NXP800 Activates the Unfolded Protein Response, Altering AR and E2F Function to Impact Castration-Resistant Prostate Cancer Growth

Prostate cancer is the most common cancer and the second highest cause of cancer-related death in men in the Western world and is increasing in incidence (1). The androgen receptor (AR) remains the main oncogenic driver and a major therapeutic target for advanced prostate cancer (2–5). Despite the development of multiple AR signaling inhibitors (ARSI, including abiraterone, enzalutamide, darolutamide, and apalutamide) that have improved the clinical outcome for patients with advanced disease, treatment resistance is inevitable and advanced prostate cancer remains fatal (5). The development of treatment resistance is commonly associated with reactivation of AR signaling which, in part, is driven by AR amplifications, AR-activating point mutations, and constitutively active AR splice variants, of which AR splice variant-7 (AR-V7) is the most clinically relevant (6–14). Currently approved therapies, and many of those in clinical development, elicit their mechanism of action through the AR ligand–binding domain and therefore will have limited activity against these mechanisms of resistance (5–16). Therefore, the development of innovative therapeutic strategies that abrogate persistent AR signaling through novel mechanisms of action is an urgent unmet clinical need in prostate cancer medicine.

One attractive strategy is to target co-regulators, such as the heat shock proteins (HSP), critical to AR activity (17–28). HSPs are a family of molecular chaperones that regulate protein homeostasis in response to cellular stress (29). Upon exposure to cellular stress, the transcription factor heat shock factor 1 (HSF1) upregulates the expression of multiple HSPs (including HSP90, HSP72, and HSP27) to protect the cell from proteotoxic stress (29–33). Importantly, HSP90, HSP72, and HSP27 have all been implicated in fueling AR signaling in prostate cancer (17–27). First, blocking HSP90 function in prostate cancer decreases AR nuclear translocation and enhances its degradation, leading to decreased transcriptional activity and proliferation (22–26). In addition, HSP90 inhibition decreases AR-V7 expression by regulating RNA processing, inhibiting the growth of treatment-resistant prostate cancer models (18). Second, HSP72 is critical for AR and AR-V7 stability, with HSP72 inhibition decreasing the growth of enzalutamide-resistant prostate cancer models (17, 19, 27, 28). Third, inhibition of HSP27 leads to AR and AR-V7 degradation with associated growth inhibition of treatment-resistant prostate cancer models (20, 21). Taken together, targeting the HSP family of molecular chaperones is a valid therapeutic approach to overcome persistent AR signaling in advanced prostate cancer, although concerns remain about the therapeutic window and tolerability of such HSP-targeting strategies (30, 34). Nevertheless, the oral HSP90 inhibitor pimitespib has been approved in Japan for gastrointestinal stromal tumor (35).

The pursuit of developing novel approaches targeting HSPs has resulted in the identification of inhibitors of the HSF1 pathway as an alternative therapeutic strategy (30, 36, 37). However, the majority of these inhibitors have not been optimized for selectivity or drug-like pharmaceutical properties, have limited information on potential mechanism of action, and are not at the clinical stage of development (36, 37). Therefore, although inhibition of the HSF1 pathway has been investigated in advanced prostate cancer models with promising antitumor activity, clinical translation remains challenging (38, 39). By contrast, CCT251236, which we partially optimized from a compound that we identified in a cell-based assay for inhibition of HSF1-mediated HSP72 induction following pharmacologic inhibition of HSP90, suppresses HSP72 and HSP27 transcription and shows promising activity in models of human multiple myeloma and ovarian adenocarcinoma cell line–derived xenografts (40, 41). Further multiparameter pharmaceutical optimization has led to our design of NXP800 (formerly CCT361814), which is currently undergoing clinical evaluation (NCT05226507), providing an opportunity to clinically translate this therapeutic strategy (42, 43).

Herein we present the first evaluation of NXP800 in models of advanced prostate cancer. We report that a higher Gene Ontology (GO) cellular response to heat gene expression signature in advanced prostate cancer tissue biopsies associates with increased AR signaling and worse overall survival (OS). In addition, we demonstrate that NXP800 reduces HSP72 expression and decreases AR signaling, reduces AR and AR-V7 protein expression as well as their DNA binding and transactivation, and also inhibits the growth of treatment-resistant prostate cancer models. Moreover, we report that NXP800 inhibits the growth of AR-independent prostate cancer models, indicating that its antitumor activity is not only mediated through abrogation of AR signaling. Consistent with this, we demonstrate that NXP800 activates the unfolded protein response (UPR) and inhibits key signaling pathways, including AR- and E2F-mediated transcription, which are implicated in the development and progression of castration-resistant prostate cancer (CRPC), thus providing further insights into the mechanism of NXP800-mediated antitumor activity. Finally, these mechanistic and phenotypic findings were confirmed using an NXP800-resistant model and an in vivo model of CRPC.

Taken together, our findings show that NXP800 demonstrates antitumor activity in treatment-resistant prostate cancer models, acting through a novel mechanism of action that has the potential to provide a new therapeutic strategy for hard-to-treat prostate cancer tumors with persistent AR signaling and high E2F activity, supporting its consideration for prostate cancer–specific clinical development.

All cell lines used in this study were grown in the suppliers’ recommended media at 37°C in 5% CO₂ and are detailed in Supplementary Table S1. Parental cell lines were tested for Mycoplasma using the VenorGeM OneStep PCR Kit (Cambio) and short tandem repeat profiled using the cell authentication service by Eurofins Medigenomix. All compounds used are detailed in Supplementary Table S2. These included the HSF1 pathway inhibitor clinical candidate NXP800 (Supplementary Fig. S1A) and the closely matched regioisomer CCT365248 used as an inactive control compound (Supplementary Fig. S1B), which were both synthesized in the Centre for Cancer Drug Discovery, Division of Cancer Therapeutics, The Institute of Cancer Research (43).

The plasmids used in this study were the pReceiver-M13 vector carrying a C-terminal fusion FLAG-tag (GeneCopoeia) containing either an empty cassette as control (CONTROL-FLAG) or full-length AR (AR-FL, AR-FL-FLAG). An AR-V7 C-terminal fusion FLAG-tag (AR-V7-FLAG) was generated from the corresponding AR-FL-FLAG plasmid through restriction enzyme digest (XhoI, BSTE1I; New England Biolabs) and ligation techniques as previously described (44). The resulting plasmids were verified by sequencing (Beckman Coulter Genomics). The PSA-ARE3-luciferase reporter plasmid has been previously described (45).

