Impact of Plasmin Activity on the Shelf Life and Stability of UHT Milk

Home , Chymosin

Impact of plasmin activity on the shelf life and stability of UHT milk

PhD thesis by Valentin Maximilian Rauh August 2014

Aarhus University Arla Strategic Innovation Center

Faculty of Science and Technology Milk Science & Functionality

Department of Food Science Protein Chemistry

Blichers Allé 20 Rørdrumvej 2

8830 Tjele 8220 Brabrand

Denmark Denmark

Preface

This PhD thesis concludes three years’ work at Arla Strategic Innovation Center and the Department of Food Science, Aarhus University. This industrial PhD project was a collaboration between Aarhus University, Arla Foods, Copenhagen University and Lund University. As part of the PhD study, a three months stay at Copenhagen University, Department of Food Science was carried out.

Main supervisor Associate Professor Marianne Hammershøj Department of Food Science, Aarhus University

Co-supervisors Professor Lotte B. Larsen Department of Food Science, Aarhus University

Professor Richard Ipsen Department of Food Science, Copenhagen University

Professor Marie Paulsson Department of Food Technology, Lund University

Mette Bakman Arla Strategic Innovation Center, Arla Foods

Assessment committee Senior researcher Niels Oksbjerg (chairman) Department of Food Science, Aarhus University

Dr. Thom Huppertz Dairy and Ingredient Technology, NIZO food research B.V., Ede, The Netherlands

Prof. Dr.-Ing. Habil. Jörg Hinrichs Institute of Food Science and Biotechnology, Universität Hohenheim, Stuttgart, Germany

Acknowledgements

This industrial PhD project was made possible by funding from the Danish Agency for Science, Technology and Innovation (DASTI) and Arla Foods.

I would like to thank my university supervisors Marianne Hammershøj, Lotte B. Larsen, Richard Ipsen and Marie Paulsson for the valuable discussions and support during the project. Each one of you contributed with a different specialization and experience, so that every aspect of the project, from processing, enzymes and dairy chemistry to statistics and paper writing, was covered.

The innovation department of Arla Foods underwent major changes in the last three years. With these changes, new people came into the project and others had to leave. I would like to thank all of my Arla supervisors, Carola Edsman, Anja Sundgren, Mette Bakman, Jacob H. Nielsen and Grith Mortensen for the support during the project. A special thanks to Mette, Anja and Grith for being great managers and always believing in me and the project.

I would like to thank the protein chemistry team in the Arla Strategic Innovation Center, Vibeke Mortensen, Betina Hansen, Lene B. Johansen and Mette Christensen for the countless hours of listening to and discussing results, ideas, hypotheses and the assistance with the analysis. A special thanks to Lene for introducing and teaching me the wonderful world of mass spectrometry; without this the project would have looked very different! I also would like to thank all colleagues in the Arla Strategic Innovation Center, both in Stockholm and Brabrand, for teaching me new techniques, helping me with analysis and conducting the pilot plant trials.

Thanks to my fellow PhD students both at Aarhus University, Copenhagen University and Arla for fruitful discussions and a great working environment. Thilo Berg is thanked for proofreading the thesis. Thanks to Nanna S. Villumsen and Therese Jansson for great scientific discussions and motivational talks.

I would like to thank my family and friends in Germany, Denmark and Sweden for a great time, for motivating me and showing me that there is a life outside the PhD. Thank you Vera and Thilo for your hospitality and always making me feel at home when I was in Stockholm and Copenhagen.

Valentin Rauh, August 2014

Abstract

Milk production in Europe is increasing, while the markets for dairy products are saturated and milk consumption is decreasing. The production of high quality long shelf life products will enable the dairy industry to distribute products into new markets and increase export into existing growth markets. Ultra high temperature (UHT) treatment of milk aims to give a commercially sterile product with a shelf life of three months or more without refrigeration. The shelf life of UHT milk is determined by deterioration of flavour and functional properties, which can be caused by either heat induced chemical changes or the activity of enzymes in milk. The principal enzyme in good quality milk which can survive UHT treatments is the indigenous protease plasmin. Plasmin is the active part of an enzyme system consisting of the inactive precursor plasminogen and several activators and inhibitors, which can interact with different milk constituents. The main objective of this PhD project was to investigate the effect of residual plasmin activity on physical-chemical changes in UHT milk.

Due to the complexity of the plasmin and milk system, interpretation of plasmin activity measurement results can be difficult. In the first part of this project, a spectrophotometric assay for plasmin and plasminogen derived activity was optimized and extended to turbid samples and allowed measurement of plasmin activity in fat containing milk samples. The measured plasmin activity was largely affected by the sample preparation and assay conditions. The developed assay reduces the risk of underestimation of plasmin activity in milk and can be modified for different samples and purposes.

The effect of residual plasmin activity on physical-chemical changes in milk during storage at 20 ºC were studied in direct steam infusion heat treated milk with ultra-short holding times. Plasmin activity resulted in extensive proteolysis of caseins and a bitter off flavour appeared after six weeks of storage at which time more than 60 % of αS- and β-caseins were hydrolysed. After 10 weeks of storage more than 90 % of αS- and β-caseins were hydrolysed and the milk began to form a gel. Bitterness was found to be the most important shelf life limiting factor for milk with residual plasmin activity. Identification of peptides in this UHT milk with residual plasmin activity allowed the exploration of the proteolytic pathway of plasmin in a milk system. 66 casein derived peptides were formed by the plasmin, of which 23 could have been responsible for the observed bitterness. Plasmin showed the

III highest affinity for the hydrophilic regions in the caseins and the least affinity for hydrophobic and phosphorylated regions.

Protein lactosylation, the initial stage of the Maillard reaction, was studied in different UHT milks over a storage period of six months by liquid chromatography-mass spectrometry (LC-MS) and the common marker furosine. The initial furosine concentration in direct steam infusion heat treated milk was found to be 10-15 times lower as compared to an indirectly processed UHT milk, but no significant difference in the increase of furosine concentration during storage was found between the milks. A quantitative correlation lactosylation measured by LC-MS and furosine concentration was established, allowing for a simpler and faster analysis of protein lactosylation.

The results presented in this thesis expand our understanding of the plasmin system in milk and can help to control plasmin activity in UHT milk. This gives a direct way to improve the quality and shelf life of UHT milk products. Furthermore, the fundamental findings of this thesis can be transferred and applied to other dairy products.

Sammendrag (abstract in Danish)

Mælkeproduktion i Europa er stigende, samtidig er markedet for mejeriprodukter mættet og forbruget af mælk er faldende. Fremstilling af højkvalitetsprodukter med lang holdbarhed vil gøre det muligt for mejeriindustrien at lancerede mejeriprodukter på nye markeder og samtidig øge eksporten til eksisterende vækstmarkeder.

Formålet med ultra høj temperatur (UHT) behandling af mælk er at opnå et kommercielt sterilt produkt med en holdbarhed på minimum tre måneder udenfor køl. Holdbarheden af UHT mælk begrænses af smagsforringelse og ændrede funktionelle egenskaber, der kan være forårsaget af enten varmeinducerede kemiske ændringer eller restaktivitet af mælkens naturlige enzymer. Det vigtigste enzym i mælk af god kvalitet er den naturligt forekommende protease plasmin, der kan overleve UHT behandlingen. Plasmin er den aktive del af et enzymsystem, der består af den inaktive proenzym, plasminogen, samt forskellige aktivatorer og inhibitorer, der kan interagere med de forskellige komponenter i mælken. Hovedformålet med dette PhD projekt var at undersøge effekten af restplasminaktivitet på fysisk- kemiske ændringer af UHT mælk.

Tolkning af plasminaktivitetsmålinger kan være vanskelig på grund af kompleksiteten af plasmin- og mælkesystemet. I den første del af dette projekt blev en spektrofotometrisk metode til måling af plasmin og plasminogenafledt aktivitet optimeret og udvidet til at kunne måle turbide prøvematricer, hvilket muliggør måling af plasminaktivitet i fedtholdige mælkeprøver. Den målte plasminaktivitet blev i høj grad påvirket af prøveforberedelsesmetoden og målevilkår. Udviklimngen af metoden reducerede risikoen for underestimering af plasminaktivitet i mælk og kan modificeres til måling af forskellige prøver under forskellige vilkår.

Effekten af restplasminaktivitet på fysisk-kemiske ændringer i mælk opbevaret ved 20 °C blev undersøgt. Mælken blev varmebehandlet med direkte dampinfusion i ultrakort tid. Aktiviteten af plasmin resulterede i omfattende proteolysis af kaseiner og dannelse af bitter afsmag efter seks ugers opbevaring, hvor mere end 60% af S- og β-kasein var hydrolyseret. Efter 10 ugers opbevaring var mere end 90% af S- og β-kasein hydrolyseret og mælken begyndte at gelere. Bitter smag blev fundet at være den vigtigste holdbarhedsbegrænsende faktor for mælk indeholdende restplasminaktivitet. Identifikation af peptider i UHT mælk med restplasminaktivitet gjorde det muligt, at udforske

V plamins proteolytiske nedbrydningsmønster i mælk. 66 kaseinafledte peptider blev dannet af plasmin, hvoraf 23 kan være årsag til den observerede bitterhed. Plasmins affinitet overfor spaltningssteder i kasein er anderledes i UHT mælk sammenlignet med kaseinmodelsystemer. Plasmin viste højest affinitet overfor de hydrofile regioner i kasein og mindst affinitet overfor de hydrofobe og fosforylerede regioner.

Laktosylering af proteiner, der er den indledende fase af Maillard-reaktionen, blev undersøgt i forskellige UHT mælkeprøver over en lagrignsperiode på seks måneder ved anvendelse af væskekromatografi koblet med massespektrometri (LC-MS) og markøren furosin. Den indledende furosinkoncentration i mælk varmebehandlet med direkte dampinfusion blev målt til at være 10-15 gange lavere end indirekte UHT behandlet mælk, men der blev ikke påvist en signifikant forskel i stigningen af furosin under lagring. En kvantitativ korrelation mellem laktosylering målt ved LC- MS og furosinkoncentrationen blev etableret.

Resultaterne præsenteret i denne afhandling har udvidet forståelsen af plasminsystemet i mælk og kan bidrage til at kontrollere plasminaktivitet i UHT mælk. På denne måde kan der opnås en forbedret kvalitet samt holdbarhed af UHT produkter. Yderlig kan de fundamentale resultater fra denne afhandling overføres og anvendes på andre mejeriprodukter.

Zusammenfassung (abstract in German)

Die Milchproduktion in Europa nimmt stetig zu, während der Markt für Molkereiprodukte gesättigt und der Milchkonsum rückläufig ist. Die Produktion von qualitativ hochwertigen Produkten mit langer Haltbarkeit ermöglicht der Milchindustrie neue Märkte zu erschließen und den Export in bereits existierende Wachstumsmärkte zu erhöhen. Das Ziel der Ultrahocherhitzung (UHT) von Milch ist ein kommerziell steriles Produkt mit einer Haltbarkeit von drei Monaten oder länger. Die Haltbarkeit von UHT Milch wird durch eine Verschlechterung des Geschmacks oder funktioneller Eigenschaften beinflusst, die entweder durch Hitze bedingte chemische Veränderungen oder Aktivität von Enyzmen in der Milch verursacht werden kann. Die endogene Protease Plasmin ist das wichtigste Enyzm in Milch von hoher Qualität, das die UHT Behandlung überstehen kann. Plasmin ist der aktive Teil eines Enzymsystems, das aus dem inaktivem Vorläufer Plasminogen sowie einigen Aktivatoren und Inhibitoren besteht, die mit verschiedenen Milchkomponenten interagieren können. Das Hauptziel dieses Projektes war die Untersuchung des Einflusses der Plasminaktivität auf physikalische und chemische Veräderungen in UHT Milch.

Die Interpretation von Plasminaktivitätmessungen gestaltet sich aufgrund der Komplexität der Milch und des Plasminsystems als schwierig. Der erste Teil dieses Projektes beschäftigte sich daher mit der Optimierung einer spektrophotometrischen Methode zur Messung von Plasmin- und Plasminogenaktivität. Die Methode wurde außerdem erweitert, sodass Messungen von trüben Proben möglich waren, was die Analyse von fetthaltigen Milchproben erlaubte. Die gemessene Plasminaktivität hing stark von der Probenaufbereitung und Analysekonditionen ab. Die entwickelte Methode reduziert das Risiko Plasminaktivität in Milch zu unterschätzen und kann darüberhinaus für verschiedene Proben und Zwecke modifiziert warden.

Der Effekt von Plasminaktivität auf physikalische und chemische Veränderungen in UHT Milch wärend der Lagerung bei 20 ºC wurde in einer Milch untersucht, die mit direkter Dampfinfusion mit extrem kurzer Haltezeit erhitzt wurde. Plasminaktivität führte zu erheblicher Proteolyse von Caseinen und ein bitterer Geschmack trat nach sechs Wochen Lagerung auf. Zu diesem Zeitpunkt waren mehr als 60 % der αS- and β-Caseine hydrolysiert. Nach 10 Wochen Lagerung waren mehr als 90 % der αS- and β-Caseine hydrolysiert und die Milch bildete ein Gel. Die Haltbarkeit der UHT Milch wurde hauptsächlich durch den bitteren Geschmack limitiert.

VII

Die Identifikation von Peptiden in der UHT Milch erlaubte eine Untersuchung des Proteolyseweges von Plasmin in einem Milchsystem. 66 von Caseinen abstammende Peptide wurden von Plasmin gebildet und 23 von diesen könnten den beobachteten bitteren Geschmack verursacht haben. Eine abweichende Affinität von Plasmin für Spaltstellen in der UHT Milch und Casein Modelsystemen wurde festgestellt. Plasmin zeigte die höchste Affinität für hydrophile Regionen der Caseine und die geringste Affinität für hydrophobe und phosphorylierte Regionen.

Die erste Phase der Maillard Reaktion, die Laktosylierung von Proteinen, wurde in verschiendenen UHT Milchtypen über einen Zeitraum von sechs Monaten mit Hilfe von Flüssigchromatographie mit Massenspektrometrie-Kopplung (LC-MS) sowie der Analyse der Furosinkonzentration untersucht. Die anfängliche Furosinkonzentration in Milch, die durch direkte Dampfinfusion erhitzt wurde, war 10-15 mal niedriger im Vergleich zu indirekt erhitzer UHT Milch. Es wurden jedoch keine signifikanten Unterschiede in der Zunahme der Furosinkonzentration während der Lagerung in den UHT Milchtypen gefunden. Eine quantitative Korrelation zwischen der mit LC-MS gemessenen Laktosylierung und der Furosinkonzentration konnte hergestellt werden, was die Analyse der Proteinlaktosylierung vereinfacht und verkürzt.

Die Ergebnisse in der vorliegenden Arbeit erweitern das Verständins des Plasminsystems in Milch und können dazu beitragen, Plasminaktivität in UHT Milch zu kontrollieren. Dadurch kann sowohl die Qualtität als auch die Haltbarkeit von UHT Milchprodukten verbessert werden. Darüber hinaus können die grundlegenden Erkenntnisse dieser Arbeit auf andere Milchprodukte übertragen und angewendet werden.

VIII

Table of contents

Preface ...... I

Acknowledgements ...... II

Abstract ...... III

Sammendrag (abstract in Danish) ...... V

Zusammenfassung (abstract in German)...... VII

List of included publications ...... XII

List of abbreviations...... XIII

1 Introduction ...... 1

2 Project overview ...... 2

2.1 Aim and objectives ...... 2

2.2 Experimental outline ...... 2

2.3 Overview of methods ...... 3

3 Summary of included publications ...... 6

3.1 Paper I: The determination of plasmin and plasminogen-derived activity in turbid samples from various dairy products using an optimised spectrophotometric method ...... 6

3.2 Paper II: Plasmin activity as a possible cause for age gelation in UHT milk produced by direct steam infusion ...... 6

3.3 Paper III: Plasmin Activity in UHT Milk: Relationship between Proteolysis, Age Gelation, and Bitterness ...... 7

3.4 Paper IV: Protein lactosylation in UHT milk during storage measured by liquid chromatography-mass spectrometry and furosine ...... 7

4 The plasmin system in milk ...... 8

4.1 Indigenous enzymes in milk ...... 8

4.2 The plasmin system in milk ...... 8

4.2.1 Components and localisation of the plasmin system in milk ...... 8

4.2.2 Molecular features of plasmin and plasminogen ...... 10

4.2.3 Activation of plasminogen ...... 12

4.2.4 Inhibitors of plasmin and plasminogen activators ...... 13

4.2.1 Specificity of plasmin ...... 14

4.2.2 Factors affecting the plasmin system ...... 18

4.2.2.1 Environmental factors ...... 18

4.2.2.2 Effect of milk processing on the plasmin system ...... 19

5 Measurement of the plasmin system activity ...... 23

5.1 Activity assays ...... 23

5.1.1 Plasmin and plasminogen...... 23

5.1.1.1 Effect of sample preparation on plasmin activity ...... 24

5.1.1.2 Effect of sample composition and buffers on plasmin activity ...... 25

5.1.1.3 Comparison of activity assays ...... 27

5.1.2 Plasminogen activators ...... 30

5.2 Immunoassays ...... 30

5.3 Other assays ...... 31

5.4 Proteolysis ...... 32

5.4.1 Hydrolysis of caseins ...... 32

5.4.2 Peptide formation ...... 33

5.4.3 Kinetic aspects of proteolysis during storage ...... 34

6 UHT milk ...... 39

6.1 Heat treatment of milk ...... 39

6.2 UHT milk processing ...... 39

6.2.1 Design of UHT processes ...... 39

6.2.2 UHT processing systems ...... 41

6.3 Heat-induced changes in milk ...... 43

6.3.1 Proteins...... 43

6.3.1.1 Whey proteins ...... 43

6.3.1.2 Caseins ...... 45

6.3.2 Heat inactivation of the plasmin system ...... 45

6.3.3 Effect of cysteine on plasmin and milk proteins during heat treatment ...... 49

6.3.4 Maillard reaction ...... 52

6.3.5 Lactulose ...... 53

6.4 Shelf life development of UHT milk ...... 54

6.4.1 Age gelation of UHT milk ...... 54

6.4.2 Flavour development...... 61

6.4.2.1 Chemical changes affecting flavour development ...... 61

6.4.2.2 Effect of enzymes on flavour ...... 62

6.4.3 Lactosylation of milk proteins during storage of UHT milk ...... 64

7 Conclusions ...... 68

8 Perspectives ...... 70

9 References ...... 72

List of included publications

This thesis is mainly based on the papers listed below and are referred to in the text by their roman numerals. The papers are included in the back of the thesis.

Paper I Rauh, V. M., Bakman, M., Ipsen, R., Paulsson, M., Kelly, A. L., Larsen, L. B., & Hammershøj, M. (2014). The determination of plasmin and plasminogen-derived activity in turbid samples from various dairy products using an optimised spectrophotometric method. International Dairy Journal, 38, 74-80.

Paper II Rauh, V. M., Sundgren, A., Bakman, M., Ipsen, R., Paulsson, M., Larsen, L. B., & Hammershøj, M. (2014). Plasmin activity as a possible cause for age gelation in UHT milk produced by direct steam infusion. International Dairy Journal, 38, 199-207.

Paper III Rauh, V. M., Johansen, L. B., Ipsen, R., Paulsson, M., Larsen, L. B., & Hammershøj, M. (2014). Plasmin Activity in UHT Milk: Relationship between Proteolysis, Age Gelation, and Bitterness. Journal of Agricultural and Food Chemistry, 62, 6852-6860.

Paper IV Rauh, V. M., Johansen, L. B., Ipsen, R., Paulsson, M., Larsen, L. B., & Hammershøj, M. (2014). Protein lactosylation in UHT milk during storage measured by liquid chromatography-mass spectrometry and furosine. Manuscript intended for International Journal of Dairy Technology.

XII

List of abbreviations

AU arbitrary unit α-la α-lactalbumin β-lg β-lactoglobulin BSA bovine serum albumin cn casein CLSM confocal laser scanning microscopy EDTA ethylenediaminetetraacetic acid ELISA enzyme-linked immunosorbent assay EACA ε-amino caproic acid ESI-MS electrospray ionisation mass spectrometry FAST fluorescence of advanced Maillard products and soluble tryptophan HMF 5-hydroxylmethyl-2-furfural (RP-) HPLC (reverse phase) high performance liquid chromatography LC liquid chromatography MS mass spectrometry NCN non-casein nitrogen MALDI MS matrix assisted laser desorption mass spectrometry MFGM milk fat globule membrane MS mass spectrometry PA plasminogen activator PL plasmin PP proteose peptone TCA trichloroacetic acid tPA tissue-type plasminogen activator UHT ultra high temperature uPA urokinase-type plasminogen activator UV ultra violet

XIII

1 Introduction

Almost all consumption milk is subjected to heat treatment during processing to inactivate pathogenic and spoilage microorganisms and to prolong the shelf life of the milk. Heat treatment of milk is always a balance between the inactivation of undesired components, i.e. microorganisms and enzymes, the preservation of desired properties and negative heat induced chemical effects. Depending on the severity of the heat treatment, the shelf life of milk is therefore limited by a decrease in organoleptic and functional properties, which can be caused by either spoilage microorganisms and enzymes or heat induced changes in milk. Low pasteurized milk (72 °C for 15-30 s) has an excellent flavour, however, the shelf life is only one to two weeks when refrigerated due to deterioration by spoilage microorganisms. On the other hand, UHT milk (135-150 °C for 1-10 s) is commercially sterile and has a shelf life from two months up to one year without refrigeration, but the flavour of the milk is significantly changed and further deteriorates during storage resulting in a low acceptance for UHT milk by consumers. The quality of UHT milk can be improved by optimisation of the heat treatment process. Direct UHT processes use steam to rapidly heat up the milk to the desired temperature. Thereafter the milk is rapidly cooled in a vacuum chamber. In indirect UHT processes, milk is heated in a tubular or plate heat exchanger by hot water and/or steam, which is separated from the milk by a metal surface. Milk from direct steam UHT processes has an improved flavour compared to milk from indirect UHT processes. The fast heating and cooling rates in direct steam processes reduce the severity of the heat treatment in regards to chemical changes (Datta, et al., 2002). Furthermore, the inactivation of microorganisms in the temperature range between 110-150 °C is mainly dependent on the temperature, whereas chemical changes are mainly dependent on the holding time when applying direct steam systems. Consequently, the heat treatment can be optimized by increasing the temperature while decreasing the heating time. This leads to the development of UHT processes outside the classic time temperature range operating at temperatures > 150 °C combined with ultra-short holding times of < 0.2 s (Holst, et al., 2010; Holst, et al., 2012; Huijs, et al., 2004). These processes enable the production of commercially sterile products with minimum chemical changes, but are not able to sufficiently inactivate heat resistant enzymes to ensure a long shelf life. Enzymes in milk are either indigenous enzymes already present in milk or are produced by bacteria in milk before processing. Of these enzymes, heat resistant proteolytic and lipolytic enzymes from psychotrophic bacteria and the heat resistant major indigenous protease plasmin have the largest impact on product deterioration during storage of this type of UHT milk, causing 1 off flavours and gelation (Datta & Deeth, 2003; McKellar, et al., 1984). While the problem of bacterial enzymes can largely be avoided by using high quality raw milk and minimizing the growth of bacteria in milk before processing, there is no possibility to eliminate plasmin in milk. An additional pre or post-heat treatment is usually applied to inactivate plasmin, which on the other hand will decrease the organoleptic quality of the milk (van Asselt, et al., 2008). While the heat inactivation of plasmin has been studied intensively, the effect of plasmin activity and proteolysis on UHT milk quality and the underlying mechanisms are only known to a limited extent. The main hypothesis of this project is that the shelf life of UHT milk with residual plasmin activity can be controlled, when proteolysis and quality defects by plasmin can be correlated. The second hypothesis is that the shelf life of the UHT milk can be predicted and optimised by correlating plasmin activity measured in an assay with the occurring proteolysis.

2 Project overview

2.1 Aim and objectives

The main objective of the project is to investigate plasmin mediated proteolysis and quality defects during the storage of UHT milk. The second objective is to correlate the proteolysis with plasmin activity measurements. Since plasmin is the active part of a complex enzyme system in a complex milk system, the measurement of plasmin activity depends largely on the used assay. The first objective is therefore to investigate the factors influencing plasmin activity in an assay and to develop a new or to improve existing assays.

2.2 Experimental outline

It was initially planned to study plasmin activity in whey and micellar casein besides milk. For this, a diafiltration experiment was conducted to obtain whey and casein fractions and to investigate plasmin activity during diafiltration. Due to lack of time and considerable process variation, it was decided to not further pursue this and the focus was put on milk only. The whey and casein fractions were used for initial testing of the plasmin and plasminogen derived activity assay. Different factors influencing the plasmin activity in different sample matrices were investigated during the method development of the plasmin activity assay. A pilot scale UHT experiment was carried out at the Arla Foods pilot plant in Stockholm. In total, five UHT milks, four processed with direct steam infusion and ultra-short holding times and varying pre-heat treatment and one indirectly heat treated UHT milk as reference were produced. The trial was performed in triplicate on three consecutive weeks. The milk was transported to the

Arla Foods Strategic Innovation Center in Denmark and the milk was stored at 20 °C and analysed on a weekly basis for 14-16 weeks. It was aimed for that the 4 directly processed UHT milk samples should contain four different levels of plasmin activity. However, the UHT treatment resulted in two UHT milks with similar plasmin activity and two UHT milks without plasmin activity. A correlation of proteolysis and plasmin activity was therefore not possible. The UHT milks without plasmin activity were further analysed on a monthly basis up to a storage time of six months to investigate chemical changes and protein modification during storage.

2.3 Overview of methods

A summary of the methods used in this project is given in Table 1. The detailed descriptions of the methods can be found in the included publications. The paper in which the method is described is annotated in the table.

Table 1: Overview of methods used in the project with the aim of the analysis and a short description of the principle. The paper in which the methods are described in detail are indicated in the table.

Analysed property Aim Method Principle Paper Plasmin activity Determination of plasmin Activity assay Plasmin activity is measured by the release of p-nitroaniline upon I & II activity hydrolysis of the substrate S-2251 over time using a spectrophotometer. Plasminogen derived Determination of Activity assay Plasminogen derived activity is measured after activation of I & II activity plasminogen derived plasminogen by uPA in the same way as plasmin activity. activity Colour Measurement of colour Colorimeter Colour was measured using the Lab colour space. It consists of a II changes in milk during lightness value L (scale ranges from 0 to 100, where higher L values storage equals whiter colour), the a value for the red-green axis (scale ranges from -100 to 100, where positive a value equals red and negative a values equals green) and the b value for the yellow-blue axis (scale ranges from -100 to 100, where positive b value equals yellow and negative b value equals blue). Particle size Investigation of particle size Static light Static light scattering uses the diffraction pattern of light from single II changes scattering particles into different directions by multiple detectors in a dilution. Particle size can be deduced from the diffraction pattern and the refractive index Microscopy Characterisation of gel Confocal laser Gel particles are stained with the fluorophores FITC for protein and nile II structure scanning red for fat which fluoresce green and red, respectively. The confocal microscopy microscope allows high resolution images of the gel microstructure at (CLSM) different depths of the sample and differentiation of fat and protein. Viscosity Determination of viscosity Rheometer Viscosity of milk was measured in a double cup system at 25 °C with a II in relation to gelation constant shear rate of 100 s-1for 120 s after a 15 s pre-shearing step. Furosine Analysis of initial Maillard HPLC The protein attached Amadori product lactulosyllysine is converted to IV reaction and protein furosine by acid hydrolysis at 115 °C and analysed by HPLC and diode lactosylation array detection.

Table 1: (Continued)

Analysed property Aim Method Principle Paper Protein analysis Measurement of protein LC-MS Milk proteins are dissociated by urea and sodium citrate, and reduced II composition by dithioerythritol. The proteins are separated by HPLC and detected by MS and UV. Analysis of whey protein LC-MS Caseins are precipitated by acidification to pH 4.6, and the soluble whey II denaturation proteins are separated by HPLC and detected by UV and MS. Denaturation of whey proteins is measured by the difference in concentration of pH 4.6 soluble whey proteins before and after heat treatment. Peptide analysis Identification of peptides LC-MS and The pH 4.6 soluble peptides are separated by HPLC and analysed by III tandem MS tandem MS. Both fixed fragmentation energies and dynamic fragmentation energies based on charge state and mass were used for peptide identification. Identification of LC-MS Polypeptides are identified by mass in the pH 4.6 soluble peptide II polypeptides fraction and protein composition after deconvolution of the mass spectra. Besides mass, phosphorylation patterns and genetic variants are used for identification. Protein modification Analysis of protein LC-MS Proteins are dissociated, reduced and separated by RP-HPLC. IV lactosylation Lactosylation is either detected by deconvolution of the summed mass spectra of a protein by a mass shift of +324 Da or directly in the summed mass spectra as a mass shift of +324 Da divided by the charge state of the investigated ion. Analysis of thiol-disulphide LC-MS Proteins are dissociated by urea and sodium citrate, but not reduced. - rearrangements by cysteine Rearrangement or thiol-disulphide interchange is characterized by the appearance of monomeric protein and/or appearance of cysteine attachment to proteins by a mass shift of +121 or +242 Da.

3 Summary of included publications

3.1 Paper I: The determination of plasmin and plasminogen-derived activity in turbid samples from various dairy products using an optimised spectrophotometric method

The objective of this study was to optimize and extend a spectrophotometric assay for plasmin and plasminogen-derived activity to enable measurement of turbid samples and different dairy products. The method was validated by assessing reproducibility, repeatability, level of detection and recovery of plasmin activity in different sample matrices. The level of detection and repeatability of this method were improved compared with previous spectrophotometric assays. The second objective was to explore the effect of sample preparation and dissociation of plasmin from caseins on plasmin activity. Skimming of the sample resulted in a decrease of plasmin activity by 30% in pasteurised and homogenised whole milk, while skimming did not affect plasmin activity in raw milk. Removal of the cream phase in homogenised milk samples by skimming lead to an underestimation of plasmin activity. Comparison of pasteurised milk with a micellar casein solution showed that the dissociation of plasmin and caseins by adding ε-amino caproic acid (EACA) and NaCl decreased interference by caseins in the assay, but increases inhibition of plasmin with serum-based inhibitory components.

3.2 Paper II: Plasmin activity as a possible cause for age gelation in UHT milk produced by direct steam infusion

In this study, the effect of plasmin activity in direct steam infusion heat treated milk with ultra- short holding times (>150 °C for <0.2 s) on physical-chemical changes in milk during storage was investigated. Preheating at either 72 or 95 °C for 180 s was performed. Milk pre-heated at 72°C showed extensive proteolysis, but no proteolysis of κ-casein was detected. Plasmin was identified as active protease and activation of plasminogen was observed as an increase in the rate of casein hydrolysis during storage. A bitter off flavour appeared after six weeks of storage and the milk contained less than 40% intact αS- and β-caseins when bitterness was noticed. Proteolysis in the stored samples correlated with a decrease in pH and with changes in colour. Gelation occurred after 10 weeks along with an increase in viscosity and almost complete proteolysis of αS- and β-caseins. In conclusion, plasmin mediated proteolysis was involved in the occurrence of age gelation and bitter off flavour. Bitterness preceded age gelation and was the shelf-life limiting factor.

3.3 Paper III: Plasmin Activity in UHT Milk: Relationship between Proteolysis, Age Gelation, and Bitterness

A peptidomic study on the directly heated UHT milk (>150 °C for <0.2 s) preheated at 72 °C for 180 s was performed to gain a deeper insight in the specificity and proteolytic pathway of plasmin in a milk system in relation to age gelation and the formation of bitter peptides. Sixty-six peptides from αS- and β-caseins could be attributed to plasmin activity during the storage period, of which 23 were potentially bitter. Plasmin exhibited the highest affinity for the hydrophilic regions in the caseins that most probably were exposed to the serum phase and the least affinity for hydrophobic or phosphorylated regions. The proteolytic pattern observed suggests that plasmin destabilizes the casein micelle by hydrolysing casein−casein and casein−calcium phosphate interaction sites, which may subsequently cause age gelation in UHT milk.

3.4 Paper IV: Protein lactosylation in UHT milk during storage measured by liquid chromatography-mass spectrometry and furosine

The initial stage of the Maillard reaction, protein lactosylation, occurs during the heat treatment of milk and continues during subsequent storage. We compared the initial lactosylation and rate of lactosylation during storage of milk proteins in directly and indirectly heated UHT milk using liquid chromatography (LC) coupled with electron spray injection mass spectrometry (ESI-MS). Furosine was used as an overall marker to allow a quantitative correlation of lactosylation measured by LC-ESI-MS in the UHT milks. Protein lactosylation increased during the storage period of 6 months at 20°C. The extent of protein lactosylation in the directly heated UHT milk was approximately 10-15 times lower compared to indirectly heated UHT milk. The rate of protein lactosylation during storage however was not significantly different for indirect and directly heated UHT milk. Both the initial extent and the rate of lactosylation positively correlated with the amount of lysine residues in the different proteins. A linear or exponential correlation with furosine concentration could be established for minor and major lactosylated proteins, respectively.

4 The plasmin system in milk

4.1 Indigenous enzymes in milk

Indigenous enzymes in milk have been subject to research for over a century and by now, about 70 enzymes have been identified in bovine milk (O’Mahony, et al., 2013). Indigenous enzymes, as opposed to exogenous enzymes, originate from the cow itself. The indigenous enzymes in milk originate from different sources. One source of enzymes is the blood plasma, and enzymes can enter the milk by leaky junctions between mammary cells. Another source is the mammary epithelial secretory cell cytoplasm. The milk fat globule membrane contains a large number of enzymes as it originates from the apical membrane of the mammary cell, which itself originates from the Golgi membranes. The last source of indigenous enzymes are somatic cells, which enter the milk from blood during bacterial infections (O’Mahony, et al., 2013). While most of the indigenous enzymes in milk do not have an obvious physiological role in milk, many of them can have a significant effect on the quality and properties of dairy products (Kelly & Fox, 2006). The enzymes most responsible for the deterioration of dairy products and relevant for the dairy industry are lipases and proteases. Milk contains two protease systems, plasmin and cathepsin, of which the first one is the principal protease in good quality milk, it has been extensively studied due to its high heat stability and impact on dairy products (Kelly, et al., 2006). In the following chapter, an overview of the plasmin system, factors affecting its action and its impact on dairy products are given.

4.2 The plasmin system in milk

4.2.1 Components and localisation of the plasmin system in milk

The major indigenous milk protease plasmin (EC 3.4.21.7) is a serine protease with a pH optimum of 7.5 at 37 °C (Grufferty & Fox, 1988b). Plasmin in milk is mainly present in its inactive form, plasminogen, which can be activated by tissue-type (tPA) or urokinase type (uPA) plasminogen activator (Bastian & Brown, 1996). The plasminogen concentration in fresh milk is 0.18-2.8 mg L-1 and 2-30 times higher compared to plasmin (0.1-0.7 mg L-1) (Benfeldt, Connie, et al., 1995; Ozen, et al., 2003; Richardson, B. C. & Pearce, 1981). Furthermore, the system is regulated by the presence of inhibitors for plasmin and plasminogen activators (Figure 1). Plasminogen is not expressed in the mammary gland and comparison of the amino acid sequence of plasmin and plasminogen from blood and milk show these to be identical from the two sources and it is thus demonstrated that plasmin originates from the blood (Benfeldt, Connie, et al., 1995; Berglund, et al., 1995).

Figure 1: The plasmin system in milk. Modified from Ismail & Nielsen (2010).

The physiological role of plasmin in blood is fibrinolysis, i.e. the solubilisation of fibrin clots, a precisely regulated process that mainly depends on the binding of the enzymes to their substrate fibrin (Cesarman-Maus & Hajjar, 2005). In milk, plasmin, plasminogen and tPA are mainly bound to the casein micelles (Baer, et al., 1994; Benfeldt, Connie, et al., 1995; Haissat, et al., 1994; Politis, et al., 1992; Politis, et al., 1994; Wang, L., et al., 2006; White, et al., 1995). uPA, on the other hand, has been shown to be associated with somatic cells by uPA receptors (Heegaard, Rasmussen, et al., 1994) and upon release from the somatic cells associates with the casein micelles as well (Heegaard, Rasmussen, et al., 1994; Lu, D. D. & Nielsen, 1993a; White, et al., 1995). The inhibitors of the plasmin system are located in the serum phase of milk (Christensen, S., et al., 1995; Precetti, et al., 1997).

4.2.2 Molecular features of plasmin and plasminogen

Bovine plasminogen has a calculated molecular mass of 88,092 Da comprising 786 amino acid residues and is glycosylated at two positions (Marti, et al., 1988; Schaller, et al., 1985). Plasminogen consists of an N-terminal pre-activation peptide, 5 homologous kringle structures and a C-terminal catalytic centre (Figure 2). The overall structure is stabilized by disulphide bonds. The kringle structures contain lysine binding sites with different affinities for lysine, by which plasminogen and plasmin bind to lysine residues in the casein micelles (Bastian & Brown, 1996; Benfeldt, Connie, et al., 1995). The amino acid composition of human and bovine plasminogen differs, but the position of the cysteine residues is the same and based on the sequences the kringles are homologous, indicating their structure is conserved (Bastian & Brown, 1996; Schaller, et al., 1985).

Figure 2: Schematic representation of plasminogen with the pre-activation peptide (PAP), kringle structures (K1-K5) and light chain containing the catalytic center. The cleavage sites and enzymes involved in the activation of plasminogen are indicated in red. Disulphide bonds are shown as dashed lines. Modified from Benfeldt et al. (1995).

The crystal structure of human plasminogen kringle 4 in complex with the lysine analogue ε- aminocaproic acid (EACA) is shown in Figure 3. As can be seen, the ε-amino group of lysine or

EACA is stabilized by two aspartic acid residues. Additionally, the hydrophobic C4 chain is stabilized by two phenylalanine and one tryptophan residue (Ponting, et al., 1992; Wu, et al., 1991). The binding of lysine or EACA to the kringle structures further results in a conformational change of the kringle structures (Wu, et al., 1991). Binding to lysine residues or lysine analogues also affects the entire plasminogen molecule; from a compact and folded to a looser, unfolded one (Hayes, K. D., et al., 2003; Mangel, et al., 1990; Markus, 1996). The electrostatic nature of the lysine binding sites can explain why plasmin is not dissociated from the casein micelles at low temperatures. On the other hand, addition of 1 M NaCl or at pH values < 4.6 plasmin dissociates from the casein micelles which is the cause for the higher plasmin activity in acid whey compared to sweet whey (Crudden & Kelly, 2003; Grufferty & Fox, 1988b; Kelly & McSweeney, 2003).

Figure 3: The refined structure of the ε-aminocaproic acid complex of human plasminogen kringle 4. The amino acids of plasminogen kringle 4 involved in the complexion of ε- aminocaproic acid are annotated (Wu, et al., 1991).

The catalytic centre of plasmin is located in the C-terminal region and contains the serine protease specific catalytic triad consisting of His-Asp-Ser located at the amino acid residues 598, 641 and 736, respectively (Hedstrom, 2002; Schaller, et al., 1985).

4.2.3 Activation of plasminogen

Plasminogen is activated by conversion from a single chain zymogen to a two chain active enzyme connected by disulphide bonds by cleavage of the peptide bond Arg557-Ile558 in plasminogen (Bastian & Brown, 1996). Plasmin is able to cleave plasminogen at Lys77-Arg78 which yields the so called pre-activation peptide. This cleavage results in a conformational change of plasminogen in which the Arg557-Ile558 bond is exposed. The rate of activation of plasminogen is therefore increased in the presence of even small amounts of plasmin (Bastian & Brown, 1996; Cesarman-Maus & Hajjar, 2005). Both uPA and tPA are serine proteases and both show some homology with plasmin. In the human system, both activators exist in both a single and a two-chain form, with different activation efficiency (Cesarman-Maus & Hajjar, 2005). Activation of plasminogen by tPA in blood is weak in absence of fibrin. The activation of plasminogen by tPA increases by several orders of magnitude when both are bound to fibrin in blood or casein in milk, respectively

(Cesarman-Maus & Hajjar, 2005; Markus, et al., 1993). Especially the αS2-casein dimer and κ- casein multimers are shown to accelerate the activation of plasminogen by tPA (Heegaard, Rasmussen, et al., 1994; Heegaard, et al., 1997). In contrast to tPA, uPA does not bind to fibrin and the rate of activation by uPA is not affected by the presence of fibrin (Lu, D. D. & Nielsen, 1993a; White, et al., 1995). Casein has been shown to enhance the activation of human plasminogen by uPA, which is most likely caused by the conformational change in plasminogen upon occupation of lysine binding sites as described above (Baer, et al., 1994; Markus, et al., 1993; Politis, et al., 1995). In model studies, both native and denatured α-la and denatured β-lg stimulated the activation of bovine plasminogen by human uPA. The enhanced rate of plasminogen activation was suggested to be caused by a faster release of plasmin from the plasminogen-uPA complex in the presence of whey proteins (Rippel, et al., 2004). Benfeldt et al. (1995) purified and characterised a low molecular weight plasmin/plasminogen with a molecular weight of 50 kDa. This so-called midi-plasminogen/midi-plasmin was hydrolysed between Arg343-Met344 and consisted only of kringle 4 and 5 and the light chain. The cleavage was suggested to be autocatalytic by plasmin or uPA. This midi plasminogen was especially found in the serum phase, suggesting that the loss of kringle 1-3 reduced the binding to caseins. Whether the full or midi form of plasmin and plasminogen are the predominant form in milk is unknown since the concentration of midi plasminogen could not be determined, but it was suggested that the midi-plasmin form was the major form of the bovine plasmin molecule in milk (Benfeldt, Connie, et al., 1995).

4.2.4 Inhibitors of plasmin and plasminogen activators

There are at least six inhibitors of plasmin present in milk: α2-antiplasmin, α2-macroglobulin, antithrombin III, C1-inhibitor, inter-α-trypsin inhibitor and bovine plasma-trypsin inhibitor

(Christensen, S., et al., 1995). The two most important milk inhibitors are the plasmin specific α2- antiplasmin and the non-specific α2-macroglobulin (Kelly & McSweeney, 2003). The α2- antiplasmin belongs to the inhibitor family of serpins. All serpins form an irreversible complex with serine proteases, which is followed by a cleavage of the inhibitor, and both the protease and inhibitor lose their activities (Cesarman-Maus & Hajjar, 2005; Christensen, S. & Sottrup-Jensen,

1992; Wiman & Collen, 1978). The function of α2-antiplasmin in blood is the inhibition and inactivation of potentially harmful plasmin being released from fibrin clots. The formation of the plasmin-α2-antiplasmin complex requires a free lysine binding site on plasmin. Christensen et al.

(1992) found that α2-antiplasmin inhibits both bovine plasmin and midi plasmin, suggesting that

α2-antiplasmin binds to the lysine binding site in kringle 4 of plasmin. The reaction involves a very fast first step in which α2-antiplasmin binds to plasmin and a slower second step in which the irreversible complex is formed (Christensen, U., et al., 1996; Wiman & Collen, 1978). Prado et al. (2006) found a competitive inhibition of plasmin in a plasmin inhibitor extract from milk containing α2-antiplasmin. Since the inhibitor was added to a mix of plasmin and substrate, the measured inhibition was most likely only the first step of α2-antiplasmin binding to plasmin. Based on their molecular weights, plasmin and plasmin inhibitor are present in an equimolar concentration in fresh milk (Prado, et al., 2006; Precetti, et al., 1997). Two types of fast reacting plasminogen activator inhibitors (PAI-I and PAI-II) have been identified in milk (Kelly & McSweeney, 2003). Besides specific protease inhibitors, plasmin can be inhibited by whey proteins. Both native and denatured forms of β-lactoglobulin act as inhibitors of plasmin activity towards a synthetic substrate and casein (Bastian, et al., 1993). Besides β-lg, also α-lactalbumin (α-la) and bovine serum albumin (BSA) inhibited plasmin as well. The variant β-lg A is shown to have a higher inhibitory effect than β-lg B (Politis, et al., 1993). This difference is surprising since the two variants differ only in two amino acid residues. Whether this finding is genuine or an artefact of the assay still needs verification. The mechanism of inhibition of plasmin by whey proteins, however, is still unknown. In case of a competitive inhibition of plasmin by β-lg, even small changes in the amino acid sequence could affect the binding of β-lg to the substrate binding site and explain the findings of Politis et al (1993), which has been shown on the catalytic efficiency of human plasmin towards similar peptide sequences (Hervio, et al., 2000). It is important to note

13 that the inhibitory effect of the whey proteins is different when synthetic substrates or casein is used as a substrate (Hayes, M. G., et al., 2002).

4.2.1 Specificity of plasmin

Plasmin shows the same specificity as trypsin and hydrolyses peptide bonds on the C-terminal side of lysine and, to a lesser extent, after arginine (Kelly & McSweeney, 2003). The primary substrate for plasmin in milk is β-casein. Plasmin hydrolyses β-casein primarily at Lys28-Lys29,

Lys105-His106 and Lys107-Glu108 which results in the so-called proteose peptones and γ-caseins, derived from C- and N-terminal part of the β-casein molecule (Eigel, et al., 1984). The nomenclature and sequence of these polypeptides are given in Table 2.

Table 2: γ-caseins and proteose peptones (PP) of β-casein by plasmin (Eigel, et al., 1984). Peptide Sequence

PP 8-fast 1-28 PP 8 slow 29-105, 29-107 PP component 5 1-105, 1-107

γ1-casein 29-209

γ2-casein 106-209

γ3-casein 108-209

Cleavage occurs also at Lys113-Tyr114 and Arg183-Asp184 (Kelly & McSweeney, 2003). In Paper

III, we did not find any cleavage by plasmin at Arg183-Asp184 in β-casein, but rather at Lys169-

Ala170 and Lys176-Ala177. Furthermore, the peptides identified in Paper III showed that plasmin hydrolysed β-casein in UHT milk at Lys29-Ile30, Lys32-Phe33, Lys48-Ile49, Lys97-Ala98 and Lys99-

Glu100 (Figure 4), suggesting that the proteose peptone fraction is more heterogeneous than previously reported.

Figure 4: Primary structure of the bovine β-casein variant A2 5P with plasmin cleavage sites described in literature and Paper III indicated by arrows (Kelly & McSweeney, 2003). The cleavage site at Arg183-Asp184 was not found in Paper III and is indicated red.

The storage time after which the peptides were identified and the rate of peptide formation indicated that the plasmin initially hydrolysed at Lys105-His106 and Lys107-Glu108 and sequentially continued to hydrolyse β-casein further around these cleavage sites at Lys97-Ala98, Lys99-Glu100 and Lys113-Tyr114 (Paper III). The only potential cleavage site in β-casein, which is hydrolysed by trypsin, but not plasmin, seems to be Arg202-Gly203 (Bumberger & Belitz, 1993).

Plasmin hydrolyses αS2-casein at approximately the same rate as β-casein (Kelly & McSweeney,

2003). The major previously reported cleavage sites include Lys21-Gln22, Lys24-Asn25, Arg114-

Asn115, Lys149-Lys150, Lys150-Thr151, Lys181-Thr182, Lys188-Ala189 and Lys197-Thr198 (Le Bars & Gripon, 1989; Visser, et al., 1989). By now, several additional cleavage sites have been identified from peptides occurring in the indigenous peptide profile of milk (Baum, Fedorova, et al., 2013; Dallas, et al., 2013). In Paper III, apart from known cleavage sites, we identified six additional plasmin specific cleavage sites in αS2-casein that have not been described before (Figure 5). The reason for the relatively scarce reports on cleavage sites in the literature for αS2-casein may be attributed to its relatively low concentration in milk and the dimeric nature of αS2-casein, which complicates peptide analysis of αS2-casein hydrolysates. The high and varying degree of phosphorylations on αS2-casein further impedes mass spectrometry analysis of αS2-casein peptides due to a reduced concentration of identical peptides varying in phosphorylations and the reduced ionisation and fragmentation ability of phosphorylated peptides (Baum, Ebner, et al., 2013; Tauzin, et al., 2003).

