Development of Novel Inhibitors Targeting the D-Box of the DNA Binding Domain of Androgen Receptor

The inhibition of the androgen receptor (AR) is an established strategy in prostate cancer (PCa) treatment until drug resistance develops either through mutations in the ligand-binding domain (LBD) portion of the receptor or its deletion. We previously identified a druggable pocket on the DNA binding domain (DBD) dimerization surface of the AR and reported several potent inhibitors that effectively disrupted DBD-DBD interactions and consequently demonstrated certain antineoplastic activity. Here we describe further development of small molecule inhibitors of AR DBD dimerization and provide their broad biological characterization. The developed compounds demonstrate improved activity in the mammalian two-hybrid assay, enhanced inhibition of AR-V7 transcriptional activity, and improved microsomal stability. These findings position us for the development of AR inhibitors with entirely novel mechanisms of action that would bypass most forms of PCa treatment resistance, including the truncation of the LBD of the AR.


Introduction
Prostate cancer (PCa) is the most prevalent malignancy in men and the second most common cause of cancer-related death in Canada and the United States (US) among males [1]. The androgen receptor (AR), a ligand-inducible transcription factor, plays a critical role in the development of PCa, where it controls the transcription of onco-driving genes [2][3][4][5][6]. The inhibition of androgen signaling remains one of the major strategies for PCa treatment; the early stages of PCa respond well to androgen deprivation therapy, surgery, and radiation. Unfortunately, most patients eventually develop a more aggressive and lethal androgen-independent form of the disease called castration-resistant PCa (CRPC) that arises through various mechanisms, including gain-of-function mutations in the AR gene and increased AR expression [7][8][9].
The development of more recent antiandrogen drugs, such as enzalutamide and abiraterone, that competitively bind the ligand-binding domain (LBD) of AR helped decrease PCa mortality and morbidity significantly [10][11][12][13]. However, resistance to these drugs could also eventually develop by diverse means, including LBD mutations and the occurrence of splice variants in the AR [9,[14][15][16]. In particular, the CRPC-prevalent AR-V7 splice variant completely lacks the LBD and becomes resistant to all the LBD-binding drugs, which comprise the whole spectrum of clinically used antiandrogens [17]. Hence, the development of LBD-independent drugs becomes an attractive strategy to combat advanced forms of PCa [18].
In previous studies [18][19][20], we have demonstrated that the DNA-binding domain (DBD) located between the N-terminal domain (NTD) and the LBD is an attractive drug In a previous study [26], we identified a druggable binding site ( Figure 1C) on the surface of the D-box and developed small molecules that disrupt the AR functional dimerization. Our lead P-box inhibitor VPC-17005 suppressed androgen signaling, demonstrated AR specificity through mutagenesis studies, and inhibited the growth of AR-positive cancer cell lines. However, VPC-17005 demonstrated rather poor metabolic stability making it a non-suitable candidate for in vivo studies and future optimization. Thus, we decided to expand the repertoire of AR DBD-dimerization inhibitors and evaluate more chemotypes by using a combination of docking, pharmacophore modeling, and a diverse set of biochemical assays.
Herein, we report the newly discovered series of chemically diverse small molecules that were characterized in numerous AR-relevant assays and shown to abrogate DBD dimerization effectively. In particular, two lead compounds, VPC-17160 and VPC-17281, demonstrated stronger inhibition of AR-V7 transcriptional activity and exhibited superior activity in mammalian two-hybrid assays compared to our previously identified lead (VPC-17005). Furthermore, compound VPC-17281 appeared to have significantly improved microsomal stability, which is critical for future hit-to-lead optimization. Overall, In a previous study [26], we identified a druggable binding site ( Figure 1C) on the surface of the D-box and developed small molecules that disrupt the AR functional dimerization. Our lead P-box inhibitor VPC-17005 suppressed androgen signaling, demonstrated AR specificity through mutagenesis studies, and inhibited the growth of AR-positive cancer cell lines. However, VPC-17005 demonstrated rather poor metabolic stability making it a non-suitable candidate for in vivo studies and future optimization. Thus, we decided to expand the repertoire of AR DBD-dimerization inhibitors and evaluate more chemotypes by using a combination of docking, pharmacophore modeling, and a diverse set of biochemical assays.
Herein, we report the newly discovered series of chemically diverse small molecules that were characterized in numerous AR-relevant assays and shown to abrogate DBD dimerization effectively. In particular, two lead compounds, VPC-17160 and VPC-17281, demonstrated stronger inhibition of AR-V7 transcriptional activity and exhibited superior activity in mammalian two-hybrid assays compared to our previously identified lead (VPC-17005). Furthermore, compound VPC-17281 appeared to have significantly improved microsomal stability, which is critical for future hit-to-lead optimization. Overall, these findings provide insights on the inhibition of DBD dimerization and provide a foundation for further development of AR-targeting drugs that bypass most of the known mechanisms of PCa drug resistance.

