Transcriptional Responses of Saccharomyces cerevisiae to Shift from Respiratory and Respirofermentative to Fully Fermentative Metabolism

Background

Adaptation to oxygen availability is crucial for all organisms. Some organisms survive only in oxygen concentrations normally present in the atmosphere, whereas for some, oxygen is toxic. Saccharomyces cerevisiae is a facultative anaerobe, able to grow in a wide range of oxygen concentrations, including complete anaerobiosis. In the presence of oxygen, S. cerevisiae utilizes respiration for efficient energy production, whereas in anaerobic conditions, S. cerevisiae has a high redox neutral carbon flux to the fermentative pathway, and energy is produced by substrate level phosphorylation. S. cerevisiae, a Crabtree-positive yeast, is able to ferment glucose even in fully aerobic conditions, allowing a high rate of sugar utilization. The ability for mixed respirofermentative metabolism in aerobic conditions and the ability to adapt to different oxygen concentrations have facilitated wide industrial applications of S. cerevisiae. However, in large-scale fermentations, especially in those of high-cell-density, provision of uniform aeration is problematic, and spatial and transient perturbations in oxygen availability may occur. Thus, the analysis of dynamic, oxygen-dependent regulation in S. cerevisiae is important not only to understand general adaptation mechanisms of living cells to various effects of oxygen, but also to understand how variations in oxygen supply may affect production processes.

Expression levels of a large number of genes and proteins differ, depending on oxygen availability, in S. cerevisiae (Bruckmann et al., 2009; de Groot et al., 2007; Kwast et al., 2002; Lai et al., 2005, 2006; ter Linde et al., 1999). In batch cultures on galactose, exposure to anoxia leads to an acute and transient upregulation of Msn2p/Msn4p-regulated genes of reserve energy metabolism and catabolic pathways and to downregulation of genes involved in cell cycle and rRNA processing (Lai et al., 2005, 2006). In batch cultures on glucose, similar transient responses to lack of oxygen are not observed (Lai et al., 2005, 2006). The stress response in galactose-grown cells has been suggested to arise from cessation of respiration, which is less significant on glucose because respiratory activity is already low under glucose repression (Lai et al., 2005). In accordance with this hypothesis, it was recently shown that exposure of galactose-grown cells to antimycin A, an inhibitor of the respiratory chain, leads to a similar transient transcriptomic response as anoxia (Lai et al., 2008). On galactose and glycerol (Dirmeier et al., 2002; Guzy et al., 2007), but not on glucose (Guzy et al., 2007), exposure to anaerobic conditions also leads to a transient oxidative stress response. This is seen as increased level of carbonylation of mitochondrial and cytosolic proteins, accumulation of 8-hydroxy-2′-deoxyguanosine in the mitochondrial and nuclear DNA, and as increased expression of SOD1 (Dirmeier et al., 2002) and elevated ROS (reactive oxygen species) levels (Guzy et al., 2007). When exposed to anoxia on acetate, a completely nonfermentable carbon source, S. cerevisiae enters a reversible state of suspended animation, in which sporulation and growth halt for the duration of anoxia but recommence when oxygen availability is restored (Chan and Roth, 2008).

Depletion of oxygen leads to transcriptional changes in S. cerevisiae that enable growth in severely oxygen-limited and anaerobic conditions. Some of these changes are mediated by depletion of heme and sterols, the syntheses of which are strictly aerobic processes (Andreasen and Stier, 1953; Hon et al., 2003; Kwast et al., 2002; Smith et al., 1996). Heme and sterol levels decline by dilution during growth in the absence of oxygen, and thus, the changes mediated by them are slow to occur. As heme levels decline, heme-dependent Hap1p and Hap2/3/4/5p transcription factors, which activate many genes needed in aerobic conditions, become deactivated. The deactivation of Hap1p leads, in turn, to deactivation of Rox1p, which in aerobic conditions represses the genes required during severe hypoxia and anaerobic conditions (Becerra et al., 2002; Ha et al., 1996; Kwast et al., 1998; Lowry and Zitomer, 1984; ter Linde and Steensma, 2002). Components of the respiratory chain have also been shown to be involved in the induction of specific hypoxic genes (David and Poyton, 2005; Guzy et al., 2007; Kwast et al., 1999). Further, as a response to anoxia, the cell wall and plasma membrane of S. cerevisiae are remodeled for import of sterols and unsaturated fatty acids (Kwast et al., 2002). Several genes encoding cell wall and plasma membrane proteins are differentially expressed in aerobic and anaerobic conditions, and the transcription factors Upc2p, Ecm22p, and Sut1p are known to play a role in the import of sterols, but the exact mechanism of the cell wall remodelling is not known (Abramova et al., 2001; Alimardani et al., 2004; Bourot and Karst, 1995; Lewis et al., 1988; Shianna et al., 2001; Snoek and Steensma, 2007).

To our knowledge, there are no previous studies on genome-wide transcriptional adaptation of fully respiratory cultures of S. cerevisiae to purely fermentative growth under anaerobic conditions. Furthermore, the adaptation of respirofermentative S. cerevisiae cultures to anaerobic conditions has previously been studied only under carbon catabolite repression in batch cultures on glucose and galactose (Lai et al., 2005, 2006). We have previously shown that in glucose-limited chemostat cultivations at low growth rate (D = 0.1 h⁻¹) S. cerevisiae grows in respiratory, respirofermentative, and fermentative metabolic modes, depending solely on the oxygen provision under the derepressed conditions (Wiebe et al., 2008). Providing 20.9% of oxygen in the chemostat inlet gas enabled fully respiratory growth, while providing 2.8, 1.0, or 0.5% oxygen led to respirofermentative growth. In the current study we initially provided glucose-limited chemostat cultures (D = 0.1 h⁻¹) of S. cerevisiae with 20.9% or 1.0% oxygen in the inlet gas and monitored the time-dependent, genome-wide transcriptional adaptation to anaerobic conditions. The experimental setup allowed the direct assessment of adaptation to anaerobicity, without the interference of carbon catabolite repression or use of different carbon sources. In other words, we were able to study the transcriptional adaptation of fully respiratory and respirofermentative cultures of S. cerevisiae to fully fermentative metabolism, and analyze whether the metabolic processes needed for the adaptation varied depending on the initial metabolic state.

