Gene expression signatures in polyarticular juvenile idiopathic arthritis demonstrate disease heterogeneity and offer a molecular classification of disease subsets.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 60, No. 7, July 2009, pp 2113–2123 DOI 10.1002/art.24534 © 2009, American College of Rheumatology Gene Expression Signatures in Polyarticular Juvenile Idiopathic Arthritis Demonstrate Disease Heterogeneity and Offer a Molecular Classification of Disease Subsets Thomas A. Griffin,1 Michael G. Barnes,1 Norman T. Ilowite,2 Judyann C. Olson,3 David D. Sherry,4 Beth S. Gottlieb,5 Bruce J. Aronow,1 Paul Pavlidis,6 Claas H. Hinze,1 Sherry Thornton,1 Susan D. Thompson,1 Alexei A. Grom,1 Robert A. Colbert,1 and David N. Glass1 Objective. To determine whether peripheral blood mononuclear cells (PBMCs) from children with recentonset polyarticular juvenile idiopathic arthritis (JIA) exhibit biologically or clinically informative gene expression signatures. Methods. Peripheral blood samples were obtained from 59 healthy children and 61 children with poly- articular JIA prior to treatment with second-line medications, such as methotrexate or biologic agents. RNA was extracted from isolated mononuclear cells, fluorescence labeled, and hybridized to commercial gene expression microarrays (Affymetrix HG-U133 Plus 2.0). Data were analyzed using analysis of variance at a 5% false discovery rate threshold after robust multichip analysis preprocessing and distance-weighted discrimination normalization. Results. Initial analysis revealed 873 probe sets for genes that were differentially expressed between polyarticular JIA patients and healthy controls. Hierarchical clustering of these probe sets distinguished 3 subgroups within the polyarticular JIA group. Prototypical patients within each subgroup were identified and used to define subgroup-specific gene expression signatures. One of these signatures was associated with monocyte markers, another with transforming growth factor ␤–inducible genes, and a third with immediate early genes. Correlation of gene expression signatures with clinical and biologic features of JIA subgroups suggested relevance to aspects of disease activity and supported the division of polyarticular JIA into distinct subsets. Conclusion. Gene expression signatures in PBMCs from patients with recent-onset polyarticular JIA reflect discrete disease processes and offer a molecular classification of disease. Supported by the NIH/National Institute of Arthritis and Musculoskeletal and Skin Diseases (grants P01-AR-048929, P30-AR47363, and P60-AR-47784), the Cincinnati Children’s Hospital Research Foundation, and the Ohio Valley Chapter of the Arthritis Foundation. Dr. Ilowite’s work was supported by Amgen. Dr. Colbert’s work was supported by a Cincinnati Children’s Hospital Medical Center Translational grant. 1 Thomas A. Griffin, MD, PhD, Michael G. Barnes, PhD, Bruce J. Aronow, PhD, Claas H. Hinze, MD, Sherry Thornton, PhD, Susan D. Thompson, PhD, Alexei A. Grom, MD, Robert A. Colbert, MD, PhD (current address: National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH, Bethesda, Maryland), David N. Glass, MD: Cincinnati Children’s Hospital Medical Center and University of Cincinnati College of Medicine, Cincinnati, Ohio; 2Norman T. Ilowite, MD: Albert Einstein College of Medicine, Bronx, New York; 3Judyann C. Olson, MD: Medical College of Wisconsin and Children’s Research Institute, Milwaukee, Wisconsin; 4David D. Sherry, MD: Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania; 5Beth S. Gottlieb, MD: Schneider Children’s Hospital, New Hyde Park, New York; 6Paul Pavlidis, PhD: University of British Columbia, Vancouver, British Columbia, Canada. Dr. Ilowite has received consulting fees, speaking fees, and/or honoraria from Abbott, Novartis, and Bristol-Myers Squibb (less than $10,000 each). Dr. Colbert has received honoraria from the Spondyloarthritis Research and Treatment Network (SPARTAN), the University of Rochester Medical Center, and Grand Rounds; Carolinas (2007) (less than $10,000 each). Address correspondence and reprint requests to Thomas A. Griffin, MD, PhD, William S. Rowe Division of Pediatric Rheumatology, Cincinnati Children’s Hospital Medical Center, 3333 Burnet Avenue, ML 4010, Cincinnati, OH 45229. E-mail: email@example.com. Submitted for publication September 30, 2008; accepted in revised form February 20, 2009. Polyarticular juvenile idiopathic arthritis (JIA) is chronic arthritis affecting more than 4 joints for more than 6 weeks, with onset before the sixteenth birthday in 2113 2114 GRIFFIN ET AL a child without other known causes of arthritis (1–3). Polyarticular JIA is divided into subtypes according to the presence or absence of rheumatoid factor (RF), with the RF⫹ subtype having positive results on tests for serum IgM-RF on 2 occasions at least 3 months apart within the first 6 months of disease. Ravelli and Martini (3) have recently proposed that RF⫺ polyarticular JIA be divided into 3 subsets: one similar to adult rheumatoid arthritis (RA), another with “dry synovitis,” and a third similar to antinuclear antibody (ANA)–positive early-onset oligoarticular JIA (3). Given this heterogeneity, it is not surprising that children with polyarticular JIA have a wide variety of disease courses and outcomes, ranging from self-limited arthritis with no longterm disabilities to relentless and destructive arthritis with severe disabilities (4). Unfortunately, our present ability to predict the disease course and outcome is limited, and treatment is typically tailored to the patient’s current disease activity, the assessment of which is also imperfect. Global gene expression profiling is a molecular technique that measures in parallel the genome-wide expression of thousands of genes in a sample of cells. This technology holds promise for dramatically advancing our knowledge of many diseases, including JIA. This approach has already provided important information regarding the classification and pathogenesis of several JIA subtypes in studies that generally used small numbers of patients with various degrees of clinical diversity (5–9). In the present study, global gene expression profiling of peripheral blood mononuclear cells (PBMCs) was used to characterize a relatively large population of children with recent-onset polyarticular JIA (both RF⫺ and RF⫹ patients) who had not been treated with