The SS18-SSX Fusion Oncoprotein Hijacks BAF Complex Targeting and Function to Drive Synovial Sarcoma

SUMMARY

Synovial sarcoma (SS) is defined by the hallmark SS18-SSX fusion oncoprotein, which renders BAF complexes aberrant in two manners: gain of SSX to the SS18 subunit and concomitant loss of BAF47 subunit assembly. Here we demonstrate that SS18-SSX globally hijacks BAF complexes on chromatin to activate an SS transcriptional signature that we define using primary tumors and cell lines. Specifically, SS18-SSX retargets BAF complexes from enhancers to broad polycomb domains to oppose PRC2-mediated repression and activate bivalent genes. Upon suppression of SS18-SSX, reassembly of BAF47 restores enhancer activation, but is not required for proliferative arrest. These results establish a global hijacking mechanism for SS18-SSX on chromatin, and define the distinct contributions of two concurrent BAF complex perturbations.

In Brief

Incorporation of the synovial sarcoma SS18-SSX fusion into BAF complexes results in concomitant eviction of BAF47. McBride et al. show that SS18-SSX retargets BAF complexes from enhancers to polycomb domains to oppose PRC2-mediated repression. Reincorporation of BAF47 upon suppression of SS18-SSX restores enhancer activation but is not required for proliferative arrest.

Graphical Abstract

INTRODUCTION

Precise regulation of gene expression is required for the establishment of differentiated tissue types that possess diverse, highly specialized functions (Dawson and Kouzarides, 2012). Modulation of chromatin architecture via DNA methylation, covalent modification of histones, and ATP-dependent chromatin remodeling, plays a critical role in maintaining appropriate gene expression (Clapier and Cairns, 2009). The mammalian SWI/SNF (BAF) ATP-dependent chromatin remodeling complex remodels nucleosome architecture, creating DNA accessibility and enabling binding of transcriptional machinery at both enhancers and promoters (Kadoch and Crabtree, 2015). Through combinatorial assembly of both ubiquitously expressed core subunits and cell-type-specific subunits, BAF complexes orchestrate gene regulation across numerous developmental contexts including pluripotency and neurogenesis (Ho and Crabtree, 2010; Ho et al., 2009; Singhal et al., 2010; Yoo et al., 2009, 2011).

Recent exome sequencing studies have revealed the major contribution of these regulators to human disease, demonstrating that genes encoding BAF complex subunits are recurrently mutated in >20% of human cancers as well as a number of intellectual disability syndromes (Kadoch and Crabtree, 2015; Kadoch et al., 2013). A link between BAF complexes and cancer was first established in malignant rhabdoid tumor (MRT), in which biallelic loss of SMARCB1, which encodes BAF47, is the singular, driving event (Versteege et al., 1998). Over the past several years, sequencing studies have demonstrated cancer-specific loss-of-function BAF complex gene mutations, such as PBRM1 mutations in clear cell renal cell carcinoma and SMARCE1 mutations in clear cell meningioma (Smith et al., 2013; Varela et al., 2011), highlighting distinct, cell-type-specific routes toward oncogenesis (Pulice and Kadoch, 2016). Recently, fusion oncoproteins such as SS18-SSX in synovial sarcoma (SS) and EWS-FLI1 in Ewing sarcoma have been shown to alter BAF complex composition and/or targeting (Boulay et al., 2017; Kadoch and Crabtree, 2013).

SS comprises 10% of all soft-tissue tumors, and nearly 100% of tumors contain the pathognomonic t(X;18)(p11.2;q11.2) chromosomal translocation that produces the hallmark SS18-SSX fusion protein. In SS18-SSX, eight C-terminal amino acids of SS18 are replaced by 78 amino acids of the C-terminus of SSX1, SSX2, or, rarely, SSX4 (Clark et al., 1994). SS has been suggested to arise from a myogenic lineage, with activation of neural genes implicated in oncogenesis (Haldar et al., 2007; Ishibe et al., 2008). SS18 is a dedicated subunit of BAF complexes, and the SS18-SSX fusion retains the ability to stably incorporate. Upon incorporation, SS18-SSX dominantly replaces wild-type (WT) SS18 and results in the eviction of the BAF47 tumor suppressor subunit (Kadoch and Crabtree, 2013). As such, SS presents a highly unique setting in which to decompose the relative contributions of SS18-SSX fusion incorporation and BAF47 subunit eviction in the development and maintenance of disease, as well as to define the unique, cancer-specific properties of SS18-SSX-containing ATP-dependent chromatin remodeling machines.

The careful balance between gene activation and repression is achieved in part through BAF complexes opposing polycomb-mediated repression (Ho et al., 2011; Kia et al., 2008). Recent studies have shown that BAF complex recruitment leads to rapid eviction of polycomb repressive complexes (PRCs) and their associated marks (Kadoch et al., 2017). PRCs repress bivalent promoters dually marked by H3K4me3 and H3K27me3 (Bernstein et al., 2006; Mikkelsen et al., 2007); indeed, more recent studies have shown that PRC2 loss specifically activates genes with bivalent promoters (Jadhav et al., 2016; Voigt et al., 2013). BAF complex subunit genes have long been categorized as trithorax members since early studies in Drosophila (Kennison and Tamkun, 1988), with the balance between Trithorax (Trx) and Polycomb (PcG) proteins implicated in the maintenance of normal development and oncogenesis (Schuettengruber et al., 2017). We have recently demonstrated that the BAF47 subunit of BAF complexes is required for BAF complex-mediated activation of enhancers and bivalent promoters, demonstrating the critical role of BAF-PRC opposition in cancer through regulation of this balance (Nakayama et al., 2017). In SS cell lines, BAF complexes have been previously shown to localize to the SOX2 locus, correlating with reduction of H3K27me3 mark occupancy and activation of SOX2 expression (Kadoch and Crabtree, 2013), suggesting a unique mechanism directly opposite of that in MRT and other BAF47-deficient sarcomas (McBride and Kadoch, 2018). We sought to define the mechanism by which BAF complexes containing the SS18-SSX fusion and associated BAF47 eviction contribute to oncogenesis in SS cell lines and patient samples.

RESULTS

SS18-SSX Globally Retargets BAF Complexes to Unique Sites with Broad Occupancy Patterns

To define the consequences of SS18-SSX protein expression and complex incorporation on the targeting of BAF complexes genome wide, we performed fusion-specific (shSSX) knockdown compared with scrambled hairpin control (shCt) in the Aska SS cell line. Upon knockdown of SS18-SSX1, protein levels of WT SS18 and BAF47 are rescued (Figures 1A and S1A–S1C), as shown previously (Kadoch and Crabtree, 2013). In these conditions, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq), using antibodies specific for SS18 (captures both WT SS18 and SS18-SSX), as well as core BAF complex subunits, BAF155 and BRG1. Unexpectedly, we found that SS18-SSX expression causes genome-wide BAF complex retargeting, encompassing both gains and losses of BAF complex target sites (Figures 1B and S1D). This extent of BAF complex retargeting conferred by SS18-SSX is distinct from the near-exclusive losses in genome-wide complex occupancy observed in BAF47-deficient tumor settings (Nakayama et al., 2017), indicating that the dual gain and loss of BAF complex localization is a distinct feature of SS.

Figure 1. SS18-SSX Drives Genome-Wide Retargeting of BAF Complexes to Broad Domains.

(A) Immunoblot for BRG1, BAF155, SS18, BAF47, and GAPDH performed on Aska SS cell whole-cell extracts in shCt and shSSX conditions (using two distinct shRNAs targeting SSX).

(B) Venn diagram of SS18 (left) and BAF155 (right) ChIP-seq peaks in shCt and shSSX conditions in Aska cells.

(C) IPs for BRG1, BAF250A (BAF complex-specific), and BAF200 (PBAF complex-specific) using nuclear extracts from Aska cells.

(D) Heatmaps of SS18, BAF155, BRG1, and BAF200 occupancy in Aska shCt and shSSX conditions over all SS18 peaks in shCt or shSSX conditions (32,077),ranked by log2(fold change) in SS18 occupancy.

(E) Example SS18, BAF155, BRG1, BAF200, RNA Pol II, ChIP-seq, and RNA-seq tracks at the FOXC1 locus in Aska cells in shCt and shSSX conditions.

(F) Distance to TSS graph for shCt-only (shSSX-lost), shCt-shSSX shared, and shSSX-only (shSSX-gained) SS18 peaks in Aska cells.

(G) Cumulative distribution function of ChIP-seq peak widths for shCt-only, shCt-shSSX shared, and shSSX-only SS18 peaks in Aska cells, and SS18 peaks in CRL7250 cells.

(H) Venn diagram of SS18 peaks in shCt and shSSX conditions in SYO1 SS cells.

(I) Heatmaps of SS18 occupancy in SYO1 shCt and shSSX conditions over all SS18 peaks in shCt or shSSX conditions (17,111), ranked by log2(fold change) in SS18 occupancy.

(J) Venn diagram of V5 peaks in V5-SS18 WT and V5-SS18-SSX1 expressing conditions in CRL7250 fibroblast cells.

(K) Example SS18 and V5 tracks at the SOX8 and CAV1 loci in Aska and SYO1 shCt and shSSX conditions, as well as CRL7250 in V5-SS18 WT and V5-SS18-SSX1 conditions.

See also Figure S1.

