Pin1 inhibition improves the efficacy of ralaniten compounds that bind to the N-terminal domain of androgen receptor

Therapies for lethal castration-resistant prostate cancer (CRPC) are an unmet medical need. One mechanism underlying CRPC and resistance to hormonal therapies is the expression of constitutively active splice variant(s) of androgen receptor (AR-Vs) that lack its C-terminus ligand-binding domain. Transcriptional activities of AR-Vs and full-length AR reside in its N-terminal domain (NTD). Ralaniten is the only drug proven to bind AR NTD, and it showed promise of efficacy in Phase 1 trials. The peptidyl-prolyl isomerase Pin1 is frequently overexpressed in prostate cancer. Here we show that Pin1 interacted with AR NTD. The inhibition of Pin1 expression or its activity selectively reduced the transcriptional activities of full-length AR and AR-V7. Combination of Pin1 inhibitor with ralaniten promoted cell cycle arrest and had improved antitumor activity against CRPC xenografts in vivo compared to individual monotherapies. These findings support the rationale for therapy that combines a Pin1 inhibitor with ralaniten for treating CRPC. Leung et al. find that the peptidyl-prolyl isomerase Pin1 targets the intrinsically disordered N-terminal domain of the androgen receptor (AR). They show that combining Pin1 inhibition with ralaniten compounds that bind to the AR N-terminal domain has enhanced antitumor activity on castration-resistant prostate cancer in xenografts, suggesting therapeutic potential.

G lobally, there were 1.3 million new cases of prostate cancer in 2018. Approximately 30% of these patients will develop recurrence after primary treatment and require hormonal therapies. Currently, all FDA-approved hormonal therapies such as abiraterone acetate and enzalutamide target the C-terminal ligand-binding domain (LBD) of full-length androgen receptor (AR). About 20-40% of castration-resistant prostate cancer (CRPC) patients have de novo resistance to these agents, and virtually all patients will acquire resistance to AR-LBD-targeted therapy within 2-3 years 1 . Most CRPC continues to depend on AR transcriptional activity to maintain tumor growth and many involve the expression of constitutively active AR splice variants (AR-Vs) that lack the LBD. The most clinically relevant AR-V is AR-V7, which is detected in one-third of CRPC patients, associated with bone metastases, and resistance to abiraterone and enzalutamide 2,3 . Common to both full-length AR and AR-Vs is the N-terminal domain (NTD) that contains all of the AR's transcriptional activity 4,5 . Thus, identifying drugs that disrupt the AR NTD should block the transcriptional activities of both fulllength AR and AR-Vs. One hurdle in finding these drugs is the intrinsic disorder of this domain.
The first and currently only drug to be tested in clinical trials that directly binds to an intrinsically disordered protein is ralaniten 6 . Ralaniten binds to Tau-5 in AR NTD 7 and inhibits the transcriptional activities of full-length AR and AR-Vs 8,9 . Proofof-concept validation for the AR NTD as a drug target and the ralaniten scaffold for developing drugs to treat CRPC was provided by the first-in-human clinical trial (NCT02606123). A second-generation ralaniten analog (EPI-7386) is currently in Phase I clinical trial (NCT04421222).
To identify approaches that would disrupt the transcriptional activity of AR NTD, we considered the peptidyl-prolyl cis/trans isomerase Pin1. This enzyme is commonly overexpressed in prostate cancer, and its expression in patient biopsies correlates with recurrence [10][11][12] . Pin1 regulates the conformation of proteins by catalyzing the isomerization of proline bonds at specific phosphorylated motifs (pSer/Thr-Pro, Pin1 motif) 13,14 . Numerous putative binding sites for Pin1 are found within the intrinsically disordered AR NTD. Pin1 is composed of a WW domain that recognizes phosphorylated substrates joined by a flexible linker to a peptidyl-prolyl isomerase (PPIase) domain to carry out its catalytic function 15,16 . In the absence of Pin1, the rate of isomerization between the cis-and trans proline is inherently slow since bond rotation is impeded by the adjacent phosphoryl group 13,17 . Proline-directed Ser/Thr kinases (CDKs and MAPKs) and phosphatases (PP2A) that recognize and modify Pin1 sites are specific to the trans-proline conformation [18][19][20] . Thus, Pin1 is the only human isomerase that recognizes a proline-directed phosphorylation site, and Pin1mediated cis/trans isomerization is proposed as a conformational switch for signal transduction 21 .
There is interest in Pin1 as a potential prognostic marker for prostate cancer, but there are no reports on whether Pin1 expression in prostate cancer influences AR signaling. Here we determine whether targeting Pin1 would be beneficial for treating CRPC. We investigated the mechanism and effect of genetic and pharmacological inhibition of Pin1 activity in CRPC and assessed the feasibility of combining a Pin1 inhibitor with ralaniten compounds using in vitro and in vivo models of CRPC.

Results
Pin1 is essential for the transcriptional activity of full-length AR. The AR NTD contains several putative binding sites for Pin1 (Fig. 1a). The levels of Pin1 protein were detected in all five of the human prostate cancer cell lines examined and were highest in androgen-sensitive LNCaP cells and lower in AR-negative PC-3 and DU145 cells (Fig. 1b). In general, Pin1 protein expression was approximately threefold higher in cell lines that express AR (LNCaP, LN95, and VCaP) than those that do not rely on AR for growth and survival (PC-3 and DU145; Supplementary Fig. 1a). In clinical samples of CRPC, transcript levels of PIN1 had a positive correlation with AR expression (Supplementary Fig. 1b). Analysis of expression of several transcriptional targets of AR in CRPC revealed positive correlations between the levels of PIN1 and KLK3 (PSA), TMPRSS2, and NKX3.1, but not FKBP5 ( Supplementary Fig. 1c-f). These results were consistent with clinical data of elevated expression of Pin1 levels in patient biopsies that are associated with an increased likelihood of developing PSA recurrence after radical prostatectomy compared to patients with low expression of Pin1 12 .
