Neuroendocrine (NE) differentiation in metastatic castration-resistant prostate cancer (mCRPC) is an increasingly common clinical feature arising from cellular plasticity. We recently characterized two mCRPC phenotypes with NE features: androgen receptor (AR)-positive NE-positive amphicrine prostate cancer (AMPC) and AR-negative small cell or neuroendocrine prostate cancer (SCNPC). Here, we interrogated the regulation of RE1-silencing transcription factor (REST), a transcriptional repressor of neuronal genes, and elucidated molecular programs driving AMPC and SCNPC biology. Analysis of prostate cancer cell lines, mCRPC specimens, and LuCaP patient-derived xenograft models detected alternative splicing of REST to REST4 and attenuated REST repressor activity in AMPC and SCNPC. The REST locus was also hypermethylated and REST expression was reduced in SCNPC. While serine/arginine repetitive matrix protein 4 (SRRM4) was previously implicated in alternative splicing of REST in mCRPC, we detected SRRM3 expression in REST4-positive, SRRM4-negative AMPC, and SCNPC. In CRPC cell lines, SRRM3 induced alternative splicing of REST to REST4 and exacerbated the expression of REST-repressed genes. Furthermore, SRRM3 and SRRM4 expression defined molecular subsets of AMPC and SCNPC across species and tumor types. Two AMPC phenotypes and three SCNPC phenotypes were characterized, denoted either by REST attenuation and ASCL1 activity or by progressive activation of neuronal transcription factor programs, respectively. These results nominate SRRM3 as the principal REST splicing factor expressed in early NE differentiation and provide a framework to molecularly classify diverse NE phenotypes in mCRPC.

Significance:

This study identifies SRRM3 as a key inducer of cellular plasticity in prostate cancer with neuroendocrine features and delineates distinct neuroendocrine phenotypes to inform therapeutic development and precision medicine applications.

Prostate cancer cell growth and survival typically relies on androgen receptor (AR) activity; therefore, therapeutic interventions targeting the AR signaling axis are cornerstones for prostate cancer therapy. However, sustained AR pathway inhibition induces selective pressures on heterogeneous tumors containing cell populations with intrinsic resistance mechanisms or plastic natures. Thus, treatment-resistant metastatic castration-resistant prostate cancer (mCRPC) arises through diverse bypass mechanisms to subvert AR-directed therapies, including AR amplification, mutation or expression of AR splice variants (1–3); adoption of alternative nuclear receptors to sustain cell growth (4); and transdifferentiation to AR-null phenotypes (5–7).

Small cell or neuroendocrine prostate cancer (SCNPC) comprises a spectrum of clinical disease states that include de novo primary SCNPC to treatment-emergent metastatic SCNPC (8). While de novo SCNPC is rare, the incidence of treatment-emergent SCNPC is rising (5, 9, 10). In addition, metastatic SCNPC is typically AR-null and histologically defined by small, round tumor cells that express neuroendocrine (NE) biomarkers synaptophysin (SYP), chromogranin A (CHGA), and/or CD56/NCAM1 (8). However, the observation of novel treatment-resistant subtypes and the diverse molecular mechanisms contributing to lineage plasticity and NE transdifferentiation have led to a more complex and incomplete understanding of SCNPC biology. For example, loss of key tumor suppressors such as TP53 and RB1 (11); upregulation of pioneer and neural transcription factors such as MYCN (12), POU3F2 (13), and SOX2 (14); and epigenomic alterations associated with EZH2 (15) have all been implicated in facilitating lineage switching. However, SCNPC is not an obligate endpoint for CRPC tumor cells with TP53 and RB1 loss (16), and clinical studies have described subsets of SCNPC with retained AR expression and activity (17). In addition, the temporal and spatial interplay of identified transcription factor programs are not well understood in SCNPC and their relevance in the context of varying genomic backgrounds to cellular plasticity remains an active area of investigation.

Transcriptome profiling has facilitated the classification of cancers into subtypes with distinct functional features. We previously employed whole-transcriptome RNA sequencing (RNA-seq) and IHC studies across mCRPC biospecimens and LuCaP patient-derived xenograft (PDX) models and identified five mCRPC molecular phenotypes; (i) AR-high prostate cancer (ARPC); (ii) AR-low prostate cancer (ARLPC); (iii) amphicrine prostate cancer (AMPC) composed of tumor cells coexpressing AR and NE programs (AR+/NE+); (iv) double-negative prostate cancer (DNPC) that lacks AR and NE gene expression (AR/NE); and (v) SCNPC (AR/NE+; ref. 7).

Our group and others have highlighted the central role of AR and RE1-silencing transcription factor (REST) programs in defining CRPC molecular phenotypes (7, 18–21). REST is a master regulator of differentiation that represses neuronal programs in non-neuronal cells. Alternative splicing of REST transcripts by serine/arginine repetitive matrix protein 4 (SRRM4) has been implicated in the loss of REST repressor activity in neurogenesis and NE+ tumors (21–23). SRRM4 incorporates exon N3c into the REST transcript, leading to the expression of a truncated REST protein (REST4) that lacks the C-terminus transcriptional repressor domain and induction of neuronal gene expression (23, 24). However, the molecular underpinnings of REST regulation in clinical mCRPC specimens and the biological impacts of attenuated REST activity as it relates to novel mCRPC phenotypes are not fully understood. The work we describe herein was designed to address these deficits.

Tissue acquisition

The Institutional Review Board of the University of Washington (Seattle, WA) approved this study (protocol no. 2341). All rapid autopsy tissues were collected from patients who signed written informed consent under the aegis of the Prostate Cancer Donor Program at the University of Washington (Seattle, WA; ref. 25). Biospecimens were obtained within 8 hours of death from patients who died of metastatic CRPC as described previously (7, 26). All PDX experiments were approved by the University of Washington Institutional Animal Care and Use Committee (protocol no. 3202-01). LuCaP PDX lines were established from specimens acquired at either radical prostatectomy or at autopsy, implanted, and maintained by serial passage in immune compromised male mice as described previously (27). All mice used for the human MYCN-expressing mouse model were maintained and all procedures were performed following protocols approved by the Weill Cornell Medicine Institutional Animal Care and Use Committee (protocol no. 2008-0019).

Cell lines

All cells were maintained at 37°C in humidified Steri-Cult CO2 incubators (Thermo Fisher Scientific). C4-2B cells (gift from L. Chung, Cedars-Sinai Medical Center, Los Angeles, CA), DU145 cells (ATCC), and 22Rv1 cells (gift from S. Plymate University of Washington, Seattle, WA) were maintained in RPMI1640 Media (Gibco, Life Technologies) with 10% FBS (Atlanta Biologicals). VCaP cells (purchased directly from ATCC; CRL-2876) were maintained in DMEM (ATCC) with 10% FBS (Atlanta Biologicals). NCI-H660 (purchased directly from ATCC; CRL-5813) cells were maintained in HITES media (ATCC) supplemented with 5% FBS. All cell lines used in this project were validated through short tandem repeat analysis using ATCC reference genomes and were routinely tested for Mycoplasma using the MycoFluor Mycoplasma Detection Kit (Invitrogen).

PCR and sequencing

First-strand cDNA synthesis was performed with 1 μg of RNA using an Advantage RT-for-PCR Kit (Clontech Laboratories). PCR was performed using either Platinum SYBR Green qPCR SuperMix-UDG (Invitrogen; qPCR and Screening PCR) or HotStarTaq Plus Master Mix (Qiagen; Sequencing PCR) on a Rotor-Gene Q (Qiagen). All primer set sequences are listed in Supplementary Table S1. Touchdown PCR was used for both screening and sequencing PCR followed by agarose gel electrophoresis (Supplementary Materials and Methods).

Transient transfections and lentiviral transductions

For ectopic expression studies, cells were transfected with Tru ORF Gold expression vectors (OriGene) for SRRM3 (catalog no.: RC226066), SRRM4 (catalog no.: SC314194), pCMV6-Entry (catalog no.: ps100001), or pCMV6-Neo (catalog no.: PCMV6NEO) using the TransIT-X2 Transfection Reagent (Mirus Bio) according to manufacturer's protocols. For siRNA knockdown studies, 22Rv1 cells were transfected with ON-TARGETplus pooled siRNA (Horizon Discovery) to SRRM3 (catalog no.: L-016790-02), SRRM4 (catalog no.: L-019322-02) or nontargeting control (catalog no.: D-001810-10) using the TransIT-siQUEST Transfection Reagent (Mirus Bio) according to manufacturer's protocols. For all transient assays, cells were seeded 24 hours prior to transfections and total RNA or protein was collected 72 hours after transfection.

For stable overexpression of ASCL1, lentiviral particles were generated and C4-2B cells were transduced at a multiplicity-of-infection of 4 with either the FU-CGW empty vector backbone or FU-ASCL1-CGW vector expressing ASCL1 (AddGene). ASCL1 cDNA (GenScript) was PCR amplified, purified, and cloned into the EcoRI site of the FU-CGW lentiviral backbone by NEBuilder HiFi DNA Assembly (New England Biolabs). A total of 72 hours after lentiviral transduction, cells were trypsinized, collected as single-cell suspensions, and placed into Advanced DMEM/F12 media supplemented with B27, 2 mmol/L GlutaMAX (Gibco), and 10 ng/mL recombinant human basic FGF and EGF. Cells were cultured for 7 days prior to protein collection for immunoblot analysis.

Immunoblot analysis

Protein from LuCaP PDX models and cell lines were obtained using the Nuclear Extract Kit (Active Motif) according to manufacturer's whole-cell protein extracts protocols. Quantification of total protein was determined using the RcDc Protein Assay (Bio-Rad Laboratories) according to manufacturer's protocols. Thirty micrograms of total protein lysate were electrophoresed on 4%–15% Bis-Tris gels (Bio-Rad Laboratories) with 1× Tris/Glycine/SDS Buffer (Bio-Rad Laboratories). The proteins were transferred to nitrocellulose membranes that were blocked with 5% Blotting-Grade Blocker (Bio-Rad Laboratories) in TBS/0.1% Tween-20 and subsequently probed with primary and secondary antibodies (Supplementary Table S2). Primary antibodies (AR, KLK3, SYP, ASCL1, and REST) have been validated and published previously using our LuCaP models, CRPC cell lines and patient specimens (7, 27–29). For SRRM3 and SRRM4 antibodies, ectopic expression and siRNA studies conducted in this report provided positive and negative controls for additional antibody validation. Proteins were visualized using Clarity Western ECL Substrate (Bio-Rad Laboratories).