PC3 cells were transfected with 0.5 μg/mL CONTROL-FLAG, AR-FL-FLAG, or AR-V7-FLAG using X-tremeGENE (Roche) as per the manufacturer’s recommendations. After 48 hours, whole-cell lysates were obtained using Pierce Immunoprecipitation (IP) Lysis Buffer (Thermo Fisher Scientific). Protein extracts (about 1.5 mg of total protein) were precleared by protein A/G Plus-Agarose (Thermo Fisher Scientific), and immune complexes were collected by incubation with 1 μg FLAG antibody (Thermo Fisher Scientific, Cat. #MA1-91878, RRID: AB_1957945) and protein A/G Plus-Agarose beads overnight at 4°C. Bead/sample complexes were collected by centrifugation and washed first in standard lysis buffer and then in 50 mmol/L triethylammonium bicarbonate solution (Sigma-Aldrich). Bound proteins were then eluted using 5% formic acid and then dried in vacuo without heating. Proteins were thoroughly redissolved in 5% acetonitrile/50 mmol/L triethylammonium bicarbonate solution before being reduced with 5 mmol/L tris(2-carboxyethyl)phosphine. Free cysteines were alkylated with 10 mmol/L 2-choloroacetamide and digested with trypsin (Promega). The digestion was quenched with neat formic acid after 4 hours. LC/MS analysis was performed on an LTQ Orbitrap Velos (Thermo Fisher Scientific) with collision-induced dissociation (CID) fragmentation and ion trap detection. Tandem mass spectra were analyzed with Mascot (version 2.3.02; RRID: SCR_014322) for peptide and protein identification using a Swiss-Prot FASTA file containing 20,305 Homo sapiens entries. Mascot was searched with a fragment ion mass tolerance of 0.6 Da and a parent ion tolerance of 5 ppm. Gln->pyro-Glu of the n-terminus, oxidation of methionine, acetylation of the n-terminus, carbamidomethylation of cysteine, and phosphorylation of serine, threonine, and tyrosine were specified as variable modifications. Scaffold (Proteome Software Inc.) was used to validate MS/MS-based peptide and protein identifications. Peptide identifications were accepted if they could be established at >95.0% probability by the PeptideProphet algorithm (46). Protein identifications were accepted if they contained at least two identified peptides. Protein probabilities were assigned by the ProteinProphet algorithm (47).

CRPC transcriptomes from the Prostate Cancer Foundation-Stand Up To Cancer (PCF-SU2C) cohort were downloaded and reanalyzed (2). CRPC transcriptomes from The Institute of Cancer Research-Royal Marsden Hospital (ICR-RMH) cohort were analyzed as previously described (48). Briefly, paired-end transcriptome sequencing reads for each of the PCF-SU2C (n = 159) and ICR-RMH (n = 95) cohorts were aligned to the human reference genome (GRCh37/hg19) using TopHat2 (v2.0.7; RRID: SCR_013035). Gene expression in fragments per kilobase of transcript per million mapped reads (FPKM) was calculated using Cufflinks (RRID: SCR_014597). CRPC transcriptomes from the University of Washington (UW) cohort were analyzed as previously described (3, 49). Briefly, sequencing reads were mapped to hg38 using STAR (v2.7.3a; RRID: SCR_004463). Gene-level abundance was quantitated using GenomicAlignments (RRID: SCR_024236) and transformed to log₂ FPKM in edgeR (RRID: SCR_012802). Single-sample enrichment scores were calculated using gene set variation analysis with the z-score method and transcriptome-wide log₂ FPKM values as the input (50). GO cellular response to heat gene expression signature was chosen for clinical analyses based on enrichment in VCaP prostate cancer cells following heat shock when screening for Molecular Signatures Database (MSigDB) pathways matching the search term “heat” and discounting the pathways that were negatively associated and those wherein none of the genes were differentially expressed after heat shock. The expression of each gene comprising the signature GO cellular response to heat (GO:0034605; Supplementary Table S3), Hallmark Androgen Response, Hallmark E2F targets, and previously described AR signatures (3, 14, 51, 52) was used to determine an aggregate expression z-score per patient, per signature. The Pearson correlation coefficient (r) and P value were calculated between the signature z-scores.

Cell lines were lysed with RIPA buffer (Pierce) supplemented with protease inhibitor cocktail (Roche) and PhosSTOP phosphatase inhibitor mix (Roche). Experiments in which puromycin incorporation was used as a measure for protein translation had an additional step of 1 μmol/L puromycin treatment for 30 minutes prior to lysis. Castration-resistant VCaP prostate cancer cell line–derived mouse xenograft lysate was obtained by mechanical homogenization and reconstituted in RIPA buffer. Protein extracts (20 μg) were separated on 4% to 12% NuPAGE Bis-Tris gel (Invitrogen) by electrophoresis and subsequently transferred onto Immobilon-P polyvinylidene difluoride membranes of 0.45-μm pore size (Millipore). Details of primary antibodies used are provided in Supplementary Table S4. Chemiluminescence was detected on the ChemiDoc Touch Imaging System (Bio-Rad).

In vitro proliferation of cell lines was measured using the CellTiter-Glo 2D Luminescent Cell Viability Assay (Promega) according to the manufacturer’s instructions, and luminescence was measured using Synergy HTX (BioTek).

Cell line RNA was extracted using the RNeasy Plus Mini Kit (QIAGEN) as per the manufacturer’s instructions. Castration-resistantVCaP prostate cancer cell line–derived mouse xenograft RNA was obtained by mechanical homogenization, reconstituted with RNeasy RLT buffer, passed through a QIAshredder tube (QIAGEN), and further processed with RNeasy Plus Mini Kit as per the manufacturer’s instructions.

cDNA was synthesized using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). qPCR was carried out using a ViiA 7 Real-Time PCR System (Life Technologies) using the TaqMan Universal PCR Master Mix (Applied Biosystems). TaqMan probes (Thermo Fisher Scientific) used are listed in Supplementary Table S5. Fold change in RNA expression levels was calculated by the comparative CT method, using the formula 2^-ΔΔCt. Results from all studies were normalized against the average of four housekeeping genes RPLP0, GAPDH, B2M, and HPRT1.

PC3 cells were seeded at a density of 5,000 cells per well in 96-well plates in phenol red–free RPMI media supplemented with 10% charcoal-stripped FBS and concurrently transfected with 0.5 μg/mL CONTROL-FLAG, AR-FL-FLAG, or AR-V7-FLAG and 0.25 μg/mL PSA-ARE3-luciferase, using X-tremeGENE HP DNA Transfection Reagent (Merck) as per the manufacturer’s recommendations and incubated overnight. The cells were then treated with vehicle (DMSO 0.1%), 250 nmol/L CCT365248, 250 nmol/L NXP800, or 3 μmol/L enzalutamide for 1 hour prior to either stimulation with 10 nmol/L dihydrotestosterone (DHT) or DHT vehicle (ethanol 0.1%). Cells were then incubated for 16 hours and lysed with Pierce IP Lysis Buffer (Thermo Fisher Scientific). Luciferase Assay Reagent (Promega) was added to lysate, and luciferase activity was measured using the BioTek Cytation 5 Cell Imaging Multimode Reader (Agilent).

22Rv1 cells were plated in hormone-deficient media for 72 hours. After which, cells were treated with CCT365248 (250 nmol/L) or NXP800 (250 nmol/L) for 1 hour, followed by stimulation with 10 nmol/L DHT or unstimulated with DHT vehicle (ethanol 0.1%) for 5 hours. Chromatin was extracted using the Chromatin Extraction Kit (Abcam), and chromatin IP (ChIP) was performed using the ChIP Kit Magnetic–One Step (Abcam) as per manufacturer’s recommendations. Briefly, cells were cross-linked with 1% fresh formaldehyde for 10 minutes at room temperature. Chromatin was sheared to 200 to 700 bp using Diagenode ultrasonicator for 30 cycles. Lysates were incubated with AR antibody (Millipore, Cat. #06-680, RRID: AB_310214). ChIP DNA was extracted with the aforementioned kit, and qPCR was performed using PowerUP SYBR (Thermo Fisher Scientific). Genes selected for ChIP-qPCR were known AR regulated genes: KLK2, KLK3, FKBP5, TMPRSS2, CHRNA2, and ANKRD30B (53). Primers were designed for AR binding sites and are listed in Supplementary Table S6. Results were expressed as mean binding as percentage of input with SD from two biological experiments with three replicates.

