Introduction

The budding yeast Saccharomyces cerevisiae has been widely used as a model organism to study the proliferative response to nutrients in eukaryotic cells. Networked processes in S. cerevisiae that contribute to this response include energy metabolism, protein biosynthesis, cell cycle progression, and gene regulation. Precise coordination of these processes is a critical feature of the extraordinary capacity of yeast cells to adjust the rate of cell division rapidly in response to nutritional changes. For example, the S. cerevisiae transcriptome undergoes a dynamic reprogramming (Wang et al. 2004) within minutes of the appearance of its preferred carbon source (C-source), glucose (Polakis and Bartley 1966). Glucose addition down-regulates expression of genes involved in utilization of reserve carbohydrates, catabolism of alternative C-sources, and respiration; a nearly comprehensive wiring diagram of this glucose repression regulatory pathway (Santangelo 2006), which can also be triggered by the fermentable sugar galactose (Polakis and Bartley 1965), is now available. Glucose-dependent repression is accompanied by simultaneous up-regulation of genes required for rapid energy production, growth, and cell cycle progression; these genes encode glycolytic enzymes, ribosomal proteins, and cyclins, respectively (Deminoff et al. 2003; Santangelo 2006).

Faster cell division upon the appearance of glucose requires careful coordination of growth with cell cycle progression (Hartwell 1974). This coordination occurs at Start in the G1 phase of the cell cycle, which is the gating event regulating the commitment to begin each new round of cell division (Hartwell et al. 1974). Numerous inputs contribute to timely passage through Start, including attainment of a critical cell size (Johnston et al. 1977), ample nutrient supply (Hartwell et al. 1974), and a critical rate of protein synthesis (Popolo et al. 1982). Impaired coordination between the cell and growth cycles causes division to occur at an abnormal size. For instance, mutations that cause premature passage through Start yield smaller cells (collectively called whi mutants). Conversely, mutations that delay passage through Start yield larger cells (collectively called lge mutants). Genome-wide cell size analysis of the yeast haploid knockout collection (Jorgensen et al. 2002) has identified ∼500 genes which, when deleted, cause a whi or lge phenotype. Many of these genes encode specific cell cycle regulators such as the well-characterized G1 cyclins (Cln1, Cln2, and Cln3) that are required for the transition at Start; other cellular processes, including protein synthesis and regulation of RNA polymerase II transcription, are also well-represented (Jorgensen et al. 2002).

Gcr1, a global regulator of RNA polymerase II transcription (Menon et al. 2005; Santangelo 2006), is required for normal glucose induction of the G1 cyclin genes, CLN1-3 (Parviz and Heideman 1998; Willis et al. 2003); in the absence of GCR1, CLN transcription is reduced, glucose-grown cells are much larger than their wild type counterparts, and a greater percentage of cells are unbudded due to a G1 delay in the cell cycle (Willis et al. 2003). Transcription of glycolytic and ribosomal protein genes (RPGs) is also impaired in gcr1Δ cells (Clifton and Fraenkel 1981; Santangelo and Tornow 1990; Tornow et al. 1993; Zeng et al. 1997; Deminoff and Santangelo 2001). Thus the Gcr1 regulator appears to target each of the predominant gene classes that are up-regulated in faster dividing glucose-grown cells (see above). Control of these genes appears to represent the fulcrum of growth/cell cycle coordination. Recent work has indicated that Gcr1 also contributes to glucose repression (Turkel et al. 2003; Sasaki and Uemura 2005). Indeed, derepression of genes normally repressed by glucose is essential in the absence of Gcr1; it is not possible to isolate gcr1Δ cells that are ρ − or ρ 0 (i.e., that lack functional mitochondria; our unpublished data). Both the activator and repressor functions of Gcr1 are proposed to operate through a recently discovered nuclear-pore mediated mechanism termed reverse recruitment (Menon et al. 2005; Santangelo 2006; and our unpublished data).

We report here an analysis of the phenotypic, transcriptomic, and metabolic consequences of Gcr1 removal in the presence of various C-sources: fermentable repressing sugars (glucose and galactose), a non-repressing fermentable sugar (raffinose), and a non-fermentable C-source (pyruvate). We found that growth of gcr1Δ cells is defective in all C-sources tested. Interestingly however, the lge phenotype of gcr1Δ cells requires the presence of a repressing sugar. We used genome-wide expression profiling to comprehensively identify loci that exhibit defective transcription in gcr1Δ cells irrespective of C-source; this analysis identified both previously known and novel Gcr1 targets. Using an established method for separating regulated changes in the transcriptional program of the cell from metabolic changes induced by glucose (Wang et al. 2004), we verified that the defect we observed in RPG expression in gcr1Δ cells is independent of both growth rate and glucose metabolism. Our genomic analysis also detected loci whose expression was increased in gcr1Δ cells; as expected, this includes a significant number of normally glucose-repressed genes. Respiration-dependent staining of mitochondria confirmed at the metabolic level that Gcr1, in addition to activating glucose-induced genes, is also required for glucose repression of mitochondrial function.

Materials and methods

Strains and growth conditions

Isogenic S. cerevisiae wild type [BY263, Mat a ade2-107 his3Δ200 leu2-Δ1 trp1Δ63 ura3-52 lys2-80 (Measday et al. 1994)], gcr1Δ [KW1433, same as BY263 except gcr1::URA (Willis et al. 2003)], and cln3Δ [GMS3503, same as BY263 except cln3::URA3 (Willis et al. 2003)] strains (S288C background) were used in this study. All strains were grown at 23°C (due to the temperature sensitive phenotype of the gcr1Δ mutant; Willis et al. 2003) as batch cultures in YEP containing the indicated C-sources. A Beckman Coulter MultiSizer 3.0 was used to measure cell density (cells/ml) and cell volume (fl); each growth curve was derived from three independent experiments. Mitochondrial staining was done on cells grown to mid-logarithmic phase in the C-source indicated.