Materials and Methods

Strain and culture conditions

Saccharomyces cerevisiae CEN.PK113-1A (MATα, URA3, HIS3, LEU2, TRP1, MAL2-8c, SUC2) was grown in 0.8 to 1 L medium in B. Braun Biotech International (Sartorius) Biostat^® CT (2.5 L working volume) bioreactors in the defined minimal medium described by Verduyn et al. (1992), with 10 g glucose L⁻¹ as carbon source, and supplemented with 10 mg ergosterol L⁻¹ and 420 mg Tween 80 L⁻¹. BDH silicone antifoam (BDH 331512K, VWR International, UK; 0.5 mL L⁻¹) was used to prevent foam production in the cultures. Chemostat cultures were maintained at D = 0.10 ± 0.02 h⁻¹, pH 5.0, 30°C, with 1.5 volume gas [volume culture]⁻¹ min⁻¹ (vvm).

Duplicate glucose-limited chemostat cultures were carried out with 20.9% (fully aerobic) or 1.0% (oxygen-limited) oxygen in the inlet gas. For cultures that received 1.0% O₂ in the gas stream, air was replaced with the equivalent volume of N₂, so that total gas flow was maintained constant. Cultures which were fed 20.9% O₂ were subject to oscillations. To prevent these, approximately 5% of the total cell concentration in the bioreactor was added to the culture as cells in mid to late exponential phase at the time when continuous medium feed was started (Zamamiri et al., 2001).

After steady states were established, the cultures were made anaerobic by replacing air (20.9 or 1.0% O₂) with 100% N₂. Samples were removed at intervals (0.2, 1, 3, 8, 24, and 79 (20.9% O₂) or 72 (1.0% O₂) h after the switch to 100% N₂) until a new steady state was achieved. These cultivations have been previously described in Wiebe et al. (2008) and biomass and metabolite analyses of the cultures were described in Wiebe et al. (2008), whereas the role of the glucose transporters during the transition was described in Rintala et al. (2008) and the transcriptomes of the initial steady states were described in Rintala et al. (2009).

Transcriptome analysis

Affymetrix microarray analysis of duplicate experiments for each of the two initial oxygen levels was performed. For the microarray analysis, the cells were collected by centrifugation (3,500 rpm, 5 min, +4°C). After centrifugation, the cell pellet was frozen in liquid nitrogen. For RNA extraction, 5–20 mg dry mass of cells were suspended in 400 μL cold (+4°C) disruption buffer (20 mM Tris-HCl, pH 7.4, 100 mM KCl, 2 mM MgCl₂, 2 mM DTT). A total 400 μL phenol-chlorophorm (50:50), 5 μL 20% (w/v) SDS and 400 μL glass beads (0.5 mm diameter; Biospec Products) were added. The cells were disrupted with a Fastprep machine (Q-Biogene), 2 × 20 s, at speed 6. After centrifugation (14,000 rpm, 15 min, +4°C), supernatant was used for total RNA extraction, using an RNeasy kit (Qiagen, Chatsworth, CA) according to the manufacturer's instructions.

Hybridizations were carried out at the Finnish DNA Microarray Center at the Turku Center for Biotechnology. A total of 2 μg total RNA was used as starting material for sample preparation. Samples were processed according to the One-Cycle Target Labeling protocol in the GeneChip Expression Analysis Manual (Affymetrix). Both before and after the amplifications the total RNA/cRNA concentrations were assessed with Nanodrop ND-1000 and the total RNA/cRNA quality was assessed using a BioRad Experion electrophoresis station. Each sample was hybridized to the GeneChip Yeast Genome 2.0 Array at +45°C overnight (16h) according to the GeneChip Expression Analysis Technical Manual (Affymetrix, Santa Clara, CA). A GeneChip Fluidics Station 450 was used to wash and stain the arrays, and a GeneChip Scanner 3000 with AutoLoader was used to scan the arrays. CEL-files were extracted with GCOS Manager 1.4.

All data analysis was done using R/Bioconductor, version 2.5.1 (R Development Core Team, 2005; Warnes, 2006). The raw data was normalized with Robust Multichip Average (RMA) normalization (Irizarry et al., 2003). Statistical differences in the expression were analyzed using linear modeling with the tools of limma package (Smyth, 2005). For each gene, a linear model was fitted by the least squares method and differential expression within pairs of experimental conditions was computed using an empirical Bayesian approach (Smyth, 2004). For correction of multiple testing errors, the Benjamini & Hochberg method controlling false discovery rate (FDR) was used (Benjamini and Hochberg, 1995). The microarray data can be accessed through GEO accession number GSE22832.

The clustering analysis of gene expression data was performed using fuzzy c-means clustering (Futschik, 2007; Futschik and Carlisle, 2005). The clustering method assigns genes to clusters with gradual membership (values between 0 and 1). For the clustering, the expression values were scaled and centred to have a mean of zero and standard deviation of 1. The parameter m, which controls the sensitivity of the clustering process to noise, was adjusted to 1.25 to prevent the detection of clusters in randomized data. The number of clusters was selected, such that no clusters were formed where all the genes would have membership values below 0.5. The enriched GO classes and KEGG metabolic pathways in the clusters were computed with the GOstats package (Falcon and Gentleman, 2007). Transcriptional regulatory motifs in the clusters were analyzed with the FIRE method (Elemento et al., 2007).

Reporter Features and Reporter Metabolites, the transcriptional regulatory sites in biochemical interaction networks, were identified with the Reporter Features and Reporter Metabolites algorithms (Oliveira et al., 2008; Patil and Nielsen, 2005). Reporter Features, here transcription factors and other regulatory proteins, were identified from the network of all known interactions between the regulators and genes by a multidimensional analysis. Absolute Pearson correlation coefficients were calculated for the time series of the normalized expression levels of all the pairs of genes connected to a regulator. Scores for the regulators were derived from the correlations of the connected genes. To obtain p-values for identification of the Reporters, the scores were tested for the null hypothesis “the regulator score is observed by chance.” Reporter Metabolites were identified from a genome-wide metabolic network of S. cerevisiae by differential analysis of the sequential time points. The scores for the metabolites were derived from the p-values for statistical significances of the differential expression of the genes connected to the metabolites in a genome-wide metabolic network between the sequential time-points as explained in detail in Patil and Nielsen (2005). The p-values to identify the Reporter Metabolites were obtained from a statistical test similarly as above.

