Simultaneous Recruitment of Coactivators by Gcn4p Stimulates Multiple Steps of Transcription In Vivo

Transcriptional activation by Gcn4p is dependent on the coactivatorsSWI/SNF, SAGA, and Srb Mediator, which are recruited by Gcn4pand stimulate assembly of the preinitiation complex (PIC) atthe ARG1 promoter in vivo. We show that recruitment of all threecoactivators is nearly simultaneous with binding of Gcn4p atARG1 and is followed quickly by PIC formation and elongationby RNA polymerase II (Pol II) through the open reading frame.Despite the simultaneous recruitment of coactivators, rapidrecruitment of SWI/SNF depends on the histone acetyltransferase(HAT) subunit of SAGA (Gcn5p), a non-HAT function of SAGA, andon Mediator. SAGA recruitment in turn is strongly stimulatedby Mediator and the RSC complex. Recruitment of Mediator, bycontrast, occurs independently of the other coactivators atARG1. We confirm the roles of Mediator and SAGA in TATA bindingprotein (TBP) recruitment and demonstrate that all four coactivatorsunder study enhance Pol II recruitment or promoter clearancefollowing TBP binding. We also present evidence that SWI/SNFand SAGA stimulate transcription elongation downstream fromthe promoter. These functions can be limited to discrete timeintervals, providing evidence for multiple stages in the inductionprocess. Our findings reveal a program of coactivator recruitmentand PIC assembly that distinguishes Gcn4p from other yeast activatorsstudied thus far.

INTRODUCTION

Top

Introduction

The activation of transcription initiation in eukaryotes involvesthe binding of a sequence-specific activator protein to an enhancer,or upstream activation/regulatory sequence (UAS/URS), followedby activator-dependent recruitment of the general transcriptionmachinery. Because packaging of promoter DNA in nucleosomesimpedes assembly of the preinitiation complex (PIC), activatorsmust recruit chromatin-modifying complexes to the promoter.These include coactivator complexes with histone acetyltransferase(HAT) activity, such as the Gcn5p-containing SAGA complex ofSaccharomyces cerevisiae, and ATP-dependent remodeling enzymes,such as SWI/SNF and RSC. Nucleosome remodeling by SWI/SNF hasalso been implicated in removing a barrier to transcriptionelongation by RNA polymerase II (Pol II) (8). The recruitmentof TATA binding protein (TBP) and Pol II is facilitated by coactivatorswhich serve as adaptors between the activator and these generaltranscription factors (GTFs). For example, SAGA complex containssubunits that interact with TBP and has been shown to mediateTBP recruitment by various activators (2, 22, 30). Similarly,the Mediator complex interacts with Pol II and has been shownto stimulate the recruitment of both Pol II and TBP (21, 25,26, 30).

The recruitment of one coactivator may stimulate the recruitmentof another, and this interdependence can be reflected in a sequentialorder of coactivator recruitment to the promoter. Such stagedrecruitment was shown for the HO gene where SWI/SNF recruitmentis necessary for, and temporally precedes, recruitment of SAGAby the activator Swi5p to the URS2 element (9). SWI/SNF is alsorequired for Mediator recruitment to HO, even though these twocoactivators are recruited simultaneously to URS1, whereas Mediatorrecruitment is dispensable for SWI/SNF recruitment to this promoter(4). At the GAL1 gene of yeast, by contrast, SAGA recruitmentby Gal4p precedes that of Mediator (6). Although SWI/SNF isrecruited to GAL1 as rapidly as SAGA, and independently of SAGAfunction, strong SWI/SNF recruitment requires Mediator and downstreamsteps in PIC assembly (19, 24). Thus, the temporal order ofcoactivator recruitment at GAL1 does not seem to reflect anobligate sequence of coactivator functions. The requirementfor SWI/SNF function as a prerequisite for SAGA recruitmentby Swi5p at HO appears to be restricted to late mitosis andapplies even to Gal4p- and Gcn4p-regulated promoters in thisphase of the cell cycle, most likely reflecting a highly condensedstate of promoter chromatin (19). In fact, both SAGA and Mediatorfunctions are required during interphase for wild-type steady-staterecruitment of SWI/SNF by Gcn4p at the target genes ARG1 andSNZ1 (43). Thus, the degree of interdependency among coactivatorsin recruitment can vary from one yeast activator to the next,and even for the same activator depending on the chromatin structureof the promoter.

A mechanism to explain SAGA-SWI/SNF interdependency has beenproposed wherein histone acetylation by Gcn5p enhances recruitmentof SWI/SNF and of SAGA itself via the bromodomains in the Snf2pand Gcn5p subunits, respectively. This mechanism was demonstratedin vitro using reconstituted nucleosome arrays, and geneticdata indicated a requirement for the Snf2p bromodomain in SWI/SNFrecruitment to the SUC2 promoter in vivo (15). However, a differentresult was obtained in another study where the Snf2p bromodomainwas dispensable for wild-type SWI/SNF recruitment to SUC2, eventhough mutations in the HAT subunits of SAGA (Gcn5p) and NuA4(Esa1p) decreased SWI/SNF recruitment in the early phase ofSUC2 induction (13). Similarly, a deletion of GCN5 did not reducesteady-state recruitment of SWI/SNF by Gcn4p (38, 43), althoughit impaired chromatin remodeling by the recruited SWI/SNF complex(38). However, it has not been determined whether the kineticsof SWI/SNF (or SAGA) recruitment by Gcn4p is influenced by theGcn5p HAT function in vivo.

A key objective of this study was to determine whether Gcn4pmounts a staged recruitment of SAGA, SWI/SNF, and Mediator,followed by PIC assembly, in which earlier events are prerequisitesfor later steps in the process. To that end, we analyzed thekinetics of Gcn4p binding, coactivator recruitment, bindingof TBP and Pol II to the promoter (PIC assembly), and the appearanceof Pol II at the 3' end of the open reading frame (ORF) duringinduction of the ARG1 gene in wild-type cells and in variouscoactivator mutants. Our results indicate that transcriptionalactivation by Gcn4p is a rapid process commencing with the immediateoccupancy of the ARG1 UAS by Gcn4p on amino acid starvation,followed quickly by nearly simultaneous recruitment of SAGA,SWI/SNF, and Mediator. These coactivators, together with RSC,coordinate the rapid recruitment of TBP and Pol II to the promoterand elongation through the ORF by Pol II without any detectabledelay in promoter clearance. Despite the nearly simultaneousrecruitment of coactivators, our mutant analyses reveal extensiveinterdependencies among the coactivators for their rapid recruitmentby Gcn4p. In particular, we provide in vivo evidence that theSAGA HAT subunit is required for wild-type kinetics of SWI/SNFrecruitment and that RSC function is needed for optimal SAGArecruitment. We also show that Mediator is strongly requiredfor activator recruitment of both SAGA and SWI/SNF throughoutthe entire course of induction.

