Targeting SWI/SNF ATPases in enhancer-addicted prostate cancer – Nature

Cell lines, antibodies, and compounds

Most cell lines were originally obtained from ATCC, DSMZ, ECACC, or internal stock. C4-2B cells were provided by E. Keller (University of Michigan). CWR-R1 cells and a series of enzalutamide-resistant prostate cancer cell lines (LNCaP_Parental, LNCaP_EnzR, CWR-R1_Parental, CWR-R1_EnzR, VCaP_Parental and VCaP_EnzR) were provided by D. Vander Griend (University of Illinois at Chicago)³¹. Bin-67 was generously provided by B. Vanderhyden (Ottawa Hospital Research Institute). All cells were genotyped to confirm their identity at the University of Michigan Sequencing Core and tested routinely for Mycoplasma contamination. LNCaP, 22RV-1, CWR-R1, PC-3, and DU145 were grown in Gibco RPMI-1640 + 10% FBS (ThermoFisher). VCaP was grown in Gibco DMEM + 10% FBS (ThermoFisher). BIN-67 cell lines were grown in custom media (20% fetal bovine serum (FBS), 40% Dulbecco’s modified Eagle’s medium, 40% Dulbecco’s modified Eagle’s medium/Ham’s F12). Sources of all antibodies are described in Supplementary Table 2. AU-15330 was synthesized by Aurigene (see Supplementary Text), dBRD9 and VZ 185 were purchased from Tocris Bioscience, and enzalutamide was purchased from Selleck Chemicals.

Computational modelling of AU-15330 – SMARCA2-BD binding

The binding model of AU-15330 in complex with SMARCA2-BD and VHL was generated using Aurigene’s proprietary computing algorithm ALMOND (algorithm for modeling neosubstrate degraders). The algorithm is developed using the ICM-Pro integrated modelling platform (http://www.molsoft.com/icm_pro.html) and trained to predict models of ternary complexes of bi-functional molecules with very short or no linkers. The process employs protein-protein docking simulation, exhaustive conformational sampling, small molecule-protein docking, and site-directed scoring of predicted ternary complex models. The computed score estimates the force of induced interactions in the predicted target–E3 ligase complex and is used as a basis for prioritization of degrader binding models. The images were prepared using PyMOL (https://www.schrodinger.com/products/pymol).

Cell viability assay

Cells were plated onto 96-well plates in their respective culture medium and incubated at 37 °C in an atmosphere of 5% CO₂. After overnight incubation, a serial dilution of compounds was prepared and added to the plate. The cells were further incubated for 5 days, and the CellTiter-Glo assay (Promega) was then performed according to the manufacturer’s instruction to determine cell proliferation. The luminescence signal from each well was acquired using the Infinite M1000 Pro plate reader (Tecan), and the data were analyzed using GraphPad Prism software (GraphPad Software).

Incucyte proliferation assays/Caspase-3/7 green apoptosis assay

A total of 4,000 cells per well were seeded in clear 96-well plates. After overnight incubation, compounds were added to the cells at logarithmic dose series. One day and 8 days after seeding, cellular ATP content was measured using CellTiterGlo (Promega). Measurements after 8 days were divided by the measurement after 1 day (that is, the T0 plate) to derive fold proliferation. For online analysis of cell growth, 4,000 cells per well were seeded in clear 96-well plates (Costar no. 3513). IncuCyte Caspase-3/7 Green Apoptosis Assay Reagent (1:1,000, Essen BioSciences no. 4440) was added, and cells were incubated at 37 °C and 5% CO₂ overnight. On the next day, compounds were added at the desired concentration using the HP digital dispenser D300, and plates were read in an Incucyte ZOOM. Every 2h, phase object confluence (percentage area) for proliferation and green object count for apoptosis were measured. Values for apoptosis were normalized for the total number of cells.

