Dynamic acetylation of all lysine-4 trimethylated histone H3 is evolutionarily conserved and mediated by p300/CBP

Abstract

Histone modifications are reported to show different behaviors, associations, and functions in different genomic niches and organisms. We show here that rapid, continuous turnover of acetylation specifically targeted to all K4-trimethylated H3 tails (H3K4me3), but not to bulk histone H3 or H3 carrying other methylated lysines, is a common uniform characteristic of chromatin biology in higher eukaryotes, being precisely conserved in human, mouse, and Drosophila. Furthermore, dynamic acetylation targeted to H3K4me3 is mediated by the same lysine acetyltransferase, p300/cAMP response element binding (CREB)-binding protein (CBP), in both mouse and fly cells. RNA interference or chemical inhibition of p300/CBP using a newly discovered small molecule inhibitor, C646, blocks dynamic acetylation of H3K4me3 globally in mouse and fly cells, and locally across the promoter and start-site of inducible genes in the mouse, thereby disrupting RNA polymerase II association and the activation of these genes. Thus, rapid dynamic acetylation of all H3K4me3 mediated by p300/CBP is a general, evolutionarily conserved phenomenon playing an essential role in the induction of immediate-early (IE) genes. These studies indicate a more global function of p300/CBP in mediating rapid turnover of acetylation of all H3K4me3 across the nuclei of higher eukaryotes, rather than a tight promoter-restricted function targeted by complex formation with specific transcription factors.

Histone acetylation is widely associated with gene regulation, specifically with active genes and a more open chromatin structure (1–3). Charge neutralization of lysines reduces the affinity of acetylated histone tails for DNA, potentially creating a more open state for transcription (3), and specific acetyllysines can form binding sites to recruit other effectors (4), suggesting distinct roles for distinct lysine residues. Most current analytical techniques provide a “snapshot” of acetylation frozen by fixation or crosslinking, with no information about the stability of modification (discussed in ref. 5). However, radiolabeling studies (6, 7) and genetic experiments in yeast (8, 9) show acetylation is highly dynamic and heterogeneous, with different populations of histones characterized by distinct rates of acetylation and deacetylation (5, 10). We have shown that all detectable K4-trimethylated histone H3 (H3K4me3) across murine nuclei, irrespective of location or association, is subject to dynamic acetylation at multiple lysine residues, becoming maximally acetylated within 10–30 min of treatment with trichostatin A (TSA) (11). Continuous turnover, by opposing lysine (K) acetyltransferase and histone deacetylase (KAT and HDAC; Fig. 1A) activities, is extremely rapid (11). The relevant KAT/HDAC enzyme system must therefore be tightly coupled to H3K4me3, and on these tails it is not restricted to any specific lysine but acetylates multiple residues (excluding the trimethylated lysine 4) to rapidly produce tetra- and pentaacetylated histones (11). Such exquisite targeting is remarkable, given the massive excess of other nucleosomes and histones available to KATs and HDACs in the nucleus.

Fig. 1.

Dynamic acetylation of all H3K4me3 is conserved in higher eukaryotes. (A) Dynamic histone acetylation is mediated by lysine acetyltransferases (KATs) and histone deacetylases (HDACs). TSA inhibits classes I, II, and IV HDACs to promote hyperacetylation of H3K4me3. (B) Human WI 38 fibroblasts (lanes 1–6), murine C3H 10T1/2 fibroblasts (lanes 7–12), or Drosophila S2 cells (lanes 13–18) were treated with trichostatin A (TSA) (33 nM; lanes 2–4, 8–10, and 14–16) or nicotinamide (NAM) (20 μM; lanes 5, 6, 11, 12, 17, and 18) as indicated. Histones were separated on acid–urea (AU) gels on which acetylation incrementally retards migration to produce the “ladder” of bands seen. Numbers indicate extent of acetylation, 0 referring to nonacetylated H3. Different exposures are shown for H3K4me3 in mouse and Drosophila (Top) due to the higher proportion of this modification in the fly (Fig. S1C for comparison). The H3K9me2 (fourth panel down) signal in mouse has been overexposed to allow detection of low levels of this modification in Drosophila. (Lanes 1, 7, and 13: control untreated cells.)

