Introduction

Ankylosing spondylitis (AS) is a chronic inflammatory disorder characterized by the inflammation of the spinal and sacroiliac joints, which initially causes bone and joint erosion and leads to ankylosis [1]. Strong genetic factors have been implicated in the etiology of AS. Human leukocyte antigen (HLA) B27 was the first genetic factor to be associated with AS, and it confers the greatest susceptibility to this disease. However, increasing evidence indicates that genes other than HLA-B27 also contribute to AS susceptibility [2].

Genome-wide association studies suggest that endoplasmic reticulum aminopeptidase 1 (ERAP1, a known aminopeptidase regulator of tumor necrosis factor receptor 1 [TNFR1] shedding), also known as ARTS1, is a potential candidate (for contributing to AS susceptibility) that is not an HLA gene [3]. ERAP1, which is located on chromosome 5q15, trims peptides for class I HLA presentation and cleaves cell-surface cytokine receptors. ERAP1 was recently found to be a key determinant of the repertoire of peptides that are presented by class I HLA molecules and initiate T-cell responses [4]. A genome-wide association study revealed that two single-nucleotide polymorphisms (SNPs) (rs27044, rs30187) of ERAP1 were significantly associated with AS, and demonstrated that rs17482078, rs10050860, and rs2287987 polymorphisms also trended toward association with AS [3]. Although some studies have demonstrated that SNPs of ERAP1 are associated with AS susceptibility, others disagree [5–23]. It would appear that allelic frequencies often differ substantially between populations, and ethnicity-specific association studies are needed to confirm genetic associations in different populations. Meta-analysis provides a powerful means of summarizing the results produced by different studies [24–26], and in the present study, we performed a meta-analysis to investigate whether ERAP1 polymorphisms are associated with susceptibility to AS in different ethnic populations.

Methods

Identification of eligible studies and data extraction

A search was performed using MEDLINE and EMBASE for studies that examined associations between the polymorphisms of ERAP1 and AS to identify articles in which the ERAP1 polymorphisms were determined in AS patients and controls (up to June 2014). Combinations of keywords such as “endoplasmic reticulum aminopeptidase 1,” “ERAP1,” “polymorphism,” “ankylosing spondylitis,” and “AS” were used as Medical Subject Heading (MeSH) and text words. References cited in the resulting studies were also investigated to identify additional studies not indexed by MEDLINE and EMBASE. Genetic association studies that determined the distributions of ERAP1 alleles in AS patients and controls were eligible for inclusion. Studies were included in the analysis if (1) they were case control studies, (2) contained original data, and (3) had sufficient data to calculate odds ratios (ORs). Studies were excluded if (1) they contained data obtained from other studies, (2) the number of alleles could not be ascertained, or (3) family members were included in the study, because such analyses are based on linkage considerations. The following information was extracted from each study: first author, year of publication, ethnic make-up of the study population, numbers of cases and controls, and the allele frequencies of the ERAP1 polymorphisms.

Evaluation of statistical associations

The allelic frequencies of the ERAP1 polymorphisms from the respective studies were determined using the allele counting method. We examined the difference between the effect of the minor allele, designated as 2, and that of the common allele, designated as 1, of each ERAP1 polymorphism. Point estimates of risk, ORs, and 95 % confidence intervals (CIs) were estimated for each study. Cochran’s Q statistic was used to assess variations or heterogeneities within each study and between studies [21]. This heterogeneity test was conducted with the null hypothesis that all the studies were analyzing the same effect. In addition, we quantified the effect of heterogeneity using I 2  = 100 % × (Q − df) / Q [22]. I 2 ranges between 0 and 100 % and represents the proportion of variability attributable to heterogeneity rather than chance between the studies. I 2 values of 25, 50, and 75 % were assigned as low, moderate, and high estimates, respectively. A fixed effects model assumes that the effects of individual genetic factors on AS susceptibility are similar for all the included studies and that observed variations between studies are caused by chance alone [23]. The random effects model assumes that different studies show substantial diversity and assesses both within-study sampling errors and between-study variances [24]. If study groups show no heterogeneity, the fixed and random effects models produce similar results, and if not, the random effects model usually produces wider CIs than does the fixed effects model. Accordingly, the random effects model was used to evaluate polymorphisms that had significant between-study heterogeneity. Statistical manipulations were performed using a Comprehensive Meta-Analysis computer program (Biostatics, Englewood, NJ, USA).

