Functional analysis to identify genes in wine yeast adaptation to low-temperature fermentation


  • Z. Salvadó,

    1.  Biotecnologia Enològica, Departament de Bioquímica i Biotecnologia, Facultat de Enologia, Universitat Rovira i Virgili, Tarragona, Spain
    Search for more papers by this author
  • R. Chiva,

    1.  Departamento de Biotecnología de los alimentos, Instituto de Agroquímica y Tecnología de Alimentos (CSIC), Valencia, Spain
    Search for more papers by this author
  • N. Rozès,

    1.  Biotecnologia Enològica, Departament de Bioquímica i Biotecnologia, Facultat de Enologia, Universitat Rovira i Virgili, Tarragona, Spain
    Search for more papers by this author
  • R. Cordero-Otero,

    1.  Biotecnologia Enològica, Departament de Bioquímica i Biotecnologia, Facultat de Enologia, Universitat Rovira i Virgili, Tarragona, Spain
    Search for more papers by this author
  • J.M. Guillamón

    1.  Biotecnologia Enològica, Departament de Bioquímica i Biotecnologia, Facultat de Enologia, Universitat Rovira i Virgili, Tarragona, Spain
    2.  Departamento de Biotecnología de los alimentos, Instituto de Agroquímica y Tecnología de Alimentos (CSIC), Valencia, Spain
    Search for more papers by this author

Jose Manuel Guillamón, Departamento de Biotecnología de los alimentos, Instituto de Agroquímica y Tecnología de Alimentos (CSIC), Avda. Agustín Escardino, 7, E-46980 Paterna, Valencia, Spain.


Aims:  To identify genes and proteins involved in adaptation to low-temperature fermentations in a commercial wine yeast.

Methods and Results:  Nine proteins were identified as representing the most significant changes in proteomic maps during the first 24 h of fermentation at low (13°C) and standard temperature (25°C). These proteins were mainly involved in stress response and in glucose and nitrogen metabolism. Transcription analysis of the genes encoding most of these proteins within the same time frame of wine fermentation presented a good correlation with proteomic data. Knockout and overexpressing strains of some of these genes were constructed and tested to evaluate their ability to start the fermentation process. The strain overexpressing ILV5 improved its fermentation activity in the first hours of fermentation. This strain showed a quicker process of mitochondrial degeneration, an altered intracellular amino acid profile and laxer nitrogen catabolite repression regulation.

Conclusions:  The proteomic and transcriptomic analysis is useful to detect key molecular adaptation mechanisms of biotechnological interest for industrial processes. ILV5 gene seems to be important in wine yeast adaptation to low-temperature fermentation.

Significance and Impact of the Study:  This study provides information that might help improve the future performance of wine yeast, either by genetic modification or by adaptation during industrial production.


Temperature fluctuations are an inevitable aspect of microbial life in exposed natural environments. However, suboptimal temperatures are also common in industrial processes. Low temperatures (10–15°C) are used in wine fermentations to enhance the production and retain flavour volatiles. In this way, white and rosé wines of greater aromatic complexity can be achieved. The improved quality of wines produced at low temperatures can be attributed to greater terpene retention, a reduction in higher alcohols and an increase in the proportion of ethyl and acetate esters in the total volatile compounds (Torija et al. 2003; Beltran et al. 2008). The optimum fermentation temperature for Saccharomyces is between 25 and 28°C. Therefore, among the difficulties inherent to wine fermentation (high concentration of sugars, low pH, presence of ethanol, nutrient deficiency, etc.), we should add a suboptimal temperature for the primary fermentation agent. Temperature affects both yeast growth and fermentation rate, with lower temperatures giving rise to a very long latency phase of up to 1 week or longer and sluggish fermentations (Meurgues 1996; Bisson 1999), which dramatically increase the duration of alcoholic fermentation, with the consequent energy expenditure.

Little is known about the molecular mechanisms that govern adaptation to cold, whereas the response of Saccharomyces cerevisiae to heat-shock stress has been investigated widely (Al-Fageeh and Smales 2006). Low temperature affects a variety of cellular processes in and the characteristics of S. cerevisiae. Previous studies have revealed that protein translational rates, cell membrane fluidity, RNA secondary structure stability, enzymatic activity, protein folding rates and heat-shock protein regulation are significantly affected (Schade et al. 2004; Aguilera et al. 2007; Tai et al. 2007; Pizarro et al. 2008). The physiological consequences of this altered state are a decreased transport, accumulation of misfolded proteins and reduced enzyme activities. The cells respond to these modifications in their physiological and biochemical state by rapidly changing processes such as protein phosphorylation and degradation, and longer-term effects involving transcriptional changes (Schade et al. 2004). To date, most studies have mainly been focused on the genome-wide transcriptional responses to cold shock (Sahara et al. 2002; Homma et al. 2003; Schade et al. 2004; Murata et al. 2005). During wine fermentations, yeasts must adapt to a new medium when they are inoculated in the grape-must (osmotic-shock, pH-shock and temperature-shock). Substantial transcriptional changes have been detected in wine yeast during the lag phase or adaptation period (Rossignol et al. 2006; Novo et al. 2007). The main responses involved the activation of some genes of the fermentation pathway and of the nonoxidative branch of the pentose pathway, and the induction of a huge cluster of genes related to ribosomal biogenesis and protein synthesis. In addition to these global transcriptional analyses, Salvadóet al. (2008) analysed the changes in the protein profile of a wine yeast during the first 24 h of fermentation after inoculation in synthetic grape-must. Protein changes mainly involved enzymes linked to nitrogen and carbon metabolism, and proteins related to cellular stress response. These changes, which occur during the lag phase, determine proper adaptation to the new medium and affect fermentative capacity and fermentation performance (Quain 1988; Blomberg 1997).

