ARTHRITIS & RHEUMATISM Vol. 56, No. 6, June 2007, pp 1954–1965 DOI 10.1002/art.22644 © 2007, American College of Rheumatology Specific Gene Expression Profiles in Systemic Juvenile Idiopathic Arthritis Emma Mary Ogilvie, Arshad Khan, Mike Hubank, Paul Kellam, and Patricia Woo Objective. Patients with systemic juvenile idiopathic arthritis (JIA) have arthritis, quotidian fevers, and other extraarticular features. This disease often remains severe and debilitating. The purpose of this study was to compare gene expression profiles in peripheral blood mononuclear cells (PBMCs) from patients with active and inactive systemic JIA to define and better understand the cause of active disease. Methods. Gene expression profiles of PBMCs were determined in cells from 9 patients with active systemic JIA and 8 patients with inactive systemic JIA. Unsupervised clustering and significance analysis were performed. We compared the systemic JIA profile with data from patients with polyarticular JIA, chronic infantile neurologic, cutaneous, articular syndrome, Kawasaki disease, and systemic lupus erythematosus to identify disease-specific genes. Quantitative reverse transcription–polymerase chain reaction of selected genes was performed on negatively selected B cells, T cells, and monocytes. Results. Unsupervised clustering of expressed genes resulted in 2 groups that corresponded to the clinical status of the patients (active and inactive disease) and was independent of their medications. A total of 286 genes were identified as significantly upregulated in patients with active disease and 86% of them were specific to systemic JIA. Interleukin-6 (IL-6) was expressed in monocytes and B cells, IL-10 in monocytes, and suppressor of cytokine signaling 3 in monocytes and T cells from patients with active disease. Conclusion. Gene expression profiles in PBMCs identified disease-specific genes in patients with systemic JIA. Cell type analyses should allow further insight into the mechanisms of the disease. Juvenile idiopathic arthritis (JIA) is a heterogeneous group of diseases. Systemic JIA, a distinct subgroup of JIA, consists of a multisystem inflammatory disease in addition to arthritis. Systemic JIA accounts for ⬃11% of all patients with juvenile arthritis (1). Features of this subgroup of patients include quotidian fevers, skin rash, lymphadenopathy, serositis, and hepatosplenomegaly. The more severe spectrum of systemic JIA has all of these features from the onset of the disease, and patients often develop severe debilitating complications. Despite recent therapeutic advances, children with persistent systemic JIA still experience early and progressive joint destruction, and ⬃30% have active disease that continues for 10 years or more after diagnosis. The etiology of systemic JIA, as with other complex autoimmune/inflammatory diseases, is thought to be through the interaction of genetic predisposition with an environmental trigger in early life. So far, no geographic clustering or any microbial association with this disease has been found to indicate a dominant environmental etiology. While HLA associations are a significant genetic factor in most of the JIA subtypes, there are few or no associations with systemic JIA (2,3). However, non-HLA genes have been reported to be associated with systemic JIA, for example, the gene for interleukin-6 (IL-6), where the more highly expressed IL-6 allele is a susceptibility gene for systemic JIA (4), and the gene for macrophage inhibitory factor, where a polymorphism is also found to be associated with systemic JIA (5). Despite these associations, the mechanisms of action of these genes in the pathogenesis of systemic JIA are not known. Measurements of cytokines have also shown that serum and synovial levels of IL-6 Supported by the Arthritis Research Campaign (grant 13895) and the Medical Research Council (grant 69716). Emma Mary Ogilvie, MSc, Arshad Khan, MSc, Mike Hubank, PhD, Paul Kellam, PhD, Patricia Woo, PhD, FRCP, FRCPCH: University College London, London, UK. Address correspondence and reprint requests to Patricia Woo, PhD, FRCP, FRCPCH, University College London, 3rd Floor Windeyer Institute, 46 Cleveland Street, London W1T 4JF, UK. E-mail: firstname.lastname@example.org. Submitted for publication July 26, 2006; accepted in revised form February 16, 2007. 1954 GENE EXPRESSION PROFILES IN SYSTEMIC JIA are increased in active systemic JIA. Whether this is the consequence of upstream immunopathologic changes is not clear. Moreover, the main cellular source of the IL-6 in serum and synovial fluid is not known. Many systemic JIA patients have systemic onset of symptoms, with little evidence of active synovitis, in the beginning. Often, persistent arthritis and other symptoms continue after the fevers subside. The most severe cases will have all disease-defining symptoms and frank polyarthritis from the beginning of the illness. It is thus likely that the pathology is not confined to, or initiated within, the joints. Peripheral blood mononuclear cells (PBMCs) are therefore likely to reflect systemic pathologic changes. Up to now, the role of PBMCs in JIA and the source of genes that lead to the systemic features during active disease have not been determined. In the present study, we compared the gene expression profiles of PBMCs from patients with active systemic JIA and patients with inactive systemic JIA in order to gain insight into which genes are up-regulated or down-regulated during the active phase of systemic JIA. We reasoned that healthy children are not necessarily a good control group for identifying the expression of specific genes responsible for disease activity, especially when the numbers of systemic JIA patients are small, since the differences observed may simply reflect variations between healthy individuals and systemic JIA patients that are not part of the disease process. Rather, by comparing samples obtained from patients with active systemic JIA and from patients with inactive systemic JIA, we control for the genetic variations inherent in systemic JIA patients in order to find specific genes that are up-regulated during the active disease process and thus gain insight into the specific