SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Objective

The development of biomarkers to predict response to therapy in polyarticular juvenile idiopathic arthritis (JIA) is an important issue in pediatric rheumatology. A critical step in this process is determining whether there is biologic meaning to clinically derived terms such as “active disease” and “remission.” The aim of this study was to use a systems biology approach to address this question.

Methods

We performed gene transcriptional profiling on children who fulfilled the criteria for specific disease states as defined by the consensus criteria developed by Wallace and colleagues. The study group comprised children with active disease (n = 14), children with clinical remission on medication (CRM; n = 9), children with clinical remission off medication (CR; n = 6), and healthy control children (n = 13). Transcriptional profiles in peripheral blood mononuclear cells (PBMCs) were obtained using Affymetrix U133 Plus 2.0 arrays.

Results

Hierarchical cluster analysis and predictive modeling demonstrated that the clinically derived criteria represent biologically distinct states. Minimal differences were seen between children with active disease and those with disease in CRM. Thus, underlying immune/inflammatory abnormalities persist despite a response to therapy. The PBMC transcriptional profiles of children whose disease was in remission did not return to normal but revealed networks of proinflammatory and antiinflammatory genes, suggesting that remission is a state of homeostasis, not a return to a normal state.

Conclusion

Gene transcriptional profiling of PBMCs revealed that clinically derived criteria for JIA disease states reflect underlying biology. We also demonstrated that neither CRM nor CR status results in resolution of the underlying inflammatory process, but that these conditions are more likely to be states of balanced homeostasis between proinflammatory and antiinflammatory mechanisms.

Juvenile idiopathic arthritis (JIA) is a term used to designate a family of childhood-onset diseases that are characterized by chronic inflammation of synovial membranes. Because the etiology of JIA is unknown, therapy remains entirely empirical, is sometimes only marginally effective, and frequently is associated with unwanted side effects.

The empirical nature of therapy for JIA is one of the most vexing problems in the field of pediatric rheumatology. A critical question, which is often asked by parents as well as physicians, is when and whether children who are doing well on medication can have treatment with those medications reduced or discontinued. Answering this question relies on 2 suppositions: 1) there is something that can be called “remission” in JIA, and 2) remission can be identified on the basis of specific clinical or laboratory features of the disease. Unfortunately, neither of these suppositions is necessarily valid. Studies in the past 10 years have shown that a significant percentage of children with polyarticular JIA experience disease flares when methotrexate is discontinued, even when disease has been stable during treatment with that drug for years (1, 2). No reliable biomarker or set of biomarkers accurately separates those children fated to experience disease recurrence as methotrexate is discontinued from those children in whom treatment with the medication can safely be discontinued.

Only recently have investigators arrived at a consensus regarding the meaning of terms such as “active disease,” “inactive disease,” and “clinical remission” (3). Although the definitions of these terms have been validated clinically, it is currently unknown whether they actually represent distinct biologic states. The development of predictive biomarkers would certainly be facilitated if these distinct disease states could be identified biologically in children with treated disease.

Because conventional biomarkers have, to date, shown limited capacity to identify remission, we elected to use genome-wide transcription profiling to determine whether the clinically derived criteria for disease states represent underlying immunobiology in children with polyarticular, IgM–rheumatoid factor (RF)–negative JIA.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Patient population and definition of disease states.

We studied 14 children with active polyarticular RF-negative JIA, as defined by the International League of Associations for Rheumatology criteria (4). Because the long-term intent of this project is to identify children who can safely stop receiving medication, all patients studied here, with the exception of those studied while their disease was in clinical remission, were receiving medication at the time of study. All patients (except those whose disease was in clinical remission) were receiving oral or subcutaneous methotrexate; in addition, 5 children were receiving subcutaneous etanercept. We studied 9 children who fit the criteria for clinical remission while receiving medication. Because this was a cross-sectional study, children were studied only once as they achieved different disease states. Finally, we studied 6 children whose disease was in remission while they were not receiving medication.

