The Meta-Analysis of Glucose and Insulin related traits Consortium (MAGIC) recently identified 16 loci robustly associated with fasting glucose, some of which were also associated with type 2 diabetes. The purpose of our study was to explore the role of these variants in South Asian populations of Punjabi ancestry, originating predominantly from the District of Mirpur, Pakistan.
Sixteen single nucleotide polymorphisms (SNPs) were genotyped in 1678 subjects with type 2 diabetes and 1584 normoglycaemic controls from two Punjabi populations; one resident in the UK and one indigenous to the District of Mirpur. In the normoglycaemic controls investigated for fasting glucose associations, 12 of 16 SNPs displayed β values with the same direction of effect as that seen in European studies, although only the SLC30A8 rs11558471 SNP was nominally associated with fasting glucose (β = 0.063 [95% CI: 0.013, 0.113] p = 0.015). Of interest, the MTNR1B rs10830963 SNP displayed a negative β value for fasting glucose in our study; this effect size was significantly lower than that seen in Europeans (p = 1.29×10−4). In addition to previously reported type 2 diabetes risk variants in TCF7L2 and SLC30A8, SNPs in ADCY5 (rs11708067) and GLIS3 (rs7034200) displayed evidence for association with type 2 diabetes, with odds ratios of 1.23 (95% CI: 1.09, 1.39; p = 9.1×10−4) and 1.16 (95% CI: 1.05, 1.29; p = 3.49×10−3) respectively.
Although only the SLC30A8 rs11558471 SNP was nominally associated with fasting glucose in our study, the finding that 12 out of 16 SNPs displayed a direction of effect consistent with European studies suggests that a number of these variants may contribute to fasting glucose variation in individuals of South Asian ancestry. We also provide evidence for the first time in South Asians that alleles of SNPs in GLIS3 and ADCY5 may confer risk of type 2 diabetes.
Citation: Rees SD, Hydrie MZI, O'Hare JP, Kumar S, Shera AS, Basit A, et al. (2011) Effects of 16 Genetic Variants on Fasting Glucose and Type 2 Diabetes in South Asians: ADCY5 and GLIS3 Variants May Predispose to Type 2 Diabetes. PLoS ONE 6(9): e24710. https://doi.org/10.1371/journal.pone.0024710
Editor: Nicholas John Timpson, University of Bristol, United Kingdom
Received: May 11, 2011; Accepted: August 17, 2011; Published: September 20, 2011
Copyright: © 2011 Rees et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Funding for the collection of DNA and clinical data from the Pakistan-resident population (DGP) was provided by Diabetes UK (project number 07/0003512; www.diabetes.org.uk). Funding for genetic analyses was provided by Eli Lilly. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have the following competing interests: This study was partly funded by Eli Lilly. Prof. Sudhesh Kumar has received lecture fees and research funding from Eli Lilly. Prof. Anthony Barnett has received honoraria from Eli Lilly for lectures and advisory work, and has received research funding from Eli Lilly. There are no patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLoS ONE policies on sharing data and materials.
Type 2 diabetes, a disease that is 4- to 6-fold more common in South Asian individuals than Europeans, is characterised by impaired glucose homeostasis resulting from a combination of beta cell dysfunction and insulin resistance. This inability to adequately regulate blood glucose levels is also linked to both micro- and macro-vascular complications. Recently, the Meta-Analysis of Glucose and Insulin related traits Consortium (MAGIC) identified 16 genetic variants robustly associated with fasting glucose in non-diabetic populations of European origin . Although a number of these single nucleotide polymorphisms (SNPs) were also associated with type 2 diabetes, several were not, suggesting that some variants may be associated with a ‘physiological’ variation in glucose levels without influencing ‘pathological’ variation and type 2 diabetes risk . As with other genetic associations, replication of these findings in datasets of different ethnic origin is an important step in helping to fine-map the aetiological variants at these loci. Our aim was to investigate the effect of these 16 SNPs on fasting glucose levels and type 2 diabetes in South Asian populations of Punjabi ancestry.
