A bypass mechanism of abiraterone‐resistant prostate cancer: Accumulating CYP17A1 substrates activate androgen receptor signaling

Abstract Background Intratumoral steroidogenesis and its potential relevance in castration‐resistant prostate cancer (CRPC) and in cytochrome P450, family 17, subfamily A, polypeptide 1 (CYP17A1)‐inhibitor treated hormone‐naïve and patients with CRPC are not well established. In this study, we tested if substrates for de novo steroidogenesis accumulating during CYP17A1 inhibition may drive cell growth in relevant preclinical models. Methods PCa cell lines and their respective CRPC sublines were used to model CRPC in vitro. Precursor steroids pregnenolone (Preg) and progesterone (Prog) served as substrate for de novo steroid synthesis. TAK700 (orteronel), abiraterone, and small interfering RNA (siRNA) against CYP17A1 were used to block CYP17A1 enzyme activity. The antiandrogen RD162 was used to assess androgen receptor (AR) involvement. Cell growth was measured by 3‐(4,5‐dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazolium bromide assay. AR‐target gene expression was quantified by reverse transcription polymerase chain reaction (RT‐PCR). Nuclear import studies using cells with green fluorescent protein (GFP)‐tagged AR were performed to assess the potential of precursor steroids to directly activate AR. Results Preg and Prog stimulated cell proliferation and AR target gene expression in VCaP, DuCaP, LNCaP, and their respective CRPC sublines. The antiandrogen RD162, but not CYP17A1 inhibition with TAK700, abiraterone or siRNA, was able to block Preg‐ and Prog‐induced proliferation. In contrast to TAK700, abiraterone also affected dihydrotestosterone‐induced cell growth, indicating direct AR binding. Furthermore, Prog‐induced AR translocation was not affected by treatment with TAK700 or abiraterone, while it was effectively blocked by the AR antagonist enzalutamide, further demonstrating the direct AR activation by Prog. Conclusion Activation of the AR by clinically relevant levels of Preg and Prog accumulating in abiraterone‐treated patients may act as a driver for CRPC. These data provide a scientific rationale for combining CYP17A1 inhibitors with antiandrogens, particularly in patients with overexpressed or mutated‐AR.

VCaP, DuCaP, LNCaP, and their respective CRPC sublines. The antiandrogen RD162, but not CYP17A1 inhibition with TAK700, abiraterone or siRNA, was able to block Preg-and Prog-induced proliferation. In contrast to TAK700, abiraterone also affected dihydrotestosterone-induced cell growth, indicating direct AR binding.
Furthermore, Prog-induced AR translocation was not affected by treatment with TAK700 or abiraterone, while it was effectively blocked by the AR antagonist enzalutamide, further demonstrating the direct AR activation by Prog.
Conclusion: Activation of the AR by clinically relevant levels of Preg and Prog accumulating in abiraterone-treated patients may act as a driver for CRPC. These data provide a scientific rationale for combining CYP17A1 inhibitors with antiandrogens, particularly in patients with overexpressed or mutated-AR. (ADT) vs ADT alone in metastatic hormone-sensitive prostate cancer. 5,6 In contrast, the selective 17,20-lyase inhibitor TAK700 (Orteronel]) 7 failed to demonstrate a survival benefit in either the pre-or postdocetaxel setting. 8,9 With novel hormonal agents now becoming the mainstay of advanced prostate cancer (PC) therapy in both hormone-naïve and castration-resistant setting, it is of importance to identify resistance mechanisms to these agents.
Clinical data have shown that a subgroup of patients with CRPC progressing on abiraterone still responds to enzalutamide, 10 suggesting that the AR signaling axis is still active in these patients despite low circulating androgen levels. 11  that not only AR gene modifications but also wild-type AR copy number gain were associated with poor response to abiraterone and impaired overall survival. In addition, expression of the ligandindependent AR variant V7 has been associated with a poor response to both enzalutamide and abiraterone. 14 Preclinical studies have postulated CYP17A1-dependent intratumoral de novo steroid synthesis as a driver of CRPC and CYP17A1-inhibitor resistant disease. These studies reported de novo dihydrotestosterone (DHT) synthesis in LNCaP and VCaP cell lines 15,16,19 and reduced AR target gene expression and DHT and T levels in CRPC xenograft tissue after abiraterone treatment. 18 In contrast, we and others have found little evidence for de novo androgen synthesis in clinical CRPC samples. Moreover, we have previously shown that androgen precursors induced cell growth and AR target gene expression in vitro, but with undetectable CYP17A1-dependent conversion into testosterone, indicating either direct AR binding or conversion rather than de novo synthesis as a driver of cell growth. 23 In the present study, we assessed if CRPC cell growth could be driven by androgen precursors (pregnenolone [Preg] and progesterone [Prog]) at clinically relevant levels found in aging men 24 as well as in patients treated with abiraterone. 25,26 We used CRPC models expressing wild-type AR as well as mutated-AR. Cell growth was studied in the presence of CYP17A1 enzyme-and AR-inhibitors. To further establish the effects of precursor steroids on AR signaling, AR translocation was evaluated using a fluorescently labeled wild-type AR.

