|Home | About | Journals | Submit | Contact Us | Français|
Gcn5 protein is a prototypical histone acetyltransferase that controls transcription of multiple yeast genes. To identify molecular functions that act downstream of or in parallel with Gcn5 protein, we screened for suppressors that rescue the transcriptional defects of HIS3 caused by a catalytically inactive mutant Gcn5, the E173H mutant. One bypass of Gcn5 requirement gene (BGR) suppressor was mapped to the REG1 locus that encodes a semidominant mutant truncated after amino acid 740. Reg1(1-740) protein does not rescue the complete knockout of GCN5, nor does it suppress other gcn5− defects, including the inability to utilize nonglucose carbon sources. Reg1(1-740) enhances HIS3 transcription while HIS3 promoter remains hypoacetylated, indicating that a noncatalytic function of Gcn5 is targeted by this suppressor protein. Reg1 protein is a major regulator of Snf1 kinase that phosphorylates Ser10 of histone H3. However, whereas Snf1 protein is important for HIS3 expression, replacing Ser10 of H3 with alanine or glutamate neither attenuates nor augments the BGR phenotypes. Overproduction of Snf1 protein also preferentially rescues the E173H allele. Biochemically, both Snf1 and Reg1(1-740) proteins copurify with Gcn5 protein. Snf1 can phosphorylate recombinant Gcn5 in vitro. Together, these data suggest that Reg1 and Snf1 proteins function in an H3 phosphorylation-independent pathway that also involves a noncatalytic role played by Gcn5 protein.
Histone acetylation is a well-studied modification of chromatin (67) and has been linked to transcriptional regulation, recombination, DNA replication, and damage repair (13). GNAT (Gcn5 protein-related N-acetyltransferases)andMYST (MOZ-Ybf2/Sas3-Sas2-Tip60) families of histone acetyltransferases (HATs) generate both targeted and global acetylation of the chromatin (78). Other HATs, such as TAF1 (formerly TAFII250) and nuclear hormone receptor coactivators, though not belonging to either family, have also been shown to playcritical chromatin-related functions via their HAT activities (78).
The Saccharomyces cerevisiae Gcn5 protein is the catalytic subunit of several chromatographically distinct HAT complexes, including SAGA, ADA (32), SALSA, and SLIK (70, 71, 85). SAGA is recruited to the promoter by certain transcriptional activators and causes promoter-specific nucleosomal hyperacetylation leading to transcriptional activation (4, 5, 48, 51, 72). The SAGA complex also performs HAT-independent functions, such as TATA binding protein (TBP) recruitment and histone deubiquitinylation (8, 9, 19, 24, 38, 44, 55, 75, 86). SAGA and SALSA/SLIK complexes share TBP-associated factors with TFIID (33). Low-resolution electron microscopic studies showed that the architectures of SAGA and TFIID complexes are highly similar (3, 11, 91, 103). TFIID is critical for mostly housekeeping gene expression, and the SAGA-dominated genes (~10% of the nuclear genes) are largely stress-induced and are under the coordinated control of multiple chromatin and transcriptional regulators (43).
Although the promoter-specific histone acetylation function of Gcn5 has been firmly established (48, 51), which molecular activities are modulated by histone acetylation remains an open question. The best-known molecular event triggered directly by histone acetylation is the recruitment of bromodomain-containing proteins (20, 45, 53, 62). Besides this, however, little is known as to what other functions may be triggered or antagonized by histone acetylation. Identification of mutations that suppress defects associated with histone hypoacetylation may reveal factors downstream of histone acetylation. Thus far, the only reported screen for suppressors rescuing gcn5 null phenotypes was a multicopy suppressor hunt identifying ARG3 (69), which is likely involved in controlling the global chromatin structure via regulating the balance of nuclear polyamine. On the other hand, Gcn5 protein is important for only a portion of yeast genes (40, 43). Suppressors that display gene specificity, instead of global effects on chromatin structure, may shed light on the molecular basis for Gcn5-mediated transcriptional activation. In our first attempt to identify the bypass of Gcn5 requirement gene (BGR) suppressors, we isolated one such mutation mapped to the REG1 gene.
REG1 (also called HEX2 and SRN1) was identified in several genetic screens of glucose repression and RNA processing (63, 65, 66, 96). Reg1 protein associates physically and functionally with an essential and multifunctional protein phosphatase 1, Glc7 (23, 61, 94), whose substrate specificity is apparently determined by association with different partners, including Reg1 protein. Mutations of REG1 cause ectopic expression of several genes under repressing conditions (21, 27, 41, 64, 97, 102). Point mutations targeted at the Glc7 interaction domain of Reg1 protein derepress ADH2 and SUC2 (23). A similar transcriptional repression defect caused by a glc7 mutation (T152K) can be suppressed by overexpressing Reg1 protein (94). These transcriptional derepression phenotypes are likely due to the inability of Glc7 to dephosphorylate the appropriate target protein(s) and consequently the ectopic increase of protein phosphorylation. Indeed, deletion of the Snf1 protein kinase suppresses the derepression defects resulting from reg1 or glc7 mutations (23, 28, 42), indicating an antagonistic relationship between the Snf1 kinase and the Reg1-Glc7 phosphatase complex. Consistent with this notion, Reg1 protein interacts directly with Snf1 protein in both yeast two-hybrid assays and affinity purification (61, 79). Furthermore, a hyperactive Snf1 protein caused by reg1Δ rescues the Spt− phenotypes of spt21Δ cells (39). Curiously, the interaction between Reg1 protein and Snf1 protein, at least within the yeast two-hybrid context, is enhanced in glucose starvation conditions (61), raising the possibility that Reg1 protein may have a positive role in Snf1 protein action under certain conditions.
