A number of non-HLA genes have been tested for association with JIA. Only a few of these have been replicated in additional cohorts. Some of these are reviewed by Phelan et al [33
]. A systematic review of the literature suggests that about 100 different non-HLA candidate loci have been investigated for associations with JIA, in over 150 tests for genetic association in different cohorts (Table ). Although ~25 % of all tests reported finding an association, independent confirmations are found for only a handful of candidate genes including PTPN22, MIF, SLC11A6, WISP3
, and TNFA
(Table ). Many studies analyze JIA as a single phenotype. While combining clinically distinct entities such as systemic JIA and oligoarticular JIA might confound the results, stratification of subjects leads to further loss of power. Finally most studies do not correct for multiple comparisons, which might result in false positive associations that cannot be ultimately confirmed. The ranges of cases in these studies varied from 33 individuals to 950 individuals with JIA, and the number of controls ranged from 40 to 1952. The median number of cases was 130, while the median number of controls was 276. Only a few studies used more than 800 cases of JIA. These observations suggest that most of these studies are underpowered to detect associations with modest odds ratios (OR) (Table ). Results of studies in other complex genetic traits suggest that most associated variants have only a modest OR of about 1.5. These observations reinforce the importance of properly performed and reported studies including the capacity to replicate.
Non-HLA candidate markers tested for association with JIA.
Non-HLA genetic genes associated with JIA that have been independently confirmed.
Sample sizes requirements for genetic association studies
Interpreting the different associations (or lack of associations) between genetic variants and phenotypes can be further complicated by differences in phenotype description. JIA comprises several sub-phenotypes with distinct clinical features and outcomes[48
]. Although the ILAR classification criteria, which are increasingly being used to describe juvenile arthritis attempts to define homogenous subtypes, there are still some challenges. For instance a child with four affected joints is classified as oligoarticular JIA, while a child with five affected joints is classified as polyarticular JIA, although clinically the difference is not significant. Similarly, a child with five affected joints and another with over 20 joints are both classified as polyarticular, although clinically they appear distinct. While this can be addressed in part, by analyzing children with JIA as a group, and then performing analyses by subtypes, this results in multiple comparisons. One approach to address this issue might be to use genetic information to stratify subjects into subtypes, and this might results in more homogenous subtypes. For instance studies of adult RA subjects frequently stratify the subjects based on presence or absence of the shared HLA-DR
]. Similarly, it might be reasonable to stratify JIA patients by HLA associations while conducting genetic studies. It also would better facilitate comparisons of genetic studies performed in different populations. Finally, this would also help to delineate associations due to LD. For instance the MHC region has extensive LD, and associations described with variants in this region could be due to LD with HLA polymorphisms. Together, these observations reinforce the need for meticulous attention to defining the phenotypes and caution with interpreting results of genetic association studies.
A recent critical review of several genes and polymorphisms tested as part of commercially available genomic profiles highlights the challenges of genetic association testing [51
]. The authors used individual gene-disease association studies as well as meta-analyses to examine polymorphisms in 56 genes. Of these 32 genes had been examined in meta-analyses of 160 polymorphism-disease association comparisons. Only 38% were found to be statistically significant, with very modest associations. Furthermore, often associations were found with phenotypes different from the original phenotype being tested for association. For instance, genes tested in cardiogenomic profiles were more frequently associated with non-cardiac diseases. It is anticipated that most disease-phenotype associations of complex traits are likely to be of modest significance, with OR <1.5 [52
]. Thus, the modest genetic effects necessitates collaboration, both by performing collaborative studies where samples are pooled, as well as by data-sharing by investigators in order to detect meaningful associations[53
Non-replication of initial associations can be due to myriad factors including a false-positive result in the first cohort, population stratification in the first cohort, inadequate power in the replication cohort(s), genotyping errors, selection biases, and true population differences. It should be noted that lack of an association in the first cohort often might discourage further studies of that gene by other investigators, even though the first cohort might not have had adequate power to truly detect a modest association. The power to detect associations also depends on the minor allele frequency, and hence for rare variants much larger samples would be necessary. It is feasible to increase the power of genetic studies by using large numbers of appropriate controls, thereby increasing the control to case ratio. The quality of genetic associations will be significantly enhanced by increasing sample sizes, and validating initial discoveries by collaborative studies. Following published guidelines for performing and reporting genetic association studies will improve the quality of studies and facilitate the discovery of true causal variants[54
]. For instance, providing power calculations, and if the variant under study has been investigated previously, performance of a meta-analysis would be beneficial. Ultimately, associations that are reproducible, as well as demonstration of functional consequences of genetic variation are necessary to translate the results of these studies to the diagnosis, treatment and understanding of JIA.