disease pathways. This study focused on the pathogenesis of active systemic JIA, but not that of systemic JIA in general, so any genes involved in both active and inactive systemic JIA would not be identified. We also attempted to compare the genes up-regulated in children with active systemic JIA with the genes up-regulated in children with other systemic inflammatory diseases in order to separate genes that are involved in all inflammatory processes and those that are likely to be more specific to systemic JIA. We used data from both autoimmune and autoinflammatory diseases in children (polyarticular JIA, chronic infantile neurologic, cutaneous, articular syndrome [CINCA], Kawasaki disease [KD] and systemic lupus erythematosus [SLE]) in order to determine which genes are likely to be specific to systemic JIA. Importantly, we also found that there are cell type– 1955 specific expressions of selected genes, suggesting that the pathology of systemic JIA is manifested through multiple cell types, which can be observed in the peripheral blood. PATIENTS AND METHODS RNA samples. The systemic JIA patients that were recruited for the study fulfilled the International League of Associations for Rheumatology classification criteria for systemic-onset JIA (6). Patients were designated as having active disease (n ⫽ 9) if they had an elevated erythrocyte sedimentation rate (ESR) and raised levels of C-reactive protein (CRP), had synovitis in ⱖ1 joint, and had at least 1 other systemic feature. Patients were designated as having inactive disease (n ⫽ 8) if they had no joint involvement, low-to-normal levels of acute-phase proteins, and no features of systemic disease. Patients were between the ages of 1 year and 22 years, and all were Caucasian. Research was performed in compliance with the Declaration of Helsinki, and ethics approval was obtained from the Great Ormond Street Hospital for Children National Health Service Trust and the Institute of Child Health, Research Ethics Committee (reference 02RU06). Informed consent was obtained, and then 5 ml of blood was drawn. PBMCs were obtained by density-gradient centrifugation of blood over Lymphoprep (Axis-Shield, Kimbolton, UK), using endotoxin-free reagents. The PBMCs were resuspended in TRIzol (Invitrogen, Paisley, UK), and RNA was extracted according to the manufacturer’s protocol. Enrichment of B cell, T cell, and monocyte fractions. B cells, T cells, and monocytes were negatively selected from fresh PBMCs that had been separated with Lymphoprep (Axis-Shield) using negative isolation kits from Dynal (Wirral, UK). Lipopolysaccharide was excluded in the separation procedure with the use of endotoxin-free reagents. Purity of the cell fractions was confirmed using fluorescence-activated cell sorter analysis with CD19, CD3, and CD14, respectively. We obtained ⬎90% purity of cells for the B cell fraction and T lymphocyte cell fractions. For the monocyte fraction, ⬎80% purity was obtained. The cells were resuspended in TRIzol and then RNA-extracted according to the manufacturer’s protocol. Microarray procedures. Samples were prepared using the GeneChip eukaryotic small sample target labeling assay, version II, with 100 ng of RNA and an Affymetrix GeneChip IVT labeling kit (Affymetrix, Santa Clara, CA). Samples were hybridized to Affymetrix U133 Plus 2.0 arrays. Analysis of expression data. Data from the arrays were generated and exported using Affymetrix Microarray Suite software. Data were normalized according to the median. The Institute for Genomic Research MultiExperiment Viewer (available at http://www.tm4.org/mev.html) was used for clustering and significance analysis. Hierarchical clustering (7) was performed using Pearson’s correlation. Significance Analysis of Microarrays (SAM) (8) was used to find genes with significantly different levels of expression. SAM was performed with 3 alternative group designations. In the first SAM (SAM1), all samples from patients with active disease were placed in one group and all samples from patients with inactive disease in 1956 OGILVIE ET AL Table 1. Information on samples used for microarrays and for quantitative RT-PCR confirmation of microarray results* Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample Sample 1 4 5 6I 6A 7 8 12 13 14I 14A 15 16 19 20 25 27 2 3 4(2) 5(2) 8(2) 10 11 26 28 24 29 30 34 35 Disease state MTX Steroids Anti-TNF Active Active Active Inactive Active Inactive Active Active Inactive Inactive Active Inactive Active Active Inactive Inactive Inactive Inactive Inactive Active Active Active Inactive Inactive Active Active Active Active Active Active Active Yes Yes Yes Yes Yes No Yes No No Yes Yes No Yes Yes Yes Yes No Yes No Yes Yes Yes Yes Yes Yes No No Yes No Yes Yes No Yes No Yes Yes No Yes No Yes No No No No No Yes No No Yes No Yes No Yes No Yes No Yes Yes Yes Yes Yes Yes No No Infliximab No No No No No Etanercept Infliximab Infliximab No No No No No No Etanercept Etanercept No Infliximab No No Etanercept Infliximab No No No No No No Sample use Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Microarray Quantitative Quantitative Quantitative Quantitative Quantitative Quantitative Quantitative Quantitative Quantitative Cell types Cell types Cell types Cell types Cell types RT-PCR RT-PCR RT-PCR RT-PCR RT-PCR RT-PCR RT-PCR RT-PCR RT-PCR * Two patients contributed samples during periods of both active and inactive disease (samples 6A and 14A during active disease; samples 6I and 14I during inactive disease). Patients 4, 5, and 8 contributed additional samples at a later date; these additional samples are denoted with “(2).” RT-PCR ⫽ reverse transcription–polymerase chain reaction; MTX ⫽ methotrexate; anti-TNF ⫽ anti–tumor necrosis factor. another. In SAM2, the samples from active and inactive disease patients that had partitioned upon hierarchical clustering were placed into 2 groups (and the 2 samples that had not clustered with other samples from the same disease group were excluded). In SAM3, the samples were grouped according to the 2 partitioned clusters obtained from hierarchical clustering. In each case, a delta value was chosen to obtain a false discovery rate (FDR) of ⱕ10%. Differentially expressed genes were functionally