A non-canonical SWI/SNF complex is a synthetic lethal target in cancers driven by BAF complex perturbation

Abstract

Mammalian SWI/SNF chromatin remodelling complexes exist in three distinct, final-form assemblies: canonical BAF (cBAF), PBAF and a newly characterized non-canonical complex (ncBAF). However, their complex-specific targeting on chromatin, functions and roles in disease remain largely undefined. Here, we comprehensively mapped complex assemblies on chromatin and found that ncBAF complexes uniquely localize to CTCF sites and promoters. We identified ncBAF subunits as synthetic lethal targets specific to synovial sarcoma and malignant rhabdoid tumours, which both exhibit cBAF complex (SMARCB1 subunit) perturbation. Chemical and biological depletion of the ncBAF subunit, BRD9, rapidly attenuates synovial sarcoma and malignant rhabdoid tumour cell proliferation. Importantly, in cBAF-perturbed cancers, ncBAF complexes maintain gene expression at retained CTCF-promoter sites and function in a manner distinct from fusion oncoprotein-bound complexes. Together, these findings unmask the unique targeting and functional roles of ncBAF complexes and present new cancer-specific therapeutic targets.

Figure 1.. mSWI/SNF family complexes are biochemically and functionally distinct.

a. Principal component analysis (PCA) performed on fitness correlations between mSWI/SNF genes from combined genome-scale RNAi- and CRISPR-Cas9-based genetic perturbation screens (Project Achilles, Broad Institute). b. SDS-PAGE and silver stain performed on purified complexes using indicated HA-tagged subunits expressed in HEK-293T cells. Purifications performed in n=3 biologically independent experiments for each bait. c. Proteomic mass-spectrometry performed on mSWI/SNF complexes purified from HEK-293T cells expressing indicated HA-tagged mSWI/SNF subunits. d. Immunoprecipitation for endogenous SMARCA4 (pan-mSWI/SNF complex component), ARID1A (canonical BAF-specific), BRD7 (PBAF-specific), and BRD9 (ncBAF-specific) subunits in HEK-293T nuclear extracts followed by immunoblot for select subunits. Subunits in blue, red, and green represent BAF-, PBAF-, and BRD9/GLTSCR1- specific complexes, respectively. Immunoprecipitations performed in n=2 biologically independent experiments. See also Supplementary Figure 7a. e. Separation of HEK-293T nuclear extracts via 10–30% glycerol gradient density sedimentation followed by immunoblot for selected mSWI/SNF subunits. Glycerol gradient performed in n=3 biologically independent experiments. See also Supplementary Figure 7b. f. Schematic depicting biochemical subunit compositions for mammalian ncBAF, canonical BAF, and PBAF complexes.

Figure 2.. Differential localization of mSWI/SNF complexes, ncBAF, cBAF, and PBAF, on chromatin.

a. Venn diagram of peaks from BRD9, GLTSCR1, and SMARCA4 ChIP-seq experiments. b. Heatmap representing correlations between normalized ChIP-seq reads (Log2(RPM)) over a merged set of all mSWI/SNF subunit peaks. ChIP performed in n=2 independent samples for each. c. Localization of ncBAF, BAF, and PBAF complexes at the VEGFA locus. ChIP performed in n=2 independent samples for each. d. Heatmap of CentriMo log adjusted p-values for top motifs returned by MEME-ChIP analysis for each ChIP-seq experiment. ChIP performed in n=2 independent samples for each, p-values were calculated using binomial test. e. Proportion of peaks from ChIP-seq experiments using indicated antibodies overlapping CTCF ChIP-seq peaks in MOLM-13 and EoL-1 cell lines. f. Pie graphs reflecting proportion of ncBAF-, BAF-, and PBAF- specific peaks overlapping with specified chromatin features (see also Supplementary Fig. 2i). g. Example tracks demonstrating differential mSWI/SNF complex family enrichment across the AFTPH locus. ChIP performed in n=2 independent samples for each.

