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 0.01 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). The hibited proliferation (Figure 3A), consistent with our previous majority (63.5%) of these gene-proximal Chd5-peaks mapped findings that proliferation is compromised in MEFs derived to genes lacking H3K4me3-peaks (Figure S1I). Out of the large from mice engineered to have an extra copy of the genomic fraction of genes whose TSS overlaps with H3K4me3 peaks, region encompassing Chd5 (Bagchi et al., 2007). there was a depletion of Chd5 peaks that was 1.2-fold less To functionally define residues that are critical for inhibiting than what would be expected randomly. Because of the small proliferation, we assayed a series of Chd5-PHD mutants for their number of Chd5-bound regions, this depletion was just over effect on proliferation using the dox-inducible system. Whereas the threshold for statistical significance (p < 0.08, corrected inducible expression of the H3-binding-competent Chd5-PHD Fisher; Table S2). When the nucleotide sequence of the peaks mutants D346A or W384A inhibited proliferation to an extent was considered, only 32.8% of the nucleotides overlapped comparable to that observed for WT Chd5, Chd5 mutants that between H3K4me3-enriched region reads and Chd5-peaks (Fig- are compromised for H3 binding (G355A, D361A, D415A, ure 1G). Out of the small percentage of peaks mapping to the C432W, and D434A) were defective in their ability to inhibit same vicinity (36.5%; see Figure S1I), the majority (62%) were proliferation (Figures 3A and S3D). Whereas D346A inhibited spaced >100 nucleotides from each other, and most of the proliferation, this was not the case in the context of a second peaks were oriented such that the Chd5-peak was upstream of mutation at D434, indicating that perturbation of a single PHD- the H3K4me3-peak (Figures 1H and S1I). The read counts and H3 interaction in one PHD overrides the ability of the other ChIP quantitative PCR (qPCR) analysis in the vicinity of several PHD to compensate. These findings indicate that PHD-mediated representative Chd5-bound genes indicate that most Chd5- binding to unmodified H3 is essential for Chd5 to inhibit bound loci lack H3K4me3 (Figures S1J and S1K). These data proliferation. indicate that the majority of Chd5-bound loci lack H3K4me3 in vivo, in agreement with our in vitro findings. Chd5 Transcriptionally Modulates Genes Implicated in Cancer, whereas H3-Binding-Incompetent Chd5-PHD Mutation of Specific Conserved PHD Residues Mutants Fail to Do So Abrogates the Chd5-H3 Interaction Genome-wide ChIP-seq analysis indicated that the majority of To identify Chd5-PHD residues that are important for the interac- Chd5-bound peaks were located in promoters, 50 untranslated tion with unmodified H3, we compared the PHDs of Chd5 with regions (UTRs), and early introns (see Figures 1F and S1H), the PHDs of the closest family members, Chd3 and Chd4 (Fig- supporting the idea that Chd5 plays a role in transcriptional ure 2A). We used site-specific mutagenesis to generate a series modulation. Pathway analysis of the candidate Chd5-modulated of full-length Chd5 constructs with mutations in individual genes identified by ChIP-seq showed enrichment for proteins conserved PHD residues, including some amino acids corre- implicated in cancer (Figure 3B; Tables S3 and S4). To validate sponding to those that were previously characterized for human specific candidate genes as being Chd5 modulated, we CHD4 (Mansfield et al., 2011; Musselman et al., 2009, 2012; Fig- analyzed their expression in primary MEFs in which Chd5 was ure 2B). Chromatin fractions from cells expressing Flag-tagged overexpressed or knocked down (Figures 3C and S3E). Indeed, Chd5 indicated that wild-type (WT) Chd5, as well as versions expression of several Chd5 targets correlated with Chd5 expres- of Chd5 with single amino acid mutations, is associated with sion. Whereas mutants D346A and W384A were able to inhibit chromatin (Figure S2A). However, in vitro assays with unmodi- both the proliferation and expression of several Chd5-bound fied H3 peptides and with either purified Chd5-PHDs (Figures genes, mutant versions of Chd5 that were defective in their ability 2C, 2D, S2B, and S2C) or nuclear lysates expressing Flag- to inhibit proliferation (G355A, D361A, D434A, and C432W) were tagged Chd5 (WT or those with single amino acid mutations; not able to inhibit expression of Chd5-bound loci. These findings Figures 2E and S2D) indicated that several Chd5 residues are indicate that PHD-mediated H3 binding is critical for the ability of

94 Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors Figure 2. Characterization of the PHD- Histone Interaction (A) Multiple sequence alignment of PHDs of mouse Chd5, Chd4, and Chd3. The zinc-binding residues (yellow) and the Chd5-PHD residues that were mutated (green and purple) are depicted. (B) Mutant residues (green and purple) are map- ped on the CHD4-PHD1 structure (2L5U) with a modeled peptide (left) and CHD4-PHD2 struc- ture (2L75) with H3K9me3 peptide (right). (C) Peptide pull-down assays with biotinylated histone peptides and recombinant tandem PHD fingers of Chd5 containing single amino acid mutations D346A, D361A, or D434A. Whereas the Chd5-PHD mutant D346A binds the H3K4me0 peptide, mutants D361A and D434A have compromised binding. (D) FP measurements for GST-cleaved WT and mutant PHDs binding to unmodified H3 (1-21)- fluorescein peptides, represented as plots of average anisotropies of two or more experiments versus protein concentration. D361A and D434A disrupt the interaction with the H3K4me0 peptide, whereas D346A retains H3K4me0 binding. Error bars denote SEM. (E) H3 peptide pull-down assays with nuclear extracts prepared from MEFs expressing Flag- tagged WT Chd5 or the Chd5-PHD mutants, followed by immunoblotting with an anti-Flag antibody, indicate that whereas the WT Chd5 and the Chd5-D346A mutant are pulled down by H3 peptide, the Chd5-D361A and D434A mutant proteins are not. See also Figure S2.

