Alterations in Yeast Species Composition of Uninoculated Wine Ferments by the Addition of Sulphur Dioxide

Uninoculated wine fermentations are conducted by a consortium of wine yeast and bacteria that establish themselves either from the grape surface or from the winery environment. Of the additives that are commonly used by winemakers, sulphur dioxide (SO2) represents the main antimicrobial preservative and its use can have drastic effects on the microbial composition of the fermentation. To investigate the effect of SO2 on the resident yeast community of uninoculated ferments, Chardonnay grape juice from 2018 and 2019 was treated with a variety of SO2 concentrations ranging up to 100 mg/L and was then allowed to undergo fermentation, with the yeast community structure being assessed via high-throughput meta-barcoding (phylotyping). While the addition of SO2 was shown to select against the presence of many species of non-Saccharomyces yeasts, there was a clear and increasing selection for the species Hanseniaspora osmophila as concentrations of SO2 rose above 40 mg/L in fermentations from both vintages. Chemical analysis of the wines resulting from these treatments showed significant increases in acetate esters, and specifically the desirable aroma compound 2-phenylethyl acetate, that accompanied the increase in abundance of H. osmophila. The ability to modulate the yeast community structure of an uninoculated ferment and the resulting chemical composition of the final wine, as demonstrated in this study, represents an important tool for winemakers to begin to be able to influence the organoleptic profile of uninoculated wines.


Introduction
Wine is a complex beverage, produced through the interplay of grape and microbial metabolomes during the process of fermentation. While the majority of modern commercial wine fermentation is performed using inoculated commercial strains of the major wine yeast Saccharomyces cerevisiae, a significant proportion of commercial wine fermentations are now being performed using uninoculated grape must. In these situations, the fermentation is conducted by a consortium of wine yeast and bacteria that establish themselves either from the grape surface or from the winery via shared equipment or other vectors such as insects [1].
In the very early stages of fermentation, apiculate yeasts, and yeast-like fungi which reside on the surface of intact grape berries or winery equipment and include the genera Aureobasidium, Rhodotorula,

Results
In order to evaluate the effect of SO 2 addition on the yeast population structure, triplicate uninoculated fermentations were established in Chardonnay grape juice across two consecutive vintages. The effect of these different SO 2 concentrations on wine volatile composition was also evaluated.

SO 2 Addition Affects Yeast Population Structure
In vintage 2018, the grape juice was treated with one of five different concentrations of total SO 2 (0, 40, 60, 80 and 100 mg/L). In addition to its antimicrobial effect, SO 2 is also a powerful antioxidant [17]. To differentiate between the antimicrobial and antioxidant effects of the SO 2 addition, an alternate antioxidant, glutathione (GSH, 250 mg/L), was also assessed for its effects on the yeast community structure.
The progress of each ferment was tracked via sugar consumption ( Figure S1), with samples taken immediately after SO 2 or GSH addition (T1), at 90% of sugar remaining (T2), 50% sugar remaining (T3) and 10% sugar remaining (T4), for meta-barcoding analysis using the fungal Internal Transcribed Spacer (ITS) region [18,19]. The addition of GSH did not affect the duration of fermentation, however, SO 2 had a significant impact on the length of time required for the fermentation to reach completion, with two of the 100 mg/mL treatments requiring five to seven days longer than the control ferments (26 day fermentation) and one of the 100 mg/mL treatments becoming stuck with 13 g/L of residual sugar.
Across the 18 samples from 2018 (6 treatments in triplicate), Operational Taxonomic Units (OTUs) that could be assigned to a total of 26 fungal genera were detected that exceeded 0.01% of the total abundance in at least one sample (Figure 1; full results in Table S1). Triplicate samples were shown to be highly concordant for each combination of SO 2 concentration and timepoint ( Figure S2). The highest level of fungal diversity was observed at the T1 timepoint, while Hanseniaspora, Metschnikowia, Saccharomyces and Torulaspora dominated the fermentations from T2 through T4, accounting over 95% of the total ITS reads ( Figure 1).