Figure 3.. ncBAF components are selective synthetic lethal dependencies in synovial sarcoma and rhabdoid tumor cell lines.

a. Waterfall plots for CERES scores across n=393 cell lines screened using CRISPR-Cas9 (Project Achilles). SS (orange) and MRT (blue) cell lines are indicated. Dashed line represents median dependency. b. BRD9 sensitivity across n=387 cell lines in Project DRIVE. Two-sided Fisher’s exact test -log10(pvalue) for BRD9 sensitivity (ATARIS score<-.75) against median z-score in each cancer type. Annotations with FDR<0.1 are colored in red. c. Heatmap of CERES scores in SYO-1 (SS18-SSX-driven SS) and SW982 (histological SS mimic without SS18-SSX translocation) ranked by difference in dependency. d. (Top) Immunoblot on total cell lysates from SYO-1 cells in each knockdown condition (n=2 biologically independent experiments); (Bottom) Proliferation experiments in SYO-1 representative of n=3 biologically-independent experiments (with separate lentiviral infection and cell number calculations) performed. Each data point represents mean +/- SD, p-value calculated by two-sided t-test on day 19. See also Supplementary Figure 7c, Supplementary Table 2. e. (Left) Chemical structure of dBRD9 degron compound; (Right) Immunoblot on total cell lysates from SYO-1 cells treated with DMSO vehicle control or dBRD9 (500nM) for 3 days (n=2 biologically independent experiments). See also Supplementary Figure 7d. f. Proliferation experiments in SYO-1 representative of n=3 biologically-independent experiments (separate treatments and cell number calculations). Each data point represents mean +/- SD, p-value calculated by two-sided t-test on day 8. See also Supplementary Table 2. g. Heatmap of gene expression changes of significantly changed genes (p-adjusted<.001 and log2(fc)>.59; Benjamini-Hochberg) in any one of the four treatments. Genes were k-means clustered into 2 groups, samples were clustered hierarchically. n=2 biological replicates for each RNA-seq experiment. h. (Top) Immunoblot on total cell lysates from TTC1240 malignant rhabdoid tumor cells treated with either DMSO vehicle control or dBRD9 (250nM) (n=2 biologically independent experiments); (Bottom) Proliferation experiments in TTC1240 representative of n=3 biologically-independent experiments (separate treatments and cell number calculations). Each data point represents mean +/- SD, p-value calculated by two-sided t-test on day 7. See also Supplementary Figure 7e, Supplementary Table 2. i. Representative colony formation assay performed on SYO-1 cells treated with dBRD9, BI-7273, or lenalidomide as a control (n=3 biologically independent experiments).

Figure 4.. ncBAF subunit domains underlie complex-specific synthetic lethalities.

a. Schematic for tiled gRNA CRISPR-Cas9 screening performed across mSWI/SNF subunit genes in SYO-1 SS cells. b. Tiled CRISPR screening for the GLTSCR1 gene in SYO-1 cells. c,d. Tiled CRISPR screening for the BRD9 gene (c) and BRD7 gene (d) in SYO-1 cells. e. Alignment of GLTSCR1 amino acid sequences across species. GLTSCR domain is highlighted. f. Alignment of amino acid sequences for BRD9 and BRD7 across species. Bromodomain and DUF3512 are highlighted. g. (Top) Construct design for GLTSCR domain experiments in HEK-293T cells. (Bottom) Immunoprecipitation of V5-tagged constructs followed by immunoblot (n=2 biologically independent experiments). See also Supplementary Figure 7f. h. (Left) Construct design for C-terminal swap experiments for BRD9 and BRD7 in HEK-293T cells (BD = bromodomain). (Right) Immunoprecipitation of BRD9, BRD7, and BRD7(B9C) and BRD9(B7C) C-terminal swap variants followed by immunoblot in HEK-293T cells (n=2 biologically independent experiments). See also Supplementary Figure 7g.

Figure 5.. ncBAF is not required for SS18-SSX1-mediated gene expression and primarily regulates fusion-independent sites.