Transcriptomic analysis

Steady-state microarray analyses were done by isolating total RNA (Wang et al. 2004) from isogenic gcr1Δ and wild type cultures grown to early logarithmic phase. Time course microarrays to analyze the immediate response to glucose were done by growing the same two strains in YEP containing 3% pyruvate as the carbon source. A reference sample was taken immediately prior to the addition of glucose (time zero); further aliquots of the cultures were harvested 20 and 60 min after glucose addition, and total RNA was extracted for all samples. For all microarray analyses, the quality of RNA was tested by using an Agilent Bioanalyzer 2100 with RNA Nano 6000 LabChips. Samples were labeled with Cy3-CTP or Cy5-CTP by using a low input fluorescent linear amplification kit (Agilent Technologies, Palo Alto, CA, USA). Labeled cRNA was purified with the RNeasy MinElute kit (Qiagen, Valencia, CA, USA) and hybridized to yeast 60-mer oligonucleotide arrays (Agilent Technologies) according to the manufacturer’s instructions. Slides were scanned at 10 μm resolution with two-line averaging using an Axon GenePix 4200A scanner and GenePix 6.0 software. Ratio-based and LOWESS normalization as well as statistical analysis were done in Acuity 4.0 (Molecular Dynamics, Piscataway, NJ, USA). Misregulated genes in gcr1Δ were identified by an average expression change of at least twofold (i.e., a log2 ratio of ≤−1 or ≥1) relative to the isogenic wild type value. Statistical significance of the genes identified by this analysis was confirmed by performing a paired, one-tailed t-test against a control array; genes referred to as statistically significant have a p < 0.001. Hierarchical clustering was done by using a centered Pearson similarity metric; K-means cluster analysis was done by using a Euclidean squared similarity metric. Steady-state microarrays were done in four replicate experiments, including both biological replicates and dye-swap technical replicates. Time course microarrays were done according to the method of Wang et al. (2004). p-values confirming the enrichment for specific MIPS functional groups were calculated as previously described (Robinson et al. 2002). All microarray data are compliant with Minimal Information About Microarray Experiments (MIAME; Brazma et al. 2001) standards, and the full datasets are available though the NCBI Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/projects/geo/). The accession number for steady state analysis in glucose-, galactose-, and raffinose-grown cells is GSE5027; the accession number for time course analysis is GSE5575.

Mitochondrial staining

To assay respiratory function via mitochondrial staining, isogenic cultures were grown to mid-logarithmic phase in YEP containing the indicated C-sources. Cells were first stained with MitoTracker Red CM-H2XRos (Invitrogen, Carlsbad, CA, USA) by adding the dye to a final concentration of 225 μM and continuing incubation for 15 min. Cells were then collected by low-speed centrifugation and resuspended to a density of 1 × 107 cells/ml in 10 mM HEPES [pH 7.4] containing the indicated C-source. Cultures grown in 3% glycerol were supplemented with a low level of glucose to stimulate growth (Banuelos and Fraenkel 1982; Sherman 2002). Mitochondrial DNA was counterstained by adding SYTO 18 (Invitrogen) to a final concentration of 10 μM and incubating at room temperature for 5 min. Stained cells were then visualized on a Zeiss 510 Meta confocal laser-scanning microscope. SYTO 18 was excited by using the Argon laser set to 488 nm; resulting fluorescence was detected with a 530–600 nm band pass filter. MitoTracker Red was excited with the HeNe laser at 543 nm; resulting fluorescence was detected with a 585 nm long pass filter. Each image was captured by using a 100× Plan-Apochromat oil objective such that 1 pixel = 0.14 μm.

Results

The gcr1Δ growth defect is C-source independent

Cells lacking the transcriptional regulator Gcr1 fail to induce rapid growth in the presence of glucose (Willis et al. 2003). Because the glycolytic genes are well-characterized targets of Gcr1 (Clifton and Fraenkel 1981; Deminoff and Santangelo 2001; and see below), we tested growth of gcr1Δ cells in C-sources which require varying degrees of reliance on glycolysis for the generation of energy, ranging from almost exclusively glycolysis-dependent (glucose) to not at all (pyruvate). We included two repressing C-sources (glucose and galactose; Polakis and Bartley 1965; Lodi et al. 1991) as well as two non-repressing C-sources (raffinose and pyruvate; Stoppani 1951; Lodi et al. 1991) in this test. The gcr1Δ growth defect is not specific to glucose but is also detectable in galactose-, raffinose-, and pyruvate-grown cultures (Fig. 1). As expected, the doubling time of wild type cells is shortest in glucose, the preferred carbon source of all organisms; in this background, we calculated the doubling time of glucose-grown wild type cells to be 1 h, 48 min. The doubling time of wild type cells grown in the presence of either galactose or raffinose was slightly more than 4 h. The doubling time of gcr1Δ cells grown in glucose, galactose, or raffinose was in all cases 8 h. This is approximately four times longer than glucose-grown wild type cells, and twice as long as galactose- or raffinose-grown wild type cells. Despite their longer doubling times, gcr1Δ cultures eventually attain cell densities in stationary phase that are comparable to the corresponding wild type cultures (Fig. 1 and data not shown).

