Activation of Ras2 or Gpa2 Recapitulates Most Glucose-Induced Transcriptional Changes
In order to examine the role of Ras2 and Gpa2 in effecting transcriptional changes in the cell in response to glucose, we measured the global transcriptional response of yeast cells immediately following induction of an activated allele of RAS2 or GPA2 (designated RAS2* and GPA2* in the figures) and compared that to the changes following glucose addition to glycerol-grown cells. To focus on signaling events, rather than the transcriptional consequences of metabolic changes in the cell, we examined the transcriptional response as it changed immediately following addition of glucose. Similarly, to examine the effects of Ras2 or Gpa2 activation, we constructed gal1 strains that carried an activated form of RAS2 or GPA2 under control of the galactose-inducible GAL10 promoter. Since gal1 strains cannot metabolize galactose, addition of galactose resulted in induction of the activated RAS2 or GPA2 allele and a small number of other galactose-inducible genes, but resulted in no changes in the metabolic state of the cell.
In a parallel set of experiments, we examined the transcriptional changes in response to glucose addition of yeast cells containing a PKA that is unresponsive to intracellular cAMP levels. The mutant PKA, referred to as tpk-w
, lacks the regulatory subunit and two of the redundant catalytic subunits, with the third catalytic subunit crippled in its activity (Cameron et al. 1988
). As a consequence, such strains possess constitutive, low-level PKA activity that is unresponsive to changes in cAMP levels in the cell. Thus, changes in cellular behavior dependent on modulation of PKA activity should be abrogated in this strain. The results of both sets of experiments are available in Table S1
The results of comparing Ras2 activation to glucose addition, provided in , indicate that most of the transcriptional changes in the cell immediately following addition of glucose to glycerol-grown cells are recapitulated by activation of Ras2. Prior to initiation of the experiment (during growth on glycerol), the expression pattern of all genes in the wild-type strain (W303 gal1) closely resembled that of the strain carrying the inducible activated Ras allele (W303 gal1 GAL10p-RAS2G19V), with only 0.4% of the genes exhibiting greater than 3-fold differences in absolute expression levels (A; r = 0.96). This reflects the isogenicity of the strains and indicates that the inducible RAS2G19V allele is not expressed under these conditions.
Glucose Stimulation and Ras2 or Gpa2 Activation Yield Similar Transcriptional Responses
Addition of glucose to wild-type cells yields a substantial and rapid change in the transcriptional profile of the cell. By 20 min postaddition, 22% of all genes changed expression by greater than 3-fold and 41% changed expression by 2-fold, with essentially the same number of genes increasing as decreasing (C). This dramatic change in the transcriptional profile was substantially recapitulated by activation of Ras. By 60 min postinduction, the profile of gene expression in the activated strain closely resembled that of the wild-type strain stimulated with glucose (D; r = 0.94). Of those genes exhibiting a change in expression levels of at least 3-fold following addition of glucose, greater than 92% of those showed at least a 2-fold change in the same direction following activation of Ras2 (F). Thus, since glucose yields activation of Ras2 and since Ras2 activation yields changes in transcription that are substantially similar to those observed following addition of glucose, we conclude that a major portion of the glucose signaling pathway regulating transcription can proceed through cAMP via Ras2.
Similar results emerge from analysis of expression changes following activation of Gpa2. Only 0.8% of all genes showed a greater than 3-fold difference in absolute expression levels between the wild-type strain and the strain carrying the inducible activated allele of Gpa2 during growth on glycerol (B; r = 0.97). The pattern of expression at 1 h following activation of Gpa2 strongly resembles that at 20 min following addition of glucose to wild-type cells (F; r = 0.93). However, the response following activation of Gpa2 under these conditions is not as robust as that following activation of Ras2 or addition of glucose. While the overall magnitude of the Ras2-induced response is essentially equivalent to that obtained by glucose addition, the overall magnitude of the Gpa2-induced response is only half that of the glucose-induced changes (F and G). Nonetheless, although somewhat muted, the pattern of transcriptional change induced by Gpa2 closely resembles that induced by glucose. These results are consistent with the hypothesis that the major role of Gpa2 in the cell is modulation of cAMP in response to the presence of a fermentable carbon source.
