Branched‐chain amino acids and type 2 diabetes: a bidirectional Mendelian randomization analysis

Abstract

ObjectiveGenetic studies have suggested that the branched‐chain amino acids (BCAAs) valine, leucine, and isoleucine have a causal association with type 2 diabetes (T2D). However, inferences are based on a limited number of genetic loci associated with BCAAs.MethodsInstrumental variables (IVs) for each BCAA were constructed and validated using large well‐powered data sets and their association with T2D was tested using a two‐sample inverse‐variance weighted Mendelian randomization approach. Sensitivity analyses were performed to ensure the accuracy of the findings. A reverse association was assessed using instrumental variables for T2D.ResultsEstimated effect sizes between BCAA IVs and T2D, excluding outliers, were as follows: valine (β = 0.14 change in log‐odds per SD change in valine, 95% CI: −0.06 to 0.33, p = 0.17), leucine (β = 0.15, 95% CI: −0.02 to 0.32, p = 0.09), and isoleucine (β = 0.13, 95% CI: −0.08 to 0.34, p = 0.24). In contrast, T2D IVs were positively associated with each BCAA, i.e., valine (β = 0.08 per SD change in levels per log‐odds change in T2D, 95% CI: 0.05 to 0.10, p = 1.8 × 10−9), leucine (β = 0.06, 95% CI: 0.04 to 0.09, p = 4.5 × 10−8), and isoleucine (β = 0.06, 95% CI: 0.04 to 0.08, p = 2.8 × 10−8).ConclusionsThese data suggest that the BCAAs are not mediators of T2D risk but are biomarkers of diabetes.

Insights into the metabolic consequences of type 2 diabetes

“…We find no evidence of a causal role of absolute triglycerides (TG) levels on T2D, although it has been described as a predictor of T2D risk 17, 42 . We replicate findings from Mosley et al 16 , which do not find evidence of a causal role of BCAAs on the risk of T2D. Almost all of the significant metabolites in the forward direction are also found to be associated in the reverse direction, highlighting the complex interplay between metabolite profiles and T2D liability.…”

“…The present work presents some limitations. We have performed bi-directional MR using a similar framework to previous studies in the UKBB based on the first release dataset of metabolite data 16, 18 , with the benefit of providing replication in the second release dataset. The use of the same outcome data for the two release datasets prevented the meta-analysis of the results.…”

“…The limited prediction of metabolite profiles may be related to the uncertain causal role in modulating T2D risk, a question investigated using Mendelian randomization (MR) 12 . One particular example is BCAAs, for which some studies have described a causal effect of valine, leucine and/or isoleucine on T2D 13–15 , while more recent evidence has shown no causal effect of BCAAs on T2D risk in the UK Biobank (UKBB) cohort 16 . Conversely, increasing evidence of a causal effect of T2D liability on metabolite profiles has been shown, with increased alanine levels caused by increased T2D liability being consistently reported 13, 17, 18 .…”

“…We first performed a bi-directional two sample MR analysis, using non-overlapping datasets for the SNP-exposure and SNP-outcome associations to limit potential bias 12 . Previous MR studies performed in the UKBB cohort have used only the first release of data for the 249 metabolites and were either limited to a few metabolites, or only investigated a single causal direction 16, 18, 26 . Secondly, we performed an interaction mQTL analysis to assess whether there is a different genetic regulation of metabolite levels between T2D patients and controls.…”