THE EFFECT OF DIOXIDE ADDITION AND ALTERNATIVE

FERMENTATION TECHNIQUES ON THE MICROBIAL COMMUNITIES AND

SENSORY PROFILES OF

By

Sydney Christian Morgan

A THESIS SUBMITTED IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

in

THE COLLEGE OF GRADUATE STUDIES

(Biology)

THE UNIVERSITY OF BRITISH COLUMBIA

(Okanagan)

August 2019

© Sydney Christian Morgan, 2019 The following individuals certify that they have read, and recommend to the College of Graduate Studies for acceptance, a dissertation entitled:

THE EFFECT OF ADDITION AND ALTERNATIVE TECHNIQUES ON THE MICROBIAL COMMUNITIES AND SENSORY PROFILES OF WINE

submitted by Sydney Christian Morgan in partial fulfillment of the requirements of the degree of Doctor of Philosophy.

Dr. Daniel Durall, Department of Biology, University of British Columbia, Okanagan Supervisor

Dr. Michael Deyholos, Department of Biology, University of British Columbia, Okanagan Supervisory Committee Member

Dr. John Klironomos, Department of Biology, University of British Columbia, Okanagan Supervisory Committee Member

Dr. Richard Plunkett, Department of Biology, University of British Columbia, Okanagan Supervisory Committee Member

Dr. David Scott, Department of Earth and Environmental Sciences, University of British Columbia, Okanagan University Examiner

Dr. Thomas Henick-Kling, School of Food Science, Washington State University External Examiner

ii

Abstract

Modern often involves the addition of sulfur dioxide (SO2), to remove potential spoilage microbes from the , and the addition of commercial wine , to ensure a successful and timely completion of alcoholic fermentation. However, consumer demand is shifting towards low-SO2 , and wines fermented by a collection of indigenous yeasts and bacteria. Many winemakers wish to produce these wines for their customers, but there is a current lack of research into the relative risks and rewards of these winemaking strategies.

We investigated these topics, and found that both SO2 addition and inoculation technique can significantly influence the fungal and bacterial communities and the sensory profiles of commercially-produced wines. conducted without

SO2 were found to contain more -derived and produced more fruity characters, while the fermentations conducted with SO2 produced wines with less-desirable characteristics. Pied de cuve inoculation had a limited effect on the wine microbial communities and was inhibited by the addition of SO2; when it persisted, it resulted in slower fermentations, and produced wines with more sweetness and body as compared to wines that did not receive the pied de cuve inoculum. At one commercial , we identified a highly diverse indigenous population of uvarum, a that is related to the main winemaking yeast, . S. uvarum was able to compete with a commercial S. cerevisiae strain in controlled fermentations, and produced wines with a unique composition of aroma-active volatile compounds, showing its potential to be used as an indigenous yeast inoculum for locally-made wines. In summary, we

iii showed that both SO2 addition and fermentation technique (uninoculated or pied de cuve inoculation) can alter the microbial communities and sensory profiles of commercial and laboratory-made wines, and reported an indigenous yeast population in the Okanagan Valley winemaking region.

iv

Lay summary

Yeasts and bacteria make wine by converting grape sugars to , and also produce hundreds of compounds that contribute to the aroma and flavour of wine. Today, most winemakers add commercial yeasts and bacteria to their grape must to ensure a successful fermentation and a consistent, predictable flavour profile. However, some winemakers are looking for minimal intervention approaches to winemaking, which can include adding fewer chemicals such as sulfur dioxide

(SO2), and letting native vineyard and winery microbes conduct fermentation. We found that wines made with different levels of SO2 contained different communities of microbes. We also found that low-SO2 wines contain more vineyard-derived yeasts and bacteria, and produce wines with more fruity aromas and flavours. In summary, we found that these minimal-intervention techniques can create wines with unique sensory profiles, and can be used successfully by winemakers to produce quality wines with more regional character.

v

Preface

A version of Chapter 2 has been published: Morgan, S.C., Scholl, C.M.,

Benson, N.L., Stone, M.L., Durall, D.M., 2017. Sulfur dioxide addition at crush alters

Saccharomyces cerevisiae strain composition in spontaneous fermentations at two

Canadian . Int. J. Food. Microbiol. 244, 96–102.

DOI:10.1016/j.ijfoodmicro.2016.12.025.

A version of Chapter 3 has been published: Morgan, S.C., Tantikachornkiat,

M., Scholl, C.M., Benson, N.L., Cliff, M.A., Durall, D.M., 2019. The effect of sulfur dioxide addition at crush on the fungal and bacterial communities and the sensory attributes of wines. Int. J. Food Microbiol. 290, 1–14.

DOI:10.1016/j.ijfoodmicro.2018.09.020.

A version of Chapter 4 has been published: Morgan, S.C., Haggerty, J.J.,

Johnston, B., Jiranek, V., Durall, D.M., 2019. Response to sulfur dioxide addition by two commercial Saccharomyces cerevisiae strains. Fermentation. 5, 1-20.

DOI:10.3390/fermentation5030069.

A version of Chapter 5 has been published: Morgan, S.C., McCarthy, G.C.,

Watters, B.S., Tantikachornkiat, M., Zigg, I., Cliff, M.A., Durall, D.M., 2019. Effect of addition and pied de cuve inoculation on the microbial communities and sensory profiles of wines: dominance of indigenous Saccharomyces uvarum at a commercial winery. FEMS Yeast Res. 19, foz049. DOI:

10.1093/femsyr/foz049.

The winery experiments described in Chapters 2, 3, and 5 were designed in consultation with my supervisor, Dr. Daniel Durall, and realized with assistance from

vi the winemakers Darryl Brooker, Taylor Whelan, Alexandra Haselich, Grant Stanley, and Matthew Fortuna. Sample collection was performed by myself, Natasha Benson

(honours student), Brittany Watters (honours student), and Chrystal Scholl

(laboratory technician). Chemical analysis was performed by myself, Natasha

Benson, and Morgan Stone (laboratory technician). S. cerevisiae and S. uvarum strain typing was performed by myself, Natasha Benson, Ieva Zigg (work-study student), Garrett McCarthy (graduate student) and Chrystal Scholl. Library preparation for Illumina was performed by myself and Mansak

Tantikachornkiat, a fellow graduate student. Illumina sequencing was performed by

Dan New and fellow technicians at IBEST at the University of Idaho. Illumina sequencing results were analyzed by myself, with assistance from Mansak

Tantikachornkiat. Technical assistance and guidance was provided by Marissa

Neuner (graduate student) and Stacey Sakakibara (visiting researcher). Sensory evaluations of wine were performed in collaboration with the Summerland Research and Development Centre (part of Agriculture and Agri-Food Canada), under the supervision of Dr. Margaret Cliff and with assistance from Kareen Stanich (sensory laboratory technician). I performed all statistical analysis and data visualization, and wrote all three manuscripts for publication.

The laboratory experiments described in Chapters 4 and 6 were conducted at the University of Adelaide in Adelaide, Australia, and were designed in consultation with Dr. Daniel Durall, as well as Dr. Vladimir Jiranek and Dr. Tommaso Watson from the University of Adelaide. I performed all experiments, including data collection and analysis, statistical analysis and data visualization, and the writing of

vii both manuscripts. Chemical analysis of the wine was performed under the invaluable guidance of Dr. Jade Haggerty. I also received training, guidance, and technical assistance from the following members of the Wine Group at the University of Adelaide: Nicholas Van Holst, Dr. Joanna Sundstrom, Louise

Bartle, Dr. Ee Lin Tek, Dr. Krista Sumby, Liang Chen, Tom Lang, Chen Liang, and

Dr. Jennifer Gardner.

viii

Table of contents Abstract ...... iii Lay summary ...... v

Preface ...... vi Table of contents ...... ix

List of tables ...... xiv

List of figures ...... xix

Acknowledgements ...... xxiv

Dedication ...... xxvii

1. Introduction ...... 1 1.1 History of winemaking ...... 1 1.2 Yeasts involved in winemaking ...... 4 1.2.1 Saccharomyces cerevisiae ...... 4 1.2.2 Non-Saccharomyces yeasts ...... 6 1.2.3 Spoilage yeasts ...... 7 1.3 Bacteria involved in winemaking ...... 9 1.3.1 Lactic Acid Bacteria ...... 9 1.3.2 Bacteria ...... 10 1.4 Microbial interactions during fermentation ...... 11 1.4.1 Yeast-yeast interactions ...... 11 1.4.2 Yeast-bacteria interactions ...... 13 1.5 Inoculation techniques ...... 14 1.6 Sulfur dioxide in winemaking ...... 17 1.7 Wine sensory profiles ...... 22 1.8 Methods of microbial identification...... 27 1.8.1 Culture-dependent identification...... 27 1.8.2 Culture-independent identification ...... 29 1.9 Objectives and predictions ...... 31 1.9.1 The effect of sulfur dioxide addition at crush on Saccharomyces cerevisiae strain composition during uninoculated fermentations ...... 31 1.9.2 The effect of sulfur dioxide addition at crush on the fungal and bacterial communities and the sensory attributes of Pinot gris wines ...... 32 ix

1.9.3 Response to sulfur dioxide addition by two commercial Saccharomyces cerevisiae strains ...... 33 1.9.4 The effect of sulfur dioxide addition and pied de cuve inoculation on the microbial communities and sensory profiles of Chardonnay wines ...... 34 1.9.5 Competition between Saccharomyces cerevisiae and Saccharomyces uvarum in controlled Chardonnay fermentations ...... 35 2. Effect of sulfur dioxide addition at crush on the Saccharomyces cerevisiae strain composition during uninoculated fermentations ...... 37 2.1 Background ...... 37 2.2 Materials and methods ...... 41 2.2.1 Study sites and experimental design ...... 41 2.2.2 Sampling ...... 42 2.2.3 Chemical analysis ...... 43 2.2.4 Yeast enumeration and isolation ...... 44 2.2.5 S. cerevisiae strain typing ...... 44 2.2.6 Sulfite resistance assay ...... 45 2.2.7 Statistical analysis ...... 46 2.3 Results and discussion ...... 47 2.3.2 S. cerevisiae relative strain abundance ...... 47 2.3.2 S. cerevisiae strain composition ...... 53 2.4 Summary ...... 56 3. The effect of sulfur dioxide addition at crush on the fungal and bacterial communities and the sensory attributes of Pinot gris wines ...... 57 3.1 Background ...... 57 3.2 Materials and methods ...... 62 3.2.1 Experimental design and sampling ...... 62 3.2.2 Chemical analysis ...... 63 3.2.3 Sample treatment with propidium monoazide ...... 65 3.2.4 DNA extraction and Illumina MiSeq library preparation ...... 66 3.2.4.1 Amplicon PCR ...... 67 3.2.4.2 Index PCR ...... 68 3.2.5 Illumina MiSeq data processing ...... 69 3.2.6 Wine bottling and sensory analysis ...... 71 3.2.7 Statistical analysis ...... 73 x

3.3 Results and Discussion ...... 76 3.3.1 Sulfur dioxide and ...... 76 3.3.2 Fungal and bacterial diversity during fermentation ...... 81 3.3.3 Fungal and bacterial abundance during fermentation ...... 84 3.3.4 Fungal and bacterial community composition ...... 91 3.3.5 Wine sensory attributes ...... 95 3.4 Summary ...... 101 4. Response to sulfur dioxide addition by two commercial Saccharomyces cerevisiae strains ...... 102 4.1 Background ...... 102 4.2 Materials and methods ...... 106 4.2.1 Experimental design ...... 106 4.2.2 Inoculation, fermentation, and sampling ...... 107 4.2.3 Residual sugar concentration ...... 108 4.2.4 Yeast abundance ...... 109 4.2.5 Sulfur dioxide determination and sulfite resistance assay ...... 109 4.2.6 determination ...... 110 4.2.7 determination ...... 110 4.2.8 Secondary metabolite analysis...... 111 4.2.8.1 Chemical standards ...... 112 4.2.8.2 Quality assurance and standard curves ...... 113 4.2.8.3 Sample preparation ...... 113 4.2.8.4 Instrumentation and parameters ...... 113 4.2.9 Strain-typing ...... 114 4.2.10 Statistical analysis ...... 114 4.3 Results and discussion ...... 117 4.3.1 Fermentation kinetics and cell density ...... 117 4.3.2 Sulfur dioxide concentration during fermentation ...... 122 4.3.3 Hydrogen sulfide production during fermentation ...... 125 4.3.4 Post-fermentation acetaldehyde concentration ...... 127 4.3.5 Yeast-derived secondary metabolite composition ...... 129 4.4 Summary ...... 135

xi

5. The effect of sulfur dioxide addition and pied de cuve inoculation on the microbial communities and sensory profiles of Chardonnay wines ...... 137 5.1 Background ...... 137 5.2 Materials and methods ...... 143 5.2.1 Experimental design and sampling ...... 143 5.2.2 Chemical analysis ...... 144 5.2.3 Saccharomyces strain identification ...... 145 5.2.3.1 Yeast isolation and DNA extraction ...... 145 5.2.3.2 Saccharomyces uvarum strain-typing ...... 145 5.2.3.3 Saccharomyces cerevisiae strain-typing ...... 147 5.2.4 High-throughput amplicon sequencing ...... 149 5.2.4.1 Sample treatment with PMA and DNA extraction ...... 149 5.2.4.2 Illumina MiSeq library preparation ...... 149 5.2.5 Illumina MiSeq data processing ...... 149 5.2.6 Wine bottling and sensory evaluation ...... 151 5.2.7 Statistical analysis ...... 153 5.3 Results and discussion ...... 156 5.3.1 Fermentation kinetics and wine chemistry ...... 156 5.3.2 Fungal communities ...... 158 5.3.3 Saccharomyces strains ...... 164 5.3.4 Bacterial communities ...... 168 5.3.5 Wine sensory attributes ...... 172 5.4 Summary ...... 178 6. Competition between Saccharomyces cerevisiae and Saccharomyces uvarum in controlled Chardonnay fermentations ...... 180 6.1 Background ...... 180 6.2 Materials and methods ...... 183 6.2.1 Experimental design ...... 183 6.2.2 Inoculation, fermentation, and sampling ...... 184 6.2.3 Sugar concentration, pH, and cell count determination ...... 186 6.2.4 Sulfur dioxide determination and sulfite resistance assay ...... 187 6.2.5 Secondary metabolite analysis...... 188 6.2.6 Statistical analysis ...... 189 xii

6.3 Results and discussion ...... 189 6.3.1 Sulfur dioxide and fermentation progression ...... 189 6.3.2 Yeast competition during fermentation ...... 192 6.3.3 Yeast-derived secondary metabolite composition ...... 197 6.4 Summary ...... 202 7. Conclusion ...... 204 7.1 Summary and general discussion ...... 204 7.2 Limitations ...... 208 7.3 Future directions ...... 211 7.4 Concluding remarks ...... 214 Literature cited ...... 215 Appendices ...... 241 Appendix A: Supplementary material for Chapter 2...... 241 Appendix B: Supplementary material for Chapter 3...... 251 Appendix C: Supplementary material for Chapter 4 ...... 257 Appendix D: Supplementary material for Chapter 5 ...... 259 Appendix E: Supplementary material for Chapter 6...... 272

xiii

List of tables Table 1.1. Selected compounds produced by wine yeasts and bacteria ...... 23 Table 2.1. The initial chemical composition ± SEM of Pinot gris must (W1) and must (W2) at cold-settling. T.A. refers to titratable acidity. Different superscript letters indicate significant differences between treatments (p < 0.05) at W2. Statistical analysis of must between treatments at W1 was not conducted as these initial samples were not independent ...... 43 Table 2.2. Relative abundance ± SEM of dominant S. cerevisiae strains in fermentations to which different levels of sulfur dioxide were added at crush at two wineries (W1 and W2). Relative abundance and standard error (SEM) values were obtained from 360 S. cerevisiae isolates (40 isolates from three fermentation stages, with three replicate fermentations per treatment). Dominant strains consist of those comprising at least 10% of the relative abundance throughout fermentation. Results in bold indicate dominance in that particular treatment ...... 52

Table 2.3. Results of one-factor PERMANOVA tests evaluating the effects of SO2 addition level at crush on S. cerevisiae strain composition throughout spontaneous fermentation at W1 and W2. Results are based on 999 unrestricted permutations of raw data. Statistical analysis was conducted separately for each winery ...... 56 Table A1. Microsatellite fingerprints of all S. cerevisiae strains identified at Winery 1 and Winery 2. Numbers represent the fragment lengths of alleles at eight loci .... 244 Table A2. Relative abundance ± SEM of all S. cerevisiae strains identified at three stages of spontaneous fermentations to which different levels of SO2 were added at crush at Winery 1. Samples were collected at the early (E), mid (M), and late (L) stages of fermentation. Relative abundance and standard error (SEM) values were obtained from 120 S. cerevisiae isolates identified at each stage (40 isolates from each replicate fermentation) ...... 246 Table A3. Relative abundance ± SEM of all S. cerevisiae strains identified at three stages of spontaneous fermentations to which different levels of SO2 were added at crush at Winery 2. Samples were collected at the early (E), mid (M), and late (L) stages of fermentation. Relative abundance and standard error (SEM) values were obtained from 120 S. cerevisiae isolates identified at each stage (40 isolates from each replicate fermentation) ...... 250 Table A4. Commercial S. cerevisiae strains used for inoculated fermentations at W1 in 2014 and previous ...... 252 Table A5. Commercial S. cerevisiae strains used for inoculated fermentations at W2 in 2014 ...... 253 Table 3.1. Chemical composition of Pinot gris grape must at Cold-settling ± SEM. Three different levels of SO2 were added to the must immediately prior to sampling

xiv

(n = 3 per treatment). Yeast assimilable nitrogen (YAN) was measured as the sum of alpha amino nitrogen (mg/L) and (mg/L) concentrations ...... 64 Table 3.2. Chemical composition of Pinot gris wine at the end of alcoholic fermentation (Late stage) ± SEM. Three levels of SO2 were added to the grape must at the cold-settling stage ...... 64 Table 3.3. Sensory attributes of wines evaluated by an expert panel of 10 judges . 73 Table 3.4. Fungal and bacterial diversity, measured as Simpson’s Index of Diversity, of Pinot gris wines to which different levels of SO2 were added at crush (n = 3 per treatment). Diversity was measured at four stages of alcoholic fermentation (Cold- settling, Early, Mid, Late) and reported ± SEM. Fungal and bacterial data were analyzed separately, and treatments were compared across all fermentation stages by performing repeated measures one-factor ANOVA and subsequent post-hoc tests when appropriate. An asterisk next to either fungal or bacterial diversity indicates significance at α = 0.05; treatments marked with different superscript letters had significantly different overall diversity across all stages of fermentation (α = 0.05) ...... 83 Table 3.5. Results of one-factor PERMANOVA tests evaluating the effects of sulfur dioxide (SO2) addition level at crush (0, 20, or 40 mg/L SO2, n = 3 per treatment) on fungal and bacterial composition throughout fermentations of uninoculated Pinot gris wines. Fungal composition was compared using three stages of fermentation (Early, Mid, Late) and bacterial composition was compared using four stages of fermentation (Cold-settling, Early, Mid, Late). Results are based on 999 unrestricted permutations of raw data. Statistical analysis was conducted separately for fungi and bacteria. Results marked with an asterisk are significant at α = 0.05 ...... 92 Table B1. Composition of sensory standards provided to the panel to aid with their wine assessments. Food standards were prepared in 50 mL neutral Sola-Nero (Arterra Wines, Mississauga, ON Canada) unless otherwise indicated ...... 254 Table B2. Partial statistical output of “panelperf” function evaluating the sensory panel’s performance in its ability to discriminate between wines of different treatments (SensoMineR package in RStudio (version 3.4.4). A panel of 10 wine experts was used for the sensory evaluation of wines to which three different levels of sulfur dioxide (SO2) had been added at crush (n = 3 per treatment). Panelists evaluated each of the nine wines in duplicate tasting sessions. P-values are reported from three-factor ANOVA [product (wines), panelist, sensory session (tasting replicate)], for each of the 18 sensory attributes. An asterisk indicates significance at α = 0.05 ...... 255 Table B3. Fungal community composition (based on 6,000 sequences per sample) of wines fermented with different levels of sulfur dioxide (SO2) added at crush (0, 20,

xv or 40 mg/L SO2), reported ± the standard error of the mean (SEM). Three replicate fermentations were conducted for each treatment for a total of nine barrels. Samples were taken from each barrel at four stages of fermentation (Cold-settling, Early, Mid, Late) ...... 256 Table B4. Bacterial community composition (based on 10,000 sequences per sample) of wines fermented with different levels of sulfur dioxide (SO2) added at crush (0, 20, or 40 mg/L SO2). Three replicate fermentations were conducted for each treatment for a total of nine barrels. Samples were taken from each barrel at four stages of fermentation (Cold Settling, Early, Mid, Late) ...... 258 Table 4.1. List of yeast-derived secondary metabolites measured, as well as their characteristic aromas (Acree and Arn, 2004a; Sun et al., 2018b) and odour detection thresholds (µg/L) in 10% (v/v) ethanol (Ferreira et al., 2000; Francis and Newton, 2005; Guth, 1997; Peinado et al., 2006, 2004; Salo, 1970; Siebert et al., 2005; compiled by Haggerty et al., 2016) ...... 112 Table 4.2. Results of a two-factor PERMANOVA evaluating the effects of yeast strain and SO2 addition on the volatile secondary metabolite profiles of Pinot gris wines. Results with an asterisk (*) are significant at α = 0.05 ...... 130 Table 4.3. Relative quantity (± SEM) of volatile compounds in wines fermented by different commercial yeast strains and in the presence of different sulfite levels (n = 4 per treatment). Compounds with an asterisk (*) indicate significant results from a two-factor ANOVA, and different superscript letters indicate significant differences among treatments (α = 0.05). Each compound/compound group was analyzed separately ...... 132 Table C1. List of yeast-derived volatile secondary metabolites as analyzed by HS- SPME GC-MS ...... 260 Table C2. Multilocus genotypes of Strain 1 and Strain 2. Numbers represent the fragments lengths of allele 1 (A1) and allele 2 (A2) at eight microsatellite loci ..... 261 Table 5.1. Experimental design. Each of the four treatments was replicated in triplicate ...... 144 Table 5.2. Primer sequences and characteristics of eleven microsatellite loci used for S. uvarum strain identification (Masneuf-Pomarede et al., 2016; Zhang et al., 2015) ...... 147 Table 5.3. Primer sequences and characteristics of eleven microsatellite loci used for S. cerevisiae strain identification (Legras et al., 2005; Pérez et al., 2001; Richards et al., 2009) ...... 148 Table 5.4. Sensory attributes of wines evaluated by an expert panel of 12 judges ...... 153

xvi

Table 5.5. Compositional analyses of stainless steel barrel-fermented Chardonnay wines to which different sulfite and inoculation treatments were applied at crush. Values are the mean ± SEM (n = 3 per treatment). Superscript letters represent the results of two-factor ANOVA and subsequent post-hoc tests performed for each analysis (α = 0.05); different superscript letters within a row indicate significantly different means ...... 157 Table 5.6. Fungal, bacterial and Saccharomyces strain diversity, measured as Simpson’s Index of Diversity (1-D), of stainless steel barrel-fermented Chardonnay to which different SO2 and inoculation treatments had been applied. Diversity ± SEM was measured at four stages of fermentation for the fungal and bacterial community (Cold-settling, Early, Mid, and Late), and across three stages for the Saccharomyces strains (Early, Mid, and Late) (n = 3 per treatment). Diversity was analyzed separately for the fungal, bacterial, and Saccharomyces data, and treatments were compared across fermentation stages by performing repeated measures two-factor ANOVA and subsequent post-hoc tests when appropriate. Treatments marked with different superscript letters had significantly different overall diversity across all stages (α = 0.05) ...... 159

Table 5.7. Results of two-factor PERMANOVA evaluating the effects of SO2 addition (0 or 40 mg/L) and inoculation treatments (Uninoculated or Pied de cuve inoculation) on the microbial communities of fungi, bacteria, and Saccharomyces strains. Results with an asterisk (*) are significant at α = 0.05 ...... 161 Table D1. Composition of sensory standards provided to the panel to aid with their wine evaluations. Food standards were prepared in 100 mL neutral Sola white wine (Arterra Wines, Mississauga, ON, Canada) unless otherwise indicated. Each panelist was given 5 mL of each aroma standard (and 10 mL of the acidity flavour standard) in a small labeled plastic cup sealed with a lid to contain the volatile compounds. Panelists were encouraged to refer back to these standards throughout the evaluation as needed ...... 262 Table D2. Fungal percent relative abundance (± SEM) at four stages of alcoholic fermentation (Cold-settling (C), Early (E), Mid (M), Late (L)), as well as the pied de cuve inoculum (P). Samples were taken from stainless-steel barrel-fermented Chardonnay to which different inoculation and SO2 treatments had been applied (n = 3 per treatment). Relative abundance was calculated from 20,000 sequences per sample. Sequences were identified to the species level unless otherwise indicated ...... 263 Table D3. Microsatellite identities of the representative multilocus genotypes (MLGs) of indigenous Saccharomyces uvarum strains isolated from stainless steel barrel- fermented Chardonnay at a commercial winery in British Columbia, Canada ...... 266 Table D4. Bacterial percent (%) relative abundance (± SEM) at four stages of alcoholic fermentation (Cold-settling (C), Early (E), Mid (M), Late (L)), as well as the xvii pied de cuve inoculum (P). Samples were taken from stainless-steel barrel- fermented Chardonnay to which different inoculation and SO2 treatments had been applied (n = 3 per treatment). Relative abundance was calculated from 8,500 sequences per sample. Sequences were identified to the genus level unless otherwise indicated ...... 272 Table D5. Partial statistical output of “panelperf” function, evaluating the sensory panel’s performance in its ability to discriminate between wines of different treatments (SensoMineR package in RStudio version 3.5.1). A panel of 12 wine experts evaluated 18 pre-determined sensory attributes of the wines, which were subjected to different sulfite and inoculation treatments (n = 2 per treatment). Panelists evaluated each wine twice, in duplicate tasting sessions, with wines served in random order. P-values reported are from three-factor ANOVA [product(wines), panelist, sensory session (tasting session)] for each of the attributes. Two judges were removed from the analysis of sulfur aroma and flavour: one judge neglected to evaluate one or more wines for this descriptor, and the other mistakenly evaluated the aroma and flavour of as opposed to sulfur compounds ...... 274 Table 6.1. Experimental design indicating the inoculation ratios and fermentation temperatures of laboratory-scale Chardonnay fermentations. Inoculation ratios are indicated in each case as the ratio of S. cerevisiae : S. uvarum. For every treatment, the total concentration of the yeast inoculum was 2.5 × 106 cells/mL. Each treatment contained three replicate fermentations ...... 184 Table 6.2. List of yeast-derived secondary metabolites measured, as well as their characteristic aromas (Acree and Arn, 2004a; Sun et al., 2018b) and odour detection thresholds (µg/L) in 10% (v/v) ethanol (Ferreira et al., 2000; Francis and Newton, 2005; Guth, 1997; Peinado et al., 2006, 2004; Salo, 1970; Siebert et al., 2005; compiled by Haggerty et al., 2016) ...... 188 Table E1. Relative quantity (± SEM) of volatile compounds in wines fermented by yeasts inoculated at different ratios (S. cerevisiae : S. uvarum) and fermented at different temperatures (n = 3 per treatment)...... 275

xviii

List of figures Figure 2.1. Commercial vs unknown strain abundance. Relative abundance (± SEM) of commercial and unknown S. cerevisiae strains in fermentations to which different levels of sulfur dioxide were added at crush at two wineries (W1 and W2). Relative abundance was determined from three fermentation stages, and each treatment had three replicate fermentations ...... 49 Figure 2.2. Percent composition of dominant S. cerevisiae strains at three stages of uninoculated fermentation at (A) Winery 1 and (B) Winery 2. Samples were taken during the Early (E), Mid (M), and Late (L) stages of fermentation. Dominant strains are those strains comprising at least 10% of the percent composition throughout the course of fermentation in at least one treatment. Values for each stage are the means of three replicate fermentations, where 40 S. cerevisiae isolates were identified for each sample for a total of 120 isolates per stage of fermentation. For variation between samples of the same treatment, please refer to Tables A.2 and A.3 ...... 50 Figure 2.3. S. cerevisiae strain composition. These PCoA ordinations describe the composition of S. cerevisiae strains during uninoculated fermentations with different levels of initial SO2 addition at (A) W1 and (B) W2. At W1, either 0, 20, or 40 mg/L SO2 was added at crush to Pinot gris must, and at W2, either 20 or 40 mg/L SO2 was added at crush to Pinot noir must. Individual data points represent the S. cerevisiae strain composition of a single sample (obtained using a Bray-Curtis dissimilarity index, calculated from the relative strain abundance of 40 S. cerevisiae isolates per sample). Each fermentation vessel was sampled at three stages of fermentation, and each treatment was replicated in triplicate, for a total of nine samples per treatment. One-factor PERMANOVA tests indicated significant differences in strain composition between treatments at each winery ...... 55 Figure 3.1. Residual sugar concentrations (° ±SEM) measured throughout AF of Pinot gris wines to which three different levels of sulfur dioxide had been added at crush (n = 3 per treatment). Fermentations from all treatments were complete within 11 days. A one-factor repeated measures ANOVA was performed, and no significant differences among treatments were found (F(2,6) = 3.90, p = 0.08) ...... 77

Figure 3.2. Free (A) and total (B) sulfur dioxide (SO2) concentrations ± SEM measured throughout AF of Pinot gris wines to which three different levels of total SO2 had been added at crush (n = 3 per treatment). SO2 was added in the form of metabisulfite (KMS) immediately prior to the sample taken at the Cold- settling stage of fermentation ...... 79 Figure 3.3. Relative abundance of fungi (based on 6,000 sequences per sample) present in wines fermented with three levels of sulfur dioxide (SO2) added at crush (n = 3 per treatment). Samples were taken at four stages of alcoholic fermentation.

xix

For variation among samples of the same treatment and for the identities of minor fungi, please refer to Table B.3 ...... 85 Figure 3.4. Relative abundance of bacteria (based on 10,000 sequences per sample) present in wines fermented with three levels of sulfur dioxide (SO2) added at crush (n = 3 per treatment). Samples were taken at four stages of alcoholic fermentation. For variation among samples of the same treatment and for the identities of minor bacteria, please refer to Table B.4 ...... 90 Figure 3.5. Principal coordinates analysis (PCoA) ordination representing the fungal community composition of wines fermented with three levels of sulfur dioxide (SO2) added at crush: 0 mg/L SO2 (black), 20 mg/L SO2 (dark grey), or 40 mg/L SO2 (light grey) (n = 3 per treatment). Dimension 1 (Dim 1) explains 49.3% of total variation and Dimension 2 (Dim 2) explains 6.0% of variation. Samples were taken at three stages of alcoholic fermentation, for a total of nine samples per treatment. One- factor PERMANOVA tests indicated significant differences in fungal community composition among treatments (P ≤ 0.03 for all comparisons) ...... 93 Figure 3.6. Principal coordinates analysis (PCoA) ordination representing the bacterial community composition of wines fermented with three levels of sulfur dioxide (SO2) added at crush: 0 mg/L SO2 (black), 20 mg/L SO2 (dark grey), or 40 mg/L SO2 (light grey) (n = 3 per treatment). Dimension 1 (Dim 1) explains 57.7% of total variation and Dimension 2 (Dim 2) explains 4.2% of variation. Samples were taken at four stages of alcoholic fermentation, for a total of 12 samples per treatment. A one-factor PERMANOVA test indicated no significant differences in bacterial community composition among treatments (F(2,33) = 0.02, p = 0.99) ...... 95 Figure 3.7. Radar plot depicting the normalized relative intensity of sensory attributes of Pinot gris wines fermented with three different levels of sulfur dioxide (SO2) added at crush. Radar plots were created using the average of three replicate wines per SO2 treatment, evaluated in duplicate by a panel of 10 wine experts. Values were standardized separately for each sensory attribute. For visual clarity, variation among treatments for each attribute was not included ...... 96 Figure 3.8. Principal component analysis (PCA) ordination of the sensory profiles of wines fermented with three different levels of sulfur dioxide (SO2) added at crush (n = 3 per treatment). Plots depict: (A) the variables factor map, and (B) the individuals factor map. Sensory attributes were analyzed by a panel of 10 wine experts. Attributes ending with “A” indicate aroma attributes, and those ending in “F” indicate flavour attributes ...... 97 Figure 4.1. Daily residual sugar levels (g/L ± SEM) of controlled fermentations conducted by different yeast strains (n = 4 per treatment). Fermentations were conducted in the presence of either (A) 0 mg/L SO2 or (B) 50 mg/L SO2. Treatments with different superscript letters indicate significantly difference fermentation progressions (two-factor repeated measures ANOVA, α = 0.05) ...... 118 xx

Figure 4.2. Total yeast abundance (CFU/mL ± SEM) during controlled fermentations conducted by different yeast strains (n = 4 per treatment). Fermentations were conducted in the presence of either (A) 0 mg/L SO2 or (B) 50 mg/L SO2. Treatments with different superscript letters indicate significantly different yeast abundance dynamics throughout fermentation (two-factor repeated measures ANOVA, α = 0.05) ...... 121

Figure 4.3. SO2 levels (mg/L) throughout fermentations conducted by different yeast strains and in the presence of different levels of initial SO2. Wine samples from each of four replicates were combined for SO2 analysis due to volume constraints. Plots represent (A) Total SO2 produced by yeasts throughout fermentations containing 0 mg/L SO2; (B) Total SO2 levels throughout fermentations containing 50 mg/L SO2; (C) Free SO2 levels throughout fermentations containing 50 mg/L SO2 ...... 124

Figure 4.4. H2S production (ppm ± SEM) by different yeast strains when fermented in the presence of 50 mg/L SO2. Treatments with different superscript letters indicate significant differences in overall H2S production dynamics (two-factor repeated measures ANOVA, α = 0.05) ...... 126 Figure 4.5. Acetaldehyde production (g/L ± SEM) by different yeast strains when fermented in the presence of 0 or 50 mg/L SO2. Treatments with different letters indicate significant differences in acetaldehyde concentration (two-factor ANOVA, α = 0.05) ...... 129 Figure 4.6. PCoA visualizing the production of volatile secondary metabolites by different yeast strains in controlled Pinot gris fermentations containing either 0 mg/L or 50 mg/L SO2 (n = 4 per treatment). Plots depict (A) the individual factors map and (B) the variable factors map showing secondary metabolites as vectors. On the individual factors map, each point represents the entire volatile secondary metabolite composition of a single wine sample. Points that are closer together contain more similar volatile profiles than points that are further apart. On the variable factors map, the length of the variable vector reflects that variable’s relative contribution in producing this ordination. Variable vectors that point towards a sample are positively correlated with that sample, and those that point away from a sample are negatively correlated. Variable vectors that appear at a ~90 angle are not correlated with each other, while those that appear at < 90 and > 90 angles are positively and negatively correlated with each other, respectively ...... 131 Figure 5.1. Progression of alcoholic fermentation, measured as residual sugar content (Brix ± SEM), of Chardonnay to which different SO2 and inoculation treatments were applied at crush (n = 3). The wines from the 0 mg/L Pied de cuve treatment conducted alcoholic fermentation at a significantly slower rate than the other three treatments (p < 0.001) ...... 157 Figure 5.2. Relative abundance (± SEM) of the dominant fungi present in Chardonnay at four stages of fermentation, as well as the composition of the pied de xxi cuve inoculum, based on 20,000 sequences per sample. SO2 and inoculation treatments were as follows (n = 3 per treatment): (A) 0 mg/L Uninoculated; (B) 40 mg/L Uninoculated; (C) 0 mg/L Pied de cuve; (D) 40 mg/L Pied de cuve. Any fungal taxa that did not achieve at least 10% relative abundance in any one sample were termed Minor Fungi ...... 162 Figure 5.3. Relative abundance (± SEM) of Saccharomyces strains present at three stages of fermentation of Chardonnay must. SO2 and inoculation treatments were as follows (n = 3 per treatment): (A) 0 mg/L Uninoculated; (B) 40 mg/L Uninoculated; (C) 0 mg/L Pied de cuve; (D) 40 mg/L Pied de cuve. Any S. uvarum strains that did not achieve at least 10% relative abundance in any one sample were termed Minor S. uvarum strains ...... 166 Figure 5.4. PCoA ordination of the Saccharomyces strain composition in wines to which different levels of SO2 were added at crush, and to which different inoculation treatments were applied. Individual data points represent the composition of Saccharomyces strains (both S. uvarum and S. cerevisiae) in a single sample. Samples were collected at three stages of alcoholic fermentation and each treatment contained three replicates, for a total of nine samples per treatment .... 168 Figure 5.5. Relative abundance (± SEM) of the dominant bacteria present in Chardonnay at four stages of fermentation, as well as the composition of the pied de cuve inoculum. SO2 and inoculation treatments were as follows (n = 3 per treatment): (A) 0 mg/L Uninoculated; (B) 40 mg/L Uninoculated; (C) 0 mg/L Pied de cuve; (D) 40 mg/L Pied de cuve. Any bacterial taxa that did not achieve at least 10% relative abundance in any one sample were termed Minor Bacteria ...... 170 Figure 5.6. Radar plot depicting the standardized relative intensity of sensory attributes of Chardonnay wines to which different levels of SO2 and different inoculation practices were implemented at crush (n = 2 per treatment). Wines were evaluated by a panel of 12 industry experts. Values were standardized separately for each sensory attribute. Attributes with a asterisk (*) were found to be significantly different among treatments (p < 0.05) ...... 174 Figure 5.7. Principal component analysis (PCA) of the sensory attributes of Chardonnay wines to which different levels of SO2 and different inoculation practices were implemented at crush (n = 2 per treatment). Wines were evaluated by a panel of 12 industry experts. The numbers in brackets indicate the two replicate wines within each treatment. The first dimension (Dim 1) represents 75.94% of total variance, while the second dimension (Dim 2) represents 8.63% of total variance 175 Figure 6.1. Daily residual sugar (g/L) levels (± SEM) of controlled Chardonnay fermentations to which different inoculation ratios (S. cerevisiae : S. uvarum). Fermentations were conducted at (A) 24 C and (B) 15 C ...... 192

xxii

Figure 6.2. Percent relative abundance (± SEM) of S. cerevisiae and S. uvarum, as well total yeast count (CFU/mL ± SEM) during controlled Chardonnay fermentations conducted at two temperatures, to which the following inoculation ratios (S. cerevisiae : S. uvarum) were applied: (A) 1:0, (B) 10:1, (C) 1:1, (D) 1:10, (E) 0:1...... 194 Figure 6.3. PCoA visualizing the production of volatile secondary metabolites by yeasts in controlled Chardonnay fermentations to which different inoculation ratios (S. cerevisiae : S. uvarum) and temperature treatments (24 or 15 C) were applied (n = 3 per treatment). Plots depict (A) the individual factors map and (B) the variable factors map showing secondary metabolites as vectors. Points that are closer together contain more similar volatile profiles than points that are further apart. On the variable factors map, the length of the variable vector reflects that variable’s relative contribution in producing this ordination. Variable vectors that point towards a sample are positively correlated with that sample, and those that point away from a sample are negatively correlated. Variable vectors that appear at a ~90 angle are not correlated with each other, while those that appear at < 90 and > 90 angles are positively and negatively correlated with each other, respectively ...... 199

xxiii

Acknowledgements

There are so many people without whom my PhD endeavour would not have been nearly as successful, and I would like to acknowledge and thank these people here. First, to my supervisor Dr. Daniel Durall, thank you for meeting with me over seven years ago and agreeing to take me on as an honours student. I didn’t know it then, but that first winery experiment in 2013 would change the course of my life and ignite my passion for research, wine, and of course wine research. I will be forever grateful for your guidance and mentorship these last many years. To my committee,

Dr. Richard Plunkett, Dr. John Klironomos, and Dr. Michael Deyholos, thank you for your support and for lending me your expertise when I needed it.

A community of people helped make my stint as a visiting researcher at the

University of Adelaide a success. Special thanks to Dr. Vladimir Jiranek, who provided me with lab space, equipment, and experimental design assistance, and to

Dr. Tommaso Watson and Nicholas Van Holst for training me. Thank you to the other researchers who assisted me with sampling, training, guidance, and much more: Dr. Joanna Sundstrom, Louise Bartle, Dr. Ee Lin Tek, Dr. Paul Grbin, Liang

Chen, Dr. Krista Sumby, Tom Lang, Dr. Jennifer Gardner, Chen Liang, and many more. Special thank you to Dr. Jade Haggerty for her invaluable help with chemical analysis.

Thank you to the past and present members of the Durall wine microbiology group: Chrystal Scholl, Mansak (Ben) Tantikachornkiat, Marissa Neuner, Britney

Johnston, Morgan Stone, Stacey Sakakibara, Brianne Newman, Garrett McCarthy,

Sarah Lyons, and especially my two honours students Natasha Benson and Brittany

xxiv

Watters. Your support, both academic and otherwise, has been instrumental in my success. It was a joy to work alongside you and I value the friendships we have forged.

Thank you to the many people who have given me the opportunity to teach throughout my degree: Dr. Janet Kluftinger, Dr. Simone Castellarin, Dr. Blythe

Nilson, and Dr. Brendan D’Souza of the University of British Columbia, and Dr.

Gabriel Balint and Jonathan Rouse of Okanagan College.

A team of people also provided me with assistance over the years. Thank you to the winemakers of Cedar Creek, 50th Parallel, and Mission Hill: Darryl Brooker,

Alexandra Haselich, Taylor Whelan, Grant Stanley, and Matthew Fortuna. Thank you to Dr. Margaret Cliff and Kareen Stanich from the sensory lab at the

Summerland Research and Development Centre, and thank you to Dan New from

IBEST at the University of Idaho. Thank you also to Dr. Miranda Hart and Dr. Mark

Rheault for equipment, and to Barb Lucente for helping keep the Biology department running.

I would also like to thank my family, Mom, Dad, and Carmen. Mom and Dad, thank you for everything you have given me. Your unwavering support has opened so many opportunities for me that I never could have dreamed about. I am so grateful to have you cheering me on every step of the way. Carmen, you have been many things to me over the years—guinea pig for my bed-hopping experiments, confidante, roommate—but you’ve always been my best friend and my amazing little sister.

xxv

Finally, thank you to my partner Mehrbod. I can’t begin to list the ways you’ve improved my life. Your love, sense of humour, unwavering belief in my abilities, and of course your bioinformatics expertise are just some of the reasons I am grateful for you. I look forward to our future as “Dr. and Dr.”

xxvi

Dedication

For great-grandma Christian, the first woman in my family to get a university degree, and for granny Barb, whose legacy of love and compassion lives on in our hearts.

xxvii

Chapter 1: Introduction

In this thesis I explored the role of sulfur dioxide (SO2) addition and alternative fermentation techniques on the microbial communities and sensory profiles of wine. In this chapter, I provide an overview of winemaking, including different winemaking techniques and the microorganisms that conduct fermentation.

Next, I summarize the role of SO2 in winemaking, and methods that wine yeasts may use to survive in the presence of SO2. I then discuss the role of yeasts and bacteria in modulating wine sensory profiles, as well as current methods of studying the microbial communities and populations in wine. Finally, I introduce the objectives and predictions of my thesis, and outline the specific goals of each research chapter.

1.1 History of winemaking

Wine has been involved in human society for millennia. The earliest evidence of intentional winemaking is from over 6,000 years ago in the Caucasus mountains, in what is today known as Armenia (Owen, 2011). Modern wine were originally domesticated in the Middle East, and from there winemaking quickly spread to Northern Africa and Europe, following major civilizations including the

Phoenicians, the Ancient Greeks, and the Ancient Romans (This et al., 2006). Under the Roman Empire, grape growing and winemaking expanded throughout Europe, often along trading routes (i.e. river valley systems). During the Crusades, the

Catholic Church further spread wine production to the rest of Europe, and it was during this time that many of the grape we still use today appeared, including and Chardonnay (This et al., 2006).

1

These areas⎯ the Middle East, Europe, and Northern Africa⎯ are known as

Old World winemaking regions. Winemaking first came to the New World⎯ the

Americas, South Africa, Australia, New Zealand, and Asia⎯ as early as the 1500s, when Spanish missionaries brought grapes to Mexico and popularized winemaking there (Teeter, 2015; This et al., 2006). From Mexico, winemaking spread north and south to the rest of the Americas. Wine made its way to South Africa in the late

1600s, and vine clippings were brought to Australia with the “First Fleet” of British prisoners in 1788. However, these vines did not survive, and more vines were sent to Australia and New Zealand in the 1800s (Clarke, 2004; This et al., 2006).

Although the European grapevine species vinifera is the most commonly used species for wine production, many enologically important species native to

North America exist. In the 1800s, North American grapevine clippings were brought back to Europe, and with them came Daktulosphaira vitifoliae, more commonly known as Phylloxera (Granett et al., 2001; Ordish, 1987). This aphid-like sap- sucking pest attacks the roots of grapevines, causing deformations and secondary infections, and quickly decimated the V. vinifera grapevine population of Europe

(Ordish, 1987). However, North American grapevine species, including Vitis riparia,

Vitis rupestris, and Vitis aestivalis, are naturally resistant to Phylloxera, and a solution to the Phylloxera epidemic was soon realized: susceptible V. vinifera varietals were grafted onto the roots of the naturally-resistant North American grapevine species. To this day, there is still no cure for Phylloxera, and grafting remains the only effective means of controlling this disease (Granett et al., 2001).

2

A general distinction is often made between Old World and New World wines.

Old World wines tend to be characterized by lower , higher acidity, and earthy aromas and flavours, while New World wines tend to be characterized by higher alcohol, lower acidity, and fruity aromas and flavours (Puckette, 2018). While many claim this is due to the soil the grapes grow in, this distinction is actually a result of climatic differences. In general, New World winemaking regions tend to be in hotter climates, which results in riper grapes with lower acidity. Riper grapes contain more sugar, which get converted into higher levels of alcohol, and hot climates/ripe grapes also lead to a more fruit-forward expression of the wines.

Cooler climates, such as those found in most of the regions, favour the expression of green and earthy characters, and also help retain the acidity levels in the grape.

Wine has been made in the Okanagan Valley of British Columbia, Canada, for over 150 years. The French priest and missionary Father Charles Pandosy planted the first grape vines in 1859. Originally, the North American grapevine species Vitis labrusca was planted, but these original vines have since been replaced with V. vinifera varietals (Robinson and Harding, 2015). When the current wine industry in the Okanagan Valley was first established in the late 1960s, most of the grapes being grown were lesser-known German varietals, including Ehrenfelser and , due to the similarities in climates between the two regions. Today, however, most grape varietals planted in the valley are noble European varietals, including Merlot, Pinot noir, Pinot gris, Chardonnay, and . Because of the north-south orientation of the valley, a wide range of grape varietals can be grown,

3 including heavy reds such as Merlot and in the south, and cool- climate grapes such as Pinot noir and Riesling in the north. Furthermore, the valley sits in the rain shadow of the Cascade mountain range, providing a hot, sunny, dry climate in which wine grapes thrive. Today, the Okanagan Valley is the second- largest winemaking region in Canada and is the second-largest producer of icewine in the world (Schreiner, 2009).

1.2 Yeasts involved in winemaking

Yeasts are the most important microorganisms involved in winemaking, as they conduct alcoholic fermentation (AF), the conversion of grape sugars into ethanol and (CO2). Three general groups of wine yeasts exist:

Saccharomyces cerevisiae, non-Saccharomyces yeasts, and spoilage yeasts.

1.2.1 Saccharomyces cerevisiae

S. cerevisiae is the dominant yeast in winemaking, and usually conducts the bulk of AF. This yeast has evolved alongside human activity for millennia, and thousands of different strains of S. cerevisiae are involved in the production of , , cider, wine, and other fermented foods and beverages (Cavalieri et al.,

2003; Sicard and Legras, 2011). S. cerevisiae has been called the first

‘domesticated’ microbe, and there is evidence that the genetic diversity of wine yeasts has been strongly influenced by human behaviour and technology (Fay and

Benavides, 2005; Legras et al., 2007; Liti et al., 2009). Over time, strains have been selected for their fermentative abilities, as well as their abilities to enhance other wine qualities, such as absorbing , thereby improving the colour of red wines (Suárez-Lepe and Morata, 2012). These strains were eventually

4 commercialized and sold as active dry yeast (ADY). Although hundreds of commercial ADY strains are available, research has shown that these commercial strains are less genetically diverse than S. cerevisiae strains associated with natural or clinical environments (Cubillos et al., 2009; Fay and Benavides, 2005).

Furthermore, commercial ADY strains show strong evidence of genetic redundancy or inbreeding, and some strains sold by different suppliers have been found to be genetically identical (Borneman et al., 2016; Fernández-Espinar et al., 2001).

S. cerevisiae is predominantly a winery-resident yeast, and is rarely isolated from healthy grapes or the vineyard (Mortimer and Polsinelli, 1999). The natural origins of S. cerevisiae are still under investigation, but strains can be found in association with trees and wasps, which may act as natural yeast reservoirs

(Blackwell and Kurtzman, 2016; Dashko et al., 2016; Jouhten et al., 2016; Sampaio and Gonçalves, 2008; Stefanini et al., 2012). While Drosophila fruit flies are often thought to be the most common wine yeast vector, they are only naturally associated with non-Saccharomyces yeast species, not with S. cerevisiae (Hoang et al., 2015; Lam and Howell, 2015). Most S. cerevisiae strains found in wine are diploid, but can sometimes be triploid, polypoid, aneuploid, or, rarely, haploid

(reviewed in Sipiczki, 2011). Little is currently known about the mating behaviour of

S. cerevisiae in nature, as it does not often undergo sexual reproduction in winery environments, instead favouring asexual reproduction via budding. However, new research suggests that the intestines of wasps may provide a succession of environments that promote outbreeding between strains of S. cerevisiae, and even between closely-related Saccharomyces species (Stefanini et al., 2016b).

5

1.2.2 Non-Saccharomyces yeasts

Other yeasts relevant to winemaking fall under the broad umbrella of ‘non-

Saccharomyces’ yeasts. Confusingly, this group also includes some members of the genus Saccharomyces, including Saccharomyces uvarum, which can be found at all stages of winemaking (Demuyter et al., 2004; Magyar and Tóth, 2011; Rementeria et al., 2003). Non-Saccharomyces yeasts are generally associated with and grapes, and many different genera can be found in these environments, including basidiomycetous yeasts such as Cryptococcus spp. and Rhodotorula spp., as well as ascomycetous yeasts such as Hanseniaspora spp., spp.,

Metschnikowia spp., spp., Torulaspora spp., and many others (Bisson and

Joseph, 2009; Jolly et al., 2014). The composition of non-Saccharomyces yeasts on grape surfaces varies considerably from region to region around the globe (Barata et al., 2012; Gayevskiy and Goddard, 2012; Ocón et al., 2010). Depending on the health and maturity of the grape, different yeasts will dominate the grape berry.

Damaged or broken berries invite the growth of more strongly fermentative yeasts such as Torulaspora spp. and even Saccharomyces spp., as well as yeasts associated with wine spoilage, such as bruxellensis (Barata et al.,

2012; Mortimer and Polsinelli, 1999). Basidiomycetous yeasts thrive in the nutrient- poor environments of immature grape berries, but as the berries ripen these yeasts are succeeded by ascomycetous yeasts, which are the dominant yeasts in freshly pressed grape juice (must) at the onset of AF (Barata et al., 2012; Bisson and

Joseph, 2009). Climate, soil conditions, vineyard management, interactions with other organisms, and nutrient availability all significantly impact the microbial

6 communities on grape surfaces, which in turn will affect which organisms are present in the grape must (Barata et al., 2012; Gayevskiy and Goddard, 2012).

Non-Saccharomyces yeasts often begin the process of AF, but are usually replaced by strains of S. cerevisiae within three to four days (Ciani et al., 2010); S. cerevisiae then completes AF. This succession of yeast species and strains is partly due to the fact that in general, non-Saccharomyces yeasts are poor fermenters and intolerant to ethanol concentrations above 5-7% (v/v), while S. cerevisiae is a fast- growing and efficient fermenter, and can tolerate ethanol up to 16-18% (v/v)

(Domizio et al., 2007; Fleet, 2003).

Non-Saccharomyces yeasts were traditionally viewed as spoilage microorganisms, so their rapid removal from the fermenting must was desirable.

However, there is wide variation amongst non-Saccharomyces yeasts in terms of fermentative capabilities and the production of flavour- and aroma-active secondary metabolites, and many winemakers are now turning to these yeasts to improve the complexity of their wines (Fleet, 2003; Jolly et al., 2014; Liu et al., 2016; Romano et al., 2003). Some species of non-Saccharomyces yeasts, including Torulaspora delbrueckii and S. uvarum, have shown great fermentative potential, and are able to complete AF while producing desirable secondary metabolites that can improve the sensory profile of wine (Azzolini et al., 2015, 2012; Bely et al., 2008; Contreras et al.,

2015; Demuyter et al., 2004).

1.2.3 Spoilage yeasts

Some yeasts are almost always undesirable in winemaking, and are therefore considered spoilage yeasts. The most famous and most destructive spoilage yeast

7 involved in winemaking is Brettanomyces bruxellensis (teleomorph Dekkera bruxellensis, commonly known as “brett”). B. bruxellensis produces compounds such as 4-ethylphenol, 4-vinylphenol, and 4-ethyl guaiacol, whose smells include barnyard, horse sweat, band-aid, and smoke (Romano et al., 2009). B. bruxellensis can survive in wine in a viable but non culturable (VBNC) state, allowing it to evade detection by culture-dependent methods (Serpaggi et al., 2012). The second-most dangerous spoilage yeast is bailii (Loureiro and Malfeito-

Ferreira, 2003), which is very resistant to most forms of preservatives, including

SO2, acetic acid, and ethanol, and is a widespread spoilage yeast in the food industry (Kuanyshev et al., 2017). Z. bailii can result in cloudiness and sediment formation as well as the production of an unpleasant yeasty odour in wine (Erickson and McKenna, 2000). Other spoilage yeasts, including Schizosaccharomyces pombe and Saccharomycodes ludwigii, are only rarely isolated from contaminated wines (Loureiro and Malfeito-Ferreira, 2003). Some non-Saccharomyces yeasts can become spoilage yeasts if left unchecked. An overgrowth of Hanseniaspora spp. can lead to stuck fermentations and the production of high concentrations of acetic acid

(vinegar) and (nail polish/solvent aroma), and other species, including

Pichia spp. and Metschnikowia spp., can become film-forming yeasts under certain conditions (Loureiro and Malfeito-Ferreira, 2003). Finally, even S. cerevisiae can sometimes be considered a spoilage yeast, as its presence can lead to the re- fermentation of wine in the bottle when residual sugar is present (Divol et al., 2006).

8

1.3 Bacteria involved in winemaking

Bacteria are present in all wine fermentations, but are particularly important for the production of red wines and some white wines that undergo (MLF). Many species and strains of bacteria co-exist and interact with yeasts during the winemaking process (Barata et al., 2012; Fleet, 2003), and the most important wine bacteria can be grouped into two main categories: lactic acid bacteria and acetic acid bacteria.

1.3.1 Lactic acid bacteria

Lactic acid bacteria (LAB) are the most important bacteria involved in fermentations, because they conduct MLF, the conversion of the tart-tasting L-malic acid to the softer L-lactic acid, and are associated with relatively high expressions of malate-specific (Alexandre et al., 2004; Edwards and Jensen, 1992). MLF is desirable for almost all red wines and some white wines, especially Chardonnay, as it can enhance the wine’s sensory profile and improve its microbial stability

(Izquierdo Cañas et al., 2012; Lonvaud-Funel, 1999; Sun et al., 2013). MLF is usually accompanied by low-level citric acid fermentation, which produces acetic acid and , and can give the wine a buttery character. Oenococcus oeni

(formerly Leuconostoc oenos) is the most important LAB, and conducts the bulk of

MLF. Other enologically important LAB include species of Lactobacillus,

Leuconostoc, and Pediococcus (Lonvaud-Funel, 1999). LAB, particularly O. oeni, are generally winery-resident bacteria, and are rarely found on grape surfaces or in the vineyard (Barata et al., 2012). Environmental bacteria, including species of

Enterococcus, Enterobacter, Bacillus, and Serratia, are more commonly associated

9 with vineyard environments but are not active participants during fermentation

(Barata et al., 2012; Godálová et al., 2016).

MLF usually takes place after AF has completed, because yeast activity, SO2 levels, and ethanol concentrations tend to inhibit the growth and activity of LAB during AF (Arnink and Henick-Kling, 2005; Fang and Dalmasso, 1993). Once AF is finished, most wine yeasts have perished, and the lack of competition in the fermentation vessel allows LAB to become more active and commence MLF. MLF may be conducted spontaneously (uninoculated) by winery-resident strains of O. oeni, or commercial strains may be intentionally inoculated into the wine. Inoculated

MLF involves either adding O. oeni strains to the wine following the completion of AF

(termed sequential inoculation) or they can be added prior to the start of AF, usually approximately 24 h after yeast inoculation (termed co-inoculation) (Abrahamse and

Bartowsky, 2012; Antalick et al., 2013).

1.3.2 Acetic acid bacteria

Acetic acid bacteria (AAB) are spoilage microorganisms and are undesirable in fermentations. If allowed to grow, they can produce high concentrations of acetic acid (the largest component of volatile acidity in wine), which gives the wine a sour, vinegary flavour above 1.2-1.4 g/L (Drysdale and Fleet, 1988). While some acetic acid is produced by almost all wine microorganisms, AAB produce significantly larger amounts. The main AAB in winemaking are species of Acetobacter and

Gluconobacter, which can be found both in the vineyard as well as in the winery

(Barata et al., 2012; Bokulich et al., 2013; Godálová et al., 2016). However, AAB are

10 aerobic bacteria, and if wine is produced and stored under anaerobic conditions, they are unable to grow or contaminate the wine.

1.4 Microbial interactions during fermentation

During fermentation, wine microorganisms compete with each other for space and nutrients. However, they can also interact in positive ways that can improve wine quality and lead to the production of wines with unique sensory profiles.

Because wine yeasts are the most important microorganisms involved in wine production, I will outline the ways in which wine yeasts interact with each other, as well as ways they interact with bacteria. Filamentous fungi may also be present in grape must (Fleet, 2003), but they do not play an important role in fermentation, and as such their interactions with enologically important microorganisms will not be discussed.

1.4.1 Yeast-yeast interactions

Generally, a higher diversity of wine yeasts is observed at the onset of AF, before S. cerevisiae dominates the fermentation. The early stages of AF are characterized by a succession of yeast species and strains, during which time many different yeasts are able to interact with one another. Eventually, dominant yeasts such as S. cerevisiae will overtake the fermentation, and strains of S. cerevisiae are often the only yeasts identified at the end of AF. S. cerevisiae grows quickly in grape must, reducing the nutrients available for other yeasts. S. cerevisiae also produces many compounds that interact negatively with other yeasts, including ethanol, short chain fatty acids (SCFAs), killer toxins, and other inhibitory peptides (Bisson, 1999;

Fleet, 2003; Schmitt and Breinig, 2006). The process of AF also produces CO2,

11 which creates an anaerobic environment in which many microorganisms cannot survive. The production of ethanol by yeasts during AF is the main mechanism by which the growth of other yeasts is inhibited. The ability of certain yeasts to produce killer toxins is determined by the possession of cytoplasmic double-stranded DNA viruses (Schmitt and Breinig, 2006). Killer toxins cause cell death in susceptible yeasts through various mechanisms, depending on the specific virus encoding the toxin, and killer activity has been identified in many species of wine yeasts, including

S. cerevisiae, Z. bailii, H. uvarum, T. delbrueckii, and Lachancea thermotolerans

(formerly Kluyveromyces thermotolerans) (Heard and Fleet, 1987; Schmitt and

Breinig, 2006, 2002). S. cerevisiae has also been shown to inhibit the growth of non-

Saccharomyces yeasts by a cell-cell contact mechanism as well as the production of antimicrobial peptides (Albergaria and Arneborg, 2016; Kemsawasd et al., 2015;

Nissen et al., 2003).

Yeasts may also interact in a positive manner during fermentation, although positive interaction mechanisms are less common. Yeast biomass is produced throughout fermentation as a result of yeast , and can provide amino acids, , and other nutrients for other yeasts to utilize (Alexandre and Guilloux-

Benatier, 2006). The hulls of dead yeasts are also able to adsorb compounds that may be toxic to living cells, such as metal ions, volatile , and grape phenolic compounds (Fleet, 2003; Lavigne-Cruège and Dubourdieu, 1996; Vasserot et al.,

1997). Furthermore, species of non-Saccharomyces yeasts that produce proteolytic and pectolytic enzymes can release nutrients that other yeasts may utilize

(Charoenchai et al., 1997).

12

1.4.2 Yeast-bacteria interactions

The interactions between wine yeasts and bacteria during fermentation may be antagonistic or complementary. Many studies have reported the abilities of some

S. cerevisiae strains to inhibit LAB, including O. oeni (Alexandre et al., 2004 and references therein). This antagonism is likely a result of many factors, including competition for nutrients, the ethanol- and SO2-sensitivity of LAB, and the production of proteinaceous compounds and medium-chain fatty acids (MCFAs) by yeasts

(Alexandre et al., 2004; Comitini et al., 2005; Reguant et al., 2005). These interactions are often strain-specific, however, and many S. cerevisiae strains are considered to be beneficial for LAB when MLF is desired. Yeasts may release mannoproteins, which can lead to an increase in available nutrient content for LAB as well as bind MCFAs and prevent their inhibitory activities (Diez et al., 2010).

Furthermore, yeast autolysis towards the end of AF may provide nutrients to LAB and increase the availability of nitrogen and nucleic acids in the wine (Alexandre et al., 2004; Balmaseda et al., 2018; Fornachon, 1968).

Non-Saccharomyces yeasts also interact with bacteria during fermentation, and the presence of non-Saccharomyces yeasts may actually be beneficial to LAB: fermentations containing more non-Saccharomyces yeasts usually have lower levels of compounds that inhibit LAB, such as ethanol and MCFAs, and higher levels of compounds that favour MLF, including citric acid and (Balmaseda et al.,

2018; Nardi et al., 2019). Interactions between yeasts and bacteria during fermentation can also significantly alter the sensory profile of wine (Fleet, 2003;

Knoll et al., 2012; Rossouw et al., 2012).

13

1.5 Inoculation techniques

There are many techniques winemakers use to help ensure the completion of

AF while introducing uniqueness into their wines. Inoculated fermentations, where the winemaker inoculates the grape must with one or more commercial ADY strains, are popular at large commercial wineries where winemakers are looking to speed up fermentation, reduce the risk of stuck or sluggish fermentations, and produce a consistent product from year to year (Alexandre and Charpentier, 1998; Heard and

Fleet, 1985). It is assumed that the commercial ADY strain(s) added to the must will take control and dominate the fermentation. However, this is not always the case, and it has been observed that aggressive commercial strains used in previous vintages may establish themselves as winery residents and out-compete the inoculated strains (Barrajón et al., 2009; Lange et al., 2014).

Uninoculated, or spontaneous, fermentations are the of winemaking, where the must is allowed to ferment naturally, with no intentional addition of selected commercial strains. Although this technique carries the risk of spoilage or stuck fermentations (Constantí et al., 1997), some winemakers are reverting back to this process, as it can produce wines with a higher aroma and flavour complexity (Egli et al., 1998; Medina et al., 2013; Vilanova and Sieiro, 2006).

Uninoculated fermentations tend to allow for a higher diversity of yeast species and strains to participate in AF than do inoculated fermentations (Csoma et al., 2010), and can result in the production a wider range of aromatic metabolites (Fleet, 2003;

Romano et al., 2003). Many assume that the strains of S. cerevisiae participating in these uninoculated fermentations are of indigenous origin. While this seems to be

14 the case in some winemaking regions (Gayevskiy and Goddard, 2012; Santamaría et al., 2008), in many others, including the Okanagan Valley, uninoculated fermentations are often dominated by commercial strains used previously at their respective wineries (Blanco et al., 2011; Clavijo et al., 2011; Constantí et al., 1997;

Hall et al., 2011; Scholl et al., 2016). Nevertheless, even in these regions, indigenous strains of S. cerevisiae may have more of an advantage in uninoculated fermentations than in inoculated fermentations, since no aggressive commercial strain has been directly added to compete with them.