methotrexate, biologic agents, or other disease-modifying antirheumatic drugs (DMARDs). The goals of applying this technology to JIA are to advance the understanding of disease pathogenesis, improve the assessment of disease activity, predict the response to medications, and foresee the long-term outcomes. The present study takes a step toward these goals by defining gene expression signatures that appear to be associated with distinct disease processes in subgroups of children with polyarticular JIA. PATIENTS AND METHODS Patients and collection of clinical data. Sixty-one children with polyarticular JIA classified according to the criteria of the International League of Associations for Rheu- matology (2) were recruited at 5 clinical sites: 24 from Cincinnati Children’s Hospital Medical Center (CCHMC), 16 from Schneider Children’s Hospital, 9 from Children’s Hospital of Philadelphia, 6 from Toledo Children’s Hospital (Toledo, OH), and 6 from Children’s Hospital of Wisconsin. Of these 61 patients, 46 were taking scheduled nonsteroidal antiinflammatory drugs, 3 were taking prednisone, and none had ever been treated with methotrexate, other DMARDs, or biologic agents. Informed consent was obtained, and clinical data were collected. The following disease activity measures were determined: erythrocyte sedimentation rate (ESR), count of joints with active disease (tender and limited range of motion, and/or swollen), Childhood Health Assessment Questionnaire (C-HAQ), physician’s global assessment of disease activity, and patient’s/parent’s global assessment of well-being. All JIA patients were tested for RF, and if the first test showed positive results, a second test was performed at least 3 months later. Most JIA patients were also tested for anti–cyclic citrullinated peptide (anti-CCP) antibodies and ANAs (Table 1). If frozen serum was available and RF or anti-CCP testing was not performed at the collecting center, these tests were performed at CCHMC. The specific joints involved with arthritis at the time of sampling were documented. Fifty-nine healthy control subjects were recruited at CCHMC. Twenty-nine controls provided blood samples as part of a demographically representative population of Cincinnati that was recruited and screened to serve as healthy controls. Thirty controls provided blood samples and limited clinical data when blood was drawn at the CCHMC Test Referral Center for clinical purposes. These 30 subjects were screened by questionnaire and were deemed to be free of inflammatory disease. Sample processing. PBMCs were isolated by Ficoll gradient centrifugation and then frozen at ⫺80°C in TRIzol reagent (Invitrogen, Carlsbad, CA) at the collecting centers, with documentation of both the time of phlebotomy and the time of freezing. The time to freezing, which was defined as the interval between these 2 times (expressed in minutes), was recorded for each sample. Frozen samples were batched and shipped on dry ice to CCHMC, where they were stored at ⫺80°C until RNA was extracted and purified on RNeasy columns (Qiagen, Germantown, MD). RNA quality was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) according to CCHMC Affymetrix GeneChip Core protocols. NuGEN Ovation Version 1 (NuGEN Technologies, San Carlos, CA) 1-round amplification was used, with 20 ng of starting RNA, to produce fluorescence labeled complementary DNA that was assayed by 54,675 probe sets on Affymetrix Human Genome U133 Plus 2.0 GeneChips (Affymetrix, Santa Clara, CA) and scanned with an Agilent G2500A GeneArray Scanner. A universal standard was prepared from a pooled mixture of RNA from 35 healthy adult volunteer donors (6) and was included in each batch of 12 samples to provide technical replicates for assessing batch-to-batch variations. Flow cytometry. Aliquots of PBMCs from 143 JIA patients and healthy controls (45 with polyarthritis, 21 with oligoarthritis, 25 with enthesitis-related arthritis, 6 with psoriatic arthritis, 12 with systemic arthritis, and 34 healthy controls) were frozen in 10% DMSO at the collecting centers, GENE EXPRESSION SIGNATURES IN SUBSETS OF POLYARTICULAR JIA 2115 Table 1. Characteristics of the polyarticular JIA subgroups* Age, mean ⫾ SD years No. (%) female No. (%) RF positive No. (%) anti-CCP positive§ No. (%) ANA positive¶ ESR, median (IQR) mm/hour** No. of joints with active disease, median (IQR) C-HAQ score, median (IQR)†† Physician’s global assessment of disease activity, by 10-point Likert scale, median (IQR) Patient’s/parent’s global assessment of well-being, by VAS, mean ⫾ SD mm‡‡ No. (%) with involvement of small joints Group A (n ⫽ 9) Group B (n ⫽ 17) Group C (n ⫽ 11) Other polyarticular JIA (n ⫽ 24) 11.3 ⫾ 3.6 7 (78) 5 (56)‡ 5 (56) 3 (33) 26 (10–30) 9 (9–14) 1.063 (0.375–1.875) 6 (5–7) 9.8 ⫾ 4.4 16 (94) 5 (29) 5 (29) 7 (41) 22 (18–49) 12 (8–21) 0.813 (0.375–1.375) 7 (5–9) 11.3 ⫾ 4.6 8 (73) 0 2 (20) 4 (36) 15 (5–28) 9 (8–11) 0.500 (0.125–1.000) 4 (4–7) 7.0 ⫾ 4.5† 21 (88) 4 (17) 6 (30) 15 (65)# 17 (11–28) 8 (5–11) 0.375 (0.000–1.250) 5 (4–7) 65 ⫾ 33 62 ⫾ 26 57 ⫾ 26 67 ⫾ 23 8 (89) 16 (94) 10 (91) 21 (88) * Hierarchical clustering of probe sets distinguished 3 subgroups of polyarticular juvenile idiopathic arthritis (JIA) patients (groups A–C), as well as patients who did not cluster into any subgroup (other polyarticular JIA). RF ⫽ rheumatoid factor (positive results on 2 occasions measured at least 3 months apart within the first 6 months of disease); IQR ⫽ interquartile range; VAS ⫽ visual analog scale (0–100 mm [poor to well]). † P ⫽ 0.018 versus groups A, B, and C combined, by one-way analysis of variance. ‡ P ⫽ 0.008 versus group C, by Fisher’s exact test. § Anti–cyclic citrullinated peptide (anti-CCP) was not tested in 1 patient in group C and 4 patients in the other polyarticular JIA group. ¶ Antinuclear antibody (ANA) was not tested in 1 patient in the other polyarticular JIA group. # P ⫽ 0.026 versus groups A, B, and C combined, by Fisher’s exact test. ** Erythrocyte sedimentation rate (ESR) was not tested in 2 patients in group A, 1 in group C, and 6 in the other polyarticular JIA group. †† The Childhood Health Assessment Questionnaire (C-HAQ) was not administered to 1 patient each in group A, group B, and the other polyarticular JIA group. The range of possible scores is 0–3. ‡‡ Patient’s/parent’s global assessment of well-being was not assessed in 1 patient each in groups A and B, and in 2 patients in the