Hybrid Androgen Receptor Inhibitors Outperform Enzalutamide and EPI‐001 in in vitro Models of Prostate Cancer Drug Resistance

Androgen receptor targeted therapies for prostate cancer have serious limitations in advanced stages of the disease. While resistance to the FDA‐approved enzalutamide is extensively documented, novel therapies based on epichlorohydrin scaffolds (EPI) are currently in clinical trials, but display suboptimal pharmacokinetics. Herein, we report the synthesis and biological characterisation of a novel class of compounds designed through covalently linking enzalutamide and EPI‐001 through various triazole based linkers. The compounds display an 18 to 53 fold improvement in the cell killing potency towards C4‐2b prostate cancer (PCa) cells compared to the gold standards of therapy, enzalutamide and EPI‐001. The most promising compounds were proven to exhibit their toxicity exclusively through androgen receptor (AR) mediated pathways. This work sets the basis for the first class of hybrid AR inhibitors which successfully combine two drug moieties – EPI‐001 and enzalutamide – into the same molecule.


Introduction
The androgen receptor (AR) is a hormone-activated transcription factor and is the main driver of PCa. AR activity in healthy, non-cancerous epithelial cells promotes the development and maintenance of the male reproductive system and has a wider role in other biological processes. However, dysregulation of this signalling can result in the activation of oncogenic transcription programmes that can trigger uncontrolled proliferation of cells, and thus transformation to an aggressive phenotype culminating in tumour formation. [1] While PCa patients with localised disease are treated with focal therapies or radical prostatectomy, treatment of metastatic disease requires androgen-deprivation therapy (ADT). Despite a high rate of initial response, almost all patients progress to a more advanced and incurable disease known as Castration-Resistant Prostate Cancer (CRPC). It is well known that AR signalling is maintained or restored in CRPC despite low serum levels of circulating androgens achieved by ADT. [2] Patients will then receive second-generation AR inhibitors such as enzalutamide which competitively binds to the AR ligandbinding domain (LBD), outcompeting dihydrotestosterone and thus inhibiting downstream AR signalling and causing PCa cancer to regress.
Enzalutamide is an FDA-approved non-steroidal antiandrogen for metastatic CRPC, which is administered together with ADT as well as monotherapy for CRPC. [3] A crucial limitation of the treatment is the development of resistance, with AR signalling becoming unresponsive to enzalutamide. [4] Apart from enzalutamide, a variety of steroidal and non-steroidal anti androgens have been developed to date. [5] Notably, resistance to enzalutamide arises primarily due to AR variants (AR-Vs) which lack the LBD and are able to sustain AR signalling in the presence of enzalutamide. [6] Overall, these limitations highlight a need for complementary strategies to inhibit the AR.
The priority as an alternative to LBD inhibition has been inhibiting the N-terminal domain (NTD). This campaign has yielded a promising class of bisphenol A derived compounds, under the generic name of EPI analogues. Two EPI derivatives with undisclosed structures are currently in clinical trials for CRPC, highlighting their therapeutic potential. [7] The compound EPI-001 represents a mixture of four stereoisomers which are able to inhibit protein-protein interactions between AR-NTD and RAP74, [8] an essential component of the AR transcriptional machinery. More specifically, EPI-001 binds to a region in the AR-NTD called transactivation-unit 5 on the AR NTD, corresponding to residues 361-537. [9] Because EPIs bind to the AR-NTD, they effectively inhibit a broad range of AR-Vs, most of which are implicated in the development of CRPC. [10] Notably, EPI inhibitors inhibit constitutively active, hormone independent AR variants that lack their LBD, as well as AR variants which have acquired gain-of-function mutations in the LBD. [11] Even if EPIs are efficient at inhibiting CRPC specific splice variants that lack their LBD, they display poor pharmacokinetics properties and half inhibitory concentrations (IC 50 ) in the high micromolar range. [12][13][14] Hence, patients suffer from excessive pill burden. It has been shown that patients have significantly lower circulating doses than what would be required in vitro for optimal therapeutic concentrations. [15] Hence, structural modifications are needed in order to improve EPI's inhibition profile.
This work explores the in vitro potential of new hybrid compounds that simultaneously target two sites of the AR, the N-terminal domain (NTD) and the ligand binding domain (LBD) (Figure 1). A dual target strategy by covalently linking enzalutamide to another drug has been recently successfully implemented for the development of dual inhibitors between enzalutamide and etinostat, a histone deacetylase inhibitor. [16] Similarly, there are extensive precedents for the success of heterobifunctional molecules, such as PROTACs, [17] bifunctional therapeutics [18] and multitarget compounds. [19] Moreover, significant efforts in moving away from conventional active site targeting have been attempted up to date. [20,21] As such, five compounds which covalently link enzalutamide and EPI-001 with different linker lengths were synthesised in order to explore the potential dual domain inhibition. Various triazole-PEG linkers were chosen, aiming to cover a range of different linker lengths. Triazole linkers are known to be biocompatible and have been widely used. [22,23] This strategy could also improve binding affinity through an entropy driven effect. Furthermore, covalently linking two pharmacophores could help overcome resistance to conventional enzalutamide only based therapy.

