Redundant Mechanisms Are Used by Ssn6-Tup1 in Repressing Chromosomal Gene Transcription in Saccharomyces cerevisiae *

The Ssn6-Tup1 corepressor complex regulates many genes in Saccharomyces cerevisiae . Three mechanisms have been proposed to explain its repression functions: 1) nucleosome positioning by binding histone tails; 2) recruitment of histone deacetylases; and 3) direct interference with the general transcription machinery or activators. It is unclear if Ssn6-Tup1 utilizes each of these mechanisms at a single gene in a redundant manner or each individually at different loci. A systematic analysis of the contribution of each mechanism at a native promoter has not been reported. Here we employed a genetic strategy to analyze the contributions of nucleosome positioning, histone deacetylation, and Mediator interference in the repression of chromosomal Tup1 target genes in vivo . We exploited the fact that Ssn6-Tup1 requires the ISW2 chromatin remodeling complex to es-tablish nucleosome positioning in vivo to disrupt chromatin structure without affecting other Tup1 repression functions. Deleting ISW2 , the histone deacetylase gene HDA1 , or genes encoding Mediator subunits individually caused slight or no derepression of RNR3 and HUG1 . However, when Mediator mutations were combined with (cid:1) isw2 or (cid:1) hda1 mutations, enhanced transcription was observed, and the strongest level of derepression was observed in triple (cid:1) isw2 / (cid:1) hda1 /Mediator mutants. The increased transcription in the mutants was not due to the loss of Tup1 and to (Amersham Biosciences) capillary blotting. UV-cross-linking and 4-h prehybridization at radioactively labeled gene-specific Nuclease Mapping— Nuclei out essentially as described previously (13, 1 of cells was grown in YPAD to an A 600 of around 1.0, and digested with Zymolyase T100 (Seikagaku). nuclei were isolated by differential centrifugation and resuspended in digestion buffer according to the size of the nuclei pellet, and digested by 0, 2, 4, and 8 units/ml of micrococcal nuclease (MNase, Worthington) for 10 min at 37 °C. purified DNA was digested with PstI restriction enzyme, and the products were detected by Southern Blotting using a 200-bp probe specific for the PstI fragment Genomic DNA digested was loaded

The Ssn6-Tup1 complex represses more than 180 genes in Saccharomyces cerevisiae controlled by different pathways (1) and is, thus, considered a global corepressor of transcription. It does not directly bind to DNA, but is targeted to promoters by sequence-specific DNA binding repressors (for review see Ref. 2). Although Ssn6 and Tup1 form a complex, it is largely accepted that Tup1 contributes the bulk of the repression activity and Ssn6 acts as an adaptor. Recruitment of Ssn6 to promoter via the LexA DNA binding domain can repress tran-scription of a reporter plasmid in a Tup1-dependent manner, while repression by LexA-Tup1 is independent of Ssn6 (3)(4)(5). Furthermore, overexpression of TUP1 can partially suppress the mating defect of a ⌬ssn6/⌬tup1 double mutant, whereas overexpression of SSN6 cannot (6).
A second Ssn6-Tup1 repression model, which is not mutually exclusive with the first, is by its recruitment of histone-modifying enzymes. Ssn6-Tup1 interacts with the class I histone deacetylases (HDACs) 1 Rpd3, Hos1, and Hos2 (20,21) and the class II HDAC Hda1 (18). Recruitment of Ssn6-Tup1 to promoters causes histone hypoacetylation, which can be reversed by deleting RPD3/HOS1/HOS2 or HDA1 (5,18,20). A recent paper showed that a triple ⌬rpd3/⌬hos1/⌬hos2 mutant has increased histone acetylation and reduced cross-linking of Tup1 to promoters (19). Thus HDAC recruitment may form a positive feedback loop in establishing and expanding a repressive chromatin domain: to repress transcription locally upon recruitment by Ssn6-Tup1 and facilitate the spreading of Tup1 into adjacent regions.
In addition to the chromatin-dependent repression mechanisms described above, Ssn6-Tup1 can also function by blocking activators and interfering with the transcription machinery (2,22,23). Ssn6-Tup1-dependent repression of transcription from a plasmid bearing an ␣2/Mcm1 binding site has been observed in vitro (24,25). In support of the basal machinery interference mechanism, mutations in various components of the Mediator sub-complex of the RNA polymerase II holoenzyme cause derepression of Ssn6-Tup1-regulated plasmid reporter systems (for review see Ref. 26), and direct interactions between Ssn6-Tup1 and the Mediator subunits Hrs1/Med3 (27,28), Srb7 (29), and Srb10 (30) have been reported. However, the effects of Mediator mutations on native chromosomal genes are less clear. Even in combination, Mediator Srb8 -10 mutations caused no defect in the repression of multiple native Tup1-dependent genes, and the combination of Mediator mutations with histone tail mutations did not affect repression of ANB1 (31,32).
