Dietary Patterns, Circulating Metabolome, and Risk of Type 2 Diabetes

Circulating metabolites may reflect biological homeostasis and have been linked to dietary intakes and human health, and may hold the promises to facilitate objective assessments of intakes and metabolic response to diets. Here, we integrated metabolomic, genetic, and metagenomic data from five longitudinal cohorts comprising 21,474 participants of diverse ethnic backgrounds, to develop metabolomic signatures for popular dietary patterns (i.e., three guideline-based diets, three plant-based diets, and two mechanism-based diets) and systematically investigated their clinical relevance. Applying machine-learning models in two deeply-phenotyped lifestyle validation studies, we identified eight metabolomic signatures (each included 37 to 66 metabolites) significantly correlated with their respective dietary pattern indices, consistently across multiple independent validation cohorts (r = 0.11–0.38; P < 8.06×10⁻⁹). These signatures included shared metabolites between diets (e.g., up to 67% among guideline-based diets, including hippuric and 3-indolepropionic acid), and metabolites unique to specific diets (e.g., N6,N6,N6-trimethyllysine to proinflammatory diet). In multivariable-adjusted analyses of 5 prospective cohorts (1,832 incident cases during up to 27 years of follow-up), the metabolomic signatures of healthful diets (i.e., Mediterranean and healthful plant-based diets) were associated with lower T2D risk (HR: 0.82–0.90; P < 3×10⁻⁶), while signatures for unhealthy diets (e.g., proinflammatory and hyperinsulinemia diets) were associated with higher T2D risk (HR: 1.23–1.26; P < 2×10⁻¹⁵); these associations were further supported by Mendelian randomization analysis incorporating genetic data. Finally, through genome-wide and taxa-wide associating analyses, we identified 15 genetic loci – including those involved in fatty acid and energy metabolism (e.g., FADS1/2 and CERS4 ; P < 5×10 ^-8 ), and 39 gut microbial species – including those relevant to butyric acid metabolism (e.g., E. eligens and F. pranusnitzii ; FDR < 0.05), significantly associated with the metabolomic signatures of diets. Genetic variants and gut microbial diversity explained up to 19.1% and 10.6% of the variation in these signatures, respectively, underscoring a potential role of host genetics and gut microbiota in dietary metabolism. In conclusion, our study identified metabolomic signatures reflecting both intakes and individual metabolic response to various diets and are associated with future T2D risk. These signatures may facilitate individualized dietary assessments and risk stratification in future nutritional research.