annotated using Ensembl v38 software, April 2006 release (available at http://www.ensembl.org/). Comparison with genes up-regulated in other inflammatory diseases. We compared the list of genes obtained from our study with the list of up-regulated genes obtained from published reports on polyarticular JIA (9), CINCA (comparison of CINCA patients who were treated versus those who were not treated with anakinra [Aksentijevich I: personal communication]), KD (10), and pediatric SLE (11). Because the Affymetrix chips that were used were not identical in all these studies, we reexamined all of the gene annotations. Genes that could be accurately identified in each study, using Ensembl identifiers where possible, were used to identify a common set of genes across all studies and were subsequently used to determine those that were differentially expressed in more than one disease. Quantitative reverse transcription–polymerase chain reaction (RT-PCR). Quantitative RT-PCR was performed including a DNA ablation step (QuantiTect Reverse Transcription Kit; Qiagen, Crawley, UK), followed by quantitative PCR (with QuantiTect Primer assays and QuantiTect SYBR Green PCR kit; Qiagen) and run on the Applied Biosystems 7000 Sequence Detection System (Applied Biosystems, Warrington, UK). We compared a number of endogenous control genes since it was recently shown that even some housekeeping genes vary when analyzed by real-time PCR. We tested ␤-actin, large ribosomal protein P0, eukaryotic translationinitiation factor, and hypoxanthine phosphoribosyltransferase 1. We chose large ribosomal protein P0 as our control gene because of its consistent expression level across our samples. Statistical analysis of the quantitative RT-PCR results was performed using the Mann-Whitney U test. RESULTS Characteristics of PBMC samples. PBMCs were derived from blood samples obtained from 9 patients during active disease and 8 patients during quiescent disease. The patients were all treated with 1 of the GENE EXPRESSION PROFILES IN SYSTEMIC JIA 1957 Figure 1. Clustering of samples according to disease state in patients with active and inactive systemic juvenile idiopathic arthritis. A, Samples grouped by disease status through hierarchical clustering of all genes. Two patients contributed samples during periods of both active and inactive disease (samples 6A and 14A during active disease; samples 6I and 14I during inactive disease). The samples clustered according to whether the disease was active or inactive, with 2 exceptions (sample 6A and sample 7). The sample clustering was not dependent on the medications taken or the age of the patients. Patients in group 1 were taking methotrexate (MTX) and steroids, those in group 2 were taking MTX and anti–tumor necrosis factor (anti-TNF), those in group 3 were taking MTX only, those in group 4 were taking anti-TNF and steroids, and those in group 5 were taking nonsteroidal antiinflammatory drugs only. B, Hierarchical clustering of the 286 genes identified as significantly differentially expressed by Significance Analysis of Microarrays (SAM), with samples grouped according to disease status. Shown at the right are the functional groups into which the genes that are up-regulated during active disease were assigned using Ensembl. The colored scale bar at the bottom indicates the fold change in gene expression across the groups. 1958 following 5 groups of medications: group 1 consisted of methotrexate (MTX) and steroids (3 patients with active and 1 with inactive disease), group 2 consisted of MTX and anti–tumor necrosis factor (anti-TNF) (2 patients with active and 2 with inactive disease), group 3 consisted of MTX only (3 patients with active and 1 with inactive disease), group 4 consisted of anti-TNF and steroids (1 patient with inactive disease), and group 5 consisted of nonsteroidal antiinflammatory drugs only (1 patient with active and 3 with inactive disease). The patients with active diseases were generally receiving higher doses of steroids, but similar doses of MTX and anti-TNF (Table 1). Two individuals contributed samples during both active and inactive stages of disease; these were samples 6I (inactive) and 6A (active) and samples 14I (inactive) and 14A (active). White blood cell (6.69–11.22 ⫻ 109/liter), lymphocyte (1.81–3.2 ⫻ 109/liter), and neutrophil (3.86– 7.22 ⫻ 109/liter) counts were within the normal range for all patients, except for patient 12, who had active disease and was receiving no medications (22.58 ⫻ 109/liter, 4.73 ⫻ 109/liter, and 16.78 ⫻ 109/liter, respectively). There was no statistically significant difference in the cell counts between the samples obtained from patients with active disease and those obtained from patients with inactive disease (data not shown). Clustering of samples from patients with active and inactive systemic JIA into 2 groups, as determined by unsupervised hierarchical clustering. To identify gene expression differences between active and inactive systemic JIA, we performed Affymetrix gene expression profiling. Following data filtering and normalization, we explored the data structuring using unsupervised hierarchical clustering (7). This showed that the majority of the samples from patients with clinically active disease grouped into one cluster, which was distinct from the cluster for the patients with inactive disease (Figure 1A). Gene expression profiles from sample 14 during either active or inactive disease (sample 14A or 14I) clustered according to the clinically defined disease state, indicating that active and inactive disease differs within as well as between individual patients. For the other samples that were taken at 2 time points from the same patient (samples 6I and 6A), the one obtained during inactive disease clustered with the other samples from patients with inactive disease, but the one obtained during active disease clustered with neither the active nor the inactive disease samples. This individual is discussed below. The data have been submitted to ArrayExpress (accession no. E-MEXP-987). OGILVIE ET AL Figure 2. Overlap between genes that were unequivocally upregulated in patients with active systemic juvenile idiopathic arthritis (JIA) and in