Neuroendocrine prostate cancer (NEPC) is a lethal subtype of prostate cancer arising mostly from adenocarcinoma via neuroendocrine transdifferentiation following androgen deprivation therapy. Mechanisms contributing to both NEPC development and its aggressiveness remain elusive. In light of the fact that hyperchromatic nuclei are a distinguishing histopathologic feature of NEPC, we utilized transcriptomic analyses of our patient-derived xenograft (PDX) models, multiple clinical cohorts, and genetically engineered mouse models to identify 36 heterochromatin-related genes that are significantly enriched in NEPC. Longitudinal analysis using our unique, first-in-field PDX model of adenocarcinoma-to-NEPC transdifferentiation revealed that, among those 36 heterochromatin-related genes, heterochromatin protein 1α (HP1α) expression increased early and steadily during NEPC development and remained elevated in the developed NEPC tumor. Its elevated expression was further confirmed in multiple PDX and clinical NEPC samples. HP1α knockdown in the NCI-H660 NEPC cell line inhibited proliferation, ablated colony formation, and induced apoptotic cell death, ultimately leading to tumor growth arrest. Its ectopic expression significantly promoted NE transdifferentiation in adenocarcinoma cells subjected to androgen deprivation treatment. Mechanistically, HP1α reduced expression of androgen receptor and RE1 silencing transcription factor and enriched the repressive trimethylated histone H3 at Lys9 mark on their respective gene promoters. These observations indicate a novel mechanism underlying NEPC development mediated by abnormally expressed heterochromatin genes, with HP1α as an early functional mediator and a potential therapeutic target for NEPC prevention and management.

Significance: Heterochromatin proteins play a fundamental role in NEPC, illuminating new therapeutic targets for this aggressive disease. Cancer Res; 78(10); 2691–704. ©2018 AACR.

Neuroendocrine prostate cancer (NEPC) has become a clinical challenge in the management of castration-resistant prostate cancer (CRPC). Although de novo cases are rare, NEPC as a special subtype of CRPC (∼10%–20%) is thought to occur via NE transdifferentiation of prostate adenocarcinomas in response to androgen deprivation treatment (ADT), resisting dependence on AR signaling as an adaptive response. As next-generation AR pathway inhibitors (ARPI) such as enzalutamide (EnZ) and abiraterone have made substantial improvements in managing CRPC adenocarcinomas in recent years (1), it is expected that the incidence of NEPC will further increase (2). Unfortunately, the overall median survival of NEPC with small cell feature is less than 1 year, primarily due to its aggressiveness and limited available treatment options (3). As such, a better understanding of the mechanisms underlying NEPC development remains much needed in order to develop more effective therapeutics.

One distinct histopathologic feature of NEPC cells is the frequent manifestation of hyperchromatic nuclei with finely dispersed chromatin and inconspicuous nucleoli, a phenomenon known as “salt and pepper” chromatin (3, 4). This is in contrast to prostatic adenocarcinoma cells, which tend to have enlarged nuclei with prominent nucleoli (3, 4). This special hematoxylin-staining characteristic suggests a distinct NEPC heterochromatin pattern (5). Heterochromatin is a condensed and transcriptionally inert chromosome conformation, regulated epigenetically by precise and dedicated machineries (6, 7). Multiple studies have demonstrated that epigenetic regulation is one major mechanism underlying NEPC development (8–11). An NEPC-specific heterochromatin gene signature can thus shed light to help better understand the disease and identify novel therapeutic targets. Another distinct feature of NEPC is the lineage alteration associated with a decrease or loss of crucial adenocarcinoma lineage–specific transcription factors, AR, FOXA1, and RE1 silencing transcription factor (REST; refs. 12–15). This leads to the repression of AR-regulated genes (e.g., prostate-specific antigen) and gain of neuroendocrine markers [e.g., neural cell adhesion molecule 1 (NCAM1/CD56), neuronal-specific enolase (NSE), chromogranin A (CHGA), and synaptophysin (SYP); refs. 12–14]. Although recent studies have reported that EZH2 inhibits AR expression and SRRM4 mediates REST splicing (8, 16), the mechanisms underlying the constant transcriptional suppression of AR and REST are still not well understood.

One of the major hurdles in studying NEPC is the lack of clinically relevant models, but substantial progress has been achieved recently in modeling NEPC development. Employing genetically engineered mouse (GEM) models, ectopic gain of N-myc and concomitant loss of Rb1 and Tp53 have both been demonstrated to induce de novo NEPC-like tumors (8, 11). Potent ADT using abiraterone in Tp53/Pten double-deficient mice also promoted overt neuroendocrine (NE) transdifferentiation of luminal adenocarcinoma cells (17). Employing engineered human primary cells or cell lines, N-myc, SOX2, BRN2, and SRRM4 overexpressions have all been shown to promote NE transdifferentiation (16, 18–20). Our laboratory has established over 45 high-fidelity patient-derived xenograft (PDX) models of prostate cancer including 5 NEPCs (www.livingtumorlab.ca; ref. 21). Among them, LTL331/331R is the first-in-field and unique PDX model of adenocarcinoma-to-NEPC transdifferentiation. Upon host castration, the primary adenocarcinoma (LTL331) initially regresses but relapses within few months as typical NEPC (LTL331R; ref. 21). Importantly, the whole transdifferentiation process observed in the LTL331/331R model is predictive of disease progression and is fully recapitulated in the donor patient (22), suggesting a strong clinical relevance. Because an overwhelming majority of expressed genes accounting for the NE phenotype in terminal NEPC may obscure the real drivers of disease progression required at earlier stages, we focused on identifying the early changes induced by ADT before NEPC is fully developed, with an additional secondary criterion that these changes persist into terminal NEPC. To this end, we developed a time series of the LTL331 model, for which samples were harvested at multiple points during the transition period following host castration in order to monitor the entire transdifferentiation process and discover potential drivers with early expression changes (22).

