|Home | About | Journals | Submit | Contact Us | Français|
The Srb mediator is an important transcriptional coactivator for Gcn4p in the yeast Saccharomyces cerevisiae. We show that three subunits of the Gal11/tail domain of mediator, Gal11p, Pgd1p, and Med2p, and the head domain subunit Srb2p make overlapping contributions to the interaction of mediator with recombinant Gcn4p in vitro. Each of these proteins, along with the tail subunit Sin4p, also contributes to the recruitment of mediator by Gcn4p to target promoters in vivo. We found that Gal11p, Med2p, and Pgd1p reside in a stable subcomplex in sin4Δ cells that interacts with Gcn4p in vitro and that is recruited independently of the rest of mediator by Gcn4p in vivo. Thus, the Gal11p/Med2p/Pgd1p triad is both necessary for recruitment of intact mediator and appears to be sufficient for recruitment by Gcn4p as a free subcomplex. The med2Δ mutation impairs the recruitment of TATA binding protein (TBP) and RNA polymerase II to the promoter and the induction of transcription at ARG1, demonstrating the importance of the tail domain for activation by Gcn4p in vivo. Even though the Gal11p/Med2p/Pgd1p triad is the only portion of Srb mediator recruited efficiently to the promoter in the sin4Δ strain, this mutant shows high-level TBP recruitment and wild-type transcriptional induction at ARG1. Hence, the Gal11p/Med2p/Pgd1p triad may contribute to TBP recruitment independently of the rest of mediator.
Transcription initiation by RNA polymerase II (Pol II) is dependent on a set of general transcription factors (GTFs), including TATA binding protein (TBP), which recognize the core promoter and facilitate initiation from the correct start site. The stimulation of transcription by activator proteins is dependent on coactivators that serve as adaptors to facilitate recruitment of Pol II and GTFs and also to activate Pol II function. Other coactivators stimulate preinitiation complex (PIC) assembly by remodeling the nucleosome structure of the promoter (21). The Srb mediator is an important coactivator in Saccharomyces cerevisiae, consisting of 20 or more distinct polypeptide subunits, which can be found associated with Pol II in a holoenzyme complex (11, 16, 23, 26). Purified mediator can support stimulation of Pol II function by activators in vitro, and it enhances basal transcription and phosphorylation of the C-terminal domain (CTD) of the largest subunit of Pol II (Rpb1p) by the general factor TFIIH (reviewed in reference 30).
The Srb2p and Srb4p-Srb11p subunits of Srb mediator were identified genetically as suppressors of truncations of the Rpb1p CTD (18). Transcription in cell extracts is highly dependent on Srb4p, Srb2p, and Srb6p (35, 41), and a temperature-sensitive srb4 mutation impairs transcription from most promoters in vivo (42). Thus, these Srb proteins carry out important general functions in transcription initiation (35). A number of mediator subunits were identified genetically by mutations that impair activation or repression of specific genes, implicating Srb mediator in transcriptional regulation in vivo. The remaining mediator subunits were identified biochemically from large-scale purification of a native mediator complex (30).
The mediator can be divided roughly into three modules. The Srb4 module consists of eight subunits (Srb2p, Srb4p to Srb6p, mediator protein 6 [Med6], Med8, Med11, and Rox3p) (15, 17, 23), of which four are essential (30). This module can interact physically with the Rpb1p CTD, TBP, and TFIIB, and it supports basal, but not activated, transcription in vitro (15), although Srb4p itself was identified as a recruitment target for the activator Gal4p (17). It was proposed that the Srb4 module corresponds to the head domain of mediator, as visualized in three-dimensional reconstructions of electron micrographs, which makes several contacts with Pol II in the holoenzyme complex (8). The Med9/Med10 module contains several essential subunits and, together with Rgr1 and Nut1p, has been equated with the middle domain of mediator (8). The Gal11 module, or tail domain, contains Gal11p, Sin4p, Med2p, and Pgd1p/Hrs1p, none of which is essential (30). The tail domain is tethered to the rest of mediator through the Rgr1p CTD (8, 25), and deletion of SIN4 leads to loss of Gal11p, Med2p, and Pgd1p from the mediator during purification. The last three proteins also were missing from mediator complexes purified from strains lacking only Gal11p, Med2p, or Pgd1p (23, 29, 33); however, dissociation of all three subunits was not observed at early steps in the purification (29). Considering that distinct in vivo phenotypes are associated with single deletions of GAL11, MED2, or PGD1 (29, 40), elimination of single subunits of the Gal11p/Med2p/Pgd1p triad probably does not disrupt the entire tail domain in vivo.
There is accumulating evidence that the mediator tail domain is an important target of transcriptional activators. Purified mutant holoenzymes lacking Gal11p, Pgd1p, and Med2p are impaired for transcriptional activation by Gal4p, VP16, and Gcn4p, and they fail to bind these activators in vitro. Moreover, all three activators bind to recombinant Gal11p, and Gcn4p also interacts with recombinant Pgd1p in vitro (23, 33). Deletions of internal Gal11p segments that impair binding to Gal4p in vitro produce commensurate activation defects in vivo, consistent with the idea that Gal11p is a direct target for Gal4p. It is unclear whether this is the case for Gcn4p, however, as deletions of Gal11p, Pgd1p (33), or Med2p (29) had little effect on transcriptional activation by Gcn4p in vivo. More recently, however, we observed reductions in the activation of Gcn4p-dependent reporters or target genes in mutants lacking each of these tail subunits (32, 40). Interestingly, a med10 Ts− mutation had strong effects on activation by Gcn4p but almost no effect on Gal4p function (12). Thus, Gal4p and Gcn4p may require distinct subsets of mediator subunits for optimal activation in vivo (29).
In this study, we investigated the requirements for tail domain subunits in the interaction of Srb mediator with recombinant Gcn4p in vitro and in the recruitment of mediator subunits by Gcn4p to target promoters in vivo. The in vitro experiments indicate that the four tail domain subunits, Gal11p, Pgd1p, Med2p, and Sin4p, along with the head domain subunit Srb2p, all contribute to a stable interaction of mediator with Gcn4p. By chromatin immunoprecipitation (ChIP) experiments, we found that each of these proteins is also required for high-level recruitment of a mediator head domain subunit by Gcn4p in vivo. Interestingly, Gal11p, Med2p, and Pgd1p reside in a stable subcomplex in sin4Δ cells that can bind to Gcn4p in vitro and is recruited by Gcn4p to target genes in vivo independently of the rest of mediator. Finally, we present evidence suggesting that the Gal11p/Med2p/Pgd1p triad can promote TBP recruitment independently of the rest of mediator in sin4Δ cells.
