Risk of developing type 2 diabetes (T2D) is heritable, with over 403 independent genetic variants having been identified primarily through analysis of common variants in European ancestry populations. To further illuminate the genetic architecture of T2D, we conducted a whole genome sequence (WGS) association study (>38x depth) of common, low frequency, and rare variants in 9,663 individuals with T2D and 35,050 controls of diverse ancestry from NHLBI’s Trans-Omics for Precision Medicine (TOPMed) study. We conducted single variant analyses and SKAT tests of rare variants (MAF<1%) annotated with pancreatic islet specific regulatory annotation to define variant sets. In single variant analyses, a total of 15 regions were significantly associated (a=5e-08) with T2D, seven of which are novel associations (in or near ODF2L, LMAN2, KCNV1, CBR1, VLDLR-AS1, LINC01052 and NKX2-5). These new associations were either ancestry specific or rare variants. For instance, a variant not present in other ancestral groups, near KCNV1, was associated with increased risk of T2D (MAF 5.5%, P= 1.57e-08, OR=1.43) in African ancestry. Set based testing of rare variants identified seven regions associated with T2D in at least one ancestry. Six of these were noncoding: in African ancestry CBR1 gene centric variants (cumulative MAC[cMAC]=623, P = 5.85e-09) and an active promoter region in proximity to MRPL46 and MRPS11 (cMAC=3,218, P = 4.89e-07); and in Asian ancestry an enhancer hub containing NXPH3 (cMAC=3,317, P = 3.42e-05) and an active enhancer region linked to RP11-135A1.2 and HES1 (cMAC=57, P = 4.87e-07) were associated with T2D. KLRC4 predicted deleterious coding variants were associated with T2D in the European ancestry group (cMAC=25, P = 3.63e-06). In large scale WGS analyses, we extended our knowledge of the genetic basis of T2D through the inclusion of ancestrally diverse samples, variant typing without imputation, and use of tissue specific regulatory annotation to identify new T2D loci.

Background: Using whole genome sequence association analysis of HbA1c (a test used to diagnose type 2 diabetes (T2D) and estimate glycemic control) in 3,123 African-Americans (AA) from the Trans-Omics for Precision Medicine (TOPMed) program, we confirmed the known AA HbA1c signal at the G6PD locus on chromosome X (rs1050828, p.Val98Met; minor allele frequency (MAF)=0.12, β=-0.41, P=8.4x10-205). This SNP lowers HbA1c independently of glycemia and so likely cause underdiagnosis of T2D when screening AA populations with HbA1c. We sought to identify additional distinct or sex-specific associations in this region through conditional and sex-stratified analyses.
Methods: Restricting the analyses to the ±500kb window flanking G6PD, we performed conditional analysis on rs1050828 using linear mixed-effect models adjusted for age, study, sex, and an empirical kinship matrix to account for relatedness. Associations with P<1.7x10-5 (0.05/2,886 variants with a minor allele count > 20) were considered distinct from rs1050828. We then performed stratified analyses in males (N=1,269) and females (N=1,854) and meta-analyzed the sex-stratified results. Analyses were conducted on the Analysis Commons using GENESIS.
Results: We identified a second distinct variant, rs76723693 (p.Leu353Pro, r2=0.0006, D’=1), that was rare (prevalence 0.5%) and had a slightly larger effect on HbA1c than rs1050828 (β=-0.44, P=2.2x10-9). The rs76723693 variant (G-allele frequency 0.5%) was associated with an absolute decrease in HbA1c of 0.98%-units (95% CI 0.51–1.44) per allele in hemizygous men and 0.46%-units (95% CI 0.26–0.66) in heterozygous women; whereas, the rs1050828 variant (T-allele frequency 12%) was associated with an absolute decrease in HbA1c of 0.88%-units (95% CI 0.81–0.95) per allele in hemizygous men and 0.34%-units (95% CI 0.30–0.39) in heterozygous women. We detected heterogeneity between males and females for rs1050828 (Phet=1.9x10-18) but not for rs76723693 (Phet=0.20). Both rs1050828 and rs76723693 are missense and putative pathogenic (ClinVar) seen in people with G6PD deficiency.
Conclusion: We are reporting for the first time the association with HbA1c of a rare G6PD variant, distinct from rs1050828. Like rs1050828, rs76723693 lowers HbA1c and causes G6PD deficiency. Rarer G6PD variants, distinct from rs1050828, may also cause underdiagnosis of T2D when screening AA populations with HbA1c. The clinical impact of rs76723693 on T2D diagnosis by HbA1c has yet to be explored.