Patients ranged in age from 3 years to 18 years and had had polyarticular JIA for 6 months to 12 years at the time of sampling. Blood was obtained at the time of routine clinical monitoring, under normal sanitary conditions; topical anesthesia with 2.5% lidocaine/2.5% prilocaine cream was provided to all children prior to the procedure.

Disease states were defined according to the consensus criteria developed by Wallace and colleagues (5), as follows: active disease (AD), which defines children with synovitis and/or fever, rash, lymphadenopathy, splenomegaly, uveitis, an elevated erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP) level, or a physician's global assessment score indicating active disease; inactive disease (ID), which defines children, including those who are and those who are not receiving medication, with no evidence of synovitis, the absence of fever, rash, lymphadenopathy, and splenomegaly, no active uveitis, a normal ESR and CRP level, and a physician's global assessment score indicating no active disease; clinical remission on medication (CRM), which defines children who are receiving medication and have inactive disease and in whom that state has been maintained for 6 continuous months; and clinical remission (CR), which defines children who are not receiving medication and have inactive disease and in whom that state has been maintained for 12 continuous months.

Healthy control subjects.

The control group comprised 13 healthy children, ages 3–15 years, who were undergoing elective surgery for noninflammatory conditions (e.g., minor orthopedic procedures) or were being seen for routine health maintenance in the Oklahoma University Children's Physicians general pediatrics clinic. Healthy children were excluded from the control group if they had experienced fever (temperature ≥38°C) in the 48 hours prior to phlebotomy. Topical anesthesia with 2.5% lidocaine/2.5% prilocaine was applied to the phlebotomy site in all children for at least 30 minutes before the procedure. The participation of all human subjects was reviewed and approved by the University of Oklahoma Health Sciences Center Institutional Review Board.

Specimens and specimen handling.

Whole blood was drawn into 10-ml citrated CPT tubes (catalog no. 362760; Becton Dickinson, Franklin Lakes, NJ). Peripheral blood mononuclear cells (PBMCs) were separated from granulocytes and red blood cells by density-gradient centrifugation and then were collected and placed immediately in TRIzol reagent (Invitrogen, Carlsbad, CA).

RNA isolation, labeling, hybridization, and scanning.

Total RNA extractions from TRIzol reagent were carried out according to the manufacturer's directions and were further purified by passage through RNeasy mini columns (Qiagen, Valencia, CA), according to manufacturer's protocols. Final RNA preparations were suspended in RNase-free water. The RNAs were quantified spectrophotometrically. RNA integrity was assessed using capillary gel electrophoresis (Agilent 2100 Bioanalyzer; Agilent, Palo Alto, CA) to determine the ratio of 28S:18S ribosomal RNA in each sample. A ratio greater than 1.0 was used to define samples of sufficient quality, and only samples with a ratio above this limit were used for microarray studies. Complementary DNA (cDNA) synthesis, hybridization, and staining were performed as specified by Affymetrix (Santa Clara, CA) using Affymetrix Human Genome U133 Plus 2.0 Arrays, an Affymetrix automated GeneChip Fluidics Station 450, and an Affymetrix Scanner 3000 7G.

Statistical analysis.

All preprocessing of Affymetrix array data was performed in the R/Bioconductor package, Affy. The raw Affymetrix Perfect Match probes were normalized by the robust multichip analysis method combined with median polish (6). The marginal data distributions were adjusted through quantile normalization. The resulting normalized values were imported into JMP Genomics software version 3.2 (Cary, NC), where they were then log transformed. Genes were filtered using the log expression variation filter to screen out genes that are not likely to be informative, based on the variance of each gene across the arrays. In this case, the filter was set to exclude genes that fell below the 50th percentile of gene variance. We identified genes that were differentially expressed between the 2 classes by using Student's 2-sample t-test (7). We used Student's t-test to provide a false discovery rate (FDR) of 5% (8). The FDR is the proportion of the list of genes claimed to be differentially expressed that are false-positive identifications.