Materials and Methods
Informed written consent was obtained from all study participants and the studies were approved by the Birmingham East, North and Solihull Research Ethics Committee (UKADS participants) and the Baqai Institute of Diabetology and Endocrinology Institutional Review Board (DGP participants). UKADS registered clinical trial number; ISRCTN38297969.
Participants in the investigation (Table 1) were from two studies of South Asian individuals; the United Kingdom Asian Diabetes Study (UKADS; 857 participants with type 2 diabetes, 417 without) ,  and the Diabetes Genetics in Pakistan study (DGP; 821 participants with type 2 diabetes, 1167 without) . Participants were all of Punjabi ancestry, confirmed over three generations, and originated predominantly from the District of Mirpur, Pakistan. Diagnosis of type 2 diabetes was established using World Health Organisation criteria . Normoglycaemic control subjects were recruited from the same geographical region as subjects with type 2 diabetes. In the UKADS control group normal glucose tolerance was defined in the majority of participants as random blood glucose <7 mmol/l. In a small number of UKADS controls (n = 22), normoglycaemia was defined as fasting plasma glucose <6.1 mmol/l and 2 hr plasma glucose <7.8 mmol/l on a 75 g OGTT; due to small sample size these data were not included in any fasting glucose analyses. Normal glucose tolerance in the DGP control group was defined as fasting whole blood glucose ≤5.6 mmol/l. Genomic DNA was extracted either from venous blood using the Nucleon® protocol (Nucleon Biosciences, Coatbridge, UK) (UKADS) or from saliva using the Oragene® DNA sample collection kit and extraction protocol (DNA Genotek Inc., Ontario, Canada) (DGP). The clinical details of individuals from the two study populations are shown in Table 1.
All subjects (UKADS, n = 1274; DGP, n = 1988; total, n = 3262) were genotyped for the 16 SNPs using the KASPar method (KBioscience, Hoddesdon, UK). Genotyping success rates were >96% for each SNP. Approximately 10% of samples were genotyped as blind duplicates resulting in an error rate of <1% for each SNP. Genotype counts in the two study populations are shown in Table S1.
Deviation from Hardy-Weinberg equilibrium (HWE) in the non-diabetic groups was tested using an exact test implemented in Haploview . None of the SNPs deviated significantly from HWE after adjustment for multiple testing. Linear regression was used to test for association between SNPs and fasting glucose levels in the DGP control group only, adjusting for age, sex and BMI. Four DGP control participants had missing BMI data; the sample size for the fasting glucose analysis was therefore 1163. Fasting plasma glucose estimates were calculated from fasting whole blood measurements using a conversion factor of 1.15. Association between SNPs and type 2 diabetes was tested in the UKADS and DGP study populations separately, using logistic regression, adjusting for age and sex. Inverse variance weighted meta-analysis, implemented in Metan, was used to combine ORs from the UKADS and DGP study populations. Only 256 UKADS controls had available BMI data. As BMI made little difference to the type 2 diabetes association (combined UKADS/DGP OR difference ≤0.015 for all SNPs), to maintain sample size and power the main results are not adjusted for BMI. Heterogeneity of ORs (between UKADS and DGP study populations, and between the combined UKADS/DGP cohort and previously published studies) was assessed using Cochran's Q statistics. None of the studied SNPs displayed significant heterogeneity of ORs between the UKADS and DGP study populations after correcting for multiple testing (Table 2). For all single-locus analyses in this study, two levels of statistical significance are referred to; nominal significance (p<0.05) and study-wide significance (p<3.13×10−3, corrected for 16 independent tests). Genetic risk scores (GRSs) were constructed to investigate the additive effect of multiple SNPs on fasting glucose levels and type 2 diabetes. For all GRS analyses, only those individuals successfully genotyped at ≥12 SNPs (nmax = 3231) were included. Firstly an allele count GRS (acGRS) was constructed. For each individual, the average number of risk alleles per successfully genotyped SNP (total number of observed risk alleles/number of successfully genotyped SNPs) was multiplied by the total number of SNPs included in the GRS. This produced a GRS approximating a simple risk allele count where all SNPs were successfully genotyped. Two weighted GRSs (wGRS) were also produced, one for the fasting glucose analyses and one for the type 2 diabetes analyses. Risk alleles of each SNP were weighted by a SNP-specific β-value, obtained from the MAGIC study , and the weighted allele count was divided by the mean MAGIC-derived β-value for the successfully genotyped SNPs. GRSs for the type 2 diabetes analyses were also estimated without the TCF7L2 and SLC30A8 variants, as these SNPs have previously displayed association with the disease in the UKADS and DGP study populations , . All statistical analyses were performed using Stata IC Version 10.1 (Stata Corporation, College Station, TX, USA). Power was calculated using Genetic Power Calculator (http://ibgwww.colorado.edu/~pshaun/gpc/)  and Quanto version 1.2 (http://hydra.usc.edu/gxe) , assuming a significance level (α) of 0.05.