genesis in vivo (Bagchi et al., 2007). Therefore, we assessed whether cells expressing functionally compromised Chd5-PHD mutants were prone to Ras- induced transformation and tumorigen- esis in vivo. rtTA MEFs coexpressing RasG12D and dox-inducible constructs encoding WT or mutant versions of Chd5 to transcriptionally modulate its targets, including genes Chd5 were injected subcutaneously into two cohorts of athymic that have been implicated in Wnt signaling, chromatin remodel- nude mice. One cohort of mice received regular food, whereas ing, and cell-cycle regulation, highlighting Chd5’s role in path- the other cohort received a dox-containing diet to induce ways that were previously implicated in tumor suppression expression of WT Chd5 or the Chd5-PHD mutants, and (Hers et al., 2011; Musgrove et al., 2011; Yadav et al., 2009; tumorigenesis was monitored (Table S5). Whereas expression Yao et al., 2011). of WT Chd5 did not lead to tumor formation, cells expressing mutant Chd5 formed tumors with a severity that correlated The Inability of Chd5-PHD Mutants to Bind H3 Causes inversely with their ability to bind H3, i.e., D361A and D434A Oncogenic Transformation and Robust Tumorigenesis caused robust tumor development, and fewer and significantly In Vivo smaller tumors formed in response to D346A (Figures 3D and The above findings indicate that PHD-mediated H3 binding is S3F). Furthermore, tumors that arose from MEFs expressing essential for Chd5 to inhibit proliferation and to transregulate Chd5-PHD mutants that cannot bind H3 (D361A and D434A) genes encoding components of cancer-associated pathways. expressed Chd5 target genes at higher levels relative to the We next investigated the extent to which perturbation of the smaller tumors that developed from MEFs expressing WT PHD-mediated Chd5-H3 interaction leads to cancer. We previ- Chd5 or D346A (Figure 3E). These findings demonstrate that ously reported that Chd5-compromised MEFs were sensitized perturbation of the Chd5-H3 interaction dramatically enhances to oncogenic transformation and thus predisposed to tumori- tumorigenesis in vivo.

Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors 95 Figure 3. Abrogation of PHD-Mediated Chd5-H3 Binding Causes an Inability to Inhibit Proliferation, Transcriptional Derepression, and Tumorigenesis In Vivo (A) Proliferation of MEFs expressing Chd5 or PHD mutants. Induction of WT Chd5 and the Chd5-PHD1-D346A mutant inhibits cellular proliferation, whereas Chd5-PHD1-D361A, Chd5-PHD2-D434A, and Chd5-PHD1-D346A/PHD2-D434A fail to do so. MEFs expressing dox-inducible WT Chd5 or PHD mutants dox- treated (+dox, blue squares) or untreated (dox, red circles) are quantitated. Data are represented as mean ± SD. Student’s two-tailed t test (*p = 2.4E-07 and **p = 4.02E-06). (B) Chd5-bound loci encode proteins involved in cancer pathways. The Kyoto Encyclopedia of Genes and Genomes (KEGG) and Ingenuity Pathway Analysis were used to classify Chd5-bound genes. (legend continued on next page)

96 Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors The Chd5-H3 Interaction Is Essential for Inhibition 4D). This demonstrates that induction of WT Chd5 suppresses of Proliferation, Induction of Differentiation, tumorigenesis in vivo. and Suppression of Tumor Growth of Human Neuroblastoma Cells DISCUSSION Given that CHD5 inactivation has been implicated in a variety of human cancers, including neuroblastomas (Fujita et al., 2008; Since the time that CHD5 was first reported as a tumor Garcia et al., 2010, 2012), we asked whether induction of WT suppressor mapping to human 1p36 (Bagchi et al., 2007), its or mutant Chd5 in a CHD5-deficient context could enforce tumor inactivation has been documented in a diverse array of human suppression in human cancer cells. We chose the human neuro- cancers (Agrawal et al., 2011; Berger et al., 2011; Gorringe blastoma cell line SK-N-AS, which has a 1p36.33-1p36.2 dele- et al., 2008; Jones et al., 2010; Koyama et al., 2012; Lang tion (Kaghad et al., 1997) encompassing the CHD5 locus. et al., 2011; Li et al., 2011; Mokarram et al., 2009; Mulero-Nav- Endogenous CHD5 is absent or expressed at very low levels in arro and Esteller, 2008; Okawa et al., 2008; Robbins et al., this cell line (Figure S4A; Garcia et al., 2010). To establish SK- 2011; Cancer Genome Atlas Research Network, 2011; Wang N-AS cells in which we could express Chd5 in a regulated et al., 2009b; Zhao et al., 2011), which indicates that CHD5 regu- fashion, we generated SK-N-AS cells that stably express rtTA, lates processes that are fundamental for cancer prevention. and introduced dox-inducible WT or mutant versions of Chd5 Therefore, defining how CHD5 protects against tumorigenesis using retroviral infection (Figures S4B and S4C). Although induc- may affect the treatment of a variety of human cancers. Although tion of both WT Chd5 and D346A inhibited proliferation (Fig- it belongs to a protein family that includes members implicated in ure 4A), we observed only an 17% reduction, compared with chromatin remodeling, the mechanism by which CHD5 exerts its the striking 54% inhibition of proliferation in MEFs (see Fig- tumor-suppressive role has been largely unexplored. Here, we ure 3A). This finding is consistent with the fact that p53, a down- demonstrate that the ability of Chd5 to bind unmodified H3 is stream effector of Chd5-mediated tumor suppression (Bagchi essential for tumor suppression. et al., 2007), is inactivated in SK-N-AS cells (Goldschneider By focusing on the mechanism by which CHD5 interacts with et al., 2006). Interestingly, human neuroblastoma cells express- chromatin, we discovered that the tandem PHDs mediate ing WT Chd5 had differentiated features, such as a flat epithelial- binding to N-terminally unmodified H3. PHDs are modules that like morphology and increased expression of the neuronal were initially identified in plant homeodomain (PHD) proteins; marker MAP2, which contrasted markedly with the phenotype subsequently, PHD-containing proteins were discovered in of cells expressing either control vector or the H3 