In recent years, consumer demand for new wine products has led to the development of new winemaking techniques, including the increasingly common practices of sequential and co-inoculations with multiple yeast strains to produce new flavour and aroma profiles (Barrajón et al., 2011; Bely et al., 2008; Benito et al.,

2014; Ciani et al., 2006; Gobbi et al., 2013; King et al., 2010). Sequential inoculation usually involves the initial inoculation of a non-Saccharomyces yeast with stronger fermentative capabilities, such as T. delbrueckii or L. thermotolerans, followed by the inoculation of a strain of S. cerevisiae one or more days later (Gobbi et al., 2013;

Izquierdo Cañas et al., 2011; Medina et al., 2013; Nardi et al., 2019; Sadoudi et al.,

2012). This technique allows the non-Saccharomyces yeast a longer period of time to grow, begin fermentation, and contribute the wine’s sensory profile before being out-competed by S. cerevisiae. Co-inoculations usually involve the inoculation of multiple strains of S. cerevisiae at the same time (Barrajón et al., 2011; King et al.,

2010, 2008), but can occasionally involve the concurrent inoculation of S. cerevisiae with a non-Saccharomyces yeast (Cheraiti et al., 2005; Gobbi et al., 2013; Kim et al.,

15

2008). When MLF is desired, O. oeni strains can be inoculated post-AF, as is traditional, or can be co-inoculated with S. cerevisiae at the beginning of fermentation (Antalick et al., 2013; Nehme et al., 2010; Rossouw et al., 2012).

Sequential and co-inoculations with non-Saccharomyces yeasts, with multiple S. cerevisiae strains, or even with bacterial strains, can all result in wines with significantly different chemical and sensory profiles as compared to wines that use only one S. cerevisiae strain as the inoculant (Antalick et al., 2013; Comitini et al.,

2011; Gustafsson et al., 2016; King et al., 2010, 2008). Sequential and co- inoculation are popular amongst winemakers looking to enhance the complexity of their wines while still allowing for greater control over the progression of fermentation.

Another alternative fermentation technique takes the contribution of vineyard- derived non-Saccharomyces yeasts to a new level. Pied de cuve inoculation, a

French term that translates to ‘foot of the tank,’ is a seeding procedure involving any indirect yeast inoculation using already-fermenting must. This was the original method of inoculation, where yeasts from the bottom of a fermenting tank would be used to inoculate freshly-pressed grape must in order to “kick-start” AF. Winemakers adopting this method today choose one of two paths: either they crush a small portion of grapes and allow this must to begin ‘spontaneous’ fermentation in the vineyard, away from the influence of winery-resident yeast strains, or they use winery fermentations as their inoculum. In either case, once the pied de cuve must has begun fermentation, it can be used to inoculate new tanks or barrels. Research into pied de cuve fermentations is extremely scarce (Clavijo et al., 2011b; Li et al.,

16

2012; Marti-Raga et al., 2015; Ubeda Iranzo et al., 2000), and the term pied de cuve is used to describe a variety of inoculation techniques. Often pied de cuve refers to a fermentation originally inoculated with a commercial ADY strain that is then used as the inoculum for a second fermentation (Li et al., 2012; Ubeda Iranzo et al., 2000).

Two previous studies make reference to pied de cuve fermentations where an uninoculated fermentation carried out in the winery is used as the inoculum for future fermentations (Clavijo et al., 2011b; Piao et al., 2015). Although one study has been conducted on fermentations carried out in the vineyard (Sturm et al., 2006), to my knowledge no studies have conducted pied de cuve fermentations where the inoculum began fermentation spontaneously in the vineyard but was subsequently transferred to the winery. This leaves a large gap in knowledge in this area of research, and considerable uncertainty to winemakers looking to enhance the contribution of their local vineyard yeasts in their fermentations.

1.6 Sulfur dioxide in winemaking

Many microorganisms associated with wine spoilage are brought into the winery with the grapes. One method of controlling the growth of undesirable microbes during and after fermentation is the addition of sulfur dioxide, which has been a staple additive in winemaking since the nineteenth century (Bush et al.,

1986). Today, SO2 is usually added as potassium metabisulfite (K2S2O5), abbreviated as KMS, and some form of SO2 is used in almost all commercially- produced wines. It is worth noting that a small amount of SO2 is produced by yeasts, so no fermentation is ever truly sulfite-free. SO2 assists winemakers in a variety of ways, acting as both an antimicrobial agent and antioxidant, controlling the growth of

17 unwanted microorganisms as well as preventing browning and other oxidation reactions. When SO2 is added at crush, the main yeasts present in the grape must are the non-Saccharomyces yeasts brought in from the vineyard. These yeasts are therefore more strongly affected by SO2 addition than S. cerevisiae, which is usually not present in must at this early stage of winemaking (Constantí et al., 1998; Henick-

Kling et al., 1998; Zott et al., 2008). As a result, there has been much research into the effects of this initial SO2 addition on the total yeast community as well as its effects on spoilage microorganisms (Bokulich et al., 2014; Constantí et al., 1998;

Henick-Kling et al., 1998; Takahashi et al., 2014). However, little thought has been given to whether this addition may directly or indirectly affect the abundance, diversity, and composition of S. cerevisiae strains (Constantí et al., 1998; Egli et al.,

1998), and thus further research is need on this topic.

Sulfur dioxide has three main species that are in equilibrium with each other

(Equation 1):

- 2- SO2 ↔ HSO3 ↔ SO3 [Equation 1]

- 2- where SO2 refers to molecular SO2, HSO3 refers to bisulfite, and SO3 refers to

sulfite (pK1 = 1.86; pK2 = 7.20) (Neta and Huie, 1985).

- At a wine pH of 3-4, sulfur dioxide exists mostly as HSO3 , but some is in the form of molecular SO2. In this thesis, unless explicitly stated, the term ‘SO2’ will refer to all forms of sulfur dioxide present in grape must. Sulfur dioxide is also present in grape must in free and bound forms. The bisulfite and sulfite ions readily interact with many components of must, which may be of grape, yeast, or bacterial origin

(Divol et al., 2012). In particular, SO2 binds strongly to acetaldehyde, the final

18 electron acceptor in AF, present in significant concentrations during fermentation

(Jackowetz and Mira de Orduña, 2012; Maier et al., 1986). SO2 that is bound to these must constituents is referred to as ‘bound’ sulfur dioxide, and SO2 that is present in the must but not bound is referred to as ‘free’ sulfur dioxide. Only the free form of SO2 is active as an antioxidant and antimicrobial agent. Furthermore, although both molecular SO2 and bisulfite are able to inhibit the growth of microorganisms, molecular SO2 is 100-500 times more effective than bisulfite, because it has no charge and is able to pass freely through the cell membrane

(Divol et al., 2012).

Once inside the cell, the higher internal pH of the cell converts the molecular

SO2 to bisulfite, as well as a small amount of sulfite. This conversion reduces the intracellular concentration of molecular SO2, facilitating more uptake. Bisulfite can also be taken into the cell though a carrier-mediated proton symport (Park and

Bakalinsky, 2004). Sulfur dioxide uptake into yeast and bacterial cells depends on the external pH and temperature of the must: lower pH and higher temperature are the favoured conditions for microbial inhibition (Divol et al., 2012; Egli et al., 1998).

An immediate repercussion of SO2 entry into the cell is the inhibition of glyceraldehyde-3- dehydrogenase, which results in an instant arrest in (Hinze and Holzer, 1986; Maier et al., 1986). SO2 can also bind to and inhibit , co-enzymes, co-factors, and various metabolites, as well as cause

DNA point mutations, all of which drastically reduce cellular function and can result in cell death (Divol et al., 2012).

19

Because sulfites became commonly used in the wine industry over two centuries ago, wine yeasts have since evolved various mechanisms of sulfite resistance. As a general rule, bacteria are the most sensitive to SO2, and non-

Saccharomyces yeasts tend to be more susceptible than S. cerevisiae, although strain variation does occur (Constantí et al., 1998; Divol et al., 2012; Henick-Kling et al., 1998). Wine yeasts have evolved four main mechanisms of evading the effects of SO2: producing acetaldehyde; reducing SO2 via incorporation into the sulfur biosynthesis (SAAB) pathway; expelling SO2 from the cell via specialized sulfite-efflux pumps; and entering a VBNC state (Divol et al., 2012). Acetaldehyde production is the most effective means of binding SO2 in wine and grape juice, as one mole of acetaldehyde can effectively bind one mole of SO2, preventing it from acting as an antimicrobial agent (Maier et al., 1986). However, high concentrations of acetaldehyde can produce wines with a bruised-apple or aroma, and have been linked to alcohol-induced headaches and alcohol sensitivity (Swiegers et al.,

2005; Wall et al., 1997). Another mechanism by which yeasts evade the effects of

SO2 is via the SAAB pathway. Yeast cells manufacture sulfur-containing amino acids, and bisulfite is one of the intermediates in the SAAB pathway. When molecular SO2 enters the cell and is converted to bisulfite, it can then be incorporated into this pathway and reduced to form S2-. This sulfite is then either used to produce sulfur-containing amino acids, or is exported from the cell as hydrogen sulfide (H2S). The amount of H2S produced is linked to the levels of nitrogen and sulfur-containing amino acids already available to the yeast; low nitrogen availability will prevent yeasts from producing sulfur containing amino acids,

20 and the presence of sulfur-containing amino acids in the environment reduces the need for yeasts to produce their own, both resulting in increased H2S production

(Aranda et al., 2006). The production of H2S by yeasts is undesirable, because it has a low sensory detection threshold and can result in ‘reduced’ wines that express rotten egg or cooked cabbage characters (Siebert et al., 2009). S. cerevisiae also possesses the transmembrane Ssu1p, encoded by the gene SSU1 and regulated by the FZF1 transcription factor (Avram et al., 1999; Park and Bakalinsky,

2000). Ssu1p functions as a specialized sulfite efflux pump, and can export SO2 back outside the yeast cell before it is able to cause any damage. This response is desirable to winemakers, because it allows SO2 to continue to act as an antimicrobial agent on more sensitive microorganisms and does not contribute to the production of off-flavours or aromas. Finally, many yeasts and bacteria will enter a

VBNC state as a final defense when exposed to high concentrations of SO2

(Agnolucci et al., 2013, 2010; Divol and Lonvaud-Funel, 2005). Unlike dormant cells,

VBNC cells are still metabolically active, and can still impact wine quality.

Sulfur dioxide addition is widespread in the food and beverage industry, and at low concentrations it is generally harmless to most individuals (Vally et al., 2009).

However, some people with asthma are highly sensitive to sulfites in foods and beverages, and even healthy individuals will experience airway constriction when inhaling SO2 in sufficient concentrations (Freedman, 1980; Yang and Purchase,

1985). Inhalation and ingestion of SO2 can cause symptoms ranging from bronchospasms, coughing, wheezing, and hives, to anaphylactic shock, depending on the sensitivity of the individual and the concentration of sulfites (Freedman, 1980;

21

Gershwin et al., 1985; Vally et al., 2009). In 1985 it was estimated that up to 5-11% of North Americans may be sensitive to metabisulfites in foods and beverages, one third of whom are not diagnosed asthmatics (Yang and Purchase, 1985). A study conducted in 2000 stated that asthma attacks triggered by wine consumption is self- reported by up to 30% of asthmatic patients (Vally et al., 2000); however, a follow-up study by the same team of researchers suggests that this value may be significantly overestimated (Vally and Thompson, 2001). In Canada, the legal limit of SO2 in bottled wine is 70 mg/L free SO2, and 350 mg/L total SO2. These limits have not been changed for over thirty years (Yang and Purchase, 1985). Sulfites are currently added to almost all commercial wines, especially at the bottling stage, because it prevents the growth of spoilage microbes during and after fermentation and ensures that wines sold in the marketplace maintain desired flavours. Despite increasing research into alternatives for SO2 during winemaking, including the use of UV irradiation, colloidal silver complex, and , SO2 remains the most popular and most effective preservative (Delfini et al., 2002; Falguera et al., 2013;

Fracassetti et al., 2015; Fredericks et al., 2011; Guerrero and Cantos-Villar, 2015;

Izquierdo-Canas et al., 2012). However, increasing our knowledge about the effects of SO2, both on fermentative microorganisms and on human health, will ensure that the addition of SO2 is kept to a minimum while maintaining wine quality.

1.7 Wine sensory profiles

Wine aroma and flavour depend on many factors, including grape , , fermentation vessel, and microbial interactions (Deed et al., 2017; Fleet,

2003; Reynolds et al., 2013; Schlosser et al., 2005). Terroir is a French term, and

22 originally referred to the idea that the soil in which the grapes were grown was the defining characteristic of the wines they produced. Today, terroir encompasses much more than simply the soil, and it is known that the climate, vineyard aspect, local microbes, and viticultural and winemaking practices all contribute significantly to the aroma and flavour of wine. Over 1000 volatile compounds have been discovered during alcoholic fermentation; at least 400 of these are products of yeast metabolism, the rest being largely due to varietal aroma compounds originating from the grapes themselves or compounds modified and extracted by yeasts and bacteria during fermentation (Romano et al., 2003). Some commonly produced compounds, along with their detection thresholds, can be found in Table 1.1.

Table 1.1 Selected compounds produced by wine yeasts and bacteria.

Compound Sensory attribute(s) Detection threshold Acetaldehyde Bruised apple/sherry 0.100 g/La Acetic acid Vinegar 0.28 g/La Buttery 0.15 g/Lb Diacetyl Buttery 0.0002-0.0028 g/Lc Ethyl acetate Fruity/solvent 0.0075 g/La Sweet/viscous 5.2 g/La Isoamyl alcohol Nail polish/fusel 0.03 g/La Isobutanol Fruity/fusel 0.04 g/La Lactic acid Creamy 1.12 g/Ld Malic acid Tart/green apple 0.85 g/Ld 4-ethyl guiacol Smoke/phenolic 0.00011 g/Lb 4-ethyl phenol Barnyard/medicinal 0.0006 g/Lb 4-vinyl phenol Medicinal/band-aid 0.00002 g/Lb a (Swiegers et al., 2005) b (Bartowsky and Pretorius, 2009) c (Martineau et al., 1995) d (Francesco et al., 2017)

23

Since S. cerevisiae is usually the dominant yeast during alcoholic fermentation, it makes the greatest contributions to the development of wine sensory attributes. However, many other yeast species also contribute to the development of wine flavour and aroma, including H. uvarum, M. pulcherrima, I. orientalis, L. thermotolerans, T. delbrueckii, and S. uvarum, among others (Fleet, 2003; Izquierdo

Cañas et al., 2011; Jolly et al., 2014; Kim et al., 2008; Romano et al., 2003; Tosi et al., 2009; Varela et al., 2017). The contributions of non-Saccharomyces yeasts are usually made at the beginning of AF, before they are out-competed and replaced by

S. cerevisiae. The compounds produced by these yeasts may positively or negatively influence wine quality, depending on the compound produced as well as its concentration.

H. uvarum has been known to produce high levels of both acetoin and ethyl acetate in wine (Herraiz et al., 1990; Romano et al., 2003). Ethyl acetate is a commonly-produced compound that contributes to volatile acidity, and at low concentrations may lend a sweet, fruity aroma to the wine. However, at higher concentrations, the aroma becomes unfavourable, adopting a nail polish or vinegar smell. Acetoin, along with diacetyl, produces a creamy, buttery flavour in wines, which can be pleasant at low concentrations, but at high concentrations is undesirable. Different strains of H. uvarum also produce varying amounts of acetic acid, an undesirable compound in wine that is the major contributor of volatile acidity

(VA) and produces a vinegar aroma (Romano et al., 2003).

Other non-traditional yeasts can have different effects on the wine profile.

Fermentations conducted by S. uvarum tend to have less alcohol, acetic acid, and

24 acetaldehyde, as well as more glycerol, malic acid, and ethyl acetate than fermentations conducted by S. cerevisiae (Castellari et al., 1994; Gamero et al.,

2013; Magyar and Tóth, 2011; Sipiczki et al., 2001). Acetic acid and acetaldehyde are considered undesirable in high concentrations. Glycerol can increase the sweetness and viscosity of wine, and malic acid lends a tart, green apple flavour, which can be desirable in some wines. Due to these qualities, S. uvarum shows promise for use in wines where lower alcohol content is desired, either through single inoculation or in co-inoculations with S. cerevisiae. Other non-Saccharomyces yeasts, such as T. delbrueckii, L. thermotolerans, and M. pulcherrima, are capable fermenters and have also been used in co-fermentations with S. cerevisiae in order to lower the alcohol content and improve the complexity of wines (Ciani and

Comitini, 2011; Loira et al., 2014; Varela et al., 2017).

Production of flavour- and aroma-active compounds, including higher alcohols, organic acids, and , varies between yeast species, and even between strains of the same species. Different yeast strains may produce different compounds, or, more commonly, different concentrations of common compounds

(Barrajón et al., 2011; Fleet, 2003; Herraiz et al., 1990). The strongest differentiators for S. cerevisiae have been found to be isobutanol, isoamyl alcohol, acetaldehyde, and acetic acid production (Romano et al., 2003). Spoilage microorganisms, including B. bruxellensis, are of particular concern to winemakers, as they produce undesirable volatile compounds that can result in wine faults. B. bruxellensis produces 4-ethylphenol, 4-vinylphenol, and 4-ethylguiacol, which impart off-flavours and aromas described as medicinal, barnyard, and animal, and smoke.

25

Bacteria also play a role in the development of wine aroma and flavour. The main contribution of bacteria is during MLF, where LAB convert the tart, green-apple taste of malic acid to the softer and smoother lactic acid. Citric acid fermentation is also performed by most LAB, which can produce diacetyl and can lend a buttery aroma and flavour to the wine. AAB produce high concentrations of acetic acid, and can raise the volatile acidity of a wine above acceptable levels if left unchecked.

The addition of SO2 can also alter the sensory characteristics of wine. Wine colour is affected by SO2 addition, and its antioxidant activity can prevent browning reactions in white wines (Coetzee et al., 2013). Additionally, SO2 addition can protect phenolic compounds and volatile thiols from oxidation, both of which contribute to a wine’s sensory profile (Coetzee et al., 2013; Ivanova et al., 2011). If

SO2 is added in high enough concentrations, it may itself lend an unpleasant odour to wines, which is often described as the smell of a freshly-lit match.

Many factors influence the final sensory profile of a wine. Grape varietal is said to be the main determining factor, but yeast/bacterial inoculation, fermentation vessel, microbial interactions during fermentation, and natural environmental factors that contribute to the concept of terroir all play significant roles in determining the way a wine smells and tastes. It is the culmination of all of these factors that give each wine its unique sensory attributes, and why wines from every corner of the globe continue to be enjoyed by millions each day.

26

1.8 Methods of microbial identification

The microbial communities and populations present in wine fermentations can be identified in many different ways. Broadly, the two main methods of microbial identification rely on either culture-dependent or culture-independent techniques.

1.8.1 Culture-dependent identification

Culture-dependent identification techniques were originally the only means available for observing and identifying the microorganisms present in must and wine.

The original method of microbial identification involved plating must or wine samples onto differential and selective media, and used colony morphology and metabolic activity to distinguish between species of yeasts or bacteria (Bisson and Joseph,

2009). Today, these methods have been eclipsed by molecular techniques such as polymerase chain reaction (PCR) paired with Sanger-sequencing or restriction fragment length polymorphism (RFLP) analysis. Fungi, including yeasts, can be identified to the species level by amplifying portions of the internal transcribed spacer (ITS) or the D1/D2 domains of the rRNA gene. (Barata et al., 2012; Constantí et al., 1998; Gayevskiy and Goddard, 2012; Nisiotou and Nychas, 2007).

Because many different strains of S. cerevisiae can be involved in a single fermentation, identifying S. cerevisiae isolates to the strain level is often desired.

Many strain-typing techniques are currently available, mostly PCR-based methods such as random amplified polymorphic DNA (RAPD) PCR, mitochondrial DNA

(mtDNA) restriction analysis, amplified fragment length polymorphism (AFLP) analysis, Fourier-transform infrared spectroscopy (FT-IR), single nucleotide polymorphism (SNP) analysis, and profiling of microsatellite DNA (Borneman et al.,

27

2016; Erkus et al., 2016; Fernández-Espinar et al., 2003; Fleet, 2007; Javier Gallego et al., 2005; Kümmerle et al., 1998). Microsatellites are also called simple sequence repeats (SSR) or short tandem repeats (STR), and are characterized by short, repeating DNA segments of between two and six base pairs. Microsatellites are typically found in non-coding regions of the , and are therefore subject to high rates of mutations. This leads to hypervariability within many microsatellite loci, which can be used to identify yeast isolates to the strain level (Field and Wills, 1998;

Legras et al., 2005; Pérez et al., 2001; Richards et al., 2009). Although microsatellite analysis is most commonly used to strain-type S. cerevisiae in wine (Barrajón et al.,

2011; Scholl et al., 2016; Schuller et al., 2012; Sturm et al., 2006; Zhang et al.,

2010), they have also been used in the field of wine research to strain-type other relevant yeast species, including L. thermotolerans and S. uvarum (Hranilovic et al.,

2017; Masneuf-Pomarede et al., 2016, 2007; Zhang et al., 2015).

While these molecular techniques are much more accurate than colony morphology and metabolic tests, they still carry the same biases of all culture- dependent methods in that they are limited by the kinds of microorganisms they can detect. Many microorganisms will be unintentionally excluded from this type of analysis, due to a number of factors. Media composition and incubation conditions are necessarily ideal for only a small subset of the microbial community. Many microorganisms are present in very low abundance, and due to the serial dilution required to obtain individual colonies for identification, these low-abundance microbes are unlikely to be isolated and identified. Finally, some microorganisms that can significantly influence the quality of wine are often present in a VBNC state

28 towards the end of fermentation, and cannot be detected by culture-based methods.

However, these microbes are still metabolically active and can lead to wine spoilage if left undetected.

1.8.2 Culture-independent identification

The biases of culture-dependent identification techniques outlined above can lead to severely underestimating the microbial diversity present in wine fermentations. Therefore, when the identification of the entire microbial consortium of a wine environment is desired, culture-independent identification methods can provide more accurate results. Culture-independent techniques include PCR- denaturing gradient gel electrophoresis (DGGE), quantitative PCR (qPCR), and high-throughput amplicon sequencing techniques such as Roche454, Illumina, Ion

Torrent, and PacBio. Many of these techniques are currently popular among wine researchers that conduct molecular analyses, because they provide a broader view of the microbial community dynamics during fermentation (Bokulich et al., 2014;

Bokulich and Mills, 2012; Hierro et al., 2006; Prakitchaiwattana et al., 2004; Wang et al., 2015a; Zarraonaindia et al., 2015). Culture-independent methods also have an advantage over culture-dependent methods in that they allow samples to be taken directly from grape, must, or wine, and can identify organisms in a VBNC state.

Despite these advantages, culture-independent techniques have a number of drawbacks, and it may be useful to wine researchers to employ both culture- dependent and culture-independent techniques when analyzing wine microbial community dynamics. Methods such as qPCR and PCR-DGGE can only identify organisms that are above a certain numerical threshold in a sample, and may miss

29 identifying microbes that are present in very low numbers (Barata et al., 2012). The qPCR method also requires primers that are universal, or specific to certain genera or species, but some of these primers have not yet been developed, and if this method is used exclusively to identify the microbial community, some organisms will be missed simply because they are not being targeted (Bisson and Joseph, 2009).

The closed-system conditions of wine fermentations create increasingly inhospitable environments for microorganisms, as ethanol accumulates and nutrient supplies deplete. By the end of AF, a significant percentage of yeast and bacteria are no longer living, but their cells and DNA remain in the wine (Branco et al., 2012). Many culture-independent techniques are unable to distinguish DNA from living and dead cells, and as a result these techniques may overestimate the contribution of some microbes at certain stages of fermentation. In order to prevent these dead organisms from being identified, DNA-binding dyes such as propidium monoazide

(PMA) can be added to samples prior to DNA extraction (Andorrà et al., 2010;

Tantikachornkiat et al., 2016). PMA binds irreversibly to the exposed DNA of dead cells when exposed to light, and prevents this DNA from amplifying during PCR.

Finally, culture-independent techniques do not involve the isolation of individual yeast or bacterial colonies, and thus further investigation of specific microbes of interest is not an option with these techniques. High-throughput amplicon sequencing is a useful tool for the characterization of whole microbial communities, but as of now it is unable to identify microorganisms to the strain level.

Due to these limitations, it is currently useful to use culture-dependent techniques in combination with culture-independent techniques in order to gain a better

30 understanding of the microbial community and population dynamics at play during wine production.

1.9 Objectives and predictions

The goal of this thesis was to investigate the effects of SO2 addition and alternative fermentation techniques on the microbial communities, yeast populations, and sensory profiles of wines. This was achieved via five main research objectives, each discussed in separate research chapters in this thesis. These objectives are outlined below, along with specific predictions and the current knowledge gaps in the field.

1.9.1 The effect of sulfur dioxide addition at crush on Saccharomyces cerevisiae strain composition during uninoculated fermentations

SO2 is commonly added at crush to remove potential spoilage microbes entering the must from the vineyard and to help promote the dominance of S. cerevisiae yeasts. While most wine SO2 research has focused on the effects of SO2 on spoilage microbes, few studies have been conducted investigating its potential to disrupt the S. cerevisiae populations involved in fermentations (Constantí et al.,

1998; Egli et al., 1998; Henick-Kling et al., 1998; Suzzi and Romano, 1982).

Furthermore, most of the studies that have been conducted on this topic either did not include replication or failed to directly compare the populations of sulfite-added versus sulfite-free fermentations, leaving a significant gap in knowledge in this area.

The objective of the study discussed in Chapter 2 was to investigate how different concentrations of SO2 added at crush (0, 20, and 40 mg/L) could alter the composition of S. cerevisiae strains performing uninoculated fermentations at two

31 commercial wineries in the Okanagan Valley. Based on previous research conducted in the Okanagan Valley (Hall et al., 2011; Scholl et al., 2016), I predicted that the fermentations would be dominated by commercial strains of S. cerevisiae, even though they were not inoculated with commercial ADY. I also predicted that the composition of the S. cerevisiae population would differ at each winery, and that different SO2 addition concentrations would result in different yeast populations, even within each winery, based on previous research that found unique microbial populations among wineries (Scholl et al., 2016) as well as differences in yeast population as a result of SO2 addition (Suzzi and Romano, 1982).

1.9.2 The effect of sulfur dioxide addition at crush on the fungal and bacterial communities and the sensory attributes of uninoculated Pinot gris wines

Non-Saccharomyces yeasts were traditionally thought to be spoilage microbes, but evidence suggests that they can play an important role in improving the complexity of wine and contributing to the expression of terroir. Because of the potential for non-Saccharomyces yeasts to shape the aroma and flavour profile of the wine, many winemakers are opting to add less or no SO2 at crush and to let the must ferment without yeast inoculation, but more research is needed to determine what the risks and outcomes of these practices may be. Most previous research that investigated these topics used laboratory-scale fermentations and/or culture- dependent methodology, which could significantly underestimate microbial diversity, as SO2 is known to induce a VBNC state in some wine microorganisms (Andorrà et al., 2008; Bokulich et al., 2014; Constantí et al., 1998; Egli et al., 1998; Henick-Kling et al., 1998; Pateraki et al., 2014; Takahashi et al., 2014).

32

The overall objective of the study discussed in Chapter 3 was to investigate how different concentrations of SO2 added at crush (0, 20, and 40 mg/L) could alter the microbial communities conducting uninoculated fermentations of Pinot gris wines, and to investigate how changes in these microbial communities could lead to differences in the perceived sensory profiles of these wines. I predicted that the diversity and composition of the fungal and bacterial communities would differ among the three SO2 treatments, and that the resulting wines would differ in their sensory attributes.

1.9.3 Response to sulfur dioxide addition by two commercial Saccharomyces cerevisiae strains

Commercial S. cerevisiae strains are selected for their abilities to efficiently conduct alcoholic fermentation and withstand a variety of stressful environments, including SO2 (Piškur et al., 2006). However, because wine yeasts have different methods of surviving in the presence of SO2, it stands to reason that different commercial strains may preferentially employ different methods of SO2 resistance.

Previously, most studies that investigated the response by wine yeasts to SO2 have focused on one of these mechanisms at a time (Aranda et al., 2006; Goto-

Yamamoto et al., 1998; Nardi et al., 2010; Park and Hwang, 2008; Stratford et al.,

1987), without investigating the ways that these mechanisms may work together to provide sulfite protection. Furthermore, some of these studies used differences in gene expression as the sole method of inferring differences in yeast resistance mechanisms, but changes in gene expression are not always correlated with changes in protein translation or metabolic activity (de Sousa Abreu et al., 2009;

33

Hoppe, 2012; Schrimpf et al., 2009; Vogel and Marcotte, 2012). Additionally, few studies have investigated the effects of SO2 addition on the production of volatile secondary metabolites by S. cerevisiae (Boroski et al., 2017; Santos et al., 2015;

Sun et al., 2016).

The objective of the study discussed in Chapter 4 was to observe the responses of two genetically distinct commercial S. cerevisiae strains to the presence (50 mg/L) or absence (0 mg/L) of SO2 in fermentations of Pinot gris conducted under laboratory conditions. All of the main mechanisms of SO2 resistance were evaluated. Specifically, fermentation kinetics, yeast abundance,

SO2 concentration, H2S production, acetaldehyde production, and secondary metabolite composition as a result of SO2 addition were compared between the two strains. I predicted that the genetic distance between these two strains would result in different responses to SO2, and subsequently in the production of different volatile secondary metabolites that are important for the modulation of wine aroma.

1.9.4 The effect of sulfur dioxide addition and pied de cuve inoculation on the microbial communities and sensory profiles of Chardonnay wines

While much research has been conducted comparing inoculated and uninoculated fermentations, very little research has investigated pied de cuve inoculation methods (Benucci et al., 2016; Clavijo et al., 2011b; Li et al., 2012;

Moschetti et al., 2016; Ubeda Iranzo et al., 2000), and to my knowledge no studies have been conducted that focus specifically on the method of pied de cuve inoculation that involves adding vineyard-specific yeasts and bacteria in order to enhance the contribution of native microorganisms to the fermentation.

34

The objective of the study discussed in Chapter 5 was to investigate the combined effects of SO2 addition (0 or 40 mg/L) and alternative fermentation techniques (uninoculated or pied de cuve inoculation) on the fungal and bacterial communities, the Saccharomyces strains, and the sensory profiles of commercial

Chardonnay wines. The design of this study was a balanced 2 × 2 factorial, resulting in four SO2/inoculation treatments. I predicted that SO2 addition would significantly alter the diversity and composition microbial communities conducting fermentation, as well as the sensory profiles of the resulting wines, based on results from previous related studies (Bokulich et al., 2014; Henick-Kling et al., 1998). Because the pied de cuve inoculum was expected to be composed of sulfite-sensitive vineyard yeasts and bacteria, I also predicted that the pied de cuve inoculum would only survive and contribute to the fermentations when SO2 was not added at crush.

1.9.5 Competition between Saccharomyces cerevisiae and Saccharomyces uvarum in controlled Chardonnay fermentations

The cryotolerant yeast S. uvarum has previously been associated with fermentations of wine conducted at lower temperatures (Dellaglio et al., 2003;

Demuyter et al., 2004; Naumov et al., 2002, 2000; Sipiczki et al., 2001), even though no commercial strains of S. uvarum were available until extremely recently. These indigenous strains of S. uvarum could provide solutions for winemakers looking to inoculate indigenous yeasts to their fermentations, in order to enhance the local character or terroir of their wines while maintaining a consistent and reproducible product. Previously, research has been conducted to investigate the enological potential of S. uvarum (Castellari et al., 1994; Gamero et al., 2013; Magyar and

35

Tóth, 2011; Sipiczki et al., 2001), but only two studies have specifically tested the competitive abilities of S. uvarum when co-inoculated with S. cerevisiae (Alonso-del-

Real et al., 2017; Cheraiti et al., 2005). However, neither study used indigenous strains of S. uvarum, and neither study used different co-inoculation ratios, leaving a gap in knowledge surrounding the competitive abilities of indigenous S. uvarum under different inoculation conditions.

The objective of the study discussed in Chapter 6 was to observe the competitive interactions between an indigenous strain of S. uvarum and a commercial strain of S. cerevisiae when co-inoculated at different initial ratios and fermented at different temperatures. Laboratory-controlled fermentations were conducted at both 15 C and 24 C with the following S. cerevisiae : S. uvarum inoculation ratios: (1) 1:0, (2) 10:1, (3) 1:1, (4) 1:10, (5) 0:1. Fermentation kinetics, total yeast abundance, yeast relative abundance, and yeast-derived volatile secondary metabolite profiles were evaluated. I predicted that the strain with the highest initial concentration in each treatment would end up dominating, regardless of fermentation temperature. When inoculated at an equal concentration, I predicted that the S. cerevisiae strain would dominate at 24 C, but the S. uvarum strain would dominate at 15 C. Finally, I predicted that each strain would produce different volatile metabolite profiles, and that fermentation temperature would also affect the production of these metabolites.

36

Chapter 2: Effect of sulfur dioxide addition at crush on the Saccharomyces cerevisiae strain composition during uninoculated fermentations

2.1 Background

Sulfur dioxide (SO2) is added at different stages during the winemaking process as both an antioxidant and an antimicrobial agent. Due to its antimicrobial properties, its addition to grape must at crush, immediately before the onset of alcoholic fermentation (AF), can reduce or remove potential spoilage microorganisms brought into the winery on grape surfaces, due to the greater sensitivity of non-Saccharomyces yeasts and bacteria to SO2 (Bokulich et al., 2014;

Constantí et al., 1998; Henick-Kling et al., 1998). This addition can therefore create an environment that promotes the rapid colonization of the grape must by

Saccharomyces cerevisiae, which is generally more sulfite-resistant and is the dominant yeast in wine fermentations.

S. cerevisiae is predominantly a winery-resident yeast, and is either added to the must intentionally by the winemaker as a commercial active dry yeast (ADY) strain, or enters the must via grape surfaces, air, or winery equipment. S. cerevisiae is rarely found in high numbers on healthy grapes or in must prior to the onset of AF

(Mortimer and Polsinelli, 1999); instead, vineyard-derived non-Saccharomyces yeasts dominate the must during these early stages of winemaking (Domizio et al.,

2007; Zott et al., 2008). Uninoculated fermentations tend to allow for a higher diversity of yeast species and strains to participate in AF than do inoculated fermentations (Csoma et al., 2010), resulting in the production of a wider range of aromatic metabolites (Fleet, 2003; Romano et al., 2003). During uninoculated

37 fermentations, strains of S. cerevisiae enter the must without human assistance, and like inoculated fermentations, by the end of fermentation they are usually the only yeast species remaining (Heard and Fleet, 1985; Zott et al., 2008). Many strains of

S. cerevisiae may be present in a single fermentation; however, there are mixed results regarding the ability of different strains to dominate uninoculated fermentations. Some researchers suggest that only a small number of strains can usually dominate in uninoculated fermentations (Blanco et al., 2006; Frezier and

Dubourdieu, 1992), whereas other researchers have observed many strains of S. cerevisiae present throughout fermentation, with no single strain dominating (Egli et al., 1998; Gayevskiy and Goddard, 2012; Torija et al., 2001). In the past, it was presumed that the S. cerevisiae strains conducting uninoculated fermentations were indigenous, originating from the local environment. Although this seems to be the case in some wine regions (Gayevskiy and Goddard, 2012; Santamaría et al.,

2008), in many others, including the Okanagan Valley wine region of Canada, uninoculated fermentations have been shown to be dominated by commercial strains previously and concurrently used at their respective wineries (Blanco et al.,

2011; Clavijo et al., 2011a; Constantí et al., 1997; Hall et al., 2011; Scholl et al.,

2016).

Even though S. cerevisiae is rarely dominant in grape must at the time of initial SO2 addition, it is the dominant species during AF, and different S. cerevisiae strains contribute differently to the production of flavour-active metabolites (Fleet,

2003; Romano et al., 2003). Although different concentrations of SO2 have been shown to change species composition of yeasts (Bokulich et al., 2014; Constantí et

38 al., 1998; Henick-Kling et al., 1998; Takahashi et al., 2014), few studies have investigated the effects that SO2 has on S. cerevisiae composition at the strain level

(Constantí et al., 1998; Egli et al., 1998; Henick-Kling et al., 1998; Suzzi and

Romano, 1982), and in these studies it is often difficult to extract information regarding population differences among SO2 treatments. Constantí et al. (1998) found that the dominant strain in spontaneous (uninoculated) fermentation was a commercial strain used previously at the experimental winery, regardless of SO2 addition level. However, the experimental treatments in this study were not replicated, which prevented the use of inferential statistics. Egli et al. (1998) found that spontaneous fermentations had the highest S. cerevisiae strain diversity, and that each fermentation had a unique composition of S. cerevisiae strains. However, they did not compare the S. cerevisiae strain profiles of sulfite-added and sulfite-free fermentations. Henick-Kling et al. (1998) also lacked comparisons of strain composition between fermentations of different SO2 treatments. Suzzi and Romano

(1982) identified Saccharomyces strains using metabolic, rather than molecular methods. Although they did make important observations regarding the behaviour of

Saccharomyces strains in sulfite-added versus sulfite-free musts, they could not statistically test whether the fermentations with different SO2 concentrations contained different yeast compositions. Moreover, the level of SO2 added was also very high (100 mg/L), limiting the applicability of this study to the wine industry.

Finally, some researchers, studying the effects of SO2 on wine yeasts, have used the term ‘population’ to refer to species-level analysis (Constantí et al., 1998;

Henick-Kling et al., 1998; Pateraki et al., 2014), which may confuse readers looking

39 to gain information on strain-level effects. In this study, we refer to ‘populations’ exclusively as groups of yeast strains belonging to the same species, and

‘communities’ as groups of yeasts composed of different species.

Our study builds on the important results of previously conducted work, while attempting to overcome some limitations of those studies. Here, we specifically define our use of the term population; we conducted the research in commercial wineries, to maximize the applicability of the study to the winemaking industry; and we performed each treatment in triplicate, to permit the use of inferential statistics and account for fermentation-to-fermentation variability. The objective of this study was to determine the effect of different levels of SO2 addition at crush (0, 20, and 40 mg/L SO2) on the relative abundance and composition of S. cerevisiae populations at the strain level during uninoculated fermentations. We predicted, based on previous research conducted in the Okanagan Valley wine region (Hall et al., 2011;

Scholl et al., 2016), that the majority of strains isolated would be of commercial origin, regardless of the amount of SO2 added. However, we also predicted that the relative abundance of these commercial strains would differ based on the amount of

SO2 added at crush; we predicted this difference in relative abundance would contribute to differences in overall strain composition between treatments. We based these predictions on previous findings showing that different S. cerevisiae strains have differing abilities, in terms of SO2 resistance and production, mechanisms of competition, and growth rates (de Ullivarri et al., 2014; Egli et al., 1998; Nadai et al.,

2016; Suzzi et al., 1985; Weeks, 1969).

40

2.2 Materials and methods

2.2.1 Study sites and experimental design

This study was conducted during the 2014 at two commercial wineries, both located on the eastern side of Okanagan Lake in the Okanagan

Valley wine region of Canada. Cedar Creek Estate Winery, here called Winery 1

(W1), conducted its first vintage in 1987 and is a medium-sized winery, producing

30-40,000 cases (270,000-360,000 L) of wine annually. Its winemakers conduct both inoculated and uninoculated fermentations of many grape varietals, including Pinot gris, which is the varietal used in this study at this winery. 50th Parallel Estate

Winery, here called Winery 2 (W2), is a small winery, producing approximately

10,000 cases (90,000 L) of wine annually. The 2014 vintage was the first year fermentations were conducted in a new facility at this winery, although the first vintage on-site was conducted in 2011. W2 also conducts both inoculated and uninoculated fermentations, and produces wines from six different varietals, with a focus on Pinot noir, which is the varietal used in this study at this winery. We chose two varietals with different winemaking protocols and wineries with vastly different winemaking and inoculation histories in order to observe whether the same trends could be observed at both wineries in spite of these differences. However, because of the extent of variation between the experiments conducted at W1 versus W2, we will not statistically compare the results of one winery to the other, but instead we will use both wineries to illustrate trends.

At each winery, the grapes for experimental fermentation were sourced from a single vineyard. Grapes were harvested and crushed/pressed according to

41 standard viticultural practices in British Columbia, Canada. At W1, Pinot gris must was first crushed and pressed into a large stainless steel tank, and then transferred into nine new 225 L oak barrels (Alain Fouquet, French Cooperage, medium toast), which were steam-cleaned prior to the addition of the grape must. The three SO2 addition levels tested at W1 were 0, 20, and 40 mg/L SO2 (n = 3 per treatment). At

W2, the Pinot noir must was crushed directly into nine 1.5 kL macrobins, which were also steam-cleaned prior to the addition of the grape must. Only two SO2 addition levels were tested at W2: 20 and 40 mg/L SO2 (n = 3 per treatment). At both wineries, SO2 was added to the grape must in the form of potassium metabisulfite

(KMS), and cold settling was conducted in the presence of SO2 (if added). The must was allowed to undergo AF uninoculated, without the addition of any commercial

ADY strains. Fermentations at both wineries were conducted at cellar temperature.

At W1, the Pinot gris fermentations began at 13.9 °C ± 0.03 but to 20.8 °C ±

0.09 by mid-fermentation. At W2, the Pinot noir fermentations began at 10.3 °C ± 0.4 and rose to 25.0 °C ± 0.8 by mid-fermentation. Due to the logistical limitations of conducting experiments with industrial partners, it was not possible to perform all

SO2 treatment levels at both wineries. However, because the standard addition of

SO2 at both wineries is 40 mg/L, our aim was to determine whether adding lower levels of SO2 at crush would result in differences in S. cerevisiae strain composition; as such, we considered 40 mg/L SO2 as the control in this experiment.

2.2.2 Sampling

Samples for chemical analysis and °Brix levels were taken from the grape must prior to the beginning of AF (Table 2.1). Samples for microbial analysis were

42 taken at three stages of fermentation: Early (14-19 °Brix), Mid (7-12 °Brix), and Late

(<2 °Brix). Samples were collected aseptically in sterile 50 mL centrifuge tubes and were immediately transported to the laboratory.

Table 2.1. The initial chemical composition ± SEM of Pinot gris must (W1) and Pinot noir must (W2) at cold-settling. T.A. refers to titratable acidity. Different superscript letters indicate significant differences between treatments (p < 0.05) at W2. Statistical analysis of must between treatments at W1 was not conducted as these initial samples were not independent.

Treatment °Brix Temp (°C) pH T.A. (g/L) W1 0 mg/L 22.5 ± 0.0 13.97 ± 0.04 3.1 ± 0.0 9.2 ± 0.03 W1 20 mg/L 22.6 ± 0.02 13.87 ± 0.02 3.0 ± 0.0 9.5 ± 0.1 W1 40 mg/L 22.6 ± 0.0 13.87 ± 0.02 3.0 ± 0.03 9.4 ± 0.3

a a a a W2 20 mg/L 23.1 ± 0.06 10.5 ± 0.7 3.4 ± 0.03 12.3 ± 0.2

b a b a W2 40 mg/L 22.5 ± 0.1 10.0 ± 0.0 3.2 ± 0.03 12.0 ± 0.4

2.2.3 Chemical analysis

Temperature and °Brix were measured throughout fermentation, using a portable Anton Paar© density meter (Saint Laurent, Canada) at W1. At W2, °Brix at crush were measured using an ATAGO® Pocket Refractometer PAL-1 (Tokyo,

Japan), and on subsequent days using a hydrometer. Temperature was measured using a thermometer at W2. For both wineries, pH was measured using a pH meter

(Fisher ScientificTM accumetTM XL 150 Benchtop Meter) and titratable acidity was measured via titration with 0.01N NaOH.

43

2.2.4 Yeast enumeration and isolation

Each fermentation sample was diluted in series, plated on solid YEPD media

(10 g/L yeast extract, 10 g/L bacterial peptone, 20 g/L dextrose, 20 g/L granulated agar), and incubated at 28 °C for two days. Samples from the must before the onset of fermentation were not plated, because S. cerevisiae is not usually found at this stage (Zott et al., 2008). Furthermore, our objective was not to determine whether any S. cerevisiae strains, indigenous or otherwise, were present in the must prior to the addition of SO2 and the onset of fermentation, but instead to identify the strains able to persist throughout and potentially dominate the fermentations. For each sample, plates containing 30-300 yeast colonies were used, and 72 yeast colonies were randomly chosen and subsequently isolated onto YEPD media. The DNA of these isolates was extracted and amplified using multiplex PCR on eight microsatellite loci (outlined in section 2.2.5 below). Of the 72 yeast colonies isolated from each sample, 40 colonies that were positively identified as S. cerevisiae strains were randomly selected for further analysis. Especially in the early stage of fermentation, a number of the isolated yeast colonies could not be identified through microsatellite analysis, and were considered to be non-Saccharomyces yeasts, which can often persist into the later stages of fermentation (Domizio et al., 2007;

Heard and Fleet, 1985).

2.2.5 S. cerevisiae strain typing

DNA from each S. cerevisiae isolate was extracted in preparation for strain identification using a water DNA extraction method (Scholl et al., 2016). Multiplex

PCR was performed on the following eight microsatellite loci to identify S. cerevisiae

44 isolates to the strain level: C4, C8, C3, C11, YML091c, YPL009c, YOR267c, and

YLR177w. These loci are mostly unlinked, with the exception of C3 and C8 (both located on Chromosome VII), and C4 and YOR267c (both located on Chromosome

XV) (Legras et al., 2005; Richards et al., 2009). The following were added to each well of a 96-well PCR plate: 0.03 μL DNA-grade water, 1.6 μL GoTaq reaction mix

(5 buffer), 0.47 μL dNTPs (10 mM each), 5.25 μg Bovine Serum Albumin (BSA),

18.13 pmol MgCl2, 6.5 pmol C4 primers, 2.2 pmol C8 primers, 0.45 pmol C3 primers,

1.25 pmol C11 primers, 4.0 pmol YML091c primers, 0.75 pmol YPL009c primers,

0.6 pmol YOR267c primers, 0.45 pmol YLR177w primers, 0.65 U GoTaq DNA

Polymerase, and 1.5 μL DNA template (post-extraction), for a total volume of 8.3 μL per well. Equal volumes of forward and reverse primers were used for each primer set. PCR, fragment analysis, and genetic fingerprinting were performed as outlined by Scholl et al. (2016). GenAlEx version 6.1 software was used to calculate the probability that two unrelated strains would have identical multilocus genotypes

(Peakall and Smouse, 2012, 2006). This probability was determined to be one in 1.2

 107 (probability of identity = 1.7  10-9).

2.2.6 Sulfite resistance assay

Eight dominant commercial S. cerevisiae strains were selected for a sulfite resistance assay. YEPD + TA plates were prepared by buffering the media to pH 3.5 with L-tartaric acid; precisely 18 mL media was poured onto each plate. To create plates containing 0, 1.0, and 2.0 mM Na2SO3, an appropriate amount of 0.25 M

Na2SO3 was spread onto the plates and allowed to set overnight (Park et al.,

1999a). Yeast colonies were grown in liquid YEPD media at 28 °C overnight before

45 being spread-plated (n = 3 per treatment). Plates were incubated for 48 h at 28 °C and checked for growth inhibition.

2.2.7 Statistical analysis

The nature of this study necessitated that the two wineries involved be analyzed separately, as there were too many variables between the study sites, including grape varietal used, fermentation vessel type, winery age, and inoculation histories. Therefore, the data collected from each winery were not combined during statistical analysis, but instead used to illustrate trends. A Bray-Curtis dissimilarity index was calculated using untransformed S. cerevisiae strain abundance for each sample, from which a Principal Coordinates Analysis (PCoA) was generated in order to visualize the spatial distribution of strains among the SO2 treatments at each winery. Permutational analysis of variance (PERMANOVA) tests, using Bray-Curtis dissimilarity and Type I sums of squares, were performed to test for differences in strain composition between SO2 treatments. PERMDISP tests, using Bray-Curtis dissimilarity and calculating deviation from centroid, indicated no violation of homogeneity of multivariate dispersion at either W1 (F(2,24) = 2.4, p = 0.14) or W2

(F(1,16) = 0.39, p = 0.58). Because the design of this study includes both repeated measures and multiple comparisons, the p-values obtained from the performed

PERMANOVA tests may overestimate differences among treatments. Therefore, we used the PCoA ordinations to visualize distances between samples of different treatments (Knight et al., 2015; Lorion and Kennedy, 2009). Test statistics (F values for PERMDISP and Pseudo-F values for PERMANOVA) were calculated based on

999 permutations of raw data. Where appropriate, pairwise comparisons between

46 treatments were performed, also based on 999 permutations of raw data. S. cerevisiae strain composition was analyzed and visualized using PRIMER version 6 software with PERMANOVA+ add-on (Plymouth, USA) (Clarke and Gorley, 2006).

Visual representation of S. cerevisiae strain abundance was performed using

GraphPad Prism version 6 software (San Diego, USA).

2.3 Results and discussion

2.3.1 S. cerevisiae relative strain abundance

A total of 66 unique S. cerevisiae strains were identified between the two wineries, of which 24 were commercial strains and 42 were of unknown origin

(Appendix A, Table A.1). At W1, 45 different strains were identified (21 commercial and 24 unknown), and at W2, 32 different strains were identified (12 commercial and

20 unknown) (Tables A.2 and A.3). Sixty-four to ninety-eight percent of the S. cerevisiae strains isolated during uninoculated fermentation were commercial ADY strains (Figure 2.1). Most of the commercial strains identified were those that were previously or concurrently used by their respective wineries (Tables A.4 and A.5).

The importance of inoculation history has been emphasized previously (Beltran et al., 2002; Hall et al., 2011; Scholl et al., 2016). The abundance of unknown strains was greater at W2 than at W1. Because W2 was conducting its first vintage in a new facility during this experiment, it is probable that this winery did not yet have winery- resident strains, so the uninoculated fermentations at W2, as compared with W1, were much more vulnerable to other environmental conditions, such as inoculated fermentations being conducted in close proximity and contamination by yeasts from outside sources. In future vintages, it will be interesting to note whether the

47 abundance of unknown strains at this winery diminishes with prolonged use of commercial strains, or whether some of these potentially indigenous strains will establish themselves as competitive winery residents. The finding that commercial strains dominated over unknown strains appeared to be independent of SO2 addition level at both wineries, indicating that the dominance of commercial strains over unknown or indigenous strains is more strongly determined by a winery’s inoculation practices than by fermentation practices such as using uninoculated fermentations or by adding SO2 in different concentrations. Nevertheless, the level of SO2 added at crush did significantly impact the presence and absence, as well as the relative abundance, of the dominant strains identified throughout the fermentations analyzed. Dominant strains are those that comprise ≥ 10% of the relative strain abundance throughout fermentation, therefore likely having the greatest impact on fermentation-derived sensory attributes (Sabate et al., 1998; Scholl et al., 2016). At

W1, each of the three SO2 treatments was dominated by four strains, all of them commercial. At W2, the treatments had either two or three dominant strains, one of which was of unknown origin (Table 2.2). Only one dominant strain at W1 (DV10) was shared by all three treatments, while two at W2 ( and RC212) were shared by both treatments. Both Lalvin® DV10 and Lalvin® RC212 were being used concurrently at W1 and W2, respectively, so it is not surprising that these yeast strains made their way into all experimental fermentations. Interestingly, the commercial strain Syrah had not been used at W2, but it was one of the dominant strains isolated at W2. It is possible that this strain was being used at nearby wineries, and was brought into contact with these fermentations via human or insect

48 vectors (Christiaens et al., 2014; Goddard et al., 2010; Knight and Goddard, 2014;

Stefanini et al., 2012). The other dominant strain identified at W2 (UN25) was in the

20 mg/L SO2 treatment, and was an unknown strain, potentially indigenous in origin.

A new facility with a limited history of commercial ADY use, such as W2, is more likely to have dominant indigenous S. cerevisiae strains (Beltran et al., 2002). Figure

2.2 demonstrates the relative abundance of dominant strains found across the different stages of fermentation at W1 and W2.

Figure 2.1. Commercial vs unknown strain abundance. Relative abundance (± SEM) of commercial and unknown S. cerevisiae strains in fermentations to which different levels of sulfur dioxide were added at crush at two wineries (W1 and W2). Relative abundance was determined from three fermentation stages, and each treatment had three replicate fermentations.

49

Figure 2.2. Percent composition of dominant S. cerevisiae strains at three stages of uninoculated fermentation at (A) Winery 1 and (B) Winery 2. Samples were taken during the Early (E), Mid (M), and Late (L) stages of fermentation. Dominant strains are those strains comprising at least 10% of the percent composition throughout the course of fermentation in at least one treatment. Values for each stage are the means of three replicate fermentations, where 40 S. cerevisiae isolates were identified for each sample for a total of 120 isolates per stage of fermentation. For variation among samples of the same treatment, please refer to Tables A.2 and A.3.

50

At W1, some of the commercial strains were able to dominate in the fermentations of some treatments, but not others (Table 2.2). For example, the strains Syrah and K1-V1116 were dominant strains in the two treatments where SO2 was added (20 and 40 mg/L), but were unable to dominate in the 0 mg/L SO2 treatment. Supertuscan was a dominant commercial strain in the 0 and 20 mg/L SO2 treatments, but was unable to dominate at 40 mg/L. Strains 3001 and ICV-D47 showed opposite, dose-dependent relationships with SO2. For example, 3001 was a dominant strain when 0 mg/L SO2 was added at crush, but its relative abundance decreased in a dose-dependent manner with increasing SO2 addition levels, while the relative abundance of ICV-D47 increased with increasing SO2 addition levels, and was a dominant strain at 40 mg/L SO2. Although more research is needed to confirm this, we speculate that those strains better adapted to musts containing higher levels of SO2 at crush, such as Syrah, ICV-D47, and K1-V1116, may be faster-growing strains, better able to capitalize on the decreased competition from non-Saccharomyces yeasts that were removed during SO2 addition. Similarly, we speculate that the strains best suited to the musts to which no SO2 was added, such as CY3079 and 3001, are likely slower-growing yeasts but ultimately more competitive against the vineyard microorganisms that were not removed with SO2 addition.

51

Table 2.2. Relative abundance (% ± SEM) of dominant S. cerevisiae strains in fermentations to which different concentrations of sulfur dioxide were added at crush at two wineries (W1 and W2). Relative abundance and standard error (SEM) values were obtained from 360 S. cerevisiae isolates (40 isolates from three fermentation stages, with three replicate fermentations per treatment). Dominant strains consist of those comprising at least 10% of the relative abundance throughout fermentation. Results in bold indicate dominance in that particular treatment. Spaces denoted by the “-” symbol indicate that the specific strain was not identified in that treatment.

Winery 1 Winery 2 Strain 0 mg/L 20 mg/L 40 mg/L 20 mg/L 40 mg/L Enoferm Syrah 3.3 ± 2 19.7 ± 5 11.7 ± 3 14.7 ± 1 26.4 ± 6 Lalvin ICV-D47 0.28 ± 0.3 2.8 ± 0.7 12.2 ± 3 - - Lalvin DV10 35.0 ± 2 19.4 ± 6 25.8 ± 6 - - Lalvin CY3079 11.4 ± 2 3.6 ± 2 5.6 ± 1 6.9 ± 2 5.3 ± 2 Lalvin K1-V1116 2.2 ± 0.7 15.3 ± 6 18.9 ± 4 0.30 ± 0.3 0.30 ± 0.3 Lalvin RC212 8.1 ± 3 2.8 ± 1 7.5 ± 1 45.3 ± 6 49.4 ± 3 Premium Supertuscan 10.8 ± 1 10.3 ± 2 2.8 ± 2 - - Vitilevure 3001 13.6 ± 0.3 6.9 ± 1 3.3 ± 0.5 - - UN25 - - - 18.6 ± 6 2.5 ± 1 Minor commercial strains 12.8 ± 2 15.6 ± 3 12.8 ± 2 8.1 ± 2 9.4 ± 1 Minor unknown strains 2.5 ± 1 3.6 ± 1 2.8 ± 1 6.1 ± 2 6.7 ± 2

To test whether differences in SO2 resistance contributed to the differential abilities of these strains to dominate in different treatments, the eight commercial strains that dominated different treatments (Table 2.2) were selected for a sulfite resistance assay by plating on YEPD + TA (pH 3.5) containing either 0, 1.0, or 2.0 mM Na2SO3. Only two strains were affected by these levels of sulfite: strains

CY3079 and DV10. Strain CY3079 was completely inhibited from growing at 2.0 mM, and growth was slowed at 1.0 mM, as evidenced by a decrease in the size of colonies (by approximately half) on those plates. Strain DV10 showed diminished growth at 2.0 mM. The results for strain CY3079 seem to correlate with the dominance pattern observed for this strain, as it was only able to dominate the fermentations where no SO2 was added. However, this trend was not extended to 52 strain 3001, which was the other strain that dominated solely in the 0 mg/L SO2 treatment. Furthermore, DV10, which was affected by 2.0 mM Na2SO3, was the most abundant strain in all fermentations at W1. In the case of DV10, it seems that the dominance of this strain in these treatments was simply due to its increased abundance in the winery environment at the time these experimental fermentations were conducted. The lack of correlation between our fermentation results and the relative sulfite resistance of the dominant strains suggests that sulfite resistance is not the driving force behind these differences in dominance patterns. We therefore hypothesize that the patterns of dominance patterns that we observed were more the result of many combined factors, which could include speed of cell replication at the onset of fermentation, differential abilities to obtain and sequester nutrients, the production of killer toxins, or other competitive traits. More research is needed in order to test these alternative hypotheses.

2.3.2 S. cerevisiae strain composition

At each winery, the S. cerevisiae strain composition of every sample was analyzed by performing one-factor PERMANOVA tests, and visualized using a

Principal Coordinates Analysis (PCoA) (Figure 2.3). Composition, unlike diversity, takes into account not only the relative abundance of individual strains, but also the identity of each strain. Plotting composition, using a PcoA, allows for visual distinction in multidimensional space of samples with different strain profiles. In this way, samples that contain similar strains but different relative strain abundance will appear as separate points; the more dissimilar the relative abundance of each strain, the more distant the two points will be. One-factor PERMANOVA tests

53 indicated significant differences in strain composition at both W1 (p = 0.001; p ≤

0.002 for all pairwise comparisons) and W2 (p = 0.001) (Table 2.3). At W1, different

SO2 addition treatments resulted in completely unique S. cerevisiae strain compositions throughout fermentation (Figure 3.3A). At W2, there was only a single- sample overlap in strain composition between treatments (Figure 3.3B). At W2, there were few dominant strains, and those that were dominant were mostly shared between the two SO2 treatments. It is therefore likely that much of the variation in strain composition at W2 can be attributed to differences in the presence/absence and relative abundance of minor S. cerevisiae strains (Table A.3). The similarity in dominant strains, as well as, the limited use of commercial strains at this winery, may have contributed to the overlap in strain composition observed at W2.

Conversely, the unique strain profiles of each SO2 treatment at W1 are likely attributed to differences in the relative abundance of dominant strains, which varied greatly between treatments (Table 2.2).

54

Figure 2.3. S. cerevisiae strain composition. These PCoA ordinations describe the composition of S. cerevisiae strains during uninoculated fermentations with different levels of initial SO2 addition at (A) W1 and (B) W2. At W1, either 0, 20, or 40 mg/L SO2 was added at crush to Pinot gris must, and at W2, either 20 or 40 mg/L SO2 was added at crush to Pinot noir must. Individual data points represent the S. cerevisiae strain composition of a single sample (obtained using a Bray-Curtis dissimilarity index, calculated from the relative strain abundance of 40 S. cerevisiae isolates per sample). Each fermentation vessel was sampled at three stages of fermentation, and each treatment was replicated in triplicate, for a total of nine samples per treatment. One-factor PERMANOVA tests indicated significant differences in strain composition between treatments at each winery.

55

Table 2.3. Results of one-factor PERMANOVA tests evaluating the effects of SO2 addition level at crush on S. cerevisiae strain composition throughout spontaneous fermentation at W1 and W2. Results are based on 999 unrestricted permutations of raw data. Statistical analysis was conducted separately for each winery.

Source d.f. SS MS Pseudo-F p

W1 Treatment 2 12691 6345.7 9.1695 0.001

W1 Residual 24 16609 692.04

W1 Total 26 29300

W2 Treatment 1 2271.9 2271.9 6.0929 0.001

W2 Residual 16 5966 372.87

W2 Total 17 8237.8

Results in bold are significant at p < 0.05

2.4 Summary

Our results clearly indicate that SO2 addition level at crush had a significant effect on S. cerevisiae populations in uninoculated fermentations. The results of our research are of interest to both the scientific community as well as to winemakers looking to add lower levels of SO2 to their uninoculated fermentations. Our results can also be applicable to winemakers that are trying to achieve more complex wines through the management of strain composition during uninoculated fermentations.

Nevertheless, additional research is necessary in order to determine whether these differences in S. cerevisiae strain composition throughout fermentation translate into tangible differences in chemical and sensory profiles of wine.

56

Chapter 3: The effect of sulfur dioxide addition at crush on the fungal and bacterial communities and the sensory attributes of Pinot gris wines

3.1 Background

Sulfur dioxide (SO2) has been used in winemaking for centuries, acting as both an antioxidant and an antimicrobial agent. It is often added at crush, prior to the start of alcoholic fermentation, to prevent the growth of unwanted microorganisms that enter the grape must from the vineyard or winery equipment. SO2 is almost always added post-fermentation as well as at bottling to act as a preservative agent.

Using excessive amounts of SO2 in winemaking can be undesirable from both a health standpoint and from an enological perspective, where the addition of too much SO2 can negatively impact the sensory attributes of a wine (Guerrero and

Cantos-Villar, 2015; Yang and Purchase, 1985). Because of these reasons, there has been a consumer-driven push in recent years for SO2 alternatives in winemaking; however, SO2 remains the most effective antioxidant and preservative available (Falguera et al., 2013; Guerrero and Cantos-Villar, 2015; Izquierdo-Canas et al., 2012).

Saccharomyces cerevisiae, the dominant yeast in winemaking, tends to be more resistant to SO2 addition than bacteria and non-Saccharomyces yeasts

(Bokulich et al., 2014; Constantί et al., 1998; Henick-Kling et al., 1998). S. cerevisiae is found in very low numbers on healthy grapes (Mortimer and Polsinelli, 1999), and therefore non-Saccharomyces yeasts dominate the must stage before the onset of alcoholic fermentation (cold-settling). The predominance of these yeasts as well as bacteria is generally not favored and winemakers can therefore choose two methods

57 to prematurely remove them: the addition of sufficiently high levels of SO2 at crush, and/or the inoculation of the must with a commercial S. cerevisiae strain, which will usually out-compete the vineyard yeasts. While non-Saccharomyces yeasts were originally thought to be exclusively spoilage organisms, a substantial and growing body of evidence has pointed to the ability of non-Saccharomyces yeasts to play important roles in the expression of varietal aromas, as well as the production of unique sensory-active secondary by-products that can increase the complexity of a wine and the expression of terroir (Ciani et al., 2010; Fleet, 2003; Jolly et al., 2014;

Romano et al., 2003; Viana et al., 2008). For these reasons, many winemakers are opting to add less or no SO2 at crush, or to let their musts ferment uninoculated

(spontaneously). However, more research needs to be conducted in order to fully understand the implications of these decisions to allow winemakers to make informed decisions in the context of uninoculated and low-SO2 winemaking.

Previous research has investigated these topics (Constantί et al., 1998; Egli et al., 1998; Henick-Kling et al., 1998; Suzzi and Romano, 1982; Takahashi et al.,

2014), but the introduction of new molecular technologies, that allow for a more accurate and thorough evaluation of the microorganisms involved in winemaking, necessitates further research into this area. High-throughput amplicon sequencing technologies such as Illumina MiSeq, among others, have enabled the detection of microorganisms in wine fermentations that were previously undetectable using culture-dependent techniques. It was previously thought that non-Saccharomyces yeasts were unable to survive in conditions exceeding 3-4% (v/v) ethanol, but culture-independent molecular identification has shown that non-Saccharomyces

58 yeasts and bacteria may survive until the end of alcoholic fermentation and in turn may be contributing significantly to the aroma and flavour profile of the wine

(Bokulich et al., 2014; Kioroglou et al., 2018; Stefanini et al., 2016a). These microorganisms may be present in too low an abundance to be identified through culture-dependent methods, they may be unable to grow on the media most commonly used for yeast or bacterial isolation, or they may be present in the fermentation in a viable but nonculturable (VBNC) state (Agnolucci et al., 2010; Divol et al., 2012). New research suggests that S. cerevisiae may produce metabolites that decrease the culturability of non-Saccharomyces yeasts (Wang et al., 2016), necessitating the use of culture-independent techniques to accurately identify the full yeast communities present in fermentations. Due to these reasons, using culture- dependent analysis when attempting to evaluate the entire microbial community in a wine sample may underestimate microbial diversity and overestimate the importance of a few species or genera (Serpaggi et al., 2012).