In Silico Identification of VPC-17160 and VPC-17281
Our earlier lead VPC-17005 demonstrated good AR inhibition but sub-optimal pharmacokinetic properties with low microsomal stability (t 1/2 = 14 min). We utilized the structure of rat AR DBD dimer bound to the DR3 oligonucleotide (PDB (Protein Data Bank) code: 1R4I) to dock drug-like compounds from the ZINC15 database into the previously identified binding site [21,27]. We then developed a pharmacophore model based on the binding pose of VPC-17005 (Supplementary Figure S1). The pharmacophore model consists of one hydrogen-bond acceptor and one aromatic ring feature, as these interactions were noted to be crucial in our previous study. We then selected 240 compounds based on this pharmacophore model that were purchased and tested for AR inhibition activity.

VPC-17160 and VPC-17281 Reduces AR-FL and AR-V7 Transcriptional Activities
Compounds were first screened for their inhibition of AR-V7 transcriptional activity. To do so, we created a PC3 cell-based population that stably encodes (i) a doxycyclineinducible AR-V7 cDNA sequence (pLIX402 tet-inducible promoter, puromycin selection), and (ii) the ARR3tk-nanoLuciferase reporter construct. Hence, nanoluciferase production was directly correlated with AR-V7 activity. This allowed the screening of the compounds without using the error-prone method of transiently transfecting V7 and reporter constructs. The inhibition efficacy of compounds was measured using doxycycline treatment along with an increasing dose of compounds. From this screen, a panel of compounds was identified with some anti-dimerization activity (Supplementary Table S1). Two hit compounds, VPC-17160 and VPC-17281, showed complete AR-V7 inhibition at 10 µM ( Figure 2D). To rule out direct effects of the compounds on luciferase expression or PC3 cell viability, we used a PC3-dox inducible luciferase cell line in parallel. Using a concentrationdependent titration, we next established that their IC 50 values for the inhibition of the AR-V7 transcription were 6 µM for both compounds ( Figure 2A). This data indicates a 2-fold improvement in the potency of AR-V7 inhibition in comparison with the parental compound VPC-17005 (IC 50 = 10 µM).
In addition, these compounds were also tested for their ability to inhibit full-length AR transcriptional activity in LNCaP cells using an enhanced green fluorescent protein (eGFP) AR transcriptional assay, where the expression of eGFP is under the direct control of an androgen-responsive probasin-derived promoter [28]. Compounds VPC-17160 and VPC-17281 showed significant inhibition of AR-FL expression at 12 µM concentration with estimated IC 50 of 2 and 5 µM, respectively (Figure 2B and Supplementary Table S1).
To further validate compounds as AR inhibitors, we tested their activity by quantifying the effect on the production of endogenous prostate-specific antigen (PSA) in PCa cell lines. PSA is an AR-regulated serine protease and is widely used as a biomarker for PCa [29]. As expected, VPC-17160 and VPC-17281 induced a significant decrease in secreted PSA levels in LNCaP cells with corresponding IC 50 values established at 2 µM and 6 µM, respectively ( Figure 2C). In addition, these compounds were also tested for their ability to inhibit full-length AR transcriptional activity in LNCaP cells using an enhanced green fluorescent protein (eGFP) AR transcriptional assay, where the expression of eGFP is under the direct control of an androgen-responsive probasin-derived promoter [28]. Compounds VPC-17160 and VPC-17281 showed significant inhibition of AR-FL expression at 12 μM concentration with estimated IC50 of 2 and 5 μM, respectively (Figure 2B and Supplementary Table S1).
To further validate compounds as AR inhibitors, we tested their activity by quantifying the effect on the production of endogenous prostate-specific antigen (PSA) in PCa cell lines. PSA is an AR-regulated serine protease and is widely used as a biomarker for PCa [29]. As expected, VPC-17160 and VPC-17281 induced a significant decrease in secreted PSA levels in LNCaP cells with corresponding IC50 values established at 2 μM and 6 μM, respectively ( Figure 2C).