binding- yeast, flies, and humans, especially in chromatin-associated impaired PHD mutants (Chd5-D361A or Chd5-D434A), which and nuclear proteins. Although the zinc-binding motifs of PHD had the classic stem-like neuroblastoma morphology with prolif- motifs are well conserved, diversity in the ligand-binding resi- erative foci and lower expression of MAP2 (Figures 4B and S4D). dues generates versatility in their interaction partners. PHDs Cells expressing the Chd5-D346A mutant had a partially differ- are histone readers that control gene expression cascades by entiated morphology. This analysis indicated that PHD-mediated recruiting multiprotein complexes consisting of chromatin regu- H3 binding is essential for Chd5 to inhibit proliferation and to lators and transcription factors. Many PHDs specifically bind the induce differentiation of human neuroblastoma cells. N terminus of H3, with different PHDs recognizing H3K4me2/3 To assess the tumorigenic potential of Chd5-expressing versus H3K4me0, H3R2me0 versus H3R2me2, methylation at human neuroblastoma cells, we injected cells subcutaneously H3K9 or H3K36, or acetylation at H3K14 (Sanchez and Zhou, into two different cohorts of athymic nude mice: one receiving 2011). Our findings are in agreement with a recent report (Oliver normal food and one receiving dox-containing food (Table S6; et al., 2012) demonstrating that PHDs of CHD5 are most homol- Figure S4E). Whereas expression of WT Chd5 resulted in an ogous to H3K4me0-readers, including those of the BRAF35- 30% reduction in overall tumor volume, expression of the HDAC complex protein BHC80 (Lan et al., 2007), Autoimmune H3-binding-compromised Chd5-PHD mutants (Chd5-D361A or Regulator (AIRE; Koh et al., 2008; Org et al., 2008), Tripartite Chd5-D434A) failed to reduce tumor growth, and instead Motif-containing protein 24 (TRIM24) (Tsai et al., 2010), DNA enhanced overall tumor volume by 30%–50% (Figures 4C and (cytosine-5)-Methyltransferase 3-like protein (DNMT3L) (Ooi

(C) Chd5 transcriptionally represses target gene expression, whereas H3-binding mutants fail to do so. qRT-PCR analyses of RNA derived from MEFs with either enhanced (WT+) or compromised (Chd5-KD) Chd5, vector control, or expressing Chd5-PHD mutants indicates that whereas both WT Chd5 and Chd5 PHD mutant D346A repress target gene expression, Chd5 PHD mutants that are unable to bind H3 (D361A, D434A, and D346A/D434A) fail to do so. Data were normalized with actin, presented as mean ± SD. Fold change was calculated by comparing MEFs with (+) and without () dox. Asterisk indicates Student’s two-tailed t test (p value < 0.001). (D) Chd5 PHD mutants that cannot bind H3 form robust tumors in vivo. MEFs expressing RasG12D and either dox-inducible WT Chd5 or Chd5 PHD mutants were injected subcutaneously into athymic nude mice, and tumor formation was monitored. Y axis denotes tumor volume; X axis denotes the days after dox was provided in the diet. Kinetics of tumor growth of untreated (red circles) and treated (blue squares) mice are shown, with data represented as mean ± SD. (NB: Although some injection sites developed lesions, these lesions were extremely small [*]. The large tumor made it necessary to sacrifice this mouse at day 26 [**]). See Table S5 for detailed tumor data. (E) Tumors developing from cells expressing Chd5 PHD mutants (D361A and D434A) have derepression of Chd5 target genes. Tumor RNA was analyzed by qRT-PCR. Data were normalized with actin, and the fold change was derived by comparing tumors with the small lesion that developed in MEFs expressing WT Chd5. Data are presented as mean ± SEM. See also Figure S3, Table S3, and Table S4.

Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors 97 Figure 4. The Chd5-H3 Interaction Is Essential for Chd5 to Inhibit Proliferation, Induce Differentiation, and Suppress Tumor Growth of Human Neuroblastoma Cells In Vivo (A) Proliferation of SK-N-AS cells expressing rtTA together with either WT Chd5 or the PHD mutants. Induction of WT Chd5 and the Chd5-PHD1-D346A mutant inhibits cellular proliferation, whereas no such effects are observed with an additional mutation in D434A or with Chd5-PHD1-D361A mutants. MEFs expressing dox-inducible WT Chd5 or PHD mutants treated with dox (+dox, blue squares) or not treated (dox, red circles) are quantitated. Data are represented as mean ± SD. Y axis denotes cell numbers; X axis denotes the days after dox treatment. Student’s two-tailed t test (*p = 0.004; **p = 0.008). (B) Morphology (upper panel) and immunofluorescence staining showing expression of the neuron-specific differentiation marker MAP2 (lower panel) of human neuroblastoma cells (SK-N-AS) expressing vector control, Chd5, or PHD mutants. SK-N-AS cells expressing WT Chd5 acquire a differentiated cellular morphology and a corresponding increase in MAP2 expression compared with vector or Chd5-PHD mutant expressing SK-N-AS cells, which grow as polygonal cells in focal clusters. Cells expressing the D346A mutant have an intermediate differentiation phenotype. Scale bar: 100 mm. (C) Expression of WT Chd5 in SK-N-AS cells reduces tumorigenesis in vivo much more efficiently than does expression of vector or the Chd5 PHD mutants that are defective in H3 binding. SK-N-AS cells expressing dox-induced control vector, WT Chd5, or Chd5-PHD mutants were injected subcutaneously into athymic

