YEAST RESEARCH. Comparative metabolic footprinting of a large number of commercial wine yeast strains in Chardonnay fermentations.

Transcription

1 RESEARCH ARTICLE Comparative metabolic footprinting of a large number of commercial wine yeast strains in Chardonnay fermentations Chandra L. Richter 1, Barbara Dunn 2, Gavin Sherlock 2 & Tom Pugh 1 1 E. & J. Gallo Winery, Modesto, CA, USA; and 2 Department of Genetics, Stanford University, Stanford, CA, USA Correspondence: Chandra L. Richter, E. & J. Gallo Winery, P.O. Box 1130, Modesto, CA 95353, USA. Tel.: ; fax: ; chandra.richter@ejgallo.com Received 29 November 2012; revised 6 March 2013; accepted 6 March Final version published online 12 April DOI: / Editor: Isak Pretorius Keywords wine yeast; alcoholic fermentation; metabolic footprint; volatile aroma composition; Saccharomyces cerevisiae; Chardonnay. Abstract Wine has been made for thousands of years. In modern times, as the importance of yeast as an ingredient in winemaking became better appreciated, companies worldwide have collected and marketed specific yeast strains for enhancing positive and minimizing negative attributes in wine. It is generally believed that each yeast strain contributes uniquely to fermentation performance and wine style because of its genetic background; however, the impact of metabolic diversity among wine yeasts on aroma compound production has not been extensively studied. We characterized the metabolic footprints of 69 different commercial wine yeast strains in triplicate fermentations of identical Chardonnay juice, by measuring 29 primary and secondary metabolites; we additionally measured seven attributes of fermentation performance of these strains. We identified up to 1000-fold differences between strains for some of the metabolites and observed large differences in fermentation performance, suggesting significant metabolic diversity. These differences represent potential selective markers for the strains that may be important to the wine industry. Analysis of these metabolic traits further builds on the known genomic diversity of these strains and provides new insights into their genetic and metabolic relatedness. YEAST RESEARCH Introduction Evidence for the production of fermented beverages by Saccharomyces yeasts, mostly S. cerevisiae, dates at least as far back as 7000 BC, making it one of the world s oldest biotechnological processes (Cavalieri et al., 2003; McGovern, 2003; McGovern et al., 2004). Fermentation technologies, and their associated yeasts, have comigrated with humans and agriculture to spread throughout the world. Saccharomyces yeasts, originally present as wild strains in soil, fruits, or tree sap, were transported by humans, along with their associated sugar source, that is, grape or grain. This likely imposed a selective pressure on these domesticated yeasts to functionally specialize for the production of specific alcoholic beverages on differing substrates and/or the production of particular desirable flavor compounds. There is now a sizable collection of domesticated strains of S. cerevisiae associated with, and specialized for, specific industrial habitats, and strains isolated from a given habitat are often genetically similar to each other (Legras et al., 2005, 2007; Camarasa et al., 2011; Sicard & Legras, 2011). Grape juice (or must), which has a low ph ( ) and high osmolarity (sugars of g L 1 ), is an environment that is highly unfavorable for the survival of most microorganisms, but to which wine yeasts are well adapted. In addition, SO 2 is routinely added at the winery to concentrations of mg L 1 to prevent oxidation and microbial spoilage, also contributing to the harsh conditions (Pizarro et al., 2007). The onset of fermentation makes the condition even less hospitable for most organisms: the environment becomes anaerobic, ethanol concentrations begin to rise, and nutrients are depleted. However, due to being well adapted to this changing environment, S. cerevisiae rapidly becomes the dominant species in the fermentation. Evolutionary adaptation of the stress response to enable S. cerevisiae to better survive in the harsh environment of fermenting grape juice is not well understood, but has likely resulted in the divergence of S. cerevisiae wine strains from nonwine S. cerevisiae strains (Legras et al., 2007; Dunn et al., 2012). Because winemaking has been practiced for several millennia (Frankel, 1998; Cavalieri et al., 2003; McGovern, 2003), there has been ample

