TCF7L2 is reproducibly associated with type 2 diabetes in various ethnic groups: a global meta-analysis
- 743 Downloads
TCF7L2 variants have been consistently associated with type 2 diabetes (T2D) in populations of different ethnic descent. Among them, the rs7903146 T allele is probably the best proxy to evaluate the effect of this gene on T2D risk in additional ethnic groups. In the present study, we investigated the association between the TCF7L2 rs7903146 polymorphism and T2D in Moroccans (406 normoglycemic individuals and 504 T2D subjects) and in white Austrians (1,075 normoglycemic individuals and 486 T2D subjects). Then, we systematically reviewed the association of this single nucleotide polymorphism (SNP) with T2D risk in a meta-analysis, combining our data with data from previous studies. The allelic odds ratios (ORs) for T2D were 1.56 [1.29–1.89] (p = 2.9 × 10−6) and 1.52 [1.29–1.78] (p = 3.0 × 10−7) in Moroccans and Austrians, respectively. No heterogeneity was found between these two different populations by Woolf test (χ2 = 0.04, df = 1, p = 0.84). We found 28 original published association studies dealing with the TCF7L2 rs7903146 polymorphism in T2D. A meta-analysis was then performed on 29,195 control subjects and 17,202 cases. No heterogeneity in genotypic distribution was found (Woolf test: χ2 = 31.5, df = 26, p = 0.21; Higgins statistic: I2 = 14.1%). A Mantel–Haenszel procedure was then performed to provide a pooled odds ratio (OR) of 1.46 [1.42–1.51] (p = 5.4 × 10−140). No publication bias was detected, using the conservative Egger’s regression asymmetry test (t = −1.6, df = 25, p = 0.11). Compared to any other gene variants previously confirmed by meta-analysis, TCF7L2 can be distinguished by its tremendous reproducibility of association with T2D and its OR twice as high. In the near future, large-scale genome-wide association studies will fully extend the genome coverage, potentially delivering other common diabetes-susceptibility genes like TCF7L2.
TCF7L2 rs7903146 T allele has been consistently associated with type 2 diabetes (T2D) in individuals of European [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14], Asian [7, 15, 16, 17], and African descent [7, 18], so far. Despite comprehensive genotyping efforts across this gene locus [1, 2, 4, 18], this single nucleotide polymorphism (SNP) always showed the strongest association with T2D. Furthermore, Helgason et al.  reported that the rs7903146 T allele is probably the ancestral allele and concluded that if this allele is not itself the causative variant, then the unidentified functional variant it tags is likely to lie outside the screened locus. Thus, genotyping this polymorphism is probably the best proxy to evaluate the effect of this gene on T2D risk in additional ethnic groups, and combining data related to this SNP is likely to be sufficient for a meta-analysis of the global contribution of the gene on T2D. In the present study, we examined the association between the TCF7L2 rs7903146 polymorphism and T2D in two additional populations of different ethnic descents, Central Europeans and, for the first time, North Africans. Then, we systematically reviewed the association of this polymorphism with T2D risk in a meta-analysis combining our data with those from previous studies.
Clinical characteristics (means ± SD) of the T2D case-control studies
Sex ratio (male/female)
Age at examination (years)
55 ± 12
58 ± 11
51.5 ± 6.0
56.5 ± 9.6
Age at diagnosis (years)
51 ± 12
49.3 ± 9.5
27.2 ± 5.3
28.0 ± 4.7
26.4 ± 4.0
30.7 ± 6.3
Genotypic distributions of rs7903146 among Moroccan and Austrian populations
Allelic OR [95% CI]
2.9 × 10−6
3.0 × 10−7
We found 28 original published association studies dealing with the TCF7L2 rs7903146 polymorphism in T2D. Unfortunately, the data available on Mexican-American population were not detailed enough to include them in the present analysis . From these studies, a total of 27,705 control individuals and 16,200 subjects with T2D have been analyzed for disease association between T2D and the T allele variant. If we include our own study cohorts (Austrians and Moroccans),the overall total would add up to 29,195 control subjects and 17,202 cases. Considering that all publications were consistent with a multiplicative model of inheritance, the relative risk for T2D was only estimated by allelic odds ratios. No heterogeneity in genotypic distribution was found (Woolf test: χ2 = 31.5, df = 26, p = 0.21; Higgins statistic: I2 = 14.1%).
We positively replicated the association of TCF7L2 variation and T2D in two populations of different ethnic descent, Central Europeans and, for the first time, North Africans. The most striking result is the very stable relative risk (around 50%) conferred by the rs7903146 T allele in these two geographically, ethnically, and environmentally diverse populations. The meta-analysis of 27 different studies confirms this finding with a resulting global OR of 1.46 [1.42–1.51], suggesting that in any tested human population the effect of TCF7L2 is very similar. The absence of heterogeneity between studies is also indicative of a universal contribution of this gene to T2D, the population-attributable risk being only driven by the prevalence of the T allele in a specific ethnic group. This situation is unique as previous candidate genes for T2D have always shown some degree of discrepancy between populations [20, 21, 22, 23, 24, 25, 26, 27, 28, 29]. TCF7L2 has moved rapidly from a novel positional candidate gene to a reference gene for T2D susceptibility [1, 30]. In most ethnic groups, except for Eastern Asians [16, 17], a highly frequent T allele (ranging from 18–35% for controls to 22–45% for cases) offers the possibility to get enough statistical power for association studies . Consequently, almost all published case-control studies were able to detect an association with T2D, except for those with few participants or without clear ethnic belonging [4, 7]. However, even among Caucasian populations there are substantial differences in T allele frequency, e.g., 17% difference between Finns  and Moroccans (this study).