Results

Gene transcription after transition from fully aerobic or oxygen-limited to anaerobic conditions

Statistical analysis of microarray data obtained from samples taken 0.2 to 79/72 h after a switch from fully aerobic (20.9% inlet O₂) or oxygen-limited (1.0% inlet O₂) glucose-limited chemostat cultures of S. cerevisiae revealed 3811 (initially fully aerobic) and 3,701 (initially oxygen-limited) genes that responded to the change in oxygen concentration (p < 0.01). Of these genes, 3022 were common to both starting conditions.

In the cultures that were initially oxygen-limited, the switch to nitrogen feed resulted in a change in expression of 2,287 genes within 0.2 h, whereas in the initially fully aerobic cultures, expression of only 1,320 changed during 0.2 h, increasing to 2,374 genes after 1 h (Fig. 1A and B). In both experiments, most of the genes whose expression had changed in 3 to 8 h had already started to change before 3 h. The lowest number of responsive genes was observed at 8 h after the cultures became anaerobic, and after 8 h a new set of genes responded. After 24 h, there were 597 and 707 genes differentially expressed, compared to the initial steady state, that had not shown earlier changes in expression in the cultures initially oxygen-limited and fully aerobic, respectively. Gene expression in the final anaerobic steady state of cultures that had been either initially fully aerobic or oxygen-limited, differed for only 3 genes (p-value <0.05) and was similar to that observed previously (Rintala et al., 2009).

The expression patterns of the significantly changing genes in the two experiments were compared using Pearson correlation (Fig. 1C). Although oxygen-limited cultures exhibited a greater initial response to anaerobic conditions than fully aerobic cultures, 1,169 genes had a correlation of >0.9 in their expression patterns. The expression patterns of genes that had previously been analyzed with the TRAC method (Wiebe et al., 2008) were compared with the patterns observed in the Affymetrix analysis of the present study. Of the 67 genes analyzed in both studies only 48 had significant changes in their expression levels. Of these, 41 showed good correlation (>0.6) between the two methods following the shift from fully aerobic to anaerobic conditions. Of the seven genes with correlation <0.6, five had less than twofold differences in their expression levels. The ones that had more than twofold differences in their expression, but low correlation between the TRAC and the Affymetrix methods, were CIT2 and CIT3.

Clustering and analysis of transcriptional regulation of the clusters

The data was clustered using fuzzy c-means clustering (Futschik and Carlisle, 2005). This clustering method does not require prefiltering of genes and thus does not discard potentially interesting genes that do not respond strongly. In addition, all genes are assigned to clusters with a membership value between zero and one, which can be used to determine the level of coregulation. The resulting clustering of gene expression data is presented in Figures 2 and 3, and the most significantly (p < 0.01) overrepresented GO categories and KEGG pathways in the clusters are presented in additional data files: Supplementary Table S1 and Table S2.

Analysis of the gene expression data of the experiment where 20.9% oxygen in the feed gas was replaced with nitrogen revealed 24 clusters containing 70–294 genes with membership value higher than 0.5, that is, belonging most strongly to the corresponding cluster (Fig. 2). The promoter sequences and 3′UTRs of the genes in these clusters were analyzed using FIRE software (Elemento et al., 2007) (see additional data file: Supplementary Fig. S1). The analysis revealed eight transcription factor binding site motifs and four 3′UTR motifs, some of which had significant co-occurrence and/or spatial colocalization patterns. The gene expression data of cultures in which 1.0% oxygen in the feed gas was replaced with nitrogen revealed 22 clusters containing 71–372 genes with membership value higher than 0.5 (Fig. 3). Analysis of the promoter sequences and 3′UTRs of genes in these clusters using FIRE software (see additional data file: Supplementary Fig. S2) revealed 14 transcription factor binding site motifs and 8 3′UTR motifs, some of which had significant co-occurrence and/or colocalization patterns. A more detailed description of the clustering results for data obtained from these cultures, and of the analysis of the regulatory elements identified in the genes of these clusters, is provided below.

Transient downregulation upon oxygen depletion: processes related to growth (replication, transcription, translation) and mitochondrial function

In the initially fully aerobic cultures, genes in clusters 4, 13, 20, and 22 were transiently downregulated between 0.2 and 8 h (Fig. 2). These clusters differed in the time at which the expression was at its lowest, which was 0.2 h (cluster 4), 1 h (clusters 20 and 22), or 3 h (cluster 13). GO categories enriched in clusters 4, 20, and 22 were largely overlapping and particularly included genes involved in amino acid and purine metabolism, ribosomal biogenesis, RNA processing, biogenesis of RNA polymerases and genes related to cell cycle and DNA replication and repair (Supplementary Table S2). Genes encoding components of cytosolic ribosomes were statistically enriched in clusters 20 and 22, and seven of these genes were found also in cluster 4. The transcription factor binding site motifs enriched in these clusters also overlapped (Supplementary Fig. S1). The PAC site, related to the regulation of the genes encoding components of ribosome biogenesis, was overrepresented in clusters 4, 20, and 22, the XBP1 site, related to the regulation of genes encoding activities of stress response, in clusters 20 and 22, and the RAP1 site, related to the regulation of the genes encoding components of ribosome biogenesis, in cluster 22. In cluster 4, a putative 3′UTR motif [(A/U/G)(G/U)AUAGA], was enriched; however, this motif was not enriched in clusters 20 and 22, being, in fact, underrepresented in cluster 22. Instead, genes in clusters 20 and 22 were enriched in another putative 3′UTR motif [UAUA(A/C)(G/U)A]. As with clusters 4, 13, 20, and 22, cluster 6, containing genes that were downregulated at 0.2 h but that had fully recovered at 1 h, contained genes involved in mRNA processing and splicing and in rRNA processing. A total of 57% of the genes in cluster 6 contained the same 3′UTR motif [(A/U/G)(G/U)AUAGA] that was overrepresented in cluster 4. In the whole genome, this site is enriched in 3′UTRs of genes involved in RNA processing and ribosome biogenesis.

