Introduction

Systemic lupus erythematosus (SLE) was a complex autoimmune disease characterized by pathogenic autoantibody production that could affect multiple organs [1]. The morbidity rate in SLE varied across different races, and it was considered that non-white racial groups were more frequently affected, for instance, 31–70/100,000 across China and 200/100,000 in African population, while it was 7–71/100,000 in Europeans [2,3,4]. Although the pathogenesis was not completely clear, it was generally known that genetic and environmental factors interacting with each other contribute to SLE risk [5]. Previous linkage analyses had identified some regions in genome for SLE, eg:1p13 [6], 1q25.1 [7], 4q24 [8], 19q13.2 [5]; however, linkage analysis could only identify large genomic regions harboring lots of susceptibility genes. The candidate gene association studies had identified some precise susceptibility genes related to SLE, involving PDCD1 [9], IRF-5 [1], FCGR2B [4], IL18 [10]. The genome-wide association studies (GWAS), which involved common variations through the entire genome, is a powerful tool to identify SLE-associated genes [7, 8, 11,12,13].

SLE-associated genes could be classified into three categories: B and T cell function-related genes [14], interferon (IFN) regulatory genes [15], and DNA repair genes [16]. A recent study had revealed that IFIH1 and TYK2 might induce the production of interferon [14].

The interferon-induced Helicase C domain 1 (IFIH1) gene, localized at chromosome 2p24.3, was a sensitive sensor of dsRNA that triggered IFN-α signaling pathway [14]. Elevated IFN-I was an obvious phenomenon in SLE patients, therefore, interferon regulatory genes were essential in the SLE pathological mechanism [17]. A recent study [18] indicated that IFIH1 rs1990760 was associated with an increased level of IFN-induced gene expression in SLE patients, in response to a given amount of serum IFN-α. IFIH1 had been identified to be associated with autoimmune diseases, such as, type I diabetes [19], Graves’ disease [20], and psoriasis [21]. Particularly, IFIH1 was identified as a susceptible gene for type I diabetes in a genome-wide association study [22].

TYK2 was located at chromosome 19p13.2, which was a part of JAK kinase, and phosphorylated and activated by combination of IFNAR and IFN-α [23] and then they signaled through the JAK-STAT (Janus kinase-signal transducer and activator of transcription) pathway to enhance expression of IFN-stimulated genes [24]. Several genetic studies were performed to analyze the relationship of TYK2 gene with SLE risk [1, 25,26,27,28,29], but the results were inconclusive. Lee et al. [30] found that TYK2 rs2304256 had a significant association with SLE in Europeans (P < 1 × 10−8), but not Asians. However, another study showed a significant association between SLE and TYK2 rs2304256 in Chinese (P = 1.85 × 10−5) [29].

IL-10 was not only a vital immunoregulatory cytokine produced by various immune cells, but also a gene related to B- and T-cells function [31]. IL-10 is located at the chromosome 1q31–1q32, which was a SLE linkage region [32]. The best-characterized signaling pathway of IL-10 was the JAK-STAT system, which could induce a series of immune responses using activated B- and T-cell [33]. Several studies have found an association between IL-10 and SLE in the past few years [31, 34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56], but the results were inconsistent.

A number of studies had tested the relationship between IFIH1, TYK2, IL-10 and SLE, but as we demonstrated above, the results were inconsistent. Therefore, we performed a comprehensive meta-analysis to evaluate the relationship of IFIH1 (rs1990760), TYK2 (rs2304256, rs280519, rs12720270), IL-10 (rs1800896, rs1800871, rs1800872) and SLE using different association models.

Materials and methods

Identification of eligible studies

To investigate the association between the three genes and SLE risk, we searched relevant articles in PubMed database and Web of Science and China National Knowledge Infrastructure (CNKI) as well as Cochrane and Embase before December 2017, using the following keywords: “TYK2”, “polymorphism”, “IFIH1” or “interferon-induced helicase C”, “IL-10” or “interleukin-10”, “SLE” or “systemic lupus erythematosus”.

Inclusion and exclusion criteria

The included studies were consistent with following conditions: (a) case/control study design; (b) providing SNP genotype distribution or allele frequency in the paper; (c) providing sample size in the paper; (d) patients were diagnosed through American College of Rheumatology criteria for SLE [57]. Accordingly, the exclusion criteria for the publications were as follows: (a) family study design; (b) repeated or overlapped publications; (c) literature reviews; (d) conference or meeting abstracts.

