ORIGINAL RESEARCH article
Sec. Fungal Genomics and Evolution
Flor Yeasts Rewire the Central Carbon Metabolism During Wine Alcoholic Fermentation
- 1Unité de Recherche Œnologie EA 4577, USC 1366 INRA, Bordeaux INP, ISVV, Université de Bordeaux, Bordeaux, France
- 2Biolaffort, Bordeaux, France
- 3Université de Strasbourg, CNRS, GMGM UMR 7156, Strasbourg, France
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.
Deciphering how the considerable phenotypic diversity observed at the species level is controlled by genetic variation is an important and non-trivial goal in biology. Improving knowledge regarding genotype–phenotype relationship provides information on evolution and adaptation mechanisms (Olson-Manning et al., 2012) and is precious in many biological fields like medicine (Minikel et al., 2020) or food industry (McCouch, 2004; Marullo et al., 2006; Sharmaa et al., 2015). Unraveling the genetic basis of adaptation highlights how organisms adapt to new selection pressures triggered by climate change, new pathogens or drugs and vaccines (Olson-Manning et al., 2012; Alföldi and Lindblad-Toh, 2013). Domestication is a specific case of adaptation with important phenotypic changes emerging from human artificial selection. Domesticated organisms are a great opportunity to study adaptation as there is a better knowledge of their adaptive history through their well-characterized phenotypic properties and selective environments (Ross-Ibarra et al., 2007; Gladieux et al., 2014). The identification of genes and molecular mechanisms leading to adaptation against domestication is also very useful in genetic selection in order to improve traits of economic interest and bringing phenotypic novelty to domesticated species (McCouch, 2004).
The yeast Saccharomyces cerevisiae rapidly emerged as an excellent model to study genotype–phenotype relationship (Brem et al., 2002; Steinmetz et al., 2002) and plenty of quantitative genetic studies were carried out in this species to study epistasis (Sinha et al., 2006), missing heritability (Bloom et al., 2013), gene-environment interaction (Smith and Kruglyak, 2008; Bhatia et al., 2014; Yadav et al., 2016; Peltier et al., 2018b) or impact of rare variants (Bloom et al., 2019; Fournier et al., 2019). S. cerevisiae was subjected to multiple domestication events in association with a large number of human associated environments (wine, beer, bread etc.) leading to distinct phylogenetic groups (Sicard and Legras, 2011; Legras et al., 2018; Peter et al., 2018). Several genetic marks of adaptation were identified such as gene loss of function (Will et al., 2010), translocations (Pérez-Ortín et al., 2002; Zimmer et al., 2014), introgressions (Novo et al., 2009; Marsit et al., 2015), and SNPs (Peltier et al., 2019) (see for review: Giannakou et al., 2020). Flor and wine yeasts are both associated with wine making environment and form two distinct but closely related phylogenetic groups (Legras et al., 2018). While both groups are able to efficiently perform wine fermentation, flor yeasts used in Sherry-like wines have the specific ability to shift to oxidative metabolism and form a velum covering wine surface after fermentation (Legras et al., 2016). Differences in genomic content between wine and flor yeast were observed and the impact of allelic variations involved in biofilm formation were proposed as a feature of genetic adaptation (Fidalgo et al., 2006; Coi et al., 2017). Other functional adaptation hallmarks related to active gluconeogenesis, response to osmotic pressure and metal transport were predicted by a population genomic approach but have not been demonstrated yet at the gene level (Coi et al., 2017).
Recent global warming caused the steady increase of sugar content in grape juices leading to higher ethanol concentration in wine with several issues regarding consumer health and wine quality (Keller, 2010; Kutyna et al., 2010; Mira de Orduña, 2010). Therefore, there is a growing demand for the development of new technologies to reduce alcohol content in wine. In this context, several institutions have attempted a biological approach in order to select new strains of S. cerevisiae with a lower fermentation yield. Various strategies were implemented such as adaptive evolution (Kutyna et al., 2012; Tilloy et al., 2015), interspecific breeding (da Silva et al., 2015), and genetic engineering (Ehsani et al., 2009; Rossouw et al., 2013). Here, we aim at finding out undescribed natural genetic variations controlling the central carbon metabolism in order to modulate the efficiency of sugar into ethanol conversion (Fermentation yield). By applying a Quantitative Trait Loci (QTL) mapping approach, we investigated the genetic determinism of three traits (glycerol production, CO2 production and malic acid consumption) that shape the carbon balance in enological conditions. Glycerol and CO2 production are direct markers of ethanol production. During glucose fermentation one molecule of CO2 is produced per molecule of ethanol produced. Regarding glycerol, a part of the glycolytic flux is deviated toward glycerol production, impeding total conversion of the glucose to ethanol. Malic acid consumption is also linked to ethanol production via the malo-ethanolic fermentation (Volschenk et al., 2003).
Our study is based on the analysis of a progeny obtained by crossing two strains derived from wine starters. A deeper analysis of parental genomes showed that, unexpectedly, one of the parental strains results to have a mosaic genome inherited from both wine and flor yeasts while the second parental strain belongs to the wine group. This admixture has promoted an important phenotypic variability impacting the central carbon metabolism of the F1 progeny. A total of 14 QTLs were identified and the effect of eight of them were experimentally validated down to the gene level. Six genes (PMA1, PNC1, PYC2, SDH2, MAE1, and MSB2), among which three are directly involved in central carbon metabolism (SDH2 in tricarboxylic acid cycle (TCA)), MAE1 in pyruvate metabolism and PYC2 in gluconeogenesis pathways, show allelic variations highly specific to flor yeasts group. Linked to these validated genes, further proteomic analyses highlighted different metabolic regulations between the parental strains for TCA and glyoxylate shunt. Altogether, these results support the hypothesis that allelic variations between wine and flor yeasts generate important phenotypic differences and could be considered as hallmarks of adaptation for different growth strategies on the wine biotope. These results also show that flor yeasts constitute a great reservoir of genetic variation to bring phenotypic novelty in commercial yeast starter to cope for new challenges as global warming (Mira de Orduña, 2010) and new viticultural practices (Kontoudakis et al., 2011).
Biometric Study of the Glycerol, CO2 and Malic Acid
In order to explore the genetic determinism of central carbon metabolism during wine alcoholic fermentation, the previous dataset of fermentation traits measured within a QTL mapping population was used (Peltier et al., 2018b). This population was obtained by mating two fully homozygous strains (SB and GN) derived from the sporulation of wine starters (Actiflore BO213 and Zymaflore VL1, respectively). A total of 94 meiotic segregants were obtained though sporulation of a single hybrid (SBxGN) (Figure 1) and phenotyped in three environmental conditions using a small-scale fermentation dispositive and enzymatic assays to measure fermentation kinetics traits and endpoint concentration of several metabolites, including glycerol and CO2 production. All segregants were sequenced and a genetic map of 3,433 biallelic markers was built in order to identify the genetic factors controlling these phenotypes (Supplementary Table 1). In the present study, an additional phenotyping effort was achieved by measuring malic acid consumption in the same conditions.
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.
Carbon balance was evaluated by measuring the main organic compounds assimilated and/or produced for each of the 94 segregants at the end of the alcoholic fermentation (Supplementary Table 2). According to the must, the fermentation yield computed ranged between 0.45 and 0.48 g/g which is close to values observed in other studies (Tilloy et al., 2014) (Supplementary File 1). An analysis of variance demonstrated a significant genetic (strain) impact on the fermentation yield (17% of the total variance explained). This integrative trait is mostly shaped by the quantitative variation of three metabolites: glycerol, malic acid, and CO2 that were partially correlated (Supplementary Figure 1). Glycerol and CO2 (which is stoichiometrically linked to ethanol) are de novo synthetized by yeast catabolism; their concentrations are expressed in g/L. The final concentration of CO2 produced is expressed hereafter as CO2max. The final concentration of malic acid depends on its initial amount in grape must which differs according to the grape juice. Since this organic acid is partially metabolized by yeast, the strain contribution was normalized by computing the percentage of Malic Acid Consumed (MAC%). For each trait, parental strains SB and GN are significantly different with important gaps for glycerol and MAC% (Wilcoxon test, p < 0.05). Indeed, SB produces 1.6 g/L more glycerol (+30%) and consumes 28% more malic acid than GN. Although significant, phenotypic difference for CO2 is more moderate. These differences are consistent with previous results showing that SB is the top strain for glycerol production and malic acid consumption compared to a panel of commercial starters (Peltier et al., 2018a). For glycerol and CO2max segregant values are in the same range than these commercial starters. However, 20 segregants are transgressing commercial strains values for MAC% with a consumption increased by 24% for the top segregant, showing potential for genetic improvement of commercial starters. Each trait had a high overall heritability (Supplementary Table 3) and displayed a bell-shaped distribution with number of segregants showing transgressive values respect to parental strains (Supplementary Figure 2). These broad biometric observations highlighted a polygenetic control of each trait with a positive contribution of both parental strains.
