Phenotypic Plasticity – Alternate Transcriptional Programs Driving Treatment Resistant Prostate Cancer

Abstract

Androgen deprivation therapy (ADT) that antagonizes androgen receptor (AR) signaling has made significant increases to overall survival of prostate cancer patients. However, ADT is not curative, and patients eventually progress to castration resistant disease (CRPC). It has become evident that a subset of prostate cancers acquire ADT resistance through mechanisms independent of AR alteration or reprogramming of AR signaling. This approximately involves a quarter of prostate cancers progressing on ADT. Collectively, these tumors evolve via phenotypic plasticity and display the activation of developmental and stemness gene signatures as well as transitional programs including an epithelial–mesenchymal phenotype. Currently, no successful treatments exist for prostate cancer patients to inhibit or reverse prostate tumor progression that utilizes mechanisms of epi-plasticity. This overview will discuss epigenetic mechanisms that mediate phenotypic plasticity and the potential for targeting the epigenome to create a novel direction for combination strategies involving epigenetic therapy to provide durable response.

I. INTRODUCTION

Prostate cancer is the second leading cause of cancer death in men in the United States.¹ Androgen deprivation therapy (ADT) by use of androgen receptor (AR) targeting drugs (such as enzalutamide, darolutamide) or androgen synthesis targeting drugs (abiraterone acetate) remains the standard of care treatment for metastatic prostate cancer patients. Although these therapies are initially effective, patients eventually develop castration resistant prostate cancer (CRPC). The majority of CRPC tumors remain dependent on AR signaling through several mechanisms including AR amplification, mutation, and constitutive expression of AR splice variants.^2,3 However, approximately a quarter of CRPC patients develop ADT resistance by lineage plasticity, which largely involves the reprogramming of AR positive adenocarcinoma to AR negative/low treatment-emergent neuroendocrine prostate cancer (t-NEPC). More recently, data has emerged to demonstrate that prostate tumors can evade ADT via lineage plasticity but maintain expression and function of the AR, highlighting the growing spectrum of lineage plastic related mechanisms underlying disease progression of prostate cancers. For this review, we will refer to this spectrum of prostate cancers as lineage plastic prostate cancer (LP-PCa).^4,5

LP-PCa is an aggressive subset of prostate tumors with high proliferation rate and rapid tumor dissemination.⁶ LP-PCa is a result of cellular reprogramming that leads to a loss of identity, acquiring an alternative lineage state that may occur directly through transdifferentiation of luminal adenocarcinoma or through epithelial mesenchymal transition (EMT) leading to an intermediate stem-like state. Lineage plasticity has added a further level of complexity to prostate cancer tumors, which rather than being defined as distinct disease states, are represented as a continuum depending upon the expression of AR pathway genes and neuroendocrine (NE) markers. LP-PCa tumors have been defined as hybrid or amphicrine tumors (AR+NE+) or double negative tumors (AR-NE-).⁷ The LP-PCa phenotype is usually represented histologically by small cells with prominent and hyperchromatic nuclei, limited cytoplasm with eosinophilic granules, and expression of neuroendocrine markers such as chromogranin (CHGA), synaptophysin (SYP), and neuron specific enolase (NSE/ENO2).^8,9 Collectively, LP-PCa have few genomic alterations that distinguish them from CRPC except for MYCN and AURKA amplification, deletion or mutation of TP53, and loss of RB1. Further, the genetic aberration TMPRSS2-ERG (T:E) fusion that is expressed in approximately 50% of all AR positive prostate cancers has been detected in NEPC samples, strengthening the origin of LP-PCa to be from adenocarcinoma CRPC tumors utilizing lineage switching to evade ADT.^10–12 Moreover, these molecular changes allow the prostate cancer cell to escape therapy through transdifferentiation from its initial adenocarcinoma cell identity to one that resembles a neuroendocrine cell identity. Further, without significant changes to tumor mutational burden, it has become apparent that this cell state change is mediated by nonmutational epigenetic programming.¹³ With this, a number of clinical trials involving epigenetic therapies have been initiated over time treating prostate cancer patients (Table 1). However, without a full understanding and identification of epigenetic reprogramming dependencies, it will be difficult to execute the full therapeutic benefits for the use of epigenetics therapies to inhibit and/or reverse therapeutic resistance governed by phenotypic plasticity. In this review, we will give a comprehensive overview of the various transcriptional, epigenetic mechanisms and cross talk between them that drives LP-PCa. Depending upon the knowledge gained from very recent studies (discussed in this review) the functional roles of various epigenetic modifiers in LP-PCa have been summarized in Fig. 1.