patients with other inflammatory diseases. Shown are the number of overlapping genes between systemic JIA (sJIA) versus Kawasaki disease (KD) and systemic lupus erythematosus (SLE), versus polyarticular JIA, versus chronic infantile neurologic, cutaneous, articular syndrome (CINCA), versus KD, and versus SLE. Table 2 lists the individual genes for all comparison groups except for systemic JIA versus CINCA. Hierarchical clustering of systemic JIA patients independent of medications. The location within the cluster of samples from patients in the various medication groups was examined in order to ensure that medication was not the cause of the observed pattern of clustering. The clustering of samples into active and inactive disease groups was not influenced by the medications they were taking (Figure 1A). Significance analysis of patients in the various medication groups also showed that there was no significant difference in gene expression between the different medication groups. There were 2 exceptions to this robust separation of disease status by gene expression. Sample 7 did not cluster as an inactive disease sample. However, the patient from whom this sample was obtained did not fulfill all of the clinical criteria for inactive disease; only clinical measurements were obtained, and the ESR and CRP levels were not available. Sample 6A was clinically designated as active disease, but the gene expression profile clustered with the inactive disease samples. However, it was noted that this sample showed up-regulation of a subset of the genes associated with active disease samples, suggesting that sample 6A is at the midpoint of the spectrum of disease. Importantly, sample 6I, which was taken from the same individual 21 weeks before during a clinically inactive phase, did not cluster with GENE EXPRESSION PROFILES IN SYSTEMIC JIA 1959 Table 2. Genes that overlap between systemic JIA and other inflammatory diseases* Affymetrix ID Overlap groups, gene name Systemic JIA, KD, and SLE Adrenomedullin Cytidine deaminase Haptoglobin S100 calcium-binding protein A12 S100 calcium-binding protein A9 S100 calcium-binding protein P Systemic JIA and polyarticular JIA Early growth response 1 Glutamate-ammonia ligase (glutamine synthase) Interleukin-8 receptor ␤ Microtubule-associated monoxygenase, calponin, and LIM domain–containing 2 Serpin peptidase inhibitor, clade B (ovalbumin), member 2 Thrombomodulin Systemic JIA and KD Adrenomedullin Cytidine deaminase Clusterin Cytochrome P450 family 1 subfamily B polypeptide 1 Cysteine-rich secretory protein LCCL domain–containing 2 Dysferlin Early growth response 1 Coagulation factor V Formyl peptide receptor 1 H2.0-like homeobox 1 (Drosophila) Haptoglobin Interleukin-8 receptor ␤ Microtubule-associated monoxygenase, calponin, and LIM domain–containing 2 Membrane-spanning 4-domains subfamily A member 4 S100 calcium-binding protein A12 S100 calcium-binding protein A9 S100 calcium-binding protein P Systemic JIA and SLE Adrenomedullin Cytidine deaminase Glutamate-ammonia ligase (glutamine synthase) Haptoglobin Matrix metalloproteinase 9 S100 calcium-binding protein A12 S100 calcium-binding protein A9 S100 calcium-binding protein P Gene symbol 202912_at 205627_at 206697_s_at 205863_at 203535_at 204351_at ADM CDA HP S100A12 S100A9 S100P 201693_s_at 217202_s_at 207008_at 212473_s_at 204614_at 203887_s_at EGR1 GLUL IL8RB MICAL2 SERPINB2 THBD 202912_at 205627_at 222043_at 202435_s_at 221541_at 218660_at 201693_s_at 204714_s_at 205118_at 214438_at 206697_s_at 207008_at 212473_s_at 219607_s_at 205863_at 203535_at 204351_at ADM† CDA† CLU† CYP1B1† CRISPLD2† DYSF† EGR1 F5† FPR1 HLX1† HP† IL8RB MICAL2 MS4A4A† S100A12† S100A9† S100P† 202912_at 205627_at 242281_at 206697_s_at 203936_s_at 205863_at 203535_at 204351_at ADM CDA GLUL HP MMP9 S100A12 S100A9 S100P * JIA ⫽ juvenile idiopathic arthritis; KD ⫽ Kawasaki disease; SLE ⫽ systemic lupus erythematosus. † One of 13 genes found by Abe et al (10) to be expressed in the monocyte fraction. sample 6A, but fell into the cluster with the other inactive disease samples. This result further supports the fact that sample 6 represented an individual who had an episode of active disease later, but was disease-free at the time of this sampling. Significant up-regulation of many genes in active systemic JIA. In order to identify genes that were significantly differentially expressed between active and inactive systemic JIA, we used SAM (8) and the clinical classifications of disease status (Figure 1A). The 3 groups of significance analyses (SAM1, SAM2, and SAM3) showed a total of 286 genes that were significantly up-regulated in patients with active disease (Fig- ure 1B). SAM1 (clinically defined active disease samples compared with inactive disease samples) showed 163 genes that were significantly up-regulated in active disease (⌬ ⫽ 0.685, 10% FDR). SAM2 (active disease samples, excluding sample 6A, and inactive disease samples, excluding sample 7) showed 203 genes that were significantly up-regulated in active disease (⌬ ⫽ 0.806, 6.5% FDR). SAM3 (samples grouped based on the 2 clusters into which they were partitioned during unsupervised hierarchical clustering) showed 129 genes that were significantly up-regulated in active disease (⌬ ⫽ 0.79, 5.3% FDR). As expected, there was considerable overlap among the 3 analyses. Nevertheless, 113 1960 genes were identified as being specific to an analysis group, with 54 unique genes in SAM1, 31 in SAM2, and 28 in SAM3. Of the 286 up-regulated genes, 129 could be annotated, allowing assignment of genes into functional groups. (Data on genes that were up-regulated in systemic JIA according to their functional annotation are available upon request from the corresponding author.) The main functional groups cover a wide range of processes. These are listed in Figure 1B. Likely disease-specific expression of many of the 286 up-regulated genes in active systemic JIA. It is possible that all studies of inflammatory disease simply identify sets of genes that are inflammation-specific rather than disease-type–specific, and since no crosscomparison studies have been undertaken, we wished to determine whether this was the case. We compared our systemic JIA