Data availability

Transcriptome analyses of the University of Washington Rapid Autopsy (UWRA) mCRPC, SU2C mCRPC, LuCaP PDX models and cell lines transfected with siRNAs to REST or a nontargeting control were conducted using previously published RNA-seq datasets (7, 30). TCGA PanCancer and small cell lung cancer transcriptome datasets were accessed and analyzed through cBioPortal (31–33). RNA-seq data and EPIC array methylation data generated in this report from mCRPC biospecimens and CRPC cell lines have been deposited at Gene Expression Omnibus (GEO) and can be accessed through accession number GSE158599 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE158599). The single-cell RNA-seq (scRNA-seq) data from human MYCN mouse model ventral prostates were deposited on GEO and can be accessed through accession number GSE158468 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE158468). Details of experimental procedures for other methods including, RT-PCR, IHC, RNA ISH, transcriptomic and epigenomic profiling, data analysis, and statistical analysis are included in the Supplementary Materials and Methods. In addition, all primer set sequences and antibodies used in this report are listed in Supplementary Tables S1 and S2, respectively.

Alternative splicing of REST to REST4 variants is a hallmark of AMPC and SCNPC

To determine the landscape of REST splice variants in mCRPC, we conducted PCR analysis using representative patient and LuCaP PDX specimens from normal prostate (NP; n = 5), ARPC (n = 5), AMPC (n = 5), ARLPC (n = 5), DNPC (n = 5), and SCNPC (n = 4). The molecular phenotypes of the UWRA mCRPC biospecimens and LuCaP PDX models were identified previously (7). The literature suggests that the REST transcript could have up to 45 different splice variants (variants S1-S45) that are expressed in tissue- and context-dependent circumstances (34). However, we focused our analysis on all REST splice variants that include neuron-specific exon N3c using established PCR primers and protocols (Fig. 1A; ref. 34). PCR analysis in conjunction with amplicon sequencing confirmed that REST splice variants encoding full-length REST proteins (exons 1a or 1b, 2, 3 and 4; variants S1 and S11) were detected in all specimens (Fig. 1A and B). Furthermore, REST splice variants with exon 2 skipping were also present in all specimens in conjunction with exon 1a (variant S5). While PCR analysis using primers for exon 1c (E1cF1/E4R1) detected faint bands in some mCRPC specimens, we could not confirm the presence of REST splice variants incorporating exon 1c through sequencing. In addition, neither PCR nor sequencing detected variants with exon 3 skipping that may encode REST proteins with diminished nuclear translocation (Fig. 1A and B). However, this does not exclude exon 3 skipping as a potential mechanism for REST activity attenuation in NE+ mCRPC. Importantly, we confirmed that exon N3c was present exclusively in AMPC and SCNPC specimens and that it was incorporated in multiple alternatively spliced REST transcripts that included either exon 1a or 1b, or had exon 2 skipping (variants S3, S7, and S12; Fig. 1A and B). Thus, our data suggest that alternative splicing of REST transcripts to variants encoding REST4 is a common event in mCRPC tumors with NE features. A summary of all confirmed REST splice variants identified in the 29 biospecimens are noted in Supplementary Table S3.

Figure 1.

Alternative REST splicing and REST locus hypermethylation occur in NE+ mCRPC. A, Top, REST exons with the location of PCR primers (black arrows). Bottom, summary of the identified REST splice variants in normal prostate (NP; n = 5), AR-high prostate cancer (ARPC; n = 5), AR-low prostate cancer (ARLPC; n = 5), double-negative prostate cancer (DNPC; n = 5), amphicrine prostate cancer (AMPC; n = 5), and small cell or neuroendocrine prostate cancer (SCNPC; n = 4). B, This is a composite image of PCR analyses of representative patient specimens from NP and the five mCRPC phenotypes using forward primers specific to REST exon 1a (E1aF1), exon 1b (E1bF1), exon 1c (E1cF1), and exon 2 (E2F1). All forward primers were paired with a reverse primer specific to REST exon 4 (E4R1). C, Boxplot of REST mRNA expression from the UWRA mCRPC cohort (ARPC, n = 88; AMPC, n = 14; ARLPC, n = 11; DNPC, n = 10; SCNPC, n = 14). D, Methylation analysis of the REST locus in C4–2B and NCI-H660 cells using EPIC arrays. DMR, differentially methylated region. E, COMPARE-MS heatmap of REST methylation in LuCaP PDX models. Results are expressed as Ct values of the samples of interest normalized to Ct values of the SSSI control (in vitro fully methylated male genomic DNA) and are colored according to scale. F, Plot of REST expression in LuCaP PDX models (n = 2 for each model). Black, ARPC models; gray, SCNPC models. ****, P < 0.0001; one-way ANOVA with Tukey multiple comparisons test.

Figure 1.

Alternative REST splicing and REST locus hypermethylation occur in NE+ mCRPC. A, Top, REST exons with the location of PCR primers (black arrows). Bottom, summary of the identified REST splice variants in normal prostate (NP; n = 5), AR-high prostate cancer (ARPC; n = 5), AR-low prostate cancer (ARLPC; n = 5), double-negative prostate cancer (DNPC; n = 5), amphicrine prostate cancer (AMPC; n = 5), and small cell or neuroendocrine prostate cancer (SCNPC; n = 4). B, This is a composite image of PCR analyses of representative patient specimens from NP and the five mCRPC phenotypes using forward primers specific to REST exon 1a (E1aF1), exon 1b (E1bF1), exon 1c (E1cF1), and exon 2 (E2F1). All forward primers were paired with a reverse primer specific to REST exon 4 (E4R1). C, Boxplot of REST mRNA expression from the UWRA mCRPC cohort (ARPC, n = 88; AMPC, n = 14; ARLPC, n = 11; DNPC, n = 10; SCNPC, n = 14). D, Methylation analysis of the REST locus in C4–2B and NCI-H660 cells using EPIC arrays. DMR, differentially methylated region. E, COMPARE-MS heatmap of REST methylation in LuCaP PDX models. Results are expressed as Ct values of the samples of interest normalized to Ct values of the SSSI control (in vitro fully methylated male genomic DNA) and are colored according to scale. F, Plot of REST expression in LuCaP PDX models (n = 2 for each model). Black, ARPC models; gray, SCNPC models. ****, P < 0.0001; one-way ANOVA with Tukey multiple comparisons test.

Close modal

Next, we sought to identify other potential mechanisms capable of suppressing REST activity in mCRPC. While we verified the presence of REST splice variants that include exon N3c in AMPC and SCNPC specimens, the relative abundance of N3c variants in SCNPC tumor specimens were markedly lower compared with AMPC. RNA-seq from the UWRA mCRPC cohort determined that SCNPC tumors had significantly lower REST transcript expression (P < 0.0001) compared with the other molecular phenotypes (Fig. 1C). To ascertain whether DNA methylation influenced REST and AR expression in SCNPC, we subjected the biospecimens used for PCR analysis to loci-specific methylation analysis using the EpigenDX platform. Predetermined CpGs along the AR locus and encompassing the REST transcriptional start site and 5′-untranslated region (UTR; i.e., exons 1a, 1b, and 1c) showed no significantly altered methylation changes between mCRPC phenotypes (Supplementary Tables S4 and S5). However, methylated CpGs around the translational start site within REST exon 2 demonstrated significant (P < 0.05) and progressive demethylation with loss of AR expression (Supplementary Table S5). Using an orthogonal approach, we employed EPIC array analysis to identify genome-wide methylated CpGs in ARPC C4-2B and SCNPC NCI-H660 cells. The REST locus contained a differentially methylated region (DMR) between the 5′UTR and exon 2, with increased DNA methylation detected in NCI-H660 cells (Fig. 1D). Furthermore, combination of methylated-DNA precipitation and methylation-sensitive restriction enzymes (COMPARE-MS) supported the observation of CpG hypermethylation of the REST DMR in SCNPC LuCaP PDX models compared with ARPC LuCaP PDX models (Fig. 1E). Methylation within the REST DMR associated with REST transcript expression in four of the five SCNPC LuCaP PDX models (Fig. 1F). Attenuated REST transcript expression did not appear to be correlated with methylation for LuCaP 49. Interestingly, SCNPC LuCaP 93 had low levels of DNA methylation and high expression levels of REST transcript. However, we previously confirmed that the principal REST transcripts expressed in LuCaP 93 encode REST4 and that REST activity is suppressed despite high REST transcript expression (7, 21). Together, our data suggest that attenuation of REST activity regardless of AR status, either through alternative splicing or through reduced transcript expression, is a pivotal step for NE conversion in mCRPC.

SRRM3 is expressed in AMPC and SCNPC tumors and associates with attenuated REST activity

While characterizing the AMPC LuCaP 77CR PDX model (27), we noticed that not all 77CR tumors exhibited alternative splicing of REST transcripts to include exon N3c (Fig. 2A). To scrutinize disparate 77CR tumors, we conducted immunoblot analysis of REST4-positive and REST4-negative 77CR tumor lysates (Fig. 2B). We observed a REST isoform at 120 kDa and two isoforms at 200 kDa with our REST C-terminus antibody in LuCaP 77CR tumors. Comparatively, there were decreases in all three REST isoforms expressing the C-terminus repressor domain in the REST4-positive 77CR tumor (ET 232). In addition, AR protein expression appeared similar between 77CR tumors, but SYP protein expression was drastically increased in the REST4-positive 77CR tumor, indicating loss of REST-mediated transcriptional repression. Furthermore, RNA ISH (BaseScope) of additional 77CR tumor specimens using probes specific to SRRM4 and REST exon N3c confirmed the heterogeneity of REST splicing in tumors with a subset of tumors positive for REST4 and a subset of tumors negative for REST4 (Fig. 2C). Of note, REST exon N3c readouts were detected throughout the tumor cell population in REST4-positive tumors, suggesting that alternative splicing to REST4 was adopted by the majority of tumor cells. Strikingly, BaseScope analysis also revealed that SRRM4 transcript expression was absent in REST4-positive LuCaP 77CR tumors (Fig. 2C). Moreover, BaseScope substantiated our previous observations that the SCNPC LuCaP 93 PDX model had high transcript expression of SRRM4 and REST exon N3c, and that ARPC LuCaP PDX models were negative for SRRM4 and inclusion of REST exon N3c (Supplementary Fig. S1A). IHC analysis of REST4-positive LuCaP 77CR PDX specimens determined that SYP and CHGA were uniformly and strongly expressed, whereas REST4-negative tumors expressed SYP and CHGA focally (Fig. 2D). IHC for AR and PSA/KLK3 revealed strong AR and PSA expression throughout tumor specimens irrespective of REST4 status. RNA-seq of LuCaP 77CR tumors established that the SRRM4 transcript was absent in all tumor specimens and that REST4-positive tumors had considerably higher expression of well characterized REST-repressed genes (SYP, CHGA, SNAP25, SCG3) compared with REST4-negative tumors (Fig. 2E). As expected, genes associated with AR-active tumors (AR, NKX3-1, and KLK3) were highly expressed and had similar expression levels in all 77CR tumors. Moreover, genes encoding transcription factors and drivers of the SCNPC phenotype (SOX2, NKX2-1, POU3F2, and LMO3) had low expression in all 77CR tumor specimens. Furthermore, alignment data from the 77CR RNA-seq profiles supported our previous PCR analyses and resolved REST splicing events around the 5′-UTR (exon 1a and 1b) irrespective of REST4 status but detected alternative splicing around exon N3c only in the REST4-positive tumors (Supplementary Fig. S1B). These data suggest that NE differentiation and loss of REST activity in AMPC 77CR tumors are associated with alternative splicing and that a splicing factor distinct from SRRM4 is responsible for converting REST to REST4 in AMPC 77CR tumors.