Linkage Analysis Brings Out a Linkage Hotspot With Pleiotropic Effect
In a previous work exploring genetic x environment interactions, five QTLs were associated with CO2max and glycerol production in the SBxGN offspring (Peltier et al., 2018b). Here, we aimed to identifying supplemental QTL controlling MAC% that was newly phenotyped. A linkage analysis was performed and significantly associated nine QTLs to this trait. Therefore, a total of 14 QTL are involved in CO2max, glycerol and MAC% (Figure 2; Supplementary Table 4). The effects of parental alleles are shown in the Supplementary Figure 3. Intriguingly, a large region of the chromosome VII (387–716 kb) was associated with all the considered traits. This linkage hotspot is almost entirely above the significance threshold for at least one trait and four distinct linkage peaks can be distinguished. This hotspot encompasses one major QTL, the locus VII_415 (Chr VII, position 415,719), influencing the glycerol production (LOD score >10) which explains more than 10% of total variance. Interestingly, for this cross, a sharper region of chromosome VII (50 kb) was previously associated with kinetic traits during second fermentation of sparkling wines (Martí-Raga et al., 2017). Three genes of this large QTL (PDR1, PMA1 and MSB2) were demonstrated to have an important phenotypic impact in this condition. Here, the QTL VII_482 linked to MAC% is located in the PMA1 coding sequence (479,910 482,666).
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).
Multiple Quantitative Trait Genes Control Glycerol Production and Malic Acid Consumption
Candidate genes neighboring the QTL peak within a 20 kb window were considered through their functional annotation and by checking for ns-SNPs within parental strains sequences using the algorithm SnpEff (Supplementary Table 5) (Sherman and Salzberg, 2020). We selected also the three genes (PDR1, PMA1 and MSB2) previously validated for second fermentation traits that are located near the major hotspot of chromosome VII in the present work. This leads to consider 11 candidate genes that could impact the traits investigated. Their effects were interrogated by a Reciprocal Hemizygosity Analysis (RHA) (Steinmetz et al., 2002). The impact of parental alleles was compared in alcoholic fermentation test using the same fermentation protocol. In addition, ethanol content (% Vol) was estimated by infrared reflectance rather than enzymatic assay (see methods). The effect of four candidate genes impacting CO2max and/or glycerol was tested. They belong to the two major QTLs found in term of variance explained: ADE6 (VII_616), MSB2 (VII_512), PDR1 (VII_482), PNC1 (VII_415). The RHA was carried out in the M15_sk condition with two sugar concentration levels (219 and 265 g/L) using at least five biological replicates for each condition. Sugar spiking would emphasize the phenotypic differences related to CO2 and ethanol production. The most obvious effects were obtained for glycerol production for genes ADE6, MSB2, and PCN1 for which hemizygous hybrids are significantly different (Wilcoxon test, p < 0.1) (Figure 3A). These three genes are located in a region of 200 kb along the chromosome VII hotspot demonstrating that distinct genetic factors in this region control the glycerol production.
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%.
Intriguingly, the sugar content modulated the phenotypic responses of hemizygous hybrids. Indeed, in sugar-spiked grape must (M15_265), alleles ADE6GN enhanced glycerol production of 12%, while the allele MSB2GN has an enhancer effect only in the original M15 grape must (219 g/L of initial sugar). The allelic forms ADE6GN, PNC1SB promote the glycerol production and their effects are those observed in the SBxGN progeny (Supplementary Table 4; Supplementary Figure 3). In contrast, the MSB2GN allele produced more glycerol which is not observed in the segregating progeny (Supplementary Figure 4). This opposite effect has been previously described for the same gene for another phenotype and could be due to the complex genetic architecture of chromosome VII (Martí-Raga et al., 2017). The difference observed in glycerol production for ADE6, PNC1 and MSB2 did not impact either the CO2max or the ethanol content (Supplementary Figure 5).
In the same way, seven candidate genes belonging to six QTLs affecting MAC% were evaluated: MAE1 (XI_381), MCH1 and GPM2 (IV_356), PYC2 (II_669), PMA1 (VII_482), SDH2 (XII_53) and YBL036c (II_152). Fermentations were carried out in both M15 and SB14. RHA revealed a significant effect for the genes MAE1, PMA1, PYC2 and YBL036c (Figure 3B) (Wilcoxon test, p < 0.05). The alleles of MAE1, PYC2 and YBL036c inherited from the parental strain SB consumed respectively 25, 19, and 45% more malic acid than those inherited from GN. In contrast, the PMA1GN allele consumed 18% more malic acid than PMA1SB. This gene, encoding for the plasma membrane ATPase, has been previously linked to the maintenance of pH homeostasis during wine fermentation and is located in the center of chromosome VII hotspot (Martí-Raga et al., 2017). Unexpectedly, a significant effect of PNC1 on MAC% was also observed and the hemizygote hybrid harboring the PNC1SB allele consumes 15% more malic acid than PNC1GN (Figure 3B) (Wilcoxon test, p < 0.05). The genomic position of PNC1 is about 50 kb from the nearest QTL peak for (MAC% VII_482), however the other causative genes (PMA1, MSB2, ADE6) associated with the chromosome VII hotspot may have altered the precision of our linkage analysis. The difference in MAC% for MAE1, PMA1, PYC2 and YBL036c did not impact CO2max (Supplementary Figure 5).
Beside the validation of these five genes on MAC%, reciprocal hemizygous analysis of SDH2 suggested its potential contribution on malic acid consumption. Although the hemizygous are not statistically different, a strong haploinsufficiency effect in both hemizygous hybrids was observed affecting either MAC% (−14%) and fermentation kinetics by doubling the fermentation duration (Supplementary Figure 6). Intriguingly, this haploinsufficiency was only present in M15 grape juice. Two factors suspected to have an impact on this haploinsufficiency were tested (initial malic concentration and pH) in synthetic grape juice (SGJ) by adjusting these two initial values to either M15 or SB14 levels. An haploinsufficiency similar to that in M15 was found in all four conditions even in the one mimicking SB14 conditions (Supplementary Figure 6). No significant interaction between the level of haploinsufficiency and pH and malic acid was found (ANOVA, p > 0.1). These findings suggest that SDH2 has a great impact on fermentation rate and MAC% during grape juice fermentation. However, since the RHA test was limited by the haploinsufficiency effect our experiments failed to clearly demonstrate the impact of parental allelic variations.
Altogether, these functional analyses validated the role of eight Quantitative Trait Gene (QTG). Four of them play a direct role in the central metabolism encoding enzymes involved in oxidoreductive reactions of carbohydrate metabolism (MAE1, PYC2, PNC1, SDH2). Two others are key regulators of osmotic (MSB2) and pH (PMA1) homeostasis. The RHA also revealed that ADE6 and YBL036c contribute to the phenotypic difference between the parental strains for glycerol production and malic acid consumption, respectively (Figure 3). However, their functional connection with the metabolic pathway of glycerol and malic acid is more difficult to address at this stage.
