|Home | About | Journals | Submit | Contact Us | Français|
Histone variants are emerging as key regulatory molecules in cancer. Here we report a novel role for the H2A.Z isoform H2A.Z.2 as a driver of malignant melanoma. H2A.Z.2 is highly expressed in metastatic melanoma, correlates with decreased patient survival, and is required for cellular proliferation. Our integrated genomic analyses reveal that H2A.Z.2 controls the transcriptional output of E2F target genes in melanoma cells. These genes are highly expressed and display a distinct signature of H2A.Z occupancy. We identify BRD2 as an H2A.Z interacting protein, whose levels are also elevated in melanoma. We further demonstrate that H2A.Z.2 regulated genes are bound by BRD2 and E2F1 in a H2A.Z.2-dependent manner. Importantly, H2A.Z.2 deficiency sensitizes melanoma cells to chemotherapy and targeted therapies. Collectively, our findings implicate H2A.Z.2 as a mediator of cell proliferation and drug sensitivity in malignant melanoma, holding translational potential for novel therapeutic strategies.
Malignant melanoma is the most lethal form of skin cancer with rising incidence and remains largely incurable. While advances in immune and targeted therapies have made tremendous progress recently (Chapman et al., 2011; Kaufman et al., 2013), they are effective only in distinct subsets of patients or result in the emergence of drug resistance (Lito et al., 2013). Thus, investigation of alternative approaches is essential.
Recent studies have shed light on the importance of epigenetic regulation in melanoma biology. Key roles for BRD4 (Segura et al., 2013), histone methyltransferases SETDB1 (Ceol et al., 2011) and EZH2 (Zingg et al., 2015), and the histone variant macroH2A (Kapoor et al., 2010) have been reported. Relevant to the present study, histone variants and their chaperones are emerging as key regulatory molecules in cancer (Vardabasso et al., 2013).
H2A.Z is a highly conserved H2A variant, with only 60% identity to canonical H2A, and is expressed and incorporated into chromatin throughout the cell cycle (Bonisch and Hake, 2012). While somewhat confounded by species-specific functions and context-dependencies, the role of H2A.Z in transcriptional regulation is well established (Svotelis et al., 2009). H2A.Z is enriched at gene promoters, as well as other regulatory regions, generally exerting a positive role on transcription (Hu et al., 2012; Obri et al., 2014).
Two distinct H2A.Z isoforms, H2A.Z.1 and H2A.Z.2, have been identified in the vertebrate genome as products of two non-allelic genes, H2AFZ and H2AFV, respectively (Dryhurst et al., 2009; Horikoshi et al., 2013; Matsuda et al., 2010). While differing by only three amino acids at the protein level, H2A.Z.1 and H2A.Z.2 are encoded by distinct nucleotide sequences. Isoform-specific functions remain unclear, and H2A.Z.1 mouse knock out studies suggest that the two genes are non-redundant (Faast et al., 2001). In the context of tumorigenesis, H2A.Z is overexpressed in breast, prostate and bladder cancers, where in some cases, it regulates proliferation (reviewed in Vardabasso et al., 2013). However, these studies either focused solely on H2A.Z.1, or did not clearly distinguish between isoforms.
Here we report a distinct role for H2A.Z.2 in melanoma. H2A.Z.2 is highly expressed in melanoma and drives proliferation by promoting expression of E2F target genes. These cell cycle regulatory genes are highly expressed and acquire a unique signature of H2A.Z occupancy – high promoter enrichment and gene body depletion. We further identified the BET (bromodomain and extraterminal domain) protein BRD2 as an H2A.Z interacting protein, whose levels are also elevated in melanoma specimens. Depletion of H2A.Z.2 results in reduced histone acetylation, BRD2 and E2F1 levels, and impairs recruitment of BRD2 and E2F1 to its target genes. Moreover, H2A.Z.2 deficiency cooperates with BET or MEK inhibition to induce melanoma cell death. Hence, our studies suggest that targeting H2A.Z deposition may be effective therapeutically in combination with existing or emerging therapies for melanoma.
