Article Text
Abstract
Objective Spondyloarthritis is a group of articular disorders sharing a genetic background. Polymorphisms in the IL-1 gene cluster have previously been associated with ankylosing spondylitis (AS), a subset of spondyloarthritis. This study examined the association between several of these polymorphisms and the whole spondyloarthritis. Particular attention was devoted to genotype–phenotype correlations.
Methods Seven single-nucleotide polymorphisms (SNP) and a variable number tandem repeat located in the IL-1 gene cluster were genotyped in 185 independent spondyloarthritis trios. Family-based association test (FBAT) was computed using the FBAT software. Analysis was carried in spondyloarthritis as a whole and also in AS. A case–control replication study was performed for four of the SNP, in an independent sample of 414 spondyloarthritis and 264 controls. A combined analysis of both studies was performed.
Results The SNP rs2856836 in IL1A was significantly associated with spondyloarthritis (p=0.009) and AS (p=0.010) in the family study. The case–control study revealed an association between another IL1A variant (rs1894399) and AS (p=0.035), and between IL1F10.3 (rs3811058) and spondyloarthritis (p=0.041). By combining family and case–control studies an association between AS and IL1A was confirmed (rs1894399, p=0.024), whereas non-AS was more significantly associated with IL1F10.3 (p=0.0043). Family-based and case–control studies revealed significant association between the two most frequent haplotypes combining the four SNP of the replication study and both spondyloarthritis (p=0.0054 and p=0.038) and AS phenotypes (p=0.018 and 0.0036).
Conclusion This study is the first to demonstrate an association between several polymorphisms located in the IL-1 gene cluster and spondyloarthritis as a whole. The IL1A locus was strongly associated with AS phenotype, whereas IL1F10 was associated with non-AS.
Statistics from Altmetric.com
Spondyloarthritis is a major group of inflammatory joint disorders, sharing clinical features and genetic background.1 It is classically divided into subtypes according to the predominant clinical presentation: ankylosing spondylitis (AS), the prototypical form that requires advanced radiographic sacroiliitis as a diagnostic criterion; psoriatic arthritis; reactive arthritis; inflammatory bowel disease (IBD)-associated arthritis; and in the absence of any distinctive feature, undifferentiated spondyloarthritis. The human leucocyte antigen (HLA)-B27 allele is the most important genetic factor predisposing to spondyloarthritis. However, half of the genetic susceptibility to spondyloarthritis is predicted to depend on genetic factors lying outside of the major histocompatibility complex (MHC) region.1 2 Recent breakthrough in the identification of other genetic factors predisposing to spondyloarthritis came from several case–control genome-wide association studies (GWAS) that identified several coding and non-coding single-nucleotide polymorphisms (SNP) significantly associated with AS, in genes such as ARTS1/ERAP1, IL23R, IL1R2 and ANTXR2.3,–,5 Starting from a genome-wide linkage approach, our group has also recently identified a new SPA2 locus significantly linked to spondyloarthritis.6 Further studies of this locus allowed us to identify a strongly associated haplotype located near to the TNFSF15 gene on 9q32,7 and another significantly associated SNP (rs3181357) in the neighbour TNFSF8 gene.8
Among others susceptibility genes encoded outside the MHC locus, the interleukin (IL)-1 gene cluster is a 360 kb region on chromosome 2 (2q13), which was first identified as a putative susceptibility locus for AS in linkage study with a suggestive logarithm of odds score of 2.5.9 Several polymorphisms located in genes of the IL-1 cluster have previously been tested for their association with AS in case–control or family-based association studies, yielding either positive of negative results.10,–,13 Some studies have reported an association of the allele 2 of a variable number of tandem repeats (VNTR) located in intron 2 of the IL1RN gene or of other alleles located through this gene with AS,13 14 whereas another study reported negative results in a different population.10 Association with polymorphisms lying in other genes of the cluster, including IL1A, IL1B and IL1F10 have also been reported in several populations.11 15 16 Finally, a recent prospective meta-analysis of the IL-1 gene complex concluded that IL1A was the most strongly associated gene in the cluster. Noteworthy, however, this case–control study combined data from diverse ethnic groups with heterogeneous results, and none of the polymorphisms was consistently associated across groups.17 Most of the foregoing studies focused on the restricted AS phenotype. Only one of them analysed the whole spondyloarthritis group in a French population.10 It failed to reveal any association with the IL-1 gene cluster. However, its power was too limited to allow definitive conclusions. Anyway, it raised the possibility that association with the IL-1 cluster could apply to the AS subtype rather than the entire group of spondyloarthritis. This could be especially the case if IL-1 gene cluster polymorphisms were influencing the anatomical damage detected on sacroiliac joint radiographs and taken as a hallmark of AS for patient selection.