Figure 2.

REST alternative splicing occurs without SRRM4 expression in the AMPC LuCaP 77CR PDX model. A, PCR analysis of LuCaP 77CR tumor specimens (n = 5) using primers specific to REST that span exon N3c. B, Immunoblot of whole tumor extracts from 77CR tumor specimens. ACTB was used as a loading control. C, BaseScope analyses of LuCaP 77CR tumor sections using probes specific to SRRM4 and REST exon N3c (REST4). White arrows point to examples of positive chromogenic FastRED readouts. D, IHC of LuCaP 77CR tumors represented in C. Bars, 20 microns. E, RNA-seq heatmap of LuCaP 77CR specimens depicting AR (top), REST repressed genes (middle), and transcription factors associated with SCNPC (bottom). Results are expressed as log2 FPKM and are colored according to scale.

Figure 2.

REST alternative splicing occurs without SRRM4 expression in the AMPC LuCaP 77CR PDX model. A, PCR analysis of LuCaP 77CR tumor specimens (n = 5) using primers specific to REST that span exon N3c. B, Immunoblot of whole tumor extracts from 77CR tumor specimens. ACTB was used as a loading control. C, BaseScope analyses of LuCaP 77CR tumor sections using probes specific to SRRM4 and REST exon N3c (REST4). White arrows point to examples of positive chromogenic FastRED readouts. D, IHC of LuCaP 77CR tumors represented in C. Bars, 20 microns. E, RNA-seq heatmap of LuCaP 77CR specimens depicting AR (top), REST repressed genes (middle), and transcription factors associated with SCNPC (bottom). Results are expressed as log2 FPKM and are colored according to scale.

Close modal

To discover novel splicing factors contributing to the expression of REST-repressed transcripts, we conducted differential expression analysis between REST4-positive and REST4-negative 77CR biospecimens (Fig. 3A). The top upregulated differentially expressed genes in REST4-positive 77CR tumors included SYP, CHGA, and several neurogenic molecules that are known to be REST-repressed in CRPC, including CHGB, CHRNB2, VGF, UNC13A, and CPLX2 (7, 21). Notably, we identified upregulated SRRM3 expression in REST4-positive 77CR tumors. While SRRM3 has been reported to be an RNA splicing factor important for vertebrate central nervous system development (35, 36), its expression and activity in carcinogenesis and tumor progression are not defined. Thus, we used RNA-seq to examine the expression of SRRM3 and REST-repressed genes associated with SRRM3 expression in ARPC and SCNPC LuCaP PDX models (Fig. 3B). Importantly, SRRM3 expression and REST-repressed genes were strongly upregulated in SCNPC LuCaP PDX models and were absent in ARPC LuCaP PDX models. In addition, immunoblot analysis of LuCaP PDX models determined that REST4-positive 77CR and SCNPC tumor specimens displayed robust SRRM3 protein expression, whereas ARPC tumors, REST4-negative 77CR and DNPC 173.2 did not express SRRM3 protein (Fig. 3C).

Figure 3.

SRRM3 is expressed in PDX models, CRPC cell lines, and patient tumors with NE differentiation and associates with attenuated REST activity. A, RNA-seq heatmap of top differentially expressed genes in LuCaP 77CR tumors ± REST4 expression. B, RNA-seq heatmap of LuCaP PDX models. Bottom panel of genes are SRRM3/REST-associated genes. C, Immunoblot of whole tumor extracts from LuCaP PDX models. ACTB was used as a loading control. ns, nonspecific band. D, RNA-seq heatmap of CRPC cell lines. All heatmaps are shown as log2 FPKM and are colored according to scale. E, Immunoblots of whole cell extracts from C4-2B, VCaP, 22Rv1, and NCI-H660 cells. Blots were probed with indicated antibodies and ACTB was the loading control. ns, nonspecific band. F, PCR analysis of CRPC cell lines using primers specific to REST. G and H, Boxplots of SRRM3 (G) and SRRM4 (H) expression in ARPC (n = 88), AMPC (n = 14), and SCNPC (n = 14) mCRPC specimens from UWRA RNA-seq. Each point represents a single tumor. *, P < 0.05; ****, P < 0.0001; ns, not significant; one-way ANOVA with Tukey multiple comparisons test.

Figure 3.

SRRM3 is expressed in PDX models, CRPC cell lines, and patient tumors with NE differentiation and associates with attenuated REST activity. A, RNA-seq heatmap of top differentially expressed genes in LuCaP 77CR tumors ± REST4 expression. B, RNA-seq heatmap of LuCaP PDX models. Bottom panel of genes are SRRM3/REST-associated genes. C, Immunoblot of whole tumor extracts from LuCaP PDX models. ACTB was used as a loading control. ns, nonspecific band. D, RNA-seq heatmap of CRPC cell lines. All heatmaps are shown as log2 FPKM and are colored according to scale. E, Immunoblots of whole cell extracts from C4-2B, VCaP, 22Rv1, and NCI-H660 cells. Blots were probed with indicated antibodies and ACTB was the loading control. ns, nonspecific band. F, PCR analysis of CRPC cell lines using primers specific to REST. G and H, Boxplots of SRRM3 (G) and SRRM4 (H) expression in ARPC (n = 88), AMPC (n = 14), and SCNPC (n = 14) mCRPC specimens from UWRA RNA-seq. Each point represents a single tumor. *, P < 0.05; ****, P < 0.0001; ns, not significant; one-way ANOVA with Tukey multiple comparisons test.

Close modal

We also examined the expression of SRRM3 and REST-repressed genes in AR-active C4-2B, VCaP, and 22Rv1 CRPC cell lines and in AR-null SCNPC NCI-H660 cells. Transcriptome analysis determined that SRRM3 and REST-repressed transcripts were upregulated in VCaP, 22Rv1, and NCI-H660 cells relative to C4-2B cells (Fig. 3D). Of note, SRRM4 transcript expression was upregulated in 22Rv1 cells only. Immunoblot analysis determined that REST protein with intact C-terminus repressor domain was considerably reduced in VCaP, 22Rv1, and NCI-H660 cells compared with C4-2B cells (Fig. 3E). Mirroring the RNA-seq data, SRRM3 and SYP proteins were detected in VCaP, 22Rv1, and NCI-H660 cells, and SRRM4 protein was detected in 22Rv1 cells only. In addition, we established that REST splice variants with exon N3c inclusion occurred intrinsically in VCaP, 22Rv1, and NCI-H660 cells (Fig. 3F). Thus, attenuated REST activity and alternative splicing of REST to REST4 in VCaP and NCI-H660 cells occurs independent of SRRM4. Next, we examined the UWRA mCRPC RNA-seq for SRRM3 and SRRM4 expression in ARPC, AMPC, and SCNPC tumors. Although SRRM3 and SRRM4 transcripts were significantly upregulated in AMPC (P < 0.0001) and SCNPC (P < 0.0001) compared with ARPC, only SRRM4 transcript expression was significantly higher (P < 0.05) in SCNPC compared with AMPC (Fig. 3G and H). Notably, SRRM4 expression was more heterogeneous compared with SRRM3 expression in AMPC and SCNPC specimens. Transcriptome analysis of the SU2C mCRPC cohort yielded similar results for SRRM3 and SRRM4 expression, but mean SRRM3 expression was significantly higher (P < 0.05) in SCNPC compared with AMPC (Supplementary Fig. S2A and S2B). Together, these data further illustrate that VCaP and 22Rv1 cells are in vitro models of AMPC and suggest that SRRM3 is widely expressed across CRPC phenotypes with NE differentiation.

