Targeting Protein Arginine Methyltransferase 5 Suppresses Radiation-induced Neuroendocrine Differentiation and Sensitizes Prostate Cancer Cells to Radiation

Prostate cancer is the second leading cause of cancer death among American men (1). Radiotherapy (RT) is a potentially curative treatment for localized prostate cancer (2, 3). Despite radiotherapy treatment, 10%–15% of patients with low-risk disease and up to 50%–60% of patients with high-risk disease experience recurrence within 5 years, among which approximately 20%–30% of patients die within 10 years (4–6). Currently, androgen deprivation therapy (ADT), which targets androgen receptor (AR) signaling and decreases the efficiency of DNA double-strand break (DSB) repair (7, 8), is used as a radiosensitization approach to treat high-risk disease (2, 4–6). However, the effect of ADT remains limited and the adverse effects of ADT significantly diminish the quality of life in treated patients (9). Thus, developing better radiosensitization approaches for prostate cancer radiotherapy is urgently needed.

Protein arginine methyltransferase 5 (PRMT5) is an emerging epigenetic regulator and type II methyltransferase that can symmetrically dimethylate arginine residues (e.g., histones H4R3, H3R2, H3R8, and H2AR3) to regulate transcription of its target genes (10, 11). PRMT5 also posttranslationally methylates non-histone protein substrates such as EGFR, p53, and Raf to modulate their functions. As such, PRMT5 regulates many cellular processes such as cell proliferation, differentiation, RNA splicing, and DNA damage response (DDR), and its overexpression correlates with disease progression and poor clinical outcomes in many types of human cancer (10, 11). We have demonstrated that PRMT5 is overexpressed in prostate cancer cells and human prostate cancer tissues to promote prostate cancer cell growth and confer therapy resistance (12–14). Mechanistically, PRMT5 epigenetically promotes transcription of AR and DNA DSB repair genes by catalyzing symmetric dimethylation of H4R3. In addition, PRMT5 may also function as an oncogene to promote prostate cancer cell growth and confer resistance to therapies independent of its epigenetic regulation of AR and DSB repair genes (15).

Neuroendocrine differentiation (NED) in prostate cancer is a process by which prostate adenocarcinoma cells transdifferentiate into neuroendocrine-like (NE-like) cells (16, 17). NED is an emerging mechanism of resistance to cancer treatments (18, 19) and contributes to disease progression and the development of treatment-induced neuroendocrine prostate cancer (NEPC), a highly aggressive and end-stage disease (20–24). By mimicking clinical radiotherapy, we demonstrated that prostate cancer cells undergo NED in response to fractionated ionizing radiation (FIR; refs. 25–27). Because PRMT5 can epigenetically activate transcription of AR and DSB repair genes to promote DSB repair (13, 14) and because the acquisition of radioresistance during the first 2 weeks of FIR is an essential phase toward FIR-induced NED, we evaluated the role of PRMT5 in FIR-induced NED. Here, we demonstrate that PRMT5 and its cofactors pICln and MEP50 are essential mediators of FIR-induced NED. Significantly, targeting PRMT5 in vitro and in prostate cancer xenograft tumors in mice validated PRMT5 as a therapeutic target for prostate cancer radiosensitization.

LNCaP (CLS, catalog no. 300265/p761_LNCaP, RRID:CVCL_0395), DU145 (CLS, catalog no. 300168/p708_DU-145, RRID:CVCL_0105), PC-3 (CLS, catalog no. 300312/p1699_PC-3, RRID:CVCL_0035), and VCaP (ATCC, catalog no. CRL-2876, RRID:CVCL_2235) cells were purchased from ATCC and cultured according to ATCC's instructions and as described previously (26, 27). Upon arrival, the cell lines were immediately expanded and aliquots were prepared and stored in liquid nitrogen. Cells were maintained for no longer than 30 passages or no longer than 3 months. LNCaP Cell line authentication was performed by IDEXX BioResearch (IMPACT I). The absence of Mycoplasma contamination was routinely verified using LookOut PCR Mycoplasma Detection Kit (Sigma). Knockdown cell lines were generated using the pLKO-Tet-On system (RRID:Addgene_90243) as described previously (28). The pLKO-Tet-On plasmid for short hairpin RNA (shRNA) expression was obtained from Addgene (58), and shRNA sequences that target PRMT5 #1 (referring to #1832), PRMT5 #2 (referring to #1577), scramble control (SC), MEP50, and pICln were used for the construction of plasmids for stable cell line generation as described previously (12–14). Stable cell lines with Dox-inducible expression of PRMT5-targeting shRNA (LNCaP-shPRMT5 #1, LNCaP-shPRMT5 #2, or LNCaP-shSC) were established from individual clones and characterized previously (12–14). Stable cell lines with Dox-inducible expression of MEP50-targeting shRNA (LNCaP-shMEP50) and pICln-targeting shRNA (LNCaP-shpICln) were established from individual clones and characterized previously (14). A list of all cell lines used in this study is included Supplementary Table S1.