SB Is a Mosaic Strain Derived From Flor and Wine Yeasts
QTL mapping is a useful strategy for identifying natural genetic variations that shape phenotypic diversity between two strains. However, in most of the cases, the causative mutations identified are rare and specific to one parental strain (Bloom et al., 2019; Fournier et al., 2019; Peltier et al., 2019) due to the clonal structure of S. cerevisiae population (Peter et al., 2018). This impairs the identification of more general mechanisms of adaptation resulting to natural selection. In order to have a more precise idea of the evolutive relevance of QTL identified, SB and GN genomes were compared to those of 403 wine related strains previously released (Legras et al., 2018; Peter et al., 2018). A phylogenetic tree was generated using 385,678 SNPs discriminating the 403 wine strains plus the parental strains SB and GN. This collection of strains encompasses wine (n = 358) and flor (n = 47) strains that form distinct groups as previously described (Coi et al., 2017; Legras et al., 2018) (Supplementary Table 6). Interestingly, SB is genetically close to the flor group while GN is quite similar to the wine group (Figure 4A). Consequently, the two parental strains used in this study are quite distant with a sequence divergence of 0.19% (~22,000 SNPs). The relatedness of SB genome with the flor group was deeply investigated by selecting a subset of 5,086 SNPs highly specific to the flor yeast group. Those SNPs have a frequency difference higher than 90% between flor and wine yeast groups. The strain SB harbors 44.3% of flor yeast specific alleles while GN only has 1.7% of them. Their distribution across the SB genome is not uniform (Figure 4B). Indeed, long portions of chromosomes have inherited 100% flor-specific alleles (Chr II) while other portions are totally exempt of them (Chr VIII). This analysis demonstrated that SB is a mosaic strain between wine yeast and flor yeast, a feature shared with some others wine starters (Coi et al., 2017).
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.
Intriguingly, nine of the 14 QTLs mapped are located in flor specific chromosomic portions. This is the case of a large stretch within chromosome VII encompassing four causative genes (PNC1, MSB2, PMA1, ADE6) that displays the genomic signature of flor yeasts. A similar observation can be made for chromosome II in which three QTLs were identified (Figure 4B). During their domestication, flor yeasts accumulated numerous mutations leading to an adaptation to grow on wine surface (Coi et al., 2017). In order to narrow such natural genetic variations, we listed the pool of ns-SNP discriminating SB and GN in the sequence of causative genes. For those SNPs, allelic frequencies of flor and wine groups were computed (Table 1). In ADE6, ns-SNPs listed are scarcely found whatever the group. The low allelic frequency of such polymorphisms would reflect recent mutations which is a common feature of the S. cerevisiae population. In contrast, for the other genes PMA1, PNC1, PYC2, SDH2, MAE1, and MSB2, the SB alleles are highly specific to flor yeast group while GN alleles are specific to the wine group. Therefore, these flor-specific alleles would have promoted the wide phenotypic variability of carbon metabolism observed in SBxGN progeny and more broadly are explaining phenotypic differences between flor and wine yeasts.
SB Proteome Reveals Peculiar Metabolic Regulations Functionally Connected With Some Causative Genes
Flor yeasts are able to grow on the wine surface at the end of the alcoholic fermentation. By creating biofilm rafts, they are able to resist to high ethanol content in harsh conditions (Legras et al., 2016). For ensuring their development, they activate particular metabolic pathways (active neoglucogenesis and respiration metabolism) that are the opposite of those developed by wine yeasts during the alcoholic fermentation. Such metabolic differences have been previously reported at the metabolomic and the proteomic levels (Alexandre, 2013; Moreno-García et al., 2015a,b; David-Vaizant and Alexandre, 2018). In order to have a broad overview of the metabolic peculiarities of the SB strain, we reanalyzed a proteomic dataset previously generated in our laboratory (Albertin et al., 2013; Blein-Nicolas et al., 2013, 2015). Data explored were obtained by quantifying the proteome of 25 S. cerevisiae strains, including SB and GN, during the fermentation of a sauvignon blanc grape juice by a shotgun proteomics approach. Samples were collected at mid-point in triplicate allowing the quantification of 1,110 proteins commonly expressed (Supplementary Table 7). A global Principal Component Analysis (PCA) demonstrates that SB is strongly discriminated by the two principal axes accounting for 34% of the total inertia suggesting an outlier protein abundance respect to 24 other strains (Figure 5A). Indeed, the Abundance Fold Change Ratio (AFCR) of SB and GN vs. the 24 other strains were compared for each of the 1,100 proteins quantified. SB displays a much distinct profile since 12.9% of its proteome reach a two folds change abundance (log2(AFCR) ± 1.0) while only 2.9% of GN proteins reach this threshold (Supplementary Figure 7). Thus, proteome variance of SB and GN are 0.504 vs. 0.143, respectively (variance F test, p < 1 × 10−16). This analysis demonstrated that SB has a particular proteome compared to GN and even to other S. cerevisiae strains.
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).
In order to analyze the origin of this discrepancy, we deeply compared SB and GN using the 1,264 proteins quantified in both strains (Supplementary Table 7). This comparative analysis reveals a set of 207 proteins with an ACFR higher than 2 (Supplementary Table 8). Within this set, a significative enrichment was found for mitochondrial proteins which represent 33% of the pool (χ2 test = 2 × 10−5). We sought functional interactions between the eight causative genes identified and the set of 207 differentially expressed proteins by performing a STRING analysis (Szklarczyk et al., 2019) (see Methods). Three of the six interaction networks computed clearly linked four QTG with proteins differentially expressed (Figure 5B). The main cluster, linked to the causative genes PYC2 and MAE1, encompassed 31 proteins including many enzymes related to pyruvate and citrate metabolism (Mls1p, Leu9p, Ach1p, Mdh3p, Dld1p, Dld2p, Ald5p, Cyb2p, Cit1p, Cit2p). The fold change abundance of such proteins suggests the existence of differential metabolic regulations between SB and GN. For instance, three of the four S. cerevisiae enzymes (Dld1p, Dld2p and Cyb2p) involved in the lactate metabolism are at least 2.5 less abundant in SB. These proteins are supposed to be repressed by glucose and anaerobiosis and participate to the oxidation of lactate into pyruvate (Becker-Kettern et al., 2016). Other proteins, belonging to the glyoxylate shunt and TCA, were differentially quantified (two-fold change ratio). Interestingly, the oxidative branch of TCA and the glyoxylate shunt (i.e., Mls1p, Dal7p, Cit1p, Cit2p, Aco2p) are broadly more abundant in SB while proteins participating to the reductive branch of TCA (i.e., Fum1p, Mdh1, Sdh2p) are more abundant in GN (Supplementary Figure 8A). These metabolic pathways are directly connected with two causative genes identified in this study MAE1 and PYC2 that controls MAC%. Strikingly, the cytosolic malate synthase Mls1p catalyzing the condensation of glyoxylate and acetyl CoA in L-malate is seven-folds more abundant in SB [log2(AFCR)>2.8] and would directly enhance its cytosolic pool of malic acid. These noteworthy variations of proteins abundance are not due to a singular contrast between SB and GN proteomes but reflect a clear specificity of SB central metabolism regulation. Indeed, the AFCR computed between SB and the 24 other S. cerevisiae strains (average value) is very similar to the AFCR of SB vs. GN (Pearson cor. test <10−13) (Supplementary Figure 8B). This analysis suggests that the peculiar proteome of SB would be due to its unusual mosaic origin encompassing large stretches of flor yeast genome.
Identification of New Allelic Variations With Potential to Improve Wine Yeast Starters
This work aimed to identify natural genetic variations that possibly modulate the catabolism of carbon sources during wine fermentation. From an applied point of view, this goal is particularly relevant for wine industry in order to cope with two main negative effects of global warming: (i) the rise of ethanol content and (ii) the reduction of the total acidity of wines. This general trend is due to the increasing concentration of sugars coupled with a drop of malic acid content in grape juices around the world (van Leeuwen and Darriet, 2016). By applying a QTL mapping strategy, eight Quantitative Trait Genes (QTG) impacting the carbon balance during the wine fermentation were identified. This study allows the identification of natural allelic variations controlling two remarkable phenotypes: the glycerol production and the percentage of malic acid consumed (MAC%). The schematic relationships of their respective proteins in the yeast metabolism map are shown on Figure 6. Despite their effect on glycerol or malic acid consumption, reciprocal hemizygosity assay fails to identify candidate genes that significantly decrease the wine final ethanol content. This could be explained by the slight effect of individual QTG and the lack of precision of the ethanol measurement method. Regarding glycerol production, PNC1 has the larger effect with a difference of 0.85 g/L between hemizygotes. This would theoretically result in a hardly measurable change of 0.05% on the final ethanol content. However, aggregating favorable alleles in one strain using marker assisted selection should lead to substantial effects (Vion et al., 2021). Regarding malic acid, its consumption is not only linked to ethanol production and can be directed to other metabolic pathways like the aspartate biosynthesis and therefore attenuates its impact on ethanol production. Direct measure of the segregants ethanol production would help to identify QTL with higher and more direct impact on ethanol production. However, no low-volume methods were enough precise and high throughput to be in accordance with the scale of this experiment. Beyond the objective of decreasing wine ethanol content, the QTG that modulate malic acid consumption are also of great interest. Indeed, as malic acid is a major contributor of wine acidity, these QTG have a significant impact on wine pH and are already used in marker assisted selection programs to control wine acidity (Vion et al., 2021).
