SMARCA4 loss reprograms p300 chromatin occupancy to subvert p53-mediated transcriptional repression in ovarian small cell carcinoma

Abstract

Small cell carcinoma of the ovary, hypercalcemic type (SCCOHT) is a rare, aggressive cancer driven by biallelic inactivation of SMARCA4 (BRG1), the ATPase subunit of the SWItch/Sucrose Non-Fermentable (SWI/SNF) chromatin remodeling complex. Despite its aggressiveness, SCCOHT exhibits a low mutation burden and retains wild-type p53 tumor suppressor. Using an integrative omics approach, we show that SMARCA4 loss redistributes the p300 acetyltransferase to promoters of cell cycle progression genes, increasing histone H3 lysine 27 acetylation (H3K27ac) and promoting transcription that sustains tumor growth. This opposes p53-mediated transcriptional repression at these loci, where co-occupancy of p53 and histone deacetylase HDAC2 are associated with decreased H3K27ac and suppressed gene expression. SMARCA4 restoration or pharmacologic inhibition of p300 suppresses SCCOHT growth, an effect reversed by p53 deletion and accompanied by reactivation of these cell cycle progression genes. Our findings uncover a chromatin-based mechanism whereby SMARCA4 loss subverts p53-mediated transcriptional repression through reprogramming p300 genomic occupancy to sustain oncogenic growth, highlighting p300 as a potential therapeutic target in SCCOHT.

Small cell carcinoma of the ovary, hypercalcemic type, is a rare, aggressive cancer caused by SMARCA4 inactivation and features wild-type p53. Here, the authors suggest that loss of SMARCA4 alters chromatin and reprograms p300 binding, undermining p53’s transcriptional repression and promoting cancer growth.

Introduction

The SWItch/Sucrose Non-Fermentable (SWI/SNF) chromatin remodeling complexes regulate transcription by using energy from ATP hydrolysis to modulate chromatin accessibility. Each complex contains an ATPase subunit, either SMARCA4 (BRG1) or its paralog SMARCA2 (BRM), together with a set of accessory proteins that define their distinct identities: canonical BRG1/BRM-associated factor (cBAF), polybromo-associated BAF (PBAF), and GLTSCR1/BAF53a-containing BAF (GBAF) or non-canonical BAF (ncBAF)^1–3. Mutations in SWI/SNF subunits are found in ~ 25% of all human cancers, highlighting their critical roles in tumorigenesis, with SMARCA4 being among the most frequently mutated subunits across diverse cancer types^4,5. We and others previously uncovered that SMARCA4 inactivating mutations are found in ~ 100% of the cases of small cell carcinoma of the ovary, hypercalcemic type (SCCOHT)^6–8, a rare, lethal ovarian cancer subtype affecting young women⁹. Although SCCOHT tumors also lack SMARCA2 protein due to epigenetic silencing^10,11, SMARCA4 inactivation is considered as the sole genetic driver of this malignancy⁸. Despite extensive characterization of SCCOHT, the mechanisms by which SMARCA4 loss promotes tumor development are still poorly understood.

A central question remains whether SWI/SNF loss drives tumorigenesis primarily by impairing DNA damage repair (DDR) pathways or by causing aberrant transcription¹². Numerous studies demonstrated that alterations in SWI/SNF components, such as SMARCA4 inactivation^13–15, promote genomic instability in various cell contexts, consistent with a role in DDR. In addition, we recently found that SMARCA4 deficiency increases R-loop levels in SCCOHT, which sequester BRCA1 for R-loop resolution, limiting its availability for DDR and resulting in PARP inhibitor sensitivity¹⁶. These findings would be consistent with a model that SMARCA4 loss contributes to SCCOHT development by disrupting genome integrity. However, SCCOHT primary tumors exhibit few genetic mutations, minimal chromosomal alterations, and retain wild-type (WT) p53 tumor suppressor^6–8,17. It is possible that BRCA1 redistribution, rather than complete loss, still permits sufficient DNA repair along with intact p53, thus preventing the accumulation of mutations in SCCOHT. These observations suggest that the altered DDR may not be the primary driver of SCCOHT and support an alternative model in which SMARCA4 loss drives SCCOHT development through activation of oncogenic transcriptional programs². In line with this, SMARCA4 deletion has been associated with gene-expression changes that promote cell proliferation¹⁸, one of the cancer hallmarks¹⁹. However, the exact mechanisms through which SMARCA4 loss drives this oncogenic transcriptional reprogramming remain unclear.

Here, through integrated functional genetic screening, transcriptomic profiling, and genome-wide chromatin state analysis, we reveal that SMARCA4 loss subverts p53-mediated transcriptional repression of a core set of cell cycle progression genes, through reprogramming the histone acetyltransferase p300 to drive transcriptional activation of these loci. Furthermore, we provide preclinical evidence supporting p300 inhibition as a potential therapeutic strategy for SCCOHT.

Results

SMARCA4 loss upregulates p53-repressed cell cycle genes

The monogenic, genomically quiescent nature of SCCOHT makes it an ideal model for uncovering genetic dependencies driven by SMARCA4 loss, which may, in turn, reveal the mechanisms sustaining oncogenic growth. To systematically identify such vulnerabilities, we analyzed genome-wide CRISPR/Cas9 knockout screens from the Cancer Dependency Map (DepMap, https://depmap.org) across 35 ovarian cancer cell lines. Among these, three were SCCOHT cell lines (BIN-67, SCCOHT-1, COV434), while the remaining 32 were SMARCA4/2-proficient. Consistent with our previous findings^20,21, CDK4 was the top gene whose knockout was selectively lethal to SCCOHT cells (Fig. 1a). This analysis also identified MDM2, an E3 ubiquitin ligase targeting p53 for degradation, as another top synthetic lethal target in SCCOHT cells. Validating this, SCCOHT cell lines, but not the non-transformed fallopian tube epithelial control cell lines (hTERT-immortalized FT237, SV40-immortalized FT190), were highly sensitive to treatment of idasanutlin, a selective inhibitor of MDM2²², in both cell viability (Fig. 1b) and colony formation (Supplementary Fig. 1a and Supplementary Data 1) assays. As expected, this growth suppression was accompanied by elevated p53 expression in all SCCOHT cell lines (Supplementary Fig. 1b), whereas TP53 knockout in these cells conferred near-complete resistance to idasanutlin treatment (Supplementary Fig. 1c–e and Supplementary Data 1). While prolonged restoration of SMARCA4 strongly suppresses SCCOHT cell proliferation^11,21, short-term SMARCA4 re-expression using a doxycycline inducible system conferred resistance to idasanutlin (Fig. 1c, d), supporting the causal role of SMARCA4 loss for MDM2 dependency in SCCOHT. These data suggest that SMARCA4 loss may promote tumorigenesis in part through countering p53 tumor suppressor function.

Fig. 1. MDM2 inhibition is synthetic lethal with SMARCA4-loss in SCCOHT.

a Ranking of genes by their differential dependency between SMARCA4/2-deficient (A4/2^Def; n = 3) and proficient (A4/2^Pro; n = 32) ovarian cancer cell lines using the CERES gene effect data from DepMap genome-wide CRISPR-knockout screens (two-tailed t test). Each dot denotes a gene. b Cell viability assay of indicated SMARCA4 (A4)-proficient vs SCCOHT cell lines treated with idasanutlin for 5 days (viability) or 24 h at 0.5 µM (immunoblots). c, d Cell viability assay (c) and immunoblots (d) of BIN-67 cells, ± doxycycline (Dox)-inducible SMARCA4 re-expression, treated with idasanutlin for 5 days. b, c n  =  3 independent experiments. Error bars, mean ± SEM. e Normalized enrichment score (NES) comparison of the RNAseq analysis for pathways enriched in BIN-67 cells upon idasanutlin treatment (0.5 µM, 24 h) or SMARCA4 (A4) re-expression. Significant pathways were defined by p-value < 0.05 (one-sided weighted Kolmogorov-Smirnov test) and NES cutoff ± 1.75, and are colored green (idasanutlin), blue (SMARCA4), or pink (both). f Dot-plot representing RNA-seq logFC for SMARCA4 restoration (y axis) and idasanutlin (Ida) treatment (x axis; 24 h, 0.5 µM), colored by p53 regulations score²⁴ - pink is negatively regulated, and black is positively regulated. g Average profile plot (up) and heatmap (down) of p53 chip at p53 bound regions that have negative p53 regulation score and are downregulated by SMARCA4 and idasanutlin. Rows represent individual peaks; the scale bar represents signal intensity. h Bar-plot showing the top 10 GO terms ranked by p-value for the effect of SMARCA4 overexpression and idasanutlin treatment (24 h, 0.5 µM) on the 101 genes in (h). d–h representative data of 3 independent experiments are shown.

To understand the underlying mechanisms, we performed RNA-seq experiments in BIN-67 cells following either idasanutlin treatment (24 h) or SMARCA4 re-expression. Gene Ontology (GO) analysis on our RNA-seq data revealed a negative normalized enrichment score (NES) for 155 pathways downregulated by both SMARCA4 restoration and p53 stabilization (Supplementary Fig. 1f). The top 20 of these pathways are associated with proliferation, notably Mitotic Sister Chromatid Segregation as the most significantly repressed pathway (Fig. 1e, bottom left quadrant). In contrast, only 25 pathways were positively enriched by both treatments (Supplementary Fig. 1f), including Extracellular Encapsulation Structure and other Extracelluar Matrix-related terms (Fig. 1e, top right quadrant), which is in line with previous findings²³. These observations suggest that while SMARCA4 and p53 co-regulate both negatively and positively enriched pathways, the predominant effect is negative regulation.