Redundant Signaling Pathways Control Glucose- Regulated Genes
To analyze the pattern of transcriptional response to glucose addition and cAMP induction, we used a partitional clustering algorithm to group genes on the basis of their behavior over all 32 samples analyzed (Heyer et al. 1999
). Prior to clustering, the expression levels of each gene over the 32 samples were normalized by subtracting from each value the average expression of that gene over all experiments and dividing by the standard deviation of the expression values. This procedure emphasizes the pattern of response of each gene over the experiments, rather than the absolute levels of response. This process yielded 144 clusters ranging in size from seven to 506 members each. By hierarchical clustering (Eisen et al. 1998
), these clusters were further organized into groups on the basis of the similarity of their patterns, yielding eight major classes exhibiting significant change in some respect over the course of the experiments. These classes, encompassing approximately 50% of all genes, are summarized in , and the corresponding pattern of expression is shown in . The list of genes in each class is provided in Table S3
Functional Enrichment among Genes Clustered by Response to Glucose and Ras Activation
Expression Patterns of Clustered Genes
In general, glucose addition yielded a rapid change in expression of genes, which remained unchanged or tended back to starting conditions at later times. We interpret this behavior to indicate that the initial response, seen at the 20 min timepoint, generally represents the response of genes to the signal initiated by addition of glucose. The later deviation from that initial response represents either adaptation of the signaling process or readjustment of expression as a consequence of the change in metabolism of the cell. In contrast, gene expression in response to activation of Ras2 or Gpa2 generally showed a lag of 20 min, followed by a monotonic change in expression over the remainder of the experiment. This is consistent with the expectation that the effects of induction of Ras2 or Gpa2 can be seen only after the new activated protein is transcribed and translated. Further, since under these conditions no significant changes in metabolism occur, the change in expression is due solely to activation of the signaling pathway. This reinforces the notion that the initial response of the cell to glucose is a signaling response, since the pattern of this monotonic change at later times, following activation of Ras2 or Gpa2, generally matches the initial response of those genes to glucose addition.
If those genes induced by glucose and by activation of Ras2 are regulated by glucose solely through the Ras2–Gpa2–cAMP pathway, then we would anticipate that glucose-induced transcriptional alteration would be abrogated in a tpk-w strain. This is the case for a subset of glucose-affected genes (classes II and VI), indicating the existence of a glucose signaling pathway that relies solely on the Ras signaling pathway. Inversely, a subset of genes is activated (or repressed) by glucose in both the wild-type and tpk-w strains but is unaffected by activation of Ras2 or Gpa2, indicating the existence of a Ras2-independent glucose signaling pathway (class III). However, the vast majority of genes that respond to glucose are affected by Ras2 activation and also respond in the tpk-w background (classes I and V). This suggests that the majority of glucose-responsive genes are regulated by redundant pathways, one of which requires Ras2 and the other one(s) of which is Ras2 independent. Thus, the major transcriptional response of glucose addition diverges prior to activation of Ras2, but converges before gene activation. This is elaborated further in the Discussion.
Ras and Gpa2 Signal Exclusively through PKA
To assess the extent to which the effects on transcription of Ras2 activation are mediated by PKA, we examined the pattern of expression following activation of Ras2 in tpk-w cells compared to that in Tpk+ cells. For those genes whose induction or repression by Ras2 is exerted through PKA, the tpk-w mutations would be expected to abrogate that response. In we plot the change in expression of each gene 60 min after galactose addition to the GAL10p-RAS2V19 tpk-w strain versus the change in expression of each gene 60 min after galactose addition to the GAL10p-RAS2V19 strain. As evident, almost all genes fail to respond to Ras2 activation in the tpk-w background. Of the 789 genes (out of 4,037 analyzed) in this experiment whose expression increased by more than 2-fold at 60 min following addition of galactose to the GAL10p-RAS2V19 strain, only 16 (2%) also showed increased expression through activation of Ras2 in the tpk-w background. Similarly, of the 1,121 genes whose expression decreased by more than 2-fold following activation of Ras2 in a wild-type background, only five (0.5%) also showed decreased expression in the tpk-w background. Repetition of these experiments using cDNA microarrays and direct Northern blot analysis of candidate genes failed to confirm that expression of any gene was altered by Ras induction in a tpk-w background (data not shown). Thus, we conclude that the entirety of the transcriptional response to Ras2 activation is mediated through PKA.
Ras and Gpa2 Affect Transcription Exclusively through PKA
The results are similar for Gpa2 activation. As noted above, the response to Gpa2 activation is not as robust as that to Ras2 activation, and, as noted in , the attenuation of the response to Gpa2 induction in a tpk-w strain is not as obvious as that seen with Ras2. Of the 444 genes in this experiment whose expression increased 2-fold or more in response to Gpa2 activation in a wild-type background, 75 (17%) also showed increased expression in the tpk-w background. Similarly, of the 831 genes whose expression decreased by 2-fold or more, 24 (3%) also showed decreased expression in the tpk-w background. However, multiple replicates of this experiment using cDNA microarrays failed to identify any gene consistently altered in transcription by Gpa2 in a tpk-w background. Thus, as with Ras, the vast majority, if not all, of Gpa2-responsive genes are regulated exclusively through PKA.
