A two-sample Mendelian randomization study explores metabolic profiling of different glycemic traits

Abstract

We assessed the causal relation of four glycemic traits and type 2 diabetes liability with 167 metabolites using Mendelian randomization with various sensitivity analyses and a reverse Mendelian randomization analysis. We extracted instruments for fasting glucose, 2-h glucose, fasting insulin, and glycated hemoglobin from the Meta-Analyses of Glucose and Insulin-related traits Consortium (n = 200,622), and those for type 2 diabetes liability from a meta-analysis of multiple cohorts (148,726 cases, 965,732 controls) in Europeans. Outcome data were from summary statistics of 167 metabolites from the UK Biobank (n = 115,078). Fasting glucose and 2-h glucose were not associated with any metabolite. Higher glycated hemoglobin was associated with higher free cholesterol in small low-density lipoprotein. Type 2 diabetes liability and fasting insulin were inversely associated with apolipoprotein A1, total cholines, lipoprotein subfractions in high-density-lipoprotein and intermediate-density lipoproteins, and positively associated with aromatic amino acids. These findings indicate hyperglycemia-independent patterns and highlight the role of insulin in type 2 diabetes development. Further studies should evaluate these glycemic traits in type 2 diabetes diagnosis and clinical management.

Fig. 1. Heatmap of associations of genetically predicted glycemic traits (fasting glucose, 2-h glucose, HbA1c and fasting insulin) and of genetic liability to type 2 diabetes with amino acids, fatty acids, and various low-molecular weight metabolites.

Circulating metabolites include: amino acids, aromatic amino acids, branched-chain amino acids, fatty acids, biomarkers of fluid balance, glycolysis related metabolites, inflammation, and ketone bodies. The estimates were obtained from Mendelian randomization analyses using the inverse variance weighted method. Asterisks depict statistical significance (p < 0.0004). BCAA branched-chain amino acid.

Fig. 2. Heatmap of associations of genetically predicted glycemic traits (fasting glucose, 2-h glucose, HbA_1c and fasting insulin) and of genetic liability to type 2 diabetes with cholesterol metabolites.

Measures of cholesterol metabolites include: apolipoproteins, cholesterol, cholesteryl esters, free cholesterol, lipoprotein particle concentrations, lipoprotein particle sizes, other lipids, phospholipids, total lipids, and triglycerides. The estimates were obtained from Mendelian randomization analyses using the inverse variance weighted method. Asterisks depict statistical significance (p < 0.0004). LDL low-density lipoprotein, HDL high-density lipoprotein, VLDL very low-density lipoprotein.

Fig. 3. Heatmap of associations of genetically predicted glycemic traits (fasting glucose, 2-h glucose, HbA_1c and fasting insulin) and of genetic liability to type 2 diabetes with lipoprotein subclasses in very small, small, medium, large, very large VLDL, and chylomicrons and extremely large VLDL.

The estimates were obtained from Mendelian randomization analyses using the inverse variance weighted method. Asterisks depict statistical significance (p < 0.0004). VLDL very low-density lipoprotein.

Fig. 4. Heatmap of associations of genetically predicted glycemic traits (fasting glucose, 2-h glucose, HbA_1c and fasting insulin) and of genetic liability to type 2 diabetes with lipoprotein subfractions in small, medium, and large LDL, in small, medium, large, and very large HDL, and those in IDL.

The estimates were obtained from Mendelian randomization analyses using the inverse variance weighted method. Asterisks depict statistical significance (p < 0.0004). LDL low-density lipoprotein, HDL high-density lipoprotein, IDL intermediate-density lipoprotein, VLDL very low-density lipoprotein.