2 Metabolomic perspective on wine fermentations 395 opportunity for divergence of wine yeast strains from other industrial and nonindustrial S. cerevisiae strains. Indeed, large-scale genomic studies have shown this divergence from other industrial, laboratory, and pathogenic strains of S. cerevisiae (Legras et al., 2007; Borneman et al., 2011; Dunn et al., 2012). It is likely these genomic differences allow wine yeast to maintain metabolic activity throughout the entire winemaking process. To date, there have been few studies examining the fermentation performance and metabolic diversity within wine yeast during wine fermentation. Starting approximately years ago, wine yeast strains began to be isolated from various geographical regions and sold commercially. Today, most commercial wine fermentations are inoculated with a pure culture of a single commercial yeast strain rather than allowing a natural or wild (noninoculated) fermentation to occur. These strains, likely derived from strains domesticated hundreds or thousands of years earlier, were isolated from fermentations with certain desirable sensory traits (metabolic footprint) and/or predictable behavior (fermentation kinetics) (Pretorius, 2000). Currently, there are more than 200 commercially produced wine yeast strains available to winemakers. Domestication, commercialization, and subsequent dissemination by humans have likely influenced the genetic structure of wine yeast strains (Pretorius, 2000; Dunn et al., 2005, 2012; Legras et al., 2007). Genomic differences can lead to phenotypic differences between strains, resulting in unique fermentation behavior and/or sensory characteristics [reviewed in (Pretorius et al., 2012)]. The sensory perception of wine is due to a specific metabolite profile, that is, the small molecules that contribute to the aroma, flavor, and mouthfeel. Grape juice is composed primarily of water, sugar, organic acids, and minor components, which can be used as a source of nutrients for yeast during fermentation. The wine resulting from fermentation of juice is composed primarily of water, ethanol, glycerol, organic acids, and many minor compounds contributing to flavor, aroma, and mouthfeel (Pizarro et al., 2007). These flavor, aroma, and mouthfeel compounds either are present in the grape and remain unmodified during fermentation or are synthesized during the winemaking process by yeast, typically from precursors in the grape juice. The effect of a specific yeast strain s cellular metabolism on its environment results in a unique set and concentration of metabolites; the concentration of many of these metabolites can be measured, giving a metabolic footprint or metabolite profile. For a given starting juice, this footprint can change due to the yeast strain used or the environmental conditions of the fermentation. Previous studies have shown that different S. cerevisiae wine yeast strains fermenting the same juice under identical conditions can yield very different wines due to differences in the metabolic footprint (Gardner et al., 1993; Romano et al., 1994, 2003; Benitez et al., 1996; Mortimer, 2000; Remize et al., 2000; Torrea & Ancin, 2002; Fleet, 2003). However, most of these studies compared only a small number of yeast (< 10) and/or a small number of metabolites (< 20), often in synthetic must or media, thereby not giving a true representation of a grape fermentation. In this study, we determined, in triplicate, the fermentation profiles and metabolic footprints (a total of 29 attributes) of 69 commercial wine yeast strains in Chardonnay grape juice under identical fermentation conditions. This allowed us to better understand the phenotypic relatedness of these strains, compared with the genomic relatedness we previously reported (Dunn et al., 2012). We observed variation in both the fermentation kinetics and metabolic footprints for these strains. We found groups of strains with related phenotypes (both fermentation and metabolite attributes) and were able to correlate these similarities to genomic similarities, providing insight into the diversity of this unique class of industrial strains. Materials and methods Yeast strains The S. cerevisiae wine yeast strains used in this study, listed in Table 1, were obtained from commercial yeast suppliers. Single colonies for each were isolated and then preserved at 80 C for reuse, to ensure pure cultures were used for each strain and that the same single colony isolate was used for all inoculations. Preparation of juice Chardonnay juice (at 24 Brix or 240 g L 1 glucose and fructose) was collected during harvest 2007 and stored frozen in 60-L drums until use. Each 60-L drum was thawed for 4 days at 4 C. After thawing, the juice was pad-filtered (2 lm; 3M Purification, Inc, Tustin CA) and sterile-filtered (0.2 lm; Pall Corp, Covina, CA). The filtered juice was added to sterile spinner flasks (Bellco, Vineland, NJ) in 1.5 L volumes. Preparation of fermentation starter cultures A small amount of the single-colony isolate of each commercial yeast strain was transferred from its frozen stock to an YPD plate (1% Bacto yeast extract, 2% Bacto peptone, 2% glucose, 1.5% agar) and incubated for 2 days at 28 C. One colony from the plate was used to inoculate 5 ml FEMS Yeast Res 13 (2013) ª 2013 Federation of European Microbiological Societies