The data were exported to Excel (Microsoft, Redmond, WA), where averages of the classes were used to calculate expression ratios. Genes that were simultaneously differentially expressed (<5% FDR), had a ratio 2-fold or larger, and for which the minimum normalized average intensity was >64 units in at least 1 group were retained for further analysis. Unsupervised hierarchical clustering was performed in Spotfire (Tibco, Sommerville, MA), using Ward's minimum variance method (9). Differences between cluster groups were tested using a chi-square test. P values less than 0.05 were considered significant.

Predictive modeling.

To predict group membership (i.e., disease state), a so-called “one-versus-many” approach was taken (10). Using this approach, the data first were broken into 2 groups for every predictive outcome. For example, subject 1 was either a control or not a control. This process was repeated for every variable (e.g., subject 1 either had active disease or did not have active disease). After all variables were dichotomized, each binary variable created was modeled using a logistic regression of the differentially expressed genes selected previously. Model terms were selected through a forward stepwise procedure. The concordance statistic was used to select the best model. Additionally, there were 2 restrictions. First, all terms in the model were statistically significant at α = 0.05. Second, due to the small sample size, a maximum of 5 terms were allowed in a single model.

Once the models were created, individuals were scored and assigned group membership. Every logistic regression was given a propensity score as belonging to a given group. Every individual was scored in all 4 models, and the model with the highest score determined classification. For example, a given patient entering the model might receive 4 scores: 5.2 for control, 2.3 for CRM, 3.9 for CR, and 2 for AD. Because the score of 5.2 is the highest, this patient would be classified as a control. In an attempt to avoid overfitting, we performed a 5-fold cross-validation of our model.

Physiologic pathway modeling.

Pathways of potential interactions between gene products were generated by placing only the statistically significantly differentially expressed genes between groups into Ingenuity Pathways Analysis software (Ingenuity Systems, Redwood City, CA). Each Affymetrix gene identifier was mapped to its corresponding gene object in the Ingenuity knowledge base. These “focus” genes were overlaid onto a global molecular network developed from information contained in the Ingenuity knowledge base. Networks of these focus genes were then algorithmically generated based on their “connectivity,” which was derived from known interactions between the products of these genes.

Real-time reverse transcription−polymerase chain reaction (RT-PCR) validation.

Total RNA was prepared as described above. Primers were designed with a 60°C melting temperature and a length of 15–28 nucleotides to produce PCR products with lengths between 50 bp and 150 bp, using Primer Express 2.0 software (Applied Biosystems, Foster City, CA). First strand cDNA was generated from 1.8 μg of total RNA per sample, using OmniScript reverse transcriptase according to manufacturer's directions (Qiagen). Complementary DNA was diluted in water at a ratio of 1:20. PCRs were run with 4 μl of cDNA template in 20-μl reactions in duplicate on an ABI 7000 Sequence Detection System, using ABI SYBR Green PCR Master Mix (Applied Biosystems) and gene-specific primers at a concentration of 0.2 μM each. The temperature profile consisted of an initial step at 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds, 60°C for 1 minute, and then a final melting curve analysis with a ramp from 60°C to 95°C over 20 minutes. Gene-specific amplification was confirmed by a single peak, using ABI Dissociation Curves software (Applied Biosystems). Average Ct values for GAPDH (run in parallel reactions to the genes of interest) were used to normalize average Ct values for the gene of interest. These values were used to calculate averages for each group (normal or patient subsets), and the relative ΔCt was used to calculate fold-change values between the groups.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Corroboration of array results by PCR.

Two genes were chosen from each of the comparisons (i.e., AD versus CRM, CRM versus CR, CR versus healthy controls) for corroboration of the array data. (The results are summarized in a table that can be viewed online at http://peds.ouhsc.edu/section_rheumatology.asp.) In all cases, the directional change (JIA versus controls) identified by the arrays were corroborated by real-time PCR analysis. These data are a subset of a larger group of RT-PCR corroborations (28 genes) for these same patients, comparing other disease states (e.g., AD versus CR). For all genes tested, those that were differentially overexpressed or underexpressed on microarrays were similarly overexpressed or underexpressed by quantitative PCR analysis.

