Determining direct binders of the Androgen Receptor using a high-throughput Cellular Thermal Shift Assay

Androgen Receptor (AR) is a key driver in prostate cancer. Direct targeting of AR has valuable therapeutic potential. However, the lack of disease relevant cellular methodologies capable of discriminating between inhibitors that directly bind AR and those that instead act on AR co-regulators has made identification of novel antagonists challenging. The Cellular Thermal Shift Assay (CETSA) is a technology enabling confirmation of direct target engagement with label-free, endogenous protein in living cells. We report the development of the first high-throughput CETSA assay (CETSA HT) to identify direct AR binders in a prostate cancer cell line endogenously expressing AR. Using this approach, we screened a pharmacology library containing both compounds reported to directly engage AR, and compounds expected to target AR co-regulators. Our results show that CETSA HT exclusively identifies direct AR binders, differentiating them from co-regulator inhibitors where other cellular assays measuring functional responses cannot. Using this CETSA HT approach we can derive apparent binding affinities for a range of AR antagonists, which represent an intracellular measure of antagonist-receptor Ki performed for the first time in a label-free, disease-relevant context. These results highlight the potential of CETSA HT to improve the success rates for novel therapeutic interventions directly targeting AR.


Mass spectrometry measurement of intracellular Enzalutamide
Intracellular Enzalutamide concentrations were determined by mass spectrometry 1-3 .
CWR22Pc-R1-AD1 cells were treated with a range of Enzalutamide doses for 2 h. Cells were harvested and washed 3x in PBS to remove extracellular compound before resuspension in 300 µL PBS and lysis by 3 cycles of 20 second sonication followed by 20 second recovery on ice. Cell lysates were stored at -80˚C prior to analysis by LCMS utilising a Waters Xevo TSQ (WAA697) and an Acquity UPLC system from Waters consisting of sample manager (L12USM631G), Acquity PDA (K12UPD606A), Column Manager (K12CMP412G) and Binary Solvent Manager (L12BUR860M). The Waters Xevo was operated in positive ion Electrospray (ESI) mode to detect the parent ion with m/z 465.1. Alongside samples a calibration curve was analysed spiked with 10, 7.5, 5, 2.5, 1, 0.5, 0.1 µM Enzalutamide in PBS. Calibration curve chromatograms were extracted, smoothed, integrated and fitted with a linear regression to give a standard curve to enable calculation of Enzalutamide concentration in test samples.
Across all experiments r2 was >0.990 and the mean error of the QC samples of 17%.

Generation of a CETSA HT assay by AlphaScreen ®
To generate a high-throughput compatible CETSA assay, thermostable (soluble) AR was quantified using an AlphaScreen ® endpoint, negating the need for physical seperation of the soluble thermostable and insoluble thermally aggregated AR populations, as previously described 4 . Thermostable AR was quantified using AlphaScreen ® technology (PerkinElmer) whereby close proximity of a donor bead and an acceptor bead allow for transfer of a singlet S3 oxygen from the donor to the acceptor, exciting a fluorophore to emit light at 520-620 nm 5 .
Two primary anti-AR antibodies derived from mouse and rabbit were applied to bind AR simultaneously at separate epitopes. Secondary AlphaScreen ® antibodies were combined with the primary antibodies in the form of AlphaScreen ® donor beads functionalised with antibodies recognising mouse antibody IgG (PerkinElmer AS104D), and AlphaScreen ® acceptor beads functionalised with antibodies recognising rabbit antibody IgG (PerkinElmer AL104C). Productive AlphaScreen ® signal will only be observed in the presence of all of the following: correctly folded, soluble AR; mouse-derived anti-AR antibody; rabbit-derived anti-AR antibody; anti-Mouse AlphaScreen ® donor; anti-Rabbit AlphaScreen ® acceptor.

AR functional RT-qPCR assay
Compound was dosed into a 384w plate (Corning) and CWR22Pc-R1-AD1 cells, supplemented with 1 nM DHT, were seeded at 1.5 x 10 4 cells/well in 25 µL volume. Phenol red free RPMI1640 media supplemented with 10% charcoal-stripped serum was used throughout. Following 48 h incubation, plates were imaged using a 10x optics with an IncuCyte ZOOM (Essen BioScience) to calculate cellular confluency. Media was removed using an ELX405 CW washer (BioTek) and 12 µL RealTime ready Cell Lysis Kit supplemented with RNAse inhibitor (Roche) was added per well. Cells were lysed by 8 minute incubation with repetitive 2 min shakes at 100 rpm followed by 2 min clarification at 300xg. RealTime ready RNA Virus Master Mix (Roche) was prepared following manufacturer's instructions, supplemented with primers and probes for FKBP5 (FAM, IDT Ref Seq # NM_001145776(3)) and β-Actin (VIC, IDT Ref Seq # NM_001101(1)). 4.5 µL/ well was added to a 384w PCR plate (4titude). Following mixing, 0.5 µL of lysate was transferred to the PCR plate using a Bravo liquid handling system (Agilent). PCR plates were sealed and analysed using a LightCycer ® II 480 (Roche). Reverse transription was performed for 8 min at 50˚C before 30 sec denaturation at 95˚C. 40 PCR cycles were performed consisting of 30 sec at 95˚C, 20 sec at 60˚C, fluorescence acquisition, and 1 sec at 72˚C. Calculated Cq values were exported for analysis.    Figure S5). Compound target engagement with AR was measured in CWR22Pc-R1-AD1 cells by CETSA HT as; i) the ability of a compound to compete out a fixed dose (1 nM) of DHT, and thus to induce thermal destabilisation at 46°C (black circles) as in Figure 2c; ii) The ability of a compound to influence thermal stability of AR at 46˚C in the absence of exogenous agonist (grey triangles) as in Figure 2b. Compound treatments were 2 h, data from one technical repeat. Selected compounds were also analysed in the absence of a heat shock to differentiate changes in AR thermal stability from changes in total cellular AR (Supplementary Figure S6).

S14
Supplementary Figure S5: Additional data for AR-driven transcriptional responses. Selected compounds from the pharmacology training set were analysed for reduction in mRNA levels of an AR-responsive transcript FKBP5 within CWR22-Pc-R1-AD1 cells, measured as an increase in Cq value by RT-qPCR and multiplexed with a β-Actin housekeeping gene. A measure of cell viability and proliferation of the same cells was performed by quantifying confluency using an IncuCyte ZOOM (plates imaged immediately before RT-qPCR analysis).
Compound treatment was 48 h, data from one technical repeat.

S15
Supplementary Figure S6: CETSA HT controls in the absence of heat-shock to differentiate AR target engagement from AR degradation. Selected compounds were tested in the CETSA HT assay in the absence of exogenous DHT and without the 46˚C heat shock. In this format the assay essentially measures total AR following 2 h compound incubation. Data from one technical repeat, N.D. = not determined. AR antagonists Enzalutamide, Hydroxyflutamide and MK-2866 thermally destabilise AR as a consequence of target engagement (Supplementary Table S1), but do not reduce total AR. Niclosamide reduces total AR, as do the Hsp90 inhibitors Onalespib, Tanespimycin and NVP-AUY-922. The Menin inhibitor MI-503, the TrkA kinase inhibitor Entrectinib, the BRD4 inhibitor JQ1, the p23 inhibitor Ailanthone and the AR degrader ACS-J9 do not reduce total AR within 2 hours.