Targeting SWI/SNF ATPases in enhancer-addicted prostate cancer

The switch/sucrose non-fermentable (SWI/SNF) complex has a crucial role in chromatin remodelling1 and is altered in over 20% of cancers2,3. Here we developed a proteolysis-targeting chimera (PROTAC) degrader of the SWI/SNF ATPase subunits, SMARCA2 and SMARCA4, called AU-15330. Androgen receptor (AR)+ forkhead box A1 (FOXA1)+ prostate cancer cells are exquisitely sensitive to dual SMARCA2 and SMARCA4 degradation relative to normal and other cancer cell lines. SWI/SNF ATPase degradation rapidly compacts cis-regulatory elements bound by transcription factors that drive prostate cancer cell proliferation, namely AR, FOXA1, ERG and MYC, which dislodges them from chromatin, disables their core enhancer circuitry, and abolishes the downstream oncogenic gene programs. SWI/SNF ATPase degradation also disrupts super-enhancer and promoter looping interactions that wire supra-physiologic expression of the AR, FOXA1 and MYC oncogenes themselves. AU-15330 induces potent inhibition of tumour growth in xenograft models of prostate cancer and synergizes with the AR antagonist enzalutamide, even inducing disease remission in castration-resistant prostate cancer (CRPC) models without toxicity. Thus, impeding SWI/SNF-mediated enhancer accessibility represents a promising therapeutic approach for enhancer-addicted cancers. PROTAC degrader–induced SWI/SNF inactivation abolishes DNA accessibility at enhancer elements of oncogenes and also tempers supra-physiologic expression of driver transcription factors, resulting in potent inhibition of tumour growth in mouse models.

The switch/sucrose non-fermentable (SWI/SNF) complex has a crucial role in chromatin remodelling 1 and is altered in over 20% of cancers 2,3 . Here we developed a proteolysis-targeting chimera (PROTAC) degrader of the SWI/SNF ATPase subunits, SMARCA2 and SMARCA4, called AU-15330. Androgen receptor (AR) + forkhead box A1 (FOXA1) + prostate cancer cells are exquisitely sensitive to dual SMARCA2 and SMARCA4 degradation relative to normal and other cancer cell lines. SWI/SNF ATPase degradation rapidly compacts cis-regulatory elements bound by transcription factors that drive prostate cancer cell proliferation, namely AR, FOXA1, ERG and MYC, which dislodges them from chromatin, disables their core enhancer circuitry, and abolishes the downstream oncogenic gene programs. SWI/SNF ATPase degradation also disrupts super-enhancer and promoter looping interactions that wire supra-physiologic expression of the AR, FOXA1 and MYC oncogenes themselves. AU-15330 induces potent inhibition of tumour growth in xenograft models of prostate cancer and synergizes with the AR antagonist enzalutamide, even inducing disease remission in castration-resistant prostate cancer (CRPC) models without toxicity. Thus, impeding SWI/SNF-mediated enhancer accessibility represents a promising therapeutic approach for enhancer-addicted cancers.
In eukaryotic cells, DNA is wrapped around histone octamers (referred to as nucleosomes), which form a physical barrier to DNA-based processes 4 . Thus, gene expression is regulated by modifying physical accessibility of the DNA through nucleosomal remodelling and, when in an accessible state, through binding of transcription factors 5,6 . In this regulatory context, non-coding genomic elements called enhancers have emerged as central hubs serving as integrative platforms for transcription factor binding and activation of lineage-specific gene programs 7,8 . The enhancer elements can lie within untranslated or distal intergenic regions and make looping interactions with their target gene promoters to potentiate RNA polymerase II (PolII)-mediated transcription 9,10 .
In cancer, genetic alterations invariably lead to an aberrant transcriptional state that is often wired through expansion and remodelling of the enhancer landscape 11,12 . This includes de novo commissioning of new enhancers (neo-enhancers) by reprogramming of pioneer factor cistromes 13 , enhancer hijacking via structural rearrangements 14,15 , and/ or abnormal enhancer-promoter interactions via alterations in chromatin topology 16 -all to enable hyper-expression of driver oncogenes. Although there has been intense interest in therapeutically targeting aberrant enhancer function in cancer, the molecular machinery responsible for enhancer maintenance and/or activation remains poorly characterized.
