Yeast Diversity and Persistence in Botrytis-Affected Wine Fermentations

Abstract

Culture-dependent and -independent methods were used to examine the yeast diversity present in botrytis-affected (“botrytized”) wine fermentations carried out at high (∼30°C) and ambient (∼20°C) temperatures. Fermentations at both temperatures possessed similar populations of Saccharomyces, Hanseniaspora, Pichia, Metschnikowia, Kluyveromyces, and Candida species. However, higher populations of non-Saccharomyces yeasts persisted in ambient-temperature fermentations, with Candida and, to a lesser extent, Kluyveromyces species remaining long after the fermentation was dominated by Saccharomyces. In general, denaturing gradient gel electrophoresis profiles of yeast ribosomal DNA or rRNA amplified from the fermentation samples correlated well with the plating data. The direct molecular methods also revealed a Hanseniaspora osmophila population not identified in the plating analysis. rRNA analysis also indicated a large population (>10⁶ cells per ml) of a nonculturable Candida strain in the high-temperature fermentation. Monoculture analysis of the Candida isolate indicated an extreme fructophilic phenotype and correlated with an increased glucose/fructose ratio in fermentations containing higher populations of Candida. Analysis of wine fermentation microbial ecology by using both culture-dependent and -independent methods reveals the complexity of yeast interactions enriched during spontaneous fermentations.

Numerous studies have examined the succession of yeasts and bacteria that occurs during the fermentation of nonsterile musts (15, 16). In general, yeasts predominate during the alcoholic fermentation, where the low pH and nutritional content of the juice select for yeast growth. Several aspects of the microbial ecology present in wine fermentations warrant investigation. Foremost is the fact that indigenous non-Saccharomyces yeasts and indigenous bacteria are potential causes of stuck and sluggish wine fermentations (3). Winemakers often increase this possibility by seeking to reduce the overall amount of the sulfur dioxide (SO₂) used in winemaking, which is employed, in part, to eliminate indigenous microbial populations. In addition, there has been a renewed interest in defining the microbial dynamics of spontaneous (uninoculated) fermentations in order to better understand and control fermentation behavior and its subsequent impact on wine flavor (13). Recently several groups have examined various non-Saccharomyces yeasts as potential adjuncts (or alternatives) to Saccharomyces cerevisiae in an effort to modify wine flavor and improve product quality (23, 24, 36).

A diverse population of yeasts, including species of Hanseniaspora (anamorph Kloeckera), Metschnikowia, Candida, Pichia, and Kluveromyces, are often present in the initial stages of most wine fermentations (16). These non-Saccharomyces yeasts typically grow for several days before the fermentation is dominated by one or more S. cerevisiae strains and a concurrent increase in ethanol concentration occurs (4). Candida and Hanseniaspora species have been shown to persist throughout wine fermentations, albeit at a lower level than S. cerevisiae strains (26, 27). Growth or persistence of individual yeast species within wine fermentations is most likely determined by differential sensitivities to temperature, ethanol, and sulfur dioxide as well as a number of other factors (15). Lower fermentation temperatures (between 10 and 20°C) have been shown to encourage growth and/or persistence of Kloeckera and Candida species (25), most likely due to increased ethanol tolerance of these yeasts at lower temperatures (20). Recently, Kloeckera and Candida were shown to possess growth rates comparable to that of S. cerevisiae at lower temperatures (10°C) (7).

Sweet white wines are commonly made from grapes infected with Botrytis cinerea (noble rot). Infection of the grape with B. cinerea results in concentration of grape sugar, which gives must from botrytis-affected (“botrytized”) grapes a characteristically high initial sugar content. Relatively few studies have examined the microbial diversity within botrytis-affected wine fermentations. Most have noticed a significant increase in weakly fermentative yeast (such as Kloeckera and Candida species) and acetic acid bacterial populations compared to those in fermentations of non-botrytis-affected musts (12, 18, 28). During the fermentation of botrytis-affected musts, the indigenous bacteria and non-Saccharomyces yeast populations decrease as Saccharomyces species dominate (18, 28). By their nature sweet wines possess residual sugar, and thus fermentations are prematurely stopped, often by judicious use of SO₂. Recently the production of gluconic acid, 5-oxofructose, and dihydroxyacetone by the acetic acid bacteria in botrytis-affected musts was shown to reduce the effective concentration of SO₂, making these wines more difficult to stabilize against further microbial growth (2).