Evidence for an Anti-Dimer Mechanism
To validate that this activity occurred via anti-homodimer inhibition rather than androgen displacement at the LBD, we measured the ability of VPC-17160 and VPC-17281 to displace a fluorescent-tagged androgenic ligand (FluormoneAL Green) from the AR-LBD. Even at high concentrations (50 μM), no significant hormone displacement was observed for compound VPC-17281 (Table 1). We observed around 25% of displacement for compound VPC-17160 at 10 μM, a concentration higher than the IC50 of the compound in transcription assays. In contrast, parental compound VPC-17005 was displacing 75% of the androgen at 10 μM, suggesting that this compound was having some of its efficacy through the ligand-binding site. This would explain why VPC-17160 and VPC-17281 are more potent inhibitors than VPC-17005 in the AR-V7 transcription assay.

Evidence for an Anti-Dimer Mechanism
To validate that this activity occurred via anti-homodimer inhibition rather than androgen displacement at the LBD, we measured the ability of VPC-17160 and VPC-17281 to displace a fluorescent-tagged androgenic ligand (FluormoneAL Green) from the AR-LBD. Even at high concentrations (50 µM), no significant hormone displacement was observed for compound VPC-17281 (Table 1). We observed around 25% of displacement for compound VPC-17160 at 10 µM, a concentration higher than the IC 50 of the compound in transcription assays. In contrast, parental compound VPC-17005 was displacing 75% of the androgen at 10 µM, suggesting that this compound was having some of its efficacy through the ligand-binding site. This would explain why VPC-17160 and VPC-17281 are more potent inhibitors than VPC-17005 in the AR-V7 transcription assay. Next, the dimerization of full-length AR was assessed by a mammalian two-hybrid (M2H, [30]) assay, where target proteins were fused with Gal4-DBD or VP16 domains. Co-expression of Gal4-DBD-AR (pB-AR) and VP16-AR (pA-AR) resulted in luciferase activity from the Gal4 responsive PG5-NLuc reporter, signifying dimerization (Figure 3). High concentrations of VPC-17005 (>20 µM) were required for significant reduction in AR and AR-V7 homo-dimerization. In contrast, VPC-17160 and VPC-17281 induced full reduction of dimerization at 20 µM concentration, suggesting a better efficacy of these compounds to inhibit dimerization.
Next, the dimerization of full-length AR was assessed by a mammalian two-hybrid (M2H, [30]) assay, where target proteins were fused with Gal4-DBD or VP16 domains. Coexpression of Gal4-DBD-AR (pB-AR) and VP16-AR (pA-AR) resulted in luciferase activity from the Gal4 responsive PG5-NLuc reporter, signifying dimerization (Figure 3). High concentrations of VPC-17005 (>20 μM) were required for significant reduction in AR and AR-V7 homo-dimerization. In contrast, VPC-17160 and VPC-17281 induced full reduction of dimerization at 20 μM concentration, suggesting a better efficacy of these compounds to inhibit dimerization.