The animal experiment was performed in the Biological Evaluation Facility of the Purdue University Center for Cancer Research and was approved by the Purdue University Animal Care and Use Committee. Forty male non-obese diabetic-Rag1(null)-γ chain(null) (NRG) mice (6–8 weeks old) were randomly divided into two groups for implantation of single cell–derived doxycycline (Dox)-inducible stable cell lines, LNCaP-shPRMT5 and LNCaP-shSC, to establish xenograft tumors (13, 14). After tumor volumes reached approximately 200–300 mm³, xenograft tumors were subjected to FIR (5 Gy/fraction, 2 fractions/week) while mice were fed with Dox (1 mg/mL in drinking water) to induce expression of shRNAs. The cotreatment (FIR and knockdown) was terminated at the end of 4 weeks (cumulative 40 Gy) and mice resumed normal drinking water without Dox. Tumor volume was measured every 2–3 days using (½ × Length × Width × Height) without blinding method. When tumors reached approximately 1,500 mm³, mice were sacrificed and tumors were resected for IHC analysis. The experiment was terminated when all mice in the control group (LNCaP-shSC) died of tumor burden (>1,500 mm³). All other details including power calculation, tumor recurrence, survival, tissue processing, and IHC analysis are provided in Supplementary Materials and Methods.

All studies involving patients with prostate cancer were conducted in full accordance with the guidelines for Good Clinical Practice and the Declaration of Helsinki. Patients with prostate cancer treated at Duke Cancer Institute Urology Clinic were diagnosed for biochemical failure after radiotherapy by following the ASTRO or Phoenix criteria. Biopsy specimens were reviewed by a GU specialized pathologist to confirm the presence of recurrent prostate cancer. Patients treated at NCI were enrolled with signed informed consent for participation in an Institutional Review Board–approved trial that prospectively collects tissues for research at the time of diagnostic biopsy for biochemical recurrence after prior radiotherapy for prostate cancer (NCT01834001). A total of 11 pairs of treatment-naïve (Pre-RT) and recurrent prostate cancer (Post-RT) specimens from the same patients were identified for IHC analysis of PRMT5 expression. Semiquantification of PRMT5 expression was performed and paired Student t test was used to determine the difference in PRMT5 expression between Pre-RT specimens and Post-RT specimens.

All other methods used in this article have been described previously (12–14, 25–27). All methods including the above two new methods are provided in detail in Supplementary Materials and Methods.

All lentiviral knockdown plasmids and the stable cells lines generated in this study are available upon request addressed to Chang-Deng Hu at hu1@purdue.edu. RNA sequencing (RNA-seq) data have been deposited in Gene Expression Omnibus (accession no: GSE163109).