Relatively few studies have employed direct (culture-independent) methods for determination of viable yeast and bacterial populations. Most have developed PCR or probe techniques to directly assay wine samples for specific bacterial or yeast populations (21, 37, 38). Millet and Lonvaud-Funel (31) employed epifluorescence to directly identify viable but nonculturable bacterial populations in wine. Given that persistence of metabolically active but nonculturable populations of yeasts in wine fermentations may affect fermentation performance (as well as final product flavor), a better understanding of these populations is critical. We have previously developed methods for direct analysis of yeasts present in wine fermentations by using denaturing gradient gel electrophoresis (DGGE) of ribosomal DNA (rDNA) amplicons (10, 11). In this work both culture-dependent (plating) and culture-independent (PCR-DGGE and reverse transcription-PCR [RT-PCR]-DGGE) methods were employed in order to characterize the impact of temperature on the yeast diversity in commercial sweet white wine fermentations.

MATERIALS AND METHODS

Wine fermentations.

Dolce wine fermentations (Dolce Winery, Oakville, Calif.) were carried out with 1999 Napa Valley Semillon grape juice. Grapes were spray inoculated with a stock B. cinerea strain (anamorph of Botryotinia fuckeliana) approximately 35 days prior to harvest. Vino Super liquid pectinase (DSM Food Specialties, Delft, The Netherlands) at 34 ml/ton and Color Pre (Scott Labs, Petaluma, Calif.) at 73 ml/ton were added to the press pan. After pressing, the juice was clarified with polyvinyl polypyrrolidone (International Specialty Products, Wayne, N.J.) at 0.25 lb/1,000 gal and Bentonite (Great Western Chemicals, Bakersfield, Calif.) at 8 lb/1,000 gal. The juice was treated with 300 mg of lysozyme (Scott Labs) per liter and potassium metabisulfite to achieve 94 mg of SO₂ per liter. A total of 3.3 g of tartaric acid per liter was added to adjust the pH. The pH and titratable acidity of the juice after pressing were 3.57 and 6.3 g/liter, respectively. Dolce fermentations were carried out by indigenous yeasts in four new French oak barrels (Seguin-Moreau Cooperage, Napa, Calif.). Two of the barrels were held at ambient cellar temperature (approximately 18°C). Two barrels were radiantly heated until the temperature of the fermenting juice reached 28°C. Samples were aseptically removed from each barrel immediately following batonage. One milliliter of each sample, in duplicate, was centrifuged at 2,000 × g for 5 min, washed in 1 ml of 4°C water, recentrifuged at 2,000 × g for 5 min, and frozen at −50°C for later PCR-DGGE analysis. Cell pellets used for RT-PCR-DGGE analysis were immersed in 300 μl of RNAlater (Ambion Inc., Austin, Tex.) prior to freezing. One milliliter of each sample was sterile filtered through a 0.45-μm-pore-size Millex-HV filter (Millipore S.A., Molsheim, France) and frozen at −50°C for later high-pressure liquid chromatography (HPLC) analysis. Aseptic measurements of wine temperature, air temperature, and densitometric soluble solids were performed daily on one barrel from each treatment.

Monoculture fermentations of S. cerevisiae and Candida sp. strain EJ1 were carried out in 250 ml of Chardonnay juice which was sterile filtered through a 0.45-μm-pore-size Millex-HV filter. Fermentations were initiated with a 0.1% inoculation. One-milliliter samples were removed daily, sterile filtered through a 0.45-μm-pore-size Millex-HV filter, and frozen at −20°C for later analysis.

Microbiological characterization.

Samples were plated in duplicate on Wallerstein laboratory nutrient agar (WLN) and lysine medium agar (LM) (Difco Laboratories, Detroit, Mich.). Colony morphotypes were differentiated visually as described previously (6) and counted. Several isolates (n = 6 to 8) of each colony morphotype were saved at 4°C for sequence analysis and to serve as DGGE controls.

Nucleic acid extraction.

For the DNA preparation, the cell pellet samples were resuspended in 1 ml of an 8-g/liter NaCl solution and transferred to a microcentrifuge tube containing 0.3 g of 0.5-mm-diameter glass beads (BioSpec Products Inc., Bartlesville, Okla.). The cell-bead mixture was centrifuged at 11,600 × g for 10 min at 4°C, and the supernatant was discarded. The cell-bead mixture was resuspended in 300 μl of breaking buffer (2% Triton X-100, 1% sodium dodecyl sulfate, 100 mM NaCl, 10 mM Tris [pH 8], 1 mM EDTA [pH 8]) and 300 μl of phenol-chloroform-isoamyl alcohol (50:48:2). The cells were then homogenized in a bead beater instrument (Fast Prep; Bio 101, Vista, Calif.) three times for 45 s each at a speed setting of 4.5. The mixture was then centrifuged at 11,600 × g for 10 min at 4°C, and the aqueous phase was removed to another microcentrifuge tube. The DNA was then further purified by using the DNeasy Plant minikit (Qiagen, Valencia, Calif.) according to the manufacturer's instructions.

