Prostate cancer (PCa) is a leading cause of cancer incidence amongst men in US, with an estimated 268,490 new cases in 2022. It is one of the leading causes of cancer-related deaths among males, with an estimated 34,500 deaths in 2022 (Siegel et al., 2022). Prostate cancer is a hormone-dependent cancer that relies on androgens acting via binding to its receptor, androgen receptor (AR) (Knudsen and Scher, 2009). Considering the crucial role of AR signaling, inhibition of AR signaling via ablating androgens is the goal of first line of PCa treatment, referred to as androgen deprivation therapy (ADT). ADT results in cancer regression initially. However, in a significant fraction of patients, 2–3 years after ADT, the disease progresses to castration-resistant prostate cancer (CRPC) (Shen and Abate-Shen, 2010). The treatment options for CRPC are limited though over last several years, several new agents have been introduced in the clinics for treating non-metastatic and metastatic CRPC such as second generation of AR pathway inhibitors (API) enzalutamide (MDV3100/ENZ) and abiraterone (ABI) (Knudsen and Scher, 2009; Scher et al., 2012). Though these agents have contributed to better survival rates, drug resistance is a major clinical challenge. Drug resistance results from heterogeneous molecular mechanisms such as AR bypass signaling or complete AR independence (Watson et al., 2015; Culig, 2017). A subset of API-resistant tumors undergo a reversible lineage trans-differentiation known as neuroendocrine differentiation (NED), that results in altered expression of lineage markers. As a result of NED, PCa cells show decreased expression of AR and increased expression of neuroendocrine (NE) lineage markers including enolase 2 (ENO2), chromogranin A (CHGA) and synaptophysin (SYP) (Aggarwal et al., 2014; Aggarwal and Small, 2014). These PCa variants, referred to as therapy-induced neuroendocrine prostate cancer (NEPC), are highly aggressive, often metastasizing to soft tissues such as liver, lung and central nervous system despite low serum PSA levels relative to disease burden (Aggarwal et al., 2014). Therefore, NEPC is associated with poor survival rates (Aggarwal et al., 2014). Therapy-induced NEPC may encompass a spectrum of histological states ranging from adenocarcinomas with mixed NE histology to pure small cell neuroendocrine carcinoma (SCNC), which is similar to small cell cancers of other organs such as small cell lung cancer (SCLC) (Epstein et al., 2014). Normal human prostate can be reprogrammed to SCNC using a common set of defined oncogenic drivers, namely dominant negative p53 (TP53DN), myrAkt1, RB1-shRNA, c-Myc and Bcl-2 (Park et al., 2018). NEPC, like SCLC and other small cell neuroendocrine cancers (SCNC) exhibit common morphological and histological features such as high nuclear to cytoplasm ratios, granular chromatin and frequent mitotic figures (Klimstra et al., 2015). NEPC along with other SCNC tumors share genome-wide expression, methylation and copy number expression patterns and have common vulnerabilities that were reported to be similar to that of hematological malignancies (Balanis et al., 2019).