profile with the profiles for polyarticular JIA (9), CINCA (Aksentijevich I: personal communication), KD (10), and SLE (11) by calculating the degree of overlap between the genes that were assigned as differentially regulated in these studies and that were unequivocally the same gene. Nearly all of the genes that were found to be up-regulated in our active disease group were specific to systemic JIA. Only 40 of our 286 genes were also up-regulated in a number of other inflammatory diseases as compared with published data (Figure 2). Six genes were found to overlap in systemic JIA, KD, and SLE (Table 2), but not in polyarticular JIA. Polyarticular JIA and systemic JIA had 6 overlapping genes (Table 2), CINCA and systemic JIA had 35 (data not shown), systemic JIA and KD had 17 (Table 2), and SLE and systemic JIA had 8 (Table 2). Quantitative RT-PCR confirmation of diseaseassociated gene expression in an independent sample of systemic JIA patients. To ensure that the gene expression profiles accurately reflected gene expression levels, we assessed representative genes that were up-regulated in samples from patients with active systemic JIA by quantitative RT-PCR, namely, IL-10 and suppressor of cytokine signaling 3 (SOCS-3). IL-10 was chosen because it is an antiinflammatory gene that may not have been expected to be more up-regulated in active versus inactive disease, given that IL-10 is a potent antiinflammatory cytokine and, intuitively, one might expect more of it in inactive disease. SOCS-3 was chosen because it is induced by both IL-10– and IL-6–mediated signaling; the genes for both of these cytokines have been previously shown by us to be associated with systemic JIA (4,12,13). OGILVIE ET AL Figure 3. Relative expression levels of genes for interleukin-10 (IL10) and suppressor of cytokine signaling 3 (SOCS-3). The 17 samples, 9 from patients with active juvenile idiopathic arthritis (JIA) (}) and 8 from patients with inactive JIA (■), used for microarray analysis were analyzed by quantitative reverse transcription–polymerase chain reaction for the expression of A, IL-10 and B, SOCS-3. Expression levels were calculated relative to the control gene large ribosomal protein P0. Additional samples were obtained from 5 patients with active JIA (}) and 4 patients with inactive JIA (■), and the expression of C, IL-10 and D, SOCS-3 was determined. Horizontal bars show the mean. P values were calculated using the Mann-Whitney U test. The levels of expression of these genes were tested by quantitative RT-PCR on the 17 active and inactive disease samples used for the microarrays and, importantly, on 9 additional samples. Of these 9 samples, 5 were taken during active disease and 4 during the quiescent stage (Table 1). Quantitative RT-PCR of the samples used for gene expression profiling confirmed the microarray results for IL-10 and SOCS-3, showing up-regulation in active systemic JIA. Quantitative RTPCR of the additional independent samples from systemic JIA patients further confirmed the up-regulation of the IL-10 and SOCS-3 genes during active disease (Figure 3). To support the findings of our analysis suggesting that genes from each of the different inflammatory gene expression patterns are disease-specific, we investigated the expression of interferon-stimulated transcription factor 3␥ (ISGF-3␥), which was previously found not to be differentially expressed in systemic JIA, polyarticular JIA, CINCA, or KD, but to be up-regulated in SLE (11). GENE EXPRESSION PROFILES IN SYSTEMIC JIA We performed quantitative RT-PCR on patients with active systemic JIA to determine if the expression of ISGF-3␥ in systemic JIA was consistent with our gene expression cross-comparisons. The quantitative RTPCR confirmed that there was no difference in expression in our systemic JIA patients (data not shown). The level of ISGF-3␥ in all of our systemic JIA patients was low. We believe, therefore, that the interferon response is not a general inflammatory disease response, but is specific to SLE. Expression of IL-6 by B cells and monocytes. IL-6 was not differentially expressed between patients with active and inactive systemic JIA. There were many reasons to think that IL-6 expression would not differ in these samples. First, levels of IL-6 have a circadian pattern, and not all samples were obtained at the same time of day. Second, levels of IL-6 rise and fall in parallel with body temperature; not all of our patients with active JIA had a fever, and those who did were not all sampled at the same stage of the fever. Third, it was thought that, like other acute-phase proteins such as CRP, the IL-6 produced during active systemic JIA may come from the liver and the reticuloendothelial system including the synovium, making PBMCs the incorrect cell type to assess. The possibility exists that IL-6 is expressed by a subgroup of PBMCs and may be important for some of the unique systemic features of this disease. To determine if the lack of differential expression of IL-6 in PBMCs was due to masking of the signal from specific and less numerous cell types, we used quantitative RT-PCR to examine IL-6 expression in enriched fractions of B cells, T cells, and monocytes from 5 systemic JIA patients with clinically active disease (1 individual was not Caucasian) and 5 normal healthy controls. Levels of expression of IL-6 were assessed to determine the source of IL-6 in samples obtained from patients with active disease. IL-6 was not expressed by T cells. We found that IL-6 was expressed at a very low level by monocytes in healthy controls and at a significantly higher level in systemic JIA patients with active disease. B cells in healthy controls expressed higher levels of IL-6 than did monocytes. Active systemic JIA B cells from a couple of the patients expressed even higher levels of IL-6 than did B cells from healthy controls (Figure 4). Thus circulating mononuclear cell, and perhaps B cell, expression of IL-6 could be important in the clinical manifestations of systemic JIA. Up-regulated gene expression in multiple cell types. The balance of proinflammatory and antiinflammatory cytokines is considered to be in disequilibrium during active disease in systemic JIA, and