By probing a panel of primary and metastatic melanoma cell lines we detected increased levels of H2A.Z protein in metastatic cells (Figure 1A). Immunoblotting of histones extracted from benign nevi and melanoma specimens revealed increased H2A.Z in melanoma tissues (Figure 1B). We also investigated H2A.Z levels in human primary melanocytes induced to senesce via serial passaging (replicative senescence) and BRAFV600E (oncogene-induced senescence) (Duarte et al., 2014). We observed diminished H2A.Z upon both modes of senescence (Figure S1A). Together, these data link global levels of H2A.Z to cellular proliferation.
To assess whether H2A.Z expression is regulated transcriptionally, as well as to examine the individual H2A.Z isoforms (which is not possible with currently available antibodies), we performed quantitative RT-PCR (qRT-PCR) using isoform-specific primers. H2A.Z isoforms are decreased in human melanocytes induced to senescence (Figure S1A). Conversely, in a panel of benign nevi and melanoma specimens, we observed increased H2A.Z.1 and H2A.Z.2 mRNA in melanoma (Figure 1C). H2A.Z.1 and H2A.Z.2 mRNA levels were also increased in cell lines derived from metastatic versus primary melanoma Figure S1B). Analysis of published transcriptional data is consistent with these findings (Talantov et al., 2005; Riker et al., 2008; Xu et al., 2008) (Figure S1C). Finally, in a cohort of patients followed clinically for 3 years after excision of metastatic lesions (Bogunovic et al., 2009), patients with high H2A.Z.1 and H2A.Z.2 showed significantly lower survival (Figure 1D). Collectively, these findings suggest that H2A.Z isoforms have a functional role in melanoma progression.
We next performed quantitative copy number analysis of H2A.Z.1 and H2A.Z.2 in nevi and metastases by qPCR and detected copy gains for both (Figure 1E). The Cancer Genome Atlas (TCGA) reports increased copy number in 13% and 52% of cutaneous melanomas for H2A.Z.1 and H2A.Z.2, respectively, which correlates with increased mRNA levels (Figure S1D). Fluorescent In Situ Hybridization (FISH) of melanoma cell lines corroborated these findings (Figure S1E).
Next we investigated the functional consequences of depleting H2A.Z.1 and H2A.Z.2 in melanoma cell lines. Using sequence-specific shRNAs for H2A.Z isoforms, we established stable SK-mel147 (NRASQ61R), WM266-4 (BRAFV600D), and 501mel (BRAFV600E) cell lines targeting either H2A.Z.1 or H2A.Z.2. Knock down was monitored by qRT-PCR and/or immunoblot (Figure S2A–C). As H2A.Z.1 is the predominant isoform in melanoma (via RNA-sequencing, below), its knock down can be appreciated at the protein level, while H2A.Z.2 knock down is obscured by H2A.Z.1 (Figure S2A, B).
We observed that loss of H2A.Z.2, but not H2A.Z.1, reduced proliferation in all cell lines (Figure 2A, B, S2D). To confirm these variant-specific effects, we generated cells stably expressing H2A.Z.1 or an shRNA-resistant H2A.Z.2 that were infected with sh_Z.2 and a control (sh_scr). Only those cells expressing an shRNA-resistant H2A.Z.2 were able to overcome the proliferation defect induced by sh_Z.2 (Figure 2C). Interestingly, HeLa cells depleted of H2A.Z isoforms did not show proliferation defects (Figure S2E). Thus, H2A.Z isoforms exert distinct and non-redundant functions in melanoma cells.
H2A.Z.2 knock down induced a G1/S cell cycle arrest (Figure 2D, E, S2F, G), accompanied by hypophosphorylation of Rb and decreased levels of Cyclins E and A (Figure 2F). This phenotype was not consistent with cellular senescence, as the expression of cyclin-dependent kinase (CDK) inhibitors (Figure S2H) and β–galactosidase activity (data not shown) were not increased. Moreover, we observed minimal cell death (Figure S2I). Next, SK-mel147 cells were arrested at early S phase via double thymidine block and subsequently released. Both control and H2A.Z.1 knock down cells progressed through S and G2 phases, and at 10 hours ~30% of the cells re-entered G1. However, H2A.Z.2-depleted cells remained largely arrested for the entire duration of the assay (Figure 2G). These findings suggest that H2A.Z.2 loss causes delayed entry into S phase. These data are strikingly similar to htz1 (H2A.Z) mutant budding yeast, which shows delayed DNA replication and cell cycle progression (Dhillon et al., 2006).