Recent studies have uncovered alterations in genes encoding constituent subunits of the SWI/SNF complex in over 20% of human cancers 2 . SWI/SNF is a multi-subunit chromatin-remodelling complex that uses energy from ATP hydrolysis to reposition or eject nucleosomes at non-coding regulatory elements, thereby enabling free DNA access for the transcriptional machinery 1 . In SWI/SNF-mutant tumours, the residual complex is thought to enable oncogenic transcriptional programs and speculated to be a viable therapeutic target [17][18][19] . Although inhibitors and degraders of ATPase and BRD7-BRD9 SWI/SNF subunits have been recently developed [20][21][22] , to our knowledge, no studies have comprehensively assessed the therapeutic efficacy of SWI/SNF inactivation across a wide spectrum of cancers. To this end, we have developed and characterized a highly-selective PROTAC degrader of both SWI/ SNF ATPase subunits-SMARCA2 (BRM) and SMARCA4 (BRG1)-that are required for the nucleosomal-remodelling functions of SWI/SNF complexes.
We found enhancer-binding transcription factor-addicted cancers (for example, AR-FOXA1-driven prostate cancer) to be exquisitely and preferentially sensitive to SWI/SNF ATPase degradation, which triggered an instantaneous, specific loss of physical accessibility and transcription factor binding at enhancer elements, thereby disrupting enhancer-wired oncogenic gene programs. To our knowledge, this study is the first preclinical proof of concept that targeted obstruction of chromatin accessibility at enhancer elements may be a potent therapeutic strategy in transcription factor-addicted tumours.

Results
We developed the PROTAC degrader, AU-15330, comprising a bait moiety that binds the bromodomain in SMARCA2 and SMARCA4 and a ligand moiety for the von Hippel-Lindau (VHL) ubiquitin ligase (Fig. 1a, Extended Data Fig. 1a). AU-15330 also binds to the secondary SWI/SNF module component PBRM1, which relies on the ATPase module for assembly onto the core complex 23 . Although it binds to the same bromodomain in target proteins as the PROTAC degrader ACBI1 20 , AU-15330 comprises a distinct linker structure that largely dictates a PROTAC's target selectivity and degradation kinetics 24 . Treatment of several cell lines with AU-15330 led to time and dose-dependent degradation of SMARCA2, SMARCA4 and PBRM1 (Fig. 1b). Mass spectrometry-based proteomics analysis confirmed SMARCA2, SMARCA4 and PBRM1 as the only significantly downregulated proteins (Extended Data Fig. 1b). Of note, we detected no change in the abundance of other bromodomain-containing proteins or non-targeted SWI/SNF subunits (Extended Data Fig. 1c, d). SWI/SNF complexes have been shown to assemble in a modular manner, with the ATPase module being the last to bind to the SMARCC1 (also known as BAF155)-containing core complex 23 . Accordingly, SMARCC1 nuclear immunoprecipitation followed by mass spectrometry showed no changes in the sequential assembly of the core and secondary modules but revealed detachment of ATPase module subunits upon AU-15330 treatment (Extended Data Fig. 1e).
Using a panel of normal and cancer cell lines from 14 distinct lineages, we found AR and FOXA1-driven prostate cancer cells to be preferentially sensitive to AU-15330 (half-maximal inhibitory concentrations (IC 50 ) < 100 nM; Fig. 1c, Extended Data Fig. 1f, g, Supplementary Table 1). AR − FOXA1 − prostate cancer cells showed moderate sensitivity (IC 50 between 100-400 nM), whereas normal and non-neoplastic prostate cells were resistant (IC 50 > 1,000 nM) to AU-15330. We observed a similar cytotoxicity profile for ACBI1 and BRM014, an allosteric dual inhibitor of SMARCA2 and SMARCA4 ATPase activity 25 (Extended Data Fig. 1h, i). Notably, AR + FOXA1 + prostate cancer cells were more sensitive to these inhibitors than SMARCA4-null cancer cell lines. Several MYC-driven multiple myeloma cells and oestrogen receptor-and/or AR-positive breast cancer cells were also acutely sensitive to AU-15330 (Fig. 1c, Extended Data Fig. 1j, k).