Western blot and immunoprecipitation

Cell lysates were prepared in RIPA buffers (ThermoFisher Scientific) supplemented with cOmpleteTM protease inhibitor cocktail tablets (Sigma-Aldrich), and total protein was measured by Pierce BCA Protein Assay Kit (ThermoFisher Scientific). An equal amount of protein was resolved in NuPAGE 3 to 8%, Tris-Acetate Protein Gel (ThermoFisher Scientific) or NuPAGE 4 to 12%, Bis-Tris Protein Gel (ThermoFisher Scientific) and blotted with primary antibodies. Following incubation with HRP-conjugated secondary antibodies, membranes were imaged on an Odyssey CLx Imager (LiCOR Biosciences). Immunoprecipitations were performed in LNCaP and VCaP cells treated as described. 600 μg of nuclear extracts isolated using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (ThermoFisher Scientific) were immunoprecipitated with SMARCC1, AR, FOXA1, or ERG antibodies according to the manufacturer’s protocol. Eluted proteins were subjected to western blot or mass spectrometry analysis. For all immunoblots, uncropped and unprocessed images are provided in Supplementary Figure 1.

RNA isolation and quantitative real-time PCR

Total RNA was isolated from cells using the Direct-zol kit (Zymo), and cDNA was synthesized from 1,000 ng total RNA using Maxima First Strand cDNA Synthesis Kit for PCR with reverse transcription (RT–PCR) (Thermo Fisher Scientific). Quantitative real-time PCR (qPCR) was performed in triplicate using standard SYBR green reagents and protocols on a QuantStudio 5 Real-Time PCR system (Applied Biosystems). The target mRNA expression was quantified using the ΔΔCt method and normalized to ACTB expression. All primers were designed using Primer 3 (http://frodo.wi.mit.edu/primer3/) and synthesized by Integrated DNA Technologies. Primer sequences are listed in Supplementary Table 2.

CRISPR knock-out and inducible shRNA knockdown

Guide RNAs (sgRNAs) targeting the exons of human SMARCA2/BRM or SMARCA4/BRG1 were designed using Benchling (https://www.benchling.com/). Non-targeting sgRNA, SMARCA2/BRM or SMARCA4/BRG1-targeting sgRNAs were cloned into lentiCRISPR v2 plasmid according to published literature³²; lentiCRISPR v2 plasmid was a gift from F. Zhang (Addgene plasmid #52961). LNCaP cells were transiently transfected with lentiCRISPR v2 encoding non-targeting or pool of three independent SMARCA2/BRM or SMARCA4/BRG1-targeting sgRNAs. Twenty-four hours after transfection, cells were selected with 1 μg ml⁻¹ puromycin for three days. Western blot was performed to examine the knock-out efficiency. The sgRNA sequences are listed in Supplementary Table 2.

ATAC-seq and analysis

ATAC-seq was performed as previously described³³. In brief, 50,000 cancer cells treated with AU-15330 or ZBC-260[30] were washed in cold PBS and resuspended in cytoplasmic lysis buffer (CER-I from the NE-PER kit, Invitrogen, cat. no. 78833). This single-cell suspension was incubated on ice for 5–8 min (depending on the cell line) with gentle mixing by pipetting every 2 min. The lysate was centrifuged at 1,300g for 5 min at 4 °C. Nuclei were resuspended in 50 μl of 1× TD buffer, then incubated with 2–2.5 μl Tn5 enzyme for 30 min at 37 °C (Nextera DNA Library Preparation Kit; cat. no. FC-121-1031). Samples were immediately purified by Qiagen minElute column and PCR-amplified with the NEBNext High-Fidelity 2X PCR Master Mix (NEB; cat. no. M0541L) following the original protocol³³. qPCR was used to determine the optimal PCR cycles to prevent over-amplification. The amplified library was further purified by Qiagen minElute column and SPRI beads (Beckman Coulter; cat. no. A63881). ATAC-seq libraries were sequenced on the Illumina HiSeq 2500 (125-nucleotide read length, paired end).

