|Home | About | Journals | Submit | Contact Us | Français|
NuA4, the major H4 lysine acetyltransferase (KAT) complex in Saccharomyces cerevisiae, is recruited to promoters and stimulates transcription initiation. NuA4 subunits contain domains that bind methylated histones, suggesting that histone methylation should target NuA4 to coding sequences during transcription elongation. We show that NuA4 is cotranscriptionally recruited, dependent on its physical association with elongating polymerase II (Pol II) phosphorylated on the C-terminal domain by cyclin-dependent kinase 7/Kin28, but independently of subunits (Eaf1 and Tra1) required for NuA4 recruitment to promoters. Whereas histone methylation by Set1 and Set2 is dispensable for NuA4's interaction with Pol II and targeting to some coding regions, it stimulates NuA4-histone interaction and H4 acetylation in vivo. The NuA4 KAT, Esa1, mediates increased H4 acetylation and enhanced RSC occupancy and histone eviction in coding sequences and stimulates the rate of transcription elongation. Esa1 cooperates with the H3 KAT in SAGA, Gcn5, to enhance these functions. Our findings delineate a pathway for acetylation-mediated nucleosome remodeling and eviction in coding sequences that stimulates transcription elongation by Pol II in vivo.
Nucleosomes inhibit transcription initiation by RNA polymerase II (Pol II) by impeding assembly of the preinitiation complex (PIC) at the promoter. Transcriptional activators bind to upstream activation sequence (UAS) elements and recruit ATP-dependent chromatin remodeling complexes to displace or evict nucleosomes from promoter regions as a means of stimulating PIC assembly. Neutralization of the positive charges on Lys residues by acetylation weakens histone interactions with DNA. Moreover, acetylation provides recognition sites for subunits of remodeling complexes that harbor bromodomains (BDs). This enables KAT complexes to enhance nucleosome displacement by remodeling complexes in vitro, and it may underlie their roles in histone eviction from promoters in vivo (4, 22, 29, 53).
Nucleosomes also impede elongation by Pol II, and histones are evicted from coding sequences in a manner directly correlated with the transcription rate (51, 60). The mechanism of cotranscriptional nucleosome eviction is not well understood. The histone chaperone Asf1 and the chromatin remodeling complex SWI/SNF have been implicated in nucleosome disassembly during elongation in Saccharomyces cerevisiae cells (50, 52). As during initiation, histone acetylation by KAT complexes could stimulate the passage of elongating Pol II by altering DNA-histone contacts (42) or by enhancing recruitment of chromatin remodeling complexes, as demonstrated for RSC and its ability to facilitate elongation through a reconstituted mononucleosome in vitro (9). This elongation-promoting activity of histone acetylation is not well documented in vivo, however, and until recently it was thought that KAT complexes are confined to the UAS and promoter regions.
We and others have found that the H3 KAT complex SAGA occupies the coding sequences of transcribed genes (21, 62, 64) and have obtained evidence that SAGA stimulates transcription elongation. Elimination of SAGA subunits, including the KAT Gcn5, more strongly impairs expression of long versus short transcription units driven by the GAL1 promoter (21, 34, 40). Nucleosome depletion in the GAL1 open reading frame (ORF) is blunted in gcn5Δ cells, which might contribute to the reduced processivity of Pol II detected during transcription of a long (~8-kb) ORF driven by the GAL1 promoter (21). Moreover, SAGA subunit Sus1 acts to couple transcription elongation to mRNA nuclear export factors (40, 49) and is instrumental in tethering GAL1 to the nuclear periphery (8).
While SAGA acetylates H3 and H2B, the KAT subunit of NuA4, Esa1, is responsible for most of the H4 and H2A acetylation in vivo (15, 30). Esa1 is the only essential KAT in budding yeast and resembles Gcn5 in broadly affecting expression of many genes (5, 16-18, 35, 46). Esa1 resides in a small complex known as picNuA4, along with Epl1 and Yng2. These three subunits also exist as a subcomplex of NuA4, which contains 10 other subunits, many of which also reside in other chromatin modifying complexes (15).
NuA4 differs from picNuA4 in its ability to interact tightly with activators (6) and can be recruited to chromatin templates in vitro (58), suggesting that NuA4 mediates targeted H4/H2A acetylation at promoters while picNuA4 catalyzes nontargeted acetylation genome-wide. Chromatin immunoprecipitation (ChIP) studies revealed that Esa1 occupies the promoters of many genes to a degree that correlates with transcriptional output (48) and is recruited by various activators in vivo (5, 18, 35, 46, 48). While it is not entirely certain whether the UAS occupancies of Esa1 in these studies reflect recruitment of NuA4 or picNuA4, an epl1 mutation that weakens association of Esa1 with NuA4 blocks PHO5 activation, implying that NuA4 is the stimulatory complex at this gene (39). Similarly, it was reported that NuA4 integrity requires Eaf1 (3, 33), and eaf1Δ reduces transcription of certain Esa1-dependent genes (3).
Recruitment of Esa1 is associated with increased acetylation of promoter nucleosomes (18, 35, 46, 59), but how this stimulates transcription is not fully understood. Both esa1 and gcn5Δ mutations reduce SWI/SNF recruitment to promoters in vivo, although the role of Snf2's BD in this effect is unresolved (18, 20, 23). Esa1 also stimulates recruitment of the FACT complex to the ARG3 promoter, which is crucial for ARG3 transcription (5). At MET16, it appears that NuA4 stimulates Pol II promoter escape by antagonizing the inhibitory function of the chromatin remodeling complex ISW1 (35).
