Genome-wide monitoring of wine yeast gene expression during alcoholic fermentation

Authors


Abstract

The transcriptome of a wine yeast was monitored throughout an alcoholic fermentation under conditions mimicking an enological environment. Major changes in gene expression occurred during fermentation, affecting more than 2000 genes, as the yeast adapted to changing nutritional, environmental and physiological conditions. The genes of many pathways are regulated in a highly coordinated manner, and genes involved in the key metabolic pathways of fermentation are strongly expressed. We showed that, during fermentation of a synthetic medium mimicking a natural must in which growth arrest was caused by nitrogen exhaustion, entry into the stationary phase triggered major transcriptional reprogramming. Many TOR target genes involved in nitrogen utilization or other functions are induced at this stage, suggesting that this signalling pathway plays a critical role in changes in gene expression in response to nitrogen depletion. Entry into stationary phase is a key physiological event and is followed by a general stress response. The superimposition of multiple stresses, including starvation and ethanol stress, gives rise to a unique stress response, involving hundreds of genes encoding proteins involved in various cellular processes, many of unknown function. Copyright © 2003 John Wiley & Sons, Ltd.

Introduction

The wine yeast strains currently used in wine fermentation were selected from natural, spontaneous fermentations and are specifically adapted to the oenological environment. These industrial yeasts are closely related to Saccharomyces cerevisiae laboratory strains but have distinct physiological properties rendering them suitable for wine fermentation (reviewed in Barre et al., 1993; Pretorius, 2000). They are specially adapted for efficient fermentation in grape musts with high sugar content (140–260 g/l), high alcohol content (up to 15% v/v), low pH (3.0–3.5), often with limiting amounts of nitrogen, lipids and vitamins, added sulphites (40–100 mg/l) and anaerobic conditions (Fleet and Heard, 1993). These yeasts are often polyploid or aneuploid and may also harbour rearranged chromosomes (Bakalinsky and Snow, 1990; Bidenne et al., 1992; Ibeas and Jimenez, 1996; Rachidi et al., 1999). These genomic reorganizations may provide mechanisms for increasing the copy number and/or expression of important genes, but the molecular basis of the properties of industrial yeasts is largely unknown. Such knowledge is crucial for improvements in the fermentation processes and defining targets for the genetic improvement or selection of wine yeasts.

Functional genomic approaches, such as microarray profiling, are powerful tools for the analysis of gene expression at the scale of the genome, providing a comprehensive view of yeast physiology (DeRisi et al., 1997; Holstege et al., 1998; Spellman et al., 1998). Global gene expression analyses are frequently carried out for laboratory yeasts, but only a few studies have dealt with industrial yeasts, including wine yeasts (Cavalieri et al., 2000; Backhus et al., 2001; Hauser et al., 2001; reviewed in Perez-Ortin et al., 2002). If we are to understand how wine yeasts cope with their stressful environment, we need to study expression under wine-making conditions, which differ considerably from laboratory conditions. Throughout the wine fermentation process, the yeast is subjected to multiple stresses, including high osmotic pressure, acidity, nutrient deprivation, starvation and high alcohol concentration. Some of these stresses occur sequentially, whereas others occur simultaneously (Bauer and Pretorius, 2000). Little is known about the mechanisms by which wine yeasts regulate gene expression to cope with changing, stressful conditions. We monitored gene expression at the level of the entire genome, during alcoholic fermentation in synthetic medium simulating a natural must. We found that strong transcriptional reprogramming took place during alcoholic fermentation, affecting more than 2000 genes, and that these changes were triggered by changes in nutritional, environmental and physiological conditions. We showed that these expression changes were highly coordinated and that master regulatory pathways played a key role in coordinating these changes in gene expression. We also assessed the impact on gene expression of the various stresses faced by the yeast during alcoholic fermentation.

Materials and methods

Strain

We used the oenological S. cerevisiae strain EC1118 from Lallemand S.A., which is widely commercialized.

Medium

Fermentation experiments were carried out with synthetic must MS300 miming a standard natural must previously described (Bely et al., 1990). This medium contains 200 g/l glucose, mineral salts (750 mg/l KH2PO4, 500 mg/l K2SO4, 250 mg/l MgSO4.7H2O, 155 mg/l CaCl2.2H2O, 200 mg/l NaCl, 4 mg/l MnSO4.H2O, 4 mg/l ZnSO4, 1 mg/l CuSO4.5H2O, 1 mg/l KI, 0.4 mg/l CoCl2.6H2O, 1 mg/l H3BO3, 1 mg/l NaMoO4.2H2O), vitamins (20 mg/l myo-inositol, 2 mg/l nicotinic acid, 1.5 mg/l calcium panthothenate, 0.25 mg/l thiamine HCl, 0.25 mg/l pyridoxine HCl, 0.003 mg/l biotin), 300 mg/l of assimilable nitrogen (ammoniacal nitrogen and α-amino nitrogen) provided by a mixture of 19 amino acids (612.6 mg/l L-proline, 505.3 mg/l L-glutamine, 374.4 mg/l L-arginine, 179.3 mg/l L-tryptophan, 145.3 mg/l L-alanine, 120.4 mg/l L-glutamic acid, 78.5 mg/l L-serine, 759.2 mg/l L-threonine, 48.4 mg/l L-leucine, 44.5 mg/l L-aspartic acid, 44.5 mg/l L-valine, 37.9 mg/l L-phenylalanine, 32.7 mg/l L-isoleucine, 32.7 mg/l L-histidine, 31.4 mg/l L-methionine, 18.3 mg/l L-tyrosine, 18.3 mg/l L-glycine, 17.0 mg/l L-lysine, and 13.1 mg/l L-cysteine) corresponding to 180 mg nitrogen and 460 mg/l ammonium chloride (corresponding to 120 mg nitrogen). The medium contained anaerobic factors (15 mg/l ergosterol; 5 mg/l sodium oleate), added to the medium in 1 ml Tween 80/ethanol (50/50, v/v). The pH was buffered at 3.3 with NaOH.