TABLE 1:

Clinical trials for Epigenetic therapies in prostate cancer

Target	Drug/Agent	Clinical trial ID	Phase	Condition or Disease	Clinical status
EZH2	PF-06821497	NCT03460977	I	Refractory and relapsed CRPC	Recruiting; no results reported
	CPI-1205 + ARPI (enzalutamide or abiraterone /prednisone)	NCT03480646	I/II	mCRPC	Recruiting; no results reported
	Tazemetostat	NCT02875548	II	Advanced solid tumors	Ongoing
	CPI-1205 + immunotherapy (ipilimumab)	NCT03525798	I/II	Advanced solid tumors	Recruiting; no results reported
	EPZ-6438 (Abiraterone/prednisone or enzalutamide)	NCT04179864	I	mCRPC	Recruiting
	SHR3680 (Novel Anti-Androgen)	NCT03741712	I/II	mCRPC	Recruiting
	DS3201 (Ipilimumab)	NCT04388852	I	mCRPC	Recruiting
BET	ZEN003694	NCT02705469	I	mCRPC	Completed; dose confirmation
	I: ZEN003694 II:ZEN003694 + enzalutamide	NCT04145375	I/II	mCRPC	Enrolling by invitation
	I: ZEN003694 II: Enzalutamide	NCT02711956	Ib/IIa	mCRPC	Active; not recruiting
	Pembrolizumab (day 1) ZEN-3694 + enzalutamide (days 1–21)	NCT04471974	II	mCRPC	Not yet recruiting (2020/8 start)
	GS-5829 + enzalutamide	NCT02607228	Ib/II	mCRPC with ARPI	Completed
	MK-8628	NCT02698176	I	CRPC	Terminated
	GSK525762 (Abiraterone/prednisone or enzalutamide)	NCT03150056		mCRPC	Active; not recruiting
	PLX2853 (Abiraterone/prednisone or olaparib)	NCT04556617	I/II	mCRPC	Recruiting
	MK-8628	NCT02259114	lb	mCRPC	Completed
	ABBV-075	NCT02391480	I	mCRPC	Completed
DNMT	Azacitidine for injectable suspension	NCT00384839	II	PCa to hormonal therapy	Completed
	5-AZA +ATRA	NCT03572387	Pilot study	PCa with PSA-only recurrence	Recruiting
	Azacitidine + docetaxel + prednisone	NCT00503984	I/II	mPC	Terminated
	Enzalutamide + decitabine	NCT03709550	Ib/n	mCRPC	Not yet recruiting (2020/8 start)
	Guadecitabine + pembrolizumab	NCT02998567	I	CRPC solid tumors	Active; not recruiting
	Azacitidine (Sodium phenylbutyrate)	NCT00006019	II	mCRPC or nomnetastatic CRPC	Completed
LSD1	INCB059872	NCT02712905	I/II	Advanced malignancies; NEPC	Terminated
	Phenelzine	NCT02217709	II	Nomnetastatic recurrent prostate Cancer	Active; not recruiting
	Phenelzine + docetaxel	NCT01253642	II	PCa with progressive disease	Terminated

FIG. 1:

Cross talk between genetic and epigenetic mechanisms driving prostate cancer progression from a castration-sensitive prostate cancer (CSPC) to lineage plastic prostate cancer (LP-PCa), which involves either a treatment related NEPC (t-NEPC) or epithelial mesenchymal phenotype (EMT). The t-NEPC phenotype is characterized by the presence of transcription factors including ASCL1, INSM1, NEUROD1, BRN2, SOX2, and N-Myc whereas the EMT phenotype is marked by upregulation of SLUG, SNAIL, ZEB1, and ZEB2 and downregulation of E-cadherin. A large number of epigenetic modifiers are involved in LP-PCa progression including EZH2, LSD1, DNMT1, BRD4, SIR2, SET8, and SWI/SNF complex, which are potential targets to treat LP-PCa. EZH2 also drives a t-NEPC independently of its methylase function by acting as a co-transcriptional activator on being phosphorylated as a binding partner with AR. ARSI, androgen receptor signaling inhibitors.