In high-resolution quantitative studies of c-fos and c-jun, histone H3 lysine-9 acetylation (H3K9ac) peaks at the promoter and 5′ end with a gap at the start site, most likely due to the absence of a nucleosome, and decreases before the end of the gene (12). This distribution precisely mirrors that of H3K4me3 across these genes (12), also shown, albeit at lower resolution, in several chromosome-wide mapping studies in yeast (13, 14), Drosophila (15), mouse (16), and human (17, 18). In TSA-treated quiescent cells, H3K4me3 across this region of c-fos and c-jun becomes rapidly hyperacetylated (11) even though the genes remain inactive. In fact, hyperacetylation inhibits physiological gene induction, challenging the link between state of acetylation and transcription and suggesting that turnover is the important factor. Consistent with this, genome-wide mapping of KATs and HDACs places these enzymes together at many gene loci (18), and a requirement for HDAC activity in gene expression has been reported (reviewed in ref. 19).

We show here that dynamic acetylation targeted to H3K4me3 is conserved in human and Drosophila as well as mouse cells. RNA interference studies in Drosophila indicate that depletion of any single HDAC does not abolish TSA-sensitive acetylation of H3K4me3. By contrast, knockdown of a single KAT, dCBP, severely reduced dynamic acetylation of H3K4me3. A new small-molecule p300/cAMP response element binding (CREB)-binding protein (CBP) inhibitor, C646 (20), was used to confirm its role mediating dynamic H3K4me3 acetylation in Drosophila and mouse and to study its function in inducible gene activation. We conclude that dynamic acetylation targeted to all H3K4me3 is evolutionarily conserved, mediated by p300/CBP, and essential for RNA polymerase II association and protooncogene induction. These studies throw light on the role that p300/CBP plays in gene regulation, indicating a more dynamic, global function across all H3K4me3-containing promoters in human, mouse, and Drosophila.

Results

All K4-Trimethylated Histone H3 in Mouse, Human, and Drosophila Cells Is Subject to Dynamic Acetylation.

All H3K4me3, but not H3 methylated at lysine 9 or bulk H3, in murine nuclei is TSA hypersensitive (11). This is visualized by Western blots of histone H3 ladders on acid–urea (AU) gels (Fig. 1B). TSA hypersensitivity is defined by the rapid appearance of tetra- and pentaacetylated histones (labeled bands 4 and 5) and complete loss of less acetylated forms (bands 1 and 2). Total conversion from lower to hyperacetylated forms occurs within 15 min, even at very low concentrations (33 nM) of inhibitor (Fig. 1B, H3K4me3, compare lanes 7 and 8). Thus, rapid turnover of acetylation occurs at multiple residues on H3K4me3, excluding the trimethylated lysine 4. Turnover is mediated exclusively by TSA-sensitive HDACs, because nicotinamide, which inhibits sirtuin-like HDACs, does not elicit this effect (Fig. 1B, lanes 7, 11, and 12). Further, hypersensitivity is specific to H3K4me3, as mono- and dimethylated lysine 4 do not show the same rapid response to TSA (11). Recognition of the K4me3 epitope by this antibody is not affected by acetylation at lysine 9 (Fig. S1 A and B), indicating that these observations, particularly the complete loss of lower bands, are not an artifact of antibody specificity or occlusion. Histone H3K79me2 and H3K36me3, which are enriched in the transcribed regions of many genes (reviewed in ref. 21), become acetylated much more slowly than H3K4me3 (Fig. 1B, lanes 7–10). Notably, even after 2-h HDAC inhibition, lower acetylated forms comparable to those seen in untreated cells remain and no maximally hyperacetylated form is detectable (Fig. 1B, compare lanes 7 and 10).

To investigate whether dynamic acetylation of H3K4me3 is evolutionarily conserved, histones from human WI 38 fibroblasts (Fig. 1B, lanes 1–6) and Drosophila S2 cells (Fig. 1B, lanes 13–18) were analyzed. Specificity of rapid hyperacetylation for H3K4me3, compared with other active (H3K79me2 and H3K36me3) and silent (H3K9me2 and H3K27me3) modifications, is absolutely maintained in all three organisms. Apart from H3K9me2, which in Drosophila strikingly appears as a single band resistant to acetylation and unresponsive to TSA after a 2-h treatment, all modifications show virtually identical responses between mouse, human, and fly (Fig. 1B, compare lanes 1–4, 7–10, and 13–16). As in the mouse, nicotinamide has no effect (Fig. 1B, lanes 1, 5, and 6 and 13, 17, and 18). Continuous turnover of acetylation targeted to H3K4me3 is therefore evolutionarily conserved in higher eukaryotes.

