Association Study of the Genetic Polymorphisms of the Transcription Factor 7-Like 2 (TCF7L2) Gene and Type 2 Diabetes in the Chinese Population

The studied population comprised 760 normoglycemic control subjects and 760 type 2 diabetic patients from a Han Chinese population in Taiwan. The glucose-tolerant (determined by 75-g oral glucose tolerance tests) control subjects were recruited from routine health examinations, and type 2 diabetic patients were recruited from the metabolism clinics of the National Taiwan University Hospital. Type 2 diabetic patients were diagnosed based on the criteria of the American Diabetes Association (28) with fasting plasma glucose ≥126 mg/dl or 2-h plasma glucose ≥200 mg/dl during an oral glucose tolerance test. Patients with ages of onset <35 years were excluded. Total cholesterol and triglycerides were measured in fasting samples using an autoanalyzer (Hitachi 7250 special; Hitachi, Tokyo, Japan). A1C was measured by high-performance liquid chromatography (Primus, Wichita, KS). Written informed consent was obtained from every participating subject, and the study was approved by the institutional review board of the National Taiwan University Hospital.

From HapMap CHB databank (public data release 21 a/phase II, Jan. 2007), 112 SNPs that had minor allele frequencies exceeding 1% without violation of Hardy-Weinberg equilibrium were identified. To identify common haplotype tagging SNPs, these eligible SNPs were entered into the Tagger program implemented in Haploview version 3.32 (29) (http://www.broad.mit.edu/mpg/haploview/) with a minor allele frequency threshold of 20% and an r² of 0.8. Thirteen tag SNPs (rs6585194, rs7919409, rs11196219, rs4918792, rs10749127, rs11196224, rs7085532, rs17130188, rs10787475, rs12775879, rs290489, rs290487, and rs290481) were selected, which captured 92.2% of the variance for haplotypes composed of all SNPs with minor allele frequencies ≥20% across the TCF7L2 gene. Because nearly all SNPs within the exon 4 LD block have minor allele frequencies <5% in the HapMap CHB databank, we additionally selected eight SNPs with known associations with type 2 diabetes (rs4506565, rs12243326, rs7895340, rs7904146, rs12255372, rs7079711, rs11196192, and rs11196213) within this region. Genotyping was performed using the GenomeLab SNPstream genotyping platform (Beckman Coulter) and its accompanying SNPstream software suite (30).

A Hardy-Weinberg equilibrium test was performed for each sequence variant for the control group before marker-trait association analysis. Tests for the associations of each SNP and haplotype with type 2 diabetes were estimated by using the Haploview software (29). Nominal two-sided P values were reported and were corrected for multiple testings by permutation 10,000 times. For haplotype construction, genotype data of both case and control groups were used to estimate intermarker LD by measures pairwise D′ and r² and define LD blocks. We used the solid spine of the LD method implemented in the Haploview software to define an LD block with an extended spine if D′ was >0.8 (29).

For the sliding window association analysis of the haplotypes composed of the adjacent 2–10 SNPs, we used the score test developed by Schaid et al. (31) and implemented in the Haplo.Stats package (http://www.mayo.edu/hsr/Sfunc.html). This method allows the adjustment for possible confounding variables and provides both global tests and haplotype-specific tests. The global P value for the associations of haplotypes composed of the adjacent 2–10 SNPs with type 2 diabetes were estimated with simulations 10,000 times. A P value <0.05 was considered statistically significant. We also performed multivariate analysis by considering age, sex, and BMI as covariates using the Haplo.Stats program. The population attributable risk fraction (PAF) was estimated with the data from the control group, calculated as follows: 1 − {1/[p² OR_homo + 2p(1 − p)OR_hetero + (1 − p)²]}, where p is the risk-allele frequency, OR_homo is the odds ratio (OR) for homozygotes, and OR_hetero is the OR for heterozygotes.

We genotyped a total of 1,520 unrelated subjects (including 760 type 2 diabetic case and 760 control subjects) from a Han Chinese population in Taiwan. The Han Chinese is the largest ethnic group in Taiwan, comprising ∼98% of the general population. The clinical and biochemical characteristics are summarized in Table 1.