To further understand the observed proliferation defect, we characterized the transcriptional profile of H2A.Z.2-deficient cells. We used Affymetrix microarrays for SK-mel147 and WM266-4 cells depleted of either H2A.Z.1 or H2A.Z.2 (Figure 3A, Table S1). Interestingly, the majority of genes were downregulated (Figure 3A, B), with only 35 overlapping genes between H2A.Z.1 and H2A.Z.2 knock down in SK-mel147 cells (Figure 3B). Similar expression data was observed for WM266-4 cells (Figure S3A, B, Table S1).
Consistent with the observed phenotype, functional annotation revealed that H2A.Z.2-regulated genes are enriched for cell cycle regulators (Figure 3C, D, Figure S3C). This is in contrast to H2A.Z.1-regulated genes, which are enriched for immunological pathways (Figure S3D). This is in line with the lack of cell cycle defects observed upon H2A.Z.1 knock down, and implicates a distinct role for H2A.Z.1 in melanoma.
Gene Set Enrichment Analysis (GSEA) and transcription factor (TF) analysis further demonstrated that the H2A.Z.2-regulated genes are associated with transcriptional hallmarks of advanced melanoma and are targets of the E2F family, including E2F1 and E2F4 (Figure 3E, F, S3E). Furthermore, qRT-PCR analysis revealed that E2F target gene expression correlates with H2A.Z.2 levels in human melanoma (Figure 3G). Given that E2F1 and E2F4 promote melanoma progression and metastasis (Alla et al., 2009; Ma et al., 2008), these results implicate concerted H2A.Z.2-E2F function in melanoma progression.
We next performed native chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) of H2A.Z to determine its genomic occupancy in melanoma cells (Figure 4). ChIP-seq of SK-mel147 cells stably expressing N-terminally eGFP-tagged H2A.Z.1 or H2A.Z.2 (along with eGFP-H2A as a control) was also carried out (Figure S4A), and exhibited highly overlapping genome-wide occupancy patterns with endogenous H2A.Z (Figure S4B, C) as well as with each other (Figure 4SD). Therefore, we utilized endogenous H2A.Z ChIP-seq for further analyses.
Amongst H2A.Z-bound sites, 14% lie within promoters and 29% in gene bodies (Figure 4A). By integrating RNA-sequencing and ChIP analyses in SK-mel147 cells, we found that H2A.Z promoter levels positively correlate with expression (Figure 4B), as previously reported (Barski et al., 2007; Hu et al., 2012). Intriguingly, gene body occupancy shows a striking negative correlation with gene expression (Figure 4B). Finally, in line with previous reports (Barski et al., 2007; Hu et al., 2012; Obri et al., 2014), our results show that half of the H2A.Z-bound sites lie within intergenic regions (Figure 4A).
Next we integrated H2A.Z ChIP-seq with H2A.Z.2 downregulated genes. Taking into account both promoter and gene body occupancy, we defined four classes of genes: H2A.Z-bound and H2A.Z.2 downregulated (Class I), H2A.Z-bound but not H2A.Z.2 downregulated (Class II), H2A.Z.2 downregulated but not H2A.Z-bound (Class III), and neither downregulated nor bound (Class IV) (Figure 4C, Table S2). Gene ontology analyses revealed that Class I is enriched for cell cycle genes, while Class II is enriched for metabolic processes (Figure S4E). Accordingly, ChIP Enrichment Analysis (ChEA2) (Kou et al, 2013) uncovered distinct TF binding profiles for each class, with only Class I showing enrichment for E2Fs (Figure 4D). By examining H2A.Z distribution, we found that Class I genes were significanly enriched at the promoter, and depleted within the gene body (Figure 4E, F). Conversely, many Class II genes lie within broader H2A.Z domains (Figure 4F). Expression of Class I genes is significantly higher than Class II genes in melanoma (Figure 4G), consistent with our findings in Figure 4B.
