Bypassing Drug Resistance Mechanisms of Prostate Cancer with Small Molecules that Target Androgen Receptor–Chromatin Interactions

In advanced prostate cancer, the main therapeutic target is the androgen receptor (AR), a transcription factor that regulates expression of genes required for tumor growth (1, 2). The AR shares a domain organization with other members of the nuclear receptor (NR) family, consisting of an N-terminal domain (NTD), a central DNA-binding domain (DBD), and C-terminal ligand-binding domain (LBD; ref. 3). Binding of endogenous androgens, such as testosterone or DHT, to the LBD promotes AR nuclear localization, dimerization via the DBD, and engagement with androgen response elements (ARE) located in enhancers of target genes (1, 2). Antiandrogens in current clinical use, including enzalutamide, block AR activation by functioning as competitive antagonists of the ligand-binding pocket (LBP) on the LBD (1, 4, 5). Drug resistance can occur when mutations in the LBP render these molecules ineffective, when cells begin synthesizing constitutively active variant forms of the AR that lack the LBD, such as AR-V7 (6–9), and/or when cells display AR gene amplification and overexpression of AR protein (10). This drug-resistant state, termed castration-resistant prostate cancer (CRPC) is responsible for nearly all prostate cancer–specific deaths.

Although various cell signaling pathways and AR-associated cofactors also influence CRPC progression, their value as targets for prostate cancer treatment is still unclear. On the other hand, it is clear that the AR remains a master regulator of genes required for tumor growth in CRPC. Accordingly, the AR represents the most direct target to explore inhibition with small molecules that attack molecular functions beyond steroid binding at the LBP. For example, the AR-NTD–targeted inhibitor EPI-506 was derived by high-throughput screening (11) and is now being tested in clinical trials. Despite the clinical potential of EPI-506, its drug design is challenged by intrinsic disorder in the AR-NTD (2), which cannot be analyzed by X-ray crystallography. Hence, rational drug design is only possible for the AR-DBD and LBD domains where crystallographic information is available (1, 12, 13). In this regard, drug development efforts at the AR-LBD have been focused on interfering with hormone binding (4, 14) or blocking cofactor recruitment sites such as the binding function 3 (BF3) surface-exposed pocket (15–17).

The AR-DBD is composed of two zinc fingers, the first of which contains the P-box recognition helix responsible for ARE specificity and the second, which contains the D-box site responsible for dimerization (13). In previous studies, we analyzed the crystal structure of the AR-DBD:DNA complex to identify a surface-exposed pocket near the protein–DNA interface that could be targeted with small molecules (18, 19). The lead compound, VPC-14449, potently inhibited full-length AR in LNCaP cells, downregulated PSA expression and suppressed growth of xenografts in mouse models (19). However, the mechanism of action of VPC-14449 has not been fully defined. To this end, the purpose of this study was to investigate VPC-14449 treatment in diverse CRPC cell lines exhibiting clinically relevant drug resistance mechanisms, including expression of mutated AR or constitutively active AR variants lacking the LBD. VPC-14449 suppressed cell viability, AR transcriptional activity, and target gene expression in every model system tested. Mechanistically, VPC-14449 reduced the ability of full-length AR as well as AR variants to interact with chromatin, although the effects on full-length AR were more pronounced. Collectively, these findings highlight the potential for AR DBD–specific inhibition to overcome drug resistance in CRPC.

VPC-14449 (4-(4-(2,4-bromo-1H-imidazol-1-yl)thiazol-2-yl)morpholine) was synthesized by Life Chemicals. A correction to the published structure of VPC-14449 was made in 2017 (20, 21). Pyrvinium pamoate was purchased from Enamine. Enzalutamide was obtained from Haoyuan Chemexpress. R881 synthetic androgen was obtained from Sigma.

Full-length human AR (hAR_WT) was encoded on a pcDNA3.1 expression plasmid. Mutations in the AR-LBD or -DBD in hAR_WT were generated using a QuikChange Kit (Agilent). PIPE cloning (19) was used to replace the AR-DBD with the rat glucocorticoid receptor (GR) DBD, yielding ARw/GRDBD as outlined in Supplementary Table S1. A plasmid encoding the GST-tagged rat C552A AR-DBD mutant was described previously (13).

