Small-Molecule Inhibition of Androgen Receptor Dimerization as a Strategy against Prostate Cancer

The clinically used androgen receptor (AR) antagonists for the treatment of prostate cancer (PCa) are all targeting the AR ligand binding pocket (LBP), resulting in various drug-resistant problems. Therefore, a new strategy to combat PCa is urgently needed. Enlightened by the gain-of-function mutations of androgen insensitivity syndrome, we discovered for the first time small-molecule antagonists toward a prospective pocket on the AR dimer interface named the dimer interface pocket (DIP) via molecular dynamics (MD) simulation, structure-based virtual screening, structure–activity relationship exploration, and bioassays. The first-in-class antagonist M17-B15 targeting the DIP is capable of effectively disrupting AR self-association, thereby suppressing AR signaling. Furthermore, M17-B15 exhibits extraordinary anti-PCa efficacy in vitro and also in mouse xenograft tumor models, demonstrating that AR dimerization disruption by small molecules targeting the DIP is a novel and valid strategy against PCa.


General Methods and Materials
All the chemical reagents and reaction solvents were purchased commercially and of research grade or better. All the reactions were monitored by thin-layer chromatography (TLC) and then visualized with an UV lamp (254 nm). All final compounds were purified to ≥95% by HPLC with UV detection at 254 nm. NMR spectra were acquired on a Bruker AVANCE III 500 spectrometer with TMS as the internal standard in DMSO-d6 or CDCl3.

General Procedure for Synthesis of Compounds 3a-3l
To a stirring suspension of N-Bromosuccinimide (6.60 mmol) and p-Toluenesulfonic acid (3.00 mmol) in 20 mL MeCN was added acetophenone (1a-1l, 6.00 mmol). This solution was stirred at 55 °C for 1 h, then cooled to room temperature, and water (20 mL) was added. The mixture was extracted with EtOAc (3 × 30 mL), and the combined organic layer was washed with brine (2 × 50 mL), dried over anhydrous Na2SO4, and concentrated To a solution of compound 3a-3l (2.50 mmol) in 15 mL MeCN was added trifluoromethanesulfonic acid (2.50 mmol) under a nitrogen atmosphere. This solution was heated to reflux for 1 h, then cooled to room temperature, and water (20 mL) was added. The mixture was extracted with EtOAc (3 × 30 mL), and the combined organic layer was washed with brine (2 × 50 mL), dried over anhydrous Na2SO4, and concentrated to give a residue, which was purified by column-chromatograph (EtOAc/PE) to give the desired product.
A suspension of compound 5a-5l (1.50 mmol) and potassium thiocyanate (1.65 mmol) in 10 mL MeCN was stirred at 60 °C for 30 min, and then 5-methylisoxazol-3amine (1.50 mmol) was added and stirred for another 1 h. After cooling to room temperature, water (20 mL) was added to the mixture and was extracted with EtOAc (3 × 30 mL). The combined organic layer was washed with brine (2 × 50 mL), dried over anhydrous Na2SO4, and concentrated to give a residue, which was purified by columnchromatograph (EtOAc/PE) to give the desired product.
The mixture was extracted with EtOAc (3 × 10 mL). The combined organic layer was washed with water (3× 50 mL) and then brine (2 × 50 mL), dried over anhydrous Na2SO4, and concentrated to give a residue, which was purified by column-chromatograph
The reaction mixture was cooled to room temperature, alkalized with saturated aqueous NaHCO3 to pH = 10 and filtrated. The filtrate was extracted with EtOAc (3 × 20 mL), and the combined organic layer was washed with brine (2 × 30 mL), dried over anhydrous Na2SO4, and concentrated to give a residue, which was purified by columnchromatograph (DCM/MeOH) to give compound M17-B20.

Construction of the AR W751R and AR F754V Mutated Models
The AR W751R and AR F754V homodimer models were generated by mutating W751 to R and F754 to V respectively in the crystal structure of the wild type AR LBD dimer in complex with an agonist DHT in each monomer (PDB entry: 5JJM) by the mutate module in Schrö dinger (v.2015). 1 The structures were then prepared by the Protein Preparation Wizard module in Schrö dinger, including removing all non-bonded hetero-atoms and water molecules, adding missing side chains, and optimizing the structure to relieve steric clashes.

