Genetically modified S. cerevisiae strain VTT-C-99318 (CEN.PK2-1D ura3::XYL1 XYL2 his3::XKS1 kanMX) 22 was derived from CEN.PK2-1D (MATα, leu2-3/112, ura3-52, trp1-289, his3Δ1, MAL2-8^c, SUC2) 23 and contains the genes XYL1 and XYL2 of P. stipitis encoding xylose reductase and xylitol dehydrogenase, respectively, integrated into the URA3 locus. XYL1 is expressed under a PGK1 promoter and XYL2 under a modified ADH1 promoter 24. In addition, the strain contains an additional copy of the endogenous xylulokinase encoding gene of S. cerevisiae with the modified ADH1 promoter 24 and integrated into the HIS3 locus. The integration of XYL1 and XYL2 was described by Toivari et al., and that of XKS1 by Richard et al. 825.

VTT-C-99318 was grown at 30°C in 1.8-litre bioreactors (Chemap CMF Bioreactor, Chemap AG, Volketswil, Switzerland) with a working volume of 1.5 litres. The pH was kept at 5.5 by automatic addition of 2 M KOH, and the stirrer speed at 500 rpm. Mass flow controllers (Bronkhorst High-Tech B.V, Ruurlo, Netherlands) regulated the total gas flow at 0.5 standard litres per minute (SLPM). The composition of the fermentor off-gas was analysed using a QMG 421C quadrupole mass spectrometer (Balzers Pfeiffer Scandinavia AB, Sweden) as described previously 14. The medium was synthetic complete (SC) (modified from Sherman, 1983) 26) supplemented with 0.5 ml silicone antifoam l^-1 (AnalaR BDH, UK) and 50 g l^-1 xylose or 50 g l^-1glucose. Triplicate cultures (F0, F1, and F2) were carried out on both carbon sources. Cultures were started with an initial OD₆₀₀ of 1 to 1.5 from pre-cultures grown for 20 h at 30°C on SC with 30 g/L glucose.

The 25 ml samples for transcriptome and proteome analyses were harvested by centrifugation (2 min, 5000 g) after 5 h and 24 h growth for glucose cultures and after 72 h growth for xylose cultures and frozen in liquid nitrogen. Cell growth was measured as OD₆₀₀ and cell dry mass as described previously 14. Extracellular concentrations of glucose, xylose, xylitol, ethanol, acetate and glycerol were analysed by high-performance liquid chromatography as described previously 14.

Samples from the F2 cultivations on each carbon source were hybridised three times on Affymetrix Yeast Genome S98 arrays (hybridisations denoted as H2.1, H2.2 and H2.3; see [Additional files 1 and 2]) and the samples from the cultivations F0 and F1 once (hybridisations denoted as H0 and H1, respectively; see [Additional files 1 and 2]). Proteome analyses were carried out with samples taken at the same time points as the samples for the transcriptome analyses.

Affymetrix hybridisations and analyses were carried out according to the protocols provided by Affymetrix, Inc. Double-stranded cDNA was synthesised from 5 μg total RNA. An in vitro transcription (IVT) reaction was then carried out to produce biotin-labelled cRNA from the cDNA. The cRNA was fragmented and hybridised to a Test3 array for quality control and then to the Yeast Genome S98 array for 16 h at 45°C. The arrays were washed and stained with streptavidin-phycoerythrin in an Affymetrix Fluidics Station 400 and scanned (Agilent G2500A GeneArray Scanner). The data were extracted with Affymetrix MicroArray Suite 5.0 software.

Array data were pre-processed using the Robust Multichip Average (RMA) 2728 implemented in the Bioconductor (version 1.5.0) 29 extension to the R software environment for statistical computing and graphics (version 2.0.1). The RMA-method computes log₂ scale expression values from cell intensity files using the RMA model of probe-specific correction of perfect match (PM) probes and quantile normalisation 28. A visual inspection of the scatter plots of the RMA pre-processed expression values, and the Pearson correlations between the replicate arrays did not indicate outlying arrays [Additional files 3, 4, 5].

