Introduction

Until recently, the progress in searching for genetic variants that predispose to type 2 diabetes (T2D) was rather slow. Only the P12A variant of the PPARG gene and the E23K variant of the KCNJ11 gene had been associated, and confirmed, with this disease in various large-scale studies performed in different populations.1 Recent identification of the TCF7L2 (transcription factor 7-like 2) gene polymorphisms as major determinants of T2D risk2 became a long-awaited breakthrough in finding true T2D-susceptibility genes. Indeed, the original observation of Grant et al2 has been successfully replicated in more than 20 studies to date. TCF7L2 has therefore been proposed as ‘the biggest story in diabetes genetics since HLA.’3

Recently, Sladek et al4 reported the results of a large genome-wide association (GWA) study that identified three novel loci that contribute substantially to susceptibility for T2D, and they also confirmed the association of TCF7L2 with T2D. One of the new loci located on chromosome 10q contains genes known to have roles in the development (the hematopoietically expressed homeobox, HHEX)) and function (insulin-degrading enzyme (IDE)) of the pancreas.5, 6 Interestingly, the HHEX gene also encodes a transcription factor that is involved in Wnt signaling, a fundamental pathway for cell growth and development,7, 8 as does TCF7L2. As replication is always a critical issue in confirming the validity of genetic association studies, we have examined whether the variants near the HHEX gene contribute to the risk of T2D in a Dutch population.

Subjects and methods

Breda study

DNA was available from 501 T2D patients from the Breda study (Table 1).9 All patients were diagnosed according to the WHO criteria (random plasma glucose level >11.1 mmol/l or a fasting plasma glucose level >7.0 mmol/l). The clinical characteristics of the patients (HbA1c, plasma cholesterol, HDL-cholesterol and triglycerides) were available, as well as the level of obesity in each individual, as represented by the BMI (defined as weight in kg divided by height in m squared). The control cohort of 920 subjects comprised of healthy blood bank donors of Dutch Caucasian origin.10 All participants gave their written informed consent, and the Breda study was approved by the Medical Ethics Committee of the University Medical Center Utrecht.

Table 1 Clinical characteristics of the Dutch T2D case and control samples
Full size table

Genotyping

Single-nucleotide polymorphisms (SNPs) rs7923837 and rs1111875 were genotyped using TaqMan assay-on-demand (C_31982553_10 and C_11214581_10 assays; Applied Biosystems, Nieuwerkerk a/d IJssel, The Netherlands). Assays were performed according to the manufacturer's specifications and the genotypes were analyzed using a TaqMan 7900HT (Applied Biosystems). The DNA samples were processed in 384-well plates. Each plate contained 8 negative controls and 16 genotyping controls (4 duplicates of 4 different samples obtained from the Centre d’Etude du Polymorphisme Humain (CEPH)). The genotype success rates were 97.5% for rs7923837 and 98.4% for rs1111875. There were no discordances in the genotypes of any of the CEPH samples. The controls were in the Hardy–Weinberg equilibrium (χ2=0.05, P=0.83 for rs7923837 and χ2=0.53, P=0.46 for rs1111875).

Statistical analysis

The genotype frequencies were tested for the Hardy–Weinberg equilibrium by χ2 analysis. To test for association of genotypes and T2D, genotype-based odds ratios (ORs) with 95% confidence intervals (CIs) were calculated using a logistic regression model. For both SNPs, the allele that showed an increased frequency in the T2D patients was taken as the risk allele and the other allele as the reference. Differences in haplotype distribution in cases and controls were tested for significance using a two-sided χ2 test. Haplotypes of the two SNPs were estimated using the COCAPHASE package of the UNPHASED program.11 The D′ and r2 between the HHEX SNPs were calculated with the same package. Power calculation was completed using Quanto software12 (http://hydra.usc.edu/gxe/). The population-attributable risk was calculated for both SNPs using the corresponding allele frequencies for this marker as explained by Greenland and Rothman.13 The ‘Prioritizer’ method14 was used to investigate the T2D-susceptibility loci reported by Sladek et al.4 All statistical analyses were performed using the SPSS program, version 13.0 for Windows (SPSS Inc., Chicago, IL, USA).

Results

In the HHEX gene, we genotyped two SNPs, rs7923837 and rs1111875, that were reported by Sladek et al.4 to be strongly associated with T2D. We found significant association between T2D and the two SNP variants in our case–control study, with the same risk allele as reported in the original study (Table 2). The major alleles of both polymorphisms were overrepresented among T2D patients compared with controls (66.7 vs 64.1%, P=0.16 for rs7923837 and 64.6 vs 60.4%, P=0.027 for rs1111875). For both rs7923837 and rs1111875, both heterozygous and homozygous carriers for the major allele were associated with an increased risk for T2D compared with non-carriers (OR: 1.60, 95% CI: 1.09–2.35, P=0.016 and OR: 1.53, 95% CI: 1.04–2.25, P=0.033, respectively, for rs7923837, and OR: 1.68, 95% CI: 1.18–2.39, P=0.004 and OR: 1.68, 95% CI: 1.16–2.41, P=0.005, respectively, for rs1111875). For both rs7923837 and rs1111875, we see an equal risk for T2D from one or two copies of the risk alleles in the HHEX gene, implying a dominant effect.