Snf1 protein acts as a cellular fuel gauge controlling responses to nutritional crises (37). The animal homologues of Snf1 protein are activated by AMP and are referred to as AMP-activated protein kinases. In plants, Snf1 protein-related kinases (SnRKs) fall into three large families, SnRK1, SnRK2, and SnRK3 (36). Snf1 protein, AMP-activated protein kinases, and SnRKs are the catalytic α subunits of a trimeric complex composed of a scaffold β protein and a regulatory γ subunit. In addition to bridging the α and γ subunits, the β protein contributes to substrate selection as well. The γ subunit of the yeast Snf1 complex is encoded by SNF4 (14). At least three yeast genes encode the β subunits (26, 104). Snf1 protein plays critical roles in controlling transcription of carbohydrate transporter and metabolism genes (80). Overexpression of Snf1 protein also causes early aging, increased rRNA recombination, and loss of rRNA locus silencing (56), a set of functions reportedly linked to histone H3 hyperphosphorylation. Indeed, several proteins can be phosphorylated by Snf1 protein in response to glucose starvation, including Reg1 protein (79), Mig1 (92), and histone H3 (60). The histone H3 phosphorylation activity of Snf1 protein has been linked directly to transcriptional activation and TBP recruitment (58, 59). Ser10 phosphorylation facilitates acetylation by increasing the affinity between Gcn5 protein and H3 (15, 18, 60). Both modifications are important for the expression of the INO1 gene in yeast (59, 60). In addition, genetic interactions between Snf1 protein and Srb/mediator proteins (49, 84) and TBP (83) were reported. Whether these general transcriptional factors can be phosphorylated by Snf1 protein is unclear.
In this work, evidence that a gain-of-function BGR allele for Reg1 protein likely adopts a novel function in facilitating transcription of HIS3 is presented. This function appears to require a functional Snf1 kinase. However, H3 phosphorylation does not play a critical role for the suppression, nor is it important for normal HIS3 activation. A unique allele specificity for a particular mutant Gcn5 protein is shared by the Reg1 suppressor and overproduction of Snf1 protein. Indeed, both Snf1 and Reg1 suppressor can be copurified with Gcn5 from yeast, linking these three proteins functionally and physically.
Yeast strains used in this work are listed in Table Table1.1. All genetic methods were according to reference 81. Yeast transformation was done with the lithium acetate method (29). Plasmids used in this work are listed in Table Table22.
To introduce gcn5 point mutations into the genome, the BamHI-HindIII fragment from wild-type or mutant GCN5 was inserted into the same sites ofYIplac211 (30) to generate pMK284. Constructs pMK284E173H and pMK284F221A were linearized with NgoMIV and transformed into yeast. Integration results in two copies of GCN5 separated by the YIplac211 sequence containing a URA3 marker. 5-Fluoroorotic acid (5-FOA) selection and genomic PCR were used to obtain and verify the desired E173H and F221A mutations.
The HIS3-lacZ reporter was introduced to yeast by transforming the StuI-linearized pMK334 that generates URA+ integrants. pMK334 was constructed by inserting the EcoRI-DraI lacZ fragment of pLKC482 (90) into the EcoRI-HindIII sites of YIplac211, resulting in pMK333. An EcoRI-BglII fragment containing the HIS3 promoter was isolated from pMK231 where a BglII site was introduced at the 5′ end of HIS3 open reading frame (ORF) and inserted into the EcoRI-BamHI sites of pMK333. A unique StuI site within the URA3 gene was used for integrative transformation. All subsequent integrants were grown in the absence of uracil to maintain the integrated sequence.
To knock out the SNF1 gene, two disruptors were constructed. snf1Δ-1::LEU2 was generated by two-step subcloning. First, an ApaLI-HindIII fragment upstream of the SNF1 ORF was inserted into the XbaI-HindIII sites of pJJ252 (47) to create pMK452. The 3′ flanking region of the SNF1 gene, an HpaI-SacI fragment obtained by PCR, was inserted into the BamHI-SacI sites of pMK452 to obtain pMK453. In the other disruptor (pYL45, snf1Δ-2::TRP1), the PstI-HindIII fragment of SNF1 was first inserted into pBluescript KS+ (pMK449). The AflII-BglII 200-bp fragment corresponding to amino acids 109 through 176 of SNF1 in pMK449 was replaced with the EcoRI-BglII fragment of pJJ248 containing the TRP1 gene (47). To create snf1 deletion strains, the HindIII-BamHI fragment of pMK453 or the EcoRI-BamHI fragment of pYL45 was obtained by restriction digestion before yeast transformation.
To introduce the REG1(1-740) allele, plasmid pYL31 was constructed by replacing the ClaI-BglII fragment of pKD97 (23) with a ClaI-XhoI-digested PCR product that contains the open reading frame of REG1 up to amino acid residue 740 followed immediately by a stop codon. The ClaI-KpnI fragment of pYL31 was cloned into HindIII-KpnI sites of YIplac211 to obtain pYL35. To replace the entire REG1 ORF with REG1(1-740), pYL35 was linearized by SnaBI and integrated into the REG1 locus by homologous recombination. The correct transformants were subjected to 5-FOA selection. Genomic PCR confirmed the correct genotype.
The reg1Δ strains were generated by introducing a PCR fragment containing the KanMX6 cassette flanked by REG1 sequences outside the ORF (10, 99). G418-resistant transformants were examined by genomic PCR to confirm the reg1Δ genotype.
To create and test histone H3 mutations, strain JHY205 (2) was first made HIS3+ by replacing the his3Δ1 allele with the BamHI fragment of pJJ217 (47) that contains the entire HIS3 gene, resulting in yDA10. Each histone H3 mutation was generated by the Quikchange method (Stratagene), using pJH33 as the template. All mutations were confirmed by sequencing.