RNA was extracted by using the Concert Plant RNA reagent (Invitrogen, Carlsbad, Calif.). The cell pellet was resuspended in 500 μl of the reagent and vortexed at maximum speed for 30 s. After 5 min at room temperature, the suspension was centrifuged at 11,600 × g for 2 min and the supernatant was transferred to a new microcentrifuge tube. One hundred microliters of 5 M NaCl and 300 μl of chloroform were added, vortexed, and centrifuged at 11,600 × g for 10 min at 4°C, and the aqueous phase was removed to another microcentrifuge tube. RNA was precipitated with 500 μl of isopropanol and centrifugation at 11,600 × g for 10 min at 4°C. The RNA pellet was rinsed with 70% ice-cold ethanol, dried under vacuum at room temperature, and resuspended in 50 μl of RNase-free water. RNA samples were treated with RNase-free DNase (Roche Diagnostics, Indianapolis, Ind.) at 37°C for a minimum of 1 h to remove coextracted DNA.

PCR amplification.

For colony morphotype identification, the D1-D2 region of the 26S rRNA gene was amplified by PCR with primer NL1 (5′-CGCCCGCCGCGCGCGGCGGGCGGGGCGGGGGCCATATCAATAAGCGGAGGAAAAG-3′) (the GC clamp sequence is underlined) and the reverse primer NL4 (5′-GGTCCGTGTTTCAAGACGG-3′) (29). Colony PCR was performed in a final volume of 100 μl containing 10 μl of 10× PCR buffer (Promega Corp., Madison, Wis.); 1.5 mM MgCl₂; 0.1 mM (each) dATP, dCTP, dGTP, and dTTP; 0.1 mM primers; 1.25 U of Taq DNA polymerase (Promega); and approximately 1 mg of whole yeast cells. The reactions were run for 30 cycles; denaturation was at 95°C for 60 s, annealing was at 52°C for 45 s, and extension was 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. For DGGE analysis of fermentation samples, primers NL1 and LS2 (5′-ATTCCCAAACAACTCGACTC-3′) (10) were used for PCR amplification. PCR was performed in a final volume of 50 μl containing 5 μl of PCR buffer; 2.0 mM MgCl₂; 0.2 mM (each) dATP, dCTP, dGTP, and dTTP; 0.2 mM primers, 1.25 U of Taq DNA polymerase (Promega); and 2 μl of the extracted DNA (approximately 20 ng). The reactions were run for 30 cycles; denaturation was at 95°C for 60 s, annealing was at 52°C for 45 s, and extension was 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. RT-PCR was performed with RevertAid Moloney murine leukemia virus reverse transcriptase (Promega). One microliter of total RNA (approximately 0.1 μg) was mixed in 10 μl of DNase- and RNase-free sterile water containing 0.5 μg of primer LS2 and incubated at 70°C for 5 min. Immediately after chilling in ice, a mixture of 25 mM Tris-HCl (pH 8.3), 25 mM KCl, 2 mM MgCl₂, 5 mM dithiothreitol, a 1 mM concentration of each deoxynucleoside triphosphate, and 20 U of RNase inhibitor (Roche) was transferred in the reaction tube. After 5 min at 37°C, 1 μl of reverse transcriptase was added, followed by incubation at 42°C for 60 min and treatment at 70°C for 10 min to stop the reaction. Three microliters of the synthesized cDNA was used for the PCR as described previously. Products were analyzed by standard agarose gel electrophoresis (1), stained with 0.5 μg of ethidium bromide per ml, visualized under UV transillumination, and photographed with a Multimage light cabinet (Alpha Innotech Corporation, San Leandro, Calif.).

DGGE analysis.

The DCode universal mutation detection system (Bio-Rad, Hercules, Calif.) was used for sequence-specific separation of PCR products and for the comparison of migrations of isolate PCR products. PCR samples were applied directly onto 8% (wt/vol) polyacrylamide gels in a running buffer containing 40 mM Tris-acetate-2 mM Na₂EDTA · H₂O (pH 8.5) (TAE) and a denaturing gradient from 20 to 60% of urea and formamide. The electrophoresis was performed at a constant voltage of 120 V for 6 h with a constant temperature of 60°C. After electrophoresis, the gels were stained in 1.25× TAE containing SYBR Gold (reconstituted according to the directions of the manufacturer [Molecular Probes, Eugene, Oreg.]) and photographed under UV transillumination. Bands of interest were excised directly from the gels by using a sterile blade, mixed with 40 μl of water, and incubated overnight at 4°C. Two microliters of this solution was used to reamplify the PCR product with the NL1-LS2 primer pair (11).

RNA hybridization.