The results of most ChIP studies have been interpreted to indicate that Esa1 and Epl1 are restricted to promoters (46, 48). NuA4 contains two chromodomains and one plant homeobox domain (PHD) that have been shown to recognize methylated histones (24, 25, 54). Because histone methylation occurs cotranscriptionally and is enriched in the coding sequences (55), it could be that NuA4 is recruited to transcribed coding sequences and functions during elongation, possibly in conjunction with SAGA. Although there is evidence that Esa1 catalyzes H4 acetylation in coding sequences (46), there is no evidence that NuA4 stimulates elongation in vivo.
In this study, we show that NuA4 is recruited cotranscriptionally to coding sequences by a mechanism involving both phosphorylation of the Pol II CTD and histone methylation. The presence of NuA4 in the coding region mediates increased H4 acetylation, enhanced recruitment of chromatin remodelers, and cotranscriptional histone eviction, and it stimulates the rate of Pol II elongation, all in conjunction with the KAT subunit of SAGA.
All strains used in this study are listed in Table Table1.1. The wild-type strain BY4741 or deletion derivatives described previously (61) were purchased from Research Genetics. All deletions were verified by PCR. The myc-tagged strains were generated as described previously (57) and verified by PCR analysis of chromosomal DNA and Western blot analysis with anti-myc antibodies. Strains containing the chromosomal PGAL1-YLR454w locus were created as described previously (21) and verified by PCR and Northern analysis. Strain DG92 was created by sporulating strain DG41 and selecting for spores isogenic to BY4741 except for eaf1Δ::kanMX4 and EPL1-myc13::HIS3 by tetrad analysis. Strains containing esa1-L254P were made by integrating AflII-digested pLP949 (13), plating Ura+ transformants on 5-fluoroorotic acid, and screening for Ts− Ura− colonies. Mutant esa1 strains were confirmed by Western blotting of whole-cell extracts (WCE) for reduced levels of H4Ac after incubation at 36°C and complementation of the Ts− phenotype by plasmid-borne ESA1. Strains with set1Δ::hphMX4 were created as described previously (19), amplifying the hphMX4 cassette of pAG32 with primers 5′-CTAGCATAGGTAACATTCCTTATTTGTTGAATCTTTATAAGAGGTCTCTGCGTTTAGAGACGTACGCTGCAGGTCGAC-3′ and 5′-TTTGCTGGAAAGCAACGATATGTTAAATCAGGAAGCTCCAAACAAATCAATGTATCATCGATCGATGAATTCGAGCTCG-3′. Loss of SET1 was confirmed by PCR and Western blotting of whole-cell extracts for the absence of trimethylated Lys4 in histone H3. Strains DG326 and DG329 were made by amplifying eaf1Δ::kanMX4 from strain 34196 with primers EAF1-A and EAF1-D, integrating the PCR product into BY4741 or H3632, and selecting for kanamycin resistance. The deletion was confirmed by PCR.
ChIP assays were carried out as described previously (21). Locations of ChIP PCR primers are listed in Table Table2.2. New primers for this study included the following: HIS4 UAS, 5′-GATGGGTCATAATACAGCAGA-3′, and 5′-GACTGATAAAAAAACGTGAGTCAC-3′; HIS4 3′ ORF, 5′-ACACTGCAACCTTCCAAAAG-3′ and 5′-TCAACCCAAGCTTACTCATTC-3′; GAL1 UAS, 5′-GCTCATTGCTATATTGAAGTACGG-3′ and 5′-AATCTTTATTGTTCGGAGCAGTG-3′; GAL1 TATA, 5′-AGTAACCTGGCCCCACAAAC-3′ and 5′-AAAGTGGTTATGCAGCTTTTCC-3′; GAL1 5′ ORF, 3′-GCTGGGGTGGTTGTACTGTT-3′ and 5′-ATAGACAGCTGCCCAATGCT-3′.
Coimmunoprecipitation experiments were carried out using whole-cell extracts as described previously (65) with the antibodies described below. Western blot analysis was conducted using whole-cell extracts made by trichloroacetic acid extraction as described previously (45) with the antibodies described below. Measurement of Pho5 enzymatic activity for gene length-dependent accumulation of mRNA (GLAM) ratio determinations was carried out as described previously using transformants of the appropriate strains harboring plasmids p1989 (empty URA3 vector), pSCh202 (PGAL1-PHO5 URA3), or pSCh209 (PGAL1-PHO5-LAC4 URA3) (34). Northern blotting of total RNA was carried out as described previously (43) using probes to the GAL1 ORF and the 5′-half of the PGAL1-YLR454w ORF.
The following antibodies were used for ChIP, coimmunoprecipitation analysis, or Western blot analysis: mouse monoclonal anti-Myc (Roche), rabbit monoclonal anti-Esa1 (ab4466; Abcam), mouse monoclonal anti-Rpb3 (Neoclone), anti-phospho-Ser5 Rpb1 (H14; Covance), anti-phospho-Ser2 Rpb1 (H5; Covance), mouse anti-Rpb1p (8WG16; Covance), rabbit polyclonal anti-H3 (ab1791; Abcam), mouse monoclonal anti-H4 (ab31827; Abcam), rabbit anti-H4Ac (06-866; Upstate), rabbit monoclonal anti-trimethyl (Lys4) histone H3 (05-745: Upstate), rabbit anti-Gcd6 (12), mouse anti-FLAG (Sigma).