To further refine the subset of mammalian SWI/SNF family complexes (BAF and PBAF) bound and retargeted by SS18-SSX in SS, we performed co-immunoprecipitation experiments in Aska SS cells. We found that SS18-SSX co-immunoprecipitated with BRG1 (core ATPase of both complexes) and BAF250A (BAF-specific), but not with BAF200 (PBAF-specific), affirming that SS18-SSX selectively incorporates into BAF complexes and does not bind PBAF complexes (Figure 1C) (Middeljans et al., 2012). We found that BAF200 was not markedly retargeted genome wide by SS18-SSX, and that PBAF complexes exhibited a promoter-proximal distribution similar to RNA polymerase II (Pol II) (Figures 1D, 1E, and S1E). However, PBAF complexes targeted to activated SS18-SSX target genes, suggesting recruitment downstream of SS18-SSX-mediated BAF complex targeting, in order to regulate promoters in a manner similar to the related RSC complexes (Krietenstein et al., 2016).

We designated SS18 sites as either shCt-only (8,610), shCt-shSSX shared (9,546), or shSSX-only (13,921) based on the fold change in SS18 occupancy between shCt and shSSX conditions (Figure 1D). We found that shCt-only sites exhibit an increased promoter-proximal localization compared with the shSSX-only gained sites, and that shared shCt-shSSX sites largely reflect retained promoter regulation unperturbed by SS18-SSX (Figure 1F). Intriguingly, we found that fusion target sites were significantly broader (median = 2,236 bp) than shared shCt-shSSX (638 bp) or shSSX-only (583 bp) sites, and that these broad peaks are a unique feature of SS18-SSX targeting that is not shared by WT BAF complexes in fibroblasts (Figure 1G). We observed similar global retargeting of BAF complexes to broad domains in SS18-SSX2-containing SYO1 SS cells (Figures 1H, 1I, S1F, and S1G).

We also expressed V5-SS18 WT or V5-SS18-SSX1 in human fibroblasts (CRL7250), and observed fusion-mediated retargeting to sites shared with SS18-SSX targets in Aska cells, accompanied by de novo DNA accessibility and gene expression changes, as assessed by assay for transposase-accessible chromatin sequencing (ATAC-seq) and RNA sequencing (RNA-seq), respectively (Figures 1J and S1H–S1N). Changes in BAF complex peak width and locus-specific occupancy directed by SS18-SSX incorporation are exemplified at the SOX8 and CAV1 loci in SS and fibroblast lines (Figure 1K). These results demonstrate a highly cancer-specific, gained genome-wide targeting of BAF complexes conferred by SS18-SSX, which is distinct from the targeting of WT BAF complexes.

Concordant, Direct Activation of Oncogenic Target Loci by SS18-SSX-Bound BAF Complexes

We next aimed to determine the concordance of SS18-SSX-mediated gene regulation across distinct SS cell lines. From RNA-seq profiles of six SS cell lines, we selected HSSY2, a monophasic SS18-SSX1 line, and SYO1, a biphasic SS18-SSX2 line, as cell lines most trancriptionally distinct from Aska cells (Figure S2A). We performed RNA-seq in shCt and shSSX conditions in HSSY2 and SYO1 cells, and identified 568 genes with significant changes in expression shared in all three cell lines (Figures 2A, 2B, and S2B–S2D). We found that 528 genes (92.9%) were concordantly regulated in the same direction across the three cell lines upon SS18-SSX knockdown (Figures 2C and 2D). In addition to genes consistently downregulated (i.e., HOXC10 and BCL2), we also observed a set of genes, including PAX7 and SOX2, which are specifically regulated in a subset of SS lines (Figures 2D and 2E). Gene ontology (GO) analysis shows enrichment of embryonic gene sets among the concordantly downregulated genes, implicating these pathways in mediating oncogenesis (Figures 2F and S2E).

Figure 2. Concordant Direct Regulation by SS18-SSX-Containing BAF Complexes across SS Cell Lines.

(A) Immunoblot for SS18, BAF47, and GAPDH performed on whole-cell extracts from Aska (biphasic, SS18-SSX1), SYO1 (biphasic, SS18-SSX2), and HSSY2 (monophasic, SS18-SSX1) cells in shCt and shSSX conditions.

(B) Venn diagram of significantly changed genes in Aska, SYO1, and HSSY2 cells by RNA-seq upon SS18-SSX knockdown.

(C) Hierarchical clustering of log2(fold change) in gene expression (RNA-seq) for Aska, SYO1, and HSSY2 cells in shCt and shSSX conditions over the shared 568 significantly changed genes, as shown in (B). n = 2 biological duplicates for each condition. Significantly upregulated (n = 351) or downregulated (n = 177) genes across all three cell lines are noted.

(D and E) Log2(fold change) (log2FC) in expression of genes that are consistently regulated (D) and differentially regulated (E) across Aska, SYO1, and HSSY2 cells in shCt and shSSX conditions. n = 2 biological duplicates for each condition.

(F) GO term analysis of shared upregulated genes (red) and downregulated genes (blue) following SS18-SSX knockdown. Number of genes in each category indicated in parentheses.

(G) Plot of –log10(p value) versus log2FC in gene expression in Aska shCt versus shSSX conditions. Significantly changed shCt-only SS18 targets (blue) and shSSX-only SS18 targets (red) are labeled with density plot (right) reflecting SS18 ChIP-seq density in each condition.

(H) Pie chart of significantly downregulated genes upon SS18-SSX knockdown in Aska cells (n = 830) showing the proportion of downregulated genes that are unique shCt-only SS18 (SS18-SSX) targets (blue) and downstream secondary targets (grey) in Aska cells.

(I) GO term analysis of SS18-SSX primary target genes (n = 289) and secondary non-target genes (n = 541) downregulated upon SS18-SSX knockdown (n = 830) in Aska. Number of genes in each category indicated in parentheses.

(J) Overlap of primary SS18-SSX targets in Aska and SYO1 cells significantly downregulated upon SS18-SSX knockdown.

(K) Example SS18 ChIP-seq and RNA-seq tracks for Aska and SYO1 in shCt and shSSX conditions at the SOX8 and PAX3 loci.

See also Figure S2.

We next sought to define the set of primary gene targets of SS18-SSX-containing BAF complexes activated by the fusion. Comparing defined shCt-only and shSSX-only SS18 peaks (Figure 1D) with transcriptional changes, we found that significantly changed targets of SS18-SSX-containing BAF complexes (n = 394) are disproportionately downregulated, whereas the significantly changed targets of SS18 WT (n = 176) are disproportionately upregulated upon fusion knockdown (Figure 2G). Of 830 genes downregulated in Aska, we found that 289 (34.8%) were primary targets of SS18-SSX and 541 (65.2%) were downstream secondary targets (Figure 2H). GO term analysis identified significant enrichment of neural and embryonic pathway among primary SS18-SSX targets, whereas cell-cycle pathways were enriched among secondary targets (Figure 2I). These results demonstrate that BAF complex retargeting and restoration of gene repression following fusion knockdown plays a critical role in halting oncogenic gene expression programs mediated by SS18-SSX.

To characterize both conservation and heterogeneity among SS18-SSX primary targets, we performed these analyses in another SS line (SYO1) and identified 406 primary targets activated by SS18-SSX (Figures S2F–S2H). Using the set of SS18-SSX sites (as defined by shCt-only SS18 peaks) in Aska and SYO1, we identified 1,934 SS18-SSX sites as shared fusion targets in both cell lines, and found that these peaks are significantly broader (median width = 7,789 bp) than fusion peaks specific to either line (Figures S2I and S2J), indicating that the broadest domains targeted by SS18-SSX are shared across SS lines. We identified 978 genes that are targets of SS18-SSX in both Aska and SYO1; however, we found that only 87 of these genes are primary targets activated by the SS18-SSX fusion in both cell lines (Figures 2J, 2K, and S2K). These results suggest a mechanism dictating the differential activation of target genes within broad domains across distinct cell lines. Collectively, these studies define gene expression features mediated specifically by SS18-SSX targeting of BAF complexes, indicating the direct role of these complexes in the activation of neural andembryonic pathways that promote cell-cycle regulation and proliferation in SS.

SS Is Genomically Stable and Transcriptionally Distinct from Cancers Driven by BAF Complex Loss-of-Function Mutations

Given the unique targeting and gene activation mediated by SS18-SSX-containing BAF complexes, we sought to determine whether the SS18-SSX oncogenic fusion is the sole driving genetic perturbation. We performed whole-exome sequencing of 18 matched primary SS-normal pairs, and combined these data with 10 TCGA SS exome datasets (Cancer Genome Atlas Research Network, 2017) (Table S1). We compared the mutation frequency observed in SS patient samples with a spectrum of cancers (Cancer Genome Atlas Research Network, 2011; Lawrence et al., 2014). We find that SS is a genomically quiet malignancy lacking any identifiable recurrent secondary mutations (Figures 3A, S3A, and S3B). The SS mutational burden (median = 0.589 mutations/kb exome) is slightly higher than that of MRT (median = 0.067 mutation/kb exome), which has the lowest mutational burden of any cancer sequenced to date (Lawrence et al., 2013; Lee et al., 2012). These data suggest that the SS18-SSX oncogenic fusion is the singular driving genetic event in SS.

Figure 3. SS Is Transcriptionally Distinct from Other Malignancies with Hallmark BAF Complex Perturbations.

(A) Each dot represents a tumor-normal pair from whole-exome sequencing. The y-axis indicates the non-synonymous mutation frequency in each sample. Tumor types are ordered by their median mutation frequency, indicated by the horizontal yellow line. SS samples are highlighted in red. Abbreviations are for tumor samples in (Lawrence et al., 2014) and (Cancer Genome Atlas Research Network, 2011).