Fig. 1
figure 1

Growth of gcr1Δ cells is impaired in all carbon sources. Isogenic strains were inoculated at a density of 1 × 106 cells/ml and incubated at 23°C to observe growth phenotypes. Wild type (WT; open symbols) and gcr1Δ (closed symbols) cells were grown in YEP containing either 2% glucose (a, circles), 3% galactose (a, triangles), 3% raffinose (b, circles) or 3% pyruvate (b, triangles). Error bars represent standard deviation

Full size image

Using these values, we calculated the doubling time of wild type cells to be sixfold shorter in the presence of glucose, and threefold shorter in the presence of either galactose or raffinose, than in the presence of pyruvate. The doubling time of gcr1Δ cells, however, is only twofold shorter in glucose, galactose, or raffinose than the doubling time of gcr1Δ cells grown in pyruvate. Therefore, as previously reported for glucose (Willis et al. 2003), the addition of galactose or raffinose to the culture medium has no negative effect on growth; gcr1Δ cells simply lack the wild type response to added sugars and consequently exhibit no more than a modest increase in growth rate.

Genome-wide expression profiling of gcr1Δ cells: reduced transcription of known and novel Gcr1 targets in both repressing and derepressing C-sources

A high-throughput analysis of expression in gcr1Δ cells using filter hybridization technology has previously been reported (Lopez and Baker 2000). However, microarrays printed onto solid glass supports offer more efficient hybridization, greater sensitivity, and better reproducibility than filter-based arrays (Southern et al. 1999); this technical advance, in combination with improved methods for data processing and analysis (Slonim 2002), means that reliable quantitative data can now be obtained from microarray analysis. We therefore did microarray analyses to profile expression in gcr1Δ cells. This analysis has already been reported for pyruvate-grown cells (Santangelo 2006) and glycerol–lactate grown cells (Sasaki and Uemura 2005), so we analyzed cultures grown in glucose, galactose, and raffinose.

We identified 116 genes that exhibited a statistically significant reduction in expression in all C-sources in the absence of Gcr1 (Fig. 2a; Table 1). An analysis of the promoters of these genes shows that 86 contain a strong match to the consensus Rap1 binding site (RMACCA; Zhu and Zhang 1999) and/or interact with Rap1 in a chromatin immunoprecipitation (ChIP) assay (Lieb et al. 2001; Lee et al. 2002), while 101 have a consensus Gcr1 binding site (CWTCC; Zhu and Zhang 1999; a complete list of these genes, cross-referenced to a previously published analysis of the genome-wide location of Rap1/Gcr1 binding sites, is given in web Table S1). Not surprisingly, most of the 116 (∼72% of the total) were either glycolytic or RPGs (Table 1). This result agreed with our previous identification of individual Gcr1 target genes (Deminoff and Santangelo 2001). Eleven of these genes [FBA1, TPI1, TDH1, TDH2, TDH3, ENO1, ENO2, PGK1, GPM1, CDC19 (PYK1), and ADH1] were also identified as showing a twofold or greater decrease in expression in one or both of two previous large scale analyses from other labs (Lopez and Baker 2000; Sasaki and Uemura 2005). One of these studies (Sasaki and Uemura 2005) also identified 57 RPGs as having reduced expression in gcr1Δ cells; however, the names of the individual genes were not included in the published data, so we are unable to assess the exact degree of correlation between that study and ours.

Fig. 2
figure 2

Transcriptomic analysis identifies down-regulated targets in gcr1Δ. Microarrays were done by competitive hybridization of RNAs from gcr1Δ and wild type cells grown in YEP containing either 2% glucose, 3% galactose, or 3% raffinose. a Hierarchical clustering was done by using a centered Pearson similarity metric on genes that exhibited at least a twofold decrease in expression in the mutant. The scale at the bottom shows color intensity relative to the mean log2 ratio of four replicate arrays in each carbon source. Genes whose expression was found to be equivalently defective in all C-sources, as defined by calculating ratios of the log2 expression values in glucose-, galactose-, and raffinose-grown cells and selecting for a value of 1 (±10%), are indicated by arrows. Asterisks denote dubious ORFs (see text for details). b K-means cluster analysis of the down-regulated target genes shown in a, generated by using a Euclidean squared similarity metric. K-means clusters are shown individually, with color intensity indicating the mean log2 ratio. The expression profile of the genes contained in each cluster is depicted by the overlaid line graph. Points on the line (white circles) represent the average of the log2 ratio for the genes in the cluster in either glucose-, galactose-, or raffinose-grown cells, respectively; averages are −2.4, −2.1, and −2.2 for cluster I; −1.5, −2.0, and −1.4 for cluster II; −2.2, −2.1, and −1.2 for cluster III; −2.4, −1.8, and −1.2 for cluster IV

Full size image
Table 1 MIPS functional classification of differentially regulated genes in gcr1Δ cells grown on glucose, galactose, or raffinose
Full size table

Since asynchronous cultures were profiled we did not expect to (and in fact did not) detect the significant Gcr1-dependence of cyclin gene transcription in glucose-grown cells (Willis et al. 2003). We did, however, identify a number of novel genes as exhibiting Gcr1-dependent transcription (Fig. 2a), including YBR187W [a previously uncharacterized open reading frame (ORF)] and PET122 (which encodes a mitochondrial translation factor). The promoters of both genes contain consensus-binding sites for Rap1 and Gcr1, and Rap1 interacts with the promoter of YBR187W (web Table A). As previously noted (Lopez and Baker 2000), there are two ORFs, now listed as dubious, that overlap extensively with Gcr1 targets: YKL153W, which overlaps GPM1; and YCR013C, which overlaps PGK1. We have identified a third such dubious ORF, YLL044W, which overlaps the Gcr1 target RPL8B. Interestingly, these three ORFs are differentially regulated in wild type and gcr1Δ cells (Fig. 2a); the design of our arrays, which use single-stranded 60-mer oligonucleotide probes complementary only to the target RNA, eliminates the possibility that this difference is attributable to cross-hybridization. Recent work from several groups using high-density tiling arrays has demonstrated the existence of independent transcription units from non-coding genes in yeast (David et al. 2006; Samanta et al. 2006). The 402 naturally occurring antisense transcripts detectable in both total and poly(A) RNA samples included YKL153W, YCR013C, and YLL044W (David et al. 2006) suggesting the intriguing possibility that these ORFs may contribute to the regulation of GPM1, PGK1, and RPL8B.