All prostate cancer cell lines were obtained in 2013, authenticated by IDEXX Laboratories in August 2014, and tested weekly for mycoplasma contamination. LNCaP (ATCC, CRL-1740), C4-2 (provided by Dr. Christopher J. Ong, Assistant Professor, University of British Columbia, Department of Surgery & Department of Urologic Sciences), and MR49F/MR49C (provided by Dr. Amina Zoubeidi, Assistant Professor, University of British Columbia, Department of Urologic Sciences; ref. 22) were androgen starved for 2 days in RPMI1640 media supplemented with 5% charcoal stripped serum (RPMI + CSS) prior to transfection. 22Rv1 cells (ATCC, CRL-2505) were maintained in RPMI + 10% CSS. Cells were seeded at 5,000 cells/well (96-well, 24 hours) and transfected with 50 ng/well ARR₃tk-Luc plasmid, 2 ng/well SV40 Renilla-luc and 0.15 μL/well TransIT20/20 reagent (TT20, Mirus) for 48 hours. Cells were treated with compounds (0.1% DMSO vehicle final) and 0.1 nmol/L R1881 androgen (22Rv1 = no R1881) for 24 hours, followed by cell lysis and luminescence measurement as described previously (19). PC-3 cells (ATCC, CRL-1435) were transfected with hAR_WT, mutant, or ARw/GRDBD chimera plasmids (50 ng), ARR₃tk-Luc (50 ng), and TT20 (0.3 μL/well) as described above. Lysates (40 μL) from luciferase assays were separated by 10% SDS-PAGE and probed with polyclonal AR-N20 AR-NTD specific (Santa Cruz Biotechnology) antibody, polyclonal anti-actin antibody (Sigma), and polyclonal anti-PARP/anti-cleaved PARP antibody (Sigma). Cell viability tests were performed by culturing cells as described above (96-well plate, 5,000 cell/well), treating with compounds/R1881 for 72 hours and incubating with 30 μL/well Promega CellTiter 96 AQueous Cell Proliferation MTS Assay (1 hour), followed by ABS492 nm measurement. Secreted PSA measurements from LNCaP cells were performed as described previously (19).

Rat AR-DBD (550-637) C552A was expressed and purified as described previously (13) with the addition of a gel filtration step carried out on a Superdex S75 column in a modified PBS buffer (pH 7.0) containing 200 mmol/L NaCl and 1 mmol/L DTT (PBS200D). Eluted protein was concentrated using a spin filter (MWCO 3,000) and stored at −80°C. The C552A mutant is less prone to aggregation at high concentrations and is fully functional in binding DNA.

Titrations were performed with a VP-ITC calorimeter (MicroCal, Inc.) at 25°C and analyzed with Origin 7 software. To maximize the weak signal resulting from VPC-14449 protein binding, several buffers were tested, with the optimal isothermal titration calorimetry (ITC) buffer determined to be PBS (pH 7.0), 200 mmol/L NaCl. Prior to titration, reducing agents were removed from the AR-DBD using a Superdex S-75 column equilibrated in ITC buffer. Titrations against VPC-14449 were carried out in ITC buffer supplemented with DMSO (2% v/v final) in both protein and ligand solutions. AR-DBD (1.0 mmol/L) was loaded into the titration syringe and injected into VPC-14449 (0.1 mmol/L) that was loaded into the cell. Aliquots (10 μL; 5-minute intervals) were injected, discarding heat from the first injection during analysis. The measured heat of dilution with protein (1 mmol/L) titrated into ITC buffer was subtracted from the titration of AR-DBD into VPC-I4449 to yield the ligand-binding isotherm. Titration of the AR-DBD against a synthetic AR response element was performed (40 mmol/L HEPES pH 7.4, 100 mmol/L NaCl) using a DNA duplex containing the two AGAACA half-sites arranged as direct repeats (top strand: 5′-TTC AGAACA TCA AGAACC A, hexameric half-sites underlined).