II. CHROMATIN REMODELING, EPIGENETIC REGULATORS, AND PHENOTYPIC PLASTICITY IN PROSTATE CANCER

A. Epithelial Mesenchymal Transition (EMT)

LP-PCa exhibits reactivation of developmental programs that can be associated with induction of epithelial to mesenchymal transition (EMT).¹⁴ EMT is a form of epi-plasticity utilized by prostate cancer cells to evade androgen blockade and permits the acquisition of increased potential for invasion and metastasis.^15–17 Several transcriptional factors including SNAIL, SLUG and zinc finger E-Box-binding homeobox 1 and 2 (ZEB1 and ZEB2) and TWIST1 regulate EMT by downregulating E-cadherin and upregulating various mesenchymal markers such as fibronectin, vimentin, and N-cadherin.^18–25 Overexpression of SNAIL not only induces EMT but enhances the expression of neuroendocrine differentiation markers like chromogranin A and ENO2, demonstrating a molecular link between EMT and activation neuronal gene expression associated with neuroendocrine prostate cancer (NEPC).^18,20

Regulators of chromatin remodeling play an important role in EMT in various cancers including PCa.²⁶ It has been recently demonstrated that over expression of a histone methyl transferase multiple myeloma SET domain (MMSET) in normal prostate cells (RWPE-1) was capable of inducing TWIST1 expression by direct activation of TWIST 1 gene expression.²⁷ Furthermore, decreased expression of E-cadherin has been linked to methylation of its promoter, which is a classical hallmark of EMT.^28,29 Enhancer of zeste homolog-2 (EZH2) is another important epigenetic factor that plays a crucial role in EMT by repressing the expression of E-cadherin and disabling homolog 2 -interacting protein (DAB2IP).³⁰ Additional studies demonstrated that loss of DAB2IP induced EMT via increased expression of PROX1, stabilization of HIF-1α, and up-regulation of vimentin and metallopeptidases.³¹ Byles et al. have demonstrated that Sirtuin 1 (SIRT1), a nicotinamide adenine dinucleotide (NAD)-dependent histone deacetylase, is a positive regulator of EMT in prostate cancer. It was shown that ZEB1 recruits SIRT1 to the E-cadherin promoter and deacetylates histone H3, which leads to reduced RNA polymerase II binding and leading to E-cadherin transcriptional loss.³² Similarly, Hou et al. illustrated the functional role of histone H4K20-specific methyl transferase SET8 in promoting EMT and tumor metastasis in prostate cancer.³³ SET8 was shown to physically associate with ZEB1 through binding to the promoter of E-cadherin and vimentin (Fig. 1).³³

Analysis conducted in a cohort of heavily treated human CRPC by single-cell sequencing, reported the existence of NEPC as well as mesenchymal (EMT-like) populations within single biopsies highlighting EMT as one of the major players of lineage plasticity in LP-PCa.³⁴ This study also conducted single cell sequencing using prostates of Pten:Rb1 double knockout and Pten:Rb1:Trp53 triple knockout⁻ genetically engineered mouse models (GEMMs) of LP-PCa to identify a single stem like luminal epithelial (L2) population that drives a lineage plasticity program leading to EMT-like (Vim+ Ncam1+), NEPC (Ascl1+), Pou2f3+, Tff3+ sub populations as reported earlier in small cell lung cancer (SCLC). Interestingly, this L2 population was enriched in inflammatory related gene signatures that have been linked to altering cell-to-cell signaling enabling cell plasticity.^34,35 Another recent study by Karthaus et al. found that co-deletion of tumor suppressor genes Rb1/Trp53 by lentiviral transduction of prostate organoids also led to a lineage plastic phenotype as evident from reduced expression of luminal lineage genes such as Nkx3.1, Folh1, Dpp4, Krt18, and Krt8 but increased expression of mesenchymal genes such as Zeb2, Vim, and Snai1.³⁶ Single-cell RNA-sequencing using organoids from these GEM models showed EMT (SMAD2 signature), Janus kinase (JAK)–signal transducer and activator of transcription (STAT) and fibroblast growth factor receptor (FGFR) signaling as top candidate pathways driving lineage plasticity in these prostate cancer models.³⁶ A parallel study led by Deng et al. using LNCaP/AR cell lines with loss of TP53/RB1, TP53/RB1/SOX2 and SOX2 overexpression, and Rb1:Tp53 mouse organoids shows that JAK-STAT signaling can promote LP-PCa and EMT phenotype.³⁷ Specifically loss of TP53/RB1 results in over expression of SOX2 in prostate cancer cells, and activates reprograming of H3K27ac, H3K27me3, and H3K4me3 gene targeting resulting in activation of JAK1, implicating JAK-STAT signaling as a potential target for treatment of LP-PCa.^37–39 Similarly, Zou et al. illustrated using GEMMs with co-deletion of Pten and Tp53 that these tumors failed to respond to abiraterone and display accelerated tumorigenesis and advance to a CRPC phenotype with focal regions of neuroendocrine differentiation/lineage plasticity which is driven by higher expression of SOX11.⁴⁰ In vivo lineage tracing studies carried in these mice showed that neuroendocrine regions arise from the transdifferentiation of the luminal adenocarcinoma.⁴⁰ Together, genetic alternations including loss of function of tumor suppressor genes RB1 and TP53 result in significant chromatin remodeling resulting in alternate transcriptional programs ultimately leading to cellular plasticity. Deep analysis of this plasticity phenotype by genomic and proteomic single-cell analysis will enable a more complete understanding and identification of therapeutic targets for validation.