The closed-system conditions of winemaking mean that as alcoholic fermentation progresses, the availability of nutrients decreases concomitantly with an increase in alcohol content and creates a progressively inhospitable environment for the microorganisms present. Towards the end of fermentation, the amount of dead yeasts and bacteria that can no longer contribute to the fermentation accumulate significantly (Branco et al., 2012). To prevent these organisms from misrepresenting the viable microbial community during analysis, DNA-binding dyes such as propidium monoazide (PMA) can be added to samples prior to DNA extraction to prevent the amplification of DNA from dead cells (Andorrà et al., 2010;

59

Tantikachornkiat et al., 2016). When microbes die in fermentation, the integrity of their cell membranes becomes compromised, allowing PMA to enter dead cells and bind to genomic DNA. When exposed to light, PMA binds irreversibly to the DNA, preventing it from being amplified during polymerase chain reaction (PCR). This current study is the first of its kind to evaluate the living microbial communities (via the use of PMA) of commercial wine fermentations with respect to SO2 addition at crush. To our knowledge, only one other study has used high-throughput amplicon sequencing to evaluate the effects of SO2 addition on fungal and bacterial communities during alcoholic fermentation (Bokulich et al., 2014).

This current study builds upon the design and results of seven important and relevant studies, three published in 1998, one published in 2008, and three published in 2014 (Andorrà et al., 2008; Bokulich et al., 2014; Constantí et al., 1998;

Egli et al., 1998; Henick-Kling et al., 1998; Pateraki et al., 2014; Takahashi et al.,

2014). While these studies form the basis of our understanding of uninoculated and/or sulfite-free fermentations, our research attempts to fill some of the gaps of these studies and to update knowledge of the topic using current molecular technologies. Constantί et al. (1998) investigated the combined effects of SO2 addition and commercial yeast inoculation but neither included biological replicates in their experimental design nor evaluated the effects of these treatments on the sensory attributes of the resulting wines. Both Henick-Kling et al. (1998) and Egli et al. (1998) also investigated the combined effects of SO2 addition and commercial yeast inoculation. These two studies evaluated the sensory attributes of the wines produced, but only Henick-Kling et al. (1998) compared sulfited and unsulfited wines

60 during sensory analysis. Neither study included enough biological replicates to allow for the use of inferential statistics. Furthermore, all three experiments published in

1998 were scaled down to between

80 L and 12 L fermentations and were conducted away from commercial wineries, thus limiting their direct applicability to the commercial winemaking process. These studies used a combination of culture-based methods to identify yeasts to the species and sometimes to the strain level, and while some of these techniques are still used today, the advent of culture-independent analysis such as high-throughput amplicon sequencing has allowed for the identification of rare and VBNC yeasts and bacteria in fermentations. Takahashi et al. (2014) and Pateraki et al. (2014) compared culture-dependent and culture-independent methods of evaluating microbial diversity, using denaturing gradient gel electrophoresis (DGGE) as the culture-independent method, but did not conduct a sensory evaluation of the wines.

Andorrà et al. (2008) used both DGGE and quantitative PCR (qPCR) to identify the fungal and bacterial communities in sulfited and unsulfited wines, but did not replicate treatments. Bokulich et al. (2014) used Illumina MiSeq to evaluate the fungal and bacterial communities in fermentations to which a range of SO2 concentrations were added, and observed changes in the bacterial, but not the fungal, community in response so SO2 addition. Each treatment was replicated in triplicate, but a sensory evaluation of the wines was not performed. All three studies from 2014 were also conducted at experimental scales (< 1 L, 14 L, and 19 L, respectively), and Takahashi et al. (2014) evaluated only inoculated fermentations.

61

The objectives of this study were to: 1) determine the effect of different levels of SO2 addition at crush (0, 20, and 40 mg/L SO2) on the relative abundance and the composition of fungal and bacterial communities present throughout uninoculated

(spontaneous) fermentations; and 2) evaluate the effect of SO2 addition on the wine sensory attributes of Pinot gris wines fermented at a commercial winery in British

Columbia, Canada. Each treatment was replicated in triplicate in new 225 L oak barrels, and the fungal and bacterial communities were determined using Illumina

MiSeq sequencing; samples were treated with PMA addition to identify only the living community. We expected that the diversity and composition of the fungal and bacterial communities would differ among the three SO2 treatments, and that the resulting wines would differ in their sensory attributes.

3.2. Materials and methods

3.2.1 Experimental design and sampling

This study was conducted during the 2014 vintage at Cedar Creek Estate

Winery, a medium-sized commercial winery located on the east side of Okanagan

Lake in British Columbia, Canada. This winery produces 30,000-40,000 cases

(270,000-360,000 L) of wine annually, and conducts both inoculated and uninoculated (spontaneous) fermentations of many grape varietals.

In this study, uninoculated fermentations of Pinot gris were evaluated. Grapes were sourced from a single vineyard associated with the winery, and were harvested and crushed/pressed according to standard viticultural practices in British Columbia,

Canada. The grape must was first crushed and pressed into a large stainless steel tank, and then transferred into nine new 225 L French oak medium-toast barrels

62

(Alain Fouquet & Associates Inc, Napa, CA, USA), which were steam-cleaned prior to the addition of the grape must. SO2 was added in three concentrations: 0, 20, and

40 mg/L SO2 (n = 3 per treatment) in the form of potassium metabisulfite (KMS).

Alcoholic fermentation progressed uninoculated (spontaneously) without the addition of any commercial yeast strains. Fermentations were conducted at cellar temperature, beginning at 13.9 °C ± 0.03 and rising to 20.8 °C ± 0.09 by the middle of fermentation.

3.2.2 Chemical analysis

Samples for chemical analysis were taken from the grape must during the cold-settling stage (Table 3.1) and again at the end of alcoholic fermentation (Table

3.2). Residual sugar levels, measured as °Brix, were evaluated daily throughout alcoholic fermentation. Depending on the pH of the wine, SO2 can be free or can be bound to various compounds present in the grape must (Divol et al. 2012). Both free and total SO2 levels were determined at four stages of fermentation: Cold-settling (>

22 °Brix), Early (14 - 19 °Brix), Mid (7 - 12 °Brix), and Late (< 2 °Brix). Samples were collected aseptically in sterile 50 mL plastic centrifuge tubes and were immediately transported to the laboratory on ice for processing.

63

Table 3.1. Chemical composition of Pinot gris grape must at Cold-settling ± SEM. Three different levels of SO2 were added to the must immediately prior to sampling (n = 3 per treatment). Yeast assimilable nitrogen (YAN) was measured as the sum of alpha amino nitrogen (mg/L) and ammonia (mg/L) concentrations.

Chemical composition ± SEM Chemical parameter 0 mg/L SO2 20 mg/L SO2 40 mg/L SO2 Temperature (°C) 14.0 ± 0.07 13.9 ± 0.03 13.9 ± 0.03 pH 3.25 ± 0.006 3.26 ± 0.003 3.27 ± 0.003 Molecular SO2 (mg/L) N/A 0.21 ± 0.04 0.54 ± 0.01 Residual sugar (°Brix) 22.5 ± 0.0 22.6 ± 0.03 22.6 ± 0.0 YAN (mg/L) 242.6 ± 6.5 234.5 ± 11 215.5 ± 7.2 Volatile acidity (g/L) 0.10 ± 0.003 0.10 ± 0.009 0.10 ± 0.003 Total acidity (g/L) 7.3 ± 0.03 7.3 ± 0.07 7.2 ± 0.03 Malic acid (g/L) 2.7 ± 0.06 2.7 ± 0.07 2.6 ± 0.0

Table 3.2. Chemical composition of Pinot gris wine at the end of alcoholic fermentation (Late stage) ± SEM. Three levels of SO2 were added to the grape must at the cold-settling stage.

Chemical composition ± SEM Chemical parameter 0 mg/L SO2 20 mg/L SO2 40 mg/L SO2 Temperature (°C) 18.4 ± 0.1 18.7 ± 0.06 18.9 ± 0.09 pH 3.03 ± 0.006 2.97 ± 0.03 2.98 ± 0.03 Molecular SO2 (mg/L) 0.019 ± 0.01 0.025 ± 0.02 0.015 ± 0.008 Residual sugar (°Brix) -1.0 ± 0.1 -0.2 ± 0.7 -0.2 ± 0.1 Volatile acidity (g/L) 0.25 ± 0.009 0.26 ± 0.01 0.21 ± 0.02 Total acidity (g/L) 7.7 ± 0.06 8.1 ± 0.2 8.0 ± 0.2 Malic acid (g/L) 1.6 ± 0.09 1.8 ± 0.09 1.8 ± 0.1 Ethanol content (%) 12.1 ± 0.03 11.7 ± 0.3 12.0 ± 0.3

Temperature and residual sugar levels were measured using a portable

Anton Paar© density meter (Saint Laurent, QC, Canada). All other chemical

TM parameters (excluding SO2 determination) were measured using an OenoFoss wine analyzer (Foss, Hilleroed, Denmark) following manufacturer protocols. SO2 levels were determined using an aeration, oxidation, distillation, and titration

64 procedure described by Zoecklein et al. (1995). Briefly, SO2 in the wine samples was distilled using aspiration from an acidified solution (20 mL fermentation sample mixed with 10 mL 25% phosphoric acid) into a hydrogen peroxide trap (10 mL hydrogen peroxide, three drops of an indicator solution (Zoecklein et al., 1995), and one drop of 0.01 N sodium hydroxide). For free SO2 determination, the acidified wine sample was kept in an ice bath so only the volatile SO2 in the sample would be distilled. For total SO2 determination, the sample was heated in a water bath to release bound SO2. Samples were aerated for 15 minutes, and then the hydrogen peroxide sample was titrated by hand using 0.01 N sodium hydroxide.

3.2.3 Sample treatment with propidium monoazide

Samples for microbial analysis were taken at four stages of fermentation:

Cold-settling (> 22 °Brix), Early (14 - 19 °Brix), Mid (7 - 12 °Brix), and Late (< 2

°Brix). Samples were collected aseptically in sterile 50 mL plastic centrifuge tubes and were immediately transported to the laboratory on ice for processing. Aliquots of

10 mL from each sample were transferred to a sterile 15 mL plastic centrifuge tube.

Samples were centrifuged at 4,000 rpm for 5 minutes, and the supernatant was discarded. The pellet was then re-suspended in 10 mL molecular-grade water and centrifuged at 4,000 rpm for 3 minutes before discarding the supernatant. A cell membrane recovery step was added for samples taken at the Early, Mid, and Late stages of fermentation to ensure the presence of alcohol would not decrease the membrane integrity of living cells in these samples (Goldstein, 1986). In this step, the pellet was re-suspended in 10 mL freshly made, autoclaved YEPD broth (20% yeast extract, 10% peptone, 10% dextrose) and placed horizontally on ice for 2

65 hours. After 2 hours, the samples were centrifuged at 4,000 rpm for 3 minutes and the supernatant was discarded. The pellet was washed with molecular-grade water, centrifuged, and the supernatant was discarded. The remaining steps of the described protocol were then completed for all samples. Next, the pellet was re- suspended in 2.25 mL molecular-grade water by pipetting up and down 10 times. A

1.1 µL of 20 mM PMA aliquot was added to each sample to achieve a total concentration of approximately 6-7 mM (Tantikachornkiat et al., 2016). Samples were placed horizontally on ice and left in the dark for 10 minutes before being exposed to light (550 W halogen lamp) for 8 minutes on an oscillating platform.

Exposure to light allows the PMA to bind irreversibly to any exposed DNA in the sample, and results in bound DNA being excluded from amplification during PCR.

Samples were centrifuged, and the pellets were transferred to 2.0 mL microcentrifuge tubes. Samples were washed with 1.5 mL molecular-grade water twice and then stored at -20 °C until further processing.

3.2.4 DNA extraction and Illumina MiSeq library preparation

Total DNA from PMA-treated fermentation samples was extracted using an

Omega E.Z.N.A.® Stool DNA Kit (Omega Biotek, Norcross, GA, USA), following the manufacturer protocol with the following modifications: 200 mg of 0.1 mm glass disruptor beads (Scientific Industries, Inc., Bohemia, NY, USA) were used in place of the provided beads; samples were homogenized using a TissueLyser II mechanical bead beater (Qiagen®, Hilden, Germany). Extracted DNA samples were stored at -

20 °C until further processing. Sample library preparation for Illumina MiSeq

66 sequencing was conducted by a two-step PCR process consisting of ‘amplicon’ and

‘index’ PCR reactions, described below.

3.2.4.1 Amplicon PCR

For bacteria, the V3/V4 region of the 16S ribosomal ribonucleic acid (rRNA) gene was amplified using F341 and R805 primers (Herlemann et al., 2011) with CS1 and CS2 linker sequences on the forward and reverse primers, respectively. These overhanging linker sequences are necessary for the second PCR reaction, ‘index’

PCR, which attaches the Illumina MiSeq adapter sequence and unique 8 nucleotide barcode to each sample. Amplicon PCR was prepared using the following reaction mix: 6.8 µL molecular-grade water; 3.0 µL GoTaq reaction mix (5 buffer); 0.3 µL dNTP (10 mM each); 0.24 µL BSA (10 mg/mL); 0.45 µL forward and reverse primers each (10 µM); 1.5 µL MgCl2 (25 mM); 0.16 µL GoTaq DNA Polymerase (5 units/µL);

2.1 µL DNA template, for a total reaction volume of 15 µL per well. Amplicon PCR for bacteria was performed using the following program: 95 °C for 3 minutes (1 cycle); 95 °C for 40 sec, 53 °C for 40 sec, 72 °C for 1 min (32 cycles); 72 °C for 7 minutes (1 cycle). All PCR reactions were performed on an Applied Biosystems

Veriti 96-Well Fast Thermal Cycler (Foster City, CA, USA).

For fungi, the ITS1 region of the rRNA gene was amplified using BITS and

B58S3 primers (Bokulich and Mills, 2013) with CS1 and CS2 linker sequences on the forward and reverse primers, respectively. Amplicon PCR was prepared using the following reaction mix: 4.67 µL molecular-grade water; 2.5 µL GoTaq Reaction

Mix (5 buffer); 0.25 µL dNTP (10 mM each); 0.2 µL BSA (10 mg/mL); 0.25 µL forward and reverse primer each (10 µM); 1.25 µL MgCl2 (25 mM); 0.13 µL GoTaq

67

DNA Polymerase (5 units/µL); 3.0 µL DNA template, for a total reaction volume of

12.5 µL per well. Amplicon PCR for fungi was performed using the following program: 95 °C for 2 minutes (1 cycle); 95 °C for 40 sec, 55 °C for 40 sec, 68 °C for

1 min (32 cycles); 68 °C for 5 minutes (1 cycle).

3.2.4.2 Index PCR

For both bacteria and fungi, a Gel Logic 400 Imaging System (Mandel,

Rochester, NY, USA) was used to visualize and confirm PCR amplification in a 1.5% agarose gel containing SYBRTM Safe DNA gel stain (Life Technologies, Carlsbad,

CA, USA). Depending on the strength of the band visualized on the gel, the amplicon products were diluted using molecular-grade water before Index PCR was performed. The Index PCR primers contained the CS1/CS2 linker sequence, an 8 nucleotide barcode sequence, and the P5/P7 Illumina adapter sequence. Each sample received a unique combination of forward and reverse primer barcodes.

Index PCR was prepared using the following reaction mix: 20.5 µL molecular-grade water; 10 µL GoTaq Reaction Mix (5 buffer); 1 µL dNTP (10 mM each); 1.5 µL BSA

(10 mg/mL); 2.5 µL forward and reverse primers each (2 µM); 9 µL MgCl2 (25 mM);

0.5 µL GoTaq DNA Polymerase (5 units/µL); 2.5 µL DNA template, for a total reaction volume of 50 µL per well. Index PCR was performed using the following program: 95 °C for 1 min (1 cycle); 95 °C for 30 sec, 62 °C for 30 sec, 68 °C for 1.5 min (12 cycles); 68 °C for 5 min (1 cycle). The PCR products were visualized in a

1.5% agarose gel containing SYBRTM Safe DNA gel stain. Index PCR was deemed successful when the amplicon length was extended by 69 bp (Illumina MiSeq P5/P7 adapter + barcode length) as compared to the Amplicon PCR product length.

68

Samples were submitted to the IBEST Genomics Resources Core facility at the University of Idaho (Moscow, ID, USA) for quantification, normalization, pooling, and sequencing. Paired-end sequencing (300 bp length) was performed on an

Illumina MiSeq Desktop Sequencer (Illumina® Inc., San Diego, CA, USA).

3.2.5 Illumina MiSeq data processing

The open-source bioinformatics pipeline Quantitative Insights Into Microbial

Ecology (QIIME1 and QIIME2) was used for identifying the microbial communities using gene sequencing data (Bolyen et al., 2018; Caporaso et al., 2010). Forward and reverse barcode files were combined using the “extract_barcodes.py” function in

QIIME1 (version 1.8). All other processing of bacterial and fungal sequences was performed in QIIME2 q2cli version 2017.12.

For bacterial sequence data, “demux” was used to demultiplex samples

(https://github.com/qiime2/q2-demux, accessed 2019-05-27). DADA2 (using the

“data2 denoise-single” command) was used to de-noise and correct Illumina- sequenced amplicon errors in forward-read sequences only (Callahan et al., 2016).

Sequences were truncated at 260 bp. Paired-end assembly was not conducted because reverse sequence reads were determined to be low quality based on observations from the “demux summarize” command. The “feature-table” plug-in was used to create a feature table and a table of representative sequences

(McDonald et al., 2012). A Naïve Bayes classifier was trained on the target region of the amplicon primer sets using Greengenes software (version 13.5) (DeSantis et al.,

2006; McDonald et al., 2012) using the “q2-feature-classifier” plugin

(https://github.com/qiime2/q2-feature-classifier, accessed 2019-05-27).

69 was assigned to the genus level. MAFFT-aligned sequences (Katoh and Standley,

2013) were used to produce a phylogenetic tree using FastTree software (version

2.1.7) (Price et al., 2009). The “q2-taxa” plug-in (https://github.com/qiime2/q2-taxa, accessed 2019-05-27) was used to filter out mitochondrial and chloroplast contamination, and to remove sequence variants that could not be identified to the order level or lower as well as those that appeared with a total frequency of less than 100. Samples were then rarefied to a sampling depth of 10,000 sequences

(based on the sample with the lowest number of sequences) before being exported from QIIME2 for further analysis.

For the fungal sequence data, a similar protocol was used as the one described above, with the following changes. Paired-end assembly of fungal sequences was performed; forward reads were truncated at 206 bp and reverse reads were truncated at 180 bp when performing the de-noising step using DADA2

(Callahan et al., 2016). Sequences were classified to the species-level using a 99% threshold classifier created using UNITE software (version 7.2) (Kõljalg et al., 2013).

The “q2-taxa” plug-in (https://github.com/qiime2/q2-taxa, accessed 2019-05-27) was used to remove sequence variants that could not be identified to the order level or lower as well as those that appeared with a total frequency of less than 100.

Samples were then rarefied to a sampling depth of 6,000 sequences (based on the sample with the lowest number of sequences) before being exported from QIIME2 for further analysis. In one instance, the name of the identified fungal organism was changed from the classification given by UNITE; sequences identified as

Saccharomyces bayanus are termed Saccharomyces uvarum in this study.

70

According to the UNITE database (version 7.2), S. bayanus and S. uvarum are synonymous, but in recent years S. uvarum has been re-instated as a separate species (Nguyen and Gaillardin, 2005). We are confident in the assessment that sequences from this research belong to S. uvarum based on culture-dependent sequencing data from these fermentations (data not included here).

3.2.6 Wine bottling and sensory analysis

Post-alcoholic fermentation, 6 L of wine was transferred from each barrel into

2 L glass bottles with screw caps (three bottles per barrel). The bottles were purchased new and cleaned with a 5% citric acid and KMS solution. Long plastic tubing was cleaned with the same solution and was used to rack the wine from the barrels to the bottles. Sulfur dioxide (60 mg/L SO2, added as 120 mg/L KMS) was added to each bottle to act as a preservative. The glass bottles were filled to the top with wine and a carbon dioxide/nitrogen (CO2/N2) gas mixture was added to the headspace before the lid was screwed on to remove any . These bottles were kept in the dark in a cooled cellar for approximately two months, with after one month, before being transferred to clean 750 mL wine bottles with screw caps supplied by the winery. The three 2 L glass bottles from each wine barrel were combined into a 19 L glass carboy cleaned with a 5% citric acid and KMS solution.

From the carboy, six 750 mL wine bottles were filled from each barrel replicate (for a total of 54 wine bottles from nine barrels). Wine was filled to the neck of the bottle and N2 gas was added to the headspace to remove any oxygen. Free and total SO2 were measured at this stage and it was determined that each bottle had between 7.4 and 14.3 mg/L free SO2 at the time of bottling.

71

Sensory evaluations of the wines were conducted in the spring of 2015 at the

Summerland Research and Development Centre (Summerland RDC) in

Summerland, British Columbia, Canada, adhering to standard sensory evaluation protocols and using a panel of 10 industry wine experts (Cliff and Dever, 1996;

Guinard and Cliff, 1987; King et al., 2013). Each wine was evaluated in duplicate in a blind study. The wines were labeled with three-digit random numbers, and served in random order using a William’s design, to control for first-order bias and carryover effects (Williams and Arnold, 1991). A short break of 5-10 minutes was taken between tasting sessions. Wine was served in 210 mL International Standards

Organization (ISO) wine glasses in 30 mL aliquots. The judges first evaluated the perceived aroma of the wines, followed by the flavour and mouthfeel attributes, using sensory characteristics defined during a prior roundtable discussion (Table

3.3). Sensory standards for panelist reference were created for each of the attributes listed in Table 3.3, the composition of which can be found in Appendix B (Table B1).

All wine samples were expectorated and panelists were instructed to rinse their mouths with water between wines. The intensity of each attribute was evaluated on an unstructured 100 unit line scale, using Compusense five© sensory evaluation software (Compusense, Inc, Guelph, ON, Canada). Evaluations were conducted individually in separate booths lit by red lighting.

Since a preliminary bench-testing revealed that no colour differences were present among the wines, colour assessments were not conducted. The use of human subjects in this study was approved by the Agriculture and Agri-Food

72

Canada (AAFC) Human Research Ethics Committee (Certificate of Approval

2015D004) and the UBC Okanagan Research Ethics Board.

Table 3.3. Sensory attributes of wines evaluated by an expert panel of 10 judges.

Aroma attributes Flavour attributes Other attributes Citrus aroma Citrus flavour Acidity Pome aroma Pome flavour Astringency Tropical fruit aroma Tropical fruit flavour Body Spice aroma Spice flavour Length of aftertaste Vanilla aroma Vanilla flavour Toasty/smoky aroma Toasty/smoky flavour Wood aroma Wood flavour

3.2.7 Statistical analysis

Wine chemical parameters were analyzed with RStudio software (version

3.4.4) and reported ± the standard error of the mean (SEM). Sulfur dioxide measurements at crush were compared by performing a one-factor analysis of variance (ANOVA) on total SO2 measurements using the “aov” function. Normality was assessed visually, and homogeneity of variance was assessed using the

“leveneTest” function from the car package (version 3.0-0), which indicated no significant differences among treatments (F(1,4) = 0.36, p = 0.58). Fermentation progression (°Brix throughout fermentation) was compared among the three SO2 treatments by performing a repeated-measures one-factor ANOVA on rank- transformed data using the “aov” function.

Fungal and bacterial diversity were analyzed separately in RStudio (version

3.4.4). Simpson’s Index of Diversity was calculated using the “diversity” function in the vegan package (version 2.5-1) and reported ± SEM. Diversity was also 73 evaluated for among treatments (Cold-settling, Early, Mid, Late) for each of the SO2 treatments, by performing a repeated-measures one-factor ANOVA using the “aov” function. Normality was assessed visually and Levene’s test indicated no violation of the assumption of heterogeneity in variance for either fungal diversity (F(2,6) = 2.60, p = 0.09) or bacterial diversity (F(2,6) = 2.40, p = 0.10). When appropriate, a Tukey post-hoc test, adjusted for multiple comparisons using the Holm method, was performed to evaluate differences among treatments using the “lme” and “glht” functions in the nmle (version 3.1-137) and multcomp (version 1.4-8) packages, respectively (Hothorn et al., 2008). Fungal and bacterial relative abundance was visualized by creating stacked bar charts using GraphPad Prism® software (version

7) (La Jolla, CA, USA). Fungi or bacteria that represented fewer than 100 sequences in every sample were grouped together and termed “minor fungi” or

“minor bacteria.”

Fungal and bacterial composition data were analyzed separately in RStudio

(version 3.4.4). A Bray-Curtis dissimilarity index was calculated from untransformed fungal/bacterial abundance for each sample. Multivariate homogeneity of group dispersions (PERMDISP) was analyzed utilizing the “betadisper” and “permutest” functions (vegan package) using Bray-Curtis dissimilarity and calculating deviation from centroid. Permutational analysis of variance (PERMANOVA) tests, using Bray-

Curtis dissimilarity, were performed on fungal and bacterial composition data to test for differences in composition among treatments, using the “adonis” function (vegan package). When appropriate, pairwise post-hoc tests were performed, adjusted for

74 multiple comparisons with the Holm method and using the “pairwise.adonis” function in the pairwiseAdonis package (version 0.0.1).

When analyzing fungal community composition, the PERMDISP test indicated significant differences among treatments (F(2,24) = 3.50, p = 0.04).

However, PERMANOVA tests are robust to unequal variances among treatments

(Anderson and Walsh, 2013), so no data transformations were made prior to performing the PERMANOVA test. Fungal community composition was initially analyzed using data from all four stages of fermentation, but the communities of the

Cold-settling stage were extremely similar among all treatments and were also dramatically different from the communities present during alcoholic fermentation

(Early, Mid, and Late stages). Therefore, this stage was removed from the analysis to more accurately assess the effect of SO2 addition during fermentation. For the bacterial community composition, a PERMDISP test indicated no significant differences among treatments (F(2,33) = 0.68, p = 0.50) with regards to multivariate dispersion. Because the bacterial communities were consistent throughout fermentation and did not change from the Cold-settling stage to the alcoholic fermentation stages (as the fungal communities did), samples from all four stages were included in the analysis of bacterial community composition.

Test statistics (F values for PERMDISP and Pseudo-F values for

PERMANOVA) were calculated based on 999 permutations of raw data. The design of this study involved repeated measures, and therefore not all data were independent of one another, potentially leading to an overestimation of treatment differences as a result of the PERMANOVA tests. Therefore, principal coordinates

75 analysis (PCoA) ordinations were used to visualize distances between samples of different treatments (Knight et al., 2015; Lorion and Kennedy, 2009). These PCoA ordinations were created using the “wcmdscale” and “ordihull” functions using Bray-

Curtis dissimilarity (vegan package).

Statistical analysis of sensory evaluation data was performed in RStudio

(version 3.4.4). The “panelperf” function in the SensoMineR package (version 1.23) was used to evaluate the sensory panel’s performance in its ability to discriminate among products (wines), using three-factor ANOVA. A principal component analysis

(PCA) was performed using the “averagetable” and “pca” functions in order to visualize the relationship between the sensory attributes and wines from different

SO2 treatments. A radar plot was generated in Excel 2016 to visualize sensory differences among the wines from different SO2 treatments; it was converted to a high-resolution image using IrfanView software (version 4.51). Values for the radar plot were standardized for each sensory attribute grouping in Excel using the

“STANDARDIZE” function, based on the mean and standard deviation of a set of data. Standardizing involves re-scaling each variable to have a mean of zero and a standard deviation of one. This can improve the visualization of differences among treatments for each attribute, and enable comparisons of data.

3.3. Results and discussion

3.3.1 Sulfur dioxide and wine chemistry

The chemical composition of the Pinot gris must at the Cold-settling stage

(Table 3.1) and at the end of alcoholic fermentation (Table 3.2) was similar for all treatments and was within an expected range for cool-climate Pinot gris wines.

76

Alcoholic fermentation progressed at a steady rate for all treatments, and although the fermentations to which 0 mg/L SO2 was added at crush completed one day before the other two treatments, the difference in fermentation rate among treatments was non-significant (F(2,6) = 3.90, p = 0.08) (Figure 3.1). This result is in accordance with the findings of Henick-Kling et al. (1998), who also found no differences in fermentation rate in response to SO2 addition.

Figure 3.1. Residual sugar concentrations (°Brix ± SEM) measured throughout AF of Pinot gris wines to which three different levels of sulfur dioxide had been added at crush (n = 3 per treatment). Fermentations from all treatments were complete within 11 days. A one-factor repeated measures ANOVA was performed, and no significant differences among treatments were found (F(2,6) = 3.90, p = 0.08).

The pH of the grape must was an average of 3.26 across all treatments at the

Cold-settling stage (Table 3.1). By the end of alcoholic fermentation, the average pH of the wines had dropped to just under 3.0 (Table 3.2). The grapes used for this study were grown in a cool-climate wine region towards the northern-most limit for wine grape production, and the pH observed here is not unexpected for white wine

77 grapes from such regions (Knoll et al., 2012). However, we note that low pH musts can mitigate the risk of contamination by spoilage yeasts and bacteria, especially when conducting uninoculated and sulfite-free fermentations, which are at an inherently higher risk for stuck or spoiled fermentations. Bacteria, including lactic acid and acetic acid bacteria, have trouble growing in musts with low pH (Bartowsky,

2009). High pH can promote the growth of bacteria such as Pediococcus spp. and can also volatilize nitrogen-heterocyclic compounds, produced by heterofermentative lactic acid bacteria, which can give wines a mousy character

(Bartowsky, 2009; Costello et al., 2001; Costello and Henschke, 2002). When dealing with musts of higher pH, wineries utilizing low-intervention winemaking techniques such as uninoculated or sulfite-free fermentations may experience an increased risk of microbial spoilage.

Both free and total SO2 were measured throughout alcoholic fermentation

(Figure 3.2). Interestingly, the initial total SO2 measurements made for the 20 mg/L

SO2 and 40 mg/L SO2 treatments were slightly lower than expected based on the level of addition; however, no significant difference was found between the two treatments in terms of expected versus observed total SO2 (one-factor ANOVA,

F(1,4) = 0.36, p = 0.58), indicating that both treatments had total SO2 levels that were lower than expected by the same amount. Therefore, it was likely that natural

SO2 loss and not SO2 addition or sampling error led to this discrepancy. Between the time of sampling and the time of SO2 determination, several hours had elapsed in the case of some samples, which could have contributed to SO2 loss. Samples

78 were also collected in plastic centrifuge tubes, which could have adsorbed some volatile compounds, including free SO2.

Figure 3.2. Free (A) and total (B) sulfur dioxide (SO2) concentrations ± SEM measured throughout AF of Pinot gris wines to which three different levels of total SO2 had been added at crush (n = 3 per treatment). SO2 was added in the form of potassium metabisulfite (KMS) immediately prior to the sample taken at the Cold- settling stage of fermentation.

While total SO2 is useful for confirming the initial concentration of SO2 added, free SO2 is a more useful measurement when studying the relationship between SO2 addition and microbiological composition. Free SO2, in combination with pH 79 measurements, can be used to determine the molecular SO2 concentration using the Henderson-Hasselbalch equation (Divol et al., 2012; Zoecklein et al., 1995), which is the most effective form of SO2 in terms of its antimicrobial activity (Divol et al., 2012). An effective concentration of molecular SO2 for antimicrobial use has been estimated at between 0.2 mg/L and 0.8 mg/L, depending on the publication, owing to the susceptibility of each particular must to spoilage based on chemical profiles such as pH, temperature and microbial load (Howe et al., 2018; King et al.,

1981; Margalit and Crum, 2004). Both sulfited treatments (20 mg/L SO2 and 40 mg/L

SO2) achieved a minimum effective concentration at the Cold-settling stage (Table

3.1). The concentration of free SO2 at cold-settling was approximately half the concentration of total SO2 for the two treatments to which SO2 was added (Figure

3.2). However, by the Early stage of fermentation, free SO2 dropped to negligible amounts in all treatments and remained very low throughout the rest of fermentation.

Total SO2, however, dropped sharply after cold-settling in the two sulfited treatments

(due to the loss of free SO2), but then increased between the Early and Mid stages of fermentation in all three treatments. This was likely due to the production of SO2 by S. cerevisiae yeasts during fermentation, which primarily produce bound SO2

(Andorrà et al., 2017; Suzzi et al., 1985). Total yeast counts, based on culture methods, were highest in the 40 mg/L SO2 treatment and lowest in the 0 mg/L SO2 treatment (data not shown), which may explain the different amounts of bound SO2 produced among the treatments during alcoholic fermentation.

Sulfur dioxide is added prior to the onset of alcoholic fermentation (and prior to the addition of commercial yeasts in inoculated fermentations) for the removal of

80 potential spoilage microorganisms coming in from the vineyard (Divol et al., 2012).

Bacteria are particularly sensitive to SO2, and previous research has noted that non-

Saccharomyces yeasts are more SO2-sensitive than S. cerevisiae yeasts (Constantí et al., 1998; Divol et al., 2012; Henick-Kling et al., 1998). Therefore, SO2 resistance is likely to play a much larger role in the ability of vineyard microorganisms to persist during fermentation than it is for strains of winery-resident yeasts such as S. cerevisiae. Because free (and molecular) SO2 decreases to a very low level soon after its addition, S. cerevisiae strains that subsequently enter a fermentation, whether intentionally or not, are likely less affected by the additions of SO2. For these winery-resident yeast strains, it is likely to be other factors, such as growth rate, range of nutrient consumption, and other competitive abilities such as the production of killer toxins that determine whether these yeasts are able to persist in a fermentation. In Chapter 2, we investigated the S. cerevisiae strains present in the fermentations from this study, and identified no patterns between SO2 resistance and yeast strain dominance in any SO2 treatment.

3.3.2 Fungal and bacterial diversity during fermentation

Fungal community diversity throughout fermentation was significantly lower in the 40 mg/L SO2 treatment than in the 0 and 20 mg/L SO2 treatments (Table 3.4). At the Cold-settling stage, all treatments contained a similar fungal diversity, as was expected given the samples were taken directly after the SO2 treatment had been applied, leaving little time for the SO2 addition to alter the fungal community.

Diversity decreased in all treatments throughout fermentation, but most drastically decreased in the 40 mg/L SO2 treatment. This was likely due to the higher level of

81

SO2 addition at crush that removed all but the most resistant vineyard yeasts.

Alcohol tolerance of the vineyard yeasts likely did not play a role in the observed decrease in diversity in the 40 mg/L SO2 treatment, because the wines from all three treatments finished alcoholic fermentation (< 0 °Brix at the time of the last sample), and reached a similar ethanol concentration (Table 3.2). While limited research is available on the resistance of non-Saccharomyces yeasts to increasing ethanol concentrations using current identification technologies (however, see Wang et al.,

2015), the previous literature emphasizes S. cerevisiae as the most important yeast in winemaking at least in part due to its high ethanol tolerance (Bisson and Joseph,

2009).

82

Table 3.4. Fungal and bacterial diversity, measured as Simpson’s Index of Diversity, of Pinot gris wines to which different levels of SO2 were added at crush (n = 3 per treatment). Diversity was measured at four stages of alcoholic fermentation (Cold- settling, Early, Mid, Late) and reported ± SEM. Fungal and bacterial data were analyzed separately, and treatments were compared across all fermentation stages by performing repeated measures one-factor ANOVA and subsequent post-hoc tests when appropriate. An asterisk next to either fungal or bacterial diversity indicates significance at α = 0.05; treatments marked with different superscript letters had significantly different overall diversity across all stages of fermentation (α = 0.05).

Simpson’s Index of Diversity ± SEM a a b Fungal Diversity* 0 mg/L SO2 20 mg/L SO2 40 mg/L SO2 Cold-settling 0.59 ± 0.02 0.57 ± 0.005 0.60 ± 0.01 Early 0.42 ± 0.03 0.48 ± 0.03 0.30 ± 0.08 Mid 0.48 ± 0.02 0.21 ± 0.02 0.11 ± 0.03 Late 0.43 ± 0.03 0.35 ± 0.05 0.16 ± 0.03

Bacterial Diversity 0 mg/L SO2 20 mg/L SO2 40 mg/L SO2 Cold-settling 0.46 ± 0.04 0.52 ± 0.06 0.46 ± 0.07 Early 0.52 ± 0.05 0.48 ± 0.03 0.52 ± 0.05 Mid 0.47 ± 0.04 0.50 ± 0.05 0.44 ± 0.04 Late 0.53 ± 0.05 0.47 ± 0.03 0.50 ± 0.06

Bacterial community diversity throughout fermentation seemed to be unaffected by either the SO2 treatment applied at crush or the stage of alcoholic fermentation (Table 3.4). Because PMA was added to the fermentation samples before DNA extraction, the fungi and bacteria identified represented those that were living at the time of sampling. Therefore, this precluded the possibility that the bacterial diversity remained unchanged throughout fermentation due to only dead bacteria being present in the sample (either due to SO2 addition, being out- competed, or susceptibility to alcohol). This result is in contrast with the results obtained by Bokulich et al. (2014), who found decreased bacterial alpha-diversity in 83 uninoculated, sulfite-free fermentations as compared to uninoculated, sulfited fermentations, likely caused by the overgrowth of a few dominant bacterial species.

Sun et al. (2016) also noted a decrease in bacterial abundance in response to increasing SO2 concentrations in strawberry wine fermentations, although this study involved much higher levels of SO2 addition and used culture-dependent identification techniques.

3.3.3 Fungal and bacterial abundance during fermentation

At the Cold-settling stage, all SO2 treatments contained a variety of different fungi (Figure 3.3; Table B3), most notably Aureobasidium pullulans, Cladosporium sp., and a small amount of Hanseniaspora sp. A. pullulans is a ubiquitous yeast-like fungus that is commonly associated with vineyard and winery environments, and has been found throughout the winemaking process (Takahashi et al., 2014; Varela and Borneman, 2017). An attempt was made to identify all fungal taxa to the species-level using the UNITE database (version 7.2), although this was not always possible. In cases where species-level identification could not be obtained, the most specific taxonomic classification available was assigned.

84

Figure 3.3. Relative abundance of fungi (based on 6,000 sequences per sample) present in wines fermented with three levels of sulfur dioxide (SO2) added at crush (n = 3 per treatment). Samples were taken at four stages of alcoholic fermentation. For variation among samples of the same treatment and for the identities of minor fungi, please refer to Table B3.

We note the presence in all treatments of low levels of certain fungi with the potential to produce mycotoxins: namely, Aspergillus flavus and Penicillium spp.

Both genera of fungi are commonly found on wine grapes and other food crops

(Freire et al., 2017). Mycotoxins produced by these fungi include aflatoxin B1 and ochratoxin A, among others, which can present a health risk to consumers if present in a food product in high enough concentrations (Inoue et al., 2013; Mateo et al.,

2007). Low levels of ochratoxin A (< 4.5 ng/mL) are common in commercially produced wines (Mateo et al., 2007). However, Canadian wines have been shown to contain a lower occurrence of ochratoxin A, along with lower levels of contamination, than imported wines sold in Canada (Soleas et al., 2001). In the study conducted by

Soleas et al. (2001), which included 96 Canadian wines, only three red wines and none of the white wines contained > 0.05 µg/L ochratoxin A. The process of 85 fermentation itself has been shown to reduce the presence of mycotoxins in wine, either through conversion of the mycotoxin to a less toxic form (Inoue et al., 2013) or possibly through their adsorption to yeasts/ and eventual removal from the wine

(Petruzzi et al., 2014). In total, Aspergillus and Penicillium spp. made up approximately 5% of the total fungal community at the Cold-settling stage, and less than 0.2% of the total fungal community during alcoholic fermentation (Table B3).

The distribution of these fungi were even across all treatments, indicating that the lack of SO2 added to the 0 mg/L SO2 treatments did not contribute to the presence or abundance of these fungi. Although mycotoxin levels were not measured for this study, we do not anticipate mycotoxin production to be of concern here.

After Cold-settling, the three stages of alcoholic fermentation (Early, Mid, and

Late) contained a consistent assemblage of fungi within each treatment, although the relative abundance of major yeasts differed markedly among SO2 treatments

(Figure 3.3). In all treatments, S. cerevisiae dominated the fermentations. This was an expected result, as even in uninoculated fermentations, winery-resident S. cerevisiae strains are known to dominate in most commercial settings, especially in the Okanagan winemaking region of Canada (Hall et al., 2011; Morgan et al., 2017;

Scholl et al., 2016). The quick succession from vineyard yeasts during the Cold- settling stage to the dominance of S. cerevisiae from the Early stage of fermentation onwards suggests that an abundance of winery-resident strains were present in the environment and were able to quickly enter the musts and outcompete the remaining vineyard yeasts. Indeed, at the winery in question where this research was conducted, most of the fermentation vessels are inoculated with commercial

86 yeast strains, and more than 40 different commercial strains had been used over the five years leading up to this study (Morgan et al., 2017). An analysis of the strain- level composition of S. cerevisiae strains in these fermentations was previously discussed in Chapter 2, where we found that approximately 98% of the strains identified were of commercial origin, regardless of SO2 addition level at crush.

Saccharomyces uvarum was also observed in all treatments in low concentrations. This yeast is not typically used as a commercial starter, and only one commercial strain, Lallemand Velluto BMV58TM, was available at the time of publishing. However, S. uvarum has been identified in wineries around the globe, generally in cool-climate winemaking regions; even when musts are uninoculated, this yeast can sometimes dominate fermentations and out-compete winery-resident

S. cerevisiae (Demuyter et al., 2004; Naumov et al., 2002, 2000). The third yeast of note is Hanseniaspora sp. (not identified to the species-level), which appeared in all fermentations but with striking differences in relative abundance between treatments. In the 0 mg/L SO2 treatment, Hanseniaspora represented up to 20% of the total fungal community, and maintained this abundance through the end of fermentation. The abundance of Hanseniaspora decreased with increasing initial

SO2 addition, and in the 40 mg/L SO2 treatment it represented no more than 5% of the total fungal community. As mentioned above, the yeasts and fungi identified represent those that were living at the time of sampling. The Hanseniaspora yeasts identified at the Late stage of fermentation (Figure 3.3) were therefore living, even at approximately 12% (v/v) alcohol. Hanseniaspora yeasts are uncommonly identified at these late stages of fermentation using culture-dependent identification

87 techniques such as plating and incubating (Combina et al., 2005; Constantí et al.,

1997; Ocón et al., 2010; Shütz and Gafner, 1993). Before the advent of culture- independent molecular technologies such as Illumina sequencing (among others), it was assumed that these vineyard yeasts could not tolerate high levels of alcohol.

Our results, however, are in accordance with results from other studies conducted using high-throughput amplicon sequencing technologies, which have identified vineyard yeasts such as Hanseniaspora spp. at late stages of alcoholic fermentation

(Bokulich et al., 2014; Pateraki et al., 2014; Stefanini et al., 2016a). These yeasts may be in a viable but nonculturable (VBNC) state, potentially induced into this state by metabolites produced by S. cerevisiae, which is why they are not identified using plating and culturing methods (Wang et al., 2016, 2015b). There is evidence that other yeasts such as Brettanomyces bruxellensis can be induced into a VBNC state by SO2 (Capozzi et al., 2016), but within the context of our results, it seems that SO2 at higher levels was effective in eliminating much of the Hanseniaspora population

(Figure 3.3). Even in a VBNC state, these yeasts are metabolically active and able to contribute to the final chemical and sensory profiles of the wines.

The bacterial community remained consistent throughout alcoholic fermentation, and was unaffected by either SO2 addition level or the stage of fermentation. Enterococcus was the most commonly identified genus, representing approximately 70% of the bacterial community, followed by a member of the

Bacillaceae family, representing approximately 15% of the bacterial community

(Figure 3.4, Table B4). Both Enterococcus and Bacillus spp. bacteria produce lactic acid and have been previously identified in wine fermentations, although

88

Enterococcus spp. have only been previously identified during malolactic fermentation in red wines (Bokulich et al., 2012; Capozzi et al., 2011; Dündar, 2016;

Pérez-Martín et al., 2014). Enterococci in particular are noted to be resistant to a wide range of environmental stresses, including elevated ethanol concentration

(Capozzi et al., 2011; Giraffa, 2002). Some strains of Enterococcus faecium isolated from have been shown to produce bacteriocins, toxins produced by one bacterium to inhibit the growth of closely related bacteria (Dündar, 2016).

Enterococci are potential spoilage bacteria due to their ability to produce tyramine, a vasoactive amine responsible for headaches in some people, and likely enter fermentations due to contamination via or winery equipment (Dündar, 2016).

Although traditionally associated with the mammalian gastrointestinal tract, these bacteria are quite ubiquitous and have been isolated from soil, water, and fermented foods, as well as on plants, vegetables, and fruits (Giraffa, 2002). It should be noted that food-borne Enterococcus spp. have not been implicated in any clinical

Enterococcus infections (Giraffa, 2002; Pérez-Martín et al., 2014).

89

Figure 3.4. Relative abundance of bacteria (based on 10,000 sequences per sample) present in wines fermented with three levels of sulfur dioxide (SO2) added at crush (n = 3 per treatment). Samples were taken at four stages of alcoholic fermentation. For variation among samples of the same treatment and for the identities of minor bacteria, please refer to Table B4.

Notably absent from the bacterial communities of these wines are common members of the lactic acid bacteria group (Table B4). Specifically, sequences identified to the genera Lactobacillus and Leuconostoc as well as the order

Lactobacillales were present in these fermentations, but due to their extremely low abundance and/or lack of specificity in taxonomic identification, they were filtered out before analysis (see Materials and methods section 3.2.5). Lactobacillus spp. do not grow well at low pH, and the addition of SO2 can delay their growth (Edwards et al.,

1993); both of these factors may have contributed to their low abundance and eventual removal from this study.

90

3.3.4 Fungal and bacterial community composition

Fungal community composition was affected by SO2 addition levels (Table

3.5), with each treatment having a significantly different composition (P ≤ 0.03 for all pairwise comparisons). Samples from the 40 mg/L SO2 treatment contained unique assemblages of fungi, while the 0 mg/L SO2 and 20 mg/L SO2 treatments contained some composition overlap among samples (Figure 3.5). Composition takes into account not only the relative abundance but also the identity of different organisms.

Therefore, samples that contain similar fungal species but with different relative abundance appear as separate points on the PCoA ordination, and the more dissimilar the identity and relative abundance of these fungi, the more distant two points will be from each other. We found a similar result in Chapter 2 when assessing the S. cerevisiae strain composition of these same fermentations; each

SO2 treatment contained a unique assemblage of yeast strains and significantly different population composition. Henick-Kling et al. (1998) and Takahashi et al.

(2014) noted differences in yeast community composition in response to SO2 addition levels, which is in accordance with the results reported here.

91

Table 3.5. Results of one-factor PERMANOVA tests evaluating the effects of sulfur dioxide (SO2) addition level at crush (0, 20, or 40 mg/L SO2, n = 3 per treatment) on fungal and bacterial composition throughout fermentations of uninoculated Pinot gris wines. Fungal composition was compared using three stages of fermentation (Early, Mid, Late) and bacterial composition was compared using four stages of fermentation (Cold-settling, Early, Mid, Late). Results are based on 999 unrestricted permutations of raw data. Statistical analysis was conducted separately for fungi and bacteria. Results marked with an asterisk are significant at α = 0.05.

Fungal Composition df SS MS Pseudo-F p SO2 Treatment 2 0.23807 0.119033 16.298* 0.001* Residual 24 0.17529 0.007304 Total 26 0.41335 Bacterial df SS MS Pseudo-F p Composition SO2 Treatment 2 0.00020 0.0000999 0.015358 0.992 Residual 33 0.21457 0.0065022 Total 35 0.21477

92

Figure 3.5. Principal coordinates analysis (PCoA) ordination representing the fungal community composition of wines fermented with three levels of sulfur dioxide (SO2) added at crush: 0 mg/L SO2 (black), 20 mg/L SO2 (dark grey), or 40 mg/L SO2 (light grey) (n = 3 per treatment). Dimension 1 (Dim 1) explains 49.3% of total variation and Dimension 2 (Dim 2) explains 6.0% of variation. Samples were taken at three stages of alcoholic fermentation, for a total of nine samples per treatment. One- factor PERMANOVA tests indicated significant differences in fungal community composition among treatments (P ≤ 0.03 for all comparisons).

No significant differences in bacterial community composition were found between treatments (Table 3.5, F(2,33) = 0.02, p = 0.99), and there was significant overlap between treatments in the PCoA ordination (Figure 3.6). Interestingly, the only other study to evaluate fungal and bacterial communities using the same

Illumina sequencing platform noted differences in the bacterial communities, but not

93 in the fungal communities, in response to a gradient of SO2 concentrations (Bokulich et al., 2014). This discrepancy is possibly due to the fact that these fermentations were conducted outside of a commercial winery environment, and had different environmental factors influencing the progression of fermentation. It is likely that the use of PMA in this study also contributed to these observed differences. Because this research evaluated only relative abundance and not total abundance of bacteria and fungi, it is possible that the bacterial communities in these fermentations were present in low abundance and had low metabolic activity during fermentation. It is also known that yeast activity can affect the behavior of bacteria during both alcoholic and malolactic fermentation (Reguant et al., 2005). Therefore, it is possible that the differences in yeast communities between this study and that of Bokulich et al. (2014) resulted in different interactions with the bacterial communities, leading to the observed differences in bacterial community performance.

94

Figure 3.6. Principal coordinates analysis (PCoA) ordination representing the bacterial community composition of wines fermented with three levels of sulfur dioxide (SO2) added at crush: 0 mg/L SO2 (black), 20 mg/L SO2 (dark grey), or 40 mg/L SO2 (light grey) (n = 3 per treatment). Dimension 1 (Dim 1) explains 57.7% of total variation and Dimension 2 (Dim 2) explains 4.2% of variation. Samples were taken at four stages of alcoholic fermentation, for a total of 12 samples per treatment. A one-factor PERMANOVA test indicated no significant differences in bacterial community composition among treatments (F(2,33) = 0.02, p = 0.99).

3.3.5 Wine sensory attributes

A sensory panel of 10 wine experts found significant differences among treatments in two of the 18 sensory attributes evaluated⎯ citrus aroma and pome fruit flavour (Table B2); these two attributes are among the most commonly identified attributes in cool-climate white wines, including Pinot gris. Wines from the 0 mg/L

95

SO2 treatment were noted to be higher in citrus aroma than the wines from the 20 mg/L SO2 and 40 mg/L SO2 treatments (Figure 3.7). Pome fruit flavour was associated with both the 0 mg/L SO2 and 20 mg/L SO2 treatments (and especially with the 20 mg/L SO2 treatment) but not with the 40 mg/L SO2 treatment. Although not significant, there was a trend towards a more fruit-forward expression (citrus, pome, and tropical fruits) in the wines from the 0 mg/L SO2 treatment, with fruity attributes being less dominant in the 40 mg/L SO2 wines, and somewhere in between for the 20 mg/L SO2 wines (Figures 3.7 & 3.8). In this study, no wine defects (sensorial or visual) were detected in wines from any treatment.

Figure 3.7. Radar plot depicting the normalized relative intensity of sensory attributes of Pinot gris wines fermented with three different levels of sulfur dioxide (SO2) added at crush. Radar plots were created using the average of three replicate wines per SO2 treatment, evaluated in duplicate by a panel of 10 wine experts. Values were standardized separately for each sensory attribute. For visual clarity, variation among treatments for each attribute was not included.

96

Figure 3.8. Principal component analysis (PCA) ordination of the sensory profiles of wines fermented with three different levels of sulfur dioxide (SO2) added at crush (n = 3 per treatment). Plots depict: (A) the variables factor map, and (B) the individuals factor map. Sensory attributes were analyzed by a panel of 10 wine experts. Attributes ending with “A” indicate aroma attributes, and those ending in “F” indicate flavour attributes.

97

Because wine yeasts are known to differ in their production of secondary by- products, it is likely that the differences in fungal community composition contributed to the sensory differences observed among treatments, including those attributes that trended towards differences but were not significantly different in this study. In the Cold-settling stage, the dominant fungi were Aureobasidium pullulans and a member of the genus Cladosporium. A Hanseniaspora yeast appeared in all treatments and remained in the fermentations until the end of alcoholic fermentation, although its relative abundance was much higher in the fermentations to which low or no SO2 had been added (Figure 3.3). Non-Saccharomyces yeasts including

Hanseniaspora and A. pullulans are able to influence grape varietal aromas by producing several different enzymes such as β-glucosidases and proteases that can react with grape precursor compounds (Varela and Borneman, 2017).

Hanseniaspora spp., including H. uvarum, H. guilliermondii, H. vinae, and H. osmophila, are known to contribute significantly to a wine’s sensory profile, producing elevated levels of acetoin and ethyl acetate and decreased levels of higher alcohols (Jolly et al., 2014; Romano et al., 2003). Acetoin produces a creamy, buttery aroma at sufficient concentrations, and ethyl acetate can range from a pleasant and fruity aroma at lower concentrations to a solvent aroma when concentrations exceed its preference threshold (Liu et al., 2015). S. uvarum was present in all treatments in low abundance, and was able to survive until the end of fermentation. S. uvarum is a competent fermenter itself, and is characterized by its cryotolerance and its ability to produce elevated levels of glycerol, which can improve the mouthfeel of a wine (Magyar and Tóth, 2011). S. cerevisiae dominated

98 the fermentations from all SO2 treatments, likely making the most significant contributions to the wine sensory profiles. We previously showed in Chapter 2 that the S. cerevisiae strains in these fermentations were almost exclusively commercial strains used previously by the winery, and that different commercial strains were able to dominate the fermentations of different SO2 treatments. Other previous studies have also shown that there is significant variability in the ability of different S. cerevisiae strains to produce compounds such as isobutanol, acetaldehyde, n- propanol, and isoamyl alcohol (Fleet, 2003; Romano et al., 2003). Both H. uvarum and S. cerevisiae strains are also highly variable in their production of acetic acid, an undesirable by-product (Romano et al., 2003).

Our results correspond with those of other studies that evaluated sulfited and sulfite-free wines. Studies investigating SO2 alternatives such as ozone, gallic tannin and ascorbic acid, or hydroxytyrosol have also noted increased fruitiness in wines where no SO2 has been added, as compared with sulfited wines (Bellincontro et al.,

2017; Boroski et al., 2017; Raposo et al., 2016). It should be noted that the comparisons made in these studies were between SO2 and SO2 alternatives, rather than simply SO2 addition at different levels, as was investigated in this study.

Henick-Kling et al. (1998) found a decrease in fruitiness and an increase in undesirable off-flavours when wines were not sulfited, but this result was confounded by the influence of inoculated wine treatments, which scored highest in attributes such as sweaty, oxidized, and paper. When only uninoculated treatments were considered, the results more closely resembled those from this study, with increased fruitiness and overall flavour intensity corresponding to lower SO2 addition

99 levels. It has been suggested that this increase in fruitiness could be a result of the production of compounds such as ethyl acetate, ethyl decanoate, ethyl dodecanoate, or ethyl octanoate/ethyl hexanoate by wine yeasts (Liu et al., 2016), or simply a result of the ability of SO2 to neutralize aromas (Guerrero and Cantos-Villar,

2015; Ribereau-Gayon et al., 2006).

We suggest that future studies be conducted using a higher number of biological replicates and/or a more neutral fermentation vessel, as well as with a more aromatic wine varietal such as Riesling or , which would allow for better elucidation of sensorial differences among wines. Previous studies have shown that when using wooden barrels in winemaking, microbes from previous fermentations can survive in the pores of wood even after cleaning and can contaminate future fermentations (González-Arenzana et al., 2013). For this reason, new oak barrels were utilized in this study to ensure that the microbial communities that were identified throughout fermentation were a true reflection of the SO2 treatments being applied and not confounded by yeasts and bacteria harbored in the barrels from previous vintages. This resulted in all wines scoring very high in oak- related attributes such as toasty/smoky, unprocessed wood, spice, and vanilla attributes. It is therefore likely that some sensory differences attributed to differences in SO2 addition level may have been masked by the intensity of the oak-related attributes. To maintain both microbiological and sensorial integrity, future research of this kind, where oak barrel fermentation/ageing is not a factor, should be conducted using stainless steel barrels or tanks.

100

3.4. Summary

This research used high-throughput amplicon sequencing (Illumina MiSeq), combined with the addition of propidium monoazide (PMA), to accurately capture the living fungal and bacterial communities present throughout uninoculated fermentations of Pinot gris wines to which different levels of sulfites (0, 20, or 40 mg/L SO2) had been added at crush. SO2 addition at crush significantly altered the fungal communities, with lower levels of SO2 resulting in fermentations with higher diversity and a greater abundance of vineyard-associated yeasts. Bacterial community composition was unaffected by either SO2 addition or fermentation stage. A sensory evaluation of the finished wines found significant differences in two key attributes (citrus aroma and pome fruit flavour), with treatments that received lower levels of SO2 displaying increased fruitiness. Wines from all treatments were of acceptable quality and no undesirable sensory traits (off-flavours or aromas) were identified. These results are of interest to both the academic and winemaking communities.

101

Chapter 4: Response to sulfur dioxide addition by two commercial

Saccharomyces cerevisiae strains

4.1 Background

Sulfur dioxide (SO2) has been used as an antimicrobial and antioxidant in winemaking for hundreds of years (Bush et al., 1986). It is added at different times throughout the winemaking process, including prior to inoculation with commercial yeast strains, in order to remove any potential spoilage microorganisms from the grape must. Commercial yeasts are usually strains of Saccharomyces cerevisiae, and are expected to be more tolerant to SO2 than non-commercial yeasts and bacteria. However, different yeasts may respond differently to SO2, and it is important for winemakers to know which yeasts respond in favourable ways so they may select appropriate yeasts to use as inoculants.

Wine yeasts have four main methods of surviving in the presence of SO2: (1) entering a viable but non-culturable (VBNC) state; (2) expelling SO2 from the cell via specialized sulfite-efflux pumps; (3) reduction of SO2 via incorporation into the sulfur amino acid biosynthesis (SAAB) pathway; and (4) production of acetaldehyde (Divol et al., 2012). If the level of SO2 added to the must is too high, wine yeasts, including both spoilage yeasts such as Brettanomyces bruxellensis and fermentative yeasts such as S. cerevisiae, may enter a VBNC state where they are still metabolically active but are not detected by culture-dependent methods (Agnolucci et al., 2013,

2010; Divol and Lonvaud-Funel, 2005). Such a response is rare for commercial yeasts, because they are selected for their ability to withstand stressful environments, including musts or wines containing SO2 (Piškur et al., 2006).

102

Additionally, the concentration of SO2 added at crush is often not high enough to induce such a state in all but the most vulnerable wine microorganisms. Sulfite efflux is a desirable response to SO2 by commercial wine yeasts because the SO2 is not converted into an undesirable form nor is it bound and rendered inactive, so it is able to resume its antimicrobial activity on more susceptible microorganisms. Sulfite efflux is conducted via the specialized Ssu1p sulfite efflux pump, and is encoded by the SSU1 gene and regulated by the FZF1 transcription factor (Aa et al., 2006;

Avram et al., 1999; Nardi et al., 2010; Park and Bakalinsky, 2000). Expression of

SSU1, or its more resistant allele SSU1-R, can be induced in the presence of SO2, and its expression is often constitutively higher in more SO2-resistant strains (Goto-

Yamamoto et al., 1998; Nardi et al., 2010). When SO2 enters the cell, the higher

- internal pH converts molecular SO2 to bisulfite (HSO3 ), which can be incorporated into the SAAB pathway. Once it is in this pathway, bisulfite is reduced to S2- and then either used to produce sulfur-containing amino acids or exported from the cell as hydrogen sulfide (H2S). This response to SO2 is undesirable for winemakers and wine consumers alike, because H2S has a very low detection threshold and can lend a rotten egg or cooked cabbage aroma to the wine (Siebert et al., 2009). Finally, acetaldehyde is produced by yeasts as an intermediate in many metabolic pathways, including alcoholic fermentation. Acetaldehyde has an extremely high affinity for SO2, with one mole of acetaldehyde able to bind approximately one mole of SO2 (Divol et al., 2012). Bound SO2 is no longer active as an antimicrobial agent, so this is an effective method of sulfite resistance. However, high levels of acetaldehyde in wine can give the wine a bruised-apple or sherry-like aroma

103

(Swiegers et al., 2005). Like the sulfite efflux pump, highly resistant yeast strains tend to have higher constitutive production of acetaldehyde, even in the absence of

SO2 (Casalone et al., 1992; Stratford et al., 1987).

Previous studies have been instrumental in determining the mechanisms by which S. cerevisiae strains respond to the presence of SO2 (Aranda et al., 2006;

Goto-Yamamoto et al., 1998; Nardi et al., 2010; Park and Hwang, 2008; Stratford et al., 1987). However, most of these studies have focused on one of these mechanisms at a time, and few investigated the ways that these mechanisms may work together to provide sulfite protection (Nadai et al., 2016). Furthermore, many

SO2 studies used differences in gene expression as the sole method of inferring differences in yeast resistance mechanisms, but changes in gene expression are not always correlated with changes in protein translation or metabolic activity (de Sousa

Abreu et al., 2009; Hoppe, 2012; Schrimpf et al., 2009; Vogel and Marcotte, 2012).

Additionally, few studies have investigated the effects of SO2 addition on the production of volatile secondary metabolites by S. cerevisiae strains (Boroski et al.,

2017; Santos et al., 2015; Sun et al., 2016), and to our knowledge, no studies have compared both the sulfite-resistance mechanisms and volatile compound production of individual commercial yeast strains. Here, we attempt to address these gaps in knowledge by conducting an investigation into the potential sulfite-resistance responses of two commercial strains of S. cerevisiae, as well as their production of volatile compounds that are important to the sensory profile of wines.

The objective of this study was to observe the responses of two commercial

S. cerevisiae strains to the presence (50 mg/L) or absence (0 mg/L) of SO2 in

104 fermentations of Pinot gris juice conducted under laboratory conditions. Specifically, fermentation kinetics, yeast abundance, SO2 concentration, H2S production, acetaldehyde production, and secondary metabolite composition as a result of SO2 addition were compared between the two strains. The two yeast strains selected for comparison were Lalvin® QA23 (referred to here as Strain 1) and Lalvin® BRL97

(referred to here as Strain 2). Both strains are commonly used to inoculate fermentations in the Okanagan Valley and elsewhere in the world, and were found to be genetically distinct. QA23 was originally isolated in Portugal and is recommended for white wine production, while BRL97 was isolated from Italy and is recommended for red wine production. However, Lalvin® BRL97 was recently shown to be genetically equivalent to Lalvin® ICV-D47 via SNP analysis (Borneman et al., 2016), and both strains have identical multilocus genotypes as determined via microsatellite analysis (see Appendix A). Interestingly, ICV-D47 is recommended for white wine production, and in Chapter 2, the abundance of ICV-D47 in uninoculated Pinot gris fermentations increased in a dose-dependent manner with increasing SO2 concentration. Both QA23 and BRL97/ICV-D47 are killer-active (Howell et al., 2006), and possess unique enological properties, including β-glucosidase production

(QA23) and colour stabilization (BRL97) (www.lallemandbrewing.com). Therefore, we predicted that these strains would differ in both their susceptibility to SO2 and in their response mechanisms to SO2. We further predicted that these differences in yeast behaviour would also result in differences in their production of volatile secondary metabolites, both between strains and between SO2 treatments within one strain.

105

4.2 Materials and methods

4.2.1 Experimental design

This experiment was conducted in 2017 at the School of Agriculture, Food, and Wine at the University of Adelaide. Two commercial yeast strains—Lalvin®

QA23 and Lalvin® BRL97, referred to here as Strain 1 and Strain 2, respectively— were inoculated into sterile-filtered Pinot gris juice containing either 0 or 50 mg/L

SO2 (n = 4 per treatment). In the 50 mg/L SO2 treatments, SO2 was added as 100 mg/L potassium metabisulfite (K2S2O5). Sixteen single-strain fermentations of 100 mL were conducted at 24 C. Because of the number of parameters being measured in this experiment and the sample volumes required for some measurements, it was not feasible to measure everything at once. To address this issue, this experiment was conducted twice, with the objective of measuring different aspects of the fermentations during Run 1 and Run 2.

Run 1 was performed in 250 mL GL45 glass Schott bottles (DWK Life

Sciences, Wertheim, Germany) sealed with custom 3D-printed airlocks (filled with 3 mL sterile water) containing a central sampling port closed with a silicone septum

(https://www.carbon3d.com/case-studies/tthandadelaide/). Each fermentation flask also contained a small, sterilized magnetic stir bar which was used to maintain an even mixture of the fermentations. These fermentations were conducted in a customized Freedom EVO® automated fermentation unit (Tecan Group, Ltd.,

Männedorf, Switzerland) with temperature-controlled panels, auto-sampling capabilities, and individual magnetic stir plates for each flask (350 rpm) (described further in Peter et al., 2018). The flasks were randomly arranged within the Freedom

106

EVO®. These fermentations were carried out for 14 days, and the following parameters were measured: yeast abundance (CFU/mL) and residual sugar (g/L + fructose) throughout fermentation, as well as the production of volatile secondary metabolites, measured from samples collected at the end of the fermentations.