H2A.Z ChIP-seq in normal human melanocytes revealed that the Class I signature is not detectable, as H2A.Z is not depleted in the gene body (Figure 4H). Consistent with this, Class I genes are expressed at significant lower levels in melanocytes than melanoma cells, while expression of all other classes remains largely unchanged (Figure 4G).
Collectively, our analyses revealed that H2A.Z.2 regulated/H2A.Z bound genes (Class I) have unique features in melanoma cells. They are E2F targets, highly expressed, and enriched for H2A.Z at the promoter and depleted in the gene body. These features do not apply to H2A.Z.1-downregulated genes (Figure S4F, G, Table S3), suggesting a unique chromatin signature at genes that regulate melanoma cell proliferation.
To further decipher H2A.Z function, we investigated the factors that interact with H2A.Z-containing nucleosomes in melanoma cells. To this aim, we utilized unbiased label-free quantitative mass spectrometry (MS) (Eberl et al., 2013). Chromatin isolated from SK-mel147 cells stably expressing eGFP, eGFP-H2A, eGFP-H2A.Z.1 or eGFP-H2A.Z.2 was digested to mononucleosomes (Figure S5A), immunoprecipitated (Figure S5B) and LC-MS/MS analyzed. We found significant enrichment of ~45 H2A.Z interactors as compared to H2A-containing nucleosomes (Figure 5A, S5C). The majority of these proteins were found in both H2A.Z variant IPs, including members of the H2A.Z histone chaperone complex, SRCAP (Billon and Cote, 2012).
We identified BRD2 to be enriched in H2A.Z.1- and H2A.Z.2-containing nucleosomes (Figure 5A, B). BET proteins (BRD2, BRD3, BRD4 and BRDT) bind to acetylated lysine residues in histones (LeRoy et al., 2008; LeRoy et al., 2012), and function as scaffolds to recruit chromatin modifying enzymes and TFs, thereby coupling histone acetylation to transcription (reviewed in (Belkina and Denis, 2012)). While BRD2 and BRD4 are both overexpressed in melanoma (Segura et al., 2013), only BRD2 interacts with H2A.Z-containing nucleosomes (Figure 5A and data not shown). We next tested whether hyperacetylation of histones would enhance the interaction between BRD2 and H2A.Z isoforms in melanoma cells. Treatment with the HDAC inhibitor trichostatin A (TSA) resulted in increased histone H4 and H2A.Z acetylation and increased BRD2 chromatin association (Figure 5C). Furthermore, the BRD2-H2A.Z interaction was enhanced (Figure 5C). By probing primary and metastatic melanoma cell lines (as in Figure 1A), we observed hyperacetylation of H4 and H2A.Z, and high levels of BRD2 in metastatic cells (Figure 5D). Collectively, these results are consistent with the fact that BRD2 is recruited to chromatin by a combination of acetylated H4 (H4ac) and H2A.Z (Draker et al., 2012).
Through immunohistochemistry (IHC) of BRD2 in a cohort of patient samples including benign nevi, thick primary melanoma and metastatic melanoma, we detected a significant increase of BRD2 in primary and metastatic melanoma specimens as compared to dermal melanocytes of nevi (Figure 5E). Next, we investigated BRD2 knock down in multiple melanoma cell lines and observed proliferation defects via G1/S arrest (Figure 5F, G, S5D–F). BRD2 knock down altered gene expression of selected E2F targets (Figure 5H, S5G). Collectively, BRD2 knock down recapitulated the phenotype observed upon H2A.Z.2 loss, suggesting that H2A.Z.2 and BRD2 work together to promote Class I transcription.