Hierarchical cluster and 1 versus many analyses.

Hierarchical cluster analysis between groups, using the differentially expressed genes in PBMCs from microarray data, demonstrated that each of the different disease states, as defined by the consensus conference (5), could be distinguished from each other. As shown in Figure 1, gene expression profiles largely distinguished control from patient samples, with control samples clustered toward the left and samples from children with active disease clustered toward the right side of the largest cluster, which contained 37 of the samples (P = 6.3 × 10−4, by Fisher's exact test). An additional 7 samples (far right cluster in Figure 1) contained 4 CR, 1 CRM, and 2 additional AD samples. Most CRM samples clustered within the control or AD subclusters.

thumbnail image

Figure 1. Hierarchical cluster analysis of differentially expressed genes in peripheral blood mononuclear cells. C = control samples; CRM = clinical remission on medication; CR = clinical remission off medication; A = active disease.

Download figure to PowerPoint

Predictive modeling revealed a unique set of 10 genes across all 4 models, the expression levels of which accurately predicted the disease state (Table 1). The concordance analysis between the observed clinical state and the predicted clinical state by microarray data revealed that a correct diagnosis could be made in 42 of 52 individuals (80%) (Table 2).

Table 1. Genes discriminating the disease state
Affymetrix probe IDGene descriptionGene symbol
AI692879Discs, large homolog 1 (Drosophila)DLG1
AU155361Trophoblast-derived noncoding RNATncRNA
AV655640CCAAT/enhancer binding protein δCEBPD
BF511231Tissue factor pathway inhibitor (lipoprotein-associated coagulation inhibitor)TFPI
JO4152Tumor-associated calcium signal transducer 2TACSTD2
NM_002261Killer cell lectin-like receptor subfamily C, member 3KLRC3
NM_003514Histone cluster 1, H2amHIST1H2AM
NM_005160Adrenergic, β, receptor kinase 2ADRBK2
NM_006242Protein phosphatase 1, regulatory (inhibitory) subunit 3DPPP1R3D
NM_006877Guanosine monophosphate reductaseGMPR
Table 2. Cross-validation of disease states*
PredictedObserved
ControlPoly CRPoly CRMActive disease
  • *

    Poly = polyarticular; CR = clinical remission off medication; CRM = clinical remission on medication; AD = active disease.

Control14111
CR0401
CRM1250
AD00319

Network modeling.

When children with active disease were compared with children who had achieved clinical remission while receiving medication, we found 23 genes that were differentially expressed between the 2 groups, 22 of which were overexpressed in children with active disease. (A table annotating these genes and relative expression levels can be viewed online at http://peds.ouhsc.edu/section_rheumatology.asp.) As described above, all of these patients were receiving medication. In silico modeling of the array data was informative. Analysis of these differentially expressed genes (Figure 2) revealed a single network of interferon-γ (IFNγ)–, interleukin-6 (IL-6)–, and IL-4–regulated genes that we (11) and other investigators (12) have identified as important elements of JIA immunopathology. This physiologic model suggests that reaching CRM status is achieved by suppression of these IL-6–, IL-4–, and IFNγ-regulated networks. The single gene that showed decreased expression in children with active disease was the aldehyde dehydrogenase A1 family member (ALDH1A1), which is known to regulate sex steroid hormones and to be IL-1 responsive (13). It is noteworthy that insulin also appears as a central mediator in this network, which is an interesting finding given the emerging data demonstrating critical “cross-talk” between tumor necrosis factor α (TNFα)– and insulin-regulated pathways (14).

thumbnail image

Figure 2. Single network derived from Ingenuity analysis of differentially expressed genes, comparing children with polyarticular juvenile idiopathic arthritis (JIA) with active disease and children with JIA who had achieved clinical remission while receiving medication. This network consists largely of genes that show increased expression in children with active disease (red).