LNCaP and C4-2 cells were androgen starved (48 hours, RMPI + CSS) before 24-hour treatment with DMSO (0.1%), DMSO + 1 nmol/L R1881, 5 μmol/L enzalutamide, or 5 μmol/L VPC-14449. MR49F cells were cultured in RPMI1640 media containing 5% FBS and treated with DMSO (0.1%), 5 μmol/L enzalutamide, or 5 μmol/L VPC-14449. 22Rv1 cells were maintained in RPMI + 10% CSS and treated with 50 μmol/L enzalutamide or VPC-14449 (48 hours) or DMSO control. R1-AD1 and R1-D567 isogenic cell lines expressing full-length AR or ARv567es (23), respectively, were starved as described above, followed by stimulation with 1 nmol/L R1881 (R1-AD1), or without androgen (R1-D567), and compound treatment (24 hours R1-AD1; 48 hours R1-D567). RNA was extracted from each treatment group using TRIzol (Thermo Fisher Scientific) and used for cDNA synthesis (Superscript II, Invitrogen). RT-PCR (125 ng cDNA, 5 μmol/L primers, Roche FastStart SYBR Green) was performed on an ABI ViiA7 thermocycler (GAPDH control). RT-PCR primers are shown in Supplementary Table S1.

Fraction of approximately 1–2 × 10⁷ of cells from 15-cm dishes were performed as described previously (24). Each cell line was cultured the same way as for RT-PCR experiments (androgen starvation) except that VPC-14449 was administered for 24 hours at various concentrations followed by 4-hour treatment with DHT (1 nmol/L) prior to biochemical fractionation of cytosol, nuclear, and chromatin fractions. Western blots for AR detection were performed with anti-AR-N20 (Santa Cruz Biotechnology, ARN20) or anti-ARV7 (Precision, AG10008) specific antibodies. Equal loading of samples and confirmation of successful subcellular fractionation was performed with an mAb against cytosol α-tubulin (Cell Signaling Technology, DM1A), an mAb against nuclear Lamin A/C (Cell Signaling Technology, 4C11), and an mAb directed against histone-H3 protein (Abcam, ab32356).

Chromatin immunoprecipitation (ChIP) was performed for three independent biological replicate experiments as described previously (25) with the exception that: R1-AD1 and R1-D567 cells were seeded at 5 × 10⁶ cells/10-cm plate in RPMI1640, settled for 48 hours, and then treated with vehicle (0.1% DMSO) or VPC-14449 (50 μmol/L) in RPMI1640 for 24 hours. Cells were refed with RPMI1640 + 10% CSS supplemented with 50 μmol/L VPC-14449 (or DMSO control) and 1 nmol/L DHT (or ethanol control) for an additional 4 hours prior to cross-linking. Nuclear pellets were sonicated on ice for 10 cycles at 45% amplitude (10-second on/off pulse for 1 minute, 2-minute rest) using a 450 Sonifier (Branson). Lysates were immunoprecipitated with AR-N20 (Santa Cruz Biotechnology) and Protein A/G Plus agarose beads preblocked with Salmon sperm DNA (Millipore). DNA was purified by PCR purification (Qiagen). Samples for chromatin immunoprecipitation sequencing (ChIP-seq) were processed exactly as samples for ChIP-PCR with the exception that blocking of protein A/G beads was performed with tRNA (Sigma). Independent biological replicate experiments were performed for ChIP-seq in R1-AD1 (n = 2) and R1-D567 (n = 3) cells. Two nanograms of DNA (ChIP-enriched and input) was used for library creation with a Truplex ChIP Sample Preparation Kit (Rubicon Genomics). ChIP-seq libraries were sequenced at the U of M Genomics Center using an Illumina HiSeq2000 with 1 × 50 bp settings. Datasets were deposited into the GEO database (accession number = GSE96084).

