SEARCH

SEARCH BY CITATION

Keywords:

  • alcoholic fermentation;
  • gene expression;
  • proteome;
  • wine;
  • yeast

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Aims:  Although wine yeast gene expression has been thoroughly investigated only few data are available on the evolution the proteome during alcoholic fermentation. This work aimed at specifying the change in proteome during fermentation and to assess its connection with transcriptome.

Methods and Results:  The proteome of a wine yeast was monitored by 2-D gel electrophoresis throughout alcoholic fermentation. Proteome was analysed in exponential growth and stationary phase. Among 744 spots, detected we observed significant changes in abundance with 89 spots displaying an increase in intensity and 124 a decrease. We identified 59 proteins among the most regulated and/or the most expressed. Glycolysis and ethanol production, amino acid and sulfur metabolism were the most represented functional categories. We found only a weak correlation between changes in mRNA and protein abundance, which is strongly dependent on the functional category.

Conclusions:  There are substantial changes in protein abundance during alcoholic fermentation, but they are not directly associated with changes at transcript level suggesting that mRNA is selectively processed and/or translated in stationary phase.

Significance and Impact of the Study:  These data show that proteome is a relevant level of analysis to gain insight into wine yeast adaptation to alcoholic fermentation.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The budding yeast Saccharomyces cerevisiae has been extensively studied as a model organism. It is also widely used in biotechnology and food industries and is especially used in fermentation industries. Selected industrial strains are now used predominantly in those areas and are subject to increasingly intensive research.

Alcoholic fermentation produces specific and stressful environmental conditions, requiring particular fermentation properties present in enological yeasts, but not in laboratory strains. The distinct physiological properties of enological yeasts specific to wine making imply that results obtained with laboratory strains and/or under laboratory conditions are not systematically transposable to enological yeast.

The popular ‘omics’ technologies have been largely set up and used in the model organism, Saccharomyces cerevisiae, and these approaches have been extended to industrial strains. Transcript regulation in wine yeast strains has been investigated intensively by microarray and SAGE (serial analysis of gene expression) in industrial-like conditions not only to decipher transcript technological properties (Backhus et al. 2001; Erasmus et al. 2003; Marks et al. 2003; Rossignol et al. 2003; Varela et al. 2005; Beltran et al. 2006; Rossignol et al. 2006; Mendes-Ferreira et al. 2007a,b; Novo et al. 2007; Jimenez-Marti and del Olmo 2008; Marks et al. 2008), but also as a model for evolution (Cavalieri et al. 2000; Infante et al. 2003; Townsend et al. 2003). Analysis of mRNA transcription levels alone is not sufficient to describe a biological system as mRNA is not the final product and does not determine translational and posttranslational regulation mechanisms.

Enological yeast mRNA expression is well defined for a wide array of industrial conditions, but few Enological studies have been conducted at the proteome level. Only three proteomic studies have been performed on enological yeast: (i) one comparing 2 enological strains presenting differences in fermentation behaviour (Zuzuarregui et al. 2006); (ii) one following protein production under conditions not truly related to those present during industrial wine fermentation (Trabalzini et al. 2003) and (iii) at the time of writing, Salvado et al. (2008) described the protein profile during the early lag phase after rehydratation of active dried yeast under low temperature condition, relevant to some technological practices. For systems biology, transcriptome and proteome regulation can be integrated. Comparative analysis of transcript and protein levels has been performed mainly with laboratory yeast strains (Griffin et al. 2002; Washburn et al. 2003; Daran-Lapujade et al. 2004; Kolkman et al. 2006; Zuzuarregui et al. 2006; de Groot et al. 2007). Zuzuarregui et al. (2006) compared transcript and protein levels between two oenological yeasts, but no studies have been performed on a large-scale fermentation process.

We previously used microarrays to monitor an industrial wine yeast transcriptome throughout alcoholic fermentation in enological conditions (Rossignol et al. 2003). In this study, we analysed the proteome by two-dimensional gel electrophoresis. We analysed the same strain under identical conditions to those used in our previous study (Rossignol et al. 2003), including similar sampling time points to compare directly the transcriptome and proteome of a wine yeast strain in oenological fermentation.