Download figure to PowerPoint

When children who had achieved CRM status were compared with children with CR status, we observed the persistence of linked proinflammatory networks in children with CRM, as shown in Figures 3A and B. In all, 39 genes distinguished these 2 patient groups. (A table annotating these genes and relative expression levels can be viewed online at http://peds.ouhsc.edu/section_rheumatology.asp.) Although it is impossible to determine how or whether these expression patterns are altered by medication (patients with CRM are still receiving medication, while patients with CR are not), it is worth noting that these networks consist of genes regulated by known leukocyte proinflammatory regulators (e.g., Jun, NF-κB) (Figure 3) as well as IFNγ- and TNFα-regulated genes, as we have previously reported (10). This finding may explain the tendency to misclassify CRM as AD, as shown in Table 2.

thumbnail image

Figure 3. Overlapping gene networks derived from Ingenuity analysis of the transcriptional profile of peripheral blood mononuclear cells (PBMCs), comparing children with juvenile idiopathic arthritis (JIA) who had achieved clinical remission while receiving medication (CRM) and those whose disease was in clinical remission while they were not receiving medication. Genes overexpressed in children with CRM status are shown in red. Note the clusters of genes regulated by the leukocyte activators Jun and NF-κB (left) and interferon-γ and tumor necrosis factor α (right). This suggests that, even during CRM, there is still an active proinflammatory response in PBMCs from patients with JIA.

Download figure to PowerPoint

The gene expression profile of PBMCs did not normalize in children whose disease was in remission, as indicated in the hierarchical cluster analysis (Figure 1). Genes in PBMCs that were differentially expressed between children with CR and control subjects included 74 up-regulated and 8 down-regulated genes. (A table annotating these genes and relative expression levels can be viewed online at http://peds.ouhsc.edu/section_rheumatology.asp.) Ingenuity analysis revealed 4 interconnected gene networks. The structure of the largest of these networks (Figure 4A) demonstrated genes that are known mediators of leukocyte activation (e.g., Jun and other MAPKs) (15, 16) as well as markers of inflammation (e.g., matrix metalloproteinases) (17). These genes are networked with transforming growth factor β1 (TGFβ1), which, depending on its physiologic context, is generally regarded as a negative regulator of inflammation and an important mediator of immune tolerance (16, 17).

thumbnail image

Figure 4. Comparison of patients with juvenile idiopathic arthritis (JIA) whose disease was in clinical remission and who were not receiving medication (CR) and healthy control subjects. A, The largest of the 4 functional gene networks derived from Ingenuity analysis of the transcriptional profile of children with JIA who had achieved CR status and healthy control children. Genes associated with leukocyte activation (e.g., Jun and other MAPKs) are networked with markers of leukocyte activation (e.g., matrix metalloproteinases). These genes, in turn, are counterregulated by genes known to modulate inflammation (e.g., transforming growth factor β). Genes overexpressed in patients with CR are highlighted in red, and genes underexpressed in patients with CR are shown in green. B and C, Persistence of tumor necrosis factor α–regulated hubs (B) and interleukin-4–regulated hubs (C) when children whose disease was in remission were compared with healthy control subjects. D, A fourth network, consisting of clusters of genes regulated by β-estradiol and dihydrotestosterone.

Download figure to PowerPoint

These findings reveal that remission is not a return to normalcy but, rather, a physiologic state in which proinflammatory elements are countered or kept in check. This interpretation is supported by the networks revealed in Figures 4B and C, which show the persistence of gene networks regulated by TNFα (Figure 4B) and IL-4 (Figure 4C), both of which are known to be involved in the immunopathology of JIA (18, 19). The fourth network (Figure 4D) consisted of genes regulated by both β-estradiol and dihydrotestosterone, which is an interesting finding in light of the known role of estrogens in regulating inflammation (20) and the female preponderance among patients with polyarticular JIA. This same network consists of genes regulated by CCAAT/enhancer binding protein α, a member of a family of proteins previously implicated in regulating IL-1β (21) and other aspects of inflammation, including regulation of cytokine expression within the rheumatoid synovium (22, 23). This transcription factor is also known to play a role in estrogen-mediated regulation of cytokine production (24, 25).