SRRM3 and SRRM4 alternatively splice REST transcripts and promote NE differentiation across tumor types

To interrogate SRRM3 function and establish REST as a direct target of SRRM3-mediated alternative splicing in CRPC, we ectopically expressed SRRM3 and SRRM4 in AR-active C4-2B and AR-null DU145 cells. PCR analysis confirmed that REST exon N3c expression was increased in both C4-2B and DU145 cells after transient transfection with either SRRM3 or SRRM4 (Fig. 4A and B). Immunoblot verified that SRRM3 and SRRM4 proteins were expressed and that both SRRM3 and SRRM4 induced SYP protein expression in C4-2B cells (Fig. 4A and B). Despite considerable SRRM3 and SRRM4 protein expression in DU145 cells, we were unable to detect corresponding increases in SYP protein. However, qPCR indicated that SRRM3 and SRRM4 ectopic expression bolstered the expression of REST-repressed genes in C4-2B cells, whereas only SYP expression was increased in DU145 cells (Supplementary Fig. S3A and B). Congruently, we transfected AMPC 22Rv1 cells with siRNAs to SRRM3 (siSRRM3), SRRM4 (siSRRM4) or a nontargeting control (siNCT) and examined impacts on REST expression and function. While ablated SRRM3 or SRRM4 protein alone increased full-length REST protein and inhibited SYP protein expression, knockdown of both SRRM3 and SRRM4 (siCombo) substantially rescued REST protein expressing the C-terminus repressor domain and markedly reduced SYP protein expression (Fig. 4C). Interestingly, increased full-length REST expression after SRRM4 knockdown associated with decreased SRRM3 protein expression, suggesting that REST activity may regulate SRRM3 in CRPC cell lines. SRRM3/4 have been suggested to be under the transcriptional control of REST in breast cancer and neurogenesis (23, 37). To determine whether REST regulates SRRM3 in CRPC, we performed transcriptome analysis of AR-expressing and AR-null cell lines transfected with siRNA to REST (siREST) or a nontargeting siRNA (siNCT). REST knockdown in all cell lines led to upregulation of SRRM3 expression and REST-repressed genes, whereas SRRM4 expression was not altered (Supplementary Fig. S3C). In search of other potential regulators of SRRM3 expression, we used the C4-2B and NCI-H660 EPIC array data to identify DMRs encompassing the SRRM3 locus. We identified hypermethylation of a DMR within the first intron of SRRM3 that was associated with SRRM3 expression in NCI-H660 cells (Supplementary Fig. S3D). In addition, COMPARE-MS analyses of ARPC and SCNPC LuCaP PDX models supported this observation and showed that hypermethylated CpGs within the DMR were associated with SRRM3 expression in SCNPC (Supplementary Fig. S3E). To further investigate this association, we performed motif and transcription factor binding site analyses using the MEME Suite TomTom tool (38) and observed that regulatory sequences associated with PRDM6, PAX5, and AR binding were significantly enriched in the SRRM3 DMR sequence.

Figure 4.

SRRM3 and SRRM4 induce REST alternative splicing to attenuate REST activity and alternatively splice the same transcripts. A and B, C4-2B and DU145 cells transfected with either SRRM3 (A), SRRM4 (B) expression vectors, or CMV negative control vectors. Top panels (above dotted lines) are representative immunoblots using indicated antibodies and bottom panels (below dotted lines) are PCR analyses using primers specific to REST exon N3c alternative splicing. C, Immunoblots of 22Rv1 cells transfected with siRNAs to SRRM3, SRRM4, SRRM3, and SRRM4 (siCombo) or nontargeting siRNA (siNCT). Blots were probed with primary antibodies as indicated. ns, nonspecific band. D and E, PCR analysis of LSD1 (D) and BHC80 (E) alternative splicing in C4-2B or DU145 cells transfected with SRRM3 expression vector or CMV control. Results show amplicons from biological duplicates and the location of primers are depicted as black arrows.

Figure 4.

SRRM3 and SRRM4 induce REST alternative splicing to attenuate REST activity and alternatively splice the same transcripts. A and B, C4-2B and DU145 cells transfected with either SRRM3 (A), SRRM4 (B) expression vectors, or CMV negative control vectors. Top panels (above dotted lines) are representative immunoblots using indicated antibodies and bottom panels (below dotted lines) are PCR analyses using primers specific to REST exon N3c alternative splicing. C, Immunoblots of 22Rv1 cells transfected with siRNAs to SRRM3, SRRM4, SRRM3, and SRRM4 (siCombo) or nontargeting siRNA (siNCT). Blots were probed with primary antibodies as indicated. ns, nonspecific band. D and E, PCR analysis of LSD1 (D) and BHC80 (E) alternative splicing in C4-2B or DU145 cells transfected with SRRM3 expression vector or CMV control. Results show amplicons from biological duplicates and the location of primers are depicted as black arrows.

Close modal

To identify other targets of SRRM3-mediated alternative splicing, we examined alternative splice variants of LSD1/KDM1A and BHC80/PHF21A that are induced by SRRM4 activity and contribute to NE conversion (39, 40). PCR analysis determined that inclusion of exon 8a in LSD1 transcripts as well as exon 14a from the mutually exclusive exons 14 and 14a in BHC80, were increased in C4-2B and DU145 cells ectopically expressing SRRM3 and SRRM4 (Fig. 4D and E; Supplementary Fig. S3F). These data demonstrate that SRRM3 and SRRM4 can have analogous functions in CRPC.

To examine the broader biological impacts associated with SRRM3 and SRRM4 expression in AR-expressing and AR-null CRPC models, we subjected C4-2B and DU145 cells ectopically expressing SRRM3 or SRRM4 to RNA-seq and gene set enrichment analysis (GSEA). Transcriptome analysis showed that either SRRM3 or SRRM4 expression in C4-2B cells resulted in significant upregulation (up > 1.5-fold, P < 0.05) of known REST-repressed genes, including SYP, SNAP25, UNC13A, CHRNB2, and SCG3 (Fig. 5A; Supplementary Fig. S4A). Moreover, there were no significant alterations in expression of AR and AR associated genes or SCNPC transcription factors SOX2, NKX2-1, POU3F2, and LMO3. Of the 146 significantly upregulated genes in C4-2B cells expressing SRRM3 and of the 107 significantly upregulated genes in C4-2B cells expressing SRRM4, there were 45 upregulated genes in common (Fig. 5B). Conversely, the DU145 cells were less amenable to SRRM3 and SRRM4 expression as few REST-repressed genes were significantly upregulated (up > 1.5-fold, P < 0.05; Fig. 5A; Supplementary Fig. S4B). Of all gene transcripts significantly upregulated (up > 1.5-fold, P < 0.05) in DU145 cells expressing SRRM3 (n = 129) or SRRM4 (n = 187), 9 genes were in common between SRRM3 and SRRM4 (Fig. 5B). While the 5 unanimously upregulated genes in both cell lines expressing SRRM3 or SRRM4 were REST-repressed genes (SYP, PAX5, UNC13A, VGF, RAB39A), CHGA was not significantly altered by SRRM3 or SRRM4 expression in either cell line, suggesting a multifactorial regulatory mechanism (Supplementary Table S6). GSEA of all significantly altered genes (FDR < 0.05) using the MSigDB C3-Transcription Factor Target database identified the REST pathway as the top significantly altered pathway in both C4-2B cells and DU145 cells expressing either SRRM3 or SRRM4, indicating that inactivation of REST repression is a primary function of SRRM3 and SRRM4 activity (Supplementary Fig. S4C). Gene ontology using the Cellular Component pathways determined that SRRM3 and SRRM4-expressing cells had increased normalized enrichment scores (NES) for pathways controlling synaptic vesicle formation, transport, and release (Fig. 5C). Thus, despite AR expression and activity in C4-2B cells, SRRM3 or SRRM4 alone is sufficient for inducing biological pathways that control neuronal secretory functions that are typical of NE+ CRPC cells. These GSEA results were recapitulated in LuCaP 77CR tumors with high SRRM3 expression, strongly supporting the physiologic relevance of our approach (Fig. 5D). Together, our data indicate that SRRM3 and SRRM4 drive conversion of ARPC to AMPC through alternative splicing and attenuation of REST and that expression of either splicing factor alone is not sufficient for complete SCNPC conversion in vitro.

Figure 5.

Ectopic expression of SRRM3 or SRRM4 significantly alters REST-regulated gene expression and promotes NE cellular plasticity in C4-2B and DU145 cells. A, RNA-seq heatmap of C4-2B and DU145 cells transfected with either SRRM3, SRRM4, or CMV expression vectors. Results are shown as log2 FPKM and are colored according to scale. B, Venn diagram showing the overlap of significantly upregulated genes (FC > 1.5; P < 0.05) between C4-2B and DU145 cells expressing SRRM3 or SRRM4 vectors compared with CMV control. C and D, GO Cellular Component summary plots of differentially expressed genes in CRPC (C) cells from A and LuCaP 77CR (D) PDX tumors with high or low SRRM3 expression. Pathways with FDR < 0.05 in at least one group are shown and arrows point to overlapping NE pathways.

Figure 5.

Ectopic expression of SRRM3 or SRRM4 significantly alters REST-regulated gene expression and promotes NE cellular plasticity in C4-2B and DU145 cells. A, RNA-seq heatmap of C4-2B and DU145 cells transfected with either SRRM3, SRRM4, or CMV expression vectors. Results are shown as log2 FPKM and are colored according to scale. B, Venn diagram showing the overlap of significantly upregulated genes (FC > 1.5; P < 0.05) between C4-2B and DU145 cells expressing SRRM3 or SRRM4 vectors compared with CMV control. C and D, GO Cellular Component summary plots of differentially expressed genes in CRPC (C) cells from A and LuCaP 77CR (D) PDX tumors with high or low SRRM3 expression. Pathways with FDR < 0.05 in at least one group are shown and arrows point to overlapping NE pathways.

Close modal

To glean further biological insights into SRRM3 expression in other tumor types, we queried The Cancer Genome Atlas (TCGA) Pan-Cancer RNA-seq datasets. Of the 32 cancer datasets, adrenocortical carcinoma (ACC), pheochromocytoma and paraganglioma (PCPG) and uveal melanoma had the highest median SRRM3 expression (Supplementary Fig. S5A). On the other hand, PCPG, low-grade gliomas (LGG), and glioblastoma multiforme (GBM) had the highest median SRRM4 expression (Supplementary Fig. S5B). Furthermore, we stratified transcriptome profiles from small cell lung cancer (SCLC) tumors based on high and low SRRM3 expression (Supplementary Fig. S6; ref. 33). Importantly, we determined that SRRM3high SCLC had significant upregulation of SRRM4 (P = 6.34 × 10–4), SYP (P = 7.80 × 10–7), CHGA (P = 1.21 × 10–7), SNAP25 (P = 1.31 × 10–7), and UNC13A (P = 2.29 × 10–7), while REST expression was significantly decreased (P = 3.14 × 10–9). Together, these data strongly suggest that SRRM3 activity and REST attenuation are key to neural crest-derived tumors and exacerbate NE biology in SCLC.