K-means cluster analysis divides the 116 genes that meet the cutoff for significant reduction in gcr1Δ cells in all C-sources into four distinct patterns of expression (Fig. 2b). Of these 116 genes, only 20 display little or no difference in the degree to which transcript levels are reduced in glucose-, galactose-, or raffinose-grown cells (cluster I, Fig. 2b). This cluster of 20 genes displays a significant functional enrichment for genes involved in glycolysis and gluconeogenesis (p = 3.71 × 10−6) as well as for genes involved in amino acid biosynthesis (p = 0.006). Genes whose dependence on Gcr1 was equivalent in all C-sources (arrows in Fig. 2a) includes the novel targets SUN4, UTR2, SAM1, and PET122, two glycolytic genes (CDC19 and GPM1), and an RPG (RPS14B). However, for most of the genes in cluster I, the ratio of gcr1Δ to wild type expression appears to be slightly lower in glucose-grown cells than in galactose- or raffinose-grown cells. Cluster II shows that another 25 genes have a significantly lower gcr1Δ to wild type ratio in galactose-grown cells than in glucose- or raffinose-grown cells; this cluster displays a significant functional enrichment for genes involved in glycolysis and gluconeogenesis (p = 9.49 × 10−6) as well as for genes involved in ribosome biosynthesis (p = 1.56 × 10−4).

For most genes affected by GCR1 deletion, the difference in expression relative to wild type was more apparent in glucose- and galactose-grown cells than in raffinose-grown cells (71 genes total, shown in clusters III and IV, Fig. 2b). These two clusters are both significantly enriched for RPGs (p < 1 × 10−14). Cluster III also includes the negative regulator of translation ASC1 and the uncharacterized essential gene YBR187W. Expression of YBR187W during the diauxic shift (DeRisi et al. 1997) and in response to glucose (Wang et al. 2004) also places it in a cluster with RPGs. Ybr187W has been shown to physically interact (Ito et al. 2001; Krogan et al. 2006) with Rpa12, Nop12, Mrd1, and Bms1, all of which participate in the synthesis and processing of rRNA (Gelperin et al. 2001; Wegierski et al. 2001; Wu et al. 2001; Jin et al. 2002; Prescott et al. 2004), as well as with Rrs1, which is a regulator of ribosome synthesis (Tsuno et al. 2000). The established method of assigning function to a gene based on its expression profile (Wu et al. 2002), and physical interactions, allow us to make the novel assignment of YBR187W to a role in ribosome biogenesis. We therefore give YBR187W the name GDT1, for Gcr1-Dependent Translation factor 1.

An average of the log2 ratios of all RPGs that meet the cutoff for significant down-regulation in the gcr1Δ mutant suggests that the defect in expression is the same within error for glucose-, galactose-, or raffinose-grown cells. However, an average for the genes in either Cluster III or IV alone shows that for a subset of RPGs, the ratio of gcr1Δ to wild type expression is lower in glucose- or galactose-grown cells than in raffinose-grown cells. Since it is commonly accepted that RPG expression is tightly linked to growth rate, this result is interesting and somewhat puzzling, as the doubling time of gcr1Δ cells grown in any of these three carbon sources is indistinguishable.

Global expression analysis of gcr1Δ cells shows that the misregulation of RPGs is detectable 20 min after glucose addition and is not a response to slow growth

Expression of RPGs is elaborately controlled; it is known to be coordinately regulated and linked to both cell growth and secretory function (Kief and Warner 1981; Mizuta and Warner 1994). RPGs are also subject to extensive post-transcriptional regulation and feedback control (Dabeva et al. 1986; Presutti et al. 1991; Li et al. 1995, 1996; Fewell and Woolford 1999). Furthermore, given the importance of protein synthetic capacity for supporting the accumulation of new cell mass, it is not surprising that when the deletion of any one of the genes encoding a factor that contributes to RPG expression is viable (GCR1, SFP1, FHL1, IFH1, or HMO1), it results in a pronounced slow growth phenotype (Clifton and Fraenkel 1981; Blumberg and Silver 1991; Hermann-Le Denmat et al. 1994; Cherel and Thuriaux 1995; Lu et al. 1996). This presents a challenge for understanding the transcriptional regulation of these genes, since any effect on expression may be attributed to a variety of causes. To eliminate any metabolic or growth-related effects on expression, we examined the immediate transcriptional response of gcr1Δ cells to the appearance of glucose.

As a reference, we compared wild type and gcr1Δ cells grown in pyruvate (0 min, Fig. 3). Within 20 min of glucose addition, 41 RPGs were found to be misregulated relative to the absence of glucose (0 min vs. 20 min, Fig. 3), as identified by K-means clustering. Nineteen of these genes were down-regulated (asterisks in Fig. 3), while 22 were up-regulated (diamonds in Fig. 3). However, most of the observed up-regulation is transient; 60 min after glucose addition, there is reduced expression of 90 of 116 genes that are down in gcr1Δ cells in steady state; this includes all of the RPGs shown in Fig. 2 except RPL41B. This transcriptomic effect occurs before the onset of any change in growth, suggesting that the altered expression seen in this experiment is likely to be a direct result of GCR1 deletion. RPG transcripts that remain less than twofold reduced 60 min after glucose addition are RPS25A, RPS27A, RPS28A, RPS30A, RPS30B, RPL10, RPL41B, RPL15B, and RPP2B (Fig. 3 and data not shown). RPS25A, RPS27A, RPS28A, RPS30A, RPS30B, RPL10, and RPL15B are not in the set of 116 genes identified by our previous analysis (Fig. 2a), while RPL41B was at the borderline of significance in galactose- and raffinose-grown cells. Transcription of RPS28A and RPP2B is known to be Rap1-independent (Mager and Planta 1991), supporting the idea that this experimental design allows us to distinguish clearly between those RPGs that are directly affected by Gcr1 and those that are influenced by coordinate regulation.

