Mig1 and Mig2 are proteins with similar zinc fingers that are required for glucose repression of SUC2 expression. Mig1, but not Mig2, is required for repression of some other glucose-repressed genes, including the GAL genes. A second homolog of Mig1, Yer028, appears to be a glucose-dependent transcriptional repressor that binds to the Mig1-binding sites in the SUC2 promoter, but is not involved in glucose repression of SUC2 expression. Despite their functional redundancy, we found several significant differences between Mig1 and Mig2: (1) in the absence of glucose, Mig1, but not Mig2, is inactivated by the Snf1 protein kinase; (2) nuclear localization of Mig1, but not Mig2, is regulated by glucose; (3) expression of MIG1, but not MIG2, is repressed by glucose; and (4) Mig1 and Mig2 bind to similar sites but with different relative affinities. By two approaches, we have identified many genes regulated by Mig1 and Mig2, and confirmed a role for Mig1 and Mig2 in repression of several of them. We found no genes repressed by Yer028. Also, we identified no genes repressed by only Mig1 or Mig2. Thus, Mig1 and Mig2 are redundant glucose repressors of many genes.
THE yeast Saccharomyces cerevisiae has adopted mechanisms to ensure that it efficiently utilizes glucose, its preferred carbon source. One way it achieves this is to repress transcription of genes whose products are dispensable in cells growing on high levels of glucose (for reviews, see Johnston and Carlson 1992; Trumbly 1992; Ronne 1995), such as genes required for utilization of carbon sources other than glucose (e.g., GAL, SUC, MAL), for gluconeogenesis (e.g., FBP1, PCK1), for enzymes of the Krebs cycle and respiration (e.g., CYC1, COX6), for high-affinity glucose transporters (e.g., HXT2), and for genes involved in sporulation, proteolysis, and peroxisomal function.
Repression of many glucose-repressed genes is executed by Mig1, a zinc-finger DNA-binding protein (Nehlin and Ronne 1990; Kleinet al. 1998). It represses transcription by recruiting the general repressors Ssn6 and Tup1 (Keleheret al. 1992; Treitel and Carlson 1995). Mig1 function is regulated at the level of its nuclear localization: in the absence of glucose, it is located in the cytoplasm; addition of glucose causes it to move rapidly into the nucleus (DeVitet al. 1997). Snf1, a protein kinase required for expression of many glucose-repressed genes (Celenza and Carlson 1984, 1986; Schuller and Entian 1987), inhibits Mig1 in the absence of glucose, probably by phosphorylating it, which causes Mig1 to move to the cytoplasm (Treitel and Carlson 1995; DeVitet al. 1997).
Mig2 and Yer028 contain two Cys2His2 zinc fingers very similar to those of Mig1 (Figure 1), and to the zinc fingers of the mammalian Krox20/Egr and Wilms’ tumor proteins (Bohmet al. 1997). Like Mig1, Mig2 represses transcription in response to glucose through Ssn6 and Tup1 (Lutfiyya and Johnston 1996). Mig1 and Mig2, like their mammalian homologs, bind to a GC-rich sequence (Nehlin and Ronne 1990; Lutfiyya and Johnston 1996). Mig1 has an additional requirement for an adjacent AT-rich sequence (Lundinet al. 1994). It is not clear if Mig2 has a similar sequence requirement for DNA binding. Neither the DNA-binding site, nor the function of Yer028 is known.
Mig1-binding sites reside in the promoters of many glucose-repressed genes, and a role for Mig1 in repression of several of these has been confirmed (for review, see Kleinet al. 1998): GAL1 (Nehlinet al. 1991; Flick and Johnston 1992), GAL4 (Griggs and Johnston 1991), SUC2 (Nehlin and Ronne 1990; Vallier and Carlson 1994), CAT8 (Hedgeset al. 1995), and MAL61, MAL62, and MAL63 (Huet al. 1995; Kleinet al. 1996; Wang and Needleman 1996). Mig1 appears to be the sole repressor of the GAL genes, because glucose repression of the GAL genes is almost completely relieved in a strain lacking MIG1 (Griggs and Johnston 1991; Nehlinet al. 1991; Flick and Johnston 1992). Mig1 and Mig2 collaborate to repress SUC2 expression (Vallier and Carlson 1994; Lutfiyya and Johnston 1996). Several other glucose-repressed genes contain Mig1-binding sites in their promoters, but the expression of several of them is not affected by disrupting MIG1 (Mercadoet al. 1991; Ronne 1995). Mig2 and Yer028 are good candidates for regulators of these genes because of the similarity of their zinc-finger domains to those of Mig1 (Figure 1; Bohmet al. 1997). In an attempt to understand the specific roles played by these three repressors, we analyzed several aspects of their function and identified the genes they regulate.