Motivated by previous findings that SAGA is recruited to coding regions and that Gcn5 stimulates elongation in vivo, we sought to implicate the KAT activity of NuA4 in transcription elongation. We asked first whether impairing the KAT activities of NuA4 and SAGA confers increased sensitivity to 6-azauracil (6-AU), an inhibitor that impedes elongation by lowering nucleotide pools (32, 36). 6-AU sensitivity (6-AUs) is not a definitive indicator of elongation defects but is produced by mutations in various elongation factors (47). We examined a Ts− mutation in ESA1 (esa1-L254P) that reduces histone H4 acetylation in bulk chromatin at the restrictive temperature of 36°C but has little effect at the permissive temperature of 30°C (Fig. (Fig.1A)1A) (13). (Henceforth, we refer to this allele simply as esa1). Interestingly, the esa1 mutation conferred 6-AUs at 36°C but not at 30°C (Fig. (Fig.1B).1B). gcn5Δ also confers 6-AUs and, importantly, esa1 exacerbates this phenotype at 30°C (Fig. (Fig.1B).1B). Thus, H3 and H4 acetylation by the KAT subunits of NuA4 and SAGA contribute to 6-AU resistance, possibly by stimulating a common aspect of transcription elongation.
To test this interpretation, we employed the GLAM assay to reveal defects in elongation. This assay compares the expression of reporters containing 1.5-kb or 4.5-kb PHO5 transcription units, both under the GAL1 promoter. Mutations in elongation factors produce a greater reduction in the long versus short reporter transcript and the encoded Pho5 enzyme activities (decreased GLAM ratio) (34). Consistent with previous findings (34), gcn5Δ lowered the GLAM ratio by a factor of ~2 after 1 h of incubation at 36°C. Although esa1 did not reduce the GLAM ratio in otherwise-wild-type (WT) cells under these conditions (Fig. (Fig.1C),1C), its effect on H4 acetylation was modest after only 1 h at 36°C (Fig. (Fig.1A).1A). Nevertheless, esa1 further reduced the GLAM ratio in the gcn5Δ background (Fig. (Fig.1C),1C), suggesting that Esa1 and Gcn5 act independently to stimulate transcription elongation.
In a second gene length-dependent transcription assay, we compared production of the ~8-kb YLR454w transcript from the chromosomal YLR454w gene placed under the control of the GAL1 promoter (32) to that of the ~1.5-kb native GAL1 mRNA (driven by the same promoter). After 2 h at 36°C, both the esa1 and gcn5Δ mutations reduced the ratio of YLR454w to GAL1 mRNA ~30% and 60%, respectively (Fig. (Fig.1D).1D). The fact that the esa1 mutant shows a defect in this assay but not in the GLAM assay might be due to the longer length of the YLR454w ORF versus the PHO5-LAC4 ORF (8 versus 4.5 kb) or to differences in the nucleotide sequence between the two long ORFs. Importantly, the double mutant displayed a more severe reduction in this ratio than either single mutant (Fig. (Fig.1D),1D), further supporting the idea that NuA4 and SAGA KAT subunits make independent contributions to the efficiency of elongation.
We explored next whether the NuA4 and SAGA KAT subunits stimulate the rate of transcription elongation by analyzing the kinetics of Pol II elongation in the ~8-kb PGAL1-YLR454w ORF. Upon glucose addition to cells growing in galactose, Pol II recruitment to the GAL1 promoter is blocked and preexisting elongating Pol II molecules finish transcribing the ORF. The kinetics of Pol II runoff during this last wave of elongation provide a measure of the elongation rate in vivo (32). In our experiments, cells were grown in the noninducing carbon source raffinose at 30°C, incubated for 1 h at 36°C (to inactivate the esa1 product), induced with galactose for another hour at 36°C, and processed for ChIP analysis of Rpb3 occupancy at various times after adding glucose to repress initiation. The background Rpb3 occupancies after 40 min in glucose were subtracted from the occupancies at earlier time points in glucose and normalized to the occupancies measured in galactose for different locations across PGAL1-YLR454w (Fig. (Fig.2A),2A), yielding the results shown in Fig. 2B to E for the different strains. (This approach corrects for PIC assembly defects by setting the maximal Rpb3 occupancies in galactose to unity and occupancies after 40 min in glucose to zero for every strain. Analysis of histograms produced without this correction, following the methods of Mason and Struhl , yielded the same conclusions that are drawn below from the data in Fig. Fig.22.)
As expected, following addition of glucose to WT cells, Pol II disappeared from the promoter and 5′ end of the PGAL1-YLR454w ORF more rapidly than from the 3′ end of the ORF. Thus, after 2 min in glucose, there was a large decline in Rpb3 occupancy at the promoter and a smaller decline 2 kb downstream from the promoter, but no decline at locations 4, 6, or 8 kb from the promoter (Fig. (Fig.2B).2B). By 4 min in glucose, only ~20 to 40% of the amounts of Rpb3 present in galactose remained associated with the PGAL1-YLR454w ORF, and by 6 min these figures dropped to only 5 to 13% of the galactose-associated Rpb3 levels (Fig. (Fig.2B2B).
Remarkably, the rate of Pol II runoff appeared to be significantly lower in the esa1 mutant (Fig. (Fig.2C).2C). After 4 min in glucose there were only 30% and 20% decreases in Rpb3 occupancy at the 4-kb and 6-kb locations, respectively, compared to the 70% and 60% reductions at these locations in WT (compare Fig. 2B and C). After 6 min in glucose, Rpb3 occupancies were still ~2-fold higher at all locations in the mutant versus WT. The results for the gcn5Δ mutant are less conclusive; however, we consistently observed higher-than-WT Rpb3 occupancies at all locations 4 min and 6 min after the addition of glucose, suggesting a small reduction in the Pol II elongation rate for this mutant as well (Fig. (Fig.2D).2D). Importantly, the gcn5Δ esa1 double mutant exhibits the largest elongation rate defect of all three mutants. These findings suggest that the KAT subunits of both SAGA and NuA4 stimulate the rate of elongation by Pol II. (It appeared that loss of Rpb3 from the promoter was faster in the double mutant [Fig. [Fig.2E]2E] than in the WT or single mutants [Fig. 2B to D]. Perhaps a fraction of the PICs are nonproductive and relatively less stable in the double mutant.)