The development of CP50, CP89, and CP142 patient-derived xenografts (PDX) have been previously described (44, 52, 54, 55). CP129 was developed from the same patient as CP89 but from a biopsy taken later in the treatment timeline. PDX organoid (PDX-O) were derived from PDX and PDX-O proliferation studies performed as previously described (44). CellTiter-Glo 3D Luminescent Cell Viability Assay (Promega) was used to assay growth of PDX-O according to the manufacturer’s instructions, and luminescence was measured using Synergy HTX (BioTek).

Prostate cancer cell line and castration-resistant VCaP prostate cancer cell line–derived mouse xenograft RNA quality was analyzed using the Agilent TapeStation RNA ScreenTape. About 500 ng of total RNA from each sample was first used in the NEBNext rRNA Depletion Kit, followed by the NEBNext Ultra II Directional RNA Library Prep Kit, according to the manufacturer’s instructions. Library quality was confirmed using the Agilent TapeStation High Sensitivity DNA ScreenTape. The libraries were quantified and normalized by qPCR using the KAPA Library Quantification Kit (Roche). Library clustering was performed on a cBot with Illumina HiSeq PE Cluster Kit v3. The libraries were sequenced as paired-end 101 bp reads on an Illumina HiSeq 2500 with an Illumina HiSeq SBS Kit v3. Base calling and quality scoring were performed using Real-Time Analyses (version 1.18.64) and FASTQ file generation and demultiplexing using CASAVA (RRID: SCR_001802). All FASTQ files were trimmed of adapter sequences using Trimmomatic (v0.39; RRID: SCR_011848). For prostate cancer cell line models (VCaP, LNCaP95, and 22Rv1, and newly developed 22Rv1 resistant sublines), paired-end reads were aligned to the human reference genome (GRCh38/hg38) using the STAR aligner (v2.7.7a; RRID: SCR_004463). For castration-resistant VCaP prostate cancer cell line–derived mouse xenograft, paired-end reads were aligned both to the human reference genome (GRCh38/hg38) and the mouse reference genome (GRCm38/mm10) using STAR. Aligned reads from both genomes were then used with XenofilteR (v1.6) to select only high-confidence human alignments. Subsequent to alignment, raw gene expression counts were computed using the Subread package (v2.0.1; RRID: SCR_009803) and used for differential expression analysis with the DESeq2 package (v1.32.0; RRID: SCR_015687). Normalized gene expression values were reported as calculated using DESeq2, using normalization to the geometric mean across replicates. Normalized counts were obtained for each model and experiment individually. The log fold change was obtained from DESeq2 with the parameters pAdjustMethod set to “false discovery rate” (FDR) and with independent filtering switched off. The differentially expressed genes were filtered for significance (adjusted P < 0.05), and the respective resulting lists of differentially expressed genes were used for pathway analysis with the GSEAPreranked algorithm from the clusterProfiler package (v4.0.5; RRID: SCR_016884) with default parameters. The gene set enrichment analysis was performed using gene sets comprising the H (Hallmark) collection from the MSigDB v7.1 (51). All normalized enrichment scores (NES) and FDRs presented for pathway analysis show the changes induced by NXP800 as compared with the closely matched inactive chemical control analog CCT365248–treated models or between an NXP800-resistant 22Rv1 prostate cancer cell subline and the corresponding controls or between control and heat shock.

To develop a resistant subline, 22Rv1 cells were treated with increasing concentrations of NXP800. The 22Rv1 NXP800 subline (NXP800-R) was defined by the ability to maintain growth in the presence of 2.5 μmol/L NXP800. Control cell lines were developed by exposing 22Rv1 cells to the inactive, matched chemical control compound CCT365248 (inactive-C line) or 0.1% DMSO vehicle (vehicle-C line).

VCaP prostate cancer cell line–derived mouse xenografts were established by subcutaneous injection of VCaP prostate cancer cells into the flank of uncastrated male nonobese diabetic SCID gamma (NSG) mice. Mice were then castrated 10 to 12 days prior to treatment commencement. Once tumors had established castration-resistant growth, animals were randomized into vehicle (10% DMSO/90% hydroxypropyl-β-cyclodextrin/citrate buffer, pH 5) or 35 mg/kg NXP800 (in vehicle) one-time-daily treatment groups. For short-term studies, mice were treated for 5 days by oral gavage, after which tumors were harvested for pharmacodynamic analyses. For long-term treatment, mice were treated for 38 days, by oral gavage. In all studies, the tumor volume (measured by caliper), animal body weight, and conditions were monitored at least two times weekly.

All tissue blocks were freshly sectioned and considered for IHC analyses of cleaved caspase 3 (Cell Signaling Technology, Cat. #9661, RRID: AB_2341188) or Ki-67 (Agilent, Cat. #M7240, RRID: AB_2142367). Cleaved caspase 3 and Ki-67 IHC was carried out as previously reported (18). Ki-67 and cleaved caspase 3 were reported as a percentage of cells that were positive for protein expression determined by a pathologist (B. Gurel) blinded to clinical data.

Bioinformatic analyses are detailed in associated sections. For the GO cellular response to heat gene expression signature clinical analyses, patients were divided into two groups (≤80th vs. >80th percentile). OS was defined as the time from CRPC biopsy to date of death or last follow-up (censored). OS was estimated using the Kaplan–Meier method and log-rank test for significance. Hazard ratio (HR) with 95% confidence intervals (CI) and P values for univariate Cox survival models are shown. The unpaired Student t tests were used to determine significant differences between independent groups in all the following analyses. For prostate cancer cell lines and PDX-Os, the difference in growth, and the difference in RNA expression, subsequent to drug treatments was reported. Furthermore, the difference in PSA-ARE3-luciferase stimulation by various AR constructs in response to DHT after drug treatments was calculated, as well as AR DNA binding in response to DHT after drug treatments. Finally, the difference in growth, RNA, and protein expression in castration-resistant VCaP prostate cancer cell line–derived mouse xenograft studies after drug treatments was reported. Statistical analyses were performed using GraphPad Prism v7 (GraphPad Software; RRID: SCR_002798). All experimental replicates and statistical analyses performed are detailed in figure legends. Statistical significance was prespecified at P ≤ 0.05. No adjustment for multiple testing has been made.

All clinical CRPC biopsy transcriptome analyses were performed on already published cohorts (2, 3, 48). All PDX models providing PDX-O were derived from patients who were treated at The Royal Marsden Hospital, had provided written informed consent, and were enrolled in institutional protocols approved by the ethics review committee. All mouse work was carried out in accordance with The ICR guidelines, including approval by The ICR Animal Welfare and Ethical Review Body, in compliance also with the UK Animals (Scientific Procedures) Act 1986 and the National Cancer Research Institute guidelines for the welfare and use of animals in cancer research (56).

The cell line and castration-resistant VCaP prostate cancer cell line–derived mouse xenograft RNA sequencing (RNA-seq) data are available under accession number PRJEB81176 from The European Nucleotide Archive. In addition to publicly deposited data, all data are contained within presented figures, and any further reasonable data access requests can be submitted to the corresponding authors.

