A Poised Initiation Complex Is Activated by SNF1*♦

Snf1, the yeast AMP kinase homolog, is essential for derepression of glucose-repressed genes that are activated by Adr1. Although required for Adr1 DNA binding, the precise role of Snf1 is unknown. Deletion of histone deacetylase genes allowed constitutive promoter binding of Adr1 and Cat8, another activator of glucose-repressed genes. In repressed conditions, at the Adr1-and Cat8-dependent ADH2 promoter, partial chromatin remodeling had occurred, and the activators recruited a partial preinitiation complex that included RNA polymerase II. Transcription did not occur, however, unless Snf1 was activated, suggesting a Snf1-dependent event that occurs after RNA polymerase II recruitment. Glucose regulation persisted because shifting to low glucose increased expression. Glucose repression could be completely relieved by combining the three elements of 1) chromatin perturbation by mutation of histone deacetylases, 2) activation of Snf1, and 3) the addition of an Adr1 mutant that by itself confers only weak constitutive activity.

The general model for eukaryotic transcription activation is that a preinitiation complex forms by ordered recruitment of specific and general transcription factors to a promoter (see reviews in Refs. [1][2][3]. Early in activation, promoter-specific factors bind, possibly with the aid of general chromatin remodeling factors (4,5). Bound activators recruit coactivators including histone modifiers and chromatin remodeling complexes such as SAGA 2 or SWI/SNF. The resulting alterations to chromatin and interactions between bound activator, Mediator complex, and general transcription factors ultimately leads to the recruitment of TBP and RNA Pol II (5)(6)(7).
Although the model implies transcription when all components of the preinitiation complex are in place, several cases are known in which Pol II is recruited to a promoter but does not proceed through the open reading frame without an additional signal (8 -10). The classic example is the Drosophila hsp70 heat shock gene. In the absence of an activating heat shock and with very little bound heat shock factor, Pol II binds the promoter and initiates transcription, only to pause ϳ25 nucleotides from the initiation site. Within seconds after heat shock, additional heat shock factor binds, elongation proceeds through the open reading frame (11), and more Pol II is recruited (10). This mechanism allows for a rapid transcriptional response to a potentially lethal environmental challenge.
Reacting to a change in carbon source does not have the urgency of responding to heat shock, and activation of the glucose-repressed genes does not involve a paused polymerase (see review of glucose repression in Ref. 12). We have, however, detected a polymerase complex bound to a glucose-repressed promoter in a strain deleted for the histone deacetylase (HDAC) genes, HDA1 and RPD3. In HDAC mutants, the Adr1 and Cat8 activators, which normally bind to glucose-repressed promoters only in low glucose conditions, bound constitutively. SNF1, which encodes the yeast AMP kinase homolog that is regulated by environmental stresses, was required for the constitutive activator binding as it is for binding in glucosederepressing conditions (13)(14)(15). At the Adr1-and Cat8-regulated ADH2 promoter, a complex was assembled that contained Pol II and components of SAGA, SWI/SNF, TFIIB, and Mediator. Despite the presence of Pol II, transcription was very low. We used the opportunity of a Pol II complex that appeared to be poised for transcription at a glucose-regulated promoter to determine that there are two subsequent steps in the relief of glucose repression, at least one of which can be triggered by activation of Snf1.

EXPERIMENTAL PROCEDURES
Strains and Primers-Strains are in supplemental Table 1. Yeast strains were grown as described in Sherman (16). Repressing medium contained 5% glucose; derepressing medium contained 0.05% glucose. Epitope tagging, gene deletion, and marker swapping were according to Knop et al.,Cross, respectively, with the exception of SNF1, which was deleted with EcoRI-cut pST70 (20). A multicopy SPT15 (TBP) plasmid was a gift from Steve Hahn's laboratory.