We presumed that the additive effects of gcn5Δ and esa1 on transcription elongation derive from simultaneously reducing H3 and H4 acetylation of ORF nucleosomes. To demonstrate that Esa1 mediates H4 acetylation in GAL1 coding sequences during transcription induction, we conducted ChIP analysis of H4 acetylated on lysines 5, 18, 12, and 16 and normalized these data for H3 occupancy to correct for reductions in nucleosome density associated with transcriptional activation. After 1.5 h at 36°C, we observed a reduced H4Ac/H3 ratio in esa1 versus WT cells at both UAS and 3′ ORF sequences at GAL1 during induction, whereas gcn5Δ cells displayed elevated H4Ac/H3 ratios at both locations (Fig. (Fig.3B),3B), as previously reported (1). (The increased H4 acetylation in gcn5Δ cells might be attributable to reduced recruitment of the Set3/Hos2 histone deacetylase complex [Set3C], as we showed previously that trimethylated H3-Lys4 levels decrease in gcn5Δ cells  and others reported that H3-Lys4 methylation stimulates Set3C recruitment .) Importantly, esa1 reduced the H4Ac/H3 ratios dramatically in the gcn5Δ esa1 double mutant compared to the gcn5Δ single mutant (Fig. (Fig.3B).3B). Similar results were obtained after 4.5 h at 36°C except for a more dramatic reduction in H4Ac/H3 ratios in esa1 cells (Fig. (Fig.3C).3C). (We confirmed that only background signals [occupancies of ~1] were observed when ChIP assays were conducted without any antibodies, for all strains examined in Fig. Fig.3C,3C, and for both raffinose- and galactose-grown cells.) Considering that H4Ac in bulk histones is only partially reduced in esa1 cells after 2 or 4 h at 36°C (Fig. (Fig.1A),1A), these ChIP results suggest that Esa1 is responsible for the majority of H4Ac present in the coding sequences and UAS during GAL1 induction.
It might seem surprising that we observed little or no increase in the H4Ac/H3 ratios in WT cells on galactose induction (Fig. (Fig.3B).3B). However, we showed previously that Gcn5-dependent H3 acetylation also remains unchanged, or declines, in the ARG1 coding sequences during transcriptional induction by Gcn4. This was attributed to the fact that Gcn4 recruits multiple histone deacetylase complexes (HDACs) to the ARG1 ORF (21). We found subsequently that multiple HDACs also are recruited to the GAL1 UAS and ORF on galactose induction (C. K. Govind and A. G. Hinnebusch, unpublished observations). Hence, we examined the H4Ac/H3 ratio in a triple mutant lacking HDAs Hda1, Hos2, and Rpd3. In this mutant, the H4Ac/H3 ratios were increased under noninducing conditions and to an even greater extent under inducing conditions, at both the UAS and ORF (Fig. 3B and C). Thus, it appears that cotranscriptional acetylation by Esa1 is opposed by one or more HDACs to maintain an optimal level of H4 acetylation across the GAL1 gene.
We asked next whether the reduction in H4 acetylation in esa1 cells dampens histone eviction at GAL1 during transcription induction. In WT cells after 1.5 h at 36°C, galactose induction reduced H3 occupancies at the UAS and 3′ ORF of GAL1 by ~70% and ~50%, respectively, of their values in noninducing medium (Fig. (Fig.3D,3D, WT, UAS, and 3′ ORF). These results are in accordance with other studies conducted at 30°C for GAL1, GAL7, and GAL10 (21, 51). Both gcn5Δ and esa1 mutants exhibited significantly smaller reductions in H3 occupancy on galactose induction in the UAS and ORF than occurred in the WT (Fig. (Fig.3D).3D). These defects are exacerbated in the gcn5Δ esa1 double mutant, with H3 eviction being largely eliminated in the UAS and ORF (Fig. (Fig.3D).3D). Thus, histone acetylation by Esa1 and Gcn5 contributes independently to nucleosome eviction from both promoter and coding sequences during GAL1 transcription. The decreased histone eviction in the gcn5Δ esa1 double mutant versus the single mutants might contribute to the relatively more severe elongation defects observed for the double mutant in the experiments described above (Fig. (Fig.11 and and22).
Because histone eviction from coding sequences is associated with high-level transcription and Pol II occupancy, we considered the effects of gcn5Δ and esa1 mutations on Pol II occupancies in the GAL1 ORF. We observed a significant reduction in the ORF occupancy of Rpb3 (~40%) only in the gcn5Δ esa1 double mutant (Fig. (Fig.3E,3E, ,3′3′ ORF), likely owing to its defect in PIC assembly (Fig. (Fig.3E,3E, TATA). The lower density of elongating Pol II molecules could contribute to the diminished H3 eviction from the ORF observed in the gcn5Δ esa1 mutant. On the other hand, the decreased histone eviction observed in this strain could be a factor in reducing the density of elongating Pol II molecules.