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

The development of clinically useful prognostic genomic biomarkers in JIA requires the ability to discern specific disease states (e.g., active disease, remission) as a first step. If clinical remission while receiving medication is indistinguishable from active disease, for example, it is unlikely that this technology will provide much assistance. That is, it is possible that the clinically derived consensus criteria for disease status do not reflect underlying disease biology. This study tested and confirmed the hypothesis that these disease states are distinguishable at the molecular level using gene expression profiling and thus provides an important first step in biomarker development.

We report here that gene profiling and hierarchical cluster analysis can distinguish different disease states, although there is more “blurring” of the groups at the biologic level than at the clinical level (Figure 1). This suggests that underlying cellular abnormalities persist in PBMCs from patients with JIA, even when treatment is successful in controlling symptoms. A plausible (but not currently provable) explanation for this observation is that the synovium is a more critical target of drug action than has previously been supposed. This hypothesis is supported by what we observed in children with inactive disease who had not received medication for at least 6 months (that is, children who had achieved CR status). Remission, as reflected at the molecular level, is clearly not a return to a normal immune/inflammatory state. Rather, the gene expression profiles suggest that remission is a state of homeostasis in which antiinflammatory (e.g., TGFβ-driven) mechanisms balance the dysregulatory elements that lead to chronic inflammation.

It is difficult to determine how accurate or predictive our models of the specific disease states in polyarticular JIA are, until we test them in an independent cohort. However, the use of the 5-fold cross-validation does provide validation that the model is not overfit. For example, the major misclassification occurred in the CRM group, in which 3 patients classified as CRM were predicted to have active disease. It is possible that our model is correct and that disease in these children is not in remission on a molecular level. This hypothesis is supported by Ingenuity modeling, which suggests that networks of proinflammatory genes are still active, even in these children who have achieved a state of remission. Under any circumstances, we will need to follow up these children over time to determine whether our cross-sectional model has prognostic capabilities.

Taken together, these findings explain 2 observations that have puzzled physicians caring for children with polyarticular JIA for many years. First, our data explain at least conceptually why recurrences or flares are so common when medications are tapered or discontinued in children who seem to be doing well: underlying abnormalities at the gene expression level are still present, even if such abnormalities are not reflected in standard clinical measures such as the ESR, the serum CRP level, the hemoglobin concentration, or the white blood cell count. These findings also explain why disease recurrences are common: remission is still a biologically abnormal state. Although it is impossible to speculate on what extrinsic factors might disrupt the complex homeostatic mechanisms that are reflected during disease remission, it is reasonable to hope that a longitudinal analysis of a large cohort of children will be highly informative.

It is important to point out that many of the pathologic networks visualized in these studies demonstrate the structure of scale-free systems (26), as we have previously seen in neutrophils from patients with JIA (27). That is, the network structures demonstrate areas of high connectivity between some genes (designated “hubs” in systems biology) and other genes showing only limited connectivity to the system (“nodes”). Furthermore, the meta-structure of the collected profiles, especially in neutrophils, demonstrated modularity, which is another feature of cellular–physiologic systems (28).

These findings have interesting implications both for our understanding of pathogenesis and for elucidating new targets of therapy. From the standpoint of pathogenesis, we note that the pathologic structures revealed on Ingenuity are organized and are therefore as likely to represent physiologic adaptation to an externally applied force as they are an unraveling of basic biologic processes (e.g., the distinction between self and non-self). From the standpoint of therapy, it is useful to mention one of the primary characteristics of scale-free systems: they are highly resistant to perturbation at their peripheral nodes but vulnerable to attack at their hubs (29) (think of what happens to air traffic when inclement weather disrupts flights into and out of Atlanta or Chicago). This means that successful new treatments for JIA will have to focus on pathophysiologic structures, not specific genes. A gene that is expressed “20-fold above controls” is not necessarily a promising target if it represents a peripheral node. A gene that shows no differential expression at all might be a promising target if it represents a system hub.