ChIP, ␤-Galactosidase Activity Assay, and Real-time Quantitative PCR-Chromatin immunoprecipitation and gene-specific PCR with gel electrophoresis were performed as described (21). Antibodies were from Santa Cruz Biotechnology Inc. (9E10 anti-Myc and F7 anti-HA) and Abcam (8WG16 anti-Pol II and Ab5131 anti-CTD phosphoserine 5). For expression analysis, RNA was isolated by hot phenol extraction (22) and converted to cDNA with a SuperScript III kit (Invitrogen) according to the manufacturer's directions. ChIP and cDNA were quantified by real-time quantitative PCR (qPCR) with an MJResearch Chromo4 system, using ABI or Quantace SYBR Master Mix, according to the manufacturers' instructions. Primer sequences are available on request. ChIP data were analyzed using the method of Steger et al. (23) or Bryant and Ptashne (6). ␤-Galactosidase assays were performed as described in Guarente (24).
Immunoprecipitations and Western Blots-Anti-HA and anti-Myc antibodies were from Santa Cruz Biotechnology. Anti-Adr1 was from Dombek et al. (25). Immunoprecipitations were carried out as in Strahl-Bolsinger et al. (26), without DNase I treatment and using 2 g of monoclonal anti-HA (F-7) or 6 g of monoclonal anti-Myc (9E10) per mg of lysate. The method of Kushnirov (27) was used for non-immunoprecipitated Western blot samples. Western blots were performed with the Odyssey system (Licor), using 1:500 -1:1000 diluted polyclonal anti-HA (Y-11) or monoclonal anti-Myc (9E10) as primary antibody.
Supercoiling Assay-The Ϫ640 to ϩ135 region of the ADH2 gene was cloned into the multiple cloning site of pALTL, a modified version of pALT (28), resulting in pLLTY1. pLLTY1 was digested with EcoRI, and the resulting ϳ2.5-kbp DNA fragment (which contained the cloned promoter, TRP1 and ARS1) was ligated and used to transform the different yeast strains to Trp prototrophy. Mid-log phase yeast cells from 50-ml cultures grown in either repressing or derepressing conditions were pelleted and washed once in cold water, and the cell pellet was frozen on dry ice and stored at Ϫ70°C. DNA was isolated from cell pellets using glass beads and phenol as outlined by Hoffman (29). DNA was electrophoresed on a 1% agarose, 1ϫ Tris-borate-EDTA gel. Both gel and running buffer contained 15 g/ml chloroquine. Gels were run at a constant 30 V for 22-36 h. Standard Southern analysis techniques were used for gel blotting and probing with a P 32 -labeled 400-bp fragment from pLLTY1. Imaging of blots and the quantitation of band intensities employed a GE Healthcare Storm 840 PhosphorImager and ImageQuant 5.4.
Nucleosome Scanning Assay (NuSA)-200 ml of repressed or derepressed cell cultures was processed using the procedures in the yeast culture, micrococcal nuclease digestion, protein degradation, and DNA purification steps as outlined in Liu et al. (30). Specifically, repressed cells were incubated with zymolyase for 15 min at 30°C, whereas derepressed cells were incubated for 45 min. After the RNase A digestion, samples were analyzed on 2% agarose gel to determine the extent of digestion, and only samples that were highly enriched in mononucleosomal DNA were used. The DNA representing the mononucleosomal fraction was isolated using a Qiagen gel extraction kit. DNA samples were diluted 1:300 -1:500 and used in qPCR reactions to quantify the presence of a specific amplicon. The protection value set for each amplicon corresponds to the -fold enrichment of that amplicon in the mononucleosomal DNA when compared with the undigested sample and normalized to CEN3 values. qPCR primers were designed to cover the promoter of ADH2 and FBP1 with amplicons averaging 100 bp in size (sequences available upon request).

Mutations in HDACs Cause Activator Binding in Non-activating Conditions-In glucose-starved cells, Adr1 and Cat8
contribute to the activation of dozens of glucose-repressed genes (21,31). Neither factor is detected at target promoters in repressed cells by ChIP, but binding is detected after derepression in low glucose (32). In strains deleted for HDA1 and RPD3, Adr1 binds to the ADH2 promoter, even in repressing conditions, without activating transcription (see below and see Ref. 33). We extended this analysis and found that in the HDAC mutants, both Adr1 and Cat8 bound constitutively to the promoters of several Adr1-and Cat8-regulated genes, specifically ACS1, ADH2, ADY2, ATO3, FBP1, JEN1, ICL1, MDH2, MLS1, and PUT4 ( Fig. 1 and supplemental Fig. 1).
