We present a method to directly characterize the yeast diversity present in wine fermentations by employing denaturing gradient gel electrophoresis (DGGE) of polymerase chain reaction (PCR)-amplified 26S ribosomal RNA (rRNA) genes. PCR-DGGE of a portion of the 26S rRNA gene was shown to distinguish most yeast genera associated with the production of wine. With this method the microbial dynamics in several model wine fermentations were profiled. PCR-DGGE provided a qualitative assessment of the yeast diversity in these fermentations accurately identifying populations as low as 1000 cells ml−1. PCR-DGGE represents an attractive alternative to traditional plating schemes for analysis of the microbial successions inherent in the fermentation of wine.
Wine fermentations are typically carried out by a complex evolution of microorganisms involving both yeasts and bacteria. Numerous studies have examined the succession of yeasts and bacteria that occurs during the fermentation of non-sterile musts [1–7]. In general, yeasts predominate during the alcoholic fermentation, where the low pH and nutritional content of the juice itself selects for yeast growth. A diverse population of yeasts including species of Kloeckera, Metschnikowia, Candida, Hanseniaspora and Saccharomyces are often present in the initial stages of most wine fermentations . The non-Saccharomyces yeasts typically grow for several days before the fermentation is dominated by one or more Saccharomyces cerevisiae strains along with a concurrent increase in ethanol concentration . The predominance of S. cerevisiae in this setting is a likely result of its high ethanol tolerance as compared with other yeasts present in the wine environment [8,10,11]. Non-Saccharomyces yeasts such as Kloeckera sp., Metschnikowia sp., Candida sp. and Hansenula sp. have been shown to persist in wine fermentations, albeit at a lower level than S. cerevisiae strains [3,12,13].
To date, most studies on the microbial constituents of fermenting wine have employed classical microbiological methods of enrichment and isolation to differentiate the various yeast and bacterial strains [8–10]. Upon isolation, the accumulated strains were characterized taxonomically using traditional microbiological methods  or molecular methods [15–19]. The necessity of enrichment and isolation for strain characterization, in addition to being time-consuming, is fraught with potential biases . While this approach has proven useful, it often fails to characterize those microorganisms for which culturing is problematic or impossible .
Recent advances in the field of molecular microbial ecology have brought forth a variety of new tools to directly assess the microbial diversity present in natural habitats  without the need for microbial enrichment. One common strategy is to sample the DNA (or RNA) of a mixed microbial community and use it as a template for both the assessment of community structure and to reveal individual constituents . Typically, these strategies examine key molecular loci, such as ribosomal RNAs (rRNAs) from which phylogenetic relationships to known microorganisms can be inferred . This tactic has the advantage of eliminating the necessity for strain isolation, thereby negating the potential biases inherent in microbial enrichment. Indeed, studies which have employed such direct analysis have repeatedly demonstrated a tremendous variance between cultivated and naturally-occurring species, thereby dramatically altering perceptions on the true microbial diversity present in various habitats .
Recently, denaturing gradient gel electrophoresis (DGGE) has been employed to differentiate rRNA genes directly purified from complex microbial communities . To date, polymerase chain reaction (PCR)-DGGE has been employed primarily to examine bacterial diversity (or less frequently, fungal diversity) in various natural habitats . In this report we evaluated PCR-DGGE for direct characterization of the yeast population dynamics inherent in laboratory wine fermentations.
Materials and methods
Yeast strains and growth conditions
The yeast strains used in this study: S. cerevisiae (UCD#522), Metschnikowia pulcherrima (UCD#125), Candida ethanolica (UCD#7) and Kloeckera apiculata (UCD#1000) were provided from the Culture Collection of the Department of Viticulture and Enology, University of California, Davis, CA, USA. The strains were grown in YEPD broth (1% yeast extract, 2% bacteriological peptone, 2% glucose) at 30°C.