It has been realized that therapy-induced NEPC represents a continuum of treatment-induced changes at molecular level resulting from a series of genetic and epigenetic alterations (Beltran et al., 2011; Lotan et al., 2011; Beltran et al., 2016; Dardenne et al., 2016; Maina et al., 2016). Analyses show that PCa NE states are derived via clonal evolution from CRPC-adenocarcinomas (Beltran et al., 2016). The key genetic events driving this transition include loss of the tumor suppressors retinoblastoma (RB1) and mutation or loss of tumor protein 53 (TP53) (Tan et al., 2014; Beltran et al., 2016). Mouse models support that dual loss of TP53 and RB1 are important steps in the development of poorly differentiated NE tumors of the prostate (Zhou et al., 2006). Normal human prostate epithelial cells could be transformed into SCNC via expression of RB1 shRNA and dominant negative p53 (Park et al., 2018). In addition, phosphatase and tensin homolog (PTEN) loss (Beltran et al., 2011; Lotan et al., 2011; Beltran et al., 2016; Dardenne et al., 2016; Maina et al., 2016) and frequent TMPRSS2-ERG gene rearrangements (Lotan et al., 2011) have been reported in NEPC. Furthermore, EZH2 overexpression and amplifications of NMYC and Aurora Kinase A (AURKA) (Beltran et al., 2011; Beltran et al., 2016; Dardenne et al., 2016; Maina et al., 2016) are cardinal alterations that have been associated with NEPC. AURKA is a cell cycle kinase that stabilizes N-Myc oncoprotein and prevents N-Myc degradation (Beltran et al., 2011; Dardenne et al., 2016; Lee et al., 2016). The overall mutational spectrum of NEPC has been reported to be similar to that of adenocarcinomas (Beltran et al., 2016; Davies et al., 2020). The lack of unique genomic alterations in NEPC as compared to adenocarcinomas imply an important role of epigenetic mechanisms (Davies et al., 2020) Studies suggest that epigenetic mechanisms, including DNA methylation and histone methylation alongwith transcriptional regulation play a huge role in the evolution of NEPC states. It is recognized that as tumors progress towards NEPC states, there is an increase in activity of stem-cell associated transcriptional programs (Smith et al., 2015; Lee et al., 2018; Smith et al., 2018). Recent years have seen a spurt in research delineating the epigenetic and transcriptional mechanisms underlying emergence of PCa NED states via stem-like states. In this article, we review the role of various transcription factors (TF) and chromatin modifiers that play a role in lineage reprogramming of prostate adenocarcinomas to NEPC under the selective pressure of various AR-targeted therapies (Figure 1). As studies shed light on the mechanistic basis of neuroendocrine differentiation in PCa, it is being recognized that these states are complex and heterogeneous. Based on the expression of TFs, the heterogeneity underlying these states is being dissected and has led to stratification of NEPC tumors. We highlight this stratification in this article. With the increased understanding, novel drugs/therapies are being tested in NEPC models. Here we review the current state of these therapies.

It has been recognized that AR activity is varied in metastatic CRPC and can stratify distinct phenotypes (Labrecque et al., 2019). Five different molecular phenotypes have been reported: 1) AR high CRPC (ARPC); 2) AR low CRPC (ARLPC); 3) amphicrine PC (AMPC) that comprises of tumor cells co-expressing AR and NE gene programs; 4) SCNC that comprises of tumor cells expressing NE genes and negative for AR expression; 5) double negative PCa (DNPC) that is AR and NE double null (Labrecque et al., 2019). In SCNC, canonical AR signaling is typically lost/decreased (Beltran et al., 2016) though clinical studies have described SCNC subsets with retained AR expression and activity (Aggarwal et al., 2018). Though end state SCNC lack AR, more than 50% of treatment resistant PCa with NE features retain nuclear AR without activation of canonical AR signaling (Aggarwal et al., 2018; Labrecque et al., 2019; Alumkal et al., 2020). A recent study showed elegantly that AR cistrome undergoes alteration upon lineage reprogramming as a result of androgen deprivation (Davies et al., 2021). They reported that AR activity was maintained as prostate tumors adopt alternative NE lineage, with changes in chromatin architecture guiding AR transcriptional rerouting. AR cooperates with EZH2 to regulate lineage plasticity, with AR and EZH2 co-occupying the reprogrammed AR cistrome to transcriptionally modulate stem cell and neuronal gene networks (Davies et al., 2021) (Figure 1). In another study, AR has been shown to physically interact with Kruppel-like factor (KLF5) on chromatin sites upon endocrine therapy for CRPC. KLF5 is induced upon therapy in CRPC patients, binding to chromatin sites with AR co-occupying chromatin sites that drive opposing transcriptional programs. KLF5 is a stem cell transcription factor that promotes basal epithelial cell phenotypes concomitant with anchorage-independent growth and increased cell migration (Che et al., 2021). Furthermore, AR activity can be directly regulated by epigenetic modifiers such as histone deacetylase SIRT1 (Davies et al., 2020) that deacetylates AR and thereby prevents association with p300. SIRT1 is upregulated in NEPC and increased SIRT1 expression induces PCa NED by activating the Akt pathway (Ruan et al., 2018). In addition, the interaction between Akt and SIRT1 is independent of N-Myc and can drive NEPC development even when N-Myc is blocked (Ruan et al., 2018). Also, AR can directly recruit histone modifiers that influence chromatin states and gene expression (Davies et al., 2020). LSD1, H3-K4 demethylase is an important AR regulator that causes suppression of canonical AR transcriptional activity (Davies et al., 2020). LSD1 undergoes epigenetic reprogramming in CRPC that activates a subset of cell cycle genes, including CENPE, a centromere binding protein and mitotic kinesin. LSD1/AR binding and transactivation get reprogrammed by RB1 loss (Liang et al., 2017; Davies et al., 2020). These studies point to the crucial role of AR in promoting lineage reprogramming associated with PCa NE states.