HSPs have been shown to bind and activate both AR and AR-V7 in prostate cancer (17–28). To determine those proteins (including HSPs) that bind both AR and AR-V7, we conducted FLAG-tagged immunoprecipitation–mass spectrometry (IP-MS) on AR-negative PC3 prostate cancer cells overexpressing FLAG-tagged AR-FL and AR-V7 proteins (Supplementary Fig. S2A). Consistent with their described role in AR signaling, members of the 70-kDa HSP family are among the most prominent binding partners of both AR-FL and AR-V7 (Supplementary Fig. S2B; refs. 17, 19, 27, 28).

Because HSPs play a key role in cellular responses to stress, we next interrogated the impact of different cellular stresses (AR targeting, heat shock, and radiation) on AR-FL and AR-V7 protein expression in VCaP prostate cancer cells (Supplementary Fig. S2C). Consistent with previous reports, AR targeting with enzalutamide led to an increase in AR-V7 protein expression (Supplementary Fig. S2C; ref. 57). Interestingly, heat shock, with associated induction of HSP72 protein expression, was as effective as enzalutamide therapy at inducing AR-V7 protein expression (Supplementary Fig. S2C). By contrast, radiotherapy, with associated phosphorylation of replication protein A, did not impact AR-V7 protein levels (Supplementary Fig. S2C). Taken together, these data indicate that 70-kDa HSPs not only bind AR-FL and AR-V7 but their physiological induction by heat shock also associates with increased AR-V7 protein expression. This supports further investigation of targeting cellular response to heat stress components as a therapeutic strategy to overcome persistent AR signaling in lethal prostate cancer.

To further interrogate the clinical significance of these preclinical observations, we tested the association between GO cellular response to heat gene expression signature (GO:0034605; Supplementary Table S3) and OS in the two CRPC transcriptome cohorts (PCF-SU2C, n = 141; ICR-RMH, n = 94, Fig. 1A, B, G, and H; refs. 2, 48). The GO cellular response to heat gene expression signature was chosen on the basis of being the most significantly enriched MSigDB pathway matching the search term “heat” in VCaP prostate cancer cells after heat shock (Supplementary Fig. S2D). Patients with a higher (>80th percentile) GO cellular response to heat gene expression signature had poorer prognosis in both the PCF-SU2C [median OS 22.0 vs. 33.5 months, HR 2.23 (95% CI 1.14-4.34), P = 0.003] and ICR-RMH [median OS 7.1 vs. 12.4 months, HR 1.58 (0.81–3.05), P = 0.11] cohorts when compared with those patients with a lower (≤80th percentile) GO cellular response to heat gene expression signature (Fig. 1A, B, G, and H). Next, we determined whether the GO cellular response to heat gene expression signature was associated with previously reported AR signatures in three independent cohorts of patients with CRPC in transcriptome analysis (PCF-SU2C, n = 159; ICR-RMH, n = 95; UW, n = 185, Fig. 1A and G; Supplementary Fig. S3A; refs. 2, 3, 14, 48, 51, 52). The GO cellular response to heat gene expression signature significantly associated with the Hallmark Androgen Response (r = 0.51, P < 0.001), AR signature (r = 0.45, P < 0.001), Nelson Response to Androgen Up (r = 0.47, P < 0.001), and AR-V7 signature (r = 0.42, P < 0.001) in the PCF-SU2C cohort (Fig. 1C–F). Consistent with this, the GO cellular response to heat gene expression signature significantly associated with the Hallmark Androgen Response (r = 0.83, P < 0.001), AR signature (r = 0.77, P < 0.001), Nelson Response to Androgen Up (r = 0.77, P < 0.001), and AR-V7 signature (r = 0.74, P < 0.001) in the ICR-RMH cohort (Fig. 1I–L). In addition, in the UW cohort, the GO cellular response to heat gene expression signature significantly associated with the Hallmark Androgen Response (r = 0.39, P < 0.001), AR signature (r = 0.39, P < 0.001), Nelson Response to Androgen Up (r = 0.29, P < 0.001), and AR-V7 signature (r = 0.4, P < 0.001; Supplementary Fig. S3B–S3E). Taken together, these data demonstrate that the GO cellular response to heat gene expression signature associates with oncogenic AR signaling and identifies patients with advanced prostate cancer with poorer prognosis, further supporting the targeting of cellular response to heat stress as a therapeutic strategy in lethal prostate cancer.

Given the role that HSPs play in AR activity, and the demonstration that the GO Cellular Response to Heat gene expression signature associates with oncogenic AR signaling and poorer prognosis in advanced prostate cancer, we next explored whether targeting the HSF1 pathway, which regulates the canonical heat shock stress response and has been shown to modulate the AR, could represent a potential therapeutic strategy for the treatment of CRPC (18, 20, 28, 39, 58–61). To this end, we used NXP800, a clinical stage HSF1 pathway inhibitor that we derived from a phenotypic screen to detect inhibitors of the HSF1-mediated stress pathway (40, 43). Consistent with our published results in human ovarian cancer cells (43), NXP800 (Supplementary Fig. S1A), but not its closely matched inactive chemical control compound CCT365248 (Supplementary Fig. S1B), inhibited the induction of HSP72 protein expression in response to HSF1 activation, mediated by HSP90 inhibition, in a concentration-dependent manner in VCaP, LNCaP95 and 22Rv1 prostate cancer cells (Supplementary Fig. S4A–S4C). In addition, NXP800 reduced both basal, and HSP90 inhibitor-induced, HSP72 protein expression levels, in contrast to its inactive chemical control CCT365248, confirming its important effect in blocking HSF1-activated HSP expression in prostate cancer cells (Supplementary Fig. S4D).

Next, we determined the impact of NXP800 on the growth of both AR signaling inhibitor (ARSI)-sensitive (VCaP) and ARSI-resistant (LNCaP95 and 22Rv1) prostate cancer cell lines. Five days of treatment with NXP800, but not its inactive chemical control CCT365248, significantly inhibited the growth of all three AR-expressing cell lines, including LNCaP95 and 22Rv1 that also express AR-V7, and are resistant to enzalutamide, suggesting that NXP800 treatment may overcome endocrine resistance (Fig. 2A–C). Considering the link between HSPs and AR activity, we next investigated the impact of NXP800 on AR signaling, demonstrating a decrease in AR-FL and AR-V7 protein expression, and also in expression of AR-responsive genes (KLK2, KLK3, TMPRSS2, and FKBP5), across all three prostate cancer cell lines (Fig. 2A–C). Having demonstrated that NXP800, but not its inactive control CCT365248, suppressed AR-responsive genes across multiple prostate cancer cell lines, we further investigated the impact of NXP800 on AR biology. Thus, we determined the impact of NXP800, its inactive chemical control CCT365248, and also enzalutamide, on transfected AR-FL or AR-V7 transactivation of a PSA-ARE3-driven luciferase reporter assay system in AR negative PC3 prostate cancer cells (Fig. 2D). Both NXP800 and enzalutamide, but not the inactive chemical control for NXP800, CCT365248, inhibited DHT-mediated AR-FL transactivation (Fig. 2D). In addition, unlike its inactive chemical control CCT365248 and enzalutamide, NXP800 inhibited AR-V7 transactivation (Fig. 2D). These data support previous reports that HSPs are critical for both AR-FL and AR-V7 activity (13, 14, 18, 20, 28). Next, using chromatin-immunoprecipitation technology with an AR antibody pull-down in 22Rv1 cells, we demonstrated that NXP800, but not its inactive chemical control CCT365248, reduced AR binding at multiple AR-responsive genes including KLK2, KLK3, FKBP5, TMPRSS2, CHRNA2, and ANKRD30B (Fig. 2E; ref. 53). These data are in contrast to enzalutamide, which inhibits AR signaling in enzalutamide-responsive VCaP cells but has little impact on AR signaling in enzalutamide-resistant LNCaP95 and 22Rv1 cells when compared with NXP800 (Fig. 2A–C; Supplementary Fig. S5A–S5C).