B. Crosstalk between Epigenome and Transcription Factors

Recent evidence suggests that epigenetic changes mediated by chromatin modifying enzymes that govern acetylation, deacetylation, methylation, and demethylation of DNA and histones play a pivotal role to mediate progression to LP-PCa. Earlier studies that focused on determining the drivers of LP-PCa have been based on alternation of somatic DNA events including mutations and copy number loss or gain and differential gene expression analysis.^10,38,39,41 Epigenomic profiling of histone modifications (H3K27ac, H3K4me3 and H3K27me3) in human LP-PCa and select LuCaP prostate cancer patient derived xenografts showed a significant reprogramming of a master transcriptional regulator Forkhead box 1 (FOXA1) to a majority of neuroendocrine specific regulatory elements driving proliferation and expression of neuroendocrine lineage defining genes in LP-PCa.⁴² These findings were validated by increased H3K27ac and FOXA1 binding to neuroendocrine regulatory elements when ectopic expression of lineage plasticity transcription factors ASCL1 (Achaete-scute homolog 1) and NKX2–1 (Nk2-homeobox1) was performed in LN-CaP cell lines.⁴² Cejas et al. reported a shared epigenetic program driving lineage plasticity across various neuroendocrine carcinomas, including NEPC (LP-PCa), SCLC, Merkel cell carcinoma (MCC), and gastrointestinal neuroendocrine carcinoma (GINEC).⁴³ Analysis by ATAC-seq and RNA-seq respectively revealed significant overlap of common chromatin states independent of each cancer’s anatomical location. Shared master regulators include ACSL1, neurogenic differentiation 1 (NEUROD1), and SOX2.⁴³ Within human NEPC and SCLC samples it is demonstrated that ASCL1 and NEUROD1 expression is mutually exclusive and are drivers of intra-tumoral heterogeneity. Specifically to SCLC, ASCL1 and NEUROD1 are shown to define heterogeneity in neuroendocrine tumors by regulating differential gene set expression, implying that each ASCL1 and NEUROD1 cooperate with unique chromatin remodeling factors to drive their respective alternate transcriptional programs.⁴⁴ These unique ASCL1 and NEUROD1 transcriptional programs were demonstrated to occur due to distinct H3K27ac genome wide de novo deposition (Fig. 1).⁴⁴ These data support the association of de novo cis-regulatory elements harboring H3K27ac as a determinant of lineage plasticity in various neuroendocrine carcinomas. Because EZH2 is responsible for H3K27me3 deposition, it is still unclear as to the extent of total regulatory elements silenced by EZH2 action, or another regulator of heterochromatin, that become accessible during lineage plasticity mediated chromatin remodeling. This current unresolved question could offer important mechanistic insight towards earlier identification of prostate adenocarcinomas advancing to LP-PCa.

Pomerantz et al. recently reported that during progression to metastasis, prostate cancer cells reprogram the AR cistrome.⁴⁵ In addition, the authors showed the reactivation of latent regulatory elements that are active during fetal prostate organogenesis. This indicates that prostate cancer reactivates developmental programs upon metastatic progression. While small cell prostate cancer lacks AR, it has been reported that more than 50 percent of treatment resistant prostate tumors harboring neuroendocrine features retain nuclear AR without the activation of canonical AR signaling.^4,46,47 In addition, studies have reported that the epigenetic regulator EZH2 can function as a co-activator of AR through a poly-comb independent mechanism.^48,49 Davies et al. showed that in response to androgen receptor inhibition, prostate tumors retain AR activity and adopt alternative lineage transcriptomes. This study also showed that during this process of lineage transition that the chromatin architecture is remodeled to support AR transcriptional rerouting. A key determinant for this is EZH2, as it is required to co-occupy target genomic sites involved in the reprogrammed AR cistrome.⁵ Based on the studies discussed above, it can be said that AR might play an important in role the development of LP-PCa.

