SAGA and Rpd3 Chromatin Modification Complexes Dynamically Regulate Heat Shock Gene Structure and Expression*

The chromatin structure of heat shock protein (HSP)-encoding genes undergoes dramatic alterations upon transcriptional induction, including, in extreme cases, domain-wide nucleosome disassembly. Here, we use a combination of gene knock-out, in situ mutagenesis, chromatin immunoprecipitation, and expression assays to investigate the role of histone modification complexes in regulating heat shock gene structure and expression in Saccharomyces cerevisiae. Two histone acetyltransferases, Gcn5 and Esa1, were found to stimulate HSP gene transcription. A detailed chromatin immunoprecipitation analysis of the Gcn5-containing SAGA complex (signified by Spt3) revealed its presence within the promoter of every heat shock factor 1-regulated gene examined. The occupancy of SAGA increased substantially upon heat shock, peaking at several HSP promoters within 30–45 s of temperature upshift. SAGA was also efficiently recruited to the coding regions of certain HSP genes (where its presence mirrored that of pol II), although not at others. Robust and rapid recruitment of repressive, Rpd3-containing histone deacetylase complexes was also seen and at all HSP genes examined. A detailed analysis of HSP82 revealed that both Rpd3(L) and Rpd3(S) complexes (signified by Sap30 and Rco1, respectively) were recruited to the gene promoter, yet only Rpd3(S) was recruited to its open reading frame. A consensus URS1 cis-element facilitated the recruitment of each Rpd3 complex to the HSP82 promoter, and this correlated with targeted deacetylation of promoter nucleosomes. Collectively, our observations reveal that SAGA and Rpd3 complexes are rapidly and synchronously recruited to heat shock factor 1-activated genes and suggest that their opposing activities modulate heat shock gene chromatin structure and fine-tune transcriptional output.

The heat shock response is a regulated transcriptional response to elevated temperature and other environmental insults. It is essential for the viability of all organisms. In mammals, physiological stresses such as fever, inflammation, infection, ethanol toxicity, and tissue ischemia are countered by the heat shock response, which is regulated by heat shock factor 1 (Hsf1), 2 an evolutionarily conserved, trimeric transcriptional activator (reviewed in Ref. 1). Concomitant with its conferring protection from environmental stress, Hsf1 plays important roles in suppressing neurodegeneration (2) and in enhancing carcinogenesis (3). In many organisms, including Homo sapiens, Drosophila melanogaster, and Saccharomyces cerevisiae, Hsf1 is localized within the nucleus under both control and stressful conditions (4 -6). In the budding yeast S. cerevisiae, Hsf1 binds its high affinity heat shock-response elements (HSEs) under noninducing conditions (7), from which it locally opens chromatin structure and promotes constitutive transcription (8 -11). In addition, the protein inducibly binds low affinity HSEs, contributing to the substantial increase in target gene transcription that occurs in response to stress (4,(12)(13)(14)(15). Interestingly, the ability of Hsf1 to open promoter chromatin is conserved in human cells (16). Genome-wide localization analysis in S. cerevisiae has identified ϳ165 genes as in vivo targets of heat shock-activated Hsf1 (15). These genes encode proteins involved in surprisingly diverse functions, including protein folding, degradation, and trafficking; energy generation; maintenance of cell integrity; cell signaling; and transcription.
S. cerevisiae serves as an excellent model system to study regulation of the heat shock response. Yeast is genetically tractable and can survive in variable environments, and many fundamental mechanisms of the stress response are conserved with higher eukaryotes. Yeast Hsf1 is unusual among activators in that it can stimulate significant levels of basal transcription. Moreover, it can bypass the requirement for a number of general transcription factors and co-activators in stimulating activated transcription. Such bypassed factors include TFIIA, Taf9 (constituent of both TFIID and SAGA), Kin28 (TFIIH kinase), and the Med17/Srb4 and Med22/Srb6 subunits of Mediator (17)(18)(19)(20). By contrast, TFIIE and the chromatin remodeling enzyme Swi/Snf are critically required for Hsf1mediated transactivation (21)(22)(23)(24).
One of the gene targets of yeast Hsf1, HSP82, has been particularly well studied (reviewed in Refs. 25,26). The HSP82 promoter is assembled into an accessible, nucleosome-free chromatin structure established and maintained by Hsf1, which constitutively binds a high affinity HSE within the UAS. Upon heat shock, Hsf1 cooperatively and inducibly binds to two additional, low affinity HSEs; concomitant with this is a significant increase in both TBP and pol II binding and a 10 -20-fold increase in transcription (12)(13)(14). Accompanying these dynamic alterations in factor-DNA interactions, histones are rapidly evicted not only from the promoter and upstream region but throughout the coding region of the gene (22). Interestingly, this domain-wide nucleosomal disassembly at HSP82 can take place independently of Swi/Snf (22), although the kinetics of histone eviction are delayed in Swi/Snf mutants (23,24).
Mediator is a second transcriptional co-regulator implicated in yeast heat shock gene regulation. Mediator integrates signals from sequence-specific activators and repressors and transduces them to the general transcriptional machinery (reviewed in Ref. 27). A screen for extragenic suppressors of a transcription defect in the hsp82-⌬HSE1 gene, lacking the high affinity Hsf1 site, uncovered recessive mutations in six Mediator subunits (4). Of these, five mapped either to its Middle module (Med10/Nut2, Med7, and Med21/Srb7) or to interfaces between the Middle/Head or Middle/Tail modules (Med19/Rox3 and Med14/Rgr1, respectively). Interestingly, these loss-of-function mutations enhanced the noninduced expression of wild-type (WT) heat shock genes while simultaneously limiting the extent of induced expression (4). These observations together with earlier ones (28) indicate that the Middle module of Mediator plays an important role in both constitutive and activated heat shock gene transcription.
Here, we use a combination of gene knock-out, in situ mutagenesis, ChIP, and expression assays to investigate the role played by histone modification complexes, particularly histone acetyltransferases (HATs) and histone deacetylase complexes, in regulating HSP82 and other Hsf1 target genes. A potential role for these complexes has been suggested by several previous studies. For example, a large increase in nucleosomal histone acetylation was observed within the promoters of certain Hsf1-regulated genes (SSA3 and CUP1) in response to a 20-min heat shock (29). More recent work has shown that core histones are transiently hyperacetylated within the promoter regions of HSP12, HSP26, HSP82, and SSA4 in response to acute thermal stress (22,30). By contrast, when cells are exposed to osmotic stress, HSP12 nucleosomes are deacetylated concomitant with a large increase in transcription (31). Thus, the precise roles played by HATs and histone deacetylase complexes in HSP gene regulation are unclear and prompted the investigation we report here.