As seen for the fermentation kinetics, the GSH addition did not affect the overall population structure relative to the control samples, however, the addition of SO 2 had a significant, but differential effect on the four main genera observed across the samples (Figure 1). Metschnikowia displayed the highest sensitivity to SO 2 , with 40 mg/L completely inhibiting the detection of this genus by the T2 timepoint. Torulaspora was shown to have a higher abundance at 40 mg/L relative to 0 mg/L, however, this genus was progressively inhibited by higher concentrations of SO 2 in a gradient from 60 through to 100 mg/L, at which point it was completely inhibited at timepoint T2. Hanseniaspora and Saccharomyces were both shown to be tolerant across all the tested SO 2 concentrations, with Hanseniaspora increasing in its total proportion relative to the other genera as the concentration of SO 2 was increased.
Fermentation 2020, 6, x FOR PEER REVIEW 3 of 14 Figure 1. Genus-level metabarcoding analysis of community response to SO2 addition. Vintage 2018 Chardonnay juice was treated with increasing concentrations of total SO2 (mg/L) or glutathione (GSH, 250 mg/L) as an alternate antioxidant. Ferment samples were taken at four timepoints (T1, at crush; T2, 10% sugar utilization; T3, 50% sugar utilization; T4, 90% sugar utilization) and subjected to ITS metabarcoding. Only genera that exceeded 0.1% abundance in at least one sample are shown.
As seen for the fermentation kinetics, the GSH addition did not affect the overall population structure relative to the control samples, however, the addition of SO2 had a significant, but differential effect on the four main genera observed across the samples (Figure 1). Metschnikowia displayed the highest sensitivity to SO2, with 40 mg/L completely inhibiting the detection of this genus by the T2 timepoint. Torulaspora was shown to have a higher abundance at 40 mg/L relative to 0 mg/L, however, this genus was progressively inhibited by higher concentrations of SO2 in a gradient from 60 through to 100 mg/L, at which point it was completely inhibited at timepoint T2. Hanseniaspora and Saccharomyces were both shown to be tolerant across all the tested SO2 concentrations, with Hanseniaspora increasing in its total proportion relative to the other genera as the concentration of SO2 was increased.
As fungal ITS sequencing generally affords the ability to define OTUs to the species level, the genus level counts were partitioned into species-level units to determine the effect of SO2 concentration on the abundance of individual species. There were 29 species that exceeded 0.1% of the total abundance in any sample, with the genus Hanseniaspora displaying the highest number of individual species (n = 4). While the addition of SO2 was shown to increase the overall abundance of Hanseniaspora at the genus level, there was a far more complex response profile when species designations were taken into account ( Figure 2). Rather than a general increase in all species of Hanseniaspora, two species, H. uvarum and H. opuntiae, were the dominant species when SO2 was absent (GSH and SO2 0 mg/L treatments). However, the addition of 40 mg/L of SO2 resulted in a drastic shift in the species composition such that H. osmophila was the sole representative of this genus Chardonnay juice was treated with increasing concentrations of total SO 2 (mg/L) or glutathione (GSH, 250 mg/L) as an alternate antioxidant. Ferment samples were taken at four timepoints (T1, at crush; T2, 10% sugar utilization; T3, 50% sugar utilization; T4, 90% sugar utilization) and subjected to ITS metabarcoding. Only genera that exceeded 0.1% abundance in at least one sample are shown.
