Thomas Henick-Kling, Cornell University, Department of Food Science and Technology, New York State Experiment Station, Geneva, NY 14456, USA (e-mail: email@example.com).
To study the impact of yeast populations on wine flavour and to better understand yeast growth dynamics, wines were produced by the (i) indigenous microflora, (ii) vigorous yeast starter EC1118 and (iii) slowly fermenting yeast Assmannshausen. Sensory analysis revealed that wines differed depending on the fermentation type. However, these yeast-related differences did not exceed the varietal character. Both added starter cultures clearly dominated the Saccharomyces population from the middle of fermentation onwards. The starter cultures differed in their repression of indigenous non-Saccharomyces yeast. EC1118 limited growth of non-Saccharomyces yeasts more strongly than Assmannshausen. Sulphite addition further repressed growth of non-Saccharomyces yeasts. On completion, more than one Saccharomyces strain was present in each fermentation, with the largest variety in the non-inoculated and the smallest in the EC1118-inoculated fermentation. Results from the two genetic assays, karyotyping, and PCR using δ-primers were not fully equivalent, limiting the usefulness of δ-PCR in studies of native Saccharomyces yeasts.
Recent work comparing the effects of different starter cultures and indigenous yeasts has shown that there are significant differences in the chemical composition of the resulting wines ( Mora et al. 1990 ; Longo et al. 1992 ; Gafner et al. 1993 ; Lema et al. 1996 ). However, very little has been done measuring sensory differences using taste panels. Studies which simply measure concentrations of aroma-active compounds without considering sensory thresholds and without considering sensory responses by tasters are not satisfactory. Sensory detection thresholds and difference thresholds should be considered when comparing differences in chemical concentrations. Sensorily perceptive differences should be the start of investigations. Some data on the effect of yeast strains have been contradictory, possibly due to the use of grape cultivars with neutral flavour characteristics and because the yeast populations were not stringently analysed ( Lorenzini 1994; Kunkee & Vilas 1994). If the yeast effect on wine flavour is due to its action on flavour precursors (mainly glycosides) from the grape rather than due to flavour-active compounds produced by the yeasts themselves, then different yeasts would produce similar wines from a neutral grape cultivar such as the Thompson’s Seedless used by Kunkee & Vilas (1994).
The study reported here addressed the issue of the dominance of added yeast starter cultures and their effect on the sensory character of the wine. A rapidly fermenting yeast culture was compared with a slowly fermenting one and with a non-inoculated fermentation. The selective effect of SO2 on yeast growth was also considered. The two white grape cultivars, Riesling and Chardonnay, were chosen as the fermentation base as they are known for producing distinctive wines, and their major sensory varietal characteristics arise from different kinds of grape flavour precursors ( Williams et al. 1989 ; Sefton et al. 1993 ; Seung & Noble 1993). Yeast populations during fermentation were followed by plating on selective media and by the genetic analyses, karyotyping and PCR-fingerprinting, which allowed the precise identification and tracking of individual strains during such mixed fermentations. Sensory evaluations of the finished, aged wines were used with yeast population and compositional data to identify correlations and patterns between microbial growth, wine chemistry and sensory attributes.