other polyarticular JIA group. shipped to CCHMC on dry ice, and stored in liquid nitrogen. These samples represented a subset of patients involved in a larger study we conducted, in which we compared gene expression profiles in PBMCs from patients with various JIA subtypes with those in healthy controls (10), and included 45 of the 61 patients with polyarticular JIA studied here. Sixteen polyarticular JIA patients were not sampled for flow cytometry due to phlebotomy volume limits relative to age. After thawing and washing PBMCs in fluorescenceactivated cell sorter buffer (phosphate buffered saline with 0.2% bovine serum albumin), cells were stained with the following monoclonal antibodies in appropriate combinations (all from BD Biosciences, San Jose, CA): peridinin chlorophyll A protein (PerCP)–conjugated CD3, fluorescein isothiocyanate (FITC)–conjugated CD8a, allophycocyanin (APC)–conjugated CD4, phycoerythrin (PE)–conjugated CD15, FITC-conjugated CD16, APC-conjugated CD33, FITC-conjugated CD34, and PerCP-conjugated CD45. Stained cells were analyzed on a FACSCalibur flow cytometer using CellQuest software (both from BD Biosciences). PBMC subpopulations of interest were captured by standardized polygonal gates. Linear correlation analysis was performed between the average normalized expression of the 50 probe sets of signatures I, II, and III (see Results), with the proportions of PBMC subpopulations determined by flow cytometry. Statistical analysis. Data were imported into GeneSpring GX 7.3.1 software (Agilent Technologies) with robust multichip analysis (11) preprocessing, referenced to the median of each gene’s adjusted value across all samples, followed by distance-weighted discrimination normalization to adjust for batch-to-batch variations (12). Probe sets for differentially expressed genes were identified by analysis of variance (ANOVA) at a 5% false discovery rate (13). Tukey’s post hoc testing identified probe sets for genes that were differentially expressed between JIA patients and healthy controls. The complete microarray dataset has been deposited in the Gene Expression Omnibus at the National Center for Biotechnology Information (NCBI) and is accessible through GEO Series accession number GSE13849. RESULTS General approach to data analysis. In a preceding study of gene expression differences between various subtypes of JIA and controls, which is published elsewhere in this issue of Arthritis & Rheumatism (10), we used ANOVA to identify 873 probe sets for genes that were differentially expressed in patients with RF⫺ polyarticular JIA (n ⫽ 45) as compared with healthy control subjects (n ⫽ 59). That comparison was the starting point for the present study, in which we used an iterative approach, with several rounds of identifying probe sets for genes that were differentially expressed between comparison groups, each time refining the analysis based on hierarchical clustering of differentially expressed genes, ultimately defining prototypical gene expression 2116 signatures. The first iteration included RF⫹ polyarticular JIA patients, the next involved subgroups of polyarticular JIA, and the final comparisons involved small cohorts of prototypical patients within each subgroup. Prototypical gene expression signatures were then applied to all study subjects to assess the association of each signature with clinical and biologic features within the entire population. Identification of polyarticular JIA subgroups clustered by PBMC gene expression patterns. For the first comparison, RF⫹ polyarticular JIA patients were included, since it was observed that many RF⫹ patients coclustered with a subset of RF⫺ patients rather than clustering separately, suggesting that RF status did not clearly distinguish a subset of polyarticular JIA assessed by PBMC gene expression profiling. Figure 1 shows an expression heat map of the 873 starting probe sets (10) in 61 polyarticular JIA patients (both RF⫹ and RF⫺) and 59 healthy controls. Subjects were ordered by hierarchical clustering, using Pearson’s correlation with average linkage. Three clusters, predominantly composed of JIA patients, are shown in the boxed areas of Figure 1. The JIA patients within these 3 clusters comprise groups A, B, and C. An alternative approach to clustering was tested in which 1,523 probe sets identified by ANOVA as being differentially expressed between all polyarticular JIA patients (RF⫺ and RF⫹; n ⫽ 61) and controls were used (data not shown). This approach also demonstrated 3 groups (A, B, and C), with few differences in the group assignments as compared with the original clustering. Most importantly, all prototypical patients for each group remained coclustered, and thus, the prototype signatures that were ultimately derived from all probe sets were identical using either approach. The next round of comparisons identified gene expression differences between each JIA subgroup (groups A, B, and C) and the healthy controls. Probe sets for genes that were identified by t-test using a 5% false discovery rate (multiple testing correction) as being differentially expressed comprised 2,111 for group A (1,025 up-regulated and 1,086 down-regulated), 6,713 for group B (2,900 up-regulated and 3,813 downregulated), and 11,485 for group C (5,319 up-regulated and 6,166 down-regulated). Notably, comparison of controls with the 24 polyarticular JIA patients that did not cluster into any subgroup did not identify any probe sets for differentially expressed genes, even when using a 20% false discovery rate. Comparisons among the groups indicated some overlap, particularly between groups B and C (582 probe sets between groups A and C, 756 GRIFFIN ET AL Figure 1. Expression heat map for 873 probe sets identified by analysis of variance as being differentially expressed between rheumatoid factor (RF)–negative patients with polyarticular juvenile idiopathic arthritis (JIA; n ⫽ 45) and healthy controls (n ⫽ 59). The relative intensity of red or blue indicates higher or lower expression as compared with the mean of all samples for that probe set (yellow). A total of 61 patients with either RF⫹ or RF⫺ polyarticular JIA and 59 healthy controls are represented. Polyarticular JIA patients are indicated by the solid circles on the left. Subjects are ordered by gene expression clustering using Pearson’s correlation with average linkage. Boxed areas indicate the 3 clusters containing