The Eaf3/5/7 Subcomplex Stimulates NuA4 Interaction with Methylated Histone H3 Lys-36 and RNA Polymerase II*

NuA4 is the only essential lysine acetyltransferase complex in Saccharomyces cerevisiae, where it has been shown to stimulate transcription initiation and elongation. Interaction with nucleosomes is stimulated by histone H3 Lys-4 and Lys-36 methylation, but the mechanism of this interaction is unknown. Eaf3, Eaf5, and Eaf7 form a subcomplex within NuA4 that may also function independently of the lysine acetyltransferase complex. The Eaf3/5/7 complex and the Rpd3C(S) histone deacetylase complex have both been shown to bind di- and trimethylated histone H3 Lys-36 stimulated by Eaf3. We investigated the role of the Eaf3/5/7 subcomplex in NuA4 binding to nucleosomes. Different phenotypes of eaf3/5/7Δ mutants support functions for the complex as both part of and independent of NuA4. Further evidence for Eaf3/5/7 within NuA4 came from mutations in the subcomplex leading to ∼40% reductions in H4 acetylation in bulk histones, probably caused by binding defects to both nucleosomes and RNA polymerase II. In vitro binding assays showed that Eaf3/5/7 specifically stimulates NuA4 binding to di- and trimethylated histone H3 Lys-36 and that this binding is important for NuA4 occupancy in transcribed ORFs. Consistent with the role of NuA4 in stimulating transcription elongation, loss of EAF5 or EAF7 resulted in a processivity defect. Overall, these results reveal the function of Eaf3/5/7 within NuA4 to be important for both NuA4 and RNA polymerase II binding.

Acetylation is one of the key histone modifications involved in activation of transcription. Acetylation of lysine residues reduces the positive charge on the histones, weakening the electrostatic interactions between histones and DNA (1). It has also been shown that acetylation stimulates interaction between bromodomain-containing chromatin remodeling complexes and nucleosomes, leading to histone eviction (2,3). Nucleosomal acetylation is important for both transcription initiation and elongation. Lysine acetyltransferase (KAT) 2 mutations that reduce acetylation have been shown to result in both transcription initiation and elongation defects (3,4).
Although acetylation is required for efficient transcription, too much acetylation can be deleterious. Hyperacetylation has been shown to be associated with cryptic transcription (5). Thus, it is essential to control the level of acetylation at a gene. Both KATs and histone deacetylase complexes (HDACs) are recruited to transcribed genes to dynamically control acetylation (3,4,6). HDACs prevent runaway acetylation and cryptic transcription, allowing for nucleosome reassembly behind the transcribing RNA polymerase II (Pol II).
Eaf3, the yeast homolog of Drosophila MSL3 (7) and human MRG15 (8), is an important regulator of acetylation because it is a subunit of both the NuA4 KAT and Rpd3C(S) HDAC (5,7). NuA4 acetylates histones H4 and H2A (9) as well as 95 other non-histone targets (10). The complex stimulates transcription of a number of genes (3,11,12) and is also involved in DNA repair and cell cycle progression (13,14).
NuA4 contains 13 subunits, 5 of which are essential. The catalytic subunit Esa1 is the only essential KAT in yeast (9). The subunits of NuA4 are organized into four subcomplexes held together by Eaf1, the only protein unique to NuA4 (15). The piccolo NuA4 (picNuA4) complex containing Esa1, Yng2, and Epl1 exists as both a part of NuA4 and an independent complex thought to be responsible for global basal levels of H4 acetylation (16). Yng2 stimulates Esa1 KAT activity and is actually stabilized by NuA4-mediated acetylation (17,18). The other NuA4 subcomplexes share subunits with the SAGA, INO80/ SWR1, and Rpd3C(S) complexes. Eaf3 is found in a subcomplex with Eaf5 and Eaf7, with Eaf5 tethering the subcomplex to NuA4 and Eaf7 connecting Eaf3 and Eaf5 (15,19). Eaf3, Eaf5, and Eaf7 may also function as an independent complex named TINTIN (19,20).
In the context of Rpd3C(S), Eaf3 works with Rco1 to enable the complex to bind di-and trimethylated H3K36 in coding sequences (21). Rpd3C(S) has also been shown to be recruited to coding sequences through interaction with the C-terminal domain of Pol II subunit Rpb1 (CTD) phosphorylated on serine 5 (Ser(P)-5) (6). The complex may also be recruited to promoters stimulated by H3K4 di-and trimethylation (22). Rpd3C(S) functions to deacetylate histones following Pol II passage to allow nucleosome reassembly and prevent cryptic initiation (23). Rpd3 can deacetylate all histone lysines except for histone H4 Lys-16 (24).
We have previously shown that NuA4 interaction with nucleosomes and recruitment to some coding sequences is stimulated by H3K4 dimethylation and H3K36 di-and trimethylation. Although NuA4 binding to chromatin was stimulated by H3 methylation, NuA4 was able to bind to both histone tail peptides and nucleosomes lacking H3 methylation in vitro, suggesting that NuA4 can interact with nucleosomes with a range of histone modifications (25). Additionally, NuA4 recruitment to coding sequences is stimulated by Ser(P)-5, although NuA4 seems to be able to interact with multiple phosphorylated forms of the CTD (3).