Our kinetic analyses of PIC formation in coactivator mutantsshow that TBP recruitment per se is not sufficient for wild-typepromoter occupancy by Pol II, suggesting that all four coactivatorsenhance Pol II recruitment or promoter clearance after bindingof TBP to the promoter. We also uncovered functions for SWI/SNFand SAGA in transcription elongation downstream of the promoter.Together, our results provide a detailed picture of the activationmechanism for Gcn4p that differs significantly from that describedfor other activators, and they extend the range of functionsstimulated by the participating coactivators in vivo.

MATERIALS AND METHODS

Top

Materials and Methods

Chromatin immunoprecipitation was conducted as described previously(30) using the following strains containing ADA2-myc describedpreviously (31): HQY392 (wild type), HQY666 (snf2 ${Delta}$ ), HQY501 (rsc2 ${Delta}$ ),HQY573 (rox3 ${Delta}$ ), HQY420 (gcn5 ${Delta}$ ), HQY418 (ada1 ${Delta}$ ), and HQY503 (gcn4 ${Delta}$ ).These strains were derived from deletion strains purchased fromResearch Genetics, and the presence of all deletions was verifiedby PCR analysis of genomic DNA. Cells were grown in syntheticcomplete (SC) medium (35). Anti-myc antibodies and anti-Rpb1pantibodies (8WG16) were purchased from Santa Cruz Biotechnologiesand AbCam, respectively. Antibodies against Gcn4p were producedpreviously (K. Natarajan and A. G. Hinnebusch, unpublished observations).Antibodies against Snf6p and TBP were kindly provided by JosephReese, and Gal11p antibodies were generously supplied by MarkPtashne.

RESULTS

Top

Results

Rapid, sequential binding of Gcn4p, coactivators and GTFs to the ARG1 promoter in starved cells. We used the chromatin immunoprecipitation (ChIP) assay to measurethe kinetics of Gcn4p binding and recruitment of coactivators,TBP, and Pol II to the ARG1 promoter following induction ofGcn4p synthesis in a strain with a myc-tagged version of SAGAsubunit Ada2p. Aliquots of the same chromatin preparations fromthis strain were immunoprecipitated with antibodies againstGcn4p, TBP, SWI/SNF subunit Snf6p, Pol II subunit Rpb1p, andMediator subunit Gal11p, and with myc antibodies (against myc-Ada2p)to evaluate recruitment of all of these factors to ARG1 in thesame population of induced cells. (We use the terms "recruitment"and "binding" to signify occupancy, representing the net outcomeof factor association and dissociation from the UAS or promoter.)

In brief, cells growing exponentially at 25°C in complete(SC) medium were treated with sulfometuron-methyl (SM) to inhibitbiosynthesis of isoleucine and valine and thereby induce Gcn4p.GCN4 expression is rapidly induced at the translational levelby starvation for any amino acid (16). (The cells were grownat 25°C to reduce the rate of Gcn4p binding and factor recruitment,because at 30°C we observed nearly identical rates of recruitmentfor all factors under study [data not shown]). Aliquots of theculture were collected at time intervals ranging from 2.5 minto 120 min postinduction and treated with formaldehyde to cross-linkfactors to chromatin in the living cells. An aliquot of cellswas removed prior to SM addition and cross-linked to providea noninduced control sample, and an isogenic gcn4 ${Delta}$ ADA2-myc strainwas cross-linked after 120 min of SM treatment to evaluate thebackground level of Gcn4p-independent factor binding at ARG1.The DNA extracted from immunoprecipitated chromatin fragmentswas subjected to PCR analysis with various primers to quantifythe UAS or TATA sequences in the promoter or sequences at the3' end of the ARG1 ORF (Fig. 1A). Immunoprecipitations wereconducted on at least two chromatin samples prepared from replicatecultures, and the PCR analysis of coprecipitated DNA sequenceswas conducted in duplicate or triplicate for each immunoprecipitatedsample.

View larger version (59K):
[in this window]
[in a new window]
FIG. 1. ChIP analysis of the kinetics of Gcn4p binding and recruitment of coactivators, TBP, and Pol II to ARG1 on induction of Gcn4p by SM. GCN4 ADA2-myc strain HQY392 was cultured at 25°C in SC medium and treated with SM at 0.5 µg/ml. Aliquots of cells were cross-linked with formaldehyde at the indicated times and processed for ChIP analysis of factor binding to ARG1 using primers to amplify the UAS- or TATA-containing sequences, or ORF sequences, at ARG1 indicated in the schematic shown in panel A. Coding sequences of the POL1 gene were amplified to control for nonspecific immunoprecipitation. Chromatin fragments were immunoprecipitated with the antibodies shown to the right of each panel, and the DNA was extracted from the immunoprecipitates (IP) and from 5% of the corresponding input chromatin (Inp) samples. A 1,000-fold dilution of the Inp and the undiluted IP DNA samples were PCR amplified in the presence of [³³P]dATP, and the PCR products were resolved by polyacrylamide gel electrophoresis and visualized by autoradiography. In parallel, gcn4 ${Delta}$ ADA2-myc strain HQY503 was treated with SM for 120 min and handled identically (last lane). The PCR-amplified fragments were quantified with a phosphorimager, and the ratios of the ARG1 signals to the POL1 signals in the IP samples were normalized for the corresponding ratios for the Inp samples to yield the "relative % IP ARG1/POL1" for each sample.

Representative data showing the kinetics of Gcn4p binding andthe recruitment of Mediator, SAGA, SWI/SNF, TBP, and Pol IIsubunits to ARG1 over the first 40 min of induction by SM arepresented in Fig. 1B to D. The fraction of the total ARG1 sequencesin the input chromatin sample that was precipitated with eachantibody was calculated and normalized for the fraction of coprecipitatingPOL1 ORF sequences in order to correct for nonspecific precipitationof chromatin fragments. The mean values and standard errorsfor these normalized ratios, referred to as relative % IP ARG1/POL1values, were calculated from the multiple PCR measurements foreach time point and plotted in the graphs presented in Fig.2. The relative % IP ARG1/POL1 values can be equated with therelative occupancy of the factor under consideration at ARG1.

View larger version (27K):
[in this window]
[in a new window]
FIG. 2. Kinetic ChIP analysis suggests sequential binding of Gcn4p, coactivators, and TBP/Pol II at the ARG1 promoter on induction of Gcn4p. The results of replicate ChIP experiments conducted as exactly as described in Fig. 1 were quantified and the mean "relative % IP ARG1/POL1" values with standard errors were plotted as a function of time for each factor, indicated along the y axes. Note that different y axis scales were employed to normalize the peak heights for the different curves.

As shown below, ChIP signals of greatly different magnitudewere obtained for the various factors binding at ARG1. It isunclear whether these differences reflect real distinctionsin factor occupancy, different efficiencies of cross-linkingto chromatin, or dissimilar efficiencies of immunoprecipitation.Thus, it could be misleading to compare the absolute rates ofincrease in promoter occupancy from one factor to the next.Hence, we used different scales to plot the data in Fig. 2 tonormalize the peak occupancy values for the set of factors underconsideration. To compare the kinetics of factor recruitment,we evaluated the times required to reach the half-maximal andmaximum occupancies for each factor during the first 40 minof induction.