predominantly JIA patients, and the JIA patients from these clusters comprise groups A, B, and C, as indicated on the right. The JIA patients in groups A, B, and C who served as prototypes for defining gene expression signatures I, II, and III, respectively, are indicated by the solid circles on the right. probe sets between groups A and B, and 2,555 probe sets between groups B and C). GENE EXPRESSION SIGNATURES IN SUBSETS OF POLYARTICULAR JIA 2117 Table 2. Top 25 probe sets of the 50 probe sets that defined gene expression signatures I, II, and III* Signature I Signature II Signature III Probe set Gene symbol Fold 1 Probe set Gene symbol Fold 1 Probe set Gene symbol Fold 1 1555728_a_at 205568_at 216951_at 214021_x_at 241981_at 210873_x_at 208791_at 206655_s_at 218660_at 210119_at 233749_at 206026_s_at 212651_at 211163_s_at 206343_s_at 207808_s_at 214469_at 1555659_a_at 206493_at 237563_s_at 231711_at 216243_s_at 229967_at 226303_at 214073_at MS4A4A AQP9 FCGR1A ITGB5 FAM20A APOBEC3A CLU GP1BB DYSF KCNJ15 MSN TNFAIP6 RHOBTB1 TNFRSF10C NRG1 PROS1 HIST1H2AE TREML1 ITGA2B LOC440731 – IL1RN CMTM2 PGM5 CTTN 4.63 4.27 4.09 4.08 3.89 3.79 3.75 3.62 3.55 3.45 3.32 3.28 3.27 3.27 3.21 3.13 3.09 3.08 3.08 3.02 3.02 2.97 2.96 2.95 2.94 216248_s_at 38037_at 205239_at 230170_at 208078_s_at 201694_s_at 242904_x_at 242397_at 202768_at 1557285_at 241824_at 214696_at 204470_at 1556874_a_at 209189_at 1568665_at 204014_at 226578_s_at 237082_at 1559203_s_at 204794_at 243213_at 207630_s_at 202861_at 201465_s_at NR4A2 HBEGF AREG OSM SNF1LK EGR1 – OLR1 FOSB LOC653193 FOSL2 MGC14376 CXCL1 RKHD2 FOS RNF103 DUSP4 DUSP1 DDEF1 KRAS DUSP2 STAT3 CREM PER1 JUN 16.38 10.36 7.93 7.93 6.59 6.35 5.55 5.53 5.47 4.92 4.41 4.38 4.37 4.32 4.25 4.19 4.18 4.17 3.99 3.70 3.65 3.61 3.51 3.38 3.28 239162_at 237001_at 240544_at 1557811_a_at 239448_at 1556865_at 244860_at 241154_x_at 232522_at 1559723_s_at 206548_at 1561166_a_at 235701_at 1561167_at 238544_at 215597_x_at 227062_at 240254_at 232882_at 1557555_at 238812_at 1556493_a_at 244682_at 242801_at 1569578_at DAPK1 NIBP ZFAND3 – SMAD3 PACSIN2 – MTSS1 TCF7L2 C9orf3 FLJ23556 FOXP1 R3HDM2 ETV6 IGF1R MYST4 TncRNA TNIK FOXO1A MAD1L1 ZA20D3 JMJD2C CAMSAP1 WWOX ANKRD11 8.05 7.71 7.35 7.23 7.20 6.87 6.78 6.74 6.73 6.58 6.39 6.17 6.16 6.16 6.10 5.99 5.97 5.85 5.80 5.71 5.56 5.53 5.51 5.49 5.48 * Fold 1 represents the ratio of the geometric mean of the prototype samples for each group (signatures I, II, and III, corresponding to groups A, B, and C, respectively) to the geometric mean of the 49 signature-free healthy controls (see Results for details). Characteristics of polyarticular JIA subgroups. Table 1 shows a comparison of the JIA patients in groups A, B, and C, as well as the JIA patients with polyarticular disease who did not cluster into any subgroup (other polyarticular JIA). Notable characteristics include a greater proportion of RF⫹ patients and a trend toward a greater proportion of anti-CCP⫹ patients in group A, a younger age and greater proportion of ANA⫹ patients in the other polyarticular JIA group, and a trend toward higher ESR, C-HAQ scores, and physician’s global assessment scores in groups A and B. Generally, there was proportionate representation of patients from each clinical center in groups A and C. Distribution of the 9 patients in group A from 5 centers was 1, 2, 3, 1, and 2. Distribution of the 11 patients in group C was 5, 2, 2, 2, and 0. Conversely, group B had a disproportionate distribution (3, 11, 1, 1, and 1), with 11 of 17 patients coming from 1 center. Identification of prototypical patients for each polyarticular JIA subgroup. For the final round of comparisons, prototypical patients were identified within each subgroup that maximized the gene expression signature for that subgroup and minimized overlap with other subgroups. Patients were included if the geometric mean of the 50 most up-regulated probe sets in that subgroup was at least 2-fold greater than that in the controls. Patients were excluded if they did not remain coclustered using several alternative hierarchical clustering methods (Pearson’s correlation, standard correlation, and distance correlation) or if the geometric mean of the 50 most up-regulated probe sets for more than 1 subgroup was increased more than 2-fold over that in the controls. Using this method, we identified 4 prototypical patients for group A and 5 prototypical patients each for groups B and C (solid circles on the right side of Figure 1). Determination of probe sets that define subgroupspecific gene expression signatures. Probe sets for genes differentially expressed between prototypical patients in each subgroup and the healthy controls were identified using t-tests with a 5% false discovery rate. For this comparison, 10 controls that had a more than 2-fold increase in the geometric mean of the 50 most upregulated probe sets for any JIA subgroup (A, B, or C) were excluded, yielding a cohort of 49 signature-free controls. This comparison produced lists of 4,098 probe sets for group A, 7,290 for group B, and 16,462 for group C (Supplementary Tables 1, 2 and 3, available on the 2118 Arthritis & Rheumatism Web site at http://www3. interscience.wiley.com/journal/76509746/home). Gene expression signatures for each subgroup (signatures I, II, and III for groups A, B, and C, respectively) were defined as the 50 most up-regulated probe sets from these lists, excluding redundant probe sets for the same genes (using the probe set with the lowest P value) and excluding probe sets that were up-regulated more than 1.5-fold in any of the other prototypical groups. Down-regulated probe sets were not included, since these probe sets had patterns similar to those of the up-regulated probe sets in each subgroup, and inclusion of down-regulated probe sets would not have significantly changed the discriminatory power of the signatures, although they may provide important clues to understanding the processes represented by each signature and are therefore included in Supplementary Tables 1–3. The top 25 probe sets for signatures I, II, and III are presented in Table 2, and all 50 probe sets for each signature are presented in Supplementary Tables 1–3 (available on the Arthritis & Rheumatism Web site at http://www3.interscience.wiley.com/journal/ 76509746/home). Quantification of subgroup-specific gene expression signatures. Signatures I, II, and III were quantified for all patients by