Genetic Associations with Diabetes: Meta-Analyses of 10 Candidate Polymorphisms

Aims The goal of our study is to investigate the combined contribution of 10 genetic variants to diabetes susceptibility. Methods Bibliographic databases were searched from 1970 to Dec 2012 for studies that reported on genetic association study of diabetes. After a comprehensive filtering procedure, 10 candidate gene variants with informative genotype information were collected for the current meta-anlayses. Using the REVMAN software, odds ratios (ORs) with 95% confidence intervals (CIs) were calculated to evaluate the combined contribution of the selected genetic variants to diabetes. Results A total of 37 articles among 37,033 cases and 54,716 controls were involved in the present meta-analyses of 10 genetic variants. Three variants were found to be significantly associated with type 1 diabetes (T1D): NLRP1 rs12150220 (OR = 0.71, 95% CI = 0.55–0.92, P = 0.01), IL2RA rs11594656 (OR = 0.86, 95% CI = 0.82–0.91, P<0.00001), and CLEC16A rs725613 (OR = 0.71, 95% CI = 0.55–0.92, P = 0.01). APOA5 −1131T/C polymorphism was shown to be significantly associated with of type 2 diabetes (T2D, OR = 1.27, 95% CI = 1.03–1.57, P = 0.03). No association with diabetes was showed in the meta-analyses of other six genetic variants, including SLC2A10 rs2335491, ATF6 rs2070150, KLF11 rs35927125, CASQ1 rs2275703, GNB3 C825T, and IL12B 1188A/C. Conclusion Our results demonstrated that IL2RA rs11594656 and CLEC16A rs725613 are protective factors of T1D, while NLRP1 rs12150220 and APOA5 −1131T/C are risky factors of T1D and T2D, respectively.


Introduction
The prevalence of diabetes is soaring up in the recent decades. The global number of diabetes patients was 173 million in 2002 and will increase to 350 million by 2030. As a group of metabolic diseases characterized with high blood sugar, most diabetes is caused by either a lack of insulin for type 1 diabetes (T1D) or a blockage in the insulin signaling pathway for type 2 diabetes (T2D). The classical symptoms of diabetes consist of polyuria, polydipsia, polyphagia and weight loss. Diabetes also causes damages to blood vessels and capillaries that may eventually lead to coronary heart diseases and blindness, respectively.
T1D and T2D are two major types of diabetes [1]. Microbial exposures and sex hormones together with lifestyle factors have been shown to be important to the development of this complex disease [2,3]. Besides environmental factors, twin studies have demonstrated a strong heritability for diabetes [4,5] and insulin related phenotypes [6][7][8]. A handful of candidate genes have been found for both the risk and complex traits of the two major types of diabetes [3,[9][10][11][12].
T1D is an autoimmune disease. Little or no insulin is produced by pancreatic beta cells that may be mistakenly attacked after an infection or some other triggers. The present meta-analyses of T1D focus on four immunomodulatory genes including IL2RA, NLRP1, IL12B and CLEC16A. IL2RA gene encodes the a-chain of IL-2 receptor (IL-2R) complex which acts as an important modulator to regulate T-cell immune response [13]. NLRP1 gene encodes a member of the Ced-4 family of apoptosis proteins that can stimulate innate immunity [14]. IL12B gene encodes a subunit of an important immunomodulatory cytokine, IL-12. IL-12 induces production from NK and T cells of interferon c (IFN-c) which favors Th1 cell differentiation [15]. CLEC16A encodes C-type lectin domain family 16 (CLEC16A) protein highly expressed on B-lymphocytes, natural killer (NK) and dendritic cells [16].