We previously reported an in vitro and in vivo model of FIR-induced NED of prostate cancer cells, in which cells underwent NED and transdifferentiated into NE-like cells after a cumulative dose of up to 40 Gy FIR (2 Gy/fraction, 5 fractions/week; refs. 19, 25–27). To determine the clinical relevance of FIR-induced NED, we performed RNA-seq analysis on prostate cancer cells treated with a cumulative dose of 40 Gy FIR. Comparing irradiated (FIR+) LNCaP cells with non-irradiated cells (FIR−), we identified 2,493 differentially expressed genes (DEGs) including 1,346 upregulated and 1,147 downregulated genes (Fig. 1A). Gene Ontology (GO) analysis identified NED, cell–cell signaling, and ion channels as upregulated GOs in FIR-induced NE-like cells while cell-cycle progression, cell division, homologous recombination (HR), and response to ionizing radiation (IR) were significantly enriched in downregulated genes after FIR (Supplementary Fig. S1). Similar results were obtained by gene set enrichment analysis (GSEA) on all genes rather than DEGs (Supplementary Fig. S2). These findings are consistent with previous reports: (i) FIR induces NED (19, 25–27), (ii), NE-like cells secrete signaling molecules to support tumor growth and cell survival (29, 30), (iii) NE-like cells express ion channels to facilitate intracellular/extracellular signaling pathways (31, 32), (iv) NE-like cells are resistant to apoptosis and do not divide (33–35), (v) non-proliferating cells utilize non-homologous end joining (NHEJ) over HR (36), and (vi) NE-like cells survive FIR treatment (19, 25–27).

To further confirm the NEPC-associated features, we performed qRT-PCR to examine the expression of the 12-gene signature of NEPC: nine upregulated and three downregulated genes (37). Eight of nine genes except for EZH2 were significantly upregulated in FIR-induced NE-like cells (Fig. 1B). Three additional NED-associated genes (BRN2, NEUROD1, and NSE) were also upregulated. However, the three downregulated NEPC signature genes (AR, REST, SPDEF) showed no significant change or a slight increase in FIR-induced NE-like cells. Given that HR was downregulated in FIR-induced NE-like cells, we also examined the expression of recently identified PRMT5 target genes involved in HR and NHEJ, and confirmed that most of PRMT5-regualted HR genes were down-regulated whereas the expression of NHEJ genes were unaltered (Fig. 1B).

We further analyzed gene expression profiles from clinical NEPC tissues and compared them with prostate adenocarcinoma tissues to identify DEGs and associated pathways. The majority of GO functions and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways associated with NED (22/24), cell–cell signaling (15/23), and ion channel regulation (20/22) were activated in both FIR-induced NE-like cells and clinical NEPC (Fig. 1C). Conversely, biological processes and pathways involved in cell-cycle progression and cell division were inhibited in FIR-induced NE-like cells whereas they were activated in NEPC (Fig. 1D). This was expected as clinical NEPC is highly aggressive/proliferative while NE-like cells do not proliferate. Collectively, the gene expression profiling suggests that FIR-induced NE-like cells possess some key features of NEPC and hence it can represent a clinically relevant NED model for mechanistic study and therapeutic evaluation.

Because a single dose of IR (2 Gy) induces PRMT5 expression to promote DSBs repair (14), we next determined whether PRMT5 protein expression is altered during the course of FIR-induced NED. FIR dose-dependently increased PRMT5 expression in LNCaP cells (Fig. 2A and B). Similar FIR-induced upregulation of PRMT5 protein was also observed in other prostate cancer cell lines DU145, PC-3, and VCaP (Supplementary Fig. S3). We also examined the expression level of PRMT5 in previously isolated LNCaP radiation-resistant sublines (IRR; ref. 26). These sublines were isolated from colonies that regrew 3 months after 40 Gy FIR treatment and were cross-resistant to radiation, androgen depletion, and docetaxel treatments. Comparing with parental cells (wt), all four sublines (IRR 1–4) showed elevated PRMT5 expression (Fig. 2C and D). To assess the clinical relevance of these observations, we performed IHC analysis of PRMT5 protein expression in paired primary treatment-naïve prostate cancer tissues (Pre-RT) and recurrent prostate cancer tissues (Post-RT) from the same patients. Indeed, PRMT5 expression was significantly elevated in Post-RT tissues compared to Pre-RT tissues (Fig. 2E and F). Even though several Pre-RT tissues already showed high PRMT5 expression (e.g., patient 1), a further increase in PRMT5 expression was observed in their Post-RT tissues. Nine Post-RT tissues showed very high PRMT5 expression score (50–60). Taken together, these results suggest that PRMT5 may play an important role in FIR-induced NED, acquisition of radioresistance, and tumor recurrence following radiotherapy.