In this study, we have compared changes in the proteome profile of a commercial wine yeast during the lag phase and initial exponential phase at two fermentation temperatures: optimum temperature (25°C) and a restrictively low temperature (13°C). This value of low temperature was used because, in a previous study with fermentations at different temperatures (Llauradóet al. 2005), the best aromatic wine profile was obtained at 13°C. We identified nine proteins as representing the most significant changes in proteomic maps during the first 24 h of fermentation at low and standard temperature (25°C). A transcription analysis was performed to correlate with the proteomic experiment. Knockout and overexpressing strains were constructed and tested to evaluate their ability to trigger fermentation. This study provides information that might help improve the future performance of wine yeast, either by genetic modification or by adaptation during industrial production.

Materials and methods

Yeast strains and growth conditions

The commercial wine yeast S. cerevisiae QA23 (Lallemand S.A., Montreal, Quebec, Canada) was used in this study because it shows a good fermentative behaviour at both low and optimum temperature (Llauradóet al. 2005). This strain was cultivated in a synthetic must (SM) prepared according to Riou et al. (1997) but with 200 g l−1 of reduced sugars (100 g l−1 glucose and 100 g l−1 fructose). The yeast assimilable nitrogen (YAN) content was 300 mg N l−1; ammoniacal nitrogen (NH4Cl), 120 mg N l−1; and amino acids, 180 mg N l−1 (expressed as mg N l−1: 4·65 Asp, 11·39 Glu, 10·40 Ser, 47·87 Gln, 3·05 His, 3·40 Gly, 8·87 Thr, 29·60 Arg, 22·90 Ala, 1·51 Tyr, 2·41 Cis, 5·29 Val, 2·93 Met, 11·95 Trp, 3·20 Phe, 3·47 Ile, 5·14 Leu and 1·62 Lys). This active dry wine yeast (ADWY) was rehydrated in water, following the manufacturer’s recommendations (30 min at 37°C). After counting under the microscope, the appropriate dilution of the rehydrated QA23 was transferred to SM to achieve an initial cell concentration of 2 × 106 cells ml−1. Cell suspension was incubated at 13 and 25°C with slight shaking to obtain homogeneous nutrient distribution in laboratory-scale fermentations: 2-l bottles filled with 1·8-l medium and fitted with closures that enabled carbon dioxide release and samples to be removed. Samples were collected after 1, 4 and 24 h for the proteomic analysis and at time 20 min, 40 min, 1 h, 2 h, 4 h and 24 h for the transcription analysis. The number of viable cells was monitored by plating on YPD medium.

Construction of deletion mutant and overexpressing strains

To simplify the generation of mutant strains, the derivative haploid hoQA23 of the wine strain was constructed by disruption of the HO gene and substitution by the KanMX4 cassette (Walker et al. 2003). The transformants were sporulated, and the spores were selected for their geneticin resistance. The haploid state of spores, which had not re-diploidized owing to successful disruption of HO, was verified by their failure to sporulate, by PCR determination of their MAT locus constitution (Huxley et al. 1990) and by flow cytometry (Bradbury et al. 2005). After a screening of twenty of these HO disruptants, the haploid strain most similar to the parental wine strain in terms of viability and fermentation capacity was selected to construct the mutants. The KanMX marker of the selected haploid strain hoQA23 was excised using the Cre-lox system. This strain was transformed with the plasmid Yep351-Cre-Cyh (Güldener et al. 1996), which carries the positive marker CYHR, conferring resistance to cycloheximide, and the CRE gene under the control of the inducible GAL1 promoter. Expression of the Cre recombinase was induced by shifting cells from YPD to YPGal (galactose) medium.

Knockout strains used in this study were constructed using the KanMX cassette, flanked by 50 nucleotides with homologous sequences within the ORF of each target gene (Güldener et al. 1996). Primers used for each gene are shown in Table 1. These primers were used to amplify the KanMX sequence from pUG6 plasmid and obtain the corresponding cassette for each target gene. Derivative haploid hoQA23 transformation was performed using a lithium acetate-based method, as described in Gietz and Woods (2002). After transformation, strain selection was carried out using geneticin (G418) added to YPD solid media at a concentration of 200 mg l−1. Knockout strains were confirmed by PCR and real-time PCR (rt-qPCR).

Table 1.   Primers used in this study
Target genePrimer (FW)Primer (RV)
  1. *Primers used for knockout cassette amplification. Underlining indicates homology to the loxP-kanMX4-loxP cassette from plasmid pUG6. The remainder sequences of the primers are homologous to the flanking region of the deleted ORF.

  2. †Primers used for overexpressing strain construction. Underlining indicates the recombination sequences homologues to the plasmid ends.

  3. ‡Primers used for transcriptomic analysis with rt-qPCR.