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.
The Flor Yeast Origin of the Parental Strain SB Is Likely Involved in the Diversity of Carbon Catabolism in the SBxGN Progeny
This study was carried out using two meiotic segregants (SB and GN) derived from commercial starters widely used in wine industry (Actiflore BO213 and Zymaflore VL1, Laffort, France). Such commercial starters have been selected in the past for their technological properties by sampling spontaneous wine fermentations (P Marullo, pers com). Unexpectedly, we find out that the SB genome has a mosaic structure inherited from two distinct groups of S. cerevisiae population: the wine and the flor yeasts (Peter et al., 2018). Around 40% of the SB genome is flor specific suggesting that BO213, the parental strain of SB, would be an F1-hybrid resulting from the cross of a flor yeast and a wine yeast, as previously observed for others wine commercial strains related to the Champagne group (Coi et al., 2017).
Wine yeasts are adapted to a fast development on grape must in competition with numerous other species in a sugar rich environment and many natural allelic variations related to their adaptation to grape juice have been described in the past (Peltier et al., 2019). In contrast, flor yeasts are adapted to survive in wine, a sugar-depleted environment containing high ethanol degree and low oxygen. Thus, flor yeasts would have accumulated specific genetic variations for coping with this harsh environment. Many efforts have been made for identifying such adaptation signatures especially concerning the development of the flor velum. This biofilm-like growth is essential for reaching the wine surface and to get oxygen which is mandatory for catabolizing ethanol and producing energy (Legras et al., 2016). Allelic variations specific to flor yeasts have been detected by using comparative genomics and the role of two genes (SFL1 and RGA2) participating in the regulation of FLO11 has been demonstrated (Coi et al., 2017). In the SBxGN cross, wine and flor specific alleles segregate providing the opportunity to study the phenotypic impact of gene pools that have undergone independent evolutionary routes with different selective pressures. Indeed, 9 of the 14 QTL identified are located in flor specific regions allowing the molecular validation of six genes (PMA1, PNC1, PYC2, SDH2, MAE1, and MSB2) characterized by flor specific alleles. This suggest that part of the allelic variations involved in the adaptive divergence between wine and flor yeast had been captured. In this study one parental strain unexpectedly appeared to be a mosaic strain between wine and flor yeasts and revealed specific genetic adaptations. In the future, incorporation of pure strains belonging to the flor yeast lineage in the experimental design will help to capture a broader part of these adaptations. Among these adaptations some may have detrimental effects in classic wine fermentation. In this context, the QTL mapping experimental design allow to track them down to the gene level and to select those that can bring positive phenotypic impact to classic wine starters.
Identified Genes Are Closely Linked to Pathways Discriminating Flor and Wine Yeast Metabolisms
First, MSB2 encodes a signaling mucin protein acting as a stress or nutrient deprivation receptor (Cullen and Sprague, 2012). Msb2p is associated with the transmembrane osmosensor Sho1p and transmits the signal to the downstream components of the monomeric G-proteins Rho involved in both filamentous growth (FG) and the high osmolarity glycerol (HOG) pathways (Tatebayashi et al., 2007). HOG pathway plays a key role for adaptation against high osmolarity levels by increasing the production of glycerol (Hohmann, 2009), the second more abundant metabolite of fermenting yeast after ethanol. The comparative analysis of MSB2 sequence reveals a unique ns-SNP between the parental strains (Table 1). The SB allele S529F is specific to flor yeasts and lowers the glycerol production with respect to the GN allele. The MSB2S529F allele has a predicted deleterious effect that would impact the signal transduction of both HOG and FG MAPK pathways. Such pathways share common components but are induced by different stimuli and provides specific responses (Pitoniak et al., 2009). The essential Rho protein Cdc42p has been described to stimulate glycerol production by triggering the MAPK Hog1p (Hohmann, 2009). Cdc42p is threefold less abundant in SB which is consistent with the hypothesis of a low Msb2p activity in this background. In contrast the non-essential GTPase Rho3p also involved in cell polarity is three times more abundant in SB. Interestingly, the abundance fold ratio of Rho3p and Cdc42p are specific to SB (compared to others S. cerevisiae strains) and might be related to the filamentous growth specificities of flor yeast required for the velum formation.
A flor-specific allele was also found in the sequence of PNC1 which encodes for a nicotinamidase that converts nicotinamide to nicotinic acid. Pnc1p, which is induced by the osmotic stress, restores redox balance by regenerating NAD+ from nicotinamide via the NAD+ salvage pathway (Ghislain et al., 2002; Effelsberg et al., 2015). RHA reveals that the allele PNC1SB enhances both the glycerol production and the MAC%. A direct functional link exists between PNC1 and glycerol biosynthesis since this protein is co-imported in the peroxisome with Gpd1p, a major controlling enzyme of glycerol biosynthesis (Nevoigt and Stahl, 1997). Under osmotic stress, their overexpression saturates the peroxisome importation system and therefore this protein became cytosolic and active (Effelsberg et al., 2015). The role of Pnc1p in MAC% is more complex to explain and might be linked to the NAD+/NADH+ homeostasis itself that is tightly controlled (Bakker et al., 2001). This organic acid can be oxidized in pyruvate (by the malic enzyme Mae1p) or in oxaloacetate (by malate dehydrogenases). Thus, an active malic acid consumption would increase the intracellular levels of NADH+ requiring an increase of glycerol production for regenerating the NAD+ pool.
Another flor yeast specific allele impacting MAC% is MAE1 that encodes for the mitochondrial malic enzyme that catalyzes the oxidative decarboxylation of malate to pyruvate (Boles et al., 1998) achieving the malo-ethanolic fermentation (Volschenk et al., 2003). Interestingly, MAE1 was also reported to influence the formation of higher alcohols, fusel acids, and acetate esters in another mapping population where the same SNP is segregating (MAE1I605V) (Eder et al., 2018). These data suggest that this allelic variation would have pleiotropic consequences in an enological context, affecting the malic acid consumption as well as the biosynthesis of relevant wine volatile compounds.
A second pleiotropic gene to be discussed is PMA1 which encodes for a membrane ATPase the major regulator of cytoplasmic pH and plasma membrane potential. During wine fermentation, pH has a great impact on intracellular malic acid diffusion and consumption (Salmon, 1987; Delcourt et al., 1995; Saayman and Viljoen-Bloom, 2017). Indeed, malic acid charge is strongly dependent of the wine pH since the pka1 of this diacid is 3.54. Bellow a pH value of 3.4, the entry of a malic acid molecule in the cytoplasm result to a net proton influx that must be pumped over for maintaining pH homoeostasis with an energy cost of 1 ATP per molecule. In the present work, the QTL VII_482 related to PMA1 has the strongest effect observed with a positive impact of the GN allele on malic acid consumption. Previously, we demonstrated that PMA1 inheritance influences fermentation kinetics with a strong interaction with the pH of the medium. Indeed the GN and SB alleles increase the fermentation rate when the pH are 3.3 and 2.8, respectively (Martí-Raga et al., 2017). These fine grain gene-environment interactions might result from the consumption level of malic acid in relation with the pH of wine.