High-copy-number and single-copy plasmids harboring GCN4-HA, pHQ1239, and p2382, respectively, were described previously (40). Yeast strains used in this study are listed in Table Table1.1. Wild-type (WT) and single-deletion mutants from the Saccharomyces Genome Deletion Project (10) were purchased from Research Genetics. The mutant alleles were verified previously by PCR amplification of genomic DNA and by complementation of mutant phenotypes with plasmid-borne WT alleles (40). All double mutants were generated by genetic crosses between the corresponding single mutants. The coding sequences for myc13 were inserted at the 3′ ends of the coding sequences of various chromosomal genes as a myc13-HIS3 cassette, as described previously (40). The presence of the myc-tagged alleles was verified by colony PCR and by Western analysis using antimyc antibodies. All of the tagged strains were tested for resistance to sulfometuron-methyl (SM), a sensitive indicator of the extent of transcriptional activation by Gcn4p (40). We observed no decrease in growth on nutrient complete medium and no increase in sensitivity to SM associated with myc tagging of GAL11, SRB6, or MED2 in strains containing WT alleles of all other mediator subunits and also in the strains harboring deletions of single mediator subunits (data not shown). Thus, it appears that the myc tags on Gal11p, Srb6p, and Med2p have little or no effect on the integrity and function of Srb mediator in vivo. The same was true for the SIN4-myc strains except that the srb2Δ SIN4-myc and srb5Δ SIN4-myc mutants are more sensitive to SM than are the corresponding srb2Δ SIN4 and srb5Δ SIN4 strains (data not shown). Thus, the myc tag on Sin4p may produce a small reduction in coactivator function when combined with elimination of Srb2p or Srb5p from Srb mediator. Deletion of GCN4 in the myc-tagged strains was carried out with plasmid pHQ1240 and verified as described previously (43).
Bacterial extracts containing glutathione S-transferase (GST) proteins and whole-cell extracts (WCEs) from yeast strains grown in yeast extract-peptone-dextrose medium (38) were used for GST pull-down assays as described previously (9) except that yeast cells were broken in a buffer containing 50 mM Tris-HCl (pH 7.5), 50 mM HEPES (pH 7.9), 10 mM MgSO4, 100 mM NH4SO4, 12.5 mM potassium acetate, 5 mM EGTA, 2.5 mM dithiothreitol, 0.01% NP-40, 20% glycerol, 1 μg of pepstatin A/ml, 1 mM AEBSF [4-(2-aminoethyl) benzenesulfonyl fluoride], and 2× Complete protease inhibitor without EDTA (Roche), and the lysates were cleared by centrifuging at 16,000 × g for 15 min and then for 30 min at 4°C in a microcentrifuge. Coimmunoprecipitation assays were conducted using the same yeast WCEs and antimyc antibodies conjugated to agarose (Santa Cruz Biotechnology). The immune complexes were washed and collected as in GST pull-down assays (9). Rabbit polyclonal antisera used in Western analysis were previously described for Srb7p (13), Med1p (1), Srb2p and Srb5p (41), Tra1p (2), Taf9p (9), Snf5p (4), and Med4p and Med8p (28), as was rat polyclonal antiserum for Rgr1p (22). Taf12p antibodies were kindly provided by J. Reese. Mouse monoclonal antibodies against the myc epitope (9E10) were obtained from Roche. Goat polyclonal antiserum for Ada2p was from Santa Cruz Biotechnology. Hemagglutinin (HA)-tagged proteins were detected with anti-HA monoclonal antibodies from Santa Cruz Biotechnology. Total RNA was extracted and subjected to Northern analysis as described previously (31).
ChIP assays were conducted as previously described (40) with the following modifications. Cultures were grown to optical density at 600 nm of 1.0 to 1.5 in synthetic complete medium (SC) lacking uracil (38). The cells were collected by centrifugation, washed with sterile H2O, and resuspended in SC lacking uracil, arginine, isoleucine, and valine and containing 0.5 μg of SM/ml. The SM-treated cells were grown for 3 h before adding formaldehyde. Cross-linked chromatin was sonicated for 12 cycles of 30 s at 4°C with at least 30 s cooling on ice per cycle and immunoprecipitated with antibodies against the myc epitope (Roche), TBP (provided by J. Reese), or Rpb1p (18WG16; Abcam, Inc.) as appropriate. The primers used to amplify the ARG1 upstream activation sequence (UAS), POL1 open reading frame (ORF) (40), SNZ1 UAS (43), and ARG1 TATA element (34) were described previously. The PCR amplification involved denaturation at 94°C for 4 min followed by 27 cycles of 94°C for 30 s, 52°C for 30 s, 65°C for 1 min, and a final extension for 5 min at 65°C.
In previous studies, it was shown that mediator complexes purified from pgd1Δ or gal11Δ single mutants, which lacked Gal11p, Med2p, and Pgd1p, failed to bind recombinant Gcn4p in vitro (23, 29, 33). As indicated above, it seemed possible that the native mediator complexes present in these mutants might not lack all three subunits of the Gal11p/Med2p/Pgd1p triad. Accordingly, we prepared WCEs from mutants with single deletions of these or other mediator subunits and analyzed the binding of the mutant complexes to recombinant GST-Gcn4p. In addition to the mutants lacking a tail subunit (gal11Δ, pgd1Δ, med2Δ, and sin4Δ mutants), we also analyzed three mutants lacking head domain subunits (srb2Δ, srb5Δ, and rox3Δ mutants) which showed defects in activation by Gcn4p in vivo (Gcn− phenotype) and the med1Δ mutant, which showed WT activation by Gcn4p (40). (According to the Saccharomyces Genome Database at http://www.yeastgenome.org, it was reported that ROX3 is essential. The rox3Δ strain employed here, produced by the Saccharomyces Deletion Project , has strong slow-growth and Gcn− phenotypes, which are complemented by an episomal copy of ROX3 .) Mediator binding to GST-Gcn4p was assayed by Western analysis using antibodies against mediator subunits on fractions bound to glutathione affinity resin. In these assays, we employed the WT GST-Gcn4p fusion and a mutant derivative with 10 alanine substitutions in the hydrophobic clusters of the Gcn4p activation domain (10Ala fusion). These mutations were shown previously to impair transcriptional activation by Gcn4p in vivo (9, 14, 32).
As expected, all mediator subunits assayed in the WT extract showed substantial binding to WT GST-Gcn4p, greatly reduced binding to the 10Ala mutant protein, and no detectable binding to GST alone (Fig. (Fig.1A,1A, WT lanes). The binding of most mediator subunits (considered as a fraction of the input amounts in the extracts) in the gal11Δ, pgd1Δ, med2Δ, rox3Δ, and med1Δ extracts was similar to that seen in the WT extract; however, a substantial reduction in binding occurred in the sin4Δ extract (Fig. (Fig.1A).1A). In the srb2Δ and srb5Δ extracts, only the binding of Srb5p and Srb2p, respectively, was seriously impaired, suggesting that these two subunits are interdependent for stable association with mediator. This is consistent with the fact that Srb2p and Srb5p interact directly with one another in vitro (17). Thus, except for deletion of Sin4p, the deletions of single mediator subunits had limited effects on mediator binding to GST-Gcn4p in vitro.