After the pre-processing, TIGR MultiExperiment Viewer (version 3.1) was used to analyse the expression values 30. The replicate values were averaged and genes which were differentially expressed between the glucose repressed and derepressed samples and the xylose samples were detected using one-way ANOVA at a p-value 0.01 based on 1000 permutation tests 31. The expression values of significantly expressed genes (1439) were then mean centred, the biological replicate arrays were averaged and the values clustered using K-means with Euclidean distance 32. Clustering was performed several times with varying parameters and the most optimal result was obtained with 13 clusters. Since the 13 clusters had 8 distinct expression profiles, and furthermore, as combining the clusters with similar expression profiles lowered the p-values of best gene ontology (GO) classes 33, the clusters were manually reorganised into 8 clusters.

5–10 mg dry mass of cells was suspended in 150 μl of 10% (v/v) trichloroacetic acid (TCA, Merck, USA) in a 1.5 ml microcentrifuge tube. 500 μl glass beads (0.5 mm diameter; Biospec Products, USA) were added and the tubes shaken in a MiniBeadbeater 8 (Biospec Products) at homogenisation speed three times for 30 seconds. The tubes were cooled on ice between each homogenisation step. The supernatants were collected and proteins were precipitated by adding 600 μl of -20°C acetone and incubating 30 min on ice. Precipitated proteins were collected by centrifugation for 30 min, 13 000 rpm, at 4°C, rinsed once with 600 μl of -20°C acetone, and re-suspended in 450 μl of 7 M urea (Promega, USA), 2 M thiourea (Fluka, USA), 4% (w/v) CHAPS (Fluka), 1% (w/v) Pharmalytes 3–10 (Pharmacia, Sweden) and 1% (w/v) DTT (Sigma) by gently shaking for 20 min at room temperature. Supernatants were collected by centrifugation for 5 min, 13 000 rpm. The protein concentration of the supernatants was determined by the Non-Interfering Protein Assay (Geno Technology, Inc., USA) and the samples were stored at -70°C prior to isoelectric focusing.

The isoelectric focusing and the second dimension on 11% (w/v) SDS-PAGE were carried out as described earlier 15. After electrophoresis the gels were fixed for 1.5 h in 30% (v/v) ethanol and 0.5% (v/v) acetic acid, and stained either only with Sypro Ruby (Molecular Probes, USA) or with phosphoprotein-specific Pro-Q Diamond (Molecular Probes) followed by Sypro Ruby, according to instructions of the manufacturer. The stained gels were scanned with a resolution of 100 microns on a Typhoon instrument (GE Healthcare, USA). The gel images were analysed using the PDQuest software (Bio-Rad, USA). The gel patterns from different gels were automatically matched with some additional manual editing, and the quantities of matching spots in different gels were compared. For each condition, average spot quantities were calculated from two to four gels of each sample taken from each of the triplicate glucose and xylose cultivations. The quantification of the resolved proteins was normalised to the total optical density in each gel image. The data from 2-DE (two-dimensional electrophoresis) gels were analysed using the TIGR MultiExperiment Viewer (version 3.1) 30. Proteins which had different abundance in glucose compared with xylose cultures were identified using the one-way ANOVA at a p-value of 0.01 31. The 220 protein-spot intensity values obtained were mean centred, the spots from replicate gels averaged, and the signal intensity values for identified protein spots (in total 70) were clustered using hierarchical clustering with Euclidean distance and average linkage 34.

The degree of phosphorylation was calculated by dividing the Pro-Q Diamond signal of a protein spot with the total protein amount of the spot determined by the Sypro Ruby signal. The averages and standard errors of the mean (SEM) were calculated from six replicate gels (2 replicate gels per sample from each of the three fermentations).

Excised gel spots were destained in a freshly prepared solution of 30 mM potassium ferricyanide and 100 mM sodium thiosulfate (1:1) and then dehydrated in acetonitrile (ACN). For protein alkylation, the dried gel pieces were incubated in 5 mM tris (2-carboxyethyl) phosphine and 55 mM iodoacetamide (Sigma, USA) in dark at room temperature for 1 h, followed by washing with 100 mM ammonium hydrogen carbonate (NH₄HCO₃) and dehydration with ACN. The gel pieces were rehydrated in 10 to 15 μl 10 mM NH₄HCO₃, 10% (v/v) ACN containing 0.01 μg/μl sequencing grade trypsin (Promega, USA), and digested overnight at 37°C. To elute the tryptic peptides from the gel, the pieces were incubated twice in 150 μl 66% ACN (v/v) and 0.1% (v/v) trifluoroacetic acid (TFA; Fluka) for 30 min at 37°C. The peptide eluents were dried to a minimal amount of liquid in a SpeedVac, and suspended in 5 μl of 0.1% (v/v) TFA.