Considering the potential critical role of IL-1, along with other recently identified genes in disease susceptibility,18 we wished to search for an association between spondyloarthritis and the IL-1 gene family cluster, by combining both family-based and case–control approaches in an ethnically homogeneous French population. Our second objective was to analyse whether specific genetic associations were dependent on a particular phenotype (AS vs spondyloarthritis).
Patients and methods
Patients
One hundred eighty-five independent trios, each consisting of an affected child and both parents, were selected for the family part of this study from a collection of French families recruited by Groupe Français d’Étude Génétique des Spondylarthropathies, as previously described.19 In a case–control panel, independent French cases were recruited through the Rheumatology clinic of Ambroise Paré Hospital (Boulogne-Billancourt), or through the national self-help patients' organisation Association Française des Spondylarthritiques. Independent French controls were obtained from the Centre d’Etude du Polymorphisme Humain, or were recruited as healthy spouses of cases with no affected children.
The diagnosis of spondyloarthritis was made according to the classification criteria of Amor et al20 and/or the European Spondylarthropathy Study Group.21 All patients were interviewed, examined and had pelvic radiographs assessed by qualified rheumatologists. Within the group of spondyloarthritis, AS was diagnosed according to the modified New York criteria.22 Regarding extra-articular manifestations, the diagnosis of psoriasis required the presence of typical lesions and/or clinical diagnosis established by a dermatologist. The diagnosis of anterior uveitis required examination by an ophthalmologist. IBD diagnosis (including Crohn's disease and ulcerative colitis) was based on endoscopic and histological examination of the gut. Reactive arthritis was diagnosed according to the criteria published by Willkens et al.23 Finally, undifferentiated spondyloarthritis was diagnosed when spondyloarthritis criteria were fulfilled, without any of the foregoing diagnoses. Three affections traits were considered for association analyses: spondyloarthritis (all patients), AS and non-AS (all patients but AS).
This study was approved by the local ethics committee of Cochin Hospital (Paris, France) and of Ambroise Paré Hospital (Boulogne Billancourt, France). A written informed consent was obtained from each participant.
Genotyping of polymorphisms
Seven SNP through the IL-1 gene cluster were selected based on previously reported positive association with AS,12 17 24 and genotyped using melting curve analyses (LightCycler System, Roche, Meylan, France). Both PCR primers and hybridisation probes were synthetised by Tib MolBiol (Berlin, Germany). The Ensembl database (http://www.ensembl.org/index.html) was used for annotation and positioning of the polymorphisms (table 1).
The variable copy number of a 86-bp repeat sequence within the IL1RN gene intron 2 determines the polymorphism resulting in five distinctive alleles. Alleles of the IL1RN 86-bp VNTR were characterised following PCR amplification and separation by electrophoresis on 2% agarose gels. PCR was performed using five units of Taq DNA polymerase (Beckman Coulter, Villepinte, France) in the following buffer: 1.65 mM magnesium chloride, 9.9 mM Tris HCl, 50 mM potassium chloride and 0.1% Triton ×100, 0.03125 mM of each nucleotide, 40 ng of DNA template and 25 pmol of each primer: 5′-CTCAGCAACACTCCTAT-3′ (forward primer) and 5′-TCCTGGTCTGCAGGTAA-3′ (reverse primer). PCR conditions were 96°C for 1 min and then 30 cycles of denaturing at 94°C for 1 min, annealing at 60°C for 1 min, followed by a final elongation time of 2 min at 70°C. PCR products were analysed on a 2% Seakem agarose gel (FMC, Le Perray en Yvelines, France). Five distinctive alleles were detected corresponding to two, three, four, five and six copies of the 86-bp sequence.
