Genome-wide association study and trans-ethnic meta-analysis identify novel susceptibility loci for type 2 diabetes mellitus

Abstract

Background: The genetic basis of type 2 diabetes (T2D) is under-investigated in the Middle East, despite the rapidly growing disease prevalence. We aimed to define the genetic determinants of T2D in Qatar.

Methods: Using whole genome sequencing of 11,436 participants (2765 T2D cases and 8671 controls) from the population-based Qatar Biobank (QBB), we conducted a genome-wide association study (GWAS) of T2D with and without body mass index (BMI) adjustment.

Results: We replicated 93 known T2D-associated loci in a BMI-unadjusted model, while 96 known loci were replicated in a BMI-adjusted model. The effect sizes and allele frequencies of replicated SNPs in the Qatari population generally concurred with those from European populations. We identified a locus specific to our cohort located between the APOBEC3H and CBX7 genes in the BMI-unadjusted model. Also, we performed a transethnic meta-analysis of our cohort with a previous GWAS on T2D in multi-ancestry individuals (180,834 T2D cases and 1,159,055 controls). One locus in DYNC2H1 gene reached genome-wide significance in the meta-analysis. Assessing polygenic risk scores derived from European- and multi-ancestries in the Qatari population showed higher predictive performance of the multi-ancestry panel compared to the European panel.

Conclusion: Our study provides new insights into the genetic architecture of T2D in a Middle Eastern population and identifies genes that may be explored further for their involvement in T2D pathogenesis.

Keywords: DYNC2H1; TCF7L2; GWAS; Polygenic risk score; SNP; T2D; Type 2 diabetes Mellitus.

Fig. 1

Manhattan Plots and Q-Q Plots for T2D BMI-unadjusted and BMI-adjusted Models. Manhattan plot (A, C) and Q-Q plot (B, D) for BMI-adjusted (A, B) and for BMI-unadjusted (C, D) GWAS. Manhattan plots represent SNPs (dots) plotted on x-axis in accordance with chromosome position against their corresponding -log10(P). The red horizontal line represents the threshold for genome-wide significance (P < 5 × 10^− 08). Q-Q plots represent the quantile distribution of observed P values versus the quantile distribution of expected P values for all SNPs

Fig. 2

Comparison of Allele Frequencies and Effect sizes (BETA) of Replicated SNPs Identified in the GWAS Catalog and QBB Cohort. Correlation of allele frequencies for replicated loci between QBB and GWAS catalog; A-B. BMI-unadjusted and C-D. BMI-adjusted models. Blue dots represent SNPs with similar direction of effect size, while red dots represent SNPs with opposite direction of effect size

Fig. 3

Regional Association Plots of Novel Loci showing Genome-wide Significant Association with T2D. The plot shows the chromosomal position of (A) rs143508949 and (B) rs2510095 plotted against -log₁₀P values (based on NCBI human genome build 38). The novel SNP (diamond) and other detected SNPs (circle) are color coded based on LD (square correlation (r2)). (C) The lead SNP rs2510095 is a significant eQTL for DYNC2H1 in Thyroid tissues. Box and whiskers plot shows DYNC2H1expression by rs2510095 genotypes

Fig. 4

The predictive performances and translatability of T2D polygenic risk scoring panels derived from European- and multi-ancestries. The predictive performances of PGS derived from the European ancestry (PGS000036) and multi-ancestry (PGS000804) when applied to participants from QBB cohort and other ancestries was assessed by the area under the receiver operating curves (AUC) values. (A) Bar plots show the predictive performances of the European-ancestry PGS when applied to QBB participants and Europeans [35], and (B) the predictive performances of multi-ancestry PGS when applied to participants from QBB, European, African and Hispanic or Latin American ancestries [32] in the basic model (without covariate adjustment). (C) Bar plots show the predictive performances of the European-ancestry PGS when applied to QBB and European [34] ancestries and (D) the predictive performances of multi-ancestry PGS when applied to participants from QBB, European, African and Hispanic or Latin American ancestries in the full model (with covariate adjustment). Error bars represent 95% confidence interval for AUC. *Predictive performance of the two PGS showed statistical significance (P < 0.05) in all tested conditions