Increased Activator Levels in HDAC Mutants Do Not Account for Binding-For both Adr1 and Cat8, binding under repressed conditions was unexpected. CAT8 expression is dependent on Snf1 (14,34), which is inactive in vitro when isolated from glucose-grown cells (35). Expression of ADR1 is not SNF1-dependent (36), but Adr1 binding during derepression requires SNF1 (13). HDAC mutations might increase Adr1 and Cat8 protein levels, resulting in binding by simple mass action, so we examined Adr1 and Cat8 levels by Western blot.
Adr1 and Cat8 were present in glucose-grown cells, although as expected from previous results (34,37,38), their levels increased dramatically upon derepression (Fig. 2, note that Cat8 samples were concentrated by immunoprecipitation before blotting). Levels of both factors were elevated in ⌬hda1⌬rpd3 strains (Fig. 2, A and B). Previous work showed that a multicopy ADR1 strain overproduces Adr1 but does not show binding in repressed conditions (13). Fig. 2C shows that although Adr1 in HDAC mutants was elevated, by comparison with the multicopy strain, these levels alone were not sufficient for binding.
Snf1 Is Required for Activator Binding in HDAC Mutants-The ⌬hda1⌬rpd3 mutants appeared to overcome the regula- tion of Adr1 and Cat8 in two ways: they increased protein levels, and they allowed DNA binding in repressing conditions. Normally, SNF1 is part of Adr1 and Cat8 regulation, so we quantified the DNA binding of Adr1 and Cat8 in the absence of SNF1. In all cases tested, binding in repressed ⌬hda1⌬rpd3 strains was reduced when an additional ⌬snf1 mutation was introduced ( Fig. 1 and supplemental Fig. 1). The effect of ⌬snf1 on Adr1 and Cat8 binding in ⌬hda1⌬rpd3 strains was comparable with the effect on binding in wild-type RPD3 and HDA1; some Snf1-independent binding of Adr1 remained (data not shown and (13), whereas Cat8 binding was dramatically diminished, possibly because of its dual dependence on Snf1 for transcription and for post-translational modification.
Initiation Complex Components Bind ADH2 in HDAC Mutants-Since activator binding is an early step in the recruitment of coactivators such as SAGA, SWI/SNF, and Mediator, we tested for binding of representative subunits of these complexes in repressed ⌬hda1⌬rpd3 cells using ADH2 as our model. ChIP analysis detected components of SAGA (Gcn5), SWI/SNF (Snf2 and Snf5), TFIIB (Sua7), Mediator (Med15, Med18, and Med14), and Pol II (Rbp1) at the ADH2 promoter in repressed HDAC knock-out strains (Fig. 3A, compare background levels in striped bars with binding in ⌬hdac in stippled bars). Comparable results for other genes known to be bound by Adr1 and Cat8 (data not shown for ADY2, FBP1, JEN1, and MLS1) suggested that these complexes were also present at other Adr1-and Cat8-dependent promoters. TBP did not appear to be in the complex, consistent with earlier findings (33), although binding above background was difficult to detect. Consistent with the TBP results, binding of the TBP-interacting protein Sua7 was also relatively low in repressed HDAC mutants. Deletion of ADR1 or SNF1 reduced Pol II occupancy at ADH2 in HDAC mutants by 50% relative to ADR1 or SNF1 wild-type strains (Fig. 3A, last four bars).