To replicate our family-based association results in an independent case–control study, we used a French collection of 414 independent cases and 264 controls. We genotyped a set of four SNP, representing the most associated markers in the foregoing study, using a SNPlex genotyping method, corresponding of a production-scale 48-plex assay (Applied Biosystems, Courtaboeuf, France). Automatic allele assignment was achieved with the Gene-Mapper software v4.0 (Applied Biosystems), with the rules-clustering algorithm. Three of the SNP, rs2856836, rs17561, rs1894399, which were in strong linkage disequilibrium (LD) with each other, were located in the IL1A gene, whereas rs3811058 was located in the IL1F10 gene.
Statistical analysis
All genotypes were screened for Mendelian inheritance using the family-based association testing (FBAT) package v2.0.0c and PedCheck version 1.1. The departure from Hardy–Weinberg equilibrium in unrelated individuals was assessed using Pedstats version 0.6.11. We used an additive model of FBAT to compute the statistical test, and gave a p value corresponding to this test. Two affection traits were considered successively: spondyloarthritis and AS. Regarding other subphenotypes (psoriais, uveitis, IBD), our dataset had not sufficient power to allow meaningful comparisons.
Association in the case–control study was assessed using a χ2 test. As the family and the case–control replication samples were independent of each other, it made sense to perform an association analysis of the pooled sample. This test was performed using the Cochran–Mantel–Haenszel method.25 Before combining case–control and family studies, genetic homogeneity of the population was assessed for each SNP using the heterogeneity Q test.26 27 Haplotype-specific association was assessed for preselected SNP. FBAT of haplotypes was performed with Haplotype-based association testwhich is an elaboration of the FBAT programme. In the case–control sample, the same haplotypes were analysed with the ‘hap-assoc’ option of PLINK.
As our study was a replication of already published associations and many marker pairs were in high LD, a Bonferroni correction for multiple testing could be overly conservative. Indeed, the Bonferroni correction is a method proposed to account for the problem of multiple comparisons whenever several tests are performed in a single study, by dividing the crude significance level to be reached by the number of tests performed. The underlying assumptions are that the null hypothesis of ‘no association’ is a priori the most likely to be true and that all the tests are independent of each other. It comes at the cost of increasing the probability of producing false negatives, and then reducing statistical power. All p values are shown without correction for multiple comparisons and a p value less than 0.05 was considered statistically significant.
Measures of LD were determined using Haploview. This program uses an expectation-maximisation algorithm to calculate the maximum likelihood estimates of haplotype frequencies, given genotype measurement.
Results
Association between IL-1 gene cluster polymorphisms and spondyloarthritis
In the family-based association study, 185 spondyloarthritis trios were genotyped for seven SNP (table 1) and a pentallelic VNTR in the IL-1 cluster region (data not shown). The 185 spondyloarthritis patients consisted of 115 (62.8%) men and 70 (37.8%) women, of whom 93% were HLA-B27 positive. Axial clinical manifestations were present in 97% of the cases, and 57% of them fulfilled AS criteria. A past or present history of peripheral arthritis, psoriasis, uveitis and IBD existed for 46%, 26%, 23% and 5% of the cases, respectively.
The results of the association test under the FBAT allelic additive model between each individual polymorphism and spondyloarthritis or AS are shown in table 2. Regarding the whole group of spondyloarthritis, there was a significant association with IL1A SNP rs2856836 (p=0.009) and with IL1F10 SNP rs3811058 (p=0.04). There was a non-significant trend for association with both other SNP located within IL1A (table 2). None of the three other SNP located either in IL1B or in IL1RN showed any evidence of association (table 2); neither did the IL1RN-VNTR analysis exhibit evidence for linkage or association with spondyloarthritis (data not shown).
In the AS subpopulation, there was also a significant association with IL1A SNP rs2856836 (p=0.01).