As fungal ITS sequencing generally affords the ability to define OTUs to the species level, the genus level counts were partitioned into species-level units to determine the effect of SO 2 concentration on the abundance of individual species. There were 29 species that exceeded 0.1% of the total abundance in any sample, with the genus Hanseniaspora displaying the highest number of individual species (n = 4). While the addition of SO 2 was shown to increase the overall abundance of Hanseniaspora at the genus level, there was a far more complex response profile when species designations were taken into account ( Figure 2). Rather than a general increase in all species of Hanseniaspora, two species, H. uvarum and H. opuntiae, were the dominant species when SO 2 was absent (GSH and SO 2 0 mg/L treatments). However, the addition of 40 mg/L of SO 2 resulted in a drastic shift in the species composition such that H. osmophila was the sole representative of this genus at 40 mg/L of SO 2 . The relative abundance of this species increased substantially as SO 2 levels were raised, producing the overall increase in Hanseniaspora that was observed at the genus level. at 40 mg/L of SO2. The relative abundance of this species increased substantially as SO2 levels were raised, producing the overall increase in Hanseniaspora that was observed at the genus level.  (T1, at crush; T2, 10% sugar utilization; T3, 50% sugar utilization; T4, 90% sugar utilization) and subjected to ITS metabarcoding. Only species that exceeded 0.1% abundance in at least one sample are shown. For those OTU where a species-level designation was not possible, the genus-level taxonomic classification of the OTU was used. Only members of the genus Hanseniaspora are colored.
A second set of fermentations were established in the subsequent year (2019) using a finer set of SO2 treatment intervals (0, 10, 20 and 40 mg/L). Consistent with the observations from the 2018 vintage, the SO2 addition affected fermentation kinetics, particularly for the 40 mg/L treatments ( Figure S1). The 2019 ferments displayed a different overall yeast diversity compared with the 2018 samples, with a lack of OTUs that could be assigned to Metschnikowia and prominent contributions from OTUs assigned to Candida spp., which increased over the range of SO2 concentration used, and Kazachstania spp. that were present at up to 20 mg/L of SO2 ( Figure 3). However, when the species level contributions were investigated, there were clear similarities between the two vintages in the dynamics of the OTUs assigned to the genus Hanseniaspora ( Figure 4). As also seen in 2018, Hanseniaspora was represented by the greatest number of species designations in 2019. The 2019 ferments also displayed a clear species shift that was associated with the use of SO2, with the proliferation of H. uvarum, H. opuntiae and H. vineae all being inhibited by SO2 in a concentrationdependent manner, while the proportion of H. osmophila was shown to be enhanced by the addition of 40 mg/L of SO2. Species-level meta-barcoding analysis of community response to the SO 2 addition. Vintage 2018 Chardonnay juice was treated with increasing concentrations of total SO 2 (mg/L) or glutathione (GSH, 250 mg/L) as an alternate antioxidant. Ferment samples were taken at four timepoints (T1, at crush; T2, 10% sugar utilization; T3, 50% sugar utilization; T4, 90% sugar utilization) and subjected to ITS metabarcoding. Only species that exceeded 0.1% abundance in at least one sample are shown. For those OTU where a species-level designation was not possible, the genus-level taxonomic classification of the OTU was used. Only members of the genus Hanseniaspora are colored.
A second set of fermentations were established in the subsequent year (2019) using a finer set of SO 2 treatment intervals (0, 10, 20 and 40 mg/L). Consistent with the observations from the 2018 vintage, the SO 2 addition affected fermentation kinetics, particularly for the 40 mg/L treatments ( Figure S1). The 2019 ferments displayed a different overall yeast diversity compared with the 2018 samples, with a lack of OTUs that could be assigned to Metschnikowia and prominent contributions from OTUs assigned to Candida spp., which increased over the range of SO 2 concentration used, and Kazachstania spp. that were present at up to 20 mg/L of SO 2 ( Figure 3). However, when the species level contributions were investigated, there were clear similarities between the two vintages in the dynamics of the OTUs assigned to the genus Hanseniaspora ( Figure 4). As also seen in 2018, Hanseniaspora was represented by the greatest number of species designations in 2019. The 2019 ferments also displayed a clear species shift that was associated with the use of SO 2 , with the proliferation of H. uvarum, H. opuntiae and H. vineae all being inhibited by SO 2 in a concentration-dependent manner, while the proportion of H. osmophila was shown to be enhanced by the addition of 40 mg/L of SO 2 . . Genus-level meta-barcoding analysis of community response to the SO2 addition. Vintage 2019 Chardonnay juice was treated with increasing concentrations of total SO2 (mg/L). Ferment samples were taken at four timepoints (T1, at crush; T2, 10% sugar utilization; T3, 50% sugar utilization; T4, 90% sugar utilization) and subjected to ITS metabarcoding. Only genera that exceeded 0.1% abundance in at least one sample are shown. . Genus-level meta-barcoding analysis of community response to the SO 2 addition. Vintage 2019 Chardonnay juice was treated with increasing concentrations of total SO 2 (mg/L). Ferment samples were taken at four timepoints (T1, at crush; T2, 10% sugar utilization; T3, 50% sugar utilization; T4, 90% sugar utilization) and subjected to ITS metabarcoding. Only genera that exceeded 0.1% abundance in at least one sample are shown.
