Discovery of JNJ-63576253, a Next-Generation Androgen Receptor Antagonist Active Against Wild-Type and Clinically Relevant Ligand Binding Domain Mutations in Metastatic Castration-Resistant Prostate Cancer

This article is featured in Highlights of This Issue, p. 761

An estimated 31,620 men died from prostate cancer in 2019, representing 10% of all cancer-related deaths (1). Nevertheless, over the past 20 years, mortality has decreased nearly 50% (1), partly attributable to the development of androgen receptor (AR) pathway inhibitor therapy, abiraterone acetate plus prednisone (AAP; ref. 2), and second-generation AR antagonists, enzalutamide (MDV-3100; ref. 3) and apalutamide (ARN-509; ref. 4). Patient progression while on these therapies has been associated with numerous resistance mechanisms. These include mutations specific to the AR signaling pathway (AR aberrations): gene (5, 6) or enhancer (7) amplification, genomic structural rearrangement (8), splice variant isoform expression (9), ligand binding domain (LBD) point mutations (10, 11), adrenal and intraprostatic androgen synthesis (12–14), and glucocorticoid receptor bypass (15, 16). Treatment-emergent small-cell neuroendocrine (17) and double-negative (18) prostate cancer, in which prostate lineage specificity is lost in response to AR pathway inhibition, have arisen more recently as mechanisms through which drug resistance can occur.

Mutations that occur in the LBD in response to AR pathway inhibitors are one mechanism through which therapeutic resistance arises. While absent in nonmetastatic castration-sensitive patients (19), LBD mutations are enriched in heavily treated metastatic castration-resistant prostate cancer (mCRPC; refs. 5, 20), suggesting that they are beneficial to the propagation of clonal populations harboring them. In cellular models, the T878A mutation, endogenous to LNCaP, activates AR signaling and proliferation in response to treatment with flutamide (21). Similarly, bicalutamide, DHEA, and corticosteroids act as agonists in the presence of W742C/L (22), H875Y (23), and L702H (24) mutations, respectively.

In 2013, the AR F877 L mutation was reported to cause an antagonist-to-agonist switch in response to treatment with both enzalutamide (11) and apalutamide (10) that increased transcriptional activation of the AR and led to drug resistance. In the clinic, AR F877 L was detected in the plasma circulating tumor DNA (ctDNA) of 3 of 29 progressing patients enrolled in the ARN-509 phase I trial. Combined with expansion cohort data, 14% of patients possessed a point mutation in the LBD at progression (25), demonstrating an unmet need.

In this article, we characterize JNJ-63576253 as a potent and selective next-generation AR pathway inhibitor of AR wild-type (WT), AR F877L, and other clinically detected LBD mutations. We demonstrate that this molecule inhibits transcriptional activity in reporter assays, cellular proliferation, and AR downstream target gene expression. Furthermore, we show that in an enzalutamide-resistant LNCaP F877 L xenograft model, JNJ-63576253 causes tumor growth inhibition (TGI). JNJ-63576253 is currently being evaluated in a phase II clinical trial (NCT02987829).

HepG2 (ATCC, catalog No. HB-8065, RRID:CVCL_V331), VCaP (ATCC, catalog No. CRL-2876, RRID:CVCL_2235), and PC3 (ATCC, catalog No. CRL-7934, RRID:CVCL_0035) were obtained from the ATCC. No further authentication was performed. LNCaP AR/cs and LNCaP F877L, originally derived from LNCaP.FGC (ATCC, catalog No. CRL-1740, RRID:CVCL_1379), were obtained from Aragon Pharmaceuticals. The authenticity of the LNCaP lines was verified by short tandem repeat analysis at Analytic Biological Services. All cell cultures were verified Mycoplasma free weekly using the Lonza MycoAlert kit (LT07-418). Cellular assays utilized cultures that had been passaged 10 or fewer times. The chemical synthesis of JNJ-63576253 is discussed in article by Zhang and colleagues (31). Enzalutamide was synthesized according to published procedures and stored at a stock concentration of 10 nmol/L in DMSO. R1881 was obtained from Sigma and stored at a stock concentration of 10 mmol/L in DMSO. Stock compounds were kept in a dessicator and replaced every 2 weeks. Plasmids were obtained from Aragon Pharmaceuticals (Supplementary Fig. S5).