Materials and methods
Riesling grapes were donated by Lamoreaux Landing winery, Lodi, NY; Chardonnay came from the NY State Agricultural Experiment Station vineyards, Geneva, NY. The musts were treated with 0 and 50 mg l−1 sulphite after pressing. Diammonium hydrogen phosphate was added to the Riesling must at 1 g l−1. After settling at 2 °C for 24 h, the white juices were racked into 19 l glass carboys and inoculated with Assmannshausen (AMH), EC 1118 (EC), or no yeast starter. Both starter cultures are Saccharomyces cerevisiae strains; EC was formerly referred to as Saccharomyces bayanus, ‘Prise de Mousse’, and is sold by Lallemand Inc. unter the name EC 1118. Both starter cultures were donated by Lallemand Inc., Montréal. This point was designated as the initiation of fermentation in non-inoculated musts. Fermentations were carried out at 16 °C. Riesling fermentations with AMH yeast were also carried out at 21 °C to determine temperature effects on the proportionate growth of Saccharomyces and non-Saccharomyces yeasts. Fermentation progress was followed daily by changes in soluble solids (°Brix). After adjusting the sulphite concentration to 30 mg l−1 free SO2, all wines were cold-stabilized at −4 °C for 2 months. Prior to bottling, 0·5 mg l−1 Cu (as CuSO4) was added to all carboys inoculated with EC or AMH yeast to remove traces of hydrogen sulphide odour; no treatment was necessary in the non-inoculated wines. The wines did not undergo malolactic fermentation. All treatments were carried out in duplicate.
Enumeration of yeast populations
Fermenting musts were sampled at several points during fermentation for indigenous and inoculated yeasts by plating on the following selective media: Lysine Agar (LA) for yeasts other than Saccharomyces, and Ethanol Sulphite Agar (ESA) for Saccharomyces yeasts, as described in Henick-Kling et al. (1998) .
DNA isolation was performed as described by Schütz & Gafner (1993). Saccharomyces yeast identification was achieved by genetic karyotyping and by PCR analysis. Chromosomes were separated using CHEF DR® III system (Bio-Rad) as described by Schütz & Gafner (1993). Agarose concentration (Bio-Rad; Molecular Biology Grade) was 1%. Run parameters were: 60 s switch time for 15 h, followed by a 90 s switch time for 9 h at 6 V cm−1 at an including angle of 120°.
PCR reactions were performed in total volumes of either 25 μl, 50 μl or 100 μl. Between 5 and 20 μl yeast cell lysate were used in the reaction, depending on the total volume of 20 μl, 50 μl or 100 μl, and the conditions were 1 × Taq Polymerase buffer (Promega) 1·5 mmol l−1 MgCl2, 200 μmol l−1 dNTPs and 1 μmol l−1 of each primer. PCR reactions were carried out with a Robocycler 40 (Stratagene, La Jolla, CA, USA). Reaction parameters were: 1 cycle, 94 °C for 1 min; 30 cycles, 94 °C for 1 min, 50 °C for 2 min, 72 °C for 1 min; 1 cycle, 72 °C for 5 min. The oligonucleotides δ1 (5′-CAAAATTCACCTATA/ATCTCA-3′) and δ2 (5′-GTGGATTTTTATTCCAACA-3′) were used to amplify genomic DNA between the δ sequences ( Ness et al. 1993 ). Yeast cell lysates were prepared by incubating a loopful of yeast cells in 500 μl lysis buffer (10 mmol l−1 Tris-HCl, pH 9; 50 mmol l−1 KCl; 0·01% gelatine; 1·5 mmol l−1 MgCl2; 0·1% Triton X-100; 25 μg proteinase K; 20 mmol l−1 1,4-dithiothreitol; 12·5 mg l−1 SDS) at 37 °C for 60 min. After heating at 95 °C for 10 min and centrifugation for 2 min, the supernatant fluid was separated from the pellet and kept at −20 °C. Separation of the PCR fragments was performed on a 3% (w/v) NuSieve® 3:1 agarose gel (FMC, Philadelphia PA, USA) using 1 × TAE buffer (40 mmol l−1 Tris acetate, 1 mmol l−1 EDTA, pH 8) containing 200 ng ml−1 ethidium bromide. A 100 bp ladder (Promega) was run alongside the samples as a molecular weight marker. DNA was visualized by u.v. transillumination and processed using the Gel Doc 1000 Video Gel Documentation System (Bio-Rad).
Riesling and Chardonnay wines were analysed for residual sugar, pH, alcohol and browning (A) by standard methods ( Zoecklein et al. 1990 ), organic acids by HPLC ( Martineau et al. 1995 ), and glycerol, ammonia and acetaldehyde by enzymatic methods (Boehringer). Free amino nitrogen (FAN) was determined according to the method of Dukes & Butzke (1997).