Associations with fasting glucose
Of the 16 SNPs studied, 12 displayed β-values with the same direction of effect as that seen in Europeans, and the SLC30A8 rs11558471 variant displayed a nominally significant association with fasting glucose levels (β = 0.063 [95% CI: 0.013, 0.113] p = 0.015) (Figure 1). No evidence for heterogeneity of effect sizes between our study and a previous study of Europeans was apparent for 15 of the studied variants. In contrast, the MTNR1B rs10830963 effect size was markedly lower in the current study compared with the Europeans from the MAGIC study  and this heterogeneity achieved study-wide significance (p = 1.29×10−4). The acGRS showed a stronger association with fasting glucose than did the wGRS (β = 0.013 [95% CI: 0.000, 0.025] p = 0.046 and β = 0.007 [95% CI: −0.004, 0.018] p = 0.204 respectively) (Figure 1).
South Asians from the current study (filled diamonds, n = 1163) and Europeans from the MAGIC  study (unfilled diamonds, n≤76,558). acGRS = allele count GRS, wGRS = weighted GRS. ES = effect size, the per-risk allele change in fasting glucose (mmol/l). The GCK rs1799884 SNP was used as a proxy for the rs4607517 variant reported in Dupuis et al  (r2 = 1.0 in CEU HapMap samples).
Associations with type 2 diabetes
None of the 16 studied variants displayed strong evidence for heterogeneity of effect on type 2 diabetes risk between the current study and the MAGIC study. In addition to the TCF7L2 and SLC30A8 variants, which have previously demonstrated association with type 2 diabetes in the UKADS/DGP study populations , , alleles of the ADCY5 rs11708067 (OR = 1.23 [95% CI: 1.09, 1.39] p = 9.10×10−4) and GLIS3 rs7034200 (OR = 1.16 [95% CI: 1.05, 1.29] p = 3.49×10−3) SNPs conferred risk of the disease in this study (Figure 2). The strength of these associations reached study-wide significance for the ADCY5 variant; the GLIS3 SNP failed to reach this threshold by a narrow margin. Both risk score measures were associated with type 2 diabetes, with the wGRS displaying a stronger association (OR = 1.04 [95% CI: 1.02, 1.06] p = 1.00×10−5) than the acGRS (OR = 1.05 [95% CI: 1.02, 1.08] p = 0.001) (Figure 2). The strength of association of both risk scores was greatly attenuated by the removal of the previously associated TCF7L2 and SLC30A8 variants (wGRS; OR = 1.01 [95% CI: 0.990, 1.04] p = 0.261, acGRS; OR = 1.03 [95% CI: 1.00, 1.06] p = 0.071).
South Asians from the current study (filled diamonds, n = 3262) and Europeans from the MAGIC  study (unfilled diamonds, n≤127,667). acGRS = allele count GRS, wGRS = weighted GRS. ES = effect size, the per-risk allele odds ratio. The GCK rs1799884 SNP was used as a proxy for the rs4607517 variant reported in Dupuis et al  (r2 = 1.0 in CEU HapMap samples).
