Evidence of bidirectional relationship between type 2 diabetes and depression; a Mendelian randomization study

Abstract

Major depressive disorder (MDD) and type 2 diabetes (T2D) represent two global health challenges with a high degree of co-occurrence. Here, we aim to investigate the causal relationship between MDD and T2D in diverse ancestries using Mendelian randomization (MR) in GWAS summary statistic and individual level (UK Biobank (UKB)) data. We assessed the bi-directional causal relationship between: (a) MDD and T2D and (b) MDD and glycaemic biomarkers (e.g. TG:HDL-C ratio, a measure of insulin resistance, fasting glucose) in non-diabetic individuals. In UKB we also tested the role of T2D on treatment resistant depression (TRD). We used multivariable MR (MVMR) to assess the role of body mass index (BMI) in the MDD to T2D relationship. Our results demonstrated that a doubling in MDD genetic liability was associated with 1.14 higher odds of T2D (95% CI:1.09, 1.19), whilst a doubling in T2D genetic liability associated with 1.02 higher odds of MDD (95% CI:1.01, 1.03). Consistent effect estimates were observed in the UKB when stratifying by sex and suggested a role for T2D in TRD. T2D GWAS derived clusters of genetic variants highlighted the importance of specific pathways in the MDD relationship, including variants raising T2D risk via body fat (OR:1.04; 95% CI:1.02, 1.06), obesity mediated insulin resistance (OR:1.06; 95% CI:1.04, 1.09) and residual glycaemic (OR: 1.02; 95% CI:1.00, 1.04) pathways. MVMR with BMI attenuated the bidirectional relationship between MDD and T2D, particularly from MDD to T2D. Genetic liability to MDD was also associated with higher TG:HDL-C ratio in individuals without T2D (β:0.11; 95% CI:0.08, 0.14). We provide evidence of bidirectional causal association between MDD and T2D, with MDD strongly predicting insulin resistance and T2D. T2D predicted both MDD and TRD and highlighted the importance of obesity and body fat pathways in the T2D to MDD relationship.

Fig. 1. Summary of 1-sample and 2-sample UVMR results.

a T2D exposure > MDD outcome: 2-sample UVMR (IVW estimates) showing the odds of MDD per doubling in the genetic liability of T2D. The figure includes all T2D SNPs from the GWAS (Full T2D SNPs) as well as the 7 subsets of SNPs hypothesised to act via specific pathways (Note: subset of liver lipid metabolism has not been shown in the figure due to large confidence intervals), b T2D exposure > Treatment resistant depression (TRD) outcome: 1-sample UVMR in UK-biobank showing the odds of TRD (GP records) per doubling in the genetic risk of T2D using genetic risk score (GRS) for all T2D SNPs and its 8 subsets. The odds of TRD were significantly higher using full T2D GRS and in body fat and obesity subset, while lower odds for MDD in Beta cell PI negative subset. *p value <0.05 (Abbreviations used: GWS genome wide significance, SNP single nucleotide polymorphism, MDD major depressive disorder, IVW inverse variance weighed, PI proinsulin, neg negative, pos positive, UVMR univariable Mendelian randomization, TRD treatment resistant depression, GRS genetic risk score, GP general practitioner) (Note: All SNPs: includes all T2D variants at GWS in the primary T2D GWAS as an exposure; excluding GWS SNPs: removes any variants that are associated at GWS with the outcome (i.e. MDD); excluding SNPs related to T2D & MDD associated traits: searching the GWAS catalogue for GWS associations involving the exposure variant or nearby variants in LD, to identify and removing the variants that are associated with related traits (e.g. BMI, body fat, anxiety, neuroticism etc.) to minimize any potential pleiotropic effect).

Fig. 2. Findings from 1-sample UVMR & MVMR and 2-sample UVMR analysis.

a MDD exposure > T2D outcome: 2-sample UVMR (IVW estimates) showing the odds ratios for T2D per doubling in MDD genetic liability. The different models presented represent including all MDD variants, excluding those at GWS with T2D and excluding all SNPs with known associations with MDD/T2D related traits, b 1-sample UVMR and MVMR (adjusted for BMI) in UK-biobank in all individuals. *p value <0.05 (Abbreviations used: GWS genome wide significance, SNP single nucleotide polymorphism, MDD Major depressive disorder, IVW Inverse variance weighed, UVMR univariable Mendelian randomization, MVMR multivariable Mendelian randomization, BMI body mass index, MHQ mental health questionnaire, GP general practitioner, GRS genetic risk score) (Note: All SNPs: includes all MDD variants at GWS in the primary MDD GWAS as an exposure; excluding GWS SNPs: removes any variants that are associated at GWS with the outcome (i.e. T2D); excluding SNPs related to MDD & T2D associated traits: searching the GWAS catalogue for GWS associations involving the exposure variant or nearby variants in LD, to identify and removing the variants that are associated with related traits (e.g. BMI, body fat, anxiety, neuroticism etc.) to minimize any potential pleiotropic effect).

Fig. 3. Summary of results from 1-sample and 2-sample UVMR analysis.

a MDD exposure > T2D biomarkers outcome: 2-sample UVMR results (IVW) representing the change in the T2D biomarker per doubling in MDD genetic liability, b major depression exposure > TG:HDL-C ratio outcome: 1-sample UVMR in UK-biobank shows increase in TG:HDL-C ratio in all and non-T2D individuals in sex stratified analysis per doubling in genetic liability to major depression (GP derived) GRS. *p value <0.05 (Abbreviations used: MDD major depressive disorder, IVW inverse variance weighed, ISI insulin sensitivity index, BMI body mass index, adj adjusted, HbA1c glycated haemoglobin, TG-HDL-C triglycerides to high density lipoprotein cholesterol ratio, UVMR univariable Mendelian randomization, T2D type 2 diabetes, GP general practitioner, GRS genetic risk score, TG-HDL-C Triglycerides to high density lipoprotein cholesterol ratio).

Fig. 4. 2-sample MR results (IVW) representing the odds of MDD per unit change in genetically instrumented biomarkers.

T2D biomarkers did not predict the risk of MDD (MDD major depressive disorder, IVW inverse variance weighed, ISI insulin sensitivity index, BMI body mass index, adj adjusted, HbA1c glycated haemoglobin, TG-HDL-C triglycerides to high density lipoprotein cholesterol ratio).