Next, to further explore the utility of NXP800 in CRPC, we evaluated treatment in four PDX-Os derived from biopsies of four patients with metastatic CRPC (from three separate patients; Fig. 3A; refs. 44, 52, 54, 55). NXP800, but not the inactive chemical control CCT365248, inhibited growth in all models studied (Fig. 3A). In contrast, AR-targeting with enzalutamide demonstrated varying responses across models, with growth inhibition only seen at the highest concentrations explored (Fig. 3A). Taken together, these data demonstrate that NXP800 abrogates AR signaling, including AR-V7 transactivation, with associated growth reduction in ARSI-resistant prostate cancer cell lines and patient-derived prostate cancer models, suggesting that it may provide an attractive strategy to overcome persistent AR signaling and treatment resistance in lethal prostate cancer.

To investigate whether the growth inhibitory effects of NXP800 were exclusively through AR signaling abrogation, we interrogated AR-positive (VCaP, LNCaP, LNCaP95, and 22Rv1) and AR-negative (PC3 and DU145) prostate cancer cell lines. NXP800, but not the inactive chemical control CCT365248, significantly inhibited the growth of all prostate cancer cell lines, independent of their AR status, suggesting that growth inhibition did not occur through AR signaling inhibition alone (Fig. 3B).

Having demonstrated that NXP800 has growth inhibitory effects in AR-dependent and AR-independent prostate cancer models, we next investigated further the effects of the drug by performing RNA-seq on VCaP, LNCaP95, and 22Rv1 prostate cancer cells subsequent to NXP800 treatment (Fig. 3C–E; Supplementary Tables S7–S9). Consistent with our targeted AR studies, 100 and 250 nmol/L NXP800 (in contrast to its inactive chemical control, CCT365248, at 250 nmol/L) substantially impacted AR signaling with de-enrichment of the Hallmark Androgen Response in VCaP (100 nmol/L, NES −1.87, FDR <0.001; 250 nmol/L, NES −1.60, FDR 0.0134), LNCaP95 (100 nmol/L, NES −1.53, FDR 0.0165; 250 nmol/L, NES −1.87, FDR 0.0012), and 22Rv1 (100 nmol/L, NES −2.29, FDR <0.001; 250 nmol/L, NES −1.88, FDR <0.001) prostate cancer cells (Fig. 3C–E; Supplementary Tables S7–S9). Interestingly, exposure to NXP800 (in contrast to its inactive control compound, CCT365248, at 250 nmol/L) also led to activation of the UPR in prostate cancer cells (Fig. 3C–E; Supplementary Tables S7–S9). Thus, treatment with 100 and 250 nmol/L NXP800 (but not its inactive control compound, CCT365248, at 250 nmol/L) led to the enrichment of the Hallmark UPR in VCaP (100 nmol/L, NES 1.61, FDR 0.0232; 250 nmol/L, NES 1.86, FDR <0.001) and 22Rv1 (100 nmol/L, NES 1.31, FDR 0.1276; 250 nmol/L, NES 1.48, FDR 0.0373), but not significantly in LNCaP95 (100 nmol/L, NES 1.38, FDR 0.0941; 250 nmol/L, NES 1.18, FDR 0.2443) prostate cancer cells (Fig. 3C–E; Supplementary Tables S7–S9). In addition, other signaling pathways including Hallmark MYC Targets V1 and V2, Hallmark E2F Targets and Hallmark G₂–M Checkpoint, which are all implicated in the development and progression of CRPC, were consistently de-enriched in response to NXP800 treatment (Fig. 3C–E; Supplementary Tables S7–S9).

Next, having identified through our RNA-seq analyses that treatment with NXP800 increased the UPR gene expression signature and decreased E2F targets, these effects were investigated further (Fig. 3F and G). Given the decrease in E2F targets, we further interrogated the expression of “activating” E2Fs within our RNA-seq analyses, demonstrating that E2F1 to E2F3 RNA expression was significantly decreased in response to NXP800 treatment across all prostate cancer cells studied (Fig. 3F; refs. 62, 63). In addition, orthogonal validation by Western blot analysis demonstrated that E2F1 protein expression was reduced in response to NXP800 treatment across all prostate cancer cells studied (Fig. 3G). Furthermore, the GO cellular response to heat gene expression signature that is associated with oncogenic AR signaling and poorer prognosis in patients with advanced prostate cancer, also significantly associated with Hallmark E2F Targets in PCF-SU2C (r = 0.77, P < 0.001) and ICR-RMH (r = 0.79, P < 0.001) CRPC transcriptome cohorts (Fig. 3H; refs. 2, 48). Finally, treatment with 250 nmol/L NXP800, but not its inactive chemical control CCT365248 at this concentration, resulted in increased eIF2α phosphorylation and ATF4, ATF6, and IRE1 protein expression in VCaP, LNCaP95, and 22Rv1 prostate cancer cells (Fig. 3G).

Taken together, these data confirmed that treatment with NXP800 suppresses AR signaling and provided novel insights into the mechanism of action of NXP800 in prostate cancer cells, with activation of the UPR and inhibition of AR, MYC, and E2F activity—all reported to be key to the development and progression of CRPC (5, 64–66).

Having determined that NXP800 activates the UPR gene signature and blocks AR and E2F activity in multiple prostate cancer models, we investigated whether, at least in part, these molecular mechanisms could be linked. Consistent with these discoveries, we further confirmed that treatment of VCaP, LNCaP95 and 22Rv1 prostate cancer cells with NXP800, but not the inactive chemical control CCT365248, increased eIF2α phosphorylation and ATF4, ATF6, and IRE1 protein expression. In addition, we demonstrated that treatment with NXP800 decreased incorporation of puromycin into nascent proteins, indicating reduced overall protein synthesis (Fig. 4A). To further explore whether the molecular mechanisms and phenotype described were a consequence of the activation of the UPR, we next determined whether the small-molecule ISRIB—which promotes the assembly, stabilization, and activation of the eukaryotic translation-initiation factor eIF2B decamer and allows protein synthesis to continue despite eIF2α phosphorylation subsequent to cell stress (67–69)—could rescue the observed NXP800-mediated phenotype (Fig. 4B–D; Supplementary Fig. S6).

The addition of ISRIB in VCaP, LNCaP95, and 22Rv1 prostate cancer cells resulted in a partial rescue of NXP800-mediated growth inhibition (Fig. 4B; Supplementary Fig. S6A). In addition, ISRIB partially reversed the impact of NXP800 on ATF4, ATF6, and IRE1 protein expression, decrease in protein expression of AR-FL/AR-V7 and E2F1, and inhibition of overall protein synthesis (Fig. 4C). Furthermore, ISRIB partially rescued NXP800-mediated suppression of AR-responsive genes (KLK2, KLK3, TMPRSS2, and FKBP5), whereas the ISRIB/CCT365248 combination had no effect on these targets (Fig. 4D; Supplementary Fig. S6B). Together with the gene expression data, our findings provide further evidence that the NXP800-mediated phenotype in prostate cancer models is, at least in part, driven by activation of the UPR.