To test this hypothesis, we performed ChIP-seq of BRD2 in SK-mel147 cells, and found a similar genomic distribution as H2A.Z (Figure 4A, S6A). We next determined the extent of co-localization of BRD2 and H2A.Z genome-wide and found that BRD2 peaks overlap with H2A.Z largely at promoters (Figure 6A). Promoters of Classes I and II genes are bound by BRD2 (Figure 6B, S6B), consistent with the fact that BRD2 interacts with both H2A.Z isoforms. However, H2A.Z and BRD2 have the highest enrichment at Class I genes (Figure 4E, ,6C6C).
Because our in silico analyses predicted that Class I genes are E2F targets (Figure 4D), and BRD2 interacts with E2F1 to mediate its recruitment to chromatin (Denis et al., 2006; Sinha et al., 2005), we next queried whether Class I genes are bound by E2F1. ChIP-seq of E2F1 in SK-mel147 cells showed this was indeed the case (Figure 6B, C, S6B). Taken together, these data suggest that H2A.Z.2 works cooperatively with BRD2 and E2F1 (Figure 6D) to promote high levels of transcription at Class I genes in melanoma. Next, we probed a panel of benign nevi and melanoma tissues for H2A.Z, BRD2, E2F1 and H4ac, and found evidence of the H2A.Z-BRD2-E2F axis in melanoma specimens (Figure 6E). This reinforces the relevance of our findings for melanoma disease.
The findings above prompted us to investigate BRD2 and E2F1 levels upon H2A.Z silencing. We observed marked reduction of BRD2 and E2F1 levels upon H2A.Z.2, but not H2A.Z.1, knock down across melanoma cell lines (Figure 6F, S6C). BET family members are not transcriptionally regulated by H2A.Z.2 (Figure S6D), suggesting that H2A.Z.2 stabilizes these factors. This was paralleled by a dramatic loss of H4 and H3 acetylation (Figure 6F and S6C, E). These data suggest that BRD2 and E2F1 chromatin recruitment to Class I genes, mediated by histone acetylation, is impaired in H2A.Z.2-deficient cells. Thus, we performed ChIP-qPCR for BRD2 and E2F1 in either control or H2A.Z isoform-depleted cells, and found that BRD2 and E2F1 recruitment to Class I promoters is dependent on H2A.Z.2 (Figure 6G). Overall, H2A.Z.2 deficiency results in dramatic alterations in chromatin structure, thereby clearly distinguishing it from H2A.Z.1. Overall, our ChIP studies demonstrate that H2A.Z.2, BRD2 and E2F1 work cooperatively to promote high levels of transcription at cell cycle promoting genes in melanoma.
Because we observed loss of histone acetylation upon H2A.Z.2 knock down, we queried whether H2A.Z.2 depletion might potentiate the effects of BET inhibitors (BETi). BETi prevent the acetyllysine binding of bromodomains with high affinity, and are effective agents in a number of tumors (Dawson et al., 2012; Segura et al., 2013). We first assessed the sensitivity of melanoma cells to JQ1 (Filippakopoulos et al., 2010) (Figure 7A), and found a dose-dependent growth inhibitiory effect in the majority of cell lines tested (Figure S7A–C). Cells treated with JQ1 for 4 days accumulated in G2/M (Figure S7B), with minor induction of apoptosis (data not shown).
Next we investigated whether H2A.Z.2 knock down cooperates with JQ1 to enhance the antiproliferative effect of melanoma cells. Whereas JQ1 treatment or H2A.Z.2 knock down alone induced growth arrest (Figure S7B, 2D–G), the combination resulted in cell death, in both BRAF and NRAS mutant lines (Figure 7B, S7D). Functional annotation of the SK-mel147 transcriptome upon JQ1 treatment (Table S4) revealed enrichment in developmental processes, distinguishing it from cell cycle annotation associated with H2A.Z.2 silencing (see Figure 3C, S7E). Consistent with this synergy, we observed that the transcriptional profiles of H2A.Z.2 knock down and BETi show minimal overlap (Figure 7C, Table S4), suggesting that while H2A.Z.2-regulated genes are BRD2 and E2F1 targets, the mode of action of JQ1 is largely distinct from H2A.Z.2. While H2A.Z and BRD2 are enriched on promoters of JQ1 regulated genes (although significantly less so than on H2A.Z.2-regulated genes), E2F1 binding is absent (Figure 7D). JQ1 regulated genes are instead targets of distinct TFs (Figure S7F).
