Introduction

Type 2 diabetes and obesity have reached epidemic proportions worldwide and the prevalence is increasing at rising rates, especially in developing countries.1, 2 Both genetic and environmental factors with overlapping etiological mechanisms are believed to be involved in the development of obesity and type 2 diabetes. FTO (fat mass- and obesity-associated gene) is one of the best examples of diabetogenic gene mediating its effect through obesity. FTO was first identified as susceptibility gene for obesity3 and later body mass index (BMI)-dependent association with type 2 diabetes was revealed in European population.4 Among the variants known to influence obesity, FTO variants had maximal effect in Europeans.5 Subsequently, a number of studies confirmed the association of common variants of FTO in population of European and non-European ethnicity.6, 7, 8, 9, 10 Independent studies in Asian population11, 12, 13, 14, 15 had yielded conflicting results, but a recent meta-analysis of common variants of FTO in East Asian population confirmed their association with obesity and type 2 diabetes.16

However, association of common variants of FTO in Indian population remains inconclusive. Two studies from Pune–Mysore17 and North Indian Sikhs18 demonstrated a BMI-independent association of rs9939609 with type 2 diabetes, but failed to replicate its association with obesity. A recent study on a South Indian population from Chennai19 demonstrated BMI-dependent association of rs8050136 with type 2 diabetes and also with obesity. These contradictions in association results for variants in FTO with obesity and type 2 diabetes demand their further validation in Indian population.

In this study, we evaluated association of previously associated common variants of FTO in intron 1 (rs9939609, rs1421085, rs8050136 and rs9930506) and additional variants (rs1861867—intron 2, rs9926180—intron 7, rs2540769—intron 8 and rs708277—3′-UTR) of FTO with type 2 diabetes and obesity in North Indians of Indo-European ethnicity.

Materials and methods

Study participants

Study involved participation of 5364 subjects from North India belonging to Indo-European ethnicity residing in and around Delhi, in two stages. In stage 1, association analysis was performed in 2115 subjects comprising of 1073 type 2 diabetes patients and 1042 controls, who were recruited before September 2008. All type 2 diabetes patients of stage 1 were recruited from the Clinic (Endocrinology Clinic, All India Institute of Medical Sciences), whereas control subjects were collected by organizing ‘Diabetes Awareness Camps’. In stage 2, further replication was assessed in an independent sample set of 3249 individuals recruited after September 2008 that included 1401 type 2 diabetes patients (recruited from Clinic and ‘Diabetes Awareness Camps’) and 1848 controls (enrolled from ‘Diabetes Awareness Camps’). Type 2 diabetes was diagnosed as per the World Health Organization criteria.20 Control subjects had age ⩾40 years, no family history of diabetes with glycosylated hemoglobin level ⩽6.0% and fasting glucose level ⩽6.11 mmol l−1. Detailed inclusion and exclusion criteria for type 2 diabetes patients and control subjects have been provided earlier.21, 22 Informed consent was obtained from all the participants and the study was approved by the Ethics committee of the participating institutions. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Clinical measurements

All subjects were extensively characterized for anthropometric and quantitative metabolic traits. Anthropometric measurements including height, weight, waist circumference (WC), hip circumference, systolic and diastolic blood pressure were taken as per standardized protocols. BMI and waist-to-hip ratio (WHR) were calculated. Biochemical measurements including levels of glycosylated hemoglobin, fasting plasma glucose, 2-h post-load plasma glucose, fasting plasma insulin, C-reactive protein, C-peptide, total cholesterol (TC), high-density lipoprotein-cholesterol, low-density lipoprotein-cholesterol, triglyceride, blood urea, creatinine and uric acid were performed using standard laboratory assays as described earlier.21 Homeostasis model assessment for insulin resistance was calculated using the formula: (fasting plasma insulin (mU l−1) × fasting plasma glucose (mmol l−1))/22.5 as reported previously.23 General characteristics of study population are presented in Table 1.

Table 1 General characteristics of the study populations
Full size table

SNP selection

Four single-nucleotide polymorphisms (SNPs) from intron 1 (rs1421085, rs8050136, rs9939609, rs9930506) that were previously shown to be associated with obesity/type 2 diabetes were selected. In addition, four more SNPs (rs1861867—intron 2, rs9926180—intron 7, rs2540769—intron 8 and rs708277-3′—UTR) were selected from the database (NCBI dbSNP Build 129) to cover the gene at regular intervals. These SNPs were prioritized based on their minor allele frequency (MAF) (>0.05), polymorphic state in at least two population and localization in functionally important region.