Figure 5: Primary structure of bovine αS2-casein A 11P with plasmin cleavage sites reported in literature and Paper III are indicated by arrows (Baum, Fedorova, et al., 2013; Dallas, et al., 2013; Le Bars & Gripon, 1989). The novel plasmin cleavage sites in αS2-casein reported in Paper III are indicated as grey arrows.

The activity of plasmin towards αS1-casein is lower compared to β-casein (Andrews & Alichanidis, 1983; Gazi, et al., 2014), which was also observed in Paper II. The major cleavage sites found for isolated αS1-casein are Arg22-Phe23, Arg90-Tyr91, Lys102-Lys103, Lys103-Tyr104,

Lys105-Val106, Lys124-Glu125 and Arg151-Gln152 (Le Bars & Gripon, 1993; McSweeney, et al.,

1993). In Paper III, we could not find indications for a cleavage at Arg151-Gln152 (Figure 6),

16 which is most likely due to a difference in the accessibility of cleavage sites on isolated and micellar casein, respectively. The region of αS1-casein, which includes Arg151-Gln152, is hydrophobic and could be inaccessible for plasmin in a micellar system. The importance of hydrophobic domains of the caseins was also shown in Paper III for the N-terminal peptide of

αS1-casein formed by plasmin αS1-casein f(1-22). Relative quantification of peptide concentrations measured by UV and corrected for their calculated molar absorption coefficient further showed that Arg22-Phe23 is not a major cleavage site in micellar casein, but that the initial hydrolysis of αS1-casein is in the region 90-105 and Lys124-Glu125 (Paper III). This is in line with the findings in Paper II, where αS1-casein f(1-124) and αS1-casein f(125-199) were the two large polypeptides present early in the storage period.

Figure 6 Primary structure of bovine αS1-casein B 8P with plasmin cleavage sites identified in literature and Paper II are indicated by arrows (Baum, Fedorova, et al., 2013; Le Bars & Gripon, 1993; McSweeney, et al., 1993). Dashed arrows indicate cleavage sites only found for isolated αS1-casein (Le Bars & Gripon, 1993; McSweeney, et al., 1993).

Plasmin has little activity on κ-casein compared to the other caseins. Reports on κ-casein hydrolysis by plasmin vary and the specificity has not been determined (Bastian & Brown, 1996; Dalsgaard, et al., 2007; Kelly & McSweeney, 2003). The low activity could be caused by the presence of carbohydrate moieties on κ-casein and its polymeric nature similar to αS2-casein (Bastian & Brown, 1996). Only few studies have looked into plasmin activity on whey proteins. In general, whey proteins are rather resistant to plasmin hydrolysis, although hydrolysis of BSA, lactoferrin, β-lg and α-la has been reported (Dalsgaard, et al., 2007; Hayes, M. G., et al., 2002; Kelly & McSweeney, 2003). In the peptidomic study conducted in Paper III, we identified 6 peptides originating from

β-lg. The cleavage sites were Lys8-Gly9, Lys71-Thr72, Lys75-Ile76, Lys71-Thr72, Lys83-Ile84 and

Lys91-Val92, which were described previously by (Caessens, et al., 1999). Plasmin hydrolysis of α-la has been reported in model studies (Dalsgaard, et al., 2008). The resistance of whey proteins towards hydrolysis is most likely due to their globular structure. The β-lg hydrolysed in the UHT milk of Paper II and Paper III was partially denatured and thus the structure was altered and easier to hydrolyse. The low levels of α-la denaturation in the UHT milk could explain the absence of α-la peptides.

4.2.2 Factors affecting the plasmin system

4.2.2.1 Environmental factors Activity of the plasmin system in milk can be affected by either changes in the transport of its components from blood to milk or differences in the activation of plasminogen (Kelly & McSweeney, 2003). Furthermore, presence of inhibitors of either plasmin or plasminogen activator inhibitors will affect the net activity of plasmin present. Different breeds of cattle show variations in plasmin activity with milk from Jersey cows having a lower plasmin activity than milk from Friesian and Montbeliard cows (Benslimane, et al., 1990; Richardson, B. C., 1983b). Plasmin activity has also been found to be increased in milk from older cows (Kelly & McSweeney, 2003). The stage of lactation can largely affect plasmin activity. Plasmin activity decreases and plasminogen activity increases in milk from the early lactation phase after calving (Benslimane, et al., 1990; Kelly & McSweeney, 2003). In late lactation milk, both plasmin, plasminogen and plasminogen activator activity are increased (Baldi, et al., 1996; Richardson, B. C., 1983b). The increased levels of plasmin and activation of plasminogen in late lactation milk have been linked to involution of the mammary gland (Politis, Lachance, et al., 1989; Politis, 1996). Plasmin, plasminogen and plasminogen activator levels are increased in mastitic milk (Kelly & McSweeney, 2003). This could be caused by an increased transfer of plasmin system components from the blood into the milk during mastitis due to compromised barrier between milk and blood as well as elevated plasminogen activator levels attributed to increased levels of somatic cells (Heegaard, Christensen, et al., 1994; Heegaard, Rasmussen, et al., 1994; Politis, Ng Kwai Hang, et al., 1989). In general, there is a positive correlation of plasmin activity with the somatic cell count (Albenzio, et al., 2004; Kennedy & Kelly, 1997; Politis, Ng Kwai Hang, et al., 1989). Besides mastitis, bacteria present in milk can affect the plasmin system as well. Proteases from psychotrophic bacteria are able to release plasmin and plasminogen from the casein micelles into the serum phase and are capable of inactivating plasmin (Fajardo-Lira & Nielsen, 1998; Frohbieter, et al., 2005). While most bacteria do not activate plasminogen, a protease from 18

Bacillus polymyxa shows plasminogen activator like activity (Larson, et al., 2006; Nielsen, S. S., 2002; Verdi & Barbano, 1991). A protease from Pseudomonas fluorescens does not activate plasmin itself, but is capable of increasing the catalytic efficiency of uPA by converting it into its two chain form (Guinot-Thomas, et al., 1995).

4.2.2.2 Effect of milk processing on the plasmin system Plasmin and plasminogen activators can be active during cold storage of milk resulting in increased proteolysis and levels of plasmin activity (Ismail & Nielsen, 2010; Somers, et al., 2002). Storage of milk at 5 °C, on the other hand, results in decreased levels of plasmin activity caused by autolysis of plasmin compared to milk stored at room temperature (Crudden, Fox, et al., 2005). Autolysis has previously been observed for isolated plasmin fractions and autolysis can be reduced by addition of lysine analogues or providing a protein network for plasmin (Saint- Denis, et al., 2001b; Ueshima, et al., 1996). The plasmin concentration used in these studies however is much higher compared to concentration of plasmin in milk and the casein in milk provides a suitable protein network to stabilize plasmin against autolysis. Most studies reporting autolysis of plasmin in milk use raw milk (or pressure treated milk) in their experiments (Crudden, Fox, et al., 2005; Guinot-Thomas, et al., 1995; Huppertz, et al., 2004). The binding of plasmin to casein is a reversible process and the decrease in plasmin activity could also be attributed to irreversible inhibition of released plasmin by plasmin inhibitors. No evidence for autolysis or irreversible inhibition exists in literature, but could help to understand the dynamics of the plasmin system during the storage of milk. Identification of either plasmin and plasminogen fragments or the presence of the 10 kDa peptide originating from the N-terminal part of α2-antiplasmin upon inhibition/inactivation of plasmin in stored raw milk could solve this question. Release of uPA from somatic cells could occur during storage of raw milk and prolonged storage could result in increased levels of uPA in milk. The effect of milk separation, bactofugation and spore filtration on uPA release from somatic cells has not been studied yet, but is expected to significantly affect the levels of uPA in milk. Heat treatment of milk greatly affects the plasmin system. Plasmin, plasminogen and plasminogen activators are generally considered as heat stable, while the inhibitors of plasmin and plasminogen activators are relatively heat labile. Low-pasteurisation conditions (72-75°C for 15 s) do not significantly affect plasmin, plasminogen and plasminogen activators (Ismail & Nielsen, 2010). In fact, pasteurisation results in an increased plasmin activity due to the inactivation of plasmin inhibitors and plasminogen activator inhibitors (Prado, et al., 2006;

Richardson, B. C., 1983a). Plasminogen activation during cold storage of pasteurized milk is thus increased compared to raw milk (Lu, R., et al., 2009). Plasmin inhibitors have been shown to have a higher heat stability than plasminogen activator inhibitors (Prado, et al., 2006). We examined the effect of heat treatment at 72 °C on plasmin activity in a small experiment as part of the method development for the plasmin assay (Paper I). Plasmin activity in the milk and the serum phase decreased slightly (Figure 7). While the recovery rate of added plasmin in the milk was constant during the heat treatment, the recovery rate in the serum increased by almost 10 % with increasing holding time. The increase in recovered activity most likely represented the inactivation of plasmin inhibitors during the heat treatment. The change in recovered activity fitted well to a first order reaction, indicating that this method could be used for a kinetic description of inactivation of plasmin inhibitors. Since the focus of the PhD project was on UHT milk, the inactivation of plasmin inhibitors was not pursued further in the project.

Figure 7: Effect of heat treatment at 72 °C for up to 195 s on plasmin activity (A) and recovery rate of added plasmin (B) in skimmed raw milk (□) and the serum phase (●) obtained by ultracentrifugation (105 rcf for 60 min) after the heat treatment. Bars indicate standard deviation, n = 3;

The increased plasmin activity after pasteurisation could also be attributed to the denaturation of plasminogen (temperature range 50.1-61.7 °C), which increases the activation by uPA (Burbrink & Hayes, 2006). Inactivation of plasmin and plasminogen during heat treatment mainly occurs via thiol-disulphide interchange reactions with denatured β-lg (Rollema & Poll, 1986). The thermal inactivation of plasmin, plasminogen and plasminogen activators will be covered in detail in section 6.3.2.

High pressure homogenisation has been shown to inactivate plasmin and plasminogen (Hayes, M. G. & Kelly, 2003). On the other side, more recent studies have indicated that these first findings were an artefact of the plasmin assay used in this study, which included a centrifugation step to remove the cream (and plasmin attached to it). High pressure homogenisation seems to have only a little effect on plasmin activity itself at elevated pressure (20 % plasmin inactivation at 200 MPa; Iucci, et al., 2008). High pressure treatment on the other hand has been shown to significantly affect the plasmin system. High pressure treatment at room temperature up to 400 MPa does not inactivate plasmin (García-Risco, et al., 2003; Huppertz, et al., 2004) and treatment at 300-400 MPa even promotes plasmin activity (Huppertz, et al., 2004). Plasmin activity decreases after high pressure treatments at 600 MPa. The increase in plasmin activity could be attributed to disruption of casein micelles and transfer of plasmin system components into the serum phase, while the inactivation of plasmin at higher pressures could be caused by pressure induced denaturation of β-lg (García- Risco, et al., 2003; Hinrichs & Rademacher, 2004; Huppertz, et al., 2004; Moatsou, et al., 2008). Studies on the pressure-temperature stability of isolated plasmin indicated a stabilizing effect of plasmin at pressures >600 MPa and temperatures <50 °C (Borda, et al., 2004). Cheeses made from milk concentrated by ultrafiltration have reduced plasmin activity and reduced rates of proteolysis (Benfeldt, C., 2006). The decrease is partially ascribed to the retention of whey proteins and inhibitors of plasmin and plasminogen activators during the ultrafiltration and subsequent inclusion in the cheese curd. The time-temperature conditions and presence of air could have affected the plasmin system, i.e. the inactivation of plasminogen activators (Benfeldt, C., 2006). During microfiltration and diafiltration, most plasmin and plasminogen remain in the retentate and exhibit an increased activity since whey proteins and inhibitors are removed by the microfiltration (Aaltonen & Ollikainen, 2011). We found similar trends in a pilot scale extended diafiltration (10 diafiltration steps, 20°C, pasteurised skim milk).

Plasmin activity varied during the diafiltration process, which could mostly be due to varying casein concentrations in the casein retentate (Figure 8A). After correction for casein concentration, plasmin activity increased by 35-37 % after five diafiltration steps. More than 95 % of whey proteins were removed from the casein retentate after four diafiltration steps. The plasmin activity in the permeate was much lower compared to the retentate and stayed constant during the filtration. The recovery rate in the permeate on the other hand decreased with increasing whey protein concentration, indicating that small amounts of plasmin dissociated form casein micelles and permeated. The recovery rate in the casein retentate increased rapidly (Figure 8B) until the whey proteins were removed. The further increase in recovery rate could be attributed to a slower permeation of plasmin inhibitors with a higher molecular weight than whey proteins. As reported by Aaltonen & Ollikainen (2011), we also found a linear correlation between whey protein concentration and plasmin activity.

Figure 8: Effect of diafiltration of pasteurized milk on plasmin activity (A) and recovery rate of plasmin activity (B) in the casein/retentate (□) and serum whey/permeate fraction (●). Bars indicate standard deviation, n = 3.

5 Measurement of the plasmin system activity

There are several possibilities to measure enzymes in milk. Enzyme concentration is usually measured by immunoassays such as enzyme-linked immune sorbent assay (ELISA), which are (in some cases) resistant to inhibitory and interfering compounds. A second possibility are activity measurements which aim to measure the effective enzyme activity in samples and take inhibitory effects into account. A third possibility is the measurement of products produced by the enzyme (Kelly & Fox, 2006). The most optimal method depends on the aim of the measurement. ELISA gives the concentration of the enzyme, but might not differentiate between active and inactive enzymes. Activity measurements usually use synthetic substrates and are largely affected by the buffers in use and the sample preparation. Analysis of enzyme reaction products might give the most unadulterated view of the effect of enzyme activity in a product, but the analysis of these products can be complex and, in the case of low enzyme activity, very time consuming. In the following, different ways of measuring the plasmin system will be presented and discussed.

5.1 Activity assays

5.1.1 Plasmin and plasminogen

Activity assays for plasmin use a synthetic substrate, i.e. small peptides containing a plasmin sensitive bond. Upon cleavage, a fluorescent or coloured product is released, which then can be quantified by measuring fluorescence or light absorbance (Figure 9). Synthetic substrates for plasmin are N-Suc-Ala-Phe-Lys-7-amido-4-methyl coumarin (fluorescence), D-Val-Leu-Lys-p- nitroanilide (S-2251, colorimetric) and HD-Norleucyl-hexahydrotyrosyl-lysine-p-nitroanilide (Spectrozyme-PL, colorimetric).

Figure 9: Schematic representation of the release of yellow p-nitroalinine by cleavage of S- 2251 by plasmin.

Plasminogen (derived) activity is measured as plasmin activity after activation of plasminogen. The difference between plasmin and plasmin plus plasminogen derived activity is used to calculate plasminogen activity. Besides the different substrates, a variety of buffers and sample preparations are used, which can affect the outcome of the measurement. It is obvious that there are difficulties in comparing results obtained by different assays and, maybe even more important, to infer to the characteristics of the plasmin system in milk from these measurements. This problem was already addressed by Kelly et al. (2006), who recommended a standardisation and comparison of plasmin activity assays. A considerable part of this PhD study dealt with the effect of buffer, sample composition and preparation on the outcome of plasmin activity measurements. Some of the key results were published in Paper I.

5.1.1.1 Effect of sample preparation on plasmin activity Spectrophotometric and fluorescence analyses generally require a clear sample with as little scattering and quenching as possible. In milk, the main components that interfere with these measurements are casein micelles and fat. Accordingly, most described assays use sodium citrate or ethylenediaminetetraacetic acid (EDTA) to dissociate the casein micelles by complexion of calcium and calcium phosphate on skimmed samples. These additions do not dissociate the casein micelles completely, but results in small particles with sizes between 15 and 50 nm (McMahon & Oommen, 2013). Plasmin and plasminogen are most likely still bound to these particles, since the pH values of sodium citrate and EDTA solutions are usually above pH 4.6. In a preliminary study in the present work, no significant differences between the treatment with ε-amino caproic acid (EACA) or sodium citrate on plasmin activity were observed (results not shown), concluding that both reagents can be used. A different approach is the incubation of the sample with substrate and the use of Clarifying Reagent to remove the turbidity before measurement (Saint-Denis, et al., 2001b). Politis et al. (1993) used EACA to release plasmin and plasminogen from the casein micelles and removed them by ultracentrifugation, resulting in a clear sample containing only plasmin and plasminogen. In order to avoid turbidity caused by fat, plasmin analysis is either carried out on skim milk samples (Rollema, et al., 1983) or includes a centrifugation step to remove the fat (Richardson, B. C. & Pearce, 1981). Using the assay described in Paper I, we were able to measure plasmin activity without removal of the fat phase. This was possible by correcting for turbidity using a second wavelength that represented the background (Prado, et al., 2006). The results showed that removal of the fat did not affect plasmin activity in unhomogenised raw milk, but the plasmin

24 activity decreased, when the fat was removed from pasteurized, homogenized milk. Upon homogenisation, the milk fat globule membrane (MFGM) becomes to consist to a considerable degree of caseins and whey proteins (Cano-Ruiz & Richter, 1997; Kessler, 2002). Removal of the cream phase thus results in the removal of plasmin associated with caseins on the milk fat globules. The same effect was observed when the surfactant Tween was added prior to the removal of fat of homogenized milk (Iucci, et al., 2008). As a result, the decrease in plasmin activity during high pressure homogenisation was found to be an artefact of the sample preparation rather than a real inactivation of plasmin (Hayes, M. G. & Kelly, 2003; Iucci, et al., 2008). This shows how important the consideration of interaction of plasmin, milk components and milk processing are for the design of an appropriate plasmin activity assay. In milk samples subjected to extreme heat treatment at 90 °C for 5 -30 min, large protein aggregates were formed and interfered with the assay presented in Paper I. These aggregates sedimented during the performance of the assay within 2 h and due to the correction for turbidity, an apparent increase in absorbance was observed in these samples. Although no plasmin activity is to be expected in these samples, the use of Tween and centrifugation of the sample would be a more appropriate sample preparation. At the time of the assay development, we were unfortunately not aware of the work of Iucci et al. (2008).

5.1.1.2 Effect of sample composition and buffers on plasmin activity In addition to the previously discussed inhibition of plasmin by whey proteins and plasmin inhibitors, the plasmin activity in an assay is also affected by the caseins. Caseins are the natural substrate for plasmin in milk and thus compete with the synthetic substrate for the active site of plasmin. They have been shown to act as competitive inhibitors against the synthetic substrate, which can be overcome by increasing the substrate concentration (Bastian, et al., 1991). The state of caseins (monomeric, small aggregates or intact micelles) might affect the interfering effect due to the differences in their mobility, diffusion rate and accessibility. Saint-Denis et al. (2001a) conclude that the main inhibitory components in the plasmin assay are the caseins. This is based on a linear decrease of plasmin activity in a dilution of pasteurized milk (containing native whey proteins and inhibitors) with UHT milk (denatured whey proteins, no inhibitors) compared to a non-linear decrease of plasmin activity, when pasteurized milk is diluted with buffer, similar to the findings reported in Paper I. Our results in Paper I suggest that the whey proteins are the main inhibitory components and that the difference in inhibition by native and denatured whey protein might not be significant, when analysing milk samples. Additionally, both native and denatured whey proteins have been shown to inhibit plasmin activity (Bastian, et al., 1993; Politis, et al., 1993). This together with our results

More recent studies indicate that the inhibitory effect of whey proteins on plasmin activity is affected by the choice of substrate, as some substrates may interact with the whey proteins (Rippel, et al., 2004). The substrate, S-2251, is shown to be able to bind to whey proteins by hydrophobic interactions, while another substrate for plasmin activity, Spectrozyme-PL, is not affected by whey proteins due to its more hydrophilic nature (Rippel, et al., 2004). Using the fluorescent substrate N-Suc-Ala-Phe-Lys-7-amido-4-methyl coumarin, the whey proteins have either an inhibitory or promoting effect at pH 6.5 or pH 5.2, respectively (Hayes, M. G., et al., 2002). These studies were carried out in model systems with plasmin and isolated whey proteins. For the milk and casein system used in our study, it could be expected that the casein particles, which have considerable hydrophobic domains, would be able to adsorb the substrate S-2251 due to their higher concentration. However, the dilution of a casein solution with buffer in Paper I did not show a significant concentration dependent decrease in plasmin activity, suggesting that the inhibitory effects measured in our study reflect real inhibition rather than substrate binding of whey proteins.

Figure 10: Measured activity of added plasmin to skimmed raw milk (○), pasteurized skim milk (●) and UHT milk (□). Activity of added plasmin was measured in the assay buffer described in Paper I containing gelatine to stabilize plasmin. Error bars indicate standard deviation, n = 3.

Apart from these factors, the buffers used for plasmin analysis have a large impact on the results. Buffers for plasmin analysis often contain NaCl and lysine derivatives like EACA to dissociate plasmin from the caseins and to enhance plasmin activity (Bastian, et al., 1991). We showed in 26

Paper I that dissociation of plasmin from caseins resulted in a decreased interference from the caseins in the assay, but on the other hand increased the inhibition of plasmin by inhibitory components. Since EACA reduces the inhibition of plasmin by α2-antiplasmin, the major inhibitory components were most likely the whey proteins. An inhibitory effect of plasmin inhibitors cannot be completely excluded, as can be seen in Figure 10. Addition of plasmin to different milk samples showed that for raw and pasteurized milk, a certain amount of plasmin activity disappeared, when added in low concentrations, while a constant inhibition/interference was found for UHT milk (Figure 10). This disappearance could be caused by inhibition/inactivation of added plasmin by plasmin inhibitors.

5.1.1.3 Comparison of activity assays Table 3 gives an overview of plasmin and plasminogen derived activity assays described and used in literature. The early plasmin assay by Richardson & Pierce (1981) and by Rollema et al. (1983) uses a simple sample preparation, i.e. dissociation of caseins and addition of substrate in a buffer. The methods differ only by the used substrate, N-Suc-Ala-Phe-Lys-7-amido-4-methyl coumarin and S-2251 in the assays described by Richardson & Pearce (1981) and Rollema et al. (1983), respectively, and dissociation of plasmin from caseins EACA in the assay used by Rollema et al (1983). The fluorescence assay by Richardson & Pearce (1981) shows a higher sensitivity and better level of detection compared to other colorimetric assays. Fluorescence assays, on the other hand, require a more careful sample handling (i.e. light exposure, differences in pH and turbidity). Both assay types require a skimming of the sample and are affected by inhibitors of plasmin and interference from caseins.

Table 3: Overview of different plasmin and plasminogen derived activity assays described in literature. If not indicated otherwise, the sample preparation refers to milk samples.

Analysis Measurement Buffer Sample preparation Reference

Plasmin, Fluorescence 0.4 M sodium  Dissociation of caseins by sodium citrate Richardson plasminogen Ex. 380 nm, citrate  Centrifugation/skimming to obtain clear & Pearce Em. 460 nm sample (1981)  Mixing with buffer and addition of substrate

Plasmin, Spectrometric, 40 mM Tris-HCl,  Addition of skimmed milk to buffer Rollema et plasminogen Abs. 405 nm pH 7.4, 100 mM containing substrate al. (1983) EDTA, 50 mM EACA

Plasmin Spectrometric, 50 mM Tris, 110  Model study Bastian et Abs. 405 nm mM NaCl, 3 mM  Mixing plasmin with buffer and reagents al. (1991) EACA, pH 7.4

Plasmin plus Spectrometric, 50 mM Tris-HCl,  Centrifugation to obtain casein/plasmin Politis et plasminogen Abs. 405 nm pH 8, 110 mM pellet al. (1993) NaCl, 50 mM  Resuspending pellet in extraction buffer EACA  Centrifugation to remove caseins  Mixing with substrate buffer

Plasmin, Fluorescence, 100 mM Tris-  Mixing of sample with extraction buffer Saint- plasminogen Ex. 370 nm, HCl, pH 8, 0.4 M (EACA and NaCl) Denis et al. Em. 440 nm NaCl, 8 mM  Mixing of sample with substrate buffer (2001a) EACA  Removal of turbidity and stopping of enzymatic reaction by Clarifying Reagent  No information on fat content of milk

Plasmin Spectrometric, 50 mM Tris, 0.1  Model study Prado et al. Abs. 405 nm M NaCl, Tween  Mixing plasmin with buffer and reagents (2006) Abs. 490 nm 80, pH 7.6  Correction for turbidity

Plasmin Fluorescence 0.4 M sodium  Dissociation of caseins by sodium citrate Iucci et al Ex. 380 nm, citrate  Centrifugation/skimming to obtain clear (2008) Em. 460 nm Tween sample in presence of Tween

Plasmin, Spectrometric, 0.4 M sodium  Dissociation of caseins by sodium citrate Paper I plasminogen Abs. 405 nm citrate, pH 9; 100  Dissociation of caseins and plasmin by Abs. 490 nm mM Tris-HCl, pH EACA and NaCl 8, 0.4 M NaCl, 8  Addition of substrate mM EACA

Politis et al. (1993) proposed an assay in which both inhibition and interference are eliminated. The assay includes a first ultracentrifugation step to remove serum based inhibitory components. The casein pellet is treated with EACA and NaCl to release plasmin, which is separated from the caseins in a second ultracentrifugation step before plasmin activity is measured. By this, plasmin and plasminogen derived activity can be measured without the interference and inhibition of milk components. The assay is very time consuming and it is questionable, if the assay is applicable for different sample matrices and processed products. Furthermore, the number of sample preparation steps will introduce sources of variation. The highest sensitivity and lowest level of detection is achieved by Saint-Denis et al. (2001) using high substrate and sample concentrations. Plasmin is dissociated from the casein micelles and incubated with the substrate. Turbidity is removed by addition of Clarifying reagent. The drawback of this assay is the point wise measurement compared to a continuous measurement used in the assays previously described. The increase in fluorescence is measured at three time points, which increases the number of measurements, time intensity and a poorer precision of the increase in fluorescence. Saint-Denis et al. (2001) also propose a correction factor for the interference of caseins referring to apparent and real activity. Since the whey proteins also do have an impact on plasmin activity, this recovery rate correction includes not only the interference by caseins, but also inhibitory effects. Since the inhibition and interference cannot be distinguished, the term recovery rate may be more correct than real activity. The assay used by Prado et al. (2006) and Rippel et al. (2004) is used in model studies, but includes a second wavelength measurement to correct for turbidity, which increases the sensitivity and LOD of colorimetric assays compared to Rollema et al. (1983). Iucci et al. (2008) use Tween to recover plasmin associated to fat globules in homogenized milk samples, which enables the measurement of plasmin activity in homogenized samples. The assay presented in Paper I allows for measurement of plasmin activity in turbid samples and does not require a centrifugation step. This enables analysis of plasmin activity in different dairy products and processed milk samples. All the described assays measure plasmin activity, but the term itself is ill defined. It is used to describe the concentration of active enzyme in a sample (Politis, et al., 1993), the activity corrected for interfering effects of caseins (Bastian, et al., 1991; Saint-Denis, et al., 2001b), or the activity of plasmin including interfering and inhibitory effects (Richardson, B. C. & Pearce, 1981; Rollema, et al., 1983). Due to the complexity of the plasmin system and the changes during activity assays (dissociation of casein micelle, release of plasmin, removal of components), conclusions on plasmin in milk based on activity measurement should be considered carefully.

5.1.2 Plasminogen activators

Plasminogen activator assays are based on the addition of plasminogen and the resulting increase in plasmin activity due to activation of plasmin by plasminogen activators. Both fibrin and synthetic substrates for plasmin have been used in literature to measure plasminogen activator activity (Korycha-Dahl, et al., 1983; Lu, D. D. & Nielsen, 1993c). Most assays use plasminogen activator fractions isolated from milk. This has the advantage that a differentiation between uPA and tPA is possible. uPA, but not tPA is inhibited by amiloride, while tPA activity on the other hand is extremely low in the absence of fibrin or presence of EACA (Lu, D. D. & Nielsen, 1993c; Vassalli & Belin, 1987; White, et al., 1995). Saint-Denis et al. (2001) describe an assay to measure plasminogen activator activity directly in milk, but no differentiation between uPA and tPA is possible. Compared to the analysis of extracts, the assay by Saint-Denis et al. (2001) is closer to the actual situation in milk. Addition of amiloride to the assay could allow for a distinction between tPA and uPA. It is however unclear, if the caseins could enhance tPA sufficiently in milk to measure tPA activity (Prado, et al., 2007). Model studies with human plasminogen showed that synthetic substrates containing lysine residues can promote plasminogen activation by uPA. At the same time, the substrate can bind to lysine binding sites on plasminogen, resulting in a conformational change in plasminogen and releasing plasminogen from a protein matrix. This would greatly decrease activation of plasminogen by tPA, (Kolev, et al., 1995). Plasminogen of various origins (human, bovine or ovine) is used for plasminogen activator assays, making a comparison between different assays almost impossible. Additionally, the available plasminogen preparations are composed of mixtures of different forms and purities (plasminogen with or without preactivation peptide, preparations containing EACA or lysine). The heterogeneity of bovine plasminogen (midi or full) and the potential effect on activation kinetics of bovine plasminogen has not been investigated yet and should be subject of future studies.

5.2 Immunoassays

ELISA with mono- and polyclonal antibodies has been used to quantify plasminogen and plasmin (Baer, et al., 1994; Benfeldt, Connie, et al., 1995; Dupont, et al., 1997; Haissat, et al., 1994; Politis, et al., 1992). Differentiation between plasmin and plasminogen was achieved by using an antibody specific for plasminogen and plasmin and an antibody specific for plasminogen only, although a complete quantification was not possible (Dupont, et al., 1997). Benfeldt et al. (1995) report a poor response of midi plasminogen using ELISA.

ELISA and western blotting have also successfully been used for the identification of plasmin inhibitors in milk (Christensen, S., et al., 1995; Precetti, et al., 1997).

5.3 Other assays

A special form of activity assay used a ferrocenyl peptide with a plasmin sensitive bond bound to a gold electrode (Ohtsuka, et al., 2009). Cleavage of the peptide by plasmin resulted in a redox signal, which could enable the direct measurement of plasmin in a sample. Fourier-transform infrared spectroscopy has been used to differentiate plasmin and plasminogen (Ozen, et al., 2003). Subtraction of caseinate and whey protein spectra coupled with multivariate data analysis allows for quantification of both plasmin and plasminogen in the presence of these proteins. The lowest concentration of the enzymes in the calibration range is 1000-fold higher than the natural concentration in milk, limiting its application for dairy products.

5.4 Proteolysis

Plasmin activity can also be measured by its action on the caseins. Measurement of proteolysis gives a real measurement of plasmin activity in milk, but underlying mechanisms such as activation of plasminogen and inhibitory effects can often not be differentiated. Furthermore, generation of proteolysis products is also influenced by substrate folding, accessibility and substrate concentration.

5.4.1 Hydrolysis of caseins

Plasmin activity is usually measured by the hydrolysis of β-casein. Proteolysis can be measured by different techniques. The most commonly used techniques include gel electrophoresis and liquid chromatography. Gel electrophoresis has the advantage that the caseins and its larger cleavage products are separated very well, but it is mainly used for qualitative purposes. Compared to liquid chromatography, the analysis is more time intensive and shows a lower dynamic range (Chove, et al., 2011). Reverse phase HPLC coupled with MS was used for the analysis of casein hydrolysis in Paper II. Figure 11 shows a comparison of UHT milk protein composition after 0 and 12 weeks of storage from the present study.

Figure 11: Reverse phase HPLC chromatogram of UHT milk with plasmin activity after 0 (bottom) and 12 weeks (top) of storage. Caseins (cn) and plasmin hydrolysis products are indicated. A detailed description for the HPLC method is given in Paper II.

After 12 weeks of storage, almost no residual intact β- and αS1-casein was present in the UHT milk with plasmin activity. Quantification of casein hydrolysis was only possible for β-casein A1 and αS1-casein 9P due to substantial overlap of the proteolysis products with the other variants and αS2-casein; a complication also reported by Gazi et al. (2014). The use of the MS signal for the quantification of intact casein is impaired by day-to-day variation of the signal intensity, 32 limited linearity of the signal, potential ion suppression of overlaying peaks and casein modification during storage (Paper IV). Capillary electrophoresis has been used for the analysis of caseins, but has the same limitation in terms of overlay of eluting proteins (Recio, et al., 1996).

5.4.2 Peptide formation

Instead of measuring the hydrolysis of intact caseins, proteolysis products can also be used to measure plasmin activity. Measurement of peptide formation can be divided into unspecific and specific methods. Unspecific methods include e.g. measurement of free amino terminals, like fluorescamine assay or the OPA (o-phthaldialdehyde) method (Nielsen, P. M., et al., 2001), and the measurement of pH 4.6 or TCA soluble material. Specific methods determine the concentration of plasmin specific proteolysis products. Unspecific methods and hydrolysis of caseins can be used for analysis of plasmin activity, when the presence of other proteases can be ruled out. In many dairy products with active plasmin, other proteases are likely to be present as well. In cheese, chymosin and peptidases from starter cultures are present, while in UHT milk proteases from psychotrophic bacteria can be present. In these cases, plasmin specific peptides should be chosen to evaluate plasmin activity.

Figure 12: Reverse phase HPLC chromatogram of UHT milk with plasmin activity after 0 (bottom) and 12 (top) wk of storage. Some plasmin specific proteolysis products are indicated. A detailed description for the HPLC method is given in Paper II. Most plasmin proteolysis products are precipitated in trichloroacetic acid (TCA) extracts, while they are soluble at pH 4.6 (Datta & Deeth, 2003). The main plasmin hydrolysis products found in pH 4.6 soluble extracts are the proteose peptones (Figure 12). In contrast, γ-caseins associate with intact caseins and precipitate at pH 4.6. Quantification of the large proteose peptones β-casein f(1-105) and f(1-107) is difficult due to their similar retention time and overlay of different

33 genetic variants. Additionally, these proteose peptones are further hydrolysed by plasmin. The β- casein f(1-28) is formed fast and does not show genetic variation compared to β-casein f(1-105) and f(1-107). The presence of β-casein f(1-25) in Paper III indicates that β-casein f(1-28) can be further hydrolysed, although at a relatively low rate. Recent studies suggest that αS2-casein f(1- 21) and f(1-24) are suitable peptides to evaluate plasmin activity in milk (Cattaneo, et al., 2014).

Both peptides are formed fast by plasmin, but αS2-casein f(1-24) is rapidly hydrolysed to f(1-21)

(Paper III). Thus, αS2-casein f(1-21) is maybe the most suitable peptide for the evaluation of plasmin activity in milk, since it is formed quickly and is not hydrolysed further by plasmin.

5.4.3 Kinetic aspects of proteolysis during storage

Measurement of plasmin activity in an assay is relatively easy. Simple hydrolysis usually follows first order reaction kinetics and the substrate is present in a high concentration, which ensures that only the linear part of the kinetic is measured during the assay. The assay time is relatively short (from 10 min to 3 h); hence, changes in activity due to activation of plasminogen are insignificant. When following plasmin activity and proteolysis during storage of UHT milk activation of plasminogen, substrate affinity and substrate depletion need to be taken into account for a kinetic evaluation. It was initially aimed for that the UHT milks used for Paper II-IV would contain four different plasmin activity levels, which would enable a comparison and correlation of plasmin activity measured by the plasmin activity assay of Paper I with proteolysis. A combination of the two analytical methods would allow for a reliable prediction of plasmin activity and proteolysis in UHT milk during storage. Unfortunately, the experiment resulted in two UHT milks without measurable plasmin activity and two UHT milks with identical plasmin activities. Kinetic modelling of proteolysis and plasmin activity in milk with different plasmin activities was therefore not possible. The UHT milk with plasmin activity used in Paper II and III permitted a basic comparison and modelling of proteolysis and plasmin activity in this specific milk.

Figure 13: Plasmin (●) and plasminogen derived (○) activity during storage of Milk72 from Paper II. Lines represent first order kinetic fits. Error bars indicate standard deviation, n = 3. As discussed in Paper II, plasmin activity increased during storage while plasminogen derived activity decreased. Both the increase in plasmin and decrease of plasminogen could be fitted well to a first order kinetic model (Figure 13). The kinetic parameters for plasmin increase and plasminogen decrease were not identical. The rate of plasminogen decrease was higher than the increase in plasmin activity. As can be seen from Figure 13, all plasminogen was activated after approximately six weeks of storage. The kinetic fit for the increase in plasmin activity therefore most likely also included changes in the inhibition/interference factors in the assay, such as a decreased level of intact caseins.

The decrease in β-casein (Paper II) and increase in β- and αS2-casein N-terminal peptides (Paper III) showed a distinct sigmoidal shape. Figure 14 reveals the decrease of intact β-casein and the development of β-casein f(1-28) during storage.

Figure 14: Development of β-casein f(1-28) (A) and decrease of intact β-casein (B) during storage. Lines represent kinetic fits using a polynomial and first order reaction kinetic.

The sigmoidal shape originates from the activation of plasminogen and thus increased rate of proteolysis in the initial part of the storage period and a decrease in the rate of proteolysis during the final part of the storage period due to substrate depletion. The kinetic of peptide formation or hydrolysis of caseins should therefore include an additional factor to represent the change in enzyme concentration. The activation of plasminogen follows a first order reaction kinetic. A normal first order reaction kinetic for product formation is:

(1) In the case of plasminogen activation, the rate constant k itself changes in a first order reaction as well (Figure 14) and could be described as:

(2) kmax represents the maximum velocity of the reaction, similar to Cmax in (1), and is dependent on the rate of plasminogen activation depicted as kact. This results in complex reaction kinetics, which are almost impossible to model due to the presence of two exponential functions. Plasminogen activator assays use high amounts of added plasminogen, which allows a kinetic interpretation of plasminogen activation using a polynomial expression rather than a first order kinetic as indicated in Figure 15: (3) The factor a in equation (3) represents the change in reaction rate due to the activation of plasminogen, while factor b represent the initial plasmin activity present in the sample. The

36 proteolysis could therefore be described by using equation (3) for the initial proteolysis and equation (1) for the remaining storage time (Figure 14). A different approach is to assume a linear kinetic for proteolysis in combination with a first order kinetic to include the increase in plasmin activity: (4) Both peptide formation and casein hydrolysis can be described by equation (4) by changing the prefix of k. Applying this fit allows an accurate description of the reaction until the substrate depletion phenomenon appears (Figure 15).

Figure 15: Development of β-casein f(1-28) (A) and decrease of intact β-casein (B) during storage. Lines represent kinetic fits using equation (4).

This approach might be more suitable for prediction purposes since the quality of the milk would be unacceptable when substrate depletion occurs. Bitterness and gelation occurred after 70 % and > 90 % of β-casein, respectively, was hydrolysed (Paper II). When sufficient kinetic data is provided, a correlation of the plasmin and plasminogen derived activity measured by the activity assay with the kinetic of proteolysis could be established, allowing the prediction of plasmin activity and proteolysis in UHT milk.

The kinetics of αS1-casein hydrolysis and the αS1-casein N-terminal peptide f(1-22) in Paper II and III showed a different kinetic behaviour compared to β-casein and β-casein f(1-28) respectively. The decrease in intact αS1-casein in the UHT milk showed an extended “lag” phase of approximately two weeks. This could have been caused by a reduced affinity and accessibility of αS1-casein for plasmin. A similar reduced affinity was seen for αS1-casein f(1-22), but in comparison to αS1-casein f(106-124), the lag phase was identical even though αS1-casein f(106- 124) requires two cleavages by plasmin to be formed. This suggests that plasmin has a reduced affinity for the cleavage at Arg22-Phe23 and a different peptide might be more suitable for the

37 modelling of the proteolysis. The C-terminal peptide αS1-casein f(194-199) was formed slowly and was only present in low concentrations and could not be used for modelling either. The peptide αS1-casein f(106-124) would thus be most suitable to be used for modelling, but the precursor formation would have to be included into the kinetics, resulting in one or two additional variables, hence making the modelling complex.

The decrease in intact αS2-casein could not be assessed due to overlap with proteose peptones eluting at the same time (Paper II) and a quantitative assessment using the MS signal is impaired by the heterogeneity in phosphorylations of αS2-casein. As discussed before, the N-terminal peptides of αS2-casein formed by plasmin hydrolysis are present in high concentrations and αS2- casein f(1-21) can be used for modelling.

6 UHT milk

6.1 Heat treatment of milk

The main aim of heat treatment of milk is to partially or fully inactivate microorganisms and enzymes to guarantee consumer safety and prolong the shelf life of milk. Increasing the time and/or temperature of the heat treatment results in increased inactivation of microorganisms and enzymes, but can negatively affect other milk constituents. Heat treatment of milk is therefore often a balance between inactivation of undesired components and preservation of desired properties. Heat treatments can be categorized by their time temperature combinations, which are defined by norms and legislation. Pasteurisation can either be high temperature, short time (HTST) with heating at 72-75 °C for 15-30 s, or low temperature, long time as e.g. thermisation at 62-65 °C for 30 min. Pasteurisation results in the inactivation of pathogen microorganisms and gives a negative result when tested for alkaline phosphatase. Any other time temperature combination resulting in the same effect can be applied as well. Milk subjected to high temperature pasteurisation is heat treated at temperatures > 85 °C and should give lactoperoxidase negative test results (Kessler, 2002). Ultra high temperature (UHT) treatment is defined as high temperature (>135 °C) for a short time, usually 135-150 °C for 1-10 s. The aim of UHT treatment is to achieve commercial sterility, i.e. free from microorganisms that can grow under storage conditions (EC, 2005). The effect of UHT treatment on relevant milk constituents will be discussed later in this section.

6.2 UHT milk processing

6.2.1 Design of UHT processes

The inactivation rate of microorganisms at a given temperature is usually measured by the decimal reduction time, D, which is the time necessary to reduce the number of microorganisms by a factor of 10. The temperature effect of the inactivation of microorganisms can be expressed by the z value and the Q10 value. The z value is defined as the change in temperature necessary to increase D by a factor of 10. The Q10 value on the other hand is defined as the change in the inactivation rate by a temperature increase of 10 °C (Kessler, 2002). Chemical effects occurring during heat treatment can be described using similar parameters as described above. Based on these relationships, lines of equal effects, i.e. time temperature combinations resulting in the same inactivation of microorganisms or chemical changes can be defined to design UHT processes.

Kessler & Horak (1981) introduced the biological effect B* and the chemical effect C* to determine time temperature regions for UHT milk processing. A B* value of one equals the 9 decimal reduction of natural thermophilic spores, which results in a commercial sterile product (Kessler & Horak, 1981; Kessler, 2002). The negative effect of heat treatment on milk constituents can be characterized by the chemical effect or C* value. A C* value of one equals the chemical effect of boiling the product for 1 min. Similar to the B* value, a C* value can be used to define the acceptable limit of chemical changes. In milk, this limit correlates with a loss of 3 % thiamine and corresponds to a C* value of one (Kessler, 2002). Based on these two factors, the optimal time temperature conditions for UHT processes are in an area where B* > 1 and C* < 1 (Figure 16). In the last decade, UHT processes with ultra-short holding times have been developed that are beyond these time temperature regimes concerning the holding time. It has been shown that UHT treatment > 150 °C with holding times of 0.2 s result in a sufficient spore reduction to obtain a commercially sterile product (Huijs, et al., 2004; van Asselt, et al., 2008). The UHT process used in Paper II-IV (> 150 °C, < 0.2 s) has been validated by Arla on full scale and gave a commercially sterile product.

Figure 16: Bacteriological B* and chemical C* effect and UHT processing area. Adapted from Kessler (2002).

6.2.2 UHT processing systems

UHT processing systems can be divided into direct and indirect systems. In indirect systems, the product is heated by counter flow of cold milk and hot water or steam in a tubular or plate heat exchanger. In direct systems, the product is heated by directly mixing it with steam under pressure, which is later removed using vacuum cooling (Burton, 1994a; Kessler, 2002). A comparison of a direct, indirect and the direct UHT process used in the trials described in Paper II-IV is shown in Figure 17.

Figure 17: Schematic representation of time temperature profiles of indirect (dotted line), direct (dashed line) and direct/ultra-short time (solid) UHT processes. Holding time prior (indirect) and after (direct) UHT treatment represent the homogenisation step.

As can be seen from Figure 17, a large advantage of direct systems is the very fast heating rate, which is only a fraction of a second and that results in minimal chemical changes to the product compared to indirect systems. One of the disadvantages of using steam to directly heat the product are that high quality steam is required, and so far the energy recovery in the process is poor. The product is therefore usually heated to 60-80 °C by indirect heat exchangers. While in indirect systems, the energy recovery is approximately 90 %, it is only 45 % when using direct systems, making direct UHT systems costlier to operate (Burton, 1994a; Kessler, 2002; Malmgren, 2007). The homogenisation step can be placed before or after the UHT treatment in indirect systems, while it is always placed after the UHT treatment in direct systems to avoid the formation of aggregates (Burton, 1994a; Datta, et al., 2002).

Direct UHT systems can be further distinguished into steam injection and steam infusion systems (Figure 18). In steam injection systems, the steam is “injected” as small bubbles into the product. In steam infusion systems, the product falls as a thin stream or film through a steam pressured chamber. Steam infusion is generally considered to be gentler than steam injection. During steam injection, steam bubbles condensate and collapse leading to cavitation and a homogenisation effect, whereas the steam condensates on the surface of the falling product in steam infusion systems. A homogenisation effect can be observed in direct steam infusion systems as well, most likely caused by cavitation occurring during the vacuum cooling (Burton, 1994a; Datta, et al., 2002; Hougaard, et al., 2009). Fouling is reduced in steam infusions systems since in contrast to steam injection systems the product does not come into contact with surfaces hotter than itself (Datta, et al., 2002).

Figure 18: Tetra Pak steam injector nozzle (left) and APV steam infusion chamber (right). Pictures from www.tetrapak.com and www.apv.com.

The UHT trials described in Paper II-IV were performed in the UHT pilot plant (APV/SPX, Silkeborg, Denmark) at the Arla Strategic Innovation Center in Stockholm. The pilot plant (Figure 19) is able to run indirect, steam infusion and steam injection UHT processes with a capacity of 150 l/h. Both tubular and plate heat exchangers can be used to heat or cool the product. Holding times can be controlled by an external holding cell (15-300 s). The pilot plant is equipped with a two-stage homogenizer with a maximum pressure of 40 MPa. In the direct UHT process, the temperature for the flash vacuum cooling was set to 8 °C lower than the pre-heat temperature to obtain the same dry matter content. Bottling of the milk samples was done in a sterile bench and every batch was checked for sterility according to (ISO-4833, 2003).

Figure 19: The UHT pilot plant at the Arla Strategic Innovation Center Stockholm.