We observed a low level of Gcn4p binding to the ARG1 UAS innoninduced cells at a level $~$ 2.5-fold above the background signalmeasured in the gcn4 ${Delta}$ strain (Fig. 1B, GCN4 at 0 min versus gcn4 ${Delta}$ ,and 2A, 0 min). Following treatment with SM, there was an immediateincrease in Gcn4p binding to the ARG1 UAS that peaked by 15to 20 min at a level $~$ 8-fold above the noninduced value (Fig.2A). The level of Gcn4p binding declined somewhat at later timepoints, reaching a steady-state level $~$ 5-fold higher than thenoninduced value by 120 min (see Fig. 3A for time points between40 and 120 min). The results of Western analysis (data not shown)revealed a rapid increase in the steady-state level of Gcn4pthat paralleled the increase in Gcn4p binding to the ARG1 promoterobserved in ChIP assays for the first 30 to 40 min of induction.This rapid induction of Gcn4p is in accord with the fact thatGCN4 expression is controlled at the translational level. However,Western analysis did not reveal any decline in Gcn4p levelsat later time points that would coincide with the decrease inGcn4p occupancy at ARG1 observed in ChIP assays between 30 minand 2 h of induction (see Fig. 3A). This decline in Gcn4p bindingat ARG1 may reflect a decrease in nuclear import, decreasedaffinity of Gcn4p for the UAS, or degradation of Gcn4p at thepromoter.

View larger version (17K):
[in this window]
[in a new window]
FIG. 3. Coactivator mutations do not substantially alter the kinetics or extent of Gcn4p binding at the ARG1 UAS. Kinetic ChIP analysis was conducted as described for Fig. 1 using the wild-type (WT) and mutant strains with the relevant genotypes listed in the inset of each graph. Antibodies against Gcn4p were employed for the ChIP assays using primers to measure its binding to the ARG1 UAS.

To determine the rate of PIC assembly following Gcn4p bindingto the UAS, we analyzed the recruitment of TBP and Pol II tothe region surrounding the TATA element at ARG1 (Fig. 2B). (Therelative % IP ARG1/POL1 values in gcn4 ${Delta}$ cells were close to unityfor these and other factors under study, indicating promoteroccupancy below the detection limit of the assay in the absenceof activator Gcn4p.) The increase in TBP occupancy required $~$ 10 min to reach its maximum rate and also peaked $~$ 10 min laterthan did Gcn4p binding. Consistent with this, half-maximal bindingof TBP occurred at $~$ 15 min, about 8 min later than that of Gcn4p.The recruitment of Pol II to the TATA region followed kineticssimilar to that seen for TBP binding to the promoter in showinga relatively slow initial rate of binding before reaching itsmaximum rate by $~$ 10 min and peaking 10 min later than did Gen4p(Fig. 2B). Thus, there is no detectable delay between the recruitmentof TBP and Pol II to the ARG1 promoter by Gcn4p. (Note thatthe ChIP signals for Pol II binding to the TATA region may bea composite of Pol II bound to the promoter and Pol II transcribingthe 5' end of the ORF.)

Recent experiments indicate that levels of Pol II occupancywithin the ORF coincide closely with the amounts of accumulatedmRNA transcripts produced from the same gene (20, 42). Therefore,we measured binding of Pol II to the 3' end of the ARG1 ORFto evaluate the increase in rate of transcription elongationduring induction by Gcn4p. The results shown in Fig. 2C indicatethat Pol II occupancy at the 3' end of the ARG1 ORF increasedwith kinetics similar to that observed for binding of TBP andPol II to the promoter (Fig. 2A and B), requiring $~$ 10 min toreach its maximum rate of binding and peaking about 12 min afterGcn4p reached its highest value. (We observed no significantbinding of Gcn4p or TBP to the 3' end of the ARG1 ORF [datanot shown], indicating that the Pol II signal measured for thislocation did not include Pol II molecules bound to the promoter,as might occur with inadequate shearing of chromatin.) Underphysiological conditions, mRNA synthesis occurs at a rate of1,200 to 1,500 nucleotides/min (17), such that transcriptionof the ARG1 ORF should require less than a minute. The factthat Pol II occupancy at the 3' end of the ORF showed no lagwith respect to PIC assembly implies that promoter clearanceoccurs almost instantaneously following PIC assembly and thattranscription proceeds at the expected rate with no significantimpediments to elongation by Pol II under inducing conditionsat ARG1. Bryant and Ptashne made a similar observation for inductionof the GAL1 promoter by Gal4p (6). The fact that Pol II occupancyin the ORF was $~$ 3-fold higher than at the promoter (cf. Fig.2B and C) implies that multiple elongating polymerase moleculescan be associated with ORF fragments of $~$ 500 bp (the averagesize of our sheared chromatin fragments) whereas only a singlePol II molecule at a time can occupy the promoter.

As would be expected from the rapid rate of PIC formation andtranscription elongation induced by Gcn4p binding at ARG1, recruitmentof the coactivators SAGA, Mediator, and SWI/SNF to the UAS occurredvery quickly following induction of Gcn4p (Fig. 2D to F). Incontrast to the results for TBP, the binding of these coactivatorsoccurred at roughly the maximum initial rate at the earliesttime point we assayed. However, their binding appeared to bebiphasic, increasing rapidly for the first 15 min in parallelwith the increase in Gcn4p binding, and then increasing moreslowly over the next 15 to 25 min to reach their peak valuesat 30 min (SWI/SNF) or 40 min (SAGA) (Fig. 2D and E) (see Fig.4A and C for the entire 120-min time course of SAGA and SWI/SNFrecruitment). Despite these biphasic kinetics, both SAGA andSWI/SNF achieved half-maximal binding at $~$ 10 min, four to fivemin preceding the half-maximal binding of TBP and Pol II. Recruitmentof Mediator also began quickly and increased rapidly with apparentlymonophasic kinetics, peaking $~$ 5 min after Gcn4p and $~$ 6 min precedingmaximum TBP binding (Fig. 2F). Like SAGA and SWI/SNF, the half-maximalbinding of Mediator occurred at $~$ 10 min. Together, these findingssuggest that the first phase of SWI/SNF and SAGA recruitmentand the monophasic recruitment of Mediator occur very quicklyafter binding of Gcn4p to the UAS and may precede the increasein PIC formation at the promoter by a short time interval of5 to 10 min. Additional evidence supporting the notion thatcoactivator recruitment precedes PIC assembly is discussed below.

View larger version (29K):
[in this window]
[in a new window]
FIG. 4. Coactivator mutations alter the kinetics and extent of SWI/SNF and SAGA recruitment by Gcn4p at the ARG1 UAS. Kinetic ChIP analysis was conducted as described in Fig. 1 using the wild-type (WT) and mutant strains with the relevant genotypes listed in the inset of each graph. Antibodies against myc-tagged Ada2p (A and B), Snf6p (C and D), or Gal11p (E and F) were employed for the ChIP assays using primers to measure binding of these factors to the ARG1 UAS. In panel A, myc-Ada2p binding in WT and snf2 ${Delta}$ cells is indicated by the scale along the left-hand y axis, while the results for gcn5 ${Delta}$ refer to the scale on the right-hand y axis.