In this study, we identified a heterochromatin molecular signature that is commonly upregulated in NEPC. Among the signature genes, we discovered that the upregulation of heterochromatin protein 1α (HP1α, encoded by CBX5), a gene prominently associated with constitutive heterochromatin and mediating concomitant gene silencing, was an early event in our LTL331/331R NE transdifferentiation model. HP1α expression was increased within weeks following castration but before emergence of NE genes, gradually reaching its highest level in terminally developed NEPC. HP1α is also widely expressed in clinical samples of NEPC. We show that HP1α is essential for NEPC cell proliferation, survival, and tumor growth, and its elevated expression promotes ADT-driven NE-differentiation in prostatic adenocarcinoma cells. HP1α ectopic expression reduces expression of AR and REST, two crucial transcription factors silenced in NEPC, and enriches the repressive histone mark H3K9me3 on their respective gene promoters.

PDXs and clinical datasets

All PDX tumor lines were grafted in NSG mice as previously described (21). This study followed the ethical guidelines stated in the Declaration of Helsinki, specimens were obtained from patients with their informed written consent form following a protocol (#H09-01628) approved by the Institutional Review Board of the University of British Columbia (UBC). Animal studies under the protocol #A17-0165 were approved by UBC's Animal Care and Use Committee. The LTL331-castrated tissues and the NEPC-relapsed LTL331R tissues were harvested at different time points after host castration (22). Transcriptomic analysis for all PDX tumors, with the exception of the LTL331-331R castration time-series samples, was performed using GE 8 × 60K microarray, as previously described (21). Transcriptomic analysis of the LTL331-331R time-series was done using RNA-sequencing data.

The clinical cohorts used in this study are as follows: RNA-seq data for the Beltran 2011 cohort (30 adenocarcinomas, 6 NEPCs) were from Weill Medical College of Cornell University (23); RNA-Seq data for the Beltran 2016 (34 CRPC-adenocarcinomas, 15 NEPCs), the Grasso 2012 (31 CRPC-adenocarcinomas, 4 NEPCs), and the Kumar 2016 (149 CRPC-adenocarcinomas) cohorts were accessed through the cBioPortal (10, 24–26); microarray data for the Varambally 2015 cohort (6 CRPC-adenocarcinomas) were accessed from the Gene Expression Omnibus (GEO) database (GSE3325; ref. 27).

Human prostate cancer specimens

Prostate cancer specimens were obtained from the Vancouver Prostate Centre Tissue Bank (103 hormone-naïve primary prostatic adenocarcinomas, 120 CRPC-adenocarcinomas, and 8 NEPCs). This study followed the ethical guidelines stated in the Declaration of Helsinki, specimens were obtained from patients with their informed written consent form following a protocol (#H09-01628) approved by the Institutional Review Board of the UBC. Tissue microarrays (TMA) of duplicate 1 mm cores were constructed manually (Beecher Instruments). NEPC specimens are histologically either small cell carcinoma or large cell neuroendocrine carcinoma with low or negative AR expression and positive CHGA expression as determined by IHC staining.

Cell line culture and reagents

NCI-H660 and 293T cells were obtained from the ATCC. Cells were authenticated with the fingerprinting method at Fred Hutchinson Cancer Research Centre. Mycoplasma testing was routinely performed at the Vancouver Prostate Centre (VPC). NCI-H660 cells were cultured in RPMI-1640 medium (Hyclone) with supplements as follows: 5% FBS (GIBCO), 10 nmol/L β-estradiol (Sigma), 10 nmol/L Hydrocortisone (Sigma), and 1% Insulin-Transferrin-Selenium (Thermo Fisher). V16D (20) cells were maintained in RPMI-1640 containing 10% FBS. 293T was kept in DMEM (Hyclone) with 5% FBS. For in vitro NE phenotype induction in V16D cells, cells were starved with phenol-red–free RPMI-1640 (GIBCO) containing 10% charcoal-stripped serum (CSS; GIBCO) for 24 hours, and then cultured in the same media or with the addition of 10 μmol/L EnZ (Haoyuan Chemexpress) for another 14 days (20).

Bioinformatics analysis

As previously described, gene set enrichment analysis (GSEA; http://software.broadinstitute.org/gsea/index.jsp) was used in this study to determine whether a defined set of genes shows significant, concordant differences between two biological phenotypes (e.g., prostate cancer adenocarcinoma vs. NEPC) or two sample groups (e.g., shC vs. KD; ref. 28). All GSEA analyses in this study used whole transcriptomic data without expression level cut-off as expression datasets. Normalized values were used for profiling data from PDX tissues, H660 and V16D stable cell lines, and the three NEPC mouse models (GSE90891, GSE92721, and GSE86532); raw read numbers were used for RNA-seq data from the Beltran 2011 cohort; preranked gene list based on P value from lowest to highest was used for the Beltran 2016 cohort (8, 10, 11, 17, 21, 23). Unbiased analysis was performed using the latest MSigDB database for each collection (29). Phenotype permutation was applied when the sample size grouped in one phenotype was more than five. Otherwise, gene set permutation was employed. False discovery rate (FDR) q values were calculated using 1,000 permutations, and a geneset was considered significantly enriched if its normalized enrichment score (NES) has an FDR q below 0.25.

Ingenuity pathway analysis (IPA) was performed as previously described (20). Differentially expressed genes were selected based on the criteria that their standard deviation between samples of one group (e.g., three HP1α knockdown lines) is below 0.5, and their Student t test P value between groups (e.g., shC vs. KD) is below 0.05.

For the NE scores and the heterochromatin score, the weight for each gene in the score list was first calculated by taking the log2 of the P value of all genes and multiplying it by the sign of the fold change. The weight for each gene in the study by Lee and colleagues was obtained directly from their study (18). The P values for the genes contributing to heterochromatin score were calculated by comparing NEPC PDXs with all other adenocarcinoma PDXs with the Student t test. A score was then assigned to each sample by multiplying the weight with the normalized RNA expression value, adding all values for each gene, log-transforming the absolute value of the sum value, and multiplying the sign of the sum value (10, 11, 18).

The heatmap was constructed using the Gplots and heatmap.2 packages in R. Hierarchal clustering was used to determine sample similarity.

Statistical analysis and data representation

Statistical analysis was done using the Graphpad Prism software. The Student t test was used to analyze statistical significance between groups in discrete measurements, whereas two-way ANOVA was used for continuous measurements. The Kaplan–Meier method was used to estimate curves for relapse-free survival, and comparisons were made using the log-rank test. Pearson correlation [with 95% confidence intervals (CI)] was used to measure the linear correlation between two variables. Any difference with P values lower than 0.05 is regarded as statistically significant, with *, P <0.05; **, P < 0.01; and ***, P < 0.001. Graphs show pooled data with error bars representing SEM obtained from at least three replicates and SD of values from clinical samples. Lines in scatter plots represent the median value of measurements from multiple samples and mean with 95% CI for scores from multiple samples. Box plot with whiskers showing 5% to 95% percentile values was used to represent analysis of IHC scores.

