Since the time that CHD5
was first reported as a tumor suppressor mapping to human 1p36 (Bagchi et al., 2007
), its inactivation has been documented in a diverse array of human cancers (Agrawal et al., 2011
; Berger et al., 2011
; Gorringe et al., 2008
; Jones et al., 2010
; Koyama et al., 2012
; Lang et al., 2011
; Li et al., 2011
; Mokarram et al., 2009
; Mulero-Navarro and Esteller, 2008
; Okawa et al., 2008
; Robbins et al., 2011
; TCGA, 2011
; Wang et al., 2009b
; Zhao et al., 2011
), which indicates that CHD5 regulates processes fundamental to cancer prevention. Therefore, defining how CHD5 protects from tumorigenesis may impact treatment of a diverse array of human cancers. Although belonging to a protein family that includes members implicated in chromatin remodeling, the mechanism whereby CHD5 exerts its tumor suppressive role had remained unexplored. Here we demonstrate that the ability of Chd5 to bind unmodified H3 is essential for tumor suppression.
By focusing on the mechanism whereby CHD5 interacts with chromatin, we discovered that the tandem PHDs mediate binding to N-terminally unmodified H3. PHDs are modules initially identified in plant homeodomain proteins; subsequent PHD-containing proteins were discovered in yeast, fly, humans, especially in chromatin-associated and nuclear proteins. Although the zinc binding motifs of PHD motifs are well conserved, diversity in the ligand binding residues generates versatility in their interaction partners. PHDs are histone ‘readers’ that control gene expression cascades by recruiting multi-protein complexes consisting of chromatin regulators and transcription factors. Many PHDs specifically bind the N-terminus of H3, with different PHDs recognizing H3K4me2/3 vs. H3K4me0, H3R2me0 vs. H3R2me2, methylation at H3K9 or H3K36, or acetylation at H3K14 (Sanchez and Zhou, 2011
). Our findings are in agreement with a recent report (Oliver et al., 2012
) demonstrating that PHDs of CHD5 are most homologous to H3K4me0-readers, including those of the BRAF35–HDAC complex protein (BHC80) (Lan et al., 2007
gulator (AIRE) (Koh et al., 2008
; Org et al., 2008
otif-containing protein 24
(TRIM24) (Tsai et al., 2010
), and DN
ike protein (DNMT3L) (Ooi et al., 2007
), and CHD4 (Mansfield et al., 2011
; Musselman et al., 2009
; Musselman et al., 2012
). These PHDs lack the aromatic cage characteristic of PHDs that specifically bind H3K4me2/3.
We identified key residues of Chd5 PHDs that are conserved among the close family members Chd3 and Chd4 and are essential for mediating the H3 interaction. Mutation of these residues (D361 in Chd5-PHD1, as well as D415 or D434 in Chd5-PHD2) abrogates the Chd5: H3 interaction, compromising Chd5’s cellular role in inhibiting proliferation, inducing differentiation, and suppressing tumorigenesis. Perturbation of H3K4me0 PHD-readers has been associated with several human malignancies. For example, mutations in AIRE are associated with autoimmune polyendocrinopathy-candiasis-ectodermal dystrophy (APECED) (Koh et al., 2008
; Org et al., 2008
), and TRIM24 expression correlates inversely with survival of breast cancer patients (Tsai et al., 2010
). Our findings herein define Chd5 as an H3-interacting protein and provide the first functional link between the CHD class of H3K4me0 PHD-readers and suppression of tumorigenesis.
Previous work indicated that CHD5 facilitates expression of a tumor suppressive network including p16 and p19 encoded by the Cdkn2a
locus (Bagchi and Mills, 2008
; Bagchi et al., 2007
). Whereas Chd5 loss enhances proliferation by compromising expression of p16/Rb and p19/p53-mediated tumor suppressive pathways, gain of the genomic interval encompassing Chd5
compromises proliferation by excessively activating these pathways. Excessive activation of p16/Rb and p19/p53-mediated tumor suppressive pathways causes apoptosis, cellular senescence, and neonatal death that are dependent upon p16, p19, and p53. Here we demonstrate for the first time that inducible expression of wild-type Chd5 inhibits proliferation, and that mutant versions of Chd5 that are not able to bind H3 fail to do so. We found that in addition to binding Cdkn2a
, Chd5 binds and regulates expression of multiple loci across the genome, the majority of which lack the active H3K4me3 mark that we found abrogates Chd5 binding in vitro.
The finding that Chd5 peaks are often upstream of adjacent H3K4me3 peaks suggests that Chd5 binding facilitates recruitment of additional protein complexes that deposit the H3K4me3 mark characteristic of transcriptionally active genes. While our findings extend our previous studies that first linked Chd5 to Cdkn2a
(Bagchi et al., 2007
), here we show that an extensive number of additional cancer-associated loci are bound and regulated by Chd5. These include genes encoding proteins implicated in chromatin dynamics and cancer-associated pathways. Thus, in addition to regulating Cdkn2a
—a locus pivotal in tumorigenesis—Chd5 modulates expression of multiple genes that regulate pathways that impinge upon the tumorigenic process.
The finding that CHD5
status is a prognostic indicator of survival following anti-cancer therapy for gallbladder carcinoma, neuroblastoma and ovarian cancer (Du et al., 2012
; Garcia et al., 2010
; Koyama et al., 2012
; Wong et al., 2011
) suggests that CHD5-modulated pathways are effective targets for anti-cancer therapies. The heterozygous nature of CHD5
mutations in human cancer leaves open the possibility that therapies that induce expression of wild-type CHD5
could enforce tumor suppression. To this end, it will be important to determine the extent to which the wild-type allele is silenced by DNA methylation in human cancers. Our finding that expression of mutant versions of Chd5 defective in H3 binding has a dominant negative effect on endogenous Chd5/CHD5 leading to enhanced tumorigenesis reminiscent of Chd5 loss, cautions that effective therapeutic strategies will need to specifically induce expression of wild type, but not mutant versions of CHD5
. In addition, CHD5 levels must be carefully controlled to avoid deleterious effects, as even one extra copy of Chd5
causes excessive apoptosis and embryonic lethality (Bagchi et al., 2007
). Given that we found that tumor growth could be inhibited in human neuroblastoma cells even in the context of p53-deficiency, it is likely that CHD5’s multi-faceted ways to enforce tumor suppression will prove useful for regulating diverse types of cancers, as well as those with a myriad of combinations of genetic lesions. The fact that CHD5 is a member of the TrxG group of proteins that opposes PcG-mediated gene expression cascades (Mills, 2010
) suggests that strategies inhibiting PcG-mediated chromatin dynamics are effective for enforcing CHD5 activity without having the deleterious effects of inducing apoptosis or cellular senescence. Our finding that Chd5 inhibits expression of the oncogenic PcG protein Bmi1 indicates that Chd5 inhibits PcG-mediated chromatin dynamics at multiple levels.
In summary, this work defines a specific histone mark that Chd5 binds which is required for its ability to regulate transcriptional cascades, to inhibit proliferation, to induce differentiation, and to efficiently suppress tumorigenesis in vivo. These findings implicate Chd5 as a member of the newly appreciated class of unmodified H3-binding proteins, providing the first mechanistic insight into Chd5-mediated tumor suppression.