Histone H4 R3 methylation at HNF4 target promoters correlates with HNF4 occupancy during intestinal cell differentiation
To identify potential coactivators involved in HNF4-dependent activation of its target genes in their natural cellular milieu, we have utilized a model system based on the ability of CaCo-2 cells to differentiate into enterocytes and to express HNF4 upon reaching confluence (Soutoglou and Talianidis, 2002
). Accordingly, levels of HNF4 dramatically increased after CaCo-2 cells reached confluence () and apoAI and another previously characterized HNF4 target gene α1-antitrypsin (α1-AT) (Soutoglou and Talianidis, 2002
) were induced (), while levels of β-actin remained constant.
Histone H4 R3 methylation at HNF4 target promoters correlates with HNF4 occupancy during differentiation of CaCo-2 cells
In ChIP analyses of cells sampled at various times after reaching confluence (), HNF4 could be detected at both the apoAI and α1-AT promoters by day 2. Its occupancy reached maximal levels by day 6, in agreement with the corresponding protein levels (). Importantly, HNF4 occupancy correlated well with an increase in R3 methylation of histone H4 (H4) of promoter associated nucleosomes at both the apoAI and α1-AT genes. Since PRMT1 is the only known protein arginine methyltransferase responsible for this modification (Strahl et al., 2001
; Wang et al., 2001
) our results establish that this cofactor is implicated in transcriptional regulation by HNF4. No significant change in PRMT1 levels was detected ().
To assess if methylation events are critical for gene expression during CaCo-2 cell differentiation, cells were treated with the methyltransferase inhibitors AdOx and MTA for 24 hr prior to harvesting at day 4 (). While a 4–5 fold stimulation of HNF4 target genes was detected in the untreated cells (lane 1 vs. lane 2) the inhibitor treatment abolished the induction of these genes, suggesting that methylation events are critical for the differentiation of CaCo-2 cells. The treatment showed preferential effects on HNF4 target genes as there was no effect on β-actin or GAPDH levels.
HNF4 interacts directly with PRMT1 through its DNA binding domain
The preceding suggests that PRMT1 is recruited to responsive promoters through HNF4 and implies potential interaction between PRMT1 and HNF4. To look for evidence of such an interaction in vivo, we cotransfected 293T cells with expression plasmids for PRMT1 and HNF4, and immunoprecipitated HNF4 from the derived extracts. Western blots indicated that PRMT1 was co-precipitated with HNF4 (, lane 5). We also performed GST pull-down experiments using purified proteins () and found that PRMT1 showed a strong specific interaction with GST fused to full-length HNF4 (residues 1-455; lane 3 vs. lane 2). To identify the HNF4 domain that interacts with PRMT1 we tested various GST-HNF4 fusion derivatives. Progressive deletions of the HNF4 C-terminus confirmed that the interaction was mediated through the N-terminus of HNF4 (residues 1-116, lane 7). More precise mapping revealed that the interaction relies on the DBD of HNF4 (residues 50-103, lane 16). Furthermore, a GST fusion (residues 1-82, lane 14) lacking the second zinc finger of the DBD was not able to interact with PRMT1, indicating that the second zinc finger plays a crucial role in this interaction.
HNF4 interacts directly with PRMT1 through its DNA binding domain and methylates HNF4 at R91
HNF4 is methylated by PRMT1 at R91
Since PRMT1 displays a broad spectrum with respect to its methylation substrates (Bedford and Richard, 2005
), and given the direct interaction of PRMT1 with the DBD of HNF4, we asked if the latter might be a substrate. First, we addressed whether HNF4 is methylated in vivo. We overexpressed epitope-tagged HNF4 and PRMT1 in 293T cells and determined if they could be metabolically labelled following treatment with S-adenosyl-L-[methyl-3
H]-methionine (SAM). In these experiments, PRMT1 served both as a negative control (since it had no HNF4) and as a positive control for the efficacy of the in vivo methylation reaction (since it is known to be automethylated). Fluorography of the affinity-purified proteins () revealed that both HNF4 (lane 4) and PRMT1 (lane 3) were methylated in vivo. In addition to PRMT1, other unidentified proteins that copurify specifically with PRMT1 were also methylated (, lane 3). Furthermore, treatment with MTA and AdOx abolished HNF4 methylation in vivo (). Despite multiple attempts, we were unable to confirm these results using mass spectroscopic methods, perhaps because of the abundance of arginine and lysine residues in HNF4 DBD and the resulting large number of small tryptic peptides that could not be easily scored.
