Defining Splicing Factor Requirements for Androgen Receptor Variant Synthesis in Advanced Prostate Cancer

At presentation, prostate cancer growth and progression are androgen dependent, with androgen receptor (AR) signaling being a key disease driver. As such, first-line therapies typically aim to attenuate this signaling axis by reducing available androgens using androgen deprivation therapy (ADT) to starve the AR of the ligand (1, 2). Although patients are typically responsive to first-line hormone therapies, patients eventually relapse, and the disease often progresses to an androgen-independent, more aggressive form of the disease termed castration-resistant prostate cancer (CRPC; ref. 3). Along with chemotherapeutics, second-generation hormonal therapies, such as the anti-androgen enzalutamide, which effectively inhibits the AR by binding to the C-terminal ligand-binding domain, are effective in upward of 50% of patients with CRPC (4–6). Critically, the AR signaling axis remains active in this advanced disease setting, in part through several aberrations of the AR signaling cascade, including AR gene amplification, gain-of-function AR mutations, and generation of alternatively spliced forms of the AR, termed AR variants (AR-V; refs. 1, 7–9). AR-Vs, including the most common isoform AR-V7, diminish sensitivity to ADT, are expressed in up to 75% of ADT-treated patients, but rarely in those with primary prostate cancer (<1%), and significantly diminish overall survival of this patient cohort (10–12). Therefore, therapeutically targeting AR-Vs represents a key unmet clinical need in the treatment of CRPC.

Unlike the full-length AR (FL-AR) isoforms, AR-Vs lack the ligand-binding domain while retaining the transcriptionally potent N-terminal transactivation domain and DNA-binding domain (13, 14). Hence, AR-Vs remain constitutively active at castrate levels of androgens and are refractory to the current repertoire of anti-androgens and ADT (9, 12, 15). Due to the unstructured nature of the N-terminal transactivation domain, therapeutically exploiting this region of AR has been proved difficult, whereas the highly conserved DNA-binding domain across the steroid hormone receptor family makes selective targeting of this region a challenge. However, encouragingly, there has been recent success in this area (16, 17). An alternative route to AR-V signaling blockade gaining traction is to identify and exploit splicing processes that are critical for AR-V generation, which could represent tractable therapeutic targets for patients with AR-V–positive CRPC (18–20).

Splicing of exon 3 to a number of distinct 3′ cryptic exons (CE) located within intron 3 generates a number of clinically relevant AR-Vs, including AR-V1 and AR-V7, with the latter, for example, containing CE3 (14, 21, 22). At present, our current understanding of the splicing factors that control CE inclusion at the expense of constitutive exons 4 to 8, which encode the FL-AR C-terminus, is limited. Candidate-based approaches have identified splicing regulators required for the inclusion of CE3 to produce mature AR-V7 transcripts, including U2AF65 and ASF/SF2 (23). Splicing factor SFPQ is highly expressed in metastatic CRPC samples and upregulates expression of SF3B2, a component of the SF3B complex (24), which has been shown to be a critical determinant of AR-V7 expression (25). Like SF3B2, depletion of SF3B3, another component of the SF3B complex, has been shown to inhibit growth of AR-V–positive CWR22Rv1 cells (24) as a consequence of diminished FL-AR and AR-V7 mRNA expression. More recently, we have shown that RBMX enhances splicing of exon 3 to several cryptic exons encoding AR-V7, AR-V6, and AR-V9, as well as canonical exons encoding FL-AR (26).

While of interest, a global analysis of splicing factors important for AR-V synthesis has yet to be performed, which is a major knowledge gap and could be vital to comprehensively define splicing factor involvement in maturation of AR-V transcripts. We therefore applied a bespoke CRISPR library screen, encompassing 211 splicing factors, to interrogate target knockout on AR-V abundance in prostate cancer cells and highlight splicing regulators critical for generating AR-Vs. MFAP1, a component of the spliceosome B complex (27, 28), and CWC22, which is part of the exon-joining complex (29–31), were identified and subsequently validated across a number of prostate cancer models to regulate AR-V and AR-V/FL-AR transcript generation, respectively. Individual depletion of both targets selectively diminished growth of prostate cancer cell lines, but not those derived from normal prostate epithelia, supporting the concept that these are nonessential genes that may offer a therapeutic window upon blockade. Critically, MFAP1 controls expression of AR-dependent and -independent gene signatures, including canonical AR target genes and those involved in the DNA damage response (DDR), which we show sensitizes prostate cancer cells to ionizing radiation (IR) upon MFAP1 knockdown. Overall, we have comprehensively profiled global splicing factor regulation of AR-V synthesis and subsequently validated MFAP1 and CWC22 as key regulators of pathogenic AR transcript splicing that could represent new therapeutic targets in the advanced stages of prostate cancer.

