Assessing the Mechanisms Responsible for Differences between Nitrogen Requirements of Saccharomyces cerevisiae Wine Yeasts in Alcoholic Fermentation

Abstract

Nitrogen is an essential nutrient for Saccharomyces cerevisiae wine yeasts during alcoholic fermentation, and its abundance determines the fermentation rate and duration. The capacity to ferment under conditions of nitrogen deficiency differs between yeasts. A characterization of the nitrogen requirements of a set of 23 strains revealed large differences in their fermentative performances under nitrogen deficiency, and these differences reflect the nitrogen requirements of the strains. We selected and compared two groups of strains, one with low nitrogen requirements (LNRs) and the other with high nitrogen requirements (HNRs). A comparison of various physiological traits indicated that the differences are not related to the ability to store nitrogen or the protein content. No differences in protein synthesis activity were detected between strains with different nitrogen requirements. Transcriptomic analysis revealed expression patterns specific to each of the two groups of strains, with an overexpression of stress genes in HNR strains and a stronger expression of biosynthetic genes in LNR strains. Our data suggest that differences in glycolytic flux may originate from variations in nitrogen sensing and signaling under conditions of starvation.

INTRODUCTION

Alcoholic fermentation for wine making is conducted in an environment that is stressful (high ethanol content, nutrient starvation, low pH) for Saccharomyces cerevisiae yeasts. The availability of diverse nutrients in grape musts (vitamins, sterols, unsaturated fatty acids, pantothenic acid, or nitrogen) is often limited, restricting yeast cell growth. Nitrogen limitation is by far the most frequent growth limitation encountered during wine fermentation (1). Various nitrogen sources are available in grape musts, including ammonium ions, amino acids, and peptides. Only ammonium ions and the alpha-amino nitrogen (free amino acids except for proline) are used by yeasts in fermentation. Grape musts have diverse nitrogen compositions, and the concentration of assimilable nitrogen ranges from 60 to 500 mg liter⁻¹ (2). The nitrogen content is dependent on many factors, including rootstock, climate, vine growing conditions, grape variety, and grape processing (3). Nitrogen availability in musts is a major determinant of the maximal fermentation rate (1) and also affects the size of the total yeast cell population; however, the effect on the fermentation rate is mainly independent of the population size (2). Indeed, low assimilable nitrogen concentrations can lead to slow fermentations (4–10). Under enological conditions, musts are considered nitrogen limited when the assimilable nitrogen content is below 150 mg liter⁻¹ (2, 11).

The ability to ferment under conditions of low nitrogen availability differs between yeast strains. This variability is manifested by differences in the capacity to maintain the fermentation rate during the stationary phase. Indeed, most alcoholic fermentation occurs during the stationary phase, and the ability to ferment strongly during this phase has a large effect on the total fermentation time. These differences of performance between yeast strains have been reported to be the nitrogen requirement for alcoholic fermentation (1, 12). Various approaches have been used to assess nitrogen requirements, for example, measuring the nitrogen flux required to support a constant fermentation rate (13, 14). According to these measures, wine yeasts can be classified as high- or low-nitrogen-requiring strains.

The stationary phase, critical for fermentation, results from nitrogen depletion from yeast; yeast cells adjust their metabolism, and the fermentation rate drops more or less progressively (14). The molecular mechanisms leading to the slowdown of the fermentation flux during starvation are still unclear. Several studies have reported a link between the drop of the fermentation rate and sugar transporter activity: nitrogen depletion and the shutdown of protein synthesis result in the catabolic inactivation and degradation of sugar transporters (10). However, the details of the consequences for the fermentation rate of such regulation during starvation are not known. The onset of nitrogen starvation triggers massive physiological changes involving autophagy, nitrogen recycling systems, a decrease of protein synthesis, and the reorientation of the carbon flux to favor glycogen and trehalose storage. Substantial transcriptional remodeling associated with the entry into starvation during wine alcoholic fermentations has been described (15, 16) and includes the development of a general stress response. In addition, specific depletion of amino acids induces genes encoding enzymes of amino acid biosynthesis (17–19). Many of these transcriptional changes are controlled by the TOR pathway, which senses the nitrogen status of the cell and adapts the nitrogen metabolism to nutrient availability (20, 21). This control allows transition from the use of rich sources of nitrogen to that of poor sources of nitrogen. Entry into stationary phase is also characterized by changes in protein abundance, which generally decreases, despite some neosynthesis of stationary-phase-specific or stress-induced proteins (16). This neosynthesis implies mechanisms for recycling nitrogen compounds, such as autophagy and ribophagy. Piggott et al. (22) have shown that autophagy is critical for yeast fitness in alcoholic fermentation, illustrating the role of nitrogen reallocation under starvation conditions.

