A Single Nucleotide Polymorphism within DUSP9 Is Associated with Susceptibility to Type 2 Diabetes in a Japanese Population

Aims The DUSP9 locus on chromosome X was identified as a susceptibility locus for type 2 diabetes in a meta-analysis of European genome-wide association studies (GWAS), and GWAS in South Asian populations identified 6 additional single nucleotide polymorphism (SNP) loci for type 2 diabetes. However, the association of these loci with type 2 diabetes have not been examined in the Japanese. We performed a replication study to investigate the association of these 7 susceptibility loci with type 2 diabetes in the Japanese population. Methods We genotyped 11,319 Japanese participants (8,318 with type 2 diabetes and 3,001 controls) for each of the 7 SNPs–rs5945326 near DUSP9, rs3923113 near GRB14, rs16861329 in ST6GAL1, rs1802295 in VPS26A, rs7178572 in HMG20A, rs2028299 near AP3S2, and rs4812829 in HNF4A–and examined the association of each of these 7 SNPs with type 2 diabetes by using logistic regression analysis. Results All SNPs had the same direction of effect (odds ratio [OR]>1.0) as in the original reports. One SNP, rs5945326 near DUSP9, was significantly associated with type 2 diabetes at a genome-wide significance level (p = 2.21×10−8; OR 1.39, 95% confidence interval [CI]: 1.24−1.56). The 6 SNPs derived from South Asian GWAS were not significantly associated with type 2 diabetes in the Japanese population by themselves (p≥0.007). However, a genetic risk score constructed from 6 South Asian GWAS derived SNPs was significantly associated with Japanese type 2 diabetes (p = 8.69×10−4, OR  = 1.06. 95% CI; 1.03−1.10). Conclusions/interpretation These results indicate that the DUSP9 locus is a common susceptibility locus for type 2 diabetes across different ethnicities, and 6 loci identified in South Asian GWAS also have significant effect on susceptibility to Japanese type 2 diabetes.


Introduction
Diabetes mellitus affects more than 300 million individuals worldwide, and its prevalence is progressively increasing, particularly in East Asian and Western Pacific regions [1]. Type 2 diabetes is the most common form of the disease, and the combination of insulin resistance in peripheral tissues and impaired insulin secretion from pancreatic b cells is believed to contribute to the development and progression of type 2 diabetes.
Although the precise mechanisms responsible for the onset of the disease are still unknown, multiple genetic factors may cooperate with environmental factors to confer susceptibility to type 2 diabetes [2].
An expanded meta-analysis of the existing European GWAS data (DIAGRAM+: Diabetes Genetics Replication and Meta-analysis+) has identified 12 loci, including 11 autosomal loci and the DUSP9 locus on chromosome X, as a strong susceptibility locus for type 2 diabetes [15]. DUSP9 encodes a member of the family of mitogen-activated protein kinase phosphatase 4, MKP4, which was also shown to play pivotal roles in regulating insulin actions [29], [30], [31]. Although the DUSP9 locus is considered an important locus for conferring susceptibility to insulin resistance, the contribution of this locus to type 2 diabetes susceptibility in the Japanese population has not yet been evaluated.
In this study, we aim to evaluate the association of 7 loci, which have not been evaluated in Japanese populations, with susceptibility to type 2 diabetes in a Japanese population: 1 single nucleotide polymorphism (SNP) in DUSP9 on chromosome X and 6 South Asian GWAS-derived loci-GRB14, ST6GAL1, VPS26A, HMG20A, AP3S2, and HNF4A.

Participants and DNA Preparation
We enrolled 8,318 type 2 diabetes patients who regularly visited the outpatient clinics of the Shiga University of Medical Science, Kawasaki Medical School, St. Marianna University, Juntendo University, and the University of Toyama or who were registered in BioBank Japan [24]. Diabetes mellitus was diagnosed by the World Health Organization (WHO) criteria [32], and type 2 diabetes was defined by a gradual adult onset. Patients who tested positive for antibodies to glutamic acid decarboxylase or who were diagnosed with mitochondrial disease or maturity-onset diabetes of the young were excluded from the present study. We also recruited 3,001 controls who underwent annual health check-ups at Keio University, St. Marianna University, or Toyama University Hospital or from the general Japanese population registered in the Japanese SNP database [24]. Obesity was defined as individuals with BMI $25 according to the criteria of Japan society for the study of obesity [33].