LNCaP F877 L and LNCaP AR/cs were transfected with a consensus sequence androgen response element (ARE) Firefly Luciferase Reporter (Qiagen) and selected under Puromycin (Thermo Fisher Scientific) to generate stable cell lines, referred to as LNCaP F877 L ARE Luc and LNCaP AR/cs ARE Luc.

HepG2 cells were cotransfected with an AR response element firefly luciferase reporter and AR-VP16 construct for 24 hours. For AR-VP16 assays, 25,000 cells from the pooled transfection were seeded into compound and 90 pmol/L R1881-spotted white and clear bottom 96-well plates and incubated for 48 hours. AR F877L-VP16 assays utilized 1 nmol/L R1881. LNCaP reporter lines were seeded at a density of 10,000 cells per well, incubated overnight, then treated for 24 hours with compound in the presence of 0.1 nmol/L R1881. After treatment, cells were assayed using the SteadyGlo-luciferase Kit (Promega) and read on an EnVision plate reader in luminescence mode. All reporter assays were conducted in media containing charcoal-stripped FBS. Raw data were analyzed in GraphPad Prism using the variable slope four-parameter nonlinear regression and normalized to vehicle (0% activity) and R1881 (100% activity) treatment.

LNCaP AR/cs, LNCaP F877L, VCaP, and PC3 were seeded at densities of 5,000 or 250 cells per well into 96-well plates and incubated overnight. Following compound treatment, the cells were incubated for 6 days before assessing proliferation using the CellTiter-Glo Kit (Promega) and read on an EnVision plate reader in luminescence mode. All proliferation assays were conducted in media containing charcoal-stripped FBS. Raw data were analyzed in GraphPad Prism using the variable slope four-parameter nonlinear regression and normalized to compound treatment day samples (0% activity) and R1881 (100% activity) treatment.

A total of 1 × 10⁶ LNCaP AR/cs or LNCaP F877 L were seeded into each well of a 6-well plate in RPMI1640 medium supplemented with 10% charcoal-stripped serum for 24 hours. The cells were treated with 0.1 nmol/L R1881 plus compound in dose response for 24 hours following treatment. The cells were harvested and processed using a Qiagen RNeasy-Plus Kit utilizing the QiaCube system. cDNA was synthesized using the High Capacity cDNA Reverse Transcriptase Kit (Thermo Fisher Scientific) and run on a Viia7 RT-PCR Instrument (Thermo Fisher Scientific) using Thermo TaqMan primers (KLK3 primer: Hs02576345_m1 and FKBP5 primer: Hs01561006_m1). Statistical analyses, where applicable, were performed using ANOVA in GraphPad Prism.

Cells were seeded at 10,000 cells per well in PDL-coated 96-well plates (Greiner 655946) in androgen-deprived medium: RPMI1640 without phenol red (Gibco, 32404-014), 5% charcoal-stripped serum (Sigma, F6765), and 2 mmol/L l-Glutamine (Sigma, G7513) for 24 hours. Dose–response testing was done at 0.0114, 0.0343, 0.103, 0.308, 0.925, 2.78, 8.35, and 25 μmol/L with constant 0.5% DMSO and in the presence and absence of 1 nmol/L methyltrienolone (R1881, ACC API0003347) added 1 hour after compound addition.

After 24-hour incubation at 37°C, cells were fixed with formaldehyde (5% final at room temperature) and permeabilized with ice-cold methanol. Hoechst 33258 (Invitrogen, H1399) was added at final 2 μg/mL and CellMask Deep Red at 0.5 μg/mL (Invitrogen, H32721). AR was detected by indirect immunofluorescence with a mouse anti-AR IgG (Abcam, 49450) and Alexa Fluor 488–labeled goat anti-mouse IgG secondary (Invitrogen, A-11001). PSA was detected using a rabbit anti-PSA IgG (Cell Signaling Technology, 5365S) and an Alexa Fluor 568 goat anti-rabbit IgG secondary (Invitrogen, A-11011). Plates were imaged with CV7000 (Yokogawa), using a 20× objective. Excitation lasers and emissions filters for the fluorescence labels mentioned above were 405 nm and 445/45, 635 nm and 676/79, 488 nm and 525/50, and 561 nm and 600/37, respectively, and corresponding exposure times 120, 120, 500, and 500 milliseconds. Data were collected from approximately 1,200 cells per well for the DMSO controls. The experiment was performed twice with three technical replicates for a total of six replicates per compound dose response.

