer tests Consumer testing requires identification of typical consumers of products or flavours being tested. Personnel at a work site or ‘experts’ should not be used, unless they represent the typical consumer. Therefore, it is critical to define the targeted consumers to be recruited in terms of demographics, purchase and consumption habits based on the existing knowledge of the typical consumers of a product category or based on the consumer targeted to become users of the new product or flavor. Because of the tremendous variation in preferences, a large number of target consumers (n Ͼ50) must be recruited especially when conducting quantitative tests. Geographical and cross-cultural differences may create different responses from consumers in product liking; therefore, these factors need to be considered in the screening of potential respondents and in the selection of the location where the test will take place.

11.6 STATISTICAL ANALYSIS OF DATA

11.6.1 Analytical tests 11.6.1.1 Difference tests Tables constructed from binomial distributions provide the number of correct responses needed for statistically significant differences at specific levels of confidence. Alternatively, tables based on normal distributions indicate the probability that the number of correct responses could have occurred by chance alone. Tables for determining significance levels P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

310 Food Flavour Technology

and exact probability levels are provided by Roessler et al. (1978), O’Mahony (1986), Lawless and Heymann (1998) and Brockhoff and Schlich (1998).

11.6.1.2 Analysis of variance Data from simple scaling tests, DA or TI studies are evaluated by analysis of variance (ANOVA) to determine whether the samples are significantly different. If the judges have scored the samples more than once, the ANOVA can determine whether judges are reproducible and consistent with each other (i.e. rate the attribute in the same manner). In ANOVA, the main effects, such as judges, samples and replications, can be considered fixed (results apply only to this experiment) or mixed (the same results would apply if other cases or judges were tested). Unless the experiment is focused on individual judges (such as comparing responses of PROP tasters vs non-tasters) instead of testing differences among samples, sensory data are analysed by mixed model ANOVA. In mixed models, judges are treated as a random factor, while the samples and replications are fixed (O’Mahony, 1986; Lundahl and McDaniel, 1988; Lawless and Heymann, 1998). When significant differences are found, tests such as Duncan’s multiple range test or Fischer’s least significant difference (LSD) identify which products differed significantly from each other, as illustrated in Fig. 11.4. In Table 11.1, the summaries of mixed model ANOVAs of intensity scores for berry and apricot aromas are shown for a study in which 21 judges rated 28 wines in duplicate. In most sensory tests, judges use different parts of the scale and are a significant source of variation. For both terms, this occurred and can be ignored. However, if there is a significant judgeÐproduct interaction, it means that the judges are using the term inconsistently (Stone et al., 1974). In the mixed model ANOVA,this error is accounted for. Thus, it can be reported that there was a significant difference in intensity across wines for the berry attribute, despite the significant judgeÐwine interaction. Discussions of analysis of sensory data with emphasis on judges’ performance can be found elsewhere (Gay and Mead, 1992; Schlich, 1994).

Table 11.1 Summary of mixed model analyses of variance of two aroma terms: F-ratios, degrees of freedom, signiﬁcance levels and interpretation.

F-ratios Degrees Source of of variation freedom Berry Apricot Interpretation

Wine (W)273.58*** 1.81** Wines differ signiﬁcantly in intensity of ‘berry’ and ‘apricot’ aromas Judge (J) 20 14.63*** 11.43*** Judges use different part of the scale (not a problem) Replication (R) 1 1.23 2.56 Replications do not differ W × R 27 1.04 0.91 No difference between replications J × R 20 1.05 1.09 Judges did not vary differently between replications W × J 540 1.33*** 1.06 Some judges rated ‘berry’ differently from others Judges rated ‘apricot’ in the same way across wines

Asterisks (**)and(***) denote signiﬁcance at p Ͻ 0.01 and p Ͻ 0.001, respectively. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 311

11.6.1.3 Power Power is the probability that the experimental results are correct: you have identified a real difference or you did not fail to find a real difference. Testing of power is seldom done, but it should be done since it validates the results or indicates that you have too low a power to allow you to rely on your results. Most frequently, data from difference tests or analyses of variance are considered significantly different if ␣ (the probability that the results could have occurred by chance alone) is less than 5% (this is also expressed as p Ͻ 0.05). Few studies test for ␤, which is the chance that you did not find a real difference. Power is determined by three factors: (1) the number of judges and replications, (2) the magnitude of the differences among the samples (effect size) and (3) the ␣-level chosen. Power can be increased by using more judges and conducting more replications. Experiments conducted with a very low number of judges almost always have very low power. In most studies, the size of the differences among the samples cannot be changed. However, trained, sensitive judges can perceive very small differences and are more reproducible, thus increasing power over untrained judges. For more detail, see Cohen (1988) or Lipsey (1990).

11.6.1.4 Principal component analysis: an exploratory tool Principal component analysis (PCA) is an efficient way of looking at large amounts of data. For example, we may have many samples for which we have a large number of chemical variables (such as concentrations of 100 volatiles) or we may have intensity ratings for a number of attributes. A PCA can reduce the information contained in the large number of variables to two (or three) principal components (PCs) or factors. A first factor or PC is extracted. The first PC is the combination of the original variables that best explains the variation among the samples. The second PC is then derived to account for the maximum variation in the remaining unexplained variability, and so on. The relationship of the original variables and the samples can then be seen by their projection on these few PCs. For example, a PCA of the sensory profile data plotted in Fig. 11.5 was conducted. In Fig. 11.7 the loadings for the sensory terms are shown as vectors and the factor scores of the samples as points for the first two PCs. The first PC (which accounts for 73.4% of the total variation) contrasts samples high in cooked onion and bell pepper aroma on the left against those on the right, which are low in these two terms but high in popcorn, roasted, nutty and brown sugar. The second PC separates samples at pH 7, which were higher in burnt aroma from the rest. The small angles between cooked onion and bell pepper indicate that they are correlated with each other and negatively correlated with the brown sugar and nutty terms, which are also correlated with each other. Consistent with the flavour profiles in Fig. 11.5, the pH 9 microwaved sample is high in brown sugar and low in bell pepper and cooked onion, whereas pH 7 and pH 2 microwaved samples are the converse. It must be emphasised that PCA is an exploratory method that facilitates data interpretation. Normally, PCA is not used to test whether samples differ significantly. As shown in Fig. 11.4, one should apply the LSD values to ratings for individual terms to determine which samples differ significantly in each attribute or do a canonical variates analysis (Heymann and Noble, 1989).

11.6.2 Consumer tests Data collected from paired preference test, preference ranking or hedonic rating are analysed with the same statistical techniques used for their equivalent analytical tests. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

312 Food Flavour Technology

Fig. 11.7 Principal component analysis of descriptive analysis data for heated 0.25 mol/L cysteine-HCl and 0.25 mol/L glucose solutions. Projection of the attributes (vectors) and factor scores for sample means and for average values for three replications on principal components I and II. , 0.4 sample; ←,0.2 term; ♦, thermal and microwaved; ,pH9;•,pH7;,pH2;, means (Song, 1990).

In the 1990s, sensory scientists and sensometricians developed a series of methodologies utilising multivariate statistics to identify sensory attributes, driving consumers’ likes and dislikes, called preference mapping. Consumers do not have the vocabulary to accurately describe flavour notes. For example, the terms ‘dry’, ‘bad’, ‘bitter’, ‘astringent’ and ‘sour’ are used interchangeably to describe sour or bitter or astringent products. Thus, to identify the highly preferred flavours or identify the flavours preferred by different market segments, preference mapping methods are used. Using these techniques, consumers’ preference ratings are statistically related to the DA data (Schlich, 1995; McEwan, 1996), leading to the identification of sensory attributes driving consumer preferences and allowing reverse-engineering products to an optimum formula appealing to consumers (Moskowitz, 1994; Schlich et al., 2003).

11.7 RELATING SENSORY AND INSTRUMENTAL FLAVOUR DATA

Several approaches have been utilised to seek causal and predictive relationships between sensory and instrumental data. Multiple regression analysis has been used to predict intensity of specific sensory attributes in terms of the ‘aroma-significant’ volatile components. How- ever, the volatiles are usually highly correlated; thus several solutions to each equation are possible and the method is of little use. PCA is a better tool for analysis of highly correlated data, but it should not be done on a data set in which both instrumental and sensory variables are included since the results are not meaningful. There are many multivariate methods for relating sensory and instrumental data in which the configurations of sensory and instrumental data matrices are compared. Calculation of the RV coefficient has been used as a measure of similarity between the matrices of sensory P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 313

data and that of the instrumental data (Schlich and Guichard, 1989; Schlich et al., 1987). PCA of instrumental variables (PCAIV) selects a small subset of volatiles that yield the conﬁguration closest to that of the PCA of the sensory data (de la Presa Owens et al., 1998). Procrustes analysis is a technique by which the spaces derived by PC analyses are matched (Dijksterhuis and Gower, 1991). The relationship of chemical and sensory loadings in the consensus conﬁguration indicates which volatiles are related to each sensory attribute, as shown in an analysis of 24 Bordeaux wines (Williams et al., 1984). The method of partial least squares (PLS) analysis of latent variables (Martens and Martens, 1986) is used to uncover relationships between sensory and instrumental data sets. The technique indicates how well variables in one data set predict or model the variation among variables in the second. PLS was used to relate the DA ratings of heated cysteine- HCl/glucose solutions with their headspace volatiles. The PLS loadings for the sensory terms and the volatiles are shown for PLS factors 1 and 2 in Fig. 11.8a. In Fig. 11.8b, the factor

Fig. 11.8 Partial least squares regression of sensory descriptive analysis data and headspace volatiles data for heated 0.25 mol/L cysteine-HCl and 0.25 mol/L glucose solutions. (a) Factor scores for sample means and for average values for three replications for factors I and II. ♦, thermal and microwaved; ,pH 9; •,pH7;,pH2;, means (Song, 1990). (b) Factor loadings for attributes and volatiles for factors I and II (percentage of explained variation shown in bold for instrumental data and not bold for sensory data). P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

314 Food Flavour Technology

scores for the means and three replications are plotted for each sample. This two-dimensional solution models 84% of the variation in the volatile data and 66% of the sensory variation. PLS factor 1 separates the pH 9 microwaved sample and the thermally heated one from the pH 2 and pH 6 microwave-heated solutions, while the second factor separates the pH 9 microwaved from the thermal. The 2-ethylfuran (2efr) is not weighted heavily, indicating that it does not model variation between the data sets, whereas the rest of the volatiles were loaded heavily and contribute to modelling variation in both data sets. The bell pepper and cooked onion aromas are most closely associated with unknown peak 5, while the burnt note is associated with the 3-methyl (3mfr) and 2-acetylfuran (2acf) and peaks 25 and 26. Analogously to the selection of odour-active compounds by GC-O, these multivariate methods reveal compounds that contribute to aroma. If a large number of samples are evaluated, then the patterns found by PLS, Procrustes or PCAIV can be robust.

11.8 SUMMARY

Sensory analysis can provide objective, quantitative information about the sensory properties of flavours by the use of properly designed tests, trained judges and appropriate sensory protocols. With more complete and accurate information, and the strengthened link between flavour chemistry and sensory science, our understanding of the relationship between flavour stimuli and perceived flavour should help explain the mechanisms of flavour perception. Consumer tests done properly can increase the chance of market success especially when marketing and sensory science work together.

REFERENCES

Acree, T.E., Barnard, J. and Cunningham, D.G. (1984) The analysis of odour-active volatiles in gas chromatographic effluents. In: Analysis of Volatiles. Methods and Applications (ed. P. Schreier), Walter de Gruyter, New York, pp. 251Ð268. Amerine, M.A., Pangborn, R.M. and Roessler, E.B. (1965) Principles of Sensory Evaluation of Foods, Academic Press, New York. Barbe, J.C., Pineau, B. and Ferreira, A.C.S. (2008) Instrumental and sensory approaches for the characterization of compounds responsible for wine aroma. Chem. Biodivers. 5, 1170Ð1183. Bartoshuk, L.M., Duffy, V.B. and Miller, I.J. (1994) PTC/PROP tasting: anatomy, psychophysics, and sex effects. Physiol. Behav. 56, 1165Ð1171. Bartoshuk, L.M., Fast, K., Duffy, V.B., Prutkin, J.M., Synder, D.J. and Green, B.G. (2000) Magnitude matching and a modified LMS produce valid sensory comparisons for PROP studies. Abstract from Society for the Study of Ingestive Behavior, Dublin, 2000. Appetite 35, 277. Beal, A.D. and Mottram, D.S. (1993) An evaluation of the aroma characteristics of malted barley by free- choice profiling. J. Sci. Food Agric. 61, 17Ð22. Berodier, F., Stevenot, C. and Schlich, P. (1997) Description of the flavour of Comte cheese. Lebens-Wissens. Technol. 30, 298Ð304. Bonnans, S. and Noble, A.C. (1993) Effect of sweetener type and of sweetener and acid level on temporal perception of sweetness, sourness and fruitiness. Chem. Senses 18, 273Ð283. Brockhoff, P.B. and Schlich, P. (1998) Handling replications in discrimination tests. Food Qual. Pref. 9, 303Ð312. Butzke, C. and Suprenant, A. (1998) Cork Sensory QC Manual, University of California, Division of Agricricultural and Natural Resources, Oakland, CA. Carpenter, R.P., Lyon, D.H. and Hasdell, T.A. (2000) Guidelines for Sensory Analysis in Food Product Development and Quality Control, Aspen Publishers, Gaithersburg, MD. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 315

Cochran, W.G. and Cox, G.M. (1957) Experimental Designs,Wiley,NewYork. Cohen, J. (1988) Statistical Power Analysis for the Behavioral Sciences, Lawrence Erlbaum Associates, Hillsdale, NJ. Cubero-Castillo, E.M. (1999) The effect of compound-specific sensitivity and carry-over effects on bitterness perception, PhD thesis, University of California, Davis, CA. Cubero-Castillo, E.M. and Noble, A.C. (2001) Effect of compound sequence on bitterness enhancement. Chem. Senses 26, 419Ð424. Davidson, J.M., Linforth, R.S.T., Hollowood, T.A. and Taylor, A.J. (1999) Effect of sucrose on the perceived flavour intensity of chewing gum. J. Agric. Food Chem. 47, 4336Ð4340. Davidson, J.M., Linforth, R.S.T. and Taylor, A.J. (1998) In-mouth measurement of pH and conductivity during eating. J. Agric. Food Chem. 46, 5210Ð5214. de la Presa Owens, C. and Noble, A.C. (1997) Effect of storage at elevated temperatures on aroma of Chardonnay wines. Am. J. Enol. Viticult. 48, 310Ð316. de la Presa Owens, C.P., Schlich, P., Wada, K. and Noble, A.C. (1998) Using sensory and instrumental data to interpret the effect of storage at elevated temperatures on aroma of Chardonnay wines. Ann. N. Y. Acad. Sci. 855, 854Ð859. Deliza, R. and MacFie, H. (1996) The generation of sensory expectation by external cues and its effect on sensory perception and hedonic ratings: a review. J. Sens. Stud. 11, 103Ð128. Diamond, J. and Lawless, H.T. (2001) Context effects and reference standards with magnitude estimation and the labeled magnitude scale. J. Sens. Stud. 16, 1Ð10. Dijksterhuis, G.B. and Gower, J.C. (1991) Interpretation of generalized procrustes analysis and allied methods. Food Qual. Pref. 3, 67Ð87. Eggert, J. and Zook, K. (ed.) (1986) Physical Requirement Guidelines for Sensory Evaluation Laboratories, ASTM Special Publication 913, American Society for Testing and Materials, Philadelphia, PA. Findlay, C.J., Castura, J.C., Schlich, P. and Lesschaeve, I. (2006) Use of feedback calibration to reduce the training time for wine panels. Food Qual. Pref. 17, 266Ð276. Gay, C. and Mead, R. (1992) A statistical appraisal of the problem of sensory measurement. J. Sens. Stud. 7, 205Ð228. Giovanni, M.E. and Pangborn, R.M. (1983) Measurement of taste intensity and degree of liking of beverages by graphic scales and magnitude estimation. J. Food Sci. 48, 1175Ð1182. Green, B.G. (1989) Capsacin sensitization and desensitization on the tongue produced by brief exposures to a low concentration. Neurosci. Lett. 107, 173Ð178. Green, B.G. and Lawless, H.T. (1991) The psychophysics of somatosensory chemoreception in the nose and mouth. In: Smell and Taste in Health and Disease (eds T.V. Getchell, R.L. Doty, L.M. Bartoshuk and J.B. Snow), Raven Press, New York, pp. 235Ð254. Green, B.G., Shaffer, S. and Gilmore, M. (1993) Derivation and evaluation of a semantic scale of oral sensation magnitude with apparent ratio properties. Chem. Senses 18, 683Ð702. Guadagni, D.B., Miers, J.C. and Venstrom, D. (1968) Methyl sulfide concentration, odour intensity and aroma quality in canned tomato juice. Food Technol. 22, 1003Ð1006. Guth, H. (1997) Quantitation and sensory studies of character impact odourants of different white wine varieties. J. Agric. Food Chem. 45, 3027Ð3032. Hall, M.J., Bartoshuk, L.M., Cain, W.S. and Stevens, J.C. (1975) PTC blindness and the taste of caffeine. Nature 253, 442Ð443. Hanaoka, K., Sieffermann, J.M. and Giampaoli, P. (2000) Effects of the sniffing port air makeup in gas chromatographyÐolfactometry. J. Agric. Food Chem. 48, 2368Ð2371. Hanaoka, K., Vallet, N., Giampaoli, P., Heyd, B. and MacLeod, P. (2001) Possible influence of breathing on detection frequency and intensity rating in gas chromatographyÐolfactometry. Food Chem. 72, 97Ð103. Heymann, H., Holt, D.L. and Cliff, M.A. (1993) Measurement of flavor by sensory descriptive techniques. In: Flavor Measurement (eds C.-T. Ho and C.H. Manley), Marcel Dekker, New York, pp. 113Ð132. Heymann, H. and Noble, A.C. (1989) Comparison of canonical variate and principal component analyses of wine descriptive analysis data. J. Food Sci. 54, 1355Ð1358. Hootman, R.C.E. (1992) Manual on Descriptive Analysis Testing for Sensory Evaluation, American Society for Testing and Materials, Philadelphia, PA. Hornung, D.E. and Enns, M.P. (1984) Independence and integration of olfaction and taste. Chem. Senses 9, 97Ð106. Husson, F., Le Dien, S. and Pages` (2001) Which value can be granted to sensory profiles given by consumers? Methodology and results. Food Qual. Pref. 12, 291Ð296. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