The clinical relevance of FIR-induced NED, the role of PRMT5 in DSB repair, and the upregulation of PRMT5 protein during FIR-induced NED prompted us to evaluate whether targeting PRMT5 can inhibit FIR-induced NED and radiosensitize prostate cancer cells. In our FIR-induced NED model in prostate cancer cells (LNCaP, DU145, and PC3; refs. 19, 25–27), there appears to have two phases during the 40 Gy FIR treatment: (i) radioresistance acquisition and (ii) NED. The RA phase mainly occurs during the first 2 weeks (0–20 Gy) when the majority of cells die while the surviving cells begin to display NE-like morphologic changes and express NE-associated proteins. The NED phase occurs during the final 2 weeks (20–40 Gy) and NE-like protein expression and morphology increases. The morphologic changes of FIR-induced NED are shown in Fig. 3A (control) as reported previously (26). Using a previously characterized LNCaP-shPRMT5 cell line (13, 14), we assessed how targeting PRMT5 affects the different phases of FIR-induced NED and cell survival. LNCaP-shSC cells, and LNCaP-shPRMT5 cells without Dox treatment were used as controls.

Consistent with the role of PRMT5 in DSB repair (14), PRMT5 knockdown during the RA phase only (0–20 Gy, Dox+) killed almost all cells and prevented surviving cells from undergoing NED in response to subsequent FIR (Fig. 3A–C; Supplementary Figs. S4 and S5). Importantly, PRMT5 knockdown during the NED phase only (30–40 Gy, Dox+) reversed NE-like morphology (shrinkage of neurite length) and killed cells that had already underwent NED. At the end of FIR treatment, less than 0.05% of cells survived when PRMT5 was targeted during any phase. In comparison, approximately 15% cells survived in controls. To corroborate this finding, we performed similar experiments by inhibiting PRMT5 catalytic activity with our PRMT5 inhibitor BLL3.3 (12–14, 38) and confirmed similar results as PRMT5 knockdown (Fig. 3D–F; Supplementary Figs. S4 and S5). Similar results were also obtained in DU145 cells treated with BLL3.3 (Supplementary Fig. S6). Consistent with morphological changes, the NED biomarker neuron specific enolase (NSE) expression was already detectable after 10 Gy of FIR whereas FIR-induced NSE expression was significantly inhibited by PRMT5 knockdown in two different Dox-inducible PRMT5 knockdown cell lines (Fig. 3G–J). Because PRMT5 knockdown virtually killed all cells by 20–40 Gy of FIR, we were unable to evaluate the inhibitory effect of PRMT5 knockdown on the expression of NED biomarkers such as NSE and chromogranin A (CgA). Given that PRMT5 knockdown only slows down cell growth without inducing cell death in AR-expressing prostate cancer cells (Supplementary Fig. S7; refs. 12, 13) and that exogenously expressed PRMT5 significantly prevented cells from FIR-induced cell death even during the first week irradiation (Supplementary Fig. S8), these results together demonstrate that PRMT5 is required for the acquisition of radioresistance, induction of NED, maintenance, and survival of NE-like cells. Thus, targeting PRMT5 can effectively sensitize prostate cancer cells to radiation.

Because PRMT5 subcellular localization is important to its biological functions (10, 11), we next determined whether FIR may alter its subcellular localization during the course of FIR-induced NED. In order to assess FIR-induced NE-like cells exclusively, we extended the FIR dose to 70 Gy and found that almost all survived cells were NE-like (Fig. 4A). Consistent with morphologic changes, FIR also dose-dependently increased the percentage of CgA+ cells, a biomarker of NE-like cells (16, 26) (Fig. 4B; Supplementary Fig. S9). Up to 40 Gy FIR, almost all survived cells are CgA+.

After confirming the dose-dependent induction of NED by FIR, we assessed the subcellular localization of PRMT5. Consistent with the dose-dependent induction of PRMT5 in Western blotting analysis, immunocytochemistry analysis conferred similar results (Fig. 4C; Supplementary Fig. S9). Although FIR increased both nuclear and cytoplasmic PRMT5 staining, we observed a slight increase in the nuclear:cytoplasmic (N:C) ratio at higher cumulative doses of FIR (where almost all surviving cells are NE-like), suggesting that nuclear PRMT5 may be particularly important in NE-like cells.