To substantiate the above observations, we examined whether the previously established list of p53 direct and indirect target genes (n = 3509), derived from 16 datasets surveying genes both regulated and bound by p53 at their loci²⁴, is also controlled by SMARCA4. First, we validated that the expression changes of these 3509 genes in BIN-67 cells upon idasanutlin treatment strongly correlated with p53 regulation scores²⁴ (Fig. 1f). Notably, a subset of genes (n = 159) with negative p53 regulation scores, indicative of p53-mediated repression, was downregulated by both SMARCA4 restoration and idasanutlin treatment (Fig. 1f). This supports that SMARCA4 restoration and p53 stabilization converge on shared gene repression programs in SCCOHT. To corroborate these results in BIN-67 cells, we conducted RNA-seq in two additional SCCOHT cell lines, COV434 and SCCOHT-1, with idasanutlin treatment (24 h) and reanalyzed the RNA-seq data in BIN-67, COV434 and SCCOHT-1, -/+ SMARCA4 restoration²⁵. Idasanutlin treatment in all three SCCOHT cell lines resulted in comparable downregulation of these 159 negatively regulated genes identified above (Supplementary Fig. 1g–i). Similarly, SMARCA4 restoration showed analogous downregulation of this gene set in these SCCOHT cells (Supplementary Fig. 1j–l). To assess whether p53-mediated repression is direct or indirect, we performed RNA-seq in BIN-67 and COV434 cells upon short-term idasanutlin treatment. Comparison of 1 h and 3 h treatments showed that robust p53 protein stabilization and induction of p53 canonical target gene CDKN1A (p21)^26–30 occurred by 3 h, which was therefore selected for subsequent RNA-seq analysis (Supplementary Fig. 1m, n). Notably, only ~ 40% of the 159 genes were downregulated at this early timepoint in both cell lines (Supplementary Fig. 1o, p). These results suggest that the complete repression observed at 24 h time point primarily depends on secondary, likely indirect, effects downstream of p53 activation. Nevertheless, our chromatin immunoprecipitation with sequencing (ChIP-seq) for p53 in BIN-67 cells revealed strong p53 occupancy at these 159 gene loci, although their repression likely involves indirect downstream mechanisms (Fig. 1g). Furthermore, the top 10 negatively enriched GO terms derived from these 159 genes upon SMARCA4 re-expression or idasanutlin treatment were predominantly associated with cell cycle regulation (Fig. 1h). These results suggest that SMARCA4 loss results in aberrant expression of cell cycle progression genes that can be repressed by p53 stabilization with idasanutlin.

SMARCA4 loss subverts p53-mediated repression of cell cycle progression genes via H3K27ac redistribution

Given that most proliferation-related pathways were negatively enriched by SMARCA4 re-expression and p53 stabilization and the substantial overlap in downregulated genes by both conditions, we hypothesized that SMARCA4-mediated growth suppression requires functional p53. Consistent with this, TP53 knockout in BIN-67 cells didn’t confer a growth advantage but partially rescued the growth suppression induced by restoration of SMARCA4 (Fig. 2a, b and Supplementary Data 1). This is in line with the patient survival data of the TCGA Pan-Cancer Atlas cohort (n = 10,967), stratified into four groups based on copy-number alterations and mutation status of TP53 and SMARCA4 (Fig. 2c), where patients with SMARCA4-deficient tumors, regardless of TP53 status, had the worst survival outcome among all groups, while those with tumors retaining SMARCA4, but aberrant in TP53, exhibited intermediate overall survival. These results support a model where SMARCA4-mediated tumor suppression depends, at least in part, on functional p53.

Fig. 2. SMARCA4 loss subverts p53-mediated repression of cell cycle genes via H3K27ac redistribution.

a, b Colony formation assay (a) and immunoblots (b) of BIN-67 cells with or without SMARCA4 (A4) re-expression in parental or TP53^KO clone. Representative results from three independent experiments. c Survival of Pan Cancer TCGA patient cohort stratified by SMARCA4 (A4) and TP53 status (log-rank test). Prof, proficient; def, deficient. TP53^prof A4^prof (n = 3420), TP53^prof A4^def (n = 324), TP53^def A4^prof (n = 882), TP53^def A4^def (n = 404). Log-rank (Mantel-Cox) test, **p- value 0.0012, ****p-value < 0.0001. d RNA-seq heatmap with simple clustering method and Manhattan clustering distance for SMARCA4 overexpression, TP53^KO, TP53^KO and SMARCA4 overexpression and idasanutlin treatment (0.5 µM, 24 h) in BIN-67 cells. Scale bar indicates z-scores. e UMAP of Normalized enrichment scores (NES) of pathways enriched in the RNA-seq conditions in (h). f Dot-plot showing the effect of SMARCA4 overexpression, idasanutlin treatment, TP53^KO and SMARCA4 overexpression plus TP53^KO on the top 10 downregulated GO terms ranked by NES upon SMARCA4 overexpression. p-values calculated by one-sided weighted Kolmogorov-Smirnov test (g) Representation of common and individual regions bound by H3K27ac in BIN-67 expressing control vector (Ctrl), SMARCA4, TP53^KO and SMARCA4 plus TP53^KO, with display of H3K27ac peak intensity (color) and percentage of peaks for each region (dot size). h Average profile plot for G and H H3K27ac regions described in panel (g). i Dot-plot showing the effect of SMARCA4 overexpression, idasanutlin treatment, TP53^KO and SMARCA4 overexpression plus TP53^KO on the top 10 downregulated GO terms ranked by NES of SMARCA4 overexpression of the genes in proximity to the H3K27ac regions described in panel (h). p-values calculated by one-sided weighted Kolmogorov-Smirnov test. d–i Representative data of 3 independent experiments are shown.

To further dissect the role of p53 in mediating SMARCA4 suppressor function, we conducted RNA-Seq studies in BIN-67 parental and TP53 knockout (TP53^KO) cells, before and after SMARCA4 restoration, alongside parental cells treated with idasanutlin. As shown in Fig. 2d, hierarchical clustering of transcriptomes confirmed that SMARCA4 restoration repressed a distinct gene expression signature that was partially recapitulated by idasanutlin treatment, consistent with the pathway analysis observation in Fig. 1g; in contrast, TP53^KO, either alone or combined with SMARCA4 re-expression, displayed an opposing expression profile. UMAP visualization of GO analysis showed that the largest cluster of pathways (n = 156) was negatively enriched by SMARCA4 and idasanutlin but positively enriched in TP53^KO cells, regardless of SMARCA4 status (Fig. 2e). Notably, the top 10 negatively enriched GO terms by SMARCA4 re-expression in parental cells were primarily linked to cell cycle progression, such as regulation of chromosome separation, mitotic sister chromatid segregation, and metaphase chromosome alignment (Fig. 2f, blue). These same 10 terms were also negatively enriched in parental cells treated with idasanutlin (green), but all positively enriched in TP53^KO cells (black) and almost all (9/10) positively enriched in TP53^KO with SMARCA4 re-expression (pink) (Fig. 2f). These results indicate that SMARCA4 loss increases transcription of cell cycle progression genes, subverting p53-mediated repression to drive SCCOHT.

To test whether transcriptional interplay between SMARCA4 and p53 reflects changes in chromatin activity, we performed ChIP-seq for H3K27ac, a histone modification that marks active enhancers and transcription start sites (TSS)^31–33, in BIN-67 cells with four different genotypes used in our above transcriptomic study: control, SMARCA4-restored, TP53^KO, and SMARCA4-restored with TP53^KO. In line with previous findings^34,35, SMARCA4 re-expression led to a genome-wide increase in H3K27ac signal, particularly at enhancers, while mildly decreased levels at promoter and TSS regions (Supplementary Fig. 2a, b). Since our above transcriptomic data show that SMARCA4-dependent repression of cell cycle progression genes is rescued by p53 loss, we hypothesized that this effect is mediated by changes in H3K27ac at these specific gene loci rather than by global alterations. We therefore performed a comparative peak analysis and categorized H3K27ac-enriched loci based on their presence in BIN-67 cells across all four genotypes to identify SMARCA4-sensitive regulatory regions (Fig. 2g). Among these, we focused on two regions: row G, where H3K27ac peaks were present in control, TP53^KO, and SMARCA4-restored with TP53^KO cells, but lost in SMARCA4-restored cells; and row H, where H3K27ac peaks were present across all conditions with a subset having significantly reduced intensity upon SMARCA4 re-expression while remaining high in all other conditions (Supplementary Data 2). Together, H3K27ac peaks in G and the above-described subset in H account for 11.03% of the total H3K27ac signal in the control condition, representing regions where SMARCA4 specifically suppresses H3K27ac only in the presence of p53 (Fig. 2g, h).

Next, we asked whether these chromatin changes were consistent with the SMARCA4-p53 transcriptional interplay observed in Fig. 2c. We used the Genomic Regions Enrichment of Annotations Tool (GREAT, https://great.stanford.edu/great/public/html/) to annotate the genes in proximity to these p53-dependent SMARCA4-repressive regions and analyzed their mRNA expression from our RNAseq data sets described above in Fig. 2d. The follow up GO analysis of genes showing consistent changes in chromatin state and mRNA expression revealed a strong enrichment for pathways involved in the positive regulation of cell cycle in the top 10 GO terms (Fig. 2i), which closely mirrored those identified in the whole transcriptome analysis (Fig. 2f) - all negatively enriched in BIN-67 cells upon SMARCA4 restoration or idasanutlin treatment but positively enriched in TP53^KO cells regardless of SMARCA4 status. Together, these results indicate that SMARCA4 loss leads to elevated H3K27ac at these cell cycle progression gene loci to promote their transcription, subverting p53-mediated repression at these loci to promote SCCOHT proliferation.