Image analysis was performed with a custom script written in PerkinElmer Acapella. The Hoechst channel was used to segment the nuclei, and the CellMask Deep Red channel to define the cell area. “AR levels” and “PSA levels” were defined as the mean intensities in the whole-cell region in the AR and PSA channel, respectively. AR translocation was defined as the ratio of nuclear/whole-cell total intensity in the AR channel. The three measurements were each aggregated per well by taking the median over the cells in the well. Visual QC and data normalization were performed in Phaedra (26).

All experimental procedures were conducted in accordance with Janssen Institutional Animal Care and Use Committee and U.S. Department of Agriculture regulations. For the Hershberger assay, prepubertal, ages 42–45 days, castrated male rats (Sprague Dawley, RRID:RGD_737903) were housed in standard caging. For LNCaP xenograft studies, 6- to 8-week-old castrated male SCID Hairless Outbred (SHO) were housed in sterile ventilated caging. All animals were maintained under aseptic and under pathogen-free conditions. The animal holding room provided 12 hours of alternating light and dark cycles and met the standards of the Association for Assessment and Accreditation of Laboratory Animal Care specifications. Reverse osmosis water and autoclaved food were supplied ad libitum. Drugs were administered by individual body weight for each rat or by per fixed dose for each mouse. JNJ-63576253 was formulated in 20% hydroxypropyl beta cyclodextrin; enzalutamide in 1% carboxymethyl cellulose, 0.1% Tween80, and 5% DMSO; and Testosterone propionate in corn oil.

Rats were administered JNJ-63576253 orally, once daily at 10, 30, or 50 mg/kg twice daily at a volume of 5 mL/kg together with testosterone propionate at 0.4 mg/kg once daily at a volume of 0.5 mL/kg, and subcutaneously for 10 days (n = 6/group). On study day 0 (predose), days 4 or 5 (1–2 hours postdose), and day 11 (24 hours after last dose), blood was collected as part of the necropsy for three animals per cohort (n = 6/group). At necropsy (day 11), the following androgen-sensitive male accessory organs were removed and weighed: Cowper gland, seminal vesicles and coagulating glands (SVCG), glans penis, ventral prostate, and levator ani-bulbocavernosus (LABC).

Mice were injected subcutaneously with LNCaP AR F877 L–mutant or WT AR human prostate tumor spheroids. The spheroids were generated by adding 2 × 10⁵ cells to 0.5 mL of Cultrex BME (Trevigen) and incubating for 1 week at 37°C with the surface complete media (RPMI1640 plus 10% HI FBS) changed daily. Approximately 0.5 mL was injected subcutaneously into the right flank of each mouse. When tumors were established (∼250 mm³), mice were randomized into experimental groups (n = 10/group) and treated orally with JNJ-63576253 at 30 or 50 mg/kg or enzalutamide at 30 mg/kg orally, once daily for 3 to 4 weeks. The tumor take rate was 50% to 70%. Terminal blood and tumor samples were collected at 2, 4, 7, or 24 hours after last dose of vehicle or JNJ-63576253 treatment. Tumor volumes and body weights were recorded twice per week using a digital caliper and analytic scale, respectively. Tumor volumes were calculated using the formula: tumor volume (mm³) = (a × b²/2); where “a” represents the length, and “b” the width.

For the Hershberger assay, the effects of JNJ-63576253 on change in weight of five different androgen-sensitive organs (ASOs): Cowper gland, SVCG, glans penis, ventral prostate, and LABC were monitored. Statistically significant suppression of ASOs is required in two of five organs for a compound to be classified as an antiandrogen. Analysis was performed by t test/Mann–Whitney. Performance criteria was determined by calculating the maximum coefficient of variation (%CV) for each of five ASO tissues. Maximum %CV has been established for androgenic/antiandrogenic effects on dependent tissues for the castrate model based endocrine disrupter screening program validation studies (27). If %CVs are not homogenous, then nonparametric statistical procedures are utilized. For the efficacy studies, tumor volume and body weight data were graphically represented and statistically analyzed utilizing GraphPad Prism software (RRID:SCR_002798). Statistical significance was evaluated for AR antagonist small molecule–treated groups versus vehicle-treated controls on the last day of the study using an one-way or two-way ANOVA with Dunnett multiple comparisons posttest or using a t test for single comparisons. Differences between groups were considered significant when the P value was ≤0.05.