it was inter- 1961 Figure 4. Expression of interleukin-6 (IL-6) in B cells and monocytes from patients with active systemic juvenile idiopathic arthritis (JIA). B cells (}), T cells (‚), and monocytes (■) from 5 patients with active systemic JIA (samples 24, 29, 30, 34, and 35 in Table 1) and 5 healthy control subjects were analyzed by quantitative reverse transcription– polymerase chain reaction for the expression of IL-6. Expression levels were calculated relative to the control gene large ribosomal protein P0. In healthy control subjects, IL-6 is produced by B cells; in patients with active systemic JIA, IL-6 is produced by monocytes and B cells. The level of IL-6 in B cells is much higher than that in monocytes. Horizontal bars show the mean value in monocytes. The P value represents the comparison of monocyte levels in patients versus controls, by Mann-Whitney U test. esting to see the up-regulation of a regulatory and antiinflammatory cytokine, IL-10, in samples from patients with active disease that were significantly above the level in samples from patients with inactive disease (Figure 3). Since dendritic cells and monocytes, as well as regulatory T cells and B cells, can all produce IL-10, we performed quantitative RT-PCR to measure the expression of IL-10 in each cell fraction in order to gain more insight into the disease mechanisms. We also examined the expression of SOCS-3, which is known to be up-regulated by both IL-10 and IL-6 intracellular signaling. Finding the source of SOCS-3 may help to indicate the cause of SOCS-3 production and the role of SOCS-3 in disease. We found that IL-10 was expressed by the monocyte fraction in samples from patients with active systemic JIA at a much higher level than by the monocyte fraction in samples from healthy controls. There was a very low level of IL-10 in the B cell and T cell fractions; however, since these fractions were only 90% pure, we cannot rule out the possibility that the presence of IL-10 was due to a small number of monocytes present in these samples (Figure 5). SOCS-3 was highly expressed by monocytes and T cells from 4 of 5 samples from systemic JIA patients with active disease. The levels of SOCS-3 in B cells from these patients were, however, negligible. 1962 OGILVIE ET AL Figure 5. Expression of interleukin-10 (IL-10) in monocytes and of suppressor of cytokine signaling 3 (SOCS-3) in monocytes and T cells. B cells (}), T cells (‚), and monocytes (■) from 5 patients with active systemic juvenile idiopathic arthritis (JIA) (samples 24, 29, 30, 34, and 35 in Table 1) and 5 healthy control subjects were analyzed by quantitative reverse transcription–polymerase chain reaction for the expression of A, IL-10 and B, SOCS-3. Expression levels were calculated relative to the control gene large ribosomal protein P0. IL-10 is expressed in monocytes (A), and SOCS-3 is expressed in monocytes and T cells (B). Horizontal bars show the mean value in monocytes. P values represent comparisons of monocyte levels in patients versus controls, by Mann-Whitney U test. Similarly, there was a very low level of expression of SOCS-3 in all cell types in samples from the healthy controls (Figure 5). DISCUSSION This is the first study to identify a gene expression signature that discriminates between active and inactive systemic JIA. Despite clinical similarities to other diseases, most of these expression differences are specific to systemic JIA. We have shown that many genes are up-regulated in active systemic JIA and have implications in the pathology and clinical manifestations of systemic JIA. Unsupervised clustering of the genes from all samples resulted in 2 distinct groups that correspond to their clinical classification as active and inactive disease. The clinical definitions were such that patients at opposite ends of the clinical spectrum were selected, and the gene expression profiles of these samples partitioned into clusters that corresponded to this clinical definition, rather than the type of medication being taken. There were no significant differences in white blood cell, lymphocyte, or neutrophil counts between patients with inactive and active disease, so the upregulated genes are unlikely to be due to a difference in cell frequencies within the PBMCs. This suggests, therefore, that the gene expression profiles are indicative of pathologic processes, rather than a confounding by other effects. PBMCs are increasingly being used for staging diseases (14,15). PBMCs are relatively easy to obtain and, hence, provide a good starting point for analyses of gene expression in systemic diseases. There are many examples in the literature where models of disease have been determined from PBMC expression data, but we believe much information is lost when this is not examined in the context of the cell type. Some studies have observed cell-specific signatures within PBMC expression data and then examined that cell type within disease groups, but there seems to be no precedent for examining more than one cell type within a disease. It is likely that a more detailed analysis of the involvement of discrete cell types is essential before disease mechanisms can be inferred. In the case of children, only small amounts of blood can be obtained, so deriving pure cell fractions from the same sample is more difficult. Nevertheless, we have successfully used negative selection of B cell, T cell, and monocytic cell types to examine the expression of selected genes that were up-regulated in our gene expression analysis. Of particular interest here is the increased production of IL-6 by monocytes and B cells in samples from patients with active disease as compared with samples from healthy controls. Monocytes and dendritic cells are a significant source of IL-6, and this is the first time that circulating monocytic cells have been shown to produce significant amounts of IL-6 in patients with systemic JIA. The role of B cells in systemic JIA has not been considered hitherto, and further work will need to be done to determine whether the high levels of IL-6 production in a couple of our patients can be replicated GENE EXPRESSION PROFILES IN SYSTEMIC JIA in a larger sample. Expression of IL-6 by these