Computational deconvolution of atherosclerotic plaques cell composition and its clinical association

Abstract

Atherosclerotic plaques are highly complex and heterogeneous, with a large number of extracellular components and cell types involved. Traditional analysis, like histology assessment of the plaques, have been useful to characterize their vulnerability, leading to a binary risk stratification of patients. However, there is a need for other analysis to assess the plaque composition at higher resolution. Single-cell (sc) RNA sequencing (RNA-seq) studies have identified the cell composition of human carotid plaques, but the costs of this technology forces patient cohorts to be too small for clinical studies. Recently, deconvolution has been introduced to infer the cell type proportions of samples with bulk RNA-seq data using a cell type reference derived from the sc RNA-seq data. We aimed to perform deconvolution of human carotid plaques from a large cohort of patients that underwent endarterectomy, to get their cell type proportions and associate these with sex (with linear regression) and atherosclerosis severity, including symptoms (with logistic regression) and future events after surgery (with Cox proportional hazard model). We first benchmarked different deconvolution procedures regarding the data processing and the method selection. We detected that reducing the number of cell types from the sc reference improved the deconvolution performance, but all methods resulted in very different cell abundances. However, proportions from macrophages and smooth muscle cells (SMCs) were consistent among methods and, more importantly, with plaque histology. We finally observed female plaques to have a tendency to host a greater SMC percentage (β = 1.86 [-0.25, 3.96], p = 0.084), and macrophages being associated with symptomatic plaques (odds ratio = 1.02 [1.01, 1.03], p = 0.003) and major cardiovascular events after endarterectomy (hazard ratio = 1.02 [1.00, 1.03], p = 0.012). Our results suggest that obtaining deconvolved cell proportions of plaques is not straightforward for all the cell types, but macrophage and SMC proportions were consistent. While macrophages were did not differ significantly from male to female plaques, the latter showed an increased SMC content. Macrophages, in turn, might constitute a risk factor for atherosclerosis severity, and more research on the subtypes that are responsible for this should follow.