Accession numbers

All PDX microarray profiles are available at www.livingtumorlab.ca and accessible under the accession number GSE41193 in the GEO database. RNA-seq profiles for the LTL331/331R castration time-series have been deposited to the European Nucleotide Archive and are available under the accession number ENA: PRJEB9660. Microarray profiles for cell lines are deposited (GSE105033).

NEPC has a distinctive heterochromatin gene signature

To determine if a specific gene expression program contributes to the hyperchromatic histologic feature of NEPC cells, we first performed an unbiased GSEA using our latest PDX collection and two NEPC clinical cohorts. Transcriptomic profiles from our high-fidelity PDX models (18 adenocarcinomas vs. 4 NEPCs), the Beltran 2011 cohort (30 adenocarcinomas vs. 6 NEPCs), and the Beltran 2016 cohort (34 CRPC-adenocarcinomas vs. 15 NEPCs; refs. 10, 21, 23) all demonstrate that heterochromatin-associated genes are significantly enriched in NEPC compared with adenocarcinoma (Fig. 1A–C; Supplementary Fig. S1A). We further analyzed three GEM models mimicking the development of NEPC. Interestingly, heterochromatin-associated genes are also significantly enriched as long as NE differentiation occurs, driven by either Pten/Rb double knockout (DKO) or Pten/Rb/Tp53 triple knockout (TKO), Pten/Tp53 knockout with potent ADT, or N-Myc overexpression (Fig. 1D–F; refs. 8, 11, 17).

To further decipher the genes contributing to the heterochromatin feature of NEPC, we analyzed the leading edge genes derived from the GSEA of PDX and clinical NEPC samples (Fig. 1A–C). Because the NEPCs in our PDX collection are all typical neuroendocrine carcinomas without mixed adenocarcinoma tissues, the upregulated genes identified in the NEPC PDXs and validated in the clinical cohorts can be considered heterochromatin signature genes. In addition, three polycomb-group (PcG) genes, EZH2, RBBP4, and RBBP7, are also included in this gene signature as their upregulated expressions in NEPC have been previously reported (9, 23). They have also been reported to play important roles in heterochromatin formation (7), but are omitted from the GSEA geneset (GO:0000792) containing other PcG members. As such, 36 genes were identified and selected to form a heterochromatin gene panel, which could successfully distinguish NEPC from adenocarcinoma through hierarchical clustering similar to two reported NE score gene panels (Fig. 1G; Supplementary Fig. S1B–S1E; refs. 10, 18). From this heterochromatin gene panel, a weighted score was calculated for each sample in the PDX and clinical cohorts following a similar strategy utilized in recent studies (10, 11, 18). Notably, similar to NEPC samples having a distinctly higher NE score compared with adenocarcinomas, NEPC samples also have a significantly higher heterochromatin score compared with adenocarcinoma (Fig. 1H). Importantly, the heterogeneity among PDX samples is much smaller than that observed in any clinical cohorts regardless of the scoring system employed, further demonstrating that our PDX collection both clearly reflects the clinical features of NEPC and adenocarcinoma and provides a distinct system to study prostate cancer subtypes. Together, these analyses indicate that a heterochromatin-related gene expression signature is a unique feature of NEPC tumors.

Expression of HP1α is upregulated early and steadily during NEPC development

Inspired by the critical functional role of heterochromatin structure in modulating cell behavior and gene expression (6, 7), we hypothesized that these NEPC-specific heterochromatin signature genes could contribute to NEPC development. As such, we attempted to identify a potential NEPC driver gene from our 36 heterochromatin signature genes. In addition to a final upregulation in terminal NEPC, expression of the candidate driver gene would also be increased early prior to full NEPC development. We thus analyzed the RNA-seq profile of the LTL331/331R NE transdifferentiation model with samples from castration-induced dormant time points prior to full NEPC relapse (21, 22). Ranked by the gene expression difference between pre- and post-castration samples, elevated expression of HP1α was found to be an early and dramatic event occurring upon host castration. Conversely, expressions of other genes that have been reported to be associated with NEPC such as EZH2 and CBX2 (9, 23) were only upregulated in the fully developed NEPC sample (LTL331R; Fig. 2A). Interestingly, we also noticed that Hp1α expression gradually increased as the NE phenotype progressed from partial to more dominantly overt in the Pten/Tp53 CRPC mouse model (Supplementary Fig. S2A), lending further support that elevated HP1α expression is an early event in NEPC development.

To validate our findings, we performed qRT-PCR to detect HP1α expression at the mRNA level and IHC staining and Western blotting to detect HP1α expression at the protein level in our PDX collection. Both the mRNA and protein levels of HP1α are significantly increased in NEPC PDXs (Fig. 2B–D; Supplementary Fig. S2B). Furthermore, we confirmed HP1α expression in the LTL331/331R castration time-series samples. Consistent with the RNA-seq data, both the mRNA and protein expression of HP1α increased upon host castration, with increased mRNA observed as early as 1 week after castration and elevated protein expression observed 3 weeks after castration (Fig. 2E–G). Taken together, these data demonstrate that elevated expression of HP1α is an early and consistent event throughout NEPC development and could be a potential driver of NEPC.

HP1α expression is upregulated in clinical NEPC samples and correlates with poor prognosis

To determine the clinical relevance of elevated HP1α expression in NEPC, we analyzed RNA expression profiles from three individual clinical cohorts containing NEPC samples. In the Beltran 2011 cohort, HP1α expression was about 4 times higher in NEPCs (n = 6) compared with adenocarcinomas (n = 30; Fig. 3A; ref. 23). In the Beltran 2016 cohort, HP1α expression was also significantly upregulated in NEPCs (n = 15) compared with CRPC adenocarcinoma samples (n = 34) by around 2 folds (Fig. 3B; ref. 10). In another cohort consisting of 35 metastatic CRPC samples, four samples with NEPC features expressed significantly higher HP1α mRNA than other typical adenocarcinomas (Fig. 3C; ref. 24). To better understand the relationship between HP1α expression and the NEPC phenotype at a protein level, we performed IHC for HP1α with the VPC clinical prostate cancer TMAs containing 103 primary adenocarcinomas, 120 CRPC adenocarcinomas, and 8 typical NEPC samples. Consistent with its mRNA level, HP1α protein was significantly overexpressed in NEPCs (mean = 2.45) compared with primary adenocarcinomas (mean = 1.66, P = 0.011) and CRPC adenocarcinomas (mean = 1.60, P = 0.014; Fig. 3D and E). These data together demonstrate that HP1α is highly expressed in NEPC.