ChIP-seq reads were mapped to human genome version hg19 using BWA v0.7.10 (26) with default parameters. Reads with mapping quality <1 were discarded. ChIP-seq peak calling was performed using MACS v2 (27) with merged BAM files from two biological replicates of DHT-treated R1-AD1 cells immunoprecipitated with AR-N20 and input DNA from R1-AD1 cells as negative control. Default parameters were used with FDR cutoff of 0.01, yielding 3,556 AR-binding sites. ChIP-seq heatmaps representing peak-centered AR or AR-V binding at these 3,556 AR-binding sites was generated using a python-based script on raw reads from merged duplicate files from R1-AD1 cells and merged triplicate files from R1-D567 cells and visualized ±3 kbp. Histograms of tag density around 3,556 called AR-binding sites called in DHT-treated R1-AD1 cells were centered by peak summit and logarithm normalized average coverage in units of number of reads per million sequenced reads was counted within 3 kbp relative to the peak center. Visualization of mapped ChIP-seq reads in BAM format was performed at the gene track level using Integrated Genomics Viewer (IGV 2.3, Broad Institute; ref. 28) with data merged from biological replicates. Data were normalized as reads per million mapped reads.

We previously characterized novel AR-DBD–specific compounds (VPC-14228/VPC-14449) predicted to bind near the P-box recognition helix (18). We also demonstrated that the corresponding inhibitory activity by both compounds was selective for the AR over other NRs, and that specific point mutations in the AR-DBD rendered the molecules less effective (18, 19). Although these findings provided evidence for an on-target interaction, the compound effects were mostly explored with transiently expressed AR, and downstream effects were only tested in the androgen-dependent LNCaP cell line (19).

To assess the activity of the lead inhibitor (VPC-14449) in models of clinical CRPC, AR transactivation and cell viability experiments were performed in LNCaP, C4-2, MR49F, and 22Rv1 cell lines, which were chosen to explore the ability of VPC-14449 to bypass drug resistance conferred by LBP mutations (hydroxyflutamide-resistant T877A in LNCaP/C4-2; enzalutamide-resistant F876L in MR49F; ref. 22) or via expression of AR variants lacking the LBD (AR-V7 in 22Rv1). Luciferase reporter assays showed that endogenous AR transactivation was inhibited by VPC-14449 in LNCaP, C4-2, and MR49F cells stimulated with the synthetic androgen R1881 (sub μmol/L IC₅₀, Fig. 1A), with less potency against AR-V activity in 22Rv1 (no R1881, low μmol/L IC₅₀, Fig. 1A). In addition, high levels of enzalutamide increased reporter activity in MR49F cells, but had no effect in 22Rv1 cells (Fig. 1A, gray). Western blot analysis confirmed that AR protein levels were not affected by the compound (Fig. 1A, bottom) and toxic effects of VPC-14449 over 24 hours treatment were ruled out by the lack of PARP cleavage in all cell types, in contrast to a strong signal for PARP cleavage in cells treated with an alternative AR-DBD inhibitor, pyrvinium pamoate (Fig. 1A, bottom; ref. 29). Cell viability assays demonstrated that VPC-14449 suppressed growth in every tested cell line, while enzalutamide resistance was apparent in both MR49F and 22Rv1 cells (Fig. 1B). Similarly, VPC-14449 inhibited AR transactivation and cell viability in a second enzalutamide-resistant cell line (MR49C, F876L mutation; ref. 30; Supplementary Fig. S1). Importantly, the viability of AR-negative PC-3 cells was not affected by any of the compounds, indicating the specific action of enzalutamide or VPC-14449 on the AR signaling pathway (Supplementary Fig. S1). These data demonstrate that VPC-14449 can inhibit AR signaling across five CRPC cell lines, each reflecting resistance to therapies targeting the AR-LBD.

We originally proposed that the VPC-14449–DBD interaction depends on Q592 and Y594 side chains (19), but did not explore proximal residues predicted to also facilitate binding. Thus, to identify other key residues, we mutated K591-L595 (KQKYL) of the AR and performed reporter assays in PC-3 cells. Three point mutants, Q592D, K593D, and Y594D, weakened inhibition of the full-length AR (1 μmol/L VPC-14449), revealing the central lysine as another critical side chain for ligand recruitment (Fig. 2A). In addition, simultaneous mutations to the binding site were introduced with a chimera construct in which the AR-DBD was replaced with the GR DBD, where the KQKYL sequence is replaced by QHNYL. Reporter assays with overexpressed AR showed that the chimera remained R1881-inducible and sensitive to enzalutamide inhibition, but resisted VPC-14449 inhibition compared with the wild-type AR (Fig. 2B).