C. Role of LSD1 in LP-PCa

Lysine-specific demethylase 1A (LSD1, KDM1A), is a flavin-dependent monoamine oxidase that regulates gene expression in stem cells and can demethylate both H3K4me1/2 and H3K9me1/2 therefore having dual functions to repress and activate gene expression in multiple cancers.⁵⁰ In breast cancer phosphorylated LSD1 (serine 112) has been shown to facilitate epithelial–mesenchymal plasticity and a metastatic phenotype by targeting H3K4 at the CDH1 (E-cadherin) promoter to repress transcription.⁵¹ In SCLC, Augert et al. showed LSD1 inhibition led to reduced lineage plasticity gene expression and complete tumor regression in FHSCO4, a SCLC patient derived xenograft model.⁵² Further, it was demonstrated that LSD1 inhibition reprogrammed SCLC neuroendocrine differentiation towards a non-neuroendocrine SCLC “inflammatory” phenotype. LSD1 inhibition was mechanistically shown to inhibit the mRNA binding protein, ZFP36L1, which led to degradation of SOX2 and insulinoma-associated protein 1 (INSM1) mRNA.⁵³ These studies highlight LSD1 as an emerging therapeutic target for SCLC.⁵⁴ In prostate cancer, LSD1 regulates AR transcriptional activity by H3K4 demethylation of canonical AR target genes.⁵⁵ Additionally, Laing et al. determined LSD1 reprogramming to a subset of cell cycle related genes including centromere binding protein (CENPE). In this study in CRPC with loss of RB1 expression, CENPE gene regulation was dependent upon co-binding of LSD1 and AR, suggesting an important role for LSD1 in driving LP-PCa.⁵⁶ Sehrawat et al. illustrated LSD1 promotes a LP-PCa phenotype independently of its demethylase function by interacting with a binding protein ZNF217 (co-activator) to promote a LP-PCa phenotype through increased expression of stemness and cell survival genes.⁵⁷ In addition, LSD1 suppression by inhibitors that target the LSD1-ZNF217 interaction showed greater anti-tumor effect when combined with ADT in vitro, thus highlighting the potential for LSD1 as a therapeutic target in LP-PCa by reversing ADT resistance. A recent study by Gao et al. revealed LSD1 mediated demethylation regulates chromatin binding of FOXA1 in prostate cancer.^57,58 LSD1 inhibition alone and in combination with ADT globally disrupted FOXA1 and AR binding in CRPC models.⁵⁸ Interestingly, LSD1 has been shown to coordinate with EZH2 at bivalent chromatin domains to regulate cell plasticity via concerted repression of developmental genes.⁵⁹ A splice variant of LSD1, named LSD1+8a, which is a result of splicing mediated by the splicing factor SRRM4 (Serine/Arginine Repetitive Matrix 4), was shown to be expressed only in NEPC-patient derived xenograft models (LuCaP 49, LuCaP 93, LuCaP 145.1, LuCaP 145.2, LTL 352, and LTL 331R).⁶⁰ This study showed that LSD1+8a and SRRM4 co-regulate a number of genes including interferon induced protein with tetratricopeptide repeats (IFIT3), bone marrow stromal cell antigen 2 (BST2), tripartite motif containing 22 (TRIM22) that drive LP-PCa.⁶⁰ More recently, increased minor splicesome (MiS) dependent splicing was shown to control prostate cell fate. MiS was primarily mediated by the small nuclear (snRNA) U6atac and activation led to progression of CRPC and NEPC phenotypes and resistance to ADT. Reversal of ADT resistance was significantly enhance by inhibition of U6atac expression.⁶¹ These data imply the importance of spliceosome mediated mechanisms and their potential as novel therapeutic targets to inhibit and/or reverse LP-PCa (Fig. 1).

D. EZH2 as Master Regulator in LP-PCa

The Polycomb group proteins (PcG) consist of two main complexes, polycomb repressive complex 1 (PRC1) and polycomb repressive complex 2 (PRC2), that repress transcription and are known to silence tumor suppressor genes in cancer, which include homeobox genes involved in differentiation.^62–64 EZH2 is a subunit of PRC2 complex that catalyzes the trimethylation of histone H3 on Lys 27 (H3K27) and involved in multiple cancer types.⁶⁵ In prostate cancer, EZH2 expression is upregulated via transcription and/or amplification and associated with prostate cancer initiation and progression.^65,66 Furthermore, numerous studies have reported that upregulation of EZH2 is a driver of LP-PCa progression.^67–69 Clermont et al. reported the generation of a neuroendocrine associated repression signature (NEARS). NEARS demonstrated significant enrichment of polycomb (including EZH2) targeted genes, highlighting polycomb function as an important mechanism in LP-PCa.⁶⁹ Further evidence supporting the role of EZH2 as a key regulator of epigenetic rewiring in LP-PCa was shown by GEMM models driven by either overexpression of N-Myc or deletion of Rb1 and/or Tp53.^70,71 Both models had significant transcriptional activation of neuroendocrine and stem cell gene signatures, including significant up-regulation of EZH2. The models were resistant to ADT because of the ability to shut down AR expression/function; however, genetic and chemical inhibition of EZH2 resulted in induced AR expression and sensitivity to ADT. Functionally, EZH2 co-operates with lineage-driving transcription factors to regulate gene expression and coordinate lineage specification. In LP-PCa, EZH2 directly binds with N-Myc to transcriptionally repress genes that enforce an AR-driven adenocarcinoma state.⁷¹ Conditional expression of N-Myc (along with activated AKT) in prostate epithelial cells is sufficient to induce neuroendocrine differentiation, which is dependent on the cooperation of N-Myc with EZH2.^71,72 Furthermore, another study identified N-Myc at neuronal lineage-associated gene promoters harboring both repressive H3K27me3 and active H3K4me3 histone marks (termed bivalent chromatin). EZH2 was required to maintain the bivalency at N-Myc-bound genes, and its knockdown led to de-enrichment of neuronal-associated pathways in LP-PCa.⁷³