In several prostate cancer cell lines, we detected substantial expression of both SWI/SNF ATPases, which were rapidly degraded in a dose-dependent manner by AU-15330 (Extended Data Fig. 2a, b). Concordantly, AU-15330 attenuated the growth of these cells and induced apoptotic cell death, while having no anti-proliferative effect on benign or non-neoplastic prostate cells (grey bars, Fig. 1c) at parallel doses (Extended Data Fig. 1f, 2c-e). Treatment with either the bromodomain ligand alone (AU-15139) or an inactive epimer of AU-15330 (AU-16235) had no effect on target protein levels or  Article cancer cell survival and growth (Extended Data Figs. 1f, g, 2f, g). Next, competition of AU-15330 with a free VHL ligand (VL285), but not with thalidomide, reversed degradation of SWI/SNF targets (Extended Data Fig. 2g) and rescued the growth inhibitory effect in a dose-dependent manner (Extended Data Fig. 2h). Furthermore, pre-treatment of VCaP cells (an AR + FOXA1 + prostate cancer cell line model) with bortezomib (a proteasome inhibitor) or MLN4924 (a NEDD8-activating enzyme inhibitor) hindered target protein degradation, indicating that AU-15330 requires the proteasome machinery and ubiquitination cascade for its action (Extended Data Fig. 2g).
As SWI/SNF complexes actively remodel nucleosomal DNA packaging, we profiled the effect of AU-15330 on physical chromatin accessibility using the assay for transposase-accessible chromatin followed by sequencing (ATAC-seq). We detected a rapid and near-complete loss in chromatin accessibility at more than 30,000 sites in VCaP cells with as little as 1 h of AU-15330 treatment (Fig. 2a), which is within minutes of SMARCA2 and SMARCA4 degradation (Extended Data Fig. 3a); approximately 25,000 genomic sites showed little to no change in nucleosomal density (Extended Data Fig. 3b). Similar profound changes in chromatin accessibility were not observed upon treatment with a BRD4 degrader (ZBC-260; Extended Data Fig. 3a, b). In our genetic models using CRISPR-Cas9 and shRNA-mediated target inactivation, we detected a significant compaction of the chromatin only upon concurrent loss of both SWI/SNF ATPases (Extended Data Fig. 3c, d). More than 90% of the AU-15330-compacted sites were within distal regulatory regions, which were enriched for enhancers, whereas the retained sites were predominantly within promoters (Fig. 2b). De novo motif and binding analysis for the regulation of transcription (BART) analyses of AU-15330-compacted sites identified DNA-binding elements for major oncogenic transcription factors in prostate cancer, including AR, FOXA1, HOXB13 and ERG (Extended Data Fig. 3e, f). As expected, retained promoter sites showed enrichment for PolII and E2F motifs (Extended Data Fig. 3g). Interrogation of chromatin changes in LNCaP cells upon AU-15330 treatment reproduced these findings (Extended Data Fig. 4a-c).
Concurrent with the loss of accessibility, chromatin immunoprecipitation followed by sequencing (ChIP-seq) revealed a decrease in chromatin binding of AR, FOXA1, and ERG in VCaP cells within 1 h of AU-15330 treatment (Fig. 2c, Extended Data Fig. 4d, e). We also detected disappearance of the characteristic 'valley' pattern in the H3K27Ac ChIP-seq signal, indicating the movement of flanking nucleosomes towards the centre of AU-15330-compacted enhancers (Fig. 2d). At early time points, we detected no loss in the abundance of the H3K27Ac mark; however, it was significantly depleted 24 h after AU-15330 treatment (Extended Data Fig. 4f). Similar results were observed upon AU-15330 treatment of LNCaP cells (Extended Data Fig. 4g, h). Loss of AR, FOXA1 and H3K27Ac ChIP signals was evident at enhancer sites of the classical AR target gene KLK3 (Extended Data Fig. 4i). We found AR, FOXA1, ERG and SMARCC1 to co-occupy a large fraction of H3K27Ac-marked regulatory elements (Extended Data Fig. 5a-c). Furthermore, multiple core SWI/SNF components were present in the mass spectrometry-based datasets of AR, FOXA1, and ERG interactomes (Extended Data Fig. 5d), which we confirmed by reciprocal co-immunoprecipitation assays (Extended Data Fig. 5e). This positions SWI/SNF complexes as common chromatin cofactors of the oncogenic transcriptional machinery in prostate cancer cells.