CWR22Rv1 (CRL-2505), VCaP (CRL-2876), LNCaP (CRL-1740; RRID: CVCL_0395), HEK293FT (CRL-3216; RRID: CVCL_6911), and RWPE-1 (CRL-11609; RRID: CVCL_3791) cell lines were purchased from ATCC. CWR22Rv1-AR-EK cells are a CRISPR-engineered cell line generated by the host laboratory from CWR22Rv1 cells (32). CWR-R1-AD1 and CWR-R1-D567 (RRID: CVCL_ZC61) cells were a kind gift from Scott Dehm, University of Minnesota (33); LNCaP-95 (RRID: CVCL_ZC87) were a kind gift from Adam Sharp (Institute of Cancer Research); and PNT1-A and PNT2-C2 cells were a kind gift from Jan Trapman, Erasmus MC. All cells were maintained in RPMI-1640 (R5886, Sigma-Aldrich), with the exception of VCaP cells, which were grown in DMEM (D5030, Sigma-Aldrich), and LNCaP-95, which were grown in phenol red–free RPMI-1640, supplemented with 10% (v/v) FCS (Sigma) and 2 mmol/L L-glutamine (Sigma), hereby referred to as full media. To assess AR-V activity (CWR22Rv1 parental and VCaP cell lines), cells were seeded in RPMI-1640 or DMEM, respectively, supplemented with 10% (v/v) dextran-coated, charcoal-stripped FCS and 2 mmol/L L-glutamine, hereby referred to as steroid-depleted media. Cell lines were cultured in a humidified incubator (MCO-20AIC, Sanyo) at 37°C under 5% CO₂. Mycoplasma tests were carried out in-house every 2 months and subjected to regular short tandem repeat profiling, and cells were never grown past 25 passages in culture.

siRNA and single-guide RNA (sgRNA; Supplementary Tables S9 and S10) were purchased from Sigma-Aldrich in lyophilized form, resuspended to 50 mmol/L in sterile RNase/DNase-free water, and were stored at −20°C in aliquots. Lipofectamine RNAiMAX Transfection Reagent (Invitrogen) was used for siRNA and sgRNA (composed of composite crRNA/tracrRNA) delivery into the cells, according to the manufacturer’s instructions. Cells were incubated for 72 hours at 37°C prior to harvest for optimal gene knockdown/knockout.

The inducible Cas9-expressing pTLCV2 vector was purchased from Addgene (Addgene plasmid #52961) together with second-generation lentiviral packaging and envelope plasmids psPAX2 and pMD2.G (Addgene plasmids #12260 and #12259). Cells were transfected with mammalian expression plasmids using TransIT-LT1 (Mirus Bio) at a ratio of 1 mg DNA:3 mL TransIT-LT1 as described in (32).

For generation of the CWR22Rv1-AR-EK-iCas9 derivative, CWR22Rv1-AR-EK cells were seeded at a density of 2 × 10⁵ cells per well and transduced with 500 μL of the pTLCV2-Cas9–containing virus for 24 hours at 37°C. Cells were then incubated with fresh media for a further 24 hours prior to the addition of 1 μg/mL puromycin continually for >3 weeks for selecting clonal populations that stably integrated the Cas9 construct and demonstrating Cas9 expression in response to 1 μg/mL doxycycline.

RNA was isolated from cells using TRIzol reagent (Life Sciences, Invitrogen) according to the manufacturer’s handbook as described in (32). The resultant RNA was subjected to cDNA synthesis using the M-MLV Reverse Transcriptase system (Promega). cDNA was diluted by adding 130 μL RNase/DNase-free water for analysis by RT-qPCR, incorporating SYBR Green Dye 1 (Life technologies) and custom primers purchased from Sigma-Aldrich (Supplementary Table S11), using a QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific). Ct values were normalized to RPL13A transcript using the ∆∆CT method.

Following protein separation by gel electrophoresis, proteins were transferred to a Hybond ECL nitrocellulose membrane (GE Healthcare) prior to Western blot analysis as described in (34) using the antibodies listed in Supplementary Table S12.

Unless stated otherwise, all graphical data are representative of the mean of three independent biological repeats, and error bars represent ±SEM. The type of statistical testing is indicated in figure legends, typically either one-way ANOVA, two-way ANOVA, or an unpaired t test. All statistical analyses were conducted using Prism 8 software. The P values are interpreted as follows: *, P < 0.05; **, P < 0.01; ***, P <0.001; ns, not significant.

RNA sequencing (RNA-seq) data generated in this study are available in Gene Expression Omnibus at GSE248502. All other data are available upon request from the corresponding author. Additional methods can be found in Supplementary Materials accompanying this article.