EZH2 has also been reported to cooperate with other chromatin remodeling enzymes required for lineage plasticity and LP-PCa. For example, EZH2 has been linked with DNA methyltransferase (DNMT) activity and may pre-mark genes for de novo DNA methylation.^74,75 This occurs potentially via a scaffolding mechanism mediated by the long ncRNA HOTAIR.⁷⁶ DNMT1 is overexpressed in NEPC indicating its role in LP-PCa.⁷⁷ Use of the transgenic adenocarcinoma of the mouse prostate (TRAMP) mice model by McCabe et al. showed that DNMT1 plays a pivotal role in prostate tumorigenesis.⁷⁸ Furthermore, Beltran et al. demonstrated that NEPC patient samples displayed differential DNA methylation patterns and increased EZH2 expression when compared to CRPC patient samples providing strong rationale for coordinated chromatin remodeling by these epigenetic mechanisms.⁷⁹ Of note, EZH2 and DNMT complex is not deeply studied in prostate cancer; however, in breast cancer it has been shown that EZH2 targets DNMT1 to gene promoters involved in Hippo signaling to enhance epithelial–mesenchymal plasticity and stem-like phenotypes.⁸⁰ Furthermore, EZH2 also complexes with a H3K36me2 methyltransferase, nuclear receptor-binding SET domain protein 2 (NSD2), and is shown to be overexpressed in LP-PCa.⁸¹ NSD2 reprograms the epigenome by repositioning the binding distribution of EZH2.⁸² NSD2 activity is also regulated by EZH2, which is needed for EZH2-mediated epigenetic reprogramming of PCa.⁸³ Given the important role of EZH2 in LP-PCa, it is an important therapeutic target the requires continued investigation.^71,84 The chromodomain protein CBX2 is a member of PRC1, and binds to H3K27me3 via its chromo-domain.⁶⁹ Furthermore, CBX2 represses gene transcription at target loci, and can carry out this function in a PRC2/EZH2 dependent and independent manner.^85,86 These results are further supported by recent transcriptomic analysis of an independent patient cohort consisting of treatment emergent NEPC and CRPC samples, and previously mentioned work from Clermont et al. showing concurrent upregulation of CBX2 and EZH2 could be important partners coordinating chromatin remodeling to drive progression towards LP-PCa.^46,69 Mechanistically, it has been demonstrated that RB1 recruits EZH2 to specific genome coordinates for H3K27me3 deposition, and RB1 loss predisposed animal models to lymphoma by derepression of these EZH2 targeted genomic locations.⁸⁷ While not proven in prostate cancer, it could be proposed that loss of RB1 allows for EZH2 promiscuity driving an alternate EZH2 interactome and chromatin remodeling leading to phenotypic plasticity (Fig. 1).