EXPERIMENTAL PROCEDURES
Yeast Strain Construction-Genotypes of strains used in this study are provided in Table 1. Construction of DSG101, harboring the hsp82-P2 allele, has been described (13). Construction of DSG125 (hsp82-⌬URS1) was done analogously, in which a combination of oligonucleotide-directed mutagenesis and two-step gene transplacement were used to replace the URS1 sequence TGAGCGGTTA (spanning Ϫ195 to Ϫ186 with respect to the ATG start codon) with GCTTATCGAT (sequence confirmed by genomic PCR fol-lowed by DNA sequencing). The latter sequence has been shown to lack promoter activity in yeast (32). Strains SBK900, SBK901, and SBK902 were created by transforming the HSP82 strain EAS2001 with pRS316, pGAL-ESA1, and pGAL-ESA1 E338Q , respectively. The pGAL-ESA1 and pGAL-ESA1 E338Q plasmids were kindly provided by Shelley Berger (Wistar Institute). To permit expression of Myc-tagged histone H4, strains SLY101 and CBY152 were transformed with pNOY436 (Myc-HHF2-TRP1-CEN6-ARS4).
S288C knock-out strains spt20⌬::KAN-MX and set2⌬::KAN-MX were obtained from the ResGen strain collection (gift of Kelly Tatchell). TAP-tagged RCO1, SAP30, SIN3, and SPT3 strains were purchased from Open Biosystems (Huntsville, AL). TAP-tagged strains were evaluated for growth phenotypes and found in each case to grow identically to the nontagged isogenic control. SBK903 and SBK904 were constructed by allelic replacement of SAP30 in strains SLY101 and DSG125, respectively, with a DNA fragment comprising SAP30-TAP:: HIS3 (PCR amplified from genomic DNA isolated from the S288C derivative, TAP-SAP30). SBK905 and SBK906 were constructed by allelic replacement of RCO1 with RCO1-TAP::HIS3 in an analogous manner. Each amplicon was purified by Zymo-Clean (ZymoResearch, Inc.) and sequenced.
Cultivation, Heat Shock, and Recovery Conditions-Typically, yeast strains were cultivated at 30°C to early log phase (A 600 ϭ 0.3 to 0.6) in rich YPD broth supplemented with 0.03 mg/ml adenine. Heat shock induction was achieved by an instantaneous 30 -39°C upshift by addition of an equal volume of pre-warmed (57°C) medium to the culture (typically 50 ml (Northern) or 100 ml (ChIP)) in a vigorously shaking 39°C water bath for the times indicated. To attain an instantaneous 39 -23°C downshift, 20-min heat-shocked cultures were mixed with an equal volume (typically 200 ml) of pre-chilled (4°C) medium and then transferred to a shaking 23°C water bath for 20 min. To terminate heat shock induction, either sodium azide was added to a final concentration of 10 mM (Northern) or formaldehyde to a 1% concentration (ChIP) to each culture.
To evaluate the role of Esa1, HSP82 gene expression was measured in an ESA1 ϩ strain containing either a galactoseinducible ESA1 ϩ or ESA E338Q episomal gene. Cultures were grown at 30°C in SDC-Ura to A 600 ϭ 0.4 and split into 2 aliquots. One aliquot was reserved as the glucose control; the other was washed twice with sterile water, suspended in SGC-Ura, and grown at 30°C for an additional 12 h. The galactoseinduced cultures were split into 2 aliquots as follows: (i) nonheat shock (NHS), subjected to additional shaking at 30°C for 20 min; and (ii) heat shock (HS), subjected to instantaneous 30 -39°C shift for 20 min by the addition of pre-warmed (57°C) medium.
Expression Analysis-␤-Galactosidase assays were performed as described previously (4). For Northern analysis, RNA was isolated, purified, electrophoresed, blotted to Gene-Screen, and sequentially hybridized to HSP82-and SCR1-specific probes as described previously (22) except ImageQuant TL (version 2003.02; Amersham Biosciences) was used to quantify phosphor images.
Chromatin Immunoprecipitation-ChIP was performed essentially as described (22). Immunoprecipitations (IPs) were achieved by adding the following antibodies to 300 l of sonicated chromatin lysate: 3 l of histone antiserum, 3 l of Myc monoclonal antibody 9E10 (Santa Cruz Biotechnology), 4 l of TAP antiserum (Open Biosystems, Huntsville, AL), 5 l of Hsf1 antiserum (11), or 3 l of pol II C-terminal domain antiserum (33). Histone antibodies specific for the following epitopes were obtained from Millipore Corp. (Temecula, CA): histone H3 (unacetylated isoform, catalog number 06-755) and di-acetyl histone H3 (Lys-9 and Lys-14) (catalog number 06-599). Following IP, reversal of protein-DNA cross-links, and purification of DNA, 2 l (out of 30 l total) was used as template in a multiplex PCR. For Hsf1-immunoprecipitated DNA, 1 l was used as template. Histone and Hsf1 IP DNA templates were subjected to 23 cycles of a program consisting of 1 min at 93°C, 1 min at 60°C, and 34 s at 72°C. TAP and pol II IP DNA samples were subjected to 28 cycles. PCR products were precipitated, electrophoresed on 7% TBE polyacrylamide gels, dried, exposed to a Phosphor Screen, and quantified on a Storm 860 PhosphorImager utilizing ImageQuant TL version 2003.02 software. Linearity of the PCR was tested by assaying serial dilutions of individual templates.
For histone ChIPs, mock IP signal was subtracted from each amplified locus. Differences in PCR amplification efficiency of one genomic locus relative to another were normalized using the following formula: Q locus ϭ IP locus /Input locus . Abundance of each histone isoform at a given locus was quantified relative to its abundance at the PHO5 promoter, which served as an internal recovery control (14), namely Q locus /Q PHO5 . (The PHO5 promoter is composed of a well characterized array of nucleosomes whose expression is not affected by heat shock.) For TAP ChIPs, an equivalent volume of TAP antibody was added to the nontagged control strain to obtain the mock IP signal. The mock signal was subtracted from each specific IP, and gel background (a much smaller value) was subtracted from ARS504. Abundance of the TAP-tagged factor was calculated (ϭ Q locus / Q ARS504 ). For Hsf1 ChIPs, gel background was subtracted from each IP signal prior to calculating Q locus . In the Hsf1 ChIPs, primers for HSP82 3Ј-UTR and PHO5 promoter were included in the multiplex PCRs as negative controls, but the corresponding amplified loci were not included in the quantification because the signal arising from each was negligible (for example, see Fig. 1C). Instead, the Q promoter value for the WT strain under NHS conditions was set at 1.0, with the other combinations of strain/cultivation state normalized relative to it. For pol II ChIPs, pre-immune signal was subtracted from each specific IP signal, and gel background (a much smaller value) was subtracted from ARS504 prior to calculation of pol II abundance (ϭ Q locus /Q ARS504 ) In Fig. 9, B and D, the ratio of the abundance of diacetylated H3 was quantified relative to total H3 present in the same chromatin samples.