Enhancer of zeste homolog 2 (EZH2), the catalytic subunit of Polycomb repressive complex 2 (PRC2), has been reported to be overexpressed in NEPC (11–13). EZH2 mediates the deposition of repressive histone modification, trimethylation of histone H3 at lysine 27 (H3K27m3) that represses lineage-specifying factors, thereby maintaining pluripotency (Lee et al., 2006) (Figure 1). It has been shown that EZH2 cooperates with oncogenic N-Myc to drive transcriptional programs leading to the evolution of PCa NE states (Dardenne et al., 2016). EZH2 is required for maintenance of bivalent chromatin states (with repressive as well as activating histone modifications) at N-Myc bound neuronal lineage-associated gene promoters. EZH2 knockdown leads to de-enrichment of neuronal-associated pathways in NEPC organoids (Berger et al., 2019). EZH2 depresses the TF SOX2 as a consequence of the functional loss of RB1 (Kareta et al., 2015; Ku et al., 2017). An interesting study shows the role of EZH2 in repressing microRNA, miR-708 via binding to its promoter region (Shan et al., 2019). EZH2 is activated by transcription factor 4 (TCF4), a key component of Wnt/β-catenin signaling. EZH2-mediated repression of miR-708 leads to de-repression of neuronatin (Ryu et al., 2013) and stem cell factor CD44 (Saini et al., 2012). Furthermore, EZH2 activity is coupled to cAMP-response element binding protein (CREB) activation in prostate cancer (Zhang et al., 2018). Activation of CREB/EZH2 axis facilitates epigenetic repression of anti-angiogenic Thrombospondin 1 (TSP1), leading to angiogenesis and NE induction in PCa xenografts (Zhang et al., 2018). In agreement with a crucial role of EZH2 in PCa NED, EZH2 inhibitors are potential agents for NEPC treatment and have been shown to re-sensitize tumors to AR-signaling inhibitors in CRPC (Clermont et al., 2015; Beltran et al., 2016).

The cellular heterogeneity of prostate tumors progressing from AR-positive adenocarcinoma to AR negative NEPC allows stratification of this disease into various subtypes. Recent studies show that evolution to NE states involve an inherent hierarchical TF network. Single cell studies on CRPC and NEPC has shown that NEPC is driven by constitutive regulation of pioneering transcription factors ASCL1 and FOXA1 alongwith selective regulation of NKX2-2 or POUF3F2 and SOX2. This expression profile could categorize NEPC into subtypes based on transcriptomic mechanisms with NE1 expressing NKX2-2 and NE2 subtype expressing POU3F2 and SOX2 (Wang et al., 2022).

Cejas et al. (2021) reported that NEPC can be stratified on the basis of expression of the neuronal transcription factors ASCL1 and NEUROD1 , like SCLC (Rudin et al., 2019). ChIP-Seq analyses of these two TFs in NEPC models showed that there are thousands of highly conserved binding sites with both overlapping and differential binding sites. De novo motif analyses of these binding sites showed that these binding sites possess the respective consensus motifs for ASCL1 and NEUROD1. Further, ASCL1 binding motif was found to be enriched with motif for NKX2 (Cejas et al., 2021), in keeping with reports that NKX2-1 TF is 16-fold highly expressed in the ASCL1 subtype of NE tumors (Borromeo et al., 2016). NEUROD1 binding motif was enriched for EBF and LHX motifs corresponding to the higher expression of TFs EBF and LHX in NEUROD1 subtype. Gene set enrichment analyses to identify pathways represented in ASCL1 and NEUROD1 subtypes showed that ASCL1 subtype was enriched in GO pathways of response to cytokines while NEUROD1 subtype showed enrichment for brain development pathways (Cejas et al., 2021). While the NE PDX models show a mutually exclusive expression of ASCL1 and NEUROD1, human NE tumors were found to exhibit a complex tumor structure with subtypes co-existing as separate sub-populations within the same tumor (Cejas et al., 2021).