Genotyping

Genotyping in stage 1 was performed by Golden Gate assay using Illumina platform (http://www.illumina.com/documents/products/workflows/workflow_goldengate_assay.pdf). Stringent quality control criteria were met to qualify an SNP for further analysis, including genotype confidence score >0.25, call frequency >0.9, GenTrans score >0.6, cluster separation score >0.4, MAF >0.05 and Hardy–Weinberg equilibrium P>0.01 in controls.24 A total of 147 samples (7.2%) were also genotyped in duplicate and an error rate of <0.01 was estimated. The variant rs708277 did not qualify the quality control (GenTrans score <0.6) and was excluded from further analysis.

In stage 2, genotyping of rs9926180 (associated in stage 1) and rs9939609 (established obesity/type 2 diabetes variant in stage 2) was performed in 1047 cases and 1038 controls using iPLEX (Sequenom, San Diego, CA, USA). Approximately 5% (N=110) of the samples of stage 2 were genotyped in duplicates and a concordance of 99.99% was observed. In addition, 810 controls were genotyped for all the seven SNPs of stage 1 and 354 cases for rs9939609 and rs9926180, to increase power of the study. These samples were genotyped by single base extension using SNaPshot ddNTP Primer Extension Kit (Applied Biosystems, Foster City, CA, USA) and electrophoresis of extended probes on an ABI Prism 3730 Genetic Analyzer (Applied Biosystems). The genotype calls were determined using GeneMapper. Approximately 6% of the samples were genotyped in duplicates and a concordance of 99.99% was observed. SNaPshot assay for two SNPs rs1861867 and rs2540769 failed in this phase.

Statistical analysis

Statistical analyses were performed using PLINK v. 1.07 (http://pngu.mgh.harvard.edu/purcell/plink/25) and SPSS v. 17.0 (SPSS, Chicago, IL, USA), unless specified otherwise. Statistical power of the study was calculated under log-additive model, assuming 10% population risk using Quanto (http://hydra.usc.edu/gxe/) (Supplementary Tables 1 and 2, Supplementary Figure 1). Hardy–Weinberg equilibrium for genotypes was checked by χ2 analysis. Linkage disequilibrium between SNPs was estimated using Haploview v. 4.1.26 Single marker- and haplotype-based association with obesity and type 2 diabetes was assessed by logistic regression analysis assuming log-additive model. A P-value of <0.01 (α=0.05/5) was considered significant after Bonferroni correction for the number of independent SNPs investigated. All quantitative trait values were inverse normally transformed to achieve normal distribution and then analyzed using linear regression. Means and standard deviations are presented in inverse normal units of the parameters in the tables. For quantitative traits association, a P-value of <0.00053 (α=0.05/(5 × 19)) was considered significant (association for five independent loci with 19 traits). Association with obesity and quantitative traits was performed only in control subjects.

Meta-analysis of stage 1 and 2 results were performed by combining summary data of two study population both under fixed and random models. Similarly, the summary data of previous studies on Indians for association with type 2 diabetes17, 18, 19 and this study were combined for meta-analysis. Association of variants with type 2 diabetes, obesity and quantitative traits were also performed by combining data for two study population and adjusting for study population in addition to other covariates. Non-parametric Mann–Whitney U-test was performed to compare the median values of clinical traits of two study population. Allele frequencies of cases and controls of two study population were compared by equality of proportions Z-test (Supplementary Table 3).

All analyses performed were adjusted for age, sex and BMI as appropriate. The odds ratios (ORs) and 95% confidence interval (CI) were calculated with respect to minor alleles and uncorrected P-values are presented in the manuscript.

Results

Association with type 2 diabetes

In stage 1, rs9926180 (intron 7) was associated with type 2 diabetes that remained significant after adjusting for sex, age and BMI (ORadj=1.24, 95% CI=1.07–1.43, Padj=0.005) and even after correcting for multiple tests (Table 2). However, association of rs9926180 could not be replicated in another independent study population in stage 2. Meta-analysis and combined analysis of two stages also did not show association of rs9926180 with type 2 diabetes (Table 4). None of the previously reported SNPs showed either BMI-dependent or -independent association with type 2 diabetes in our population. Interestingly, the most widely studied variant rs9939609 also failed to show any association with type 2 diabetes in our study population. Similar results were obtained when association analysis were performed by adjusting for sex, age and BMI as covariates and only BMI as a covariate (Table 2).