Fermentations were run in 10-l bench-top fermenters (New Brunswick Scientific Co., Edison, NJ, USA) at a temperature of 25°C. Fermentations were carried out in 6 l of Triple M medium , or Triple M medium containing half the normal concentration of ‘yeast nitrogen base w/o amino acids and ammonium sulfate’ (YNB; Difco Laboratories, Sparks, MD, USA). Triple M medium was employed to allow easy modulation of medium nutrients. Previously we observed alterations in yeast population dynamics by changes in the YNB component of Triple M medium (D.A. Mills, unpublished data). Yeast fermentations were run anaerobically. No attempt was made to control or monitor dissolved oxygen.
One-ml aliquots from individual overnight cultures of S. cerevisiae, K. apiculata, C. ethanolica and M. pulcherrima, grown in Triple M medium, were mixed and used to inoculate the fermentations resulting in a final concentration of about 104–105 cells ml−1 of each strain. During the fermentation, 40-ml samples were collected at least twice a day. Serial dilutions were made in a solution of 8% NaCl, 1% peptone, plated, in duplicate, on WL Nutrient agar (Difco Laboratories) and incubated at room temperature (∼22°C) for 4–5 days. The different yeast strains were enumerated on the basis of the different color and colony morphology . The fermentations were followed for about 16 days. Glucose, fructose and ethanol were assayed in duplicate using enzymatic analysis kits from Boehringer Mannheim (Indianapolis, IN, USA) as described by the manufacturer.
For the isolation of DNA from yeast controls and Triple M medium , DNA was purified using standard procedures  with the following modifications. Depending on the broth cell density, 2–20 ml fermentation broth was centrifuged at 16 000×g for 10 min at 4°C and the cell pellet (approximately 100 μl volume) was resuspended in 200 μl of breaking buffer (2% Triton X-100, 1% SDS, 100 mM NaCl, 10 mM Tris pH 8, 1 mM EDTA pH 8). The cells were homogenized in a bead beater instrument (Fast Prep™, Bio101, USA) with 0.3 g of glass beads (0.5 mm in diameter; BioSpec Products, Bartlesville, OK, USA) in the presence of 200 μl phenol/chloroform/isoamyl alcohol (50:48:2). Two hundred μl TE (10 mM Tris, 1 mM EDTA pH 7.6) was added and the bead/cell mixture was centrifuged for 10 min at 16 000×g at 4°C after which the aqueous phase was collected. The DNA was precipitated with 2.5 volumes of 100% ethanol and centrifuged at 16 000×g at 4°C for 10 min and the pellet washed with 70% ethanol, dried and resuspended in 50 μl of sterile distilled water containing 2 IU RNase (Sigma, USA). The sample was then incubated at 37°C for 30 min before storage at −20°C.
DNA amplification and primers
Approximately 250 nucleotides of the 5′- end region of the 26S rRNA gene was amplified by PCR using the primer NL1, 5′-CGC CCG CCG CGC GCG GCG GGC GGG GCG GGG GCC ATA TCA ATA AGC GGA GGA AAA G-3′ (the GC clamp sequence is underlined)  and a reverse primer LS2, 5′-ATT CCC AAA CAA CTC GAC TC-3′ (corresponding to nucleotide positions 266 to 285 on the S. cerevisiae 26S RNA gene (GenBank accession number M19229)). PCR was performed in a final volume of 50 μl containing 10 mM Tris–HCl, 50 mM KCl, 1.5 mM MgCl2, 0.2 mM each dATP, dCTP, dGTP and dTTP, 0.2 mM of the primers, 1.25 IU Taq-DNA polymerase (Perkin Elmer, Cetus, USA) and 2 μl of the extracted DNA (approximately 10 ng). The reactions were run for 30 cycles: denaturation was at 95°C for 60 s, annealing at 52°C for 45 s and extension at 72°C for 60 s. An initial 5 min denaturation at 95°C and a final 7 min extension at 72°C were used. Products were analyzed on 2% agarose gels containing 0.5 μg ml−1 ethidium bromide, visualized under UV light and photographed with the Multimage™ Light Cabinet (Alpha Innotech Corporation, San Leandro, CA, USA).