Although the yeast response to nitrogen status in alcoholic fermentation has been described, the molecular mechanisms underlying the differences in nitrogen requirements between strains are unknown. We report here an investigation of a set of wine yeast strains with different nitrogen requirements with the aim of elucidating the underlying molecular mechanisms. We combined a physiological analysis with a transcriptomic study and found that differences in nitrogen requirement may be linked to differences in the signaling in response to nitrogen starvation.

MATERIALS AND METHODS

Yeast strains and inoculation.

The eight S. cerevisiae strains used in this study were acquired from various laboratory and industrial sources and are listed in Table 1. When no published data were available, species identification was checked by internal transcribed spacer restriction with HaeIII (23). Among these strains, EC1118 is a reference wine yeast strain, since its genome has been fully sequenced and characterized (24). The strains were stored and plated on yeast extract-peptone-dextrose (YEPD) plates for single-colony isolation. To inoculate experimental cultures, a preculture was prepared by transferring a single colony from the agar plate to 50 ml YEPD medium in a 250-ml Erlenmeyer flask. This inoculum was grown overnight in an incubator at 28°C with shaking (280 rpm). One aliquot from this preculture was used to inoculate fermentors to a density of 1 × 10⁶ cells ml⁻¹.

TABLE 1.

Saccharomyces cerevisiae strains used in this study

Synthetic culture medium.

The yeast was cultured in the synthetic medium (SM) described by Bely et al. (2). It contains 200 g liter⁻¹ glucose, and its amino acid composition simulates the nitrogen content of a standard grape juice. For these fermentations, we used a medium containing only glucose to overcome the already known effects of variations in fructose uptake (25), consideration of which was out of the scope of the present study. The standard SM, called “normal medium,” contained 425 mg liter⁻¹ assimilable nitrogen (SM425), and another SM with a lower nitrogen content (100 mg liter⁻¹ assimilable nitrogen [SM100]) was also used. In both SM425 and SM100, the nitrogen source consisted of ammonium salts (30%) and alpha-amino acids (70%), which, except for proline, were considered assimilable nitrogen.

Fermentation conditions, kinetics, and cycloheximide treatment.

The fermentors (1.2 liters containing 1 liter of medium) were closed with fermentation locks (CO₂-bubbling outlets filled with water). The filling conditions were controlled, and fermentations were carried out under isothermal conditions (28°C) with permanent stirring (300 rpm). All fermentations were run in duplicate or triplicate.

The amount of CO₂ released during fermentation was calculated from automatic measurements (taken every 20 min) of the fermentor weight (26). The CO₂ production rate was calculated by polynomial smoothing of the last 10 measurements of fermentor weight loss. The numerous acquisitions of weight and the precision of weighing (0.1 to 0.01 g) allowed the CO₂ production rate to be calculated with accuracy (1).

To test the involvement of protein synthesis, cell cultures in SM100 were treated at the beginning of the stationary phase with cycloheximide (100 μg ml⁻¹; Sigma, St. Louis, MO). Cycloheximide is an inhibitor of eukaryotic protein synthesis and acts by blocking the elongation factor 2-mediated translocation of peptidyl-tRNA from the A site to the P site on the 60S ribosomal subunit. The efficiency of cycloheximide inhibition was verified by measuring the rate of l-[³⁵S]methionine incorporation in a control strain as a marker of protein synthesis.

Quantification of yeast strain nitrogen requirement.

One approach to determining the yeast nitrogen requirement is to add throughout the fermentation in nitrogen-limited medium the amount of assimilable nitrogen necessary to maintain the CO₂ production at a constant rate during the stationary phase (13, 14). Fermentations were performed in SM100. When the CO₂ production rate decreased, diammonium phosphate (DAP; 300 mg liter⁻¹, equivalent to 63 mg liter⁻¹ assimilable nitrogen atoms) was added using a Gilson peristaltic pump. The amount of nitrogen added to maintain the CO₂ production rate constant is proportional to the fermentation rate. The effects of the nitrogen additions were quantified by plotting the amount added against the amount of CO₂ produced and calculating the mean nitrogen consumption between 10 and 50 g liter⁻¹ of CO₂ produced (P_10-50).