Mediator and RSC, but not SWI/SNF or Gcn5p, stimulate SAGA recruitment by Gcn4p. Having observed almost simultaneous recruitment of SAGA, SWI/SNF,and Mediator at ARG1, we next asked whether these coactivatorsare interdependent for their rapid recruitment by Gcn4p. Forexample, wild-type recruitment of SAGA may depend on the simultaneousrecruitment of SWI/SNF or Mediator in addition to the knowndirect interaction of SAGA with the Gcn4p activation domain(11). In this event, mutations in subunits of Mediator or SWI/SNFwould impair the rate or extent of SAGA recruitment. Accordingly,we measured the kinetics of SAGA recruitment in a rox3 ${Delta}$ mutant,lacking a critical subunit of Mediator, or in a snf2 ${Delta}$ mutant,lacking the ATPase subunit of SWI/SNF. We also examined theeffects of deleting the HAT subunit of SAGA or disrupting theSAGA complex in gcn5 ${Delta}$ and ada1 ${Delta}$ mutants, respectively (14), onrecruitment of SWI/SNF, Mediator, and SAGA itself. Because theRSC complex is recruited by Gcn4p (37), we further investigatedwhether deletion of the RSC2 subunit affects SAGA recruitment.Rsc2p is a nonessential RSC subunit containing two bromodomains(7), shown previously to be required for transcriptional inductionof ARG1 by Gcn4p (30).

First, we examined the effects of these coactivator mutationson the kinetics of Gcn4p binding to the ARG1 UAS. As shown inFig. 3A and B, binding of Gcn4p in the five coactivator mutantsoccurred at similar rates and peaked within 5 min of the peaktime observed in the isogenic wild-type strain. Moreover, noneof the mutations significantly reduced the steady-state levelof Gcn4p binding observed at 120 min (Fig. 3A and B), in agreementwith our previous findings (30). The small variations in Gcn4pbinding observed in the mutant strains cannot account for themajor defects in coactivator recruitment or PIC formation describedfor these mutants below.

When we analyzed the mutants for SAGA recruitment, we foundthat the rox3 ${Delta}$ and rsc2 ${Delta}$ mutations led to severe reductions inSAGA recruitment throughout the entire time course, whereassnf2 ${Delta}$ produced only a slight reduction in SAGA recruitment (Fig.4A and B). Thus, even though Gcn4p can directly interact withSAGA, it requires functions of Mediator and RSC to achieve themaximal rate and extent of SAGA recruitment at ARG1. It wasshown in vitro that acetylation of histone H3 by Gcn5p stimulatesSAGA binding to promoter nucleosomes via the bromodomain inGcn5p (15). However, we found that neither the rate nor extentof SAGA recruitment was significantly reduced by deletion ofGCN5. In fact, SAGA recruitment was substantially higher thanwild type in the gcn5 ${Delta}$ strain at all time points (note the differencein scale for gcn5 ${Delta}$ in Fig. 4A). We conclude that recruitmentof SAGA by Gcn4p is stimulated by Mediator and RSC but independentof both SWI/SNF remodeling function and histone acetylationby SAGA throughout the whole time course of induction.

Mediator and both Gcn5p and a non-HAT function of SAGA are required for rapid recruitment of SWI/SNF. In contrast to the results obtained for SAGA, deletion of GCN5had a significant effect on SWI/SNF recruitment by Gcn4p between $~$ 15 min and 120 min of induction (Fig. 4C). As deletion of GCN5does not eliminate any other subunits from the SAGA complex(41), it is likely that the loss of SAGA HAT activity in gcn5 ${Delta}$ cells impairs SWI/SNF recruitment in this mutant. Disruptionof SAGA by ada1 ${Delta}$ produced an even stronger defect in SWI/SNFrecruitment, extending over a wider time interval than did gcn5 ${Delta}$ (Fig. 4C). Thus, both the HAT subunit (disrupted by gcn5 ${Delta}$ ) anda non-HAT function of SAGA (disrupted by ada1 ${Delta}$ ) contribute toSWI/SNF recruitment by Gcn4p. The results in Fig. 4D on therox3 ${Delta}$ mutant show that SWI/SNF recruitment is also strongly dependenton Mediator throughout the entire time course of induction.Interestingly, in rsc2 ${Delta}$ cells, SWI/SNF was recruited more rapidly,and to a greater extent, than in RSC2 cells during the first100 min of induction (Fig. 4D). Thus, SWI/SNF recruitment isnot dependent, and may be antagonized, by RSC during the earlyand middle stages of ARG1 induction by Gcn4p.

The results described above indicate that Mediator is criticallyrequired for recruitment of both SAGA and SWI/SNF by Gcn4p.By contrast, we found that inactivation of SWI/SNF by snf2 ${Delta}$ ledto somewhat greater-than-wild-type levels of Mediator recruitmentduring the initial stages of induction and had no effect by120 min (Fig. 4E). Disruption of SAGA by ada1 ${Delta}$ had only a modesteffect on the kinetics of Mediator recruitment that was limitedto the interval between 20 to 30 min of induction (Fig. 4F).Thus, both the rate and extent of Mediator recruitment by Gcn4pto ARG1 are largely independent of both SAGA and SWI/SNF atARG1.

Coactivators enhance rapid PIC formation and promoter clearance or elongation. Having observed that Gcn4p recruits SWI/SNF, SAGA, and Mediatorprior to PIC formation, we wished to examine the effects ofmutations in these coactivators on the kinetics of PIC assemblyand transcription elongation. The snf2 ${Delta}$ mutation had little effecton the kinetics or extent of TBP recruitment (Fig. 5A), butit produced a marked reduction in Pol II binding to the promoterregion between 40 to 80 min of induction (Fig. 5B) and reducedPol II occupancy in the 3' end of the ORF throughout the wholetime course of induction (Fig. 5C). The fact that Pol II occupancyin the 3' end of the ORF was reduced in snf2 ${Delta}$ cells during thefirst 30 min of induction (Fig. 5C) without any decrease inthe promoter occupancy of TBP and Pol II (Fig. 5A and B) stronglysuggests that inactivation of SWI/SNF function impairs promoterclearance or elongation by Pol II. Because we cannot distinguishbetween defects in promoter clearance or elongation with thisassay, we will use the term elongation to refer to all stepsbetween the end of PIC assembly and transcription termination.

View larger version (44K):
[in this window]
[in a new window]
FIG. 5. Coactivator mutations alter the kinetics and extent of TBP and Pol II occupancy of the promoter and of Pol II occupancy in the ORF at ARG1. Kinetic ChIP analysis was conducted as described in Fig. 1 using the wild-type (WT) and mutant strains with the relevant genotypes listed in the inset of each graph. Antibodies against TBP (A, D, and G) or Pol II subunit Rpb1p (B, C, E, F, H, and I) were employed for the ChIP assays using the appropriate primers to measure binding of TBP and Pol II to the promoter (A, B, D, E, G, and H) or Pol II occupancy in the ORF (C, F, and I). In panel D, TBP binding in WT and ada1 ${Delta}$ cells is indicated by the scale along the left-hand y axis, while the results for gcn5 ${Delta}$ refer to the scale on the righthand y-axis.