A probe specific to Candida sp. strain EJ1 was generated by amplifying a portion of the D2 region of the 26S rDNA gene with the universal primer NL1 (29) and a primer, C1 (5′-TACCGCATTTATCTTCCCCC-3′), internal to the 26S rRNA D1 loop of Candida sp. strain EJ1 (L. Cocolin and D. A. Mills, submitted for publication). DNA was amplified in a 50-μl final volume containing 10 mM Tris-HCl (pH 8), 50 mM KCl, 1.5 mM MgCl₂, 0.2 mM deoxynucleoside triphosphates, 0.2 μM primers, 1.25 U of Taq polymerase (Promega), and 5 μl of extracted Candida sp. strain EJ1 DNA (about 10 to 50 ng of total DNA). The PCR cycle parameters were 30 cycles of denaturation at 95°C for 1 min, annealing at 60°C for 45 s, and extension at 72°C for 1 min. An initial denaturation at 95°C for 5 min and a final extension at 72°C for 7 min were also used. The amplified product (156 bp) was labeled by incorporating digoxigenin-UTP (Roche) into the PCR mixture as described by the manufacturer.

A total of 1 μg of RNA purified from select fermentation samples was applied to Zeta-probe GT membranes (Bio-Rad) by using a BioDot slot blot apparatus (Bio-Rad) as indicated by the manufacturer. Hybridization and detection were performed by using a digoxigenin chemiluminescence kit (Roche) as indicated by the manufacturer. Control RNA samples were purified from active cultures of S. cerevisiae and Candida sp. strain EJ1.

Sequence analysis.

The PCR products from the colony PCR of the isolates were purified by using a Wizard PCR purification kit (Promega) and then sent to a commercial sequencing facility (Davis Sequencing, Davis, Calif.) for sequencing. Reamplified DGGE bands were similarly purified and sequenced. Sequence compilation and comparison were performed by use of Genetics Computer Group sequence analysis software with the BLAST program.

HPLC analysis.

The samples and standards were run on a Hewlett-Packard 1100 series instrument with a cation H guard column, two 30-cm Aminex HPX-87H columns (Bio-Rad), and a Hewlett-Packard 1047A refractive index detector. The column was eluted with 1.5 mM sulfuric acid at a flow rate of 0.6 ml/min and a column temperature of 50°C. An external standard solution at three dilution levels was used. The variabilities in the fructose and glucose assays were 2.1 and 2.2%, respectively.

Nucleotide sequence accession number.

The partial 26S rDNA sequence of Candida sp. strain EJ1 was deposited in GenBank under accession number AY078348.

RESULTS

Fermentation characteristics.

There were distinct differences in fermentation characteristics for the treatments at ambient and heated temperatures (Fig. 1 and 2, respectively). Day 0 represents the day that the wine was put into barrels. The relatively small change in wine temperature (5°C) in the ambient-temperature treatment throughout the fermentation contrasted with the large change in wine temperature (15°C) in the heated treatment. The rates of sugar consumption, ethanol production, and glycerol production were higher in the heated fermentations than in the ambient-temperature fermentations (Fig. 1 and 2). The average maximum rate of glucose and fructose consumption was 41 g/liter/day for the heated treatment versus 21 g/liter/day for the ambient-temperature treatment. The average maximum rate of ethanol production in the heated treatment was 19 g/liter/day, versus 8 g/liter/day in the ambient-temperature treatment, and the average maximum production rate of glycerol was 2.6 g/liter/day in the heated treatment versus 0.7 g/liter/day in the ambient-temperature treatment. Minimal differences in acetic acid formation were noted between treatments. In contrast, the relative proportions of glucose and fructose varied between the ambient-temperature and heated fermentations. At the beginning of both fermentations, the glucose/fructose ratio was 0.92, and during the maximum rate of fermentation, it approached 1.00. After 21 days, the glucose/fructose ratio in the ambient-temperature barrels was an average of 0.91, while, in the heated barrels, it was 0.71.

FIG. 1.

Yeast, chemical, and temperature profiles for the Dolce fermentation carried out at ambient temperature. Fermentations were carried out in duplicate barrels, and a representative data set is shown.

FIG. 2.

Yeast, chemical, and temperature profiles for the Dolce fermentation carried out at high temperature. Fermentations were carried out in duplicate barrels, and a representative data set is shown.

Yeast population dynamics.

In both treatments, six distinguishable yeast isolates were identified on WLN medium (6). Partial 26S rDNA sequence analysis (29) of representative isolates indicates the six morphotypes to be S. cerevisiae, Hanseniaspora uvarum, Pichia kluyveri, Metschnikowia pulcherrima, Kluyveromyces thermotolerans, and a Candida strain (herein called Candida sp. strain EJ1) which could not be assigned to a known species. In the early stages of the fermentations WLN medium was used for enumeration of the six morphotypes. However, as the fermentation progressed, lower populations of non-Saccharomyces species became increasingly difficult to enumerate in the presence of a high population of Saccharomyces. Therefore, at later stages in the fermentation LM (30) was also employed, starting at day 12 in the ambient-temperature treatment and day 8 in the heated treatment. Only two colony morphotypes were observed on LM at these latter stages. Representative isolates were restreaked on WLN medium. The resultant morphologies on WLN medium were shown to be consistent with Candida sp. strain EJ1 and K. thermotolerans isolates.