calculating the average of the geometric mean of the fold increase over control values for the 50 probe sets that comprised each signature. A value of 1 equates to the average signature observed in the 49 signature-free controls. The magnitudes of each signature in every JIA patient are graphed in Figure 2, using the same order of patients as in Figure 1 for comparison. As expected, patients in group A predominantly expressed signatures I, and patients in group C predominantly expressed signature III. In contrast, patients in group B not only expressed signature II, but many patients in group B also expressed significant amounts of either signature I or III, suggesting overlap between groups B and A and between groups B and C. Overall, the numbers of JIA patients that exhibited at least a 1.5-fold increase in a particular signature were 16 (26.2%) for signature I, 17 (27.9%) for signature II, and 20 (32.8%) for signature III. Quantification of the gene signatures enabled correlation with the clinical characteristics and the abundance of PBMC subsets determined by flow cytometry. Correlation of signature I expression with rheumatoid factor and monocytes. Group A contained the highest proportion of RF⫹ patients (5 of 9 patients) (Table 1), which suggests that signature I expression correlated with RF positivity. To test this, the average GRIFFIN ET AL Figure 2. Comparison of gene expression signatures I, II, and III in patients with polyarticular juvenile idiopathic arthritis (JIA) and healthy controls. The magnitude of each signature in each of the 120 study subjects was calculated as the average fold increase in expression of the 50 probe sets that defined each signature over the mean of those 50 probe sets in the cohort of 49 signature-free controls. Vertical broken lines indicate the 2-fold increase over the mean in the controls; baselines indicate the mean in the controls (value of 1). Subjects are identical to, and in the same order as, those shown in Figure 1. Polyarticular JIA patients are indicated by the solid circles on the left: single symbols are rheumatoid factor (RF)⫺; double symbols are RF⫹. Boxed areas indicate the 3 clusters containing predominantly JIA patients (groups A, B, and C, as indicated on the right). The JIA patients in groups A, B, and C who served as prototypes for defining gene expression signatures I, II, and III, respectively, are indicated by the solid circles on the right. GENE EXPRESSION SIGNATURES IN SUBSETS OF POLYARTICULAR JIA 2119 Figure 3. Correlation of gene expression signatures with rheumatoid factor (RF) status, monocyte lineage markers, and time to freezing of peripheral blood mononuclear cells (PBMCs) from patients with juvenile idiopathic arthritis (JIA). A, Average magnitude of each gene expression signature (I, II, and III) in RF⫹ (n ⫽ 14) and RF⫺ (n ⫽ 47) polyarticular JIA patients. Values are the mean and SD. Only signature I gene expression was statistically significantly different in RF⫹ versus RF⫺ patients, by t-test. B, Correlation coefficients for gene expression signature I (solid bars), signature II (open bars), and signature III (shaded bars) in relation to monocyte lineage subsets, as determined by flow cytometry. C, Average time between isolation of PBMCs and freezing (time to freezing) of samples from polyarticular JIA patients in group B (n ⫽ 16), all other polyarticular JIA patients (n ⫽ 44), and healthy controls (n ⫽ 59). Time to freezing was not documented for 1 patient in group B. Values are the mean and SD. D, Plot of time to freezing of PBMC samples from each subject against the magnitude of signature II gene expression for that subject. Solid circles indicate polyarticular JIA patients in group B; open circles indicate all other study subjects, including the healthy controls. magnitude of each signature was compared between RF⫹ and RF⫺ JIA patients, and indeed, RF⫹ patients had statistically greater signature I expression than did RF⫺ patients (Figure 3A). Expression of signature II did not differ between RF⫹ and RF⫺ patients, and signature III expression trended toward being greater in RF⫺ patients, which is consistent with the absence of any RF⫹ patients in group C (Table 1). RF⫹ polyarticular JIA is more likely than RF⫺ polyarticular JIA to involve joint damage (14), raising the possibility that expression of signature I is an indicator of active joint injury, which is also consistent with group A trending toward higher ESR and C-HAQ scores than the other subgroups (Table 1). Along these lines, the genes that comprise signature I suggest that this signature comes from monocytes (15), and the presence of signature I may indicate that these cells are either mediating or responding to joint damage. In particular, there was increased expression of monocyte markers FCGR1A (CD64) (4.09-fold) (Table 2) and CD14 (2.33-fold) (Supplementary Table 1) in prototypical group A patients (15). While the flow cytometry analysis did not include CD64 or CD14 markers, other markers of monocyte lineage subsets (16) correlated with signature I better than with the other signatures, demonstrating an association of signature I with increased abundance of monocyte populations (Figure 3B). Correlation of signature II gene expression with prolonged sample processing. The biologic basis for the expression of signature II genes was first suggested by the observation that patients from 1 clinical center were overrepresented in this subgroup, which led to the discovery that samples from patients in group B tended to have had longer processing times, as measured by the time to freezing. The average time to freezing for group B samples was significantly greater than that for samples from other polyarticular JIA patients or healthy controls (Figure 3C). However, prolonged time to freezing was 2120 GRIFFIN ET AL Figure 4. Correlation of signature III expression with CD8b expression and with the abundance of CD8⫹ T cells in patients with juvenile idiopathic arthritis (JIA). A, Relative expression of CD8b mRNA in polyarticular JIA patients in group C (n ⫽ 11), all other polyarticular JIA patients (n ⫽ 50), and healthy controls (n ⫽ 59). Values are the mean and SD. B, Percentage of CD3⫹CD8⫹ peripheral blood mononuclear cells (PBMCs) in polyarticular JIA patients in group C (n ⫽ 8), all other polyarticular JIA patients (n ⫽ 36), and healthy controls (n ⫽ 28), as determined by flow cytometry. Samples