Two other genes with a direct connection with malic acid metabolism were illuminated. The gene PYC2 involved in gluconeogenesis pathway encodes for a pyruvate carboxylase that converts pyruvate to oxaloacetate (Stucka et al., 1991; Walker et al., 1991). During fermentation, pyruvate carboxylase is the sole source of oxaloacetate playing an essential role in aspartate biosynthesis, TCA turnover, and malic acid biosynthesis (Huet et al., 2000). Indeed, PYC2 overexpression enhances malic acid production in a bioengineering context (Bauer et al., 1999). We hypothesized that the allelic variants of SB may have reduced the Pyc2p activity reducing the biosynthetic flux of malic acid from pyruvate. To cope with this reduction, a first metabolic alternative would be the de novo synthesis of malic acid from the glyoxylate shunt. This is consistent with the high abundance of the malate synthase (more than seven-folds) observed in SB respect to GN. A second metabolic alternative would be a strongest uptake from the external media which is the hallmark of the SB strain.
Finally, a surprising effect of SDH2 deletion was observed. This gene encodes for a subunit of the succinate dehydrogenase complex (complex II) ensuring electron transfer from succinate to ubiquinone. This TCA cycle step is involved in the mitochondrial respiratory chain and is mostly inactive during the alcoholic fermentation (Camarasa et al., 2003) due to oxygen depletion and catabolic repression (Klein et al., 1998; Kwast et al., 1998). Indeed, under sake brewing conditions, the CO2 production rate was not impacted in double mutants Δsdh1, Δsdh2 (Kubo et al., 2000). These commonly admitted results contrasted with the strong haploinsufficiency effect of SDH2 deletion observed for MAC% and fermentation kinetics in M15 medium (Supplementary Figure 6). Although we could not measure a significant difference between hemizygous hybrids, the strong haploinsufficiency observed suggests that the succinate dehydrogenase complex would play an unsuspected physiological role in this specific background. Interestingly STRING analysis reveals that five proteins functionally associated to SDH2 are differentially synthetized between SB and GN. These proteins belong to the respiratory complexes II, III and IV. Thus, complex II (Sdh1p and Sdh2p) is less abundant in SB while proteins belonging to complex III (Qcr10p) and IV (Cox2p and Cox12p) are more abundant. Due to the functional importance of protein stoichiometry in such complexes, abundance change in few proteins would impact the residual activity of the respiratory chain. Therefore, the functional understanding of the succinate dehydrogenase complex during alcoholic fermentation will require further analyses that are not the purpose on the present paper.
Identified QTG Contribute to Divergent Central Carbon Metabolism
Flor yeasts exhibit an active gluconeogenesis and respiration catabolism during velum development that impact their proteomics response (Alexandre, 2013; Legras et al., 2016) (Moreno-García et al., 2015b). However, to our knowledge, a comparative proteomic study between flor and wine yeast was never achieved. Since the SB strain harbor 40% of the genomic signature of a flor yeast, we supposed that this strain could exhibit particular flor yeast features at the proteomic level. Indeed, SB showed a specific proteomic pattern compared to all the other strains. As SB has a mosaic genome, a comparison with pure strains from the flor yeast lineage will enhance our knowledge regarding these differences. Several proteins involved in pyruvate and gluconeogenesis were differentially quantified. Many of them have been previously described as specific signature of velum development, confirming that SB specific pattern is a marker of flor yeast specificities (Moreno-García et al., 2015b).
By implementing a STRING analysis, we attempted to retrace a functional link between the eight QTG identified and the proteomic variations observed between parental strains. This indirect analysis would bridge the gap between specific flor yeast variations and the overall proteomic discrepancy of the SB strain. Three causative genes (MSB2, SDH2 and PYC2) harboring flor specific alleles were functionally connected with three protein clusters (Figure 5B). PYC2 and SDH2 are directly involved in central carbon metabolism playing an essential role in gluconeogenesis and respiration, respectively. The first controls the unique way for producing glucose from ethanol since the pyruvate kinase catalyzed an irreversible reaction (Pronk et al., 1996). The second belongs to the succinate dehydrogenase which is inactivated during the fermentation and that constitutes the first step of respiration chain (complex II) which is essential for producing energy in aerobic conditions. A contrasted regulation between the oxidative and reductive branch of TCA was observed in the strain SB (Supplementary Figure 8A) promoting the idea that succinate dehydrogenase activity would participate to the regulation of TCA proteome. Although this hypothesis remains to be validated by further experiments, we hypothesized that the specific flor alleles Sdh2K158E and Pyc2Q373K carried by SB strain might impact the overall proteomic response of this strain by controlling key steps of gluconeogenesis and TCA cycle.
Materials and Methods
Yeast Strains and Culture Media
All the strains used in this study belong to the yeast species S. cerevisiae. SB and GN strains are monosporic clones derived from industrial wine starters, VL1 and Actiflore BO213, respectively. Generation of the SBxGN and segregant populations were described by Peltier et al. (2018b). Briefly, F1-hybrids were obtained by manual crossing with micromanipulator. After sporulation on ACK (2% potassium-acetate, 2% agar) media, monosporic clones were isolated by micromanipulation. Yeast was cultured at 30°C in yeast YPD media (10 g/L yeast extract, 20 g/L peptone and 20 g/L glucose) and solidified with 2% agar when required. The strains were stored long term in YPD with 50% of glycerol at −80°C.
The two grape juices used, Merlot of vintage 2015 (M15) and Sauvignon Blanc of vintage 2014 (SB14), were provided by Vignobles Ducourt (Ladaux, France) and stored at −20°C. Before fermentation, grape juices were sterilized by membrane filtration (cellulose acetate 0.45 μm Sartorius Stedim Biotech, Aubagne, France). Fermentations were carried out as previously described (Peltier et al., 2018a). Briefly, fermentations were run at 24°C in 10 ml screw vials (Fisher Scientific, Hampton, New Hampshire, USA) with 5 ml of grape must. Hypodermic needles (G 26–0.45 × 13 mm, Terumo, Shibuya, Tokyo, Japan) were inserted through the septum for CO2 release. Two micro-oxygenation conditions were used by applying or not constant orbital shaking at 175 rpm during the overall fermentation. For this data, three fermentation conditions were used: SB14 with shaking (SB14_Sk), M15 with shaking (M15_Sk) and M15 without shaking (M15). Fermentation progress was estimated by regularly monitoring the weight loss caused by CO2 release using a precision balance. The amount of CO2 released over time was modeled by local polynomial regression fitting with the R-loess function setting the span parameter to 0.45. From this model CO2max parameter was extracted: maximal amount of CO2 released (g/L) and the end of the fermentation. Fermentation conditions were described by (Peltier et al., 2018b). Glycerol and malic acid concentration were determined by enzymatic assay (Peltier et al., 2018b) using K-GCROLGK and K-LMAL-116A enzymatic kits (Megazyme, Bray, Ireland), following the instructions of the manufacturer.
The QTL mapping analysis was performed with the R/QTL package (Broman et al., 2003) on the data collected in the three environmental conditions by using the Haley-Knott regression model that provides a fast approximation of standard interval mapping (Haley and Knott, 1992). The analysis is taking in account environment and cross as an additive covariate, aiming to identify QTL robust to environment and cross factor:
where yi is the phenotype for individual i, μ the average value, βgi the QTL genotype for individual i, Aγz the matrix of environment covariates (y = M15_Sk, SB14_Sk, M15) and ϵ the residual error. For each phenotype, a permutation test of 1,000 permutations tested the significance of the LOD score obtained, and a 5% FDR threshold was fixed for determining the presence of QTLs (Churchill and Doerge, 1994). The QTL position was estimated as the marker position with the highest LOD score among all markers above the threshold in a 30 kb window.
Hemizygous Hybrids Construction
For each QTL, candidate genes were sought in a 30 kb windows around the QTL position with the maximal LOD score. Genes with non-synonymous SNPs and/or with a function related to the trait of interest were retained. Candidate genes were validated by reciprocal hemizygosity analysis according to (Steinmetz et al., 2002) using SBxGN hybrid. Deletion cassettes were obtained by PCR amplification of the disruption cassette plus 500 pb of the flanking regions using as genomic template the genomic DNA of the strains Y04691, Y03717, Y04878, Y03751, Y04405, Y01529, Y03062 of the EUROSCARF collection (http://euroscarf.de), which contain disruption cassettes for the following genes: ADE6, GPM2, MAE1, MCH1, PNC1, PYC2, SDH2, YBL036C, respectively. Primers used for strains construction are listed in Supplementary File 2. Reciprocal hemizygotes for MSB2, PDR1 and PMA1 were previously constructed with the same strategy by (Martí-Raga et al., 2017).