We also examined the expression and subcellular location of two PRMT5 cofactors MEP50 and pICln and observed that FIR increased their expression (Fig. 4D and E; Supplementary Fig. S9). Interestingly, pICln became predominantly nuclear localized upon FIR treatment whereas its expression in the cytoplasm was largely unaltered during FIR. The increase in MEP50 expression was less drastic than that of pICln; however, the subcellular distribution in cells treated with FIR was similar to pICln (predominantly nuclear). To substantiate this finding, we performed subcellular fractionation in non-irradiated cells and in FIR-irradiated NE-like cells and obtained similar results (Supplementary Fig. S10). Taken together, these results suggest that increased nuclear levels of PRMT5 and its cofactors pICln and MEP50 may be important to FIR-induced NED.

The dose-dependent induction of PRMT5, MEP50, and pICln proteins by FIR suggests that MEP50 and pICln may contribute to PRMT5-mediated NED upon FIR. We followed the same targeting strategy as shown in Fig. 3 to evaluate the effect of pICln or MEP50 knockdown on FIR-induced NED and cell killing by using our single cell–derived Dox-inducible knockdown stable cell lines: LNCaP-shpICln and LNCaP-shMEP50 (12, 14). Knockdown of pICln at any phase sensitized LNCaP cells to FIR (Fig. 5A and B; Supplementary Fig. S4). During the RA phase, pICln knockdown was more effective than PRMT5 knockdown to prevent NED with a better killing effect (Fig. 5A and B). Contrastingly, during the NED phase, pICln knockdown was less effective than PRMT5 knockdown at reversing NE-like morphology and killing cells that had already undergone NED (Fig. 5A and C; Supplementary Figs. S4 and S5). This suggests that pICln is more important to the RA phase. Interestingly, although MEP50 is not involved in DSB repair (14), MEP50 knockdown during any phase sensitized LNCaP cells to FIR (Fig. 5D and E; Supplementary Fig. S4). MEP50 knockdown on cells that had already undergone NED reversed NE-like morphology more effectively and faster than PRMT5 or pICln knockdown (Fig. 5D and F; Supplementary Fig. S5). Collectively, these results suggest that pICln may be particularly important to survival by preventing NED during the RA phase whereas MEP50 may be critical to the induction and maintenance of NED during FIR-induced NED.

We next determined whether targeting PRMT5 can sensitize prostate cancer xenograft tumors to radiation in a preclinical mouse model by following our previous protocol (Fig. 6A; ref. 27). We used Dox-inducible LNCaP-shPRMT5 and LNCaP-shSC cells to establish xenograft tumors. Once tumors grew to approximately 200–300 mm³, mice were given Dox-containing drinking water to establish and maintain PRMT5 knockdown (shPRMT5), or express SC (shSC) in the tumors. Following 3-day pretreatment with Dox-containing water (enough time for efficient PRMT5 knockdown), tumors were subjected to their first IR treatment. Tumors were treated with FIR (5 Gy/fraction, 2 fractions/week) for a cumulative dose of 40 Gy while mice were continuously fed with Dox-containing drinking water. After completion of the 4-week cotreatment with FIR and Dox, mice resumed normal drinking water. Tumor growth was monitored twice weekly and mice with tumors larger than 1,500 mm³ were sacrificed because of tumor burden.

Control shSC tumors continued to grow during FIR to an average of approximately 135% of their original volume (Fig. 6B). About 65% of these shSC tumors were larger following 40 Gy FIR. However, only one of these shSC tumors was larger than 400 mm³ (cutoff for tumor recurrence), indicating that during FIR, the growth of shSC tumors was mostly suppressed. Conversely, 100% of the shPRMT5 tumors shrunk following the FIR to an average size of approximately 25% of the original volume (Fig. 6B), indicating that PRMT5 knockdown sensitized the prostate cancer tumors to FIR treatment. The shSC tumors began to regrow almost immediately following the completion of FIR/Dox cotreatment and continued until sacrificed because of tumor burden. Strikingly, the tumor volume for 17 of 20 mice in the shPRMT5 group became less than 100 mm³ following FIR and remained negligible until termination of the experiment. As only tumor volumes from live mice were used to calculate the average tumor volume, the average tumor volume of the shSC group began to fluctuate when mice with large tumors begun to die (arrow on Fig. 6B).