Ethics Statements
All participants agreed to the protocol of this study and provided written informed consent. DNA samples were obtained from the peripheral blood of each participant, and the study protocol was approved by the ethics committees of RIKEN Yokohama Institutes and each of the participating institutes, i.e. Shiga University of Medical Science, Kawasaki Medical School, St. Marianna University, Juntendo University, the University of Toyama, and Keio University.

SNP Genotyping
We selected 6 autosomal SNPs identified by GWAS in individuals of South Asian ancestry, which were reported by Kooner et al [27]. These included rs3923113 near the growth factor receptor-bound protein 14 gene (GRB14), rs16861329 in the ST6 beta-galactosamide alpha-2,6-sialyltranferase 1 gene (ST6GAL1), rs1802295 in the vacuolar protein-sorting 26 homolog A gene (VPS26A), rs7178572 in the high-mobility group 20A gene (HMG20A), rs2028299 near the adaptor-related protein complex 3 sigma 2 subunit gene (AP3S2), and rs4812829 in the hepatocyte nuclear factor 4 alpha gene (HNF4A). We also examined 1 Xchromosomal SNP, rs5945326, near the dual-specificity phosphatase 9 gene (DUSP9), which was identified in a European GWAS reported by Voight et al [15]. The 11 autosomal SNPs also reported by Voight et al. [15] have been already evaluated in the present samples [20], and not examined in the present study.
Genotyping was performed using the multiplex-polymerase chain reaction (PCR) invader assay [34]. Genotyping success rates for all SNPs were over 98% (98.3% for rs3923113, 98.8% for rs16861329, 98.8% for rs1802295, 99.1% for rs7178572, 99.1% for rs2028299, 98.9% for rs4812829, 98.9% for rs5945326), and concordance rates for all SNPs in 95 randomly selected duplicated samples are 100%. Since our samples have been collected in 6 distinct study groups, we compared the risk allele frequencies of the 7 SNPs among the 6 groups. The results indicated, risk allele frequencies of each SNPs were not different among the 6 groups, except for rs5945326, of which risk allele frequency in group 6 control was lower than those in other controls (Table S1).

Statistical Analysis
We performed Hardy-Weinberg equilibrium (HWE) tests according to the method described by Nielsen et al [35]. Differences in the genotype distribution of each SNP between cases and controls were evaluated by logistic regression analysis with or without adjusting for age, sex, and body mass index (BMI). The association of each SNP with quantitative traits, BMI, fasting plasma glucose (FPG), the homeostasis model assessment of b-cell function (HOMA-b), and the HOMA of insulin resistance (HOMA-IR) [36], [37] was evaluated by multiple linear regression analysis. Because values of these traits in the present Japanese population showed a skewed distribution, we used log-transformed values for the analyses. Genotypes of autosomal SNPs were scored using an additive model (0, 1, and 2 for the homozygous of noneffect allele, heterozygous, and homozygous of effect allele, respectively). For evaluating X-chromosome SNP, men were coded as homozygous, and all men and homozygous women were included in the association study (0 for the homozygous of noneffect allele and 1 for the homozygous of effect allele, assuming Xchromosome inactivation in females.).
Statistical analyses were performed using StatView software (SAS Institute, Cary, NC, USA). The level of significance was determined by the Bonferroni's method for correcting multiple testing errors, and a p value of ,0.007 (0.05 divided by 7) was considered statistically significant. Table 1 Genotype distributions of all autosomal SNPs in the present study were in accordance with the Hardy-Weinberg equilibrium proportions ( Table 2). The results of the association study for each SNP with type 2 diabetes in the present Japanese population are shown in Table 3. All SNPs had the same direction of effect (odds ratio [OR] .1.0) as in the original reports [15], [27]. Of these, 1 SNP, rs5945326 near DUSP9, was significantly associated with type 2 diabetes, and the association reached a genome-wide significance level (p,5610 28 ). A subsequent sex-stratified analysis for rs5945326 showed no remarkable difference in the effect size between men and women (OR in women: 1.45, 95% confidence interval [CI]: 1.1221.87 OR in men: 1.38, 95%CI: 1.2121.57 Table S2). One SNP, rs7178572 in HMG20A, had a nominal association with type 2 diabetes (p,0.05, Table 3); however, this association was not significant after Bonferroni's correction. The remaining 5 SNPs were not associated with type 2 diabetes in the Japanese population (p$0.05). We did not observe a significant association of 6 autosomal SNPs with type 2 diabetes in a dominant or recessive association model (Table S3, Table S4). However, a genetic risk score (GRS) constructed from the 6 SNPs was significantly associated with type 2 diabetes in the present Japanese population (Table 3).

