SUPERVISED AND UNSUPERVISED LEARNING TO DEFINE THE CARDIOVASCULAR RISK OF PATIENTS ACCORDING TO AN EXTRACELLULAR VESICLE MOLECULAR SIGNATURE

Abstract

Objective: Secreted extracellular vesicles (EVs) are membrane-bound nanoparticles released from cells. Since their content reflect the specific signatures of cellular activation and injury, EVs display a strong potential as biomarkers in the cardiovascular (CV) field. We aimed at dissecting a specific EV signature able to stratify patients according to their CV risk and likelihood to develop fatal CV events. Design and method: A total of 404 patients were included in the analysis. For each subject, we evaluated several CV risk indicators (age, sex, BMI, hypertension, hyperlipidemia, diabetes, coronary artery disease, chronic heart failure, chronic kidney disease, smoking habit, organ damage) and the likelihood of fatal CV events at 10 years, according to the SCORE charts of the European Society of Cardiology. Serum EVs were isolated by immuno-capture and analyzed for the expression of 37 EV surface antigens by flow cytometry. Unsupervised and supervised learning algorithms were applied for clustering patients according to CV risk. Results: Based on expression levels of EV antigens, unsupervised learning classified patients into three clusters (cluster I, 288 patients; cluster II, 86 patients; cluster III, 30 patients). Prevalence of hypertension, diabetes, chronic heart failure and organ damage (defined as left ventricular hypertrophy and/or microalbuminuria) progressively increases from cluster I to cluster III, with an average 6.9-fold increase. Several EV antigens, including markers from platelets (CD41b-CD42a-CD62P), leukocytes (CD1c-CD2-CD3-CD4-CD8-CD14-CD19-CD20-CD25-CD40-CD45-CD69-CD86), and endothelium (CD31-CD105) were independently associated to the CV risk indicators and correlated to age, blood pressure, glucometabolic profile, renal function, and SCORE risk. EV specific signature obtained by supervised learning allowed the accurate classification of patients according to their 10-year risk of future CV events, as estimated with the SCORE risk charts. Conclusions: EV profiling, obtainable from minimally-invasive blood sampling, may be integrated into CV risk stratification, displaying a potential role in the tailored management of these patients.