Biomass, nitrogen, and protein determinations.

The yeast population was characterized by determining the number of cells after sonication using an electronic particle counter (Multisizer 3 Coulter Counter; Beckman Coulter) and by measuring the dry weight.

The total nitrogen concentration was determined by distillation and back-titration according to the Kjeldahl mineralization method (27). The residual ammonium concentration in the medium was measured by spectrophotometry (Enzytec fluid ammonia; R-Biopharm, Darmstadt, Germany) with glutamate dehydrogenase (GLDH) according to the manufacturer's instructions. Residual free amino acids in the medium were assayed by cation-exchange chromatography followed by postcolumn derivatization with ninhydrin (Biochrom 30; Biochrom, Cambridge, United Kingdom). The total available nitrogen (TAN) and the residual nitrogen (RN) present in the medium at the end of the fermentation were determined as the sums of ammoniacal nitrogen and nitrogen from amino acids determined before and after fermentation, respectively. Total yeast-assimilable nitrogen (TYAN) was calculated as the difference between TYAN and RN.

Total cell protein was determined with a bicinchoninic acid (BCA) protein assay (BCA method), which is a modified biuret method. In brief, proteins were extracted from cells by incubation with 50% (vol/vol) dimethyl sulfoxide for 1 h at 100°C and quantified with a BCA assay kit (Sigma-Aldrich, France). Bovine serum albumin (BSA) was used as the standard for calibration.

Protein synthesis determination.

Protein synthesis rates were estimated by measurement of the incorporation of l-[³⁵S]methionine into the acid-precipitable protein fraction; 100 μl of culture medium (containing 5 × 10⁶ cells) was added to 10 μl of l-[³⁵S]methionine (1 μCi). Incorporation was stopped either immediately (controls) or after incubation for 30 min at 25°C by adding 20 μl of the samples (cells plus l-[³⁵S]methionine) to 1 ml of cold 10% trichloroacetic acid and 100 μl of BSA (1 mg BSA ml⁻¹) as a carrier protein to precipitate the proteins. The samples were incubated for 30 min on ice. The precipitable material was collected by filtration through a 0.2-μm-pore-size Nuclepore filter and washed with 30 ml of cold 10% trichloroacetic acid. The radioactivity was measured in a liquid scintillation counter.

RNA extraction, labeling, and microarray hybridization.

The eight strains were subjected to whole-genome expression analysis. EC1118 was considered the reference strain. Two biological replicates were performed for each strain. Total RNA was isolated from cultures at 45 g liter⁻¹ of CO₂ production (50% fermentation progress) by the TRIzol method as described by Chomczynski and Sacchi (28). For each RNA extraction, 1 × 10⁹ cells were pelleted by centrifugation (3,000 rpm, 2 min) and lysed mechanically by vortexing with glass beads (diameter = 0.3 mm) in 400 μl TRIzol (Gibco BRL) at 4°C for 8 min. The liquid phase was collected, and TRIzol was added to give a final volume of 4 ml; the samples were incubated for 5 min at room temperature, and 800 μl chloroform was added. The mixture was vortexed and incubated for 2 min at temperature room and then centrifuged (10,000 rpm, 15 min). An equal volume of cold (−20°C) isopropanol was added to the aqueous phase to precipitate the RNA. The samples were incubated for 10 min at room temperature and centrifuged (10,000 rpm, 10 min). The resulting nucleic acid pellet was washed twice with 2 ml 75% ethanol–diethyl pyrocarbonate-treated water and then dissolved in 150 μl of nuclease-free water (Qiagen). The RNA in 100-μg aliquots of these preparations was purified with an RNeasy minikit (Qiagen) following the RNA cleanup protocol, including membrane DNase digestion. Fluorescent cDNAs were synthesized from 100 ng of total RNA using a one-color RNA spike-in kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's recommendation (indirect labeling of mRNAs with cyanine 3 dCTP dye). Labeled cDNA was purified with the RNeasy minikit (Qiagen).