Consistent with the results from the sin4Δ extract, we observed essentially no mediator binding to GST-Gcn4p in the three extracts prepared from double mutants lacking Sin4p and one of the other tail subunits, Gal11p, Pgd1p, or Med2p (Fig. (Fig.1B).1B). Importantly, a strong reduction in binding also was observed for the gal11Δ med2Δ double mutant (Fig. (Fig.1B,1B, lanes 17 to 20), and only slightly less severe binding defects occurred with the double mutants lacking Pgd1p and Med2p or Pgd1p and Gal11p (lanes 5 to 8 and 13 to 16). Note that the binding of SAGA and SWI/SNF subunits to GST-Gcn4p occurred at nearly WT levels in all of the double mutants, indicating that the binding defects were specific for mediator subunits. (Although Taf9p and Taf12p are shared by SAGA and TFIID, we showed previously that TFIID does not bind to GST-Gcn4p in these assays (32). It is possible, however, that the binding of Tra1p to GST-Gcn4p reflects interaction of the NuA4 complex, rather than SAGA, with GST-Gcn4p.)
The findings in Fig. 1A and B can be explained by proposing that the sin4Δ mutation leads to dissociation of the other three tail subunits from mediator (29), impairing all interactions between Gcn4p and the tail domain. By contrast, the gal11Δ, med2Δ, and pgd1Δ single mutations would leave the other mediator tail subunits intact, and the mutant tail domains lacking only one of these subunits would still interact effectively with GST-Gcn4p. Because the interaction with GST-Gcn4p is lost in all three double mutants lacking two subunits of the Gal11p/Med2p/Pgd1p triad, it is possible that Gcn4p must contact any two of these subunits simultaneously for stable interaction with mediator in vitro. Alternatively, Gcn4p might contact only one of the three subunits, but this interaction requires an indirect contribution from one of the other two subunits of the triad.
Deleting the head domain subunit Srb2p together with the tail subunit Pgd1p, Gal11p, or Sin4p also greatly impaired the binding of mediator to GST-Gcn4p (Fig. (Fig.1C1C and data not shown), comparable to what was observed in the double mutants lacking two tail subunits (Fig. (Fig.1B).1B). By contrast, a double deletion of head domain subunits Srb2p and Srb5p had little effect on the binding of the other mediator subunits to GST-Gcn4p (Fig. (Fig.1C,1C, lanes 5 to 8). These results suggest that the binding of Srb mediator to Gcn4p involves overlapping contributions from the four tail subunits and the head subunit Srb2p.
It was important to verify that the impaired binding to GST-Gcn4p that occurred with the mutants lacking two mediator subunits did not result from disruption of the remainder of mediator. To this end, we tagged the C terminus of Sin4p with 13 myc epitopes in the mutants described above containing SIN4. Extracts from the resulting SIN4-myc strains were immunoprecipitated with myc antibodies, and the immune complexes were probed for various mediator subunits. All six mediator subunits we tested from the head or middle domains were immunoprecipitated with myc antibodies from the SIN4-myc strain, but not from the untagged SIN4 strain (Fig. (Fig.2A).2A). Importantly, we observed WT association of all head and middle domain subunits with myc-Sin4p in the three double mutants lacking two tail subunits (gal11Δ med2Δ, pgd1Δ med2Δ, and pgd1Δ gal11D mutants) and in the srb2Δ gal11Δ and srb2Δ pgd1Δ double mutants (Fig. (Fig.2B).2B). Hence, we conclude that the strong defects in binding to GST-Gcn4p observed for these five double mutants (Fig. (Fig.1)1) probably do not result from disruption of the mediator head or middle domains.
Because an otherwise intact mediator complex lacking the tail domain can be purified from a sin4Δ strain (8, 29), we expected to find that head and middle domain subunits would also remain associated with one another in the sin4Δ extract. To verify this prediction, we myc tagged the head subunit Srb6p in SIN4 and sin4Δ strains and conducted coimmunoprecipitation analysis as described above. As expected, the middle domain subunits Med1p, Rgr1p, and Srb7p and the head subunit Srb2p coimmunoprecipitated with myc-Srb6 in both the SIN4 and sin4Δ strains and also in srb2Δ cells (Fig. (Fig.2C).2C). We additionally myc tagged Gal11p to confirm that deleting SIN4 would lead to dissociation of other tail subunits from the head and middle portions of mediator (29). As expected, myc-Gal11p was dissociated from the head and middle domain subunits in the sin4Δ extract (Fig. (Fig.2D,2D, cf. lanes 4 to 6 and 7 to 9). We also found that myc-Gal11p failed to coimmunoprecipitate with middle and head subunits in extracts lacking the tail subunits Pgd1p or Med2p, but not in the extract lacking head subunit Srb2p (Fig. (Fig.2D,2D, lanes 10 to 19 versus 4 to 6). Results essentially identical to those shown for the GAL11-myc strains shown in Fig. Fig.2D2D were obtained with a panel of strains harboring myc-tagged Med2p (Fig. (Fig.2E2E).
The dissociation of myc-Gal11p from the rest of mediator in the med2Δ and pgd1Δ extracts seems at odds with our finding that binding of mediator to GST-Gcn4p was more severely impaired in extracts of the gal11Δ med2Δ and gal11Δ pgd1Δ double mutants than in extracts of the med2Δ and pgd1Δ single mutants (Fig. 1A and B), which implied that Gal11p was still associated with mediator in the med2Δ and pgd1Δ single mutants. To explain this discrepancy, we propose that adding the myc13 tag to Gal11p weakens its association with other tail subunits and that combining this alteration with the elimination of Pgd1p or Med2p from the tail domain leads to dissociation of myc-tagged Gal11p from the rest of mediator. An analogous explanation would apply to the failure of myc-tagged Med2p to coimmunoprecipitate with other mediator subunits from gal11Δ or pgd1Δ extracts (Fig. (Fig.2E)2E) even though the GST pull-down data in Fig. Fig.11 strongly suggested that Med2p still resides in the mediator complexes found in gal11Δ or pgd1Δ extracts.