Although levels of PIN1 mRNA were elevated in clinical samples of neuroendocrine prostate cancer that have low or no expression of AR ( Supplementary Fig. 1g, h) 22 , there was no association between levels of PIN1 and AR ( Supplementary  Fig. 1i). Thus, the lower expression of Pin1 protein detected in DU145 and PC-3 cells may be a cell line-specific phenomenon. These clinical data show that Pin1 was expressed in both CRPC and neuroendocrine prostate cancer, and importantly that levels of PIN1 mRNA in CRPC were associated with expression of AR and its downstream transcriptional targets.
The catalytic activity of Pin1 can be regulated by phosphorylation by death-associated protein kinase 1 (DAPK1) and polo-like kinase 1 (PLK1), which inactivates the isomerase domain and enhances protein stability, respectively 23,24 . In prostate cancer cell lines, there was a positive correlation between protein levels of DAPK1 and Pin1 but no correlation between PLK1 and Pin1 expression ( Supplementary Fig. 1j, l). Analysis of clinical samples indicated that DAPK1 was expressed in CRPC and neuroendocrine prostate cancer, while PLK1 was weakly expressed in CRPC ( Supplementary Fig. 1k, m).
The impact of the loss of Pin1 expression on AR transcriptional activity was evaluated in Pin1 knockdown experiments using two Pin1-targeting siRNAs that target separate regions of the human PIN1 transcript (Fig. 1c, d). Transcriptional activity of full-length AR was assessed in LNCaP cells using two wellcharacterized AR-driven reporter gene constructs. Androgen induction of PSA(6.1 kb)-and PB-luciferase activities were significantly decreased with reduced expression of Pin1 compared to activities measured in cells transfected with a non-targeting control siRNA (Fig. 1e, f). Pin1 knockdown did not inhibit the activities of non-AR-driven reporters, cytomegalovirus (CMV) or AP-1-luciferase reporters (Fig. 1g,h), thereby suggesting selectivity of Pin1 expression on AR transcriptional activity.
To determine if pharmacological inhibitors of Pin1 would yield similar results to genetic knockdown of Pin1, two known inhibitors of the PPIase domain of Pin1 were employed, juglone and all-trans retinoic acid (ATRA). Consistent with previous reports 25,26 , both juglone and ATRA inhibited the isomerase activity of Pin1 in a dose-dependent manner ( Supplementary  Fig. 2a, b). 13-cis-retinoic acid (13cisRA) did not inhibit the isomerase activity of Pin1 ( Supplementary Fig. 2c), which supports specificity of the all-trans conformer as expected. Consistent with the knockdown of Pin1, both pharmacological inhibitors of Pin1 reduced the induction of PSA-and PBluciferase activity by androgen in a dose-dependent manner (Fig. 2a, b). The IC 50 values for juglone were 10.9 µM (SE = 1.11 µM, R 2 = 0.85) for PSA and 6.45 µM (SE = 1.15 µM, R 2 = 0.97) for PB, whereas ATRA was more potent and had an IC 50 of 4.60 µM (SE = 1.05 µM, R 2 = 0.74) for PSA and 3.57 µM (SE = 1.05 µM, R 2 = 0.91) for PB. The 13cisRA conformer was less potent than ATRA (Fig. 2a, b), with IC 50 values of 6.22 µM and 5.42 µM for PSA and PB, respectively, which is consistent with the conversion of 13cisRA to ATRA by cellular retinoid isomerases 27 . Also consistent with Pin1 knockdown, Pin1 inhibitors displayed selectivity for blocking AR transcriptional activity without impeding the activities of the non-AR-driven CMV-and AP-1luciferase reporters (Fig. 2c, d).
To confirm the selective effects of Pin1 inhibitors on endogenous gene expression, protein levels of PSA and transcript levels of other well-characterized target genes of full-length AR were assessed. Pin1 inhibitors blocked the androgen-induced eightfold increase in PSA protein without impacting the levels of AR protein (Fig. 2e, f). Thus, the mechanism of impeding the     Fig. 3a-d). The androgen-induced levels of FKBP5 transcript that were efficiently blocked by antiandrogen enzalutamide were poorly blocked by Pin1 inhibitor, thereby suggesting gene-specific responses ( Supplementary Fig. 3d). In the presence of androgen, levels of AR and PIN1 transcripts were not significantly affected by either antiandrogen or Pin1 inhibitor ( Supplementary Fig. 3e, f). There is one putative Pin1-binding motif (Ser650-Pro651) within the AR hinge region, which harbors a bipartite nuclear localization signal required for AR nuclear translocation 28 and could potentially be impacted by inhibiting Pin1. To investigate this, cells that express a YFP-AR fusion protein were treated with Pin1 inhibitors, and cellular localization of AR was determined. As expected in the absence of androgen, YFP-AR was mainly cytosolic and then predominantly nuclear following androgen treatment (Fig. 2g). In the absence of androgen, AR remained cytosolic in the presence of Pin1 inhibitors, contrary to bicalutamide and enzalutamide which increased AR nuclear localization as previously reported 9,29 (Fig. 2g, h). Pin1 inhibitors did not prevent androgen-induced nuclear translocation of AR (Fig. 2g, h). These results suggest that the mechanism of impeding AR transcriptional activity by Pin1 inhibitors does not involve blocking AR nuclear translocation or affect the androgen-induced function of the bipartite nuclear localization signal adjacent to a potential Pin1-binding site in the hinge region.
