Meta-analyses of IL1A polymorphisms and the risk of several autoimmune diseases published in databases

Background Based on published data, we aimed to quantitatively elucidate the possible genetic influence of rs17561 G/T and rs1800587 C/T polymorphisms of the IL1A (interleukin 1 alpha) gene in the susceptibility to several autoimmune diseases. Methods A series of meta-analyses were carried out. After database searching, we utilized our inclusion/exclusion criteria to screen and include the eligible studies. Passociation (P value of association test), Bonferroni-corrected Passociation value; false discovery rate (FDR)-corrected Passociation, ORs (odd ratios), and 95% CI (confidence interval) were generated to assess the magnitudes of genetic relationships. Results A total of 35 eligible articles were included. Pooled analysis data of both rs17561 G/T and rs1800587 C/T in the overall population indicated a negative association between cases of autoimmune diseases and negative controls (all Passociation>0.05, Bonferroni-corrected Passociation>0.05, FDR-corrected Passociation>0.05). Similar results were found in most subgroup analyses (all Passociation>0.05, Bonferroni-corrected Passociation>0.05, FDR-corrected Passociation>0.05), apart from the rs1800587 in the Graves’ disease subgroup, which showed an increased risk in some cases, compared with controls, under the models of allele T vs. C, carrier T vs. C, CT+TT vs. CC, and CT vs. CC (all Passociation<0.05, Bonferroni-corrected Passociation<0.05, FDR-corrected Passociation>0.05, OR>1). Conclusion Based on the available data, C/T genotype of the rs1800587 polymorphism within IL1A gene may be associated with an increased Graves’ disease risk. We did not see evidence regarding a positive role for rs1800587 or rs17561 in the risk of other autoimmune diseases, such as systemic sclerosis or rheumatoid arthritis. These conclusions still merit further data support and molecular exploration.


Introduction
Human autoimmune diseases are a group of pathologies that cause clinical damage or destruction of body tissue due to an immune response to its own antigens [1,2]. There are many types of autoimmune diseases, such as SSC (systemic sclerosis), JIA (juvenile idiopathic arthritis), BD (Behcet's disease), RA (rheumatoid arthritis), MS (multiple sclerosis), GD (Graves' disease), SLE (systemic lupus erythematosus), and TID (type 1 diabetes) [1,2]. A few cytokine genes have been reported to be linked to the autoimmune disease [2][3][4].
Several meta-analyses have reported an association between IL1A rs17561, rs1800587 polymorphisms and the presence of various autoimmune diseases, including systemic lupus erythematosus [16,17], rheumatoid arthritis [18], multiple sclerosis [19] and Graves' disease [20]. However, the genetic relationship between IL1A SNPs and the risk of other autoimmune diseases, including systemic sclerosis and type 1 diabetes, has not been reported. In the present study, we probed the genetic role of IL1A gene SNPs rs17561 and rs1800587 in the risk of autoimmune diseases using quantitative synthesis of overall meta-analysis followed by subgroup analyses.

Methods
The meta-analysis was conducted per the PRISMA (preferred reporting items for systematic reviews and meta-analyses) guidelines [21]. S1 File illustrates the meta-analysis on genetic association studies checklist, and S2 File shows the PRISMA 2009 checklist.

Database searching
We obtained potentially suitable articles by systematically searching three databases (up to April 2018): PubMed, WOS (Web of Science), and Embase (Excerpta Medica Database). The search terms were shown in S3 File.

Article screening
The following screening items were used to exclude publications: duplicates, reviews, letters, meta-analysis, abstracts or posters, and studies with unrelated data. Each study should have investigated an association between IL1A gene polymorphisms and autoimmune disease risk. The genotype frequency data could be extracted from both case and control groups. We also performed a chi-square-based Q-test to confirm that the genotype distribution of control group was consistent with HWE (Hardy-Weinberg Equilibrium).

Data extraction
Detailed data, including the first author name, publication year, SNP, disease type, genotype frequency, genotyping assay, and ethnicity, were extracted and summarized independently. Conflicting data were discussed with all authors, and missing data were requested by e-mail. We also used the Newcastle-Ottawa Scale (NOS) system to assess the study quality and generate an NOS score. An NOS score < 5 means the study was poor quality, and such studies were excluded.

