Integration of Proteomic and Metabolomic Plasma Profiles in a Community-Based Cohort

Introduction: Circulating proteins and metabolites play key roles as markers and effectors of cardiometabolic diseases. We hypothesized that integrating these data sets may identify novel circulating protein-metabolite associations and provide new insight into molecular pathways that underlie cardiometabolic and vascular disease. Methods: Aptamer-based proteomic profiling of 1,129 analytes in plasma samples from 923 Framingham Heart Study Offspring participants (mean age 56 +/- 12 years, 52% women) was performed using the Somascan platform. Targeted metabolomic profiling of 217 analytes was performed by liquid chromatography tandem mass spectrometry, as previously reported. Pearson correlations between log-transformed protein and metabolite concentrations were assessed using PROC CORR in SAS, adjusting for age, sex, and the homeostatic model assessment. Results: We identified over 5,000 unadjusted protein-metabolite associations with high levels of significance (Bonferroni-corrected threshold P-value -7 ). Our analysis identified protein-metabolite relationships known to define key biological pathways. For example, Apolipoprotein E (ApoE) was most highly associated with triacylglyerols (TAGs) of varying fatty acid chain length and degrees of saturation, the most significant of which was a TAG with 50 carbons and 4 double bonds (P=5x10 -48 ). Similarly, thyroid stimulating hormone was most strongly negatively associated with thyroxine (P-value 4x10 -18 ). By contrast, we identified hundreds of novel correlations between plasma levels of metabolites and key cardiometabolic proteins such as Apo E (for example, with the branched chain amino acids valine, P=2x10 -11 ; leucine, P=7x10 -10 ; and isoleucine, P=1x10 -9 ), PCSK9 (with 3-aminoisobutyric acid, P=7x10 -11 ), and adiponectin (with 2-aminoadipic acid, P=1x10 -26 ). These associations remained highly significant when adjusted for age, sex, and insulin resistance. Conclusions: The integration of protein and metabolite profiling data sets from large community-based cohorts can be used to identify novel protein-metabolite associations. These associations may offer new tools to elucidate the molecular pathways that underlie cardiometabolic and vascular disease.