For co-crystallisation of peptides with the matrix, an aliquot of the peptide solution was mixed with a saturated matrix solution in a ratio of 1:10, and 1 to 2 μl was dispensed on a MALDI target and let to dry at room temperature. The matrix solution used was prepared by dissolving 17 mg α-cyano-4-hydroxycinnamic acid (Sigma) in 1 ml of 33% (v/v) ACN and 0.1% (v/v) TFA.

MALDI-TOF analyses were carried out with Autoflex (Bruker Daltonics, Bremen, Germany), equipped with a nitrogen pulsed laser (337 nm), operated in positive mode. Typically, mass spectra were acquired by accumulating spectra of 400 laser shots. External calibration was performed for molecular assignments using a peptide calibration standard (Bruker Daltonik GmbH, Leipzig, Germany).

Protein identifications were performed by searching the peptide masses against the National Center for Biotechnology Information (NCBI) non-redundant database using Prowl's ProFound – Peptide Mapping (Rockefeller University) 35, Protein Prospector ms-fit (University of California, San Francisco) 36 and Matrix Science's Mascot – Peptide Mass Fingerprint (Matrix Science Ltd, UK) 37. Protein identifications by peptide mass fingerprinting were further evaluated by comparing the calculated and observed molecular masses and pI-values, as well as the number of peptides matched and the percentage of sequence coverage.

Three aerobic batch fermentations were carried out both on 50 g l^-1 glucose and on 50 g l^-1 xylose to compare the yeast transcriptome and proteome of cells growing on xylose with that of glucose repressed and glucose derepressed cells. Samples of the xylose-grown cells were harvested at 72 h from the start of the xylose cultures with 32 ± 1 g l^-1 of residual xylose present. Samples of the glucose repressed cells were harvested at 5 h from the start of the glucose cultures with 37 ± 2 g l^-1 of residual glucose present. Samples of the glucose derepressed cells were harvested at 24 h from the start of the glucose cultures containing no glucose but 13 ± 1 g l^-1 of accumulated ethanol.

The volumetric consumption and production profiles of xylose, glucose, biomass, xylitol and ethanol (g l^-1) of the cultures are shown in Figure 1. In addition, some acetate and glycerol (less than 2 g l^-1 each) were produced in the glucose cultures, and less than 4 g l^-1acetate, 3 g l^-1 glycerol and 2 g l^-1 ethanol were measured in the xylose cultures. The doubling time of the cells in the xylose cultures was 28 ± 2 h (during 5 to 72 h), and in the glucose cultures 1.8 ± 0.1 h (during 2 to 5 h) and 73 ± 3 h (during 24 to 53 h).

The 1439 genes, which were expressed differently (based on ANOVA at a p-value 0.01) between the glucose repressed (Glc5h), glucose derepressed (Glc24h) and xylose (Xyl72h) samples were organised into eight separate clusters with distinct expression profiles shown in Figure 2. The top three gene ontology classes of each cluster and the number of genes found in each class are listed in Table 1. All genes in each cluster of Figure 2 are shown in additional files [Additional files 6, 7, 8, 9, 10, 11, 12, 13].

70 out of 547 protein spots separated on the 2-DE gels had different abundances between the glucose repressed (Glc5h), glucose derepressed (Glc24h) and xylose (Xyl72h) samples and were identified by MALDI-TOF peptide mass mapping. They represented 55 different proteins, with 12 proteins present as multiple protein spots on the gels (Table 2) [Additional files 14 and 15]. 73% of the responses at the protein level correlated qualitatively with the microarray data (Table 2).

One of the first responses of S. cerevisiae to glucose is the induction of genes encoding hexose transporters (Hxt) that are also used for xylose uptake 3839. The signal for the presence of glucose is mediated via the glucose sensor proteins Snf3p and Rgt2p, responding to low and high glucose concentrations, respectively 4041. The expression of SNF3 is repressed at high concentrations of glucose via Mig1p (42 and references therein). In the xylose-grown cells, the transcript levels of SNF3 were higher than in the glucose repressed cells, but lower than in the glucose derepressed cells (Fig. 2, cluster 2), while RGT2 had its highest expression in the xylose-grown cells (Fig. 2, cluster 4). The expression of RGT2 is reported to be independent of glucose concentration 43, but in the present study derepression was observed (Fig. 2, cluster 4).