A global screen of splicing factors important for AR-V transcript maturation was required to fill an important knowledge gap and furnish new therapeutic targets for AR-V–positive prostate cancer. To that end, we created a bespoke CRISPR library consisting of four sgRNAs per gene, targeting 211 core splicing factors, and splicing regulators (Supplementary Table S1). These were incorporated into a high-throughput confocal microscopy pipeline, assessing total AR-V protein abundance following gene knockout. Our proprietary CWR22Rv1-derivative cell line CWR22Rv1-AR-EK, which only expresses AR-Vs (32) and hence is an ideal model system to conduct the study, was further modified to homogenously express Cas9 in response to doxycycline (termed CWR22Rv1-AR-EK-iCas9; Supplementary Fig. S1A–S1C). Doxycycline-induced Cas9 activity was subsequently validated using an AR-targeting sgRNA, which markedly decreased AR-V levels compared with a scrambled sgRNA control, as a consequence of effective AR gene knockout, and significantly diminished canonical AR target gene expression (Supplementary Fig. S2A–S2C). Importantly, CRISPR-mediated knockout of AR decreased CWR22Rv1-AR-EK-iCas9 cell growth to levels comparable with receptor knockdown in parental and CWR22Rv1-AR-EK-iCas9 derivatives (Supplementary Fig. S2D and S2E), indicating efficient CRISPR activity in this new cell line. To optimize immunofluorescent detection of AR levels for our high-throughput assay, a 1:1,000 dilution of an N-terminal AR-targeting antibody was identified as optimal to successfully detect flux to AR abundance in CWR22Rv1-AR-EK-iCas9 cells transfected with control or AR-targeting sgRNAs (Supplementary Fig. S3A). Furthermore, selection of an Alexa Fluor 647 secondary antibody enabled the acquisition of higher-resolution images and increased the dynamic range of AR detection over other tested antibodies (Supplementary Fig. S3B and S3C). Consistent with this, the detection of reduced AR levels in cells transfected with either AR-targeting sgRNAs or siRNAs confirms that our immunofluorescence assay provides a clear high-throughput screening window to assay AR-V abundance and is appropriate to assess the impact of splicing factor gene knockout in the subsequent screening assays (Supplementary Fig. S3D–S3F).

In addition to the 211 splicing factors contained within the screen, we incorporated the previously validated sgAR-1 as a positive control for depleted AR protein abundance. An additional positive control of three PLK1-targeted sgRNAs (sgPLK1-Pool), which were shown to cause a significant reduction in cell number equivalent to siRNA depletion (Supplementary Fig. S4A–S4C), was included to provide a positive control for target knockout lethality (see Supplementary Table S1 for the full list of screen targets). The nontargeting scrambled sgRNA was used to define average steady-state nuclear AR-V abundance to which relative change in AR-V levels upon splicing factor gene knockout was calculated.

The optimized screening protocol (as shown in Fig. 1A) was conducted in four independent experiments and utilized a 96-hour assay window to maximize the impact of gene knockout as evidenced by significantly diminished AR staining and cell growth by sgAR-1 and sgPLK1-Pool controls, respectively (Fig. 1B). The top 50 screening hits which reduced AR-V protein levels are shown in the heatmap (Fig. 1C). Of these, 10 screening hits demonstrated a significant decrease in the AR-V protein level compared with control across the four independent experimental repeats (Fig. 1D), including SFPQ, which had been previously identified as a regulator of AR-V7 synthesis (24). MFAP1, SLU7, and CWC22 were further evaluated and compared with the positive control SFPQ and the negative control QKI, which showed negligible effects on AR levels across the four screen repeats. Depletion of these five targets in several PC cell lines, including VCaP, CWR-R1-AD1, and CWR-R1-D567 (Supplementary Fig. S5A), demonstrated varied impacts on AR abundance. MFAP1 and CWC22 depletion selectively downregulated AR-V and AR-V/FL-AR isoforms, respectively, in CWR22Rv1 derivatives and VCaP cells, but not in CWR-R1-AD1 cells, to levels similar to SFPQ (Fig. 1E; Supplementary Fig. S5B–S5E), whereas knockdown of SLU7 had no impact on AR abundance across all tested prostate cancer cell lines and was consistent with QKI (Supplementary Fig. S5B–S5F). This latter finding suggested that the observed reduction in AR levels upon SLU7 knockout in the CRISPR screen was potentially off-target. Importantly, MFAP1 and CWC22 knockdown in CWR-R1-D567 cells failed to markedly affect respective expression of the TALEN-engineered AR-V567es, an AR-V which does not require selective splicing activity to be generated (Supplementary Fig. S5F), suggesting that these splicing regulators may directly control AR transcript processing rather than downstream translation. Crucially comparing AR abundance and cell proliferation readouts from the CRISPR screen indicated that both MFAP1 and CWC22 were among a small number of splicing factors that diminished AR and cell growth to levels equivalent to AR knockout (Supplementary Fig. S6A and S6B). Hence, MFAP1 and CWC22 were studied further to define their involvement in regulating FL-AR and AR-V expression in prostate cancer.

Consistent with our preliminary siRNA experiments, MFAP1 and CWC22 knockout using the deconvoluted sgRNA pools markedly depleted AR-V protein abundance in CWR22Rv1-AR-EK-iCas9 cells (Fig. 1E), coincident with reduced expression of both splicing factors by CRISPR-mediated gene editing (Supplementary Fig. S7A and S7B). Subsequent transfection of additional MFAP1- and CWC22-targeting siRNAs across several FL-AR–expressing and FL-AR/AR-V–positive and –negative prostate cancer cell lines, including VCaP, LNCaP and LNCaP-95, supported our findings that MFAP1 depletion selectively reduces AR-V protein levels without impacting FL-AR, whereas CWC22 knockdown consistently reduced both FL-AR and AR-Vs (Fig. 2A; Supplementary Figs. S8A–S8E and S9A–S9E). Importantly, mRNA levels of several clinically relevant AR-Vs, including AR-V1, AR-V6, AR-V7, and AR-V9, were shown to be downregulated by both MFAP1 and CWC22 knockdown in CWR22Rv1 derivatives and VCaP cells, suggesting that these splicing factors may be important for AR-V transcript synthesis and/or maturation (Fig. 2B–D). In support of this, CRISPR-induced knockout of MFAP1 and CWC22 caused significant reduction of AR-V1 and -V7 transcripts (Fig. 2E). FL-AR transcript levels were found to be reduced by knockdown of both splicing factors, which, for MFAP1, was unexpected, given it had no effect on FL-AR protein levels. Irrespective of this, however, respective depletion of MFAP1 and CWC22 was found, in the main, to significantly decrease expression of canonical AR target genes UBE2C, CCNA2, and TMPRSS2 in CWR22Rv1 derivatives and VCaP cells grown under steroid-depleted conditions (Fig. 2F–H). We speculate that this is a direct result of downregulated expression of AR-V and FL-AR isoforms in response to diminished levels of the two splicing factors.