Publicly available sequences of yeasts from wine and flor genetic groups were retrieved from (Legras et al., 2018; Peter et al., 2018) and are listed in Supplementary Table 6. A matrix of 385,678 SNPs was generated with GenotypeGVCFs from GATK after gvcf files were constructed as detailed in (Peter et al., 2018). This matrix was used to build a neighbor-joining tree using the ape and SNPrelate R packages. Flor and wine yeast genetic groups were determined according to (Legras et al., 2018; Peter et al., 2018) and correspond to the flor genetic group and the Wine/European (subclade 4), respectively. Flor yeast specific alleles were defined as alleles with a frequency difference of 90% between flor and wine genetic groups.
Proteomic Data Reanalysis
The dataset used for reanalyzing proteome specificities of the strain SB correspond to the Supplementary Table 5 published by Blein-Nicolas et al. (2015). This dataset compassed the proteomes of 66 Saccharomyces strains quantified during the alcoholic fermentation of a Sauvignon blanc grape juice at two temperatures. Among those strains, 28 S. cerevisiae strains constituting a half-diallel design of 7 parental strains of different origins and 21 F1-hybrids. In that study the parental strains SB and GN were referenced as E2 and E3, respectively. A subset portion of this large data set was reanalyzed for narrowing down the proteomic specificities of the strain SB. Only the proteome corresponding to S. cerevisiae strains measure at 26°C were kept. Indeed, proteomic data for the strain E2 (SB) at 18°C were not available. In addition, we removed the proteomes of the strains W1, EW21 and EW31 due to the lower number of proteins quantified (<900) respect to the other strains. By applying these filters, we analyzed the abundance of 1,100 proteins commonly quantified in 25 S. cerevisiae strains including GN and SB. In addition, the list of the 1,264 proteins specifically detected between SB and GN was listed in the Supplementary Table 7. The abundance values indicated in Supplementary Table S7 are the average of three biological replicates where 90% of the data points have a CV% lower than 5.37. The Abundance Fold Change Ratio (AFCR) of the strains SB and GN were expressed in log2 for an easier comparison. An arbitrary AFCR threshold of ±1 was used for selected proteins having a relevant abundance change, this basic threshold is widely used in the proteomics literature. The Supplementary Table 8 provides the list of the 207 proteins selected in the set of the 1,264 proteins common to SB and GN. Proteins with a differential abundance between SB and GN were used for computing a STRING analysis in order to find out functional connections with the eight genes validated in this study. The permanent link of such analysis is the following https://version-11-0.string-db.org/cgi/network.pl?networkId=pEeVlh8dPgJJ. The interaction classes interrogated were “experiments” and “databases” with the highest confidence score.
All the statistical and graphical analyses were carried out using R software (R Core Team, 2018).
The lato sensu heritability H2 was estimated for each phenotype as follows:
where σP2 is the variance of progeny population in each environmental conditions, explaining both the genetic and environmental variance of the phenotype measured, whereas σE2 is the median of the variance of replicates in each environmental conditions, explaining only the environmental fraction of phenotypic variance.
Data Availability Statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.
EP and PM contributed to the study conception and design. EP and CV carried out laboratory experiments. EP, CV, OA, AF, JS, and PM performed data analysis and interpretation. EP wrote the first draft of the manuscript. AF, PM, and JS contributed and revised the manuscript. All authors have read and approved the final manuscript.
This work was funded by Région d'Aquitaine (https://www.nouvelle-aquitaine.fr). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Conflict of Interest
EP was employed by Biolaffort. PM and CV are employed by Biolaffort.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The handling editor CTH declared a past co-authorship with one of the authors JS.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
The authors thank Justine Pape, Dylan Dos Reis, and Elodie Kaminski that helped managing fermentations.
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/ffunb.2021.733513/full#supplementary-material
Supplementary Figure 1. Correlation between traits. Data is normalized according to environment. Each dot represents the average value of an individual in one of the three phenotypic condition. Correlation coefficient and P value of Spearman's correlation test is indicated. CO2max is negatively correlated with glycerol and positively correlated with MAC% (Spearman test, p < 0.01). However, rho values observed are quite low (<0.2) because the variation in CO2 production is balanced by glycerol production and malic acid consumption.
Supplementary Figure 2. Distribution of traits. (Left) Distribution of the progeny according to trait and media is represented. Dashed vertical line represent parental average value. (Right) Data is normalized according to environment. Distribution of the progeny in all media, according to trait and cross. Dashed vertical line represent parental average value.
Supplementary Figure 3. QTL effect in population. Effect of each QTL according to parental inheritance. Each dot represents the phenotypic value of one individual and are colored according to their marker inheritance. Bigger points represent the mean of the population.
Supplementary Figure 4. Discrepancy for MSB2. (A) Effect of the marker associated to MSB2 in the offspring. Each dot represents the phenotypic value of one individual and are colored according to their marker inheritance. (B) Result of RHA test for MSB2. The represented value is from at least 5 biological replicates. The level of significance is indicated as follows: *p ≤ 0.1. **p ≤ 0.05. ***p ≤ 0.01. Solid lines of kinetic curves represent the mean and the shadow the standard error.
Supplementary Figure 5. Impact of targeted genes on CO2max and ethanol. 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.
Supplementary Figure 6. SDH2 hemizygotes show a substantial haploinsufficiency according to media. The represented value is from at least 5 biological replicates. A Wilcoxon–Mann–Whitney test was applied to assess the significance of the phenotypic difference between wild type and hemizygote. The level of significance is indicated as follows: *p ≤ 0.1, **p ≤ 0.05, and ***p ≤ 0.01. Solid lines of kinetic curves represent the mean and the shadow the standard error.
Supplementary Figure 7. SB proteome exhibit a strongest variability than GN respect to 24 others S. cerevisiae proteomes. The plot represents the distribution of the Abundance Fold Change Ratio (expressed in log2) of the strains SB and GN respect to the average values of 24 other strains. The variance of SB and GN computed for the 1,110 proteins indicated a highest variability of the SB proteome (F-test analysis <1·10−7).
Supplementary Figure 8. Abundance of proteins belonging to the oxidative and reductive branches of TCA in SB respect to GN and others S. cerevisiae strains. (A) Abundance fold ratio of quantified proteins belonging to the TCA and the glyoxylate shunt; red and green colors indicated over and under expressed proteins in the SB strain vs. GN (left box) or vs. the average value of 24 S. cerevisiae strains (right box). (B) Correlation between the AFCR (log2) of SB vs. GN and SB vs. 24 S. cerevisiae strains for the commonly expressed proteins.
Supplementary File 1. Assessment of the alcoholic fermentation yield and variability of carbon use in wine fermentation.
Supplementary File 2. Hemizygotes construction.
Supplementary Table 1. Genotype data of offspring.
Supplementary Table 2. Phenotype data of offspring.
Supplementary Table 3. Heritability.
Supplementary Table 4. QTL list.
Supplementary Table 5. Candidate genes.
Supplementary Table 6. Strains used for phylogeny analysis.
Supplementary Table 7. Protein dataset.
Supplementary Table 8. Protein difference SB vs. GN.