A homology model of AR LBD was built using a glucocorticoid receptor (GR) as a template and induced-fit docking calculations were performed using Schrödinger Suite 2014, using default settings (28), followed by manual docking and rotamer selection to further fine tune a “bicalutamide-like” pharmacophore of the A-ring having the CN and CF3 groups. The modeled binding mode of JNJ-63576253 suggests that this molecule can stabilize a putative “antagonistic” open helix 12 conformation (orange tubes), while an “agonistic” closed conformation of helix 12 (cyan ribbons), as depicted by superimposed agonistic conformation of AR LBD, would have severe clashes with such a binding mode (29). With this model, we have hypothesized that a potential agonistic conformational switch introduced by the F877 L mutation (or other, pharmacologically similar LBD mutations) would be incompatible with binding of JNJ-63576253, likely due to the presence of an extended and bulkier piperidine moiety in this molecule (compared with enzalutamide).

JNJ-63576253 (Fig. 1A) was identified following extensive optimization of existing Janssen chemical matter (30, 31) with the aim of restoring activity against enzalutamide- and apalutamide-resistant cellular models driven by the F877 L and other clinically relevant LBD point mutations (10, 11). In biochemical assays, JNJ-63576253 displayed superior activity to enzalutamide in WT AR competitive radioligand binding assays with the synthetic androgen R1881. Competition for binding to estrogen and glucocorticoid receptors was undetectable or had IC₅₀ value greater than 30 μmol/L (Fig. 1B), representing an approximately 1,000-fold selectivity versus other nuclear hormone receptors. Utilizing molecular docking models informed by published data for other nuclear receptors, we confirmed that mutation from phenylalanine to leucine at AR amino acid 877 reduces steric hindrance in the ligand binding pocket. Subsequent binding of antagonists, such as enzalutamide, produces an agonist conformation rather than antagonist conformation (Fig. 1C). This mechanism was similar to that which has been described computationally for enzalutamide and the flutamide- and bicalutamide-associated mutations at T878A and W742C/L (32–34).

To evaluate AR functional activity in cells, we utilized a fusion construct of AR and the VP16 virion phosphoprotein transactivation domain as a tool to sensitively discriminate ligand-receptor agonism (10, 35). VP16 drives constitutive nuclear translocation and transactivation in the absence of coactivator recruitment (36). In HepG2 transcriptional reporter models, JNJ-63576253 completely inhibited transiently transfected VP16-AR F877 L (IC₅₀ = 15 nmol/L) in the presence of 90 pmol/L R1881. In contrast, enzalutamide failed to reach 50% inhibition at any tested concentration up to 30 μmol/L (Fig. 2A). JNJ-63576253 and enzalutamide displayed similar inhibition of AR WT VP16 in the presence of 90 pmol/L R1881 (Fig. 2C). In VP16-AR F877 L cells, enzalutamide elicited activation of the AR in the absence of R1881 at concentrations as low as 1 nmol/L, reaching 40% activity at 3 μmol/L (Fig. 2B). JNJ-63576253 did not significantly activate the AR in the absence of R1881, reaching 5% activity at 10 μmol/L with no activity at 30 μmol/L. These in vitro concentrations approximate tumor concentrations measurable at efficacious doses in pharmacodynamics and xenograft studies (Figs. 5 and 6).

We next sought to assess the activity of JNJ-63576253 in LNCaP, a model of prostate adenocarcinoma. In cells stably transfected with AR F877 L and an ARE-driven firefly luciferase reporter, JNJ-63576253 completely inhibited AR-mediated transactivation in the presence of 100 pmol/L R1881 (IC₅₀ = 99 nmol/L). In contrast, enzalutamide acted as an incomplete antagonist, activating AR signaling at concentrations greater than 1 μmol/L (Fig. 2D). Reporter activity reached 154% of control stimulation with 100 pmol/L R1881, the concentration that induces maximal proliferation in this model (Supplementary Fig. S4). Both compounds were comparably effective at inhibiting AR transactivation in LNCaP AR/cs (37) stably expressing the ARE reporter at low nanomolar IC₅₀ concentrations (Fig. 2E). In addition, JNJ-63576253 was capable of effectively inhibiting transactivation of other clinically relevant AR LBD mutations in transiently transfected HepG2 (Fig. 2F). We demonstrate that the F877 L single mutation is sufficient to induce potent agonism in HepG2 transient expression models (Supplementary Fig. S3).