To further evaluate the NXP800-mediated phenotype, we generated an NXP800 (NXP800-R)–resistant 22Rv1 prostate cancer subline derivative that maintained growth at high concentrations (2.5 μmol/L; Fig. 5A). Alongside the generation of NXP800-R, control cell lines were developed by exposing 22Rv1 cells to the inactive chemical control compound CCT365248 (inactive-C) or to 0.1% DMSO vehicle (vehicle-C). Characterization of these sublines demonstrated that the NXP800-R subline is indeed resistant to NXP800 treatment at high concentrations (up to 10 μmol/L) in contrast to both the vehicle-C and inactive-C sublines that respond to NXP800 in a concentration–responsive manner similar to the parental 22Rv1 cell line (Fig. 5B). In addition, in contrast to the vehicle-C and inactive-C control and parental cell 22Rv1 cell lines, NXP800 treatment had no effect on basal, or HSP90 inhibitor–induced, HSP72 protein expression in the NXP800-R subline (Supplementary Fig. S7).

Next, we further investigated the mechanism of action of NXP800 using our NXP800-R and inactive-C sublines. Consistent with our results in the 22Rv1 parental line, NXP800 treatment in the inactive-C subline led to increased eIF2α phosphorylation and ATF4, ATF6, and IRE1 protein expression, reduced protein synthesis, decreased AR signaling with reduced AR-FL/AR-V7 protein expression and AR DNA binding, and decreased E2F1 protein expression (Fig. 5C; Supplementary Fig. S8A–S8C). By contrast, treatment with NXP800 in the NXP800-R subline had limited/no effect on these molecular markers identified in both the NXP800-treated inactive-C and 22Rv1 parental lines (Fig. 5C; Supplementary Fig. S8A–S8C). In addition, the NXP800-R subline had a reduced basal growth rate with lower basal protein synthesis, AR signaling, and E2F1 protein expression, when compared with the inactive-C subline (Fig. 5A–C; Supplementary Fig. S8B).

Furthermore, when comparing RNA-seq analyses across these models, consistent with our previous studies described earlier, treating inactive-C sublines with 250 nmol/L NXP800 led to a significant enrichment of Hallmark UPR and de-enrichment of Hallmark E2F Targets with significantly lower “activating” E2F (E2F1–3) RNA levels when compared with treatment with 250 nmol/L of the inactive chemical control CCT365248 (Fig. 5D; Supplementary Table S10). By contrast, treatment of the NXP800-R subline with 250 nmol/L NXP800 did not show any changes in hallmark signaling pathways, and the RNA levels of “activating” E2Fs were minimally affected (Fig. 5E; Supplementary Table S11). Finally, only Hallmark E2F Targets and Hallmark MYC Targets V1 and V2 were significantly de-enriched when comparing inactive-C and NXP800-R sublines, with “activating” E2F (E2F1–3) RNA levels being significantly lower in the NXP800-R subline (Fig. 5F; Supplementary Table S12).

Taken together, these data support our earlier findings, confirming that NXP800 antiproliferative activity is mediated, in part, through the activation of the UPR with associated blockade of key signaling pathways implicated in CRPC progression.

Having demonstrated that NXP800 inhibits the growth of ARSI-resistant prostate cancer models by activating the UPR with associated suppression of key signaling pathways required for the development and progression of CRPC in vitro, we next evaluated the impact of the drug on a castration-resistant VCaP prostate cancer cell line–derived mouse xenograft. VCaP prostate cancer cell line–derived mouse xenografts in vivo that had developed castration resistance were treated orally with 35 mg/kg NXP800 or vehicle, either for 5 consecutive days (pharmacodynamic analyses) or over a 38-day period (efficacy analyses). The regimen used incorporated dose interruptions to allow for recovery from any weight loss due to NXP800 treatment, when compared with vehicle (Supplementary Fig. S9A–S9D). Importantly, NXP800 treatment significantly reduced VCaP prostate cancer cell line–derived mouse xenograft tumor growth in 38 days, also reducing the time for tumors to reach 200% volume when compared with the vehicle [median time to 200%: 32 days vs. not reached, HR 5.4 (95% CI 1.4–20.9), P = 0.003; Fig. 6A and B]. Evaluation of VCaP prostate cancer cell line–derived mouse xenograft tumors after 5 days of NXP800 treatment by RNA-seq demonstrated significant de-enrichment of Hallmark E2F Targets together with significantly decreased “activating” E2F (E2F1–3) RNA levels when compared with vehicle controls (Fig. 6C and D). Although these RNA-seq analyses did not demonstrate significant changes in the Hallmark Androgen Response or Hallmark UPR signatures, further orthogonal analyses did show increases in eIF2α phosphorylation, and ATF4 protein expression, but limited impact on AR-FL, AR-V7, and AR signaling (Fig. 6E; Supplementary Fig. S10A and S10B; Supplementary Table S13). Finally, consistent with the inhibition of xenograft tumor growth observed, NXP800 treatment led to increased cleaved caspase 3, indicating apoptosis, and reduced Ki-67, indicating decreased proliferation, when compared with the vehicle control (Fig. 6F and G; Supplementary Fig. S10C). Overall, these data demonstrate that NXP800 activates the UPR and suppresses key signaling pathways, including the E2Fs, to inhibit the growth of castration-resistant VCaP prostate cancer cell line–derived mouse xenografts in vivo, supporting its consideration for prostate cancer–specific clinical development.

Despite the development of multiple new treatments for advanced prostate cancer, treatment resistance in the clinic is inevitable and commonly driven by the reactivation of AR signaling (6–14). The development of innovative therapeutic approaches that suppress AR signaling through novel mechanisms of action remains an urgent unmet clinical need for prostate cancer medicine. One attractive strategy is to target AR co-regulators such as HSPs that have been shown to be critical for AR signaling in prostate cancer models (17–28). The studies herein provide further evidence to support the biological importance of HSPs in advanced prostate cancer, demonstrating that both AR and AR-V7 bind members of the 70-kDa HSP family. In addition, heat shock induction of HSP72 protein expression is as effective as enzalutamide treatment at increasing AR-V7 protein expression. These data in prostate cancer cells indicate that components of the molecular response to heat stress facilitate AR splicing, persistent AR signaling, and treatment resistance and help validate intervention in this pathway as a potential therapeutic approach. We further describe preclinical therapeutic and mechanistic studies to support the potential for the clinical-stage developmental HSF1 pathway inhibitor in prostate cancer.

Studies evaluating the clinical relevance of individual HSPs in prostate cancer have been limited to small patient cohorts and lack comprehensive interrogation of lethal disease (27). To address this, and the fact that multiple HSPs and HSF1 have been implicated in AR signaling, we explored the clinical relevance of the GO cellular response to heat gene expression signature (which includes multiple HSPs and HSF1) in advanced prostate cancer (39, 61). These studies demonstrate across multiple transcriptome cohorts of patients with mCRPC that the GO cellular response to heat gene expression signature associates with previously derived AR and AR-V7 signatures as well as worse clinical outcomes (3, 14, 51, 52). These data further support targeting HSPs to overcome persistent AR signaling and prostate cancer treatment resistance.