(B) Tumor types characterized by BAF complex gene perturbations used in RNA-seq clustering analyses in (C–E). Subunit affected and numbers of cases analyzed are represented.

(C) Principal component analysis of the top 5% most variable genes for tumor samples in (B) with SS highlighted in green.

(D) Z score expression of genes correlated most (|PCC| > 0.5) positively (top, red, n = 346) or negatively (bottom, blue, n = 110) with PC1, from top 5% most variable genes, with samples ranked by PC1 scores. Example genes are labeled on the right.

(E) GO term analysis of SS (346 genes, red) and other BAF-mutant tumor (110 genes, blue) gene signatures. Number of genes in each category indicated in parentheses.

(F) NMF clustering of SS RNA-seq samples (n = 64) reveals two major transcriptional subgroups.

(G) Box plot of gene expression levels (reads per kilobase million [RPKM]) for subgroup markers SOX2, MYC, PAX7, and PAX3, with groups corresponding to those in (F). ***p <0.001, **p <0.01, *p <0.05 by two-tailed t test. For boxplots, box represents interquartile range, the horizontal line inside the box represents the median, whiskers extend to extremes, and individual values are shown.

(H) Scatterplots reflecting exclusivity of MYC and PAX7 gene expression in SS primary samples, with groups corresponding to those in (F).

(I) Binary calls for TMA immunohistochemical staining (black, positive; gray, negative) for SOX2, MYC, PAX7, and PAX3 in SS TMA cases (n = 186).

(J) Representative images of immunohistochemical stains for SOX2, MYC, PAX7, and PAX3 on SS TMA. Scale bar, 25 μm.

See also Figures S3 and S4, Tables S1 and S2.

To characterize the transcriptional signatures in SS compared with other BAF complex-mutated tumors, we performed comprehensive RNA-seq on 64 primary SS specimens containing either SSX1 or SSX2 fusions (Table S2). SSX1 or SSX2 expression was entirely restricted to the exons involved in the fusion gene (Figures S3C and S3D). We did not find global transcriptional distinctions between SS18-SSX1 and SS18-SSX2 cases (Figure S3E). In addition, we identified alternative fusion breakpoints in SSX1, each of which retained additional exons of SSX1 along with the canonical exons encoding the 78 amino acids of SSX (Figures S3F and S3G). These results demonstrate globally similar transcriptional profiles among SS18-SSX1 and SS18-SSX2 cases, and establish that SSX1/2 expression in SS tumors results exclusively from the fusion gene.

To define the transcriptional features unique to SS, we compared these RNA-seq profiles with those of other cancers with hallmark BAF complex perturbations, including MRT, epitheliod sarcoma (EpS), renal medullary carcinoma (RMC), small-cell carcinoma of the ovary, hypercalcemic type (SCCOHT), and SMARCA4-deficient thoracic sarcomas (SA4DTS) (Le Loarer et al., 2015) (Figure 3B). Using an unbiased principal component analysis, we find that SS is transcriptionally distinct from other BAF complex-mutated sarcomas analyzed (Figures 3C and S3H). Correlating PC1(the principal component that distinguishes SS from other malignancies) with most variably expressed genes, we identified 346 genes specifically expressed in SS and 110 genes specifically expressed in other BAF complex-mutated cancers (Figure 3D). SS-specific genes include known markers such as TLE1 (Terry et al., 2007), and important regulators of neural development (i.e., SOX8, FOXC1) and myogenesis (i.e., PAX7) that we identified as primary SS18-SSX targets in cell lines (Figures 2J, S3I, and S3J). GO term analysis identified key regulators of developmental and mesenchymal programs as SS-specific genes (Figure 3E), further highlighting unique transcriptional pathways that are aberrantly activated in SS. These results establish that the SS transcriptional landscape is distinct from other BAF complex-mutated sarcomas and cancers.

Owing to the heterogeneous SS18-SSX-mediated gene activation observed in SS cell lines, we sought to characterize the degree to which this heterogeneity in gene expression is observed in primary samples. We performed NMF clustering on whole-transcriptome profiles of primary SS samples (n = 64) and found that these data best segregate into two subgroups as determined by the cophenetic coefficient and silhouette consensus (Figures 3F and S4A–S4C). We find that markers of this subgrouping include key myogenic genes PAX3 and PAX7, as well as putative oncogenes SOX2 and MYC (Figure 3G). In addition, we find mutual exclusivity in expression between MYC and PAX7 as well as MYC and SOX2, and find that these exclusivities correspond to the subgroupings (Figures 3H and S4D). To affirm these results, we characterized the protein expression of these putative markers in 186 patient-derived SS specimens (Barrott et al., 2016) (Figure S4E). We found that these markers are expressed at the protein level across SS tumors, and relate to one another in a manner similar to that observed at the RNA level (Figures 3I, 3J, and S4F), with 89.8% (167/186) of all SS specimens exhibiting expression of at least one of these four core markers. While we did not observe significant differences in survival based on the status of these markers between subgroups (Figure S4G), we did find that that expression of SOX2 and PAX7 is more frequently observed in primary cases, whereas MYC expression is more frequently observed in metastatic cases (Figure S4H). Heterogeneous expression of myogenic progenitor markers such as PAX7 and PAX3 suggest a window for SS18-SSX-mediated oncogenic activation during myogenic differentiation, which is consistent across cell lines and patient samples (Buckingham and Rigby, 2014).

SS18-SSX Directs BAF Complexes to Broad Polycomb-Repressed Domains to Activate Bivalent Genes

We next sought to characterize the mechanism by which SS18-SSX-containing BAF complexes mediate SS-specific gene activation. Given previous findings that BAF complexes target to SOX2 at which H3K27me3 levels are reduced in an SS cell line (Kadoch and Crabtree, 2013), we performed ChIP-seq for H3K27me3 as well as for the PRC2 subunit SUZ12 in Aska cells. Unexpectedly, we found a significant degree of co-occupancy between BAF complexes and PRC2 complexes genome wide at broad domains, which is exemplified at the promoters of PAX3 and SOX2 (Figures 4A and S5A). SS18-SSX knockdown results in retargeting of BAF complexes away from SUZ12 and H3K27me3 sites, suggesting that SS18-SSX is responsible for mediating this co-occupancy (Figures 4B and S5B). To confirm this unique, SS-specific pattern, we performed ChIP-seq for H3K27me3 and SUZ12 in human fibroblasts, and found that while WT SS18-marked BAF complexes exhibit minimal co-occupancy with PRC2, SS18-SSX targets BAF complexes to PRC2-marked sites (Figures 4C and S5C). These results demonstrate an unexpected co-occupancy of BAF complexes and PRC2 complexes that is mediated by SS18-SSX targeting, and is not a feature observed for WT BAF complexes (Ho et al., 2009).

Figure 4. SS18-SSX Directs BAF Complexes to Broad Polycomb Domains to Activate Bivalent Genes.

(A) Example tracks of co-occupancy of BAF complexes (SS18, BRG1, and BAF155) and PRC2 complexes (SUZ12 and H3K27me3), with activation marks (H3K4me3, RNA Pol II, and RNA-seq) at the PAX3 locus in Aska cells.

(B) Venn diagram of SS18 peaks in shCt and shSSX conditions with SUZ12 peaks in shCt condition in Aska cells.

(C) Venn diagram of V5 peaks in V5-SS18 WT and V5-SS18-SSX1 conditions with SUZ12 peaks in V5-SS18 WT condition in CRL7250 fibroblast cells.

(D) Metagene plots of Aska shCt-only SS18 sites (n = 8,610) split by broad (≥10 kb width) and narrow (<10 kb width) in shCt and shSSX conditions for SS18, SUZ12, H3K27me3, and H3K4me3 occupancy.

(E) Venn diagram of bivalent genes (dually marked by H3K4me3 and H3K27me3) in shCt and shSSX conditions in Aska cells.

(F) Breakdown of SS18-SSX primary target genes and secondary non-target genes by promoter status in shCt and shSSX conditions in Aska cells.

(G) Example SS18, BAF155, SUZ12, H3K27me3, H3K4me3, RNA Pol II, and RNA-seq tracks at the FOXC1, ZIC5, and ZIC2 loci shows SS18-SSX-mediated gene activation and that RNA Pol II recruitment is dependent on basal H3K4me3 levels.

See also Figure S5.