As noted above, Pol II occupancy of the promoter was impairedin snf2 ${Delta}$ cells between 40 to 80 min of induction without anydecrease in TBP recruitment (cf. Fig. 5A and B). Thus, the reducedPol II occupancy in the promoter cannot be a secondary effectof impaired TBP binding to the TATA element. One explanationfor this finding could be that SWI/SNF stimulates Pol II recruitmentto the promoter by a mechanism downstream of TBP binding. However,elongating Pol II molecules bound at the 5' end of the ORF maycontribute to Pol II occupancy in the promoter. Accordingly,SWI/SNF might stimulate promoter clearance instead if we assumethat the stalled Pol II molecules dissociate from the promoter.Supporting this last assumption, kin28 mutations, which reducephosphorylation of serine-5 in the C-terminal domain (CTD) ofPol II by TFIIH and thereby impair promoter clearance, leadto reduced ChIP signals for Pol II in the promoters of yeastgenes (27, 33). Hence, it is possible that snf2 ${Delta}$ impairs promoterclearance rather than Pol II recruitment following TBP bindingbetween 40 and 80 min of induction.

Deletion of GCN5 produced no reduction in TBP recruitment butled to a strong decrease in Pol II occupancy in the 3' end ofthe ORF throughout most of the time course of induction. Infact, TBP recruitment occurred at higher-than-wild-type levelsin the gcn5 ${Delta}$ mutant (note the different scales in Fig. 5D). Similarto snf2 ${Delta}$ , the gcn5 ${Delta}$ mutation reduced Pol II occupancy in the ORFat the 15 and 20 min time points (Fig. 5F) without producingany significant reduction in TBP or Pol II occupancy in thepromoter (Fig. 5D and E). This last comparison implies a defectin transcription elongation in gcn5 ${Delta}$ cells early in the inductionprocess. Between 20 and $~$ 80 min of induction, gcn5 ${Delta}$ also resembledsnf2 ${Delta}$ in reducing Pol II occupancy of the promoter without anydefect in TBP recruitment (cf. Fig. 5D and E). Thus, as concludedabove for snf2 ${Delta}$ , elimination of the SAGA HAT subunit Gcn5p impairseither Pol II recruitment or promoter clearance between 20 and80 min of induction. The similarity in the phenotypes of thesnf2 ${Delta}$ and gcn5 ${Delta}$ mutants shown in Fig. 5 is consistent with thefact that GCN5 is required for optimal SWI/SNF recruitment (Fig.4C) and for the remodeling function of recruited SWI/SNF (38).

Similar to our findings on snf2 ${Delta}$ and gcn5 ${Delta}$ , the ada1 ${Delta}$ mutationreduced Pol II occupancy in the ORF much more than it reducedPol II occupancy in the promoter during the whole time courseof induction, consistent with a defect in elongation (cf. Fig.5E and F). Moreover, ada1 ${Delta}$ produced a greater reduction in thepromoter occupancy of Pol II versus TBP between 20 and 40 minof induction (cf. Fig. 5D and E). Thus, ada1 ${Delta}$ additionally impairsPol II recruitment or promoter clearance downstream of TBP bindingto the TATA element. The fact that ada1 ${Delta}$ and gcn5 ${Delta}$ displayed thesame defects at steps following TBP recruitment is consistentwith the fact that Gcn5p is a component of SAGA and ada1 ${Delta}$ disruptsSAGA function. However, whereas gcn5 ${Delta}$ leads to higher-than-wild-typeTBP recruitment, ada1 ${Delta}$ reduced TBP binding at the promoter. Thisdifference is consistent with previous conclusions that a non-HATfunction of SAGA is instrumental in TBP recruitment (2, 22,30). It also helps to explain why ada1 ${Delta}$ produced a stronger reductionin ARG1 transcription than did gcn5 ${Delta}$ , nearly eliminating detectableassociation of Pol II with the ORF (Fig. 5F). Note also thatada1 ${Delta}$ more strongly impairs recruitment of SWI/SNF than doesgcn5 ${Delta}$ (Fig. 4C).

The rsc2 ${Delta}$ mutation resembles the other mutations discussed thusfar in producing marked reductions in Pol II promoter occupancy(Fig. 5B) and transcription elongation (Fig. 5C). However, rsc2 ${Delta}$ led to an earlier decline in Pol II promoter occupancy thandid snf2 ${Delta}$ or gcn5 ${Delta}$ , evident by 10 to 15 min of induction, andthus more closely resembles ada1 ${Delta}$ in this respect. The rsc2 ${Delta}$ mutantalso more closely resembled ada1 ${Delta}$ in producing a moderate reductionin TBP recruitment throughout much of the time course (Fig.5A), which may contribute to the low-level promoter occupancyof Pol II in this mutant (Fig. 5B). However, the strong decreasein Pol II promoter occupancy at 40 and 60 min in rsc2 ${Delta}$ cellscannot be explained by decreased TBP recruitment at these timepoints (cf. Fig. 5A and B), leading us to propose that rsc2 ${Delta}$ additionally impairs Pol II recruitment or promoter clearancefollowing TBP binding. It is noteworthy that ada1 ${Delta}$ and rsc2 ${Delta}$ ledto similar reductions in TBP and Pol II occupancy in the promoter(cf. Fig. 5A and B and 5D and E) but that ada1 ${Delta}$ produced a muchstronger decrease in Pol II occupancy in the ORF (cf. Fig. 5C and F).This comparison suggests that ada1 ${Delta}$ impairs transcriptionelongation to a greater extent than does rsc2 ${Delta}$ .

The rox3 ${Delta}$ mutation in Mediator led to significant reductionsin TBP binding to the promoter beginning after $~$ 15 min of inductionand extending throughout most of the remainder of the time course(Fig. 5G). It appeared that the uninduced level of TBP bindingto the promoter was elevated in this mutant. Strikingly, PolII occupancy of both the promoter and the ORF was almost completelyeliminated at all time points in rox3 ${Delta}$ cells (Fig. 5H and I).Thus, among the five mutations analyzed here, rox3 ${Delta}$ had the greatestcombined effect on TBP and Pol II recruitment to the promoter(PIC formation) and on transcription elongation. Because TBPbinding occurs at wild-type or higher levels in the first 15min of induction, but Pol II recruitment is strongly defectiveduring this interval, it follows that Mediator also stimulatesPol II recruitment or promoter clearance downstream of TBP binding(cf. Fig. 5G and H). Given the strong defect in PIC formationin the rox3 ${Delta}$ mutant, it is not possible to determine whetherthe severe reduction in Pol II occupancy of the ORF in thismutant (Fig. 5I) additionally involves a defect in elongation.The severe phenotype of rox3 ${Delta}$ shown in Fig. 5 is consistent withthe fact that this mutation impairs recruitment of both SWI/SNFand SAGA (Fig. 4A to D) in addition to impairing Mediator function.