RESULTS
Experimental Strategy-To identify co-regulators of Hsf1, we investigated the consequences of deleting candidate factors on the transcription of wild-type HSP82 and a well characterized promoter mutant, hsp82-P2 (8,13). hsp82-P2 bears a doublepoint mutation within the high affinity heat shock element (illustrated in Fig. 1A) that results in a 10-fold reduction in basal transcription and a 3-fold loss in activated transcription (Fig. 1B). In addition, Hsf1 binding to the mutated UAS HS is reduced ϳ10-fold under NHS conditions and ϳ2-fold under 15 min of HS-induced conditions (Fig. 1C, compare lanes 7 and 8 with lanes 3 and 4; results are quantified in the histogram). Thus, the 2-bp P2 mutation does not alter the fundamental interaction of Hsf1 with the hsp82 promoter; the protein is detectable under NHS conditions, and its binding is enhanced upon HS and reduced upon attenuation (chronic HS). Yet, its abundance relative to that seen at HSP82 ϩ is reduced under all three states. Given this, we reasoned that hsp82-P2 might facilitate identification of the co-activators of Hsf1 because these factors would be less likely functionally redundant, and their role in stimulating Hsf1-activated transcription therefore more apparent.
Set1 and Set2 Are Dispensable for Normal Levels of HSP82 Transcription-Current models suggest important roles for both Set1 and Set2 histone methyltransferases in regulating pol II transcription. Therefore, we began our investigation by examining their roles in HSP82 expression. Set1 is the catalytic subunit of COMPASS and is responsible for the mono-, di-, and tri-methylation of H3 K4. H3 K4 tri-methylation correlates with pol II gene activation in yeast (34,35) and is found at the transcription start sites of a majority of human protein-encoding genes (36). Set2 methylates H3 Lys-36 within nucleosomes of active gene coding regions where it physically associates with Ser-2-phosphorylated, elongating pol II (37). Di-and tri-methylated H3 Lys-36 in turn recruits the Rpd3(S) complex to locally deacetylate histones and thereby repress synthesis of cryptic transcripts (38 -40). Surprisingly, despite these important in vivo activities, the transcriptional output of both HSP82 and hsp82-P2 is largely unaffected by deletion of either SET1 or SET2 (Fig. 2, A and B). The dispensability of Set1 is consistent with the absence of a detectable net increase in H3 K4 methyl-FIGURE 1. A 2-bp mutation within HSE1 increases the dynamic range of hsp82 expression as well as Hsf1 promoter occupancy. A, HSP82 promoter sequence, numbered relative to the ATG start codon (ϩ1), and the P2 and ⌬URS1 promoter mutations evaluated in this study. ARE, ancillary repression element; STRE, stress-response element. B, Northern analysis of HSP82 and hsp82-P2 (strains SLY101 and DSG101, respectively; depicted are means Ϯ S.E. of four independent assays). NHS, cells maintained at 30°C; HS, cells shifted from 30 to 39°C for 20 min. Transcript levels of HSP82 were assigned as 100 for NHS and 2000 for HS (13,14,22); hsp82-P2 mRNA levels are normalized relative to those of HSP82. C, in vivo cross-linking analysis of Hsf1 at HSP82 and hsp82-P2 at the indicated times following heat shock. Depicted is a PAGE analysis of multiplex PCRs of ChIPs using an Hsf1-specific polyclonal antibody. Input, DNA equivalent to 1% of the sonicated chromatin used in each IP sample (lanes 3-5 and 7-9). Mock, immunoprecipitation of an equivalent aliquot of chromatin mediated by Pansorbin cells alone. Depicted below is a bar graph summary of relative Hsf1 abundance with vertical lines representing the standard error; n ϭ 4. Note that a level of 1.0 is Ͼ50-fold higher than the amount detected at either the PHO5 promoter or the HSP82 3Ј-UTR, neither of which bears HSEs.
ation within the HSP82 promoter and ORF nucleosomes during heat shock induction (22).
Gcn5 and Esa1 HATs Play Important Roles in Regulating HSP82 Transcription-To investigate a potential role for HATs in regulating HSP82, we used mutagenesis to abolish or compromise their function. We initially evaluated the role of Gcn5, the catalytic subunit of SAGA and SAGA-like co-activator complexes (41,42), as well as the functionally distinct ADA complex (43). Gcn5 preferentially acetylates H3 in vitro (44) and H3 and H2B in vivo (45), and the SAGA complex has been shown to be a versatile co-activator of gene transcription in yeast (46). Indeed, in cells lacking Gcn5, activated hsp82-P2 transcription is reduced Ͼ10-fold; in addition, basal HSP82 transcription is reduced 5-fold, although induced HSP82 tran-script levels are not significantly affected (Fig. 2, A and B). Consistent with a role for SAGA, deletion of SPT20, encoding a SAGA-specific structural subunit, has a comparable effect to deletion of GCN5 on HSP82 transcription ( Fig. 2A). By contrast, deletion of SAS3, which encodes a functionally related, H3-specific HAT (47,48), has little or no effect on the transcription of either hsp82 allele (Fig. 2, A and B). Likewise, depletion of Sas3 has no effect on the previously characterized domain-wide displacement of nucleosomes (signified by loss of histone H4) at the heat shock-induced HSP82 gene (Fig. 2C) (22).
We next tested the role of Esa1, the catalytic subunit of NuA4, in HSP82 transcriptional regulation. Esa1 has previously been shown to regulate the transcription of both constitutively and inducibly regulated genes (49). In vitro, it preferentially acetylates the N termini of H4, H2A, and the H2A variant, Htz1 (50 -52). Interestingly, both H2A and H4 are rapidly acetylated within the HSP82 promoter in response to instantaneous heat shock (22). As Esa1 is encoded by an essential gene, it could not be deleted. Instead, we assessed its role in regulating HSP82 by overexpressing the dominant negative Esa1 E338Q allele. The E338Q point mutation maps to a domain in Esa1 that is conserved among HATs, including Gcn5/PCAF, and disrupts Esa1 catalytic activity (53). In cells grown in medium containing 2% galactose, HSP82 transcription is induced 8-fold upon acute heat shock (Fig. 2D, vector alone). Overexpression of Esa1 ϩ slightly diminishes HSP82 transcript levels, whereas overexpression of Esa1 E338Q substantially decreases gene induction (Fig. 2D, white bars; p Ͻ 0.0002, two-tailed t test). These results imply a role for Esa1, and by extension NuA4, in activating HSP82 and suggest the possibility that NuA4 acts directly to regulate Hsf1 target genes. Consistent with this idea, ChIP assays have demonstrated Esa1 recruitment to the promoter regions of heat shock-induced SSA3, SSA4, and HSP104 (49,54).