Labreque et al. stratified NEPC based on expression of RNA splicing factors serine/arginine repetitive matrix protein 4 (SRRM4) and SRRM3 (Labrecque et al., 2021b) that control alternative splicing of REST. REST is a repressor of neuronal genes that is downregulated in NEPC (Flores-Morales et al., 2019). SRRM4 is involved in alternative splicing of REST wherein it incorporates exon N3c into the REST transcript leading to the expression of truncated REST protein (REST4) that lacks the C-terminal repressor domain that leads to loss of its repressor activity and induction of neuronal genes (Shimojo et al., 2013; Li et al., 2017). SRRM3 have overlapping functions with SRRM4 and mediates alternative splicing of REST to REST4 (Labrecque et al., 2021b). Based on SRRM4 and SRRM3 expression, three molecular subtypes of SCNC were characterized that were found to be progressively neuronal: 1) SCNC-1: These tumors express SRRM3, driving alternative splicing of REST to REST4. These SCNC tumors co-express ASCL1; 2) SCNC-2: This subtype of SCNC express SRRM3 and SRRM4 alongwith neuronal MYCN, NEUROD1 and NEUROD4 expression. These tumors express significantly upregulated transcription factors such as ZNF536, ISL1, PAX6 and MYT1L; 3) SCNC-3: A subtype of SCNC that express SRRM3, SRRM4 accompanied with MYCN and NEUROD6. The neural transcription factors upregulated in SCNC-3 tumors included BARHL1, LHX3, UNCX, POU3F3, POU4F2, POU4F3 and ATOH1. This study suggests that the biology of SCNC-1 is strikingly different from that of SCNC-2 and -3 and is expected to respond differently to therapies. Since SCNC-1 is N-Myc null, it is not expected to respond to AURKA inhibitor Alisertib that acts by disrupting AURKA-N-Myc interaction (Beltran et al., 2019). Further, SCNC-2 tumors could represent tumors transitioning to SCNC-3 (Ku et al., 2017). These studies highlight the heterogeneity of NEPC and suggest that therapeutic strategies need to be designed accordingly.

Currently, the standard of care for NEPC is platinum based drugs that show a short duration of response and are administered to all NEPC patients (Aparicio et al., 2013). With an increased understanding of the epigenetic basis of NEPC, therapeutic startegies targeting epigenetic mechanisms are being tested for targeting NEPC effectively (Table 1). Table 2 lists the clinical trials in NEPC focused on TFs, chromatin modifiers and other proteins. EZH2 inhibitors are potential agents for NEPC treatment and have shown promise in pre-clinical studies in reversing the NE phenotype established by N-Myc (Dardenne et al., 2016; Berger et al., 2019) or RB1/TP53 loss. EZH2 inhibitors PF-06821,497 and CPI-1205 are currently being tested in clinical trials in CRPC (NCT03460977 and NCT03480646, respectively) (Davies et al., 2020). CPI-1205 is being tested in combination with enzalutamide or abiraterone/prednisone and has shown promising activity (Davies et al., 2020). Considering preclinical evidence for EZH2 inhibition leading to inhibition of NE differentiation, EZH2 is an important target for NEPC (Ku et al., 2017). These inhibitors have been shown to re-sensitize tumors to AR-signaling inhibitors in CRPC (Clermont et al., 2015; Beltran et al., 2016).

Targeting other epigenetic regulators such as BET family are also being investigated in advanced CRPC (Asangani et al., 2014) and may have rationale in NEPC (Beltran and Demichelis, 2021; Conteduca et al., 2021). BET inhibition using compound JQ1 disrupts the recruitment of AR to target gene sites (Asangani et al., 2014) BET inhibition blocks E2F1/BRD4-regulated program and decreases growth of NEPC tumor models. Further, a subset of t-NEPC patient tumors with high activity of this program showed decreased tumor growth in a BETi clinical trial (Kim et al., 2021). Aggarwal et al. (2020) tested the activity of BETi ZEN-3694 in a phase IIb/IIa study in metastatic CRPC patients with resistance to AR signaling pathway inhibitors and reported acceptable tolerability and potential efficacy in these patients. Importantly, tumors with low baseline AR activity appeared to derive greater benefit than patients with high baseline AR activity (Aggarwal et al., 2020).