Association studies between JIA and non-HLA variants have been reviewed previously [33
]. The following are some genetic associations that have been confirmed in more than one cohort. TNFA
, the gene encoding the pro-inflammatory cytokine TNF-α has been the focus of several association studies of autoimmune diseases. Associations between TNFA
and different JIA subtypes have been previously reported. An association between early onset oligoarticular JIA and TNF variants were reported by Epplen et al[56
]. Another study found associations between persistent oligoarticular JIA and TNFA
]. An association between a SNP at position -857 of TNFA
and sJIA was reported by Date et al[45
]. Schmeling et al found an association between TNFA
variants and psoriatic arthritis polyarticular JIA subtypes[58
]. In a case control study of children with JIA, Miterski et al found no associations with TNFA
SNPs, but found an association with a microsatellite in the TNFA
]. Thus, while variants in TNFA appear to be associated with JIA, there are clearly differences between various studies with regard to the associated variant(s) and/or associated subtype. It should be noted that some of the associations could be due to LD, as the TNFA
loci are both contained in the MHC region.
The gene encoding macrophage migration inhibitory factor (MIF) has also been investigated in several cohorts of patients with JIA. A functional SNP at position -173 has been associated with JIA in the study by Donn et al, in 526 British children with JIA and 259 ethnicity-matched controls[60
]. This finding was replicated using an independent cohort of JIA trios, using TDT [61
]. However, this SNP did not show an association in the studies by Miterski et al, and Berdeli et al, although it should be noted that both of these cohorts had smaller number of subjects [59
]. Of interest, the latter study did find that carriage of -173C allele was a strong predictor of poor outcome of JIA[62
]. It has previously been shown that this allele is a predictor of poor outcome in sJIA [63
]. A recent study of patients with oligoarticular JIA that had been followed for at least 5 years confirmed that -173C was a predictor of poor response to intra-articular corticosteroid treatment [64
]. Together, these studies imply that MIF variants may be associated with susceptibility to JIA, as well as the phenotype of JIA but should be noted that MIF-173C
was not associated with adult RA in a replication study[65
gene (formerly called NRAMP1
) is important in natural resistance to various intracellular infections mediated by macrophages. Variants in this gene have also been found to be associated with JIA in at least two cohorts [66
]. In the first study of Latvian and Russian children with JIA, a functional promoter microsatellite polymorphism (allele 3) was found to be associated with susceptibility to JIA (OR 2.26)[66
]. Conversely, another variant, allele 2, was associated with protection from JIA. Interestingly, these alleles demonstrate opposite associations in infection and autoimmunity, with allele 3 conferring susceptibility to autoimmunity and protection from tuberculosis. Conversely, allele 2 confers protection from autoimmunity while increasing susceptibility to tuberculosis. These suggest that balancing natural selection might be acting upon this locus. In a subsequent study, Runstadler et al studied both of these SLC11A1
alleles as well as three SNPs and an insertion/deletion polymorphism in the region [67
]. A six-marker haplotype demonstrated a strong association with JIA as a group, as well as in oligoarticular and polyarticular subtypes. Haplotypes that did not contain allele 3 identified in the study by Sanjeevi et al, were also found to show significant association suggesting that variants in this region might have independent associations with JIA. These two studies support a role for SLC11A1
variants in JIA susceptibility.