In the cultures that were initially oxygen-limited, clusters 3, 20, and 21, which were transiently downregulated between 0.2 and 3 h, were enriched in genes related to ribosome biogenesis and assembly, rRNA processing, amino acid, and purine metabolism, and DNA replication and repair (Fig. 3 and Supplementary Table S2). In addition, genes related to DNA replication and repair and cell cycle were enriched in cluster 17, which was also downregulated between 0.2 and 3 h. Although these transient responses were similar to those seen when the fully aerobic cultures became anaerobic they started slightly earlier and had recovered earlier in the cultures that had been oxygen limited (in 1 to 3 h) than in those with initial oxygen of 20.9%, in which most of these genes were still downregulated at 8 h. Genes in clusters 3 and 20 were enriched in transcription factor binding site motifs PAC and RRPE, the PUF4 3′UTR motif, involved in the regulation of genes encoding ribosomal proteins, and in a putative 3′UTR motif (AUAGA). Both of the 3′UTR motifs showed significant colocalization. In addition, genes in cluster 21 were enriched in binding sites of Xbp1p, and Rap1p transcription factors, the RRPE motif, the PUF4 3′UTR, and a putative 3′UTR (UCCGUAC) motif. Genes in cluster 17 were enriched in binding sites for Swi4p and Xbp1, transcription factors regulating genes encoding activities related to cell cycle and stress response. Similar to cluster 6 of the initially fully aerobic cultures, genes of cluster 7 in the initially oxygen-limited cultures were transiently downregulated at 0.2 h after the cultures became anaerobic and the level of transcription had fully recovered at 1 h. Cluster 7 contained genes related to amino acid metabolism, purine biosynthesis, transcription, chromatin remodeling, and splicing. This cluster was enriched in binding sites of transcription factor Bas1p, regulating genes encoding activities related to purine and histidine biosynthesis, and the PUF5 3′UTR motif, associated with mRNAs encoding nuclear components. Genes in cluster 2 were transiently downregulated 3 h and were enriched in genes related to cell cycle, DNA packaging, organelle organization and biogenesis, ribosomal protein, and rRNA transport. Genes in cluster 2 were enriched in binding sites for Swi4p transcription factor and PUF5 3′UTR motif.

In the initially fully aerobic cultures, genes in cluster 15 were temporally downregulated but showed upregulation in the anaerobic steady state. The genes in this cluster were related to mitochondrial translation (57 genes) and protein targeting to mitochondria (MAS1, TOM6, TOM20, TIM9, TIM13, POR1). The 3′UTR site PUF3, which regulates the genes encoding mitochondrial proteins, was overrepresented in this cluster. In the initially oxygen-limited cultures, cluster 4 had a trend of downregulation during 1–3 h and a full recovery at 8 h. Among the genes of this cluster, 68 and 14 genes encoded activities involved in mitochondrial translation and mitochondrial protein import, respectively. This cluster also contained genes encoding additional mitochondrial membrane proteins (SHE9, ADK2, ARH1, YGR012W, DPM1, OMS1, MSS51, YMR166C) and genes related to respiration (COQ10, DIA4, MAM33, YMR293C, COX23, ATP12, ATP11, COX17). Genes in this cluster were also enriched in the PUF3 3′UTR motif and additionally two putative 3′UTR motifs, [CC(C/U)GUA(A/U)] and [(C/U)CT(A/U)GUA]. A closer evaluation of the two putative 3′UTR motifs revealed that they always preceded the PUF3 motif and represented the first part of it.

In the initially fully aerobic cultures, the genes involved in sulphate assimilation and methionine biosynthesis (MET1, MET2, MET3, MET8, MET14, MET17, MET22, ECM17, YLL058W) and sulphate uptake (SUL2) showed downregulation at 0.2 h and had fully recovered at 1 h (cluster 6). However, other genes related to sulphur amino acid metabolism (MET32, MUP1, MET4, BDS1, MET28) were upregulated during 1–72 h (cluster 10). In the initially oxygen-limited cultures, several genes related to methionine biosynthesis and sulphate assimilation (MET10, MET17, SAM2, CYS3, MET14, MET1, OAC1, MET18) also responded by transient downregulation, but with lowest level of transcription at 1 h and full recovery only at 8 h (cluster 21).

Transient upregulation upon oxygen depletion: processes of protein degradation

Protein degradation mechanisms were transiently upregulated as a response to lack of oxygen both in the initially fully aerobic and oxygen-limited cultures. In the initially fully aerobic cultures, genes of vacuolar transport (cluster 9), proteasome and autophagy (cluster 5), and protein catabolism (cluster 3) showed transient upregulation at 0.2 and 1 h (Fig. 2 and Supplementary Table S2).

When the initially oxygen-limited cultures became anaerobic, genes of vacuolar transport (cluster 18), protein targeting to vacuole (cluster 9), vacuolar protein catabolic process (cluster 22), autophagy (clusters 9 and 18), ubiquitin-mediated proteolysis (cluster 10), and proteasome (cluster 14) showed transient upregulation during 0.2–3 h (Fig. 3 and Supplementary Table S3). In addition, clusters 15 and 16, which showed upregulation in the anaerobic steady state compared to the initial steady state, were enriched in genes of protein targeting to the vacuole and proteasome, respectively. Genes in clusters 9, 18, and 22 were enriched in binding sites of transcription factors Msn2/4p and Gis1p, regulating genes encoding activities related to stress response, whereas genes in clusters 14 and 16 were enriched in binding sites of Rpn4p, regulating genes encoding components of the proteasome. Genes in cluster 15 were enriched in a putative transcription factor binding site [TA(A/T)ACGA].