Although we have shown that NuA4 can bind nucleosomes with multiple modifications as well as the CTD, little is known about what subunits are important for these interactions. Cryo-EM data with unmodified nucleosomes showed that NuA4 binding along the edge and face of the nucleosome and picNuA4 binding to recombinant nucleosomes involves both Yng2 and Epl1 (26). The isolated recombinant Yng2 plant homeobox domain preferentially bound di-and trimethylated H3K4 peptides (27,28). The Yaf9 YEATS domain may be involved in binding acetylated histone H3 (29). Arp4 has been shown to bind nucleosomes in vitro (30). The Esa1 chromodomain has been shown to preferentially bind the unmodified H3 tail over the tail carrying methylated Lys-9 (31).
Although Eaf3 has been shown to be involved in stimulating Rpd3C(S) and TINTIN binding to H3K36me2/3 and TINTIN binding to the Pol II CTD (19,32), its role in NuA4 interaction with nucleosomes has not been thoroughly characterized. To better understand the function of the Eaf3/5/7 subcomplex, we examined its role in NuA4 binding to nucleosomes, recruitment to transcribed genes, and effect on H4 acetylation.

Results
Previous work has demonstrated that Eaf3 can bind methylated H3K36 in the context of Rpd3C(S) (21) and that the Eaf3/ 5/7 complex can bind both methylated H3K36 and the Pol II CTD (19). We hypothesized that these three proteins are the reason that NuA4 binding to nucleosomes is stimulated by H3K36 methylation and that they may also contribute to NuA4-Pol II interactions. We initially wanted to understand how much of the function of Eaf3, Eaf5, and Eaf7 occurred as part of NuA4 and how much was as an independent complex.
Eaf3/5/7 Functions as Part of NuA4 -To better understand how much of the function of the Eaf3/5/7 complex was as part of NuA4 and how much was as an independent complex, we compared the phenotypes of strains missing EAF3, EAF5, and EAF7 individually or in pairs ( Fig. 1A) with strains affecting NuA4 integrity (eaf1⌬) (15) or KAT activity (yng2⌬ and esa1 L254P ) ( Fig. 1B) (13,17). Because ESA1 is essential, we are using a temperature-sensitive mutant with an L254P mutation, which will henceforth be referred to as esa1. We expected eaf1⌬, yng2⌬, and esa1 cells to represent loss of NuA4 function. Both eaf1⌬ and yng2⌬ cells exhibited a slow growth phenotype on YPD at 30°C (Fig. 1B, YPD 30°C). This observation is consistent with the idea that Eaf1, Yng2, and Esa1 are required for NuA4 integrity and overall function, whereas the primary function of Eaf3/5/7 within NuA4 is to stimulate binding to the subset of nucleosomes methylated on H3K36.
The overlapping but not identical phenotypes of eaf3/5/7⌬ and eaf1⌬/yng2⌬/esa1 mutants reflect the presence of Eaf3 in Rpd3C(S), NuA4, and a possible independent Eaf3/5/7 complex. The fact that eaf3/5/7⌬ mutants showed phenotypes similar to but less severe than those of eaf1⌬/yng2⌬/esa1 at 42°C and on formamide, MMS, and rapamycin suggests that these three proteins function as part of NuA4 and that their loss impairs NuA4 function less than loss of subunits that affect overall complex integrity or KAT activity. Finally, the different phenotypes seen for eaf5⌬ and eaf7⌬ as compared with eaf1⌬/ yng2⌬/esa1 on caffeine suggest that they probably do function in both an independent complex and NuA4.
Whereas the effect of eaf3⌬ only became apparent in the absence of Rpd3C(S), the other eaf3/5/7⌬ mutants all reduced H4 acetylation ϳ40% (Fig. 2, A (cf. lanes 1-7) and C). The loss of acetylation in these mutants suggests that they play a role in NuA4-mediated H4 acetylation. The fact that these mutations all decrease H4 acetylation to the same degree is consistent with Eaf3/5/7 working together as a subcomplex within NuA4. Whereas the eaf3/5/7⌬ mutants reduced H4 acetylation ϳ40%, the eaf1⌬ and yng2⌬ mutants that affect NuA4 integrity and KAT activity reduced acetylation ϳ85% (Fig. 2, A (cf. lanes 1, 10, and 11) and C). This is consistent with the relative magnitude of the phenotypes observed in these two sets of mutants as well as their functions in NuA4.
At 30°C, esa1 cells had an H4 acetylation defect similar to that of eaf3/5/7⌬ mutants but showed a significantly greater defect (ϳ90%) after 4 h at 37°C (Fig. 2, A, B (cf. lanes 1 and 12), and C). Again, consistent with the phenotypes at 37°C, none of the other mutants showed a significant difference in H4 acetylation between 30 and 37°C. The greater H4 acetylation defects in eaf1⌬, yng2⌬, and esa1 cells, as compared with eaf3/5/7⌬ mutants, reflect the fact that Eaf1, Yng2, and Esa1 affect overall NuA4 integrity or KAT activity, whereas the Eaf3/5/7 subcomplex is probably important for NuA4 interaction with the subset of chromatin containing methylated H3K36.