DISCUSSION

Top

Discussion

Simultaneous, but interdependent, recruitment of SWI/SNF, SAGA, and Mediator by Gcn4p to ARG1. Our kinetic analysis of wild-type cells indicates that recruitmentof the three coactivators SWI/SNF, SAGA, and Mediator to theARG1 UAS occurs almost simultaneously with Gcn4p binding. Recruitmentof SAGA and SWI/SNF by Gcn4p begins instantaneously on inductionof Gcn4p but proceeds more slowly and, thus, requires 10 to20 min longer to reach the maximum levels than does bindingof Gcn4p. Although recruitment of Mediator may begin more slowly,it reaches its maximum value sooner than does SAGA and SWI/SNF,peaking only about 5 min later than Gcn4p.

The binding of TBP and Pol II at the promoter increased at arelatively slow rate during the first 5 to 10 min of induction,such that both factors attained their half-maximal values $~$ 5to 10 min later than did the three coactivators. Similar kineticswere observed for the appearance of Pol II at the 3' end ofthe ARG1 ORF, indicating no detectable lag between PIC assemblyand transcription elongation. It may seem contradictory thatno recruitment of TBP was detected until 10 min of inductionwhile small increases in Pol II occupancy of the promoter andORF were reproducibly detected 5 min earlier (Fig. 1C and Dand 2B and C). This may be attributed to the fact that ChIPsignals were considerably lower for TBP versus Pol II (Fig.2B), so that low-level binding of TBP at 5 min was below thedetection limit of the assay.

Although recruitment of the coactivators, Pol II, and TBP allshowed significant increases by 5 to 10 min after inductionof Gcn4p, it appeared that TBP and Pol II required 5 to 10 minlonger than the coactivators to reach their maximum rates andpeak values of binding at ARG1. These kinetic data suggest thatrecruitment of coactivators and the stimulation of PIC formationby Gcn4p occur in two distinct stages, as depicted schematicallyin Fig. 6 (stages I and II). This conclusion is supported byour findings that SAGA and Mediator are required for wild-typePIC formation and transcription at the earliest time pointsof induction. Although SWI/SNF is not required for PIC formationduring the first 20 min, it is necessary for optimum transcriptionrates at the earliest time points. Importantly, we also showedrecently (31) that recruitment of SWI/SNF, SAGA, and Mediatorare unaffected by deletion of the TATA element at ARG1, whereasPIC assembly is severely impaired by this promoter mutation(30). Thus, taken together, our data strongly suggest that recruitmentof SWI/SNF, SAGA, and Mediator precedes PIC assembly at theARG1 promoter during induction by Gcn4p.

View larger version (26K):
[in this window]
[in a new window]
FIG. 6. A model for simultaneous interdependent recruitment of SAGA, SWI/SNF, and Mediator by Gcn4p and multiple functions of these coactivators in PIC assembly and elongation. Transcriptional activation is divided into the two stages that were temporally resolved in wild-type cells. The top two panels depict the first stage in transcriptional activation involving binding of Gcn4p to the UAS and the nearly simultaneous recruitment of SAGA, SWI/SNF, and Mediator. Even though binding of Gcn4p and recruitment of SAGA and SWI/SNF commence immediately on induction of Gcn4p synthesis, binding of Gcn4p is depicted as the first step because mutational inactivation of these coactivators (or of Mediator) does not affect Gcn4p binding to the UAS. Although recruitment of RSC is also depicted here, the kinetics of its recruitment have not been determined. Interdependencies among the coactivators are summarized using arrows color coded for the stimulatory factor. The bottom three panels depict the second stage in activation involving TBP recruitment, Pol II recruitment or promoter clearance downstream of TBP binding, and elongation. While these three steps were not temporally resolved, genetic analysis reveals that Pol II recruitment is dependent on TBP binding to the TATA element and that the coactivators independently stimulate these reactions in the manner depicted by the color-coded arrows. See text for further details.

We believe that Gcn4p binding to the UAS is the first eventin transcriptional activation of ARG1. Even though the recruitmentof SAGA and SWI/SNF begins immediately on induction of GCN4translation, Gcn4p is the first factor to reach its maximumlevel of binding. Furthermore, none of the coactivators canbe recruited to ARG1 in gcn4 ${Delta}$ cells (37) (Fig. 1), whereas mutationalimpairment of the coactivators did not significantly alter therate or extent of Gcn4p binding to the UAS (Fig. 3). Thus, wepropose that Gcn4p binding to the UAS precedes the recruitmentof these coactivators at ARG1 (Fig. 6, step 1).

The fact that recruitment of the three coactivators occurs almostsimultaneously could be taken as an indication that Gcn4p independentlyrecruits each coactivator. Indeed, recombinant Gcn4p can bindspecifically to all three coactivators in vitro (11, 23, 29,39). Consistent with this idea, we found that inactivating SWI/SNFhad only a slight impact on the kinetics and end point of SAGArecruitment and did not affect the rate of Mediator recruitmentto ARG1. Similarly, Mediator recruitment was nearly unaffectedby disrupting SAGA integrity with the ada1 ${Delta}$ mutation. BecauseMediator recruitment to ARG1 is independent of both SWI/SNFand SAGA, the interaction of Mediator with the Gcn4p activationdomain may be sufficient for its wild-type recruitment to theUAS at ARG1 (Fig. 6, step 2).

In contrast to the independent recruitment of Mediator, therecruitment of both SWI/SNF and SAGA by Gcn4p is criticallydependent on Mediator (Fig. 6, step 2), being severely impairedin the rox3 ${Delta}$ mutant at the earliest time points. This last conclusionmay seem at odds with our finding that recruitment of Mediatorseemed to lag slightly behind that of SWI/SNF (and possiblySAGA) during the first few minutes of induction in wild-typecells (Fig. 2D to F). However, the basal level of Mediator recruitmentprovided by the uninduced level of Gcn4p may support the earliestphase of SWI/SNF and SAGA recruitment by the Gcn4p that is inducedby amino acid starvation.

We recently compared the effects of coactivator mutations onthe steady-state recruitment of SWI/SNF, SAGA, and Mediatorat three different Gcn4p target genes, ARG4, SNZ1, and ARG1.Recruitment of SAGA and SWI/SNF after 2 h of starvation wasfound to be strongly dependent on Mediator at all three genes,in accordance with the findings presented here for ARG1. Thus,the importance of Mediator in recruitment of SWI/SNF and SAGAmay be a general attribute of Gcn4p. However, the ability ofGcn4p to recruit Mediator independently of SAGA observed atARG1 (in both studies) did not extend to ARG4 and SNZ1 (31).Hence, it is possible that SAGA and Mediator are mutually interdependentfor their recruitment by Gcn4p at many target genes and thatsome feature of the ARG1 promoter, or another transcriptionfactor that binds at ARG1, allows for SAGA-independent recruitmentof Mediator by Gcn4p at this gene.