Data extraction

We collected information including authors, year of publishing, numbers of case and control, ethnicity, and numbers of genotype distribution. If there were no SNP genotype data, we calculated the numbers by the sample size and allele frequency.

Quality assessment

Study quality was evaluated systematically using the Newcastle–Ottawa Scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses. The scale assigns a score of zero to nine stars to each article, whereby a greater number of stars indicate a higher quality study. Using this ‘star system’, each included study was judged on three broad perspectives: the selection of the study groups; the comparability of the groups; and the ascertainment of outcome of interest. Apart from that, Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement was an guideline for reporting observational studies, we also have applied it to observational studies that we have included in the present study.

Statistical analysis

Pooled ORs and 95% confidence intervals (CIs) were used to evaluate the strength of association between polymorphism and SLE risk for every eligible study. Hardy–Weinberg equilibrium (HWE) was tested in control using Chi square test [58] and it was considered as HWE deviation when P value was less than 0.05 [59]. We performed HWE test for all included studies, and removed studies that were deviated from HWE in controls. Cochran’s Q statistics was used to evaluate within- and between-study heterogeneities. And the effect of heterogenicity can be calculated with the following formula: I2 = 100% × (Q − df)/Q [60]. I2 statistic was used to quantify the extent of heterogenicity, which ranges from 0 to 100%. Mostly, if I2 < 25%, this indicated that each study was homogeneous, then the fixed model could be applied. When 25% < I2 < 75%, the heterogenicity between studies could not be ignored. If I2 > 75%, it evinced that there was high heterogenicity between studies. Apart from I2 statistic, the P value was also important for test of homogeneity. In our meta-analysis, when Pheterogeneity > 0.1, the fixed model would be chosen, otherwise the random model was applied. We carried out sensitivity analysis to evaluate the effect of an individual study on pooled OR. Egger’s and Begg’s test were used to evaluate the publication bias [61]. The software program STAT12.0 was used in data analysis.

Result

Studies inclusion and characteristics

Figure 1 showed the literature searching process. 548 articles (15 for IFIH1, 26 for TYK2, 507 for IL-10) were found in PubMed database and CNKI, of which 500 studies were excluded by means of reading the title and abstract, therefore, 43 studies were included in our meta-analysis (7 for IFIH1, 7 for TYK2, 29 for IL-10). Characteristics of the included studies were summarized in Table 1. The Newcastle–Ottawa Scale (NOS) used for assessing the quality of studies was also shown in Table 1 and the scores ranged from 4 to 6. No article presented poor quality according to the Newcastle–Ottawa scale. The result of STROBE statement was presented in the TableS1. The STROBE checklist provided guidance on how to report observational research well, but not an instrument to evaluate the quality of observational research [62].

Fig. 1
figure 1

Flow diagram of literature search

Full size image
Table 1 The characteristics of studies included in our meta-analysis
Full size table

The association between IFIH1 gene polymorphism and SLE

After testing the HWE deviation, 7 studies including 9288 patients and 24,040 controls were included in the meta-analysis of IFIH1 rs1990760 polymorphism and SLE. Test of heterogeneity in overall population was not significant (Pheterogeneity = 0.184), therefore, the fixed model was applied. In the allelic test, rs1990760 showed a significant association with SLE in overall population (T vs C, OR 1.135, 95% CI 1.094–1.179, P = 2.56 × 10−11, Table 2; Fig. 2a), which reached genome-wide level of significance (P < 5 × 10−8). Stratification analysis by ethnicity demonstrated an association between IFIH1 rs1990760 and SLE in Europeans (T vs. C, OR 1.148, 95% CI 1.071–1.230, P = 9.23 × 10−5, Table 2; Fig. 2a), but not Asians (T vs. C, OR 1.140, 95% CI 0.972–1.336, P = 0.107, Table 2; Fig. 2a). We also tested the recessive and dominant model for the T allele, and the result showed significance in both dominant model and recessive model in overall population (CT + TT vs CC, OR 1.203, 95% CI 1.128–1.284, P = 1.8 × 10−8, Table 2, Fig S1a; TT vs CT + CC, OR 1.163, 95% CI 1.098–1.231, P = 2.6 × 10−7, Table 2, Fig S1b) and in Europeans (CT + TT vs CC, OR 1.233, 95% CI 1.074–1.416, P = 2.88 × 10−3 Table 2, Fig S1a; TT vs CT + CC, OR 1.178, 95% CI 1.070–1.297, P = 8.38 × 10−4, Table 2, Fig S1b). In Brazilians, no association was between the rs1990760 and SLE under allelic model (T vs. C, OR 1.041, 95% CI 0.832–1.302, P = 0.723, Table 2; Fig. 2a).