Run 2 was performed using 250 mL glass Erlenmeyer flasks modified to include a sampling port which was closed with a silicone septum, and sealed with a glass airlock containing no water but fitted with H2S gas detection tubes (see section

4.2.6 below). These fermentations were performed in a temperature-controlled room on two shaker beds (350 rpm). The flasks were randomly arranged such that two replicates from each treatment were placed on each shaker. These fermentations were carried out for nine days, and the following parameters were measured: free and total SO2 (mg/L) and H2S (ppm) levels throughout fermentation, as well as acetaldehyde (g/L), measured from samples collected at the end of the fermentations.

4.2.2 Inoculation, fermentation, and sampling

Pinot gris juice, pressed from grapes collected in South Australia (2015 vintage), was filtered consecutively through 0.4 µm and 0.22 µm filters. An aliquot of

100 µL of the sterile-filtered juice was plated onto YEPD agar and incubated at 28-

30 C to confirm sterility. Prior to inoculation, one colony of each yeast strain was aseptically transferred to separate flasks containing 40 mL liquid YEPD media.

These flasks were sealed with parafilm and incubated aerobically overnight at 28-30

C. A small homogeneous subsample from each of these flasks was diluted 15

107 times, 5% propidium iodide (1 mg/mL) was added to fluorescently stain any dead cells (Valentine et al., 2019), and the number of living cells were counted using a hemocytometer to determine the cells/mL of each flask. Approximately 2.5 × 106 cells/mL were inoculated from the YEPD media into a starter culture containing 45% liquid YEPD, 45% sterile-filtered Pinot gris juice, and 10% sterile water. This starter culture was incubated overnight at 28-30 C, after which the cell concentration was determined via hemocytometer, and 2.5 × 106 cells/mL were inoculated into the fermentation flasks containing sterile-filtered Pinot gris juice. At each step, non- inoculated flasks containing the same starter culture media were incubated to ensure no microbial contamination.

4.2.3 Residual sugar concentration

Samples for residual sugar (g/L glucose + fructose) were taken every day.

For each sample, 100 µL of fermenting juice was taken from each flask using a sterilized sampling needle. Residual sugar concentration (g/L glucose + fructose) was determined enzymatically. Fermentation samples were diluted and adjusted to a final volume of 200 µL for analysis in 96-well microplates. Residual sugars were quantified enzymatically via spectrophotometry on a microplate reader with microplate stacker attachment (Tecan Group, Ltd., Männedorf, Switzerland), using hexokinase + glucose-6-phosphate dehydrogenase and phosphoglucose isomerase enzymes (Megazyme Inc., Chicaco, IL, USA). Plate preparation and enzymatic/spectrophotometric analyses were performed robotically.

108

4.2.4 Yeast abundance

Samples for total yeast abundance (CFU/mL) were taken on days 1, 3, 5, 7,

10, 12, and 14. Samples were serially diluted and plated using an automated spiral plater (Don Whitley Scientific Ltd., Bingley, UK) before being incubated for 48 h at

28-30 C. After incubation, yeast colonies were counted using ProtoCOL 3

(Synbiosis©, Cambridge, UK), an automated colony counting and zone measuring software program.

4.2.5 Sulfur dioxide determination and sulfite resistance assay

Total and free SO2 were determined using an aeration, oxidation, distillation, and titration procedure described previously in Chapter 3 (section 3.2.2). Samples for SO2 determination were taken prior to the experiment, to confirm that the juice did not contain any SO2 already, as well as on days 1, 2, 4, 5, and 9 of the fermentations. On each sampling day, 10 mL samples from each replicate were combined into one 40 mL sample. This is because of the volume required for the

SO2 determination procedure. Therefore, each treatment contained a single data point on each sampling day. Molecular SO2 was calculated from the free SO2 concentration and the pH of the juice using the Henderson-Hasselbach equation

(Divol et al., 2012; Zoecklein et al., 1995). Juice pH was measured using a

CyberScan 1100 pH meter (Eutech Instruments Pte. Ltd., Singapore).

The relative sensitivity of each yeast to SO2 was determined using a sulfite resistance assay (Park et al., 1999b). Autoclaved YEPD media was buffered to pH

3.5 with tartaric acid, and precisely 18 mL media was poured onto 25 mL sterile plastic petri dishes and left to harden. Appropriate amounts of 0.5 M sodium sulfite

109

(Na2SO3) were spread onto the plates in order to create the following concentrations: 1, 1.0, 1.5, 2.0, and 2.5 mM Na2SO3. These plates were left to set overnight at room temperature. The following day, dilutions of each yeast inoculum containing approximately 100 living cells/100 µL were spread onto plates at each

Na2SO3 concentration (n = 3) and incubated for 48 h at 28-30 C, and then visually assessed for growth inhibition.

4.2.6 Hydrogen sulfide determination

The production of H2S throughout fermentation was analyzed using Kitagawa

Precision Gas Detector Tubes (no. 120SF) (Kawasaki, Japan). Each tube was cut at both ends so that gas would be allowed to travel through the tube. The bottom end of each glass tube was inserted into a 3-4 cm piece of clear rubber tubing, and subsequently inserted into the end of the glass airlock, ensuring no gas could escape the fermentation without passing through the H2S detector tube. The level of

H2S produced in each fermentation was monitored on days 1, 2, 3, 4, 5, and 9 of the experiment.

4.2.7 Acetaldehyde determination

Acetaldehyde concentration (g/L) was determined at the end of the nine day fermentations via high-performance liquid chromatography (HPLC) (Valentine et al.,

2019). This was performed on an Agilent 1100 Series HPLC (Agilent Technologies

Australia, Mulgrave, VIC, Australia) fitted with an Aminex HPX-87H column (Bio-Rad

Laboratories, Hercules, CA, USA), with 2.5 mmol/L H2SO4 as the mobile phase, flowing at a rate of 0.5 mL/min, and measured by refractive index (G1362A refractive index detector, Agilent). Before injection, samples were filtered on an AcroPrepTM

110

Advance 96-well filter plate with a 0.2 µm polytetrafluoroethylene (PTFE) membrane

(Pall©, Port Washington, NY, USA). Acetaldehyde standards of known concentration were included for quantification. Data analysis was performed using Agilent

ChemStation (version B.01.03) software.

4.2.8 Secondary metabolite analysis

Relative quantification of 26 secondary metabolites produced by yeasts during fermentation (Table 4.1) was performed using a combination of headspace- solid phase microextraction (HS-SPME) and gas chromatography-mass spectrometry (GC-MS). These compounds were chosen because they are produced by wine yeasts during alcoholic fermentation and are aroma-active, meaning they contribute to the aroma profile of the wine. At the end of the 10 day fermentations, the wines were centrifuged and decanted into 22 mL glass vials with screw caps containing a polytetrafluoroethylene (PTFE) lining, leaving no headspace, and were stored in the dark at 4 C.

111

Table 4.1. List of yeast-derived secondary metabolites measured, as well as their characteristic aromas (Acree and Arn, 2004a; Sun et al., 2018b) and odour detection thresholds (µg/L) in 10% (v/v) ethanol (Ferreira et al., 2000; Francis and Newton, 2005; Guth, 1997; Peinado et al., 2006, 2004; Salo, 1970; Siebert et al., 2005; compiled by Haggerty et al., 2016).

Category Metabolite Abbreviation Aroma Threshold (µg/L) Ethyl acetate EA fruity, solventab 7,500 Ethyl propanoate EP fruity, solventab 1,800 Ethyl 2-methylpropanoate E2MP fruitya 15 Ethyl butanoate EB apple, strawberrya 20 Ethyl 2-methylbutanoate E2MB apple, strawberrya 1-18 Ethyl esters Ethyl 3-methylbutanoate E3MB apple, strawberrya 3 Ethyl hexanoate EH fruity, fermented peara 5-14 Ethyl octanoate EO fruity, fattyb 2 Ethyl decanoate ED grapeb 200 Ethyl dodecanoate1 EDD soapy, esteryb 25,6195 2-methylbutyl acetate2 2MBA fruityb 160 3-methylbutyl acetate2 3MBA bananab 30 Acetates Hexyl acetate HA fruity, herbyb 670 2-phenylethyl acetate 2PEA rose, honeyb 250 2-methylbutanoic acid3 2MBAcid , cheeseb 1,500 3-methylbutanoic acid3 3MBAcid sweat, rancidb 33.4 Acids Hexanoic acid1 HAcid sour, vinegar-likea 420 Octanoic acid OA sweat, cheeseab 500 Decanoic acid DA soura, rancid/fattyb 1,000-8,100 2-methyl propanol 2MP roasted nutsa, solventb 40,000 2-methyl butanol4 2MB onionb 65,000 3-methyl butanol4 3MB roasted nutsa, whiskyb 30,000 Alcohols Hexanol HOH green, floralab 8,000 2-phenylethanol 2PE honey, spice, floralb 10,000-14,000 Benzyl alcohol Benz sweet, floralb 900,000 Methionol MOH cooked potatoa 1,000 1EDD and HA had overlapping retention times 22MBA and 3MBA had overlapping retention times 32MBAcid and 3MBAcid had overlapping retention times 42MB and 3MB had overlapping retention times 5Odour threshold for EDD was determined in 43% (v/v) ethanol (Sun et al. 2018)

4.2.8.1 Chemical standards

Stock solutions for all compounds were prepared from high-purity stocks from: Aldrich (Milwaukee, WI, USA) (2-phenylethyl acetate (99%), ethyl decanoate

( 99%), hexyl acetate (99%), ethyl octanoate ( 99%), ethyl hexanoate (≥ 99%), 2- methyl butyl acetate (99%), ethyl 2-methylbutanoate (99%), ethyl 3-methylbutanoate

(99%), ethyl butanoate (99%), benzyl alcohol (98%), 2-methyl butanol (≥ 99%), methionol (≥ 98%), ethyl 2-methyl propanoate (99%), ethyl acetate (99.9%), 112 hexanoic acid (≥ 99.5%), hexanol (≥ 99%), ethyl dodecanoate (≥ 98%), ethyl propanoate (99%), 2-methyl butanoic acid (≥99%), 3-methyl butanoic acid (99%));

Sigma (St. Louis, MO, USA) (2-phenyl ethanol (≥ 98%), 3-methylbutyl acetate (≥

99%), octanoic acid (≥ 99% )); Unilab (Mandaluyong City, Philippines) (1-octanol (≥

95%)); Fluka (Buchs, SG, Switzerland) (3-methyl butyl acetate (≥ 99.7%); and

Chemsupply (Gilman, SA, Australia) (ethyl 3-methyl butanoate (≥ 99.7%), 3-methyl butanol (≥ 99.8%), decanoic acid (≥ 99.5%) 2-methylpropanol (≥ 99.5%).

4.2.8.2 Quality assurance and standard curves

Concentrated ‘category stock solutions’ of these compounds (Table 4.1) were pre-made and dissolved in ethanol (Haggerty et al., 2014). A stock solution (10 mg/L in 10% ethanol) of the internal standard 1-octanol was also created. Eight dilutions of the stock solution, corresponding to the following percent concentrations, were used to create standard curves for each of the evaluated compounds: 100%

(concentrated stock solution), 40%, 20%, 4%, 1%, 0.2%, 0.04%, and 0.01%.

4.2.8.3 Sample preparation

Duplicates of each sample were made following a general SPME procedure:

2.0 g sodium chloride, 1.5 mL sample, 8.4 mL Milli-Q® water (Merck Millipore©,

Darmstadt, Germany), and 100 µL internal standard (10 mg/L 1-octanol) were added to a 20 mL SPME crimp cap vial, closed with a magnetic PTFE/silicone crimp cap seal, and mixed thoroughly.

4.2.8.4 Instrumentation and parameters

Samples were analyzed in duplicate via GC-MS with HS-SPME extraction on an Agilent Technologies© 7890A GC coupled with a 5975C MS detector (Santa

113

Clara, CA, USA). Prior to analysis, samples were agitated at 250 rpm for 10 min at

50 C, then extracted with a 65 µm 24 gauge polydimethylsiloxane/divinylbenzene

(PDMS/DVB) fiber (Supelco®, Bellefonte, PA, USA) for 10 min. The GC inlet was fitted with a SPME injection sleeve (0.75 mm ID) (Supelco®) and the inlet temperature was set to 250 C, with pressure at 184.84 kPa and a split ratio of 0.6:1 with a helium carrier gas at 1.02 mL/min. Compound separation was performed on a

60 m × 0.25 mm VF-WAXms column (Agilent Technologies©). The oven program was as follows: 1.5 min at 50 C, then ramped at a rate of 5 C/min up to 200 C, then at a rate of 10 C/min to 240 C, which was held for 3.5 min (total run time 39 min). All samples were run under selected ion monitoring (SIM) mode with corresponding dwell times (Table C.1). The transfer line was held at 250 C. Source temperature was held at 230 C and electron impact ionization was performed at 70 eV. Quadrupole temperature was held at 150 C. To ensure instrumentation stability, a QC injection of the third standard curve solution was performed at three points during each analysis run.

4.2.9 Strain-typing

To confirm that the strains used in this study were genetically distinct, the multilocus genotypes of Strain 1 and Strain 2 were determined via fragment analysis of eight hypervariable microsatellite loci. Strain-typing was performed as described in Chapter 2 (section 2.2.5). These multilocus genotypes can be viewed in Table C2.

4.2.10 Statistical analysis

A standard curve for each compound of interest measured via HS-SPME GC-

MS was created in Excel 2016, and a linear line of best fit was applied to each

114 volatile compound of interest. R2 values above 0.99 were confirmed for all compounds, except ethyl acetate, which had an R2 value of 0.93 (data not shown).

This indicated that no non-target compounds were being measured along with the compound of interest.

Residual sugar, yeast abundance, H2S production, and acetaldehyde production data were all statistically analyzed using RStudio (version 3.5.1) Two- factor repeated measures ANOVA were performed to test for differences in residual sugar (g/L glucose + fructose) and yeast abundance (CFU/mL) among treatments throughout fermentation. These ANOVA were performed using the “aov” function.

When appropriate, Tukey HSD post-hoc tests were performed to conduct pairwise comparisons among treatments using the “lm” function in the nlme package, and the

“glht” and “cld” functions in the multcomp package, adjusted for multiple comparisons using the Holm method. The assumption of homogeneity of variance was assessed using the “leveneTest” function from the car package (version 3.0-0), which indicated a significant difference among treatments for yeast abundance

(F(3,93) = 10.1, p = < 0.001) but not for residual sugar (F(3,220) = 2.04, p = 0.11).

Yeast abundance data were normally distributed, but residual sugar data were not.

However, because ANOVA are generally robust to violations of assumptions when sample sizes are equal among groups, no data transformations were performed.

The production of H2S (ppm) by yeasts in response to SO2 addition was compared among yeast strains by performing a one-factor repeated-measures

ANOVA on untransformed data using the “aov” function. Because no H2S was produced in the 0 mg/L treatments, these samples were removed from analysis.

115

Levene’s test indicated no violation of the assumption of homogeneity of variance

(F(1,34) = 1.31, p = 0.26). A Tukey HSD test (using the “lm,” “glht,” and “cld” functions), adjusted for multiple comparisons using the Holm method, was used to perform pairwise comparisons among strains.

Acetaldehyde production was assessed by performing a two-factor ANOVA on untransformed data using the “aov” function. The assumption of homogeneity of variance was assessed using the Levene test, which indicated no significant difference among treatments (F(3,12) = 0.559, p = 0.65). A Tukey HSD test (using the “lm,” “glht,” and “cld” functions), adjusted for multiple comparisons using the

Holm method, was used to perform pairwise comparisons among treatments.

The relative quantity of secondary metabolites among treatments was assessed by performing two-factor ANOVA on untransformed data using the “aov” function. Each compound was analyzed separately. When appropriate, Tukey HSD tests (using the “nlme” and “glht” functions), adjusted for multiple comparisons using the Holm method, performed pairwise comparisons among treatments. The composition of secondary metabolites was analyzed using Primer v.6 software with

PERMANOVA+ add-on (Plymouth, MA, USA) (Clarke and Gorley, 2006). A two- factor permutational analysis of variance (PERMANOVA), using Euclidean distance and Type II sums of squares, was performed in order to test for differences in secondary metabolite composition among treatments. A PERMDISP test, using

Euclidean distance and calculating deviation from centroid, indicated no violation of the assumption homogeneity of multivariate dispersion (F(3,12) = 2.86, p = 0.31).

Test statistics (F values for PERMDISP and Pseudo-F values for PERMANOVA)

116 were calculated based on 999 permutations of normalized data. A principal coordinate analysis (PCoA) was generated in order to visualize differences in secondary metabolite composition among treatments. Variable vectors were plotted using Pearson’s correlation.

4.3 Results and discussion

4.3.1 Fermentation kinetics and cell density

Fermentation kinetics, as measured by residual sugar content (g/L glucose + fructose), differed as a result of both yeast strain and SO2 level (Figure 1). In the

® ® control (when no SO2 was added), both Strain 1 (Lalvin QA23) and Strain 2 (Lalvin

BRL97) completed fermentation at the same rate (Figure 1A). However, when SO2 was added, Strain 1 performed fermentation at a slower rate (Figure 1B). Strain 2 conducted alcoholic fermentation at the same rate at both SO2 concentrations. This suggests that Strain 1 is more sensitive to SO2 addition than Strain 2. This was confirmed by a sulfite resistance assay (see Section 4.2.5), where Strain 2 showed no decrease in colony size at any concentration of Na2SO3, and Strain 1 showed signs of growth inhibition beginning at 1.5 mM Na2SO3 (data not shown).

117

Figure 4.1. Daily residual sugar levels (g/L ± SEM) of controlled fermentations conducted by different yeast strains (n = 4 per treatment). Fermentations were conducted in the presence of either (A) 0 mg/L SO2 or (B) 50 mg/L SO2. Treatments with different superscript letters indicate significantly difference fermentation progressions (two-factor repeated measures ANOVA, α = 0.05).

118

Previous studies (Andorrà et al., 2008; Bokulich et al., 2014; Henick-Kling et al., 1998) have found no differences in fermentation rate in response to SO2 addition, and one study (Constantí et al., 1998) found an increase in fermentation rate with added SO2. However, these studies were conducted using non-sterile grape juice, because the objective was to follow changes in yeast community dynamics as a result of SO2 addition level. In the case of Constantί et al. (1998) it is likely that the addition of SO2 removed a significant portion of the non-

Saccharomyces yeast community, thereby reducing competition for S. cerevisiae.

Contrastingly, in the case of our study, the progression of fermentation in each flask was dependent upon the SO2 resistance of a single yeast strain. Furthermore, even though Strain 1 did perform AF at a slower rate when SO2 was added, it was still able to complete fermentation successfully within 12 days, and no fermentations in this study could be considered slow or sluggish.

Yeast abundance (CFU/mL) also differed by both yeast strain and SO2 level

(Figure 2). Yeast abundance increased at the start of fermentation, and remained high until Day 5-7, after which it began to decrease for all treatments. In the control fermentations, Strain 1 had the highest abundance, but when SO2 was added, Strain

2 had the highest abundance. Overall, the yeast abundance kinetics of Strain 1 did not differ between the two SO2 treatments (as determined by a repeated-measures

ANOVA), but even so, when SO2 was added, the growth Strain 1 was delayed, remaining low until Day 5. This is likely due to the increased SO2 sensitivity of Strain

1, and may have caused the slower fermentation rate in the SO2 treatment as compared with the control. Cocolin and Mills (2003) also observed an initial

119 decrease in S. cerevisiae cell concentration in response to SO2 addition, but this initial decrease did not affect its ability to complete fermentation within a similar time frame as the fermentations conducted without SO2 addition. Strain 2, on the other hand, exhibited increased growth in the presence of SO2, reaching much higher total yeast abundance in the SO2 treatment as compared to the control. Previous studies, conducted with non-sterilized grape must, have found that SO2 addition can both stimulate yeast growth (Pateraki et al., 2014; Morgan et al., unpublished) or inhibit it

(Cocolin and Mills, 2003; Henick-Kling et al., 1998), depending on the must characteristics, and also likely on the yeast strain(s) used. In accordance with our

Strain 1 results, Andorrà et al. (2008) also noticed no difference in overall yeast abundance in fermentations inoculated with QA23 (our Strain 1) and fermented with or without SO2 addition.

The observation of both strains being present at every stage of fermentation in the SO2 treatment suggests that neither of the yeast strains entered a VBNC state. Although SO2 is able to induce a VBNC response in S. cerevisiae (Divol and

Lonvaud-Funel, 2005), it is relatively rare to have commercial S. cerevisiae strains exhibiting this mechanism. In addition, the concentration of SO2 used in this study

(50 mg/L) is commonly added to commercial fermentations, and was likely much too low to induce a VBNC state in commercial S. cerevisiae strains.

120

Figure 4.2. Total yeast abundance (CFU/mL ± SEM) during controlled fermentations conducted by different yeast strains (n = 4 per treatment). Fermentations were conducted in the presence of either (A) 0 mg/L SO2 or (B) 50 mg/L SO2. Treatments with different superscript letters indicate significantly different yeast abundance dynamics throughout fermentation (two-factor repeated measures ANOVA, α = 0.05).

121

4.3.2 Sulfur dioxide concentration during fermentation

Sulfur dioxide concentrations changed over the course of fermentation for all treatments. When no SO2 was added, total SO2 was 0 mg/L on Day 1, increased between Day 1 and Day 2 due to SO2 production by yeasts, and then decreased gradually over the rest of the fermentation for both yeast strains (Figure 4.3A). Both strains produced similar total SO2 levels. The treatments that received 50 mg/L SO2 contained ~46 mg/L total SO2 on Day 1 (Figure 4.3B). The slight decrease in the observed total SO2 concentration from what was added is likely due to the fact that more than one hour had elapsed between the time when the SO2 was added to the juice and the time when it was measured, leading to natural SO2 loss. In the SO2- added fermentations, total SO2 decreased consistently over the course of the nine day experiment for both strains, and both strains had similar total SO2 patterns.

While total SO2 is useful for confirming SO2 addition levels and for monitoring

SO2 production by yeasts, free SO2 is more useful for evaluating antimicrobial activity. Free SO2 is the portion of total SO2 that is available to act as an antimicrobial agent, and molecular SO2, which has the most antimicrobial activity, can be calculated from the concentration of free SO2 and the pH of the juice (Divol et al., 2012; Zoecklein et al., 1995). In the 50 mg/L SO2 treatments, approximately

28 mg/L of the SO2 added was present in its free form on Day 1 (Figure 4.3C). The pH of the juice on Day 1 was 3.159, and molecular SO2 was calculated to be ~1.2 mg/L in these fermentations. Free SO2 represented approximately 60% of the total

SO2 concentration on Day 1, which is a larger proportion than is regularly found in grape must after SO2 addition (Henick-Kling et al., 1998; Morgan et al., 2019;

122

Pateraki et al., 2014). This is likely because this juice was sterile-filtered before the

SO2 was added, and there were few grape constituents and only one microbe (as opposed to a diversity of microorganisms) present to bind the SO2. By Day 2, the fermentations inoculated with Strain 2 did not contain any free SO2, but the fermentations inoculated with Strain 1 still contained approximately 21 mg/L. By Day

4, free SO2 had decreased to ~ 0 mg/L for both treatments, and remained at this concentration until the end of the experiment. Free SO2 was not produced by either strain in the 0 mg/L SO2 treatments.

123

Figure 4.3. SO2 levels (mg/L) throughout fermentations conducted by different yeast strains and in the presence of different levels of initial SO2. Wine samples from each of four replicates were combined for SO2 analysis due to volume constraints. Plots represent (A) Total SO2 produced by yeasts throughout fermentations containing 0 mg/L SO2; (B) Total SO2 levels throughout fermentations containing 50 mg/L SO2; (C) Free SO2 levels throughout fermentations containing 50 mg/L SO2.

The extended presence of free SO2 in the fermentations conducted by Strain

1, but not by Strain 2, suggests that these strains may have different responses to

SO2. It is possible that Strain 1 has a higher constitutive expression of FZF1 or

124

SSU1/SSU1-R, which encode and regulate the sulfite efflux pump Ssu1p (Goto-

Yamamoto et al., 1998; Park and Bakalinsky, 2000). A higher expression of these genes could lead to more free SO2 being exported back outside the yeast cell. This response to SO2 is desirable for winemakers, because it allows the exported SO2 to act as an antimicrobial agent against other microorganisms which may be spoilage- causing. Studies have shown that these genes display a high degree of polymorphism, and different yeast strains have highly variable expressions of these genes (Goto-Yamamoto et al., 1998; Nardi et al., 2010). However, it is also possible that the prolonged presence of free SO2 seen in these treatments was simply a result of decreased yeast activity during the first days of the experiment, due to increased sulfite sensitivity; Strain 1 did display decreased cell growth during the first few days of fermentation when SO2 was added (see section 4.3.1 above). A detailed exploration of the transcriptome and proteome of these two strains, particularly with regard to genes and gene products involved in sulfite efflux, could help conclusively determine the reasons for the differences in sulfite response observed here.

4.3.3 Hydrogen sulfide production during fermentation

Hydrogen sulfide production differed between strains when SO2 was added

(Figure 4.4), but no H2S was produced by either strain in the absence of SO2 (data not shown). In the SO2 treatments, both strains produced similar levels of H2S by the end of the nine day fermentations, but their rates of production differed. Neither strain produced H2S by Day 2, but by Day 3 H2S production began to differ. Between

Day 2 and Day 3, Strain 2 produced an average of 152 ppm. After this rapid

125 increase, H2S production slowed for the remainder of the fermentation. Strain 1 did not produce any H2S until Day 5, after which time H2S production increased to an average of 215 ppm by Day 9.

Figure 4.4. H2S production (ppm ± SEM) by different yeast strains when fermented in the presence of 50 mg/L SO2. Treatments with different superscript letters indicate significant differences in overall H2S production dynamics (two-factor repeated measures ANOVA, α = 0.05).

- When SO2 enters the yeast cell, it is converted to bisulfite (HSO3 ), which may then be incorporated into the SAAB pathway and reduced to S2- by the sulfite reductase , encoded by the MET5/MET10 genes. This S2- may be used to produce sulfur-containing amino acids, or may be exported outside the cell as H2S.

The fate of S2- in this pathway is in part determined by the levels of sulfur-containing amino acids already present in the yeast cell, as higher methionine levels in the cell have been correlated with decreased SO2 resistance (Aranda et al., 2006). The production of H2S by yeasts is also highly strain-dependent (Linderholm et al.,

2008), and is considered an undesirable response to SO2, because H2S has a very

126 low detection threshold (1.1-1.6 µg/L), above which it can produce a rotten egg or cooked cabbage smell in wine (Siebert et al., 2009; Smith et al., 2015). The delay in

H2S production by Strain 1 in response to SO2 addition is possibly due to its decreased cell density at the beginning of fermentation as a result of sulfite sensitivity, since both strains eventually produced similar total H2S over the course of fermentation. Because H2S production by yeasts is related to the presence of free

(and molecular) SO2 in the environment, the higher relative proportion of free SO2 in these experimental fermentations likely resulted in higher-than-normal H2S production by these yeast strains. Furthermore, H2S is very volatile, and our methodology in this study involved determining only the total amount of H2S produced by each strain, not the final concentration of H2S in the wines at the end of the fermentation. Anecdotally, none of these wines expressed any aromas related to

H2S, and it is likely that the final concentration of H2S in the wines was below its detection threshold.

4.3.4 Post-fermentation acetaldehyde concentration

The addition of SO2 did not alter the production of acetaldehyde for either strain, but overall, Strain 2 produced more acetaldehyde than Strain 1 (Figure 4.5).

Acetaldehyde has a high affinity for SO2, and can bind to SO2 outside the cell and effectively prevent its antimicrobial actions (Divol et al., 2012). At low levels (< 125 mg/L), acetaldehyde may benefit a wine’s sensory profile, but at higher concentrations it can cause the wine to smell of bruised apple and sherry

(Bartowsky and Pretorius, 2009; Swiegers et al., 2005; Zoecklein et al., 1995).

Therefore, high acetaldehyde production by wine yeasts is considered undesirable.

127

In white wine, acetaldehyde levels range from 11-493 mg/L (Liu and Pilone, 2000).

The acetaldehyde levels found in this current study fall in the middle of this range

(100 - 250 mg/L), and were slightly above the sensory threshold for acetaldehyde, though likely not high enough in any treatment to result in a fault. Some previous studies have found acetaldehyde production by yeasts to increase in response to

SO2 addition (Divol et al., 2006; Weeks, 1969), and others have found increased acetaldehyde production to be a general trait of sulfite-resistant strains, even in the absence of SO2 (Casalone et al., 1992; Stratford et al., 1987). The higher constitutive production of acetaldehyde by Strain 2 could explain our finding that

Strain 2 was more resistant to SO2 than Strain 1. It is currently unclear whether the increased production of acetaldehyde is a defense mechanism used by yeasts to survive in the presence of SO2, or simply a result of the presence of SO2 causing enzymatic inhibition of the products of acetaldehyde (Divol et al., 2012).

128

Figure 4.5. Acetaldehyde production (g/L ± SEM) by different yeast strains when fermented in the presence of 0 or 50 mg/L SO2. Treatments with different letters indicate significant differences in acetaldehyde concentration (two-factor ANOVA, α = 0.05).

4.3.5 Yeast-derived secondary metabolite composition

The composition of the 26 volatile secondary metabolites measured (Table

4.1) was affected by both yeast strain and SO2 treatment (Table 4.2), and a PCoA ordination was created to visualize the spatial distribution of samples with respect to their volatile profiles (Figure 4.6). This PCoA showed that separation of samples by

SO2 treatment and by yeast strain was achieved with the second principal component (PCO2), representing 29.8% of total variation. When no SO2 was added,

Strain 1 and Strain 2 produced similar volatile profiles (Figure 4.6A). Additionally, the overall composition of volatile secondary metabolites of the wines fermented by

Strain 1 was not significantly altered by the addition of SO2, although the relative 129 abundance of some specific metabolites did vary (Table 4.3, discussed below).

However, the production of volatile compounds by Strain 2 was affected by SO2 addition, and more variation was observed among replicates for the wines fermented by this strain. A similar result was also observed by Santos et al. (2015), who found different compositions of volatile compounds between wines produced with and without SO2. In that study, the volatile composition was also analyzed via HS-SPME

GC-MS, although most compounds were combined into representative groups for analysis. It is also unclear in that study which commercial S. cerevisiae strain was used for inoculation, or what concentration of SO2 was added at crush. A study conducted by Howell et al. (2006) found that single-strain fermentations conducted by QA23 (our Strain 1) and ICV-D47 (genetically equivalent to our Strain 2) produced unique volatile profiles. In that study, the volatile profiles produced by

QA23 were more variable among replicates than the profiles produced by ICV-D47, which is in contrast with our findings. However, it is unclear whether these fermentations were conducted in the presence of SO2 or not.

Table 4.2. Results of a two-factor PERMANOVA evaluating the effects of yeast strain and SO2 addition on the volatile secondary metabolite profiles of Pinot gris wines. Results with an asterisk (*) are significant at α = 0.05.

Source df SS MS Pseudo-F p Yeast strain (Y) 1 79.963 79.963 5.3215 0.002* Sulfite addition (S) 1 43.755 43.755 2.9119 0.009* Y × S 1 25.965 25.965 1.7279 0.13 Residual 12 180.32 15.026 Total 15 330

130

Figure 4.6. PCoA visualizing the production of volatile secondary metabolites by different yeast strains in controlled Pinot gris fermentations containing either 0 mg/L or 50 mg/L SO2 (n = 4 per treatment). Plots depict (A) the individual factors map and (B) the variable factors map showing secondary metabolites as vectors. On the individual factors map, each point represents the entire volatile secondary metabolite composition of a single wine sample. Points that are closer together contain more similar volatile profiles than points that are further apart. On the variable factors map, the length of the variable vector reflects that variable’s relative contribution in producing this ordination. Variable vectors that point towards a sample are positively correlated with that sample, and those that point away from a sample are negatively correlated. Variable vectors that appear at a ~90 angle are not correlated with each other, while those that appear at < 90 and > 90 angles are positively and negatively correlated with each other, respectively.

131

Both Strain 1 and Strain 2 increased their production of 2/3-methylbutyl acetate, hexyl acetate, 2-phenylethyl acetate, and 2-phenyl ethanol in the presence of SO2 (Table 4.3). Strain 2 also showed an increase in the production of 2-methyl propanol, 2/3-methyl butanol, and methionol, and a decrease in the production of hexanoic acid/ethyl dodecanoate when SO2 was added. In general, the production of volatile compounds was higher in the SO2-added wines, especially with regards to the wines fermented by Strain 2. Strain 1 was less affected by SO2 addition in terms of the production of volatile compounds.

Table 4.3. Relative quantity (± SEM) of volatile compounds in wines fermented by different commercial yeast strains and in the presence of different sulfite levels (n = 4 per treatment). Compounds with an asterisk (*) indicate significant results from a two-factor ANOVA, and different superscript letters indicate significant differences among treatments (α = 0.05). Each compound/compound group was analyzed separately.

Compound(s) Strain 1 Strain 1 Strain 2 Strain 2 (0 mg/L SO2) (50 mg/L SO2) (0 mg/L SO2) (50 mg/L SO2) EA 4.10 ± 0.13 4.21 ± 0.15 4.64 ± 0.41 3.96 ± 0.20 EP 0.206 ± 0.012 0.198 ± 0.0070 0.246 ± 0.074 0.150 ± 0.016 E2MP* 0.0805 ± 0.0054a 0.0858 ± 0.0031a 0.103 ± 0.013ab 0.131 ± 0.011b EB 0.334 ± 0.0079 0.316 ± 0.018 0.421 ± 0.10 0.272 ± 0.0032 E2MB 0.0257 ± 0.0017 0.0220 ± 0.00077 0.0273 ± 0.0036 0.0328 ± 0.0035 E3MB 0.0245 ± 0.0051 0.0216 ± 0.0034 0.0347 ± 0.0064 0.0329 ± 0.0064 2MP* 1.02 ± 0.066a 1.07 ± 0.027ab 0.928 ± 0.11a 1.35 ± 0.074b 2/3MBA* 0.730 ± 0.020a 1.04 ± 0.031b 0.738 ± 0.041a 1.21 ± 0.042c 2/3MB* 54.2 ± 1.7ab 55.7 ± 1.9ab 47.8 ± 1.4a 58.5 ± 3.1b EH 11.1 ± 0.41 13.1 ± 0.82 13.2 ± 2.6 10.2 ± 1.3 HA* 0.322 ± 0.020a 0.567 ± 0.046b 0.386 ± 0.020a 0.581 ± 0.039b HOH 3.57 ± 0.11 3.37 ± 0.048 3.45 ± 0.13 3.47 ± 0.11 EO 38.8 ± 3.0 45.0 ± 4.2 36.4 ± 4.2 40.0 ± 2.3 ED 11.7 ± 1.6 10.8 ± 1.2 9.61 ± 3.0 10.1 ± 0.90 2/3MBAcid* 0.213 ± 0.0098ab 0.196 ± 0.0057a 0.236 ± 0.025ab 0.277 ± 0.014b MOH* 0.0656 ± 0.0091a 0.0781 ± 0.0088a 0.0480 ± 0.0074a 0.122 ± 0.0076b 2PEA* 1.28 ± 0.038b 1.65 ±0.061c 0.843 ± 0.018a 1.35 ± 0.012b HAcid/EDD* 2.44 ± 0.18ab 2.06 ± 0.13ab 2.66 ± 0.31b 1.71 ± 0.095a Benz 0.0159 ± 0.0011 0.0158 ± 0.0011 0.0150 ± 0.0013 0.0155 ± 0.00069 2PE* 35.1 ± 1.4b 41.84 ± 1.8c 22.3 ± 0.70a 32.2 ± 1.5b OA* 4.85 ± 0.16ab 5.64 ± 0.063b 4.77 ± 0.33a 5.46 ± 0.10ab DA 5.51 ± 0.19 5.71 ± 0.29 4.41 ± 1.1 4.84 ± 0.074

132

Two of the four replicate wines fermented by Strain 2 in the SO2 treatment were positively correlated with the production of 2/3-methylbutyl acetate, hexyl acetate, methionol, 2-phenylethyl acetate, 2-phenylethanol, and ethyl octanoate

(Figure 4.6). These compounds produce a range of aromas, including both desirable

(fruity/banana, floral/rose, honey, herby) and undesirable (fatty, and cooked potato) aromas (Table 4.1), and can often be found above their sensory thresholds in wines fermented by S. cerevisiae (Capece et al., 2013; Herraiz et al., 1990; Mateo et al.,

2001; Nicolini et al., 2011; Peinado et al., 2004; Saberi et al., 2012). The other two replicate wines fermented by Strain 2 in the SO2 treatment were found to be more similar to the wines fermented by Strain 1 in the presence of SO2, which were positively correlated with the production of ethyl decanoate (grape) and decanoic acid (sour/rancid/fatty), and negatively correlated with the production of ethyl acetate

(fruity/solvent), ethyl hexanoate (fruity/fermented pear), and ethyl 3-methylbutanol

(apple/strawberry) (Figure 4.6). These compounds are also often found above their sensory thresholds in wine (Capece et al., 2013; Garcia et al., 2002; Mateo et al.,

2001; Nicolini et al., 2011; Peinado et al., 2004; Saberi et al., 2012; Ubeda Iranzo et al., 2000), and because ethyl acetate was negatively correlated with these wines, it is likely that it was found at a lower concentration, which can lend a pleasant fruity aroma to the wine.

In general, the wines produced from the SO2 control treatment (no SO2 added) were negatively correlated with the production of compounds such as 2- methyl propanol, ethyl 2-methylpropanoate, ethyl 2-methylbutanoate, and 2/3- methylbutanoic acid. These compounds give wine roasted nuts/solvent, fruity,

133 apple/strawberry, and butter/cheese/sweat/rancid aromas, respectively (Table 4.1).

A single replicate wine fermented by Strain 2 in the control treatment was positively correlated with the production ethyl butanoate (apple/strawberry), and negatively correlated with the production of octanoic acid (sweat/cheese) (Figure 4.6). The wines fermented without SO2 were therefore generally found to be negatively correlated with the production of compounds that produce undesirable aromas, such as octanoic acid, 2/3-methylbutanoic acid, and 2-methyl propanol. Previous studies that have investigated the effects of SO2 addition at crush on the sensory profiles of wines have also noted that in general, wines fermented without SO2 tend to contain more fruity aromas, while wines fermented with SO2 contain higher intensities of undesirable characteristics (see Henick-Kling et al., 1998 and results from Chapter

3).

Few other studies have investigated the effect of SO2 addition on the volatile profiles of wines fermented by different S. cerevisiae strains (Boroski et al., 2017;

Santos et al., 2015; Sun et al., 2016). Partially in accordance with our results,

Santos et al. (2015) also observed differences in the composition of volatile compounds in wines produced with and without SO2, and Sun et al. (2016) also noted an increase in the production of 2-phenylethanol when SO2 was added to strawberry wine fermentations. However, contrary to our results, Boroski et al.

(2017) noted a general decrease in the relative production of volatile compounds in response to SO2 addition in Chardonnay. It is likely that the production of volatile secondary metabolites in response to SO2 addition is highly dependent on the

134 composition of the juice being fermented as well as the specific yeast strains involved in the fermentation.

4.4 Summary

This study investigated the responses of two different commercial S. cerevisiae strains to the presence (50 mg/L) and absence (0 mg/L, control) of SO2 in terms of fermentation kinetics and yeast-derived volatile secondary metabolites.

Strain 2 was found to be more resistant to SO2 than Strain 1, which exhibited decreased growth and a slower fermentation rate when SO2 was added. However, both yeast strains successfully completed fermentation with or without SO2 addition, and neither yeast strain exhibited a VBNC response to the 50 mg/L SO2 added. Free

SO2 levels differed between strains when 50 mg/L SO2 was added, with free SO2 remaining in the juice for a longer period of time in the fermentations conducted by

Strain 1. The rate of H2S production also differed between the two yeast strains in response to SO2 addition, but both strains ultimately produced similar levels of H2S by the end of the fermentations. Strain 2 expressed constitutively higher acetaldehyde production, potentially explaining at least part of the difference in SO2 resistance observed between the two strains. Based on these results, sulfite efflux and acetaldehyde production were the most different between the two strains, but it is likely that a combination of multiple mechanisms were involved in sulfite resistance for both strains. Research into the transcriptome and proteome of these two specific strains may help provide a more clear understanding of which mechanism(s) these strains rely on most heavily for sulfite resistance. Overall, the production of volatile compounds by both yeast strains was higher when SO2 was

135 added, and the wines produced without SO2 were generally negatively correlated with the production of undesirable volatile compounds. Strain 1 produced a consistent wine regardless of whether SO2 was added, and the volatile profile of these wines were not different between SO2 treatments. Conversely, wines fermented by Strain 2 were much more variable in their volatile profiles, and SO2 significantly affected the composition of these yeast-derived secondary metabolites.

These results may help winemakers make informed decisions when selecting commercial yeasts for inoculation into fermentations containing different levels of

SO2, and when specific wine characteristics are desired.

136

Chapter 5: Effect of sulfite addition and pied de cuve inoculation on the microbial communities and sensory profiles of Chardonnay wines

5.1. Background

Wine production involves the interaction of many microbial species, including both yeasts and bacteria. These microorganisms may reside in the vineyard, and enter the winery along with the grapes at harvest. They may also be winery-resident strains that survive on winery equipment and subsequently enter fermentations.

Wine yeasts are often grouped into two categories: Saccharomyces cerevisiae and non-Saccharomyces yeasts. S. cerevisiae is generally considered to be a winery- resident yeast, and is rarely found in significant numbers in the vineyard or on healthy grapes (Mortimer and Polsinelli, 1999). While some wineries around the world rely on the presence of indigenous strains of S. cerevisiae to conduct alcoholic fermentation (AF), most wineries use commercial strains of S. cerevisiae to inoculate their musts, which can ensure a timely and consistent progression of AF.

Conversely, non-Saccharomyces yeasts are associated with both vineyard and winery environments, and this group is comprised of many different species, including Hanseniaspora spp., Metschnikowia spp., Candida spp., Pichia spp.,

Torulaspora spp., and many others. Confusingly, this group also contains some species belonging to the Saccharomyces genus, including Saccharomyces uvarum.

Most non-Saccharomyces species are poor fermenters and cannot complete AF without the assistance of S. cerevisiae, but recent studies have highlighted their importance in producing wines with increased aroma and flavour complexity (Jolly et al., 2014; Liu et al., 2016). There are, however, some non-Saccharomyces yeasts

137 that have shown great fermentative potential, and have been observed conducting and completing alcoholic fermentation; among these yeasts are T. delbrueckii

(Azzolini et al., 2015, 2012; Bely et al., 2008) and S. uvarum (Contreras et al., 2015;

Demuyter et al., 2004). S. uvarum has been associated with cider production

(Almeida et al., 2014; Coton et al., 2006; Suárez Valles et al., 2007), some traditional fermented beverages (Rodríguez et al., 2017, 2014), and low-temperature wine fermentations around the world (Demuyter et al., 2004; Naumov et al., 2000;

Zhang et al., 2015). S. uvarum produces lower concentrations of acetic acid and higher concentrations of glycerol compared to S. cerevisiae (Magyar and Tóth,

2011), and has also been recommended for the production of reduced-alcohol wines

(Varela et al., 2017). However, currently there is only one commercial strain of S. uvarum available (Velluto BMV58®).

For many years, S. uvarum was known academically as S. bayanus var. uvarum (Masneuf-Pomarede et al., 2007; Naumov et al., 2002, 2001, 2000) or simply S. bayanus (Coton et al., 2006; Suárez Valles et al., 2007). More recently, S. uvarum has been re-classified as a pure species within the Saccharomyces sensu stricto complex, and S. bayanus is now understood to be a hybrid between S. eubayanus and S. uvarum (Borneman and Pretorius, 2015; Nguyen and Gaillardin,

2005). Unfortunately, due to the different names this species has had, and because

S. uvarum and S. bayanus were once considered synonymous, it is difficult to trace which studies truly investigated S. uvarum. Furthermore, many commercial yeasts that claim to be strains of S. bayanus have been since genetically characterized as

138 strains of S. cerevisiae, further obscuring the history of research into S. uvarum

(Fernández-Espinar et al., 2001).

S. uvarum, along with other non-Saccharomyces yeasts, can contribute to the production of flavour-active secondary metabolites. Research has shown that fermentations conducted with a variety of yeast species and strains are more complex than single-strain fermentations (Bagheri et al., 2018; King et al., 2010;

Zara et al., 2014). In recent years, some winemakers have begun to look for ways to naturally increase the diversity of yeasts in their fermentations. The most common method is to allow the grape must to ferment uninoculated, or spontaneously, without the addition of any commercial yeast strains. The goal of uninoculated fermentations is to encourage non-commercial yeasts to conduct AF. These yeasts are expected to be a combination of vineyard-derived non-Saccharomyces yeasts and indigenous strains of S. cerevisiae (Combina et al., 2005; Frezier and

Dubourdieu, 1992; Vigentini et al., 2014), although some studies have found the fermentations consisting of a mixture of commercial strains (Blanco et al., 2011;

Santamaría et al., 2005; Scholl et al., 2016). Another method involves intentionally inoculating vineyard-derived non-Saccharomyces yeasts prior to the onset of AF.

Pied de cuve is a French term that refers to two different methods of indirect inoculation: (1) an inoculum made from must that is already fermenting in the winery

(Clavijo et al., 2011b; Li et al., 2012; Ubeda Iranzo et al., 2000); or (2) an inoculum made from spontaneously-fermenting must in the vineyard, away from the influence of winery-resident yeast strains (Sturm et al., 2006). In this study, we use the second definition of pied de cuve inoculation, which is intended to increase the contribution

139 of non-Saccharomyces yeasts from the vineyard by intentionally adding them in higher numbers to increase their competitive edge at the onset of AF. Yeasts are not the only microorganisms present on grapes in the vineyard; bacteria are also found on wine grapes (Barata et al., 2012), and these bacteria will also become part of the microbial community of the pied de cuve inoculum.

The bacterial community in wine has been less thoroughly studied than the yeast community (Bokulich et al., 2014, 2012; Cappello et al., 2016; Park, 2014;

Portillo et al., 2016), due to the fact that bacteria do not conduct AF. Bacteria are, however, responsible for conducting malolactic fermentation (MLF), the conversion of L-malic acid to L-lactic acid. Malolactic fermentation usually takes place after AF has completed and the yeast activity in the fermentation vessel has died down, reducing the competition for strains of lactic acid bacteria (LAB) (Alexandre et al.,

2004). MLF reduces the acidity and can potentially increase the buttery character of the wine (Styger et al., 2011). This is desirable for most red wines and some white wines, including Chardonnay. The inevitable interaction between yeasts and bacteria necessitates an investigation into the entire microbial communities present in fermenting must.

One current method for analyzing the entire microbial community is via high- throughput amplicon sequencing technologies, such as the Illumina sequencing platform. These technologies do not require culture-based methods, and as such they are able to identify a much wider range of microorganisms in a sample, including rare and viable but not culturable organisms. However, these technologies can be prone to an over-estimation of certain taxa due to dead microorganisms

140 within the specific medium (e.g. wine must in the later stages of fermentation). To address this, some researchers add a DNA-binding dye called propidium monoazide

(PMA), which binds to and inhibits the amplification of DNA from dead cells during

PCR (Bokulich and Mills, 2012; Morgan et al., 2019; Tantikachornkiat et al., 2016;

Vendrame et al., 2013). In this way, the entire living community of microorganisms in a sample can be captured. Unfortunately, high-throughput amplicon sequencing technologies are not yet able to accurately distinguish between strains within a species, and in wine research strain-level dynamics are particularly important, because different strains are known to contribute differently to the sensory profile of a wine (King et al., 2010; Romano et al., 2003). Additionally, culture-independent methods do not allow researchers to isolate microbes of interest for further research.

Therefore, it is important to conduct culture-dependent analysis alongside these newer technologies. Microsatellites, or short sequence repeats, have been used extensively to distinguish between strains of wine yeasts, including S. cerevisiae

(Legras et al., 2005; Pérez et al., 2001; Richards et al., 2009) and S. uvarum

(Masneuf-Pomarede et al., 2016; Zhang et al., 2015).

Very little research has been conducted on pied de cuve inoculation (Benucci et al., 2016; Clavijo et al., 2011b; Francesca et al., 2016; Li et al., 2012; Marti-Raga et al., 2015; Moschetti et al., 2016; Ubeda Iranzo et al., 2000), and almost no research has been conducted on the method that involves inoculation with must that begins spontaneously fermenting in the vineyard. Furthermore, most of these studies involved the addition of sulfur dioxide (SO2) to the grape must (Benucci et al., 2016; Clavijo et al., 2011b; Li et al., 2012; Moschetti et al., 2016; Ubeda Iranzo

141 et al., 2000), which could obscure results, because non-Saccharomyces yeasts, potentially present in the pied de cuve inoculum, are known to be sensitive to SO2

(Constantí et al., 1998; Divol et al., 2012; Henick-Kling et al., 1998). Here we attempt to narrow this gap in knowledge by testing the effects of vineyard-derived pied de cuve inoculation, paired with different SO2 addition treatments. Finally, few studies have described the diversity of indigenous strains of S. uvarum at a commercial winery (Dellaglio et al., 2003; Demuyter et al., 2004; Masneuf-Pomarede et al., 2016; Naumov et al., 2002, 2000; Zhang et al., 2015), and to our knowledge this is the first study to do so for a winery in North America.

The objective of this study was to determine the effects of SO2 addition at crush and vineyard-derived pied de cuve inoculation on the microbial communities and sensory attributes of commercially-produced Chardonnay wines. Based on the literature, we predicted that SO2 addition would significantly alter the microbial communities conducting AF, as well as the sensory profiles of these wines (Bokulich et al., 2014; Henick-Kling et al., 1998; Morgan et al., 2019). Because vineyard- derived yeasts are more sensitive to SO2 than Saccharomyces yeasts (Constantí et al., 1998; Divol et al., 2012; Henick-Kling et al., 1998), we expected that the pied de cuve inoculation treatment would affect the microbial communities and sensory attributes of the wines that it was added to, but only in the treatments that did not also receive SO2.

142

5.2 Materials and methods

5.2.1 Experimental design and sampling

This experiment was conducted at Mission Hill Family Estate Winery in British

Columbia, Canada. This winery is one of the largest in British Columbia and conducts both uninoculated (spontaneous) and inoculated fermentations of many grape varietals. All samples were taken during the 2015 vintage from stainless steel barrel-fermented Chardonnay. The Chardonnay grapes were sourced from a single vineyard located in Oliver, British Columbia (Vineyard 8). Prior to AF, the must underwent cold-settling in a large stainless steel tank before its distribution into 12 separate new 300 L stainless steel barrels (InovaWine, La Pocatière, QC, Canada), which were steam-cleaned prior to the addition of the experimental grape must.

Stainless steel barrels were chosen because, unlike oak barrels, they do not impart their own sensory character to the wine, obscuring treatment-specific differences in the wines. To each barrel, 150 ppm Lallemand® Fermaid KTM complex yeast nutrient was added prior to the start of AF.

The design of this experiment was a 2 × 2 factorial, where the two factors were (1) SO2 addition and (2) pied de cuve inoculation (Table 5.1). For the SO2 addition factor, two levels of SO2 were added to the Chardonnay must: 0 mg/L or 40 mg/L SO2, added as 80 mg/L potassium metabisulfite (KMS). For the pied de cuve inoculation factor, the must was either left uninoculated, or was inoculated with a pied de cuve mixture of non-Saccharomyces yeasts originating from the vineyard.

Each treatment was performed in triplicate. To achieve the pied de cuve mixture,

Chardonnay grapes from the same vineyard (Vineyard 8) were hand-harvested 2-3

143 days prior to the main harvest. These grapes were crushed by hand in new, cleaned plastic buckets (n = 6), and left to begin fermentation in the vineyard. These fermentations did not come into contact with the winery environment until they were used to inoculate the pied de cuve treatments three days later. Three of the six replicates were selected by the winemakers for their organoleptic qualities, and the other three were discarded. The three selected replicates were combined into one container, and approximately 500 mL of the inoculum was added to the six stainless steel barrels that were designated to the pied de cuve treatments. Samples were taken at four stages of fermentation, as defined by their Brix level: Cold-settling (>

21 Brix , prior to the start of AF), Early (14-19 Brix), Mid (7-12 Brix), and Late (< 2

Brix). Samples were collected in sterile 50 mL centrifuge tubes and were transported on ice back to the laboratory for processing.

Table 5.1. Experimental design. Each of the four treatments was replicated in triplicate.

Uninoculated Pied de cuve inoculation 0 mg/L SO2 0 mg/L Uninoculated 0 mg/L Pied de cuve 40 mg/L SO2 40 mg/L Uninoculated 40 mg/L Pied de cuve

5.2.2 Chemical analysis

Sugar concentrations were measured throughout fermentation, and specific compositional analyses were measured after AF was complete. These compositional analyses included: pH, glucose and fructose, ethanol, titratable acidity, malic acid, and volatile acidity. All compositional analyses were measured 144 using a WineScanTM wine analyzer (Foss, Hilleroed, Denmark). The concentration of glucose, fructose, titratable acidity, malic acid, and volatile acidity were reported in g/L, while ethanol was expressed as % by volume (% v/v).

5.2.3 Saccharomyces strain identification

5.2.3.1 Yeast isolation and DNA extraction

Yeast isolation and DNA extraction was performed as described previously in

Chapter 2. Briefly, samples were plated at three stages of AF, and the DNA from 72 randomly-selected yeast colonies per sample was extracted using a water-extraction method and stored at -20 C.

5.2.3.2 Saccharomyces uvarum strain-typing

Identification of S. uvarum strains was performed using a multiplex PCR reaction that targeted eleven microsatellite loci: L1, L2, L3, L4, L7, L8, L9, NB1,

NB4, NB8, and NB9 (Table 5.2). Two separate Multiplex PCR were performed on the extracted DNA from each yeast isolate. Primers were diluted to a starting concentration of 10 µM (L1, L2, L3, NB1, NB4, NB8, NB9) or 15 µM (L4, L7, L8, L9) before being added to the PCR mix, and equal volumes of forward and reverse primers were added. The first multiplex PCR recipe contained primers for L2 and L9, as follows (volumes are per-reaction): 1 µL post-extraction DNA template (5-50 ng),

5 µL Qiagen® Multiplex PCR Master Mix (2×), 0.24 µL L2 primers (forward and reverse primers each for every primer set), 0.24 µL L9 primers, and topped with

DNA-grade water for a final reaction volume of 10 µL. The second multiplex PCR recipe contained the primers for the other nine loci, as follows (volumes are per- reaction): 1 µL post-extraction DNA template (5-50 ng), 5 µL Qiagen® Multiplex PCR

145

Master Mix (2×), 0.12 µL L1 primers (forward and reverse primers each for every primer set), 0.3 µL L3 primers, 0.03 µL L4 primers, 0.035 µL L7 primers, 0.03 µL L8 primers, 0.04 µL NB1 primers, 0.12 µL NB4 primers, 0.16 µL NB8 primers, 0.16 µL

NB9 primers, and topped with DNA-grade water for a final reaction volume of 10 µL.

PCR for both primer mixes was performed on an Applied Biosystems Veriti Thermal

Cycler (Foster City, CA, USA) with the following parameters: 97°C for 4 min (1 cycle); 95°C for 30 s, 54°C for 60 s, 72°C for 2 min (34 cycles); 72°C for 10 min (1 cycle). Post-PCR samples were submitted to Fragment Analysis and DNA

Sequencing Services at the University of British Columbia (Kelowna, BC, Canada) for fragment analysis on a 3130xl DNA sequencer (Applied Biosystems, Foster City,

CA, USA). GeneMapper 4.0 software (Applied Biosystems, Foster City, CA, USA) was used to determine the fragment size at each locus, and the resulting multilocus genotype (MLG) of each isolate was compared to that of others using Bruvo’s genetic distance (Bruvo et al., 2004), which considers mutational distances in microsatellites. Bruvo’s distance was calculated in RStudio (version 3.5.1) using the

‘poppr’ package (version 2.8.1) (Kamvar et al., 2015, 2014). A distance threshold of

0.3 was applied, and any MLGs that fell within this distance were considered to belong to the same strain. Isolates that only partially amplified were re-analyzed and subsequently excluded from analysis upon a second amplification failure. Isolates that did not amplify were considered to be potential S. cerevisiae strains, and were subjected to a secondary S. cerevisiae strain typing analysis, described in section

5.2.3.3 below.

146

Table 5.2. Primer sequences and characteristics of eleven microsatellite loci used for S. uvarum strain identification (Masneuf-Pomarede et al., 2016; Zhang et al., 2015).

Locus Motif Chromosome Forward primer sequence (5’-3’) Reverse primer sequence (5’-3’) Dye name location L1 GT X CGTGTTGAAGACATAATTG AATCTGAACGACAGGAAT 6-FAM L2 ATT II TGCCCTTCTTATTCTTGT GAAAATATCAACGCATTAAA 6-FAM L3 TA XI GTATGCATCACTATTTTTCG AATTTGGTAATTTGAATGTG 6-FAM L4 CTG XI GGACACTAGAGTTCGTCTCG GCCACCACTATCAGTTCG VIC L7 TC XII GTAGAATTCACCACAGGTC CCGTATATAAAACAGCACTT NED L8 GTT VIII CACGGCAATCAGCACATTT TGAAGTTTCATCATCGGCAA NED L9 ATT IX AAAAAGCAACCTTAAAAGCAACA CTTTACGTAGGCTCATGGCA VIC NB1 ATG X GTGCTCCATGGACTTGTATGAAGCAA GTTCGTTACCTTCAGTGCTC 6-FAM NB4 TGT X GTGCTCGACATTGTAAAAGCACAGCA ACGGGGCTTCTCTAGATATT 6-FAM NB8 TGT XVI GTGCTCTGCATGAAAGATTGTAAAGG TCCACAACGATATCAAGACA 6-FAM NB9 AT XV GTGCTCAAACAAGAAACTGTGGTCGT TGCTTTAATTTCAAGAAACA 6-FAM

5.2.3.3 Saccharomyces cerevisiae strain-typing

Identification of S. cerevisiae strains was performed using a multiplex PCR reaction that targeted eleven hyper-variable microsatellite loci: C3, C4, C5, C8, C11,

SCY, YLR, YML, YPL, ScAAT3, and AT4 (Table 5.3). Multiplex PCR of these 11 loci was performed on the extracted DNA from yeast isolates that did not amplify using the S. uvarum primers. All primers were diluted to a starting concentration of 10 µM before being added to the PCR mix, and equal volumes of forward and reverse primers were added. The multiplex PCR recipe was as follows (volumes are per- reaction): 1 µL post-extraction DNA template (5-50 ng), 5 µL Qiagen® Multiplex PCR

Master Mix (2×), 0.0504 µL C3 primers (forward and reverse primers each for every primer set), 0.36 µL C4 primer, 0.068 µL C5 primer, 0.28 µL C8 primers, 0.075 µL

C11 primers, 0.08 µL SCY primers, 0.072 µL YLR primers, 0.42 µL YML primers,

0.19 µL YPL primers, 0.088 µL ScAAT3 primers, 0.175 µL AT4 primers, and topped with DNA-grade water for a final reaction volume of 10 µL. PCR was performed on an Applied Biosystems Veriti Thermal Cycler (Foster City, CA, USA) with the following parameters: 95°C for 15 min (1 cycle); 94°C for 30 s, 54°C for 90 s, 72°C

147 for 30 s (34 cycles); 68°C for 45 min (1 cycle). Post-PCR samples were submitted to

Fragment Analysis and DNA Sequencing Services at the University of British

Columbia (Kelowna, BC, Canada) for fragment analysis on a 3130xl DNA sequencer

(Applied Biosystems, Foster City, CA, USA). GeneMapper 4.0 software (Applied

Biosystems, Foster City, CA, USA) was used to determine the fragment size at each locus, and the resulting genetic fingerprint of each isolate was compared to the fingerprints of > 200 commercial strains present in a database curated at the

University of British Columbia. If a strain contained two or more loci that did not match those of another strain, it was considered to be a unique strain; fingerprints that did not match any strain in the database were termed ‘unknown’ strains.

Isolates that only partially amplified were re-analyzed and subsequently excluded from analysis upon a second amplification failure. Isolates that did not amplify at all were considered to be non-Saccharomyces yeasts. Although 72 yeast colonies were initially isolated, not all ended up being Saccharomyces strains, and those that were identified as other than S. uvarum or S. cerevisiae were excluded from analysis.

After strain identification, each sample was rarefied to 52 isolates.

Table 5.3. Primer sequences and characteristics of eleven microsatellite loci used for S. cerevisiae strain identification (Legras et al., 2005; Pérez et al., 2001; Richards et al., 2009).

Locus Motif ORFa Forward primer sequence (5’-3’) Reverse primer sequence (5’-3’) Dye name coordinates C3 CAA YGL139w GTGTCTCTTTTTATTTACGAGCGGGCCAT AAATCTCATGCCTGTGAGGGGTAT 6-FAM C4 TAA/ YOL109w AGGAGAAAAATGCTGTTTATTCTGACC TTTTCCTCCGGGACGTGAAATA VIC TAG C5 GT YFR028c GTGTCTTGACACAATAGCAATGGCCTTCA GCAAGCGACTAGAACAACAATCACA VIC C8 TAA YGL014c GTGTCTCAGGTCGTTCTAACGTTGGTAAAATG GCTGTTGCTGTTGGTAGCATTACTGT NED C11 GT YJR044c TTCCATCATAACCGTCTGGGATT TGCCTTTTTCTTAGATGGGCTTTC PET SCY GTT YOR267c TACTAACGTCAACACTGCTGCCAA GGATCTACTTGCAGTATACGGG PET YLR CAG YLR177w CTTAAACAACAGCTCCCAAA ATGAATCAGCGCATCAGAAAT PET YML AAT YML091c GTGTCTAAGCCTCTTCAAGCATGAC GTGTCTGGACAATTTTGCCACCTTA NED YPL CTT YPL009c AACCCATTGACCTCGTTACTATCGT TTCGATGGCTCTGATAACTCCATTC FAM ScAAT3 AAT YDR160w GTGTCTGAGGAGGGAAATGGACAG GCCTGAAGATGCTTTTAG NED AT4 TA YLL049w GCAACATAATGATTTTGAGGT GTGTCTTGTGTGAGCATAGTGGAGAA 6-FAM aOpen Reading Frame

148

5.2.4 High-throughput amplicon sequencing

5.2.4.1 Sample treatment with PMA and DNA extraction

Samples for high-throughput amplicon sequencing (Illumina MiSeq) were taken at four stages of AF: Cold-settling, Early, Mid, and Late. However, only the samples from the Early, Mid, and Late stages received PMA. This is because we considered it likely that most microorganisms would be living during the Cold-settling stage, due to an abundance of nutrients and absence of ethanol, and we also felt it was important to capture the entire microbial community present at the onset of AF.

Aliquots of 10 mL were aseptically taken from each sample and transferred to sterile

15 mL plastic centrifuge tubes. Samples were prepped and treated with approximately 6 mM PMA and DNA was extracted from these samples as described previously in Chapter 3.

5.2.4.2 Illumina MiSeq library preparation

Sample library preparation was conducted using a two-step PCR procedure consisting of ‘Amplicon’ and ‘Index’ PCR reactions, as described previously in

Chapter 3.

5.2.5 Illumina MiSeq data processing

The open source bioinformatics pipeline Quantitative Insights Into Microbial

Ecology (QIIME1 and QIIME2) (Bolyen et al., 2018; Caporaso et al., 2010) was used to analyze the high-throughput metagenomic sequencing data, as described previously in Chapter 3.

For the fungal sequence data, paired-end assembly of forward and reverse reads was performed; forward reads were truncated at 206 bp and reverse reads

149 were truncated at 180 bp. Sequences were classified to the species level using a

99% threshold classifier made with the UNITE (version 7.2) database (Kõljalg et al.,

2013). Sequence variants that could not be classified to the order levels or lower, and those that appeared with a total frequency of < 100 sequences, were excluded from analysis. Samples were rarefied to 20,000 sequences before being exported from QIIME2 for further analysis. We noted that most of the sequences identified were classified as by UNITE. However, based on our culture-dependent analysis of these samples, we are confident that these sequences actually belong to Saccharomyces uvarum. S. bayanus and S. uvarum are synonymous in the UNITE (version 7.2) database, and in the past this was believed to be true, but in recent years S. uvarum has been re-instated as a separate and pure species within the Saccharomyces sensu stricto complex

(Borneman and Pretorius, 2015; Nguyen and Gaillardin, 2005).

For the bacterial sequence data, paired-end assembly was not performed, as the reverse reads were found to be of low quality. Forward reads were truncated at

260 bp, and taxonomy was assigned to sequences using a Greengenes (version

13.5) classifier (DeSantis et al., 2006; McDonald et al., 2012), which was trained on the target region of the amplicon primer sets. Sequence variants that could not be classified to the order levels or lower, and those that appeared with a total frequency of < 100 sequences, were excluded from analysis. Contamination from chloroplast and mitochondrial DNA was also removed. Samples were rarefied to 8,500 sequences before being exported from QIIME2 for further analysis.