circulating white blood cells could be important in the pathogenesis of the multisystem manifestations of systemic JIA. Neutrophil gene expression was not investigated here, but should also be considered in future studies to build a comprehensive picture of integrated cellular function in systemic diseases. The finding of an up-regulation of antiinflammatory IL-10 in active systemic JIA as compared with inactive systemic JIA is counterintuitive. It is possible that the IL-10 is being expressed in response to the inflammation but that it is not effective against the disease. IL-10 is expressed by macrophages and regulatory T cells, among others, and analysis of cell type– specific expression would inform us of the pathways involved. We found that IL-10 was expressed mainly by monocytes from patients with active systemic JIA, but only at low levels in T cells. This may indicate the need for further investigation of the state of immune cell suppression in systemic JIA. Production of SOCS-3 is stimulated by both IL-10 and IL-6. We found that SOCS-3 was produced in both monocytes and T cells in patients with active systemic JIA. Further investigation of other related genes expressed by each of these cell types and subsequent pathway analysis will help to explain the role of these two genes in systemic JIA. Assignment of our gene list into functional groups indicated that genes with many diverse functions were up-regulated in active systemic JIA. Once expression in cell types has been determined, it will be possible to see which cell types are expressing genes from which functional groups. This may then inform us about the pathways involved in the pathogenesis of active systemic JIA. Recently, Pascual et al (16) identified an IL-1 signature through gene expression profiling of healthy PBMCs after in vitro incubation with systemic JIA sera. Their work subsequently showed up-regulation of IL-1␤ and IL-1 receptor type II expression in their systemic JIA patients as compared with healthy controls. We failed to identify an IL-1 signature in the genes upregulated in active PBMCs from our systemic JIA patients. Of the 17 genes Pascual and colleagues found to be up-regulated in systemic JIA, only 2, the genes for serpin peptidase inhibitor, clade B (ovalbumin), member 2 (SERPINB2) and for thrombomodulin (THBD), were found to be up-regulated in our patients with active systemic JIA. Pascual also showed that sera from febrile patients were more efficient at inducing IL-1␤ secretion in PBMCs from healthy individuals than sera from afebrile patients. This suggests that IL-1 is more related 1963 to fever, rather than to the cause of the full clinical manifestations of systemic JIA. Furthermore, the positive results from 11 patients treated with IL-1 signaling blockade (anakinra) have not been universally replicated by other groups of investigators. For example, in an ongoing study of our patients, only 1 of 6 patients has responded to IL-1 blockade (17), and Quartier et al (18) have presented data showing that only about one-half of their systemic JIA patients responded to IL-1 blockade. There are many pediatric systemic inflammatory diseases in which gene expression studies have been performed. This allows us to examine whether the genes that are up-regulated in these diseases are common to all the diseases or whether there are some diseasespecific genes. For this reason, we compared our data with other published data. There are differences in the Affymetrix microarrays used, summarizing methods, types of analysis, and types of annotation in the different data sets. Ideally, raw data from these studies would be compared; however, such data are not publicly available. It is nevertheless worthwhile to compare what is available, using the Ensembl ID of the published genes and not the gene names, which can change because the Ensembl database is regularly updated. By comparing genes that are unequivocally the same from these data sets, we have identified 246 genes that are up-regulated during active systemic JIA as compared with inactive systemic JIA, but were not apparently up-regulated in other inflammatory systemic diseases in children. This indicates there is very likely to be up-regulation of disease activity–specific genes in systemic JIA. Like systemic JIA, KD is a multisystem inflammatory illness of the young, with fevers, arthritis, rash, and coronary artery involvement. Abe et al (10) compared PBMC and monocytic gene expression in patients with KD before and after intravenous infusion of highdose immunoglobulin, which reduces inflammation and fever. Thirteen of the genes that were overexpressed in both systemic JIA and KD were expressed in the monocyte fraction (see Table 2). Comparing the expression in monocytes and T cells using quantitative RT-PCR showed that IL-6 was expressed by monocytes and not by T cells (10). We have shown that in systemic JIA, B cells produce even more IL-6 than monocytes, suggesting that B cells may also be important for IL-6 expression in KD. Abe and colleagues also showed that IL-10 was produced by T cells and monocytes, but T cells from our patients did not produce IL-10. Together, these data emphasize the importance of investigating multiple cell types in complex diseases and suggest that gene expres- 1964 sion and function are not only cell-specific, but also disease-specific. Our comparison of the published gene expression data for polyarticular JIA (9) with our gene expression data for systemic JIA did not show many overlaps. This could be due to the use of earlier versions of the Affymetrix arrays in the polyarticular JIA study, but may also be due to the fact that this mirrors the clinical impression that systemic JIA is very different from the other types of JIA. Since we specifically identified genes that were differentially up-regulated in active versus inactive systemic JIA, there may be susceptibility genes involved in the pathogenesis of systemic JIA that are not revealed by this comparison. Comparison with other inflammatory diseases or subtypes of JIA would identify whether some of the genes are unique to systemic JIA. We attempted this using currently available information and found that there is