https://orcid.org/0000-0002-2896-4296
Abstract
In budding yeast, fermentation is the most important pathway for energy production. Under low-glucose conditions, ethanol is used for synthesis of this sugar requiring a shift to respiration. This process is controlled by the transcriptional regulators Cat8, Sip4, Rds2 and Ert1. We characterized Gsm1 (glucose starvation modulator 1), a paralog of Rds2 and Ert1. Genome-wide analysis showed that Gsm1 has a DNA binding profile highly similar to Rds2. Binding of Gsm1 and Rds2 is interdependent at the gluconeogenic gene FBP1. However, Rds2 is required for Gsm1 to bind at other promoters but not the reverse. Gsm1 and Rds2 also bind to DNA independently of each other. Western blot analysis revealed that Rds2 controls expression of Gsm1. In addition, we showed that the DNA binding domains of Gsm1 and Rds2 bind cooperatively in vitro to the FBP1 promoter. In contrast, at the HAP4 gene, Ert1 cooperates with Rds2 for DNA binding. Mutational analysis suggests that Gsm1/Rds2 and Ert1/Rds2 bind to short common DNA stretches, revealing a novel mode of binding for this class of factors. Two-point mutations in a HAP4 site convert it to a Gsm1 binding site. Thus, Rds2 controls binding of Gsm1 at many promoters by two different mechanisms: regulation of Gsm1 levels and increased DNA binding by formation of heterodimers.
Introduction
In the budding yeast Saccharomyces cerevisiae, glucose is the carbon source of choice and fermentation is the major pathway for energy production, even under aerobic conditions (Crabtree effect). However, with glucose becoming scarce, ethanol produced during fermentation is used as a carbon source, a process requiring a shift to a respiration mode. Other nonfermentable carbon sources, such as lactate, acetate or glycerol, can also be used [for a review, see (1)]. These gluconeogenic carbons ultimately are incorporated into the gluconeogenic pathway and used for the production of glucose 6-phosphate, an essential metabolic precursor for numerous cellular processes (2). The shift from fermentative to nonfermentative metabolism results in profound reprogramming of gene expression (3,4). For instance, increased expression of genes for gluconeogenesis and the glyoxylate cycle is observed following a shift to ethanol, while expression of some fermentation genes is reduced under these conditions. In agreement with these changes of gene expression, glucose starvation results in a shift of histone acetylation marks from growth promoting genes to gluconeogenic and fatty acid metabolism genes (5).
Regulation of the expression of gluconeogenic genes is mainly controlled by a subclass of zinc finger proteins, the zinc cluster proteins (6). These transcriptional regulators are unique to fungi and amoeba and are characterized by the presence of a highly conserved cysteine-rich motif involved in DNA recognition. The cysteine residues coordinate folding of the DNA binding domain through binding to two zinc atoms. The DNA binding domain also contains a variable linker region involved in the DNA binding specificity and a dimerization region consisting of a coiled coil that mediates protein–protein interactions for homo- or heterodimerization (6). Zinc cluster proteins preferentially recognize CGG triplets arranged as inverted (CGGNXCCG), direct (CGGNXCGG) or everted repeats (CCGNXCGG) (6). For example, Gal4 binds to an inverted repeat consisting of two CGG triplets spaced by 11 bp (CGGN11CCG) (7), while Hap1 recognizes a direct repeat (CGGN6CGG) (8–10). Leu3 and Pdr3 recognize an everted repeat with a spacing of 4 bp (CCGN4CGG) or 0 bp (CCGCGG), respectively (11,12). A dimerization domain consisting of a coiled coil allows high-affinity DNA binding of these homodimers. Variation in the spacing (‘NX’) between the CGG triplets further increases the repertoire of binding sites (6). Finally, binding of zinc cluster proteins as monomers has also been reported (13).
The master regulator Snf1 (14–17), a homolog of AMP-activated protein kinases, is activated by low-glucose conditions leading to the phosphorylation of a number of substrates, including DNA binding proteins such as Mig1 and the zinc cluster proteins Cat8, Sip4 and Rds2 (1). Inactivation of the Mig1 repressor by phosphorylation results in derepression of a number of target genes such as CAT8. Activation of Cat8 and Sip4 by phosphorylation results in upregulation of the expression of gluconeogenic and glyoxylate cycle genes. Both Cat8 and Sip4 bind to carbon source response elements (CSREs) found in the promoters of target genes such as PCK1 and FBP1 encoding the key gluconeogenic enzymes phosphoenolpyruvate carboxykinase and fructose 1,6-bisphosphatase, respectively (18–20). Cat8 also controls the expression of SIP4 (20). Furthermore, the zinc finger (C2H2 motif) activator Adr1 controls the expression of genes involved in ethanol utilization and peroxisomal proliferation (21).
The zinc cluster genes RDS2, ERT1 and GSM1 encode highly related paralogs. For example, using the full-length Gsm1 polypeptide as a query for a BLAST search, E-values are 8 × 10−26 and 2 × 10−14 for Ert1 and Rds2, respectively. By performing phenotypic studies on zinc cluster genes, we identified a number of phenotypes associated with a deletion of the open reading frame (ORF) RDS2 such as a lack of growth with a nonfermentable carbon source (a phenotype related to the strain background) (22,23). Chromatin immunoprecipition and DNA microarray (ChIP-chip) experiments showed that Rds2 binds to only 7 promoters with cells grown in the presence of glucose, while it binds to many additional genes (total of 43) when ethanol is used as a carbon source (24). Strikingly, the genes bound by Rds2 are involved in gluconeogenesis (e.g. PCK1) and related pathways such as the glyoxylate shunt and the tricarboxylic acid cycle. Importantly, we showed that Rds2 acts as a transcriptional activator of gluconeogenic genes, while it is a repressor of the negative regulators of gluconeogenesis (24). Rds2 also controls expression of the regulatory gene HAP4 that encodes the activating and limiting subunit of the heterotetrameric activator complex Hap2/3/4/5 involved in activation of respiration genes (25). In addition, activation of Rds2 is correlated with its hyperphosphorylation by the Snf1 kinase (24).
By performing ChIP-chip analysis with the ERT1 (ethanol regulated transcription factor 1) gene product, we showed that this zinc cluster protein is an additional regulator of gluconeogenesis (26). In analogy to Rds2, Ert1 is bound to a limited number of genes under glucose conditions (total of 11; e.g. PCK1). With ethanol as a carbon source, enrichment of Ert1 is observed at many additional genes (>70) involved in gluconeogenesis and mitochondrial function (e.g. FMP43 encoding a subunit of the pyruvate mitochondrial transporter). Our results also demonstrated that Ert1 is redundant with other regulators of gluconeogenesis. Interestingly, we also showed that Ert1 acts as a repressor of PDC1 encoding pyruvate decarboxylase, an important enzyme for fermentation. Thus, during a shift from fermentation to respiration, Ert1 has dual functions: it represses a key fermentation gene, while it activates gluconeogenic genes.
The Timmers’ group has shown that, under glucose conditions, the zinc cluster protein Gsm1 (glucose starvation modulator 1) binds to a very limited number of promoters such as PCK1 and HAP4 (27). However, ChIP-chip analysis was not performed with ethanol as a carbon source. In this study, we show that, in the presence of ethanol, Gsm1 binds to over 600 promoters, including those of the gluconeogenic genes PCK1 and FBP1. Strikingly, the binding profile of Gsm1 is very similar to Rds2, suggesting that these factors bind to DNA as heterodimers. ChIP-chip analysis of Gsm1 in a strain carrying a deletion of RDS2 and the converse, Rds2 in a Δgsm1 background, showed a complex pattern of DNA binding. For example, Gsm1 and Rds2 bind to the FBP1 promoter in an interdependent manner, while at PCK1 the presence of Rds2 is required for binding of Gsm1 but not the reverse. In contrast, both factors bind independently of each other at the HAP4 promoter. These observations can be explained by the fact that Rds2 partially controls Gsm1 levels. In addition, an electrophoretic mobility shift assay (EMSA) using the purified DNA binding domains of Gsm1 and Rds2 showed that these factors bind cooperatively by formation of heterodimers. We also showed that Rds2 and Ert1, but not Gsm1, bind cooperatively to the PDC1 and HAP4 promoters. These experiments revealed a novel mode of DNA recognition for zinc cluster proteins. Our results unravel a complex pattern of DNA recognition by three paralogs.
Materials and methods
Strains and media
Strains and oligonucleotides used in this study are listed in Supplementary Tables S1 and S2, respectively. Media were prepared as described (28). Tagging of GSM1 at its C-terminus with three HA epitopes was performed by amplifying HA-URA3-HA sequences using oligos PETGSM1-1 and PETGSM1-2. The polymerase chain reaction (PCR) product was transformed in strain BY4741 (29) followed by selection on plates lacking uracil. The URA3 marker was removed by internal recombination as described (30). Double-deletion strains (SC205, SC206, SC207, SC237, SC238 SC247, SC248 and SC249) were generated using single-deletion strains (31) that were transformed with a PCR product containing the Candida glabrata HIS3 gene flanked by sequences complementary to the 5′ and 3′ regions of the GSM1 ORF followed by selection on minimal plates lacking histidine. The PCR product was obtained by using oligonucleotides GSM1-CgHIS3-1 and GSM1-CgHIS3-2 with C. glabrata DNA (strain ATCC 66032) as a template. Similarly, the double-deletion strains SC208 and SC209 were obtained by mating single-deletion strains SC102A and SC193 followed by sporulation and selection for double deletions in a BY4741 background. A similar approach was used to generate SC201 and SC202 using SC105A and SC193. The triple-deletion strains SC239 and SC240 were obtained by transforming SC202 with a PCR product generated with oligos RDS2-CgHIS3-1 and RDS2-CgHIS3-2 using C. glabrata DNA as a template. The diploid strain yCW was generated by crossing strains yCW1476 and yCW1477 (32).