Targeting Pin1 inhibits transactivation of AR NTD. The remaining putative Pin1-binding sites are within the intrinsically disordered AR NTD that contains transcription activation units, Tau-1 and Tau-5, to mediate androgen-dependent and androgenindependent transactivation, respectively 5,30 . The loss of AR-LBD shifts the location of transcriptional activity from Tau-1 with androgen, to Tau-5 in the absence of LBD 30 . To determine whether Pin1 impacts transactivation of AR NTD, we employed a transactivation assay that consists of a construct encoding the human AR 1-558 of the NTD fused to a Gal4DBD (Gal4-ARN) that lacks AR-LBD. Forskolin-and interleukin-6 (IL-6)-induced transactivation of AR NTD 31,32 were both blocked by treatment with an AR NTD inhibitor, ralaniten (EPI-002, positive control), as previously reported 9,33,34 (Fig. 3a, b). ATRA and knockdown of Pin1 expression blocked both forskolin-and IL-6-induced transactivation of Gal4-ARN, whereas juglone only inhibited IL-6-induced transactivation ( Fig. 3a-d).
To map the region in the AR NTD necessary for Pin1 inhibitors to mediate an effect on AR transactivation, AR NTD fragments were tested. IL-6 induced transactivation of Gal4-AR 234-391 and Gal4-AR 392-558 fragments while the Gal4-AR 1-233 fragment was not inducible and had minimal activity (Fig. 3e). Pin1 inhibitor juglone inhibited transactivation of AR 234-391 and intact AR NTD (1-558), but only had a modest effect on AR 392-558. Importantly, juglone did not inhibit the activity of Gal4DBD (empty vector) or AR 1-233. These findings suggest that inhibiting the Pin1 PPIase domain selectively blocks IL-6induced transactivation of AR NTD by involving the region of AR encoded by amino acids 234-391.
To determine whether putative Pin1 motifs located in the AR 234-391 region (Ser256-Pro257 and Ser308-Pro309) were involved in transactivation, we assessed the activity of Gal4-AR 234-391 constructs carrying proline to glycine mutations. IL-6 could still induce transactivation of the mutant AR 234-391 constructs (P257G, P309G, and P257G/P309G), but their fold-inductions were overall reduced by 33-48% (Fig. 3f). Juglone inhibited the P257G and P309G mutants to levels that were comparable to wild-type but had no effect on the P257G/P309G double mutant. Collectively these results suggest that both the Ser256-Pro257 and Ser308-Pro309 motifs were sufficient to regulate transactivation of AR NTD.
We performed pull-down assays to determine whether Pin1 interacted with AR NTD. Pin1 interaction was detected with AR 1-558, AR 234-391, and AR 392-558, but not with AR 1-233 (Fig. 3g). With the AR 234-391 fragment, Pin1 interaction was detected despite mutation of either putative Pin1-binding motifs (P257G or P309G), but interaction was minimal with the P257G/ P309G double mutant (Fig. 3h). These results imply that Pin1 can interact with either Ser256-Pro257 or Ser308-Pro309 on AR NTD, which is consistent with transactivation assays (Fig. 3f). Co-immunoprecipitation assays confirmed interaction between endogenous full-length AR and Pin1, regardless of androgen ( Supplementary Fig. 4a). We also detected interaction between Pin1 and AR-V7, following ectopic expression of AR-V7 in LNCaP cells ( Supplementary Fig. 4b). Androgen-independent transactivation of AR by IL-6 signaling involves interaction between signal transducer and activator of transcription 3 (STAT3) and residues 234-558 of the AR NTD 31 . This overlaps with the region required for Pin1 interaction (Fig. 3g). Thus, we assessed if the interaction between AR and STAT3 would be impacted by a Pin1 inhibitor. Endogenous STAT3 complexes were co-immunoprecipitated from LNCaP cells treated with juglone and IL-6. Ralaniten was included as a positive control for inhibiting STAT3 interaction with AR NTD 33 . Juglone reduced the amount of AR interacting with STAT3 by~50%, which was similar to the amount decreased by ralaniten (Fig. 3i). Juglone did not prevent phosphorylation of MAPK (isoforms p44 and p42) or STAT3 (Tyr705 and Ser727) (Fig. 3j, k). These data suggest that targeting Pin1 inhibits the interaction between the AR NTD and STAT3 by a mechanism that does not involve preventing IL-6-induced phosphorylation of STAT3 or MAPK.