MATERIALS AND METHODS
Yeast strains, media, and transformations: All strains used in this study are derived from S288C (Table 1). Yeast cells were grown at 30° in standard medium: YEP (rich) medium, or synthetic (minimal) medium lacking the appropriate amino acids (Roseet al. 1990). Yeast transformations were done as described by Schiestl and co-workers (Schiestl and Gietz 1989; Schiestlet al. 1993).
Gene disruptions were done using PCR products, as previously described (Baudinet al. 1993; Niedenthalet al. 1996). Briefly, yeast were transformed to His+ or G418R with a HIS3- or KanMX-containing PCR product that included at each end 45 bp upstream and downstream of the region to be disrupted. This resulted in replacement of the target gene (from translation START to STOP codons) with HIS3, or KanMX. HSL1, YCL024, and GIN4 were disrupted in diploid strain YM4919 using the following oligos: OM1150 and OM1151 (HSL1), OM1153 and OM1154 (YCL024), OM1156 and OM1157 (GIN4). The gene disruption was verified by a PCR with primers flanking the disrupted gene (OM1152, HSL1; OM1155, YCL024; OM1158, GIN4) and a primer in HIS3 (OM483). The mig1::KanMX disruption was made as described above using oligos OL937 and OL938; primers OL939 and OM1117 were used in a PCR to confirm the correct disruption. Disruptions of MIG2 and YER028 were described previously (Lutfiyya and Johnston 1996). Other gene disruptions included snf1::URA3, made using the construct pBM2225 cut with BamHI and HindIII, and ura3::LYS2, made with pBM2265 cut with HindIII.
Plasmids: Standard procedures for the manipulation of plasmid DNA and transformation into bacteria were followed (Sambrooket al. 1989). Escherichia coli DH5α was used as the host for all plasmids. All promoter fusions to lacZ (except pBM3190, M. DeVit, unpublished data) were made as follows: the promoter region of each gene (approximately 1 kb of sequence upstream from the ATG, including the ATG) was amplified from genomic DNA by a PCR with the oligonucleotides listed in Table 2. Several independent PCR products for each were combined, digested with BamHI and EcoRI, and cloned between the BamHI and EcoRI sites of YEp357R (pBM2640; Myerset al. 1986). The YDR516 and HXK1 PCR products were cut with BamHI only and inserted into the BamHI site of YEp357R. All constructs were sequenced to confirm that the ATG was in-frame with lacZ.
The Mig1-binding sites were subcloned into the reporter plasmid pBM2832 (Ozcan and Johnston 1996), which has the upstream activation sequence (UAS) fragment of the LEU2 gene and the TATA box of HIS3 (with part of the HIS3 coding region) fused to lacZ in YEp356. Two single-stranded oligonucleotides consisting of each Mig1-binding site (OM1041+OM1042, OM1043+OM1044, OM268+OM286, OM270+ OM271; see Table 2) were annealed for 15 min at 37°. The double-stranded oligos, which have EcoRI “sticky” ends, were then cloned into the EcoRI site of pBM2832. The resulting inserts were sequenced to determine the number of inserts. The LexA1-87-Yer028-encoding plasmid, pBM3613, was made as follows. The YER028 coding region (starting at the ATG) was amplified from genomic DNA by a PCR with oligonucleotides OM1341 and OM1342 as primers. Several independent PCR products were combined, digested with EcoRI and BamHI, and cloned between the EcoRI and BamHI sites of pSH2-1 (vector containing the lexA DNA-binding domain, amino acids 1-87; see Hanes and Brent 1989).
The plasmid containing Mig2 (amino acids 81-381) fused to GFP-β-galactosidase (pBM3691) was created by recombination in yeast. MIG2 was amplified by the PCR using oligonucleotides OM1538 and OM1539 as primers and pBM3091 as template. This product was amplified in a PCR using primers OM1540 and OM1493 to provide homologous sequence for recombination (OM1493 adds 45 nucleotides identical to sequence 5′ of the BamHI site in pBM3098 that is immediately 5′ of the GFP coding sequence in this plasmid; OM1540 adds 42 nucleotides identical to the sequence 3′ to the BamH1 site). This PCR product and BamHI cut pBM3098 (provided by Jim Haseloff, MRC Laboratory of Molecular Biology) were cotransformed into yeast strain YM4342. Plasmids from extracts of Ura+ transformants were transformed into bacteria for amplification and analysis to confirm the presence of the MIG2 coding sequence. Plasmid pBM3691 was retransformed into yeast strain YM4342 for visualization of GFP by fluorescent microscopy (DeVitet al. 1997).
The CYC1-lacZ reporters used in Figure 4 are pLGΔ312s (Guarente and Hoar 1984), which has lacZ under control of the wild-type CYC1 promoter, and JK1621 (Keleheret al. 1992), which is identical to pLGΔ312s, except with four LexA-binding sites inserted 5′ of the UAS. In Figure 5, the GAL1-lacZ reporters are pLR1Δ1, which contains the lacZ gene under the control of the GAL1 promoter with the UAS deleted (Westet al. 1984) and pSH18-8, which is derived from pLR1Δ1, but has four LexA-binding sites replacing the UAS (R. Brent, personal communication).
