Findings

Our understanding of the genetic basis of juvenile idiopathic arthritis (JIA) has recently increased, but still lags behind many other autoimmune diseases. This is largely due to the paucity of DNA collections internationally. While association of JIA with variation in the major histocompatibility complex (MHC) is well-established[1], over the last five years, there have been a number of reports of new JIA susceptibility loci that lie outside this region. These findings have resulted from candidate gene approaches (for example, examining genes known to be associated with rheumatoid arthritis), and more recently, from a limited number of genome-wide association study (GWAS) approaches. The vast majority of reports describe discovery and replication findings generated from two large sample collections, from the UK and the US[2]. Thus, although solid evidence for association with JIA has usually been described, for many of the identified loci, further replication in an entirely independent JIA sample would be beneficial in confirming their contribution to disease risk.

To address this, we performed a single nucleotide polymorphism (SNP) replication study within a new sample of JIA cases and healthy hospital-based child controls, collected so far as part of the ongoing Australian CLARITY study (ChiLdhood Arthritis Risk factor Identification sTudY)[3]. We selected 56 SNPs from 42 gene regions. Candidacy was based on published evidence of association, or a trend towards association, with total JIA, or one or more of its subtypes. These genes included ADAD1-IL2-IL21[4–6], AFF3[5], ANGPT1[6], ATXN2[7], BACH2[7], c12orf30[6, 8], c3orf1[9], CCR5[10], CD14[11], CD226[5], CD247[2], CLEC16A[12], COG6[6], CTLA4[5], ERAP1[13], IL12A[7], IL15[9], IL23R[6, 13], IL2RA[2, 6, 14], IL7R[5], JMJD1C-REEP3[9], KIF5A[15], LPP[7], MBL2[16], MEFV[17], NLRP3[17], NOD2[17], NRBF2-EGR2[9], PRKCQ[15], PSTPIP1[17], PTPN2[2, 6], PTPN22[6, 18], RANTES (CCL5)[19], STAT4[6, 8, 15], TNFA[20], TNFAIP3[8, 15], TRAF1-C5[15, 21], and VTCN1[22]. Additional SNPs from four other loci not attributed to any gene in the original publication but lying closest to the genes DCN1, FHIT, HUNK, and SLITRK5[22] were also selected. Genotyping was performed on a total of 324 JIA cases (mean age 9.7 years, 67.3% female) and 568 controls (mean age 7.8 years, 40.7% female) using the Sequenom MassARRAY system (assay design details available from the authors). After quality control pruning (removal of SNPs with Hardy Weinberg Equilibrium p < 0.01, and SNPs or samples with < 90% genotyping success rate) genotypes were successfully generated for 51 SNPs at 38 gene loci in 318 cases and 556 controls. SNPs excluded from analysis are listed in Additional file1. Allelic, genotypic, additive (Cochrane-Armitage test for trend), dominant and recessive association analyses were performed using PLINK v1.07[23]. Given the a priori evidence for association of each SNP, no adjustments for multiple testing were made.

Outcomes of the association analyses in CLARITY cases and controls, including p values for the most highly significant test and allelic test for all SNPs analysed is shown in Table 1. Figure 1 shows a forest plot of the allelic odds ratios generated in the CLARITY sample compared to previously published odds ratios (ORs) for the majority of SNPs tested. Some SNPs were not included in this figure, including SNPs for which an OR was not presented in the original report, or where SNPs were previously associated specifically with rarer JIA subtypes. Our data demonstrates clear evidence of replication (defined as p < 0.05 for any test, ORs or case-control allele frequency differences in a direction consistent with previous reports) for SNPs at loci containing the genes ATXN2, c12orf30, c3orf1, PTPN22, STAT4, and TRAF1-C5. Possible evidence of replication (defined as p < 0.2 for any test, ORs or case-control allele frequency differences in a direction consistent with previous reports) was also generated for SNPs near AFF3, CD226, MBL2, PSTPIP1, and RANTES (CCL5). For IL15, we found a significant association, but in the opposite direction to that previously reported. The IL15 SNP is an A/T transversion with frequencies of both alleles close to 50%, and thus there is the possibility of allele reversal. Although our minor T allele frequency in controls was entirely consistent with that found in the prior study, it is difficult to be sure that the IL15 assays used in both studies were based on the same DNA strand. We therefore cannot be sure if our data supports replication. Additionally for CTLA4, considered a general autoimmunity susceptibility gene but with conflicting association results for JIA[5, 24], we did not generate any evidence of replication.

Figure 1
figure 1

Forest Plot comparing previously published odds ratios (ORs) with CLARITY ORs for genes examined via nearby SNPs. Where more than one previously published OR was identified for a SNP, the OR based on the largest sample size was used. Some examined genes were excluded from this figure (see text for explanation). For genes with multiple SNPs (eg ADAD1-IL2-IL21) the SNPs are presented in the same order as listed in Table 1. * No gene attribution in original publication, closest gene by UCSC Genome Browser listed.

Full size image
Table 1 Clarity SNP association results, full sample
Full size table

We then performed a sensitivity analysis in which we included only cases (n = 200) and controls (n = 341) of European ancestry using the stringent definition of self-reported European ancestry of all four of the child’s grandparents. Non-European participants, along with participants for whom full grandparent data was not provided, were excluded. The outcome of this re-analysis is shown in Additional file1: Table S1 and Additional file1: Figure S1. In general, the results of this European-only analysis were not materially different to the full sample analysis (taking into account reduced statistical power resulting from a reduction of sample size), with one exception. ATXN2 appeared influenced by ethnicity, with an opposite, non-significant, direction of effect in the European subgroup.

In conclusion, we have provided independent replication data for JIA susceptibility loci that have previously been identified using a generally limited number of international sample collections. We have confirmed association of JIA with SNPs close-by to c12orf30, c3orf1, PTPN22, STAT4, and TRAF1-C5; and we have provided further support for the association of SNPs close-by to AFF3, CD226, MBL2, PSTPIP1, and CCL5. A limitation of our study was our relatively small sample size; our full dataset had 80% power to detect an OR of 1.4 for an allele at 20% frequency in the population at an alpha of 0.05. Given that many of the published ORs were less than 1.4, and that a number of the SNPs analysed had minor allele frequencies less than 20%, we cannot, from our current data, exclude association of any of the SNPs examined. Our current sample size also precluded detailed subtype-specific association analyses. Large collaborative GWAS efforts would be beneficial in confirming the outstanding genes, and providing further novel insights into the breadth of genetic loci involved in JIA susceptibility.