Culture conditions

Active dried yeasts were rehydrated in 50 g/l glucose at 37 °C for 30 min and used to inoculate MS300 medium at a density of 1 × 106 cells/ml.

Fermentations were performed at 24 °C, in 1.2 l fermenters equipped with locks to maintain anaerobiosis and with constant stirring. CO2 production was monitored by automatic weighing of the fermentor every 20 min, to determine weight loss. The rate of CO2 production was calculated by a method based on polynomial smoothing. The number of cells was determined with an electronic particle counter (Beckman Coulter).

mRNA extraction and reverse transcription

Total RNA was extracted using Trizol reagent (Gibco BRL, Life Technologies). Cells sample corresponding to 90 DO units were pelleted by centrifugation (5000 × g for 5 min) in two microcentrifuge tubes, resuspended in 400 µl Trizol and broken by vortexing for 4 min with 300 µl glass beads. The two extracts were pooled and the total volume was adjusted to 8 ml with Trizol reagent. After incubation for 5 min at room temperature, 1.6 ml chloroform was added to separate the aqueous and the organic phase with a brief agitation. After incubation for 3 min at room temperature the solution was centrifuged at 15 000 × g for 15 min and the aqueous phase was recovered. The RNA was precipitated by addition of an equal volume of cold (−20 °C) isopropyl alcohol and centrifugation at 10 000 × g for 10 min. mRNA was purified using the PolyATract mRNA Isolation System (Promega) according to the manufacturer's instructions, except that we doubled the SSC concentration of each buffer. All the water used was treated with DEPC.

Fluorescently labelled cDNA probes were synthesized with the CyScribe First-Strand cDNA Labelling Kit, containing Cy3-dUTP and Cy5-dUTP fluorescent labels (Amersham Pharmacia Biotech), using 1.5 µg mRNA instead of 1 µg. Probes were purified with the QIAquick Nucleotide Removal Kit (Qiagen) according to the manufacturer's instructions.

Hybridization

We used CMT yeast S288c microarray slides (Corning), which have spots for 6138 genes. Hybridization was carried out as previously described (Alexandre et al., 2001).

Slides were prehybridized by incubation with Dig Easy Buffer (Roche) for 45 min at 42 °C, washed with water and dried with argon under pressure. Probes were concentrated by evaporation under vacuum to a volume of 10 µl. The probe solution was completed by adding 30 µl Dig Easy Buffer and 2 µl 10 mg/ml boiled salmon sperm DNA. This solution was then heated for 5 min at 95 °C before hybridization with the slides. Slides were covered with a coverslip and placed in a hybridization chamber (Corning) at 37 °C overnight with the probe solution.

Washing of slides

The slides were washed for 5 min in 2× SSC, 0.1% SDS at 42 °C, then for 10 min in 0.1× SSC, 0.1% SDS at room temperature. Finally, they were washed four times in 0.1× SSC buffer for 2 min each, at room temperature. The slides were dried with argon under pressure before scanning. Scans were performed with the GenePix 4000B microarray scanner (Axon Instruments).

Microarray analysis

Microarray experiments were performed twice for each stage, once with Cy3 labelling and once with Cy5 labelling. Ratios were normalized using the GenePix normalization factor (Axon instruments), calculated by a linear-based ratio method:

equation image

where Ri is the ratio for the ith feature. We have applied a quality filtering as described by Tseng et al. (2001) to discard unreliable data. Briefly, we used for each gene the coefficient of variation of the ratio (i.e. standard deviation divided by the mean) as a quality indicator. For each gene we have constructed a windowing subset by selecting 50 genes whose ratios are closer to this gene. If the coefficient of variation of this gene is within the top 10% in its windowing subset, then we considered the data on this gene as unreliable. Most of the data discarded corresponded to relatively low intensity spots (about 60% of genes that failed to pass this quality filtering had intensity value below 500 (GenePix aribitrary units). Application of this method eliminated 6.0–7.5% of the data at each stage. The ratios used in tables and graphics correspond to averaged log2 value. Cluster analysis were performed with PermutMatrix software developed by G. Caraux, based on the Eisen et al. (1998) hierarchical clustering, using euclidean distance, Ward's method and log2 transformed ratio for processing.

Determination of amino acid profiles in fermentation media

Fermentation medium (10 ml) was mixed with 50 ml of 96% (v/v) ethanol and allowed to stand for 48 h at −20 °C. The mixture was centrifuged (20 000 × g, 20 min) and the supernatant was dried under vacuum and the residue resuspended in 0.2 N lithium citrate buffer (pH 2.2). Amino acids were separated by ion-exchange chromatography on an anionic Ultropac-8 lithium-form resin (Amersham Pharmacia Biotech) with a Chromakon 400 (Kontron) and a Biochrom 20 analyser (Amersham Pharmacia Biotech). Amino acids were detected by reaction with ninhydrin.

Determination of trehalose and glycogen concentrations

Trehalose and glycogen concentrations were determined as previously described (Roustan et al., 2002). The cells were washed three times with saline (50 g/l NaCl in water), centrifuged, and the cell pellets used for the extraction of carbohydrates. Trehalose was extracted from cells by incubation with 8 ml 0.5 M TCA for 1 h at 0 °C, followed by washing with 8 ml 0.5 M TCA and then with 4 ml water. The three supernatents were pooled and water was added to a final volume of 25 ml. We added 2.5 ml of anthrone solution (250 ml of sulphuric acid, 50 ml H2O, 570 mg anthrone) to 0.25 ml of the assay mixture and heated the resulting mixture at 100 °C for 12 min. We measured OD at 625 nm and trehalose concentration was determined by interpolation using a standard curve for trehalose.