Our findings that Mediator has an important function in promotingSAGA recruitment by Gcn4p differs from previous results obtainedfor Gal4p. No reduction in SAGA recruitment to GAL1 was producedby mutations in Mediator subunits srb4-ts (3) or gal11 ${Delta}$ (6),even though srb4-ts abolishes PIC formation at GAL1 (26). Acritical requirement for Mediator in SWI/SNF recruitment byGcn4p is also distinctive, in that recruitment of SWI/SNF tothe HO gene by Swi5p seems to be independent of Mediator (4).Although SWI/SNF recruitment by Gal4p is dependent on Mediator,it additionally requires PIC formation at GAL1 (24), whereaswe found that SWI/SNF recruitment is independent of the TATAbox and PIC formation (31). Thus, Mediator likely plays a moredirect role in promoting SWI/SNF recruitment by Gcn4p than forGal4p.

Recruitment of SWI/SNF by Gcn4p is dependent on SAGA in additionto Mediator (Fig. 6, step 2), and our finding that SWI/SNF recruitmentwas impaired in gcn5 ${Delta}$ cells has important implications for thehistone code hypothesis (36). Previous results of in vitro experimentsindicated that histone acetylation by SAGA enhances recruitmentof SAGA and SWI/SNF via the bromodomains in Gcn5p and Snf2p(15). However, recruitment of SWI/SNF in vivo by the activatorsSwi5p (9) and Gal4p is independent of Gcn5p HAT function andthe integrity of SAGA (24). Moreover, deletion of Gcn5p reducedSWI/SNF recruitment only in a narrow time window during SUC2induction (13). In contrast to these findings, we found thatgcn5 ${Delta}$ impairs SWI/SNF recruitment by Gcn4p during most of theinduction period, but especially at the time of maximum SWI/SNFrecruitment. This result provides the strongest evidence todate that histone acetylation by Gcn5p enhances associationof SWI/SNF with a UAS element in vivo.

The reason why we previously failed to notice the dependenceon Gcn5p for steady-state SWI/SNF recruitment (43) may be thatGcn4p was being overexpressed in our former study. The greater-than-wild-typelevel of Gcn4p binding to the UAS may have suppressed the reductionin SWI/SNF recruitment that was observed here at 120 min ofinduction in gcn5 ${Delta}$ cells with native levels of Gcn4p. Furthermore,the fact that gcn5 ${Delta}$ produced no decrease in SWI/SNF recruitmentto a synthetic PHO5 promoter containing Gcn4p binding sites(38) suggests that the stimulatory effect of histone acetylationon SWI/SNF recruitment can also vary with the chromatin structureof the promoter.

Our finding that SAGA recruitment was not diminished by gcn5 ${Delta}$ (Fig. 3C) seems at odds with the idea that the Gcn5p bromodomainenhances recruitment of SAGA to acetylated promoter histonesin vitro (15). Similarly, Syntichaki et al. observed that deletionof the Gcn5p bromodomain did not diminish Gcn5p-dependent histoneacetylation at the synthetic UAS_Gcn4-PHO5 promoter (38). Thus,there is no direct evidence that SAGA recruitment is enhancedby binding of the Gcn5p bromodomain to acetylated histone tailsin vivo. In fact, our finding that SAGA recruitment was elevatedin gcn5 ${Delta}$ cells raises the possibility that H3 acetylation isrequired for release of SAGA from the UAS during induction byGcn4p.

Comparing the effects of ada1 ${Delta}$ and gcn5 ${Delta}$ mutations on the kineticsof SWI/SNF recruitment indicates that a function of SAGA apartfrom Gcn5p HAT activity is required for optimal SWI/SNF recruitment,consistent with our earlier observations concerning steady-stateinduction by Gcn4p (43). It is currently unknown how SAGA supportsSWI/SNF recruitment by Gcn4p independently of histone acetylationby Gcn5p. In the case of Gal4p and the activator responsiblefor RNR3 induction, the recruitment of SWI/SNF is dependenton PIC assembly (24, 34). However, because SWI/SNF recruitmentto ARG1 is unaffected by deletion of the TATA element (31),we conclude that stimulating PIC formation does not representthe non-HAT SAGA function that enhances SWI/SNF recruitmentby Gcn4p. Furthermore, because ada1 ${Delta}$ produced only a small reductionin Mediator recruitment to ARG1, it seems unlikely that thedecrease in SWI/SNF recruitment at this gene is an indirectconsequence of impaired Mediator recruitment in ada1 ${Delta}$ cells.

Based on our results with the snf2 ${Delta}$ mutant, we surmised thatchromatin remodeling is dispensable for SAGA recruitment. However,surprisingly, high-level SAGA recruitment by Gcn4p is dependenton RSC (Fig. 6, step 2). To our knowledge, this is the firstevidence that chromatin remodeling by RSC stimulates the recruitmentof a coactivator. Presumably, a basal amount of RSC recruitedby the uninduced level of Gcn4p contributes to the immediate,Rsc2p-dependent increase in SAGA recruitment that occurs oninduction of Gcn4p. Interestingly, rsc2 ${Delta}$ led to elevated SWI/SNFrecruitment, suggesting that these two chromatin remodelingfactors compete for binding sites at ARG1. Given that Rsc2pand Snf2p both contain bromodomains, rsc2 ${Delta}$ may eliminate competitionbetween RSC and SWI/SNF for binding to acetylated nucleosomesas a means of increasing SWI/SNF recruitment.

The chromatin remodeling activity of SWI/SNF is dispensablefor recruitment of both Mediator and SAGA to ARG1 over the timecourse of induction. Similar observations have been made forthe ARG4 and SNZ1 genes at steady state (31). In fact, we foundthat snf2 ${Delta}$ impairs only Pol II recruitment and elongation atARG1. These findings stand in sharp contrast to the requirementfor SWI/SNF in recruitment of both SAGA (9) and Mediator tothe HO gene by Swi5p (4). As noted above, this requirement seemsto reflect the condensed state of chromatin during mitosis andalso applies to Gcn4p and Gal4p during mitosis (19).

Functions of coactivators in Pol II recruitment and elongation. A key function of coactivators is to promote PIC assembly. Ourkinetic analysis of coactivator mutants indicates that RSC,SAGA and Mediator all stimulate recruitment of TBP to the promoter(Fig. 6, step 3), as the rsc2 ${Delta}$ , ada1 ${Delta}$ , and rox3 ${Delta}$ mutations producedquantitative reductions in TBP binding to the TATA element atvarious times during the induction of ARG1. These findings arein general agreement with our previous analysis of steady-stateTBP binding to the ARG1 promoter (30). However, the strongerreductions in TBP binding observed in that study may be attributableto the fact that we analyzed strains containing a myc₁₃ tagon the C terminus of TBP, which slightly impairs TBP functionand may exacerbate the effects of coactivator mutations on TBPrecruitment.