Overexpressing strains were obtained using pGREG505 vector within a galactose-inducible promoter (pGAL1), as described in Jansen et al. (2005). Briefly, the wine yeast hoQA23 was co-transformed with the SalI digested pGREG505 plasmid together with the PCR-amplified target gene, flanked by recombination sequences homologues to the plasmid ends (primers are shown in Table 1). The lithium acetate transformation method was also used (Gietz and Woods 2002). The cells carrying the multicopy plasmid were selected by growing the transformants in YPD with geneticin (200 mg l−1). Correct plasmid ligation and insertion was confirmed by PCR. During overexpression constructions, HIS3 fragment, within pGREG505 plasmid, was replaced by target gene. For overexpressing strains, the reference or control strain used in the different experiments was the haploid strain hoQA23, with the plasmid pGREG505 without the HIS3 fragment (hoQA23pGREG). The overexpression of these genes was tested by rt-qPCR. All the strains (mutant and overexpressing) constructed in this study are shown in Table 2.

Table 2.   Original and constructed strains in this study
QA23MATa/MATαWine commercial strain
hoQA23MATα; YDL227C::kanMX4Derivative wine haploid strain
Δgdh1hoQA23; YOR375C::kanMX4GDH1 mutant strain
Δhsp12hoQA23; YFL014W::kanMX4HSP12 mutant strain
Δhsp26hoQA23; YBR072W::kanMX4HSP26 mutant strain
Δpdc1hoQA23; YLR044C::kanMX4PDC1 mutant strain
Δtdh1hoQA23; YJL052W::kanMX4TDH1 mutant strain
Δyhb1hoQA23; YGR234W::kanMX4YHB1 mutant strain
Δure2hoQA23; YNL229C::kanMX4URE2 mutant strain
hoQA23-pGREGhoQA23-pGREG505Haploid strain with empty plasmid
pGal1-GDH1hoQA23-pGREG GDH1GDH1 overexpressing strain
pGal1-HSP12hoQA23-pGREG HSP12HSP12 overexpressing strain
pGal1-HSP26hoQA23-pGREG HSP26HSP26 overexpressing strain
pGal1-ILV5hoQA23-pGREG ILV5ILV5 overexpressing strain
pGal1-PDC1hoQA23-pGREG PDC1PDC1 overexpressing strain
pGal1-TDH1hoQA23-pGREG TDH1TDH1 overexpressing strain
pGal1-YHB1hoQA23-pGREG YHB1YHB1 overexpressing strain

Protein extraction and two-dimensional electrophoresis

Protein extracts were prepared as described in Blomberg (2002). Briefly, this was as follows: cell suspension was vortexed for 4 × 30 s with glass beads containing PMSF as a protease inhibitor (placed on ice between vortexing) and subsequently boiled for 5 min with SDS/mercaptoethanol buffer. Following nuclease treatment of the cells, protein contents of the extract were estimated using a 2-D quant kit (Amersham Pharmacia Biotech, Uppsala, Sweden). Soluble proteins were run in the first dimension using a commercial horizontal electrophoresis system (Multiphor II; Amersham Pharmacia Biotech). Forty-five micrograms of protein from whole-cell lysates was mixed with immobilized pH gradient (IPG) rehydration buffer (8 mol l−1 urea/2% NP-40/10 mmol l−1 DTT; final volume of 500 μl) and loaded onto polyacrylamide strips. The IPG strips with a nonlinear pH gradient 3–10 were allowed to rehydrate overnight. Samples were run at 50 μA per strip; in the first step voltage was ramped to 500 V over a period of 5 h, kept at 500 V for 5 h more, ramped again to 3500 V over a period of 9·5 h, and finally kept at 3500 V for 5 h. After the first dimension, IPG strips were then equilibrated twice for 15 min in equilibration solution (0·05 mol l−1 Tris–HCl, pH 8·8, 6 mol l−1 urea, 30% v/v glycerol and 2% w/v SDS), first with 65 mmol l−1 DTT (reduction step), and finally with 135 mmol l−1 iodoacetamide (alkylation). The second dimension was carried out in a vertical electrophoresis system (Ettan DALTsix; Amersham Pharmacia Biotech) in a 12·5% (26 cm × 20 cm × 1 mm) polyacrylamide gel where proteins were separated based on molecular size. Electrophoresis conditions were 1 W per gel until the dye front reached the bottom of the gel. Sets of three gels were used for each sampling time.

Silver staining protocol and image analysis

The staining protocol was as described by Blomberg (2002). Gels were scanned using an Image Scanner UMAX, Amersham (300 dpi, 12-bit image), enabling the acquisition of spot intensities in pixel units. Images were analysed using ImageMaster 2D Platinum (GE Healthcare, Little Chalfont, UK). Normalization was performed with the aforementioned software based on the total gel density to compensate differences in images caused by variations in experimental conditions (e.g. protein loading or staining). Spots detection was performed with ImageMaster 2D platinum-automated spot detection algorithm. The gel image with the greatest number of spots and the best protein pattern was chosen as a reference template for the image analysis. Spots in the standard gel were then matched across each 2DE gel at a reference temperature (25°C) for each sampling time (1, 4 and 24 h after inoculation in the media). Matching features of the software were used to relate and compare the sets of gels. Finally, to achieve maximum reliability and robustness of the results, the Student’s t-test was performed allowing us to identify those sets of proteins displaying a statistically significant difference with the confidence level set at 95%.

