Abstract Body: Introduction: Coronary artery disease (CAD) is a leading cause of death worldwide. It is now generally accepted that residual risk for CAD resides in genetic variation that regulates cellular processes of smooth muscle cells (SMCs) in the vessel wall.

Methods: We applied CRISPRi Perturb-seq using split-pool barcoding (SPLiT-seq) in immortalized human coronary artery SMCs, targeting 971 genome wide association study (GWAS) nominated genes with 3,163 guides across 954,000 cells. In vitro Perturb-seq transcriptomic effects were measured using single-cell RNA-seq. Explanted coronary arteries from cardiac transplant recipients was collected to generate a high quality scMultiome human coronary artery cell atlas (n=9). Together, this provided orthogonal references to map gene programs using consensus non-negative matrix factorization (cNMF) with robust cross-validation of gene programs. In vivo Perturb-seq via SMC-targeted AAV delivery is performed for mechanistic validation and characterization of in vivo effects.

Results: cNMF decomposition performed on the Perturb-seq and human scMultiome atlas revealed 50 in vitro, and 40 in vivo gene programs, respectively. In total, 84% of our candidate genes demonstrated statistically robust knockdown. This allowed systematic integration of GWAS variants to gene linking followed by analysis of perturbation-associated differential gene signatures with SMC-specificity and CAD risk gene enrichment that converged across both cNMF references. Together, this enabled quantitative ranking of functional disease genes by their impact on transcriptional circuits. To extend upon this framework, SMC-targeted in vivo Perturb-seq demonstrated robust knockdown across 19 target genes using 123 guides, enabling rapid simultaneous characterization of cellular functions. Integration of these analyses highlighted convergence of CAD-associated perturbations on risk associated vascular regulatory programs such as TGFB signaling and nominated ZEB1 as a high-confidence mediator of CAD genetic risk. Transcriptomic analysis of in vivo effects revealed a shift in SMC cellular state marked by ectopic activation of epithelial markers and upregulation of immune signaling.

Conclusion: We combine SPLiT-seq CRISPRi screening with multi-modal gene program references for robust hypothesis generation and in vivo Perturb-seq for rapid functional validation of CAD risk associated loci, identifying ZEB1 as a novel regulator of SMC identity and disease risk.