The impairment of insulin signaling in T2D is complex. Insulin signaling is involved in both glucose and lipid metabolism. In the present meta-analyses of T2D, we selected 2 genes in glucose metabolism (SLC2A10 and CASQ1), 2 genes in lipid metabolism (APOA5 and KLF11), and 2 genes in signal transduction (ATF6 and GNB3). SLC2A10 gene encodes a member of the facilitative glucose transporter family with an effect on maintaining glucose homeostasis. CASQ1 gene encodes acidic glycoprotein calsequestrin 1 (CASQ1) that is a calcium storage protein and calcium is considered to regulate the expression of the insulin-responsive glucose transporter GLUT4 [17]. APOA5 is located on human chromosome 11q23, in the APOA1/APOC3/APOA4 gene cluster [18]. KLF11 encodes Kruppel-like factor 11 with the function of regulating hepatic lipid metabolism [19]. ATF6 encodes UPR transducer unfolded protein that is related to the endoplasmic reticulum stress in the b-cell pathogenesis of type 2 diabetes [20]. GNB3 encodes the b3 subunit of hetero-trimeric G proteins in insulin signaling [21].
Associations between single-nucleotide polymorphisms (SNPs) of the above 10 genes and diabetes (including T1D and T2D) have been reported in different ethnic populations [16,. Here we performed a series of meta-analyses for these SNPs whose allelic frequencies are often substantially different among multiple ethnic populations. The goal of our study is to evaluate the overall contribution of these SNPs to diabetes susceptibility in combined populations using a meta-analysis approach.

Search Strategy and Study Selection
An initial search was performed through online databases including PubMed, Embase, SpingerLink, Web of Science, Chinese National Knowledge Infrastructure (CNKI), and Wanfang. The keywords comprise the terms including ''diabetes'' together with ''SNP'' or ''polymorphism'' or ''variants'' or ''mutation''. The selection of studies in our meta-analysis was abided by the criteria as follows: (1) case-control studies; (2) selected studies have sufficient data to calculate ORs with the corresponding 95% CIs; (3) every polymorphism has at least 3 independent datasets from the retrieved articles; (4) selected polymorphisms have not been addressed in previous meta-analysis of diabetes. Finally, the current meta-analysis involved a total of 10 genetic variants comprising NLRP1 rs12150220, IL2RA rs11594656, CLEC16A rs725613,APOA5 21131T/C, SLC2A10 rs2335491, ATF6 rs2070150, KLF11 rs35927125, CASQ1 rs2275703, GNB3 C825T, and IL12B 1188A/C.

Statistical Analysis
All the analyses were performed in Review Manager (version 5.0, The Cochrance Collaboration [58]). The combined ORs and the corresponding 95% CIs were calculated and demonstrated in the forest plots using the fixed or the random effects model. Heterogeneity was measured in our meta-analysis using Cochran's Q and the inconsistency index (I 2 ) statistic [59]. Funnel plots were used to detect whether there were obvious publication bias among the involved studies. An I 2 value of equal to or greater than 50% indicates a substantial heterogeneity among the studies in the meta-analysis that used a random-effect model for the analysis. For I 2 value less than 50%, a fixed-effect model will be applied for the meta-analysis. The combined ORs and the corresponding 95% CIs were calculated using the fixed-effect model or the randomeffect model if I 2 is less than 50%. P values less than 0.05 were considered to be significant.

Results
As shown in Figure 1, our initial search for the genetic studies of diabetes retrieved 6,452 articles from PubMed, Embase, Web of Science, CNKI and Wanfang from 2000 to 2012. Among them, 4,021 studies were involved with genes reported in previous metaanalyses and thus discarded for further analysis. A total of 504 articles were again filtered out because they failed to accumulate at  (Figures 2 and 3, and Table 3), including rs11594656 of IL2RA gene (I 2 = 0%), rs12150220 of NLRP1 gene (I 2 = 0%), 1188A/C of IL12B gene (I 2 = 0%), 21131T/C of APOA5 gene (I 2 = 1%), rs2335491 of SLC2A10 gene (I 2 = 0%), rs2070150 of ATF6 (I 2 = 0%), and rs35927125 of KLF11 gene (I 2 = 7%). No visual bias was showed in the meta-analyses of these 7 SNPs ( Figure 4 and Table 3). Our data also demonstrated a significant heterogeneity of the rest 3 SNPs that comprise rs725613 of CLEC16A gene (I 2 = 69%), rs2275703 of CASQ1 gene (I 2 = 65%), and C825T of GNB3 gene (I 2 = 82%). Therefore random-effect tests were applied for the meta-analyses of the above 3 SNPs. Their funnel plots were demonstrated in Figure 4 and no visual bias was observed for the 3 meta-analyses.