Results

Studies included in the meta-analysis

Electronic and manual literature searches resulted in the identification of 87 studies, and 22 of these studies were selected for a full-text review based on their titles and abstracts [9–16, 25–28]. Two studies were excluded because one used the data analyzed in another study and another did not have extractable allele data. Thus, 20 studies met the inclusion criteria (Table 1) [5–23, 27]. However, one of these studies contained data on three groups [13] and two studies had data on two different groups [17]. These groups were treated independently. Thus, for this meta-analysis, we considered 22 separate comparisons, which comprised, in total, 19,902 AS patients and 39,750 controls. Due to the limited number of studies on polymorphisms, 11 types of meta-analyses were performed (Table 2). Selected characteristics of the 21 studies are summarized in Table 1.

Table 1 Characteristics of the individual studies included in the meta-analysis
Full size table
Table 2 Meta-analysis of associations between ERAP1 polymorphisms and AS
Full size table

Meta-analysis of the association between rs30187, rs27044, rs10050860, rs2287987, rs17482078, and rs26653 polymorphisms and AS susceptibility

The meta-analysis identified a significant association between AS and the minor alleles of the rs30187 polymorphism in the overall population (OR = 1.255, 95 % CI = 1.147–1.373, P = 8.0 × 10−8) (Table 2, Fig. 1). Stratification by ethnicity led to the identification of a significant association between this polymorphism and AS in Europeans (OR = 1.283, 95 % CI = 1.237–1.331, P < 1.0 × 10−9), but not in East Asians and Middle Easterners (Table 2, Fig. 1). Meta-analyses of results of the studies on rs27044, rs10050860, rs2287987, rs17482078, and rs26653 polymorphisms showed that these polymorphisms had the same pattern as rs30187 in the overall population and in the European group. All of these polymorphisms were associated with AS susceptibility in Europeans (Table 2).

Fig. 1
figure 1

ORs and 95 % CIs of individual studies and pooled data for the minor allele versus the common allele of the ERAP1 rs30187 polymorphism for susceptibility to AS overall (a) and each ethnic group (b)

Full size image

Meta-analysis of the relationship between the rs27434, rs27037, rs27980, and rs27582 polymorphisms and AS susceptibility

Meta-analysis revealed a significant association between the minor alleles of the rs27434 and rs27037 polymorphisms and the susceptibility to AS in Europeans (OR = 1.214, 95 % CI = 1.131–1.304, P < 1.0 × 10−9; OR = 1.238, 95 % CI = 1.159–1.323, P < 1.0 × 10−9, respectively) (Table 2, Fig. 2). The rs27037 polymorphism was significantly associated with AS susceptibility in both European and Asian patients (Table 2). Meta-analysis showed a significant association between susceptibility to AS and the minor alleles of the rs27980 and rs27582 polymorphisms in East Asians (OR = 0.904, 95 % CI = 0.818–0.999, P = 0.047; OR = 0.871, 95 % CI = 0.772–0.982, P = 0.024, respectively) (Table 2). However, these polymorphisms had not been studied in Europeans.

Fig. 2
figure 2

ORs and 95 % CIs of individual studies and pooled data for the minor allele versus the common allele of the ERAP1 rs27037 polymorphism for susceptibility to AS in the overall population (a) and in each ethnic group (b)

Full size image

Meta-analysis of the relationship between rs27529 and AS susceptibility

No association was found between the rs27529 polymorphism and AS susceptibility by meta-analysis (OR = 1.007, 95 % CI = 0.644–1.573, P = 0.977) in the overall population (Table 2). Due to the limited number of available studies, stratification by ethnicity was not performed for this meta-analysis.

Heterogeneity and publication bias

Heterogeneity was found between the studies during the meta-analyses for the association between AS susceptibility and some ERAP1 polymorphisms. However, heterogeneity was not found during the meta-analyses of the rs10050860, rs27037, rs17482078, rs27980, and rs27582 polymorphisms. Publication bias causes a disproportionate number of positive studies and poses a problem for meta-analyses. However, we found no evidence of publication bias for all study subjects (Egger’s regression test P values >0.1).

Discussion

AS is strongly associated with HLA-B27, but only 1–5 % of HLA-B27-positive individuals develop AS, which suggests that other genes contribute to the risk of AS development [2]. A previous genome-wide association study revealed that ERAP1 polymorphisms are associated with AS at a high level of statistical significance [3], but more recent studies, in different ethnic groups, have produced mixed results [5–23].