Effect on the Growth of Prostate Cancer Cell Lines
VPC-17160 and VPC-17281 were further evaluated for their potential to inhibit the growth of prostate cancer cell lines LNCaP (AR-FL dependent) and 22rv1 (AR-V7 dependent). As anticipated, all compounds effectively inhibited the growth of the AR-dependent LNCaP cell line. Compound VPC-17281 showed an IC50 of 25 μM in the LNCaP cell line. Interestingly, compound VPC-17281 caused a significant reduction in the growth of 22rv1 cells (IC50 = 10 μM; see Figure 4), which suggests that this compound truly targeted the truncated variant of the AR.

Effect on the Growth of Prostate Cancer Cell Lines
VPC-17160 and VPC-17281 were further evaluated for their potential to inhibit the growth of prostate cancer cell lines LNCaP (AR-FL dependent) and 22rv1 (AR-V7 dependent). As anticipated, all compounds effectively inhibited the growth of the AR-dependent LNCaP cell line. Compound VPC-17281 showed an IC 50 of 25 µM in the LNCaP cell line. Interestingly, compound VPC-17281 caused a significant reduction in the growth of 22rv1 cells (IC 50 = 10 µM; see Figure 4), which suggests that this compound truly targeted the truncated variant of the AR.  We then tested the metabolic stability of these molecules using human microsomes. Additionally, the measured half-life in microsomes increased by approximately six-fold for compound VPC-17281 (T1/2 > 90 min) compared to VPC-17005 (T1/2 = 14 min), whereas compound VPC-17160 showed poor stability with a half-life of 3 min. This could explain the lower potency of this compound in cell growth assays, as it is metabolized quickly in the media.

Discussion
Prostate cancer growth is primarily driven by AR signaling, which could be effec- We then tested the metabolic stability of these molecules using human microsomes. Additionally, the measured half-life in microsomes increased by approximately six-fold for compound VPC-17281 (T 1/2 > 90 min) compared to VPC-17005 (T 1/2 = 14 min), whereas compound VPC-17160 showed poor stability with a half-life of 3 min. This could explain the lower potency of this compound in cell growth assays, as it is metabolized quickly in the media.

Discussion
Prostate cancer growth is primarily driven by AR signaling, which could be effectively depleted with androgen-competing chemicals. All drugs currently on the market act through binding to the ligand-binding domain (LBD) of the AR. Unfortunately, a large number of patients develop resistance to androgen displacement treatment. Thus, the development of drugs that target a different site on the surface of AR (such as the dimerization site) remains an attractive strategy to combat drug resistance.
Previously we reported VPC-17005, a small molecule inhibitor that targets the dimerization interface on the AR DBD that demonstrated promising AR inhibition in androgensensitive LNCaP and enzalutamide-resistant MR49F cells in vitro [26]. However, further use of VPC-17005 was limited by poor metabolic stability and limited potency against the AR-V7. Thus we have utilized the structure of VPC-17005 as a template to develop further improved AR dimerization inhibitors that demonstrate enhanced target affinity and drug-like properties. From this screen combining docking and pharmacophore modeling with wet-lab experiments, we identified two leads-VPC-17281 and VPC-17160-that outperformed the parent compound. These new AR DBD P-box inhibitors are likely to be more specific as their cross-interaction with the LBD site was minimal (according to the DHT displacement assay). Furthermore, this observation was in agreement with the enhanced activity of VPC-17281 and VPC-17160 against the AR-V7 truncated form of the receptor. The docking models we developed indicate that the enhanced affinity towards the DBD dimerization site could result from a larger hydrophobic surface area of the newly developed compounds and the increased number of hydrophobic contacts with the protein ( Figure 5). Other additional protein-ligand interactions include a hydrophobic contact formed with Leu595 by both VPC-17160 and VPC-17281 but not VPC-170005 ( Figure 5). In addition, these molecules are generally bulkier than the parent compound, and, thus, they are less likely to be accommodated by the ligand-promiscuous LBD site. The higher affinity of VPC-17281 for the receptor could also be justified by the formation of an additional hydrogen bond between the hydroxyl group and Asn611 as predicted by the binding pose ( Figure 5). Overall, the developed pharmacophore model successfully identified compounds that bind the dimerization site and could be used for further screenings. Furthermore, both lead compounds VPC-17160 and VPC-17281 inhibited the growth of 22rv1 and LNCaP cancer cell lines, proving the antineoplastic effects of the dimer inhibition. Importantly, the profound activity of VPC-17281 on the AR-V7 dependent 22rv1 cell line provides additional assurance as to the effectiveness of our lead inhibitor and its applicability for targeting castration-resistant prostate cancer cells. However, it is important to mention that VPC-17281 also demonstrated some activity in the PC3 cell line, which suggests off-target effects and possible toxicity. In contrast, VPC-17160 showed no effect on the PC3 cell lines suggesting its selectivity to AR and could represent the best Furthermore, both lead compounds VPC-17160 and VPC-17281 inhibited the growth of 22rv1 and LNCaP cancer cell lines, proving the antineoplastic effects of the dimer inhibition. Importantly, the profound activity of VPC-17281 on the AR-V7 dependent 22rv1 cell line provides additional assurance as to the effectiveness of our lead inhibitor and its applicability for targeting castration-resistant prostate cancer cells. However, it is important to mention that VPC-17281 also demonstrated some activity in the PC3 cell line, which suggests off-target effects and possible toxicity. In contrast, VPC-17160 showed no effect on the PC3 cell lines suggesting its selectivity to AR and could represent the best starting template for further drug optimization.
It should also be noted that microsomal stability represents one of the key pharmacokinetic properties of a candidate drug since a low microsomal half-life usually indicates poor bioavailability. VPC-17281 showed a greatly improved microsomal stability, while VPC-17160 was readily metabolized in microsomes. These results indicate that further optimization of the compounds is necessary to overcome the issues of toxicity and low microsomal stability.

