Quantitative analysis of wine yeast gene expression profiles under winemaking conditions

Authors

  • Cristian Varela,

    1. Departamento de Ingeniería Química y Bioprocesos, Facultad de Ingeniería, Pontificia Universidad Católica de Chile, Santiago, Chile
    Current affiliation:
    1. The Australian Wine Research Institute, PO Box 197, Glen Osmond, Adelaide, SA 5064, Australia.
    Search for more papers by this author
  • Javier Cárdenas,

    1. Departamento de Ingeniería Química y Bioprocesos, Facultad de Ingeniería, Pontificia Universidad Católica de Chile, Santiago, Chile
    Search for more papers by this author
  • Francisco Melo,

    1. Departamento de Genética Molecular y Microbiología, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, Santiago, Chile
    Search for more papers by this author
  • Eduardo Agosin

    Corresponding author
    1. Departamento de Ingeniería Química y Bioprocesos, Facultad de Ingeniería, Pontificia Universidad Católica de Chile, Santiago, Chile
    • Pontificia Universidad Católica de Chile, Casilla 306 Correo 22, Santiago, Chile.
    Search for more papers by this author

Abstract

Wine fermentation is a dynamic and complex process in which the yeast cell is subjected to multiple stress conditions. A successful adaptation involves changes in gene expression profiles where a large number of genes are up- or downregulated. Functional genomic approaches are commonly used to obtain global gene expression profiles, thereby providing a comprehensive view of yeast physiology. We used SAGE to quantify gene expression profiles in an industrial strain of Saccharomyces cerevisiae under winemaking conditions. The transcriptome of wine yeast was analysed at three stages during the fermentation process, mid-exponential phase, and early- and late-stationary phases. Upon correlation with the yeast genome, we found three classes of transcripts: (a) sequences that corresponded to ORFs; (b) expressed sequences from intergenic regions; and (c) messengers that did not match the published reference yeast genome. In all fermentation phases studied, the most highly expressed genes related to energy production and stress response. For many pathways, including glycolysis, different transcript levels were observed during each phase. Different isoenzymes, including hexose transporters (HXT), were differentially induced, depending on the growth phase. About 10% of transcripts matched non-annotated ORF regions within the yeast genome and could correspond to small novel genes originally omitted in the first gene annotation effort. Up to 22% of transcripts, particularly at late-stationary phase, did not match any known location within the genome. As the available reference yeast genome was obtained from a laboratory strain, these expressed sequences could represent genes only expressed by an industrial yeast strain. Further studies are necessary to identify the role of these potential genes during wine fermentation. Copyright © 2005 John Wiley & Sons, Ltd.

Introduction

Yeasts are subjected to multiple and changing stress conditions during alcoholic fermentation, which is a dynamic and complex process. These conditions include hyperosmotic shock, nutrient limitation and starvation, temperature variations and ethanol toxicity. Wine yeasts have evolved mechanisms to sense and respond to environmental changes and thus maintain metabolic activity and cellular integrity (Bauer and Pretorius, 2000). The stress conditions that occur during wine fermentation affect the yeast's metabolism, producing a metabolic response that in turn enables the cell to adapt to the new environmental conditions. A successful adaptation implies a metabolic reorganization in order to maintain cellular activity (vitality). Sometimes, this reorganization leads to a decrease in the fermentation rate, generating sluggish fermentations (Walker, 1998). When the adaptation mechanisms fail, the reduced viability is followed by a decrease in vitality, resulting in stuck fermentations. A successful adaptation also involves changes in gene expression profiles where a large number of genes are up- or downregulated. Therefore, obtaining the specific expression profiles of wine yeast during wine fermentation would allow understanding, at the molecular level, of both the biological process of wine fermentation and the regulation of gene expression in response to changes in the environment (Perez-Ortin et al., 2002).

Among molecular biology techniques that enable the whole transcriptome to be studied, only DNA microarrays have been used to describe the transcriptome of wine yeast studied in standard laboratory culture media, chemically defined wine media and grape must (Backhus et al., 2001; Erasmus et al., 2003; Hauser et al., 2001; Marks et al., 2003; Rossignol et al., 2003). Unfortunately, DNA microarrays are not fully quantitative and results from different experiments cannot be compared directly. From a global point of view of cell metabolic regulation, where genomic information could be complemented with gene expression profiles, total protein concentration or data from intracellular metabolites, it is of key importance to use methods that enable quantification of all the transcripts present within the yeast cell. Serial analysis of gene expression (SAGE), a technique developed by Velculescu et al. (1995), enables the analysis of thousands of expressed genes and a total quantification of each transcript.

SAGE is an experimental technique designed to gain a direct and quantitative measure of gene expression (Velculescu et al., 1995). This technique has already been used to quantify the yeast transcriptome (Kal et al., 1999; Velculescu et al., 1997), as well as to study the expression profiles in different cellular types (Blomberg and Zuelke, 2004; Chen et al., 1998; Norman et al., 2004). SAGE is based mainly on two principles: representation of mRNAs by short (9–10 bp) nucleotide sequences (tags); and linkage of these tags for cloning, which allows an efficient sequencing analysis (Stollberg et al., 2000; Yamamoto et al., 2001). The expression level of each transcript is quantified by the frequency that a particular tag is observed. Unlike DNA microarrays, SAGE does not require prior knowledge of the genes to be analysed; indeed, SAGE allows efficient identification of novel transcripts or novel genes in the genome (Chen et al., 2000; Saha et al., 2002).