It was reported that histone acetylation stimulates the recruitment and activity of chromatin remodeling complex RSC in promoting Pol II elongation through a mononucleosome in vitro (9). Hence, we wondered if the elongation and histone eviction defects in the gcn5Δ and esa1 mutants reflect impaired RSC recruitment to coding sequences. Accordingly, we examined RSC occupancies at GAL1 by ChIP analysis of the ATPase subunit Sth1. In WT cells, myc-tagged Sth1 was not detected above background levels in the UAS (data not shown) but was readily observed in the 3′ ORF region at GAL1, dependent on galactose induction (Fig. (Fig.4A,4A, WT, Gal). After 1.5 h at 36°C, the esa1, gcn5Δ, and gcn5Δ esa1 mutants showed greatly reduced Sth1 occupancies of the GAL1 3′ ORF. As shown in Fig. Fig.3E,3E, the Rpb3 occupancy in the GAL1 3′ ORF was not reduced by the esa1 and gcn5Δ mutations under these conditions; hence, the decreased Sth1 occupancies in these mutants do not reflect decreased levels of elongating Pol II. As we observed no differences in Sth1 levels in these strains by Western analysis (Fig. (Fig.4B),4B), the ChIP results suggest that both H3 acetylation and H4 acetylation enhance RSC recruitment to the GAL1 ORF in vivo.
Previous work showed that histone acetylation by NuA4 or SAGA stabilized the SWI/SNF interaction with a nucleosome array in vitro (23) and that SWI/SNF is recruited to coding sequences in vivo (52). We found that, similar to RSC, the occupancy of the SWI/SNF subunit Snf6 in GAL1 coding sequences was reduced in esa1 cells after 1.5 h at 36° (Fig. (Fig.4C)4C) without a decrease in Snf6 protein levels (Fig. (Fig.4B).4B). There was a much smaller reduction in Myc-Snf6 occupancy in gcn5Δ cells, however, suggesting that SWI/SNF recruitment in vivo depends more heavily on acetylation of H4 than of H3.
Having found that RSC is recruited cotranscriptionally, we asked whether mutations affecting RSC function impair elongation. RSC is found as two complexes, identical except for the presence of Rsc1 in the RSC1 complex and Rsc2 in the RSC2 complex. We found that rsc1Δ and rsc2Δ cells are 6-AUs, with a somewhat stronger phenotype for the rsc2Δ mutant (Fig. (Fig.5A).5A). Combining the rsc2Δ and esa1 mutations did not lead to an obvious increase in 6-AUs compared to the rsc2Δ mutant at the permissive temperature (30°C) (Fig. (Fig.5A).5A). In addition to 6-AUs, rsc1Δ and rsc2Δ both conferred defects in the gene length-dependent transcription assays described above, provoking a greater reduction in levels of long versus short transcripts (Fig. 5B and C). Again, we did not observe a more severe defect in the rsc2Δ esa1 double mutant compared to the rsc2Δ single mutant. Interestingly, ChIP analysis of PGAL1-YLR454w under inducing conditions revealed that rsc2Δ mutants show a small but significant decrease in Pol II occupancy toward the 3′ end of the coding sequence (Fig. 5D to F), suggesting a moderate processivity defect (32). These results are consistent with the idea that NuA4 facilitates transcription elongation in part by stimulating RSC recruitment to coding sequences.
Having found that H4 acetylation by NuA4 stimulates elongation, it was crucial to determine whether NuA4 is recruited cotranscriptionally to coding sequences and interacts with elongating Pol II. To this end, we compared the occupancies of myc-tagged NuA4 subunits at the UAS and coding sequences of genes induced by Gcn4. Starving cells for Ile and Val with sulfometuron methyl (SM) evokes a rapid increase in occupancies of Gcn4 and coactivators at the UASs of its target genes (20). Here we found that the UAS occupancies of NuA4 subunit Eaf1 were ~4-fold higher in WT versus gcn4Δ cells at ARG1 (Fig. 6B and C, UAS), and similar results were obtained for ARG4 (Fig. (Fig.6F).6F). Gcn4-dependent recruitment to the UASs of both genes was likewise observed for Eaf5 and Eaf7, subunits of the proposed “Rpd3(S) module of NuA4” absent from picNuA4 (Fig. 6D, E, G, and H). As Eaf1 is thought to be the only subunit unique to NuA4 (3, 33), our results clearly establish that NuA4 is recruited by Gcn4 to the UAS elements of its target promoters.
Importantly, Eaf1 and other NuA4 subunits also displayed Gcn4-dependent occupancies of the 3′ ends of the coding regions of both genes, comparable to those measured for the cognate UAS elements (Fig. 6B to H, ,3′3′ ORF). We also conducted ChIP analysis of NuA4 at GAL1 in cells grown with noninducing or inducing carbon sources. The results indicated greater occupancies of Eaf1, Epl1, and Esa1 in both the 3′ ORF and UAS regions of GAL1 under inducing conditions, particularly for the coding sequences (Fig. 7A to C). Hence, NuA4 occupancy of coding sequences is associated with transcriptional induction by Gal4 or Gcn4. We extended our findings on NuA4 recruitment to include the constitutively expressed genes ADH1 and PMA1 (see Fig. 9F and G, below), leading us to conclude that NuA4 occupancy of coding sequences is not limited to inducible genes.
Because it was reported that Eaf1 is required for NuA4 integrity, we examined the effect of deleting EAF1 on the occupancy of NuA4/picNuA4 subunit Epl1 at Gcn4 target genes. Interestingly, UAS occupancy was reduced to nearly background levels in eaf1Δ cells (Fig. 8A to C), indicating that most Epl1 is recruited to the UAS by Gcn4 as a subunit of NuA4. Surprisingly, however, 3′ ORF occupancy of Epl1 was unaffected by eaf1Δ at all three genes (Fig. 8A to C). One explanation could be that Epl1 in picNuA4 can still associate with the coding sequences when dissociated from NuA4 in the eaf1Δ mutant. However, Eaf5, present in NuA4 but not picNuA4, shows the same behavior as Epl1, with eaf1Δ reducing Eaf5 occupancy at the UAS but not the coding sequences (Fig. 8D and E). Northern blot analysis of ARG1, ARG4, and HIS4 revealed that expression of these genes is unaffected in eaf1Δ cells (data not shown). This finding, plus the fact that ORF occupancies of Epl1 and Eaf5 in eaf1Δ cells are Gcn4 dependent (Fig. 8A to E), confirms that NuA4 recruitment to the coding sequences is cotranscriptional in eaf1Δ cells. Hence, our findings suggest that Eaf1 is required for NuA4 recruitment by Gcn4 to the UAS, but not for cotranscriptional recruitment of NuA4 to the coding sequences.