To test if the effects of MFAP1 and CWC22 downregulation manifested in reduced prostate cancer cell proliferation, as evidenced from our initial CRISPR screening data, an extensive panel of prostate cancer and normal prostate epithelial cells was subjected to siRNA-mediated knockdown. Depletion of both MFAP1 and CWC22 caused a significant reduction in CWR22Rv1 and CWR22Rv1-AR-EK cell growth comparable with AR knockdown, suggesting that the impact of MFAP1 and CWC22 depletion phenocopies loss of AR [Fig. 3A and B (SRB assay); Supplementary Fig. S10A–S10C (cell counts)]. In VCaP cells, MFAP1 depletion had no significant impact on cell growth, which is a likely consequence of selective depletion of AR-Vs, but not FL-AR, in this AR gene–amplified cell line (Fig. 3C), whereas loss of both AR-Vs and FL-AR upon CWC22 knockdown caused a significant decrease in cell proliferation. In contrast, knockdown of MFAP1 and CWC22 failed to affect the growth of CWR-R1-AD1 and CWR-R1-D567, which is consistent with Western blot analysis demonstrating no impact of splicing factor depletion on FL-AR and AR-V567es isoforms (Fig. 3D and E). Furthermore, AR-negative PC3 cells, but not normal prostate epithelial PNT1A and PNT2C-2, showed sensitivity to diminished MFAP1 and CWC22 expression, implicating roles in RNA metabolism outside of the AR signaling cascade (Fig. 3G and H). Importantly, the lack of response in the PNT1A and PNT2C-2 cells may highlight a therapeutic window in which MFAP1 and CWC22 targeting could selectively impact CRPC cells without toxicity in normal prostate tissue. We subsequently validated the effect of diminished MFAP1 and CWC22 levels on prostate cancer cell growth using CRISPR-mediated knockout of both splicing factors in our CWR22Rv1-AR-EK-iCas9 cell line (Fig. 3I), as well as demonstrating that knockdown of each causes arrest at the G₁ phase of the cell cycle at the expense of G₂–M (Fig. 3J). Interestingly, depletion of MFAP1 modestly enhanced the antiproliferative effect of the second-generation anti-androgen enzalutamide on CWR22Rv1 cells (Fig. 3K) and also diminished growth of an enzalutamide-resistant VCaP cell line (Fig. 3L), providing a rationale for targeting AR-V splicing regulators in treatment-resistant CRPC. Overall, our findings provide evidence that in AR-positive cells, MFAP1 and CWC22 regulate AR isoform generation at the level of transcript synthesis and/or maturation, which subsequently impacts canonical AR signaling and prostate cancer cell growth.

To provide more mechanistic insights into the role of the two splicing factors in mature AR transcript synthesis, we first assessed the effect of knockdown of MFAP1 and CWC22 on endogenous unspliced AR pre-mRNA levels in CWR22Rv1 cells (Fig. 4A). This experiment was designed to provide insights into whether these splicing factors regulated AR gene transcription or at a point downstream during transcript maturation. Intriguingly, although a marked downregulation of mature AR-V7 transcript levels was observed, there was no apparent impact on AR pre-mRNA levels (Fig. 4B) in response to MFAP1 and CWC22 depletion (Fig. 4C and D), suggesting that these splicing factors function downstream of transcription and may regulate splicing.

To test this, we subsequently conducted minigene-based AR splicing analysis in AR-negative PC3 cells (Fig. 4E) as performed in (26). Here, cells were individually depleted of MFAP1 and CWC22, and the effect on splicing of ectopic AR-V7– (exon 3–CE3) and FL-AR–encoding (exon 3–exon 4) exons was analyzed by RT-qPCR. As shown in Fig. 4F, splicing of exons encoding AR-V7 was significantly diminished by knockdown of both MFAP1 and CWC22 (Fig. 4G and H), whereas FL-AR–encoding exon splicing was markedly and selectively depleted by CWC22, which is largely consistent with our findings in CWR22Rv1 and VCaP PC cells. These data support the concept that MFAP1 and CWC22 are involved in regulating splicing events required for AR-V and FL-AR generation. Because of our interest in defining the role of splicing factors in the selective generation of AR-Vs, we decided to limit our subsequent molecular characterizations to MFAP1 only.