Table 2 Association study of the rs7923837 and rs1111875 SNPs with T2D
Full size table

In our control population, the two SNPs were found to be in strong linkage disequilibrium (LD) with each other (D′=0.93, r2=0.74). The haplotype analysis did not reveal a risk haplotype that provided stronger evidence for association with T2D than each SNP individually (data not shown).

Discussion

Our data confirm the association of the common variants of rs7923837 and rs1111875 near the HHEX gene, identified by Sladek et al,4 with an increased risk of T2D in a Dutch population. The risk allele frequencies for both SNPs were similar to those reported in the original study. Both SNPs were significantly associated with T2D risk under a dominant genetic model. The population-attributable risks due to the at-risk alleles of rs7923837 and rs1111875 were estimated to be 33 and 36%, respectively.

Our study had 60% power to detect the ORs reported by Sladek et al4 for rs11117875 with a significance level of 0.05 and assuming an allele frequency of the risk allele of 0.60 and a log-additive model, as was reported by Sladek et al.4 Similarly, for rs7923837, we had 71% power to detect the reported ORs by Sladek et al.4 It needs to be noted that in the current study we used control subjects randomly selected from blood bank donors. Thus, we cannot exclude the possibility that some of these individuals may have diabetes or will develop this condition in later life. However, this would result in an underestimation of the ‘true’ effect size of HHEX on susceptibility to T2D in our study.

Although the HHEX gene is an attractive diabetes gene from the locus on chromosome 10q, we cannot exclude the possibility that one of the other genes within the locus – IDE of KIF11 – is the causal one because of the extended LD. We reasoned that it is likely that the causal genes would be functionally related in some way and we used the Prioritizer program14 to investigate how genes within these loci interact with each other and also with the known T2D genes PPARG and TCF7L2. In short, Prioritizer is a bioinformatics tool that is specifically designed to prioritize candidate genes from associated regions in the genome based on the function of the genes. Prioritizer is based on the assumption that disease genes in a specific disorder are usually functionally related, and it uses various resources, such as Reactome, Gene Ontology (GO), KEGG, etc, to identify interacting genes (a detailed description of the methods used by this tool is described by Franke et al14). Results from Prioritizer indicate that HHEX is predicted to interact with PPARG and TCF7L2, whereas KIF11 and IDE cannot be easily related to any of these genes, which suggests that HHEX is the most likely causal candidate (Figure 1). For each interaction, evidence for the interaction can be looked up at the accompanying website, http://www.genenetwork.nl, along with the sources where this evidence came from. Inspection of these interactions reveals that the predicted interactions between PPARG, TCF7L2 and HHEX are based on the fact that GO terms are shared between these genes. As this program uses existing databases and does not generate experimental evidence, the results should be interpreted with caution. With this analysis, we merely wish to emphasize that among the possible candidate genes, given the association of the two SNPs and the LD patterns in this genomic region, HHEX is the most likely candidate.

Figure 1
figure 1

Prioritizer analysis of the T2D-susceptibility loci identified by Sladek et al.4 Two loci (one on chromosome 10q and one on chromosome 11q) reported by Sladek et al4 to be significantly associated with T2D contained multiple genes. Owing to LD, we could not unequivocally determine which genes (IDE, KIF11 and HHEX from 10q and EXT2 and ALX4 from 11q) were the causal ones. We reasoned that it is likely that the causal genes would be functionally related in some way and we used the Prioritizer14 to investigate how genes within these loci interact with each other and also with the known T2D genes PPARG and TCF7L2. The results indicated that HHEX and ALX4 are predicted to interact with each other and with PPARG and TCF7L2, whereas the other positional candidate genes, IDE, KIF11 and EXT2, cannot be easily related to any of these genes. This suggests that HHEX and ALX4 are the most likely causal candidate genes from a biological perspective.

Full size image

We have previously reported the association of a TCF7L2 variant with T2D in the same study sample.15 TCF7L2 has recently been established as a major determinant of diabetes risk.3 Involvement in the Wnt signaling pathway affecting β-cell development and/or function has been proposed as the most likely mechanism for the role of TCF7L2 in the pathogenesis of diabetes.7 The HHEX gene encodes a transcription factor that is also involved in the Wnt signaling pathway8 and that has been shown to be essential for pancreas development.5 It might well be interesting to study a possible epistatic interaction between these two genes. As interaction studies require large study populations, it is not possible to answer this question with the Breda cohort. Functional studies and studies in larger cohorts will be required to investigate the possible epistasis between HHEX and TCF7L2, and to understand the role both genes may play in the pathogenesis of T2D.

In conclusion, we have confirmed that variants of the HHEX gene contribute to the risk of T2D in a Dutch population. The association of the rs7923837 and rs1111875 variants at HHEX with T2D risk has recently been confirmed in three large GWA studies.16, 17, 18, 19 These replications of the association reported by Sladek et al3 support the view that such studies can be successful in identifying novel pathways and susceptibility genes for T2D.