Plasmids for protein expression
Plasmid pGST-Gsm1 (a.a. 1–189) was constructed by amplifying the DNA binding domain of Gsm1 by PCR using oligos ZL (GSM1)-1 and ZL (GSM1)-2 and genomic DNA isolated from strain BY4741. The PCR product was cut with BamHI and MfeI and subcloned into the plasmid pGEX-f (11) cut with BamHI and EcoRI. The expression vector pGST-Rds2 (a.a. 1–97) was obtained by cutting plasmid pGST-Rds2 (a.a. 1–152) (24) with NdeI, filling with Klenow and ligating. This plasmid allows expression of the DNA binding domain of Rds2 (a.a. 1–152) with 8 extra amino acids due to a frameshift. The plasmid pGST-Ert1 (a.a. 1–152) has been described previously (26).
Plasmids for SYRTH analysis
Plasmid pGST-Gsm1 (a.a. 1–619) was constructed by amplifying the GSM1 ORF by PCR using oligos GSM1-GIB-1 and GSM1-GIB-2, while plasmid pGST-Rds2 (a.a. 1–447) was constructed by amplifying the RDS2 ORF by PCR using oligos RDS2-GIB-1 and RDS2-GIB-2 using as a template DNA isolated from the strain BY4741. The PCR products were subcloned separately into pGEX-f (11) cut with BamHI and EcoRI using Gibson assembly. The GSM1 ORF was amplified by PCR from the plasmid GST-Gsm1 (a.a. 1–619) using the oligos SYRTH-GSM1-GIB-3 and SYRTH-GSM1-GIB4. The RDS2 ORF was amplified by PCR from the plasmid GST-Rds2 (a.a. 1–447) using the oligos SYRTH-RDS2-GIB-3 and SYRTH-RDS2-GIB-4. The PCR products were digested with DpnI and purified on a PCR column. Using plasmids pCW778 and pYL040 as vectors (32), pCW778-Gsm1, pYL040-Gsm1 and pYL040-Rds2 were constructed by Gibson assembly, while pCW778-Rds2 was obtained by using T4 DNA ligase. Empty vectors were constructed by digesting plasmids pCW778 and pYL040 (32) with SmaI and religated using T4 DNA ligase.
Spotting assays
Strains were grown overnight in YPD medium at 30°C. Cells were harvested, washed twice, resuspended in water and diluted to OD600= 1.0. Then, cells were serially diluted before being spotted on minimal plates supplemented with casamino acids (0.2%) as well as leucine, methionine, histidine and uracil (each at 0.004%) and 2% glucose or 2% of a nonfermentable carbon source (ethanol or lactate or glycerol). Plates were incubated at 30°C for 1 day (glucose) or 4–5 days (ethanol, lactate, glycerol).
ChIP-chip assays
Strains were grown in YPD to an OD600 ≈ 0.7. Cells were washed twice in water and resuspended in rich medium containing 2% glucose (YPD) or containing 2% ethanol (YEP-2% ethanol) and grown for 3 h. Cross-linking, immunoprecipitation, DNA purification, DNA labeling, hybridization and scanning were performed as described (33). The microarrays used, custom designed by Agilent Technologies, were described before (34) and contain a total of ∼180 000 Tm-adjusted 60-mer probes covering the entire yeast genome with virtually no gaps between probes.
ChIP-chip data analysis
Identification of binding sites for Rds2 and Gsm1
One round of Gaussian smoothing (standard deviation of 100 bp) was applied to the log2 binding ratio data, reporting a value every 10 bp. The maxima that are above 0.8 were identified. Significantly enriched regions were identified using the P-values associated with each probe as described (35). Finally, the maxima that overlap with the significant regions were retained and considered as significant binding sites. This method identified 632, 629, 627 and 471 binding sites for the Rds2, Gsm1, Rds2/Δgsm1 and Gsm1/Δrds2 datasets, respectively.
Overlap between datasets
The overlap between the binding sites identified in the different datasets was calculated using the University of California at Santa Cruz TableBrowser (36). The binding sites, extended to 200 bp, were used as input and the ‘Intersection’ function of the TableBrowser was used to create bed files only containing the desired overlaps.
Effect of factor deletion on occupancy
To determine the magnitude of the effect of deleting one transcription factor on the occupancy of the other, the average log2 ratio was calculated over each binding site in each dataset and these values were subtracted to create a fold change between wild-type and mutant. To do this, we first trimmed the data to keep only the data overlapping the Rds2 or Gsm1 wild-type binding sites. This was done in the TableBrowser using the ‘Intersection’ tool. For example, a file was intersected with another file to create a bedgraph (using the ‘DATA VALUE format’ as output option) containing the Rds2 data with only the values that are 200 bp centered on its 632 binding sites. This generated trimmed bedgraphs (note that these contain one value at every bp, so 200 values per binding site). Next, the values for the mutants were subtracted from those of the wild-type. The 200 values for each binding site were averaged, leading to 1 fold change value for each region. The regions were then sorted based on the fold change. Next, the ‘coordinate’ files were created to be used in VAP (37) to create heatmaps of factor occupancy over binding sites sorted on their dependence on the other factor. Heatmaps were created using VAP and Treeview (38) (to build images from outfile generated by VAP) using the above ‘coordinate’ files and the smoothed data files as inputs.
Reproducibility of the ChIP-chip experiments
Computed Pearson correlation (r) between the different replicates was >0.8 for all experiments (at 300 bp resolution) except for Gsm1 in a Δrds2 strain, which gives 0.57. The lower correlation between the replicates of this experiment could be explained by the fact that binding of Gsm1 is quite reduced in a Δrds2 strain. There is therefore a reduced signal-to-noise ratio contributing to a weaker correlation.
Gene ontology (GO) term analysis was performed using the GO Term Finder (version 0.86) implementation in the Saccharomyces Genome Database (https://www.yeastgenome.org).
Immunoblotting of whole cell extracts
To assess the bulk levels of Rds2 and Gsm1 within WT/Δgsm1 and WT/Δrds2 cells, respectively, at different times after ethanol treatment, immunoblotting experiments were performed on whole cell extracts obtained as follows. Cells were grown in YPD to an OD600 ≈ 0.7, washed twice in water, resuspended in YEP-2% ethanol and grown for 1.5 and 3 h. For each sample, 50 ml of cell cultures were pelleted and resuspended in 700 μl lysis buffer (50 mM HEPES–KOH, pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, protease and phosphatase inhibitors). Cells were then lysed by bead beating for 5 min. Soluble extracts were recovered by centrifugation. Sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) loading buffer, 5× (250 mM Tris–HCl, pH 6.8, 10% SDS, 50% glycerol, 0.05% bromophenol blue, 5% β-mercaptoethanol) was added to a final 1× concentration to 40 μl of each sample, incubated at 95°C for 5 min, briefly centrifuged and analyzed by SDS–PAGE and western blot analysis using mouse monoclonal anti-HA (Santa Cruz, sc-7392, 1:1000 dilution) and mouse monoclonal anti-Pgk1 (Abcam, ab113687, 1:30 000 dilution) antibodies. Blots were developed using donkey anti-mouse IRDye 680RD antibody (LI-COR Biosciences, LIC-926-68072) according to the manufacturer’s instructions and scanned on the Odyssey infrared imaging system.
Electrophoretic mobility shift assays
Expression, purification of GST proteins and thrombin cleavage to remove the GST moiety were performed as described (11,39). For DNA–protein binding, the reaction was performed in 4 mM Tris–HCl (pH 8), 40 mM NaCl, 25 μg/ml sonicated salmon sperm DNA and 10 pmol/ml labeled probe. DNA binding quantification was performed with a phosphorimager (Storm 840, Amersham Biosciences). DNA–protein signals were subtracted from background (lanes lacking protein) and ratios of mutant to wild-type signals were determined.