SAGA Occupies Both the Promoter and the Coding Region of HSP82 upon Heat Shock-We turned our focus to SAGA and asked whether it physically interacts with the HSP82 upstream region. Previous work has suggested that this complex is recruited to gene promoters via direct interaction with DNAbound activators (55,56) and that this is mediated by the Tra1 subunit of SAGA, both in vivo and in vitro (57)(58)(59). Consistent with this, SAGA has been detected within the UAS/promoter regions of a number of activated yeast genes (54,60). Given the importance of SAGA in sustaining the basal transcription of HSP82 (Fig. 2, A and B), we hypothesized that it would be present within the uninduced promoter where it could contribute to the hyperacetylated and DNase I-hypersensitive state of the chromatin (7,9,14,61) (see also Fig. 9 below). Indeed, SAGA (signified by its conserved Spt3 subunit) is present at the uninduced HSP82 promoter (Fig. 3A, 0 min, solid bar). Moreover, following heat shock, SAGA abundance is rapidly, and significantly, increased within the first 45 s (p Ͻ 0.008). This increase correlates with the rapid net increase in H2A, H3, and H4 acetylation seen in response to heat shock (22,30). Interestingly, enhanced levels of SAGA persist at the promoter for at least 15 min, a time when nucleosomes are largely displaced (Fig. 3C). This observation suggests that the role of SAGA in regulating HSP82 may extend beyond acetylation of histones.
In addition to increased occupancy within the promoter under heat-inducing conditions, Spt3 notably occupies the HSP82 ORF and 3Ј-UTR (Fig. 3A, striped and speckled bars), and this is detectable within the first 30 -45 s. Interestingly, the 5Ј-to-3Ј occupancy of SAGA along the gene is sequential, as indicated by the kinetics of Spt3 accumulation at the promoter, ORF, and 3Ј-UTR (Fig. 3A, also compare D-F, red curve). The delayed kinetics of occupancy by Spt3 of the 3Ј-UTR relative to its occupancy at the promoter argues that its cross-linking to the downstream positions is not a consequence of gene looping, which has been observed at a number of activated yeast genes and which may occur at HSP82 as well. Instead, comparison of the kinetics of Spt3 occupancy with those of pol II (Fig. 3, A and   B) suggests that, among other possibilities, SAGA might travel with elongating pol II or be recruited by the elongation complex (discussed further below).
Interestingly, nucleosomal abundance (as signified by histone H3) inversely correlates with the abundance of both SAGA and pol II (Fig. 3C, also cf. D-F). During the 30-s to 2-min interval, for example, abundance of Spt3 and pol II peaks at 5-10-fold their initial levels and at the same time H3 levels are reduced 70 -80%, similar to the previously observed behavior of H4 (Fig. 2C) (22). Together, the data demonstrate an inverse relationship between SAGA/pol II and that of nucleosomes within the heat shock-induced HSP82 gene.
SAGA Is Likewise Recruited to the ORFs of SSA3 and CPR6 but Is Largely Restricted to the Promoter Regions of SSA4 and HSP104-We next asked whether SAGA was detectable within the coding regions of other heat shock-induced, Hsf1-regulated genes. At SSA3, Spt3 is detectable within both the promoter and ORF under noninducing conditions. Following heat shock, its abundance increases rapidly, tripling within 30 s and quintupling within 2 min (Fig. 4A). Unlike the case with HSP82, there is no discernible lag in SAGA recruitment to the ORF. pol II recruitment is robust and equally rapid to both promoter and ORF (Fig. 4B), as assayed by Rpb1 ChIP of the same chromatin samples. These results, together with those above, are consistent with recruitment of SAGA to the coding region being mediated by elongating pol II.
We next investigated HSP104, a gene also chiefly regulated by Hsf1. In contrast to what was seen at either HSP82 or SSA3, SAGA principally occupies the HSP104 promoter. Only a small amount associates with the ORF, and only in response to heat shock (Fig. 4D, see legend). pol II, on the other hand, enters the HSP104 coding region within 30 s of heat shock, and its occupancy sequentially peaks at positions arrayed 5Ј to 3Ј across the gene (Fig. 4E). Examination of histone H3 occupancy reveals that its abundance at the HSP104 promoter is reduced nearly 90% within 2 min and is Ͼ80% depleted at the mid-ORF (ϩ1311) region within 15 min, despite the absence of SAGA. The paradoxical increase in histone H3 detected at position ϩ2665 (Fig. 4F, red bars) may reflect the opposing, stabilizing activity of a Sin3-Rpd3 complex, whose recruitment to the HSP104 ORF and 3Ј-UTR is robust (see Fig.  8D, below), and may be a consequence of heat shock-induced nucleosomal assembly of ARS1206 (see physical map, Fig. 4D).
In light of these unexpected differences in the pattern of SAGA occupancy, we extended our survey to two other Hsf1 targets, CPR6, a heat-inducible gene that encodes an Hsp82associated immunophilin, and SSA4, which like SSA3 encodes an Hsp70 chaperone. As shown in Fig. 5A, the Spt3 occupancy profile for CPR6 resembles that of HSP82 and SSA3, low within the noninduced promoter yet rapidly and significantly enhanced in response to heat shock. Moreover, this enhancement is seen within both promoter and coding regions, is significant within 30 s (p Ͻ 0.05), and peaks within 2 min. In contrast, SAGA occupancy of SSA4 is largely restricted to its promoter, to which Spt3 binds robustly following heat shock (Fig. 5B). SAGA is also detectable within the coding region of the gene but at a much reduced level (Ͻ15% of its promoter occupancy) and with less rapid kinetics of recruitment than seen at CPR6.
These results, in combination with those discussed above, indicate that association of SAGA with the coding region differs among the target genes of Hsf1.
Finally, we asked whether SAGA is recruited to heat-activated genes under the dual regulation of Hsf1 and Msn2/Msn4. As can be seen in Fig. 5, C and D, Spt3 is recruited to heat- activated HSP26 and HSP12, as detected by ORF-localized probes. However, it fails to be recruited to the constitutively transcribed PMA1 gene (data not shown), consistent with previous observations that SAGA primarily regulates stress-responsive genes (62).