Given the hallmark breadth of SS18-SSX-containing BAF complex peaks, we sought to determine if there is a relationship between these unique sites and PRC2-mediated repression in SS. We divided shCt-only SS18 sites (SS18-SSX targets) into broad (>10 kb width, n = 791) and narrow (≤10 kb width, n = 7,819) peaks and performed metagene analysis for BAF and PRC2 complexes over these sites. While knockdown of SS18-SSX results in decreased BAF complex occupancy at both broad and narrow sites, we found marked increases in SUZ12 and H3K27me3 occupancy selectively at broad sites without changes in protein levels (Figures 4D, S1A, and S5D). The differential signals for SUZ12 and H3K27me3 are more modest compared with SS18, owing to the fact that not every SS18 broad site exhibiteddetectablePRC2 occupancy, possibly due to the substantially slower rate of heterochromatin formation by PRC2 (Kadoch et al., 2017); as such, the SUZ12 and H3K27me3 levels captured by ChIP-seq in the shSSX condition may not be fully restored at broad sites on the timescale of these experiments. Knockdown of SS18-SSX restored the expected anti-correlation between SUZ12 and gene expression that was not observed in the shCt condition, suggesting that SS18-SSX disrupts PRC2-mediated repression (Figure S5E). These results challenge previous suggestions that PRC2 is a dependency in SS (Kawano et al., 2016; Shen et al., 2016; Su et al., 2012), as treatment of Aska, SYO1, and HSSY2 cell lines with the EZH2 inhibitor EPZ005687 failed to inhibit cellular proliferation (Figures S5F and S5G). In contrast to PRC2 and its mark, our analysis showed no significant changes in H3K4me3 occupancy over either narrow or broad SS18 regions (Figure 4D), but showed marked decreases in RNA Pol II recruitment to both broad and narrow regions (Figure S5D), suggesting that the mechanism of SS18-SSX-mediated bivalent gene activation involves RNA Pol II recruitment rather than pause release (Williams et al., 2015). These results point toward BAF complexes as the critical mediator of polycomb opposition, while PBAF complexes are recruited downstream of BAF complex-mediated activation to active promoters within broad domains of BAF complex occupancy (Figures 1D and 1E). Taken together, these studies demonstrate that SS18-SSX-containing BAF complexes actively target and serve as the critical mediator of polycomb opposition at broad domains, leading to recruitment of RNA Pol II and gene activation.

Importantly, we found that suppression of SS18-SSX, and hence restoration of WT BAF complex composition and retargeting, resulted in a gain of repressed bivalent genes (dually marked by H3K4me3 and H3K27me3) (Figure 4E), owing to increases in H3K27me3, while H3K4me3 targets remain constant (Figure S5H). We found substantial increases in bivalent (and hence repressed) genes upon fusion knockdown––from 21.5% (62/289) to 45.3% (131/289)––the greatest increase of any promoter category, among primary SS18-SSX targets, while secondary genes did not exhibit similar changes (Figure 4F). Examination of all genes activated by SS18-SSX (i.e., those downregulated upon SS18-SSX knockdown) revealed that changes in gene expression strongly correlate with baseline H3K4me3 levels at gene promoters in the presence of SS18-SSX (Figure S5I), consistent with previous findings in settings of PRC2 loss (Jadhav et al., 2016). These results account for heterogeneous gene activation of SS18-SSX targets observed within and across cell lines, as demonstrated at the FOXC1 and ZIC2/5 loci (Figure 4G). Here, SS18-SSX-mediated BAF complex occupancy and PRC2 displacement occur at similar levels across these genes; however, H3K4me3 levels directly mirror the level of gene expression activation and RNA Pol II recruitment. Importantly, we found that variable H3K4me3 levels also accounted for heterogeneous activation of genes such as PAX7 and SOX2 (Figures S5J and S5K). These results demonstrate that SS18-SSX targets BAF complexes to broad polycomb domains to actively oppose polycomb-mediated repression, and subsequently recruit RNA Pol II to activate genes in a manner proportionate to H3K4me3 levels.

BAF47-Mediated Rescue of Enhancer Activation Is Not Required for Proliferative Arrest upon SS18-SSX Suppression

We have previously shown that SS18-SSX drives two concurrent changes in BAF complex composition: (1) the gain of 78aa of SSX to the SS18 subunit, and (2) destabilization of the BAF47 subunit, raising the possibility that these tumors are driven in a manner similar to MRT by loss of BAF47 (Figures 5A, S6A, and S6B) (Arnold et al., 2013; Kohashi et al., 2010). We sought to decouple the contributions of SS18-SSX knockdown and BAF47 reincorporation in the suppression of oncogenic gene expression and proliferation in SS. We used CRISPR-Cas9-mediated gene knockout to generate Aska cells lacking BAF47 (denoted BAF47Δ). In BAF47Δ−1 (clone no. 1) cells, no detectable BAF47 expression was recovered in shCt or shSSX conditions, compared with naive WT Aska cells (Figure 5B). This experimental system enables us to examine the effects SS18-SSX knockdown in the presence or absence of BAF47 restoration, and thereby decouple the role of each event in SS cell chromatin regulation, gene expression, and proliferation.

Figure 5. BAF47 Reincorporation Restores Enhancer Activation but Is Dispensable for Proliferative Arrest.

(A) Schematic depicting BAF complex perturbations in SS (SS18-SSX1/2/4) and malignant rhabdoid tumor (SMARCB1^−/−).

(B) Input blot of whole-cell extracts from Aska WT and BAF47-deficient (BAF47Δ) cells, in shCt and shSSX conditions.

(C) Venn diagram of SS18 peaks in shCt and shSSX conditions in Aska WT and BAF47Δ cells.

(D) Heatmaps of SS18 and H3K27ac ChIP-seq occupancy and ATAC-seq chromatin accessibility in Aska WT and BAF47Δ cells in shCt and shSSX conditions over all SS18 peaks in Aska WT cells (32,077) ranked by log2(fold change) in SS18 occupancy in Aska WT cells.

(E) Example SS18, H3K27ac, and ATAC-seq tracks at the CAV1 locus in Aska WT and BAF47Δ cells in shCt and shSSX conditions.

(F and G) Log2(fold change) in gene expression for Aska WT and two BAF47Δ clones in shCt and shSSX conditions over all significantly changed genes (2,350) in Aska WT day 7 (F) and highlighted example genes (G), ranked by mean fold change across all samples.

(H) Example SS18 and RNA-seq tracks at the HOXC cluster in Aska WT and BAF47Δ cells in shCt and shSSX conditions.

(I) Proliferation of Aska WT and BAF47Δ cells in shCt and shSSX conditions. Data are shown as means ±SD for n = 3 experiments: ***p <0.001 by two-tailed t test.

See also Figure S6.

We performed ChIP-seq for SS18 and H3K27ac, as well as ATAC-seq, in WT and BAF47Δ−1 Aska cells in shCt and shSSX conditions. We observed that SS18-SSX-mediated targeting to broad sites was abated upon knockdown of SS18-SSX in BAF47Δ cells in a manner similar to that in the WT cell setting (Figure 5C). In the WT context, SS18-SSX knockdown restored BAF complex targeting to distal sites to mediate accessibility and enhancer activation (Figures 5C, 5D, S6C, and S6D). Surprisingly, we found that, in the absence of BAF47, BAF complexes were not retargeted to these enhancer sites and that these sites remained inaccessible (Figures 5C, 5D, S6C, and S6D). Attenuated enhancer activation is exemplified at the CAV1 and VIM loci (Figures 5E and S6E). ChIP-seq for SS18 in CRL7250 fibroblasts affirmed that rescued enhancer sites are present in a WT cell setting (Figure S6F). These data align with the widespread BAF47-mediated enhancer activation observed in MRT experimental systems (Nakayama et al., 2017). Our results demonstrate that SS18-SSX hijacks BAF complex-mediated activation from enhancers to broad polycomb domains, demonstrating a dual gain and loss of gene activation that occurs uniquely in SS.

To determine the role for BAF47 in SS18-SSX-mediated gene expression, we performed RNA-seq on two BAF47Δ Aska clones, as well as WT Aska cells, at day 3 post-infection (as opposed to day 7 for all other samples), to observe early effects of SS18-SSX suppression (Figures S6G and S6H). We saw a global attenuation of expression changes in BAF47Δ clones as compared with WT Aska cells, with conserved directionality over all significantly changed genes (Figures 5F, 5G, S6I–S6K). Specifically, we found that expression changes in BAF47Δ clones (at day 7 post-SS18-SSX knockdown) best resemble the changes observed in WT Aska cells at day 3 post-SS18-SSX knockdown (Figures 5F, 5G, S6I–S6K). These results suggest that BAF47 plays a role in restoring WT gene regulation, likely by enabling active displacement and retargeting of oncogenic BAF complexes to normal sites upon SS18-SSX knockdown, as observed by ChIP-seq (Figure 5H).

Finally, we sought to determine if SS cells undergo proliferative arrest in the absence of BAF47 rescue. Importantly, proliferative arrest upon SS18-SSX knockdown was preserved in BAF47Δ Aska cells and this result was confirmed in BAF47Δ SYO1 cells, indicating that BAF47 rescue and BAF47-mediated enhancer activation are not required for proliferative arrest, as they are in MRT (Figures 5I, S6M, and S6N). These results demonstrate that proliferative arrest upon suppression of SS18-SSX expression is not dependent on BAF47 rescue, enabling functional decoupling of the roles for SS18-SSX loss and BAF47 rescue in the genome-wide retargeting observed in SS.

DISCUSSION

We demonstrate that a global hijacking event underlies synovial sarcomagenesis, with SS18-SSX directing BAF complexes away from enhancers to activate broad, polycomb-repressed domains, leading to oncogenic activation of bivalent genes (Figure 6A). SS18-SSX-mediated eviction of BAF47 results in enhancer decommissioning, but rescue of these enhancers upon SS18-SSX knockdown is dispensable for proliferative arrest, suggesting that the key mechanism driving SS is the oncogenic gene activation at bivalent genes by the highly unique genome-wide targeting of BAF complexes by the SSX 78 amino acids (Figure 6B). These data present evidence that BAF complexes can be co-opted from one activating role to another, providing a broadly relevant mechanism by which fusion oncoprotein- and transcription factor-mediated recruitment can direct BAF complexes in development and disease.

Figure 6. Model for SS18-SSX-Mediated Hijacking of BAF Complexes in SS.