Paired-end .fastq files were trimmed and uniquely aligned to the GRCh38/hg38 human genome assembly using Novoalign (Novocraft) (with the parameters -r None -k -q 13 -k -t 60 -o sam –a CTGTCTCTTATACACATCT), and converted to .bam files using SAMtools (version 1.3.1). Reads mapped to mitochondrial or duplicated reads were removed by SAMtools and PICARD MarkDuplicates (version 2.9.0), respectively. Filtered .bam files from replicates were merged for downstream analysis. MACS2 (2.1.1.20160309) was used to call ATAC-seq peaks. The coverage tracks were generated using the program bam2wig (http://search.cpan.org/dist/Bio-ToolBox/) with the following parameters: –pe–rpm–span–bw. Bigwig files were then visualized using the IGV (Broad Institute) open-source genome browser, and the final figures were assembled using Adobe Illustrator.

De novo and known motif enrichment analysis

All de novo and known motif enrichment analyses were performed using the HOMER (v.4.10) suite of algorithms43. Peaks were called by the findPeaks function (-style factor -o auto) at 0.1% false discovery rate; de novo motif discovery and enrichment analysis of known motifs were performed with findMotifsGenome.pl (–size given–mask). The top 10 motifs from the results are shown, and motifs were generally ascribed to the protein family instead of specific family members (unless known).

RNA-seq and analysis

RNA-seq libraries were prepared using 200–1,000 ng of total RNA. PolyA+ RNA isolation, cDNA synthesis, end-repair, A-base addition, and ligation of the Illumina indexed adapters were performed according to the TruSeq RNA protocol (Illumina). Libraries were size selected for 250–300 bp cDNA fragments on a 3% Nusieve 3:1 (Lonza) gel, recovered using QIAEX II reagents (QIAGEN), and PCR amplified using Phusion DNA polymerase (New England Biolabs). Library quality was measured on an Agilent 2100 Bioanalyzer for product size and concentration. Paired-end libraries were sequenced with the Illumina HiSeq 2500, (2 × 100 nucleotide read length) with sequence coverage to 15–20M paired reads.

Libraries passing quality control were trimmed of sequencing adaptors and aligned to the human reference genome, GRCh38. Samples were demultiplexed into paired-end reads using Illumina’s bcl2fastq conversion software v2.20. The reference genome was indexed using bowtie2-build, and reads were aligned onto the GRCh38/hg38 human reference genome using TopHat2³⁴ with strand-specificity and allowing only for the best match for each read. The aligned file was used to calculate strand-specific read count for each gene using HTSeq-count (version 0.13.5)³⁵. EdgeR (version 3.34.1)³⁶ was used to compute differential gene expression using raw read-counts as input. Heatmaps were generated using the ComplexHeatmap³⁷ package in R. For gene enrichment analysis (GSEA), we first defined ERG and FOXA1 gene signatures from VCaP or LNCaP cells treated with control siRNA or siRNA targeting ERG³⁸ or FOXA1 (generated in this study) containing 250 significantly downregulated genes. For AR and MYC, the Hallmark gene signatures were used. These gene signatures were used to perform a fast pre-ranked GSEA using fgsea bioconductor package³⁹ in R. We used the function fgsea to estimate the net enrichment score and p-value of each pathway, and the plotEnrichment function was used to plot enrichment for the pathways of interest.