The RNR3 gene encodes for the large subunit of the enzyme ribonucleotide reductase (RNR), which plays a crucial role in DNA replication and repair. RNR3 and the other RNR genes (RNR1, 2, and 4) are under the tight control of the Ssn6-Tup1 complex (13,33,34). The corepressor complex is recruited to the RNR genes by the sequence-specific DNA-binding protein Crt1, which recognizes the DNA damage response elements (DREs) in the upstream repression sequence (URS) (19,35). Crt1 interacts with Ssn6-Tup1 in vitro and recruits it to promoters in the absence of DNA damage, and the activation of the DNA damage checkpoint pathway causes the release of Crt1 and Tup1 from the promoter (14,35). Nucleosomes are precisely positioned over RNR3 in the repressed state, and inducing the cell with the DNA-damaging agent MMS causes the release Crt1-Ssn6-Tup1 from the URS and chromatin remodeling, suggesting a role for chromatin structure in RNR3 gene regulation (13,14,35). Likewise, deleting CRT1, SSN6, or TUP1 disrupts the nucleosome array over RNR3 (13,14). Recent work from our laboratory suggested that nucleosome positioning by Ssn6-Tup1 requires the chromatin remodeling/nucleosome spacing complex ISW2. Deleting the gene coding for the catalytic subunit of the ISW2 complex, ISW2, disrupts nucleosome positioning without causing significant transcription derepression of the DNA damage-inducible RNR3 or stress-inducible ENA1 genes (14), suggesting that pathways in addition to nucleosome positioning repress Tup1regulated genes.
In this study, we used DNA damage-inducible genes as our primary models to analyze systematically the contribution of each Ssn6-Tup1 repression mechanism in transcriptional regulation. We find that Tup1 utilizes multiple redundant mechanisms to repress transcription, which include nucleosome positioning, histone deacetylation, and Mediator interference. We propose that Ssn6-Tup1 has developed multiple mechanisms to function as a "global" corepressor, and different groups of genes have developed different strategies to utilize Ssn6-Tup1 in repression.

MATERIALS AND METHODS
Strains and Media-The strains used in this study are listed in Table  I. Gene deletions were constructed by one-step replacement using PCRgenerated cassettes (36). Detailed information on their construction will be provided upon request. In all cases, cells were grown in 2% peptone, 1% yeast extract, 20 g/ml adenine sulfate, and 2% dextrose (YPAD) at 30°C. The induced cells were treated, at the A 600 of 0.6 -1.0, with methyl methane sulfate (MMS) at a concentration of 0.03% for 2-2.5 h.
RNA Isolation and Northern Blot-RNA isolation was carried out as previously reported (37). 10 -15 ml of yeast cell grown in YPAD, treated or untreated with MMS, were harvested by centrifugation, washed with STE buffer (10 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA), and resuspended in RNA preparation buffer (1% SDS, 100 mM Tris-HCl, pH 7.5, 10 mM EDTA, 500 mM NaCl). RNA was released by glass bead disruption in the presence of 150 l of phenol/chloroform, extracted once more with phenol/chloroform, precipitated with ethanol, and dissolved in diethyl pyrocarbonate-treated water. RNA concentration was then quantified by measuring its absorption at 260 nm. 20 g of total RNA was separated on 1% (1.6% for ANB1 analysis) formaldehyde agarose This study gels and transferred to nylon membrane (Amersham Biosciences) by capillary blotting. After UV-cross-linking and 4-h prehybridization at 65°C, radioactively labeled gene-specific probes were added. Micrococcal Nuclease Mapping-Nuclei preparation was carried out essentially as described previously (13,38). In brief, 1 liter of cells was grown in YPAD to an A 600 of around 1.0, harvested, and digested with Zymolyase T100 (Seikagaku). The nuclei were isolated by differential centrifugation and resuspended in digestion buffer according to the size of the nuclei pellet, and digested by 0, 2, 4, and 8 units/ml of micrococcal nuclease (MNase, Worthington) for 10 min at 37°C. The purified DNA was digested with PstI restriction enzyme, and the products were detected by Southern Blotting using a 200-bp probe specific for the PstI fragment (13). Genomic DNA digested by (PstI plus MluI) and (PstI plus EagI) was loaded as the marker.
Chromatin Immunoprecipitation-Chromatin immunoprecipitation was performed as described in previous publications with minor changes (39,40). 50 ml of cells was grown in YPAD media to A 600 of 0.5-1.0 and cross-linked for 15 min at 23°C by the addition of formaldehyde to 1%. The induced cells were treated at A 600 of 0.7 with 0.03% MMS and incubated for 2 h before cross-linking. Cross-linking was terminated by the addition of glycine (125 mM) followed by incubation at room temperature for 15 min. Extracts were prepared by glass bead disruption, and the chromatin was sheared into fragments averaging in size of 300 -400 bp. The extracts were then clarified by centrifugation, and 200 l was used for immunoprecipitation. 1 l of di-acetylated H3 antibody (Upstate), anti-TBP polyclonal antiserum, anti-myc (9E10, Covance), 8WG16 monoclonal antibody (Covance), and 1/300 diluted anti-Tup1 polyclonal antibody were used routinely. The immune complexes were isolated on 25 l of protein A-Sepharose CL-4B beads (Amersham Biosciences) and washed, and the DNA was eluted. After reversing the cross-links at 65°C overnight, the immunoprecipitated and input DNA were analyzed by semi-quantitative PCR. The PCR products were analyzed on agarose gels, stained with ethidium bromide, scanned with a Typhoon system (Molecular Dynamics), and quantified by ImageQuaNT software (Amersham Biosciences). The amplified immunoprecipitated DNA was normalized to the input samples. Results are averages and standard errors from three to six batches of independently prepared extracts.

