CBP physically associates with an HMT
In order to test whether CBP physically interacts with other histone-modifying activities, we immunoprecipitated CBP from HeLa cell nuclear extracts with a polyclonal CBP-specific antibody. As expected, immunoprecipitation of CBP led to the co-immunoprecipitation of HAT activity (Figure 1A, left panel, CBP), which is specific since no HAT activity was co-immunoprecipitated with an irrelevant antibody directed against the p107 protein. We then tested immunoprecipitates for HMT activity by incubating them with purified histones and 3H-labelled S-adenosyl methionine, followed by a standard filter-binding assay. Strikingly, a high level of HMT activity was also found in CBP, but not in p107 immunoprecipitates (Figure 1A, right panel).
Figure 1. CBP is associated with HMT activity, which methylates free histones and polynucleosomal substrates. (A) HeLa cell nuclear extracts (25 μl) were immunoprecipitated with the indicated antibody (anti-CBP: A-22, Santa Cruz; irrelevant: anti-p107, C-18, Santa Cruz). Immunoprecipitates were then assayed for HAT activity (left panel) or HMT activity (right panel) using purified histones as substrate. (B) As in (A) except that the CBP antibody (A-22) was pre-incubated with an increasing amount (2 and 5 μg) of its specific competitor peptide (peptide A-22 P, Santa Cruz) or of an irrelevant peptide (from P/CAF, Eurogentec) before being used for immunoprecipitation. As a control, immunoprecipitation with the irrelevant anti-p130 antibody (C-20, Santa Cruz) was performed. (C) As in (A) with 17.5 μl of nuclear extracts, except that polynucleosomes were used as substrates instead of histones. The HMT activity measured using purified histones was lower than in (A) due to variation from one batch of HeLa nuclear extracts to the other.
Download figure to PowerPoint
To analyse further the specificity of this interaction, we performed peptide competition experiments (Figure 1B). We found that the peptide against which the antibody was raised (CBP), but not an irrelevant peptide (Irr), could efficiently inhibit the immunoprecipitation of HMT activity by the anti-CBP antibody.
Taken together, these results indicate that in live cells CBP is physically associated with an HMT activity.
Immunoprecipitation of endogenous CBP under stringent conditions (in the presence of 0.1% SDS) immunoprecipitated HAT activity but no HMT activity (data not shown), suggesting that CBP is not an HMT by itself. Indeed, recombinant bacterially produced glutathione S-transferase (GST)–CBP did not harbour any intrinsic HMT activity (data not shown). However, GST–CBP was able to retain specifically HMT activity from HeLa nuclear extracts (data not shown), thus confirming our immunoprecipitation results. We also noticed that the HMT activity recruited by GST–CBP was consistently lower than after immunoprecipitation of endogenous CBP. This probably reflects the fact that the endogenous CBP–HMT complex has to be first disrupted before its reassociation on recombinant CBP, and could suggest that the interaction between CBP and the HMT is not direct.
Another important question to address was to determine whether the CBP-associated HMT activity could methylate nucleosomal histones. To this aim, purified polynucleosomes were prepared (data not shown) and were used as a substrate in an HMT assay. Interestingly, we found that polynucleosomal histones were methylated by the CBP-associated HMT activity (Figure 1C). However, the HMT activity, similarly to HAT activity, was ∼3- to 4-fold lower on nucleosomes than on free histones (data not shown).
Taken together, these results indicate that in live cells, CBP physically associates with an HMT, which can methylate polynucleosomal substrates. Moreover, they suggest the existence of a complex that harbours both HAT and HMT activities. Interestingly, immunoprecipitation of pCAF, which is not associated with CBP/p300 in HeLa cells (Ogryzko et al., 1998), did not result in the co-immunoprecipitation of significant HMT activity (our unpublished result). Thus, the ability to interact with an HMT is specific for CBP/p300 among HATs.
The CBP-associated HMT methylates specifically K4 and K9 of histone H3
In live cells, methylation of histones has been reported to occur on histone H3 or histone H4 N-terminal tails. In order to test the substrate specificity of the CBP-associated HMT activity, we analysed methylated histones by SDS–PAGE followed by Coomassie blue staining (Figure 3A, left panel) or autoradiography (Figure 3A, right panel). Our results showed that only histone H3 was methylated by the CBP-associated HMT activity (Figure 3A). Remarkably, we could not detect any methylation at all on histone H4, which is also known to be methylated in live cells. When a peptide derived from the histone H3 N-terminal tail was used as a substrate instead of histones, we found that it was very efficiently methylated by the CBP-associated enzyme (Figure 3B, H3 peptide). Again, this methylation was highly specific, as no activity could be seen when a peptide derived from the N-terminal tail of histone H4 was used as a substrate (H4 peptide). Note that both substrates could be acetylated by CBP (data not shown), indicating that the lack of methylation of the histone H4 peptide is not due to a general defect of this peptide.
Figure 3. The CBP-associated HMT methylates the histone H3 N-terminal tail. (A) HeLa nuclear extracts immunoprecipitated with the anti-CBP antibody or the anti-p107 (ctrl) were assayed for the presence of HMT activity using 1 μg of purified histones as substrate. Histones were then separated on an 18% SDS–polyacrylamide gel and detected by Coomassie blue staining (left panel) or by autoradiography (right panel). (B) As in (A), except that a peptide derived from the histone H3 or H4 N-terminal tail was used as a substrate. (C) The CBP-associated HMT methylates lysines 4 and 9, which are methylated in vivo. Histone H3 labelled by CBP-associated HMT activity was subjected to microsequencing analysis and parallel scintillation counting. Lysine residues methylated by the CBP-associated HMT and methylated in vivo are in bold. Acetylated lysines are underlined.