Disruption of the molecular interaction between AURKA and N-Myc by Alisertib (MLN8237) has been examined as a therapeutic modality for NEPC (Beltran et al., 2019). A phase-II clinical trial of Alisertib reported modest clinical benefit of Alisertib in NEPC patients while a second phase I/II trial in combination with abiraterone in CRPC-NE had to be terminated due to toxicity and non-significant clinical benefit (Davies et al., 2020).

Targeting of TFs that mediate transcriptional changes during lineage reprogramming such as BRN2, BRN4, MYCN, ASCL1, FOXA1 and ONECUT2 can be exploited as a therpeutic strategy for NEPC. However, TFs are challenging to target directly (Beltran and Demichelis, 2021). BRN2 inhibition via siRNA or CRISPR/Cas9 or small molecule inhibitor was found to reduce proliferation in NEPC cell lines (Daksh et al., 2019). CSRM617 is a small molecular ONECUT2 inhibitor developed on the basis of a three-dimensional model of ONECUT2. This is found effective in reducing proliferation and tumor volume by inducing apoptosis of cancer cells with high ONECUT2 expression (Cully, 2018).

With the recent studies highlighting the enormous heterogeneity of PCa NE states and stratification based on expression of distinct TFs, the transcriptional programs of these subtypes are likely distinct. This emphasizes the requirement for specific therapeutic strategies focused on targeting distinct master regulators of each subtype for effective therapy. Considering the upregulation of Delta-like protein 3 (DLL3) in NEPC cases as compared to CRPC-adenocarcinomas, a humanized DLL3 antibody has been exploited for therapeutic targeting of NEPC (2019, Puca et al., 2019; Thoma, 2019). This therapeutic modality is likely to be effective in the ASCL1 subtype of NEPC (Cejas et al., 2021). AURKA inhibition may be more efficacious in the NEUROD1 subtype, similar to the case in SCLC (Mollaoglu et al., 2017). Therapies targeting CEACAM5 such as carcinoembryonic antigen-related cell adhesion molecule CEACAM5 antibody-drug conjugate (DeLucia et al., 2021) is likely to be efficacious in ASCL1 subtype of NEPC since this subtype was reported to be enriched in CEACAM 1,5,6,7 (Cejas et al., 2021). Effective therapeutic strategies need to be designed that would target the different subpopulations to avoid emergence of outgrowth of the resistant subpopulations (Cejas et al., 2021). There is a need to develop biomarker-driven treatment strategies for NEPC. Furthermore, owing to inter and intra-patient heterogeneity, it will be critical to gain better insights into the underlying biology of these tumors and then design combinatorial therapeutic strategies targeting factors important in specific tumor subsets (Labrecque et al., 2021a). Though progress is being made in this direction, we still have a long way to go before these therapies can be effectively implemented in the clinic.

In conclusion, transcription factors and chromatin modifiers play an integral role in driving neuroendocrine trans-differentiation in prostate cancer. Pioneer transcription factors such as ASCL1, FOXA1 and SOX2 can engage on their respective target sites on nucleosomal DNA and initiate transcriptional regulation at sites of closed chromatin. Further, interplay between transcription factors such as AR, FOXA1, BRN2 and BRN4 and epigenetic modifiers such as EZH2 drive chromatin states underlying lineage reprogramming in PCa. The complexities underlying emergence and maintenance of PCa NE states is being recognized and deciphered. With an increased understanding of the temporal and spatial interplay of transcription factors and their associated gene expression programs in PCa NED, better therapeutic strategies can be designed for targeting NEPC. Though considerable progress have been made in this direction in last few years, we are still far from understanding the complexy regulatory interplay. Further, it is important to unravel the interplay between these TFs, epigenetic modifiers and tumor microenvironment to gain a holistic understanding of the disease that would be required for devising effective therapeutic modalities.