In-gel protein digestion

Protein spots of interest were excised manually and then digested automatically using a Proteineer DP protein digestion station (Bruker-Daltonics, Bremen, Germany), according to Shevchenko et al. (1996) with minor variations. Gel plugs were reduced with 10 mmol l−1 dithiothreitol (Amersham Biosciences, Sweden) in 50 mmol l−1 ammonium bicarbonate (99·5% purity; Sigma, St Louis, MO, USA) and alkylated with 55 mmol l−1 iodoacetamide (Sigma Chemical) in 50 mmol l−1 ammonium bicarbonate. The gel pieces were then rinsed with 50 mmol l−1 ammonium bicarbonate and acetonitrile (gradient grade; Merck, Germany) and dried under a stream of nitrogen. Modified porcine trypsin (sequencing grade; Promega, Madison, WI, USA) at a final concentration of 13 ng μl−1 in 50 mmol l−1 ammonium bicarbonate was added to the dry gel pieces and digestion proceeded at 37°C for 6 h. Finally, 0·5% trifluoroacetic acid (99·5% purity; Sigma Chemical) was added for peptide extraction.

MALDI-MS(/MS) and database search

An aliquot of the above digestion solution was mixed with an aliquot of α-cyano-4-hydroxycinnamic acid (Bruker-Daltonics) in 33% aqueous acetonitrile and 0·1% trifluoroacetic acid. This mixture was deposited onto a 600-μm AnchorChip MALDI probe (Bruker-Daltonics) and allowed to dry at room temperature. MALDI-MS(/MS) data were obtained using an Ultraflex time-of-flight mass spectrometer (Bruker-Daltonics) equipped with a LIFT-MS/MS device (Suckau et al. 2003). Spectra were acquired in the positive-ion mode at 50-Hz laser frequency, and 100–1500 individual spectra were averaged. For fragment ion analysis in the tandem time-of-flight (TOF/TOF) mode, precursors were accelerated to 8 kV and selected in a timed ion gate. Ionized fragments generated by laser-induced decomposition of the precursor were further accelerated by 19 kV in the LIFT cell, and their masses were analysed after passing the ion reflector. Measurements were in part taken using post-LIFT metastable suppression, which allowed the removal of the precursor and metastable ion signals produced after extraction from the second ion source. Peptide mass mapping data were analysed in detail using flexAnalysis software (Bruker-Daltonics). MALDI-TOF mass spectra were calibrated internally using two trypsin autolysis ions with m/= 842·510 and m/= 2211·105; for MALDI-MS/MS, spectra were calibrated with fragment ion spectra obtained for the proton adducts of a peptide mixture covering the 800–3200 m/z region. MALDI-MS and MS/MS data were combined through the MS BioTools program (Bruker-Daltonics), to search the NCBInr database using Mascot software (Matrix Science, London, UK; Perkins et al. 1999).

RNA extraction and cDNA synthesis

Total RNA was isolated from yeast samples as described by Sierkstra et al. (1992) and resuspended in 50 μl of DEPC-treated water. Total RNA suspensions were purified using the High Pure Isolation kit (Roche Applied Science, Mannheim, Germany), following the protocol provided by the manufacturer. RNA concentrations were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA), and the quality of RNA was verified electrophoretically on 0·8% agarose gels. Solutions and equipment were treated so that they were RNase free, as outlined in Amberg et al. (2005).

Total RNA was reverse-transcribed with Superscript™ II RNase H Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) in a GenAmp PCR System 2700 (Applied Biosystem, Foster City, CA, USA). Then, 0·5 μg of Oligo (dT)12–18 Primer (Invitrogen) was used with 0·8 μg of total RNA as template in a reaction volume of 20 μl. Following the protocol provided by the manufacturer, after denaturation at 70°C for 10 min, cDNA was synthesized at 42°C for 50 min. Finally, the reaction was inactivated at 70°C for 15 min.

Real-time quantitative PCR

The PCR primers used in this study are listed in Table 1. The primers were all designed with the available GenBank sequence data and the Primer Express software (Applied Biosystems) in accordance with the Applied Biosystems guidelines for designing PCR primers for quantitative PCR with the exception of the housekeeping gene ACT1, previously described by Beltran et al. (2004). All amplicons were shorter than 100 bp, which ensured maximal PCR efficiency and, therefore, the most precise quantification.

For each gene, a standard curve was made with yeast genomic DNA. DNA extraction was performed as described by Querol et al. (1992), digested by RNase and isolated by twofold phenol–chloroform extractions and ethanol precipitation. Concentration was determined using a GeneQuant spectrophotometer (Pharmacia, Uppsala, Sweden). Serial tenfold dilutions of DNA were conducted to yield DNA concentrations from 400 to 4 × 10−2 ng μl−1. These dilution series were amplified (in triplicate) by SYBR PCR for each gene to obtain standard curves. The standard curve displays the Ct value vs. log 10 of the starting quantity of each standard. The starting quantity of the unknown samples was calculated against the standard curve by interpolation. Gene expression levels are shown as the concentration of the studied gene normalized with the concentration of the housekeeping ACT1 gene.

The rt-qPCR reaction was performed using SYBR® Green I PCR (Applied Biosystems). The 25-μl SYBR PCR reactions contained 300 nmol l−1 of each PCR primer, together with 1 μl cDNA (or 5 μl of each DNA serial dilution for standard tubes) and one time SYBR master mix (Applied Biosystems).

All PCRs were mixed in 96-well optical plates (Applied Biosystems) and cycled in a PE Applied Biosystems 5700 thermal cycler under the following conditions: 50°C for 2 min, 95°C for 10 min and 40 cycles at 95°C for 15 s and at 60°C for 60 s. All samples were analysed in triplicate, and the expression values were averaged using analysis software (Applied Biosystems).

