Association Studies of Variants in the Genes Involved in Pancreatic β-Cell Function in Type 2 Diabetes in Japanese Subjects

• HWE, Hardy-Weinberg equilibrium
• KATP channel, ATP-sensitive K+ channel
• LD, linkage disequilibrium
• SNP, single nucleotide polymorphism

Impaired insulin secretion and insulin resistance both contribute to the pathogenesis of type 2 diabetes. The former is a characteristic feature of type 2 diabetes, especially in Asians including Japanese (1), and genes encoding proteins critical in pancreatic β-cell function are therefore particularly good candidate susceptibility genes for type 2 diabetes for this population. Studies of maturity-onset diabetes of the young in humans (2) and knockout mice (3) have shown that mutations of transcription factors required for development, differentiation, and maintenance of the pancreatic β-cells can cause diabetes. Pancreatic β-cell ATP-sensitive K⁺ channels (K_ATP channels) are crucial in the regulation of insulin secretion by coupling cell metabolism to membrane electrical activity. The pancreatic β-cell K_ATP channel comprises two subunits, the inwardly rectifying potassium channel Kir6.2 (KCNJ11) and the sulfonylurea receptor SUR1 (ABCC8) (4). Mutations in the genes (ABCC8 and KCNJ11) can cause familial persistent hyperinsulinemic hypoglycemia of infancy (5) and permanent neonatal diabetes (6). Several polymorphisms in these genes also have been reported to be associated with type 2 diabetes in populations with distinct ethnic backgrounds (7–20). However, a large-scale association study of these genes has not been performed in type 2 diabetes in the Japanese population. Here, we report on the association of variants in genes encoding various transcription factors and pancreatic β-cell K_ATP channel subunits with type 2 diabetes in a large Japanese cohort.

A case-control association study using 1,590 Japanese diabetic subjects and 1,244 nondiabetic control subjects was performed. All subjects were genotyped for 33 variants of 12 genes including transcription factors (TCF1, TCF2, HNF4A, ISL1, IPF1, NEUROG3, PAX6, NKX2–2, NKX6–1, and NEUROD1) and β-cell K_ATP channel subunits (KCNJ11 and ABCC8) (Table 1).

Results of Hardy-Weinberg equilibrium (HWE) tests are shown in Table 1 of the online appendix (available at http://diabetes.diabetesjournals.org). All genotypes were in HWE, except for departures in cases at TCF2_SNP (single nucleotide polymorphism) 5 rs2689, TCF2_SNP6 rs2688, and NKX2-2_SNP1 (+856 TGA) and in controls at NKX6-1_SNP1 rs1017560 and ABCC8_SNP1 rs1799854 (online appendix Table 1). Although none of these are significant with correction for multiple comparisons, we reanalyzed several of the variants, including NKX2-2_SNP1 (+856 TGA) and NKX6-1_SNP1 rs1017560, and confirmed that there was no typing error for these variants. We also tested whether the observed departures were consistent with the genotype frequencies expected for a genetic disease model (21). The genotype distributions for TCF2_SNP5 rs2689, TCF2_SNP6 rs2688, NKX2-2_SNP1 (+856 TGA), and ABCC8_SNP1 rs1799854 are consistent with genetic models that best fit these data. In contrast, the departure from HWE observed in the control samples for NKX6-1_SNP1 rs1017560 is not consistent with any genetic model for disease. Thus, the observed departure from HWE in controls at NKX6-1_SNP1 rs1017560 is likely to be a chance observation. The remaining departures are unlikely to be attributable to genotyping errors and are consistent with the possibility that the selection of case and control samples from a population in HWE at a susceptibility locus (at the test marker or a polymorphism in strong linkage disequilibrium [LD]) has generated genotype distributions with the observed departures from HWE.

Among the 33 variants of 12 genes, 6 variants (TCF2_SNP4 rs1016991, TCF2_SNP6 rs2688, HNF4A_SNP3 rs745975, NKX2-2_SNP2 rs3746741, NKX6-1_SNP1 rs1017560, and ABCC8_SNP1 rs1799854) showed at least nominally significant associations (P < 0.05) with type 2 diabetes (Table 1 and online appendix Table 2). ABCC8_SNP1 rs1799854 showed the strongest association (P = 0.0073) with diabetes among the SNPs examined in this study. By further analysis of the variant, the T/T genotype was found in 454 (28.6%) and 298 (24.0%) subjects in the diabetic and control groups, respectively, a significant difference in the frequency of individuals with the T/T genotype between the two groups (C/C + C/T vs. T/T, P = 0.0068) (online appendix Table 2). The odds ratio (OR) for the T/T genotype was 1.27 (95% CI 1.07–1.50; C/C + C/T vs. T/T), indicating that the T/T genotype in ABCC8_SNP1 rs1799854 is associated with type 2 diabetes in Japanese subjects.