were available for flow cytometry from only 72 of the 120 study subjects. Values are the mean and SD. C, Plot of the relative expression of CD8b mRNA in each study subject against the magnitude of signature III expression for that subject. Solid circles indicate polyarticular JIA patients in group C; open circles indicate all other study subjects, including the healthy controls. D, Plot of the percentage of CD3⫹CD8⫹ PBMCs in each study subject against the magnitude of signature III expression for that subject. Solid circles indicate polyarticular JIA patients in group C; open circles indicate all other study subjects, including the healthy controls. not sufficient to produce signature II expression, since a number of control and JIA samples did not exhibit this signature despite having a relatively prolonged time to freezing (Figure 3D), which suggests differential susceptibility of samples to the length of processing. Interestingly, the data graphed in Figure 3D also suggest that if all samples were processed promptly, with a time to freezing of ⬃60 minutes, signature II expression would not have been observed at all. Signature II gene expression did not correlate with any PBMC subpopulations analyzed by flow cytometry (all correlation coefficients ⬍0.3), indicating that this signature was not associated with the abundance of any particular PBMC subset. Rather, the association with longer time to freezing suggests that signature II expression was due to activation of PBMCs after phlebotomy. Correlation of signature III expression with reduced CD8ⴙ T cells and increased plasmacytoid dendritic cells. As noted above, each JIA subgroup had both up-regulated and down-regulated genes, and up- regulated genes were sufficient to define the gene expression signatures. Nevertheless, down-regulated genes are likely to provide valuable information regarding disease processes, and along these lines, CD8b was noted to be among the most highly down-regulated genes in group C patients (0.44-fold as compared with controls) (Supplementary Table 3). CD8a was also decreased, although not as strongly as CD8b (0.76-fold as compared with controls) (Supplementary Table 3). Since the abundance of CD8⫹ T cells had been measured by flow cytometry, it was of interest to assess the correlation of CD8b expression with the proportion of CD8⫹ T cells in group C patients. Indeed, both CD8b expression and CD8⫹ T cell abundance were significantly lower in group C patients compared with all other polyarticular JIA patients or with healthy controls (Figures 4A and B). Additionally, CD8b expression and abundance of CD8⫹ T cells trended to be lower in patients who exhibited strong expression of signature III (Figures 4C and D). The correlation coefficient for the GENE EXPRESSION SIGNATURES IN SUBSETS OF POLYARTICULAR JIA expression of signature III with the abundance of CD8⫹ T cells was –0.14, which supports an inverse relationship. In contrast, signature III expression exhibited its strongest positive correlation with Lin–BDCA-4⫹ plasmacytoid dendritic cells (17), with a correlation coefficient of 0.30. These results suggest that reduced CD8⫹ T cells and increased plasmacytoid dendritic cells are important features of the disease process represented by signature III expression, which may involve the action of transforming growth factor ␤ (TGF␤), as discussed below. DISCUSSION Global gene expression analyses of PBMCs from a large cohort of patients with polyarticular JIA who had not been treated with DMARDs or biologic agents revealed striking subgroups within an otherwise uniform collection of patients with recent-onset polyarthritis. These gene expression profiles demonstrated biologically significant heterogeneity within the JIA population, with distinct subgroups of patients expressing distinct gene expression signatures. These signatures appeared to be manifestations of biologic processes that are occurring in some, but not all, patients, and in the long run, they may prove to be valuable tools for classifying and managing polyarticular JIA. Three distinct gene expression signatures with variable expression among polyarticular JIA patients were identified. Signature I was strongly expressed in many RF⫹ JIA patients, but was also expressed in a number of RF⫺ JIA patients (Figures 2 and 3A). In fact, signature I may prove useful for identifying a subset of RF⫺ JIA patients that have a disease phenotype similar to that in RF⫹ patients, as described by Ravelli and Martini (3). Signature I contains many genes specifically expressed in monocytes, and thus, this signature may be a manifestation of increased abundance and/or activation of peripheral blood monocytes. A number of genes in signature I have been identified as being regulated in monocytes from patients with other rheumatic diseases. For example, FCGR1A (CD64), the third-ranked gene in signature I, was shown by Wijngaarden et al (18) to be dramatically downregulated in monocytes from patients with RA following methotrexate treatment. Likewise, MS4A4A, the topranked gene in signature I, and FCGR1A were the top 2–ranked cell surface marker genes in isolated monocytes that Abe et al (19) found to be significantly down-regulated in Kawasaki disease following treatment with intravenous immunoglobulin. Additionally, MS4A4A, CLU, DYSF, and IL8RB were found by Ogilvie 2121 et al (7) to be up-regulated in PBMCs from patients with active systemic JIA. A very interesting comparison with 25 genes that have been reported to be up-regulated in RA (20) showed that 16 of these genes were also up-regulated in group A prototypes, including CD14, AQP9, and several S100 proteins, while in contrast, 14 of these up-regulated RA genes were down-regulated in group C prototypes, emphasizing the contrast between groups A and C (see Supplementary Tables 1 and 3, available on the Arthritis & Rheumatism Web site at http://www3. interscience.wiley.com/journal/76509746/home). Thus, signature I may be indicative of monocyte activity in a number of autoimmune inflammatory diseases, and it may be useful for assessing disease activity and monitoring response to treatments in these diseases. Rigorous protocols were followed for isolating and freezing PBMCs as quickly as possible to minimize the effects of processing on gene expression. Nevertheless, some of the observed gene expression differences that contribute to signature II were associated