Binding of Activators in Repressed HDAC Mutants Does Not Fully Activate Transcription-Despite the binding of activators, coactivators and Pol II, Northern blots had shown little or no ADH2 RNA from repressed ⌬hda1⌬rpd3 cells (33,39). Quantitation with a sensitive qPCR technique detected some transcript from ADH2 in a repressed HDAC mutant, but expression was less than 1% of wild-type derepressed levels. Expression from other Adr1-and Cat8-dependent promoters was generally less than 10% of wild-type derepressed levels ( Fig.  4) with the exception of POT1 (29%) and ICL2 (36%). Fig. 4 shows data for full-length transcripts, but the same results were obtained when one gene (ADY2) was analyzed for production of shorter transcripts using primers to detect the 3Ј end, middle, and 5Ј end of the RNA (supplemental Table 2). RNA abundance was unchanged regardless of the primers used to detect it, suggesting that Pol II was not pausing or stalling in the middle of the gene. Expression of ADR1 and CAT8 was consistent with the protein levels seen in Fig. 2.
Since TBP is a very late addition to the preinitiation complex (6) and ChIP assays showed no difference between TBP at the ADH2 promoter in wild-type and ⌬hda1⌬rpd3 strains, we supplied the cells with an excess of TBP from a high copy plasmid with SPT15 (TBP) under ADH1 promoter control. This resulted in high levels of TBP by Western blot 3 but did not 3 R. Chang, unpublished.   Table  1. For wild type, each strain was compared with its isogenic HDA1RPD3 parent. ChIP for TBP was from strains carrying an N-terminally Myc-tagged SPT15 on a multicopy plasmid. Data are expressed as the ratio of ChIP/input normalized to a non-expressed telomeric region. The highest value of each set is 100%. %IP, percentage of total immunoprecipitated, normalized to background. B, phosphorylation state of the Pol II CTD. Strains were wild type (CTYTY18) and ⌬hda1⌬rpd3 (CTYTY46). Data are presented as described for panel A and are the average of three biological replicates. p ϭ 0.25 for hypophosphorylated Pol II at ADH2 in repressed and derepressed ⌬hdac.
To evaluate the state of the polymerase poised at the ADH2 gene, we did ChIP for phosphorylation of serine 5 of the CTD, a modification suggestive of initiation (40). The paused polymerase at the Drosophila hsp70, hsp26, GAP, TUB, and Actin5C genes is Ser-5-P-modified (10). In repressed ⌬hda1⌬rpd3 cells, Ser-5-P was detected at the ADH2 promoter at 17% of derepressed levels (Fig. 3B). Hypophosphorylated Pol II, detected with an antibody against unphosphorylated CTD, was present at 42% of derepressed levels, comparable with the total Pol II seen in Fig. 3A. Since different antibodies have different precipitation efficiencies, we could not compare ChIP values directly, but the poised polymerase appeared to be a mixture in which only a fraction was in the Ser-5-P state, measured relative to derepressed levels. By the same criterion, a greater fraction was hypophosphorylated. As a control, CTD Ser-5-P was examined in repressed and derepressed wild-type cells, at the 5Ј end of the housekeeping gene ACT1. Levels decreased in low glucose, consistent with reduced transcription during slower cell growth (Fig. 3B).
ADH2 Promoter Chromatin Structure Is Altered in Repressed HDAC Mutants-Transcription of Adr1-dependent genes upon activation in low glucose is associated with an Adr1-dependent chromatin remodeling event at the promoter (41,42). We used two techniques, supercoiling analysis and NuSA, to determine whether activator binding without low glucose signaling in a ⌬hda1⌬rpd3 strain affected the ADH2 promoter chromatin structure. The supercoiling analysis relies on the fact that the amount of negative supercoiling on an isolated DNA plasmid is proportional to the number of nucleosomes assembled on the DNA in vivo (43). Thus by determining the change in the distribution of topoisomers between samples, one can determine the change in nucleosome density. The Ϫ640 to ϩ135 region (relative to the ATG initiation codon) of the ADH2 gene was cloned into a multicopy TRP1/ARS1 yeast plasmid (28). In this plasmid, the cloned promoter, TRP1 and ARS1, can be released by digestion with EcoRI, ligated, and used to transform yeast to Trp prototrophy. The 2.5-kbp episome has the same chromatin architecture as the chromosomal ADH2 promoter, and Adr1dependent loss of nucleosome density is observed in derepressing conditions. 4 When the topoisomer distribution was compared in the wild-type strain between repressed, 2.5-h derepressed, or 5-h derepressed, a downward shift representing loss of one nucleosome occurred between 2.5 and 5 h of derepression (Fig. 5, A and B). This shift in the distribution of topoisomers was not observed when DNA from a ⌬snf1 mutant was examined after growth in repressing and derepressing conditions (Fig. 5A, lanes 7 and 8). However, using a ⌬hda1⌬rpd3 strain, the topoisomer distribution seen in the repressed and 2.5-h derepressed samples was similar to the distribution seen at 5 h of derepression in the wild-type strain, indicating that the nucleosome density in the ⌬hda1⌬rpd3 strain in high glucose resembles the wild-type density in low glucose (Fig. 5A, lanes 4  and 5, and B). Upon derepression, there is a greater loss of nucleosomes in the ⌬hda1⌬rpd3 strain (Fig. 5A, lane 6, and B).