Discussion
In the present study, a comprehensive systematic overview of genetic association studies was performed for the susceptibility of T1D and T2D. We scrutinized all the candidate case-control studies to identify the eligible SNPs with at least three independent datasets. Our meta-analyses of 10 polymorphisms showed significant evidence for 3 T1D-associated SNPs (NLRP1 rs12150220, IL2RA rs11594656, and CLEC16A rs725613) and 1 T2D-associated SNP (APOA5 21131T/C). Our meta-analyses were unable to find significant associations of the rest 6 SNPs with T1D or T2D. Moreover, power analysis showed that there might be a lack of power for the meta-analysis of SLC2A10 rs2335491 (1,455 cases and 1,083 controls, 39%) under a moderate risk of diabetes (OR = 1.2). These might partly explain our failure to observe significant results for the meta-analyses of some polymorphisms.
After Bonferroni correction, only the association of SNP rs11594656 with T2D remains significant. However, false discovery rate (FDR) test, a less conservative correction for multiple hypothesis testing, shows that the q values are 5.11E-5 for IL2RA rs11594656, 0.051 for NLRP1 rs12150220, 0.026 for CLEC16A rs725613, and 0.051 for APOA5 21131T/C. This  suggests the robustness of our positive results in the meta-analyses, although we can't exclude a chance of false positive results for NLRP1 rs12150220, CLEC16A rs725613 and APOA5 21131T/C. Sensitivity analysis demonstrated there were no significant differences of four significant genetic variants after exclusion, suggesting that the results of our meta-analyses was robust. In addition, we have performed a comprehensive analysis for the Fst values of the involved SNPs. Our results show there are moderate ethnic differences for ATF6 rs2070150 (Fst = 0.13), although there are minimal heterogeneity among the involved studies from different ethnic groups (I 2 = 0). And KLF11A rs35927125 is monomorphic in Asians, however, its minor allele frequency in Caucasian populations ranges from 8.8-12.2% (Fst = 0.0232). On the contrary, there were little ethnic difference for CLEC16A rs725613 and GNB3 rs5443 (Fst ,0.1), although there exist large heterogeneity in the involved studies (I 2 .60%). For CASQ1 rs2275703, the heterogeneity might come from the discrepancies of the samples in the case-control studies.
Pancreatic b-cell inflammation and apoptosis plays a pivotal role in the pathogenesis of T1D [60]. As a member of the Ced-4 family of apoptosis proteins, NLRP1 is an important mediator of programmed cell death [61]. NLRP1 plays a pivotal role in the pathogenesis of some inflammatory diseases [62,63]. In the present research, we combined three independent datasets and performed a meta-analysis to evaluate the association between NLRP1 rs12150220 polymorphism and T1D susceptibility. Although large ethnic differences of allele frequency were found for NLRP1 rs12150220 (T allele frequency: 47-53.1% in Caucasians versus 66.4% in Brazilians), minimal heterogeneity was observed in the meta-analysis of this polymorphism (I 2 = 0%). Our results support NLRP1 rs12150220 as a protective genetic factor of T1D and  Table 3. Additional characteristics of individual T1D and T2D studies in the meta-analyses. s provide hints to clarify the mechanistic role of NLRP1 gene in the pathogenesis of T1D. Evidence from both genetic and animal model studies has shown a crucial role of IL-2/IL-2RA in the pathogenesis of T1D [37,[64][65][66][67]. IL-2/IL-2RA regulates CD4 + CD25 + regulatory T cells so as to maintain immune homeostasis [67]. IL-2RA rs12722495 was shown to contribute to the risk of T1D by lowering IL-2 signaling and diminishing the function of CD4 + CD25 + regulatory T cells [64]. Interestingly, there is a significant association of IL-2RA rs11594656 as a protective factor with the risk of T1D in Polish population [22]. These two SNPs were 24.726 kb away and not in the same linkage disequilibrium block. Expansion of CD4 + CD25 + FOXP3 + regulatory T cells through maternal insulin treatment was shown to reduce the risk of T1D in children [68]. Increased resistance to CD4 + CD25hi regulatory T cell-mediated suppression was showed in T1D patients [69]. Our meta-analysis established a significant association between IL-2RA rs11594656 polymorphism and T1D, although the influence of rs11594656 polymorphism on the regulation of IL-2RA gene remains to be unveiled in the future.