316 Food Flavour Technology

Jaeger, S.R. (2006) Non-sensory factors in sensory science research. Food Qual. Pref. 17, 132Ð144. Karrer, T. and Bartoshuk, L.M. (1991) Capsaicin desensitization and recovery on the human tongue. Physiol. Behav. 49, 757Ð764. King, S.C., Meiselman, H.L., Hottenstein, A.W., Work, T.M. and Cronk, V. (2007) The effects of contextual variables on food acceptability: a confirmatory study. Food Qual. Pref. 18, 58Ð65. Labbe, D., Schlich, P., Pineau, N., Gilbert, F. and Martin, N. (2009) Temporal dominance of sensations and sensory profiling: a comparative study. Food Qual. Pref. 20, 216Ð2211. Lawless, H.T. and Heymann, H. (1998) Sensory Evaluation of Food: Principles and Practices, Chapman & Hall, New York. Le Guen, S., Prost, C. and Demaimay, M. (2000) Critical comparison of three olfactometric methods for the identification of the most potent odourants in cooked mussels (Mytilus edulis). J. Agric. Food Chem. 48, 1307Ð1314. Lesschaeve, I. (2006) The use of sensory descriptive analysis to gain a better understanding of consumer wine language. In: Proceedings of the 3rd International Wine Business and Marketing Research Conference (ed. AWBR), ENSAM, Montpellier, France. Lesschaeve, I., Norris, L.N. and Lee, T.H. (2002) Defining and targeting consumer preferences. In: Proc. 11th Aust. Wine Ind. Tech. Conf. (eds R.J. Blair, P.J. Williams and P.B. H¿j), 7Ð11 October 2001, Adelaide, SA, Aust. Wine Ind. Tech. Conf. Inc. Linforth, R.S.T., Baek, I. and Taylor, A.J. (1999) Simultaneous instrumental and sensory analysis of volatile release from gelatine and pectin/gelatine gels. Food Chem. 65, 77Ð83. Linforth, R.S.T. and Taylor, A.J. (1993) Measurement of volatile release in the mouth. Food Chem. 48, 115Ð120. Lipsey, M.W. (1990) Design Sensitivity: Statistical Power for Experimental Research, Sage Publications, Newbury Park, CA. Lundahl, D.S. and McDaniel, M.R. (1988) The panelist effect Ð fixed or random? J. Sens. Stud. 3, 113Ð 121. MacFie, H. and Bratchell, N. (1989) Designs to balance the effect of order of presentation and first-order carry-over effects in Hall tests. J. Sens. Stud. 4, 129Ð148. Marks, L.E., Stevens, J.C., Bartoshuk, L.M., Gent, J.F., Rifking, B. and Stone, V.K. (1988) Magnitude- matching: the measurement of taste and smell. Chem. Senses 13, 63Ð87. Martens, M. and Martens, H. (1986) Partial least squares regression. In: Statistical Procedures in Food Research (ed. J.R. Piggot), Elsevier Applied Science, London, pp. 292Ð359. McDaniel, M.R., Miranda-Lopez, R., Watson, B.T., Michaels, N.J. and Libbey, L.M. (1990) Pinot noir aroma: a sensory/gas chromatographic approach. In: Flavours and Off-flavours (ed. G. Charalambous), Elsevier Science, Amsterdam, pp. 23Ð25. McEwan, J.A. (1996) Preference mapping for product optimization. In: Multivariate Analysis of Data in Sensory Science (eds T. Naes and E. Risvik), Elsevier Science, New York, pp. 71Ð102. McNeill, K.L., Sanders, T.H. and Civille, G.V. (2000) Using focus groups to develop a quantitative consumer questionnaire for peanut butter. J. Sens. Stud. 15, 163Ð178. Meilgaard, M.S., Civille, G.V. and Carr, B.T. (2006) Sensory Evaluation Techniques, 4th edn, CRC Press, Boca Raton, FL. Meilgaard, M.S., Reid, D.S. and Wuborski, K.A. (1982) Reference standards for beer flavour terminology system. Am. Soc. Brew. Chem. 40, 119Ð128. Mela, D.J. (1989) Bitter taste intensity: the effect of tastant and thiourea taster. Chem. Senses 14, 131Ð135. Moskowitz, H.R. 1994. Food Concepts and Products: Just in Time Development, Food & Nutrition Press, Trumbull, CT. Munoz,˜ A.M., Civille, G.V. and Carr, B.T. (1992) Sensory Evaluation in Quality Control,VanNostrand Reinhold, New York. Murray, J.M. and Delahunty, C.M. (2000) Selection of standards to reference terms in a cheddar-type cheese flavour language. J. Sens. Stud. 15, 179Ð199. Network, E.S. (1996) European Sensory and Consumer Study. A Case Study on Coffee, Campden & Chor- leywood Food Research Association, Chipping Camden, UK. Noble, A.C. (1996) TasteÐaroma interactions. Trends Food Sci. Technol. 7, 439Ð444. Noble, A.C., Arnold, R.A., Buechsenstein, J., Leach, E.J., Schmidt, J.O. and Stern, P.M. (1987) Modification of a standardized system of wine aroma terminology. Am. J. Enol. Viticult. 38, 143Ð146. Noble, A.C., Matysiak, N.L. and Bonnans, S. (1991) Factors affecting the time-intensity parameters of sweetness. Food Technol. 45(11), 121Ð126. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

Sensory methods of ﬂavour analysis 317

Norris, M.B. (1982) Salivary and gustatory responses in aqueous solutions as a function of anionic species, titratable acidity and pH. MSc thesis, University of California, Davis, CA. O’Mahony, M. (1986) Sensory Evaluation of Food: Statistical Methods and Procedures, Marcel Dekker, New York. Opet, J.M. (1989) Effect of caffeine, ethanol, and sucrose on temporal perception of menthol. M.S. thesis, University of California, Davis, CA. Ott, A., Fay, L.B. and Chaintreau, A. (1997) Determination and origin of the aroma impact compounds of yogurt flavour. J. Agric. Food Chem. 45, 850Ð858. Pagliarini, E., Lembo, P. and Bertuccioli, M. (1991) Recent advancements in sensory analysis of cheese. Int. J. Food Sci. 2, 85Ð89. Pfeiffer, J.F., Boulton, R.B. and Noble, A.C. (2000) Modeling the sweetness response using time-intensity data. Food Qual. Pref. 11, 129Ð138. Pineau, N., Cordelle, S. and Schlich, P. (2004) Methode« d’acquisition, de codage et d’analyse de profiles sensoriels temporels. 8eme` journees« Agro-industrie et methodes« statistiques, March, 10Ð12, 87Ð94. Pollien, P., Fay, L.B., Baumgartner, M. and Chaintreau, A. (1999) First attempt of odorant quantitation using gas chromatography-olfactometry. Anal. Chem. 71, 5391Ð5397. Prescott, J., Norris, L.N., Kunst, M. and Kim, S. 2005. Estimating a ‘consumer rejection threshold’ for cork taint in white wine. Food Qual. Pref. 16, 345Ð349. Prutkin, J., Duffy, V.B., Etter, L., Gardner, E., Lucchina, L.A., Snyder, D.J., Tie, K., Weiffenbach, J. and Bartoshuk, L.M. (2000) Genetic variation and inferences about perceived taste intensity in mice and men. Physiol. Behav. 69, 161Ð173. Roessler, E.B., Pangborn, R.M., Sidel, J.L. and Stone, H. (1978) Expanded tables for estimating significance in paired-preference, paired-difference, duo-trio and triangle tests. J. Food Sci. 43, 940Ð947. Schlich, P. (1993) Uses of change-over designs and repeated measurements in sensory and consumer studies. Food Qual. Pref. 4, 223Ð235. Schlich, P. (1994) Graphical representation of assessors performances. J. Sens. Stud. 9, 157Ð169. Schlich, P. (1995) Preference mapping: relating consumer preferences to sensory or instrumental measurements. In: Bioflavour’95 (eds P. Etievant and P. Schreier), INRA Editions, Paris, pp. 135Ð150. Schlich, P. and Guichard, E. (1989) Selection and classification of volatile compounds of apricot using the RV coefficient. J. Agric. Food Chem. 37, 142Ð150. Schlich, P., Issanchou, S., Guichard, E., Etievant, P. and Adda, J. (1987) RV coefficient: a new approach to select variables in PCA and get correlations between sensory and instrumental data. In: Flavour Science and Technology (eds M. Martens, G.A. Dalen and H. Russwurm), Wiley, New York, pp. 469Ð473. Schlich, P., Lespinasse, N. and Navez, B.D.S. (2003) Preference mapping of tomato varieties from the French market. A PrefMaX consumer segmentation validated by a two-year study. In: Proceedings of the 5th Pangborn Sensory Science Symposium (eds H. Meiselman, A.V. Cardello and R. Bell), Elsevier Science, Boston, MA. Shortread, G.W., Richards, P., Swan, J.S. and Burtles, S. (1979) The flavour terminology of Scotch whiskey. Brewers Guardian, November, 2Ð6. Song, W.L. (1990) Sensory and chemical evaluation of aroma formed via Maillard reaction upon microwave irradiation and conventional heating, MSc thesis, University of California, Davis, CA. Stahl, W.H. (ed.) (1978) Compilation of Odour and Taste Threshold Values Data, American Society for Testing and Materials, Philadelphia, PA. Stone, H. and Sidel, J. (1985) Sensory Evaluation Practices, Academic Press, New York. Stone, H., Sidel, J., Oliver, S., Woolsey, A. and Singleton, R.C. (1974) Sensory evaluation by quantitative descriptive analysis. Food Technol. 28(11), 24Ð34. Thomson, D.M.H. and McEwan, J.A. (1988) An application of the repertory grid method to investigate consumer perceptions of foods. Appetite 10, 181Ð193. Ullrich, F. and Grosch, W. (1987) Identification of most intense volatile flavour compounds formed during autoxidation of linoleic acid. Z. Lebensm.-Unters.-Forsch. 184, 277Ð282. Van Ruth, S.M., Roozen, J.P. and Posthumus, M.A. (1995) Instrumental and sensory evaluation of the flavour of dried French beans (Phaseolus vulgaris) influenced by storage conditions. J. Sci. Food Agric. 69, 393Ð401. Williams, A.A. and Arnold, G.M. (1985) A comparison of the aromas of 6 coffees characterised by conventional profiling, free-choice profiling and similarity scaling methods. J. Sci. Food Agric. 36, 204Ð214. Williams, A.A. and Langron, S.P. (1984) The use of free-choice profiling for the evaluation of commercial Ports. J. Sci. Food Agric. 35, 558Ð568. P1: SFK/UKS P2: SFK c11 BLBK221-Taylor/Linforth November 21, 2009 8:24 Printer Name: Yet to Come

318 Food Flavour Technology

Williams, A.A., Rogers, C., and Noble, A.C. (1984) Characterization of ﬂavouring alcoholic beverages. In: Flavour Research of Alcoholic Beverages. Instrumental and Sensory Analysis (eds L. Nykanen and P. Lehtonen), Foundation for Biotechnical and Industrial Research, Helsinki, pp. 235Ð254. Yantis, J.E. (ed). (1992) The Role of Sensory Analysis in Quality Control, American Society for Testing and Materials, Philadelphia, PA. Zellner, B.D., Dugo, P., Dugo, G. and Mondello, L. (2008) Gas chromatography-olfactometry in food ﬂavour analysis. J. Chromatog. A 1186, 123Ð143. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

12 Brain imaging Luca Marciani, Sally Eldeghaidy, Robin C. Spiller, Penny A. Gowland and Susan T. Francis

12.1 INTRODUCTION

Flavour scientists use a wide range of increasingly sophisticated analytical tools to characterise their samples and assess chemical senses at the receptor level, varying from ‘classic’ bench instrument techniques to more recent methods of determining the availability of tastants and volatiles at the receptor level, such as mouth swabbing and in-nose atmospheric pressure chemical ionisation mass spectrometry (Taylor et al., 2000). Sensory scientists assess the final perceptual output, for example using panels of trained assessors. Until recently, there has remained a frustrating investigational gap at the interface between these two sciences: flavour input and output scores reported by assessors. This is because the interface occurs at the subject’s brain, which has effectively been a ‘black box’. Signals from taste, olfactory and somatosensory receptors travel via nerve afferents to primary, secondary and the higher-order brain areas responsible for detecting stimuli and forming associations, decision-making and emotional responses to the flavour experienced. Improved understanding of the neural pathways involved in flavour processing has been facilitated by parallel advances with animal models and human studies. Invasive single-cell recordings and surface electroencephalography (EEG) have facilitated the development of animal models. These studies provide data on cell reactivity to characterise a particular cortical region with millisecond temporal resolution, but they give little detail on how information is processed at several levels. More recently in the past decade, an explosion in the availability and use of non-invasive human brain mapping techniques has allowed serial studies of hierarchical cortical processing in alert healthy human volunteers. Interest in applying brain imaging techniques to study the chemical senses is rapidly growing and to date over 400 studies using different methodologies have been published. The majority of these studies explore the cortical representation of individual sensory attributes of uni-modal stimuli, but there has been recent growth in the study of how the ‘cross-modal’ integration of sensory inputs in flavour perception is represented in the human brain. This chapter provides a brief outline to human brain anatomy, the cortical pathways used in the experience of taste, aroma and texture and their integration. Functional brain imaging techniques are reviewed, with particular focus on functional magnetic resonance imaging (fMRI) and aspects of experimental design, data acquisition and data processing. This is followed by a brief review of brain imaging studies of the cortical representation of taste and aroma, from early studies investigating the cortical response to pleasant and aversive taste stimuli to recent studies on multi-modal perception and the cortical representation of the oral perception of fat. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

320 Food Flavour Technology

12.2 CORTICAL PATHWAYS OF TASTE, AROMA AND ORAL SOMATOSENSATION

Various types of sensory receptors are responsible for the detection of ﬂavour molecules and sample texture. The transduction of taste and olfaction occurs via chemo- and sensory receptors, inducing a change in the electrochemical membrane potential. This then leads to depolarisation and generation of action potentials in afferent nerve ﬁbres. These potentials then project, via separate neuronal pathways, into the brain to be decoded. The following sections describe the pathways involved in gustatory (taste), olfactory (odour), and somatosensory processing and their integration.

12.2.1 Basic brain anatomy and function The brain is the most complex organ within the nervous system. Brain structure and function are briefly outlined here, but for a more detailed discussion the reader is referred to Kandel et al. (2000) and Bear et al. (2007). The brain and spinal cord constitute the central nervous system. The extensive network of sensory and motor neurons that transmits sensory information from receptors on the body’s surface, internal organs and muscles to the brain forms the peripheral nervous system (PNS). The main pathway for transferring information between the brain and the PNS is the spinal cord. The brain consists of four principal sections – the brainstem, cerebellum, diencephalon and cerebrum – as illustrated in Fig. 12.1. The brainstem is the lower extension of the brain, which directly connects the brain with the spinal cord. Every message transmitted between the brain and spinal cord passes through the medulla oblongata – a part of the brainstem. The majority of information entering the brain via nerve fibres of the PNS is transmitted by nerve fibres, which decussate (cross over from one side to another) within the spinal cord or the medulla of the brainstem before being conveyed to the cerebrum. Consequently, the right side (hemisphere) of the cerebrum is generally concerned with the left side of the body, whilst the left hemisphere governs the right side of the body. The cerebellum is located below the cerebrum and behind the brainstem, and plays an important role in the integration

(a) (b) Cerebrum Insula lobe Frontal lobe

Corpus Frontal lobe callosum Parietal lobe

Occipitial lobe

Diencephalon Occipitial lobe Brainstem Cerebellum Temporal lobe

Fig. 12.1 Principal sections in the brain and the major lobes in the cerebrum: (a) sagittal and (b) axial sections. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 321

of sensory perception. The diencephalon lies above the brainstem and embodies the thalamus and hypothalamus. The thalamus is an important relay station for sensory information and is discussed below with respect to each of the sensory modalities. The cerebrum is the largest part of the brain, constituting nearly 90% of the total brain mass, and lies over the other brain structures. It has been mapped out on the basis of the cortical cytoarchitectural, or organisation of cells, into 52 numbered Brodmann’s areas. Functionally, the cerebrum is divided into specific areas that interpret sensory impulses. The cerebrum consists of three major components: the cerebral cortex, which is the cerebrum’s outer layer composed of grey matter made up of nerve cell bodies; deep to the cortex lies the limbic system (including amygdala, hippocampus and cingulate gyrus) and the basal ganglia. The cerebrum is divided into left and right hemispheres. The two hemispheres have the same general appearance and are separated by a deep fissure connected at the base by the corpus callosum, a thick bundle of white matter fibres that transfers information between the two hemispheres. Each hemisphere of the cerebrum is subdivided into four major lobes: frontal, parietal, temporal and occipital lobes. A fifth lobe, the insula, is buried between the temporal and frontal lobes (Fig. 12.1b).