(A) SS18-SSX directs BAF complexes to broad polycomb domains at which they oppose polycomb-mediated repression to activate bivalent genes in a manner dependent on basal H3K4me3 levels. Upon suppression of the SS18-SSX fusion oncoprotein, BAF complexes return to distal sites, mediating enhancer (and corresponding gene) reactivation.

(B) In the absence of BAF47, SS18-SSX suppression results in similar restoration of gene silencing at broad polycomb domains, as well as proliferative arrest;however, enhancer accessibility and activation are not restored.

We previously showed that SS18-SSX incorporation results in destabilization of BAF47 from BAF complexes (Kadoch and Crabtree, 2013); however, the relative contribution of SS18-SSX incorporation and BAF47 destabilization was unclear. Our results demonstrate that gained targeting by SS18-SSX is the critical event in SS, and that SS18-SSX targets BAF complexes to broad polycomb domains to activate bivalent genes in a highly cancer-specific manner. We observe that restoration of BAF47 mediates enhancer activation, yet this is dispensable for SS cell proliferative arrest compared with SS18-SSX suppression, suggesting that gain-of-function activation by BAF complexes is the critical mechanism underlying oncogenic gene expression and proliferation. Interestingly, these results are opposite to those observed in BAF47-deficient sarcomas (Nakayama et al., 2017), suggesting that the utility of EZH2 and other PRC2 inhibitors is not translatable to SS, as has begun to be appreciated in the clinic (Schoffski et al., 2017).

Our results demonstrate a mechanism in which a trithorax family protein complex activates bivalent genes in a cancer-specific manner (Kennison and Tamkun, 1988; Schuettengruber et al., 2017). We show that SS18-SSX expression mirrors PRC2 loss at bivalent genes, driving gene activation through opposition of PRC2 and its deposition of H3K27me3, in a manner dependent on the underlying H3K4me3 levels (Jadhav et al., 2016). Consequently, this suggests that altering H3K4 methylation levels may modulate SS18-SSX-mediated activation of bivalent genes and provides the motivation to explore inhibiting H3K4me3 placement by MLL2/COMPASS complexes to silence oncogenic gene expression in SS (Hu et al., 2013, 2017). These broad PRC2-marked domains have been identified as critical regulators of neural crest development, during which decreases in PRC2 causes activation of bivalent genes (Minoux et al., 2017). Our results suggest that these broad domains may be a hallmark to the SS cell of origin and therefore not observed in fibroblasts, providing a potential explanation for the specific developmental state(s) permissive of SS18-SSX-mediated oncogenesis.

Previous studies have shown that an increase in bivalent genes occurs during myogenic differentiation, and that H3K4me3 levels attenuate upon aging of muscle stem cells (Liu et al., 2013). Our results suggest that SS18-SSX-targeted BAF complexes undermine the gain in H3K27me3 that occurs during myogenic differentiation, and implicate a window of susceptibility in myogenic progenitor cells undergoing active proliferation during which SS18-SSX activates neural and embryonic genes to induce oncogenic proliferation in a manner akin to neural crest migration, leading to tumor development. In addition, following SS18-SSX knockdown, we did not observe increases in terminal myogenic marker expression, but did find activation of some genes implicated in myogenesis. Several myogenic genes, such as PAX7 and PAX3, also play critical roles in neural crest migration as well as mesoderm specification, indicative of a developmental history for the tissue of origin that is primed to active neural and embryonic gene set. This suggests that SS18-SSX suppression returns SS cells to a quiescent intermediary that is not indicative of terminal differentiation, such as is the case in rhabdomyosarcoma, in which the PAX-FOXO fusion co-opts terminal differentiation transcription factors (Gryder et al., 2017). This window of susceptibility within myogenic differentiation permissive of SS18-SSX-induced tumor formation suggests a molecular basis for the transcriptional heterogeneity among fusion gene targets in primary tumors and is important to consider in the identification of therapeutically relevant biomarkers in this disease.

SS18-SSX-directed targeting of BAF complexes to broad polycomb domains provides strong motivation to identify the physical features of the interaction between the SSX 78aa tail and specific features of the histone landscape. Our results suggest that SSX is targeted to a mark or protein that is concomitant with broad PRC2 domains, which may inform therapeutic vulnerabilities via disruption of this oncogenic recruitment. Previous observations of polycomb compaction domains at HOX clusters (Kundu et al., 2017), and our observed SS18-SSX-mediated activation of HOX genes suggest that the SSX tail may facilitate BAF complex entry into discrete polycomb regions to enable chromatin engagement and gene activation, in a similar manner to rapamycin-induced recruitment to the Oct4 locus (Kadoch et al., 2017). Further work is needed to characterize precisely how SSX targets BAF complexes to these domains; however, we demonstrate interdependent roles of SSX targeting and the endogenous chromatin landscape in SS18-SSX-mediated oncogenic gene activation.

In summary, our results demonstrate that SS18-SSX hijacks BAF complexes from enhancers to broad polycomb domains, at which active opposition of polycomb-mediated repression drives expression of bivalent genes. These studies differentiate SS18-SSX from any other BAF complex-mediated cancer, including EWS-FLI1 fusion in Ewing sarcoma, which targets BAF complexes selectively to GGAA repeat enhancers (Boulay et al., 2017). Results in these disease contexts demonstrate that polycomb opposition and enhancer activation can each be driving in malignancy. Furthermore, our results suggest that one function of BAF complexes can be co-opted to facilitate another, presenting an opportunity to decompose the contribution of each specific function in mediating both oncogenesis as well as differentiation and development.

STAR★METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Cigall Kadoch (cigall_kadoch@dfci.harvard.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell Lines and Cell Culture

Six SS cell lines, two MRT cell lines, one EpS cell line, and four cell lines without BAF perturbations were used in this study (Table S3). The Aska and Yamato cell lines were a generous gift from Kazuyuki Itoh, Norifumi Naka, and Satoshi Takenaka (Osaka University, Japan), the YaFuSS cell line was a gift from Junya Toguchida (Kyoto University, Japan), the SYO1 cell line was a gift from Akira Kawai (National Cancer Center Hospital, Japan), and the HSSY2 cell line was purchased from RIKEN. The CRL7250 human fibroblast cell line was a kind gift from Drs. Berkeley Gryder and Javed Khan (National Cancer Institute, Bethesda, MD). All cell lines were cultured in DMEM medium (Gibco) supplemented with 10–20% fetal bovine serum, 1% Glutamax (Gibco), 1% Sodium Pyruvate (Gibco) and 1% Penicillin-StreptoMYCin (Gibco) and maintained in a humidified incubator at 37°C with 5% CO₂.

Human Data and Tumor Data Selection

Molecular data were obtained from patients in three manners: 1). Two institutions (MD Anderson and NCI) provided synovial sarcoma (n=64 for RNA-seq and n=18 for WES) and epithelioid sarcoma (n=6 for RNA-seq) sequencing data. The samples and anonymous clinical data used were obtained and analyzed under protocols approved by the local Institutional Review Board (IRB) or exempted by the HHS Office of Human Subjects Research Protections determination per NIH policy. 2). Data from ten synovial sarcoma cases for WES was obtained from The Cancer Genome Atlas Project (TCGA) (Cancer Genome Atlas Research Network, 2017). Local Institutional Review Boards (IRBs) at the TCGA tissue source sites reviewed protocols to approve submission of cases. TCGA Project Management collected necessary human subjects documentation to ensure the project complies with 45-CFR-46 (the ‘‘Common Rule’’). The program has obtained documentation from every contributing clinical site to verify that IRB approval has been obtained to participate in TCGA. 3). Other tumor sequencing data was obtained as publically available resources as indicated in the Method Details section.

Synovial Sarcoma Tissue Microarray (TMA)

Formalin-fixed, paraffin-embedded synovial sarcoma specimens collected at UTMDACC between 1988 and 2014 were reviewed for inclusion on the TMA (Barrott et al., 2016). The samples and anonymous clinical data used from MD Anderson were obtained and analyzed under protocols approved by the local Institutional Review Board (IRB). A single paraffin block and associated H&E were selected for each patient (n=192). Two adjacent 1.0 mm cores from each sample were placed into a recipient block using an ATA-100 Advanced Tissue Arrayer (Chemicon International, Temecula, CA).

METHOD DETAILS

Gene Knockdown and Stable Gene Expression Constructs

Constitutive expression of shRNA hairpins targeting SS18-SSX (1: 5’-AGAAAGCAGCTGGTGATTTAT-3’, 2: 5’-CAGTCACTGACAGT TAATAAA-3’) or a scramble non-targeting control (5’-CCTAAGGTTAAGTCGCCCTCGCTCGAGCGAGGGCGACTTAACCTTAGG-3’) was achieved using lentiviral infection of the pLKO.1 vector, selected with puromycin (2 μg/ml). Constitutive expression of SS18 wild-type (SS18) and SS18-SSX1 with V5 N-terminus tag in CRL7250 cell line was achieved using lentiviral infection of an EF1alpha-driven expression vector (modified from Clonetech, dual Promoter EF-1a-MCS-PGK-Puro), selected with puromycin (2 μg/ml).

Lentiviral Generation

Lentivirus was produced by PEI (Polysciences) transfection of HEK293T LentiX cells (Clontech) with gene delivery vector co-transfected with packaging vectors pspax2 and pMD2.G as previously described (Kadoch and Crabtree, 2013). Viral supernatants were harvested 72 hr post-transfection and concentrated by ultracentrifugation at 20,000 rpm for 2.5 hr at 4°C. Virus containing pellets were resuspended in PBS and added dropwise to cells in the presence of 10 μg/mL polybrene. Selection of lentivirally-infected cells was achieved with puromycin used at 2 μg/ml. Knockdown efficiency and overexpression was measured by western blot analysis.