We identified 59 proteins and evaluated transcript and protein regulation, as well as their correlation. Overall, we provide a two-dimensional protein map of an industrial yeast in enological-like conditions.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Yeast strain, media and culture conditions

The industrial S. cerevisiae oenological strain EC1118 from Lallemand S.A. (Toulouse, France) was used in all fermentation experiments. Rehydration conditions of active dried yeasts were as previously described (Rossignol et al. 2003). Fermentation experiments were carried out with the well defined synthetic must MS300, miming a standard natural must described previously (Rossignol et al. 2003). This medium contains 200 g l−1 of glucose and 300 mg l−1 of assimilable nitrogen provided by a mixture of ammonia and amino acids. Fermentations were performed at 24°C, in 1·2 l fermenters equipped with locks to maintain anaerobiosis and with constant stirring. Fermentations were monitored with an on-line monitoring system based on an automatic weighing of the fermentor every 20 min, to determine weight loss and the rate of CO2 production. This method gives a highly accurate measure of fermentation progress. Position of sampling for protein analysis was chosen according to the transcript profile during fermentation (Rossignol et al. 2003). These experiments and two-dimensional gel electrophoresis were carried out in the same period as the transcript profiling experiments previously published (Rossignol et al. 2003).

Protein extraction and two-dimensional gel electrophoresis

Protein extraction and two-dimensional gel electrophoresis were carried out as previously described in Boucherie et al. (Boucherie et al. 1995) with the following modifications: the first dimension gel contained 3·6% acrylamide/bisacrylamide (Pharmacia Biotech), 9·5 mol l−1 urea, 3·57% CHAPS, 4% carrier ampholytes (1/5 Pharmalytes 5–6; 2/5 Pharmalytes 5–8; 2/5 Pharmalytes 3–10), 0·3% TEMED (Boehringer Mannheim) and 0·3% ammonium persulfate (Boehringer Mannheim). The second dimension gels were run at 5 W for 15 h. After migration, gels were fixed in fixation buffer (50% ethanol; 7·5% acetic acid) for 1 h and rinsed with distilled water, several times. The method allows the separation of proteins with a pI between 3·8 and 6·8 and a molecular weight between 15 and 180 kDa. For image acquisition, proteins were labelled using Sypro Ruby (Molecular Probes). Briefly, gels were incubated in Sypro Ruby overnight, protected from light and washed for 30 min three times in a wash solution (10% ethanol, 7% acetic acid). Gels were scanned using an FX Molecular Imager scanner using 100 μm resolution (Bio-Rad). For MALDI-TOF protein identification, gels were stained with Coomassie Blue. Briefly, gels were stained with a solution of 4/5 Coomassie blue 1/5 methanol for 2 h and rinsed in a solution containing 10% acetic acid and 25% methanol.

Quantitative gel analysis

Quantitative gel comparisons were performed with Proteomweaver 1.0 software (Definiens, München, Germany). We ran at least three gels for each stage and, for comparison, we selected the two best quality gels for each stage, showing the best resolution and intensity scale. Spot detection is mainly automatic and the software allows automatic intensity normalization between gels. The integrated intensity ratios were calculated and analysed and significance was evaluated using t-test. For analysing general regulation, ratios >2 with a P-value <0·05 were considered significant. Spots with relative intensities <0·05 in the two conditions were not evaluated for ratios.

Protein identification

Proteins were identified by mass spectrometry as previously described (Joubert et al. 2001). Briefly, selected protein spots were cut out from the gels and trypsin-digested. Proteins were identified by MALDI-TOF and peptide masses obtained were analysed with MS-Fit software (http://prospector.ucsf.edu/mshome.htm) and mascot software (http://www.matrixscience.com/search_form_select.html) for protein identification.

Alternatively, proteins were identified by gel matching with a two-dimensional gel electrophoresis map of a lager beer yeast grown in industrial conditions (Kobi et al. 2004). This two-dimensional gel reference map was produced with the same protocole and apparatus as two-dimensional gels presented in this work and allows obvious gel matching.

Identification of functions and processes over- or under-represented

We use the GoToolBox web application (Martin et al. 2004) to identify gene ontology (GO) processes that are over-represented in the list of identified proteins. The full gene set from Saccharomyces Genome Database (SGD; http://www.yeastgenome.org) was used as a reference. Some of the categories have been merged in the Table 1 for clarity.