Inhibition of the acetyltransferase p300 is synthetic lethal with SMARCA4-loss in SCCOHT

Based on the above findings, we hypothesized that SCCOHT cells depend on key epigenetic regulators to maintain elevated histone H3K27 acetylation at the cell cycle progression gene loci. To systematically test this, we performed pooled CRISPR/Cas9 knockout screens in SCCOHT-1 and IOSE80, an immortalized but non-transformed ovarian epithelial cell line, using a previously validated sgRNA library targeting 496 epigenetic modifiers³⁶. As shown in Fig. 3a, these screens identify EP300, encoding for the key acetyltransferase p300 that acetylates histone H3K27, as a top candidate whose knockout was selectively lethal in SCCOHT-1 but not in IOSE80 cells. Our finding was independently supported by the publicly available CRISPR knockout screen data from DepMap previously analyzed in Fig. 1a, where EP300 was also among the top candidate genes whose knockout was selectively lethal to SCCOHT cells but not to SMARCA4/2-proficient cells (Fig. 3b). CREB binding protein (CBP), a p300 paralog that also acetylates H3K27³⁷, was not a synthetic lethal target in SCCOHT cells in both screens (Fig. 3a, b), suggesting a unique role of p300 in SCCOHT. Together, these results indicate that p300 may play a critical role in maintaining elevated H3K27ac levels at cell cycle progression gene loci in SCCOHT cells.

Fig. 3. Inhibition of p300 acetyltransferase is synthetic lethal with SMARCA4-loss.

a Differential gene dependency in IOSE80 (SMARCA4/2-proficient) and SCCOHT-1 cells from pooled CRISPR screens with a sgRNA knockout library against epigenetic regulators (n = 494). The relative abundance of sgRNA after 14 days of culturing compared to seeding (LFC) was determined by next-generation sequencing. b Ranking of genes by their differential dependency between SMARCA4/2-deficient (A4/2^Def; n = 3) and proficient (A4/2^Pro; n = 32) ovarian cancer cell lines using the CERES gene effect data from DepMap genome-wide CRISPR-knockout screens (two-tailed t test). c, d Colony formation assays (c) and immunoblots (d) of indicated cell lines expressing ctrl or independent shRNAs targeting EP300. e–g Cell viability assay (e), colony formation assay (f), and immunoblots (g) of indicated cell lines treated with A-485 for 5, 1, or 10–15 days, respectively. h Cell viability assay showing inducible SMARCA4 (A4) expression (Dox 1 µg/ml) in BIN-67 confers resistance to 5 days of A-485 treatment (immunoblot Extended Data Fig. 1b). i, j Immunoblots (i) and cell viability assay (j) in FT190 cells with indicated SMARCA4/2 perturbations treated with A-485 for 5 days. c, d, f, g, i Representative data of 3 independent experiments are shown. e, h, j, n  =  3 independent experiments. Error bars, mean ± SEM.

Validating the above screen results, EP300 knockdown using three independent shRNAs strongly suppressed the growth of all three SCCOHT cell lines with less effects on the SMARCA4-proficient controls (Fig. 3c, d and Supplementary Data 1). In keeping with this, treatment of A-485, a recently developed inhibitor of p300 acetyltransferase activity³⁸, robustly suppressed SCCOHT cell growth with little effect on SMARCA4-proficient controls (Fig. 3e–g and Supplementary Data 1). Furthermore, short-term restoration of SMARCA4 in BIN-67 cells conferred resistance to A-485 (Fig. 3h), while induction of SMARCA4 expression with a low dose of doxycycline conferred resistance to A-485 in long-term colony formation assays (Supplementary Fig. 3a, b and Supplementary Data 1). In contrast, SMARCA4 knockout and SMARCA2 knockdown, either alone or in combination, sensitized non-transformed FT190 cells to A-485 (Fig. 3i, j), suggesting that loss of either SWI/SNF ATPase induces p300 dependency. The above results demonstrate that the acetyltransferase function of p300 is essential for SCCOHT, which may be exploited therapeutically using existing acetyltransferase inhibitors against p300.

SMARCA4 loss redirects p300 to cell cycle progression gene loci to promote their expression

To examine the role of p300 in the SMARCA4-p53 regulatory axis, we performed ChIP-seq to assess p300 occupancy at H3K27ac-marked sites in BIN-67 cells across the four genotypes described above (control, SMARCA4-restored, TP53^KO, and SMARCA4-restored with TP53^KO). In contrast to the global gain of H3K27ac observed upon SMARCA4 re-expression (Supplementary Fig. 2a), global p300 occupancy was markedly reduced following SMARCA4 restoration (Supplementary Fig. 4a), suggesting that other histone acetyltransferases are likely responsible for the global gain of H3K27ac. Notably, TP53^KO, either alone or combined with SMARCA4, partially reversed this reduction of p300 occupancy (Supplementary Fig. 4a). Furthermore, p300 peaks were greatly depleted by SMARCA4 restoration at both enhancer and promoter-TSS regions marked by H3K27ac (Supplementary Fig. 4b, c). Notably, SMARCA4 induction slightly increased p300 protein levels (Supplementary Fig. 4d), indicating that the reduced p300 chromatin occupancy is not attributed to decreased expression. Together, these results suggest that the reduced p300 occupancy upon SMARCA4 restoration may be responsible for the reduction of H3K27ac at cell cycle progression gene loci.

To further assess the contribution of SMARCA4 restoration to global changes in p300 occupancy, we clustered p300-bound peaks into three categories: depleted, common, and gained (Fig. 4a). Depleted peaks, which constituted the majority, corresponded to regions where p300 binding was completely lost upon SMARCA4 restoration. Common peaks retained p300 binding in both control and SMARCA4-restored cells, although peak intensity was reduced in the latter. Gained peaks represented a minor subset of sites that acquired p300 binding upon SMARCA4 restoration (Fig. 4a). Comparison of our p300 ChIP-seq analysis to publicly available SMARCA4 and H3K27ac ChIP-seq data sets in the same BIN-67 cells³⁵ showed SMARCA4 binding at p300-depleted regions following re-expression of either WT or the ATPase-deficient SMARCA4 mutant (T910M) (Supplementary Fig. 4e). However, only restoration of WT SMARCA4, but not the ATPase-deficient mutant SMARCA4 (T910M), resulted in H3K27ac reduction in these regions, which closely recapitulated p300 occupancy changes (Fig. 4a). This suggests that p300-mediated acetylation at these loci is countered by the SMARCA4 ATPase activity. Consistent with this, our H3K27ac ChIP-seq in BIN-67 cells revealed a similar trend of H3K27ac signal changes upon SMARCA4 re-expression and A-485 treatment (Supplementary Fig. 4e). Together, these findings indicate that SMARCA4 restoration drives widespread loss of p300 binding and H3K27ac deposition at these regulatory regions through its ATPase activity.

Fig. 4. SMARCA4 loss redistributes p300 to cell cycle progression gene loci to promote their expression.

a Left, p300 ChIP-seq heatmap in BIN-67 expressing control or doxycycline-inducible SMARCA4 (A4; 1 µg/ml, 24 h). Peak clustering of p300 occupancy changes upon SMARCA4 restoration: depleted, unchanged, gained. Right, H3K27ac ChIP-seq heatmap³⁵ for clusters defined in the left panel for control, SMARCA4-restored, and ATPase-inactive SMARCA4 (T910M). Rows represent individual peaks; scale bars represent signal intensity. b, c Volcano plots of differentially expressed genes (DEG) from RNA-seq in SCCOHT cell lines treated with A-485 (1 µM, 24 h) (b) or expressing inducible SMARCA4³⁴ (c). Pink and purple boxes indicate filtering criteria schematically; exact thresholds are described in (d). d Venn diagram illustrating the overlap between p300-depleted peaks (q-Value < 5 × 10 − 6) (a), DEG from the A-485 RNA-seq data (b) (LFC threshold of ± 0.5), and DEG from the SMARCA4 restored RNA-seq data (c) (LFC threshold of ± 0.5). e Dotplot showing the effect of A-485 treatment (1 µM, 24 h), SMARCA4 overexpression, idasautlin (Ida) treatment (0.5 µM, 24 h) in parental or TP53^KO BIN-67 cells on the top 5 downregulated GO terms, ranked by Normalized Enrichment Score (NES) of A-485 treatment in the 417 genes obtained from the Venn diagram overlap in (d). p-values calculated by one-sided weighted Kolmogorov-Smirnov test. f Changes in E2F2 mRNA levels normalized to ACTB expression in parental or TP53^KO BIN-67 cells upon A-485 (1 µM, 24 h) or idasanutlin (0.5 µM, 24 h) treatment. g Colony formation assay of indicated cell lines expressing control or independent shRNAs targeting E2F2. h ChIP-seq heatmaps of transcription start site (TSS) regions clustered into two groups: upper, the 417 genes obtained from the Venn diagram overlap in (d); lower, all genes (73686 genes, GENCODE 48 release). Displayed p300 ChIP-seq in BIN-67 expressing control vector or inducible SMARCA4 (Doxycycline, 1 µg/ml), SMARCA4 and H3K27ac ChIP-seq in BIN-67 expressing control vector, SMARCA4, or catalytically dead SMARCA4 (T910M), as well as H3K27ac ChIP-seq in parental BIN-67 cells ± A-485 treatment (1 µM, 24 h). a–e, g, h representative data of 3 independent experiments generated in this study are shown. f n  =  4 independent experiments. Error bars, mean ± SEM.

We next asked whether this loss of p300 binding underlies the SMARCA4-mediated suppression of cell cycle progression genes observed in Fig. 2c. We used the GREAT tool to identify genes proximal to the above described p300-depleted peaks and examined their overlap with transcriptional targets. To this end, we first performed RNA–seq in three SCCOHT cell lines (BIN-67, COV434, SCCOHT-1) before and after A-485 treatment to define p300-regulated genes (Fig. 4b). These were compared to SMARCA4 regulated genes derived from published RNA–seq of the same cell lines expressing inducible SMARCA4³⁴ (Fig. 4c). Genes commonly regulated by A-485 and SMARCA4 in all three lines (|log₂ fold change| > 0.5) were intersected with the genes near p300-depleted peaks. This analysis identified 417 genes regulated by both p300 and SMARCA4 and associated with reduced p300 occupancy upon SMARCA4 restoration (Fig. 4d).