Albertin, W., Marullo, P., Bely, M., Aigle, M., Bourgais, A., Langella, O., et al. (2013). Linking post-translational modifications and variation of phenotypic traits. Mol. Cell. Proteom. MCP 12, 720–735. doi: 10.1074/mcp.M112.024349
Bakker, B. M., Overkamp, K. M., Van Maris, A. J, a, Kötter, P., Luttik, M., a, H., et al. (2001). Stoichiometry and compartmentation of NADH metabolism in Saccharomyces cerevisiae. FEMS Microbiol. Rev. 25, 15–37. doi: 10.1111/j.1574-6976.2001.tb00570.x
Bauer, J., Luttik, M. A. H., Flores, C.-L., Dijken, J. P., Pronk, J. T., and Niederberger, P. (1999). By-product formation during exposure of respiring Saccharomyces cerevisiae cultures to excess glucose is not caused by a limited capacity of pyruvate carboxylase. FEMS Microbiol. Lett. 179, 107–113. doi: 10.1111/j.1574-6968.1999.tb08715.x
Becker-Kettern, J., Paczia, N., Conrotte, J. F., Kay, D. P., Guignard, C., Jung, P. P., et al. (2016). Saccharomyces cerevisiae forms D-2-hydroxyglutarate and couples its degradation to D-lactate formation via a cytosolic transhydrogenase. J. Biol. Chem. 291, 6036–6058. doi: 10.1074/jbc.M115.704494
Bhatia, A., Yadav, A., Zhu, C., Gagneur, J., Radhakrishnan, A., Steinmetz, L. M., et al. (2014). Yeast growth plasticity is regulated by environment-specific Multi-QTL interactions. G3 Genes Genomes Genet. 4, 769–777. doi: 10.1534/g3.113.009142
Blein-Nicolas, M., Albertin, W., da Silva, T., Valot, B., Balliau, T., Masneuf-Pomarède, I., et al. (2015). A systems approach to elucidate heterosis of protein abundances in yeast. Mol. Cell. Proteom. MCP 14, 2056–2071. doi: 10.1074/mcp.M115.048058
Blein-Nicolas, M., Albertin, W., Valot, B., Marullo, P., Sicard, D., Giraud, C., et al. (2013). Yeast proteome variations reveal different adaptive responses to grape must fermentation. Mol. Biol. Evol. 30, 1368–1383. doi: 10.1093/molbev/mst050
Bloom, J. S., Boocock, J., Treusch, S., Sadhu, M. J., Day, L., Oates-Barker, H., et al. (2019). Rare variants contribute disproportionately to quantitative trait variation in yeast. ELife 8:49212. doi: 10.7554/eLife.49212
Boles, E., de Jong-Gubbels, P., and Pronk, J. T. (1998). Identification and characterization of MAE1, the Saccharomyces cerevisiae structural gene encoding mitochondrial malic enzyme. J. Bacteriol. 180, 2875–2882. doi: 10.1128/JB.180.11.2875-2882.1998
Camarasa, C., Grivet, J. P., and Dequin, S. (2003). Investigation by 13C-NMR and tricarboxylic acid (TCA) deletion mutant analysis of pathways of succinate formation in Saccharomyces cerevisiae during anaerobic fermentation. Microbiology 149, 2669–2678. doi: 10.1099/mic.0.26007-0
Coi, A. L., Bigey, F., Mallet, S., Marsit, S., Zara, G., Gladieux, P., et al. (2017). Genomic signatures of adaptation to wine biological ageing conditions in biofilm-forming flor yeasts. Mol. Ecol. 26, 2150–2166. doi: 10.1111/mec.14053
da Silva, T., Albertin, W., Dillmann, C., Bely, M., la Guerche, S., Giraud, C., et al. (2015). Hybridization within saccharomyces genus results in homoeostasis and phenotypic novelty in winemaking conditions. PLoS ONE 10:e0123834. doi: 10.1371/journal.pone.0123834
Delcourt, F., Taillandier, P., Vidal, F., and Strehaiano, P. (1995). Influence of pH, malic acid and glucose concentrations on malic acid consumption by Saccharomyces cerevisiae. Appl. Microbiol. Biotechnol. 43, 321–324. doi: 10.1007/BF00172832
Eder, M., Sanchez, I., Brice, C., Camarasa, C., Legras, J. L., and Dequin, S. (2018). QTL mapping of volatile compound production in Saccharomyces cerevisiae during alcoholic fermentation. BMC Genom. 19:166. doi: 10.1186/s12864-018-4562-8
Effelsberg, D., Cruz-Zaragoza, L. D., Tonillo, J., Schliebs, W., and Erdmann, R. (2015). Role of pex 21p for piggyback import of Gpd1p and Pnc1p into peroxisomes of Saccharomyces cerevisiae. J. Biol. Chem. 290, 25333–25342. doi: 10.1074/jbc.M115.653451
Ehsani, M., Fernández, M. R., Biosca, J. a, and Dequin, S. (2009). Reversal of coenzyme specificity of 2,3-butanediol dehydrogenase from Saccharomyces cerevisae and in vivo functional analysis. Biotechnol. Bioeng. 104, 381–389. doi: 10.1002/bit.22391
Fournier, T., Saada, O. A., Hou, J., Peter, J., Caudal, E., and Schacherer, J. (2019). Extensive impact of low-frequency variants on the phenotypic landscape at population-scale. ELife 8:49258. doi: 10.7554/eLife.49258
Gladieux, P., Ropars, J., Badouin, H., Branca, A., Aguileta, G., De Vienne, D. M., et al. (2014). Fungal evolutionary genomics provides insight into the mechanisms of adaptive divergence in eukaryotes. Mol. Ecol. 23, 753–773. doi: 10.1111/mec.12631
Huet, C., Menendez, J., Gancedo, C., and François, J. M. (2000). Regulation of pyc1 encoding pyruvate carboxylase isozyme I by nitrogen sources in Saccharomyces cerevisiae. Eur. J. Biochem. 267, 6817–6823. doi: 10.1046/j.1432-1033.2000.01779.x
Keller, M. (2010). Managing grapevines to optimise fruit development in a challenging environment: a climate change primer for viticulturists. Aust. J. Grape Wine Res. 16, 56–69. doi: 10.1111/j.1755-0238.2009.00077.x
Kontoudakis, N., Esteruelas, M., Fort, F., Canals, J. M., De Freitas, V., and Zamora, F. (2011). Influence of the heterogeneity of grape phenolic maturity on wine composition and quality. Food Chem. 124, 767–774. doi: 10.1016/j.foodchem.2010.06.093
Kubo, Y., Takagi, H., and Nakamori, S. (2000). Effect of gene disruption of succinate dehydrogenase on succinate production in a sake yeast strain. J. Biosci. Bioeng. 90, 619–624. doi: 10.1016/S1389-1723(00)90006-9
Kutyna, D. R., Varela, C., Henschke, P., Chambers, P. J., and Stanley, G. (2010). Microbiological approaches to lowering ethanol concentration in wine. Trends Food Sci. Technol. 21, 293–302. doi: 10.1016/j.tifs.2010.03.004
Kutyna, D. R., Varela, C., Stanley, G., Borneman, R, Henschke, P., and Chambers, P. J. (2012). Adaptive evolution of Saccharomyces cerevisiae to generate strains with enhanced glycerol production. Appl. Microbiol. Biotechnol. 93, 1175–1184. doi: 10.1007/s00253-011-3622-7
Legras, J. L., Galeote, V., Bigey, F., Camarasa, C., Marsit, S., Nidelet, T., et al. (2018). Adaptation of S. cerevisiae to fermented food environments reveals remarkable genome plasticity and the footprints of domestication. Mol. Biol. Evol. 35, 1712–1727. doi: 10.1093/molbev/msy066
Marsit, S., Mena, A., Bigey, F., Sauvage, F.-X., Couloux, A., Guy, J., et al. (2015). Evolutionary advantage conferred by an eukaryote-to-eukaryote gene transfer event in wine yeasts. Mol. Biol. Evol. 2015:msv057. doi: 10.1093/molbev/msv057
Martí-Raga, M., Peltier, E., Mas, A., Beltran, G., and Marullo, P. (2017). Genetic causes of phenotypic adaptation to the second fermentation of sparkling wines in Saccharomyces cerevisiae. G3 Genes Genomes Genetics 7, 399–412. doi: 10.1534/g3.116.037283
Minikel, E. V., Karczewski, K. J., Martin, H. C., Cummings, B. B., Whiffin, N., Rhodes, D., et al. (2020). Evaluating drug targets through human loss-of-function genetic variation. Nature 581, 459–464. doi: 10.1038/s41586-020-2267-z