Eaf3/5/7 Stimulates NuA4 Interaction with Nucleosomes and Pol II in Vivo-The decrease in H4 acetylation that we observed in eaf3/5/7⌬ mutants could be due to the subcomplex stimulating NuA4 activity or interaction with chromatin. Previous work on Rpd3C(S) and the independent Eaf3/5/7 complex has shown that Eaf3 contributes to their interaction with methylated H3K36 (19,21). We used coimmunoprecipitation (co-IP) FIGURE 2. The Eaf3/5/7 subcomplex stimulates H4 acetylation. Shown is H4 acetylation in bulk histones. A, strains BY4741, DGY392, 259, 2940, DGY831, DGY827, DGY828, 6209, DGY844, DGY326, DGY304, and DGY150 were grown to stationary phase in SC, and whole cell extracts were subjected to Western blotting analysis with antibodies against H4-Ac and the H4 C terminus (H4). B, yeast were treated as in A, except they were diluted 1:10 in fresh SC and grown for 4 h at 37°C before making whole cell extracts. C, films were scanned, and band intensity was quantitated using ImageJ. H4 acetylation was calculated as the ratio of H4-Ac to H4 band intensity. Columns represent an average of six independent experiments. Error bars, S.D. All mutant strains except for eaf3⌬ had significantly different levels of H4-Ac than WT as determined by Student's t test (*, p Ͻ 0.01). The rco1⌬eaf3⌬ cells had significantly lower levels of H4-Ac compared with rco1⌬ cells ( †, p Ͻ 0.01), as determined by Student's t test. In esa1 cells, significantly less H4-Ac was found at 37°C compared with 30°C (#, p Ͻ 0.01) as determined by Student's t test.
to examine NuA4-nucleosome interactions in the presence and absence of Eaf3/5/7 subunits. We compared the interaction between Myc-Eaf1 or Myc-Rli1 and histone H3. Because Eaf1 is the only subunit unique to NuA4 and is important for complex integrity (15), the presence of Eaf1 should be indicative of the presence of the entire NuA4 complex. Rli1 is a ribosome-associated protein that functions in the cytoplasm and has no role in transcription (38) that served as a negative control for nonspecific Myc tag interactions.
We previously proposed a two-stage mechanism for NuA4 recruitment to coding sequences in which NuA4 first binds the Pol II CTD and then preferentially binds to nucleosomes in coding sequences stimulated by H3K4 and H3K36 methylation (25). Both Rpd3C(S) and the independent Eaf3/5/7 complex were shown to bind the Pol II CTD (6,19). Thus, reduced binding to nucleosomes in eaf3/5/7⌬ mutants could be due to reduced interaction with Pol II. We examined binding of NuA4 to Pol II by pulling down Myc-Eaf1 and measuring the amount of coprecipitated CTD phosphorylated on either serine 2 (Ser(P)-2) or serine 5 (Ser(P)-5) or Rpb3, another Pol II subunit. NuA4-Pol II binding decreased significantly in all eaf3/5/7⌬ mutant strains to the same degree as NuA4-H3 binding (Fig. 3, C and D). It is interesting that the same subcomplex is able to stimulate NuA4 interaction with both nucleosomes and Pol II, although the Eaf3 chromodomain has only been shown to bind methylated H3K36 (32).
Eaf3/5/7 Stimulates NuA4 Binding to Methylated H3K36 in Vitro-NuA4 interaction with nucleosomes in coding sequences may be dependent on recruitment to coding sequences through binding to the Pol II CTD. It is possible that NuA4 interaction with nucleosomes in vivo was reduced in eaf3/5/7⌬ mutants because of defects in Pol II binding. To examine whether Eaf3/5/7 is actually involved in NuA4nucleosome interactions, we performed in vitro nucleosome pull-downs. Chromatin containing TAP-H4 from whole cell extracts of HHF2-TAP or untagged strains was immobilized on IgG-coated beads. These beads containing immobilized chromatin were washed and incubated with whole cell extracts from EAF1-Myc strains. Binding of NuA4 to immobilized chromatin was measured by Western blotting. To confirm that we were actually measuring NuA4 binding to chromatin and not to free histones or H3-H4 dimers, we performed Western blotting on the IgG bead-bound fraction after incubation with TAP-H4 whole cell extracts. We were able to detect all four histones, including both TAP-tagged and untagged H4 (Fig. 4C), suggesting that we were looking at nucleosomes, rather than just free histones.
Consistent with the co-IP results, NuA4 binding to chromatin from WT HHF2-TAP cells was decreased ϳ50% in all three eaf3/5/7⌬ mutants (Fig. 4, A (cf. lanes 6 -10) and B (WT)). This interaction should not require Pol II-mediated recruitment to coding sequences, so this result supports the hypothesis A and C, whole cell extracts of mid-log phase SC cultures of strains DGY3, DGY400, DGY741, DGY835, DGY837, and YDH353 were immunoprecipitated using anti-H3 (A) or anti-Myc (C) antibodies. Immune complexes were subjected to Western blotting analysis with anti-Myc, anti-H3, and anti-Rpb3 antibodies as well as antibodies against the Ser(P)-5 (S5p) and Ser(P)-2 (S2p) forms of the Pol II CTD. B, films were scanned, and band intensity was determined using ImageJ. Relative binding was calculated as the ratio of Myc pellet to input band intensities divided by the band intensity for the H3 pellet. Binding in mutants was normalized to WT binding. Columns represent the mean of eight independent experiments. Error bars, S.D. All mutants significantly decreased NuA4-H3 interactions, as determined by Student's t test (*, p Ͻ 0.01). D, relative binding was calculated similarly to B except dividing the pellet to input ratio by the Myc pellet band intensity for each coprecipitated protein. Columns represent the mean of 15 independent experiments, whereas error bars show the S.D. All mutations significantly reduced interaction between Myc-Eaf1 and histone H3 or Pol II as determined by Student's t test (*, p Ͻ 0.01).