Pin1 inhibitors block the proliferation of prostate cancer cells driven by full-length AR and AR-Vs. AR drives the proliferation of most prostate cancer cells. To determine whether blocking AR transcriptional activity by Pin1 inhibitors leads to decreased proliferation, BrdU incorporation assays were utilized in a battery of cell lines. Androgen-induced proliferation of LNCaP cells depends on functional full-length AR. LN95 cells express fulllength AR and AR-Vs, including AR-V7, but their proliferation is driven by AR-Vs and is not increased by androgen and thus are resistant to AR-LBD inhibitors. PC-3 and DU145 cells are devoid of functional AR and are androgen-independent. The effects of Pin1 inhibitors were compared to enzalutamide (AR-LBD inhibitor) and ralaniten (AR-NTD inhibitor) as controls. Pin1 inhibitors attenuated androgen-induced proliferation of LNCaP cells similar to validated AR antagonists (Fig. 4a). Pin1 inhibitors also attenuated the proliferation of LN95 cells along with ralaniten, whereas the AR-LBD inhibitor enzalutamide had no effect (Fig. 4b). These results are consistent with the proliferation of this cell line mediated by truncated AR-Vs that lack AR-LBD and with Pin1 regulating AR NTD. Of the Pin1 inhibitors tested, ATRA had better selectivity for blocking AR-dependent proliferation compared to juglone and did not impact the proliferation of either PC-3 or DU145 cells (Fig. 4c, d). Collectively these results support that ATRA has selectivity for blocking proliferation mediated by full-length AR or AR-Vs. Based on these findings, we focused on ATRA as a pharmaceutical inhibitor of Pin1 since its effects were consistent with knockdown of Pin1, whereas juglone has known cytotoxic effects on non-specific targets involved in mitosis 35 .
Pin1 inhibitors and knockdown block AR-V7 transcriptional activity. Pin1 interacted with the AR NTD (Fig. 3g) and reduced expression or activity of Pin1 impeded both AR NTD transactivation ( Fig. 3a-d) and AR-V-driven proliferation of LN95 cells (Fig. 4b). To directly determine if Pin1 is required for AR-V7 transcriptional activity, the V7BS 3 -luciferase reporter that is driven by three tandem repeats of an AR V7-specific promoter element from the UBE2C gene 36 was tested. As expected, V7BS 3 luciferase was highly induced by ectopic expression of AR-V7, which could not be blocked by AR-LBD inhibitor enzalutamide (Fig. 5a). Consistent with blocking AR NTD transcriptional activity, ralaniten, Pin1 inhibitors, and Pin1 knockdown all decreased AR-V7-induced reporter activity (Fig. 5a, b). Western blot analyses revealed similar levels of AR-V7 between the different treatments and siRNAs (Fig. 5c, d). Collectively, these results imply that the isomerase activity of Pin1 impacts AR-V7 transcriptional activity through the AR NTD.
Combination therapy with ralaniten and targeting Pin1. Developing better treatments for CRPC is urgently needed, especially one that inhibits the AR NTD to block the transcriptional activities of AR-Vs. A double strike against the AR NTD by combination therapy targeting Pin1 together with ralaniten was therefore investigated as a possible therapeutic approach. Since ATRA is approved clinically for other indications with known toxicity profiles, we focused on this compound in combination with ralaniten and a second-generation analog EPI-7170 that has improved potency 37 . The impact of a suboptimal concentration of ATRA (5 μM) on dose-response curves for ralaniten and EPI-7170 for inhibiting androgen-induced PSA(6.1 kb)-luciferase activity was determined. The IC 50 of ralaniten for PSA(6.1 kb) luciferase was 7.91 ± 1.15 μM (R 2 = 0.94), and the curve shifted to the left with ATRA and an improved IC 50 of 4.45 ± 1.28 µM (R 2 = 0.83) (Fig. 5e). A similar trend was observed with EPI-7170, which had an IC 50 of 0.74 ± 1.12 μM (R 2 = 0.96) as monotherapy and an IC 50 of 0.42 ± 1.16 μM (R 2 = 0.93) with ATRA (Fig. 5f). For both ralaniten compounds, the inhibitory responses in combination with ATRA exceeded the predicted additive effects determined by the Bliss independence model 38 (Fig. 5g, h). Excluding the highest concentrations (40 µM of ralaniten and 4 µM of EPI-7170), which inhibited~98% of reporter activity, the Transactivation assays performed in LNCaP cells expressing a vector encoding the human AR NTD fused to a Gal4 DNA-binding domain (Gal4-ARN). Cells were treated with vehicle (DMSO), EPI-002 (25 μM), juglone (20 μM), or ATRA (10 μM), and then incubated with IL-6 (a 50 ng/mL) or forskolin (b 25 μM) for 24 h, n = 8 independent experiments. c, d Gal4-ARN activity in cells incubated with Pin1 siRNA or a non-targeting control siRNA and then stimulated with IL-6 or forskolin for 24 h, n = 3. Transactivation assays performed using (e) Gal4-ARN fragments (n = 6), or (f) Gal4-AR 234-391 constructs (n = 3) carrying proline to glycine mutations, and treated with juglone (20 µM) and IL-6 (50 ng/mL) for 24 h. Pull-down assays performed on LNCaP cells transfected with an expression vector encoding a polyhistidine-tagged (g) AR NTD or (h) AR 234-391 construct and then incubated with IL-6 (50 ng/mL) for 6 h. Input samples show Pin1 expression from lysates before the pull-down. i Co-immunoprecipitation assay showing AR protein co-immunoprecipitated with STAT3 from LNCaP cells incubated with EPI-002 (EPI, 35 μM), juglone (JUG, 30 μM), or vehicle (VEH, DMSO), and IL-6 (50 ng/mL) for 6 h. j, k Levels of phosphorylated MAPK (isoforms p44 and p42) and phosphorylated STAT3 (Tyr705 and Ser727) from LNCaP cells pre-treated with juglone (20 µM) and then stimulated with IL-6 (50 ng/mL) for 15 min. Error bars show the mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001; ns not significant. mean combination index (CI) for ralaniten and EPI-7170 was 0.86 and 0.91, respectively. This signifies that the ralaniten compounds were moderately synergistic with ATRA.

Effect of Pin1 inhibitors and combinations on the cell cycle.