Repression of DNA Damage-inducible Genes in the Absence of
Nucleosome Positioning-The RNR3 gene is an Ssn6-Tup1-dependent gene whose regulation correlates with nucleosome positioning and disruption (13). Our previous studies indicate that disrupting nucleosome positioning is not sufficient to achieve a high level of derepression or preinitiation complex assembly, suggesting that other mechanisms carry out repression in the absence of nucleosome positioning (14). Thus, we set out to define what pathways mediate Tup1 repression when nucleosome positioning is disrupted. Fig. 1A shows that deleting the gene encoding the catalytic subunit of the ISW2 complex, ISW2 (41), disrupts nucleosome positioning over the promoter and into the coding sequence of RNR3. Moreover, the pattern is indistinguishable from that of digested naked DNA or that of a ⌬crt1 mutant, suggesting that nucleosome positioning is completely disrupted in the ⌬isw2 mutant. Specifically, the TATA box and promoter is exposed to micrococcal nuclease digestion (Fig. 1A). On the other hand, transcription of RNR3 was only moderately increased ( Fig. 1B), about a 2.5-fold increase, versus a 40-to 50-fold increase in fully derepressed cells (ϩMMS). Importantly, RNR3 can be fully derepressed in the ⌬isw2 cells, suggesting that the failure to observe a high level of transcription in these cells is not caused by an indirect effect on the derepression/activation machinery. A similar transcription phenotype was observed at HUG1, another gene controlled by Crt1-Ssn6-Tup1 and the DNA damage checkpoint pathway (42). Like RNR3, the chromatin structure at HUG1 is dependent upon ISW2 (14). These results suggest that disruption of nucleosome positioning is insufficient to cause a high level of derepression of these DNA damage-inducible genes.
The Histone Deacetylase Hda1 Is Required for Full Repression of the DNA Damage-inducible Genes-Because one mech-anism of Tup1-mediated repression involves the recruitment of histone deacetylases (HDACs), we explored a role for HDA1 in repression. Hda1 is a class II HDAC that interacts with Tup1 (18), and the global gene expression profiles of ⌬hda1 and ⌬tup1 mutants overlap considerably (43). Derepression of RNR3 by MMS treatment results in a 3-fold increase in the acetylation of lysines 9 and 14 of H3 ( Fig. 2A, upper left) (40). We found that deleting HDA1 resulted in an increase in H3 acetylation to a level comparable to that of the induced condition ( Fig. 2A, upper left), suggesting that Hda1 is required for deacetylating histone H3 at the RNR3 promoter. However, only slight, if any, increase in RNR3 transcription was observed in FIG. 1. Analysis of the chromatin structure and transcription in a ⌬isw2 mutant. A, MNase nucleosome mapping of the RNR3 gene in wild type (WT), ⌬isw2, and ⌬crt1 cells. M is a genomic DNA molecular marker. ND is MNase-digested naked DNA. The location of DNA damage response elements (DREs) is indicated by the lines between nucleosomes Ϫ1 and Ϫ2. The TATA box is located within nucleosome Ϫ1. RNR3 is fully derepressed, and chromatin is disrupted in a ⌬crt1 strain (Li and Reese (13)). B, Northern blot analysis of RNR3 and HUG1 mRNA in MMS-treated or untreated wild type and ⌬isw2 cells. The small cytoplasmic RNA scR1, transcribed by RNA polymerase III, is the loading control. The numbers below the RNR3 and HUG1 blots are the quantification of the signals, normalized to the scR1 loading control, with the value of the untreated wild type cell arbitrarily set as 1.
FIG. 2. Analysis of HDA1 in the repression of RNR3 and HUG1. A, ChIP assay monitoring the level of histone H3 (K9, 14) acetylation and the cross-linking of Tup1, TBP, and RNA polymerase II (Rpb1) to the RNR3 promoter. Cells were treated, or not, with 0.03% MMS for 2 h prior to cross-linking. A primer pair corresponding to the URS (Ϫ448 to Ϫ236) was used to amplify DNA cross-linked to Tup1 and to the promoter (Ϫ179 to ϩ8) for the others. Each immunoprecipitation was normalized to the input and the amount of DNA precipitated from untreated wild type cells, which was arbitrarily set as 1. B, Northern blot of RNR3 and HUG1 mRNA. The results from untreated cells are shown in the upper panel (lanes 1-8), and those from cells treated with MMS are in the lower panel (lanes 9 -16). The signals were quantified and normalized as described in the legend of Fig. 1. C, MNase mapping of nucleosome positioning at RNR3 in ⌬hda1 mutants. The experiments were conducted as described in the legend to Fig. 1A and under "Materials and Methods." the ⌬hda1 mutant, as compared with that of MMS-treated cells (Fig. 2B, compare lanes 1, 3, and 9). A slightly higher level of derepression was observed at HUG1, about 3-to 4-fold, but this level was significantly below that of MMS-treated cells. Because certain HDACs may play a role in transcription activation (44,45; and see below), it is a formal possibility that the failure to observe a high level of derepression in the ⌬hda1 cells results from the muting of transcription due to the positive effects of HDACs. However, data arguing against this are the robust MMS-induced increase in RNR3 and HUG1 transcription in the ⌬hda1 mutant (Fig. 2B, lane 11; and see below). To support the Northern blot data, we performed chromatin immunoprecipitation (ChIP) assays to monitor the cross-linking of TBP and RNA polymerase II (Rpb1), to analyze preinitiation complex (PIC) formation. A 3-and 8-fold increase in TBP and Rpb1 cross-linking, respectively, was detected in MMS-treated versus untreated cells. In contrast, no significant change was detected in the ⌬hda1 mutant ( Fig. 2A, bottom). Thus, mimicking histone H3 acetylation increases by this genetic strategy failed to activate the gene or cause PIC formation. We next examined if Tup1 is present at RNR3 in the ⌬hda1 mutant, and found that Tup1 cross-linking was not reduced, suggesting it is capable of repressing transcription though histone H3 acetylation is high.