(legend continued on next page) 98 Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors et al., 2007), and CHD4 (Mansfield et al., 2011; Musselman et al., et al., 2010; Koyama et al., 2012; Wong et al., 2011) suggests 2009, 2012). These PHDs lack the aromatic cage characteristic that CHD5-modulated pathways are effective targets for anti- of PHDs that specifically bind H3K4me2/3. cancer therapies. The heterozygous nature of CHD5 mutations We identified key residues of Chd5 PHDs that are conserved in human cancer raises the possibility that therapies that induce among the close family members Chd3 and Chd4, and are expression of WT CHD5 could enforce tumor suppression. To essential for mediating the H3 interaction. Mutation of these address this issue, it will be important to determine the extent residues (D361 in Chd5-PHD1, as well as D415 or D434 in to which the WT allele is silenced by DNA methylation in human Chd5-PHD2) abrogates the Chd5-H3 interaction, compromising cancers. Our finding that expression of mutant versions of Chd5 Chd5’s cellular role in inhibiting proliferation, inducing differenti- defective in H3 binding has a dominant negative effect on endog- ation, and suppressing tumorigenesis. Perturbation of H3K4me0 enous Chd5/CHD5, leading to enhanced tumorigenesis reminis- PHD readers has been associated with several human malignan- cent of Chd5 loss, cautions that effective therapeutic strategies cies. For example, mutations in AIRE are associated with auto- will need to specifically induce expression of WT but not mutant immune polyendocrinopathy-candidiasis-ectodermal dystrophy versions of CHD5. In addition, CHD5 levels must be carefully (Koh et al., 2008; Org et al., 2008), and TRIM24 expression corre- controlled to avoid deleterious effects, as even one extra copy lates inversely with survival of breast cancer patients (Tsai et al., of Chd5 causes excessive apoptosis and embryonic lethality 2010). Our findings herein define Chd5 as an H3-interacting (Bagchi et al., 2007). Given our finding that tumor growth could protein and provide a functional link between the CHD class of be inhibited in human neuroblastoma cells even in the context H3K4me0 PHD readers and suppression of tumorigenesis. of p53 deficiency, it is likely that CHD5’s multifaceted ways to Previous work indicated that CHD5 facilitates expression of enforce tumor suppression will prove useful for regulating a tumor-suppressive network that includes p16 and p19 en- diverse types of cancers, including those that involve combina- coded by the Cdkn2a locus (Bagchi and Mills, 2008; Bagchi tions of genetic lesions. The fact that CHD5 is a member of the et al., 2007). Whereas Chd5 loss enhances proliferation by Trithorax-group (TrxG) proteins, which oppose Polycomb- compromising expression of p16/Rb and p19/p53-mediated group (PcG)-mediated gene expression cascades (Mills, 2010), tumor-suppressive pathways, gain of the genomic interval en- suggests that strategies that inhibit PcG-mediated chromatin compassing Chd5 compromises proliferation by excessively dynamics could enforce CHD5 activity effectively without the activating these pathways. Excessive activation of p16/Rb deleterious effects of inducing apoptosis or cellular senescence. and p19/p53-mediated tumor-suppressive pathways causes Our finding that Chd5 inhibits expression of the oncogenic PcG apoptosis, cellular senescence, and neonatal death, which are protein Bmi1 indicates that Chd5 inhibits PcG-mediated chro- dependent on p16, p19, and p53. Here, we demonstrate for matin dynamics at multiple levels. that inducible expression of WT Chd5 inhibits proliferation, but In summary, this work defines a specific histone mark that mutant versions of Chd5 that are not able to bind H3 fail to do is bound by Chd5 and is required for its ability to regulate tran- so. We found that in addition to binding Cdkn2a, Chd5 binds scriptional cascades, inhibit proliferation, induce differentiation, and regulates the expression of multiple loci across the genome, and efficiently suppress tumorigenesis in vivo. These findings the majority of which lack the active H3K4me3 mark that we implicate Chd5 as a member of the newly appreciated class of found abrogates Chd5 binding in vitro. The finding that Chd5 unmodified H3-binding proteins, and provide mechanistic peaks are often upstream of adjacent H3K4me3 peaks suggests insight into Chd5-mediated tumor suppression. that Chd5 binding facilitates the recruitment of additional protein complexes that deposit the H3K4me3 mark, which is character- EXPERIMENTAL PROCEDURES istic of transcriptionally active genes. While our findings extend our previous studies that first linked Chd5 to Cdkn2a (Bagchi Plasmid Construction and Retroviral Expression System et al., 2007), here we show that an extensive number of addi- Mouse Chd5 cDNA from KK DNAFORM and Geneservice Ltd (Clone ID: M5C1079M20) was used as a template for generating PCR products to clone tional cancer-associated loci are bound and regulated by the PHDs and the full-length Chd5 cDNA. The PHDs were cloned into pGEX- Chd5. These include genes encoding proteins that have been 6P1 (Clontech). The dox-inducible retroviral expression system was generated implicated in chromatin dynamics and cancer-associated path- by cloning full-length Chd5 into pmCherry-C1 (Clontech) at the BglII/SalI ways. Thus, in addition to regulating Cdkn2a, a pivotal locus in restriction site, removing cherry-Chd5 by NheI/SalI cleavage, and ligating to tumorigenesis, Chd5 modulates the expression of multiple a XbaI/XhoI-cut modified version of TtRMPV-PGK-HygroR (Zuber et al., genes that regulate pathways that impinge upon the tumorigenic 2011) or TtRMPV-PGK-PuroR plasmid (see Figure S3A). The pMSCV-GFP- IRES-mNrasG12D plasmid (Zuber et al., 2009) was provided by S. Lowe. Muta- process. tions were generated by site-directed mutagenesis, and all plasmids were The finding that CHD5 status is a prognostic indicator of confirmed by sequencing. The small hairpin RNA (shRNA) construct used for survival following anticancer therapy for gallbladder carcinoma, knocking down Chd5 (shChd5-WZ) was cloned into the MLP retroviral vector. neuroblastoma, and ovarian cancer (Du et al., 2012; Garcia Retroviral infection is described in the Extended Experimental Procedures.

nude mice, and tumor formation was monitored. The Y axis denotes the fold increase of tumor volume over a 17-day period. Mice that were untreated (red) or treated with dox (blue) are shown. Data are represented as mean ± SEM. (D) Percentage change of tumor growth in dox-treated mice compared with untreated mice. Expression of WT Chd5 reduces tumorigenesis by 31%, whereas expression of Chd5-PHD mutants D361A and D434A results in a 30%–50% increase in tumorigenesis. See also Figure S4 and Table S6.

Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors 99 Antibodies (Roche). To allow comparison among primer sets, unprecipitated input Anti-Chd5 (M-182: sc-68389 [Santa Cruz] and in-house-raised polyclonal anti- samples from each condition were serially diluted and used as standards for body Chd5-232), anti-H3K4me3 (07-473; Millipore), anti-H3K9me3 (07-442; all PCRs. The ChIP-Seq analysis is described in Extended Experimental Millipore), normal rabbit IgG (2729; Cell Signaling), anti-histidine (631212; Procedures. Clontech), anti-b-actin (A5441; Sigma), and anti-MAP2 (ab11267; Abcam) were used. Cellular Assays MEFs and SK-N-AS cells were assessed for proliferation by plating 5 3 104 Recombinant Protein Production and Purification cells and 2.5 3 104 cells on 6 cm dishes or six-well plates, respectively. Cells Chd5-PHD glutathione S-transferase (GST) fusion proteins were produced by were grown in Dulbecco’s modified Eagle’s medium (DMEM) containing inducing 1–2 L of bacterial cultures (in Luria Bertani [LB] medium containing 100 mg/ml Hygromycin B (Roche) and 0.5 mg/ml Puromycin (Sigma-Aldrich),

50 mM zinc chloride) with 0.1 mM IPTG (A600 = 0.6), followed by a 14 hr incu- and were either untreated or treated with 0.2 mg/ml dox (Sigma). Cells were bation at 18C. Bacterial pellets were lysed in 100 mM Tris pH 7.5, 400 mM harvested and each plate was counted in triplicate using a Z1 Coulter particle NaCl, 5 mM dithiothreitol (DTT), 0.1 mg/ml lysozyme, and protease inhibitor counter (Beckman-Coulter) or crystal violet staining. Graphs and SEs were ob- tablets (Roche), and cleared lysates were treated with polyethyleneimine tained with Prism software. Tumorigenesis assays in athymic nude mice were before purification. The GST-PHDs were purified by binding with gluta- performed as previously described (Bagchi et al., 2007; Hemann et al., 2004). thione-agarose (Sigma) and eluted with reduced glutathione (Sigma). The Briefly, MEFs (0.5 3 106 cells) were injected subcutaneously into irradiated eluate was dialyzed in 50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 1 mM DTT athymic nude mice, and dox was administered in food and drinking water overnight and used for biochemical assays. On-column cleavage of GST (2 mg/ml) for 7 days and continued with dox only in food. Matrigel-mixed was performed with PreScission Protease (GE Healthcare), and GST-cleaved 1.6 3 106 SK-N-AS cells were injected subcutaneously into two cohorts of irra- purified PHD was dialyzed with assay buffer. diated athymic nude mice. One group received a normal diet and the other group received the dox diet 2 days prior to injections and continued on the In Vitro Histone Peptide Array, Peptide Pull-Down, and Fluorescence dox diet for the rest of the experiment. Tumor development was monitored Polarization Assays via measurements by a blinded observer, and tumors were harvested and Peptide microarray experiments were performed as previously described (Koh processed for subsequent analysis. et al., 2008). Biotinylated-peptide pull-down assays using recombinant proteins were performed as previously described (Shi et al., 2006). C-terminal ACCESSION NUMBERS biotinylated peptides (Table S8) were purchased from Millipore. Pull-down assays using nuclear extracts prepared from WT mouse brain using the The sequencing data have been deposited in the Short Read Archive (http:// Dignam protocol (Dignam et al., 1983) were performed as previously described www.ncbi.nlm.nih.gov/sra) under accession number SRA062358. (Wysocka et al., 2006) with some modifications. Briefly, brains were dissected from C57BL/6 mice, crushed in liquid nitrogen, and lysed according to the SUPPLEMENTAL INFORMATION Dignam protocol. Nuclear extracts were precleared with streptavidin beads (Amersham) and incubated with peptide that had been prebound to streptavi- Supplemental Information includes four figures, eight tables, and Extended din beads for 3 hr at 4C in binding buffer (20 mM HEPES, pH 7.9 at 4C, 20% Experimental Procedures and can be found with this article online at http:// glycerol, 0.15 M KCl, 0.5 mM DTT, 0.2% (v/v) Triton X-100, and protease inhib- dx.doi.org/10.1016/j.celrep.2012.12.009. itor). Approximately 5 mg of peptide was used per pull-down. Beads were washed eight times and the bound proteins were subjected to Coomassie staining and western blot analyses. LICENSING INFORMATION Fluorescence polarization (FP) experiments were carried out using a Biotek Synergy 4 plate reader (Biotek) at 30C as previously described (Jacobs et al., This is an open-access article distributed under the terms of the Creative 2004). Fluorescein-labeled peptide (10 nM, synthesized as ARTKQTAR Commons Attribution-NonCommercial-No Derivative Works License, which KSTGGKAPRKQLAK-Flu; Peptide Protein Research Ltd.) was incubated for permits non-commercial use, distribution, and reproduction in any medium, 30 min at 30C with varying amounts of purified GST-cleaved PHDs in binding provided the original author and source are credited. buffer (50 mM Tris, pH 7.0, 150 mM NaCl, 50 mM ZnCl2, 5 mM DTT). Protein concentration was measured by absorbance spectroscopy (PHD: ACKNOWLEDGMENTS ε280 = 23460 M1cm1; PHD1: ε280 = 15,845 M1cm1; PHD2: ε280 = 6,990 M1cm1). Peptide concentration was determined using absorbance We thank members of the Mills laboratory for advice and assistance, W. Li for spectroscopy (extinction coefficient for fluoresceinated peptide ε492 = generating the Chd5 shRNA (shChd5-WZ), S. Lowe for providing plasmids and 68,000 M1cm1). Binding curves were analyzed using the anisotropy from rtTA mice, S. Hearn for help with microscopy, E. Wang and C. Vakoc for advice two or three independent experiments. regarding ChIP-qPCR, and L. Bianco and her staff for help with tumor studies. This work was supported by The SASS Foundation for Medical Research ChIP Assays and ChIP-Seq Analyses (S.P.), NIH RO1 CA127383 (A.A.M.), NIH RO1 GM079641 (O.G.), and an Ellison Chromatin was prepared from WT MEFs and used for immunoprecipitation. Senior Scholar in Aging Award (O.G.). ChIP analyses were performed as described previously (Zeng et al., 2006) with some modifications. For these analyses, 1 3 107 cells were used per Received: May 5, 2012 immunoprecipitation and were sequentially cross-linked with ethylene glyco- Revised: November 6, 2012 lbis[succinimidyl succinate] (EGS; Thermo Scientific) for 30 m, followed by Accepted: December 13, 2012 1% formaldehyde for 10 m at room temperature. Cross-linking was quenched Published: January 10, 2013 with glycine, cells were washed and lysed, and chromatin was sheared by sonication. Chromatin was immunoprecipitated with Dyna-beads preconju- REFERENCES gated with primary antibodies specific for normal rabbit IgG, Chd5 (M-182), or H3K4me3. The immunoprecipitated samples were washed and eluted, Agrawal, N., Frederick, M.J., Pickering, C.R., Bettegowda, C., Chang, K., Li, and cross-linking was reversed with 200 mM NaCl for 8 hr. Samples were di- R.J., Fakhry, C., Xie, T.X., Zhang, J., Wang, J., et al. (2011). Exome sequencing gested with RNase A and treated with proteinase K before phenol/chloroform of head and neck squamous cell carcinoma reveals inactivating mutations in extraction, and then used for ChIP-Solexa sequencing (performed at the CSHL NOTCH1. Science 333, 1154–1157. Genomics Shared Resource) or qPCR. Samples were analyzed in triplicate by Bagchi, A., and Mills, A.A. (2008). The quest for the 1p36 tumor suppressor. real-time PCR using SYBR green (Quanta Biosciences) on the LightCycler 480 Cancer Res. 68, 2551–2556.