The second technique used to detect changes in the chromatin structure was a NuSA. NuSA quantifies the micrococcal nuclease sensitivity of a DNA sequence in vivo and was used to detect nucleosome positioning and density at the ADH2 promoter region. The monosomal DNA resulting from micrococcal nuclease digestion was analyzed using an array of tiled amplicons spanning the ADH2 promoter region and qPCR. The NuSA for wild-type repressed cells showed the three previously described nucleosomes: N-2 at the 5Ј end of the nucleosome free region, N-1 covering the TATA box, and Nϩ1 covering the translational start site (42) (Fig. 5C). When wild-type cells were shifted to derepressing conditions, two changes in the chromatin were seen. An overall reduction of nucleosome density was represented by a decrease in the amplitude of the nucleosome protection peaks and a small but reproducible 3Ј shift in the position of N-1 (Fig. 5C, inset). Nucleosome density in repressed ⌬hda1⌬rpd3 mutants was comparable with derepressed wild type; however, the small shift in N-1 position was not seen (Fig. 5C). This agrees with data from Verdone et al. (33), showing increased acetylation and micrococcal nuclease sensitivity of the ADH2 promoter in the HDAC mutants. The FBP1 promoter was also analyzed by NuSA (data not shown), and the same trend was seen. There was an overall decrease in nucleosome density in derepressed chromatin when compared with repressed chromatin, and the nucleosome density in the repressed ⌬hda1⌬rpd3 mutant more closely resembled the wild-type derepressed than the wild-type repressed samples.   DECEMBER 28, 2007 • VOLUME 282 • NUMBER 52

JOURNAL OF BIOLOGICAL CHEMISTRY 37311
Snf1 Activation and a Non-phosphorylatable Adr1 Contribute to Activation-The repressed ⌬hda1⌬rpd3⌬ strains presented a unique opportunity for investigating late stage steps in activation of glucose-repressed genes. RNA Pol II and associated transcription factors appeared to be poised at ADH2 and other Adr1-and Cat8-dependent promoters. Some chromatin remodeling had occurred, and possibly some initiation, but glucose repression still persisted.
We hypothesized that one or more final steps, perhaps involving the kinase activity of Snf1, were required for full derepression. Snf1 is kept inactive by the Reg1.Glc7 phosphatase complex (44,45), so REG1 was knocked out in ⌬hda1⌬rpd3 and isogenic HDA1 RPD3 strains, and RNA was measured by qPCR. As seen previously (25,46,47), 5 in a ⌬reg1 strain, the Snf1-dependent SUC2 gene was almost fully activated on glucose (80 -115% of wild-type derepressed), and ADH2 was expressed to ϳ7% of wild-type derepressed levels ( Fig. 6A and data not shown). Low glucose increased ADH2 expression nearly 5-fold, to 33% of wild-type derepressed levels. The generation time of the ⌬reg1 strain was more than double that of wild type, so slow growth may be responsible for low ADH2 expression. Adding the ⌬reg1 deletion to an ⌬hda1⌬rpd3 strain showed a clear synergistic effect. Expression of Adr1-and Cat8-dependent genes was as high or higher than a derepressed ⌬reg1HDA1RPD3 strain. Glucose repression still prevailed for all genes tested except SUC2 as further activation was seen after removing glucose (Fig. 6B and data not shown for the Adr1-or Cat8-bound ADY2, ICL2, FBP1, and MLS1 genes).