The gene encoding IL-1α, a potent pro-inflammatory cytokine, has been investigated in several JIA cohorts for a genetic association. Originally an association was described between a SNP in the promoter region of IL1A
in Norwegian children with JIA[68
]. The association was most pronounced in those with early onset oligoarticular JIA. However this finding was not replicated in two cohorts from the UK in subsequent studies, although one study investigating a number of polymorphisms demonstrated an association that was not significant when corrected for multiple comparisons [69
]. A recent two-staged association study identified several variants in the IL-1 gene cluster, and some in the IL-1 receptor cluster were associated with systemic JIA [71
]. IL-6 is another potently pro-inflammatory cytokine shown to be elevated in children with JIA[16
]. A functional SNP in the promoter of the IL6
gene was shown to be associated with sJIA by Fishman et al. The effect was pronounced in children under 5 years of age[75
]. This finding was not replicated in a smaller sJIA cohort by Donn et al[69
]. However, using an international cohort of families of children with sJIA, Ogilvie et al did find a significant association between the IL6-174 SNP and sJIA, but in children with onset >5 years[76
]. An association between a promoter variant in IRF1
, the gene encoding interferon regulatory factor and JIA was reported by Donn et al in a cohort of 417 cases with JIA and 276 controls[69
]. However, when they repeated the study using different controls and additional cases, no significant differences in allele or genotype frequencies of several IRF1
variants were observed. Together, these studies illustrate the difficulties of replicating significant positive findings from initial cohorts in subsequent larger replication studies.
An association between a SNP (G84A) in the gene encoding the Wnt-1 inducible signaling pathway protein 3 (WISP3
) and polyarticular JIA was described by Lamb et al[77
]. After observing a positive association in the initial cohort of 159 cases and 263 controls, the finding was replicated in an independent cohort of 181 cases and 355 controls. In addition they used parent-child trios for TDT, which demonstrated a trend towards over-transmission of the 84A allele. This finding has not been replicated in other cohorts to our knowledge.
JIA and RA are examples of autoimmune diseases mediated by Th1-immune responses. Synovial T-cells express high levels of the Th1-chemokine receptor (CCR5). A 32 basepair deletion in the open reading frame of the gene encoding CCR5 (Δ32) has been shown to be protective against RA in a meta-analysis of RA association studies[78
]. We have shown that a variant in the promoter region of CCR5
, C-1835T is significantly under-transmitted to children with early onset JIA, and with oligoarticular JIA[79
was also tested in ~700 simplex and multiplex JIA families and found to be under-transmitted to children with early onset JIA. These two variants did not appear to be in LD, and the haplotype that did not contain either of the "protective" variants was significantly over-transmitted. In a study of Norwegian adults with RA and children with JIA, Lindner et al tested CCR5-Δ32
, and failed to replicate these results[80
]. When they repeated the meta-analysis of RA association studies including their results, the negative association between RA and CCR5-Δ32
was less pronounced, but still statistically significant (OR 0.8, p < 0.007). When they compared children with JIA and the controls, although there was a trend towards a negative association, it was not statistically significant (OR 0.82, p < 0.15). The authors in that study did not investigate other CCR5
variants, including C-1835T.
The gene PTPN22
encodes a lymphoid tyrosine phosphatase (LYP) involved in inhibition of T-cell activation. A SNP within the PTPN22
gene (C1858T) has been shown to be associated with multiple autoimmune phenotypes including RA, T1DM, and SLE[65
]. To date, there have been four studies of this polymorphism in JIA cohorts [84
]. The study by Seldin et al did not find an association between 230 Finnish probands with JIA and this PTPN22
]. In contrast, Viken et al reported an association between JIA and PTPN22
C1858T using 320 Norwegian cases[86
]. Another study by Hinks et al also confirmed an association between PTPN22
and JIA using 661 JIA cases from the UK[84
]. A fourth study using Czech JIA patients also confirmed the association between the C1858T variant and JIA[87
]. A pooled analysis of these four studies confirms that carriage of the T allele increases the susceptibility to JIA, although the magnitude of the association is only modest, with an OR of ~1.3. Thus, PTPN22
has been shown to have an association with JIA.