Flor Yeasts Rewire the Central Carbon Metabolism During Wine Alcoholic Fermentation

Abstract

The identification of natural allelic variations controlling quantitative traits could contribute to decipher metabolic adaptation mechanisms within different populations of the same species. Such variations could result from human-mediated selection pressures and participate to the domestication. In this study, the genetic causes of the phenotypic variability of the central carbon metabolism of Saccharomyces cerevisiae were investigated in the context of the enological fermentation. The genetic determinism of this trait was found out by a quantitative trait loci (QTL) mapping approach using the offspring of two strains belonging to the wine genetic group of the species. A total of 14 QTL were identified from which 8 were validated down to the gene level by genetic engineering. The allelic frequencies of the validated genes within 403 enological strains showed that most of the validated QTL had allelic variations involving flor yeast specific alleles. Those alleles were brought in the offspring by one parental strain that contains introgressions from the flor yeast genetic group. The causative genes identified are functionally linked to quantitative proteomic variations that would explain divergent metabolic features of wine and flor yeasts involving the tricarboxylic acid cycle (TCA), the glyoxylate shunt and the homeostasis of proton and redox cofactors. Overall, this work led to the identification of genetic factors that are hallmarks of adaptive divergence between flor yeast and wine yeast in the wine biotope. These results also reveal that introgressions originated from intraspecific hybridization events promoted phenotypic variability of carbon metabolism observed in wine strains.

Keywords: QTL; Saccharomyces cerevisiae; alcoholic fermentation; flor yeast; linkage analysis; quantitative genetic; wine fermentation; yeast.

Figure 1

Experimental design. (A) Overview of yeast central carbon metabolism during fermentation with the main carbon input and output. (B) Segregant population, genetic map and phenotypic conditions used for QTL mapping.

Figure 2

Linkage analysis leads to the identification of 14 QTLs. Linkage analysis results for the CO₂max, Glycerol and MAC% for chromosome with at least one QTL. Horizontal lines represent the threshold of significance according to permutation test (FDR = 5%). Vertical lines highlight QTL peaks. Gray shadow encompasses the previously identified QTL hotspot containing PDR1, MSB2 and PMA1 (Martí-Raga et al., 2017).

Figure 3

Results of the reciprocal hemizygosity analysis. Boxplot are colored according to the allele present in the hemizygous hybrids (blue = both, red = GN and green = SB) and represented the dispersion of at least five biological replicates. A Wilcoxon–Mann–Whitney test was applied to assess the significance of the phenotypic difference between hemizygotes. The level of significance is indicated as follows: ^*p ≤ 0.1, ^**p ≤ 0.05, ^***p ≤ 0.01 and ^****p ≤ 0.001. (A) RHA result for glycerol. (B) RHA result for MAC%.

Figure 4

SB is closely related to flor yeasts. (A) Dendrogram using 385,678 SNPs from 405 wine strains. Flor yeasts group is highlighted. (B) Percentage of specific allele own by SB along the genome is represented by a gradient from dark blue (0%) to light blue (100%). Gray portions represent genome tracks without any flor yeast specific allele. SB is aneuploid for chromosome IX and therefore is not considered in this analysis. The 20 QTLs mapped are shown with red dots (some of them are overlapping) and validated genes are shown in green.

Figure 5

Proteomic analysis reveals the outlier behavior of the SB strain. (A) We reanalyzed a proteomic dataset previously obtained by shotgun quantitative proteomics (Blein-Nicolas et al., 2015). Yeast samples of 25 S. cerevisiae strains including SB and GN were collected at mid fermentation of a Sauvignon blanc grape juice. A set of 1,110 proteins common to all the strain was selected for analyzing strain relationships by a principal component analysis. The first two components representing 34% of the total inertia illustrate that the proteome of the strain SB (blue point) is quite divergent from the other S. cerevisae strains including GN (red point). (B) The functional interactions between 207 differentially expressed proteins and the eight QTG validated in this study was interrogated by using STRING algorithm. The three clusters encompassed 2, 4 and 31 proteins showing a strong functional interaction with the four causative genes PYC2, MAE1, MSB2 and SDH2 (black crosses). Active interactions were computed using the STRING algorithm on the base of experimental data and annotated database with a minimal interaction score of 0.8. Proteins were colored according to their mitochondrial origin (red), their involvement in pyruvate metabolism (blue) or in neo glucogenesis (green).

Figure 6

Relative position of the eight QTG in the metabolic map of S. cerevisiae. The metabolic relationships between the eight causative genes identified in this study is presented. Genes impacting glycerol production are represented in green while genes impacting MAC% are presented in blue.