The CFU plating results indicate some common trends among the two treatments. In all four barrels only six prominent morphotypes were identified. Moreover, within each treatment the six morphotypes reached the same approximate maximum CFU population. Several prominent differences were also observed in the plating results obtained for the two treatments (Fig. 1 and 2). As expected, Saccharomyces populations reached a maximum density of 10⁷ CFU per ml faster in the heated treatment (5 days) than in the ambient-temperature treatment (7 days). In addition, non-Saccharomyces yeasts persisted far longer in the ambient-temperature barrel fermentations than in the heated barrel fermentations. In the ambient-temperature treatment Candida populations remained quite high, around 10⁷ CFU per ml, for much of the fermentation (21 days) and for nearly 18 days after establishment of an equal population of Saccharomyces. A significant drop in the Candida sp. was observed only near the end of the ambient-temperature fermentation, as the ethanol concentration progressed above 100 g/liter. Like Candida, Kluyveromyces populations persisted throughout the ambient-temperature fermentation, although they did so at a lower level (fewer than 10⁶ CFU/ml) and dropped consistently around 3 days after the establishment of a dominant Saccharomyces population and as the ethanol concentration progressed above 60 g/liter. Several other non-Saccharomyces yeasts present in the initial stages of the ambient-temperature fermentation did not persist throughout the fermentation, including Pichia (present for ∼6 days), Hanseniaspora (present for ∼8 days), and Metschnikowia (present for ∼10 days) spp.

The heat treatment had a significant impact on the non-Saccharomyces yeasts, with most populations becoming undetectable on plating medium earlier than in the ambient-temperature treatment. On WLN medium, the non-Saccharomyces yeasts ceased to appear after approximately 7 days with Pichia observed for only 2 days, Hanseniaspora for 5 days, Metschnikowia for 5 days, Candida for 6 days, and Kluyveromyces for 7 days. After day 7, significantly lower populations of Kluyveromyces and Candida were detected on LM throughout the remainder of the fermentation. With the exception of Pichia, which was eliminated earlier, the non-Saccharomyces populations in the heated fermentation rapidly decreased once a peak temperature of 30°C was achieved and the ethanol concentration progressed above 50 g/liter.

PCR and RT-PCR-DGGE profiles.

The results obtained from DGGE analysis of the ambient- and high-temperature fermentations are shown in Fig. 3 and 4, respectively. Both the RT-PCR and PCR-DGGE profiles of the yeast populations roughly mirror the CFU data. DGGE bands were identified by comigration with PCR products generated from isolated strains and by direct DNA sequencing of DGGE bands. Several common trends were observed in the DGGE profiles. First, the DGGE patterns clearly demonstrate the longer persistence of non-Saccharomyces yeasts in the ambient-temperature fermentation than in the heated fermentation. Second, as yeast populations fell below ∼10⁴ CFU/ml, the cognate DGGE bands became faint or disappeared. This threshold is likely the result of a larger quantity of Saccharomyces DNA in these samples outcompeting the smaller amounts of template from the non-Saccharomyces yeasts for amplification of the rDNA (14).

FIG. 3.

DGGE analysis of 26S rDNA (PCR) or rRNA (RT-PCR) products obtained directly from the samples taken from the Dolce fermentation carried out at ambient temperature. Lane designations indicate the time of fermentation sampling (days). Bands marked with an asterisk were excised, reamplified, sequenced, and identified by sequence analysis. Abbreviations: H.u., H. uvarum; H.o., H. osmophila; K.t., K. thermotolerans; C., Candida sp. strain EJ1; S.c., S. cerevisiae.

FIG. 4.

DGGE analysis of 26S rDNA (PCR) or rRNA (RT-PCR) products obtained directly from the samples taken from the Dolce fermentation carried out at high temperature. Lane designations indicate the time of fermentation sampling (days). Bands marked with an asterisk were excised, reamplified, sequenced, and identified by sequence analysis. Abbreviations: H.u., H. uvarum; H.o., H. osmophila; K.t., K. thermotolerans; C., Candida sp. strain EJ1; S.c., S. cerevisiae.

Several populations of yeasts were readily observed in the DGGE profiles from day 1, including H. uvarum, Hanseniaspora osmophila, Candida sp. strain EJ1, and K. thermotolerans. A band corresponding to M. pulcherrima was not seen in DGGE gels even though the population was above 10⁵ CFU/ml. This is likely due to poor amplification of M. pulcherrima DNA with the NL1-LS2 primer set (data not shown). In addition, no DGGE bands corresponding to Pichia species were revealed, most likely due to the relatively low number of CFU present. A band corresponding to H. osmophila was identified even though H. osmophila was not revealed in the plating analysis. Thus, the DGGE results indicated a mixed Hanseniaspora population consisting of a minimum of two species that were not differentiated on the WLN medium.