We confirmed that full-length HNF4 (residues 1-455) is efficiently methylated by PRMT1 in vitro (, lane 1; compare top and bottom panels). PRMT1 methylated all constructs containing the DBD (including residues 1-116, 1-174, and 50-116; lanes 3, 4 and 5) but failed to methylate a construct containing the LBD (residues 128-380, lane 6) or the first N-terminal 24 residues (lane 2). Finer mapping revealed that residues in the second zinc finger (see ) are needed for methylation, as construct 1-82, which lacked this motif, was not methylated (lane 9). Next we mutated the three arginines located within the second zinc finger. Mutants R100Q (lane 12) or R88Q (lane 14) were efficiently methylated by PRMT1. By contrast, mutant R91Q (lane 13) displayed reduced (by circa 80%) HNF4 methylation efficiency. Similarly, mutant R91W (lane 15) completely abolished methylation within the HNF4 DBD. Although the residual methylation that remains in the R91Q mutant suggests the possible existence of secondary methylation sites, these results clearly establish that the major PRMT1 methylation site within the HNF4 DBD is R91.
HNF4 methylation increases its binding to DNA
HNF4 methylation increases its ability to bind DNA
To assess the functional consequences of HNF4 methylation we compared the ability of wild type HNF4 and a methylation mutant (R91W) to transactivate a reporter plasmid containing the apoAI enhancer. As shown in , mutant HNF4 displays a 60% decrease in the transactivation potential relative to the wild-type. To rule out effects on HNF4 subcellular localization we performed immunolocalization experiments in 293T transfected with wild type or mutant (R91W) HNF4 or in HepG2 cells treated with MTA and AdOx (Fig. S1). Neither the mutant nor the treatment changed the localization of HNF4, which remained exclusively nuclear.
To verify if the reduced transactivation by the R91 mutant reflects compromised binding activity we quantified the promoter occupancy by ChIP. shows that the binding of the R91W mutant to the responsive element of the reporter plasmid is reduced by 25%. However, a quantitative dot blot analysis of the HNF4 in the immunoprecipitated material indicated that the transfected mutant protein accumulates to a circa 4-fold higher level than wild-type (, inset). Thus, once corrected for this difference, our data reveal that methylation at R91 significantly increases HNF4 binding to cognate sites in vivo.
To further confirm the previous result we purified the overexpressed wild-type and mutant (R91W) HNF4 proteins and tested their ability to bind their cognate element in a gel shift experiment (). The R91W mutant (lanes 7–10) showed a significant decrease in DNA binding capacity compared to the wild-type (lanes 3–6). To verify that the decrease in binding in the mutant protein is due to absence of eukaryote-specific post-translational modification(s) and not due to gross structural changes, we also checked the DNA binding capacity of corresponding proteins that had been expressed in bacteria in control experiments. These non-modified recombinant proteins bind DNA with similar affinities (compare lanes 12 to 15 and lanes 16 to 19), indicating that the decreased binding observed in the case of the mutant HNF4 expressed in 293T cells is caused by the lack of a postranslational modification.
Next we tested if in vitro methylation of HNF4 by PRMT1 affects its DNA binding. A biotinylated template containing four copies of the apoAI site A was used either directly (, lanes 1 to 7) or after chromatinization (lanes 8 to 13) to test HNF4 binding affinity in the presence or absence of PRMT1 and SAM. Micrococcal nuclease digestion () indicated that our template was able to accommodate two nucleosomes. Under all the conditions tested, HNF4 failed to recruit PRMT1 to the template (). Accordingly, we could not detect any specific complex between DNA, HNF4 and PRMT1 in gel shift experiments (data not shown). Presence of PRMT1 seemed to impair HNF4 binding to the immobilized probe, probably by steric blockage of the DBD (lane 4 vs. lane 3; lane 10 vs. lane 9). Nonetheless, at least on the chromatinized template, efficient stabilization of HNF4 binding (against interference from PRMT1) was seen in the presence of SAM (lane 13 vs. lane 10). This differential effect may be a reflection of the weaker affinity of HNF4 for chromatin compared to naked DNA. Therefore, methylation of the HNF4 DBD leads to stabilization of binding to chromatin targets.