3 396 C.L. Richter et al. Table 1. Origins and sources of Saccharomyces cerevisiae wine strains studied. Origins were obtained from yeast supplier s catalog. A singlecolony isolate was collected from each commercial yeast preparation and preserved at 80 C for reuse; the same single-colony isolate was used to inoculate all fermentations. Yeast strain Catalog name Supplier Origin 43 Uvaferm 43 Lallemand Inter Rhone Anchor Inter Rhone 4F9 Fermicru 4F9 DSM Nantes, France 58W3 58W3 Vinquiry Alsace, France 71B Lalvin 71B Lallemand Narbonne, France AWRI 350 AWRI 350 Maurivin AWRI AWRI 796 AWRI 796 Maurivin South Africa AWRI R2 Maurivin R2 Maurivin Bordeaux, France BA11 BA11 Lallemand Estacao Vitivinicola de Baraida BDX Enoferm BDX Lallemand Pastuer Institute, Paris, France BGY Burgundy Lallemand Burgundy, France BM45 Lalvin BM45 Brunello Lallemand University of Siena BP 725 BP 725 Maurivin France BRL97 BRL97 Barolo Lallemand University of Torino CSM CSM Lallemand ITV Bordeaux CY3079 Lalvin CY3079 Lallemand Bourgogne D254 Lalvin ICV-D254 Lallemand ICV, Rhone Valley D47 Enoferm IVC-D47 Lallemand Cotes du Rhone D80 Lalvin ICV-D80 Lallemand Cote Rotie, Rhone Valley DV10 DV10 Lallemand Champagne EC1118 Lalvin EC-1118 Lallemand Champagne Elegance Maurivin Elegance Maurivin Portugal EpernayII Maurivin EP 2 Maurivin Epernay, France F15 Zymaflore F15 Laffort Medoc F33 Actiflore C (F33) Scott Labs/Laffort Laffort Research Laboratory FA1 FA1 Scott Labs/Lallemand Fermichamp Fermichamp DSM Alsace, France ICV-GRE Lalvin ICV-GRE Lallemand Rhone Valley IOC IOC Epernay K1 Lalvin V1116 Lallemand Montpellier L2056 Rhone L2056 Lallemand Cotes du Rhone L2226 Enoferm L2226 Lallemand Cotes du Rhone L2323 Lalvin L2323 Lallemand Rhone Valley Lalvin AC Lalvin AC Lallemand Loire LVCB Fermicru LVCB DSM Casablanca Valley, Chile N96 N96 Anchor South Africa NT112 NT 112 Anchor Stellenbosch, South Africa NT116 NT 116 Anchor Stellenbosch, South Africa NT202 NT 202 Anchor Stellenbosch, South Africa NT45 NT 45 Anchor South Africa NT50 NT 50 Anchor Stellenbosch, South Africa PC Premier Cuvee Lesaffre France PDM Maurivin PDM Maurivin Primeur Maurivin Primeur Maurivin INRA Narbonne, France QA23 Enoferm QA23 Lallemand UTAD in Portugal R2 R2 Lallemand Sauternes region of Bordeaux RC212 Lalvin RC212 Lallemand Burgundy Rhone 4600 Rhone 4600 Lallemand Cotes du Rhone S-101 St. Georges S-101 Lesaffre/Springer Beaujolais S-102 C.K. S-102 Lesaffre/Springer Val de Loire S-325 U.C.L.M. S-325 Lesaffre/Springer Spain S-377 U.C.L.M. S-377 Lesaffre/Springer Spain SAUV L3 Sauvignon L3 Maurivin Bordeaux, France Simi White Simi White Lallemand

4 Metabolomic perspective on wine fermentations 397 Table 1. Continued Yeast strain Catalog name Supplier Origin SYR Syrah Lallemand Cotes du Rhone T306 T306 Lallemand Hunter Valley, NSW Australia T73 T73 Lallemand La Universidad de Velencia of Spain UCD522-L Lallemand UCD 522 Lallemand Pastuer Institute, Paris, France UCD522-M Mauri UCD522 Maurivin UC Davis VIN13 VIN 13 Anchor Stellenbosch, South Africa VIN7 VIN 7 Anchor Stellenbosch, South Africa VL1 Zymaflore VL1 Scott Labs/Laffort Bordeaux Institute of Oenology VL2 Zymaflore VL2 Scott Labs/Laffort VL3C Zymaflore VL3 Scott Labs/Laffort Bordeaux Institute of Oenology VR5 Fermicru VR5 DSM Burgundy, France W372 WE 372 Anchor Stellenbosch, South Africa WE14 WE 14 Anchor South Africa Williams Selyem Williams Selyem Vinquiry Sonoma, CA X5 Zymaflore X5 Scott Labs/Laffort Laffort Research Laboratory YPD broth, for a total of three 5-mL cultures, and incubated for 2 days at 28 C. After 2 days, the 5-mL cultures were used to inoculate three different flasks each with 100 ml YPD broth and incubated for 2 days at 28 C. The cells were harvested by centrifugation, washed in 200 ml sterile 0.9% sodium chloride solution, and resuspended in 25 ml of 0.9% sodium chloride. Cell concentration of the yeast slurry was obtained using a Z2 Coulter Particle Counter (Beckman Coulter, Hialeah, FL), and the juice was inoculated with the amount of yeast slurry to give an initial concentration of cells ml 1. Fermentation protocol Fermentations were performed in spinner flasks in 1.5 L volumes at 18 C with a constant agitation of 25 r.p.m. Samples were collected daily to measure cell concentration (Z2 Coulter Particle Counter) and four chemical attributes: sugar (glucose and fructose UV Method, Randox, Antrim, UK), ethanol, and two measurements of yeast assimilable nitrogen (YAN): NOPA (nitrogen by OPA) and ammonia (Ammonia Enzymatic UV, Randox, Antrim, UK). Analytical methods Upon completion of fermentation, the wines were coldsettled (48 h at 1.7 C/35 F) and analyzed for standard wine chemical attributes, including the following: residual sugar, ethanol (GC column 80/120 Carbopack B AW 5% Carbowax 20M, Supelco, Bellefonte, PA), ph, malic acid (L-Malic Acid, Megazyme, County Wicklow, Ireland), volatile acidity (Bergmeyer & Mollering, 1974), glycerol (Glycerol Assay, Megazyme, County Wicklow, Ireland), total SO 2 (FIAstar 5000, Foss, Hoganas, Sweden), and free SO 