H3K4me3 Is Dynamically Acetylated on Lysine 9, 14, and 18.

Sequential ChIP studies have shown H3K4me3 and H3K9ac on the same nucleosomes at the promoters of a number of genes, including c-fos and c-jun (11, 22). To investigate coexistence of modifications on individual histone molecules rather than nucleosomes, we developed a protocol to immunodeplete free histones from mouse fibroblasts using antibodies against H3K4me3. Unbound material was analyzed on SDS (Fig. 2A) or AU (Fig. 2B) gels to monitor any corresponding loss of acetylation and blots quantified by scanning (Fig. 2 A, ii and B, ii). Depletion of H3K4me3 has very little effect on the total amount of H3K9ac left in the unbound fraction (Fig. 2A, compare lanes 1 and 4 with 2 and 5), suggesting that the majority of K9ac occurs on H3 that is not K4 trimethylated. However, reciprocal depletion using anti-H3K9ac antibodies dramatically decreases the amount of H3K4me3 left unbound to approximately 20% of input (Fig. 2A, compare lanes 1 and 4 with 3 and 6). Taken together, these data show that much of the H3K4me3 population is acetylated at lysine 9, but the vast majority of H3K9ac in the mouse nucleus occurs on H3 tails that are not trimethylated at lysine 4. Although genome-wide studies show colocalization of H3K4me3 and H3K9ac (15, 16, 18), these peaks constitute an exceedingly minute fraction of the genome. A relatively low level of H3K9 acetylation at other locations, not detectable as major peaks, can account for the majority of K9ac that does not occur on K4-trimethylated H3 tails.

Fig. 2.

The most highly acetylated histone H3 in TSA-treated cells is trimethylated at lysine 4. (A, i) Purified free histones from untreated (lanes 1–3) or TSA-treated (33 nM, 30 min; lanes 4–6) C3H 10T1/2 cells were immunodepleted with antibodies against H3K4me3 (lanes 2 and 5) or H3K9ac (lanes 3 and 6). Unbound material was resolved by SDS/PAGE and subjected to Western blotting with antibodies against total H3 (Bottom), H3K9ac (Middle), or H3K4me3 (Top). (Lanes 1 and 4: input material before immunodepletion.) (A, ii) Blots from A, i were quantified using ImageJ and normalized to total H3. Data (mean of three biological replicates, plotted ± SEM) are presented relative to input under untreated or TSA-treated conditions (lanes 1 and 4 from A, i). (B, i) Purified free histones from control (lanes 1 and 2) or TSA-treated (33 nM, 30 min; lanes 3 and 4) C3H 10T1/2 cells were immunodepleted with anti-H3K4me3 antibody, resolved on acid–urea gels, and subjected to Western blotting using total H3, anti-H3K4me3, or anti-acetyl H3 antibodies as indicated on the Left of each panel. (Lanes 1 and 3: input material; lanes 2 and 4: immunodepleted fraction.) (B, ii) Total staining for each modification in blots from B, i was quantified using ImageJ, with normalization to total H3 levels. Data (mean of three biological replicates, plotted ± SEM) are presented relative to input under untreated or TSA-treated conditions (lanes 1 and 3 from B, i).

Analysis of H3K4me3-immunodepleted fractions (Fig. 2A) on AU gels showed that depletion of H3K4me3 has no obvious effect on H3K9ac in untreated cells (Fig. 2B, i, compare lanes 1 and 2, and Fig. 2B, ii), but a striking effect on the most highly acetylated H3K9ac bands following acute HDAC inhibition (Fig. 2B, i, compare lanes 3 and 4 and bands 4 and 5). These bands (tetra- and pentaacetylated H3) are preferentially removed by H3K4me3 depletion (Fig. 2B, i, lanes 3 and 4; quantified as 79.5% (± 11.1%) reduction in band 5). Thus, the most highly and dynamically acetylated histone H3 at lysine 9 is distinguished by trimethylation at lysine 4.