Agilent 8x15k gene expression microarrays with specific designs were obtained from Agilent Technologies (Santa Clara, CA) and used for microarray hybridization. The array design was based on Agilent ID 038619 with the addition of 39 genes from the new regions of EC1118 (24) (available in the Gene Expression Omnibus [GEO] database under accession number GPL17690). Six hundred nanograms of labeled cRNA was hybridized on the slide for 17 h in a rotating oven (10 rpm) at 65°C using an Agilent Technologies kit, according to the manufacturer's instructions, and then washed. A Genepix 4000B scanner (Axon Instrument Inc.) was used for array digitalization, and data were analyzed with Genepix Pro7 software (Axon Instruments).

Statistical analyses.

R.2.14.2 software (29) was used for statistical analyses. To obtain a general overview of the profile of the high- and low-nitrogen-requiring strains, principal component analysis (PCA) was performed using the FactoMineR package (30) with the following parameters: maximal fermentation rate (R_max), fermentation rate at 50% sugar consumption (R₅₀), total fermentation duration (F_d), and the quantified nitrogen requirement (QNr). The limma package (31–33) was used to normalize and analyze the microarray data (quantile method for normalization between arrays). To identify differential gene expression between strains with high and low nitrogen requirements (HNRs and LNRs, respectively), a modified t test was applied by filtering on confidence at a P value of <0.05, using the Benjamini and Hochberg false discovery rate as a multiple-testing correction of the t test P values (34). Only genes with a log fold difference of greater than 1 (positive or negative) were taken into consideration. The genes for which expression differed were grouped according to Gene Ontology (GO) process terms using the Funspec program with the Bonferroni correction at a P value cutoff of 0.05 (33). Hierarchical clustering was used to group the selected genes by similarity of expression profiles using the Cluster program, version 3.0 (applying centered correlation and complete-linkage clustering), and the results were loaded into the Java TreeView (version 1.1.5r2) program for data display (35, 36).

Microarray data accession numbers.

The complete data set is available through the Gene Expression Omnibus (GEO) database (accession number GSE50626). The microarray description is under GEO accession number GPL17690.

RESULTS

Yeast nitrogen requirements during stationary phase.

Twenty-three Saccharomyces cerevisiae strains from various origins were screened to determine their abilities to ferment in a must with a low nitrogen content (see Fig. S1 in the supplemental material). Their nitrogen requirements were quantified (by calculating the average slope between 10 and 50 g liter⁻¹ of CO₂ released after nitrogen addition) according to the constant-rate fermentation (CRF) procedure described by Manginot et al. (14) (see Materials and Methods). The distribution of the nitrogen requirement phenotype showed substantial diversity between the strains; the distribution of the strains according to nitrogen requirement was a continuum, indicating that it is a quantitative trait involving several genes. These results are consistent with the findings of Manginot et al. (14). Eight of the 23 strains were selected for their extreme nitrogen requirements (low and high): 5 strains (4CAR1, L2868, EC1118, Fermiflor, Zymasil) exhibited a low nitrogen requirement (P_10-50, close to 1 mg N g⁻¹ CO₂; Fig. 1), and 3 strains (K1M, 7013, MTF1782) showed a high nitrogen requirement (P_10-50, between 2.2 mg and 2.5 mg N g⁻¹ CO₂; Fig. 1). These selected strains were thus representatives of strains with low nitrogen requirements (LNRs) or high nitrogen requirements (HNRs). The fermentation kinetics for the eight strains on a nitrogen-limited medium (SM100) were studied (Fig. 2). The results showed that all the strains exhibited low fermentation rates and long durations of fermentation. The R_max did not differ greatly between the strains and did not correlate with the nitrogen requirement; similarly, the cell population size was not correlated with the nitrogen requirement of the strains (see Fig. S2 in the supplemental material). However, during the stationary phase, after the growth had stopped, the fermentation profiles of the strains diverged substantially, such that by 70 h of fermentation the two groups of strains were clearly segregated: the strains with low nitrogen requirements exhibited a higher fermentation rate than strains with high nitrogen requirements. This is consistent with the classification of the strains on the basis of the measurement of nitrogen requirements by ammonium addition. We verified that the HNR strains did not exhibit any intrinsic fermentation capacity defect by examining their behavior in nitrogen-rich must (SM425): under these conditions, there were only small differences in the fermentation profiles between the strains (see Fig. S3 in the supplemental material).

FIG 1.

Comparison of nitrogen requirements of eight wine yeast strains. The measure is expressed in mg nitrogen g⁻¹ CO₂ on the basis of the amount of nitrogen necessary to support a constant fermentation rate between 10 and 50 g liter⁻¹ CO₂ (P_10-50). The reproducibility of this technique was tested for two strains (L2868 and MTF1782), and variations are indicated on the graph with error bars.