As an important control, we saw no changes in chromatin binding of CTCF in AU-15330-treated cells (Extended Data Fig. 6a-d).
Global transcriptomic profiling with RNA sequencing (RNA-seq) revealed significant downregulation of AR and FOXA1-regulated genes in multiple prostate cancer cells, as well as ERG-regulated transcripts in ERG fusion-positive VCaP cells. We also detected significant loss in the expression of MYC target genes with AU-15330 (Fig. 2e, Extended Data Fig. 6e, f). The global AU-15330 gene signature was highly concordant with transcriptional changes associated with ARID1A loss (Extended Data Fig. 6g). However, neither BRD7 nor BRD9 degradation alone attenuated the expression of classical AR, FOXA1 and ERG target genes or the MYC gene to an extent comparable to AU-15330, suggesting that canonical SWI/SNF (cBAF) complexes are the primary cofactors of oncogenic enhancer-binding transcription factors (Extended Data Fig. 6h-j). The expressions of AR, MYC and FOXA1 genes themselves are frequently amplified in advanced prostate cancer by copy amplification and/or enhancer duplication 15,26,27 . We found that AU-15330 markedly decreased expression of AR, FOXA1, MYC and TMPRSS2-ERG transcripts to 40-60% of their baseline expression (Extended Data Fig. 7a), with parallel decreases at the protein level (Fig. 3a). More severe transcriptional attenuation of these oncogenes was noted upon BRD4 degradation by ZBC-260, with AU-15330 specifically abolishing expression of additional driver oncogenes (Extended Data Fig. 7b), again suggesting a distinct mechanism of action for AU-15330-mediated anti-tumour cytotoxicity. Similar results were observed in genetic-inactivation models (Extended Data Fig. 7c).
The hyper-expression of oncogenes like AR, FOXA1 and MYC in cancer has been shown to be wired through looping interactions with multi-enhancer clusters 15,26,27 , often referred to as super-enhancers. Several such regulatory clusters were identified in cis-proximity of the AR, MYC and TMPRSS2-ERG genes (Fig. 3b), and AU-15330 treatment led to immediate compaction of these sites and loss of H3K27Ac, AR and FOXA1 ChIP-seq signal at the super-enhancers (Fig. 3c, Extended Data Fig. 7d). To detect changes in the interaction of super-enhancers with their target gene promoters, we performed H3K4me3 (active promoter mark) and H3K27Ac Hi-C coupled with ChIP-seq (HiChIP-seq) upon AU-15330 treatment. SWI/SNF inactivation markedly disrupted the three-dimensional looping interactions of cis-enhancers with the AR gene promoter (Fig. 3d, Extended Data Fig. 8a). Similar attenuation of enhancer-promoter interactions was detected by H3K27Ac HiChIP-seq at the FOXA1 locus (Extended Data Fig. 8b), which is recurrently rearranged in advanced prostate cancer 15 . Aggregate peak analyses (APA) of enhancer-promoter interactions showed a marked attenuation of contact strength and/or frequency starting as early as 2 h after AU-15330 treatment, that is, within 1 h of SMARCA2 and SMARCA4 degradation (Fig. 3e, Extended Data Fig. 8c). At these early time points, we did not detect a significant decrease in H3K27Ac signal at the compacted enhancer sites (Fig. 2d), strongly suggesting that physical chromatin accessibility and transcription factor binding serve as primary determinants of functional enhancer-promoter interactions. Of note, we found no change in the looping interactions between CTCF-bound elements (Extended Data Fig. 8d, e). Together, these data show that SWI/SNF ATPase inactivation specifically leads to genome-wide collapse of the AR, FOXA1, ERG and MYC-activated core enhancer circuitry in prostate cancer cells.
