© Richard Griffin, Shutterstock images®
Understanding of childhood arthritis has been advanced by several studies of global gene and protein expression profiling. Two studies of transcriptional profiling of total peripheral blood mononuclear cells (PBMCs) and one of synovial fluid mononuclear cells (SFMCs) suggest 'signatures' that could either distinguish types of arthritis or even predict disease course,1, 2, 3 while a proteomic study of synovial fluid has suggested that subtypes can be distinguished by protein expression.4 The term juvenile idiopathic arthritis (JIA) describes a heterogeneous group of autoimmune disorders, which manifest as arthritis before the age of 16 years. Current JIA classification reflects varying degrees of severity and the presence of other symptoms.5 Oligoarticular JIA, which affects 
4 joints in the first 6 months of disease, is the most common presentation: it can either self-remit or be controlled with minimal intervention (persistent oligoarticular JIA), or may extend to affect >4 joints (extended oligoarticular JIA). Children presenting with 
5 affected joints have polyarticular JIA, and are further subdivided into those with and without rheumatoid factor (RF+ and RF-, respectively). Other JIA subtypes reflect the presence of psoriasis (psoriatic-related arthritis), enthesitis (enthesitis-related arthritis), or quotidian fever and rash with other systemic features (systemic JIA). Although this classification system is a valuable tool for defining cases, it does not offer methods to predict prognosis or whether a particular treatment will be successful. Considerable heterogeneity in disease course and treatment response exists, both between and within subtypes of JIA. Predicting disease course and response to medications would offer tangible benefit to patients; these are the goals of several current research efforts.
Barnes et al.2 sought to identify subtype-specific genes by performing gene expression analysis of PBMCs from 136 patients with early-stage JIA and 59 healthy controls; the majority of the JIA patients had not yet received treatment. Great efforts were made to ensure rapid preparation of samples and quality control between experiments, which are vital issues in such studies. Supervised clustering revealed many genes differentially expressed in JIA patients compared with controls; these differences were most marked in patients with systemic JIA, which parallels earlier studies.6 Hierarchical clustering using differentially expressed transcripts confirmed heterogeneity within each subtype. Interestingly, roughly one quarter of oligoarticular, polyarticular and enthesitis-related arthritis patients clustered with controls, which suggests that PBMCs may not provide an adequate 'window' on disease process. Pathway analysis on 163 possible canonical pathways (defined by the software, Ingenuity Pathway Analysis7) suggested that 46 pathways were over-represented in JIA; again, differences between subtypes were seen. Whether unsupervised (or 'blind') clustering would have correctly classified the children according to subtype was not discussed. A caveat to this study is that the authors chose to group together children whose disease was initially mild but extended to more joints after at least 6 months (extended oligoarticular JIA5) with children with polyarticular JIA, thereby missing a potential opportunity to uncover novel mechanisms of disease extension. Provided the clinical data are available, performing such analysis in the future should be possible.
The issue of predictors of extension in patients with oligoarticular JIA was the focus of another recent study,3 where gene expression profiling was performed on SFMCs from children with oligoarthritis at first joint aspiration—before DMARD or steroid administration—and analyzed on the basis of outcome at 1 year, when the child had either persistent or extended oligoarticular disease. The demonstration that 344 genes distinguish these two groups indicates that molecular pathways involved in this deterioration are activated in the joint before the clinical event is apparent. Interestingly, the complement pathway was implicated in disease extension, and was also over-represented in JIA patients versus controls in the study of Barnes et al.2
The study from Griffin et al.1 focused on subtypes of polyarticular JIA (RF+ and RF-), and found three distinct expression signatures. The first contained transcripts expressed by monocytes. This group had the highest proportion of RF+ patients, although the signature was not exclusive to these patients. The authors suggest that this signature may relate closely to adult rheumatoid arthritis, a hypothesis that might be testable in silico by comparing their results with microarray studies performed in patients with rheumatoid arthritis using the same technology. Another signature, found in children who were all RF-, was characterized by genes that are inducible by transforming growth factor (TGF)-
. TGF- is thought to suppress immune responses both directly and through the action of regulatory T cells.8 This group also had fewer peripheral blood CD8+ T cells, a lower expression of CD8-specific transcripts and a relatively milder disease phenotype than the other groups. This finding had parallels with the Hunter et al.3 study, where oligoarthritis patients with low numbers of synovial fluid CD8+ T cells were least likely to develop extended arthritis and a high CD4:CD8 synovial T cell ratio was a predictor of favorable outcome.3 The third gene expression signature described by Griffin et al.1 serves as a warning about experiments on ex vivo biological material. This signature, which was associated with immediate-early immune responses, brought together patients and controls whose samples had the longest time between phlebotomy and the freezer, and reminds us that biological activity continues even after a sample is removed from the patient. A further group of patients in the Griffin et al.1 study who did not fall into any of these signatures was noteworthy for having a high proportion of cases positive for antinuclear antibody.
Predicting disease course and response to medications would offer tangible benefit to patients...
In a recent profiling study of JIA using a proteomics approach, Gibson et al.4 identified larger extracellular proteins in synovial fluid and looked for those that differed between oligoarticular and polyarticular subtypes, as well as between oligoarticular patients who had or had not extended by 1 year. The quantity of a protein did not differ with classification or outcome, but rather different isoforms or post-translational modifications of the same proteins were observed. This is the first study to identify proteins that could predict the future of a child recently diagnosed with JIA.
These studies are just the start of what global analysis may be able to offer toward improving treatment and revealing the underlying pathophysiology, and might mark the beginning of a new era of 'predictive molecular tools' for JIA. However, these approaches are expensive and labor intensive. In addition, all of these data may include false positives, and therefore need to be replicated in fresh cohorts. The next challenge is to ensure meticulous, protocol-driven sample collection, maximal use of information by public sharing of data, and use of top-quality data analysis methods in order to enable the validation of novel predictive biomarkers that will ultimately reduce the burden of juvenile arthritis for children and their families.