Molecular Dynamics (MD) Simulations for Dimers of AR WT , AR W751R and AR F754V
The three prepared structures were used as the initial structure for the MD simulations.
The partial charges of DHTs were obtained using the restrained electrostatic potential (RESP) fitting technique based on the electrostatic potentials computed at the Hartree-Fock (HF) SCF/6-31G* level by Gaussian16. 2,3 Afterwards, the topology and parameter files of DHTs were generated by using the antechamber module in Amber 18. 2,4 The molecular mechanics (MM) parameters from the ff14SB and GAFF2 force fields were assigned to the proteins and DHTs, respectively, using the LEaP module of Amber 18 5, 6 .
Each system was solvated into a cubic TIP3P water box with 20 Å away from the surface of the protein complex. Finally, an appropriate number of chloride ions were added to neutralize each system.
A prior multistage equilibration strategy was used to remove unfavorable contacts for each system. First, each system was submitted to 5,000 steps of steepest descent MM minimization, followed by 5,000 steps of conjugate gradient MM minimization to the solvent molecules. Then, the side chains were optimized by 5,000 steps of steepest descent and 5,000 steps of conjugate gradient MM minimizations. Thereafter, 5,000 steps of steepest descent and 5,000 steps of conjugate gradient MM minimizations were conducted to the water box. Afterwards, each system was heated to 300 K over a period of 0.2 ns, and then equilibrated over 1.5 ns in the NPT ensemble (T = 300 K and P = 1 bar). Finally, each system was submitted to 500 ns NPT (T = 300 K and P = 1 bar) MD simulations. In all the stages, the temperature was controlled by the Langevin temperature equilibration scheme with a collision frequency of 2.0 ps − 1 and the pressure was controlled by using a Berendsen barostat. 7,8 The particle mesh Ewald (PME) algorithm was applied to handle the long-range electrostatic interactions under periodic S22 boundary condition and a cutoff of 8 Å was used for the van der Waals interactions. 9 The SHAKE method was employed to constrain all covalent bonds of hydrogen atoms. 10 The simulation time step was set as 2 fs and the snapshot was recorded every 10 ps.

SBVS Protocol
The The binding poses of the clustered compounds were manually inspected and filtered.
Finally, 476 potential compounds were purchased from ChemDiv chemical library for in vitro bioassays.

MD Simulations for AR LBD Monomer Bound with M17
The docked structure of the AR LBD bound with M17 was selected as the initial structure for the 1,000 ns long-time MD simulations. The partial charges of M17 were calculated by Gaussian16. 2,3 Then, the antechamber module in Amber 18 package were employed to generate the topology and parameter files of M17. 2,4 The LEaP module in Amber 18 were employed to assign the ff14SB and GAFF2 force fields to the AR LBD and M17, respectively. 5, 6 Other system parameters of preparation, minimization, heating, equilibrium procedures were same as methods of 'Molecular Dynamics (MD) Simulations for Dimers of AR WT , AR W751R and AR F754V '. The system was submitted to three independent 1,000 ns MD simulations and the trajectories from 800-1,000 ns with 1,000 snapshots were applied for binding free energy calculation and per-residue decomposition based on MM/GBSA method as our previous reported. 11,12 Cell Culture   Cell lines of LNCaP, 22RV1, DU145, PC3, C4-2, NIH-3T3, A549, NCI-H1299,  were cultured in a humidified, 5% CO2-containing atmosphere incubator.

AR Transcriptional Activity Assay
The AR transcriptional activities were determined using the constructed LNCaP-ARR2PB-eGFP cell as previously reported. 11,12 The cells of LNCaP-ARR2PB-eGFP were starved with 5% charcoal-stripped serum (CSS) in RPMI-1640 media for 5 days, followed by seeding into a 96-well plate (3.

Competitive Ligand Binding Assay
The binding affinities of M17 and M17-B15 to the LBP of AR LBD were determined using PolarScreen™ Androgen Receptor Competitor Assay Kit (#A15880, Invitrogen Life Technologies) according to the user manual.

Secreted Prostate-Specific Antigen (PSA) Assay
After the AR transcriptional activity assay was completed, 300 μl of the media supernatant including the secreted PSA was measured using IMMULITE ® 2000 XPi Immunoassay System (Siemens Ltd., Erlangen, Germany).