Statistical association analysis
Stata/SE 12.0 software (StataCorp, USA) was used. To evaluate the strength of genetic relationships, P association , pooled ORs (odd ratios), and 95% CI (confidence interval) were generated referring to relevant publications [22][23][24][25][26]. The P association value was then adjusted by the Bonferroni and false discovery rate (FDR) correction method [27], using R software version 3.4.3. Bonferroni and FDR-corrected P association <0.05 from the association test was considered statistically significant. Six comparison models were utilized: allele T vs. G for rs17561, allele T vs. C for rs1800587 (allele); carrier T vs. G for rs17561, carrier T vs. C for rs1800587 (carrier); TT vs. GG for rs17561, TT vs. CC for rs1800587 (homozygote); GT vs. GG for rs17561, CT vs. CC for rs1800587 (heterozygote); GT+TT vs. GG for rs17561, CT+TT vs.CC for rs1800587 (dominant); TT vs. GG+GT for rs17561, and TT vs. CC+CT for rs1800587 (recessive). We also performed the subgroup analyses according to the characteristics of ethnicity, disease type, and control source.   Abtahi, 2015  rs1800587  SSc  82  72  16  98  98  21  PB  PCR-SSP  seven  Asian   Aggarwal, 2012  rs1800587  JIA  42  47  5  93  78  14  PB  PCR-RFLP  seven  Asian   Akman, 2008  rs1800587  BD  32  17  4  19  22  7  Q statistics with P heterogeneity (P value of heterogeneity) and I 2 tests with I 2 values were conducted to assess heterogeneity among the studies. When P heterogeneity was >0.05 and the I 2 value was <50%, the absence of high heterogeneity was inferred, and a fixed-effects model (Mantel-Haenszel method) was applied. Otherwise, a random-effects model (DerSimonian and Laird method) was utilized.

Sensitivity analysis and bias evaluation
We performed sensitivity analysis to test whether the pooled results were stable. In sensitivity analysis, the effect of each study on the pooled ORs was evaluated as each included study was excluded one-by-one. We also performed Begg's test and Egger's test to evaluate publication bias. P values of Begg's test and Egger's test, namely P Begg and P Egger , below 0.05 indicate the absence of publication bias.

Study characteristics
As shown in Fig 1, we searched three databases, identified a total of 240 articles [PubMed (n = 53), WOS (n = 81), Embase (n = 106)], and subsequently removed 45 duplicate articles. Then, 150 articles were excluded by our screening criteria. Assessing the eligibility of the remaining 45 articles, ten articles were removed, because seven did not contain complete genotype data and three were not consistent with HWE. Eventually, a total of 35 articles [8-15, 20, 28-53] were included, and none exhibited poor quality (all NOS score > 5). We list the characteristics of these studies in Table 1.

Meta-analysis of rs17561
Eleven case-control studies with 2,561 cases and 2,099 controls were enrolled for the meta-analysis of the IL1A rs17561 G/T polymorphism. As shown in Table 2, compared with controls, no increased risk was detected in any of the cases under six comparison models, including allele T vs. G [P association (P value in test of association) = 0.576, Bonferroni-corrected P association = 1.000, FDR-corrected P association = 0.703]; carrier T vs. G (P association = 0.586, Bonferroni-corrected P association = 1.000, FDR-corrected P association = 0.703); TT vs. GG (P association = 0.909, Bonferroni-corrected P association = 1.000, FDR-corrected P association = 0.909); GT vs. GG (P association = 0.419, Bonferroni-corrected P association = 1.000, FDR-corrected P association = 0.703); GT+TT vs. GG (P association = 0.438, Bonferroni-corrected P association = 1.000, FDR-corrected P association = 0.703); TT vs. GG+GT (P association = 0.043, Bonferroni-corrected P association = 1.000, FDR-corrected P association = 0.258). Forest plot data of the meta-analyses under different models are provided in Fig 2 and S1-S5 Figs. We also performed subgroup analyses by ethnicity and disease types. Similar negative results were obtained under different comparison models (all P association >0.05, Bonferroni-corrected P association >0.05, FDR-corrected P association >0.05, Table 3), except for the Asian (P association = 0.024, Bonferronicorrected P association = 0.144, FDR-corrected P association = 0.048) and PB (P association = 0.043, Bonferroni-corrected P association = 0.258, FDR-corrected P association = 0.043) subgroups under the TT vs. GG+GT model. These data suggested that the IL1A rs17561 G/T polymorphism seems not be related to a risk for autoimmune disease overall.

Meta-analysis of rs1800587
A total of 31 case-control studies with 7,381 cases and 4,049 controls were used for meta-analysis of the IL1A rs1800587 C/T Polymorphisms. Pooled data from the overall population (  data, the C/T genotype of IL1A rs1800587 C/T polymorphism is more likely to be statistically associated with an increased risk of Graves' disease, but not other autoimmune diseases, such as systemic sclerosis, juvenile idiopathic arthritis, rheumatoid arthritis, multiple sclerosis and systemic lupus erythematosus.