ChIP–seq and data analysis

Chromatin immunoprecipitation experiments were carried out using the HighCell# ChIP-Protein G kit (Diagenode) as per the manufacturer’s protocol. Chromatin from 5 × 10⁶ cells was used for each ChIP reaction with 10 μg of the target protein antibody. In brief, cells were trypsinized and washed twice with 1× PBS, followed by cross-linking for 8 min in 1% formaldehyde solution. Crosslinking was terminated by the addition of 1/10 volume 1.25 M glycine for 5 min at room temperature followed by cell lysis and sonication (Bioruptor, Diagenode), resulting in an average chromatin fragment size of 200 bp. Fragmented chromatin was then used for immunoprecipitation using various antibodies, with overnight incubation at 4 °C. ChIP DNA was de-crosslinked and purified using the iPure Kit V2 (Diagenode) using the standard protocol. Purified DNA was then prepared for sequencing as per the manufacturer’s instructions (Illumina). ChIP samples (1–10 ng) were converted to blunt-ended fragments using T4 DNA polymerase, Escherichia coli DNA polymerase I large fragment (Klenow polymerase), and T4 polynucleotide kinase (New England BioLabs (NEB)). A single adenine base was added to fragment ends by Klenow fragment (3′ to 5′ exo minus; NEB), followed by ligation of Illumina adaptors (Quick ligase, NEB). The adaptor-ligated DNA fragments were enriched by PCR using the Illumina Barcode primers and Phusion DNA polymerase (NEB). PCR products were size-selected using 3% NuSieve agarose gels (Lonza) followed by gel extraction using QIAEX II reagents (Qiagen). Libraries were quantified and quality checked using the Bioanalyzer 2100 (Agilent) and sequenced on the Illumina HiSeq 2500 Sequencer (125-nucleotide read length).

Paired-end, 125 bp reads were trimmed and aligned to the human reference genome (GRC h38/hg38) with the Burrows-Wheeler Aligner (BWA; version 0.7.17-r1198-dirty)⁴⁰. The SAM file obtained after alignment was converted into BAM format using SAMTools (version 1.9). MACS2 (version 2.1.1.20160309) callpeak was used for performing peak calling with the following option: ‘macs2 callpeak–call-summits–verbose 3 -g hs -f BAM -n OUT–qvalue 0.05’. For H3K27ac data, the broad option was used. Using deepTools (version 3.3.1) bamCoverage, a coverage file (bigWig format) for each sample was created. The coverage was calculated as the number of reads per bin, where bins are short consecutive counting windows. While creating the coverage file, the data was normalized with respect to each library size. ChIP peak profile plots and read-density heat maps were generated using deepTools, and cistrome overlap analyses were carried out using the ChIPpeakAnno (version 3.0.0) or ChIPseeker (version 1.29.1) packages in R (version 3.6.0).

HiChIP library preparation and data analysis

HiChIP assay was performed on 5×10⁶ DMSO or AU-15330 treated VCaP cells. Frozen cells were resuspended in 1× PBS and crosslinked with 3 mM DSG and 1% formaldehyde. Washed cells were digested with 0.5 µl MNase in 100 μl of nuclease digest buffer with MgCl₂. Cells were lysed with 1× RIPA, and clarified lysate (approximately 1,400 ng) was used for ChIP. The antibody amount used per ChIP and vendor information are as follows: CTCF: 1.14 μg of Cell Signaling cat. no. 3418; H3K4me3: 3.4 μg of Cell Signaling cat. no. 9751; H3K27ac: 0.4 μg of Cell Signaling cat. no. 8173. The Protein A/G bead pulldown, proximity ligation, and libraries were prepared as described in the Dovetail protocol (Dovetail HiChIP MNase Kit). Libraries were sequenced on an Illumina HiSeq 4000.

Raw fastq files were aligned using BWA mem (version 0.7.17-r1198-dirty) with the −5SP options with an index containing only the main chromosome from the human genome release hg38 (available from the UCSC genome). The aligned paired reads were annotated with pairtools (version 0.3.0) parse (https://github.com/open2c/pairtools) with the following options–min-mapq 40–walks-policy 5unique–max-inter-align-gap 30 and the–chroms-path file corresponding to the size of the chromosome used for the alignment index. The paired reads were further processed to remove duplicated reads, sorted with unaligned reads removed with the pairtools sort and the pairtools dedup tools with the basic option to produce an alignment file in the bam format as well as the location of the valid pair. The valid pairs were finally converted to the .cool and .mcool format using the cooler cload and cooler zoomify tools (version 0.8.11)⁴¹ and to the .hic format using the juicer tool (version 1.22.01)⁴².