100 Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors Bagchi, A., Papazoglu, C., Wu, Y., Capurso, D., Brodt, M., Francis, D., Bredel, Jacobs, S.A., Fischle, W., and Khorasanizadeh, S. (2004). Assays for the deter- M., Vogel, H., and Mills, A.A. (2007). CHD5 is a tumor suppressor at human mination of structure and dynamics of the interaction of the chromodomain 1p36. Cell 128, 459–475. with histone peptides. Methods Enzymol. 376, 131–148. Baker, L.A., Allis, C.D., and Wang, G.G. (2008). PHD fingers in human Jones, S., Wang, T.L., Shih, IeM., Mao, T.L., Nakayama, K., Roden, R., Glas, diseases: disorders arising from misinterpreting epigenetic marks. Mutat. R., Slamon, D., Diaz, L.A., Jr., Vogelstein, B., et al. (2010). Frequent mutations Res. 647, 3–12. of chromatin remodeling gene ARID1A in ovarian clear cell carcinoma. Science Barski, A., Cuddapah, S., Cui, K., Roh, T.Y., Schones, D.E., Wang, Z., Wei, G., 330, 228–231. Chepelev, I., and Zhao, K. (2007). High-resolution profiling of histone methyl- Kaghad, M., Bonnet, H., Yang, A., Creancier, L., Biscan, J.C., Valent, A., Minty, ations in the human genome. Cell 129, 823–837. A., Chalon, P., Lelias, J.M., Dumont, X., et al. (1997). Monoallelically expressed Berger, M.F., Lawrence, M.S., Demichelis, F., Drier, Y., Cibulskis, K., Siva- gene related to p53 at 1p36, a region frequently deleted in neuroblastoma and chenko, A.Y., Sboner, A., Esgueva, R., Pflueger, D., Sougnez, C., et al. other human cancers. Cell 90, 809–819. (2011). The genomic complexity of primary human prostate cancer. Nature Koh, A.S., Kuo, A.J., Park, S.Y., Cheung, P., Abramson, J., Bua, D., Carney, D., 470, 214–220. Shoelson, S.E., Gozani, O., Kingston, R.E., et al. (2008). Aire employs Bouazoune, K., Mitterweger, A., La¨ ngst, G., Imhof, A., Akhtar, A., Becker, P.B., a histone-binding module to mediate immunological tolerance, linking chro- and Brehm, A. (2002). The dMi-2 chromodomains are DNA binding matin regulation with organ-specific autoimmunity. Proc. Natl. Acad. Sci. modules important for ATP-dependent nucleosome mobilization. EMBO J. USA 105, 15878–15883. 21, 2430–2440. Koyama, H., Zhuang, T., Light, J.E., Kolla, V., Higashi, M., McGrady, P.W., Cancer Genome Atlas Research Network. (2008). Comprehensive genomic London, W.B., and Brodeur, G.M. (2012). Mechanisms of CHD5 Inactivation characterization defines human glioblastoma genes and core pathways. in neuroblastomas. Clin. Cancer Res. 18, 1588–1597. Nature 455, 1061–1068. Lan, F., Collins, R.E., De Cegli, R., Alpatov, R., Horton, J.R., Shi, X., Gozani, O., Cancer Genome Atlas Research Network. (2011). Integrated genomic Cheng, X., and Shi, Y. (2007). Recognition of unmethylated histone H3 lysine 4 analyses of ovarian carcinoma. Nature 474, 609–615. links BHC80 to LSD1-mediated gene repression. Nature 448, 718–722. Dignam, J.D., Lebovitz, R.M., and Roeder, R.G. (1983). Accurate transcription Lang, J., Tobias, E.S., and Mackie, R. (2011). Preliminary evidence for involve- initiation by RNA polymerase II in a soluble extract from isolated mammalian ment of the tumour suppressor gene CHD5 in a family with cutaneous mela- nuclei. Nucleic Acids Res. 11, 1475–1489. noma. Br. J. Dermatol. 164, 1010–1016. Du, X., Wu, T., Lu, J., Zang, L., Song, N., Yang, T., Zhao, H., and Wang, S. Li, M., Zhao, H., Zhang, X., Wood, L.D., Anders, R.A., Choti, M.A., Pawlik, T.M., (2012). Decreased expression of chromodomain helicase DNA-binding protein Daniel, H.D., Kannangai, R., Offerhaus, G.J., et al. (2011). Inactivating muta- 5 is an unfavorable prognostic marker in patients with primary gallbladder tions of the chromatin remodeling gene ARID2 in hepatocellular carcinoma. carcinoma. Clin. Transl. Oncol., Jul 24 [Epub ahead of print]. Nat. Genet. 43, 828–829. Flanagan, J.F., Mi, L.Z., Chruszcz, M., Cymborowski, M., Clines, K.L., Kim, Y., Lusser, A., Urwin, D.L., and Kadonaga, J.T. (2005). Distinct activities of CHD1 Minor, W., Rastinejad, F., and Khorasanizadeh, S. (2005). Double chromodo- and ACF in ATP-dependent chromatin assembly. Nat. Struct. Mol. Biol. 12, mains cooperate to recognize the methylated histone H3 tail. Nature 438, 160–166. 1181–1185. Mansfield, R.E., Musselman, C.A., Kwan, A.H., Oliver, S.S., Garske, A.L., Fujita, T., Igarashi, J., Okawa, E.R., Gotoh, T., Manne, J., Kolla, V., Kim, J., Davrazou, F., Denu, J.M., Kutateladze, T.G., and Mackay, J.P. (2011). Plant Zhao, H., Pawel, B.R., London, W.B., et al. (2008). CHD5, a tumor suppressor homeodomain (PHD) fingers of CHD4 are histone H3-binding modules with gene deleted from 1p36.31 in neuroblastomas. J. Natl. Cancer Inst. 100, preference for unmodified H3K4 and methylated H3K9. J. Biol. Chem. 286, 940–949. 