Heterogeneity, bias and sensitivity
Apart from the TT vs. GG+GT comparison of rs17561, larger heterogeneity was detected ( Table 2, I 2 >50.0% or P heterogeneity >0.05), and random effect models were utilized. In addition, as shown in Table 2

Discussion
Previously, the rs1800587 C/T SNP of IL1A gene was reported to not be linked to the risk or severity of systemic lupus erythematosus in a Spanish population [12], juvenile idiopathic arthritis in an Iranian population [15], and juvenile idiopathic arthritis in the UK [13]. IL1A rs17561 SNP was not associated with rheumatoid arthritis susceptibility in a Mexican population [14]. However, the IL1A rs1800587 and rs17561 SNPs were also reported to be associated with the risk of systemic sclerosis in a Japanese population [8]. The rs1800587 C/T SNP of IL1A gene has been related to susceptibility to systemic sclerosis in a Slovak Caucasian population [9], Graves' ophthalmopathy in an Iranian population [10], and Graves' disease in a Tunisian population [11]. Therefore, we first comprehensively explored the association between IL1A rs17561 and rs1800587 SNPs and the risk of overall autoimmune diseases using metaanalysis and subgroup analyses by characteristics of ethnicity, disease type and source of control.
In 2013, a meta-analysis was reported [17], investigating the genetic relationship between IL1A rs1800587 and rs17561 SNPs and the risk of systemic lupus erythematosus based on four case-control studies from three articles [12,42,48], which did not provide strong evidence for an association. In 2014, data from another meta-analysis containing four studies from three articles [12,48,51] supported a potential association for rs1800587 in Europeans [16]. In this  study, we added another case-control study [53] to the subgroup meta-analysis of systemic lupus erythematosus for rs1800587, and observed a negative association.
In one meta-analysis of rheumatoid arthritis susceptibility [18], there were four case-control studies [32,35,38,54] for rs1800587 and three case-control studies [34,39,43] for rs17561. No positive association between IL1A rs1800587 and rs17561 SNPs and the risk of rheumatoid arthritis was observed [18]. Here, we included more data for our updated meta-analysis and removed one case-control study [54], in which the genotype distribution of control group did not fulfill Hardy-Weinberg equilibrium. Seven case-control studies [14,34,36,38,39,41,43] were enrolled for the subgroup analysis of rs17561, and six case-control studies [14,32,35,36,38,46] were used for rs1800587. Our pooled data with enhanced statistical power also indicated that the IL1A rs1800587 and rs17561 SNPs were not linked to the risk of rheumatoid arthritis, which was consistent with previous data [18].  [33,37,44,50,55] for a meta-analysis of rs1800587 SNP and two case-control studies [47,56] for meta-analysis of rs17561 SNP. However, negative association was reported for both s1800587 and rs17561 [19]. Here, due to the limitation of Hardy-Weinberg equilibrium, one case-control study [55] was excluded from our subgroup meta-analysis of rs1800587. We also found that the rs1800587 SNP was not linked to the risk of multiple sclerosis.
In 2010, Liu et al. investigated the genetic relationship between IL1A rs1800587 SNP and risk of Graves' disease via a meta-analysis and found a positive association in an Asian population [20]. Here, our data in the subgroup meta-analysis of Graves' disease showed similar results. It is possible that the rs1800587 SNP within the 5'-flanking regulatory region of IL1A gene affects the normal production, secretion or function of interleukin-1.
Some limitations exist in our meta-analysis. First, we did not obtain strong evidence regarding the effect of rs1800587 and rs17561 SNPs for the risk of different types of autoimmune diseases, due to the limited number of included independent case-control studies. Only two casecontrol studies [30,40] were included in the subgroup of Graves' disease under the homozygote and recessive models. Second, even though no remarkable publication bias was detected by our Begg's test and Egger's test, larger heterogeneity existed in the majority of comparisons. We observed a decreased level of heterogeneity in some subgroup analyses by disease type, such as the "rheumatoid arthritis, RA" subgroup of rs17561 and "multiple sclerosis, MS" subgroup of rs1800587. The factor of specific disease type may be involved in the source of heterogeneity. Further relevant researches with larger sample sizes were required. Third, we only acquired suitable case-control studies published in English. The outcome may be affected by the inclusion of unpublished articles, or articles published in another language. Fourth, it is worth analyzing the combined influence of different SNPs or cytokine genes, when more casecontrol studies become available.
Taken together, based on published articles in databases, our meta-analysis suggested that the rs1800587 polymorphism, rather than rs17561, within the IL1A gene, may be a genetic risk factor for Graves' disease. However, IL1A rs17561 or rs1800587 polymorphism seems not to be statistically linked to the risk of other analyzed autoimmune diseases, such as systemic sclerosis, juvenile idiopathic arthritis, rheumatoid arthritis, multiple sclerosis and systemic lupus erythematosus.