Although HP1α is dramatically upregulated in NEPC, we also noticed that some adenocarcinomas also expressed higher levels of HP1α than others (Fig. 3D). We then investigated whether the expression of HP1α is associated with poor patient prognosis, thus possibly being a predisposing factor to poor outcome. Among the 103 primary adenocarcinoma samples in the VPC cohort, there were 37 samples with clinical follow-up information. The Kaplan–Meier analysis showed that patients with high expression of HP1α (IHC score ≥ 2) had a significantly shorter disease-free survival time (HR = 4.627, P = 0.0074) than low HP1α-expressing patients (IHC score < 2; Fig. 3F). For CRPC patients, the group with high expression of HP1α (top 1/3) had a significantly shorter overall survival time after the first hormonal therapy (HR = 2.961, P = 0.021), implying that HP1α positively correlates with poor prognosis upon hormonal therapy (Fig. 3G). Overall, these analyses indicate that increased expression of HP1α is a poor prognostic factor in advanced prostate cancer.

HP1α knockdown inhibits NEPC cell proliferation and induces apoptosis, leading to tumor growth arrest

Considering that HP1α is significantly upregulated in NEPC, we proceeded to investigate its function in NEPC cells. NCI-H660 is a unique and typical NEPC cell line (30). Consistent with the high HP1α expression observed in clinical NEPC samples, HP1α is also expressed at the highest level in H660 cells compared with seven other prostate cancer cell lines at both RNA and protein levels (Supplementary Fig. S3A). We thus constructed stable H660 cell lines with HP1α knocked down using lentiviral-delivered shRNAs (Fig. 4A). Upon HP1α knockdown, the overall proliferation of H660 cells was significantly inhibited, as evaluated by both crystal violet staining and cell counting (Fig. 4B; Supplementary Fig. S3B). We further performed colony formation assays to evaluate the reproductive and survival ability of single cells. Remarkably, HP1α knockdown was able to ablate the colony formation ability of H660 cells (Fig. 4C). We then proceeded to analyze in greater detail the potential cellular mechanisms underlying the attenuation of cell growth mediated by HP1α knockdown. An EdU incorporation assay showed that knockdown of HP1α led to a significant reduction of EdU-positive, DNA synthesis-active cells (Fig. 4D). Furthermore, an apoptosis assay and FACS analysis indicated that HP1α knockdown also promoted cell early-apoptosis and final death (Fig. 4E and F). Molecularly, the apoptosis markers cleaved-caspase-3 and cleaved-PARP1 were both upregulated in knockdown cells (Fig. 4G); multiple machineries involved in DNA damage repair and cell-cycle progression were also impaired in HP1α knockdown cells as determined by GSEA and IPA from stable cell line transcriptomic profiles (Fig. 4H; Supplementary Fig. S3C and S3D). Thus, HP1α depletion inhibited NEPC cell proliferation and induced apoptosis in vitro.

We then assessed the functional impact of HP1α knockdown on tumor growth in vivo. Compared with control H660 cells, xenograft tumor growth of the KD2 stable cell line with 70% HP1α knockdown (Fig. 4A; Supplementary Fig. S3E) was observed to be dramatically inhibited and much slower, as determined by continuous tumor volume measurements and final fresh tumor weights (Fig. 4I and J). The other two stable lines (KD1, KD3) with 80% to 95% HP1α knockdown were not able to generate sufficient cell numbers for in vivo tumor formation assays. Taken together, the in vitro and in vivo studies establish that HP1α knockdown could both inhibit NEPC cell proliferation and induce apoptosis, leading to NEPC tumor growth arrest.

HP1α promotes NE transdifferentiation of prostatic adenocarcinoma cells following ADT

Because HP1α was identified as a potential early driver of NEPC development following hormone therapy, we further investigated whether HP1α could enhance NE phenotype in ADT-induced NE differentiation of adenocarcinoma cells similar to our LTL331/331R NE transdifferentiation PDX model. We used V16D cells, a LNCaP-derived CRPC cell line (20), to construct stable cell lines with HP1α ectopic expression (Fig. 5A). Upon ADT using CSS or EnZ treatment for 14 days, both control and HP1α-overexpressing V16D cells were assessed for the expression of terminal NE markers. HP1α overexpression significantly promoted the expression of NCAM1 (CD56) and NSE, as shown by both qRT-PCR and Western blotting (Fig. 5B and C). HP1α overexpression alone was not able to induce NE transdifferentiation (Fig. 5B and C). Further transcriptomic analysis of control and HP1α-overexpressing V16D cells upon EnZ treatment revealed that HP1α overexpression consistently upregulated the expression of a panel of NEPC marker genes and enriched neuronal-associated signaling pathways (Fig. 5D and E; Supplementary Fig. S4A). Pearson correlation analyses with transcriptomic data from multiple CRPC clinical cohorts (i.e., the Beltran 2016, the Kumar 2016, the Grasso 2012, and the Varambally 2005) further demonstrated that high expression of HP1α is positively correlated with the expression of terminal NE markers (i.e., NCAM1, NSE, CHGA, CHGB) in advanced prostate cancer (Fig. 5F and G; Supplementary Fig. S4B and S4C; refs. 10, 24, 25, 27). Overall, these analyses suggest that HP1α is a potential functional driver promoting NE transdifferentiation.