with prolonged processing times. Still, this signature does not appear to have arisen solely as a function of processing. It was also dependent on immunologic differences between subjects, since this signature was observed almost exclusively in JIA patients and not in healthy controls, and many of those JIA patients also expressed signature I and/or signature III genes. Signature II contains many immediate early genes, including several FOS-related genes, JUN, and EGR1 (21), which is consistent with very recent cellular responses. The association of signatures I and III with antigen-presenting cells (monocytes and plasmacytoid dendritic cells, respectively) suggests that signature II expression may develop after phlebotomy in samples that have pathologically primed antigen-presenting cells that are able to readily stimulate other cells in the test tube environment, leading to rapid induction of immediate early genes. Furthermore, the pathologic relevance of signature II is supported by the presence of many genes that have previously been associated with autoimmune arthritis. For example, NR4A2 (also called NURR1), the first-ranked gene in signature II, is markedly up-regulated in RA synovium (22). Likewise, OSM, the third-ranked gene in signature II, is detectable in JIA synovial fluid, and it was shown to induce joint inflammation in a murine adenoviral gene–transfer model (23). Thus, signature II may provide valuable information regarding pathologic processes in the samples subjected to prolonged processing that may not otherwise have been observed had the processing been performed more promptly. 2122 Signature III was associated with the absence of RF (Figure 3) and appears to be distinct from signature I. In fact, few patients expressed both signatures I and III together (Figure 2), suggesting that these signatures are manifestations of independent biologic processes. Like group A patients, the average age of the group C patients was greater than that of the entire population of patients with polyarticular JIA (mean ⫾ SD 9.2 ⫾ 4.6 years), supporting the concept that signature III expression identified a distinct subset of JIA patients. Additionally, group C patients showed a trend toward lower ESR, lower C-HAQ scores, and lower physician’s global assessment of disease activity (Table 1), all of which are consistent with signature III expression identifying a less inflammatory subset of polyarticular JIA, possibly similar to the “dry synovitis” subset described by Ravelli and Martini (3). Moreover, signature III was associated with low numbers of CD8⫹ T cells (Figure 4) and an increased abundance of blood dendritic cell antigen 4 (BDCA-4)– positive plasmacytoid dendritic cells, and it contains many genes that are inducible by TGF␤ and are potentially involved in mediating or regulating the action of TGF␤. These include the first-ranked gene, DAPK1, which is a proapoptotic protein that can be induced by TGF␤ via SMAD activation (24). Other signature III genes that are TGF␤-inducible include SMAD3, BCL2, MAPK1, and FOXO3A (25–28). These observations suggest that TGF␤ may be responsible for reducing levels of CD8⫹ T cells via a proapoptotic influence and increasing plasmacytoid dendritic cells via an activating effect. Notably, increased levels of TGF␤ have been detected in synovial fluid from patients with juvenile arthritis (29), where it has been conjectured to serve an immunosuppressive role in countering synovial inflammation (30). Interestingly, while one might conjecture that exposure of PBMCs to TGF␤ occurs in the inflamed synovium, another intriguing possibility is that it occurs via regulatory T cells that express TGF␤ (31). Many of the polyarticular JIA patients did not express signatures I, II, or III (24 of the 61 patients evaluated [39%]). This group of patients was statistically significantly younger and had a higher rate of ANA positivity than the rest of the polyarticular JIA patients. It is uncertain whether these patients represent a unified group, but one can speculate that they do not have the same disease phenotype as patients that expressed signature I and/or signature III, and given their age and ANA status, it is likely that many of these patients comprise a polyarticular JIA subset that closely resembles early-onset ANA-positive oligoarticular JIA, as GRIFFIN ET AL described by Ravelli et al (3,32). It is also noted that a few healthy control subjects clustered into group A, B, or C. While control subjects were deemed free of inflammatory disease based on a screening questionnaire, there was no assurance that every control subject was completely healthy, and this may account for coclustering of some controls with the JIA patients. In summary, we have demonstrated PBMC gene expression signatures in a large population of children with recent-onset polyarticular JIA that correlate with differences in disease characteristics. These signatures support the subclassification of polyarticular JIA offered by Ravelli and Martini (3), with signature I identifying both RF⫹ and RF⫺ patients with disease similar to that of adult RA, signature III identifying a less inflammatory disease subset, and patients with ANA⫹ early-onset disease expressing neither signature. Thus, these signatures offer a molecular classification of polyarticular JIA and may prove to be valuable tools for assessing disease activity, predicting response to medications, and forecasting long-term outcomes. ACKNOWLEDGMENTS We acknowledge and appreciate the contributions of the following individuals: Lori Luyrink, Shweta Srivastava, Ndate Fall, and Sarah Croswell (research assistants, CCHMC); Wendy Bommer and Anne Johnson (clinical research coordinators, CCHMC); Lukasz Itert (data processing and management, CCHMC); Jesse Gillis (data preprocessing, University of British Columbia); Marsha Malloy (study coordinator, Children’s Hospital of Wisconsin); Beth Martin (study coordinator, Toledo Children’s Hospital); Marilyn Orlando (study coordinator, Schneider Children’s Hospital); Sara Jane Wilson (study coordinator, Children’s Hospital of Philadelphia); and Jeremy Zimmermann (data collection, Children’s Hospital of Wisconsin). AUTHOR CONTRIBUTIONS All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Griffin 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 conception and design. Griffin, Barnes, Ilowite, Olson, Sherry, Gottlieb, Aronow, Thompson, Grom, Colbert, Glass. Acquisition of data. Griffin, Barnes, Ilowite, Olson, Sherry, Gottlieb, Thornton, Thompson, Grom, Colbert, Glass. Analysis and interpretation of data. Griffin, Barnes, Aronow, Pavlidis, Hinze, Thornton, Thompson, Grom, Colbert, Glass. REFERENCES 1. 