We then examined the expression of these 417 genes across RNA-seq data sets in BIN-67 cells with differential TP53/SMARCA4 status and idasanutlin treatment (Fig. 2b) and added two additional conditions: A-485 treatment and A-485 combined with TP53^KO. GO analysis of the top 5 negatively enriched terms upon A-485 treatment revealed pathways primarily involved in cell cycle regulation, sister chromatid segregation and regulation of nuclear division, all of which were also negatively enriched upon idasanutlin treatment and SMARCA4 restoration (Fig. 4e). In contrast, TP53^KO positively enriched these pathways and rescued the negative NES scores of SMARCA4 overexpression and A-485 treatment (Fig. 4e). Notably, E2F2 was among the top 20 of 242 genes downregulated by A-485, SMARCA4, and idasanutlin, but elevated upon TP53^KO alone or in combination with A-485 or SMARCA4 (Supplementary Fig. 4f), which we validated by qRT-PCR (Fig. 4f). Furthermore, E2F2 knockdown using three independent shRNAs strongly suppressed the growth of SCCOHT lines but had much less impact on SMARCA4/2-proficient ovarian controls (Fig. 4g, Supplementary Data 1 and Supplementary Fig. 4g), supporting the critical role of this p300-activated and p53-repressed cell cycle axis in promoting SCCOHT proliferation.

Consistent with the transcriptional regulation, ChIP–seq analysis in BIN-67 cells revealed high occupancy of p300, H3K27ac, and SMARCA4³⁵ specifically at the TSS of the 417 gene loci (Fig. 4h, upper panel) compared to all coding and non-coding genes TSS genome-wide (Fig. 4h, lower panel). Focusing on these 417 gene loci, H3K27ac was suppressed upon A-485 treatment; furthermore, SMARCA4 restoration led to strong reduction of p300 occupancy at these loci (Fig. 4h). Complementing these findings, we reanalyzed a publicly available ChIP–seq dataset in BIN-67 cells³⁵ showing that both WT and T910M mutant SMARCA4 were recruited to these 417 gene loci upon restoration, but only WT SMARCA4 was able to reduce H3K27ac in these regions (Fig. 4h). These results indicate that SMARCA4-mediated reprogramming of p300 chromatin occupancy is essential for controlling H3K27ac levels at these loci, which requires SMARCA4 ATPase activity. Together, these findings indicate that SMARCA4 loss upregulates these cell cycle progression genes by increasing p300 occupancy at their loci, driving SCCOHT proliferation.

SMARCA4 loss alters chromatin states at cell cycle progression gene loci

To determine whether the effects of SMARCA4 loss extend beyond gained p300 occupancy at individual cell cycle progression genes to broader chromatin state remodeling in SCCOHT, we applied the ChromHMM pipeline³⁹ to investigate this. We integrated ATAC-seq, RNA-seq, and ChIP-seq data for p300 (Fig. 4a), SMARCA4, and activating histone marks (H3K27ac, H3K4me1) in BIN-67 cells before and after restoration of SMARCA4 (Fig. 2e)^21,34,35, as well as ChIP-seq data for repressive histone marks (H3K9me3, H3K27me3) available in parental BIN-67 cells³⁵. From this integrated analysis, we defined five distinct chromatin states and inferred the biological function of the associated genomic regions, based on the combination of chromatin marks, chromatin accessibility, and transcriptional activity in their spatial context (Fig. 5a and Supplementary Fig. 5a–g). State 1 (S1) represents a quiescent or inactive chromatin state, lacking enrichment for any activation or repression marks, chromatin accessibility, or transcription. S2 corresponds to a repressed state, predominantly marked by H3K27me3, indicative of Polycomb-mediated silencing⁴⁰. S3 denotes an enhancer state, characterized by the presence of H3K4me1 along with varying levels of p300, H3K27ac, and ATAC-seq signal, suggesting poised or active enhancer elements. S4 marks active TSSs, showing strong enrichment for H3K27ac, H3K4me1, p300, and ATAC signal at annotated TSS regions, along with detectable transcriptional activity. S5 reflects a transcribed state, defined primarily by RNA-seq signal extending across gene bodies and exons, with lower enrichment for chromatin accessibility or histone activation marks at promoters.

Fig. 5. SMARCA4 loss alters chromatin states at cell cycle progression gene loci.

a Chromatin states region clustering using the ChromHMM³⁹ pipeline displaying average chromatin states of BIN-67 expressing control vector or SMARCA4 (A4). A total of 5 states were identified through genomic profiling of 7 chromatin features in BIN-67 before or after restoration of SMARCA4: total RNA-seq, ATAC-seq, and p300, H3K27ac, H3K4me1, H3K9me3 and H3K27me3 ChIP-seq. Darker heatmap colors indicate higher enrichment for each chromatin feature in that state. b–d Chromatin state transition maps for occupancies of p300 (b), SMARCA4 (c), H3K27ac (d) following SMARCA4 restoration in BIN-67 cells. For each state-state change, circle size depicts the relative amount of that state change compared to the initial genome-wide state representation ([genomic bp initial → final]/[genomic bp initial]), and color indicates the proportion bound by p300, SMARCA4, H3K27ac. e Dotplot showing the effect of SMARCA4 overexpression, idasautlin treatment, TP53^KO, SMARCA4 overexpresion plus TP53^KO, A-485 treatment (1 µM, 24 h) and A-485 treatment (1 µM, 24 h) plus TP53^KO on the top 10 downregulated GO terms, ranked by Normalized Enrichment Score (NES) in response to A485 treatment, in the genes associated with the state change from S4 to S1 and S2 upon SMARCA4 restoration from (b). p-values calculated by one-sided weighted Kolmogorov-Smirnov test. f, g Colony formation assay (f) and immunoblots (g) of BIN-67 parental or TP53^KO cells ± A-485 treatment (f, 15 days; g, 24 h). Representative results from three independent experiments. h, i Cell cycle analysis by propidium iodide staining in BIN-67 parent (h) and TP53^KO (i) cells treated with A-485 (1 µM, 24 h) or idasanutlin (0.5 µM, 24 h). a–g Representative data of 3 independent experiments generated in this study are shown. h, i n = 3 independent experiments. Error bars, mean ± SEM.

To visualize how SMARCA4 restoration alters p300 binding and H3K27ac across these chromatin states, we examined ChromHMM-predicted transitions between SMARCA4-deficient (Ctrl) and SMARCA4-restored (A4) conditions (Supplementary Fig. 5a). Notably, SMARCA4 restoration induced the transition of a set of active promoter states (S4) in control cells to either quiescent (S1) or repressed chromatin (S2) in SMARCA4-restored cells, which were accompanied by a marked loss of p300 binding (Fig. 5b), visualized as blue shading for the decrease of log₂(fold change) of p300 occupancy in SMARCA4-restored cells relative to the control. In the same regions, SMARCA4-bound loci increased (Fig. 5c, red shading), while the H3K27ac occupancy decreased (Fig. 5d, blue shading), which was also observed upon p300 inhibition with A-485 (Supplementary Fig. 5h). Together, these data demonstrate that SMARCA4 restoration reprograms the chromatin landscape by evicting p300 from defined active promoter regions, leading to reduced histone H3K27 acetylation and transcriptional suppression of nearby genes that are otherwise aberrantly active when SMARCA4 is lost.

To validate whether these chromatin state transitions result in transcriptional repression of cell cycle progression genes, we identified nearby genes (2892) underlying above described S4 to S1/2 transition and analyzed their expression in the RNA-Seq datasets described in Fig. 4e. GO analysis of these entire 2892 genes once again revealed negative enrichment in cell cycle–related pathways in the top 10 GO terms in response to A-485, which were also negatively enriched upon SMARCA4 overexpression or idasanutlin treatment (Fig. 5e). In keeping with our previous observations, TP53^KO rescued the repression induced by SMARCA4 and A-485 treatment (Fig. 5e). Furthermore, of the 2892 genes, 555 were suppressed by A-485, idasanutlin treatment or SMARCA4 restoration (Supplementary Fig. 5i), including E2F2 that we have shown to be regulated by p300, p53, and SMARCA4 in Fig. 4f, g. Consistent with this, TP53^KO in BIN-67 and SCCOHT-1 cells rescued the A-485-induced growth suppression and RB phosphorylation (Fig. 5f, g, Supplementary Fig. 5j, k and Supplementary Data 1).

Furthermore, we conducted cell cycle analysis in BIN-67 and SCCOHT-1 cells in response to A-485 or idasanutlin treatments. In both cell lines, A-485 significantly induced G1 arrest, accompanied by a reduction in S and G2/M phases (Fig. 5h, i and Supplementary Fig. 6a–n). Similarly, idasanutlin also caused G1 arrest in these cells, associated with a reduction in S or S and G2/M phases (Fig. 5h, i and Supplementary Fig. 6a–n). These results indicate that both A-485 and idasanutlin induce cell cycle arrest primarily at the G1 phase, consistent with the growth suppression phenotype and reduced expression of cell cycle progression genes. Moreover, TP53^KO partially rescued this elevated G1 phase induced by A-485 in both BIN-67 and SCCOHT-1 models (Supplementary Fig. 6o, p), while abolishing the cell cycle effects caused by idasanutlin observed in parental cells as expected (Fig. 5h, i and Supplementary Fig. 6g–n, q, r). Notably, TP53^KO had variable effects on A-485-induced impact on other phases of the cell cycle, depending on the cell models tested. These results showed that A-485 treatment primarily induced G1 arrest in SCCOHT cells that is partially p53 dependent. Together, these findings support a central role for SMARCA4 loss in promoting a transcriptionally active chromatin state at these cell cycle progression gene loci, in part by recruiting p300, which counters their repression by p53.