Synthetic strategy
Five compounds were synthesised exploring various linker types and sizes between an enzalutamide and an EPI-001 moiety. The strategy aimed at linking enzalutamide through a previously explored growing vector to the EPI moiety. [24] The EPI moiety was functionalised in a way that preserved its essential chloride moiety, [25] while the opposite end was used as a growing vector for linker attachment. The compounds were synthesised as detailed in Figure 2b. A family of alkyne-alcohols 5 a-5 e was reacted with the mono-chloro derivative 2 in an erbium (III) catalysed epoxide ring opening to yield 3 a-3 e. A difficult hydrolysis of enzalutamide 6 yielded carboxylic acid 7 which was coupled to an azido-containing linker to yield 8. Copper catalysed azide-alkyne cycloadditions were chosen as the last step to yield the final compounds 9 a-9 e.

Cellular toxicity assays
The five compounds were tested against a C4-2b prostate cancer cell line. The C4-2b cell line is sequentially derived from the androgen independent C4-2 and androgen sensitive LNCaP cell lines that originated from a patient with metastatic carcinoma. C4-2b are a more aggressive sub-line of these cells. Unlike LNCaP, C4-2b expresses several AR-Vs (Supplementary Figure 1), this displaying enzalutamide resistance, making it a promising testing system for our compounds. [26,27] Cellular toxicity studies showed that compounds 9 a-9 e display an 18 to 53 fold improvement in the half lethal concentration (LC 50 ) compared to enzalutamide and EPI-001 against the C4-2b prostate cancer cell line (LC 50 s ranging between 1.7 and 4.6 μM vs. 63.5 μM for enzalutamide and 84.8 μM for EPI-001) ( Figure 3a). Enzalutamide displayed a nonsigmoidal response over the interval of tested concentrations, in accordance to previous findings , [28] . [29] EPI was found to have an LC 50 of 84.4 μM, in accordance with literature estimates. [14] An equimolar cocktail of enzalutamide and EPI-001 outperformed both EPI-001 and enzalutamide (LC 50 = 38.1 μM), presumably due to the genetic heterogeneity of the C4-2b cell line and prevalence of variants (Supporting Information Figure 1). [30] Crucially, compounds 9 a-9 e outperform both enzalutamide and EPI-001 by factors of 18 to 53, highlighting the success of the hybridisation strategy.
To evaluate whether the cell toxicity was AR mediated, we performed complementary cytotoxicity studies on PC-3 cells. PC-3 cells are metastatic adenocarcinoma cells that express very low levels of AR [31] and therefore both enzalutamide and EPI-001 are not toxic at concentrations lower than 100 μM. [32] Compounds 9 a-9 e were therefore expected to not display any significant toxicity towards the PC-3 cell line.
The results presented in Figure 3 show that 9e has nonsignificant toxicity towards the PC-3 cells, while 9a has significant non-AR mediated toxicity with an LC 50 of 1.6 μM. It is therefore likely that 9a's performance in C4-2b cells is non-AR mediated given its identical LC 50 values in these two cells lines (1.6 μM and 1.7 μM respectively). Notably, compounds 9 b-9 d displayed negligible toxicity towards the PC-3 cell line, suggesting that their toxicity is AR dependent (Figure 3b).

Luciferase reporter assays
Aiming to prove the direct interaction of the dual inhibitors with the AR, target engagement luciferase assays were performed. AR-null CV-1 monkey kidney cells were transfected with a Gal 4 -AR-FL plasmid construct that can be induced by androgens and can interact with Gal 4 -DNA binding sites present on the co-transfected reporter construct. This highly sensitive system reports directly on the AR-mediated transcriptional activity. In this assay, synthetic androgen metribolone (R1881) treatment (1 nM) led to a four-fold increase in luciferase activity, indicating androgen mediated AR activation, as expected. Cotreatment with enzalutamide at a concentration of 5 μM led to a significant decrease in luciferase activity, while EPI-001 was not able to inhibit gene transcription at the tested concentration (5 μM), in accordance to previous findings. [32] Results show that 9b, 9d and 9e significantly inhibit the transcriptional activity of a Gal4-full length androgen receptor (Gal4-AR-FL) construct (Figure 4a), further confirming that these compounds can inhibit the AR transcriptional activity. Compound 9 c was not able to inhibit gene transcription, suggesting its effects on cell growth are not AR mediated. Overall, the compounds were less potent than enzalutamide in engaging the luciferase reporter, which we hypothesise could be due to membrane permeability issues.