FIG 2.

Fermentation profiles of eight different wine yeast strains with contrasted nitrogen requirements in a nitrogen-limited medium (SM100).

Analysis of relationships between traits.

We investigated the relationships between the phenotypic traits of the eight yeasts by principal component analysis (PCA) (Fig. 3A and B). We considered four measures to build the variable factor map: the maximum fermentation rate (R_max), the fermentation rate at 50% sugar consumption (R₅₀), the total fermentation duration (T_f), and the quantified nitrogen requirement (QNr). The projection on the two principal axes preserves 95% of the information and explains, respectively, 88% and 7% of this variation. The PCA of kinetic measures allowed the discrimination of strains according to their nitrogen requirement. Also, there was a strong correlation between the strain nitrogen requirement and total fermentation duration, in agreement with the observations of Bely et al. (2). PCA also indicated that the nitrogen requirement and total fermentation duration were not correlated with R₅₀ and R_max. These results are consistent with the notion that the ability of cells to maintain a high fermentation rate in stationary phase is linked to their nitrogen requirement.

FIG 3.

PCA of kinetic traits (T_f, QNr, R_max, R₅₀) (A) and individual strains (B). The values were obtained from the profiles of fermentation performed on SM100 at 24°C. Dim, dimension.

Comparisons of physiological traits.

We examined whether differences in nitrogen requirements were associated with variations in the cellular content of yeast cells. We measured the total nitrogen, protein content, and dry mass of yeast cells at various stages of the fermentation process (see Fig. S4 in the supplemental material). For all the measures, differences between the LNR and HNR groups were small, and they were not statistically significantly different. There was no clear segregation of the nitrogen content between LNR and HNR strains (see Fig. S4A in the supplemental material). There were also no significant differences between strains in the protein content at midfermentation (see Fig. S4B in the supplemental material) or at 80% fermentation progression (see Fig. S4C in the supplemental material). At the end of fermentation, there were small differences, and although several LNR strains displayed a slightly higher protein content, this was not the case for the Fermiflor strain. A higher protein content was thus not a general feature of LNR strains. For all strains, dry mass values increased between midfermentation and the end of fermentation, in agreement with previous reports (see Fig. S4D and E in the supplemental material) (37). However, HNR and LNR strains showed slightly different behaviors, with a larger increase in dry mass for HNR strains than LNR strains during the stationary phase. The increase in dry mass was consistent with the storage of carbohydrates (glycogen, trehalose) at this stage, as previously demonstrated by Manginot et al. (14). We also checked the utilization of the nitrogen sources by the strains. All the nitrogen sources (except proline) were metabolized by all the strains during the first 20 h of fermentation (data not shown). This excludes the possibility that the differences in nitrogen requirements were consequences of differences in the uptake of nitrogen sources.

Assessment of the role of protein synthesis.

The ability of yeast strains to maintain a high fermentation rate during the stationary phase can be dependent on their capacity to synthesize proteins. We therefore examined the effects of inhibiting protein synthesis on the fermentation kinetics. We added the protein synthesis inhibitor cycloheximide to cultures of the strains on SM100 at the beginning of stationary phase (at 47 h of fermentation); as cell growth had already stopped by this time, any effects of the drug could not be due to inhibition of growth. Cycloheximide modified the fermentation profiles, confirming the involvement of protein synthesis. Cycloheximide triggered a large and rapid decrease of the fermentation rate of LNR strain L2868 immediately after its addition (Fig. 4). Its effect on the fermentation rate of HNR strain MTF1782 was smaller, and the kinetic profile remained similar to that of the control (Fig. 4). Furthermore, the kinetic profiles of strains L2868 and MTF1782 become very similar after cycloheximide addition. It was thus possible that the differences in behavior were due to more protein synthesis in LNR strains than in HNR strains. We therefore compared the protein synthesis activities by measuring the rate of l-[³⁵S]methionine incorporation into L2868 and MTF1782 (see Fig. S5 in the supplemental material). Protein synthesis was very similar in the two strains (difference, 5%). The differences in kinetic profiles are therefore not caused by differences in overall protein synthesis activity. We also confirmed that cycloheximide inhibited protein synthesis under these conditions. These results suggested the involvement of other mechanisms in the differences of the glycolytic flux between strains during the stationary phase of alcoholic fermentation.

FIG 4.