In this study we investigated the effects of 16 SNPs on fasting glucose levels and type 2 diabetes risk in two South Asian populations of Punjabi ancestry. Only the SLC30A8 rs11558471 variant was nominally associated with fasting glucose levels. Twelve of the 16 SNPs displayed positive β-values, however, suggesting that a number of these variants may be true determinants of fasting glucose levels in our study populations (Figure 1), even though their effects were too small to be accurately detected in our modestly-sized cohort. Comparing the effect sizes observed in this study to those reported in Europeans highlights some potential differences. Three of the six variants most strongly associated with fasting glucose in the MAGIC study (MTNR1B rs10830963, DGKB/TMEM195 rs2191349 and ADCY5 rs11708067) have negative β-values in the current study (Figure 1), and this disparity reached statistical significance for the MTNR1B variant. The observed differences in effect size are probably the reason that the acGRS was nominally associated with fasting glucose levels, but the wGRS (weighted using European-derived β-values) was not.
It is of interest to note that variants within or near GCK, GCKR, G6PC2 and MTNR1B have been shown to be associated with fasting glucose levels in Indian Asians, with similar effect sizes to those seen in Europeans . The GCK and GCKR SNPs studied in Indian Asians (rs4607519 and rs1260326 respectively) are in strong linkage disequilibrium (LD) with the SNPs genotyped in this study (r2≥0.96 in GIH HapMap  samples; Gujerati Indians in Houston, Texas) and the G6PC2 rs560887 SNP was genotyped in both studies. The fasting glucose association results we present for GCK and GCKR are very similar to those demonstrated in both the Europeans from the MAGIC  study (Figure 1) and Indian Asians . Although our estimate of effect size for the G6PC2 rs560887 variant is low (Figure 1), it is not statistically different from that seen in Europeans. In contrast, the MTNR1B rs10830963 variant displayed an effect size lower than that seen in Europeans, at study-wide significance. The MTNR1B variant reported as being associated with fasting glucose in Indian Asians  (rs2166706) is only in moderate LD (r2 = 0.45) with the variant (rs10830963) genotyped in our South Asian cohort and reported as the sentinel MTNR1B SNP in the MAGIC study ; if different LD patterns result in rs2166706 being in tighter LD with the aetiological variant than rs10830963 in South Asians, this may explain some of the observed discrepancy in effect size. It must be noted, however, that this LD estimate is taken from CEU (Utah residents with Northern and Western European ancestry from the CEPH collection) HapMap samples as rs10830963 has not been genotyped in the GIH HapMap samples. In addition, imputation analyses in the Indian Asian study estimated that the strongest association signal for fasting glucose in this population was in fact the rs10830963 SNP , although this imputation was not able to utilise South Asian-specific LD patterns. It is interesting to note that a recent study of Indian Sikhs demonstrated that a low frequency variant (rs1374645) was associated with glucose levels, whereas rs10830963 was not . Further studies of MTNR1B SNPs and their association with glucose levels in South Asians may be useful in fine-mapping the aetiological variant.