It was of interest to ascertain whether myc-Gal11p remains associated with Med2p and Pgd1p in a stable subcomplex in sin4Δ cells. To determine this, we tagged Med2p and Pgd1p with three tandem HA epitopes in the sin4Δ GAL11-myc and SIN4 GAL11-myc strains and conducted coimmunoprecipitation assays. The mediator head and middle subunits coimmunoprecipitated with myc-Ga111p from the SIN4 extract, but not from the sin4Δ extract (Fig. (Fig.3A,3A, lanes 10 to 15 versus 1 to 9), in agreement with the results described above (Fig. (Fig.2D).2D). Importantly, HA-Med2p and HA-Pgd1p coimmunoprecipitated with myc-Gal11p at high levels from both sin4Δ and SIN4 extracts (Fig. (Fig.3A,3A, cf. lanes 4 to 9 and 10 to 15). Similarly, when HA antibodies were used, only myc-Gal11p coimmunoprecipitated with HA-Pgd1p from the sin4Δ extract, whereas all of the mediator subunits coimmunoprecipitated with HA-Pgd1p from the SIN4 extract (Fig. (Fig.3A,3A, lanes 16 to 18 versus 19 to 21). (The apparent increase in association of HA-Pgd1p and HA-Med2p with myc-Gal11p in the sin4Δ strain versus the SIN4 strain, evident in lanes 5, 8, 11, and 14 of Fig. Fig.3A,3A, was not observed consistently, as seen by comparing lanes 17 and 20 of Fig. Fig.3A.)3A.) Based on the results shown in Fig. Fig.3A,3A, we conclude that myc-Gal11p forms a stable subcomplex with HA-Med2p and HA-Pgd1p in vivo in the absence of Sin4p.
We next addressed whether the myc-Gal11p/HA-Med2p/HA-Pgd1p subcomplex can interact with Gcn4p in vitro. As shown in Fig. Fig.3B,3B, myc-Gal11p, HA-Med2p, and HA-Pgd1p all showed strong binding to GST-Gcn4p in sin4Δ extracts at levels comparable to that seen in the corresponding SIN4 extracts. As expected, the binding of five subunits from the head and middle domains was abolished in the sin4Δ extracts. These results show that the myc-Gal11p/Pgd1p/Med2p subcomplex in sin4Δ extracts can interact specifically with GST-Gcn4p. It could be argued that Gal11p, Med2p, and Pgd1p bind to GST-Gcn4p in the sin4Δ extract as components of the Paf1 complex, as Gal11p has been found associated with this alternative mediator complex (39). However, our previous finding that myc-Paf1p in cell extracts does not bind specifically to GST-Gcn4p (40) is at odds with this possibility. Moreover, we found here that myc-Paf1p does not coimmunoprecipitate with HA-Med2p or HA-Pgd1p (data not shown).
We next asked whether the mediator tail subunits and Srb2p are required for recruitment of mediator by Gcn4p to target promoters in vivo. By ChIP analysis, we showed previously that Gcn4p can recruit myc-tagged mediator subunits Srb6p, Gal11p, and Sin4p to the ARG1 UAS when Gcn4p synthesis is induced by starvation for isoleucine and valine with the inhibitor SM (40). Hence, we set out to determine the effects of deleting single subunits of Srb mediator on the recruitment of myc-Srb6p to ARG1 in SM-treated cells.
We first conducted ChIP analysis on six pairs of yeast strains, all harboring the SRB6-myc allele, containing either all other WT mediator genes or a single deletion of SIN4, GAL11, PGD1, MED2, or SRB2. One member of each pair contained gcn4Δ while the other carried episomal GCN4, and all strains were grown in the presence of SM to induce Gcn4p synthesis in the GCN4 strains. In agreement with previous results, the ARG1 UAS was immunoprecipitated with myc-Srb6p at a level approximately fivefold higher in the GCN4 strains than in the gcn4Δ strains containing all WT mediator subunits (Fig. 4A and B, WT lanes). By contrast, we observed low-level recruitment of myc-Srb6p to ARG1 in all five GCN4 strains lacking single mediator subunits (Fig. 4A and B). The amount of Gcn4p-dependent binding of myc-Srb6p to the UAS in each mutant was expressed as a percentage of that measured in WT to yield the values listed below the histogram in Fig. Fig.4B.4B. The results show that all five subunit deletions reduced Gcn4p-dependent recruitment of myc-Srb6p at ARG1 to ≤10% of the WT level. Deleting each tail domain subunit, or Srb2p, also had a marked effect on recruitment of myc-Srb6p by Gcn4p to the SNZ1 UAS (Fig. (Fig.4C4C).
Western analysis of myc-Srb6p revealed no significant differences in myc-Srb6p levels between the WT and mutant strains analyzed in Fig. Fig.44 (data not shown). Moreover, myc-Srb6p coimmunoprecipitated at similar levels with head and middle domain subunits from extracts of these WT and mutant strains (Fig. (Fig.2C2C and data not shown). Thus, the reductions in myc-Srb6p recruitment in these mutants do not result from disruption of mediator or decreased steady-state levels of myc-Srb6p.
It was also important to determine whether the binding of Gcn4p itself to the target genes was affected by deletions of mediator subunits. To address this possibility, we introduced a plasmid-borne GCN4-myc allele into the gcn4Δ strains containing mediator subunit deletions (or the otherwise WT gcn4Δ strain) and conducted ChIP assays to measure myc-Gcn4p binding to the ARG1 promoter. All of the mutants displayed myc-Gcn4p binding to the ARG1 UAS at levels greater than or equal to that seen in the corresponding WT strain (34) (data not shown). Thus, we conclude that the mediator tail subunits and Srb2p are required for recruitment of the Srb mediator head domain by promoter-bound Gcn4p.
Having shown that the Gal11p/Pgd1p/Med2p subcomplex is dissociated from the rest of mediator in the sin4Δ extract and that it can interact specifically with GST-Gcn4p in vitro, we conducted ChIP assays to determine whether Gcn4p can recruit this subcomplex to target promoters in sin4Δ cells. We began by analyzing strains containing GAL11-myc (Fig. (Fig.4D).4D). Remarkably, we saw no reduction in recruitment of myc-Gal11p to ARG1 in the sin4Δ strain and only a modest reduction in binding (to ~80% of WT) in the pgd1Δ and srb2Δ mutants (Fig. (Fig.4E).4E). Only the med2Δ mutation strongly impaired recruitment of myc-Gal11p to ARG1. Similar results were obtained at SNZ1 (Fig. (Fig.4F)4F) except that sin4Δ, pgd1Δ, and srb2Δ had relatively stronger effects on myc-Gal11p binding at this gene. Nevertheless, these three mutations all produced much smaller reductions in recruitment of myc-Gal11p than of myc-Srb6p at SNZ1 (cf. Fig. 4E and F to B and C). Deletion of MED2 greatly impaired myc-Gal11p binding at SNZ1, as observed at ARG1. We verified by Western analysis that the steady-state levels of myc-Gal11p in WT and med2Δ extracts are similar (data not shown). The findings in Fig. 4F and G imply that high-level recruitment of myc-Gal11p by Gcn4p is relatively independent of Sin4p and other mediator subunits and may depend only on Med2p.