12.2.2 Central gustatory pathways The electrical signals generated in the gustatory axons (thin elongated fibres that wind themselves around taste receptors cells) are transmitted to the brain by three cranial nerves: cranial nerve VII (facial nerve), cranial nerve IX (glossopharyngeal nerve) and cranial nerve X (vagus nerve). Cranial nerve VII is innervated by taste buds on the anterior two-thirds of the tongue and palate and transmits information regarding the identity and quantity of tastants. Cranial nerve IX innervates the posterior third of the tongue and is mostly responsive to swallowing of tastants, whilst cranial nerve X is innervated by taste buds on the epiglottis and pharynx. These cranial nerves, together with cranial nerve V (trigeminal nerve), relay somatosensory information from the afferents that innervate the tongue and relate information on physical properties of food such as pressure and temperature. Cranial nerves VII, IX and X enter the brainstem and then unite at the rostral portion of the nucleus of the solitary tract (NST) in the medulla oblongata, which is also responsible for swallowing and salivatory reflexes. From here taste projects to the cortex via two main pathways (Fig. 12.2a). In the first taste pathway, the signal projects directly to the ventral posterior medial (VPM) nucleus of the thalamus. The VPM nucleus contains a topographic representation of the oral cavity. Signals then pass from the VPM and terminate in the dorsal part of the anterior insula in the frontal operculum (FO), which is the primary gustatory cortex, and in the Brodmann’s areas of the somatosensory cortex (areas 1, 2, 3a and 3b) in the post-central gyrus. The insula cortex is a triangular region beneath the frontal, parietal and temporal lobes of the cerebral cortex (Fig. 12.2b) subdivided into three principal areas (anterior, middle and posterior) (Kandel et al., 2000) comprising Brodmann’s areas 13–16. The insula is richly connected to the cerebral cortex, the basal ganglia and the limbic structures. The anterior insula is connected to the frontal lobe, and receives a direct projection from the VPM nucleus of the thalamus and the central nucleus of the amygdala. The posterior insula is connected to the parietal and temporal lobes, receiving input from the VPM nucleus of the thalamus. The insula projects to cortical areas and receives reciprocal afferent projections from gustatory, olfactory and somatosensory areas. The insula is thought to be involved P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

322 Food Flavour Technology Amygdala Thalamus ACC Insula Post-central gyrus Post-central (b) (b) Brain areas correlated with taste perception. CN X CN IX CN VII OFC Amygdala Secondary gustatory cortex ACC Amygdala Hypothalamus Taste buds Taste Taste buds Taste and palate) Taste buds Taste (epiglottis and pharynx) (posterior third of tongue) (anterior of tongue two-thirds Anterior insula/ frontal operculum Post-central gyrus Post-central Primary gustatory cortex Nucleus of pons in medulla (a) Schematic diagram of central gustatory (taste) pathways. VPM of thalamus Nucleus of solitary tract (a) Fig. 12.2 P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 323

in conscious taste perception, including identifying taste qualities. The insula then projects to the secondary taste cortex (Baylis et al., 1995), the central nucleus of the amygdala and anteriorly to the caudolateral orbitofrontal cortex (OFC). The OFC is part of the ventral surface of the prefrontal cortex, deﬁned by Brodmann’s areas 10, 11 and 47. It receives inputs from taste, olfactory, somatosensory, visual and auditory stimulation, and also receives inputs via the amygdala, thalamus, cingulate cortex, hypothalamus, hippocampus and dorsolateral prefrontal cortex. The OFC projects to temporal lobe, hippocampus, hypothalamus and cingulate cortex. The OFC has been shown from electrophysiology studies to act as a higher order taste centre involved in encoding affective (emotional) values of the taste stimulus. It is believed to be modulated by motivational state, responding to hunger and not to satiety (Del Parigi et al., 2002). It has also been postulated that the lateral part of the OFC may play a role in hedonic judgements (Zald and Pardo, 2000). The second taste pathway is from the NST in the medulla to the nucleus of the pons in the cerebellum, which then projects to the limbic system including amygdala, anterior cingulate cortex (ACC) and lateral hypothalamus. The limbic system is thought to be involved in affective pleasure or displeasure to particular tastes. The cingulate cortex is located medially above the corpus callosum (Fig. 12.2b) and is subdivided into several Brodmann’s areas (the main areas relevant to ﬂavour being Brodmann’s areas 23, 25, 31 and 32). The ACC has been shown to play an important role in autonomic and cognitive functions including working memory and anticipation of reward. It has also been shown to have a ‘hedonic’ response to emotions induced by odour- or pain-related stimuli. The amygdala nuclei are a major component of the limbic system and are located at the tip of the temporal lobe. The amygdala receives multiple projections from the olfactory bulb, hypothalamus, thalamus, prefrontal, temporal and occipital lobes, insula and olfactory cortices, cingulate and brainstem including NST. It projects to frontal, prefrontal, insula, cingulate and orbitofrontal cortices, as well as thalamus and hypothalamus. The amygdala’s main role is the formation and storage of memories associated with emotional events. In addition to the negative emotional reaction, it also plays a major role in pleasure and reward (such as to the taste of food or odour) and in the control of food intake (JonesGotman et al., 1997; Zald and Pardo, 1997; Zald et al., 1998). Figure 12.2b shows those brain areas correlated with taste perception.

12.2.3 Central olfactory pathways Odours are perceived through the mouth (retronasally) or through the nose (orthonasally). Following exposure to an odour, electrical signals generated in the olfactory epithelium are transmitted by the cranial nerve I (olfactory nerve) to the olfactory bulb. In addition to the olfactory system, some odours such as ammonia, menthol or chlorine directly stimulate cranial nerve V (trigeminal nerve) and are projected to the somatosensory cortex (see Section 12.2.4). The olfactory sense is unique among sensory systems in that the olfactory bulb projects directly to the primary olfactory cortex (POC) without passing through the thalamus (Fig. 12.3a). The POC is located in the ventral forebrain (Fig. 12.3b) and consists of the piriform cortex, anterior cortical amygdaloid nucleus (ACo nucleus), periamygdaloid cortex, anterior olfactory nucleus and olfactory tubercle. The POC is concerned with conscious perception and odourant discrimination. The piriform cortex is the largest and most prominent component of the olfactory cortex. It is located along the lateral olfactory tract on the caudolateral part of the orbital cortex and projects directly to higher order areas including P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

324 Food Flavour Technology cortex Amygdala (b) (b) Primary olfactory Hippocampus Orbital cortex Insula Ventral striatum Entorhinal Amygdala Thalamus frontal cortex Hypothalamus (b) Primary olfactory cortex. cortex (POC) Primary olfactory Olfactory tract Olfactory bulb (a) Schematic diagram of central olfactory pathways. Olfactory receptors (a) Fig. 12.3 P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 325

the OFC, insula and thalamus. It has been suggested that information is represented in a topographical fashion in the piriform cortex (Gottfried et al., 2006). From the POC, olfactory information projects either directly to the secondary olfactory cortex (insula, ventral striatum, entorhinal cortex and OFC) or via a relay in the mediodorsal nucleus of the thalamus. Thalamic connections are thought to serve as a conscious mechanism for odour perception. The POC also projects directly to the hypothalamus and to the limbic system, including hippocampus, and the amygdala (Fig. 12.3a). The entorhinal cortex, Brodmann’s area 28, is the major source of input to the hippocampus, which, in turn, projects to the cortex. The limbic system controls the emotion, learning and behaviour responses to olfaction. As for taste stimuli, the lateral OFC and amygdala are thought to play a role in hedonic judgements.

12.2.4 Central oral somatosensory pathways Cranial nerves IX and X convey oral somatosensory sensations from the posterior third of the tongue and the pharynx, in addition to conveying taste information. Sensory information from the anterior two-thirds of the tongue and the nasal cavity is conveyed by cranial nerve V (the trigeminal nerve), the largest cranial nerve. Its primary function is sensory, in addition to certain motor functions, such as biting, chewing and swallowing. Somatosensory information transmitted by cranial nerves VII, IX and X projects to the NST, which then, together with information from the trigeminal nerve, converges in the thalamus (Fig. 12.4a). In addition to projections to the NST, information from the trigeminal nerve projects to the trigeminal nucleus. This conveys sensations such as tactile, proprioceptive, pain and heat, as well as impulses from the stretch receptors of mastication muscles. Sensory information from the trigeminal nucleus unites with those from the NST in the VPM nucleus of the thalamus, which in turn projects to the primary sensory cortex (SI) in the post-central gyrus of the parietal lobe and the secondary somatosensory cortex (SII) (Fig. 12.4b). Information from SI projects to the secondary somatosensory cortex (SII), AAC and insula. SI and SII are concerned with receiving and processing somatosensory information together with the insula, OFC and ACC. SI spans the post-central gyrus, deﬁned by Brodmann’s areas 1, 2, 3a and 3b, and receives projections from the VPM and ventral posterior lateral (VPL) nuclei of the thalamus, and projects to the primary motor cortex and the associated somatosensory areas (including SII). SI has a topographical representation of the whole body surface dependent on the number of neurons, referred to as the ‘somatosensory homunculus’. The pharynx, tongue, jaw and lips occupy a large portion of the homunculus and are represented in the most ventral portion. SII is located close to SI in the upper bank of the lateral sulcus, Brodmann’s areas 40 and 43. It receives inputs mainly from SI and directly from the thalamus. It projects back to SI (areas 1 and 3b), the primary motor cortex, and the posterior insula (Fig. 12.4). Somatosensory association areas are found in the superior/posterior parietal lobe, Brodmann’s areas 5 and 7. They receive inputs from the SI and the thalamus, and play an important role in reinforcing the perception and identiﬁcation of shape, size and texture of the stimulus.

12.2.5 Interaction and association of stimuli Accumulated evidence from psychophysics, electrophysiology and more recently brain imaging suggests cross-modal interactions between taste, aroma and somatosensation of foods. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

326 Food Flavour Technology cortex Secondary somatosensory SII). cortex Primary (b) somatosensory SII OFC ACC Insula cortex Primary motor CN X (b) Primary and secondary somatosensory areas (SI and CN V CN IX Sl and pharynx Trigeminal nucleus Trigeminal Posterior third of tongue Anterior two-thirds of tongue in medulla VPM of thalamus (a) Schematic diagram of oral somatosensory pathways. Nucleus of solitary tract (a) Fig. 12.4 P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 327

Since these sensory systems described above are anatomically dissociated, it is likely that interactions occur at the central rather than peripheral receptor level (Noble, 1996). The independent presentation of taste and retronasal/orthonasal odour stimuli produces overlap- ping activation in insula, OFC, amygdala and ACC. Several psychophysical studies have investigated taste–aroma interactions (Frank and Byram, 1988; Dalton et al., 2000; Hort and Hollowood, 2004) and show the degree of interaction depends on the congruency of the taste–aroma pairing. These psychophysical findings support evidence from primate electrophysiological studies, which have demonstrated taste–aroma cross-modal integration. Rolls and Baylis (1994) showed the presence of bimodal taste–aroma neurons in the OFC that respond selectively to both taste and aroma stimulations when presented independently (uni- modal) or simultaneously (combined). These findings provide evidence of the convergence of taste and aroma inputs to produce flavour perception.

12.3 IMAGING OF BRAIN FUNCTION

Although animal models, electrophysiology and neurosurgery have contributed much to the knowledge of the brain, it is the recent ability to non-invasively image brain function in response to various tasks in healthy, conscious individuals that has opened up new and exciting avenues for research. The magnetoelectric techniques of electroencephalography (EEG) and magnetoencephalography (MEG), and the haemodynamic techniques of positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) are described below.

12.3.1 Methodologies to image brain function EEG is a non-invasive technique that directly measures the electrical activity caused by currents associated with neuronal firing. It is the oldest brain mapping technique, with the first recordings demonstrated in animals in the late-nineteenth century. Electrical signals (potentials) are detected by electrodes (commonly between 32 and 128 electrodes) placed on the scalp. The potentials are thought to reflect the summation of electrical fields generated from large populations of pyramidal neurons that are aligned perpendicular to the local cortical surface. EEG can measure the potentials evoked by a series of well time-locked stimuli, called event-related potentials (Luck, 2005), or changes in the oscillatory power in the EEG signal. EEG measurements have temporal resolution of the order of milliseconds; however, their spatial localisation is poor, with signals arising mostly from the cortex close to the scalp. Typically, a large number of repeats are necessary for good signal-to-noise ratio (SNR). Within these limitations, EEG is non-invasive, relatively cheap and has been applied to study gustatory and olfactory responses (Kobal, 1985; Kobal et al., 1992; Brauchli et al., 1995; Mizoguchi et al., 2002). MEG measures the magnetic field generated by the electrical current flow within active neurons (Cohen, 1968). These magnetic fields are very small, typically 50–500 fT. They can be detected using a large array of superconducting quantum interference devices, typically approximately 300, arranged around the head to provide whole brain coverage. MEG is completely non-invasive and has excellent submillisecond temporal resolution. A limitation of MEG is its poor spatial resolution as localisation relies on solving the ‘inverse problem’, i.e. localising the source of the MEG signal recorded around the scalp, though the spatial resolution of MEG is higher than of EEG. As for EEG, MEG signals are better resolved from P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

328 Food Flavour Technology

more superficial cortex than from deep structures in the brain. In contrast to EEG, which can detect electric fields from current sources oriented tangentially and radially to the skull, MEG detects only tangential current sources with radially oriented sources providing no useful MEG signal. MEG has been used to investigate cortical activity in response to gustation (Kobayakawa et al., 1996, 1999; Saito et al., 1998; Yamamoto et al., 2003). Simultaneous EEG and MEG recordings have been performed by Mizoguchi et al. (2002). PET was developed in the mid-1970s (Phelps et al., 1975) and is now an established technique in clinical oncology, but also provides an indirect haemodynamic measure of brain activation. PET is considered to be invasive neuroimaging experiment, since it involves an injection of a radioactive label (a biologically active molecule containing a short-lived radioactive, positron-emitting nucleus such as 15O) into a vein. On decay, the nucleus emits two photons (gamma rays) with equal energy, which travel in opposite directions. These are detected by an array of scintillator detectors, and data are reconstructed to create a three- dimensional image showing activated brain areas. Cerebral oxygen, glucose metabolisms, regional cerebral blood flow (CBF) – all increase in the cortical areas activated by a task, and these can be monitored by changes in the local concentration of different labels. PET has been used to determine changes in regional blood flow in response to taste and aroma (Kinomura et al., 1994; Gautier et al., 1999; Savic et al., 2000; Zald et al., 2002). PET has various limitations; radiation exposure limits its repeated use and it has also low spatial resolution and poor temporal resolution due to the short half-life of the radioactive label. The focus of the following section will be on fMRI. We encourage the reader to refer to the excellent textbooks available that provide a more rigorous and detailed discussion of fMRI (Baert et al., 2000; Huettel et al., 2004; Jezzard et al., 2004).

12.3.2 Functional magnetic resonance imaging fMRI using the Blood Oxygenation Level Dependent (BOLD) contrast is now the most widely used method for mapping brain activity. Shortly after the ﬁrst observations of BOLD in animals in the late-1980s by Ogawa and colleagues (Ogawa et al., 1990), this contrast was successfully used to generate functional images in the human brain (Bandettini et al., 1992; Kwong et al., 1992). Since then BOLD has rapidly evolved, opening a new era in human brain mapping, which has led to thousands of papers to date. fMRI measures brain activity indirectly by detecting changes in cerebral haemodynamics that are associated with neural activation; the origin of this signal is described further in Section 12.3.2.1. fMRI has the advantage of being a non-invasive technique based on endogenous contrast, allowing serial studies. It has high spatial resolution (typically ∼3to 5 mm isotropic). The main limitation of fMRI is that it arises from the vascular response to neuronal activity, which limits its inherent temporal resolution. Furthermore, the exact physiological origin of BOLD contrast is incompletely understood and the subject of much research to date (Hoge et al., 1999; Buxton et al., 2004).