Next, we pharmacologically characterized AU-15330 in animal models of advanced prostate cancer. Notably, prolonged AU-15330 treatments showed no evident toxicity in immuno-competent mice (Extended Data Fig. 9, Supplementary Text, Supplementary Table 3). We first employed the VCaP castration-resistant prostate cancer (CRPC) model (VCaP-CRPC) to assess the efficacy of AU-15330. As expected, treatment of castrated male mice bearing the VCaP-CRPC xenografts with enzalutamide (an AR antagonist) showed moderate anti-tumour efficacy; however, treatment with AU-15330 led to potent inhibition of tumour growth, triggering disease regression in more than 20% of animals ( Fig. 4a Treatment with AU-15330 was similarly effective in inhibiting the growth of enzalutamide-resistant cell lines, including derivatives of VCaP and LNCaP cells (Extended Data Fig. 11k-l). The combinatorial regimen also markedly inhibited tumour growth in MDA-PCa-146-12, a patient-derived xenograft (PDX) model that is inherently resistant to enzalutamide (Extended Data Fig. 12a-c). We further established a CRPC variant of the MDA-PCa-146-12 PDX by tumour implantation into castrated mice (Extended Data Fig. 12a). Even in this highly aggressive model, the combinatorial regimen induced significant tumour growth inhibition, causing regression in more than 30% of animals ( Fig. 4e, Extended Data Fig. 12d). In all arms of these studies, we detected no changes in animal body weights (Extended Data Fig. 12e, f). There was also no sign of goblet cell depletion in the gastrointestinal tract (Extended Data Fig. 12g), no defect in germ cell maturation and no testicular atrophy (Extended Data Fig. 12h, i) in AU-15330-treated mice-all of which have been reported as toxicities of therapies targeted towards BET proteins 28 .

Discussion
We report AU-15330 as a novel, highly specific and VHL-dependent PRO-TAC degrader of SWI/SNF ATPase components (SMARCA2, SMARCA4 and PBRM1) that shows preferential cytotoxicity in enhancer-binding transcription factor-addicted cancers at low nanomolar concentrations. Our study identifies the SWI/SNF complex as a transcriptional dependency in AR/FOXA1-driven prostate cancer. Mechanistically, we show that complete inactivation of SWI/SNF ATPase induces a rapid, near-complete and targeted loss of chromatin accessibility at the core-enhancer circuitry of AR, FOXA1, MYC and ERG, thereby attenuating their cancer-promoting transcriptional programs and tempering GAPDH is used as a loading control, and is probed on a representative immunoblot. This experiment was repeated independently twice. b, H3K27Ac ChIP-seq signal rank-ordered list of super-enhancers in VCaP cells with select cis-coded driver oncogenes denoted (HOMER). c, Normalized read density of ATAC-seq at super-enhancers (n = 32,545 sites) in VCaP cells treated with DMSO or AU-15330 (1 μM) for 1 or 4 h (two-sided t-test). In box plots, the centre line shows median, box edges mark quartiles 1-3, and whiskers span quartiles 1-3 ± 1.5 × interquartile range. d, H3K4me3 HiChIP-seq heat maps within the AR gene locus in VCaP cells with or without AU-15330 (1 μM) treatment for 4 h (bin size = 25 kb). ATAC-seq read-density tracks from the same treatment conditions are overlaid. Grey highlights mark enhancers; blue highlights the AR promoter. Loops indicate read-supported cis interactions within the locus. IR, interaction reads. e, APA plots for H3K4me3 and H3K27Ac HiChIP-seq data for all possible interactions between putative enhancers and gene promoters in VCaP cells with or without with AU-15330 treatment (1 μM, 4 h).
Article the enhancer-wired supra-physiologic expression of driver oncogenes (Fig. 4f). These findings are in line with those from recent studies that have used chemical and/or genetic approaches to show that continuous SWI/SNF-remodelling activity is needed to retain enhancers in an open, nucleosome-free conformation 29,30 . To our knowledge, this is the first study to demonstrate that physical chromatin accessibility can be modulated at non-coding regulatory elements as a novel therapeutic strategy in cancer treatment. Thus, recently developed SWI/ SNF ATPase inhibitors and degraders add to the growing arsenal of chromatin-targeted therapeutics for directly combating enhancer addiction in human cancers, warranting assessments of their their safety and efficacy in clinical trials.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-021-04246-z.