To obtain additional evidence that the Gal11p/Pgd1p/Med2p triad can be recruited independently of the rest of mediator, we conducted ChIP analysis of strains containing myc-tagged Med2p. Our coimmunoprecipitation analysis had indicated that myc-Med2p was dissociated from the rest of mediator by the sin4Δ, gal11Δ, or pgd1Δ mutation (Fig. (Fig.2E).2E). In ChIP analysis, we observed strong Gcn4p-dependent recruitment of myc-Med2p to ARG1 and SNZ1 in the sin4Δ strain but weak recruitment of myc-Med2p in the gal11Δ and pgd1Δ mutants (Fig. 5A and B). The identical conclusion emerged from ChIP analysis of strains containing GCN4 on a single-copy versus multicopy plasmid, the only difference being that Gcn4p-dependent recruitment was uniformly lower in the strains harboring single-copy GCN4 (Fig. (Fig.5A).5A). (Using a GCN4-myc allele, we showed that the binding of myc-Gcn4p at ARG1 is approximately twofold higher when myc-Gcn4p is expressed from a multicopy versus single-copy plasmid .) The data in Fig. 5A and B indicate that Med2p can be recruited efficiently in the sin4Δ mutant as a component of the isolated tail domain, dependent on Gal11p and Pgd1p. The fact that recruitment of myc-Med2p shows a greater requirement for Pgd1p than does recruitment of myc-Gal11p may be explained by our observation that myc-Med2p levels are generally lower in the pgd1Δ strain than in WT and gal11Δ strains (Fig. (Fig.2E2E and data not shown).
The results above showed that recruitment of the tail subunits myc-Gal11p and myc-Med2p exhibits little requirement for Sin4p whereas recruitment of head subunit Srb6p is strongly dependent on Sin4p. We extended this distinction to a second head subunit by conducting ChIP assays on SIN4 and sin4Δ strains containing myc-tagged Srb5p. As shown in Fig. Fig.5C,5C, recruitment of myc-Srb5p to ARG1 was greatly impaired by the sin4Δ mutation, just as we observed for myc-Srb6p (Fig. (Fig.4B).4B). This supports the idea that recruitment of the head domain is strongly dependent on its association with the tail domain of mediator.
The coimmunoprecipitation analysis of the SIN4-myc strains described above had indicated that myc-Sin4p remained associated with the rest of mediator in the presence of the med2Δ, gal11Δ, pgd1Δ, or srb2Δ mutation (Fig. (Fig.2B).2B). Hence, we expected to find that recruitment of myc-Sin4p by Gcn4p in vivo would have the same subunit requirements observed for myc-Srb6p. In accordance with this prediction, recruitment of myc-Sin4p to ARG1 was strongly impaired by the med2Δ, gal11Δ, and pgd1Δ mutations, whether GCN4 was present on single- or multicopy plasmids (Fig. (Fig.5D).5D). Interestingly, Srb2p appears to be dispensable for myc-Sin4p binding at ARG1 but is required for high-level myc-Sin4p recruitment at SNZ1 (Fig. (Fig.5E).5E). The last findings suggest that Srb2p may be less critical than the tail subunits for Srb mediator recruitment by Gcn4p. In addition, there appears to be more-stringent requirements for mediator recruitment at SNZ1 than at ARG1, similar to what we observed previously for recruitment of SWI/SNF by Gcn4p at these two genes (43).
Having concluded that deletions of each of the four tail domain subunits impaired the recruitment of myc-Srb6p, we compared the effects of these deletions on transcriptional activation by Gcn4p. We showed previously that the gal11Δ, med2Δ, and pgd1Δ strains show increased sensitivity to SM, consistent with impaired transcriptional activation of the ILV2 gene by Gcn4p. They also were defective for induction of a lacZ reporter gene containing tandem Gcn4p binding sites upstream of the CYC1 promoter, showing induction levels of only ~30% of WT. However, the sin4Δ mutant showed nearly WT resistance to SM and was not defective for induction of the Gcn4p-dependent lacZ reporter (40).
In the present study, we conducted Northern analysis of ARG1 and SNZ1 mRNAs using the same inducing conditions employed for ChIP assays (Fig. (Fig.6A).6A). The levels of transcripts were quantified from two or more independent cultures and the mean values were plotted as a fraction of the WT value in the histogram shown in Fig. Fig.6A6A (see legend for details). Among the single mutants, only the med2Δ strain showed strong reductions in the mRNA levels for both genes, decreasing induction of ARG1 and SNZ1 mRNAs by ~90 and 70%, respectively (Fig. (Fig.6A).6A). The pgd1Δ single mutation reduced the induction of ARG1 and SNZ1 mRNAs by 20 and 40%, respectively, whereas the gal11Δ mutation decreased ARG1 mRNA by 30% but did not impair induction of SNZ1 mRNA. Consistent with our previous findings, sin4Δ had little or no effect on the induction of either transcript. Thus, Med2p seems to play the most critical role, while Sin4p is dispensable for high-level transcription of these genes under inducing conditions.
Interestingly, the pgd1Δ and sin4Δ single mutants showed a marked derepression (approximately ninefold and fourfold, respectively) of ARG1 mRNA under noninducing conditions (Fig. (Fig.6B).6B). In fact, sin4Δ cells showed nearly the same levels of ARG1 mRNA under noninducing and inducing conditions (cf. Fig. 6A and B). Consistent with this, we found previously that sin4Δ produced a fivefold derepression of the Gcn4p-dependent CYC1-lacZ reporter described above and that pgd1Δ produced an approximately fivefold derepression of a HIS3-GUS reporter under noninducing conditions (40). The gal11Δ mutation also conferred a moderate (approximately twofold) derepression of both ARG1 and SNZ1 mRNAs in the uninduced cells (Figs. (Figs.6B).6B). Thus, it seems that inactivation of Sin4p, Pgd1p, or Gal11p derepresses promoter activity at low levels of Gcn4p binding to the UAS under noninducing conditions, even though Pgd1p and Gal11p are required for optimal transcriptional activation by Gcn4p under inducing conditions.
We also conducted Northern analysis on double mutants lacking two different subunits of the tail domain. All of the double mutants involving med2Δ had less-severe induction defects than did the med2Δ single mutant (Fig. (Fig.6A).6A). Thus, gal11Δ, pgd1Δ, and sin4Δ partially overcome the activation defect conferred by med2Δ, possibly through their derepressing effects on promoter function. On the other hand, sin4Δ exacerbated the modest activation defects conferred by the pgd1Δ and gal11Δ mutations (Fig. (Fig.6A).6A). The sin4Δ gal11Δ double mutant also showed a synthetic slow-growth phenotype on nonstarvation medium (data not shown). The last findings suggest that there are additive defects in recruitment or coactivator functions of mediator produced by simultaneously deleting Sin4p and Pgd1p or Gal11p which outweigh the derepression of promoter activity produced by eliminating these tail subunits individually.
