Dynamic changes in microbiota and mycobiota during spontaneous 'Vino Santo Trentino' fermentation

Abstract

Vino Santo is a sweet wine produced from late harvesting and pressing of Nosiola grapes in a small, well-defined geographical area in the Italian Alps. We used metagenomics to characterize the dynamics of microbial communities in the products of three wineries, resulting from spontaneous fermentation with almost the same timing and procedure. Comparing fermentation dynamics and grape microbial composition, we show a rapid increase in a small number of wine yeast species, with a parallel decrease in complexity. Despite the application of similar protocols, slight changes in the procedures led to significant differences in the microbiota in the three cases of fermentation: (i) fungal content of the must varied significantly in the different wineries, (ii) Pichia membranifaciens persisted in only one of the wineries, (iii) one fermentation was characterized by the balanced presence of Saccharomyces cerevisiae and Hanseniaspora osmophila during the later phases. We suggest the existence of a highly winery-specific 'microbial-terroir' contributing significantly to the final product rather than a regional 'terroir'. Analysis of changes in abundance during fermentation showed evident correlations between different species, suggesting that fermentation is the result of a continuum of interaction between different species and physical-chemical parameters.

Figure 1

Sample details and selection. A. Must samples were collected from the three wineries (Poli, Pedrotti and Pisoni) once a week every week from the beginning of fermentation until completion, occurring on the 88th day of fermentation at all the wineries. Samples were named with ‘T’ (time) and the number of days passing from the beginning of fermentation (0–88). Each sample was analysed using PCR‐RFLP (see details in materials and methods), and the samples most representative of each winery (see clustering in different parts of the dendrogram in B) were selected for meta‐taxonomic analysis (highlighted in red). B. Dendrograms were built on distances calculated on the ITS1‐5.8S–ITS2 regions PCR‐RFLP profiles. Amplified ITS1‐5.8S–ITS2 regions were digested with both HaeIII and HinfI enzymes. The most parsimonious trees were found on 1/0 (presence/ absence) band profiles with p enny software (p hylip). Finally, a consensus dendrogram was obtained with consense (p hylip) and drawn with figtree. Label colours: red = Poli winery must samples, blue = Pisoni winery must samples, green = Pedrotti winery must samples. Asterisks indicate samples that were selected for meta‐taxonomic analysis by 454‐pyrosequencing.

Figure 2

Fermentation dynamics. A. Fungal DNA quantification. Total fungal DNA was quantified through qRT‐PCR by using universal primers amplifying the ITS1 region. Standard curves were constructed using PCR products of the ITS1 rDNA of a metagenomic sample. Error bars: standard deviations of three technical replicates. B. Chao1 estimator of the number of fungal OTUs observed in the samples using meta‐taxonomic analysis. C. Pearson correlation r of fungal populations at T0 against each other at time points. D. Weighted u nifrac distances of fungal population at T0 with respect to fungal populations at the other time points. Lines represent only guidelines.

Figure 3

Sample fungal population diversity. Distribution of variables over the first two components of the PCoA carried out on weighted u ni f rac distances. Left: first two coordinates of the PCoA. Arrows connect consecutive time points. Right: the same PCoA plot with the eight most abundant OTUs (showing the highest average relative abundance across all the samples) overlaid as coloured points with a size proportional to the mean relative abundance of the taxon across all samples. Genus coordinates were calculated as the weighted average across sample coordinates. Grey dots in the right plot represent the sample coordinates; arrows in the left plot show the progression of time.

Figure 4

Vino Santo mycobiota. A. Average relative abundance (bars) and number of winery‐specific species (dashed lines) during the three phases of fermentation, early (0–7 days, T0‐T7), intermediate (14–21 days, T14‐T21) and late (54–77 days, T54‐T77). The fungal species present in fermentation at all the wineries make up the ‘Vino Santo core’. Filled bars indicate the average relative abundance of the Vino Santo fermentation core. The error bars indicate the standard error of the mean. The names of the species belonging to various identified groups are superimposed on (or written above) the corresponding bars; B. Absolute abundance correlation matrix of genera not having linear progression with time (Spearman P ≥ 0.05). Couples of correlated genera were selected when Spearman correlation resulted in r > 0.7 and P < 0.05. Crossed squares indicate non‐significant pairs (P ≥ 0.05). C. Correlation network of species selected as being correlated with each other in terms of relative abundance and having a linear progression with time. Pairs of correlated species were selected when Spearman correlation resulted in r > 0.7 and P < 0.05. Significantly correlated pairs of species were then selected as significantly correlated with time (linear regression of both species having P < 0.05).

Figure 5

Spearman's correlations with fungal OTU relative abundances and chemical factors. A. computed using the entire dataset. B. computed using data divided by winery. Crossed squares indicate non‐significant pairs (P ≥ 0.05).