Variants in TCF7L2 and SLC30A8 have previously been associated with type 2 diabetes in our Punjabi populations , . In addition to these variants, SNPs in ADCY5 and GLIS3 were associated with the disease in the current study. To our knowledge, this is the first time that either of these SNPs has been implicated in type 2 diabetes development in a South Asian population. The ADCY5 gene encodes adenylate cyclase 5, an enzyme that catalyses the generation of cAMP, a second messenger vital in a number of biological processes. It has been demonstrated in a large meta-analysis of Europeans that the rs11708067 SNP within ADCY5 is strongly associated (p≤3.6×10−8) with type 2 diabetes, fasting glucose levels and HOMA-B (a measure of β-cell function) but is not associated with HOMA-IR (a measure of insulin resistance) , . This suggests that variants within this gene may exert their effect on disease risk through β-cell dysfunction and insulin secretion. In addition, a SNP within the ADCY5 gene (rs9883204), in LD with the variant investigated in this study, is associated with foetal growth and birth weight . Insulin is an important growth factor in utero, potentially providing a common mechanism linking reduced foetal growth with increased risk of type 2 diabetes. The GLIS3 gene encodes the transcription factor GLIS family zinc finger 3 isoform, a protein that regulates target gene transcription and has been shown to play a key role in β-cell generation in mice , . Rare functional mutations within the GLIS3 gene lead to a syndrome of neonatal diabetes and congenital hyperthyroidism , and the rs7020673 SNP within the gene is robustly associated with type 1 diabetes . The GLIS3 rs7034200 SNP only displayed a weak association with type 2 diabetes in the MAGIC study , although there is evidence that this variant confers risk of the disease in a Chinese population . As with the fasting glucose analyses, some potential disparity was apparent between our South Asian type 2 diabetes association results and those reported by the MAGIC study  (Figure 2), although the relatively small size of our cohort makes any differences difficult to quantify statistically. It is interesting to note that, excluding the TCF7L2 SNP, three of the four variants most strongly associated with type 2 diabetes in the MAGIC study  (MTNR1B rs10830963, PROX1 rs340874 and GCKR rs780094) have odds ratios of less than one in our South Asian populations (Figure 2).
The lack of association of many of the studied variants with fasting glucose and type 2 diabetes in our study cohort could be due to small sample size and low statistical power (Table 3). For the analysis of type 2 diabetes, this study had >80% power to detect the effect of just the TCF7L2 SNP. The study was underpowered to detect the effect of any SNP on fasting glucose levels, assuming similar effect sizes to those seen in European populations (Table 3). Although the statistical evidence for heterogeneity is weak, it is also possible that the studied variants have different effect sizes in our Punjabi populations compared with Europeans. This may be due to differences in LD patterns between the two ethnic groups, as disease-associated SNPs derived from GWA studies are typically not aetiological variants. Previous findings that the MTNR1B and G6PC2 variants genotyped in this study are associated, either directly or through imputation, with glucose levels in Indian Asians , however, make it unclear whether any potential differences in LD patterns are likely to contribute to our observed lack of association at these loci.
In conclusion, our study of Punjabi populations demonstrated that 12 of 16 variants displayed β-values for fasting glucose with the same direction of effect as that seen in Europeans. In addition, we provide evidence that alleles of SNPs in ADCY5 and GLIS3 may confer risk of type 2 diabetes, the first time that this has been reported in South Asian populations.
The authors would like to thank all the diabetic patients and control subjects for agreeing to participate in this study. We are grateful to Ms. Shanaz Mughal, Mrs. Kam Johal, Dr. Anthony Dixon, Dr. Sri Bellary and Mrs. Tahera Mehrali within the UK, and Dr. Asher Fawwad, Dr. Waheed Iqbal and Dr. Abdul Razzaq in Pakistan for recruiting the study subjects and coordinating data collection. The United Kingdom Asian Diabetes Study was conceived and managed by Professor Anthony Barnett, Professor Sudhesh Kumar and Dr. Paul O'Hare.
Conceived and designed the experiments: MAK. Performed the experiments: SDR. Analyzed the data: SDR. Wrote the paper: SDR MAK. Contributed to discussion and reviewed/edited the manuscript: MZIH JPO SK ASS AB AHB.
- 1. Dupuis J, Langenberg C, Prokopenko I, Saxena R, Soranzo N, et al. (2010) New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat Genet 42: 105–116. ng.520 [pii];10.1038/ng.520 [doi].
- 2. Bellary S, O'Hare JP, Raymond NT, Gumber A, Mughal S, et al. (2008) Enhanced diabetes care to patients of south Asian ethnic origin (the United Kingdom Asian Diabetes Study): a cluster randomised controlled trial. Lancet 371: 1769–1776.
