A genome-wide cross-trait analysis identifies shared loci and causal relationships of type 2 diabetes and glycaemic traits with polycystic ovary syndrome

Abstract

Aims/hypothesis: The link underlying abnormal glucose metabolism, type 2 diabetes and polycystic ovary syndrome (PCOS) that is independent of BMI remains unclear in observational studies. We aimed to clarify this association using a genome-wide cross-trait approach.

Methods: Summary statistics from the hitherto largest genome-wide association studies conducted for type 2 diabetes, type 2 diabetes mellitus adjusted for BMI (T2DM_adjBMI), fasting glucose, fasting insulin, 2h glucose after an oral glucose challenge (all adjusted for BMI), HbA_1c and PCOS, all in populations of European ancestry, were used. We quantified overall and local genetic correlations, identified pleiotropic loci and expression-trait associations, and made causal inferences across traits.

Results: A positive overall genetic correlation between type 2 diabetes and PCOS was observed, largely influenced by BMI (r_g=0.31, p=1.63×10^-8) but also independent of BMI (T2DM_adjBMI-PCOS: r_g=0.12, p=0.03). Sixteen pleiotropic loci affecting type 2 diabetes, glycaemic traits and PCOS were identified, suggesting mechanisms of association that are independent of BMI. Two shared expression-trait associations were found for type 2 diabetes/T2DM_adjBMI and PCOS targeting tissues of the cardiovascular, exocrine/endocrine and digestive systems. A putative causal effect of fasting insulin adjusted for BMI and type 2 diabetes on PCOS was demonstrated.

Conclusions/interpretation: We found a genetic link underlying type 2 diabetes, glycaemic traits and PCOS, driven by both biological pleiotropy and causal mediation, some of which is independent of BMI. Our findings highlight the importance of controlling fasting insulin levels to mitigate the risk of PCOS, as well as screening for and long-term monitoring of type 2 diabetes in all women with PCOS, irrespective of BMI.

Keywords: Genome-wide cross-trait analysis; Insulin resistance; Mendelian randomisation; Polycystic ovary syndrome; Type 2 diabetes.

Fig. 1

Illustration of the genome-wide cross-trait analysis design. We first quantified overall and local genetic correlation, then identified specific pleiotropic loci and detected expression–trait associations and finally inferred causal relationships. Genome-wide genetic correlation analysis: https://github.com/bulik/ldsc; local genetic correlation analysis: https://huwenboshi.github.io/hess/; cross-trait meta-analysis: http://hal.case.edu/~xxz10/zhu-web/; Mendelian randomisation: https://mrcieu.github.io/TwoSampleMR/; transcriptome-wide association analysis: http://gusevlab.org/projects/fusion/

Fig. 2

Estimates of the causal effects of genetically predicted type 2 diabetes and glycaemic traits on PCOS. The boxes denote the point estimates of the causal effects and the error bars denote the 95% CIs. The IVW approach was used in the primary analysis and the MR-Egger and weighted median approaches were used in sensitivity analyses. The ORs for PCOS were scaled to the per unit increase in log OR of type 2 diabetes and per unit increase in glycaemic traits. T2DM, type 2 diabetes mellitus