H2A.Z.2 loss not only enhances sensitivity of melanoma cells to BETi, but also to chemo- and targeted therapies (MEKi) used clinically for melanoma (Figure 7E). Collectively, these data suggest that H2A.Z.2 is a critical mediator of melanoma drug sensitivity and regulating its deposition may serve as an important target for novel therapeutic strategies.
While histone variants and their chaperones have emerged as critical players in cancer biology (Vardabasso et al., 2013), our mechanistic understanding remains limited. Here we report a unique role for H2A.Z.2 in driving melanoma cell proliferation and drug sensitivity. To our knowledge this is the first study to identify a specific role for H2A.Z.2 in any tumor type. This may suggest a melanoma-specific role for H2A.Z.2 in driving proliferation, and it will be of interest to learn if H2A.Z.2 plays a similar role in other tumors. Importantly, we do not exclude a role for H2A.Z.1 in melanoma, as it is also upregulated and correlates with shorter patient survival.
Because H2A.Z isoforms have distinct roles in melanoma, we hypothesized they may have unique interaction partners and genomic occupancy. Our studies indicate that H2A.Z.1 and H2A.Z.2 share genomic occupancy patterns and interact with similar histone chaperone complexes. However, it is clear that H2A.Z.2 is critical for promoting cell cycle progression in melanoma, and acts distinctly from H2A.Z.1. Our data strongly suggest that a unique property of H2A.Z.2 is to promote and/or maintain BRD2, E2F1 and histone acetylation levels. While the exact mechanism remains unclear, H2A.Z.2 likely acts together with histone acetylation to recruit co-activators and TFs, such as BRD2 and E2F1 respectively, to promote expression of cell cycle regulators (see below).
Our analyses revealed that H2A.Z.2 promotes the expression of E2F target genes. In melanoma cells, these genes are characterized by a unique signature of H2A.Z occupancy - highly enriched at the TSS and depleted within the gene body - and this pattern associates with high gene expression levels. Our findings are in line with previous observations in plants and yeast (Coleman-Derr and Zilberman, 2012; Sadeghi et al., 2011; Zilberman et al., 2008); for example, H2A.Z is excluded from the bodies of actively transcribed genes in Arabidopsis (Coleman-Derr and Zilberman, 2012; Zilberman et al., 2008). Intriguingly, the DREAM complex was recently reported to promote H2A.Z gene body incorporation to repress cell cycle progression genes in C. elegans (Latorre et al., 2015). Together, these studies suggest that H2A.Z is differentially distributed across promoters and gene bodies at distinct subsets of genes to regulate their expression levels.
Our study has identified BRD2 as an H2A.Z-interacting protein in malignant melanoma. Work by Denis and colleagues initially demonstrated that BRD2 has oncogenic potential: BRD2 transforms mouse fibroblasts in the context of oncogenic Ras (Denis et al., 2000), and Eμ-BRD2 transgenic mice develop B-cell lymphoma and leukemia (Greenwald et al., 2004). In fact, BRD2 has a crucial role in cell cycle control, and by interaction with E2F1, it regulates the expression of cyclins and other cell cycle regulatory genes (Denis et al., 2000; Denis et al, 2006; Sinha et al., 2005).
