Metabolomic and genetic associations with insulin resistance in pregnancy

Abstract

Aims/hypothesis: Our study aimed to integrate maternal metabolic and genetic data related to insulin sensitivity during pregnancy to provide novel insights into mechanisms underlying pregnancy-induced insulin resistance.

Methods: Fasting and 1 h serum samples were collected from women in the Hyperglycemia and Adverse Pregnancy Outcome study who underwent an OGTT at ∼28 weeks' gestation. We obtained targeted and non-targeted metabolomics and genome-wide association data from 1600 and 4528 mothers, respectively, in four ancestry groups (Northern European, Afro-Caribbean, Mexican American and Thai); 1412 of the women had both metabolomics and genome-wide association data. Insulin sensitivity was calculated using a modified insulin sensitivity index that included fasting and 1 h glucose and C-peptide levels after a 75 g glucose load.

Results: Per-metabolite and network analyses across the four ancestries identified numerous metabolites associated with maternal insulin sensitivity before and 1 h after a glucose load, ranging from amino acids and carbohydrates to fatty acids and lipids. Genome-wide association analyses identified 12 genetic variants in the glucokinase regulatory protein gene locus that were significantly associated with maternal insulin sensitivity, including a common functional missense mutation, rs1260326 (β = -0.2004, p = 4.67 × 10^-12 in a meta-analysis across the four ancestries). This SNP was also significantly associated with multiple fasting and 1 h metabolites during pregnancy, including fasting and 1 h triacylglycerols and 2-hydroxybutyrate and 1 h lactate, 2-ketoleucine/ketoisoleucine and palmitoleic acid. Mediation analysis suggested that 1 h palmitoleic acid contributes, in part, to the association of rs1260326 with maternal insulin sensitivity, explaining 13.7% (95% CI 4.0%, 23.3%) of the total effect.

Conclusions/interpretation: The present study demonstrates commonalities between metabolites and genetic variants associated with insulin sensitivity in the gravid and non-gravid states and provides insights into mechanisms underlying pregnancy-induced insulin resistance. Graphical abstract.

Keywords: GCKR; Genome-wide association studies; Insulin sensitivity; Metabolomics; Pregnancy.

Fig. 1

Significant associations of fasting and 1 h metabolite levels and per-metabolite change following a glucose load with insulin sensitivity in meta-analysis. Significant associations of metabolites are shown based on p<0.05 after FDR adjustment in the fully adjusted model (model 4), which included field centre, sample storage time, mean arterial pressure, maternal age, neonatal sex, gestational age and maternal BMI at OGTT and parity. The red metabolites were positively associated with insulin sensitivity and the black metabolites were inversely associated. AA, amino acid; AC, acylcarnitine; Asn/Asx, asparagine/aspartic acid; Cho, carbohydrate; FA, fatty acid; G1P, glycerol 1-phosphate; GC/TCA, glycolysis/tricarboxylic acid cycle; Glu/Glx, glutamine/glutamic acid; Leu/Ile, leucine/isoleucine; NM/2AA/NE, N-methylamine/2-aminobutanoic acid/N-ethylglycine; OA, organic acid; OHpro, hydroxyproline; Pur/Pyr, purine or pyrimidine

Fig. 2

Sub-networks of fasting metabolites (a) and 1 h metabolites (b) associated with maternal insulin sensitivity. Nodes are coloured by metabolite class and sized by FDR-adjusted p values in the fully adjusted model (model 4), with large nodes referring to metabolites significantly associated with insulin sensitivity and small nodes referring to metabolites correlated with an individually significant metabolite. Blue shading represents spinglass communities within the estimated network. The solid lines between two nodes represent dependencies for intra-cluster metabolites, and the red dashed lines represent dependencies for metabolites across clusters. AA, amino acid; AC, acylcarnitine; Asn/Asx, asparagine/aspartic acid; Cho, carbohydrate; FA, fatty acid; G1P, glycerol 1-phosphate; Glu/Glx, glutamine/glutamic acid; Leu/Ile, leucine/isoleucine; NM/2AA/NE, N-methylamine/2-aminobutanoic acid/N-ethylglycine; OA, organic acid; OHpro, hydroxyproline; Pur/Pyr, purine or pyrimidine

Fig. 3

Manhattan plot for maternal insulin sensitivity for the meta-analysis across the four ancestry groups. The red line indicates genome-wide significance (p<5×10⁻⁸)

Fig. 4

Mediation analysis of the role of 1 h palmitoleic acid in mediating the relationship between rs1260326 and maternal insulin sensitivity. a, The association between rs1260326 and 1 h palmitoleic acid; b, the association between 1 h palmitoleic acid and maternal insulin sensitivity; c, the direct association between rs1260326 and maternal insulin sensitivity after adjustment for 1 h palmitoleic acid; ab+c, the total effect of rs1260326 on maternal insulin sensitivity. All associations were adjusted for field centre (for European ancestry mothers), the first two principal components, gestational age (weeks), maternal age, BMI, height, mean arterial pressure, parity, smoking (yes/no), alcohol intake status (yes/no) and fasting plasma glucose for fasting metabolites (1 h plasma glucose for 1 h metabolites) at the OGTT