The 2μm SNF1 construct pYL41 was created by cloning the BamHI-PstI fragment containing the entire transcription unit of SNF1 into EcoRI-PstI sites of YEplac112 (30). Deletion of the general control-responsive element (GCRE) was as described previously (51).
pMK547Gcn5 with an N-terminal hemagglutinin (HA) tag was created by cotransforming XbaI-linearized pMK547, derived from pAB8 with the Gal4 DNA binding domain deleted (34), and a PCR-amplified GCN5 open reading frame. The Gcn5-TAP fusion construct (pYL54) was generated by a strategy essentially equivalent to QuikChange mutagenesis protocol (Roche) except that the mutagenic primers were PCR-amplified TAP sequence (74) flanked with sequences around the stop codon of GCN5. pMK144 (52) was the template for mutagenesis and insertion of the TAP sequence. pYL67, a plasmid derived from pBS1479 (74) by replacing the TAP sequence with eight Myc repeats, was severed as PCR template to amplify the Myc::TRP1 cassette with flanking sequence correlated with residue 740 or the stop codon of REG1. Gel-purified PCR products were transformed into yeast cells to generate Reg1-Myc fusions.
Gcn5 protein amino acids 19 to 348 lacking the bromodomain were cloned into pET21a and expressed as a His-tagged protein (pMK515). The desired point mutations were generated by the Quikchange method (Stratagene) and verified by sequencing. The recombinant protein was induced in the BL21 strain by adding 1 mM (final concentration) IPTG (isopropyl-β-d-thiogalactopyranoside) when cell culture reached an optical density at 600 nm (OD600) of 0.5/ml. Cell cultures were grown at 37°C for 3 h. Extraction and protein affinity purification were done according to reference 50.
Kinase assays were done with the above Gcn5 protein incubated with glutathione S-transferase (GST)-Snf1 (wild-type or K84R) expressed and purified from yeast according to reference 35. The GST-SNF1 constructs were kindly provided by D. Thiele (Duke University).
The yeast genomic DNA library (#21) containing the mTn-lacZ/LEU2 intervening sequence was provided by M. Snyder (Yale University) (77). The DNA was prepared by cesium chloride gradient and digested by NotI before transforming into yMK995. Ten micrograms of the library DNA was digested and isolated by phenol-chloroform extraction and ethanol precipitation. Approximately 26,000 LEU+ transformants were replica plated to synthetic complete (SC)-His medium containing 20 mM 3-amino-1,2,4-triazole (3-AT) and incubated at 37°C for 3 to 5 days. 3-AT-resistant colonies were further transferred to nitrocellulose membranes, and the lacZ level was tested according to reference 1. Colonies that showed blue color on the lacZ filter assays were grown in SC-Leu medium overnight and transferred to yeast extract-peptone-dextrose (YPD) (representing the repressed condition) or synthetic minimal medium (SD) containing 20mM 3-AT for 4 h. Yeast cells (20ml; OD600 of 0.1/ml) were then harvested by centrifugation (10,000 × g for 5 min at 4°C), washed, and suspended in extraction buffer (0.3 M sorbitol, 0.1 M NaCl, 5 mM MgCl2, 10 mM Tris HCl, pH 7.4, 5 mM EDTA, Complete protease inhibitor cocktail [Roche]). Whole-cell extracts were prepared by vigorous agitation with glass beads using a bead beater (Biospec Products). β-Galactosidase activity was quantified according to reference 1. One clone, renamed yMK1055 henceforth, repetitively showed elevated lacZ expression in response to amino acid starvation and was further studied. yMK1055 was backcrossed to yMK1075 before 3-AT tests. To verify that a single mTn insertion event was responsible for the BGR phenotypes, yMK1055 was crossed to yMK1085. The diploid strain was subjected to sporulation and tetrad dissection; all trp− segregants were tested for cosegregation of 3-AT resistance and leucine prototrophy. Recommended procedures were employed to map the integration site of the mTn-lacZ fragment (http://ygac.med.yale.edu/mtn/insertion_libraries.stm). Namely, yeast genomic DNA was isolated, digested by EcoRI, and subjected to intramolecular ligation prior to bacterial transformation. Plasmid DNA was isolated from Escherichia coli cells and sequenced across the junction between REG1 and mTn-lacZ/LEU2 using a primer specific to lacZ.
Yeast cells were grown in appropriate selection media until the OD600 reached 0.5. Cells were then pelleted by centrifugation (5,000 ×g, 5 min, 4°C) and transferred to either YPD (for basal expression) or SD supplemented with required nutrients and 40 mM 3-AT (for induced expression). Cell suspensions were further incubated at 37°C for 2 to 3 h before harvesting for RNA preparation. Although these relatively harsh conditions for induction were not essential, such treatment generally generated more consistent results in HIS3 activation in our strain background. Procedures for RNA preparation and Northern blot hybridization were described previously (52).
To test the interaction between Gcn5 and Snf1 proteins, a GST-Snf1 expression construct (35) or just GST (pYL44) was transformed to the strain carrying pMK547Gcn5. Purification of GST-Snf1 wasas described previously (35). Glutathione Sepharose 4B (30 μl; Amersham) was added to whole-cell extracts purified from 1.5 × 109 cells and incubated at 4°C for 3 h under constant rocking. Beads were pelleted and washed twice with HEMGT buffer (25 mM HEPES, pH 7.9, 12.5 mM MgCl2, 150 mM NaCl, 10% glycerol, 0.1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1× Complete protease inhibitor cocktail [Roche]) followed by two more washes with HEMGT buffer containing 300 or 500 mM NaCl. The bound fractions were eluted by sodium dodecyl sulfate (SDS) loading buffer and resolved by 8% SDS-polyacrylamide gel electrophoresis (SDS-PAGE). The copurified HA-Gcn5 protein was detected by anti-HA antibodies (12CA5; Roche).