Results
We were interested in better elucidating the role of Gsm1, in particular with cells grown in the presence of ethanol by performing ChIP-chip experiments. To this end, GSM1 was tagged with three HA epitopes at its natural chromosomal location using homologous recombination. In agreement with van Bakel et al. (27), we observed very limited binding of Gsm1 (<10 promoters) under glucose conditions (data not shown). For example, Gsm1 was detected at the promoters of the PCK1 and the HAP4 genes. In contrast to glucose conditions, Gsm1 is found at 629 promoters in the presence of ethanol, including PCK1 and FBP1. Strikingly, Gsm1 and Rds2 show a highly similar pattern of binding with an 81% overlap (Figure 1A). GO analysis showed similar hits for Gsm1 and Rds2 (Supplementary Table S3), such as generation of precursor metabolites and energy as well as transmembrane transport, suggesting that these factors perform common processes. The similar binding of Gsm1 and Rds2 suggests that these two factors may bind to DNA as heterodimers. To test this hypothesis, we performed ChIP-chip analysis with tagged Gsm1 expressed in a Δrds2 strain and the converse, tagged Rds2 expressed in a Δgsm1 background. Results show that binding of Gsm1 is largely dependent on Rds2, while Rds2 relies on Gsm1 to a lesser extent for binding to DNA (Figure 1B). For example, binding of Gsm1 and Rds2 is interdependent at the FBP1 promoter, while at PCK1 Rds2 is required for binding of Gsm1 but not the reverse (Figure 1C). In addition, Gsm1 and Rds2 bind to GDB1 and GUP2 to a large extent independently of each other (Table 1). Rds2 (but not Gsm1) binds to the PDC1 promoter, while Gsm1 (but not Rds2) is found at some other promoters (Table 1). A list of selected promoters bound by Gsm1 and/or Rds2 is given in Table 1. In summary, of the six possible types of binding, five were observed (Table 2). Overall, when looking at common Rds2/Gsm1 binding sites, results show that Gsm1 is dependent on Rds2 for binding at 87% of promoters. In contrast, Rds2 DNA occupancy relies on Gsm1 at only 22% sites (Figure 1B). Thus, a complex interplay between Gsm1 and Rds2 is observed. Requirement of Rds2 for Gsm1 DNA binding could be explained by the fact that, in a Δrds2 strain, Gsm1 would be expressed at lower levels. To address this possibility, we performed western blot analysis using total extracts obtained at 0, 1.5 and 3 h after a shift from glucose to ethanol. HA-Rds2 levels remained unchanged during the time course as assayed in a wild-type and a Δgsm1 strain (Figure 2, top panel). In contrast, Gsm1-HA levels increased in a time-dependent manner in a wild-type strain, in agreement with the observation that GSM1 messenger RNA levels are increased in the presence of ethanol (4). However, Gsm1-HA levels were clearly reduced at 1.5 and 3 h in a Δrds2 strain. Thus, a requirement of Rds2 for binding of Gsm1 can be (partly) explained by reduced Gsm1 levels upon deletion of RDS2. However, our EMSAs clearly show interdependent binding of Gsm1 and Rds2, as well as Ert1 and Rds2 (see below). In addition, we also observe similar binding of Gsm1 at GDB1, GUP2 and MDM31 in a Δrds2 strain (Table 1).
Gsm1 and Rds2 genome-wide binding sites: complex interplay between Gsm1 and Rds2. ChIP-chip experiments were performed with HA-tagged Gsm1 in wild-type and Δrds2 strains. Conversely, experiments were done with HA-tagged Rds2 in wild-type and Δgsm1 strains. All ChIP experiments were performed with biological duplicates that were highly correlated and the figure shows results from their weighted averaging (see the ‘Materials and methods’ section). (A) Venn diagram showing the extensive overlap between the Rds2 (n = 632) and Gsm1 (n = 629) binding sites. (B) Heatmap of Gsm1 (panels 1 and 2) and Rds2 (panels 4 and 5) binding in a 2-kb window around their respective binding sites. The data for each factor are shown in wild-type cells and in cells where the other factor gene was deleted. A differential binding map is also shown for both factors (panels 3 and 6). The binding sites were ordered based on the magnitude of the effect of the deletion (top binding sites are the least affected by the deletion and the bottom ones are the most affected). (C) Binding of Gsm1 and Rds2 is shown at the FBP1 (left) and PCK1 promoters (right). Gray rectangles correspond to ORFs with arrows showing the direction of transcription. Strains used for ChIP-chip analysis are indicated on the left. The Y axis (log2 scale) gives the enrichment (binding) for a given factor. (D) Binding of Gsm1 and Rds2 at a region of chromosome XI. Binding of these factors at the HAP4 promoter is also shown (right side).
Selected Gsm1 and Rds2 targets under ethanol conditions
| Enrichment (given in log2 values) for | |||||
|---|---|---|---|---|---|
| Gene | Function (after yeastgenome.org) | RDS2 | RDS2 in a Δgsm1 strain | GSM1 | GSM1 in a Δrds2 strain |
| PCK1 | Phosphoenolpyruvate carboxykinase; key enzyme in gluconeogenesis | 5.7 | 6.8 | 3 | 1.6 |
| PDC1 | Major of three pyruvate decarboxylase isozymes; key enzyme in alcoholic fermentation | 6.6 | 6.9 | NS | NS |
| PTK1 | Putative serine/threonine protein kinase; regulates spermine uptake | 5.2 | 2.8 | 3.7 | 2.1 |
| MAE1 | Mitochondrial malic enzyme; catalyzes the oxidative decarboxylation of malate to pyruvate | 5.2 | 6 | NS | NS |
| AQR1 | Plasma membrane transporter of the major facilitator superfamily | 5.1 | 5.4 | NS | NS |
| HAP4 | Subunit of the heme-activated, glucose-repressed Hap2/3/4/5 CCAAT-binding complex | 5 | 3 | 5.6 | 2.5 |
| SOM1 | Subunit of the mitochondrial inner membrane peptidase | 4.9 | 2.1 | 4.9 | 3.1 |
| GAT1 | Transcriptional activator of nitrogen catabolite repression genes | 4.6 | 5 | 2.5 | 1.2 |
| OLE1 | Delta(9) fatty acid desaturase; required for monounsaturated fatty acid synthesis | 4.5 | 2.2 | 4.6 | 1.9 |
| FBP1 | Fructose 1,6-bisphosphatase; key regulatory enzyme in the gluconeogenesis pathway | 4.1 | 1 | 4.4 | 1 |
| API2 | Putative protein of unknown function | 4.5 | 3.3 | 4.7 | 2.9 |
| YDR095C | Dubious ORF | 4.4 | 2.2 | 4.5 | 2.1 |
| LCL1 | Putative protein of unknown function; deletion mutant is fluconazole resistant | 4.4 | 2.5 | 4.4 | 2.1 |
| YPL062W | Putative protein of unknown function | 4.2 | 3 | 4.2 | 2.4 |
| ATP3 | Gamma subunit of the F1 sector of mitochondrial F1F0 ATP synthase | 2.8 | 1 | 1.3 | NS |
| BRP1 | Putative protein of unknown function | 4.1 | 4.3 | 3.9 | 2.1 |
| PUS2 | Mitochondrial tRNA:pseudouridine synthase | 3.9 | 3 | 4 | 2.5 |
| GLN1 | Glutamine synthetase; synthesizes glutamine from glutamate and ammonia | 3.8 | 3.5 | 3.7 | 2 |
| CIT1 | Citrate synthase; catalyzes the condensation of acetyl coenzyme A and oxaloacetate to form citrate | 3.5 | 2 | 3.5 | 1.4 |
| CYC1 | Cytochrome c, isoform 1; electron carrier of mitochondrial intermembrane space | 3.4 | 1.7 | 3.5 | 1.7 |
| GDB1 | Glycogen debranching enzyme; required for glycogen degradation | 1.5 | 1.4 | 1.5 | 1.2 |
| GUP2 | Probable membrane protein; possible role in proton symport of glycerol | 1.5 | 1.6 | 2.8 | 2.2 |
| MDM31 | Mitochondrial protein that may have a role in phospholipid metabolism | NS | NS | 1.5 | 1.3 |
| Enrichment (given in log2 values) for | |||||
|---|---|---|---|---|---|
| Gene | Function (after yeastgenome.org) | RDS2 | RDS2 in a Δgsm1 strain | GSM1 | GSM1 in a Δrds2 strain |
| PCK1 | Phosphoenolpyruvate carboxykinase; key enzyme in gluconeogenesis | 5.7 | 6.8 | 3 | 1.6 |
| PDC1 | Major of three pyruvate decarboxylase isozymes; key enzyme in alcoholic fermentation | 6.6 | 6.9 | NS | NS |
| PTK1 | Putative serine/threonine protein kinase; regulates spermine uptake | 5.2 | 2.8 | 3.7 | 2.1 |
| MAE1 | Mitochondrial malic enzyme; catalyzes the oxidative decarboxylation of malate to pyruvate | 5.2 | 6 | NS | NS |
| AQR1 | Plasma membrane transporter of the major facilitator superfamily | 5.1 | 5.4 | NS | NS |
| HAP4 | Subunit of the heme-activated, glucose-repressed Hap2/3/4/5 CCAAT-binding complex | 5 | 3 | 5.6 | 2.5 |
| SOM1 | Subunit of the mitochondrial inner membrane peptidase | 4.9 | 2.1 | 4.9 | 3.1 |
| GAT1 | Transcriptional activator of nitrogen catabolite repression genes | 4.6 | 5 | 2.5 | 1.2 |
| OLE1 | Delta(9) fatty acid desaturase; required for monounsaturated fatty acid synthesis | 4.5 | 2.2 | 4.6 | 1.9 |
| FBP1 | Fructose 1,6-bisphosphatase; key regulatory enzyme in the gluconeogenesis pathway | 4.1 | 1 | 4.4 | 1 |
| API2 | Putative protein of unknown function | 4.5 | 3.3 | 4.7 | 2.9 |
| YDR095C | Dubious ORF | 4.4 | 2.2 | 4.5 | 2.1 |
| LCL1 | Putative protein of unknown function; deletion mutant is fluconazole resistant | 4.4 | 2.5 | 4.4 | 2.1 |
| YPL062W | Putative protein of unknown function | 4.2 | 3 | 4.2 | 2.4 |
| ATP3 | Gamma subunit of the F1 sector of mitochondrial F1F0 ATP synthase | 2.8 | 1 | 1.3 | NS |
| BRP1 | Putative protein of unknown function | 4.1 | 4.3 | 3.9 | 2.1 |
| PUS2 | Mitochondrial tRNA:pseudouridine synthase | 3.9 | 3 | 4 | 2.5 |
| GLN1 | Glutamine synthetase; synthesizes glutamine from glutamate and ammonia | 3.8 | 3.5 | 3.7 | 2 |
| CIT1 | Citrate synthase; catalyzes the condensation of acetyl coenzyme A and oxaloacetate to form citrate | 3.5 | 2 | 3.5 | 1.4 |
| CYC1 | Cytochrome c, isoform 1; electron carrier of mitochondrial intermembrane space | 3.4 | 1.7 | 3.5 | 1.7 |
| GDB1 | Glycogen debranching enzyme; required for glycogen degradation | 1.5 | 1.4 | 1.5 | 1.2 |
| GUP2 | Probable membrane protein; possible role in proton symport of glycerol | 1.5 | 1.6 | 2.8 | 2.2 |
| MDM31 | Mitochondrial protein that may have a role in phospholipid metabolism | NS | NS | 1.5 | 1.3 |
NS, not significant.