Enzyme assays: β-Galactosidase assays were carried out in permeabilized cells grown to mid-log phase as described previously (Yocumet al. 1984), except that cell densities (OD600) and ONPG (OD420) were quantified in microtiter plates on a Molecular Devices (Sunnyvale, CA) plate reader. Yeast were grown in minimal medium lacking uracil (or uracil and histidine) and containing either 4% glucose (repressing conditions) or 5% glycerol and 0.05% glucose (nonrepressing conditions) to mid-log phase (OD600 ∼1.0). Activities are given in Miller units and are the average of at least four assays of at least two independent transformants. Cells were prepared from exponentially growing cultures for invertase assays. Repressed cultures were grown overnight in media containing 4% glucose; for derepression, cells were shifted to media containing 5% glycerol and 0.05% glucose for 2.5 hr (YEP media) or 3 hr (synthetic media). Secreted invertase was assayed in whole cells as described by Goldstein and Lampen (1975) and Celenza and Carlson (1984), except that ABTS [2,2-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid)] was substituted for o-dianisidine (0.53 mg per reaction), and acid was not added at the end of the reaction. Reactions were incubated at room temperature for 30-60 min to allow the color to develop. Tubes were spun 2-3 min and absorbance was measured at 420 nm in microtiter plates on a Molecular Devices plate reader.
In vitro binding site selection: Oligonucleotide OM1159 (5′GTAAAACGACGGCCAGTGGATCCataaaaatgcggggaaGAAT TCCCTGTGTGAAATTGTTATCC 3′) was designed so that it would contain a 16-nucleotide degenerate internal region flanked on the 5′ end by a 23-nucleotide sequence of the M13 forward primer followed by a BamHI site, and at the 3′ end by an EcoRI site followed by a 26-nucleotide sequence complementary to the M13 reverse primer. The internal region was synthesized by adding at each synthesis step a mixture containing 79% of the wild-type nucleotide of the SUC2-A site and 7% of each of the other 3 nucleotides. A total of 10 pmol of OM1159 was labeled and converted to double strands in a 20-μl reaction containing (final concentration): 1× Taq buffer (Boehringer Mannheim, Indianapolis), 50 μm 3 dNTP mix (minus A), 4 μm dATP, 10 pmol reverse primer (OM558), 20 μCi of [α-32P]dATP, 5 units Taq DNA polymerase (Boehringer Mannheim). This was incubated for 1 min at 94°, 3 min at 51°, and 9 min at 72° for one cycle. The reaction was chased with 50 μm cold dATP for 9 min at 72° and purified on a NucTrap push column (Stratagene, La Jolla, CA). The labeled double-stranded DNA was then purified on a 10% polyacrylamide gel and eluted from the acrylamide by agitation in a buffer containing 0.5 m ammonium acetate, 10 mm MgAc, 1 mm EDTA, and 0.1% SDS, at 37° for 4-15 hr. After elution, the DNA was ethanol precipitated, dried, and resuspended in 100 μl of ddH2O. The specific activity was 3 × 105 cpm/pmol.
For the gel shift assays, approximately 1 × 105 cpm of DNA (0.3 pmol) was incubated with 1-5 μl of Mig2 or Mig1 for 10 min at 4° in a 25-μl reaction in the following buffer: 50 mm Tris-HCl (pH 7.5), 10% glycerol, 35 mm MgCl2, 200 mm KCl, 10 μm ZnSO4, 2.5 mm DTT, and 0.5 μg of poly(dI:dC). Mig1 and Mig2 proteins were produced in E. coli as previously described (Lutfiyya and Johnston 1996). Protein-DNA complexes were separated on a nondenaturing 6% (30:0.8) polyacrylamide gel (containing 3% glycerol) run at 4° at 13 V/cm in 0.5× Tris-borate-EDTA buffer (Sambrooket al. 1989), after which the gel was exposed to X-OMAT (Eastman Kodak, Rochester, NY) film for 10 hr at -70°. The shifted band was excised and the DNA eluted as described above. After elution, the DNA was extracted once with phenol/chloroform, once with chloroform, ethanol precipitated, washed once with 70% ethanol, dried, and resuspended in 20 μl of ddH2O.
To amplify the DNA recovered from the gel shift experiment, 1 μl was used in a PCR using primers OM259 (universal forward primer) and OM558 (universal reverse primer). Reactions were incubated for 1 min at 94°, 1 min at 51°, and 1 min at 72° for 25 cycles. The PCR products were digested with BamHI and EcoRI, cloned between the BamHI and EcoRI sites of pBluescript SK+ (Stratagene), and the resulting plasmids were sequenced on an ABI 373A automated sequencer using dye-labeled terminators.