Table 1.   List of the proteins identified, ordered by functional categories Thumbnail image of Thumbnail image of

Transcriptome and proteome comparison

In the case of multiple-spot proteins identified, intensity values were summed (Table 1). This usually leads to a small increase in the intensity value of the main spot due to addition of the minor spot values. Thus, no P values were taken into account in transcript and protein regulation comparisons as multiple spots could not generate a single P value. The nonparametric spearman’s rank correlation coefficient method (Rs) used to estimate the correlation between transcriptome and proteome was calculated with Winstat® software (Fitch Software, Krozingen, Germany).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The proteome of an industrial yeast in oenological conditions

The fermentation conditions as well as the industrial strain (EC1118) used in this study are identical to those reported in (Rossignol et al. 2003) for transcriptome analysis and the sampling stages for proteome analysis have also been chosen according to those previously described, making these data directly comparable. Culture samples from two stages of yeast growth showing strong differences in transcript regulation were subjected to two-dimensional gel electrophoresis. The first sampling stage (reference stage) is located at the start of the exponential phase (Rossignol et al. 2003). The second sampling stage (stationary stage) consisted of cells in the stationary phase, which face 6% ethanol (Rossignol et al. 2003). We also sampled a third intermediate stage (Rossignol et al. 2003), which is located at the end of the growth phase. Preliminary results indicated few differences between this later stage and the stationary stage (data not shown). Thus, we focused our analysis on the reference and stationary stage. Figure 1a shows the two-dimensional gel proteome map at the start of the growth (reference stage) and Fig. 1b shows the two-dimensional gel proteome map at the stationary stage. We detected up to 744 spots, present in the two stages analysed. Our analysis detected 89 spots for which the intensity was at least twofold stronger in stationary phase than in reference stage and 124 spots for which it was at least twofold weaker. Typical examples of regulated spots are shown in Fig. 1c. Protein identification was performed by selecting the spots with the highest or most variable abundance at the stages analysed. Two methods were used for identification: mass spectrometry and gel matching. Interestingly, the two-dimensional gel map obtained was much more similar to industrial brewery yeast gel map (Kobi et al. 2004), and to a lower extent to distilling strain gel map (Hansen et al. 2006), than those from laboratory strains (Boucherie et al. 1995) in term of spot intensity and distribution. Eighty-four spots were identified among the most intense and the most regulated spots. Thirty-four spots were identified by mass spectrometry and 50 by gel matching. Among the spots identified by mass spectrometry, 10 were first identified by gel matching and confirmed by mass spectrometry (Hxk2, Eno1, Pdc1, Adh1, Tps1, Asn1, Met17, Met6, Yef3, Vma2). Overall, spots identified correspond to 59 proteins, as several spots corresponded to the same protein. The proteins were identified and listed by function (Table 1). The functional category most represented is glycolysis and ethanol production, with 10 proteins identified corresponding to 21 spots. Enolase 2 and alcohol dehydrogrenase 1 were the two most expressed proteins in our strain and under our experimental conditions. These proteins represent the major part of the proteins detectable at the reference and stationary stages, with 28% and 36% of the overall spot intensity, respectively. Most of these proteins are induced between the reference and stationary stage (Hxk2, Glk1, Fba1, Tpi1, Eno1, Adh1) and only Pdc1 is repressed. The proteins involved in glycerol production (Rhr2, Hor2, Gpd1) and other carbohydrate metabolic pathways (Tps1, Ino1, Zwf1) all show a similar induction pattern.

image

Figure 1.  Two-dimensional gels of (a) the reference stage and (b) the stationary stage. Proteins were visualized with Sypro Ruby. Proteins identified are annotated on the gels. Underlined protein names correspond to proteins identified by mass spectrometry. (c) magnified regions of gel containing examples of spots of some of the most regulated proteins between reference and stationary stages.