Selected Gsm1 and Rds2 targets under ethanol conditions
| Enrichment (given in log2 values) for | |||||
|---|---|---|---|---|---|
| Gene | Function (after yeastgenome.org) | RDS2 | RDS2 in a Δgsm1 strain | GSM1 | GSM1 in a Δrds2 strain |
| PCK1 | Phosphoenolpyruvate carboxykinase; key enzyme in gluconeogenesis | 5.7 | 6.8 | 3 | 1.6 |
| PDC1 | Major of three pyruvate decarboxylase isozymes; key enzyme in alcoholic fermentation | 6.6 | 6.9 | NS | NS |
| PTK1 | Putative serine/threonine protein kinase; regulates spermine uptake | 5.2 | 2.8 | 3.7 | 2.1 |
| MAE1 | Mitochondrial malic enzyme; catalyzes the oxidative decarboxylation of malate to pyruvate | 5.2 | 6 | NS | NS |
| AQR1 | Plasma membrane transporter of the major facilitator superfamily | 5.1 | 5.4 | NS | NS |
| HAP4 | Subunit of the heme-activated, glucose-repressed Hap2/3/4/5 CCAAT-binding complex | 5 | 3 | 5.6 | 2.5 |
| SOM1 | Subunit of the mitochondrial inner membrane peptidase | 4.9 | 2.1 | 4.9 | 3.1 |
| GAT1 | Transcriptional activator of nitrogen catabolite repression genes | 4.6 | 5 | 2.5 | 1.2 |
| OLE1 | Delta(9) fatty acid desaturase; required for monounsaturated fatty acid synthesis | 4.5 | 2.2 | 4.6 | 1.9 |
| FBP1 | Fructose 1,6-bisphosphatase; key regulatory enzyme in the gluconeogenesis pathway | 4.1 | 1 | 4.4 | 1 |
| API2 | Putative protein of unknown function | 4.5 | 3.3 | 4.7 | 2.9 |
| YDR095C | Dubious ORF | 4.4 | 2.2 | 4.5 | 2.1 |
| LCL1 | Putative protein of unknown function; deletion mutant is fluconazole resistant | 4.4 | 2.5 | 4.4 | 2.1 |
| YPL062W | Putative protein of unknown function | 4.2 | 3 | 4.2 | 2.4 |
| ATP3 | Gamma subunit of the F1 sector of mitochondrial F1F0 ATP synthase | 2.8 | 1 | 1.3 | NS |
| BRP1 | Putative protein of unknown function | 4.1 | 4.3 | 3.9 | 2.1 |
| PUS2 | Mitochondrial tRNA:pseudouridine synthase | 3.9 | 3 | 4 | 2.5 |
| GLN1 | Glutamine synthetase; synthesizes glutamine from glutamate and ammonia | 3.8 | 3.5 | 3.7 | 2 |
| CIT1 | Citrate synthase; catalyzes the condensation of acetyl coenzyme A and oxaloacetate to form citrate | 3.5 | 2 | 3.5 | 1.4 |
| CYC1 | Cytochrome c, isoform 1; electron carrier of mitochondrial intermembrane space | 3.4 | 1.7 | 3.5 | 1.7 |
| GDB1 | Glycogen debranching enzyme; required for glycogen degradation | 1.5 | 1.4 | 1.5 | 1.2 |
| GUP2 | Probable membrane protein; possible role in proton symport of glycerol | 1.5 | 1.6 | 2.8 | 2.2 |
| MDM31 | Mitochondrial protein that may have a role in phospholipid metabolism | NS | NS | 1.5 | 1.3 |
| Enrichment (given in log2 values) for | |||||
|---|---|---|---|---|---|
| Gene | Function (after yeastgenome.org) | RDS2 | RDS2 in a Δgsm1 strain | GSM1 | GSM1 in a Δrds2 strain |
| PCK1 | Phosphoenolpyruvate carboxykinase; key enzyme in gluconeogenesis | 5.7 | 6.8 | 3 | 1.6 |
| PDC1 | Major of three pyruvate decarboxylase isozymes; key enzyme in alcoholic fermentation | 6.6 | 6.9 | NS | NS |
| PTK1 | Putative serine/threonine protein kinase; regulates spermine uptake | 5.2 | 2.8 | 3.7 | 2.1 |
| MAE1 | Mitochondrial malic enzyme; catalyzes the oxidative decarboxylation of malate to pyruvate | 5.2 | 6 | NS | NS |
| AQR1 | Plasma membrane transporter of the major facilitator superfamily | 5.1 | 5.4 | NS | NS |
| HAP4 | Subunit of the heme-activated, glucose-repressed Hap2/3/4/5 CCAAT-binding complex | 5 | 3 | 5.6 | 2.5 |
| SOM1 | Subunit of the mitochondrial inner membrane peptidase | 4.9 | 2.1 | 4.9 | 3.1 |
| GAT1 | Transcriptional activator of nitrogen catabolite repression genes | 4.6 | 5 | 2.5 | 1.2 |
| OLE1 | Delta(9) fatty acid desaturase; required for monounsaturated fatty acid synthesis | 4.5 | 2.2 | 4.6 | 1.9 |
| FBP1 | Fructose 1,6-bisphosphatase; key regulatory enzyme in the gluconeogenesis pathway | 4.1 | 1 | 4.4 | 1 |
| API2 | Putative protein of unknown function | 4.5 | 3.3 | 4.7 | 2.9 |
| YDR095C | Dubious ORF | 4.4 | 2.2 | 4.5 | 2.1 |
| LCL1 | Putative protein of unknown function; deletion mutant is fluconazole resistant | 4.4 | 2.5 | 4.4 | 2.1 |
| YPL062W | Putative protein of unknown function | 4.2 | 3 | 4.2 | 2.4 |
| ATP3 | Gamma subunit of the F1 sector of mitochondrial F1F0 ATP synthase | 2.8 | 1 | 1.3 | NS |
| BRP1 | Putative protein of unknown function | 4.1 | 4.3 | 3.9 | 2.1 |
| PUS2 | Mitochondrial tRNA:pseudouridine synthase | 3.9 | 3 | 4 | 2.5 |
| GLN1 | Glutamine synthetase; synthesizes glutamine from glutamate and ammonia | 3.8 | 3.5 | 3.7 | 2 |
| CIT1 | Citrate synthase; catalyzes the condensation of acetyl coenzyme A and oxaloacetate to form citrate | 3.5 | 2 | 3.5 | 1.4 |
| CYC1 | Cytochrome c, isoform 1; electron carrier of mitochondrial intermembrane space | 3.4 | 1.7 | 3.5 | 1.7 |
| GDB1 | Glycogen debranching enzyme; required for glycogen degradation | 1.5 | 1.4 | 1.5 | 1.2 |
| GUP2 | Probable membrane protein; possible role in proton symport of glycerol | 1.5 | 1.6 | 2.8 | 2.2 |
| MDM31 | Mitochondrial protein that may have a role in phospholipid metabolism | NS | NS | 1.5 | 1.3 |
NS, not significant.