To determine whether Srrm3 and Srrm4 are expressed in murine prostate cancer cells, we employed transcriptome analysis using a genetically engineered mouse model (GEMM) expressing human MYCN (Pb-Cre+; Ptenf/f; MYCN+; Rb1f/f). RNA-seq of ventral prostate tumors revealed that Srrm3 and Srrm4 expression were significantly upregulated (P < 0.05) in poorly differentiated tumors from both castrated and intact mice compared with adenocarcinomas from intact mice (Supplementary Fig. S7). Next, we conducted scRNA-seq to more clearly identify cell populations that expressed Srrm3 and Srrm4 (Fig. 6A). The scRNA-seq of 8-week-old ventral prostates (n = 3) revealed distinct populations of MYCN-expressing cells, including a subpopulation expressing SCNPC features (Ar and MYCN+, Syp+ or ChgA+ or Insm1+; Fig. 6B; Supplementary Fig. S8A). Importantly, cells with Srrm3 or Srrm4 expression or Srrm3 and Srrm4 coexpression clustered with the SCNPC-like cell cluster (Fig. 6C–E). Differential expression analysis between Srrm3+ or Srrm4+ SCNPC-like cells and ARPC-like cells (i.e., Ar+ and Syp) and Enrichr gene ontology (GO) analysis of significantly altered genes (P < 0.05) determined that nervous system development (P = 9.37 × 10–15) was the top enriched pathway in Srrm3+ or Srrm4+ SCNPC-like cells compared with ARPC-like cells (Fig. 6F; Supplementary Table S7). In addition, AMPC-like cells (i.e., Ar+/Syp+) have been reported in CRPC GEMM models but the molecular events leading to this cellular state are not well understood (11, 41). We detected Srrm3+ and Srrm4+ cells with AMPC-like profiles in the scRNA-seq data (Supplementary Fig. S8B–S8D). Moreover, AMPC-like cells demonstrated significant (P < 0.05) upregulation of REST-repressed genes compared with ARPC-like cells (Supplementary Fig. S8E). In agreement with C4-2B and LuCaP 77CR results from Fig. 5, Enrichr GO analysis determined that chemical and synaptic transmission (P = 1.54 × 10–19) was the top enriched pathway for differentially expressed genes (P < 0.05) in AMPC-like cells compared with ARPC-like cells (Supplementary Fig. S8F; Supplementary Table S8). These data indicate that attenuation of REST activity via Srrm3 and Srrm4 expression occurs in Ar and Ar+ cells with NE differentiation in a CRPC GEMM model. In addition, the AMPC molecular phenotype is not unique to human CRPC cells.

Figure 6.

Srrm3 and Srrm4 are expressed in SCNPC-like tumor cells from a MYCN-driven CRPC GEMM model. A, t-SNE plot and clustering of combined scRNA-seq data from 8-week-old Pb-Cre+; Ptenf/f; MYCN+; Rb1f/f prostate tumors (n = 3). Data comprised of 22,461 cells. B, Identification of SCNPC-like cells that are MYCN+ and Ar as well as positive for at least one NE marker (Syp+, Chga+, Insm1+). C and D, Expression levels of Srrm3 (C) and Srrm4 (D) in scRNAseq data. Data are represented as log2 expression of the normalized read lengths and are colored according to scale. E, Identification of cells that are Srrm3 and Srrm4 double positive. F, Enrichr gene ontology analysis of the significant differentially expressed genes (P < 0.05) in SCNPC-like cells that are Srrm3+ or Srrm4+ compared with ARPC-like cells (Ar+/Syp). Results are ranked according to P value.

Figure 6.

Srrm3 and Srrm4 are expressed in SCNPC-like tumor cells from a MYCN-driven CRPC GEMM model. A, t-SNE plot and clustering of combined scRNA-seq data from 8-week-old Pb-Cre+; Ptenf/f; MYCN+; Rb1f/f prostate tumors (n = 3). Data comprised of 22,461 cells. B, Identification of SCNPC-like cells that are MYCN+ and Ar as well as positive for at least one NE marker (Syp+, Chga+, Insm1+). C and D, Expression levels of Srrm3 (C) and Srrm4 (D) in scRNAseq data. Data are represented as log2 expression of the normalized read lengths and are colored according to scale. E, Identification of cells that are Srrm3 and Srrm4 double positive. F, Enrichr gene ontology analysis of the significant differentially expressed genes (P < 0.05) in SCNPC-like cells that are Srrm3+ or Srrm4+ compared with ARPC-like cells (Ar+/Syp). Results are ranked according to P value.

Close modal

SRRM3 and SRRM4 expression delineate diverse biological classes of AMPC and SCNPC

The transcriptome analyses of the UWRA and SU2C mCRPC cohorts demonstrated that SRRM3 was upregulated in the majority of AMPCs but not all, and that SRRM4 expression was more robust in SCNPCs compared with AMPCs (Fig. 3G and H; Supplementary Fig. S2A and S2B). These findings suggest that SRRM3 and SRRM4 could stratify mCRPC biospecimens with biologically distinct NE programs. Thus, we scrutinized the SU2C mCRPC polyA RNA-seq dataset (n = 270; ref. 30) as well as transcriptomes from the UWRA mCRPC and LuCaP PDX models and focused on transcription factors driving differentiation states. First, we segregated tumor transcriptional profiles based on the expression of luminal fate transcription factors and biomarkers (AR, NKX3-1, KLK2, KLK3, TARP) and broadly expressed NE biomarkers (SYP, CHGA, CHGB, SCG3, SEZ6) and this revealed ARPC, AMPC, ARLPC, DNPC, and SCNPC molecular phenotypes across all datasets (Supplementary Fig. S9; Supplementary Fig. S10A and S10B). Next, we stratified AMPC tumors in the SU2C/UWRA datasets using SRRM3 and SRRM4 expression and identified AMPC tumors with SRRM3/4 expression and AMPC tumors without SRRM3/4 expression (Fig. 7A and B). Notably, AMPC tumors typically expressed SRRM3 transcript alone or in conjunction with SRRM4 but not SRRM4 alone (Fig. 7A and B). As expected, differential expression analysis between ARPC tumors and AMPC-SRRM3+ tumors in the SU2C dataset revealed that significantly upregulated genes (FDR < 0.05, log2FC > 1) in AMPC-SRRM3+ tumors included REST-repressed genes (ACTL6B, CHRNB2, SNAP25, UNC13A, XKR7) and broad NE biomarkers (Supplementary Fig. S11A). However, differential expression analysis in the SU2C dataset between ARPC and SRRM3/4-null AMPC tumors revealed ASCL1 as the top differentially expressed gene (FDR < 0.05, log2FC > 1; Supplementary Fig. S11B). Interestingly, the majority of REST-repressed genes associated with SRRM3 expression were not altered in AMPC-ASCL1+ (SRRM3/4-null) mCRPC tumors. To determine whether ASCL1 could drive an alternative AMPC phenotype, we stably expressed ASCL1 in C4-2B cells and examined genes associated with AMPC-ASCL1+ tumor specimens. Immunoblot analysis demonstrated that ASCL1 induced SYP and INSM1 protein expression, whereas SRRM3, SRRM4 and full-length REST proteins were not altered (Fig. 7C). Moreover, AR protein expression and activity (STEAP1) remained intact, suggesting that C4-2B cells expressing ASCL1 acquire the molecular features of AMPC-ASCL1+ patient tumors.

Figure 7.

A and B, SRRM3 and SRRM4 identify molecular subtypes of NE+ mCRPC. RNA-seq heatmaps of CRPC metastases from Stand Up to Cancer (SU2C; n = 84; A) and University of Washington Rapid Autopsy (UWRA; n = 49; B) segregated according to differential expression of transcription factors and NE biomarkers from literature. Results are expressed as log2 FPKM (expression) or as GSVA (activity scores) and are colored according to scale. Amphicrine, red; AR-low, purple; double-negative, blue; small cell or neuroendocrine, gold. C, Immunoblot of C4-2B cells transduced with constructs encoding ASCL1 or GFP (control). GAPDH was used as a loading control and blots show lysates from two independent experiments. ns, nonspecific band. D, GSEA summary of the Hallmark pathways in AMPC-SRRM3+, AMPC-ASCL1+, SCNPC-1, SCNPC-2, and SCNPC-3 compared with ARPC in the SU2C mCRPC dataset. Pathways with FDR < 0.05 in at least one group are shown. E and F, Venn diagrams of SCNPC (E) and AMPC (F) molecular phenotypes depicting unique and overlapping upregulated genes (FDR < 0.05) relative to ARPC.

Figure 7.

A and B, SRRM3 and SRRM4 identify molecular subtypes of NE+ mCRPC. RNA-seq heatmaps of CRPC metastases from Stand Up to Cancer (SU2C; n = 84; A) and University of Washington Rapid Autopsy (UWRA; n = 49; B) segregated according to differential expression of transcription factors and NE biomarkers from literature. Results are expressed as log2 FPKM (expression) or as GSVA (activity scores) and are colored according to scale. Amphicrine, red; AR-low, purple; double-negative, blue; small cell or neuroendocrine, gold. C, Immunoblot of C4-2B cells transduced with constructs encoding ASCL1 or GFP (control). GAPDH was used as a loading control and blots show lysates from two independent experiments. ns, nonspecific band. D, GSEA summary of the Hallmark pathways in AMPC-SRRM3+, AMPC-ASCL1+, SCNPC-1, SCNPC-2, and SCNPC-3 compared with ARPC in the SU2C mCRPC dataset. Pathways with FDR < 0.05 in at least one group are shown. E and F, Venn diagrams of SCNPC (E) and AMPC (F) molecular phenotypes depicting unique and overlapping upregulated genes (FDR < 0.05) relative to ARPC.

Close modal

While characterizing SRRM3 and SRRM4 expression in AMPC tumors in the SU2C and UWRA datasets, we also observed that SCNPC tumors typically expressed SRRM3, whereas SRRM4 expression was limited to a subset of SCNPC tumors and was associated with MYCN expression (Fig. 7A and B). Differential expression analysis between SCNPC tumors lacking SRRM4 expression (SCNPC-1) and ARPC in the SU2C dataset revealed that REST-repressed genes and ASCL1 were the top significantly upregulated genes (FDR < 0.05, log2FC > 1) in SCNPC-1 (Supplementary Fig. S11C). Of note, the NEUROD basic-Helix-Loop-Helix (bHLH) transcription factor family (NEUROD1, NEUROD2, NEUROD4, NEUROD6) have various functions in neuronal fate, and NeuroD1 specifically classifies a subset of SCLC tumors (42, 43). While NEUROD1 and NEUROD4 were expressed in SRRM4/MYCN expressing SCNPC tumors in the SU2C dataset, only a subset of SRRM4/MYCN tumors expressed NEUROD6 transcript (Fig. 7A and B). Differential expression analysis between ARPC and SCNPC tumors with upregulated SRRM4, MYCN, NEUROD1, and NEUROD4 (SCNPC-2) revealed significantly upregulated neural transcription factors (FDR < 0.05; log2FC > 1) such as ZNF536, ISL1, PAX6, and MYT1 L (Supplementary Fig. S11D). Conversely, differentially expressed genes between ARPC and SCNPC tumors expressing NEUROD6 (SCNPC-3) in the SU2C dataset identified BARHL1, LHX3, UNCX, POU3F3, POU4F2, POU4F3, and ATOH1 as significantly upregulated (FDR < 0.05, log2FC > 1) neural transcription factors in SCNPC-3 tumors (Supplementary Fig. S11E). Notably, POU4F3 and ATOH1 were not highly expressed in SCNPC-3 in the UWRA or LuCaP PDX datasets; however, this observation is based on a single mCRPC sample, and the PDX line was derived from this same metastasis. Nevertheless, for all three SCNPC molecular phenotypes the expression of a set of defined transcription factors were recapitulated in all three RNA-seq datasets.