Another activation event that might be absent in the ⌬hda1⌬rpd3 mutants involves Adr1. Although the precise mechanism is unknown, mutations in the region of the phosphorylated serine 230 in Adr1 relieve glucose repression of ADH2, leading to up to 10% of fully derepressed levels (48,49). We confirmed this by qPCR on transcripts from an ⌬adr1 strain with the ADR1 S230A allele (Adr1 c ) on a CEN plasmid. Comparable with earlier results (25), the ADR1 c allele caused constitutive expression of ADH2 but to 2% or less than wildtype derepressed levels (Fig. 6A). It had a hyperactivating effect in derepressing conditions, even for FBP1 and MLS1 (data not shown), which are not strongly Adr1-dependent, although Adr1 can be detected at their promoters (32). Adr1 c had a greater than additive effect when combined with the ⌬hda1⌬rpd3 mutations, and hyperactivation was seen when glucose was depleted (Fig. 6B).
Dombek et al. (25) showed that combining the ADR1 c allele and ⌬reg1 caused constitutive ADH2 expression to 20% of wildtype derepressed levels, indicating that Reg1 and Adr1 c act in separate pathways. Lack of full expression suggested other influences on expression. We added the dominant (48, 49) ADR1 c allele on a plasmid to the ⌬reg1 strains used in Fig. 6A and found that it had the expected synergistic effect at ACS1, ADH2, and FBP1 ( Fig.  6B and data not shown). At ADH2 and ACS1, expression in glucose was 300% of wild-type derepressed levels, yet glucose repression was still operational because shifting cells to low glucose gained further expression.  . Nucleosome density at the ADH2 promoter is lower in ⌬hda1⌬rpd3 when compared with wild type. A, supercoiling analysis on strains containing a plasmid with the ADH2 promoter was performed as described under "Experimental Procedures." Samples from wild-type (LLTY4) and ⌬hda1⌬rpd3 (LLTY66) strains were repressed (R) and 2.5-or 5-h derepressed (DR). Samples from ⌬snf1 (LLTY69) were repressed or derepressed for 5 h. The asterisk indicates nicked DNA, and the arrowhead indicates the location of band 1. Under the conditions used, when the mean distribution of topoisomers shifts toward the bottom of the gel, the nucleosome density on the plasmid has decreased. B, topoisomer distributions were quantified using phosphorimaging. The data are presented as the total intensity for each of eight bands. Since repressed and 2.5-h derepressed within a strain was the same for both wild type and ⌬hda1⌬rpd3, the averages of these two samples were graphed (R) with the error bars indicating the range between the two averaged samples. Top and bottom indicate the location on the gel. C, the NuSA assay described under "Experimental Procedures" was used to determine the nucleosome positioning and density for the ADH2 promoter. Each point shows the relative protection for each PCR amplicon, quantitated by real-time PCR and normalized to a centromeric control. From left to right, peaks show the Ϫ2, Ϫ1, and ϩ1 nucleosomes. The inset illustrates the N-1 shift in wild-type derepressed sample. The same data were used for the inset as for the main graph, except that each point was graphed as the percentage of the highest value within each strain. Samples were wild-type (TYY201) repressed (R) or derepressed for 4 h (DR) or ⌬hda1⌬rpd3 (CTYTY44) repressed.