Glycogen was extracted from the cell pellets after trehalose extraction. The cells were suspended in 1 vol 8 N HCL and 4 vols DMSO and incubated at 60 °C for 30 min. The solution was cooled and neutralized with 1 vol 8 N NaOH and the volume was adjusted with 0.11 M citrate buffer, pH 4.5, to dilute the DMSO to 1/10. The appropriate dilution of the sample, in a final volume of 1 ml citrate buffer, was incubated overnight at 37 °C with 25 µl α-amyloglucosidase (Roche). The amount of glucose released into the supernatant was determined using the Sigma Diagnostic Glucose Kit (Sigma Diagnostic) according to the manufacturer's instructions.

Results

Overview

We monitored gene expression in the industrial strain EC1118 throughout the alcoholic fermentation of a synthetic must mimicking a typical natural must, [high sugar content (200 g/l glucose), low assimilable nitrogen content (300 mg/l) and acidic conditions (pH 3.3)]. Strict anaerobiosis was not imposed but the fermentation conditions were largely anaerobic due to the design of the fermentor and the effect of CO2 production. We determined the fermentation profile for strain EC1118 after inoculation with dried cells and grown in such conditions (Figure 1). The growth phase was short and about 60% of the sugars were fermented when the cells were in stationary phase. The fermentation rate (dCO2/dt) peaked (Vmax) just before entry into stationary phase and gradually declined thereafter until the end of the fermentation, when sugar reserves were totally exhausted and ethanol concentration had reached 12% (v/v). The accumulation of ethanol followed exactly the same time course as the cumulative release of CO2.

Figure 1.

Fermentation kinetic of EC1118 yeast strain in MS300 medium. The six time-points of sampling for RNA extraction are indicated by arrows and dotted lines. The RNA from stage 1 were used as reference for all hybridization experiments. The CO2 released is directly correlated with the ethanol production. The ethanol concentration in the medium [E] can be obtained using the relation E (g/l) = 1.011 mmath image (g/l) + 2.7 with an error <3%. The residual glucose in the fermentation medium can be also calculated in a similar manner

Gene expression profiles were examined at six time points during the fermentation (Figure 1). Stage 1, which corresponds to the start of the growth phase, was chosen as the reference point for all microarray hybridization experiments. The two first stages corresponded to growing cells, and stage 3 corresponded to cells just entering stationary phase. At stages 4–6, the cells were no longer proliferating and ethanol concentration was increasing. At stage 6, fermentation was almost complete and only a small amount of sugar (5 g/l) was present. Hybridization experiments were performed twice, except at stage 6, when experiments were performed in triplicate because the data obtained with these cells were less reliable, probably due to extreme physiological conditions [3 days of starvation, 11.7% (v/v) ethanol]. The relevant data are available from our website: www.ensam.inra.fr/spo/yeastgenomic.

A large number of genes displayed up- or downregulation at the various stages (Table 1) during fermentation. If a stringent threshold of a tripling in expression was applied, 1090 genes were found to be upregulated and 920 downregulated at at least one stage during the fermentation. The wine yeast transcriptome was found to be stable during the growth phase but displayed marked changes on entry into stationary phase. Only 99 genes displayed changes in expression between stages 1 and 2 (growth phase), whereas the expression of 668 new genes was induced between stages 2 and 3. Further regulations occurred later during the stationary phase (Figure 2). Changes in expression were only transient for some genes, but most of the genes displayed overall up- or downregulation during fermentation, as illustrated by hierarchical clustering of the 2010 regulated genes (Figure 3). Groups of genes involved in a common pathway tended to be clustered in a similar profile, consistent with coordinated regulation of the genes of the corresponding pathways during fermentation. Some such genes were readily identified, such as genes encoding proteins involved in protein biosynthesis (major cluster A), nucleotide or amino acid biosynthesis, all of which were downregulated, and the PAU, AAD and FIT genes and genes encoding proteins involved in methionine biosynthesis (MET cluster), which were upregulated.

Figure 2.

Number of up- or downregulated genes with a factor higher than 3 at each stage. The histograms discriminate the genes as they appear regulated for the first time at each stage, as indicated in the key

Figure 3.

Hierarchical clustering of up- and downregulated genes (1090 and 920 genes, respectively) during the fermentation. A, cytoplasmic ribosomal proteins; B, rRNAs; C, nucleotide metabolism; D, amino acid metabolism; E, subtelomeric genes (unknown function); F, methionine biosynthesis; G, AAD (aryl alcohol dehydrogenase), FIT (iron uptake) genes; H, PAU (putative role in anaerobic sterol uptake) genes. The colour scale at the bottom represents the expression ratio x-fold repressed in green and x-fold induced in red

Table 1. Fermentation parameters for each stage
  1. Number of up- or downregulated genes for three ranges of ratio for each stage. The ratios were in all cases obtained by comparison with stage 1.