Determination of fermentation activity

Fermentation activity of deletion mutant and overexpressing strains was analysed after inoculation in a synthetic grape-must (described above) with a population size of 2 × 106 cells ml−1. To obtain this inoculum, the deletion mutant strains were grown in YPD overnight at 28°C and the overexpressing strains were grown in YPGal with geneticin to induce synthesis of the overexpressed gene at 28°C. Fermentation volume for all the strains was 400 ml of media in 500-ml bottles. The consumption of the initial 15% of sugars was monitored by measuring the relative density of the media (g l−1), using a Densito 30PX densitometer (Mettler Toledo, Schwerzenbach, Switzerland).

Intracellular amino acid determination

A total of 108 cells cultured in YPGal were harvested and washed twice with miliQ water. After that, cells were suspended in 500 μl of miliQ water and incubated at 100°C for 15 min, centrifuged at 15 700 g for 10 min at 4°C. The supernatant was collected and used for analysis.

Amino acids and ammonium ion were determined simultaneously using the method of Gómez-Alonso et al. (2007). Briefly, samples (400 μl) were derivatized by 15 μl of diethylethoxymethylenemalonate (Fluka, Buchs, Switzerland) in the presence of 700 μl of 1 mol l−1 borate buffer (pH 9), 300 μl of methanol and 10 μl of L-2-aminoadipic acid (Internal Standard, 1 g l−1) over 30 min in an ultrasound bath. Then, the samples were treated at 80°C for 2 h. The analyses were performed on an Agilent 1100 Series HPLC (Agilent Technologies, Waldbronn, Germany) comprising a quaternary pump, an autosampler and a multiple wavelength detector at 269, 280 and 300 nm. Nitrogen compound separation of sample (50 μl) was carried out using a 4·6 × 250 mm, 5 μm ACE C18-HL column (Symta, Madrid, Spain) with a guard column (ACE5 C18-HL) through a binary gradient (Gómez-Alonso et al. 2007) at a flow of 0·9 ml min−1. The different nitrogen compounds were identified according to the retention time of corresponding standards and were quantified using the internal standard method.

Mitochondrial staining

DiOC6 (Invitrogen) was used to stain mitochondria of living cells (Koning et al. 1993). Staining protocol and DiOC6 working concentration were according to the manufacturer recommendations. Briefly, 106 cells ml−1 were resuspended in 10 mmol l−1 HEPES buffer, pH 7·4, containing 5% glucose. DiOC6 was added up to a final concentration of 175 nmol l−1. Cells were observed under fluorescence microscopy Leica DM4000 B (Leica Microsystems, Wetzlar, Germany) after a room-temperature incubation period of 15 min.

Statistical analysis

Proteomic and transcription data are the result of three independent fermentations at 13 and 25°C. Statistical test for 2DE image analysis was performed using the Student’s t-test module of ImageMaster 2D Platinum, with confidence level set at 95%. Significant differences among strains were checked by means of an analysis of variance, using the one-way anova module of the statistical software package spss 13.0 (IBM Corp., Armonk, NY). Fermentations with the mutant and overexpressing strains were also repeated at least three times.

The software MeV 4.8.0 (Multiple Experiment Viewer; Madison, WI, USA) has been used to perform a heat-map representing the ratio between gene transcription and protein concentration at both temperatures.


Effect of fermentation temperature on changes in proteome profile

We have compared the changes in proteome profile of the commercial wine strain QA23 after inoculation in a synthetic grape-must at two different temperatures, 13 and 25°C. The differences in protein composition between temperatures were obtained after the comparison between gels from the same time-points after inoculation (1, 4 and 24 h) and by applying a restrictive statistical analysis to obtain a reduced and reliable set of proteins with clear differences between both temperatures (only proteins whose concentration was modified at least twofold) (Fig. 1). There were no significant differences in proteins after 1 h, and only Hsp12p presented a higher cellular concentration after 4 h at low temperature (Fig. 1). The remaining proteins with statistically significant differences were observed in the 24-h sample, including Hsp12p. Most of these proteins can generally be grouped into three main categories: glucose metabolism (Adh1p, Pdc1p and Tdh1p), stress response (Hsp12p, Hsp26p and Yhb1p) and nitrogen metabolism (Gdh1p and Ilv5p). The elongation factor Eft2p, involved in ribosomal translocation during protein synthesis, also showed significant differences between both temperatures.

Figure 1.

 Proteins with significant differences of concentration between both temperatures at the same time-points after inoculation (1, 4 and 24 h). Positive and negative values represent higher and lower concentration at 13°C after 24 h, respectively. * Protein with different amount in the 4 h sampling point.

Transcription analysis of selected genes

To check whether differences in protein concentration were also correlated with differences in transcription activity, we also analysed the relative expression of some selected genes encoding these proteins at both temperatures. Several representatives of each functional category were chosen. The genes analysed were as follows: GDH1, HSP12, HSP26, ILV5, PDC1, TDH1 and YHB1 (Fig. 2a). We have also represented the ratio 13°C/25°C of these relative gene activities in a heat-map (Fig. 2b). To make easier the comparison between transcripts and protein levels, the last column of this heat-map also represents the same ratio of the protein concentration (fold change of Fig. 1).

Figure 2.