As the fermentations progressed, a band corresponding to S. cerevisiae became visible at day 3 of the heated-fermentation profile (Fig. 4) and day 5 (Fig. 3) of the ambient-temperature profile, corresponding to populations of Saccharomyces in each fermentation of between 10⁴ and 10⁵ CFU/ml. PCR-DGGE and RT-PCR-DGGE profiles clearly show Candida species persisting throughout both fermentations. Candida populations identified within previous Dolce vintages appear as a characteristic DGGE doublet (11). It remains to be determined if this doublet is the result of a PCR artifact or indicates the presence of more than one strain of Candida within Dolce fermentations. Unlike the situation with other non-Saccharomyces yeasts, Candida bands were visible in both RT-PCR and PCR-DGGE profiles after CFU population levels had dropped below 10⁴ CFU/ml. This was particularly obvious in the heated trial, where CFU populations of 10² to 10³ CFU/ml resulted in clear DGGE bands.

Both PCR-DGGE and RT-PCR-DGGE profiles indicate K. thermotolerans populations persisting throughout the ambient-temperature fermentation and nearly disappearing as the population fell below 10⁴ CFU/ml. The cognate K. thermotolerans DGGE bands in the heated fermentation disappeared rapidly after the fermentation temperature reached 30°C and after the population fell below 10⁴ CFU/ml. Surprisingly, the PCR and RT-PCR-DGGE results differed in persistence of Hanseniaspora bands. In the ambient-temperature fermentation (Fig. 3), H. osmophila bands were present in the RT-PCR-DGGE sample for 21 days, while the PCR-DGGE profile indicated only an H. osmophila band for 11 days. Similarly, in the high-temperature fermentation (Fig. 4), the RT-PCR results failed to reveal an H. uvarum population in the first 4 days of the fermentation, whereas the PCR-DGGE profile clearly demonstrated an H. uvarum presence.

A viable but nonculturable Candida population?

One explanation for the persistence of a visible DGGE band emanating from a Candida population of below 10⁴ CFU/ml was that a higher population of viable but nonculturable cells existed. To examine this, a specific probe was designed to the Candida sp. strain EJ1 26S rDNA sequence and used semiquantitatively in rRNA slot blot analysis. As seen in Fig. 5, RNA purified from heated fermentation samples taken on days 12, 14, 15, and 45 revealed a strong presence of Candida sp. strain EJ1 rRNA at a level higher than predicted by the CFU analysis. As a comparison, RNA isolated from 10⁶ Candida sp. strain EJ1 cells (determined by number of CFU) was probed, resulting in a less intense signal. These results suggest that a metabolically active population of Candida sp. strain EJ1 persists throughout the Dolce fermentation at a substantially higher level than can be revealed by plating analysis.

FIG. 5.

RNA slot blot with a Candida sp. strain EJ1-specific probe. (A) RNA samples extracted directly from samples taken from the Dolce fermentation carried out at high temperature. Lane designations indicate the day of fermentation sampling and the S. cerevisiae RNA and blank controls. One microgram of total RNA was blotted onto the membrane. (B) Serial dilutions of Candida sp. strain EJ1 RNA. Lane designations indicate the amount of total RNA blotted onto the membrane. (C) Total RNA purified from serial dilutions of active Candida sp. strain EJ1 cells. Lane designations indicate the total number of cells from which RNA was extracted.

Candida sp. strain EJ1 is extremely fructophilic.

Examination of the final fructose concentrations present in the heated and ambient-temperature fermentations suggested that the persistence of non-Saccharomyces yeasts, particularly a high population of Candida sp. strain EJ1, resulted in a lowering of the final fructose concentration and a higher glucose/fructose ratio. Separate monoculture fermentations of the Candida sp. strain EJ1 in sterile Chardonnay juice revealed an extremely fructophilic phenotype (Fig. 6), in which no glucose was consumed even after the fructose was completely exhausted. In contrast, monoculture fermentations of the S. cerevisiae strain isolated from the Dolce fermentations exhibited a clear glucophilic phenotype.

FIG. 6.

Glucose and fructose consumption curves for fermentations of S. cerevisiae and Candida sp. strain EJ1 in sterile Chardonnay juice.

Yeast diversity in inoculated Dolce fermentations.

Dolce fermentations are also carried out with an S. cerevisiae starter culture inoculum. Given the strong presence of non-Saccharomyces yeasts in the uninoculated fermentations carried out at ambient temperatures, we predicted that a similar persistence of non-Saccharomyces yeasts would exist in the inoculated fermentations. Figure 7 shows a PCR-DGGE profile of select samples taken from an inoculated barrel during the 1999 Dolce fermentation. A strong DGGE band corresponding to Saccharomyces is visible early in the fermentation, as would be expected given the inoculation. DGGE bands corresponding to Hanseniaspora, Kluveromyces, and Candida populations are also observed until day 5, day 7, and day 12, respectively. The persistence of Candida and Kluveromyces populations in inoculated Dolce fermentations mirrors the situation in uninoculated fermentations, suggesting that a large initial population of Saccharomyces does not dramatically alter the fundamental yeast dynamics present within these fermentations.