Upregulation for anaerobic steady state: cell wall components, remodelling of cell wall, sterol and amino acid uptake, and iron/cation homeostasis

In the initially fully aerobic cultures, genes involved in sterol transport (DAN1-3 TIR1-4, PAU2,3,4,5,8,9,14,18, OSH6, UPC2, PDR11, HES1, AUS1) and genes related to cation homeostasis and iron transport (FRE6, FET4, ENB1, CUP5, ATX2, COT1, TIS11, IZH3, SMF3, IRC7, SIT1, FIT2, ARN1) were upregulated during 0.2–8 h after the change in oxygen provision and reached their new steady-state values, with higher expression levels, within 24 h (cluster 19). In the initially oxygen-limited cultures, genes related to sterol uptake and biosynthesis (DAN1–3, TIR1–4, PAU8,9,14,2 OSH6, UPC2, ERG8, ERG26, ERG7, ERG2, PDR11, DAP1, ARE1, SUT2, NCP1, ERG9, ERG27, ERG24, HES1, ERG28, AUS1, KES1) and cation homeostasis (IRC7, SMF3, IZH4, VMA2, SRO77, COT1, CUP5) remained unchanged for the first 3 h after the change of oxygen provision, but were upregulated to their new anaerobic steady-state values during 8 to 24 h (cluster 1).

Genes encoding plasma membrane phosphate transporters (PHO86, PHO84, PHO89) were transiently upregulated 0.2 h after the switch to nitrogen in the initially oxygen-limited cultures (cluster 19). PHO84 was 30-fold, PHO89 fourfold, and PHO86 1.5-fold upregulated at 0.2 h. Although expression of PHO84 and PHO89 returned to their original levels within 3 h, and PHO84 was twofold downregulated in the anaerobic steady state, PHO86 remained 1.5-fold upregulated in the anaerobic steady state. Although not found in any of the clusters, PHO84 and PHO89 were also transiently upregulated during the first 3 h after the fully aerobic cultures became anaerobic and subsequently downregulated in the anaerobic steady state.

Genes encoding amino acid transporters (BAP3, GNP1, DIP5, TAT1) were highly upregulated during the adaptation to anaerobic conditions and in the anaerobic steady state in the initially fully aerobic cultures (cluster 7). A similar but weaker trend was observed in the initially oxygen-limited cultures. Several other genes involved in nitrogen transport showed transient upregulation, while the proline permease (encoded by PUT4) was downregulated during the adaptation and in the anaerobic steady state (additional data file: Supplementary Fig. S3).

Downregulation for anaerobic steady state: fatty acid oxidation, peroxisome biogenesis, oxidative phosphorylation, TCA cycle, and PPP

In the initially fully aerobic cultures, genes related to fatty acid oxidation and peroxisomal biogenesis in cluster 1 were downregulated within 24 h (PEX19, CRC1, PEX2, PEX8, PEX3, PEX18), whereas genes related to fatty acid oxidation and peroxisomal biogenesis (FOX2, POT1, PCD1, PXA2, PIP2, IDP3, PXA1, POX1, PEX11, SPS19, DCI1, ECI1, PXA2, PEX14, PEX5, PEX11, PCS60) and response to oxidative stress (SOD2, POS5, UBA, MCR1, CTT1, CTA1) in cluster 24 showed downregulation already at 3 h. In the initially oxygen-limited cultures, genes in clusters 13 and 18 were downregulated as the cells approached the anaerobic steady state, but the genes in cluster 18 were transiently upregulated before the final downregulation. Cluster 18 contained genes of fatty acid oxidation and peroxisomal biogenesis (POT1, PXA2, PIP2, CRC1, OAF1, PXA1, SPS19, MDH3, ECI1, PEX27, PEX15, PEX30, PEX2, PEX22, PEX3), genes related to response to oxidative stress (HYR1, CCP1, SRX1, GAD1, PRX1, GPX1, TRR2, ORF YCL033C, HSP12), genes of oxidative phosphorylation (NDE2, ATP4, GSM1, ATP3, SDH2), genes of the TCA cycle (KGD1, LPD1, ACO1, IDP2), and the pentose phosphate pathway (SOL4, GND2, TKL2). Cluster 13 contained genes of fatty acid oxidation (FOX2, CTA1, POX1, TES1, PSC60, DCI1, INP1), oxidative phosphorylation (ATP19, QCR9, ATP2, QCR8, QCR5, COX12, COX6, QCR2, ATP20, SDH1, SDH4, COX9, ATP14, QCR10, ATP16, COX8, COX4, COX5a, COR1, NDI1, ATP15, COX13, ATP18, ATP1, COX7, CYT1, QCR7, SDH3, ATP7, ATP5), the TCA cycle (MDH1, FUM1, KGD2, PYC2, IDH2, LSC2, CIT1), and genes related to response to reactive oxygen species (SOD2, SOD1, CTA1, POS5, GRX2, MCR1). Genes in cluster 13 were enriched in two putative 3′UTR motifs [(A/U)AUAUUC and A/C)UUUAU(G/U)(A/U)], and in binding sites of Ume6p transcription factor, regulating genes encoding activities related to the cell cycle, and a putative transcription factor [A(A/T)C(C/T)CCG]. Cluster 14, genes of which showed transient upregulation during 0.2 to 8 h, contained genes of glutathione metabolism (GPX2, GTO1, GTT1, GLO1, GTT2) and carnitine metabolism (YAT1, YAT2).

Transient upregulation of reserve energy metabolism

In the initially fully aerobic cultures, genes of glycogen and trehalose metabolism were transiently upregulated during 0.2 to 8 h. Cluster 9 included genes related to synthesis of glycogen (PCL6, GLG1, GSY2, GLC8) and synthesis of trehalose (TPS1), whereas cluster 12 included genes related to synthesis of glycogen (GLC3, PIG2), mobilization of glycogen (GDB1, GPH1), and mobilization of trehalose (NTH1). In the initially oxygen-limited cultures, genes related to synthesis of glycogen (GSY2), synthesis of trehalose (TPS1), mobilization of glycogen (GDB1), and mobilization of trehalose (NTH1) were transiently upregulated between 0.2 and 3 h. In addition, genes related to synthesis of glycogen (PCL6, GAC1, PCL7, REG1, PCL8, RIM11), synthesis of trehalose (TPS2, TSL1), and mobilization of trehalose (NTH2, TPS2) in cluster 9 had a trend of transient upregulation between 0.2 and 1 h. Binding sites for Msn2/4p and Gisp1 transcription factors were enriched in both clusters of oxygen-limited cultures, and in addition, binding sites for Ume6p and a putative transcription factor [A(A/T)C(C/T)CCG] were enriched in cluster 9. Cluster 18, which was transiently upregulated before final downregulation, contained genes of glycogen metabolism (BMH2, GLC3, GLG2, SGA1, GLG1, GLC8, GPH1). In this cluster, binding sites for Msn2/4p, Gisp1 and Ume6p, and a 3′UTR motif [(A/U)AUAUUC] were enriched.

