The majority of genetic variants significantly associated with glycemic traits reside in the non-coding genome, with many causal variants still unknown. Whole genome sequence association (WGSA) analysis allows us to (1) fine-map known and novel loci without depending on imputation, (2) assess enrichment of sub-significant associations across non-coding annotations, and (3) discover novel rare variant associations. Here, we present an initial WGSA of fasting insulin (FI) and fasting glucose (FG) levels in phase 1 TOPMed data (N=7121) with deep (>30x) sequence coverage in five cohorts, three European-ancestry (EA): Framingham Heart Study, N=3209; Old Order Amish Studies, N=980, Cleveland Family Study, N=197, and two African-American (AA): Jackson Heart Study, N=2487, Cleveland Family Study, N=248. For each cohort, we used linear mixed effects models (as implemented in EMMAX or MMAP software) adjusting for sex, age and BMI, with empirical kinship for relatedness and population structure. We restricted to variants with minor allele count > 5 in more than 1 cohort and meta-analyzed within and across ancestry with METAL. In EA+AA analysis, common (minor allele frequency [MAF]>5%) variant associations (P<5x10-8) were identified for FG at known loci: MTNR1B (rs10830963, P=2.5x10-16; rs12792753, P=1.4x10-8), GCK (rs4607517, P=1.2x10-10, and 13 additional variants), and G6PC2 (rs560887, P=5.4x10-10). At MTNR1B, multi-ancestry fine-mapping reduced the significant FG associations from 46 (EA) to 2 (AA+EA). Novel FI associations were seen in rare (MAF<1%) variants: 18q12.1 (rs146884135, P=2.1x10-9), 4p12 (rs193168677, P=1.2x10-8), 1q25.3 (rs550837507, P=3.8x10-8), and DCLK3 (rs573417731, P=4.8x10-8). These results, which are being extended with phase 2 TOPMed cohorts (expected total N=27830), highlight the value of multi-ancestry WGSA to refine known and discover novel associations for complex traits.

Background
A1C GWAS have identified 17 common genetic variants of which 11 act through erythrocyte pathways and may interfere with A1C diagnostic accuracy. Genetic effect size and minor allele frequency (MAF) may differ under differential erythrocytic selection pressures. We used WGS data to determine ancestral differences in associations with A1C across the allelic frequency spectrum in white and black populations.

Methods
We performed single nucleotide variant (SNV) WGS association analyses with A1C in 4,889 non-diabetic individuals (2,466 Framingham Heart Study (FHS) whites; 2,523 Jackson Heart Study (JHS) blacks) within the NHLBI Trans-Omics for Precision Medicine Program (TOPMed). We conducted mixed-effect linear regressions with an empirical kinship matrix adjusted for sex, age and principal components in the 11 GWAS loci (+/-100kb around the lead SNV). Within each locus, we performed conditional analyses on the lead SNV in FHS to identify distinct signals (P≤0.005). We performed fixed-effect meta-analysis across both ancestries using METAL and Fisher’s exact test to evaluate ancestral differences in MAF.

Results
The HK1 locus (10q22) was associated with A1C at genome-wide significance (rs72805692, MAFFHS=0.1, βFHS=-0.07; MAFJHS=0.02, βJHS=-0.11, PMeta=4x10-9). We identified signals in 5 of the 11 loci (SPTA1, HBS1L(MYB), HK1, CDT1, HIST1H4A/HFE) at P<0.001 in at least one cohort. We observed no ancestral differences in effect sizes for the strongest signal (Cochran test, PQ≥ 0.1); though MAFs were generally higher in whites compared to blacks (PFisher 0.4 to <2x10-16). At 4 loci, we identified 5 low frequency and rare distinct signals with marked ancestral differences in MAF (PFisher<0.01).

Conclusion
Genetic variation at erythrocytic A1C GWAS loci exhibits ancestral differences in MAF. Understanding the extent to which such ancestral differential genetic effects interfere with A1C accuracy is a next step to reduce ethnic health disparities in diabetes.

Background
A1C GWAS have identified 17 common genetic variants of which 11 act through erythrocyte pathways and may interfere with A1C diagnostic accuracy. Genetic effect size and minor allele frequency (MAF) may differ under differential erythrocytic selection pressures. We used WGS data to determine ancestral differences in associations with A1C across the allelic frequency spectrum in white and black populations.
Methods
We performed single nucleotide variant (SNV) WGS association analyses with A1C in 4,889 non-diabetic individuals (2,466 Framingham Heart Study (FHS) whites; 2,523 Jackson Heart Study (JHS) blacks) within the NHLBI Trans-Omics for Precision Medicine Program (TOPMed). We conducted mixed-effect linear regressions with an empirical kinship matrix adjusted for sex, age and principal components in the 11 GWAS loci (+/-100kb around the lead SNV). Within each locus, we performed conditional analyses on the lead SNV in FHS to identify distinct signals (P≤0.005). We performed fixed-effect meta-analysis across both ancestries using METAL and Fisher’s exact test to evaluate ancestral differences in MAF.
Results
The HK1 locus (10q22) was associated with A1C at genome-wide significance (rs72805692, MAFFHS=0.10, βFHS=-0.07; MAFJHS=0.02, βJHS=-0.11, PMeta=4x10-9). We identified signals in 5 of the 11 loci (SPTA1, HBS1L(MYB), HK1, CDT1, HIST1H4A/HFE) at P<0.001 in at least one cohort. We observed no ancestral differences in effect sizes for the strongest signal (Cochran test, PQ≥ 0.1); though MAFs were generally higher in whites compared to blacks (PFisher 0.4 to <2x10-16). At 4 loci, we identified 5 low frequency and rare distinct signals with marked ancestral differences in MAF (PFisher<0.01).
Conclusion
Genetic variation at erythrocytic A1C GWAS loci exhibits ancestral differences in MAF. Understanding the extent to which such ancestral differential genetic effects interfere with A1C accuracy is a next step to reduce ethnic health disparities in diabetes.