11779–11791. Garcia, I., Mayol, G., Rodrı´guez, E., Sun˜ ol, M., Gershon, T.R., Rı´os, J., Cheung, Marfella, C.G., and Imbalzano, A.N. (2007). The Chd family of chromatin N.K., Kieran, M.W., George, R.E., Perez-Atayde, A.R., et al. (2010). Expression remodelers. Mutat. Res. 618, 30–40. of the neuron-specific protein CHD5 is an independent marker of outcome in Matthews, A.G., Kuo, A.J., Ramo´ n-Maiques, S., Han, S., Champagne, K.S., neuroblastoma. Mol. Cancer 9, 277. Ivanov, D., Gallardo, M., Carney, D., Cheung, P., Ciccone, D.N., et al. (2007). Garcia, I., Mayol, G., Rı´os, J., Domenech, G., Cheung, N.K., Oberthuer, A., RAG2 PHD finger couples histone H3 lysine 4 trimethylation with V(D)J recom- Fischer, M., Maris, J.M., Brodeur, G.M., Hero, B., et al. (2012). A three-gene bination. Nature 450, 1106–1110. expression signature model for risk stratification of patients with neuroblas- Mills, A.A. (2010). Throwing the cancer switch: reciprocal roles of polycomb toma. Clin. Cancer Res. 18, 2012–2023. and trithorax proteins. Nat. Rev. Cancer 10, 669–682. Godfried, M.B., Veenstra, M., Valent, A., Sluis, Pv., Vouˆ te, P.A., Versteeg, R., Mokarram, P., Kumar, K., Brim, H., Naghibalhossaini, F., Saberi-firoozi, M., and Caron, H.N. (2002). Lack of interstitial chromosome 1p deletions in Nouraie, M., Green, R., Lee, E., Smoot, D.T., and Ashktorab, H. (2009). Distinct clinically-detected neuroblastoma. Eur. J. Cancer 38, 1513–1519. high-profile methylated genes in colorectal cancer. PLoS ONE 4, e7012. Goldschneider, D., Horvilleur, E., Plassa, L.F., Guillaud-Bataille, M., Million, K., Mulero-Navarro, S., and Esteller, M. (2008). Chromatin remodeling factor Wittmer-Dupret, E., Danglot, G., de The´ , H., Be´ nard, J., May, E., and Douc- CHD5 is silenced by promoter CpG island hypermethylation in human cancer. Rasy, S. (2006). Expression of C-terminal deleted p53 isoforms in neuroblas- Epigenetics 3, 210–215. toma. Nucleic Acids Res. 34, 5603–5612. Murawska, M., and Brehm, A. (2011). CHD chromatin remodelers and the Gorringe, K.L., Choong, D.Y., Williams, L.H., Ramakrishna, M., Sridhar, A., Qiu, transcription cycle. Transcription 2, 244–253. W., Bearfoot, J.L., and Campbell, I.G. (2008). Mutation and methylation Musgrove, E.A., Caldon, C.E., Barraclough, J., Stone, A., and Sutherland, R.L. analysis of the chromodomain-helicase-DNA binding 5 gene in ovarian cancer. (2011). Cyclin D as a therapeutic target in cancer. Nat. Rev. Cancer 11, Neoplasia 10, 1253–1258. 558–572. Hemann, M.T., Zilfou, J.T., Zhao, Z., Burgess, D.J., Hannon, G.J., and Lowe, Musselman, C.A., Mansfield, R.E., Garske, A.L., Davrazou, F., Kwan, A.H., S.W. (2004). Suppression of tumorigenesis by the p53 target PUMA. Proc. Oliver, S.S., O’Leary, H., Denu, J.M., Mackay, J.P., and Kutateladze, T.G. Natl. Acad. Sci. USA 101, 9333–9338. (2009). Binding of the CHD4 PHD2 finger to histone H3 is modulated by cova- Hers, I., Vincent, E.E., and Tavare´ , J.M. (2011). Akt signalling in health and lent modifications. Biochem. J. 423, 179–187. disease. Cell. Signal. 23, 1515–1527. Musselman, C.A., Ramı´rez, J., Sims, J.K., Mansfield, R.E., Oliver, S.S., Denu, Jacobs, S.A., and Khorasanizadeh, S. (2002). Structure of HP1 chromodomain J.M., Mackay, J.P., Wade, P.A., Hagman, J., and Kutateladze, T.G. (2012). bound to a lysine 9-methylated histone H3 tail. Science 295, 2080–2083. Bivalent recognition of nucleosomes by the tandem PHD fingers of the

Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors 101 CHD4 ATPase is required for CHD4-mediated repression. Proc. Natl. Acad. novel direct and selective binding modules of histone H3 methylated at either Sci. USA 109, 787–792. lysine 4 or lysine 36. J. Biol. Chem. 282, 2450–2455. Nielsen, P.R., Nietlispach, D., Mott, H.R., Callaghan, J., Bannister, A., Kouzar- Sims, J.K., and Wade, P.A. (2011). SnapShot: chromatin remodeling: CHD. ides, T., Murzin, A.G., Murzina, N.V., and Laue, E.D. (2002). Structure of the Cell 144, 626–626. HP1 chromodomain bound to histone H3 methylated at lysine 9. Nature 416, Sjo¨ blom, T., Jones, S., Wood, L.D., Parsons, D.W., Lin, J., Barber, T.D., 103–107. Mandelker, D., Leary, R.J., Ptak, J., Silliman, N., et al. (2006). The consensus Okawa, E.R., Gotoh, T., Manne, J., Igarashi, J., Fujita, T., Silverman, K.A., coding sequences of human breast and colorectal cancers. Science 314, Xhao, H., Mosse, Y.P., White, P.S., and Brodeur, G.M. (2008). Expression 268–274. and