We are still a long way from the ultimate goal of developing gene expression–based disease biomarkers that will direct therapy in polyarticular JIA. What this study has done is confirm that clinical remission (both during treatment with medication and without medical treatment) is a biologically distinct state. Furthermore, we have demonstrated that clinical remission in the absence of treatment with medication is not a normal state but represents a homeostatic condition in which proinflammatory and antiinflammatory mechanisms appear to be in balance. Answering the critical questions about biomarkers will require the study of large groups of children prospectively, a task that we have already undertaken.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES

Dr. Jarvis 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. Knowlton, Frank, Jarvis.

Acquisition of data. Jiang, Frank, Aggarwal, Wallace, McKee, Chaser, Tung, Smith, Chen, Osban, O'Neil, Centola, McGhee, Jarvis.

Analysis and interpretation of data. Knowlton, Jiang, Frank, O'Neil, Centola, Jarvis.

Manuscript preparation. Knowlton, Jiang, Frank, Aggarwal, Wallace, O'Neil, Jarvis.

Statistical analysis. Knowlton.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. REFERENCES
  • 1
    Foell D, Frosch M, Schulze zur Wiesch A, Vogl T, Sorg C, Roth J. Methotrexate treatment in juvenile idiopathic arthritis: when is the right time to stop? Ann Rheum Dis 2004; 63: 2068.
  • 2
    Gottlieb BS, Keenan GF, Lu T, Ilowite NT. Discontinuation of methotrexate treatment in juvenile rheumatoid arthritis. Pediatrics 1997; 100: 9947.
  • 3
    Wallace CA, Huang B, Bandeira M, Ravelli A, Giannini EH. Patterns of clinical remission in select categories of juvenile idiopathic arthritis. Arthritis Rheum 2005; 52: 355462.
  • 4
    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: 3902.
  • 5
    Wallace CA, Ruperto N, Giannini E, for the Childhood Arthritis and Rheumatology Research Alliance (CARRA), the Pediatric Rheumatology International Trials Organization (PRINTO), and the Pediatric Rheumatology Collaborative Study Group (PRCSG). Preliminary criteria for clinical remission for select categories of juvenile idiopathic arthritis. J Rheumatol 2004; 31: 22904.
  • 6
    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: 24964.
  • 7
    Simon RM, Korn EL, McShane LM, Radmacher MD, Wright GW, Zhao Y. Design and analysis of DNA microarray investigations. New York: Springer-Verlag; 2003.
  • 8
    Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 1995; 57: 289300.
  • 9
    Ward JH. Hierarchical grouping to optimize an objective function. J Am Stat Assoc 1963; 58: 23644.
  • 10
    Allwein EL, Schapire RE, Singer Y, Kaelbling P. Reducing multiclass to binary: a unifying approach for margin classifiers. J Mach Learn Res 2001; 1: 11341.
  • 11
    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: R1532.
  • 12
    Kutukculer N, Caglayan S, Aydogdu F. Study of pro-inflammatory (TNF-α, IL-1α, IL-6) and T-cell-derived (IL-2, IL-4) cytokines in plasma and synovial fluid of patients with juvenile chronic arthritis: correlations with clinical and laboratory parameters. Clin Rheumatol 1998; 174: 28892.
  • 13
    Rae MT, Niven D, Ross A, Forster T, Lathe R, Critchley HO, et al. Steroid signalling in human ovarian surface epithelial cells: the response to interleukin-1α determined by microarray analysis. J Endocrinol 2004; 183: 1928.