H3K18-acetylated histones do not become noticeably hyperacetylated upon TSA treatment, but the appearance of a faint band above bulk H3K18ac following TSA treatment, which is lost following H3K4me3 depletion (Fig. 2B, compare lanes 1, 3, and 4) suggests that the small proportion of the K18ac population that is marked by K4me3 is targeted for dynamic acetylation. In contrast, depletion of H3K4me3 reduces the H3K14ac population to less than 40% input (Fig. 2B), suggesting that a significant proportion of histone H3 acetylated at lysine 14 is also trimethylated at lysine 4. These data demonstrate that hyperacetylated H3K4me3 molecules are acetylated at K9, K14, and K18, suggesting that these three lysines are dynamically acetylated on K4-trimethylated histone H3.

Analysis of HDACs and KATs Shows That p300/CBP Mediates Acetylation of H3K4me3 in Drosophila and Mouse.

The Drosophila genome encodes five potentially TSA-sensitive HDACs—dHDACs 1 (also known as dRpd3), 3, 4, 6 (also known as dHDAC2), and X (23). We found redundancy among these enzymes in mediating deacetylation of histone H3K4me3 (Fig. S2). dsRNA-mediated knockdown of dHDAC1 produced some increased basal acetylation of H3K4me3 in control cells, but none of the individual HDAC knockdowns affected the TSA-induced hyperacetylation of H3K4me3 (Fig. S2). Even allowing for the incomplete nature of dsRNA-mediated knockdown (Fig. S2C), these studies are in agreement with a previous study (23) showing that deacetylation in Drosophila is mediated redundantly by multiple HDACs.

By contrast, our studies on KATs identified a single enzyme responsible for dynamic acetylation of H3K4me3. We again used Drosophila cells in which KAT enzyme families are smaller; dCBP (dKAT3) is homologous to mammalian CBP (KAT3A) and p300 (KAT3B), and dGCN5 (dKAT2) to GCN5 (KAT2A) and p300/CBP-associated factor (PCAF) (KAT2B) in mammals. Specific knockdown of these two transcripts was verified by qRT-PCR (Fig. S3A).

Knockdown of dGCN5 has little effect on the basal acetylation of H3K4me3 compared with the mock knockdown targeting lacZ (Fig. 3A, compare lanes 1 and 5). Further, TSA-induced hyperacetylation of H3K4me3 is unimpaired in the dGCN5 knockdown (Fig. 3A, compare lanes 2 and 6). In contrast, knockdown of dCBP reveals a key role for this enzyme in dynamic histone H3K4me3 acetylation. Loss of dCBP has little effect on the basal level of H3K4me3 acetylation (Fig. 3A, compare lanes 3 and 5). However, TSA-induced hyperacetylation is severely reduced, especially the appearance of the most acetylated forms (bands 4 and 5), leaving a predominance of triacetylated H3K4me3 in the dCBP knockdown, compared with tetra- and pentaacetylated forms in both dGCN5 and lacZ knockdowns (Fig. 3A, compare lanes 2, 4, and 6). Although the response to TSA is clearly diminished after dCBP knockdown, some H3K4me3 TSA sensitivity is detectable in these cells, which could come from residual dCBP remaining after knockdown (Fig.S3A) or from other enzymes involved in targeting H3K4me3.

Fig. 3.

p300/CBP is required for TSA-induced hyperacetylation of H3K4me3. (A) Drosophila S2 cells were treated with dsRNA targeting dGCN5 (lanes 1 and 2), dCBP (lanes 3 and 4), or lacZ (nontargeting control; lanes 5 and 6) as described. Histones from untreated (lanes 1, 3, and 5) or TSA-treated (33 nM, 30 min; lanes 2, 4, and 6) cells were resolved on acid–urea gels and probed using antibodies against H3K4me3 (Upper) or total H3 (Lower). Blots are representative of three replicate experiments. (B) Structures of p300/CBP inhibitor (C646) and control compound (C37). (C) Quiescent S2 cells were treated for 60 min with p300/CBP inhibitor (C646, 30 μM; lanes 3 and 4) or control compound (C37, 30 μM; lanes 5 and 6) with TSA (33 nM; lanes 2, 4, and 6) added where indicated for the final 30 min. Histones were resolved on acid–urea gels and probed for H3K4me3 (Upper) or total H3 (Lower). (Lanes 1 and 2: control and TSA-treated cells.) (D) Quiescent C3H 10T1/2 cells were treated and analyzed exactly as described above for Drosophila S2 cells in Fig. 3C.