FIG. 7.

DGGE analysis of 26S rDNA (PCR) products obtained directly from the samples taken from the Dolce fermentation inoculated with an S. cerevisiae starter culture. Lane designations indicate the time of fermentation sampling (days). Abbreviations: H.u., H. uvarum; K.t., K. thermotolerans; C., Candida sp. strain EJ1; S.c., S. cerevisiae.

DISCUSSION

The presence of non-Saccharomyces yeasts in wine fermentations has been documented extensively (15, 35). The non-Saccharomyces yeasts most often associated with wine fermentations are Hanseniaspora and Candida species and, to a lesser extent, Pichia, Kluveromyces, and Metschnikowia species, among others (17). These yeasts may affect wine fermentations both directly, through production of off-flavors, and indirectly by modulating the growth or metabolism of the dominant Saccharomyces population (4). Significant growth of non-Saccharomyces yeasts early in fermentations has been associated with off-character production in wines and/or stuck and sluggish fermentations (3).

Previous analysis of Dolce fermentations indicated a rich diversity of yeasts present in the initial stages of the fermentations (A. Heisey, personal communication). Prior PCR-DGGE analysis revealed the persistence of a Candida population throughout the fermentations (11). In this work both direct molecular and indirect plating methods were employed to characterize the diversity within ambient-temperature (∼20°C) and transiently heated (∼30°C) Dolce fermentations. In both trials similar arrays of yeast genera were observed. Saccharomyces, Hanseniaspora, Pichia, Metschnikowia, Kluyveromyces, and Candida species were present at similar levels in the initial stages of both fermentations. While the initial population size of Candida was high (10⁶ to 10⁷ CFU/ml) for both fermentations, it was similar to those of non-Saccharomyces yeast populations observed in other botrytis-affected wine fermentations (18). As expected, Saccharomyces dominated both ambient-temperature and heated fermentations. Moreover, Candida and Kluyveromyces populations grew slightly and persisted at detectable levels throughout both fermentations. C. stellata and K. thermotolerans have been observed to persist in a similar fashion in some Majorcan wine fermentations (32). Given the strong persistence of the non-Saccharomyces yeasts in the uninoculated Dolce fermentation, it is not surprising to find a similar persistence within inoculated fermentations (Fig. 7).

This work clearly demonstrates the impact that fermentation temperature has on growth of non-Saccharomyces yeasts. In the heated fermentations, most non-Saccharomyces populations disappeared dramatically as assessed by both direct molecular and CFU plating analyses. The most obvious explanation for the decrease in non-Saccharomyces yeasts in the heated trials is the combination of higher temperatures and ethanol (5). Alternatively, the non-Saccharomyces yeasts might not have been able to adapt as quickly to the more rapid increase in ethanol concentration in the heated fermentations compared to a more gradual ethanol production witnessed in the ambient-temperature trial. Regardless, the detrimental effect of temperature and ethanol is illustrated by a comparison of the persistence of Candida populations in both trials. In the heated barrels the Candida CFU population decreased dramatically as the temperature of the fermenting juice reached ∼30°C and as the ethanol concentration rose above 50 g/liter. In the ambient-temperature trial, a high Candida CFU population persisted for 21 days, during which the juice temperature never rose significantly above 20°C. In that fermentation, Candida species survived ethanol concentrations of approximately 100 g/liter before a significant decline in the number of CFU occurred. This result agrees with previous studies that demonstrated an increased tolerance of ethanol by Candida strains at lower fermentation temperatures (20). Other studies on temperature effects on non-Saccharomyces yeasts demonstrated no significant growth rate differences between various strains of Kloeckera (Hanseniaspora), Candida, and Saccharomyces species (7). However, during the Dolce fermentations, none of the non-Saccharomyces yeast populations, in either trial, increased in population size by more than an order of magnitude (as determined by CFU per milliliter). Thus, a transient temperature increase in the heated fermentations may have affected non-Saccharomyces yeast persistence rather than growth.