Next, we examined the expression of transcription factors SOX2, PROX1, LHX2, and LMO3 that have been reported to be upregulated in SCNPC (7, 44, 45). While SOX2, PROX1, LHX2, and LMO3 were generally expressed in SCNPC tumors, SOX2 and PROX1 also demonstrated marked expression in transitioning/mixed DNPC tumors (i.e., NElow; Fig. 7A and B). Finally, we assessed the expression of neural cell surface antigens NCAM1, NRCAM, and L1CAM and determined that SCNPC generally upregulated the expression of all antigens (Fig. 7A and B). Together, these data suggest that PROX1 and SOX2 are expressed in early NE differentiation programs and further support that NCAM1 is a useful biomarker for metastatic SCNPC in CRPC (46, 47).

To associate biological programs in NE+ mCRPC tumors expressing diverse transcription factors, we conducted GSEA using all significantly altered genes (FDR < 0.05) compared with ARPC for AMPC-SRRM3+, AMPC-ASCL1+, SCNPC-1, SCNPC-2, and SCNPC-3. Analysis of the Hallmark pathways demonstrated that genes contributing to mitotic spindle, G2–M checkpoint and E2F targets were significantly enriched compared with ARPC in AMPC-SRRM3+, SCNPC-1 and SCNPC-2, but were diminished in AMPC-ASCL1+ and SCNPC-3 (Fig. 7D). These data suggest that the proliferative capacity of NE+ phenotypes may be different, but larger cohorts are needed to confirm these observations. Congruently, the LuCaP 173.1 PDX model recapitulates SCNPC-3 and tumor growth is slower when compared with other SCNPC LuCaP PDX models, noting that LuCaP 173.1 represents our only PDX model with this phenotype (Supplementary Fig. S12A and S12B). Likewise, GSEA of the GO Biological Process pathways demonstrated that the top enriched pathways for SCNPC-1 were involved in cell cycle and proliferation, whereas SCNPC-2 and SCNPC-3 were enriched for highly differentiated neuronal pathways (Supplementary Table S9). Importantly, SCNPC-3 tumors displayed enrichment for multiple GO Biological Process pathways involved in sensory neuron biology and glutamate signaling, indicating a more neuronally differentiated phenotype compared with SCNPC-1 tumors. Shared and unique upregulated genes (FDR < 0.05, log2FC > 2) in SCNPC-1, SCNPC-2, and SCNPC-3 validated the transcription factors defining the molecular subtypes (Fig. 7E; Supplementary Table S10). Conversely, GO Biological Process pathways revealed that AMPC-SRRM3+ and AMPC-ASCL1+ were both enriched for overlapping pathways involving neurotransmitter and synaptic vesicle regulation (Supplementary Table S11). However, significantly upregulated genes (FDR < 0.05, log2FC > 2) encoding secreted neurogenic molecules and synaptic vesicle proteins were shared between AMPC-SRRM3+ and AMPC-ASCL1+ (Fig. 7F; Supplementary Table S12). These data indicate that the AMPC phenotype can arise through convergent biological pathways regulated by REST or ASCL1. Moreover, SCNPC may exist as a spectrum with dedifferentiated and rapidly proliferating tumors (SCNPC-1), potentially leading to more differentiated SCNPC tumors resembling glutamatergic sensory neurons (SCNPC-3).

Widespread and long-term use of AR signaling inhibitors are changing the molecular and phenotypic landscapes of prostate cancer. In addition, cellular plasticity and the emergence of novel mCRPC phenotypes has led to new and evolving features of mCRPC biology. We and others have identified dysregulation of the REST pathway as a critical step for treatment-resistance and NE differentiation in prostate cancer (7, 18–21). However, the molecular underpinnings of REST regulation in clinical mCRPC specimens and the relationship of attenuated REST activity to the spectrum NE tumor types were not well understood. In this article, we used mCRPC biospecimens and LuCaP PDX models to characterize REST alternative splice variants across ARPC, AMPC, ARLPC, DNPC, and SCNPC phenotypes (7). Our approach identified alternative splicing of REST to REST4 and DNA methylation of the REST locus as key biological mechanisms contributing to attenuated REST activity in mCRPC tumors with NE features. In support of our data, REST methylation, reduced REST mRNA expression and the expression of REST4 splice variants are noted in SCLC (22, 48). However, SRRM4 has been reported as the principal RNA splicing factor responsible for inserting exon N3C to the REST transcript and converting REST to REST4 variants outside of the central nervous system (23, 49, 50). Importantly, we discovered SRRM3 expression in AMPC and SCNPC tumors and leveraged in vitro functional studies and transcriptome analyses to demonstrate that SRRM3 has overlapping functions with SRRM4 and alternatively splices REST to REST4. In addition, we confirmed that SRRM3 expression and activity extends across NE cancer subtypes and in a SCNPC mouse model that expresses human MYCN. Altogether, our data greatly expand the biological significance of REST attenuation in tumors with NE differentiation and identifies diverse mechanisms regulating REST activity.

A major goal of the current study was to characterize SRRM3 expression across tumor types and use upregulation of REST-repressed genes as a surrogate for SRRM3 function. This approach was supported through our in vitro experiments validating that SRRM3 and SRRM4 induce alternative REST splicing and upregulation of REST-repressed genes in CRPC. Although stable expression of SRRM4 has previously been shown to induce SYP protein expression and transdifferentiation of DU145 cells to a SCNPC-like phenotype (51), we were unable to recapitulate SCNPC progression in DU145 cells using a transient system. While we confirmed that both SRRM3 and SRRM4 induced alternative splicing of REST, LSD1, and BHC80 to neuronally expressed variants in C4-2B and DU145 cells, we speculate that epigenomic alterations in DU145 cells require further engagement for complete SCNPC transdifferentiation. Nevertheless, our analysis of transcriptome datasets in CRPC tumors and cell lines, SCLC and in a GEMM model of CRPC suggest that the vast majority of NE+ cells upregulate SRRM3 expression and attenuate REST activity to initiate cellular transformation. Indeed, SCNPC NCI-H660 cells have intrinsic REST4 expression and express SRRM3 only. Importantly, our data provide evidence that the AMPC phenotype could be a biological midpoint between the ARPC to SCNPC disease continuum and that SRRM3 is preferentially expressed in AR-active tumors and earlier in disease progression compared with SRRM4. In support of this hypothesis, we demonstrated that SRRM3 and SRRM4 expression are upregulated in neural crest cell tumors (ACC, PGPL, GBM, LGG, and uveal melanoma) in TCGA PanCancer dataset. Interestingly, the transcriptomes of the neural crest–derived tumors cluster together and a 15-gene signature common to all tumors includes SYP, CHGB, STMN3, BEX1, and EEF1A2 (52). These genes were identified as significantly upregulated in C4-2B cells expressing SRRM3/SRRM4 and through differential expression analysis in the SRRM3-expressing LuCaP 77CR tumors. In addition, alternative REST splicing and REST4 variants have been observed in neuroblastoma, pheochromocytoma, and glioma (53, 54). Thus, our data cumulatively support that de-differentiation to a neural crest cell state is likely a common event in advanced treatment-resistant tumors with the capacity for NE differentiation.

Small cell NE tumors across cancer types share common molecular features and therapeutic vulnerabilities (44, 55). However, our dissection of transcription factors and biological pathways contributing to NE differentiation suggest that a more complex NE biology should be appreciated for therapeutic considerations. Indeed, we validated VCaP and 22Rv1 cells as in vitro models of AMPC. While both cell lines express the clinically relevant NE biomarkers SYP and CHGA, alternatively splice REST to REST4 and upregulate REST repressed neurogenic molecules, these cell lines have long been considered in vitro models of ARPC. Although VCaP cells respond to AR pathway inhibitors (56), 22Rv1 cells have markedly reduced PSA expression and are refractory to enzalutamide (57). How these molecular features relate to treatment-response in the patient population will require further investigation. In parallel, transcriptome analysis using the SU2C and UWRA mCRPC specimens and the LuCaP PDX models identified molecular subsets of AMPC and SCNPC that greatly inform the biology of NE transdifferentiation. We recognize that the number of SCNPC tumors within the SU2C, UWRA, and LuCaP PDX model transcriptome datasets are relatively small and further validation of our findings will be required. In addition, the potential examination of mixed tumor cell populations inherent to bulk RNA-seq, the temporal considerations of single biopsies and the issues of multifactor regulation for some NE biomarkers (e.g., CHGA) need to be appreciated. However, the consistency of transcription factor profiles across multiple mCRPC datasets lends strength to our analysis. Principally, we characterized two AMPC phenotypes driven by either REST attenuation or ASCL1 activity and three progressively neuronal SCNPC phenotypes characterized by (i) SRRM3 and ASCL1 expression only (SCNPC-1), (ii) SRRM3, SRRM4, MYCN, and NEUROD1 expression (SCNPC-2) and (iii) SRRM3, SRRM4, MYCN and NEUROD6 expression (SCNPC-3; Supplementary Fig. S13). Congruently, molecular phenotypes of SCLC can be distinguished by ASCL1 and NEUROD1 activity and it has been shown that SCLC can progress from an ASCL1 phenotype to a NEUROD1 phenotype through MYC activity (42, 43, 58). In addition, ASCL1 activity can drive neuroendocrine reprogramming in some CRPC models (29). While SCNPC-2 tumors could be transitioning to SCNPC-3 tumors that resemble glutamatergic sensory neurons, our data suggest that the biology of SCNPC-1 tumors are strikingly different from SCNPC-2 and SCNPC-3 and could respond differently to therapies. For example, a recent clinical trial using the aurora kinase A inhibitor alisertib in SCNPC did not meet its primary endpoint of 6-month radiographic progression-free survival (59). However, the study identified a subset of responders and nominated MYCN overactivity as a biomarker for therapeutic benefit. While future testing of alisertib in stratified SCNPC preclinical models is needed, our data predict that SCNPC-1 tumors (MYCN-null) would be refractory to alisertib therapy. Thus, our profiling efforts delineate the heterogeneity observed in the SCNPC patient population and could be translated to future clinical trials for more appropriate patient selection.

