Abstract MP33: Inflammatory proteomic signatures of metabolically unhealthy obesity

Background: Obesity is a risk factor for CVD incidence and mortality. Accumulating evidence implicates viscerally located adiposity may be particularly relevant for the progression of cardiovascular risk factors, with chronic inflammation as a plausible causal pathway. However, prior research has lacked precise phenotyping to characterize the relationship between visceral fat and inflammatory markers. Objective: Leveraging DXA body composition imaging and plasma proteomics measured by the Olink 384 inflammation panel, we aimed to (i) characterize phenotypes of metabolically glycemic unhealthy obesity and (ii) develop related novel inflammatory proteomic signatures ( i.e . “inflammotypes”). Methods: We analyzed 639 participants (mean age 63.7) from the VITamin D and OmegA-3 TriaL (VITAL) with DXA and proteomics at baseline. We defined four metabolic obesity phenotypes for each participant, as the sum of standardized values of visceral adipose tissue (VAT) mass and one of four glycemic metabolic traits, including fasting glucose, HbA1c, HOMA-IR, and LPIR (a lipid-based score predictive of incident diabetes). We also derived categorical phenotypes by cross-classifying VAT (high/low) with each glycemic trait (unhealthy/healthy), using cut-points that yielded the highest AUC. We then used elastic net regression models to identify the inflammotypes related to each of the four metabolic obesity phenotypes. Finally, we performed an external validation in the COcoa Supplement and Multivitamin Outcomes Study (COSMOS; N=371, mean age 69.0). Results: The continuous metabolic obesity measures were related to unique inflammotypes consisting of 86 to 102 proteins, with 21 proteins (including HGF, IL1RN, ISM1, PON3, and HSD11B1) shared across all phenotypes. For the categorical phenotypes, we identified 19, 22, and 3 proteins differentially expressed among metabolically unhealthy obesity subgroups, as defined for HbA1c, HOMA-IR, and LPIR metabolically unhealthy, respectively, but none were specific to fasting glucose. The inflammotypes were highly correlated with the corresponding phenotypes in COSMOS (Pearson correlation coefficients ranged from 0.63 to 0.82). Conclusion: We derived and externally validated novel proteomic inflammotypes of metabolically unhealthy obesity. These inflammotypes will be evaluated for the associations with subsequent risk of cardiometabolic outcomes in prospective longitudinal cohorts, to elucidate visceral adiposity-related pathways of CVD risk.