Analysis of transcriptional responses in context of regulatory networks using Reporter Features analysis

We further analyzed the transcriptional response in context of the network of all known interactions between transcription factors/other regulatory proteins and genes and performed a multidimensional Reporter Features analysis (Oliveira et al., 2008; Patil and Nielsen, 2005). From the network, the analysis identified regulators (i.e., transcription factors or regulatory proteins) whose surrounding genes had expression profiles with significantly high correlation during the time course of adaptation to the anaerobic steady state. Thus, it identified transcription factors and other regulatory proteins that most probably determined the expression profiles of the sets of responding genes. Regulatory networks of the Reporter Regulators (with Reporter p-values ≤0.01) of the present data are shown in Figures 4 and 5. The open reading frames associated with the regulators are presented in Additional data files: TR21_multiD_reporters_p01.sif and TR21_multiD_reporters_p01.sif.

The regulatory network of Reporter Regulators identified for the initially fully aerobic cultures switching to anaerobicity contained 29 regulatory proteins and a total of 297 nodes of regulators and genes (Fig. 4). The regulatory network of Reporter Regulators identified for the initially oxygen-limited cultures switching to anaerobicity contained 28 regulatory proteins and a total of 327 nodes of regulators and genes (Fig. 5). The two regulatory networks shared the stress response regulators Msn2/4p, Hsf1p and Hog1p, the growth-related regulators Basp1, Rap1p, Ifh1p, Gts1p, Rsc30p, and Esa1p, and the protein degradation-related regulators Rpt6p and Snf7p. Additionally, both networks contained regulators of genes of fatty acid β-oxidation Oaf1p and Pip2p, the Upc2p regulator of genes of sterol biosynthesis, the carbon-source responsive factor Adr1p, and the Met1p regulator of methionine biosynthesis.

The Gcr1p activator of glycolytic genes, the cAMP-dependent protein kinase Tpk2, and the Snf1p and Snf4p protein kinases were identified as Reporter Regulators only for the initially fully aerobic cultures. On the other hand, the heme-activated transcription factors of the Hap2/3/4/5p complex, the Sfp1p regulator of ribosome biogenesis and the Mbp1p regulator of cell cycle were active regulators of only the initially oxygen-limited cultures. The Hap1p regulator was specific in the regulation of the initially oxygen-limited cultures, with a Reporter p-value threshold of 0.01, but with a threshold of 0.05 it was identified as an active regulator of both cultures.

Analysis of the transcriptional response in context of the metabolic network

Because the cultures studied were different in their initial metabolic state, being either fully respiratory or respirofermentative, the transcriptional response of the cultures to sudden oxygen depletion was also studied in the context of a genome-wide metabolic network with the Reporter Metabolites algorithm (Patil and Nielsen, 2005). This algorithm identified the metabolites in the metabolic network whose surrounding enzymes, that is, the genes encoding them, had significantly differential expression at different times. The identified Reporter Metabolites (threshold reporter p-value ≤0.05) in the pathways of central carbon metabolism of S. cerevisiae after sudden depletion of oxygen from fully aerobic and oxygen-limited cultures are shown in Figure 6. The open reading frames associated with the metabolites of central carbon metabolism are presented in Additional data file: central_carbon_metabolism.sif

During the first interval, 0-0.2 h, the initially oxygen-limited cultures showed Reporter Metabolites in upper glycolysis and in the pentose phosphate pathway, whereas in the initially fully aerobic cultures Reporter Metabolites were not observed in the pentose phosphate pathway until after 0.2 h. In the latter cultures, likewise between 0.2 and 1 h, Reporter Metabolites were also observed in the glyoxylate cycle. Between 1 and 3 h after the oxygen depletion, the initially fully aerobic cultures showed Reporter Metabolites in the TCA cycle. Additionally, the redox cofactor NADH was identified as a Reporter Metabolite after 24 h, when the anaerobic steady state was established, independent of the initial metabolic state, but in the initially fully aerobic cultures NADH was identified as a Reporter also in the earlier phase of adaptation, between 1 and 3 h. In the initially oxygen-limited cultures, the cofactor NADPH was identified as a Reporter Metabolite between 1 and 3 h after the switch to anaerobic conditions.

Discussion

In response to oxygen depletion, both the fully aerobic and oxygen-limited cultures showed transient downregulation, but full recovery to the level of the initial steady state of genes encoding activities related to growth and cell proliferation. As a function of time, the expression profiles of these genes differed in the two culture sets, the initially oxygen-limited cultures responding more rapidly. However, the overall response was very similar. In addition, the same transcription factors and regulators, related to growth and cell proliferation were identified in the initially fully aerobic and the oxygen-limited cultures when the results of the FIRE and Reporter analyses were combined.

The downregulation of genes encoding activities involved in these processes is likely due to the decrease observed in the specific growth rate to 0.06 h⁻¹ almost immediately after the shift to anaerobiosis (Wiebe et al., 2008). This hypothesis is supported by the earlier observation that genes related to cell cycle, DNA replication and repair, rRNA processing, and protein synthesis are transiently downregulated as a response to anoxia in batch cultivations on galactose, when the metabolism is strongly respiratory and specific growth rate is reduced as a response to anoxia, but not on glucose when the metabolism is partially fermentative, even in the presence of oxygen, and no change in specific growth rate is observed (Lai et al., 2005, 2006). However, it is interesting that although the specific growth rate was below 0.1 h⁻¹ for approximately 15 h, the level of transcription of especially the ribosomal genes had returned to that observed in the initial steady state within 3 to 8 h. The transcription of genes encoding ribosomal proteins is positively correlated to the specific growth rate in steady-state chemostat cultures (Fazio et al., 2008; Regenberg et al., 2006), but it has been suggested that these genes are regulated by the external environment rather than the specific growth rate (Levy et al., 2007). Signalling in response to environmental changes may have developed to enable a faster response than feedback regulation by metabolic pathways could mediate (Levy et al., 2007). Supporting this hypothesis, Zaman et al. (2009) recently concluded that growth-rate specific transcription in yeast generally results from cells sensing their nutritional environment.