In summary, our data highlight an unrecognized mechanism of ARPC to AMPC or SCNPC conversion that hinges on a SRRM3-REST regulatory axis. In addition, AMPC and SCNPC encompass multiple molecular phenotypes with convergent NE differentiation pathways. Future studies examining the pathology of novel NE+ molecular subtypes would shed further light on the clinical impacts of diverse tumor characteristics. In the meantime, SRRM3 and SRRM4 could be used as biomarkers to distinguish distinct NE+ phenotypes to aid in clinical trial design and patient tumor stratification.

J.K. Lee reports personal fees from Mission Bio and grants from Immunomedics outside the submitted work. D. Li reports grants from Department of Defense Idea Development Award, NIH, and The Pacific Northwest Prostate Cancer SPORE during the conduct of the study. J.J. Alumkal reports grants from NCI and Department of Defense during the conduct of the study; personal fees and other support from Janssen Biotech, Merck Sharp & Dohme, Dendreon, Astellas, and Bristol Myers Squibb outside the submitted work. E. Corey reports grants from NIH and CDMRP during the conduct of the study, and other support from Forma Therapeutics, Janssen Research & Development, KronosBio, Bayer Pharmaceuticals, and Arvinas outside the submitted work. P.S. Nelson reports personal fees from Janssen, Bristol Myers Squibb, Merck, and Venable outside the submitted work. No disclosures were reported by the other authors.

M.P. Labrecque: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. L.G. Brown: Investigation, writing–review and editing. I.M. Coleman: Data curation, formal analysis, investigation, visualization, writing–review and editing. B. Lakely: Investigation, methodology, writing–review and editing. N.J. Brady: Formal analysis, investigation, writing–review and editing. J.K. Lee: Formal analysis, investigation, writing–review and editing. H.M. Nguyen: Methodology, writing–review and editing. D. Li: Investigation, writing–review and editing. B. Hanratty: Formal analysis, writing–review and editing. M.C. Haffner: Formal analysis, investigation, writing–review and editing. D.S. Rickman: Formal analysis, investigation, writing–review and editing. L.D. True: Supervision, writing–review and editing. D.W. Lin: Resources, funding acquisition, writing–review and editing. H.-M. Lam: Formal analysis, investigation, writing–review and editing. J.J. Alumkal: Investigation, writing–review and editing. E. Corey: Methodology, writing–review and editing. P.S. Nelson: Conceptualization, supervision, funding acquisition, writing–review and editing. C. Morrissey: Conceptualization, resources, supervision, funding acquisition, methodology, project administration, writing–review and editing.

The authors thank the patients and their families, Celestia Higano, Evan Yu, Elahe Mostaghel, Heather Cheng, Mike Schweizer, Bruce Montgomery, Funda Vakar-Lopez, Martine Roudier and the rapid autopsy teams in the Urology and Laboratory Medicine & Pathology Departments at the University of Washington. They would also like to thank the Animal Facility and the Comparative Medicine Animal Caregivers for assistance with the LuCaP PDX work. This work was supported by a Department of Defense Idea Development Award-Partnering-PI (W81XWH-17-1-0414; W81XWH-17-1-0415 to M.P. Labrecque, L.G. Brown, I.M. Coleman, B. Lakely, H.M. Nguyen, D. Li, L.D. True, H.M. Lam, E. Corey, P.S. Nelson, and C. Morrissey), Department of Defense Idea Award (W81XWH2010405 to J.J. Alumkal), R01 CA251245 (J.J. Alumkal), NCI Drug Resistance and Sensitivity Network (P50 CA186786-07S1 to J.J. Alumkal), R01 CA234715 (I.M. Coleman, D. Li, P.S. Nelson), the Pacific Northwest Prostate Cancer SPORE (P50CA97186 to M.P. Labrecque, L.G. Brown, I.M. Coleman, B. Lakely, H.M. Nguyen, D. Li, L.D. True, D.W. Lin, H.-M. Lam, J.J. Alumkal, E. Corey, P.S. Nelson, and C. Morrissey), the Department of Defense Prostate Cancer Biorepository Network (W81XWH-14-2-0183 to L.G. Brown, B. Lakely, H.M. Nguyen, L.D. True, H.-M. Lam, E. Corey, and C. Morrissey), Scientific Computing Infrastructure at Fred Hutch funded by ORIP (S10OD028685 to I.M. Coleman, B. Hanratty), P01 CA163227 (I.M. Coleman, P.S. Nelson), P30 CA015704 (I.M. Coleman, D. Li, P.S. Nelson), U.S. Department of Defense Prostate Cancer Research Program (W81XWH-20-1-0111 to M.C. Haffner), the Safeway Foundation (M.C. Haffner), the STTR (I.M. Coleman, B. Hanratty, P.S. Nelson), Institute for Prostate Cancer Research (M.P. Labrecque, L.G. Brown, I.M. Coleman, B. Lakely, H.M. Nguyen, L.D. True, D.W. Lin, E. Corey, P.S. Nelson, and C. Morrissey), and the Richard M. LUCAS Foundation (H.M. Nguyen and E. Corey).