Six months after the end of alcoholic fermentation, all wines were evaluated by a panel of 11 experienced tasters. All panellists (male:female ratio was 8:3) were experienced in sensory testing of wines and included oenologists, viticulturists, sensory scientists and professional wine-makers. The wines were first compared in a blind tasting in which the panellists developed a list of descriptors. Sampling was done in randomized complete blocks, each treatment replicate being sampled twice by each panellist. Sensory characteristics were quantified on a 1–10 scale using descriptors developed through Free Choice Profiling ( Meilgaard et al. 1991 ). The resulting scores were normalized and compared by difference tests, and their significant differences were following analysis of variates using SAS System software.
Variation of fermentation length and yeast growth depending on starter culture, sulphite treatment and temperature
Riesling and Chardonnay grape musts with either no, or 50 mg l−1 sulphite added at pressing were fermented (i) by the indigenous microflora without added Saccharomyces yeast starter cultures, (ii) with the addition of the slowly fermenting yeast starter Assmannshausen, and (iii) with the addition of the fast fermenting yeast starter EC1118. The lag phases (i.e. the time between racking and start of the alcoholic fermentation) and total length of fermentation varied depending on the grape must, type of the used yeast, sulphite addition and fermentation temperature. Overall, fermenting Chardonnay grape musts took between 15 and 21 d longer than Riesling, with fermentations considered to be completed when residual sugar concentration was below 5 g l−1 ( Fig. 1). Non-inoculated musts fermented in 32 d for Riesling and 47 d for Chardonnay, inoculation with AMH decreased fermentation times to 19 d for Riesling and 40 d for Chardonnay, whereas fermentations using EC finished in 13 d for Riesling and 34 d for Chardonnay ( Fig. 1).
Sulphite addition had the greatest effect when used in non-inoculated musts, reducing total fermentation time by as much as 10 d. Sugar content decreased slightly during the lag phase of non-sulphited musts but stayed unchanged when sulphite was added. This effect correlated with higher populations of approximately 107 cfu ml−1 of non-Saccharomyces strains in unsulphited Riesling musts, in contrast to must treated with 50 mg l−1 sulphite in which they were reduced between 10- and 100-fold ( Fig. 2). However, in musts inoculated with AMH or EC, utilization of sugar during the initial part of fermentation was not affected by sulphite addition. In general, decline in non-Saccharomyces yeast cell numbers correlated with Saccharomyces yeast growth and ethanol accumulation ( Fig. 2).
It has been reported from experiments with synthetic media ( Sharf & Margalith 1983) and filter-sterilized media ( Heard & Fleet 1988) that temperature may affect the growth of various yeasts differently. To test for the effect of fermentation temperature in a non-sterile grape must containing an indigenous microflora, Riesling musts were inoculated with AMH and fermented at 21 °C and 16 °C, respectively, with and without added sulphite ( Fig. 3). Sulphite addition did not affect the fermentation rate at 21 °C but at 16 °C, fermentation in the non-sulphited must took 1 week longer than in the sulphited must. Without SO2 and at the lower temperature, the ratio of non-Saccharomyces to Saccharomyces yeasts was lower. In the non-sulphited must at 16 °C, the number of Saccharomyces yeasts was approximately 20 times lower than the non-Saccharomyces yeast population. Saccharomyces yeasts dominated only with added sulphite and at the higher temperature. Sulphite additions at 16 °C and 21 °C suppressed growth of non-Saccharomyces yeasts.