2 (FIAstar 5000, Foss, Hoganas, Sweden). The finished wines were analyzed for yeast-produced metabolites, including the following: fusel alcohols, acetaldehyde, and ethyl acetate (A.O.A.C., 1997). To measure additional aroma compounds, samples were analyzed using 6890 Series Gas Chromatograph with a J&W Scientific DB-5MS High-Resolution Gas Chromatography Column [adapted from (Soleas et al., 2002)]. Screening for polygalacturonase activity Select strains were screened for polygalacturonase activity using a plate assay described by Louw et al. (Louw et al., 2010). Vin7, Epernay II, NT45, and NT50, grown overnight in YPD (1% yeast extract, 2% peptone and 2% dextrose), were spotted on polygalacturonase plates (1.25% polygalacturonic acid (Sigma-Aldrich), 0.67% yeast extract, 1% dextrose, 20% agar) and incubated at 28 C for 3 days. Degradation was visualized by washing colonies off with distilled water and staining plates with 6M HCl. Statistical analysis Statistical analyses, including principal component analysis (PCA) and agglomerative hierarchical clustering, were performed using XLSTAT software, version (Addinsoft, New York, NY) and The UNSCRAMBLER, version 10.1 (Camo Software, Woodbridge, NJ). Results To investigate phenotypic differences among 69 commercial wine yeast strains, we characterized their fermentation performance and metabolic footprints in filter-sterilized Chardonnay juice (Table 1). The FEMS Yeast Res 13 (2013) ª 2013 Federation of European Microbiological Societies

5 398 C.L. Richter et al. fermentations were conducted using standard winemaking practices in triplicate for each strain, using a single juice source to minimize variation based on juice chemistry. We measured fermentation kinetics as well as concentrations of metabolic byproducts and volatile organoleptic compounds that are either produced or volatilized by the yeast. Fermentation kinetics We observed significant variation in the fermentation kinetics between the different strains. The length of fermentation, measured from the time of yeast inoculation to the point at which all fermentable sugars were consumed, varied from 5 to 18 days (Fig. 1, Table 2). The length of time required to ferment 50% of the available sugar (120 g L 1 ) was similar among all strains tested, ranging from 2.3 to 4.2 days (Fig. 1, Table 2). However, the range became larger during the next stage of fermentation, with the consumption of 83% of the sugars (40 g L 1 ) ranging from 4.0 to 11.3 days (Fig. 1, Table 2). Variability among the strains was the most pronounced during consumption of the final 17% of sugars (dryness, < 2gL 1 ). In the early stages of fermentation, the yeasts were actively dividing, increasing in cell concentration until day three or four, at which point cell division ceased and the cell concentration remained stable for the remainder of the fermentation ( 10% reduction in the final fermentation phase; data not shown). Most strains reached a maximum cell concentration of cells ml 1, although concentrations ranged from to cells ml 1 (Table 2). There was a weak positive correlation between maximum cell concentration and the duration of fermentation (r = 0.49). There are likely several factors contributing to this observation; for example, cell concentration does not always correlate to total cell biomass or dry weight (Prescott, 1975). The cell biomass was not measured in this study but may have had a stronger influence on fermentation rate than maximum cell concentration. In addition to consuming sugars, yeast will consume nitrogen in the form of free amino acids and ammonia, the combination of which is referred to as YAN. Assimilable nitrogen is required for S. cerevisiae growth in a grape juice fermentation. Although there are many sources of nitrogen in grape juice, only ammonia and nonproline amino acids can be assimilated. Nitrogen assimilation is important for yeast growth and completion of fermentation; individual yeast strains have been shown to have unique assimilation patterns (Crepin et al., 2012). The strains used in this study consumed 55 90% of the available YAN, with a mean of 76.8% (Table 2). Nitrogen consumption contributes, in part, to total biomass, and there was a weak positive correlation (r = 0.46, P < 0.001) between the maximum cell concentration and the percentage of YAN consumed by each strain. Fig. 1. Comparison of fermentation rates for 69 Saccharomyces cerevisiae wine yeast strains fermenting Chardonnay grape juice. Fermentation rate is displayed as time to consume sugar; Time 120 sugar is the time for each strain to consume 50% of the available sugar, Time 40 g L 1 sugar is the time for each strain to consume 83%, and Time Dryness is the time for each strain to consume all available sugar (dryness < 2.0 g L 1 sugar); note that this data are presented in numerical form in Table 2. Fermentations were performed in triplicate. Error bars represent one standard deviation. EPII: Epernay II.