Types of binding modes observed for Gsm1 and Rds2
| Type of binding | RDS2 in a wild-type strain | RDS2 in a Δgsm1 strain | GSM1 in a wild-type strain | GSM1 in a Δrds2 strain |
|---|---|---|---|---|
| Interdependent, e.g. FBP1 | + | +/− | + | +/− |
| Largely independent of each other, e.g. GDB1 and GUP2 | + | + | + | + |
| Rds2 only, e.g. PDC1 | + | + | +/− | +/− |
| Gsm1 only, e.g. MDM31 | +/− | +/− | + | + |
| Rds2-dependent, e.g. PCK1 | + | + | + | +/− |
| Gsm1-dependent, not observed | + | +/− | + | + |
| Type of binding | RDS2 in a wild-type strain | RDS2 in a Δgsm1 strain | GSM1 in a wild-type strain | GSM1 in a Δrds2 strain |
|---|---|---|---|---|
| Interdependent, e.g. FBP1 | + | +/− | + | +/− |
| Largely independent of each other, e.g. GDB1 and GUP2 | + | + | + | + |
| Rds2 only, e.g. PDC1 | + | + | +/− | +/− |
| Gsm1 only, e.g. MDM31 | +/− | +/− | + | + |
| Rds2-dependent, e.g. PCK1 | + | + | + | +/− |
| Gsm1-dependent, not observed | + | +/− | + | + |
Types of binding modes observed for Gsm1 and Rds2
| Type of binding | RDS2 in a wild-type strain | RDS2 in a Δgsm1 strain | GSM1 in a wild-type strain | GSM1 in a Δrds2 strain |
|---|---|---|---|---|
| Interdependent, e.g. FBP1 | + | +/− | + | +/− |
| Largely independent of each other, e.g. GDB1 and GUP2 | + | + | + | + |
| Rds2 only, e.g. PDC1 | + | + | +/− | +/− |
| Gsm1 only, e.g. MDM31 | +/− | +/− | + | + |
| Rds2-dependent, e.g. PCK1 | + | + | + | +/− |
| Gsm1-dependent, not observed | + | +/− | + | + |
| Type of binding | RDS2 in a wild-type strain | RDS2 in a Δgsm1 strain | GSM1 in a wild-type strain | GSM1 in a Δrds2 strain |
|---|---|---|---|---|
| Interdependent, e.g. FBP1 | + | +/− | + | +/− |
| Largely independent of each other, e.g. GDB1 and GUP2 | + | + | + | + |
| Rds2 only, e.g. PDC1 | + | + | +/− | +/− |
| Gsm1 only, e.g. MDM31 | +/− | +/− | + | + |
| Rds2-dependent, e.g. PCK1 | + | + | + | +/− |
| Gsm1-dependent, not observed | + | +/− | + | + |
Rds2 controls Gsm1 levels. Top panel: wild-type and Δgsm1 strains expressing tagged Rds2 were used to prepare extracts at 0, 1.5 and 3 h following a shift to ethanol for western blot analysis. Pgk1 was used as a loading control. Bottom panel: same as above except that tagged Gsm1 was used and tested in a wild-type and a Δrds2 strain. The western blots were generated three times and a representative blot is shown.
Gsm1 and Rds2 physically interact with each other
To start investigating whether Rds2 and Gsm1 may regulate each other’s ability to bind DNA, we first tested whether Gsm1 and Rds2 physically interact with each other. Since Rds2 and Gsm1 are DNA binding proteins (see below), use of the classical two-hybrid assay is not optimal. As a result, we used a modified two-hybrid assay called SYRTH (for Ste11/Ste50 yeast related two-hybrid) that has been shown to work with transcriptional regulators (32). This system allows detection of cytoplasmic protein–protein interactions and it relies on the observation that Ste11 must interact with Ste50 through SAM domains for activation of the high osmotic stress pathway. In this modified two-hybrid assay, factor X is fused to Ste11 lacking the SAM domain, while factor Y is fused to Ste50 also lacking this domain. Interaction of protein X with protein Y brings together Ste11 and Ste50 for activation of the high osmotic pathway and growth in hyperosmotic medium, the readout of the assay. A strain carrying deletions of the STE11 and STE50 genes was transformed with various combinations of expression vectors. Transformants were grown on selective medium, diluted and spotted on minimal plates containing galactose for induction of the expression of the proteins of interest. In the absence of osmotic stress (0 mM NaCl), growth was similar for all combinations of expression vectors tested (Figure 3, left panel). In the presence of 500 mM NaCl, no growth was observed with cells expressing mutants STE11 and STE50 or a single fusion protein. However, co-expression of Gsm1–Ste11 with Rds2–Ste50 resulted in growth under osmotic stress (Figure 3, right panel). Thus, Gsm1 and Rds2 interact with each other in this assay. However, no growth was observed when testing the converse: Rds2–Ste11 with Gsm1–Ste50. This may be due to alternate conformations of fusion proteins that prevent an interaction. Since the SYRTH assay relies on cytoplasmic interactions, our results strongly suggest that Gsm1 and Rds2 can interact with each other even in the absence of DNA.
Gsm1 interacts with Rds2 in a modified two-hybrid assay. Strain yCW (Supplementary Table S1) was transformed with empty expression vectors for mutated STE11 or STE50 or with vectors for expression of fusion proteins. Cells were grown in minimal selective medium lacking uracil and histidine and diluted to OD600 = 0.1. Cells were further diluted 25, 100, 400 and 2000 times and 10 μl of cells were spotted on minimal plates containing 2% galactose with (right panel) or without 500 mM NaCl (left panel). Expression vectors are indicated on the left part of the figure. Spotting in this figure is a representative experiment out of three.
Deletion of GSM1 along with other genes involved in gluconeogenesis does not result in growth defects
We were interested in determining whether deletion of GSM1 results in reduced growth with a nonfermentable carbon source (ethanol, lactate, glycerol). We also tested strains carrying deletions of GSM1 in combination with other genes encoding regulators of gluconeogenesis as well as a triple-deletion strain (Δgsm1 Δrds2 Δert1). Strains were grown in rich medium containing glucose (YPD), washed with water, serially diluted and spotted on plates containing glucose or ethanol as a carbon source. Under glucose conditions, growth was similar for all tested strains (Figure 4, first panel). As expected from previous observations (18,26), a severe growth defect was observed with a strain lacking CAT8 under ethanol conditions. However, deletion of GSM1 by itself or in combination with other regulatory genes of gluconeogenesis did not affect growth. Similarly, deletion of the three paralogs (GSM1, RDS2 and ERT1) did not impair growth (Figure 4, second panel). Similar results were obtained when testing growth on the nonfermentable carbon sources glycerol and lactate (Figure 4, third and fourth panels). These results suggest a high level of redundancy among transcriptional regulators of gluconeogenesis.
Gsm1 does not interact genetically with genes encoding regulators of gluconeogenesis. Cells were grown overnight in rich medium (YPD) and washed twice in water. Cells were diluted 25, 100, 400 and 1600 times and 10 μl of cells were spotted on minimal plates supplemented with casamino acids (0.2%) as well as leucine, methionine, histidine and uracil (each at 0.004%) and 2% glucose (left panel), 2% ethanol (second panel), 2% lactate (third panel) and 2% glycerol (fourth panel). Plates were incubated at 30°C for 1 day (glucose), 5 days (ethanol) or 4 days (lactate and glycerol). For double- and triple-deletion strains, two independent isolates were tested. Spotting in this figure is a representative experiment out of four.
Gsm1 by itself binds in vitro to the FBP1 promoter and cooperatively with Rds2
In order to better understand the mode of binding of the three paralogs, a series of EMSAs were performed with the DNA binding domains of Gsm1 (a.a. 1–189), Rds2 (a.a. 1–97) and Ert1 (a.a. 1–152). For simplicity, we will refer to these polypeptides as Gsm1, Rds2 and Ert1. The DNA binding domains were bacterially expressed as fusions with GST, purified on glutathione beads and the GST moiety removed by cleavage with thrombin. The binding site of Gsm1 at the FBP1 promoter was identified by performing EMSAs. Gsm1 was used with a set of overlapping probes encompassing the FBP1 promoter from −800 to −350 bp relative to the ATG. Results show that Gsm1 binds only to probe ‘J’ that overlaps CSRE #1 (Figure 5A and B). No binding of Gsm1 was detected with the probe ‘OP’ containing the CSRE and sequences upstream of this element or with the CSRE itself (Figure 5B). However, when using a probe (‘X’) containing the CSRE and sequences extending at the 3′ end (Figure 5A), a DNA–protein complex was detected (Figure 5B). In summary, the binding site for Gsm1 maps to a 33-bp fragment (probe ‘X’) overlapping CSRE #1.
Gsm1 binds in vitro to the FBP1 promoter. (A) The FBP1 promoter is schematically shown on top with rectangles corresponding to CSREs. Probes (A–K) were used in an EMSA with the purified DNA binding domain of Gsm1 (a.a. 1–189). (B) An EMSA performed with probes J, OP, CSRE #1 and X. Sequence of probe X is given in Figure 6A; also see (26). This is a representative experiment out of three.