| Cell proliferation assays
For cell proliferation assays, 5,000 cells per well were plated in 96well dishes in their respective medium with DCC. After overnight attachment, the synthetic androgen R1881, steroids, and compounds were added to reach the indicated concentrations in a final volume of 200 μL. After 9 days, cell proliferation was assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT)assay as described previously. 30

| Steroids and compounds
For cell culture assays, RD162, a nonsteroidal antiandrogen was used (Merck, Oss, Netherlands). It is closely related to and was selected from the same drug-screen as MDV3100 (enzalutamide). In vitro and in vivo, it has equal potency in AR-antagonism as enzalutamide and no significant difference in bioavailability in preclinical testing. 31 For AR binding assays, enzalutamide (Axon Medchem, Groningen, The Netherlands) was used because of its current use in clinical practice.  Table S1).

| CYP17A1 knockdown
After overnight attachment, cells were transfected with CYP17A1 or nontargeting small interfering RNA (siRNA; On-TARGETplus SMARTpool siRNA; Dharmacon, Lafayette, LA) using Lipofectamine RNAiMax (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. Twenty-four hours after transfection, the medium was replaced by DCC medium with indicated steroids. RNA was isolated after 48 hours or proliferation determined at day 6.

| Gene expression analysis
For quantitative polymerase chain reaction (qPCR) studies, RNA was isolated using RNA-Bee (TEL-TEST Inc, Friendswood, TX) from cultures treated for 48 hours with indicated compounds/steroids, 24 hours after seeding in DCC medium at 100.000 cells per well.
Reverse transcriptase and qPCR runs were performed as described previously 21 Table S2. Gene expression was calculated as fold expression over housekeeping genes GAPDH or PBGD and vehicle treated cells.

| Nuclear AR import studies
Nuclear translocation of the AR has been studied in time as well as in end-point measurements using fluorescence confocal microscopy on MOLL ET AL.

| 939
PC346C cells stably expressing enhanced green fluorescent protein (EGFP)-AR. 29 To measure the effect of a concentration range of Preg and Prog, cells were seeded in a glass bottom 96-well plate in culture medium supplemented with the charcoal-stripped serum to avoid premature AR activation. Sixteen hours before imaging enzalutamide, TAK700, abiraterone (1 μM), and DMSO carrier only as control were added. Subsequently, 4 hours before imaging potential AR translocation was initiated using 0, 1, 10, and 100 nM Preg or Prog, and with 0.1 and 1 nM R1881 as the positive control, and nuclei were stained with Hoechst for reference. Cells were imaged using the Opera Phenix HCS system equipped with an x40 water immersion objective.
Hoechst and EGFP were exited using 405 and 488 nm lasers and were visualized using 435 to 480 nm and 500 to 550 nm band-pass filters. EGFP intensities were measured in the nuclear (nuc) and the peri-nuclear (cyto) regions. Nuclear translocation of the AR was expressed by nuclear signal intensity/(nuclear signal intensity+cytoplasmatic signal intensity), after background subtraction.
The ratio of AR nuclear localization was expressed as: For the analysis of AR-translocation dynamics, cells were seeded on glass coverslips in six-well plates. After overnight attachment, cells were treated with TAK700 (3 µM) or vehicle for 12 hours and subsequently transferred to a live-cell chamber and maintained at 37°C and 5% CO 2 .   (Table S3). Preg and Prog at levels of 1 nM and upstimulated cell growth in all CRPC clones tested ( Figure 1A). RD162 effectively blocked 10 nM Prog-and Preg-induced cell growth ( Figure 1B), indicating that the proliferative effects of Preg and Prog were AR-driven. Despite complete CYP17A1 inhibition as demonstrated in H295R cells, TAK700 could not inhibit Preg-and Prog-induced cell growth in these CRPC clones in concentrations up to 10 μM, which is approximately 150 times the IC50 in our in vitro conditions ( Figure 1C). Preg-and Prog-activated AR was confirmed by upregulated expression of the AR-target gene PSA in VCaP BIC-B even in the presence of TAK700 ( Figure S2). Similar results were obtained for LNCaP ( Figure S3), and PC346C CRPC cells that are characterized by an overexpressed wild-type AR (PC346C FLU1) or by T877A AR mutation (PC346C FLU2; Figure S4). Of note, CYP17A1 messenger RNA (mRNA) could not be detected in LNCaP nor in PC346C (Table   S3). Also, we were unable to detect CYP17A1 protein in VCaP despite detectable mRNA levels. (Figure S5). The AR-driven effects in the presence of TAK700 indicate that growth of these CRPC clones, despite upregulated CYP17A1 mRNA, was not dependent on the increased activity of de novo steroidogenesis.