In light of these challenges, and to support the development of novel therapeutic strategies to exploit this important prostate cancer biological feature, we evaluated NXP800, a clinical-stage drug optimized from a hit in a phenotypic screen aimed to discover HSF1 pathway inhibitors (based on the ability to block HSP90 inhibitor-mediated HSP72 induction), which is currently being evaluated in a phase Ib clinical trial in ovarian cancer (NCT05226507; refs. 40, 43). The studies herein demonstrate that NXP800 reduced basal and HSF1-mediated, HSP90 inhibitor–induced, HSP72 protein expression levels in prostate cancer cell lines, consistent with its mechanistic development rationale and preclinical therapeutic activity in sensitive preclinical models of ovarian cancer (40, 70). Importantly, NXP800 inhibited AR/AR-V7 transactivation and reduced AR DNA binding and expression of AR-responsive genes, as well as decreased the growth of prostate cancer cell lines and patient-derived models with common ARSI-resistant mechanisms, including AR amplification and AR-V7 protein expression. These data support the consideration of NXP800 for prostate cancer–specific clinical development.

Interestingly, NXP800 demonstrated anticancer activity in AR-negative prostate cancer cell lines, suggesting that the antitumor activity of the drug is not only due to blocking AR signaling. Unbiased analyses of RNA-seq data across multiple prostate cancer cell lines treated with NXP800 demonstrated a predominant increase in the UPR gene signature. Further orthogonal validation of this identified that NXP800 increased eIF2α phosphorylation and ATF4, ATF6, and IRE1 protein expression. It is important to note that the eIF2α axis has been previously shown to regulate ATF4, ATF6, and IRE1 protein expression, and although we have not fully interrogated the activation status of the three arms of the UPR, we have identified that the NXP800-mediated phenotype may be driven through the eIF2α axis, and this warrants further investigation in the future (71, 72). Considering those pathways beyond the UPR, RNA-seq analyses demonstrated that treatment with NXP800 impacted other key pathways, including AR, MYC, and E2F activity, implicated in prostate cancer treatment resistance (63, 66, 73). These data also support the study of NXP800 in subjects suffering from mCRPC.

Importantly, activation of the UPR has previously been reported to be both beneficial and detrimental to the survival of cancer cells (74–78). Although the UPR is generally viewed as protective for prostate cancer cells, treatment with ONC201, a UPR activator, led to the induction of the UPR and cell death in prostate cancer models, which is consistent with our studies (74, 77, 78). Considering these data, we next investigated whether the NXP800-mediated increase in eIF2α phosphorylation and ATF4, ATF6, and IRE1 protein expression was beneficial or detrimental to these prostate cancer models studied. To do this, we used the small-molecule eIF2B stabilizer/activator ISRIB to determine whether the NXP800-induced phenotype can be rescued (67). ISRIB did rescue some of the impact of NXP800-mediated increases in components of the UPR, including ATF4, ATF6, and IRE1 protein expression, which caused a reduction of growth inhibition across most of our models. In addition, NXP800-induced suppression of AR-responsive genes, E2F1 to E2F3 RNA expression, and E2F1 protein expression was attenuated by ISRIB, consistent with our hypothesis that these changes, including ATF4, ATF6, and IRE1 protein induction, may be driven by NXP800-mediated activation of the eIF2α axis (71, 72). It will be important that follow-on studies deconvolute the mechanism by which NXP800 activates the eIF2α axis, including investigation of the eIF2α kinases (such as PERK, PKR, GCN2, and HRI), and interrogate whether NXP800 activates the IRE1 and ATF6 pathways, which may all play a role in the observed NXP800-mediated phenotype, as this will further support the clinical development of this therapeutic strategy (79).

Consistent with our ISRIB studies, NXP800 had limited to no impact on eIF2α phosphorylation and ATF4, ATF6 and IRE1 protein expression, or effects on AR, E2F and MYC in our NXP800-resistant 22Rv1 prostate cancer subline, further supporting that activation of the eIF2α axis plays a role in the NXP800-mediated phenotype observed. Overall, these data suggest that NXP800, through a novel mechanism of action, offers the potential for clinical antitumor activity against prostate tumors not only with AR splice variant expression but also with high E2F and MYC activity, supporting clinical evaluation following progression on the first AR-targeting therapy in which both AR-dependent and AR-independent mechanisms of resistance have emerged (6–14, 80).

The work herein does have limitations that we acknowledge. The translation of therapeutic strategies targeting HSPs and proteostasis to the clinic has been challenged by pharmacologic and tolerability limitations, which may be relevant to NXP800, although the oral HSP90 inhibitor pimitespib has been approved in Japan (35). Predictive biomarkers of response to our proposed new therapeutic strategy also now have to be developed, and we have not pursued this to date, even though we envision that it may include patients with tumors exhibiting AR-V7 expression, high activating E2Fs, and/or high MYC. In addition, our in vivo studies, although supported by multiple in vitro prostate cancer cell line and patient-derived prostate cancer models, remain limited. Our clinical RNA-seq data focused on signatures derived from CRPC transcriptome cohorts, and we did not explore expression of single HSPs at the RNA or protein level. However, this removed the challenges of preanalytic variables that can confound IHC analyses and supported the identification of signatures that provide a more accurate representation of cellular processes that are rarely represented by single autonomous molecules. The development and interrogation of our NXP800-resistant 22Rv1 prostate cancer subline has not identified the mechanism of acquired resistance to NXP800 despite the near complete reversal of the NXP800-mediated phenotype. However, our investigative studies were limited to RNA-seq analyses, and these models provide a valuable resource for further mechanistic work to investigate drivers of NXP800 resistance and identify rational synergistic combinations to further support the clinical development of NXP800 in lethal prostate cancer. Critically, activation of the UPR can be beneficial and detrimental to prostate cancer cells, and although we and others have confirmed that this can drive prostate cancer cell death, as this may be context specific, it will be important that further studies elucidate those specific tumors that will respond to such therapeutic strategies (74, 77, 78). Importantly, we have identified pharmacodynamic markers, such as AR, AR-V7, PSA, phospho-eIF2α, E2F1, and HSP72, as well as gene expression signatures that can be incorporated into clinical pharmacokinetic/pharmacodynamic studies as part of a Pharmacological Audit Trail (81).

Taken together, these data, alongside other studies that have interrogated HSF1 pathway inhibition in prostate cancer models, support the biological and clinical relevance of HSPs in mCRPC. Importantly, this study with NXP800, a clinical-stage developmental HSF1 pathway inhibitor currently being trialed in ovarian cancer, supports its clinical evaluation as a novel therapeutic strategy for patients suffering from mCRPC, whose tumors have evidence of spliced AR, increased E2F1 to E2F3, and MYC function (42).