  • 14
    Hotamisligil GS. Inflammatory pathways and insulin action. Int J Obes Relat Metab Disord 2003; 27 Suppl 3: S535.
  • 15
    Dong C, Davis RJ, Flavell RA. MAP kinases in the immune response. Annu Rev Immunol 2002; 20: 5572.
  • 16
    Sigal LH. Basic science for the clinician 43: the mitogen-activated protein kinase family in inflammatory signaling. J Clin Rheumatol 2007; 13: 969.
  • 17
    Parks WC, Wilson CL, Lopez-Boado YS. Matrix metalloproteinases as modulators of inflammation and innate immunity. Nat Rev Immunol 2004; 4: 61729.
  • 18
    Wahl SM, Chen W. Transforming growth factor-β-induced regulatory T cells referee inflammatory and autoimmune diseases. Arthritis Res Ther 2005; 7: 628.
  • 19
    Kriegel MA, Li MO, Sanjabi S, Wan YY, Flavell RA. Transforming growth factor-β: recent advances on its role in immune tolerance. Curr Rheumatol Rep 2006; 8: 13844.
  • 20
    De Jager W, Hoppenreijs EP, Wulffraat NM, Wedderburn LR, Kuis W, Prakken BJ. Blood and synovial fluid cytokine signatures in patients with juvenile idiopathic arthritis: a cross-sectional study [published erratum appears in Ann Rheum Dis 2008;67:280]. Ann Rheum Dis 2007; 66: 58998.
  • 21
    Wedderburn LR, Robinson N, Patel A, Varsani H, Woo P. Selective recruitment of polarized T cells expressing CCR5 and CXCR3 to the inflamed joints of children with juvenile idiopathic arthritis. Arthritis Rheum 2000; 43: 76574.
  • 22
    Nilsson BO. Modulation of the inflammatory response by estrogens with focus on the endothelium and its interactions with leukocytes. Inflamm Res 2007; 56: 26972.
  • 23
    Massaad C, Paradon M, Jacques C, Salvat C, Bereziat G, Berenbaum F, et al. Induction of secreted type IIA phospholipase A2 gene transcription by interleukin-1β: role of C/EBP factors. J Biol Chem 2000; 275: 2268694.
  • 24
    Miyazawa K, Mori A, Yamamoto K, Okudaira H. Transcriptional roles of CCAAT/enhancer binding protein-β, nuclear factor-κB, and C-promoter binding factor 1 in interleukin (IL)-1β-induced IL-6 synthesis by human rheumatoid fibroblast-like synoviocytes. J Biol Chem 1998; 273: 76207.
  • 25
    Nishioka K, Ohshima S, Umeshita-Sasai M, Yamaguchi N, Mima T, Nomura S, et al. Enhanced expression and DNA binding activity of two CCAAT/enhancer-binding protein isoforms, C/EBPβ and C/EBPδ, in rheumatoid synovium. Arthritis Rheum 2000; 43: 15916.
  • 26
    Galien R, Evans HF, Garcia T. Involvement of CCAAT/enhancer-binding protein and nuclear factor-κB binding sites in interleukin-6 promoter inhibition by estrogens. Mol Endocrinol 1996; 10: 71322.
  • 27
    Galien R, Garcia T. Estrogen receptor impairs interleukin-6 expression by preventing protein binding on the NF-κB site. Nucleic Acids Res 1997; 25: 24249.
  • 28
    Jeong H, Tombor B, Oltval ZN, Barabasi AL. The large-scale organization of metabolic networks. Nature 2000; 407: 6514.
  • 29
    Jarvis JN, Petty HR, Tang Y, Frank MB, Tessier PA, Dozmorov I, et al. Evidence for chronic, peripheral activation of neutrophils in polyarticular juvenile rheumatoid arthritis. Arthritis Res Ther 2006; 8: R154.
  • 30
    Dong J, Horvath S. Understanding network concepts in modules. BMC Syst Biol 2007; 1: 24.
  • 31
    Albert R, Jeong H, Barabasi AL. Error and attack tolerance of complex networks. Nature 2000; 406: 37882.