150

5.2.6 Wine bottling and sensory evaluation

At the end of AF, 6 L of wine was transferred from each barrel into clean 2 L glass bottles with screw caps (three bottles per barrel), as described previously in

Chapter 3. Briefly, the wine was racked from the barrels to the bottles, 50 mg/L SO2

(150 mg/L KMS) was added to each bottle as a preservative, any headspace was displaced with nitrogen (N2) gas, and the wines were kept in a cool, dark cellar for approximately three months before being transferred to clean 750 mL wine bottles

(six wine bottles per barrel). The wine was filled to the neck of the bottle and N2 gas was added to the headspace to remove any oxygen before the bottles were corked using new corks and a manual floor corker.

Sensory evaluation of the wines was conducted in 2016 at the Summerland

Research and Development Centre (Summerland RDC) in Summerland, British

Columbia, Canada, adhering to standard sensory evaluation protocols and using a panel of 12 industry wine experts (Cliff and Dever, 1996; Guinard and Cliff, 1987;

King et al., 2013). Two wines from each treatment were randomly selected for sensory evaluation, for a total of eight wines (n = 2 per treatment), which were each evaluated in duplicate. Such an approach minimized sensory fatigue and improved the accuracy and reliability of the evaluations. The sensory panel consisted of 12 judges evaluating the wines for their intensity of aroma, flavour, and mouthfeel attributes. The wines were labeled with random three-digit numbers and served in random order using a William’s design to control for first-order bias and carryover effects (Williams and Arnold, 1991). A short break of 5-10 minutes was taken once all the wines had been tasted, and then the wines were evaluated a second time

151 with a different tasting order and different three-digit codes, to ensure panelist consistency. Wines were served in 210 mL International Standards Organization

(ISO) wine glasses in 30 mL aliquots, covered with 6 cm plastic petri dishes. The panelists evaluated the perceived aroma of the wines, followed by the perceived flavour and mouthfeel attributes, using sensory characteristics defined during a prior roundtable discussion (Table 5.4). Sensory standards for panelist reference were created for each of the attributes listed in Table 5.4, and the composition of these standards can be found in Appendix D (Table D1). Panelists were instructed to expectorate all wine samples and rinse their mouths with water between each wine

(panelists were given the option of still or sparkling water). The intensity of each attribute was evaluated on a 15 cm scale. The scale contained ‘low’ and ‘high’ markers located 1.5 cm from each end, but was otherwise unstructured. Panelists marked the scale with a vertical line at the appropriate intensity, and wrote the three- digit sample number above the marking. If wine samples were perceived to have equal intensity for a particular attribute, a single line was drawn and multiple three- digit sample numbers were written above the line. The distance from the left-hand end of the scale (representing zero intensity) to each line was measured in mm for every wine, attribute, panelist, and tasting session. The use of human subjects in this study was approved by the Agriculture and Agri-Food Canada (AAFC) Human

Research Ethics Committee (Certificate of Approval 2016F003) and the UBC

Okanagan Research Ethics Board (Certificate of Approval H16-00628).

152

Table 5.4. Sensory attributes of wines evaluated by an expert panel of 12 judges.

Aroma attributes Flavour attributes Other attributes Citrus aroma Citrus flavour Sweetness Pome fruit aroma Pome fruit flavour Acidity Tropical fruit aroma Tropical fruit flavour Body Melon aroma Melon flavour Length of aftertaste Floral aroma Floral flavour Sulfur aroma Sulfur flavour Earthy aroma Earthy flavour

5.2.7 Statistical analysis

Fermentation kinetics and wine chemistry data were analyzed in RStudio

(version 3.5.1). Residual sugar (Brix) was evaluated over 20 days of fermentation and fermentation progression was compared among treatments by performing a repeated-measures two-factor ANOVA using the “aov” function with ranked data

(data were not normally distributed). A Tukey HSD post-hoc test was subsequently performed to evaluate differences among treatments using the “lme” and “glht” functions in the nlme (version 3.1-137) and multcomp (version 1.4-8) packages

(Hothorn et al., 2008). Fermentation progression was plotted using GraphPad

Prism® software (version 7) (La Jolla, CA, USA). The compositional analyses of the finished wine were compared among treatments by performing two-factor ANOVA and subsequent post-hoc tests when appropriate. Normality was assessed visually and Levene’s test indicated no violation of the assumption of homogeneity of variance for any of the compositional analyses (data not shown).

All microbial data collected (Saccharomyces strain data, fungal Illumina sequencing data, and bacterial Illumina sequencing data) were analyzed separately, due to the variation in sample collection and sample processing. Microbial diversity was analyzed in RStudio (version 3.5.1). Simpson’s Index of Diversity (1-D) was 153 calculated using the “diversity” function in the vegan package (version 2.5-1) and reported ± the standard error of the mean (SEM). Diversity was evaluated at four stages of fermentation for the fungal and bacterial communities, and at three stages of fermentation for the Saccharomyces strains. Diversity was compared among treatments by performing repeated-measures two-factor ANOVA using the “aov” function. Normality was assessed visually and Levene’s test indicated no violation of the assumption of homogeneity of variance for bacterial diversity (F(3,44) = 0.22, p =

0.88), fungal diversity (F(3,44) = 1.27, p = 0.30), or Saccharomyces strain diversity

(F(3,32) = 2.25, p = 0.10). If appropriate, a Tukey honest significant difference

(HSD) post-hoc test was performed to evaluate differences among treatments.

Permutational analysis of variance (PERMANOVA) tests, using Bray-Curtis dissimilarity and Type III sums of squares, were performed to test for differences in microbial community composition among treatments; when appropriate, pairwise comparisons among treatments were performed. The Bray-Curtis dissimilarity index was calculated using untransformed abundance data. PERMDISP tests, using Bray-

Curtis dissimilarity and calculating deviation from centroid, indicated no violation of the assumption of homogeneity of multivariate dispersion for the bacterial community (F(3,44) = 2.20, p = 0.16), the fungal community (F(3,44) = 0.01, p =

0.98), or the Saccharomyces strain community (F(3,32) = 0.84, p = 0.55). Test statistics (F values for PERMDISP and Pseudo-F values for PERMANOVA) were calculated based on 999 permutations of residuals under the full model. Because the design of this study includes both repeated measures and multiple comparisons, the p-values obtained from these PERMANOVA may overestimate differences

154 among treatments. Therefore, when appropriate, we also generated a Principal

Coordinates Analysis (PCoA) in order to visualize the spatial distribution of the microbial community among samples and treatments. Microbial community composition was analyzed and visualized using PRIMER-e© (version 6) software with PERMANOVA+ add-on (Auckland, New Zealand) (Clarke and Gorley, 2006).

Relative abundance of fungi, bacteria, and yeasts was visualized by creating stacked bar charts using GraphPad Prism® (version 7) software (La Jolla, CA, USA).

Statistical analysis of sensory evaluation data was performed in RStudio

(version 3.5.1). The “panelperf” function in the SensoMineR package (version 1.23) was used to evaluate the sensory panel’s performance in its ability to discriminate among products (wines), using three-factor ANOVA. In order to visualize the relationship between the sensory attributes and wines from different inoculation treatments, a principal component analysis (PCA) was performed using the

“averagetable,” “PCA,” and “plot.PCA” functions in the FactoMineR package (version

1.41). A radar plot was generated in Excel 2016 to visualize differences in sensory profiles between the wines; it was converted to a high-resolution image using

IrfanView software (version 4.51). For each sensory attribute, values were standardized using the “STANDARDIZE” function, which is based on the mean and standard deviation of a set of data. Standardizing involved re-scaling each variable to have a mean of zero and a standard deviation of one, which can facilitate the comparison of attributes among treatments.

155

5.3 Results and discussion

5.3.1 Fermentation kinetics and wine chemistry

Alcoholic fermentation progressed at different rates for the different treatments (Figure 5.1). Both treatments to which SO2 was added (40 mg/L

Uninoculated and 40 mg/L Pied de cuve) conducted AF at an identical rate, and completed AF within 13 days (Figure 5.1). The 0 mg/L Uninoculated treatment was slightly slower and completed AF within 17 days, although the difference in fermentation rate among these three treatments was not significant. The 0 mg/L

Pied de cuve treatment, however, had a significantly different fermentation progression compared to that of the other three treatments, with a lag phase of seven days and a much slower rate of fermentation (Figure 5.1). It is likely that the pied de cuve inoculum, which comprised vineyard-derived yeasts, was the cause of this delay in the onset of fermentation in the 0 mg/L Pied de cuve treatment.

Vineyard-derived non-Saccharomyces yeasts are known to have lower fermentative capabilities (Varela and Borneman, 2017), and adding more of these non-fermenting yeasts may have prevented winery-resident yeasts from entering and beginning AF.

The reason this fermentation delay was not observed in the 40 mg/L Pied de cuve treatment was probably because the added SO2 was successfully killing the non-

Saccharomyces yeasts from the pied de cuve inoculum, negating their effect; non-

Saccharomyces yeasts are known to be more sensitive to SO2 than Saccharomyces yeasts (Constantí et al., 1998; Divol et al., 2012; Henick-Kling et al., 1998).

156

Figure 5.1. Progression of alcoholic fermentation, measured as residual sugar content (Brix ± SEM), of Chardonnay to which different SO2 and inoculation treatments were applied at crush (n = 3). The wines from the 0 mg/L Pied de cuve treatment conducted alcoholic fermentation at a significantly slower rate than the other three treatments (p < 0.001).

Table 5.5. Compositional analyses of stainless steel barrel-fermented Chardonnay wines to which different sulfite and inoculation treatments were applied at crush. Values are the mean ± SEM (n = 3 per treatment). Superscript letters represent the results of two-factor ANOVA and subsequent post-hoc tests performed for each analysis (α = 0.05); different superscript letters within a row indicate significantly different means.

0 mg/L 40 mg/L 0 mg/L 40 mg/L Uninoculated Uninoculated Pied de cuve Pied de cuve pH 3.53 ± 0.02a 3.62 ± 0.06a 3.56 ± 0.04a 3.56 ± 0.13a Glucose (g/L) 0.03 ± 0.03a 0.00 ± 0.00a 0.10 ± 0.00a 0.10 ± 0.00a Fructose (g/L) 2.70 ± 0.72b 0.60 ± 0.15b 9.40 ± 0.93a 1.67 ± 1.1b Ethanol (% v/v) 13.60 ± 0.06a 13.70 ± 0.06a 13.20 ± 0.06b 13.60 ± 0.12a Titratable Acidity (g/L) 6.07 ± 0.15b 7.17 ± 0.09a 6.27 ± 0.03b 7.20 ± 0.17a Malic Acid (g/L) 0.56 ± 0.02b 2.34 ± 0.01a 0.43 ± 0.06b 2.38 ± 0.02a Volatile Acidity (g/L) 0.20 ± 0.01a 0.19 ± 0.00a 0.20 ± 0.00a 0.20 ± 0.01a

At the end of AF, pH, glucose, and volatile acidity concentrations were similar among all treatments (Table 5.5), while fructose, ethanol, titratable acidity, and malic acid differed among the treatments. In the 0 mg/L Pied de cuve treatment, fructose 157 was significantly higher and ethanol was significantly lower, indicating that AF did not fully complete in these wines. Titratable acidity and malic acid were significantly lower in both treatments that did not receive SO2 (0 mg/L Uninoculated and 0 mg/L

Pied de cuve)—the malic acid was mostly degraded by the end of AF in these wines, indicating that bacteria in both treatments were conducting MLF at the same time the yeasts were conducting AF. The finding that malic acid was not degraded in either of the 40 mg/L treatments suggests that the bacteria responsible for this malic acid conversion were particularly sensitive to SO2; bacteria in general are more sensitive to SO2 than yeasts, including LAB (Fang and Dalmasso, 1993; Sun et al.,

2016).

5.3.2 Fungal communities

Fungal community diversity was not significantly different among treatments (Table

5.6). Fungal diversity varied throughout AF, with the highest diversity observed at the Cold-settling stage. This is expected, because the Cold-settling stage often includes many different species of yeast and fungi that subsequently perish during

AF. This may be due to competition or lack of tolerance to ethanol, and leaves only a few fermentative species to conduct AF.

158

Table 5.6. Fungal, bacterial and Saccharomyces strain diversity, measured as Simpson’s Index of Diversity (1-D), of stainless steel barrel-fermented Chardonnay to which different SO2 and inoculation treatments had been applied. Diversity ± SEM was measured at four stages of fermentation for the fungal and bacterial community (Cold-settling, Early, Mid, and Late), and across three stages for the Saccharomyces strains (Early, Mid, and Late) (n = 3 per treatment). Diversity was analyzed separately for the fungal, bacterial, and Saccharomyces data, and treatments were compared across fermentation stages by performing repeated measures two-factor ANOVA and subsequent post-hoc tests when appropriate. Treatments marked with different superscript letters had significantly different overall diversity across all stages (α = 0.05).

Fungi Cold-settling Early Mid Late 0 mg/L Uninoculateda 0.66 ± 0.03 0.25 ± 0.10 0.41 ± 0.04 0.23 ± 0.06 40 mg/L Uninoculateda 0.44 ± 0.01 0.11 ± 0.05 0.34 ± 0.01 0.29 ± 0.05 0 mg/L Pied de cuvea 0.65 ± 0.03 0.19 ± 0.05 0.22 ± 0.04 0.29 ± 0.10 40 mg/L Pied de cuvea 0.45 ± 0.02 0.27 ± 0.09 0.34 ± 0.02 0.39 ± 0.10 Bacteria Cold-settling Early Mid Late 0 mg/L Uninoculateda 0.50 ± 0.04 0.59 ± 0.06 0.59 ± 0.05 0.46 ± 0.05 40 mg/L Uninoculateda 0.51 ± 0.05 0.56 ± 0.05 0.53 ± 0.04 0.47 ± 0.04 0 mg/L Pied de cuvea 0.49 ± 0.04 0.56 ± 0.06 0.55 ± 0.05 0.55 ± 0.01 40 mg/L Pied de cuvea 0.49 ± 0.05 0.50 ± 0.06 0.52 ± 0.04 0.47 ± 0.06 Saccharomyces Cold-settling Early Mid Late 0 mg/L Uninoculateda - 0.92 ± 0.02 0.93 ± 0.01 0.92 ± 0.004 40 mg/L Uninoculateda - 0.92 ± 0.01 0.94 ± 0.01 0.92 ± 0.02 0 mg/L Pied de cuveb - 0.88 ± 0.01 0.84 ± 0.04 0.88 ± 0.02 40 mg/L Pied de cuvea - 0.93 ± 0.01 0.95 ± 0.01 0.93 ± 0.01

The overall fungal community composition did not differ among treatments, when compared across all fermentation stages (Table 5.7). However, the composition of the Cold-settling samples differed markedly from the composition of the three fermentation stages (Early, Mid, and Late) for all treatments (Figure 5.2).

The relative abundance of individual taxa was not statistically evaluated among treatments, but some trends were observed. For all treatments, Aureobasidium pullulans was the most abundant species identified both at the Cold-settling stage and in the pied de cuve inoculum. A. pullulans is commonly associated with vineyard and winemaking environments, and is a ubiquitous environmental yeast-like fungus 159

(Takahashi et al., 2014; Varela and Borneman, 2017). In the two 0 mg/L treatments

(0 mg/L Uninoculated and 0 mg/L Pied de cuve), the second-most abundant species was Aspergillus flavus, followed by one or more species of Hanseniaspora. In the treatments to which SO2 was added (40 mg/L Uninoculated and 40 mg/L Pied de cuve), however, A. flavus was present in low relative abundance, indicating that it was at least partly inhibited by the SO2 addition at crush. A. flavus was also present in low relative abundance in the pied de cuve inoculum, which could suggest that the increase in A. flavus seen in the 0 mg/L treatments was a result of rapid growth in the vineyard between the time when the pied de cuve inoculum was harvested and when the entire vineyard was harvested. Although A. flavus is known to produce mycotoxins, which have potentially harmful health effects, it is a commonly-observed fungus on wine grapes, and Canadian wines have been shown to contain lower levels of mycotoxin contamination than imported wines (Freire et al., 2017; Mateo et al., 2007; Soleas et al., 2001). Additionally, the process of fermentation reduces the presence of any mycotoxins, through denaturation or adsorption to the lees (Inoue et al., 2013; Petruzzi et al., 2014). Mycosphaerella tassiana was also observed in the pied de cuve inoculum, but was present in extremely low abundance at the Cold- settling stage for all treatments. M. tassiana is also known as Davidiella tassiana, and is a black mould of cereals. M. tassiana is an antagonist of , which causes grey mould, and is also sensitive to inhibition by yeasts (Hilber-

Bodmer et al., 2017). Other common wine yeasts were identified in low relative abundance, including Kazachstania aerobia, Wickerhamomyces anomalus, Candida spp., and Torulaspora delbrueckii (Table D2).

160

Table 5.7. Results of two-factor PERMANOVA evaluating the effects of SO2 addition (0 or 40 mg/L) and inoculation treatments (Uninoculated or Pied de cuve inoculation) on the microbial communities of fungi, bacteria, and Saccharomyces strains. Results with an asterisk (*) are significant at α = 0.05.

Community Source d.f. SS MS Pseudo-F p

SO2 addition (S) 1 510.68 510.68 0.26698 0.759 Inoculation treatment 1 34.654 34.654 0.01811 0.965 (I) Fungi S × I 1 299.85 299.85 0.15676 0.784 Residual 44 84162 1912.8 Total 47 85007 SO2 addition (S) 1 745.51 745.51 6.4831 0.008* Inoculation treatment 1 13.262 13.262 0.11553 0.911 (I) Bacteria S × I 1 1.2496 1.2496 0.01086 0.969 Residual 44 5059.7 114.99 Total 47 5819.7 SO2 addition (S) 1 14802 14802 9.923 0.001* Inoculation treatment 1 2161.7 2161.7 1.4492 0.065 Saccharomyces (I) strains S × I 1 2113 2113 1.4166 0.075 Residual 32 47733 1491.6 Total 35 66809

161

Figure 5.2. Relative abundance (± SEM) of the dominant fungi present in Chardonnay at four stages of fermentation, as well as the composition of the pied de cuve inoculum, based on 20,000 sequences per sample. SO2 and inoculation treatments were as follows (n = 3 per treatment): (A) 0 mg/L Uninoculated; (B) 40 mg/L Uninoculated; (C) 0 mg/L Pied de cuve; (D) 40 mg/L Pied de cuve. Any fungal taxa that did not achieve at least 10% relative abundance in any one sample were termed Minor Fungi.

162

The three stages of AF (Early, Mid, and Late) were dominated by S. uvarum, comprising between 52-97% of the relative abundance of every sample, and maintaining dominance through to the end of AF (Figure 5.2). S. cerevisiae was present in all treatments and all fermentation stages (0.1-25% relative abundance, depending on fermentation stage and treatment), but was unable to out-compete S. uvarum. This dominance of S. uvarum was not affected by the SO2 addition or the pied de cuve inoculation treatments. This result was very unexpected, and to our knowledge this is the first report of S. uvarum dominating and completing fermentations at a commercial winery in North America. S. uvarum has previously been observed dominating fermentations in France (Demuyter et al., 2004; Naumov et al., 2000), Hungary (Naumov et al., 2002; Sipiczki et al., 2001), Slovakia (Naumov et al., 2002), and northern Italy (Dellaglio et al., 2003), and has also been identified in fermentations in Canada (see Chapter 3 results) and New Zealand (Zhang et al.,

2015). S. uvarum is described as a cryotolerant (Almeida et al., 2014) or cryophilic

(Naumov et al., 2000) yeast, and has been shown to grow well at lower temperatures (Dellaglio et al., 2003; Sipiczki et al., 2001). It is most commonly identified in wine regions with cooler climates, and is almost always associated with low-temperature fermentations. In this study, fermentations were carried out at cellar-temperature, with the grape must starting at 12.9 C during the Cold-settling stage, and reaching a maximum temperature of 19.6 C during the Mid stage of AF.

No heat was applied to these fermentations; instead, the increase in temperature observed was a result of increased yeast metabolic activity during AF. The winery where this study was conducted had decades earlier been used for cider production.

163

Because S. uvarum has been associated with cider fermentations (Almeida et al.,

2014; Coton et al., 2006; Suárez Valles et al., 2007), it is possible that indigenous strains of S. uvarum established themselves as residents of this facility during this time. It is also possible that these strains originated in the vineyard, as a very small proportion of S. uvarum (0.10 ± 0.05%) was identified in the pied de cuve inoculum

(Table D2). However, more research is needed to determine the origins of this yeast.

5.3.3 Saccharomyces strains

Yeast isolates belonging to both S. uvarum and S. cerevisiae were identified to the strain level. Because these strains were all part of the 72 yeast isolates per sample that were strain-typed, they were treated as a group when rarefying and when analyzing the diversity and composition of these strains. Therefore, we will refer to the combination of S. uvarum and S. cerevisiae strains as ‘Saccharomyces strains’ for convenience. Saccharomyces strain diversity was significantly lower in the 0 mg/L Pied de cuve treatment, as compared to the other three treatments (Table 5.6).

Nevertheless, strain diversity was very high for all treatments, and this diversity did not decrease throughout AF; this meant a high diversity of Saccharomyces strains conducted AF and were able to persist through to the end of AF without one strain out-competing the rest. Surprisingly, 150 unique S. uvarum strains were identified across the 12 fermentations sampled in this study (Table D3). Previous studies have reported no more than 89 unique S. uvarum genotypes isolated from must, wine, or other sources (Dellaglio et al., 2003; Demuyter et al., 2004; Masneuf-Pomarede et al., 2016, 2007; Naumov et al., 2002, 2000; Tosi et al., 2009; Zhang et al., 2015).

164

However, these studies used less robust strain-typing techniques and fewer yeast isolates (a maximum of 114), even though some studies analyzed yeast isolates from around the world, not just from a single environment. Our study strain-typed

1860 S. uvarum isolates using 11 microsatellite loci, the most of any study involving

S. uvarum. A relatively conservative cut-off of 0.3 was used when calculating

Bruvo’s distance, so we are confident in our strain assignment. Previously, a cut-off of 0.25 has been used to strain-type S. cerevisiae (Martiniuk et al., 2016). We also observed a much higher occurrence of heterozygosity as compared to previous studies: 42.7% of the strains identified in this study contained at least one heterozygous locus, with five being the average number of heterozygous loci.

Previous studies have identified heterozygosity in 28.8% (Masneuf-Pomarede et al.,

2016), 23.1% (Zhang et al., 2015), and 0% (Masneuf-Pomarede et al., 2007) of S. uvarum strains associated with grapes and wine. L9 was by far the most distinguishing locus, and contained 15 different alleles. All other loci contained 3-4 alleles, with the exception of L8, which contained six. Of the 150 strains identified, only four were able to reach a relative abundance of at least 10% in any sample

(Figure 5.3). Two of these strains (strains 1 and 3) were homozygous for all loci, and two other strains (strains 2 and 4) were heterozygous at five and four loci, respectively.

165

Figure 5.3. Relative abundance (± SEM) of Saccharomyces strains present at three stages of fermentation of Chardonnay must. SO2 and inoculation treatments were as follows (n = 3 per treatment): (A) 0 mg/L Uninoculated; (B) 40 mg/L Uninoculated; (C) 0 mg/L Pied de cuve; (D) 40 mg/L Pied de cuve. Any S. uvarum strains that did not achieve at least 10% relative abundance in any one sample were termed Minor S. uvarum strains.

166

The composition of the Saccharomyces strains differed by SO2 addition level, but not by Pied de cuve treatment (Table 5.7). A PCoA ordination showed a clear separation between samples from different SO2 treatments, with only a single- sample overlap (Figure 5.4). Although it was not statistically significant (p = 0.065), there was a trend towards a separation of strain composition between the 0 mg/L

Uninoculated and the 0 mg/L Pied de cuve treatments (Figure 5.4). This trend was not observed between the 40 mg/L Uninoculated and the 40 mg/L Pied de cuve treatments, which had completely overlapping strain compositions. In the 0 mg/L

Uninoculated and 0 mg/L Pied de cuve treatments, S. uvarum strain 1 was the most commonly identified strain (Figure 5.3). S. uvarum strain 1 was much less abundant in the 40 mg/L Uninoculated and 40 mg/L Pied de cuve treatments; instead, S. uvarum strain 2 and S. uvarum strain 4 had increased relative abundance (Figure

5.3). More research should be conducted to investigate the oenological potential and origins of these dominant S. uvarum strains. We do not expect that the S. uvarum strains present in this winery are commercial, because only one commercial S. uvarum strain existed when this study was conducted (Velluto BMV58TM), which has never been used at this winery, and to our knowledge, this strain was also not available in Canada when this study was being conducted.

Five strains of S. cerevisiae were identified across all fermentations: two commercial strains (Vitilevure® 3001TM and Lalvin® ICV D254TM) as well as three strains of unknown origin. The strain 3001 was used in the winery during the 2015 vintage (the year this study was conducted), and D254 was last used during the 2014 vintage.

Previous studies have identified D254 as a competitive winery resident, able to enter

167 and dominate uninoculated fermentations (Lange et al., 2014). However, all S. cerevisiae strains were isolated in extremely low abundance, and collectively never made up more than 5.8% relative abundance in any sample.

Figure 5.4. PCoA ordination of the Saccharomyces strain composition in wines to which different levels of SO2 were added at crush, and to which different inoculation treatments were applied. Individual data points represent the composition of Saccharomyces strains (both S. uvarum and S. cerevisiae) in a single sample. Samples were collected at three stages of alcoholic fermentation and each treatment contained three replicates, for a total of nine samples per treatment.

5.3.4 Bacterial communities

Bacterial community diversity was not significantly different among treatments

(Table 5.6). Diversity was also constant throughout AF, possibly because bacteria

168 are known to be less active than yeasts during AF, and usually only begin to increase their growth and metabolism during MLF or after the death of yeasts, when there is less competition and when nutrient levels increase due to yeast autolysis

(Arnink and Henick-Kling, 2005).

Even though the bacterial community diversity did not differ significantly among treatments, the composition of the bacterial community was significantly affected by

SO2 addition level, but not by pied de cuve inoculation (Table 5.7). These results are similar to those found by Bokulich et al. (2014), who found differences in the bacterial community, but not the fungal community, in response to SO2 addition. In all treatments, a member of the genus Enterococcus dominated, followed by a member of the family Bacillaceae (Figure 5.5). This result has been reported previously in spontaneous fermentations of Pinot gris at a winery in British Columbia

(see Chapter 3 results). Enterococci have also been identified in red wine fermentations around the world (Bokulich et al., 2012; Capozzi et al., 2011; Dündar,

2016; Pérez-Martín et al., 2014), and are generally considered to be environmental microorganisms, isolated from many different sources, similar to the yeast-like fungus Aureobasidium pullulans (Giraffa, 2002; Prasongsuk et al., 2018). They are also potential spoilage bacteria in wine, as they can produce tyrosine, which can cause headaches when consumed (Dündar, 2016). However, food-associated

Enterococcus spp. have never been associated with clinical Enterococcus infections

(Giraffa, 2002; Pérez-Martín et al., 2014).

169

Figure 5.5. Relative abundance (± SEM) of the dominant bacteria present in Chardonnay at four stages of fermentation, as well as the composition of the pied de cuve inoculum. SO2 and inoculation treatments were as follows (n = 3 per treatment): (A) 0 mg/L Uninoculated; (B) 40 mg/L Uninoculated; (C) 0 mg/L Pied de cuve; (D) 40 mg/L Pied de cuve. Any bacterial taxa that did not achieve at least 10% relative abundance in any one sample were termed Minor Bacteria.

170

The most notable bacteria in this study could only be identified to the family- level (Enterobacteriaceae) when using the Greengenes database (Figure 5.5).

Because we were unable to perform paired-end assembly of the forward and reverse bacterial sequences in this study, the sequences that were used for taxonomic assignment via the Greengenes database were relatively short (260 bp), which limited the specificity in taxonomic assignment. However, when we compared the taxonomy table generated in QIIME2 to our representative sequences, we determined that three amplicon sequence variants were grouped together to form the Enterobacteriaceae group seen in Figure 5.3. We determined via NBCI BLAST® that > 97% of the sequences within this group belonged to the species Tatumella ptyseos (Accession no. LS483499.1), with the remaining < 3% belonging to two other Tatumella species: T. saanichensis (Accession no. JQ726633.1) and T. terrea

(Accession no. AB907784.1). The Tatumella spp. in this study were present in the pied de cuve inoculum, as well as the two 0 mg/L SO2 treatments (0 mg/L

Uninoculated and 0 mg/L Pied de cuve), but were seemingly inhibited by the addition of SO2. This observation is supported by Takahashi et al. (2014), who noted that T. terrea growth was inhibited by 75 mg/L SO2 added at crush. Tatumella spp. were first identified in association with wine grapes in 2011 in Greece (Nisiotou et al., 2011), and have since been identified in wine fermentations from Italy

(Maragkoudakis et al., 2013), Japan (Takahashi et al., 2014), Portugal (Pinto et al.,

2015), and Slovakia (Godálová et al., 2016), although this is the first time Tatumella spp. have been reported in North America. While T. ptyseos and T. terrea have been previously identified in wine fermentations around the globe, to our knowledge

171 this is the first time T. saanichensis has been associated with wine grapes and wine fermentations. T. saanichensis has previously only been identified in association with cocoa bean fermentations, and it was suggested that it may be involved in the production of lactic acid (Papalexandratou et al., 2013). Interestingly, the two treatments that contained Tatumella spp. (0 mg/L Uninoculated and 0 mg/L Pied de cuve) were also the two treatments that underwent MLF at the same time as AF

(Table 5.5). Because the presence of Tatumella spp. was the only notable difference in bacterial community composition among the treatments of this study, we suggest that these species/strains of Tatumella may be responsible for the conversion of malic acid to lactic acid in the 0 mg/L Uninoculated and 0 mg/L Pied de cuve treatments. More traditional lactic acid bacteria, including Leuconostoc sp. and

Lactococcus sp., were identified in these fermentations, but were present in very low relative abundance, and their numbers did not differ among treatments (Table D4).

This suggests that these traditional LAB were not responsible for MLF in the 0 mg/L treatments.

5.3.5 Wine sensory attributes

A sensory panel, consisting of 12 judges, evaluated the wines from all four treatments (n = 2 per treatment). The panel identified significant differences in 14 of the 18 attributes (p < 0.05): citrus aroma/flavour, pome fruit aroma/flavour, tropical fruit aroma/flavour, floral flavour, earthy aroma/flavour, sulfur aroma/flavour, acidity, sweetness, and body. Results from the three-factor ANOVA performed can be viewed in Appendix D (Table D5). The two wines that received SO2 at crush (40 mg/L Uninoculated and 40 mg/L Pied de cuve) had extremely similar sensory

172 profiles, and were rated as being higher in citrus flavour, earthy aroma and flavour, sulfur aroma and flavour, and acidity (Figures 5.6 and 5.7). Alternatively, while the 0 mg/L Uninoculated and 0 mg/L Pied de cuve treatments were more similar to each other than they were to the treatments that received SO2, they did have distinct sensory profiles themselves. This suggests that for the 0 mg/L Pied de cuve treatment, the pied de cuve inoculum was able to alter the sensory profile of the wines when it was not inhibited by the addition of SO2. The 0 mg/L Pied de cuve treatment was rated highest for body and sweetness, possibly due to the residual fructose that was not completely fermented in this treatment (Table 5.2).

Interestingly, this treatment was also rated to be lowest in citrus flavour and acidity, and highest in floral flavour (Figure 5.6). These five attributes (body, sweetness, citrus flavour, acidity, and floral flavour) distinguished the 0 mg/L Pied de cuve wines from the 0 mg/L Uninoculated wines, which were otherwise evaluated quite similarly: both the 0 mg/L Uninoculated and the 0 mg/L Pied de cuve treatments were rated to be highest in pome fruit aroma/flavour and tropical fruit aroma/flavour. This more fruit-forward character in the wine is consistent with previous findings where juice or must did not receive SO2 at crush (Henick-Kling et al., 1998 and results from

Chapter 3).

173

Figure 5.6. Radar plot depicting the standardized relative intensity of sensory attributes of Chardonnay wines to which different levels of SO2 and different inoculation practices were implemented at crush (n = 2 per treatment). Wines were evaluated by a panel of 12 industry experts. Values were standardized separately for each sensory attribute. Attributes with an asterisk (*) were found to be significantly different among treatments (p < 0.05).

174

Figure 5.7. Principal component analysis (PCA) of the sensory attributes of Chardonnay wines to which different levels of SO2 and different inoculation practices were implemented at crush (n = 2 per treatment). Wines were evaluated by a panel of 12 industry experts. The numbers in brackets indicate the two replicate wines within each treatment. The first dimension (Dim 1) represents 75.94% of total variance, while the second dimension (Dim 2) represents 8.63% of total variance.

175

The decreased acidity observed in the 0 mg/L Pied de cuve and 0 mg/L

Uninoculated wines was due to the lower concentrations of titratable acidity and malic acid present in these wines at the end of fermentation (Table 5.2). This is because these wines also underwent spontaneous MLF, presumably conducted by

Tatumella bacteria (see previous discussion in section 5.3.3). MLF converts the tart- tasting malic acid to the softer, less acidic lactic acid, hence the decrease in acidity.

Interestingly, ‘butter’ was not a sensory attribute chosen as a representative descriptor for these wines, despite the fact that it is often associated with wines that have undergone MLF (Bartowsky and Henschke, 2004). Citric acid fermentation, often occurring alongside MLF, produces diacetyl, which is responsible for the buttery taste and smell in wine. It is possible that the Tatumella spp., which presumably conducted MLF in these wines, did not conduct citric acid fermentation concurrently, although more research is needed to confirm this. Unfortunately, citric acid was not monitored in this study. Although the 0 mg/L Pied de cuve and 0 mg/L

Uninoculated wines had similar titratable acidity and malic acid concentrations

(Table 5.2), the 0 mg/L Pied de cuve wines were perceived as being less acidic

(Figure 5.6), likely due to the higher concentration of fructose present in these wines

(Stampanoni, 1993).

The yeasts and fungi present in these fermentations likely contributed to the sensory profiles of these wines. A. pulluans and Hanseniaspora spp., both present at the Cold-settling stage, can produce β-glucosidase and protease enzymes, which can react with grape precursor compounds and contribute to the enhancement of varietal aromas (Varela and Borneman, 2017). Different S. cerevisiae strains can

176 produce different concentrations of aroma-active compounds such as isobutanol, acetaldehyde, n-propanol, acetic acid, and isoamyl alcohol (Fleet, 2003; Romano et al., 2003). S. uvarum is known to produce less acetic acid and acetaldehyde and more glycerol, , isoamyl alcohol, isobutanol, and ethyl acetate as compared to S. cerevisiae (Castellari et al., 1994; Magyar and Tóth, 2011; Sipiczki et al., 2001). S. uvarum has also displayed higher esterase activity than S. cerevisiae at low fermentation temperatures, which can impact the presence of fruity aromas (Tosi et al., 2009). Like S. cerevisiae, the relative production of these compounds in S. uvarum also varies greatly by strain (Castellari et al., 1994). At lower concentrations, as is found in most wines, glycerol and isoamyl alcohol provide a sweet taste and smell, respectively, and isobutanol and ethyl acetate provide an alcohol/solvent and fruity aroma, respectively. S. uvarum tends to be a less efficient fermenter than S. cerevisiae, meaning it produces less ethanol and more secondary metabolites per gram of sugar (Magyar and Tóth, 2011), and has been suggested for use in wines where a reduced alcohol content is desired

(Contreras et al., 2015; Varela et al., 2017). Fermentations conducted with a higher diversity of species and strains are generally considered to be more complex and desirable than single-strain fermentations (Albertin et al., 2017; Suárez-Lepe and

Morata, 2012). Fermentations with two or more S. cerevisiae strains can produce wines with significantly different sensory profiles as compared to single-strain fermented wines (King et al., 2010, 2008). Although this has not been studied yet with regards to strains of S. uvarum, it is likely that the same holds true. More

177 research is needed to determine the specific contribution of this S. uvarum consortium to the wines’ sensory profiles.

5.4 Summary

This study investigated the effects of SO2 addition and vineyard pied de cuve inoculation on the microbial communities and sensory attributes of commercially- produced Chardonnay wines. Using modern molecular techniques, including microsatellite analysis as well as high-throughput amplicon sequencing combined with PMA addition, we were able to identify the breadth and depth of the consortium of bacteria and fungi present in these fermentations. SO2 addition level was found to significantly affect the bacterial community; Tatumella spp. were able to survive when no SO2 was added, likely resulting in MLF being conducted at the same time as AF in these treatments. SO2 addition also significantly altered the

Saccharomyces strain composition, with different S. uvarum strains playing dominant roles in the fermentations with different SO2 treatments. The pied de cuve inoculum was only able to survive when no SO2 was added, and its effects were limited—wines from this treatment underwent a considerable lag period before commencing AF, and did not fully complete AF, resulting in wines with higher sweetness and body and lower acidity than the wines from the other three treatments. Both the 0 mg/L Uninoculated and 0 mg/L Pied de cuve wines were considered fruitier, with higher pome fruit aroma/flavour, tropical fruit aroma/flavour, and citrus aroma than those that did not receive SO2. The wines from both treatments that received SO2 at crush were rated higher in the less desirable characteristics, including sulfur aroma/flavour and earthy aroma/flavour. The

178 successful use of a pied de cuve inoculum composed of vineyard-derived yeasts and bacteria is highly dependent on the microbial composition of the inoculum, and we recommend that winemakers take precautions when performing these inoculation techniques. For example, performing a pied de cuve inoculation on a small subset of fermentations that can later be used for blending and increasing complexity. For the first time, we report the dominance of a highly genetically diverse population of indigenous S. uvarum in a commercial winery in North America; research is on-going to investigate the origins and oenological potential of these yeast strains.

179

Chapter 6: Competition between Saccharomyces cerevisiae and

Saccharomyces uvarum in controlled Chardonnay fermentations

6.1 Background

Although Saccharomyces cerevisiae is usually the dominant yeast in winemaking, other species, including Saccharomyces uvarum, are able to conduct

(and complete) alcoholic fermentation; these yeasts may even compete with S. cerevisiae for dominance in commercial wineries. S. uvarum has been identified dominating fermentations in winery environments around the world, usually associated with cool-climate regions and low-temperature fermentations (Dellaglio et al., 2003; Demuyter et al., 2004; Naumov et al., 2002, 2000; Sipiczki et al., 2001).

Fermentations conducted by S. uvarum tend to have less alcohol, acetic acid, and acetaldehyde, as well as more glycerol, succinic acid, malic acid, isoamyl alcohol, isobutanol, and ethyl acetate, as compared to fermentations conducted by S. cerevisiae (Castellari et al., 1994; Gamero et al., 2013; Magyar and Tóth, 2011;

Sipiczki et al., 2001). Additionally, because these fermentations are usually conducted at lower temperatures, they can produce wines with more balanced aroma profiles (Deed et al., 2017).

While the progression of fermentation usually involves one yeast strain dominating by the end of fermentation, this is not always the case, and we observed many different S. uvarum strains co-fermenting Chardonnay must at commercial winery in Chapter 5. However, some of these strains were found in higher proportion than others, and one such strain, termed ‘S. uvarum strain 1,’ was present with a relative abundance of up to 46% in some fermentations. This indigenous S. uvarum

180 strain was apparently able to out-compete the other S. uvarum strains, as well as any winery-resident S. cerevisiae strains present, to achieve this position in the fermentations. This was the first report of such an occurrence in North America, and a further investigation into the enological potential and competitive abilities of this yeast strain was warranted.

Traditionally, fermentations were carried out uninoculated, or spontaneously, with yeasts from the vineyard and winery environments conducting fermentation.

Usually these fermentations involved a succession of yeast species and strains, and resulted in one or a few main strains dominating by the end of fermentation.

However, the high diversity of yeasts throughout fermentation resulted in wines with more complex aroma and flavour profiles. Today, most winemakers choose to inoculate their fermentations with a single commercial yeast strain⎯ these inoculated fermentations pose less of a risk of stuck or spoiled fermentations, but they can also lack the complexity of multi-yeast fermentations (Bagheri et al., 2018; King et al.,

2010; Zara et al., 2014). Increasingly, consumers are interested in unique wines, as well as wines that are more fruit-forward and contain less alcohol. Indigenous strains of S. uvarum may provide a solution to these challenges, particularly if used in combination with other yeast strains, because they are able to ferment at lower temperatures, which can anecdotally preserve fruity aromas, and are also known to produce wines with lower alcohol contents (Magyar and Tóth, 2011). We must first, however, understand how indigenous S. uvarum strains behave in fermentation and in competition with other yeasts.

181

Previously, research has been conducted to investigate the enological potential of S. uvarum (Castellari et al., 1994; Gamero et al., 2013; Magyar and

Tóth, 2011; Sipiczki et al., 2001), but to our knowledge only two studies have been conducted that investigate the ability of this species to compete against strains of S. cerevisiae commonly found in the winery environment (Alonso-del-Real et al., 2017;

Cheraiti et al., 2005). However, neither study investigated different co-inoculation ratios, and the S. uvarum strain used by Cheraiti et al. (2005) was actually a S. cerevisiae-S. uvarum hybrid. Additionally, both studies used a synthetic fermentation medium as opposed to real wine grape juice, and Cheraiti et al. (2005) investigated only one fermentation temperature (24 C). Here, we conducted controlled

Chardonnay fermentations of an indigenous S. uvarum strain, co-inoculated at different ratios with a commercial S. cerevisiae strain. Because S. uvarum is a cryotolerant yeast, it was also important to test this yeast’s competitive abilities at the upper and lower temperature ranges for white wine fermentations.

The overall objective of this study was to observe the competitive interactions between an indigenous strain of S. uvarum and a commercial strain of S. cerevisiae in controlled Chardonnay fermentations. This was achieved by inoculating these yeasts at different initial ratios, and following their relative abundance over the course of the 10 day fermentations. Additionally, each inoculation ratio was fermented at two temperatures, to observe whether a lower temperature increased the competitive abilities of the cryotolerant S. uvarum strain. At the end of the experiment, wine samples were taken for volatile metabolite analysis, to determine whether each yeast produced a unique volatile profile. We predicted that the yeast

182 strain with the highest initial concentration in each treatment would end up dominating the fermentation regardless of fermentation temperature. When inoculated at an equal concentration, we predicted the S. cerevisiae strain would dominate at a higher fermentation temperature, but the S. uvarum strain would dominate at a lower fermentation temperature. Finally, we predicted that each strain would produce different volatile metabolite profiles, and that fermentation temperature would also affect the production of these metabolites.

6.2 Materials and methods

6.2.1 Experimental design

This experiment was conducted in 2016 at the School of Agriculture, Food, and Wine at the University of Adelaide. A commercial strain of S. cerevisiae (Lalvin®

QA23) and an indigenous strain of S. uvarum (SU01) were inoculated into sterile- filtered Chardonnay juice in five different ratios in order to observe their competitive abilities under controlled fermentations (Table 6.1). Fermentations were conducted at both 24 C and 15 C, for a total of 10 treatments (n = 3 per treatment). The S. uvarum strain selected (SU01) was originally isolated from stainless steel barrel- fermented Chardonnay in 2015 at Mission Hill Family Estate Winery in the

Okanagan Valley wine region of Canada. SU01 (called ‘S. uvarum strain 1’ in

Chapter 5) was the most frequently identified strain in these fermentations, which were dominated by S. uvarum, even though this winery uses many different commercial S. cerevisiae strains. QA23 was selected because it is recommended for white wine fermentations, including Chardonnay, and is marketed as being able

183 to ferment at low temperatures. QA23 was also used to inoculate fermentations at

Mission Hill in 2015, the year SU01 was identified.

Table 6.1. Experimental design indicating the inoculation ratios and fermentation temperatures of laboratory-scale Chardonnay fermentations. Inoculation ratios are indicated in each case as the ratio of S. cerevisiae : S. uvarum. For every treatment, the total concentration of the yeast inoculum was 2.5 × 106 cells/mL. Each treatment contained three replicate fermentations.

Ratio Temperature 1:0 10:1 1:1 1:10 0:1 24 C 1:0 (24 C) 10:1 (24 C) 1:1 (24 C) 1:10 (24 C) 0:1 (24 C) 15 C 1:0 (15 C) 10:1 (15 C) 1:1 (15 C) 1:10 (15 C) 0:1 (15 C)

6.2.2 Inoculation, fermentation, and sampling

Frozen Chardonnay juice from Blewitt Springs, South Australia (2013 vintage) was thawed and filtered, first through a 0.4 µm filter and then through a 0.22 µm filter. An aliquot of 100 µL of the sterile-filtered juice was plated onto YEPD agar and incubated at 28-30 C to confirm sterility.

Prior to inoculation, one colony of each yeast strain was aseptically transferred to separate flasks containing 50% autoclaved liquid YEPD media and

50% Chardonnay juice, to acclimatize the yeasts to the Chardonnay medium. These flasks were sealed with parafilm and incubated aerobically overnight at 28-30 C. A small homogeneous subsample from each of these flasks was diluted 10 times, 5 % propidium iodide (1 mg/mL) was added to fluorescently stain any dead cells

(Valentine et al., 2019), and the number of living cells were counted using a hemocytometer to determine the cells/mL of each flask. Once the cell concentration of each yeast starter culture was determined, they were inoculated into fermentation 184 flasks containing 100 mL sterile Chardonnay juice in different amounts to achieve the ratios outlined in Table 6.1 (n = 3 per treatment). The amount of each starter culture added was also adjusted so that the total density of living yeast cells in each fermentation flask was 2.5 × 106 cells/mL at the start of the experiment.

The fermentation flasks consisted of 250 mL GL45 glass Schott bottles (DWK

Life Sciences, Wertheim, Germany) sealed with custom 3D-printed airlocks (filled with 3 mL sterile water) containing a central sampling port closed with a silicone septum (https://www.carbon3d.com/case-studies/tthandadelaide/, accessed 2019-

05-27). The fermentation flasks also contained a small, sterilized stir bar which were used to maintain an even mixture of the fermentations. Laboratory-scale fermentations were conducted in a customized Freedom EVO® automated fermentation unit (Tecan Group, Ltd., Männedorf, Switzerland) with temperature- controlled panels, auto-sampling capabilities, and individual magnetic stir plates for each flask (described further in Peter et al., 2018). Fermentations at the two temperatures were conducted simultaneously, and within each temperature treatment, the order of the fermentation flasks was randomized. An uninoculated fermentation flask containing only sterile-filtered Chardonnay juice was used as a negative control at each temperature to monitor for any microbial contamination. No contamination was observed over the course of this experiment. The fermentations were conducted for 10 days before being terminated.

185

6.2.3 Sugar concentration, pH, and cell count determination

Samples for residual sugar (g/L glucose and fructose) and yeast ratio/cell count determination were taken on days 1, 3, 5, 7, and 10. For each sample, 100 µL fermenting juice was taken from each flask using a sterilized sampling needle.

Residual sugar concentration (g/L glucose + fructose) was determined enzymatically. Fermentation samples were diluted and adjusted to a final volume of

200 µL for analysis in 96-well microplates. Residual sugars were quantified enzymatically via spectrophotometry on a microplate reader with microplate stacker attachment (Tecan Group, Ltd., Männedorf, Switzerland), using hexokinase + glyceraldehyde-6-phosphate dehydrogenase and phosphoglucose isomerase enzymes (Megazyme Inc., Chicaco, IL, USA). Plate preparation and enzymatic/spectrophotometric analyses were performed robotically. The pH of the

Chardonnay juice pre-fermentation was determined using a CyberScan 1100 pH meter (Eutech Instruments Pte. Ltd., Singapore).

Samples taken for yeast ratio and cell count determination were serially diluted and sheared through a thin 0.65 × 90 mm spinal needle (Terumo®, Tokyo,

Japan) with a syringe to break up any clumps of cells and prevent multiple cells from growing into a single colony and being mis-reported. The samples were then spread plated onto Wallerstein Laboratory Nutrient (WLN) agar (Oxoid Ltd., Scoresby, VIC,

Australia) before being incubated for 48 h at 28-30 C. WLN media allows for the visual distinction between S. cerevisiae and S. uvarum colonies: S. cerevisiae appears as medium-sized, cream-coloured colonies, while S. uvarum appears as

186 small, forest green colonies. In this way, both total colony forming units (CFU) /mL as well as the ratio of QA23:SU01 could be determined for each sampling day.

6.2.4 Sulfur dioxide determination and sulfite resistance assay

Because the Chardonnay juice used in this experiment was obtained in 2013 and frozen for three years before use, we considered it likely that SO2 was added to the juice before freezing. Four juice aliquots of 50 mL were taken prior to the start of the experiment for SO2 measurements. Total and free SO2 were determined using the aeration, oxidation, distillation, and titration procedure described previously in

Chapter 3 (section 3.2.2). Molecular SO2 was calculated from the free SO2 concentration and the pH of the juice using the Henderson-Hasselbach equation

(Divol et al., 2012; Zoecklein et al., 1995). Because SO2 was found in the juice, the relative sensitivity of each yeast to SO2 was determined using a sulfite resistance assay (Park et al., 1999b). Autoclaved YEPD media was buffered to pH 3.5 with tartaric acid, and precisely 18 mL media was poured onto 25 mL sterile plastic petri dishes and left to harden. Appropriate amounts of 0.5 M sodium sulfite (Na2SO3) were then spread onto the plates in order to create the following concentrations: 1,

1.0, 1.5, 2.0, and 2.5 mM Na2SO3. These plates were left to set overnight at room temperature. The following day, dilutions of each yeast inoculum containing approximately 100 living cells/100 µL were spread plated onto plates at each

Na2SO3 concentration (n = 3) and incubated for 48 h at 28-30 C, and then visually assessed for growth inhibition.

187

6.2.5 Secondary metabolite analysis

Relative quantification of 26 secondary metabolites produced by yeasts during fermentation (Table 6.2) was performed using a combination of headspace- solid phase microextraction (HS-SPME) and gas chromatography-mass spectrometry (GC-MS), as described previously in Chapter 4. Samples were taken at the end of the 10 day experiment.

Table 6.2. List of yeast-derived secondary metabolites measured, as well as their characteristic aromas (Acree and Arn, 2004a; Sun et al., 2018b) and odour detection thresholds (µg/L) in 10% (v/v) ethanol (Ferreira et al., 2000; Francis and Newton, 2005; Guth, 1997; Peinado et al., 2006, 2004; Salo, 1970; Siebert et al., 2005; compiled by Haggerty et al., 2016).

Category Metabolite Abbreviation Aroma Threshold (µg/L) Ethyl acetate EA fruity, solventab 7,500 Ethyl propanoate EP fruity, solventab 1,800 Ethyl 2-methyl propanoate E2MP fruitya 15 Ethyl butanoate EB apple, strawberrya 20 Ethyl 2-methyl butanoate E2MB apple, strawberrya 1-18 Ethyl esters Ethyl 3-methyl butanoate E3MB apple, strawberrya 3 Ethyl hexanoate EH fruity, fermented peara 5-14 Ethyl octanoate EO fruity, fattyb 2 Ethyl decanoate ED grapeb 200 Ethyl dodecanoate1 EDD soapy, esteryb 25,6195 2-methyl butyl acetate2 2MBA fruityb 160 3-methyl butyl acetate2 3MBA bananab 30 Acetates Hexyl acetate HA fruity, herbyb 670 2-phenylethyl acetate 2PEA rose, honeyb 250 2-methyl butanoic acid3 2MBAcid butter, cheeseb 1,500 3-methyl butanoic acid3 3MBAcid sweat, rancidb 33.4 Acids Hexanoic acid1 HAcid sour, vinegar-likea 420 Octanoic acid OA sweat, cheeseab 500 Decanoic acid DA soura, rancid/fattyb 1,000-8,100 2-methyl propanol 2MP roasted nutsa, solventb 40,000 2-methyl butanol4 2MB onionb 65,000 3-methyl butanol4 3MB roasted nutsa, whiskyb 30,000 Alcohols Hexanol HOH green, floralab 8,000 2-phenyl ethanol 2PE honey, spice, floralb 10,000-14,000 Benzyl alcohol Benz sweet, floralb 900,000 Methionol MOH cooked potatoa 1,000 1EDD and HA had overlapping retention times 22MBA and 3MBA had overlapping retention times 32MBAcid and 3MBAcid had overlapping retention times 42MB and 3MB had overlapping retention times 5Odour threshold for EDD was determined in 43% (v/v) ethanol (Sun et al. 2018)

188

6.2.6 Statistical analysis

A standard curve for each compound of interest measured via HS-SPME GC-

MS was created in Excel 2016, and a linear line of best fit was applied to each volatile compound of interest. R2 values above 0.99 were confirmed for all compounds, except ethyl acetate, which had an R2 value of 0.93 (data not shown).

This indicated that no non-target compounds were being measured along with the compound of interest.

Chi square goodness-of-fit tests were performed on the relative abundance

(%) of S. cerevisiae and S. uvarum CFU/mL data from samples plated on Day 1 of the experiment, in order to confirm that the initial inoculation ratios were as expected

(per treatment). Sugar consumption, yeast counts (CFU/mL), and yeast relative abundance throughout fermentation were visualized ± the standard error of the mean (SEM) in Excel 2016 and converted to high-resolution images using IrfanView software (version 4.51).

The composition of secondary metabolites was analyzed using Primer v.6 software with PERMANOVA+ add-on (Plymouth, MA, USA) (Clarke and Gorley,

2006). A principal coordinate analysis (PCoA) was generated in order to visualize differences in secondary metabolite composition among treatments. Variable vectors were plotted using Pearson’s correlation.

6.3 Results and discussion

6.3.1 Sulfur dioxide and fermentation progression

The Chardonnay juice used in this study was found to contain 32.6 ± 0.40 mg/L total SO2 and 5.62 ± 0.02 mg/L free SO2. This juice was crushed and pressed

189 in 2013 and sulfited before being frozen, so the juice was expected to contain some residual SO2. The pH of the juice was 3.20, and the molecular portion of SO2 was therefore calculated to be approximately 0.22 mg/L. Molecular SO2 is the most active antimicrobial portion of SO2 (Divol et al., 2012), so it is important to calculate when conducting microbial studies. Effective microbial inhibition by molecular SO2 has been estimated to be 0.2-0.8 mg/L, depending on the publication and the susceptibility of the grape juice to spoilage (Howe et al., 2018; King et al., 1981;

Margalit and Crum, 2004). Saccharomyces yeasts are known to be more resistant to

SO2 than vineyard-derived non-Saccharomyces yeasts, with commercial S. cerevisiae yeasts being the most resistant, so it is unlikely that ~0.2 mg/L SO2 would be effective in inhibiting the growth of either yeast used in this study. Indeed, a sulfite resistance assay confirmed that both yeasts were able to grow on media containing up to 2.5 mM Na2SO3 (~160 mg/L total SO2), and their growth, determined by colony size after 48 h incubation, slowed at similar rates with increasing Na2SO3 concentration (data not shown).

Fermentation progression, as measured by total sugar concentration (g/L glucose + fructose) on each day, proceeded at similar rates within each fermentation temperature (Figure 6.1). The fermentations conducted at 24 C were mostly complete at the end of the 10 day experiment, with the 1:0 and 10:1 treatments

(dominated by S. cerevisiae) containing an average of < 3 g/L residual sugar. The other three treatments (1:1, 1:10, and 0:1) contained an average of 11-22 g/L residual sugar, but from the sugar consumption trajectory it is likely that these fermentations would have also completed AF within the next 2-3 days. Conversely,

190 none of the fermentations conducted at 15 C were complete within the 10 day timeframe of the experiment. Regardless of inoculation ratio, all fermentations were sluggish, and contained an average of 75-92 g/L residual sugar on Day 10 (Figure

6.1A). It is possible that the fermentations conducted at 15 C would have become stuck before completing alcoholic fermentation, regardless of which yeast was conducting the fermentation. This was an unexpected result, because S. uvarum is considered a cryotolerant yeast, and is usually found in low-temperature fermentations (Almeida et al., 2014; Demuyter et al., 2004; Naumov et al., 2000).

Additionally, QA23 is described as being ‘capable of fermenting juice at low temperatures (15 C) to dryness’ (https://www.lallemandbrewing.com). However, in most commercial cellars, barrel fermentations begin at cellar-temperature (12-14 C) but then increase to 18-20 C over the course of AF (see Chapters 3 and 5), due to the metabolic activity of the yeasts raising the temperature in the barrel naturally.

Artificially holding the temperature at 15 C, as was the case in this controlled experiment, may not accurately reflect the behaviour of these yeasts in an applied setting. Gamero et al. (2013) also conducted fermentations with a variety of

Saccharomyces strains at both 12 C and 28 C, and noticed that all of the strains, even the cryotolerant S. uvarum strains, took longer to complete fermentation at 12

C. On average, the 28 C fermentations took less than seven days to complete, while the 12 C fermentations took more than 18 days. It is possible, therefore, that the fermentations conducted at 15 C in this current study would all have completed if given enough time.

191

Figure 6.1. Daily residual sugar (g/L) levels (± SEM) of controlled Chardonnay fermentations to which different inoculation ratios (S. cerevisiae : S. uvarum). Fermentations were conducted at (A) 24 C and (B) 15 C.

6.3.2 Yeast competition during fermentation

In all inoculation treatments and at both fermentation temperatures, total yeast count (CFU/mL) grew to a peak density by Day 5-7, and then began to decline

(Figure 6.2). This growth trajectory is typical of closed systems, although it should be

192 noted that for the fermentations conducted at 15 C, this decrease in cell density began before sugar consumption had reached 50% (Figure 6.1B). The 0:1 fermentations, conducted only by the S. uvarum strain SU01, contained the highest

CFU/mL throughout fermentation, and the 1:1 treatments, which were inoculated with an equal ratio of QA23 and SU01, contained the lowest CFU/mL. It is likely that the equal inoculation ratio in the 1:1 treatment created an environment that promoted competition between the two yeasts, which could have resulted in the yeasts diverting energy to nutrient sequestration and yeast antagonism mechanisms rather than cell growth. There was a trend towards increased cell density at 15 C in the treatments dominated by SU01 (1:10 and 0:1; Figures 6.2D and 6.2E). This result is supported by previous studies that also showed an increase in S. uvarum growth at lower temperatures (Dellaglio et al., 2003; Sipiczki et al., 2001).

Interestingly, this increase in growth did not correspond to an increase in fermentation performance, as all treatments still exhibited sluggish fermentations at

15 C.

193

Figure 6.2. Percent relative abundance (± SEM) of S. cerevisiae and S. uvarum, as well total yeast count (CFU/mL ± SEM) during controlled Chardonnay fermentations conducted at two temperatures, to which the following inoculation ratios (S. cerevisiae : S. uvarum) were applied: (A) 1:0, (B) 10:1, (C) 1:1, (D) 1:10, (E) 0:1.

194

Chi square goodness-of-fit tests confirmed that the initial QA23:SU01 inoculation ratios did not differ from the expected ratio within each treatment (χ² =

0.38 - 8.02, p = 0.064 - 0.43). The 1:0 and 0:1 treatments were conducted completely by QA23 and SU01, respectively, as they were the only strains inoculated into each respective treatment at either temperature. The 10:1 treatment was quickly dominated by QA23 at both temperatures, and QA23 maintained its dominance throughout fermentation (Figure 6.2B). When inoculated at an equal ratio

(1:1), QA23 was still able to out-compete SU01 at both temperatures. The 1:1 (24

C) treatment did contain more SU01 on Day 1 than QA23, but even still, by Day 3

QA23 dominated with > 90% relative abundance (Figure 6.2C), and by Day 10 SU01 represented < 1% of the relative yeast abundance. SU01 was more competitive at

15 C than at 24 C, and in the 1:1 (15 C) treatment SU01, although still sub- dominant, was able to maintain approximately 25% of the relative abundance through to the end of the fermentations, likely contributing significantly to the volatile compound profiles of these wines. Like the 10:1 fermentations, the 1:10 fermentations were dominated by the yeast that was initially inoculated at a higher ratio - in this case, SU01 (Figure 6.2D). However, in the 1:10 (24 C) treatment,

QA23 was able to maintain its initial inoculation proportion of approximately 10% throughout fermentation, and even increased its relative abundance slightly by Day

10. As expected, SU01 was more competitive at the lower fermentation temperature, and QA23 was only able to maintain a 1-2% relative abundance on average throughout fermentation in the 1:1 (15 C) treatment. Unlike the 10:1 treatments, where SU01 was completely eliminated by Day 10 at both fermentation

195 temperatures, QA23 was never completely eliminated in the 1:10 treatments, highlighting its competitive abilities even when inoculated at a disadvantageous ratio.

A recent study conducted by Alonso-del-Real et al. (2017) also noted that a commercial S. uvarum wine strain (Velluto BMV58®) was more competitive at lower temperatures in co-fermentations with a commercial S. cerevisiae strain (1:1 inoculation ratio). They also found, however, that BMV58 was able to maintain a 1:1 ratio at 20 C, and was able to dominate (> 90% relative abundance) at 12 C, which

SU01, used in this current study, was unable to do. This could be a result of differences in competitive abilities between BMV58 and SU01, but could also be a result of the different S. cerevisiae strains used. Alonso-del-Real et al. (2017) used

Lalvin T73TM as the commercial S. cerevisiae strain, which has a much higher optimal fermentation temperature range (18-35 C) than QA23 (14-28 C), and is recommended for red wine production. Therefore, it is expected that T73 would have reduced competitive abilities when inoculated into fermentations that were significantly below its optimal temperature range.

S. cerevisiae has many competitive traits that allow it to dominate wine fermentations, chief among them being its ability to conduct AF vigorously and efficiently, and its ability to withstand high concentrations of ethanol. In traditional wine fermentations, the vineyard-derived non-Saccharomyces yeasts that begin the process of fermentation will quickly be out-competed by S. cerevisiae, as many are unable to withstand high ethanol concentrations (Domizio et al., 2007; Fleet, 2003).

However, in this study, both QA23 and SU01 conducted alcoholic fermentation at

196 similar rates at both fermentation temperatures, so it is unlikely that fermentation rate or ethanol tolerance contributed to the difference in competitive abilities observed between these two strains. S. cerevisiae also dominates fermentations by inhibiting the growth of other species through the production of killer toxins (Schmitt and Breinig, 2006) and through its rapid growth rate that can result in a depletion of available nutrients. QA23 does produce killer toxins, but it is unclear which class(es) of toxins are produced by this strain, or whether S. uvarum is sensitive to killer toxins. To our knowledge, there is currently only one report of S. uvarum containing potential killer plasmids (Ivannikova et al., 2007), although this is likely because little research has been conducted to investigate this topic, not because this yeast does not have a killer phenotype. More research is needed in order to determine the specific competitive mechanisms at play between these two strains.

6.3.3 Yeast-derived secondary metabolite composition

Both inoculation ratio and fermentation temperature seemingly affected the composition of the 26 yeast-derived volatile secondary metabolites measured in these wines. Due to the design of this experiment, samples for secondary metabolite analysis were taken when the fermentations were terminated after 10 days. This means that the samples taken from the 24 C treatment were taken near the end of fermentation, while the samples taken from the 15 C treatment were taken mid- fermentation. Because this discrepancy could potentially lead to an overestimation of differences between the two temperature treatments, we have chosen not to conduct a statistical analysis of these results but have chosen instead to highlight trends observed among treatments.

197

The two treatments dominated by SU01 (0:1 and 1:10) had similar volatile profiles, and the composition of volatile secondary metabolites in these treatments did not seem to be affected by fermentation temperature (Figure 6.3A). Conversely, the treatments dominated by QA23 were likely significantly affected by fermentation temperature (Figure 6.3A). At 24 C, the 1:0 and 10:1 treatments were extremely similar, and the 1:1 treatment, which was also dominated by QA23 at 24 C, was also similar to these treatments. At 15 C, however, the 1:1 treatment contained approximately 25% SU01 throughout fermentation, and had a different volatile profile than the 1:0 and 10:1 treatments, which were similar to each other, but still distinct from the treatments that fermented at 24 C (Table 6.4 and Figure 6.3A).

This suggests that in the 1:1 (15 C) treatment, SU01 was present at a high enough concentration to significantly contribute to the production of volatile secondary metabolites. These results also suggest that fermentation temperature may have a greater effect on the fermentative behaviour of QA23 than SU01.

198

Figure 6.3. PCoA visualizing the production of volatile secondary metabolites by yeasts in controlled Chardonnay fermentations to which different inoculation ratios (S. cerevisiae : S. uvarum) and temperature treatments (24 or 15 C) were applied (n = 3 per treatment). Plots depict (A) the individual factors map and (B) the variable factors map showing secondary metabolites as vectors. Points that are closer together contain more similar volatile profiles than points that are further apart. On the variable factors map, the length of the variable vector reflects that variable’s relative contribution in producing this ordination. Variable vectors that point towards a sample are positively correlated with that sample, and those that point away from a sample are negatively correlated. Variable vectors that appear at a ~90 angle are not correlated with each other, while those that appear at < 90 and > 90 angles are positively and negatively correlated with each other, respectively.