In addition to performing the nucleosome pull-down assays with chromatin from WT HHF2-TAP cells, we also used chromatin from cells lacking the histone methyltransferases Set1 and Set2. Set1 is responsible for H3K4 methylation (39), and Set2 is responsible for H3K36 methylation (40). Thus, set1⌬ cells have no H3K4 methylation, set2⌬ cells have no H3K36 methylation, and set1⌬set2⌬ cells lack methylation on both lysines. If Eaf3/5/7 is the only part of NuA4 responsible for the complex binding to methylated H3K36, eaf3/5/7 mutants should reduce NuA4 binding to chromatin from WT and set1⌬ cells but not set2⌬ cells.
WT NuA4 bound most strongly to WT chromatin. Loss of H3K4 methylation, H3K36 methylation, or both marks significantly reduced interactions by ϳ50% (Fig. 4, A (cf. lane 6) and B). This is consistent with previous work showing that both methylation marks stimulate NuA4 interaction with nucleosomes (25). Deletion of any of the Eaf3/5/7 subcomplex subunits significantly reduced NuA4 binding to chromatin from set1⌬ and set1⌬set2⌬ cells but not from set2⌬ cells (Fig. 4, A (cf.  lanes 6 -10) and B). This result suggests that the nucleosomal binding target for Eaf3/5/7 within NuA4 is methylated H3K36. The 50% reduction in WT NuA4 binding to chromatin from set2⌬ cells is probably due to loss of the interaction between H3K36 and the Eaf3/5/7 module, because no further reduction in binding was observed in strains lacking Eaf3/5/7 subunits.
Because Eaf3/5/7 stimulates NuA4 binding to methylated H3K36, Eaf3 is likely to adopt the same conformation in NuA4 as it does in Rpd3C(S) and in the independent Eaf3/5/7 complex.
Eaf3/5/7 Stimulates Binding of NuA4 to Di-and Trimethylated H3K36 -Although the co-IP and nucleosome pull-down results suggested that Eaf3/5/7 stimulates NuA4 binding to methylated H3K36, it was not clear whether the subcomplex preferentially targeted the mono-, di-, or trimethylated form. Dimethylation is enriched in the middle of ORFs, and trimethylation is more prevalent toward the 3Ј-end of ORFs (41). NuA4 occupancy is observed along the full length of the ORF (3), so it is likely to bind to nucleosomes with multiple histone modifications. We have previously shown that NuA4 can bind to unmodified nucleosomes but that its binding is stimulated by H3K36 di-and trimethylation (25). It is likely that this binding is through the Eaf3/5/7 subcomplex.
To test the specific methylation marks recognized by the Eaf3/5/7 subcomplex, we performed a peptide pull-down experiment using biotinylated peptides corresponding to amino acids 1-21 or 21-44 of histone H3 containing unmodified, or mono-, di-, or trimethylated Lys-4 or Lys-36. These peptides were immobilized on streptavidin-coated beads and incubated with TAP-Eaf1 NuA4 complex purified from WT, eaf3⌬, eaf5⌬, or eaf7⌬ cells. During the purification process, protein A is cleaved from the TAP tag, leaving just Eaf1 fused to the calmodulin binding domain (CBD-Eaf1) (42). Binding to the peptides was analyzed by Western blotting using antibodies against the TAP tag (CBD-Eaf1), Esa1, and Yaf9. The fact that the amounts of CBD-Eaf1, Esa1, and Yaf9 bound to the peptides varied in parallel suggests that NuA4 is intact in these assays. This is consistent with previous reports that loss of the Eaf3/5/7 subcomplex does not affect the association of other subcomplexes with Eaf1 (19).
We initially performed these binding assays in buffer containing 500 mM (H3K4 peptides) or 400 mM (H3K36 peptides) NaCl to be able to differentiate NuA4 binding to the different methylation states. We were able to detect NuA4 binding above background to all of the peptides except trimethylated H3K4 (Fig. 5). NuA4 binding to the unmodified and mono-and dimethylated H3K4 peptides was ϳ9-fold over background (Fig.  5C). Interactions with the unmodified or methylated H3K36 peptide were ϳ6-fold above background. Further methylation of H3K36 (di-or trimethylation) strongly stimulated binding, increasing it by another 6-fold (Fig. 5D). Consistent with its role in Rpd3C(S), loss of Eaf3 reduced NuA4 binding to H3K36me2/3 by ϳ65% (Fig. 5D). NuA4 lacking Eaf3 did not display a defect in binding to any of the other peptides.
Eaf3 is probably the subunit of NuA4 responsible for interaction with methylated H3K36, and it has been shown that Eaf7 and Eaf5 are required for tethering Eaf3 to the complex (19). Thus, Eaf3 should be lost from NuA4 in eaf5⌬ and eaf7⌬ mutants, resulting in H3K36me2/3 binding defects similar to those observed with NuA4 lacking Eaf3. To test this hypothesis, we repeated the peptide pull-downs, this time carrying out binding in buffer containing 200 mM NaCl. In the lower salt buffer, NuA4 bound to all of the H3(1-21) peptides, regardless of methylation state ϳ5-6-fold over background (Fig. 6, A  (WT, lanes 1-11) and B). NuA4 interaction with the H3  peptides was also unaffected by Lys-36 methylation, with NuA4 binding ϳ4-fold over background to all four peptides (Fig. 6, A  (WT lanes 12-19) and B).