To directly assess VPC-14449 binding to the AR DBD, we employed ITC with the recombinant rat AR DBD. Because of the high protein concentration used in the titrations, heat of dilution from the protein was significant; therefore, this value was subtracted from the binding isotherm in subsequent analyses. Titrating 1 mmol/L AR-DBD into 0.1 mmol/L VPC-14449 indicated a modest interaction between protein and compound, with a calculated binding constant (K_D) of approximately 64 μmol/L and N-value of 0.53 (Fig. 2C). Control experiments measuring the affinity between the AR-DBD and an oligonucleotide harboring an ARE revealed a strong interaction (56.8 nmol/L), demonstrating that the protein was properly folded and competent for DNA binding (Supplementary Fig. S2).

Given that VPC-14449 inhibited T877A and F876L AR mutations (Fig. 1A), we hypothesized that VPC-14449 would bypass other clinically relevant mutations in the AR-LBD. To test this, we considered a set of mutations (L701H, W741C, W741L, H874Y) that are induced by antiandrogen treatment based on their frequent detection in tissues and circulation of CRPC patients, but are absent in patients with localized treatment-naïve prostate cancer (31–33). As patient-derived cell lines harboring these mutations were not available, we performed luciferase reporter assays with transient overexpression of wild-type or mutant AR. Consistent with the DBD as the site of action, the transcriptional activity of every tested AR-LBD mutant was inhibited, to various degree, by VPC-14449 (Fig. 3A).

Because VPC-14449 and enzalutamide target distinct sites on the AR, we next tested for possible combinatorial effects. Cell viability and secreted PSA measurements in LNCaP cells simultaneously treated with doses of enzalutamide and VPC-14449 corresponding with their respective IC₂₅ values produced approximately 50% inhibition in each assay (Fig. 3B), indicating an additive effect of coadministration.

Given that VPC-14449 inhibited transcriptional activity of full-length AR and AR variants, we next analyzed the consequences of VPC-14449 on expression of AR gene targets in the panel of CRPC cell lines. R1881-dependent expression of KLK3 (PSA), TMPRSS2, and FKBP5 in LNCaP and C4-2 cells was suppressed by 5 μmol/L VPC-14449 or enzalutamide (Fig. 4A and B). R1-AD1 cells, which were derived from heterogeneous CWR-R1 cell line and express full-length AR (23), displayed a similar VPC-14449-mediated inhibition of FKPB5 and FASN, two genes under AR regulation in that cell line (Fig. 4C; refs. 23, 25). The GR-regulated gene, FKPB52, served as a negative control in all RT-PCR experiments and was unaffected by treatments with androgens or AR inhibitors.

MR49F cells contain approximately 45% of AR protein bearing the F876L mutation (23). To avoid excessive stimulation of wild-type AR, MR49F cells were cultured in FBS-supplemented media containing naturally occurring androgens. VPC-14449 suppressed target gene expression, with either no effect (KLK3, TMPRSS2) or agonist effect (FKBP5) resulting from enzalutamide treatment (Fig. 4D). To determine whether the AR and AR-V7–responsive M-phase mitotic gene UBE2C (34) was affected by VPC-14449, we performed RT-PCR analysis in 22Rv1 cells cultured under castrate conditions to prevent hormone stimulation of the endogenous AR. Significant reduction of UBE2C expression was observed with 50 μmol/L VPC-14449 (Fig. 4E) without any apparent effect of enzalutamide. Similar results were observed with R1-D567 cells that express only the truncated ARv567es variant (23), with FKBP5 and FASN expression moderately suppressed by VPC-14449 (Fig. 4F). Taken together, these findings demonstrate that VPC-14449 blocks the transcriptional program driven by mutant forms of the AR, or truncated AR variants.