In order to compare the 2018 and 2019 data, the metabarcoding time course results were analyzed by Bray-Curtis dissimilarity analysis ( Figure 5). The T1 samples from across both vintages were broadly similar, with the ferments characterized by the presence of non-fermentative genera such as Aureobasidium, Cladosporium and Epicoccum. As observed in the abundance plots, ferments progressed towards being dominated by S. cerevisiae (Axis 1), however, there was a clear division between samples with 40 mg/L or more of added SO 2 , which deviated along Axis 2 towards H. osmophila, while the samples with less than 40 mg/L were dominated by the signal from H. uvarum. Thus, despite differences in the overall microbial populations of the wild fermentations performed across the two vintages, both displayed consistent alterations in the microbial community due to the amount of SO 2 addition.  In order to compare the 2018 and 2019 data, the metabarcoding time course results were analyzed by Bray-Curtis dissimilarity analysis ( Figure 5). The T1 samples from across both vintages were broadly similar, with the ferments characterized by the presence of non-fermentative genera such as Aureobasidium, Cladosporium and Epicoccum. As observed in the abundance plots, ferments progressed towards being dominated by S. cerevisiae (Axis 1), however, there was a clear division between samples with 40 mg/L or more of added SO2, which deviated along Axis 2 towards H. osmophila, while the samples with less than 40 mg/L were dominated by the signal from H. uvarum. Thus, despite differences in the overall microbial populations of the wild fermentations performed across the two vintages, both displayed consistent alterations in the microbial community due to the amount of SO2 addition. . Species-level metabarcoding analysis of community response to the SO 2 addition. Vintage 2019 Chardonnay juice was treated with increasing concentrations of total SO 2 (mg/L). Ferment samples were taken at four timepoints (T1, at crush; T2, 10% sugar utilization; T3, 50% sugar utilization; T4, 90% sugar utilization) and subjected to ITS metabarcoding. Only species that exceeded 0.1% abundance in at least one sample are shown. For those OTU where a species-level designation was not possible, the genus-level taxonomic classification of the OTU was used. Only members of the genus Hanseniaspora are colored.

SO 2 Addition Influences Wine Volatile Composition
Given the significant effect of SO 2 addition on the microbial community structure, it was of interest to understand whether these different SO 2 treatments were also associated with changes to the chemical composition of the wine. This was assessed through an analysis of the volatile yeast metabolites known to contribute to the aromatic profile of wine. Of the 39 aroma compounds analyzed, 18 displayed a significant difference (ANOVA, p ≤ 0.001) in concentration across one of the SO 2 regimes in either 2018 or 2019 (Table S2). Of these, ten analytes displayed more than a 1.5-fold decrease in at least one of the SO 2 treatments, while seven displayed an increase of the same magnitude ( Figure 6). Three analytes, 2-methylpropanol (decreasing in response to SO 2 ) and 2-phenylethyl acetate and hexyl acetate (increasing), displayed the same effect across the 2018 and 2019 vintages. In all cases, there was a significant effect at 40 mg/L SO 2 . Furthermore, in most situations in which a significant difference in analyte concentration was observed across multiple SO 2 regimes, there was a correlation between SO 2 concentration and the magnitude of change. The largest change in analyte concentration was observed for the desirable aroma compound 2-phenylethyl-acetate, with the 100 mg/L SO 2 treatment in 2018 displaying over two orders of magnitude more of this metabolite than the control, and the 2019 40 mg/L SO 2 treatment having over nine times as much 2-phenylethyl-acetate as the control (Table S2). More generally, higher SO 2 concentrations resulted in decreases in short chain acetates and higher alcohols and increases in 6-carbon and 8-carbon esters and acids. There was no effect of SO 2 concentration on low molecular weight volatile sulphur compound production.