6.3 Heat-induced changes in milk

6.3.1 Proteins

6.3.1.1 Whey proteins The whey proteins are a heterogeneous group of proteins that are found in the serum phase after precipitation of caseins by acidification or renneting. The effect of heat treatment on the two major whey proteins, β-lg and α-la, has been researched intensively in milk and isolated proteins under different conditions. When looking at the denaturation of whey proteins, a distinction between reversible and irreversible denaturation has to be made. Differential scanning calorimetry measurements show that β-lg and α-la unfold and lose their tertiary globular structure at 75-80 °C and 65 °C, respectively (Paulsson, et al., 1985; Tolkach, 2007). This denaturation of the whey proteins is highly reversible. β-lg contains five cysteine residues of which four form disulphide bridges and one free cysteine residue. In its native form, the free cysteine residue is buried in the globular structure of β-lg. Upon unfolding of β-lg at higher temperatures (> 75-80 °C), the free cysteine residue and hydrophobic residues become exposed and are able to react via thiol-thiol, thiol- disulphide and hydrophobic interactions with other proteins; as a result, b-lg denatures irreversibly and forms aggregates (Anema, 2008; Sawyer, 2003). Isolated, α-la is very resistant to irreversible heat induced denaturation (Wang, Q., et al., 2006). However, it contains 4 disulphide

43 bonds and the irreversible denaturation is accelerated in the presence of β-lg (Anema, 2008; Tolkach, 2007). The kinetics of irreversible heat induced denaturation of α-la and β-lg consist of two reactions with different activation energies. The Arrhenius plots for α-la and β-lg show a distinct discontinuity at 80 and 90 °C, respectively. Below 80 and 90 °C, the activation energies are high, indicating that a large number of bonds are broken and the rate limiting factor for irreversible denaturation is the unfolding of the whey proteins. Above 80 and 90 °C, the activation energy is much lower and the whey proteins are completely unfolded and the speed limiting factor is the aggregation of the whey proteins (Anema, 2008; Dannenberg & Kessler, 1988; Tolkach, 2007). Besides the formation of aggregates between whey proteins, β-lg can react with κ-casein on the surface of the casein micelle via thiol-disulphide interchange reactions (Anema, 2007). Irreversibly denatured whey proteins precipitate at pH 4.6 and the degree of denaturation of whey proteins is usually measured as the concentration of pH 4.6 soluble whey proteins before and after heat treatment. While whey protein denaturation is almost negligible during HTST pasteurisation at 72 °C for 15 s, UHT treatments can result in almost complete denaturation of whey proteins. β-lg denaturation is lower in UHT milk processed with direct systems (35-80 %) than indirect systems (79-100 %) due to the faster heating rate in direct systems (Burton, 1994a; Datta, et al., 2002). Association of whey proteins with the casein micelles is lower in direct systems compared to indirect systems. This can also be attributed to the fast heating rate, i.e. short holding time. Basically, all whey proteins are instantly unfolded during direct UHT processing, giving the reactive monomers a greater opportunity to react with themselves than associating with the casein micelles (Anema, 2008). The whey protein denaturation in the UHT milks produced for Paper II-IV is shown in Table 4. At a pre-heat temperature of 72 °C, the direct steam infusion step alone resulted in 34 % β-lg and less than 10 % α-la denaturation. Prolonging the pre-heat treatment to 180 s at 72 °C increased the degree of whey protein denaturation only slightly. However, increasing the pre-heat temperature to 95 °C drastically increased the degree of whey protein denaturation. Analysis of the process showed that the main reason for the increase was caused by the difference in flash cooling temperature (8 °C lower than the pre-heat temperature) and the higher temperature in the following homogenisation step, resulting in an increased heat load (approximately 87 °C for 15- 20 s). In the indirect UHT milk and the direct UHT milk 95°C/180 s + > 150°C /< 0.2 s, β-lg is almost completely denatured (Table 4). There is, however, a significant difference in the α-la denaturation, indicating that the heat damage was still lower in the direct UHT compared to the indirect.

Table 4: Whey protein denaturation in direct (pre-heat treatment + direct steam infusion step) and indirect UHT milk used for Paper II-IV. Whey protein denaturation was measured as the difference in pH 4.6 soluble whey proteins before and after heat treatment. Values for β-lg are an average of β-lg A and B. Different superscripts indicate significant differences (P < 0.05).

Whey protein Heat treatment denaturation [%]

β-lg α-la

72°C/5 s + > 150°C /< 0.2 s 33.9 ± 2.7a 9.4 ± 3.2a 72°C/180 s + > 150°C /< 0.2 s 36.7 ± 3.2a 13.1 ± 3.2a 95°C/5 s + > 150°C /< 0.2 s 86.6 ± 2.1b 26.6 ± 4.0b 95°C/180 s + > 150°C /< 0.2 s 94.1 ± 1.6c 45.5 ± 3.3c Indirect 140 °C/4 s 93.6 ± 1.5c 70.0 ± 6.1d

6.3.1.2 Caseins Compared to the whey proteins, the caseins and casein micelles are very heat stable and the micellar conformation is preserved even during severe heat treatments (Walstra, et al., 2006). During severe heat treatments, polymerisation of caseins can occur, which could be caused by the formation of lysinoalanine, isopeptide formation or the Maillard reaction (Zin El-Din & Aoki, 1993). The most important heat induced change of caseins during heat treatment is an increase in the casein micelle size due to the association of denatured whey proteins (Anema & Li, 2003a; Anema, 2007; Crudden, Oliveira, et al., 2005a). The increase in casein micelle size depends largely on the degree of whey protein denaturation. Only small changes are observed when less than 80 % of the whey proteins are denatured, while a fast and larger increase in micelle size is observed at higher degrees of denaturation (Anema & Li, 2003b).

6.3.2 Heat inactivation of the plasmin system

Inactivation by heat of plasmin and plasminogen in milk has been closely linked to the denaturation of β-lg and thiol-disulphide interchange reactions. Plasmin inactivation in the absence of β-lg is decreased (Rollema & Poll, 1986) and a positive correlation between the concentration of β-lg and the rate of plasmin activation has been found (Saint-Denis, et al., 2001a). Proof of the thiol-disulphide reaction as the main driving force for plasmin inactivation was provided by the increased inactivation of plasmin upon addition of cysteine (Durkee & Hayes, 2008; Lu, R., et al., 2009; Metwalli, et al., 1998; Rollema & Poll, 1986) and decreased plasmin inactivation upon blocking the free thiol group of β-lg by addition of KIO3 (Kelly & Foley, 1997). The heat inactivation kinetics of plasmin and plasminogen in various dairy systems has been investigated by several authors (Alichanidis, et al., 1986; Crudden, Oliveira, et al., 45

2005a, 2005b; Metwalli, et al., 1998; Rollema & Poll, 1986; Saint-Denis, et al., 2001a). A selection of kinetic parameters is summarized in an Arrhenius plot in Figure 20.

Figure 20: Arrhenius plot for the heat inactivation of plasmin in milk (red line) from Saint- Denis et al (2001) and plasmin+plasminogen in milk (green line) and a casein micelle solution (black line) from Rollema et al. (1986). The Arrhenius plot for heat denaturation of β-lg is indicated (dashed line) using the kinetic parameters of Dannenberg & Kessler (1988).

In milk, inactivation of plasmin and plasminogen shows a similar kinetic pattern to that of β-lg denaturation, with a discontinuity at around 95 °C, indicating that the inactivation of plasmin and plasminogen consists of two reactions with different activation energies as well. Older studies report the discontinuity at 77 °C (Driessen & van der Waals, 1978; Metwalli, et al., 1998). The difference in the kinetic reported by Rollema et al. (1986) and Saint-Denis et al. (2001b) above 90 °C could be attributed to the different assays used and the fact that the kinetics of Rollema et al. (1986) are a combined kinetic for plasmin and plasminogen. Plasminogen exhibited a lower heat stability compared to plasmin below 90 °C, but a higher stability above 90 °C is found by Rollema et al. (1986), while Saint-Denis et al. (2001b) show comparable heat stabilities of plasmin and plasminogen. Comparing the rate of β-lg denaturation with plasmin inactivation, it becomes obvious that the rate for plasmin inactivation below 90 °C is much faster compared to the denaturation rate of β-lg. Above 90 °C, the rates are very similar and, depending on the applied inactivation kinetic, β-lg denaturation is actually faster than plasmin inactivation above 120 °C according to Saint-Denis et al. (2001b). As previously discussed, interaction of β-lg with the casein micelle is less pronounced at higher temperatures. Since plasmin and plasminogen are 46 associated with the casein micelle, the same mechanism could apply for the inactivation of plasmin and plasminogen. A reported comparison of plasmin inactivation and β-lg association with the casein micelle suggested that the interaction of plasmin with β-lg and the β-lg association with the casein micelle was faster than aggregation of β-lg in the temperature range between 65-80 °C (Crudden, Oliveira, et al., 2005a). Direct UHT treatment with a pre-heat at 72 °C for 180 s and direct steam infusion at >150 °C for < 0.2 s resulted in an almost equal degree of β-lg denaturation and plasmin activation (Paper II). In comparison to the inactivation kinetic of plasmin in milk, the activation energy for the inactivation of plasmin and plasminogen in a casein micelle solution is constant and the inactivation rate is decreased by a factor of 5-10 in the casein micelle solution (Rollema & Poll, 1986). The same effect has been observed for the denaturation kinetic of α-la in absence of β-lg (Tolkach, 2007; Wang, Q., et al., 2006). Based on the two step reaction kinetic consisting of an unfolding and aggregation reaction, Tolkach (2007) concluded that the unfolding of α-la should still result in a slight discontinuity of the activation energy in absence of β-g. While this is not the case for isolated α-la, Metwalli et al. (1998) found a discontinuity in the inactivation kinetic of isolated plasmin at 77 °C, indicating that the unfolding of plasmin has an effect on the inactivation of plasmin. The difference to the results of Rollema et al. (1986) could be due to the presence of caseins, which affect the inactivation of plasmin as well. Figure 21 gives a good overview of the factors influencing plasmin inactivation in a milk system. Plasmin inactivation in the presence of sodium caseinate is reduced compared to isolated plasmin. Caseins are also able to protect plasmin against inactivation in the presence of β-lg, but not against inactivation by cysteine (Metwalli, et al., 1998). The protective effect of caseins against inactivation of plasmin is not fully understood yet. In absence of free thiol groups, caseins could bind to the lysine binding sites of plasmin, which has been shown to enhance the heat resistance of plasmin against inactivation by EACA (Grufferty & Fox, 1988a). Caseins have also been shown to act as chaperones and could reduce the heat induced denaturation and inactivation or assist in a refolding after heat treatment (He, et al., 2011; Morgan, et al., 2005; Treweek, et al., 2011; Zhang, et al., 2005). This could especially be the case for the sodium caseinate used by Metwalli et al. (1998). Metwalli et al. (1998) on the other hand showed that temperature-activity of plasmin is not affected by caseins, suggesting that caseins do not protect plasmin from unfolding. However, the unfolding and inactivation of plasmin in the absence of free thiol groups does not necessarily include the unfolding of the light chain containing the active centre, but could also involve aggregation or structural changes in the kringle structures.

Figure 21: Effect of two milk proteins on heat-induced inactivation of plasmin in a milk salt solution. (A) Effect of sodium caseinate (+), β-lg (▴ ) and β-lg+caseinate (♦); (B) of sodium caseinate (+), cysteine (□), cysteine+caseinate (○), on inactivation of plasmin (control, ●) at 77 °C. From Metwalli et al. (1998).

In the presence of free thiol groups, a different protective effect of caseins seems to take place. The fact that caseins are able to protect plasmin against inactivation by β-lg, but not by cysteine implies a protection via a steric hindrance of the reaction between β-lg and plasmin by the caseins. In this project, an experiment was planned to study the heat stability of plasmin associated with caseins and plasmin released by NaCl and EACA. Using the protocol by Politis et al. (1993) to dissociate plasmin, we could not detect any significant difference between the inactivation rate of dissociated and associated plasmin at 80 °C in skim milk. Ultracentrifugation of the milk, however, showed that approximately 50 % of plasmin was still attached to the caseins. Due to lack of time, this could not be investigated further. Clarification of the protective effect of caseins on plasmin inactivation could help to develop processes with improved plasmin inactivation. The fat content of milk influenced the heat inactivation of plasmin using lenient steam infusion heat treatment (Dickow, 2011). A possible explanation for this could be the relatively high number of free thiol groups on the MFGM, which could catalyse and enhance thiol-disulphide interchange reactions during heat treatment (Lee & Sherbon, 2002). Plasmin inactivation in whey was found to be very dependent on the type of whey. Plasmin inactivation in acid whey was much slower compared to sweet whey, which could partially be attributed to the difference in pH (Crudden, Oliveira, et al., 2005b). Interestingly, neither the inactivation kinetic of plasmin nor of

β-lg in acid whey showed the characteristic discontinuity found for milk (Crudden, Oliveira, et al., 2005b). Results on the heat inactivation kinetic of plasminogen activators are inconsistent. Overall, it is known that plasminogen activators show slightly higher heat stability compared to plasmin (Lu, D. D. & Nielsen, 1993b; Saint-Denis, et al., 2001a). While Saint-Denis et al. (2001b) find the same kinetic for plasminogen activators compared to plasmin with a discontinuity at 90 °C, whereas Lu & Nielsen (1993) report a single reaction with constant activation energy. While Saint-Denis et al. (2001a) assay the plasminogen activators directly in milk, whereas Lu & Nielsen (1993) use a milk model system containing buffer and casein. The absence of β-lg can explain the difference in the inactivation kinetic. uPA and tPA contain 12 and 17 disulphide bonds respectively, indicating that the inactivation mechanism is most likely similar to plasmin and plasminogen. Heat inactivation of tPA and uPA shows that the latter is more heat stable (Prado, et al., 2007). This could be due to the higher number of disulphide bonds in tPA compared to uPA. Similar to plasmin activity, the term ‘inactivation’ of plasmin system components is ambiguous. Thiol-disulphide interchange reaction of β-lg with the kringle structure of tPA would result in a largely decreased activity since tPA would no longer be able to bind to the caseins. The kringle structure of uPA on the other hand is not important for the binding to caseins and a reaction with β-lg would most likely not affect its activity. An assay using fibrin to enhance tPA activity (Prado, et al., 2007) thus could lead to false assumptions regarding inactivation or reduced activity of tPA. The same applies for the inactivation of plasmin. Disruption of the kringle structures by thiol-disulphide interchange reactions could release plasmin and inhibit its action on caseins, similar to the effect of EACA on plasmin activity towards fibrin. The activity in an assay using a small synthetic substrate would however not be affected.

6.3.3 Effect of cysteine on plasmin and milk proteins during heat treatment

The addition of cysteine to milk has been proposed in several studies to improve the inactivation of plasmin, however these studies do not specify the impact of cysteine addition on flavour and milk proteins (Durkee & Hayes, 2008; Rollema & Poll, 1986). The effect of cysteine or other reducing agents on milk proteins during heat treatment is largely unknown and has only been studied in relation to acid gel formation after severe heating conditions (Goddard, 1996; Nguyen, et al., 2012; Nguyen, et al., 2013). In order to evaluate the practical use of cysteine for improved plasmin inactivation, the potential changes in milk proteins need to be considered. In an initial study, low concentrations of cysteine up to 1.5 µM were added to pasteurized skim milk and heat

49 treated at 80 °C with holding times up to 60 s in a water bath. Plasmin inactivation increased with increasing cysteine concentration as expected (Figure 22A).

Figure 22: Effect of addition of 0 (●), 0.5 (○), 1 (▼) and 1.5 (∆) µM cysteine on the inactivation of plasmin (A), monomeric κ-casein (B) and denaturation of β-lg (C) and α-la (D) during heat treatment at 80 °C. The sample needed approximately 10 s to heat up and the unheated sample is marked as -10 s.

A considerable inactivation of plasmin already occurred during heating up of the sample in the first 10 s. In milk with 0.5 µM cysteine, a shift in reaction kinetic was observed after 30 s, which indicated that the cysteine available for plasmin inactivation had reacted; thereafter inactivation occurred at a similar rate as in the control sample without cysteine. The protein composition was analysed as described in Paper II, with the exception that no reducing agent was added to the urea and sodium citrate buffer. In samples containing cysteine, a peak appeared at the retention time of unreduced κ-casein and was confirmed by MS-analysis as a monomeric κ-casein with a mass shift of -2 Da compared to the reduced κ-casein, indicating an intramolecular disulphide bond. The concentration hereof depended both on the holding time and on the cysteine concentration (Figure 22B). This indicated a decrease of multimeric κ-casein by increased formation of intramolecular disulphide bonds catalysed by cysteine as reported by Nguyen et al (2012). In the sample with 1.5 µM cysteine, the concentration of monomeric κ-casein decreased 50 after 60 s holding time and a different peak appeared which was identified to be κ-casein with two cysteine residues attached, suggesting a time and cysteine concentration dependent formation of monomeric κ-casein and/or monomeric κ-casein with two cysteine residues. Surprisingly, the degree of β-lg denaturation was only slightly affected by the presence of cysteine (Figure 22C). Milk heat treated with higher cysteine concentrations showed slightly increased β-lg denaturation, but no difference could be seen for 0.5 µM cysteine addition compared with the control milk. In contrast to β-lg, the degree of α-la denaturation increased with increasing cysteine addition (Figure 22D). The difference in the denaturation of β-lg and α-la could be related to their structure and number of cysteine residues. The main mechanism for α-la denaturation in milk is the thiol-disulphide interchange of α-la with the free thiol group of β-lg and the presence of additional free thiol groups accelerates the denaturation (Tolkach, 2007). The relatively small change in the denaturation of β-lg could be attributed to the free thiol group of β- lg that might act as a catalyst to regain the native structure by releasing cysteine. Reaction of cysteine with the free thiol group of β-lg resulted in denaturation of β-lg since β-lg with an attached cysteine was present in total protein, but not in the pH 4.6 soluble material. The differently treated milks were not evaluated sensorically, but all milk samples with added cysteine had a distinct sulphuric smell. An immediate repetition of the trial was not possible since the analysis of unreduced heat treated milk samples rendered the HPLC column unusable and no appropriate column was available. The trial showed that the addition of cysteine significantly affects milk proteins. In regards to a practical use of cysteine for the heat inactivation of plasmin in milk products, these effects need to be further studied and characterized. While the modifications and denaturation of whey proteins might not be a crucial factor for UHT milk, the modifications of κ-casein could affect the stability of the casein micelle and result in sedimentation or other quality defects.

6.3.4 Maillard reaction

The Maillard reaction is a chemical reaction of reducing sugars and amino groups. In milk, the Maillard reaction occurs mainly during the heat treatment, but also during subsequent storage, although at a much lower rate (see section 6.4.3). The main reducing sugar in milk is lactose and it reacts mainly with lysine residues on proteins since the level of free amino acids in milk is relatively low (Van Boekel, 1998). The Maillard reaction is sometimes divided into an early, an advanced and a final stage.

Figure 23: Formation of the Amadori product lactulosyllysine by the reaction of a protein bound lysine with lactose. From Siciliano et al. (2013).

The early stage of the Maillard reaction, depicted in Figure 23, is the condensation of the reducing sugar with the amino group and the formation of a Schiff’s base, which following Amadori rearrangement gives the Amadori product lactulosyllysine (Van Boekel, 1998). In the advanced stage of the Maillard reaction, the Amadori product breaks down into numerous products. The final stage consists of the condensation of amino and sugar compounds to polymerized proteins and brown pigments. The Maillard reaction can significantly affect the quality of milk and milk products. Blockade of the lysine residues reduces the nutritional value and digestibility (Dalsgaard, et al., 2007; Van Boekel, 1998). Various flavour compounds and compounds with either beneficial (antioxidative, antimicrobial) or harmful (mutagenic, allergenic) properties can be formed (Siciliano, et al., 2013; Van Boekel, 1998). Colour changes occur when brown pigments are formed. Additionally, the functionality of proteins can be changed by lactosylation, such as solubility and thermal stability (Wang & Ismail, 2012). The heat load applied during UHT processing is usually not sufficient to produce the final Maillard reaction products and colour changes. The main Maillard reaction product found in UHT milk is the Amadori product. Some advanced Maillard reaction products, such as 5-hydroxyl-2-methylfurfural (HMF), are present in UHT milk, but only in a limited amount (Birlouez-Aragon, et al., 1998; Claeys, et al., 2003, 2004).

Direct analysis of the Amadori product is difficult, as it can only be measured after enzymatic breakdown of the lactosylated proteins. The most common way to measure protein lactosylation is to measure furosine, a product that is formed by acid hydrolysis of the Amadori product (Erbersdobler & Somoza, 2007; Van Boekel, 1998). A different approach is the boiling of the lactosylated proteins in oxalic acid to induce the formation of HMF. Both methods have the disadvantage that only a part of the Amadori product is actually converted to furosine and HMF, and HMF itself is already present in milk. In addition, the sample preparation is a very time intensive process. The formation of lactulosyllysine results in a mass change of the unmodified protein of +324 Da and can be measured by mass spectrometry (Figure 23). Several studies investigated the lactosylation of whey proteins and caseins (Czerwenka, et al., 2006; Meltretter, et al., 2009; Scaloni, et al., 2002; Siciliano, et al., 2013). While a qualitative assessment of protein lactosylation can easily be achieved, quantification of the extent of lactosylation using MS can be challenging. The quantification of protein lactosylation will be discussed in detail in section 6.4.3.

6.3.5 Lactulose

Besides whey protein denaturation, furosine and HMF, lactulose is often used as marker for the heat treatment of UHT milk (Datta, et al., 2002). Lactulose is an isomeric form of lactose and is formed during heat treatment. The reaction is catalysed by either amino groups of the milk proteins or by inorganic salts (Burton, 1994b; Datta, et al., 2002). Lactulose is a stable product and not present in raw milk. However, the formation of lactulose is very pH dependent (Burton, 1994b).

6.4 Shelf life development of UHT milk

6.4.1 Age gelation of UHT milk

The term ‘age gelation’ has been used in literature to describe the phenomena of physical instability of sterilized milk. Age gelation observed by different researchers differs greatly in gels appearance and the cause and/or mechanism by which they are formed (Nieuwenhuijse & van Boekel, 2003). The mechanisms of age gelation have been divided into a proteinase-induced, non-enzymatic-induced and a two stage process consisting of a combination of proteolysis followed by physical chemical changes (Aroonkamonsri, 1996; Harwalkar, 1992; Manji & Kakuda, 1988). An overview of causes and mechanisms of age gelation by Nieuwenhuijse & van Boekel (2003) is shown in Figure 24.

Figure 24: Simplified scheme of various pathways of protein destabilisation in milk and milk products. From Nieuwenhuijse & van Boekel, 2003).

Non-enzymatic age gelation in normal UHT milk is extremely slow, while fast in concentrated milks (Manji & Kakuda, 1986; Nieuwenhuijse & van Boekel, 2003). The proposed mechanism for age gelation at medium and high storage temperatures in concentrated UHT milks is the

54 reaction between β-lg and κ-casein during the heat treatment. This complex can dissociate from the casein micelle during subsequent storage and form a gel network (McMahon, 1996). In age gelation of concentrated UHT milk, protein dissociation and casein micelle crosslinking are regarded as factors causing age gelation. The hypothesis that proteolysis merely facilitates the release of β-lg-κ-casein complexes from the casein micelle is untenable (Datta & Deeth, 2001, 2003; Malmgren, 2007; McMahon, 1996). Various studies, including Paper II, show that protease induced age gelation is actually impaired by increasing heat treatment. Furthermore, Paper II showed that age gelation is possible with almost no β-lg associated to κ-casein, findings in accordance to (Kelly & Foley, 1997; Manji & Kakuda, 1988). Protease induced age gelation can be caused both by plasmin and heat stable proteases originating from psychotrophic bacteria. In contrast to plasmin, bacterial proteases are able to hydrolyse κ- casein and can, in a similar fashion as chymosin, result in a rennet-curd like gel (Nieuwenhuijse & van Boekel, 2003; Snoeren, et al., 1979). Recent studies showed that bacterial proteases, besides hydrolysing κ-casein, also are able to destabilize casein micelles by proteolysis in general (Gaucher, et al., 2009; Gaucher, et al., 2011). Bacterial proteases show a more unspecific proteolysis pattern, which usually results in small peptides and can be distinguished from plasmin peptides by different RP-HPLC preparation methods. While both peptides formed by bacterial proteases and plasmin are soluble at pH 4.6, the larger plasmin derived peptides are not soluble in TCA (Datta & Deeth, 2003). The role of plasmin in age gelation has been subject of intensive research, but it has not been unequivocally proven that plasmin can cause age gelation (O’Mahony, et al., 2013). Although plasmin has been shown to be involved in age gelation, the conditions under which gelation occurs vary considerable and so far there is no direct correlation between the onset time of gelation and level of proteolysis (Manji & Kakuda, 1988). As the study described in Paper II shows, age gelation can occur in UHT milk with residual plasmin activity. No age gelation takes place when heat treatment conditions result in complete inactivation of plasmin and plasminogen, or when plasmin activity is inhibited (de Koning, et al., 1985; Kohlmann, et al., 1988; Manji & Kakuda, 1986). Age gelation in relation to plasmin therefore occurs more frequent when using direct UHT systems compared to indirect systems (Datta, et al., 2002; Manji & Kakuda, 1986). A first crucial factor for age gelation caused by plasmin is the rate of proteolysis. A high rate of proteolysis, either by high residual plasmin activity or addition of plasmin, plasminogen or somatic cell extracts results in a solution of polypeptides, but not in gelation. Addition of different concentrations of plasmin shows that high levels of plasmin activity result in 2-phase separation with clarification of milk and creaming (Crudden, Afoufa-Bastien, et al., 2005; Kohlmann, et al., 1991). Addition of low concentration of plasmin or plasminogen on the other 55 hand results in gel formation (Kohlmann, et al., 1991). Similar effects are shown, when somatic cell extracts are added to milk (Kelly & Foley, 1997). A second impact factor is the storage temperature of UHT milk. Gelation is slow at low storage temperatures, most rapid at 20-25 °C, and does not or very late occur at storage at > 30 °C (Harwalkar, 1992; Malmgren, 2007; Manji & Kakuda, 1986; Nieuwenhuijse & van Boekel, 2003). Since proteolysis was on-going in these UHT milks, it is assumable that proteolysis by plasmin was too fast to allow gel formation at storage temperatures > 30 °C (Nieuwenhuijse & van Boekel, 2003). Manji & Kakuda (1988) tried to separate the effect of proteolysis from the effect of storage temperature and time by pre- hydrolysing milk with plasmin to different degrees before storage. They found that gelation occurred faster at higher storage temperatures, but did not find a correlation between the time of gelation and the degree of hydrolysis. Plasmin induced age gelation is therefore most likely a two-step process, consisting of the proteolysis itself and physical-chemical changes in the casein structure. In the UHT trial conducted for Paper II and III, two UHT milks with plasmin activity were produced: The UHT milk described in Paper II with a pre-heat treatment of 72 °C for 180 s and a UHT milk pre-heat treated at 72 °C for 5 s. While two replicates of the latter UHT milk were identical and showed comparable plasmin activities and proteolysis compared to the milk used for Paper II and III, one replicate showed decreased plasmin activity and reduced proteolysis of caseins and peptide formation (Figure 25).

Figure 25: Hydrolysis of αS1-casein (A), development of β-casein f(1-28) (B), plasmin activity (C) and viscosity development (D) in three replicate UHT milks (●,○,□) heat treated at >150 °C for > 0.2 s with a pre-heat treatment of 72 °C for 5 s during storage.

The time point of gelation, analysed as viscosity increase, in this replicate was also delayed by 2- 3 weeks compared to the other two replicates (Figure 25D) and occurred when the same level of

αS1-casein had been hydrolysed (Figure 25A). This indicates that at least in a narrow range, the extent of proteolysis does have an effect on the time point of gelation. In conclusion, the plasmin induced age gelation depends mainly on the following critical parameters at a given storage temperature:

 Minimum degree of proteolysis to sufficiently destabilize the casein micelle.  Minimum storage time for changes in casein micelle structure to form a gel which mainly depends on the storage temperature.  Maximum degree of proteolysis so that the peptides are no longer able to form a gel matrix any longer.

Although this is a very simplified model, these three factors could explain the different findings in relation to storage time and temperature and degree of proteolysis upon gelation. A schematic illustration of the model is shown in Figure 26:

Figure 26: Schematic illustration of the critical parameters for plasmin induced age gelation as indicated by dashed lines: Maximum degree of proteolysis (1), minimum degree of proteolysis (2) and minimum storage time (3). Other factors could influence these parameters as well. Proteolysis correlated well with a decrease in pH, which could affect the calcium equilibrium in milk and the stability of the casein micelle. Whether such changes are sufficient to affect casein micelle stability is unknown and the effect of calcium on age gelation by plasmin is limited (Nieuwenhuijse & van Boekel, 2003). Another factor could be the activation of plasminogen and the previously discussed change in the rate of proteolysis, which makes the comparison of UHT milk systems and model systems with added plasmin difficult. Destabilisation of the casein micelle could also change the proteolytic pattern of plasmin, as indicated in Paper III, and proteolysis and destabilisation could have large influences on each other. In contrast to age gelation by bacterial proteases, the mechanism of casein micelle destabilisation and gel formation as a consequence of plasmin activity is largely unknown. Electron micrographs of plasmin induced gels showed partially disintegrated casein micelles and the gel was made of threads and knots of protein (Aroonkamonsri, 1996; de Koning, et al., 1985).The presence of homogenized fat globules with a size distribution similar to casein micelles interfered with the particle size measurement of casein micelles in the study presented in Paper II. We could therefore not investigate changes in the particle size distribution of casein micelles, which could indicate disintegration caused by proteolysis. An attempt was made to interpret the observed decrease in Lab-colour values with changes in casein micelle structure. Dilution of UHT milk

58 without proteolysis with water resulted in a decrease of the b-value, i.e. less yellow, while the L- and a-values remained constant. Dilution with sodium citrate and dissociation of the casein micelles resulted in a decrease of L-, a-, and b-values similar to the UHT milk with plasmin activity, which could indicate a disruption and dissociation of casein micelles due to proteolysis. The initial gel is usually formed at the bottom of the container and has been described as fragile and custardy (Aroonkamonsri, 1996; Datta & Deeth, 2003; Kelly & Foley, 1997). The gel found in the UHT milk described in Paper II showed similar features as can be seen in Figure 27.

Figure 27: Effect of Storage time on gel formation in the bottom of the bottles in 3 replicates of UHT milk in Paper II. Ages of milk: 13, 12 and 11 weeks (from left to right). Notice the increase in gel volume over time.

The gel at the bottom of the bottle increased in volume during the storage time and after approximately 20 weeks, the entire bottle was gelled. Although the UHT milk was not analysed further, some bottles were kept for visual inspection and after 25 weeks, a clear phase separation between the gel and the serum phase was observed. According to Nieuwenhuijse & van Boekel (2003), gelation is dependent on the dissociation and aggregation of peptides since non-casein nitrogen (NCN) increases upon proteolysis by plasmin. Analysis of pH 4.6 soluble peptides showed that main polypeptides found in the serum were proteose peptones. The only large polypeptide deriving from αS1-casein was f(125-199) and it was associated with caseins. Comparison of protein and peptide compositions of gel and liquid phases showed a similar composition, but the gel contained lower concentrations of κ-casein. The main bond type to stabilize a plasmin induced gel seems to be hydrophobic interaction (Aroonkamonsri, 1996; Nieuwenhuijse & van Boekel, 2003). 59

In Paper II, we compared the gelation mechanism of plasmin induced gelation with acid gelation of milk, which is caused by reduction of charge repulsion and solubilisation of colloidal calcium phosphate (Lucey, et al., 2001; Lucey & Singh, 2003). The plasmin induced gel found in the UHT milk exhibited a similar microstructure as found in acid gels (Figure 28). The proteolytic pattern of plasmin on the other hand showed that plasmin hydrolysed caseins preferably in hydrophilic domains and at the border of hydrophilic-hydrophobic or hydrophilic-phosphoserine regions (Paper III).

Figure 28: CLSM micrographs of gel particles extracted from gelled UHT milk after 13 weeks of storage at different magnifications. Protein is stained green, fat is stained red. Fat globule clusters and their incorporation in the gel network can be seen. From Paper II. This indicates that plasmin could destabilize the casein micelle by hydrolysing around regions, which stabilize the casein micelle by calcium phosphate and hydrophobic interactions. It could be hypothesized that this proteolysis pattern could result in a disruption of the casein micelle, when enough stabilizing regions are hydrolysed. Loss of the casein micelle structure could subsequently result in a rearrangement of the polypeptides to a gel like network. The exact mechanism of plasmin induced age gelation still remains unknown. The UHT trial conducted in this project aimed to investigate the effect of plasmin on UHT milk in general and it was not known beforehand, whether gelation actually would occur. In order to investigate the

60 relation between proteolysis and gelation, UHT milks with a broad range of residual plasmin activities should be produced. The level of residual plasmin activity could be easier controlled by choosing one pre-heat temperature and changing the holding time to create appropriate residual plasmin activities. To study gelation mechanisms, especially destabilisation of casein micelles, the use of skim milk as feed would allow to observe changes in particle size. The release of caseins or peptides in relation to proteolysis could be achieved by separating the serum phase from casein micelles by ultracentrifugation and analysis of both fractions for peptides and intact caseins. A deeper investigation of the gel by either addition of citrate or cooling or transferring the gel into a hydrophobic medium could allow insight to, whether calcium or hydrophobic bonds stabilize the gel.

6.4.2 Flavour development

6.4.2.1 Chemical changes affecting flavour development

Freshly prepared UHT milk has a distinctive cooked, sulphurous flavour, which disappears during storage. At the same time, an aged and stale flavour develops (Figure X). In addition, flavour attributes such as ‘heated’ and ‘oxidized’ develop. The flavour of UHT milk is commonly perceived as poor and is a problem for the dairy industry since the customer acceptance for UHT milk is low (Colahan-Sederstrom & Peterson, 2004; Perkins, M. L. & Deeth, 2001).

Figure 29: Diagrammatic representation of flavour changes in UHT milk during storage. Adapted from Burton (1994). The initial cooked flavour arises from volatile sulphur components, which are formed during the UHT treatment. The major source of these sulphur components are free thiol groups, particularly of β-lg. The MFGM is also rich in free thiol groups and the concentration of volatile sulphur

61 components increases with increasing fat content (Al-Attabi, et al., 2009; Vazquez-Landaverde, et al., 2006). In addition, several volatiles can be formed during the Maillard reaction with methionine and cysteine (Al-Attabi, et al., 2009). The main components responsible for the cooked flavour are methanethiol, dimethylsulphide and dimethyldisulphide. Hydrogen sulphide itself has a relatively high odour activity value, but can react further to other sulphur volatiles (Vazquez-Landaverde, et al., 2006; Zabbia, et al., 2011). Disappearance of cooked flavour is dependent on the concentration of oxygen in UHT milk. In the presence of oxygen, volatile sulphur components are rapidly oxidized and the cooked flavour decreases (Al-Attabi, et al., 2009; Burton, 1994a). Directly processed UHT milk usually has a lower initial cooked flavour compared to indirectly processed UHT milk. This is due to the lower overall heat load in directly processed UHT milk and some of the volatile components can be removed during the vacuum cooling step (Datta, et al., 2002). On the other hand, most of the oxygen is removed during vacuum cooling as well and the cooked flavour disappears slower compared to indirectly processed UTH milk (Datta, et al., 2002; Valero, et al., 2001). High levels of oxygen accelerate the appearance of stale and oxidized flavour. Stale and oxidized flavours have been linked to the presence of lipid oxidation products, such as methyl ketones and aldehydes (Calvo & de la Hoz, 1992; Perkins, Melinda L., et al., 2005; Valero, et al., 2001; Vazquez-Landaverde, et al., 2005). Aldehydes are regarded to have a higher impact on flavour than methyl ketones due to their lower sensory threshold (Datta, et al., 2002). Oxidation of proteins can also lead to an unpleasant oxidized or ‘activated flavour’ (Burton, 1994a). Methyl ketones also form in the Maillard reaction and contribute to the heated flavour of UHT milk. Other Maillard reaction products contributing to the flavour are lactones, diacetyl, maltol, vanillin and benzaldehyde (Calvo & de la Hoz, 1992; Contarini, et al., 1997; Jansson, Jensen, et al., 2014).

6.4.2.2 Effect of enzymes on flavour

Flavour development of UHT milk also depends on raw milk quality. Psychotrophic bacteria are able to produce heat resistant lipases, whose activity can lead to a rancid flavour (Burton, 1994a). Proteolysis can indirectly affect the flavour of UHT milk. Increased levels of free amino acids accelerate the Maillard reaction and lead to a higher concentration of Strecker degradation products (Jansson, Clausen, et al., 2014). Several studies reported the occurrence of bitterness before gelation of milk in stored UHT milks (Burton, 1994a; McKellar, et al., 1984; Nieuwenhuijse & van Boekel, 2003). While many studies

62 focussed on age gelation of milk, studies on bitterness in milk are scarcer. Since bitterness occurs before gelation, it is the shelf life limiting factor and of higher importance for the dairy industry. Bacterial proteases are found to cause bitterness at very low levels of proteolysis (McKellar, 1981). These proteases cleave relatively unspecific and are able to hydrolyse hydrophobic domains of the caseins, which are the largest source of bitter peptides (Gaucher, et al., 2011; Lemieux & Simard, 1992). The appearance of bitterness thus has mostly been connected to the presence of bacterial proteases. The appearance of bitterness in the UHT milk described in Paper II was linked to plasmin since the milk was as fresh as possible, of excellent microbial quality and proteolysis of bacterial proteases was not detected during storage (Paper III). Huijs et al. (2004) report plasmin as the cause for bitterness in directly processed UHT milk with ultra-short holding time, similar to the process used in Paper II. Bitterness is also observed in directly processed UHT milk with high somatic cell count and plasmin activity (Santos, et al., 2003; Topçu, et al., 2006). The milk also contained a high level of psychotrophic bacteria so the exact cause of bitterness cannot be differentiated. The literature only describes one sensory evaluation of peptides produced by plasmin and it is done on isolated β-casein (Harwalkar, et al., 1993). Proteolysis of β-casein caused an astringent flavor and was linked to the formation of γ-caseins. Peptides from tryptic digests of β-casein were described as ’astringent, tickling, burning and bitter’, verifying that proteolysis of β-casein by plasmin could lead to bitter peptides (Bumberger & Belitz, 1993). A common way to estimate the bitterness of a peptide is the Q-rule. Bitterness has for long been associated with the hydrophobicity of a peptide. The Q-value is an average value of the hydrophobicity of the amino acid side chains divided by the number of amino acids (Ney, 1971). Peptides with a Q-value < 1300 cal res-1 are not bitter while peptides with Q > 1400 cal res-1 are generally considered as bitter. No prediction of bitterness is possible for Q-values between 1300- 1400 cal res-1 (Lemieux & Simard, 1992; Ney, 1971). In Paper III, we identified 23 potentially bitter peptides based on the Q-rule. With the exception of β-casein f(49-97), the potential bitter peptides originating from β-casein found in the UHT milk were not described as bitter in tryptic digests (Bumberger & Belitz, 1993). Interestingly, the most bitter peptide found in tryptic digests of β-casein, β-casein f(203-209), was not identified in UHT milk (Bumberger & Belitz, 1993). The C-terminal region of β-casein is very hydrophobic and it was suggested that it is inaccessible for plasmin in a milk system. Most of the peptides, also previously proposed bitter peptides, derived from hydrophobic regions of αS1- and αS2-casein. (Habibi‐Najafi, et al., 1996; Le Bars & Gripon, 1989; Maehashi & Huang, 2009).

Although the Q-rule allows for an initial assessment of potential bitter peptides, it was not possible to ascribe the observed bitterness in the UHT milk to certain peptides. An exception to the Q-rule is that it is only valid up to a molecular weight of 6000 Da (Lemieux & Simard, 1992; Ney, 1971). In order to identify the sensory properties of these potential bitter peptides, they would need to be either extracted from milk or synthesized and evaluated sensorically. This would have to be done in a milk system to include matrix effects, which could partially mask bitter taste (Toelstede & Hofmann, 2008). Determination of the formation rate and quantification of confirmed bitter peptides could allow a prediction of bitterness and consequently the shelf life of UHT milk.

6.4.3 Lactosylation of milk proteins during storage of UHT milk

As indicated in section 6.3.4, the Maillard reaction takes place during storage of UHT milk at ambient temperatures and results in an increase in protein lactosylation during storage. The rate of protein lactosylation depends on the storage temperature with increased rates of protein lactosylation at higher storage temperatures (Bosch, et al., 2008; Corzo, et al., 1994; Nangpal & Reuter, 1990). Nangpal & Reuter (1990) reported a linear increase in furosine concentration in UHT milk during storage at 20 and 30 °C, with an increase in furosine of 1 mg L-1 per week, similar to the rate of furosine increase reported for the UHT milks in Paper IV (Figure 30). The increase in furosine over time was found to be independent of the initial furosine concentration (Paper IV). At higher storage temperatures, the increase in furosine is not linear any more at prolonged storage times as lactulosyllysine reacts further and intermediate and late Maillard reaction products are formed (Le, et al., 2011; Nangpal & Reuter, 1990).

Figure 30: Furosine development in directly heated UHT milk (> 150 °C for < 0.2 s) pre-heated at 95 °C for 5 (●) or 180 s (○) and in indirect UHT milk (140 °C for 4 s, □) during storage at 20 °C. Error bars indicate standard deviation, n = 3; from Paper IV. 64

Assessment of protein lactosylation by measurement of furosine includes an 18 h acid boiling step to form furosine followed by analysis of furosine by HPLC, making this analysis expensive and time intensive. Lactosylation can also be observed by LC-MS using the protein composition method described in Paper II & IV. This method requires a one hour sample preparation step and a HPLC run time of 20 min per sample and can give a more detailed view on the Maillard reaction since the lactosylation of different proteins can be observed.

Figure 31: Summed mass spectra (retention time 11.0-13.2 min) of unmodified and lactosylated αS1-casein 8 P in indirect UHT milk (140 °C for 4 s) after 6 months of storage at 20 °C. Inset is showing a zoom of the most abundant ions. The m/z shifts due addition of lactose are indicated. From Paper IV.

Lactosylation renders a protein more hydrophilic and lactosylated proteins elute earlier in reverse phase HPLC than non lactosylated proteins (Czerwenka, et al., 2006; Losito, et al., 2007; Losito, et al., 2010). This results in a broadening of peaks in the UV chromatogram of proteins (Paper II & IV). A chromatographic separation of lactosylated proteins, which would allow for quantification of lactosylation using the UV signal, is hard to achieve since the change in hydrophobicity upon lactosylation is little. Analysis of summed mass spectra of a peak containing non lactosylated and lactosylated proteins on the other hand allows a relative comparison of the MS intensities of the different lactosylated proteins. The comparison can either be performed on specific ions (Figure 31) or on deconvoluted spectra (Figure 32)

Figure 32: Deconvoluted mass spectrum of αS1-casein 8 P in indirect UHT milk (140 °C for 4 s) after 0 (A) and 6 (B) months of storage at 20 °C. Mass shifts due to addition of lactose are indicated. From Paper IV.

Exploration of milk protein lactosylation has mainly been focused on whey proteins, especially pH 4.6 soluble whey proteins (Siciliano, et al., 2013). This limits the characterisation of protein lactosylation to milk products in which a sufficient amount of pH 4.6 soluble whey proteins are present (Meltretter, et al., 2009). In products where protein lactosylation actually could affect the overall product quality, the heat treatment is so severe that most or all of the whey proteins are denatured, such as the indirectly processed UHT milk in Paper IV or milk powders. Furthermore, denaturation and unfolding of whey proteins could affect the kinetic of lactosylation due to increased susceptibility of lysine residues (Losito, et al., 2010). Still for a limited range of heat treatments a quantitative relationship between lactosylation of whey proteins and heat markers could be established. The used heat markers were whey protein denaturation as well as the fluorescence of advanced Maillard products and soluble tryptophan (FAST) index (Losito, et al., 2010; Meltretter, et al., 2009). While these markers reflect the severity of the heat treatment, they do not necessarily correlate with the extent of the Maillard reaction. The only direct correlation of milk protein lactosylation measured by MS and furosine concentration was established for the lactosylation of β-casein in processed cheese (Steffan, et al., 2006). Other studies on lactosylation of caseins measured by MS are only qualitative (Johnson, et al., 2011; Scaloni, et al., 2002). Quantitative analysis of milk protein lactosylation can be challenging since reports on response factors of differently lactosylated protein species differ and quantitative comparison with furosine is lacking in most studies. For proteins in their native state, such as whey proteins and lysozyme, the response factor was found to be not affected by lactosylation of proteins when analysed with

66 both electrospray ionisation MS (ESI-MS) and matrix assisted laser desorption (MALDI) MS (Czerwenka, et al., 2006; Meltretter, et al., 2009; Yeboah & Yaylayan, 2001). For reduced caseins without any notable tertiary structure however, a lower response factor for lactosylated proteins was found compared to non lactosylated caseins (Scaloni, et al., 2002). In Paper IV, both lactosylated species of caseins and whey proteins exhibited a reduced response factor in MS intensity, although the UV chromatogram showed no loss of protein and was constant throughout the storage period. Although the response factors were not identical for different lactosylated protein species in Paper IV, a linear regression of the relative area distribution of the proteins was possible. The rate of lactosylation measured as decrease in unlactosylated protein correlated with the number of lysine residues in protein for both caseins and whey proteins. The decrease of unlactosylated proteins correlated well with the increase in furosine during storage, allowing a quantitative estimation of protein lactosylation by LC-MS measurements in the covered range of furosine. The differences in response factors could be attributed to the structure of proteins. The globular structure of whey proteins results in an overall decreased ionisation. A comparison of mass spectra of pH 4.6 soluble whey proteins and reduced whey proteins in low pasteurized milk samples showed large differences. The most abundant charge state for reduced β-lg and α-la were 17 and 13 compared to 11 and 6 for pH 4.6 soluble whey proteins. With increasing charge state, a potential reduction in the ionisation efficiency of a lactosylated lysine residue would become more pronounced compared to lower charge states, where ionisable residues are present in excess. Analysis of the response factors of lactosylated pH 4.6 soluble whey proteins and comparison to the response factors after reduction of the same samples could give an insight in the underlying mechanism. A different approach to determine the extent of lactosylation milk is the analysis of lactosylated peptides present in milk (Pinto, et al., 2012). Plasmin derived peptides and especially β-casein f(1-28) were found to be well suited to detect lactosylation in milk as they are already present in raw milk. A high concentration of lactosylated β-casein f(1-28) was detected in the peptidomic study of UHT milk with plasmin activity in Paper III, but the concentration of β-casein f(1-28) was not high enough in the UHT milks without plasmin activity in Paper IV to obtain reproducible results.

7 Conclusions

The studies presented in this thesis have provided insight into the measurement of plasmin activity and the effect of plasmin activity on the shelf life development of UHT milk. The tested conditions for measuring plasmin activity in an assay showed that the measured plasmin activity is not only dependent on sample composition and treatment, but also on sample preparation and buffer systems used. The term activity has been used in literature to describe plasmin concentration, real activity and apparent activity including interfering components, but should actually be referred to as the rate of hydrolysis of a synthetic substrate under given conditions. There is no right or wrong way to measure activity, but one should be aware of all factors that influence the activity in an assay in order to take decisions on assay conditions to be used and to draw conclusions from activity measurements on changes in activity or the characteristics of plasmin in milk. Considering these factors allows the design of assays that measure an activity, which in the best way reflects the research aim. Although it is more time and work intensive than an activity assay, following proteolysis by plasmin gives a direct insight to the effect of plasmin activity in UHT milk. Identification of peptides allowed identification of several marker peptides for plasmin activity, which were not further hydrolysed during the presently applied storage period. Furthermore, differences in affinity of plasmin for specific cleavage sites lead to a time dependent appearance of peptides, which can potentially be used to characterize different stages of proteolysis. It is possible to model the proteolysis of β-casein on a protein and peptide level, and of αS2-casein on a peptide level with the analytical tools available in this project. Unfortunately, we could only model the kinetic at one plasmin concentration and were not able to validate the model. In terms of quality defects, plasmin activity could be linked to the occurrence of bitterness and age gelation in UHT milk in this study. The identification of (potential) bitter peptides provides a starting point for further studies on the sensory threshold of bitter peptides, which would allow the establishment of critical peptide concentrations that can limit the shelf life of UHT milk. Proteolysis by plasmin can cause age gelation of UHT milk, but the underlying mechanisms are still not fully understood. The gelation mechanism is most likely a two-step process and the proteolysis pattern leading to gelation can provide a hypothetical mechanism of destabilisation of the casein micelle by proteolysis. The quality defects in the UHT milk investigated in this study occurred only after substantial proteolysis with 60-70 % of αS1- and β-casein being hydrolysed for the milk to taste bitter and more than 90 % to cause gelation. While the extent of proteolysis necessary for gelation and bitterness could not be answered for different residual plasmin levels, this study gave an insight

68 in the effect of plasmin activity in milk and provided a setup to explore this further. The results show that low levels of plasmin activity in UTH milk can be acceptable, when the quality defects are matched with the shelf life of the UHT milk.