Even though we cannot temporally resolve the recruitment ofTBP and Pol II to the promoter in wild-type cells, we believethat Pol II recruitment follows TBP binding (Fig. 6, steps 3and 4) because deletion of the TATA element nearly abolishedPol II recruitment at ARG1 (30). Therefore, any coactivatormutations that impair TBP recruitment by Gcn4p should indirectlyreduce Pol II recruitment. In accordance with this prediction,the rsc2 ${Delta}$ , ada1 ${Delta}$ , and rox3 ${Delta}$ mutations all reduced promoter occupancyby Pol II at ARG1. However, these three mutations, along withsnf2 ${Delta}$ and gcn5 ${Delta}$ , reduced Pol II recruitment during certain phasesof ARG1 induction where there was little or no defect in TBPbinding. One interpretation of these last findings is that SWI/SNF,RSC, SAGA, and Mediator promote Pol II recruitment by a mechanismdistinct from a stimulatory effect on TBP binding (Fig. 6, step4). This uncoupling of Pol II recruitment from TBP binding wasparticularly obvious for the snf2 ${Delta}$ and gcn5 ${Delta}$ mutants in whichTBP recruitment occurred at wild-type or higher levels throughoutthe time course, while Pol II promoter occupancy was impairedduring the later stages of induction. We made a similar findingpreviously for the srb10 ${Delta}$ mutation, which decreased steady-statepromoter occupancy by Pol II at ARG1, ARG4, and SNZ1 withoutreducing TBP recruitment by Gcn4p (30).

In vitro studies by Hahn and colleagues indicate that followinginitiation of transcription most of the general factors andMediator remain bound to the promoter in a Scaffold complex,which facilitates the efficient recruitment of Pol II, TFIIB,and TFIIF in subsequent rounds of transcription (32, 44). BecauseTBP is a component of the Scaffold, the functions of coactivatorsin recruiting Pol II to the Scaffold would necessarily be downstreamof TBP recruitment. Thus, the defects in Pol II occupancy thatoccur without reductions in TBP binding at ARG1 could reflectrequirements for Mediator, SAGA, SWI/SNF, or RSC in reinitiationof transcription by the Scaffold complex in vivo.

An alternative interpretation for this phenotype, however, isa defect in promoter clearance following PIC assembly. Whileit might be expected that a failure to make the transition frominitiation to elongation would lead to accumulation of Pol IIat the promoter, the kin28-ts16 mutation in the CTD kinase activityof TFIIH strongly reduces Pol II promoter occupancy withoutaffecting TBP binding (27, 33). This result may indicate thatPol II stalled at the promoter in kin28 cells dissociates rapidly;alternatively, it is possible that most of the ChIP signal forPol II at promoters derives from early elongation complexeswhich are eliminated by kin28-ts16 (27). In any event, becausethe coactivator mutations we analyzed produce the same phenotypedescribed for kin28-ts16 cells, they may impair promoter clearancerather than Pol II recruitment (Fig. 6, step 4).

The physical association between Mediator and Pol II could underliean adaptor mechanism for direct recruitment of Pol II or providea means of stimulating promoter clearance via CTD phosphorylation(28). Alternatively, Mediator could act indirectly in thesefunctions by stimulating recruitment of SAGA or SWI/SNF. Anadaptor function for SAGA in Pol II recruitment is possible,but most previous work describes its interactions with TBP (12,22). The similar effects of gcn5 ${Delta}$ and snf2 ${Delta}$ , and the fact thatGcn5p promotes SWI/SNF recruitment (this study) and remodelingfunction (38), suggest that the SAGA HAT may indirectly stimulatePol II recruitment or promoter clearance by enhancing SWI/SNFactivity. Nucleosome remodeling of the promoter downstream ofthe TATA element by SWI/SNF (or RSC) could enhance Pol II recruitmentor its escape from the promoter.

Finally, our kinetic analysis of coactivator mutants has uncoveredfunctions for SWI/SNF and SAGA in stimulating elongation byPol II (Fig. 6, step 5). Thus, we found that snf2 ${Delta}$ and gcn5 ${Delta}$ produceda strong reduction in Pol II occupancy in the 3' end of theORF early in the induction period ( $~$ 20 min) at a time when TBPand Pol II were present at the promoter at nearly wild-typelevels. In addition, because ada1 ${Delta}$ produced a much stronger reductionin transcription than did rsc2 ${Delta}$ , despite comparable reductionsin Pol II promoter occupancy, it appears that a non-HAT functionof SAGA is also required for elongation. It might be expectedthat defects in elongation would lead to a queuing of Pol IImolecules in the beginning of the ORF and a higher-than-wild-typePol II occupancy in the promoter, which were not observed. Itis possible, however, that stalled Pol II molecules are cleared,e.g., by elongation factor TFIIS (Dst1p), or that the predictedaccumulation of Pol II in the promoter is offset by reductionsin Pol II recruitment produced by the coactivator mutations.

It was shown previously that activators can stimulate elongationas well as initiation (1, 5). In the case of mammalian activatorHSF1, there is strong circumstantial evidence that nucleosomeremodeling by SWI/SNF, recruited by HSF1, is instrumental inovercoming an elongation barrier in the 5' end of the hsp70gene (reference 8 and references therein). Our findings on Gcn4pprovide the first direct evidence that SWI/SNF and SAGA arerequired for a wild-type efficiency of elongation in vivo. Theseresults are consistent with genetic interactions in yeast betweenTFIIS and subunits of SWI/SNF (10) and with genetic and physicalinteractions between SAGA subunit Spt8p and TFIIS (40). Moreover,the recruitment of SWI/SNF by Gcn4p to plasmid-borne HIS3 producesa labile chromatin domain that extends from the promoter intothe coding sequences (18). Thus, chromatin remodeling of theORF by SWI/SNF may account for its stimulatory role in elongationat ARG1. Enhancing the recruitment or remodeling function ofSWI/SNF could, in turn, explain the positive role of SAGA inelongation apart from the Spt8p-TFIIS interaction.

In summary, our findings reveal a strong dependence of Gcn4pon Mediator for recruitment of SAGA and SWI/SNF, on Gcn5p anda non-HAT function of SAGA in SWI/SNF recruitment, and on RSCfor recruitment of SAGA to ARG1. This network of interdependencies(Fig. 6, step 2) represents one of the most extensive set ofcooperative interactions governing coactivator recruitment describedthus far, and it differs in important respects from that describedfor other yeast activators. The kinetic analysis of PIC assemblyand transcription elongation in the coactivator mutants hasallowed us to implicate SAGA, SWI/SNF, RSC, and Mediator instimulating Pol II recruitment or promoter clearance downstreamof TBP binding, and we uncovered functions for SAGA and SWI/SNFin stimulating Pol II elongation apart from their known activitiesin PIC assembly.

ACKNOWLEDGMENTS

We are very grateful to Joe Reese, Mark Ptashne, and KrishnamurthyNatarajan for generous gifts of antibodies.

FOOTNOTES

* Corresponding author. Mailing address: NIH, Building 6A, Room B1A-13, Bethesda, MD 20892. Phone: (301) 496-4480. Fax: (301) 496-6828. E-mail: ahinnebusch{at}nih.gov.

Present address: National Cancer Center, 809 Madul-Dong, Ilgan-Gu,Goyang-Si, Gyeonggi-do 411-769, Republic of Korea.

REFERENCES

Top