Many well-defined Pin1 substrates are mitotic proteins that are involved in regulating cell cycle progression 39 . AR-V7 is reported to regulate the expression of a distinct subset of cell cycle-related genes 40 , including UBE2C. To determine whether ATRA alters the expression of cell cycle proteins, we performed western blot analyses from LN95 cells treated with ATRA (0-10 µM) or the inactive 13cisRA conformer. ATRA binds directly to the catalytic pocket of Pin1, which inhibits the PPIase domain and induces degradation of Pin1 protein 25 . Consistent with this mechanism, a dose-dependent decrease in Pin1 levels was measured with ATRA ( Fig. 6a,b). 13cisRA did not impact the cellular levels of Pin1, thereby supporting the specificity of ATRA for Pin1. ATRA significantly decreased the expression of cyclin D1 and cyclin A2 (G 1 and S cyclins, respectively), but not cyclin B1 (M cyclin), and selectively increased the levels of cyclin-dependent kinase (CDK) inhibitor p27 Kip1 but not p21 Cip1 (Fig. 6b). Together these findings demonstrate the on-target inhibition of Pin1 by ATRA and reveal a perturbation of proteins involved in facilitating the G 1 /S transition.
Combinations of ATRA and EPI were tested for their effect on the cell cycle by flow cytometry. Analysis of BrdU incorporation and DNA content showed that 35-40% of LN95 cells were active in S phase with vehicle (Fig. 6c, column 1). ATRA reduced the S phase population in a dose-and time-dependent manner, which was accompanied by a concomitant increase in G 1 cells (Fig. 6c,  columns 1-3). Combinations of ATRA (5 and 10 μM) with ralaniten further reduced the S phase, where ralaniten with 5 μM of ATRA showed a comparable response on S phase to the combination with 10 μM of ATRA (Fig. 6c, columns 4-6). These analyses suggest that ATRA and ralaniten were synergistic in blocking the cell cycle, possibly by arresting cells in G 1 and blocking the transition to S phase. Consistent with this notion, ATRA inhibited the ability of LN95 cells to form colonies in a dose-dependent manner ( Supplementary Fig. 5a). Ralaniten also significantly reduced the number of colonies, and combination with 10 µM of ATRA had an improved response compared to either monotherapy ( Supplementary Fig. 5b). Together these results suggest that combination therapy of ATRA and ralaniten blocks the cell cycle and may irreversibly impair cell division.
Since combination of ATRA and ralaniten markedly increased the number of cells in G 1 , we sought to determine whether the response was associated with senescence by measuring senescence-associated β-galactosidase (SA β-gal) activity 41,42 . ATRA significantly increased the percentage of SA β-galpositive cells for LNCaP and LN95 cells, but not for PC-3 and DU145 cells (Supplementary Fig. 6a, b). Expression of p16 INK4A , a marker of replicative senescence and aging, was not induced by ATRA and was overall weakly expressed ( Supplementary Fig. 6c). LN95 cells treated with a combination of ralaniten and ATRA showed intense SA β-gal staining and developed a morphology that is associated with senescent cells (appearing larger and more flattened) (Fig. 6d). Approximately 65% of cells treated with combination therapy stained positive for SA β-gal, which was higher than for ralaniten (31%) or ATRA (37%) monotherapies (Fig. 6e). The CI of ATRA and ralaniten for producing a senescence response was 0.86, which indicates that the combination is slightly synergistic. Western blot analyses revealed a loss of F-box protein Skp2 (S-phase kinase-associated protein 2) and upregulation of CDK inhibitor p27 Kip1 by combination of ralaniten and ATRA (Fig. 6f). These results are consistent with the induction of a senescence response in prostate cells that depend on AR function for proliferation 43,44 . The levels of retinoblastoma protein (Rb) and phospho-Rb (serine residues 807 and 811) were also decreased by both monotherapies and combination therapy (Fig. 6g). Collectively, these findings suggest that targeting the AR NTD with ATRA and ralaniten may induce G 1 arrest and senescence in prostate cancer cells that express AR-Vs, by a mechanism involving the Skp2/p27 Kip1 pathway and Rb inactivation but not requiring p16 INK4A .
Combination therapy inhibits the in vivo growth of enzalutamide-resistant CRPC xenografts. To evaluate the efficacy of combination therapy with ATRA and ralaniten in vivo, we employed LN95-D3 cells (clonal subline) 45 . Consistent with the parental LN95 cells, enzalutamide had no effect on the cell cycle distribution of the LN95-D3 subline in the absence of androgen, whereas treatment with ATRA or ralaniten compounds reduced the S phase and increased the G 1 population (Fig. 7a, columns  1-3). Furthermore, combination therapy of ATRA with either ralaniten or EPI-7170 further decreased the S phase. This synergy was specific to combinations of ATRA and ralaniten compounds and was not apparent with combination of ATRA and enzalutamide (Fig. 7a, column 4). These data support that proliferation of LN95-D3 cells was resistant to enzalutamide but remains sensitive to ralaniten compounds that target the AR NTD, which was consistent with the cells being driven by AR-Vs.
A pilot study was conducted to study the effects of ATRA monotherapy on the growth of LN95-D3 CRPC xenografts. The growth of LN95-D3 tumors was markedly reduced in mice bearing ATRA slow-release pellets compared to placebo (Supplementary Fig. 7a). On-target activity of ATRA was validated by a decrease in Pin1 levels in xenografts from mice treated with ATRA pellets compared to placebo (Supplementary Fig. 7b). Moreover, we measured levels of serum PSA, a widely used and well-characterized clinical biomarker for CRPC progression and patient responses to therapy. PSA expression in sera was reduced on average by 27% in animals treated with ATRA compared to placebo, albeit the responses varied between animals and did not reach statistical significance ( Supplementary Fig. 7c). Notably, there was a positive correlation between serum PSA and tumor volume ( Supplementary Fig. 7d), which implies that serum PSA was a reliable surrogate marker for tumor progression of the LN95-D3 xenograft model.
