Metabolic signatures of insulin resistance in 7,098 young adults

Abstract

Metabolite associations with insulin resistance were studied in 7,098 young Finns (age 31 ± 3 years; 52% women) to elucidate underlying metabolic pathways. Insulin resistance was assessed by the homeostasis model (HOMA-IR) and circulating metabolites quantified by high-throughput nuclear magnetic resonance spectroscopy in two population-based cohorts. Associations were analyzed using regression models adjusted for age, waist, and standard lipids. Branched-chain and aromatic amino acids, gluconeogenesis intermediates, ketone bodies, and fatty acid composition and saturation were associated with HOMA-IR (P < 0.0005 for 20 metabolite measures). Leu, Ile, Val, and Tyr displayed sex- and obesity-dependent interactions, with associations being significant for women only if they were abdominally obese. Origins of fasting metabolite levels were studied with dietary and physical activity data. Here, protein energy intake was associated with Val, Phe, Tyr, and Gln but not insulin resistance index. We further tested if 12 genetic variants regulating the metabolites also contributed to insulin resistance. The genetic determinants of metabolite levels were not associated with HOMA-IR, with the exception of a variant in GCKR associated with 12 metabolites, including amino acids (P < 0.0005). Nonetheless, metabolic signatures extending beyond obesity and lipid abnormalities reflected the degree of insulin resistance evidenced in young, normoglycemic adults with sex-specific fingerprints.

FIG. 1.

Associations of amino acids with HOMA-IR across tertiles of waist circumference for men and women. Linear regression and ANCOVA models were adjusted for age and waist circumference, total cholesterol, HDL cholesterol, and triglycerides. Association magnitudes are in units of 1-SD increased HOMA-IR per 1-SD increase in amino acid level, and error bars indicate SE. Associations were meta-analyzed for the two cohorts (n = 7,098). *P < 0.05; ♦P < 0.0005 for association of amino acid with HOMA-IR. P < 0.05; ⊗P < 0.0005 for amino acid × waist interaction, indicating different slopes across tertiles of waist circumference.

FIG. 2.

Associations of dietary composition and physical activity with circulating metabolites. All associations were adjusted for age and sex. Physical activity associations are shown with additional adjustment for HOMA-IR as well. Association magnitudes are in units of 1-SD change in metabolite concentrations per 1-SD change in lifestyle measure. Error bars indicate 95% CIs and numbers indicate P values of association. Protein, fat, and carbohydrate energy intake is per total energy intake. Dietary energy intake was derived from 48-h dietary interviews (n = 911), and physical activity was quantified as MET index based on questionnaires (n = 6,223). Av., average.

FIG. 3.

Associations of genetic variants regulating metabolite levels with HOMA-IR (A) and the strongest circulating metabolite measure (B). Associations for rs1260326 in GCKR with the metabolites before (C) and after (D) adjustment for triglycerides. Error bars indicate 95% CIs and numbers indicate P values of association. All associations were adjusted for sex, age, waist, and population structure and meta-analyzed for the two cohorts (n = 6,343). Association magnitudes are in units of 1 SD HOMA-IR or metabolite concentration per allele copy. Av., average.