- 3. Rees SD, Hydrie MZ, Shera AS, Kumar S, O'Hare JP, et al. (2011) Replication of 13 genome-wide association (GWA)-validated risk variants for type 2 diabetes in Pakistani populations. Diabetologia. 10.1007/s00125-011-2063-2 [doi].
- 4. Alberti KG, Zimmet PZ (1998) Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med 15: 539–553.
- 5. Barrett JC, Fry B, Maller J, Daly MJ (2005) Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 21: 263–265.
- 6. Rees SD, Bellary S, Britten AC, O'Hare JP, Kumar S, et al. (2008) Common variants of the TCF7L2 gene are associated with increased risk of type 2 diabetes mellitus in a UK-resident South Asian population. BMC Med Genet 9: 8.
- 7. Purcell S, Cherny SS, Sham PC (2003) Genetic Power Calculator: design of linkage and association genetic mapping studies of complex traits. Bioinformatics 19: 149–150.
- 8. Gauderman W, Morrison J (2006) QUANTO 1.1: A computer program for power and sample size calculations for genetic-epidemiology studies, http://hydra.usc.edu/gxe.
- 9. Chambers JC, Zhang W, Zabaneh D, Sehmi J, Jain P, et al. (2009) Common genetic variation near melatonin receptor MTNR1B contributes to raised plasma glucose and increased risk of type 2 diabetes among Indian Asians and European Caucasians. Diabetes 58: 2703–2708. db08-1805 [pii];10.2337/db08-1805 [doi].
- 10. The International HapMap Consortium (2003) The International HapMap Project. Nature 426: 789–796.
- 11. Been LF, Hatfield JL, Shankar A, Aston CE, Ralhan S, et al. (2011) A low frequency variant within the GWAS locus of MTNR1B affects fasting glucose concentrations: Genetic risk is modulated by obesity. Nutr Metab Cardiovasc Dis. S0939–4753(11)00022-6 [pii];10.1016/j.numecd.2011.01.006 [doi].
- 12. Voight BF, Scott LJ, Steinthorsdottir V, Morris AP, Dina C, et al. (2010) Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet 42: 579–589.
- 13. Freathy RM, Mook-Kanamori DO, Sovio U, Prokopenko I, Timpson NJ, et al. (2010) Variants in ADCY5 and near CCNL1 are associated with fetal growth and birth weight. Nat Genet 42: 430–435. ng.567 [pii];10.1038/ng.567 [doi].
- 14. Kang HS, ZeRuth G, Lichti-Kaiser K, Vasanth S, Yin Z, et al. (2010) Gli-similar (Glis) Kruppel-like zinc finger proteins: insights into their physiological functions and critical roles in neonatal diabetes and cystic renal disease. Histol Histopathol 25: 1481–1496.
- 15. Kang HS, Kim YS, ZeRuth G, Beak JY, Gerrish K, et al. (2009) Transcription factor Glis3, a novel critical player in the regulation of pancreatic beta-cell development and insulin gene expression. Mol Cell Biol 29: 6366–6379. MCB.01259-09 [pii];10.1128/MCB.01259-09 [doi].
- 16. Senee V, Chelala C, Duchatelet S, Feng D, Blanc H, et al. (2006) Mutations in GLIS3 are responsible for a rare syndrome with neonatal diabetes mellitus and congenital hypothyroidism. Nat Genet 38: 682–687. ng1802 [pii];10.1038/ng1802 [doi].
- 17. Barrett JC, Clayton DG, Concannon P, Akolkar B, Cooper JD, et al. (2009) Genome-wide association study and meta-analysis find that over 40 loci affect risk of type 1 diabetes. Nat Genet 41: 703–707. ng.381 [pii];10.1038/ng.381 [doi].
- 18. Hu C, Zhang R, Wang C, Wang J, Ma X, et al. (2010) Variants from GIPR, TCF7L2, DGKB, MADD, CRY2, GLIS3, PROX1, SLC30A8 and IGF1 are associated with glucose metabolism in the Chinese. PLoS One 5: e15542. 10.1371/journal.pone.0015542 [doi].