Most, if not all, genes undergo both chromatin disruption and enhanced histone acetylation upon induction, and RNR3 is no exception (Figs. 1 and 2) (40). Thus, we wondered if simultaneously disrupting nucleosome positioning and increasing histone acetylation by deleting ISW2 and HDA1, respectively, would cause an enhanced level of derepression of RNR3 and HUG1. The level of derepression was compared in a double ⌬isw2/⌬hda1 mutant and the single mutants, and we found that the double mutant had higher levels of derepression than either of the single mutants. HUG1 was more sensitive to the mutations than RNR3, displaying a higher level of derepression. However, the levels of transcription were significantly below that of MMS-treated cells. This suggests that even though both ISW2 and HDA1 play a role in the repression of DNA damage-inducible genes, a significant level of repression remains when nucleosome positioning is disrupted and histone H3 is acetylated.
Next, we explored the role of RPD3 in the repression of RNR3 and HUG1. ISW2 collaborates with RPD3 in the repression of many genes in vivo, and a microarray analysis reports that RNR3 is up-regulated about 1.6-fold in a ⌬isw2/⌬rpd3 mutant (46). However, deletion of RPD3 did not cause derepression of RNR3 or HUG1, and in fact, reduced the basal and MMS-induced levels of gene expression (Fig. 2B, compare lanes 1, 5, 9, and 13). The MMS-induced levels of transcription in the ⌬rpd3 mutant are reduced approximately 3-to 6-fold for RNR3 and HUG1, respectively. Moreover, the induction of RNR3 and HUG1 was reduced in all mutants containing the ⌬rpd3 mutation (Fig. 2B, lanes 13-16; also see below, Fig. 3A). A positive role of the class I HDACs in gene expression agrees with recent reports (44,45) and our own findings. 2 These results argue against the requirement for RPD3 in the repression of DNA damage-inducible genes, but for a role in their activation. A double ⌬hda1/⌬rpd3 mutant showed no more transcription than the ⌬hda1 single mutant (lane 7). Surprisingly, the double ⌬isw2/⌬rpd3 mutant did not show a significant level of derepression, and in fact, the ⌬rpd3 mutation suppressed the weak derepression observed in the ⌬isw2 background (lane 6). This is consistent with RPD3 playing a positive role in the transcription of DNA damage-inducible genes but is different from the microarray data (46). We speculate that strain differences or limitations in detecting small changes in expression using the microarray technique explain the discrepancy.
Histone Acetylation Is Insufficient to Disrupt Chromatin Structure over RNR3-Histone acetylation neutralizes positive charges on the histone tails and could weaken DNA-histone interactions, making the chromatin structure more permissive for transcription. It is unknown if increasing histone acetylation can itself disrupt chromatin structure, so we examined the chromatin structure of RNR3 in a ⌬hda1 strain by MNase mapping. In contrast to the dramatic chromatin remodeling observed in the ⌬isw2 strain, no evidence for disrupted nucleosome positioning was detected in the ⌬hda1 mutant (Fig. 2C, lanes 3-10). Specifically, the regularly positioned nucleosomes embedding the TATA box and coding sequence were intact in the ⌬hda1 mutant. So the increase in histone acetylation caused by the ⌬hda1 mutation did not affect the ability of Ssn6-Tup1 to position nucleosomes at RNR3, and steps in addition to histone acetylation are required for chromatin remodeling.
Because the ⌬isw2/⌬hda1 double mutant showed higher RNR3 transcription than the ⌬isw2 single deletion (Fig. 2B), we wondered if ⌬hda1 has an additive effect on the chromatin structure change caused by the ⌬isw2 mutation. However, a comparison between ⌬isw2 and ⌬isw2/⌬hda1 strains showed no detectable difference in the digestion pattern over the TATA box and the transcribed region (Fig. 2C, compare lanes 12 and  13 with 6 and 7). Thus the further increase in transcription caused by deleting HDA1 in the ⌬isw2 background is not accompanied by enhanced disruption in nucleosome positioning.
Role of Mediator Revealed by Altering Chromatin-mediated Repression-Activation of RNR3 correlates with a disruption in nucleosome positioning and an increase in histone acetylation, which were mimicked genetically by deleting ISW2 and HDA1, respectively. However, a significant level of repression of RNR3 and HUG1 remained in either mutant, or when the two mutations were combined. This result suggests that Ssn6-Tup1 represses these genes by one or more mechanisms in addition to nucleosome positioning and histone deacetylation. A possibility is repression through the Mediator. Thus, we examined the involvement of Mediator in the repression of RNR3 and HUG1. Mutants of Mediator components that have been proposed by others to be involved in Ssn6-Tup1 repression were analyzed (29, for review see Ref. 26). As shown in Fig. 3 (A and B), mutation of MED3, SIN4, CSE2, GAL11, NUT1, or SRB7 had little to no effect on the level of repression of RNR3 or HUG1. Only the ⌬srb10 and ⌬srb11 strains showed a reproducible derepression of RNR3 (ϳ2.5-to 3-fold). Thus, even though single Mediator mutants cause derepression of some reporter genes and a few native genes (see below and "Discussion"), most caused no dramatic change in the transcription of the DNA damage-inducible genes.