We investigated the effectiveness of JNJ-63576253 at inhibiting cellular proliferation in several models of human prostate adenocarcinoma. In LNCaP F877 L overexpression models, JNJ-63576253 treatment in the presence of 100 pmol/L R1881 completely inhibited proliferation at concentrations greater than 3 μmol/L with an IC₅₀ value of 197 nmol/L. As predicted by the reporter data, enzalutamide displayed incomplete antagonist activity at concentrations greater than 3 μmol/L (Fig. 3A). Both compounds were nearly equivalent at inhibiting proliferation in the LNCaP AR/cs line with IC₅₀ value of approximately 250 nmol/L (Fig. 3B). In the VCaP model of prostate adenocarcinoma, which contains an amplified AR, TMPRSS2-ERG fusion, and expresses AR-V7, enzalutamide and JNJ-63576253 displayed nanomolar inhibition of proliferation (IC₅₀ <100 nmol/L; Fig. 3C).

Next, we used RT-PCR to assess the ability of JNJ-63576253 and enzalutamide to inhibit the transcription of two canonical AR downstream target genes, KLK3 and FKBP5. Utilizing the LNCaP AR F877 L–overexpressing model, both compounds inhibited transcription in the presence of R1881. However, the inhibitory effect of enzalutamide was blunted with increasing concentrations, such that KLK3 and FKBP5 expression at 30 μmol/L was 71% and 133%, respectively, of 100 pmol/L R1881-only treatment. These results contrast with a reduction in KLK3 and complete inhibition of FKBP5 gene expression by JNJ-63576253 (Fig. 4A and B). We discovered that in the absence of ligand, enzalutamide at 30 μmol/L was capable of recapitulating 86% and 141% of 100 pmol/L R1881-induced KLK3 and FKBP5 expression, respectively. JNJ-63576253 alone produced 25% of 100 pmol/L R1881-induced KLK3 expression at 30 μmol/L and no agonism for FKBP5 (Fig. 4C and D). These data demonstrate that in an F877 L–mutated AR, enzalutamide acts as a partial antagonist in the presence of R1881 and an agonist in its absence. In contrast, JNJ-63576253 maintained antagonist activity both in the presence and absence of R1881. Both enzalutamide and JNJ-63576253 inhibited 100 pmol/L R1881-mediated AR downstream target gene expression in the LNCaP AR/cs model. There was no statistically significant difference between enzalutamide or JNJ-63576253 at identical concentrations (Supplementary Fig. S2).

AR nuclear translocation occurs following ligand binding and intramolecular dimerization (38). The first-generation antiandrogen bicalutamide was shown to reduce the ratio of nuclear-to-cytoplasmic AR in LNCaP by 50%, whereas, enzalutamide and apalutamide inhibited nuclear AR to approximately 10% of total (36). To assess inhibition of nuclear translocation, we employed an immunofluorescence imaging assay to measure AR protein localization in LNCaP following compound treatment for 24 hours. JNJ-63576253 reduced nuclear localization in the presence of R1881 to 9% of total protein at 8.25 μmol/L, while enzalutamide reduced it to 35% (Supplementary Fig. S8A and S8C). PSA protein was reduced, indicating inhibition of AR downstream gene transcription. Employing an enzalutamide-resistant LNCaP line containing a heterogeneous mixture of mutations, AR nuclear localization was reduced to 30% of total AR protein by JNJ-63576253 at 8.25 μmol/L in contrast to 64% by enzalutamide (Supplementary Fig. S8B). Both compounds reduced AR protein at μmol/L concentrations in the absence of significant cell death (Supplementary Fig. S8D).

To confirm in vivo androgen antagonism–dependent activity of JNJ-63576253, castrated male rats were treated with testosterone propionate (0.4 mg/kg, s.c.) and either vehicle or JNJ-63576253 at 10, 30, or 50 mg/kg orally in the Hershberger assay. Administration of JNJ-63576253 with testosterone propionate (0.4 mg/kg) at either 30 or 50 mg/kg was well-tolerated and resulted in statistically significant inhibition of weight gain in ASOs (Fig. 5A). Both SVCG and ventral prostates were significantly reduced (<49% or <48%, respectively) as compared with the testosterone propionate–treated group (P < 0.05). At the lowest 10 mg/kg dose level, JNJ-63576253 failed to inhibit testosterone propionate–induced ASO weight gain. Notably, at 50 mg/kg, testosterone propionate–stimulated ASO development was completely inhibited to that of the vehicle-treated castration controls. Furthermore, the degree of ASO weight suppression correlated with plasma drug concentrations (Fig. 6B) and drug exposures appeared linear across the dose range (Fig. 6C). These data support the minimally effective dose of 30 mg/kg and a maximally effective dose of 50 mg/kg for JNJ-63576253 in the Hershberger assay.