Fig. 3
figure 3

The gcr1Δ defect in expression of RPGs can be detected as early as 20 min after carbon source up-shift. Time course microarrays were done by competitive hybridization of RNAs from gcr1Δ and wild type cells grown in YEP in the absence of glucose (0 min) or after glucose addition (20 and 60 min). The scale at the bottom shows color intensity relative to the log2 ratio of expression in gcr1Δ and wild type cells. Genes whose expression exhibits a decrease (indicated by asterisks) or an increase (indicated by diamonds) relative to wild type 20 min after glucose addition were identified by K-means cluster analysis (data not shown)

Full size image

Our analysis shows that 73 of 138 total RPGs in S. cerevisiae exhibit a statistically significant reduction in expression in glucose-, galactose-, and raffinose-grown gcr1Δ cells (Fig. 2a, web Table A); for all but one of the 73, the reduction in expression is growth-rate independent (Fig. 3). Of these 72 genes, 68 have a strong Rap1 consensus-binding site and/or are known to be Rap1-bound in vivo (web Table S1). This group of 68 genes, whose reduction in expression is both growth rate-independent and statistically significant in cells grown on a non-repressing carbon source, represents our high-confidence dataset for RPGs that are affected by Gcr1.

The large size (lge) phenotype of gcr1Δ cells is specific to growth on a repressing sugar

Yeast mutants defective in either protein biosynthesis or carbon metabolism have been shown to exhibit cell size phenotypes (Jorgensen et al. 2002). Glucose-grown gcr1Δ cells are known to exhibit a lge phenotype; this phenotype is linked to both low levels of CLN mRNA and a severely reduced protein synthetic rate (Willis et al. 2003). We show here that expression of some Gcr1-dependent RPGs varies in a C-source dependent manner (Fig. 2b), suggesting the possibility that in gcr1Δ cells, protein biosynthetic rates and/or cell cycle progression may also vary with C-source. If this is the case, then gcr1Δ cell size might also be expected to vary, since the doubling time of gcr1Δ cells remains the same in all C-sources. We therefore measured the size of gcr1Δ cells in different C-sources to test whether the lge phenotype was C-source independent or instead might vary in a C-source dependent manner. We first measured wild type cells, which are known to undergo a slight increase in cell size during rapid growth on glucose (Johnston et al. 1979; Tokiwa et al. 1994). This was indeed observed as a modest shift of the cell volume peak for glucose compared to cells grown in other C-sources (Fig. 4a). The lge phenotype of gcr1Δ cells, which is pronounced in glucose, is at least as severe in galactose (Fig. 4b). Significantly, there was no detectable cell size phenotype during growth on the non-repressing C-sources raffinose and pyruvate (Fig. 4b). To confirm that this effect is specific to Gcr1 removal we measured cln3Δ cells, which are defective in G1 progression and well-characterized as having a lge phenotype in glucose (Nash et al. 1988). Importantly, unlike gcr1Δ cells, cln3Δ cells are larger to approximately the same extent in each of the C-sources tested (Fig. 4c).

Fig. 4
figure 4

gcr1Δ cells have a carbon source dependent large size phenotype. Cell size distributions from wild type (WT, a) and gcr1Δ cells (b) growing in YEP containing 2% glucose, 3% galactose, 3% raffinose, or 3% pyruvate as the carbon source were collected during mid-logarithmic phase. Cell volume (fl) is plotted relative to number percent of the population. Each determination represents ∼10,000 cells. c The mean cell volume from gcr1Δ and cln3Δ cells is plotted as a percentage of the corresponding wild type value for each of the four carbon sources. Each bar represents the average of duplicate samples collected at three time points during mid-logarithmic growth. Error bars denote standard deviation and a dashed line representing the wild type value (set to 100%) is included for reference

Full size image

Gcr1 removal results in both a genome-wide defect in glucose repression and an increase in the expression of stress response genes

Our microarray analysis also identified 118 genes whose transcription exhibited a statistically significant increase in glucose-, galactose-, and raffinose-grown gcr1Δ cells (Fig. 5a). Only 28 of these genes either contain a consensus Rap1 site in their promoter (Zhu and Zhang 1999) and/or are bound by Rap1 (Lieb et al. 2001; Lee et al. 2002), although 97 contain a consensus Gcr1 binding site (Zhu and Zhang 1999; a complete list of these genes is given in web Table S2). The motif discovery algorithm MEME (Bailey et al. 2006) did not identify any other sequence elements common to the promoters of all 118 genes. Since a CT box alone is insufficient for Gcr1 to function in vivo (Buchman et al. 1988; Zeng et al. 1997), the influence of GCR1 deletion on some of these genes may occur as a result of changes in highly networked regulatory pathways. Notable exceptions whose promoters contain both a Rap1 site and a consensus CT box are the transcription factor TYE7, the stationary-phase induced gene SNZ1, the glucose-regulated stress response factor HSP12, the transketolase TKL2, and the inosine monophosphate dehydrogenase IMD2, which catalyzes the first step of GMP biosynthesis.