On average, 99.08% of attempted genotypes were successful (success rates ranging from 98.1 to 99.9% for each SNP) except for rs6585194, in which genotyping failed in all samples. The concordance rate of genotyping duplication was 99.93%. Graphical representation of SNPs in relation to the exon-intron structure is shown in Fig. 1. All SNPs were in introns. The genomic position, nucleic acid composition, and minor allele frequencies of the 20 genotyped SNPs are summarized in Table 2.

To determine the extent of LD in the TCF7L2 gene, genotype data of both case and control groups were used to estimate intermarker LD. Standardized pairwise LD coefficients D′ and r² between markers were estimated (Table 3). Six LD blocks were identified across the gene (Fig. 1). LD block 1 spanning over a part of intron 3, the whole exon 4, and a part of the intron 4 corresponds to the exon 4 LD block described previously by Grant et al. (6), which contains the variants of the strongest associations with type 2 diabetes in the populations of European ancestry.

The minor allele frequencies of SNPs located in LD block 1 were low (0.4–2.6%, Table 2) in this Chinese population compared with populations of European ancestry. The genetic variations within this region were rare, with one single haplotype (A-C-T-T-G-G) accounting for 95% of all chromosomes (Table 4). We did not detect any association of type 2 diabetes with any SNP within this LD block, including rs7903146 and rs12255372 (Table 2). There was no association of haplotypes in LD block 1 with type 2 diabetes (global P = 0.47, best haplotype-specific P = 0.39) (Table 4). Sliding window analysis for the haplotypes composed of adjacent SNPs (2–10 SNPs in length) also failed to reveal any association with type 2 diabetes within this region (Table 5).

However, we found that SNP rs290487 (minor allele frequency 39.03%), located 150 kb downstream of rs7903146, was significantly associated with type 2 diabetes (nominal P = 0.0021) (Table 2). The association remained significant after correction for multiple testing (permuted P = 0.03) (Table 2). The allelic OR of the rs290487 C-allele carriers was 1.26 (95% CI 1.08–1.46) (Table 2). The genotypic OR for the heterozygous genotype (CT) was 1.36 (1.08–1.71, P = 0.0063) and 1.51 (1.10–2.07, P = 0.0085) for homozygous genotype (CC), compatible with either dominant or additive genetic models (P = 0.0017 and 0.0059, respectively) and corresponding to a PAF of 18.7%. The SNP rs290487 was located in an LD block (LD block 6) spanning from intron 7 to intron 13 (14 kb in length) (Fig. 1). Another SNP rs290481 in this block was also associated with type 2 diabetes with marginal significance (nominal P = 0.021, permuted P = 0.27) (Table 2). The haplotypes within LD block 6 were significantly associated with type 2 diabetes (global P = 0.012) (Table 4). Among them, the haplotype T-A was associated with a reduced risk (OR 0.83 [95% CI 0.71–0.96], permuted P = 0.033) (Table 4), and the haplotype C-G was associated with an increased risk of type 2 diabetes (1.24 [1.06–1.44], permuted P = 0.014) (Table 4). These associations remained significant after adjustment for sex, age, and BMI (Tables 2 and 4), suggesting their effects on diabetes were not primarily mediated through adiposity. Sliding window analysis also revealed consistently significant associations of haplotypes containing rs290487 with type 2 diabetes (best global P = 0.002) (Table 5).

The polymorphism of TCF7L2 (rs7903146) is the strongest single genetic variant associated with type 2 diabetes (6,23–26) and has been convincingly replicated in multiple populations (7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26). In this study, we aimed to explore the effect of the TCF7L2 polymorphisms in a Han Chinese population. We found that the type 2 diabetes risk–conferring alleles in the exon 4 LD block described previously by Grant et al. (6), such as rs7903146 T and rs12255372 T, were rare in this Chinese population. We did not detect any association of genetic variants within this region with type 2 diabetes in this Chinese population. However, given their low allele frequencies of 2.3% and assuming the genotype relative risks of 1.45 and 2.41 in heterozygous and homozygous carriers, respectively (6), and the diabetes prevalence of 8% in the Chinese population in Taiwan, enrollment of ∼1,700 case subjects would be necessary for the case-control study to have 80% power (32). Thus, a sample size at least twice larger than our study would be needed to detect the effect of these genetic variants.