Opposing chromatin regulation by SMARCA4 loss-induced p300 activation and p53-mediated HDAC2 repression

While p53 functions primarily as a transcription activator^26–28, many studies^41–44 have suggested a repressor function, although the mechanisms remain controversial^24,26. Notably, among the above-described 555 cell-cycle progression genes downregulated by SMARCA4 restoration, A-485, and idasanutlin treatment in BIN-67 cells, 144 are predicted to be regulated directly or indirectly by p53²⁴, and the majority of these (112/144) are suppressed by p53 stabilization (Supplementary Fig. 6s, t). Together, these findings support a role for p53 in repressing a subset of genes that are also suppressed upon SMARCA4 restoration and p300 inhibition. To understand the mechanisms underlying p53-dependent repression at these loci, we examined the protein interaction network of p53 using BioGRID (https://thebiogrid.org). Analysis of interaction frequency revealed histone deacetylases (HDACs), HDAC1 and HDAC2, among the top-ranking potential interactors (no other HDACs among the top 1000) (Fig. 6a). Given our above observations that TP53 knockout restored acetylation and transcription at cell cycle progression genes (Fig. 2e–g), we hypothesized that p53-dependent repression at these loci may involve HDAC1/2 activity opposing p300-mediated H3K27 acetylation. Confirming the BioGRID analysis, p53 was detected in HDAC2 immunoprecipitate from BIN-67 nuclear extract in both control and A-485 treated conditions (Fig. 6b). Similarly, we detected p53-HDAC1 interaction, although to a lesser extent compared to p53-HDAC2 (Supplementary Fig. 7a). Furthermore, treatment with the pan-HDAC inhibitor trichostatin A restored E2F2 expression (Fig. 6c) and rescued cell viability in the presence of A-485 (Fig. 6d), suggesting that HDAC inhibition recapitulates, at least in part, the functional effects of TP53^KO.

Fig. 6. SMARCA4 loss reprograms p300 chromatin occupancy opposing p53–mediated HDAC2 repression.

a Dot plot visualization of candidate p53 interactors identified from BioGRID (y-axis, log2 of the total number studies showing each interaction; x-axis, log2 BioGRID rank of each interactor). b Immunoblot of HDAC2 nuclear immunoprecipitation in BIN-67 parental or TP53^KO clone −/+ A-485 treatment (1 µM, 24 h). c n  =  4 biological replicates. Changes in E2F2 mRNA levels normalized to ACTB expression in BIN-67 cells, −/+ pretreatment with Trichostatin A (TSA; 100 nM, 48 h), followed by further treatment with A-485 (1 µM, 24 h) or idasanutlin (0.5 µM, 24 h). Two-way ANOVA. d n  =  3 independent experiments. Cell viability assay of BIN-67 cells −/+ Trichostatin A (100 nM; 6 days), followed by of A-485 treatment for 5 days. e Chip-seq heatmaps of in BIN-67 parental and TP53^KO cells expressing either control vector or SMARCA4 (A4) for p300, H3K27ac, p53, HDAC2, HDAC1, in the chromatin region identified in Fig. 1i. Rows represent individual peaks; scale bars represent signal intensity. f Genome browser visualization of the E2F2 gene body (box highlighting the promoter) for the different conditions described in panel (e). g, h Tumor volumes of two SCCOHT patient-derived xenograft (PDX) models, NRTO-1 (g) and XVOA12721 (h), both treated with Vehicle (n = 5) or A-485 (100 mg/kg, n = 5). Two-way ANOVA. ****p-value < 0.0001. n = 5 independent experiments. i Representative images (left) and quantification (right) of H3K27ac (top) and Ki67 (middle) immunohistochemistry staining, as well as Hematoxylin and Eosin (H&E, left bottom) staining for SCCOHT PDX XVOA12721 tumors treated with either Vehicle or A-485. Scale bar, 100 µM scale. Two-tailed Welch’s t test. **p-value 0.0041, ****p-value < 0.0001. Error bars, mean ± SEM. Representative results from n = 5 independent experiments.

To link HDAC1/2 and p300 function to the transcriptional interplay between SMARCA4 and p53 in regulating cell cycle progression genes, we conducted ChIP-seq to systematically profile their genomic occupancies, together with the H3K27ac dataset described in Fig. 2g, in BIN-67 cells with differential p53/SMARCA4 status. We focused on the SMARCA4-repressive, cell cycle progression gene loci (comprising 11.03% of total H3K27ac peaks as identified in Fig. 2f), where H3K27ac levels were strongly suppressed by SMARCA4 but elevated upon TP53^KO (Fig. 6e), in line with a repressing role of p53 at these loci. Consistently, we detected co-occupancies of HDAC1/2 and p53 in these same regions.

Knockout of TP53 in parental BIN-67 cells markedly reduced HDAC2 occupancy, accompanied with elevated H3K27ac mark, supporting that p53 is required for HDAC2 occupancy at these loci for repression. In line with this, SMARCA4 restoration resulted in the loss of p53 binding associated with reduced HDAC2 occupancy at the same loci. HDAC1 occupancy was also reduced by TP53^KO, in a lesser degree compared to HDAC2. Notably, this reduction of HDAC1 occupancy upon p53 loss, but not that of HDAC2, was rescued when SMARCA4 was restored, suggesting that SMARCA4 retains HDAC1 in these loci independent of p53, as restoration of SMARCA4 evicted p53 binding in these cell cycle progression gene loci. While this SMARCA4-dependent mechanism of HDAC1 binding remains to be further investigated, this could contribute to the net decrease in H3K27ac levels upon SMARCA4 restoration alone: SMARCA4 led to concomitant reduction of p300 and HDAC2 occupancy, two factors with opposing effects on acetylation (compare A4 to Ctrl). Contrasting SMARCA4 restoration alone, TP53^KO in SMARCA4-restored cells retained p300 binding in these regions, which may in part explain the persistent acetylation in these cell cycle progression gene loci in TP53^KO/A4 cells. Genome browser visualization of the E2F2 locus further illustrates this regulatory interplay (Fig. 6f). Together, these results suggest that SMARCA4 controls p300 and HDAC1/2 occupancy at cell cycle progression genes-associated chromatin to suppress H3K27ac. This leads to reduced gene expression associated with oncogenic cell cycle progression when SMARCA4 is present. Consequently, loss of SMARCA4 redirects p300 to these gene loci, opposing p53-mediated repression, resulting in elevated H3K27ac and aberrant cell cycle progression gene expression program to sustain SCCOHT growth.

Our data established that p300 plays a critical role promoting SCCOHT through aberrant cell cycle activation. To explore the therapeutic potential of targeting p300 in this aggressive cancer, we treated two different patient-derived xenograft (PDX) models of SCCOHT with A-485 or the vehicle control. Consistent with our in vitro data (Fig. 3), this single-agent treatment elicited strong anti-tumor activity (Fig. 6g, h), causing no noticeable toxicity, including stable body weight (Supplementary Fig. 7b, c). In addition, we observed reduced H3K27ac levels and the proliferation marker Ki67 upon A-485 treatment (Fig. 6i, j and Supplementary Fig. 7d). These results suggest that p300 inhibition with A-485 could represent a rational therapeutic strategy for patients affected by SCCOHT. Since PDX models lack an intact immune system, it would be valuable to develop genetically engineered mouse models of SCCOHT or humanized PDX models to better evaluate the potential clinical utility of A-485 in immunocompetent settings.

Discussion

We uncovered a mechanistic interplay between SMARCA4 loss, p300 reprogramming, and p53 suppression that converges on promoting the expression of cell cycle progression genes in SCCOHT. These findings also provide a rationale for targeting p300 as a potential treatment strategy for this lethal cancer affecting young women.

SCCOHT is an aggressive malignancy driven by SMARCA4 inactivation. Paradoxically, primary SCCOHT tumors harbor a quiescent genome and retain wild-type p53^45,46, which is mutated or lost in approximately 50% of all human primary tumors. This suggests that the tumor suppressor function of p53 is inhibited in SCCOHT. Our complementary results using MDM2 inhibitor stabilizing p53, as well as p53 knockout in SCCOHT cells, indicate that the cell cycle progression is the key target pathway suppressed by p53 in this context. By integrating gene expression and chromatin profiling, we demonstrate that both SMARCA4 and p53 control the expression of key cell cycle progression genes by regulating H3K27ac levels at their gene promoters. Loss of SMARCA4 redirects p300 to these loci to increase H3K27ac and promote their expression; concomitantly, SMARCA4 loss also increased occupancy of p53 at these promoters, which is accompanied with increased HDAC2 occupancy at these same loci to reduce H3K27ac, suppressing their gene expression. The net increase of H3K27ac at these sites and subsequent activation of corresponding cell cycle progression genes suggest that the increased p300 occupancy outplays HDAC2. Thus, SMARCA4 loss sufficiently subverts p53-mediated repression through reprogramming p300, leading to oncogenic cell cycle progression in SCCOHT (Supplementary Fig. 7e).

The tumor suppressor function of p53 is generally attributed to its role as a transcriptional activator, which induces gene expression programs that can lead to gene activation or indirect repression, the latter often mediated through its transcriptional target p21^26–30. While our results indicate that p53 occupies these cell cycle progression gene loci alongside higher HDAC2 occupancy, our current data is not sufficient to conclude the direct repression activity of p53. Similarly, even though nuclear p53-HDAC2 interaction was detected by immunoprecipitation in SCCOHT cells, whether this is direct or indirect remains to be further investigated. Notably, HDAC inhibition restored E2F2 expression and rescued SCCOHT cell viability in the presence of A-485; however, the broader contribution of HDAC2 to the repression of other cell cycle progression genes requires additional studies.

This bypass of p53-mediated repression of cell cycle progression genes caused by SMARCA4 loss may explain why primary SCCOHT tumors retain wild-type p53 despite of their aggressive nature. Notably, most atypical teratoid/rhabdoid tumors (ATRTs) - which are closely related to SCCOHT and driven by loss of another key SWI/SNF subunit, SMARCB1 - also exhibit genomic stability, retain WT p53 and are sensitive to MDM2 inhibition⁴⁷. Mechanisms similar to what we uncovered in SCCOHT may also be in play for ATRT, which warrant future studies.