8 Perspectives

The research effort on plasmin and the plasmin system in milk is considerable, and with every published study, some questions are answered, but new and often more profound questions arise. The same applies for this study. The presented results give an insight into the effect of proteolysis on UHT milk, but cannot answer the mechanism of age gelation or establish a direct connection between plasmin activity, proteolysis and quality defects. In retro perspective, the UHT trial conducted in this study could have been planned differently. Applying the same preheat temperature and only varying the holding time would have allowed a more specific control over residual plasmin activities in the UHT milk and maybe could have been predicted using kinetic data in literature. Sensory evaluation of the UHT milk by a trained sensory panel would have given a more detailed view on off flavour and bitterness caused by plasmin. Most potential bitter peptides were present in low concentrations and could not be quantified using the UV signal. Quantification of these peptides can be improved by optimizing the chromatographic separation or adapting the sample preparation for smaller hydrophilic peptides. An alternative is the use of an internal standard for the LC-MS analysis, which then allows quantification using the MS signal over a certain range. In order to study the gelation phenomenon, a skim milk system and dynamic light scattering instead of static light scattering should be used in order to investigate changes in the casein micelle structure. Furthermore, release of peptides and caseins from the casein fraction could be used as indicators for casein micelle destabilisation. A repetition of the UHT storage experiment in this study under the conditions mentioned above could enable the correlation of plasmin activity, proteolysis and quality defects and allow the dairy industry to control the effects of plasmin activity better. As indicated in Paper I, the dissociation of plasmin and caseins affect the interaction of plasmin with inhibitory components. A dissociation of plasmin and caseins during processing, for example by pressure, could result in an increased inhibition and/or inactivation of plasmin by whey proteins and inhibitors. Although we could not see differences in the heat inactivation of plasmin in the presence of NaCl and EACA, it could be hypothesized that the dissociation of plasmin and caseins could negate the protective effect of caseins towards the heat inactivation of plasmin. Beside these follow up studies to this project, several more basic aspects of plasmin still need more detailed attention. The concentration of plasminogen activators and the higher heat stability of plasmin compared to plasminogen make them an important factor to limit proteolysis in milk and dairy products. The influence of storage conditions of raw or pasteurized milk on the

70 activation of plasminogen before further processing has only been partially studied and could be a starting point to reduce plasminogen activation. During the separation of milk into skim milk and cream, somatic cells and other large particles are separated as well. Most of the uPA is initially bound to somatic cells and the effect of the separation on the stability of somatic cells and release of uPA is unknown. As uPA is considered the major plasminogen activator and has a higher heat stability compared to tPA, a reduction of uPA concentration in milk would reduce proteolysis in milk and milk products. Pre-processing of UHT milk yields several possibilities, which could result in a minimisation of the plasmin system load in the final product, which could positively affect product quality.

9 References

Al-Attabi, Z., D'Arcy, B. R., & Deeth, H. C. (2009). Volatile sulphur compounds in UHT milk. Crit Rev Food Sci Nutr, 49, 28-47.

Albenzio, M., Caroprese, M., Santillo, A., Marino, R., Taibi, L., & Sevi, A. (2004). Effects of somatic cell count and stage of lactation on the plasmin activity and cheese-making properties of ewe milk. Journal of Dairy Science, 87, 533-542.

Alichanidis, E., Wrathall, J. H., & Andrews, A. T. (1986). Heat stability of plasmin (milk proteinase) and plasminogen. Journal of Dairy Research, 53, 259-269.

Andrews, A. T., & Alichanidis, E. (1983). Proteolysis of caseins and the proteose-peptone fraction of bovine milk. Journal of Dairy Research, 50, 275-290.

Anema, S. G., & Li, Y. (2003a). Effect of pH on the Association of Denatured Whey Proteins with Casein Micelles in Heated Reconstituted Skim Milk. Journal of Agricultural and Food Chemistry, 51, 1640-1646.

Anema, S. G., & Li, Y. (2003b). Association of denatured whey proteins with casein micelles in heated reconstituted skim milk and its effect on casein micelle size. Journal of Dairy Research, 70, 73-83.

Anema, S. G. (2007). Role of κ-Casein in the Association of Denatured Whey Proteins with Casein Micelles in Heated Reconstituted Skim Milk. Journal of Agricultural and Food Chemistry, 55, 3635-3642.

Anema, S. G. (2008). Chapter 8 - The whey proteins in milk: thermal denaturation, physical interactions and effects on the functional properties of milk. In T. Abby, B. Mike, M. B. Harjinder SinghA2 - Abby Thompson & S. Harjinder (Eds.), Milk Proteins (pp. 239-281). San Diego: Academic Press.

Aroonkamonsri, J. (1996). The Role of Plasmin and Bacterial Proteinases on Age Gelation of UHT Milk (Pseudomonas fluorescens, Pseudomonas fragi, Pseudomonas luteola, Vibrio fluvialis). PhD Thesis, University of Guelph, Canada.

Baer, A., Ryba, I., & Collin, J. C. (1994). Binding of bovine plasminogen to immobilized casein and its activation thereon. International Dairy Journal, 4, 597-616.

Baldi, A., Savoini, G., Cheli, F., Fantuz, F., Senatore, E., Bertocchi, L., & Politis, I. (1996). Changes in plasmin-plasminogen-plasminogen activator system in milk from Italian Friesian herds. International Dairy Journal, 6, 1045-1053.

Bastian, E. D., Brown, R. J., & Ernstrom, C. A. (1991). Casein Interference in Bovine Plasmin Assays Using a Synthetic Substrate. Journal of Dairy Science, 74, 4119-4124.

Bastian, E. D., Hansen, K. G., & Brown, R. J. (1993). Inhibition of Plasmin by β-Lactoglobulin Using Casein and a Synthetic Substrate. Journal of Dairy Science, 76, 3354-3361. 72

Bastian, E. D., & Brown, R. J. (1996). Plasmin in milk and dairy products: an update. International Dairy Journal, 6, 435-457.

Baum, F., Ebner, J., & Pischetsrieder, M. (2013). Identification of multiphosphorylated peptides in milk. Journal of Agricultural and Food Chemistry, 61, 9110-9117.

Baum, F., Fedorova, M., Ebner, J., Hoffmann, R., & Pischetsrieder, M. (2013). Analysis of the endogenous peptide profile of milk: identification of 248 mainly casein-derived peptides. J Proteome Res, 12, 5447-5462.

Benfeldt, C., Larsen, L. B., Rasmussen, J. T., Andreasen, P. A., & Petersen, T. E. (1995). Isolation and characterization of plasminogen and plasmin from bovine milk. International Dairy Journal, 5, 577-592.

Benfeldt, C. (2006). Ultrafiltration of cheese milk: Effect on plasmin activity and proteolysis during cheese ripening. International Dairy Journal, 16, 600-608.

Benslimane, S., Dognin-Bergeret, M. J., Berdague, J.-L., & Gaudemer, Y. (1990). Variation with season and lactation of plasmin and plasminogen concentrations in Montbeliard cows' milk. Journal of Dairy Research, 57, 423.

Berglund, L., Andersen, M. D., & Petersen, T. E. (1995). Cloning and characterization of the bovine plasminogen cDNA. International Dairy Journal, 5, 593-603.

Birlouez-Aragon, I., Nicolas, M., Metais, A., Marchond, N., Grenier, J., & Calvo, D. (1998). A Rapid Fluorimetric Method to Estimate the Heat Treatment of Liquid Milk. International Dairy Journal, 8, 771-777.

Borda, D., Indrawati, Smout, C., Van Loey, A., & Hendrickx, M. (2004). High pressure thermal inactivation kinetics of a plasmin system. Journal of Dairy Science, 87, 2351-2358.

Bosch, L., Alegrı´a, A., Farré, R., & Clemente, G. (2008). Effect of storage conditions on furosine formation in milk–cereal based baby foods. Food Chemistry, 107, 1681-1686.

Bumberger, E., & Belitz, H. D. (1993). Bitter taste of enzymic hydrolysates of casein. I. Isolation, structural and sensorial analysis of peptides from tryptic hydrolysates of beta-casein. Zeitschrift fur Lebensmittel-Untersuchung und -Forschung, 197, 14-19.

Burbrink, C. N., & Hayes, K. D. (2006). Effect of thermal treatment on the activation of bovine plasminogen. International Dairy Journal, 16, 580-585.

Burton, H. (1994a). Properties of UHT-Processed Milk. In Ultra-High-Temperature Processing of Milk and Milk Products (pp. 254-291): Springer US.

Burton, H. (1994b). Chemical and Physical Changes in Milk at High Temperatures. In Ultra- High-Temperature Processing of Milk and Milk Products (pp. 44-76): Springer US.

Caessens, P. W. J. R., Daamen, W. F., Gruppen, H., Visser, S., & Voragen, A. G. J. (1999). β- Lactoglobulin Hydrolysis. 2. Peptide Identification, SH/SS Exchange, and Functional Properties of Hydrolysate Fractions Formed by the Action of Plasmin. Journal of Agricultural and Food Chemistry, 47, 2980-2990.

Calvo, M. M., & de la Hoz, L. (1992). Flavour of heated milks. A review. International Dairy Journal, 2, 69-81.

Cano-Ruiz, M. E., & Richter, R. L. (1997). Effect of Homogenization Pressure on the Milk Fat Globule Membrane Proteins. Journal of Dairy Science, 80, 2732-2739.

Cattaneo, S., Stuknytė, M., Pellegrino, L., & De Noni, I. (2014). Targeted peptides for the quantitative evaluation of casein plasminolysis in drinking milk. Food Chemistry, 155, 179-185.

Cesarman-Maus, G., & Hajjar, K. A. (2005). Molecular mechanisms of fibrinolysis. British Journal of Haematology, 129, 307-321.

Chove, L. M., Grandison, A. S., & Lewis, M. J. (2011). Comparison of methods for analysis of proteolysis by plasmin in milk. The Journal of Dairy Research, 78, 184.

Christensen, S., & Sottrup-Jensen, L. (1992). Bovine α2-antiplasmin N-Terminal and reactive site sequence. FEBS Letters, 312, 100-104.

Christensen, S., Wiegers, T., Hermansen, J., & Sottrup-Jensen, L. (1995). Plasma-derived protease inhibitors in bovine milk. International Dairy Journal, 5, 439-449.

Christensen, U., Bangert, K., & Thorsen, S. (1996). Reaction of human α2-antiplasmin and plasmin Stopped-flow fluorescence kinetics. FEBS Letters, 387, 58-62.

Claeys, W. L., Van Loey, A. M., & Hendrickx, M. E. (2003). Kinetics of hydroxymethylfurfural, lactulose and furosine formation in milk with different fat content. Journal of Dairy Research, 70, 85-90.

Claeys, W. L., Van Loey, A. M., & Hendrickx, M. E. (2004). Intrinsic time temperature integrators for heat treatment of milk. Trends in Food Science & Technology, 13, 293-311.

Colahan-Sederstrom, P. M., & Peterson, D. G. (2004). Inhibition of Key Aroma Compound Generated during Ultrahigh-Temperature Processing of Bovine Milk via Epicatechin Addition. Journal of Agricultural and Food Chemistry, 53, 398-402.

Contarini, G., Povolo, M., Leardi, R., & Toppino, P. M. (1997). Influence of Heat Treatment on the Volatile Compounds of Milk. Journal of Agricultural and Food Chemistry, 45, 3171- 3177.

Corzo, N., López-Fandiño, R., Delgado, T., Ramos, M., & Olano, A. (1994). Changes in furosine and proteins of UHT-treated milks stored at high ambient temperatures. Zeitschrift für Lebensmittel-Untersuchung und Forschung, 198, 302-306.

Crudden, A., & Kelly, A. L. (2003). Studies of plasmin activity in whey. International Dairy Journal, 13, 987-993.

Crudden, A., Afoufa-Bastien, D., Fox, P. F., Brisson, G., & Kelly, A. L. (2005). Effect of hydrolysis of casein by plasmin on the heat stability of milk. International Dairy Journal, 15, 1017-1025.

Crudden, A., Fox, P. F., & Kelly, A. L. (2005). Factors affecting the hydrolytic action of plasmin in milk. International Dairy Journal, 15, 305-313.

Crudden, A., Oliveira, J. C., & Kelly, A. L. (2005a). Kinetics of changes in plasmin activity and proteolysis on heating milk. Journal of Dairy Research, 72, 493-504.

Crudden, A., Oliveira, J. C., & Kelly, A. L. (2005b). Kinetic studies of the thermal inactivation of plasmin in acid or sweet whey. International Dairy Journal, 15, 1245-1253.

Czerwenka, C., Maier, I., Pittner, F., & Lindner, W. (2006). Investigation of the Lactosylation of Whey Proteins by Liquid Chromatography−Mass Spectrometry. Journal of Agricultural and Food Chemistry, 54, 8874-8882.

Dallas, D. C., Guerrero, A., Parker, E. A., Garay, L. A., Bhandari, A., Lebrilla, C. B., Barile, D., & German, J. B. (2013). Peptidomic Profile of Milk of Holstein Cows at Peak Lactation. Journal of Agricultural and Food Chemistry, 62, 58-65.

Dalsgaard, T. K., Nielsen, J. H., & Larsen, L. B. (2007). Proteolysis of milk proteins lactosylated in model systems. Mol Nutr Food Res, 51, 404-414.

Dalsgaard, T. K., Heegaard, C. W., & Larsen, L. B. (2008). Plasmin Digestion of Photooxidized Milk Proteins. Journal of Dairy Science, 91, 2175-2183.

Dannenberg, F., & Kessler, H.-G. (1988). Reaction Kinetics of the Denaturation of Whey Proteins in Milk. Journal of Food Science, 53, 258-263.

Datta, N., & Deeth, H. C. (2001). Age Gelation of UHT Milk—A Review. Food and Bioproducts Processing, 79, 197-210.

Datta, N., Elliott, A. J., Perkins, M. L., & Deeth, H. C. (2002). Ultra-high-temperature (UHT) treatment of milk: comparison of direct and indirect modes of heating Australian Journal of Dairy Technology, 57, 211-227.

Datta, N., & Deeth, H. C. (2003). Diagnosing the cause of proteolysis in UHT milk. LWT - Food Science and Technology, 36, 173-182.

75 de Koning, P. J., de Wit, J. N., & Driessen, F. M. (1985). Age-thinning and gelation in concentrated and unconcentrated UHT-sterilised skim milk. Netherlands Milk Dairy Journal, 39, 71-87.

Dickow, J. A. (2011). Lenient Steam Injection heat treatment of milk and whey protein concentrate - Impact on enzymatic activities, protein denaturation and functional properties. PhD Thesis, Aarhus University.

Driessen, F. M., & van der Waals, C. B. (1978). Inactivation of Native Milk Proteinase by Heat Treatment. Netherlandss Milk Dairy Journal, 32, 245-254.

Dupont, D., Bailly, C., Grosclaude, J., & Collin, J.-C. (1997). Differential titration of plasmin and plasminogen in milk using sandwich ELISA with monoclonal antibodies. Journal of Dairy Research, 64, 77-86.

Durkee, D., & Hayes, K. D. (2008). Effects of cysteine addition and heat treatment during non-fat dry milk processing on the plasmin enzyme system. Dairy Science & Technology, 88, 81- 94.

EC. (2005). Commission Regulation (EC) No 2074/2005 of 5 December 2005 laying down implementing measures for certain products under Regulation (EC) No 853/2004 of the European Parliament and of the Council and for the organisation of official controls under Regulation (EC) No 854/2004 of the European Parliament and of the Council and Regulation (EC) No 882/2004 of the European Parliament and of the Council, derogating from Regulation (EC) No 852/2004 of the European Parliament and of the Council and amending Regulations (EC) No 853/2004 and (EC) No 854/2004. Official Journal of the European Union, 48, 27-59.

Eigel, W. N., Butler, J. E., Ernstrom, C. A., Farrell, H. M., Harwalkar, V. R., Jenness, R., & Whitney, R. M. (1984). Nomenclature of Proteins of Cow's Milk: Fifth Revision. Journal of Dairy Science, 67, 1599-1631.

Erbersdobler, H. F., & Somoza, V. (2007). Forty years of furosine – Forty years of using Maillard reaction products as indicators of the nutritional quality of foods. Molecular Nutrition & Food Research, 51, 423-430.

Fajardo-Lira, C. E., & Nielsen, S. S. (1998). Effect of Psychrotrophic Microorganisms on the Plasmin System in Milk1. Journal of Dairy Science, 81, 901-908.

Frohbieter, K. A., Ismail, B., Nielsen, S. S., & Hayes, K. D. (2005). Effects of Pseudomonas fluorescens M3/6 Bacterial Protease on Plasmin System and Plasminogen Activation. Journal of Dairy Science, 88, 3392-3401.

García-Risco, M. R., Recio, I., Molina, E., & López-Fandiño, R. (2003). Plasmin Activity in Pressurized Milk. Journal of Dairy Science, 86, 728-734.

Gaucher, I., Molle, D., Gagnaire, V., Leonil, J., Rousseau, F., & Gaucheron, F. (2009). Destabilization of commercial UHT milks: proteolysis and changes in milk particles. Milchwissenschaft, 64, 43-47. 76

Gaucher, I., Tanguy, G., Fauquant, J., Jardin, J., Rousseau, F., Robert, B., Madec, M., & Gaucheron, F. (2011). Proteolysis of casein micelles by Pseudomonas fluorescens CNRZ 798 contributes to the destabilisation of UHT milk during its storage. Dairy Science & Technology, 91, 413-429.

Gazi, I., Vilalva, I. C., & Huppertz, T. (2014). Plasmin activity and proteolysis in milk protein ingredients. International Dairy Journal, 38, 208-212.

Goddard, S. J. (1996). Effect of thiol reagents on the acid-heat-induced gelation of high-heat skim milk. Journal of Dairy Research, 63, 639-642.

Grufferty, M. B., & Fox, P. F. (1988a). Heat stability of the plasmin system in milk and casein systems. New Zealand Journal of Dairy Science and Technology, 23, 143-152.

Grufferty, M. B., & Fox, P. F. (1988b). Milk alkaline proteinase. Journal of Dairy Research, 55, 609-630.

Guinot-Thomas, P., Al Ammoury, M., Le Roux, Y., & Laurent, F. (1995). Study of proteolysis during storage of raw milk at 4 °C: Effect of plasmin and microbial proteinases. International Dairy Journal, 5, 685-697.

Habibi‐Najafi, M. B., Lee, B. H., & Law, B. (1996). Bitterness in cheese: A review. Critical Reviews in Food Science and Nutrition, 36, 397-411.

Haissat, S., Marchal, E., Montagne, P., Humbert, G., Bene, M. C., Faure, G., & Linden, G. (1994). Quantitative Characterization of Bovine Plasminogen Binding to Caseins. Analytical Biochemistry, 222, 472-478.

Harwalkar, V. R. (1992). Age gelation of sterilized milks. In P. F. Fox (Ed.), Advanced Dairy Chemistry Volume 1: Proteins (pp. 691-734). London: Elsevier Science Publishers Ltd.

Harwalkar, V. R., Cholette, H., McKellar, R. C., & Emmons, D. B. (1993). Relation Between Proteolysis and Astringent Off-Flavor in Milk. Journal of Dairy Science, 76, 2521-2527.

Hayes, K. D., Ozen, B. F., Nielsen, S. S., & Mauer, L. J. (2003). FTIR determination of ligand- induced secondary and tertiary structural changes in bovine plasminogen. Journal of Dairy Research, 70, 461-466.

Hayes, M. G., McSweeney, P. L. H., & Kelly, A. L. (2002). The influence of native and heat- denatured whey proteins on enzyme activity. 1. Plasmin. Milchwissenschaft, 57, 208-211.

Hayes, M. G., & Kelly, A. L. (2003). High pressure homogenisation of milk (b) effects on indigenous enzymatic activity. Journal of Dairy Research, 70, 307-313.

He, J.-S., Zhu, S., Mu, T.-H., Yu, Y., Li, J., & Azuma, N. (2011). αs-casein inhibits the pressure- induced aggregation of β-lactoglobulin through its molecular chaperone-like properties. Food Hydrocolloids, 25, 1581-1586.

Hedstrom, L. (2002). Serine Protease Mechanism and Specificity. Chemical Reviews, 102, 4501- 4524.

Heegaard, C. W., Christensen, T., Rasmussen, M. D., Benfeldt, C., Jensen, N. E., Sejrsen, K., Petersen, T. E., & Andreasen, P. A. (1994). Plasminogen activators in bovine milk during mastitis, an inflammatory disease. Fibrinolysis, 8, 22-30.

Heegaard, C. W., Rasmussen, L. K., & Andreasen, P. A. (1994). The plasminogen activation system in bovine milk: Differential localization of tissue-type plasminogen activator and urokinase in milk fractions is caused by binding to casein and urokinase receptor. Biochimica et Biophysica Acta (BBA) - Molecular Cell Research, 1222, 45-55.

Heegaard, C. W., Andreasen, P. A., Petersen, T. E., & Rasmussen, L. K. (1997). Binding of plasminogen and tissue-type plasminogen activator to dimeric αs2-casein accelerates plasmin generation. Fibrinolysis and Proteolysis, 11, 29-36.

Hervio, L. S., Coombs, G. S., Bergstrom, R. C., Trivedi, K., Corey, D. R., & Madison, E. L. (2000). Negative selectivity and the evolution of protease cascades: the specificity of plasmin for peptide and protein substrates. Chemistry & Biology, 7, 443-452.

Hinrichs, J., & Rademacher, B. (2004). High pressure thermal denaturation kinetics of whey proteins. Journal of Dairy Research, 71, 480-488.

Holst, H. H., Gunther, W. S., Andersen, J., & Lundgren, K. (2010). Long shelf life milk and milk-related products, and a process and milk processing plant for their manufacture. Patent application WO 2010085957. Arla Foods amba, Viby, Denmark.

Holst, H. H., Sundgren, A., & Rauh, V. M. (2012). Lactose-reduced milk-related product, and a process and milk processing plant for its manufacture. Patent application WO 2012010699. Arla Foods amba, Viby, Denmark.

Hougaard, A. B., Hammershøj, M., Vestergaard, J. S., Poulsen, O. L. E., & Ipsen, R. H. (2009). Instant infusion pasteurisation of bovine milk. I. Effects on bacterial inactivation and physical–chemical properties. International Journal of Dairy Technology, 62, 484-492.

Huijs, G., Van Asselt, A. J., Verdurmen, R. E. M., & De Jong, P. (2004). High speed milk. Dairy Industries International, 69, 30-32.

Huppertz, T., Fox, P. F., & Kelly, A. L. (2004). Plasmin activity and proteolysis in high pressure- treated bovine milk. Lait, 84, 297-304.

Ismail, B., & Nielsen, S. S. (2010). Invited review: Plasmin protease in milk: current knowledge and relevance to dairy industry. Journal of Dairy Science, 93, 4999-5009.

ISO-4833. (2003). Microbiology of food and animal feeding stuffs - horizontal method for the enumeration of microorganisms - Colony-count technique at 30°C: Geneva, Switzerland: International Standardisation Organisation.

Iucci, L., Lanciotti, R., Kelly, A., & Huppertz, T. (2008). Plasmin activity in high-pressure- homogenised bovine milk. Milchwissenschaft, 63, 68-70.

Jansson, T., Clausen, M. R., Sundekilde, U. K., Eggers, N., Nyegaard, S., Larsen, L. B., Ray, C., Sundgren, A., Andersen, H. J., & Bertram, H. C. (2014). Lactose-Hydrolyzed Milk Is More Prone to Chemical Changes during Storage than Conventional Ultra-High- Temperature (UHT) Milk. Journal of Agricultural and Food Chemistry.

Jansson, T., Jensen, S., Eggers, N., Clausen, M. R., Larsen, L. B., Ray, C., Sundgren, A., Andersen, H. J., & Bertram, H. C. (2014). Volatile component profiles of conventional and lactose-hydrolyzed UHT milk—a dynamic headspace gas chromatography-mass spectrometry study. Dairy Science & Technology, 94, 311-325.

Johnson, P., Philo, M., Watson, A., & Mills, E. N. C. (2011). Rapid Fingerprinting of Milk Thermal Processing History by Intact Protein Mass Spectrometry with Nondenaturing Chromatography. Journal of Agricultural and Food Chemistry, 59, 12420-12427.

Kelly, A. L., & Foley, J. (1997). Proteolysis and storage stability of UHT milk as influenced by milk plasmin activity, plasmin/β-lactoglobulin complexation, plasminogen activation and somatic cell count. International Dairy Journal, 7, 411-420.

Kelly, A. L., & McSweeney, P. L. H. (2003). Indigenous Proteinases in Milk. In P. F. Fox & P. L. H. McSweeney (Eds.), Advanced Dairy Chemistry—1 Proteins (pp. 495-521): Springer US.

Kelly, A. L., & Fox, P. F. (2006). Indigenous enzymes in milk: A synopsis of future research requirements. International Dairy Journal, 16, 707-715.

Kelly, A. L., O’Flaherty, F., & Fox, P. F. (2006). Indigenous proteolytic enzymes in milk: A brief overview of the present state of knowledge. International Dairy Journal, 16, 563-572.

Kennedy, A., & Kelly, A. L. (1997). The influence of somatic cell count on the heat stability of bovine milk plasmin activity. International Dairy Journal, 7, 717-721.

Kessler, H. G., & Horak, P. (1981). Objektive Beurteiling der UHT-Milcherhitzung durch Normierung bakteriologischer und chemische Effekte. Milchwissenschaft, 36, 129-133.

Kessler, H. G. (2002). Food and bio process engineering: dairy technology (5 ed.). Munich, Germany: Verlag A. Kessler.

Kohlmann, K. L., Nielsen, S. S., & Ladisch, M. R. (1988). Effect of Serine Proteolytic Enzymes (Trypsin and Plasmin), Trypsin Inhibitor, and Plasminogen Activator Addition to Ultra- High Temperature Processed Milk. Journal of Dairy Science, 71, 1728-1739.

Kohlmann, K. L., Nielsen, S. S., & Ladisch, M. R. (1991). Effects of a Low Concentration of Added Plasmin on Ultra-High Temperature Processed Milk. Journal of Dairy Science, 74, 1151-1156.

Kolev, K., Owen, W. G., & Machovich, R. (1995). Dual effect of synthetic plasmin substrates on plasminogen activation. Biochimica et Biophysica Acta (BBA) - Protein Structure and Molecular Enzymology, 1247, 239-245.

Korycha-Dahl, M., Dumas, B. R., Chene, N., & Martal, J. (1983). Plasmin Activity in Milk. Journal of Dairy Science, 66, 704-711.

Larson, N. K., Ismail, B., Nielsen, S. S., & Hayes, K. D. (2006). Activity of Bacillus polymyxa protease on components of the plasmin system in milk. International Dairy Journal, 16, 586-592.

Le Bars, D., & Gripon, J. C. (1989). Specificity of plasmin towards bovine αs2-casein. Journal of Dairy Research, 56, 817-821.

Le Bars, D., & Gripon, J., C. (1993). Hydrolysis of αs1-casein by bovine plasmin. Lait, 73, 337- 344.

Le, T. T., Bhandari, B., & Deeth, H. C. (2011). Chemical and Physical Changes in Milk Protein Concentrate (MPC80) Powder during Storage. Journal of Agricultural and Food Chemistry, 59, 5465-5473.

Lee, S. J., & Sherbon, J. W. (2002). Chemical changes in bovine milk fat globule membrane caused by heat treatment and homogenization of whole milk. Journal of Dairy Research, 69, 555-567.

Lemieux, L., & Simard, R., E. (1992). Bitter flavour in dairy products. II. A review of bitter peptides from caseins: their formation, isolation and identification, structure masking and inhibition. Lait, 72, 335-385.

Losito, I., Carbonara, T., Monaci, L., & Palmisano, F. (2007). Evaluation of the thermal history of bovine milk from the lactosylation of whey proteins: an investigation by liquid chromatography–electrospray ionization mass spectrometry. Analytical and Bioanalytical Chemistry, 389, 2065-2074.

Losito, I., Stringano, E., Carulli, S., & Palmisano, F. (2010). Correlation between lactosylation and denaturation of major whey proteins: an investigation by liquid chromatography– electrospray ionization mass spectrometry. Analytical and Bioanalytical Chemistry, 396, 2293-2306.

Lu, D. D., & Nielsen, S. S. (1993a). Isolation and Characterization of Native Bovine Milk Plasminogen Activators. Journal of Dairy Science, 76, 3369-3383.

Lu, D. D., & Nielsen, S. S. (1993b). Heat Inactivation of Native Plasminogen Activators in Bovine Milk. Journal of Food Science, 58, 1010-1012.

Lu, D. D., & Nielsen, S. S. (1993c). Assays for Native Plasminogen Activators in Bovine Milk. Journal of Dairy Science, 76, 3362-3368.

Lu, R., Stevenson, C. D., Guck, S. E., Pillsbury, L. A., Ismail, B., & Hayes, K. D. (2009). Effect of various heat treatments on plasminogen activation in bovine milk during refrigerated storage. International Journal of Food Science & Technology, 44, 681-687.

Lucey, J. A., Tamehana, M., Singh, H., & Munro, P. A. (2001). Effect of heat treatment on the physical properties of milk gels made with both rennet and acid. International Dairy Journal, 11, 559-565.

Lucey, J. A., & Singh, H. (2003). Acid Coagulation of Milk. In P. F. Fox & P. L. H. McSweeney (Eds.), Advanced Dairy Chemistry—1 Proteins (pp. 1001-1025): Springer US.

Maehashi, K., & Huang, L. (2009). Bitter peptides and bitter taste receptors. Cellular and Molecular Life Sciences, 66, 1661-1671.

Malmgren, B. (2007). Evaluation of UHT milk processed by direct steam injection and steam infusion technology. Doctoral thesis, Lund University, Sweden.

Mangel, W., Lin, B., & Ramakrishnan, V. (1990). Characterization of an extremely large, ligand- induced conformational change in plasminogen. Science, 248, 69-73.

Manji, B., & Kakuda, Y. (1986). Effect of Storage Temperature on Age Gelation of UHT-Milk Processed by Direct and Indirect Heating Systems. Canadian Institute of Food Science and Technology Journal, 19, xlv.

Manji, B., & Kakuda, Y. (1988). The Role of Protein Denaturation, Extent of Proteolysis, and Storage Temperature on the Mechanism of Age Gelation in a Model System. Journal of Dairy Science, 71, 1455-1463.

Markus, G., Hitt, S., Harvey, S. R., & Tritsch, G. L. (1993). Casein, a powerful enhancer of the rate of plasminogen activation. Fibrinolysis, 7, 229-236.

Markus, G. (1996). Conformational changes in plasminogen, their effect on activation, and the agents that modulate activation rates — a review. Fibrinolysis, 10, 75-85.

Marti, T., Schaller, J., Rickli, E. E., Schmid, K., Kamerling, J. P., Gerwig, G. J., van Halbeek, H., & Vliegenthart, J. F. G. (1988). The N-and O-linked carbohydrate chains of human, bovine and porcine plasminogen. European Journal of Biochemistry, 173, 57-63.

McKellar, R. C. (1981). Development of Off-Flavors in Ultra-High Temperature and Pasteurized Milk as a Function of Proteolysis1. Journal of Dairy Science, 64, 2138-2145.

McKellar, R. C., Froehlich, D. A., Butler, G., Cholette, H., & Campbell, C. (1984). The effect of uncooled storage on proteolysis bitterness and apparent viscosity in ultra-high- temperature milk. Canadian Institute of Food Science and Technology Journal, 17, 14-17.

McMahon, D. J. (1996). Age-gelation of UHT milk: Changes that occur during storage, their effect on shelf life and the mechanism by which age-gelation occurs. In Heat treatments and alternative methods (pp. 315-325). Brussels, Belgium: International Dairy Federation. 81

McMahon, D. J., & Oommen, B. S. (2013). Casein Micelle Structure, Functions, and Interactions. In P. L. H. McSweeney & P. F. Fox (Eds.), Advanced Dairy Chemistry (pp. 185-209): Springer US.

McSweeney, P. L. H., Olson, N. F., Fox, P. F., Healy, A., & Højrup, P. (1993). Proteolytic specificity of plasmin on bovine αs1‐Casein. Food Biotechnology, 7, 143-158.

Meltretter, J., Birlouez-Aragon, I., Becker, C.-M., & Pischetsrieder, M. (2009). Assessment of heat treatment of dairy products by MALDI-TOF-MS. Molecular Nutrition & Food Research, 53, 1487-1495.

Metwalli, A. A. M., de Jongh, H. H. J., & van Boekel, M. A. J. S. (1998). Heat inactivation of bovine plasmin. International Dairy Journal, 8, 47-56.

Moatsou, G., Bakopanos, C., Katharios, D., Katsaros, G., Kandarakis, I., Taoukis, P., & Politis, I. (2008). Effect of high-pressure treatment at various temperatures on indigenous proteolytic enzymes and whey protein denaturation in bovine milk. Journal of Dairy Research, 75, 262-269.

Morgan, P. E., Treweek, T. M., Lindner, R. A., Price, W. E., & Carver, J. A. (2005). Casein Proteins as Molecular Chaperones. Journal of Agricultural and Food Chemistry, 53, 2670-2683.

Nangpal, A., & Reuter, H. (1990). Reference diagram for furosine content in UHT milk. Kieler Milchw. Forsch., 45, 77-86.

Ney, K. H. (1971). Voraussage der Bitterkeit von Peptiden aus deren Aminosäurezu- sammensetzung. Zeitschrift fur Lebensmittel-Untersuchung und -Forschung, 147, 64-68.

Nguyen, N. H. A., Anema, S. G., Havea, P., Guyomarc’h, F., & Wong, M. (2012). Effect of adding low levels of β-mercaptoethanol on the disulphide bonds of κ-casein and β- lactoglobulin solutions. International Dairy Journal, 26, 78-82.

Nguyen, N. H. A., Wong, M., Havea, P., Guyomarc’h, F., & Anema, S. G. (2013). The protein interactions and rheological properties of skim milk heated in the presence of low levels of reducing agent. Food Chemistry, 138, 1604-1609.

Nielsen, P. M., Petersen, D., & Dambmann, C. (2001). Improved method for determining food protein degree of hydrolysis. Journal of Food Science, 66, 642-646.

Nielsen, S. S. (2002). Plasmin System and Microbial Proteases in Milk: Characteristics, Roles, and Relationship. Journal of Agricultural and Food Chemistry, 50, 6628-6634.

Nieuwenhuijse, J. A., & van Boekel, M. A. J. S. (2003). Protein Stability in Sterilised Milk and Milk Products. In P. F. Fox & P. L. H. McSweeney (Eds.), Advanced Dairy Chemistry—1 Proteins (pp. 947-974): Springer US.

O’Mahony, J. A., Fox, P. F., & Kelly, A. L. (2013). Indigenous Enzymes of Milk. In P. L. H. McSweeney & P. F. Fox (Eds.), Advanced Dairy Chemistry (pp. 337-385): Springer US.

Ohtsuka, K., Maekawa, I., Waki, M., & Takenaka, S. (2009). Electrochemical assay of plasmin activity and its kinetic analysis. Analytical Biochemistry, 385, 293-299.

Ozen, B. F., Hayes, K. D., & Mauer, L. J. (2003). Measurement of plasminogen concentration and differentiation of plasmin and plasminogen using Fourier-transform infrared spectroscopy. International Dairy Journal, 13, 441-446.

Paulsson, M., Hegg, P. O., & Castberg, H. B. (1985). Thermal stability of whey proteins studied by differential scanning calorimetry. Thermochimica Acta, 95, 435-440.

Perkins, M. L., & Deeth, H. C. (2001). A survey of australian consumers' attitudes towards UHT milk (Vol. 56). Melbourne, Australia: Dairy Industry Association of Australia.

Perkins, M. L., D'Arcy, B. R., Lisle, A. T., & Deeth, H. C. (2005). Solid phase microextraction of stale flavour volatiles from the headspace of UHT milk. Journal of the Science of Food and Agriculture, 85, 2421-2428.

Pinto, G., Caira, S., Cuollo, M., Fierro, O., Nicolai, M., Chianese, L., & Addeo, F. (2012). Lactosylated casein phosphopeptides as specific indicators of heated milks. Analytical and Bioanalytical Chemistry, 402, 1961-1972.

Politis, I., Lachance, E., Block, E., & Turner, J. D. (1989). Plasmin and Plasminogen in Bovine Milk: A Relationship with Involution? Journal of Dairy Science, 72, 900-906.

Politis, I., Ng Kwai Hang, K. F., & Giroux, R. N. (1989). Environmental Factors Affecting Plasmin Activity in Milk. Journal of Dairy Science, 72, 1713-1718.

Politis, I., Barbano, D. M., & Gorewit, R. C. (1992). Distribution of Plasminogen and Plasmin in Fractions of Bovine Milk. Journal of Dairy Science, 75, 1402-1410.

Politis, I., Zavizion, B., Barbano, D. M., & Gorewit, R. C. (1993). Enzymatic Assay for the Combined Determination of Plasmin Plus Plasminogen in Milk: Revisited. Journal of Dairy Science, 76, 1260-1267.

Politis, I., White, J. H., O’hare, K., Zavizion, B., Gilmore, J., & Caler, W. (1994). Distribution of Plasminogen Activator Forms in Fractions of Goat Milk. Journal of Dairy Science, 77, 2900-2906.

Politis, I., White, J. H., Zavizion, B., Goldberg, J. J., Guo, M. R., & Kindstedt, P. (1995). Effect of Individual Caseins on Plasminogen Activation by Bovine Urokinase-Type and Tissue- Type Plasminogen Activators. Journal of Dairy Science, 78, 484-490.

Politis, I. (1996). Plasminogen Activator System: Implications for Mammary Cell Growth and Involution. Journal of Dairy Science, 79, 1097-1107.

Ponting, C. P., Marshall, J. M., & Cederholm-Williams, S. A. (1992). Plasminogen: a structural review. Blood Coagulation & Fibrinolysis, 3, 605-614.

Prado, B. M., Sombers, S. E., Ismail, B., & Hayes, K. D. (2006). Effect of heat treatment on the activity of inhibitors of plasmin and plasminogen activators in milk. International Dairy Journal, 16, 593-599.

Prado, B. M., Ismail, B., Ramos, O., & Hayes, K. D. (2007). Thermal stability of plasminogen activators and plasminogen activation in heated milk. International Dairy Journal, 17, 1028-1033.

Precetti, A. S., Oria, M. P., & Nielsen, S. S. (1997). Presence in Bovine Milk of Two Protease Inhibitors of the Plasmin System. Journal of Dairy Science, 80, 1490-1496.

Recio, I., de Frutos, M., Olano, A., & Ramos, M. (1996). Protein Changes in Stored Ultra-High- Temperature-Treated Milks Studied by Capillary Electrophoresis and High-Performance Liquid Chromatography†. Journal of Agricultural and Food Chemistry, 44, 3955-3959.

Richardson, B. C., & Pearce, K. N. (1981). The determination of plasmin in dairy products. New Zealand Journal of Dairy Science and Technology, 16, 209-220.

Richardson, B. C. (1983a). The Proteinases of Bovine Milk and the Effect of Pasteurization on Their Activity. New Zealand Journal of Dairy Science and Technology, 18, 233-245.

Richardson, B. C. (1983b). Variation of the concentration of plasmin and plasminogen in bovine milk with lactation N. Z. J. Dairy Sci. Techno!. 18, 247-252.

Rippel, K. M., Nielsen, S. S., & Hayes, K. D. (2004). Effects of Native and Denatured Whey Proteins on Plasminogen Activator Activity. Journal of Dairy Science, 87, 2344-2350.

Rollema, H. S., Visser, S., & Poll, J. K. (1983). Spectrophotometric assay of plasmin and plasminogen in bovine milk. Milchwissenschaft, 38, 214-217.

Rollema, H. S., & Poll, J. K. (1986). The alkaline milk proteinase system : kinetics and mechanism of heat-inactivation. Milchwissenschaft, 41, 536-540.

Saint-Denis, T., Humbert, G., & Gaillard, J. L. (2001a). Heat inactivation of native plasmin, plasminogen and plasminogen activators in bovine milk: a revisited study. Lait, 81, 715- 729.

Saint-Denis, T., Humbert, G., & Gaillard, J. L. (2001b). Enzymatic assays for native plasmin, plasminogen and plasminogen activators in bovine milk. Journal of Dairy Research, 68, 437-449.

Santos, M. V., Ma, Y., Caplan, Z., & Barbano, D. M. (2003). Sensory Threshold of Off-Flavors Caused by Proteolysis and Lipolysis in Milk. Journal of Dairy Science, 86, 1601-1607.

Sawyer, L. (2003). β-Lactoglobulin. In P. F. Fox & P. L. H. McSweeney (Eds.), Advanced Dairy Chemistry—1 Proteins (pp. 319-386): Springer US.

Scaloni, A., Perillo, V., Franco, P., Fedele, E., Froio, R., Ferrara, L., & Bergamo, P. (2002). Characterization of heat-induced lactosylation products in caseins by immunoenzymatic and mass spectrometric methodologies. Biochimica et Biophysica Acta (BBA) - Proteins and Proteomics, 1598, 30-39.

Schaller, J., Moser, P. W., Dannegger-Muller, G. A., Rosselet, S. J., Kampfer, U., & Rickli, E. E. (1985). Complete amino acid sequence of bovine plasminogen. Comparison with human plasminogen. Eur J Biochem, 149, 267-278.

Siciliano, R. A., Mazzeo, M. F., Arena, S., Renzone, G., & Scaloni, A. (2013). Mass spectrometry for the analysis of protein lactosylation in milk products. Food Research International, 54, 988-1000.

Snoeren, T. H. M., van Riel, J. A. M., & Both, P. (1979). Proteolysis during the storage ofUHT- sterilised whole milk. I. Experiments with milk heated by the direct system for 4 seconds at 142°C. Netherlands Milk Dairy Journal, 33, 31-39.

Somers, J. M., Guinee, T. P., & Kelly, A. L. (2002). The effect of plasmin activity and cold storage of cheese milk on the composition, ripening and functionality of mozzarella-type cheese. International Journal of Dairy Technology, 55, 5-11.

Steffan, W., Balzer, H. H., Lippert, F., Sambor, B. C., Bradbury, A. G. W., & Henle, T. (2006). Characterization of casein lactosylation by capillary electrophoresis and mass spectrometry. European Food Research and Technology, 222, 467-471.

Tauzin, J., Miclo, L., Roth, S., Mollé, D., & Gaillard, J. L. (2003). Tryptic hydrolysis of bovine αS2-casein: identification and release kinetics of peptides. International Dairy Journal, 13, 15-27.

Toelstede, S., & Hofmann, T. (2008). Sensomics Mapping and Identification of the Key Bitter Metabolites in Gouda Cheese. Journal of Agricultural and Food Chemistry, 56, 2795- 2804.

Tolkach, A. (2007). Thermisches Denaturierungsverhalten von Molkenproteinfraktionen: selektive Denaturierung - Fraktionierung mit Membranen - Reaktions-, Auffaltungs- und Aggregationskinetik: Verlag Dr. Hut.

Topçu, A., Numanoğlu, E., & Saldamlı, İ. (2006). Proteolysis and storage stability of UHT milk produced in Turkey. International Dairy Journal, 16, 633-638.

Treweek, T. M., Thorn, D. C., Price, W. E., & Carver, J. A. (2011). The chaperone action of bovine milk αS1- and αS2-caseins and their associated form αS-casein. Archives of Biochemistry and Biophysics, 510, 42-52.

Ueshima, S., Okada, K., & Matsuo, O. (1996). Stabilization of plasmin by lysine derivatives. Clinica Chimica Acta, 245, 7-18.

Valero, E., Villamiel, M., Miralles, B., Sanz, J., & Martı nez-Castro, I. (2001). Changes in flavour and volatile components during storage of whole and skimmed UHT milk. Food Chemistry, 72, 51-58. van Asselt, A. J., Sweere, A. P. J., Rollema, H. S., & de Jong, P. (2008). Extreme high- temperature treatment of milk with respect to plasmin inactivation. International Dairy Journal, 18, 531-538.

Van Boekel, M. A. J. S. (1998). Effect of heating on Maillard reactions in milk. Food Chemistry, 62, 403-414.

Vassalli, J.-D., & Belin, D. (1987). Amiloride selectively inhibits the urokinase-type plasminogen activator. FEBS Letters, 214, 187-191.

Vazquez-Landaverde, P. A., Velazquez, G., Torres, J. A., & Qian, M. C. (2005). Quantitative Determination of Thermally Derived Off-Flavor Compounds in Milk Using Solid-Phase Microextraction and Gas Chromatography. Journal of Dairy Science, 88, 3764-3772.

Vazquez-Landaverde, P. A., Torres, J. A., & Qian, M. C. (2006). Quantification of Trace Volatile Sulfur Compounds in Milk by Solid-Phase Microextraction and Gas Chromatography– Pulsed Flame Photometric Detection. Journal of Dairy Science, 89, 2919-2927.

Verdi, R. J., & Barbano, D. M. (1991). Effect of Coagulants, Somatic Cell Enzymes, and Extracellular Bacterial Enzymes on Plasminogen Activation. Journal of Dairy Science, 74, 772-782.

Visser, S., Slangen, K. J., Alting, A. C., & Vreeman, H. J. (1989). Specificity of bovine plasmin in its action on bovine αS2-casein. Milchwissenschaft, 44, 335-339.

Walstra, P., Wouters, J. T. M., & Geurts, T. J. (2006). Heat Treatment. In Dairy Science and Technology (pp. 225-272). Boca Raton, USA: Taylor & Francis.

Wang, L., Hayes, K. D., & Mauer, L. J. (2006). Fluorescent Labeling Study of Plasminogen Concentration and Location in Simulated Bovine Milk Systems. Journal of Dairy Science, 89, 58-70.

Wang, Q., Tolkach, A., & Kulozik, U. (2006). Quantitative Assessment of Thermal Denaturation of Bovine α-Lactalbumin via Low-Intensity Ultrasound, HPLC, and DSC. Journal of Agricultural and Food Chemistry, 54, 6501-6506.

Wang, Q., & Ismail, B. (2012). Effect of Maillard-induced glycosylation on the nutritional quality, solubility, thermal stability and molecular configuration of whey proteinv. International Dairy Journal, 25, 112-122.

White, J. H., Zavizion, B., O'Hare, K., Gilmore, J., Guo, M. R., Kindstedt, P., & Politis, I. (1995). Distribution of plasminogen activator in different fractions of bovine milk. Journal of Dairy Research, 62, 115-122.

Wiman, B., & Collen, D. (1978). On the Kinetics of the Reaction between Human Antiplasmin and Plasmin. European Journal of Biochemistry, 84, 573-578.

Wu, T. P., Padmanabhan, K., Tulinsky, A., & Mulichak, A. M. (1991). The refined structure of the epsilon-aminocaproic acid complex of human plasminogen kringle 4. Biochemistry, 30, 10589-10594.

Yeboah, F. K., & Yaylayan, V. A. (2001). Analysis of glycated proteins by mass spectrometric techniques: qualitative and quantitative aspects. Food / Nahrung, 45, 164-171.