Next, we tested whether inhibition of androgen antagonism–dependent activity, as measured by the degree and duration of testosterone propionate–induced ASO weight gain, correlated with suppression of tumor growth in vivo. Male SHO castrated mice bearing established human prostate LNCaP AR/cs tumors were orally administered vehicle alone or JNJ-63576253 at 30 mg/kg once daily for 3 weeks. As shown in Fig. 6A, daily administration of JNJ-63576253 elicited 78% TGI (P < 0.05) as compared with the vehicle-treated controls. The effect of JNJ-63576253 on tumor growth was further evaluated in a second human prostate cancer xenograft model. Mice bearing established human LNCaP F877 L tumors were orally administered vehicle alone or JNJ-63576253 at 30 or 50 mg/kg once daily for 3 weeks (Fig. 6B). Administration of JNJ-63576253 resulted in a statistically significant TGI of >58% (P < 0.01) as compared with the vehicle-treated control mice. In contrast, no efficacy was observed with enzalutamide treatment at 30 mg/kg (Fig. 6C). Daily dosing with JNJ-63576253 was well-tolerated at all dose levels tested with no adverse effect on body weight (Supplementary Fig. S7). Plasma and tumor concentrations of JNJ-63576253 were determined at 2, 4, 7, and 24 hours after dosing. The peak concentration was 1,745 or 3,875 ng/mL and trough levels of drug was 250 or 310 ng/mL (for 30 or 50 mg/kg, respectively; Fig. 6D and E). Drug concentration was higher in the tissue with tumor to plasma of approximately 21- to 25-fold for both dose levels.

Despite therapeutic advances in the past decade, an estimated 31,620 men died from prostate cancer in the United States in 2019 (1), representing a significant unmet need. First-generation therapies (e.g., flutamide; ref. 39 and bicalutamide; ref. 40) were thought to ablate AR transcriptional activity by either directly preventing androgen binding to the ligand binding pocket of the AR or by reducing circulating plasma levels of androgens (e.g., leuprorelin; ref. 41). Studies that emphasized the continued importance of the AR in disease progression through gene amplification (42) and antagonist-to-agonist conversion by amplification (37) laid the scientific rationale for the development of AAP, enzalutamide, and apalutamide. In the SU2C-PCF mCRPC cohort, 85% of patients progressing on these therapies possessed AR aberrations when including structural variants and the recently discovered amplified upstream putative enhancer (5, 6, 8). These data illustrate that metastatic tumors are enriched in AR mutations and provide a strong rationale for the continued development of more effective inhibitors.

In response to this unmet need, we developed JNJ-63576253, a next-generation AR antagonist that displayed robust inhibition in WT and LBD-mutated, enzalutamide-resistant models of prostate cancer. JNJ-63576253 attenuated ASO growth in the rat Hershberger model (Fig. 5), which demonstrated oral bioavailability and tissue exposure. We showed that in an enzalutamide-resistant LNCaP F877 L xenograft model, JNJ-63576253 caused TGI (Fig. 6). In a 93-cell line panel, JNJ-63576253 and enzalutamide displayed similar antiproliferative activity profiles (Supplementary Fig. S6). In the KINOMEscan assay, used to detect activity against a panel of 468 kinases, JNJ-63576253 was not identified as a hit against any target (ref. 43; Supplementary Fig. S7). On the basis of these data, we predict that JNJ-63576253 will be tolerated in the clinic and possesses a similar profile to other marketed AR antagonists.

JNJ-63576253 is effective in models that possess AR amplification (LNCaP AR/cs) and AR-V7 expression (VCaP). AR amplification is enriched in mCRPC compared with treatment-naïve patient populations (5, 6, 19), and AR-V7 has been associated with resistance to enzalutamide and apalutamide (9). In the clinic, AR-V7 is often expressed in the context of AR-FL copy-number amplification and patients within this subpopulation often respond to apalutamide, suggesting that the full-length isoform drives transcriptional activity (44). We hypothesized that this mechanism may explain the activity of JNJ-63576253 in the VCaP model. In contrast, AR structural isoforms may arise due to AR genomic rearrangement or deletion (8) and have been linked to enzalutamide resistance in vitro (45). JNJ-63576253 was not tested in models that express these structural variants, such as the 22Rv1 or R1-D567 (46).