Stage 123  456
Cells 10° cells/ml1440115140140140
CO2 g/l 0625 456989
Ethanol % v/v 00.83.3  69.111.7
ExpressionRatioNumber of regulated genes
Upregulated≥10668 965580
 ≥5 and <102210815193164
 ≥8 and <540211272225304
Downregulated≥3 and <522211269224238
 ≥5 and <106147179131118
 ≥10323 65135129
Total upregulated 68387519373548
Total downregulated 31381513490485

Many biosynthetic pathways are downregulated as fermentation progresses

The genes downregulated during fermentation encoded proteins of various functional categories. Many were involved in biosynthetic pathways or in growth-associated functions (DNA replication, secretion, cell cycle control). Several metabolic pathways displayed remarkably coordinated regulation. The largest group of downregulated genes comprised 316 genes encoding proteins involved in protein synthesis: ribosomal proteins; translation factors; ribosomal RNA transcription and processing; tRNA synthesis or aminoacyl-tRNA-synthase (Figure 4). The 117 genes encoding cytoplasmic ribosomal proteins showed a high degree of co-regulation and their expression decreased strongly on entry into stationary phase (stage 3). This highly coordinated regulation is consistent with a major decrease in protein synthesis capacity on entry into stationary phase. Genes encoding mitochondrial ribosomal proteins showed little change in expression, even though mitochondria are poorly developed under fermentation conditions. A similar difference in the response of mitochondrial and cytoplasmic ribosomal protein genes was observed following inhibition of the TOR pathway by rapamycin treatment (Cardenas et al., 1999). Genes encoding proteins involved in amino acid biosynthesis (68 genes) were increasingly downregulated as fermentation progressed. All amino acid synthesis genes were affected, with the exception of those corresponding to the methionine biosynthetic pathway. Genes of the nucleotide biosynthetic pathway, which formed a specific cluster of early downregulated genes, displayed a distinct pattern. The massive downregulation of these biosynthetic pathways is consistent with adaptation according to the yeast's growth requirements.

Figure 4.

Expression profiles of downregulated genes whose products are involved in specific metabolic pathways and functional categories. The genes involved in the ergosterol metabolism and whose products act upstream of the farnesyl pyrophosphate are displayed in red and those acting downstream in black. The colour scale at the bottom represents the expression ratio x-fold repressed in green and x-fold induced in red

Most of the genes encoding proteins involved in ergosterol biosynthesis were downregulated during fermentation, although the precise expression profile depended on the position of the corresponding protein in the pathway (Figure 4). Levels of mRNA for proteins acting downstream from farnesyl pyrophosphate declined later than those for proteins acting upstream, probably due to positive transcriptional control of these genes by anaerobiosis (Kwast et al., 2002). In the conditions used, ergosterol was supplied in the medium and anaerobiosis prevented the ergosterol biosynthetic pathway from functioning. However, under industrial conditions, with natural musts that may have low sterol contents, the capacity to synthesize ergosterol may be essential to protect the yeast against ethanol stress (Alexandre et al., 1994). Ergosterol synthesis can occur if oxygen is added during fermentation. Oxygen is frequently added during industrial wine fermentations, to increase yeast viability and fermentation quality. The efficiency of oxygen additions has been shown to decrease as the fermentation progresses (Sablayrolles, 1996). The general decrease in the expression of ergosterol biosynthesis genes is consistent with this observation.

Genes involved in key metabolic pathways or functions are induced during the growth phase

Several metabolic pathways of known or suspected importance in wine fermentation displayed high levels of remarkably coordinated upregulation during the growth phase. We analysed 15 PAU/TIR genes, which encode putative cell wall proteins. All were strongly induced during the growth phase (some by factors of more than 100) and reached high levels of expression (Figure 5A). Given the high degree of similarity between these genes, it is difficult to identify reliably the PAU gene from which a given transcript was produced (Rachidi et al., 2000). The induction pattern was consistent with the regulation of these genes by anaerobiosis, as previously described (Rachidi et al., 2000). TIR1,2,3 genes displayed similar but not identical patterns of regulation, probably reflecting the involvement of different hypoxia regulation pathways (Abramova et al., 2001). Although the function of the Pau/Tir proteins is still unclear, it has been suggested that they are involved in sterol transport (Wilcox et al., 2002). Interestingly, two recently identified sterol carriers (Wilcox et al., 2002), PDR11 and AUS1, displayed a similar profile (Figure 5C). The high level of induction of these genes with known or suspected involvement in sterol uptake is consistent with the notion that efficient incorporation of sterols into membranes is of critical importance in wine fermentation. The large size of the PAU/TIR gene family (27 members) is also consistent with the proteins encoded by these genes playing a key role in industrial and natural conditions. The regulation of ergosterol carriers by anaerobiosis (Wilcox et al., 2002) seems to be specially designed to ensure high levels of expression during fermentation. However, as maximal levels of expression were not reached until stationary phase, ergosterol uptake capacity seems to be disconnected from cell growth and membrane formation. Other genes involved in lipid metabolism and regulated by anaerobiosis—such as HES1, which encodes a protein involved in ergosterol trafficking, and YSR3, which encodes a protein involved in sphingolipid metabolism—displayed a similar pattern of rapid induction and high levels of expression. The anaerobiosis-regulated OLE1 gene, which encodes a fatty acid desaturase, showed little change in expression during fermentation, with high levels of expression observed throughout (data not shown).

Figure 5.

(A) Expression profiles of upregulated genes whose products are involved in specific metabolic pathways and functional categories. (B) Graphical representation of gene expression profiles of PAU genes. (C) Graphical representation of gene expression profiles of sterol transporters. (D) Graphical representation of gene expression profiles of thiamine metabolism genes. (E) Graphical representation of gene expression profiles of methionine metabolism genes. The missing points correspond to unreliable data

Ten genes of the thiamine biosynthesis or utilization pathways were coordinately induced during the growth phase and continued to be strongly expressed throughout the fermentation (Figure 5A, D). The upregulation of this set of genes indicates coordinated and rapid derepression of the whole thiamine pathway. The cluster includes genes encoding proteins with unclear functions (PET18, SNZ2) thought to be involved in thiamine biosynthesis (Llorente et al., 1999; Rodriguez-Navarro et al., 2002). Most of these genes displayed high levels of expression throughout the fermentation. This early induction of the pathway was surprising because thiamine was present in the medium in amounts (0.74 µM) previously shown to be repressive (Praekelt et al., 1994). This may indicate modification to the control of thiamine biosynthesis gene expression in wine yeast. Such modifications may be related to the critical role played by thiamine in wine fermentation and the high level of activity of enzymes using thiamine diphosphate and pyruvate decarboxylase in particular. Thiamine availability in musts has been shown to have a strong impact on the kinetics of fermentation (Bataillon et al., 1996). Genes encoding proteins involved in biotin metabolism were also co-regulated and induced early in the fermentation (Figure 5A), despite the abundance of biotin in the must (3 µM). This upregulation of biotin biosynthesis genes during wine fermentation has been reported in previous studies and may reflect a large requirement for this vitamin under wine fermentation conditions (Backhus et al., 2001).