Ninety-five percent of the shSC tumors recurred within 85 days following the end of FIR/Dox treatment (Fig. 6C). Only 15% (3/20) of shPRMT5 tumors recurred with significant delay. It took 141–155 days following the end of FIR treatment for these three shPRMT5 tumors to recur. The first death in the shSC group was observed at 85 days after the end of FIR treatment, followed by 50% of the others mice with shSC tumors died after another 28 days (Fig. 6D). In contrast, 85% of the mice with shPRMT5 tumors survived throughout the experiment. The other 3 mice with shPRMT5 tumors succumbed much later than their shSC counterparts (162–178 days following the end of FIR treatment). They survived on average about 3 months longer than mice with shSC tumors.

Consistent with our previous observation that FIR induces NED in LNCaP xenograft tumors (27), the percentage of CgA+ cells in recurrent shSC tumors was significantly elevated when compared with shPRMT5 tumors (Fig. 6E). These results confirmed our in vitro observation that targeting PRMT5 can effectively inhibit FIR-induced NED also applies in vivo.

We also assessed the expression and subcellular localization of PRMT5, pICln, and MEP50 proteins. All three proteins were highly expressed in almost all recurrent tumors (Fig. 6E; Supplementary Fig. S11A). Comparing shSC and shPRMT5 tumors, the only obvious trend in subcellular localization was that pICln was highly expressed in the nucleus of the three recurrent shPRMT5 tumors. We further determined that the percentage of CgA+ cells correlated with expression of both nuclear PRMT5 and nuclear MEP50 but not nuclear pICln expression in shSC recurrent tumors (Fig. 6F; Supplementary Fig. S11B), suggesting that nuclear PRMT5 and MEP50 protein may promote FIR-induced NED. Indeed, co-immunofluorescence analysis confirmed that nuclear PRMT5 or MEP50 expression, but not pICln, positively correlated with the expression of CgA in the same cells (Supplementary Fig. S12). Collectively, our results provide in vivo evidence that targeting PRMT5 is an effective radiosensitization approach.

Treatments of hormone naïve prostate cancer with the first-line therapies such as radiotherapy and ADT can induce a subset of prostate adenocarcinoma cells to undergo NED and differentiate into NE-like cells, which are further selected to progress to NEPC after second-generation ADT (abiraterone and enzalutamide; refs. 18–24). Although ADT-induced NED and castration-resistant NEPC have been extensively studied, radiation-induced NED and NEPC remain largely uninvestigated. We previously demonstrated that FIR induces NED in prostate cancer cells and prostate xenograft tumors in mice, and that a subset of patients with prostate cancer show elevated plasma CgA levels following radiotherapy (26, 27). To better characterize FIR-induced NE-like cells, we profiled gene expression changes upon FIR and identified DEGs associated with FIR-induced NED. Comparing with the expression profile from clinical NEPC tissues (39), we confirmed that FIR-induced NE-like cells share some upregulated pathways such as NED, cell–cell signaling, and ion channel regulation. Unlike clinical NEPC, FIR-induced NE-like cells exhibit downregulation of cell cycle and cell division, consistent with the fact that clinical NEPC is highly proliferative whereas NE-like cells are non-proliferating (16, 33, 40). In summary, our comparative transcriptomic analysis provides evidence that FIR-induced NED represents a clinically relevant model for mechanistic study and therapeutic evaluation.

On the basis of our findings that PRMT5 cooperates with pICln to promote DSB repair and that targeting PRMT5 sensitizes prostate cancer cells to radiation (14), we assessed the role of PRMT5 and its cofactors pICln and MEP50 in FIR-induced NED and confirmed their role in this process. First, FIR induced PRMT5 protein expression dose-dependently during the course of FIR-induced NED in multiple prostate cancer cell lines. The protein expression of pICln and MEP50 was similarly elevated. Second, targeting PRMT5 via Dox-inducible knockdown or pharmacological inhibition with BLL3.3 in both LNCaP and DU145 cells effectively sensitized prostate cancer cells to FIR. During any singular phase (RA or NED), targeting PRMT5 killed more than 99.5% of cells, comparable with the effect of PRMT5 targeting during the entire 4-week FIR treatment. Similar results were obtained when MEP50 or pICln was knocked down. Third, Dox-induced knockdown of PRMT5 during FIR-induced NED in LNCaP xenograft tumors in mice significantly sensitized xenograft tumors to FIR, delayed tumor recurrence, decreased recurrence rate, and subsequently prolonged the overall survival. Fourth, 15% of recurrent shSC tumor cells are CgA+, consistent with our previous observation (27). However, CgA+ cells in the three recurrent tumors from the shPRMT5 group were extremely rare, suggesting that the reduced tumor recurrence associated with PRMT5 knockdown is likely due to the inhibition of FIR-induced NED. These three recurrent tumors likely developed resistance via NED-independent mechanisms.