Table 2 Meta-analysis of the association between IFIH1 polymorphism and SLE risk
Full size table
Fig. 2
figure 2figure 2figure 2

a Forest plot for meta-analysis of IFIH1 rs1990760 and SLE risk (T vs. C). b Forest plot for meta-analysis of TYK2 rs2304256 and SLE risk (AC/CC vs. AA). c Forest plot for meta-analysis of IL-10 rs1800896 and SLE risk (GG vs. GA/AA). d Forest plot for meta-analysis of IL-10 rs1800896 and SLE risk (GG vs. AA)

Full size image

The association between TYK2 gene polymorphisms and SLE

After testing the HWE deviation, there were 4058 patients and 5985 controls (6 studies). TYK2 rs2304256_C allele increased SLE risk in Europeans under three models (C vs. A, OR 1.434, 95% CI 1.203–1.710, P = 5.82 × 10−5, Table 3, Fig.S2a), (CC + AC vs. AA, OR 1.599, 95% CI 1.264–2.022, P = 8.85 × 10−5, Table 3; Fig. 2b; CC vs. AC + AA, OR 1.512, 95% CI 1.233–1.855, P = 7.19 × 10−5, Table 3, Fig.S2b). When coming to overall population, TYK2 rs2304256_C polymorphism showed a significant association with SLE under the recessive model (CC vs AC + AA, OR 1.314, 95% CI 1.074–1.608, P = 8.05 × 10−3, Table 3, Fig.S2b) and a marginal association under the allelic model (C vs A, OR 1.294, 95% CI 1.054–1.589, P = 0.014 Table 3, Fig.S2a). However, the C allele of rs2304256 did not confer a significant risk for the development of SLE in Asians under any models in this meta-analysis (Table 3).

Table 3 Meta-analysis of the association between TYK2 polymorphism and SLE risk
Full size table

Totally, 1380 patients and 3274 controls (3 studies), and 1608 patients and 8183 controls (3 studies) were recruited for rs12720270 and rs280519 after testing the HWE deviation, respectively. As shown in Table 3, there was no association between rs12720270 and rs280519 polymorphism and SLE under any models in overall population. Stratification analyses by ethnicity were not performed for the two SNPs, because all studies for rs12720270 were from Asian population considering the Hardy–Weinberg equilibrium, and for rs280519, there were only two articles done on Asians and one on Europeans.

The association between IL-10 gene polymorphism and SLE

SNP rs1800896, rs1800871, rs1800872 were regarded as tags of promoter of IL-10 gene [78, 79]. We recruited 25 studies for IL-10 rs1800896 polymorphism after testing the HWE deviation, including 3332 patients and 4612 controls. There was a slight association in overall population of three models with rs1800896 (allelic model: Table 4, Fig S3a; recessive model: Table 4; Fig. 2c; dominant model: Table 4, Fig S3b). Under recessive model, the result showed that rs1800896 had a marginal association with Asians (GG vs GA + AA, OR 2.623, 95% CI 1.346–5.115, P = 4.65 × 10−3, Table 4; Fig. 2c), but not Europeans. In the dominant model, we found an association between rs1800896 polymorphism and SLE in Europeans (GG + GA vs AA, OR 1.375, 95% CI 1.072–1.764, P = 1.21 × 10−2, Table 4, Fig S3b), but not in Asians.

Table 4 Meta-analysis of the association between IL-10 polymorphism and SLE risk
Full size table

We also calculated homozygous OR for genotypes (GG vs. AA) (overall: OR 1.539, 95% CI 1.142–2.072, P = 4.65 × 10−3, Europeans: OR 1.499, 95% CI 1.066–2.106, P = 1.98 × 10−2, Asians: OR 3.032, 95% CI 1.325–6.938, P = 8.54 × 10−3, Fig. 2d), which revealed that GG homozygote may increase the SLE risk. Furthermore, we analyzed the heterozygous OR for genotype GA vs. AA (overall: OR 1.229, 95% CI 0.980–1.541, P = 0.075, Europeans: OR 1.343, 95% CI 1.044–1.727, P = 2.2 × 10−2, Asians: OR 1.285, 95% CI 0.857–1.926, P = 0.226, Fig S3c).