On the other hand, consistent with their positive role in transcription and the requirement for Mediator as a whole for RNR3 expression (40), certain Mediator mutants caused defects in the induction of RNR3 and HUG1 by MMS (Fig. 3, A  and B, lower panels). The defect was observed in only a subset of Mediator mutants, and interestingly, RNR3 and HUG1 responded differently even though they are both regulated by Crt1-Ssn6-Tup1. For instance, deletion of SRB10, SRB11, or CSE2 had little to no effect on the MMS-induced level of RNR3 mRNA, but a clear effect on HUG1.
The contributions of Mediator in repression could be masked by the chromatin-mediated mechanisms, such as histone deacetylation and nucleosome positioning. To better understand whether Mediator is indeed targeted by Ssn6-Tup1 in the regulation of DNA damage-inducible genes, we combined the Mediator mutations with ⌬hda1 or ⌬isw2 mutations, and examined RNR3 and HUG1 transcription. Transcription of both genes in the ⌬hda1 or ⌬isw2 mutants was enhanced by additional mutations in Mediator components, with the effect of ⌬med3, ⌬sin4, and ⌬nut1 stronger on HUG1 and ⌬srb10 and ⌬srb11 stronger on RNR3 (Fig. 3, A and B, upper panels). In parallel, the ⌬rpd3 mutation was also examined and showed no further derepression in combination with ⌬med3, ⌬srb10, or  (29). The others are in BY4741 or BY4705 background. The signals were quantified and normalized as described in the legend of Fig. 1. C, MNase mapping of RNR3 in ⌬srb10 mutants. Experiments were conducted as described in Fig. 1A and under "Materials and Methods." ⌬sin4 ( Fig. 3A, lanes 6, 9, and 12 versus lanes 4, 7, and 10, respectively) but a reduced level of induction in these strains (lower panel, ϩMMS). This is again consistent with a positive role for Rpd3 in the activation of DNA damage-inducible genes and suggests that the additive effect of the Mediator mutations is specific to the HDA1 mutation. Despite subtle differences in the response of each gene to the mutations described here, these results indicate that the Mediator plays a role in repression, but in most cases the effects are only apparent when chromatin-mediated repression is perturbed by disrupting nucleosome positioning (⌬isw2) or increasing histone H3 acetylation (⌬hda1).
Because deletion of SRB10 had the strongest effect on the expression of RNR3 of all the Mediator mutations examined here, the chromatin structure over RNR3 promoter was mapped in ⌬srb10, ⌬srb10/⌬hda1, and ⌬srb10/⌬isw2 strains. The data shown in Fig. 3C indicate that deleting SRB10, even in combination with ⌬hda1, did not cause a detectable change in nucleosome positioning or chromatin structure at RNR3. More subtle defects or changes in a fraction of cells cannot be ruled out, however, and may be likely. Thus, the increase in RNR3 transcription in ⌬srb10 and ⌬srb10/⌬hda1 strains was not from extensive defects in nucleosome positioning. Moreover, the digestion pattern of the ⌬srb10/⌬isw2 double mutant was similar to that of a ⌬isw2 mutant (compare Fig. 3C to Fig.  1A), further suggesting that the repression defect caused by ⌬srb10 is independent of changes in chromatin structure.
ISW2, Hda1, and Mediator Collaborate in Repression of the DNA Damage-inducible Genes-All of the double mutants examined thus far displayed levels of derepression significantly lower than that observed in MMS-treated cells or in a ⌬tup1 mutant. Next, we constructed strains with all three classes of mutations and analyzed the level of derepression of RNR3 and HUG1. We focused on mutations that had the strongest effect on transcription as single or double mutants. As Fig. 4 shows, consistent with experiments described above, derepression was slight or undetectable in the single mutants and the double ⌬isw2/⌬hda1 mutant had higher transcription than the single mutants. Two of the three triple mutants displayed an elevated level of HUG1 RNA compared with the ⌬isw2/⌬hda1 double deletion, which is even higher than a ⌬tup1 mutant and close to that observed in MMS-treated wild type cells (compare lanes 8 and 9 with 12 and 13). The noted exception is the ⌬isw2/⌬hda1/ ⌬srb10 mutant, which expressed HUG1 at a level no different from the double ⌬isw2/⌬hda1 mutant. For RNR3, deleting selected Mediator components in the ⌬isw2/⌬hda1 mutant background reproducibly increased expression (compare lanes  8 -10 with 7), but the level of derepression in the triple mutants was still a fraction of that in a ⌬tup1 mutant (ϳ20 -30%). Given the significant level of repression of RNR3 remaining in the triple ⌬isw2/⌬hda1/Mediator mutants, Tup1 may also repress through additional pathways or gene products. Alternatively, because we know that Mediator is important for the activation of RNR3, the attenuated level of derepression levels in the triple mutant may result from a balance between the repressive affects of the Mediator and its requirement for activation. For instance, it has been proposed that Srb7 plays a dual role in the repression and activation of the GAL genes (29).