surprisingly little overlap between the up-regulated genes in our systemic JIA patients and those in diseases that have systemic inflammation as a feature and are clinically similar. This is the first indication in the literature that there is expression of systemic JIA–specific genes and that they are not common to all inflammatory diseases. In another study examining gene expression in polyarticular JIA (19), an interferon-␥ signature was found to be present in the cohort; however, the data are difficult to interpret. The microarrays used contained probes for only 2,382 genes, and many unconventional analyses and alternative gene lists were produced. We have shown that the interferon signaling molecule ISGF-3␥ is not differentially up-regulated in the expression profiles of patients with active systemic JIA, suggesting that the interferon pathway is not involved in systemic JIA. Bennett et al (11) also showed that while the interferon signature was present in their SLE patients, it was absent in their cohort of systemic JIA patients. We further confirmed by quantitative RT-PCR the lack of up-regulation of ISGF-3␥ in the arrays of cells from patients with active systemic JIA, thus lending support for our method of comparing gene expression data from PBMCs in different pediatric inflammatory diseases. In conclusion, this study is the first to demonstrate that there are disease activity–specific genes in systemic JIA. We have also shown that analysis of gene expression in different cell types is important if one is to infer disease mechanisms from gene expression profiling data. OGILVIE ET AL ACKNOWLEDGMENTS We would like to thank Ivona Aksentijevich for allowing us to compare our data with her as-yet-unpublished CINCA data and Virginia Pascual and Jacques Banchereaux for providing us with the SLE expression data. We would like to thank Mary Collins for helpful comments on the manuscript, Ana Gutierrez for help with the processing of samples, Mark Fife for advice on the experiments, and Nipurna Jina for performing the microarray hybridizations. AUTHOR CONTRIBUTIONS Dr. Woo had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study design. Ogilvie, Kellam, Woo. Acquisition of data. Ogilvie, Hubank. Analysis and interpretation of data. Ogilvie, Khan, Hubank, Kellam, Woo. Manuscript preparation. Ogilvie, Khan, Kellam, Woo. Statistical analysis. Ogilvie, Khan, Kellam. REFERENCES 1. Schneider R, Laxer RM. Systemic onset juvenile rheumatoid arthritis. Baillieres Clin Rheumatol 1998;12:245–71. 2. Glass DN, Giannini EH. Juvenile rheumatoid arthritis as a complex genetic trait [review]. Arthritis Rheum 1999;42:2261–8. 3. Thomson W, Barrett JH, Donn R, Pepper L, Kennedy LJ, Ollier WE, et al. Juvenile idiopathic arthritis classified by the ILAR criteria: HLA associations in UK patients. Rheumatology (Oxford) 2002;41:1183–9. 4. Fishman D, Faulds G, Jeffery R, Mohamed-Ali V, Yudkin JS, Humphries S, et al. The effect of novel polymorphisms in the interleukin-6 (IL-6) gene on IL-6 transcription and plasma IL-6 levels, and an association with systemic-onset juvenile chronic arthritis. J Clin Invest 1998;102:1369–76. 5. Donn R, Alourfi Z, De Benedetti F, Meazza C, Zeggini E, Lunt M, et al. Mutation screening of the macrophage migration inhibitory factor gene: positive association of a functional polymorphism of macrophage migration inhibitory factor with juvenile idiopathic arthritis. Arthritis Rheum 2002;46:2402–9. 6. Petty RE, Southwood TR, Baum J, Bhettay E, Glass DN, Manners P, et al. Revision of the proposed classification criteria for juvenile idiopathic arthritis: Durban, 1997. J Rheumatol 1998;25:1991–4. 7. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 1998;95:14863–8. 8. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001;98:5116–21. 9. Barnes MG, Aronow BJ, Luyrink LK, Moroldo MB, Pavlidis P, Passo MH, et al. Gene expression in juvenile arthritis and spondyloarthropathy: pro-angiogenic ELR⫹ chemokine genes relate to course of arthritis. Rheumatology (Oxford) 2004;43:973–9. 10. Abe J, Jibiki T, Noma S, Nakajima T, Saito H, Terai M. Gene expression profiling of the effect of high-dose intravenous Ig in patients with Kawasaki disease. J Immunol 2005;174:5837–45. 11. Bennett L, Palucka AK, Arce E, Cantrell V, Borvak J, Banchereau J, et al. Interferon and granulopoiesis signatures in systemic lupus erythematosus blood. J Exp Med 2003;197:711–23. 12. Ogilvie EM, Fife MS, Thompson SD, Twine N, Tsoras M, Moroldo M, et al. The –174G allele of the interleukin-6 gene GENE EXPRESSION PROFILES IN SYSTEMIC JIA 13. 14. 15. 16. confers susceptibility to systemic arthritis in children: a multicenter study using simplex and multiplex juvenile idiopathic arthritis families. Arthritis Rheum 2003;48:3202–6. Fife MS, Gutierrez A, Ogilvie EM, Stock CJ, Samuel JM, Thomson W, et al. Novel IL10 gene family associations with systemic juvenile idiopathic arthritis. Arthritis Res Ther 2006;8:R148. Whitney AR, Diehn M, Popper SJ, Alizadeh AA, Boldrick JC, Relman DA, et al. Individuality and variation in gene expression patterns in human blood. Proc Natl Acad Sci U S A 2003;100: 1896–901. Griffiths MJ, Shafi MJ, Popper SJ, Hemingway CA, Kortok MM, Wathen A, et al. Genomewide analysis of the host response to malaria in Kenyan children. J Infect Dis 2005;191:1599–611. Pascual V, Allantaz F, Arce E, Punaro M, Banchereau J. Role of interleukin-1 (IL-1) in the pathogenesis of systemic onset juvenile 1965 idiopathic arthritis and clinical response to IL-1 blockade. J Exp Med 2005;201:1479–86. 17. Ozen S, Hoffman HM, Frenkel J, Kastner D. Familial Mediterranean fever (FMF) and beyond: a new horizon. Fourth International Congress on the Systemic Autoinflammatory Diseases held in Bethesda, USA, 6–10 November 2005. Ann Rheum Dis 2006; 65:961–4. 18. Quartier Lequerre T, Rosellini D, Sibilia J, Debandt M, KonePaut I, Job-Deslandre C. Anakinra in systemic-onset juvenile idiopathic arthritis and adult onset Still’s disease [abstract]. Ann Rheum Dis 2006;65 Suppl II;107. 19. Jarvis JN, Dozmorov I, Jiang K, Frank MB, Szodoray P, Alex P, et al. Novel approaches to gene expression analysis of active polyarticular juvenile rheumatoid arthritis. Arthritis Res Ther 2004;61: R15–32.