Download figure to PowerPoint

The amino acid metabolism category had the highest number of proteins identified (12 proteins). Most of these proteins were more expressed in the stationary stage than in the reference stage (Leu2, His4, Asn1, Cys3, Gdh1, Met17, Sam2) and a few were repressed (Lys20, Ilv5). Met6, Met17 and Sam2, involved in sulphate assimilation, were highly expressed under our fermentation conditions. They represent 7% of the total number of spots quantified on the gels in the stationary stage. Quantitatively, this is the second most important metabolic pathway after glycolysis. This metabolic pathway is important in wine yeast because its function determines the production of H2S, which is an undesirable by-product given its negative sensorial impact. Interestingly, proteins of the sulfur pathway are among the most expressed. They are involved in steps located downstream of H2S production and are thought to be potentially critical for H2S release. A high activity of these enzymes may be one condition enabling yeast to produce only low amounts of H2S, thus leading to its selection. All of the proteins involved in protein synthesis identified are down-regulated in the stationary phase during the fermentation, whereas Prb1, a protein which participated in protein degradation, is induced. This correlates with the decrease in protein synthesis observed at this stage of fermentation (Salmon 1989) and with the decrease of mRNA abundance of ribosomal proteins previously described (Rossignol et al. 2003). At the stationary stage, assimilable nitrogen is depleted from the medium (Rossignol et al. 2003) and cells probably undergo amino acid recycling. Moreover, induction of Prb1p during grain fermentation in industrial distilling fermentors has already been described (Hansen et al. 2006). This is also consistent with the idea that most protein degradation during starvation depends on vacuolar proteases (Egner et al. 1993).

Several proteins involved in stress response have been identified and show various types of regulation. Most are repressed (Ssa2, Sti1, Sse1, Hsp78, Tom70) and two are induced: Hsp26 (heat shock protein) and Ahp1 (thioredoxin peroxidase). Most chaperones involved in protein folding (cytoplasmic or mitochondrial) are down-regulated, probably in relation to the decrease in protein synthesis, whereas heat-shock proteins with specific functions are induced. Trabalzini et al. (2003) reported a similar pattern with an enological yeast in YPD modified medium at a later time point. Several Hsp proteins were down-regulated, whereas Hsp26 and oxidative stress proteins were induced. Hsp26 was more highly expressed in a strain with a good fermentation profile than in a strain leading to sluggish fermentation at the entry into stationary phase in industrial conditions (Zuzuarregui et al. 2006). The authors have suggested that this protein is important for adaptation to the stationary phase in alcoholic fermentation. Authors ended with the same conclusion for the product of YPR127w (Zuzuarregui et al. 2006), which is also induced in the present work (Table 1).

Correlation with transcriptome regulation

The choice of the two-dimensional gel sample times was driven by the microarray result already obtained to get comparable data. We analysed regulation ratios between reference and stationary stage and relative intensity values at the stationary stage for transcripts and proteins of the 59 proteins identified in this study (Table 1). The ratios of proteins and RNA expression between the stationary stage and the reference stage show a weak but positive correlation (Rs = 0·385; P-value = 0·001), as illustrated in Fig. 2.

image

Figure 2.  RNA expression ratios vs protein expression ratios plot. Values are expressed in log2.

Download figure to PowerPoint

Classifying proteins by functional category reveals a pattern for regulation correlation. For example, the functional category of protein synthesis has the best correlation between transcript and protein regulation. Expression of all the proteins identified and their respective transcripts decreased between the reference stage and the stationary stage. Similarly, proteins involved in protein degradation simultaneously to their respective transcripts.

Proteins and transcripts in the glycolysis category show no correlation to opposite correlation between the reference and stationary stages. Most of the proteins identified are induced, suggesting that regulation takes place at the translational level for this pathway. The amino acid biosynthesis category presents a strong opposite correlation between mRNA and protein regulations for most of the proteins identified (7 on 12), and only four show clear co-regulation (Lys20, Ilv5, Aro9 and Met6).

We have also attempted to correlated abundance data, although two-colour microarrays are not optimal for quantitative measurements of mRNA abundance. However, we observed a positive and significant correlation between the abundance of transcripts and the abundance of the corresponding proteins in stationary phase (Rs = 0·647; P-value = 1·9 E-8) as well as at the reference stage.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We characterized the proteome of a wine yeast strain in enological conditions. The strain EC118 was chosen in a previous work (Rossignol et al. 2003) because of its efficient fermentation properties and because it is a widely used strain. We carried out proteome map comparisons with similar studies on industrial strains; our map matches more specifically the industrial yeast proteome map generated in industrial conditions (Kobi et al. 2004; Hansen et al. 2006) than that generated under laboratory conditions (Trabalzini et al. 2003; Brejning et al. 2005). This underlines a specific profile of industrial fermentation, with a strong abundance of glycolysis and ethanol synthesis proteins. The over-representation of proteins of this pathway is greater than that obtained in laboratory conditions (Fraenkel 2003). Gene transcript levels for this metabolic pathway are also among the highest. Even if transcript and protein regulation are not clearly correlated, this metabolic pathway is the most or one of the most expressed at the transcript and protein level. Indeed, high glycolytic enzyme levels are considered critical for wine yeast fermentation capacity.