In this work, we used SAGE to quantify the gene expression profiles of the wine yeast Saccharomyces cerevisiae EC1118 under winemaking conditions. To the best of our knowledge, this is the first report where the transcriptome of an industrial strain of S. cerevisiae is quantitatively assessed. The differential expression of genes related to stress response, the discovery of potential novel genes and the expression of oxygen-dependent pathways under anoxic conditions, are discussed.

Material and methods

Yeast strain and culture medium

An isolate of S. cerevisiae EC1118 (Lalvin, Zug, Switzerland), a widely used commercial wine strain, was used throughout this study. Initial seed cultures were grown in YPD medium at 28 °C under aerobic conditions. MS300 artificial must, which simulates a standard grape juice, was used in bioreactor fermentations (Salmon and Barre, 1998). MS300 medium was modified by increasing total sugar concentration to 240 g/l and including equal parts of glucose and fructose (Varela et al., 2004). Assimilable nitrogen content of MS300 medium was 300 mg N/l supplied as ammonia and amino acids.

Growth conditions

A 50 l Bioengineering bioreactor (Bioengineering, Wald, Switzerland) with a 35 l working volume was inoculated to an initial density of 1 × 106 cells/ml. Cells were washed with 0.9% NaCl to eliminate any remaining nitrogen from the rich media prior to inoculation. The temperature was maintained at 28 °C and the pH at 3.5. Nitrogen was used to sparge the medium for 30 min (250 ml/min) before inoculation to eliminate any oxygen. Agitation was kept at 100 rpm to hold cells in suspension. Carbon dioxide production, in addition to nitrogen sparging and weak agitation, ensured anaerobic conditions throughout the experiment. Three independent experiments were carried out to obtain fermentation profiles.

Analytical techniques

Carbon dioxide evolution in the bioreactors was determined with a Gallus 1000 volumetric flux transductor (Schlumberg, Buenos Aires, Argentina). Culture samples were taken periodically to establish the fermentation status. These samples were analysed to determine the dry cell weight, the cell number, and the concentrations of glucose, fructose, organic acids, amino acids, ammonia and free amino acid nitrogen. Dry cell weight was estimated by filtering cells and washing them twice with distilled water and then drying the preparation to a constant weight at 85 °C. Cell numbers were estimated microscopically using a Neubauer chamber (Brand, Wertheim, Germany). Glucose and fructose concentrations were measured by high-performance liquid chromatography (HPLC), using a Waters high-performance carbohydrate cartridge (Varela et al., 2004). Organic acids, glycerol and ethanol concentrations were measured by HPLC, using a Bio-Rad HPX-87H column (Varela et al., 2003). Amino acids were derivatized with Waters AccQ Fluor reagent, and then the concentrations were measured by HPLC, using an AccQ Tag amino acid analysis column in accordance with the instructions of the manufacturer (Waters, Milford, MA). The ammonia concentration was measured enzymatically, using glutamate dehydrogenase (Sigma, St. Louis, MO). The concentration of free amino acid nitrogen was determined by using the σ-phthaldehyde/N-acetyl-L-cysteine spectrophotometric assay (NOPA) procedure (Dukes and Butzke, 1998). Viability was measured by using a LIVE/DEAD yeast viability kit (Molecular Probes, Eugene, OR) as previously described (Varela et al., 2004).

mRNA extraction and SAGE procedure

The yeast transcriptome was analysed at three stages of the fermentation process: mid-exponential phase, and early- and late-stationary phases. Total RNA was extracted from the samples using Trizol reagent (Life Technologies, Carlsbad, CA) following the manufacturer's instructions. Total RNA was used to obtain the SAGE libraries using the I-SAGE kit (Invitrogen, Carlsbad, CA) as described by the manufacturer. Briefly, mRNA was converted into double-stranded copy DNA (cDNA) using a biotinylated oligo(dT) primer. cDNA was digested with NlaIII and 3′ cDNAs fragments were isolated, using streptavidin paramagnetic beads. 3′ cDNAs were split into two pools and SAGE linkers A and B were ligated to pools 1 and 2, respectively. SAGE tags were released with BsmF1 and blunt-ended with T4 polymerase. The tags from pools 1 and 2 were then ligated to each other. The ligation product was amplified with 28 cycles of a polymerase chain reaction (PCR) and digested with NlaIII. Ditags were isolated from a 12% polyacrylamide gel, concatemerized and cloned into pZero digested with SphI. Clones containing at least 500 bp inserts were sent for sequencing to Agencourt Bioscience (Beverly, MA).

SAGE data analysis

SAGE data analysis involved four parts: (a) determination of tag abundance; (b) identification of each observed tag (genome mapping); (c) determination of statistically significant differences in expression levels; and (d) cluster analysis. For tag abundance determination, analysis of sequencing data was performed using the software package SAGE 2000 (Velculescu et al., 1995; Zhang et al., 1997). For tag identification we generated a database with tags from the whole yeast genome (manuscript in preparation). The annotated tags database contained all the 14 bp sequences in the yeast genome that started with the NlaIII restriction site (5′-CATG-3′). All potential tags were linked to the existing gene annotations from the NCBI database. Then, the observed tags were merged with the constructed database and the expression of different genes was determined. The criterion used to map the tags was 100% identity matches within the gene or up to 500 bases downstream (towards the 3′ end), unless an ORF was annotated before that downstream length was accomplished. When different tags originating from the same gene were found, these tags were added to calculate expression levels. It is noteworthy that the resulting database allowed the annotation of 47% more tags than those originally classified as NORFs and NIDs, when employing the tag annotated database available from SaccharomycesGenome Database (SGD) (Cherry et al., 1998), which contains only experimentally obtained sequence tags.