Case–control replication study
To expand the foregoing positive results, we genotyped the four most relevant polymorphisms (three located in IL1A and one in IL1F10) in an independent French case–control cohort (table 3). The 414 spondyloarthritis cases consisted of 200 (48.3%) men and 214 (51.7%) women, of whom 79% were HLA-B27 positive. Axial clinical manifestations were present in 95.6% of the cases and 51% of them fulfilled AS criteria. A past or present history of peripheral arthritis, psoriasis, uveitis and IBD existed in 38%, 29%, 25% and 7% of the cases, respectively. We observed a significant association with IL1F10.3 (rs3811058) in the whole spondyloarthritis population (p=0.041). Association with this SNP was even more significant in the ‘non-AS’ patients (p=0.023), whereas it was not significant in the AS subgroup (table 3). Conversely, a significant association was detected in AS patients with IL1A SNP rs1894399 (p=0.035), whereas no individual SNP located in IL1A was significantly associated with spondyloarthritis as a whole (table 3). The combined analysis of family-based and case–control studies reinforced the significance of association between spondyloarthritis and IL1F10.3 (rs3811058) (p=0.0043), and showed a significant association of one IL1A SNP (rs1894399) with both the spondyloarthritis (p=0.012) and AS (p=0.026) subphenotype (table 4). The heterogeneity test was significant for SNP rs2856836 in the whole spondyloarthritis population (table 4; Q=0.016), thus preventing interpretation of the combined results for this SNP, which was found to be associated with AS (p=0.024).
Haplotypic study
When the causal variant is not identified, haplotype-based analysis could be more powerful for association studies in which there is LD in the region of interest. Therefore, the four SNP genotyped in family-based and case–control studies were combined in a haplotypic association analysis (table 5). The four most frequent haplotypes were generated, accounting for 96.5% of all haplotypes. Of most interest, the most frequent H1 haplotype (CTAT) was significantly associated with spondyloarthritis (p=0.0054) as well as AS (p=0.018) in a family-based study, whereas the second most frequent haplotype (TGGT) was found to be associated in case–control studies with both spondyloarthritis (p=0.0383) and AS (p=0.0036) (table 5).
Discussion
The present study is the first to report an association between the IL-1 gene cluster and the whole group of spondyloarthritis. One single family study previously performed with a similar objective reported negative results, but it was not sufficiently powered to reveal an association of genes with a small effect.10 We took advantage of the availability of trios to perform a family-based transmission disequilibrium test, which is more robust than case–control association tests, because it avoids population stratification bias. Based on the combined analysis of family and case–control replication studies, both IL1A (p=0.012) and IL1F10 (p=0.0043) were found to be associated with spondyloarthritis. Moreover, the two most frequent haplotypes combining three SNP in IL1A and one in IL1F10 were found to be associated with both spondyloarthritis and AS phenotypes.
Several studies have already investigated the association between the IL1A, IL1B and IL1RN genes and AS, reporting conflicting results in different populations.12,–,15 28 29 Overall, the most consistent associations have been reported with IL1A variants. In particular, a recent prospective meta-analysis of the IL-1 gene complex studied the same polymorphisms as we did in the present study. Its results implied that rs1894399 situated in the IL1A gene was the most strongly associated with AS.17 This SNP was also the most significantly associated with spondyloarthritis, and the only one associated with AS in our family study.
On the other hand, both our family and case–control studies revealed an association of the IL1F10.3 marker (rs3811058) with spondyloarthritis, but not with AS. In our case–control and pooled analysis, this marker was also significantly associated with non-AS, whereas it was not associated with AS in the previously published meta-analysis study.17 Interestingly, a study of psoriatic arthritis reported two regions in the IL-1 cluster that appeared to influence the susceptibility to psoriatic arthritis independently of each other; the first one spanned through IL1A and the second into IL1F10.30 Altogether, these results may indicate a role for IL1F10 in modulating the phenotypic presentation of spondyloarthritis. Consistent with such an interpretation both IL1F10 and IL1F5 are thought to result from duplications of IL1RN, and their products are expected to exert antagonistic effects similar to those of IL-1Ra. Although the function of IL-1F10 is still poorly understood, IL-1F10 binds IL-1R1 suggesting that it may regulate IL-1α/β.31
We found no association either with IL1RN VNTR, nor with the two SNP located in the IL1RN gene that codes for IL-1Ra. Previous positive findings at this locus might be due to LD, as extensive LD was seen across the IL-1 gene cluster, especially between IL1F10 and IL1RN, which are physically close to each other.
