Identifying and Validating Genetic Variants Associated with Cardiac Differentiation and Disease

About This Grant

PROJECT SUMMARY Congenital heart disease (CHD) is a heritable, incurable group of birth defects that impact 1% of infants that is likely caused by mutations in the gene regulatory networks (GRNs) that govern heart development. However, the development of therapeutics is hindered by our incomplete understanding of the genetic etiologies underlying CHD. Discovery of the intrinsic GRNs that govern the vast array of cell types essential for heart development is the first step towards the development of therapeutic interventions. By integrating advances in artificial intelligence with genome engineering, emerging evidence from our team suggests that CHD is an oligogenic disorder driven by latent genetic interactions (GIs) between coding and noncoding variants that affect differentiation of multiple non-myocyte cell types critical for organogenesis. In preliminary studies, we employed a base-resolution neural network trained on human fetal heart chromatin accessibility data to identify noncoding variation from CHD probands from multiple cohorts, including NIH-funded programs, that is predicted to disrupt cis-regulatory element (cRE) activity suggesting these regions can act as enhancers. This analysis identified thousands of noncoding variants that are predicted to tune the regulatory activity of the cRE in discrete cell types. To overcome experimental hurdles faced by prior CHD studies that prevented broad association of genetic variants to discrete cardiac phenotypes, we next generated a time-resolved single cell multiome (RNA+ATAC-seq) that includes 200,000 cells during human induced pluripotent stem cell (hiPSC) derived cardiac organoid (cardioid) differentiation. We confirmed this system recapitulates the dynamic, nascent stages of heart development and includes cell states found in the early embryo, creating a robust model system for dissecting developmental mechanisms relevant to CHD. In this proposal, our goal is to connect machine learning tools with the cardioid differentiation model to nominate, characterize and validate complex genetic mechanisms essential for heart development and disease. First, we will employ a single-cell massively parallel reporter assay in cardioids to quantify cell type-specific effects of prioritized noncoding variants, as well as the directionality and magnitude of the variant’s effect (Aim 1). Next, we will infer novel disease-causing genes by creating a cell type and differentiation stage resolved enhancer-gene linkage atlas from our cardioid single-cell multiome data (Aim 2). Finally, we will employ naturally occurring ancestral variation from divergent source populations to identify GIs involved in heart development and disease (Aim 3). Variants uncovered in each aim will subsequently be inserted into endogenous loci in hiPSCs to identify variants that independently disrupt cardiac differentiation versus those that require perturbation of a known CHD signaling axis to perturb differentiation. This proposal rigorously integrates cutting-edge advancements in machine learning and functional genomics to elucidate fundamental GRNs implicated in cardiac differentiation, representing a crucial initial step toward understanding the genetic etiologies of CHD.