a. Immunoblot for ncBAF components in HA-SS18 and HA-SS18-SSX complex purifications (n=1 biologically independent experiment). See also Supplementary Figure 7h. b. Heatmap of significantly downregulated genes (p-adjusted<1e-3 and log2(fold change)< -0.59) in shSS18-SSX (7 days post infection) and dBRD9 (3 and 6 days) conditions in SYO-1 cells k-means clustered into 4 groups. n=2 biological replicates for each RNA-seq experiment, p-adjusted values are Benjamini-Hochberg adjusted Wald p-values. c. GSEA of RNA-seq data for shSS18-SSX and dBRD9 (day 6) conditions in (b). Specific pathways and gene sets are indicated. d. (Top) Immunoblot on CRL7250 whole cell lysates described in Figure S5B (n=1 biologically independent experiment); (Bottom) Heatmap of log2(fold change) of gene expression in CRL7250 fibroblast cells treated with DMSO, dBRD9, or dBRD9 followed by lentiviral introduction of V5-SS18 or V5-SS18-SSX (n=2 biological replicates for each RNA-seq experiment). Genes included were expressed (>1 RPKM) and had a log2(fold change) of at least +/-.59 in at least one of the conditions. Genes were k-means clustered into 2 groups and samples were clustered hierarchically. See also Supplementary Figure 7i. e. Heatmap of ChIP-seq read density of SS18, BRD9, and H3K4me3 over SS18 sites in SYO-1 cells (shScr (control hairpin) and shSSX (targeting SS18-SSX) conditions), clustered into 3 groups. f. Box plot of log2(fold change) in gene expression of genes closest to fusion-dependent sites (n=595) in shSS18-SSX and dBRD9 conditions. n=2 independent samples for each ChIP performed, significance was calculated by two tailed t-test. Box represents interquartile range (IQR), bar in center shows data median. Minima and maxima shown extend to from the box +/- 1.5*IQR. g. Pie chart representing chromatin landscape (fusion-dependent, fusion-independent promoter, fusion-independent distal) of the nearest BRD9 peak to the top 500 most downregulated genes. h. Violin plot of CERES scores for genes that changed with a significance of p-adjusted<1e-3 after 6 days of dBRD9 treatment in SYO-1 cells. P-value between conditions calculated by two tailed t-test. Violin plot represents kernel density estimation with data quartiles represented as lines, the data median is shown as a dot.

Figure 6.. ncBAF is required for maintenance of gene expression and retains co-localization with promoters and CTCF in SMARCB1-deficient cancers.

a. Venn diagram of BRD9 and SMARCA4 ChIP-seq peaks in TTC1240. b. Bar plot of proportion of SMARCA4 peaks that overlap with a BRD9 peak in SS, MRT and mSWI/SNF-intact hematopoietic cancer cell lines. n=2 independent ChIPs, p-value calculated with two tailed t-test. c. Proportion of MRT-specific super-enhancers (SE) (Chun et al.) overlapping a BRD9 peak in TTC1240. d. log2(fold change) in SMARCA4 occupancy against mean occupancy between DMSO and dBRD9 treated TTC1240 cells. Highlighted, occupancy change of FDR<5e-2. n=2 biological replicates per ChIP, FDR values are multiple test corrected Wilcox test p-values. e. Tracks showing BRD9, H3K27ac, and SMARCA4 +DMSO and +dBRD9 occupancy in TTC1240 at SPARCL1 (n=2 independent samples for each ChIP performed). f. Spike-in normalized heatmap of SMARCA4 and BRD9 ChIP occupancy across lost SMARCA4 sites in TTC1240 upon dBRD9 treatment ranked by SMARCA4 occupancy in DMSO condition. g. Boxplot of H3K27ac ChIP occupancy in TTC1240 cells at SMARCA4 sites lost or retained upon dBRD9 treatment. Box represents interquartile range (IQR), bar in center shows data median. Minima and maxima shown extend to from the box +/- 1.5*IQR with data falling outside of that range shown as points. h. Volcano plot of gene expression changes in TTC1240 with dBRD9 treatment (7 days). Blue indicates genes with TSS within 100kb of a lost site. MRT disease-associated genes (Chun et al.) with TSS<100kb from peak are labeled. n=2 biological replicates for each RNA-seq experiment, p-adjusted values are Benjamini-Hochberg adjusted Wald p-values. i. Histograms of log2(fold change) in SMARCA4 ChIP occupancy across SMARCA4 peaks in TTC1240 and MOLM-13 after dBRD9 treatment. j,k. SMARCA4 (j) and BRD9 (k) peak distribution in BAF-perturbed (TTC1240 and SYO-1) and BAF-wild-type (EoL-1, MOLM-13, Jurkat) settings. l, m. Model for ncBAF dependency in cancers driven by cBAF perturbations. Perturbations in the core BAF functional module (SMARCB1, SMARCE1, ARID1A/B, with the exception of the ATPase subunits) result in loss of cBAF gene regulatory function and reliance on ncBAF for gene expression maintenance at hallmark ncBAF promoters and CTCF sites (l). Chemical or biologic ncBAF disruption results in loss of gene expression maintenance (m).