Effects on the transcriptional activity of endogenous AR
With these results in hand, we focused on investigating the effect of compounds 9 b-9 e on endogenous AR mediated transcription. The KLK3 is a gene tightly regulated by the androgen receptor, with its transcript level highly sensitive to the AR activity [33] Hence, RT-qPCR experiments were performed in order to investigate the KLK3 gene expression levels at a concentration of 5 μM for the drugs, in C4-2b cells. The synthetic androgen, R1881, was used at a concentration of 1 nM. Enzalutamide acted as a positive control at a concentration of 5 μM, while EPI-001 was not able to inhibit the KLK3 gene transcription at the tested concentration, in accordance with previous findings. [32] Compound 9 b was able to significantly inhibit the KLK3 gene transcription at the tested concentrations ( Figure 4b). While compound 9e failed to induce a statistically significant change in KLK3 gene expression (p = 0.1186), compound 9 b was able to cause a significant change (p = 0.0397). Overall, the compounds were less potent than enzalutamide in activating downstream transcription, which we hypothesise could be due to membrane permeability issues.

Discussion
Following the results of the robust biological assays, compound 9 b has emerged as the lead candidate compound from our study. With an LC 50 of 1.6 μM in C4-2b cells and negligible toxicity in PC-3 cells (> 75 % cell viability at 10 μM), compound 9 b was able to inhibit AR mediated gene transcription in the luciferase assay, proving its direct target engagement with the AR. Moreover, the compound was able to significantly inhibit the transcription of the KLK3 gene, further proving that its effects are specifically mediated via AR inhibition. We hypothesise that the marked improvement in cell toxicity could be due to an entropic effect, whereby the linker increases the effective local concentration of the second inhibitor that binds to the androgen receptor. This theory is supported especially by the fact that the dual inhibitors outperform an equimolar cocktail of EPI-001 and enzalutamide by a factor of 80.

Conclusions
We have synthesised a novel class of a previously unreported type of AR inhibitors. An 18 to 53 fold improvement in cell killing potency (LC 50 ) was obtained for all tested compounds. Cellular toxicity for 9 b-9 e was proven to be induced via AR inhibition given their lack of toxicity towards PC-3 cells.
Compound 9 e outperformed all tested compounds in target engagement dual luciferase reporter assays (p < 0.0001), indicating its potent and selective mechanism of action. KLK3 RT-qPCR assays found compound 9 b to be the most promising candidate (p = 0.0397).
Altogether, these results suggest that the architecture of compound 9 b is promising for further development and investigation. In summary, this research demonstrates the synthesis and characterisation of compound 9 b as a first-inclass AR inhibitor that is more effective and specific at inhibiting the growth of C4-2b cells compared to standard of care drug enzalutamide. This research has high potential to lead the

General experimental techniques
All experiments were performed in oven-dried glassware and under an atmosphere of nitrogen, unless stated otherwise. Commercial starting materials were used without further purification. Dry solvents were distilled from mixtures containing CaH 2 or LiAlH 4 as drying agents. Yields refer to spectroscopically and chromatographically pure compounds unless otherwise specified. Analytical thin layer chromatography (TLC) was carried out on glass Merck Kieselgel 60 F254 plates. The plates were visualised under direct UV irradiation (254 nm). R f values are quoted to the nearest 0.1. Preparative thin layer chromatography was performed on commercially available Analtech plates. Flash column chromatography was undertaken on silica gel 60 (230-.400 mesh) under a positive air pressure. The eluent systems are reported as % (v/v) of the solvent components.
Reverse phase column chromatography was carried out using a Combiflash Rf200 automated chromatography system with Redis-ep® reverse-phase C18-silica flash columns (20-40 μm). Preparative high-performance liquid chromatography (HPLC) was performed on an Agilent 1260 infinity machine. The samples were eluted using a Supelcosil ABZ + PLUS column (250 mm × 21.2 mm, 5 μm). Infrared spectra were recorded on a Perkin Elmer Spectrum One FT-IR spectrometer which is fitted with attenuated total reflectance (ATR) sampling accessory. The absorption maxima (ν) higher than 500 cm À 1 are quoted in wavenumbers (cm À 1 ) and are presented with the aid of abbreviations: w, weak; m, medium; s, strong; br, broad. Data are reported as: wavenumbers, assignment. 1 H NMR spectra were recorded under an internal deuterium lock at rt on Bruker Advance III HD (400 MHz, 500 MHz, 700 MHz; Smart probe). Assignments are supported by 1 H-1 H COSY, 1 H-13 C HSQC and 1 H-13 C HMBC spectra. Chemical shifts (δ) are given in ppm quoted to the nearest 0.01 ppm (δ H ). The residual solvent peaks are 7.26 for CDCl 3 , 5.32 for CD 2 Cl 2 , 3.31 for CD 3 OD and 2.51 for (CD 3 ) 2 SO. Coupling constants for mutually coupling protons are reported in Hertz, rounded to the nearest 0.1 Hz. Data are reported as: chemical shift, multiplicity (br, broad; s, singlet; d, doublet; t, triplet; q, quartet; m, multiplet; or a combination of them), coupling constants, number of nuclei. Spectra were processed using TopSpin v.4.0.6(Bruker). Carbon magnetic resonance spectra were recorded using an internal deuterium lock at rt on Bruker Avance III HD (101 MHz) with broadband proton decoupling. Chemical shifts (δ C ) are quoted to the nearest 0.1 ppm and the solvent reference peaks (in ppm) are 77.2 (CDCl 3 ), 53.5 (CD 2 Cl 2 ), 49.1 (CD 3 OD), 33.0 (CD 3 ) 2 SO). Fluorine magnetic resonance spectra were recorded using an internal deuterium lock at rt on Bruker Avance III HD (101 MHz) with broadband proton decoupling. Chemical shifts (δ C ) are quoted to the nearest 0.1 ppm. For fluorine containing compounds, data are reported as: chemical shift, multiplicity, coupling constant. For the other compounds, data are reported as chemical shifts. Spectra were processed using TopSpin v.4.0.6(Bruker).
High resolution mass spectrometry (HRMS) measurements were performed on a Waters LCT Premier Time of Flight mass spectrometer, with errors within � 5 ppm.