We were interested in determining the DNA sequences recognized by Gsm1 more precisely. Zinc cluster proteins bind preferentially to CGG triplets (6). Five CGG triplets are found in probe ‘X’ (Figure 6A). In order to define the triplets important for Gsm1 binding, probes containing mutations in these triplets were used in an EMSA (Figure 6B). Mutations in CGG triplets 1, 4 and 5 did not affect binding of Gsm1 (Figure 6B, lanes 6, 12 and 14). However, mutating triplet 2 or 3 abolished binding of Gsm1 (Figure 6B, lanes 8 and 10). These results allowed mapping with more accuracy the DNA site recognized by Gsm1: two CGG triplets, with one contained within the CSRE and one located downstream of the CSRE.
A short DNA element is sufficient for binding of Gsm1 by itself or cooperatively with Rds2 at the FBP1 promoter. (A) Sequence of the FBP1 probe ‘X’ is shown on the top of the figure with CGG triplets indicated with arrows and designated ‘a’ to ‘e’. The CSRE is shown as a gray area. Lowercase letters correspond to nucleotides that were mutated. (B) An EMSA was performed with the purified DNA binding domain of Gsm1 (a.a. 1–189) (2 μl) with probe X and mutants (‘a’ to ‘e’) with nucleotide changes indicated on top of the panel. (C) An EMSA was performed with the purified DNA binding domains of Gsm1 (a.a. 1–189) and Rds2 (a.a. 1–97) using probe X. Lane 1: probe by itself. Lanes 2–5: 0.025, 0.5, 1.0 and 2.0 μl of Gsm1 (a.a. 1–189). Lanes 6–9: 0.5, 1.0, 2.0 and 4.0 μl of Rds2 (a.a. 1–97). Lanes 10–13: same as lanes 6–9 except that 0.5 μl of Gsm1 was also added. Lanes 14–17: same as lanes 2–5 except that 4.0 μl of Rds2 was also added. This is a representative experiment out of three.
Our ChIP-chip data demonstrated that binding of Gsm1 and Rds2 to the FBP1 promoter is interdependent in vivo (Figure 1C). We tested whether this observation could be recapitulated in vitro using an EMSA. With increasing amount of Gsm1, a DNA–protein complex is observed only at the highest concentration used in the EMSA (Figure 6C, lane 5), while no Rds2–DNA complex is seen with this protein by itself (Figure 6C, lanes 6–9). In contrast, a DNA–protein complex is observed when mixing Gsm1 and Rds2 (Figure 6C, lanes 10–13 and 14–17). For example, a strong binding is observed in the presence of Gsm1 and Rds2, while no complex is detected with the same amounts of Gsm1 or Rds2 by themselves (Figure 6C, compare lane 16 with lanes 4 and 9), suggesting cooperative binding of Gsm1 and Rds2. In summary, Gsm1 and Rds2 bind cooperatively to a region of the FBP1 promoter.
Mutations that prevent binding of Gsm1 also abolish cooperative binding with Rds2
We further refined our analysis of sequences important for binding of Gsm1 to DNA by testing additional mutations (Figure 7A and B). Results show that mutations at positions ‘a’, ‘b’, ‘d’, ‘e’, ‘f’, ‘g’ and ‘o’ to ‘s’ did not significantly affect Gsm1 binding (Figure 7B; for quantification, see Figure 7D). In contrast, mutations at positions ‘h’, ‘i’, ‘k’ and ‘m’ almost completely abolished DNA binding of Gsm1, while mutations at positions ‘c’, ‘l’ and ‘n’ reduced DNA binding by at least 2-fold. When combining these data with those of Figure 6B, results show that eight nucleotides are important for DNA binding of Gsm1 with the sequence CCGGNGtTa (nucleotides in lowercase letters corresponding to positions less critical for Gsm1 binding). We then determined which nucleotides are important for cooperative binding of Gsm1 and Rds2 (Figure 7C; for quantification, see Figure 7D). Unexpectedly, scanning mutations revealed that the cooperative binding pattern of Gsm1 and Rds2 was very similar to the one observed with Gsm1 by itself. For example, mutations ‘h’, ‘i’, ‘k’ and ‘m’ that were critical for binding of Gsm1 also greatly reduced cooperative binding of Gsm1 and Rds2. However, mutation at position ‘c’ reduced binding of Gsm1 by 2-fold without affecting cooperative binding (Figure 7D). Our analysis of Gsm1 and Rds2 suggests that these factors do not bind to adjacent sites but rather to a short stretch (9 bp) of common DNA sequences. Similar observations were made with Rds2 and Ert1 at the PDC1 and HAP4 promoters (see below).
Mutational analysis suggests that Gsm1 and Rds2 do not bind at FBP1 to adjacent sites but rather to a short common DNA stretch. (A) Sequence of the FBP1 probe ‘Z1’ (a shorter version of probe ‘X’) is shown on the top of the figure with CGG triplets indicated with arrows and numbered 1–4. The CSRE is shown as a gray area. Lowercase letters (‘a’ to ‘s’) correspond to nucleotides that were mutated. Black dots correspond to nucleotides that were shown to be critical for binding of Gsm1 (Figure 5). (B) An EMSA was performed with the purified DNA binding domain of Gsm1 (a.a. 1–189) with probe ‘Z1’ and mutants (‘a’ to ‘s’) with nucleotide changes indicated on top of the panel. (C) An EMSA was performed with the purified DNA binding domains of Gsm1 (a.a. 1–189) and Rds2 (a.a. 1–97) using probe ‘Z1’. Lane 1: probe by itself. Lanes 2 and 3, 0.25 μl Gsm1 (a.a. 1–189). and 4.0 μl Rds2 (a.a. 1–97), respectively. Lanes 4–23: 0.25 μl of Gsm1 (a.a. 1–189) and and 4.0 μl Rds2 (a.a. 1–97). (D) Quantitative binding of Gsm1 (black rectangles) or Gsm1–Rds2 (gray rectangles) to mutants of probe ‘Z1’ relative to a wild-type probe. Data are an average of three independent experiments.
Cooperative binding in vitro of Ert1 and Rds2 at the PDC1 promoter
As stated above, Gsm1, Rds2 and Ert1 are highly conserved paralogs raising the possibility that Rds2 may favor binding of Ert1 at specific promoters, as observed with Gsm1. Using ChIP-chip analysis, we previously showed that Ert1 and Rds2 (but not Gsm1, this study) are detected at PDC1, a gene encoding the fermentation enzyme pyruvate decarboxylase (24,26). Thus, these two factors may act together for binding at this promoter. We have shown that the DNA binding domain of Ert1 recognizes a sequence located at position −780 to −761 relative to the ATG of the PDC1 gene (26). We performed an EMSA using a probe corresponding to these DNA sequences (probe ‘H’) and the purified DNA binding domains of Ert1 and Rds2. With increasing concentrations of Rds2, a smear is observed at the edge of the lanes, indicative of a nonstable Rds2–DNA complex (Figure 8, lanes 2–5). With Ert1, a weak signal is observed only at the highest concentration of protein used (Figure 8, lane 9). In contrast, mixing of Rds2 and Ert1 results in binding at concentrations not observed in the presence of a single protein (e.g. compare lanes 4, 8 and 12 in Figure 8). Thus, Rds2 binds cooperatively with Gsm1 at the FBP1 promoter as well as with Ert1 at the PDC1 promoter.
![Ert1 and Rds2 bind cooperatively to the PDC1 promoter. An EMSA was performed with the purified DNA binding domains of Ert1 (a.a. 1–152) and Rds2 (a.a. 1–97) using PDC1 probe ‘H’ [sequence shown in Figure 9A; also see (26)]. Lane 1: probe alone. Lanes 2–5: 0.5, 1.0, 2.0 and 4.0 μl of Rds2 (a.a. 1–97). Lanes 6–9: 0.025, 0.05, 0.1 and 0.2 μl of Ert1 (a.a. 1–152). Lanes 10–13: same as lanes 2–5 except that 0.1 μl of Ert1 (a.a. 1–152) was also added. Lanes 14–17: same as lanes 6–9 except that 4.0 μl of Rds2 (a.a. 1–97) was also added. This is a representative experiment out of two.](https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/nar/52/5/10.1093_nar_gkad1185/2/m_gkad1185fig8.jpeg?Expires=1747875532&Signature=HIm1Dyvi3kt5g29UzMthwPh4s71oU~P14i~EJEMFAOEIygAnaAVEdeK~3ls9q183f1qcxvfOl8nw7hK-g56wsQwl9S3yvGMNDm3zXJKaMnrCLwBitfXC9oc95bbKht4JH7y8WzifCho6VOJJ8DeiXdMawgJ-HYmULNZF8bgvAotmjxRfwT4mQI7sZUfUe3PhTQt-34C8~3JspwdWAxny4vBJxj3tBI1GmbYTFQ9rA-xXm1-qSnNM9I3NvwnKfZ3WF82GVOPOtaVSTTGDfByVIxate36uNju58CWHOPkNlbOpd9HrSjZBZVTxbJIlTodPLgPv937RFz2H9lnWmBG5Ww__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
Ert1 and Rds2 bind cooperatively to the PDC1 promoter. An EMSA was performed with the purified DNA binding domains of Ert1 (a.a. 1–152) and Rds2 (a.a. 1–97) using PDC1 probe ‘H’ [sequence shown in Figure 9A; also see (26)]. Lane 1: probe alone. Lanes 2–5: 0.5, 1.0, 2.0 and 4.0 μl of Rds2 (a.a. 1–97). Lanes 6–9: 0.025, 0.05, 0.1 and 0.2 μl of Ert1 (a.a. 1–152). Lanes 10–13: same as lanes 2–5 except that 0.1 μl of Ert1 (a.a. 1–152) was also added. Lanes 14–17: same as lanes 6–9 except that 4.0 μl of Rds2 (a.a. 1–97) was also added. This is a representative experiment out of two.