Effects of addition of cycloheximide at 10 g liter⁻¹ h⁻¹ of CO₂ produced on the fermentation in SM100. A strain with a low nitrogen requirement (L2868) was compared with a strain with a high nitrogen requirement (MTF1782).

Transcriptomic analysis.

Gene expression is a major determinant of phenotypic diversity (17) and may contribute to differences in strain fermentation properties. The analysis of gene expression patterns can provide insights into the molecular and physiological mechanisms associated with differences in fermentation capacities. We therefore examined the gene expression profiles of the eight selected strains. The transcriptomes were analyzed in cells fermenting in a nitrogen-limited medium (SM100) and harvested at stationary phase (when 45 g liter⁻¹ of CO₂ had been produced and the ethanol content corresponded to 6%). At this stage, the strains exhibited differences in fermentation rates. In addition, the transcriptome has been shown to be stable at this stage when growth has stopped for all cells and nutrients have been depleted (15). We compared the global expression patterns by microarray analysis, using two biological replicates for each strain. Intraclass coefficients of correlation (ICC) were 94.6% to 99.1% (data not shown), indicating a good reproducibility of the data.

Assessing variations in gene expression in the yeast population.

Four hundred fifty-four genes were differentially expressed between the two groups of strains (HNR and LNR strains), with an absolute log fold change of greater than 1; 248 genes were overexpressed in LNR strains, and 206 genes were overexpressed in HNR strains. The set of genes overexpressed in HNR strains was enriched in genes related to the stress response and encoding proteins involved in the production of reserve carbohydrates, including trehalose and glycogen (TPS1, TPS2, TPS3, UGP1, TLS1, PGM2, GLG1, GLG2) (Fig. 5A); in response to nitrogen starvation, these strains overexpressed genes involved in the metabolism of amino acids, such as arginine (ARG1), serine (CHA1), alanine (ALD2), and glutathione (GLO1), and also genes involved in glutamate biosynthesis (GAD1, GDH2) and sulfate metabolism (MET5, SUL2). Conversely, the set of genes overexpressed in LNR strains included many encoding proteins involved in protein synthesis (Fig. 5B). This suggests that in LNR strains, genes encoding rRNA processing were less susceptible to stress caused by nitrogen starvation.

FIG 5.

Analysis of transcriptomic data (mean of two repetitions): enrichment of gene functions among strains with high nitrogen requirements (A) and strains with low nitrogen requirements (B). The values correspond to the number of genes in the Funspec functional category (Bonferroni correction and a P value of 0.05). The P values are indicated for each category. LSU, large subunit.

Hierarchical clustering analysis.

A global comparison of the expression patterns of all the strains revealed 2,484 genes that were differentially expressed (adjusted P < 0.05). The aim of this analysis was to observe the tendency of gene expression in a strain group. Hierarchical clustering of the gene expression profiles identified eight clusters of expression profiles for which there was a tendency in segregation between LNR and HNR strains (see Fig. S6 in the supplemental material). In fact, the segregation was not totally perfect between the two groups of strains. For example, the Fermiflor LNR strain showed similarity to HNR strains in cluster II and a part of cluster III (Fig. 6) and a neutral position for other clusters. Interestingly, the EC1118 strain, also classified as an LNR strain, presented an HNR transcriptomic response for all the clusters. For the others strains, observations generally showed four clusters (clusters I, II, III, and IV) including groups of genes overexpressed in HNR strains (Fig. 6). Cluster I contained genes involved in proteasome degradation (PRE2, PRE3, PRE6), and cluster II was enriched in genes involved in the ubiquitin protein degradation pathway. Cluster III was composed of two small subclusters with similar profiles and included genes involved in storage of reserves, such as trehalose and glycogen (TPS1, TPS2, TSL1, PGM2, GLC8). Cluster IV contained genes associated with autophagy. Thus, protein degradation and nitrogen recycling both appear to be more active in HNR than LNR strains. Indeed, the four clusters contained genes controlled by nitrogen catabolite repression (NCR) and the TOR pathway (VID30 in cluster I, GDH3 in cluster II, PUT1 in cluster III, ATG14 in cluster IV). Four clusters (clusters V, VI, VII, and VIII) included genes overexpressed in LNR strains. They contained genes involved in ribosome biogenesis, for example, genes of the RPS (ribosomal protein, small subunit) and RPL (ribosomal protein, large subunit) families, and families of genes involved in RNA processing (UTP [U three protein], NOP [nucleolar protein], IMP [component of the U3 small nucleolar ribonucleoprotein], RRP [ribosomal RNA processing], and RPF [ribosomal protein factor] family genes). These data indicate that there is a very common mechanism: a stronger expression of genes involved in protein and ribosome synthesis, transcription, and RNA metabolism in the LNR strains than the HNR strains. However, the expression data obtained for strain EC1118 revealed a different behavior and suggested that additional mechanisms are involved. EC1118 transcriptomic data present an exception, since it has a physiological profile of an LNR strain but a transcriptomic profile similar to that of HNR strains. This suggests that the EC1118 fermentation ability under nitrogen-deficient conditions originates from a mechanism different from that for the other strains. There is also evidence that the fermentation rate does not determine the transcriptomic profile of the strain.