We recently reported the discovery and characterization of C646 (Fig. 3B), a small molecule inhibitor of p300/CBP identified by virtual ligand screening of a library of compounds using the crystal structure of the bisubstrate inhibitor Lys-CoA bound to p300 (20). C646 produced 86% inhibition of p300 in vitro at 10 μM, and less than 10% inhibition of the KATs PCAF, GCN5, Sas, and Moz (20). It was used in parallel with a control compound, C37 (Fig. 3B), chemically similar to C646 but inactive against p300/CBP (20). Consistent with dCBP knockdown, pretreatment of S2 cells with C646 completely abrogates TSA-induced hyperacetylation of H3K4me3 histones (Fig. 3C, compare lanes 3 and 4 with 1 and 2 and 5 and 6), confirming that this enzyme mediates dynamic H3K4me3 acetylation. Note that addition of dCBP inhibitor after treatment with TSA to enhance acetylation produces no loss of hyperacetylated H3K4me3 (Fig. S3B, compare lanes 2 and 3), so loss of the TSA-sensitive effect cannot be explained by C646-induced deacetylation. Furthermore, this effect is conserved in higher eukaryotes, as pretreating mouse fibroblasts with p300/CBP inhibitor completely abolishes TSA-induced hyperacetylation of H3K4me3 (Fig. 3D). The fact that RNA interference and chemical inhibition, two widely different approaches to inhibit p300/CBP, efficiently block dynamic acetylation of H3K4me3 in mouse and fly proves that this enzyme is responsible and the phenotype is conserved.

p300/CBP Is Required for TSA- and Stimulus-Induced Acetylation Across the Promoters of c-fos and c-jun.

We have previously described quantitative distributions of histone modifications, including H3K4me3 and H3K9ac, across c-fos and c-jun in mouse fibroblasts (12). We used quantitative ChIP to map p300/CBP KAT activity, defined by sensitivity of histone acetylation to inhibition by C646, across these genes, with gapdh and β-globin^maj (hbb) used to represent constitutively active and transcriptionally silent genes (Fig. 4). As reported previously (11), treatment with TSA for 30 min strongly enhances H3K9ac at c-fos, c-jun, and gapdh (Fig. 4A, dark blue bars). These increases are greatest at the promoter (c-fos −260, c-jun −966) and 5′ end (c-fos +444, c-jun +1,119) of these genes, identifying continuous KAT and HDAC activity at these nucleosomes. Dynamic acetylation is independent of transcription, as c-fos and c-jun are not expressed under these conditions and pretreatment with the transcriptional inhibitor DRB (Fig. 4A, purple bars) has no effect on TSA-stimulated H3K9ac. TSA-stimulated acetylation is completely lost in cells pretreated with C646 (Fig. 4A, cream-colored bars) showing that this is mediated by p300/CBP. This is not a secondary effect of transcriptional inhibition induced by the compound (Fig. 5) as pretreatment with DRB has no effect on acetylation (Fig. 4A, purple bars). Note that ChIP using anti-p300 antibody shows that C646 treatment does not disrupt its localization at c-fos or c-jun (Fig. S4A).

Fig. 4.

p300/CBP KAT activity is required for stimulus- and TSA-induced H3K9 acetylation at c-fos and c-jun independent of transcription. Control C3H 10T1/2 cells (dark blue bars) or cells pretreated with p300/CBP inhibitor (C646, 30 μM, 30 min; cream-colored bars) or DRB (20 μg/mL, 10 min; purple bars), were incubated for an additional 30 min with TSA (33 nM) or EGF (50 ng/mL). Crosslinked, MNase-released nucleosomes were immunoprecipitated using antibodies against H3K9ac. Bound DNA analyzed by real-time PCR using primers to c-fos, c-jun, gapdh, and hbb (primer positions indicated above each panel) and compared with input. Data are the mean of at least three biological replicates, plotted ± SEM. Schematic representations of these genes are shown, indicating the locations of PCR primers used in Figs. 4 and 5B and Figs. S4 and S7. Boxes are exons; closed boxes are coding regions; open boxes are 5′ and 3′ untranslated regions; circles are polyadenylation sites.