Rpd3 Histone Deacetylase Is Rapidly Recruited to the HSP82 Promoter and ORF upon Heat Shock, and This Is Facilitated by URS1-We next investigated the role of the Rpd3 histone deacetylase, which is recruited to pol II genes as part of either the Rpd3(L) or Rpd3(S) complex (38,39,63). Rpd3(L) is primarily recruited to gene promoters; Rpd3(S), as discussed above, principally associates with gene coding regions. Previous work suggests that Rpd3(L) is targeted to promoters by sequencespecific DNA-binding proteins such as Ume6 or by co-repres- . SAGA associates with both the promoter and coding region of SSA3 but is restricted to the promoter of HSP104. A, abundance of Spt3-TAP at the promoter and ORF of SSA3 at t ϭ 0 min and the indicated times following heat shock. B, occupancy of pol II, as assayed in Fig. 3. C, H3 abundance at SSA3, as assayed in Fig. 3. D, cross-linking analysis of Spt3-TAP at four indicated regions within HSP104 at t ϭ 0 min and the indicated times following heat shock. The increase in Spt3-TAP abundance seen within the coding region ϩHS does not appear to be significant (p Ͼ 0.1). E, occupancy of pol II at HSP104 as in B. F, abundance of H3 at HSP104 as in C. Depicted for all panels except C are means Ϯ S.E.; n ϭ 4 (C, means Ϯ S.D.; n ϭ 2). Midpoint coordinates of each amplicon are provided. sors such as Sin3 or Tup1. This recruitment typically leads to transcriptional repression (64,65), presumably by inhibiting one or more steps involved in initiation. On the other hand, Rpd3(S) is thought to be recruited to actively transcribed regions independently of sequence-specific DNA-binding proteins, and such recruitment may contribute to the suppression of genome-wide histone acetylation (66). Delivery of Rpd3(S) to active coding regions is achieved through its chromodomainand PHD finger-containing subunits, Eaf3 and Rco1, respectively, that recognize Set2 di-and tri-methylated H3 Lys-36 residues (40,67). The resultant deacetylation of transcribed regions by Rpd3(S) suppresses spurious intragenic transcription initiation (38). In addition, deacetylated nucleosomes could potentially inhibit subsequent rounds of pol II elongation, thereby reducing bona fide gene transcript levels. Indeed, methylation of H3 Lys-36 residues generates a genetic requirement for positively acting elongation factors (39).
To test the role of Rpd3, we asked whether an RPD3 deletion impacted upon hsp82 expression. We observed that under non-inducing conditions HSP82 is derepressed 3-fold by an rpd3⌬ mutation, whereas hsp82-P2 is derepressed an order of magnitude (Fig. 6, A and B, black bars). Induced transcription is also significantly elevated for each allele (increased nearly 70%). These results indicate that Rpd3-containing complexes repress both HSP82 and hsp82-P2 expression.
In an effort to distinguish between the contributions of targeted versus nontargeted Rpd3 complexes, we investigated the role of URS1, a regulatory element located between the UAS HS and TATA box of HSP82 (see Fig. 1A) and a putative recognition site of Ume6 (the sequence, TGAGCG-GTTA, bears an 8/10 match to the URS1 consensus, TGGGCGGCTA (68)). The URS1 motif regulates genes involved in meiosis (64,68). To investigate the role of this cis-element in mitotically growing cells, we used homologous recombination to replace it with an equivalent length of inert DNA (altering no other DNA sequence), creating a chromosomal allele termed hsp82-⌬URS1 (Fig. 1A). Consistent with URS1 mediating transcriptional repression, the FIGURE 5. SAGA recruitment to the Hsf1 target genes CPR6 and SSA4. A, occupancy of Spt3-TAP at the promoter, ORF, and 3Ј-UTR of CPR6 at times both prior to and following heat shock. B, Spt3-TAP abundance at the promoter, ORF, and 3Ј-UTR of SSA4. Midpoint coordinates of each amplicon are indicated. Depicted for all panels are means Ϯ S.E.; n ϭ 4. The increased level of Spt3 detected within the ORF and 3Ј-UTR of SSA4 ϩHS appears significant (p Ͻ 0.001). C and D, SAGA is recruited to the Hsf1-and Msn2/Msn4-co-regulated genes, HSP26 and HSP12. Spt3-TAP abundance at HSP26 or HSP12 was monitored under the conditions indicated (n ϭ 4 Ϯ S.E. or n ϭ 2 Ϯ S.D., respectively).
⌬URS1 substitution led to substantial increases in both basal and induced hsp82 expression (Fig. 6A).
To demonstrate that Rpd3 complexes are in fact recruited to HSP82, and to investigate whether URS1 plays a role, we conducted ChIP using HSP82 and hsp82-⌬URS1 strains expressing TAP-tagged Rco1 or Sap30 (subunits specific to Rpd3(S) or Rpd3(L), respectively), or Sin3 (common to both) (38,39). Surprisingly, under noninducing conditions, Sin3 is undetectable within either the HSP82 promoter or coding region (Fig. 7A, 0 min). Nonetheless, an Rpd3-containing complex is likely present at the promoter because both Sap30 and Rco1 are detected (Fig. 7, B and C, discussed further below). The apparent absence of Sin3 at the noninduced HSP82 5Ј-end may stem from epitope masking, or alternatively, Rpd3 complexes at noninduced HSP82 may lack Sin3. Whatever the explanation, following a 45-s heat shock, Sin3 is readily detected at all three regions, and its occupancy continues to remain high, and essentially unchanged, during the initial 15 min of heat shock. Thereafter, Sin3-containing complexes dissociate from HSP82, particularly following a 20-min recovery. Interestingly, occupancy of Sin3 closely parallels that of pol II, particularly over the ORF (plotted in Fig. 7, D-F), and this likely reflects the presence of Rpd3(S) (see below). Also noteworthy, recruitment of Rpd3 complexes, like that of SAGA, positively correlates with histone H3 eviction. In fact, the bulk of Sin3 is recruited when histone eviction is most rapid (namely, between 0 and 45 s heat shock at the promoter and 0 -2 min elsewhere) (Fig. 7, D-F).
As expected, Rpd3(L) (Sap30) is detected within the HSP82 promoter under noninducing conditions (Fig. 7B). Unexpectedly, given its negative role in regulating HSP82 transcription, Sap30 abundance increases in response to acute heat shock. Sap30 also efficiently cross-links to the 3Ј-UTR, although it shows little tendency to associate with the ORF of the gene. (The low ChIP signal within the ORF could arise from Sap30 bound to the promoter and 3Ј-UTR and thus be due to a ChIP "echo.") Rpd3(S) (Rco1) is also unexpectedly abundant at the HSP82 promoter, particularly during the first 45 s of heat shock (Fig. 7C). Nonetheless, preferential association with the promoter dissipates, and beginning at 2 min, Rco1 occupancy of the 5Ј-, middle, and 3Ј-positions of the gene is roughly equivalent, paralleling occupancy of Sin3.