This last conclusion implies that NuA4 is not fully disrupted by elimination of Eaf1 in vivo. Supporting this deduction, eaf1Δ does not reduce coimmunoprecipitation of picNuA4 subunit Esa1 with myc-tagged Eaf5 (Fig. (Fig.8F,8F, lanes 1 to 6 and 13 to 15). As expected, eaf1Δ also does not affect coimmunoprecipitation of Esa1 with another picNuA4 subunit, myc-Epl1 (Fig. (Fig.8F,8F, lanes 7 to 12). Remarkably, eaf1Δ does eliminate coimmunoprecipitation of Esa1 with FLAG-tagged Tra1 (Fig. (Fig.8F,8F, lanes 16 to 24), indicating that eaf1Δ leads to dissociation of Tra1 without fully disrupting the NuA4 complex. Considering evidence that Tra1 provides a direct contact for Gcn4 in NuA4 (7), its absence from the complex in eaf1Δ cells can account for the reduced UAS occupancy of NuA4 subunits in this mutant. By contrast, it appears that the partial NuA4 complex present in eaf1Δ cells can be recruited efficiently to coding sequences in the absence of Gcn4-Tra1 interactions at the UAS. Presumably, other NuA4 subunits interact with histones, Pol II, or other elongation factors to mediate the ORF association of NuA4.
To demonstrate conclusively that association of NuA4 with coding sequences depends on transcription, we impaired PIC assembly by deleting the TATA element at ARG1. This mutation (TATAΔ) eliminates Pol II at the promoter, and it reduces, but does not abolish, Pol II occupancy in the coding sequences, presumably owing to a cryptic promoter in the ORF (43, 44). Importantly, TATAΔ reduces the occupancies of NuA4 subunits Eaf1 and Epl1 in the ORF, but not at the UAS (Fig. 9A and B), confirming that NuA4 associates with ARG1 coding sequences cotranscriptionally. The fact that NuA4 UAS occupancy is unaffected by TATAΔ implies that NuA4 recruitment to the UAS is independent of PIC assembly.
The C-terminal domain (CTD) of Pol II subunit Rpb1, comprised of tandem repeats of the heptad Y1S2P3T4S5P6S7, is a scaffold for recruiting various factors involved in the elongation phase of transcription in a manner stimulated by phosphorylation of CTD residues (41). CTD phosphorylation on Ser5 by cyclin-dependent kinase 7/Kin28 of TFIIH at the promoter stimulates cotranscriptional recruitment of the H3-Lys4 methyltransferase Set1 (38), elongation factor Paf1C (43), and SAGA (21), among other factors, early in the elongation phase of transcription (41). Hence, we asked whether NuA4's association with ARG1 coding sequences is reduced by impairing Kin28 activity with the kin28-ts16 mutation.
As expected (41, 43), WT cells exhibited higher occupancies of Ser5-phosphorylated Rpb1 (Ser5P) in the promoter and 5′ORF than in the 3′ORF of ARG1. Furthermore, the Ser5P occupancies are strongly reduced in kin28-ts16 cells at the restrictive temperature (Fig. (Fig.9C).9C). This effect is associated with a moderate decline (<2-fold) in Rpb3 occupancy in the ORF but not the promoter (Fig. (Fig.9D),9D), consistent with decreased promoter clearance or escape by Pol II (55). The 3′ ORF occupancy of Eaf1 was reduced to a greater extent than was Pol II occupancy in the kin28-ts16 mutant, leading to a significant reduction in the Eaf1/Rpb3 ratio at the 3′ end of the ARG1 ORF (Fig. (Fig.9E).9E). Importantly, we saw a similar decrease in the NuA4:Rpb3 occupancy ratios at the ADH1 coding sequences in kin28-ts16 cells (Fig. (Fig.9F).9F). At this gene, kin28-ts16 strongly reduced the occupancy of Ser5P without significantly decreasing Pol II (Rpb3) occupancy in the ORF (data not shown). Together, these findings indicate that cotranscriptional recruitment of NuA4 is stimulated by Ser5 CTD phosphorylation by Kin28.
In support of this last conclusion, we found that Rpb1 phosphorylated on the CTD at Ser5 or Ser2 (Ser2P) specifically coimmunoprecipitated with myc-Eaf1 from WCE (Fig. (Fig.9H).9H). As a negative control in this experiment, we used a strain expressing myc-tagged Rli1, a ribosome-associated protein that is not expected to interact with chromatin or Pol II. This allows us to control for any nonspecific interactions of Pol II or histones with the anti-myc-coated resin. These results suggest that NuA4 interacts directly with the elongating form of Pol II. A smaller proportion of hypophosphorylated Rpb1 was also specifically associated with myc-Eaf1 (compare input and pellet signals), which might reflect Pol II-NuA4 interaction at the promoter preceding CTD phosphorylation (Fig. (Fig.9H9H).