Zabbia, A., Buys, E. M., & De Kock, H. L. (2011). Undesirable Sulphur and Carbonyl Flavor Compounds in UHT Milk: A Review. Critical Reviews in Food Science and Nutrition, 52, 21-30.

Zhang, X., Fu, X., Zhang, H., Liu, C., Jiao, W., & Chang, Z. (2005). Chaperone-like activity of β- casein. The International Journal of Biochemistry & Cell Biology, 37, 1232-1240.

Zin El-Din, M., & Aoki, T. (1993). Polymerization of casein on heating milk. International Dairy Journal, 3, 581-588.

Aaltonen, T., & Ollikainen, P. (2011). Effect of microfiltration of milk on plasmin activity. International Dairy Journal, 21, 193-197.

Paper I

Rauh, V. M., Bakman, M., Ipsen, R., Paulsson, M., Kelly, A. L., Larsen, L. B., & Hammershøj, M. (2014)

The determination of plasmin and plasminogen-derived activity in turbid samples from various dairy products using an optimised spectrophotometric method

International Dairy Journal, 38, 74-80

International Dairy Journal 38 (2014) 74e80

Contents lists available at ScienceDirect

International Dairy Journal

journal homepage: www.elsevier.com/locate/idairyj

The determination of plasmin and plasminogen-derived activity in turbid samples from various dairy products using an optimised spectrophotometric method

Valentin M. Rauh a,b,*, Mette Bakman a, Richard Ipsen c, Marie Paulsson d, Alan L. Kelly e, Lotte B. Larsen b, Marianne Hammershøj b a Arla Foods Strategic Innovation Centre, Rørdrumvej 2, DK-8220 Brabrand, Denmark b Aarhus University, Dept. of Food Science, Faculty of Science and Technology, Blichers Allé 20, DK-8830 Tjele, Denmark c University of Copenhagen, Department of Food Science, Rolighedsvej 30, DK-1958 Frederiksberg C, Denmark d Department of Food Technology, Engineering and Nutrition, Lund University, SE-221 00 Lund, Sweden e School of Food and Nutritional Sciences, University College Cork, Cork, Ireland article info abstract

Article history: A spectrophotometric assay for plasmin and plasminogen-derived activity in dairy products was opti- Received 13 June 2013 mised and extended to determine plasmin and plasminogen-derived activity in turbid samples of dairy Received in revised form products. The method was validated by assessing reproducibility, repeatability, level of detection and 24 March 2014 recovery of plasmin activity in different sample matrices. Plasmin activity in raw milk was not affected by Accepted 26 March 2014 skimming, but decreased by 30% in pasteurised and homogenised whole milk, leading to an underes- Available online 26 April 2014 timation of plasmin activity. The effects of dissociation of plasmin and caseins by ε-aminocaproic acid (EACA) plus NaCl on the plasmin activity were investigated. Comparison of pasteurised milk with a micellar casein solution showed that the dissociation of plasmin and caseins on adding EACA and NaCl decreases interference by caseins, but increases inhibition of plasmin with serum-based inhibitory components. The level of detection and repeatability of this method for plasmin activity analysis were improved compared with previous spectrophotometric assays. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction PL readily hydrolyses b- and aS2-casein, and, to a lesser extent, aS1- casein (Grufferty & Fox, 1988). PL activity may be linked to the The major indigenous proteinase in milk, plasmin (PL; EC formation of unclean and bitter off-flavours in milk, and age gela- 3.4.21.7), is the active part of a complex enzyme system. Its tion of UHT milk (Harwalkar, Cholette, McKellar, & Emmons, 1993; zymogen, plasminogen (PLG), is present in raw, bovine milk at a 2e Kelly & Foley, 1997). On the other hand, hydrolysis of caseins by PL 30 times higher concentration than PL (Ismail & Nielsen, 2010), and plays an important role in the initial ripening of Swiss cheese types is activated by two families of plasminogen activators (PA), (Bastian & Brown, 1996). urokinase-type PA (uPA) and tissue-type PA (tPA). The activity of PL Various methods to measure PL and PLG-derived activities have and PA is regulated by plasmin inhibitors (PIs), such as a2-anti- been described (Politis, Zavision, Barbano, & Gorewit, 1993; plasmin, and plasminogen activator inhibitors (Grufferty & Fox, Richardson & Pearce, 1981; Rollema, Visser, & Poll, 1983; Saint- 1988; Ismail & Nielsen, 2010; Korycka-Dahl, Dumas, Chene, & Denis, Humbert, & Gaillard, 2001). The determination of PL activ- Martal, 1983; Lu & Nielsen, 1993; Precetti, Oria, & Nielsen, 1997). ity in these assays is based on the hydrolysis of a specific substrate, The level of PL in milk can vary and depends on environmental which upon cleavage releases a chromogenic (Rollema et al., 1983) factors, such as stage of lactation and somatic cell count (Larsen or fluorogenic product (Richardson & Pearce, 1981; Saint-Denis et al., 2010; Nicholas, Auldist, Molan, Stelwagen, & Prosser, 2002). et al., 2001). The different sample preparations used in these assays, however, cause varying results and different interfering effects have been reported. Two substrates (S-2251 and Spectrozyme PL) have been reported to enhance PLG activation (Kolev, Owen, & * þ Corresponding author. Tel.: 45 87332783. ε E-mail address: [email protected] (V.M. Rauh). Machovich, 1995), while -aminocaproic acid (EACA), a lysine http://dx.doi.org/10.1016/j.idairyj.2014.03.008 0958-6946/Ó 2014 Elsevier Ltd. All rights reserved. V.M. Rauh et al. / International Dairy Journal 38 (2014) 74e80 75 derivative often used to enhance PL activity, inhibits activation of casein pellet was reconstituted with UF permeate to original volume human PLG (Alkjaersig, Fletcher, & Sherry, 1959; Cesarman-Maus & and gently stirred at 4 C overnight. Samples of Swedish Herrgård Hajjar, 2005). EACA dissociates PL and PLG from caseins by binding cheese (>28% fat, >6 months) andparmigiano Reggiano (>32% fat, >6 to the lysine-binding sites on PL and PLG (Korycka-Dahl et al.,1983). months) were obtained from a local supermarket. Danish Havarti Furthermore, it has been shown that the dissociation of human cheese (>60% fat, 1 week) was produced in the pilot plant at Arla PLG by EACA also inhibits activation of PLG by tPA and uPA Foods Strategic Innovation Center (Brabrand, Denmark). (Cesarman-Maus & Hajjar, 2005; Collen, 1987; Sun et al., 2002). On the other hand, the natural substrate of PL in milk, the caseins, can 2.3. Basic sample preparation and measurement of PL and PLG- act as competitive inhibitors towards a chromogenic substrate used derived activity in the assays (Bastian, Brown, & Ernstrom, 1991a) and affect the activation of PLG (Heegaard, Andreasen, Petersen, & Rasmussen, Milk samples (1 mL) were mixed with 250 mL 0.4 M tri-sodium 1997; Markus, Hitt, Harvey, & Tritsch, 1993). Whey proteins are citrate buffer, pH 8.9, and shaken for 15 min to dissociate the also able to inhibit PL activity (Hayes, McSweeney, & Kelly, 2002; casein micelles. Sample preparation for cheese was adapted from Politis et al., 1993), but the exact mechanism has not been that of Richardson and Pearce (1981) with modifications as follows. described so far; b-lactoglobulin and bovine serum albumin have A mass of 5 g of grated cheese was dissolved in 45 mL 0.4 M tri- shown a higher inhibitory effect than a-lactalbumin (Politis et al., sodium citrate buffer, pH 8.9, and stirred for 30 min at room tem- 1993). perature, followed by holding for 30 min at 45 C, and finally, Another consideration in terms of assay suitability and perfor- 15 min at room temperature to separate the fat phase. The solution mance is that, especially for fluorescence assays, the turbidity of was then centrifuged at 7500 g for 15 min, and the supernatant milk can severely interfere with the measurement. In most assays below the cream phase was removed and centrifuged again under described, an extensive sample preparation protocol is needed to the same conditions. avoid these interferences (Politis et al., 1993; Rollema et al., 1983). The citrate-treated milk or cheese samples were diluted 1:1 (v/ In case of fluorescence methods, a high dilution of the sample v) with an assay buffer as described by Saint-Denis et al. (2001) (Richardson & Pearce, 1981) or an additional clarification treatment containing 0.1 M TriseHCl, 8 mM EACA, 0.4 M NaCl, pH 8, and (clarifying reagentÔ) before measurement (Saint-Denis et al., mixed for 15 min to dissociate PL and PLG from the caseins. 2001) is required. These preparations may not only reduce the For determination of PLG-derived activity, PLG in the citrate- sensitivity of the assay, but could also further complicate the treated samples was activated by a 1:1 (v/v) dilution with an comparison of PL activities measured by different sample prepa- assay buffer containing 200 Plough units mL 1 urokinase and rations, substrates and detection methods (Kelly, O’Flaherty, & Fox, incubated for 60 min at 37 C(Korycka-Dahl et al., 1983; Rollema 2006). et al., 1983; Saint-Denis et al., 2001). The aim of this study was to combine the spectrophotometric PL and PLG-derived activity were determined by measuring the assay for PL and PLG-derived activity described by Rollema et al. rate of hydrolysis of the chromogenic substrate S-2251 (Rollema (1983) with the sample preparation proposed by Saint-Denis et al., 1983) and release of p-nitroaniline. To compensate for et al. (2001) and expand it to turbid, fat-containing milk types lower dilution of 1:5 v/v the sample in the final mixture compared and different sample matrices including cheeses. A further aim was with 1:10 (v/v) in the assay described by Rollema et al. (1983) and, to study the effect of skimming on PL activity in different milk types to increase the sensitivity of the assay, the substrate concentration and to characterise the inhibitory effects of casein and serum was increased from 0.6 mM S-2251 (Rollema et al., 1983)to2mM in -based components on the assay, especially the effect of dissocia- the final mixture. The substrate, S-2251, was dissolved in 0.1 M Trise tion of PL and caseins by EACA and NaCl on PL and PLG-derived HCl buffer, pH 8 to a concentration of 4 mM. activity in different sample types. The sample solution (50 mL) was transferred to a microtitre plate and the reaction was initiated by addition of 50 mL of the substrate 2. Materials and methods solution. Absorbance was read at 405 nm and 490 nm at 37 C using an ELISA plate reader (Bio-Tek EL 808 and Bio-Tek Synergy Mx, 2.1. Materials BioTek Instruments Inc., Wisconsin, IL, USA) at intervals of 2e 10 min for 60e180 min depending on the level of PL activity in the PL from bovine serum was obtained from Roche Diagnostics sample. Substrate volumes per sample were minimised through (Hvidovre, Denmark) and urokinase (Thrombolysin) was from use of Costar 3695 clear half-area polystyrene 96-well plates Immuno Danmark A/S (Copenhagen, Denmark). The ε-amino- (Corning Inc., New York, NY, USA). To correct for turbidity, the caproic acid (EACA) was obtained from SigmaeAldrich Denmark background absorbance values (490 nm) were subtracted from the ApS (Brøndby, Denmark) and H-D-valyl-L-leucyl-L-lysyl-4- absorbance values at 405 nm, corresponding to colour develop- nitroanilide (S-2251) from Haemochrom Diagnostica GmbH ment due to release of p-nitroaniline. Correction for turbidity using (Essen, Germany). the absorbances at 540 and 630 nm was also tested. The increase in absorbance, indicated as DA405 nme490 nm as a function of time, dA/ 2.2. Origin of samples dt, was converted into PL units using a standard curve prepared with commercial PL solution from a PL stock solution (0.5 U in Tris The pasteurised (minimum of 72 C for 15 s) skim milk (SM), buffer, pH 8) with assay buffer additionally containing 4 mg mL 1 homogenised and pasteurised whole milk with 3.5% fat (WM), and gelatine to stabilise PL. The correlation between PL activity and UHT milk (0.5% fat) used in this study were commercially available absorbance was linear in the range 15 mUmL 1 to 1 mU mL 1 milk products obtained from a local supermarket, stored at 5 Cand (R2 ¼ 0.999). No increase over time in absorbance was observed in a analysed before expiry date. Raw bulk milk (RM) was obtained from blank sample containing assay buffer with 4 mg mL 1 gelatine only. Arla Foods (Brabrand Dairy Plant, Brabrand, Denmark). A whey- protein-free milk serum (UF-permeate) was prepared by ultrafiltra- 2.4. Method validation tion of SM using 5-kDa spiral wound membranes (Koch Membrane Systems HFK-328). A micellar casein solution (MCS) was prepared by The linearity of the assay was evaluated by measuring PL activity ultracentrifugation of SM at 100,000 g for 60 min at 25 C. The in undiluted SM as well as 1:1 and 1:10 (v/v) dilutions of SM and 3 76 V.M. Rauh et al. / International Dairy Journal 38 (2014) 74e80 different wavelength to correct for turbidity (490, 540 and 630 nm). been shown previously to dissociate PL from casein micelles (Saint- The wavelength (490, 540 or 630 nm) used for turbidity correction Denis et al., 2001). had no significant effect on the slope, dA/dt, of the curves, for The effect of EACA þ NaCl on PL and PLG-derived activities in RM reference, 490 nm was thus selected. Linearity and turbidity of the and SM was evaluated by diluting the sodium-citrate-treated assay for fat containing milks was assessed with WM and RM using samples (1:1 v/v) with either 0.1 M TriseHCl, pH 8, or 0.1 M Trise absorbance at 490 nm to correct for turbidity. HCl, pH 8, containing 8 mM EACA and 0.4 M NaCl. Repeatability of the assay was evaluated by measuring PL and The effect of pre-incubation with EACA þ NaCl on PL activity was PLG-derived activity of a SM sample for a total of 14 times. compared for a SM and a MCS sample prepared from the same SM. Reproducibility was assessed by measuring PL and PLG-derived A dilution series of SM and MCS was prepared by addition of 0.1 M activity in frozen aliquots of SM, RM, Danish Havarti cheese, TriseHCl buffer, pH 8, to a final concentration of SM or MCS of 100, Swedish Herrgård cheese and Parmigiano Reggiano on 4 occasions 80, 60, 40 and 20% (v/v). From each of these dilutions, 1 mL was (measured on 4 different plates on 4 different days). taken and a volume of 250 mL 0.4 M tri-sodium citrate buffer, pH 8.9 Limit of detection was compared with that determined by Saint- was added. Each sample was shaken for 15 min before further Denis et al. (2001) by diluting SM with 0.1 M TriseHCl, pH 8. The limit dilution (1:1, v/v) with either 0.1 M TriseHCl, pH 8, or 0.1 M Trise of detection was the highest dilution of SM with a significantly higher HCl, pH 8, containing 8 mM EACA and 0.4 M NaCl. PL activity was activity than a blank consisting of assay buffer with 4 mg mL 1 gelatine. measured as described in Section 2.3.

2.5. Effect of skimming on plasmin activity in milk 2.8. Statistical analysis

The effect of fat removal by centrifugation on the development Analysis of variance (ANOVA) and separation of means was of absorbance in the PL assay was investigated using RM and WM carried out using MATLAB (Mathworks, Natick, MA, USA). One- with SM as control. Prior to skimming, a volume of 2.5 mL 0.4 M tri- way-ANOVA was used to determine significant differences sodium citrate buffer, pH 8.9 was added to 10 mL of RM, WM or SM (P 0.05) among groups. Differences between the means and mean and the solutions were mixed for 15 min. One mL of each solution separation were determined using TukeyeKramer multiple means was taken out and stored at 4 C to ensure the same experimental comparison tests (P 0.05 and 0.01). conditions, while the remaining volume was further prepared for skimming by centrifugation at 10,000 g for 10 min at 4 C. The 3. Results and discussion upper fat-containing layer was discarded and 1 mL of the clear liquid was used for further analysis. Both the un-skimmed and 3.1. Method validation skimmed samples were further diluted 1:1 (v/v) 0.1 M TriseHCl, 8mM EACA, 0.4 M NaCl, pH 8 and PL activity was measured as The linearity of the assay was evaluated by measuring PL activity described in Section 2.1. Three independent batches of each milk in undiluted SM as well as 1:1 and 1:10 (v/v) dilutions of SM with (RM, WM and SM) were used for the experiment and sample 0.1 M TriseHCl buffer, pH 8. The reference wavelength measure- preparation was performed in duplicate. ments showed that the turbidity in undiluted and diluted SM samples was not constant throughout the assay. Within the first 2.6. Recovery of activity of added PL and inhibitory effects of milk 20 min, the turbidity decreased rapidly in an exponential manner, components which led to a non-linear increase in the DA405 nme490 nm in that time period (Fig. 1), and this increase appeared to be independent Recovered activity of PL was measured by addition of a defined of sample concentration. The first 20 min of the measurements amount of PL with known activity to samples, along with the assay were therefore excluded from the calculations of the slope. e buffer containing 0.1 M Tris HCl, 8 mM EACA, 0.4 M NaCl, pH 8, to Following this, the absorbance, DA405 nme490 nm increased linearly investigate possible inhibitory and interfering effects in the assay. (20e120 min) in both the diluted and undiluted SM samples Samples tested were RM, SM, UHT milk, MCS as well as Havarti, (R2 > 0.99). Herrgård and Parmigiano Reggiano cheese. Furthermore, to provide a reference system, which does not contain the PL inhibitors pre- 1 sent in milk, a gelatine solution (4 mg mL ) containing 0.1 M Trise HCl buffer, pH 8 was prepared. Gelatine has earlier been shown to be a suitable agent for stabilising PL (Saint-Denis et al., 2001). The measured activity of PL in the gelatine system was set as 100% recovered activity. The effect of sample concentration on recovered activity and inhibitory effects was evaluated by diluting a UHT milk sample with 0.1 M TriseHCl buffer, pH 8, to a final concentration of UHT milk of 100, 75, 50 and 25% (v/v) and a constant amount of added PL. PL activity in the UHT milk dilution was compared with PL activity in dilutions of a SM sample to which UHT milk had been added to a final proportion of SM in the mixture of 100, 75, 50 and 25% (v/v). PL activity was measured as described in Section 2.3.

2.7. Effect of dissociation of plasmin by EACA and NaCl in presence and absence of whey proteins on PL and PLG-derived activity Fig. 1. Plasmin activity measured as development in absorbance DA405 nme490 nm due to substrate hydrolysis and release of 4-nitroaniline in undiluted (B), 1:1 (,) and 1:10 To further investigate the effect of milk components, PL was (7) dilution of pasteurised skim milk with 0.1 M TriseHCl buffer, pH 8 (average of 3 þ dissociated from casein by the use of EACA and NaCl (EACA NaCl). determinations standard deviation). Lines are linear ﬁts for the time period 20e EACA þ NaCl at concentrations of 4 mM and 2 M, respectively, have 120 min to guide the eye. V.M. Rauh et al. / International Dairy Journal 38 (2014) 74e80 77

The development of absorbance at 405 nm and at 490 nm in RM, Table 1 m 1 a WM and SM are shown in Fig. 2A. The assay with RM was only Plasmin and plasminogen-derived activity ( UmL ) in milk and cheese samples. stable up to 2 h, after which time creaming occurred, and no further Sample Plasmin activity Plasminogen-derived measurement was possible. Due to the presence of fat, the initial activity absorbance at 405 and 490 nm of RM and WM was 4e5 times Raw milk 390 47 3203 130 higher compared with SM. The absorbance at 405 nm in SM and Pasteurised skim milk 462 13 3259 37 WM increased slightly during the incubation time, but decreased in Havarti (1 week old) 4095 167 46,940 600 > RM, which may be due to the onset of creaming. At the same time, Herrgård ( 6 months) 10,225 309 25,841 958 Parmigiano Reggiano 8579 228 25,728 255 the absorbance at 490 nm decreased for both RM and SM, while it a Results are shown as means standard deviation of 4 determinations on 4 remained almost constant for SM. Fig. 2B shows the development 1 1 D occasions. PL and PLG derived activity are shown in mUmL for milk and in mUg in absorbance A405 nme490 nm after correction for turbidity. As can for cheeses. be seen, the absorbance increased linearly in all milk samples, and there was only a minor difference in the initial absorbance. The with previous chromogenic methods, where the level of detection correction for turbidity using a reference wavelength at 490 nm was reported by Rollema et al. (1983) to be between 6 and 10% of PL thus enables measurements in turbid samples. activities in bulk milk, but does not reach the level of detection of Reproducibility of the assay and performance in different sample fluorogenic methods (0.1% according to Saint-Denis et al., 2001). matrices was assessed by measurement of PL and PLG-derived activity in frozen aliquots of RM, SM,1 week old Danish Havarti cheese, 3.2. Effect of skimming by centrifugation on plasmin activity in milk Herrgård cheese and Parmigiano Reggiano on 4 occasions (Table 1). SM showed a higher PL activity than RM, while the PLG-derived Previous methods for measurement of PL activity have been activity was similar in SM and RM. The PLG/PL ratio was 8.2 for RM based on measurements in skimmed milk (Rollema et al., 1983), and 7.1 for SM, which is in the reported range for milk (Ismail & and have either included a centrifugation step to remove the cream Nielsen, 2010). Due to the fat content, the reproducibility in RM phase (Politis et al., 1993; Richardson & Pearce, 1981) or do not was poorer compared with SM, with a relative standard deviation of specify the fat content of the milk tested (Saint-Denis et al., 2001). 12.1% for PL activity and 4.1% for PLG-derived activity, compared To investigate the effect of skimming on the development of with 2.8% and 1.1% for SM. PLG-derived activity in the Havarti cheese absorbance in the assay, samples of WM and RM were analysed was approximately 11.5 times higher than the PL activity and was before and after skimming, with SM as a control. approximately 2 times higher than the PL activity in the Herrgård Table 2 shows the measured PL activities before and after skim- cheese (Table 1). The PLG to PL ratio of the Havarti cheese is within ming and the percentage difference in PL activity due to skimming. the range of previously reported values of approximately 3 The PL activity in un-skimmed WM was w25% higher compared with (Benfeldt, 2006)to21(Bastian, Hansen, & Brown, 1991b). the SM, but decreased significantly by almost 30% after skimming To our knowledge, PL and PLG-derived activity in Herrgård (P 0.01), to a similar level as in the SM, while skimming had no cheese has not been described previously. PLG-derived activity in significant effect on the PL activity in RM and SM (P 0.05). The Parmigiano Reggiano was similar to that in the Herrgård cheese, measurements in RM provide further proof that the described while PL activity was lower. The relative standard deviations of the method is suitable for measurement in turbid fat-containing samples. PL and PLG-derived activity in the cheeses (Table 1) were from 3.0 The decrease in PL activity in WM after skimming could be to 4.1% for PL activity and from 1.3 to 3.7% for PLG-derived activity. attributed to the removal of PL associated with caseins on the milk In comparison, Richardson and Pearce (1981) reported a repro- fat globules. The native milk fat globule membrane fraction of raw ducibility of 9% for a Swiss-type cheese (n ¼ 9). Repeatability of the milk contains a low level of PL activity, probably due to casein assay was assessed by measuring a SM sample 14 times. The contamination (Benfeldt, Larsen, Rasmussen, Andreasen, & repeatability, measured as % relative standard deviation, for PL and Petersen, 1995; Politis, Barbano, & Gorewit, 1992). Upon homoge- PLG-derived activity was 2% and 1%, respectively, compared with nisation, the composition of the milk fat globule membrane 2% reported for PL activity by Saint-Denis et al. (2001) and 10% (MFGM) changes and up to 70% of the reformed membrane consists reported for PL activity by Rollema et al. (1983). of caseins and casein micellar fragments (Cano-Ruiz & Richter, The lowest PL activity significantly higher than a blank sample 1997). The high initial PL activity and the decrease in PL activity (P < 0.05) was detected in a 100-fold diluted sample of SM, indicating in the WM upon skimming indicates that a large portion of PL could an overall limit of detection of 1%. This is an improvement compared be associated with the caseins in the MFGM after homogenisation

Fig. 2. Development of (A) absorbance at 405 (open symbols) and 490 nm (ﬁlled symbols) in raw milk (B,C), whole pasteurised milk (,,-) and pasteurised skim milk (>,A) and of (B) DA405 nme490 nm. 78 V.M. Rauh et al. / International Dairy Journal 38 (2014) 74e80

Table 2 Table 3 PL activity in three independent batches (A-C) of pasteurised, homogenised whole Recovered activity of added PL in different milk and cheese types.a milk with 3.5% fat (WM), raw bulk milk (RM) and pasteurised skim milk (SM) before and after skimming by centrifugation.a Sample Recovered activity (%) Raw milk 38.8 2.9a Milk sample Before skimming After skimming Percentage activity b Pasteurised skim milk 43.8 5.3 (mUmL 1) (mUmL 1) after skimming UHT milk 50.3 2.7c WM Micellar casein 66.9 1.5d A 703 33 509 10 72.4 5.1 Havarti 70.4 1.8a B 683 35 491 5 71.9 5.2 Herrgård 64.8 0.8b C 679 24 482 2 70.9 3.6 Parmigiano Reggiano 65.5 3.3b RM a Results are shown as means of 3 determinations standard deviation; A 558 6 561 3 100.4 1.1 values without common superscripts are significantly different (P 0.05). B 636 9 694 3 109.2 1.9 C 712 18 646 10 90.8 2.9 SM The recovered activity of added PL in SM, RM, MCS and UHT A 549 15 523 8 95.1 3.2 milk, and Danish Havarti, Swedish Herrgård and Parmigiano B 534 8 552 9 103.3 2.1 Reggiano cheese, are shown in Table 3. The recovered activity in the C 558 15 563 1 101.0 2.7 milk samples was correlated with the heat treatment of the sam- a Results are means of 2 determinations standard deviation. ples with RM showing the lowest recovered activity and UHT milk the highest. The increase in recovered activity with increasing heat and explain the similar PL activities of the skimmed WM and SM. treatment is most likely due to inactivation of PI, which has been The higher PL activity in WM compared with SM is in contrast to shown to be heat-labile (Prado, Sombers, Ismail, & Hayes, 2006). fi earlier ndings, which reported that 40% of PL is inactivated during The recovery rate in MCS was only approximately 67%, due to a standard single-stage homogenisation of milk (Hayes & Kelly, competitive inhibition of the synthetic substrate by caseins (Bastian fi 2003). Initial experiments to con rm PL activity in the reformed et al., 1991a). Compared with the milk samples, MCS showed a 16e fi MFGM of homogenised cream were not possible due to dif culty of 29% higher recovered activity, reflecting the inhibitory effect of dispersing of the washed cream in the presently used buffer sys- serum-based PI and whey proteins on PL activity. tems. In conclusion, skimming and centrifugation of the sample Bastian et al. (1991a) showed that activity of PL towards S-2251 prior to analysis facilitates the measurement of PL in RM samples, is competitively inhibited by caseins and, by increasing the sub- but leads to an underestimation of the PL activity in homogenised strate concentration to 0.4 mM, eliminated the inhibitory effect of milk. 5mgmL 1 casein. In this assay, a casein concentration of approximately 2.6 mg mL 1 in the MCS and >4 times higher substrate 3.3. Recovery of activity of added PL and inhibitory effects of milk concentration compared with those used by Bastian et al. (1991a) components had an inhibitory effect on the PL activity. The differences in these results may arise from the source of PL and higher PL con- Fig. 3A shows the PL activity in a dilution of SM with UHT milk centrations used for the experiment. In this study, the activity of the and the activity of a fixed amount of PL added to a dilution of UHT indigenous PL associated with the casein fraction was measured, with buffer. PL activity in the dilution of SM with UHT milk whereas Bastian et al. (1991a) used commercial PL from bovine decreased linearly with increasing proportion of UHT milk added plasma (final concentration 26.8 nM) at a concentration 3e17 times with the same casein and whey protein concentration. Lowering higher than that present in milk (1.6e8.2 nM). the whey protein and casein concentration increased measured PL Recovered activity in cheeses was similar to that for MCS, activity, as can be seen from the increase in activity of the added PL although the protein concentration in the final mixture was in UHT milk with decreasing proportion of added UHT milk. The approximately double that of the milk samples. Protein content in recovered activity of the added PL in UHT milk increased from the cheeses varied due to the removal of fat and the recovered 48.1 1.3% in the sample containing UHT milk only to 71.2 0.7% in activity in the cheese samples increased with increasing fat con- the dilution containing 25% UHT milk (Fig. 3B), which shows the tent. Although the fat content in the Danish Havarti cheese was inhibitory and interfering effects of caseins and denatured whey approximately twice as high as for the other cheese types, the proteins (Bastian, Hansen, & Brown, 1993; Hayes et al., 2002) on the recovered activity was only slightly higher. The Havarti cheese was PL activity. only one week old and the hydrolysis of caseins during the ripening

Fig. 3. PL activity (A) in pasteurised skim milk diluted with UHT milk without detectable PL activity (B) and activity of added PL in UHT milk (0.5% fat; >) diluted with 0.1 M Trise HCl buffer, pH 8; percentage (B) of recovered activity of added PL to a dilution series from 0 to 100% of commercial UHT milk with 0.1 M TriseHCl buffer, pH 8. Results are shown as average of 3 determinations standard deviation. V.M. Rauh et al. / International Dairy Journal 38 (2014) 74e80 79 of the cheeses could cause a reduced inhibitory effect on the syn- Table 4 thetic substrate as they could be too small to block the active centre Effect of dissociation of PL and caseins by EACA plus NaCl on PL and PLG derived activity in skimmed raw milk and pasteurised skim milk.a or not have a cleavage site for PL and thus not bind at all. Hydro- lysed caseins and the resulting peptides may have a reduced Sample Plasmin Plasminogen-derived Plasminogen/plasmin activity activity (mUmL 1) ratio inhibitory effect on the synthetic substrate. (mUmL 1) The recovered activities for milk and micellar casein in Table 3 a are in the same range as previously reported results. Saint-Denis Raw milk þ EACA 438 4 n.d. et al. (2001) reported a recovery of 30e40% in diluted (1:4, v/v) and NaCl Raw milk EACA 443 6a n.d. pasteurised milk, while Richardson and Pearce (1981) reported a and NaCl recovery of 40% in a 1:16 (v/v) diluted 3% solution of sodium Pasteurised skim 459 6b 3367 87a 7.3 0.2a caseinate. This shows that the described method has a similar milk þ EACA sensitivity and sufficient substrate concentration to compensate and NaCl Pasteurised skim 422 6c 3036 47b 7.2 0.2a for inhibitory effects compared with previously described milk EACA methods, although a direct comparison is difficult due to the and NaCl different buffers, dilutions and samples used for the a Results are shown as mean 3 determinations standard deviation (n.d., not determination. determined); values without common superscripts for each row are significantly different (P < 0.05). 3.4. Effect of dissociation of plasmin by EACA and NaCl in the presence and absence of whey proteins on PL and PLG-derived activity containing 20% SM, showing that dilution of serum-based inhibitors increases PL activity. For the SM samples, on the other PL in milk is known to be associated with caseins via lysine- hand, the relatively small difference in PL activity compared with binding sites that are not affected by addition of trisodium citrate the MCS samples indicates that the interfering effect of the caseins (Politis et al., 1993; Rollema et al., 1983). EACA has previously been is overshadowed by a change in the interaction of PL with serum- shown to dissociate PL and PLG from caseins by binding to the based inhibitory components (e.g., whey proteins). With lysine-binding sites of PL and PLG, and enhance measured PL ac- increasing sample dilution, the percentage difference between tivity (Korycka-Dahl et al., 1983). The lysine-binding sites of PL are samples with free and bound PL increased from an initial of 8% in also of great importance for the physiological function of human PL the 100% SM to 19% in the 20% SM. This shows that the inhibitory in blood and the fibrinolytic system. Binding of PL to fibrin ensures effect of serum-based components decreases with increasing a close proximity of the enzyme to its substrate and also protects PL dilution of SM. Since the interaction of PL and a -antiplasmin is reduced in from irreversible inhibition by a2-antiplasmin, a major indigenous 2 PL inhibitor, which requires a free lysine-binding site and a free presence of EACA (Christensen, Sottrup-Jensen, & Christensen, catalytic centre of PL to interact and inhibit PL (Cesarman-Maus & 1995) and pasteurisation conditions largely inactivate PIs (Prado Hajjar, 2005; Collen, 1987). The dissociation of PL and caseins et al., 2006), the observed inhibition is most likely due to an could therefore reduce the competitive inhibition of the synthetic enhanced inhibition of PL by whey proteins. This protective effect of substrate by caseins and affect the interaction of PL with PI and caseins towards inhibition of PL by whey proteins in milk has not to whey proteins. our knowledge been documented previously, although this was The dissociation of PL and caseins by EACA þ NaCl increased PL suggested by Politis et al. (1993). Whether the dissociation of PL activity in SM significantly (P > 0.05) by approximately 9% and 11% and caseins also enhances the inhibition of PL in a native milk for PL and PLG-derived activity, respectively (Table 4). The PLG/PL environment and reduces proteolysis of casein was not the focus of ratio of the free and bound SM was not significantly different, this study, but should be further investigated to allow a correlation indicating that EACA þ NaCl did not affect the activation of PLG by of proteolysis and PL activity. uPA under the chosen assay conditions. The dissociation of PL did not lead to a significant increase in PL activity in RM, suggesting that the dissociation increased inhibition of PL by serum-based inhibitors. To further investigate the inhibitory effect of serum-based inhibitory components on free and bound PL, EACA and NaCl were added together to a dilution series of SM and MCS. Fig. 4 shows measured PL activity as a function of the percentage of SM or MCS in the dilutions in the presence or absence of EACA þ NaCl. The PL activity in the MCS did not decrease proportionally with the casein content. This suggests that the inhibitory effect of casein increases with concentration, as previously reported by Bastian et al. (1991a). Dissociation of PL and caseins in the undiluted MCS increased PL activity by 44%. This shows that the dissociation of PL and caseins greatly reduces the competitive inhibition of the synthetic substrate by caseins in the assay. The difference in PL activity between free and bound PL in MCS was constant in all dilutions, indicating that the same competitive inhibition applies at different casein concentrations. Fig. 4. Effect of adding EACA þ NaCl on the plasmin activity in SM (,,-) and a Compared with MCS, the increase in PL activity upon dissoci- micellar casein solution (B,C). Milk and the micellar casein solution were diluted with TriseHCl buffer, pH 8. The samples were mixed 1:1 with TriseHCl buffer without ation of PL and caseins in SM was less pronounced. In contrast to (dashed line, filled symbols) or with 6 mM EACA and 0.4 M NaCl (solid line, open the MCS, the difference in PL activity increased with increasing symbols) and assayed for plasmin activity (average of 3 determinations standard dilution from 9% in the undiluted SM to 24% in the sample deviation). 80 V.M. Rauh et al. / International Dairy Journal 38 (2014) 74e80

4. Conclusion Grufferty, M. B., & Fox, P. F. (1988). Milk alkaline proteinase. Journal of Dairy Research, 55, 609e630. Harwalkar, V. R., Cholette, H., McKellar, R. C., & Emmons, D. B. (1993). Relation The method presented enables the measurement of PL and PLG- between proteolysis and astringent off-flavor in milk. Journal of Dairy Science, derived activity in turbid samples, such as homogenised whole 76, 2521e2527. milk, micellar casein solutions and cheese extracts, using a micro- Hayes, M. G., & Kelly, A. L. (2003). High pressure homogenisation of milk (b) effects on indigenous enzymatic activity. Journal of Dairy Research, 70, 307e313. titre plate real-time kinetic method with high substrate. The per- Hayes, M. G., McSweeney, P. L. H., & Kelly, A. L. (2002). The influence of native and formance of the assay was determined in relation to limit of heat-denatured whey proteins on enzyme activity. 1. Plasmin. Milchwissen- detection, repeatability, reproducibility, recovery of added PL and schaft, 57, 208e211. Heegaard, C. W., Andreasen, P. A., Petersen, T. E., & Rasmussen, L. K. (1997). Binding the effect of milk concentration on the linearity of the assay. When of plasminogen and tissue-type plasminogen activator to dimeric as2-casein pasteurised, homogenised milk was centrifuged, approximately accelerates plasmin generation. Fibrinolysis and Proteolysis, 11,29e36. 30% of the PL activity was removed, in contrast to raw, whole milk, Ismail, B., & Nielsen, S. S. (2010). Invited review: plasmin protease in milk: current knowledge and relevance to dairy industry. Journal of Dairy Science, 93, 4999e where the level did not decrease after skimming. The dissociation 5009. of PL from the caseins by simultaneous addition of EACA and NaCl Kelly, A. L., & Foley, J. (1997). Proteolysis and storage stability of UHT milk as mainly affected interactions of PL with inhibitory components influenced by milk plasmin activity, plasmin/b-lactoglobulin complexation, present in whey, e.g., PIs and whey proteins, which reduced PL plasminogen activation and somatic cell count. International Dairy Journal, 7, 411e420. activity by approximately 30e40%. Further studies will be required Kelly, A. L., O’Flaherty, F., & Fox, P. F. (2006). Indigenous proteolytic enzymes in milk: to clarify the effect of homogenisation on the PL system and PL a brief overview of the present state of knowledge. International Dairy Journal, e activity and the mechanism of inhibition of PL by whey proteins. 16, 563 572. Kolev, K., Owen, W. G., & Machovich, R. (1995). Dual effect of synthetic plasmin substrates on plasminogen activation. Biochimica et Biophysica Acta (BBA) e Acknowledgements Protein Structure and Molecular Enzymology, 1247, 239e245. Korycka-Dahl, M., Dumas, B. R., Chene, N., & Martal, J. (1983). Plasmin activity in milk. Journal of Dairy Science, 66, 704e711. The Danish Agency for Science, Technology and Innovation Larsen, L. B., Hinz, K., Jørgensen, A. L. W., Møller, H. S., Wellnitz, O., (DASTI) is thanked for financial support of the study. We thank Bruckmaier, R. M., et al. (2010). Proteomic and peptidomic study of proteolysis Vibeke Mortensen from Arla Foods Strategic Innovation Centre for in quarter milk after infusion with lipoteichoic acid from Staphylococcus aureus. Journal of Dairy Science, 93,5613e5626. excellent technical assistance and Prof. Jacob Holm Nielsen (Uni- Lu, D. D., & Nielsen, S. (1993). Isolation and characterization of native bovine milk versity of Copenhagen) for helpful discussions. plasminogen activators. Journal of Dairy Science, 76, 3369e3383. Markus, G., Hitt, S., Harvey, S. R., & Tritsch, G. L. (1993). Casein, a powerful enhancer of the rate of plasminogen activation. Fibrinolysis, 7, 229e236. References Nicholas, G. D., Auldist, M. J., Molan, P. C., Stelwagen, K., & Prosser, C. G. (2002). Effects of stage of lactation and time of year on plasmin-derived proteolytic Alkjaersig, N., Fletcher, A. P., & Sherry, S. (1959). ε-Aminocaproic acid: an inhibitor of activity in bovine milk in New Zealand. Journal of Dairy Research, 69, 533e540. plasminogen activation. Journal of Biological Chemistry, 234, 832e837. Politis, I., Barbano, D. M., & Gorewit, R. C. (1992). Distribution of plasminogen and Bastian, E. D., & Brown, R. J. (1996). Plasmin in milk and dairy products: an update. plasmin in fractions of bovine milk. Journal of Dairy Science, 75, 1402e1410. International Dairy Journal, 6,435e457. Politis, I., Zavizion, B., Barbano, D. M., & Gorewit, R. C. (1993). Enzymatic assay for Bastian, E. D., Brown, R. J., & Ernstrom, C. A. (1991a). Casein interference in bovine the combined determination of plasmin plus plasminogen in milk: Revisited. plasmin assays using a synthetic substrate. Journal of Dairy Science, 74,4119e4124. Journal of Dairy Science, 76, 1260e1267. Bastian, E. D., Hansen, K. G., & Brown, R. J. (1991b). Activation of plasmin with Prado, B. M., Sombers, S. E., Ismail, B., & Hayes, K. D. (2006). Effect of heat treatment urokinase in ultrafiltered milk for cheese manufacture. Journal of Dairy Science, on the activity of inhibitors of plasmin and plasminogen activators in milk. 74, 3669e3676. International Dairy Journal, 16, 593e599. Bastian, E.D., Hansen, K.G., & Brown, R.J. (1993). Inhibitionofplasmin by b-lactoglobulin Precetti, A. S., Oria, M. P., & Nielsen, S. S. (1997). Presence in bovine milk of two using casein and a synthetic substrate. Journal of Dairy Science, 76,3354e3361. protease inhibitors of the plasmin system. Journal of Dairy Science, 80, 1490e Benfeldt, C. (2006). Ultrafiltration of cheese milk: effect on plasmin activity and 1496. proteolysis during cheese ripening. International Dairy Journal, 16,600e608. Richardson, B. C., & Pearce, K. N. (1981). The determination of plasmin in dairy Benfeldt, C., Larsen, L. B., Rasmussen, J. T., Andreasen, P. A., & Petersen, T. E. (1995). products. New Zealand Journal of Dairy Science and Technology, 16, 209e220. Isolation and characterization of plasminogen and plasmin from bovine milk. Rollema, H. S., Visser, S., & Poll, J. K. (1983). Spectrophotometric assay of plasmin International Dairy Journal, 5,577e592. and plasminogen in bovine milk. Milchwissenschaft, 38,214e217. Cano-Ruiz, M. E., & Richter, R. L. (1997). Effect of homogenization pressure on the Saint-Denis, T., Humbert, G., & Gaillard, J. L. (2001). Enzymatic assays for native milk fat globule membrane proteins. Journal of Dairy Science, 80,2732e2739. plasmin, plasminogen and plasminogen activators in bovine milk. Journal of Cesarman-Maus, G., & Hajjar, K. A. (2005). Molecular mechanisms of fibrinolysis. Dairy Research, 68,437e449. British Journal of Haematology, 129, 307e321. Sun, Z., Chen, Y.-H., Wang, P., Zhang, J., Gurewich, V., Zhang, P., et al. (2002). The Christensen, S., Sottrup-Jensen, L., & Christensen, U. (1995). Stopped-flow fluores- blockage of the high-affinity lysine binding sites of plasminogen by EACA cence kinetics of bovine alpha 2-antiplasmin inhibition of bovine midiplasmin. significantly inhibits prourokinase-induced plasminogen activation. Biochimica Biochemical Journal, 305,97e102. et Biophysica Acta (BBA) e Protein Structure and Molecular Enzymology, 1596, Collen, D. (1987). Molecular mechanisms of fibrinolysis and their application to 182e192. fibrin-specific thrombolytic therapy. Journal of Cellular Biochemistry, 33,77e86.

