Using molecular classification to predict gains in maximal aerobic capacity following endurance exercise training in humans

Timmons JA, Knudsen S, Rankinen T, Koch LG, Sarzynski M, Jensen T, Keller P, Scheele C, Vollaard NB, Nielsen S, Akerstrom T, MacDougald OA, Jansson E, Greenhaff PL, Tarnopolsky MA, van Loon LJ, Pedersen BK, Sundberg CJ, Wahlestedt C, Britton SL, Bouchard C. Using molecular classification to predict gains in maximal aerobic capacity following endurance exercise training in humans. J Appl Physiol 108: 1487-1496, 2010. First published February 4, 2010; doi:10.1152/japplphysiol.01295.2009.-A low maximal oxygen consumption ((V) over dotO(2max)) is a strong risk factor for premature mortality. Supervised endurance exercise training increases (V) over dotO(2max) with a very wide range of effectiveness in humans. Discovering the DNA variants that contribute to this heterogeneity typically requires substantial sample sizes. In the present study, we first use RNA expression profiling to produce a molecular classifier that predicts (V) over dotO(2max) training response. We then hypothesized that the classifier genes would harbor DNA variants that contributed to the heterogeneous (V) over dotO(2max) response. Two independent preintervention RNA expression data sets were generated (n = 41 gene chips) from subjects that underwent supervised endurance training: one identified and the second blindly validated an RNA expression signature that predicted change in (V) over dotO(2max) (predictor genes). The HERITAGE Family Study (n = 473) was used for genotyping. We discovered a 29-RNA signature that predicted (V) over dotO(2max) training response on a continuous scale; these genes contained similar to 6 new single-nucleotide polymorphisms associated with gains in (V) over dotO(2max) in the HERITAGE Family Study. Three of four novel candidate genes from the HERITAGE Family Study were confirmed as RNA predictor genes (i.e., reciprocal RNA validation of a quantitative trait locus genotype), enhancing the performance of the 29-RNA-based predictor. Notably, RNA abundance for the predictor genes was unchanged by exercise training, supporting the idea that expression was preset by genetic variation. Regression analysis yielded a model where 11 single-nucleotide polymorphisms explained 23% of the variance in gains in (V) over dotO(2max), corresponding to similar to 50% of the estimated genetic variance for (V) over dotO(2max). In conclusion, combining RNA profiling with single-gene DNA marker association analysis yields a strongly validated molecular predictor with meaningful explanatory power. (V) over dotO(2max) responses to endurance training can be predicted by measuring a similar to 30-gene RNA expression signature in muscle prior to training. The general approach taken could accelerate the discovery of genetic biomarkers, sufficiently discrete for diagnostic purposes, for a range of physiological and pharmacological phenotypes in humans.