Esa1 and Eaf3 each contain a chromodomain, and Yng2 contains a plant homeobox domain, which can interact with methylated histones (24, 25, 54). Consistent with this, the H3-Lys4 and H3-Lys36 methyltransferases Set1 and Set2, respectively, were found to enhance Esa1 recruitment and H4-Lys18 acetylation at the MET16 promoter (35). Considering that H3 methylation occurs cotranscriptionally, Set1 and Set2 could also be critical for NuA4 recruitment to coding sequences. However, we observed no significant effect of a set1 set2Δ double mutation on Eaf1 or Epl1 occupancies at the 3′ end of ARG1 (Fig. 10A and B). Moreover, we observed only moderate reductions in Eaf1 occupancy in the GAL1 ORF during induction by galactose, and in the coding sequences of constitutively transcribed ADH1 (Fig. 10C and D). To examine this issue more broadly, we determined the effects of eliminating Set1/Set2 on coimmunoprecipitation of bulk Ser5-phosphorylated Rpb1 and histone H3 with Myc-Eaf1 from WCE. Interestingly, Eaf1 association with H3 was strongly reduced, whereas Eaf1 association with Ser2P was not diminished by set1 set2Δ (Fig. 10E, lanes 2 and 5). This finding suggests that NuA4 is recruited to coding sequences by the phosphorylated Pol II CTD, and its subsequent association with nucleosomes is stabilized by H3 methylation. Supporting this hypothesis, we found that H4 acetylation in the ARG1 ORF was reduced in set1Δ set2Δ cells (Fig. 10F), consistent with reduced NuA4 KAT activity in the coding sequences despite the high-level NuA4 occupancy observed at ARG1 in the absence of both H3 methyltransferases (Fig. 10A and B).
In this paper, we provide evidence for a pathway of acetylation-mediated nucleosome remodeling and eviction in transcribed coding sequences that stimulates elongation by Pol II in vivo. We obtained several lines of evidence that H3 acetylation by SAGA and H4 acetylation by NuA4 both contribute to efficient elongation in vivo. Thus, esa1 and gcn5Δ each confer sensitivity to 6-AU and provoke gene length-dependent defects in transcription, and these phenotypes are exacerbated in a double mutant impaired for both KATs. Using an in vivo assay to measure the rate of Pol II elongation, we found that esa1 reduces the rate of elongation and exacerbates a smaller reduction in elongation rate in gcn5Δ cells. Mutations in numerous elongation factors have been examined with this assay, including Paf1C subunits, Spt4, TFIIS, THO/TREX subunits, CTK1, and SWI/SNF, but none reduced the Pol II elongation rate (32). Thus, apart from a mutation in Rpb1 of Pol II itself, these are the first mutations described in yeast that appear to reduce the elongation rate in vivo. It is possible that the effects of the esa1 mutation on transcription elongation involve impaired recruitment of SAGA. However, the fact that esa1 exacerbates the transcription elongation defects observed in gcn5Δ cells (lacking the SAGA KAT) indicates that NuA4 and SAGA KAT activities contribute independently to transcription elongation. Consistent with this conclusion, we found that eliminating Gcn5 does not impair recruitment of NuA4 to Gcn4 target genes (data not shown).
Because nucleosomes impede Pol II elongation and are evicted during transcription, mutations in chromatin modifying factors might be expected to reduce the elongation rate by reducing nucleosome disassembly (32). Indeed, we made the novel finding that loss of H4 acetylation in esa1, in addition to loss of H3 acetylation in gcn5Δ, dampens nucleosome eviction from the GAL1 ORF during transcription. Histone acetylation could stimulate nucleosome eviction by weakening histone-DNA interactions or creating a binding surface for BD-containing chromatin remodeling complexes. Previous genome-wide localization of RSC identified primarily Pol II promoters and Pol III genes as binding sites (14, 37). However, we found that RSC occupancies at GAL1 are substantially higher in the ORF versus UAS. While this work was in progress, RSC was shown to be recruited exclusively to the coding sequences of stress-induced genes regulated by mitogen-activated protein kinase Hog1 (31). Hence, ORF association by RSC could be a characteristic of highly inducible genes. SWI/SNF has also been shown to occupy coding sequences and affect Pol II occupancy in the ORF (20, 52), but the role of histone acetylation in its cotranscriptional recruitment was unknown.
We found that recruitment of RSC and SWI/SNF to the coding regions of GAL1 was enhanced to varying degrees by H3 and H4 acetylation by Gcn5 and Esa1. These results provide in vivo evidence supporting the idea that acetylated nucleosomes provide an optimal substrate for cotranscriptional recruitment of RSC and SWI/SNF via their BD-containing subunits. Previous work showed that Rsc4 tandem BDs bind H3 tail peptides acetylated on Lys14 and suggested that H3-Lys14 acetylation by Gcn5 is more critical than H4 acetylation by Esa1 for nucleosome binding of Rsc4 BDs (26). It could be that BDs in other RSC subunits are more dependent on acetylated H4 to explain the strong effects of esa1 on RSC recruitment to GAL1 observed here. Indeed, SAGA and NuA4 both stimulate RSC recruitment to a nucleosomal template in vitro (9). The association of RSC with coding sequences might also be enhanced by direct association of RSC with elongating Pol II via Rsc4 interaction with Rpb5 (56).