The DCode™ Universal Mutation Detection System (Bio-Rad, Hercules, CA, USA) was used for sequence specific separation of PCR products. PCR samples were applied directly onto 8% (w/v) polyacrylamide gels in a running buffer containing 40 mM Tris-acetate, 2 mM Na2EDTA·H2O, pH 8.5 (TAE) and a denaturing gradient from 30 to 60% of urea and formamide. The electrophoresis was performed at a constant voltage of 120 V for 4 h with a constant temperature of 60°C. After electrophoresis the gels were stained in 1.25×TAE containing SYBR Green (reconstituted according to the manufacturers directions; Molecular Probes, Eugene, OR, USA) and photographed under UV transillumination.
Differentiation of wine yeast genera by PCR-DGGE
Numerous studies have demonstrated the use of rRNA sequence analysis for differentiating microbial species [31–33]. Recently a variety of yeast species were differentiated by analysis of partial sequences of the large subunit rDNA . These studies examined the 5′ end of the large subunit rDNA encompassing the D1 and D2 expansion domains . To amplify a portion of this region for the differentiation of wine yeasts by DGGE, we used the previously designed NL-1 forward primer  and a new LS2 reverse primer (see Section 2). These primers amplify a product of approximately 250 bp covering most of the D1 expansion loop . As indicated in Fig. 1, NL-1/LS2 amplification products from a variety of wine-associated yeast genera can be easily differentiated by DGGE however, they do not differentiate different species of the same genera.
PCR-DGGE profile of model fermentations
To evaluate PCR-DGGE for profiling the microbial successions inherent in wine fermentations, we examined mixed culture fermentations of model wine medium inoculated with four yeasts: S. cerevisiae, M. pulcherrima, C. ethanolica and K. apiculata. This mixture of microorganisms was chosen for several reasons. First, each of these genera is routinely isolated from the early stages of most wine fermentations [8,10]. Second, these four species are easily differentiated on WL plating medium thereby facilitating plate counting of each strain from a mixed culture . Third, NL-1/LS2 amplification products for all four species are readily separated by DGGE (Fig. 1).
The initial fermentation employed standard Triple M medium (fermentation A). Plating results (Fig. 2A) indicated that populations of M. pulcherrima, C. ethanolica and K. apiculata increased dramatically within the first 50 h (106 to 107 cfu ml−1) while the growth of S. cerevisiae lagged slightly behind reaching similar levels within 75 h. Populations of M. pulcherrima and K. apiculata began to decline after approximately 100–130 h coinciding with the accumulation of ethanol at 15–22 g l−1. Neither M. pulcherrima nor K. apiculata could be observed on plating medium after 240 h. C. ethanolica was shown to reach the highest concentration (∼107 cfu ml−1) and persisted throughout the fermentation. The population of C. ethanolica only began to decrease with a coincident increase of S. cerevisiae near the end of the fermentation. Interestingly, the available glucose was not completely consumed within 400 h in this fermentation.
As shown in Fig. 2B, the PCR-DGGE profile of fermentation A accurately monitored the presence or absence of specific strains throughout the fermentation. All four strains could be detected during the early stages of the fermentation. Moreover, as populations of M. pulcherrima and K. apiculata began to fall below 103 cfu ml−1, at approximately 240 h into the fermentation, the corresponding DGGE band disappeared (Fig. 2B). In general the DGGE bands of M. pulcherrima and K. apiculata persisted for several hours beyond our ability to detect viable cells of either strain on plating media. This could be due to the cells entering a viable non-culturable (VNC) state or could possibly reflect the presence of M. pulcherrima and K. apiculata DNA in the fermentation as a result of cell lysis.
