(DCP078) METABOLOME-WIDE MENDELIAN RANDOMIZATION FOR YOUTH-ONSET TYPE 2 DIABETES

Background: Several studies focus on circulating metabolites to identify early biomarkers for youth type 2 diabetes (T2D), but are hampered by increased costs and unmeasured confounding. Mendelian randomization (MR) offers an instrumental variable-based approach, to study causal associations between biomarkers and diseases, using single nucleotide polymorphisms (SNPs) identified in genome-wide studies (GWAS).

METHODS AND RESULTS: We investigated causal effects on youth T2D of metabolites with available SNP-instruments in four GWAS (Kettunen et al., Lotta et al., Long et al., Shin et al.), using two-sample MR. Effects of these SNPs on pediatric T2D were derived from the multi-ethnic ProDiGY GWAS (3,006 cases/6,061controls) or its European subset (664 cases/1434 controls). MR effects were estimated using Wald ratios for metabolites with a single SNP-instrument or the Inverse Variance Weighted (IVW) method in multi-instrument MRs. Results express odds ratios (OR) and 95% confidence intervals (CI) for youth T2D per standard deviation increase in the genetically-predicted metabolite level. Steiger testing evaluated the directionality of the MR associations. Multivariable MR (MVMR) tested effects of the metabolites on T2D conditioning on childhood BMI.Among 586 unique metabolites tested in the multiethnic MR, 26 metabolites demonstrated a causal effect on T2D (FDR P-values < 0.1), most of which clustered in the glycerophospholipid class. Among these, PC aa C32:2 and PC ae C36:2 had a protective effect (OR = 0.83, 95% CI: 0.77-0.89; OR = 0.89, 95% CI: 0.82-0.96). Steiger testing supported causality in the expected direction for 22 metabolites, and indicated reverse causality for 4 metabolites (HDL diameter, caproate (6:0), alanine, serum.C). MVMR analyses for 5 metabolites showed that, conditioning on childhood BMI, the MR associations of HDL.D, Serum.C and PC aa C34:3 with T2D disappeared, while those of PC aa C32:2 and PC ae C36:2 remained.
In the European MR analysis, 8 of the 586 metabolites showed a causal effect on T2D (FDR P-value < 0.1), 6 of which are glycerophospholipids. 1-stearoyl-2-linoleoyl-GPE(18:0/18:2) had a protective effect on T2D (OR=0.90; 95% CI: 0.85-0.95), whereas the other 7 metabolites increased T2D risk, with 1-Docosahexaenoyl-GPE(22:6) associating with the highest risk (OR = 1.22; 95% CI: 1.08-1.37). Steiger testing supported the expected direction of the causal association for all 8 metabolites. MVMR was not performed due to insufficient number of MR instruments.

Conclusion: Sensitivity analyses with pleiotropy-robust MR methods will be performed, as well as colocalization analyses to test whether the candidate metabolites share causal SNPs with youth T2D. Finally, using MetaboAnalyst, we will seek to characterize predominant metabolic pathways in pediatric T2D.