Fig. 5.

p300/CBP acetyltransferase activity is required for immediate-early gene induction. (A) Control C3H 10T1/2 cells (lanes 1–3) or cells pretreated for 30 min with p300/CBP inhibitor (C646, 30 μM; lanes 4–6), control compound (C37, 30 μM; lanes 7–9), or TSA (33 nM; lanes 10–12), or pretreated for 30 min with TSA, then C646 added with EGF (lanes 13–15). After pretreatment, cells were stimulated with EGF (50 ng/mL) for a further 30 (lanes 2, 5, 8, 11, and 14) or 60 min (lanes 3, 6, 9, 12, and 15). Total RNA was probed by Northen blotting for expression of c-fos (Top), c-jun (Middle) and gapdh (Bottom). (Lanes 1, 4, 7, 10, and 13: non-EGF treated controls.) (B) Control C3H 10T1/2 cells (dark blue bars) or cells pretreated with p300/CBP inhibitor (C646, 30 μM, 30 min; cream-colored bars), TSA (33 nM, 30 min; red bars), or TSA then C646 (30 min TSA alone, followed by C646 for a further 30 min; purple bars), were incubated for an additional 15 min with EGF (50 ng/mL). Crosslinked, MNase-released nucleosomes were immunoprecipitated using antibodies against RNA polymerase II. Bound DNA was analyzed by real-time PCR using primers to specific positions along c-fos, c-jun, gapdh, and hbb (positions indicated above each panel; shown schematically in Fig. 4) and compared with their representation in the input DNA before immunoprecipitation. Data are the mean of two biological replicates, plotted ± SEM.

Stimulation with EGF to activate expression of c-fos and c-jun (11) produces an increase in H3K9ac in a distribution similar to that seen with TSA (Fig. 4A, compare dark blue bars). Pretreatment with p300/CBP inhibitor before stimulation with EGF abrogates any increase in H3K9ac (Fig. 4A, cream-colored bars), confirming that p300/CBP KAT activity is also responsible for stimulus-induced acetylation at c-fos and c-jun. Interestingly, whereas EGF-induced H3K9 acetylation at c-fos is unaffected by DRB pretreatment, the increase is lost at c-jun (Fig. 4A, purple bars), suggesting that stimulus-induced acetylation is dependent on transcription at this gene. EGF treatment has no effect on H3K9ac levels at the nonresponsive gapdh or hbb genes, demonstrating that acetylation at c-fos and c-jun is specifically targeted to these genes by EGF-stimulated signaling mechanisms. In agreement with previous work at c-fos and c-jun (11, 12) and genomewide analyses (13, 15–18), high levels of H3K4me3 colocalize with peaks of H3K9ac induction, and where levels of H3K4me3 are lower, there is reduced basal and inducible H3K9ac (Fig. S4B).

p300/CBP Acetyltransferase Activity Is Required for Immediate-Early (IE) Gene Expression.

To ask whether p300/CBP KAT activity is required for gene expression, the effect of C646 on c-fos and c-jun induction was analyzed. Inhibition of p300/CBP produces complete loss of induction of both genes (Fig. 5A, compare lanes 1–3 with 4–6), whereas the control compound C37 has no effect. The inhibitor does not act by interfering with signaling to these genes, as ERKs are phosphorylated efficiently in the presence of C646 (Fig. S5A, compare lanes 3 and 4 with 5 and 6). Further, inhibition of gene induction is virtually instantaneous and observed when the compound is added simultaneously with, or even soon after, stimulation (Fig. S5B, compare lanes 1–3 with 4–16). Specificity of inhibition is confirmed by the fact that the compound does not affect global transcription, as assessed by bromouridine labeling (Fig. S6).

To understand how p300/CBP inhibition ablates c-fos and c-jun induction, we analyzed the distribution of RNA polymerase II (Pol II) at these genes (Fig. 5B, dark blue bars). Highest levels of Pol II are recovered at the transcriptional start site of c-fos (−79) and c-jun (−57), with levels decreasing downstream (c-fos +119, +1,443; c-jun +153, +1,119). Gene induction with EGF increases Pol II levels at all sites to high levels, reflecting an extremely active but short-lived burst of transcription, in contrast to lower levels recovered at gapdh. Treatment with C646 strongly disrupts Pol II localization at c-fos and c-jun irrespective of transcriptional induction (Fig. 5B, cream-colored bars), although it does not affect the presence of p300 at these genes (Fig. S4A). This suggests that p300/CBP KAT activity is required for Pol II occupancy at these genes.

