- A sizeable portion of the heritability of type 2 diabetes remains unexplained by loci identified using additive-model genome-wide association studies (GWAS)
- This study, the largest recessive-model GWAS of type 2 diabetes conducted to date, combined data from seven discovery cohorts (33,139 cases and 279,507 controls) and two replication cohorts (28,336 cases and 62,253 controls)
- 51 loci reached genome-wide significance, of which five had never been detected in additive analyses, and one of them, rs115018790, was replicated in both cohorts (meta-analysis OR, 2.56; P=1 × 10-16)
- The phenotypic effects of rs115018790 extended to type 2 diabetes-related traits such as levels of triglycerides and LDL, HDL and total cholesterol
- Future recessive GWAS of type 2 diabetes and glycemic traits are expected to detect additional associations that may improve polygenic risk scores and perhaps identify new therapeutic targets
Over the past decade, hundreds of genetic loci associated with type 2 diabetes mellitus have been identified by large genome-wide association studies (GWAS). Most of these studies have relied on additive models, which assume a uniform, linear increase in risk for each copy of a major allele.
Subscribe to the latest updates from Endocrinology Advances in Motion
An additive model is computationally simple, but it can miss genetic risk variants with recessive effects, particularly low-frequency variants. Researchers at Massachusetts General Hospital recently conducted the largest GWAS meta-analysis to date that used a recessive model and found at least one new variant that has substantial effects on lipids and the risk of diabetes.
Massachusetts General Hospital's Josep M. Mercader, PhD, a researcher in the Diabetes Unit and Center for Genomic Medicine and group leader at the Broad Institute of MIT and Harvard, Jose C. Florez, MD, PhD, chief of Mass General's Endocrine Division and the Diabetes Unit and co-director of the Metabolism Program at the Broad Institute, Aaron Leong, MD, MSc, an endocrinologist in the Diabetes Unit and researcher in the Center for Genomic Medicine, Mark O'Connor, MD, former research fellow, Varinderpal Kaur and Philip Schroeder of the Diabetes Unit, and colleagues detail their findings in Diabetes.
The discovery sample consisted of 33,139 participants with type 2 diabetes and 279,507 control subjects from the UK Biobank, the Mass General Brigham Biobank and five cohorts known collectively as 70K for T2D. The researchers compared their results with the two largest additive GWAS to date that provide summary statistics, one published in Nature Genetics in 2018 and the other published in that journal in 2020.
51 loci reached genome-wide significance in the recessive model, of which five were distinct from previously reported additive signals. Two of the novel variants had minor allele frequency (<5%) and were associated with more than double the risk of diabetes in homozygous carriers:
- rs115018790—OR, 2.63 (P=3 × 10−13), located within an intron of the PELO and ITGA1 genes on chromosome 5
- rs140453320—OR, 6.94 (P=5 × 10−9), located within an intron of the ADAMTS6 gene on chromosome 5
By way of comparison, most type 2 variants identified via additive analyses have ORs <1.1.
In two separate cohorts (28,336 subjects with diabetes and 62,253 control subjects), three of the novel variants replicated, one did not and there was insufficient power to study the fifth. rs115018790 replicated in both cohorts (meta-analysis OR, 2.56; P=1 × 10−16). rs140453320 could be assessed only in one cohort.
The Lead Variant
The researchers took the most significant variant, rs115018790, into further analyses:
- The co-localization analysis confirmed a link between rs115018790 and reduced PELO expression in the pancreas, adipose tissue, skeletal muscle, and other tissues
- In a phenome-wide association study of the UK Biobank, homozygotes for rs115018790 had a 20% increase in triglycerides, 10% decrease in LDL cholesterol, lower levels of HDL and total cholesterol, higher levels of albumin, and lower levels of C-reactive protein
- rs115018790 was also associated with non-biomarker phenotypes, including a variety of hematologic features, increased frequency of alcohol intake and metformin use
- The effect of rs115018790 on both type 2 diabetes and biomarkers was significantly stronger in men than women
- A metabolome-wide association study linked rs115018790 to changes in multiple lipid-related phenotypes, particularly decreased cholesterol percentage and increased triglyceride percentage in large LDL particles in men, and for both parameters, there was evidence of causal mediation
The Merits of Non-additive Models
In previous work published in Nature Communications, Mass General researchers conducted non-additive GWAS for several age-related diseases in a cohort dubbed GERA. They identified a rare variant associated with type 2 diabetes that was associated with reduced expression of the PELO gene, just as rs115018790 was in the current study.
Little is known about the connection between PELO and diabetes, but these new findings show that a variant linked to reduced expression of the gene has substantial effects on lipid metabolism as well as diabetes.
Future recessive GWAS of type 2 diabetes and glycemic traits may detect additional associations that will provide a better understanding of diabetes pathophysiology, enhance the predictive power of polygenic risk scores and perhaps identify new therapeutic targets.
Learn more about the Center for Genomic Medicine
Refer a patient to the Diabetes Unit