Type 1 Diabetes Genetics Consortium (T1DGC): Multi-Ethnic ImmunoChip Study

Type 1 diabetes (T1D) is characterized by an autoimmune attack that destroys the insulin-producing pancreatic beta cells and is driven by diverse genetic and environmental factors. Between 8-100 chromosome regions are known to contain variants that alter T1D risk, substantially fewer than many other common diseases. Less is known about the contribution of T1D risk alleles in non-European populations, despite recent increases in T1D diagnoses in multiple non-European ancestry groups. To address these gaps in T1D genetics, we doubled the sample size from the previous largest T1D association study, genotyped ancestrally diverse T1D cases, controls, and affected families with the Illumina ImmunoChip array, and imputed additional variants using a large haplotype reference panel. After quality filtering, a total of 61,427 participants and 140,333 genotyped ImmunoChip variants were included in analysis, providing dense coverage in 188 autosomal regions ("ImmunoChip regions") and sparse genotyping in other regions of the genome. Each participant was assigned to one of five ancestry groups using k-means clustering of ImmunoChip genotype principal components: European (EUR, N = 47,319), African Admixed (AFR, N = 4,290), Finnish (FIN, N = 6,991), East Asian (EAS, N = 588) and Other Admixed (AMR, N = 2,239). In total, 16,159 T1D cases, 25,386 controls (unrelated) and 6,143 trio families (i.e., an affected child and both parents) were genotyped and included in association analyses. Imputation with the National Heart, Lung, and Blood Institute (NHLBI) Trans-Omics for Precision Medicine (TOPMed) multi-ethnic reference panel was used to improve discovery power and fine-mapping resolution. After imputation, the number of variants in ImmunoChip regions with imputation R-squared > 0.8 and MAF > 0.005 in each ancestry group was 322,084 (AFR), 166,274 (EUR), 163,612 (FIN), 137,730 (EAS) and 188,550 (AMR). Imputed genotypes were compared in the same participants who have whole genome sequencing data, resulting in high concordance. Unrelated cases and controls (N = 41,545) were analyzed initially, assuming an additive inheritance model. T1D trio families (N = 6,143 trios, some trio families were multiplex and analyzed as multiple trios) were analyzed separately, with meta-analysis of case-control and trio results used for discovery of regions associated with T1D (p < 5×10-8). Summary-level data at each SNP (genotype and imputed) for type 1 diabetes (outcome) are provided in dbGaP.]]>
We utilized samples and data from a variety of sources to establish a cohort of 61,427 participants, including 16,159 T1D cases, 25,386 controls and 6,143 trio families (parents with a T1D child). Genotypes obtained from the ImmunoChip were imputed to the TOPMed Imputation Server and used for association analyses. To be included in the T1DGC Multi-Ethnic ImmunoChip Study, families, trios and case-control units had to meet specific criteria. For an affected sibling pair (ASP) family to be enrolled into the T1DGC, the family had to have at least one ASP available for sampling, with preference given to families with availability of one or both biological parents and at least one unaffected sibling. For trio families, both parents were necessary to include with the single T1D case. For cases, the individual had to be diagnosed with type 1 diabetes (T1D) with onset < 16 years and required use of insulin within 6 months of diagnosis with no concomitant disease or disorder associated with diabetes. The same diagnostic criteria hold for the case in trio families, and the proband in ASP families. In ASP families, the affected sibling could be included with onset < 35 years. For controls, the participant had no evidence of type 1 diabetes.]]>
The T1DGC utilized cases, controls, affected sibpair and trio families of European Caucasian ancestry to conduct the original GWAS meta-analysis of type 1 diabetes (Barrett JC et al, Nat. Genet 2009; 41, 703–707. PMID: 19430480 PMCID: PMC2889014). The significant regions identified for T1D were combined with other collections of European Caucasian ancestry and genotyped on the custom fine-mapping array (ImmunoChip) to refine loci, establish credible sets of SNPs, and determine the likely functional impact using ENCODE and RoadMap data (Onengut-Gumuscu S et al, Nat. Genet 2015; 47, 381–386. PMID: 25751624 PMCID: PMC4380767). As analysis of ImmunoChip data in samples of African ancestry had demonstrated overlap in loci but not always the same credible sets of variants in a locus in European Caucasian populations (Onengut-Gumuscu S et al, Diabetes Care 2019; 42, 406-415. PMID: 30659077 PMCID: PMC6385701), the effort to conduct the fine-mapping in populations of diverse genetic ancestry was undertaken. The current data represent the results of multi-ethnic fine-mapping for type 1 diabetes.]]>