For Gcn5-Reg1(1-740) copurification, whole-cell extracts from cells carrying pYL54 and C-terminally Myc-tagged Reg1 or Reg1(1-740) protein were prepared with the bead-beating method in FA lysis buffer (50 mM HEPES, pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% Na deoxycholate, 1 mM phenylmethylsulfonyl fluoride, 1× Complete protease inhibitor cocktail [Roche]). Lysates from 3 × 109 cells were incubated with 30 μl of immunoglobulin G Sepharose 6 (Amersham) for 2 h at 4°C. After three washes with FA lysis buffer, the beads were boiled in SDS loading buffer and resolved by 8% SDS-PAGE. Western blots were conducted with an anti-c-Myc antibody (Roche).
To screen for extragenic suppressors that rescue transcriptional defects caused by loss-of-function mutations of Gcn5 protein, we introduced a single mutation to the catalytic domain of Gcn5 protein. This mutation, E173H, is a Glu-to-His mutation at residue 173. Since Gcn5 protein participates in more than one complex, the use of this mutant likely maintains the integral architecture of these complexes (103). Previously, an E173Q mutation was shown by others to drastically reduce the in vitro and in vivo activity of Gcn5 protein (54, 73, 93). However, this E173Q mutant in our hands maintained significant activities in HIS3 expression after it was integrated back to the native GCN5 locus (data not shown). We thus designed a Glu-to-His mutation. With the slight positive charge of histidine under physiological pH, a more drastic reduction of the catalytic power of Gcn5 protein was expected (54, 89). The HAT activity of a bacterially expressed E173H mutant was tested using chicken histones as the substrates. As predicted, this mutation significantly reduced the in vitro HAT activity of Gcn5 protein (Fig. (Fig.1A).1A). To test in vivo functions, the E173H allele was integrated to the native GCN5 locus to replace the wild-type allele. In parallel, another well-characterized F221A allele (52) was integrated in the same manner. Both alleles were controlled by the native GCN5 cis elements. Yeast strains bearing the wild-type, complete knockout, F221A, or E173H allele of GCN5 were then tested for responses to amino acid starvation. Each strain was patched to YPD medium and then replica plated to synthetic complete medium lacking histidine and supplemented with various concentrations of 3-AT, a competitive inhibitor of the His3 protein. Very minor growth defects were seen in gcn5− strains when assayed at 30°C in medium supplemented with 3-AT (Fig. (Fig.1B).1B). However, when these cells were incubated at 37°C, 3-AT induced obvious growth defects of all three gcn5− strains. None of these cells were temperature sensitive (compare growth on YPD and SC-His without 3-AT). The clear growth defects of gcn5− cells provide a platform for suppressor screening.
We further modified the E173H mutant strain by introducing a HIS3-lacZ reporter to the ura3-52 locus. Insertion of HIS3-lacZ did not change the cellular sensitivity to 3-AT (Fig. (Fig.1B,1B, bottom two patches). This lacZ reporter, under the control of the HIS3 promoter, was also activated by amino acid starvation (Fig. (Fig.1C)1C) and hence offered a convenient means to verify the 3-AT-resistant suppressor phenotypes.
To identify suppressors, we used a minitransposon (mTn)-based mutagenesis approach (77). In this method, the mTn-lacZ/LEU2 sequence was integrated into a yeast genomic DNA library via transposition. Yeast DNA fragments along with the interrupting sequence were excised from the plasmid pool and transformed into yeast. Each mTn sequence integrated to the chromatin via homologous recombination between the flanking yeast sequence and the corresponding genomic locus. LEU+ transformants were replica plated to 20 mM 3-AT medium and grown at 37°C. All 3-AT-resistant clones were then screened for increased expression of β-galactosidase induced by amino acid starvation. From approximately 26,000 LEU+ transformants, we identified 1 such colony (Fig. (Fig.1C).1C). Northern data clearly showed that the HIS3 expression was upregulated in this suppressor strain compared with the parental gcn5 E173H cells (Fig. (Fig.1D).1D). Similar complementation in transcription was seen in HIS1, HIS6 (not shown), and HIS4 (Fig. (Fig.2C)2C) as well. Genetic assays showed that a single mTn insertion event was responsible for the suppression phenotypes (data not shown and see Materials and Methods).
To map the mutation, we rescued and cloned the mTn insertion along with the flanking yeast sequences (see Materials and Methods). DNA sequencing across the junction revealed that the mutagenic fragment had inserted to the coding region of REG1 (Fig. (Fig.2A),2A), resulting in in-frame fusion of lacZ to residue 740 of Reg1 protein. While the expression of the Reg1-mTn-lacZ fusion protein may have contributed to some β-galactosidase activity shown in Fig. Fig.1C,1C, the HIS3 transcript quantification results (Fig. (Fig.1D)1D) unequivocally demonstrated the rescue of gcn5− defects. Nonetheless, since the in-frame fusion of lacZ added a large mass to the truncated Reg1 protein, we were curious whether the β-galactosidase fusion was necessary for the suppression. By integrative transformation, we replaced the chromosomal copy of the native REG1 with one that is truncated after residue 740 (without the lacZ fusion) and tested whether this “clean” REG1(1-740) allele was able to suppress the gcn5 E173H mutation. Figure Figure2B2B shows that with a truncated Reg1(1-740) protein, the gcn5 E173H cells also exhibited significant resistance to 3-AT (Fig. (Fig.2B,2B, row 5) and restoration of HIS3 expression (Fig. (Fig.2B,2B, lane 7), although the original mTn-lacZ insertion consistently showed better growth than Reg1(1-740). The in-frame fusion of β-galactosidase enhanced but was not essential for suppression efficacy.
We further deleted the REG1 gene and found that the suppression phenotypes were lost (Fig. (Fig.2B,2B, left panel, row 4, and right panel, lane 4). Interestingly, in the presence of a functional Gcn5 protein, deleting the REG1 gene does not seem to affect HIS3 expression (Fig. (Fig.1D,1D, lanes 7 and 8). Together, these results showed that a Reg1(1-740) truncated protein is essential and sufficient for suppressing the gcn5 E173H mutation in HIS3 transcription.
