Abstract MP12: Proteomic Predictors Of High-density Lipoprotein Cholesterol Response To Regular Exercise

Introduction: Regular exercise beneficially increases plasma HDL-C levels at the group level. However, variation in individual HDL-C responses to exercise highlight a need for predictive biomarkers of exercise response. Hypothesis: We hypothesized that baseline abundance of circulating proteins is predictive of HDL-C response to exercise and that identified proteins are part of a complex biological network of exercise response. Methods: We measured over 5,000 circulating proteins using an aptamer-affinity based platform (SomaScan) in 667 black and white adults from the HERITAGE Family Study. Fasting plasma HDL-C was measured at baseline and following 20 weeks of supervised endurance exercise training. To predict exercise induced changes in HDL-C using baseline abundance of circulating proteins, models were created using LASSO regression and a 70/30 training test data split with 10-fold cross validation. Biological pathways, networks, and functions involving proteins identified in predictive modeling were investigated by ingenuity pathway analysis (IPA) and integrated molecular pathway level analysis (IMPaLA). Results: Regular exercise significantly increased HDL-C in the sample by 1.5 ± 4.6 mg/dL (p<0.0001), however marked inter-individual differences in response were present (range: -19.5 to +17.4 mg/dL). LASSO regression of circulating proteins only yielded a model of 120 proteins with similar but stronger predictive power to a model of 19 clinical traits (root mean square error = 4.52 and 5.3 mg/dL respectively). LASSO regression of both clinical and proteomic predictors resulted in a final model of baseline HDL-C and 116 circulating proteins, with an improved root mean square error of 4.11 mg/dL. Furthermore, this panel of 116 proteins was able to explain 40.0% of the variance in exercise induced changes in plasma HDL-C, while clinical predictors alone (including baseline HDL-C) explained only 3.9%. Pathway analysis of these 116 proteins identified several biological processes including pathways involved in the progression towards atherosclerosis, angiogenesis, mTOR signaling, and mitochondrial fatty acid synthesis. Conclusions: Circulating proteins may allow for prediction of exercise induced changes in HDL-C. Additionally, proteins predictive of HDL-C response to exercise are associated with important biological pathways and may provide insights into the molecular mechanisms of the benefits of regular exercise.