Our loss-of-function approach revealed that the chromatin association and total levels of BRD2, E2F1 and histone acetylation are H2A.Z.2- dependent. This is in line with the fact that BRD2's preference for H2A.Z-containing nucleosomes is mediated by a combination of hyperacetylated H4, and features on H2A.Z itself (Draker et al., 2012), and that histone acetyltransferase (HAT) activity is contained within BRD2 nuclear complexes (Sinha et al., 2005). Furthermore, we find evidence of an H2A.Z-BRD2-E2F axis in melanoma tissues. Accordingly, our ChIP analyses show that BRD2 and E2F1 are enriched at promoters of Class I genes and that H2A.Z.2 is required for recruitment of these factors to these E2F targets. Interestingly, Draker at al., found that recruitment of BRD2 to androgen receptor (AR)-regulated genes in prostate cancer cells is dependent on H2A.Z.1. Thus, BRD2 may associate with distinct TFs and H2A.Z isoforms to achieve oncogenic gene transcription in different tumor types.
Collectively, we envision that H2A.Z.2 recruits BRD2 and E2Fs, along with HAT activity, to E2F target genes in melanoma cells. This in turn results in increased expression of cell cycle genes, and ultimately promotes proliferation (Figure 7F). Our findings implicate the H2A.Z.2-BRD2-E2F1 axis as a driver of melanoma progression. Of these molecules, BRD2 represents a key target for therapy.
Metastatic melanoma is notoriously refractory to conventional cancer therapies, and remains largely resistant to current targeted therapies (Lito et al., 2013). Here we show that in combination with BET inhibition, H2A.Z.2 depletion is effective in inducing cell death. Because a tool to disrupt H2A.Z deposition is currently lacking, it is plausible that combining BETi with a potent inhibitor of HAT activity will potentiate melanoma cell death (Figure 7F). This combination may not only evict BET proteins from chromatin, but cause additional destabilization of BET proteins and their associated TFs due to loss of acetylation (Figure 7F). It will be of interest to create BRD-specific inhibitors, if achievable, as our study suggests that BRDs function distinctly in disease. Finally, our findings implicate H2A.Z.2 as a mediator of cell proliferation and drug sensitivity in malignant melanoma. Owing to the fact that histone modification and deposition are reversible processes, our study holds therapeutic potential for this highly intractable neoplasm.
Primary (WM115, WM1789, WM39), metastatic (SK-mel147, WM266-4, 501mel, A375, SK-mel2, SK-mel28, SK-mel239, SK-mel5, M14, WM165-1) melanoma cell lines, HeLa cells, and human melanocytes were cultured as described in Extended Experimental Procedures. Lentiviral vectors and shRNAs used for the generation of stable cell lines are described in Extended Experimental Procedures. Infections were performed according to standard procedures (Kapoor et al., 2010).
Chromatin fractionation and acid extraction of histones were performed as described (Kapoor et al., 2010) and in Extended Experimental Procedures. Antibodies used in this study are listed in Extended Experimental Procedures.
Approval to collect melanoma specimens was granted by Mount Sinai Biorepository Cooperative and the New York University Interdisciplinary Melanoma Cooperative Group (project number HSD08-00565 and IRB number 10362, respectively). Approval to collect benign nevi was granted by ISMMS (Icahn School of Medicine at Mount Sinai) Division of Dermatopathology (project number 08-0964).
For RNA extraction, qRT-PCR, and primers, see Extended Experimental Procedures. RNA amplification, labeling and hybridization to Human Gene 1.0 ST Arrays (Affymetrix) were performed as described previously (Wiedemann et al., 2010), and data processed in R/bioconductor (www.bioconductor.org). For data analysis, see Extended Experimental Procedures.
For proliferation curves, cells were counted up to 7 days and normalized to cell counts at day 1. Colony formation assay was performed by seeding cells at low density and allowing growth for 2 weeks. Cells were washed with phosphate buffered saline, fixed in 10% methanol/acetic acid solution and stained with 1% crystal violet. Flow cytometry experiments were performed as described in Extended Experimental Procedures.
Chromatin from SK-mel147 cells was digested with micrococcal nuclease (MNase) and used for ChIP with H2A.Z (Abcam ab4174) and GFP Trap Beads (Chromotek), essentially as described (Hasson et al., 2013). SK-mel147 cells stably expressing control or isoform-specific shRNAs were cross-linked for 10’ with 1% formaldehyde and immunoprecipitated with BRD2 and E2F1 antibodies (Bethyl Laboratories A302-583A and Santa Cruz sc-193, respectively) as described in Extended Experimental Procedures. Sequencing libraries were generated and barcoded for multiplexing as described (Hasson et al., 2013) and libraries were submitted for 100-bp, single-end Illumina sequencing on a HiSeq 2500. For data processing and analysis, see Extended Experimental Procedures.