Fig. 5
figure 5

Transcriptome analysis identifies up-regulated targets in gcr1Δ. Microarrays were done by competitive hybridization of RNAs from gcr1Δ and wild type cells grown in YEP containing either 2% glucose, 3% galactose, or 3% raffinose. a Hierarchical clustering was done on genes that exhibited at least a twofold increase in expression in the mutant by using a centered Pearson similarity metric. The scale at the bottom shows color intensity relative to the mean log2 ratio of four replicate arrays. b K-means cluster analysis of the up-regulated target genes shown in a, generated by using a Euclidean squared similarity metric. K-means clusters are shown individually, with color intensity indicating the mean log2 ratio. The expression profile of the genes contained in each cluster is depicted by the overlaid line graph. Points on the line (white circles) represent the average of the log2 ratio for the genes in the cluster in glucose-, galactose-, or raffinose-grown cells, respectively; averages are 2.0, 1.9, and 1.9 for cluster I; 2.9, 2.5, and 1.5 for cluster II; 3.6, 1.9, and 2.0 for cluster III

Full size image

Multiple processes that are normally either silent or weakly expressed in glucose- or galactose-grown cells (Table 1), including alternative C-compound metabolism and the pentose phosphate pathway, were significantly enriched among the genes in this dataset. This was not surprising, since Gcr1 is known to participate in glucose repression (Turkel et al. 2003; Sasaki and Uemura 2005), and five of these genes (GLK1, DLD3, CIT1, CIT3, and CIT2) were previously identified as having significantly increased expression in gcr1Δ cells (Sasaki and Uemura 2005). Additionally, there was a significant enrichment in this dataset for genes that are functionally annotated to the stress response (YRO2, GRX3, HSP78, HSP12, DAK2, CTT1, DAK1, SNO1, SNZ1, PAI3, YGP1, DDR2, YOR054C, and ATH1). However, more than half of these 14 genes are also up-regulated in response to the diauxic shift (DeRisi et al. 1997). It has recently been suggested that activation of some component of the stress response may be required for the switch between fermentative and oxidative metabolism (Brauer et al. 2005). Since gcr1Δ cells display no significant increase in thermotolerance (data not shown), it seems likely that the enrichment for stress response genes is linked to changes in carbon metabolism (see below).

Using K-means cluster analysis, we divided the 118 genes that exhibit a significant increase in expression into three groups. Slightly less than half (47 genes, shown in cluster I, Fig. 5b) show little or no difference in expression between glucose-, galactose-, or raffinose-grown cells. This cluster is functionally enriched for genes involved in nucleotide metabolism (p = 1.01 × 10−4), and specifically for genes involved in ribonucleotide metabolism.

Genes in clusters II and III have a significantly higher gcr1Δ to wild type ratio in glucose-grown cells than in galactose- or raffinose-grown cells; these clusters display a significant functional enrichment for genes involved in C-compound metabolism (p = 5.27 × 10−4) and C-compound and carbohydrate utilization (p = 6.89 × 10−6), respectively. Additionally, cluster II is functionally enriched for genes that are part of the pentose-phosphate pathway (p = 1.19 × 10−3). These results are strongly suggestive of a genome-wide alteration in carbon metabolism in gcr1Δ cells.

Comparison of glucose signaling in wild type and gcr1Δ cells

To place the effect of GCR1 deletion on the cells’ transcriptional program into a larger biological context, we compared our microarray data to a previously published analysis of the glucose response in wild type cells (Wang et al. 2004). There is a strong inverse correlation in the expression of nutrient responsive genes in wild type and gcr1Δ cells upon the appearance of glucose. Twenty minutes after glucose addition, transcription of 1,651 (1,067 + 584) genes is repressed at least twofold by wild type cells. Of these 1,651 genes, 584, or 35%, show a twofold or greater increase in transcription in gcr1Δ cells over the same interval of time (Fig. 6a); in contrast, only 131, or 8%, decrease (data not shown).

Fig. 6
figure 6

A comparison of the glucose response in wild type and gcr1Δ cells. a and b Venn diagrams showing a comparison between the transcriptomic response of gcr1Δ to glucose and the previously reported (Wang et al. 2004) wild type response to glucose. Thirty-five percent of the genes that are repressed in wild type cells in response to glucose are expressed at a higher level in gcr1Δ cells (a) and 34% of the genes that are induced in wild type cells are expressed at a lower level in gcr1Δ cells (b)

Full size image

The intersection between the set of genes that is repressed in wild type cells and the set of genes that is up-regulated in gcr1Δ cells is most significantly enriched for genes involved in energy metabolism, C-compound and carbohydrate utilization, respiration, C-compound and carbohydrate metabolism, and the TCA cycle (Table 2), consistent with a relief from glucose repression in gcr1Δ cells. Other functional categories enriched in the intersection include stress response genes and genes that participate in amino acid catabolism. Genes involved in amino acid metabolism and biosynthesis are also significantly enriched among the 996 genes up-regulated in gcr1Δ cells in response to glucose (Fig. 6a; Table 2); this is intriguing in light of a recent bioinformatics-based study that suggested Gcr1 may act downstream of Cha4 (MacIsaac et al. 2006), a zinc-finger transcription factor required for the catabolism of serine and threonine (Holmberg and Schjerling 1996).

Table 2 Comparison of glucose repressed genes in wild type cells with glucose induced genes in gcr 1Δ cells
Full size table

The inverse correlation described above also holds true for genes that increase in wild type relative to those that decrease in the gcr1Δ mutant 20 min after the appearance of glucose. About 1,103 (727 + 376) genes are induced twofold or more in wild type cells in response to glucose, 34% of which are down-regulated twofold or more in gcr1Δ cells. However, only 75, or 7%, are induced in gcr1Δ cells (data not shown). The 376 genes that represent the intersection between these two sets function almost exclusively in rRNA and tRNA biosynthesis (Fig. 6b; Table 3); these RNA components of the translational machinery are necessary to support maximum cell growth. Glucose addition induces a further set of rRNA and tRNA genes in wild type cells (Table 3). Interestingly, cell cycle genes are among those whose expression is decreased in the gcr1Δ mutant in response to glucose (Table 3), consistent with an imbalance in growth and cell cycle progression (Willis et al. 2003).