In our study, the strongest associations concerned haplotypes rather than individual SNP, suggesting either that the markers studied herein were not themselves causally associated with spondyloarthritis, or that spondyloarthritis association with the IL-1 genes is complex, resulting from a combination of LD between the genes and of functional interactions between their products.32
Several GWAS have recently identified non-MHC genetic loci associated with AS, but not the IL-1 locus.3 5 In addition to the fact that these studies were restricted to the AS phenotype, the genetic approach was different and could explain discrepancies. GWAS case–control association aims to capture a substantial proportion of common variation on the human genome by using LD. However, even though such GWAS have been very successful in identifying genes, their power to detect susceptibility locus of low magnitude effect is incomplete. On the other hand, power limitations and study design differences could explain differences in the results observed between our family-based and case–control studies.
Spondyloarthritis is classically subdivided into AS, which is characterised by predominant axial skeletal involvement and advanced radiographic sacroiliitis, psoriatic arthritis, IBD-associated arthritis, reactive arthritis and undifferentiated spondyloarthritis. A familial aggregation among these subtypes has been demonstrated and suggested a share of predisposing genetic factors.19 On the other hand, specific genes could influence the phenotypic heterogeneity of spondyloarthritis. Indeed, in a previous family study, we have shown familial aggregation of two distinct phenotypes.33 This is the reason we stratified our analysis to distinguish between AS and non-AS phenotypes. However, we also decided not to expand further genotype–phenotype correlations, because of the limited number of patients presenting with other subphenotypes, such as those defined by extra-articular manifestations.
IL-1 plays key roles in innate as well as in adaptive immunity, to activate acute inflammatory responses. The original members of the IL-1 family of genes are IL1A and IL1B, coding for IL-1α and IL-1β, respectively, and IL1RN coding for the IL-1 receptor antagonist (IL-1Ra). With respect to the pathogenesis of AS, IL-1 is secreted by activated macrophages in the inflamed synovium and initiates the recruitment of immune cells and the progression of inflammation. Moreover, it stimulates bone resorption.34 Therefore it is plausible that IL1A may be associated with either disease susceptibility and/or sacroiliitis, as suggested by its negative association in patients without definite radiographic sacroillitis (non-AS). More recently, several others members of the IL-1 family have been identified with an important role in innate and adaptive immune responses. Other members of the family named IL1F5 to IL1F10 lie between IL1A and IL1RN, and are widely expressed. Recent evidence implicated IL-1F5, IL1F6, IL1F8 and IL1F9 in skin disease, particularly in psoriasis.10 11 On the other hand, IL1F10 appears to be preferentially expressed in the basal epithelium of the skin, and presents a genomic structure and amino acid sequence close to IL-1Ra and IL1F5, suggesting an antagonistic role.12
IL-1 genes may have weak effects on spondyloarthritis susceptibility individually, but may nevertheless be important in combination with other genetic factors, due to epistasis or other interactions. It is noteworthy that IL-1 is known to interact with the products of other genes recently identified as involved in the susceptibility to spondyloarthritis: IL1R2, IL23R and TNFSF15.3 7 Indeed, IL-1 combines with IL-23 to mediate inflammation via γδT cells and T-helper 17 cells, which are considered a key T-cell subset involved in several chronic inflammatory diseases, including spondyloarthritis.35 Furthermore, the expression of TL1A, the protein encoded by TNFSF15, is highly inducible by IL-1α.8 36
In conclusion, our study demonstrated an association between several polymorphisms located in the IL-1 gene cluster and spondyloarthritis as a whole, and showed that the IL1A locus was strongly associated with the AS phenotype, whereas IL1F10 was rather associated with non-AS. Our results confirm that the IL-1 gene cluster is a susceptibility locus for spondyloarthritis, and indicate a possible influence of this cluster on the spondyloarthritis phenotype.
References
Footnotes
DM and AK contributed equally to the work.
-
Competing interests None.
-
Patient consent Obtained.
-
Ethics approval This study was approved by the local ethics committees of Cochin Hospital (Paris, France) and of Ambroise Paré Hospital (Boulogne Billancourt, France).
-
Provenance and peer review Not commissioned; externally peer reviewed.