Examining the MFAP1-regulated transcriptome was next undertaken to define its role in regulating global gene expression and splicing activities in prostate cancer. Critically, given that MFAP1 knockdown reduces AR-V levels, it was important to distinguish differentially expressed genes (DEG) driven by AR-Vs from those directly impacted by MFAP1 depletion. We therefore incorporated an AR-V knockdown arm, alongside MFAP1 depletion, into our RNA-seq workflow in CWR22Rv1 cells (Supplementary Fig. S11A) and validated target knockdown at the protein and transcript levels prior to sequencing (Supplementary Fig. S11B–S11D). Reassuringly, although our triplicate samples for each experimental arm showed good overlap, we also observed robust interexperimental arm separation, which was indicative of distinct gene expression profiles among scrambled control, MFAP1 knockdown, and AR-V knockdown (Supplementary Fig. S11E). Specifically, AR-V and MFAP1 depletion resulted in 13,578 and 13,578 DEGs, respectively; of these, 648 and 1,004 DEGs demonstrated a statistically significant 1.5-fold change (log₂ fold change 0.58; P-adjusted <0.05; Fig. 5A; Supplementary Fig. S11F; Supplementary Table S2).

Given that we would be initially assessing an overlap between the MFAP1 and AR-V statistically significant DEGs (SDEG), we began by validating our CWR22Rv1-derived AR-V transcriptome by comparing with a previously generated AR-V signature from CWR22Rv1-AR-EK cells (32). Reassuringly, we observed a 34% overlap in DEGs between AR-V knockdown in CWR22Rv1 (siCE3) and AR depletion in the AR-V–only CWR22Rv1-AR-EK derivative (siAREx1), supporting the utilization of these gene sets to study transcriptional interplay with MFAP1 (Supplementary Fig. S12A–S12C; see Supplementary Table S3 for overlapping and non-overlapping gene lists). Importantly, examination of the MFAP1 transcriptome indicated respective 26% and 24% overlaps with the AR-V–driven gene expression signature from CWR22Rv1 (siCE3) and CWR22Rv1-AR-EK (siAREx1) cells (Fig. 5B; Supplementary Fig. S13A and S13B). Consistent with this finding, gene set enrichment analysis indicated that MFAP1 depletion significantly downregulated AR hallmark and bespoke AR and some, but not all, AR-V7 gene sets (Fig. 5C; Supplementary Fig. S13C; Supplementary Table S4; ref. 35). As such, several AR-V7 target genes, such as ZWINT, PLK1, and TPX2, were markedly downregulated in response to MFAP1 depletion (Supplementary Fig. S13D), whereas several hallmark and Kyoto Encyclopedia of Genes and Genomes pathways controlled by MFAP1, including mitotic spindle and DNA replication, overlapped with those regulated by AR-Vs (Fig. 5D and E; Supplementary Fig. S14A, S14B and S15A, S15B; Supplementary Table S5A and S5B). These findings support the concept that MFAP1 controls an AR-V–driven transcriptional signature largely by modulating AR-V levels, via transcript processing, as opposed to a direct impact on AR transcriptional co-regulation.

Crucially, 74% of the MFAP1-regulated transcriptome was found to be independent of its effect on AR-Vs, with 451 and 463 SDEGs shown to be downregulated by MFAP1 depletion in CWR22Rv1 and CWR22Rv1-AR-EK cells, respectively, that are distinct from AR-V–driven genes (Supplementary Fig. S13A and S13B). As shown in Fig. 5F, analysis of these AR-V–independent MFAP1-regulated genes (see Supplementary Table S6 for gene list) identified involvement in a number of metabolic and biosynthetic pathways, including arginine, proline, and pyruvate metabolism, mTOR signaling, and cell cycle, consistent with our previous data indicating a cytostatic effect of MFAP1 knockdown in prostate cancer cells.

MFAP1 is an abundant component of the spliceosome B complex and by interacting with U5 (hPRP6) and U4/U6 (hPRP3) proteins, is thought to play an important structural role in enabling downstream splicing activity (27, 28). Hence, it was important to first assess the wider impact of MFAP1 depletion on global splicing decisions in CWR22Rv1 cells. To that end, differential splicing analysis was conducted using SUPPA2 (36), which provides a ∆proportion spliced in value that represents the abundance of alternatively spliced transcripts between our scrambled control and MFAP1-knockdown experimental arms. As shown in Fig. 6A and B, depletion of MFAP1 resulted in a total of 8,776 splicing events with a cutoff of ± 0.2 ∆proportion spliced in, 397 of which were considered significant (P value < 0.05), with alternative first exon and exon skipping identified as the predominant changes to transcript composition in response to MFAP1 manipulation (Supplementary Table S7A and S7B). Running the genes associated with the significant splicing alterations through Enrichr pathway analysis (37) identified several cellular processes potentially impacted by MFAP1 depletion, including RNA binding and RNA polymerase activity (Fig. 6C).