Protein preparation: The entire Yer028 protein was fused to the bacterial MalE protein by amplifying YER028 in a PCR with oligonucleotides OM1341 and OM1342 as primers, combining several independent reactions, digesting them with EcoRI and BamHI, and inserting the fragment between the EcoRI and BamHI sites of pMAL (New England Biolabs, Beverly, MA), generating pBM3643. Cells were grown, and protein was purified on a maltose affinity column according to the manufacturer’s protocol.
DNA-binding assays: A labeled probe of the SUC2 promoter region was made by combining PCR products from several independent PCR reactions (using as primers OM526 and OM534), digesting with EcoRI, purifying the digested product on a nondenaturing 10% polyacrylamide gel, and labeling with [32P]dATP by filling in with the Klenow fragment of DNA polymerase I (Sambrooket al. 1989). Oligonucleotide probes were annealed, labeled with [32P]dATP by filling in the sticky ends with the Klenow fragment of DNA polymerase I, and purified on a NucTrap push column (Stratagene). The oligonucleotides used were SUC2-A (OM1041 and OM1042) and SUC2-B (OM1043 and OM1044); Mig1 binds to both of these sites (Nehlin and Ronne 1990; Lutfiyya and Johnston 1996). The gel shift assay was carried out as described above.
Computer methods: searching the yeast genome for Mig1-binding sites: A pattern search program, RNABOB (http://genome.wustl.edu/eddy/#rnabob/), was used to search for Mig1 binding sites in a database consisting of only the regions between the predicted open reading frames (ORFs) in the yeast genome (S. Eddy, personal communication). The search was limited to sites similar to the SUC2-A site: AAAAA T GCGGGG. The criteria used were: (1) any sites that matched exactly in the GC-box and allowed up to three changes in the AT-box, or (2) any sites that had a GC-box sequence of GTGGGG, or CCGGGG, and any one change in the AT-box. This allowed for flexibility in the AT-box, while maintaining a strict requirement in the GC-box for sites that Mig1 is known to bind to (Lundinet al. 1994).
Screening an array of yeast genes for Mig1, Mig2, and Yer028 regulated genes: Total yeast RNA was isolated from YM4797 (wild type) and YM4804 (mig1Δ mig2Δ yer028Δ) cultures grown in rich medium containing 4% glucose, as described previously (Elderet al. 1983). Poly(A)+ RNA was prepared using the poly(A) spin mRNA isolation kit from New England Biolabs. Fluorescently labeled cDNA was prepared from the mRNA and hybridized to glass slides to which a DNA array of approximately 3325 yeast ORFs was attached (Derisiet al. 1997).
Mig1 and Mig2 function is regulated differently: The Snf1 protein kinase inhibits Mig1 under nonrepressing conditions (Vallier and Carlson 1994), probably by phosphorylating it, leading to its exit from the nucleus (Treitel and Carlson 1995; DeVitet al. 1997). Because of this, SUC2 expression is abolished in a snf1Δ mutant (Figure 2, line 5; also see Vallier and Carlson 1994). Mig1 is clearly responsible for this, because deletion of MIG1 (Figure 2, line 7), but not MIG2 (Figure 2, line 6) restores expression. Snf1 appears not to inhibit Mig2 function. This is clearly seen in a snf1 mig1 mutant (line 7), in which Mig2 is primarily responsible for the 10-fold glucose repression of SUC2 expression observed in this mutant (compare lines 2 and 4 or lines 7 and 8): Mig2-mediated repression and derepression of SUC2 expression are not affected by loss of SNF1 (compare lines 2 and 7). Thus, it appears that Snf1 does not inactivate Mig2. Nevertheless, Mig2 function is regulated in response to glucose (lines 2 and 7), suggesting that the regulator of Mig2 (X in Figure 2A) responds to glucose. Similar results were reported by Vallier et al. (1994).
We thought that the protein responsible for inactivating Mig2 under nonrepressing conditions could be one of the Snf1 homologs encoded in the yeast genome. Loss of the protein that inactivates Mig2 would cause Mig2 to repress SUC2 expression even under derepressing conditions (in the absence of glucose, as loss of SNF1 causes Mig1 to repress under these conditions). However, disruption of the genes encoding three of the closest homologs of Snf1 (HSL1, YCL024, and GIN4) in a mig1Δ snf1Δ strain had no effect on SUC2 expression (Table 3). Thus, none of these three Snf1 homologs appears to regulate Mig2.
Regulation of Mig1 and Mig2 function occurs by different mechanisms. The nuclear localization of Mig1 is regulated by glucose (DeVitet al. 1997), but Mig2 is located in the nucleus both in the presence and absence of glucose (Figure 3). This is consistent with the observation that Snf1, whose action causes Mig1 to move to the cytoplasm, does not regulate Mig2 activity.