 Transcription analysis of selected genes. (a) Relative gene expression of the selected genes at different stages of alcoholic fermentation at 25°C (filled circles) and 13°C (open circles). Changes in gene activity are shown relative to the expression of the 20-min sampling time-point at 25°C (set as value 1). (b) Heat-map depicts the ratio of the gene expression and protein concentration at 13°C/25°C. Positive (red) and negative (green) values represent higher and lower gene expression (protein concentration) at 13°C, respectively.

Generally speaking, we detected a positive correlation between both transcription and translation (Fig. 2b), despite an evident delay in some cases between increased gene activity and protein concentration. The mRNA of the stress proteins, HSP12 and HSP26, decreased at both temperatures from the first sampling time-point until 24 h after inoculation. This decrease was faster at 25°C than at 13°C, which resulted in increasing values of the ratio 13°C/25°C. Accordingly, Hsp12p was not detected at 25°C, while it was still detectable at 13°C after 24 h. At the last sampling time-point (24 h), HSP26 gene activity increased at 25°C (ratio 25°C/13°C of 18·30). This increase in gene activity also correlated with a protein concentration 3·8 times higher at 25°C than at 13°C (Fig. 1). Conversely, the other stress gene YHB1 showed a stepped increased initially and a decrease after 24 h at 25°C. At 13°C, this increase was smaller initially, although this gene was more active after 24 h. Again, this higher transcriptional activity correlated with a higher protein concentration at 13°C after 24 h.

The PDC1 gene showed an up-regulation after yeast inoculation at both temperatures. This increase in activity occurred faster at 25°C, reaching a maximum in the 4-h sample and dropping again in the 24 h-sample. Conversely, the maximum activity in cells growing at low temperature was detected in the 24-h sample. In this case, protein differences were observed 24 h after inoculation and did not correlate with RNA levels because a higher concentration was detected at 25°C. Meanwhile, TDH1 showed a clear correlation between RNA levels and protein concentration after 24 h of inoculation. Tdh1p concentration was 2·6-fold higher at 25°C, while the TDH1 RNA level was over 18 times higher in the same cells growing at 25°C. At the other sampling time-points, this gene was down-regulated after inoculation and this decrease was faster at 25°C.

Transcription levels of nitrogen metabolism-related genes showed different profiles. Maximum GDH1 activity was detected in the first samples after inoculation. However, this gene activity was higher at 25°C than at 13°C, which correlated with higher protein concentrations at this temperature. ILV5 showed an up-regulation trend at all sampling time-points, with higher activity at 25°C. However, protein concentration was higher at low temperature in the 24-h sample.

Phenotypic evaluation of deletion mutant and overexpressing strains

Once proteins with differential concentration at both temperatures had been identified, the importance of the genes encoding these proteins was determined by deleting or overexpressing these genes in a wine strain, with the exception of Δilv5, which was unviable in this genetic background. To simplify the generation of mutant strains, we constructed the derivative haploid hoQA23 of the wine strain. This haploid strain did not differ from the wild-type strain QA23 in terms of growth and fermentation rate (data not shown). Transcriptional activity in all mutant strains was determined during the preculture in YPD, and as expected, no transcriptional activity was detected. Moreover, the overexpressed genes were controlled by a GAL promoter, which was strongly induced during the preculture of the overexpressing strains (preculture in YPGal). However, these genes were repressed when strains were inoculated in the synthetic grape-must, owing to the high glucose concentration. As mentioned above, we were interested in the adaptation of the wine yeast to the grape-must after inoculation. Thus, with this strategy, the overexpressing strains are enriched in a specific protein during preculture, allowing a better adaptation and giving rise to the differential phenotype during the first hours of wine fermentation. Transcriptional activity in all overexpressing strains was also determined during the preculture in YPGal, and all of them showed an increase in activity ranging from 12 to 15 times more than the control strain with the empty plasmid pGREG505 (data not shown).

Both mutant and overexpressing strains were phenotypically evaluated according to their fermentation capacity in the first hours of fermentation. To this aim, it was determined the time needed to consume 15% of total sugars in a synthetic grape-must during fermentation at 25 and 13°C (Fig. 3). Curiously, all the tested strains showed worse fermentation performance than the reference strains (hoQA23 for the mutants and hoQA23pGREG for the overexpressing strains) at 25°C. Most of these strains (both mutant and overexpressing) also fermented worse at 13°C, but some of them improved their fermentation capacity at this temperature. The only strain showing significant better fermentation performance was the pGAL1-ILV5.

Figure 3.

 Phenotypic evaluation of deletion mutant and overexpressing strains. Relative values to reach 15% of sugar metabolization by tested strains in synthetic must at (a) 25 and (b) 13°C. Time zero represents the amount of time required by the reference strains (hoQA23 for the mutants and hoQA23pGREG for the overexpressing strains) to reach 15% of sugar consumption. Negative values represent a shorter time, and therefore, positive values correspond to longer time. *Significant differences by anova with P-value ≤0·05.

Effect of ILV5 overexpression on nitrogen metabolism and mitochondrial organization

As the pGAL1-ILV5 strain showed an improved fermentation activity, we decided to perform a more in-depth phenotypic analysis of this strain. ILV5 is a gene that encodes a mitochondrial protein (acetohydroxyacid reductoisomerase) involved in branched-chain amino acid (BCA) biosynthesis of leucine, isoleucine and valine (Zelenaya-Troitskaya et al. 1995). Thus, we analysed the effect of ILV5 overexpression on the intracellular amino acid profile. Surprisingly, BCA concentrations were not modified compared to the control strain hoQA23pGREG, whereas glutamate, glutamine, arginine and proline were found in higher concentrations in this control strain. On the other hand, alanine and γ-aminobutyric acid (GABA) were present in higher concentrations in the pGAL1-ILV5 strain (data not shown).

