Functional Characterization of mSWI/SNF Complexes using Perturb-seq

Abstract

Mammalian SWI/SNF (mSWI/SNF or BAF) complexes are multi-subunit chromatin remodeling machines that use the power derived from ATP hydrolysis to mobilize nucleosomes and facilitate chromatin accessibility. mSWI/SNF complexes exist in three forms: canonical BAF (cBAF), polybromo-associated BAF (PBAF), or non-canonical BAF (ncBAF), each of which has a distinct subunit composition, genomic localization pattern, and activity on chromatin. mSWI/SNF subunits are mutated in greater than 20% of human cancers as well as in a variety of neurodevelopmental disorders, which has motivated the study of their activities across a wide range of disease settings. While studies have begun to define the roles for individual mSWI/SNF subunits in various specific tissue or disease contexts, this heterogeneity in genetic backgrounds has hindered efforts to comparatively assess mSWI/SNF subunit functions. To this end, we performed a CRISPR-Cas9 knockout screen followed by single cell RNA sequencing (Perturb-seq), to probe both the individual functions of subunits as well as their combinatorial logic upon perturbation of multiple subunits. We combined these insights with single cell chromatin accessibility profiles and bulk chromatin binding profiles to define complex-, module-, and subunit-specific roles in the regulation of diverse gene sets and pathways. We further identify the contribution of each subunit (both subcomplex-specific and shared) to unique subcomplex activity using a logistic regression classifier. Finally, we mined RNA-seq profiles of TCGA-cataloged primary tumors and identified tumors with high correlation to mSWI/SNF perturbation signatures that lack mutations in any BAF subunit. We probed their mutational landscapes and identified transcription factors that mediate expression of the highly correlated gene sets to identify potential convergent mechanisms of gene expression signatures. In addition to this large-scale effort, a focused study of the roles of the paralogous mSWI/SNF subunits ARID1A and ARID1B was performed to elucidate their similarities and differences in functional domains, gene targeting roles, and effects on mSWI/SNF complex composition. Using CRISPR-Cas9 domain scanning, the critical role for a previously uncharacterized N-terminal region of ARID1B was identified. Additionally, we found a preference for SMARCD paralog integration based upon ARID1 paralog integration, as well as identified different chromatin binding patterns of ARID1A- and ARID1B-containing complexes. Taken together, this body of work represents a multidisciplinary effort to comparatively and individually assess the functional roles of mSWI/SNF subunits.