Gcn5, in the context of SAGA complex, is recruited to the HIS3 promoter by the transcriptional activator Gcn4 that binds the cognate cis element, GCRE (51). To test whether the Reg1(1-740) suppressor exerted its function via Gcn4-GCRE or a novel Gcn4-independent mechanism, we replaced the GCRE 5′ to the HIS3 gene with an irrelevant sequence (51) and determined whether the suppression was affected. Comparison of mRNA transcribed from the GCRE-less HIS3 with the wild-type HIS4 control clearly showed that the GCRE was essential for Reg1(1-740)-mediated suppression (Fig. (Fig.2C),2C), indicating that Reg1(1-740) protein modulates an activity downstream of the normal Gcn4 functions.
One possible mechanism for the observed suppression is restoration of the HAT activity of the Gcn5 E173H mutant protein. To see if this was the case, we conducted chromatin immunoprecipitation using an antibody against histone H3 acetylated at Lys9 and/or 14 (52). Figure Figure2D,2D, lane 1, shows the expected hyperacetylation at the +1 nucleosome of HIS3 in the presence of a wild-type Gcn5 protein (51). The TATA element and the transcription initiation sites are within the +1 nucleosome. Promoter hyperacetylation was lost in the E173H background (Fig. (Fig.2D,2D, lane 5). When REG1 was replaced with the REG1(1-740) suppressor allele, H3 remained hypoacetylated at the HIS3 promoter (Fig. (Fig.2D,2D, lane 7), suggesting that the canonical nucleosomal H3 acetylation by the Gcn5 acetyltransferase was not affected in the REG1(1-740) background.
To characterize the Reg1(1-740) suppressor further, we first tested whether this allele was dominant or recessive. Figure Figure3A3A shows that the REG1+/REG1(1-740) heterozygote retained a growth advantage over the parental gcn5 E173H REG1+ homozygous strain at both 30 and 37°C. The strength of the suppression appeared to be somewhat weaker than the haploid strain, suggesting that the Reg1(1-740) protein was a gain-of-function, semidominant suppressor.
Besides E173H, the F221A mutation also causes quantitative defects of Gcn5 functions in vitro and in vivo (Fig. (Fig.1B)1B) (52). This mutation selectively impairs the ability of Gcn5 protein to bind acetyl coenzyme A (acetyl-CoA) (57, 76, 88), which is a prerequisite step for histone tail binding (88). It is thus likely that histone tails within the vicinity of the SAGA complex remain free from binding by the F221A mutant protein. In contrast, based on the studies of the E173Q mutant (89) and our yeast two-hybrid tests comparing different disabled gcn5 mutants (M.-H. Kuo, unpublished data), the E173H allele most likely prolongs its association with both substrates because the catalytic process stalls after the ternary complex is formed. Because of the possible differential effects on histone tail accessibility, we tested whether E173H, F221A, and a knockout allele of GCN5 responded differently to Reg1(1-740) suppressor.
To test the allele specificity, the gcn5Δ and gcn5 F221A strains were engineered so that REG1 was replaced with the REG1(1-740) allele, and the resultant strains were tested on 3-AT plates (Fig. (Fig.3B).3B). Neither the F221A nor the complete knockout allele of gcn5 was rescued by Reg1(1-740) (Fig. (Fig.3B,3B, rows 6 and 8). The strong preference for the E173H allele suggests that Reg1 protein may interact directly with Gcn5 protein or may control Gcn5 at a step(s) subsequent to the formation of the Gcn5-acetyl-CoA-histone ternary complex.
We further tested whether Reg1(1-740) protein rescued other gcn5− phenotypes. Figure Figure3C3C shows that all three gcn5− mutants exhibited severe growth retardation in yeast extract-peptone-glycerol and yeast extract-peptone-ethanol media, as previously reported (58). Moreover, gcn5− cells were also sensitive to 100 mM hydroxyurea, an inhibitor of DNA replication. However, neither the mTn allele nor the clean truncation of Reg1(1-740) protein was able to suppress any of these defects. The failure to suppress other phenotypes, such as caffeine sensitivity, and the inability to use galactose or sucrose were also observed (data not shown). In contrast, the sporulation defects (12) of a gcn5 E173H homozygous strain were partially rescued (not shown). We conclude that the Reg1(1-740) protein suppresses only a subset of Gcn5 target genes.
Reg1 protein is best known to inhibit the kinase activity of Snf1 protein and consequently prevents the expression of many genes when glucose is abundant (see the introduction), a function termed glucose repression. Snf1 protein derepresses the expression of these genes via several mechanisms, including histone H3 phosphorylation (59, 60). Phosphorylated H3 was shown to bind Gcn5 protein at a higher affinity (15, 17). It thus seems plausible that the hypoacetylation phenotype caused by gcn5 mutations may be compensated for by hyperphosphorylation of H3, which either helps anchor Gcn5 protein to the HIS3 promoter or by itself provides an environment suitable for stronger HIS3 expression.
To examine the link between Gcn5, Snf1, and H3 phosphorylation, we first tested whether Snf1 protein was involved in HIS3 expression. To this end, we created two deletion alleles of SNF1. The snf1Δ-1::LEU2 mutant had the entire ORF replaced with a LEU2 marker. However, this marker was incompatible with the mTn-lacZ-LEU2 insertion mutant; we thus created another allele, snf1Δ-2::TRP1, that was truncated after amino acid 108. Figure Figure44 shows that both snf1 mutations caused obvious growth defects on 3-AT plates (Fig. (Fig.4A,4A, rows 2 to 4) as well as impaired HIS3 expression (Fig. (Fig.4B,4B, lane 3, and data not shown), demonstrating that the Snf1 protein was also a critical transcriptional regulator for HIS3. We next tested whether the Snf1 protein was important for the suppression. Deleting SNF1 from the original gcn5 E173H REG1::mTn-lacZ suppressor strain significantly attenuated the suppression phenotypes (Fig. (Fig.4A,4A, rows 6 and 7, and B, lane 5). Thus, Snf1 protein is critical for normal and Reg1(1-740)-mediated HIS3 activation.