RGT1, encoding a transcription factor of HXT genes had, like SNF3, higher transcript levels in the xylose-grown cells than in the glucose repressed cells, but lower than in the glucose derepressed cells (Fig. 2, cluster 2). The expression of RGT1 is reported not to be regulated in response to glucose (42 and references therein), but the expression profile in cluster 2 suggests that its expression was derepressed in the absence of glucose, and to a lesser extent also on xylose (Fig. 2). Consistently, HXT2 and HXT3, which are repressed by RGT1, had highest expression in the presence of glucose and lowest in the glucose derepressed cells (Fig. 2, cluster 1). In addition to its function in repressing HXT genes in the absence of glucose 44, Rgt1p is required for repression of HXK2 at low levels of glucose together with the transcription cofactor Med8p 45. The expression of RGT1 and HXK2 was as expected: while RGT1 was expressed at its highest level in the glucose derepressed cells, HXK2 had its lowest expression in these cells (Fig. 2, cluster 1). MTH1 encoding a corepressor of RGT1 had its highest expression in the xylose-grown cells (Fig. 2, cluster 4). Mth1p is involved in maintaining the repression of HXT genes when glucose is not present 46. Nevertheless, in the cells grown on xylose, HXT2 and HXT3 had higher transcript levels compared with the glucose derepressed cells (Fig. 2, cluster 1), HXT6 compared with the glucose repressed cells (Fig. 2, cluster 2), and HXT4 and HXT16 compared with both the glucose repressed and derepressed cells (Fig. 2, clusters 7 and 4, respectively).

In xylose-grown cells, the genes encoding activities of the TCA and glyoxylate cycles, respiration and gluconeogenesis were expressed more strongly than in glucose repressed cells but less strongly than in derepressed cells (Fig. 2, cluster 2). On glucose these genes are repressed via the pathway involving Snf1p, Mig1p and Hxk2p 47. In more detail, cluster 2 contained genes encoding enzymes of respiration (e.g. COX4, COX5a, COX7, COX14-15, COR1, QCR2, QCR7, QCR9, ATP3-5, ATP16), gluconeogenesis (e.g. FBP1, PCK1), the TCA and glyoxylate cycles (e.g. KGD1-2, SDH1-4, FUM1, MDH1, CIT1-3, MLS1, ICL1), alcohol catabolism (e.g. ADH2), and trehalose and glycogen synthesis (e.g. TSL1, GLC1-3, GSY2) (Table 1). In addition, genes encoding transcriptional regulators of the aforementioned genes, such as ADR1, CAT8, HAP4, SIP1-2 and 4, and REG2 48 were present in cluster 2 (Fig. 2). In the proteome analysis of the corresponding samples, a similar trend was seen in the abundance of proteins with respiratory functions, such as ATP synthases Atp1p and Atp2p, ubiquinol-cytochrome-c reductase (Cor1p; two isoforms detected), isocitrate dehydrogenase Idh2p (expression of IDH2 was comparable in the glucose repressed and xylose-grown cells; Fig. 2, cluster 8), and the α-subunit of succinyl-CoA ligase Lsc1p (Table 2) [Additional file 15].

Distinct profiles in the abundance of transcripts and enzyme levels of the glycolytic and ethanol pathways were observed. HXK2 and PGI1 had lower expression in the xylose-grown cells compared with the glucose repressed cells (Fig. 2, cluster 1). On the other hand, the xylose-grown cells had higher amounts of Pgk1p, Gpm1p and Adh1p, compared with the glucose repressed cells (Table 2) [Additional file 15]. In addition, Eno1p and Adh2p (whose encoding gene had higher transcript abundance on xylose compared with the glucose repressed cells; Fig. 2, cluster 2) were most abundant in the xylose-grown cells (Table 2) [Additional file 15]. The latter two proteins are typically synthesised on non-fermentable carbon sources 4950.