We next compared levels of recruitment of TBP and Pol II by Gcn4p in the med2Δ and sin4Δ mutants. ChIP analysis with antibodies against TBP or the Pol II subunit Rpb1p showed an approximately threefold-higher level of binding by TBP and Rpb1p to the ARG1 promoter and an approximately threefold greater association of Rpb1p with the ARG1 ORF in the WT strains containing episomal GCN4 than in those carrying gcn4Δ (Fig. (Fig.7,7, WT bars). These results demonstrate that Gcn4p recruits TBP and Pol II to the promoter in the course of stimulating ARG1 transcription, as also shown recently (34). In the med2Δ strain, we observed strong reductions in recruitment of TBP and Rpb1p to the promoter and decreased Rpb1p occupancy in the ARG1 ORF (Fig. (Fig.7),7), all consistent with the impaired induction of ARG1 mRNA seen in this mutant (Fig. (Fig.6A).6A). Thus, Med2p is required for the stimulation of PIC assembly by Gcn4p at ARG1. In accordance with the WT induced level of ARG1 mRNA in the sin4Δ strain (Fig. (Fig.6A),6A), we observed only a small reduction in TBP recruitment to the promoter (Fig. (Fig.7A)7A) and no reduction in Rpb1p association with the ARG1 ORF (Fig. (Fig.7C)7C) in this mutant. Thus, it appears that deletion of Sin4p and the attendant dissociation of the tail domain from the rest of mediator have little impact on the ability of mediator to promote TBP recruitment and transcription initiation. Considering that the Gal11p/Med2p/Pgd1p triad is the predominant entity containing Med2p that is recruited to ARG1 in sin4Δ cells (Fig. (Fig.44 and and5),5), these findings could indicate that this triad can carry out Med2p-dependent stimulation of TBP recruitment independently of the rest of Srb mediator. This interesting possibility is discussed further below. We will also address the fact that Pol II occupancy at the promoter, but not in the coding region, of ARG1 was reduced by the sin4Δ mutation (cf. Fig. 7B and C).
The results of our in vitro experiments indicate that the tail subunits Gal11p, Pgd1p, and Med2p and the head subunit Srb2p make overlapping contributions to the binding of Srb mediator in cell extracts to recombinant Gcn4p. Whereas single deletions of tail subunits Gal11p, Pgd1p, and Med2p had little effect on mediator binding to GST-Gcn4p, a strong reduction in binding occurred when any pair of these subunits was deleted simultaneously. Consistent with this, deletion of SIN4 alone nearly abolished the binding of mediator to GST-Gcn4p in vitro (Fig. (Fig.1).1). It was shown previously (29) and confirmed here by coimmunoprecipitation analysis that the mediator complex in sin4Δ extracts lacks Gal11p, Pgd1p, and Med2p in addition to Sin4p, accounting for its negligible interaction with GST-Gcn4p. Mediator binding to GST-Gcn4p also was strongly impaired when either Gal11p, Pgd1p, or Med2p was deleted together with head subunit Srb2p. Coimmunoprecipitation experiments showed that the head and middle domains of mediator were still associated with Sin4p in all three of these double mutants lacking Srb2p (Fig. (Fig.2B).2B). Hence, the simplest explanation for our in vitro binding data is that Gal11p, Pgd1p, Med2p, and Srb2p make independent contributions to the interaction of mediator with Gcn4p, none of which is uniquely required for stable association between mediator and GST-Gcn4p in vitro. The interaction is severely impaired, however, when the contributions of any two of these four subunits are eliminated simultaneously.
Others have shown that recombinant Gcn4p cannot bind to mediator complexes purified from gal11Δ, med2Δ, or pgd1Δ single mutants. Our results do not contradict these previous findings because the purified complexes in those studies lacked all three subunits, Gal11p, Med2p, and Pgd1p (23, 33), and thus were equivalent to the mediator complex present in our sin4Δ extracts. Park et al. reported that GST-Gcn4p binds to recombinant forms of Pgd1p/Hrs1p and Gal11p, but not to Med2p (33). It is possible that deleting Med2p together with Pgd1p or Gal11p leads to loss of the third member of this triad from the mediator. This could explain our finding that eliminating Med2p exacerbates the binding defects caused by removing Gal11p or Pgd1p without having to invoke a direct contact between Med2p and Gcn4p.
Our coimmunoprecipitation analysis showed that myc-Gal11p remains associated with HA-Pgd1p and HA-Med2p in a sin4Δ extract in which these subunits are dissociated from the rest of mediator (Fig. (Fig.3A).3A). Although interaction of head and middle domain subunits with GST-Gcn4p was impaired by sin4Δ, the binding of myc-Gal11p, HA-Pgd1p, and HA-Med2p to GST-Gcn4p was retained in sin4Δ extracts (Fig. (Fig.3B).3B). These findings indicate that the stable myc-Gal11p/Med2p/Pgd1p triad present in sin4Δ cells can interact effectively with recombinant Gcn4p in vitro.
In view of the last results, it might be expected that Sin4p would be the only subunit essential for recruitment of mediator head and middle domains by Gcn4p in vivo, since the other three tail subunits and Srb2p appeared to make redundant contacts with GST-Gcn4p in vitro. However, our ChIP analysis showed that single deletions of each tail subunit, or of Srb2p, strongly impaired recruitment of the head subunit myc-Srb6p by Gcn4p to ARG1 and SNZ1 (Fig. 4A to C). Similar results were obtained for recruitment of myc-Sin4p, because deleting any of the other three tail subunits greatly reduced its recruitment to both target genes (Fig. 5D and E). To reconcile these in vivo findings with our in vitro binding data, it could be proposed that loss of a contact between Gcn4p and mediator resulting from single deletions of GAL11, MED2, PGD1, or SRB2 can be overcome in vitro by mass action at the high concentrations of recombinant GST-Gcn4p used in the binding assays. The concentration of Gcn4p may be lower in living cells, given the short half-life of Gcn4p in vivo (19). In addition, Gcn4p must compete with other transcriptional activators for mediator binding, and the concentrations of these competing proteins will be far lower than that of recombinant GST-Gcn4p in the in vitro assays.
Srb2p was less critical than the tail subunits for recruitment of myc-Sin4p to SNZ1, and it was nearly dispensable for myc-Sin4p recruitment to ARG1 (Fig. 5D and E). To explain the fact that deletion of Srb2p greatly reduced recruitment of myc-Srb6p (Fig. 4A to C) but had relatively little effect on myc-Sin4p recruitment, it could be proposed that myc tagging Srb6p and deleting Srb2p have additive effects on the interaction of Gcn4p with the head domain in which both of these subunits reside. The compound effects on the head domain produced by combining srb2Δ with SRB6-myc would be equivalent to a single deletion of GAL11, MED2, or PGD1 in weakening the interaction of Srb mediator with Gcn4p in vivo. By contrast, the head domain alteration produced by srb2Δ in the SIN4-myc strain would be insufficient to reduce mediator recruitment to ARG1. By this model, Srb2p is less critical than the tail subunits for recruitment of WT mediator by Gcn4p in vivo.