MIG1 and MIG2 expression are regulated differently: Expression of MIG1 and MIG2 is regulated differently. MIG1 expression is repressed about 12-fold in the presence of glucose; MIG2 is expressed constitutively (Table 4). Mig1 and Mig2 are primarily responsible for repression of MIG1 expression: repression is slightly relieved in mig1Δ, mig2Δ, and yer028Δ single mutants, and almost completely relieved in a mig1Δ mig2Δ strain. Yer028 plays no significant role in repression of MIG1, because deleting YER028 in the mig1Δ mig2Δ mutant has little effect on expression. Thus, Mig1, together with Mig2, represses its own expression in the presence of glucose.
YER028 may encode a glucose-dependent transcriptional repressor: Because Yer028 possesses a DNA-binding domain very similar to those of Mig1 and Mig2, and is also similar to Mig2 outside of the zinc-finger region, it seemed likely that Yer028 has a function similar to Mig1 and Mig2. Indeed, a LexA-Yer028 chimeric protein is a glucose-dependent repressor of gene expression (Figure 4, lines 1-4). Like Mig1 and Mig2, Yer028 also requires Ssn6 and Tup1 to repress gene expression (Figure 4, lines 6 and 8). In the absence of TUP1 and SSN6, LexA-Yer028 activates transcription (Figure 5, lines 1 and 3). This is like LexA-Mig1 (Treitel and Carlson 1995), but unlike LexA-Mig2 (Lutfiyya and Johnston 1996).
Yer028 binds to the Mig1-binding sites in the SUC2 promoter: To determine if Yer028 binds to the same DNA sequence as Mig1 and Mig2, it was produced in E. coli (see materials and methods) and assayed for binding to the Mig1-binding sites in the SUC2 promoter (Figure 6). Yer028 binds well to a fragment of the SUC2 promoter containing both Mig1-binding sites and also to oligonucleotides of the individual Mig1-binding sites in the SUC2 promoter (SUC2-A and SUC2-B).
Function of individual Mig1 DNA-binding sites in vivo: Mig1 and Mig2 have different relative affinities for different binding sites in the GAL1 and SUC2 promoters, which may explain the different sensitivities of these genes to the two repressors (Lutfiyya and Johnston 1996). To determine if the affinities of Mig1 and Mig2 for various binding sites in vitro are correlated with the amount of transcriptional repression caused by those sites in vivo, their ability to repress expression through the Mig1-binding sites in the GAL1 and SUC2 promoters was tested (Figure 7). The SUC2-A and GAL1-A sites cause the most repression, consistent with the observation that Mig1 has the highest affinity for these sites and that Mig2 has the highest affinity for SUC2-A. The SUC2-B and GAL1-C sites, to which Mig1 and Mig2 bind less strongly, directed much less repression [though two copies of these sites caused 20- to 100-fold repression (data not shown)]. Mig1 plays the major role in repression caused by all four sites (line 2); Mig2 is responsible for most, or all, of the remaining repression (line 4). Thus, the relative affinities of Mig1 and Mig2 for their binding sites, measured in vitro, correlates well with their ability to repress transcription through these sites in vivo.
Two other observations are notable in the experiment presented in Figure 7. First, GAL1-A causes a small but probably significant amount of glucose repression in a mig1Δ mig2Δ yer028Δ strain (data not shown), suggesting that another protein binds to this site to mediate repression. This is similar to the situation for GAL1, whose expression is ∼3-fold repressed by glucose in the triple mutant strain (data not shown). Second, in the absence of MIG1 and MIG2, SUC2-A causes 5 to 10-fold activation on glycerol, suggesting that an activator binds to this site (data not shown). This is consistent with the observations of others (Sarokin and Carlson 1984; Bu and Schmidt 1998).
Defining binding site specificities for Mig1 and Mig2: In an attempt to define the binding sites preferred by Mig1 and Mig2, sequences bound by each protein were selected in vitro (materials and methods; see Horwitz and Loeb 1986; Oliphant and Struhl 1987; Wright and Funk 1993). An oligonucleotide with a degenerate sequence of the SUC2-A strong Mig1-binding site (Figure 7A) was synthesized, with each position of the binding site seeded with 79% of the wild-type nucleotide and 7% of each of the other three nucleotides. Oligonucleotides bound by Mig1 or Mig2 were selected in a gel mobility shift experiment (Figure 8A), eluted from the gel, cloned in a plasmid, and their sequence determined.
A total of 147 sites were identified for Mig2 and 106 for Mig1. The results, summarized in Figure 8B, show that the two proteins prefer almost identical binding sites. The most conserved sequence element is a GC-rich sequence, and the requirements in the GC-box appear to be almost identical for the two proteins. It is known that Mig1 requires for binding an AT-rich sequence directly upstream of the GC-box (Lundinet al. 1994), which is evident from this experiment, because only 12% of the sites bound by Mig1 contain two or more G’s or C’s within the AT-box (Figure 8C). In contrast, 22% of the sites bound by Mig2 contain two or more G’s and C’s within the AT-box: more G’s and C’s are found in positions 2, 3, 4, and 7 as compared to Mig1. Thus, Mig2 may not have as strict a requirement as Mig1 for the AT-rich sequence for binding. We conclude that Mig1 and Mig2 probably recognize the same sequences. Thus, the different effects of these proteins on the GAL1 and SUC2 genes are probably not due to differences in binding site specificity but rather to their affinities for different sites.