There was no association of other variants in ABCC8 and KCNJ11 with diabetes (Table 1 and online appendix Table 2). These include the E23K variant in KCNJ11 (KCNJ11_SNP1 rs5219: for the K allele, OR 1.08 [95% CI 0.97–1.21], P = 0.15). To determine the extent of LD between the four variants in ABCC8 and KCNJ11, we calculated D′ and r² (Table 2). There was modest LD between ABCC8_SNP2 rs4148643 and ABCC8_SNP3 rs757110. Strong LD was found between ABCC8_SNP3 rs757110 and KCNJ11_SNP1 rs5219. In the latter, we tested two-locus haplotypes having a frequency of >5% for association with diabetes and found no association of any of the haplotypes with diabetes (data not shown).

We examined the genes involved in pancreatic β-cell function (transcription factors and K_ATP channel subunits) in relation to type 2 diabetes in a large cohort of Japanese subjects. The study included 2,834 subjects, the largest case-control study so far conducted on these variants in a Japanese population. For disease susceptibility allele frequencies in the range of 0.3–0.5, our sample had >99% power to detect a susceptibility gene with a genotype relative risk in the range of 1.5–1.85 (for any genetic model of inheritance). For allele frequencies in this range, we had >80% power to detect susceptibility genes with genotype relative risk in the range of 1.25–1.55. Power was similarly good for dominant models with lower susceptibility allele frequencies (0.1–0.3) or recessive models with higher susceptibility allele frequencies (0.5–0.9). The sample was reasonably powered (>90%) to detect recessive susceptibility alleles at low frequencies (0.1–0.3) for higher genotype relative risks (2.1–4.0) but was not sufficiently powered to detect very common (>0.7) dominant susceptibility genes (genotype relative risk >100).

ABCC8_SNP1 rs1799854 (exon 16 −3c/t variant) was significantly associated with type 2 diabetes, primarily due to increased frequency of T/T homozygotes among patients. Since this variant is located in the 3′ splice site, it might impair normal splicing. Alternatively, the variant could be in strong LD with an unidentified functional variant in the unscreened region harboring the ABCC8 gene. There have been two case-control association studies (22,23) conducted for the variant in Japanese populations, both of which found no association of this variant with type 2 diabetes. However, because these studies were based on a relatively small number of subjects (167 subjects in 22; 456 subjects in 23), their power to detect associations is limited. In Caucasians, several studies (7,8,11–13,17) have reported association of the variant with type 2 diabetes, although other studies found no association of the variant with type 2 diabetes (14–16). On the other hand, several studies have reported an association of the E23K variant in KCNJ11 (KCNJ11_SNP1 rs5219 in this study) with type 2 diabetes in Caucasians (9,10,16,18). Recent meta-analyses (19,20) of the variant support this association. Although our present study finds no association of the E23K variant with diabetes in Japanese subjects (for the K allele, OR 1.08 [95% CI 0.97–1.21], P = 0.15), 95% CI around the OR overlaps in meta-analysis of European populations, suggesting that our results are not inconsistent with the previous studies on the E23K variant in KCNJ11.

The International HapMap Project aims to determine the common patterns of DNA sequence variation in the human genome (24). In the initial phase of the project, genetic data are being gathered from four populations with African, Chinese, Japanese, and European ancestry. Twenty of 33 SNPs used in this study were genotyped on Japanese subjects in the HapMap project, providing important information for determining whether the genes of interest are associated with type 2 diabetes in a Japanese cohort. To clarify the relationships between our SNPs and those of the HapMap, the patterns of LD between SNPs for each gene are shown in online appendix Fig. 1. Among nine genes (TCF1, TCF2, HNF4A, ISL1, PAX6, NKX6-1, NEUROD1, ABCC8, and KCNJ11) that have a relatively large number of genotyped SNPs in the HapMap, four genes (ISL1, NKX6-1, NEUROD1, and KCNJ11) show a relatively strong LD, while five genes (TCF1, TCF2, HNF4A, PAX6, and ABCC8) show a weak LD across each gene. For the former genes, our data provide considerable information on the association of genes of interest with type 2 diabetes. However, for the latter genes, we could provide only partial information on their association with type 2 diabetes. In contrast, there are none or few genotyped SNPs in the HapMap for three genes (IPF1, NEUROG3, and NKX2-2). For these genes, our data provide valuable information on both the SNPs and their association with type 2 diabetes.

Associations of HNF4A variants in the upstream promoter region with type 2 diabetes have recently been reported in several populations (25–27). Using Japanese samples in the HapMap data, we calculated LD between our SNPs (rs717247 and rs745975) and those (rs1884614, rs2144908, and rs4810424) showing association with type 2 diabetes. However, LD was not found (online appendix Fig. 1), indicating that the association of our SNPs with type 2 diabetes is different from that of other studies. In this study, modest associations (P < 0.05) with type 2 diabetes were also detected for several of the candidate genes examined (TCF2_SNP4 rs1016991, TCF2_SNP6 rs2688, NKX2-2_SNP2 rs3746741, and NKX6-1_SNP1 rs1017560). As there has been no large association study for these variants, these associations need to be confirmed by further replication studies. Nevertheless, this is the largest association study so far conducted on these genes in Japanese subjects, providing valuable information not only for this population, but also for comparison with other ethnic groups.