Paper II

Rauh, V. M., Sundgren, A., Bakman, M., Ipsen, R., Paulsson, M., Larsen, L. B., & Hammershøj, M. (2014)

Plasmin activity as a possible cause for age gelation in UHT milk produced by direct steam infusion

International Dairy Journal, 38, 199-207

International Dairy Journal 38 (2014) 199e207

Contents lists available at ScienceDirect

International Dairy Journal

journal homepage: www.elsevier.com/locate/idairyj

Plasmin activity as a possible cause for age gelation in UHT milk produced by direct steam infusion

Valentin M. Rauh a,b,*, Anja Sundgren c, Mette Bakman a, Richard Ipsen d, Marie Paulsson e, Lotte B. Larsen b, Marianne Hammershøj b a Arla Foods Strategic Innovation Centre, Rørdrumvej 2, DK-8220 Brabrand, Denmark b Aarhus University, Dept. of Food Science, Faculty of Science and Technology, Blichers Allé 20, DK-8830 Tjele, Denmark c Arla Foods Strategic Innovation Centre, Lindhagensgatan 126, SE-10546 Stockholm, Sweden d University of Copenhagen, Department of Food Science, Rolighedsvej 30, DK-1958 Frederiksberg C, Denmark e Dept of Food Technology, Engineering and Nutrition, Lund University, SE-22100 Lund, Sweden article info abstract

Article history: The effect of enzymatic activity in direct steam infusion heat treated milk with ultra-short holding times Received 20 September 2013 (>150 C for <0.2 s) on age gelation during storage was investigated. Preheating at either 72 or 95 C for Received in revised form 180 s was performed. Milk pre-heated at 72 C showed extensive proteolysis and exhibited bitter off 18 December 2013 ﬂavour and contained <40% intact a - and b-caseins after 6 weeks storage at 20 C. No proteolysis of k- Accepted 18 December 2013 S casein was detected. Plasmin was identiﬁed as active protease and activation of plasminogen was Available online 8 January 2014 observed as an increase in the rate of casein hydrolysis. Proteolysis in the stored samples correlated with a decrease in pH and with changes in colour. Gelation occurred after 10 weeks along with an increase in viscosity and extensive proteolysis of aS- and b-caseins. In conclusion, plasmin activity was involved in age gelation and bitterness caused by proteolysis was the shelf-life limiting factor. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction compared with indirect methods, due to the rapid heating and cooling of the milk (Newstead, Paterson, Anema, Coker, & Wewala, Ultra-high temperature (UHT) treatment of milk is a heating 2006). This may be solved by inclusion of a preheat treatment step process at very high temperatures for short holding times, which (Newstead et al., 2006; Van Asselt, Sweere, Rollema, & de Jong, renders the milk commercially sterile and gives a product with a 2008) or by increasing the temperature during UHT treatment long shelf life at ambient temperatures. UHT treatments are usually (Topçu, Numanoglu, & Saldamlı,2006). carried out at temperatures of 140 C for 4e6 s. Both indirect and The cause for age gelation in UHT milk is not known, but it has direct heating systems (steam injection or infusion) have been been suggested that release of the b-lactoglobulin-k-casein com- applied to increase the shelf-life of milk. More recently, UHT plex from the casein micelle is involved (McMahon, 1996). This treatments with temperatures >150 C and ultra-short holding complex is formed during heat treatment by the denaturation of b- times (<0.2 s) have been shown to result in a sufﬁcient reduction of lactoglobulin (b-lg) and subsequent aggregation with k-casein via micro-organisms, while doing minimal heat damage to milk con- disulphide bonds. The released complex aggregates and forms a gel. stituents, such as vitamins and whey proteins (De Jong, 1996). Proteolysis accelerates or can cause the release of the b-lactoglob- A major limiting factor for the shelf life of UHT milk is proteol- ulin-k-casein complex by hydrolysis of the caseins that anchor the ysis of caseins, which can be caused by residual heat-stable bac- b-lactoglobulin-k-casein in the casein micelle (Datta & Deeth, 2001; terial proteases, indigenous milk proteases or combinations of McMahon, 1996). While a certain level of proteolysis is required to these. Such proteolysis can cause formation of bitter off-ﬂavours, induce age gelation, the relation between gelation time and extent sedimentation and age gelation of UHT milk (Datta & Deeth, of proteolysis is controversial (Manji & Kakuda, 1988; Newstead 2003; Kelly & Foley, 1997). These problems more often occur in et al., 2006). directly heated UHT milk, since the heat load is reduced, when Plasmin (PL; EC 3.4.21.7), the major indigenous protease in milk, has been closely linked to the age gelation of UHT milk (Kelly & Foley, 1997; Kohlmann, Nielsen, & Ladisch, 1991; Manji & Kakuda, b a * Corresponding author. Tel.: þ45 87332783. 1988; Newstead et al., 2006). PL readily hydrolyses - and S2- E-mail address: [email protected] (V.M. Rauh). casein, and to a lesser extent, aS1-casein (Grufferty & Fox, 1988). The

0958-6946/$ e see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.idairyj.2013.12.007 200 V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207 level and activity of PL in milk can vary and depends on biological volume 6.8 mL) on a TA Discovery HR-2 rheometer (TA Instruments, factors, such as stage of lactation and somatic cell count (Larsen New Castle, DE, USA). Measurements were performed at 25 C. After et al., 2010; Nicholas, Auldist, Molan, Stelwagen, & Prosser, 2002). the temperature was reached, a 15 s pre-shearing step was con- PL is the active part of a complex enzyme system. Its zymogen, ducted before viscosity was recorded for 120 s at a shear rate of plasminogen (PLG), is present in raw, bovine milk at a 2e30 times 100 s 1. Measurements were performed in triplicate. higher concentration than PL (Ismail & Nielsen, 2010; Korycka- The volume based particle size distribution of the milk was Dahl, Dumas, Chene, & Martal, 1983; Rauh et al., submitted for obtained with a Master Sizer 3000 (Malvern Instruments Ltd., publication), and is activated by two families of plasminogen acti- Malvern, UK). Distilled, degassed water was used as dispersant. The vators (PA), urokinase-type PA (uPA) and tissue-type PA (tPA). The refractive index was set to 1.45 for particles and to 1.33 for the activity of PL and PA is regulated by plasmin inhibitors (PIs), such as dispersant. Milk was added under agitation at 1000 rpm to the a2-antiplasmin, and plasminogen activator inhibitors (PAIs) dispersion unit until laser obscuration was in the range of 5.5e8.0% (Grufferty & Fox, 1988; Ismail & Nielsen, 2010; Lu & Nielsen, 1993; and particle size distributions were recorded. Measurements were Precetti, Oria, & Nielsen, 1997). performed in duplicate. The components of the PL system show very different thermal After 10 weeks storage, visually detectable gel fragments were stabilities. While PL, PLG and PA are heat-stable enzymes, the in- placed on a microscope slide and stained by adding 1 mL of a 0.02% hibitors of the PL system are generally considered as heat labile fluorescein-5-isothiocyanate (FITC; Merck, Darmstadt, Germany) (Prado, Ismail, Ramos, & Hayes, 2007; Prado, Sombers, Ismail, & solution in acetone for proteins and a 0.02% Nile red (9- Hayes, 2006; Richardson, 1983). Due to this complexity it is diffi- diethylamino-5H-benzophenoxazine-5-one; Merck) solution in cult to predict the actual PL activity and the consequence of pro- acetone for fat. After the acetone was fully evaporated, the samples teolysis for heat treatment of milk under novel and untested time- were examined using a 40 and 100 oil immersion objective on temperature conditions. a Leica DMIRE2 inverted confocal laser scanning microscope (Leica The aim of the present study was to investigate the effect of Microsystems GmbH, Heidelberg, Germany) fitted with an Ar/Kr plasmin activity on proteolysis and physical-chemical changes in laser for FITC and a He 543/594 nm laser for Nile red. FITC was milk treated with direct steam infusion heating and ultra-short excited at 485 nm and emission was recorded between 500 and holding times during subsequent storage. 545 nm. Nile red was excited at 543 nm and the emitted signal recorded from 560 to 700 nm. 2. Materials and methods PL and PLG derived activity were measured by the rate of hydrolysis of the chromogenic substrate S-2251 (Rollema, Visser, & 2.1. Materials Poll, 1983) and release of p-nitroaniline as described previously (Rauh et al., submitted for publication). PLG was activated by One day old low pasteurised milk (72 C for 15 s) with 1.5% fat was addition of uPA (Immuno Danmark A/S, Copenhagen, Denmark). obtained from Arla Foods Stockholm Dairy (Stockholm, Sweden) and The rate of increase in absorbance intensity during the assay is processed in the UHT pilot plant (SPX, Silkeborg, Denmark) at Arla proportional to the PL or PLG derived activity. Strategic Innovation Centre (Stockholm, Sweden). The milk (200 L) Analysis of protein composition in the milk was performed as was pre-heated at 72 C (Milk72) or 95 C (Milk95) for 180 s in a described by Bonfatti, Grigoletto, Cecchinato, Gallo, and Carnier tubular heat exchanger before being subjected to the direct steam (2008) with the following modifications. A sample of 200 mLof infusion heat treatment (>150 C for <0.2 s). Pre-heat treatment milk (w/w) was frozen at 20 C until further use. conditions were chosen to obtain a milk with (Milk72) and without Disulphide bonds were reduced by the addition of 20 mL1M (Milk95) residual PL activity. After flash cooling, the milk was dithiothreitol and 1 mL 100 mM tri-sodium citrate, 6 M urea at 30 C homogenised using a two-stage homogeniser (160/40 bar), further for 60 min. The samples were centrifuged at 9300 g at 5 C for cooled down to 20 C by a tubular heat exchanger and packed 10 min. From the clear phase, 200 mL were used for the analysis. A aseptically in 300 mL glass bottles. The trials were performed during volume of 5 mL was injected into the LCeMS system. The column three consecutive weeks. The milk was stored in the dark at 20 Cina was a BioSuiteÔ C18 PA-B (C18, 3.5 mm; 2.1 mm 250 mm, Waters, climate chamber for 14e16 weeks. Samples, i.e., one 300 mL bottle Milford, MA, USA). Buffer A was Milli-Q water with 0.5% (v/v) tri- per treatment, were collected once every week for further analysis. fluoroacetic acid (TFA) and buffer B acetonitrile with 0.1% (v/v) TFA. Column temperature was 47 C and a linear gradient from 34.8% to 2.2. Analytical methods 46.5% buffer B from 2 to 16.5 min was applied at a flow rate of 0.35 mL min 1 with UV detection at 214 nm. MS detection was Each batch of processed milk was tested for sterility according to performed with a quadrupole time-of-flight mass spectrometer ISO 4833 (ISO, 2003). The milk was visually inspected for sedi- with (Q-TOF-MS; Agilent 6530 Aqurate mass detector, Agilent mentation, gelation and colour. Milk72, Milk95 and fresh low pas- Technologies, Santa Clara, CA, USA) and MS scans were continu- teurised milk were tasted for perceived bitterness by five panellists. ously recorded between a mass to charge ratio (m/z) of 300 and The pH of the milk was measured at room temperature bya pH meter 3500. Data analysis was made with Mass Hunter software (Agilent (Knick Elektronische Messgeräte GmbH & Co. KG., Berlin, Germany). Technologies). Measurements were performed in duplicate. Measurements were performed in duplicates. Colour development For the analysis of pH 4.6 soluble peptides and native whey was measured using Hunters Lab colour space with a colorimeter proteins, volumes of 20 mL of milk were adjusted to pH 4.6 with 1 M Minolta CR-A70 (Minolta Camera Co., Ltd, Osaka, Japan). The L-, a-, HCl and centrifuged at 11,000 g for 10 min at 4 C. From the clear and b-values reflect lightness (0 ¼ black, 100 ¼ white), redness supernatant, 1.5 mL was taken out and frozen at 20 C until (100 ¼ green, 100 ¼ red) and yellowness (100 ¼ blue, further use. Prior to analysis, 1 mL of the sample was centrifuged at 100 ¼ yellow), respectively. The instrument was calibrated with a 9300 g at 5 C for 10 min and the supernatant used for the analysis Minolta standard plate (standard values: Y ¼ 92.4; x ¼ 0.3161; by injecting 10e20 mL to the LC system, depending on the peptide y ¼ 0.3325). Measurements were performed in triplicate. Viscosity concentration. The column was a Xbridge BEH 300 (C18, 3.5 mm; of the milk was measured using a double gap concentric cylinder 2.1 mm 250 mm, Waters) operated at 45 C. Buffer A was Milli-Q geometry (inside cup diameter 40.0 mm, inside bob diameter water with 0.1% (v/v) TFA and buffer B acetonitrile with 0.5% (v/v) 40.8 mm, cylinder height 59.5 mm, operating gap 2.0 mm, sample TFA. A linear gradient was applied with 100% buffer A from 3 min, V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207 201

100%e45% buffer A at 45 min at a flow rate of 0.3 mL min 1 with UV of storage. A similar development has been shown for direct UHT detection at 214 nm. The degree of denaturation of whey proteins treatments with ultra-short holding time (Huijs et al., 2004). No was calculated as the relative difference in the UV peak area of pH bitterness was detected in Milk95. 4.6 soluble whey proteins before and after heat treatment. Mea- After 10 weeks storage, a visible difference in the colours be- surements were performed in duplicate. tween Milks72 and 95 was observed, with Milk72 being less opa- que than Milk95. After 10 weeks, a gel layer was found at the 3. Results and discussion bottom of the bottle of Milk72, which increased in thickness in the following week. Milk95 showed no gelation or sedimentation 3.1. Influence of heat treatment during the storage period. The gel was soft and fragile. Along with this gel layer, floating gel particles appeared in Milk72 after 11 The heat treatment rendered the milk sterile as no microbial weeks storage. After 16 weeks storage, Milk72 was completely activity was detected during storage. The degree of whey protein gelled and syneresis could be observed. The gel formation and denaturation and residual PL and PLG derived activities for Milk72 appearance was similar to the description of a plasmin induced gel and Milk95 were measured one day after the heat treatment and found in UHT milk by Kelly and Foley (1997). In contrast to other are shown in Table 1. In comparison with other direct heat treat- similar studies (Newstead et al., 2006; Topçu et al., 2006), Milk72 ments applying ultra-short holding times reported earlier, Milk72 showed no sedimentation. This difference may arise from the showed a higher degree of denaturation for b-lg. Previously re- shorter storage time needed for gelation in this study. Milk72 ported values for b-lg denaturation were 19e21% at 150e170 C already gelled after 80 d of storage, while Newstead et al. (2006) (Huijs, Van Asselt, Verdurmen, & De Jong, 2004) and 15% and 28% at and Topçu et al. (2006) reported gelation after 150e180 d of 150 C(Dickow, 2011) for skim and whole milk, respectively, both storage. using a steam injection system. This difference can be explained by In addition to the gel formation, a creamy layer appeared on the the additional b-lg denaturation occurring during the pre-heat top of Milk72 after 80 d of storage, a phenomenon that has been treatment at 72 C for 180 s amounting to approximately 2e4% described for milk in connection with plasmin activity by (Tolkach & Kulozik, 2007) and additionally approximately 4% dur- Kohlmann et al. (1991). This was most apparent in the samples ing the homogenisation step following the direct heat treatment adjusted to pH 4.6: due to the homogenisation, small fat globules (García-Risco, Ramos, & López-Fandiño, 2002). Milk95 showed could be observed in the supernatant of Milk95 after centrifugation, >90% denaturation of b-lg and denaturation of a-lactalbumin (a-la) while the supernatant of Milk72 was clear with a creamy layer on in Milk95 was approximately 3.5 times higher than in Milk72. top. Residual PL and PLG derived activities in Milk72 were 30.9 2.3% and 14.0 2.7%, respectively, whereas no residual PL or 3.3. Physical-chemical changes during storage PLG derived activity could be detected in Milk95. Denaturation of b- lg and inactivation of PL and PLG are known to be closely correlated The pH of Milk72 and Milk95 (Fig. 1) decreased during storage, (Ismail & Nielsen, 2010; Newstead et al., 2006). The higher heat load but the decrease was more pronounced in Milk72, by approxi- and degree of b-lg denaturation of Milk72 correlates with the lower mately 0.2 pH-units. The larger decrease in Milk72 can be related to residual PL and PLG derived activity compared to the residual levels the residual PL activity in the sample and proteolysis after lysine of 50% and 37% PL and PLG derived activity, respectively, found after residues (Kelly & Foley, 1997; Kohlmann et al., 1991). Interestingly, steam injection (150 C, Dickow, Nielsen, & Hammershøj, 2012). the pH did not decrease further after the onset of gelation after 10 Interestingly, it seems that in direct UHT treatments with ultra- weeks storage. short holding times, PLG appears to have lower heat stability The colour values of Milk95 only decreased slightly during the than PL (Table 1). Kinetic studies on the heat inactivation of PL and storage period, while the colour values for Milk72 decreased PLG as well as direct UHT treatments show a similar or higher heat gradually, becoming less white (decreased L-value), more greenish stability of PLG compared with PL at temperatures between 90 and (higher negative a-value), and less yellow (lower b-value). As can 142 C(Manji & Kakuda, 1986; Rollema & Poll, 1986; Saint-Denis, be seen in Fig. 2, Milk95 showed higher initial a and b colour values, Humbert, & Gaillard, 2001). This suggests that a different heat which can be attributed to the higher extent of Maillard reaction inactivation kinetic for PL and PLG applies for direct UHT treat- due to the more severe heat treatment (Kneifel, Ulberth, & Schaffer, ments with ultra-short holding times compared with other UHT treatments and previous kinetic studies.

3.2. Appearance and taste

In Milk72 a bitter ﬂavour was noticeable after 6 weeks that rapidly intensiﬁed, and the milk was extremely bitter after 7 week

Table 1 Whey protein denaturation and residual plasmin (PL) and plasminogen (PLG) derived activity in Milk72 and Milk95 after indirect pre-heat treatment and direct steam infusion (>150 C for >0.2 s).a

Milk Indirect pre-heat Whey protein denaturation (%) Residual activity (%) treatment b-lg a-la PL PLG

Milk72 72 C/180 s 36.7 3.2 13.1 2.2 30.9 2.3 14.0 2.7 Milk95 95 C/180 s 94.1 1.6 45.5 3.3 n.d. n.d.

a The degree of denaturation of whey proteins (b-lg, b-lactoglobulin; a-la, a- lactalbumin) was calculated as the relative difference in the UV peak area of pH 4.6 Fig. 1. pH development in milk subjected to direct steam infusion (>150 C for <0.2 s) soluble whey proteins before and after heat treatment: results are shown as and pre-heat treated at 72 C()or95C(,) for 180 s during storage at 20 C. Results average standard deviation of three replicates; n.d., non-detectable. are shown as average standard deviation of three replicates. 202 V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207

Fig. 2. Development of Lab colour values and viscosity in milk subjected to direct steam infusion (>150 C for <0.2 s) and pre-heat treated at 72 C()or95C(,) for 180 s during storage at 20 C. Results are shown as average standard deviation of three replicates.

1992). Similar to the pH value, the colour values levelled out with the fat globules to the particle size distribution. After 11 weeks the onset of gelation, and showed a tendency to increase again for storage, larger particles between 8 and 10 mm could be observed. In the a and b colour values. The difference in the L value of Milk72 the following week, the population of the larger particles increased and Milk95 correlates well with the observed visible difference in while the casein micelle/fat globule distribution became narrower opacity and could indicate a change in the casein micelles proper- and contributed to a relatively smaller amount of the particles. The ties. A reduction of the L value has been correlated with an increase large particles were most likely gel fragments. Due to the stirring in casein micelle size after extensive hydrolysis with PL (Crudden, during the measurement, the gel particles broke down to smaller Afoufa-Bastien, Fox, Brisson, & Kelly, 2005) as well with a dissoci- fragments resulting in a relatively narrow size distribution. The ation of casein micelles upon tri-sodium citrate addition broadening of the distribution towards higher sizes after 14 weeks (O’Sullivan, Kelly, & Fox, 2002). The latter could also be associated reflects the increased firmness of the gel. with PL mediated proteolysis as partially disintegrated casein micelles have been found in PL-treated milk (Aroonkamonsri, 1996). The development in viscosity of Milk72 and 95 is shown in Fig. 2. Viscosity of both milks was constant between 2.0 and 2.5 mPa s during the first 9 weeks of storage. After 10 weeks storage, the onset of gelation caused the viscosity of Milk72 to increase and it further increased rapidly in the following week up to 13.6 2.6 mPa s. The viscosity of Milk95 did not change during the applied storage period. A similar observed increase in viscosity with the onset of gelation has been reported previously by Kohlmann et al. (1991) and Kelly and Foley (1997). During the first 10 weeks storage, the particle size distribution did not change significantly (data not shown). Fig. 3 shows a representative volume based particle size distribution of particles in Milk72 between 0 and 14 weeks storage. Due to the homogenisation of the milk, the individual size distributions of casein micelles and fat globules could not be distinguished. The 10 week old milk shows a mono-modal distribution consisting of casein micelles and homogenised fat globules with an average diameter of Fig. 3. Volume based particle size distribution of milk pre-heat treated at 72 C for approximately. 190 nm. The change in casein micelles observed in 180 s and subjected to direct steam infusion (>150 C for <0.2 s) after 0 (continuous the colour measurement could not be seen in the particle size line), 8 (short-dashed line), 10 (dotted line), 12 (long-dashed line), and 14 (dot-dash distribution and is most likely overshadowed by the contribution of line) weeks storage at 20 C. V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207 203

The microstructure of the gel found in Milk72 was visualised by extensive proteolysis during the storage period, while the overall confocal laser scanning microscopy (see Supplementary material). chromatographic pattern for Milk95 remained unchanged. From The gel appeared to have a loose, non-homogeneous structure with the UV chromatograms, no statement about the development in liquid filled cavities, similar to the microstructure of yoghurts and aS2-casein could be made due to overlay with upcoming poly- acid milk gels (Pereira, Matia-Merino, Jones, & Singh, 2006; Torres, peptides (Fig. 5 AeD). After 4 weeks storage, the b-casein A1 peak Amigo Rubio, & Ipsen, 2012). The micrographs show small fat decreased the most, followed by a slight decrease in aS1-casein and globules (0.5e2 mm), that are partially present in clusters and b-casein A2. The peak of b-casein A1 disappeared almost completely incorporated in the gel structure. The diameter of the fat globules is after 8 weeks storage, along with a large decrease in the aS1-casein in line with the observed particle size distributions. The presence of peak. After 12 weeks storage, no intact aS1-casein with 9 phos- agglutinated fat globules could account for the observed cream phorylations (P) and b-casein A1 was present in the UV chro- layer in Milk72. matogram (Fig. 5D). Over the storage period, the k-casein peaks seemed to disappear, but developed similarly in both milk types, 3.4. Enzymatic activity and proteolysis during storage ruling out proteolysis as a cause for this change in k-casein profile. Instead, the disappearance can be attributed to a broadening of the 3.4.1. Plasmin and plasminogen derived activity peaks due to modifications of the proteins, likely by the Maillard The development in PL and PLG derived activities in Milk72 are reactions induced by the heat treatment (Gaucher, Mollé, Gagnaire, shown in Fig. 4. As can be seen, PLG derived activity decreased & Gaucheron, 2008). For Milk95, the same tendency could be seen rapidly, while PL activity increased. This indicates activation of PLG for aS2-casein. into PL during storage, and no PLG derived activity was detectable Interestingly, significant differences in the peak development of 1 2 after 6 weeks storage. A similar development has been reported in b-casein A and A as well as for aS1-casein with 8P and 9P were direct heated UHT milk (Manji & Kakuda, 1986). The increase in PL observed. Therefore, the mass spectra of peaks 4e9 were analysed activity could be fitted well to a first order reaction kinetic after 8 weeks storage to investigate the components present in the (k ¼ 0.1828, R2 ¼ 0.995), while PLG derived activity on the other hand peaks. The results are summarised in Table 2 and show that pro- decreased from 0 to 6 week in a linear fashion (y ¼104.7x þ 603; teose peptones 5 and 8 (slow) from b-casein A2 are co-eluting with 2 R ¼ 0.997). The inhibitors of the PL system are heat labile: under aS2-casein along with a polypeptide from aS1-casein f(1e124). The pasteurisation conditions (74.5 C for 15 s) PAI is completely inac- corresponding polypeptide aS1-casein f(125e199) was co-eluting tivated, while about 36% of PI is inactivated (Prado et al., 2006). PA with aS1-casein 8P. While the peaks of aS1-casein 9P and b-casein have been shown to have a similar or even higher heat stability than A1, either were not or only slightly contaminated with poly- 2 PL and PLG (Lu & Nielsen, 1993; Saint-Denis et al., 2001), suggesting peptides, g1- and g2-caseins were co-eluting with b-casein A and that a significant proportion of PA remained in Milk72 after the heat g3-casein with a-la. This explains the difference in the peak 1 2 treatment. This, together with the inactivation of PAI, explains the development of aS1-casein 8 and 9P as well as b-casein A and A .A fast rate of PLG activation in Milk72. After 4 weeks storage, the PL comparison of the mass spectra (data not shown) showed that the activity in Milk72 was comparable to the PL activity found in the low rate of hydrolysis was similar for the two genetic variants A1 and A2 1 pasteurised milk (632 28 mUmL ) prior to heat treatment. The of b-casein and for the differently phosphorylated aS1-caseins. For 1 gelation of Milk72 had no apparent effect on PL activity. this reason, aS1-casein 9P and b-casein A were chosen for a relative The PL and PLG derived activities developed differently and PL quantification of the proteolysis based on the UV chromatograms. activity continued to increase after depletion of PLG. This devel- Proteolysis in UHT milk can be caused by PL and heat resistant opment can be ascribed to other factors affecting PL activity, such as proteases from, e.g., psychotropic bacteria. The milk used in this a reduced competitive inhibition of PL by caseins towards the study was one day old low-pasteurised milk, which makes it un- synthetic substrate in the analysed milk samples (Bastian, Brown, & likely that the extensive proteolysis in Milk72 was caused by bac- Ernstrom, 1991) due to proteolysis. terial proteases. The appearance of proteose peptones and g- caseins suggested that PL was the main actor of proteolysis. Bac- 3.4.2. Changes in protein profiles during storage terial proteases are more unspecific in their cleavage sites and are In Fig. 5, a comparison of the protein profiles of Milk72 and 95 capable of hydrolysing k-casein (Datta & Deeth, 2001). In Milk72 after 0, 4, 8 and 12 weeks storage is shown. Milk72 showed very and Milk95 no proteolysis of k-casein was observed, further strengthening the indications that PL was the main active protease in Milk72. 1 The hydrolysis of aS1-casein 9P and b-casein A in Milk72 during 1 storage are shown in Fig 6.BothaS1-casein 9P and b-casein A decrease in an inverse sigmoidal manner. Hydrolysis of casein by PL can usually be described by an exponential decay function (Newstead et al., 2006). In Milk72, a lag phase in the first 2e3 weeks storage preceded the exponential decay. This lag phase can be attributed to the activation of PLG to active PL as shown in Fig. 4. With increasing PL activity, the rate of hydrolysis increased. 1 The b-casein A was hydrolysed faster than aS1-casein 9P, as also reported in the literature (Grufferty & Fox, 1988). After 9 weeks storage, more than 95% of b-casein A1 was hydrolysed, while the same degree of hydrolysis for aS1-casein 9P was reached after 11 week. The rate of hydrolysis of aS2-casein by PL has been shown to be similar to b-casein (Grufferty & Fox, 1988). The MS spectrum of the a -casein peak at different storage times (data not shown) showed Fig. 4. Plasmin () and plasminogen (6) derived activity in milk pre-heat treated at S2 a 72 C for 180 s and subjected to direct steam infusion (>150 C for <0.2 s) during that S2-casein was almost completely hydrolysed after 9 weeks storage at 20 C. Results are shown as average standard deviation of three replicates. storage, indicating the same rate of hydrolysis as b-casein. Gelation 204 V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207

Fig. 5. Protein proﬁles of milk subjected to direct steam infusion (>150 C for <0.2 s) and pre-heat treated at 72 C (Milk72, solid line) or 95 C (Milk95, dashed line) for 180 s after 0 (A), 4 (B), 8 (C) and 12 (D) weeks storage at 20 C. The observed proteins at week 0 are k-casein (peaks 1e3), aS2-casein (4), aS1-casein with 8 (5) or 9 (6) phosphorylations, b- casein A1 (7) and A2 (8), a-lactalbumin (9), b-lactoglobulin B (10) and A (11).

1 occurred after 95% b-casein A and aS1-casein 9P had been indicated, the gel had a relatively loose structure, which contained hydrolysed. a high proportion of liquid phase. This shows that k-casein and The protein composition of the gel formed had the same pattern whey proteins were not incorporated into the gel network. To the as the liquid (Fig. 7). The relative concentration of residual intact a- authors’ knowledge, the composition of a plasmin induced gel has and b-caseins and the polypeptides in the gel were higher than in not been reported previously. the liquid, while the concentrations of k-casein and whey proteins Bitterness in Milk72 was noticeable after 6 weeks storage, at a were similar in the gel and in the liquid. As the micrographs time point when approximately 70% of aS1-casein and approximately 75% of b-casein (and presumably aS2-casein) was hydrolysed. Bitterness caused by PL has been attributed to peptides from Table 2 Probable identitya of components in milk pre-heat treated at 72 C for 180 s and subjected to direct steam infusion (>150 C for <0.2 s) after 8 weeks of storage at 20 C from protein proﬁle in peaks 4e9 (see Fig. 5A).

Peak no. Deconvoluted Theoretical Mass error ID mass (Da) mass (Da) (ppm)

4 8717.61 8717.91 34.9 b-CN A2 f(29e105) 12,178.26 12,178.28 2.3 b-CN A2 f(1e105)

14,995.20 14,995.15 3.2 aS1-CN 8P f(1e124) 25,229.33 25,229.35 5.7 aS2-CN 11P 25,309.29 25,309.33 1.0 aS2-CN 12P 25,389.32 25,389.31 0.2 aS2-CN 13P 5 8639.36 8639.61 29.0 aS1-CN 8P f(125e199) 23,615.68 23,615.75 2.9 aS1-CN 8P 6 23,695.58 23,695.73 6.4 aS1-CN 9P 7 24,023.15 24,024.25 4.4 b-CN A1 20,563.65 20,563.88 11.0 b-CN A1 f(29e209) 12,679.69 12,680.03 26.3 b-CN f(98e209) 8 11,824.75 11,824.96 20.4 b-CN f(106e209) 20,523.57 20,523.86 13.9 b-CN A2 f(29e209) 23,984.04 23,984.23 7.8 b-CN A2 Fig. 6. Hydrolysis of a -casein 9P () and b-casein A1 (6) in milk pre-heat treated at 9 11,559.45 11,559.64 16.3 b-CN f(108e209) S1 72 C for 180 s and subjected to direct steam infusion (>150 C for <0.2 s) during 14,187.00 14,187.12 8.3 a-lactalbumin storage at 20 C expressed as the relative change in peak area of UV absorption at a Identity is based on mass ﬁngerprinting using average mass and reduced pro- 214 nm. The arrow indicates the onset of gelation. Results are shown as teins and peptides: CN, casein; P, phosphorylated. average standard deviation of three replicates. V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207 205

activity measurement. The decreased rate of proteolysis was most likely due to a limited accessibility of remaining substrate for PL in the gel matrix and depletion of preferred cleavage sites. In Fig. 9, the development in the total peak area of pH 4.6 soluble peptides during storage is shown. The peak area of Milk72 developed in a sigmoidal way and appeared to almost inversely mirror 1 the hydrolysis of aS1-casein 9P and b-casein A (Fig. 8). After a 2e3 week lag phase, the peak area increased rapidly and levelled out after the onset of gelation. This is in line with the almost complete 1 hydrolysis of aS1-casein 9P and b-casein A after 11 weeks, showing that no more material soluble at pH 4.6 was formed. The total peak area of Milk95 increased only slightly during the storage period, conﬁrming that no signiﬁcant proteolysis occurred.

3.4.4. Correlation of proteolysis and gel formation Proteolysis mediated by PL can cause gelation and increases Fig. 7. Protein profiles of gel particles (solid line) and liquid phase (dashed line) in milk transparency of milk and the outcome of proteolysis seem to pre-heat treated at 72 C for 180 s and subjected to direct steam infusion (>150 C for <0.2 s) after 13 weeks storage at 20 C. depend largely on the extent of PL activity and rate of proteolysis in the milk. Large amounts of added PL have earlier been reported to cause clarification of milk (Crudden et al., 2005), while low b a -casein (Harwalkar, Cholette, McKellar, & Emmons, 1993) and S2- amounts or indigenous residual PL has caused gelation of milk casein (Le Bars & Gripon, 1989). The extensive proteolysis necessary (Kelly & Foley, 1997; Kohlmann et al., 1991; Newstead et al., 2006). to produce bitterness indicates that either (1) the sensory threshold In this study, at the onset of gelation of Milk72 more than 95% of of the formed bitter peptides is high or bitterness was masked by a- and b-caseins were hydrolysed. Previous studies (Kelly & Foley, the matrix, or (2) that the cleavage positions that result in bitter 1997; Kohlmann et al., 1991; Newstead et al., 2006) showed that peptides are not preferred by PL and the bitter peptides were almost complete proteolysis of b-casein was reached before a gel formed after the majority of the preferred cleavage positions had was formed, while a-casein on the other hand was either still been hydrolysed. present to some extent (Kelly & Foley, 1997) or almost completely hydrolysed (Kohlmann et al., 1991). The low level of b-lg denatur- fi 3.4.3. Changes in peptide pro les during storage ation in Milk72, compared to other UHT treatments, and the low fi The development in the peptide pro le of Milk72 (Fig. 8) concentration of b-lg and k-casein in the formed gel suggests that e showed a rapid increase of peaks with 31 34 min retention time the formation and release of a b-lactoglobulin-k-casein complex fi between 0 and 4 weeks storage. These peaks were identi ed as was not the main cause of gelation. Most studies on PL in milk have proteose peptones. After 8 weeks storage, more peptides appeared focussed on b-casein as an indicator of proteolysis. In relation to age between 17 and 30 min. Most of the peptides eluted late in the RP- gelation, it is important to consider both hydrolysis of aS- and of b- UHPLC, which has been reported previously to correlate with PL caseins. A part of b-casein has been shown to be highly mobile in fi activity (Datta & Deeth, 2003). The peptide pro le changed slightly the casein micelle (Dalgleish, 2011) and hydrolysis of b-casein could between 12 and 14 weeks storage, although the total peak area of be responsible for the reported increase in hydrophobicity fi the peptide pro le of Milk72 remained constant in the same time (Aroonkamonsri, 1996), while the hydrolysis of a -casein could be fi S1 period. The most profound change in the peptide pro le between responsible for loss of micellar integrity, which in combination 12 and 14 weeks storage was seen in the proteose peptone peaks. could lead to the age gelation of milk. Manji and Kakuda (1988) b 1 e The proteose peptone -casein A f(1 105) eluting at 33.0 min suggested gelation to be a two-stage process, which requires b 1 e decreased and -casein A f(29 105) eluting at 32.7 min increased some degree of proteolysis before time-dependent physical- (peptides are indicated by arrows in Fig. 8). This shows that PL was chemical changes lead to the gelation of milk. This could be a still active after gelation of Milk72, as already suggested by the PL

Fig. 9. Development of pH 4.6 soluble peptides of milk subjected to direct steam Fig. 8. Peptide profile of pH 4.6 soluble peptides of milk pre-heat treated at 72 C for infusion (>150 C for <0.2 s) and pre-heat treated at 72 C (Milk72, )or95C 180 s and subjected to direct steam infusion (>150 C for <0.2 s) at 0, 4, 8, 12 and 14 (Milk95, ,) for 180 s during storage at 20 C expressed as total peak area after weeks storage at 20 C. Arrows indicate proteose peptones 8 and 5 of b-casein A1 integration of UV absorbance at 214 nm. Results are shown as average standard f(29e105) and f(1e105) at 32.7 min and 33.0 min, respectively. deviation of three replicates. 206 V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207 reason for the different observations on age gelation of milk caused separation and quantification of bovine milk protein genetic variants. Journal of e by PL. If PL activity in the milk is high, the polypeptides could be Chromatography A, 1195,101 106. Crudden, A., Afoufa-Bastien, D., Fox, P. F., Brisson, G., & Kelly, A. L. (2005). Effect of further degraded and inhibit the formation of a gel, before physical- hydrolysis of casein by plasmin on the heat stability of milk. International Dairy chemical changes will take place. In milk with only little residual PL Journal, 15,1017e1025. Dalgleish, D. G. (2011). On the structural models of bovine casein micelles-review activity, physical-chemical changes induced by proteolysis could e fi and possible improvements. Soft Matter, 7, 2265 2272. occur prior to achieving a degree of proteolysis suf cient to cause a Datta, N., & Deeth, H. C. (2001). Age gelation of UHT milkda review. Food and gelation, which could be the case for observed sedimentation prior Bioproducts Processing, 79,197e210. to or instead of gelation (Kelly & Foley, 1997; Newstead et al., 2006). Datta, N., & Deeth, H. C. (2003). Diagnosing the cause of proteolysis in UHT milk. LWT e Food Science and Technology, 36,173e182. This may also explain why the degree of proteolysis and the onset of De Jong, P. (1996). Modelling and optimization of thermal processes in the dairy in- gelation have not been directly correlated yet. dustry. Doctoral Thesis. Delft, The Netherlands: Delft University of Technology. The results obtained in the present study suggest a gelation Dickow, J. A. (2011). Lenient steam injection heat treatment of milk and whey protein concentrate e Impact on enzymatic activities, protein denaturation and functional mechanism that is similar to acid gelation, with the difference that properties. PhD Thesis. Aarhus, Denmark: Aarhus University. proteolysis instead of acid causes a destabilisation of the casein Dickow, J. A., Nielsen, M. T., & Hammershøj, M. (2012). Effect of Lenient Steam micelle. Acidification of milk leads to a reduced repulsion of casein Injection (LSI) heat treatment of bovine milk on the activities of some en- micelles due to a reduced negative charge of k-casein, followed by a zymes, the milk fat globule and pH. International Journal of Dairy Technology, 65,191e200. rearrangement of the casein micelles by solubilisation of calcium García-Risco, M. R., Ramos, M., & López-Fandiño, R. (2002). Modifications in milk due to the pH drop that leads to the formation of a gel network proteins induced by heat treatment and homogenization and their influence on e (Lucey & Singh, 2003). While the gelation by plasmin depends susceptibility to proteolysis. International Dairy Journal, 12,679 688. Gaucher, I., Mollé, D., Gagnaire, V., & Gaucheron, F. (2008). Effects of storage tem- mainly on the rate of hydrolysis, the gel properties of acid gels perature on physico-chemical characteristics of semi-skimmed UHT milk. Food depend largely on the rate of acidification (Anema, 2008). Since the Hydrocolloids, 22,130e143. pH in Milk72 is only decreasing slightly, the repulsion by k-casein is Grufferty, M. B., & Fox, P. F. (1988). Milk alkaline proteinase. Journal of Dairy Research, 55, 609e630. still present in Milk72 and inhibits aggregation of casein micelles, Harwalkar, V. R., Cholette, H., McKellar, R. C., & Emmons, D. B. (1993). Relation which explains the absence of k-casein in the gel network. between proteolysis and astringent off-flavor in milk. Journal of Dairy Science, To fully understand the mechanism of PL induced age gelation, 76, 2521e2527. Huijs, G., Van Asselt, A. J., Verdurmen, R. E. M., & De Jong, P. (2004). High speed milk. these two stages would need to be investigated separately, e.g., by Dairy Industries International, 69,30e32. following changes in casein micelle structure in milk with different Ismail, B., & Nielsen, S. S. (2010). Invited review: plasmin protease in milk: current degrees of proteolysis. knowledge and relevance to dairy industry. Journal of Dairy Science, 93, 4999e 5009. ISO. (2003). Microbiology of food and animal feeding stuffs - horizontal method for the 4. Conclusions enumeration of microorganisms e Colony-count technique at 30 C. ISO 4833. Geneva, Switzerland: International Standardisation Organisation. Kelly, A. L., & Foley, J. (1997). Proteolysis and storage stability of UHT milk as Our results indicate that PL mediated proteolysis can be a likely influenced by milk plasmin activity, plasmin/b-lactoglobulin complexation, cause for age gelation in UHT milk produced with direct steam plasminogen activation and somatic cell count. International Dairy Journal, 7, infusion (>150 C, <0.2 s). Activation of PLG to PL could be observed 411e420. Kneifel, W., Ulberth, F., & Schaffer, E. (1992). Tristimulus colour reflectance mea- as an increase in PL activity and an increase in the rate of casein surement of milk and dairy products. Lait, 72, 383e391. hydrolysis and peptide formation. The observed changes in colour Kohlmann, K. L., Nielsen, S. S., & Ladisch, M. R. (1991). Effects of a low concentration and pH correlated with the proteolysis. The main shelf life limiting of added plasmin on ultra-high temperature processed milk. Journal of Dairy e factor for the UHT milk used was the appearance of bitterness, Science, 74,1151 1156. Korycka-Dahl, M., Dumas, B. R., Chene, N., & Martal, J. (1983). Plasmin activity in which preceded age gelation. Further studies are necessary to milk. Journal of Dairy Science, 66, 704e711. investigate the peptide formation by PL and to identify peptides Larsen, L. B., Hinz, K., Jørgensen, A. L. W., Møller, H. S., Wellnitz, O., related to the development of bitterness. Bruckmaier, R. M., et al. (2010). Proteomic and peptidomic study of proteolysis in quarter milk after infusion with lipoteichoic acid from Staphylococcus aureus. Journal of Dairy Science, 93,5613e5626. fi a Acknowledgements Le Bars, D., & Gripon, J.-C. (1989). Speci city of plasmin towards bovine S2-casein. Journal of Dairy Research, 56,817e821. Lucey, J. A., & Singh, H. (2003). Acid coagulation of milk. In P. F. Fox, & The Danish Agency for Science, Technology and Innovation P. L. H. McSweeney (Eds.) (3rd ed.)Proteins: Vol 1. Advanced dairy chemistry (pp. (DASTI) is thanked for financial support of the study. Many thanks 1001e1025). New York, NY, USA: Kluwer Academic-Plenum Publishers. to Jens Röschmann and Lena Bergström for the help in the Arla Lu, D. D., & Nielsen, S. (1993). Isolation and characterization of native bovine milk plasminogen activators. Journal of Dairy Science, 76, 3369e3383. Foods pilot plant and Ulf Andersen, Bettina Mikkelsen (Arla Stra- Manji, B., & Kakuda, Y. (1986). Effect of storage temperature on age gelation of UHT- tegic Innovation Center) and Isabel Celigueta (Nestlé Product milk mrocessed by direct and indirect heating systems. Journal of Dairy Science, e Technology Center) for the help with the confocal microscopy. 69, 2994 3001. Manji, B., & Kakuda, Y. (1988). The role of protein denaturation, extent of proteolysis, and storage temperature on the mechanism of age gelation in a model Appendix A. Supplementary data system. Journal of Dairy Science, 71, 1455e1463. McMahon, D. J. (1996). Age-gelation of UHT milk: Changes that occur during storage, their effect on shelf life and the mechanism by which age-gelation occurs. Brussels, Supplementary data related to this article can be found at http:// Belgium: International Dairy Federation. dx.doi.org/10.1016/j.idairyj.2013.12.007 Newstead, D. F., Paterson, G., Anema, S. G., Coker, C. J., & Wewala, A. R. (2006). Plasmin activity in direct-steam-injection UHT-processed reconstituted milk: effects of preheat treatment. International Dairy Journal, 16,573e579. References Nicholas, G. D., Auldist, M. J., Molan, P. C., Stelwagen, K., & Prosser, C. G. (2002). Effects of stage of lactation and time of year on plasmin-derived proteolytic Anema, S. G. (2008). Effect of temperature and rate of acidification on the rheo- activity in bovine milk in New Zealand. Journal of Dairy Research, 69, 533e540. logical properties of acid skim milk gels. Journal of Food Processing and Preser- O’Sullivan, M. M., Kelly, A. L., & Fox, P. F. (2002). Influence of transglutaminase vation, 32,1016e1033. treatment on some physico-chemical properties of milk. Journal of Dairy Aroonkamonsri, J. (1996). The role of plasmin and bacterial proteinases on age gela- Research, 69, 433e442. tion of UHT milk (Pseudomonas fluorescens, Pseudomonas fragi, Pseudomonas Pereira, R., Matia-Merino, L., Jones, V., & Singh, H. (2006). Influence of fat on the luteola, Vibrio fluvialis). PhD Thesis. Guelph, ON, Canada: University of Guelph. perceived texture of set acid milk gels: a sensory perspective. Food Hydrocol- Bastian, E. D., Brown, R. J., & Ernstrom, C. A. (1991). Casein interference in bovine loids, 20, 305e313. plasmin assays using a synthetic substrate. Journal of Dairy Science, 74,4119e4124. Prado, B. M., Ismail, B., Ramos, O., & Hayes, K. D. (2007). Thermal stability of plas- Bonfatti, V., Grigoletto, L., Cecchinato, A., Gallo, L., & Carnier, P. (2008). Validation of minogen activators and plasminogen activation in heated milk. International a new reversed-phase high-performance liquid chromatography method for Dairy Journal, 17, 1028e1033. V.M. Rauh et al. / International Dairy Journal 38 (2014) 199e207 207

Prado, B. M., Sombers, S. E., Ismail, B., & Hayes, K. D. (2006). Effect of heat treatment Rollema, H. S., Visser, S., & Poll, J. K. (1983). Spectrophotometric assay of plasmin on the activity of inhibitors of plasmin and plasminogen activators in milk. and plasminogen in bovine milk. Milchwissenschaft, 38,214e217. International Dairy Journal, 16, 593e599. Saint-Denis, T., Humbert, G., & Gaillard, J. L. (2001). Heat inactivation of native Precetti, A. S., Oria, M. P., & Nielsen, S. S. (1997). Presence in bovine milk of two plasmin, plasminogen and plasminogen activators in bovine milk: a revisited protease inhibitors of the plasmin system. Journal of Dairy Science, 80, 1490e study. Lait, 81,715e729. 1496. Tolkach, A., & Kulozik, U. (2007). Reaction kinetic pathway of reversible and irre- Rauh, V., Bakman, M., Ipsen, R., Paulsson, M., Kelly, A. L., Larsen, L. B., et al. (2013). versible thermal denaturation of blactoglobulin. Lait, 87,301e315. _ The determination of plasmin and plasminogen-derived activity in turbid Topçu, A., Numanoglu, E., & Saldamlı, I. (2006). Proteolysis and storage stability of samples from various dairy products using an optimized spectrophotometric UHT milk produced in Turkey. International Dairy Journal, 16, 633e638. method. International Dairy Journal. submitted for publication. Torres, I. C., Amigo Rubio, J. M., & Ipsen, R. (2012). Using fractal image analysis to Richardson, B. C. (1983). The proteinases of bovine milk and the effect of pasteur- characterize microstructure of low-fat stirred yoghurt manufactured with ization on their activity. New Zealand Journal of Dairy Science and Technology, 18, microparticulated whey protein. Journal of Food Engineering, 109,721e729. 233e245. Van Asselt, A. J., Sweere, A. P. J., Rollema, H. S., & de Jong, P. (2008). Extreme high- Rollema, H. S., & Poll, J. K. (1986). The alkaline milk proteinase system: kinetics and temperature treatment of milk with respect to plasmin inactivation. Interna- mechanism of heat-inactivation. Milchwissenschaft, 41, 536e540. tional Dairy Journal, 18,531e538. Supplementary data

S. 1. Microstructure of gel particles in milk pre-heat treated at 72 °C for 180 s and subjected to direct steam infusion (>150 °C for <0.2 s) after 13 wk of storage at 20 °C. Green areas indicate protein stained with FITC and red areas fat stained with Nile red (red). The scale is indicated in the micrographs.

Paper III

Rauh, V. M., Johansen, L. B., Ipsen, R., Paulsson, M., Larsen, L. B., & Hammershøj, M. (2014)

Plasmin Activity in UHT Milk: Relationship between Proteolysis, Age Gelation, and Bitter- ness

Journal of Agricultural and Food Chemistry, 62, 6852-6860

Errata First paragraph in materials and methods on page 6853, first column, should say:

Milk Processing. Three batches of 1 day old pasteurized (72 °C for 15 s) and homogenized milk (1.5% fat) were obtained from Arla Foods Stockholm Dairy (Sweden) during three consecutive weeks. The milk was preheat treated at 72 °C for 180 s in a tubular heat exchanger before being subjected to a direct steam infusion heat treatment at >150 °C for <0.2 s (UHT pilot plant, APV/SPX, Silkeborg, Denmark). After flash cooling to 67 °C, the milk was homogenized (pressure 160/40 bar), further cooled to 20 °C, and packed aseptically in 300 mL glass bottles. The UHT milk was stored in the dark at 20 °C in a climate chamber for 14 weeks. One bottle was used once every week of the storage period for analysis. The sterility of each batch of processed milk was tested according to ISO 483324.