Because derepression can be caused by a loss of Tup1 binding to promoters, we used the ChIP assay to verify that Tup1 remained at the promoters of RNR3 and HUG1 in the mutants. As reported previously (14), the cross-linking of Tup1 to RNR3 increased significantly in strains containing the ⌬isw2 mutation (Fig. 5A, left, bar 2). This was observed in all mutants containing a deletion of ISW2, such as ⌬isw2/hda1 and the triple mutants (Fig. 5A, left, bars 7-10). Surprisingly, deleting ISW2 did not increase the cross-linking of Tup1 to the HUG1 promoter (Fig. 5A, right, bar 2), or the RNR2 gene (data not shown), which are also regulated by Crt1-Ssn6-Tup1. Thus, the increased cross-linking of Tup1 to RNR3 in ⌬isw2 mutants is gene-specific. What accounts for the difference is not clear, although the different promoter structures at each locus could be a reason. The excessive recruitment of Tup1 might also explain why RNR3 was only partially derepressed in the triple mutants, whereas derepression of HUG1 was comparable to that of ⌬tup1 or DNA damage induction. The promoter-specific sensitivity of Tup1 cross-linking was not unique to ISW2 deletion. Deleting SRB10 reduced the cross-linking of Tup1 to RNR3 somewhat (about 30%), but was not obviously affected at HUG1 or RNR2 (Fig. 5A; and data not shown) (19). The reduction in Tup1 cross-linking to RNR3 in the ⌬srb10 single mutant was slight but clearer in the ⌬isw2 background where Tup1 cross-linking is elevated in the first place (Fig. 5A, left, bar 10). It needs to be pointed out that, in the ⌬isw2/hda1/srb10 mutation, which caused the strongest RNR3 transcription, Tup1 cross-linking is 2-fold above that in the wild type strain. Furthermore, in the other two Mediator mutants, ⌬med3 and ⌬sin4 did not affect Tup1 cross-linking to RNR3 or HUG1 (Fig.  5A), although they enhanced the transcription when combined with ⌬isw2 or ⌬hda1 mutations. The cross-linking data suggest that the increased transcription in the mutants was not due to the loss of Tup1 at the promoters, but rather resulted from compromised repression functions of the corepressor. Importantly, the presence of Tup1 at the promoter in the various mutants argues that the derepression observed in these mutants is not caused by indirect effects on the promoter, such as inducing a DNA damage signal, because Tup1 is released under such conditions ( Figs. 2A and 5A) (14).
We next examined if the increase in RNR3 and HUG1 expression correlated with increased TBP recruitment. With few exceptions, the trend in the increase in gene expression in the  1-12) and MMS-treated (lanes [13][14][15][16][17][18][19][20][21][22][23][24] cells. ScR1 is a loading control. Quantification is shown below each blot and is expressed relative to untreated wild type cells on the left (lane 1), which was arbitrarily set to 1.0. mutants correlated with increased cross-linking of TBP to the promoters. The most striking example of this was observed at HUG1, which showed the strongest derepression in the mutants. For instance, the ⌬hda1/⌬isw2/⌬med3 mutant displayed the highest level of transcription, equal to that of MMS-treated cells, and had the strongest increase in TBP cross-linking (Fig.  5, B and C, right panel, compare bars 9 and 12). Likewise, derepression of RNR3 was partial in the triple mutants, and the level of TBP cross-linking to RNR3 in these strains was not as high as in fully induced cells (ϩMMS). Surprisingly, we observed an increase in TBP cross-linking to RNR3 and HUG1 in the ⌬sin4 mutant, even though no obvious derepression was observed. The cause of this is unknown, but this observation is consistent with genetic evidence that implies that SIN4 may negatively regulate the binding of TBP to the HO promoter (47,48). With few exceptions the increase in TBP cross-linking correlates with the mRNA levels detected by Northern blot, both of which occur without a significant decrease in Tup1 cross-linking to the promoters. These results imply that Tup1 is not sufficient to block TBP recruitment, but it requires ISW2, HDA1, and Mediator to do so. Furthermore, because TBP occupancy largely mirrored that of transcription, it is unlikely that these mutants raised the level of mRNA by only affecting RNA stability or transcription elongation.  For RNR3 (left), primer pairs corresponding to the URS (Ϫ448 to Ϫ236) and promoter (Ϫ179 to ϩ8) were used to amplify DNA cross-linked to Tup1 and TBP, respectively. For HUG1 (right), cross-linking of Tup1 was detected using primers flanking the DREs (Ϫ308 to Ϫ15) and TBP cross-linking was analyzed using primers flanking the TATA box (Ϫ141 to Ϫ15). C, bar graph quantifying the levels of transcription of RNR3 and HUG1 from the blot presented in Fig. 4.

Redundant Repression Mechanisms Are Also Observed at
Other Ssn6-Tup1-regulated Genes-Ssn6-Tup1 regulates genes controlled by multiple cellular pathways. To address if the redundancy of repression mechanisms is unique to DNA damage-inducible genes we examined the expression of additional Ssn6-Tup1 target genes. Similar to RNR3, nucleosomes are positioned at the anoxia gene ANB1 in a Tup1-dependent manner (11), and deletion of ISW2 disrupts its chromatin structure without affecting Tup1 cross-linking to the promoter (Ref. 14 and data not shown). We examined ANB1 mRNA levels in the panel of mutants and observed a similar derepression pattern to that of the DNA damage-inducible genes. The expression of ANB1 was not significantly affected by any single mutation of ⌬isw2, ⌬hda1, ⌬sin4, ⌬med3, ⌬srb10, ⌬cse2, and ⌬rpd3 (Fig. 6,  A and B). However, derepression was significant in the ⌬hda1/ ⌬isw2, ⌬hda1/⌬srb10, and the ⌬hda1/⌬isw2/mediator triple mutants, where a level of 40 -55% of that observed in a ⌬tup1 mutant was observed (Fig. 6, A and B). Thus, ANB1 is also repressed via multiple mechanisms/pathways.
