Trithorax monomethylates histone H3K4 and interacts directly with CBP to promote H3K27 acetylation and antagonize Polycomb silencing

Abstract

Trithorax (TRX) antagonizes epigenetic silencing by Polycomb group (PcG) proteins, stimulates enhancer-dependent transcription, and establishes a 'cellular memory' of active transcription of PcG-regulated genes. The mechanisms underlying these TRX functions remain largely unknown, but are presumed to involve its histone H3K4 methyltransferase activity. We report that the SET domains of TRX and TRX-related (TRR) have robust histone H3K4 monomethyltransferase activity in vitro and that Tyr3701 of TRX and Tyr2404 of TRR prevent them from being trimethyltransferases. The trx(Z11) missense mutation (G3601S), which abolishes H3K4 methyltransferase activity in vitro, reduces the H3K4me1 but not the H3K4me3 level in vivo. trx(Z11) also suppresses the impaired silencing phenotypes of the Pc(3) mutant, suggesting that H3K4me1 is involved in antagonizing Polycomb silencing. Polycomb silencing is also antagonized by TRX-dependent H3K27 acetylation by CREB-binding protein (CBP). We show that perturbation of Polycomb silencing by TRX overexpression requires CBP. We also show that TRX and TRR are each physically associated with CBP in vivo, that TRX binds directly to the CBP KIX domain, and that the chromatin binding patterns of TRX and TRR are highly correlated with CBP and H3K4me1 genome-wide. In vitro acetylation of H3K27 by CBP is enhanced on K4me1-containing H3 substrates, and independently altering the H3K4me1 level in vivo, via the H3K4 demethylase LSD1, produces concordant changes in H3K27ac. These data indicate that the catalytic activities of TRX and CBP are physically coupled and suggest that both activities play roles in antagonizing Polycomb silencing, stimulating enhancer activity and cellular memory.

Keywords: CBP; Drosophila; H3K4 monomethylation; Trithorax.

Fig. 1.

The catalytically inactive trx^Z11 and trr³ mutants exhibit reduced H3K4me1. (A) Quantitative westerns of H3K4me1 and H3K4me3 levels, normalized to total H3, in third instar salivary glands from trx^Z11 homozygotes and wild-type (Oregon-R) controls. Standard errors were calculated from three independent samples. (B-E) Immunostaining of H3K4me1 and H3K4me3 (B) on polytene chromosomes from wild type (wt, Oregon-R), trx^Z11 mutant and TRX overexpressers (o/e), (C,D) in stage 10 trx^Z11 and trr³/Y mutant embryos and wild-type controls (Oregon-R in C, w¹¹¹⁸ in D) and (E) in trr³/Y; trx^Z11/trx^Z11 double-mutant embryos. Arrows point to wild-type (w¹¹¹⁸) and double-mutant embryos. DAPI staining for trx^Z11 (B, bottom right) indicates embryo position. Note that trx^Z11 polytene chromosomes in B appear under-replicated compared with wild type. Images were taken under the same settings as wild-type control. See also supplementary material Fig. S1D for H3K4me3 immunostaining in trr³/Y mutant and wild-type embryos (stage 13).

Fig. 2.

The TRX SET domain is a histone H3K4 monomethyltransferase. (A-C) Western blots (WB) of H3K4me1/2/3 reaction products of HMT assays conducted with purified wild-type and mutant GST-TRX-C (see constructs in D) and recombinant histone H3 and H3K4R mutant substrates. Coomassie Blue-stained GST fusion proteins and H3 substrates are shown below westerns. In lane 7 of C, recombinant H3K4me2 or H3K4me3 (20 ng/lane) was used as positive antibody control (additional controls in supplementary material Fig. S6). Note that use of the H3K4R mutant as substrate or the absence of the methyl donor SAM in HMT assays completely abolished the H3K4me1 signal (lane 3 in B and C), indicating the specificity of the HMT assay. (D) Summary of GST-TRX-C fragments and mutants tested in HMT assays.

Fig. 3.

The SET domains of TRR and human MLL1 and MLL2 are robust H3K4 monomethyltransferases. In vitro HMT assays (as in Fig. 2) of wild-type and mutant GST-TRR-C421 (A), human GST-MLL1-C251 (B) and GST-MLL2-C219 (C). Asterisk indicates the degraded form of GST-TRR-C421 in Coomassie Blue-stained gel (bottom of A). Lane 6 in A contains the GST-TRX-C751 Y3701F switch mutant for comparison with the corresponding TRR-C Y2404F mutant. Recombinant H3K4me1, H3K4me2 or H3K4me3 was used as positive control in B and C. The weak western signal in lane 3 of A (third panel down) is likely to be due to non-specific binding of the anti-H3K4me3 antibody to H3K4R.

Fig. 4.