To see how Gcn5 and Snf1 proteins may genetically interact to activate HIS3, we examined whether overexpressing one of these two enzymes can rescue the HIS3 expression defects caused by a mutation of the other. While a 2μm multicopy GCN5 construct was unable to rescue the 3-AT hypersensitivity of the snf1Δ-2::TRP1 strain (Fig. (Fig.4C,4C, left panel), overproduction of Snf1 protein effectively rescued the E173H allele of gcn5 (Fig. (Fig.4C,4C, right panel, compare rows 3 and 4). On the other hand, overproduction of a catalytically inactive mutant of Snf1, K83R, failed to rescue the gcn5− phenotypes (data not shown), suggesting that the kinase activity was essential for the suppression. Intriguingly, neither deletion nor the F221A allele of gcn5 responded to the multicopy SNF1 plasmid. Thus, the Snf1 multicopy suppressor displays an allele specificity similar to that of Reg1(1-740). Furthermore, in the presence of a functional GCN5, overproduction of Snf1 protein yielded higher resistance to 3-AT (Fig. (Fig.4C,4C, right panel, row 2), very similar to the GCN5+ REG1(1-740) strain (Fig. (Fig.3B,3B, row 2).
Taking together the above results, as well as the reports that Reg1 and Snf1 interact genetically and physically for transcription of several inducible genes (see the introduction), it seems likely that Snf1 may be part of the mechanism by which Reg1(1-740) protein suppresses the E173H mutant allele.
To test whether H3 Ser10 phosphorylation contributes to the BGR phenotypes, we used a yeast strain in which both copies of each of the four core histone genes had been deleted (2). Viability of the cells was supported by a low-copy-number plasmid bearing wild-type histone genes and a URA3 marker. The desired histone mutations can be introduced into an otherwise identical construct containing a LEU2 nutrient marker. After transforming the latter plasmid that delivered the specific histone mutation(s), the wild-type histone genes were shuffled out by 5-FOA selection, leaving the mutant allele as the sole copy for histone expression. Additionally, GCN5 and REG1 were replaced with the E173H and REG1(1-740) alleles, respectively. 3-AT resistance was then compared among different LEU+ Ura− strains as shown in Fig. Fig.5.5. In this genetic background, Reg1(1-740) also effectively rescued the E173H mutant. However, the S10A mutation did not impose a discernible effect on cellular growth (Fig. (Fig.5,5, compare rows 3 and 4), ruling out a critical role played by phosphorylated Ser10 alone. Within the amino-terminal tail domain of histone H3, Ser28 and Ser31 share sequence similarity with Ser10 (7ARKSTGG and 25ARKSAPSTGG). Although Snf1 protein has not been shown to phosphorylate either serine residue, Ser28 can be phosphorylated by the Aurora family kinases for chromatin condensation during mitosis (16, 31, 82). We were curious about the possibility that the Snf1 kinase activity might “spill over” to these two residues in the REG1(1-740) strain. Thus, a triple Ser-to-Ala mutant, S10A/S28A/S31A, was introduced to the gcn5 E173H REG1(1-740) background. These cells still exhibited robust growth in the presence of 3-AT (Fig. (Fig.5,5, row 6), further supporting the notion that H3 phosphorylation was unlikely to be the driving force for the observed BGR phenotypes. Consistent with this, neither single nor triple Ser-to-Ala mutations exacerbated the 3-AT hypersensitivity caused by the E173H mutation in a REG1+ background (Fig. (Fig.5,5, compare rows 2, 8, and 10). We therefore conclude that Ser10 phosphorylation, though important for activation of several other genes, does not contribute appreciably to Gcn5 and Snf1 protein-mediated HIS3 expression. Thus, Snf1 protein most likely controls HIS3 expression by a novel, H3 phosphorylation-independent mechanism(s).
While preventing H3 phosphorylation imposes no apparent effect on the Reg1(1-740) protein-generated suppression, we were nonetheless interested in knowing whether a constitutively phosphorylated H3 would be sufficient to bring about a chromatin environment that suppresses the gcn5 E173H transcriptional defects. Toward this end, Ser10, Ser28, and Ser31 were replaced by aspartate or glutamate that mimicked the negatively charged phosphorylation state. Cellular growth in the presence of 3-AT was then assessed. While a single S10E mutation yielded very few differences in REG1 or REG1(1-740) background (Fig. (Fig.5,5, rows 5 and 9), the triple acidic mutation clearly brought about stronger resistance to 3-AT (Fig.5, rows 7 and 11). Since this phenotype was independent of the REG1 status, we conclude that constitutive negative charges at the amino terminus of H3 represent another bypass of Gcn5 requirement suppressor.
The above data place both Reg1 and Snf1 proteins to the regulatory circuitry of HIS3 and likely other amino acid starvation-inducible genes. The ability of Reg1(1-740) protein and overproduced Snf1 kinase to rescue preferentially the E173H mutant suggests an intriguing possibility that Gcn5 protein is a functional target for the Snf1 kinase. To test this hypothesis, we purified a wild-type and a catalytically inactive (K84R) GST-Snf1 protein from yeast (35) and incubated these two preps with recombinant Gcn5 protein expressed in E. coli. [γ-32P]ATP was included in the reactions to track the phosphorylation status of Gcn5. Figure Figure6A6A shows that Gcn5 protein was indeed phosphorylated in the presence of the wild-type Snf1 protein. The K84R mutation effectively diminished Gcn5 phosphorylation, indicating that Snf1 protein was responsible for Gcn5 protein phosphorylation.