Moreno-García, J., García-Martínez, T., Millán, M. C., Mauricio, J. C., and Moreno, J. (2015a). Proteins involved in wine aroma compounds metabolism by a Saccharomyces cerevisiae flor-velum yeast strain grown in two conditions. Food Microbiol. 51, 1–9. doi: 10.1016/j.fm.2015.04.005
Moreno-García, J., García-Martínez, T., Moreno, J., and Mauricio, J. C. (2015b). Proteins involved in flor yeast carbon metabolism under biofilm formation conditions. Food Microbiol. 46, 25–33. doi: 10.1016/j.fm.2014.07.001
Novo, M., Bigey, F., Beyne, E., Galeote, V., Gavory, F., Mallet, S., et al. (2009). Eukaryote-to-eukaryote gene transfer events revealed by the genome sequence of the wine yeast Saccharomyces cerevisiae EC1118. Proc. Natl. Acad. Sci. USA. 106, 16333–16338. doi: 10.1073/pnas.0904673106
Peltier, E., Bernard, M., Trujillo, M., Prodhomme, D., Barbe, J. C., Gibon, Y., and Marullo, P. (2018a). Wine yeast phenomics: a standardized fermentation method for assessing quantitative traits of Saccharomyces cerevisiae strains in enological conditions. PLoS ONE 13:e0190094. doi: 10.1371/journal.pone.0190094
Peltier, E., Sharma, V., Raga, M. M., Roncoroni, M., Bernard, M., Yves, G., et al. (2018b). Dissection of the molecular bases of genotype × environment interactions: a study of phenotypic plasticity of Saccharomyces cerevisiae in grape juices. BMC Genomics 19:772. doi: 10.1186/s12864-018-5145-4
Peltier, E., Friedrich, A., Schacherer, J., and Marullo, P. (2019). Quantitative trait nucleotides impacting the technological performances of industrial Saccharomyces cerevisiae strains. Front. Genet. 10:683. doi: 10.3389/fgene.2019.00683
Pérez-Ortín, J. E., Querol, A., Puig, S., and Barrio, E. (2002). Molecular characterization of a chromosomal rearrangement involved in the adaptie evolution of yeast strains. Genome Res. 12, 1533–1539. doi: 10.1101/gr.436602
Peter, J., De Chiara, M., Friedrich, A., Yue, J. X., Pflieger, D., Bergström, A., et al. (2018). Genome evolution across 1,011 Saccharomyces cerevisiae isolates. Nature 556, 339–344. doi: 10.1038/s41586-018-0030-5
Pitoniak, A., Birkaya, B., Dionne, H. M., Vadaie, N., and Cullen, P. J. (2009). The signaling mucins Msb2 and Hkr1 differentially regulate the filamentation mitogen-activated protein kinase pathway and contribute to a multimodal response. Mol. Biol. Cell 20, 3101–3114. doi: 10.1091/mbc.e08-07-0760
Pronk, J. T., Steensma, H. Y., and Van Dijken, J. P. (1996). Pyruvate metabolism in Saccharomyces cerevisiae. Yeast 12, 1607–1633. doi: 10.1002/(SICI)1097-0061(199612)12:16<1607::AID-YEA70>3.0.C.O.;2-4
R Core Team (2018). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available online at: https://www.r-project.org/ (accessed September 30, 2021).
Ross-Ibarra, J., Morrell, P. L., and Gaut, B. S. (2007). Plant domestication, a unique opportunity to identify the genetic basis of adaptation. Proc. Natl. Acad. Sci. USA. 104, 8641–8648. doi: 10.1073/pnas.0700643104
Rossouw, D., Heyns, E. H., Setati, M. E., Bosch, S., and Bauer, F. F. (2013). Adjustment of trehalose metabolism in wine Saccharomyces cerevisiae strains to modify ethanol yields. Appl. Environ. Microbiol. 79, 5197−5207. doi: 10.1128/AEM.00964-13
Sharmaa, A., Lee, J. S., Dang, C. G., Sudrajad, P., Kim, H. C., Yeon, S. H., et al. (2015). Stories and challenges of genome wide association studies in livestock - a review. Asian Austral J. Anim. Sci. 28, 1371–1379. doi: 10.5713/ajas.14.0715
Steinmetz, L. M., Sinha, H., Richards, D. R., Spiegelman, J. I., Oefner, P. J., McCusker, J. H., et al. (2002). Dissecting the architecture of a quantitative trait locus in yeast. Nature 416, 326–330. doi: 10.1038/416326a
Stucka, R., Dequin, S., Salmon, J. M., and Gancedo, C. (1991). DNA sequences in chromosomes 11 and VII code for pyruvate carboxylase isoenzymes in Saccharomyces cerevisiae: analysis of pyruvate carboxylase-deficient strains. MGG Mol. General Genet. 229, 307–315. doi: 10.1007/BF00272171
Szklarczyk, D., Gable, A. L., Lyon, D., Junge, A., Wyder, S., Huerta-Cepas, J., et al. (2019). STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 47, D607–D613. doi: 10.1093/nar/gky1131
Tatebayashi, K., Tanaka, K., Yang, H.-Y., Yamamoto, K., Matsushita, Y., Tomida, T., et al. (2007). Transmembrane mucins Hkr1 and Msb2 are putative osmosensors in the SHO1 branch of yeast HOG pathway. EMBO J. 26, 3521–3533. doi: 10.1038/sj.emboj.7601796
Tilloy, V., Cadière, A., Ehsani, M., and Dequin, S. (2015). Reducing alcohol levels in wines through rational and evolutionary engineering of Saccharomyces cerevisiae. Int. J. Food Microbiol. 213, 49–58. doi: 10.1016/j.ijfoodmicro.2015.06.027
Tilloy, V., Ortiz-Julien, A., and Dequin, S. (2014). Reduction of ethanol yield and improvement of glycerol formation by adaptive evolution of the wine yeast Saccharomyces cerevisiae under hyperosmotic conditions. Appl. Environ. Microbiol. 80, 2623–2632. doi: 10.1128/AEM.03710-13
Vion, C., Peltier, E., Bernard, M., Muro, M., and Marullo, P. (2021). Marker assisted selection of malic-consuming Saccharomyces cerevisiae strains for winemaking. Efficiency and limits of a QTL's driven breeding program. J. Fungi 7:304. doi: 10.3390/jof7040304
Walker, M. E., Val, D. L., Rohde, M., Devenish, R. J., and Wallace, J. C. (1991). Yeast pyruvate carboxylase: identification of two genes encoding isoenzymes. Biochem. Biophys. Res. Commun. 176, 1210–1217. doi: 10.1016/0006-291X(91)90414-3
Will, J. L., Kim, H. S., Clarke, J., Painter, J. C., Fay, J. C., and Gasch, A. P. (2010). Incipient balancing selection through adaptive loss of aquaporins in natural Saccharomyces cerevisiae populations. PLoS Genet. 6:e1000893. doi: 10.1371/journal.pgen.1000893
Zimmer, A., Durand, C., Loira, N., Durrens, P., Sherman, D. J., and Marullo, P. (2014). QTL dissection of lag phase in wine fermentation reveals a new translocation responsible for Saccharomyces cerevisiae adaptation to sulfite. PLoS ONE 9:e86298. doi: 10.1371/journal.pone.0086298
Keywords: yeast, flor yeast, QTL, wine fermentation, quantitative genetic, linkage analysis, alcoholic fermentation, Saccharomyces cerevisiae
Citation: Peltier E, Vion C, Abou Saada O, Friedrich A, Schacherer J and Marullo P (2021) Flor Yeasts Rewire the Central Carbon Metabolism During Wine Alcoholic Fermentation. Front. Fungal Biol. 2:733513. doi: 10.3389/ffunb.2021.733513
Received: 30 June 2021; Accepted: 31 August 2021;
Published: 18 October 2021.
Edited by:Chris Todd Hittinger, University of Wisconsin-Madison, United States
Reviewed by:Amparo Querol, Consejo Superior de Investigaciones Científicas (CSIC), Spain
Jose Paulo Sampaio, New University of Lisbon, Portugal
Copyright © 2021 Peltier, Vion, Abou Saada, Friedrich, Schacherer and Marullo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Emilien Peltier, email@example.com