We attributed the ability of enzalutamide to inhibit LNCaP F877 L–overexpressing lines at lower than physiologic concentrations (less than 1 μmol/L; Figs. 2D and 3A) to the interplay of endogenous and mutant AR protein within the cell. In the HepG2 AR F877 L reporter model, in which no WT protein is present, inhibition of AR signaling by enzalutamide was mitigated at all concentrations in comparison with JNJ-63576253 (Fig. 2A). In addition, in the LNCaP F877 L xenograft model, a clear difference in TGI was observed between JNJ-63576253 and enzalutamide (Fig. 6C). These results highlight that in interpreting antagonist-to-agonist switch by the F877 L mutation, an understanding of in vivo physiologic concentrations must be considered when interpreting in vitro assay data (Fig. 6C).

In contrast to a previous report, we discovered that the ability of enzalutamide to activate AR F877 L was not dependent on the presence of the T878A mutation (47). This discrepancy may be explained by the different cellular overexpression models used to transfect HepG2 cells and resulting variations in transcriptional machinery between prostate and non-prostate cell lines. Molecular dynamics suggest that binding of enzalutamide to both the F877 L and F877L/T878A mutations induces an agonist conformation (34). Variations in reporter assay IC₅₀ values may be explained by the different AR constructs and R1881 concentrations utilized. These deviations in assays and methods were employed to draw more relevant comparisons with the previously published preclinical characterization of apalutamide (36).

AR LBD mutations have been previously correlated with overall and progression-free survival (PFS) in the post-docetaxel (48), but not in the treatment-naïve space (49). Furthermore, their detection in mCRPC (5) and absence in localized disease (19) suggest that their emergence is a consequence of androgen deprivation therapy (ADT) treatment-acquired resistance. Prior studies have linked the occurrence of LBD mutations with therapeutic intervention. The T878A mutation, in particular, has been detected in patients progressing on flutamide (50) and AAP or ketoconazole (51). AR T878A detection in ctDNA was associated with shorter overall survival (OS) compared with WT tumors in patients that relapsed while on AAP (52). In cellular models possessing the T878A or W742C/L mutations, both flutamide (21) and bicalutamide (22), respectively, are capable of rescuing AR signaling activity in the absence of exogenous androgen.

The recently discovered F877 L mutation has been detected in patients treated with the second-generation antiandrogens, enzalutamide or apalutamide (10, 25, 53, 54). Like the T878A mutation, F877 L enables agonism of the AR by second-generation antiandrogens in the absence of androgen and partial antagonism in the presence of androgen (refs. 10, 11; Supplementary Fig. S1). Joseph and colleagues were the first group to describe the F877 L mutation in three of 27 patients progressing after multiple cycles of apalutamide in the clinic (10). Data from the expansion cohort confirmed an increase in the mutational frequency of F877 L during treatment (25). Interestingly, in this study two of five patients who presented with de novo AR F877 L prior to therapy responded to apalutamide, with corresponding decreases in PSA and prolonged time to radiographic progression. These results imply that the F877 L mutation is not a mechanism of primary resistance; however, these patients possessed a low (<0.1%) mutation fraction at baseline, suggesting that a critical threshold may be required. At progression, the F877 L mutational fraction for these two patients had increased more than 10-fold. Continuing this trend, results from the SPARTAN phase III registration trial of patients with localized disease reported that AR LBD mutations were present in 1.9% of ctDNAs tested at baseline, but increased to 8.8% by end-of-treatment (44). However, PFS2 and OS were not associated with apalutamide treatment in patients that acquired these mutations. Indeed, the mutation rate was similar in both apalutamide plus ADT and ADT-alone arms (7/69 and 5/69, respectively), suggesting that ADT, not apalutamide, drives the selection of LBD point mutations. In contrast, results from the TITAN phase III trial evaluating the mCRPC population reported that acquisition of LBD mutations was increased by end-of-treatment in 13.3% (2/15) of patients on apalutamide plus ADT in contrast to 6.82% on ADT alone (3/44; ref. 55). In this cohort, the detection of AR aberrations, in general, was associated with both decreased PFS2 and OS. We speculate that in the localized setting, apalutamide effectively abrogates AR signaling and suppresses bulk tumor growth. However, clonal populations possessing LBD mutations and/or other mutations eventually expand until they sufficiently dominate the tumor microenvironment to induce clinical progression. The fitness of metastatic tumors, possessing a greater mutational burden, is enhanced through further AR aberrations, including LBD mutations, producing a significant effect on PFS2 and OS outcomes.