Genes encoding proteins of the sulphur amino acid biosynthetic pathway formed a specific cluster (16 genes) of genes that were transiently upregulated during yeast growth (Figure 5E). This highly coordinated expression of the enzymes of the sulphur assimilation pathway is consistent with tight control by common transcription factors (reviewed in Thomas and Surdin-Kerjan, 1997). The coordinated control of MET genes depends principally on the activator Met4 and its associated factors Cbf1, Met28, Met31 and Met32. The early induction of these genes is consistent with the use of low, subrepressive amounts of methionine in the medium (200 µM) and the rapid use of this amino acid by the yeast (see Figure 7A). The MET genes were downregulated when the yeast stopped growing, consistent with MET gene expression being tightly correlated to cell growth (Patton et al., 2000). The MET30 repressor was induced only later in stationary phase (not with the MET cluster), suggesting that S-adenosylmethionine (which induces MET30) accumulation did not occur before the yeast stopped growing and was not involved in the downregulation of the MET gene cluster. Expression of the MET genes peaked just before the time at which the release of H2S is generally observed during wine fermentation (at the end of growth) (Jiranek et al., 1995). The flux of sulphur through the sulphate assimilation pathway is of key technological importance in wine-making because the release of excessive amounts of H2S may have a negative impact on the sensorial qualities of the wine. Fine-tuning of the pathway is clearly required to ensure that only small amounts of this metabolite are produced. Coordinated expression of the genes corresponding to this pathway is the first level of such tight control and may reflect the need for yeasts producing low levels of H2S for wine-making.

Weak regulation of carbohydrate metabolism genes

The behaviour of the genes encoding proteins involved in sugar metabolism is of prime importance under alcoholic fermentation conditions. Most of the genes encoding proteins involved in glycolysis and fermentation were strongly expressed throughout the fermentation (data not shown). The main cases of significant changes in expression corresponded to switches from one isoform to another and to the induction of stress-responsive isoforms (Gasch et al., 2000). This was the case for HXK1, GLK1, PGM2 and GPM2 (Figure 6A), and led to a shift from Hxk2 to Hxk1, which may favour fructose utilization at the end of the fermentation. Fructose is the main sugar remaining at the end of fermentation and the Hxk1 isoform has a higher affinity than the Hxk2 isoform for fructose. Stress regulation probably also accounts for the induction of three ALD genes (ALD2, ALD3, ALD4) on entry into stationary phase, while ALD6, the major isoform involved in acetate formation, is downregulated (Remize et al., 2000). Although major changes in the expression of the HXT genes encoding hexose carriers are known to occur (Riou et al., 1997 and unpublished results), they could not be accurately addressed here, given the high level of similarity between HXT genes, which might result in cross-hybridization.

Figure 6.

(A) Gene expression profiles of isoforms involved in carbohydrate metabolism. The expression ratios are depicted using the colour code described for Figure 3. (B) Expression profiles of genes known to be submitted to carbon catabolic repression which display an induction during the fermentation. The expression ratios are depicted using the colour code described for Figure 3

The GPD1 gene was the most strongly expressed isoforms of glycerol-3-phosphate dehydrogenase (expressed more than 10 times more strongly than GPD2), consistent with the demonstrated osmotic stress response of yeast (Perez-Torrado et al., 2002). Surprisingly, the GCY1 and DAK1 genes, which encode proteins involved in glycerol utilization, were also upregulated on entry into stationary phase. Co-expression of the genes encoding the proteins responsible for the synthesis and breakdown of glycerol indicate the probable existence of a futile cycle for controlling redox equilibrium by converting NADH to NADPH (reviewed in Blomberg, 2000).

Genes encoding proteins involved in respiration, the TCA cycle and neoglucogenesis, all of which are known to be subject to carbon catabolite repression, were generally expressed at very low levels, consistent with repression by the high concentration of glucose during most of the alcoholic fermentation. However, several genes (SUC2, MAL31, KGD1, QCR8, CIT1, MLS1, POT1) displayed levels of expression higher than expected for glucose-repressed genes and were induced at growth arrest or during the stationary phase (Figure 6B). This indicates that some genes may be partially derepressed during fermentation, despite high sugar levels. This regulation may involve the modulation of carbon repression by starvation and/or nitrogen depletion because growth arrest was triggered by nitrogen exhaustion (see below). These observations are consistent with those of Backhus et al. (2001), indicating the relief from carbon repression of various genes in response to low levels of nitrogen during wine fermenation. The impact of this regulation on yeast metabolism in conditions of fermentation is unclear.