Intrigued by the in vitro phosphorylation results, we further tested whether Gcn5 and Snf1 proteins interacted in vivo. To this end, we epitope tagged Gcn5 with HA at its amino terminus. Two yeast strains expressing GST-Snf1 or GST were transformed with the HA-GCN5 construct and subjected to one-step purification with a glutathione matrix. After extensive washing, the bound materials were resolved by SDS-PAGE and probed with an anti-HA antibody. Figure Figure6B6B shows apparent copurification of the HA-Gcn5 protein with GST-Snf1 but not GST alone. Literally identical results were obtained in reciprocal experiments (i.e., immunoprecipitation with the anti-HA antibody, followed by Western analyses to quantify Snf1 protein in the precipitate) (not shown), confirming the in vivo association between Gcn5 and Snf1 proteins.
We then asked whether Reg1 protein also associated with Gcn5 protein. Figure Figure6C6C shows that a Myc-tagged Reg1(1-740) protein was also present in the crude preparation of an epitope-tagged Gcn5 protein. Intriguingly, the full-length Reg1-Myc protein was not detected under the same condition (Fig. (Fig.6C,6C, first two lanes), consistent with the gain-of-function trait of the Reg1(1-740) suppressor protein.
The histone acetyltransferase activity of Gcn5 protein is critical for the expression of multiple yeast genes. Point mutations that eliminate the HAT activity of Gcn5 protein cause defects in promoter acetylation and in transcriptional activation of such model genes as HIS3 and PHO5 (7, 52, 73, 100). While these results provide solid evidence that Gcn5 protein uses its HAT activity to activate transcription, microarray studies also showed that a gcn5 knockout strain has transcriptional defects in more genes than does a strain expressing a catalytically inactive mutant (43), suggesting that Gcn5 protein may perform noncatalytic roles in gene expression. Indeed, Jacobson and Pillus showed that a catalytically inactive Gcn5 protein counteracts transcriptional silencing at subtelomeric loci (46). Such noncatalytic functions of Gcn5 protein may be unveiled by characterizing point mutations that abrogate the catalytic power of Gcn5 protein but permit other functions to be exerted. This notion seems to be consistent with the data presented in this work. For example, HIS3 and HIS4 expression are effectively rescued by the Reg1(1-740) suppressor (Fig. (Fig.1D1D and and2C)2C) in the E173H but not the knockout background. No restoration of histone H3 acetylation was detected, suggesting one possibility that the noncatalytic function of the E173H allele is selectively enhanced by Reg1(1-740) protein. This function is likely synergistic with its catalytic counterpart, as more pronounced resistance to 3-AT is exhibited by GCN5+ REG1(1-740) and GCN5+ multicopy SNF1 strains (Fig. (Fig.3B3B and and4C4C).
It is also intriguing that the F221A allele is refractory to Reg1(1-740) and higher doses of Snf1 protein. Several other suppressors that are currently characterized by us do not show such unique allele specificity (Y. Liu, X. Xu, and M.-H. Kuo, unpublished data). Molecularly, E173H and F221A mutations abrogate the HAT activity of Gcn5 via different mechanisms and may have different impacts on histone tails. F221A impairs acetyl-CoA binding (57, 88, 93), whereas E173H blocks the nucleophilic attack on the bound acetyl-CoA (89). Association of acetyl-CoA is prerequisite to histone tail binding (88, 89). After the transfer of the acetyl group to histone within the ternary complex, the acetylated histone dissociates first and then follows the consumed coenzyme A. Thus, blocking the association between Gcn5 and acetyl-CoA by the F221A mutation likely prevents Gcn5 protein from binding to the substrate histone, rendering the latter susceptible to other unregulated or untimely chromatin binding and modulating activities. The E173H mutation, on the other hand, may lock Gcn5, acetyl-CoA, and the histone tail in a ternary complex, thus preventing possible usage or modifications of the histone tail by other activities. In addition, it remains a strong possibility that Gcn5 protein uses nonhistone protein substrates (68). If so, the retention of one of these proteins by the E173H mutant enzyme may exacerbate the histone hypoacetylation defects.
Furthermore, only a subset of defects associated with gcn5− mutants can be rescued by Reg1(1-740) (Fig. (Fig.3C).3C). Together, it is highly likely that Gcn5 uses multiple mechanisms to activate transcription in a target gene (or transcriptional activator)-dependent manner.
Reg1 protein is a regulatory subunit for Glc7, an essential and multifunctional type I protein phosphatase (95). Reg1 protein also interacts with several other proteins, including Snf1 (61, 79) and the yeast 14-3-3 homologues, Bmh1 and Bmh2 proteins (22). The binding domains for these proteins are all within the first 500 amino acids that are conserved among Reg1 protein homologues (22, 23, 61). These domain are preserved in our REG1(1-740) suppressor allele, suggesting that the prototypical functions of Reg1 protein are not impaired by the C-terminal truncation.
The Reg1(1-740) protein lacks about one third of the total length. The truncation occurs immediately before a stretch of acidic residues (15 of 19 residues are Asp or Glu), and the deleted portion is rich in serine, threonine, and acidic residues (16% Ser, 4.4% Thr, 8.8% Asp, and 7.3% Glu). Little is known about the molecular functions or potential partners of this part of the Reg1 protein. Preliminary sequence search reveals no clear homologues to this region across species (data not shown). Contrary to the gain-of-function BGR phenotypes, this C-terminal region is dispensable for glucose repression. For example, Dombek et al. showed that the C-terminal deletion of Reg1 protein (up to residue 693) does not cause appreciable derepression of ADH2 or SUC2 (23). Shirra and Arndt reported that a Reg1 protein missing the last 80 amino acids is able to fully complement a recessive reg1-326 mutant (83). Indeed, we have no evidence of transcriptional derepression of those glucose-repressible genes in the REG1(1-740) background (Y. Liu and M.-H. Kuo, unpublished). It is possible that the carboxyl-terminal third of Reg1 protein interacts with a negative regulator(s), or another region of Reg1 protein in cis, that restricts specifically the HIS3 expression-related functions of Reg1 protein. Perhaps this negative regulator selectively controls the residual non-HAT function of the E173H mutant of Gcn5 protein. Upon deleting this Ser-Thr-Asp-Glu-rich domain, the negative effect of this regulator diminishes, hence unleashing the non-HAT function of Gcn5 protein for HIS3 activation. This view is consistent with the affinity purification data (Fig. (Fig.6)6) that the Reg1(1-740) but not the full-length Reg1 protein can be copurified with an epitope-tagged Gcn5 protein.