Based on these findings, a larger study was undertaken to compare monotherapies to combination therapy of ATRA and EPI-7170 in castrated mice bearing LN95-D3 xenografts. The animals were implanted with a 21-day release placebo or ATRA (5 mg) pellet and then treated daily with either EPI-7170 (30 mg/ kg) or vehicle (CMC). After 21 days of treatment, tumor growth was significantly reduced in the animals treated with ATRA, EPI-7170, and combination, relative to the vehicle/placebo-control (Fig. 7b). The combination therapy was superior to individual monotherapies in reducing the final tumor volume and reached statistical significance earlier than the monotherapies. There were no significant differences in body weight of the animals during the study, and the treatments were well-tolerated (Fig. 7c).
Xenograft tumors harvested at the end of the study were variable in appearance, but several of the tumors from the combination group were remarkably reduced in size and white in appearance (Fig. 7d, e). Western blot analyses of tumor lysates confirmed the on-target activity of ATRA with decreased levels of Pin1 and its target cyclin D1 (Fig. 7f, g). Consistent with in vitro data (Fig. 6f), tumor levels of Skp2 were reduced by EPI-7170, combination therapy, and to a modest extent by ATRA (Fig. 7f,  h). The expression of full-length AR and AR-Vs varied between tumors, but overall levels and the ratio of AR-V to full-length AR were not significantly impacted by treatment. Analysis of whole blood collected throughout the study revealed a 50% decrease in serum PSA from animals treated with combination therapy compared to the control ( Supplementary Fig. 8), where 6 out of 9 animals on combination therapy showed stable or regression of PSA during the study (Supplementary Fig. 8d). Sections of harvested xenograft were assessed by immunohistochemistry for proliferation antigen Ki-67 and expression of PSA and AR. Consistent with ATRA and EPI-7170 decreasing tumor volume, ATRA, EPI-7170, and combination therapy decreased the number of Ki-67-positive tumor cells in xenografts (Fig. 8a, b). PSA expression was also decreased by ATRA, EPI-7170, and combination therapy, compared to the control (Fig. 8a, c). The levels of AR protein were similar in xenografts from animals treated with EPI-7170 or combination therapy and was decreased slightly by ATRA (Fig. 8a, d). Collectively, these in vivo data with in vitro data support that combination therapy of a ralaniten analog (EPI-7170) with a Pin1 inhibitor (ATRA) inhibits the growth of CRPC tumors driven by AR-Vs.

Discussion
Increasing evidence suggests that AR-Vs underlie resistance of CRPC to second-generation AR-LBD inhibitors 2,46-48 . This has led to an interest in finding therapies that target the AR NTD. Unfortunately, this domain is intrinsically disordered. Only one drug, ralaniten, has been tested clinically that directly binds to any intrinsically disordered target, thereby emphasizing the hurdles in developing drugs to these difficult targets. Ralaniten showed clinical signs of efficacy but had poor pharmacokinetics that led to excessive pill burden 49 . A second-generation ralaniten analog with improved stability, EPI-7386, is currently in Phase I evaluation. Ralaniten binds residues 354-448 7 predominantly within Tau-5 (residues 360-485). In response to androgen, transcriptional activity of full-length AR resides in Tau-1 (residues 101-370). The loss of AR-LBD with AR-Vs shifts the site of transcriptional activity from Tau-1 with androgen to Tau-5 in the absence of androgen 30 . Multiple hits against the AR NTD that encompass both Tau-1 and Tau-5 are proposed to yield better therapeutic responses for CRPC that express both full-length AR and AR-Vs 6 .
Pin1 motifs are predicted across the entire AR NTD and could potentially impact transactivation of AR NTD. This cis/transproline isomerase is frequently elevated in prostate cancer and associated with disease recurrence [10][11][12] . We found that Pin1 interacted with AR NTD predominantly within Tau-1 (residues 234-391) and that genetic and pharmacological inhibition of Pin1 selectively disrupted several key aspects of AR signaling. These included inhibition of: (1) transcriptional activities of fulllength AR and AR-V7; (2) transactivation of AR NTD; (3) protein-protein interactions; and (4) AR-and AR-V-dependent proliferation of prostate cancer cells. A combination of pharmacological Pin1 inhibitor with ralaniten compounds was synergistic for blocking cell cycle progression and promoted senescence of enzalutamide-resistant LN95 cells whose proliferation is driven by AR-Vs. Consistent with in vitro data, this combination therapy had superior antitumor effects in vivo compared to monotherapies.
Here we provide evidence that genetic and pharmacological inhibition of Pin1 selectively impeded AR transcriptional activity and AR-driven proliferation. Support for selectivity was drawn from clinical data supporting the association between PIN1 levels and the levels of AR and its target genes KLK3/PSA, TMPRSS2, and NKX3.1. Non-AR-driven reporters were not blocked by Pin1 inhibitors or knockdown contrary to reporters driven by fulllength AR and AR-V7. Gene expression analyses showed a sensitivity of AR-regulated genes to inhibiting Pin1 activity. Ontarget activity against Pin1 was measured in harvested CRPC xenografts. Similarly, inhibiting Pin1 induced known biological responses that occur with the loss of AR function that included attenuating progression from G 1 to S phase of the cell cycle and increased cellular senescence 50 .