Growth arrest is caused by nitrogen depletion and is associated with major changes in transcription

Under oenological conditions, nitrogen often limits yeast growth. Monitoring of nitrogen sources showed that the assimilable nitrogen was completely consumed during the growth phase (Figure 7A). The amino acids were assimilated with variable patterns, depending on yeast preferences. Proline was not utilized because oxygen is required for its metabolism (Henscke and Jiranek, 1993). With the exception of cysteine, which was not utilized by our strain, all the nitrogen sources considered assimilable under wine fermentation conditions (‘assimilable nitrogen’ is defined as nitrogen in the form of ammonia and α-amino nitrogen of amino acids other than proline) were utilized. The timing of nitrogen utilization and cell growth was consistent with growth arrest being triggered by exhaustion of the assimilable nitrogen. Growth arrest was found to be associated with major changes in the expression of genes encoding proteins involved in nitrogen use. The main feature of these changes was the induction of a large set of genes involved in the metabolization of poor, alternative nitrogen sources (Figure 7E). Genes required for the utilization of proline (PUT1, PUT2, but not PUT4, encoding the proline carrier), allantoin (DAL1, DAL2, DAL3, DAL4, DAL5, DAL7, DAL80, DAL81, DAL82) and urea (DUR1,2) were induced (Figure 7B, C). Genes encoding nitrogen permeases (GAP1, CAN1, DUR3, MEP2, PTR2) or proteins involved in management of the glutamate pool (GDH2, GLN1, GAD1, UGA1) were also upregulated at growth arrest. All these genes are subject to nitrogen catabolite repression (NCR) and are controlled by the TOR pathway (Cardenas et al., 1999). The Tor kinases have been shown to regulate the expression of NCR-sensitive genes by controlling binding of the GATA-type transcription factor Gln3 to the repressor Ure2. The changes in expression observed at growth arrest perfectly matched those described after inhibition of the TOR pathway by rapamycin treatment (Cardenas et al., 1999). Tor proteins have been shown to control the transcription of genes according to nutrient availability and to enable yeast to shift from high- to low-quality nitrogen sources (Hardwick et al., 1999; Kuruvilla et al., 2001). This situation corresponds to the stage in fermentation when all the preferred nitrogen sources are exhausted and only proline remains. The absence of oxygen resulted in a lack of proline utilization, despite the induction of several genes involved in proline metabolism. The coordinated upregulation of genes encoding proteins involved in nitrogen utilization is consistent with control by the TOR pathway. Although genes encoding proteins involved in nitrogen metabolism were derepressed only when good nitrogen sources were exhausted, the NCR-sensitive AGP1 gene, which encodes an amino acid carrier (Regenberg et al., 1999), escaped this repression. This gene was strongly expressed during the growth phase but only weakly expressed thereafter (data not shown), indicating that other factors had an effect on its expression.

Figure 7.

Genetic and physiological response elicited by depletion of assimilable nitrogen. (A) Time-course evolution of amino acids and assimilable nitrogen during the fermentation. The 19 amino acids and the NH4+ were monitored. Only some representative profiles are displayed. Ammoniacal nitrogen and α-amino nitrogen except proline are considered as assimilable nitrogen (Ass. N.). In this calculation of assimilable nitrogen only one atom of nitrogen per molecule of amino acid is taken into account. (B) Expression profiles of genes whose products are involved in allantoin metabolism. (C) Expression profiles of genes whose products are involved in amino acid utilization, nitrogen permeases and glutamate pool management. (D) Expression profiles of genes whose products are involved in protein degradation. (E) Scheme of glutamate pool management and poor nitrogen source utilization. The genes indicated in red are upregulated during the fermentation process (transcriptional regulators of allantoin metabolism genes are not represented). (F) Glycogen and trehalose accumulation kinetics. (G) Expression profiles of genes whose products are involved in glycogen metabolism. (H) Expression profiles of genes whose products are involved in trehalose metabolism

Other TOR target genes, including the vacuolar protease genes PRB1 and PEP4, and genes encoding proteins involved in autophagy, such as APG1 and AUT7, were also induced at growth arrest (Figure 7D). The concomitant induction of TFS1 and PAI3, which encode protease inhibitors, suggests that protein degradation is tightly controlled at the post-translation level. The induction of this set of genes is consistent with active nitrogen recycling through vacuolar proteolysis during starvation.

Entry into stationary phase is associated with other cellular events that may also be controlled by TOR. We monitored levels of the storage carbohydrates glycogen and trehalose during fermentation (Figure 7F). Yeast cells began to accumulate storage carbohydrates from growth arrest. Genes encoding proteins involved in the metabolism of glycogen and trehalose were induced during the growth phase and their expression peaked in stationary phase (Figure 7G, H). There was a short time-lag between gene induction and carbohydrate storage. Genes encoding proteins involved in the synthesis and breakdown of glycogen and trehalose were coordinately induced. The slight delay in carbohydrate storage and the simultaneous induction of genes encoding proteins involved in the synthesis and breakdown of these carbohydrates suggest that homeostasis was achieved by post-transcriptional control. This should enable the yeast to fine-tune its carbohydrate reserves rapidly in response to cellular signals. The TOR pathway is thought to control glycogen and trehalose accumulation at the post-translational level, in response to nitrogen depletion (Francois and Parrou, 2001). Under these fermentation conditions, carbohydrates begin to accumulate when the assimilable nitrogen is depleted; therefore, TOR is also likely to trigger the storage of carbohydrates at growth arrest in response to nitrogen exhaustion.

The stress response is primarily associated with the stationary phase

Many yeast genes are known to be regulated by stress and global expression analyses have shown that a large set of genes respond in a similar manner to many stresses (Causton et al., 2001; Gasch et al., 2000). The corresponding groups of genes have been referred to as ESR (environmental stress response) or CER (common environmental response) genes, and these two groups overlap considerably. We examined the expression profiles of the 367 genes upregulated in the ESR + CER groups: 213 genes were upregulated during fermentation. Many stress genes were induced on entry into stationary phase or shortly afterwards. Only 17 ESR/CER genes were upregulated at stage 2 vs. 145 ESR/CER genes at stage 4. This is consistent with the stationary phase constituting a condition of stress (Werner-Washburne et al., 1993), probably amplified during fermentation, due to the accumulation of ethanol.