199

In general, when fermented at 24 C the QA23-dominated fermentations produced higher amounts of volatile compounds (Table E2). The SU01-dominated fermentations were less variable at the two fermentation temperatures, and generally produced similar amounts of each compound at both 15 C and 24 C.

However, the fermentations dominated by SU01 did produced higher amounts of a number of compounds as compared to QA23-dominated fermentations, regardless of fermentation temperature, including ethyl 2-methylpropanoate, 2-methyl propanol,

2-phenylethanol, 2-phenylethyl acetate, hexyl acetate/ethyl dodecanoate, ethyl 2- methylbutanoate, ethyl 3-methylbutanoate, and 2/3-methylbutanoic acid (Appendix

E). Previous studies have also highlighted differences in volatile compound production between species and strains of yeasts (Bagheri et al., 2018; Ciani et al.,

2006; Jolly et al., 2014; Romano et al., 2003), and specifically with regards to S. cerevisiae and S. uvarum (Gamero et al., 2013; López-Malo et al., 2013; Stribny et al., 2015). A study conducted by López-Malo et al. (2013) found differences in metabolite production by QA23 and a non-commercial strain of S. uvarum when fermented at 12 C, and also found that QA23 produced different amounts of metabolites when fermented at 12 C versus 28 C, supporting our own results.

However, Gamero et al. (2013) conducted fermentations with three S. uvarum strains at two temperatures (12 and 28 C), and contrary to our results, they did note differences in volatile compound production between the two temperatures for each strain. Many of these changes were consistent for the three S. uvarum strains, but some compounds, including ethyl caprylate and 2-phenyl ethyl acetate, varied by strain.

200

The treatments dominated by SU01 (0:1 and 1:10) were correlated with the production of 10 volatile compounds (Figure 6.3B), which produce both desirable and undesirable aromas (Table 6.2) Among these, ethyl 2-methyl propanoate

(fruity), ethyl 3-methyl butanoate (apple/strawberry), ethyl 2-methyl butanoate

(apple/strawberry), 2-methyl propanol (roasted nuts/solvent), and 2-phenyl ethanol

(honey/spice/floral) have been found to be produced above their detection thresholds by S. uvarum (Tosi et al., 2009; Varela et al., 2017). At 24 C, the 1:0 and

10:1 treatments (dominated by QA23) were most closely associated with the production of ethyl octanoate (fruity, fatty) and ethyl decanoate (grape), which have been shown to be produced above their sensory thresholds by S. cerevisiae

(Comitini et al., 2011), as well as benzyl alcohol (sweet, floral). All of these compounds have desirable sensory characteristics at lower concentrations. The 1:1 treatment was also dominated by QA23 at 24 C, but was correlated with the production of decanoic acid (sour, rancid/fatty), ethyl hexanoate (fruity, fermented pear), and ethyl butanoate (apple, strawberry). Decanoic acid is not often found above its sensory threshold (Comitini et al., 2011), but ethyl hexanoate and ethyl butanoate are (Comitini et al., 2011; Varela et al., 2017). At 15 C, however, the 1:0 and 10:1 treatments were correlated with the production of hexanol (green, floral) and hexyl acetate (fruity, herby). Both of these compounds contribute to an overall

‘green’ character in wine, and previous research has also found that wines fermented at lower temperatures contain more ‘green’ characters (Deed et al.,

2017). The 1:1 (15 C) treatment was not positively correlated with the production of any specific compounds, but seemed to be negatively correlated with the production

201 of 2/3-methyl butanol (onion, roasted nuts, whisky), 2/3-methyl butyl acetate (fruity, banana), and ethyl acetate (fruity, solvent) (Figure 6.3).

6.4 Summary

This study investigated the enological potential and competitive abilities of an indigenous strain of S. uvarum. Fermentations conducted at a higher temperature

(24 C) progressed at a quicker rate as compared to fermentations conducted at a lower temperature (15 C), regardless of which yeast strain was conducting the fermentation. S. cerevisiae was found to be more competitive than S. uvarum, and dominated fermentations inoculated at a 10:1 and 1:1 (S. cerevisiae : S. uvarum) ratio. However, S. uvarum was more competitive at 15 C than at 24 C, and was able to maintain a 25% relative abundance in the 1:1 ratio fermentations conducted at the lower temperature. When inoculated at a 1:10 ratio, S. uvarum was able to dominate, although S. cerevisiae was not completely eliminated at either temperature. The volatile profiles of these wines corresponded to the composition of yeast strains in the fermentations: the fermentations dominated by S. uvarum had distinct volatile profiles from those dominated by S. cerevisiae. Fermentation temperature only affected the production of volatile compounds in the S. cerevisiae- dominated fermentations. However, because the samples for secondary-metabolite production were taken at different stages of fermentation (mid-fermentation for the

15 C treatments and late-fermentation for the 24 C treatments), it is possible that some differences between fermentation temperature are being over-estimated in this experiment. Follow-up experiments that compare samples from the same stage of fermentation are recommended. The fermentations conducted by S. uvarum were

202 correlated with the production of the highest number of secondary metabolites, potentially producing a more complex wine. Because fermentation temperature did not seem to affect the production of secondary metabolites of this S. uvarum strain, it has the potential to produce a consistent product even under different fermentation conditions. This study highlights the potential for an indigenous S. uvarum strain from the Okanagan Valley to produce unique and quality wines.

203

Chapter 7: Conclusion

7.1 Summary and general discussion

The overall goal of this thesis was to investigate the topics of SO2 additions and alternative fermentation techniques in wine, and their effects on microbial communities and wine sensory profiles. We accomplished this via five main research objectives.

In Chapter 2, we characterized the S. cerevisiae populations in uninoculated fermentations conducted at two commercial wineries. These fermentations had received different additions of sulfur dioxide at crush (0, 20, or 40 mg/L SO2), which resulted in significantly different S. cerevisiae populations conducting fermentation.

We found that some strains increased or decreased their abundance in the fermentations in a dose-dependent manner, depending on the concentration of SO2 that was added to the must at crush. Furthermore, even though all the fermentations analyzed were uninoculated, we found that commercial ADY strains made up the majority of S. cerevisiae strains identified at both wineries, usually corresponding with commercial strains used previously or concurrently at their respective wineries.

This study demonstrated that SO2 addition at crush can indirectly influence which S. cerevisiae strains end up dominating the fermentation, even though S. cerevisiae is not usually the intended target of SO2 additions by winemakers. These findings are of particular interest to winemakers who may be looking to manage the composition of yeast strains in their fermentations by modifying the concentration of SO2 they add to their fermentations.

204

In Chapter 3, we used Illumina sequencing in combination with PMA addition to capture the living fungal and bacterial communities present during uninoculated fermentations of Pinot gris wines to which different concentrations of SO2 (0, 20, or

40 mg/L) were added at crush. Although the bacterial community was not significantly influenced by SO2 addition, and all treatments were dominated by S. cerevisiae, we found a large Hanseniaspora yeast population that was able to persist until the end of alcoholic fermentation. Its relative abundance was significantly higher in the fermentations to which low or no SO2 was added, likely because Hanseniaspora spp. tend to be sensitive to SO2. These wines also exhibited differences in sensory attributes among SO2 treatments, with two key wine sensory attributes (citrus aroma and pome fruit flavour) being higher in intensity in the wines to which low or no SO2 was added at crush. These differences in sensory attributes were likely the result of a combination of factors, including the S. cerevisiae strain population, the fungal community, and the chemical composition of the wines. This research provided an in-depth look into the fungal and bacterial communities during alcoholic fermentation, and gives a better understanding of the microbial community response to SO2 additions at crush. This was the first time that high-throughput amplicon sequencing, in combination with PMA addition, had been used to identify both the fungal and bacterial communities of wine fermentations in a commercial setting.

In Chapter 4, we investigated the responses of two commercial S. cerevisiae strains to SO2 addition under laboratory conditions. We found that Strain 2 was more resistant to SO2 than Strain 1, possibly due in part to the higher constitutive

205 expression of acetaldehyde by Strain 2. However, both strains completed AF successfully, even in the presence of SO2. Overall, the production of volatile compounds by both yeasts was higher when SO2 was added, and the wines produced without SO2 were generally negatively correlated with the production of undesirable volatile compounds. In terms of the production of volatile compounds,

Strain 1 was less affected by SO2 addition than Strain 2, which showed significantly different volatile profiles with response to SO2, and more variation within each replicate. These results may help winemakers make more informed decisions when selecting commercial yeast strains for inoculation into fermentations containing different concentrations of SO2, and when specific wine characteristics are desired.

In Chapter 5, we conducted a study at a large winery that investigated how decisions made by winemakers at crush can affect the microbial communities and sensory attributes of commercially-produced wines. We tested the effects of inoculation techniques (uninoculated or pied de cuve inoculation) as well as sulfite addition (0 or 40 mg/L SO2) on the fungal and bacterial communities,

Saccharomyces strains, and sensory attributes of Chardonnay wines. We used

Illumina sequencing in combination with PMA addition to accurately capture the entire living fungal and bacterial communities, and used microsatellite strain-typing to identify strains of S. cerevisiae and S. uvarum present in the fermentations. When no SO2 was added at crush, the wines underwent AF and MLF simultaneously.

Tatumella bacteria were also present in significant numbers only in the barrels to which no SO2 was added, and we hypothesize that these bacteria were responsible for the degradation of malic acid observed in these treatments. This is the first report

206 of Tatumella in North American wineries, and the first report of Tatumella saanichensis in wine. Additionally, all fermentations were dominated by a highly genetically diverse indigenous population of S. uvarum, and 150 unique S. uvarum strains were identified, with differences in strain composition as a result of SO2 addition. Although S. uvarum has been reported dominating low-temperature wine fermentations in Europe, this is the first report of such an occurrence in North

America, and the diversity of S. uvarum strains observed in this study represents the most diverse population of S. uvarum reported anywhere. The wines that did not receive SO2 at crush were rated more fruity and less acidic than the wines that received SO2, which expressed less-desirable traits such as earth and sulfur aroma/flavour. The pied de cuve inoculum had a limited effect on the microbial community and only persisted when no SO2 was added; these wines did not fully complete AF, and were rated to be higher in body and sweetness, and lower in acidity, than the other wines. This was also the first study conducted that investigated a vineyard-derived pied de cuve inoculation technique, and provides an important stepping stone for other researchers looking to investigate this topic.

In Chapter 6, we selected the most abundant indigenous S. uvarum strain from the study conducted in Chapter 5 to test its competitive abilities in co- fermentations with a commercial S. cerevisiae strain under controlled conditions.

Laboratory-scale fermentations of Chardonnay were conducted with S. uvarum and

S. cerevisiae co-inoculated in different initial ratios, as well as by themselves.

Additionally, because S. uvarum is a cryotolerant strain, we conducted these fermentations at two temperatures (15 C and 24 C). We found that S. uvarum was

207 only able to out-compete S. cerevisiae when inoculated at a higher initial ratio. S. uvarum was more competitive at a lower fermentation temperature, but performed alcoholic fermentation better at a higher temperature. The production of volatile secondary-metabolites by S. uvarum was not affected by fermentation temperature, but for S. cerevisiae, fermentation temperature significantly altered the volatile profiles produced. Additionally, the volatile profiles of the wines corresponded to the composition of yeast strains in the fermentations: the fermentations dominated by S. uvarum had distinct volatile profiles from those dominated by S. cerevisiae. This is the first study conducted that investigated the competitive abilities of an indigenous strain of S. uvarum in competition with a commercial strain of S. cerevisiae, and provides evidence that indigenous S. uvarum strains could be used in co- inoculations with S. cerevisiae in order to produce unique wines with more regional character.

7.2 Limitations

As with any research project, the studies conducted for this thesis came with limitations. When conducting research in an applied field, the main considerations when designing an experiment are often its relevance to the industry, and the applicability of the results. Therefore, all of the studies I conducted for this thesis came with trade-offs between the ability to control all variables, as expected academically, and the ability to conduct research that winemakers and wine industry professionals could extrapolate from for their own purposes. In order to balance these trade-offs, I conducted three research projects (Chapters 2, 3, and 5) at industry-scale and in industrial environments, and two laboratory-scale studies

208

(Chapters 4 and 6), which allowed me to investigate mechanistic effects, and the contributions of individual yeasts to fermentations.

In Chapter 2, we were unable to achieve a balanced design between the two wineries - three SO2 addition concentrations were evaluated at Winery 1 (0, 20, and

40 mg/L SO2), while only two were evaluated at Winery 2 (20 and 40 mg/L SO2).

This was a compromise that was necessary in order to collaborate successfully with each winery - the winemaker at Winery 2 was concerned about the possibility of wine spoilage if no SO2 was added, and as the head winemaker of a small winery, he was unwilling to risk 1,500 L of quality grape must for the sake of this experiment.

In Chapter 3, we conducted all fermentations in brand-new French oak barrels, all from the same cooperage. We chose new barrels because previous research has shown that used oak barrels can harbour yeast and bacteria between vintages, which may subsequently contaminate future fermentations (González-

Arenzana et al., 2013). Because we were interested in how the microbial communities of wine change in response to SO2 addition, it was important to maintain the microbiological integrity of this experiment. However, new oak barrels impart a strong oaky, toasty, and smoky quality on the wines produced within them, which likely obscured any treatment-specific sensory differences in these wines. In this study, only two sensory attributes were considered to be significantly different among treatments, although many sensory attributes approached significance. In all of the wines evaluated in this study, the unprocessed wood and smoky/toasty characteristics were the sensory attributes with some of the highest intensities.

Because the oak barrels themselves were not a part of this experiment, we

209 suggested that future studies be conducted in other vessels such as stainless steel barrels, which can be sterilized and which do not impart a specific quality on the wines they produce. We followed this advice for the study conducted in Chapter 5, where we used new stainless steel barrels for all treatments and found significant differences in 14 of the 18 sensory attributes evaluated.

For Chapter 4, we originally intended to compare the chemical responses of the yeast strains with the relative expression of genes related to these responses.

To this end, RNA was extracted from fermentation samples two hours after the yeasts were inoculated into juice containing 0 or 50 mg/L SO2. These RNA samples were converted to cDNA, and shipped from Australia to Canada. We designed primers targeting 20 different genes previously found to be related to SO2-resistance

(Kim et al., 2012; Nadai et al., 2016; Nardi et al., 2010), as well as four reference genes. Unfortunately, only a small amount of RNA was originally extracted from these fermentations, and so little of the target cDNA was present in the samples shipped to Canada that the investigation into the gene expression profiles of these yeasts became infeasible.

In Chapter 5, we used the Illumina MiSeq platform to perform paired-end sequencing of the fungal and bacterial communities present in these wines.

Unfortunately, for the bacterial community, the reverse sequences were of low quality, and ended up being discarded; we were therefore left with much shorter, forward-read-only sequences to analyze. Because of these shorter sequences, and the limited nature of curated databases such as Greengenes, the most important bacteria in this study could only be identified to the family level using the

210 recommended protocol. We were, however, able to identify these sequences to the species-level using NCBI-Blast. The field of ‘mycobiome’ research is newly- emerging, and as such new methods of analyzing the fungal community are constantly being developed. Between the time when the data from Chapter 5 was analyzed to the time that this thesis was written, a new and more accurate de- noising technique was developed that takes into account the variability of the ITS region among fungal taxa (https://benjjneb.github.io/dada2/ITS_workflow.html). It is possible that the identified composition of the fungal community in these treatments would be different if such methods were used to analyze these data.

For Chapter 6, we were originally interested in investigating more co- inoculation ratios of S. cerevisiae and S. uvarum than were included in this chapter, but we were limited by the amount of juice available. To get around this limitation, we first attempted to conduct these fermentations using chemically-defined grape juice media (CDGJM), which has been used for many laboratory-based studies

(Haggerty et al., 2016; Peter et al., 2018; Valentine et al., 2019). However, we noticed that the indigenous S. uvarum strain used in this study was unable to properly grow and ferment in this CDGJM, giving it an unfair disadvantage when co- inoculated with S. cerevisiae, which grew well in this synthetic medium. It is likely that CDGJM was designed specifically to study S. cerevisiae in controlled fermentations, and that the nutrient requirements of S. uvarum are slightly different.

7.3 Future directions

The introduction of high-throughput amplicon sequencing and other molecular technologies has greatly expanded the potential scope of wine microbiology

211 research. Traditional assumptions of the wine research field, including the assumption that poorly fermentative yeasts such as Hanseniaspora spp. could not survive in wine through to the end of alcoholic fermentation, have been challenged in this thesis and will continue to be challenged by future research.

The effects of SO2 addition on wine yeasts has been studied for decades, but there are always more avenues to pursue in this field. While specific mechanisms of

SO2 resistance have been investigated, a more holistic approach could help us understand how these different mechanisms may work together to provide SO2 resistance. Furthermore, most SO2 resistance studies have focused on either spoilage yeasts such as B. bruxellensis or on S. cerevisiae, and there is still a lack of knowledge into the mechanisms of SO2 resistance used by non-Saccharomyces yeasts. Future research should expand the focus of SO2 resistance studies to include these enologically important yeasts.

Currently, there is a significant lack of research surrounding the topic of pied de cuve inoculation, and because many winemakers are turning to alternative fermentation techniques to produce unique wines with local character, knowledge surrounding the potential implications of these techniques is important. To our knowledge, this thesis provides the first investigation into the effects of vineyard- initiated pied de cuve inoculation. Future research should include a comparison of different pied de cuve methods, from vineyard-initiated inoculation, to inoculums sourced from currently-fermenting vessels, to those sourced from lees post- fermentation. For vineyard-initiated pied de cuve experiments, it is imperative to winemakers to know what the optimal length of time the pied de cuve starter should

212 be left in the vineyard for before being used for inoculation. It would also be interesting to know whether using over-ripe grapes for the pied de cuve may give rise to inoculums that favour more strongly-fermentative yeasts.

Because we identified such a highly diverse population of Saccharomyces uvarum fermenting Chardonnay at a local winery, it is likely that this population has its natural origins somewhere in the Okanagan Valley. A thorough investigation into the potential origins of this species is necessary, from uncultivated natural environments to agricultural land (including vineyards), to winery equipment and fermentations from other wineries around the valley. This future work should include consultation with the Okanagan Syilx First Nation, as there is some evidence that the First Peoples of this valley performed simple fermentation, which is a potential source of these yeasts. The S. uvarum strain we used for our controlled fermentations in Chapter 6 showed great potential for making consistent wines at various fermentation temperatures, and could be used in the future for single or co- inoculated fermentations. It is likely that other S. uvarum strains show similar enological potential, and a detailed exploration of this species for use in winemaking could be of value to the winemakers of the Okanagan Valley. Because the S. uvarum strain used in Chapter 6 was unable to successfully grow and ferment in

CDGJM, future work should include the development of either species-specific

CDGJM or a CDGJM that encompasses the nutrient requirements of all wine microorganisms.

213

7.4 Concluding remarks

The results of this thesis provide important insights into the effects of both

SO2 additions and alternative fermentation techniques on the microbial communities and sensory profiles of wines. Using a combination of culture-dependent and culture-independent technologies, we were able to accurately identify the entire living fungal and bacterial communities present in commercial wine fermentations, as well as identify S. cerevisiae and S. uvarum populations to the strain level. We showed that the concentration of SO2 addition at crush can alter the fungal, bacterial, and Saccharomyces composition of these fermentations, leading to differences in wine sensory profiles. For the first time in North America, we reported a genetically diverse population of S. uvarum dominating and completing alcoholic fermentations at a large commercial winery. We also identified a group of Tatumella bacteria that survived only when no SO2 was added to the fermentations and likely conducted malolactic fermentation at the same time that alcoholic fermentation was occurring. In laboratory-based studies, we highlighted the differences in response to

SO2 of two commonly-used commercial S. cerevisiae strains, and evaluated the enological potential and competitive abilities of an indigenous S. uvarum strain fermented at different temperatures. We found that the indigenous S. uvarum strain produced consistent volatile profiles regardless of fermentation temperature, and was more competitive at a lower fermentation temperature. This research can provide winemakers with the tools they need to confidently make decisions surrounding SO2 addition and inoculation techniques.

214

Literature cited

Aa, E., Townsend, J.P., Adams, R.I., Nielsen, K.M., Taylor, J.W., 2006. Population structure and gene evolution in Saccharomyces cerevisiae. FEMS Yeast Res. Rev. 6, 702–715. doi:10.1111/j.1567-1364.2006.00059.x Abrahamse, C.E., Bartowsky, E.J., 2012. Timing of malolactic fermentation inoculation in Shiraz grape must and wine: influence on chemical composition. World J. Microbiol. Biotechnol. 28, 255–265. doi:10.1007/s11274-011-0814-3 Acree, T., Arn, H., 2004. Flavornet [WWW Document]. Flavornet Hum. odor Sp. URL http://flavornet.org/ (accessed 3.29.19). Agnolucci, M., Cristani, C., Maggini, S., Rea, F., Cossu, A., Tirelli, A., Nuti, M., 2013. Impact of sulphur dioxide on the viability, culturability, and volatile phenol production of Dekkera bruxellensis in wine. Ann. Microbiol. 64, 653–659. doi:10.1007/s13213-013-0698-6 Agnolucci, M., Rea, F., Sbrana, C., Cristani, C., Fracassetti, D., Tirelli, A., Nuti, M., 2010. Sulphur dioxide affects culturability and volatile phenol production by Brettanomyces/Dekkera bruxellensis. Int. J. Food Microbiol. 143, 76–80. doi:10.1016/j.ijfoodmicro.2010.07.022 Albergaria, H., Arneborg, N., 2016. Dominance of Saccharomyces cerevisiae in alcoholic fermentation processes: role of physiological fitness and microbial interactions. Appl. Microbiol. Biotechnol. 100, 2035–2046. doi:10.1007/s00253- 015-7255-0 Albertin, W., Zimmer, A., Miot-Sertier, C., Bernard, M., Coulon, J., Moine, V., Colonna-Ceccaldi, B., Bely, M., Marullo, P., Masneuf-Pomarede, I., 2017. Combined effect of the Saccharomyces cerevisiae lag phase and the non- Saccharomyces consortium to enhance wine fruitiness and complexity. Appl. Microbiol. Biotechnol. 101, 7603–7620. doi:10.1007/s00253-017-8492-1 Alexandre, H., Charpentier, C., 1998. Biochemical aspects of stuck and sluggish fermentation in grape must. J. Ind. Microbiol. Biotechnol. 20, 20–27. doi:10.1038/sj.jim.2900442 Alexandre, H., Costello, P.J., Remize, F., Guzzo, J., Guilloux-Benatier, M., 2004. Saccharomyces cerevisiae-Oenococcus oeni interactions in wine: Current knowledge and perspectives. Int. J. Food Microbiol. 93, 141–154. doi:10.1016/j.ijfoodmicro.2003.10.013 Alexandre, H., Guilloux-Benatier, M., 2006. Yeast autolysis in – a review. Aust. J. Grape Wine Res. 12, 119–127. doi:10.1111/j.1755- 0238.2006.tb00051.x Almeida, P., Gonçalves, C., Teixeira, S., Libkind, D., Bontrager, M., Masneuf- Pomarède, I., Albertin, W., Durrens, P., Sherman, D.J., Marullo, P., Hittinger, C.T., Gonçalves, P., Sampaio, J.P., 2014. A Gondwanan imprint on global diversity and domestication of wine and cider yeast Saccharomyces uvarum. Nat. Commun. 5, 4044. doi:10.1038/ncomms5044 215

Alonso-del-Real, J., Lairón-Peris, M., Barrio, E., Querol, A., 2017. Effect of temperature on the prevalence of Saccharomyces non cerevisiae species against a S. cerevisiae wine strain in wine fermentation: competition, physiological fitness, and influence in final wine composition. Front. Microbiol. 8, 1–15. doi:10.3389/fmicb.2017.00150 Anderson, M.J., Walsh, D.C.I., 2013. PERMANOVA , ANOSIM , and the Mantel test in the face of heterogeneous dispersions: What null hypothesis are you testing? Ecol. Monogr. 83, 557–574. doi:10.1890/12-2010.1 Andorrà, I., Esteve-Zarzoso, B., Guillamón, J.M., Mas, A., 2010. Determination of viable wine yeast using DNA binding dyes and quantitative PCR. Int. J. Food Microbiol. 144, 257–262. doi:10.1016/j.ijfoodmicro.2010.10.003 Andorrà, I., Landi, S., Mas, A., Guillamón, J.M., Esteve-Zarzoso, B., 2008. Effect of oenological practices on microbial populations using culture-independent techniques. Food Microbiol. 25, 849–856. doi:10.1016/j.fm.2008.05.005 Andorrà, I., Martín, L., Nart, E., Puxeu, M., Hidalgo, C., Ferrer-Gallego, R., 2017. Effect of grape juice composition and nutrient supplementation on the production of sulfur dioxide and carboxylic compounds by Saccharomyces cerevisiae. Aust. J. Grape Wine Res. 24, 260–266. doi:10.1111/ajgw.12325 Antalick, G., Perello, M.C., De Revel, G., 2013. Co-inoculation with yeast and LAB under winery conditions: Modification of the aromatic profile of merlot wines. South African J. Enol. Vitic. 34, 223–232. Aranda, A., Jiménez-Martí, E., Orozco, H., Matallana, E., Del Olmo, M., 2006. Sulfur and adenine metabolisms are linked, and both modulate sulfite resistance in wine yeast. J. Agric. Food Chem. 54, 5839–5846. doi:10.1021/jf060851b Arnink, K., Henick-Kling, T., 2005. Influence of Saccharomyces cerevisiae and Oenococcus oeni strains on successful malolactic conversion in wine. Am. J. Enol. Vitic. 56, 228–237. Avram, D., Leid, M., Bakalinsky, A.T., 1999. Fzf1p of Saccharomyces cerevisiae is a positive regulator of SSU1 transcription and its first finger region is required for DNA binding. Yeast 15, 473–480. doi:10.1002/(SICI)1097- 0061(199904)15:6<473::AID-YEA388>3.0.CO;2-Y Azzolini, M., Fedrizzi, B., Tosi, E., Finato, F., Vagnoli, P., Scrinzi, C., Zapparoli, G., 2012. Effects of Torulaspora delbrueckii and Saccharomyces cerevisiae mixed cultures on fermentation and aroma of Amarone wine. Eur. Food Res. Technol. 235, 303–313. doi:10.1007/s00217-012-1762-3 Azzolini, M., Tosi, E., Lorenzini, M., Finato, F., Zapparoli, G., 2015. Contribution to the aroma of white wines by controlled Torulaspora delbrueckii cultures in association with Saccharomyces cerevisiae. World J. Microbiol. Biotechnol. 31, 277–293. doi:10.1007/s11274-014-1774-1 Bagheri, B., Zambelli, P., Vigentini, I., Bauer, F.F., Setati, M.E., 2018. Investigating the effect of selected non-Saccharomyces species on wine ecosystem function and major volatiles. Front. Bioeng. Biotechnol. 6, 1–12. 216

doi:10.3389/fbioe.2018.00169 Balmaseda, A., Bordons, A., Reguant, C., Bautista-Gallego, J., 2018. Non- Saccharomyces in wine: Effect upon Oenococcus oeni and malolactic fermentation. Front. Microbiol. 9, 1–8. doi:10.3389/fmicb.2018.00534 Barata, A., Malfeito-Ferreira, M., Loureiro, V., 2012. The microbial ecology of wine grape berries. Int. J. Food Microbiol. 153, 243–259. doi:10.1016/j.ijfoodmicro.2011.11.025 Barrajón, N., Arévalo-Villena, M., Rodríguez-Aragón, L.J., Briones, a., 2009. Ecological study of wine yeast in inoculated vats from La Mancha region. Food Control 20, 778–783. doi:10.1016/j.foodcont.2008.10.002 Barrajón, N., Capece, A., Arévalo-Villena, M., Briones, A., Romano, P., 2011. Co- inoculation of different Saccharomyces cerevisiae strains and influence on volatile composition of wines. Food Microbiol. 28, 1080–6. doi:10.1016/j.fm.2011.02.016 Bartowsky, E.J., 2009. Bacterial spoilage of wine and approaches to minimize it. Lett. Appl. Microbiol. 48, 149–156. doi:10.1111/j.1472-765X.2008.02505.x Bartowsky, E.J., Henschke, P.A., 2004. The ‘buttery’ attribute of wine — diacetyl — desirability, spoilage and beyond. Int. J. Food Microbiol. 96, 235–252. doi:10.1016/j.ijfoodmicro.2004.05.013 Bartowsky, E.J., Pretorius, I.S., 2009. Microbial Formation and Modification of Flavor and Off-Flavor Compounds in Wine, in: Biology of Microorganisms on Grapes, in Must and in Wine. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 209– 231. doi:10.1007/978-3-540-85463-0_11 Bellincontro, A., Catelli, C., Cotarella, R., Mencarelli, F., 2017. Postharvest ozone fumigation of grapes to prevent the use of sulfites and to increase in wine. Aust. J. Grape Wine Res. 23, 200–206. doi:10.1111/ajgw.12257 Beltran, G., Torija, M.J., Novo, M., Ferrer, N., Poblet, M., Guillamón, J.M., Rozès, N., Mas, A., 2002. Analysis of yeast populations during alcoholic fermentation: A six year follow-up study. Syst. Appl. Microbiol. 25, 287–293. doi:10.1078/0723-2020-00097 Bely, M., Stoeckle, P., Masneuf-Pomarède, I., Dubourdieu, D., 2008. Impact of mixed Torulaspora delbrueckii-Saccharomyces cerevisiae culture on high-sugar fermentation. Int. J. Food Microbiol. 122, 312–320. doi:10.1016/j.ijfoodmicro.2007.12.023 Benito, S., Palomero, F., Galvez, L., Morata, A., Calderon, F., Palmero, D., Suarez- Lepe, J.A., 2014. Quality and composition of red wine fermented with Schizosaccharomyces pombe as sole fermentative yeast, and in mixed and sequential fermentations with Saccharomyces cerevisiae. Food Technol. Biotechnol. 52, 376–382. Benucci, I., Liburdi, K., Cerreti, M., Esti, M., 2016. Characterization of active dry 217

wine yeast during starter culture (pied de cuve) preparation for sparkling wine production. J. Food Sci. 81, M2015–M2020. doi:10.1111/1750-3841.13379 Bisson, L.F., 1999. Stuck and sluggish fermentations. Am. J. Enol. Vitic. 50, 107– 119. Bisson, L.F., Joseph, C.M.L., 2009. Yeasts, in: König, H., Unden, G., Fröhlich, J. (Eds.), Biology of Microorganisms on Grapes, in Must and in Wine. Springer- Verlag, Berlin, pp. 47–60. Blackwell, M., Kurtzman, C.P., 2016. Social wasps promote social behavior in Saccharomyces spp. Proc. Natl. Acad. Sci. 113, 1971–1973. doi:10.1073/pnas.1600173113 Blanco, P., Orriols, I., Losada, a., 2011. Survival of commercial yeasts in the winery environment and their prevalence during spontaneous fermentations. J. Ind. Microbiol. Biotechnol. 38, 235–239. doi:10.1007/s10295-010-0818-2 Blanco, P., Ramilo, A., Cerdeira, M., Orriols, I., 2006. Genetic diversity of wine Saccharomyces cerevisiae strains in an experimental winery from Galicia (NW Spain). Antonie van Leeuwenhoek, Int. J. Gen. Mol. Microbiol. 89, 351–357. doi:10.1007/s10482-005-9038-6 Bokulich, N. a., Ohta, M., Richardson, P.M., Mills, D. a., 2013. Monitoring seasonal changes in winery-resident microbiota. PLoS One 8. doi:10.1371/journal.pone.0066437 Bokulich, N.A., Joseph, C.M.L., Allen, G., Benson, A.K., Mills, D.A., 2012. Next- generation sequencing reveals significant bacterial diversity of botrytized wine. PLoS One 7, 3–12. doi:10.1371/journal.pone.0036357 Bokulich, N.A., Mills, D.A., 2013. Improved selection of Internal Transcribed Spacer- specific primers enables quantitative, ultra-high-throughput profiling of fungal communities. Appl. Environ. Microbiol. 79, 2519–2526. doi:10.1128/AEM.03870-12 Bokulich, N.A., Mills, D.A., 2012. Next-generation approaches to the microbial ecology of food fermentations. BMB Rep. 45, 377–389. Bokulich, N.A., Swadener, M., Sakamoto, K., Mills, D.A., Bisson, L.F., 2014. Sulfur dioxide treatment alters wine microbial diversity and fermentation progression in a dose-dependent fashion. Am. J. Enol. Vitic. 66, 1–21. doi:10.5344/ajev.2014.14096 Bolyen, E., Rideout, J.R., Dillon, M.R., Bokulich, N.A., Abnet, C.C., Al-Ghalith, G.A., Alexander, H., Alm, E.J., Arumugam, M., Asnicar, F., Bai, Y., Bisanz, J.E., Bittinger, K., Brejnrod, A., Brislawn, C.J., Brown, C.T., Callahan, B.J., Caraballo-Rodríguez, A.M., Chase, J., Cope, E.K., Da Silva, R., Dorrestein, P.C., Douglas, G.M., Durall, D.M., Duvallet, C., Edwardson, C.F., Ernst, M., Estaki, M., Fouquier, J., Gauglitz, J.M., Gibson, D.L., Gonzalez, A., Gorlick, K., Guo, J., Hillmann, B., Holmes, S., Holste, H., Huttenhower, C., Huttley, G.A., Janssen, S., Jarmusch, A.K., Jiang, L., Kaehler, B.D., Kang, K. Bin, Keefe, C.R., Keim, P., Kelley, S.T., Knights, D., Koester, I., Kosciolek, T., Kreps, J., 218

Languille, M.G., Lee, J., Ley, R., Liu, Y.-X., Loftfield, E., Lozupone, C., Maher, M., Marotz, C., Martin, B.D., McDonald, D., McIver, L.J., Melnik, A. V, Metcalf, J.L., Morgan, S.C., Morton, J.T., Naimey, A.T., Navas-Molina, J.A., Nothias, L.F., Orchanian, S.B., Pearson, T., Peoples, S.L., Petras, D., Preuss, M.L., Pruesse, E., Rasmussen, L.B., Rivers, A., Robeson, M.S.I., Rosenthal, P., Segata, N., Shaffer, M., Shiffer, A., Sinha, R., Song, S.J., Spear, S.R., Swafford, A.D., Thompson, L.R., Torres, P.J., Trinh, P., Tripathi, A., Turnbaugh, P.J., Ul-Hasan, S., Van der Hooft, J.J., Vargas, F., Vàzquez-Baeza, Y., Vogtmann, E., Von Hippel, M., Walters, W., Wan, Y., Wang, M., Warren, J., Weber, K.C., Williamson, C.H., Willis, A.D., Xu, Z.Z., Zaneveld, J.R., Zhang, Y., Zhu, Q., Knight, R., Caporaso, J.G., 2018. QIIME 2 : Reproducible, interactive, scalable, and extensible microbiome data science. PeerJ. doi:https://doi.org/10.7287/peerj.preprints.27295v2 Borneman, A.R., Forgan, A.H., Kolouchova, R., Fraser, J.A., Schmidt, S.A., 2016. Whole genome comparison reveals high levels of inbreeding and strain redundancy across the spectrum of commercial wine strains of Saccharomyces cerevisiae . G3 Genes, , Genet. 6, 957–971. doi:10.1534/g3.115.025692 Borneman, A.R., Pretorius, I.S., 2015. Genomic insights into the Saccharomyces sensu stricto complex. Genetics 199, 281–291. doi:10.1534/genetics.114.173633 Boroski, M., Crupi, P., Tamborra, P., Antonacci, D., Toci, A.T., 2017. Influence of winemaking techniques with low sulphur dioxide on wine varieties Chardonnay, Pinot and . J. Food Nutr. Res. 56, 326–334. Branco, P., Monteiro, M., Moura, P., Albergaria, H., 2012. Survival rate of wine- related yeasts during alcoholic fermentation assessed by direct live/dead staining combined with fluorescence in situ hybridization. Int. J. Food Microbiol. 158, 49–57. doi:10.1016/j.ijfoodmicro.2012.06.020 Bruvo, R., Michiels, N.K., D’Souza, T.G., Schulenburg, H., 2004. A simple method for the calculation of microsatellite genotype distances irrespective of ploidy level. Mol. Ecol. 13, 2101–2106. doi:10.1111/j.1365-294X.2004.02209.x Bush, R.K., Taylor, S.L., Busse, W., 1986. A critical evaluation of clinical trials in reactions to sulfites. J. Allergy Clin. Immunol. 78, 191–202. Callahan, B.J., McMurdie, P.J., Rosen, M.J., Han, A.W., Johnson, A.J.A., Holmes, S.P., 2016. DADA2: High resolution sample inference from Illumina amplicon data. Nat. Methods 13, 581–583. doi:10.1038/nmeth.3869.DADA2 Capece, A., Siesto, G., Romaniello, R., Lagreca, V.M., Pietrafesa, R., Calabretti, A., Romano, P., 2013. Assessment of competition in wine fermentation among wild Saccharomyces cerevisiae strains isolated from grapes in Tuscany region. LWT - Food Sci. Technol. 54, 485–492. doi:10.1016/j.lwt.2013.07.001 Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., Fierer, N., Peña, A.G., Goodrich, J.K., Gordon, J.I., Huttley, G. a, Kelley, S.T., Knights, D., Koenig, J.E., Ley, R.E., Lozupone, C. a, Mcdonald, D., 219

Muegge, B.D., Pirrung, M., Reeder, J., Sevinsky, J.R., Turnbaugh, P.J., Walters, W. a, Widmann, J., Yatsunenko, T., Zaneveld, J., Knight, R., 2010. QIIME allows analysis of high- throughput community sequencing data. Nat. Publ. Gr. 7, 335–336. doi:10.1038/nmeth0510-335 Capozzi, V., Di Toro, M.R., Grieco, F., Michelotti, V., Salma, M., Lamontanara, A., Russo, P., Orrù, L., Alexandre, H., Spano, G., 2016. Viable But Not Culturable (VBNC) state of Brettanomyces bruxellensis in wine: New insights on molecular basis of VBNC behaviour using a transcriptomic approach. Food Microbiol. 59, 196–204. doi:10.1016/j.fm.2016.06.007 Capozzi, V., Ladero, V., Beneduce, L., Fernández, M., Alvarez, M.A., Benoit, B., Laurent, B., Grieco, F., Spano, G., 2011. Isolation and characterization of tyramine-producing Enterococcus faecium strains from red wine. Food Microbiol. 28, 434–439. doi:10.1016/j.fm.2010.10.005 Cappello, M.S., Zapparoli, G., Logrieco, A., Bartowsky, E.J., 2016. Linking wine lactic acid bacteria diversity with wine aroma and flavour. Int. J. Food Microbiol. 243, 16–27. doi:10.1016/j.ijfoodmicro.2016.11.025 Casalone, E., Colella, C.M., Daly, S., Gallori, E., Moriani, L., Polsinelli, M., 1992. Mechanism of resistance to sulphite in Saccharomyces cerevisiae. Curr. Genet. 22, 435–440. doi:10.1007/BF00326407 Castellari, L., Ferruzzi, M., Magrini, A., Giudici, P., Passarelli, P., Zambonelli, C., 1994. Unbalanced wine fermentation by cryotolerant vs. non-cryotolerant Saccharomyces strains. Vitis 33, 49–52. Cavalieri, D., McGovern, P.E., Hartl, D.L., Mortimer, R., Polsinelli, M., 2003. Evidence for S. cerevisiae fermentation in ancient wine. J. Mol. Evol. 57. doi:10.1007/s00239-003-0031-2 Charoenchai, C., Fleet, G.H., Henschke, P.A., Todd, B.E.N.T., 1997. Screening of non- Saccharomyces wine yeasts for the presence of extracellular hydrolytic enzymes. Aust. J. Grape Wine Res. 3, 2–8. doi:10.1111/j.1755- 0238.1997.tb00109.x Cheraiti, N., Guezenec, S., Salmon, J.-M., 2005. interactions between Saccharomyces cerevisiae and Saccharomyces uvarum in mixed culture under enological conditions. Appl. Environ. Microbiol. 71, 255–60. doi:10.1128/AEM.71.1.255-260.2005 Christiaens, J.F., Franco, L.M., Yaksi, E., Verstrepen, K.J., Christiaens, J.F., Franco, L.M., Cools, T.L., Meester, L. De, Michiels, J., Wenseleers, T., 2014. The fungal aroma gene ATF1 promotes dispersal of yeast cells through insect vectors. Cell Rep. 9, 425–432. doi:10.1016/j.celrep.2014.09.009 Ciani, M., Beco, L., Comitini, F., 2006. Fermentation behaviour and metabolic interactions of multistarter wine yeast fermentations. Int. J. Food Microbiol. 108, 239–245. doi:10.1016/j.ijfoodmicro.2005.11.012 Ciani, M., Comitini, F., 2011. Non-Saccharomyces wine yeasts have a promising role in biotechnological approaches to winemaking. Ann. Microbiol. 61, 25–32. 220

doi:10.1007/s13213-010-0069-5 Ciani, M., Comitini, F., Mannazzu, I., Domizio, P., 2010. Controlled mixed culture fermentation: A new perspective on the use of non-Saccharomyces yeasts in winemaking. FEMS Yeast Res. 10, 123–133. doi:10.1111/j.1567- 1364.2009.00579.x Clarke, K., Gorley, R., 2006. PRIMER v6: User Manual/Tutorial. Clarke, O., 2004. ’s companion: an essential guide for all lovers of Australian wine. Harcourt, Orlando, FL. Clavijo, A., Calderón, I.L., Paneque, P., 2011a. Effect of the use of commercial Saccharomyces strains in a newly established winery in Ronda (Málaga, Spain). Antonie van Leeuwenhoek, Int. J. Gen. Mol. Microbiol. 99, 727–731. doi:10.1007/s10482-010-9514-5 Clavijo, A., Calderón, I.L., Paneque, P., 2011b. Yeast assessment during alcoholic fermentation inoculated with a natural “pied de cuve” or a commercial yeast strain. World J. Microbiol. Biotechnol. 27, 1569–1577. doi:10.1007/s11274-010- 0609-y Cliff, M.A., Dever, M.C., 1996. Sensory and compositional profiles of British Columbia Chardonnay and Pinot noir wines. Food Res. Int. 29, 317–323. doi:10.1016/0963-9969(96)83271-6 Cocolin, L., Mills, D.A., 2003. Wine yeast inhibition by sulfur dioxide: a comparison of culture-dependent and independent methods. Am. J. Enol. Vitic. 54, 125– 130. Coetzee, C., Lisjak, K., Nicolau, L., Kilmartin, P., du Toit, W.J., 2013. Oxygen and sulfur dioxide additions to Sauvignon blanc must: Effect on must and wine composition. Flavour Fragr. J. 28, 155–167. doi:10.1002/ffj.3147 Combina, M., Elía, A., Mercado, L., Catania, C., Ganga, A., Martinez, C., 2005. Dynamics of indigenous yeast populations during spontaneous fermentation of wines from Mendoza, Argentina. Int. J. Food Microbiol. 99, 237–243. doi:10.1016/j.ijfoodmicro.2004.08.017 Comitini, F., Ferretti, R., Clementi, F., Mannazzu, I., Ciani, M., 2005. Interactions between Saccharomyces cerevisiae and malolactic bacteria: Preliminary characterization of a yeast proteinaceous compound(s) active against Oenococcus oeni. J. Appl. Microbiol. 99, 105–111. doi:10.1111/j.1365- 2672.2005.02579.x Comitini, F., Gobbi, M., Domizio, P., Romani, C., Lencioni, L., Mannazzu, I., Ciani, M., 2011. Selected non-Saccharomyces wine yeasts in controlled multistarter fermentations with Saccharomyces cerevisiae. Food Microbiol. 28, 873–882. doi:10.1016/j.fm.2014.05.007 Constantí, M., Poblet, M., Arola, L., Mas, A., Guillamón, J.M., 1997. Analysis of yeast population during alcoholic fermentation in a newly established winery. Am. J. Enol. Vitic. 48, 339–343. 221

Constantí, M., Reguant, C., Poblet, M., Zamora, F., Mas, A., Guillamon, J.M., 1998. Molecular analysis of yeast population dynamics: effect of sulphur dioxide and inoculum on must fermentation. Int. J. Food Microbiol. 41, 169–175. Contreras, A., Curtin, C., Varela, C., 2015. Yeast population dynamics reveal a potential ‘collaboration’ between Metschnikowia pulcherrima and Saccharomyces uvarum for the production of reduced alcohol wines during Shiraz fermentation. Appl. Microbiol. Biotechnol. 99, 1885–1895. doi:10.1007/s00253-014-6193-6 Costello, P.J., Henschke, P.A., 2002. Mousy off-flavor of wine: Precursors and biosynthesis of the causative N-Heterocycles 2-ethyltetrahydropyridine, 2- acetyltetrahydropyridine, and 2-acetyl-1-pyrroline by Lactobacillus hilgardii DSM 20176. J. Agric. Food Chem. 50, 7079–7087. doi:10.1021/jf020341r Costello, P.J., Lee, T.H., Henschke, P., 2001. Ability of lactic acid bacteria to produce N-heterocycles causing mousy off-flavour in wine. Aust. J. Grape Wine Res. 7, 160–167. doi:10.1111/j.1755-0238.2001.tb00205.x Coton, E., Coton, M., Levert, D., Casaregola, S., Sohier, D., 2006. Yeast ecology in French cider and black olive natural fermentations. Int. J. Food Microbiol. 108, 130–135. Csoma, H., Zakany, N., Capece, A., Romano, P., Sipiczki, M., 2010. Biological diversity of Saccharomyces yeasts of spontaneously fermenting wines in four wine regions: Comparative genotypic and phenotypic analysis. Int. J. Food Microbiol. 140, 239–248. doi:10.1016/j.ijfoodmicro.2010.03.024 Cubillos, F.A., Vásquez, C., Faugeron, S., Ganga, A., Martínez, C., 2009. Self- fertilization is the main sexual reproduction mechanism in native wine yeast populations. FEMS Microb. Ecol. 67, 162–170. doi:10.1111/j.1574- 6941.2008.00600.x Dashko, S., Liu, P., Volk, H., Butinar, L., Piškur, J., Fay, J.C., 2016. Changes in the relative abundance of two Saccharomyces species from oak forests to wine fermentations. Front. Microbiol. 7, 1–12. doi:10.3389/fmicb.2016.00215 de Sousa Abreu, R., Penalva, L.O., Marcotte, E.M., Vogel, C., 2009. Global signatures of protein and mRNA expression levels. Mol. Biosyst. 5, 1512–26. doi:10.1039/b908315d de Ullivarri, M.F., Mendoza, L.M., Raya, R.R., 2014. Killer activity of Saccharomyces cerevisiae strains: partial characterization and strategies to improve the biocontrol efficacy in winemaking. Antonie van Leeuwenhoek, Int. J. Gen. Mol. Microbiol. 106, 865–878. doi:10.1007/s10482-014-0256-7 Deed, R.C., Fedrizzi, B., Gardner, R.C., 2017. Influence of fermentation temperature, yeast strain, and grape juice on the aroma chemistry and sensory profile of Sauvignon blanc wines. J. Agric. Food Chem. 65, 8902–8912. doi:10.1021/acs.jafc.7b03229 Delfini, C., Gaia, P., Schellino, R., Strano, M., Pagliara, A., Ambrò, S., 2002. Fermentability of grape must after inhibition with dimethyl dicarbonate (DMDC). 222

J. Agric. Food Chem. 50, 5605–5611. doi:10.1021/jf0256337 Dellaglio, F., Zapparoli, G., Malacrinò, P., Suzzi, G., Torriani, S., 2003. Saccharomyces bayanus var. uvarum and Saccharomyces cerevisiae succession during spontaneous fermentations of Recioto and Amarone wines, Annals of Microbiology. Demuyter, C., Lollier, M., Legras, J.-L., Le Jeune, C., 2004. Predominance of Saccharomyces uvarum during spontaneous alcoholic fermentation, for three consecutive years, in an Alsatian winery. J. Appl. Microbiol. 97, 1140–1148. doi:10.1111/j.1365-2672.2004.02394.x DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L., Keller, K., Huber, T., Dalevi, D., Hu, P., Andersen, G.L., 2006. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072. doi:10.1128/AEM.03006-05 Diez, L., Guadalupe, Z., Ayestarán, B., Ruiz-Larrea, F., 2010. Effect of yeast mannoproteins and grape on the growth of wine lactic acid and acetic acid bacteria. J. Agric. Food Chem. 58, 7731–7739. doi:10.1021/jf100199n Divol, B., du Toit, M., Duckitt, E., 2012. Surviving in the presence of sulphur dioxide: strategies developed by wine yeasts. Appl. Microbiol. Biotechnol. 95, 601–13. doi:10.1007/s00253-012-4186-x Divol, B., Lonvaud-Funel, A., 2005. Evidence for viable but nonculturable yeasts in botrytis-affected wine. J. Appl. Microbiol. 99, 85–93. doi:10.1111/j.1365- 2672.2005.02578.x Divol, B., Miot-Sertier, C., Lonvaud-Funel, A., 2006. Genetic characterization of strains of Saccharomyces cerevisiae responsible for “refermentation” in Botrytis-affected wines. J. Appl. Microbiol. 100, 516–526. doi:10.1111/j.1365- 2672.2005.02818.x Domizio, P., Lencioni, L., Ciani, M., Di Blasi, S., Pontremolesi, C., Sabatelli, M.P., 2007. Spontaneous and inoculated yeast populations dynamics and their effect on organoleptic characters of Vinsanto wine under different process conditions. Int. J. Food Microbiol. 115, 281–289. doi:10.1016/j.ijfoodmicro.2006.10.052 Drysdale, G.S., Fleet, G.H., 1988. Acetic acid bacteria in winemaking: a review. Am. J. Enol. Vitic. 39, 143–154. Dündar, H., 2016. Bacteriocinogenic potential of Enterococcus faecium isolated from wine. Probiotics Antimicrob. Proteins 8, 150–160. doi:10.1007/s12602-016- 9222-1 Edwards, C.G., Jensen, K.A., 1992. Occurrence and characterization of Lactic Acid Bacteria from Washington State wines: Pediococcus spp. Am. J. Enol. Vitic. 43, 233–238. Edwards, C.G., Powers, J.R., Jensen, K.A., Weller, K.M., Peterson, J.C., 1993. Lactobacillus spp. from Washington state wines: Isolation and characterization. 223

J. Food Sci. 58, 453–458. Egli, C.M., Edinger, W.D., Mitrakul, C.M., 1998. Dynamics of indigenous and inoculated yeast populations and their effect on the sensory character of Riesling and Chardonnay wines. J. Appl. Microbiol. 85, 779–789. Erickson, J.P., McKenna, D.N., 2000. Zygosaccharomyces, in: Robinson, R.K., Batt, C.A., Patel, P.D. (Eds.), Encyclopedia of . Academic Press, London. Erkus, O., de Jager, V.C.L., Geene, R.T.C.M., van Alen-Boerrigter, I., Hazelwood, L., van Hijum, S.A.F.T., Kleerebezem, M., Smid, E.J., 2016. Use of propidium monoazide for selective profiling of viable microbial cells during Gouda cheese ripening. Int. J. Food Microbiol. 228, 1–9. doi:10.1016/j.ijfoodmicro.2016.03.027 Falguera, V., Forns, M., Ibarz, A., 2013. UV-vis irradiation: An alternative to reduce SO2 in white wines? LWT - Food Sci. Technol. 51, 59–64. doi:10.1016/j.lwt.2012.11.006 Fang, T.J., Dalmasso, J.P., 1993. Antimicrobial activity of sulfur dioxide to certain lactic acid bacteria isolated from wines. Chinese J. Microbiol. Immunol. 26, 116–131. Fay, J.C., Benavides, J. a., 2005. Evidence for domesticated and wild populations of Saccharomyces cerevisiae. PLoS Genet. 1, 0066–0071. doi:10.1371/journal.pgen.0010005 Fernández-Espinar, M.T., Barrio, E., Querol, A., 2003. Analysis of the genetic variability in the species of the Saccharomyces sensu stricto complex. Yeast 20, 1213–1226. doi:10.1002/yea.1034 Fernández-Espinar, M.T., López, V., Ramón, D., Bartra, E., Querol, A., 2001. Study of the authenticity of commercial wine yeast strains by molecular techniques. Int. J. Food Microbiol. 70, 1–10. doi:10.1016/S0168-1605(01)00502-5 Ferreira, V., López, R., Cacho, J.F., 2000. Quantitative determination of the odorants of young red wines from different grape varieties. J. Sci. Food Agric. 80, 1659–1667. doi:10.1002/1097-0010(20000901)80:11<1659::AID- JSFA693>3.0.CO;2-6 Field, D., Wills, C., 1998. Abundant microsatellite polymorphism in Saccharomyces cerevisiae, and the different distributions of microsatellites in eight prokaryotes and S . cerevisiae, result from strong mutation pressures and a variety of selective forces. PNAS 95, 1647–1652. Fleet, G., 2003. Yeast interactions and wine flavour. Int. J. Food Microbiol. 86, 11– 22. doi:10.1016/S0168-1605(03)00245-9 Fleet, G.H., 2007. Yeasts in foods and beverages: Impact on product quality and safety. Curr. Opin. Biotechnol. 18, 170–175. doi:10.1016/j.copbio.2007.01.010 Fornachon, J.C.M., 1968. Influence of different yeasts on the growth of lactic acid bacteria in wine. J. Sci. Food Agric. 19, 374–378. doi:https://doi.org/10.1002/jsfa.2740190705 224

Fracassetti, D., Gabrielli, M., Costa, C., Tom??s-Barber??n, F.A., Tirelli, A., 2015. Characterization and suitability of polyphenols-based formulas to replace sulfur dioxide for storage of sparkling white wine. Food Control. doi:10.1016/j.foodcont.2015.09.005 Francesca, N., Gaglio, R., Alfonzo, A., Settanni, L., Corona, O., Mazzei, P., Romano, R., Piccolo, A., Moschetti, G., 2016. The wine: Typicality or mere diversity? The effect of spontaneous fermentations and biotic factors on the characteristics of wine. Agric. Agric. Sci. Procedia 8, 769–773. doi:10.1016/J.AASPRO.2016.02.064 Francesco, C., Malfeito Ferreira, J. de C.P., Mota, M. da S.G., 2017. Taster responses to acidity in white wine. Francis, I.L., Newton, J.L., 2005. Determining wine aroma from compositional data. Aust. J. Grape Wine Res. 11, 114–126. doi:10.1111/j.1755- 0238.2005.tb00283.x Fredericks, I.N., du Toit, M., Krügel, M., 2011. Efficacy of ultraviolet radiation as an alternative technology to inactivate microorganisms in grape and wines. Food Microbiol. 28, 510–7. doi:10.1016/j.fm.2010.10.018 Freedman, B.J., 1980. Sulfur dioxide in foods and beverages: its use as a preservative and its effect on asthma. Br. J. Dis. Chest 74, 128–134. Freire, L., Passamani, F.R.F., Thomas, A.B., Nassur, R. de C.M.R., Silva, L.M., Paschoal, F.N., Pereira, G.E., Prado, G., Batista, L.R., 2017. Influence of physical and chemical characteristics of wine grapes on the incidence of Penicillium and Aspergillus fungi in grapes and ochratoxin A in wines. Int. J. Food Microbiol. 241, 181–190. doi:10.1016/j.ijfoodmicro.2016.10.027 Frezier, V., Dubourdieu, D., 1992. Ecology of yeast strains Saccharomyces cerevisiae during spontaneous fermentation in Bordeaux winery. Am. J. Enol. Vitic. 43, 375–380. Gamero, A., Tronchoni, J., Querol, A., Belloch, C., 2013. Production of aroma compounds by cryotolerant Saccharomyces species and hybrids at low and moderate fermentation temperatures. J. Appl. Microbiol. 114, 1405–1414. doi:10.1111/jam.12126 Garcia, A., Carcel, C., Dulau, L., Samson, A., Aguera, E., Agosin, E., Günata, Z., 2002. Influence of a mixed culture with Debaryomyces vanriji and Saccharomyces cerevisiae on the volatiles of a wine. J. Food Sci. 67, 1138–1143. doi:10.1111/j.1365-2621.2002.tb09466.x Gayevskiy, V., Goddard, M.R., 2012. Geographic delineations of yeast communities and populations associated with vines and wines in New Zealand. Int. Soc. Microb. Ecol. J. 6, 1281–1290. doi:10.1038/ismej.2011.195 Gershwin, M.E., Ough, C., Bock, A., Fletcher, M.P., Nagy, S.M., Tuft, D.S., 1985. Grand rounds: adverse reactions to wine. J. Allergy Clin. Immunol. 75, 411– 420.