12.3.2.1 BOLD response During brain activity, neural demand increases leading to increased glucose and oxygen consumption. Increased neuronal activity is associated with a local haemodynamic response, involving vessel dilation and an increase in both cerebral blood ﬂow (CBF) and cerebral blood volume (CBV). However, the increase in blood ﬂow overcompensates the oxygen P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 329

Fig. 12.5 Schematic representation of haemodynamic changes, which lead to a BOLD signal.

delivery needed to match increased oxygen consumption (CMRO2) (Fig. 12.5). Conse- quently, the local ratio of oxyhaemoglobin to deoxyhaemoglobin concentration increases in activated brain regions. The MRI signal is strongly influenced by the oxygenation state of the blood, as oxygenated and deoxygenated bloods have very different magnetic properties (Pauling and Coryell, 1936). It is this which leads to the endogenous BOLD contrast. Haemoglobin is the protein within red blood cells to which oxygen binds. When oxygen is bound to haemoglobin this forms oxyhaemoglobin. Oxyhaemoglobin has magnetic properties similar to the surrounding brain tissue (diamagnetic), whilst deoxyhaemoglobin, with no bound oxygen molecules, is paramagnetic, distorting the magnetic field (inducing magnetic field inhomogeneity) around vessels. This field inhomogeneity induced by deoxyhaemoglobin increases the rate of intravoxel spin dephasing (decreasing the apparent spin–spin transverse relaxation time, T2*), leading to a reduction in the MRI signal in both the blood vessels and surrounding tissue. The magnitude of this distortion increases with the amount of paramagnetic deoxyhaemoglobin. In cortical regions at rest, there is a relatively high deoxyhaemoglobin concentration, on activation the increase in oxygenation lengthens T2* resulting in a slight increase in MRI signal (brighter image in activated region) (typically 1–5% at 3 T depending on the area of the brain). The BOLD signal is typically observed using a gradient-echo echo-planar imaging (GE-EPI) MRI sequence, which is T2*-weighted. The time course of the observed BOLD signal is known as the haemodynamic response function (HRF). The HRF has been suggested to comprise three distinct components: an initial dip, a positive peak and a post-stimulus undershoot (Fig. 12.6). The initial dip has been reported to occur immediately after stimulation, to be small in amplitude and to last for 1–2 seconds (Menon et al., 1995). Its existence is still under debate, and it is thought to reflect an immediate increase in oxygen consumption in the capillary bed close to the site of neuronal activity. This is followed by a large positive rise that peaks after approximately 6 seconds. This is the main feature of the HRF and results from the increased CBV and CBF, leading to oversupply of oxygen-rich blood and an increase in BOLD signal. This positive BOLD signal change is the component typically assessed using BOLD fMRI and can reach 5% at 3 T. The width of the peak and hence the shape of the HRF are proportional to the stimulus duration. After the end of the stimulus, the amplitude of the BOLD signal declines and displays a negative post-stimulus undershoot before returning to baseline; this can take of the order of 30 seconds. This post-stimulus undershoot is thought to arise from deoxygenated P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

330 Food Flavour Technology

Positive BOLD response

1 Post-stimulus

Signal change (%) undershoot

0 6 12 18 24 Time (second) initial dip

Stimulus

Fig. 12.6 Schematic representation of the BOLD HRF.

blood concentration remaining elevated, due to a mismatch either between CBV and CBF or between CBV and oxygen consumption (Kruger et al., 1996; Mandeville et al., 1998).

12.3.2.2 Data acquisition The basic concept of an fMRI experiment is for a person to lie inside an MRI scanner and perform a task whilst BOLD images are rapidly acquired to indirectly detect the increase in neuronal activity. A set of BOLD images covering the whole brain (a brain volume) is typically collected every 1–3 seconds, and (to increase sensitivity) hundreds of brain volumes are acquired during many repeats of a stimulus, with the entire fMRI paradigm lasting around 10–20 minutes. The signal intensity at each spatial position (voxel) within the image is compared to a model of the expected BOLD response to the paradigm, and any areas with signal changes correlated to the task can then be determined using statistical analyses. Signal averaging and statistical processing are required due to the difficulty in detecting low (1–5%) BOLD signal changes against a background physiological noise of a similar magnitude (Purdon and Weisskoff, 1998). BOLD sensitivity increases supralinearly with magnetic strength, which is one of the reasons for the current demand for ultra-high-field MRI scanners. Two main MRI sequence parameters play a key role in fMRI data acquisition. The first is the echo time (TE), which is the time between the excitation pulse and the signal collection in the EPI sequence. In GE-EPI, the optimal echo time to maximise BOLD sensitivity is approximately equal to the T2* of grey matter (Fera et al., 2004). This is typically 30–40 milliseconds at 3 T, but different brain areas can have very different T2* values, such as the OFC and insula, which play an important role in flavour processing (Hagberg et al., 2002). Therefore, the common method of acquiring images at a single echo time (TE) in conventional fMRI studies may limit the BOLD sensitivity to a narrow range of T2* values and to specific brain areas. Figure 12.7 demonstrates that a short echo time enhances the BOLD sensitivity in areas of short T2*, such as the orbitofrontal cortex (a secondary taste cortex with short T2*), whilst other cortical areas of interest, which have relatively long T2*, P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 331

(a)

Improved BOLD Signal dropout in sensitivity in OFC OFC

Short echo time Long echo time

(b) 0.09 T2* = 25 millisecond 0.08 T2* = 45 millisecond 0.07

0.06

0.05

0.04

0.03 Signal change (a.u)

0.02

0.01

0 0 20 40 60 80 100 120 140 TE (millisecond)

Fig. 12.7 (a) Areas with a short T2* such as the OFC require a short TE to optimise sensitivity. (b) The effect of echo time on the BOLD signal change for areas with short and long T2*.

such as the primary taste cortex, require a longer echo time to optimise BOLD sensitivity. The use of multi- or dual-echo EPI sequence where two or more echo times are acquired following a single excitation pulse provides increased sensitivity across cortical areas with both short and long T2* (Poser et al., 2006; Gowland and Bowtell, 2007). The images generated from each echo can then be combined using a weighted summation. The use of a dual-echo EPI sequence has been used to study ﬂavour processing in brain regions including the OFC and P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

332 Food Flavour Technology

amygdala (Marciani et al., 2006). This is the method that is favoured in our laboratory to study ﬂavour perception. The second key sequence parameter is the repetition time, TR, the time between two successive acquisitions of the same brain images (brain volumes). TR will dictate the temporal resolution with which the BOLD signal change can be sampled and must be traded against the required volume coverage (how many images through the brain can be sampled). It is worth remembering that the choice of spatial resolution will impact also on the available signal-to-noise ratio (SNR) and the acquisition time (TA) and hence the temporal resolution of the experiment.

12.3.2.3 Paradigm design The BOLD response yields a small percentage signal change, which is also affected by scanner noise and physiological noise (due to respiration and heart beat). Thus, an fMRI paradigm requires many repeats (trials or cycles) of a paradigm to increase the contrast-to- noise ratio. fMRI paradigms use block, event-related (ER) or mixed designs (Amaro and Barker, 2006) (Fig. 12.8). A typical block design consists of two or more conditions, which are presented for discrete periods (epochs). The simplest design consists of a block of stimulus presentation (on period typically of 10–15 seconds) followed by a block of rest (off or baseline period), to allow the haemodynamic response to return to baseline before another block of stimuli is presented. The alternation of two conditions is shown in Fig. 12.8 as a ‘stimulus-rest (SR) block’, in which a ‘cycle’ corresponds to one epoch of each condition. Figure 12.8 illustrates five cycles, but many cycles are repeated in a typical fMRI experiment. The signal acquired during the on period is then compared to the off period (or a different stimulus condition). Thus, the resulting BOLD signal will reflect the magnitude of neural activity in one condition versus another. This design dominated the early years of fMRI experimentation, as it has high statistical power (Price et al., 1999) and produces relatively large BOLD signal changes. However, it is difficult to sample the shape of the HRF using such paradigms, and repeated delivery of a stimulus may lead to adaptation effects, particularly relevant when studying flavour processing. In the mid-1990s, a second type of paradigm making use of the fast image acquisition of fMRI emerged, that of ER designs (Humberstone et al., 1997; Friston et al., 1998). In this design, stimuli are presented as isolated brief events of stimuli (e.g. for 1 second) separated in time from one another. The time between the start of each stimuli presentation is termed the interstimulus interval (ISI). The main advantage of ER compared to the block design is the ability to detect transient variations in individual haemodynamic responses, allowing the temporal characterisation of BOLD signal changes to each event, and due to the shortened nature of an ER design it is possible to study many stimuli in a randomised order. Inevitably, the ER design does lead to a reduction in statistical power. Many studies are now performed using a mixed design, which is a hybrid of these two techniques using a short stimulus (3–5 seconds) duration. This method is ideal for studies of flavour as this reduces the stimulus delivery to the subject and limits stimulus adaptation effects. Rather than deliver a single stimulus of a given level (intensity, concentration), it is possible to perform tasks or deliver stimuli at different levels in a parametric design. The idea of increasing the stimulus intensity associated with a particular task, without modifying its intrinsic nature, is that only areas associated with the parameter being manipulated (e.g. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 333

S R S R S R S R S R Block design

SSSSS Event-related design

Mixed design RRR

Fig. 12.8 Illustration of block, event-related and mixed designs. S, stimulus phase; R, rest phase.

concentration of stimulus) will show a correlated increase in activity (Buchel et al., 1998). This allows the separation of stimulus intensity related areas from those areas involved in the ‘maintenance’ of a task (Jansma et al., 2000). A ﬁnal consideration for fMRI designs is temporal sampling of the BOLD signal. In block designs the on/off period is typically set to be an integer value of the TR period. Thus, all volumes in consecutive cycles are collected with the same delay from stimulus presentation (‘time-locked’). However, for mixed and ER designs the temporal resolution of fMRI acquisition is often improved by using ‘jittering’. ‘Jittering’ refers to the use of different delays (offsets) between the acquisition of a brain volume and the start of stimulus presentation, resulting in different time points being sampled at each cycle of stimulus presentation. This can be achieved by using an ISI that is not a multiple of the TR.

12.3.2.4 Data analysis Having collected data for an fMRI study in, for example, 15 subjects, the researcher will have collected over a half a million fMRI images. The volume of data acquired and to be processed in an fMRI experiment places high demands on computing power and speciﬁcation, commonly requiring parallel computing infrastructure. Extracting meaningful brain mapping information from these images is a challenge, which requires a series of post-processing steps to identify those areas of the brain (voxels) that show activity (a time course) correlated with the paradigm. fMRI data analysis can be divided into two stages: pre-processing and statistical analysis. Pre-processing involves a series of steps aimed at eliminating non-task- related variability in the experimental data; statistical analysis identiﬁes voxels, which show a response correlated with the stimulus. There is a multitude of software packages now available from different laboratories for data analysis, which adopt common steps, such as the Oxford centre for functional magnetic resonance imaging of the brain Software Library (FSL) and Analysis of Functional Neuroimages (AFNI). P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

334 Food Flavour Technology

The first problem encountered with an fMRI data set is that the image slices in a volume are acquired in a TR period (typically 2–3 seconds) in ascending, descending or interleaved spatial ordering. This results in each slice in a given volume being acquired at a slightly different time with respect to the other slices of the same volume. For example, a volume acquisition of the brain consisting of 26 slices may be acquired in a TR period of 2.6 seconds, leading to the last slice of the volume being acquired 2.5 seconds after the first slice of that brain volume. This introduces substantial time differences into the information contained in that brain volume. These need to be corrected prior to the haemodynamic modelling process (described below), which assumes that all the slices acquired within a given volume are captured at the same time. This correction is termed slice timing and applies a time shift to the time series of each voxel within each slice to a reference slice (usually the first or the middle slice acquired). Following this correction, all the time series contained within each slice in a volume will appear as if they were acquired concurrently in time. The second data processing step is realignment. An fMRI experiment typically lasts several minutes and sometimes lasts up to an hour. Subjects cannot lie in the MR scanner with their head perfectly still for the whole duration of the scan session, and therefore there is inevitably movement, ranging from a fraction of a millimetre up to several millimetres, particularly for patients. This leads to a spatial shift in the location of a given voxel across the fMRI time series, which can be particularly problematic when head motion is correlated with the task (for example, due to swallowing during a flavour task). These movements will introduce motion-related signal changes (rather than haemodynamic signal changes) in the fMRI time series that may be indistinguishable from the BOLD response to the task, leading to false activation (Ashburner and Friston, 2001). To eliminate such head movements, the series of image volumes collected during an fMRI experiment can be spatially realigned to each other (motion correction). Common software packages, such as those mentioned above, typically use automated image registration algorithms (Woods et al., 1998), which apply a 6-parameter (3 for translation and 3 for rotation) rigid body transformation to the images. The realignment parameters should be plotted to display the movement of each subject throughout the fMRI experiment, allowing rejection of data sets that contain large motion (for example, a subject who moves by more than one image voxel) or motion strongly correlated with the task. Figure 12.9 illustrates an example of realignment parameters. These realignment parameters can then be used as regressors of no interest in subsequent statistical analysis. To assess a group fMRI response it is necessary to combine the subjects’ data together. However, the study cohort usually comprises many subjects, whose individual brains differ in shape, size and orientation, making it impossible to draw inferences between and across subjects. The solution to this problem is to transform each subject’s images into standard brain space (a brain template), a process known as spatial normalisation. Normalisation applies a 12-parameter realignment algorithm (3 for translation, 3 for rotation, 3 for resizing and 3 for shearing of the brain tissue) to match the brain to a chosen template, typically the Montreal Neurological Institute template space. Despite the spatial normalisation process, cortical regions may still have small differences in location between subjects due to differences in the individuals’ cortical organisation. Ap- plying spatial smoothing to the data overcomes this effect. Spatial smoothing is accomplished by convolving each voxel with a 3D Gaussian kernel of a specific full-width-half-maximum. Typically, smoothing of two to three times the voxel size is used. Spatial smoothing provides further advantages. If the spatial smoothing filter matches the size of the cortical activity, the SNR of the statistical maps will increase, but if larger than the size of the expected activated P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 335 x y z x y z 500 500 400 400 than 9 mm) throughout the fMRI 300 300 200 200 Rotation (degrees) Translation (mm) 100 100 0 0 6 4 2 0 -2 -4 5 0 -5 10 (b) x y z x y z 500 500 moved less than 1 mm, (b) subject with gross movements (more the study. 400 400 300 300 200 200 Translation (mm) Rotation (degrees) 100 100 Example realignment parameters for (a) subject who 0 0 6 4 2 0 -2 -4 5 0 -5 10 (a) experiment whose data were subsequently excluded from Fig. 12.9 P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

336 Food Flavour Technology

region it can reduce the underlying signal. Further, the application of spatial smoothing is necessary to satisfy the requirement of Gaussian random field theory (Friston et al., 1995) used in the statistical analyses of fMRI data (described in more detail below). In addition, high-pass temporal filtering can be applied to each voxel’s time series to remove unwanted slowly varying signal components, such as physiological noise (cardiac and respiration) and scanner drift. The cut-off window must be chosen carefully to avoid removing signals of interest together with the noise. In addition, there are now methods to attempt to correct physiological noise at source, such as the retrospective correction of physiological motion effects in fMRI (RETROICOR) (Glover et al., 2000; Frank et al., 2001). Following pre-processing, the fMRI data are interrogated to identify the brain regions that show significant signal changes in response to the applied stimuli. Most fMRI analyses are carried out on each voxel independently (univariant approach) and tested against the null hypothesis that they are not activated. Generally, mathematical functions are used to represent the empirical ‘shape’ of the BOLD response. The general linear model (GLM) (Friston et al., 1994) is one of the methods most commonly used to analyse fMRI data. In the GLM, the expected BOLD responses are linearly modelled by convolving the stimulus waveform function with a basis function of a predefined or ‘canonical’ HRF. In a block design, the stimulus waveform is represented by a boxcar function, while in an event-related design the stimulus waveform is represented by a series of stick functions. The GLM is formulated in matrix notation as Y = ␤X + ε, where Y is the vector of observed data, X is the design matrix, ␤ is the vector of the parameters to be estimated and ε is the vector of residual errors, which are assumed to be independent and normally distributed. The design matrix X represents the experimental design (the model). A design matrix can be illustrated visually as shown in Fig. 12.10. Each row in the design matrix corresponds to one time series

Fig. 12.10 An example of a design matrix modelled with the HRF only. Each row represents a point in time (volume) and columns correspond to experimental conditions and confounds. The ﬁrst column represents the experimental condition (with white lines depicting stimulus on periods) and the ﬁnal six columns show the motion parameters as regressors of no interest. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 337

Basis function Predicted BOLD Boxcar function response Canonical HRF Signal intensity

Time (second)

Temporal derivative

Fig. 12.11 Schematic representation of the effect of the convolution of the canonical HRF and TD with a boxcar function. The TD (dotted line) can be seen to shift the latency of the response.

of BOLD signal (volume), and each column corresponds to an experimental condition (onset time for stimulus) and/or to confounds such as the estimated motion parameters. These are included as regressors of no interest to remove signal fluctuations correlated to confounds. The parameter vector ␤ is then obtained for each voxel by independently fitting the model to each voxel’s time series. For each column in the design matrix, the parameter that best fits the data is obtained. A good fit between the model and the data will indicate those brain areas that are activated. Different brain regions may have different HRFs, being more delayed or broadened due to their local vasculature or response time. By convolving the data with only the canonical HRF, areas that are activated but display latency and/or dispersion may be missed. To model such variations and maximise statistical power, temporal and/or dispersion derivatives can be used in addition to the canonical HRF. Figure 12.11 illustrates a schematic representation of the convolution of a ‘canonical’ HRF or a ‘canonical HRF plus its temporal derivative (TD)’ basis functions with a boxcar function. The TD delays the HRF. In Marciani et al. (2006), the effect of these two model designs on statistical power was investigated for flavour stimuli. It was found that the canonical HRF plus TD gave higher statistical power across all brain areas than the canonical only. It is possible to add a third ‘dispersion’ basis function to broaden the HRF; however, care must be taken when increasing the number of basis functions as this can increase the number of false-positive activations. Parametric tests (T-orF-statistics depending on the number of basis functions in the model) are then performed on a voxel level to test the statistical significance of activated voxels in response to the given task. The resulting T (or F) values of the activated voxels are then commonly converted to Gaussian form (distribution) to represent Z and uncorrected probability P scores. It is important to note that fMRI analysis performs a significant number of statistical tests on a very large number of data points. To minimise the number of false positives, it is necessary to apply statistical correction for multiple comparisons to fMRI results. Two methods are commonly used: the family-wise error rate correction (Friston et al., 1995) and the false discovery rate correction (Genovese et al., 2002). The family-wise error rate controls the chance of any false positives, using random field theory, while the false discovery rate is less stringent and controls the fraction of false positives. The processing steps described above will then generate an ‘activation map’, typically called a statistical parametric map (SPM), of the subjects’ activity. Those areas surviving a defined statistical threshold are presented P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

338 Food Flavour Technology

in a colour scale superimposed on the image template space (fMRI people often call these coloured areas ‘the blobs’). Processing to this point is known as ‘first-level analysis’. The majority of fMRI analyses proceed to ‘second-level analysis’ to interrogate the data to answer questions based on group statistics. Group statistics generally combine data from 15 to 20 subjects. There are two statistical approaches commonly used in fMRI analysis to group subjects: ‘fixed effects’ and ‘random effects’ analyses. Fixed effects analyses assume that all subjects are affected similarly to the stimulus, only taking into account within subject variability, and thus can only be used to make inferences about the particular subjects studied. In fixed effects analysis, a large design matrix is formed from all subject’s datasets. Random effects analysis takes into account the variability across subjects, in addition to within subject variability, allowing inferences to be made of the population from which the subjects are drawn. In random effects analysis, a design matrix and a statistical analysis are performed for each subject (within subject variability) and the generated SPMs from each subject are pooled for a second-level analysis where the variance is computed across subjects. Having calculated SPMs of activated areas, regions of interest (ROIs) analysis can provide a useful method of quantifying relative brain activation within a given area. This involves extraction of the BOLD time-series signal plots from given ROI in the brain using various tools such as the MarsBaR toolbox (Brett et al., 2002). The locations of these ROIs can be chosen either from apriorianatomical knowledge guided by hypotheses or from spatial information on local maxima of activation gained using the SPMs themselves. This type of ROI analysis is becoming increasingly popular since it generates a set of numerical results for each location and stimulus that can be plotted in bar charts and interrogated for significant differences using much simpler statistics such as ANOVAs and paired t-tests, making results more accessible to wider audiences.