Article

pubs.acs.org/JAFC

Plasmin Activity in UHT Milk: Relationship between Proteolysis, Age Gelation, and Bitterness Valentin M. Rauh,*,†,‡ Lene B. Johansen,† Richard Ipsen,§ Marie Paulsson,∥ Lotte B. Larsen,‡ and Marianne Hammershøj‡

† Arla Foods Strategic Innovation Centre, Rørdrumvej 2, DK-8220 Brabrand, Denmark ‡ Faculty of Science and Technology, Department of Food Science, Aarhus University, Blichers Allé20, DK-8830 Tjele, Denmark § Department of Food Science, University of Copenhagen, Rolighedsvej 30, DK-1958 Frederiksberg, Denmark ∥ Department of Food Technology, Engineering and Nutrition, Lund University, SE-221 00 Lund, Sweden

*S Supporting Information

ABSTRACT: Plasmin, the major indigenous protease in milk, is linked to quality defects in dairy products. The specificity of plasmin on caseins has previously been studied using purified caseins and in the indigenous peptide profile of milk. We investigated the specificity and proteolytic pathway of plasmin in directly heated UHT milk (>150 °C for <0.2 s) during 14 ° α β weeks of storage at 20 C in relation to age gelation and bitter peptides. Sixty-six peptides from S- and -caseins could be attributed to plasmin activity during the storage period, of which 23 were potentially bitter. Plasmin exhibited the highest affinity for the hydrophilic regions in the caseins that most probably were exposed to the serum phase and the least affinity for hydrophobic or phosphorylated regions. The proteolytic pattern observed suggests that plasmin destabilizes the casein micelle by hydrolyzing casein−casein and casein−calcium phosphate interaction sites, which may subsequently cause age gelation in UHT milk. KEYWORDS: age gelation, bitterness, casein, plasmin, proteolysis, UHT milk

■ INTRODUCTION and the major whey protein, β-lactoglobulin (β-lg). The fi β α α speci city of PL toward -, S1- and S2-CN in model studies Proteolysis by plasmin (PL), the major indigenous milk fi 15−17 fi protease, can greatly affect the quality of dairy products. using puri ed caseins has been described. The speci city of PL toward caseins in a micellar system has been studied in Proteolysis of caseins by plasmin leads to the formation of fi 18,19 bitter off-flavours in milk and has been linked to the age relation to the endogenous peptide pro le of milk, but not gelation of UHT milk, making it a critical factor for the shelf life in relation to extensive proteolysis and quality defects of dairy − fi of these products.1 3 While PL is an important factor for the products. The proteolytic speci city of PL could give an insight ripening characteristics of some cheeses,4 PL-induced proteol- into the mobility and interaction of PL with the casein micelles. fi 5,6 Age gelation of UHT milk is previously suggested to be ysis can also reduce cheese yield and curd rmness. κ β PL is the active part of a complex enzyme system in milk caused by complexation of -CN with denatured -lg during consisting of, apart from PL itself, its zymogen, plasminogen the heat treatment of milk. A subsequent release and (PLG), which is present in a 6−10 times higher concentration aggregation of this complex during storage of UHT milk could then result in gelation.20 Proteolysis is hypothesized to than PL, as well as plasminogen activators (PA) and inhibitors κ −β of both PL and PA.7 In contrast to the inhibitors, both PL, facilitate the release of the -CN -lg-complex by hydrolyzing − PLG, and PA exhibit a high thermal stability.8 10 Heat the caseins that anchor the complex on the surface of the casein micelle.21 treatment of milk thus promotes PL-induced proteolysis due ff to both reduced inhibitory capacity for residual PL and to In a recent study, we investigated the e ect of PL activity on physicochemical changes in milk heated by direct steam increased activation of PLG into PL. The caseins (CN) in milk 1 are structurally present as colloidal particles (casein micelles), infusion, where we found that proteolysis by residual PL rich in calcium phosphate nanoclusters which are surrounded resulted in a decrease in pH and whiteness of the milk during by the phosphorylated caseins. The casein micelles are storage. Simultaneously, PL activity increased during the storage period, while PLG-derived activity decreased. This internally stabilized by ionic interactions with calcium β α α phosphate and hydrophobic interactions between caseins, correlated with an increase in the rate of -, S1-, and S2-CN ff fl while the protruding κ-casein on the surface sterically stabilizes hydrolysis. Bitter o - avor developed after 7 weeks of storage, the casein micelle externally against aggregation.11 In milk, PL and age gelation of the milk was observed after 11 weeks. At is associated with the casein micelle, ensuring a close proximity to its substrate.12,13 The most susceptible casein for PL in milk Received: May 5, 2014 is β-CN, giving rise to the well described γ-CN fraction and Revised: June 23, 2014 14 α proteose peptones. PL also readily cleaves S2-CN and to a Accepted: June 25, 2014 α κ lesser extent S1-CN, but has little or no activity toward -CN Published: June 25, 2014

© 2014 American Chemical Society 6852 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article α Table 1. Presence and Identity of pH 4.6 Soluble Peptides from S1-CN Formed by Plasmin in Directly Heated UHT Milk; Peptides with Q Values >1400 Are Potentially Bitter and Are Shown in Bold

storage time peptidea mass day 1 4 weeks 8 weeks 14 weeks Q valueb α − S1-CN f(1 7) 874.55 x x x 1777 α − S1-CN f(1 22) 2615.48 x x x 1297 α − S1-CN f(23 34) 1383.72 x x x 1833 α − S1-CN f(23 36) 1640.86 x x 1718 α − S1-CN f(35 42) 945.51 x x 1030 α − S1-CN f(80 90) 1336.67 x x x x 1054 α − S1-CN f(80 100) 2585.37 x 1342 α − S1-CN f(80 102) 2826.54 x 1396 α − S1-CN f(91 100) 1266.70 x x 1660 α − S1-CN f(91 102) 1507.88 x x 1710 α − S1-CN f(91 103) 1635.97 x x 1694 α − S1-CN f(91 105) 1927.14 x 1759 α − S1-CN f(100 105) 678.45 x x 1753 α − S1-CN f(101 124) + P 2916.53 x 1358 α − S1-CN f(103 124) + P 2675.35 x x x 1304 α − S1-CN f(104 119) + P 1950.95 x 1339 α − S1-CN f(104 124) + P 2547.26 x x x 1294 α − S1-CN f(106 119) + P 1659.79 x x 1218 α − S1-CN f(106 124) + P 2256.10 x x x 1201 α − S1-CN f(120 124) 614.32 x x 1152 α − S1-CN f(125 132) 909.47 x 756 α − S1-CN f(194 199) 747.36 x x x 1703 aPhosphorylations are indicated after the peptide. bQ value was calculated according to Ney (1971) in cal mol−1. Presence of peptides at different storage times is indicated by ‘x’. this time more than 90% of α- and β-CNs were hydrolyzed. of-flight mass spectrometer (Q-TOF MS, Agilent 6530 Accurate Mass The relatively low degree of β-lg denaturation in the UHT milk Q-TOF LC/MS, Agilent Technologies) with a Jetstream interface. μ of 36.7 ± 3.2% indicated that the age gelation may be caused by Depending on the peptide concentration, 10 or 20 L of the sample ff κ was injected to the uHPLC-Q-TOF-MS system. The column was an adierent mechanism than complexation of -CN with μ × Xbridge BEH 300 (rp-C18, 3.5 m 2.1 mm 250 mm) (Waters, denatured β-lg. At this level of β-lg denaturation, only a very ° ff β κ 22,23 Milford, MA, USA) operated at 45 C. Bu er A was Milli-Q water with limited fraction of -lg is associated with -CN. 0.05% trifluoroacetic acid (TFA) (v/v and buffer B acetonitrile with The aim of the present study was to analyze the peptide 0.1% (v/v) TFA. A linear gradient was applied with 100% buffer A formation and specificity of PL in this UHT milk and to relate from 3 min, 100% to 45% buffer A at 45 min at a flow rate of 0.3 mL the peptides found with the previously observed bitter off-flavor min−1. Peptides were monitored at 214 nm. and age gelation of the UHT milk during storage. The Q-TOF MS was run in 4 GHz high resolution mode. The Jetstream interface added a sheath gas at 350 °C with a flow of 8 L −1 ■ MATERIALS AND METHODS min . A nozzle voltage of 100 V was applied to improve the ionization. The drying gas temperature was set to 325 °Cataflow rate Milk Processing. Three batches of 1 day old pasteurized (72 °C of 10 L min−1. The capillary voltage was set to 2.5 kV in the first time for 15 s) and homogenized milk (1.5% fat) were obtained from Arla segment (3.0−24.0 min) and to 4.0 kV in the second time segment Foods Stockholm Dairy (Sweden) during three consecutive weeks. (24.0−35.2 min) to achieve optimal ionization for both small and large The milk was preheat treated at 74 °C for 180 s in a tubular heat peptides. Spectra in the mass range from 300 to 3200 m/z were exchanger before being subjected to a direct steam infusion heat recorded in positive mode with a resolution of 20000 in the Q-TOF. treatment at >150 °C for <0.2 s (UHT pilot plant, APV/SPX, Tandem mass spectra of the two most abundant ions with a charge Silkeborg, Denmark). After flash cooling to 67 °C, the milk was state of 1, 2, or 3 in each MS spectrum were obtained in collision- homogenized (pressure 160/40 bar), further cooled to 20 °C, and induced dissociation (CID) mode (collision gas nitrogen, isolation packed aseptically in 300 mL glass bottles. The UHT milk was stored width 1.3 m/z (narrow) at 5 different fixed collision energies and in the dark at 20 °C in a climate chamber for 14 weeks. One bottle was flexible collision energies optimized for larger peptides. The fixed used once every week of the storage period for analysis. The sterility of collision energies were 10, 20, 30, 50, and 100 V, and the flexible each batch of processed milk was tested according to ISO 4833.24 energy settings were based on charge state and m/z (singly charged: Sample Preparation. A volume of 15 mL of milk was taken from offset 18 V, slope 3 V m/z−1; doubly charged: offset 2 V, slope 2.2 V each bottle and adjusted to pH 4.6 with 1 M HCl followed by m/z−1; triply charged: offset 3.5 V, slope 1.5 V m/z−1). Mass spectra centrifugation at 11000g for 10 min at 4 °C. The clear supernatant was and UV chromatograms were analyzed by MassHunter software recovered and stored at −20 °C until further analysis. At the time of (version B 6.01, Agilent Technologies).The MS spectra were analysis, the samples were thawed for 60 min at room temperature. investigated for potential peptides formed by plasmin, which were One mL of the sample was centrifuged at 13200g at 5 °C for 10 min not identified by MS/MS, using the “Find by Formula” feature in and used for the analysis. Sample preparation was performed in MassHunter. Matching ions and isotopic patterns were further duplicate. evaluated with regard to their retention time. Peptide Analysis. Ultra-high-performance liquid chromatography Database Search. The obtained tandem mass spectra were (uHPLC) was performed on an Agilent 1290 Infinity uHPLC (Agilent analyzed with Mascot v2.4 (Matrix Science, London, UK). Tandem Technologies, Santa Clara, CA, USA) coupled to a quadrupole-time- mass spectra were searched against a custom database containing

6853 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article known genetic and PTM (post translation modification) variants of major milk proteins. The following search parameters were applied: protease: unspecific; mass tolerance for precursor ions 15 ppm and 0.6 Da for product ions; variable modifications: phosphorylations serine, oxidation methionine. Only identified peptides (expect score <0.05) that were present in at least two of the three replicates were considered. Relative Quantification of Identified Peptides. Peaks in the UV chromatogram were identified by their corresponding MS spectra and retention time of the identified peptides by MS/MS. The retention time shift between the UV and MS signal was approximately 0.25 min. Only peaks without or with only minor overlaps with other components in the MS spectra were considered. In order to compare the rate of peptide formation, the integrated UV peak area was divided by a calculated molar extinction coefficient (ε) for the peptide as 25 α described previously. Phosphorylation of serine was not taken into Figure 1. Primary structure of bovine S1-CN B-8P. Phosphoserine account. Peak areas were further normalized for the initial casein residues are indicated by circles. Open circle indicates a possible ff fi distribution to correct for the di erences in peptide concentration. phosphorylation at Ser41. Identi ed pH 4.6 soluble peptides formed by The casein composition was based on the UV peak area corrected for plasmin during storage at 20 °C are indicated by arrows. ε α α β κ the calculated ( S1-CN: 30%, S2-CN: 13%, -CN: 44%, -CN: ε −1 −1 α α 13%). Calculated in M cm were 346843 ( S1-CN), 375032 ( S2- CN), 331417 (β-CN) and 279859 (κ-CN). indicating that Lys79-His80 and Lys124-Glu125 are the two major Assessment of Bitterness. Bitterness of peptides was evaluated α α ’ 26 cleavage sites in the central region of S1-CN. The peptides S1- by applying Ney s Q-rule based on peptide hydrophobicity. An − α − α − average hydrophobicity Q of the peptide is calculated as the sum of the CN f(23 34), S1-CN f(23 36), and S1-CN f(35 42) appeared after 4−8 weeks of storage. This implies a sequential amino acids side chains hydrophobicity divided by the number of α amino acid residues of the peptide. Peptides with a Q value >1400 cal/ cleavage of the N-terminal region of S1-CN by PL beginning at −1 res are considered bitter Arg22-Phe23 and subsequently Lys34-Glu35 and Lys36-Val37. Peptides containing these cleavage sites Arg100-Leu101, Arg119- ■ RESULTS AND DISCUSSION Leu120, and Lys132-Glu133 occur only after 8 or 14 weeks of storage and seem not to be preferred by PL. The late or slow ff fi A total of 84 di erent casein-derived peptides were identi ed cleavage at these positions is most likely due to the preference by the uHPLC-Q-TOF analysis (Tables 1−4). In the 1 day old of PL for cleaving after lysine rather than after arginine29 and milk (analyzed 1 day after processing), only 11 peptides could their close proximity to the major cleavage sites makes a further be identified by MS/MS. This number is small compared to hydrolysis in the vicinity unlikely. other recent studies, which used more profound peptide The identified cleavage sites of PL and the order of 18,19 extraction and concentration methods. The main objective occurrence are in good agreement with previous studies using fi α 17,29 in this study, however, was to study the proteolysis and puri ed S1-CN with the exception of the cleavage sites at development of peptides during storage rather than the peptide Arg151-Gln152, a proposed major cleavage site by McSweeney et profile of the milk used for the UHT treatment. The majority al.,17 and at Lys -Gln . Relevant peptides and polypeptides fi α 58 59 by number of the peptides identi ed were derived from S2-CN containing the cleavage site Arg -Gln (α -CN f(125−151), α β 151 152 S1 (36 peptides), followed by S1-CN (31) and -CN (17). α -CN f(133−151), α -CN f(106−151), and α -CN f(152− κ S1 S1 S1 During the entire storage period, no peptides derived from - 199)) were not detected. The cleavage site Arg151-Gln152 is α CN were detected, indicating that indigenous proteases other located in middle of the most hydrophobic region of S1-CN than PL, such as cathepsin D, were probably heat inactivated as (Figure 4A), which has been identified as the main polymer- α could be expected due to their lower heat stability compared ization and interaction region of S1-CN via hydrophobic 30 fi α with PL and also that the raw material had not contained heat interactions. In contrast to puri ed S1-CN, this region is stable proteases from psychotrophic bacteria, which are known likely to be inaccessible for PL and Arg151-Gln152 and was also to be able to hydrolyze κ-CN.27,28 More than 75% of the not hydrolyzed by trypsin in a micellar system.31,32 The second peptides (66 peptides out of 79) could be assigned to be the not-present cleavage site, Lys58-Gln59, is placed between the α − result of PL action solely (Table 4). In the remaining 18 two phosphoserine clusters of S1-CN (residues 41 51 and identified peptides, PL was responsible for either the N- or C- 61−70). The phosphoserine residues tightly bind to the terminal cleavage in 11 of the 18 identified peptides. calcium phosphate clusters in the casein micelle, which makes Identified Peptides Formed by Plasmin. Identified this region inaccessible for enzymatic hydrolysis.31,33 α peptides derived from S1-CN and the storage time of their The largest numbers of peptides present as a result of PL α appearance is shown in Table 1. The position of the peptides activity derives from S2-CN (Table 2). After 1 day of storage, α α within in the S1-CN sequence is shown in Figure 1. The only peptides originating from the N- and C-terminal region of S2- α − α peptide present after 1 day of storage was S1-CN f(80 90), CN were present (Figure 2). The C-terminal region of S2-CN suggesting Lys79-His80 and Arg90-Tyr91 to be the initial sites has previously been shown to be exposed and easily accessible fi 34 α − cleaved by PL. After 4 weeks of storage, the peptides identi ed for hydrolysis, and the appearance of S2-CN f(182 188) and α α − originated from the N-terminal region of S1-CN (Arg22-Phe23 S2-CN f(174 181) after 4 weeks of storage suggests a and Lys7-His8), from the central, lysine-rich region (Lys102- sequential hydrolysis of the C-terminal region. Initial cleavage Lys103, Lys103-Tyr104, Lys105-Val106, and Lys124-Glu125) as well as sites are Lys197-Thr198 and Lys188-Ala189, followed by Lys181- the C-terminal end with cleavage at Lys193-Thr194. The peptides Thr182 and Lys173-Phe174. After the same storage time, peptides α − α − α − α S1-CN f(91 102), S1-CN f(91 103), and S1-CN f(91 105) originating from the central region of S2-CN were formed by fi were not identi ed until after 8 or 14 weeks of storage, hydrolysis of Arg114-Asn115, Lys136-Lys137, Lys137-Thr138, Lys149-

6854 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article α Table 2. Presence and Identity of pH 4.6 Soluble Peptides from S2-CN Formed by Plasmin in Directly Heated UHT Milk; Peptides with Q Values >1400 Are Potentially Bitter and Are Shown in Bold

storage time peptidea mass day 1 4 weeks 8 weeks 14 weeks Q valueb α − S2-CN f(1 21) + 4P 2745.99 x x x x 880 α − c S2-CNf(1 24) + 4P 3131.19 x x x x 851 α − d S2-CN f(25 45) 2380.15 x 1019 α − S2-CN f(71 80) 1245.64 x x 1245 α − S2-CN f(115 125) 1194.67 x 1331 α − S2-CN f(115 136) + 2P 2587.16 x x x 939 α − S2-CN f(115 137) + 2 P 2715.25 x x 963 α − S2-CN f(115 149) + 3 P 4162.85 x x 971 α − c S2-CN f(115 150) + 3 P 4290.94 x x x 986 α − S2-CN f(137 149) + P 1593.70 x x x 1025 α − S2-CN f(137 150) + P 1721.80 x x x 1059 α − S2-CNf(138 149) + P 1465.61 x x x 986 α − S2-CN f(138 150) + P 1593.70 x x x 1025 α − S2-CN f(150 165) 1990.12 x x 1215 α − S2-CN f(151 165) 1862.03 x x 1177 α − S2-CN f(151 166) 1990.12 x x 1197 α − S2-CN f(151 173) 2893.61 x x 1177 α − S2-CN f(153 165) 2632.88 x 1208 α − S2-CN f(166 173) 1049.60 x x 1176 α − S2-CN f(167 173) 921.50 x x 1130 α − S2-CN f(171 181) 1397.77 x x 1762 α − S2-CN f(174 181) 978.55 x x x 1846 α − S2-CN f(174 188) 1863.00 x x 1463 α − S2-CN f(182 188) 902.46 x x x 971 α − S2-CN f(182 197) 1982.06 x x x 1434 α − S2-CN f(182 207) 3214.78 x x 1644 α − S2-CN f(189 197) 1097.61 x x x 1793 α − S2-CN f(189 199) 1326.75 x x 1644 α − S2-CNf(189 207) 2330.33 x x x x 1892 α − S2-CNf(198 205) 974.59 x x 1814 α − S2-CNf(198 207) 1250.74 x x x x 1980 α − S2-CNf(200 207) 1021.60 x x x 2233 aPhosphorylations are indicated after the peptide. bQ value was calculated according to Ney (1971) in cal mol−1. cPeptide was identified by MS only. dIntramolecular disulfide bond. Presence of peptides at different storage times is indicated by ‘x’.

Lys150, and Lys150-Thr151. Earlier studies suggest that the cleavage sites Lys149-Lys150 and Lys150-Thr151 are preferred over 16 α Arg114-Asn115, but the simultaneous appearance of S2-CN − α − α − f(115 136), S2-CN f(137 149), and S2-CN f(138 150) indicate that PL readily hydrolyzed all of these cleavage sites. After 8 weeks of storage peptides originating from the region − α fi 151 173 in S2-CN were identi ed. The late appearance of these peptides indicates a sequential hydrolysis of this region from an N- and C-terminal direction after the preferred cleavage sites Lys150-Thr151 and Lys173-Phe174 are hydrolyzed. α The least preferred cleavage sites in S2-CN seem to be Arg125- Glu126 and Lys152-Leu153, which are either in close proximity to phosphoserine residues or to major cleavage sites (Figure 2). Although the region 25−114 contains eight potential cleavage α − α − sites for PL, only S2-CN f(71 80) and S2-CN f(25 45) could be identified. Masses corresponding to peptides including α Figure 2. Primary structure of bovine S2-CN A-11P. Phosphoserine the cleavage site Lys91-Phe92, Lys32-Glu33, and Lys41-Glu42 were residues are indicated by circles. The dotted line between Cys36 and not present in the milk. The not-present cleavage at Lys91-Phe92 fi fi α Cys40 indicates the intramolecular disul de bridge. Identi ed pH 4.6 is located in a hydrophobic region of S2-CN (Figure 4B), soluble peptides formed by plasmin during storage at 20 °C are which may be inaccessible for enzymatic hydrolysis.35 Cleavage indicated by arrows. of this bond can be additionally hindered by Pro93 which α 35 induces a change in the secondary structure of S2-CN. Lys32- Glu33 and Lys41-Glu42 are located in the proximity of the two

6855 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article

Table 3. Presence and Identity of pH 4.6 Soluble Peptides from β-CN Formed by Plasmin in Directly Heated UHT Milk; Peptides with Q Values >1400 Are Potentially Bitter and Are Shown in Bold

storage time peptidea mass day 1 4 weeks 8 weeks 14 weeks Q valueb β-CN f(1−25) + 4P 3121.26 x x 1006 β-CN f(1−28) + 4P 3476.48 x x x x 1057 β-CN f(1−29) + 4P 3604.58 x x x x 1072 β-CN f(29−48) + P 2559.13 x x x x 818 β-CN f(33−48) + P 2060.82 x x 614 β-CN f(98−105) 872.48 x x x 1328 β-CN f(98−107) 1137.63 x x x x 1262 β-CN f(100−105) 645.32 x x x 1238 β-CN f(100−107) 910.47 x x x 1179 β-CN f(106−113) 1012.52 x x x 1655 β-CN f(108−113) 747.36 x x x 1873 β-CN f(170−176) 779.49 x x x 1777 aPhosphorylations are indicated after the peptide. bQ value was calculated according to Ney (1971) in cal mol−1. Presence of peptides at different storage times is indicated by ‘x’. α α cysteine residues of S2-CN. The majority of the S2-CN Peptides Formed by Other Proteases. Peptides that molecules is present in a form with an intramolecular disulfide were identified in the UHT milk samples and which cannot be bond, which induces a tight loop in the protein34,36 which may assigned to be a result of PL activity alone are summarized in α β limit the accessibility of Lys41-Glu42 and Lys32-Glu33 for PL. To Table 4. Most peptides deriving from S1-CN or -CN indicate 28,38,39 our knowledge, the cleavage of Arg45-Asn46, Lys70-Ile71, Lys80- activity of cathepsins alone or in combination with PL. Ile , Arg -Glu , Lys -Ile , and Arg -Tyr has not been Peptides formed by cathepsins alone were identified after 1 day 81 125 126 166 167 205 206 α reported earlier for PL in either model studies or as part of the of storage or after 4 weeks of storage in the case of S1-CN indigenous peptide profile in milk. This could be attributed to f(1−14). Peptides formed by the combined activity of α ff cathepsins and PL on the other hand appear after 4−8 weeks the low concentration of S2-CN in milk or a di erent α 31,35 of storage. This indicates that cathepsins were active in the raw conformation of S2-CN in model systems. α or pasteurized milk, but inactivated by the UHT treatment and In comparison to S2-CN, only a relatively small number of peptides was identified derived from β-CN (Table 3). The that the polypeptides formed by cathepsins were further peptides β-CN f(1−28), β-CN f(1−29), β-CN f(29−48), and hydrolyzed by residual PL to smaller peptides. With the β-CN f(98−107) were present in the milk after 1 day of exception of cathepsin D, the heat stability of cathepsins is largely unknown. Although cathepsin D in its active form can storage. After 4 weeks of storage, all the observed peptides 40 β − β − partially survive HTST pasteurization, it is unlikely that except -CN f(33 48) and -CN f(1 25) were present. The ffi identified peptides mainly derived from cleavage around the su cient active cathepsins are present after the UHT treatment to significantly contribute to the observed proteolysis. The two major cleavage sites of PL in β-CN, Lys -His , and 105 106 peptides deriving from α -CN originate from the C-terminal of Lys -Glu (Figure 3). The peptide β-CN f(170−176) was S2 107 108 the protein. With the exception of α -CN f(201−206), the the only identified peptide originating from the hydrophobic C- S2 β peptides contained the PL cleavage site Lys197-Thr198. The terminal part of -CN and in contrast to earlier studies using fi 31,37 fi apparently unspeci c C-terminal cleavage site of the peptides trypsin, no peptides were identi ed from hydrolysis of suggests activity of carboxypeptidases in the raw milk or heat- Arg183-Asp184 and Arg202-Gly203. induced hydrolysis.41,42 Peptide Development during Storage. The presence of identified peptides after different storage periods gives a good indication of preferred and uncleaved cleavage sites of PL in the UHT milk system, but only gives limited information on the actual concentration and development of the peptides. Semiquantitative analysis of the peptides using MS spectra was not possible because their response factors are unknown and their intensity might be affected by ion suppression of coeluting components. The UV chromatograms at 214 nm were therefore used for a relative quantification and comparison of the peptides. Due to substantial overlap of large areas in the chromatogram, only a limited number of peaks could be used for such relative quantification. Quantified peptides included α − α − α − α S1-CN f(1 22), S1-CN f(106 124), S2-CN f(1 21), S2- − α − β − CN f(1 24), S2-CN f(115 150), as well as -CN f(1 28) β − Figure 3. Primary structure of bovine β-CN A2-5P. Phosphoserine and -CN f(29 48). A comparison of the UV peaks and the residues are indicated by circles. Identified pH 4.6 soluble peptides extracted ion chromatograms are shown in the Supporting formed by plasmin during storage at 20 °C are indicated by arrows. Information together with MS spectra of the peaks.

6856 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article

Table 4. Presence and Identity of pH 4.6 Soluble Casein Derived Peptides Formed by Other Proteases in Directly Heated UHT Milk; Peptides with Q Values >1400 Are Potentially Bitter and Are Shown in Bold

storage time proteaseb peptide mass day 1 4 weeks 8 weeks 14 weeks Q valuea N-terminal C-terminal α − − S1-CN f(1 8) 1011.62 x x x 1618 ? α − − S1-CN f(1 14) 1663.93 x x x 1309 CB α − − S1-CN f(1 21) 2459.38 x x x x 1324 CG α − − S1-CN f(1 24) 2909.61 x x x 1410 CB,CD α − S1-CN f(3 7) 621.40 x 1598 CB PL α − S1-CN f(24 36) 1089.59 x x 1646 CB,CD PL α − S1-CN f(25 36) 1236.66 x x 1563 CB,CD PL α − − S1-CN f(180 199) 2215.05 x x x x 1232 CG α − − S1-CN f(185 199) 1688.81 x x x 1232 ? α − S2-CN f(198 202) 556.36 x x x 1844 PL CP α − S2-CN f(198 203) 719.42 x x x 2015 PL CP α − S2-CN f(198 204) 818.49 x x 1969 PL CP α − S2-CN f(201 206) 809.44 x 2292 ? CP β-CN f(96−105) 1087.61 x x x 1180 CD,CG PL β-CN f(96−107) 1352.76 x x 1262 CD,CG PL β-CN f(106−125) 2390.16 x x 1482 PL CD β-CN f(108−125) 2125.00 x x 1535 PL CD β-CN f(163−169) 761.43 x 909 CB,CD PL aQ value was calculated according to Ney (1971) in cal mol−1. bPossible responsible proteases for N- and C-terminal cleavage are indicated: CB, cathepsin B; CD, cathepsin D; CG, cathepsin G; PL, plasmin; CP, carboxypeptidase; −, no cleavage; ?, unknown protease. Presence of peptides at diﬀerent storage times is indicated by ‘x’.

A comparison of peptides originating from the N-terminus of f(115−150) originates from two of the most preferred cleavage β α their parent proteins is shown in Figure 5. The peptide -CN sites in S2-CN and showed the highest formation rate. The − fi − α − f(1 28) had the highest formation rate. During the rst 4 5 peptide S1-CN f(106 124) developed almost in the same α − fi weeks of storage, the relative concentration increased manner as S1-CN f(1 22) within the rst 6 weeks of storage exponentially, reflecting the previously described increase in but showed a higher rate of formation from 7 weeks of storage PL activity and rate of peptide formation. From 5 weeks of on. These findings further highlight the preference of PL for storage and further on, the relative concentration of β-CN f(1− cleavage sites located in the hydrophilic regions of the caseins. 28) increased almost linearly and then leveled out after 12 Gagnaire et al. (1998) found that the proteose peptones β-CN weeks of storage. The decrease in the rate of formation could f(1−105/7) were retained in the casein micelle, while β-CN be linked to changes in the matrix induced by the starting f(29−105/7) was released into the serum phase.31 This release gelation of the milk after 11 weeks of storage. Further limits further hydrolysis of these proteose peptones by the degradation of β-CN f(1−28) to β-CN f(1−25) may also casein-associated PL, which explains why β-CN f(29−48) was reduce the concentration. Since the proteose peptones β-CN only formed in very small amounts. f(1−105) and β-CN f(1−107) were still present after 14 weeks Destabilization of Casein Micelles by Proteolysis. PL of storage (data not shown), substrate depletion can be enters the milk from the blood44 and associates itself with the excluded as a cause for the reduced rate of formation. The casein micelle.45 Whether PL is able to enter the micellar α − peptide S2-CN f(1 24) showed the same initial rate of structure or is located entirely on the surface of the casein formation as β-CN f(1−28) and leveled off after 6 weeks of micelle is unknown. It becomes obvious from Figure 3 that the α − storage. The peptide S2-CN f(1 21) exhibited a lower rate of majority of the major cleavage sites of PL within the caseins, α − α 31,32 formation than S2-CN f(1 24), but in contrast to S2-CN similar to those of trypsin, are located at the most f(1−24), continued to increase after 6 weeks of storage. This hydrophilic and serum-exposed regions of the proteins. β-CN is α − α − 46 indicates a hydrolysis of S2-CN f(1 24) to S2-CN f(1 21) mainly present in the interior of the casein micelle and is that may be caused by substrate depletion of the cleavage site proposed to stabilize water or serum channels within the casein α − ff 47,48 α Lys24-Asn25. The development of S1-CN f(1 22) di ered micelle. S2-CN is involved in the binding to calcium from the other N-terminal peptides. The initial lag phase was phosphate clusters, regions that are hypothesized to be α − 11 not observable for S1-CN f(1 22), and the overall surrounded by hydrophobic domains and likely to be concentration was significantly lower compared to that of the inaccessible for PL. The rapid cleavage, within 4 weeks of α β β α N-terminal peptides derived from S2- and -CN. The cleavage storage, of the N-terminal regions of -CN and S2-CN implies α β sites for these peptides in S2- and -CN are in very hydrophilic that PL is able to penetrate the casein micelle. In relation to a domains of the proteins (Figure 4 A−C), while the cleavage site destabilization of the casein micelle by proteolysis, it is notable α Arg22-Phe23 in S1-CN is in a hydrophobic domain involved in that the minor cleavage sites, together with some of the major casein−casein interactions.43 cleavage sites, are located at the boundary between hydrophilic, Peptides being present as the result of two proteolytic hydrophobic, or phosphorylated regions of the caseins (Figure cleavages are shown in Figure 6. They exhibited an extended lag 4A−C). Thus, PL hydrolyses around the regions essential for phase compared to the N-terminal peptides, due to the time the internal integrity and stabilization of the casein micelle, i.e. α needed to form their precursors. As discussed above, S2-CN hydrophobic interaction between caseins, interactions of

6857 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article

Figure 5. Development of casein-derived N-terminal peptides formed by plasmin in milk preheat treated at 72 °C for 180 s and subjected to direct steam infusion (>150 °C for <0.2 s) during storage at 20 °C. Relative concentration is based on peak area of UV absorbance at 214 nm corrected for the peptides theoretical molar extinction at 214 nm. α Peptide concentration is normalized for S2-CN.

α α Figure 4. Hydrophobicity distribution in S1-CN (A), S2-CN (B), and β-CN (C) with major (●) and minor (○) cleavage sites of plasmin and phosphoserine clusters (*). Classification of major and minor cleavage sites is based on Tables 1−3. Hydrophobicity was calculated according to Tanford (1962) using an average of seven amino acids with a relative weight for the amino acid center being one Figure 6. Development of casein-derived peptides formed by plasmin in milk preheat treated at 72 °C for 180 s and subjected to direct steam and 0.75, 0.5, and 0.25 for the amino acids one, two, or three positions ° ° from the center amino acid. Positive values indicate hydrophobic and infusion (>150 C for <0.2 s) during storage at 20 C. Relative negative values, hydrophilic regions. concentration is based on peak area of UV absorbance at 214 nm corrected for the peptides theoretical molar extinction at 214 nm. α Peptide concentration is normalized for S2-CN. caseins with calcium phosphate, and ionic or salt interactions between caseins. Cleavage of these regions in the casein micelle were hydrolyzed completely. This indicates that a large number would lead to a loosening of the micelle structure, which of amphiphilic and charged polypeptides were present in the correlates with the reported increase in micelle size after milk system that could form and stabilize a gel network. While incubation with PL.49 A looser structure is hypothesized to some proteolysis is needed for gelation, the extent of allow PL to reach previously inaccessible regions, which can proteolysis has not yet been related with gelation. The lack α explain the initial resistance of S1-CN to hydrolysis. of consistency in degree of proteolysis in relation to age Age gelation of UHT milk has been proposed to be a two- gelation and PL activity described in previous studies could also stage process of proteolysis and physicochemical changes that be attributed to differences in processing conditions and levels lead to the gelation.50 Proteolysis-induced age gelation of UHT of PL activity (direct or indirect UHT treatments, addition of − milk by PL which does not involve the κ-CN−β-lg -complex, PL to UHT milk).2,50 53 Different processing conditions can could occur after a critical number of the casein−casein and greatly affect the protein matrix in milk, such as the degree of casein−calcium phosphate interaction sites have been hydro- wheyproteindenaturationoncaseinmicelles,22 thereby lyzed and the remaining interactions are not sufficient to influencing the substrate accessibility and the proteolytic maintain a micellar arrangement. While at the point of gelation, pattern of PL in milk. A high PL activity could result in a α β 1 more than 95% of the S- and -CN were hydrolyzed, the fast breakdown of polypeptides to small peptides, which are peptide development showed that far from all cleavage sites unable to form a gel. Our results suggest that the proteolytic

6858 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article pathway of caseins by PL is more important for the ■ ASSOCIATED CONTENT destabilization of the casein micelles than the overall decrease *S Supporting Information in the level of intact caseins. Spectra of peptides used for relative quantification and proteins Potential Bitter Peptides. Potential bitter peptides with a in Mascot database used for identification. This material is − Q value >1400 cal mol 1 are highlighted in Table 1−4. available free of charge via the Internet at http://pubs.acs.org. Bitterness of peptides from β-CN was found in the γ-CN fraction after hydrolysis with PL,3 and analysis of tryptic digests ■ AUTHOR INFORMATION of β-CN have shown that β-CN f(203−209) is extremely Corresponding Author bitter.37 The potential bitter peptides found in the present *(V.M.R.) Phone: +45 87 33 2783. Fax: +45 87 46 6688. E- study were β-CN f(108−113) and β-CN f(170−176), which mail: [email protected]. have previously been described sensorially as tickling and Funding burning.37 The bitter peptide β-CN f(203−209) was not The Danish Agency for Science, Technology and Innovation identified in the UHT milk or was only present in a very low (DASTI) is thanked for financial support of the study. concentration. Previously described bitter peptides deriving Notes α α − from S1-CN, that can be formed by PL, are S1-CN f(23 34) The authors declare no competing financial interest. α − 54,55 α − and S1-CN f(91 100). Besides S1-CN f(91 100), all peptides from residue 91 to 100−105 found in the milk had Q ■ ACKNOWLEDGMENTS values >1400, and were potentially bitter. The smaller peptides We thank laboratory technicians Vibeke Mortensen and Betina α − α − S1-CN f(1 7) and S1-CN f(194 199) were highly hydro- Hansen at the Arla Strategic Innovation Center for assistance phobic, and could have contributed to bitterness. The peptides with the analysis and helpful discussions. α − α − α − S2-CN f(182 207), S2-CN f(189 207), and S2-CN f(198 207) have been correlated with bitterness and were also present ■ REFERENCES 16 α early in the storage period in the UHT milk. Besides S2-CN (1) Rauh, V. M.; Sundgren, A.; Bakman, M.; Ipsen, R.; Paulsson, M.; f(182−188), all peptides deriving from the C-terminal region of Larsen, L. B.; Hammershøj, M. Plasmin activity as a possible cause for α age gelation in UHT milk produced by direct steam infusion. Int. Dairy S2-CN from residue 171 and onward are potentially bitter. α J. 2014, DOI: 10.1016/j.idairyj.2013.12.007. Cleavage of Arg205-Tyr206 in S2-CN gives rise to the bitter 56 fi (2) Kelly, A. L.; Foley, J. Proteolysis and storage stability of UHT dipeptide Tyr-Leu. The identi ed potential bitter peptides milk as influenced by milk plasmin activity, plasmin/β-lactoglobulin indicate that the bitterness caused by PL might mainly be complexation, plasminogen activation and somatic cell count. Int. α α β − caused by hydrolysis of S1- and S2-CN, rather than -CN as Dairy J. 1997, 7, 411 420. previously reported.3 Several peptides formed by the combined (3) Harwalkar, V. R.; Cholette, H.; McKellar, R. C.; Emmons, D. B. action of PL and cathepsins were potentially bitter as well. Most Relation Between Proteolysis and Astringent Off-Flavor in Milk. J. α Dairy Sci. 1993, 76, 2521−2527. of these peptides derive from the N-terminal region of S1-CN α (4) Bastian, E. D.; Lo, C. G.; David, K. M. M. Plasminogen Activation and the C-terminal region of S2-CN. in Cheese Milk: Influence on Swiss Cheese Ripening. J. Dairy Sci. A bitter off-flavor was detected in the milk after 6 weeks of 1997, 80, 245−251. storage, which intensified rapidly within 1 week of storage.1 (5) Crudden, A.; Patrick; Fox, F.; Kelly, A. L. Factors affecting the The concentration and sensory thresholds of the potential hydrolytic action of plasmin in milk. Int. Dairy J. 2005, 15, 305−313. bitter peptides in the UHT milk are unknown, and the (6) Mara, O.; Roupie, C.; Duffy, A.; Kelly, A. L. The Curd-forming bitterness cannot be attributed to a specific peptide. We suggest Properties of Milk as Affected by the Action of Plasmin. Int. Dairy J. 1998, 8, 807−812. that either the overall concentration of bitter peptides reached (7) Ismail, B.; Nielsen, S. S. Invited review: Plasmin protease in milk: the sensory threshold after 7 weeks of storage or that one or current knowledge and relevance to dairy industry. J. Dairy Sci. 2010, several bitter peptides with a lower sensory threshold were 93, 4999−5009. formed very rapidly in the storage period between 5 and 7 (8) Prado, B. M.; Sombers, S. E.; Ismail, B.; Hayes, K. D. Effect of weeks. Similar to trypsin, PL has the capability to form a range heat treatment on the activity of inhibitors of plasmin and − of potential bitter peptides, but further studies are necessary to plasminogen activators in milk. Int. Dairy J. 2006, 16, 593 599. fi (9) Prado, B. M.; Ismail, B.; Ramos, O.; Hayes, K. D. Thermal con rm the bitterness of these peptides and their sensory stability of plasminogen activators and plasminogen activation in threshold. heated milk. Int. Dairy J. 2007, 17, 1028−1033. In conclusion, the special UHT process used allowed the (10) Saint-Denis, T.; Humbert, G.; Gaillard, J. L. Heat inactivation of investigation of specificity and effect of PL in a sterile milk native plasmin, plasminogen and plasminogen activators in bovine system, without causing substantial changes to the milk by the milk: a revisited study. Lait 2001, 81, 715−729. heat treatment. PL showed a different specificity in the milk (11) McMahon, D. J.; Oommen, B. S. Casein Micelle Structure, Functions, and Interactions. In Advanced Dairy Chemistry; McSweeney, system compared to previous model studies, and 6 new − α fi P. L. H., Fox, P. F., Eds.; Springer: New York, 2013; pp 185 209. cleavage sites of PL in S2-CN could be identi ed. The relative (12) White, J. H.; Zavizion, B.; O’Hare, K.; Gilmore, J.; Guo, M. R.; quantification of peptides and time of their appearance allowed Kindstedt, P.; Politis, I. Distribution of plasminogen activator in a direct comparison of formation rates of peptides to assess the different fractions of bovine milk. J. Dairy Res. 1995, 62, 115−122. affinity of plasmin to certain cleavage sites. We could show that (13) Ismail, B.; Choi, L. H.; Were, L. M.; Nielsen, S. S. Activity and nature of plasminogen activators associated with the casein micelle. J. PL is able to produce potentially bitter peptides and bitterness − was most likely caused by peptides arising from α - and α - Dairy Sci. 2006, 89, 3285 3295. S1 S2 (14) Eigel, W. N.; Butler, J. E.; Ernstrom, C. A.; Farrell, H. M.; CN. The observed proteolytic pattern indicated that PL- Harwalkar, V. R.; Jenness, R.; Whitney, R. M. Nomenclature of mediated proteolysis can destabilize the casein micelle and Proteins of Cow’s Milk: Fifth Revision. J. Dairy Sci. 1984, 67, 1599− explain proteolysis-induced age gelation of milk. 1631.

6859 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Journal of Agricultural and Food Chemistry Article γ γ γ κ α (15) Eigel, W. N. Formation of 1-A2, 2-A2 and 3-A caseins by in bridges in monomers of - and s2-casein from bovine milk. J. Dairy vitro proteolysis of β-CASEIN A2 with bovine plasmin. Int. J. Biochem Res. 1994, 61, 485−493. 1977, 8, 187−192. (37) Bumberger, E.; Belitz, H. D. Bitter taste of enzymic hydrolysates (16) Le Bars, D.; Gripon, J.-C. Specificity of plasmin towards bovine of casein. I. Isolation, structural and sensorial analysis of peptides from αs2-casein. J. Dairy Res. 1989, 56, 817−821. tryptic hydrolysates of beta-casein. Z. Lebensm. Unters. Forsch. 1993, (17) McSweeney, P. L. H.; Olson, N. F.; Fox, P. F.; Healy, A.; 197,14−19. Højrup, P. Proteolytic specificity of plasmin on bovine α -casein. Food (38) Considine, T.; Geary, S.; Kelly, A. L.; McSweeney, P. L. H. s1 α β Biotechnol 1993, 7, 143−158. Proteolytic specificity of cathepsin G on bovine s1- and -caseins. Food Chem. 2002, 76,59−67. (18) Baum, F.; Fedorova, M.; Ebner, J.; Hoffmann, R.; Pischetsrieder, ́ M. Analysis of the endogenous peptide profile of milk: identification of (39) Considine, T.; Healy, A.; Kelly, A. L.; McSweeney, P. L. H. 248 mainly casein-derived peptides. J. Proteome Res. 2013, 12, 5447− Hydrolysis of bovine caseins by cathepsin B, a cysteine proteinase − 5462. indigenous to milk. Int. Dairy J. 2004, 14, 117 124. (19) Dallas, D. C.; Guerrero, A.; Parker, E. A.; Garay, L. A.; Bhandari, (40) Moatsou, G.; Bakopanos, C.; Katharios, D.; Katsaros, G.; A.; Lebrilla, C. B.; Barile, D.; German, J. B. Peptidomic profile of milk Kandarakis, I.; Taoukis, P.; Politis, I. Effect of high-pressure treatment − at various temperatures on indigenous proteolytic enzymes and whey of holstein cows at peak lactation. J. Agric. Food Chem. 2013, 62,58 − 65. protein denaturation in bovine milk. J. Dairy Res. 2008, 75, 262 269. (20) McMahon, D. J., Age-gelation of UHT milk: Changes that occur (41) Larsen, L. B.; Hinz, K.; Jorgensen, A. L.; Moller, H. S.; Wellnitz, during storage, their eﬀect on shelf life and the mechanism by which O.; Bruckmaier, R. M.; Kelly, A. L. Proteomic and peptidomic study of proteolysis in quarter milk after infusion with lipoteichoic acid from age-gelation occurs. In Heat Treatments and Alternative Methods; Staphylococcus aureus. J. Dairy Sci. 2010, 93, 5613−5626. International Dairy Federation: Brussels, Belgium, 1996; pp 315−325. (42) Gaucher, I.; Molle,́ D.; Gagnaire, V.; Gaucheron, F. Effects of (21) Datta, N.; Deeth, H. C. Age Gelation of UHT MilkA Review. storage temperature on physico-chemical characteristics of semi- Food Bioprod. Process. 2001, 79, 197−210. skimmed UHT milk. Food Hydrocolloids 2008, 22, 130−143. (22) Crudden, A.; Oliveira, J. C.; Kelly, A. L. Kinetics of changes in (43) Farrell, H. M., Jr.; Brown, E. M.; Hoagland, P. D.; Malin, E. L. plasmin activity and proteolysis on heating milk. J. Dairy Res. 2005, 72, − Higher Order Structures of the Caseins: A Paradox? In Advanced Dairy 493 504. Chemistry1 Proteins; Fox, P. F., McSweeney, P. L. H., Eds.; Springer: (23) Anema, S. G.; Li, Y. Effect of pH on the Association of New York, 2003; pp 203−231. Denatured Whey Proteins with Casein Micelles in Heated (44) Berglund, L.; Andersen, M. D.; Petersen, T. E. Cloning and Reconstituted Skim Milk. J. Agric. Food Chem. 2003, 51, 1640−1646. ﬀ characterization of the bovine plasminogen cDNA. Int. Dairy J. 1995, (24) ISO, Microbiology of food and animal feeding stu s - horizontal 5, 593−603. method for the enumeration of microorganisms - Colony-count technique at (45) Rollema, H. S.; Visser, S.; Poll, J. K. Spectrophotometric assay of ° 30 C. ISO 4833. International Standardisation Organisation: Geneva, plasmin and plasminogen in bovine milk. Milchwissenschaft 1983, 38, Switzerland, 2003. 214−217. (25) Kuipers, B. J.; Gruppen, H. Prediction of molar extinction (46) Marchin, S.; Putaux, J. L.; Pignon, F.; Leonil, J. Effects of the coefficients of proteins and peptides using UV absorption of the environmental factors on the casein micelle structure studied by cryo constituent amino acids at 214 nm to enable quantitative reverse phase transmission electron microscopy and small-angle X-ray scattering/ high-performance liquid chromatography-mass spectrometry analysis. ultrasmall-angle X-ray scattering. J. Chem. Phys. 2007, 126, 045101. J. Agric. Food Chem. 2007, 55, 5445−5451. (47) Dalgleish, D. G. On the structural models of bovine casein (26) Ney, K. H. Voraussage der Bitterkeit von Peptiden aus deren micelles-review and possible improvements. Soft Matter 2011, 7, Aminosaurezu-sammensetzung.̈ Z. Lebensm. Unters. Forsch. 1971, 147, 2265−2272. 64−68. (48) Trejo, R.; Dokland, T.; Jurat-Fuentes, J.; Harte, F. Cryo- (27) Hachana, Y.; Kraiem, K.; Paape, M. J. Effect of plasmin, milk transmission electron tomography of native casein micelles from somatic cells and psychrotrophic bacteria on casein fractions of ultra bovine milk. J. Dairy Sci. 2011, 94, 5770−5775. high temperature treated milk. Food Sci. Technol. Res. 2010, 16,79−86. (49) Crudden, A.; Afoufa-Bastien, D.; Fox, P. F.; Brisson, G.; Kelly, (28) Larsen, L. B.; Benfeldt, C.; Rasmussen, L. K.; Petersen, T. E. A. L. Effect of hydrolysis of casein by plasmin on the heat stability of Bovine milk procathepsin D and cathepsin D: coagulation and milk milk. Int. Dairy J. 2005, 15, 1017−1025. protein degradation. J. Dairy Res. 1996, 63, 119−130. (50) Manji, B.; Kakuda, Y. The Role of Protein Denaturation, Extent α (29) Le Bars, D.; Gripon, J. C. Hydrolysis of s1-casein by bovine of Proteolysis, and Storage Temperature on the Mechanism of Age plasmin. Lait 1993, 73, 337−344. Gelation in a Model System. J. Dairy Sci. 1988, 71, 1455−1463. (30) Malin, E. L.; Brown, E. M.; Wickham, E. D.; Farrell, H. M. (51) Kohlmann, K. L.; Nielsen, S. S.; Ladisch, M. R. Effects of a Low Contributions of terminal peptides to the associative behavior of α - Concentration of Added Plasmin on Ultra-High Temperature s1 − casein. J. Dairy Sci. 2005, 88, 2318−2328. Processed Milk. J. Dairy Sci. 1991, 74, 1151 1156. (31) Gagnaire, V.; Leonil,́ J. Preferential sites of tryptic cleavage on (52) Aroonkamonsri, J. The Role of Plasmin and Bacterial the major bovine caseins within the micelle. Lait 1998, 78, 471−489. Proteinases on Age Gelation of UHT Milk (Pseudomonas fluorescens, (32) Diaz, O.; Gouldsworthy, A. M.; Leaver, J. Identification of Pseudomonas fragi, Pseudomonas luteola, Vibrio fluvialis). Ph.D. Thesis, peptides released from casein micelles by limited trypsinolysis. J. Agric. University of Guelph, Canada, 1996. Food Chem. 1996, 44, 2517−2522. (53) Newstead, D. F.; Paterson, G.; Anema, S. G.; Coker, C. J.; (33) Holt, C.; de Kruif, C. G.; Tuinier, R.; Timmins, P. A. Wewala, A. R. Plasmin activity in direct-steam-injection UHT- processed reconstituted milk: Effects of preheat treatment. Int. Dairy Substructure of bovine casein micelles by small-angle X-ray and − neutron scattering. Colloids Surf., A 2003, 213, 275−284. J. 2006, 16, 573 579. (34) Farrell, H. M., Jr.; Malin, E. L.; Brown, E. M.; Mora-Gutierrez, (54) Matoba, T.; Hayashi, R.; Hata, T. Isolation of bitter peptides A. Review of the chemistry of α -casein and the generation of a from tryptic hydrolysate of casein and their chemical structure. Agric. S2 Biol. Chem. 1970, 34, 1235−1243. homologous molecular model to explain its properties. J. Dairy Sci. (55) Hill, R.; Van Leeuwen, H. Bitter peptides from hydrolysed 2009, 92, 1338−1353. casein coprecipitate. Aust. J. Dairy Technol. 1974, 29,32−34. (35) Tauzin, J.; Miclo, L.; Roth, S.; Molle,́ D.; Gaillard, J. L. Tryptic α (56) Kim, H. O.; Li-Chan, E. C. Quantitative structure-activity hydrolysis of bovine S2-casein: identification and release kinetics of − relationship study of bitter peptides. J. Agric. Food Chem. 2006, 54, peptides. Int. Dairy J. 2003, 13,15 27. 10102−10111. (36) Rasmussen, L. K.; Højrup, P.; Petersen, T. E. Disulphide arrangement in bovine caseins: localization of intrachain disulphide