Western Blot Analysis

Standard protocols were used for the detection of proteins by immunoblot (IB) analysis and the primary antibodies used are listed in Table S4. Membranes were developed using IRDye (LI-COR Biosciences) secondary antibodies for visualization by the LI-COR Odyssey Imaging System (LI-COR Biosciences).

Cell Lysate Collection

Whole cell extract preparations were performed by washing trypsinized cells with PBS pH 7.4, resuspending in whole cell lysis buffer (PBS pH 7.4 and 1% SDS) and heating at 95°C for 3 minutes. Nuclear extract (NE) preparation and immunoprecipitation (IP) studies were performed as described previously (Ho et al., 2009). Briefly, the trypsinized cells were incubated in Buffer A (25 mM HEPES pH 7.6, 5 mM MgCl₂, 25 mM KCl, 0.05 mM EDTA, 10% glycerol and 0.1% NP40 with protease inhibitor, 1 mM DTT and 1 mM phenylmethylsulfonyl fluoride (PMSF)) for 10 minutes and the pellets were resuspended in 600 μl of Buffer C (10 mM HEPES pH 7.6, 3 mM MgCl₂, 100 mM KCl, 0.5 mM EDTA and 10% glycerol with protease inhibitor, 1 mM DTT and 1 mM PMSF) with 67 μl of 3 M (NH₄)₂SO₄ for 20 minutes. The lysates were spun down using an ultracentrifuge at 100,000 rpm at 4°C for 10 minutes. Nuclear extracts were precipitated with 200 μg of (NH₄)₂SO₄ on ice for 20 minutes and finally purified as pellets by ultracentrifugation at 100,000 rpm at 4°C for 10 minutes. The pellets were resuspended in IP Buffer (300 mM NaCl, 50 mM Tris-HCl pH 7.5, 1 mM EDTA and 1% Triton-X100 with protease inhibitor, 1 mM DTT and 1 mM PMSF) for the subsequent experiments. The details of the antibodies used for immunoblotting and IP experiments are presented in Table S4.

Immunoprecipitations

Nuclear extracts were quantified using the Bradford assay and 200 μg of protein was incubated with 2.5 μg of antibody in a final volume of 200 μl of IP Buffer overnight at 4°C. Each sample was then incubated with Protein G Dynabeads (Thermo Scientific) for two hours. Beads were washed three times with IP buffer and twice with BC100 (20 mM HEPES, 100 mM KCl, 0.2 mM EDTA, 10% Glycerol), and eluted with 20 μl of sample buffer (NuPage LDS buffer (1×) (Life Technologies) containing 100 mM DTT and water). All antibodies used are presented in Table S4.

CRISPR-Cas9 Mediated Gene Knockout of SMARCB1

The SMARCB1 locus was targeted by the Ini1 CRISPR/Cas9 KO Plasmid and Ini1 HDR Plasmid (Santa Cruz Biotechnology sc-423027; sc-423027-HDR) in Aska cells following the manufacturer’s protocol. Specifically, two million cells were co-electroporated with two plasmids (2 μg DNA/plasmid) using the Amaxa Cell Line Nucleofector Kit V. After nucleofection, cells were expanded for 48 hours and GFP⁺/RFP⁺ cells expressing both the KO/HDR plasmids were single-cell sorted using FACS (fluorescence-activated cell sorting). Single-cell clones were expanded and screened using immunoblot detection for identification of successful knockouts.

Cell Proliferation Assay

To measure cell proliferation following shRNA-mediated knockdown, 1×10⁴ cells per well were seeded in 6-well plates following 48-hour exposure to the lentivirus and 4-day selection with 2 μg/ml of puromycin, with Day 6 denoting the day cells were plated after infection and selection. The cellular confluence in three wells was then monitored using IncuCyte Live Cell Analysis System (Essen Bioscience, Ann Arbor, MI) every 24-hours post seeding. Alternatively, 4×10³ cells per well (96-well plates) were seeded at Day 7 post-infection or to measure cell proliferation following EPZ005687 compound treatment, 4×10³ cells per well (96-well plates) or 4×10⁴ cells per well (12-well plates) were seeded and then treated with 0–10 μM of EPZ005687 (Cayman Chemical) prepared in DMSO for 7–14 days. The cell viability was measured by quantifying ATP contents using the CellTiter-Glo Luminescence Cell Viability Assay (Promega) with the SpectraMax M5 Microplate Reader (Molecular Devices, San Jose, CA) or Vi-CELL Cell Counter (Beckman, Brea, CA). Statistical analyses were performed by Student’s t test.

Immunohistochemistry of Synovial Sarcoma Tissue Microarray (TMA)

Immunohistochemical analysis of BAF47/MYC/SOX2/PAX7/PAX3 expression was performed as previously described (Charville et al., 2016). Anti-BAF47 antibody (1:100 dilution) was applied following antigen retrieval using pH 8.0 EDTA buffer. Anti-MYC (1:900 dilution) and anti-SOX2 (1:75 dilution) antibodies were applied following antigen retrieval using 0.01 M citrate buffer (pH 6.1). Anti-PAX7 (1:200 dilution) and anti-PAX3 (1:50 dilution) antibodies were applied following antigen retrieval using 0.01 M citrate buffer (pH 6.0). All antibodies are listed in Table S4. Immunoreactivity was considered positive if >10% of tumor cells showed staining with appropriate nuclear chromogen localization.

Chromatin Immunoprecipitation (ChIP)

For chromatin immunoprecipitation (ChIP) experiments, indicated cells were harvested following 48 hr exposure to specified lentivirus and 5 day selection with 2 μg/ml of puromycin. ChIP experiments were performed per standard protocols (Millipore, Billerica, MA) with minor modifications. Briefly, cells were cross-linked for 10 min with 1% formaldehyde at 37°C. This reaction was subsequently quenched with 125 mM glycine for 5 min and 5–10 million fixed cells were used per ChIP experiment. Chromatin from fixed cells was fragmented by sonication with a Covaris E220 and the solubilized chromatin was incubated with the indicated antibody (listed in Table S4) overnight at 4°C. Antibody-chromatin complexes were pulled down by incubation with Protein G-Dynabeads (Thermo Scientific) for 3 hours at 4°C, washed and eluted. The samples then underwent crosslink reversal, RNase A (Roche) treatment, and proteinase K (Thermo Scientific) treatment before the captured DNA was extracted with AMP Pure beads (Beckman Coulter).

RNA Isolation from Cell Lines

Cells were harvested following 48 hr exposure to the lentivirus and either 1 day (day 3 post-infection) or 5 day (day 7 post-infection) selection with 2 μg/ml of puromycin for RNA-seq experiments. Samples for RNA-seq were prepared in biological duplicate (indicating independent lentiviral production, infection, selection, and cell culture). All RNA was collected using the RNeasy Mini Kit (Qiagen).

RNA Isolation from Tumor Samples

Fresh frozen tumor tissue was disrupted using BioMasher II Micro Tissue Homogenizers (DWK Life Sciences, Millville, NJ) and homogenized with syringes and 21-guage needles. RNA was extracted from tissue lysate using the Animal Tissues protocol from the RNeasy Mini kit or the AllPrep DNA/RNA/miRNA Universal kit (Qiagen), with final elution volumes of 50 μL.

Assay for Transposase-Accessible Chromatin Sequencing (ATAC-seq)

Cells were harvested following 48-hour exposure to lentivirus and 5 days of selection with 2 μg/ml of puromycin and ATAC-seq experiments were performed as previously described (Buenrostro et al., 2015). Samples for ATAC-seq were prepared in biological duplicate (indicating independent lentiviral production, infection, selection and cell culture) and prepared using the Nextera DNA sample Prep Kit (Illumina). Briefly, 5×10⁴ cells were incubated in hyptotonic buffer and lysis buffer, then were resuspended in Transpoase reaction mixture for 30 min at 37°C with gentle shaking followed by DNA purification.

Library Preparation, Sequencing, and Data Processing for ChIP, RNA and ATAC Samples

Library preparations for ChIP-seq and RNA-seq samples were performed in the Molecular Biology Core Facilities at the Dana-Farber Cancer Institute using the ThruPLEX DNA-seq Kit (Rubicon Genomics) or TruSeq Stranded mRNA Kit (Illumina), respectively. Sequencing was performed on an Illumina Nextseq 500 with 75bp single end for ChIP-seq and RNA-seq, and 75bp paired end for ATAC-seq.

ChIP-seq reads were mapped to the human reference genome (hg19) using Bowtie2 (Langmead and Salzberg, 2012) version 2.1.0 with parameters –k 1. Summary statistics for ChIP-seq experiments are presented in Table S5. RNA-seq reads were mapped to the human reference genome (hg19) using STAR (Dobin et al., 2013) version 2.3.1 with default parameters. Reads were converted to BAM format using samtools version 0.1.19 (Li et al., 2009).

ATAC-seq reads were processed to the human reference geome (hg19) as previously described (Buenrostro et al., 2013). Briefly, paired-end reads were trimmed to 30bp and paired using Trimmomatic 0.35 (Bolger et al., 2014), and then mapped to hg19 using Bowtie2 version 2.1.0 with parameter -X2000 (Langmead and Salzberg, 2012). Mapped reads were subsequently filtered for duplicate reads using Picard MarkDuplicates (http://broadinstitute.github.io/picard). Reads were filtered to remove reads that overlap with ENCODE blacklisted regions (ENCODE Project Consortium, 2012).