In an attempt to discern the role of URS1 in the recruitment of Rpd3 complexes, occupancy of Sap30 and Rco1 was analyzed at the hsp82-⌬URS1 allele. In response to heat shock, Rco1 recruitment to hsp82-⌬URS1 is delayed ϳ90 s compared with its recruitment to HSP82 (Fig. 7C). In addition, the persistence of Rco1 at hsp82-⌬URS1 relative to HSP82 is diminished 40% following a 60-min heat shock and Ͼ75% following a 20-min recovery. Delayed recruitment of the Rpd3(L)-specific subunit Sap30 is also seen at hsp82-⌬URS1, and cross-linking to the 3Ј-UTR is reduced (Fig. 7B). These results indicate that the URS1 element enhances the efficiency of recruitment of both Rpd3(L) and Rpd3(S) complexes to the HSP82 promoter, as well as the retention of Rpd3(S) throughout the gene.
Rpd3 Histone Deacetylase Complexes Are Also Recruited to Heat Shock-activated SSA3 and HSP104-To extend the above observations, we asked whether Sin3-containing complexes were recruited to other Hsf1 target genes. As shown in Fig. 8A, Sin3 is abundant within both the SSA3 promoter and ORF under noninducing conditions. Following heat shock, Sin3 levels are slightly increased at the promoter, yet are doubled within the ORF. This increase is sustained throughout the 60-min time course and may help stabilize coding region nucleosomes during heat shock (Fig. 4C). At HSP104, Sin3 is detectable principally at the promoter under basal conditions, and like Spt3, its occupancy is significantly increased upon heat shock (Fig. 8B, also cf. D-F). Recruitment of Sin3 to the HSP104 promoter is kinetically uncoupled from that of pol II (Fig. 8C, compare green and blue  curves), contrasting to the case with HSP82 (Fig. 7D). In addition, Sin3 is detected throughout the body of the gene, where it sequentially occupies the promoter, the ϩ132 ORF, the ϩ1311 ORF, and the ϩ2665 3Ј-UTR (Fig. 8B). Indeed, occupancy of Sin3 closely parallels that of processive pol II (Fig. 8, D-F). Paradoxically, as seen for HSP82 (Fig. 7, D-E), Sin3 abundance within the 5Ј-end of HSP104 anti-correlates with that of histone H3 (Fig. 8, C-E, compare green and black  curves).
Rpd3 Stabilizes Nucleosomes at HSP82-We next sought to determine whether the loss of Rpd3 affected either histone abundance or modification state at HSP82. It might be expected that if Rpd3 plays a role in regulating the histone acetylation state of Hsf1 target genes, then an rpd3⌬ mutation would lead to an enrichment of acetylated nucleosomes. However, although there may be a modest effect of deleting RPD3, it does not appear to be significant, even over the ORF and 3Ј-UTR (Fig. 9B, rpd3⌬ versus RPD3 15 min; p Ͼ 0.1). By contrast, an rpd3⌬ mutation does affect histone eviction, because following a 15-min heat shock, we see Ͼ90% eviction of histone H3 at the HSP82 promoter and ORF in the rpd3⌬ strain compared with 70% in the isogenic RPD3 strain (Fig. 9A; p Ͻ 0.003).
Because recruitment of both the Rpd3(L) and Rpd3(S) complexes is enhanced by URS1 (see Fig. 7, B and C) and because deletion of Rpd3 has global and possibly indirect effects on transcription and chromatin, we asked whether the ⌬URS1 mutation impacted upon either histone acetylation state or nucleosome abundance at hsp82-⌬URS1. Unlike the rpd3⌬ FIGURE 8. Sin3-containing complexes associate with both the promoter and coding region of heat shock-activated SSA3 and HSP104. A, ChIP analysis of Sin3-TAP abundance at SSA3 at 0 min and at the indicated times following heat shock. B, ChIP analysis of Sin3-TAP to the indicated regions of HSP104. A and B, means Ϯ S.E. are depicted; n ϭ 4. C-F, line graphs comparing the abundance of Spt3-TAP (red), Sin3-TAP (green), Rpb1 (blue), and histone H3 (black) within the HSP104 promoter and the indicated regions within its ORF either before or after thermal upshift or following a 20-min recovery from heat shock. Protein occupancies depicted in C-F represent composites derived from data presented in A and in D-F of Fig. 4. knock-out (Fig. 9A), deletion of URS1 does not significantly enhance nucleosome disassembly following heat shock (Fig.  9C). However, consistent with URS1-targeted recruitment of Rpd3(L), H3 acetylation increases within the hsp82-⌬URS1 mutant, particularly within the promoter (p Ͻ 0.07, Fig. 9D). These data are consistent with the idea that Rpd3 complexes stabilize nucleosomes within both the promoter and coding region of HSP82 and that the URS1 cis-element targets deacetylation activity to promoter-localized nucleosomes.
URS1 Deletion, but Not Rpd3 Depletion, Positively Affects Hsf1 Recruitment to the HSP82 Promoter-The enrichment of acetylated histones at the HSP82 promoter in an rpd3⌬ background as well as at the hsp82-⌬URS1 promoter in RPD3 ϩ cells raised the possibility that Hsf1 binding itself might also be enhanced. We thought this possible, given the following: (i) Drosophila HSF preferentially binds acetylated nucleosomes in vitro (69); and (ii) yeast Hsf1 binds its target HSEs embedded in nucleosomes in a Gcn5-dependent fashion. 3 Indeed, as shown in Fig. 10A, Hsf1 occupancy under noninducing conditions increases nearly 2-fold at hsp82-⌬URS1 compared with HSP82 (p Ͻ 0.02). However, this is not recapitulated in the isogenic rpd3⌬ strain (Fig. 10B). This suggests that the slight increase seen in promoter-associated H3 acetylation in the rpd3⌬ strain is insufficient to enhance Hsf1 binding to the WT UAS HS and further that URS1, in addition to fostering recruitment of Rpd3 complexes to the promoter, regulates HSP82 transcription by antagonizing Hsf1 occupancy. FIGURE 9. Rpd3 limits the extent to which histone H3 is evicted at HSP82, whereas URS1 suppresses H3 acetylation within the HSP82 promoter. A, histone H3 abundance at the indicated HSP82 regions as assayed by ChIP. Isogenic RPD3 and rpd3⌬ strains were maintained at 30°C or heat-shocked at 39°C for 15 min. B, abundance of diacetylated H3 (AcK9 and AcK14), quantified relative to unacetylated H3, at HSP82 in cells cultivated as in A. C and D, as in A and B, except isogenic HSP82 and hsp82-⌬URS1 strains were used. In all panels, means Ϯ S.E. are depicted; n ϭ 3 or 4. Minor differences in apparent occupancy of histones for the WT strain (SLY101) in A versus C and B versus D may arise from the different composition of primer mixture used (see "Experimental Procedures").