General synthetic procedure for 5 b-5 e
To a mixture of corresponding diol (1 eq., 3.1 M) in THF cooled to 0°C was added NaH (0.4 eq.) in small portions. A solution of propargyl bromide in toluene (6.6 M, 0.62 eq.) was added dropwise at 0°C. The mixture was subsequently stirred at room temperature for 24 h, and then H 2 O (30 mL) was added dropwise to the mixture at 0°C. The resulting mixture was extracted with CH 2 Cl 2 , dried with Na 2 SO 4 and the solvent subsequently removed in vacuo. The resulting alkynes were purified by silica gel chromatography (CH 2 Cl 2 /EtOAc and acetone/hexane mixtures) to yield the corresponding alkynes.

General synthetic procedure for 3 b-3 e
To a solution of 2 in CH 3 CN (0.26 M) was added corresponding alkyne (6.6 eq.) and the mixture stirred for 10 minutes. Er(OTf) 3 (0.2 eq.) was subsequently added and the mixture was stirred at room temperature for 24 h. The solvent subsequently removed under a stream on N 2 . The mixture was purified by reverse phase column chromatography (H 2 O/MeCN 9 : 1-1 : 9) to yield the corresponding alkynes.

Cell viability assays
AR-positive C4-2b cells were treated with compounds at a range of concentrations for 72hr in full media in 96-well plates. The range of concentrations for C4-2b were mirrored in AR-null PC-3 cells. Cell viability was determined using the CellTiter 96 Aqueous MTS assay (Promega), as per supplier's instructions whereby cells were incubated with the reagent 3-(4,5-dimethylthiazol-2-yl)-5-(3carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) and OD was measured following an incubation period. Viability was calculated by normalising to negative control wells containing media only and displaying as a percentage of DMSO-treated cells. The DMSO concentration was 0.1 % in all wells.

Luciferase assays
CV-1 cells were transfected with plasmids AR-FL-Gal4, p(UAS4)-TATA-luc and internal control, Renilla-luc reporter plasmid using Lipofectamine 2000 (Invitrogen) in a 10 cm dish following manufacturer's protocol. After 24hr, cells were treated with drugs + /À hormone in a white 96-well plate in medium containing charcoalstripped FBS. 48hr post-treatment, firefly and renilla luciferase signals were measured using the Twinlite kit (Perkin Elmer) according to manufacturer's instructions. Transactivation of the AR was calculated by normalising to the Renilla internal control and was displayed as a fold-induction compared to Vehicle no drug control wells. The DMSO concentration was 0.1 % in all wells.

RT-qPCR
Cells were treated + /À R1881 and with drug compounds/DMSO for 24 hr in 6-well plates. Cells were then harvested, and RNA extracted using the RNeasy Plus Mini Kit (Qiagen), according to the kit's instructions. RNA was then reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). For qPCR, the QuantiNova SYBR Green PCR Kit (Qiagen) was used to measure gene expression compared to the internal control gene, GAPDH. Delta-delta-ct analysis method was applied to generated values such that conditions were displayed as a fold-induction compared to Vehicle no drug control wells (containing the equivalent DMSO concentration as drug-treated wells). The DMSO concentration was 0.1 % in all wells.