We next sought to better define the role of MFAP1 in regulating AR transcript processing/maturation. Our previous pre-mRNA and minigene analyses provided evidence that MFAP1 controlled AR mRNA processing downstream of transcription and facilitated AR-V7–encoding CE3 inclusion in the ectopic minigene-derived transcript, suggesting that MFAP1 may directly regulate AR-V splicing in prostate cancer. Consistent with these findings, analysis of AR transcript composition in our RNA-seq data provided compelling evidence that in response to MFAP1 knockdown, splicing of exon 3 to CE3 encoding AR-V7 is downregulated (Fig. 6D), supporting the concept that MFAP1 is a bona fide splicing factor of AR-Vs in prostate cancer. Interestingly, we found that splicing events of FL-AR–encoding exon 3 to exon 4 were modestly upregulated upon MFAP1 depletion. This is inconsistent with our minigene experiments, in which we observed no change in exon 3 to exon 4–containing transcripts, and is in stark contrast to our RT-qPCR analyses, which showed marked depletion of FL-AR transcripts upon MFAP1 knockdown. Although ill-defined, we speculate that although knockdown of MFAP1 has a positive impact on exon 3 to exon 4 splicing, its loss may deregulate splicing of downstream FL-AR–encoding exons, which destine the aberrant transcripts to nonsense-mediated decay (NMD), resulting in an overall reduction of mRNA abundance. Finally, to support our findings that MFAP1 is an AR-V splicing regulator, and to dismiss a role for MFAP1 in regulating AR transcript stability, we assessed FL-AR and AR-V7 transcript stability in response to MFAP1 depletion using actinomycin D. As expected, steady-state AR transcript levels were significantly diminished by MFAP1 knockdown [Fig. 6E (left columns)]. However, upon actinomycin D treatment for 0 to 3 hours, which blocks de novo RNA synthesis, turnover of FL-AR and AR-V7 transcripts was not enhanced [Fig. 6D (right columns)], suggesting that MFAP1 does not regulate mature FL-AR and AR-V7 mRNA stability but controls transcript metabolism at the level of splicing.

In addition to its effect on AR transcript splicing, we observed a significant downregulation of DDR-associated gene sets, nucleotide excision repair and homologous recombination, in response to MFAP1 knockdown, which was consistent with the effects of AR-V depletion (Supplementary Table S8). Given that MFAP1 controls an AR-V transcriptome, which we had previously shown includes a DDR gene signature (32), the finding that MFAP1 regulates genes involved in DNA repair was not unexpected and is highlighted in Fig. 7A, demonstrating consistent downregulation of DDR-associated genes between depletion of MFAP1 and AR. To address if compromising MFAP1 activity diminishes repair of DNA damage, CWR22Rv1-AR-EK cells depleted of MFAP1, AR-Vs, or both were subjected to 2 Gy IR treatment prior to analysis of γH2AX foci. Repair of DNA breaks was robust at 6 and 24 hours after IR treatment in the scrambled control but was significantly compromised at 6 hours in response to individual MFAP1 and AR-V knockdown (Fig. 7B; Supplementary Fig. S16). These data support our findings that both MFAP1 and AR-Vs regulate a gene signature involved in DNA repair. Interestingly, dual knockdown of MFAP1 and AR-Vs led to a further modest reduction in γH2AX resolution at 24 hours, but this was insignificant. When this experiment was repeated in MFAP1-depleted CWR22Rv1 cells treated with and without enzalutamide, there was no marked difference in DNA repair capacity in the presence or absence of AR signaling 24 hours after IR treatment (Fig. 7C), indicating that MFAP1 and AR regulation of the DDR likely converges on shared transcriptional targets. Additional experiments conducted in the AR-negative PC3 cell line, in which MFAP1 knockdown was found to enhance the abundance of γH2AX foci at 6 and 24 hours after IR treatment, further support an involvement of MFAP1 in response to DNA damage and are consistent with the findings from AR-negative HeLa cells demonstrating reduced DDR-associated gene expression upon MFAP1 knockdown (Fig. 7D; Supplementary Fig. S17; ref. 38).

Lastly, to address if the observed compromised DDR capacity in cells individually or dually depleted of MFAP1 and AR-V manifested as diminished cell viability, clonogenic assays were performed in CWR22Rv1-AR-EK cells subjected to single or double knockdown ± IR treatment. As shown in Fig. 7E, in both the presence and absence of IR treatment, individual and dual depletion of either MFAP1 or AR-V significantly decreased prostate cancer cell viability, with the latter finding being consistent with previous studies (32, 39). Importantly, the elevated sensitivity of MFAP1-depleted cells to IR suggests that MFAP1 targeting, potentially in combination with AR-targeting agents, may offer more effective combination treatments in prostate cancer. Our findings that PC3 cells show reduced viability upon MFAP1 depletion ± IR (Fig. 7F) support this idea. Overall, our study has provided insights into the control of AR-V synthesis and defined MFAP1 as a key regulator of AR-V splicing and DDR in models of advanced prostate cancer.