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.
Lallous
N
,
Volik
SV
,
Awrey
S
,
Leblanc
E
,
Tse
R
,
Murillo
J
, et al
Functional analysis of androgen receptor mutations that confer anti-androgen resistance identified in circulating cell-free DNA from prostate cancer patients
.
Genome Biol
2016
;
17
:
10
.
2.
Sharp
A
,
Coleman
I
,
Yuan
W
,
Sprenger
C
,
Dolling
D
,
Rodrigues
DN
, et al
Androgen receptor splice variant-7 expression emerges with castration resistance in prostate cancer
.
J Clin Invest
2019
;
129
:
192
208
.
3.
Takeda
DY
,
Spisak
S
,
Seo
JH
,
Bell
C
,
O'Connor
E
,
Korthauer
K
, et al
A somatically acquired enhancer of the androgen receptor is a noncoding driver in advanced prostate cancer
.
Cell
2018
;
174
:
422
32
.
4.
Puhr
M
,
Hoefer
J
,
Eigentler
A
,
Ploner
C
,
Handle
F
,
Schaefer
G
, et al
The glucocorticoid receptor is a key player for prostate cancer cell survival and a target for improved antiandrogen therapy
.
Clin Cancer Res
2018
;
24
:
927
38
.
5.
Bluemn
EG
,
Coleman
IM
,
Lucas
JM
,
Coleman
RT
,
Hernandez-Lopez
S
,
Tharakan
R
, et al
Androgen receptor pathway-independent prostate cancer is sustained through FGF signaling
.
Cancer Cell
2017
;
32
:
474
89
.
6.
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
.
7.
Labrecque
MP
,
Coleman
IM
,
Brown
LG
,
True
LD
,
Kollath
L
,
Lakely
B
, et al
Molecular profiling stratifies diverse phenotypes of treatment-refractory metastatic castration-resistant prostate cancer
.
J Clin Invest
2019
;
130
:
4492
505
.
8.
Epstein
JI
,
Amin
MB
,
Beltran
H
,
Lotan
TL
,
Mosquera
JM
,
Reuter
VE
, et al
Proposed morphologic classification of prostate cancer with neuroendocrine differentiation
.
Am J Surg Pathol
2014
;
38
:
756
67
.
9.
Vlachostergios
PJ
,
Puca
L
,
Beltran
H
. 
Emerging variants of castration-resistant prostate cancer
.
Curr Oncol Rep
2017
;
19
:
32
.
10.
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
.
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.
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
.
13.
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
.
14.
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
.
15.
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
.
16.
Nyquist
MD
,
Corella
A
,
Coleman
I
,
De Sarkar
N
,
Kaipainen
A
,
Ha
G
, et al
Combined TP53 and RB1 loss promotes prostate cancer resistance to a spectrum of therapeutics and confers vulnerability to replication stress
.
Cell Rep
2020
;
31
:
107669
.
17.
Aggarwal
R
,
Huang
J
,
Alumkal
JJ
,
Zhang
L
,
Feng
FY
,
Thomas
GV
, et al
Clinical and genomic characterization of treatment-emergent small-cell neuroendocrine prostate cancer: a multi-institutional prospective study
.
J Clin Oncol
2018
;
36
:
2492
503
.
18.
Flores-Morales
A
,
Bergmann
TB
,
Lavallee
C
,
Batth
TS
,
Lin
D
,
Lerdrup
M
, et al
Proteogenomic characterization of patient-derived xenografts highlights the role of REST in neuroendocrine differentiation of castration-resistant prostate cancer
.
Clin Cancer Res
2019
;
25
:
595
608
.
19.
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
.
20.
Svensson
C
,
Ceder
J
,
Iglesias-Gato
D
,
Chuan
YC
,
Pang
ST
,
Bjartell
A
, et al
REST mediates androgen receptor actions on gene repression and predicts early recurrence of prostate cancer
.
Nucleic Acids Res
2014
;
42
:
999
1015
.
21.
Zhang
X
,
Coleman
IM
,
Brown
LG
,
True
LD
,
Kollath
L
,
Lucas
JM
, et al
SRRM4 expression and the loss of REST activity may promote the emergence of the neuroendocrine phenotype in castration-resistant prostate cancer
.
Clin Cancer Res
2015
;
21
:
4698
708
.
22.
Coulson
JM
,
Edgson
JL
,
Woll
PJ
,
Quinn
JP
. 
A splice variant of the neuron-restrictive silencer factor repressor is expressed in small cell lung cancer: a potential role in derepression of neuroendocrine genes and a useful clinical marker
.
Cancer Res
2000
;
60
:
1840
4
.
23.
Raj
B
,
O'Hanlon
D
,
Vessey
JP
,
Pan
Q
,
Ray
D
,
Buckley
NJ
, et al
Cross-regulation between an alternative splicing activator and a transcription repressor controls neurogenesis
.
Mol Cell
2011
;
43
:
843
50
.
24.
Chen
GL
,
Miller
GM
. 
Alternative REST splicing underappreciated
.
eNeuro
2018
;
5
:
ENEURO.0034
18.2018
.
DOI: 10.1523/ENEURO.0034-18.2018
.
25.
Morrissey
C
,
Roudier
MP
,
Dowell
A
,
True
LD
,
Ketchanji
M
,
Welty
C
, et al
Effects of androgen deprivation therapy and bisphosphonate treatment on bone in patients with metastatic castration-resistant prostate cancer: results from the University of Washington Rapid Autopsy Series
.
J Bone Miner Res
2013
;
28
:
333
40
.
26.
Roudier
MP
,
True
LD
,
Higano
CS
,
Vesselle
H
,
Ellis
W
,
Lange
P
, et al
Phenotypic heterogeneity of end-stage prostate carcinoma metastatic to bone
.
Hum Pathol
2003
;
34
:
646
53
.
27.
Nguyen
HM
,
Vessella
RL
,
Morrissey
C
,
Brown
LG
,
Coleman
IM
,
Higano
CS
, et al
LuCaP prostate cancer patient-derived xenografts reflect the molecular heterogeneity of advanced disease an–d serve as models for evaluating cancer therapeutics
.
Prostate
2017
;
77
:
654
71
.
28.
Brady
L
,
Kriner
M
,
Coleman
I
,
Morrissey
C
,
Roudier
M
,
True
LD
, et al
Inter- and intra-tumor heterogeneity of metastatic prostate cancer determined by digital spatial gene expression profiling
.
Nat Commun
2021
;
12
:
1426
.
29.
DeLucia
DC
,
Cardillo
TM
,
Ang
L
,
Labrecque
MP
,
Zhang
A
,
Hopkins
JE
, et al
Regulation of CEACAM5 and therapeutic efficacy of an anti-CEACAM5-SN38 antibody-drug conjugate in neuroendocrine prostate cancer
.
Clin Cancer Res
2021
;
27
:
759
74
.
30.
Abida
W
,
Cyrta
J
,
Heller
G
,
Prandi
D
,
Armenia
J
,
Coleman
I
, et al
Genomic correlates of clinical outcome in advanced prostate cancer
.
Proc Natl Acad Sci U S A
2019
;
116
:
11428
36
.
31.
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
.
32.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
33.
George
J
,
Lim
JS
,
Jang
SJ
,
Cun
Y
,
Ozretic
L
,
Kong
G
, et al
Comprehensive genomic profiles of small cell lung cancer
.
Nature
2015
;
524
:
47
53
.
34.
Chen
GL
,
Miller
GM
. 
Extensive alternative splicing of the repressor element silencing transcription factor linked to cancer
.
PLoS One
2013
;
8
:
e62217
.
35.
Torres-Mendez
A
,
Bonnal
S
,
Marquez
Y
,
Roth
J
,
Iglesias
M
,
Permanyer
J
, et al
A novel protein domain in an ancestral splicing factor drove the evolution of neural microexons
.
Nat Ecol Evol
2019
;
3
:
691
701
.
36.
Nakano
Y
,
Wiechert
S
,
Banfi
B
. 
Overlapping activities of two neuronal splicing factors switch the GABA effect from excitatory to inhibitory by regulating REST
.
Cell Rep
2019
;
27
:
860
71
.
37.
Lee
NS
,
Evgrafov
OV
,
Souaiaia
T
,
Bonyad
A
,
Herstein
J
,
Lee
JY
, et al
Non-coding RNAs derived from an alternatively spliced REST transcript (REST-003) regulate breast cancer invasiveness
.
Sci Rep
2015
;
5
:
11207
.
38.
Gupta
S
,
Stamatoyannopoulos
JA
,
Bailey
TL
,
Noble
WS
. 
Quantifying similarity between motifs
.
Genome Biol
2007
;
8
:
R24
.
39.
Coleman
DJ
,
Sampson
DA
,
Sehrawat
A
,
Kumaraswamy
A
,
Sun
D
,
Wang
Y
, et al
Alternative splicing of LSD1+8a in neuroendocrine prostate cancer is mediated by SRRM4
.
Neoplasia
2020
;
22
:
253
62
.
40.
Li
Y
,
Xie
N
,
Chen
R
,
Lee
AR
,
Lovnicki
J
,
Morrison
EA
, et al
RNA splicing of the BHC80 gene contributes to neuroendocrine prostate cancer progression
.
Eur Urol
2019
;
76
:
157
66
.
41.
Chiaverotti
T
,
Couto
SS
,
Donjacour
A
,
Mao
JH
,
Nagase
H
,
Cardiff
RD
, et al
Dissociation of epithelial and neuroendocrine carcinoma lineages in the transgenic adenocarcinoma of mouse prostate model of prostate cancer
.
Am J Pathol
2008
;
172
:
236
46
.
42.
Rudin
CM
,
Poirier
JT
,
Byers
LA
,
Dive
C
,
Dowlati
A
,
George
J
, et al
Molecular subtypes of small cell lung cancer: a synthesis of human and mouse model data
.
Nat Rev Cancer
2019
;
19
:
289
97
.
43.
Borromeo
MD
,
Savage
TK
,
Kollipara
RK
,
He
M
,
Augustyn
A
,
Osborne
JK
, et al
ASCL1 and NEUROD1 reveal heterogeneity in pulmonary neuroendocrine tumors and regulate distinct genetic programs
.
Cell Rep
2016
;
16
:
1259
72
.
44.
Balanis
NG
,
Sheu
KM
,
Esedebe
FN
,
Patel
SJ
,
Smith
BA
,
Park
JW
, et al
Pan-cancer convergence to a small-cell neuroendocrine phenotype that shares susceptibilities with hematological malignancies
.
Cancer Cell
2019
;
36
:
17
34
.
45.
Park
JW
,
Lee
JK
,
Sheu
KM
,
Wang
L
,
Balanis
NG
,
Nguyen
K
, et al
Reprogramming normal human epithelial tissues to a common, lethal neuroendocrine cancer lineage
.
Science
2018
;
362
:
91
5
.
46.
Chen
JF
,
Yang
C
,
Sun
Y
,
Cao
D
. 
Expression of novel neuroendocrine marker insulinoma-associated protein 1 (INSM1) in genitourinary high-grade neuroendocrine carcinomas: An immunohistochemical study with specificity analysis and comparison to chromogranin, synaptophysin, and CD56
.
Pathol Res Pract
2020
;
216
:
152993
.
47.
Yao
JL
,
Madeb
R
,
Bourne
P
,
Lei
J
,
Yang
X
,
Tickoo
S
, et al
Small cell carcinoma of the prostate: an immunohistochemical study
.
Am J Surg Pathol
2006
;
30
:
705
12
.
48.
Kreisler
A
,
Strissel
PL
,
Strick
R
,
Neumann
SB
,
Schumacher
U
,
Becker
CM
. 
Regulation of the NRSF/REST gene by methylation and CREB affects the cellular phenotype of small-cell lung cancer
.
Oncogene
2010
;
29
:
5828
38
.
49.
Shimojo
M
,
Shudo
Y
,
Ikeda
M
,
Kobashi
T
,
Ito
S
. 
The small cell lung cancer-specific isoform of RE1-silencing transcription factor (REST) is regulated by neural-specific Ser/Arg repeat-related protein of 100 kDa (nSR100)
.
Mol Cancer Res
2013
;
11
:
1258
68
.
50.
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
.
51.
Lee
AR
,
Gan
Y
,
Tang
Y
,
Dong
X
. 
A novel mechanism of SRRM4 in promoting neuroendocrine prostate cancer development via a pluripotency gene network
.
EBioMedicine
2018
;
35
:
167
77
.
52.
Crona
J
,
Backman
S
,
Welin
S
,
Taieb
D
,
Hellman
P
,
Stalberg
P
, et al
RNA-sequencing analysis of adrenocortical carcinoma, pheochromocytoma and paraganglioma from a pan-cancer perspective
.
Cancers
2018
;
10
:
518
.
53.
Ren
H
,
Gao
Z
,
Wu
N
,
Zeng
L
,
Tang
X
,
Chen
X
, et al
Expression of REST4 in human gliomas in vivo and influence of pioglitazone on REST in vitro
.
Biochem Biophys Res Commun
2015
;
463
:
504
9
.
54.
Palm
K
,
Metsis
M
,
Timmusk
T
. 
Neuron-specific splicing of zinc finger transcription factor REST/NRSF/XBR is frequent in neuroblastomas and conserved in human, mouse and rat
.
Brain Res Mol Brain Res
1999
;
72
:
30
9
.
55.
Corella
AN
,
Cabiliza Ordonio
MVA
,
Coleman
I
,
Lucas
JM
,
Kaipainen
A
,
Nguyen
HM
, et al
Identification of therapeutic vulnerabilities in small-cell neuroendocrine prostate cancer
.
Clin Cancer Res
2020
;
26
:
1667
77
.
56.
Tran
C
,
Ouk
S
,
Clegg
NJ
,
Chen
Y
,
Watson
PA
,
Arora
V
, et al
Development of a second-generation antiandrogen for treatment of advanced prostate cancer
.
Science
2009
;
324
:
787
90
.
57.
Li
Y
,
Chan
SC
,
Brand
LJ
,
Hwang
TH
,
Silverstein
KA
,
Dehm
SM
. 
Androgen receptor splice variants mediate enzalutamide resistance in castration-resistant prostate cancer cell lines
.
Cancer Res
2013
;
73
:
483
9
.
58.
Mollaoglu
G
,
Guthrie
MR
,
Bohm
S
,
Bragelmann
J
,
Can
I
,
Ballieu
PM
, et al
MYC drives progression of small cell lung cancer to a variant neuroendocrine subtype with vulnerability to Aurora kinase inhibition
.
Cancer Cell
2017
;
31
:
270
85
.
59.
Beltran
H
,
Oromendia
C
,
Danila
DC
,
Montgomery
B
,
Hoimes
C
,
Szmulewitz
RZ
, et al
A phase II trial of the Aurora kinase A inhibitor alisertib for patients with castration-resistant and neuroendocrine prostate cancer: efficacy and biomarkers
.
Clin Cancer Res
2019
;
25
:
43
51
.