22Rv1 cells, which express full-length AR as well as several truncated AR variants (AR-Vs), including AR-V7 (35), were used to analyze the effects of VPC-14449 on AR subcellular localization dynamics. Analysis of whole cell, cytosolic, and nuclear extracts revealed nuclear translocation of full-length AR in cells stimulated with DHT, whereas AR-V7 was constitutively nuclear (Fig. 5A). VPC-14449 did not alter expression or nuclear localization of AR or AR variants in these experiments (Fig. 5A; ref. 19). Separation of nuclear extracts into soluble and insoluble (chromatin bound) fractions revealed a reduction in association of full-length AR with chromatin in cells treated with >1 μmol/L VPC-14449. The effect of VPC-14449 on chromatin binding of AR-V7 was less apparent in 22Rv1 cells, even at a concentration of 50 μmol/L (Fig. 5A).

To test this mechanism more broadly, we performed chromatin fractionation in an expanded panel of AR-positive cells treated with VPC-14449. In LNCaP and R1-AD1 cells, VPC-14449 inhibited androgen-induced chromatin binding of AR at concentrations >1 μmol/L (Fig. 5B). Similar to 22Rv1 cells, higher concentrations of VPC-14449 (>10 μmol/L) were required to inhibit ARv567es chromatin binding in the R1-D567 cell line (Fig. 5C) and AR-V7 binding in CWR-R1 cells (Fig. 5D). In C4-2 (androgen insensitive) or MR49F (enzalutamide resistant) cell lines, VPC-14449 at higher concentrations (>10 μmol/L) was also required to inhibit binding of AR to chromatin (Fig. 5E).

Given the ability of VPC-14449 to prevent large-scale interactions between the AR and chromatin across multiple cell lines, we explored the ability of VPC-14449 to block specific chromatin-binding events of AR and AR variants using ChIP approaches in the isogenic R1-AD1 and R1-D567 cell lines. In R1-AD1 cells, recruitment of the full-length AR to AR-binding sites (ARBS) within or near the FASN, FKBP5, and TSC2 genes (25) was stimulated by DHT treatment, but suppressed with 50 μmol/L VPC-14449 (Fig. 5F) or 10 μmol/L enzalutamide (Supplementary Fig. S3A). However, in R1-D567 cells, VPC-14449 inhibited constitutive binding of ARv567es to the FASN and FKBP5 ARBSs, but not the TSC2 exon 37 ARBS (Fig. 5G). Enzalutamide, at a concentration of 10 μmol/L, had no effect on ARV567es binding and is consistent with the lack of an LBD of this variant (Supplementary Fig. S3B). Lowering the concentration of VPC-14449 (10 μmol/L) still resulted in robust inhibition of full-length AR binding to ARBS (Supplementary Fig. S3A), but resulted in less efficient inhibition of ARv567es binding (Supplementary Fig. S3B). Control ChIP experiments that probed chromatin regions known to be unoccupied by AR or ARv567es showed no change in signal caused by treatment with DHT or VPC-14449 (Supplementary Fig. S4). Collectively, these results demonstrate that moderate to high levels of VPC-14449 inhibit chromatin binding of diverse AR species, including mutant forms of full-length AR and truncated AR variants.

To test genome-wide changes in AR and ARv567es recruitment to ARBSs in response to VPC-14449, we performed ChIP-seq experiments in R1-AD1 and R1-D567 cells using an antibody targeted to the AR-NTD. Peak calling using data from R1-AD1 cells treated with DHT revealed 3,556 ARBSs that met an FDR cutoff of 0.01. These ARBSs displayed strong, peak-centered binding of DHT-activated AR when visualized as heatmaps (Fig. 6A). DHT dependence of AR binding was evidenced by lower heatmap signal intensity in vehicle-treated cells. Consistent with chromatin fractionation data, the DHT-stimulated recruitment of AR to these sites was inhibited by VPC-14449 treatment (Fig. 6A, left). Conversely, although visual inspection of data from R1-D567 cells provided strong evidence for peak-centered binding of ARv567es to these 3,556 ARBSs, there did not appear to be any effect of VPC-14449 at the 50 μmol/L dose tested (Fig. 6A, right). To evaluate the effects of VPC-14449 more rigorously, we plotted signal intensities for AR and ARv567es binding at these 3,556 ARBSs under the various treatment conditions. These plots confirmed that VPC-14449 reduced signal intensity of AR binding in R1-AD1 cells (Fig. 6B), but had a modest effect on ARv567es binding in R1-D567 cells (Fig. 6C).