Correlation between transcript and protein abundance has been investigated in yeast over the last few years. Most studies report insufficient correlation (Gygi et al. 1999; de Nobel et al. 2001; Daran-Lapujade et al. 2004; Brejning et al. 2005; de Groot et al. 2007), whereas others (Futcher et al. 1999; Ghaemmaghami et al. 2003) report a good correlation. By merging several sources of mRNA and protein quantification data, Greenbaum et al. (2003) showed that transcript and protein levels were more highly correlated for some localization or functional categories. We also tempted to correlate protein and mRNA abundance, even if microarrays are not optimal quantitative tools. We did observe a relatively good correlation, but we could not determine specific correlation for functional or localization categories.

In contrast, our data displayed a weak but positive correlation for regulation of expression of transcripts and proteins. Correlation of regulation of expression has been previously investigated with affymetrix and Mudpit technologies by Washburn et al. (2003). These studies report an Rs of 0·45 and spotted good correlation for several amino acid pathways. Similar analyses using glass microarray and ICAT technology (Griffin et al. 2002) reported lower but still positive correlation (Rs = 0·21). Indeed, some functional categories are reported to display correlation for regulation of expression. Even with a relatively low number of proteins identified, the correlation between mRNA and protein regulation also appears to depend on the functional category in our data. This highlights the importance of deciphering correlation data preferentially at the pathway or functional category level.

We cannot exclude that some divergence in regulation is coming from technical bias. Among others, several spots are detected for the same protein (posttranslational modifications, proteolysis, etc), which can affect the quantification; also, only a limited number of proteins were studied due to 2-D gel technical limitations. Compiling results obtained with different technologies to reduce bias led to a good correlation of the abundance level (Rs = 0·85) (Lu et al. 2007). Moreover, mis-identification by gel matching may have occurred. But using a reference map produced with the exact same method, as in our case, strongly increases the sensitivity of the gel matching method, and all spots identified by both methods in this study corroborate.

Our data are consistent with a selective translation of the mRNA as the cells enter into stationary phase. Under these conditions, this selection may be critical for the yeast, as it must reorient its protein synthesis with restricted nitrogen availability. In the stationary phase, the cells are starved for nitrogen and the protein synthesis capacity is low (Salmon 1989; Rossignol et al. 2003). The glycolytic pathway and stress protection are clearly two priorities for yeast protein synthesis. As wine yeast are selected on the basis of their ability to sustain a high fermentation activity after long periods of starvation, the high levels of glycolytic enzymes sustained under these conditions may be a specific feature of wine yeasts. Elsewhere, it is notable that the expression profile of glycolytic proteins is opposite to the yeast fermentation rate which decreases during the same period. This suggests that glycolytic enzymes are not likely to be the rate-limiting step of fermentation.

A selection of mRNA may operate in the conditions studied. Izawa et al. (2005) have recently shown that there was an accumulation of bulk polyA+ mRNA in the nucleus in parallel to the fermentation process, and obviously in relation to ethanol stress. There might be a selection for mRNA export at this level, which could explain differences in correlation for a given gene-protein couple depending on the conditions. This underlines the necessity to perform relevant experiments under appropriate conditions. Similar experiments in industrial fermentors with natural musts may probably produce slight differences, but the high variability and the undefined composition of natural must would prevent these results from being used as a reference.

The analysis of these two levels of regulation, transcript and protein, is complementary and will be helpful to obtain a better overview of the mechanisms involved in any change in environmental conditions. It will also help of decipher the complex feedback regulations that link them, and to determine where the most obvious regulation takes place for the proteins identified. In addition to the molecular-based information on enological yeast during fermentation provided by this study, the two-dimensional protein map reported here could be used as a reference for future studies and could help of define specific fermentation traits at the protein level.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We would like to thank the Dr Strub team in the Pr Van Dorsselaer lab for the trypsin Digestions and MALDI-TOF experiments.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References