Roles for Both SAGA and NuA4 in Regulating Heat Shock
Gene Transcription-Results presented in this study implicate the involvement of both SAGA and NuA4 in regulation of heat shock gene transcription. Deletion of either of two SAGA subunits, Gcn5 or Spt20, reduced HSP82 basal transcription 5-fold; moreover, induced transcription of an attenuated hsp82 allele, hsp82-P2, is diminished Ͼ10-fold by loss of Gcn5. Similarly, both Hsf1 binding and transcriptional activation of a third hsp82 allele, hsp82-⌬HSE (70), are strongly Gcn5-dependent. 3 Regulation of Hsf1 target genes by SAGA is likely to be direct given that Spt3, a core subunit of SAGA, is detected at all HSP promoter regions tested. Genes co-regulated by Hsf1 and Msn2/Msn4, specifically HSP12 and HSP26 (71), likewise appear to be under SAGA regulation, based both on Spt3 occupancy and the expression phenotypes of gcn5⌬ and spt20⌬ mutants (this study and data not shown). Targeting of SAGA to heat-activated genes is in agreement with previous work that indicated a primary role for SAGA, as opposed to the alternative TBP-containing complex, TFIID, in the transcriptional activation of stress-response genes (62). Indeed, heat shock induces assembly of SAGA at stress-induced genes (72).
With respect to the H2A-and H4-specific HAT, NuA4, we found that heat shock-activated HSP82 transcription is significantly reduced in a strain expressing a dominant negative version of its catalytic subunit, Esa1 E338Q . This, combined with previous observations of a "burst" in H2A and H4 acetylation within the HSP82 promoter during the first 45 s of heat shock (22), argues for NuA4 involvement in HSP82 regulation. Supporting this idea, ChIP localization studies have shown Esa1 recruitment to several heat shock-induced gene promoters (49,54). Therefore, both NuA4 and SAGA are likely to play important roles in heat shock-induced nucleosomal disassembly at HSP82. SAGA may also participate in creation of the nucleosome-free region spanning the HSP82 promoter under control conditions, given that SAGA is constitutively present at the promoter and that ablation of either Gcn5 or Spt20 severely reduces basal transcription. Active remodeling of the HSP82 promoter to create a nucleosome-free region is critical given that in the absence of DNA-bound Hsf1 (and the chromatin remodeling activities that it recruits), a stable dinucleosome assembles within the ϳ300-bp HSP82 upstream region, both in vivo (9,70) and in vitro. 4 Hsf1 plays a similar role in perturbing the stability and positioning of nucleosomes within the HSC82 promoter (11), and analogous roles have been recently demonstrated for the essential Myb family proteins Abf1 and Reb1 at a variety of genes (73).
SAGA Co-localizes with Elongating RNA Polymerase at Hsf1 Target Genes-A striking finding is that at three Hsf1 target genes, HSP82, SSA3, and CPR6, significant levels of SAGA are detected not only at the promoters, but within the coding regions of the genes. Initially thought to be exclusively targeted to the promoter/UAS of GAL1 (55,57,60), SAGA has been recently detected within the induced coding regions of both GAL1 and ARG1 (74,75). The mechanism underlying this has not been elucidated, however. A role for the pol II elongation complex in recruiting SAGA is suggested by our detailed kinetic assays of HSP82 and SSA3, which reveal that Spt3 occupancy closely parallels that of the large pol II subunit within the ORF and 3Ј-UTR of each gene. Detection of SAGA within heat shock coding regions is unlikely to be an artifact of gene looping. In addition to the fact that Spt3 was sequentially detected 5Ј-to-3Ј within each of these genes during a heat shock time course, mirroring detection of pol II, promoter-bound factors are not inevitably detected within HSP coding regions. For example, as demonstrated here, Hsf1 cannot be detected within the HSP82 coding region under either control or inducing conditions.
Regarding the mechanism of recruitment of SAGA to the promoter, it may be instigated by a direct interaction with   NOVEMBER 20, 2009 • VOLUME 284 • NUMBER 47 DNA-bound Hsf1 which, by analogy with two previously studied acidic activators, Gal4 and Gcn4, may engage in direct physical interaction with the Tra1 subunit of SAGA (57)(58)(59). Hsf1 could also recruit the NuA4 complex to HSP promoters through interaction of its acidic activation domain(s) with Tra1, a subunit shared in common with SAGA. In contrast, SAGA localization to heat shock gene ORFs may involve, in one scenario, direct physical interaction with the elongation complex, either by virtue of SAGA traveling with pol II from the promoter or (more likely) by SAGA reversibly associating with the transcription elongation complex as it processes down the gene. In an alternative scenario, SAGA could be recruited via interactions of its subunits with transcription-specific covalent modifications of coding region nucleosomes. However, this latter mechanism is unlikely, given the rapid and nearly complete disassembly of nucleosomes that occurs within both the HSP82 ORF and 3Ј-UTR upon heat shock induction. Definitive elucidation of how SAGA is recruited to both the promoter and coding regions of HSP82 (and like-regulated genes) remains a goal for future studies.

SAGA and Rpd3 Dynamically Regulate Yeast Heat Shock Genes
In contrast to the case with HSP82, SSA3 and CPR6, heat shock-induced Spt3 recruitment to HSP104 and SSA4 is largely confined to the promoter. However, the difference in SAGA distribution between the two sets of genes may be more apparent than real. For example, although small, the increase in Spt3 detected within the induced SSA4 coding region is significant (p Ͻ 0.001 and Ͻ0.0005 for the ORF and 3Ј-UTR, respectively). Therefore, SAGA association with the coding regions of heat shock-induced genes may be the general case. The basis for its efficient recruitment to the ORFs of select HSP genes is unknown and remains a goal for future studies.
The role of SAGA within transcribed regions is also unclear, and may partly, or primarily, be related to a function other than histone acetylation. In support of this latter possibility are the following points: (i) our failure to see an enrichment in acetylated nucleosomes within the HSP82 coding region upon heat shock (Fig. 9); (ii) the failure of a gcn5⌬ knock-out to impair domain-wide nucleosome disassembly of heat shock-induced HSP82 (22); and (iii) the fact that hsp82-P2 is strongly dependent on Gcn5 for induced expression (Fig. 2B), despite showing no detectable increase in H2A, H3, or H4 acetylation upon heat shock (22). At heat shock gene promoters, SAGA may also play a role that extends beyond nucleosomal remodeling, given that elevated levels of SAGA are detected between 2 and 15 min following thermal upshift, a time when most histone H3 (and by extension, nucleosomes) has been evicted. One possibility is that SAGA may facilitate assembly of the pre-initiation complex during both the pioneering and subsequent rounds of transcription. In support, previous work has shown that SAGA stabilizes the TBP-TATA interaction (76) and that SAGA is required for the recruitment of both TBP and pol II to Gcn4regulated gene promoters (77,78).
