SERUM PROTEOMIC PROFILING OF CAROTID ATHEROSCLEROSIS: A POPULATION OUTCOME STUDY

Abstract

Objective: Circulating proteins linked to atherosclerosis may improve cardiovascular (CV) disease prediction. Here, we applied proteomic profiling, feature selection and unbiased clustering to identify proteins associated with carotid atherosclerosis and integrate them in a protein-based classification system for prediction of future CV events. Design and method: 491 community-dwelling participants (mean age, 58±11 years; 51% women) underwent carotid ultrasonography and proteomic profiling (CVD II panel, Olink Proteomics) at baseline. Subsequently, we collected cardiovascular outcome for a median follow-up period of 10.2 years. We applied partial least squares (PLS) and a machine learning algorithm (Random Forest, RF) to identify proteins associated with carotid intima-media thickness (cIMT) and the presence of atherosclerotic plaques. Next, we assessed the association of future CV events with protein-based phenogroups defined by unbiased clustering (Gaussian Mixture modelling) based on influential proteins in PLS and RF. Results: Of 92 measured proteins, 13 were labelled as important by PLS and RF and were independently associated with cIMT and/or the presence of plaques after multivariable adjustment: pro-adrenomedullin, agouti-related protein, brother of CDO, CD40 ligand, CD84, GLO1, HB-EGF, HSP27, lipoprotein lipase, lymphotactin, PAR-1, PDGF subunit B and TGM2. Based on these proteins, the clustering algorithm subdivided the cohort into two distinct phenogroups. Compared to the first (n = 174), participants in the second phenogroup (n = 317) had: i) a more unfavourable lipid profile with higher total cholesterol and triglycerides and lower HDL cholesterol (P<0.014 for all), but no differences in other CV risk factors such as age, blood pressure, smoking status and diabetes mellitus (P>0.061 for all); ii) higher cIMT (P = 0.0020); and iii) a significantly higher risk for future CV events during follow-up (multivariable-adjusted hazard ratio (95% CI) versus cluster 1: 2.63 (1.25 to 5.56); P = 0.01). Conclusions: Carotid atherosclerosis was linked to proteins reflecting haemostasis, angiogenesis, blood pressure regulation and inflammation. Focused proteomic phenomapping adequately identified individuals at high risk for future CV events and may thus complement current CV risk stratification strategies.