Table 2 Association of variants with type 2 diabetes
Full size table

Association with obesity and its measures

For association analysis with obesity, control subjects were categorized as normal weight (BMI <23 kg m−2), overweight (BMI 23–25 kg m−2) and obese (BMI ⩾25 kg m−2) based on BMI cutoffs for Asians defined by World Health Organization expert committee27 (Supplementary Table 4). None of the variants showed association with BMI-defined obesity in stage 1 of the study (Table 3). However, in stage 2, rs1421085 (OR=1.57, 95% CI=1.21–2.04, P=0.0007) and rs8050136 (OR=1.37, 95% CI=1.06–1.77, P=0.02) showed association with obesity when we compared obese/overweight with normal weight subjects. In the same stage, the association was also observed when we compared obese with normal weight subjects (rs1421085: OR=1.65, 95% CI=1.26–2.17, P=0.0003; rs8050136: OR=1.45, 95% CI=1.11–1.91, P=0.007) with increased risk. The associations were also observed after combined analysis and meta-analysis (Table 4). As WC has been shown to be a better predictor of obesity,28 we also tested association by defining obesity using WC (>90 cm in men and >80 cm in women)28 and WHR (>0.95 in men and>0.80 in women)29 as measures of central obesity (Supplementary Table 5). The same variant rs1421085 was associated with WC-defined obesity in stage 2 (OR=1.43, 95% CI=1.11–1.85, P=0.006) and combined study population (OR=1.23, 95% CI=1.06–1.43, P=0.006). In WHR-defined obesity analysis, rs1421085 and rs9930506 showed nominal association only in stage 2 of the study. Association analysis with quantitative measures of obesity (BMI, WC and WHR) in stage 1 showed only nominal association of rs1421085 (P=0.04), rs8050136 (P=0.03) and rs9939609 (P=0.03) with WC (Supplementary Table 6). However in stage 2 (Supplementary Table 7), association with BMI and WC was observed for rs1421085 (BMI: P=0.0002; WC: P=0.001), rs8050136 (BMI: P=0.002; WC: P=0.05) and rs9930506 (BMI: P=0.04; WC: P=0.02). The variant rs9939609 also showed nominal association with WC (P=0.03). In the combined study population (Supplementary Table 8), all the above three variants were associated with BMI and WC in addition to nominal association of rs9939609 with WHR (P=0.03).

Table 3 Association of variants with BMI-defined obesity (controls only)
Full size table
Table 4 Meta-analysis for the two study populations
Full size table

Association with quantitative trait

We also examined association of FTO variants with other quantitative traits related to obesity and type 2 diabetes to see if they affect risk of obesity/type 2 diabetes by influencing them. In stage 1, rs9926180 was associated with high-density lipoprotein-cholesterol (P=0.002), and in stage 2, rs9930506 was associated with triglycerides (P=0.002). Nominal associations of FTO variants with other traits were also observed (Supplementary Table 6).

Haplotype-based association

Intron 1 variants—rs1421085, rs8050136 and rs9939609 formed a haplotype block encompassing a region of 19 kb (Supplementary Figure 2). Association analysis of this haplotype block with type 2 diabetes, obesity and BMI failed to show any association (Supplementary Table 9).

Meta-analysis with other studies

Meta-analysis of the four studies on Indians (including this study) for association with type 2 diabetes (Figure 1) showed association under the random-effect (ORrandom=1.16, P-valuerandom=0.02) and fixed-effect models (ORrandom=1.21, P-valuerandom=0.001).

Figure 1
figure 1

Forest plot of meta-analysis (fixed and random effects) for association of rs9939609 with type 2 diabetes in Indians. (a) Fixed effect and (b) random effect. *Pune study17 involved participation of both Indo-European (2814) and Dravidian Indians (961). In all studies, analysis was adjusted for sex, age and body mass index (BMI), except for Indo-Europeans (Pune), where analysis was adjusted only for BMI. For Dravidian Indians,19 rs8050136 was considered as proxy single-nucleotide polymorphism (SNP) for rs9939609 for this analysis as this SNP was not studied in Dravidian Indians. Test for heterogeneity: Q=6.571 on 3 d.f. (P=0.09). Moment-based estimate between studies variance=0.008.

Full size image

Discussion

Here we explored association of FTO variants with type 2 diabetes, obesity and related quantitative traits in North Indians. We did not find FTO variants to influence the risk of type 2 diabetes in our study population. Our preliminary study suggested a BMI-independent association of a new variant rs9926180 in intron 7 with type 2 diabetes, but that could not be replicated in an independent study population. However, variants of intron 1 (rs1421085, rs8050136 and rs9939609) were associated with obesity status and/or measures of obesity (BMI, WC and WHR). This is in accordance with previous reports of these variants for association with obesity.