Protein Expression and Purification
The

Structure Modeling of AR LBD Dimer According to the CXMS results
The structure modeling was performed using Xplor-NIH (v.3.3). 14 Crystal structure, PDB code 5V8Q, 15  During the refinement, one AR LBD subunit was fixed while the other AR LBD subunit was grouped as a rigid body and allowed to reorient. The simulation annealing protocol has been described previously. 16,17 In addition to the cross-linking distance restraints (an upper-bound of 20.4 Å between cross-linked Cβ atoms), a weak radius-ofgyration restraint and van der Waals repulsive energy terms are also included. The number of conformers has to be increased to 7 (ensemble size N = 7) to account for all cross-linking restraints. The refinement was repeated 1,200 times with different random seeds. The ensemble structures that satisfy all the restraints and have no steric clashes were selected for further analysis.

Small-Angle X-ray Scattering (SAXS) Analysis
The SAXS data for the free or antagonist-bound AR LBD were collected at the National  18 We then varied the monomer ratio stepwise, from 50% to 95%, in 5% increment, which contribution was subtracted from the experimental value at each angle for the dimer contribution. Subsequently, we used Genetic Algorithm Judging Optimization of Ensembles (GAJOE) 19 to select the dimer conformations that can best satisfy the dimer scattering. The library of dimer conformers (containing 8,300 structures) is obtained from the CXMS-based structure modeling.
Each structure contains 7 conformers that satisfy all CXMS restraints discussed above.
The selection process was repeated 100 times with default parameters in GAJOE to identify the ensemble with the smallest χ 2 value.

Luciferase Reporter Assays for FL-AR F876L and FL-AR F876L/T877A
The full-length the F876L (FL-AR F876L ) and F876L/T877A (FL-AR F876L/T877A ) mutants were constructed on the basis of the human full length wild type AR plasmid pCMV-hAR (#89078, Addgene) using the primer pairs in Table S5. The FRET assay was then tested by Olympus FV1000 equipped with a UPlanSApo 60× /1.35 NA oil objective. CFP was detected using 405 nm excitation while YFP was excited using 515 nm at moderate laser power, and the emission was detected using a 480-495 nm bandpass emission filter and 535-565 nm bandpass emission filter, respectively.
The first pair of CFPbefore and YFPbefore was collected. Then YFP was bleached by scanning

Analysis of Sequences and Structures of Nuclear Receptor Ligand Binding Domains (NR LBDs)
The sequences of 48 NR LBDs were retrieved from UniProt database. Then, these sequences were applied to construct phylogenetic tree using Mega-X (v.10.2.5). 25 The Phylogenetic tree was visualized using iTOL online tool. 26 The structures of NR LBDs were retrieved from Protein Data Bank database. The unreported structures of NR LBDs were homology modeled using the SWISS-MODEL web server. 27 The structures were aligned and the binding site were analyzed using Multiple Sequence Viewer module in Schrö dinger.

Clonogenic Assay
1.5× 10 3 cells/well of LNCaP or 22RV1 were seeded into a 6-well plate. After 1 day incubation, the cells were treated with DMSO, Enz or M17-B15. After 16-18 days incubation, the medium of each well was removed and washed thrice using PBS. Finally, the cell colonies were fixed by methanol, followed by staining with 0.1% crystal violet for 25 min.

Quantitative PCR (qPCR) for PSA, CDC20, CENPF, MKI67
LNCaP cells were cultured in medium containing 5% CSS in six-well plates and treated with Enz and M17-B15 (0.1, 1 μM, in the presence of 5 nM DHT) for 24 h, respectively, where 5 nM DHT and DMSO were separately used in parallel as controls. After 48 h incubation, EZ-10 DNAaway RNA Mini-Preps Kit (Sangon Biotech) and Hifair® Ⅲ 1st Strand cDNA Synthesis SuperMix for qPCR (YEASEN) were employed for mRNA extraction and cDNA preparation, respectively. Then, the PCR SYBR Green Master Mix (YEASEN) was employed to amplify cDNA using the prime pairs in Table S5.

Extraction of Nuclear and Cytoplasmic Fractions
LNCaP cells were cultured with 5% CSS RPMI-1640 in 4 dishes (1 × 10 6 /dish) and Extraction Reagents (#78833, ThermoFisher) were applied to fractionation and extraction of nuclear and cytoplasmic proteins according to manufacturer's protocol.

Xenograft Studies
The animal protocols were in accordance with the guidelines of the Animal Welfare

Statistical Analysis
Graphpad Prism (v.8.4) program was adopted for statistical analysis and data were presented as the mean ± standard error of mean (SEM). The value of p was determined on the basis of at least three independent assays using Student's t-test, one-way ANOVA or two-way ANOVA analysis.