Determination of statistically significant differences was performed using the SAGEstat software, which allows to compare differences between two libraries (Kal et al., 1999; Ruijter et al., 2002). Hence, the phases: mid-exponential–early-stationary, early–late-stationary and mid-exponential–late-stationary were analysed statistically as pairs. After this analysis, the total numbers of tags from the three SAGE libraries was normalized to 15 000 mRNAs to obtain the number of transcripts per cell (Kal et al., 1999).

Cluster analysis was performed using Seqexpress 1.2.1. This software allows the determination of the optimal number of clusters by applying the expectation maximization algorithm (Boyle, 2004). Transcripts were clustered according to the expression profile and expression level. Therefore, 13 clusters were defined for representing upregulation, downregulation and constant expression at several expression levels. SAGE data sets are available on the web (www2.ing.puc.cl/∼sage).

Metabolic fluxes and gene expression

Flux distributions at mid-exponential phase and early- and late-stationary phases were determined by metabolic flux analysis and have been reported previously (Varela et al., 2004). This analysis enabled quantification of fluxes through the metabolic pathways which describe the anaerobic metabolism of S. cerevisiae. Then, metabolic fluxes of every biochemical reaction were compared with the expression levels of the gene involved in that particular reaction. Two ‘entry gates’ or groups of reactions that feed a particular pathway were defined. The glycolysis entry gate grouped glucose and fructose transport reactions and the reactions responsible for the conversion of both sugars into phosphate sugars. The TCA cycle entry gate grouped the reactions involved in the formation of isocitrate from pyruvate: pyruvate dehydrogenase, citrate synthase and aconitase.

Results

Fermentation profiles

The average time to reach dryness (<4 g sugar/l wine) was 6 days (170 h) (see Figure 1A in Varela et al., 2004). S. cerevisiae EC1118 consumed glucose in preference to fructose, which is a result of differences in the transporters' affinities for these sugars (Reifenberger et al., 1997). The biomass growth curve showed an exponential phase and a stationary phase that began at 48 h, at which point the biomass concentration was 5.8 g/l. Ammonia was depleted after 24 h of culturing, and this coincided with the highest specific growth rate (0.2 h−1). Assimilable nitrogen was depleted from the medium at 48 h, which resulted in cell growth arrest. Even though all assimilable nitrogen was consumed, the cell viability remained greater than 97% until all the sugar was depleted. Ethanol synthesis occurred mainly in the stationary phase and resulted in a final ethanol concentration of 12.7% (v/v). Besides ethanol, the yeast produced a number of other products during the fermentation. The final concentration of glycerol, quantitatively the second most important product of wine fermentation, was 7.8 g/l. Other significant compounds produced by the yeast were succinic acid and acetic acid, whose final concentrations were 1.8 and 1.0 g/l, respectively.

Figure 1.

Relative distribution of gene expression levels at different stages of wine fermentation. Mid-exponential phase (white columns), early-stationary phase (grey columns) and late-stationary phase (black columns). Circles represent expression levels at log phase of S. cerevisiae YPH499 grown in YPD medium (Velculescu et al., 1997). Total tags obtained by phase: 4881 for mid-exponential; 3483 for early-stationary; 4327 for late-stationary; and 20 184 for log phase in rich medium

Gene expression profiles were examined at mid-exponential phase, and early- and late-stationary phases (Table 1). At mid-exponential phase (20 h), sugar concentration was very high (∼200 g/l), whereas ammonia was almost completely consumed (∼5 mg/l). At early-stationary phase (48 h), cell growth was arrested, since all assimilable nitrogen was depleted from the medium. At late-stationary phase (96 h), sugar was still present in the medium (∼16 g/l), whereas ethanol concentration had reached 12% (v/v).

Table 1. Fermentation values at different yeast growth phases
 Mid-exponential phaseEarly-stationary phaseLate-stationary phase
  1. Data obtained from three independent experiments.

Time (h)204896
DCW (g/l)2.0 ± 0.25.8 ± 0.15.8 ± 0.1
Residual sugar (g/l)204 ± 5.1100 ± 3.816 ± 3.9
Yeast assimilable nitrogen (mg N/l)76 ± 5.97 ± 27 ± 2
Ethanol (g/l)9.6 ± 3.350.7 ± 3.194.2 ± 3.7

Yeast transcriptome

The total number of tags obtained was 12 691 (including 4881 for mid-exponential phase, 3483 from early-stationary phase and 4327 from late-stationary phase). Analysis of the relative distribution of gene expression levels showed that only 0.4% of the messengers were expressed at 100 copies/cell or more, 11.4% at more than 10 copies/cell, whereas most transcripts (88.6%) were present at 10 copies/cell or less (Figure 1). This had already been observed for yeast cells grown in rich medium (Velculescu et al., 1997). For subsequent analyses, only tags present at least twice were considered, since those present only once are not reliable (Table 2).

Table 2. Tag numbers under winemaking conditions
 Mid-exponential phaseEarly-stationary phaseLate-stationary phase
  • 1

    Tags that correspond to a defined ORF in the yeast genome.

  • 2

    Tags that match intergenic regions within the yeast genome and do not correspond to an annotated ORF.

  • 3

    Tags that do not match to any location within the yeast genome.