Intracellular glutamine concentration has been related to the triggering of nitrogen catabolite repression (NCR). NCR involves transcriptional regulation by four members of the GATA family of transcription factors, as well as the regulatory protein Ure2 (Cooper 2002). A glutamine signal activates Ure2, which binds Gln3, retaining it in the cytoplasm and preventing the activation of nitrogen-regulated genes. Conversely, lower levels of cytoplasmic glutamine would induce dissociation of Gln3 from Ure2, dephosphorylation of Gln3, and importation of Gln3 into the nucleus, causing derepression (activation of genes involved in the transport and metabolism of the poorer nitrogen sources) (Magasanik and Kaiser 2002). Therefore, we should expect a more nitrogen-derepressed situation in the pGAL1-ILV5 strain than in the control strain hoQA23pGREG. To infer the effect of ILV5 overexpression on nitrogen regulation, we monitored expression levels of two marker genes affected by NCR, GAP1 and MEP2. These genes are strongly repressed in the presence of rich nitrogen sources in the wine yeast used in this study (Beltran et al. 2004). We also deleted the URE2 gene of the hoQA23 strain (Δure2) for use as a NCR-derepressed control (Fig. 4). As expected, both genes were repressed 1 h after inoculation of the control strain in the fermentation medium. This repression was practically nonexistent in the Δure2 strain, while these genes were also repressed in the pGAL1-ILV5 strain, although this repression was much more moderate than in the reference strain.

Figure 4.

 Relative gene expression of the NCR marker genes (a) GAP1 and (b) MEP2. Control strain hoQA23pGREG (inline image), pGAL1-ILV5 (inline image) and Δure2 (inline image) strains. These values are relative to gene expression of the sample at time 20 min in the control strain.

However, Ilv5p is considered a bifunctional protein because it is not only involved in BCA biosynthesis but also plays a role in maintaining mtDNA stabilization (Zelenaya-Troitskaya et al. 1995; Bateman et al. 2002; Macierzanka et al. 2008). We also analysed mitochondrial organization during the first 24 h of fermentation in the pGAL1-ILV5 and reference strain hoQA23pGREG by simple DiOC6 staining (Fig. 5). In the reference strain, very well-structured mitochondria could be observed in the first two samples (0 and 4 h after inoculation), whereas these mitochondria developed into unorganized structures in the latter samples (8 and 24 h). On the other hand, the pGAL1-ILV5 strain showed these unstructured mitochondria from time zero (just before inoculation).

Figure 5.

 Mitochondrial organization in the strains hoQA23-pGREG (reference strain) and pGAL1-ILV5 during the first 24 h of synthetic grape-must fermentation at 25°C (DiOC6 staining).


In this study, we were interested in the ability of yeast to adapt to the fermentation medium after inoculation. To this end, first the proteome of a commercial wine yeast strain was compared in the first hours of fermentation at optimum (25°C) and low temperature (13°C). The time frame studied comprised the first 24 h of fermentation. This period covered the lag and initial log phase of the wine yeast after inoculation. A previous study (Salvadóet al. 2008) analysed the changes occurring in the rehydrated yeast protein profile during the first hours after inoculation under oenological-like conditions at low temperatures. Salvadóet al. (2008) detected changes in around 50 proteins, which reflected degradation or modifications in proteins present in the rehydrated yeasts. In the present study, we aimed to identify proteins whose concentration depended exclusively on the fermentation temperature by comparing the proteome at the same time-points at both temperatures. In this case, we only detected nine proteins with statistically different concentrations, which are mainly involved in stress response and in glucose and nitrogen metabolism. Most of these proteins were also transcriptionally studied during the same time frame of wine fermentation to assess the connection between transcription and translation mechanisms.

In a global transcriptional study of a wine yeast strain during alcoholic fermentation, Rossignol et al. (2003) described down-regulation of the stress genes, regulated by the general stress transcription factors Msn2/Msn4, during the exponential (growth) phase and an induction of these genes on entering the stationary phase. The slower growth of cells at low temperature explained the weaker repression of the genes HSP12 and HSP26. Hsp12 was not induced at any of the time-points studied, thus leading to higher protein concentrations 4 and 24 h after inoculation at 13°C. Hsp12 is a small heat-shock protein, which is induced under different conditions, such as low and high temperatures, osmotic or oxidative stress and high sugar or ethanol concentrations (Praekelt and Meacock 1990). Recently, Pacheco et al. (2009) has shown the role played by this protein in freezing-stress tolerance of yeast. Our results also indicate a role in the adaptation to mild cold stress. The other stress protein with high concentrations at low temperatures is Yhb1. This protein plays a role in oxidative and nitrosative stress responses but so far has not been linked to cold adaptation. However, it has previously been reported (Zhang et al. 2003) that a rapid downshift in the growth temperature of S. cerevisiae from 30 to 10°C results in an increase in transcript levels of the antioxidant genes SOD1, CTT1 and GSH1. In our previous proteomic study (Salvadóet al. 2008), we also detected an increase in Cys3, a protein related to antioxidant response, during adaptation to low-temperature fermentation. The present study found stronger activity of the YHB1 gene in the first minutes and 24 h after inoculation at low temperatures. Thus, low temperature might induce stronger oxidative stress in yeast cells.