Previous work in our laboratory indicated that elimination of the YNB component (Difco Laboratories) of Triple M medium resulted in a complete lack of growth of yeast strains (data not shown). This reagent contains a variety of vitamins, trace elements and salts that aid in the propagation of yeast . In an effort to profile an altered microbial growth pattern, an additional fermentation was carried out using modified Triple M medium in which the content of the YNB component was included at one half the normal concentration (fermentation B). As indicated in Fig. 3A, the overall pattern of microbial growth in fermentation B is roughly similar to that in fermentation A. M. pulcherrima, C. ethanolica and K. apiculata grow rapidly to high cell densities (∼107 cfu ml−1) within 50 h and both M. pulcherrima and K. apiculata could not be detected on plating media as the ethanol concentration in the fermentation reached 20–30 g l−1. The pattern of microbial growth is different from fermentation A, however, in that the populations of M. pulcherrima, K. apiculata and C. ethanolica were extinguished more rapidly. M. pulcherrima and K. apiculata became undetectable approximately 50–100 h earlier (respectively) as compared to fermentation A. Moreover, C. ethanolica did not persist throughout 400 h of fermentation, becoming undetectable after ∼340 h. The likely cause for the early exit of M. pulcherrima, C. ethanolica and K. apiculata is the more vigorous ethanol production (most likely produced from S. cerevisiae) in fermentation B. Surprisingly, glucose was completely consumed within 400 h in fermentation B as compared to fermentation A. One possible rationale for this enhanced conversion by S. cerevisiae is the more rapid disappearance of M. pulcherrima, C. ethanolica and K. apiculata thereby reducing competition for limiting nutrients and/or reducing the production of inhibitory compounds by these yeasts.
As seen in Fig. 3B, the DGGE profile clearly monitored the altered microbial population changes that took place in fermentation B. DGGE bands corresponding to K. apiculata and M. pulcherrima disappeared after 138 and 174 h (respectively), coinciding with the inability to observe these strains on plating media (Fig. 3A). The disappearance of C. ethanolica is also clearly represented by the lack of a cognate band after 333 h of fermentation.
Wine fermentations represent an amalgam of microbiological activities derived from both starter and non-starter microorganisms [9,10]. The growth of non-starter (or indigenous) microorganisms can significantly impact the efficacy of starter cultures and even inhibit normal starter function . All previous attempts to characterize the microbial diversity in wine fermentations have employed standard methods of enrichment and isolation to cultivate various microbial constituents before taxonomic identification [8–10,12]. This estimate of microbial diversity is often problematic, however, since many microorganisms may not grow on standard laboratory media. Moreover, microorganisms may enter a VNC state that may be overlooked in standard plating schemes [22,39]. In this work we demonstrate that PCR-DGGE is a viable alternative to standard plating methods for a qualitative assessment of the microbial constituents in model wine fermentations. PCR-DGGE was shown to be quite sensitive, detecting subtle differences in yeast flora development and persistence in two fermentations done under slightly different conditions. In each case there was excellent agreement between the plating data and the presence or absence of a cognate DGGE band.
Given the complex nature of multitemplate PCR, the quantitation of DGGE profiles is problematic [40–43]. Other groups have attempted to quantitate cell populations by comparing DGGE band intensity to internal standards [44,45]. While no attempt was made to quantitate individual yeast populations in this work, DGGE bands were observed for populations as low as 103 cells ml−1. Therefore, the presence of a DGGE band represents a yeast population above a minimum threshold value of 103 cells ml−1 and thus identifies only the predominant yeast populations in these fermentations.
Since PCR-DGGE can be performed in a reasonably rapid fashion (1 day) and with minimal sample volume, this method may be useful for characterizing wine microbial diversity in situations where plating may be impractical, such as in studies involving a large number of fermentation replicates. Indeed, a modified version of the method presented here has been used to obtain discriminatory profiles of yeast populations in several commercial white wine fermentations (D.A. Mills, unpublished results). The use of PCR-DGGE should enable enologists to gain a better view of the microbial diversity present in a range of fermentations without the necessity for problematic plating analysis.
This work was supported by grants from the American Vineyard Foundation (L.F.B., D.A.M.), the California Competitive Grants Program for Research in Viticulture and Enology (L.F.B., D.A.M.) and the UC-Davis New Faculty Research Program (D.A.M.).