Total tags488134834327
Unique tags 614 449 625
ORF tags1449 (73%)344 (77%)354 (57%)
NORF tags262 (10%)42 (9%)136 (22%)
Non-identified tags3103 (17%)63 (14%)135 (21%)

Upon correlation with the yeast genome, three different kinds of tags were found: (a) tags that corresponded to an annotated open reading frame (ORF tags); (b) tags that corresponded to intergenic regions (NORF tags); and (c) tags that did not match the published reference yeast genome (non-identified or NID tags). At both mid-exponential and early-stationary phases, about 10% and 15% of the total tags corresponded to NORF and NID tags, respectively (Table 2). At late-stationary phase NORF and NID tags increased to 22% of the total, i.e. 136 and 135 tags, respectively. Both kinds of tags showed differential expression at early- and late- stationary phases as compared to mid-exponential phase. Thirty-three NORF tags and 10 NID tags were shown to be statistically significantly upregulated by at least five times at late-stationary phase, reaching a maximum of 134 and 51 transcripts/cell, respectively. Since these two kinds of tags do not correspond to any known ORF, they are not available in the existing DNA microarrays and therefore have not been reported to be expressed under winemaking conditions. Hence, NORF and NID tags could represent potential novel genes that are required to be expressed during wine fermentation.

In the original annotation of yeast genes, ORFs encoding proteins with less than 100 amino acids were omitted from the annotation unless evidence for the presence of a gene had been found by other direct means (Goffeau et al., 1996). Currently, only 224 known genes (3.5% of the yeast genome) encode these small proteins (Oshiro et al., 2002). Since, in other sequenced organisms, genes encoding for small proteins correspond up to 10% of their genomes, Oshiro et al. (2002) suggested that there might be an additional 400 genes encoding small proteins in the yeast genome. SAGE allows detection and identification of expressed sequences originating from any part of the genome that makes this method very useful for identifying new genes (Saha et al., 2002).

In this study, we found 197 different NORF tags expressed during wine fermentation. Although 72 tags have been reported in other yeast studies using SAGE (Kal et al., 1999; Oshiro et al., 2002; Velculescu et al., 1997), 70% of the NORF tags obtained here, especially at late-stationary phase, have not been described previously. Similarly, we found 245 different NID tags expressed throughout the fermentation.

Grouping genes according to expression profiles and expression levels, we defined 13 distinct clusters; however, more than 99% of the observed tags were grouped by the first nine clusters (Figure 2). Tags upregulated at late-stationary phase were grouped into the first three clusters, representing 24.2% of total tags. These clusters exhibited different expression levels. Thus, cluster 1 showed an expression increase from medium level (20–60 transcripts/cell) to high level (60–100 transcript/cell) and cluster 2 an increase from low level (0–20 transcripts/cell) to medium level. Cluster 3 grouped low expressed tags that increased at late-stationary phase (Figure 2A). Tags downregulated at late-stationary phase were grouped in clusters 4–6, accounting for 8.3% of total tags (Figure 2B). In cluster 4, expression levels decreased from high to low levels, whereas in cluster 5 expression was reduced from medium to low levels. Cluster 6 grouped low expressed tags that decreased at late-stationary phase. Tags expressed at constant levels throughout the fermentation were grouped in clusters 7 and 8. Cluster 7 grouped tags expressed at medium level, whilst cluster 8 grouped those tags expressed at low level. These two clusters represented 61.1% of the total tags (Figure 2C). Tags expressed at low level at mid-exponential and late-stationary phases but at higher level at early-stationary phase were grouped in cluster 9, representing 6.1% of total tags (Figure 2C). Tags upregulated and downregulated at late-stationary phase showing other expression levels were grouped into clusters 10–13. These clusters represented 0.7% of the observed tags and each cluster grouped less than four tags (data not shown). The clustered patterns observed here could be used to find common promoters that regulate gene expression at different stages of fermentation and even at different levels. Therefore, it could be possible to ‘switch on’ and ‘switch off’ exogenous (homologous and heterologous) genes at specific stages during wine fermentation.

Figure 2.

Cluster analysis. The nine most representative clusters. Clusters were defined according to expression profile and expression levels of observed tags. (A) Clusters 1–3 grouped tags upregulated at late-stationary phase (96 h). (B) Clusters 4–6 represented tags downregulated at late-stationary phase. (C) Clusters 7 and 8 grouped tags expressed at constant levels, whereas cluster 9 grouped tags upregulated at early-stationary phase (48 h)

As the target genes of most yeast transcriptional regulators have been established (Lee et al., 2002), we used this information to determine whether genes grouped in the same cluster are regulated by the same transcription factors. In clusters 1 and 2, 21% of the genes had target promoter regions for the stress response transcription factors Hsf1p and Skn7p, whereas in cluster 3 these genes accounted for 7%. The transcription factors Rap1p, Fhl1p and Gcr1p, which bind to regulatory regions of genes encoding ribosomal proteins and glycolytic enzymes, were associated to 26% of the genes grouped in clusters 4 and 5. In cluster 6, 12% of the genes were related to Fhl1p. Thus, a significant proportion of co-regulated genes in both up- and downregulated clusters were associated with the same transcription factor. Other clusters were related to several transcriptional regulators and the relevance of specific promoter regions was less evident.

At mid-exponential and early-stationary phases, the most highly expressed genes corresponded to the glycolytic genes: aldolase, 3-phosphoglycerate kinase and glyceraldehyde 3-phosphate dehydrogenase I. At late-stationary phase, however, the most highly expressed genes corresponded to genes encoding proteins related to stress response, such as heat shock proteins, metallothioneins regulated by heat shock elements (Liu and Thiele, 1996) and chaperone proteins (Table 3). Among the top 30 most highly expressed genes, a NORF tag, which corresponds to an expressed sequence tag (EST) from chromosome 12, and a NID tag, were found to be expressed at greater than 50 transcripts/cell at late-stationary phase. Although the function of these putative genes remains to be clarified, their expression at the end of the wine fermentation could be relevant.