The proteins encoded by the induced genes are involved in many cellular processes. Given the broad definition of the environment stress response, most of the genes shown to respond to nitrogen depletion or to encode proteases or proteins involved in nitrogen utilization or autophagy are considered to be members of the ESR group. The classical stress-responsive heat shock genes (HSP12, HSP26, HSP78, HSP104, HSP42 SSE2), which encode proteins primarily involved in protein folding, were strongly induced (Figure 8A). Although these genes displayed different expression profiles, all were induced towards the end of the growth period and remained induced throughout the bulk of the stationary phase. The HSP26 gene was found to be one of the most regulated genes of the genome, showing about 100-fold induction. HSP12 was strongly induced despite the high levels of glucose present at growth arrest (140 g/l) and the fact that this gene has been shown to be glucose-repressed (de Groot et al., 2000) as previously observed (Perez-Torrado et al., 2002). Various genes encoding proteins with unclear molecular function that have been shown to respond to many stresses (SIP18, SPS100, GRE1, SPI1, PST1, YNL200c, YAL61w) were also strongly upregulated, with maximal levels of expression during the stationary phase (Figure 8B). Several of these highly stress-responsive genes have been shown to be induced in wine fermentation stationary phase (Riou et al, 1997; Puig and Perez-Ortin, 2000). Many of these genes are regulated by the general stress transcription factors Msn2/Msn4 (Causton et al., 2001; Estruch, 2000) and MSN4 itself was upregulated at growth arrest (data not shown). The AAD genes, which encode putative aryl-alcohol dehydrogenases, were upregulated with expression peaking at growth arrest (Figure 8C). The induction of these genes may reflect a response to oxidative stress (Delneri et al., 1999). In addition, we observed that various genes involved in protection against oxidative stress were strongly expressed throughout fermentation (data not shown). This is consistent with the idea that yeast has to protect against this stress even under anaerobiosis.

Figure 8.

Expression profiles of stress-responsive genes. (A) Expression profiles of HSP upregulated genes. (B) Expression profiles of highly upregulated genes with unclear molecular function and previously shown to be regulated by stress. (C) Expression profiles of the AAD4 and AAD6 genes. The other AAD genes display the same profile but only AAD4 and AAD6 were shown to be regulated by oxidative stress in a laboratory strain (Delneri, 1999). (D) Expression profiles of genes involved in H+ homeostasis. (E) Expression profiles of genes involved in cell wall biogenesis. (F) Expression profiles of six subtelomeric genes of unknown function

Genes encoding proteins involved in H+ homeostasis—HSP30, PTK2 and KHA1 which encode a negative regulator of the plasma membrane ATPase, a positive regulator of this enzyme and a putative H+/K+ antiporter, respectively—were induced at or after growth arrest (Figure 8D). The control of H+ homeostasis is thought to be an important target of ethanol stress (Alexandre et al., 2001) and the simultaneous induction of a positive and a negative regulator of Pma1 is consistent with fine-tuning of H+ extrusion activity. The plasma membrane H+ ATPase gene PMA1 was strongly expressed, and its expression in the stationary phase decreased to about 50% of its initial level (data not shown). The induction of the H+/K+ exchanger gene, KHA1, is also consistent with K+ being beneficial for yeast viability during wine fermentation (Kudo et al., 1998). The exact contribution of ethanol to the induction of this group of genes, and more generally to the genetic responses observed, is difficult to determine because ethanol stress coincides with starvation. Moreover, many ethanol-responsive genes are also induced by starvation. At growth arrest (stage 3), the concentration of ethanol was only 3.3%, which is generally considered as sub-stressing. The genes induced at stage 3 therefore probably responded to starvation rather than to ethanol. The pattern of induction may also provide insight into the contribution of ethanol to the genetic response. The HSP30 expression profile, with expression continuing to increase after growth arrest, may be consistent with a response to gradually increasing ethanol concentration.

An ethanol stress response is certainly responsible for some of the uniqueness of the response observed during the stationary phase for wine fermentation. This gene response is unique because only 30% of the induced genes correspond to genes reported to be induced in previous reports of stationary phase response in different conditions (aerobic, carbon-starved yeasts) (Gasch et al., 2000). The induction of genes encoding proteins involved in cell wall biogenesis (FKS1, GSC2, SSD1, MPT5), which showed little or no regulation by starvation in other studies, may be indicative of a specific response to ethanol (Figure 8E). These genes displayed similar patterns, with late induction during fermentation, suggesting that cell wall thickening may help the yeast to cope with ethanol.

The response observed in stationary phase involved many genes encoding proteins of unknown function. These genes included a set of subtelomeric genes (YLR463c, YEL074w, YEL073c, YHL049c, YHL050c, YOL159c, YPL272c, YPL277c, YPL278c) not previously described as stress-responsive, that were strongly induced (Figure 8F).

The stress response during wine fermentation appeared to be primarily associated with the stationary phase. However, various stress-responsive genes (UBI4, HSP150, YRO2), although not differentially regulated, were expressed at high levels throughout fermentation, suggesting that they responded to a very early stress (data not shown). The initial stressors—acidity and osmotic pressure—did not trigger a permanent and general stress response. Acidity seems to trigger a response only at the beginning of fermentation, given the profile of PDR12, which can be used as a marker of the acid response. This gene is specifically induced by acid (Causton et al., 2001). In our study, it was transiently induced at stage 2 and its expression decreased thereafter (data not shown). The osmotic stress response appeared to occur early and to be restricted to a few responsive genes. The strongly osmo-responsive genes GPD1, HOR2 and HOR7 were induced early and maintained high expression levels throughout the fermentation (data not shown). Only some specific genes displayed patterns of regulation consistent with previous observations, indicating that major stress responses are only transient (Gasch et al., 2000).