FIG 6.

Cluster extracts of genes overexpressed in HNR strains. Cluster I is composed of genes involved in proteome degradation. Cluster II is enriched in genes associated with ubiquitin protein degradation. Cluster III is composed of genes involved in storage of reserves. Cluster IV is related to the autophagy pathway.

The genes overexpressed in LNR strains included 25 of the 39 new genes identified in three regions (regions A, B, and C) of the EC1118 strain genome originating from horizontal transfer (28). The A region is not differentially expressed between LNR and HNR strains (see Fig. S7 in the supplemental material). The B region is present in four strains, three LNR strains (L2868, 4CAR1, EC1118) and an HNR strain (K1M). The presence or absence of the introgressions was verified by PCR on multiple genes in the region (data not shown). Expression data showed that only LNR strains expressed genes contained in this region (see Fig. S7 in the supplemental material). Concerning the C region, three strains contained the totality of the introgression (L2868, 4CAR1, EC1118), and for two strains, we controlled the absence of these regions (7013, MTF1782). Three strains contained only three genes of the C region (Fermiflor, Zymail, K1M). Expression data showed that these three genes are expressed in all the LNR strains but also in the HNR strain K1M (see Fig. S7 in the supplemental material). These are the genes FOT1 and FOT2, corresponding to two fungal oligopeptide transporters (38), and the SEO1 gene, corresponding by sequence homology to a putative permease, a member of the allantoate transporter subfamily (28). Although the expression of these three genes in all LNR strains might suggest a role in the nitrogen requirement, a similar behavior in HNR strain K1M contradicts this hypothesis.

Correlation between gene expression and fermentation rates.

The expression of a total of 282 genes was found to correlate with the fermentation rate (R₅₀): 129 positively and 153 negatively. Only the set of positively correlated genes was enriched in particular functions (Fig. 7). From these data, high fermentation rates on nitrogen-deficient medium were associated with the overexpression of genes involved in protein synthesis, transcription, and rRNA processing. These genes belong to different clusters with genes overexpressed in LNR strains. There was no detectable functional enrichment among the set of genes negatively correlated with R₅₀; nevertheless, various genes involved in the metabolism of energy reserves (GLC2, GLC3, YPI1) are members of this group.

FIG 7.

Analysis of transcriptomic data (mean of two repetitions): enrichment of gene functions among genes positively correlated with the fermentation rate (R₅₀). The values correspond to the number of genes in the Funspec functional category (Bonferroni correction and a P value of 0.05). The P values are indicated for each category.

DISCUSSION

The fermentation capacities of yeast strains in medium with a low nitrogen content differ (12, 14, 38). We screened 23 strains of different origins for their nitrogen requirements and confirmed here the variability of this character. The nitrogen requirements of the strains varied more than 2.5-fold and were associated with diversity in fermentation duration on a nitrogen-limited medium. Through the characterization of the fermentation properties of eight strains with contrasting nitrogen requirements, we confirmed that the differences are linked to how well the strains maintain a high fermentation rate after the entry into stationary phase under conditions of nitrogen starvation. We checked that the differences in fermentation performances were not triggered by differences in the utilization of the available nitrogen sources: all the strains consumed the assimilable nitrogen sources during the first hours of the fermentation, in agreement with previous observations (39). There were only small differences between HNR and LNR strains in term of cellular contents: we did not find any correlations between strain fermentation capacity and their abilities to store nitrogen, consistent with the report by Julien et al. (40). We also could not differentiate HNR and LNR strains on the basis of protein content. The total dry mass during starvation at the end of the fermentation was higher for HNR than LNR strains, consistent with the greater accumulation of storage carbohydrate. This observation is in agreement with the findings of Manginot et al. (13), who reported that the amount of glycogen was directly correlated to the nitrogen requirement of the strains and was thus higher in HNR strains. Therefore, the only difference in cellular contents that we could find to be related to the nitrogen requirements was greater storage of carbohydrates in HNR than LNR strains.