Identifying genes regulated by Mig1, Mig2, and Yer028: Despite the fact that Mig1 and Mig2 are glucose-dependent transcriptional repressors that appear to bind to similar DNA sequences, they do not always regulate the same genes. With the hope of shedding light on this paradox, and in an attempt to identify genes regulated by Yer028, we sought to identify genes regulated by each protein. In addition to revealing the genes that are regulated by these proteins, differences in the binding sites of regulated genes might provide clues to why these repressors act on different genes.
Two approaches were taken to identify genes that are regulated by Mig1, Mig2, and Yer028. The first approach involved searching the yeast genome for known Mig1-binding sites (Nehlin and Ronne 1990; Nehlinet al. 1991; Lundinet al. 1994; Huet al. 1995). A pattern search program was used to find Mig1-binding sites in the regions between ORFs in the yeast genome (see materials and methods). More than 100 genes with candidate Mig1-binding sites in their promoters were identified. Several of these, listed in Table 5 (lines 1-7), were tested for regulation by Mig1 and Mig2. The promoter regions (∼1 kb of sequence upstream of the ATG codon) of these genes were fused to lacZ and expression assayed in wild-type and various mutant strains. The results are illustrated in Figure 9, A-E and summarized in Table 5. Four of the seven genes tested, YKR075, REG2, YDR516, and HXT13 are regulated by Mig1 and Mig2. Glucose repression of YKR075 and YDR516 expression (encoding a protein homologous to REG1 and a protein similar to Glucokinase 1, respectively) is slightly relieved by deletion of MIG1 or MIG2, and almost completely relieved when both genes are deleted (A and C). REG2 (B) and HXT13 (D) are partially derepressed in a mig1Δ mig2Δ strain, but still exhibit significant (three- to fourfold) repression in the double mutant. Yer028 plays little role in glucose repression of any of these genes, and thus cannot account for the repression remaining in the double mutant. No expression was detected for HXT15 and HXT17, which encode proteins similar to hexose transporters. Glucose repression of TPS1 (E) was not clearly affected by any of the three mutations.
To identify more comprehensively genes regulated by Mig1, Mig2, and Yer028, expression of more than half of the predicted ORFs in the yeast genome was measured in wild-type (YM4797) and mig1Δ mig2Δ yer028Δ (YM4804) strains using DNA microarrays (Derisiet al. 1997; see also materials and methods). Of the ∼3325 genes analyzed in this way, 235 exhibit expression at least twofold higher in the triple mutant compared to wild type (Table 6; see materials and methods). Thirty-eight genes have the opposite expression pattern, being at least twofold more highly expressed in the wild-type strain than in the triple mutant. Many of these genes possess at least one (usually several) potential Mig1/Mig2-binding sites. The promoter regions of 17 of the genes in the first class (Table 5, lines 9-25), and 1 of the genes in the second class (Table 5, line 8) were fused to lacZ (see materials and methods) and their expression analyzed to determine which of the three repressors affects their expression (data for 7 of these are shown in Figure 10, A-G).
Ten of the 18 genes analyzed in more detail were removed from the analysis for various reasons. Expression of 4 genes (RIM9, AHT1, YEL050, and YHR054) could not be detected by the β-galactosidase assay. Six genes yielded results that either did not agree with results from the array [expression levels were no higher in the triple mutant than in wild type (HSP82, HSP60, and YLR264)], or that varied significantly from one experiment to the next (SSA4, HSP26, and HSP30). Five of these 6 genes encode heat-shock proteins that are induced by many conditions, so it is easy to imagine that they could give variable results from experiment to experiment.
Five genes (HXK1, DOG2, YEL070, YLR042, YFL054; Figure 10, A-E) are clearly regulated by both Mig1 and Mig2 and are repressed to varying degrees by glucose; this repression is partially relieved by a mig1Δ mutation and almost completely relieved by further deletion of MIG2. For HXK1, DOG2, YEL070, and YLR042 (Figure 10, A-D), the effect of deleting MIG2 is only apparent if MIG1 is also deleted (as for SUC2). For YFL054 (Figure 10E), deletion of MIG2 by itself has a modest effect on glucose repression that is much more apparent if MIG1 is also deleted. Deletion of YER028 had little or no effect on expression of any of these genes and thus appears not to be involved in their regulation. Thus, all of these genes, like SUC2, appear to be regulated by both Mig1 and Mig2, with Mig1 being the primary repressor. No genes solely sensitive to Mig2 or to Yer028 were identified.