Direct or indirect interactions between p300 and SWI/SNF components have been previously described^48,49. For example, the SWI/SNF complex can recruit p300 to enhancer elements, maintaining H3K27ac at regulatory regions of development and differentiation genes; loss of this interaction has been proposed to drive tumorigenesis by disrupting differentiation programs⁴⁸. In keeping with previous findings^34,35, our data also show that SMARCA4 re-expression led to a genome-wide increase in H3K27ac, particularly at enhancers. However, this is not accompanied with increased p300 occupancy (Supplementary Fig. 4b), suggesting that other histone acetyltransferases are responsible for the increased H3K27ac in these enhancers in SCCOHT. Furthermore, in ARID1A-mutant endometrial cancers, loss of ARID1A results in p300-mediated H3K27 hyperacetylation and increased chromatin accessibility specifically at super-enhancers, driving invasion phenotypes⁴⁹. We also observed that loss of SMARCA4 coincides with increased p300 occupancy at H3K27ac-marked enhancers in SCCOHT cells (Supplementary Fig. 4b). However, compared with enhancer binding, a greater proportion of H3K27ac peaks at promoters and TSSs were bound by p300 upon SMARCA4 loss in SCCOHT (Supplementary Fig. 4c), which are enriched for cell cycle progression genes essential for SCCOHT survival. This difference likely reflects the fact that ARID1A-deficient cells retain SMARCA4/2 and thus maintain fully functional SWI/SNF subcomplexes that do not contain ARID1A, including PBAF and ncBAF, whereas SCCOHT cells completely lack both SWI/SNF ATPases. Thus, in SCCOHT, SMARCA4 loss drives cell proliferation not through enhancer-mediated invasion programs, but by redirecting p300 to promoters and TSSs of cell cycle progression genes to promote their expression.

More recent studies have identified a synthetic lethal interaction between dual CBP/p300 inhibition and cancers deficient in cBAF complex activity due to loss of SMARCB1, SMARCA4/2, or the SS18-SSX fusion, where p300/CBP inhibition suppresses expression of KREMEN2, a gene upregulated upon cBAF loss^50,51. While SCCOHT cells also exhibit elevated KREMEN2 expression, its downregulation in response to A-485 treatment is variable across cell lines, suggesting that, unlike in other cBAF-deficient contexts, KREMEN2 is not a central driver in SCCOHT. Instead, we find that E2F2 and a broader set of cell cycle progression genes were consistently and robustly downregulated by both SMARCA4 restoration and p300 inhibition (A-485). Our results indicate that the oncogenic role of p300 in SMARCA4-deficient SCCOHT is primarily mediated through activation of core cell cycle transcriptional programs, which is a central hallmark of cancer. Although our findings reveal that SMARCA4 loss-mediated redirection of p300 contributes to SCCOHT tumorigenesis, the precise mechanism remains unclear. One possibility is that the chromatin remodeling activity of SMARCA4 competes with p300 binding by evicting nucleosomes at p300-targeted regions, a mechanism that warrants further investigation.

In summary, we identify a mechanism by which SMARCA4 loss reprograms p300 to activate cell cycle progression genes sustaining SCCOHT proliferation, despite the presence of wild-type p53. This mechanism may act in concert with SMARCA4 loss-driven silencing of developmental genes to promote SCCOHT development. Moreover, our findings reveal an epigenetic vulnerability in SMARCA4-deficient cancers and provide a rationale for targeting p300 as a therapeutic strategy.

Methods

All research in this study complies with all relevant ethical regulations. All biohazard protocols were approved by the Environmental Health and Safety of McGill University and the Biosafety Committee of the University of British Columbia (UBC). All animal procedures were approved by the Facility Animal Care Committee (FACC) of McGill University and the Animal Care Committee of UBC, according to guidelines of the Canadian Council on Animal Care Standards (CCAC). Studies on SCCOHT patient tumors were approved by the Institutional Review Board (IRB) at McGill University, McGill IRB # A08-M61-09B, and UBC, IRB# H18-01652.

Data mining and analysis of genome-wide CRISPR screen data

CRISPR/Cas9 knockout screening data were downloaded from the DepMap Public 21Q2 dataset (https://depmap.org/portal/). The genetic background and SMARCA4/2 expression of DepMap cell lines were derived from Cancer Cell Line Encyclopedia. Cell lines were called as SMARCA4/2-dual deficient based on literature references (BIN-67, SCCOHT-1, COV434)^11,52–54. Differential dependency was calculated by comparing CERES scores between SMARCA4/2-dual deficient cell lines versus proficient lines. Significance was assessed using an unpaired two-tailed t test.

Cell culture

293 T cells were cultured with DMEM (Dulbecco’s modified Eagle medium, Thermo Fisher Scientific, Cat# 11995-065) containing 7% fetal bovine serum (Sigma, Cat# F1051), 1% penicillin/streptomycin (Thermo Fisher Scientific, Cat# 15140-122), and 2 mM L-glutamine (Thermo Fisher Scientific, Cat# 25030-081). All other cell lines were cultured in RPMI (Roswell Park Memorial Institute 1640 Medium, Thermo Fisher Scientific, Cat# 11875-093; no pyruvate) with 7% fetal bovine serum (Sigma, Cat# F1051), 1% penicillin/streptomycin (Thermo Fisher Scientific, Cat# 15140-122), and 2 mM L-glutamine (Thermo Fisher Scientific, Cat# 25030-081). Cells were maintained at 37 °C in a humidified 5% CO₂-containing incubator, and a regular Mycoplasma test was performed using Mycoalert Detection Kit (Lonza, Cat # LT07-318). BIN-67: Dr. B. Vanderhyden (Ottawa Hospital Research Institute, Ottawa) originally from Dr. S.R. Goldring (Hospital for Special Surgery, New York, originally derived from patients with ovarian carcinoma treated at the Dana-Farber Cancer Institute (Boston, MA)); SCCOHT-1: Dr. R. Hass (Medical University Hannover, Hannover, generated by Dr. R. Hass); IOSE80: Dr. N. Auersperg (The University of British Columbia, Vancouver); FT237, FT190: Dr. T.G. Shepherd (The Mary & John Knight Translational Ovarian Cancer Research Unit, Ontario, FT190 was originally provided by R. Drapkin, University of Pennsylvania, Philadelphia, PA); 293 T: ATCC. COV434: Sigma, 07071909; OVCAR3: Dr. Nelly Auersperg (University of British Columbia, Vancouver, originally from Thomas C Hamilton, Fox Chase Cancer Center). All cell lines have been validated by Short-Tandem Repeat profiling.

Compounds and antibodies

A-485 and Idasanutlin have been purchased from Med Chem Express (Houston, Texas, USA). Antibodies against HSP90 (H-114, 1:1,0000), β-Actin (Cat# sc-47778, 1:1,0000), p53 (sc-126, 1:4000), E2F2 (sc-9967, 1:1000) were from Santa Cruz Biotechnology (Dallas, TX, USA); antibodies against SMARCA2 (Cat# 11996, 1:1000), p21 (Cat# 2947S, 1:1000), pRB S795 (Cat# 9301, 1:1000) were from Cell Signaling (Danvers, MA, USA); antibodies against SMARCA4 (A300-813A, 1:1000 and ab110641, 1:5000) were from Bethyl Laboratories (Montgomery, TX, USA) and Abcam (Tornoto, ON, Canada), respectively; Antibodies against p300 (ab10485, 1:1000; and ab14984 1:1000), H3K27ac (ab4729 1:5000) and HDAC2 (ab124974, 1:1000) were from Abcam. HDAC1 (40967, 1:1000) was from Active Motif. Antibodies for immunohistochemistry are listed in the corresponding method section below.

Plasmids, Lentivirus production and infection

Individual shRNA vectors used were from the Mission TRC library (Sigma) provided by McGill Platform for Cellular Perturbation (MPCP) at McGill University: shSMARCA2#1 (TRCN0000358828); shSMARCA2#2 (TRCN0000020333); shEP300#1 (TRCN0000039884), shEP300#2 (TRCN0000009883), shEP300#3 (TRCN0000039885) shE2F2#1 (TRCN0000013798), shE2F2#2 (TRCN0000013799), shE2F2#3 (TRCN0000013800). sgRNA (ATGGTGCTGACCCCCAGGCCTT) targeting SMARCA4 and previously validated sgRNA(GAGCGCTGCTCAGATAGCGA) targeting TP53⁵⁵ were cloned into pLentiCRISPRv2 which was from Addgene (Cat# 52961). pReceiver control vector pReceiver-SMARCA4 was purchased from GeneCopoeia. pIN20 and pIN20-SMARCA4 were kindly provided by Dr. Jannik N. Andersen (The University of Texas, MD Anderson Cancer Center).

Lentiviral transduction was performed using the protocol as described at http://www.broadinstitute.org/rnai/public/resources/protocols. Briefly, 2.5 × 10⁶ 293 T cells were seeded in a six-well plate with 2 mL DMEM medium per well. 8 hours later, cells were transfected with the indicated lentiviral constructs, the packaging (psPAX2) and envelope (pMD2.G) plasmid by CaCl₂. Virus containing medium were collected (24 and 36 h after transfection) and stored at − 80 °C. Infected cells (~ 8 h for infection and ~ 20 h for recovery) were selected in medium containing puromycin or blasticidin for 2–3 days and collected immediately for the experiments.

Colony formation assays

Single-cell suspensions of all cell lines were counted and plated into 6-well plates at a density of 0.5–8 × 10⁴ cells per well. Cells were cultured in a medium containing the indicated drugs for 10–14 days (refreshed every 3 days). At the endpoints, cells were fixed with 4% formaldehyde in PBS, stained with crystal violet (0.1%w/v in water) and photographed.