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% CO 2 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.

ATAC-seq and analysis
ATAC-seq was performed as previously described 33 33 . 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 CTGTCTCTTATA-CACATCT), 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 TopHat2 34 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 ComplexHeatmap 37 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 ERG 38 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 package 39 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 × 10 6 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 5x10 6  Raw fastq files were aligned using BWA mem (version 0.7.17-r1198dirty) 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 5uniquemax-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 package 37 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 × W 2 ), 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 × 10 6 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 mm 3 ), 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 × 10 6 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 mm 3 ), 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 × 10 6 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 mm 3 ), 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 described 44 . 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 mm 3 tumours coated with 100% Matrigel to both flanks of mice. Once tumours reached ~200 mm 3 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 mm 3 , 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).

TMT mass spectrometry
VCaP cells were seeded at 5 × 10 6 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.

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 package 47 .

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

Data availability
All raw next-generation sequencing, ATAC, ChIP, RNA, and HiChIPseq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) repository at NCBI under accession code GSE171592. Source data are provided with this paper.  Fig. 7 | SWI/SNF inhibition down-regulates the expression of oncogenic drivers through disruption of promoter and super-enhancer interactions. (a, b) RNA expression (RNA-seq) heatmaps from VCaP or LNCaP cells treated with DMSO, AU-15330 (1 μM), or ZBC-260 (BRD4 degrader) for the noted durations. n = 2, biological replicates. (c) RNA expression (qPCR) of indicated genes in stable CRISPR-engineered LNCaP-sgNC (control) or LNCaP-sgSMARCA2 (SMARCA2 inactivation) cells that were treated with a non-target control shRNA or two distinct shRNAs targeting the SMARCA4 gene. Data are presented as mean +/− SD (n = 3, technical replicates) from one-of-two independent experiments. Right, immunoblots showing expression of the indicated protein in CRISPR/shRNA-engineered LNCaP cells. Vinculin is the loading control probed on a representative immunoblot. This experiment was repeated independently twice. (d) Normalized read density of AR, FOXA1 and H3K27Ac ChIP-seq signal at the super-enhancer sites (n = 1,551 sites) in VCaP cells treated with DMSO or AU-15330 (1 μM) for 4h or H3K27Ac with 24h AU-15330 (two-sided t-test). For all box plots, the center shows median, box marks quartiles 1-3, and whiskers span quartiles 1-3 ± 1.5 × IQR. Fig. 8 | Enhancer-promoter interactions at loci of oncogenic transcription factors with AU-15330. (a) ATAC-seq and ChIP-seq tracks for AR, FOXA1, and H3K27Ac within the AR gene locus in VCaP cells with or without AU-15330 treatment (1 μM for 6h for AR and FOXA1; 1 μM for 24h for H3K27Ac). (b) H3K27Ac HiChIP-seq heatmaps within the FOXA1 gene locus in VCaP cells plus/minus treatment with AU-15330 (1 μM) for 4h (bin size = 25Kb). ATAC-seq read-density tracks from the same treatment conditions are overlaid. Grey highlights mark enhancers, while the blue highlight marks the FOXA1 promoter. (c) Aggregate peak analysis (APA) plots for H3K4me3 (active promoter mark) HiChIP-seq data for all possible interactions between putative enhancers and gene promoters in VCaP cells plus/minus treatment with AU-15330 (1 μM) for noted durations. (d) APA plots for CTCF HiChIP-seq data for all possible interactions between CTCF-bound insulator elements in VCaP cells plus/minus treatment with AU-15330 (1 μM, 4h). TAD, topologically associating domain. (e) CTCF HiChIP-seq heatmaps in a gene locus at Chr14, including the FOXA1 topologically associating domain (TAD), in VCaP cells plus/minus treatment with AU-15330 (1 μM) for 4h (bin size = 100Kb). CTCF ChIP-seq read-density tracks from VCaP cells plus/minus AU-15330 treatment (1 μM) for 6h are overlaid.