Both Rpd3(L) and Rpd3(S) Complexes Are Recruited to the HSP82 Promoter Region-The histone deacetylase Rpd3 has been previously shown to be recruited to stress-responsive genes, including those induced by osmotic stress, hypoxic stress, and DNA damage (31,79,80). Interestingly, Rpd3 is involved in activation of these genes. At HSP12, Sin3 binds directly to the upstream regulatory region of the gene in response to osmotic stress, accompanied by reduced levels of acetylated histones and increased levels of pol II (31). Similarly, Rpd3 is robustly recruited to the UAS of the DNA damageinducible gene, RNR3, where it deacetylates nucleosomes (80). An Rpd3-Sin3-Sap30 complex was shown to be required for activation of the anaerobic gene DAN1, where Rpd3 remodels promoter-associated nucleosomes under inducing conditions to permit stable binding of a sequence-specific activator (79).
In contrast, we show that both basal and heat shock-induced HSP82 expression is repressed by targeted Rpd3 complexes. An Rco1-containing complex, which we interpret to be Rpd3(S), associates with the HSP82 promoter prior to heat shock. Its occupancy there dramatically increases during the initial 30 -45 s of heat shock, and this is partially mediated by the URS1 element. At the same time, Rco1 is detectable within the coding region of the gene, where it remains associated for at least 60 min. Efficient recruitment of Rpd3(S) to the HSP82 coding region is consistent with the idea that the remaining nucleosomes (Ͻ20% of the noninduced state) are methylated at H3 Lys-36. Both di-and tri-methylated H3 Lys-36 residues are bound by the Eaf3/Rco1 subunits of Rpd3(S) (40,67) and thus can trigger localized deacetylation.
In addition, a Sap30-containing complex, which we interpret to be Rpd3(L), is recruited to the promoter of HSP82 under both NHS and HS conditions, and its occupancy is likewise enhanced by heat shock. Interestingly, Sap30 also cross-links to the 3Ј-UTR in parallel with its association with the promoter but is scarce within the ORF. The distinctive 5Ј/3Ј cross-linking of Sap30 raises the possibility that HSP82 is looped, under both noninducing and inducing conditions, with its promoter engaged in physical contact with the 3Ј-end of the gene, as has been reported for other transcriptionally active yeast genes (81)(82)(83). It is intriguing that such looping, if it indeed exists, fosters dual 5Ј/3Ј cross-linking of Sap30 at all times but not of any other factor examined. Therefore, if HSP82 is in fact looped, which has yet to be demonstrated, then Rpd3(L) may participate in this process. How this may affect Rpd3(L)-mediated transcriptional repression of HSP82 is unclear.
Although not addressed by ChIP, the striking rpd3⌬ phenotype of hsp82-P2 argues for direct involvement of Rpd3 complexes in regulation of its chromatin structure. As discussed above, severely weakening the Hsf1-promoter DNA interaction at a second mutant, hsp82-⌬HSE1, results in the appearance of sequence-positioned, repressive nucleosomes upstream of hsp82 (9,70). Results presented here, in conjunction with earlier findings (22), suggest that the P2 mutation reduces hsp82 transcription 10-fold, at least in part, through Rpd3-mediated deacetylation of promoter-associated nucleosomes.
Role for Hsf1 in the Recruitment of Repressive Rpd3 Complexes-The Sin3, Sap30, and Rco1 ChIP results represent, to our knowledge, the first evidence that Rpd3 complexes are recruited to both the promoter and coding region of a gene upon its induction. Promoter recruitment of Rco1 was unanticipated given that current models, as discussed above, propose that Rpd3(S) is specifically recruited to actively transcribed coding regions where H3 Lys-36 methylated marks target Rpd3(S) complexes (38,39). It is possible that Rpd3(S) recruitment to the promoter occurs via the same mechanism, but if this were the case, it is unclear how promoterassociated nucleosomes are methylated at H3 Lys-36 residues. Alternatively, the primary recruiter of both complexes might be Hsf1 itself, given the following points: (i) Hsf1 is constitutively bound to HSE1 (this study and Refs. 7,9,25); and (ii) potent N-and C-terminal activation domains of Hsf1 are rapidly unmasked in response to temperature upshift to trans-activate target genes and promote promoter chromatin remodeling (30,84). Indeed, the abundance of Rpd3 complexes at the HSP82 promoter correlates with transcriptional output. This seems paradoxical, given that its role appears to be to suppress transcription, in contrast to its activating role at other genes. We propose that Rpd3 acts to fine-tune the heat shock transcriptional response, thereby modulating transcriptional output of HSP genes under inducing, as well as noninducing, conditions. Importantly, there is no statistical difference in the recruitment kinetics of Rpd3 and SAGA complexes to either the promoter or coding region of the induced HSP82 gene (compare Fig. 3A and Fig. 7A). As such, our data are consistent with essentially synchronous recruitment of these complexes. A model summarizing the potential roles for Rpd3 and SAGA in regulating HSP82 is schematically depicted in Fig. 11. It is interesting that the abundance of Sin3-containing complexes is highest when that of histone H3 is lowest (between 2 and 15 min of HS). Although this may indicate that Rpd3 is required to deacetylate the remaining nucleosomes (ϳ10% of those present at t ϭ 0 min), as suggested in the model, it also raises the possibility that Rpd3, like SAGA discussed above, has a role in regulating HSP82 that extends beyond nucleosomal remodeling. Future work will be required to investigate this and other interesting possibilities. FIGURE 11. Summary of SAGA, Rpd3(L), and Rpd3(S) complex occupancy at the wild-type HSP82 gene under NHS, acute HS, and prolonged HS states based on kinetic ChIP assays. The schematic emphasizes the main conclusions of this study as follows: (i) SAGA and Rpd3 complexes are both present at heat shock promoters under noninducing as well as inducing conditions; (ii) SAGA and Rpd3(S) are rapidly and synchronously recruited to select heat shock gene coding regions, including that of HSP82, in response to heat shock; and (iii) the abundance of HAT and histone deacetylase complexes is inversely proportional to nucleosomes throughout the length of the HSP82 gene under each condition. Asterisks indicate sequence-specific DNA binding subunits of Rpd3(L); Ume6 bears the pertinent DNA-binding activity toward URS1 (64).