For the generation of the aggregate peak analyses (APA) plots, we used the HiCExplorer tools (version 3.7) and the hicAggregateContacts command with–range 50000:100000–numberOfBins 30. Plots for all chromosomes were individually computed and summated to generate the global APA plots. The ComplexHeatmaps package³⁷ in R was used for the generation of the final heatmap. For the Hi-ChIP contact heatmap, .hic files were uploaded to the WashU Epigenome Browser (https://epigenomegateway.wustl.edu/), and screenshots from gene loci of interest were downloaded using the default viewing conditions.

Super-enhancer analysis

Super-enhancer regions were identified with findPeaks function from HOMER (version v.4.10)⁴³ using options “-style super -o auto”. In addition, the option “-superSlope −1000” was added to include all potential peaks, which were used to generate the super-enhancer plot (super-enhancer score versus ranked peaks). The slope value of greater than or equal to 1 was used to identify super-enhancer clusters. The input files to findPeaks were tag directories generated from alignment files in SAM format with makeTagDirectory function from HOMER.

AU-15330 and enzalutamide formula for in vivo studies

AU-15330 was added in 40% of 2-hydroxypropyl-β-cyclodextrin (HPβCD) and sonicated until completely dissolved, and then the solution was further mixed with 5% dextrose in water (D5W) to reach a final concentration of 10% HPβCD. AU-15330 was freshly prepared right before administration to mice. AU-15330 was delivered to mice by intravenous injection either through the tail vein or retro-orbital injection unless otherwise indicated. Enzalutamide was added in 1% carboxymethyl cellulose (CMC) with 0.25% Tween-80 and sonicated until homogenized. Enzalutamide was delivered to mice by oral gavage.

Human prostate tumour xenograft models

Six-week-old male CB17 severe combined immunodeficiency (SCID) mice were procured from the University of Michigan breeding colony. Subcutaneous tumours were established at both sides of the dorsal flank of mice. Tumours were measured at least biweekly using digital calipers following the formula (π/6) (L × W²), where L is length and W is width of the tumour. At the end of the studies, mice were killed and tumours extracted and weighed. The University of Michigan Institutional Animal Care and Use Committee (IACUC) approved all in vivo studies.

For the VCaP non-castrated tumour model, 3 × 10⁶ VCaP cells were injected subcutaneously into the dorsal flank on both sides of the mice in a serum-free medium with 50% Matrigel (BD Biosciences). Once tumours reached a palpable stage (~200 mm³), mice were randomized and treated with either 10, 30 mg kg⁻¹ AU-15330, or vehicle through intravenous injection 5 days per week for 3 weeks.

For the VCaP castration-resistant tumour model, 3 × 10⁶ VCaP cells were injected subcutaneously into the dorsal flank on both sides of the mice in a serum-free medium with 50% Matrigel (BD Biosciences). Once tumours reached a palpable stage (~200 mm³), tumour-bearing mice were castrated. Once tumours grew back to the pre-castration size, mice were randomized and treated with either 60 mg kg⁻¹ AU-15330 or vehicle by intranvenous injection 3 days per week, and with or without 10 mg kg⁻¹ enzalutamide by oral gavage 5 days per week for 5 weeks.

For the C4-2B non-castrated tumour model, 1 × 10⁶ cells were injected subcutaneously into the dorsal flank on both sides of the mice in a serum-free medium with 50% Matrigel (BD Biosciences). Once tumours reached a palpable stage (~100 mm³), mice were randomized and treated with either 60 mg kg⁻¹ AU-15330 or vehicle by intravenous injection 3 days per week, and with or without 30 mg kg⁻¹ enzalutamide by oral gavage 5 days per week for 4 weeks. Following the IACUC guidelines, in all treatment arms the maximal tumour size did not exceed the 2.0 cm limit in any dimension and animals with xenografts reaching that size were duly euthanized. The raw tumour volumes and/or weights from all animal efficacy studies are included in the Source Data files.