Gpr1 Is Required for Efficient Glucose Response
encodes a protein structurally related to seven-transmembrane, G-protein–coupled receptors, and both biochemical and genetic evidence suggests it regulates Gpa2 activity in response to glucose (Xue et al. 1998
; Kraakman et al. 1999
; Lorenz et al. 2000
). Accordingly, to assess the role of Gpr1 in the cell's transcriptional response to glucose, we examined the global transcriptional pattern of isogenic GPR1
strains at 20-min intervals following glucose addition to glycerol-grown cells. Further, to assess the extent to which Gpr1-mediated signaling was processed through PKA, we performed a similar time course experiment with isogenic GPR1 tpk-w
and gpr1 tpk-w
strains. The full set of data is available in Table S2
. In both experiments we found that the overall transcriptional response (both induction and repression) was attenuated, although not eliminated, in the gpr1
strain relative to the GPR1
strain. For instance, for those genes whose expression changed by more than 50% following glucose addition to the GPR1 TPK
strain, the average induction or repression ratio in the gpr1
strain was approximately half that in the GPR1
strain. K-means clustering of normalized data confirmed this general view (). For instance, cluster 1, which included 470 genes highly enriched in those involved in ribosome biosynthesis, exhibited on average induced expression in the GPR1 TPK
strain following glucose addition, but no induction in the gpr1 TPK1
strain. Similar results were observed for genes in cluster 8, and induction of genes in clusters 4 and 6 was attenuated in the gpr1
strain compared to that in the GPR1
strain. Thus, these results are consistent with the hypothesis that Gpr1 participates in glucose signaling, but is not the sole mediator of that signaling.
Loss of Gpr1 Diminishes the Glucose Response
The time course data from the tpk-w strain suggest that Gpr1 might affect multiple glucose signaling pathways. If a Gpr1-initiated signal were transmitted solely through PKA, then the pattern of gene expression following glucose addition to the gpr1 tpk-w strain would be essentially identical to that observed in the GPR1 tpk-w strain. While the correlation between the expression patterns of gpr1 tpk-w and GPR1 tpk-w (r = 0.73) is higher than that between gpr1 TPK and GPR1 TPK (r = 0.65), the patterns of expression of gpr1 tpk-w and GPR1 tpk-w, as highlighted by the cluster analysis, are similar but notably distinct (particularly in clusters 2, 4, 6, and 7). Thus, these results could suggest that Gpr1 impinges on both PKA-dependent and PKA-indepen-dent signaling pathways. Alternatively, the steady-state differences between gpr1 and GPR1 strains at the onset of the experiment could render the strains differentially responsive to glucose. This issue could be resolved by appropriate conditional alleles in GPR1 and TPK.
Ras, Gpa2, and Glucose Induce Genes in Mass Accumulation and Repress Genes in Respiration and Mitochondrial Function
We have addressed the nature of the genes regulated by glucose and Ras2 in two different but related ways. First, we asked how those genes that have been annotated as performing related functions behave on average over the set of experiments. Second, we have determined whether genes performing a common function are significantly overrepresented in any cluster of coexpressed genes. Both approaches give essentially the same results.
In , we present the average level of expression of all the genes associated with the indicated function (as annotated by the Munich Information Center for Protein Sequences [MIPS] program) relative to that at time 0 in the wild-type strain. As evident, genes required for translation are upregulated by glucose and activation of Ras2 or Gpa2. This includes genes for RNA polymerase I and III subunits, cytoplasmic tRNA synthetases, rRNA and tRNA processing enzymes, translation initiation factors, and, to a slightly lesser degree, ribosomal proteins. Similarly, genes for these functional categories are highly enriched in those clusters in which expression increases following addition of glucose to wild-type or tpk-w cells or following activation of Ras2 or Gpa2 (see ). Thus, a major portion of the transcriptional restructuring following glucose addition is directed toward enhancement of the translational machinery. Somewhat surprisingly, though, this is induced not solely by increased metabolism, but at least in part by a direct response to a signaling circuit, which is mediated at least in part by Ras2.
Functional Analysis of Glucose- and Ras-Induced Expression Changes
On the other side of the coin, genes involved in oxidative respiration, including components of the TCA cycle, oxidative phosphorylation apparatus, and ubiquinone (CoQ) synthesis, and all the genes required solely for gluconeogenesis are significantly downregulated both by glucose addition and by activation of Ras or Gpa2. These functional categories of genes are significantly overrepresented in that class of coexpressed genes that are downregulated in all conditions tested (class V). Thus, Ras2-dependent and Ras2-independent repression pathways redundantly regulate the restructuring associated with conversion from respiration to fermentation.
Several groups of genes appear to be regulated by glucose exclusively through a PKA-dependent pathway. These are genes repressed by Ras2 or Gpa2 and by glucose in the wild-type strain, but not in the tpk-w strain (class VI), and include those involved in carbohydrate storage (trehalose and glycogen) and, to a large extent, in ubiquinone synthesis. A number of genes exhibit induction by glucose in an exclusively Ras2-dependent fashion and include genes involved in ribosome biogenesis.