It is important that the suppressing power of Reg1(1-740) protein is abrogated by deleting SNF1. While this result alone does not prove that Snf1 protein acts downstream of the Reg1(1-740) suppressor, considering the well-established interaction between Reg1 and Snf1 proteins, we suggest that at least part of the suppressor function of Reg1(1-740) protein is mediated through Snf1 protein. However, we cannot rule out the existence of an intermediary step(s)/factor(s) for the suppression.
One probable factor involved in the BGR phenotype is the type 1 protein phosphatase Glc7. Reg1 is one of several regulators of the essential Glc7 enzyme. Unfortunately, our attempts to link Glc7 protein to the BGR phenotypes failed to generate conclusive data. Using several known glc7 point mutations that cause phenotypes in glycogen metabolism and/or glucose repression, we indeed found a few able to confer strong resistance to 3-AT in the absence of a functional Gcn5 protein. However, such elevated 3-AT resistance was not accompanied by increased HIS3 transcription (Y. Liu and M.-H. Kuo, unpublished). This disparity probably arises from the fact that Glc7 protein controls multiple cytoplasmic and nuclear functions (e.g., see references 87 and 101). Changes in the metabolism and flux of 3-AT may render yeast cells resistant to 3-AT with a low level of HIS3 transcription. The possible involvement of GLC7 in HIS3 regulation awaits further investigation when more mutant glc7− alleles are available.
Reg1 protein was recently shown to be purified in a complex containing two yeast 14-3-3 homologues, Bmh1 and Bmh2 proteins, and heat shock proteins Ssd1 and Ssd2 (22). Deleting BMH1 or BMH2 did not appreciably alter the ability of Reg1(1-740) protein to rescue the gcn5 E173H mutant (X. Xu and M.-H. Kuo, data not shown), indicating that these two proteins are not part of the suppression mechanism. Alternatively, functional redundancy between Bmh1 and Bmh2 proteins (92% identical) (98) may account for the lack of phenotypes in bmh1Δ and bmh2Δ strains.
Interestingly, the gain-of-function nature of the Reg1(1-740) suppressor, as well as the phenotypic similarity between Reg1(1-740) and overexpressed Snf1 protein, are at odds with the well-characterized antagonistic relationship with Snf1 protein (see the Introduction). We suggest that the functional relationship between these two proteins may be gene dependent. One precedent for this type of functional variation was reported for Spt3/8 proteins on TBP recruitment. Spt3 protein genetically and physically interacts with TBP (25). While Spt3 protein is required for TBP binding to the TATA elements of GAL1 and ADH2 (8, 9, 24, 55), it also plays a negative role in TBP-TATA interaction in other cases (7, 105).
Snf1 protein is a member of the AMP-activated protein kinase family that serves as a metabolic sensor in eukaryotic cells (37). It thus seems reasonable that Snf1 protein also contributes to the regulation of amino acid biosynthesis genes as shown in this work. Despite the functional interaction between Gcn5 and Snf1 proteins for INO1 activation (15, 17, 58-60), the H3 phosphorylation function of Snf1 protein is unlikely to be a major determinant in HIS3 expression (Fig. (Fig.5).5). However, we cannot rule out the possibility of phosphorylation at other residues or histones by Snf1 protein. In addition, genetic data showed that Srb/mediator complex and TBP are also potential substrates of Snf1 kinase (49, 83).
It is interesting that Snf1 protein can modify a recombinant Gcn5 protein and that these two proteins are copurified from yeast (Fig. (Fig.6).6). We do not yet know the site(s) modified by Snf1 protein in vitro, nor has it been tested whether Gcn5 protein is phosphorylated in vivo. The human Gcn5 protein was shown to be modified and inhibited by the DNA-dependent kinase (6). In our hands, the in vitro-phosphorylated Gcn5 also seems to exhibit a slightly lower activity on histones H3 and H4 (X. Xu and M.-H. Kuo, unpublished). However, it remains an open question as to whether a phosphorylated Gcn5 protein behaves differently within the context of native complexes.
In conclusion, combining the data presented here and those reported by others, we propose a simple model that that Reg1(1-740) protein uses its newly adopted affinity for Gcn5, while maintaining the Snf1 interaction domain (79), to mediate the interaction between Gcn5 and Snf1 proteins. When Snf1 protein is brought to the vicinity of Gcn5, phosphorylation of Gcn5 protein or another factor(s) within or near the SAGA complex may provide the noncatalytic function that rescues the E173H mutation for effective activation of a subset of Gcn5 target genes.
We are grateful to the following people for generously supplying materials: D. Almy for DA10; C. D. Allis, M. Smith, and J.-Y. Hsu for the histone knockout strain and plasmids; K. Dombek and E. Young for REG1 constructs; D. Thiele for GST-SNF1 constructs; M. Snyder for the mTn library; K. Tatchell for mutant strains of GLC7; A. Acharya for chicken nuclei; and M. Carlson for SNF1 constructs. We also thank Xuqin Wang for technical assistance. S. Triezenberg and A. Acharya are thanked for providing critical comments on the manuscript.
This work was supported by NIH R01 GM62282.