12.3.3 fMRI design for flavour processing When designing fMRI paradigms to study flavour processing, there are many considerations. The first challenge is stimulus delivery. The study of taste is challenging as it requires the delivery of a liquid to the mouth of a subject who is lying supine in an MR scanner. Published studies have used various methods to deliver tastants. These include taste chambers (Bujas et al., 1991) placed on the tip of the tongue or nozzles placed in the mouth. Until recently fMRI studies have used very small quantities of stimuli. However, taste receptor cells are broadly tuned, reflecting the need to use a stimulus delivery system that engages a wide range of receptors in the oral cavity. We developed an automated spray stimulus delivery system (Marciani et al., 2006) designed to extensively cover the whole oral cavity (Fig. 12.12). Subjects hold the spray nozzles between their lips and position the nozzles in the middle of their mouth to receive the sprayed solutions with extensive stimulation of taste receptors through the dispersion of the liquid on the tongue and other mouth surfaces. The system is placed outside the scanner room and computer-controlled to drive four 12 V electric pumps, each connected to a reservoir of stimuli, giving flow of 1 mL/second. Each pump is connected to a length of plastic tube to allow liquids to be delivered to subjects whilst lying inside the scanner. A second consideration is the design of the flavour paradigm for fMRI; specific questions include how long should the stimulus be applied for, how should the palate be cleaned between stimuli and when should the subjects swallow? As outlined in Fig. 12.8, stimuli can be delivered in a block, event-related or mixed design. Behavioural studies of flavour P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 339

Plastic tubes Reservoirs

Presentation PC

12 V pumps

12 V battery and opto-isolated Disposable plastic interfaces to presentation spray nozzles

Fig. 12.12 The stimulus spray delivery.

processing have shown that it is particularly important to consider adaptation effects when delivering sensorial stimuli, such as taste and aroma. Holding a liquid stimulus in the mouth will change both its chemical and physical properties. Olfactory receptor neurons show strong activity to short exposures (Poellinger et al., 2001), but following constant exposure to odourant for several seconds significant adaptation occurs (Dalton et al., 2000); this adaptation declines after a short time, allowing receptors to then redetect the odourants. It is therefore important to limit the on period of stimulus delivery and allow a sufficient ISI between successive deliveries of stimuli to maximise the power of detection. A further issue when designing an fMRI paradigm is the need to clear the mouth between stimuli, a wash of lime juice/water combinations or artificial saliva is typically used. Typical alternative fMRI paradigm designs used in fMRI studies of taste are shown schematically in Fig. 12.13. In taste studies subjects are usually cued to swallow at a defined point in the fMRI cycle. Until recently a delayed swallow has typically been used to minimise the effects of motion induced by the swallow on the MR images (Fig. 12.13a). However, the time of swallowing in the fMRI paradigm is particularly important for the design of flavour studies. There are two distinct routes for odourants to reach olfactory receptors within the nasal cavity: the orthonasal pathway and the retronasal pathway. The orthonasal pathway involves odourants reaching the olfactory receptors through sniffing or inhaling air through the nostrils. The retronasal pathway involves odourants released from the food travelling from the oral cavity through the posterior nostril of the nasopharynx to the olfactory receptors during food ingestion and breathing. Swallowing is essential for the perception of retronasal odourants (Buettner et al., 2001; Weel et al., 2004), where the volatiles are transported to the upper airways during the exhalation that follows the swallowing action. In most studies of P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

340 Food Flavour Technology

(a) Delayed swallow

HRF

Stimulus Swallow Rinse Swallow 0 5 15 20 25 30 Second

(b) Immediate swallow

HRF

Stimulus Rinse Rinse 0 3 21 25 29 34 Second

Swallow Swallow Swallow

Fig. 12.13 One cycle of fMRI paradigm, showing delayed swallow (a) and immediate swallow (b). If an immediate swallow is used the detected signal will be that of the tastant only, whilst in (b) both the tastant and interaction with the aroma will be modelled and detected.

retronasal olfaction, swallowing has generally been delayed to reduce head motion artefacts (Fig. 12.13a), but we have shown that the main retronasal delivery of volatiles in the fMRI paradigm occurs immediately after the swallow. This suggests that fMRI studies should be performed using a prompt swallowing after each sample delivery to obtain physiological retronasal olfactory stimulation. In our studies (Fig. 12.13b) 3 mL of the stimulus solution is delivered by spraying over a 3-second period (flow rate 1 mL/second). Two mouth rinses of 5 mL water and lime juice solution are delivered after 18 seconds over a period of 5 seconds to clear the oral cavity of lingering taste and aroma compounds (as commonly used by sensory panels). We validated our immediate swallow paradigm using atmospheric pressure chemical ionisation–mass spectrometry (see also Chapter 10). This showed that the main retronasal delivery of volatiles in the paradigm occurred immediately after the swallow. Several brain areas were found to be activated, including the insula, FO, rolandic operculum/parietal lobe, piriform, dorsolateral prefrontal cortex, AAC, ventromedial thalamus, hippocampus and medial orbitofrontal cortex. If a delayed swallow is used, the main interaction effect of taste and aroma could be lost. A further measure, which can aid in the experimental design of flavour studies, is to use surface electromyography (EMG) to define the exact time of swallow. Surface EMG typically uses a pair of MRI-compatible Ag/AgCl electrodes placed over the submental muscle complex and the difference between the two channels is assessed. Post-processing must be used to remove gradient-induced artefacts in the EMG signal collected during fMRI P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 341

acquisition; this typically uses an average template subtraction correction method (Allen et al., 2000). Figure 12.14a shows a typical artefact-corrected EMG trace for a single cycle of the fMRI paradigm. This data can then be used for accurate modelling of the time the stimulus remains in the mouth to define the ‘on’ period in the fMRI model (Fig. 12.14b and c), or the swallow trace can be used as a regressor of no interest to remove swallow-related motion artefacts. An important consideration for flavour studies is what condition the stimulus should be compared to. If flavour stimuli are compared to a baseline period, then activated areas will be those involved in flavour processing as well as those used in the ‘maintenance’ of taste and swallowing. Typically, flavour stimuli are compared with a control solution. Many studies use a tasteless and aroma-less ‘mock saliva’ solution (containing 6.25 mM KCl and 0.625 mM NaHCO3) as a control solution to mimic the somatosensory effects induced by swallowing and the presence of a liquid in the mouth. KCl and NaHCO3 are natural ionic components of the human saliva and form a better neutral control than pure water (O’Doherty et al., 2001; de Araujo et al., 2003). Studies investigating the cortical response to oral fat have used a control stimulus of carboxymethyl cellulose (CMC) (a thickener). However, care should be taken when using a ‘control’ stimulus as the control stimulus may itself give more activity than the stimulus of interest; water itself has been shown to lead to a large BOLD response, and thickeners such as CMC may themselves give rise to taste. Therefore, instead of comparing to a control stimulus, it is often better to use a parametric design (as described in Section 12.3.2.3) where graded stimuli are used (for example, different intensities of tastant or odour, or differing levels of fat concentrations).

12.3.4 Behavioural data and subject choice It is advantageous to collect behavioural data such as subjective ratings of stimuli during the fMRI scan. Scores collected during the imaging can be related to prior psychophysical characterisation of the samples or used to model the expected HRF of fMRI data, for example as regressors in the fMRI analysis (see Section 12.3.2.4). The magnetic environment of the scanner room is a challenge to the use of electronic equipment conventionally used to collect subject responses, or measure physiological parameters. Nevertheless, this can be achieved by projecting visual cues and asking subjects to press various buttons in response to a stimulus or use joysticks with visual analogue scales. This approach is desirable; however, care must be taken that this does not interfere with the primary paradigm of interest, complicating the interpretation of the fMRI data. In fMRI studies groups of 15–20 subjects are typically assessed to form a generalised population response (Friston et al., 1999). However, it may be possible to reduce this number if it is possible to reduce the heterogeneity of the group studied; for instance, the subjects’ taster status has been shown to affect the fMRI response. This is further discussed in Section 12.4.5.

12.3.5 Measurement limitations The MR environment can be challenging to study ﬂavour imaging. MRI uses a strong superconductive magnet and subjects lay in the magnet bore, which is typically a tube 60 cm wide and 120 cm long. During fMRI scanning subjects are exposed to approximately 120 dB acoustic noise as the images are acquired. The most obvious constraint of this environment is the need for subjects to adopt a supine position. This imposes difﬁculty in P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

342 Food Flavour Technology

Fig. 12.14 (a) Artefact rejected EMG traces for the swallow reﬂex acquired concurrently during a single cycle of fMRI paradigm. (b) SPM maps modelled by convolving the box function with a 3-second box function (stimulus duration) and (c) with a box function determined from EMG. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 343

stimulus delivery and fMRI paradigm design. If results from fMRI experiments are to be meaningfully extrapolated to flavour perception, it is necessary to establish whether body position has a significant impact on subjects’ sensitivity to flavour. We have demonstrated that body position has no overall effect on subjects’ ability to discriminate between two subtly different flavoured samples (Hort et al., 2009). Secondly, adequate quantities of stimuli are needed to ensure coverage of the oral cavity, and stimuli need to be presented repetitively for cortical activations to reach statistical power. However, care must be taken to limit the total volume of liquid delivered to subjects, whilst they lie supine, to avoid problems with swallowing. A further problem can be motion artefacts in fMRI images related to swallowing. As mentioned in Section 12.3.3, delayed swallowing to minimise the movement artefacts could affect flavour perception. It would be very difficult to study samples that require chewing. Finally, an important consideration is that not all subjects can participate in MR scans. Due to the large magnetic field associated with this technique, it is not suited for subjects with metal implants and devices such as pacemakers.

12.4 BRAIN IMAGING OF FLAVOUR

12.4.1 Brain imaging of taste In recent years, the use of neuroimaging techniques, particularly fMRI, has improved our understanding of the cortical representation of taste in humans. Activation of the primary taste cortex including the anterior insula and FO has been reported in many studies (Francis et al., 1999; de Araujo et al., 2003). More recently, the rolandic operculum of the parietal cortex has been identified as a part of the primary gustatory cortex (Faurion et al., 1999; Ogawa et al., 2005). O’Doherty et al. (2001) were the first to investigate the cortical response to pleasant (glucose) and aversive (salt) taste stimuli by assessing stimuli against a tasteless control stimulus; other studies (using PET) had previously used water as a control (Zald et al., 1998). In individual subject analysis, the OFC showed separate areas activated with respect to the two tastes. These findings were consistent with a PET study by Zald et al., 2002), where aversive bitter taste led to activation of the anterior OFC, whilst the caudolateral OFC responded to a sweet taste. Small et al. (Small et al., 2003) also noted activations in the OFC to sweet and aversive taste. They also investigated the neural response to taste intensity when valence was held constant and showed activation of the amygdala and mid-insula. Furthermore, the amygdala responded to both sweet and aversive taste (when intensity was held constant), with preferential activation to sweet taste, providing evidence that amygdala is not solely involved in processing aversive taste. Other activations including ACC have been reported in many fMRI studies (Francis et al., 1999; O’Doherty et al., 2001; de Araujo et al., 2003; Small et al., 2003).

12.4.2 Brain imaging of aroma Several fMRI studies have investigated the cortical representation to ortho- and retronasal olfactory stimulation. These studies report robust activation in higher order olfactory areas, including the OFC, insula and ACC (Francis et al., 1999; O’Doherty et al., 2000; Savic et al., 2000). In contrast, the POC has shown no (Yousem et al., 1997; Fulbright et al., 1998) or inconsistent activations in some studies (O’Doherty et al., 2000; Weismann et al., P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

344 Food Flavour Technology

2001). This may be due to its small structure, susceptibility to artefacts in this region or habituation effects (Poellinger et al., 2001). Cerf-Ducastel and Murphy (Cerf-Ducastel and Murphy, 2001) reported the first neural response to retronasal olfactory stimulation, using an aqueous solution containing olfactory components. The results showed activation in the same brain regions previously reported for orthonasal stimulation: piriform cortex, insula, FO, amygdala, ACC, OFC, hippocampus and entorhinal cortex. However, in addition this study showed activation of the rolandic operculum, an area not observed in the orthonasal studies. The rolandic operculum is thought to be related to swallowing and lingual stimulation, indicating that the retronasal perception may associate oral somatosensory stimulation. This area has since been reported by other studies (de Araujo et al., 2003; Small et al., 2004). Small and colleagues (Small et al., 2005) compared cortical representations of retro- and orthonasal stimulation. Their study was the first to perform a direct comparison between the ortho- and retronasal stimulation. In this study stimuli were administrated in the same (gaseous) phase (de Araujo et al., 2003). Results supported the finding of Cerf-Ducastel and Murphy of activation of the rolandic operculum, but since this retronasal stimulation was through vapour stimuli it is unlikely that the rolandic operculum was activated due to engaging somatosensory areas from tongue movement and swallowing, and rather suggests that retronasal perception may include the oral cavity component. Olfactory perception and sensitivity vary widely across individuals. Many factors alter olfactory perception, including medications, disease and hormonal disturbances. In general, two kinds of processes disturb the olfactory perception: prevention of the odourant reaching the olfactory epithelium and damage of the olfactory epithelium and damage of parts of the olfactory system. Studies have reported age-related changes in olfactory function, including identification and discrimination (Stevens et al., 1988), and intensity perception (Stevens and Cain, 1985).

12.4.3 Imaging cortical associations In recent years neuroimaging studies have begun to assess cortical associations to investigate the origin of flavour processing. de Araujo et al. (de Araujo et al., 2003) investigated taste–aroma integration using a uni-modal taste and uni-modal retronasal aroma and compared this to bimodal stimuli with combined taste and aroma ‘flavour’. Results showed significant activation to bimodal ‘flavour’ stimuli with congruent combinations (sucrose + strawberry odour) but not for incongruent combinations. Activations were found in caudal OFC (cOFC), ventral striatum, anterior insula, FO, amygdala, and ACC. Furthermore, the congruent combination showed a supra-additive response, which was greater than the sum of the activation produced by uni-modal strawberry and sucrose. Small et al. (2004) used a similar protocol and confirmed this finding for congruent (sucrose + vanilla odour) and incongruent (salty + vanilla) combinations. In addition, they highlighted other areas showing supra-additive responses including the ventrolateral prefrontal cortex and posterior parietal cortex. A more recent study by McCabe and Rolls (McCabe and Rolls, 2007) showed the supra-additive response in a combination of monosodium glutamate and vegetable odour in the OFC. It is important to note that the studies described above used retronasal odour delivery. Taste–odour integration is strongly dependent on the mode of the odour delivery (retronasal vs orthonasal). A PET study by Small (Small et al., 1997) demonstrated significant suppression in taste and orthonasal odour combination of sucrose and strawberry, suggesting that flavour perception largely results from the retronasal rather than orthonasal pathway. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 345

12.4.4 Texture and the ‘taste of fat’ Very little is known of the representation of oral somatosensory information in the human brain. To date only two fMRI studies have investigated the cortical representation to oral texture (viscosity). de Araujo and Rolls (de Araujo and Rolls, 2004) investigated the representation of oral viscosity and oral fat using a tasteless and odourless solution, carboxymethyl cellulose, and vegetable oil (pure fat). The mid- and anterior insula were activated to both oral viscosity and oral fat. Other regions responding to oral fat included the OFC, ACC and hypothalamus. The mid-insula cortex was thought to represent somatosensory properties of the oral activity (texture), whilst the ACC represented the hedonic properties. A more recent study by de Celis Alonso et al. (2007), using Manugel (an alginate gel) as a viscous stimulus, showed activation consistent with de Araujo and Rolls study. However, additional responses were found in the mid- and anterior insula, post-central gyrus and rolandic/parietal operculum.