Prog and Preg stimulate cell growth of castration-naïve VCaP and DuCaP cells via AR activation to test if Preg and Prog could also
facilitate cell growth of castration-naïve parental VCaP and DuCaP, characterized by relatively low levels of CYP17A1 (Table S3) Figure 3C). Likewise, in DuCaP, Preg-and Prog-induced cell proliferation was unaffected by CYP17A1-directed siRNA ( Figure 3D). Thick arrows indicate the preferred steps in human androgen biosynthesis and subsequent AR activation as reported in the literature. Abiraterone and TAK700 effectively inhibit steroid synthesis, but cannot prevent direct binding of the steroid precursors Preg or Prog to the AR (dotted line). Direct AR-antagonism by enzalutamide will still block activation by either DHT or Prog. AR, androgen receptor; CRPC, castration-resistant prostate cancer; CYP17A1, cytochrome P450, family 17, subfamily A, polypeptide 1; DHT, dihydrotestosterone; Preg, pregnenolone; Prog, progesterone underscore reports that intratumoral de novo steroidogenesis is not essential for CRPC growth. 36 We demonstrate that inhibition of AR-mediated cell growth by abiraterone can be (partly) explained by direct AR-antagonism, albeit at peak concentrations. This is consistent with data that abiraterone can bind to and antagonize AR in LNCaP and VCaP, 37,38 and our prior report that abiraterone can also partially block AR nuclear translocation. 29 Furthermore, the previously observed 3βHSD inhibition, 39 combined with the antiandrogenic potency and inhibition of CYP17A1, 3βHSD, and SRD5A by the abiraterone metabolite, D4-abiraterone, 40 may explain the beneficial effect of abiraterone in clinical trials relative to TAK700, which seems to lack these additional effects.
To date, two retrospective studies reported PSA response rates of 27% and 30% for enzalutamide in abiraterone progressive patients. 41,42 Interestingly, in a third retrospective study, PSA responses for enzalutamide after abiraterone vs abiraterone after enzalutamide have been reported to be higher (30% vs 6%), with a trend towards longer PFS in the first group. 43 In contrast, adding abiraterone to continued enzalutamide treatment did not result in a significant delay of PSA progression vs abiraterone alone in patients with biochemical progression on enzalutamide. 44 However, none of these studies have interrogated the upfront combination of ARantagonism with androgen synthesis inhibition.
Although our study may be limited by the use of in vitro models, it is important to note that these different cell lines were selected based on highly clinically relevant characteristics, including the absence of intratumoral CYP17A1 expression in the context of high AR expression. The studies were performed using clinically relevant levels of steroids and drugs as reported from relevant patient cohorts, to recapitulate the negative consequences of enhanced substrate levels of preg and prog in patients with CRPC treated with CYP17A1 inhibitors. The in vitro system allows for defined assessment of the potential of different steroids that will otherwise be obscured in in vivo models. These data provide basic mechanistic evidence to combine steroid synthesis inhibition with antiandrogens to fully extinguish ligand-dependent AR activation in tumors that have become hypersensitive to minute levels of androgen or alternative steroidal ligands (like accumulating progestagens due to systemic CYP17A1 inhibition) via AR amplification or mutations in the absence of intratumoral CYP17A1 activity. Thus, prospective data on PSA response and possible survival benefit of combining abiraterone with enzalutamide from the start of second-line hormonal therapy are eagerly awaited. 45

| CONCLUSIONS
In summary, our study demonstrates that in castration-naïve and CRPC cell lines, androgen precursor steroids Preg and Prog are able to directly activate wild-type and mutated-AR, independent of CYP17A1-mediated conversion into testosterone ( Figure 5). These findings may indicate a mechanism of resistance for patients progressing on CYP17A1 therapy where enzyme inhibition causes accumulation of these androgen precursors, and provide an explanation of why CYP17A1-inhibitorresistant tumors may still respond to treatment with antiandrogens.
From a clinical perspective, these data support the rationale for the combination of CYP17-inhibitors with potent antiandrogens, to effectively suppress AR activation mediated by accumulating steroidal ligands in both AR-amplified and AR-mutated tumors.