Discussion

We show here that during alcoholic fermentation, a wine yeast undergoes a specific transcriptional programme that differs in many aspects from that observed under standard laboratory conditions. During wine fermentation, the yeast adapts to changing nutritional, environmental and physiological conditions by sequentially activating or repressing large numbers of genes encoding proteins involved in many metabolic pathways. We found that genes encoding proteins for a given pathway were often regulated in a highly coordinated manner, consistent with tight control by common regulatory pathways. In many cases, the observed patterns of regulation were consistent with the known gene controls, but many genes did not respond as expected—upregulation of nitrogen- or carbon-repressed genes despite the presence of the repressing factor, for example—indicating a complex interplay of signalling and regulatory networks under these conditions. Of course, some of the changes in expression may be specific to our wine yeast strain. The underlying regulatory networks appear to enable the yeast to carry out fermentation efficiently in such conditions because the genes encoding proteins involved in critical metabolic pathways—glycolysis, ergosterol uptake, thiamine biosynthesis—were strongly expressed throughout the entire fermentation process. A substantial number of the strongly expressed genes are regulated by anaerobiosis (genes involved in sterol uptake and trafficking, cell wall proteins, fatty acid desaturation). A strong response to anaerobiosis is probably important for wine yeasts, given the strong anaerobiosis in many industrial conditions.

In wine fermentations, yeast growth is never restricted by carbon limitation, but always by another nutrient, frequently nitrogen. Our data strongly indicate that under such conditions the TOR nitrogen-sensing pathway plays a key role in modifying transcription patterns and in the control of other events associated with nitrogen depletion. TOR triggers the reorientation of nitrogen metabolism and probably also of carbohydrate storage, together with downregulation of the ribosomal protein genes. The involvement of the TOR cascade may also have other implications because of the central role played by this cascade in the control of cellular processes. Tor kinases control cell growth as a function of nutrient availability (reviewed in Schmelzle and Hall, 2000). Thus, growth arrest is probably triggered by the TOR cascade, in response to nitrogen exhaustion. Our data show that the switch from the replicative to the non-replicative state is a key physiological event in wine fermentation and is followed by a general stress response. Many stress-responsive genes are thought to be controlled by the general stress factors Msn2/Msn4, which shuttle between the cytoplasm and the nucleus. This shuttling is controlled by the TOR pathway (Beck and Hall, 1999) and protein kinase A (reviewed in Estruch, 2000). Given the role of TOR in nitrogen sensing, the inhibition of TOR by nitrogen depletion might be the initial trigger of the stress response. Other pathways, such as that involving protein kinase A, may be involved in stress signalling at growth arrest (Gorner et al., 1998). However, given the large amount of sugar, and therefore of cAMP, only cAMP-independent control of protein kinase A may operate (de Winde et al., 1997).

Our data provide insight into the temporal impact of the various stresses faced by yeast during fermentation. The initial stresses—high osmotic pressure and acidity—did not give rise to a permanent stress response. As indicated by other studies, the response to osmotic stress immediately after inoculation is rapid and transient (Perez-Torrado et al., 2002), and this seems to apply to most of the stress-responsive genes. A general stress response was established only in stationary phase. Unexpectedly, we found that ethanol did not appear to be the initial trigger of the stress response. Instead, our data suggest that ethanol stress imposes itself later on, in addition to the initial stress due to growth arrest. It is difficult to determine the precise role of ethanol in the stress response because many target genes are common to different stresses and no ethanol stress-specific genes have been identified. During the stationary phase, yeast cells integrate numerous stress signals (ethanol, osmotic, acid, nutrient depletion), and this clearly involves various signalling pathways. Physiological conditions at these stages—carbon repressed cells, anaerobiosis—are extreme and give rise to a unique stress response that involves a large variety of cellular targets. Moreover, many of the genes that respond at these stages encode proteins of unknown function, highlighting the limits of our understanding of the molecular mechanisms of wine yeast survival under such stress conditions. A substantial number of these genes were not found to be regulated by various environmental stresses in previous studies, suggesting that wine fermentation conditions may be suitable for determining the function of unknown genes. Further experiments are required to determine the elements of the described stress response that are specific to wine strains and contribute to their high level of stress resistance.

Analysis of the regulation of genes according to their chromosomal location indicated that subtelomeric regions harboured large numbers of induced genes: 112 of 320 subtelomeric genes upregulated (35% induced vs. 17% for the whole genome). When the same analysis is applied to the data from a diauxic shift analysis (de Risi et al., 1997), the amount of regulated genes are respectively 6.2% (subtelomeric) and 6.6% (whole genome). This suggests that subtelomeric genes are more strongly regulated than usual under wine fermentation conditions. These genes include various members of multigene families (PAU, AAD) that may artificially increase the number of regulated genes by cross-hybridization between homologous genes. Subtelomeric genes are thought to be new genes that have recently expanded in S. cerevisiae due to the specific adaptation of this species (Llorente et al., 2000), and many genes located at chromosome ends are involved in sugar fermentation (SUC, MAL, MEL). These chromosomal regions are known to be subject to genomic changes and variations in the gene copy number of the wine strain may contribute the regulations we observed. The regulation of subtelomeric genes during wine fermentation is consistent with the notion that human fermentation practices, such as wine making, have played an important role in the evolution of the yeast genome (Mortimer 2000). Regulated subtelomeric genes encoding proteins of unknown function may therefore be involved in the adaptation of S. cerevisiae for sugar fermentation.

The transcriptome analyses were performed in conditions representative of industrial situations. Although genomic regulation may differ in natural musts with a different nutritional status, the main patterns of gene regulation observed are probably relevant to many wine fermentation situations. The wealth of information provided by this study represents a starting point for deciphering the regulatory circuits during wine fermentation and should help us to understand the properties of wine yeasts. The regulation of many small metabolic pathways, in some cases involving the products of single genes, of value in wine making, cannot be discussed here. More detail is available from the companion website.

Acknowledgements

The authors thank Martine Pradal for technical assistance in amino acid analyses and Gille Carraux for help in cluster analysis with PermutMatrix software.

Ancillary