So far, few other SNPs have been reproducibly associated with T2D [20, 21, 22, 23, 24, 25, 26, 27, 28, 29]. The comparison between meta-analyses (shown in Fig. 2) clearly illustrates that the magnitude of the TCF7L2 effect is much higher than any other confirmed T2D candidate. The individual effects of the other variants are modest, ranging from 10 to 30%. Interestingly, no major interactions with the T allele have been found to strongly modulate T2D susceptibility, even if body mass index [5, 7, 9, 16, 18], gender , drugs , or lifestyle interventions  may modulate TCF7L2 effects. No functional significance has been attributed to the TCF7L2 T allele so far, which is also different from most T2D SNPs listed in Fig. 2. In rodents, the Wnt/β-catenin signaling pathway is important for the development of pancreas [31, 32], but the pathophysiology of the TCF7L2-associated T2D remains to be clarified in humans.
The binding of transcription factors and alternative splicing events should be studied in the intronic region where the T allele is located. Recent theoretical studies have emphasized that as few as 20 susceptibility variants on the scale of those in TCF7L2 may suffice to explain as much as 50% of the burden of the disease . In the near future, large-scale genome-wide association data will fully extend the genome coverage, potentially delivering other common diabetes-susceptibility genes like TCF7L2 .
Materials and methods
Two populations with different ethnic backgrounds were analyzed: white Austrians from central Europe  and Moroccans from North Africa.
Moroccan subjects were recruited by the Faculty of Medicine (Fes) and were subject to a standardized clinical examination at the Hassan II Hospital. Inclusion criteria for cases were: (1) T2D according to 1997 American Diabetes Association (ADA) criteria; (2) family history of diabetes in first degree relatives; (3) BMI <30 kg/m2. Inclusion criteria for controls were: (1) age at examination over 45 yr.; (2) normal fasting glucose according to 1997 ADA criteria; (3) BMI <27 kg/m2.
Austrian subjects with T2D were recruited from diabetes outpatient clinics of the Landeskliniken Salzburg and the Hallein Hospital. Patients who were <73 years of age and <63 years of age at diagnosis were included. Participants of the Salzburg Atherosclerosis Prevention Program in Subjects at High Individual Risk (Saphir) who were not using hypoglycemic medications and had fasting blood glucose levels <110 mg/dl served as control subjects. Study populations comprised only white Europeans, mainly of Bavarian or Austrian-German descent, living in the same geographic region.
The main clinical characteristics were reported in Table 1. The genetic study was approved by local Ethical Committees and informed consent was obtained from all participants.Participants were considered as normoglycemic controls when their fasting glucose concentration was lower than 6.1 mmol/l.
High-throughput genotyping for the rs7903146 variant was performed using the TaqMan® SNP Genotyping Assays (Applied Biosystems, Foster City, CA, USA). The polymerase chain reaction (PCR) primers and TaqMan probes were designed by Primer Express and optimized according to the manufacturer’s protocol. There was a 98% genotyping success rate and the genotyping error rate was assessed by sequencing 384 control and 384 T2D individuals and by re-genotyping a random 10% sample. No difference was found with the first genotyping results, thus the genotyping error rate was 0%.
Tests for deviation from Hardy–Weinberg equilibrium (HWE) and for association were performed with the De Finetti program (http://ihg.gsf.de/cgi-bin/hw/hwa1.pl). All P values were two-tailed. Meta-analysis was performed using the packages “rmeta” and “meta” of the R-Project (http://www.r-project.org).The Woolf test was first applied to test the genotypic heterogeneity between studied groups , then the Mantel–Haenszel procedure was performed to provide a pooled odds ratio with a 95% confidence interval. We performed the Higgins statistic (I2) to quantify the amount of between-study variability in effect attributable to true heterogeneity rather than chance . We also used the Egger’s regression method to test for publication bias .
To assess whether the model of inheritance was multiplicative or departing from linearity, we applied a logistic regression test with two variables corresponding to genotypes v1 (coded 0, 1, 2), reflecting a linear increase in risk, and v2 (coded 0, 1, 0), reflecting a departure from linearity. Different meta-analyses of T2D association studies are discussed in Fig. 2: PPARG [20, 21], KCNJ11[22, 23], CAPN10 [24, 25], HNF1A , ENPP1 , and IL6 [28, 29]. We only reported studies with allelic odds ratios, based on diverse ethnic groups. SPSS 14.1 software (SPSS, Chicago, IL, USA) was used for general statistical analyses.
This work was partly supported by the French Governmental “Agence Nationale de la Recherche”, by the French-Moroccan convention “CNRST-CNRS”, and the charities: “Association Française des Diabétiques” and “Programme national de recherche sur le diabète” and “Association des diabétiques de la Wilaya de Fès”. We thank Marianne Deweider and Frederic Allegaert for the DNA bank management and Stefan Gaget for his help on phenotype databases. We are indebted to all subjects who participated to this study.
- 12.Melzer D, Murray A, Hurst AJ et al (2006) Effects of the diabetes linked TCF7L2 polymorphism in a representative older population. BMC Med 4:34 DOI 10.1186/1741-7015-4-34
- 17.Hayashi T, Iwamoto Y, Kaku K et al (2007) Replication study for the association of TCF7L2 with susceptibility to type 2 diabetes in a Japanese population. Diabetologia 50:980–984 (Electronic publication ahead of print)Google Scholar
- 22.Gloyn AL, Weedon MN, Owen KR et al (2003) Large-scale association studies of variants in genes encoding the pancreatic beta-cell KATP channel subunits Kir6.2 (KCNJ11) and SUR1 (ABCC8) confirm that the KCNJ11 E23K variant is associated with type 2 diabetes. Diabetes 52:568–572PubMedCrossRefGoogle Scholar
- 35.Agresti A (2002) Categorical data analysis. Wiley, New YorkGoogle Scholar