Characteristics of the Study Participants are Shown in
We next examined the associations between these SNPs and quantitative traits related to glucose metabolism, such as HOMA-IR, HOMA-b, and FPG, by using the control samples. The Aallele of rs5945326 (near the DUSP9) had a modest effect on increased HOMA-IR and HOMA-b values ( Table 4, Table S5). None of the SNPs showed significant association with these glycemic traits in the present Japanese population, although the sample size might not be sufficiently large enough (Table S6). We further searched MAGIC (the Meta-Analyses of Glucose and Insulin-related traits Consortium) data for 6 South Asian GWAS derived SNPs. Although rs3923113 near GRB14 has modest effect on reduced insulin sensitivity (HOMA-IR, b = 0.011, se = 0.0042, p = 0.01275), we did not find significant association between the 6 SNPs and glycemic traits (Table S7).
We also examined the association between these SNPs and BMI ( Table 5, Table S8). The results indicated that rs2028299 near AP3S2 was significantly associated with BMI only when both patients and controls were used for the analysis, with adjustment for age, sex, and disease state of type 2 diabetes (rs2028299, b = 20.007, SE = 0.003, p = 0.0032; Table 5).

Discussion
In this study, we examined the association of 7 SNPs, previously identified by European or South Asian GWAS, with susceptibility to type 2 diabetes in a Japanese population and showed that rs5945326 near DUSP9, an X-chromosome SNP, was significantly associated with type 2 diabetes in the Japanese population.
Among the examined SNP loci in the present study, we found that rs5945326 near DUSP9 on the X chromosome was strongly associated with type 2 diabetes in the Japanese population. The association of rs5945326 attained a genome-wide significance (p,5.0610 28 ) in the present sample, and the risk allele (A) was in accordance with that in the original report [15]. Furthermore, the association of rs5945326 with type 2 diabetes was also observed in Pakistani [39] or in Han Chinese population [40], indicating that this locus is a common susceptibility locus for type 2 diabetes across multiple ethnic groups. Furthermore, the effect size (OR) of rs5945326-A in the present Japanese sample (OR = 1.39) was larger than that in other Japanese type 2 diabetes loci, except for KCNQ1 and TCF7L2 [17], [18], [22], [23]. This result suggests that the DUSP9 locus is one of the most important susceptibility loci for type 2 diabetes in the Japanese population. Although the risk allele frequency of rs5945326 is lower in one control group than those in other control groups, the association of rs5945326 was still statistically significant even after excluding the control (group 6 control, Table S1) (p = 2.95610 24 , OR = 1.30, 95% CI; 1.13-1.49); therefore, we thought that the present positive finding was not affected by a sample heterogeneity.
DUSP9 encodes a dual-specificity phosphatase 9 (MKP4). MKP4 is expressed in insulin-responsive tissues, including adipose, muscle, and liver tissues. Emanuelli et al. showed that MKP4 has protective effect against the development of insulin resistance because of its ability to inactivate extracellular signal-regulated kinase (ERK) or c-Jun N-terminal kinase (JNK), which are considered to be the mediators of stress-induced insulin resistance [31]. In this report, Emanuelli et al. also showed that overexpression of Mkp4 in the liver by injecting adenovirus vectors encoding Mkp4 ameliorated insulin resistance in ob/ob mice, suggesting that Mkp4/Dusp9 has favourable effects on glucose metabolism by increasing insulin sensitivity. In contrast, other research groups have shown that MKP4 decreases insulin-stimulated glucose uptake in cultured adipocytes [29], [30]. Therefore, the effects of MKP4/DUSP9 on the metabolic action of insulin may differ between different tissues or under different conditions. Regardless, MKP4 can be considered an important regulator of insulin sensitivity. In the present study, we observed that a risk allele of rs5945326 near DUSP9 tended to increase HOMA-IR (b = 0.066, p = 0.0769; Table 4), and subsequent BMI-stratified analysis showed that the effects of the rs5945326-A allele could be observed only in the non-obese group (BMI,25, b = 0.099, p = 0.0198; Table 4, Table S9) but not in obese group (BMI$25, b = -0.035, p = 0.6423); the effect of rs5945326-A on reducing insulin sensitivity might be masked by obesity-induced insulin resistance in the obese group. Although the association of rs5945326 with HOMA-IR was still not statistically significant after correction for multiple testing errors, the rs5945326-A allele may confer susceptibility to type 2 diabetes by reducing insulin sensitivity.