CLEC16A gene is located in the major histocompatibility complex class II region (16p13), and it encodes a member of the C-type lectin domain containing family. CLEC16A gene variants were associated with multiple autoimmune diseases such as T1D [38,[70][71][72]. CLEC16A gene variants were shown to be associated with the alternative splicing event in the CLEC16A transcription [73]. Our results suggested a significant association between CLEC16A rs725613 and T1D among 5,074 samples from Caucasian and Chinese populations (P = 0.01). A significant heterogeneity (I 2 = 69%) among these ethnic samples warranted a replication in additional populations.
High level of glucose was shown to induce expression of APOA5 [74,75] that is an efficient regulator of plasma triglycerides (TGs) by enhancing the catabolism of TG-rich lipoproteins [76] and prohibiting the transportation of TGs [77]. APOA5 could probably play a role in the pathogenesis of T2D by regulating the cholesterol homeostasis [44,76]. APOA5 gene variants were also reported to be associated with the lipid levels [50,78,79] and the risk of coronary heart disease [80,81] in T2D patients. Since both environmental factors [82,83] and other genes [84] interact with APOA5 gene, our significant observation for APOA5 21131T/C polymorphism may only partly explain the risk of T2D. Minimal heterogeneity among the involved studies in our meta-analysis (I 2 = 1%), however, along with previous results [85,86], in which ethnic differences were observed for the T allele frequency of APOA5 21131T/C among the studies in our meta-analysis (68.8-76% in Chinese versus 93.7% in Caucasians).
There are some limitations in the present study. Firstly, publishing bias might exist in this meta-analysis. Case-control studies with a lack of significant results were much harder to be published than those with positive findings. In addition, only publications in English and Chinese were included in the current meta-analyses. All these may distort the results in our metaanalyses. Secondly, some of the involved case-control studies [22][23][24]26,28,31,34,35,[37][38][39]41,[43][44][45][46]49,[51][52][53][54][55][56][57] didn't provide information on the exclusion of other diseases (such as coronary artery diseases, hypertension, and etc.) during recruitment. Thirdly, the effects of genetic factors on diabetes risk were confounded by other phenotypic parameters such as body mass index. Therefore, case-control studies with better design are warranted to avoid these confounding factors and replicate our findings in future. Fourthly, due to a lack of enough independent datasets, subgroup analysis and meta-regression were not applied to identify differences in effect and sources of heterogeneity. Lastly, our meta-analysis focused on gene loci with at least three independent studies, and this might prevent those gene loci in two large scale case-control studies from being included in the current meta-analysis.
In conclusion, we identify significant associations between 4 SNPs (NLRP1 rs12150220, IL2RA rs11594656, CLEC16A rs725613 and APOA5 21131T/C) and diabetes. Meta-analysis among 4,456 samples has confirmed that rs12150220 of NLRP1 gene is a risk factor of T1D in Caucasian and Brazilian populations. Meta-analysis among 37,432 samples has confirmed that rs11594656 of IL2RA gene is a risk factor of T1D in Caucasian and Japanese populations. Meta-analysis among 5,074 samples has confirmed that rs725613 of CLEC16A gene is a risk factor of T1D in Caucasian and Chinese populations. Another meta-analysis among 4,005 samples indicates that 21131T/C of APOA5 gene is a risk factor of T1D/T2D in Chinese and Caucasian populations.