Table 3. Most highly expressed genes at late-stationary phase
GeneExponential (transcripts/cell)Early-stationary (transcripts/cell)Late-stationary (transcripts/cell)Description
  • 1

    Tag that matches an intergenic region within the yeast genome and does not correspond to an annotated ORF.

  • †, *, §

    Statistically significant difference in expression levels (p < 0.05): §comparing mid-exponential and early- stationary phases; comparing early- and late-stationary phases; *comparing mid-exponential and late-stationary phases.

HSP26128 98512*Heat shock protein 26
HSP82 33 30294*Heat shock protein 82
TEF1/2 61 59199*Translational elongation
PGK1224339§186*3-Phosphoglycerate kinase
FBA1506610162*Aldolase
CUP1-1/1-2 82 77161*Copper ion binding
HSP30 18 42134*Heat shock protein 30
NORF1 24 25134*EST from chromosome 12
TIP1 64 25§131*Cold and heat shock protein
SEC3  6 <4121*Cellular transport

Yeast metabolism

Many biosynthetic pathways were downregulated as fermentation progressed, confirming the results of Rossignol et al. (2003). Genes related to cell wall organization and biogenesis, protein biosynthesis, RNA processing and protein metabolism showed moderate to low expression levels (below 60 transcripts/cell) at the beginning of the fermentation. The expression of these genes decreased throughout the fermentation. The downregulation of biosynthetic genes is thought to allow the cell to save energy and to adapt to new conditions (Gasch et al., 2000). Genes related to nucleotide metabolism showed constant expression levels and were classified into cluster 8.

Sugars

Many glycolytic genes were downregulated at the end of fermentation and therefore classified into clusters 4–6 (Table 4). Other genes (HXK1 and PGI1) showed the same expression level throughout the fermentation and were classified in cluster 7. Genes encoding glycolytic enzymes exhibited differences in expression levels of up to 150 times at the same phase. Similarly, Velculescu et al. (1997) reported differences of up to 70 times for glycolysis-related transcripts for glucose-grown yeast cells. The phosphofructokinase gene was the lowest expressed gene in glycolysis, <4 transcripts/cell, indicating that expression of this gene is tightly regulated. Surprisingly, FBA1, which encodes the following enzymatic reaction in glycolysis, was the most expressed gene under the conditions used in this study (Table 4). Genes encoding enzymes of the TCA cycle, except for ACO1 (aconitase), MDH1 (malate dehydrogenase) and SDH4 (succinate dehydrogenase), showed very low expression levels throughout fermentation. These three genes were expressed at 30–40 transcripts/cell at early-stationary phase and then downregulated at late-stationary phase.

Table 4. Expression of glycolytic genes
GeneExponential (transcripts/cell)Early-stationary (transcripts/cell)Late-stationary (transcripts/cell)Description
  • †, *, §

    Statistically significant difference in expression levels (p < 0.05):§comparing mid-exponential and early-stationary phases; comparing early- and late-stationary phases; * comparing mid-exponential and late-stationary phases.

HXK1 24 12 13Hexokinase
PGI1 <4 12 13Phosphoglucoisomerase
PFK2 <4 <4 <4Phosphofructokinase
FBA1506610162*Aldolase
TDH1 98163§103Glyceraldehyde 3-phosphate dehydrogenase I
TDH2109 46§ <4*Glyceraldehyde 3-phosphate dehydrogenase II
PGK1224339§186Phosphoglycerate kinase
GPM1 49 64 17*Phosphoglycerate mutase
ENO2 46 30 <4*Enolase
CDC19 12 47§ 34Pyruvate kinase

Expression levels of hexose transporter genes (HXT) were determined during wine fermentation (Table 5). As observed for genes involved in glycolysis, expression levels of HXT genes decreased at late-stationary phase, suggesting a possible co-regulation of glycolysis and sugar transport. HXT2 showed low expression levels throughout fermentation. HXT3 was the most highly expressed HXT gene, showing a pattern consistent with earlier results (Luyten et al., 2002). Unexpectedly, HXT6/7, described as high-affinity carriers (Reifenberger et al., 1997), were expressed throughout fermentation. However, expression of HXT6, cloned from an oenological strain into a hxt1-7Δ mutant strain, could not restore growth on a medium with low glucose concentration, whereas HXT7-only cells were shown to grow only weakly on the same medium. In contrast, expression of HXT6 or HXT7 from a laboratory strain in the hxt1-7Δ mutant enabled the yeast to grow well on low glucose (Luyten et al., 2002). This result illustrates that, despite high sequence homology, hexose carriers from oenological and laboratory yeast strains have dissimilar behaviours and could play different roles during wine fermentation.

Table 5. Expression of HXT genes
GeneExponential phase (tran-scripts/cell)Early-stationary phase (transcripts/cell)Late-station-ary phase (transcripts/cell)
  • *

    §†Statistically significant difference in expression levels (p < 0.05): § comparing mid-exponential and early-stationary phases; comparing early- and late-stationary phases; *comparing mid-exponential and late-stationary phases.