The mechanisms underlying the role of PRMT5 and its cofactors in FIR-induced NED remain to be investigated. We confirmed the upregulation of eight of nine NEPC signature genes, suggesting that PRMT5 may epigenetically regulate gene expression changes during FIR-induced NED. The induction of SOX2, BRN2, FOXA2, SRRM4, and ASCL1 is significant and interesting considering that these are master transcriptional or splicing regulators of treatment-induced NED (41–45). However, it is not feasible to use chromatin immunoprecipitation sequencing or ChIP-qPCR to evaluate whether PRMT5 is an epigenetic regulator of these critical NED genes or other master regulators of NED, because almost all cells were killed by a cumulative dose of 40 Gy FIR. Future studies utilizing novel technologies that require smaller number of cells will be pursued to identify target genes of PRMT5. Nonetheless, the potential epigenetic roles of PRMT5, pICln, and MEP50 in FIR-induced NED are supported by several pieces of evidence obtained already. First, PRMT5 cooperates with pICln to regulate the transcription of DDR genes (14). PRMT5 and pICln likely play a similar pro-survival epigenetic role in response to FIR, particularly during the RA phase. In contrast, MEP50, which is not involved in the epigenetic regulation of DDR genes, appears to be more important to facilitate differentiation. Second, elevated nuclear localization of PRMT5, pICln, and MEP50 protein was associated with an increase in the percentage of NE-like cells during FIR-induced NED. Third, nuclear PRMT5 protein expression positively correlated with nuclear MEP50 protein expression in recurrent tumors. Fourth, the nuclear expression of PRMT5 and MEP50 is highly correlated to the percentage of CgA+ cells and the expression of CgA in recurrent tumors.

It is possible that PRMT5 and cofactors may regulate FIR-induced NED via alternative mechanisms. As AR protein expression is downregulated in FIR-induced NE-like cells (26) and AR regulates the expression of SOX2, BRN2, and REST (41, 46, 47), PRMT5 may also promote FIR-induced NED via indirect regulation of these master regulators of NED. Because PRMT5 can activate transcription of AR in prostate cancer cells (12, 13), future research is needed to determine whether PRMT5 no longer activates transcription of AR in NE-like cells and in clinical NEPC. It is also worth noting that CREB is a critical regulator of both ADT- and FIR-induced NED (19, 25–27, 48). Indeed, increased nuclear-localized phosphorylated CREB was observed in FIR-induced NE-like cells (26, 27). In response to glucagon stimulation in primary hepatocytes, PRMT5 can be recruited to CREB target gene promoters to activate transcription of downstream genes (49). Thus, PRMT5 may promote FIR-induced NED via activation of CREB target genes.

The clinical significance of treatment-induced NED has recently been recognized as a mechanism of therapy resistance and a potential contributing factor to the emergence of NEPC (17, 20, 21). Whether targeting treatment-induced NED would represent a valid approach to sensitize prostate cancer cells to existing treatments and prevent disease progression remains largely unknown. On the basis of our findings here that targeting PRMT5 can effectively inhibit FIR-induced NED and sensitize prostate cancer cells to FIR in vitro and in xenograft tumors in mice, PRMT5 can be a novel radiosensitization target for prostate cancer treatment. Furthermore, PRMT5 targeting is clinically significant and impactful with respect to its remarkable effect to sensitize xenograft tumors to FIR, delay and decrease tumor recurrence, and prolong overall survival. Considering the epigenetic role of PRMT5 in transcription activation of both AR and several DDR genes (12–14), and the facts that both ADT and targeting DDR proteins are clinically relevant approaches to improve radiotherapy (2, 3, 5, 50), the potent radiosensitization effect of PRMT5 targeting in prostate cancer cells is likely due to triple targeting of AR, DSB repair, and FIR-induced NED. As we explored, targeting PRMT5 in NE-like cells (during weeks 3–4) is sufficient to reverse NE phenotype and sensitizes NE-like cells to FIR, indicating that PRMT5 may be required for the maintenance and survival of NE-like cells and/or NEPC cells in general. Hence, targeting PRMT5 could also be explored as a novel therapeutic approach to treat NEPC.