Using pre-clinical LP-PCa human models including NCI-H660, WCMC154 and murine DKO (Pte;Rb1 knockout) and TKO (Pten;Rb1;Trp53 knockout) cell line models, EZH2 inhibition was shown to re-sensitize prostate tumors to enzalutamide and reverse the neuroendocrine phenotype established by drivers of lineage plasticity, namely N-Myc or Rb1/Tp53 loss.^38,69,79 Studies by Ku et al. suggest a therapeutic approach to treat LP-PCa, by showing genetic and chemical inhibition of EZH2 could re-sensitize tumors with a LP-PCa phenotype to enzalutamide, thus showing EZH2 inhibition restores androgen receptor expression and sensitivity to antiandrogen therapy.³⁸ Furthermore, silencing EZH2 increased the expression of the luminal lineage markers and decreased the expression of neuroendocrine lineage markers. Another study by Kirk et al. using mouse models representing lethal PCa, including LP-PCa, showed a positive correlation between topoisomerase IIα (Top2α) and histone methyl transferase, EZH2.⁸⁸ A combination of etoposide (topoisomerase inhibitor) and EZH2 inhibitor (GSK126 or DZNep) not only significantly increased cell death in murine (TRAMP-C2 and Myc-CaP/AS) and human (LnCaP) prostate cancer cell lines but also extended time to tumor progression and increased therapy efficacy in vivo.⁸⁸ Encouraging results from aforementioned studies have led to a number of clinical trials using EZH2 inhibitors alone or in combination with ADT or immunotherapy to target refractory and metastatic CRPC tumors. Ongoing Phase I/b2 clinical trials are based on use of the EZH2 inhibitor CPI-1205 (NCT03480646) and EPZ-6438 (tazmetostat) (NCT04179864) in combination with enzalutamide or abiraterone/prednisone for treatment of mCRPC patients. Another ongoing phase I dose escalation study (NCT03460977) involves the use of EZH2 inhibitor (PF-06821497) in patients with advanced/metastatic CRPC. Other ongoing clinical trials involve the use of a monoclonal antibody that activates immune system by targeting CTLA-4 (ipilimumab) in combination with either EZH2 inhibitor CPI-1205 (NCT03525798) or DS3201 (NCT04388852) for treatment of patients with advanced solid tumors and mCRPC. Other clinical trials based on use of EZH2 inhibitors are mentioned in Table 1.

E. SWI/SNF Complex

Recently, the role of switch sucrose non-fermenting (SWI/SNF) also known as Brg/Brahma-associated factor (BAF) chromatin remodeling complex has been investigated in LP-PCa. Cyrta et al. reported expression of the SMARCA4 (BRG1) subunit (one of the two ATPase units of the mammalian SWI/SNF) was associated with LP-PCa, marked by increased NE marker expression, and decreased overall survival.⁸⁹ On the other hand, Tyler et al. demonstrated that loss of BRG1 sensitizes CCSP+ cells [club cell secretory protein-positive (CCSP+) bronchiolar epithelial club cell] within the lung in a cell-type dependent manner for malignant transformation and tumor progression, resulting in highly advanced dedifferentiated tumors and increased metastatic incidence. Furthermore, BRG1-deficient primary tumors lack lung lineage transcription factor activities and resemble a metastatic cell state. Loss of BRG1 impairs the function of all three classes of SWI/SNF complexes, resulting in decreased chromatin accessibility in lung lineage motifs and consequently accelerating tumor progression.⁹⁰ Specifically, BRG1 is known to interact with several other factors specific to neural differentiation, including the transcription factor NKX2.1 (also known as TTF-1) and the growth factor VGF (known as neurosecretory protein VGF), thereby indicating a role of BRG1 and thus the SWI/SNF complex in LP-PCa.⁸⁹ Another recent study reported MUC1-C (Mucin 1-C) binds directly to the E2F1 transcription factor and MUC1-C-E2F1 pathway induces expression of embryonic stem cell–specific BAF (esBAF) components: BRG1, ARID1A, BAF60a, BAF155, and BAF170 in LP-prostate cancer cells.⁹¹ Furthermore, this study also showed that MUC1-C promotes LP-PCa progression via integrating activation of E2F1 and esBAF along with induction of NOTCH1, NANOG, and stemness.