HXT2<4 8<4
HXT33021<4*
HXT6/718 8<4*

In addition to the HXT genes, TDH1 and TDH2 encoding two glyceraldehyde-3-phosphate dehydrogenase isoenzymes were found to be expressed at different levels (Table 4). Similarly, isoenzymes of glycerol 3-phosphate dehydrogenase (encoded by GPD1 and GPD2), pyruvate decarboxylase (encoded by PDC1 and PDC5), alcohol dehydrogenase (encoded by ADH1/2 and ADH6) and aldehyde dehydrogenase (encoded by ALD2/3, ALD4 and ALD6) were also differentially expressed, depending on growth phase (data not shown).

Nitrogen and amino acids

GAP1, which encodes the general amino acid permease, showed a constant expression level throughout the fermentation. Genes encoding amino acid vacuolar transporters (AVT3, AVT4 and AVT6) were mainly expressed during mid-exponential phase and then downregulated. This expression pattern is consistent with the amount of assimilable nitrogen present in the medium at each growth phase. PUT1, which encodes proline oxidase, was expressed at high levels (about 89 transcripts/cell) during the first two phases and downregulated at late-stationary phase. However, proline was not removed from the must at any growth phase. Although proline is able to induce expression of PUT1 in anaerobic conditions, oxygen is essential to allow proline degradation (Wang and Brandriss, 1987). PUT2, encoding Δ1-pyrroline 5-carboxylate dehydrogenase, the next step of the proline utilization pathway, increased significantly at early-stationary phase but was undetectable later on.

Expression of many genes related to amino acid biosynthesis, including the transcriptional activator GCN4, was downregulated at late-stationary phase (cluster 5). However, GDH2 and GDH3, which encode two glutamate dehydrogenase isoenzymes, HOM3 encoding aspartate kinase and MET30 a regulator of amino acid metabolism, were upregulated at late-stationary phase.

Trehalose and glycogen

Genes related to trehalose biosynthesis did not change significantly, whereas genes involved in trehalose degradation (ATH1 and NTH2) exhibited different expression patterns. ATH1 encodes an acid trehalase, whereas NTH2 encodes a neutral trehalase. Expression of ATH1 increased throughout fermentation, whilst NTH2 expression increased in the early-stationary phase and then decreased. We also observed the downregulation at the end of the wine fermentation of GLC3, involved in glycogen biosynthesis, whereas GPH1, involved in glycogen degradation, was expressed at a low and constant level throughout fermentation.

Sterols

The yeast cell cannot produce ergosterol in the absence of oxygen. However, many genes encoding proteins involved in ergosterol biosynthesis were expressed at mid-exponential and early-stationary phases and downregulated at the end of fermentation. A similar expression pattern was observed by Rossignol et al. (2003).

Electron transport chain

Several genes related to the electron transport chain (QRC, COX, CYC and CYT genes), inactive during wine fermentation, were also expressed. We also found a differential expression of COX6 and COX13, encoding subunit VI and VIa, respectively, of cytochrome c oxidase. Whereas COX13 was downregulated at stationary phase (half of the expression found at mid-exponential phase), COX6 was upregulated at late-stationary phase. Although the expression of these genes under anoxic conditions could seem contradictory, genes encoding different subunits of cytochrome c oxidase (COX1, COX2, COX4, COX5A, COX6, COX7, COX8 and COX9) were shown to be expressed in the absence of oxygen (Dagsgaard et al., 2001). Moreover, most cytochrome c oxidase subunits, including subunits I and II, which comprise the catalytic core, are present in anoxic cells and they are localized in the promitochondria (Dagsgaard et al., 2001). As no cytochrome c oxidase activity has been found in anaerobiosis (Rogers and Stewart, 1973), the function of this protein complex, if any, is still unknown.

Stress response

All environmental changes that trigger an adaptive response are denominated as stress. Genes regulated by stress or that respond in a similar way to many stresses have been referred to as environmental stress response (ESR) or common environmental response (CER) genes, respectively (Rossignol et al., 2003). The proteins encoded by the genes induced by stress are involved in many cellular processes. Stress-responsive heat shock genes (HSP26, HSP42, HSP60, HSP78, HSP82, SSE2), which encode proteins related to protein folding, were strongly induced (Table 6). Although these genes showed different expression levels, all were induced at late-stationary phase. HSP26 was found to be the most highly expressed gene of the genome at late-stationary phase (Table 3). HSP104 and HSP12, the latter previously described as glucose-repressed in a laboratory strain of S. cerevisiae (de Groot et al., 2000), were expressed at constant levels throughout the fermentation (Table 6).

Table 6. Expression of genes related to stress response
GeneExponential (transcripts/cell)Early-stationary (transcripts/cell)Late-stationary (transcripts/cell)Description
  • †, *

    §Statistically significant difference in expression levels (p < 0.05): § comparing mid-exponential and early-stationary phases; comparing early- and late-stationary phases; *comparing mid-exponential and late-stationary phases.

HSP12  9<4  6Protein folding
HSP2612898512*Protein folding
HSP30 1842134*H+ homeostasis
HSP42  6 8 34*Protein folding
HSP60  9 4 55*Protein folding
HSP78 <4<4 31*Protein folding
HSP82 3330294*Protein folding
HSP104 <417 <4Protein folding
NCE102 1221 31*Function unknown
PMA1/2 2125 13Membrane ATPases
PMP2 3342 83*H+ homeostasis (putative)
PST1 <4 8 <4Function unknown
PWP1 <4 4 10Function unknown
SIW14  912 <4Tyrosine phosphatase activity
SPI1 18 8 91*Function unknown
SSC1 2417 16Protein folding
SSE1  6 8 20Protein folding
SSE2 <4 4 24*Protein folding (putative)
TIR1 18<4 <4Constituent of cell wall
YGP1 3030  6*Function unknown
YRO2 2430 10Function unknown

Genes encoding proteins involved in proton homeostasis, HSP30 and PMP2 (coding for a negative and a positive regulator of the plasma membrane H+-ATPase, respectively) were upregulated at the end of the fermentation (Table 6). Simultaneous induction of positive and negative regulators of plasma membrane H+-ATPase has been described as consistent with the fine-tuning of H+ extrusion activity (Alexandre et al., 2001). The plasma membrane H+-ATPase, encoded by PMA1/2, was expressed from the beginning of the fermentation at similar levels (Table 6).

