Chd5 Requires PHD-Mediated Histone 3 Binding for Tumor Suppression
Total Page:16
File Type:pdf, Size:1020Kb
Cell Reports Article Chd5 Requires PHD-Mediated Histone 3 Binding for Tumor Suppression Shilpi Paul,1 Alex Kuo,4 Thomas Schalch,1,2,3,6 Hannes Vogel,5 Leemor Joshua-Tor,1,2,3 W. Richard McCombie,1 Or Gozani,4 Molly Hammell,1 and Alea A. Mills1,* 1Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 11724, USA 2W.M. Keck Structural Biology Laboratory, Cold Spring Harbor, NY 11724, USA 3Howard Hughes Medical Institute 4Department of Biological Sciences 5Department of Pathology Stanford University, Stanford, CA 94305, USA 6Present address: Department of Molecular Biology, University of Geneva, Geneva, Switzerland *Correspondence: [email protected] http://dx.doi.org/10.1016/j.celrep.2012.12.009 SUMMARY et al., 2011; Berger et al., 2011; Gorringe et al., 2008; Jones et al., 2010; Lang et al., 2011; Li et al., 2011; Okawa et al., 2008; Rob- Chromodomain Helicase DNA binding protein 5 bins et al., 2011; Sjo¨ blom et al., 2006; Cancer Genome Atlas (CHD5) is a tumor suppressor mapping to 1p36, Research Network, 2008, 2011) in a variety of human cancers. a genomic region that is frequently deleted in human CHD5 expression is also a favorable predictor of survival cancer. Although CHD5 belongs to the CHD family of following anticancer therapy (Du et al., 2012; Garcia et al., chromatin-remodeling proteins, whether its tumor- 2010; Koyama et al., 2012; Wong et al., 2011). suppressive role involves an interaction with chro- CHD5 is a member of the CHD protein family, a group of nine proteins (CHD1–CHD9) that are defined by dual chromodomains matin is unknown. Here we report that Chd5 binds as well as SWI/SNF-like ATP-dependent helicase motifs (Mar- the unmodified N terminus of H3 through its tandem fella and Imbalzano, 2007; Sims and Wade, 2011). For CHD1, plant homeodomains (PHDs). Genome-wide chro- these motifs have been implicated in nucleosome mobilization matin immunoprecipitation studies reveal preferen- (Lusser et al., 2005). CHD proteins have been shown to mediate tial binding of Chd5 to loci lacking the active mark transcriptional activation, repression, and elongation (Murawska H3K4me3 and also identify Chd5 targets implicated and Brehm, 2011). Although some chromodomains bind methyl- in cancer. Chd5 mutations that abrogate H3 binding ated histones (Flanagan et al., 2005; Jacobs and Khorasaniza- are unable to inhibit proliferation or transcriptionally deh, 2002; Nielsen et al., 2002), the chromodomains of CHD4 modulate target genes, which leads to tumorigenesis do not bind histone tails, but instead bind DNA directly (Boua- in vivo. Unlike wild-type Chd5, Chd5-PHD mutants zoune et al., 2002). CHD5, like its closest members CHD3 and are unable to induce differentiation or efficiently sup- CHD4, has tandem plant homeodomains (PHDs; Figure 1A). Growing evidence implicates PHDs as readers of specifically press the growth of human neuroblastoma in vivo. modified or unmodified histones (Godfried et al., 2002; Koh Our work defines Chd5 as an N-terminally unmodi- et al., 2008; Lan et al., 2007; Mansfield et al., 2011; Matthews fied H3-binding protein and provides functional et al., 2007; Musselman et al., 2009, 2012; Ooi et al., 2007; evidence that this interaction orchestrates chro- Org et al., 2008; Rajakumara et al., 2011; Saksouk et al., 2009; matin-mediated transcriptional programs critical for Shi et al., 2006, 2007; Tsai et al., 2010; Wang et al., 2009a; Wy- tumor suppression. socka et al., 2006). The PHD-mediated histone interaction appears to be functionally important because its perturbation INTRODUCTION is associated with various human diseases, including immuno- logical disorders, neurological syndromes, and cancer (Baker Rearrangements of the short arm of human chromosome 1 occur et al., 2008). frequently in a variety of human cancers, with 1p36 deletion Given CHD5’s pivotal role in human cancer, we sought to being a prevalent lesion (Bagchi and Mills, 2008). We previously elucidate its mechanism of tumor suppression by determining identified Chromodomain Helicase DNA binding protein 5 the ability of its PHDs to bind specific histone marks. Here we (CHD5) as a tumor suppressor mapping to 1p36, and discovered demonstrate that the dual PHDs of Chd5 mediate binding that it is frequently deleted in human glioma (Bagchi et al., 2007). specifically to the N terminus of H3 lacking posttranslational Recent studies have indicated that in addition to being modifications, and define this interaction as being essential for commonly deleted, CHD5 is epigenetically silenced (Koyama Chd5 to inhibit cellular proliferation, modulate gene expression, et al., 2012; Mokarram et al., 2009; Mulero-Navarro and Esteller, induce differentiation, and effectively suppress tumorigenesis 2008; Wang et al., 2009b; Zhao et al., 2011) or mutated (Agrawal in vivo. 92 Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors Figure 1. PHDs Mediate Binding of Chd5 with the N Terminus of Unmodified H3 (A) Schematic diagram of mouse Chd5: HMG-box (161–201), PHD1 (345–403), PHD2 (408–465), Chromodomains (468–646), DNA/RNA helicase C (700–1,140), Homeodomain-like (1,425–1,481), CHD-C2 (1,728–1,901). Numbers in parentheses depict amino acid number. (B) Histone peptide array probed with GST-Chd5- PHD1+2; red spots (left) and corresponding red text (right) indicate GST-specific signals. a, asymmetric; ac, acetylation; H, histone; me, methylation; ph, phosphorylation; s, symmetric. (C) Peptide pull-down assays with biotinylated histone peptides and recombinant tandem PHDs of Chd5. Chd5 PHDs bind the unmodified H3 (1- 21) peptide; trimethylation of H3K4 (but not H3K9) disrupts binding. (D) Peptide pull-down assays with biotinylated H3 peptides and nuclear extracts prepared from WT adult mouse brain, followed by immunoblotting with an anti-Chd5 antibody, show preferential binding with H3K4me0. (E) Coimmunofluorescence indicates that Chd5 (red) and H3K4me3 (green) do not overlap (Pear- son coefficient of correlation ranges from À0.01 to 0.12). Cherry-tagged Chd5 expressed in MEFs (upper panels) and endogenous Chd5 immuno- stained with anti-Chd5 antibody (lower panels). Scale bar: 15 mm. (F) The majority of Chd5 peaks identified by ChIP- seq lie within 2 kb of a TSS. (G) The majority (67.2%) of nucleotides that map to H3K4me3 and Chd5 peaks do not overlap. (H) Only 38% (represented by the gray box) of Chd5 and H3K4me3 overlapping peaks are within 100 bp of each other. Negative or positive distance indicates that the H3K4me3 peak is upstream or downstream, respectively, of the Chd5 peak relative to the transcriptional orientation of the gene. The distribution is skewed toward Chd5- bound peaks being upstream of the H3K4me3- bound peaks (positive distance), i.e., away from the gene body. See also Figure S1, Table S1, and Table S2. affinity (Figures S1B and S1C). Endoge- nous Chd5 also bound preferentially to H3K4me0 and did not bind to H3K4me3, a mark that is characteristic of transcrip- RESULTS tionally active genes (Figures 1D and S1D). In agreement with a previous report that PHDs of both CHD5 (Oliver et al., 2012) Chd5-PHDs Bind the N Terminus of Unmodified H3 and its close family member CHD4 bind N-terminally unmodified We screened histone peptide arrays with tagged purified poly- H3 (Musselman et al., 2012), our findings indicate that the PHDs peptides encompassing the tandem PHDs of Chd5 and identi- of Chd5 bind to unmodified H3, and that this interaction is dis- fied specific binding with N-terminally unmodified H3 peptides rupted by posttranslational modifications of extreme N-terminal (unmodified residues 2, 3, and 4 of H3; Figures 1B and S1A). residues of H3. We confirmed the specificity of PHD interactions with unmodi- fied H3, H3K4me1, H3K4me2, and H3K4me3 using in vitro Chd5 Preferentially Binds Genomic Loci Lacking the peptide pull-down assays (Figure 1C). Similar assays using indi- Active Mark H3K4me3 vidual PHDs indicated that although PHD1 and PHD2 had the Given that our in vitro data indicate that Chd5-PHDs do not bind same binding preference for H3K4me0, PHD2 had the highest H3K4me3, we asked whether the subnuclear pattern of Chd5 is Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors 93 reciprocal to that of H3K4me3. Using immunofluorescence in critical for facilitating the H3 interaction. These analyses suggest primary mouse embryonic fibroblasts (MEFs), we found Chd5 that PHD1-D346A, PHD2-E414A, and PHD2-E419A bind H3 expressed in two different patterns: in small dots throughout more efficiently than PHD1-D361A, PHD2-D434A, and PHD2- the nucleus, or in a punctate pattern overlapping with DAPI-en- D415A. riched regions (Figure 1E). Expression analyses indicated that whereas Chd5 and H3K4me3 did not significantly overlap, the The Chd5-H3 Interaction is Critical for Chd5’s Ability punctate pattern of Chd5 overlapped with H3K9me3, a mark of to Inhibit Cellular Proliferation heterochromatin (Figures S1E–S1G). To determine whether PHD-mediated binding to H3 is function- To determine whether this inverse correlation between Chd5 ally important for Chd5 activity, we compared full-length WT and H3K4me3 was evident at the genomic level, we used chro- Chd5 with Chd5-PHD mutants in terms of their ability to inhibit matin immunoprecipitation followed by DNA sequencing (ChIP- cellular proliferation. WT Chd5 proved to be such a potent inhib- seq) to identify Chd5-bound and H3K4me3-bound loci, and itor of proliferation that it was necessary to express it in a regu- compared their correlations across the genome. Annotation of lated fashion; therefore, we developed a tetracycline-inducible peaks to Refseq genes indicated that the majority (61.7%) of system (Zuber et al., 2011) by expressing WT and mutant Chd5-bound peaks were within 2 kilobase (kb) of the transcrip- Chd5 in primary Rosa reverse tetracycline transactivator (rtTA)- tion start site (TSS) of known genes (Figures 1F and S1H). expressing MEFs (Figures S3A–S3C). Whereas doxycycline Because H3K4me3 preferentially marks TSS (Barski et al., (dox) treatment of vector-expressing MEFs had no significant 2007), we restricted our analysis to high-confidence peaks within effect (Figure S3D), dox-induced expression of WT Chd5 in- this interval (Table S1; accession number SRA062358).