Download figure to PowerPoint
To determine the methylation sites, we undertook Edman degradation followed by sequence analysis in parallel with liquid scintillation counting of radiolabelled methylated histone H3 (Figure 3C). We found that the CBP-associated HMT activity methylates preferentially lysine 9 and to a lesser extent lysine 4 from histone H3. Scintillation counting of the membrane after the degradation of the first 27 amino acids indicated that there was hardly any radioactivity left. Again, this result confirms that the main methylation sites are located within the histone H3 N-terminal tail. Taken together, results from Figures 3 and 4 indicate that the CBP-associated HMT methylates K9 and K4 from histone H3. Interestingly, these sites are found methylated in mammalian cells (Strahl et al., 1999).
Figure 4. Acetylation by CBP is not affected by K9 methylation. HAT assays of bacterially produced GST–CBP were performed using either the unmodified histone H3 peptide (unmod.), a peptide methylated on K9 (K9-Met) or a peptide acetylated on K14 (K14-Ac) as substrates. Acetylation efficiency was calculated relative to 100% for the unmodified peptide.
Download figure to PowerPoint
These data also suggest that the CBP-associated enzyme is unlikely to be the known HMT CARM1 (Chen et al., 1999), which methylates arginines (data not shown) or SUV39H1, which has a slightly different substrate specificity (Rea et al., 2000). The possibility still remains, however, that native CARM1 or SUV39H1 could exhibit a different specificity from bacterially produced recombinant proteins, due to post-translational modifications or interactions with other cellular proteins.
K9 methylation does not influence acetylation by CBP
Histone acetylation is generally linked to transcriptional activation. We thus tested whether methylation of H3 tail by the CBP-associated enzyme could interfere with acetylation by CBP (Figure 4). A peptide containing the first 17 amino acids from histone H3 was efficiently acetylated by recombinant bacterially produced GST–CBP (Figure 4, unmod.). As expected, when a peptide acetylated on the major CBP acetylation site (K14-Ac) was used, the acetylation efficiently decreased to 25%. The residual activity probably reflects acetylation of the peptide at K4 or K9, although these lysines are not acetylated by CBP using nucleosomes as substrate (Schiltz et al., 1999). Methylation of K9, the major site methylated by the CBP-associated enzyme (see Figure 3), led to a slight reduction in the efficiency of acetylation (K9-Met), which probably reflects the fact that K9 is not acetylatable in this peptide. This result indicates that methylation on K9 does not significantly affect the efficiency of acetylation by CBP on other lysines. In the converse experiment, we also found that acetylation of histone H3 by CBP did not change the efficiency of methylation by the CBP-associated enzyme (data not shown).
In this paper, we show that the versatile co-activator CBP is physically associated with an HMT (Figure 1). This result raises the question of the possible involvement of histone methylation in transcriptional activation. The functional consequences of histone methylation have been poorly studied so far, due to the absence of anti-methylated histone antibodies and also due to the fact that it is biochemically difficult to separate methylated histones from unmethylated histones. It was recently shown that methylation of lysine 4 of histone H3 occurs only in transcriptionally active macronuclei in Tetrahymena (Strahl et al., 1999). This finding raises the possibility that histone methylation, like acetylation, could be associated with transcriptional activation. The molecular cloning of the enzymes that methylate histones, such as the recently described SUV39H1 (Rea et al., 2000), will undoubtedly be a major breakthrough in the studies on the functional consequences of histone methylation.
The discovery of a putative HMT, named CARM1, which methylates histone H3 in vitro and which can act as a nuclear receptor co-activator (Chen et al., 1999), could provide a potential link between histone methylation and transcriptional activation. CARM1 was found to be homologous to arginine methyl transferases. We found that recombinant bacterially produced CARM1 preferentially methylates arginine 17 of histone H3 in vitro (data not shown). This result confirms that CARM1 is indeed an arginine methyl transferase in vitro. However, methylation of histones on arginines is still a controversial matter and has never been demonstrated to occur in live cells (Gary and Clarke, 1998). These data cast doubt about histones as bona fide substrates for CARM1.
Another interesting question raised by our results deals with the relationship between various histone N-terminal tail modifications.
Strikingly, both histone N-terminal tail acetylation and methylation occur on lysines (Strahl and Allis, 2000). It is thus tempting to speculate that, on a particular lysine, one modification might affect the other. However, the only lysine that can be acetylated or methylated is lysine 9 of histone H3. Pre-acetylation of lysine 9 impairs its methylation by the CBP-associated enzyme (data not shown), as well as by Suv39H1 (Rea et al., 2000), suggesting that both modifications cannot occur at the same time.
Our results (Figure 4) suggest that histone H3 methylation by the CBP-associated enzyme does not significantly change the efficiency of subsequent acetylation by CBP (Rea et al., 2000). However, it was recently shown that methylation at K9 inhibits phosphorylation of histone H3 at S10 and that p300 acetylates more efficiently phosphorylated histone H3 (Cheung et al., 2000; Lo et al., 2000). Thus, histone H3 methylation by the CBP-associated enzyme could indirectly inhibit acetylation by CBP. Interestingly, CBP was shown to physically associate with the putative histone H3 kinases pp90rsk (Nakajima et al., 1996; Sassone-Corsi et al., 1999), raising the possibility that the relative amount of pp90rsk or HMT associated with CBP could be a way to regulate acetylation by CBP.