ChIP-Seq Data Analysis

Peaks were called against input reads using MACS2 (Zhang et al., 2008) version 2.1.0 at q=1e-3. Broad peak calls were used for all marks in this study. Peaks were filtered to remove peaks that overlap with ENCODE blacklisted regions (ENCODE Project Consortium, 2012), as well as peaks mapped to unmappable chromosomes (only chr1–22,X,Y included). Duplicate reads were removed using samtools rmdup for all downstream analyses (Li et al., 2009). ChIP-seq track densities were generated per million mapped reads with MACS2 2.1.0 using parameters –B –SPMR.

BAF complex sites were determined in Aska and SYO1 using the merged peak set for SS18 in shCt and shSSX conditions, using bedtools merge –d 2000 to ensure nearby broad peak regions were merged into one peak (Quinlan and Hall, 2010). In CRL7250, the same methodology was used as above but for V5 peaks in V5-SS18 and V5-SS18-SSX1 conditions. Venn diagrams were generated using the R statistical package, using the minimum number of overlapping regions for resolving multiple peak overlaps.

Metagene read densities were generated using HTSeq (Anders et al., 2015), with fragment length extended to 200bp to account for the average 200bp fragment size selected in sonication. Total read counts for each region was normalized by the number of mapped reads to give reads per million mapped reads. Metagene plots were generated using average read densities across all sites indicated for each condition—narrow metagene plots were generated around the center of the peak, whereas broad metagene plots used the full peak window, with 5kb flanking regions on each side of the peak included in visualization. Heatmaps were generated using the same HTSeq read densities as in metagene plots, sites were then ranked by mean ChIP-seq signal for the epitope and condition indicated in each figure. Heatmaps were visualized using Python matplotlib with a midpoint of 0.5 reads per million for the heatmap color scale to set the threshold for visualization.

Condition-specific BAF complex peaks were determined as below. HTSeq read densities were generated using the merged peak sets as described above over a 10kb window (SS18 for Aska, SYO1; V5 for CRL7250). From this, average RPM values over each peak-centered window were used to calculate log2FC values between conditions, and condition-specific sites were determined using a 1.5FC cutoff for specificity of condition. This was used instead of venn diagram peak calls to account for effect size of retargeting, and account for regions with depleted BAF complex occupancy but a peak called. Peak widths were determined from this merged peak set, with condition-specific peaks determined as above.

Distance to TSS for ChIP-seq peaks was determined using BEDTools closest function with hg19 refFlat TSS annotation, with small RNA genes (MIR and SNO) removed. Target genes were determined using TSS sites within 2kb of a peak. GO Term analysis was performed on the target gene sets using biological processes annotation (Gene Ontology Consortium, 2015), with a significance threshold of 1e-3.

Cell Line RNA-seq Data Analysis

RPKM values for samples were generated using GFold (Feng et al., 2012) version 1.1.0. All error bars represent Mean±SEM. Significance was assessed using the R package DESeq2 (Love et al., 2014) using raw read counts generated with Rsubread featureCounts against the hg19 refFlat annotation. Small RNA genes (MIR and SNO) were removed from all downstream analyses. Significantly changing genes were assessed with a Bonferri-corrected p value of less than 1e-3, a two-fold gene expression change (|log2FC|>1), and filtered for expressed genes (RPKM ≥ 1 in at least one condition of comparison) to determine set of significantly changing genes. For heatmap analysis of RNA-seq, log2 fold change values were plotted as generated for GSEA analyses below, or z-scores of RPKM values were plotted across all samples visualized. RNA-seq tracks were generated using bedtools genomecov –split –scale with the mapped read count to generate tracks normalized per million mapped reads.

GSEA was performed using the GSEA Preranked function of the JAVA program (http://www.broadinstitute.org/gsea) as described previously (Subramanian et al., 2005). Rank files for GSEA were generated using RPKMs for duplicate RNA-seq in each cell line, removal of short RNAs, filtering for expressed genes (minimum RPKM value for four samples >= 1), averaging replicates of each condition, then doing a log2 fold change comparison with a pseudocount of 1 in each condition, i.e. log2( ( RPKM^shSSX + 1)/(RPKM^shCt + 1) ). Log2 fold change values for RNA-seq in figures were identical to those used for GSEA analysis except non-expressed genes were included in the analysis. RPKM values for all cell lines samples generated in this study can be found in Table S2.

Tumor RNA-seq Data Analysis

Tumor RNA-seq data sets for SMARCA4-deficient sarcomas (SA4DTS), small cell carcinoma of the ovary hypercalcemic type (SCCOHT), malignant rhabdoid tumor (MRT), renal medullary carcinoma (RMC) and epithelioid sarcoma (2 of 8 samples, six were sequenced in this study) were published previously (Le Loarer et al., 2015). RPKM values for all tumor data was generated as above. Principal component analysis was performed on log2(RPKM+1) normalized expression values, using top 5% most variable genes for PCA clustering. To determine SS and MRT gene sets, genes in top 5% most variable were correlated with PC1, and genes with > 0.5 Pearson correlation coefficient (PCC) were visualized and analyzed by GO term analysis. Heatmap of PC1-correlated genes is ranked by PCC values of each gene. t-SNE was performed using the same expression set as with PCA analysis, using default parameters of the Rtsne package. For clustering of SS tumor expression data, NMF clustering was performed for 2 to 10 clusters. Cophenetic coefficient and silhouette consensus were used to determine n=2 as number of groups for best fit. Two-tailed t tests were used to determine significance of differential expression between subgroups of synovial sarcoma tumor expression data.

Fusion breakpoints were determined from RNA-seq of patient samples using exon-level expression of SSX1/2, as SSX genes are exclusively expressed in the SS18-SSX fusion. Expression pre- and post-fusion breakpoint was determined using custom reference annotation splitting refFlat annotation for SSX1/2 into pre/post-fusion breakpoint and calculating RPKM values using GFold. RPKM values for all patient samples collected in this study can be found in Table S2.

ATAC-seq Data Analysis

Biological duplicate samples were merged for all downstream analyses. Tracks of ATAC-seq were generated using bedtools genomeCoverageBed with parameters -pc -bg and scaled to be per million reads using map counts generated by samtools view function with parameters -c -f 67.

DNA Isolation and Whole Exome Sequencing of Tumors

DNA was extracted (< 30 mg) using the Frozen Tissue protocol from the QIAamp DNA Mini kit (Qiagen, Inc., Valencia, CA), with a final elution volumes between 60–100 μL. Samples were then submitted for SureSelect Human All Exon V4 library preparation (Agilent Technologies, Inc., Santa Clara, CA) and subsequent 76 bp short-read whole exome sequencing on Illumina HiSeq 2000 (Illumina, Inc., San Diego, CA). Mapping of reads to the hg19 reference genome using BWA-MEM (Li, 2013) was followed by somatic variant calling using matched peripheral blood or adjacent tissue (MuTect and Pindel) (Cibulskis et al., 2013; Ye et al., 2009). High quality variants were defined as those with a minimum tumor read depth of ≥ 20, minimum matched normal read depth of ≥ 10, and minimum alternate allele frequencies in the tumor and normal as ≥ 0.01 and ≤ 0.01, respectively. Mutation calls are presented in Table S1.

Whole Exome Data Analysis

Tumor data for other tumor types was acquired from TCGA or Broad Institute (Cancer Genome Atlas Research Network, 2011; Lawrence et al., 2013). Mutation rates were determined for each case in each tumor type, determining the number of nonsynonymous mutations in each case, and dividing by 45Mb to account for estimated length of exome in the human genome (1.5% of 3Gb). SS cases from TCGA (n=10) were combined with cases sequenced in this study (n=18) to determine SS mutation rates. Median mutation rates for each tumor type are denoted. The results published here are in whole or part based upon data generated by The Cancer Genome Atlas managed by the NCI and NHGRI. Information about TCGA can be found at http://cancergenome.nih.gov.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical comparisons between two groups for proliferation analyses were performed with GraphPad Prism software 7.0 using a two-tailed unpaired t-test. The sample size (n) is indicated in the figure legends and represents biological replicates. Details for sequence data analyses and statistical significance are described in the specific Method Details section.