12.4.5 The issue of the ‘super-tasters’ The perception of taste across individuals may be altered due to many factors. Research has shown that the population can be divided into non-tasters (about 25%), medium tasters (about 50%), and super-tasters (about 25%). This variation is thought to be due to the existence of more fungiform papillae and taste buds in some people than others (Miller, 1988; Smith and Margolskee, 2001). However, there is growing evidence that taster status co-varies with tactile somatosensory acuity, and that it is in fact the increased mechanoreceptors density, which correlates with fungiform papillae density, that determines taster status. Population classification of taster status is commonly assessed in behavioural studies by measuring subjects’ sensitivity to a bitter chemical called ‘6-n-propylthiouracil (PROP)’. Subjects are asked to taste a filter paper soaked in super-saturated PROP solution and rate for bitterness intensity on a generalised labelled magnitude scale (Bartoshuk et al., 2003). Non-tasters cannot taste the bitterness of PROP, medium tasters sense the bitterness whilst accepting it and super-tasters find the taste of PROP unacceptable. Tasters (medium- and super-tasters) are more sensitive than non-tasters to the bitterness of caffeine and to the sweetness of sucrose and some artificial sweeteners (Bartoshuk et al., 1994). Moreover, some studies have also shown that PROP tasters are also more sensitive to fat (Tepper and Nurse, 1997). We have recently shown that taster status is highly correlated with the cortical response in somatosensory areas (SI, SII, mid- and posterior insula) and reward areas (amygdala and anterior cingulate) to fat. The data showed a significant increase in BOLD response with taster status (super-taster Ͼ taster Ͼ non-taster). This finding of a strong correlation with taster status in somatosensory areas supports the findings of Essick et al. (2003) who suggest that heightened taster status is correlated with mechanoreceptor levels. It also suggests that when scanning subjects for flavour studies it is recommended to exclude non-tasters.

12.5 FUTURE TRENDS

There is still much research needed to understand ﬂavour processing in the human brain. Many technical challenges still remain and many other possible applications are still to be discovered. A future direction will be to understand both the cortical response to ﬂavour and P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

346 Food Flavour Technology

the interaction with the gut, for example to improve understanding of factors affecting the sensation of satiety and the influence of digestion on flavour perception. In terms of functional MRI, the development of ultra-high-field (7 T) scanners will improve spatial mapping and allow better differentiation of cortical areas involved in uni-modal and multi-modal processing. One of the major gains of ultra-high field is that the increased BOLD signal change will allow mapping of fMRI responses to a single trial, thus allowing the delivery of more natural stimuli and reduced adaptation effects. Multi-modal imaging, using the application of two or more imaging techniques, will enhance the ability to dissect the underlying neural processes in the human brain and improve temporal and spatial resolution. This is best achieved by combining neuroimaging techniques that measure different aspects of the neural activity, such as electrophysiological activity (EEG/MEG) with haemodynamic responses (fMRI/PET). Simultaneous EEG and fMRI (Ives et al., 1993) measurements have provided an attractive method of collecting data with both high spatial and temporal resolution, and recent work has exploited the use of simultaneous EEG and MEG (Mizoguchi et al., 2002).

REFERENCES

Allen, P.J., Josephs, O. and Turner, R. (2000) A method for removing imaging artifact from continuous EEG recorded during functional MRI. Neuroimage 12, 230–239. Amaro, E. and Barker, G.J. (2006) Study design in MRI: basic principles. Brain Cogn. 60, 220–232. Ashburner, J. and Friston, K.J. (2001) Why voxel-based morphometry should be used. Neuroimage 14, 1238–1243. Baert, A.L., Moonen, C.T.W. and Bandettini, P.A. (2000) Functional MRI, Springer-Verlag, Berlin. Bandettini, P.A., Wong, E.C., Hinks, R.S., Tikofsky, R.S. and Hyde, J.S. (1992) Time course of EPI of human brain function during task activation. Magn. Reson. Med. 25, 390–397. Bartoshuk, L.M., Duffy, V.B., Fast, K., Green, B.G., Prutkin, J. and Snyder, D.J. (2003) Labeled scales (e.g., category, Likert, VAS) and invalid across-group comparisons: what we have learned from genetic variation in taste. Food Qual. Pref. 14, 125–138. Bartoshuk, L.M., Duffy, V.B. and Miller, I.J. (1994) PTC/PROP tasting: anatomy, psychophysics and sex effects. Physiol. Behav. 56, 1165–1171. Baylis, L.L., Rolls, E.T. and Baylis, G.C. (1995) Afferent connections of the caudolateral orbitofrontal cortex taste area of the primate. Neuroscience 64, 801–812. Bear, M.F., Connors, B.W.and Paradiso, M.A. (2007) Neuroscience: Exploring the Brain. Lippincott Williams & Winkins, Baltimore, MD. Brauchli, P., Ruegg, P.B., Etzweiler, F. and Zeier, H. (1995) Electrocortical and autonomic alteration by administration of a pleasant and unpleasant odor. Chem. Senses 20, 505–515. Brett, M., Anton, J., Valabregue, R. and Poline, J. (2002) Region of Interest Analysis Using an SPM Toolbox, Eighth International Conference on Functional Mapping of the Human Brain, Sendai, Japan. Buchel, C., Holmes, A.P., Rees, G. and Friston, K.J. (1998) Characterizing stimulus-response functions using nonlinear regressors in parametric fMRI experiments. Neuroimage 8, 140–148. Buettner, A., Beer, A., Hannig, C. and Settles, M. (2001) Observation of the swallowing process by application of videofluoroscopy and real-time magnetic resonance imaging-consequences for retronasal aroma stimulation. Chem. Senses 26, 1211–1219. Bujas, Z., Szabo, S., Ajdukovic, D. and Mayer, D. (1991) Magnitude estimation of adaptation to salt using a flow chamber for stimulus delivery. Physiol. Behav. 49, 735–737. Buxton, R.B., Uludag, K., Dubowitz, D.J. and Liu, T.T. (2004) Modeling the hemodynamic response to brain activation. Neuroimage 23, S220–S233. Cerf-Ducastel, B. and Murphy, C. (2001) fMRI activation in response to odorants orally delivered in aqueous solutions. Chem. Senses 26, 625–637. Cohen, D. (1968) Magnetoencephalography: evidence of magnetic fields produced by alpha rhythm currents. Science 161, 784–786. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 347

Dalton, P., Doolittle, N., Nagata, H. and Breslin, P.A.S. (2000) The merging of the senses: integration of subthreshold taste and smell. Nat. Neurosci. 3, 431–432. de Araujo, I.E.T. and Rolls, E.T. (2004) Representation in the human brain of food texture and oral fat. J. Neurosci. 24, 3086–3093. de Araujo, I.E.T., Rolls, E.T., Kringelbach, M.L., McGlone, F. and Phillips, N. (2003) Taste-olfactory convergence, and the representation of the pleasantness of flavour, in the human brain. Eur. J. Neurosci. 18, 2059–2068. de Celis Alonso, B., Marciani, L., Head, K., Clark, P., Spiller, R.C., Rayment, P., Ablett, S., Francis, S. and Gowland, P.A. (2007) Functional magnetic resonance imaging assessment of the cortical representation of oral viscosity. J. Texture Stud. 36(6), 725–737. Del Parigi, A., Gautier, J.F., Chen, K.W., Salbe, A.D., Ravussin, E., Reiman, E. and Tataranni, P.A. (2002) Neuroimaging and obesity – mapping the brain responses to hunger and satiation in humans using positron emission tomography. Ann. N. Y. Acad. Sci. 967, 389–397. Essick, G.K., Chopra, A., Guest, S. and McGlone, F. (2003) Lingual tactile acuity, taste perception, and the density and diameter of fungiform papillae in female subjects. Physiol. Behav. 80, 289–302. Faurion, A., Cerf, B., Van De Moortele, P.F., Lobel, E., Mac Leod, P. and Le Bihan, D. (1999) Human taste cortical areas studied with functional magnetic resonance imaging: evidence of functional lateralization related to handedness. Neurosci. Lett. 277, 189–192. Fera, F., Yongbi, M.N., van Gelderen, P., Frank, J.A., Mattay, V.S. and Duyn, J.H. (2004) EPI-BOLD fMRI of human motor cortex at 1.5 T and 3.0 T: sensitivity dependence on echo time and acquisition bandwidth. J. Magn. Reson. Imaging 19, 19–26. Francis, S., Rolls, E.T., Bowtell, R., McGlone, F., O’Doherty, J., Browning, A., Clare, S. and Smith, E. (1999) The representation of pleasant touch in the brain and its relationship with taste and olfactory areas. Neuroreport 10, 453–459. Frank, L.R., Buxton, R.B. and Wong, E.C. (2001) Estimation of respiration-induced noise fluctuations from undersampled multislice fMRI data. Magn. Reson. Med. 45, 635–644. Frank, R.A. and Byram, J. (1988) Taste smell interactions are tastant and odorant dependent. Chem. Senses 13, 445–455. Friston, K.J., Fletcher, P., Josephs, O., Holmes, A., Rugg, M.D. and Turner, R. (1998) Event-related fMRI: characterizing differential responses. Neuroimage 7, 30–40. Friston, K.J., Holmes, A.P., Poline, J.B., Grasby, P.J., Williams, S.C.R., Frackowiak, R.S.J. and Turner, R. (1995) Analysis of fMRI time series revisited. Neuroimage 2, 45–53. Friston, K.J., Holmes, A.P. and Worsley, K.J. (1999) How many subjects constitute a study? Neuroimage 10, 1–5. Friston, K.J., Holmes, A.P., Worsley, K.J., Poline, J.P., Frith, C.D. and Frackowiak, R.S.J. (1994) Statistical parametric maps in functional imaging: a general linear approach. Hum. Brain Mapp. 2, 189–210. Fulbright, R.K., Skudlarski, P., Lacadie, C.M., Warrenburg, S., Bowers, A.A., Gore, J.C. and Wexler, B.E. (1998) Functional MR imaging of regional brain responses to pleasant and unpleasant odors. Am. J. Neuroradiol. 19, 1721–1726. Gautier, J.F., Chen, K.W., Uecker, A., Bandy, D., Frost, J., Salbe, A.D., Pratley, R.E., Lawson, M., Ravussin, E., Reiman, E.M. and Tataranni, P.A. (1999) Regions of the human brain affected during a liquid-meal taste perception in the fasting state: a positron emission tomography study. Am. J. Clin. Nutr. 70, 806–810. Genovese, C.R., Lazar, N.A. and Nichols, T. (2002) Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage 15, 870–878. Glover, G.H., Li, T.Q. and Ress, D. (2000) Image-based method for retrospective correction of physiological motion effects in fMRI: RETROICOR. Magn. Reson. Med. 44, 162–167. Gottfried, J.A., Winston, J.S. and Dolan, R.J. (2006) Dissociable codes of odor quality and odorant structure in human piriform cortex. Neuron 49, 467–479. Gowland, P.A. and Bowtell, R. (2007) Theoretical optimization of multi-echo fMRI data acquisition. Phys. Med. Biol. 52, 1801–1813. Hagberg, G.E., Indovina, I., Sanes, J.N. and Posse, S. (2002) Real-time quantification of T-2(*) changes using multiecho planar imaging and numerical methods. Magn. Reson. Med. 48, 877–882. Hoge, R.D., Atkinson, J., Gill, B., Crelier, G.R., Marrett, S. and Pike, G.B. (1999) Investigation of BOLD signal dependence on cerebral blood flow and oxygen consumption: the deoxyhemoglobin dilution model. Magn. Reson. Med. 42, 849–863. Hort, J. and Hollowood, T.A. (2004) Controlled continuous flow delivery system for investigating taste-aroma interactions. J. Agric. Food Chem. 52, 4834–4843. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

348 Food Flavour Technology

Hort, J., Redureau, S., Hollowood, T., Marciani, L., Eldeghaidy, S., Head, K., Busch, J., Spiller, R.C., Francis, S., Gowland, P.A. and Taylor, A.J. (2009) The effect of body position on flavour release and perception: implications for fMRI studies. Chemosens. Percept. 1(4), 253–257. Huettel, S.A., Song, A.W. and McCarthy, G. (2004) Functional Magnetic Resonance Imaging, Sinauer Associates, Sunderland, MA. Humberstone, M., Sawle, G.V., Clare, S., Hykin, J., Coxon, R., Bowtell, R., Macdonald, I.A. and Morris, P.G. (1997) Functional magnetic resonance imaging of single motor events reveals human presupplementary motor area. Ann. Neurol. 42, 632–637. Ives, J.R., Warach, S., Schmitt, F., Edelman, R.R. and Schomer, D.L. (1993) Monitoring the patients’ EEG during echo-planar MRI. Electroencephalogr. Clin. Neurophysiol. 87, 417–420. Jansma, J.M., Ramsey, N.F., Coppola, R. and Kahn, R.S. (2000) Specific versus nonspecific brain activity in a parametric n-back task. Neuroimage 12, 688–697. Jezzard, P., Matthews, P.M. and Smith, S.M. (2004) Functional MRI: An Introduction to Methods, Oxford University Press, Oxford. JonesGotman, M., Zatorre, R.J., Cendes, F., Olivier, A., Andermann, F., McMackin, D., Staunton, H., Siegel, A.M. and Wieser, H.G. (1997) Contribution of medial versus lateral temporal-lobe structures to human odour identification. Brain 120, 1845–1856. Kandel, E.R., Schwartz, J.H. and Jessel, T.M. (eds) (2000) Principles of Neural Science, McGraw-Hill, New York. Kinomura, S., Kawashima, R., Yamada, K., Ono, S., Itoh, M., Yoshioka, S., Yamaguchi, T., Matsui, H., Miyazawa, H., Itoh, H., Goto, R., Fujiwara, T., Satoh, K. and Fukuda, H. (1994) Functional-anatomy of taste perception in the human brain studied with positron emission tomography. Brain Res. 659, 263–266. Kobal, G. (1985) Gustatory evoked potentials in man. Electroencephalogr. Clin. Neurophysiol. 62, 449–454. Kobal, G., Hummel, T. and Vantoller, S. (1992) Differences in human chemosensory evoked potentials to olfactory and somatosensory chemical stimuli presented to left and right nostrils. Chem. Senses 17, 233–244. Kobayakawa, T., Endo, H., AyabeKanamura, S., Kumagai, T., Yamaguchi, Y., Kikuchi, Y., Takeda, T., Saito, S. and Ogawa, H. (1996) The primary gustatory area in human cerebral cortex studied by magnetoencephalography. Neurosci. Lett. 212, 155–158. Kobayakawa, T., Ogawa, H., Kaneda, H., Ayabe-Kanamura, S., Endo, H. and Saito, S. (1999) Spatio- temporal analysis of cortical activity evoked by gustatory stimulation in humans. Chem. Senses 24, 201– 209. Kruger, G., Kleinschmidt, A. and Frahm, J. (1996) Dynamic MRI sensitized to cerebral blood oxygenation and flow during sustained activation of human visual cortex. Magn. Reson. Med. 35, 797–800. Kwong, K.K., Belliveau, J.W., Chesler, D.A., Goldberg, I.E., Weisskoff, R.M., Poncelet, B.P., Kennedy, D.N., Hoppel, B.E., Cohen, M.S., Turner, R., Cheng, H.M., Brady, T.J. and Rosen, B.R. (1992) Dynamic magnetic-resonance-imaging of human brain activity during primary sensory stimulation. Proc. Natl. Acad. Sci. USA 89, 5675–5679. Luck, S.J. (2005) An Introduction to the Event-Related Potential Technique, MIT Press, Cambridge. Mandeville, J.B., Marota, J.J.A., Kosofsky, B.E., Keltner, J.R., Weissleder, R., Rosen, B.R. and Weisskoff, R.M. (1998) Dynamic functional imaging of relative cerebral blood volume during rat forepaw stimulation. Magn. Reson. Med. 39, 615–624. Marciani, L., Pfeiffer, J.C., Hort, J., Head, K., Bush, D., Taylor, A.J., Spiller, R.C., Francis, S. and Gowland, P.A. (2006) Improved methods for fMRI studies of combined taste and aroma stimuli. J. Neurosci. Methods 158, 186–194. McCabe, C. and Rolls, E.T. (2007) Umami: a delicious flavor formed by convergence of taste and olfactory pathways in the human brain. Eur. J. Neurosci. 25, 1855–1864. Menon, R.S., Ogawa, S., Hu, X.P., Strupp, J.P., Anderson, P. and Ugurbil, K. (1995) BOLD based functional MRI at 4 Tesla includes a capillary bed contribution: echo-planar imaging correlates with previous optical imaging using intrinsic signals. Magn. Reson. Med. 33, 453–459. Miller, I.J. (1988) Human taste bud density across adult age groups. J. Gerontol. 43, B26–B30. Mizoguchi, C., Kobayakawa, T., Saito, S. and Ogawa, H. (2002) Gustatory evoked cortical activity in humans studied by simultaneous EEG and MEG recording. Chem. Senses 27, 629–634. Noble, A.C. (1996) Taste-aroma interactions. Trends Food Sci. Technol. 7, 439–444. O’Doherty, J., Rolls, E.T., Francis, S., Bowtell, R. and McGlone, F. (2001) Representation of pleasant and aversive taste in the human brain. J. Neurophysiol. 85, 1315–1321. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