In Silico Identification Experiments
The crystal structure of rat AR DBD was downloaded from PDB database (1R4I) and prepared with Protein Preparation Wizard from Schrödinger [31,32]. The preparation included the addition of hydrogen atoms (followed by energy minimization), assignment of bond orders, and filling of the missing side chains. Compounds were derived from the ZINC15 database and then prepared with MOE (Molecular Operating Environment) [27,33]. In particular, molecules were protonated, partial charges were added, and then they were subjected to energy minimization. The docking was performed with GlideSP from Schrödinger [34]. Pharmacophore model development and screening were also performed using MOE.

Androgen Displacement Assay
Androgen displacement was assessed with the Polar Screen Androgen Receptor Competitor Green Assay Kit as per the instructions of the manufacturer.
The mammalian two-hybrid assay was performed in PC3 cells by co-transfection of 15 ng pA-AR or pB-AR plasmids, with 12.5 ng PG5-NLuc reporter. PC3_iV7_3TKNluc or PC3_iNluc cells were seeded in DMEM +5% CSS at 5000 cells/well for 24 h, followed by adding doxycycline/compounds (18 h) and luminescence measurement. LNCaP and 22rv1 cells were starved in RPMI + CSS (48 h) before cell viability tests using the same conditions as above, followed by 30 µL/well PrestoBlue™ Reagent (1 h) and fluorescence measurement (TECAN F200).

Statistical Analysis
All statistical analysis and dose-response curves were generated using GraphPad Prism 6.07 software (GraphPad Software, La Jolla, CA, USA), www.graphpad.com (accessed 1 March 2021).

Conclusions
In the current study, we discovered novel inhibitors of AR DBD dimerization through the synergetic application of virtual screening methods and experimental validation. The study was built upon a previously reported AR DBD inhibitor VPC-17005 that suffered from sub-optimal biostability and pharmacokinetics. The newly developed lead compounds, VPC-17281 and VPC-17160, belong to different chemotypes and exhibit significantly improved microsomal stability and anti-AR activity. This study broadens the repertoire of compounds targeting AR DBD dimerization and could provide useful insights for the further development of novel antiandrogens with unique and novel mechanisms of action. These drugs would be of immense value for CRPC patients that do not respond to conventional AR inhibitors.