SO2 Addition Influences Wine Volatile Composition
Given the significant effect of SO2 addition on the microbial community structure, it was of interest to understand whether these different SO2 treatments were also associated with changes to the chemical composition of the wine. This was assessed through an analysis of the volatile yeast metabolites known to contribute to the aromatic profile of wine. Of the 39 aroma compounds the 100 mg/L SO2 treatment in 2018 displaying over two orders of magnitude more of this metabolite than the control, and the 2019 40 mg/L SO2 treatment having over nine times as much 2-phenylethylacetate as the control (Table S2). More generally, higher SO2 concentrations resulted in decreases in short chain acetates and higher alcohols and increases in 6-carbon and 8-carbon esters and acids. There was no effect of SO2 concentration on low molecular weight volatile sulphur compound production.

Discussion
Winemakers are limited in their ability to influence the native microbial population of grape juice, with SO 2 addition representing the main available intervention. Previous microbiological studies have shown that species (and strains) of the major wine yeasts can respond differently to the application of SO 2 , with commercial strains of S. cerevisiae displaying diverse but higher tolerance to SO 2 [20], while "wild" yeasts display lower tolerances and are therefore thought to be broadly selected against through the Fermentation 2020, 6, 62 9 of 13 application of moderate amounts of SO 2 prior to the start of fermentation [14][15][16]21]. Through the application of ITS meta-barcoding, this study demonstrated that the addition of over 40 mg/L of total SO 2 favored the presence of the non-Saccharomyces species H. osmophila at the expense of other genera such as Metschnikowia, Torulaspora and Kazachstania. H. osmophila has previously been shown to be resistant to SO 2 concentrations of over 40 mg/L [22,23]. While previous research into the effects of SO 2 on grape juice consortia did observe the antagonistic effect of SO 2 against non-Saccharomyces yeasts, the presence of H. osmophila was not specifically reported, although Hanseniaspora yeasts were observed at levels of SO 2 above 40 mg/L [16,24,25].
Much is known regarding the molecular basis of SO 2 tolerance in S. cerevisiae, where the sulfite efflux pump SSU1 provides the main mode of resistance [26,27], but very little information is available on the main genetic determinants of SO 2 resistance in non-Saccharomyces species, although SSU1 has been suggested to have a major role in Brettanomyces bruxellensis [28]. However, recent comparative genomic studies of Hanseniaspora spp. have shown that there is a clear differentiation of this genus into two well-defined phylogenetic clades, in which one of the differentiating factors is a homolog of SSU1, which is absent in the large clade containing H. uvarum, but present in the clade containing H. osmophila and H. vineae [29][30][31]. While this may explain the different response to SO 2 of H. osmophila versus H. uvarum, there are likely many other factors that impact the response of a specific species, as T. delbrueckii, which also possesses an SSU1 homolog and displayed increased abundance at up to 40 mg/L SO 2 , and is less tolerant at higher SO 2 concentrations than H. osmophila. Likewise, H. vineae, which was present at levels similar to H. osmophila in the control ferments and contains an SSU1 homolog, did not increase in abundance in response to increases in the concentrations of SO 2 .