6 Metabolomic perspective on wine fermentations 399 Table 2. Fermentation kinetics for 69 wine yeast strains. Fermentations were carried out in triplicate in Chardonnay grape juice containing g L 1 sugar, mg L 1 YAN, g L 1 TA, and mg L 1 SO 2. Fermentation rate is displayed as time to consume sugar: Time to 120 g L 1 sugar is the time (in days) for each strain to consume 50% of the available sugar, Time to 40 g L 1 sugar is the time for each strain to consume 83%, and Time to Dryness is the time for each strain to consume all available sugar (where dryness is < 2.0 g L 1 sugar). Maximum cell concentration was the maximum concentration observed through daily tracking. Percentage of YAN and malic acid consumed during fermentation is also shown. Yeast Days to 120 g L 1 sugar Days to 40 g L 1 sugar Days to dryness Max cell density wine cells ml 1 % YAN consumed % Malic consumed Alcohol%v/v F W B AWRI R AWRI AWRI BA BDX BGY Bm BP BRL CSM CY D D D DV EC Elegance EPII F F FA Fermichamp ICV-GRE IOC K L L L Lalvin AC LVCB N NT NT NT NT NT PC PDM Primeur QA R RC Rhone FEMS Yeast Res 13 (2013) ª 2013 Federation of European Microbiological Societies

7 400 C.L. Richter et al. Table 2. Continued Yeast Days to 120 g L 1 sugar Days to 40 g L 1 sugar Days to dryness Max cell density wine cells ml 1 % YAN consumed % Malic consumed Alcohol%v/v S S S S SauvL Simi White SYR T T UCD UDC522-Mauri Vin VIN VL VL VL3C VR W WE Williams Selyem X Metabolite footprinting Large differences were seen among the strains in metabolite production. We measured 29 yeast-derived metabolic byproducts in the wine. These metabolites represent the exo-metabolome, with each compound contributing to the complex aroma and flavor of wine, including higher alcohols, ethyl esters, and acetate esters, among others (Table 3). The majority of the traits measured centered symmetrically around the mean, rather than having bimodal or multimodal distributions (Fig. 2). Camarasa et al. (2011) have previously measured 18 metabolic byproducts produced by yeast during fermentation and identified ranges of 2- to 15-fold. In this study, we measured many additional fermentation attributes and metabolic byproducts and observed differences of up to 1000-fold between strains (Fig. 2, Table 3). The impact of yeast on wine flavor is determined by the differential production of these metabolic compounds, several of which are volatile and also influence the aroma of the wine (Guth, 1997; Ferreira et al., 2000). Alcohols In a grape juice fermentation, ethanol yields are 90 95% of theoretical, with the remaining 5 10% explained by both conversion of glucose to biomass and loss of ethanol through evaporation (Konig et al., 2009). The range of ethanol produced was %v/v (Table 2). During the course of fermentation, yeast will also produce higher alcohols (also called fusel alcohols, containing > 2 carbon atoms) as secondary metabolites from amino acid synthesis. These higher alcohols are the most abundant volatile components in wine (Sumby et al., 2010). In this study, we measured active amyl alcohol, iso-amyl alcohol, iso-butyl alcohol, n-butyl alcohol, and n-propyl alcohol. These alcohol metabolites can have both positive and negative impacts on the wine, depending on the concentration (Konig et al., 2009). Total higher alcohol concentrations of 400 mg L 1 or greater can result in a pungent, solvent aroma, whereas concentrations of < 300 mg L 1 are often described as imparting a fruity aroma (Swiegers & Pretorius, 2005). Approximately 10% of the strains produce total higher alcohols > 300 mg L 1 ; none of the strains were observed to produce > 400 mg L 1. Several of the higher alcohols correlated positively with each other. Active amyl alcohol and iso-butyl alcohol are two branched-chain alcohols synthesized through the Ehrlich pathway that we found to be positively correlated (r = 0.67, P < 0.001, Fig. 3). Strains producing high amounts of both were BDX, L2056, Vin7, and BP725. Strains producing low amounts of both compounds were 43, D47, EC1118, NT112, and VL3C. Although not as abundant as active amyl alcohol, n-propyl alcohol and the total concentration of higher alcohols were also positively correlated (r = 0.76, P < 0.001, Fig. 3). Methanol, which is not a higher alcohol,