Similar to NuA4 lacking Eaf3, NuA4 purified from eaf5⌬ or eaf7⌬ strains bound to the di-and trimethylated H3K36 peptides ϳ50% less than WT NuA4 without showing reduced binding to any of the other peptides (Fig. 6B). The similar effects of eaf3⌬, eaf5⌬, and eaf7⌬ on NuA4 binding to histone tail peptides are consistent with the results of the co-IP and nucleosome pull-down experiments. Together, these results support a model in which the chromodomain of Eaf3 binds H3K36me2/3 as a component of NuA4 and in which Eaf5 and Eaf7 help tether Eaf3 to the complex.
Eaf3/5/7 Stimulates NuA4 Occupancy at Transcribed ORFs and Transcription Elongation-NuA4 recruitment to transcribed genes is stimulated by interaction with both Pol II and nucleosomes. NuA4 occupancy at transcribed genes is reduced in kin28 cells (3) as well as in set1⌬ and set2⌬ cells (25). Because of the role of the Eaf3/5/7 complex in stimulating binding to methylated H3K36 and Pol II, we hypothesized that it would also be important for NuA4 occupancy at ORFs. We analyzed the effect of the Eaf3/5/7 subcomplex on recruitment of NuA4 to GAL1, ADH1, PMA1, PYK1, ARG1, and ARG4 by ChIP. GAL1, ARG1, and ARG4 are inducible genes, whereas ADH1, PMA1, and PYK1 are constitutively expressed.
NuA4 occupancy was significantly reduced (ϳ50%) at the GAL1, ADH1, PMA1, and PYK1 ORFs but not at the promoters in eaf3⌬, eaf5⌬, eaf7⌬, and eaf7⌬eaf5⌬ cells (Fig. 7, A-D). The similar effects in the single and double mutants suggest that the Eaf3/5/7 subcomplex functions as a unit to stimulate NuA4 occupancy in coding sequences. At ARG1 and ARG4, NuA4 occupancy in the ORF was significantly decreased in eaf5⌬, eaf7⌬, and eaf7⌬eaf5⌬ cells but not in eaf3⌬ cells (Fig. 7, E and  F). The NuA4 occupancy defects were not likely to be caused by reduced transcription at these genes, because there was no accompanying decrease in Rpb3 occupancy (Fig. 7, G-L). These results support the model that Eaf3/5/7-mediated interactions with Pol II and nucleosomes methylated on H3K36 stimulate NuA4 recruitment to transcribed ORFs.
We have previously shown that NuA4 stimulates both transcription initiation and elongation (3). NuA4 occupancy is reduced in coding sequences in the absence of the Eaf3/5/7 complex, so we investigated whether there was a transcription elongation defect in eaf3⌬, eaf5⌬, and eaf7⌬ cells. To do this, we compared the production of a long and short transcript driven by the GAL1 promoter as part of the gene length-dependent accumulation of mRNA (GLAM) assay (43). This technique calculates a GLAM ratio, which is the expression of a 4.5-kb transcript in which the PHO5 gene has been fused to the Kluyveromyces lactis LAC4 gene divided by the expression of the 1.5-kb PHO5 gene as determined by phosphatase  assay. Mutants affecting transcription elongation will reduce PHO5::LAC4 expression more so than PHO5 expression, thus reducing the GLAM ratio. When we compared GLAM ratios in Eaf3/5/7 mutants, we found that the GLAM ratio increased in eaf3⌬ cells, indicating more efficient transcription elongation. In contrast, eaf5⌬ and eaf7⌬ both reduced the GLAM ratio ϳ50% (Fig. 8A), suggesting that they have a transcription elongation defect.
To further characterize the elongation defect in eaf5⌬ and eaf7⌬ cells, we measured the kinetics of Pol II occupancy at the ϳ8-kb P GAL1 -YLR454w ORF. When cells growing in galactose are switched to glucose, transcription initiation rapidly shuts off, whereas Pol II molecules in the ORF finish transcribing. Measuring the rate at which these last molecules of Pol II leave the ORF indicates the elongation rate in vivo (44). We induced WT, eaf5⌬, and eaf7⌬ cells grown in raffinose with galactose for 1 h and processed the samples for ChIP of Pol II subunit Rpb3 at multiple time points after the addition of 4% glucose. Rpb3 occupancies at each time point were normalized to Rpb3 occupancy in galactose.
Similar to previous findings in WT cells, Pol II rapidly left the ORF after the addition of galactose, starting with loss from the beginning of the ORF. Pol II was essentially completely gone after about 8 min (Fig. 8B) (3). There were no noticeable differences in the kinetics of Pol II elongation in eaf5⌬ or eaf7⌬ cells as compared with WT (Fig. 8, B-D). This suggests that the loss of the Eaf3/5/7 subcomplex does not lead to an elongation rate defect.
Interestingly, in WT cells, Pol II occupancy at the beginning of the ORF decreased before occupancy at the end of the ORF, probably due to those molecules moving from the beginning to the end of the ORF (Fig. 8B, 2 and 4 min). In eaf5⌬ and eaf7⌬ cells, Pol II occupancy across the ORF appeared to decrease at the same rate (Fig. 8, C and D). This could indicate a processivity defect, in which Pol II is not reaching the end of the ORF, and is consistent with previous observations (19). To confirm this, we examined Pol II occupancy at P GAL1 -YLR454w in galactose. We found small but significant decreases in Pol II occupancy at 6 and 8 kb in eaf5⌬ and eaf7⌬ cells as compared with WT (Fig. 8E).