Functional AR is required for the progression of cells from G 1 to S phase of the cell cycle 51,52 . Consistent with these reports, both ralaniten compounds and ATRA decreased the S phase population and increased G 1 , whereas enzalutamide had no effect on AR-V-driven proliferation. G 1 arrest was accompanied by increased levels of p27 Kip1 and the loss of cyclins D1 and A2 by ATRA which were expected in response to blocking AR function 52 . Similarly, ATRA and ralaniten both increased cellular senescence with concomitant losses of Skp2 and pRb which are known responses to a loss of AR function 50,51 . A combination of ATRA with the more potent ralaniten analog EPI-7170 was more effective than the individual monotherapies in inhibiting the in vivo growth of LN95-D3 xenograft tumors, reducing serum PSA, and was not toxic to the animals. This xenograft model expresses PSA in castrated animals and demonstrated robust growth in vivo with a tumor volume doubling time of approximately 3 days; therefore, it is a representative model for aggressive, AR-V7-positive CRPC. ATRA (tretinoin) is the only clinically used drug known to inhibit Pin1 and is FDA-approved to treat acute promyelocytic leukemia. These studies overall support that repurposing ATRA as a combination therapy with a ralaniten analog could provide a significant near-term clinical impact for the treatment of CRPC.

Methods
Cell culture. Cell lines were obtained from the following sources: LNCaP cells from Dr. Leland Chung (Cedars-Sinai Medical Centre, Los Angeles, CA); LNCaP95 (LN95) cells from Dr. Stephen Plymate (University of Washington, Seattle, WA); VCaP cells and PC-3 cells from the American Type Culture Collection (Manassas, VA); and DU145 cells from Dr. Victor Ling (British Columbia Cancer Agency, Vancouver, BC). LNCaP, VCaP, PC-3, and DU145 cells were authenticated by short tandem repeat analysis and tested to ensure they were mycoplasma-free by DDC Medical (Fisher Scientific, Ottawa, ON). Cell lines were maintained in the following culture media: LNCaP cells in RPMI-1640 media with 5% fetal bovine serum (FBS); LN95 in RPMI-1640 media with 10% charcoal-stripped serum (CSS); PC-3 and DU145 cells, in DMEM (Invitrogen, Carlsbad, CA) with 5% and 10% FBS, respectively, and supplemented with 2 mM L-glutamine and 1 mM of sodium pyruvate; VCaP cells in DMEM (Sigma-Aldrich, St. Louis, MO) with 10% FBS, 2 mM L-glutamine, and 1 mM sodium pyruvate. All cell lines were maintained at 37°C in a humidified incubator containing 5% CO 2 for not more than ten passages. Cell lines were routinely tested with the Venor GeM Mycoplasma Detection Kit (Sigma-Aldrich) to ensure that they were mycoplasma-free.
Fluorescence microscopy. AR translocation assays were performed by a modified protocol from Myung et al. 9 . Approximately 2.8 × 10 5 LNCaP cells were grown on coverslips in six-well plates and transiently transfected with 2 μg of an expression vector encoding a YFP-AR fusion protein. On the following day, cells were pretreated with the indicated compounds for 1 h and stimulated with R1881 (1 nM) or a vehicle containing ethanol for 3 h. After incubating with compounds, cells were fixed with 4% paraformaldehyde, washed twice with PBS, and then counterstained with DAPI. Slides were examined on an Axiovert 200 Fluorescence Microscope (Zeiss, Oberkochen, Germany). The ratio of the YFP signal in the nucleus relative to the cytoplasm was quantified with ImageJ software. For each image, background fluorescence was measured and subtracted from experimental values. A total of at least 50 cells from nine images containing six or more cells per field were quantified to determine the nuclear to cytoplasmic YFP ratio for each treatment group.
BrdU incorporation assay. BrdU incorporation was determined with the colorimetric BrdU ELISA kit (Roche Diagnostics), according to the manufacturer's protocol. In brief, cells were seeded in 96-well plates (LNCaP, 5000/well; LN95, 8000/well; PC-3 and DU145, 4000/wells) in media with reduced serum (0.5-1%). On the following day, cells were pre-treated with test compounds for 1 h and then incubated with 0.1 nM of R1881. After the indicated incubation times, cells were labeled with 10 μM of BrdU for 2 h and then fixed for the assay. BrdU incorporation was quantified with a VersaMax ELISA microplate reader (Molecular Devices, San Jose, CA) by measuring absorbance at 370 nm with a reference wavelength of 492 nm.
Transcriptional reporter assay for AR-V7. The V7BS 3 -luciferase plasmid, which contains three tandem repeats of an AR-V7-specific promoter element of the UBE2C gene, was a gift from Dr. Stephen Plymate (University of Washington) and has been described 36 . LNCaP cells seeded in 24-well plates were co-transfected with the V7BS 3 -luciferase reporter (0.25 μg/well), an expression vector encoding AR-V7 (0.2 μg/well), and a filler plasmid (pGL4, 0.3 μg/well). On the following day, the cells were treated with the indicated compounds for 24 h. Luciferase activity was measured for 10 s using the Luciferase Assay System (Promega) and the data were normalized to total protein concentration determined by a Bradford assay. Western blot analyses were performed to verify the expression of AR-V7.