Prostate patient-derived xenograft models

The University of Texas M. D. Anderson Cancer Center PDX series has been previously described⁴⁴. PDXs were derived from men with CRPC undergoing cystoprostatectomy using described protocols. MDA-PCa-146-12 was derived from a CRPC patient diagnosed with Gleason 5+4=9 prostate adenocarcinoma. MDA-PCa-146-12 was derived from a specimen obtained from the left bladder wall and demonstrated conventional adenocarcinoma (AR⁺). PDXs were maintained in male SCID mice by surgically implanting 2 mm³ tumours coated with 100% Matrigel to both flanks of mice. Once tumours reached ~200 mm³ in size, mice were randomized and divided into different treatment groups receiving either 60 mg kg⁻¹ AU-15330 or vehicle by subcutaneous injection 3 days per week, and with or without 10 mg/kg enzalutamide by oral gavage 5 days per week for 3 weeks. For castration-resistant MDA-PCa-146-12, tumours were established on castrated male SCID mice. Once tumours reached ~100 mm³, mice were randomized and divided into different treatment groups receiving either 60 mg kg⁻¹ AU-15330 or vehicle by intravenous injection 3 days per week, and with or without 30 mg kg⁻¹ enzalutamide by oral gavage 5 days per week for 6 weeks. Following the IACUC guidelines, in all treatment arms the maximal tumour size did not exceed the 2.0 cm limit in any dimension and animals with xenografts reaching that size were duly euthanized. The raw tumour volumes and/or weights from all animal efficacy studies are included in the Source Data files.

Histopathological analysis of organs harvested for drug toxicity

For the present study, organs (liver, spleen, kidney, colon, small intestine, prostate, and testis) were harvested and fixed in 10% neutral buffered formalin followed by embedding in paraffin to make tissue blocks. These blocks were sectioned at 4 µm and stained with Harris haematoxylin and alcoholic eosin-Y stain (both reagents from Leica Surgipath) and staining was performed on Leica autostainer-XL (automatic) platform. The stained sections were evaluated by two different pathologists using a brightfield microscope in a blinded fashion between the control and treatment groups for general tissue morphology and coherence of architecture. A detailed comprehensive analysis of the changes noted at the cellular and sub-cellular level were performed as described below for each specific tissue.

Evaluation of liver

Liver tissue sections were evaluated for normal architecture, and regional analysis for all three zones was performed for inflammation, necrosis, and fibrosis.

Evaluation of spleen

Splenic tissue sections were evaluated for the organization of hematogenous red and lymphoid white pulp regions including necrosis and fibrotic changes if any.

Evaluation of kidney

Kidney tissue sections were examined for changes noted if any in all the four renal functional components, namely glomeruli, interstitium, tubules, and vessels.

Evaluation of colon

Colonic tissue sections were examined for mucosal (epithelium and lamina propria), sub-mucosal, and seromuscular layer changes including crypt changes, goblet cells, inflammatory infiltrate granulation tissue, and mucosal ulceration. A detailed goblet cell evaluation was also performed utilizing Alcian blue staining wherein goblet cells and epithelial cells were counted in ten colonic crypt epithelia in each experimental animal of the various subgroups. Summation of all the goblet and epithelial cells was done, and a ratio of goblet cell to epithelial cell (GC ratio) was calculated per sample.

Evaluation of small intestine

Small intestine tissue sections were examined for mucosal changes such as villous blunting, villous: crypt ratio, and evaluated for inflammatory changes including intraepithelial lymphocytes, extent (mucosal, sub-mucosal, serosal), and type of inflammatory infiltrate including tissue modulatory effect.