Protein synthesis can be a determinant of yeast fermentation capacity. We confirmed that protein biosynthesis was required to maintain the fermentation rate in stationary phase. However, the rates of protein synthesis in representative HNR and LNR strains were similar and could not explain the differences in the fermentation rate between these two strains. The transcriptomic study revealed expression patterns that were common to the strains of each group (HNR and LNR strains), except for strain EC1118, indicating that their member strains displayed similar physiological responses. However, there were large differences between the two groups of strains: genes involved in synthetic processes (protein synthesis, RNA processing, etc.) were overexpressed in LNR strains, whereas genes involved in protein degradation, nitrogen recycling, and the stress response were more strongly expressed in HNR strains. Except for the EC1118 strain, it seems that a common mechanism allows distinguishing LNR and HNR strains. The HNR strains exhibited a pattern characteristic of nutrient-starved and stressed cells, whereas the LNR strains exhibited a clearly weaker starvation response. The higher expression of genes involved in ribosome biogenesis in LNR strains indicates that the TOR pathway is more active than it is in HNR strains and, therefore, that nitrogen starvation is less severe. The stress response genes controlled by Msn2/Msn4, including genes for the synthesis of the storage carbohydrates glycogen and trehalose, were expressed more strongly in the HNR than the LNR strains, and there was an increase in storage carbohydrates. This is consistent with the previous observation that the dry mass for HNR strains was greater than that for LNR strains, presumably due to the accumulation of glycogen and trehalose, as demonstrated in a previous study (14). Under nitrogen-limited conditions, the HNR yeasts were more sensitive to nitrogen starvation, resulting in a stronger stress response, weaker expression of biosynthetic genes, and a lower fermentation rate. The susceptibility to nitrogen starvation could correspond to a sensitivity threshold, which would be more important in HNR strains. This sensitivity to nitrogen could also be the result of a nitrogen dose effect in the medium. Ethanol production at this stage was identical for the two groups of strains, ruling out a possible role in the differences in stress response.

These differences may be consequences of differences in the nitrogen-sensing signaling systems. The differences in gene expression and the physiological responses are reminiscent of the observations of Watanabe et al. (41), who analyzed the effects of a mutated form of the PAS kinase RIM15 in sake yeast. They showed that the altered form of RIM15 reduced the yeast stress response and the amount of storage carbohydrate and enhanced the fermentation rate. RIM15 is involved in nitrogen signaling downstream from TOR, which senses the nitrogen status of the cell (42, 43). Nitrogen starvation leads to activation of Rim15p, resulting in induction of the quiescent program and a stress response (44, 45). This suggests that differences in sensing and nitrogen signaling between HNR and LNR strains could contribute to differences in fermentation rates during starvation. Such a mechanism is consistent with the absence of differences in the amount and synthesis of protein between the two groups of strains. How diversity in nitrogen sensing/signaling modulates the glycolytic flux is unclear, although reducing the stress response in sake yeast could increase the fermentation rate (39, 45, 46). This implies a link between nitrogen signaling and the fermentation rate that could explain the differences in behavior between the two groups of strains. The data obtained from the EC1118 strain suggest that other mechanisms may also be involved in the nitrogen requirement. Moreover, this strain possesses several introgressed non-Saccharomyces cerevisiae regions. We examined whether they could play a role in fermentative capacity under nitrogen-limited conditions. We observed that only three out the five LNR strains expressed genes of these regions, indicating that these regions are not necessary for expression of the low nitrogen requirement phenotype. However, we cannot rule out the contribution of these regions in a given strain.

Thus, HNR yeasts may be more sensitive to nitrogen starvation stress and decreasing their fermentation rate. In conclusion, the nitrogen requirements of a wine strain may correspond to a common mechanism that could be its ability to sense nitrogen starvation and develop a quiescent program that reduces the flow of energy and increases its adaptation to stress. Further work is required to identify the genes and mechanisms underlying these phenomena. The nitrogen requirement of a strain is a quantitative trait, so it may be possible to identify the genes responsible for the phenotype by a genetic approach.