Combining ⌬hda1⌬rpd3, Snf1 Activation, and Adr1 c Completely Relieves Glucose Repression-At this point, three factors had been identified that caused transcription complex assembly or partial activation of ADH2 in repressing conditions: chromatin perturbation (⌬hda1⌬rpd3, less than 1% of full derepression), Snf1 activation (⌬reg1, 7% full derepression), and hyperactivity of Adr1 (ADR1 c , 1% full derepression). Combinations of ⌬hda1⌬rpd3 with ⌬reg1, ⌬hda1⌬rpd3 with Adr1 c , or ⌬reg1 with Adr1 c suggested that they contribute separately to activation, but since ADH2 is TBP-dependent (13), TBP must have been recruited sufficiently in all of these combinations since some transcription was detected. Nonetheless, some glucose repression remained. Therefore, we combined all three factors in a ⌬reg1⌬hda1⌬rpd3 strain with Adr1 c . Glucose repression was completely eliminated, and expression of Adr1dependent genes in a ⌬reg1⌬hda1⌬rpd3/Adr1 c strain was as high in repressing conditions as in the same strain in derepressing conditions (Fig. 6C). Overall expression of glucose-repressed genes was lower than in a fully wild-type strain, possibly because growth of the triple mutant with hyperactive Adr1 was significantly slower.

Chromatin Perturbation Allows Activator and Preinitiation
Complex Binding at ADH2-Consistent with current models of coactivator recruitment, we found that when the Adr1 transcription factor bound its target promoter ADH2, it recruited a partial preinitiation complex that included Pol II, even in nonactivating conditions. Some CTD Ser-5-P could be detected, and although this is not definitive evidence of initiation, it suggested that at least some of the Pol II was transcription-competent. Binding of coactivators and Pol II after a nonspecific chromatin perturbation was consistent with recent data on the phosphate-repressible PHO5 promoter, where initial chromatin remodeling by SWI/SNF and SAGA, and later remodeling by the histone chaperone Asf1, are key events in activator binding and subsequent transcription complex assembly (4). Zhang and Reese (50) showed that disrupting a positioned nucleosome over the TATA box of the DNA damage-inducible RNR3 led to Pol II recruitment in repressed conditions, although in this case, significant transcription was achieved. Some chromatin remodeling of the ADH2 promoter had occurred in the HDAC mutant, but the TATA-containing nucleosome had not been repositioned as it is in derepressed cells.
Recruitment of coactivators and Pol II demonstrates that the Adr1 activation domain(s) are functionally accessible in repressing conditions unlike the activation domains of some inducible transcription factors (e.g. peroxisome proliferator-activated receptors). As in derepressing conditions, formation of the partial initiation complex in the HDAC mutant required activator, and activator binding required SNF1. Nevertheless, the presence of this complex was insufficient for full transcription, and the promoter was still subject to glucose repression.
Snf1 Has Activities in Glucose-grown Cells-The SNF1 requirement for binding at glucose-repressed genes in repressing conditions was surprising because Snf1 is considered to be inactive at these promoters in these conditions. Although Snf1 purified from glucose-grown cells does not show in vitro kinase activity against a peptide substrate (35), SNF1 is required for full INO1 expression when it is induced by low inositol, even in the presence of glucose (51)(52)(53). In repressed cells, Snf1 coimmunoprecipitates with its substrate Mig1 (54), indicating some interaction with target proteins, even when it may not be catalytically fully active. We have seen that ⌬snf1 increases the proportion of Adr1 that is phosphorylated on Ser-230 in both glucose-repressing and low glucose-derepressing conditions, 6 although the effect on DNA binding, coactivator interaction, and transcription activation is still under investigation. Recently, Hong and Carlson (55) have shown that the kinase activity of Snf1 is activated in glucose-grown cells that are subjected to high pH or high NaCl. Constitutive, Snf1-dependent DNA binding of Adr1 and Cat8 when promoter chromatin structure is altered might be because the ⌬hda1⌬rpd3 mutations introduce a particular cell stress that partially activates Snf1.