The generation of AR-Vs in upward of 75% of patients with prostate cancer in response to FL-AR–targeted therapies represents a key clinical challenge (11). By enabling constitutive activation of the AR signaling cascade and rendering the current repertoire of hormonal therapies ineffective, AR-Vs drive progression of disease to a more advanced and often fatal state (9, 12, 40). There remains, therefore, an unmet clinical need to effectively diminish AR-V function, which would theoretically re-sensitize advanced prostate cancer to AR pathway inhibitors and provide more durable therapies for men with prostate cancer. Although direct domain-specific targeting of AR-Vs remains challenging, inhibiting AR-V synthesis by blocking key splicing regulators represents a potentially tractable route to providing new therapeutic strategies for advanced disease (19). To date, a number of candidate-based approaches have identified splicing regulators of AR-Vs, such as U2AF65 (23), SAM68 (41), PSF (24), and RBMX (26), which has provided insights into AR transcript processing in models of advanced prostate cancer. Although a recent whole-genome CRISPR screen was conducted to identify regulators of AR signaling (42), critically, a global and unbiased screening pipeline to identify human splicing factors involved in AR-V synthesis has not been performed. Expediting such a study would help fill a key knowledge gap and furnish new therapeutic targets for novel prostate cancer interventions. With this rationale, we developed an optimized CRISPR-based screening pipeline to examine the impact of knockout of 211 individual splicing factors on AR-V generation in a newly developed CWR22Rv1-AR-EK-iCas9 derivative.

Hits that significantly downregulated AR-V levels, and impacted cell proliferation, included the previously identified AR-V splicing regulator SFPQ (24), hence validating the experimental pipeline, as well as MFAP1 and CWC22 which, to the best of our knowledge, have not been linked with AR splicing. Key validation experiments across a number of AR-V–positive and –negative prostate cancer cell line models provided evidence that whereas CWC22 regulates the protein levels of both FL-AR and AR-Vs, MFAP1 selectively regulates only AR-Vs. Consistent with diminished AR isoform abundance, depletion of these splicing factors resulted in downregulation of several canonical AR target genes and reduced cell growth of both AR-positive and -negative prostate cancer cell line models. Importantly, MFAP1 and CWC22 knockdown had nominal impact on non–cancer-derived prostate epithelial cell line models, suggesting a potential therapeutic window should these splicing factors be successfully targeted. Although we found no significant change in MFAP1 and CWC22 expression in samples of patients with prostate cancer or correlation with AR-V7 transcripts in The Cancer Genome Atlas prostate cancer cohort (Supplementary Fig. S18A and S18B), the fact that knockdown of both splicing factors was more effective in cancer versus normal prostate cells may demonstrate deregulated activity in disease independent of absolute abundance.

Further investigating how AR isoforms were reduced upon MFAP1 and CWC22 knockdown demonstrated that both splicing factors control transcript abundance of several clinically relevant AR-Vs, including AR-V7, AR-V1, AR-V6, and AR-V9, as well as FL-AR, although this did not translate to reduced FL-AR protein levels in cells lacking MFAP1, the mechanism of which remains ill-defined. Crucially, depletion of both splicing factors did not impact AR pre-mRNA levels, suggesting that these factors are acting posttranscriptionally to regulate AR-Vs. Using an AR minigene reporter plasmid containing partial introns adjacent to exon 3, CE3, and exon 4, to allow respective AR-V– (exon 3–CE3) and FL-AR–encoding (exon 3–exon 4) splicing decisions to be assessed, we found that knockdown of CWC22 downregulated both CE3 and exon 4 inclusion, whereas MFAP1 selectively reduced CE3 inclusion without impacting FL-AR–encoding events. This latter finding is at odds with the fate of endogenous FL-AR transcripts in our RT-qPCR analysis and may suggest that MFAP1 depletion has deleterious effects on the composition of wild-type AR isoforms downstream of exon 4, although this remains to be established.

CWC22 has a number of key roles in mRNA processing, including (i) formation of the exon-joining complex core (29); (ii) binding of eIF4A3 into the spliceosome C complex for pre-mRNA splicing (30); and (iii) NMD of transcripts containing premature stop codons (31). We speculate that CWC22 knockdown diminishes both FL-AR and AR-V transcripts because of impaired pre-mRNA splicing, resulting in improper intron retention and NMD of these transcripts. However, given the selective function of MFAP1 in controlling AR-V abundance, and AR-V7–encoding exon splicing activity in the minigene experiments, we chose to focus our attention on MFAP1-mediated regulation of AR-V synthesis.

MFAP1 is a component of the spliceosome B complex and interacts with U5 (hPRP6) and U4/U6 (hPRP3) proteins (27, 28, 43), suggesting a role of MFAP1 in tri-snRNP recruitment during B complex formation. Additionally, MFAP1 was found to co-immunoprecipitate with core spliceosome protein SF3B1 and co-localized with SC35, a marker of splicing speckles, implicating a role of MFAP1 in the spliceosome (38). Intriguingly, SF3B1 has been reported to control synthesis of FL-AR and AR-V7 in CRPC, and its expression was upregulated in samples of patients with advanced prostate cancer (44). As such, the observed reduction of AR-V transcripts upon MFAP1 depletion could be due to improper tri-snRNP recruitment during B complex formation and compromised initiation of pre-mRNA splicing. It is therefore possible that MFAP1 functions to facilitate recruitment of already defined AR-V splicing factors to enable synthesis of AR-V7 transcripts. Therefore, to better define its role in regulating AR pathway activity and transcript processing, global transcriptomic analysis was performed in prostate cancer cells depleted of MFAP1. Consistent with our RT-qPCR profiling of canonical AR target genes, knockdown of MFAP1 significantly downregulated the AR hallmark and bespoke AR-V7 gene sets (35), supporting the concept that MFAP1 modulates AR-V levels to indirectly impact global AR signaling.