In vitro HNF4 methylation increases its DNA binding activity on chromatin
Increased binding of methylated HNF4 in the experiment of could be explained through two different mechanisms: (a) methylation of HNF4 could reduce the affinity of PRMT1 for the HNF4 DBD, thus making more HNF4 available for DNA binding; or (b) methylation could enhance HNF4 DNA binding directly. To test the first hypothesis, the binding of GST-HNF4 to PRMT1 (with pre-incubation of the proteins in the absence or presence of SAM) was tested under increasingly stringent conditions. shows that methylation of HNF4 results in an increase of its affinity for PRMT1. Thus, the increased binding of HNF4 to chromatin following methylation cannot be explained by decreased interaction between HNF4 and PRMT1. The fact that methylation changes the affinity of PRMT1 for its own substrate made us wonder if methylation of histone tails changes the affinity for PRMT1 (). In contrast to the results obtained for HNF4, methylation of the histone H4 tail was found to result in a reduced affinity for PRMT1. This indicates that methylation can either decrease or increase the affinity for PRMT1 depending on the substrate.
Our results suggest that HNF4 DBD methylation occurs in the absence of DNA binding and that the methylation reaction and DNA binding may be mutually exclusive. We prebound HNF4 to an immobilized DNA template containing four copies of the apoAI site A and performed an in vitro methyltransferase assay using an equivalent amount of free HNF4 as control. shows that methylation of template-bound HNF4 was dramatically reduced compared to that of free HNF4 proving that DNA binding prevents methylation of the DBD. Thus, methylation of HNF4 can only occur prior to DNA binding.
PRMT1 coactivates HNF4 and displays synergistic effects with p300 and SRC-1 through its LBD
While our binding data indicate that PRMT1 cannot be directly recruited by HNF4 to a DNA template, ChIP assays indicate that recruitment of this factor is correlated with HNF4 occupancy of responsive promoters and suggest that PRMT1 plays a chromatin coactivator role in HNF4 mediated transcription. Therefore, we designed an in vitro assay system to dissect coactivator function.
Initial pilot experiments allowed us to narrow down the LBD as the major HNF4 transactivation domain. First, transient transfection in CaCo-2 cells (Fig. S2A) using GAL4 fusion derivatives of HNF4 indicated that a fusion containing residues 128-380 (LBD) was 100-fold more active compared to GAL4 DBD alone. By contrast, residues 1-128 displayed no significant transactivation activity. Similarly, in in vitro transactivation assays from a naked DNA template containing five GAL4 sites (Fig. S2B), GAL4-HNF4-LBD was able to stimulate transcription to a level that was comparable to the fold-activation elicited by GAL4-VP16.
We focused then on the possible coactivation of this domain by PRMT1 in an in vitro transcription system based on chromatinized templates reconstituted with recombinant histones (). The fact that several reports (Chen et al., 2000
; Koh et al., 2001
; Lee et al., 2002
) suggest synergy between several nuclear receptor coactivators and histone arginine methyltransferases led us to test the ability of PRMT1 to function as an HNF4 coactivator alone or in combination with purified p300 and SRC-1 ().
In vitro transcription system for chromatin templates
To validate our chromatin transcription system we first compared the ability of p300, a well-established coactivator both for HNF4 (Malik et al., 2002
) and VP16 (Kundu et al., 2000
), to stimulate GAL4-VP16- or GAL4-HNF4 LBD-dependent transcription in the presence of acetyl-CoA (). In contrast with what is observed with naked DNA, the GAL4-HNF4-LBD alone barely stimulated transcription (lane 4 vs. lane 1). However, further addition of p300 increased transcription by 8-fold (lane 5 vs. lane 4). Similarly, p300 was able to stimulate GAL4-VP16 transcription by about 6.7-fold. The p300 coactivator activity was dependent on the presence of an activator as there was no effect in the absence of GAL4-VP16 or GAL4-HNF4-LBD (lane 6).
In further analysis, we found that similar to p300 (, lane 5 vs. lane 2), SRC-1 moderately stimulated HNF4-dependent transcription when added alone (lane 3 vs. lane 2). By contrast, PRMT1 had only a weak effect on transcription on its own (lane 4). The simultaneous presence of p300 and SRC-1 led only to a level of stimulation that reflected the sum of the independent activities of these coactivators (lane 7 vs. lane 3 and lane 5). PRMT1 did not show cooperative effects with either SRC-1 (lane 6) or p300 (lane 8). However, a strong synergistic effect that was at least 30-fold over that seen with PRMT1 alone (lane 4) was evident when the three coactivators were added together (lane 9). Note that the total coactivator-dependent induction exceeds 30-fold although this could not be precisely quantitated given the essentially undetectable baseline signal (lane 2). Again, this effect was strictly dependent on the presence of GAL4-HNF4-LBD, as the three coactivators failed to stimulate transcription in the absence of the activator (lane 10).