The minor allele of rs373863828, a missense variant in CREBRF, has been associated with lower fasting glucose (FG) in Samoans. This variant is common in Polynesians (reported minor allele frequency [MAF] range: 0.096–0.259) but very rare in non-Polynesian populations (MAF<0.001). Other variation in the region may affect FG in other populations. We explore here whether rare variation in the CREBRF region is associated with FG in non-Polynesians in NHLBI’s TOPMed Program. We used whole-genome sequencing data from 11 TOPMed studies with reported FG and excluded individuals with type 2 diabetes. We divided the sample into four groups based on self-reported ancestry (African, n=6,551; Asian, n=2,309; European, n=18,013; and Hispanic, n=1,054). We used SKAT-O to test rare variation for association with FG, adjusting for age, sex, BMI, and ancestry. We selected rare SNPs in the CREBRF region (gene boundaries  50 base pairs) with a TOPMed-wide minor allele count  10 and excluded intergenic variants. Within each ancestry group, we excluded SNPs with a MAF > 0.05. First, we used a sliding window approach covering the region with 41 windows (4 kilobase (kb) with 2 kb overlap) with 28 to 348 variants per window and a median of 106. We set a conservative significance threshold of p<6×10−4 based on the number of non-overlapping windows and ancestry groups (0.05/(21×4)). We found a significant association between a European window of 263 variants and FG (p=2.9×10−4). The window comprises the last 3.3 kb of intron 8 and the first 0.7 kb of exon 9, including the stop codon. However, we did not see significant evidence of association in other ancestries, and the location of the peak association was not consistent between ancestry groups (African: p=0.014, 185 variants; Asian: p=0.023, 76 variants; Hispanic: p=1.7×10−3, 62 variants).

Second, we completed a gene-based analysis of the same CREBRF region. Our significance threshold was p<0.013 given the number of ancestry groups tested (0.05/4). In this analysis, we did not see significant evidence of association of CREBRF with FG in any ancestry (African: p=0.097, 3,319 variants; Asian: p=0.261, 1,671 variants; European: p=0.027, 4,643 variants; Hispanic: p=0.018, 1,336 variants). These results suggest that rare variation in the CREBRF region could be affecting the variance of FG in Europeans. It is possible that with more power association between rare variation and FG could be detected in other ancestries.

Type 2 diabetes (T2D) is prevalent in the Independent State of Samoa in Polynesia with an age­ standardized prevalence of 22.7% for men and 26.6% for women. Unique genetic factors may exist that contribute to this high prevalence. In addition, the effect of genes on susceptibility to T2D can differ between obese and lean cases. The Samoan population is genetically isolated, making it a prime candidate for genetics studies. We used whole­ genome sequence data generated through NHLBI’s TOPMed Program for 1,195 Samoans to create a Samoan­ specific reference panel to impute genotypes for an additional 1,897 Samoans. To identify genes and variants associated with T2D phenotypes, we completed genome­ wide association studies (GWAS) of (1) T2D, (2) T2D in obese (Polynesian cutoff of BMI≥32 kg/m2) and (3) T2D in non­obese with 9,609,170 sequenced and imputed variants, adjusting for age, sex, principal components of ancestry and empirical kinship (model 1). We also adjusted for BMI in model 2. In the T2D GWAS (n=478 cases, n=2,389 controls), a variant in a known T2D gene, PPARGC1A, was genome­ wide significant in both models 1 and 2 (p=1.30×10−8 and p=8.32×10−9, respectively). This variant is common both in other TOPMed populations (minor allele frequency (MAF)=0.376) and in Samoans (MAF=0.250). In the non­obese T2D GWAS (n=148 cases, n=1,119 controls), variants in or near three BMI­ associated genes were significantly associated with T2D in models 1 and 2 (in ADAM23, p=6.67×10−11 and p=2.08×10−11, respectively; near SATB2, p=1.37×10−10 and p=6.68×10−10; and near CREB1, p=1.17×10−9 and p=7.02×10−10), and a variant in MAP2, which has not been previously associated with T2D or related traits, was genome­ wide significant in models 1 and 2 (p=4.80×10−10 and p=5.22×10−11, respectively). These four variants are common in other TOPMed populations (MAF:0.094– 0.179), but rare in Samoans (MAF<0.009). In the obese T2D GWAS (n=330 cases, n=1,270 controls), no variants were genome­ wide significant. However, a suggestive (model 1: p=5.38×10−8) variant near known BMI gene TMEM182 is notable because this variant is rare in other TOPMed populations (MAF=0.0003), but common in Samoans (MAF=0.151). Additional exploration is necessary to further replicate these novel associations and characterize these variants. This highlights the importance of including diverse populations in genetics research to further characterize a map of disease­ associated variation.