8 Metabolomic perspective on wine fermentations 401 was also measured. All strains, except one, had approximately 40 mg L 1 methanol in the finished wine, similar to levels that have been reported previously (Rossouw et al., 2008). Vin7-produced wine contained over twofold more methanol than any other strain. Methanol can be released into wine through pectinase activity (Louw et al., 2010; Eschstruth & Divol, 2011). Although most S. cerevisiae have little to no pectinase activity, a few strains and several hybrid yeast strains have been shown to possess pectinase activity (Louw et al., 2010; Eschstruth & Divol, 2011). Vin7 is a S. cerevisiae/s. kudriavzevii hybrid and was found to have increased pectinase activity suggesting that Vin7 may possess the ability to degrade pectin thereby releasing methanol (Supporting Information, Fig. S1). Sulfur dioxide During fermentation, yeast produces sulfur dioxide, the amount of which has been shown to be strain dependent (Swiegers & Pretorius, 2007). Sulfur dioxide can help minimize oxidation events that may ruin wine, but it can also result in problematic secondary processing steps if present at levels above 100 mg L 1 (Konig et al., 2009). SO 2 production among the strains in this study varied by 10-fold, ranging from 20 to 200 mg L 1. The majority of the strains (86%) produced < 50 mg L 1 ; however, 16% produced > 50 mg L 1, and two strains, BM45 and NT112, produced greater 100 mg L 1. Glycerol, the third major component in dry wines after ethanol and water, is the final product in the glyceropyruvic fermentation pathway that results in the regeneration of NAD+ (Moreno-Arribas & Polo, 1993). The glycerol concentrations produced by these strains varied from 0.6 to 10 g L 1, with an average of 6.5 g L 1 (Table 3). Only one strain, NT50, reached the 10 g L 1 threshold that is thought to confer a desirable sweetness flavor and positive mouthfeel sensation (Remize et al., 1999). NT50 is a S. cerevisiae/s. kudriavzevii hybrid (Dunn et al., 2012) that may have elevated glycerol production due to its S. kudriavzevii heritage (see Discussion). Acetaldehyde Acetaldehyde is produced by S. cerevisiae through the oxidation of ethanol by alcohol dehydrogenase II (encoded by ADH2) (Jackowetz et al., 2011). The yeast strains characterized in this study produced between 20 and 150 mg L 1 (Table 3). Acetaldehyde will continue to increase as the wine ages through oxidation of ethanol and can reach concentrations of 400 mg L 1 (Konig et al., 2009). Acetaldehyde contributes a nutty, sherry-like quality to the wine and is not desirable in most wine styles. Yeast strains BM45 and NT112 both produced acetaldehyde in concentrations > 100 mg L 1 (Table 3). These concentrations are above the sensory threshold for acetaldehyde, and further increases may negatively affect the wine flavor. Two strains, Vin7 and W372, were very low producers (< 25 mg L 1, Table 3). Volatile acidity Volatile acidity (VA) is the measurement of volatile acids, including acetic acid, lactic acid, succinic and propionic acid, which are important in winemaking because of flavor impacts as well as legal regulations. Yeast will produce these acids through metabolic pathways. High levels of acetic acid can cause the wine to become pungent and smell like vinegar; however, low levels can improve the quality. Legally, the VA must be below 1.1 g L 1 in white wine. The yeasts in this study produced a range of g L 1 VA (Table 3). The mean VA produced was 0.4 g L 1 ; however, we did observe a few strains that produced 1.0 g L 1, very close to the legal limit. In addition to producing acids, yeast can metabolize acid. Several groups have observed that yeast will consume malic acid during fermentation in a strain-dependent manner (Pretorius, 2000; Konig et al., 2009). In this study, the yeast strains consumed anywhere from 10 to 40% of the available malic acid (Table 2). Glycerol Esters Esters are produced by yeast during fermentation and typically impart a fruity aroma to the wine. Esters are the second most abundant volatile compounds in wine and are produced by microbial esterases, alcohol acetyltransferases, and lipases (Sumby et al., 2010). Different commercial yeast strains produce variable amounts of esters and, therefore, a variety of wine aroma profiles (Lambrechts & Pretorius, 2000; Vilanova et al., 2007; Cadiere et al., 2011). We measured 14 esters important for Chardonnay varietal characteristics to investigate yeast strain differences. Ethyl acetate is the most common ester in wine and is readily formed by a reaction between ethanol and acetic acid (Sumby et al., 2010). At low levels (< 100 mg L 1 ), ethyl acetate contributes to a fruity aroma in the wine; at concentrations above 100 mg L 1, ethyl acetate is described as solvent/nail polish. Most strains in this study produced < 100 mg L 1 ethyl acetate, with a mean of 62 mg L 1 (Table 3), while Vin13 and NT116 both produced over 100 mg L 1 ethyl acetate. Factors such as nitrogen content and grape esters impact the final concentration. Specific grape varieties are associated with specific yeast- FEMS Yeast Res 13 (2013) ª 2013 Federation of European Microbiological Societies