Discussion
The NuA4 KAT complex, responsible for the targeted acetylation of histones H4 and H2A (9) among other substrates (10), plays an important role in stimulating both transcription initiation and elongation at numerous genes (3,11,12,45). NuA4 is recruited to promoters through interaction of its Tra1 subunit with activators (46), whereas its occupancy in coding sequences is stimulated by both Pol II CTD serine 5 phosphorylation and H3K4 and H3K36 methylation (3,25). We show here that the Eaf3/5/7 complex, also known as TINTIN (20), functions as a part of NuA4 to stimulate binding to both di-and trimethylated FIGURE 8. The Eaf3/5/7 subcomplex stimulates efficient transcription elongation. A, GLAM assays were performed on strains BY4741, DGY392, 259, and 2940 carrying P GAL1 -PHO5 and P GAL1 -PHO5::LAC4 reporter plasmids. Cells were grown to early log phase, and Pho5-specific activity was measured. GLAM ratios were calculated as the Pho5 activity for the long divided by the Pho5 activity for the short reporter. Columns represent the mean of two independent experiments, each with three independent cultures. Error bars, S.D. The GLAM ratio was significantly decreased in eaf5⌬ and eaf7⌬ cells as compared with WT as determined by Student's t test (*, p Ͻ 0.01). B-D, analysis of Pol II run-off from the ϳ8-kb P GAL1 -YLR454w ORF during glucose repression. WT (B), eaf5⌬ (C), and eaf7⌬ (D) strains were grown as described for Fig. 7A and treated with 4% glucose for the indicated times before cross-linking, and ChIP was performed with anti-Rpb3 antibodies and primers to amplify an intergenic region on chromosome I and P GAL1 -YLR454w. Relative occupancy was calculated, taking the ratio of band intensities of PCR products for P GAL1 -YLR454w versus the chromosome I reference and dividing by the same ratio for input samples. The values in glucose were normalized to values in galactose. Columns represent the average of three independent experiments. Error bars, S.D. E, PCRs from B-D samples in galactose were analyzed by normalizing the Rpb3 occupancy in the YLR454w ORF to occupancy at the promoter. Rpb3 occupancy at 6 and 8 kb was significantly reduced compared with WT in eaf5⌬ and eaf7⌬ strains as determined by Student's t test (*, p Ͻ 0.05).
H3K36 as well as Pol II. These interactions are important for NuA4 occupancy at transcribed ORFs, acetylation of histone H4, and efficient transcription elongation.
A role for the Eaf3/5/7 subcomplex in NuA4 interaction with methylated H3K36 and Pol II is supported by multiple lines of evidence. The fact that eaf5⌬ and eaf7⌬ mutants display less sensitivity to multiple stressors than eaf1⌬, yng2⌬, and esa1 is consistent with them working as part of NuA4 (Fig. 1). Because Eaf1 is important for NuA4 integrity and Yng2 and Esa1 are critical for KAT activity (13,15,17,47), cells with mutations in these subunits lacking intact NuA4 or full NuA4 activity are likely to have more severe phenotypes than eaf5⌬ or eaf7⌬ cells in which NuA4 is still intact. The caffeine insensitivity of eaf5⌬ and eaf7⌬ strains and sensitivity of eaf1⌬, yng2⌬, and esa1 cells are consistent with previous reports (48). The difference in phenotypes may be due to the presence of Eaf5 and Eaf7 in TINTIN or because the subunits affect different functions of NuA4. NuA4 plays roles in transcription as well as DNA repair (14). Disruption of the complex or loss of KAT activity in eaf1⌬, yng2⌬, and esa1 cells is likely to affect both processes, whereas loss of the Eaf3/5/7 subcomplex may only affect transcription. The difference in phenotypes of eaf3⌬ cells compared with the other mutants is probably due to its role in Rpd3C(S). Whereas loss of Eaf1, Yng2, or Esa1 will affect only NuA4, eaf3⌬ will have an effect on Rpd3C(S), TINTIN, and NuA4.
NuA4 is the only KAT complex in S. cerevisiae that can tetraacetylate histone H4, so any defect in NuA4 activity or ability to interact with nucleosomes should lead to decreased H4 tetraacetylation (H4-Ac). As shown previously (35), eaf3⌬ cells had similar levels of global H4-Ac in bulk histones as WT (Fig. 2). Evidence for the role of Eaf3 in NuA4 is supported by the observation that rco1⌬eaf3⌬ cells had significantly lower H4-Ac than rco1⌬ cells. Rco1 is involved in the binding of Rpd3C(S) to nucleosomes (21,36) and is important for the association of Eaf3 with the complex (5). The fact that eaf3⌬ reduced H4-Ac in an rco1⌬ background supports a role for Eaf3 in NuA4-mediated H4 acetylation.