TRX and TRR physically associate with CBP in vivo. (A) Immunoblots of TRX-N, TRR and E(Z) after IP from nuclear extract (NE) by anti-CBP (lane 3), anti-BRM (lane 4) and anti-TRX-N (lane 5) antibodies. The asterisk indicates serum background. (B) Immunoblots of TRX-C, CBP and p105 after IP by anti-TRX-C (lane 4) and anti-ESC (lane 3) antibodies. Pre-immune serum in lane 2 (A,B) and the p105 subunit of the CAF1 complex (bottom of B) serve as negative controls.

Fig. 5.

The C-terminus of TRX contains the CBP-interacting domain (CID). (A-C) Immunoblots of endogenous TrxG proteins (CBP, BRM, UTX and SNR1) pulled down from NE by GST-TRX-C (or TRR-C421 in lane 5 of A). Coomassie Blue-stained purified GST fusion proteins used in pulldown assays are shown at the bottom of B and C. (D) Summary of GST-TRX fragments tested and results of CBP pulldowns. Numbers within the CID refer to four KIX-binding motifs (see Fig. 6C).

Fig. 6.

TRX binds directly to the CBP KIX domain. (A) GST-CBP fragments were used to pull down purified TRX-CID (with a C-terminal His tag) and were Coomassie Blue stained (bottom). The TRX CID was detected by anti-His antibody (top). (B) Summary of GST-CBP fragments tested and results of TRX CID pulldown. Conserved domains including two zinc fingers (ZF1 and ZF3), KIX (also called CREB-binding domain) and bromodomain (BrD) are shown. The three enzymes TRX, UTX (H3K27 demethylase) and BRM (chromatin remodeler) in TrxG have been found to interact with CBP at KIX, ZF1 and PHD finger, respectively (Tie et al., 2012). (C) Sequence alignment of known KIX-binding motifs in MLL1 and transcription factors (Wang et al., 2012) and four putative KIX-binding motifs identified within the TRX CID. Φ is a hydrophobic residue (in bold) and X is an arbitrary residue. Phosphorylated serines in CREB (S133) and FOXO3a-CR3 (S626) are underlined.

Fig. 7.

H3K4me1 enhances acetylation of H3K27 by CBP. (A,B) In vitro CBP HAT assays were performed using recombinant H3, either unmodified or containing single mono-, di- or trimethyl-K analogs (Simon et al., 2007) at K4, K9 or K36 as substrate. Western blots were used to assay the H3K27ac reaction product (top) and H3 (bottom, loading control). All conditions in HAT assays were identical except for the histone H3 substrate. (C) Immunoblots of various histone H3 modifications in salivary gland extracts from control (w¹¹¹⁸; lane 1), lsd1^ΔN homozygotes (lane 2) and LSD1 overexpresser (expressed under the control of da-GAL4) (lane 3). Quantitative westerns of histone H3 K4me1, K27ac, K27me3 and K4me3 are shown at the bottom with the control value set as 1. Error bars indicate s.d. of normalized means from three independent replicates. (D) S2 cells were treated with LSD1 inhibitor (10, 20, 40 and 80 μM) for 3 days and then analyzed by western blot. Note that lower concentrations (10 μM and 20 μM) of inhibitor had no detectable effect on bulk H3K4me1 and H3K27ac levels (lanes 2 and 3 in B).

Fig. 8.

TRX-C colocalizes with H3K4me1, CBP and TRR at putative PRE/TREs and enhancers. (A) Heat map clustered by peaks of TRX-C, TRR, CBP, PC, H3K4me3, H3K27ac, H3K27me3 and H3K4me1 from published ChIP-chip and ChIP-seq data in Drosophila S2 cells. Each column is centered on the peak midpoint within a 10 kb region (±5 kb from midpoint). (B) Heat map of clusters 1 and 2 (putative PRE/TREs and enhancers) in A sorted by TRX-C peak intensity. Weak H3K4me3 peaks in cluster 2 suggest that some promoters might be within 5 kb of putative enhancers or PREs (see pnr and zfh1 in supplementary material Fig. S9). (C) Model of epigenetic regulation of PcG target gene expression by TRX and CBP. TRX and CBP are physically associated and constitutively present on PRE/TREs. In PcG-repressed genes (top), TRX (and TRR)-dependent H3K4me1 is only present at PRE/TREs and H3K27ac deposition is inhibited by the broad domains of H3K27me3 that spread from the PRE to promoter (P). In active genes (bottom), H3K4me1 spreads from PRE/TREs to form broad domains over neighboring cis-regulatory regions, perhaps assisted by interactions between TRX/CBP and enhancer-bound transcriptional activators. The coupling of TRX and CBP catalytic activities and the stimulatory effect of H3K4me1 on acetylation by CBP promotes the formation of similar broad domains of H3K27ac, which might also be dependent on TRX/CBP interactions with enhancer-bound activators. This prevents PRC2-mediated deposition of H3K27me3 at the PRE and its propagation from the PRE over the same broad region. TRX (and CBP) also acts at promoters (Enderle et al., 2011) of active genes. Binding of the third PHD finger of TRX to promoter-associated H3K4me3 is required for transcription and might have an indirect effect on the H3K4me3 level.