The effect on YBR101 (Figure 10F) expression of deleting MIG1 and MIG2 was unusual: glucose repression of this gene is not apparent, and expression is higher in the mig1Δ mig2Δ mutant, both in the presence and absence of glucose [SSE2 exhibits a similar pattern of expression (data not shown)]. It is unclear why Mig1 and Mig2 action on this gene is not sensitive to the levels of glucose. A higher level of expression of the other genes in glycerol-grown mig1Δ mig2Δ cells is also apparent. This is likely due to residual repression activity of Mig1 and Mig2 under nonrepressing conditions. Expression of HXT1 was reduced almost threefold in the mig1Δ mig2Δ mutant, under both inducing and noninducing conditions (Figure 10G). Thus Mig1 and Mig2 appear to be modest activators of HXT1 expression. In fact, HXT1 expression is induced by high levels of glucose (Ozcan and Johnston 1995), conditions that cause Mig1 and Mig2 to act as transcriptional repressors. While it remains to be seen if Mig1 and Mig2 bind directly to HXT1, we note that it contains Mig1-binding sites in its promoter.
Mig1 and Mig2 are redundant transcriptional repressors: Mig2 was initially identified as a repressor that collaborates with Mig1 to cause glucose repression of SUC2 expression (Lutfiyya and Johnston 1996). Because it appeared to have no effect on glucose repression of GAL1 expression, we wondered if there are other genes repressed only by Mig1, or some genes repressed only by Mig2. Among several genes in the yeast genome whose promoters contain potential Mig1-binding sites, we found none that are regulated only by Mig1 (Figure 9; Table 5). A more comprehensive survey of the yeast genome revealed many genes apparently regulated by Mig1, Mig2, and/or Yer028 (Table 6), but all of the ones we analyzed in more detail are regulated by both Mig1 and Mig2 (Figure 10). Thus, as far as we can tell, Mig2 always works in conjunction with Mig1.
However, Mig1 appears to be more important than Mig2, since it is sufficient to cause nearly full repression of most of the genes it regulates. An effect of deleting MIG2 alone is observed only for a few genes. Perhaps Mig2 binds to DNA with lower affinity than Mig1, causing it to repress gene expression only in the absence of Mig1. Alternatively, Mig2 could be regulated posttranscriptionally, so that it is present only at low levels in the presence of Mig1. In any event, it is difficult to believe that Mig2 evolved to repress gene expression only when Mig1 is absent.
DNA-binding specificities of Mig1, Mig2, and Yer028: Why does Mig1 seem to act alone at some promoters (e.g., GAL1, GAL4, HXT4, and CAT8; Griggs and Johnston 1991; Flick and Johnston 1992; Hedgeset al. 1995; Ozcan and Johnston 1995), but with Mig2 at others? One possibility is that these proteins differ in the DNA-binding sites they recognize, despite possessing very similar DNA-binding domains. This is the case for some other related DNA-binding proteins. For example, a single amino acid in the recognition helix of the Bcd homeodomain determines its DNA-binding specificity, thereby distinguishing its binding site from that of similar homeodomain proteins (Treismanet al. 1989; Gehringet al. 1994). The Elk-1 and SAP-1a ETS-domain transcription factors also recognize slightly different sites due to subtle differences in their DNA-binding domains (Shoreet al. 1996). The results of the in vitro selection for Mig1- and Mig2-binding sites (Figure 8) suggest that the two proteins bind to the same sequence, though there may be subtle differences between them that were not detected in our experiment (we recognize, however, that our experiment would not uncover Mig2-binding sites that differ greatly from the consensus Mig1-binding site). For example, it is not clear if Mig2, like Mig1, requires for binding an AT-rich sequence preceding the GC-box. Similar difficulties in determining subtle differences in binding site specificities between very similar proteins using in vitro binding site selections have been noted (Shoreet al. 1996). Mig1 and Mig2 appear to differ in their relative affinities for their sites (Lutfiyya and Johnston 1996), like two other very similar zinc-finger proteins, NGFI-A and NGFI-C (Swirnoff and Milbrandt 1995), and these affinities correlate well with the ability of these sites to repress gene expression in vivo (Figure 7).
Amino acids in the Mig1 zinc fingers important for DNA recognition can be inferred on the basis of the structure of the Mig1 homolog Zif268 (NGFI-A), bound to its recognition site (Pavletich and Pabo 1991). These amino acids (see Figure 1) are identical in Mig1, Mig2, and Yer028. Amino acids other than those contacting the DNA may also have a significant effect on binding site recognition. For example, although Krox-20 and Mig1 share similar zinc-finger domains, Krox-20 has a stricter requirement for its site than does Mig1 (Nardelliet al. 1992; Lundinet al. 1994). Differences in DNA-binding affinities of NGFI-A and NGFI-C have been shown to be determined by the protein context of the DNA-binding domain, and not by the zinc fingers themselves (Swirnoff and Milbrandt 1995). In addition, some residues important for defining the binding site specificity of Elk-1, an ETS-domain transcription factor, do not contact DNA (Shoreet al. 1996). Instead, these residues probably affect the way other residues interact with DNA. A true understanding of the different DNA-binding abilities of Mig1 and Mig2 awaits knowledge of their structures.