225

Giraffa, G., 2002. Enterococci from foods. FEMS Microbiol. Rev. 26, 163–171. doi:10.1016/S0168-6445(02)00094-3 Gobbi, M., Comitini, F., Domizio, P., Romani, C., Lencioni, L., Mannazzu, I., Ciani, M., 2013. Lachancea thermotolerans and Saccharomyces cerevisiae in simultaneous and sequential co-fermentation: A strategy to enhance acidity and improve the overall quality of wine. Food Microbiol. 33, 271–281. doi:10.1016/j.fm.2012.10.004 Godálová, Z., Kraková, L., Puškárová, A., Bučková, M., Kuchta, T., Piknová, Ľ., Pangallo, D., 2016. Bacterial consortia at different wine fermentation phases of two typical Central European grape varieties: Blaufränkisch (Frankovka modrá) and Grüner Veltliner (Veltlínske zelené). Int. J. Food Microbiol. 217, 110–116. doi:10.1016/j.ijfoodmicro.2015.10.015 Goddard, M.R., Anfang, N., Tang, R., Gardner, R.C., Jun, C., 2010. A distinct population of Saccharomyces cerevisiae in New Zealand: evidence for local dispersal by insects and human-aided global dispersal in oak barrels. Environ. Microbiol. 12, 63–73. Goldstein, D.B., 1986. Effect of alcohol on cellular membranes. Ann. Emerg. Med. 15, 1013–1018. doi:10.1016/S0196-0644(86)80120-2 González-Arenzana, L., Santamaría, P., López, R., Garijo, P., Gutiérrez, A.R., Garde-Cerdán, T., López-Alfaro, I., 2013. Microwave technology as a new tool to improve microbiological control of oak barrels: A preliminary study. Food Control 30, 536–539. doi:10.1016/j.foodcont.2012.08.008 Goto-Yamamoto, N., Kitano, K., Shiki, K., Yoshida, Y., Suzuki, T., Iwata, T., Yamane, Y., Hara, S., 1998. SSU1-R, a sulfite resistance gene of wine yeast, is an allele of SSU1 with a different upstream sequence. J. Ferment. Bioeng. 86, 427–433. Granett, J., Walker, M.A., Kocsis, L., Omer, A.D., 2001. Biology and management of grape phylloxera. Annu. Rev. Entomol. 46, 387–412. doi:10.1146/annurev.ento.46.1.387 Guerrero, R.F., Cantos-Villar, E., 2015. Demonstrating the efficiency of sulphur dioxide replacements in wine: A parameter review. Trends Food Sci. Technol. 42, 27–43. doi:10.1016/j.tifs.2014.11.004 Guinard, J.-X., Cliff, M., 1987. Descriptive analysis of Pinot noir wines from Carneros, Napa, and Sonoma. Am. J. Enol. Vitic. 38, 211–215. Gustafsson, F., Jiranek, V., Neuner, M., Scholl, C., Morgan, S., Durall, D., 2016. The interaction of two Saccharomyces cerevisiae strains affects fermentation- derived compounds in wine. Fermentation 2, 1–10. doi:10.3390/fermentation2020009 Guth, H., 1997. Quantitation and sensory studies of character impact odorants of different white wine varieties. J. Agric. Food Chem. 45, 3027–3032. doi:10.1021/JF970280A

226

Haggerty, J., Bowyer, P.K., Jiranek, V., Taylor, D.K., 2014. Comparative study on the sensitivity of solid-phase microextraction fibre coatings for the analysis of fermentation bouquet compounds. Aust. J. Grape Wine Res. 20, 378–385. doi:10.1111/ajgw.12100 Haggerty, J.J., Taylor, D.K., Jiranek, V., 2016. Monitoring volatile aroma compounds during fermentation in a chemically defined grape juice medium deficient in Leucine. Am. J. Enol. Vitic. 67, 350–355. doi:10.5344/ajev.2016.15026 Hall, B., Durall, D.M., Stanley, G., 2011. Population dynamics of Saccharomyces cerevisiae during spontaneous fermentation at a British Columbia winery. Am. J. Enol. Vitic. 62, 66–72. doi:10.5344/ajev.2010.10054 Heard, G.M., Fleet, G.H., 1987. Occurrence and growth of killer yeasts during wine fermentation. Appl. Environ. Microbiol. 53, 2171–2174. Heard, G.M., Fleet, G.H., 1985. Growth of natural yeast flora during the fermentation of inoculated wines. Appl. Environ. Microbiol. 50, 727–728. Henick-Kling, T., Edinger, W., Daniel, P., Monk, P., 1998. Selective effects of sulfur dioxide and yeast starter culture addition on indigenous yeast populations and sensory characteristics of wine. J. Appl. Microbiol. 84, 865–876. doi:10.1046/j.1365-2672.1998.00423.x Herlemann, D.P.R., Labrenz, M., Jürgens, K., Bertilsson, S., Waniek, J.J., Andersson, A.F., 2011. Transitions in bacterial communities along the 2000 km salinity gradient of the Baltic Sea. ISME J. 5, 1571–1579. doi:10.1038/ismej.2011.41 Herraiz, T., Reglero, G., Herraiz, M., Martin-Alvarez, P.J., Cabezudo, M.D., 1990. The influence of the yeast and type of culture on the volatile composition of wines fermented without sulfur dioxide. Am. J. Enol. Vitic. 41, 313–318. Hierro, N., Esteve-Zarzoso, B., González, Á., Mas, A., Guillamón, J.M., 2006. Real- time quantitative PCR (QPCR) and reverse transcription-QPCR for detection and enumeration of total yeasts in wine. Appl. Environ. Microbiol. 72, 7148– 7155. doi:10.1128/AEM.00388-06 Hilber-Bodmer, M., Schmid, M., Ahrens, C.H., Freimoser, F.M., 2017. Competition assays and physiological experiments of soil and phyllosphere yeasts identify Candida subhashii as a novel antagonist of filamentous fungi. BMC Microbiol. 17, 4. doi:10.1186/s12866-016-0908-z Hinze, H., Holzer, H., 1986. Analysis of the energy metabolism after incubation of Saccharomyces cerevisiae with sulfite or nitrite. Arch. Microbiol. 145, 27–31. Hoang, D., Kopp, A., Chandler, J.A., 2015. Interactions between Drosophila and its natural yeast symbionts—is Saccharomyces cerevisiae a good model for studying the fly-yeast relationship? PeerJ 3, e1116. doi:10.7717/peerj.1116 Hoppe, A., 2012. What mRNA abundances can tell us about metabolism. Metabolites 2, 614–31. doi:10.3390/metabo2030614 Hothorn, T., Bretz, F., Westfall, P., 2008. Simultaneous inference in general 227

parametric models. Biometrical J. 50, 346–363. doi:10.1002/bimj.200810425 Howe, P.A., Worobo, R., Sacks, G.L., 2018. Conventional measurements of sulfur dioxide (SO2) in red wine overestimate SO2 antimicrobial activity. Am. J. Enol. Vitic. 1, ajev.2018.17037. doi:10.5344/ajev.2018.17037 Howell, K.S., Cozzolino, D., Bartowsky, E.J., Fleet, G.H., Henschke, P. a., 2006. Metabolic profiling as a tool for revealing Saccharomyces interactions during wine fermentation. FEMS Yeast Res. 6, 91–101. doi:10.1111/j.1567- 1364.2005.00010.x Hranilovic, A., Bely, M., Masneuf-Pomarede, I., Jiranek, V., Albertin, W., 2017. The evolution of Lachancea thermotolerans is driven by geographical determination, anthropisation and flux between different ecosystems. PLoS One 12, 1–17. doi:10.1371/journal.pone.0184652 Inoue, T., Nagatomi, Y., Uyama, A., Mochizuki, N., 2013. Degradation of aflatoxin B1 during the fermentation of alcoholic beverages. Toxins (Basel). 5, 1219– 1229. doi:10.3390/toxins5071219 Ivannikova, Y. V, Naumova, E.S., Naumov, G.I., 2007. Viral dsRNA in the wine yeast Saccharomyces bayanus var. uvarum. Res. Microbiol. 158, 638–643. doi:10.1016/j.resmic.2007.07.008 Ivanova, V., Vojnoski, B., Stefova, M., 2011. Effect of the winemaking practices and aging on phenolic content of Smederevka and Chardonnay wines. Food Bioprocess Technol. 4, 1512–1518. doi:10.1007/s11947-011-0566-y Izquierdo-Canas, P.M., Garcia-Romero, E., Huertas-Nebreda, B., Gomez-Alonso, S., 2012. Colloidal silver complex as an alternative to sulphur dioxide in winemaking. Food Control 23, 73–81. doi:10.1016/j.foodcont.2011.06.014 Izquierdo Cañas, P.M., Palacios García, a. T., García Romero, E., 2011. Enhancement of flavour properties in wines using sequential inoculations of non-Saccharomyces (Hansenula and Torulaspora) and Saccharomyces yeast starter. Vitis - J. Grapevine Res. 50, 177–182. Izquierdo Cañas, P.M., Pérez-Martín, F., Romero, E.G., Prieto, S.S., Herreros, M.D.L.L.P., 2012. Influence of inoculation time of an autochthonous selected malolactic bacterium on volatile and sensory profile of and Merlot wines. Int. J. Food Microbiol. 156, 245–254. doi:10.1016/j.ijfoodmicro.2012.03.033 Jackowetz, J.N., Mira de Orduña, R., 2012. Metabolism of SO2 binding compounds by Oenococcus oeni during and after malolactic fermentation in white wine. Int. J. Food Microbiol. 155, 153–7. doi:10.1016/j.ijfoodmicro.2012.01.025 Javier Gallego, F., Angeles Pérez, M., Núñez, Y., Hidalgo, P., 2005. Comparison of RAPDs, AFLPs and SSR markers for the genetic analysis of yeast strains of Saccharomyces cerevisiae. Food Microbiol. 22, 561–568. doi:10.1016/J.FM.2004.11.019 Jolly, N.P., Varela, C., Pretorius, I.S., 2014. Not your ordinary yeast: Non- 228

Saccharomyces yeasts in wine production uncovered. FEMS Yeast Res. 14, 215–237. doi:10.1111/1567-1364.12111 Jouhten, P., Ponomarova, O., Gonzalez, R., Patil, K.R., 2016. Saccharomyces cerevisiae metabolism in ecological context. FEMS Yeast Res. 16, 80. doi:10.1093/femsyr/fow080 Kamvar, Z.N., Brooks, J.C., Grünwald, N.J., 2015. Novel R tools for analysis of genome-wide population genetic data with emphasis on clonality. Front. Genet. 6, 208. doi:10.3389/fgene.2015.00208 Kamvar, Z.N., Tabima, J.F., Grünwald, N.J., 2014. Poppr: An R package for genetic analysis of populations with clonal, partially clonal, and/or sexual reproduction. PeerJ 2, e281. doi:10.7717/peerj.281 Katoh, K., Standley, D.M., 2013. MAFFT multiple sequence alignment software version 7: Improvements in performance and usability. Mol. Biol. Evol. 30, 772– 780. doi:10.1093/molbev/mst010 Kemsawasd, V., Branco, P., Almeida, M.G., Caldeira, J., Albergaria, H., Arneborg, N., 2015. Cell-to-cell contact and antimicrobial peptides play a combined role in the death of Lachanchea thermotolerans during mixed-culture alcoholic fermentation with Saccharomyces cerevisiae. FEMS Microbiol. Lett. 362. doi:10.1093/femsle/fnv103 Kim, D.H., Hong, Y.A., Park, H.D., 2008. Co-fermentation of grape must by Issatchenkia orientalis and Saccharomyces cerevisiae reduces the malic acid content in wine. Biotechnol. Lett. 30, 1633–1638. doi:10.1007/s10529-008- 9726-1 Kim, D.M., Choi, S.-H., Ko, B.S., Jeong, G.-Y., Jang, H.-B., Han, J.-G., Jeong, K.-H., Lee, H.-Y., Won, Y., Kim, I.-C., 2012. Reduction of PDC1 expression in S. cerevisiae with xylose isomerase on xylose medium. Bioprocess Biosyst. Eng. 35, 183–189. doi:10.1007/s00449-011-0638-4 King, A.D., Ponting, J.D., Sanshuck, D.W., Jackson, R., Mihara, K., 1981. Factors affecting death of yeast by sulfur dioxide. J. Food Prot. 44, 92–97. doi:10.4315/0362-028X-44.2.92 King, E.S., Dunn, R.L., Heymann, H., 2013. The influence of alcohol on the sensory perception of red wines. Food Qual. Prefer. 28, 235–243. doi:10.1016/j.foodqual.2012.08.013 King, E.S., Kievit, R.L., Curtin, C., Swiegers, J.H., Pretorius, I.S., Bastian, S.E.P., Leigh Francis, I., 2010. The effect of multiple yeasts co-inoculations on Sauvignon Blanc wine aroma composition, sensory properties and consumer preference. Food Chem. 122, 618–626. doi:10.1016/j.foodchem.2010.03.021 King, E.S., Swiegers, J.H., Travis, B., Leigh Francis, I., Bastian, S.E.P., Pretorius, I.S., 2008. Coinoculated fermentations using Saccharomyces yeasts affect the volatile composition and sensory properties of Vitis vinifera L . cv . Sauvignon Blanc wines. J. Agric. Food Chem. 56, 10829–10837. doi:10.1021/jf801695h

229

Kioroglou, D., LLeixá, J., Mas, A., del Carmen Portillo, M., 2018. Massive sequencing: A new tool for the control of alcoholic fermentation in wine? Fermentation 4, 7. doi:10.3390/fermentation4010007 Knight, S., Goddard, M.R., 2014. Quantifying separation and similarity in a Saccharomyces cerevisiae metapopulation. ISME J. 9, 361–370. doi:10.1038/ismej.2014.132 Knight, S., Klaere, S., Fedrizzi, B., Goddard, M.R., 2015. Regional microbial signatures positively correlate with differential wine phenotypes: Evidence for a microbial aspect to terroir. Sci. Rep. 5, 14233. doi:10.1038/srep14233 Knoll, C., Fritsch, S., Schnell, S., Grossmann, M., Krieger-Weber, S., du Toit, M., Rauhut, D., 2012. Impact of different malolactic fermentation inoculation scenarios on Riesling wine aroma. World J. Microbiol. Biotechnol. 28, 1143– 1153. doi:10.1007/s11274-011-0917-x Kõljalg, U., Nilsson, R.H., Abarenkov, K., Tedersoo, L., Taylor, A.F.S., Bahram, M., Bates, S.T., Bruns, T.D., Bengtsson-Palme, J., Callaghan, T.M., Douglas, B., Drenkhan, T., Eberhardt, U., Dueñas, M., Grebenc, T., Griffith, G.W., Hartmann, M., Kirk, P.M., Kohout, P., Larsson, E., Lindahl, B.D., Lücking, R., Martín, M.P., Matheny, P.B., Nguyen, N.H., Niskanen, T., Oja, J., Peay, K.G., Peintner, U., Peterson, M., Põldmaa, K., Saag, L., Saar, I., Shüβler, A., Scott, J.A., Senés, C., Smith, M.E., Suija, A., Taylor, D.L., Telleria, M.T., Weiss, M., Larsson, K.-H., 2013. Towards a unified paradigm for sequence-based identification of fungi. Mol. Ecol. 22, 5271–5277. doi:10.1111/mec.12481 Kuanyshev, N., Adamo, G.M., Porro, D., Branduardi, P., 2017. The spoilage yeast : Foe or friend? Yeast 34, 359–370. doi:10.1002/yea.3238 Kümmerle, M., Scherer, S., Seiler, H., 1998. Rapid and reliable identification of food- borne yeasts by Fourier-transform infrared spectroscopy. Appl. Environ. Microbiol. 64, 2207–2214. Lam, S.S.T.H., Howell, K.S., 2015. Drosophila-associated yeast species in vineyard ecosystems. FEMS Microbiol. Lett. 362, 1–7. doi:10.1093/femsle/fnv170 Lange, J.N., Faasse, E., Tantikachornkiat, M., Gustafsson, F.S., Halvorsen, L.C., Kluftinger, A., Ledderhof, D., Durall, D.M., 2014. Implantation and persistence of yeast inoculum in Pinot noir fermentations at three Canadian wineries. Int. J. Food Microbiol. 180, 56–61. doi:10.1016/j.ijfoodmicro.2014.04.003 Lavigne-Cruège, V., Dubourdieu, D., 1996. Demonstraction and interpretation of the yeast lee ability to adsorb certain volatile thiols contained in wine. OENO One 30, 201. doi:10.20870/oeno-one.1996.30.4.1096 Legras, J.L., Merdinoglu, D., Cornuet, J.M., Karst, F., 2007. Bread, beer and wine: Saccharomyces cerevisiae diversity reflects human history. Mol. Ecol. 16, 2091–2102. doi:10.1111/j.1365-294X.2007.03266.x Legras, J.L., Ruh, O., Merdinoglu, D., Karst, F., 2005. Selection of hypervariable microsatellite loci for the characterization of Saccharomyces cerevisiae strains. 230

Int. J. Food Microbiol. 102, 73–83. doi:10.1016/j.ijfoodmicro.2004.12.007 Li, E., Liu, C., Liu, Y., 2012. Evaluation of yeast diversity during wine fermentations with direct inoculation and pied de cuve method at an industrial scale. J. Microbiol. Biotechnol. 22, 960–966. doi:10.4014/jmb.1111.11013 Linderholm, A.L., Findleton, C.L., Kumar, G., Hong, Y., Bisson, L.F., 2008. Identification of genes affecting hydrogen sulfide formation in Saccharomyces cerevisiae. Appl. Environ. Microbiol. 74, 1418–1427. doi:10.1128/AEM.01758- 07 Liti, G., Carter, D.M., Moses, A.M., Warringer, J., Parts, L., James, S.A., Davey, R.P., Roberts, I.N., Burt, A., Koufopanou, V., Tsai, I.J., Bergman, C.M., Bensasson, D., O’Kelly, M.J.T., Van Oudenaarden, A., Barton, D.B.H., Bailes, E., Nguyen, A.N., Jones, M., Quail, M.A., Goodhead, I., Sims, S., Smith, F., Blomberg, A., Durbin, R., Louis, E.J., 2009. Population genomics of domestic and wild yeasts. Nature 458, 337–341. doi:10.1038/nature07743 Liu, J., Toldam-Andersen, T.B., Petersen, M.A., Zhang, S., Arneborg, N., Bredie, W.L.P., 2015. Instrumental and sensory characterisation of Solaris white wines in Denmark. Food Chem. 166, 133–142. doi:10.1016/j.foodchem.2014.05.148 Liu, P.T., Lu, L., Duan, C.Q., Yan, G.L., 2016. The contribution of indigenous non- Saccharomyces wine yeast to improved aromatic quality of Cabernet Sauvignon wines by spontaneous fermentation. LWT - Food Sci. Technol. 71, 356–363. doi:10.1016/j.lwt.2016.04.031 Liu, S.-Q., Pilone, G.J., 2000. An overview of formation and roles of acetaldehyde in winemaking with emphasis on microbiological implications. Int. J. Food Sci. Technol. 35, 49–61. doi:10.1046/j.1365-2621.2000.00341.x Loira, I., Vejarano, R., Bañuelos, M. a., Morata, A., Tesfaye, W., Uthurry, C., Villa, A., Cintora, I., Suárez-Lepe, J. a., 2014. Influence of sequential fermentation with Torulaspora delbrueckii and Saccharomyces cerevisiae on wine quality. LWT - Food Sci. Technol. 59, 915–922. doi:10.1016/j.lwt.2014.06.019 Lonvaud-Funel, A., 1999. Lactic acid bacteria in the quality improvement and depreciation of wine, Antonie van Leeuwenhoek. López-Malo, M., Querol, A., Guillamon, J.M., 2013. Metabolomic comparison of Saccharomyces cerevisiae and the cryotolerant species S. bayanus var. uvarum and S. kudriavzevii during wine fermentation at low temperature. PLoS One 8, e60135. doi:10.1371/journal.pone.0060135 Lorion, C.M., Kennedy, B.P., 2009. Relationships between deforestation, riparian forest buffers and benthic macroinvertebrates in neotropical headwater streams. Freshw. Biol. 54, 165–180. doi:10.1111/j.1365-2427.2008.02092.x Loureiro, V., Malfeito-Ferreira, M., 2003. Spoilage yeasts in the wine industry. Int. J. Food Microbiol. 86, 23–50. doi:10.1016/S0168-1605(03)00246-0 Magyar, I., Tóth, T., 2011. Comparative evaluation of some oenological properties in wine strains of Candida stellata, Candida zemplinina, Saccharomyces uvarum 231

and Saccharomyces cerevisiae. Food Microbiol. 28, 94–100. doi:10.1016/j.fm.2010.08.011 Maier, K., Hinze, H., Leuschel, L., 1986. Mechanism of sulfite action on the energy metabolism of Saccharomyces cerevisiae. Biochim. Biophys. Acta - Bioenerg. 848, 120–130. doi:10.1016/0005-2728(86)90167-2 Maragkoudakis, P.A., Nardi, T., Bovo, B., D’andrea, M., Howell, K.S., Giacomini, A., Corich, V., 2013. Biodiversity, dynamics and ecology of bacterial community during grape marc storage for the production of grappa. doi:10.1016/j.ijfoodmicro.2013.01.005 Margalit, Y., Crum, J.D., 2004. Concepts in Wine Chemistry. Wine Appreciation Guild, San Francisco. Marti-Raga, M., Sancho, M., Guillamon, J.M., Mas, A., Beltran, G., 2015. The effect of nitrogen addition on the fermentative performance during sparkling wine production. Food Res. Int. 67, 126–135. doi:10.1016/j.foodres.2014.10.033 Martineau, B., Acree, T.E., Henick-Kling, T., 1995. Effect of wine type on the detection threshold for diacetyl. Food Res. Int. 28, 139–143. doi:10.1016/0963- 9969(95)90797-E Martiniuk, J.T., Pacheco, B., Russell, G., Tong, S., Backstrom, I., Measday, V., 2016. Impact of commercial strain use on Saccharomyces cerevisiae population structure and dynamics in Pinot noir vineyards and spontaneous fermentations of a Canadian winery. PLoS One 11, e0160259. doi:10.1371/journal.pone.0160259 Masneuf-Pomarede, I., Le Jeune, C., Durrens, P., Lollier, M., Aigle, M., Dubourdieu, D., 2007. Molecular typing of wine yeast strains Saccharomyces bayanus var. uvarum using microsatellite markers. Syst. Appl. Microbiol. 30, 75–82. doi:10.1016/j.syapm.2006.02.006 Masneuf-Pomarede, I., Salin, F., Börlin, M., Coton, E., Coton, M., Le Jeune, C., Legras, J.-L., 2016. Microsatellite analysis of Saccharomyces uvarum diversity. FEMS Yeast Res. 16, fow002. doi:10.1093/femsyr/fow002 Mateo, J.J., Jiménez, M., Pastor, a., Huerta, T., 2001. Yeast starter cultures affecting wine fermentation and volatiles. Food Res. Int. 34, 307–314. doi:10.1016/S0963-9969(00)00168-X Mateo, R., Medina, Á., Mateo, E.M., Mateo, F., Jiménez, M., 2007. An overview of ochratoxin A in beer and wine. Int. J. Food Microbiol. 119, 79–83. doi:10.1016/j.ijfoodmicro.2007.07.029 McDonald, D., Price, M.N., Goodrich, J., Nawrocki, E.P., Desantis, T.Z., Probst, A., Andersen, G.L., Knight, R., Hugenholtz, P., 2012. An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria and archaea. ISME J. 6, 610–618. doi:10.1038/ismej.2011.139 Medina, K., Boido, E., Fariña, L., Gioia, O., Gomez, M.E., Barquet, M., Gaggero, C., Dellacassa, E., Carrau, F., 2013. Increased flavour diversity of Chardonnay 232

wines by spontaneous fermentation and co-fermentation with Hanseniaspora vineae. Food Chem. 141, 2513–21. doi:10.1016/j.foodchem.2013.04.056 Morgan, S.C., Scholl, C.M., Benson, N.L., Stone, M.L., Durall, D.M., 2017. Sulfur dioxide addition at crush alters Saccharomyces cerevisiae strain composition in spontaneous fermentations at two Canadian wineries. Int. J. Food Microbiol. 244, 96–102. doi:10.1016/j.ijfoodmicro.2016.12.025 Morgan, S.C., Tantikachornkiat, M., Scholl, C.M., Benson, N.L., Cliff, M.A., Durall, D.M., 2019. The effect of sulfur dioxide addition at crush on the fungal and bacterial communities and the sensory attributes of Pinot gris wines. Int. J. Food Microbiol. 290, 1–14. doi:10.1016/j.ijfoodmicro.2018.09.020 Mortimer, R., Polsinelli, M., 1999. On the origins of wine yeast. Res. Microbiol. 150, 199–204. doi:10.1016/S0923-2508(99)80036-9 Moschetti, G., Corona, O., Gaglio, R., Squadrito, M., Parrinello, A., Settanni, L., Barone, E., Francesca, N., 2016. Use of fortified pied de cuve as an innovative method to start spontaneous alcoholic fermentation for red winemaking. Aust. J. Grape Wine Res. 22, 36–45. doi:10.1111/ajgw.12166 Nadai, C., Treu, L., Campanaro, S., Giacomini, A., Corich, V., 2016. Different mechanisms of resistance modulate sulfite tolerance in wine yeasts. Appl. Microbiol. Biotechnol. 100, 797–813. doi:10.1007/s00253-015-7169-x Nardi, T., Corich, V., Giacomini, A., Blondin, B., 2010. A sulphite-inducible form of the sulphite efflux gene SSU1 in a Saccharomyces cerevisiae wine yeast. Microbiology 156, 1686–1696. doi:10.1099/mic.0.036723-0 Nardi, T., Panero, L., Petrozziello, M., Guaita, M., Tsolakis, C., Cassino, C., Vagnoli, P., Bosso, A., 2019. Managing wine quality using Torulaspora delbrueckii and Oenococcus oeni starters in mixed fermentations of a red wine. Eur. Food Res. Technol. 245, 293–307. doi:10.1007/s00217-018-3161-x Naumov, G.I., Masneuf, I., Naumova, E.S., Aigle, M., Dubourdieu, D., 2000. Association of Saccharomyces bayanus var. uvarum with some French wines: Genetic analysis of yeast populations. Res. Microbiol. 151, 683–691. doi:10.1016/S0923-2508(00)90131-1 Naumov, G.I., Naumova, E.S., Antunovics, Z., Sipiczki, M., 2002. Saccharomyces bayanus var. uvarum in Tokaj wine-making of Slovakia and Hungary. Appl. Microbiol. Biotechnol. 59, 727–730. doi:10.1007/s00253-002-1077-6 Naumov, G.I., Nguyen, H.-V., Naumova, E.S., Michel, A., Aigle, M., Gaillardin, C., 2001. Genetic identification of Saccharomyces bayanus var. uvarum, a cider- fermenting yeast, International Journal of Food Microbiology. Nehme, N., Mathieu, F., Taillandier, P., 2010. Impact of the co-culture of Saccharomyces cerevisiae – Oenococcus oeni on malolactic fermentation and partial characterization of a yeast-derived inhibitory peptidic fraction. Food Microbiol. 27, 150–157. doi:10.1016/j.fm.2009.09.008 Neta, P., Huie, R.E., 1985. Free-radical chemistry of sulfite. Environ. Health 233

Perspect. 64, 209–217. doi:10.2307/3430011 Nguyen, H.V., Gaillardin, C., 2005. Evolutionary relationships between the former species Saccharomyces uvarum and the hybrids Saccharomyces bayanus and Saccharomyces pastorianus; reinstatement of Saccharomyces uvarum (Beijerinck) as a distinct species. FEMS Yeast Res. 5, 471–483. doi:10.1016/j.femsyr.2004.12.004 Nicolini, G., Moser, S., Román, T., Mazzi, E., Larcher, R., 2011. Effect of juice turbidity on fermentative volatile compounds in white wines. Vitis - J. Grapevine Res. 50, 131–135. Nisiotou, A.A., Nychas, G.-J.E., 2007. Yeast populations residing on healthy or Botrytis-infected grapes from a vineyard in Attica, Greece. Appl. Environ. Microbiol. 73, 2765–2768. doi:10.1128/AEM.01864-06 Nisiotou, A.A., Rantsiou, K., Iliopoulos, V., Cocolin, L., Nychas, G.-J.E., Va, D.P.R.A., 2011. Bacterial species associated with sound and Botrytis-infected grapes from a Greek vineyard. Int. J. Food Microbiol. 145, 432–436. doi:10.1016/j.ijfoodmicro.2011.01.017 Nissen, P., Nielsen, D., Arneborg, N., 2003. Viable Saccharomyces cerevisiae cells at high concentrations cause early growth arrest of non-Saccharomyces yeasts in mixed cultures by a cell-cell contact-mediated mechanism. Yeast 20, 331– 341. doi:10.1002/yea.965 Ocón, E., Gutiérrez, a. R., Garijo, P., Tenorio, C., López, I., López, R., Santamaría, P., 2010. Quantitative and qualitative analysis of non-Saccharomyces yeasts in spontaneous alcoholic fermentations. Eur. Food Res. Technol. 230, 885–891. doi:10.1007/s00217-010-1233-7 Ordish, G., 1987. The great wine blight. Pan Macmillan, London. Owen, J., 2011. Earliest known winery found in Armenian cave [WWW Document]. Natl. Geogr. News. URL https://news.nationalgeographic.com/news/2011/01/110111-oldest-wine-press- making-winery-armenia-science-ucla/ (accessed 5.1.19). Papalexandratou, Z., Lefeber, T., Bahrim, B., Lee, O.S., Daniel, H.-M., De Vuyst, L., 2013. Hanseniaspora opuntiae, Saccharomyces cerevisiae, Lactobacillus fermentum, and Acetobacter pasteurianus predominate during well-performed Malaysian cocoa bean box fermentations, underlining the importance of these microbial spe. Food Microbiol. 35, 73–85. doi:10.1016/J.FM.2013.02.015 Park, D.H., 2014. Effect of incubation temperature on variations in bacterial communities grown in fermenting meju and the nutritional quality of soy sauce. Food Sci. Biotechnol. 23, 1921–1928. doi:10.1007/s10068-014-0262-6 Park, H., Bakalinsky, A.T., 2004. Evidence for sulfite proton symport in Saccharomyces cerevisiae. J. Microbiol. Biotechnol. 14, 967–971. Park, H., Bakalinsky, A.T., 2000. SSU1 mediates sulphite efflux in Saccharomyces cerevisiae. Yeast 16, 881–888. 234

Park, H., Hwang, Y.-S., 2008. Genome-wide transcriptional responses to sulfite in Saccharomyces cerevisiae. J. Microbiol. 46, 542–548. doi:10.1007/s12275-008- 0053-y Park, H., Lopez, N.I., Bakalinsky, A.T., 1999a. Use of sulfite resistance in Saccharomyces cerevisiae as a dominant selectable marker. Curremt Genet. 36, 339–344. Park, H., Lopez, N.I., Bakalinsky, A.T., 1999b. Use of sulfite resistance in Saccharomyces cerevisiae as a dominant selectable marker. Curr. Genet. 36, 339–344. Pateraki, C., Paramithiotis, S., Doulgeraki, A.I., Kallithraka, S., Kotseridis, Y., Drosinos, E.H., 2014. Effect of sulfur dioxide addition in wild yeast population dynamics and polyphenolic composition during spontaneous red wine fermentation from Vitis vinifera cultivar . Eur. Food Res. Technol. 239, 1067–1075. doi:10.1007/s00217-014-2303-z Peakall, R., Smouse, P.E., 2012. GenAlEx 6.5: Genetic analysis in Excel. Population genetic software for teaching and research - an update. Bioinformatics 28, 2537–2539. doi:10.1093/bioinformatics/bts460 Peakall, R., Smouse, P.E., 2006. GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Mol. Ecol. Notes 6, 288–295. doi:10.1111/j.1471-8286.2005.01155.x Peinado, R.A., Mauricio, J.C., Moreno, J., 2006. Aromatic series in sherry wines with gluconic acid subjected to different biological aging conditions by Saccharomyces cerevisiae var. capensis. Food Chem. 94, 232–239. doi:10.1016/J.FOODCHEM.2004.11.010 Peinado, R.A., Moreno, J., Medina, M., Mauricio, J.C., 2004. Changes in volatile compounds and aromatic series in sherry wine with high gluconic acid levels subjected to aging by submerged yeast cultures. Biotechnol. Lett. 26, 757– 762. doi:10.1023/B:BILE.0000024102.58987.de Pérez-Martín, F., Seseña, S., Fernández-González, M., Arévalo, M., Palop, M.L., 2014. Microbial communities in air and wine of a winery at two consecutive vintages. Int. J. Food Microbiol. 190, 44–53. doi:10.1016/j.ijfoodmicro.2014.08.020 Pérez, M.A., Gallego, F.J., Martínez, I., Hidalgo, P., 2001. Detection, distribution and selection of microsatellites (SSRs) in the genome of the yeast Saccharomyces cerevisiae as molecular markers. Lett. Appl. Microbiol. 33, 461–466. doi:10.1046/j.1472-765X.2001.01032.x Peter, J.J., Watson, T.L., Walker, M.E., Gardner, J.M., Lang, T.A., Borneman, A., Forgan, A., Tran, T., Jiranek, V., 2018. Use of a wine yeast deletion collection reveals genes that influence fermentation performance under low-nitrogen conditions. FEMS Yeast Res. 18, foy009. doi:10.1093/femsyr/foy009 Petruzzi, L., Sinigaglia, M., Corbo, M.R., Campaniello, D., Speranza, B., Bevilacqua, A., 2014. Decontamination of ochratoxin A by yeasts: Possible approaches and 235

factors leading to toxin removal in wine. Appl. Microbiol. Biotechnol. 98, 6555– 6567. doi:10.1007/s00253-014-5814-4 Piao, H., Hawley, E., Kopf, S., DeScenzo, R., Sealock, S., Henick-Kling, T., Hess, M., 2015. Insights into the bacterial community and its temporal succession during the fermentation of wine grapes. Front. Microbiol. 6, 1–12. doi:10.3389/fmicb.2015.00809 Pinto, C., Pinho, D., Cardoso, R., Custódio, V., Fernandes, J., Sousa, S., Pinheiro, M., Egas, C., Gomes, A.C., 2015. Wine fermentation microbiome: a landscape from different Portuguese wine appellations. Front. Microbiol. 6, 1–13. doi:10.3389/fmicb.2015.00905 Piškur, J., Rozpedowska, E., Polakova, S., Merico, A., Compagno, C., 2006. How did Saccharomyces evolve to become a good brewer? Trends Genet. 22, 183– 186. doi:10.1016/j.tig.2006.02.002 Portillo, M. del C., Franquès, J., Araque, I., Reguant, C., Bordons, A., 2016. Bacterial diversity of and grape surface from different vineyards at Priorat wine region (Catalonia, Spain). Int. J. Food Microbiol. 219, 56–63. Prakitchaiwattana, C.J., Fleet, G.H., Heard, G.M., 2004. Application and evaluation of denaturing gradient gel electrophoresis to analyse the yeast ecology of wine grapes. FEMS Yeast Res. 4, 865–877. doi:10.1016/j.femsyr.2004.05.004 Prasongsuk, S., Lotrakul, P., Ali, I., Bankeeree, W., Punnapayak, H., 2018. The current status of Aureobasidium pullulans in biotechnology. Folia Microbiol. (Praha). 63, 129–140. doi:10.1007/s12223-017-0561-4 Price, M.N., Dehal, P.S., Arkin, A.P., 2009. FastTree: Computing large minimum evolution trees with profiles instead of a distance matrix. Mol. Biol. Evol. 26, 1641–1650. doi:10.1093/molbev/msp077 Puckette, M., 2018. The real differences: New World vs Old World wine [WWW Document]. Wine Folly. URL https://winefolly.com/review/new-world-vs-old- world-wine/ (accessed 5.1.19). Raposo, R., Ruiz-Moreno, M.J., Garde-Cerdán, T., Puertas, B., Moreno-Rojas, J.M., Zafrilla, P., Gonzalo-Diago, A., Guerrero, R.F., Cantos-Villar, E., 2016. Replacement of sulfur dioxide by hydroxytyrosol in white wine: Influence on both quality parameters and sensory. LWT - Food Sci. Technol. 65, 214–221. doi:10.1016/j.lwt.2015.08.005 Reguant, C., Carreté, R., Constantí, M., Bordons, A., 2005. Population dynamics of Oenococcus oeni strains in a new winery and the effect of SO2 and yeast strain. FEMS Microbiol. Lett. 246, 111–117. doi:10.1016/j.femsle.2005.03.045 Rementeria, A., Rodriguez, J.A., Cadaval, A., Amenabar, R., Muguruza, J.R., Hernando, F.L., Sevilla, M.J., 2003. Yeast associated with spontaneous fermentations of white wines from the “Txakoli de Bizkaia” region (Basque Country, North Spain). Int. J. Food Microbiol. 86, 201–207. doi:10.1016/S0168- 1605(03)00289-7 236

Reynolds, A.G., Taylor, G., De Savigny, C., 2013. Defining Niagara terroir by chemical and sensory analysis of Chardonnay wines from various soil textures and vine sizes. Am. J. Enol. Vitic. 64, 180–194. doi:10.5344/ajev.2013.12092 Ribereau-Gayon, P., Maujean, A., Dubourdieu, D., 2006. Handbook of Enology, 2nd ed. John Wiley and Sons Ltd., Chichester. Richards, K.D., Goddard, M.R., Gardner, R.C., 2009. A database of microsatellite genotypes for Saccharomyces cerevisiae. Antonie van Leeuwenhoek, Int. J. Gen. Mol. Microbiol. 96, 355–359. doi:10.1007/s10482-009-9346-3 Robinson, J., Harding, J. (Eds.), 2015. The Oxford Companion to Wine, 4th ed. Oxford University Press, Oxford. Rodríguez, M., Pérez-Través, L., Sangorrín, M., Barrio, E., Querol, A., Lopes, C., 2017. Saccharomyces uvarum is responsible for the traditional fermentation of apple Chicha in Patagonia. FEMS Yeast Res. 17, fow109. doi:10.1093/femsyr/fow109 Rodríguez, M.E., Pérez-Través, L., Sangorrín, M.P., Barrio, E., Lopes, C.A., 2014. Saccharomyces eubayanus and Saccharomyces uvarum associated with the fermentation of Araucaria araucana seeds in Patagonia. FEMS Yeast Res. 14, 948–965. doi:10.1111/1567-1364.12183 Romano, A., Perello, M.C., Lonvaud-Funel, A., Sicard, G., De Revel, G., 2009. Sensory and analytical re-evaluation of “Brett character.” Food Chem. 114, 15– 19. doi:10.1016/j.foodchem.2008.09.006 Romano, P., Fiore, C., Paraggio, M., Capece, A., 2003. Function of yeast species and strains in wine flavour. Int. J. Food Microbiol. 86, 169–180. doi:10.1016/S0168-1605(03)00290-3 Rossouw, D., Du Toit, M., Bauer, F.F., 2012. The impact of co-inoculation with Oenococcus oeni on the trancriptome of Saccharomyces cerevisiae and on the flavour-active metabolite profiles during fermentation in synthetic must. Food Microbiol. 29, 121–131. doi:10.1016/j.fm.2011.09.006 Sabate, J., Cano, J., Querol, A., Guillamon, J.M., 1998. Diversity of Saccharomyces strains in wine fermentations: analysis for two consecutive years. Lett. Appl. Microbiol. 26, 452–455. doi:10.1046/j.1472-765X.1998.00369.x Saberi, S., Cliff, M.A., van Vuuren, H.J.J., 2012. Impact of mixed S. cerevisiae strains on the production of volatiles and estimated sensory profiles of Chardonnay wines. Food Res. Int. 48, 725–735. doi:10.1016/j.foodres.2012.06.012 Sadoudi, M., Tourdot-Maréchal, R., Rousseaux, S., Steyer, D., Gallardo-Chacón, J.J., Ballester, J., Vichi, S., Guérin-Schneider, R., Caixach, J., Alexandre, H., 2012. Yeast-yeast interactions revealed by aromatic profile analysis of Sauvignon Blanc wine fermented by single or co-culture of non-Saccharomyces and Saccharomyces yeasts. Food Microbiol. 32, 243–253. doi:10.1016/j.fm.2012.06.006

237

Salo, P., 1970. Determining the odor thresholds for some compounds in alcoholic beverages. J. Food Sci. 35, 95–99. doi:10.1111/j.1365-2621.1970.tb12378.x Sampaio, J.P., Gonçalves, P., 2008. Natural populations of Saccharomyces kudriavzevii in Portugal are associated with oak bark and are sympatric with S. cerevisiae and S. paradoxus. Appl. Environ. Microbiol. 74, 2144–2152. doi:10.1128/AEM.02396-07 Santamaría, P., Garijo, P., López, R., Tenorio, C., Gutiérrez, A.R., 2005. Analysis of yeast population during spontaneous alcoholic fermentation: Effect of the age of the cellar and the practice of inoculation. Int. J. Food Microbiol. 103, 49–56. doi:10.1016/j.ijfoodmicro.2004.11.024 Santamaría, P., López, R., López, E., Garijo, P., Gutiérrez, A.R., 2008. Permanence of yeast inocula in the winery ecosystem and presence in spontaneous fermentations. Eur. Food Res. Technol. 227, 1563–1567. doi:10.1007/s00217- 008-0855-5 Santos, M.C., Nunes, C., Rocha, M.A.M., Rodrigues, A., Rocha, S.M., Saraiva, J.A., Coimbra, M.A., 2015. High pressure treatments accelerate changes in volatile composition of sulphur dioxide-free wine during bottle storage. Food Chem. 188, 406–414. doi:10.1016/j.foodchem.2015.05.002 Schlosser, J., Reynolds, A.G., King, M., Cliff, M., 2005. Canadian terroir: Sensory characterization of Chardonnay in the Niagara Peninsula. Food Res. Int. 38, 11–18. doi:10.1016/j.foodres.2004.07.003 Schmitt, M.J., Breinig, F., 2006. Yeast viral killer toxins: Lethality and self-protection. Nat. Rev. Microbiol. 4, 212–221. doi:10.1038/nrmicro1347 Schmitt, M.J., Breinig, F., 2002. The viral killer system in yeast: From molecular biology to application. FEMS Microbiol. Rev. 26, 257–276. doi:10.1016/S0168- 6445(02)00099-2 Scholl, C.M., Morgan, S.C., Stone, M.L., Tantikachornkiat, M., Neuner, M., Durall, D.M., 2016. Composition of Saccharomyces cerevisiae strains in spontaneous fermentations of Pinot Noir and Chardonnay. Aust. J. Grape Wine Res. 22, 384–390. doi:10.1111/ajgw.12221 Schreiner, J., 2009. The wineries of British Columbia. Whitecap Books, North Vancouver. Schrimpf, S.P., Weiss, M., Reiter, L., Ahrens, C.H., Jovanovic, M., Malmström, J., Brunner, E., Mohanty, S., Lercher, M.J., Hunziker, P.E., Aebersold, R., von Mering, C., Hengartner, M.O., 2009. Comparative functional analysis of the Caenorhabditis elegans and Drosophila melanogaster proteomes. PLoS Biol. 7, e1000048. doi:10.1371/journal.pbio.1000048 Schuller, D., Cardoso, F., Sousa, S., Gomes, P., Gomes, A.C., Santos, M. a S., Casal, M., 2012. Genetic diversity and population structure of Saccharomyces cerevisiae strains isolated from different grape varieties and winemaking regions. PLoS One 7. doi:10.1371/journal.pone.0032507

238

Serpaggi, V., Remize, F., Recorbet, G., Gaudot-Dumas, E., Sequeira-Le Grand, A., Alexandre, H., 2012. Characterization of the “viable but nonculturable” (VBNC) state in the wine spoilage yeast Brettanomyces. Food Microbiol. 30, 438–447. doi:10.1016/j.fm.2011.12.020 Shütz, M., Gafner, J., 1993. Analysis of yeast diversity during spontaneous and induced alcoholic fermentations. J. Appl. Bacteriol. 75, 551–558. doi:10.1111/j.1365-2672.1993.tb01594.x Sicard, D., Legras, J.-L., 2011. Bread, beer and wine: yeast domestication in the Saccharomyces sensu stricto complex. C. R. Biol. 334, 229–236. doi:10.1016/J.CRVI.2010.12.016 Siebert, T.E., Bramley, B., Solomon, M., 2009. Hydrogen sulfide: aroma detection threshold study in white and red wines. AWRI Tech. Rev. 183, 14–16. Siebert, T.E., Smyth, H.E., Capone, D.L., Neuwöhner, C., Pardon, K.H., Skouroumounis, G.K., Herderich, M.J., Sefton, M.A., Pollnitz, A.P., 2005. Stable isotope dilution analysis of wine fermentation products by HS-SPME-GC-MS. Anal. Bioanal. Chem. 381, 937–947. doi:10.1007/s00216-004-2992-4 Sipiczki, M., 2011. Diversity, variability and fast adaptive evolution of the wine yeast (Saccharomyces cerevisiae) genome - A review. Ann. Microbiol. 61, 85–93. doi:10.1007/s13213-010-0086-4 Sipiczki, M., Romano, P., Lipani, G., Miklos, I., Antunovics, Z., 2001. Analysis of yeasts derived from natural fermentation in a Tokaj winery. Antonie van Leeuwenhoek, Int. J. Gen. Mol. Microbiol. 79, 97–105. doi:10.1023/A:1010249408975 Smith, M.E., Bekker, M.Z., Smith, P.A., Wilkes, E.N., 2015. Sources of volatile sulfur compounds in wine. Aust. J. Grape Wine Res. 21, 705–712. doi:10.1111/ajgw.12193 Soleas, G.J., Yan, J., Goldberg, D.M., 2001. Assay of ochratoxin A in wine and beer by high-pressure liquid chromatography photodiode array and gas chromatography mass selective detection. J. Agric. Food Chem. 49, 2733– 2740. doi:10.1021/jf0100651 Stampanoni, C.R., 1993. Influence of acid and sugar content on sweetness, sourness and the flavour profile of beverages and sherbets. Food Qual. Prefer. 4, 169–176. doi:10.1016/0950-3293(93)90159-4 Stefanini, I., Albanese, D., Cavazza, A., Franciosi, E., De Filippo, C., Donati, C., Cavalieri, D., 2016a. Dynamic changes in microbiota and mycobiota during spontaneous “Vino Santo Trentino” fermentation. Microb. Biotechnol. 9, 195– 208. doi:10.1111/1751-7915.12337 Stefanini, I., Dapporto, L., Berná, L., Polsinelli, M., Turillazzi, S., Cavalieri, D., 2016b. Social wasps are a Saccharomyces mating nest. Proc. Natl. Acad. Sci. 113, 2247–2251. doi:10.1073/pnas.1516453113 Stefanini, I., Dapporto, L., Legras, J.-L., Calabretta, A., Di Paola, M., De Filippo, C., 239

Viola, R., Capretti, P., Polsinelli, M., Turillazzi, S., Cavalieri, D., 2012. Role of social wasps in Saccharomyces cerevisiae ecology and evolution. Proc. Natl. Acad. Sci. 109, 13398–13403. doi:10.1073/pnas.1208362109 Stratford, M., Morgan, P., Rose, A.H., 1987. Sulphur dioxide resistance in Saccharomyces cerevisiae and Saccharomycodes ludwigii. J. Gen. Microbiol. 133, 2173–2179. Stribny, J., Gamero, A., Pérez-Torrado, R., Querol, A., 2015. Saccharomyces kudriavzevii and Saccharomyces uvarum differ from Saccharomyces cerevisiae during the production of aroma-active higher alcohols and acetate esters using their amino acidic precursors. Int. J. Food Microbiol. 205, 41–46. doi:10.1016/J.IJFOODMICRO.2015.04.003 Sturm, J., Grossmann, M., Schnell, S., 2006. Influence of grape treatment on the wine yeast populations isolated from spontaneous fermentations. J. Appl. Microbiol. 101, 1241–1248. doi:10.1111/j.1365-2672.2006.03034.x Styger, G., Prior, B., Bauer, F.F., 2011. Wine flavor and aroma. J. Ind. Microbiol. Biotechnol. 38, 1145–1159. doi:10.1007/s10295-011-1018-4 Suárez-Lepe, J.A., Morata, A., 2012. New trends in yeast selection for winemaking. Trends Food Sci. Technol. 23, 39–50. doi:10.1016/j.tifs.2011.08.005 Suárez Valles, B., Pando Bedriñana, R., Fernández Tascón, N., Querol Simón, A., Rodríguez Madrera, R., 2007. Yeast species associated with the spontaneous fermentation of cider. Food Microbiol. 24, 25–31. doi:10.1016/j.fm.2006.04.001 Sun, J., Li, Q., Luo, S., Zhang, J., Huang, M., Chen, F., Zheng, F., Ab, X.S., Li, H., 2018. Characterization of key aroma compounds in Meilanchun sesame flavor style baijiu by application of aroma extract dilution analysis, quantitative measurements, aroma recombination, and omission/addition experiments. RSC Adv. 8, 23757. doi:10.1039/c8ra02727g Sun, S.Y., Gong, H.S., Zhao, K., Wang, X.L., Wang, X., Zhao, X.H., Yu, B., Wang, H.X., 2013. Co-inoculation of yeast and lactic acid bacteria to improve cherry wines sensory quality. Int. J. Food Sci. Technol. 48, 1783–1790. doi:10.1111/ijfs.12151 Sun, Y., Zhang, T., Lü, H., Yu, Z., Li, X., 2016. Effect of added sulphur dioxide levels on the fermentation characteristics of strawberry wine. J. Inst. Brew. 122, 446– 451. doi:10.1002/jib.342 Suzzi, G., Romano, P., 1982. Induced changes by SO2 on the population of Saccharomyces as agents of the natural fermentation of musts. Vini d’Italia 22, 138–145. Suzzi, G., Romano, P., Zambonelli, C., 1985. Saccharomyces strain selection in minimizing SO2 requirement during vinification. Am. J. Enol. Vitic. 36, 199–202. Swiegers, J.H., Bartowsky, E.J., Henschke, P. a., Pretorius, I.S., 2005. Yeast and bacterial modulation of wine aroma and flavour. Aust. J. Grape Wine Res. 11, 139–173. doi:10.1111/j.1755-0238.2005.tb00285.x 240

Takahashi, M., Ohta, T., Masaki, K., Mizuno, A., Goto-Yamamoto, N., 2014. Evaluation of microbial diversity in sulfite-added and sulfite-free wine by culture- dependent and -independent methods. J. Biosci. Bioeng. 117, 569–75. doi:10.1016/j.jbiosc.2013.10.012 Tantikachornkiat, M., Sakakibara, S., Neuner, M., Durall, D.M., 2016. The use of propidium monoazide in conjunction with qPCR and Illumina sequencing to identify and quantify live yeasts and bacteria. Int. J. Food Microbiol. 234, 53–59. doi:10.1016/j.ijfoodmicro.2016.06.031 Teeter, A., 2015. The first foreign grape in America has ties to communion [WWW Document]. Vine Pair. URL https://vinepair.com/wine-blog/the-first-foreign- grape-in-america-has-ties-to-communion/ (accessed 5.1.19). This, P., Lacombe, T., Thomas, M.R., 2006. Historical origins and genetic diversity of wine grapes. TRENDS Genet. 22, 511–519. doi:10.1016/j.tig.2006.07.008 Torija, M.J., Rozès, N., Poblet, M., Guillamón, J.M., Mas, a., 2001. Yeast population dynamics in spontaneous fermentations: Comparison between two different wine-producing areas over a period of three years. Antonie van Leeuwenhoek, Int. J. Gen. Mol. Microbiol. 79, 345–352. doi:10.1023/A:1012027718701 Tosi, E., Azzolini, M., Guzzo, F., Zapparoli, G., 2009. Evidence of different fermentation behaviours of two indigenous strains of Saccharomyces cerevisiae and Saccharomyces uvarum isolated from Amarone wine. J. Appl. Microbiol. 107, 210–218. doi:10.1111/j.1365-2672.2009.04196.x Ubeda Iranzo, J.F., González Magaña, F., González Viñas, M.A., 2000. Evaluation of the formation of volatiles and sensory characteristics in the industrial production of white wines using different commercial strains of the genus Saccharomyces. Food Control 11, 143–147. doi:10.1016/S0956- 7135(99)00086-9 Valentine, G.D.S., Walker, M.E., Gardner, J.M., Schmid, F., Jiranek, V., 2019. Brief temperature extremes during wine fermentation: effect on yeast viability and fermentation progress. Aust. J. Grape Wine Res. 25, 62–69. doi:10.1111/ajgw.12365 Vally, H., de Klerk, N., Thompson, P.J., 2000. Alcoholic : Important triggers for asthma. J. Allergy Clin. Immunol. 105, 462–467. doi:10.1067/mai.2000.104548 Vally, H., Misso, N.L.A., Madan, V., 2009. Clinical effects of sulphite additives. Clin. Exp. Allergy 39, 1643–1651. doi:10.1111/j.1365-2222.2009.03362.x Vally, H., Thompson, P.J., 2001. Role of sulfite additives in wine induced asthma: single dose and cumulative dose studies. Thorax 56, 763–769. doi:10.1136/thorax.56.10.763 Varela, C., Barker, A., Tran, T., Borneman, A., Curtin, C., 2017. Sensory profile and volatile aroma composition of reduced alcohol Merlot wines fermented with Metschnikowia pulcherrima and Saccharomyces uvarum. Int. J. Food Microbiol. 252, 1–9. doi:10.1016/j.ijfoodmicro.2017.04.002

241

Varela, C., Borneman, A.R., 2017. Yeasts found in vineyards and wineries. Yeast 34, 111–128. doi:10.1002/yea.3219 Vasserot, Y., Caillet, S., Maujean, A., 1997. Study of anthocyanin adsorption by yeast lees. Effect of some physiochemical parameters. Am. J. Enol. Vitic. 48, 433–437. Vendrame, M., Iacumin, L., Manzano, M., Comi, G., 2013. Use of propidium monoazide for the enumeration of viable Oenococcus oeni in must and wine by quantitative PCR. Food Microbiol. 35, 49–57. doi:10.1016/j.fm.2014.03.010 Viana, F., Gil, J. V., Genovés, S., Vallés, S., Manzanares, P., 2008. Rational selection of non-Saccharomyces wine yeasts for mixed starters based on formation and enological traits. Food Microbiol. 25, 778–785. doi:10.1016/j.fm.2008.04.015 Vigentini, I., Fabrizio, V., Faccincani, M., Picozzi, C., Comasio, A., Foschino, R., 2014. Dynamics of Saccharomyces cerevisiae populations in controlled and spontaneous fermentations for D.O.C.G. base wine production. Ann. Microbiol. 64, 639–651. doi:10.1007/s13213-013-0697-7 Vilanova, M., Sieiro, C., 2006. Contribution by Saccharomyces cerevisiae yeast to fermentative flavour compounds in wines from cv. Albariño. J. Ind. Microbiol. Biotechnol. 33, 929–933. doi:10.1007/s10295-006-0162-8 Vogel, C., Marcotte, E.M., 2012. Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nat. Rev. Genet. 13, 227–32. doi:10.1038/nrg3185 Wall, T.L., Peterson, C.M., Peterson, K.P., Johnson, M.L., Thomasson, H.R., Cole, M., Ehlers, C.L., 1997. Alcohol metabolism in Asian-American men with genetic polymorphisms of aldehyde dehydrogenase. Ann. Intern. Med. 127, 376. doi:10.7326/0003-4819-127-5-199709010-00007 Wang, C., Esteve-Zarzoso, B., Cocolin, L., Mas, A., Rantsiou, K., 2015a. Viable and culturable populations of Saccharomyces cerevisiae, Hanseniaspora uvarum and Starmerella bacillaris (synonym Candida zemplinina) during Barbera must fermentation. Food Res. Int. 78, 195–200. doi:10.1016/j.foodres.2015.10.014 Wang, C., Mas, A., Esteve-Zarzoso, B., 2016. The interaction between Saccharomyces cerevisiae and non-Saccharomyces yeast during alcoholic fermentation is species and strain specific. Front. Microbiol. 7, 1–11. doi:10.3389/fmicb.2016.00502 Wang, C., Mas, A., Esteve-Zarzoso, B., 2015b. Interaction between Hanseniaspora uvarum and Saccharomyces cerevisiae during alcoholic fermentation. Int. J. Food Microbiol. 206, 67–74. doi:10.1016/j.ijfoodmicro.2015.04.022 Weeks, C., 1969. Production of sulfur dioxide-binding compounds and of sulfur dioxide by two Saccharomyces yeasts. Am. J. Enol. Vitic. 32–39. Williams, A.A., Arnold, G.M., 1991. The influence of presentation factors on the sensory assessment of beverages. Food Qual. Prefer. 3, 101–107. 242

doi:10.1016/0950-3293(91)90029-E Yang, W.H., Purchase, E.C.R., 1985. Adverse reactions to sulfites. Can. Med. Assoc. J. 133, 865–868. Zara, G., Mannazzu, I., Del Caro, A., Budroni, M., Pinna, M.B., Murru, M., Farris, G.A., Zara, S., 2014. Wine quality improvement through the combined utilisation of yeast hulls and Candida zemplinina/Saccharomyces cerevisiae mixed starter cultures. Aust. J. Grape Wine Res. 20, 199–207. doi:10.1111/ajgw.12078 Zarraonaindia, I., Owens, S.M., Weisenhorn, P., West, K., Hampton-Marcell, J., Lax, S., Bokulich, N.A., Mills, D.A., Martin, G., Taghavi, S., van der Lelie, D., Gilbert, J.A., 2015. The soil microbiome influences grapevine-associated microbiota. MBio 6, e02527-14. doi:10.1128/mBio.02527-14.Editor Zhang, H., Richards, K.D., Wilson, S., Lee, S.A., Sheehan, H., Roncoroni, M., Gardner, R.C., 2015. Genetic characterization of strains of Saccharomyces uvarum from New Zealand wineries. Food Microbiol. 46, 92–99. doi:10.1016/j.fm.2014.07.016 Zhang, H.Y., Lee, S.A., Bradbury, J.E., Warren, R.N., Sheth, H., Hooks, D.O., Richards, K.D., Gardner, R.C., 2010. Yeasts isolated from New Zealand vineyards and wineries. Aust. J. Grape Wine Res. 16, 491–496. doi:10.1111/j.1755-0238.2010.00111.x Zoecklein, B.W., Fugelsang, K.C., Gump, B.H., Nury, F.S., 1995. Wine Analysis and Production. Chapman & Hall, New York. Zott, K., Miot-Sertier, C., Claisse, O., Lonvaud-Funel, A., Masneuf-Pomarede, I., 2008. Dynamics and diversity of non-Saccharomyces yeasts during the early stages in winemaking. Int. J. Food Microbiol. 125, 197–203. doi:10.1016/j.ijfoodmicro.2008.04.001

243

Appendix A

Table A.1. Microsatellite fingerprints of all S. cerevisiae strains identified at Winery 1 and Winery 2. Numbers represent the fragment lengths of alleles at eight loci.

Loci Strain C11 C3 C4 C8 YOR267c YLR177w YML091c YPL009c Fermol Arôme 214 216 115 115 247 256 140 153 288 288 122 122 283 283 301 304 Plus Fermol 194 202 121 121 238 238 143 143 280 280 122 122 302 304 301 301 Mediterranée Actiflore F15 218 218 124 124 253 253 153 153 324 324 122 122 280 280 291 291 Zymaflore FX10 222 222 121 121 253 253 140 140 324 324 122 122 274 274 301 301 Zymaflore RB2 218 218 127 127 253 253 140 140 348 348 105 105 324 324 304 304 Zymaflore RX60 216 216 115 115 247 256 140 146 312 312 122 122 298 298 276 304 Zymaflore VL1 214 216 115 115 247 256 140 153 288 288 122 122 283 298 301 304 Zymaflore VL3 214 216 115 121 247 256 140 146 280 280 105 122 295 298 301 304 Zymaflore X5 216 216 115 115 247 256 140 146 280 312 122 122 298 308 276 304 Zymaflore X16 194 216 115 121 253 256 140 153 288 344 122 122 283 304 301 301 Enoferm Syrah* 202 212 121 124 244 259 146 150 280 280 105 122 274 324 245 307 Lalvin BA11 208 210 121 121 259 259 156 156 288 288 105 105 283 283 267 267 Lalvin BM45 222 225 124 124 256 256 153 153 324 324 105 105 280 280 301 301 Lalvin CY3079 214 216 121 124 247 256 140 146 280 280 122 122 274 292 301 301 Lalvin DV10 191 216 115 115 304 304 146 146 306 312 122 122 268 317 276 291 Lalvin ICV- 187 187 121 121 250 250 150 150 280 280 122 122 274 274 304 304 D254** Lalvin ICV- D47*** 194 202 121 121 238 244 143 143 280 280 122 122 302 302 301 301 Lalvin K1-V1116 194 202 124 124 253 253 150 153 280 280 119 119 327 327 255 280 Lalvin RC212 187 220 121 121 250 256 146 150 312 358 122 122 317 317 304 307 Lalvin QA23 191 216 115 121 304 304 140 146 306 312 122 122 268 317 276 291 Premium Blanc 12 V 216 220 121 121 244 253 146 146 280 280 122 122 292 324 301 301 Premium 196 196 121 121 247 247 140 140 328 328 122 122 250 250 301 304 Supertuscan Red Star 216 239 115 121 247 247 140 153 344 344 122 122 271 271 301 310 Montrachet Vitilevure 3001 216 220 115 121 250 250 146 150 280 280 122 122 250 298 280 301 UN1 202 212 121 121 259 259 146 150 280 280 105 122 274 324 245 307 UN2 202 212 121 124 259 259 146 150 280 280 105 122 324 324 245 307 UN3 194 202 115 121 244 244 143 146 288 288 122 122 286 292 304 307 UN4 194 194 121 121 247 247 140 140 328 328 119 119 250 250 301 304 UN5 187 220 121 121 250 256 146 150 280 358 122 122 274 317 304 307 UN6 187 220 121 121 250 253 140 150 280 324 122 122 274 274 301 304 UN7 212 212 124 124 259 259 146 146 280 280 105 105 324 324 304 307 UN8 202 212 121 124 244 244 146 150 280 280 105 122 274 324 304 307 UN9 202 212 121 121 244 259 146 150 280 280 105 105 274 324 245 307 UN10 202 202 121 121 244 259 146 150 280 280 105 122 324 324 245 307 UN11 202 202 124 124 244 244 150 150 280 280 122 122 324 324 307 307 UN12 202 202 124 124 259 259 150 150 280 280 105 105 324 324 307 307 UN13 202 202 121 124 244 244 146 150 280 280 122 122 274 324 245 307 UN14 202 214 124 124 244 259 146 150 280 280 125 125 274 324 245 307 UN15 202 214 121 124 259 259 146 150 280 280 122 122 274 324 245 307 UN16 214 216 115 115 247 256 140 146 280 312 122 122 298 298 276 304 UN17 214 216 121 121 247 247 140 140 280 280 122 122 274 274 301 301 UN18 214 216 121 121 256 256 140 140 280 280 122 122 274 292 301 301 UN19 216 216 124 124 244 244 146 146 288 288 122 122 280 292 272 301 UN20 216 216 115 155 247 247 140 153 288 288 122 122 283 283 301 304 UN21 216 220 121 121 250 250 146 150 280 280 122 122 250 250 280 301 UN22 220 220 115 115 250 250 146 150 280 280 122 122 250 298 280 301 UN23 194 216 115 121 247 247 140 146 328 328 122 122 250 268 301 304 UN24 216 216 121 121 256 256 140 146 280 280 122 122 274 292 301 301 UN25 194 194 121 121 247 247 146 146 288 288 122 122 261 261 307 307 UN26 220 220 121 121 250 250 150 150 312 312 122 122 317 317 304 307 UN27 220 220 121 121 250 250 140 140 324 324 122 122 274 274 301 301 244

UN28 191 191 124 124 276 276 150 150 280 280 122 122 254 254 276 276 UN29 202 202 124 124 259 259 146 146 280 280 105 105 274 274 307 307 UN30 225 225 124 124 256 256 153 153 324 324 105 125 280 280 301 301 UN31 214 216 115 121 247 256 140 153 288 312 122 122 283 283 304 304 UN32 206 218 124 124 256 256 146 146 280 280 122 122 304 304 301 301 UN33 202 202 121 124 244 259 146 150 280 280 105 122 274 324 307 307 UN34 202 202 124 124 256 256 146 146 280 280 122 122 324 324 304 304 UN35 187 220 121 121 250 256 146 146 358 358 122 122 317 317 304 307 UN36 187 187 121 121 247 250 150 150 280 280 122 122 274 274 301 304 UN37 220 220 121 121 250 256 146 150 312 312 122 122 317 317 304 307 UN38 202 202 124 124 244 259 150 150 280 280 105 122 274 324 245 307 UN39 212 212 121 124 244 259 146 150 280 280 105 122 274 274 245 245 UN40 225 225 121 121 253 253 150 150 328 328 122 122 261 261 276 276 UN41 191 191 124 124 276 276 150 150 280 280 125 125 254 254 276 276 UN42 187 220 115 121 250 256 140 153 288 312 122 122 283 283 304 307 *Same genetic fingerprint as Lalvin® Rhône 2056 **Same genetic fingerprint as Fermol® Premier Cru ***Same genetic fingerprint as Lalvin® BRL97

245

Table A.2. Relative abundance ± SEM of all S. cerevisiae strains identified at three stages of spontaneous fermentations to which different levels of SO2 were added at crush at Winery 1. Samples were collected at the early (E), mid (M), and late (L) stages of fermentation. Relative abundance and standard error (SEM) values were obtained from 120 S. cerevisiae isolates identified at each stage (40 isolates from each replicate fermentation).

0 mg/L 20 mg/L 40 mg/L E M L E M L E M L 5.8 3.3 1.7 6.7 12.5 5.8 0.83 1.7 2.5 ± ± ± ± ± ± ± ± ± AEB® Fermol Arôme Plus 1.7 0.83 0.83 1.7 5.0 1.7 0.83 0.83 0.0 0.83 ± AEB® Fermol Mediterranée ------0.83 - 0.83 3.3 0.83 ± ± ± Laffort® Actiflore® F15 - - - 0.83 - - 2.2 - 0.83 1.7 1.7 2.5 2.5 1.7 2.5 ± ± ± ± ± ± Laffort® Zymaflore® FX10 0.83 0.83 1.4 1.4 0.83 1.4 - - - 0.83 1.7 ± ± Laffort® Zymaflore® RX60 - - - 0.83 1.7 - - - - 0.83 0.83 1.7 1.7 1.7 ± ± ± ± ± Laffort® Zymaflore® VL1 - 0.83 0.83 0.83 1.7 - - 1.7 - 1.7 1.7 1.7 ± ± 0.83 ± Laffort® Zymaflore® VL3 - - - - 1.7 0.83 ± 0.83 - 0.83 6.7 1.7 1.7 1.7 0.83 0.83 ± ± ± ± ± ± Laffort® Zymaflore® X5 2.2 0.83 0.83 - 0.83 0.83 0.83 - - 3.3 4.2 2.5 17.5 18.3 23.3 8.3 11.7 15.0 ± ± ± ± ± ± ± ± ± Lallemand Enoferm® Syrah 2.2 1.7 1.4 2.9 4.2 9.8 3.0 2.2 7.2 0.83 ± Lallemand Lalvin® BA11 ------0.83 0.83 0.83 Lallemand Lalvin® BM45 ------± ± - 246

0.83 0.83 10.8 12.5 10.8 5.0 2.5 3.3 6.7 4.2 5.8 ± ± ± ± ± ± ± ± ± Lallemand Lalvin® CY3079 1.7 8.0 3.0 3.8 1.4 0.83 0.83 3.0 0.83 37.5 26.7 40.8 19.2 22.5 16.7 22.5 31.7 23.3 ± ± ± ± ± ± ± ± ± Lallemand Lalvin® DV10 6.3 6.0 0.83 5.5 3.8 7.9 9.5 6.8 4.4 2.5 3.3 4.2 1.7 1.7 1.7 5.8 ± ± ± ± ± ± ± Lallemand Lalvin® ICV-D254 1.4 2.2 1.7 - 0.83 - 0.83 0.83 2.2 0.83 0.83 3.3 4.2 11.7 10.8 14.2 ± ± ± ± ± ± ± Lallemand Lalvin® ICV-D47 - 0.83 - 0.83 0.83 0.83 3.3 3.0 2.2 1.7 2.5 2.5 17.5 12.5 15.8 20.8 20.8 15.0 ± ± ± ± ± ± ± ± ± Lallemand Lalvin® K1-V1116 0.83 1.4 0.0 9.5 7.2 4.4 6.8 5.8 4.3 3.3 13.3 7.5 2.5 2.5 3.3 13.3 5.0 4.2 ± ± ± ± ± ± ± ± ± Lallemand Lalvin® RC212 0.83 7.4 2.5 2.5 2.5 0.83 3.0 2.9 1.7 0.83 ± Lallemand Lalvin® QA23 ------0.83 - - 0.83 0.83 ± ± Premium® Blanc 12 V - - - - 0.83 - - - 0.83 12.5 13.3 6.7 11.7 9.2 10.0 4.2 ± ± ± ± ± ± 4.2 ± ± Premium® Super Tuscan 2.9 3.0 3.0 3.0 3.0 4.3 - 1.7 3.0 10.8 13.3 16.7 7.5 4.2 9.2 2.5 4.2 3.3 ± ± ± ± ± ± ± ± ± Vitilevure® 3001 1.7 1.7 2.2 2.9 1.7 0.83 2.5 2.2 2.2 0.83 ± UN1 - - - - 0.83 - - - - 0.83 ± UN2 - 0.83 ------0.83 0.83 1.7 0.83 UN3 ± - ± ± - - ± - - 247

0.83 0.83 0.83 0.83 0.83 ± UN4 - - - - - 0.83 - - - 0.83 ± UN5 - - - 0.83 - - - - - 0.83 ± UN6 ------0.83 - - 0.83 ± UN7 - - - - 0.83 - - - - 0.83 0.83 ± ± UN8 - - - 0.83 - - 0.83 - - 0.83 ± UN9 ------0.83 - - 0.83 ± UN10 - - - - - 0.83 - - - 0.83 0.83 ± ± UN11 - - - - - 0.83 - 0.83 - 0.83 ± UN12 - - - 0.83 - - - - - 0.83 ± UN13 0.83 ------0.83 ± UN14 ------0.83 0.83 ± UN15 ------0.83 0.83 0.83 0.83 UN16 - - - ± - ± ± - - 248

0.83 0.83 0.83 0.83 ± UN17 - 0.83 ------0.83 ± UN18 - - - 0.83 - - - - - 0.83 ± UN19 - 0.83 ------0.83 ± UN20 0.83 ------0.83 ± UN21 0.83 ------0.83 ± UN22 - - 0.83 ------0.83 ± UN23 ------0.83 - - 0.83 ± UN24 ------0.83

249

Table A.3. Relative abundance ± SEM of all S. cerevisiae strains identified at three stages of spontaneous fermentations to which different levels of SO2 were added at crush at Winery 2. Samples were collected at the early (E), mid (M), and late (L) stages of fermentation. Relative abundance and standard error (SEM) values were obtained from 120 S. cerevisiae isolates identified at each stage (40 isolates from each replicate fermentation).