Cell cycle analysis by flow cytometry. Cells were grown in 10 cm dishes in 10% CSS for 3 days before the media was changed to 1.5% CSS and ATRA and/or EPI-002 were added at the indicated concentrations. After treatments, the cells were labeled with BrdU (10 μM) for 2 h and then trypsinized, washed, and fixed in 70% EtOH/PBS at −20°C. Approximately 2.5-5.0 × 10 5 of fixed cells were labeled with fluorescein (FITC)-conjugated anti-BrdU antibody (B44; BD Biosciences) and the DNA content was stained with 7-AAD (Sigma). The data were collected with a BD FACSCalibur flow cytometer (BD Biosciences). FITC and 7-AAD fluorescence were detected in the FL1 and FL3 channels, respectively, using CellQuest Pro and analyzed by FlowJo V10 software (Ashland, Oregon). Pre-gating was performed on all samples by plotting FL3-W × FL3-H to exclude doublets from the analysis.
Senescence-associated β-galactosidase staining. β-galactosidase activity at suboptimal pH (6.0) was used as a marker to detect cellular senescence 42 . Cells were seeded in four-well chamber slides at a density of 8000-10,000 cells. On the following day, the media was changed to low serum media (1.5%) and cells were treated with EPI-002 or ATRA. Three days later, the media was removed and cells were fixed with 2% formaldehyde/0.2% glutaraldehyde for 10 min at room temperature, washed with PBS, and then incubated with freshly prepared β-Gal staining solution (Cell Signaling) titrated to pH 6.0. Staining was optimal after incubating with the β-Gal solution at 37°C for 16-18 h. After staining, the slides were washed twice with PBS, once with absolute methanol, and then mounted with 50% glycerol/PBS. The percentage of β-galactosidase-positive cells were counted with ImageJ using the Cell Counter plugin. A total of nine images containing 100-300 cells were scored to determine the percentages of positive cells.
Animal studies. All experiments involving animals conform to the relevant regulatory and ethical standards and were approved by the University of British Columbia Animal Care Committee (A18-0077). Metacam (1 mg/kg, 0.05 mL/10 g of body weight) was administered subcutaneously prior to any surgery. Animals were anesthetized with isoflurane and euthanized by CO 2 . Six-eight-week-old male (NOD-scid IL2Rgamma null ) mice were maintained in the Animal Care Facility at the British Columbia Cancer Research Centre. Mice were castrated 2 weeks before they were subcutaneously inoculated in the right flank with two million LN95-D3 cells in a 1:1 volume of matrigel (Corning Discovery Labware, Corning, NY). Tumor volume was measured twice a week by digital calipers and calculated by the formula: length × width × height × 0.5236. ATRA (5 mg, 21-day release) or a matching placebo pellet purchased from Innovative Research of America (Sarasota, FL) was subcutaneously implanted in the left side of the neck of mice 3 weeks after inoculating the cells or when tumor volume was about 33 mm 3 . Starting 3 days after the implants, animals were treated daily by oral gavage with 30 mg/kg body weight of EPI-7170 or vehicle (3% DMSO/1.5% Tween-80/1% CMC). Tumors were excised 1 day after the last dose and prepared for western blot analysis and immunohistochemistry.
Immunohistochemistry. Sections were cut from formalin-fixed, paraffinembedded tumor tissue at a thickness of 5 µm, deparaffinized in xylene, and then rehydrated in alcohols and distilled water. Endogenous peroxidase was blocked by incubating with a 3% hydrogen peroxide solution for 5 min. Sections were incubated overnight at 4°C with antibodies recognizing the Ki-67 antigen (MIB-1; Dako Omnis) or antibodies specific for the AR N-terminus (PG-21; Millipore) at a dilution of 1:50. PSA was detected using a monoclonal mouse anti-human PSA antibody concentrate (ER-PR8; Dako). A negative control without the primary antibody was included for each sample on the same slide. Antigen was detected with 3,3-diaminobenzidine (DAB) and hematoxylin was used for counterstaining. Images were captured on an Axio Imager M2 microscope (Zeiss) at ×200 magnification. The percentage of Ki-67-positive cells was determined with ImageJ using the Cell Counter plugin. At least 1000 cells were scored from three images of representative tumors for each treatment group. The intensity of DAB staining for PSA was quantified by ImageJ using color deconvolution and measuring the mean gray value for the DAB image, where optical density = log(max intensity/mean intensity), and a max intensity of 255 for an 8-bit image.
Quantification of serum PSA. Serum levels for total PSA were measured by a PSA ELISA kit (Anogen, Mississauga, ON) with a lower limit sensitivity of~1 ng/mL. Whole blood was sampled from the tail vein on days indicated or by cardiac puncture immediately after euthanasia. The serum fraction was separated by centrifugation at 12,000 rpm for 3 min at room temperature and then stored at −20°C until further analysis. Serum samples were diluted tenfold for the assay and were compared to calibration standards, according to the manufacturer's protocol. The concentration of total PSA was determined by a VersaMax ELISA microplate reader by measuring absorbance at 450 nm with a reference wavelength of 650 nm.
Statistics and reproducibility. Synergistic interactions between compounds were evaluated according to the Bliss independent model for a predicted additive effect,F a þ F b À F a F b À Á , where F a and F b are the fractional responses for treatment A and B at a given dose 38 . The CI was determined by the equation Ã =F ab where F ab represents the fractional response for the combination of A and B. Drug combinations were considered to be synergistic if CI < 1, additive if CI = 0, and antagonistic if CI > 1. Statistic differences were determined by GraphPad Prism 8 (San Diego, CA) by analysis of variance (ANOVA) unless stated otherwise. We used an alpha level of 0.05 for all statistical tests, where *P < 0.05, **P < 0.01, ***P < 0.001; ns not significant. Experiments were repeated with independent biological replicates at least three times.