Yer028 appears to be a glucose-dependent transcriptional repressor: Due to the sequence similarity between Yer028 and Mig1 and Mig2, it seemed likely that Yer028 would function similarly to these proteins. Indeed, LexA-Yer028 is a glucose-regulated transcriptional repressor (Figure 4), and Yer028 binds to the same sites as Mig1 and Mig2 (Figure 6). It is surprising, then, that it does not appear to be functionally redundant with either Mig1 or Mig2. Yer028 appears to have evolved a role in regulating expression of a set of genes separate from those regulated by Mig1 and Mig2.
Because LexA-Yer028 is a glucose-activated repressor that binds to Mig1/Mig2-binding sites, we are mystified by the fact that Yer028 does not cause glucose repression of a reporter gene containing those sites (Figure 7). This is unlikely to be due to lack of expression of YER028, because it appears as a strong signal on the expression arrays (data not shown). Since the repression ability of LexA-Yer028 is regulated by glucose, we can only surmise that the DNA-binding ability of native Yer028 is inhibited in glucose-grown cells.
It is surprising that Mig1 and Mig2 appear to be more similar in function than Mig2 and Yer028, because Mig2 is more similar to Yer028 than to Mig1, both within and outside of the zinc fingers. In addition, Mig2 and Yer028 are clearly more closely related, because they arose by duplication of a chromosomal region (Wolfe and Shields 1997). Mig2 and Yer028 are also similar in size, ∼380 amino acids, compared to the 504-amino-acid Mig1 protein. It is possible that Mig2 and Yer028 function redundantly in expression of some genes not detected in our DNA microarrays.
Differential regulation of Mig1 and Mig2: Although Mig1 and Mig2 are functionally redundant repressors, there are significant differences in their regulation: MIG1 expression is glucose repressed about 10-fold by Mig1 and Mig2, while MIG2 is constitutively expressed at a higher level than MIG1 (Table 4). This is surprising considering that Mig1 and Mig2 seem to have equivalent roles in repression of many genes (Figures 9 and 10), with Mig1 apparently playing a more important role in repression.
Mig1 and Mig2 proteins also appear to be regulated differently. The Snf1 protein kinase clearly regulates Mig1 function (Vallier and Carlson 1994; DeVitet al. 1997), but does not appear to affect Mig2 (Figure 2; see also Vallier and Carlson 1994). This is consistent with the fact that Mig1 and Mig2 are also regulated at different levels; Mig1 function is regulated by its nuclear localization (DeVitet al. 1997), while Mig2 is always present in the nucleus, regardless of carbon source (Figure 3). We thought that proteins similar to Snf1 might act on Mig2, but three of the closest homologs to Snf1 (Hsl1, Gin4, and Ycl024) appear not to affect Mig2 function (Table 3). The role in regulation of Mig2 function of three other kinases more distantly related to Snf1 (Kin1, Kin2, and Kin4) remains to be tested.
Genes regulated by Mig1 and Mig2: We identified 235 genes that exhibit at least twofold higher expression in the mig1 mig2 yer028 triple mutant strain compared to the wild-type strain. These fall into several functional categories (Table 6). Almost 30% of the genes play a role either in metabolic pathways, or in mitochondrial functions. This is reasonable, because these are two major classes of genes that are known to be glucose repressed (Johnston and Carlson 1992; Trumbly 1992; Ronne 1995). All four of the genes we analyzed in more detail that have known or probable roles in metabolic pathways (HXK1, DOG2, YEL070, and YFL054) were glucose repressed by Mig1 and Mig2. Another well-represented class of genes are heat-shock or stress-response genes. About half of the Mig1/Mig2-regulated genes are of unknown and unpredictable function.
Expression of ∼11% of yeast genes (∼710) is increased by a factor of at least two as glucose is progressively depleted from the medium (i.e., are glucose-repressed genes; Derisiet al. 1997). On the basis of analysis of about half of the genes in yeast, we estimate that 10-15% of these genes are regulated by Mig1 and Mig2. Thus, while it is clear that Mig1 and Mig2 are important for glucose repression, other repressors that play a major role in glucose repression of gene expression remain to be identified.
We thank Sean Eddy for help with analysis of the yeast genome for Mig1-binding sites, and M. Carlson, R. Brent, and A. Johnson for plasmids and strains. We thank Jim Dover for making Yer028 protein. This work was supported by National Institutes of Health grant GM32540 and funds provided by the James S. McDonnell Foundation.
Communicating editor: F. Winston
- Received June 8, 1998.
- Accepted September 2, 1998.
- Copyright © 1998 by the Genetics Society of America