Decreased diversity of Saccharomyces yeast strains with the use of starter cultures
Saccharomyces yeast colonies from the middle and end of Riesling and Chardonnay fermentations were taken from ESA agar plates for identification in order to obtain information about diversity in the Saccharomyces yeast population and dominance of the inoculated yeast starter cultures ( Fig. 4). In total, 141 Saccharomyces yeast isolates were analysed by genetic karyotyping and PCR-based fingerprinting using the primers δ1 and δ2. Figure 4(a) shows all the different patterns which were found by electrophoretic karyotyping. The difficulty of discriminating between different patterns was solved by dividing the lanes into four different windows, and counting the bands and comparing their sizes in the same windows between different lanes. Data are summarized in Fig. 4(b). Twenty different Saccharomyces yeast strains were found in the fermentations. The largest variety of seven different strains was present in the spontaneous fermentations and six strains in the fermentations inoculated with AMH. The number of indigenous Saccharomyces was reduced to three strains in the EC-inoculated musts. In all fermentations, the diversity decreased towards the end of fermentation. Besides the starter cultures AMH and EC, each fermentation had its own unique strain spectrum of indigenous Saccharomyces. The non-inoculated fermentations showed presence of EC and AMH yeasts, most probably due to contamination. The inoculated yeast strains clearly dominated with 60% and over 90% in the AMH-, and over 70% and 80% in the EC-inoculated fermentations. In contrast, no clear strain dominance was observed in the non-inoculated fermentation. Whereas strain O4 predominated with over 30% in the middle of fermentation, it had decreased by the end of fermentation to less than 20%.
Figure 5 shows the δ-PCR products of the same yeast colonies. There was no perfect correlation between these two methods; some colonies that clearly differed in the chromosomal pattern of karyotyping were discriminated only by weak bands in the PCR assay as the only discrimination criterion. Moreover, for the three strains E5, D5 and K4, δ-PCR patterns were unique ( Fig. 5), even if reaction conditions were substantially modified (data not shown). This result indicates that the δ-PCR assay may fail to discriminate different yeast strains.
Grape cultivar-specific flavour effects of starter cultures
Wines from the three different fermentation types were sensorily compared by a trained taste panel. Interestingly, a very similar set of descriptors was chosen in describing the wines produced from the same grape variety ( Table 1). In the case of Riesling, comparison of the scores for the wines revealed significant differences for the descriptors ‘acetic’, ‘body’, ‘diacetyl/caramelized’, ‘overall fruitiness’, ‘melon’, ‘paper/ cardboard’, ‘pineapple/tropical’ and ‘spicy’ ( Table 1). The wines from non-inoculated fermentations had the majority of high scores for these descriptors, while those fermented with EC had the lowest scores ( Fig. 6). AMH-inoculated wines were generally between these extremes, but showed high scores for ‘overall fruitiness’ and ‘acetic’, and lowest scores for ‘diacetyl’ and ‘paper’. For Chardonnay wines, the use of different starter cultures produced significant differences for the descriptors ‘astringent/phenolic’, ‘floral’, ‘herbaceous/ vegetative’, ‘reduced sulphur/H2S’, ‘oxidized’, ‘pear’ and ‘sweaty’ ( Table 1). As observed in the Riesling fermentations, the highest scores for most flavour attributes were given to wines from non-inoculated fermentations. These wines were characterized by floral and pear aromas and by strong vegetative, reduced and sweaty aromas. Wines fermented with AMH tended to have more fruity flavours (‘floral’ and ‘pear’), and EC wines were less fruity and had stronger oxidized, astringent, and some sweaty and herbaceous flavours.
Table 1. Descriptors selected through Free-Choice Profiling for all wines
Descriptors which had significant differences between inoculation treatments (P Ð0·05) are in bold.
Statistical analysis revealed that the descriptors which were significantly different among Riesling wines were all different from those which described significant differences in Chardonnay wines ( Table 1), suggesting that the different yeasts affected the flavour compounds variety-specifically. Nevertheless, yeast strain characteristics were also observed ( Fig. 6). AMH tended to be strongly fruity and spicy in the Riesling, and floral and pear-like in the Chardonnay. EC produced the least fruity wines in this comparison. These wines tended to have ‘aldehydic/phenolic, paper, oxidized and astringent flavours. In non-inoculated wines, most flavour attributes, positive and negative, tended to be highest.