Further evaluation of VPC-14449 effects on AR and ARv567es binding was performed at the gene track level. As expected, VPC-14449 blocked DHT-dependent recruitment of AR to specific ARBS located near or within the FASN, FKBP5, and TSC2 genes in R1-AD1 cells (Fig. 6D–F). There also appeared to be only modest effects of VPC-14449 on ARv567es recruitment to FASN, FKBP5, and TSC2 ARBSs in R1-D567 cells (Fig. 6D–F). Collectively, our ChIP-seq data support the results obtained from fractionation and ChIP-PCR experiments, indicating that VPC-14449 has a more robust effect on genome-wide chromatin binding of AR compared with AR variants.

Our earlier studies characterizing AR-DBD inhibitors provided the first evidence that selectively targeting the AR-DBD interactions could be rationally achieved (18, 19). Here, we provide new evidence for the utility of DBD-specific inhibition in cell lines that reflect drug resistance mechanisms in clinical CRPC, where VPC-14449 blocked reporter expression and downregulated many known AR target genes. In addition, ITC experiments provide evidence for direct binding between VPC-14449 and the AR-DBD. The weak binding affinity observed by ITC could arise from the absence of DNA, which might be required for the exact conformation of the AR-DBD where the binding pocket for VPC-14449 is correctly formed. Indeed, VPC-14449 was rationally designed to block DNA binding based on the X-ray structure of the AR-DBD bound to double-stranded DNA.

We gained insight into the predicted compound-binding pocket by revealing new point and simultaneous mutations in the AR-DBD (K593D, KQK→QHN) that reduce, but not abolish, VPC-14449 inhibition. Despite the tolerance of these substitutions, most mutations, including S579D, F583D, and R586D, inactivate the receptor (18), indicating the structural and functional conservation of the DBD domain. For this reason, and given the lack of AR-DBD mutations in patients receiving therapy (36), we speculate that resistance could take longer to emerge from prolonged exposure to an AR-DBD–specific compound compared with enzalutamide. In addition, coadministration with enzalutamide revealed no incompatibility with VPC-14449, suggesting that dual inhibition might preemptively select against resistance causing mutations in both DBD and LBD AR domains. Accordingly, VPC-14449 retained at least partial activity toward all forms of the AR bearing clinically relevant LBD mutations in reporter assays. A subset of these AR-LBD mutants, including L702H, W742C, and W742L, required higher levels of VPC-14449 to inhibit their transcriptional activity compared with the wild-type receptor. AR proteins expressed with these specific mutations were also partially resistant to other AR inhibitors, including those directed to the BF3 site as reported previously (17), and could be due to different protein folding of each mutant or changes in the recruitment of cofactors.

VPC-14449 was less effective against transcriptional activity mediated by AR variants compared with wild-type AR as observed at the reporter assay, targeted ChIP, and genome-wide chromatin binding levels. Importantly, VPC-14449 significantly slowed the growth of variant-driven 22rv1 cells (35), although the compound only partially inhibited AR-V7 transcriptional activity most likely due to the strong AR-specific promoter of the ARR₃tk-luc construct (19) used in the reporter assay. Differential effects of VPC-14449 upon AR-Vs could stem from the absence of the LBD, which might change the conformation of the AR NTD/DBD core in a way that affects DNA recognition or the ability to bind VPC-14449. Alternatively, chromatin fractionation studies revealed that the majority of endogenously expressed AR-V7 in 22Rv1 cells is tightly associated with DNA and not present in the soluble nuclear fraction. Newly synthesized AR-V7 proteins might rapidly translocate to the nucleus and tightly associate to chromatin before VPC-14449 can exert its effect. In contrast, significant amounts of full-length AR remain in the soluble nuclear extract after R1881 stimulation, suggesting an equilibrium of free/chromatin-bound forms of the AR that may be more sensitive to compound inhibition. Given the entire set of data, we conclude that VPC-14449 does not completely separate the AR or AR variants from chromatin, but the weakening of this interaction may manifest as growth and viability defects in prostate cancer cells, likely by blocking access to specific genes regulating those processes.