Although MFAP1 depletion compromises AR/AR-V–positive prostate cancer cell growth, it remains currently unclear as to what extent this is driven by reduced AR signaling compared with other proproliferative pathways. Critically, we identified a large cohort of MFAP1-regulated transcripts totaling 74% that were independent of those controlled by AR-Vs and involved in numerous cellular pathways, including apoptosis, mTOR signaling, and cell-cycle regulation. These findings align to previous studies demonstrating that compromised MFAP1 activity impacted HeLa cell division (45) and apoptosis (38), which is also consistent with our AR-negative PC3 cell line data which demonstrated reduced growth upon MFAP1 knockdown. Although ill-defined, the effect of MFAP1 depletion on abundance of mRNAs outside of its effect on AR-Vs may be a consequence of deregulated splicing activity causing rapid transcript turnover via NMD or, as observed for the AR-V transcriptome, downregulation of transcription factors which would directly alter gene expression programs. Indeed, global splicing analysis revealed that MFAP1 depletion caused 397 significant splicing alterations with alternative first exon and skipping exon being the most common, suggesting that differential transcript composition and metabolism may expedite differential gene expression observed in cells lacking MFAP1. Importantly, a detailed analysis of AR transcripts in CWR22Rv1 cells indicated that AR exon inclusion patterns were disrupted in response to MFAP1 knockdown in which selective downregulation of AR-V7– and AR-V9–encoding exon 3 to CE spliced junctions was observed without effect on constitutive FL-AR–encoding exon 3 to exon 4 splicing. These specific changes in AR-V transcript composition provides further evidence that MFAP1 directly regulates AR mRNA processing to facilitate the synthesis of AR-Vs. Interestingly, comparing these AR splicing alterations with those observed in a recent study of the splicing factor RBMX, in which exon 3 splicing to all downstream CEs and constitutive exons was diminished in response to RBMX knockdown (26), suggests that MFAP1 works at distinct stages of splicing to RBMX even though they regulate similar global splicing activities (alternative first exon and skipping exon being most commonly altered events).

Finally, consistent with a study demonstrating that MFAP1 regulated a DDR gene signature in HeLa cells (38), we found that several DDR-associated pathways were significantly downregulated in response to MFAP1 depletion, including nucleotide excision repair, base excision repair, and homologous recombination. This was not unexpected considering MFAP1 knockdown reduces AR-V abundance, and both AR-Vs and FL-AR have been shown to regulate a transcriptional program of DNA repair genes in response to DNA damage (32, 46). That said, we identified several DDR-associated genes downregulated by MFAP1 depletion either independently of or cooperatively with AR-Vs, suggesting that targeting MFAP1 may sensitize AR-V–positive and –negative prostate cancer cells to IR. Consistent with these observations, assessment of DNA repair kinetics in response to IR in AR-V–positive CWR22Rv1-AR-EK and AR-null PC3 cells revealed that MFAP1 depletion reduced DNA repair capacity, providing tangible evidence that MFAP1 regulates the DDR irrespective of AR status. Consistent with impaired repair of IR-induced DNA breaks, cells depleted of MFAP1 reduced colony-forming ability of both AR-V–positive and AR-negative prostate cancer cell lines and can further enhance sensitization of CRPC cells to IR.

Overall, we have shown that MFAP1 and CWC22 play important roles in AR-V synthesis, with MFAP1 selectively regulating splicing decisions to enable AR-V transcript maturation in prostate cancer cells. This translates to reduced AR-V signaling and downregulation of both AR-V–dependent and –independent DDR gene signatures that could provide therapeutic vulnerabilities in CRPC by targeting these splicing processes.

C. Crafter reports employment with AstraZeneca and ownership of AstraZeneca shares. D.J. O’Neill reports employment with AstraZeneca and ownership of AstraZeneca shares. No disclosures were reported by the other authors.

L. Walker: Data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. R. Duncan: Investigation, methodology, writing–review and editing. B. Adamson: Investigation, methodology, writing–review and editing. H. Kendall: Methodology. N. Brittain: Investigation. S. Luzzi: Investigation, methodology. D. Jones: Resources, methodology. L. Chaytor: Resources, supervision, writing–review and editing. S. Peel: Resources, software, formal analysis. C. Crafter: Conceptualization, formal analysis, supervision, funding acquisition, writing–review and editing. D.J. O’Neill: Conceptualization, resources, software, formal analysis, supervision, funding acquisition, visualization, methodology, writing–review and editing. L. Gaughan: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, writing–original draft, project administration.

The authors thank Anna Dickson for helping with crRNA library resuspension and Echo dosing and Morag Hunter for her assistance with image acquisition. L. Walker is supported by a BBSRC iCase Award and Prostate Cancer Research (PCR-6955); R. Duncan is supported by the European Regional Development Fund (IIIP-NE; 25R17PO1847); B. Adamson is supported by the Ken Bell Bursary and JGW Patterson Foundation (12/21 NU009331); D. Jones is supported by Prostate Cancer Research (PCR-6955); H. Kendall is supported by The Barbour Foundation; and N. Brittain is supported by Cancer Research UK (C9380/A25138).