DATA AND SOFTWARE AVAILABILITY

The accession number for the sequence data for all cell line experiments and deidentified SS (NCI) patient RNA-seq samples is GSE108028. The accession number for the sequence data for patient-associated samples of SS and EpS (MDA) is EGA: EGAS00001002920.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse Anti-SMARCA4 (BRG1) (G-7) (IB)	Santa Cruz	Cat#sc-17796; RRID:AB_626762
Rabbit Anti-SMARCA4 (BRG1) (D1Q7F) (IP)	Cell Signaling Technology	Cat#49360; RRID:AB_2728743
Rabbit Anti-SMARCA4 (BRG1) (J1) (ChIP)	Homemade	N/A
Mouse Anti-INI1 (BAF47) (A-5) (IB)	Santa Cruz	Cat#sc-166165; RRID:AB_2270651
Mouse Anti-INI1 (BAF47) (IHC)	BD Biosciences	Cat#612111; RRID:AB_2191717
Rabbit Anti-SMARCC1 (BAF155) (H-76) (IB)	Santa Cruz	Cat#sc-10756; RRID:AB_2191997
Rabbit Anti-SMARCC1 (BAF155) (155–7) (ChIP)	Homemade	N/A
Rabbit Anti-SS18 (D6I4Z) (IB & ChIP)	Cell Signaling Technology	Cat#21792S; RRID:AB_2728667
Mouse Anti-ARID1A (BAF250A) (C-7) (IB)	Santa Cruz	Cat#sc-373784; RRID:AB_10917727
Rabbit Anti-ARID1A (BAF250A) (IP)	Bethyl	Cat#A301–041; RRID:AB_2060365
Mouse Anti-ARID2 (BAF200) (E-3) (IB)	Santa Cruz	Cat#sc-166117; RRID:AB_2060382
Rabbit Anti-ARID2 (BAF200) (PP12) (IP & ChIP)	Cell Signaling Technology	Cat#P62240 (test antibody)
Rabbit Anti-SUZ12 (D39F6 XP(R)) (ChIP)	Cell Signaling Technology	Cat#3737S; RRID:AB_2196850
Rabbit Anti-H3 (IB)	Abcam	Cat#Ab1791; RRID:AB_302613
Rabbit Anti-H3K27me3 (IB & ChIP)	Millipore	Cat#07–449; RRID:AB_310624
Rabbit Anti-H3K4me3 (15–10C-E4) (ChIP)	Millipore	Cat#05–745R; RRID:AB_1587134
Rabbit Anti-H3K27Ac (ChIP)	Abcam	Cat#ab4729; RRID:AB_2118291
Mouse Anti-GAPDH (G-9) (IB)	Santa Cruz	Cat#sc-365062; RRID:AB_10847862
Mouse Anti-V5 tag (IB)	Thermo Fisher Scientific	Cat#P/N-46–0705; RRID:AB_2556564
Rabbit Anti-V5 tag (D3H8Q) (ChIP)	Cell Signaling Technology	Cat#13202; RRID:AB_2687461
Mouse Anti-RNAPolII (ChIP)	Diagenode	Cat#C15200004; RRID:AB_2728744
Rabbit Anti-IgG (IP)	Cell Signaling Technology	Cat#2729S; RRID:AB_1031062
Rabbit Anti-MYC (IHC)	Abcam	Cat#Ab32072; RRID:AB_731658
Rabbit Anti-SOX2 (D6C9) (IHC)	Cell Signaling Technology	Cat#3579; RRID:AB_2195767
Mouse Anti-PAX7 (IHC)	Developmental Studies Hybridoma Bank	RRID:AB_2299243
Mouse Anti-PAX3 (IHC)	R&D Systems	Cat#MAB2457; RRID:AB_2159398
Biological Samples
Primary Tumor Samples	MD Anderson & NCI	N/A
Human synovial sarcoma tissue microarray	Barrott et al., 2016	N/A
Chemicals, Peptides, and Recombinant Proteins
Puromycin	Sigma-Aldrich	Cat#P8833–25MG
Blasticidin	Life Technologies	Cat#R210–01
EPZ005687	Cayman Chemical Company	Cat#1396772–26–1
Dimethyl sulfoxide	Sigma-Aldrich	Cat#D2650
DMEM, high glucose, no glutamine	Life Technologies	Cat#11960–069
PBS, pH 7.4	Life Technologies	Cat#10010–049
GlutaMAX	Life Technologies	Cat#35050–079
Trypsin-EDTA (0.25%), phenol red	Life Technologies	Cat#25200–114
Sodium Pyruvate	Life Technologies	Cat#11360–070
Penicillin-Streptomycin	Life Technologies	Cat#15140–163
Polybrene	Santa Cruz Biotechnology	Cat#sc-134220
Polyethylenimine (PEI) (MW 40,000)	Polysciences	Cat#24765
Goat Anti-Mouse IgG Antibody, IRDye 680RD Conjugated	LI-COR Biosciences	Cat#926–68070
Goat Anti-Rabbit IgG Antibody, IRDye 800CW Conjugated	LI-COR Biosciences	Cat#926–32211
HEK293T LentiX	Clontech	Cat#632180
Dynabeads Protein G	Thermo Fisher Scientific	Cat#10004D
NuPage LDS Sample Buffer (4X)	Life Technologies	Cat#NP0007
Formaldehyde	Sigma-Aldrich	Cat#F8775
Glycine	Sigma-Aldrich	Cat#G7126
RNase	Roche	Cat#11119915001
Proteinase K	Thermo Fisher Scientific	Cat#AM2546
Agencourt AMPure XP	Beckman Coulter	Cat#A63882
Critical Commercial Assays
Amaxa Cell Line Nucleofector Kit V	Lonza	Cat#VCA-1003
CellTiter-Glo Luminescent Cell Viability	Promega	Cat#G7571
RNeasy Mini Kit	Qiagen	Cat#74104
AllPrep DNA/RNA/miRNA Universal Kit	Qiagen	Cat#80224
QIAamp DNA Mini Kit	Qiagen	Cat#51304
Nextera DNA Sample Preparation Kit	Illumina	Cat#FC-121–1031
ThruPLEX DNA-seq Kit	Rubicon Genomics	Cat#R400407
TruSeq Stranded mRNA Kit	Illumina	Cat#20020595
SureSelect Human All Exon V4	Agilent	Cat#5190–4671
Deposited Data
ChIP-seq, RNA-seq, and ATAC-seq from cell lines	This study	GEO: GSE108028
RNA-seq of deidentified tumor samples (NCI)	This study	GEO: GSE108028
RNA-seq and whole exome sequencing of patient-associated tumor samples (MD Anderson)	This study	EGA: EGAS00001002920
Whole exome sequencing from TCGA and Broad studies	(Cancer Genome Atlas Research Network, 2011; Lawrence et al., 2013)	http://firebrowse.org/ http://tumorportal.org
RNA-seq of BAF complex-perturbed cancers	(Le Loarer et al., 2015)	GEO: SRP052896
Experimental Models: Cell Lines
CRL7250	ATCC	CRL-7250
Aska	Naka et al., 2010	N/A
SYO1	Kawai et al., 2004	N/A
HSSY2	Sonobe et al., 1992	N/A
Yamato	Naka et al., 2010	N/A
YaFuSS	Ishibe et al., 2005	N/A
Fuji	Nojima et al., 1990	N/A
Oligonucleotides
shRNA targeting sequence: SSX #1 (5’-AGAAAGCAGCTGGTGATTTAT-3’)	This study	N/A
shRNA targeting sequence: SSX #2 (5’-CAGTCACTGACAGTTAATAAA-3’)	This study	N/A
shRNA targeting sequence: Scramble control (5’-CCTAAGGTTAAGTCGCCCTCGCTCGAGCGAGGGCGACTTAACCTTAGG-3’)	This study	N/A
Recombinant DNA
EF-1a-MCS-PGK-Puro-V5-SS18 WT	Kadoch and Crabtree, 2013	N/A
EF-1a-MCS-PGK-Puro-V5-SS18-SSX1	Kadoch and Crabtree, 2013	N/A
Ini1 (BAF47) CRISPR/Cas9 KO Plasmid	Santa Cruz Biotechnology	Cat#sc-423027
Ini1 (BAF47) HDR Plasmid	Santa Cruz Biotechnology	Cat#sc-423027-HDR
Software and Algorithms
Bowtie2	Langmead and Salzberg, 2012	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
STAR	Dobin et al., 2013	https://github.com/alexdobin/STAR
MACS2	Zhang et al., 2008	https://github.com/taoliu/MACS
HTSeq	Anders et al., 2015	https://htseq.readthedocs.io/
GFold	Feng et al., 2012	https://bitbucket.org/feeldead/gfold
DESeq2	Love et al., 2014	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
GSEA	Subramanian et al., 2005	http://software.broadinstitute.org/gsea/index.jsp
Bedtools	Quinlan and Hall, 2010	http://bedtools.readthedocs.io/en/latest/
Samtools	Li et al., 2009	http://samtools.sourceforge.net/
Picard	Broad Institute	http://broadinstitute.github.io/picard
Trimmomatic	Bolger et al., 2014	http://www.usadellab.org/cms/?page=trimmomatic
Gene Ontology	Gene Ontology Consortium, 2015	http://geneontology.org/
BWA-MEM	Li, 2013	bio-bwa.sourceforge.net/
Mutect	Cibulskis et al., 2013	http://www.broadinstitute.org/cancer/cga/mutect
Pindel	Ye et al., 2009	gmt.genome.wustl.edu/packages/pindel

Significance.

Mammalian SWI/SNF (BAF) complexes are mutated in over 20% of human cancers, with both gain- and loss-of-function perturbations each implicated in malignancy. Here, we find that the SS18-SSX fusion oncoprotein hallmark to synovial sarcoma globally hijacks BAF complexes away from enhancers to broad polycomb domains to mediate oncogenic gene activation. BAF47 reassembly is dispensable for proliferative arrest upon SS18-SSX suppression, uncoupling the contributions of two concurrent BAF complex perturbations. These data underscore that the cancer-specific chromatin-targeting function of the SSX tail is the critical driving event in SS, and that comprehensive biochemical and functional interrogation of aberrant chromatin regulatory complexes is required to inform therapeutic approaches.

Highlights.

• SS18-SSX assembly results in concurrent gains and losses in genome-wide BAF complex targeting

• Synovial sarcoma is transcriptionally distinct from other BAF complex-driven malignancies

• SS18-SSX targets BAF complexes to broad polycomb domains to activate bivalent genes

• BAF47 reassembly activates enhancers but is dispensable for proliferative arrest