The H4-Ac reductions seen in eaf5⌬ and eaf7⌬ mutants were modest but significant. These small reductions are probably due to the ability of NuA4 to bind nucleosomes with multiple histone modifications. Even without the ability to bind H3K36me2/3, NuA4 can still acetylate hypomethylated nucleosomes or nucleosomes with H3K4 methylation. It has also been proposed that piccoloNuA4 (picNuA4), composed of Esa1, Epl1, and Yng2, is responsible for genome-wide basal H4 acetylation (16), and this complex is not affected by mutations in EAF3/5/7. Mutations that affect picNuA4 (yng2⌬ and esa1) reduce H4-Ac to a greater degree (ϳ80%) than eaf3/5/7⌬ mutations (ϳ40%). Interestingly, we observed no loss of H4-Ac in set2⌬ cells (25), which supports the idea that NuA4 can acetylate nucleosomes in the absence of H3K36 methylation, possibly by interacting with other modified histones. The decrease seen in eaf3/5/7⌬ mutants is probably due to the fact that the subcomplex stimulates NuA4 binding to both H3K36me2/3 and Pol II. Loss of NuA4 binding directly to nucleosomes with H3K36me2/3 as well as reduced presence in coding sequences as a result of decreased binding to Pol II will contribute to reduced H4-Ac.
Although the Eaf3/5/7 subcomplex is likely to exert its effects on H4-Ac by stimulating NuA4 binding to nucleosomes, it is possible that the subcomplex could directly affect NuA4 activity. NuA4 binding to nucleosomes in vivo, as measured by co-IP, was significantly reduced in eaf3/5/7⌬ mutants (Fig. 3). This suggests that the Eaf3/5/7 subcomplex's effects on H4-Ac are due to its role in NuA4-nucleosome interactions. Interestingly, NuA4 binding to Pol II was reduced to the same degree as binding to nucleosomes in all three eaf3/5/7⌬ mutants (Fig. 3, C and D). It is possible that the nucleosome binding defects are the result of decreased Pol II binding. If interaction with the CTD is required for NuA4 binding to nucleosomes in coding sequences, Pol II binding defects would then also reduce nucleosome binding. The fact that eaf3/5/7⌬ mutants reduce nucleosome binding in vitro (Fig. 4) suggests that the subcomplex directly stimulates NuA4 binding to nucleosomes and not just to Pol II.
We have previously shown that NuA4 occupancy at the ARG1 and ADH1 open reading frames is reduced in kin28-ts cells, but not in ctk1⌬ cells, although we could detect NuA4 binding to hypophosphorylated CTD as well as Ser(P)-5 and Ser(P)-2 in vivo (3). Kin28 is responsible for Ser-5 phosphorylation, whereas Ctk1 is responsible for the majority of Ser-2 phosphorylation in yeast (49). Thus, we concluded that NuA4 interaction with Pol II was stimulated by Ser(P)-5. Although we did not detect a NuA4 occupancy defect at ARG1, Eaf7 occupancy at RPS16A, RPL16B, PMA1, and PGK1 was decreased in ctk1⌬ cells (19). The occupancy defects observed in CTD kinase mutants are confounded by the fact that these mutants affect H3K4 and H3K36 methylation (50,51), which also stimulates NuA4 occupancy at transcribed genes (25). Although the co-IP results presented herein suggest that the Eaf3/5/7 complex stimulates NuA4 binding to Pol II, it is still unclear what form of the polymerase is recognized by NuA4.
Another line of evidence supporting a role for the Eaf3/5/7 subcomplex in stimulating NuA4 binding to nucleosomes came from in vitro binding assays. NuA4 lacking Eaf3/5/7 subunits bound more weakly to WT chromatin and chromatin missing H3K4 methylation than did WT NuA4 (Fig. 4). The NuA4 mutations did not affect the ability of the complex to bind chromatin without H3K36 methylation. The fact that eaf3/5/7⌬ mutations did not affect the ability of NuA4 to bind to chromatin lacking H3K36 methylation is consistent with a model in which NuA4 binds methylated H3K36 primarily through the Eaf3/5/7 subcomplex. Removing either the NuA4 binding target or the part of NuA4 that recognizes that target should have the same effect on binding.
Surprisingly, loss of any Eaf3/5/7 subunit reduced NuA4 binding to chromatin from set1⌬set2⌬ cells lacking both H3K4 and H3K36 methylation. This suggests that in the absence of H3 methylation, this subcomplex may help NuA4 bind to some other part of the nucleosome. Chromodomains, like the one in Eaf3, have been shown to bind other targets, including methylated histone H3 Lys-9, histone H3 Lys-27, histone H4 Lys-20, unmodified H3 tails, and RNA (31,52). Thus, it is possible that the Eaf3/5/7 subcomplex stimulates NuA4 binding to multiple nucleosomal targets, with the primary target being methylated H3K36.
In the nucleosome pull-down experiments, the absence of H3K4 or H3K36 methylation reduced WT NuA4 binding to chromatin to a greater degree than observed previously, when the only significant binding defect was to chromatin lacking both H3K4 and H3K36 methylation (25). The increased binding defect observed here is probably due to the fact that we used immobilized chromatin to pull down NuA4 from whole cell extracts in this study, whereas we had previously used purified NuA4. It is possible that there are other proteins in the whole cell extract that can compete with NuA4 for binding to hypomethylated nucleosomes.
The addition of the TAP tag to histone H4 may disrupt nucleosome formation or affect histone modification. We think it unlikely that the TAP tag affects nucleosome formation, because all four core histones are pulled down by the IgG beads (Fig. 4C). Even if the TAP tag alters histone modifications, this would not affect the co-IP and histone tail peptide pull-down results, which also demonstrate that Eaf3/5/7 stimulates NuA4 binding to methylated H3K36.