Table 3 Comparison of glucose induced genes in wild type cells with glucose repressed genes in gcr1Δ cells
Full size table

Mitochondrial staining reveals that gcr1Δ cells respire in the presence of repressing sugars

Since gcr1Δ cells are defective in repression of numerous genes that encode mitochondrial functions, we used a respiration-dependent mitochondrial stain, MitoTracker Red CM-H2XRos, to confirm that mitochondrial function is derepressed in the absence of Gcr1. The MitoTracker probe is added to cells in its reduced form and is not fluorescent unless it enters an actively respiring cell. In respiring cells the probe is oxidized to its fluorescent form and is sequestered in the mitochondria. To confirm its specificity we co-localized the MitoTracker signal with a fluorescent mitochondrial DNA stain (SYTO 18). When the media is supplemented with either glucose or galactose, mitochondria in gcr1Δ cells are active despite the presence of a repressing sugar; this is in stark contrast to the inactive mitochondria of wild type cells (Polakis and Bartley 1965; Fig. 7; Table 4). Mitochondrial activity in vivo therefore correlates with the genome-wide glucose repression defect that was observed with expression profiling. Further, quantitation of MitoTracker staining (Table 4) demonstrated that the percentage of respiring gcr1Δ cells grown on repressing sugars is as large as the percentage of respiring wild type cells grown on non-repressing C-sources (pyruvate and glycerol).

Fig. 7
figure 7

gcr1Δ cells respire in repressing C-sources. Wild type (WT) and gcr1Δ cells were grown in YEP containing either 2% glucose (a and d), 3% galactose (b and e), or 3% glycerol (c and f) as the carbon source. Each panel shows mid-logarithmic cells stained with the respiration-dependent fluorescent dye, MitoTracker Red CM-H2XRos (MT Red; top left quadrant of each panel) and counterstained with the mitochondrial DNA fluorescent dye, SYTO 18 (bottom left quadrant). The corresponding DIC image is shown in the top right quadrant. The bottom right quadrant shows all three of these quadrants merged into one image. Quantitation of MT Red staining for each condition is shown in Table 4. Though as expected pyruvate-grown cells have active mitochondria that can be detected in this assay (see Table 4), glycerol-grown cells are shown as the positive control because for unknown reasons they stain more brightly with MitoTracker

Full size image
Table 4 Respiration-dependent mitochondrial staining of gcr1Δ cells
Full size table

Discussion

The global regulator Gcr1 activates transcription of three classes of gene products that influence the transition through Start in the G1 phase of the cell cycle: glycolytic enzymes, ribosomal proteins, and cyclins (Clifton and Fraenkel 1981; Santangelo and Tornow 1990; Tornow et al. 1993; Zeng et al. 1997; Deminoff and Santangelo 2001; Willis et al. 2003); transcription of these genes makes up more than 75% of pol II transcripts in rapidly growing cells (Santangelo and Tornow 1990; Warner 1999). The gcr1Δ mutant, which is known to be defective in exiting G1 (Willis et al. 2003), divides at a slow rate regardless of the catabolic pathways employed or the richness of the energy source (repressing and non-repressing, as well as fermentable and non-fermentable, C-sources were tested; Fig. 1).

In the absence of Gcr1, transcription of two glycolytic genes―CDC19 and GPM1―was equally defective irrespective of the C-source (arrows in Fig. 2a). Since the product of CDC19 (which encodes pyruvate kinase) does not participate in gluconeogenesis, its reduced expression in pyruvate-grown gcr1Δ cells (Zeng et al. 1997; Santangelo 2006) likely makes no contribution to the corresponding growth defect. It is certainly possible that one or more of the remaining six genes that exhibit unconditionally altered expression in the absence of Gcr1 (RPS14B, GPM1, SAM1, PET122, SUN4, and UTR2) account entirely for the C-source independent gcr1Δ growth defect. However we favor an alternative explanation―that the failure to properly regulate expression of the remaining 227 Gcr1-dependent genes is at least partly responsible for the inability of gcr1Δ cells to muster more than a meager increase in the rate of cell division in the presence of even the richest source of energy, glucose. If this idea is correct, Gcr1 is one of but a few master regulators of growth/cell cycle coordination (see below).

We show here that slightly more than half of all RPGs are affected by GCR1 deletion when cells are grown in either a repressing or a non-repressing C-source (Fig. 2). We also show that reduced expression of RPGs in gcr1Δ cells grown in the repressing C-source glucose cannot be attributed either to slow growth or secondary metabolic effects (Fig. 3). This is consistent with earlier work showing that abolition of Gcr1 dimerization results defects in RPG expression without resulting in a slow growth phenotype (Deminoff and Santangelo 2001). However, it is in conflict with the previously published filter hybridization-based genome-wide analysis of expression in gcr1Δ cells (Lopez and Baker 2000), which reports an increase in RPGs expression in glycerol–lactate grown cells. We do not observe an increase in RPG expression in gcr1Δ cells during steady-state growth on a non-repressing C-source, as measured on either a global level [(raffinose-grown cells; (Fig. 2) or pyruvate-grown cells; (Santangelo 2006) and time zero in (Fig. 3)], or on the level of the individual gene (S1 and primer extension analysis of RPG expression in either pyruvate-grown or glycerol–lactate grown cells; our unpublished data). This discrepancy may be attributable to the different platforms used in this and the earlier study.