Three CGG triplets are found in probe ‘H’ (Figure 9A). In agreement with our previous observations (26), the EMSA shows that CGG triplet 2 is critical for binding of Ert1 (Figure 9B, lane 8, position ‘g’). Similarly, mutations at positions ‘e’, ‘f’ and ‘j’ prevented binding of Ert1, while a mutation at position ‘h’ reduced its binding (Figure 9B). We performed a similar mutational analysis under conditions where Ert1 and Rds2 bind cooperatively to DNA (Figure 9C). Strikingly, the pattern of binding was highly similar to the one observed with Ert1 by itself. However, while mutations at positions ‘k’ and ‘l’ had no effect on binding of Ert1 by itself, cooperative binding showed some reduced binding. This analysis strongly suggests that Ert1 and Rds2 do not recognize adjacent DNA sites but rather a common DNA stretch, as observed with Gsm1 and Rds2 at the FBP1 promoter.
Mutational analysis suggests that Ert1 and Rds2 do not bind at PDC1 to adjacent sites but rather to a short common DNA stretch. (A) Sequence of the PDC1 probe ‘H’ is shown on the top of the figure with CGG triplets indicated with arrows and numbered 1–3. Lowercase letters (‘a’ to ‘l’) correspond to nucleotides that were mutated. (B) An EMSA was performed with probe ‘H’ and mutants, as indicated on top of the figure; 0.4 μl of Ert1 (a.a. 1–152) was used for the EMSA. (C) Lanes 1–13: same as panel (B) except that 0.1 μl of Ert1 (a.a. 1–152) and 2.0 μl of Rds2 (a.a. 1–97) were used for the EMSA. Lanes 14 and 15: 0.1 μl of Ert1 (a.a. 1–152) and 2.0 μl of Rds2 (a.a. 1–97), respectively. This is a representative experiment out of two for Ert1 in combination with Rds2.
Cooperative binding in vitro of Ert1 and Rds2 at the HAP4 promoter
We then focused on the HAP4 promoter where Gsm1 and Rds2 can bind to some extent DNA independently of each other (Figure 1C and Table 1). Our previous ChIP-chip analysis showed that Ert1 also binds to this promoter (26). With a series of overlapping probes, we scanned the HAP4 promoter from −901 to −715 bp relative to the ATG. No binding was detected when using Gsm1 or Rds2 by themselves, while two sites were identified for Ert1 (data not shown). We focused on one site located at −855 to −822 bp relative to the ATG (probe ‘C’ in Figure 10). With Rds2 by itself, no DNA complex was detected (Figure 10A, lanes 2–5), while an Ert1–DNA complex was observed at the highest concentration of protein used (Figure 10A, lane 9). Upon mixing of Ert1 and Rds2, a complex was observed at concentrations where no complex was observed with single proteins (e.g. compare lanes 4 and 8 with lanes 12 and 16). We looked at other combinations of proteins. Results show that Gsm1 does not bind cooperatively with Rds2 at probe ‘C’ (Figure 10B), while similar observations were made with Ert1 in combination with Gsm1 (Figure 10C). In summary, cooperative binding is observed only with Ert1 and Rds2 at this site of the HAP4 promoter.
Rds2 and Ert1 (but not Gsm1) bind cooperatively to a DNA site in the HAP4 promoter. EMSAs were performed with probe ‘C’ located at the HAP4 promoter at position −855 to −822 bp relative to the ATG. The DNA sequence of the probe is shown in Figure 11A. (A) Lane 1: probe alone. Lanes 2–5: 0.5, 1.0, 2.0 and 4.0 μl of Rds2 (a.a. 1–97). Lanes 6–9: 0.025, 0.05, 0.1 and 0.2 μl of Ert1 (a.a. 1–152). Lanes 10–13: same as lanes 2–5 except that 0.1 μl of Ert1 (a.a. 1–152) was also added. Lanes 14–17: same as lanes 6–9 except that 0.5 μl of Rds2 was also added. (B) Lane 1: probe alone. Lanes 2–5: 0.1, 0.2, 0.5 and 1.0 μl of Gsm1 (1–189). Lanes 6–9: 0.25, 0.5, 1.0 and 2.0 μl of Rds2 (1–97). Lanes 10–13: same as lanes 6–9 except that 0.5 μl of Gsm1 (1–189) was also added. Lanes 14–17: same as lanes 6–9 except that 0.5 μl of Gsm1 (a.a. 1–189) was also added. (C) Lanes 1–4: 0.025, 0.05, 0.1 and 0.2 μl of Ert1 (a.a. 1–152). Lanes 5–8: 0.1, 0.2, 0.5 and 1.0 μl of Gsm1 (a.a. 1–189). Lanes 9–12: same as lanes 1–4 except that 0.5 μl of Gsm1 (a.a. 1–189) was also added. Lanes 13–16: same as lanes 5–8 except that 0.1 μl of Ert1 (a.a. 1–152) was also added. This is a representative experiment out of two.
We then determine nucleotides that are important for Ert1 binding to probe ‘C’. A series of mutants ‘a’ to ‘p’ (Figure 11A) were tested in an EMSA with Ert1 by itself (Figure 11B). Results show that mutations at positions ‘g’, ‘h’, ‘i’ and ‘k’ abolished binding of Ert1, while a mutation at position ‘j’ greatly reduced binding. We performed a similar experiment to determine which nucleotides are important for cooperative binding of Ert1 and Rds2 (Figure 11C). The pattern of binding is very similar to what was observed with Ert1 by itself. The only difference is with mutation at position ‘j’ where a stronger binding is observed with mixing of Ert1 and Rds2, although the signal is weaker as compared to a wild-type probe (compare lanes 1 and 11 in Figure 11C). In summary, our analysis of the FBP1, PDC1 and HAP4 promoters does not support a model where the proteins of interest would bind to adjacent DNA sites. Our results rather favor a model where Rds2 assists Gsm1 or Ert1 for DNA binding via a common DNA site.
Mutational analysis suggests that Ert1 and Rds2 do not bind at HAP4 to adjacent sites but rather to a short common DNA stretch. (A) Sequence of the HAP4 probe ‘C’ is shown on the top of the figure with CGG triplets indicated with arrows and numbered 1–3. Lowercase letters (‘a’ to ‘p’) correspond to nucleotides that were mutated. (B) An EMSA was performed with probe ‘C’ and mutants, as indicated on top of the figure; 0.3 μl of Ert1 (a.a. 1–152) was used for the EMSA. (C) Lanes 1–17: same as panel (B), except that 0.25 μl of Ert1 (a.a. 1–152) and 0.5 μl of Rds2 (a.a. 1–97) were used for the EMSA. Lanes 18 and 19: 0.25 μl of Ert1 (a.a. 1–152) and 0.5 μl of Rds2 (a.a. 1–97), respectively. This is a representative experiment out of three.
DNA specifies the formation of heterodimers
Our results show that the formation of specific heterodimers differs according to the promoter examined. We tested whether DNA itself specifies the formation of different heterodimeric species. Since Ert1 binds to FBP1, PDC1 and HAP4 (26), we focused on Gsm1 using the HAP4 probe ‘C’. Figure 12A shows an alignment of binding sites at the FBP1, HAP4 and PDC1 promoters. The core sequence CCCGGA is conserved between the FBP1 and the HAP4 sites. As expected from results described above (e.g. Figure 7), Gsm1 binds by itself or in combination with Rds2 to probe ‘Z1’ of the FBP1 promoter (Figure 12B, lanes 3–5). Even though ChIP-chip analysis detected Gsm1 at HAP4 (Figure 1D and Table 1), this protein does not bind to probe ‘C’ by itself or in combination with Rds2 (Figure 12B, lanes 8–10). A likely explanation for this observation is that Gsm1 binds to a different site at the HAP4 promoter. Downstream of the conserved core sequence, GTT and AAA triplets are present at the FBP1 and HAP4 sites, respectively (Figure 12A). Mutating AAA of the HAP4 probe to GTT (probe HFH-A) resulted in binding of Gsm1 by itself or in combination with Rds2 (Figure 12B, lanes 13–15). Similar results were obtained with the double mutant gAt (probe HFH-B, where A was not mutated) (Figure 12B, lanes 18–20). In conclusion, DNA by itself specifies the formation of different heterodimeric species.
DNA specifies the formation of heterodimeric species. (A) Alignment of binding sites identified in the FBP1, HAP4 and PDC1 promoters. Sequences of probes ‘Z1’, ‘C’ and ‘H’ are given in Figures 7A, 11A and 9A, respectively. Conserved nucleotides are highlighted. Mutations in the HAP4 ‘C’ site made to resemble an FBP1 ‘Z1’ site (probes HFH-A and HFH-B) are highlighted in gray. (B) Amounts of Gsm1 (a.a. 1–189) and Rds2 (a.a. 1–97) as well as the probes used in the EMSA are indicated at the bottom of the figure. This is a representative experiment out of two.
