Author(s): MARTIN, J., TARDIVEL, C., CALLOVINI, A., RABERIN, A., COCON-PERRIN, R., BULLE, L., NOTTIN, S., DE TULLIO, P., LE GOFF, C., MILLET, G.P. , Institution: INRAE, Country: FRANCE, Abstract-ID: 2133

INTRODUCTION:
The Tor des Géants is one of the most extreme mountain ultra-marathons worldwide (330 km, >24,000 m elevation gain). Athletes endure prolonged physical strain, sleep deprivation, nutritional stress, and harsh environmental conditions over several days, making this event a unique in vivo model to investigate the limits of human physiological and metabolic adaptation. Our objective was to characterize metabolomic and lipidomic adaptations during such extreme endurance exercise and to identify pre-race molecular signatures predictive of performance.

METHODS:
Forty-nine athletes (11 women, 38 men) were enrolled. Blood samples were collected pre-race, at mid-race (147 km), and at the finish (330 km). Plasma and red blood cells were isolated and stored at −80°C. Untargeted metabolomics and lipidomics were performed using liquid chromatography coupled to high-resolution mass spectrometry. In plasma, 363 metabolites and 277 lipid species were identified; 362 lipid species were characterized in red blood cells.
Machine learning approaches (random forest, partial least squares regression) were applied to pre-race data to predict finishing status and race time. System-level adaptations were explored by mapping metabolites to biological pathways and functional subsystems and visualized as knowledge graphs.

RESULTS:
A pre-race signature of 11 plasma metabolites predicted finishing probability with 81% accuracy (P = 2.5 × 10⁻⁵; ROC AUC = 0.945, repeated 10-fold cross-validation). A second 13-metabolite signature predicted race time among finishers with an average error of 2.34 ± 0.34 %. Comparable predictive performance was obtained using selected plasma lipid species.
Longitudinal analyses (PLS-discriminant analysis) revealed a marked metabolic shift from pre-race to mid-race, followed by a partial return toward baseline at the finish. Early adaptations (start to mid-race) involved oxidative stress defense, mitochondrial energy production, and gut microbiota-related metabolism. Later changes (mid-race to finish) included sterol and steroid metabolism, nucleotide metabolism, and signatures suggestive of epigenetic regulatory potential. Lipid metabolism, muscle metabolic tuning, and vascular regulation were modulated throughout the race.

CONCLUSION:
Extreme ultra-endurance exercise induces dynamic, non-linear metabolic reprogramming with distinct temporal phases. Pre-race metabolomic and lipidomic signatures show strong potential for predicting performance and may serve as tools to assess athlete preparedness. Validation is ongoing in an independent ultra-trail cohort (SwissPeaks700 & 380). The possibility of exercise-induced epigenetic modulation warrants further careful investigation. Omics coupling to clinic-physiological data is underway.