This comprehensive study focused on the microbiological and sensory effects of different vinification parameters such as grape must, sulphite addition and fermentation microflora (inoculated and non-inoculated). As observed in many local wineries, Riesling mainly differed from Chardonnay grape juice in that Riesling fermentations were completed faster. However, the difference of between 15 and 21 d, even in the fermentations inoculated with the vigorous starter culture EC, was a surprising result and is probably due to diammonium hydrogen phosphate as a nitrogen source added to the Riesling but not to the Chardonnay must in this study. Sulphite exhibits many different effects, as summarized by Fleet & Heard (1992). In the present experiments, sulphite addition to the must prior to fermentation shortened fermentation time in all cases except for the fermentations with EC, where no difference between sulphited and unsulphited must was detected. It is likely that this time difference is due to repression of non-Saccharomyces yeasts and selective growth of sulphite-tolerant Saccharomyces yeasts. It can be assumed that the reason for no effect in the fermentations with EC was that the vigorous EC starter culture restricted growth of other yeasts by itself. Generally, the sulphite was restrictive rather than selective for non-Saccharomyces growth, which was reduced 10- to 100-fold by sulphite addition. This is in agreement with the earlier findings that growth of indigenous yeasts is suppressed but not limited by sulphite addition (summarized in Fleet & Heard 1992). The fact that the amount of AMH yeasts did not exceed 106 cfu ml−1 was not reproduced in other years. However, it was usually found that maximum cell numbers were as much as 10-fold lower than EC populations. As a conclusion, with fast growing, rapidly dominating yeasts at moderate temperature (16 °C), sulphite additions to must can be omitted. However, considering sensory attributes, previous work in this group has shown that there are differences in wines produced without and with (0, 10, 20, 50 mg l−1) sulphite additions to the must ( Carrasco et al. 1994 ; Henick-Kling et al. 1998 ). In these studies, the wines fermented without any sulphite were consistently less fruity with more fermentation off-odours.
The results with AMH- and the indigenous non-Saccharomyces yeast growth at two different temperatures generally match the observations from inoculations with various yeasts by Heard & Fleet (1988) and Sharf & Margalith (1983). They suggest that temperature is a likely parameter for controlling different yeasts during fermentation. Viability of Saccharomyces yeast was much better (at least a factor of 10) at the higher temperature of 21 °C, irrespective of added sulphite, whereas non-Saccharomyces yeasts were reduced approximately 10-fold if 50 mg l−1 sulphite was present. As described earlier ( Fleet & Heard 1992), sulphite in combination with Saccharomyces yeast growth rather than on its own leads to a decrease in non-Saccharomyces yeast growth. This is probably due to ethanol producing activity. This study shows for the first time that different Saccharomyces starter cultures direct growth of indigenous non-Saccharomyces yeasts to different degrees. The starter AMH allowed a much larger population of non-Saccharomyces to develop than starter EC. The extent of the growth of non-Saccharomyces yeasts depended on the fermentation temperature and on the use of SO2 in the preparation of the must.