Evaluation of prostate

Prostate tissue sections were evaluated to note for any epithelial abnormality and stromal changes identified in all four lobes (dorsal, anterior, lateral, and ventral). Additionally, any overt inflammatory infiltrate was also examined.

Evaluation of testis

Testicular tissues were examined for the architectural assessment of seminiferous tubules (orderly maturation of germinal epithelial cells devoid of maturation arrest and Sertoli cell prominence), Leydig cells, and interstitial reaction. For an in-depth comprehensive analysis to comment upon the spermatogenesis in a semi-quantitative method, a testicular biopsy score count (Johnsen score) in 100 orderly cross-sections of seminiferous tubules in each animal of all the subgroups at 20× magnification was performed. Each of the 100 seminiferous tubules assessed was given a score (score range: 0–10), and the average score was calculated (total sum of score/100).

Alcian blue staining

Alcian blue staining was performed as per the manufacturer’s protocol (Alcian Blue Stain Kit (pH 2.5) cat. no. ab150662). Following an overnight incubation of tissue sections at 58 °C, slides were deparaffinized in xylene followed by hydration in ethanol (100%, 70%) and water for 5 min each. Slides were then incubated in acetic acid solution for 3 min followed by a 30 min incubation at room temperature in Alcian blue stain (pH 2.5). Excess Alcian blue was removed by rinsing slides in acetic acid solution for 1 min, and three water washes for 2 min each. Nuclear Fast Red solution was used as a counterstain for 5 min. Slides were subsequently washed in running tap water, dehydrated in ethanol, xylene, and mounted using EcoMount (Thermo Fisher, cat. no. EM897L).

Immunohistochemistry

Immunohistochemistry was performed on formalin-fixed paraffin-embedded 4μm sections of mouse or xenograft tissues. Slides with tissue sections were incubated at 58 °C overnight and the next day were deparaffinized in xylene, followed by serial hydration steps in ethanol (100%, 70%) and water for 5 min each. Endogenous tissue peroxidase activity was blocked by placing slides in 3% H₂0₂-methanol solution for 1 h at room temperature. Antigen retrieval was performed by microwaving slides in a solution of citrate buffer (pH 6) for 15 min, followed by blocking in 2.5% normal horse serum (Vector Laboratories, cat. no. S-2012-50) for 2 h. The slides were then incubated in the following primary antibodies overnight at 4 °C: BRG1 (Abcam cat. no. 108318, 1:100), AR (Millipore cat. no. 06-680, 1:2,000), BRM1 (Millipore Sigma cat. no. HPA029981, 1:100), FOXA1 (Thermo Fisher Scientific cat. no. PA5-27157, 1:1,000), ERG (Cell Signaling Technology cat. no. 97249S, 1:500). ImmPRESS-HRP conjugated anti-mouse–anti-rabbit cocktail from Vector Laboratories (cat. no. MP-7500-50) was used as secondary antibodies (room temperature, 1 h). Visualization of staining was done per the manufacturer’s protocol (Vector Laboratories, cat. no. SK-4100). Following DAB staining, slides were dehydrated in ethanol, xylene (5 min each), and mounted using EcoMount (Thermo Fisher, cat. no. EM897L).