Pol II recruitment since adding a REG1 deletion to the histone deacetylase deletion strains had a greater than additive effect on transcription than either of the mutations alone. We believe that Snf1 activation, rather than a pleiotropic effect of ⌬reg1, is responsible. REG1 deletion is commonly used to deregulate Snf1. Snf1 is activated through phosphorylation of its serine 210 (35, 56 -59) and inactivated through dephosphorylation by a PP1 type 1 protein phosphatase composed of Glc7 and the regulatory subunit Reg1 (35,45,56). As a PP1 phosphatase, Glc7 has many substrates (60) and therefore pleiotropic effects. Reg1, however, as the subunit that targets Glc7 to Snf1 (44,61), is fairly specific; without Reg1, Snf1 is constitutively phosphorylated and active (56,61). An allele of SNF1 with the Ser-210 codon mutated to Asp, which might be expected to mimic the phosphorylated state, abolished the kinase activity (62), so REG1 deletion remains the most effective way to constitutively activate Snf1. As confirmation to the ⌬reg1 results, we used a dominant SNF4-204 allele, which produces a Snf1-activating subunit that constitutively binds Snf1 (51). This also stimulated ADH2 expression in the HDAC mutant to 2-fold over background. 7 The Snf1-dependent post-Pol II recruitment step could be chromatin modification (63), although preliminary results showed that the histone H3-S10A mutation does not significantly affect ADH2 expression, 8 so if Snf1 modifies chromatin at ADH2, it is probably not through H3 Ser-10 phosphorylation. Snf1 could be acting through Pol II, TBP, or Mediator. By twohybrid and coimmunoprecipitation analysis, Snf1 associates with both Ctk1 (64), a component of the kinase that phosphorylates the Pol II CTD on serine 2 during elongation (65), and the Mediator components Med16, Cdk8, and CycC (66). Genetic evidence also links Snf1 to TBP (51) and to Mediator as mutations in several Mediator genes have been isolated as ⌬snf1 suppressors (67). We are currently investigating whether Snf1 is helping to recruit specific coactivators to the initiation complex in HDAC mutants. Another possible late role for Snf1 is reverse recruitment, transporting genes to regions of active transcription at nuclear pores since Snf1 localizes to the nuclear periphery during derepression (68). Further steps that are triggered by low glucose and may or may not involve Snf1 occur beyond the late Snf1-dependent step since a shift to low glucose induced further gene expression. The Adr1 activator itself is implicated since an S230A mutant allowed full expression in the ⌬reg1⌬hda1⌬rpd3 strain.
Our model (Fig. 7) is that, in repressed ⌬hda1⌬rpd3 cells, Snf1 is sufficient to allow binding of Adr1 and Cat8. Deletion of REG1 either increases the amount of active Snf1 or stimulates a different activity of Snf1 that leads to activation-competent Adr1 and Cat8. For example, Snf1 may be required for a postbinding phosphorylation of Cat8 that leads to activation of transcription. Unphosphorylated Cat8 appears to be competent to bind because Heisinger et al. (69) found that bacterially produced Cat8, which is not expected to receive its normal modifications, showed in vitro DNA binding. Phosphorylation is important to the activator function of Cat8 because a S562E point mutation in Cat8, which mimics the phosphorylated state, shows some constitutive activity (15). In support, the genes that we find to be the least Cat8-dependent, ICL2 and POT1, have the highest expression in repressed REG1⌬hda1⌬rpd3 cells (Fig. 4), suggesting that Snf1 activation is not as critical for these genes. Modifications to Adr1 are not as well characterized, but we hypothesize that we see relief of glucose repression in the ⌬reg1⌬hda1⌬rpd3/Adr1 c strain because Adr1 c mimics active Adr1, through more productive binding or interaction with coactivators, leading to full activation in repressing conditions. FIGURE 7. Snf1 acts at a step after RNA Pol II binding. As shown in the model, in repressed wild-type cells (WT R), histones are deacetylated, Cat8 and Adr1 (striped and black shapes, respectively) do not bind the ADH2 promoter, and coactivators (stippled cloud) are not recruited. In repressed ⌬hda1⌬rpd3 cells (hdac⌬ R), histones are acetylated, activator binding requires Snf1, and coactivators and Pol II (arrow) are recruited. Transcription does not occur. In repressed ⌬hda1⌬rpd3⌬reg1/Adr1 c cells, additional or alternate activation of Snf1 leads to phosphorylation of additional substrates. Combination with hyperactivity of Adr1 c allows full derepression of this strain. Circled P stands for phosphorylation; circled C indicates Adr1 C .