Cell viability assays

Cultured cells were plated into 96-well plates (0.5 k–6 k cells per well) and treated with medium containing the indicated drugs the next day. Cells were cultured for 4–7 days (refreshed twice a week), and cell viability was measured using the CellTiter-Blue® Viability Assay (Promega) by measuring the fluorescence (560/590 nm) in a microplate reader. Relative cell viability was calculated by normalizing the absorbance or fluorescence to that of the vehicle-treated controls after background subtraction. For the A-485 and trichostatin A (TSA) combination experiment, normalization was performed within each arm (A-485 conditions normalized to untreated controls; TSA + A-485 conditions normalized to TSA-treated controls).

Protein lysate preparation and immunoblots

Cells were lysed with protein sample buffer, heated at 95 °C for 5 min, and processed with Novex® NuPAGE® Gel Electrophoresis Systems (Thermo Fisher Scientific). HSP90 and beta-actin served as loading controls.

RNA isolation and qRT-PCR

Total RNA was isolated using Trizol (Invitrogen) and converted to cDNAs using the Maxima First Strand cDNA Synthesis Kit (Thermo Scientific). Quantitative real-time reverse transcription PCR (qRT-PCR) was carried out using SYBR® Green master mix (Roche) according to manufacturer protocols. Relative mRNA levels of the indicated genes were normalized to the housekeeping gene ACTB. The sequences of the primers used for qRT-PCR are as follows:

ACTB_Forward (Fwd), GTTGTCGACGACGAGCG;

ACTB_Reverse (Rev), GCACAGAGCCTCGCCTT;

E2F2_Forward (AGTGCCCGACAGGACTGAGGAC)

E2F2_Reverse (GCACCTCCTCTGGGCACAGGTA)

Chromatin immunoprecipitation and sequencing

Cells were fixed in complete media with 0.3% formaldehyde for 30 minutes at 4 °C and then quenched by adding 0.125 mol/L glycine for 5 min at room temperature and 15 min on ice. Fixed BIN-67 or BIN-67 TP53^KO cells ± restoration of SMARCA4 were pelleted and washed once with 1 × PBS before snap-freezing on dry ice. Antibodies against H3K27ac (ab4729, Abcam, 2 µg/ml), HDAC2 (ab124974, Abcam, 10 µg/ml), p300 (ab14984, Abcam, 15 µg/ml), IgG (Abcam, ab37415, 15 µg/ml), HDAC1 (40967, Active Motif, 10 µg/ml), p53 (sc-126, Santa Cruz Biotechnology, 10 µg/ml) were used for chromatin immunoprecipitation (ChIP) experiments following a protocol with micrococcal nuclease (MNase)⁵⁶. Briefly, the cell pellet was resuspended in 1 mL buffer A (25 mmol/L HEPES–NaOH, pH 7.5; 10 mmol/L KCl; 0.1% NP-40; and 1.5 mmol/L MgCl2) and homogenized. Nine volumes of buffer B (15 mmol/L HEPES–NaOH, pH 7.9; 15 mmol/L NaCl; 60 mmol/L KCl; 0.5 mmol/L phenylmethylsulfonylfluoride; and 0.32 mol/L sucrose) were added, and the mixture was centrifuged at 450 × g (at 4 °C for 7 minutes) to collect nuclei. The nuclei were resuspended in 1 mL buffer B and 3.3 µL of 1 mol/L CaCl2 and then prewarmed at 37 °C for 15 min. MNase (1 µL, New England Biolabs, M0247S) was added and incubated at 37 °C for 15 min. The reaction was quenched with 10 µmol/L EGTA on ice for 5 min. One volume of 2 × buffer C (20 mmol/L Tris–HCl, pH 8.0; 200 mmol/L NaCl; 2 mmol/L EDTA; 1 mmol/L EGTA; 0.2% Na-deoxycholate; and 1% N-lauroylsarcosine) and Triton X-100 (final 1%) was added. Samples were then passed through a 21 G needle five times and centrifuged at 13,000 × g (4 °C) for 15 min. Antibodies were added to the lysate for overnight incubation at 4 °C. Protein G Magnetic Dynabeads (Thermo Fisher Scientific) were used for pulldown. Immunoprecipitated chromatin bound to Dynabeads was washed four times with RIPA buffer (50 mmol/L HEPES-KOH, pH 7.5; 500 mmol/L LiCl; 1 mmol/L EDTA; 1% NP-40; and 0.7% Na-deoxycholate). Chromatin was eluted from Dynabeads in 100 µL elution buffer (50 mmol/L Tris-HCl, pH 8.0; 10 mmol/L EDTA; and 1% SDS) by incubating at 65 °C for 30 min. Immunoprecipitated samples and input were incubated overnight at 65 °C to denature formaldehyde cross-linking. Samples were then treated with RNase A (Thermo Fisher Scientific), followed by proteinase K (Sigma-Aldrich), before extraction using a Zymo kit (D5205). All ChIP sequencing (ChIP-seq) libraries were constructed using the NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs) according to the manufacturer’s instructions.

ChIP-seq data analysis

Publicly available ChIP-seq data for BIN-67 cells with or without SMARCA4 restoration (GSE117735³⁵) were analyzed in addition to the conditions described above. ChIP-seq data were mapped to the hg38 genome using Bowtie2 (v2.4.6)⁵⁷ with default parameters. Duplicate reads were removed with SAMtools (v1.17)⁵⁸, and peaks were called using MACS2 (v2.2.8)⁵⁹. Metaplots and heatmaps were generated with deepTools (v3.5.1)⁶⁰, with normalization performed using the bamCoverage or bamCompare functions. H–G regions (Fig. 2f) were defined based on fold-change (FC) values of MACS2 signal scores, with peaks stratified according to the following thresholds: FC in SMARCA4 versus control ≤ –0.5, FC in TP53^KO versus control ≥ – 0.2, and FC in TP53^KO and SMARCA4 versus control ≥ – 0.2. ChIP-seq density was calculated using HOMER (v4.11)⁶¹. Unique and overlapping regions across conditions were identified with the “multiinter” or “intersect” functions of BEDTools (v2.30.0). Enhancers were defined as H3K27ac peaks annotated by HOMER, excluding promoter and TSS regions. Promoter/TSS regions were annotated separately using HOMER. Gene assignments were performed with GREAT, using either default settings or the two nearest genes within ± 2 kb of the peak center. Genome browser snapshots were visualized in IGV (v2.16)⁶².

ChromHMM

ChromHMM³⁹ was applied to segment the hg38 genome in control and SMARCA4-overexpressing BIN-67 cells based on combinatorial chromatin features. Seven datasets (total RNA, ATAC, H3K4me1, H3K27ac, p300, H3K27me3, and H3K9me1) were first binarized from aligned BAM files. The genome was segmented into 200-bp intervals based on state classifications. To enable direct comparison between conditions, models were trained using the “concatenated” mode, which generates a shared chromatin state framework across datasets. Models were tested with 5–25 states, and the optimal number was chosen based on manual evaluation of biological interpretability and distinction between states. A 5-state model was selected for downstream analyses. Biological functions of each state were inferred based on combinatorial chromatin features and genome ontology annotation. Emission parameters were reordered to group states by putative functional category. Chromatin state BED files were imported into R as non-overlapping GenomicRanges objects for enrichment analyses, differential state comparisons, and visualization. In chromatin state change plots, for each state-state change, the size of the circles represents the relative amount of that state change compared to the initial genome-wide state representation ([genomic bp initial → final] / [genomic bp initial]), and the color indicates the log₂ fold change (log₂FC) of differentially bound peaks.

Transcriptome analysis

Transcriptomic profiling was performed in BIN-67 and BIN-67 TP53KO cells under the following conditions: BIN-67 cells ± SMARCA4 restoration, BIN-67 TP53KO cells ± SMARCA4 restoration, BIN-67 cells ± A-485 treatment (1 µM, 24 h), BIN-67 TP53KO cells ± A-485 treatment (1 µM, 24 h), and BIN-67 cells ± idasanutlin treatment (0.5 µM, 24 h). Publicly available transcriptomic data were also included: BIN-67 cells ± SMARCA4 restoration (GSE117735³⁵ and BIN-67, COV-434, and SCCOHT-1 cells ± SMARCA4 restoration (GSE151026³⁴). Sequencing files were downloaded from the Sequence Read Archive (SRA) and aligned to the human reference genome (hg38) using STAR (v2.6.1c)⁶³. Gene expression levels were quantified using Homer or HTseq⁶⁴ with Gencode gene annotation (GTF) files. Differentially expressed genes were identified using Edger (v4.6.3)⁶⁵. Log₂ fold changes for RNA-seq conditions were used for clustering with a simple clustering method and Manhattan clustering distance. Heatmaps were plotted with pheatmap (v1.0.13). Gene ontology analysis was performed with clusterProfiler (v4.16.0)⁶⁶, which calculates p-values by one-sided weighted Kolmogorov-Smirnov test.

Cell cycle analysis

BIN-67 and BIN-67 TP53^KO cells were treated with A-485 (1 µM, 24 h) and Idasanutlin (0.5 µM, 24 h) before harvesting. Cells were then washed with PBS containing 1% FBS and fixed with 1 ml cold 70% ethanol for 30 min on ice. After washing twice with PBS, cells were treated with 25 µg ml−1 Ribonuclease A and stained with 50 µg ml−1 propidium iodide solution for 10 min. Cell cycle distribution was analyzed using a BD FACS Aria III flow cytometer (BD Biosciences) according to the manufacturer’s instructions. Doublets were excluded based on forward and side scatter parameters, and cell cycle phases were quantified using the Watson pragmatic algorithm implemented in FlowJo software.