Salvadóet al. (2008) reported an increase in glycolytic enzymes after yeast inoculation. We have detected two of these enzymes, Pdc1 and Tdh1, in a significantly higher concentration at optimum temperature than at low temperature. Conversely to protein concentration, the transcriptional profile of these glycolytic genes showed a higher induction at low temperature in all the samples analysed, with the exception of TDH1 after 24 h at 25°C. Rossignol et al. (2009) also studied the proteome of a wine yeast strain during exponential growth and stationary phase during fermentation. As in our study, glycolysis, amino acid metabolism and stress proteins were the most represented functional categories. They also tried to correlate protein abundance with changes in mRNA using the data obtained in a previous transcriptome analysis performed under the same conditions (Rossignol et al. 2003). They concluded that proteins and transcripts in the glycolysis category did not show correlation. This lack of correlation can easily be explained by multiple translational and post-translational regulation mechanisms. However, other mechanisms such as gene autoregulation could explain why higher protein levels correlate with lower gene activity. This is the case for PDC1, which is activated in the absence of Pdc1 (Eberhardt et al. 1999).

We constructed mutants and strains overexpressing most of these genes to determine the importance of these genes in yeast growth and fermentation at low temperature. The overexpression of ILV5 improved the period required to consume the 15% of grape-must sugars, which could be related to a better adaptation to the new medium and temperature conditions after yeast inoculation. In the proteome analysis, this protein was in higher concentration at low temperature. It seems logical that the increase in this protein leads to an increase in fermentation capacity at low temperature.

Inoculation of active dry wine yeasts (ADWY) involves a shift from the respiratory metabolism during its industrial propagation to a fermentative metabolism during wine fermentation. In our previous proteome analysis (Salvadóet al. 2008), most of the mitochondrial proteins that showed significant differences 24 h after inoculation were present at lower concentrations, suggesting a transition from the respirative to the fermentative metabolism in the first hours of fermentation, which had shifted completely 1 day after inoculation. This hypothesis is supported by the mitochondrial staining performed in this study (Fig. 5). In the wild strain, the mitochondria were well-structured in the first 4 h after inoculation, but these mitochondria developed into unorganized structures after 24 h. However, the overexpression of ILV5 yielded unorganized mitochondria during the preculture phase. These disorganized mitochondria may produce a quick onset of the fermentative metabolism in the pGAL1-ILV5 strain, which would explain the quicker CO2 production in the first hours of fermentation. However, the quicker presence of unorganized mitochondria in the pGAL1-ILV5 strain does not fit with the role previously assigned to this gene in maintaining mtDNA stabilization. Zelenaya-Troitskaya et al. (1995) reported that overexpression of ILV5, at levels only two- or threefold, is sufficient to suppress the mtDNA instability phenotype of Δabf2 (a gene required for the maintenance of the wild-type (ρ+) mitochondrial DNA).

During this fermentative metabolism, the mitochondrial TCA cycle is not functional, which may alter the intracellular amino acid distribution. TCA compounds, such as oxalacetate and α-ketoglutarate, are the precursors of aspartate and glutamate, respectively. The mismatch in the nitrogen metabolism observed in the ILV5 overexpressing strain yielded a more relaxed NCR mechanism than the control strain. Intracellular glutamine and ammonium have been reported as the main triggers of this nitrogen repression (Schure et al. 2000). The lower concentration of the intracellular glutamine pool (no differences were observed in the ammonium concentration) in the pGAL1-ILV5 strain can explain this laxer NCR regulation. Owing to the link between nitrogen metabolism and aroma production during wine fermentation (Bell and Henschke, 2005), the influence of this strain to the organoleptic quality of the wines should be analysed in the future. In a recent study, Chen et al. (2011) showed the increase in isobutanol (a fusel alcohol with influence in the aroma of the wine) by simultaneous overexpression of the ILV2, ILV3 and ILV5 genes.

In conclusion, the use of proteomic and transcriptomic analysis is useful to detect key molecular adaptation mechanisms of biotechnological interest for industrial processes. Although most of these functional analyses were carried out with laboratory strains, the use of industrial strains and conditions mimicking the industrial environment are recommended. In this study, we have detected genes and proteins with different activity and concentrations at low temperature. Strains that were either mutant for or overexpressed these genes were constructed in a derivative haploid of a commercial strain to determine the importance of these genes in fermentation at low temperature. Most of these strains did not show a differential phenotype to the reference strains in fermentation activity during the first hours of fermentation, with the exception of the strain overexpressing ILV5. The mitochondria degeneration process was faster in this pGAL1-ILV5 strain, which produced an altered intracellular amino acid profile and laxer NCR regulation. It is likely that these physiological differences lead to a quicker onset of the fermentative metabolism, which explains the improved fermentation activity in the early stages. Currently, we are developing a strain overexpressing ILV5 in the wild commercial strain, using clean and safe integrating methods. Natural grape-musts will be fermented at low temperature with this genetically modified strain, and an in-depth analysis will be performed of the fitness advantage afforded to this overexpressing strain and of the final quality product.


This work was supported by project AGL2010-22001-C02-01 from the Spanish government and by project PROMETEO/2009/019 from the Valencian government.