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. GENE EXPRESSION SIGNATURES IN SUBSETS OF POLYARTICULAR JIA 2. Petty RE, Southwood TR, Manners P, Baum J, Glass DN, Goldenberg J, et al. International League of Associations for Rheumatology classification of juvenile idiopathic arthritis: second revision, Edmonton, 2001. J Rheumatol 2004;31:390–2. 3. Ravelli A, Martini A. Juvenile idiopathic arthritis. Lancet 2007; 369:767–78. 4. Lovell DJ. Update on treatment of arthritis in children: new treatments, new goals. Bull NYU Hosp Jt Dis 2006;64:72–6. 5. 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. 6. Fall N, Barnes M, Thornton S, Luyrink L, Olson J, Ilowite NT, et al. Gene expression profiling of peripheral blood from patients with untreated new-onset systemic juvenile idiopathic arthritis reveals molecular heterogeneity that may predict macrophage activation syndrome. Arthritis Rheum 2007;56:3793–804. 7. Ogilvie EM, Khan A, Hubank M, Kellam P, Woo P. Specific gene expression profiles in systemic juvenile idiopathic arthritis. Arthritis Rheum 2007;56:1954–65. 8. 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;6:R15–32. 9. Allantaz F, Chaussabel D, Stichweh D, Bennett L, Allman W, Mejias A, et al. Blood leukocyte microarrays to diagnose systemic onset juvenile idiopathic arthritis and follow the response to IL-1 blockade. J Exp Med 2007;204:2131–44. 10. Barnes MG, Grom AA, Thompson SD, Griffin TA, Pavlidis P, Itert L, et al. Subtype-specific peripheral blood gene expression profiles in recent-onset juvenile idiopathic arthritis. Arthritis Rheum 2009;60:2102–12. 11. Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 2003;4:249–64. 12. Benito M, Parker J, Du Q, Wu J, Xiang D, Perou CM, et al. Adjustment of systematic microarray data biases. Bioinformatics 2004;20:105–14. 13. Hochberg Y, Benjamini Y. More powerful procedures for multiple significance testing. Stat Med 1990;9:811–8. 14. Gilliam BE, Chauhan AK, Low JM, Moore TL. Measurement of biomarkers in juvenile idiopathic arthritis patients and their significant association with disease severity: a comparative study. Clin Exp Rheumatol 2008;26:492–7. 15. Zarev PV, Davis BH. Comparative study of monocyte enumeration by flow cytometry: improved detection by combining monocyte-related antibodies with anti-CD163. Lab Hematol 2004;10: 24–31. 16. Stec M, Weglarczyk K, Baran J, Zuba E, Mytar B, Pryjma J, et al. Expansion and differentiation of CD14⫹CD16⫺ and CD14⫹⫹ CD16⫹ human monocyte subsets from cord blood CD34⫹ hematopoietic progenitors. J Leukoc Biol 2007;82:594–602. 17. Dzionek A, Fuchs A, Schmidt P, Cremer S, Zysk M, Miltenyi S, et al. BDCA-2, BDCA-3, and BDCA-4: three markers for distinct subsets of dendritic cells in human peripheral blood. J Immunol 2000;165:6037–46. 18. Wijngaarden S, van Roon JA, van de Winkel JG, Bijlsma JW, Lafeber FP. Down-regulation of activating Fc␥ receptors on 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 2123 monocytes of patients with rheumatoid arthritis upon methotrexate treatment. Rheumatology (Oxford) 2005;44:729–34. 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. Bovin LF, Rieneck K, Workman C, Nielsen H, Sorensen SF, Skjodt H, et al. Blood cell gene expression profiling in rheumatoid arthritis: discriminative genes and effect of rheumatoid factor. Immunol Lett 2004;93:217–26. Iwaki K, Sukhatme VP, Shubeita HE, Chien KR. ␣- and ␤adrenergic stimulation induces distinct patterns of immediate early gene expression in neonatal rat myocardial cells. fos/jun expression is associated with sarcomere assembly; Egr-1 induction is primarily an ␣1-mediated response. J Biol Chem 1990;265:13809–17. McEvoy AN, Murphy EA, Ponnio T, Conneely OM, Bresnihan B, FitzGerald O, et al. Activation of nuclear orphan receptor NURR1 transcription by NF-B and cyclic adenosine 5⬘-monophosphate response element-binding protein in rheumatoid arthritis synovial tissue. J Immunol 2002;168:2979–87. De Hooge AS, van de Loo FA, Bennink MB, Arntz OJ, Fiselier TJ, Franssen MJ, et al. Growth plate damage, a feature of juvenile idiopathic arthritis, can be induced by adenoviral gene transfer of oncostatin M: a comparative study in gene-deficient mice. Arthritis Rheum 2003;48:1750–61. Jang CW, Chen CH, Chen CC, Chen JY, Su YH, Chen RH. TGF-␤ induces apoptosis through Smad-mediated expression of DAP-kinase. Nat Cell Biol 2002;4:51–8. Yang X, Letterio JJ, Lechleider RJ, Chen L, Hayman R, Gu H, et al. Targeted disruption of SMAD3 results in impaired mucosal immunity and diminished T cell responsiveness to TGF-␤. EMBO J 1999;18:1280–91. Prehn JH, Bindokas VP, Marcuccilli CJ, Krajewski S, Reed JC, Miller RJ. Regulation of neuronal Bcl2 protein expression and calcium homeostasis by transforming growth factor type ␤ confers wide-ranging protection on rat hippocampal neurons. Proc Natl Acad Sci U S A 1994;91:12599–603. Choi ME. Mechanism of transforming growth factor-␤1 signaling. Kidney Int Suppl 2000;77:S53–8. Atfi A, Abecassis L, Bourgeade MF. Bcr-Abl activates the AKT/ Fox O3 signalling pathway to restrict transforming growth factor␤-mediated cytostatic signals. EMBO Rep 2005;6:985–91. Gattorno M, Facchetti P, Ghiotto F, Vignola S, Buoncompagni A, Prigione I, et al. Synovial fluid T cell clones from oligoarticular juvenile arthritis patients display a prevalent Th1/Th0-type pattern of cytokine secretion irrespective of immunophenotype. Clin Exp Immunol 1997;109:4–11. Bucht A, Larsson P, Weisbrot L, Thorne C, Pisa P, Smedegard G, et al. Expression of interferon-␥ (IFN-␥), IL-10, IL-12 and transforming growth factor-␤ (TGF-␤) mRNA in synovial fluid cells from patients in the early and late phases of rheumatoid arthritis (RA). Clin Exp Immunol 1996;103:357–67. Nadkarni S, Mauri C, Ehrenstein MR. Anti-TNF-␣ therapy induces a distinct regulatory T cell population in patients with rheumatoid arthritis via TGF-␤. J Exp Med 2007;204:33–9. Ravelli A, Felici E, Magni-Manzoni S, Pistorio A, Novarini C, Bozzola E, et al. Patients with antinuclear antibody–positive juvenile idiopathic arthritis constitute a homogeneous subgroup irrespective of the course of joint disease. Arthritis Rheum 2005; 52:826–32.