6860 dx.doi.org/10.1021/jf502088u | J. Agric. Food Chem. 2014, 62, 6852−6860 Supplementary data

1. Proteins in custom Mascot database used for identification of peptides from tandem MS

sp|P02662|CASA1A_BOVIN Alpha-S1-casein sp|P02662|CASA1B_BOVIN Alpha-S1-casein sp|P02662|CASA1C_BOVIN Alpha-S1-casein sp|P02662|CASA1D_BOVIN Alpha-S1-casein sp|P02662|CASA1G_BOVIN Alpha-S1-casein sp|P02662|CASA1I_BOVIN Alpha-S1-casein sp|P02663|CASA2A_BOVIN Alpha-S2-casein sp|P02663|CASA2B_BOVIN Alpha-S2-casein sp|P02663|CASA2D_BOVIN Alpha-S2-casein sp|P02666|CASBA1_BOVIN Beta-casein sp|P02666|CASBA2_BOVIN Beta-casein sp|P02666|CASBA3_BOVIN Beta-casein sp|P02666|CASBB_BOVIN Beta-casein sp|P02666|CASBC_BOVIN Beta-casein sp|P02666|CASBF_BOVIN Beta-casein sp|P02666|CASBI_BOVIN Beta-casein sp|P02668|CASKA_BOVIN Kappa-casein sp|P02668|CASKB_BOVIN Kappa-casein sp|P02668|CASKC_BOVIN Kappa-casein sp|P02668|CASKE_BOVIN Kappa-casein sp|P00711|LALBAA_BOVIN Alpha-lactalbumin sp|P00711|LALBAB_BOVIN Alpha-lactalbumin sp|P02754|LACBA_BOVIN Beta-lactoglobulin sp|P02754|LACBB_BOVIN Beta-lactoglobulin sp|P02754|LACBC_BOVIN Beta-lactoglobulin sp|P02754|LACBD_BOVIN Beta-lactoglobulin sp|P02769|ALBU_BOVIN Serum albumin sp|P02769|ALBU214_BOVIN Serum albumin sp|P24627|TRFL_BOVIN Lactotransferrin sp|P01239|PRL_BOVIN Prolactin sp|P80025|PERL_BOVIN Lactoperoxidase sp|P80025|PERLS_BOVIN Lactoperoxidase Ser394 variant sp|Q29443|TRFE_BOVIN Serotransferrin sp|P79345|NPC2_BOVIN Epididymal secretory protein E1 sp|Q9XSC9|TCO2_BOVIN Transcobalamin-2 sp|P18902|RET4_BOVIN Retinol-binding protein 4 sp|P15497|APOA1_BOVIN Apolipoprotein A-I sp|P18892|BT1A1_BOVIN Butyrophilin subfamily 1 member A1 sp|P80457|XDH_BOVIN Xanthine dehydrogenase/oxidase sp|P06868|PLMN_BOVIN Plasminogen sp|Q95114|MFGM_BOVIN Lactadherin sp|P31096|OSTP_BOVIN Osteopontin sp|Q8WML4|MUC1_BOVIN Mucin-1 sp|Q8MI01|MUC15_BOVIN Mucin-15 sp|Q8MI01-2|MUC15S_BOVIN Isoform 2 of Mucin-15 sp|Q9TUM6|PLIN2_BOVIN Perilipin-2 isoform-1 sp|Q9TUM6-2|PLIN2_BOVIN_2 Isoform 2 of Perilipin-2

2. Supplementary spectra of peptides used for relative quantification

αS1-CN f(1-22) UV and EIC chromatogram MS spectrum at peak top Comments/Contamination Overlay from αS1 f(106-124)

1000 2e+5 800 873.1665 600 400 2e+5 200 0

Absorbance 214 nm 214 Absorbance

23,0 23,5 24,0 24,5 25,0 25,5 26,0 1e+5

3e+5 1309.2459

Intensity (AU) Intensity 2e+5 5e+4 655.1266 1e+5

Intensity (AU) Intensity 0

-1e+5 23,0 23,5 24,0 24,5 25,0 25,5 26,0 400 600 800 1000 1200 1400 1600 1800 2000 Retention time (min) m/z αS1-CN f(106-124) UV and EIC chromatogram MS spectrum at peak top Comments/Contamination

Overlay from αS1 f(1-22)

1000 1,2e+5

800 753.3740 600 1,0e+5 400 200 8,0e+4 1129.5582 0

Absorbance 214 nm 214 Absorbance

23,0 23,5 24,0 24,5 25,0 25,5 26,0 6,0e+4 3e+5 Intensity (AU) Intensity 4,0e+4 2e+5

1e+5 2,0e+4 565.2839

Intensity (AU) Intensity 0

-1e+5 0,0 23,0 23,5 24,0 24,5 25,0 25,5 26,0 400 600 800 1000 1200 1400 1600 1800 2000 Retention time (min) m/z

αS2-CN f(1-21) UV and EIC chromatogram MS spectrum at peak top Comments/Contamination Overlay with 550.3334+1

800 1e+5

600

400 8e+4 200

0 6e+4

Absorbance 214 nm 214 Absorbance

17 18 19 20 21 2e+5 4e+4 2e+5 (AU) Intensity 916.6757 1374.5114 1e+5 5e+4 2e+4 687.7559

Intensity (AU) Intensity 0

-5e+4 0 17 18 19 20 21 400 600 800 1000 1200 1400 1600 1800 2000 Retention time (min) m/z

αS2-CN f(1-24) UV and EIC chromatogram MS spectrum at peak top Comments/Contamination Overlay with β-CN f(106-113) (207.2661+2) 800 40000

600 1045.0690 400 30000 200

Absorbance 214 nm 214 Absorbance

17 18 19 20 21 20000 2e+5 1567.0994

2e+5 (AU) Intensity 784.0544 1e+5 10000 5e+4

Intensity (AU) Intensity 0

-5e+4 0 17 18 19 20 21 400 600 800 1000 1200 1400 1600 1800 2000 Retention time (min) m/z

αS2-CN f(115-150) UV and EIC chromatogram MS spectrum at peak top Comments/Contamination Overlay with 613.9731+6 (unknown component) 1000 60000

800 1431.9881 600 50000 400 200 40000 0

Absorbance 214 nm 214 Absorbance

21,0 21,5 22,0 22,5 23,0 23,5 24,0 30000 4e+5 1074.2425

3e+5 (AU) Intensity 20000 2e+5 859.7979

1e+5 10000

Intensity (AU) Intensity 0

-1e+5 0 21,0 21,5 22,0 22,5 23,0 23,5 24,0 400 600 800 1000 1200 1400 1600 1800 2000

β-CN f(1-28) UV and EIC chromatogram MS spectrum at peak top Comments/Contamination Overlay with 794.3810+3 (unknown component) 1,6e+5

1500 870.3845 1,4e+5 1160.1748 1000 1,2e+5 500 696.7094 1,0e+5

Absorbance 214 nm 214 Absorbance 0 23,0 23,5 24,0 24,5 25,0 25,5 26,0 8,0e+4 8e+5

6e+5 (AU) Intensity 6,0e+4 4e+5 4,0e+4 2e+5 1740.2566 2,0e+4

Intensity (AU) Intensity 0

-2e+5 0,0 23,0 23,5 24,0 24,5 25,0 25,5 26,0 400 600 800 1000 1200 1400 1600 1800 2000

β-CN f(29-48) UV and EIC chromatogram MS spectrum at peak top Comments/Contamination Overlay with 802.4830+1 (unknown component) 600 2,5e+5

400 Col 1 vs Col 2 780.5020 2,0e+5 200

0 1,5e+5

Absorbance 214 nm 214 Absorbance

17,0 17,5 18,0 18,5 19,0 19,5 20,0 2e+5 1,0e+5

2e+5 (AU) Intensity 1e+5 5,0e+4 5e+4 1281.0849

Intensity (AU) Intensity 0

-5e+4 0,0 17,0 17,5 18,0 18,5 19,0 19,5 20,0 400 600 800 1000 1200 1400 1600 1800 2000

Paper IV

Rauh, V. M., Johansen, L. B., Ipsen, R., Paulsson, M., Larsen, L. B., & Hammershøj, M. (2014)

Protein lactosylation in UHT milk during storage measured by liquid chromatography-mass spectrometry and furosine

Manuscript intended for International Journal of Dairy Technology

2 Protein Lactosylation in UHT Milk during Storage measured by Liquid Chromatography-

Mass Spectrometry and Furosine

Valentin M. Rauh*a,b, Lene B. Johansena, Mette Bakmana, Richard Ipsenc, Marie Paulssond, Lotte

B. Larsenb and Marianne Hammershøjb

a Arla Foods Strategic Innovation Centre, Rørdrumvej 2, DK-8220 Brabrand, Denmark b Aarhus University, Dept. of Food Science, Faculty of Science and Technology, Blichers Allé 20,

DK-8830 Tjele, Denmark c University of Copenhagen, Department of Food Science, Rolighedsvej 30, DK-1958 Frederiksberg

C, Denmark c Department of Food Technology, Engineering and Nutrition, Lund University, SE-221 00 Lund,

Sweden

*Corresponding author. Tel: +4587332783; FAX: +4587466688

E-mail address: [email protected]

3 1 ABSTRACT

2 The initial stage of the Maillard reaction, protein lactosylation, occurs during the heat treat-

3 ment of milk and continues during subsequent storage. We compared the initial lactosylation

4 as well as the rate of lactosylation of milk proteins during storage in UHT milk subjected to

5 directly or indirectly heat treatment using liquid chromatography (LC) coupled with electro

6 spray injection mass spectrometry (ESI-MS). Furosine content was used as an overall marker

7 to allow for a quantitative correlation of lactosylation measured by LC-ESI-MS in the UHT

8 milks. Protein lactosylation increased during the storage period of 6 months at 20°C. Both the

9 initial extent and the rate of lactosylation positively correlated with the number of lysine resi-

10 dues in the different proteins. An exponential or linear correlation with furosine concentration

11 could be established for major and minor lactosylated proteins, respectively.

4 12 1. Introduction

13 During heat treatment and following storage of UHT milk, the Maillard reaction is one of the

14 major quality deterioration factors. The formation of the Amadori product ε-lactulosyllysine

15 between lactose and the free ε-amino group of lysine results in a decreased nutritional value

16 of the milk as the lysine will no longer be bioavailable and extensive lactosylation can also

17 reduce the digestibility of proteins (Dalsgaard et al., 2007). In the past, the extent of lactosyl-

18 ation has mainly been studied using indirect methods, such as the measurement of furosine,

19 which is a product of the acid hydrolysis of lactulosyllysine (Erbersdobler and Somoza,

20 2007). More recently, mass spectrometry (MS) has been shown to be a powerful tool for the

21 analysis of milk protein lactosylation in both wet and dry systems (Siciliano et al., 2013). Pro-

22 tein lactosylation in wet milk has been studied both in the casein fraction (Scaloni et al., 2002;

23 Johnson et al., 2011), in the native whey fraction (Czerwenka et al., 2006; Meltretter et al.,

24 2009; Losito et al., 2007) and in indigenous phospho-peptides (Pinto et al., 2012). Heat

25 treatment of milk has a major impact on the whey protein fraction and severe heat treatments,

26 such as e. g. UHT (136-140 °C for 2-6 s), result in an almost complete denaturation of the

27 whey proteins (Datta et al., 2002). In previous studies, the degree of lactosylation of native

28 whey proteins could be correlated with whey protein denaturation and fluorescence

29 (Meltretter et al., 2009; Losito et al., 2010). Despite the fact that caseins are the quantitatively

30 major milk proteins, lactosylation has only been observed for the most abundant caseins: αS1-

31 casein (αS1-CN) and β-casein (β-CN). A quantitative correlation in wet milk systems with fu-

32 rosine concentration could only be achieved for a singly lactosylated β-CN in processed

33 cheese (Steffan et al., 2006). In dry milk systems, a quantitative relationship between furosine

34 concentration and lactosylated peptides originating from whey proteins was established for

35 milk powder (Le et al., 2013).

36 Furthermore, the Amadori product is continuously formed during storage of milk. The devel-

37 opment of lactulosyllysine during the storage is therefore an important parameter for shelf life

5 38 estimation of UHT milks. To our knowledge the development of protein lactosylation during

39 storage has not been studied using mass spectrometry before.

40 The aim of this study was to follow the lactosylation of the casein and whey fractions in di-

41 rectly and indirectly heat treated UHT milks by ESI-MS and furosine analysis during subse-

42 quent storage. In particular, the rate of lactosylation of different caseins and whey proteins

43 during storage was compared.

45 2. Materials and Methods

46 2.1 Materials

47 One day old low pasteurized milk (72 °C for 15 s) with 1.5 % fat was obtained from Arla

48 Foods Stockholm Dairy (Stockholm, Sweden) and processed in the UHT pilot plant (SPX,

49 Silkeborg, Denmark) at the Arla Strategic Innovation Center (Stockholm, Sweden). The milk

50 (200 L) was pre-heated at 95 °C for 5 s (Direct A) or 180 s (Direct B) in a tubular heat ex-

51 changer before being subjected to the direct steam infusion heat treatment ( >150 °C for <0.2

52 s). After flash cooling, the milk was homogenized using a two-stage homogenizer (160/40

53 bar), further cooled down to 20 °C by a tubular heat exchanger and packed aseptically in 300

54 mL glass bottles. For the indirect UHT process (Indirect), milk was heated to 80 °C, homoge-

55 nized using a two-stage homogenizer (160/40 bar) and further heated at 140 °C for 4 s. The

56 milk was cooled down to 20 °C by a tubular heat exchanger and packed aseptically in 300 mL

57 glass bottles. These trials were repeated three times during three consecutive weeks and the

58 milk was stored in the dark at 20 °C in a climate chamber for 6 months. One 300 mL bottle

59 per treatment was collected once every month for analysis.

61 2.1 Analytical methods

62 Protein content was measured by a MilkoScan FT 120 (Foss, Hillerød, Denmark). Analysis of

63 protein composition in the milk was performed as described by Bonfatti, Grigoletto, Cecchi-

6 64 nato, Gallo, and Carnier (2008) with the following modifications. A sample of 200 µL of milk

65 (w/w) was frozen at -20 °C until further use. Disulphide bonds were reduced by the addition

66 of 20 µL 0.5 M dithiothreitol and 1 mL 100 mM tri-sodium citrate, 6 M urea at 30 °C for 60

67 min. The samples were centrifuged at 9300 x g at 5 °C for 10 min. From the clear phase, 200

68 µL were used for the analysis. A volume of 5 µL was injected into the LC-MS system. The

TM 69 column was a BioSuite C18 PA-B (C18, 3.5 µm 2.1 mm x 250 mm, Waters, Milford, MA,

70 USA). Buffer A was Milli-Q water with 0.05 % trifluoroacetic acid (TFA) (v/v) and buffer B

71 acetonitrile with 0.1 % (v/v) TFA. Column temperature was 47 °C and a linear gradient from

72 34.8 % to 46.5 % buffer B from 2 to 16.5 min was applied at a flow rate of 0.35 mL min-1

73 with UV detection at 214 nm. MS detection was performed with a electrospray-time-of-flight

74 mass spectrometer with (ESI-TOF-MS; Agilent 6230 Aqurate mass detector, Agilent Tech-

75 nologies) and MS scans were continuously recorded between a mass to charge ratio (m/z) of

76 300 and 3200 in 2 GHZ extended dynamic positive mode. The TOF-MS had a drying gas

77 temperature at 325 °C at a flow rate of10 L min-1 and a nebulizer pressure of 35 psi. Sheath

78 gas temperature was 300 °C at a flow rate of 7 L min-1. Fragmentor energy was set 200 V and

79 the applied energy to the capillary was 3500 V. Data analysis was performed with Mass

80 Hunter software (Agilent Technologies, Santa Clara, CA, USA).

81 Mass spectra and UV chromatograms were analyzed by MassHunter (version B 6.01) and

82 Profinder (pre-release version B.06.00 Service Pack 1) software (Agilent Technologies, Santa

83 Clara, CA, USA).

84 Analysis of furosine was performed as follows: One mL milk sample was mixed with 3 mL of

85 10 M HCl in screw-cap tubes. Nitrogen was bubbled through the samples for 2 min to mix the

86 samples. The tubes were closed and heated at 115 ± 1 °C for 18 h. After heating, the samples

87 were cooled and filtered through a paper filter (pore size 20 µm). The filtrate was further fil-

88 tered through a 0.45 µM disposable filter, diluted 5 times in 3 M HCl and transferred into

89 HPLC vials. For furosine analysis, 10 µL of the filtrate were injected into a Dionex 300DX

7 90 system (Dionex, Germering, Germany) with a DAD detector (UV at 280 nm) using a Supelco

91 Supelcosil LC-8 column for isocratic separation with 0.06 M sodium acetate as buffer at a

92 flow rate of 1 mL min-1.

93 All measurements were performed in duplicate. Standard deviation between replicates for all

94 measurements was ranging from 5-10 %.

96 2.1 Data analysis

97 In order to achieve the optimal results, the ESI mass spectra were analyzed by 3 different

98 methods of data analysis for αS1-CN 8P. The αS1-CN 8P was chosen, as it is the most abun-

99 dant casein and MS analysis of the milk showed that αS1-CN 8P was present in only one ge-

100 netic in contrast to β-CN. Extracted ion chromatograms (EIC) of unmodified, mono-

101 lactosylated and di-lactosylated αS1-CN were extracted using the most abundant ion as quanti-

102 fier and the second abundant ion as qualifier (the used quantifier ions are shown in Figure 2).

103 The mass error was set to 100 ppm. Mass spectra were deconvoluted in MassHunter using the

104 maximum entropy setting (mass step 1 Da, signal to noise > 30 counts, minimum consecutive

105 charge states 5). Finally, the mass spectra were analyzed using the Profinder software, a batch

106 recursive deconvolution tool. A large molecular extractor algorithm (LMFE) was used. Reten-

107 tion time tolerance was set to 1 min and an m/z window of 50 ppm. All EIC from proteins

108 having 5 to 35 charge states were smoothed using a Gaussian function before integration.

109 Additionally, the UV signal at 214 nm was integrated. Due to a co-elution of αS1-CN 8P and

110 9P, both variants were integrated as one peak.

111 Results are shown as mean value ± standard deviation of the 3 replicates. Linear regression

112 analysis and ANOVA was performed with Matlab software (Mathworks, Natick, MA). Mod-

113 el? Class variables were heat treatment (1, 2, 3), storage time points (1- 6 months)? Differ-

114 ences were regarded significant at 95%-level i.e. (P<0.05).

115

8 116 3. Results and Discussion

117 3.1 Initial furosine concentration and whey protein denaturation

118 In order to evaluate the effect of heat treatment, the initial furosine concentration and degree

119 of whey protein denaturation was analyzed in the different UHT milks after one day of stor-

120 age (Table 1). As expected, the degree of whey protein denaturation increased with increasing

121 severity of the heat treatment. As previously reported, the degree of denaturation of β-lg was

122 higher compared to α-la (Dannenberg and Kessler, 1988). While there is a clear difference of

123 the heat treatment in respect to α-la denaturation for all milks, the degree of β-lg denaturation

124 in the Direct B and Indirect was not significantly different. This can be attributed to the al-

125 most complete denaturation of β-lg in the milk (Table 1). As expected, the initial furosine

126 concentrations in Direct A and Direct B were approximately 10-15 times lower compared to

127 Indirect and within the range of previously reported results (Datta et al., 2002). No furosine

128 could be detected in the low pasteurized feed. The furosine content in Direct A was slightly

129 higher than in Direct B and the pre heat treatment of 95 °C for 180 s of Direct B did not affect

130 the furosine concentration. This trend continued during the entire storage period (Figure 1).

131 The furosine content increased linearly in all UHT milks during the 6 months storage period

132 and no significant difference in the rate of formation between the milks was observed. This is

133 in line with previous findings (Corzo et al., 1994; Nangpal and Reuter, 1990) and indicates

134 that the severity of the heat treatment affects the initial extent of the lactosylation but not the

135 rate of formation during storage under similar conditions. The linear increase of furosine dur-

136 ing storage in all milks further shows that the Maillard reaction was in its initial stage during

137 the entire storage period and that degradation of lactulosyllysine to intermediate Maillard re-

138 action products, such as 5-hydroxy-2-methylfurfural (HMF), most likely only occurred to a

139 limited extent (Le et al., 2011).

140

141

9 142 3.2 Comparison of MS analysis

143 Analysis of protein lactosylation using ESI-MS was evaluated for αS1-CN 8 P using EIC, de-

144 convolution and Agilent Profinder. A representative mass spectrum of αS1-CN 8 P in Indirect

145 after 6 months of storage is shown in Figure 2. The most abundant ion for αS1-CN 8 P was at a

146 charge state +22 and the 2 ions with a m/z shift of 14.6 and 29.2 represented the mono and di-

147 lactosylated αS1-CN 8 P species. No ion for a tri-lactosylated αS1-CN 8 P could be found in the

148 mass spectrum. An EIC of αS1-CN 8 P using the ions shown in the inset of Figure 2 are shown

149 in Figure 3 A. As previously reported, the addition of lactose renders the protein more hydro-

150 philic, which results in a shift to lower retention times and a broadening of the peaks

151 (Czerwenka et al., 2006; Losito et al., 2007). The EIC of lactosylated and especially di-

152 lactosylated αS1-CN 8 P are broader compared to the unlactosyalted αS1-CN 8 P, which can be

153 caused by the lactosylation of different lysine residues, which can result in a change in the

154 retention time (Czerwenka et al., 2006). Figure 3B shows the corresponding UV signal of αS1-

155 CN 8 P and 9 P of Indirect, seen as a shoulder on the right side of the αS1-CN 8 P peak. The

156 comparison of the UV signal after 0 and 6 months of storage of Indirect shows a decrease in

157 peak height and a broadening of the peak due to the changes in retention time observed in the

158 EIC. The deconvoluted spectra of αS1-CN 8 P after 0 and 6 months of storage are shown in

159 Figure 4. After 6 months of storage, a tri-lactosylated αS1-CN 8 P could be observed in the

160 UHT milk and the intensity distribution shifted towards higher degrees of lactosylation (Fig-

161 ure 4B). The decrease in un-lactosylated αS1-CN 8 P intensity was not proportional to the in-

162 crease in the intensity of the lactosylated αS1-CN 8 P species. The UV area of αS1-CN 8 P on

163 the other hand was constant during the entire storage period in all UHT milks (± 6 %). This

164 relationship was valid for all l reduced proteins and the summed MS area decreased by ap-

165 proximately 25-40 % after 6 months of storage. The decrease in intensity indicates a reduced

166 ionizability for lactosylated proteins, which previously are reported for caseins by Scaloni et

167 al. (2002) using a single quadrupole MS (Scaloni et al., 2002). For pH 4.6 soluble un-

10 168 lacostylated and lactosylated β-lg and α-la on the other hand, an identical ionizability is sug-

169 gested for matrix assisted laser desorption (MALDI) TOF-MS and triple quadrupole MS

170 (Czerwenka et al., 2006; Meltretter et al., 2009). A shift in the charge state distribution to-

171 wards lower charge states occurred upon lactosylation of pH 4.6 soluble β-lg analysed with

172 triple quadrupole MS, which might be caused by a hindrance of protonation by the added lac-

173 tose to lysine residues (Czerwenka et al., 2006). The reported differences in the ionization of

174 un-lactosylated and lactosylated milk proteins could be ascribed to the different MS instru-

175 mentation and analysis conditions used. It is likely that the ionizability of lactosylated pro-

176 teins could depend on the charge state distribution of the protein. A reduced ionization of lac-

177 tosylated proteins might be negligible for lower charge state distribution or singly charged

178 proteins in MALDI-TOF-MS as there are other ionizable residues present in excess. However,

179 the lactosylation might have a substantial impact on the ionization of reduced and unfolded

180 proteins with a higher charge state distribution as observed in this study. A direct comparison

181 of the lactosylation of denatured and pH 4.6 soluble whey proteins was not possible in this

182 study due to the high degree of whey protein denaturation in the UHT milks (Table 1).

183 Because of the decrease and variation in MS area (measured as counts per second), the area

184 distribution of the summed areas of un-lactosylated and lactosylated proteins was used for

185 calculating the extent of lactosylation. Analysis using deconvolution and the Profinder soft-

186 ware gave similar results, while analysis using EIC showed higher degrees of lactosylation

187 due to a missing exclusion of background noise (see supplementary section). In comparison to

188 the deconvolution algorithm, Profinder resulted in an improved extraction of the ion enve-

189 lopes due to the recursive workflow and was thus used for further analysis.

190

191 3.2 Kinetics of protein lactosylation during storage

192 Di-lactosylated (for α-la and κ-CN) and tri-lactosylated proteins (for β-lg, αS1-CN, αS2-CN

193 and β-CN) were not present in the one day old UHT milks and accounted for < 2 % of the ar-

11 194 ea in the most severely lactosylated UHT milk samples, i.e. Indirect after 6 months of storage,

195 and were not considered in the data analysis. A comparison of the absolute and relative area

196 distribution of αS1-CN 8 P lactosylation during storage in the UHT milks is shown in Figure 5

197 (for other proteins see supplementary section). As discussed above, the overall area decreased

198 during storage. In Indirect, both the area of un-lactosylated and mono-lactosylated αS1-CN 8 P

199 decreased during storage. The relative change in intensity of un-lactosylated and lactosylated

200 αS1-CN 8 P, on the other hand, fitted well to a linear regression in all milks (Figure 5 bottom

201 row). On a closer look, both the relative area of un-lactosylated and mono-lactosylated αS1-

202 CN 8 P seemed to level out due to the difference in ionization efficiency of the lactosylated

203 αS1-CN 8 P and could also be fitted to an exponential function. In order to compare the lacto-

204 sylation rates of different proteins and to reduce the number of variables, a linear regression

205 was applied.

206 Table 2 shows the results of the linear regression of casein lactosylation during storage. For

2 207 αS1-CN 8 P, β-CN A and κ-CN, the degree of lactosylation was significantly higher in Direct

208 B compared to Direct A, showing that the measurement of lactosylation by LC-MS is a very

209 sensitive method for discrimination between heat treatments. As the initial furosine results

210 already indicated, the initial degree of protein lactosylation in Direct A and Direct B was

211 similar for most of the proteins. The degree of lactosylation measured by LC-MS after heat

212 treatment in Indirect was only 2-4 times higher than in Direct A and Direct B. The lower de-

213 gree of lactosylation found, compared to the furosine measurement (10-15 time higher; Figure

214 1), could be caused by the different response factors of the lactosylated proteins and the fact

215 that the two lactose moieties in di-lactosylated proteins were not taken into account. The ini-

216 tial extent of lactosylation for the different caseins was αS2-CN > αS1-CN > β-CN > κ-CN,

217 which matches the number of lysine residues (24, 14, 11 and 9) and thus the available reaction

218 sites of the caseins. The results do, to some extent, support previous findings by Scaloni et al.

219 suggesting αS1-CN and β-CN being the most lactosylated caseins (Scaloni et al., 2002).

12 220 Similar to the initial casein lactosylation, the rate of lactosylation was alike for Direct A and

221 Direct B with the exception of β-CN A2. The decrease in un-lactosylated and increase in

222 mono-lactosylated protein was more pronounced in the Direct A and Direct B than indirect. A

223 higher degree of initial lactosylation resulted in a lower rate of lactosylation. This can be as-

224 cribed to the slightly exponential nature of the curves (see Figure 5). The increase in di-

225 lactosylated proteins on the other hand was significantly higher in Indirect reflecting the more

226 advanced degree of lactosylation in this UHT milk. The αS1-CN 8 P showed the highest rate

227 of lactosylation in all UHT milks followed by β-CN, αS2-CN and κ-CN. The lower rate of lac-

228 tosylation of αS2-CN was ascribed to a high initial degree of lactosylation and reduced re-

229 sponse factors of the lactosylated αS2-CN. Additionally, the amount of phosphorylations of

230 αS2-CN varies, which results in a massive overlay of the differently phosphorylated and lacto-

231 sylated αS2-CN variants. This limits the extraction of ion envelopes and may cause ion sup-

2 232 pression. This was also reflected in the lower R –value of the linear regression of αS2-CN. The

233 rate of lactosylation for κ-CN was not significantly different between the UHT milks. This

234 could be caused by the low initial degree of lactosylation and the absence of multi-

235 lactosylated variants with accordingly lower response factors. To our knowledge, the lacto-

236 sylation of αS2-CN and κ-CN has not been reported previously.

237 The lactosylation of whey proteins during storage is shown in Table 3. The initial lactosyla-

238 tion and rate of lactosylation of the reduced whey proteins was similar to the lactosylation of

239 the reduced caseins. The initial lactosylation as well as the rate of lactosylation during storage

240 of β-lg were higher compared to those of α-la, which correlate with the amount of available

241 reaction sites on the proteins with 15 lysine residues in β-lg and 12 in α-la, and is in line with

242 previous findings (Czerwenka et al., 2006; Losito et al., 2007). The initial degree of lactosyla-

243 tion of α-la was significantly different for all UHT milks, but as with κ-CN, there was no sig-

244 nificant difference in the rate of lactosylation of α-la between the UHT milks. Previous stud-

245 ies indicate a difference in the extent of lactosylation of native and unfolded β-lg. The lysine

13 246 residues in unfolded β-lg could be easier accessible to react with lactose, and di-lactosylated

247 β-lg is the predominant form in the unfolded β-lg population (Losito et al., 2010). The rate of

248 lactosylation for β-lg during storage showed a similar trend, with a positive correlation of

249 rates of formation of dilatosylated β-lg and severity of the heat treatment and whey protein

250 denaturation (Table 3). However, the same trend was found for the caseins as well with the

251 exception of κ-casein. The caseins do not possess any notable tertiary structure, which would

252 explain this phenomenon. Thus, it cannot be concluded, whether the higher degree of whey

253 protein denaturation or simply a higher initial degree of lactosylation caused the higher rate of

254 formation of di-lactosylated β-lg in this study.

255

256 3.3 Correlation of lactosylation measured by furosine and LC-MS

257 The correlation of furosine concentration and decrease in un-lactosylated reduced protein is

258 shown in Figure 6 for β-CN, αS1-CN 8 P, α-la and β-lg. For less lactosylated proteins, such as

259 β-CN, κ-CN and α-la linear correlations between decrease in un-lactosylated proteins and in-

260 crease in furosine concentrations were found with correlation coefficients of R2=0.95, 0.92,

261 and 0.87, respectively. This is in line with Steffan et al. (2006), who found a linear correlation

262 between unmodified β-CN and furosine in processed cheese, although only a singly lactosyl-

263 ated β-CN was detected (Steffan et al., 2006). For the higher lactosylated proteins αS1-CN 8 P

264 and β-lg, the best correlations were achieved with an exponential fit resulting in R2=0.93 and

265 R2=0.95, respectively. The difference was caused by the change in response factor and the

266 higher concentration of di-lactosylated variants. Figure 6 shows that the area distribution of

267 lactosylated milk proteins measured by LC-MS can be used to quantify the extent of protein

268 lactosylation in milk during storage. Meltretter et al. (2009) previously showed that the lacto-

269 sylation of pH 4.6 soluble whey proteins correlated with several heat markers, but the limiting

270 factor for this method is the amount of pH 4.6 soluble whey proteins, which does not allow

271 assessment of severely heat treated products (Meltretter et al., 2009). Furthermore, a potential

14 272 higher degree of lactosylation for denatured whey proteins compared to native whey proteins

273 might lead to an underestimation of the degree of lactosylation when measuring on pH 4.6

274 soluble whey proteins. The analysis of reduced caseins and whey proteins in this study, on the

275 other hand, did not have this limitation. For highly heat treated products with higher degrees

276 of lactosylation compared to the UHT milk used in this study, the less lactosylated proteins

277 (β-CN, κ-CN and α-la) could serve as marker proteins, since the effect of reduced intensity of

278 lactosylated species would be less pronounced. An improved chromatographic separation

279 could further facilitate the extraction of the MS spectra due to a reduced overlap of the differ-

280 ent lactosylated species. The use of a relative area distribution of un-lactosylated and lactosyl-

281 ated protein species additionally makes this method robust against other protein modifica-

282 tions, such as oxidation, and limited proteolysis. It should be noted that the described method

283 is only suited for analysis of the initial stage of the Maillard reaction. For assessment of prod-

284 ucts containing intermediate and advanced Maillard reaction products, the protein lactosyla-

285 tion is not proportional to the extent of the Maillard reaction due to a degradation of lactulo-

286 syllysine and different methods, such as measurement of advanced glycosylation end products

287 or fluorescence accumulation, should be applied (Le et al., 2013; Van Boekel, 1998; Bosch et

288 al., 2007; Le et al., 2011).

289

290 4. Conclusion

291 In conclusion, we could show that LC-MS is a suitable analysis to investigate the initial stage

292 of the Maillard reaction in UHT milk during storage. The results obtained by LC-MS allowed

293 a quantitative correlation with the commonly used marker, furosine. In comparison to the fu-

294 rosine method, the LC-MS method allowed the analysis of the protein lactosylation of all ma-

295 jor milk proteins, giving a more detailed view of the Maillard reaction in UHT milk during

296 storage. The extent of protein lactosylation in the directly heated UHT milk was approximate-

297 ly 10-15 times lower compared to indirectly heated UHT milk. The rate of protein lactosyla-

15 298 tion during storage was constant for indirect and directly heated UHT milk and correlated

299 with the number of lysine residues in the caseins and whey proteins.

300

301 5. Acknowledgements

302 The Danish Agency for Science, Technology and Innovation (DASTI) is thanked for financial

303 support of the study. We thank laboratory technicians Vibeke Mortensen and Betina Hansen

304 at the Arla Strategic Innovation Center for the assistance with the analysis and helpful discus-

305 sions. Steve Madden and Steve Fischer at Agilent Technologies are thanked for expanding the

306 Profinder software, which enabled the analysis of intact proteins in this study.

307

308 6. References

309 Bosch, L, Alegria, A, Farre, R and Clemente, G (2007) Fluorescence and color as markers for

310 the Maillard reaction in milk–cereal based infant foods during storage. Food Chemistry 105

311 1135-1143.

312 Corzo, N, López-Fandiño, R, Delgado, T, Ramos, M and Olano, A (1994) Changes in

313 furosine and proteins of UHT-treated milks stored at high ambient temperatures. Zeitschrift

314 für Lebensmittel-Untersuchung und Forschung 198 302-306.

315 Czerwenka, C, Maier, I, Pittner, F and Lindner, W (2006) Investigation of the Lactosylation

316 of Whey Proteins by Liquid Chromatography−Mass Spectrometry. Journal of Agricultural

317 and Food Chemistry 54 8874-8882.

318 Dalsgaard, T K, Nielsen, J H and Larsen, L B (2007) Proteolysis of milk proteins lactosylated

319 in model systems. Mol Nutr Food Res 51 404-414.

320 Dannenberg, F and Kessler, H-G (1988) Reaction Kinetics of the Denaturation of Whey

321 Proteins in Milk. Journal of Food Science 53 258-263.

16 322 Datta, N, Elliott, A J, Perkins, M L and Deeth, H C (2002) Ultra-high-temperature (UHT)

323 treatment of milk: comparison of direct and indirect modes of heating Australian Journal of

324 Dairy Technology 57 211-227.

325 Erbersdobler, H F and Somoza, V (2007) Forty years of furosine – Forty years of using

326 Maillard reaction products as indicators of the nutritional quality of foods. Molecular

327 Nutrition & Food Research 51 423-430.

328 Johnson, P, Philo, M, Watson, A and Mills, E N C (2011) Rapid Fingerprinting of Milk

329 Thermal Processing History by Intact Protein Mass Spectrometry with Nondenaturing

330 Chromatography. Journal of Agricultural and Food Chemistry 59 12420-12427.

331 Le, T T, Bhandari, B and Deeth, H C (2011) Chemical and Physical Changes in Milk Protein

332 Concentrate (MPC80) Powder during Storage. Journal of Agricultural and Food Chemistry

333 59 5465-5473.

334 Le, T T, Deeth, H C, Bhandari, B, Alewood, P F and Holland, J W (2013) Quantification of

335 lactosylation of whey proteins in stored milk powder using multiple reaction monitoring.

336 Food Chemistry 141 1203-1210.

337 Losito, I, Carbonara, T, Monaci, L and Palmisano, F (2007) Evaluation of the thermal history

338 of bovine milk from the lactosylation of whey proteins: an investigation by liquid

339 chromatography–electrospray ionization mass spectrometry. Analytical and Bioanalytical

340 Chemistry 389 2065-2074.

341 Losito, I, Stringano, E, Carulli, S and Palmisano, F (2010) Correlation between lactosylation

342 and denaturation of major whey proteins: an investigation by liquid chromatography–

343 electrospray ionization mass spectrometry. Analytical and Bioanalytical Chemistry 396 2293-

344 2306.

345 Meltretter, J, Birlouez-Aragon, I, Becker, C-M and Pischetsrieder, M (2009) Assessment of

346 heat treatment of dairy products by MALDI-TOF-MS. Molecular Nutrition & Food Research

347 53 1487-1495.

17 348 Nangpal, A and Reuter, H (1990) Reference diagram for furosine content in UHT milk. Kieler

349 Milchw. Forsch. 45 77-86.

350 Pinto, G, Caira, S, Cuollo, M, et al. (2012) Lactosylated casein phosphopeptides as specific

351 indicators of heated milks. Analytical and Bioanalytical Chemistry 402 1961-1972.

352 Scaloni, A, Perillo, V, Franco, P, et al. (2002) Characterization of heat-induced lactosylation

353 products in caseins by immunoenzymatic and mass spectrometric methodologies. Biochimica

354 et Biophysica Acta (BBA) - Proteins and Proteomics 1598 30-39.

355 Siciliano, R A, Mazzeo, M F, Arena, S, Renzone, G and Scaloni, A (2013) Mass spectrometry

356 for the analysis of protein lactosylation in milk products. Food Research International 54

357 988-1000.

358 Steffan, W, Balzer, H H, Lippert, F, Sambor, B C, Bradbury, A G W and Henle, T (2006)

359 Characterization of casein lactosylation by capillary electrophoresis and mass spectrometry.

360 European Food Research and Technology 222 467-471.

361 Van Boekel, M a J S (1998) Effect of heating on Maillard reactions in milk. Food Chemistry

362 62 403-414.

18 Tables

Table 1: Degree of whey protein denaturation and furosine content in UHT milks one day after processing. The values for β-lg are an average of β-lg A and B. Different superscripts in one column indicate significant difference (P<0.05).

Whey protein Furosine Milk Heat treatment denaturation [%] (mg/100 g protein)

β-lg α-la 95°C/5 s + Direct A 86.6 ± 2.1a 26.6 ± 4.0 a 11.6 ± 5.6 a > 150°C /< 0.2 s 95°C/180 s + Direct B 94.1 ± 1.6 b 45.5 ± 3.3 b 7.6 ± 4.6 a > 150°C /< 0.2 s Indirect 140 °C/4 s 93.6 ± 1.5 b 70.0 ± 6.1 c 119.9 ± 23.0 b

19 Table 2: Initial lactosylation and rate of lactosylation of reduced caseins in directly heated UHT milk (> 150 °C for < 0.2 s) pre-heated at 95 °C for 5 s (Direct A) or 180 s (Direct B) and in indirect UHT milk (140°C for 4 s, Indirect) during storage at 20 °C. Results from linear regression (y=ax+b) shown as average ± 95 % confidence interval. Different superscripts within a row for ‘slope a’ or within a row for ‘initial lactosylation b’ indicate significant difference (P<0.05).

Protein Slope a (% month-1) Initial lactosylation b (%) R2 Direct A Direct B Indirect Direct A Direct B Indirect Direct A Direct B Indirect a a b a b c αS1-CN -4.210±0.209 -3.837±0.216 -2.730±0.491 90.10±0.74 87.87±0.76 63.82±1.72 0.99 0.99 0.87 a a b a a b αS1-CN+lac 3.449±0.216 3.110±0.221 1.631±0.380 8.19±0.76 9.59±0.78 28.67±1.12 0.98 0.98 0.85 a a b a b c αS1-CN+2lac 0.7609±0.055 0.726±0.039 1.099±0.181 1.71±0.20 2.54±0.14 7.51±0.64 0.98 0.99 0.90 a a b a a c αS2-CN -2.293±0.299 -2.368±0.846 -1.385±1.002 78.34±1.05 78.40±2.98 58.68±3.62 0.94 0.66 0.31 a ab b a a b αS2-CN+lac 1.715±0.210 1.794±0.582 0.590±0.680 15.97±0.74 16.10±2.05 29.36±2.72 0.94 0.70 0.15 a a a a a c αS2-CN+2lac 0.578±0.136 0.575±0.295 0.795±0.344 5.37±0.33 5.50±1.04 11.68±1.24 0.93 0.48 0.55 β-CN A1 -2.607±0.132a -2.892±0.406a -2.002±0.529a 90.93±0.47a 88.40±1.43b 70.69±1.91c 0.99 0.97 0.77 β-CN A1+lac 2.337±0.125a 2.537±0.334a 1.367±0.413b 7.55±0.44a 9.09±1.18b 23.94±1.41c 0.99 0.97 0.72 β-CN A1+2lac 0.270±0.036a 0.355±0.082a 0.635±0.135b 1.52±0.13a 2.51±0.29b 5.37±0.49c 0.93 0.92 0.84 β-CN A2 -2.534±0.162ac -2.076±0.176bc -1.944±0.614abc 91.20±0.57a 90.78±0.62a 70.66±1.81c 0.98 0.93 0.78 β-CN A2+lac 2.194±0.142a 1.828±0.159b 1.284±0.477b 6.65±0.50a 7.52±0.56a 23.71±1.39c 0.98 0.93 0.71 β-CN A2+2lac 0.340±0.045a 0.247±0.036b 0.660±0.146c 2.16±0.16a 1.69±0.13b 5.72±0.51c 0.93 0.82 0.83 κ-CN -1.543±0.112a -1.613±0.143a -1.280±0.321a 96.37±0.40a 95.00±0.50b 82.61±1.16c 0.98 0.97 0.98 κ-CN+lac 1.543±0.112a 1.613±0.143a 1.280±0.321a 3.63±0.40a 5.00±0.50b 17.39±1.16c 0.98 0.97 0.98

20 Table 3: Initial lactosylation and rate of lactosylation of whey proteins in directly heated UHT milk (> 150 °C for < 0.2 s) pre-heated at 95 °C for 5 (Direct A) or 180 s (Direct B) and in indirect UHT milk (140°C for 4 s, Indirect) during storage at 20 °C. Results from linear regression (y=ax+b). Different superscripts within a row for ‘slope a’ or within a row for ‘initial lactosylation b’ indicate significant difference (P<0.05).

Protein Slope a (% month-1) Initial lactosylation b (%) R2 Direct A Direct B Indirect Direct A Direct B Indirect Direct A Direct B Indirect α-la -2.626±0.580a -2.352±0.199a -1.686±0.491a 95.74±1.99a 91.80±0.70b 71.85±1.77c 0.85 0.97 0.73 α-la +lac 2.626±0.580a 2.352±0.199a 1.686±0.491a 4.26±1.99a 8.20±0.70b 28.15±1.77c 0.85 0.97 0.73 β-lg A -4.032±0.461a -3.313±0.188b -2.533±0.477c 89.27±1.62a 88.64±0.70a 65.51±1.72c 0.95 0.99 0.87 β-lg A+lac 2.893±0.462a 2.470±0.185a 1.256±0.287c 10.09±1.23a 9.83±0.60a 27.29±1.04c 0.95 0.98 0.81 β-lg A+2lac 1.140±0.081a 0.842±0.081b 1.277±0.207c 1.14±0.28a 1.53±0.30a 7.20±0.73c 0.98 0.96 0.90 β-lg B -4.004±0.257a -3.257±0.328b -2.408±0.447c 89.59±0.91a 87.61±1.20a 67.01±1.61c 0.98 0.96 0.87 β-lg B+lac 3.355±0.281a 2.745±0.303b 1.479±0.310c 9.65±0.99a 12.60±1.10b 27.63±1.12c 0.97 0.95 0.84 β-lg B+2lac 0.650±0.047a 0.512±0.064b 0.929±0.148a 0.76±0.17a 1.08±0.20a 5.36±0.53c 0.98 0.94 0.90

21 Figures

Figure 1: Furosine development in directly heated UHT milk (> 150 °C for < 0.2 s) pre-heated at 95 °C for 5 (○) or 180 s (●) and in indirect UHT milk (140 °C for 4 s, □) during storage at 20 °C.

Figure 2: Summed mass spectra (retention time 11.0-13.2 min) of unmodified and lactosylated

αS1-CN 8 P in indirect UHT milk (140 °C for 4 s) after 6 months of storage at 20 °C. Inset is showing a zoom of the most abundant ions. The m/z shifts due addition of lactose are indicated.

Figure 3: Extracted ion chromatogram of unmodified (solid line), mono-lactosylated (dashed line) and di-lactosylated (dash-dot line) αS1-CN 8 P in indirect UHT milk (140 °C for 4 s) after 6 months of storage at 20 °C using the ions shown in Figure 2 (A). Corresponding UV signal at 214 nm of αS1-CN 8 P and 9 P after 0 (solid line) and 6 (dashed line) months of storage at 20 °C (B).

Figure 4: Deconvoluted mass spectrum of αS1-CN 8 P in indirect UHT milk (140 °C for 4 s) after 0 (A) and 6 (B) months of storage at 20 °C. Mass shifts due to addition of lactose are indicated.

Fig 5: Absolute areas (top row) and area distribution (bottom row) of unmodified and lactosylated

αS1-CN 8 P in directly heated UHT milk (> 150 °C for < 0.2 s) pre-heated at 95 °C for 5 (A) or 180 s (B) and in indirect UHT milk (140 °C for 4 s, C) during storage at 20 °C.

Figure 6: Correlation of furosine content and % of unmodified protein for α-la (A), β-lg A (B), β- 2 CN A (C) and αS1-CN 8 P (D) from directly heated UHT milk (> 150 °C for < 0.2 s) pre-heated at 95 °C for 5 (●) or 180 s (○) and in indirect UHT milk (140 °C for 4 s, □). The results of a linear regression are shown.

Supplementary data

1. Summed mass spectra of unmodified and lactosylated caseins in indirect UHT milk (140 °C for 4 s) after 6 months of storage at 20 °C. Inset is showing a zoom of the most abundant ions. The m/z shifts due addition of lactose are indicated.

αS2-CN β-CN

κ-CN α-la

β-lg

2. Absolute and relative change in MS area of reduced αS2-CN during storage in Direct A (A), Direct B (B) and Indirect (C).

3. Absolute and relative change in MS area of reduced β-CN during storage in Direct A (A), Di- rect B (B) and Indirect (C).

3. Absolute and relative change in MS area of reduced κ-CN during storage in Direct A (A), Di- rect B (B) and Indirect (C).

4. Absolute and relative change in MS area of reduced α-la during storage in Direct A (A), Direct B (B) and Indirect (C).

5. Absolute and relative change in MS area of reduced β-lg during storage in Direct A (A), Direct B (B) and Indirect (C).

6. Development of summed intensity of unmodified and lactosylated reduced αS1-CN 8 P using EIC area, deconvolution intensity, profinder area and UV area at 214 nm of αS1-CN 8 P and 9 P in indirect UHT milk (140 °C for 4 s) during storage at 20 °C.