9 402 C.L. Richter et al. Table 3. Comparison of phenotypic variables comprising yeast metabolic footprint. Fermentations were carried out in triplicate in Chardonnay grape juice. Concentrations were determined in the final wine product. Yeast VA g L 1 Total SO2 mg L 1 Glycerol g L 1 Acetaldehyde mg L 1 Active amyl alcohol mg L 1 Iso-amyl alcohol mg L 1 Iso-butyl alcohol mg L 1 n-butyl alcohol mg L 1 n-propyl alcohol mg L 1 Methanol mg L 1 2-phenyl ethanol lg L methyl pentanol lg L vinyl guaiacol lg L 1 Heptanol lg L F W B AWRI R AWRI AWRI BA BDX BGY Bm BP BRL CSM CY D D D DV EC Elegance EPII F F FA Fermichamp ICV-GRE IOC K L L L Lalvin AC LVCB N NT NT NT NT NT PC PDM Primeur QA R RC Rhone S S S S SauvL Simi White SYR T T UCD UDC522-Mauri Vin VIN VL VL VL3C VR W WE Williams Selyem X

10 Metabolomic perspective on wine fermentations 403 Hexanol lg L 1 Ethyl butanoate lg L 1 Ethyl decanoate lg L 1 Ethyl hexanoate lg L 1 Ethyl laurate lg L 1 Ethyl Octanoate lg L 1 Ethyl Palmitate lg L 1 Ethyl Succinate lg L 1 Ethyl valine lg L 1 Ethyl acetate lg L 1 Hexyl acetate lg L 1 Isoamyl acetate lg L 1 Phenethyl acetate lg L 1 Nonanoicloate lg L 1 Syringaldehyde lg L FEMS Yeast Res 13 (2013) ª 2013 Federation of European Microbiological Societies

11 404 C.L. Richter et al. (a) (b) Fig. 2. Comparison of phenotypic variables (fermentation performance attributes and wine metabolite concentrations) for 69 Saccharomyces cerevisiae wine yeast strains fermenting Chardonnay grape juice. (a) Pie graph showing the breakdown of types of attributes measured in this study. (b) Distribution frequency graphs for maximum cell density and six representative wine metabolites, for 69 S. cerevisiae wine strains. Maximum cell density was the maximum concentration observed during the course of fermentation as monitored by daily measurement. Final total SO 2, acetaldehyde, ethyl acetate, glycerol, 2-phenyl ethanol, and ethyl hexanoate concentrations were measured in the finished wine. derived esters, likely due to different amino acid profiles in grape (Ferreira et al., 2000). Correlation between phenotypic traits We used the Pearson s product moment method to perform correlation analysis between the phenotypic traits assayed in this study (Fig. 3a). Consistent with previous studies, most of the variables were independent (Camarasa et al., 2011). However, a few of the variables were strongly correlated (Fig. 3). Not surprisingly, Time to 40 g L 1 sugar was strongly correlated with Time to Dryness (r = 0.91, P < 0.001). The production of SO 2 and acetaldehyde was also positively correlated (r = 0.89,

12 Metabolomic perspective on wine fermentations 405 (a) (b) Fig. 3. Relationships between metabolic variables. (a) Correlation matrix of entire dataset including fermentation kinetics and wine metabolites. (b) Relationship between specific pairs of phenotypic variables: time to dryness and time to 40 g L 1 sugar; total SO 2 concentration and acetaldehyde; n-propyl alcohol and total higher alcohols; active amyl alcohol and iso-butyl alcohol; 2-phenethyl acetate and hexyl acetate; hexyl acetate and isoamyl acetate. r values labeled on graphs. FEMS Yeast Res 13 (2013) ª 2013 Federation of European Microbiological Societies

13 406 C.L. Richter et al. P < 0.001, Fig. 3b) and highly variable among strains (Moreno-Arribas & Polo, 1993). Aldehydes are typically unstable compounds; however, SO 2 will bind acetaldehyde creating the stable adduct acetaldehyde hydroxysulfonate (Moreno-Arribas & Polo, 1993). Therefore, high-sulfiteproducing strains are acetaldehyde stabilizing strains, resulting in a correlation between these two metabolites. The total fusel alcohol concentration was positively correlated with n-propyl alcohol concentration (r = 0.76, Fig. 3b). n-propyl alcohol is the second most abundant fusel alcohol identified in this study (Table 3). It is produced as a byproduct in sterol biosynthesis (Rossouw et al., 2008). Active amyl alcohol and isobutyl alcohol are also positively correlated (r = 0.68, Fig. 3b). Active amyl alcohol and isobutyl alcohol are produced through catabolism of iso-leucine and valine, respectively (Rossouw et al., 2008). Additionally, we found