With respect to IL-10 rs1800871, 18 articles involving 1685 patients and 2500 controls were included. We included 23 studies for IL-10 rs1800872, with a total of 2520 patients and 3870 controls. In our meta-analysis, rs1800871 and rs1800872 were not associated with SLE in overall population under any model (Table 4).

Allele frequency of the 3 SNPs in meta-analysis and in the 1000 Genomes Phase3 population

In Table 5, we presented a clear distinction of allele frequencies in different ethnicities in our meta-analysis and 1000 Genomes of the 3 SNPs. In consideration of the sample size and population, the allele frequencies of IL-10 rs1800896, and TYK2 rs2304256 in this meta-analysis were consistent with the allele frequencies in 1000 Genome project AMR (Ad Mixed American), EUR (European ancestry), EAS (East Asian ancestry), respectively; however, there was discrepancy between the allele frequencies of IFIH1 rs1990760 and 1000 Genomes Project (Table 4).

Table 5 The allele frequency comparison between the meta-analysis and 1000 Genomes Project
Full size table

Publication bias and sensitivity analysis

Funnel plot and Bgger’s linear regression test has been used to detect the publication bias. We found no publication bias in our meta-analysis under allele genetic model (IFIH1 rs1990760, P = 0.858; TYK2 rs2304256, P = 0.260; TYK2 rs12720270, P = 0.296; TYK2 rs280519, P = 1; IL-10 rs1800896, P = 0.154; IL-10 rs1800871, P = 0.112; IL-10 rs1800872, P = 0.526, Fig.S4). We also performed sensitivity analysis to evaluate the influence of an individual study on pooled OR (Fig. S5), and the pooled OR was not substantially altered, when any one study was removed.

Discussion

JAK-STAT pathway was important for inflammatory conditions and autoimmune diseases including SLE [80], and the genes IFIH1, TYK2, IL-10 that were related to this pathway [17]. In the present meta-analysis, 43 studies were recruited to evaluate the association between IFIH1, TYK2, IL-10 and SLE risk. The result provided evidence that IFIH1 rs1990760, TYK2 rs2304256 and IL-10 rs1800896 were associated with SLE. In addition, ethnicity-specific meta-analysis indicated an association between the SNPs and SLE in Europeans but not in Asians; however, ethnicity-stratified results should be interpreted carefully.

As we all know, the progress of combination of IL-10 and its receptor could be divided into two steps. First, IL-10 binds to the IL-10R1. Then, the complex of IL-10/IL-10R1 changed the cytokine conformation through the molecular interaction. Under this condition, the complex could create a binding site which was used for combination of IL-10R2 [81,82,83]. Subsequently, two members of the Janus kinase family, Janus kinase 1 (on IL-10R1) and tyrosine kinase 2 (TYK2) (on IL-10R2) would bind with the IL-10/IL-10R complex, then creating docking sites for the transcription factor STAT3. Type-I IFN receptors which were found on all nucleated cells can be phosphorylated by TYK2, meanwhile, a report demonstrated that IFIH1 can increase IFN-α production [84].

IFIH1 rs1990760 was associated with increasing IFN-α level in SLE patients [14, 17]. And IFN-α binding to IFNAR would activate JAK-STAT pathway to induce corresponding gene transcription. There was one GWAS study [85] investigating SNP rs1990760 and SLE susceptibility in European, reaching the genome-wide significance (P = 4 × 10−8). Our study also illustrated that T allele of IFIH1 rs1990760 apparently increased the risk of SLE, but its effect was diverse in different ethnicities. We validated the association between rs1990760 and SLE in overall population and Europeans. These results were also consistent with the previous meta-analysis conducted by the Cen et al., which indicated that rs1990760 had a significant association with SLE risk in Europeans (overall: P = 3.18 × 10−5, European: P = 1.96 × 10−5) [86]. Simultaneously, it was worth mentioning that the study by Gateva et al. [34] was removed from our study due to inconsistency with Hardy–Weinberg equilibrium (P < 0.001), while it was included in the meta-analysis conducted by the Cen et al. [86]. In terms of the other meta-analysis produced by Silva [68], it also showed IFIH1 rs1990760 was associated with SLE (overall: P = 0.0266).

No association was found in our study performed on Asians; however, only two articles were included in the meta-analysis (1120 patients and 1246 controls), and further assessment is needed to clarify the relationship between rs1990760 polymorphism and SLE risk in Asians. Furthermore, the frequency of T allele is very different between Europeans and Asians in 1000 Genomes Phase3 (0.61 in European vs. 0.19 in Asians) (Table 5). Although the frequency of T allele was common in Asians, it was much lower than in Europeans. The discrepancy of T allele frequency might be the cause of the inconsistent result between the two populations.