20 mg/ L 40 mg/L E M L E M L AEB® Fermol Arôme Plus 1.7 ± 0.83 5.0 ± 3.8 4.2 ± 1.7 3.3 ± 2.2 0.83 ± 0.83 2.5 ± 0.0 Laffort® Zymaflore® FX10 2.5 ± 2.5 0.83 ± 0.83 - - - 0.83 ± 0.83 Laffort® Zymaflore® RB2 3.3 ± 0.83 0.83 ± 0.83 3.3 ± 0.83 9.2 ± 2.2 5.8 ± 0.83 4.2 ± 2.2 Laffort® Zymaflore® VL1 0.83 ± 0.83 - - - - - Laffort® Zymaflore® X5 ------Laffort® Zymaflore® X16 ------Lallemand Enoferm® Syrah 11.7 ± 0.83 13.3 ± 3.6 19.2 ± 0.83 22.5 ± 5.0 21.7 ± 5.8 35.0 ± 8.0 Lallemand Lalvin® CY3079 4.2 ± 3.0 6.7 ± 3.0 10.0 ± 1.4 5.8 ± 0.83 5.8 ± 3.6 4.2 ± 1.7 Lallemand Lalvin® ICV-D254 0.83 ± 0.83 - - - 0.83 ± 0.83 0.83 ± 0.83 Lallemand Lalvin® K1-V1116 - 0.83 ± 0.83 - 0.83 ± 0.83 - - Lallemand Lalvin® RC212 49.2 ± 8.3 46.7 ± 5.8 40.0 ± 5.2 49.2 ± 3.3 58.3 ± 3.0 40.8 ± 3.0 Red Star® Montrachet - 0.83 ± 0.83 - - - - UN1 0.83 ± 0.83 - - - - - UN2 - - - - 0.83 ± 0.83 - UN25 18.3 ± 4.6 18.3 ± 7.4 19.2 ± 5.5 1.7 ± 0.83 2.5 ± 1.4 3.3 ± 2.2 UN26 0.83 ± 0.83 - - - - - UN27 1.7 ± 1.7 - - - - - UN28 2.5 ± 2.5 5.0 ± 2.9 1.7 ± 1.7 1.7 ± 0.83 1.7 ± 0.83 1.7 ± 0.83 UN29 - 0.83 ± 0.83 1.7 ± 0.83 0.83 ± 0.83 - 0.83 ± 0.83 UN30 - 0.83 ± 0.83 - - - 1.7 ± 1.7 UN31 - - - 0.83 ± 0.83 - - UN32 - - - - - 0.83 ± 0.83 UN33 - - - - - 0.83 ± 0.83 250

UN34 - - - - - 0.83 ± 0.83 UN35 0.83 ± 0.83 - 0.83 ± 0.83 - - 0.83 ± 0.83 UN36 - - - 0.83 ± 0.83 - - UN37 - - - 0.83 ± 0.83 - - UN38 - - - 0.83 ± 0.83 - - UN39 - - - 0.83 ± 0.83 0.83 ± 0.83 0.83 ± 0.83 UN40 - - - - 0.83 ± 0.83 - UN41 0.83 ± 0.83 - - - - - UN42 - - - 0.83 ± 0.83 - -

251

Table A.4. Commercial S. cerevisiae strains used for inoculated fermentations at W1 in 2014 and previous vintages. Strains 2009† 2010 2011 2012 2013 2014 AEB® Fermol Arôme Plus x - - - - - AEB® Fermol Chardonnay x - - - - - AEB® Fermol Mediterranée x - - - - - AEB® Fermol Sauvignon x - - - - - Anchor® Exotics SPH - x x x x x Anchor® Vin 7 x - - - - - Anchor® Vin 13 - x x x x x Anchor® NT 116 - x x - - - Laffort® Zymaflore® F15 - - x x x x Laffort® Zymaflore® FX10 - x x x x x Laffort® Zymaflore® RB2 - x x x x x Laffort® Zymaflore® RX60 - x x x x x Laffort® Zymaflore® VL1 - x x x x x Laffort® Zymaflore® VL3 - x x - - - Laffort® Zymaflore® X5 x - - x - - Laffort® Zymaflore® X16 x - - - - - Lallemand Enoferm® AMH - x x - - - Lallemand Enoferm® Syrah/Lallemand - x - x x x Lalvin® Rhône 2056‡ Lallemand Lalvin® BA11 - x x x x x Lallemand Lalvin® BM45 x - - - - - Lallemand Lalvin® CLOS - - - x - - Lallemand Lalvin® CY3079 - x x - - - Lallemand Lalvin® DV10 - - - x x x Lallemand Lalvin® EC-1118 x - - - - - Lallemand Lalvin® ICV-D254/ AEB® Fermol x x x - - - Premier Cru‡ Lallemand Lalvin® ICV-D47/Lallemand x x x x x x Lalvin® BRL 97‡ Lallemand Lalvin® K1-V1116 x - - - - - Lallemand Lalvin® QA23 - x x - - - Lallemand Lalvin® RC212 - x x x x x Lallemand Lalvin® W15 - x x x x x Lallemand uvaferm® 228 x - - - - - Lallemand uvaferm® 43 x - - - - - Premium® Blanc 12 V - - - - x x Premium® Supertuscan - - - - x x Premium® - - - - x x Red Star® Montrachet - x x x x x Vitilevure® 3001 - - x x - - Vitilevure® 58W3 - - - x x x Vitilevure® Elixir - - - x x x † Strains in this column were used in 2009 or earlier ‡ Strains that have been found to display the same multilocus genotype

252

Table A.5. Commercial S. cerevisiae strains used for inoculated fermentations at W2 in 2014. Strains 2014 Anchor® Vin13 x SIHA® Active Yeast 7 x Lallemand Lalvin® RC212 x Lallemand Lalvin® CY3079 x Laffort® Zymaflore® RB2 x Lallemand Lalvin® QA23 x AEB® Fermol Arôme Plus x

253

Appendix B

Table B.1. Composition of sensory standards provided to the panel to aid with their wine assessments. Food standards were prepared in 50 mL neutral Sola-Nero white wine (Arterra Wines, Mississauga, ON Canada) unless otherwise indicated.

Sensory standard Composition of standard Source Citrus -one 12 cm piece of fresh -fresh lemon lemon rind, steeped overnight Pome -15 mL each of apple juice -freshly-squeezed Granny Smith and pear juice apple and pear juice -222 cm cube of canned -No Name® sliced pears in fruit pear juice from concentrate -1.0 g ascorbic acid per 50 -Natural Factors® C mL pressed juice crystals Tropical fruit -2.5 mL mango juice -SunRype® 100% no sugar added mango juice -5 mL pineapple juice -SunRype® 100% no sugar added pineapple juice Spice -2 cm stick cinnamon -Spice Cargo cinnamon sticks -1 whole clove -McCormick© GourmetTM cloves -1/32 tsp ground allspice -McCormick© Club House -Steeped for two hours ground allspice Vanilla -79 µL pure vanilla extract -President’s Choice® pure vanilla extract Toasty/smoky -1.4 g wood chips in a small -IDL Consulting Inc. French Oak cup with a lid (no liquid) large grade heavy toast (wood chips) Wood -1.3 g wood chips in a small -World Cooperage© American cup with a lid (no liquid) Oak large chips medium toast (wood chips) Acidity -0.13 g tartaric acid dissolved -DMDTM General Storage L (+) in 100 mL water tartaric acid Astringency -0.1 g alum dissolved in 150 -Valmo Laboratory Inc. alum mL water USP ( alum)

254

Table B.2. Partial statistical output of “panelperf” function evaluating the sensory panel’s performance in its ability to discriminate between wines of different treatments (SensoMineR package in RStudio (version 3.4.4). A panel of 10 wine experts was used for the sensory evaluation of wines to which three different levels of sulfur dioxide (SO2) had been added at crush (n = 3 per treatment). Panelists evaluated each of the nine wines in duplicate tasting sessions. P-values are reported from three-factor ANOVA [product (wines), panelist, sensory session (tasting replicate)], for each of the 18 sensory attributes. An asterisk indicates significance at α = 0.05.

Sensory Attribute Product Panelist Session Product: Product: Panelist: Panelist Session Session Citrus aroma 0.034* <0.001* 0.17 0.51 0.46 0.0018* Pome aroma 0.28 <0.001* 0.71 0.29 0.48 0.0013* Tropical fruit aroma 0.16 <0.001* 0.96 0.83 0.31 0.0090* Spice aroma 0.17 <0.001* 0.63 0.25 0.27 <0.001* Vanilla aroma 0.7 <0.001* 0.07 0.83 0.16 0.0077* Toasty/smoky aroma 0.15 <0.001* 0.78 0.48 0.72 <0.001* Wood aroma 0.47 <0.001* 0.24 0.90 0.70 0.0018* Citrus flavour 0.72 <0.001* 0.79 0.33 0.29 <0.001* Pome flavour 0.029* <0.001* 0.53 0.94 0.59 <0.001* Tropical fruit flavour 0.34 <0.001* 0.82 0.72 0.15 <0.001* Spice flavour 0.87 <0.001* 0.39 0.011* 0.10 <0.001* Vanilla flavour 0.22 <0.001* 0.010* 0.36 0.019* 0.089 Toasty/smoky flavour 0.87 <0.001* 0.38 0.0029* 0.34 <0.001* Wood flavour 0.39 <0.001* 0.011* 0.93 0.29 0.0043* Acidity 0.44 <0.001* 0.018* 0.29 0.65 0.0097* Astringency 0.77 <0.001* 0.81 0.89 0.49 <0.001* Body 0.44 <0.001* 0.63 0.46 0.078 <0.001* Length of Aftertaste 0.15 <0.001* 0.44 0.44 0.23 <0.001*

255

Table B3. Fungal community composition (based on 6,000 sequences per sample) of wines fermented with different levels of sulfur dioxide (SO2) added at crush (0, 20, or 40 mg/L SO2), reported ± the standard error of the mean (SEM). Three replicate fermentations were conducted for each treatment for a total of nine barrels. Samples were taken from each barrel at four stages of fermentation (Cold-settling, Early, Mid, Late).

0 mg/L SO2 20 mg/L SO2 40 mg/L SO2

Fungi CS E M L CS E M L CS E M L 13 4310 4030 4258 9 3983 5285 4735 26 4970 5654 5478 Saccharomyces ± ± ± ± ± ± ± ± ± ± ± ± cerevisiae 1 192 129 231 2 275 80 201 17 294 95 98 3552 30 62 95 3496 158 79 198 3355 219 108 163 Aureobasidium ± ± ± ± ± ± ± ± ± ± ± ± pullulans 143 9 3 11 54 6 16 32 99 34 8 20 192 1461 1583 1482 23 1490 392 860 44 327 100 180 Hanseniaspora ± ± ± ± ± ± ± ± ± ± ± ± 22 179 105 226 12 289 46 157 9 140 60 65 1338 7 13 43 1765 55 6 67 1700 138 23 61 Cladosporium ± ± ± ± ± ± ± ± ± ± ± ± 122 4 8 5 69 8 3 19 121 17 8 4 - 176 288 62 - 247 233 90 2 93 67 38 Saccharomyces ± ± ± ± ± ± ± ± ± ± uvarum 19 60 11 41 73 26 2 15 17 5 299 7 15 34 181 36 - 17 232 110 28 42 Sclerotiniaceae ± ± ± ± ± ± ± ± ± ± ± (family) 15 4 8 5 3 18 11 30 10 1 7 224 9 10 9 153 21 1 12 258 12 6 8 Aspergillus ± ± ± ± ± ± ± ± ± ± ± ± flavus 86 8 6 3 58 11 1 9 88 9 3 4 130 - - 13 126 4 3 10 131 - 4 17 Curvibasidium ± ± ± ± ± ± ± ± ± cygneicollum 5 3 18 4 3 6 11 3 4 118 - - 3 101 6 - 9 88 35 - 10 Mycosphaerella ± ± ± ± ± ± ± ± tassiana 28 3 13 3 5 10 5 4 58 - - - 105 - - 1 88 - - 2 Penicillium ± ± ± ± ± 6 23 1 23 1 ------96 9 1 Hanseniaspora ± ± ± vineae 77 7 1 20 - - - 9 - - 1 15 - - - Udeniomyces ± ± ± ± puniceus 2 1 1 5 16 - - 1 11 - - - 12 - - - Penicillium ± ± ± ± 256 spinulosum 3 1 6 8 21 - - - 9 - - - 9 - - - Rhodosporidiobolus ± ± ± colostri 11 9 9 9 - - - 12 - - - 13 - - - Rhizopogon ± ± ± abietis 3 5 6 9 ------10 - 1 - Mucor ± ± ± circinelloides 5 8 1 2 ------16 - - - Alternaria ± ± 2 16

257

Table B4. Bacterial community composition (based on 10,000 sequences per sample) of wines fermented with different levels of sulfur dioxide (SO2) added at crush (0, 20, or 40 mg/L SO2). Three replicate fermentations were conducted for each treatment for a total of nine barrels. Samples were taken from each barrel at four stages of fermentation (Cold Settling, Early, Mid, Late).

0 mg/L SO2 20 mg/L SO2 40 mg/L SO2

Fungi CS E M L CS E M L CS E M L 7159 6626 7088 6589 6637 7019 6790 7146 7125 6621 7321 6841 Enterococcus ± ± ± ± ± ± ± ± ± ± ± ± 391 484 399 412 530 302 446 282 582 447 371 503 1251 1675 1220 1513 1677 1362 1624 1164 1260 1698 1165 1511 Bacillaceae ± ± ± ± ± ± ± ± ± ± ± ± (family) 378 419 358 339 427 373 420 333 421 433 362 388 504 536 564 584 502 548 464 529 505 552 515 526 Paenibacillus ± ± ± ± ± ± ± ± ± ± ± ± 17 28 42 14 35 19 8 39 51 29 10 34 413 439 459 569 458 437 423 473 436 453 421 454 Alkaliphilus ± ± ± ± ± ± ± ± ± ± ± ± 1 35 16 36 18 29 37 27 46 5 7 22 240 258 249 270 281 244 261 235 252 271 201 228 Lactococcus ± ± ± ± ± ± ± ± ± ± ± ± 27 36 21 32 44 22 41 24 41 13 13 34 69 71 76 87 69 70 62 71 68 53 61 76 Pseudomonas ± ± ± ± ± ± ± ± ± ± ± ± 7 11 11 8 7 16 11 10 11 27 15 12 61 58 61 75 62 88 59 60 74 79 60 57 Clostridiaceae ± ± ± ± ± ± ± ± ± ± ± ± (family) 7 8 10 8 10 23 14 8 16 17 10 2 51 30 50 28 28 57 29 52 51 32 46 25 Brevibacillus ± ± ± ± ± ± ± ± ± ± ± ± 24 21 19 18 20 29 17 23 24 21 19 20 36 28 34 39 27 38 36 48 32 38 41 28 Virgibacillus ± ± ± ± ± ± ± ± ± ± ± ± 5 6 8 8 10 11 8 10 4 6 13 5 31 39 41 49 47 27 36 23 30 39 22 41 Arthrobacter ± ± ± ± ± ± ± ± ± ± ± ± 3 12 17 15 7 1 7 4 11 5 2 12 36 52 37 44 39 22 30 38 20 31 24 40 Clostridiales ± ± ± ± ± ± ± ± ± ± ± ± (order) 4 9 3 6 14 15 21 8 9 16 12 8 40 36 23 26 34 17 32 32 37 14 22 28 Bacillus ± ± ± ± ± ± ± ± ± ± ± ± 5 2 12 3 2 17 8 6 6 7 11 4 23 33 29 27 25 18 38 29 21 33 32 31 Lysinibacillus 258

± ± ± ± ± ± ± ± ± ± ± ± 4 5 3 5 3 9 6 6 3 6 3 10 24 24 29 30 25 26 24 30 23 28 18 19 Psychrobacter ± ± ± ± ± ± ± ± ± ± ± ± 6 6 7 4 3 6 2 9 1 5 2 5 13 18 - 27 7 22 22 17 11 - 26 34 Enterococcaceae ± ± ± ± ± ± ± ± ± ± (family) 13 18 27 7 22 22 17 11 26 18 13 33 - - 32 - 24 21 15 33 - 24 Paenibacillaceae ± ± ± ± ± ± ± ± (family) 13 18 13 24 21 9 33 24 6 11 13 12 12 - 13 12 8 8 7 11 Leuconostoc ± ± ± ± ± ± ± ± ± ± ± 4 1 3 1 4 4 4 4 4 4 2 7 21 4 8 6 2 8 - 5 8 4 5 Enterobacteriaceae ± ± ± ± ± ± ± ± ± ± ± (family) 7 8 4 5 6 2 2 4 8 4 3 5 - 6 4 5 17 5 1 - 6 6 Bacillaceae ± ± ± ± - ± ± ± ± ± (family) 5 6 4 5 7 5 1 6 3 3 5 - 2 7 - 4 - 3 2 - 4 Sporosarcina ± ± ± ± ± ± ± ± 3 5 2 4 4 3 2 4 3 2 6 2 3 - 3 3 3 1 2 2 Clostridium ± ± ± ± ± ± ± ± ± ± ± 2 2 3 1 1 2 1 1 1 2 2 4 0 3 4 6 - - 4 4 - - 3 Carnobacterium ± ± ± ± ± ± ± ± 4 0 3 4 3 4 4 3 1 2 2 - 3 1 2 1 2 2 3 1 Myroides ± ± ± ± ± ± ± ± ± ± ± 1 2 1 1 1 1 1 1 1 2 1 5 ± 1 2 - 3 - 1 3 3 - 1 1 Streptococcus 1 ± ± ± ± ± ± ± ± 1 1 2 1 1 1 1 1 1 - - 10 2 - - 2 4 1 - - Planococcaceae ± ± ± ± ± ± (family) 1 5 2 2 4 1 1 2 5 - 4 - - - 2 - - 3 Bacillales ± ± ± ± ± ± (order) 1 2 5 2 1 1 1 1 - - - 1 - - 2 2 1 - Enhydrobacter ± ± ± ± ± ± 1 1 1 2 2 1

259

Appendix C

Table C1. List of yeast-derived volatile secondary metabolites as analyzed by HS- SPME GC-MS.

Metabolite Rt Quantitation Confirmation Scan Dwell time (min) ion mass ion mass window (msec) (min) Ethyl acetate 3.78 43.00 29.0 0.12 40 Ethyl propanoate 4.45 43.00 57.0 0.66 20 Ethyl 2-methylpropanoate 4.54 43.00 57.0 0.85 20 Ethyl butanoate 5.51 31.00 88.0 0.72 25 Ethyl 2-methylbutanoate 5.84 102.00 87.0 0.68 20 Ethyl 3-methylbutanoate 6.04 88.00 130.0 0.48 30 2-methyl propanol 6.55 41.00 43.0 0.88 30 2-methylbutyl acetate 7.00 87.00 72.0 0.60 30 3-methylbutyl acetate 7.10 87.00 72.0 0.80 30 2-methyl butanol 9.10 42.00 56.0 1.80 100 3-methyl butanol 9.20 42.00 56.0 2.00 100 Ethyl hexanoate 9.65 88.00 99.0 1.10 100 Hexyl acetate 10.63 69.00 73.0 1.44 100 Hexanol 12.56 45.00 56.0 2.72 30 Ethyl octanoate 14.76 88.00 101.0 2.52 100 1-octanol (internal standard) 17.65 56.00 41.0 4.30 100 Ethyl decanoate 19.68 101.00 157.0 2.36 100 2-methylbutanoic acid 20.28 61.00 73.0 1.26 100 3-methylbutanoic acid 20.28 61.00 73.0 1.26 100 Methionol 21.38 106.00 61.0 1.56 100 2-phenylethyl acetate 23.60 65.00 104.0 3.20 100 Hexanoic acid 24.12 101.00 60.0 1.14 100 Ethyl dodecanoate 24.20 101.00 60.0 1.30 100 Benzyl alcohol 24.82 79.00 51.0 0.74 80 2-phenylethanol 25.56 91.00 92.0 1.42 100 Octanoic acid 28.44 60.00 73.0 2.88 100 Decanoic acid 32.33 73.00 139.0 4.66 100

260

Table C2. Multilocus genotypes of Strain 1 and Strain 2. Numbers represent the fragments lengths of allele 1 (A1) and allele 2 (A2) at eight microsatellite loci.

C11 C3 C4 C8 YOR267c YLR177w YML091c YPL009c A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 Strain 191 216 115 121 312 312 140 146 306 312 125 125 268 317 422 434 1 Strain 194 202 121 121 244 250 143 143 280 280 125 125 302 302 446 446 2

261

Appendix D

Table D1. Composition of sensory standards provided to the panel to aid with their wine evaluations. Food standards were prepared in 100 mL neutral Sola white wine (Arterra Wines, Mississauga, ON, Canada) unless otherwise indicated. Each panelist was given 5 mL of each aroma standard (and 10 mL of the acidity flavour standard) in a small labeled plastic cup sealed with a lid to contain the volatile compounds. Panelists were encouraged to refer back to these standards throughout the evaluation as needed.

Sensory Composition of standard Source standard Citrus -Two 12 cm pieces of -Fresh lemon fresh lemon rind, steeped overnight Pome fruit -40 mL apple juice -Fresh-squeezed Granny Smith apple and -40 mL pear juice Williams pear juice -1.0 g ascorbic acid per 50 -No Name® sliced pears in fruit juice from mL pressed juice concentrate -Natural Factors® Vitamin C crystals Tropical fruit -3 mL mango juice -Oasis Classics® mango juice -10 mL pineapple juice -No Name® 100% unsweetened pineapple juice -1 cm-thick banana slice -Banana Melon -0.5 µL honeydew extract -International Flavors and Fragrances (IFF) -0.5 µL cantaloupe extract Canada Ltd. Honeydew Melon -0.5 µL watermelon extract -IFF Canada Ltd. Cantaloupe -Tastemaker Natural + Artificial Watermelon Flavor Floral -0.16 µL Jasmine extract -Firmenich© Artificial Jasmine Earthy -20 g each crushed white -White mushroom, unwashed and crimini mushroom (let -Crimini mushroom, unwashed sit 1 h and then strain) Sulfur -1 tsp each chopped garlic -Colossal garlic and onion (let sit 20 min, -Red onion strain, dilute 10×) Sweetness -3.0 g Glucose dissolved in -Sigma© Life Science D-(+)-Glucose, minimum 100 mL water 99.5% Acidity -0.13 g Tartaric Acid -EMDTM General Storage L-(+)-Tartaric Acid dissolved in 100 mL water

262

Table D2. Fungal percent relative abundance (± SEM) at four stages of alcoholic fermentation (Cold-settling (C), Early (E), Mid (M), Late (L)), as well as the pied de cuve inoculum (P). Samples were taken from stainless-steel barrel- fermented Chardonnay to which different inoculation and SO2 treatments had been applied (n = 3 per treatment). Relative abundance was calculated from 20,000 sequences per sample. Sequences were identified to the species level unless otherwise indicated.

0 mg/L Uninoculated 40 mg/L Uninoculated 0 mg/L Pied de cuve 40 mg/L Pied de cuve

Fungi C E M L C E M L P C E M L C E M L Saccharomyces 1.1 84 73 86 0.30 94 79 83 0.10 1.3 89 88 82 0.10 84 79 71 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± uvarum 0.3 6 5 4 0.06 3 0.5 4 0.05 0.5 3 3 8 0.03 6 2 10 Aureobasidium 48 3.4 4.5 0.52 73 1.0 3.4 2.2 67 52 4.7 3.5 3.8 73 3.3 3.7 3.3 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± pullulans 6 1 0.5 0.3 0.8 0.3 0.5 0.3 2 5 1 4 1 2 0.5 0.5 1 Saccharomyces 0.79 11 22 12 0.17 4.8 17 15 0.21 0.60 5.1 7.5 13 0.12 12 16 25 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± cerevisiae 0.2 6 5 4 0.04 3 0.3 4 0.2 0.2 4 2 6 0.02 5 1 9 25 0.19 0.19 0.46 3.4 0.12 0.092 0.15 5.7 21 0.19 0.11 0.21 5.3 0.18 0.083 0.14 Aspergillus flavus ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 11 0.1 0.08 0.2 2 0.1 0.07 0.1 4 9 0.1 0.05 0.08 4 0.1 0.06 0.1 Hanseniaspora 11 0.20 0.79 0.24 8.1 0.77 0.65 0.33 4.5 9.2 0.34 0.52 0.59 8.1 0.26 0.34 0.21 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± spp. 5 0.2 0.3 0.2 2 0.4 0.3 0.03 4 2 0.05 0.09 0.3 3 0.03 0.06 0.1 0.37 0.14 0.062 0.042 5.15 0.033 0.032 0.047 2.2 5.1 0.077 0.14 0.23 4.4 0.017 0.047 0.028 Cladosporium sp. ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 0.2 0.07 0.04 0.04 1 0.02 0.03 0.03 0.8 0.8 0.02 0.02 0.03 0.5 0.01 0.02 0.02 Mycosphaerella 0.66 0.10 0.075 0.020 1.26 0.012 0.078 0.047 9.3 0.60 0.11 0.15 1.2 0.065 0.020 0.073 ± ± ± ± ± ± ± ± ± ± ± - ± ± ± ± ± tassiana 0.07 0.05 0.02 0.02 0.2 0.01 0.04 0.03 4 0.09 0.04 0.03 0.2 0.03 0.02 0.04 0.50 0.070 0.025 0.89 0.018 0.0083 0.012 4.3 0.14 0.060 0.035 0.062 1.1 0.040 0.03 0.04 Alternaria sp. ± ± - ± ± ± ± ± ± ± ± ± ± ± ± ± ± 0.02 0.02 0.01 0.04 0.009 0.008 0.01 2 0.1 0.02 0.02 0.009 0.2 0.005 0.02 0.008 Capnodiales 0.87 0.15 0.088 0.037 1.0 0.10 0.028 1.2 1.0 0.043 0.093 0.11 0.90 0.033 0.080 ± ± ± ± ± - ± ± ± ± ± ± ± ± ± ± - (order) 0.06 0.03 0.05 0.03 0.1 0.06 0.02 0.5 0.2 0.03 0.05 0.07 0.2 0.03 0.04 0.006 0.005 0.96 0.0050 0.83 1.2 0.81 7 0 Aspergillus niger ± ± - - ± - - - - ± - - ± - - ± ± 0.05 0.005 0.07 0.3 0.02 0.006 0.005 0.006 0.012 0.010 0.11 3.2 0.055 0.095 0.078 0.13 0.082 0.010 Kazachstania 7 ± - ± - ± - - - ± ± ± ± ± ± ± - ± aerobia 0.01 0.006 0.02 3 0.03 0.07 0.04 0.1 0.04 0.01 0.006 Sclerotiniaceae 0.91 0.69 0.80 0.80 ± - - - ± - - - - ± - - - ± - - - (family) 0.2 0.09 0.1 0.1 Guehomyces 0.80 0.32 0.34 0.76 0.30 ± - - - ± - - - ± ± - - - ± - - - pullulans 0.1 0.08 0.2 0.2 0.1

263

0.005 0.88 0.50 0.050 0.54 0.44 Curvibasidium 0 ± - - - ± - - - ± ± - - ± - - - ± cygneicollum 0.1 0.03 0.05 0.07 0.2 0.005 0.001 0.008 0.69 0.44 0.56 0.31 Candida 7 3 ± - - - ± - - - ± - - - ± - - ± ± santamariae 0.03 0.1 0.09 0.05 0.001 0.008 0.006 0.45 0.58 0.55 0.34 Wickerhamomyc 7 ± - - - ± - - - - ± - - - ± - - ± es anomalus 0.04 0.2 0.1 0.03 0.006 0.006 0.27 0.0050 0.51 0.38 0.027 0.49 Epicoccum 7 ± - - ± ± - - - ± - - - ± ± - - ± nigrum 0.2 0.005 0.1 0.1 0.01 0.06 0.006 0.30 0.25 0.65 0.2 Penicillium sp. ± - - - ± - - - - ± - - - ± - - - 0.1 0.04 0.2 0.005 0.006 0.001 0.006 0.43 0.0050 0.085 0.80 0.022 0.03 Torulaspora 7 7 7 ± - ± ± - - - ± - ± ± - - - ± ± ± delbrueckii 0.07 0.005 0.003 0.2 0.01 0.02 0.006 0.001 0.006 0.30 0.16 0.012 0.40 0.015 0.12 0.012 0.013 0.016 Candida sp. ± - - - ± - - ± - ± - - ± ± ± ± ± 0.06 0.1 0.006 0.07 0.01 0.06 0.01 0.01 0.01 Trichosporon 0.45 0.37 0.14 ± ------± - - - ± - - - asahii 0.2 0.2 0.1 0.18 0.17 0.003 0.33 0.18 Coprinellus ellisii ± - - - ± - - - 3 ± ± - - - ± - - - 0.05 0.03 0.003 0.1 0.07 0.003 0.005 0.005 0.083 0.098 0.048 0.010 0.18 0.010 Hanseniaspora 3 0 0 ± - - - ± - - - - ± ± ± ± - ± ± ± osmophila 0.04 0.05 0.02 0.01 0.08 0.01 0.003 0.005 0.005 Torulaspora 0.15 0.073 0.21 0.0050 ± - - - ± - - - - ± - - - ± - - - pretoriensis 0.02 0.05 0.08 0.005 Penicillium 0.13 0.14 0.055 0.070 ± - - - ± - - - - ± - - - ± - - - spinulosum 0.07 0.08 0.05 0.04 Pleosporaceae 0.068 0.11 0.14 - - - - ± - - - ± - - - - ± - - - (family) 0.06 0.1 0.07 0.008 0.18 0.058 0.035 0.055 Phaeosphaeria 3 ± - - - ± - - - ± - - - ± - - - ± microscopica 0.1 0.03 0.03 0.05 0.008 Candida 0.11 0.015 0.040 0.17 ± - - - ± - - - ± ± ------zeylanoides 0.07 0.01 0.04 0.04 Kluyveromyces 0.062 0.053 0.065 0.088 0.020 ± - - - ± - - - ± ± - - - ± - - - lactis 0.009 0.02 0.06 0.04 0.02

264

Rhotodorula 0.063 0.20 0.018 ± ------± - - - ± - - - graminis 0.06 0.1 0.01 Rhodosporidiobol 0.017 0.070 0.090 0.057 ± - - - ± - - - - ± - - - ± - - - us colostri 0.01 0.04 0.09 0.04 Oidiodendron 0.15 0.010 0.030 0.020 ± - - - ± - - - ± - - - - ± - - - maius 0.1 0.01 0.03 0.02 Mucor 0.027 0.11 0.082 ± ------± ± ------circinelloides 0.09 0.08 0.7 0.16 0.010 0.043 Geastrum triplex ± - - - ± - - - - ± ------0.09 0.01 0.03 0.006 0.057 0.087 0.062 Tulostoma 7 - - - ± - - - - ± - - - ± - - - ± fimbriatum 0.04 0.03 0.03 0.006 Nakazawaea 0.030 0.055 0.047 0.052 ± - - - ± - - - - ± - - - ± - - - holstii 0.03 0.03 0.04 0.04 Cystofilobasidiu 0.025 0.057 0.10 ± - - - ± - - - - ± ------m capitatum 0.02 0.05 0.1 Filobasidium 0.18 ------± ------wieringae 0.1 Pleosporales 0.16 ------± ------(order) 0.08 Eremothecium 0.012 0.075 0.035 0.18 ± - - - ± - - - - ± - - - ± - - - coryli 0.01 0.03 0.03 0.01 Neoascochyta 0.080 ------± ------desmazieri 0.08 0.003 0.015 0.048 0.027 3 - - - - ± - - - ± - - - ± - - - Erysiphe necator ± 0.01 0.04 0.01 0.003

265

Table D3. Microsatellite identities of the representative multilocus genotypes (MLGs) of indigenous Saccharomyces uvarum strains isolated from stainless steel barrel-fermented Chardonnay at a commercial winery in British Columbia, Canada.

L1 L2 L3 L4 L7 L8 L9 NB1 NB4 NB8 NB9

MLG A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 A1 A2 MLG.001 163 163 306 306 224 224 306 306 259 259 229 229 283 283 206 206 350 350 439 439 88 88

MLG.005 165 165 290 290 228 228 318 318 269 269 208 208 277 277 212 212 335 335 448 448 122 122

MLG.006 163 163 294 294 224 224 318 318 259 259 220 220 277 277 212 212 335 335 439 439 88 88

MLG.011 167 167 290 290 224 224 318 318 269 269 220 220 226 226 206 206 350 350 439 439 114 114

MLG.017 163 167 294 306 228 228 318 318 259 259 214 220 274 281 206 206 350 350 439 439 114 122

MLG.018 167 167 294 294 218 224 306 306 259 259 208 220 289 289 209 209 344 344 439 460 114 114

MLG.022 167 167 306 306 218 218 318 318 259 259 208 208 289 289 209 209 344 344 448 448 114 114

MLG.023 167 167 290 306 228 228 300 300 259 259 208 208 289 289 212 212 335 335 439 439 114 114

MLG.030 167 167 306 306 218 218 318 318 269 269 208 208 292 292 209 209 335 335 448 448 114 114

MLG.031 167 167 306 306 218 218 300 318 259 269 208 220 296 296 206 209 344 350 412 448 114 114

MLG.034 167 167 294 294 228 228 318 318 269 269 214 214 271 271 206 206 350 350 448 448 114 114

MLG.035 167 167 306 306 218 218 300 300 259 259 208 208 286 286 206 206 350 350 448 448 114 114

MLG.038 167 167 294 306 214 214 300 318 259 267 247 247 187 286 209 209 335 335 412 448 114 114

MLG.042 167 167 294 294 228 228 306 306 269 269 208 208 187 187 209 209 335 335 448 448 114 114

MLG.044 167 167 306 306 228 228 318 318 269 269 220 220 187 187 212 212 335 335 439 439 114 114

MLG.046 167 167 290 290 224 224 306 306 259 259 208 208 187 187 212 212 350 350 448 448 114 114

MLG.051 167 167 306 306 218 218 300 300 269 269 214 214 223 223 212 212 335 335 412 412 122 122

MLG.055 167 167 294 294 218 218 318 318 259 259 208 208 187 187 212 212 335 335 448 448 114 114

MLG.056 167 167 306 306 218 218 300 300 259 259 220 220 277 277 212 212 335 335 412 412 114 114

MLG.063 167 167 306 306 218 218 318 318 269 269 220 220 223 223 209 209 344 344 448 448 114 114

MLG.064 167 167 303 303 218 218 306 306 259 259 220 220 223 223 209 209 335 335 448 448 114 114

MLG.068 167 167 294 294 218 218 300 300 259 259 214 214 223 223 206 206 350 350 412 412 114 114

MLG.071 167 167 306 306 218 228 318 318 259 269 208 208 187 292 209 209 335 344 448 448 114 114

MLG.078 163 167 294 306 214 214 300 318 259 259 220 220 187 277 206 209 335 350 439 448 114 114

266

MLG.081 163 167 303 306 218 218 300 306 259 269 208 220 223 292 209 209 335 335 439 448 114 114

MLG.082 163 165 303 306 228 228 306 318 259 259 208 208 187 223 209 209 335 344 448 448 114 114

MLG.084 163 163 306 306 218 218 300 300 259 259 220 220 281 281 206 206 350 350 439 439 114 114

MLG.086 163 163 294 294 224 224 306 318 259 259 211 220 187 187 206 206 335 350 439 448 114 114

MLG.087 163 163 294 294 228 228 318 318 269 269 229 229 187 187 206 206 350 350 439 439 114 114

MLG.094 163 163 290 290 218 218 306 306 269 269 229 229 283 283 206 206 350 350 448 448 114 114

MLG.097 163 163 303 306 224 228 300 306 269 269 220 220 226 226 209 209 335 335 412 439 114 114

MLG.104 163 163 294 294 214 214 300 300 269 269 214 214 277 277 206 206 350 350 439 439 114 114

MLG.108 165 165 306 306 218 218 306 306 269 269 220 220 289 289 209 209 335 335 412 412 114 114

MLG.110 163 163 306 306 228 228 300 300 259 259 220 220 292 292 209 209 344 344 439 439 114 114

MLG.115 167 167 294 294 214 214 318 318 259 259 214 214 187 187 209 209 335 335 448 448 114 114

MLG.116 163 163 303 303 218 218 300 300 269 269 208 208 187 187 209 209 335 335 448 448 114 114

MLG.117 163 163 290 290 218 218 318 318 269 269 220 220 187 187 212 212 335 335 448 448 114 114

MLG.124 167 167 294 294 218 218 318 318 259 259 208 208 187 187 206 206 350 350 448 448 114 114

MLG.128 163 163 294 294 218 218 318 318 259 259 220 220 281 281 212 212 335 335 439 439 114 114

MLG.133 163 163 306 306 218 228 300 318 259 259 208 208 277 289 206 209 344 350 439 448 114 114

MLG.135a 163 163 306 306 228 228 300 300 259 259 220 220 277 277 206 206 350 350 439 439 114 114

MLG.136 163 163 294 294 228 228 300 300 259 259 220 220 277 277 212 212 335 335 439 439 114 114

MLG.139 163 163 303 306 228 228 300 306 259 269 208 220 223 223 206 209 335 350 439 448 114 114

MLG.142 165 165 294 294 228 228 318 318 259 259 229 229 277 281 206 206 350 350 439 439 88 88

MLG.147 163 163 294 294 218 228 306 306 269 269 220 220 277 277 206 206 350 350 412 412 114 114

MLG.150 163 163 306 306 228 228 300 300 269 269 220 220 277 277 206 206 350 350 439 439 114 114

MLG.151 163 167 294 306 228 228 300 318 259 259 208 220 277 277 206 206 350 350 439 448 114 122

MLG.160 163 163 294 294 228 228 339 339 259 259 208 208 274 274 206 206 350 350 448 448 114 114

MLG.163 163 163 294 303 218 228 306 306 259 269 208 208 223 281 206 209 335 350 439 448 114 114

MLG.172 163 163 306 306 228 228 306 306 269 269 220 220 187 187 212 212 335 335 439 439 114 114

MLG.179b 163 163 303 303 228 228 306 306 269 269 208 208 223 223 209 209 335 335 448 448 114 114

MLG.193 163 163 306 306 228 228 300 300 269 269 208 208 223 223 206 206 350 350 439 439 114 114

MLG.196 163 163 303 306 228 228 306 306 259 269 208 208 223 223 206 209 344 350 448 455 114 114

MLG.197 163 165 294 306 218 228 300 300 259 267 208 220 223 277 206 206 350 350 412 448 114 114

MLG.211 165 165 290 290 224 224 300 300 269 269 220 220 274 274 212 212 335 335 439 439 114 114

267

MLG.214 165 165 303 303 224 224 306 306 269 269 208 208 226 226 209 209 335 335 412 412 114 114

MLG.217 165 165 306 306 228 228 300 300 269 269 208 208 289 289 212 212 350 350 448 448 114 114

MLG.219 165 165 306 306 218 218 300 300 259 259 220 220 292 292 212 212 335 335 439 439 114 114

MLG.220 165 165 306 306 228 228 300 300 269 269 208 208 277 277 206 206 350 350 439 439 114 114

MLG.221 163 163 294 294 228 228 300 300 259 259 220 220 277 277 206 206 350 350 439 439 114 114

MLG.222 163 165 294 306 228 228 300 306 259 269 214 220 187 187 206 209 335 350 412 439 114 114

MLG.226 165 165 306 306 228 228 318 318 269 269 208 208 187 187 209 209 344 344 448 448 114 114

MLG.227 163 163 306 306 228 228 300 300 269 269 214 214 277 277 206 206 350 350 448 448 114 114

MLG.228 165 165 306 306 228 228 300 300 259 259 220 220 277 277 206 206 350 350 448 448 114 114

MLG.233 165 165 306 306 218 228 300 300 259 269 208 220 277 289 212 212 335 335 412 439 114 114

MLG.245 165 165 303 306 224 228 300 306 267 269 208 208 223 223 206 209 335 350 412 448 114 114

MLG.253 165 165 306 306 224 224 306 306 267 267 220 220 223 223 206 206 350 350 460 460 114 114

MLG.260 165 165 306 306 224 224 300 300 267 267 220 220 223 223 206 206 350 350 412 412 114 114

MLG.267 165 165 294 294 224 224 300 300 269 269 220 220 223 223 206 206 350 350 439 439 114 114

MLG.269 163 163 306 306 228 228 339 339 269 269 208 208 223 223 206 206 350 350 439 439 114 114

MLG.271 167 167 294 294 228 228 300 300 269 269 214 214 223 271 206 206 350 350 439 439 122 122

MLG.274 163 167 294 306 228 228 300 300 259 259 208 220 271 281 206 206 350 350 439 439 114 122

MLG.275 167 167 294 306 218 228 318 318 269 269 214 220 271 292 206 209 344 350 448 448 122 122

MLG.279 167 167 294 294 214 218 300 300 259 269 208 208 187 187 209 209 335 350 412 448 114 122

MLG.283 167 167 294 294 218 228 300 300 269 269 208 208 292 292 212 212 335 350 412 448 122 122

MLG.285 167 167 294 294 214 228 318 318 259 269 214 214 271 277 206 206 350 350 439 448 114 122

MLG.288 167 167 294 294 224 224 306 306 259 259 214 214 223 223 212 212 335 335 412 412 114 114

MLG.290 163 167 294 294 218 228 300 318 259 269 208 220 241 274 206 212 335 350 439 448 114 122

MLG.295 163 167 306 306 228 228 300 318 259 269 220 220 187 289 206 212 335 350 412 439 114 122

MLG.308 163 167 294 303 228 228 306 318 269 269 208 208 223 277 209 209 335 350 412 448 114 122

MLG.312 163 167 306 306 228 228 300 300 269 269 220 220 223 281 206 206 350 350 412 439 114 122

MLG.314 165 165 306 306 228 228 300 339 259 259 208 220 223 277 206 206 350 350 439 439 114 114

MLG.315 163 163 290 306 218 228 318 318 259 269 208 220 296 296 206 206 350 350 439 439 114 122

MLG.322 163 165 294 306 228 228 300 339 259 269 208 220 271 277 206 206 350 350 439 439 114 122

MLG.323 163 165 294 306 218 218 300 318 259 269 208 220 277 292 212 212 335 350 412 448 88 122

MLG.332 163 165 294 306 228 228 318 339 259 259 208 214 187 223 206 212 335 350 439 448 114 122

268

MLG.333 165 165 294 306 218 228 300 318 267 267 208 208 277 296 209 209 335 350 412 448 122 122

MLG.345 165 165 294 294 228 228 318 318 269 269 214 214 277 277 206 206 350 350 439 439 122 122

MLG.346 165 165 306 306 214 224 318 318 259 259 208 208 187 286 206 206 335 350 439 448 114 122

MLG.347 163 165 294 306 218 228 306 318 259 269 208 208 187 274 206 209 344 350 412 448 114 122

MLG.352 165 165 306 306 218 218 300 318 259 269 208 208 289 292 209 209 344 350 448 448 114 122

MLG.357 163 163 294 294 228 228 300 300 269 269 220 220 223 274 206 206 350 350 448 448 122 122

MLG.359 165 165 294 306 218 228 300 300 259 267 220 220 223 277 206 212 335 350 412 439 114 122

MLG.363 167 167 306 306 218 228 300 300 269 269 208 208 277 277 212 212 335 335 412 412 122 122

MLG.373 167 167 294 294 228 228 318 318 259 259 214 214 281 281 212 212 335 335 439 439 122 122

MLG.378 165 165 294 294 228 228 300 300 259 259 208 214 271 289 206 206 350 350 439 448 122 122

MLG.384 167 167 306 306 228 228 300 300 259 259 220 220 289 289 212 212 335 335 412 412 122 122

MLG.390 150 165 294 306 218 228 300 318 269 269 208 208 292 292 212 212 335 350 412 448 122 122

MLG.397 150 150 294 294 228 228 300 300 269 269 208 208 292 292 212 212 335 335 448 448 122 122

MLG.406 167 167 294 294 228 228 300 300 259 259 208 208 271 271 209 209 344 344 448 448 122 122

MLG.415 167 167 294 294 228 228 300 300 269 269 208 208 271 271 206 206 350 350 439 439 122 122

MLG.424 167 167 294 294 214 214 300 300 269 269 208 208 187 187 209 209 335 335 412 412 122 122

MLG.427 167 167 306 306 218 218 300 300 259 259 220 220 187 187 206 206 350 350 439 439 122 122

MLG.432 167 167 306 306 224 224 300 318 269 269 211 211 277 292 212 212 350 350 412 448 122 122

MLG.433 167 167 294 294 218 228 300 318 269 269 208 208 277 292 212 212 350 350 412 448 122 122

MLG.444 165 165 294 306 218 228 300 300 269 269 208 208 277 277 212 212 335 350 412 412 122 122

MLG.449 167 167 294 294 228 228 300 300 259 259 208 208 277 277 206 206 350 350 448 448 122 122

MLG.451 167 167 294 294 214 218 318 318 269 269 208 208 187 277 212 212 350 350 412 448 122 122

MLG.454 167 167 294 294 218 218 318 318 269 269 208 208 277 277 212 212 335 335 412 412 122 122

MLG.465 167 167 306 306 228 228 300 300 269 269 208 208 277 277 212 212 350 350 412 412 122 122

MLG.496 167 167 294 294 228 228 318 318 269 269 208 208 226 226 206 206 350 350 412 412 122 122

MLG.506 163 163 294 294 224 224 318 318 259 259 208 208 281 281 209 209 335 335 448 448 122 122

MLG.512 163 163 306 306 228 228 306 306 267 267 220 220 177 177 206 206 350 350 412 412 122 122

MLG.515 165 165 294 294 228 228 300 300 269 269 208 208 226 277 206 206 350 350 439 439 122 122

MLG.518 165 165 294 294 228 228 318 318 259 259 214 214 277 277 206 206 350 350 439 439 122 122

MLG.524 163 163 294 294 224 224 300 300 267 267 220 220 223 223 212 212 335 335 439 439 122 122

MLG.525 167 167 306 306 228 228 339 339 259 259 220 220 223 223 206 206 350 350 412 412 122 122

269

MLG.540 163 165 306 306 228 228 300 300 259 269 208 220 277 292 206 212 335 350 412 439 122 122

MLG.541 165 165 294 306 228 228 300 339 259 269 208 208 223 271 206 206 350 350 439 448 122 122

MLG.543 165 165 306 306 228 228 318 318 269 269 229 229 237 237 206 206 350 350 460 460 122 122

MLG.544 165 165 294 294 224 224 306 306 259 259 229 229 177 177 212 212 335 335 439 439 122 122

MLG.545 165 165 294 294 214 214 318 318 269 269 208 208 187 187 206 206 350 350 448 448 122 122

MLG.567 165 165 294 294 218 218 300 318 269 269 208 208 292 292 212 212 335 335 412 448 122 122

MLG.574 165 165 294 294 228 228 300 300 269 269 220 220 289 289 209 209 344 344 439 439 122 122

MLG.575 165 165 306 306 224 224 318 339 259 259 208 208 223 286 206 206 335 335 439 439 122 122

MLG.595 165 165 294 294 218 218 318 318 269 269 208 208 292 292 212 212 335 335 448 448 122 122

MLG.598 167 167 306 306 218 218 318 318 269 269 208 208 292 292 212 212 335 335 412 412 122 122

MLG.609 165 165 294 306 218 228 300 318 269 269 208 208 281 296 212 212 350 350 412 455 122 122

MLG.620 165 165 294 294 224 228 318 318 259 259 208 214 271 281 206 206 350 350 439 460 122 122

MLG.622 165 165 294 294 218 218 300 318 269 269 208 214 271 292 206 212 335 350 439 448 122 122

MLG.637 165 165 306 306 228 228 306 306 269 269 214 214 187 187 212 212 335 335 412 412 122 122

MLG.639 165 165 306 306 218 218 318 318 259 259 208 208 289 289 206 206 350 350 448 448 122 122

MLG.659 167 167 294 294 228 228 318 318 269 269 214 214 271 271 206 206 350 350 448 448 122 122

MLG.677 165 165 294 294 228 228 300 318 269 269 208 208 292 292 212 212 350 350 412 412 122 122

MLG.679c 165 165 294 294 218 228 300 318 269 269 208 208 277 292 212 212 350 350 412 448 122 122

MLG.682d 165 165 294 306 218 228 300 318 269 269 208 208 277 277 212 212 335 350 412 448 122 122

MLG.693 165 165 306 306 228 228 318 318 269 269 208 208 292 292 212 212 335 335 448 448 122 122

MLG.694 165 165 294 294 228 228 300 300 269 269 208 208 292 292 212 212 335 335 412 412 122 122

MLG.698 167 167 294 306 218 228 318 318 269 269 208 208 277 292 212 212 335 350 412 448 98 122

MLG.699 165 165 294 294 228 228 318 318 269 269 208 208 277 277 206 212 350 350 448 448 122 122

MLG.701 165 165 294 306 218 228 300 318 259 269 208 214 271 277 206 212 335 350 448 448 122 122

MLG.706 165 165 303 303 228 228 300 300 269 269 208 208 274 274 212 212 350 350 412 412 122 122

MLG.713 167 167 306 306 218 218 318 318 269 269 208 208 277 277 212 212 350 350 412 412 122 122

MLG.731 163 163 306 306 224 224 300 300 259 259 247 247 289 289 212 212 335 335 448 448 122 122

MLG.743 165 165 306 306 224 224 318 318 259 259 220 220 277 277 206 206 350 350 439 439 122 122

MLG.756 167 167 306 306 228 228 300 300 269 269 220 220 223 223 206 206 350 350 412 412 122 122

MLG.761 165 165 306 306 228 228 318 318 259 259 208 208 223 223 206 206 350 350 439 439 122 122

MLG.762 165 165 294 306 214 214 318 339 259 269 208 208 187 223 206 209 335 350 412 439 122 122

270

MLG.768 163 163 290 290 224 224 306 306 269 269 208 208 223 223 209 209 335 335 412 412 122 122

MLG.777 165 165 306 306 228 228 339 339 269 269 208 208 223 223 206 206 350 350 439 439 122 122 aMLG135 is referred to as “S. uvarum strain 1” in Figure 4. bMLG179 is referred to as “S. uvarum strain 3” in Figure 4. cMLG679 is referred to as “S. uvarum strain 4” in Figure 4. dMLG682 is referred to as “S. uvarum strain 2” in Figure 4.

271

Table D4. Bacterial percent (%) relative abundance (± SEM) at four stages of alcoholic fermentation (Cold-settling (C), Early (E), Mid (M), Late (L)), as well as the pied de cuve inoculum (P). Samples were taken from stainless-steel barrel- fermented Chardonnay to which different inoculation and SO2 treatments had been applied (n = 3 per treatment). Relative abundance was calculated from 8,500 sequences per sample. Sequences were identified to the genus level unless otherwise indicated.

0 mg/L Uninoculated 40 mg/L Uninoculated 0 mg/L Pied de cuve 40 mg/L Pied de cuve

Fungi C E M L C E M L P C E M L C E M L 69 59 59 72 67 62 64 70 69 70 63 64 64 69 67 66 71 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Enterococcus 4 8 6 4 4 5 4 3 4 4 7 5 1 4 5 4 5 12 12 12 12 15 19 19 15 12 12 12 14 11 15 18 18 15 Bacillaceae ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± (family) 4 2 2 4 4 5 4 3 3 3 2 2 4 4 5 4 4 5.8 5.6 5.8 5.7 5.6 7.4 5.9 5.7 5.4 5.6 6.0 6.4 4.8 5.2 6.3 6.6 5.7 Paenibacillus ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± sp. 0.2 0.6 0.7 0.3 0.5 0.2 0.2 0.1 0.3 0.1 1 0.7 0.7 0.2 0.3 0.09 0.3 5.2 3.1 3.2 3.9 4.4 4.4 3.7 3.5 2.9 5.3 3.4 3.7 2.7 4.3 3.8 3.5 3.4 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Alkaliphilus 0.1 0.3 0.5 0.2 0.3 0.1 0.2 0.3 0.3 0.2 0.5 0.5 0.4 0.1 0.2 0.05 0.2 0.1 15 15 0.19 0.23 0.0078 5.4 0.11 10 6.5 12 0.098 0.17 0.039 Enterobacteria ± ± ± ± ± - - ± ± ± ± ± ± ± ± - ± ceae (family) 0.1 6 6 0.1 0.1 0.007 5 0.05 7 4 6 0.05 0.1 0.03 2.6 2.1 1.9 2.2 2.3 2.9 3.0 2.4 2.0 2.2 2.3 2.3 2.1 2.6 2.7 2.8 2.3 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Lactococcus 0.3 0.02 0.2 0.2 0.4 0.4 0.3 0.3 0.3 0.2 0.2 0.09 0.3 0.4 0.4 0.2 0.4 0.63 0.93 0.44 0.75 1.3 1.1 1.5 1.0 0.48 0.69 0.38 0.91 0.40 0.78 0.54 1.1 1.0 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Bacillus 0.08 0.5 0.1 0.2 0.5 0.5 0.5 0.3 0.06 0.1 0.1 0.2 0.04 0.2 0.1 0.3 0.3 0.94 0.31 0.33 0.56 0.51 0.54 0.40 0.30 0.37 0.89 0.47 0.36 0.29 0.60 0.38 0.39 0.35 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Pseudomonas 0.1 0.05 0.1 0.06 0.07 0.06 0.06 0.15 0.05 0.06 0.2 0.09 0.05 0.1 0.05 0.1 0.03 0.56 0.54 0.50 0.56 0.32 0.37 0.33 0.30 0.41 0.57 0.55 0.56 0.40 0.27 0.26 0.25 0.20 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Brevibacillus 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.3 0.2 0.2 0.2 0.3 0.1 0.2 0.2 0.2 0.1 0.64 0.20 0.27 0.45 0.66 0.20 0.32 0.31 0.40 0.81 0.35 0.38 0.39 0.55 0.24 0.16 0.20 Clostridiaceae ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± (family) 0.05 0.1 0.1 0.05 0.2 0.2 0.2 0.2 0.03 0.03 0.2 0.03 0.04 0.04 0.1 0.08 0.1 0.48 0.37 0.32 0.45 0.32 0.29 0.27 0.23 0.40 0.37 0.48 0.40 0.31 0.29 0.22 0.31 0.29 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Virgibacillus 0.1 0.1 0.1 0.1 0.05 0.2 0.2 0.1 0.08 0.08 0.2 0.1 0.07 0.1 0.1 0.09 0.06 0.34 0.21 0.23 0.16 0.31 0.30 0.36 0.25 0.13 0.35 0.24 0.29 0.23 0.29 0.31 0.036 0.27 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Lysinibacillus 0.07 0.02 0.03 0.08 0.2 0.2 0.06 0.06 0.08 0.03 0.02 0.04 0.09 0.04 0.07 0.07 0.06 0.33 0.16 0.11 0.24 0.35 0.33 0.25 0.19 0.15 0.35 0.16 0.082 0.13 0.36 0.22 0.27 0.22 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Arthrobacter 0.04 0.02 0.06 0.07 0.04 0.1 0.08 0.03 0.01 0.09 0.05 0.04 0.1 0.1 0.03 0.04 0.05 0.58 0.47 0.05 0.34 0.37 0.15 0.14 0.35 0.059 0.071 0.39 0.17 0.13 Clostridiales ± ± ± - ± ± ± ± - ± ± - ± ± - ± ± (order) 0.08 0.4 0.05 0.08 0.2 0.1 0.1 0.2 0.05 0.07 0.05 0.1 0.1

272

0.33 0.10 0.11 0.18 0.22 0.16 0.15 0.082 0.16 0.26 0.18 0.13 0.098 0.19 0.13 0.12 0.090 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Psychrobacter 0.008 0.04 0.06 0.05 0.02 0.02 0.05 0.07 0.04 0.03 0.07 0.03 0.04 0.05 0.01 0.02 0.02 0.23 0.13 0.16 0.38 0.21 0.28 0.24 0.10 0.20 Paenibacillace ± ± - ± ± - ± - ± ± - - ± ± - - - ae (family) 0.2 0.1 0.1 0.2 0.2 0.1 0.2 0.1 0.2 0.31 0.16 0.094 0.18 0.20 0.094 0.28 Enterococcus ± - - ± - - - ± ± - - - ± - ± ± - (family) 0.3 0.1 0.09 0.1 0.2 0.09 0.2 0.14 0.016 0.027 0.12 0.024 0.016 0.027 0.035 0.078 0.047 0.078 0.082 0.020 0.035 ± - ± ± ± ± ± ± ± ± - ± ± ± - ± ± Leuconostoc 0.04 0.01 0.02 0.004 0.02 0.01 0.01 0.02 0.02 0.03 0.06 0.04 0.02 0.03 0.027 0.035 0.035 0.063 0.031 0.020 0.063 0.063 Carnobacteriu - - - ± ± ± ± ± - ± ± - ± - - - - m 0.02 0.03 0.03 0.04 0.03 0.02 0.03 0.06 0.063 0.031 0.027 0.012 0.039 0.075 Bacillaceae - - - - ± - - ± ± ± - - ± ± - - - (family) 0.03 0.03 0.02 0.01 0.03 0.07 0.031 0.035 0.094 0.024 0.012 0.047 Bradyrhizobiac - ± - - ± - - ± - - ± - - - - ± ± eae (family) 0.03 0.02 0.09 0.02 0.01 0.04 0.007 0.007 0.007 0.003 0.027 0.024 0.024 0.016 0.0039 0.047 0.0078 0.012 0.031 8 8 8 9 - ± ± - - ± ± ± ± ± - ± ± ± ± ± ± 0.01 0.01 0.02 0.008 0.003 0.03 0.007 0.01 0.02 Clostridium 0.007 0.007 0.007 0.003 0.003 0.043 0.016 0.016 0.031 0.016 0.020 0.024 0.020 0.016 9 - - ± ± - - ± ± ± ± ± - - ± - ± ± 0.02 0.01 0.01 0.02 0.01 0.02 0.02 0.02 0.01 Streptococcus 0.003 0.031 0.039 0.020 0.0078 0.047 ± - ± - - - ± - - - ± - - ± - - - Sporosarcina 0.03 0.03 0.02 0.007 0.03

273

Table D5. Partial statistical output of “panelperf” function, evaluating the sensory panel’s performance in its ability to discriminate between wines of different treatments (SensoMineR package in RStudio version 3.5.1). A panel of 12 wine experts evaluated 18 pre-determined sensory attributes of the wines, which were subjected to different sulfite and inoculation treatments (n = 2 per treatment). Panelists evaluated each wine twice, in duplicate tasting sessions, with wines served in random order. P-values reported are from three-factor ANOVA [product(wines), panelist, sensory session (tasting session)] for each of the attributes. Two judges were removed from the analysis of sulfur aroma and flavour: one judge neglected to evaluate one or more wines for this descriptor, and the other mistakenly evaluated the aroma and flavour of sulfites as opposed to sulfur compounds.

Sensory Product Panelist Session Product: Product: Panelist: Attribute Panelist Session Session Citrus aroma 0.024* <0.001* 0.754 0.239 0.792 0.051 Pome fruit <0.001* <0.001* 0.917 0.459 0.045* 0.037* aroma Tropical fruit 0.009* <0.001* 0.777 0.329 0.027* 0.031* aroma Melon aroma 0.574 <0.001* 0.053 0.027* 0.006* 0.003* Floral aroma 0.074 <0.001* 0.033* 0.168 0.664 <0.001* Sulfur aroma <0.001* <0.001* 0.059 0.007* 0.206 0.718 Earthy aroma 0.004* <0.001* 0.054 <0.001* 0.179 0.576 Citrus flavour 0.011* <0.001* 0.029* <0.001* 0.655 0.102 Pome fruit <0.001* <0.001* 0.224 0.847 0.324 0.002* flavour Tropical fruit 0.007* <0.001* 0.508 0.498 0.090 0.204 flavour Melon flavour 0.211 <0.001* 0.582 0.013* 0.250 0.006* Floral flavour 0.009* <0.001* 0.032* 0.024* 0.483 0.323 Sulfur flavour <0.001* <0.001* 0.430 0.002* 0.987 0.025* Earthy Flavour 0.019* <0.001* 0.046* <0.001* 0.368 0.271 Sweetness <0.001* <0.001* 0.181 0.016* 0.389 0.148 Acidity <0.001* <0.001* 0.203 <0.001* 0.970 <0.001* Body <0.001* <0.001* 0.578 0.010* 0.136 <0.001* Length of 0.455 <0.001* 0.739 0.002* 0.836 <0.001* Aftertaste * Indicates significance at α = 0.05

274

Appendix E

Table E1. Relative quantity (± SEM) of volatile compounds in wines fermented by yeasts inoculated at different ratios (S. cerevisiae : S. uvarum) and fermented at different temperatures (n = 3 per treatment).

24 C 15 C Compound(s) 1:0 10:1 1:1 1:10 0:1 1:0 10:1 1:1 1:10 0:1 7.84 7.95 6.72 6.63 6.92 5.85 5.63 5.22 7.01 7.82 EA ± ± ± ± ± ± ± ± ± ± 0.46 0.41 0.39 0.35 0.70 0.26 0.067 0.55 0.24 0.49 0.270 0.255 0.273 0.199 0.186 0.258 0.254 0.273 0.476 0.476 EP ± ± ± ± ± ± ± ± ± ± 0.0089 0.0056 0.0094 0.011 0.027 0.0069 0.010 0.018 0.27 0.023 0.0882 0.0971 0.0841 0.181 0.233 0.416 0.0381 0.0476 0.124 0.148 E2MP ± ± ± ± ± ± ± ± ± ± 0.0047 0.0061 0.0067 0.020 0.027 0.0084 0.0015 0.0022 0.0085 0.012 0.667 0.709 0.570 0.340 0.367 0.664 0.705 0.508 0.333 0.334 EB ± ± ± ± ± ± ± ± ± ± 0.074 0.029 0.057 0.016 0.042 0.036 0.035 0.037 0.032 0.052 0.0201 0.0204 0.0169 0.0276 0.0340 0.013 0.00995 0.0131 0.0367 0.0405 E2MB ± ± ± ± ± ± ± ± ± ± 0.0027 0.0018 0.0021 0.0052 0.0047 0.00053 0.00063 0.00068 0.0027 0.0028 0.0247 0.0266 0.0197 0.0277 0.0297 0.0107 0.00986 0.0117 0.0341 0.0381 E3MB ± ± ± ± ± ± ± ± ± ± 0.0035 0.0028 0.0022 0.0043 0.0077 0.0016 0.0013 0.0016 0.0039 0.0040 01.08 1.95 0.913 1.64 1.90 0.504 0.460 0.687 1.56 1.74 2MP ± ± ± ± ± ± ± ± ± ± 0.048 0.031 0.0 0.13 0.29 0.019 0.021 0.039 0.066 0.19 4.61 4.64 3.90 4.10 4.10 2.18 2.35 3.03 3.28 3.31 2/3MBA ± ± ± ± ± ± ± ± ± ± 0.33 0.29 0.13 0.20 0.67 0.15 0.056 0.21 0.18 0.43 67.4 66.5 58.4 51.0 53.2 38.0 36.0 38.5 57.8 60.9 2/3MB ± ± ± ± ± ± ± ± ± ± 2.9 1.3 3.1 4.4 8.2 0.69 0.50 2.0 0.98 4.1 22.2 22.1 20.6 11.0 10.4 21.4 22.3 18.8 11.6 11.0 EH ± ± ± ± ± ± ± ± ± ± 0.79 1.8 0.43 0.48 0.76 0.55 0.58 0.80 0.36 0.29 HA 1.43 1.44 1.44 1.07 1.02 1.51 1.78 1.97 1.33 1.26

275

± ± ± ± ± ± ± ± ± ± 0.048 0.12 0.083 0.056 0.19 0.092 0.14 0.12 0.13 0.16 2.27 2.24 2.04 1.94 2.01 2.77 2.83 2.45 2.12 2.05 HOH ± ± ± ± ± ± ± ± ± ± 0.062 0.028 0.021 0.11 0.17 0.16 0.075 0.17 0.17 0.064 99.2 111 82.3 67.6 81.9 61.5 59.1 32.7 45.3 72.8 EO ± ± ± ± ± ± ± ± ± ± 10 13 9.8 4.0 11 6.3 3.2 8.5 24 18 58.5 64.5 51.8 43.2 41.2 31.5 25.1 20.9 24.3 30.9 ED ± ± ± ± ± ± ± ± ± ± 5.1 8.1 23 1.7 6.8 8.2 2.5 4.9 12 5.1 0.175 0.170 0.145 0.186 0.222 0.116 0.110 0.134 0.314 0.312 2/3MBAcid ± ± ± ± ± ± ± ± ± ± 0.0073 0.0056 0.0029 0.022 0.036 0.0070 0.0043 0.017 0.031 0.015 0.158 0.181 0.114 0.0857 0.106 0.0388 0.0357 0.0405 0.0717 0.0878 MOH ± ± ± ± ± ± ± ± ± ± 0.015 0.017 0.0070 0.014 0.011 0.0044 0.0013 0.0030 0.011 0.012 10.5 10.9 7.47 37.6 46.0 2.09 2.17 4.79 29.1 34.2 2PEA ± ± ± ± ± ± ± ± ± ± 0.19 0.62 0.65 2.2 5.7 0.22 0.049 0.28 4.1 3.7 11.5 12.4 7.69 20.3 25.8 5.01 4.41 5.22 15.5 27.2 HAcid/EDD ± ± ± ± ± ± ± ± ± ± 1.1 1.9 3.9 0.71 6.9 1.4 0.57 1.7 9.4 6.4 0.0144 0.0135 0.0114 0.0111 0.0115 0.0106 0.0105 0.0085 0.0100 0.0106 Benz ± ± ± ± ± ± ± ± ± ± 0.0020 0.00039 0.0012 0.00048 0.0020 0.00073 0.0010 0.00070 0.00024 0.00096 58.3 61.5 53.6 117 130 28.3 27.2 35.4 106.49 119 2PE ± ± ± ± ± ± ± ± ± ± 2.8 4.0 4.2 9.5 16 1.7 0.25 2.0 0.82 5.1 8.38 9.45 7.84 6.02 6.74 9.95 10.2 5.19 5.82 9.67 OA ± ± ± ± ± ± ± ± ± ± 0.26 0.53 0.32 0.17 0.68 0.47 0.095 1.4 3.4 2.0 12.7 12.8 14.0 7.83 6.16 16.4 14.8 10.5 7.88 8.41 DA ± ± ± ± ± ± ± ± ± ± 1.6 0.61 0.60 0.18 0.47 2.2 0.40 1.3 1.6 0.097

276