It is pertinent to mention that consistent with results of the studies in other population, our previous study demonstrated association of top eight genome-wide association studies confirmed loci with type 2 diabetes in North Indian population.30 However, in this study, we failed to observe association for FTO in the same study population. Phenotypic differences between Indians and Europeans might be one of the possible reasons for observed difference in association in two populations. Indians have higher body fat and lower muscle mass for a given BMI compared with Europeans, have different fat distribution (high truncal, subcutaneous and intra-abdominal fat) and are centrally obese.31 Studies in different ethnic groups of Indian population have provided contradictory reports for association of FTO variants with type 2 diabetes and obesity. A study that recruited subjects from Western and Southern India (Pune–Mysore study),17 representing Indo-European and Dravidian population, found rs9939609 to be associated with type 2 diabetes independent of BMI (MAF in controls 0.30). Similar results were obtained in another study of Indo-European population of North Indian Sikhs18 (MAF in controls 0.31). However, these studies failed to replicate association of rs9939609 with obesity in the control subjects. A recent study in Dravidian population from India19 demonstrated FTO variant association with type 2 diabetes independent of BMI and also with obesity in the control subjects (MAF in controls 0.11). However, they did not study rs9939609; instead, they studied rs8050136 (a proxy SNP for rs9939609). We investigated both these variants (rs9939609 and rs8050136) in this study and observed them to be in strong linkage disequilibrium (r2>0.99, D′>0.99) with each other (MAF in controls 0.34). By far this is the largest study on any Indian population, investigating the role of FTO variants with obesity and type 2 diabetes. Despite being sufficiently powered, our study did not detect association of common variants of FTO with type 2 diabetes (dependent or independent of BMI). We also meta-analyzed the data of all the four studies on the Indian population (including this study) and found variants of FTO to be associated with type 2 diabetes, but with a lower effect size (Figure 1). This may be indicative of a lower effect size of FTO variants in this study population.

The contradictory findings of the studies in diverse ethnic groups from different geographical regions of India might indicate ethnic-specific effect of FTO variants in Indians. However, study design, study population and sample size can significantly affect the outcome of association studies. There was considerable heterogeneity in subject recruitment criteria, statistical analyses, sample sizes and most importantly sample collection areas among these studies on Indian population. The study in North Indian Sikhs involved participation of only 920 subjects. While the Pune study involved participation of 3775 subjects (inclusive of individuals of two different ethnicities: Indo-European and Dravidians) and for reasons unknown analysis was not adjusted for sex and age. Diet and physical activity are shown to have considerable effect in modulating the susceptibility to obesity by FTO variants.32, 33 There is tremendous diversity in diet and physical activity not only among Europeans and Indians, but also within different ethnic groups of Indian population. Hence, differences in genetic susceptibility of obesity and type 2 diabetes observed in different population might be due to modulation by population-specific environmental factors and lifestyle. However, absence of data on diet and physical activity in this study population does not allow affirmative conclusion in this regard.

Although we observed differences in various anthropometric and biochemical measures of subjects in stage 1 and stage 2, there is no genetic heterogeneity between the subjects of the two stages (unpublished data). We performed a multidimensional scaling (MDS) analysis based on 608 unlinked markers genotyped in the initial study population of the present study that clearly demonstrates that our study population belongs to one cluster (Supplementary Figure 3). In stage 1, all type 2 diabetes patients were recruited from the Clinic, and thus having longer duration of diabetes, under hypoglycemic and lipid-lowering medication that can influence quantitative measurements. Although stage 2 comprised of a combination of patients from the Clinic and ‘Diabetes Awareness Camps’ that included a proportion of patients who were newly detected with type 2 diabetes and without any kind of medication, this might result in the observed differences in glycemic measures, lipid profile and creatinine levels of patients in two stages. ‘Diabetes Awareness Camps’ in stage 1 were held in areas of low socio-economic localities, whereas stage 2 predominantly comprised of subjects recruited from camps organized in more affluent localities. We believe that difference in socio-economic localities of the subjects could explain the differences in the anthropometric and biochemical measures.

In conclusion, common variants of FTO were associated with obesity, but not with type 2 diabetes in our study population. However, it is also possible that the studied variants had small effects on type 2 diabetes, and that variants of FTO other than the ones studied are involved in conferring genetic susceptibility of type 2 diabetes and modulations of their effect through environment (diet and physical activity). Hence, further large-scale studies on Indians of different ethnicities with data for diet and physical activity and dense SNP coverage (end-to-end tag SNPs) spanning the entire gene is required to elucidate the role of FTO on the predisposition to type 2 diabetes in Indians.