Adaptation of S. cerevisiae to Fermented Food Environments Reveals Remarkable Genome Plasticity and the Footprints of Domestication

Abstract

The budding yeast Saccharomyces cerevisiae can be found in the wild and is also frequently associated with human activities. Despite recent insights into the phylogeny of this species, much is still unknown about how evolutionary processes related to anthropogenic niches have shaped the genomes and phenotypes of S. cerevisiae. To address this question, we performed population-level sequencing of 82 S. cerevisiae strains from wine, flor, rum, dairy products, bakeries, and the natural environment (oak trees). These genomic data enabled us to delineate specific genetic groups corresponding to the different ecological niches and revealed high genome content variation across the groups. Most of these strains, compared with the reference genome, possessed additional genetic elements acquired by introgression or horizontal transfer, several of which were population-specific. In addition, several genomic regions in each population showed evidence of nonneutral evolution, as shown by high differentiation, or of selective sweeps including genes with key functions in these environments (e.g., amino acid transport for wine yeast). Linking genetics to lifestyle differences and metabolite traits has enabled us to elucidate the genetic basis of several niche-specific population traits, such as growth on galactose for cheese strains. These data indicate that yeast has been subjected to various divergent selective pressures depending on its niche, requiring the development of customized genomes for better survival in these environments. These striking genome dynamics associated with local adaptation and domestication reveal the remarkable plasticity of the S. cerevisiae genome, revealing this species to be an amazing complex of specialized populations.

Fig. 1.

(A) Whole-genome genealogy inferred for 159 Saccharomyces cerevisiae strains. The tree was inferred from 313,973 SNPs using the maximum-likelihood method as implemented in RAxML with the GTRGAMMA model of sequence evolution. The tree was rooted according to the midpoint method. Support values from bootstrap replicates >90% are indicated with red dots. Main lineages or populations are indicated with colored ellipses. (B) Population structure inferred with admixture from the same data set for a different number of ancestral clusters (2–11). The best partition was obtained for K = 11 ancestral clusters. (C) Population structure obtained with Fine Structure after data phasing. Med. Oak: Mediterranean Oak

Fig. 2.

(A) Repartition of the different new regions among available genomes. (B) Graphical representation of the origin specificity of selected chromosomal regions. For each chromosomal region statistically linked to a specific set of origins, we created an individual pie graph where the percentage of different origins is represented as a colored slice. The slices representing the most frequent origin are slightly separated from the others. Region’s pies are ordered based on the function of the most abundant origin.

Fig. 3.

PCA performed on the set of variants with potential impact according to SIFT.

Fig. 4.

Manhattan plot presenting the output of different statistics for sweep detection. Points above the threshold (dotted circles) are shown in red. (A) Output of Tajima’s D, Fay and Wu H, SVD, iHS, omega for the wine strain population. (B) Output of Tajima’s D, Fay and Wu H, SVD, iHS statistics for the cheese strain population.

Fig. 5.

Plots of the Kaplan–Meier product limit estimates of the efficiency of strains from a defined origin to achieve wine fermentation. The curves show the evolution over time of the probability of incomplete fermentation.

Fig. 6.

Growth on galactose as a carbon source for strains from different populations (A) Growth kinetics of all strains: cheese strains are indicated in pink and strains of other origins in blue. (B) Mean specific growth rate per population. The global significance of the differences between groups was evaluated by an ANOVA P value <6.7×10⁻⁶, and pairwise differences were assessed by the Tukey test: Cheese—Med oak <0.04 and Cheese—other groups <0.01. Differences between groups other than cheese were not significant.