mRNAs of genes encoding subunits of ribosomes have a half-life of 22 ± 6 min in S. cerevisiae, and the average half-life of mRNAs in yeast is 30 min (Wang et al., 2002). Thus, the two- to fivefold decrease in mRNA levels within 10 min after the oxygen depletion, observed in particular with mRNAs of genes related to ribosome biogenesis and RNA processing indicates active degradation of these mRNAs. The control of degradation of mRNAs involve 3′UTRs, which also have important roles in the translation and localization of mRNAs (Grzybowska et al., 2001; Lawless et al., 2009; Saint-Georges et al., 2008; Ulbricht and Olivas, 2008). In the present study, binding sites of specific 3′UTRs motifs were found to be enriched in the genes related to growth and cell proliferation. The transient downregulation profiles of genes in which these motifs were enriched suggest that the motifs may have a role in the degradation of mRNAs. In fact, the PUF4 motif, known to be involved in the decay of mRNAs of genes related to rRNA synthesis and processing and ribosomal biogenesis (Gerber et al., 2004; Grigull et al., 2004) and PUF5, associated with mRNAs encoding nuclear components (Gerber et al., 2004), were identified in clusters of the initially oxygen-limited cultures. In addition, colocalized with PUF4, a putative 3′UTR motif AUAGA was identified in the initially oxygen-limited cultures and also in the initially fully aerobic cultures as a part of a longer putative motif [A/U/G)(G/U)AUAGA]. In addition, another putative 3′UTR motif [UAUA(A/C)(G/U)A] was identifed in the initially fully aerobic cultures. The fact that different motifs were identified under the different culture conditions is intriguing, but it remains unclear if these differences reflect the small differences in the timing of the transcriptional responses under these two conditions.

In both the initially fully aerobic and oxygen-limited cultures, transient upregulation of genes encoding activities of processes of protein degradation was observed upon oxygen depletion. In the initially oxygen-limited cultures, some of these activities continued to be upregulated in the new anaerobic steady state compared to the initial steady state. Further, two regulators related to protein degradation (Rpt6p, Snf7p) (Glickman et al., 1999; Robinson et al., 1988) were identified as regulators under both conditions. Protein degradation may be related to the transiently occurring reduced specific growth rate of the cultures. However, it may also indicate a more global remodeling of cellular functions and need for novel activities upon anaerobicity. Proteolysis plays an important role in the response to stress conditions (Hilt and Wolf, 1992), and in fact, binding sites of stress response-related transcription factors (Msn2/4p, gis7p) (Martinez-Pastor et al., 1996; Zhang et al., 2009) were identified as enriched in these genes in the initially oxygen-limited cultures. Additionally, other indications of stress response were observed in the present study. Reporter Features analysis identified stress response-related regulators (Msn2/4p, Hsf1p, Hog1p) (Martinez-Pastor et al., 1996; Sorger and Pelham, 1987; Westfall et al., 2004) in both cultures as a response to oxygen depletion and among the earlier mentioned genes related to growth and cell proliferation, the stress response-related transcription factor Xbp1p (Mai and Breeden, 1997, 2000) was enriched.

Genes encoding mitochondrial membrane proteins and proteins with activities related to mitochondrial translation or protein import to mitochondria were transiently downregulated upon introduction of anoxia in all cultures. Again, the initial metabolic state made little difference to the response; the profiles of the temporal downregulation were close to identical. Among these genes, the PUF3 motif, which promotes the degradation of, and controls the localization of mRNAs encoding mitochondrial proteins (Gerber et al., 2004; Olivas and Parker, 2000; Saint-Georges et al., 2008), was enriched. Interestingly, in the initially fully aerobic cultures the transcription of these genes recovered to a higher level than in the initial steady state. The same was observed in the comparison of different steady states with varying levels of oxygen provision: the lowest expression level of these genes was observed in the fully aerobic cultures, compared to the oxygen-limited and anaerobic cultures (Rintala et al., 2009). This may be an indication of still unknown functions of mitochondria under oxygen-limited and anaerobic conditions.

When both the fully aerobic and the oxygen-limited cultures were switched to anaerobiosis, a rapid and transient downregulation of genes encoding proteins of sulphate assimilation and methionine biosynthesis occurred. The genes encoding activities for methionine biosynthesis are known to be downregulated as a response to increase in the intracellular concentration of S-adenosylmethione (AdoMet) or methionine (Thomas et al., 1989), and several transcriptional activators regulate the expression of these genes (Blaiseau et al., 1997; Kuras et al., 1996; Thomas et al., 1992). AdoMet provides activated methyl groups for phospholipid synthesis, and for protein and histone methylation, so downregulation of these genes could be related to a general transient downregulation of biosynthesis (Kaiser et al., 2006). Methionine biosynthesis is also closely linked to the synthesis of glutathione, which is the main component in the maintenance of cellular redox balance (Lopez-Mirabal and Winther, 2008), and thus the rapid and transient changes observed in the genes encoding activities of methionine biosynthesis are likely to reflect altered redox balancing as a result of oxygen depletion (Lopez-Mirabal and Winther, 2008).

Genes encoding plasma membrane phosphate transporters Pho84p and Pho89p were upregulated immediately after the shift to anaerobic conditions in both cultures. According to Gonzales et al. (2000) the intracellular phosphate and polyphosphate levels start to increase in less than 10 min after a shift to anaerobiosis and phosphate levels stay high in anaerobic, compared to aerobic conditions. This increase results from transient increase in transport of extracellular phosphate into the cell during the first 15 min after the shift to anaerobiosis. The function of this increase is not known, but regulation of glycolytic enzymes has been suggested (Gonzalez et al., 2000), and it appears to be important in metabolic regulation in response to sudden oxygen depletion. The increase in the transcription of phosphate transporter encoding genes observed in the current study was only transient and their expression was lower in the anaerobic than in the oxygen-receiving steady states.