Detailed chemical analysis showed that the addition of SO 2 resulted in a significant increase in the concentration of key esters and particularly the aroma compound 2-phenylethyl-acetate, which increased over 9-fold, to levels well above the sensory threshold for this compound, even under modest SO 2 additions (40 mg/L). Given the microbiological shift that was observed, it is likely that this change in ester production is due to the increasing prevalence of H. osmophila in these ferments with higher SO 2 levels. This is supported by published data from fermentations established with purified non-Saccharomyces strains, in which mixtures of H. osmophila and S. cerevisiae (90:10 ratio) were shown to produce higher concentrations of acetate esters (with the exception of isoamyl acetate) and concentrations of 2-phenylethyl-acetate almost 10-fold greater than those observed using S. cerevisiae alone [32,33].
In summary, uninoculated fermentations can provide desirable complexity, however, the process lacks the ability to introduce specific fermentation characteristics through the use of commercial starter strains with distinct fermentation aroma and/or flavor profiles. The ability to modulate the yeast community structure of an uninoculated ferment, and the resulting chemical composition of the final wine, demonstrated in this study represents an important tool for winemakers to begin to be able to influence the organoleptic profile of uninoculated wines.

Fermentation
Commercial, high-solids Chardonnay juice (not pre-treated with SO 2 during harvesting or processing) was obtained directly after destemming and crushing from Yalumba wineries during the 2018 and 2019 vintages ( Table 1). The juice was transferred to 2 L Schott bottles and then treated with either potassium metabisulfite (ACE Chemical Company, Camden Park, Australia) to the appropriate final total SO 2 concentration or 250 mg of reduced glutathione (GSH, Sigma-Aldrich Sydney, Australia) as indicated. Each experiment was performed in triplicate. Bottles were sealed with airlocks and incubated at 18 • C. Ferments were assessed at least every 24 h by refractometry and sugar analysis (see below), with samples taken for meta-barcoding at four approximate sugar levels (T1, directly after treatment; T2, 90% sugar remaining; T3, 50% sugar remaining; T4, 10% sugar remaining) from an in-built sampling port.

Chemical Analysis
Titratable acidity and pH were determined using a TitraLab 840 (Radiometer) and the yeast assimilable nitrogen concentration was estimated by the NOPA + enzymatic ammonia method [40] on a Gallery Discrete Analyser (ThermoFischer, Waltham, USA) by AWRI Commercial Services (Australia).
During fermentation, [Glucose + Fructose] concentrations were determined spectrophotometrically using a Randox kit (Randox Laboratories Ltd., Crumlin, Antrim, UK) with adaptations for use in a 96-well microplate format [41]. Volatile acetates, esters and higher alcohol concentrations in the finished wines were determined using large-volume, stable-isotope dilution headspace-GC/MS analysis (Metabolomics Australia, Adelaide Australia) as adapted from [42] and as described by [43]. Volatile sulphur compounds contributing sulfidic off-aromas, were determined by gas chromatography with sulphur chemiluminescence detection (GC/SCD) [42,44]. Aroma thresholds were in wine estimated using data from Siebert et al., [42] Free SO 2 was measured in grape juice supplemented with freshly prepared PMS using the aspiration/titration method [45].

Statistical Analysis
An analysis of variance (ANOVA) was conducted using the formula aov (analyte~treatment) in R (version 3.6.3) to determine whether mean aroma-active compound concentrations (n = 3 for all treatments) differed with regard to SO 2 treatment. If ANOVA p values were less than 0.05, a multiple comparison of the analyte concentration with respect to treatment was undertaken using the function HSD.test (agricolae) to determine the grouping of the treatments at alpha = 0.05. ANOVA F 3,8 values, p values, treatment means, standard deviations and treatment group are reported in Table S2.
Supplementary Materials: The following are available online at http://www.mdpi.com/2311-5637/6/2/62/s1, Figure S1: Sugar-consumption kinetics, Figure S2: Replicate fermentation comparisons, Table S1: OTU abundance measurements, Table S2: Chemical analysis results. Funding: The AWRI is supported by Australia's grape growers and winemakers through their investment body Wine Australia with matching funds from the Australian Government.