The SU2C-PCF consortium reported that 21 of 150 (14%) patients in their mCRPC cohort possessed LBD point mutations, however, the F877 L mutation was not detected (5). Because 50% of the patients in this study had not received AAP or enzalutamide, this may have limited the F877 L mutational frequency. Seven patients possessed the T878A mutation, which has been associated with flutamide and AAP intervention, and seven with the L702H mutation, which is activated by corticosteroids, such as the prednisone component of the abiraterone treatment regimen. Prior flutamide and bicalutamide use was not disclosed in the study. Other groups have similarly reported LBD mutations in patients treated with enzalutamide and/or AAP and their association with worse clinical outcomes (52, 53).

CRPC tumors remain dependent on the AR and both AR amplification and LBD mutations have been associated with worse clinical outcomes (48). Efforts to delineate the role that each individually plays in tumor progression have been difficult. Several studies have suggested that AR amplification and LBD mutations do not occur in the same tumor (5, 54), while AR structural variants appear to be exclusive of LBD mutations, but can be found in tumors harboring amplified AR (8). Other studies have described interpatient heterogeneity, with point mutations and gene amplification arising prior to and after AR gene amplification (56, 57). These cohorts represent heterogeneous patient populations and larger studies will be required to ascertain in which context these are relevant. JNJ-63576253 is anticipated to be effective in tumor populations possessing amplified AR and LBD mutations. The role of AR variants in clinical outcomes remains unresolved (58–60). As discussed previously, we predict that JNJ-63576253 could be effective in this setting depending on the context (45).

As the prostate cancer patient population continues to evolve in the next decade, existing data suggest that the AR LBD-acquired mutation frequency is likely to increase with treatment (44, 55). Although earlier intervention delays progression, ultimately AR aberrations will emerge that correlate with a worse prognosis (61, 62). Addressing this growing unmet need, the data presented here support JNJ-63576253 as a clinical candidate with potential effectiveness against multiple AR pathway mechanisms of resistance in the subset of patients who do not respond to or are progressing on second-generation AR-targeted therapeutics. This agent is currently under investigation in a phase I/IIA study with patients diagnosed with mCRPC (NCT02987829, https://clinicaltrials.gov).

J.R. Branch reports a patent for WO 2017123542 A1 issued. V. Pande reports a patent for WO2017123542A1 pending. P.J. Connolly reports personal fees from Janssen Research & Development LLC and is a Johnson & Johnson stockholder outside the submitted work, as well as has a patent for WO2017123542A pending. Z. Zhang reports a patent for WO2017123542A pending. I. Hickson reports a patent for WO2017123542A issued, and former employment with Janssen, all work conducted as an employee of Janssen, stockholder of Janssen, and discussion and advice for TRACON. J.R. Bischoff reports other from Johnson and Johnson outside the submitted work. L. Meerpoel reports a patent for WO2017123542A issued. M.M. Gottardis reports employment with Janssen Pharmaceuticals R and D part of Johnson and Johnson where this work originated and is a shareholder of Johnson and Johnson. G. Bignan reports a patent for WO2017123542A pending. No disclosures were reported by the other authors.

J.R. Branch: Conceptualization, resources, formal analysis, investigation, writing–original draft, writing–review and editing. T.L. Bush: Investigation, writing–review and editing. V. Pande: Investigation, writing–review and editing. P.J. Connolly: Conceptualization, supervision, writing–review and editing. Z. Zhang: Conceptualization, supervision, writing–review and editing. I. Hickson: Conceptualization, supervision, writing–review and editing. J. Ondrus: Investigation, writing–review and editing. S. Jaensch: Investigation, writing–review and editing. J.R. Bischoff: Conceptualization, supervision, writing–review and editing. G. Habineza: Investigation, writing–review and editing. G. Van Hecke: Investigation, writing–review and editing. L. Meerpoel: Conceptualization, supervision, writing–review and editing. K. Packman: Supervision, writing–review and editing. C.J. Parrett: Investigation, writing–review and editing. Y.T. Chong: Investigation, writing–review and editing. M.M. Gottardis: Conceptualization, supervision, writing–review and editing. G. Bignan: Conceptualization, supervision, investigation, writing–review and editing.

The authors would like to thank Brent Rupnow, Shibu Thomas, and Margaret Yu for scientific discussion related to the article.

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.