HP1α represses AR and REST expression and enriches H3K9me3 on their promoters

We further analyzed how HP1α contributed to the NE phenotype. A major function of HP1α in heterochromatin is to repress gene expression using epigenetic machineries (31, 32), which is an observation supported by our transcriptomic analyses with HP1α knockdown and overexpressing cells (Fig. 5E; Supplementary Fig. S4D). Decreased or loss of expression of the crucial adenocarcinoma lineage–specific transcription factors AR, FOXA1, and REST is a well-established mechanism leading to NE differentiation (12–14). Notably, when HP1α was overexpressed in V16D cells, AR and REST were downregulated at both the mRNA and protein levels (Fig. 6A and B; Supplementary Fig. S5A). Consistently, a panel of AR target genes that are repressed in clinical NEPCs are also downregulated upon HP1α overexpression in V16D cells (Fig. 6C). Reciprocally, HP1α knockdown in H660 cells reactivated AR and REST mRNA expression (Supplementary Fig. S5B), though their protein levels remained undetectable due to low abundance.

We next investigated potential mechanisms underlying the downregulation of AR and REST upon HP1α overexpression. We found that, rather than affecting global heterochromatin-related genes, HP1α overexpression significantly enriched (while knockdown impaired) pericentric heterochromatin machineries as determined by GSEA (Fig. 6D; Supplementary Fig. S5C–S5E). Pericentric heterochromatin is a constitutive heterochromatin structure characterized by the repressive histone mark H3K9me3 (32). We then performed a chromatin immunoprecipitation (ChIP) assay to examine the occupancy of H3K9me3 on the promoter regions of AR and REST following HP1α modulation. Interestingly, in V16D cells, HP1α overexpression increased the occupancy of H3K9me3 on both the AR and REST promoters, whereas its knockdown in H660 cells decreased H3K9me3 enrichment on the AR promoter (Fig. 6E; Supplementary Fig. S5F). In accordance with our findings in cell lines, the pericentric heterochromatin geneset is consistently and significantly enriched in NEPCs compared with adenocarcinomas in our PDX models, multiple clinical cohorts, and GEM models (Fig. 6F; Supplementary Fig. S5G). Meanwhile, Pearson correlation analyses with multiple clinical cohorts showed that HP1α expression is negatively correlated with AR and REST expression in advanced prostate cancer samples (Fig. 6G and H; Supplementary Fig. S5H; refs. 10, 24, 25, 27). Overall, our data indicate that HP1α could regulate expression of the two crucial adenocarcinoma lineage–specific transcription factors AR and REST potentially via modulating the enrichment of H3K9me3 on their promoters.

Loss of luminal epithelial cell characteristics together with gain of small cell neuroendocrine features makes NEPC a completely different disease than typical adenocarcinoma CRPC. This transdifferentiation makes all ARPIs inapplicable despite their otherwise remarkable revolution on the treatment of advanced prostate cancers (1, 14). More critically, the application of potent ARPIs for the clinical management of CRPC could accelerate the incidence of NEPC in near future (2, 3, 33). As such, deciphering the biological and molecular mechanisms underlying NEPC development is fundamentally important for developing novel therapeutics.

Gross differences in nuclear hyperchromatic morphology between adenocarcinoma cells and NEPC cells have long been a criterion for NEPC pathologic diagnosis. Our study here unmasks, for the first time, the precise molecular basis underlying this NEPC-specific nuclear phenotype and identifies a 36-gene NEPC heterochromatin signature using multiple high-fidelity prostatic adenocarcinoma and NEPC PDXs and two prevalent clinical cohorts. Notably, our heterochromatin gene signature can significantly distinguish NEPCs from adenocarcinomas similar to two other NEPC gene signatures derived from whole-genomic differences, suggesting a crucial function of heterochromatin in NEPC and also a potential application for diagnostic purposes. Furthermore, as epigenetic machineries play a major foundational role in heterochromatin formation and function (6, 34), this 36-gene heterochromatin signature also includes 29 epigenetic factors (35). Among them, PcG genes such as EZH2 have already been demonstrated to play critical roles in NEPC (8, 9, 36). Thus, these heterochromatin genes also provide a molecular basis for NEPC development and aggressiveness. Considering that the hyperchromatic nuclear pattern in NEPC is also shared by other small cell carcinomas such as small cell lung cancer, this heterochromatin signature may be further applicable to small cell carcinomas of other tissue origins (37).

From these 36 heterochromatin genes, HP1α was identified as a potential early driver of NE transdifferentiation. The longitudinal analyses of LTL331 PDX tissues following host castration demonstrated that the upregulated expression of HP1α is an early event. This is in contrast to a number of genes previously reported to be involved in NEPC development, such as EZH2 and CBX2 (9, 23), which were only upregulated in the fully developed NEPC LTL331R tumor. Notably, our LTL331/331R model is not only clinically relevant (21, 22), but also highly reproducible in delivering the same NE transdifferentiation in 18 individual repeats without exception so long as castration is applied. This robust phenomenon indicates that NE transdifferentiation in the LTL331/331R model is lineage determined, and the early changes detected in castrated tumors may thus reflect an inevitable and not stochastic event. More importantly, we indeed demonstrate that HP1α can promote terminal NE marker expression in adenocarcinoma cells following ADT. In our study, we also noticed that although HP1α was able to repress AR expression and AR signaling under AR-driven prostatic epithelial status, it cannot function as a neural factors to directly induce the NE phenotype. Only when AR signaling is diminished by ADT can HP1α ectopic expression promote NE transdifferentiation in adenocarcinoma cells. This process both recapitulates the in vivo NE transdifferentiation phenotype occurring in the LTL331/331R model and mirrors the clinical progression of NEPC where most cases appear after hormonal therapy (3, 33). In terminal NEPC cells where HP1α mainly functions as a regulator of aggressiveness, HP1α knockdown cannot alter NE phenotype (Supplementary Fig. S6A). Therefore, HP1α could potentially serve as an early therapeutic target to interfere with disease progression before NEPC fully develops.