The extremely fructophilic nature of Candida sp. strain EJ1 suggests an additional rationale for copersistence of a Candida population with a larger Saccharomyces population in Dolce fermentations. A fructophilic Candida population would not have an impact on glucose levels consumed by a glucophilic S. cerevisiae population and therefore would not compete for substrate. An expected outcome of this apparent neutralism is a greater total consumption of fructose. This is evident in the ambient-temperature Dolce fermentation, where persistence of a higher Candida population correlated with consumption of additional fructose compared to those in the heated fermentation (Fig. 1 and 2). As the heated fermentation reached ethanol concentrations of 100 g/liter, the glucose/fructose ratio was approximately 0.71, while, at the same ethanol concentration, the ambient-temperature fermentations exhibited a much closer glucose/fructose ratio (∼0.9). Ciani and Ferraro (8) demonstrated that mixed fermentations containing C. stellata and Saccharomyces exhibited more complete utilization of sugars and postulated that the cause was the preferential utilization of fructose by highly fructophilic C. stellata species. Maintenance of an optimum glucose/fructose ratio has been suggested as a cause of reduced fermentative activity in Saccharomyces (19). By selectively consuming fructose in the Dolce fermentation, the Candida population may act in a commensal fashion, aiding the overall fermentative capacity of S. cerevisiae by increasing the glucose/fructose ratio. Whether or not this potential commensalism exists, the differences in sugar utilization between Candida and Saccharomyces species illustrate a complementing aspect of indigenous yeasts that are enriched to populate uninoculated wine fermentations.

Several groups have examined Candida species as potential adjuncts for flavor modification of wine (8, 36). Others have noted significant glycerol and acetic acid production by C. stellata when examined in monoculture (9, 36). In this work, the persistence of Candida populations in the ambient-temperature fermentations did not coincide with increased glycerol concentrations (Fig. 2 and 3). As both heated and ambient-temperature fermentations reached ethanol concentrations of 100 g/liter, the glycerol concentration was around 16 g/liter. Glycerol was produced more rapidly in heated fermentations; however, that increase coincided with a similarly rapid production of ethanol, suggesting that Saccharomyces was the producing microorganism. Neither heated nor ambient-temperature fermentations exhibited a dramatic difference in acetic acid concentrations.

Direct molecular methods are now a commonly used tool for ecological analysis, revealing the tremendous diversity of uncultured microorganisms in various habitats (33). To date, however, relatively few studies have employed such methods to characterize the microbial ecology of food and wine fermentations (22). In this study, the use of PCR-DGGE in combination with plating analysis revealed the strengths and weaknesses of the two approaches. For example, PCR-DGGE analysis revealed an additional Hanseniaspora population, H. osmophila, in the Dolce fermentations which went unnoticed in the CFU analysis. This omission was likely due to a color morphology similar to that of H. uvarum on WLN medium (both have green colonies). Interestingly, RT-PCR-DGGE analysis suggests that an active H. osmophila population of greater than ∼10⁴ cells per ml (the approximate PCR-DGGE detection limit in the presence of a higher Saccharomyces population [10]) was present throughout most of the ambient-temperature fermentation. RT-PCR-DGGE analysis did not detect an H. uvarum population, although a slight band could be detected in the PCR-DGGE analysis. The lack of an observable H. uvarum band in the RT-PCR-DGGE gel may be a result of a copy number effect, whereby the increased amount of competing yeast template rRNA in the RNA samples effectively masked an observable H. uvarum band. Neither Hanseniaspora population (H. osmophila or H. uvarum) was detected by CFU analysis in the later stages of either the ambient-temperature or heated fermentations, although a lower H. osmophila population may have been obscured in the CFU and DGGE analysis by the higher Candida population.

Direct analysis also demonstrated the persistence of a Candida population in the heated fermentation even though CFU analysis indicated a population level below the PCR-DGGE detection threshold. This contrasted with the situation with the Kluveromyces population, which disappeared from the DGGE analysis at approximately the same time that the population fell below 10⁴ CFU per ml (compare Fig. 2 and 4). Specific detection of the Candida population by RNA hybridization confirmed the existence of a Candida population of greater than 10⁶ cells per ml. Previously, Millet and Lonvaud-Funel used direct epifluorescence microscopy on aging wines and observed a dramatically higher bacterial population than was revealed by plating analysis (31). This work suggests a similar situation with indigenous yeasts in the Dolce fermentation. However, it remains to be determined if the direct RNA analysis has revealed a metabolically active Candida population or a metabolically inactive Candida population in which the cellular RNA was protected from degradation (or both).

Certain deficiencies with PCR-DGGE methodology were also revealed in this work. Detection of a Metschnikowia population was not possible even though the CFU population determined in the initial stages of either fermentation appeared to be above the detection threshold for PCR-DGGE. Amplification of the 26S rDNAs from Metschnikowia isolates obtained in this study often resulted in poor yields (data not shown), which could have resulted in an omission of Metschnikowia in DGGE profiles of fermentation samples. Previous analysis of an M. pulcherrima isolate by DGGE indicated a single resolved band (10). However, recent work by Pallman et al. (34) demonstrated a diverse number of M. pulcherrima biotypes within a single commercial wine fermentation. Thus, variability within the 26S rRNA D1-D2 region may result in poor amplification of Metschnikowia rDNA by the NL1-LS2 primer pair used in this study.