PRMT5 has emerged as an oncogene and is overexpressed in most human cancers reported to date (10, 11). Our findings here strongly suggest that PRMT5 contributes to prostate cancer radioresistance and is likely a therapeutic target for prostate cancer radiosensitization. This is also supported by our observation of a significantly elevated level of PRMT5 expression in recurrent prostate cancer tissues after radiotherapy failure. Currently, several phase I clinical trials are underway to evaluate the safety of PRMT5 inhibitors for patients with non–Hodgkin lymphoma and other solid tumors. Should these clinical trials deem PRMT5 inhibitors safe for human use, future clinical trials may assess the effectiveness of targeting PRMT5 as a radiosensitization approach for prostate cancer radiotherapy.

J. Huang reports personal fees from Kingmed Diagnostics, MoreHealth, OptraScan, Genetron, Omnitura, Vetonco, York Biotechnology, Genecode, VIVA Biotech, Sisu Pharma; grants from Zenith Epigenetics, BioXcel Therapeutics, and Fortis Therapeutics outside the submitted work. No disclosures were reported by the other authors.

J.L. Owens: Conceptualization, data curation, formal analysis, methodology, writing–original draft, writing–review and editing. E. Beketova: Data curation, methodology. S. Liu: Data curation, formal analysis, methodology. Q. Shen: Data curation. J.S. Pawar: Data curation. A.M. Asberry: Data curation. J. Yang: Data curation. X. Deng: Data curation, formal analysis, methodology. B.D. Elzey: Data curation, methodology. T.L. Ratliff: Resources. L. Cheng: Resources. R. Choo: Resources. D.E. Citrin: Resources. T.J. Polascik: Resources. B. Wang: Data curation, formal analysis, methodology. J. Huang: Formal analysis, supervision, funding acquisition. C. Li: Resources, funding acquisition. J. Wan: Conceptualization, formal analysis, supervision, funding acquisition, methodology, project administration, writing–review and editing. C.-D. Hu: Conceptualization, formal analysis, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

This study was partially supported by grants from U.S. Army Medical Research Acquisition Activity, Prostate Cancer Research Program (W81XWH-12-1-0346, W81XWH-13-1-0398, and W81XWH-16-1-0394) and NCI RO1CA212403, Indiana University Simon Comprehensive Cancer Center Near-Miss Initiative, and Purdue University Center for Cancer Research Small Grants. Primary RNA-seq analysis and TCGA data analysis were performed by the Collaborative Core for Cancer Bioinformatics (C³B) shared by Indiana University Simon Cancer Center (P30CA082709) and Purdue University Center for Cancer Research (P30CA023168) with support from the Walther Cancer Foundation. DNA sequencing was conducted by the Genomic Core Facility supported by NCI CCSG CA23168 to Purdue University Center for Cancer Research. We would like to thank the Indiana University Precision Health Initiative for their support. J.L. Owens was supported by the Indiana Clinical and Translational Sciences Institute (CTSI) Pre-Doctoral Fellowship, which was made possible with partial support from grant numbers TL1 TR001107, TL1 TR002531, UL1 TR001108, and UL1 TR002529 (A. Shekhar, PI) from the NIH, National Center for Advancing Translational Sciences, Clinical and Translational Sciences Award. We would also like to thank L. Cheng and Dr. Jeannie Poulson Plantenga for mentoring J.L. Owens as part of the CTSI Fellowship. A.M. Asberry was supported by NIH T32 grant NIH T32GM125620. We also thank Sandra Torregrosa-Allen and Melanie P. Currie for their technical support of animal studies.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.