In contrast to RNR3, HUG1, and ANB1, transcription of the flocculation gene FLO1 was strongly affected by single ⌬sin4 and ⌬srb10 mutations and was moderately affected in the ⌬cse2 mutant. Therefore, Mediator plays a more prominent role in the repression of FLO1. Derepression of FLO1 in the ⌬srb10 single mutant was almost 50% of that caused by TUP1 deletion (Fig. 6, A and B). Even though FLO1 was not as strongly derepressed in the ⌬isw2 or ⌬isw2/⌬hda1 mutants as in some Mediator mutants, the involvement of ISW2 and HDA1 was revealed, however, when combined with the Mediator mutations. The ⌬hda1 mutation strongly synergized with Mediator mutations, and the level of FLO1 message in the double Mediator/ ⌬hda1 mutants was increased to a level almost equal to that of a ⌬tup1 strain (Fig. 6B, compare lanes 6, 9, 4, and 7). Moreover, derepression was significantly enhanced when ⌬isw2 was combined with the Mediator mutation ⌬cse2 (Fig. 6B, lane 8).
Strikingly, the glucose repressed SUC2 gene was not reproducibly derepressed by any of the mutations tested ( Fig. 6A; and data not shown). On the other hand, ⌬isw2 caused a  Fig. 4. The signals were quantified and normalized as described in the legend to Fig. 1. B, Tr1 (TIF51A) RNA cross-hybridizes with the ANB1 probe and is not directly regulated by Ssn6-Tup1 (11,31). decrease in the SUC2 basal expression (Fig. 6A, lane 2), suggesting that ISW2 might play a positive role in SUC2 transcription. The same observation was reported previously, and it was proposed that the loss of ISW2 remodeling activity might result in more repressor binding sites exposed (46). Thus, the mutant combinations described here cannot derepress all Tup1-regulated genes. It is unclear if Tup1 utilizes a different compellation of factors to repress SUC2, or if SUC2 requires an additional activation step that is strongly blocked by Tup1. For instance, it was proposed that that Tup1 blocks the activator of mating type-specific genes, Mcm1, independently of its affects on chromatin (23). Collectively, the data presented above provides evidence that Tup1 represses endogenous genes by both affecting chromatin structure and interfering with mediator. In the chromatin-dependent repression pathway, both nucleosome positioning and histone deacetylation mechanisms can be utilized. Importantly, in some cases, the effects of the Mediator mutations were only obvious after weakening contributions from chromatin structure by deleting ISW2 or HDA1.

DISCUSSION
Redundancy in Ssn6-Tup1-mediated Repression-The debate over the function of the Ssn6-Tup1 complex has been ongoing for more than a decade. Multiple mechanisms/pathways have been proposed, each supported by experimental evidence generated from different reporter gene systems or target genes. Although it is generally accepted that Tup1 can repress through at least three separate pathways, Mediator interference, HDAC recruitment, and nucleosome positioning, it was not clear if each pathway was gene-specific or if multiple pathways are used at all genes in a redundant fashion. A clear analysis of the contributions of each repression pathway at a single locus has not been described. Using DNA damage-inducible genes as our primary model, we observe that progressively inactivating each of the three pathways genetically led to a corresponding increase in expression of RNR3 and HUG1. The most striking example is HUG1, where certain triple mutants displayed expression levels similar to MMS-treated cells or a ⌬tup1 mutant. Our results indicate that, not only are multiple mechanisms utilized by Ssn6-Tup1 for repression, but also they are utilized redundantly at multiple genes.
Another unanswered question is whether or not nucleosome positioning accounts for the ability of Tup1 to repress transcription or if it is a consequence of repression. A challenge to addressing this issue was separating the nucleosome positioning activity of Tup1 from the other repression mechanisms. Attempting to disrupt nucleosome positioning by mutating histone tails or deleting multiple HDACs simultaneously affects all Tup1 mechanisms, because these strategies lead to the loss of Tup1 binding to promoters (19). The use of the ⌬isw2 mutation has allowed us to analyze Tup1-mediated repression of the DNA damage-inducible RNR3 and HUG1 genes in the context of a disrupted chromatin structure, without causing the release of Tup1. We demonstrate that under this condition (⌬isw2) Tup1 represses transcription and TBP recruitment even though nucleosome positioning is disrupted over the promoter. Similar observations were made at haploid-specific genes and ANB1 (11,16), but it was unclear from those studies how Tup1 represses transcription in the absence of nucleosome positioning. The mutational analysis presented here argues that Tup1 continues to repress transcription in a disrupted nucleosomal context by recruiting HDACs to maintain low levels of histone acetylation and by interfering with Mediator function.
Ssn6-Tup1 has been proposed to block TBP binding and preinitiation complex (PIC) assembly based on the observation that Ssn6-Tup1 repressed promoters are devoid of TBP and PIC components (30,49). But it was not clear if Tup1 blocks PIC formation by direct interference or indirectly through its affects on chromatin and Mediator, or both. It is tempting to speculate that the nucleosome positioning ability of Tup1 is solely responsible for blocking TFIID recruitment, because a nucleosome prevents TBP from binding to TATA elements in vitro (50). However, Tup1 can repress transcription and block TBP binding when nucleosomes are disrupted over the promoter. The nucleosome over the TATA box of RNR3 is disrupted in the ⌬isw2 mutant, yet transcription is repressed and TBP does not cross-link (Figs. 1 and 5B) (14). Moreover, Tup1 can repress the haploid-specific genes under control of the a1-␣2 repressor without positioning nucleosomes over a CYC1-LacZ reporter construct (16), and histone tail mutations disrupt chromatin at ANB1 without increasing transcription or TBP cross-linking (11,32). Finally, moving the TATA box of STE6 into linker DNA cannot activate the gene (8). Although all of these studies suggest that Tup1 can repress transcription without positioning a nucleosome over the TATA box, it was unclear how it does so. We find that progressively inactivating different Tup1 repression mechanisms lead to a correlative increase in TBP recruitment, even though Tup1 remained bound to the promoter under these conditions. This was particularly striking at HUG1. This suggests that Tup1 indirectly inhibits TBP recruitment by its ability to recruit HDACs, position nucleosomes over promoter, and interfere with Mediator. A role for histone deacetylation in preventing TBP recruitment has been proposed. Artificially recruiting an HDAC to a reporter gene reduces TBP recruitment (51); however, Tup1mediated repression or nucleosome positioning was not addressed in this study.