One model for the role of p300/CBP in gene expression is that it acetylates nucleosomes to provide a more “open” template through which transcription can proceed (3). Although TSA treatment produces highly acetylated histones at c-fos and c-jun (Fig. 4A), it interferes with gene expression (11). However, this is not as severe as the complete loss of expression caused by inhibition of p300/CBP (Fig. 5A, compare lanes 1–3 with 4–6 and 10–12). Noting that addition of C646 after TSA treatment does not reduce the high levels of histone acetylation generated, either globally (Fig. S3B, lane 3) or at c-fos, c-jun, and gapdh (Fig. S7), we asked whether preacetylation of histones with TSA could rescue gene induction abrogated by p300/CBP inhibition. In fact, forced hyperacetylation by TSA does not restore the loss of gene induction upon p300/CBP inhibition (Fig. 5A, compare lanes 4–6 with 13–15), nor does it prevent the loss of Pol II at c-fos and c-jun (Fig. 5B, purple bars). Taken together, these findings suggest that dynamic modification is more important than static acetylation for efficient gene induction.

Discussion

We show here that dynamic acetylation of all H3K4me3 mediated by CBP is evolutionarily conserved, observed before the divergence of the single Drosophila enzyme dCBP into the paralogs p300 and CBP in mammals. CBP was discovered as a transcriptional coactivator that binds to CREB (24, 25) and p300 complements this activity (26). Whereas p300 and CBP are often considered functionally redundant, some studies support unique roles (reviewed in ref. 27). The majority of genes bound by one also show high levels of the other (18), suggesting common targeting. They interact with many transcription factors and coactivators, initially suggesting a structural role in promoter complexes (reviewed in ref. 28); p300 and CBP have been localized to c-fos and c-jun by ChIP (18, 29) and through interactions with other proteins (30–32). A purely structural role was challenged following discovery of their acetyltransferase activity (33, 34) with the catalytic domain required for transcription from chromatinized promoter constructs in vitro and in vivo (35, 36). Recent in vitro studies with reconstituted nucleosome arrays show a requirement for p300 KAT activity to allow decompaction of 30 nm chromatin, nucleosome remodelling, and transcription factor binding (37–39).

p300/CBP, Dynamic Histone Acetylation, and Gene Induction.

Consistent with previous mass spectrometry (40) and biochemical (41) work, we show that H3K4me3 tails are dynamically modified up to the pentaacetylated state, including at lysines 9, 14, and 18 (Fig. 2), This suggests that the enzyme responsible, p300/CBP, targets specific H3 tails but no specific lysine. In p300/CBP double-knockout mouse fibroblasts, forskolin-induced acetylation of lysine 5, 8, 12, and 16 of histone H4 at c-fos is inhibited (29), further suggesting no specific targeting of residues. A high level of acetylation is insufficient for efficient gene expression in vivo; treatment of cells with TSA enhances acetylation (Fig. 1) but interferes with c-fos and c-jun induction (11). Further, loss of gene expression and Pol II localization caused by p300/CBP inhibition cannot be relieved by preacetylating nucleosomes before inhibition (Fig. 5). This indicates a more dynamic role for acetylation in gene expression, suggesting that turnover may be the important factor. Analyses of quiescent cells in which c-fos and c-jun are poised but inactive and inhibition of transcription with DRB (Fig. 4A) both indicate that transcription is not required for dynamic acetylation.

Cotargeting of H3K4 Trimethylation and Dynamic Acetylation to the Promoter and 5′ End of Genes.

The finding that dynamic acetylation and H3K4me3 (Figs. 1 and 2) colocalize on the same nucleosomes across the promoter and start site of c-fos and c-jun (11) raises the question of their cotargeting. Even when the KAT–HDAC enzyme balance is drastically forced in favor of acetylation by HDAC inhibitors, strict targeting to H3K4me3 does not break down. Numerous ChIP studies have established the presence of H3K4me3, H3K9ac, p300/CBP, and HDACs at the promoter and 5′ end of many genes (17, 18, 22), suggesting widespread colocalization.

There are two classes of model by which cotargeting of H3K4me3 and rapid dynamic acetylation may occur. The first involves independent targeting to the same loci and H3 tails, as previously shown for serine-10 phosphorylation and lysine-9 acetylation (42). This suggests that the relevant enzymes may be part of a common process, and cotargeting may arise from independent DNA sequence recognition or unique interactions with the machinery of signal transduction and transcriptional regulation; p300 and CBP have been isolated in complexes containing TATA-binding protein (TBP) (43, 44) and RNA polymerase II (45–47).