To our knowledge, VPC-14449 is the only AR-specific inhibitor that is proven to directly affect chromatin binding of full-length AR or AR-Vs through the DBD. It remains to be seen whether pyrvinium, another DBD-interacting compound, operates by the same mechanism. Notably, reports have shown that pyrvinium potently inhibits the transcriptional activity of several AR variants (29). Alternatively, pyrrole-imidazole polyamides, small-molecule DNA mimics that bind specific DNA sequences (37), can block AR recruitment to certain AREs, preventing transcriptional activity and suppressing tumor growth in prostate cancer models (38). In addition to directly targeting the AR-DBD or AREs, AR or AR-V interactions with chromatin can be reduced by treatment with the bromodomain inhibitor JQ1, which targets the bromodomain and extraterminal (BET) family of chromatin readers BDR2, 3, 4, and BRDT (39). Inhibition of BRD4, which itself interacts with the AR and regulates AR expression (39), affected the chromatin structure in such a way that blocked AR and AR-Vs from interacting with ARBSs across many cell types (25). Given its combinatorial effects with enzalutamide, additive or synergistic inhibition of VPC-14449 and JQ1 is also conceivable.

In conclusion, our findings establish proof of principle for targeted inhibition of the AR DBD to overcome AR-dependent drug resistance mechanisms, prompting further preclinical development of such agents to refine their novel mechanism of action.

K. Dalal is a consultant/advisory board member for Roche Pharmaceuticals. M.E. Gleave has ownership interest (including patents) in OGX-011 and OGX 427. S.M. Dehm is a consultant/advisory board member for Alpine Bioventures, GP, LLC, Janssen Research and Development LLC, and Medivation/Astellas. A. Cherkasov is a consultant/advisory board member for Roche Pharmaceuticals. P.S. Rennie is a consultant/advisory board member for Roche Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Dalal, M. Che, N.S. Que, R. Tse, F. Ban, H. Li, M.E. Gleave, D.T. Gewirth, S.M. Dehm, A. Cherkasov, P.S. Rennie

Development of methodology: K. Dalal, M. Che, N.S. Que, K.J. Tam, M. Roshan-Moniri, E. LeBlanc, M.E. Gleave, D.T. Gewirth, P.S. Rennie

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Dalal, M. Che, N.S. Que, A. Sharma, N. Lallous, H. Borgmann, D. Ozistanbullu, R. Tse, M. Roshan-Moniri, E. LeBlanc, M.E. Gleave, D.T. Gewirth

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Dalal, M. Che, N.S. Que, A. Sharma, R. Yang, N. Lallous, H. Borgmann, D. Ozistanbullu, F. Ban, M. Roshan-Moniri, E. LeBlanc, M.E. Gleave, D.T. Gewirth, S.M. Dehm, P.S. Rennie

Writing, review, and/or revision of the manuscript: K. Dalal, M. Che, N.S. Que, N. Lallous, M. Roshan-Moniri, M.E. Gleave, D.T. Gewirth, S.M. Dehm, A. Cherkasov, P.S. Rennie, H. Li

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Che, M.E. Gleave, A. Cherkasov

Study supervision: K. Dalal, S.M. Dehm, A. Cherkasov, P.S. Rennie

We thank Dr. A. Zoubeidi for access to MR49F and MR49C cells, and Dr. C.J. Ong for C4-2 cells.

This work was funded by the Canadian Institutes of Health Research (to P.S. Rennie and A. Cherkasov). The work was also supported by a Prostate Cancer Canada - Translational Acceleration Grant (to P.S. Rennie and A. Cherkasov). The authors acknowledge support from the NIH (grant# R01CA174777; to S.M. Dehm). Financial support was further provided by a Synergistic Idea Development Award from the U.S. Department of Defense (W81XWH-14-1-0518 to S.M. Dehm; W81XWH-14-1-0519 to P.S. Rennie; W81XWH-14-1-0520 to D.T. Gewirth), and additionally the Terry Fox New Frontiers Program Project grant (TFRI project 1062; to P.S. Rennie and A. Cherkasov).

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.