Genes which encode proteins with unknown molecular functions and that respond to stress, according to SGD (NCE102, SPI1), were also upregulated at late-stationary phase (Table 6). Some genes related to stress response (PST1, PWP1, SIW14, SSC1, SSE1, TIR1, YRO2) showed similar expression levels throughout the fermentation (Table 6). Other genes, e.g. YGP1, HSP31 and HSP32/33, were downregulated at late-stationary phase, indicating a response to an early stress no longer present at the end of fermentation.

Correlation between metabolic fluxes and gene expression

In a recent study (Varela et al., 2004), we reported the redistribution of metabolic fluxes in S. cerevisiae EC1118 during wine fermentation, using the same experimental conditions described here. Although the expression of some glycolytic genes increased, we found that metabolic fluxes throughout glycolysis diminished. Genes involved in the TCA cycle were slightly upregulated, while fluxes within the cycle decreased. Entry gates to glycolysis, considering sugar transport and the hexokinase reaction, and to the TCA cycle, considering the reactions from pyruvate to isocitrate, showed the best correlations between gene expression and metabolic fluxes (r2 = 0.928 and r2 = 0.848, respectively) (Figure 3).

Figure 3.

Correlation between gene expression and metabolic fluxes at different yeast growth phases. Entry gates to glycolysis (black squares) and the TCA cycle (grey squares). The stoichiometric model included glycolysis, pentose phosphate pathway, the TCA cycle, amino acid metabolism and macromolecule biosynthesis pathways

Discussion

Stress conditions throughout the wine fermentation process, such as nutrient limitation, starvation, temperature variations and ethanol toxicity, affect the yeast's metabolism, eliciting a metabolic response. This response, which enables the cell to adapt to the new environmental conditions, implies a metabolic reorganization in which a great number of genes are up- or downregulated. In this work, SAGE was used to quantify gene expression profiles in the commercial strain S. cerevisiae EC1118 under winemaking conditions. We detected the expression of genes that have not been reported in other studies using SAGE (Kal et al., 1999; Velculescu et al., 1997), suggesting that winemaking conditions impose a substantially different environment for yeast growth.

Since the total number of tags obtained in this study was not high enough, we were unable to detect very low-abundance mRNA molecules (<2 copies/cell). The total number of tags also affects the confidence level for detecting low-abundance transcripts. However, analysis of the relative distribution of gene expression levels revealed that low-abundance mRNA classes were highly similar to those described previously using a greater number of tags obtained from a different yeast strain (Velculescu et al., 1997). Therefore, we considered the data of low-abundance transcripts reliable.

SAGE enabled the identification of novel transcripts (NORFs and NIDs) within the yeast transcriptome. Since we compared the transcriptome of a wine yeast with the published reference yeast genome, obtained from a laboratory strain of S. cerevisiae, several factors could explain the presence of non-identified tags, e.g. differences in the sequence of known genes between both yeasts and novel genes present only in the wine yeast. Indeed, a high frequency of polymorphisms has been described in S. cerevisiae (Oefner, 2002). Further research is necessary to assess the presence of new genes in S. cerevisiae EC1118.

Many of these novel transcripts were expressed at high levels and differentially regulated depending on the growth phase. Particularly at late-stationary phase, where ethanol concentration was 12% (v/v) and nitrogen had been already depleted for 2 days, we observed a high increase in the number of new transcripts. This harsh environment, very different to situations normally studied under laboratory conditions, probably triggers the expression of genes that encode critical functions for the yeast cell and that had not been discovered until now. Oshiro et al. (2002) investigated the expression and translation of several NORFs previously identified by SAGE under laboratory conditions. After growing yeasts in different growth conditions, which included treatments with hydroxyurea, nocodazole, methyl methane sulphonate, ultraviolet light, along with heat and cold shocks, they proposed the addition of 62 new genes to the yeast genome. This illustrates that in order to increase the chance of observing expression of new genes, unusual growth conditions are often necessary.

Our results have indicated that ‘new’ genes could have a significant role under winemaking conditions. Although expression patterns do not present conclusive evidence of gene function, they indicate directions for further research. Indeed, a strong functional connection between transcription factors and metabolic pathways has been described, suggesting that uncharacterized ORFs can be assigned a functional link based on co-expression with specific metabolic pathways (Ihmels et al., 2004). Characterization of these novel genes could lead to further understanding of the yeast physiology and the wine fermentation process.

Many genes involved in pathways and reactions that require oxygen for their function, such as the ERG, PUT, COX and OLE genes, were expressed in absence of oxygen. Post-transcriptional mechanisms are probably responsible for the regulation of the corresponding proteins, which would enable the cell to rapidly respond when oxygen is again available. An interesting case is the differential gene expression of subunits VI and VIa of cytochrome c oxidase. Similarly to subunits Va and Vb, which change cytochrome c oxidase activity in response to oxygen availability (Kwast et al., 1998), subunit VIa affects the activity of cytochrome c oxidase in response to ionic strength and ATP concentration (Beauvoit et al., 1999; Taanman and Capaldi, 1993). As there is no activity of cytochrome c oxidase under anoxic conditions (Rogers and Stewart, 1973), differential regulation of cytochrome c oxidase subunits with different properties by the yeast cell remains to be studied.