Consistent with our finding of RSC recruitment to coding sequences, we observed phenotypes indicating elongation defects in strains lacking the Rsc1- or Rsc2-containing forms of these complexes, including increased sensitivity to 6AU and gene length-dependent defects in transcript accumulation. Interestingly, mutational inactivation of RSC subunit Rsc9 appeared to diminish the progression of elongating Pol II through the 3′ end of the STL1 ORF, a stress-induced gene with high RSC occupancy of coding sequences, providing further evidence that RSC can stimulate elongation in vivo (31). The esa1 and rsc2Δ mutations produced nonadditive phenotypes in the gene-length dependent reporter assays, consistent with the possibility that Esa1 stimulates elongation by promoting RSC recruitment or function. However, we have not detected a decreased rate of Pol II elongation in the PGAL1-YLR454w ORF after promoter shutoff in rsc1Δ or rsc2Δ mutants (data not shown). This implies that the effect of esa1 in reducing the elongation rate does not arise simply from impaired RSC recruitment, although it must be recalled that none of these mutations eliminates RSC function. We did observe a small decrease in Pol II occupancy toward the 3′ end of the 8-kb PGAL1-YLR454w ORF under inducing conditions in rsc2Δ cells (Fig. 5D to F), suggesting a decrease in Pol II processivity (32). Like rsc2Δ, mutations in several elongation factors were shown previously to impair processivity without decreasing elongation rate (32). To explain the gene length-dependent reductions in transcription observed in rsc1Δ and rsc2Δ cells, it might be necessary to posit defects in another aspect of mRNA biogenesis during elongation in addition to Pol II processivity, such as cotranscriptional mRNP assembly.
Several studies showed that Esa1 and Epl1 are recruited to UASs, but it was unclear whether this occurs in the context of NuA4, picNuA4, or both. By ChIP analysis of Eaf1 (unique to NuA4), we provided direct evidence that NuA4 is recruited to UASs by Gal4 and Gcn4. Our unexpected findings that eaf1Δ does not completely disrupt NuA4 but impairs association of Tra1 with the rest of the complex, and also reduces NuA4 recruitment to the UAS, provide in vivo evidence for the importance of Tra1 in NuA4 recruitment by an activator. The finding in other studies that NuA4 was completely disrupted in eaf1Δ extracts (3, 33) might be explained by the fact that our coimmunoprecipitation assay involves rapid, small-scale affinity purification without an elution step, perhaps allowing us to preserve a labile, incompletely disrupted NuA4 complex.
Previous ChIP studies suggested that Esa1 and Epl1 are restricted to promoters (46, 48). Motivated by our findings that Esa1 stimulates elongation, we provided the first evidence that NuA4 subunits (including Eaf1) occupy the coding sequences of both constitutive and inducible genes at levels comparable to their corresponding UAS occupancies. It could be argued that gene looping, wherein the 5′ end 3′ ends of the gene are physically connected (2), contributes to NuA4 occupancy of coding sequences; however, our data suggest that this contribution is either small or nonexistent. First, we found that NuA4 occupancy is detected across the ARG1 ORF and is not limited to only the 5′ and 3′ ends (Fig. (Fig.6C).6C). Second, we were unable to detect the activator Gcn4 in coding sequences (data not shown), indicating that not all promoter-bound factors are detected in the ORF by our ChIP protocol. Most importantly, we found that disruption of NuA4 promoter occupancy in eaf1Δ does not affect its ORF occupancy (Fig. 8A to E). Taken together, these data suggest that NuA4 is recruited to coding sequences independent of gene looping, as we showed previously for SAGA (21).
We also established that the ORF association of NuA4 occurs cotranscriptionally and is stimulated by Ser5 CTD phosphorylation by Kin28, and we further demonstrated that NuA4 is associated in vivo with Rpb1 phosphorylated on Ser5 or Ser2, which represent elongating forms of Pol II. NuA4 also coimmunoprecipitated with hypophosphorylated Rpb1, which should represent Pol II at the promoter. We suggest that NuA4 interacts with Pol II at the promoter in a manner enhanced by its association with activators or coactivators bound to the UAS. Lacking these interactions in the coding sequences, NuA4's interaction with elongating Pol II would be enhanced by the phosphorylated CTD.
Recruitment of NuA4 to coding sequences has been suspected because it contains three subunits harboring chromodomains or PHD fingers, and H3 methylation by Set1 and Set2 occurs cotranscriptionally. Surprisingly, we found that eliminating both Set1 and Set2 produces little or no reduction in NuA4 occupancy of coding sequences at several different genes and only a moderate reduction in H4 acetylation at ARG1, as assayed by ChIP (Fig. 10C, D, and F). Furthermore, set1Δ set2Δ had no effect on NuA4's association with bulk elongating Pol II phosphorylated on Ser2 of the CTD in coimmunoprecipitation assays (Fig. 10E). By contrast, H3 methylation by Set1 or Set2 was required for measurable coimmunoprecipitation of NuA4 with H3 (Fig. 10E). One way to account for these findings is to propose that elimination of Set1 and Set2 does not abolish NuA4-H3 association, but it increases the off rate of the interaction in a manner that reduces H4 acetylation by NuA4. This last defect would be partially offset by impaired recruitment of the histone deacetylase complexes Set3C and Rpd3-S, which also depends on H3 methylation by Set1 or Set2, respectively (10, 27, 28). This scenario can account for the observed stronger reduction in NuA4-H3 association observed in our coimmunoprecipitation assays, which involve extensive washing of immune complexes, compared to ChIP assays, where interactions are preserved by cross-linking. However, our finding that set1Δ set2Δ has no detectable effect on association of NuA4 with elongating Pol II leads us to suggest a two-step mechanism for cotranscriptional association of NuA4 with nucleosomes, wherein NuA4 is recruited to sites of transcription by the phosphorylated Pol II CTD and then binds to nucleosomes containing methylated H3 tails to facilitate H4 acetylation (Fig. (Fig.11).11). This in turn weakens histone-DNA contacts and promotes recruitment of RSC and SWI/SNF through their BDs, the combination of which facilitates nucleosome eviction and enhances the rate of transcription elongation in vivo.
We thank Lorraine Pillus, Sebastián Chávez, Brehon Laurent, Kevin Struhl, Mark Solomon, Joseph Reese and Jacques Côté for strains or plasmids. We thank Fan Zhang and Hongfang Qiu for advice and help with experiments.
This work was supported by the Intramural Research Program of the NIH.
Published ahead of print on 12 October 2009.