Total RNA samples were isolated from human melanocytes and enantiomer- or JQ1-treated SK-mel147 using miRNeasy mini kit (Qiagen) following manufacturer's protocol. Sequencing libraries were prepared and data analysis performed as described in Extended Experimental Procedures.
Mononucleosomes were generated according to (Sansoni et al., 2014) and described in Extended Experimental Procedures.
See Extended Experimental Procedures for details.
Statistical tests were applied as indicated in figure legends. Asterisks as follow: *p<0.05; ** p<0.01; ***p<0.001. Boxplots represent Tukey boxplots with outliers omitted.
The authors are grateful for assistance and reagents provided by Luis F. Duarte, Avnish Kapoor, Nicholas Mills, Clemens Bönisch, Pauline Rimmele, Danielle Martinez (Baylor), and the laboratories of Robert Fisher, Stuart Aaronson, Ramon Parsons and Jay Bradner. We thank Cristina Montagna and Jidong Shan at Molecular Cytogenetic Core at Albert Einstein College of Medicine, Genomics core facility at Mount Sinai, Avi Ma’ayan for statistical support, Robert Phelps, Madelaine Haddican, Shelbi Jim On, and Giselle Singer for assistance with benign nevi collection, and Mark Lebwohl for his support. Funding was supported by a Melanoma Research Development Award (Mount Sinai) to C.V., DOD BCRP Postdoctoral fellowship (W81XWH-11-1-0018) and NYSCF Druckenmiller Fellowship to A.G.-M., Graduate fellowship from CONACyT (239663) to D.V-G, International Max Planck Research School for Life Science (IMPRS-LS) to S.P., Bundesministerium für Bildung und Forschung (FKZ01GS0861, DiGtoP consortium) to M.M., DOD Collaborative Award CA093471 (W81XWH-10-1-080), 1R01CA155234, and 1R01CA163891 to E.H., Deutsche Forschungsgemeinschaft (DFG) through the Collaborative Research Center SFB 1064 (project A10 to S.B.H. and project Z04 to T.S.), HA 5437/4-1 and the Center for Integrated Protein Science Munich (CIPSM) to S.B.H, and Worldwide Cancer Research, Hirschl/Weill-CaulierResearch Award and NCI/NIH R01CA154683 to E.B. We dedicate this study in memory of Estela Medrano.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
All microarray, RNA-seq and ChIP-seq data sets have been deposited to NCBI’s Gene Expression Omnibus (GSE59060).
AUTHOR CONTRIBUTIONSC.V. and E.B. conceived this study with guidance from S.B.H. C.V. performed all qPCR, shRNA studies and generated eGFP melanoma cell lines. C.V. and T.P. performed immunoblots and cell cycle analyses. S.B.H. and T. Straub performed and analyzed microarrays, respectively. C.V., D.H. and S.P. performed ChIP-seq experiments and data analyses was lead by C.V., A.G.-M., D.V.-G, D.H. and T. Straub. S.P. performed IPs for MS, and E.C.K. performed MS with the support of M.M. C.V. and T. Strub performed drug treatments and S.P., C.V. and T. Strub performed BRD2 immunoblots. J.D. performed BRD2 IHC and J.L.Y. and R.S. scored tissues. C.-Y.C. assisted with network analyses and A.V. provided copy number analysis support. M.F.S., B.F.-C. and E.H. provided assistance with patient data and with BETi experiments. B.F.-C. prepared samples for RNA-seq, and D.H. analyzed RNA-seq data. C.V., A.G.-M., D.V.-G., D.H., S.P, E.C.K., S.B.H. and E.B. designed experiments and interpreted results. C.V., E.B. and S.B.H. wrote the manuscript with contributions from all other coauthors.