PRMT1 displays synergistic effects with SRC-1 and p300 on HNF4-dependent in vitro transcription of chromatin templates
In control experiments with naked DNA template, in contrast to the results with the chromatin template, addition of p300 and SRC-1 inhibited HNF4-mediated transcription by 40% (, lane 1 vs. lane 2); further addition of PRMT1 reduced transcription to almost undetectable levels (lane 3) suggesting that the coactivation capacity of these factors depends on the presence of chromatin templates.
To demonstrate that PRMT1 coactivates HNF4 through methylation of histone H4 R3 we also reconstituted chromatin templates that contained recombinant H4 in which R3 had been mutated to a glutamine (H4 R3Q) in place of the wild-type () (An et al., 2004
). We compared this template relative to the wild-type for its ability to be activated in a PRMT1-dependent manner in an in vitro transcription assay in which we reduced the amounts of p300 and SRC-1 to better visualize the PRMT1 effect. As a result, HNF4-dependent transcription was only slightly stimulated by p300 and SRC-1 ( lane 1 vs. lane 2). Whereas the specific stimulation by PRMT1 was in the order of 5-fold when wild type H4 was present in the chromatin (lane 3 vs. lane 2), PRMT1 was not able to stimulate transcription from chromatin containing H4 R3Q (lane 7 vs. lane 8). We also tested the effect of omitting SAM from the transcription reactions, which reduced PRMT1 stimulation of transcription by 50% (lane 4). This confirms that the methyltransferase activity of PRMT1 is required for its role as a coactivator. (The residual activity in the absence of ectopic SAM is likely to arise from SAM in the nuclear extract.) These data convincingly prove that H4 R3 is a major target site through which PRMT1 stimulates HNF4 dependent transcription.
Basis for functional synergy between HNF4 coactivators
To assess the contribution of histone modifications in the above effects we performed HAT assays using our chromatinized template and the purified coactivators (). Consistent with previous reports (Xu and Li, 2003
), SRC-1 acetylation activity was virtually nonexistent (lanes 1 and 2). Significant HAT activity of p300 was observed and, importantly, found to be dependent on GAL4-HNF4-LBD implying that this cofactor is targeted to chromatin in an activator dependent fashion (lane 3 vs. lane 4) and that the LBD of HNF4 suffices for its recruitment to chromatin. Negligible effects on p300 HAT activity were evident upon addition of SRC-1 or PRMT1 (lanes 5 and 6). However, similar experiments to test for histone methylation with pure components did not yield a detectable signal (data not shown). That H4 R3 methylation does in fact take place during the transcription process and is dependent both on the presence of the specified coactivators (, lane 1 vs. lane 2) and GAL4-HNF4-LBD (lane 2 vs. lane 3) was confirmed by in vitro ChIP using antibodies against dimethylated H4 R3 after in vitro transcription.
Since the HNF4 LBD failed to interact directly with PRMT1 () there remains the question of how PRMT1 is recruited to the promoter by HNF4. It is generally believed that the p160 family of coactivators act as docking sites for recruitment of other coactivators to DNA-bound nuclear receptors (Xu and Li, 2003
). To evaluate the possibility that HNF4 LBD, SRC-1 and PRMT1 form a ternary complex that may facilitate the recruitment of PRMT1 to HNF4 responsive elements we conducted a GST pull down experiment. First, we confirmed that SRC-1 independently interacts with the LBD of HNF4 and with PRMT1 (). Next we performed GST pull downs to see if there is facilitated recruitment of PRMT1 to the HNF4 LBD by SRC-1. Under the conditions of this experiment, PRMT1 on its own showed a very weak interaction with HNF4 LBD (, lane 4). However, in the presence of SRC-1 (lane 5), significantly greater amounts of PRMT1 were retained indicating that HNF4, SRC-1 and PRMT1 form a ternary complex that may facilitate the recruitment of PRMT1 to the HNF4 LBD.
HNF4, PRMT1 and SRC-1 form a ternary complex
Together, our data conclusively demonstrate that PRMT1 is an HNF4 coactivator that functions at the level of the chromatin template through methylation of histone H4 R3 and that its recruitment can be facilitated through the formation of a ternary complex between HNF4, SRC-1 and PRMT1.