W. Zeng reports grants from the ICR during the conduct of the study. E. Pasqua reports personal fees from the ICR during the conduct of the study, as well as patents for WO201549535, WO2016156872, and WO2018065787 issued. N. Chessum reports personal fees from the ICR outside the submitted work, as well as a patent for WO2015049535 issued and licensed to Nuvectis Pharma. M. Cheeseman reports a patent for US11787786B2 issued to AstraZeneca UK Ltd, as well as patents for EP3283484A1 and EP3523295A1 issued to Cancer Research Technology Ltd. R. te Poele reports grants from Nuvectis Pharma and Sixth Element during the conduct of the study. M. Powers reports grants from Nuvectis Pharma and the ICR during the conduct of the study, as well as personal fees from the ICR outside the submitted work. P. Clarke reports grants from Nuvectis Pharma and Sixth Element during the conduct of the study. U. Banerji reports other support from Nuvectis Pharma during the conduct of the study; personal fees from Carrick Therapeutics, PharmEnable, Ellipsis Pharma, Dania Therapeutics, Amalus Therapeutics, and PEGASCY-GROUP; grants from Verastem Oncology and Avacta outside the submitted work; and a patent for the ICR issued. K. Jones reports grants from CRUK, Cancer Research Technology Pioneer Fund, and Battle Against Cancer Investment Trust during the conduct of the study, as well as a patent for WO2015/049535A1 issued, with royalties paid from Nuvectis Pharma. P. Workman reports personal fees from Alterome Therapeutics; grants and personal fees from Astex Pharmaceuticals, Merck KGaA, and Cyclacel Pharmaceuticals; other support from AstraZeneca, Chemical Probes Portal, CV6 Therapeutics, Epicombi Therapeutics, and Chroma Therapeutics; personal fees and other support from Nextech Invest, STORM Therapeutics, and Derwentwater Associates; grants, personal fees, and other support from Nuvectis Pharma; grants, nonfinancial support, and other support from Vivan Therapeutics; grants from Battle Against Cancer Investment Trust and Sixth Element Capital/CRT Pioneer Fund; grants and other support from the ICR; and personal fees and nonfinancial support from Nuevolution during the conduct of the study. P.S. Nelson reports grants from Janssen, as well as personal fees from Bristol Myers Squibb, Genentech, AstraZeneca, and Pfizer outside the submitted work. J.S. de Bono reports other support from AbbVie, Acai Therapeutics, Amgen, Astellas, Amunix, Bayer, Celcuity, Dark Blue Therapeutics, Duke Street Bio Ltd, GSK, Takeda, and Tango Therapeutics; personal fees from BioXcel Therapeutics, Daiichi, Dunad Therapeutics, Endeavor BioMedicines Inc., MacroGenics, MOMA Therapeutics, Nuvation Bio, One-carbon Therapeutics Inc., PAGE Therapeutics, and Tubulis GmbH; grants and other support from Crescendo, Genentech/Roche, Merck Serono, MetaCurUm, Myricx, Nurix Therapeutics, Oncternal Therapeutics, Orion Pharma, Sanofi, and Immunic Therapeutics; personal fees and other support from Novartis; and grants and personal fees from Pfizer, outside the submitted work; in addition, J.S. de Bono reports a patent for DNA damage repair inhibitors for treatment of cancer, licensed to AstraZeneca, and for 17-substituted steroids useful in cancer treatment, licensed to Janssen. A. Sharp reports other support from Sanofi, Roche Genentech, Nurix, Astellas Pharma, Merck Sharp & Dohme, DE Shaw Research, CHARM Therapeutics, Ellipses Pharma, and Droia Ventures outside the submitted work; being an employee of the ICR, which has a commercial interest in abiraterone, PARP inhibition in DNA repair defective cancers, and PI3K/AKT pathway inhibitors (no personal income); and being the CI/PI of industry-sponsored clinical trials. No disclosures were reported by the other authors.

J. Welti: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. D. Bogdan: Conceptualization, resources, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. I. Figueiredo: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. I. Coleman: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. J. Jiménez Vacas: Data curation, methodology, writing–review and editing. K. Liodaki: Data curation, methodology, writing–review and editing. F. Weigl: Data curation, methodology, writing–review and editing. L. Buroni: Data curation, methodology, writing–review and editing. W. Zeng: Data curation, methodology, writing–review and editing. I. Bernett: Data curation, methodology, writing–review and editing. C. Bertan: Data curation, methodology, writing–review and editing. T.I. Roumeliotis: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. A. Bhamra: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. J. Rekowski: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. B. Gurel: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. A.J. Neeb: Methodology, writing–review and editing. J. Ning: Methodology, writing–review and editing. D. Li: Data curation, methodology, writing–review and editing. V.S. Gil: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. R. Riisnaes: Data curation, methodology, writing–review and editing. S. Miranda: Data curation, methodology, writing–review and editing. M. Crespo: Data curation, methodology, writing–review and editing. A. Ferreira: Data curation, methodology, writing–review and editing. N. Tunariu: Resources, methodology, writing–review and editing. E. Pasqua: Resources, writing–review and editing. N. Chessum: Resources, writing–review and editing. M. Cheeseman: Resources, writing–review and editing. R. te Poele: Writing–review and editing. M. Powers: Writing–review and editing. S. Carreira: Supervision, methodology, writing–review and editing. J. Choudhary: Formal analysis, supervision, validation, investigation, methodology, writing–review and editing. P. Clarke: Resources, supervision, writing–review and editing. U. Banerji: Supervision, writing–review and editing. A. Swain: Resources, supervision, writing–review and editing. K. Jones: Resources, supervision, writing–review and editing. W. Yuan: Data curation, software, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. P. Workman: Resources, supervision, investigation, writing–review and editing. P.S. Nelson: Supervision, writing–review and editing. J.S. de Bono: Conceptualization, resources, supervision, funding acquisition, investigation, visualization, writing–original draft, project administration, writing–review and editing. A. Sharp: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing.

The authors gratefully acknowledge the patients and families of patients who contributed to this study. This work was supported by ProstateUK (Research Innovation Award to J.S. de Bono and travelling prize fellowship to J. Jiménez Vacas); the PCF (Young Investigator Awards to A. Sharp; challenge awards to A. Sharp and J.S. de Bono); MRC (Clinical Research Training Fellowship to A. Sharp; research funding to J.S. de Bono); The Academy of Medical Sciences (Starter Grant for Clinical Lecturers to A. Sharp); the Wellcome Trust (Clinical Research Career Development Fellowship to A. Sharp); the NIH and Care Research Biomedical Research Centre (Funding to A. Sharp); the NIH (P50CA097186, R01CA234715, R01CA266452, R21CA277368, and P01CA163227 to P.S. Nelson; R50CA274336 to I. Coleman); and Cancer Research UK (program grants C309/A11566 and C2739/A22897 to P. Workman; Centre Programme and Experimental Cancer Medicine Centre grants to J.S. de Bono). P. Workman is a CRUK Life Fellow and acknowledges support from CRUK Strategic Award C35696/A23187 and Infrastructure Award C309/A27413, as well as funding for the CRUK Children’s Brain Tumour Centre of Excellence (C9685/A26398/RG93685); Wellcome (Biomedical Resource and Technology Development Grant 212969/Z/18/Z to support the Chemical Probes Portal); Chordoma Foundation; Mark Foundation; Bone Cancer Research Trust; CRIS Cancer; and the ICR. J.S. de Bono is an NIH Research (NIHR) senior investigator. All authors from the ICR acknowledge infrastructural funding from CRUK to the Convergence Science Centre at the ICR and NIHR Biomedical Research Centre/Experimental Cancer Medicine Centre initiatives to the ICR and The Royal Marsden NHS Foundation trust. The views expressed in this article are those of the authors and not necessarily those of the National Health Service, the NIHR, or the Department of Health.