Immunohistochemistry

Formalin-fixed paraffin-embedded tumor tissue sections were cut at 4 μm thickness and were subjected to automated immunohistochemistry (IHC) on Ventana (Discovery Ultra) and using optimized protocols. The primary antibodies used in this study include anti-Ki67 (Abcam #16667, 1:100), and anti-H3K27ac (Abcam #4729, 1:100). Upon developing with DAB, the slides were then counterstained with haematoxylin, dehydrated, cleared and mounted with a synthetic mounting medium. The slides were digitalized using an Aperio scanner at 20x magnification. Images were analyzed and quantified using Aperio ImageScope (Leica Biosystems). The results were reported as a percentage of tissue surface area positive for protein of interest. Areas with low tumor cellularity and artifacts were not included in the analysis.

Survival analysis

Survival analyses of TCGA (n  = 10967) were performed using data derived from cBioportal (https://www.cbioportal.org/). This dataset was analyzed by first separating patients into four groups (SMARCA4 deletion/TP53 normal: SMARCA4 deletion/TP53 deletion: SMARCA4 normal/ TP53 normal: SMARCA4 normal/ TP53 deletion) according to SMARCA4 copy number status. Then, the overall survival data for each group was plotted, and statistics were analyzed with Kaplan-Meier simple survival analysis in GraphPad.

Mouse xenografts and in vivo drug efficacy studies

Written informed consent for the use of human samples was obtained to generate all human cell line-derived xenografts and PDXs. Animal experiments were performed according to standards outlined in the Canadian Council on Animal Care Standards (CCAC) and the Animals for Research Act, R.S.O. 1990, Chapter c. A.22, and by following internationally recognized guidelines on animal welfare. All animal procedures (Animal Use Protocol #7407 (McGill) and # A22-0005 (UBC)) were approved by the Institutional Animal Care Committee according to guidelines of the CCAC. Since all human cell line-derived xenografts and PDXs were derived from female patients, all animal experiments in this study used female mice. Housing conditions: Temperature: 16 degrees min − 24 degrees max; Humidity: 15% low − 60% high; Photoperiod: 7am–7 pm light, 7pm–7am dark (McGill); 6a–6 pm light, 6pm–6am dark (UBC). The animal experiments carried out at Goodman Cancer Research Institute of McGill University used 8–12-week-old in-house bred female NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice; the animal experiments carried out at British Columbia Cancer Research Institute (BCCRI) used 7–9 week-old in house bred female NRG (NOD.Rag1KO.IL2RγcKO) mice. SCCOHT PDX NRTO-1 and XVOA12721 were established, viably preserved, and studied at Goodman Cancer Research Institute of McGill University and BCCRI, respectively. Tumors were cut into pieces and then inserted into a pocket in the subcutaneous space of NSG mice. Mice were randomized to treatment arms once the average tumor volume reached 100 mm³. A-485 was prepared fresh daily from a DMSO stock in a vehicle consisting of 5% DMSO, 40% PEG-300, 5% Tween-80, and 50% saline, and administered by intraperitoneal injection at 100 mg/kg twice daily on a 5-on/1-off schedule. Tumor volume and mouse weight were measured twice to thrice weekly. Tumor volume was calculated as length*(width)²*0.52. The persons who performed all the tumor measurements were blinded to the treatment information. The maximal tumor size permitted by our Institutional Animal Care Committees is 2500 mm³, which was not exceeded in our experiments. Animals were euthanized by inhalational anesthesia with 5% isoflurane, followed by CO₂ exposure. Death was confirmed by cessation of respiration and absence of reflexes, after which cervical dislocation was performed as a secondary physical method to ensure death.

Immunoprecipitation

For HDAC2 coimmunoprecipitation, BIN-67 and BIN-67 TP53^KO cells were lysed in ice-cold hypotonic buffer for 15 min on ice (20 mmol/L Tris, pH 7.6, 10 mmol/L KCl, 1.5 mmol/L MgCl2, and EDTA-free protease and phosphatase inhibitors; Roche). NP-40 was added to a final concentration of 0.5%, and the extracts were centrifuged at 14,000 × g for 1 min at 4 °C. The nuclear pellet was resuspended in nuclease buffer (20 mmol/L Tris, pH 7.6, 150 mmol/L NaCl, 1.5 mmol/L MgCl2, 2.5 mmol/L CaCl2, and 0.5 µL phenylmethylsulfonyl fluoride 100 mmol/L) and treated with Micrococcal nuclease (MNase; New England Biolabs) for 2 h at 4 °C. The lysates were clarified by centrifugation at 14,000 × g for 10 min at 4 °C and diluted in immunoprecipitation buffer (50 mmol/L Tris, pH 7.6, 150 mmol/L NaCl, 1% NP-40, and protease and phosphatase inhibitors), and protein concentrations were determined using the Bradford reagent (Bio-Rad). Immunoprecipitations were carried out on 400 µg of protein extracts using 2 µg of the HDAC2 antibody (ab124974, Abcam, 2 µg/ml) and incubated overnight at 4 °C with rotation. Protein G magnetic beads were added and incubated at 4 °C for 2 h. The precipitated proteins were washed three times with immunoprecipitation buffer, eluted with SDS loading buffer at 95 °C for 10 min. The HDAC2 ability to immunoprecipitated p53 was analyzed by Western blot using the Novex NuPAGE Gel Electrophoresis System. BIN-67 TP53^KO cells served as a negative control instead of IgG. To avoid p53 signal to be confounded with the IgG heavy chain, we used the secondary antibody TrueBlot® ULTRA: Anti-Rabbit IgG HRP (Rockland, 18-8816-31, 1:1000), which enables detection of immunoblotted p53, without hindrance by interfering immunoprecipitating immunoglobulin heavy and light chains.

Statistics and reproducibility

GraphPad Prism 8 software was used to generate graphs and statistical analyses. Statistical significance was determined by one- and two-way ANOVA, Student’s t test, log-rank test. Methods for statistical tests, the exact value of n, and definition of error bars were indicated in figure legends, *p   <   0.05, **p   <   0.01, ***p   <   0.001, and ****p   <   0.0001. For simplicity, one-way ANOVA Brown–Forsythe and Welch tests followed by Dunnett’s test for multiple comparisons is referred as “one-way ANOVA corrected for multiple comparisons” in the figure legends. All experiments have been reproduced in at least two independent experiments unless otherwise specified in the figure legends. All immunoblots and images shown are representative of these independent experiments. In all instances, all the attempts at replicating the experiments produced similar results. No data were excluded. Sample sizes for in vitro experiments (at least three independent replications) were chosen based on the standard practices of the field. For public data set analyses, sample sizes were determined based on what was available publicaly, authors had no influence over how sample sizes were chosen for the design of these studies. Regarding the in vivo study, no statistical method was used to predetermine sample size, however, at least 5 tumors per group were used to have appropriate statistical power based on previous experience and samples were randomly distributed into groups. The person administering the drug or placebo was not blinded to the drug condition due to the complexity of the experiments and limited personnel. However, the subsequent measurements were blinded to the treatment information.

Reporting summary

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

Author contributions

G.A., K.L., A.A., S.T., Y.X., A.M., V.P., M.H., and G.M. carried-out experiments and data analysis. G.A. and A.A. performed statistical analysis. G.A. and G.M. performed bioinformatic analysis. K.P. and L.F. contributed the samples and provided pathology expertise. G.A., Y.W., and S. H. wrote the manuscript with inputs from all authors. M.P., W.D.F., Y.W., and S.H. supervised the experiments. Y.W. and S.H. conceived and oversaw the study. All authors read and approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks Benjamin Bitler, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by Canadian Institute of Health Research (CIHR) grants PJT-156233 (S.H.), PJT-438303 (S.H.), FDN-148390 (W.D.F.), and PJT-178179 (Y.W.). Y.W. is a recipient of an early career investigator award from the Ovarian Cancer Research Alliance (ECIG-2025-3-2014). S.H. was supported by a Canada Research Chair in Functional Genomics. K.L. and G. A. are supported by Canderel Studentship and Emma Ciani Studentship. We also appreciate the generous support from the British Columbia Cancer Foundation and the VGH/UBC Hospital Foundation.

Data availability

The ChIP-seq in BIN-67 expressing control vector and inducible SMARCA4 data generated in this study have been deposited in the European Nucleotide Archive (ENA) database under accession code ERP181298. The ChIP-seq in BIN-67 expressing control vector, SMARCA4, TP53^KO and SMARCA4 plus TP53^KO Control, SMARCA4 restoration data generated in this study have been deposited in the ENA database under accession code ERP181310. The RNA-seq for BIN-67, COV434 and SCCOHT-1 cells treated with A-485 data generated in this study have been deposited in the ENA database under accession code ERP181311. The RNA-seq for BIN-67 cells expressing control vector or TP53^KO with or without A-485 treatment data generated in this study have been deposited in the ENA database under accession code ERP181313. The RNA-seq for BIN-67 cells expressing control vector (Ctrl), SMARCA4, TP53^KO and SMARCA4 plus TP53^KO Control and idasanutlin treatment data generated in this study have been deposited in the ENA database under accession code ERP181314. The RNA-seq data in COV434 and SCCOHT-1 with or without idasanutlin treatment data generated in this study have been deposited in the ENA database under accession code ERP192500. The ChIP-seq and RNA-seq in BIN-67 cells for SMARCA4 restoration or T910M mutant SMARCA4 expression data reused in this study are available in the GEO database under accession codes GSE117734 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE117734] and GSE117311 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE117311], respectively³⁵. The RNA-seq in BIN-67 cells for SMARCA4 restoration data reused in this study are available in the GEO database under accession codes GSE120297 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE120297]²¹. The RNA-seq in BIN-67, COV434 and SCCOHT-1 with or without SMARCA4 inducible restoration data reused in this study are available in the GEO database under accession codes GSE151026 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE151026]³⁴. Collectively, all the reused datasets are publicly available. Gene dependency scores were analyzed using CRISPR knockout datasets available from the DepMap portal [https://depmap.org]. The remaining data are available within the Article, Supplementary Information or Source Data file. Source data are also available in the Source Data file [10.6084/m9.figshare.32180532]. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74204-8.