Whereas recruitment of head subunits Srb6p and Srb5p by Gcn4p was strongly dependent on Sin4p (Fig. 4A to C and and5C),5C), recruitment of the tail subunits myc-Gal11p and Med2p occurred at high levels in sin4Δ cells (Fig. 4D to F and 5A and B). This is the expected outcome if a large fraction of the Gal11p/Med2p/Pgd1p triad is dissociated from the head and middle domains in vivo, as we saw in vitro (Fig. (Fig.3A),3A), and if the Gal11p/Med2p/Pgd1p triad can be recruited by Gcn4p independently of the rest of mediator. Thus, these results provide the strongest evidence to date that the mediator tail subunits are direct targets of the Gcn4p activation domain in vivo. The binding of myc-Gal11p to ARG1 and SNZ1 was impaired by med2Δ (Fig. 4D to F), indicating that myc-Gal11p is dependent on its association with Med2p for efficient recruitment by Gcn4p. Because pgd1Δ produced a smaller reduction in myc-Gal11p recruitment, it seems that myc-Gal11p is less dependent on Pgd1p than on Med2p for interaction with Gcn4p. On the other hand, pgd1Δ strongly impaired myc-Med2p recruitment at both genes (Fig. 5A and B). That recruitment of myc-Med2p seems to show a greater requirement for Pgd1p than does recruitment of myc-Gal11p may be explained by the fact that myc-Med2p levels are generally lower in the pgd1Δ strain than in WT or gal11Δ strains (Fig. (Fig.2E2E and data not shown). In any event, it is clear that optimal recruitment of the Gal11p/Med2p/Pgd1p triad by Gcn4p in vivo requires the integrity of its constituent subunits. We propose that complex formation by these proteins produces an extended binding surface that accommodates simultaneous interactions with multiple hydrophobic clusters in the Gcn4p activation domain.
Considering a previous report that Gal11p is associated with the Paf1 complex (39), it could be proposed that the Gal11p/Med2p/Pgd1p triad is recruited in sin4Δ cells in the context of the Paf1 complex. We believe that this is unlikely for several reasons. First, we previously detected no binding of Paf1 to GST-Gcn4p in vitro under conditions where the binding of Srb mediator was robust (40). Second, we detected no coimmunoprecipitation of HA-Med2p or HA-Pgd1p with myc-tagged Paf1p from cell extracts under conditions where association of HA-Med2p and HA-Pgd1p with Srb mediator subunits occurred at high levels (Fig. (Fig.3A3A and data not shown). Third, there has been no previous report that Med2p and Pgd1p are associated with the Paf1 complex. Fourth, recent affinity purifications of the Paf1 complex have not identified Gal11p as a subunit, suggesting that it is a very minor constituent of the complex. Fifth, we found recently that deletion of the Cdc73p subunit of the Paf1 complex does not affect recruitment of TBP by Gcn4p to ARG1 (34), whereas deletion of Med2p clearly does (Fig. (Fig.7A).7A). Thus, the function of Med2p in TBP recruitment is not dependent on the intact Paf1 complex. Admittedly, we cannot yet eliminate the possibility that Med2p must function in TBP recruitment in association with the Paf1 complex in sin4Δ cells where the association of Med2p with Srb mediator is disrupted.
It is interesting that deleting the head subunit gene SRB2 generally had a greater effect on recruitment of the tail subunits Gal11p and Med2p than did sin4Δ (Fig. 4A to C and 5A and B). This result was unexpected because sin4Δ dissociates the Gal11p/Med2p/Pgd1p triad from the rest of mediator, whereas srb2Δ has no effect on the association of tail subunits with the rest of mediator (Fig. 2B, D, and E). Thus, it is very unlikely that srb2Δ impairs the interaction of tail subunits with one another. To account for the greater effect of srb2Δ than of sin4Δ on recruitment of Gal11p and Med2p, we suggest that dissociation of the Gal11p/Med2p/Pgd1p subcomplex from the rest of mediator in the sin4Δ mutant enhances recruitment of the isolated triad by Gcn4p, compensating for the loss of interaction between Gcn4p and Srb2p in the head domain. Perhaps incorporation of the Gal11p/Med2p/Pgd1p subcomplex into mediator occludes one of its contacts with Gcn4p and necessitates an additional interaction with the head domain via Srb2p. Indeed, single-particle electron microscopy analysis revealed physical interactions between the tail and middle domains of mediator (7) which might obscure an activator binding surface in the tail.
Deleting any of the four tail subunits of mediator strongly reduced recruitment of the head and middle domains of Srb mediator by Gcn4p to ARG1 and SNZ1. Because mediator is essential for most Pol II transcription (42), this might be predicted to produce a severe defect in transcriptional activation. However, only the med2Δ strain displayed strong reductions in the induced levels of ARG1 and SNZ1 mRNAs (Fig. (Fig.6A).6A). Our ChIP analysis showed that recruitment of TBP and Pol II to the ARG1 promoter was greatly impaired in med2Δ cells (Fig. 7A and B), accounting for the reduced occupancy of Pol II in the ORF (Fig. (Fig.7C)7C) and for the defective induction of ARG1 mRNA seen in this strain. Thus, Med2p is critically required for the PIC assembly stimulated by Gcn4p, in accordance with our other recent findings (34). Remarkably, sin4Δ had little or no effect on TBP recruitment to the promoter (Fig. (Fig.7A),7A), even though deletion of Sin4p eliminates a critical connection between the tail subunits and the rest of mediator. Considering that the Gal11p/Med2p/Pgd1p triad is the predominant mediator entity recruited to ARG1 in sin4Δ cells, it seems possible that Med2p can carry out its function in TBP recruitment in the context of the Gal11p/Med2p/Pgd1p triad independently of the rest of mediator. However, there are some caveats to this conclusion.
It seems improbable that the Gal11p/Med2p/Pgd1p triad alone could support a WT level of TBP recruitment, considering that an srb4 Ts− mutation was shown to abolish TBP recruitment by Gal4p (20, 24), plus our finding that the mediator head subunit Rox3p is critical for TBP recruitment by Gcn4p (34). Thus, there are likely to be contributions from both head and tail domains to the TBP recruitment function of mediator. In fact, we always observed a low but significant recruitment of mediator head subunits in the sin4Δ strain (Fig. 4B and C). This could result from inefficient recruitment of tailless mediator or from recruitment of a small fraction of otherwise intact mediator lacking only Sin4p which was too unstable to be detected in our coimmunoprecipitation analysis of the SRB6-myc sin4Δ strain (Fig. (Fig.2C).2C). Because we don't know how much mediator is required at the promoter to support a WT level of TBP recruitment, a small fraction of tailless or Sin4p-less mediator could make a considerable contribution to this function. Hence, it is unclear how much of the TBP recruitment that occurs in sin4Δ cells can be attributed to the free Gal11p/Med2p/Pgd1p triad. Nevertheless, it is intriguing that a recent model for the structure of a Pol II/TFIIF/TFIIB/TBP/DNA initiation complex and its interaction with mediator seems to position the tail domain of mediator in the vicinity of TFIIB and TBP (3, 5), consistent with a role in stabilizing TBP binding to the promoter.