TMT mass spectrometry

VCaP cells were seeded at 5 × 10⁶ cells on a 100 mm plate 24 h before treatment. Cells were treated in triplicate by the addition of test compounds. After 4 h, the cells were harvested and processed by using EasyPep Mini MS Sample Prep Kit (Thermo Fisher, A40006). Samples were quantified using a micro BCA protein assay kit (Thermo Fisher Scientific) and cell lysates were proteolyzed and labelled with TMT 10-plex Isobaric Label Reagent (Thermo Fisher Scientific, 90110) essentially following the manufacturer’s protocol. Briefly, upon reduction and alkylation of cysteines, the proteins were precipitated by adding 6 volumes of ice-cold acetone followed by overnight incubation at 20 °C. The precipitate was spun down, and the pellet was allowed to air dry. The pellet was resuspended in 0.1M TEAB and digested overnight with trypsin (1:50 enzyme:protein) at 37 °C with constant mixing using a thermomixer. The TMT 10-plex reagents were dissolved in 41 ml of anhydrous acetonitrile, and labelling was performed by transferring the entire digest to the TMT reagent vial and incubating it at room temperature for 1 h. The reaction was quenched by adding 8 ml of 5% hydroxylamine and a further 15 min incubation. Labelled samples were mixed together and dried using a vacufuge. An offline fractionation of the combined sample (200 mg) into 10 fractions was performed using high pH reversed-phase peptide fractionation kit according to the manufacturer’s protocol (Pierce, 84868). Fractions were dried and reconstituted in 12 ml of 0.1% formic acid/2% acetonitrile in preparation for LC–MS/MS analysis.

To obtain superior accuracy in quantitation, we employed multinotch-MS3⁴⁵ which minimizes the reporter ion ratio distortion resulting from fragmentation of co-isolated peptides during MS analysis. Orbitrap Fusion (Thermo Fisher Scientific) and RSLC Ultimate 3000 nano-UPLC (Dionex) was used to acquire the data. The sample (2 ml) was resolved on a PepMap RSLC C18 column (75 mm i.d. × 50 cm; Thermo Scientific) at the flowrate of 300 nl min⁻¹ using 0.1% formic acid/acetonitrile gradient system (2–22% acetonitrile in 150 min; 22–32% acetonitrile in 40 min; 20 min wash at 90% followed by 50 min re-equilibration) and direct spray into the mass spectrometer using EasySpray source (Thermo Fisher Scientific). The mass spectrometer was set to collect one MS1 scan (Orbitrap; 60K resolution; AGC target 2 × 10⁵; max IT 100 ms) followed by data-dependent, ‘‘Top Speed’’ (3 s) MS2 scans (collision-induced dissociation; ion trap; NCD 35; AGC 5 × 10³; max IT 100 ms). For multinotch-MS3, top 10 precursors from each MS2 were fragmented by HCD followed by Orbitrap analysis (NCE 55; 60K resolution; AGC 5 × 10⁴; max IT 120 ms, 100-500 m/z scan range). Proteome Discoverer (v2.1; Thermo Fisher) was used for data analysis. MS2 spectra were searched against SwissProt human protein database (release 11 November 2015; 42,084 sequences) using the following search parameters: MS1 and MS2 tolerances were set to 10 ppm and 0.6 Da, respectively; carbamidomethylation of cysteines (57.02146 Da) and TMT labelling of lysine and N-termini of peptides (229.16293 Da) were considered static modifications; oxidation of methionine (15.9949 Da) and deamidation of asparagine and glutamine (0.98401 Da) were considered variable. Identified proteins and peptides were filtered to retain only those that passed FDR threshold. Quantitation was performed using high-quality MS3 spectra (Average signal-to-noise ratio of 20 and <30% isolation interference).

Meta-analyses of protein interactomes

Interactome proteomics data of AR and ERG was downloaded from published literature^38,46. The FOXA1 nuclear co-immunoprecipitation/mass spectrometry experiment was performed in this study as described above. The protein interactomes of AR, ERG, and FOXA1 were ranked based on FDR at the top 10%, and the intersection was taken from these three independent studies.

Assessment of drug synergism

To determine the presence of synergy between two drug treatments, cells were treated with increasing concentrations of either drug for 120 h, followed by the determination of viable cells using the CellTiter-Glo Luminescent Cell Viability Assay (Promega). The experiment was carried out in four biological replicates. The data were expressed as percentage inhibition relative to baseline, and the presence of synergy was determined by the Bliss method using the synergy finder R package⁴⁷.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.