Tup1 Utilizes Mediator in the Repression of Chromosomal Genes-The ability of Ssn6-Tup1 to directly interfere with the basal transcription machinery was implied by the repression of a presumably nucleosome-free reporter plasmid in vitro (24,25). More evidence came from genetic approaches that showed mutations in Mediator subunits derepress Ssn6-Tup1-regulated reporter systems (for review see Refs. 2 and 26) and biochemical studies demonstrating direct interactions between Tup1 and Srb7, Med3, and Srb10 (27)(28)(29)(30). However, a genetic analysis of Mediator mutations revealed that mutating multiple Mediator components, even in combination, did not cause significant derepression of chromosomal Ssn6-Tup1-regulated genes (31). Most cases documenting derepression by Mediator mutations used promoter-lacZ reporter constructs, with few examples at natural genes (28,29,52). Thus the importance of Mediator in Ssn6-Tup1 repression at endogenous genes has not been resolved, and the artificial nature of the lacZ reporter systems could account for their sensitivity to Mediator mutations. For example, although the single Mediator mutations ⌬sin4, ⌬nut1, and ⌬nut2 caused derepression of the HO-lacZ and PHO5-lacZ reporters, transcription at the native HO and PHO5 genes were not affected (53).
Our studies may explain why Mediator mutations have stronger effects on lacZ reporter constructs, but little to no effect on natural genes. We find that weakening the contributions of chromatin structure in repression renders native Ssn6-Tup1 target genes more sensitive to Mediator mutations. It is likely that on many promoters the chromatin-mediated repression pathway is sufficient to block transcription in the Mediator mutants. The chromatin structure of some artificial reporter genes may exist in a semi-permissive state, allowing the effects of Mediator mutations to be seen. Thus, we argue that Mediator components do play a role in the repression of chromosomal Ssn6-Tup1-regulated genes, but the effects can be masked by chromatin-dependent, or other, repression mechanisms. What is the significance of Mediator repression at DNA damage-inducible genes or ANB1 where the chromatin mechanism is dominant? The chromatin pathway may be dominant in maintaining repression, but perhaps the first step in establishing repression is the inhibition of Mediator. Tup1 can attenuate transcription and PIC formation via contacts with Mediator subunits to allow for the chromatin pathway to be established and provide a permanent, and stable form of repression.
Our results are seemingly in some disagreement with studies showing that ⌬srb10, ⌬srb11, or srb7 mutations in combination with either a deletion of the H4 or H3 tail does not cause derepression of ANB1 (11,31,32). The conclusion from those studies was that disrupting chromatin structure and mutating Mediator components do not cause derepression of ANB1. However, mutating a single histone tail, as was done in the mentioned studies, may not be as disruptive to chromatin structure as deleting ISW2. Alternatively, deleting histone tails can affect activation pathways (54,55), and the derepressive effects of these mutations could be overshadowed by the positive role histone tails in activation. Our strategy, disrupting positioning using the ⌬isw2 mutation, leaves the histone tails intact and should not alter other functions of the tails. Therefore, a difference in the approaches used to disrupt chromatin structure may likely account for the discrepancy.
Gene-specific Responses to Tup1 Repression Mechanisms-Tup1-regulated genes show unequal sensitivities to different mutations and inactivating different repression pathways. RNR3, HUG1, and ANB1 appear to be more sensitive to disrupting the chromatin-dependent pathways, and most of the derepression is observed in double ⌬isw2/⌬hda1 cells with mediator mutations playing a lesser role. In contrast, FLO1 was more sensitive to Mediator mutations. Obvious derepression of FLO1 resulted from single ⌬cse2, ⌬sin4, and ⌬srb10 mutations, and Mediator and HDA1 mutations synergized strongly at this gene. Furthermore, no significant derepression was observed at SUC2 in any mutant combination. Why do genes show different sensitivities to the mutants? A possibility is that, at a certain locus, some mechanisms play a less important role in repression. For instance, the local chromatin environment of a gene can make it particularly sensitive to some of the repression mechanisms of Ssn6-Tup1 more than others, such that its expression will be insensitive to mutations disrupting the underutilized pathway(s). For example, genes whose repression relies less on nucleosome positioning will be relatively insensitive to disruptions in positioning but more sensitive to HDACs or Mediator mutations. Locus-dependent sensitivity to Mediator mutations has been described. For example, the transcription of PHO5 was not affected by a sin4 mutation at its natural locus, but relocating it to the URA3 locus rendered it sensitive (56). In addition, given the fact that the type, number, and locations of the sequence-specific repressors vary among individual promoters, the Ssn6-Tup1 complexes recruited by these repressors might adopt confirmations more favorable for utilizing some repression mechanisms, but not the others. The distinct sensitivities of the Ssn6-Tup1-regulated genes to the different mutations described here implies that different groups of genes have evolved distinct, but overlapping strategies to use Ssn6-Tup1 for repression, which may be essential for it to function as a "global" corepressor.