Genome-Wide Association with Fasting Glucose and Insulin in 20,200 African Americans Suggests New Quantitative Trait Loci and Allelic Heterogeneity at Known Loci: the African American Glucose and Insulin Genetic Epidemiology (AAGILE) Consortium. J. Meigs1, M.-F. Hivert2, A. Morris3, M. Li4, M. Ng5, J. Liu6, R. Jensen7, X. Guo8, L. Yanek9, M. Nalls9, L. Bielak10, M. Irvin11, W.-M. Chen12, P. An13, E. Kabagambe14, B. Cade15, J. Wilson16, the. MAGIC Investigators17, J. Hong18, D. Rybin18, C.-T. Liu18, the. AAGILE Investigators19 1) Gen Med Div, Massachusetts Gen Hosp, Boston, MA; 2) University of Sherbrooke, Sherbrooke, Quebec, Ca; 3) University of Oxford, Oxford, UK; 4) Johns Hopkins University, Baltimore, MD; 5) Wake Forest University, Winston-Salem, NC; 6) Womens Health Initiative, Seattle, WA; 7) University of Washington, Seattle, WA; 8) Cedars Sinai Medical Center, Los Angeles, CA; 9) NIH, Bethesda, MD; 10) University of Michigan, Ann Arbor, MI; 11) University of Alabama, Birmingham, AL; 12) University of Virginia, Charlottesville, VA; 13) Washington University, St. Louis, MO; 14) Vanderbilt University Medical Center, Nashville, TN; 15) Brigham and Women's Hospital, Boston, MA; 16) University of Mississippi Medical Center, Jackson, MS; 17) USA and Europe; 18) Boston University School of Public Health, Boston, MA; 19) USA.
High fasting glucose (FG) and insulin (FI) disproportionately affecting African Americans (AA) may have a genetic basis. We used meta-analyses (m-a) of genome-wide (g-w) association studies (GWAS) of FG and FI in AA to test whether loci identified in European (EU) ancestry individuals also are associated in AA, and to find new AA quantitative trait (QT) loci. We performed GWAS in 16 cohorts of 20,209 (FG) and 17,871 (FI, adjusted for BMI) non-diabetic AA (mean age 56 yr) using additive genetic models to test associations with 3.3M single nucleotide polymorphisms (SNPs), and combined results in METAL using inverse-variance weighted m-a. To leverage possible AA-EU heterogeneity at each SNP, we combined AA METAL results with MAGIC published results (Manning 2012, N= up to 96,496 EU in 29 cohorts) and m-a the two results files using MANTRA, a Bayesian method accounting for allelic heterogeneity among population clusters that returns a Bayes Factor, with (logBF) >6 suggesting g-w SNP-trait association. We evaluated associations for 23 known FG and 8 FI loci by testing reported EU Index SNPs (Dupuis 2010, Manning 2012) in AA and also identifying the Best SNP within +/- 250 kb of the EU Index SNP. We sought new loci for replication based on low METAL P values in AA and high MANTRA logBF in AA+EU. For 23 known FG loci, 1 EU Index (MTNR1B) and 1 AA Best SNP (GCK) were g-w significant (P<2.5x10-8) in AA, 9 Index and 23 Best SNPs were nominally significant (P<0.05), with 13 Best SNPs remaining significant after P value correction for the number of SNPs tested at the locus. At 14/23 FG loci the r2 (YRI HapMap 2) for Index vs Best SNP was <0.2, suggesting potentially independent signals. For 8 known FI loci, no Index or Best SNPs were g-w significant in AA, but 3 Index and 8 Best SNPs had P<0.05, with 5 Best SNPs remaining significant after P value correction. At 4/8 FI loci the r2 Index vs. Best SNP was <0.2. For new discovery, 14 novel AA FG and 8 novel AA FI loci had a SNP P<10-5 (in AA) and logBF >4 (in AA+EU), giving 22 high-interest SNPs now being tested for replication. We conclude that all 31 FG and FI loci known in EU show at least nominal association in AA, suggesting some genetic determinants of FG and FI are similar across AA and EU, but with apparent allelic heterogeneity at many loci. Combining fixed effects m-a in an AA sample with trans-ethnic m-a in an AA-EU sample has identified a wealth of new AA diabetes-related QT SNPs to test for replication.
You may contact the first author (during and after the meeting) at