Based upon the above conclusions as well as our previously published work, most if not all of the glycolytic and translational component genes identified in this study are likely to be directly stimulated targets of Gcr1. However, it is important to confirm that altered transcription of a gene identified with expression profiling reflects direct targeting by a given regulator, rather than a complex readout generated by a cross-talking network of transcription factors. This requires at minimum that the regulator in question be shown to interact with the putative target gene. Genome-wide location data (Lee et al. 2002) are available for all 141 transcription factors listed in the Yeast Protein Database (Costanzo et al. 2000), including Gcr1 and the other components of its activation assemblage (Rap1 and Gcr2; Deminoff and Santangelo 2001; Santangelo 2006). Genes identified in our profiling experiments for which location analysis provides confirmation of direct targeting by Gcr1 are TPI1, ADH1, CDC19, GPM1, TDH3, RPL1B, and RPS19A (Lee et al. 2002). However, at least where Gcr1 is concerned there must be numerous false negatives in the genome-wide location analysis; for example, genomic foot printing has confirmed that Gcr1 binds in vivo to ENO1, ENO2, and PGK1 (Henry et al. 1994). Though similar confirmation has been difficult to obtain for individual RPGs via ChIP with epitope-tagged Gcr1 (data not shown), it seems unlikely that RPL1B and RPS19A will prove the only bona fide direct RPG targets of the Gcr1 regulator.

With respect to its involvement in glucose repression, direct interaction between Gcr1 and the canonical glucose-repressed gene SUC2 has been established (Turkel et al. 2003; and our unpublished data). However, though we have shown that Gcr1 is required for glucose repression of mitochondrial gene function (Fig. 7), neither SUC2 nor the glucose-repressed genes detected here (Fig. 5) were identified as direct targets via genome-wide location analysis (Lee et al. 2002). In combination with the aforementioned result for putative RPG targets of Gcr1, continued acceptance of the recruitment paradigm of gene regulation leaves us with an uncomfortable choice regarding these data. Either the interaction between Gcr1 and most of its direct targets is too subtle to detect with the ChIP technique, or Gcr1 influences some glycolytic genes, RPGs, and glucose-repressed genes directly, and other genes within these same classes indirectly. What makes this even more puzzling is that most of the promoters in question have an identical profile of specific DNA binding sites (e.g., Rap1 sites in RPG promoters).

The reverse recruitment paradigm of gene regulation (Santangelo 2006) may allow us to resolve this conundrum. Gcr1 involvement in both activation (Menon et al. 2005) and repression (our unpublished data) occurs via a reverse recruitment mechanism. If target genes are attracted to a megalithic protein assemblage while active, and are released from this assemblage when inactive, whether or not a given interaction between a regulator and its putative target gene can be detected will almost certainly be subject to spurious influences. For example, a false negative may be obtained because the position of a subunit within the assemblage places it many subunits away from the DNA helix and therefore (at least in a standard ChIP assay) incapable of pulling bona fide target genes into an immunoprecipitated pellet. Clearly much work remains in characterizing the physical relationships between Gcr1, its regulatory partners and targeted chromatin. Indeed, the reverse recruitment idea postulates that nuclear suprastructure is a central feature of many (if not all) gene regulatory mechanisms.

While investigating the effects of C-source on gcr1Δ cells we also discovered two new aspects of the gcr1Δ phenotype. First, gcr1Δ cells respire in the presence of fermentable sugars; whether this altered lifestyle is a cause or effect of the changes we observe in glucose-regulated gene expression is a very interesting question. Changes in gene expression may cause mitochondria to become active. One model that fits into this category is that Rap1 and Gcr1 directly affect the expression of one or more genes, such as PET122, SNZ1, and/or HSP12 that trigger the shift from fermentative to non-fermentative metabolism. According to this model, deletion of gcr1Δ causes the switch to be flipped inappropriately, resulting in actively respiring mitochondria in glucose-grown cells. Another model is that actively respiring mitochondria are the result of reduced glycolytic flux, brought about by reduced expression of glycolytic genes in the mutant. Alternatively, actively respiring mitochondria may signal for changes in gene expression. One possibility that fits into this category is that the cell senses changes in the GDP/GTP ratio, as the tricarboxylic acid cycle produces GTP; interestingly, expression of genes required for ribonucleotide biosynthesis are increased in gcr1Δ cells.

The second new feature of the gcr1Δ phenotype reported here is that the large cell size of the mutant is dependent upon the presence of a repressing sugar. This is also true of cells lacking Sfp1, another activator of RPG expression (Jorgensen et al. 2002). However, sfp1Δ cells exhibit the exact opposite response, i.e., in the presence of a repressing sugar gcr1Δ cells are larger whereas sfp1Δ cells are smaller (Jorgensen et al. 2004; Cipollina et al. 2005). There are two different models, which are not mutually exclusive, that may resolve this apparent paradox. The first model postulates that Gcr1 and Sfp1 have qualitatively different effects on protein biosynthesis, resulting in opposite effects on cell size. Consistent with this idea, sfp1Δ and gcr1Δ cells have different polysome profiles; in sfp1Δ cells, the majority of ribosomes are associated with lower-order polysomes (Fingerman et al. 2003), whereas in gcr1Δ cells, the overall levels of both higher- and lower-order polysomes are reduced (our unpublished data). The second model proposes an additional function for Gcr1 and/or Sfp1 that influences progression through Start. For Gcr1, this second function may be an effect on CLN levels that is independent of its effect on RPGs; consistent with this idea, we have previously observed a reduction in CLN1 and CLN2 levels in a gcr1Δ cln3Δ mutant that is greater than the effect of either gcr1Δ or cln3Δ alone (Willis et al. 2003). For Sfp1, this second function might be related to its role in DNA damage (Xu and Norris 1998). A variation of this model hypothesizes that although Gcr1 and Sfp1 have similar functions in stimulating RPG expression, yeast cells require those functions at different times during the cell cycle. Indeed, it has previously been suggested that Sfp1 function may be required at the G2/M transition (Xu and Norris 1998) as well as for passage through Start. Discovering which of these two models is correct should be of critical importance in elucidating the delicate balance between growth, RPG expression, and cell cycle progression.