F. Bromodomain-Containing (BRD) Proteins

BRD proteins are chromatin readers that recognize mono-acetylated histones, trigger chromatin remodeling, and initiate transcription. Aberrant expression of bromodomains is common in cancer including metastatic prostate cancer.⁹² Kim et al. demonstrated that BRD4 (BRD family protein) and E2F1 cooperate to activate an AR-repressed, LP-PCa lineage plasticity program.⁹³ In this study the authors also showed that BET inhibition (BETi) blocks BRD4/E2F1 mediated transcription and activates AR-repressed genes, resulting in antitumor activity towards both preclinical and clinical LP-PCa tumors.⁹³ Mandigo et al. recently demonstrated that upon loss of RB1, AR complexes with E2F1 to drive a transcriptional program in prostate cancer cells allowing resistance to apoptosis.⁹⁴ Cooperation between BRD proteins and EZH2 in prostate cancer has not been deeply explored. However, studies involving other disease sites indicate that these chromatin remodeling proteins do cooperate. In bladder cancer, BRD4 was identified as an upstream regulator of EZH2 transcription in a cMYC dependent manner.⁹⁵ Pediatric high-grade gliomas, like diffuse intrinsic pontine glioma (DIPG) often demonstrate a mutant H3K27M resulting in loss of EZH2 mediated methylation and gain of H3K27ac. Further, it has demonstrated targeting EZH2 in this DIPG results in anti-tumor activity.⁹⁶ Because BRD proteins primarily interact with acetylated histones, combination of BET and EZH2 inhibition was found to be significantly superior to inhibiting each target alone.⁹⁷ As discussed above, findings from research studies demonstrate BRD4 inhibitors activate AR repressed genes. As a consequence of promising pre-clinical data BET bromodomain inhibitors are currently being tested in clinical trials either alone or in combination with ADT or immunotherapy (Table 1). A Phase I trial-NCT02705469 based on use of the BET inhibitor, ZEN003694 alone has been completed while Phase I/II trials (NCT04145375 and NCT02711956) on examining the effects of ZEN003694 in combination with enzalutamide for the treatment of mCRPC patients are ongoing. These trials conducted in mCRPC patients are promising as median radiographic progression-free survival in the overall cohort was 9 months and over 10 months in patients with prior progression on enzalutamide therapy. ZEN003694 plus enzalutamide has a favorable toxicity which may be due to accelerated production of an active metabolite ZEN-3791 as a result of pharmacokinetic interaction between enzalutamide and ZEN003694. Since combination of enzalutamide and ZEN003694 led to 2–4 fold reduction of BET target genes including MYC. These data provide clinical evidence that BET inhibition may overcome resistance mechanisms and re-sensitize patients to ADT. A very recent Phase 2 trial (NCT04471974) has just begun on determining the efficacy of ZEN003694 in combination with enzalutamide plus pembrolizumab in mCRPC patients. Another Phase I trial (NCT03150056) is aimed at use of BET inhibitor, GSK525762 in combination with ADT whereas Phase I/II trial (NCT04556617) is using PLX2853 in combination with ADT and Olaparib in mCRPC patients.

III. SUMMARY AND FUTURE DIRECTIONS

Absence of effective therapeutic options to treat LP-PCa is a major concern. Transcriptional dysregulation and chromatin remodeling stand as key mechanisms that drive LP-PCa. Recent advancements from epigenetic profiling of neuroendocrine related cancers including LP-PCa have led to a deeper understanding of epigenetic alterations and their coordination with each other to drive phenotypic plasticity via nonmutational epigenetic reprogramming. EZH2 is one of the most prominent epigenetic regulators and currently leads major interest for therapeutic development in LP-PCa. To date, preclinical studies have demonstrated the efficacy of EZH2 inhibition alone and in combination with AR directed therapies or chemotherapy to induce robust anticancer effects in LP-PCa models; however, it is yet to be determined if such combination approaches including EZH2 directed therapies will yield clinical success for patients with LP-PCa. As we move forward, we must also pay attention to alternate epigenetic targets including LSD1, SWI/SNF, BRD4, DNA methylation and fully determine their potential as therapeutic options for LP-PCa patients.

ABBREVIATIONS:

ADT

androgen deprivation therapy

AR

androgen receptor

ASCL1

achaete-scute homolog 1

BAF

Brg/Brahma-associated factor

BST2

bone marrow stromal cell antigen 2

CCSP+

club cell secretory protein-positive

CENPE

centromere binding protein

CHGA

chromogranin

CRPC

castration resistant prostate cancer

DAB2IP

disabling homolog 2-interacting protein

DIPG

intrinsic pontine glioma

DNMT

DNA methyltransferase

EMT

epithelial mesenchymal transition

EZH2

enhancer of zeste homolog 2

esBAF

embryonic stem cell specific BAF

FGFR

fibroblast growth factor receptor

FOXA1

forkhead box 1

GEMMs

genetically engineered mouse models

GINEC

gastrointestinal neuroendocrine carcinoma

IFIT3

interferon induced protein with tetratricopeptide repeats

INSM1

insulinoma-associated protein

JAK

Janus kinase

LP-PCa

lineage plastic prostate cancer

LSD1

lysine-specific demethylase 1

MCC

Merkel cell carcinoma

MMSET

multiple myeloma SET domain

NAD

nicotinamide adenine dinucleotide

NE

neuroendocrine

NEARS

neuroendocrine associated repression signature

NEUROD1

neurogenic differentiation 1

NKX2–1

NK2-homeobox1

NSD2

nuclear receptor-binding SET domain protein 2

NSE

neuron specific enolase

SRRM4

Serine/Arginine Repetitive Matrix 4

STAT

signal transducer and activator of transcription

SWI/SNF

switch sucrose non-fermenting

SYP

synaptophysin

t-NEPC

treatment-emergent neuroendocrine prostate cancer

Top2α

topoisomerase IIα

TRAMP

transgenic adenocarcinoma of the mouse prostate

TRIM22

tripartite motif containing 22

ZEB1 and ZEB2

zinc finger E-Box-binding homeobox 1 and 2