sequence analysis of candidates for the 1p36.31 tumor suppressor Tsai, W.W., Wang, Z., Yiu, T.T., Akdemir, K.C., Xia, W., Winter, S., Tsai, C.Y., gene deleted in neuroblastomas. Oncogene 27, 803–810. Shi, X., Schwarzer, D., Plunkett, W., et al. (2010). TRIM24 links a non-canonical Oliver, S.S., Musselman, C.A., Srinivasan, R., Svaren, J.P., Kutateladze, T.G., histone signature to breast cancer. Nature 468, 927–932. and Denu, J.M. (2012). Multivalent recognition of histone tails by the PHD Wang, G.G., Song, J., Wang, Z., Dormann, H.L., Casadio, F., Li, H., Luo, J.L., fingers of CHD5. Biochemistry 51, 6534–6544. Patel, D.J., and Allis, C.D. (2009a). Haematopoietic malignancies caused by Ooi, S.K., Qiu, C., Bernstein, E., Li, K., Jia, D., Yang, Z., Erdjument-Bromage, dysregulation of a chromatin-binding PHD finger. Nature 459, 847–851. H., Tempst, P., Lin, S.P., Allis, C.D., et al. (2007). DNMT3L connects unmethy- Wang, X., Lau, K.K., So, L.K., and Lam, Y.W. (2009b). CHD5 is down-regulated lated lysine 4 of histone H3 to de novo methylation of DNA. Nature 448, through promoter hypermethylation in gastric cancer. J. Biomed. Sci. 16, 95. 714–717. Wong, R.R., Chan, L.K., Tsang, T.P., Lee, C.W., Cheung, T.H., Yim, S.F., Siu, Org, T., Chignola, F., Hete´ nyi, C., Gaetani, M., Rebane, A., Liiv, I., Maran, U., N.S., Lee, S.N., Yu, M.Y., Chim, S.S., et al. (2011). CHD5 Downregulation Mollica, L., Bottomley, M.J., Musco, G., and Peterson, P. (2008). The autoim- Associated with Poor Prognosis in Epithelial Ovarian Cancer. Gynecol. Obstet. mune regulator PHD finger binds to non-methylated histone H3K4 to activate Invest. 72, 203–207. gene expression. EMBO Rep. 9, 370–376. Wysocka, J., Swigut, T., Xiao, H., Milne, T.A., Kwon, S.Y., Landry, J., Kauer, Rajakumara, E., Wang, Z., Ma, H., Hu, L., Chen, H., Lin, Y., Guo, R., Wu, F., Li, M., Tackett, A.J., Chait, B.T., Badenhorst, P., et al. (2006). A PHD finger of H., Lan, F., et al. (2011). PHD finger recognition of unmodified histone H3R2 NURF couples histone H3 lysine 4 trimethylation with chromatin remodelling. links UHRF1 to regulation of euchromatic gene expression. Mol. Cell 43, Nature 442, 86–90. 275–284. Yadav, S., Singhal, J., Singhal, S.S., and Awasthi, S. (2009). hSET1: a novel Robbins, C.M., Tembe, W.A., Baker, A., Sinari, S., Moses, T.Y., Beckstrom- approach for colon cancer therapy. Biochem. Pharmacol. 77, 1635–1641. Sternberg, S., Beckstrom-Sternberg, J., Barrett, M., Long, J., Chinnaiyan, Yao, H., Ashihara, E., and Maekawa, T. (2011). Targeting the Wnt/b-catenin A., et al. (2011). Copy number and targeted mutational analysis reveals novel signaling pathway in human cancers. Expert Opin. Ther. Targets 15, 873–887. somatic events in metastatic prostate tumors. Genome Res. 21, 47–55. Zeng, P.Y., Vakoc, C.R., Chen, Z.C., Blobel, G.A., and Berger, S.L. (2006). Saksouk, N., Avvakumov, N., Champagne, K.S., Hung, T., Doyon, Y., Cayrou, In vivo dual cross-linking for identification of indirect DNA-associated proteins C., Paquet, E., Ullah, M., Landry, A.J., Coˆ te´ , V., et al. (2009). HBO1 HAT by chromatin immunoprecipitation. Biotechniques 41, 694, 696, 698. complexes target chromatin throughout gene coding regions via multiple Zhao, R., Yan, Q., Lv, J., Huang, H., Zheng, W., Zhang, B., and Ma, W. (2011). PHD finger interactions with histone H3 tail. Mol. Cell 33, 257–265. CHD5, a tumor suppressor that is epigenetically silenced in lung cancer. Lung Sanchez, R., and Zhou, M.M. (2011). The PHD finger: a versatile epigenome Cancer 76, 324–331. reader. Trends Biochem. Sci. 36, 364–372. Zuber, J., Radtke, I., Pardee, T.S., Zhao, Z., Rappaport, A.R., Luo, W., McCur- Shi, X., Hong, T., Walter, K.L., Ewalt, M., Michishita, E., Hung, T., Carney, D., rach, M.E., Yang, M.M., Dolan, M.E., Kogan, S.C., et al. (2009). Mouse models Pen˜ a, P., Lan, F., Kaadige, M.R., et al. (2006). ING2 PHD domain links histone of human AML accurately predict chemotherapy response. Genes Dev. 23, H3 lysine 4 methylation to active gene repression. Nature 442, 96–99. 877–889. Shi, X., Kachirskaia, I., Walter, K.L., Kuo, J.H., Lake, A., Davrazou, F., Chan, Zuber, J., McJunkin, K., Fellmann, C., Dow, L.E., Taylor, M.J., Hannon, G.J., S.M., Martin, D.G., Fingerman, I.M., Briggs, S.D., et al. (2007). Proteome- and Lowe, S.W. (2011). Toolkit for evaluating genes required for proliferation wide analysis in Saccharomyces cerevisiae identifies several PHD fingers as and survival using tetracycline-regulated RNAi. Nat. Biotechnol. 29, 79–83.

102 Cell Reports 3, 92–102, January 31, 2013 ª2013 The Authors