Within the yeast transcriptome, we found different transcript levels for genes involved in the same metabolic pathway (glycolysis, TCA cycle, ergosterol biosynthesis, pentose phosphate pathway, etc.) at the same growth phase. In glycolysis, differences in transcript levels were up to 150 times at early-stationary phase between two genes, which code for sequential steps within the pathway. In addition, protein levels and protein activities for glycolytic genes have been also shown to differ greatly at the same growth phase (Fraenkel, 2003). However, despite differences in transcript and protein levels, carbon flux through glycolysis is highly coordinated under winemaking conditions, as recently showed by Varela et al. (2004). Hence, these differences show that, in order to present a fully functional metabolic pathway, the yeast cell requires different mRNA levels of the genes involved in that pathway. Probably, mRNA stability or post-transcriptional mechanisms are responsible for dissimilar transcript levels in the same pathway.

New genes introduced and overexpressed in S. cerevisiae are normally under the control of the constitutive promoter pPGK1 (Lilly et al., 2000; Verstrepen et al., 2003). However, expression of the FBA1 gene is two-fold higher than the PGK1 gene during the first two stages of fermentation (Table 4). Although this difference could be strain-dependent, the FBA1 promoter might be used to overexpress genes at greater levels and thus make the phenotype of the new strain stronger. In addition, the use of promoters specific for different growth stages could be advantageous to obtain a differential expression of new genes in wine yeasts.

Many isoenzymes involved in pathways of the central metabolism were expressed differentially, depending on the growth phase. In addition to different physiological roles, the presence of isoenzymes in the S. cerevisiae genome probably enables the yeast cell both to manage efficiently the pool of co-factors, such as NADH and NADPH, and to control the gene expression by different physiological triggers rather than different pathways in order to respond competently to new environmental conditions (Ansell et al., 1997). Alternatively, isoenzymes could be dedicated to distinct processes, using a common reaction, and thus reduce cross-talk and unwanted interactions between separate metabolic pathways (Ihmels et al., 2004).

We observed different patterns of expression for genes related to stress response, which illustrates the temporal impact of the various stresses on the yeast cell during wine fermentation. Immediately after inoculation, yeast faces hyperosmotic conditions that trigger a fast and transient osmotic stress response (Perez-Torrado et al., 2002). As the first sample in our experiments was probably taken after yeast adaptation, we did not observe any response to the high osmotic pressure present in the must. Although other individual responses are more difficult to evaluate because many genes are expressed in response to different environmental stimuli (Bauer and Pretorius, 2000), our results show that several stress-related genes are expressed from mid-exponential phase onwards. Some of these genes, many of them with unknown functions, were upregulated by growth arrest, but the highest transcript levels were found at late-stationary phase, indicating that yeasts are subjected to extreme conditions during this phase. As mentioned before, at late-stationary phase, we observed the highest number of NORFs and NIDs, suggesting that wine fermentation conditions could be appropriate for studying the function of both genes with unknown functions and novel genes. Probably the high amount of ethanol present in the medium at this phase is responsible for these physiological changes. Although more research is necessary, the different responses of genes related to stress response could be used as a gene reporter system to assess the presence of early stresses that could affect a ‘normal’ fermentation profile and impair the quality of the final product. Indeed, there is an inverse correlation between stress resistance and stuck fermentations (Ivorra et al., 1999).

Comparing metabolic fluxes and gene expression levels, many discrepancies were found, especially for genes involved in glycolysis and in the TCA cycle. Such divergences have already been reported in other integrative approaches with S. cerevisiae and other microorganisms (Bro et al., 2003; Oh and Liao, 2000). Nevertheless, entry gates to glycolysis and to the TCA cycle showed the best correlations between metabolic fluxes and transcript levels. Detailed analysis of co-regulation in central metabolic pathways has shown that the co-expressed enzymes are often arranged in a linear order, corresponding to a metabolic flow that is directed in a particular direction (Ihmels et al., 2004). Although fluxes through central metabolic pathways in S. cerevisiae are not primarily controlled at the transcriptional level (Daran-Lapujade et al., 2004), pathway entry gates probably control carbon flux throughout a specific pathway and could be regulated at both transcriptional and metabolic levels. This would in turn enable the yeast cell to optimize the flux and thus maximize metabolic efficiency under different conditions (Stelling et al., 2002). A good correlation between metabolic fluxes and gene expression was also found for pathway entry gates in Corynebacterium glutamicum (Kromer et al., 2004), suggesting that this regulation mechanism is probably common to microorganisms. Further research is necessary to evaluate the contribution of post-transcriptional mechanisms, post-translational modification, and regulation by intracellular concentrations of substrates, products, and effectors to the regulation of fluxes of central metabolic pathways.

Acknowledgements

We would like to thank Professor Sakkie Pretorius and Dr Florian Bauer for their support in early stages of this work, Dr Jan Ruijter for kindly providing the SAGEstat software, and Ana Maria Molina and Alicia Zuñiga for their contribution to this work. This work was supported by Fondo Nacional para el Desarollo Científico y Tecnológico de Chile (FONDECYT), Grants 2010087 and 1010959. Cristian Varela was supported by a doctoral fellowship from Consejo Nacional de Investigación Científica y Tecnológica de Chile (CONICYT).

Ancillary