In addition to being significantly overexpressed in clinical NEPC samples, HP1α plays a crucial role in terminal NEPC as demonstrated by thorough functional studies in the bona fide NEPC cell line NCI-H660. Although previous studies have reported the function of HP1α in breast cancer, lung cancer, and cholangiocarcinoma, its function in prostate cancer remains elusive (38–42). Our data demonstrate that HP1α knockdown in NEPC cells dramatically inhibited proliferation, completely ablated colony formation, and induced apoptotic cell death. Consequently, HP1α depletion markedly inhibited NEPC tumor growth in vivo. Alternatively, in V16D adenocarcinoma cells where HP1α overexpression promoted NE transdifferentiation, HP1α overexpression did not significantly enhance proliferation (Supplementary Fig. S6B). These data suggest that HP1α is particularly essential for NEPC malignancy, with one potential mechanism being HP1α depletion impairs mitotic machineries. Although reported to drive and maintain heterochromatin structure (31, 32, 43), HP1α is prominently associated with constitutive heterochromatins (44) as demonstrated through our study. Depletion of HP1α did not affect heterochromatin-related genes universally, but significantly impaired the pericentric constitutive heterochromatin machineries. Pericentric heterochromatin is a key element ensuring proper chromosome segregation in metaphase (6), which is also a major previously reported function of HP1α (38, 45). The abnormally enriched pericentric heterochromatin genes in NEPCs may also explain the highly proliferative feature of NEPC, for which HP1α may play a driver function. Another potential mechanism underlying the essential function of HP1α in NEPC aggressiveness is that HP1α depletion impaired DNA damage response (DDR) machineries, which is in accordance with previous studies (46, 47). Most recently, another study also reported that a DDR pathway is enriched in NEPC, contributing to NEPC cell proliferation (48). Overall, HP1α could potentially serve as a therapeutic target for effective management of developed NEPC.

Our findings also suggest a novel, HP1α-mediated mechanism of NEPC development. AR, FOXA1, and REST are the three crucial adenocarcinoma lineage–specific transcription factors maintaining luminal epithelial characteristics (12–14). Our data demonstrate that HP1α can repress AR and REST expression in adenocarcinoma cells, whereas its depletion in NEPC cell can reactivate their expression. HP1α gene was first identified to be a modulator of position effect variegation where euchromatic genes abnormally juxtaposed with pericentric heterochromatin could be silenced due to compaction into heterochromatin (49). In our study, we also found that HP1α may play an epistatic role in regulating the pericentric heterochromatin apparatus. The repressive histone mark H3K9me3 is a hallmark of pericentric heterochromatin and also a major substrate recognized by HP1α (31, 32). Given that the AR and REST genes are natively located in the vicinity of pericentric heterochromatins (Xq12 and 4q12 respectively; Supplementary Fig. S6C), they could potentially be sensitive to pericentric heterochromatin deregulation. Our data indeed showed that modulation of HP1α regulates the enrichment of H3K9me3 on the promoters of AR and REST. Considering the significant enrichment of H3K9me3-characterized pericentric heterochromatin genes in multiple NEPC models and clinical cohorts, our study suggests that the HP1α/H3K9me3 axis may partially explain the absence or loss of AR and REST expression in NEPCs. Further investigation on genome-wide occupancy of HP1α and H3K9me3 will provide valuable insights. HP1α mediates gene silencing together with other precise epigenetic machineries (31). Previous studies have shown that AR expression can also be repressed by the EZH2/H3K27me3 axis (8, 36). As such, HP1α/H3K9me3 might coordinate with EZH2/H3K27me3, establishing the complete repression of AR in NEPCs. Alternative splicing has been suggested to suppress REST in NEPCs (16). HP1α was also reported to mediate mRNA alternative splicing and exon recognition (50). As such, HP1α might be involved in mediating REST splicing as well. Our data here suggest HP1α/H3K9me3 as a new mechanism leading to the silencing of REST mRNA in most NEPC samples.

In summary, our data imply a novel mechanism underlying NEPC development: HP1α drives the abnormal formation of pericentric heterochromatin, which in turn promotes ADT-induced NE transdifferentiation via repressing AR and REST expression, and confers the malignant NEPC phenotype via promoting aggressive proliferation and cell survival. Taken together, HP1α can be considered an early and master mediator of NEPC development and aggressiveness, making it an exceptional novel therapeutic target for potentially effective treatment of NEPC.

No potential conflicts of interest were disclosed.

Conception and design: X. Ci, F. Crea, C.J. Ong, D. Lin, Y. Wang

Development of methodology: X. Ci, J. Hao, R. Wu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Ci, J. Hao, H. Xue, R. Wu, S. Qi, A.M. Haegert, A. Zoubeidi, M.E. Gleave, C.C. Collins

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Ci, J. Hao, X. Dong, R. Wu, S. Qu, F. Crea, C.C. Collins, D. Lin, Y. Wang

Writing, review, and/or revision of the manuscript: X. Ci, J. Hao, S.Y. Choi, P.W. Gout, F. Zhang, F. Crea, M.E. Gleave, D. Lin, Y. Wang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Ci, J. Hao, X. Dong, A.M. Haegert, Y. Wang