A second class of model is based on dependence of one modification on the other. For example, p300/CBP may mediate dynamic acetylation through direct or indirect recognition of trimethylated lysine 4, which may provide a binding platform or enhance KAT activity. In support of this mechanism WDR5 knockdown, which depletes H3K4 methylation (48), attenuates the TSA-induced increase in H3K9ac levels at promoters (18). Other KATs are known to be recruited to H3K4me3 to induce histone acetylation; yeast Yng1 and Yng2, which recognize H3K4me3 via their plant homeo domain (PHD) fingers (49, 50), form part of the NuA3 and NuA4 KAT complexes, respectively (51, 52), and mammalian ING4 links HBO1 acetyltransferase activity to H3 lysine-4 trimethylated nucleosomes (53, 54).

A sequential targeting mechanism is conceivable. Unmethylated CpG dinucleotides within CpG islands (reviewed in ref. 55) may primarily recruit CXXC motif-containing proteins, including the H3K4 methyltransferase MLL1 (56) and CGBP/Cfp1, which associates with H3K4 methyltransferases (57, 58). DNA binding by Cfp1 has recently been shown to restrict Setd1A and H3K4me3 to euchromatic nonmethylated CpG regions (59). Similarly, ChIP-seq analysis has shown a tight association between Cfp1 and H3K4me3 at CpG islands, and Cfp1 knockdown depletes H3K4me3 levels at nonmethylated CpGs (60). This provides a plausible mechanism to target H3K4me3 to these regions, which could then recruit p300/CBP for dynamic histone acetylation.

Materials and Methods

Cell Culture, Protein Extraction, and Analysis.

C3H 10T1/2 mouse fibroblasts grown in DMEM (10% FCS, 37 °C, 6% CO₂) were quiesced (0.5% FCS, 18–20 h) before treatment. Drosophila S2 cells were grown in Schneider's medium (10% FCS, 25 °C) and quiesced (0.1% FCS, 24 h) before treatment. Pretreatment refers to addition of inhibitor before stimulation, with inhibitor remaining on cells during stimulation. Cells were treated with TSA [10 ng/mL (33 nM); Sigma], nicotinamide (20 μM; Sigma), DRB (20 μg/mL; Sigma), EGF (50 ng/mL; Promega), and TPA (100 nM; Sigma). Synthesis and specificity of C646 (30 μM) and C37 (30 μM) are described elsewhere (20). Acid extraction of histones, resolution on SDS and acid–urea gels, and Western blotting are as described previously (12, 61). Western blots were quantified using ImageJ 1.44 (National Institutes of Health (http://rsb.info.nih.gov/ij/). Antibodies used are described in SI Materials and Methods.

Histone Immunoprecipitation.

Histones were isolated by acid extraction (61) and resuspended in 8 M urea, 5% acetic acid, then dialyzed into 10 mM Tris-HCl pH 8.0, and adjusted to RIPA buffer (61). Histones were incubated with anti-H3K4me3 or -H3K9ac serum with rotation for 2 h at 4 °C, then protein A-sepharose beads (Sigma) preblocked with BSA (0.02% final concentration after addition of histones) added for a further 2 h. The unbound fraction was dialyzed into 10 mM Tris-HCl pH 8.0. Protein was precipitated (25% TCA, 4 °C, 30 min), washed three times with ice-cold acetone, dried, and used for gel electrophoresis.

RNA Interference.

Fragments (400–600 bp) of target genes were amplified from cDNA prepared from Drosophila S2 RNA, using primers incorporating T7 RNA polymerase promoters. Primer sequences obtained from the dkfz E-RNAi Web service (http://e-rnai.dkfz.de) were used as template for dsRNA synthesis using the MEGAscript T7 kit (Ambion) and dsRNA was purified using NucAway Spin Columns (Ambion). S2 cells were plated at 2 × 10⁶ cells per 10 cm² well in 1 mL Drosophila Serum-Free Media (Invitrogen) and 15 μg dsRNA was added to each well. After 30 min, 3 mL Schneider's medium with 10% FCS was added and cells were left for 48 h before treatment and harvesting.

RNA and Chromatin Immunoprecipitation Analyses.

Total cellular RNA was analyzed by Northern blotting (11) and quantitative reverse transcription PCR (12); for protocols see SI Materials and Methods. MNase-digested crosslinked mononucleosomes were used as input for chromatin immunoprecipitation as described previously (12); for protocol see SI Materials and Methods.