High-content CRISPR screens link coronary artery disease genes to endothelial cell programs [RNAseq]

However, identifying such convergence remains challenging: each GWAS locus can have 2-20 candidate genes, the cellular programs a gene participates in are difficult to define in an unbiased fashion, and it remains unclear which genes and programs would be likely to influence disease risk. Here, we explored a new approach to address this challenge, by creating an unbiased catalog of gene programs and their regulators in endothelial cells to link variants to functions for coronary artery disease (CAD). To do so, we applied CRISPRi-Perturb-seq to knock down all expressed genes within 500 Kb of all CAD GWAS loci (2,285 genes in total) and measure their effects on the transcriptome using single-cell RNA-seq. We used consensus non-negative matrix factorization to define 60 gene expression programs—including core cellular programs, such as ribosome biogenesis, and endothelial cell-specific programs, such as flow response and angiogenesis—and link these programs to upstream regulators including transcription factors, chromatin regulators, metabolic enzymes, and signaling cascades. By combining this gene-to-program catalog with variant-to-gene maps, we find that candidate CAD genes converge onto 6 interrelated gene programs, together involving known and novel genes in 39 of 229 CAD GWAS loci. Analysis of these programs revealed that the cerebral cavernous malformations (CCM) complex—whose potential connection to CAD has not been previously explored—acts upstream to regulate other CAD genes involved in cytoskeletal organization, extracellular matrix remodeling, and cell migration. The strongest regulator of these programs is TLNRD1, a highly conserved but poorly studied gene that we show acts in the CCM pathway and regulates actin organization and endothelial cell barrier function. Together, our study nominates new genes that likely influence risk for CAD, identifies convergence of CAD risk loci into certain gene programs in endothelial cells, and demonstrates a generalizable strategy to catalog gene programs to connect disease variants to functions.