Genome-wide association for fasting plasma glucose was performed among 5,089 nondiabetic Indian Asians (fasting glucose levels <7.0 mmol/l; free from pharmacologic or dietary treatment for diabetes) (12
) and genotyped using the Illumina Hap610 BeadChip, and a separate sample of 2,385 Indian Asians was genotyped with the Illumina Hap300 BeadChip. Results for the Hap610 sample were analyzed separately, followed by meta-analysis of the results from Hap610 and Hap300 genotyping arrays. To examine the possible contribution of single nucleotide polymorphisms (SNPs) identified to raised glucose in Indian Asians, allele frequencies and effect sizes were compared with results for 4,462 European Caucasian participants of the Northern Finland Birth Cohort of 1966 (NFBC1966).
All Indian Asian participants were recruited though the London Life Sciences Prospective Population (LOLIPOP) Study. LOLIPOP is an ongoing population-based cohort study of ~30,000 Indian Asian and European Caucasian men and women aged 35–75 years and recruited from the patient records among 58 general practitioners in West London, U.K. (13
). Indian Asians were identified as having all four grandparents born on the Indian subcontinent. Response rates averaged 62%; there were no material differences between responders and nonresponders with respect to age, sex, comorbidity, and available data for risk factors. All participants gave written informed consent, including for genetic studies. The study is approved by the local Research Ethics Committee.
An interviewer-administered questionnaire was used to collect data on medical history, family history, current prescribed medication, and cardiovascular risk factors. Physical assessment included anthropometric measurements (height, weight, waist circumference, and hip circumference) and blood pressure. Blood was collected after an 8-h fast for biochemical analysis, including glucose, insulin, total and HDL cholesterol, and triglycerides, and whole blood was taken for extraction of DNA. Insulin sensitivity and pancreatic β-cell function were estimated using the homeostatic model assessment (HOMA) method (14
) where insulin sensitivity (HOMA-S) = 22.5 (fasting plasma insulin × fasting plasma glucose) and β-cell function (HOMA-B) = (20 × fasting plasma insulin)/(fasting plasma glucose − 3.5).
We studied 5,089 nondiabetic Indian Asians genotyped with the Illumina Hap610 BeadChip (Hap610 sample) and 2,385 Indian Asians genotyped with the Hap300 BeadChip (Hap300 sample). The Hap300 sample included 698 Indian Asians with type 2 diabetes, enabling testing of SNPs against type 2 diabetic case subject or control subject status.
All European Caucasians were participants of NFBC1966, a study of factors affecting preterm birth, low birth weight, and subsequent morbidity and mortality. Participants comprise 12,068 mothers and their 12,231 births in the provinces of Oulu and Lapland during 1966. At age 31 years, all individuals still living in the Helsinki area or northern Finland were asked to participate in a detailed biological and medical examination (n
= 6,007) including measurement of fasting glucose. For the current study, genotype and fasting glucose measurements were available for 4,462 NFBC1966 participants (2,116 male and 2,346 female). Study methods and characteristics of participants have been described previously (11
). The University of Oulu ethics committee approved the study. The NFBC1966 samples used were also included in the genome-wide association studies that identified common genetic variation in and around G6PC2
as determinants of glucose levels in European populations (7
Genotyping using the Illumina Hap610 BeadChip was carried out according to standard methodology. In brief, each sample was whole genome amplified, fragmented, precipitated, and resuspended in appropriate hybridization buffer. Denatured samples were hybridized on prepared HumanHap610 BeadChips for a minimum of 16 h at 48°C. After hybridization, the BeadChips were processed for the single-base extension reaction as well as staining and imaging on an Illumina Bead Array Reader. Normalized bead intensity data obtained for each sample were loaded into the Illumina Beadstudio 2.0 software, which converted fluorescence intensities into SNP genotypes. Based on 17 duplicate scans, mean genotyping concordance rate was 99.9987% (range 99.9930–99.9998%). For the 582,539 autosomal SNPs included on the Hap610 BeadChip, average SNP call rate was 99.7%, with call rates >95% for 99.7% of SNPs. Call rate was >95% for 99.1% of people; 46 people were excluded for call rates <95%. We also excluded SNPs with minor allele frequency <0.01, call rate <0.95%, or Hardy-Weinberg equilibrium of P
. This left 544,390 autosomal SNPs for the genome-wide association analysis. Methods and quality control for genotyping carried out using the Illumina Hap300 BeadChip in Indian Asians and the Illumina Hap370 BeadChip in European Caucasians have been described previously (11
Single SNP marker tests were carried out for association with fasting glucose levels in nondiabetic Indian Asians under an additive genetic model with adjustment for age and sex. Principal components analysis was used to characterize population substructure in Indian Asians genotyped on the Hap610 and Hap300 BeadChips, and the top four components were included in models (15
). Analysis of QQ plots for association over all SNPs showed good adherence to null expectations, indicating that the approach was sufficient to allow for any inflation due to population substructure (see supplemental Fig. S1, available in the online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db08-1805/DC1
). λ values for the genome-wide analyses were 1.01 in both the Illumina Hap610 and the Illumina Hap300 samples (compared with 1.07 and 1.02, respectively, for analyses without principal components). The Hap610 sample size provided 80% power to identify SNPs associated with 0.8% of population variation in glucose at genome-wide significance (P
< 5 × 10−8
We then carried out a meta-analysis of results from the Illumina Hap610 and the Illumina Hap300 samples. Imputation was done using a hidden Markov model algorithm implemented in MACH software and pooled phased haplotypes for the CEU, CHB/JPT, and YRI samples from HapMap build35, dbSNP build 125 (17
). Imputed SNPs with minor allele frequency <0.01 or low-quality score (r2
< 0.30) were removed. This generated ~1.9 million directly genotyped or imputed autosomal SNPs per participant with data available in both samples. Meta-analysis was carried out using z
scores weighted by square root of sample size, implemented in the software package METAL (www.sph.umich.edu/csg/abecasis/metal
). QQ plots showed good adherence to null expectations (λ for meta-analyzed data = 1.047; supplemental Fig. S1).
The associations of SNPs with type 2 diabetes and other phenotypic traits, and of risk allele score with glucose levels, were tested in the Illumina Hap300 data using regression analysis and an additive genetic model. Analyses were adjusted for age and sex in Indian Asians as well as for sex among European Caucasians (NFBC1966 participants are all 31 years of age). Heterogeneity of effect between Indian Asians and European Caucasians was tested by joint analysis of data for Indian Asians and European Caucasians, with incorporation of ethnicity and genotype-ethnicity interaction terms into the regression models. Two SNP regression models were used to identify whether SNPs from the same genetic locus have separate relationships with glucose levels. Glucose risk allele scores were also compared between Indian Asians and European Caucasians by independent samples t test.