Discussion
This study focused on three paralog transcription factors, the zinc cluster proteins Gsm1, Rds2 and Ert1. A number of yeast transcriptional regulators, such as zinc cluster proteins, have been shown to bind DNA as heterodimers [Pdr1/Pdr3, Pdr1/Stb5 and Oaf1/Pip2 (40–42)]. However, to our knowledge, there is no report of yeast heterodimers’ identification on a genome-wide scale. Our ChIP-chip analysis of Gsm1 and Rds2 showed a highly similar binding pattern, but it was more complex than initially anticipated with five different modes of binding (Table 2). Gsm1 and Rds2 bind in an interdependent way to the FBP1 promoter, an observation that could be recapitulated in vitro with the purified DNA binding domains of these factors in an EMSA (Figure 6). However, Rds2 is required for binding of Gsm1 to the PCK1 promoter but not the reverse, while these factors bind to a large extent independently of each other to the GDB1 and the GUP2 promoters (Table 1). The large dependence on Rds2 for Gsm1 DNA binding can be explained by the fact that Rds2 controls expression of Gsm1 (Figure 2) as well as cooperative binding, as shown with EMSAs. We also showed that another zinc cluster factor Ert1 binds in vivo (26) and in vitro (Figure 7) to the PDC1 promoter. EMSA showed that Ert1 cooperates with Rds2 (but not Gsm1) for binding to probe ‘C’ of the PDC1 promoter (Figure 9), although we cannot exclude Gsm1/Rds2 cooperative binding at a different location of the HAP4 promoter. In addition, the zinc cluster protein Sut1 also binds in vivo to the HAP4 promoter (43). Gsm1 may cooperate with Sut1 or, alternatively, it may bind by itself to this promoter. In analogy to HAP4, we showed that Rds2 binds cooperatively with Ert1 to the PDC1 promoter (Figure 8), while our ChIP-chip analysis did not detect Gsm1 at this gene.
Unexpectedly, our EMSA analysis showed that Gsm1 can bind in vitro to DNA upon a slight increase (2-fold) of its concentration (Figure 6C, compare lanes 4 and 5). It remains to determine whether Gsm1 binds to DNA as a monomer or a homodimer, but we favor the latter possibility (see below). Upon mixing Rds2 and Gsm1, we did not observe the formation of a complex with a different mobility that would provide a direct proof of heterodimerization. One strategy used to show the formation of homodimers (or heterodimers) is to mix short and long forms of the protein(s) of interest resulting in the formation of a DNA complex of intermediate mobility as compared to short–short and long–long forms (41,44). However, this strategy was not successful with Gsm1 and Rds2 using fusions to three epitopes of HA or MYC or even to a large protein (β-galactosidase) (data not shown). However, the observation that Gsm1 and Rds2 interact with each other in a modified two-hybrid assay (Figure 2), together with cooperative binding in vitro, strongly suggests the formation of heterodimers.
As stated above, Gsm1 binds in vitro to probe ‘Z1’ of the FBP1 promoter but not to probe ‘C’ of the HAP4 promoter, while Ert1 recognizes both probes (Figures 7 and 10–12 and data not shown). However, changing two nucleotides of the HAP4 probe to better match the FBP1 probe allowed binding of Gsm1 by itself or in combination with Rds2 (Figure 12) strongly suggesting that DNA specifies the formation of a given heterodimeric species. What is the mode of formation of different heterodimers among Gsm1, Rds2 and Ert1 for cooperative binding? Many transcriptional regulators have been shown to bind as heterodimers with each factor binding to adjacent DNA sequences called half-sites. For example, a number of mammalian nuclear receptors (e.g. retinoid acid receptor/retinoid X receptor, vitamin D receptor/retinoid X receptor) have monomers occupying half-DNA sites (45). Regarding yeast zinc cluster proteins, no structures of heterodimers have been reported. However, crystal structures of the DNA binding domains of Gal4, Ppr1, Put3, Hap1 and Leu3 have shown that these zinc cluster proteins interact with adjacent sites for the formation of homodimers (7,9,12,46,47).
In Figure 13, we propose a model for DNA binding of Gsm1 by itself or in combination with Rds2 at FBP1 and related sites. As stated above, Leu3 and Pdr3 recognize an everted repeat with a spacing of 4 bp (CCGN4CGG) and 0 bp (CCGCGG), respectively (11,12). Two overlapping CGG triplets in opposite directions (CCGG) are essential for Gsm1 binding (Figure 6). This corresponds to an even more compacted version of the Pdr3 and Leu3 DNA sites. Additional nucleotides (G and A) downstream of the top CGG triplet are also essential for DNA recognition. Regarding the CGG triplet located on the bottom strand, mutational analysis showed that only one nucleotide upstream of this triplet affects binding (2-fold) of Gsm1 (Figure 6B). These observations suggest that Gsm1 binds as a homodimer to an imperfect everted repeat. In contrast, mutations of the nucleotides upstream of the CGG triplets did not affect binding of Gsm1/Rds2 (Figure 6C). Similarly, we could not identify mutations downstream of the top CGG triplet that would affect binding of the Gsm1/Rds2 heterodimer but not Gsm1 by itself. Similar results were obtained at PDC1 and HAP4 (Figures 8 and 10). In summary, our data do not favor a mode of DNA binding where Gsm1 and Rds2 or Ert1 and Rds2 heterodimers would recognize adjacent half-sites. Does Rds2 contact DNA? With Rds2 by itself, no well-defined Rds2–DNA complex was observed in our various EMSAs. However, at high Rds2 concentrations, a smear is observed indicative of an unstable complex (e.g. see Figure 9C, lane 15). Thus, Rds2 has limited DNA binding capacity but acts as a modulator of DNA binding by interacting with Gsm1 or Ert1. It is likely that the weak DNA binding of Rds2 induces a conformational change of Gsm1 resulting in increased binding of these two factors. However, our data do not allow distinguishing between this model or one that would not involve a conformational change. Our unpublished data show that a shorter version of Rds2 (a.a. 1–81 instead of a.a. 1–97) does not allow cooperative binding with Gsm1 (a.a. 1–187). This suggests that a.a. 82–97 are critical for cooperative binding maybe by allowing an Rds2–Gsm1 interaction. These data suggest a novel type of DNA recognition for cooperative binding of zinc cluster proteins.
Model for binding of Gsm1 and Rds2. Gsm1 and Rds2 are shown bound to different DNA sites with nucleotides involved in Gsm1 binding in bold characters. Dark and light highlighted sectors correspond to nucleotides essential and nucleotides less critical for Gsm1 binding, respectively. Arrows correspond to overlapping CGG triplets found in opposite directions (an everted repeat). The triple black lines correspond to dimerization domains. (A) Gsm1 bound by itself to an FBP1 site. (B) Gsm1 bound as a heterodimer with Rds2 to an FBP1 site. (C) Gsm1 does not bind to HAP4 due to nucleotide differences (‘X’) at that site. (D) Gsm1 bound to a hybrid HAP4–FBP1 site. Mutations corresponding to nucleotides found at the FBP1 site are shown by lowercase letters.
Heterodimer formation increases transcription factor specificity, binding affinity to DNA and diversity of target sites. As our results have shown, Gsm1 and Ert1 by themselves are capable of binding DNA independently of each other, but in some cases the Gsm1–Rds2 heterodimer increases the efficiency of binding, recognizing novel DNA elements that Gsm1 cannot recognize by itself (e.g. PCK1). Moreover, we have identified a novel mode of DNA recognition for zinc cluster proteins with binding of two factors at a common DNA site. Heterodimerization allows the formation of complex regulatory networks essential for gene expression and, as a result, yeast can better respond to environmental cues like reduced glucose levels triggering a switch from fermentation to a respiratory mode. In summary, our results reveal a novel mode of transcriptional regulation with the formation of different heterodimers at specific promoters through recognition of a short and common DNA element. With over 175 000 known or putative zinc cluster proteins (https://www.supfam.org/SUPERFAMILY/cgi-bin/genome.cgi?sf=57701), our study will be invaluable to characterize the mode of action of these transcriptional regulators in other fungal species.
Data availability
All microarray data have been deposited into Gene Expression Omnibus archive under accession number GSE225348.
Supplementary data
Supplementary Data are available at NAR Online.
Acknowledgements
We thank Drs Cunle Wu and Malcolm Whiteway for providing strains and plasmids used in the SYRTH assay. We also thank Dr Stéphane Laporte (McGill University) for useful comments.
Funding
Natural Sciences and Engineering Research Council of Canada (to B.T.); Fonds de recherche du Québec-Nature et technologies (to B.T.); Canadian Institutes of Health Research [PJT-162334 and PJT-156383 to F.R.]. Funding for open access charge: Research Institute of the McGill University Health Centre.
Conflict of interest statement. None declared.
References
Author notes
The first two authors should be regarded as Joint First Authors.
Comments