Florez et al. (22) recently reported the associations of these variants with increased risk of developing diabetes in a large prospective cohort of the Diabetes Prevention Program. In subgroup analysis, there were also no associations of these variants with type 2 diabetes in Asians, which could at least attribute to a small sample size (128 participants) (22). Horikoshi et al. (33) and Hayashi et al. (34) independently reported significant associations between rs7903146 and type 2 diabetes in the Japanese population. However, the minor allele frequency of rs7903146 was low (3–5%) in their study. The PAF for rs7903146 is only ∼3% in the Japanese population, a much lower value than that in the populations of European ancestry (6,33,34). These data indicated that the genetic variants within the exon 4 LD block were rare among the Japanese and Chinese population, so that large sample sizes would be needed to detect the effects.

Interestingly, we identified a novel association of SNP rs290487 C located in an LD bock spanning over intron 7 to intron 13 with type 2 diabetes. In contrast to rs7903146 T, rs290487 C was common (allele frequency 39.03%) in the Chinese population. The risk allele and genotypes were associated with a significantly increased risk of type 2 diabetes, and the corresponding PAF was also substantial (18.7%) in the Chinese population, comparable with that of rs7903146 T in the populations of European origin (6). These data suggested that genetic variants of the TCF7L2 gene were also major determinants of type 2 diabetes in the Chinese population. Moreover, our results along with the published reports from other ethnic groups demonstrate that different variations of the TCF7L2 gene confer the risk of type 2 diabetes in different populations. These findings provided novel evidence supporting a role of the TCF7L2 locus in the pathogenesis of type 2 diabetes in a Chinese population.

The underlying mechanisms by which a genetic variation within the intron of the TCF7L2 gene confers susceptibility of type 2 diabetes remain to be elucidated. The LD block harboring rs290487 lies very close (<2 kb) to the 3′ end of TCF7L2 gene. The human TCF7L2 gene displays a high degree of alternative splicing near the 3′ end of the gene (35,36). The 3′ end of the TCF7L2 gene encodes the binding domains for COOH-terminal binding protein, a protein implicated in the repression of the TCF7L2 transcriptional activity. The alternative splicing leads to the synthesis of a number of isoforms with different COOH-terminal ends and binding domain activities, which were associated with altered downstream signaling activities (35,36). Furthermore, many alternative splice sites were observed either experimentally or in silico in this region (35). It is intriguing to speculate that genetic variations in this region would affect the action of TCF7L2 through the regulation of alternative splicing. Based on multispecies sequence alignment analysis using a phylogenic hidden Markov model implemented in the PhastCons program (http://genome.ucsc.edu/cgi-bin/hgGateway) (37), there is a high degree of evolutional conservation in the 3′ end of human TCF7L2 gene, which we identified for association of diabetes. Functional studies will be required to determine the relationship between sequence variations, gene expression, protein product, and function in human tissues. LD block 6 contains two nonsynonymous exonic SNPs and 37 intronic SNPs. Therefore, further extensive search within this region may be needed to identify the true risk-conferring site as well as to confirm the association.

In summary, we identified a novel association of the genetic variant SNP rs290487 in the TCF7L2 gene with type 2 diabetes in a Chinese population, in contrast to the previous reports of the SNP rs7903146 in other ethnic groups. The variant was located in an LD block close to the 3′ end of the TCF7L2 gene, a region displaying a high degree of alternative splicing. The effect of this variant was significant, with a substantial PAF in the Chinese population. We did not detect any association of the previously established risk allele (rs7904136 T and rs12255372 T) with type 2 diabetes, probably owing to their low frequency in the Chinese population. These data suggested that genetic polymorphisms of the TCF7L2 gene were associated with type 2 diabetes and are major genetic contributors of the risk of type 2 diabetes a Chinese population.

This research project was supported by grants from the National Science Council (NSC-93-B-002-005 and NSC-94-3112-B-002-019) of Taiwan.

We thank all of the patients for their participation in this study.