In our previous meta-analysis, we found that the rs27044, rs17482078, rs10050860, rs30187, and rs2287987 polymorphisms of ERAP1 were associated with the development of AS in Europeans [28]. In the present study, we updated our meta-analysis on the association between ERAP1 polymorphisms and AS. We undertook an ethnicity-specific meta-analysis of the associations between the ERAP1 polymorphisms and AS susceptibility in European, East Asian, and Middle Eastern populations. Our meta-analysis revealed a significant association between AS and the rs30187 polymorphism in the group containing all study subjects and in the group containing Europeans. Meta-analyses of results for the rs27044, rs10050860, rs2287987, rs17482078, and rs26653 polymorphisms showed the same pattern as rs30187. Our findings support associations between AS susceptibility in East Asians and the ERAP1 polymorphisms rs27037, rs27980, and rs27582, but not the rs30187, rs27434, and rs27044 polymorphisms. The rs27037 polymorphisms were significantly associated with AS susceptibility in both Europeans and Asians. In addition, meta-analysis showed a significant association between AS and the rs27980 and rs27582 polymorphisms in East Asians. However, these polymorphisms had not been studied in Europeans. The minor alleles of the rs27980 and rs27582 polymorphisms in East Asians were protective for developing AS unlike other ERAP1 polymorphisms. The reason for the difference remains unclear but may be in part explained by different linkage disequilibrium with causative polymorphisms or gene of AS. These findings suggest that ERAP1 polymorphisms are associated with the development of AS in Europeans as well as Asians. The allelic frequencies of genes may often differ substantially between populations, and an association between the polymorphisms and a particular disease may depend on ethnicity. Therefore, ethnicity-specific association studies are required to confirm genetic associations in different populations. The ORs for all associations were a little above 1 and small, while the OR for HLA B27 and the susceptibility of AS has been known as very high. However, most associations of polymorphisms in complex diseases refer to small odds ratios [29]. Our finding suggests that ERAP1 is one of the associated genes with AS susceptibility, although the associations had small effect and were not strong as much as HLA-B27.

The role of ERAP1 in the pathogenesis of AS has yet to be fully understood. ERAP1 is known to be an important determinant of the repertoire of peptides that are presented by class I HLA molecules and, consequently, initiate T cell responses. ERAP1 encodes an endoplasmic reticulum aminopeptidase that is involved in trimming peptides to optimal lengths for class I HLA presentation [4]. Thus, ERAP1 variants that may affect the aminopeptidase function could perturb peptide presentation. AS is primarily an HLA class I-mediated autoimmune disease, and more than 90 % of patients carry the HLA-B27 gene. The association between ERAP1 and AS supports the arthritogenic peptide hypothesis, which suggests that the disease is triggered by the presentation of a peptidase, on HLA-B27 molecules at the surface of antigen-presenting cells, to cytotoxic CD8-positive cells. ERAP1 cleaves cell surface receptors for pro-inflammatory cytokines, which downregulates their signaling [30]. Thus, genetic variants that alter this function of ERAP1 could have proinflammatory effects. Furthermore, ERAP1 binds directly to the extracellular domain of TNFR1 and promotes the IL-1β-mediated cleavage of its ectodomain, to generate soluble TNFR1. This ERAP1-assisted generation of extracellular TNFR1 could be crucial for regulating the bioactivity of TNF, which plays a key role in the regulation of inflammation [30]. However, it is not known whether ERAP1 polymorphisms affect the clinical manifestations of AS. Further studies are required to determine the natures of the associations between ERAP1 polymorphisms and disease severity.

The present study has some limitations that should be considered. First, publication bias, heterogeneity, and confounding factors may have distorted the meta-analysis. Second, our ethnicity-specific meta-analysis included data mainly from European and East Asian populations, and thus, our results are applicable only to these ethnic groups. Third, the majority of these studies were performed on populations of European descent, and only two to four studies were conducted on East Asians, which might mean that our investigations using the East Asian group were underpowered. Fourth, the possibility of contributions to AS susceptibility by ERAP1 haplotypes also needs to be examined by meta-analysis. Unfortunately, in the present study, we could not conduct meta-analysis on AS susceptibility and ERAP1 haplotypes due to limited data. Fifth, it would have been interesting to determine if an association exists between ERAP1 polymorphisms and HLA-B27 status, HLA-B27 activity, or the clinical features of AS, but this was not possible because the available data was limited.

In conclusion, this meta-analysis of published data confirms that the ERAP1 polymorphisms are associated with AS susceptibility in Europeans and East Asians. The allelic frequencies of genes often differ substantially between populations, and thus, further ethnicity-specific association studies are required to confirm genetic associations with AS susceptibility in different populations. This meta-analysis demonstrates the need for further studies to determine whether ERAP1 polymorphisms contribute to AS susceptibility in various ethnic groups.