Study supervision: H.H. He, C.C. Collins, Y. Wang

Other (pathology): L. Fazil

We would like to thank all members of the Y. Wang laboratory for technical support and helpful discussions. We also would like to thank Dr. Jin-Tang Dong (Emory University, Atlanta, GA) for providing the plasmids described herein. This work was supported in part by the Canadian Institutes of Health Research (Y. Wang), Terry Fox Research Institute (Y. Wang, M.E. Gleave, C.C. Collins, and C.J. Ong), BC Cancer Foundation (Y. Wang), Urology Foundation (Y. Wang), and Prostate Cancer Canada (Y. Wang). X. Ci was supported by the Prostate Cancer Canada-Movember training award. J. Hao and S. Qu were supported by the Mitacs Accelerate award. S.Y. Choi was supported by the CIHR Doctoral award.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Attard
G
,
Parker
C
,
Eeles
RA
,
Schroder
F
,
Tomlins
SA
,
Tannock
I
, et al
Prostate cancer
.
Lancet
2016
;
387
:
70
82
.
2.
Eric Jay
S
,
Rahul Raj
A
,
Jiaoti
H
,
Artem
S
,
Li
Z
,
Joshi
JA
, et al
Clinical and genomic characterization of metastatic small cell/neuroendocrine prostate cancer (SCNC) and intermediate atypical prostate cancer (IAC): results from the SU2C/PCF/AACRWest Coast Prostate Cancer Dream Team (WCDT)
.
J Clin Oncol
2016
;
34
(
15_suppl
):
5019
.
3.
Vlachostergios
PJ
,
Puca
L
,
Beltran
H
. 
Emerging variants of castration-resistant prostate cancer
.
Curr Oncol Rep
2017
;
19
:
32
.
4.
Humphrey
PA
. 
Diagnosis of adenocarcinoma in prostate needle biopsy tissue
.
J Clin Pathol
2007
;
60
:
35
42
.
5.
Fischer
AH
,
Jacobson
KA
,
Rose
J
,
Zeller
R
. 
Hematoxylin and eosin staining of tissue and cell sections
.
CSH Protoc
2008
;
2008
:
pdb prot4986
.
6.
Saksouk
N
,
Simboeck
E
,
Dejardin
J
. 
Constitutive heterochromatin formation and transcription in mammals
.
Epigenetics Chromatin
2015
;
8
:
3
.
7.
Morgan
MA
,
Shilatifard
A
. 
Chromatin signatures of cancer
.
Genes Dev
2015
;
29
:
238
49
.
8.
Dardenne
E
,
Beltran
H
,
Benelli
M
,
Gayvert
K
,
Berger
A
,
Puca
L
, et al
N-Myc induces an EZH2-mediated transcriptional program driving neuroendocrine prostate cancer
.
Cancer Cell
2016
;
30
:
563
77
.
9.
Clermont
PL
,
Lin
D
,
Crea
F
,
Wu
R
,
Xue
H
,
Wang
Y
, et al
Polycomb-mediated silencing in neuroendocrine prostate cancer
.
Clin Epigenetics
2015
;
7
:
40
.
10.
Beltran
H
,
Prandi
D
,
Mosquera
JM
,
Benelli
M
,
Puca
L
,
Cyrta
J
, et al
Divergent clonal evolution of castration-resistant neuroendocrine prostate cancer
.
Nat Med
2016
;
22
:
298
305
.
11.
Ku
SY
,
Rosario
S
,
Wang
Y
,
Mu
P
,
Seshadri
M
,
Goodrich
ZW
, et al
Rb1 and Trp53 cooperate to suppress prostate cancer lineage plasticity, metastasis, and antiandrogen resistance
.
Science
2017
;
355
:
78
83
.
12.
Wright
ME
,
Tsai
MJ
,
Aebersold
R
. 
Androgen receptor represses the neuroendocrine transdifferentiation process in prostate cancer cells
.
Mol Endocrinol
2003
;
17
:
1726
37
.
13.
Lapuk
AV
,
Wu
C
,
Wyatt
AW
,
McPherson
A
,
McConeghy
BJ
,
Brahmbhatt
S
, et al
From sequence to molecular pathology, and a mechanism driving the neuroendocrine phenotype in prostate cancer
.
J Pathol
2012
;
227
:
286
97
.
14.
Terry
S
,
Beltran
H
. 
The many faces of neuroendocrine differentiation in prostate cancer progression
.
Front Oncol
2014
;
4
:
60
.
15.
Kim
J
,
Jin
H
,
Zhao
JC
,
Yang
YA
,
Li
Y
,
Yang
X
, et al
FOXA1 inhibits prostate cancer neuroendocrine differentiation
.
Oncogene
2017
;
36
:
4072
80
.
16.
Li
Y
,
Donmez
N
,
Sahinalp
C
,
Xie
N
,
Wang
Y
,
Xue
H
, et al
SRRM4 drives neuroendocrine transdifferentiation of prostate adenocarcinoma under androgen receptor pathway inhibition
.
Eur Urol
2017
;
71
:
68
78
.
17.
Zou
M
,
Toivanen
R
,
Mitrofanova
A
,
Floc'h
N
,
Hayati
S
,
Sun
Y
, et al
Transdifferentiation as a mechanism of treatment resistance in a mouse model of castration-resistant prostate cancer
.
Cancer Discov
2017
;
7
:
736
49
.
18.
Lee
JK
,
Phillips
JW
,
Smith
BA
,
Park
JW
,
Stoyanova
T
,
McCaffrey
EF
, et al
N-Myc drives neuroendocrine prostate cancer initiated from human prostate epithelial cells
.
Cancer Cell
2016
;
29
:
536
47
.
19.
Mu
P
,
Zhang
Z
,
Benelli
M
,
Karthaus
WR
,
Hoover
E
,
Chen
CC
, et al
SOX2 promotes lineage plasticity and antiandrogen resistance in TP53- and RB1-deficient prostate cancer
.
Science
2017
;
355
:
84
8
.
20.
Bishop
JL
,
Thaper
D
,
Vahid
S
,
Davies
A
,
Ketola
K
,
Kuruma
H
, et al
The master neural transcription factor BRN2 is an androgen receptor suppressed driver of neuroendocrine differentiation in prostate cancer
.
Cancer Discov
2017
;
7
:
54
71
.
21.
Lin
D
,
Wyatt
AW
,
Xue
H
,
Wang
Y
,
Dong
X
,
Haegert
A
, et al
High fidelity patient-derived xenografts for accelerating prostate cancer discovery and drug development
.
Cancer Res
2014
;
74
:
1272
83
.
22.
Akamatsu
S
,
Wyatt
AW
,
Lin
D
,
Lysakowski
S
,
Zhang
F
,
Kim
S
, et al
The placental gene PEG10 promotes progression of neuroendocrine prostate cancer
.
Cell Rep
2015
;
12
:
922
36
.
23.
Beltran
H
,
Rickman
DS
,
Park
K
,
Chae
SS
,
Sboner
A
,
MacDonald
TY
, et al
Molecular characterization of neuroendocrine prostate cancer and identification of new drug targets