Regarding the remaining 6 loci derived from a South Asian GWAS, rs7178572 in the intron of HMG20A was nominally associated with type 2 diabetes; however, no SNPs showed a significant association with type 2 diabetes in the present Japanese population after correction for multiple testing errors. Because all SNP loci had the same direction of effect as in the original reports (OR.1.0) and the estimated powers in the present study for these unreplicated SNP loci were between 28% and 76% (Table S10), insufficient power of the present study may be the principal explanation for the differences between the present study and the original study in South Asian populations. Then we further examined a combinational effect of these 6 loci using GRS constructed from the 6 South Asian GWAS derived SNPs, and found that the GRS was significantly associated with Japanese type 2 diabetes (Table 3), suggesting these loci also have significant effect on susceptibility to Japanese type 2 diabetes.
Because control subjects in the present study were younger than type 2 diabetes patients, the control group might include several individuals who would develop the disease, and the possibility of type 2 error may be increased, although results were not affected by adjusting age. Then we evaluated the association of the 7 SNPs with type 2 diabetes using older control individuals (age$50 or  $60). The results were almost the same as original findings, although effect sizes for rs3923113 and rs16861329 on the disease susceptibility were likely increased when we selected older control individuals for the association study (Table S11). In addition, there were considerable differences in the risk allele frequencies for these loci between the Japanese and South Asian populations (Table  S12). In a subsequent study to determine the association of all 7 SNPs with BMI, we observed a significant association of rs2028299 near AP3S2 with BMI when the analysis was performed using all participants, i.e. both type 2 diabetes patients and controls. AP3S2 encodes a clathrin-associated adaptor complex. The expression of AP3S2 has been shown to be ubiquitous, and no report has yet emerged to suggest a functional role of this gene product in the pathogenesis of obesity or type 2 diabetes. There are several nearby genes (other than AP3S2) around rs2028299, including PLIN1 that lies 200 kb from rs2028299 and encodes perilipin-1, a hormonally regulated phosphoprotein that coats fat droplets [41]. Because perilipin has been shown to play pivotal roles in adipocyte lipid metabolism and has been shown to have an association with obesity in humans and experimental animal models [42][43][44][45], the present signal may reflect the effects of nearby genes, including PLIN1. However, the association of rs2028299 with BMI in the present study was modest, and a larger association study is required to clarify the association of this locus with BMI.
In conclusion, we observed a significant association of rs5945326 with type 2 diabetes in a Japanese population. The results indicate that rs5945326 is a common type 2 diabetes susceptibility locus across different ethnic groups. Although it is suggested that the 6 South Asian GWAS-derived loci have significant effect on conferring susceptibility to type 2 diabetes in the Japanese, studies with larger cohorts from the Japanese population are required for evaluating the association of the 6 loci with type 2 diabetes. Moreover, further functional studies are  necessary to elucidate the biological mechanisms of each locus for conferring susceptibility to the disease.