In both the initially fully aerobic and the oxygen-limited cultures, genes related to fatty acid oxidation and peroxisomal biogenesis, response to oxidative stress, oxidative phosphorylation, TCA cycle, and the pentose phosphate pathway were downregulated in anaerobic conditions. However, some of the genes of these pathways were transiently upregulated in the initially oxygen-limited cultures, before being downregulated to the new anaerobic steady-state levels. Similar transient upregulation of genes of oxidative phosphorylation and the TCA cycle has been previously observed during adaptation to anaerobic conditions in batch cultures on galactose, but not on glucose, suggesting that the response is linked to cessation of respiration (Lai et al., 2005, 2006). Additionally, respiratory-deficient yeast cells in aerobic conditions respond to loss of oxidative phosphorylation by upregulating genes related to peroxisomal activities, including fatty acid oxidation and anaplerotic reactions, to increase supplies of acetyl-CoA and OAA (Epstein et al., 2001). As this response was not seen in the adaptation of the initially fully aerobic cultures to anaerobiosis, it seems to be specific for the transition from respirofermentative conditions. The expression of genes encoding some of these processes was also oxygen-dependent in steady-state conditions (Rintala et al., 2009), and the expression levels may have already been maximal in 20.9% oxygen, allowing no further upregulation to occur.

As in the clustering analysis, Reporter Regulators related to stress, growth, protein degradation, fatty acid catabolism, sterol biosynthesis, and carbon source regulation were identified in the adaptation to anaerobiosis, independent of the initial metabolic state of the culture. The regulatory network of the initially fully aerobic cultures specifically contained an additional regulator of glycolysis, indicative of changes upon switching from respiration to fermentation and initiating a high specific carbon flux through the glycolytic and fermentative pathways. The key metabolic regulator kinases Snf1p/Snf4p involved in several different processes (stress response, translation, lipid and glycogen biosynthesis, glucose derepression) and Tpk2p involved in the response to nutrients and stress, identified in the initially fully aerobic cultures as Reporter Regulators, suggest a coordinated transcriptional adjustment of metabolic genes.

Central carbon metabolism encompasses the major energy-generating pathways in the cell, the respiratory and the fermentative pathways, but it also generates precursors and reducing power for biosynthetic reactions. Reporter Metabolite analysis revealed that the temporally differential expression of genes encoding activities of the central carbon metabolism as a response to oxygen depletion was dependent on the initial metabolic state of the culture. In the initially oxygen-limited cultures, sudden oxygen depletion led to a rapid response around the metabolites of the upper part of glycolysis and the pentose phosphate pathway, whereas in the initially fully aerobic cultures, the hierarchical regulation of the central carbon metabolism seemed not to be the first priority because slower responses around the metabolites of the pentose phosphate pathway, and the glyoxylate and TCA cycles were observed. Changes in the expression of genes encoding enzymes producing or consuming the metabolites are likely to lead to changes in the metabolite concentrations that are variables in the metabolic regulation of the system. However, in our previous study, the effect of oxygen depletion on the concentrations of the metabolites of central carbon metabolism was observed to be generally faster than the effect on gene expresssion (Wiebe et al., 2007). Then again, the differential regulation of especially the oxidative part of the pentose phosphate pathway was seen also in the study of Wiebe et al. (2007) as the concentration of 6-phosphogluconate was found to be dependent on the initial oxygen concentration of the cultures (Wiebe et al., 2007).

The changes observed in the expression of the genes encoding the upper glycolytic enzymes are probably not reflected in the metabolite concentrations, as glycolytic flux has been shown to be mainly controlled at the posttranscriptional level (Daran-Lapujade et al., 2004; de Groot et al., 2007). Upper glycolysis is the entry point of storage carbohydrates into metabolism, and thus its metabolites as reporters may also indicate the mobilization of storage carbon due to the reduced production of energy equivalents as respiration suddenly ceases. In fact, the concentration of trehalose 6-phosphate, an intermediate of trehalose biosynthesis, was observed to be dependent on the initial oxygen concentration by Wiebe et al. (2007). In the current study, genes related to both mobilization and storage of glycogen and trehalose were found to be transiently upregulated in both conditions. This seemingly futile response could not be explained. However, the simultaneous upregulation of genes acting in mobilization and storage of glycogen and threhalose has previously been seen as a response to stress and shown to be dependent on Msn2/4p transcription factors (Hottiger et al., 1987; Parrou et al. 1997).

Conclusions

The transcriptional responses of S. cerevisiae grown under glucose-limitation in either fully aerobic or oxygen-limited conditions (resulting in respiratory and respirofermentative metabolic states, respectively) to sudden depletion of oxygen were very similar to each other. During the adaptation to anaerobic conditions, and thus to fermentative growth and energy generation, the cells responded by transient downregulation of genes related to growth and cell proliferation. Additionally, the adaptation to anaerobiosis evoked stress response-related regulatory networks independent of the initial metabolic state of the culture. To enable global remodeling of the activities needed for the new mode of growth, a transient upregulation of genes related to protein degradation was observed.

In the initially oxygen-limited cultures the shift to anaerobiosis led to specific regulation of aerobic genes by the Hap2/3/4/5p complex and to transient upregulation of genes involved in oxidative phosphorylation, TCA cycle, fatty acid oxidation, peroxisomal biogenesis, oxidative stress, and pentose phosphate pathway. As the cell senses the decrease in the oxygen concentration, it may try to enhance the use of oxygen by upregulation of genes encoding the above mentioned activities. Although a similar transient response was not observed in the cells initially grown in the fully respiratory conditions, it may have been too short to have been observed with the sampling frequency used in this study, and thus it remains unclear whether the Hap2/3/4/5p complex controls a similar response in the transition from fully aerobic to anaerobic conditions.

Acknowledgments

We thank Pirjo Tähtinen, Eila Leino, and Tarja Laakso for excellent technical assistance. Prof. Jens Nielsen provided fruitful discussions and shared his expertise in yeast systems biology. The microarray analyses were carried out at the Finnish DNA Microarray Centre at Turku Centre for Biotechnology. The financial support of Tekes, The Finnish Funding Agency for Technology and Innovation (Project numbers 40135/04 and 40537/05) and Academy of Finland (Centre of Excellence, Industrial Biotechnology 2000–2005; project number 214568, Centre of Excellence, White Biotechnology—Green Chemistry 2000–2013; grant number 118573, and SYSBIO programme 2004–2007; project number 207435) is gratefully acknowledged.

Author Disclosure Statement

The authors declare that no conflicting financial interests exist.