.
Cancer Discov
2011
;
1
:
487
95
.
24.
Grasso
CS
,
Wu
YM
,
Robinson
DR
,
Cao
X
,
Dhanasekaran
SM
,
Khan
AP
, et al
The mutational landscape of lethal castration-resistant prostate cancer
.
Nature
2012
;
487
:
239
43
.
25.
Kumar
A
,
Coleman
I
,
Morrissey
C
,
Zhang
X
,
True
LD
,
Gulati
R
, et al
Substantial interindividual and limited intraindividual genomic diversity among tumors from men with metastatic prostate cancer
.
Nat Med
2016
;
22
:
369
78
.
26.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
27.
Varambally
S
,
Yu
J
,
Laxman
B
,
Rhodes
DR
,
Mehra
R
,
Tomlins
SA
, et al
Integrative genomic and proteomic analysis of prostate cancer reveals signatures of metastatic progression
.
Cancer Cell
2005
;
8
:
393
406
.
28.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
29.
Liberzon
A
,
Subramanian
A
,
Pinchback
R
,
Thorvaldsdottir
H
,
Tamayo
P
,
Mesirov
JP
. 
Molecular signatures database (MSigDB) 3.0
.
Bioinformatics
2011
;
27
:
1739
40
.
30.
Johnson
BE
,
Whang-Peng
J
,
Naylor
SL
,
Zbar
B
,
Brauch
H
,
Lee
E
, et al
Retention of chromosome 3 in extrapulmonary small cell cancer shown by molecular and cytogenetic studies
.
J Natl Cancer Inst
1989
;
81
:
1223
8
.
31.
Eissenberg
JC
,
Elgin
SC
. 
HP1a: a structural chromosomal protein regulating transcription
.
Trends Genet
2014
;
30
:
103
10
.
32.
Maison
C
,
Almouzni
G
. 
HP1 and the dynamics of heterochromatin maintenance
.
Nat Rev Mol Cell Biol
2004
;
5
:
296
304
.
33.
Rickman
DS
,
Beltran
H
,
Demichelis
F
,
Rubin
MA
. 
Biology and evolution of poorly differentiated neuroendocrine tumors
.
Nat Med
2017
;
23
:
1
10
.
34.
Grewal
SI
,
Jia
S
. 
Heterochromatin revisited
.
Nat Rev Genet
2007
;
8
:
35
46
.
35.
Medvedeva
YA
,
Lennartsson
A
,
Ehsani
R
,
Kulakovskiy
IV
,
Vorontsov
IE
,
Panahandeh
P
, et al
EpiFactors: a comprehensive database of human epigenetic factors and complexes
.
Database (Oxford)
2015
;
2015
:
bav067
.
36.
Kleb
B
,
Estecio
MR
,
Zhang
J
,
Tzelepi
V
,
Chung
W
,
Jelinek
J
, et al
Differentially methylated genes and androgen receptor re-expression in small cell prostate carcinomas
.
Epigenetics
2016
;
11
:
184
93
.
37.
Travis
WD
. 
Update on small cell carcinoma and its differentiation from squamous cell carcinoma and other non-small cell carcinomas
.
Mod Pathol
2012
;
25
Suppl 1
:
S18
30
.
38.
De Koning
L
,
Savignoni
A
,
Boumendil
C
,
Rehman
H
,
Asselain
B
,
Sastre-Garau
X
, et al
Heterochromatin protein 1alpha: a hallmark of cell proliferation relevant to clinical oncology
.
EMBO Mol Med
2009
;
1
:
178
91
.
39.
Cheng
W
,
Tian
L
,
Wang
B
,
Qi
Y
,
Huang
W
,
Li
H
, et al
Downregulation of HP1alpha suppresses proliferation of cholangiocarcinoma by restoring SFRP1 expression
.
Oncotarget
2016
;
7
:
48107
19
.
40.
Yu
YH
,
Chiou
GY
,
Huang
PI
,
Lo
WL
,
Wang
CY
,
Lu
KH
, et al
Network biology of tumor stem-like cells identified a regulatory role of CBX5 in lung cancer
.
Sci Rep
2012
;
2
:
584
.
41.
Itsumi
M
,
Shiota
M
,
Yokomizo
A
,
Kashiwagi
E
,
Takeuchi
A
,
Tatsugami
K
, et al
Human heterochromatin protein 1 isoforms regulate androgen receptor signaling in prostate cancer
.
J Mol Endocrinol
2013
;
50
:
401
9
.
42.
Shapiro
E
,
Huang
H
,
Ruoff
R
,
Lee
P
,
Tanese
N
,
Logan
SK
. 
The heterochromatin protein 1 family is regulated in prostate development and cancer
.
J Urol
2008
;
179
:
2435
9
.
43.
Larson
AG
,
Elnatan
D
,
Keenen
MM
,
Trnka
MJ
,
Johnston
JB
,
Burlingame
AL
, et al
Liquid droplet formation by HP1alpha suggests a role for phase separation in heterochromatin
.
Nature
2017
;
547
:
236
40
.
44.
Nielsen
AL
,
Ortiz
JA
,
You
J
,
Oulad-Abdelghani
M
,
Khechumian
R
,
Gansmuller
A
, et al
Interaction with members of the heterochromatin protein 1 (HP1) family and histone deacetylation are differentially involved in transcriptional silencing by members of the TIF1 family
.
EMBO J
1999
;
18
:
6385
95
.
45.
Abe
Y
,
Sako
K
,
Takagaki
K
,
Hirayama
Y
,
Uchida
KS
,
Herman
JA
, et al
HP1-assisted Aurora B kinase activity prevents chromosome segregation errors
.
Dev Cell
2016
;
36
:
487
97
.
46.
Luijsterburg
MS
,
Dinant
C
,
Lans
H
,
Stap
J
,
Wiernasz
E
,
Lagerwerf
S
, et al
Heterochromatin protein 1 is recruited to various types of DNA damage
.
J Cell Biol
2009
;
185
:
577
86
.
47.
Lee
YH
,
Kuo
CY
,
Stark
JM
,
Shih
HM
,
Ann
DK
. 
HP1 promotes tumor suppressor BRCA1 functions during the DNA damage response
.
Nucleic Acids Res
2013
;
41
:
5784
98
.
48.
Zhang
W
,
Liu
B
,
Wu
W
,
Li
L
,
Broom
BM
,
Basourakos
SM
, et al
Targeting the MYCN-PARP-DNA damage response pathway in neuroendocrine prostate cancer
.
Clin Cancer Res
2018
;
24
:
696
707
.
49.
Eissenberg
JC
,
James
TC
,
Foster-Hartnett
DM
,
Hartnett
T
,
Ngan
V
,
Elgin
SC
. 
Mutation in a heterochromatin-specific chromosomal protein is associated with suppression of position-effect variegation in Drosophila melanogaster
.
Proc Natl Acad Sci U S A
1990
;
87
:
9923
7
.
50.
Yearim
A
,
Gelfman
S
,
Shayevitch
R
,
Melcer
S
,
Glaich
O
,
Mallm
JP
, et al
HP1 is involved in regulating the global impact of DNA methylation on alternative splicing
.
Cell Rep
2015
;
10
:
1122
34
.