correlations between some acetate esters (Fig. 3b). Phenethyl acetate (honey, rose) correlated with hexyl acetate (fruity; r = 0.68), and isoamyl acetate (banana) correlated with hexyl acetate (fruity; r = 0.69). Phenethyl acetate is produced through catabolism of phenylalanine, and isoamyl acetate is produced through the catabolism of leucine (Rossouw et al., 2008). Hexyl acetate is produced through C6 compounds, such as hexanol, which are neutral compounds commonly found in grapes that influence wine aroma (Dennis et al., 2012). Hierarchical clustering and PCA To better understand the phenotypic relatedness between all 69 strains, we performed an agglomerative hierarchical clustering of the metabolic footprints of the yeasts. This analysis included all of the quantitative phenotypic traits, that is, the 29 yeast-produced metabolites. This clustering resulted in four groups of yeast strains, with each group containing from 6 to 30 strains (Fig. 4a). The dendrogram for each of the four groups is flat, suggesting homogeneity within the group (Fig. 4a). The compounds with the largest variability (based on standard deviation) were as follows: glycerol, ethyl hexanoate, ethyl octanoate, 2-phenyl ethanol, isoamyl acetate, and 4-vinyl guaiacol (Fig. 2 and data not shown). These compounds are all important for the aroma and mouthfeel of wine and are produced as byproducts of glycolysis (glycerol), fatty acid metabolism (ethyl hexanoate and ethyl octanoate), amino acid catabolism (2-phenyl ethanol and isoamyl acetate), and grape derived phenolic precursors (4-vinyl guaiacol). A PCA using the same 29 yeast-derived metabolites was also performed. This analysis explained 77% and 16% of the variance in the first two discriminant axes, respectively. The metabolites most responsible for the variation are 2-phenyl ethanol (a volatile phenol) and glycerol (an alcohol) along PC1 and isoamyl acetate (an acetate ester) and 4-vinyl guaiacol (a volatile phenol) along PC2. Groups one and three were spread across PC2 axis, whereas groups two and four were spread across PC1 (Fig. 4b). Several strains were shown to be phenotypically similar by both hierarchical clustering and PCA. One distinguishing characteristic between the groups was the difference in ratios of classes of aroma compounds (Fig. 4c). There are two types of esters in wine: acetate esters and ethyl esters. Variation in the ratios of these two classes of compounds was observed between the groups. Group 1 showed the highest acetate to ethyl ester ratio, whereas Group 3 showed the lowest. Ethyl acetate, the most abundant ester in wine, was excluded from the total acetate ester concentration due to the high concentrations. Included in this study is a distinct group of 14 commercial wine yeast strains commonly known among winemakers and yeast producers to be related; in addition, these strains have been shown to be genetically similar (Dunn et al., 2012). These strains belong to the Prise de Mousse (PDM) family (LVCB, DV10, Elegance, 4F9, Rhone4600, EC1118, QA23, NT116, NT112, NT202, N96, IOC , Pris De Mousse, and Premier Cuvee) and were found to be genetically distinct from other wine yeast strains, suggesting that they may be an ancestral population of closely related wine yeast (Rossouw et al., 2008, 2009, 2010, 2012; Rossouw & Bauer, 2009; Dunn et al., 2012). In this study, the PDM family spreads across the phenotypic landscape with representatives in each of the four clusters; however, the majority (eight of the 14) of the strains are found in cluster one (Fig. 4a). The remaining six strains are spread evenly across the three other clusters. Discussion Continuous selection, over potentially thousands of years, has led to a set of yeast strains optimized for winemaking. An understanding of how these yeasts influence the properties of wine flavor, aroma, and mouthfeel provides the basic knowledge necessary for appropriate strain selection. This study provides a description of the phenotypic differences for 69 yeast strains, including fermentation kinetic differences and the strains metabolic footprints, during the course of a wine fermentation using identical juice. The fermentation time was highly variable, ranging from 5 to 18 days suggesting that there may be differences in the tolerances of yeast to the extreme conditions that are present toward the latter stages of fermentation (high ethanol levels, nutrient limitations, low ph, etc.). However, all strains were eventually able to ferment all consumable sugars, a condition termed dryness in the