Integrated multiomic profiling reveals SWI-SNF subunit-specific pathway alterations and targetable vulnerabilities

Background: Mutations in subunits of SWI/SNF chromatin remodeling complex frequently occur in cancer and no targeted therapy is currently available for this patient population. To systematically characterize novel targetable vulnerabilities in SWI/SNF-deficient cancers, we integrated multiomics profiling and drug screening of tumor cell lines with publicly available molecular profiling and CRISPR screen, using newly developed bioinformatic methodologies. Methods: We generated in-house transcriptome and proteome profiling with high-throughput drug screening on a panel of isogenic HAP1 cell lines knocked-out (KO) for genes encoding SWI/SNF subunits or other chromatin remodelers. We further acquired publicly available molecular profiling and CRISPR-screen datasets from the TCGA and DepMap projects, respectively, and integrated all results. Gene set enrichment analyses were performed using an optimized pipeline based on (i) a pruned version of the Reactome pathway database, (ii) combining ROntoTools and GSEA algorithms, which scored as top performant strategy. Synthetic lethal interactions (SLi) from the DepMap CRISPR screening were identified using a new benchmarked SLi algorithm, based on expression levels of genes on interest, to consider functional deficiencies. Results: Metabolism of proteins , which had not previously been linked to SWI/SNF, was the most frequently dysregulated pathway category both in HAP1 SWI/SNF-KO mutant cell lines and in the DepMap analysis comparing models with low and high SWI/SNF subunit expression; it was the third most dysregulated in SWI/SNF-defective TCGA patient tumors (after Signal transduction and Immune system , potentially due to the presence of an immune microenvironment). High-throughput drug screen identified multiple pharmacological vulnerabilities, including inhibitors of the histone acetyltransferase CBP/EP300 or mitochondrial respiration, which were selectively cytotoxic in SWI/SNF-defective models. Our new SLi algorithm, also independently identified genetic synthetic lethality between SWI/SNF defects and EP300 or mitochondrial respiration genes. These orthogonal genetic and chemical vulnerabilities were revalidated experimentally in histotype-relevant SMARCA4-isogenic models, using two highly selective EP300 inhibitors and the mitochondrial respiratory chain complex III inhibitor Antimycin A. Conclusion: Integrating multi-omics profiling with high-throughput drug- and genetic- screening, respectively analyzed using an optimized enrichment pipeline and a newly developed SLi algorithm, enabled the identification of actionable genetic dependencies in SWI/SNF-defective cancers. These predicted synthetic lethal interactions were experimentally validated in relevant models, supporting their clinical relevance. JBS and CA contributed equally. SPV and AB contributed equally.