Brain imaging 349

O’Doherty, J., Rolls, E.T., Francis, S., Bowtell, R., McGlone, F., Kobal, G., Renner, B. and Ahne, G. (2000) Sensory-specific satiety-related olfactory activation of the human orbitofrontal cortex. Neuroreport 11, 893–897. Ogawa, H., Wakita, M., Hasegawa, K., Kobayakawa, T., Sakai, N., Hirai, T., Yamashita, Y. and Saito, S. (2005) Functional MRI detection of activation in the primary gustatory cortices in humans. Chem. Senses 30, 583–592. Ogawa, S., Lee, T.M., Kay, A.R. and Tank, D.W. (1990) Brain magnetic-resonance-imaging with contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci. USA 87, 9868–9872. Pauling, L. and Coryell, C.D. (1936) The magnetic properties and structure of hemoglobin, oxyhemoglobin and carbonmonoxyhemoglobin. Proc. Natl. Acad. Sci. USA 22, 210–216. Phelps, M.E., Hoffman, E.J., Mullani, N.A. and Terpogossian, M.M. (1975) Application of annihilation coincidence detection to transaxial reconstruction tomography. J. Nucl. Med. 16, 210–224. Poellinger, A., Thomas, R., Lio, P., Lee, A., Makris, N., Rosen, B.R. and Kwong, K.K. (2001) Activation and habituation in olfaction – an fMRI study. Neuroimage 13, 547–560. Poser, B.A., Versluis, M.J., Hoogduin, J.M. and Norris, D.G. (2006) BOLD contrast sensitivity enhancement and artifact reduction with multiecho EPI: parallel-acquired inhomogeneity-desensitized fMRI. Magn. Reson. Med. 55, 1227–1235. Price, C.J., Veltman,D.J., Ashburner, J., Josephs, O. and Friston, K.J. (1999) The critical relationship between the timing of stimulus presentation and data acquisition in blocked designs with fMRI. Neuroimage 10, 36–44. Purdon, P.L. and Weisskoff, R.M. (1998) Effect of temporal autocorrelation due to physiological noise and stimulus paradigm on voxel-level false-positive rates in fMRI. Hum. Brain Mapp. 6, 239– 249. Rolls, E.T. and Baylis, L.L. (1994) Gustatory, olfactory, and visual convergence within the primate orbitofrontal cortex. J. Neurosci. 14, 5437–5452. Saito, S., Endo, H., Kobayakawa, T., Ayabe-Kanamura, S., Kikuchi, Y., Takeda, T. and Ogawa, H. (1998) Temporal process from receptors to higher brain in taste detection studied by gustatory-evoked magnetic fields and reaction times. Ann. N. Y. Acad. Sci. 855, 493–497. Savic, I., Gulyas, B., Larsson, M. and Roland, P. (2000) Olfactory functions are mediated by parallel and hierarchical processing. Neuron 26, 735–745. Small, D.M., Gerber, J.C., Mak, Y.E. and Hummel, T. (2005) Differential neural responses evoked by orthonasal versus retronasal odorant perception in humans. Neuron 47, 593–605. Small, D.M., Gregory, M.D., Mak, Y.E., Gitelman, D., Mesulam, M.M. and Parrish, T. (2003) Dissocia- tion of neural representation of intensity and affective valuation in human gustation. Neuron 39, 701– 711. Small, D.M., Jones-Gotman, M., Zatorre, R.J., Petrides, M. and Evans, A.C. (1997) Flavor processing: more than the sum of its parts. Neuroreport 8, 3913–3917. Small, D.M., Voss, J., Mak, Y.E., Simmons, K.B., Parrish, T. and Gitelman, D. (2004) Experience-dependent neural integration of taste and smell in the human brain. J. Neurophysiol. 92, 1892–1903. Smith, D.V. and Margolskee, R.F. (2001) Making sense of taste. Sci. Am. 284, 32–39. Stevens, J.C. and Cain, W.S. (1985) Age-related deficiency in the perceived strength of 6 odorants. Chem. Senses 10, 517–529. Stevens, J.C., Cain, W.S. and Burke, R.J. (1988) Variability of olfactory thresholds. Chem. Senses 13, 643–653. Taylor, A.J., Linforth, R.S.T., Harvey, B.A. and Blake, B. (2000) Atmospheric pressure chemical ionisation mass spectrometry for in vivo analysis of volatile flavour release. Food Chem. 71, 327–338. Tepper, B.J. and Nurse, R.J. (1997) Fat perception is related to PROP taster status. Physiol. Behav. 61, 949–954. Weel, K.G.C., Boelrijk, A.E.M., Burger, J.J., Verschueren, M., Gruppen, H., Voragen, A.G.J. and Smit, G. (2004) New device to simulate swallowing and in vivo aroma release in the throat from liquid and semiliquid food systems. J. Agric. Food Chem. 52, 6564–6571. Weismann, M., Yousry, I., Heuberger, E., Nolte, A., Ilmberger, J., Kobal, G., Yousry, T.A., Kettenmann, B. and Naidich, T.P. (2001) Functional magnetic resonance imaging of human olfaction. Neuroimaging Clin. N. Am. 11, 237–250, viii. Woods, R.P., Grafton, S.T., Holmes, C.J., Cherry, S.R. and Mazziotta, J.C. (1998) Automated image registration: I. General methods and intrasubject, intramodality validation. J. Comput. Assist. Tomogr. 22, 139–152. P1: SFK/UKS P2: SFK c12 BLBK221-Taylor/Linforth December 2, 2009 17:21 Printer Name: Yet to Come

350 Food Flavour Technology

Yamamoto, C., Takehara, S., Morikawa, K., Nakagawa, S., Yamaguchi, M., Iwaki, S., Tonoike, M. and Yamamoto, T. (2003) Magnetoencephalographic study of cortical activity evoked by electrogustatory stimuli. Chem. Senses 28, 245–251. Yousem, D.M., Williams, S.C.R., Howard, R.O., Andrew, C., Simmons, A., Allin, M., Geckle, R.J., Suskind, D., Bullmore, E.T., Brammer, M.J. and Doty, R.L. (1997) Functional MR imaging during odor stimulation: preliminary data. Radiology 204, 833–838. Zald, D.H., Hagen, M.C. and Pardo, J.V. (2002) Neural correlates of tasting concentrated quinine and sugar solutions. J. Neurophysiol. 87, 1068–1075. Zald, D.H., Lee, J.T., Fluegel, K.W. and Pardo, J.V. (1998) Aversive gustatory stimulation activates limbic circuits in humans. Brain 121, 1143–1154. Zald, D.H. and Pardo, J.V. (1997) Emotion, olfaction, and the human amygdala: amygdala activation during aversive olfactory stimulation. Proc. Natl. Acad. Sci. USA 94, 4119–4124. Zald, D.H. and Pardo, J.V. (2000) Functional neuroimaging of the olfactory system in humans. Int. J. Psychophysiol. 36, 165–181. P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

Index

Absorption of flavours, 193 electronic noses, 261 Abstract flavours, 20 fast GC, 250 Acetic acid Ð pathway, 59 GC methods, 249 Acid/base fractionation, 245 GC/olfactometry, 250 advantages, 246 GC×GC methods, 250 scheme, 246 heart cutting, 250 Activated charcoal identification of compounds, 255 aroma extraction, 234 infrared spectroscopy, 256 Activity coefficients, 191, 193Ð194, 209 mass spectrometry, 257 Acyloins Ð bio formation, 100 specific GC detectors, 254 Additives in flavourings, 36 Aroma binding, 193 Adsorption of flavours, 193 Aroma extraction AEDA (Aroma Extract Dilution Analysis), 65, activated charcoal, 234 297 bias due to extraction methodology, 230, 238 AEDA (Aroma Extract Dilution Analysis) of cocoa, choice of method, 231 133 cryogenic traps, 233 AEDA (Aroma Extract Dilution Analysis) of milk distillation methods, 235 chocolate, 134 defining aims and objectives, 242 AEDA (Aroma Extract Dilution Analysis) of tea, dynamic headspace, 234 133 effect of method on recoveries, 230 AEDA Aroma Extract Dilution Analysis, 252 flavour changes with time, 244 Agaric acid fractionation, 245 limit in flavours, 35, 44 general considerations, 241 Alcohol oxidation identification of key components, 243 biotechnology, 94 internal standards, 242 Allium flavours, 162 off-odours, 243 Alpha helices, 196 purge and trap, 233 ␣-Pinene quantitative and qualitative data, 242 biotransformation, 92 recoveries from solvent extraction, 237 Amadori compound formation, 53 selection criteria, 242 Amadori rearrangement products (ARP), 54 solvent extraction, 237 Analytical challenges in flavour, 229 SPME, 238 Analytical methods stir bar, 240 choosing the appropriate method, 230 Tenax trapping, 233 trace levels of flavour compounds, 230 using volatility, 231 Anise vacuum distillation, 235 secondary metabolites, 106 Aromagrams, 250 ANOVA, 310 AromaÐtaste interactions, 301 APCI ionisation ARP (Amadori rearrangement product), 54, 61 direct MS in the liquid mode, 280 Artefacts APCI-MS, 259 due to enzyme activity, 241 Apple flavour, 166 in extraction, 241 Apple varieties, 166 thermally induced, 241 Apricot flavour, 169 Artificial flavouring Aroma analysis European definition 1974, 28 delivery from foods, 260 Atmospheric Pressure Chemical Ionisation direct mass spectrometry, 259 (APCI)ÐMS, 259, 272 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

352 Index

Batch extraction principle Carvone flavour delivery, 215 bioformation yam cells, 107 Beer fermentation Cereal products, 158 undesirable odours, 100 Cerebrum, 321 Beer flavour, 154 Character-impact compound, 65, Benzaldehyde 135Ð136 bioformation, 97 Character-impact compounds Benzopyrone process flavour, 70Ð73 limits in food and beverages, 34 CHARM analysis, 252 Biocatalysis, 90 Cheese flavour, 149 Bioformation Cheese ripening, 149 effect of solvents and solubility, 113 Cheese ripening, 150 extraction techniques, 113 Chemical dissociation flow through production, 114 effect on partition, 211 genetic engineering, 107 Chemical ionisation process constraints, 112 direct mass spectrometry, 272 production of specific racemates, 103Ð104 Cherry flavour, 20 recent reviews, 114 Chewing gum reactor type, 113 flavour delivery, 223 use of mixed organisms, 113 non-equilibrium in vivo model, 223 value of market, 114 Chewing gum and flavour, 9 Biotechnology of flavours, 89 Chitosan Biotransformation flavour delivery, 203 using plant cells, 107 Chocolate flavour, 80, 152 Blackcurrant flavour, 167 cis-3-Hexenol, 2 Blobs Citrus flavours, 171 in fMRI, 338 Citrus processing, 172 Blotters, 5 Cocoa flavour, 80 BOLD response Coffee and Maillard reaction, 68 in fMRI, 328Ð329 Coffee flavour, 154 Brain anatomy, 320 Cognitive effects, 301 Brain function, 320, 327 Compatibility of molecules, 179 Brain imaging, 319 Cone voltage methodologies, 327 fragmentation control, 274 Brassica flavours, 163 Consumer Breath-by-breath genetic engineering, 108 data analysis, 282 Consumer testing ester delivery from confectionery, 274 numbers and choice of members, 309 flavour delivery, 273 Consumer tests, 304, 311 Breath-by-breath aroma delivery, 281 methodology, 304 Building blocks Consumer view savoury, 74 natural flavours, 90 Butanedione, 60 Consumption of flavour, 136 Butter flavour, 148 Controlled flavour release, 202 Corona discharge, 273 C9 compounds Cortical pathways taste properties, 165 somatosensation, 320 CAMOLA, 54, 56 Coumarin Capillary GC columns, 249 limit in flavours, 35, 44 Capsaicin Council of Europe limit in flavours, 44 inventories concept, 34 perceptual effects, 160 1988 Directive, 31 Caramel flavour, 80 role in legislation, 28 Carbohydrate fragmentation, 58 scope of 1988 directive, 33 Carry-over effects, 307 Cranial nerves, 321 Carveol Cream flavour, 148 bioformation yam cells, 107 Cream soda, 20 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

Index 353

Cryogenic traps Distillation methods aroma extraction, 233 aroma extraction, 235 Cyclodextrin Ð as flavour reservoir, 283 Droplet size Ð emulsions, 220 Cyclodextrins, 195 Dynamic headspace Cysteine aroma extraction, 234 flavour conjugates, 145 role in meat flavour formation, 75 E number system, 39 Cytochrome P450, 93 Echo time in fMRI, 330 Dairy flavours, 147 Economics of flavour production, 146 Dalton’s law, 192 EEG Damascenone brain function measurement, 327 raspberry flavour, 17 Electronic noses Data acquisition aroma analysis, 261 in fMRI, 330 Electrospray ionisation realignment in fMRI, 334 direct MS in the liquid mode, 280 slice timing in fMRI, 334 Electrostatic polymer–flavour interactions, 184 Data analysis EMG in fMRI, 333 for determination of swallowing times, 340 sensory data, 309 Empirical modelling, 224 Data-driven models of flavour delivery/release, 207 Emulsifiers 4-Decanolide Ð bioformation, 96 effect on partition, 211 Deoxyosones, 54 Emulsion reaction to produce flavours, 55 mass transfer, 219 Descriptive analysis, 301 partition coefficient, 212 Detection limits Emulsion-based flavours, 13 GC/MS, 232 Encapsulation, 178, 202 in vivo monitoring, 269 1,2-Enolisation, 53 MS, 257 2,3-Enolisation, 58 Dicarbonyl formation, 61 Encapsulation Difference from control tests, 303 materials used, 179 Diffusion of flavours, 184, 186, 199 Enzyme modification for specific substrates, 104 Dimethyl resorcinol, 2 Enzyme technology, 101 Dimethyl sulfide raspberry flavour, 17 manipulation of reactions, 102 Direct mass spectrometry, 259 Enzymes analysis of tastants, 270 formation of artefacts during aroma isolation, assigning ions to compounds, 277 241 background, 271 from genetically modified organisms, 109, 111 breath sampling, 271Ð283 from plants, 140 calibration, 276 plant callus, 105 chemical ionisation, 272 plant cell suspensions, 105 compound identification, 277 plant cells, 104 effect of processing conditions, 287 plant sources, 104 electron-impact, 271 Episuite, 214 fruit quality, 289 Epoxidation future developments, 290 by biotransformation, 93 measuring in vivo flavour delivery, 280 Essential oil clarification, 12 miniaturisation, 290 Esterification process flavours, 286 bioformation in organic solvents, 103 reagent ions, 272 bioformation of flavours, 102 MS sampling frequency and detail, 283 Ethanol soft ionisation, 271 flavour solvent, 12 suppression, 277, 286 Ethics and regulations tastant analysis, 279 sensory testing, 308 tomato aroma, 289 Ethyl acetate Discrimination tests, 296 raspberry flavour, 17 Distal stimulus, 267 Ethyl methyl phenyl glycidate, 7 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

354 Index

Ethyl octanoate breath-by-breath, 273 partitioning behaviour, 218 chitosan, 203 European legislation complexity, 202 background, 30 diffusion, 199 current situation, 48 dynamic conditions, 197 definition of natural 2008, 47 effect of gas flow, 222 interpretation, 40 effect of saliva, 198 process flavour, 43 equilibrium conditions, 197 role of EFSA, 41 ethanolic beverages, 285 2008 definitions, 42 hard candies, 201 2008 onwards, 41 interfacial mass transfer, 198 traditional processes, 42 kinetics, 197 European list of permitted flavours, 45 liquid foods, 200 Experimental design measuring tastants in vivo, 270 sensory tests, 306 models, 207 Extraction multifactorial studies, 203 static headspace, 232 non-equilibrium model, 199 Extraction of flavours penetration theory, 199 plants, 142 semi-solid foods, 200 solid foods, 201 Fat stagnant film model, 199 effect on ice cream flavour formulations, 138 viscosity, 200 Fatty acid viscous foods, 285 bioformation, 99 Flavour delivery systems, 202 FD (dilution factor), 298 Flavour effects on texture and stability, 9 FEMA Flavour formation role in legislation, 26 secondary plant metabolic pathways, 106 Ferulic acid Flavour formulation biotransformation, 99 compounding, 15 Fick’s law, 184 number of ingredients, 15 FID flame ionisation detectors, 254 raspberry as example, 16 Film reactor, 287 temperature at consumption, 11 Flame ionisation detectors (FID), 254 Flavour glycosides, 101, 145 Flame photometric detector Ð GC, 233 as flavour release systems, 112 Flavour analysis Flavour industry complexity, 268 history, 1, 140 sensitivity, 269 value, 146 speed for in vivo monitoring, 268 Flavour intensity of mixtures, 6 Flavour balance, 6 Flavour legislation Flavour binding methods, 24 carbohydrates, 195 overview, 24 chemical forces, 193 Flavour matching, 21 competition for binding sites, 197 Flavour perception, 129, 190 lipids, 194 temporal measurement, 299 polysaccharides, 195 Flavour precursors, 139 proteins, 10, 196 enzymic hydrolysis, 101 starch, 10, 196 Flavour production, 15 to foods, 193 Flavour profiles, 302 Flavour characteristics, 3 Flavour receptors, 129 Flavour complexity, 6 modelling structure for optimum binding, Flavour composition and food matrix, 8 131 Flavour concentrations signal processing, 130 in vivo, 269 Flavour reformulation Flavour creation low and regular fat foods, 283 art or science?, 22 using direct MS measurements, 281 Flavour delivery Flavour release, 186. See flavour delivery batch extraction principle, 215 Flavour skeleton, 4 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