Although the sin4Δ mutant exhibits a nearly WT level of TBP recruitment, it shows reduced occupancy of Pol II at the ARG1 promoter. Interestingly, this apparent defect in Pol II recruitment is not associated with smaller amounts of Pol II in the ARG1 ORF (Fig. (Fig.7C).7C). To reconcile these findings, it could be proposed that inactivation of Sin4p increases the rate of promoter clearance and thereby compensates for a reduced rate of Pol II recruitment in the sin4Δ mutant. This could yield a net reduction in steady-state promoter occupancy by Pol II without significantly decreasing the transcription rate at ARG1. A tacit assumption of this explanation is that promoter clearance is a relatively slow step in WT cells. However, we cannot dismiss the possibility that a conformational change in the PIC associated with sin4Δ decreases the efficiency of Pol II cross-linking or immunoprecipitation and does not produce a genuine reduction in Pol II binding to the promoter.
How can we account for the weak phenotypes of gal11Δ and pgd1Δ mutants given that these two mutations (unlike sin4Δ) decrease the recruitment of both head and tail domain subunits by Gcn4p? One possibility is that Gcn4p normally recruits the mediator in excess of the amount needed to support WT transcription, so that the large reductions in mediator recruitment observed in the gal11Δ and pgd1Δ mutants do not produce a strong decrease in the transcription rate. Physiologically relevant levels of coactivator recruitment may be below the detection limit of the ChIP assay. Gcn4p recruits an array of other coactivators, some of which (SAGA and SWI/SNF) appear to have functions redundant with those of Srb mediator (37). Hence, certain mediator functions may be carried out by other coactivators when mediator recruitment is diminished (but not abolished) by deletion of Gal11p, Pgd1p, or Sin4p. To account for the stronger activation defect seen in med2Δ cells, we note first that med2Δ produced the most severe overall recruitment defects seen among all of the mediator mutants that we tested. Second, it is possible that deletion of Med2p indirectly impairs the functions of both Gal11p and Pgd1p, thus disabling the entire tail domain. Finally, Med2p may have a unique function that cannot be carried out by other coactivators.
An alternative explanation for the weak effects of gal11Δ and pgd1Δ on transcriptional induction, which could also explain the high-level transcription observed in sin4Δ cells, is prompted by our finding that these mutations produce partial (gal11Δ and pgd1Δ) or nearly complete (sin4Δ) derepression of ARG1 transcript levels under noninducing conditions. In addition, they suppress, rather than exacerbate, the transcription defects seen in med2Δ cells (Fig. (Fig.6B).6B). Others have reported that inactivation of Sin4p, Gal11p, or Pgd1p derepresses transcription from other promoters lacking a UAS, in mutants defective for a coactivator, or under conditions where an activator is nonfunctional (reviewed in reference 30). These findings suggest that these three tail subunits function in repressing basal promoter activity even though they are required for optimal induction by the activator (27), such that deletions of these proteins have offsetting positive and negative effects on transcription. In this view, the loss of recruitment and coactivator functions would outweigh the disruption of repressing activity only in the sin4Δ gal11Δ and sin4Δ pgd1Δ double mutants, where we observed a marked reduction in transcription. Perhaps the mediator complexes lacking only Gal11p or Pgd1p function more effectively than WT mediator in promoter clearance, as suggested above for the sin4Δ mediator. The repressor activity of Sin4p might also be related to its ability to destabilize reinitiation complexes on immobilized promoters in cell extracts (36). An intriguing possibility is that the tail domain exerts an inhibitory effect on the head or middle domains of mediator through the known physical contacts between these two domains (3, 7) in order to prevent promiscuous transcription in the absence of activators. According to this model, deletion of Sin4p, Gal11p, or Pgd1p would weaken the negative function of the tail domain and bypass the need for Gcn4p to overcome the repressor function of mediator. Presumably, Med2p does not participate in this negative regulatory function and is required only in the recruitment and coactivator functions of mediator. The fact that the sin4Δ, gal11Δ, and pgd1Δ mutations partially suppress the induction defect in med2Δ cells implies that one role of Med2p is to overcome the negative function of these other tail domain subunits in the presence of induced levels of activator Gcn4p.
Our finding that ARG1 mRNA levels are constitutively elevated in the sin4Δ mutant (Fig. (Fig.6)6) may seem at odds with the ChIP data showing strong Gcn4p-dependent increases in both TBP binding to the promoter and Pol II association with the ORF in sin4Δ cells (Fig. 7A and C). However, this discrepancy can be explained in several ways. It is possible that the high-level ARG1 transcription seen in uninduced sin4Δ cells (Fig. (Fig.6B)6B) is still dependent on the basal level of Gcn4p binding to the ARG1 promoter that occurs in nonstarvation conditions. Indeed, the uninduced level of Gcn4p makes a strong contribution to ARG1 transcription in nonstarved WT cells (6). It is also possible that the low-level recruitment of TBP and Pol II afforded by uninduced levels of Gcn4p is sufficient for high-level ARG1 transcription when the repressing functions of the mediator tail domain are inactivated by deletion of Sin4p.
In summary, our results provide strong evidence that the mediator tail domain is an in vivo target of Gcn4p. Gal11p, Med2p, and Pgd1p are necessary for efficient recruitment of the head and middle domains of mediator and are sufficient for high-level recruitment by Gcn4p as an isolated subcomplex in sin4Δ cells. We showed that Med2p is critical for TBP and Pol II recruitment, whereas sin4Δ has relatively little effect on PIC formation, even though it disrupts association of the Gal11p/Med2p/Pgd1p triad with the rest of mediator. These last findings raise the intriguing possibility that the tail subcomplex has an important coactivator function in TBP recruitment. It will be interesting to determine whether the Gal11p/Med2p/Pgd1p triad serves directly as an adaptor or facilitates the recruitment of another GTF or coactivator with a role in TBP recruitment.
We thank Rick Young, Tony Weil, Jerry Workman, Brad Cairns, Larry Myers, Roger Kornberg, Joe Reese, Michael Green, Stefan Bjorklund, and Young-Joon Kim for generous gifts of antibodies. We are grateful to Chhabi Govind, Sungpil Yoon, Hongfang Qiu, and Cuihua Hu for discussions, advice, and help in myc tagging.