Integration of spatial transcriptomics, genetics, and histomorphology for causal inference in atherosclerotic cardiovascular disease

Despite success of lipid lowering therapies, atherosclerotic cardiovascular disease (CVD) remains the leading cause of death worldwide. We use human genetics coupled to single cell profiling of atherosclerotic lesions to provide insights into new molecular and cell-specific mechanisms of plaque vulnerability and CVD risk. Although plaque histology classifications have identified key characteristics (e.g., thin fibrous caps) that relate to clinical events, traditional histology may not capture the molecular complexity and full informativeness of lesion histology. There is great need for evaluation and integration of new histomorphological (HM), cellular, genomic and spatial features of plaque vulnerability in human lesions as well as in mouse models, which fail to recapitulate key lesion hallmarks of clinical events in humans. Our preliminary data in human and mouse atherosclerotic lesions suggest that advanced machine learning (ML) algorithms can identify novel HM and spatial transcriptomic (ST) features of plaque vulnerability. We propose to: (1) in Aim 1, apply our novel hierarchical image feature ML extraction approach to high-resolution histology image data from unstable (n=300) vs. stable (n=300) carotid as well as 200 coronary and 200 peripheral lesions from the Munich Vascular Biobank (MVB) to detect novel lesion features and regions that associate with unstable vs. stable plaque status and coronary plaque types as well as major adverse cardiovascular events (MACE); (2) in Aim, 2 use the 10x Genomic Xenium platform, to perform ST and histology in the same sections of a subset of MVB lesions (n=64), map ST data to HM features, and deploy a novel super-resolution gene expression prediction algorithm to expand ST data to all 1000 MVB plaques. We will then test whether gene expression and cell-subtype features associate with MACE and if this is independent of HM features; (3) in Aim 3, integrate existing MVB SNP data, lesion bulk RNA-seq data and scRNA-seq to perform cell deconvolution analysis and infer cell subtype proportions in bulk RNA-seq data. These data will be used to identify SNPs for cell subtype-specific expression (eQTLs) and for HM (hmQTLs), and through genetic colocalization analyses with CVD GWAS SNPs to infer causality for specific genes, cell subtypes and HM features in CVD events. Each aim will also map mouse lesion HM and ST features to human data to identify features in mouse lesion that are shared with clinically-relevant HM and ST features in human plaques. This will advance greater translational fidelity of mouse models of atherosclerosis. Finally, because of emerging prominence for smooth muscle cell (SMC)-derived cells and their regulatory genes in plaque vulnerability and GWAS loci for CVD, human and mouse models will incorporate SMC-lineage tracing to validate SMC-cell origin in HM and ST features of plaque vulnerability and CVD. Our proposal provides a unique integrative platform, supported by human genetics, to implicate novel genes, cell subtypes and HM features in lesion vulnerability and CVD risk. This work will advance mechanism-based therapeutics for CVD in an era of residual CVD risk despite aggressive LDL-C lowering in high-risk patients.