Index 355

Flavour solvents, 11 Gas-solid volatile distribution, 289 Flavour stability GC detectors, 254 during processing, 10 GC/EI/APCI/MS storage, 10 assigning ions to compounds, 278 Flavour thresholds, 131 GC/EI/PTR/MS Flavouring assigning ions to compounds, 278 European definition, 28 GC/MS Flavourists sensitivity limits, 232 analyses and formulation, 2 GC/olfactometry, 244, 250, 297 approach to formulation, 1 identification of key odourants, 251 and customers, 22 optimum operating conditions, 251 reconstitution, 3 GCO (gas chromatography olfactometry), sensory panels, 8 65 training and mentoring, 8 GC×GC methods, 250 Flavours Gelatin gels dispersed, 13 modelling flavour delivery, 225 spray-dried, 14 Gels FlavrSavr modelling flavour delivery, 224 genetically modified tomato, 110 General linear model (GLM) Flory and Huggins equation, 182 fMRI, 336 fMRI, 319 Genetically modified organisms, 108 adaptation to flavour stimuli, 339 Glassy state, 178 administration of flavour stimuli, 338 Glyoxal, 58 aroma imaging, 343 G-protein receptors, 130 control solutions for flavour studies, 341 Grapefruit flavour, 167 effects on subjects, 338 GRAS, 15 flavour paradigm design, 338 comparison with EU list, 36 flavour processing, 338 role in legislation, 26 imaging cortical association, 344 Grassy flavours, 164 imaging fat ‘taste’, 345 Green flavours, 164 imaging texture, 345 Group contribution methods for estimating physical subject choice, 341 properties, 209 swallowing and aroma signal, 339 Gum acacia taste imaging, 343 encapsulant, 14 timing of stimuli paradigm, 340 Gustatory pathways, 321, 323 fMRI methodology, 328 schematic, 322 Fractionation preparative GC, 248 Harmonisation of legislation, 48 silicic acid, 247 Head movements using HPLC, 247 in fMRI, 334 Fractionation of aroma, 245 Headspace dilution acid/base, 245 modelling, 216 Fragrances from plants, 174 Heart cutting Free volume of polymer systems, 181 GC technique, 250 Fresh flavours, 164 Henry’s law, 192 Fruit flavour Hexanal manipulation through genetic engineering, continuous production from linoleic acid, 114 111 Heyns reaction, 52 FTNF Heyns rearrangement products (HRP), 54 definition, 27 High-resolution gas chromatography (HRGC), 249 Functional magnetic resonance imaging (fMRI), 319 Hildebrand equation, 180 Furaneol HPLC analysis, 254 bioformation, 96 HRF flavour enhancement, 5 in fMRI, 337 HRP (Heyns rearrangement product), 54 ␥-Decalactone pre- and post harvest, 146 Humidity Garlic, 161 effect on Maillard reaction, 287Ð288 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

356 Index

HVP (hydrolysed vegetable protein), 151 Lactones Hydration bioformation, 96 biotechnology, 94 dairy notes, 5 Hydrogen cyanide peach note, 4 limit in flavours, 35, 44 Latin square design, 307 Hydrolases Legislation for flavour bioformation, 102 natural flavours, 90 Hydrophobicity of flavours, 194 Legislation and flavour discovery, 25 4-Hydroxy-5-methyl-3(2H)-furanone (norfuraneol), Lemon oil, 9 55 LikensÐNickerson aroma extraction, 235 Limitations Identification of compounds electronic noses, 261 CAMOLA approach for Maillard products in direct fMRI measurements, 341 MS, 279 GC/olfactometry, 251 direct MS, 270, 277 HPLC analysis, 254 isobaric compounds in direct MS, identification of compounds by direct MS, 278 270 labelled precursors in direct MS, 279 model mouths, 267 Immobilised plant cells, 107 modelling, 214, 224 Impurities in flavours SPME, 239 metals, 34 Limonene natural substances, 35, 44 biotransformation, 92 Inclusion complexes - starch/aroma, 196 Line scales, 298 Infinite dilution, 192 Linear retention indices (LRI), 256 Infrared spectroscopy, 256 Lipid effect, 213 Instrumental sensory correlations, 266, 281Ð282, 312, Lipid–flavour interactions, 194 314 Lipid interactions Intensity rating tests, 298 pathway, 63 Interfacial mass transfer, 198 Lipid interactions with Maillard products, 62 Internal standards epoxy-2-alkenals, 64 aroma extraction, 242 Liposomes, 202 International preferences and tastes, 19 Lipoxygenase pathway IOFI genetic engineering, 112 guidelines on process flavours, 74 Liquid flavours, 11 opinion on legislation, 36, 48 oil soluble, 13 Ionisation energies, 272 water soluble, 11 Ionones Log P, 214 raspberry flavour, 17 effect on flavour delivery, 225 5-Inosine monophosphate (IMP), 70 predicting maximum flavour delivery, Isobaric compound identification, 278 226 Isolation techniques effect on compound recoveries, 230 MAFF Isothiocyanates, 163 UK legislation 1965, 31 Maillard flavour formation Jasmine cysteine and meat flavour, 75 raspberry flavour, 18 effect of pH, 76 JECFA effect of sugars and amino acids, 75 role in legislation, 27 effect of time/temperature, 76 effect of water, 76 Kinetics factors influencing, 74 flavour delivery, 197 optimisation, 77 Kovats indices, 256 reaction kinetics, 77 role of thiamine, 78 Labelling information for consumers sulfur sources, 79 European situation, 39 yeast, 79 Labelling requirements Maillard intermediates European situation, 38 flavour precursors, 54 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

Index 357

Maillard reaction in vivo delivery, 220 aroma compounds, 65Ð67 in vivo delivery from emulsions, 222 coffee, 68 information needed, 207 general aspects, 51 limitations, 214 general scheme, 52 mass transfer, 216 monitoring by direct MS, 278 maximum delivery in gelatin gels, 225 taste compound formation, 69 software, 214 Main effects using timeÐintensity data, 224 sensory data analysis, 310 validation, 208, 216, 227 Maltodextrin volatile transfer in the upper airway, 223 spray-dried flavours, 14 Monoterpene biotransformation, 91 Maltol MS-Nose, 273 bioformation, 96 Mushroom, 161 flavour enhancement, 5 Marangoni effect, 286 Natural flavouring Mass spectra, 258 European definition 1974, 28 Mass spectrometers lemon as example, 40 compound identification, 257 USA definition, 26 low and high resolution, 257 Natural flavours Mass transfer biotechnology, 89 effect of gas flow, 222 challenges in production, 128 Mass transfer coefficient, 222 consumer view, 90 boundary layer, 217 usage, 89 overall, 217 Natural sources Mass transport of flavour in polymer systems, 185 European definitions, 29 Mastication Natural sources of flavour, 127 flavour delivery, 201 Negative lists Matching aroma delivery legislation, 25 in low and regular fat foods, 281 NIF analysis, 253, 297 Meat flavour Non-equilibrium model patents, 55 chewing gum, 223 pathway, 57 Non-equilibrium model of flavour delivery, 199, Meaty flavour 215 sulfur compounds, 68 Norfuraneol, 55 MEG Norisoprenoid brain function measurement, 327 biotransformation, 95 Menthofuran Number of components in flavours, 229 limit in flavours, 44 Number of flavour chemicals, 132 Metal oxide aroma sensors, 261 Methane thiol Octanol/water partition coefficient, 214 bioformation, 100 Odour activity value, 251, 268 Micro-organisms Odour activity values for coffee, 134 source of enzymes, 91 Odour database, 256 Mint compounds Odour fatigue flavour, cooling and bitter thresholds, 142 in GC/olfactometry, 251 Model quality, 210 Odour thresholds Modelling in different media, 137 breath aroma concentrations, 221 for terpenoids, 132 delivery from dynamic systems, 214 Off-flavours, 138 empirical, 224 Oil fraction, 220 emulsions, 214 Olfactory pathways, 323 equilibrium conditions, 208 schematic, 324 flavour delivery, 207 On-line MS, 266 flow chart, 215 interfering factors, 270 headspace dilution method, 216, 218 Onion flavour in vitro dynamic, 208 bioformation, 107 in vivo, 208 effect of cooking, 162 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

358 Index

Oral somatosensory pathways, 325 EU conditions for manufacture, 43 schematic, 326 European definition 1995, 29 Orange oil, 172 other ingredients, 79 composition, 173 savoury, 78 Orange processing, 172 sweet products, 80 OSME analysis, 253, 297 Proline Oxygen permeability role in bread aroma, 75, 158 encapsulation systems, 185 PROP brain imaging, 345 Panellist fatigue, 298, 306 Propylene glycol, 11 Paradigm design acetal and ketal formation, 11 in fMRI, 332 Protein-based delivery systems, 203 Partial least squares, 313 Proteins Partition aroma binding, 196 equilibrium and in vivo comparison, 221 effect on allyl thiocyanate, 137 in vivo considerations, 221 Proton affinity, 272 Partition coefficient, 191, 208 ethanol, 286 air/water, 194 values for odourants, 273 air/emulsion, 212 Proton transfer reaction, 272 effect of lipid, 211 Proton transfer reactionÐmass spectrometry (PTR-MS) emulsions, 218 principles, 275 environmental experiences, 210 Psychophysics of flavour, 131 estimation using QSPR, 209 PTR-MS, 276 oil/water, 194 calculation of amounts analysed, 277 relationships between Kaw and Kow, 211 Pulegone Peach flavour, 169 limit in flavours, 35, 44 Pear flavour, 167 Purge and trap Penetration theory of flavour delivery, 199 aroma extraction, 233 Pentanedione, 58, 60 Pentylpyridine formation, 63 QSPR Perillic acid flavour reformulation, 282 bioformation, 92 gel flavour delivery model, 224 Peripheral nervous system (PNS), 320 QSPR modelling pH and effect on Strecker products, 62 general procedures, 210 Phenylacetaldehyde formation pathway, solubility, 210 62 Quality control checks Phenylethanol using aroma profiles, 244 bioformation, 98 using sensory test, 303 Physical chemistry, 191 Quality of flavours, 136 Physical properties of flavours Quantitative structureÐproperty relationship (QSPR), encapsulation, 180 195 Plant organ cultures Quassin bioformation of flavour, 105 limit in flavours, 44 Plasticisation of polymers, 181, 186 Popcorn flavour, 80 Racemate resolution, 103 Positive lists RAS legislation, 24 validation of delivery models, 214 Post-harvest changes in flavour, 145 Raspberry flavour, 168 Powdered flavours, 11 Raspberry flavour composition, 16, 18 Preference mapping, 312 Raspberry ketone Primary character, 3 bioformation, 98 Principal component analysis (PCA), Reaction kinetics 311 Maillard flavour formation, 77 Process flavour Reagent ions manufacture, 74 direct mass spectroscopy, 272 reaction scheme, 56 Recombinant DNA (rDNA) difficulty with definition, 32 plant techniques, 110 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

Index 359

Recoveries disadvantages, 248 solvent extraction, 237 fractionation of aromas, 248 SPME, 238 Smoke flavour Regions of interest (ROI) choice of woods, 36 in fMRI, 338 difficulty with definition, 33 Release of flavours, 184 2003 EU Directive, 40 Repetition time SNIF analysis, 253 in fMRI, 332 Software for modelling, 214 Restrictions on natural ingredients in flavours, 46 Solid foods Retention times flavour delivery, 201 GC compound identification, 255 Solid-phase micro-extraction (SPME), 238 Reversible flavour binding, 196 Solubility parameters synthetic and natural biopolymers, 181 SAFE (solvent-assisted flavour extraction), 243 Solvent extraction Safety of flavours, 147 aroma isolation, 237 Safrole Solvents limit in flavours, 35, 44 essential oils, 173 Salting-out effect, 195 Sorption isotherms Santonin polymer-solvent systems, 183 limit in flavours, 35, 44 Sorption properties of flavours, 182 Savoury flavours, 78 Spatial smoothing Scatchard equation, 196 in fMRI, 334 SCOOP Species-specific meat flavour, 78Ð79 flavour specification, 37 Spice flavours, 159 Scoville units, 160 Spider diagrams, 302 Secondary characteristics, 4 SPME, 238 Selected ion flow tubeÐmass spectrometry (SIFT-MS), operation, 239 275 recoveries, 238 Selected ion monitoring (SIM), 258 Spray-dried flavours, 14, 178 Semi-solid foods Stagnant film model of flavour delivery, 199 flavour delivery, 200 Starch Sensitivity of flavours, 7 aroma binding, 196 Sensory analysis, 296 Starter cultures, 149Ð150 data analysis of TI results, 300 genetically modified, 109 effect of administrative route, 300 Static headspace methods, 232 partial least squares, 313 Statistical analysis references, 298 sensory data, 309 training needs, 296 Statistical considerations Sensory quality control tests, 303 in fMRI, 332, 337 Sensory testing Steam distillation, 143 administration, 304 aroma extraction, 235 avoiding bias, 305 Stereochemical effects on flavour, 132, 141 carry-over effects, 307 Steven’s law, 267 data analysis, 309 Stimuli ethics, 308 association in brain, 325 evaluation of panellist/judge performance, 309 interaction in brain, 325 locations for consumer tests, 305 Stir bar panellist/judge training and criteria, 308 aroma extraction, 240 power required, 311 Storage protocols for sample presentation, 305 stability of flavours, 138 selection of judges, 308 Storage studies Sensory tests flavour changes, 244 experimental design, 306 Strawberry flavour, 4, 168 Sesquiterpenes Strecker aldehydes, 64 biotransformation, 95 Strecker intermediates SIFT-MS, 275 effect of humidity, 288 Silicic acid Strecker reaction, 52, 61 P1: OTA/XYZ P2: ABC ind BLBK221-Taylor/Linforth November 27, 2009 19:6 Printer Name: Yet to Come

360 Index

Strecker aldehydes, 61 TimeÐintensity Sugar fragmentation pathway, 60 use in modelling, 224 Sulfur compounds TimeÐintensity rating, 299 aroma and origin, 70 Timing of flavour delivery during eating, Superadditive effects 268 brain imaging, 344 Tomato flavour, 169 Supercritical carbon dioxide genetic modification, 110 aroma extraction, 237 Triacetin, 12 Super-tasters Trigeminal nerve, 321 brain imaging, 345 Trimethyl citrate, 12 Suppression Trimethylpyrazine, 2 ion sources, 277 Tropical fruit flavours, 170 Surface area Tutti-frutti, 20 flavour delivery, 201 Sweet process flavours, 80 Ultra-high-field scanners Sweeteners, 158 fMRI, 346 Umami taste compounds, 69 Taste compounds United States analysis by HPLC, 255 flavour legislation, 26 Maillard, 69 Taste effects in flavour formulation, 5 Vacuum distillation Taste enhancement, 5 aroma extraction, 235 Taste-modifying compounds, 69 Vagus nerve, 321 Taste thresholds Validation of models, 227 for terpenoids, 132 van der Waals forces, 193 TasteÐaroma interactions, 174 Vanilla flavour, 3, 170 TasteÐsmell interactions, 301 Vanillin TCA (thiazolidine carboxylic acids), 54 bioformation, 97 Tea flavour, 153 bioformation by chilli cells, 107 Temporal dominance of sensation (TDS), 299 bioformation from ferulic acid, Temporal filtering 108 in fMRI, 336 Vapour pressure, 192, 209 Temporal measurement of flavour perception, 299 Variation in aroma delivery Tenax trapping effect of people, 281 aroma extraction, 233 Variation in olfactory perception Terpene biosynthesis effect of people, 344 genetic modification, 111 Verbenol Theoretical models bioformation, 93 emulsion flavour delivery/release, 219 Viscosity Theoretical models of flavour delivery/release, 207 flavour delivery, 200 Theory of encapsulation, 179 Volatility Thermal desorption to extract aroma compounds, 231 aroma extraction, 236 Thermodynamics in flavour release, 191 Water clusters, 273 Thiamine Wine flavour, 156 role in Maillard reaction, 78 WONF Thiols definition, 27 aromas in process flavours, 70 Threshold tests, 297 Yeast Thujone role in Maillard reaction, 79 limit in flavours, 35, 44 Yeast extract flavour, 68 P1: SFK/UKS P2: SFK BM BLBK221-Taylor/Linforth November 28, 2009 15:28 Printer Name: Yet to Come P1: SFK/UKS P2: SFK BM BLBK221-Taylor/Linforth November 28, 2009 15:28 Printer Name: Yet to Come