Reciprocally, a number of genes exhibit induction by glucose in a completely Ras-independent fashion. As noted in , expression of members of class III increases monotonically following glucose addition, in contrast to the pattern seen with genes in other induction classes, in which an initial rapid increase in expression following glucose addition is followed by an immediate stabilization or downshift. This may indicate that these genes are upregulated as a consequence of the metabolic changes or growth acceleration attendant on glucose addition. The enrichment of genes involved in DNA replication in this category is consistent with this hypothesis.
Identification of Potential Transcription Factors Mediating the Response to Ras2 Activation
We have used a number of computational approaches to identify potential regulatory sequences and regulatory factors responsible for changes in gene expression in response to glucose and/or Ras2 activation. All of these approaches are based on the assumption that genes exhibiting a common expression pattern over all the experiments are more likely to share a common regulatory sequence or respond to a common transcription factor (see Supporting Information).
Several motifs (RRPE, PAC) and transcription factor-binding sites (Sfp1, Rap1, Fhl1) are associated with the class of genes induced by glucose through both a Ras-dependent and a Ras-independent pathway. Rap1- and Fhl1-binding sites have previously been associated with ribosomal protein genes (Lieb et al. 2001
; Lee et al. 2002
), and the enrichment of these sites in this class represents the high proportion of ribosomal protein genes in the clusters comprising this class. Similarly, the RRPE and PAC motifs have been associated with genes encoding elements of the translational machinery and with genes that are upregulated following overexpression of Sfp1 (Hughes et al. 2000
; Wade et al. 2001
; Jorgensen et al. 2002
). Thus, these three transcription factors and their associated motifs are potential loci through which glucose and/or Ras activates transcription of translation-related genes.
To evaluate whether the predicted motifs mediate Ras-activated transcription, we inserted each motif upstream of a reporter gene lacking any other upstream activation sequence (UAS) and then introduced the individual constructs into strains containing the inducible RAS2* or GPA2* alleles. As a positive control, we examined expression of the RPS18B promoter/enhancer region when it was fused to the reporter construct. As evident in , activation of Ras2 or Gpa2 resulted in a 3-fold increase in expression of the reporter construct, consistent with the observation that expression of this gene increased following induction of either RAS2* or GPA2* in our genome-wide expression analysis. Having confirmed the ability of this system to detect Ras-responsive promoters, we examined the ability of the Rap1-binding site or the RRPE or PAC element to enhance transcription in response to activation of the Ras pathway. As noted in Tables , , and , both the Rap1-binding site and the RRPE element yielded strong enhancer activity, especially when present in multiple copies. In contrast, the PAC element exhibited no enhancer activity. Further, the Rap1 enhancer activity increased modestly but consistently in glucose versus glycerol medium and following activation of Ras2 or Gpa2. Activation of Ras2 or Gpa2 also consistently yielded increased expression driven by the RRPE element. Finally, an MCB element provided modest enhancer activity that was further stimulated by growth on glucose but not by activation of Ras2 or Gpa2. This is consistent with the expression pattern of genes in the cluster in which the MCB motif is enriched.
Functional Analysis of Motifs: Potential Activator Elements
Functional Analysis of Motifs: Potential Repressor Elements
Functional Analysis of Motifs: PDR10 Element Is Not Ume6 Dependent
Several motifs were identified as correlated with repression by glucose and by Ras2 or Gpa2. These included binding sites for Rpn4, Ume6, Hap2/3/4, and Msn2/4 as well as several sequences of unknown association. We tested several of these motifs for their ability to mediate glucose- or Ras-induced transcriptional repression by inserting them between the CYC1 UAS and the promoter of a CYC1-lacZ reporter construct and examining expression under different growth conditions. Most of the known elements manifested modest repression activity that was not enhanced by growth on glucose or by Ras or Gpa2 activation. However, multiple copies of an Ume6-like element from PDR10 elicited strong glucose-enhanced repression activity. As evident from , the element caused 5- to 10-fold repression when cells were grown in glycerol and 500-fold repression when cells were grown in glucose. While this element exhibits some similarity to a Ume6-binding site, it does not mediate repression by Ume6. As noted in , deletion of UME6 (or RPN4, MIG1, MIG2, MSN4, PHD1, RGM1, STD1, RIM101, SFL1, or NRG1; data not shown) did not alleviate the repressive effects of this element, although this deletion eliminated repression effected by a known Ume6-binding site from CAT8. Repression by the PDR10 site was alleviated by deletion of TUP1 or SSN6. Thus, this element likely functions by recruiting the Tup1/Ssn6 repressor complex to the promoter through a specific DNA binding factor intermediate. Given the lack of correspondence between the sequence of the element and known regulatory motifs, the element likely represents a novel glucose repression mechanism.