The nucleosome pull-down suggested that Eaf3/5/7 stimulates NuA4 binding to methylated H3K36. Histone tail peptide pull-down experiments were used to confirm that, similar to Rpd3C(S) and TINTIN, Eaf3/5/7 stimulates NuA4 binding specifically to di-and trimethylated H3K36. These binding assays were initially performed in 500 mM NaCl for the H3(1-21) peptides and in 400 mM NaCl for the H3(21-44) peptides (Fig. 5). These high salt concentrations resulted in differential binding of NuA4 to the different methylated peptides. When binding reactions were performed in 200 mM NaCl, NuA4 bound equally to all peptides (Fig. 6). Under both salt conditions, loss of Eaf3/5/7 resulted in reduced NuA4 binding only to di-and trimethylated H3K36.
The final piece of evidence that Eaf3/5/7 stimulates NuA4 binding to methylated H3K36 was the finding that eaf3/5/7⌬ mutations reduce NuA4 occupancy in transcribed ORFs (Fig.  7). At the inducible GAL1 and constitutive ADH1, PMA1, and PYK1, NuA4 occupancy was significantly reduced by all eaf3/ 5/7⌬ mutants in the ORF but not at the promoter. At ARG1 and ARG4, eaf5/7⌬ but not eaf3⌬ mutants reduced occupancy in the ORF. Because Eaf3 is a subunit of both Rpd3C(S) and NuA4 and is important for binding of both complexes to chromatin, it is possible that these complexes compete for binding to methylated H3K36 (53). Reduced Rpd3C(S) binding in an eaf3⌬ mutant may allow more NuA4 to bind, which can potentially interact with chromatin through multiple subunits.
NuA4 occupancy at ARG1 was only reduced in set1⌬set2⌬ cells and not in the single histone methyltransferase mutants (25). If the only role of the Eaf3/5/7 subcomplex was to stimulate binding to methylated H3K36, NuA4 occupancy defects should be similar in eaf3/5/7⌬ mutants and set2⌬ cells. The fact that we did not observe reduced occupancy in set2⌬ but did in eaf5⌬, eaf7⌬, and eaf7⌬eaf5⌬ cells reflects the function of Eaf3/ 5/7 in stimulating NuA4 interaction with Pol II as well as methylated H3K36.
The function of the Eaf3/5/7 subcomplex within NuA4 seems to be to help the complex bind to chromatin containing methylated H3K36, which is predominantly found at the 3Ј-ends of genes (41). The fact that we observed a Pol II processivity defect in eaf5⌬ and eaf7⌬ cells but an elongation rate defect in esa1 cells (3) suggests that NuA4 is needed throughout the ORF to stimulate transcription. When NuA4 activity is comprised in esa1 cells, acetylation is reduced throughout the ORF, leading to a rate defect. In the absence of the Eaf3/5/7 subcomplex, the effects on NuA4 are likely to be concentrated more toward the end of the ORF, resulting in a different elongation defect.
Overall, our results provide insight into the function of the Eaf3/5/7 subcomplex as well as the mechanism by which NuA4 interacts with chromatin to participate in transcription elongation. Eaf3 consistently stimulates binding to methylated H3K36, regardless of the complex that it is a part of. It is also likely that a great deal of the function of Eaf3/5/7 is through its role in stimulating NuA4 interaction with chromatin and Pol II.

Experimental Procedures
Yeast strains used in this study are listed in Table 1. DGY392 and DGY817 were made by transforming 7143 and 2940 with NotI-digested M4755 and M4758 (54), respectively. Substitution of LEU2 or URA3 for kanMX4 was confirmed by sensitivity to G418 and growth on SC lacking leucine or uracil. EAF1-TAP strains were generated as described previously (55). Proper integration of the TAP tag was confirmed by PCR analysis of chromosomal DNA and Western blotting of WCEs with ␣-TAP antibodies. EAF1-Myc strains were created as described previously (56). The presence of Myc-Eaf1 was confirmed by PCR analysis of chromosomal DNA and Western blotting of WCEs with ␣-Myc antibodies. DGY827 and DGY828 were generated by transforming 259 and 2940 respectively with the eaf3⌬::LEU2 cassette amplified from DGY392 chromosomal DNA. Deletion of EAF3 was confirmed by PCR. DGY831 was made in a similar manner with the eaf5⌬::kanMX4 cassette amplified from 259 chromosomal DNA transformed into DGY817 and loss of EAF5 confirmed by PCR.
Coimmunoprecipitation experiments were carried out as described previously (57) with the antibodies described below. Western blots were exposed to films that were scanned, and band intensity was determined using ImageJ (58). WCEs for Western blotting analysis of histone H4 acetylation and confirmation of EAF1-Myc and EAF1-TAP strains were made by trichloroacetic acid precipitation as described previously (59).
Author Contributions-A. S. conducted most of the experiments, analyzed the results, and wrote most of the paper. P. R. conducted the histone tail peptide-binding assays. J. M. L. conducted some of the chromatin immunoprecipitation and coimmunoprecipitation experiments. L. C. conducted the experiments looking at NuA4-Pol II interaction. A. M. helped revise the paper and contributed to co-IP and ChIP experiments. S. S. also helped revise the paper and contributed to H4 acetylation Western blotting and ChIP experiments. M. J. S. assisted with spotting assays, data analysis, and revisions. D. S. G. conceived of the project, conducted the nucleosome pull-down experiments, and wrote the paper with A. S., A. M., S. S., and M. J. S.