Genetic identification allowed us to follow the succession of strains of Saccharomyces yeasts in the inoculated and non-inoculated fermentations. Our results confirm previous studies that show a succession of various strains of Saccharomyces ( Frezier & Dubourdieu 1992; Schütz & Gafner 1993). We also observed a decrease in the number of different Saccharomyces strains in fermentations with added starter cultures. It is important to realize that added Saccharomyces starter cultures do not completely dominate a fermentation. As our results show, different starter cultures can dominate to different degrees, allowing a different number of indigenous Saccharomyces strains to grow as well. It was surprising that there were even two strains other than EC at the end of the EC-inoculated fermentation, accounting for almost 20% of the population. Based on the strong growth and fermentation properties of this starter culture, it could be assumed that no other yeast strain was able to persist through to the end of fermentation. Each fermentation type selected a different mix of indigenous yeasts, indicating that the choice of a starter culture creates unique populations. This was unexpected as it was assumed that besides the added starter cultures, the indigenous yeast population would be the same in all Riesling musts on the one hand, and in all Chardonnay musts on the other. It is therefore speculated that nutritional and chemical competition plays an important role in inhibiting some yeast strains and allowing for growth of others. The appearance of EC and AMH yeasts in the non-inoculated fermentations, and of EC yeasts in the AMH inoculated fermentations, is most likely due to cross-contamination during the fermentation process, as EC and AMH yeasts were not originally present in the indigenous microflora. Even though great care was taken during the experimental wine-making not to carry yeasts from one fermentation vessel to another when checking the progress of fermentation, this contamination shows how easily yeasts are transferred from one vat to another under winery conditions.
As karyotyping may be too complex and too laborious for analysing yeast populations during winery fermentations ( Grando et al. 1994 ), PCR fingerprinting was also applied in this study using the previously described δ primers ( Ness et al. 1993 ; Coakley & Ross 1996). PCR with these primers worked for some but not all strains found in the indigenous populations. PCR agreed with differences in karyotype patterns for many Saccharomyces strains, but some strains (15%) were distinct based on the karyotype but could not be differentiated by the δ primer fingerprints. Vezinhet et al. (1994) reported about 80%δ-PCR reliability, which is in the same range as the present studies. It can therefore be argued that this method is not completely satisfactory for studying indigenous Saccharomyces yeast populations during wine-making, as there is a risk of different strains being identified as the same. However, this method can work satisfactorily when following known strains which differ in their δ fingerprint pattern. Alternatively, PCR-based discrimination as described ( Grando et al. 1994 ; Quesada & Cenis 1995; Barros Lopes et al. 1996 ; Laidlaw et al. 1996 ) could be evaluated for the purpose of studying wine yeast populations.
Higher scores in aroma description indicate greater intensity of a particular aroma rather than the desirability of a specific flavour. The wines from non-inoculated fermentations had the highest scores and those fermented with EC, the lowest. Therefore, the non-inoculated wines could be described as the most aroma intense. The quality of a wine, however, is determined by the balance of flavours and how they compliment each other, rather than by aroma intensity. Therefore, depending on the preference of the taster, the overall quality of a non-inoculated wine can be rated as very high or very low. Non-inoculated wines were characterized by large populations of non-Saccharomyces whereas the wines inoculated with EC had the smallest populations of non-Saccharomyces. Thus, large populations of nonSaccharomyces were correlated with more intense positive and negative flavour attributes. The most interesting result from the sensory evaluation was that the flavours that varied depending on the fermentation type did not overlap at all for Riesling and Chardonnay wines. Even though some of the same flavours were recognized in both Riesling and Chardonnay wines, each of the inoculations produced a different set of significantly different flavours. This suggests that yeasts, although having a considerable effect on the final flavour, do not change varietal grape character. Combining the findings of the sensory evaluation with the microbiological data from this study, the preference of many wine-makers for low fermentation temperatures in Riesling production with the intention of enhancing fruity flavours may be based directly on the enhancement of growth of non-Saccharomyces yeasts. Temperature regulation is obviously an important tool in wine-making as it shifts the population of non-Saccharomyces and Saccharomyces yeasts. Low fermentation temperatures in non-inoculated fermentations are risky because of the high probability that non-Saccharomyces yeasts with the potential for causing strong off-flavours will become dominant under these conditions.
This research was made possible by grants from the New York Wine and Grape Foundation, from Lallemand Inc., Montréal, the Swiss National Foundation, and US Federal and New York State agricultural research funds. The authors thank Lamoreaux Landing Wine Cellars, Lodi, NY, and Dr Robert M. Pool, New York State Agricultural Experiment Station, Geneva, NY, for generously providing the grapes for the experiments.