Overlaying the comparison of gene expression in the glucose repressed, derepressed and xylose-grown cells on the main metabolic network (Fig. 3) further demonstrates how the genes encoding activities of the TCA and glyoxylate cycles had higher expression profiles on xylose compared to the glucose repressed cells, whereas the opposite was observed with many genes of glycolysis and the PPP. The differences indicated in the expression of all individual genes in Figure 3 were not statistically significant based on ANOVA, however, the analysis clearly illustrates the distinct trends in the expression of genes of individual pathways in the central carbon metabolism on xylose compared to the glucose repressed and derepressed cells.

In addition to Snf3p and Rgt2p, extracellular glucose is sensed via Gpr1p, which signals the presence of glucose via Gpa2p to the cyclic AMP-protein kinase A (cAMP-PKA) signalling pathway 51. The glucose-induced activation of cAMP synthesis is subsequently repressed by glucose and therefore not considered to be operative during growth on glucose 52. Most genes encoding components of the cAMP-PKA pathway had intermediate expression in the xylose-grown cells with highest expression in the glucose derepressed cells and lowest in the glucose repressed cells [Additional file 16]. Consistently, 52% of the genes (28/54 genes) repressed via this pathway in the presence of glucose had the same intermediate expression profile in the xylose-grown cells (data not shown).

HXK1, HXT16 and SUC2, which are all repressed by glucose via the Snf1p/Mig1p-pathway 42535455, had their highest expression levels in the xylose-grown cells (Fig. 2, cluster 4). MAL11, MAL31, and MAL32, encoding activities for the utilisation of maltose and representing classical prototype genes of glucose repression by Mig1p, were also relieved from glucose repression (Fig. 2, cluster 6). This was observed despite the fact that GAL83 encoding one of the β-subunits of the Snf1 kinase complex and enabling nuclear localisation of the kinase in the presence of non-fermentable carbon sources 56 had its lowest expression in the cells grown on xylose (Fig. 2, cluster 3), as if xylose was perceived as a fermentable, repressive carbon source.

The metabolic pathway responsible for the cytoplasmic synthesis of acetyl-CoA from ethanol involves alcohol dehydrogenase 2, cytoplasmic aldehyde dehydrogenase 6, and acetyl-CoA synthase encoded by ADH2, ALD6 and ACS1, respectively. ADH2 and ACS1, and the genes encoding the two transcription factors, Adr1p and Cat8p, which regulate their expression 57 had higher transcript levels on xylose compared to the glucose repressed cells (Fig. 2, cluster 2). However, expression of ALD6 was lower on xylose compared with the glucose repressed and derepressed cells (Fig. 2, cluster 5), as was the abundance of two isoforms of Ald6p in the proteome analysis (Table 2) [Additional file 15]. ALD5 encoding a mitochondrial, NADPH-dependent isoform of acetaldehyde dehydrogenase 58, was up-regulated on xylose (Fig. 2, cluster 7). In the proteome analysis, Ald4p, another mitochondrial, NADPH-dependent isoform, had its highest abundance on xylose (Table 2) [Additional file 15]. The expression of glucose repressed ACH1, encoding a mitochondrial acetyl-CoA hydrolase 59, had higher transcription on xylose compared to the glucose repressed cells (Fig. 2, cluster 2), which was also reflected in the abundance of Ach1p (Table 2) [Additional file 15].

Interestingly, several glycolytic enzymes (i.e. Hxk2p, Pgk1p, Eno1p, Eno2p) had post-translationally modified forms with different abundances in the xylose-grown and glucose repressed and derepressed cells (Table 2) [Additional files 14, 15 and 17]. The fluorescent phosphoprotein-specific Pro-Q Diamond stain revealed reproducible, distinct patterns of phosphorylated forms of Hxk2p, Glk1p, Eno1p and Eno2p that were present in different relative quantities in the three different conditions studied (normalised to the total amount of each protein; Fig. 4) [Additional file 17]. Both Eno1p and Eno2p had three phosphorylated pI forms with different abundances in the xylose- and glucose grown cells (Fig. 4). The expression of ENO1 was expected to be repressed by glucose, but both ENO1 and ENO2 should be expressed during growth on non-fermentable carbon sources 49. Two phosphorylated isoforms were identified for Glk1p and three for Hxk2p (Fig. 4). GLK1 should be expressed during growth on non-fermentable carbon sources and Hxk2p should be dominant during growth on glucose 55. The expression of ENO1 and GLK1 did not significantly change between the three samples studied.