Overproduction of the IL-10-receptor complex would lead to more autoantibodies that could damage organs and tissues [78]. Three SNPs (rs1800896, rs1800871, rs1800872) were found in the promoter region of IL-10 gene. An association was identified between IL-10 rs1800896 and SLE risk in overall population under allelic and recessive model. To be specific, the previous meta-analysis, including 1788 patients and 2673 controls by Zhou et al., found a trend of association between rs1800896 and SLE risk in the overall population under the allelic model (allelic: OR 1.125, P = 0.054) [32], after carefully comparing our study and Zhou et al., we found that we included four additional articles [70,71,72, 74] of Chinese dataset, and expanded the sample size by approximately 1.8 times, which might have had an effect on the results.

In the subgroup analysis, under the allelic model, we have indicated a suggestive association between rs1800896 and SLE risk in Europeans, but not in Asians. Our result was consistent with other studies [31, 41, 42, 47, 49, 56], which showed rs1800896_G allele did not confer the risk of SLE in Asian. For further exploring the relationship between the rs1800896 polymorphism and SLE risk in Asians, we assessed the association model. We suggested that GG homozygote might make a contribution to increasing the risk of SLE. Besides, it is noteworthy that the frequency of G allele has shown a distinct difference betweens Asian and Europeans in the 1000 Genomes Phase3 (0.052 in Asians vs. 0.453 in Europeans). The frequency of G allele in Asians was low, and a larger sample size was needed to investigate the association between rs1800896_G allele and SLE susceptibility in Asians.

In the previous studies, no association was found between the rs1800871 and SLE risk in Asians [41, 42, 47], which was consistent with our results. But, the study by Song et al. reported an trend of association between rs1800871 and SLE in Asians (P = 0.027) [87], after checking carefully, we found that two studies (Lu et al. [75] and Chong et al. [31]), that deviated from HWE, were excluded in our meta-analysis, furthermore, we included three newly published articles [48, 49, 56] in our analysis.

In the JAK-STAT pathway, the combination of IL-10 and IL-10R complex was very important, and it should be noted that TYK2 focused on the tetrameric IL-10 receptor, than creating docking sites for the transcription factor STAT3 [88]. We preformed a meta-analysis to identify the association between TYK2 rs2304256, rs12720270, rs280519 polymorphisms and SLE, respectively. In terms of TYK2 rs2304256, a genetic association between rs2304256 and SLE was significantly detected in Europeans [1, 25, 26, 28], but not in Asians [27, 69]. Simultaneously, the previous meta-analysis [30] also reported similar results. These phenomena were consistent with the recent GWAS study [85] which demonstrated a strong association between rs2304256 and SLE (P = 2.43 × 10−12) in Europeans. To identify the relationship between rs2304256 and SLE risk in different ethnicities, we conducted population stratification analyses (Europeans and Asians). A significant association was found in Europeans (P < 0.0001), but not in Asians. Though we found no association between the rs2304256 and SLE in Asians, the ethnicity-specific association should be interpreted cautiously because only three articles for Asian population were included in our study. According to the data from 1000 Genomes Phase3, the frequency of C allele was almost threefold than A allele in Europeans (C: 0.738, A: 0.262), while the frequency of the two alleles had no big difference in Asians. The discrepancy of frequency led to the difference of results; therefore, a more larger sample size is needed to investigate the relationship between rs2304256_C allele and SLE susceptibility in Asians. In our study, TYK2 rs12720270 and TKY2 rs280519 did not confer a significant risk for the development of SLE in any of the racial subgroups.

The current meta-analysis had some inevitable limitations to be taken into account. On the one hand, although our overall sample size was large, the participant number in the specific ethnic population was relatively small, with the smallest sample size of 738 patients and 2632 controls [43]. On the other hand, it was universally acknowledged that the mechanism of SLE was sophisticated. A more precise analysis could be performed because of potential confounding factors such as gene–gene and gene–environment interaction, and other life style factors.

In summary, IFIH1 rs1990760_T and TYK2 rs2304256_C alleles had a significant association with SLE susceptibility in Europeans, but not in Asians; however, the ethnicity-specific association should be interpreted cautiously because of the different sample sizes in different ethnicities. For IL-10 rs1800896, GG homozygous might contribute to SLE in Asians. Considering the limited samples in Africans and African-Americans in this analysis, further studies are still required in a larger number of samples and diverse ethnic populations.