We studied 25 children newly diagnosed with rheumatoid factor-negative, polyarticular JRA. Diagnosis was based on accepted and validated criteria endorsed by the American College of Rheumatology (ACR) 10. Children were excluded if they had been treated with corticosteroids or methotrexate, or if they had received therapeutic doses of nonsteroidal anti-inflammatory drugs for more than 3 weeks prior to study. Patients with active disease ranged in age from 4 to 15 years and presented with proliferative synovitis of multiple joints. All had joint activity scores of at least 15 using a standard scoring system 11 based on that used in pediatric rheumatology clinical trials 12. Children followed longitudinally were designated as having a 'partial response' to therapy if they met American College of Rheumatology-30 improvement criteria from their baseline state. Children were designated to have inactive disease if there was no objective synovitis on exam, morning stiffness for not more than 20 minutes/day, and a normal erythrocyte sedimentation rate. In addition, we studied 14 of these children on more than one occasion to observe changes in gene expression pattern in response to therapy. S100A8/A9 protein levels, a marker of neutrophil-endothelial cell interactions (see below), were studied in 24 children, 20 of whom were studied on more than one occasion to observe responses to therapy.

Healthy control subjects (n = 10) were young adults (age 18 to 30) with no history of rheumatic or chronic inflammatory disease. Previously published work from our group 13 has demonstrated that such subjects are appropriate controls for gene expression studies in children with polyarticular JRA because gene expression profiles of peripheral blood buffy coats of children with polyarticular JRA revert toward patterns indistinguishable from such healthy controls after treatment.

After the execution of the informed consent process as approved by the Oklahoma University Health Sciences Center (OUHSC) Institutional Review Board, whole blood (20 cc) was drawn into sterile sodium citrate tubes containing a cell density gradient (cat no. 362761; BD Biosciences, San Jose, CA, USA) and carried immediately to the Pediatric Rheumatology Research laboratories on the OUHSC campus. Granulocytes were immediately separated from mononuclear cells by density gradient centrifugation. Centrifugation was performed at room temperature, resulting in the red cells and granulocytes' layering in the bottom of the tube. Red cells were removed from the granulocytes by hypotonic cell lysis as recommended by the manufacturer, and granulocytes were placed immediately in Trizol reagent for RNA purification. Plasma was removed and stored at -80°C until used in enzyme-linked immunosorbent assays (ELISAs) for S100 protein levels (see below). Cells prepared in this fashion are more than 98% CD66b⁺ by flow cytometry and contain no contaminating CD14⁺ cells. Granulocytes were immediately placed in Trizol reagent (Invitrogen, Carlsbad, CA, USA), and RNA was purified exactly as recommended by the manufacturer. RNA was stored under ethanol at -80°C until used for hybridisation and labeling.

The arrays used in these experiments were developed at the Oklahoma Medical Research Foundation Microarray Core Facility in collaboration with QIAGEN Operon (Alameda, CA, USA). Microarrays were produced using commercially available libraries of 70-nucleotide-long DNA molecules whose length and sequence specificity were optimised to reduce the cross-hybridisation problems encountered with cDNA-based microarrays. The microarrays had 21,329 human genes represented. The oligonucleotides were derived from the UniGene and RefSeq databases. For the genes present in this database, information on gene function, chromosomal location, and reference naming are available. All 11,000 human genes of known or suspected function were represented on these arrays. In addition, most undefined open reading frames were represented (approximately 10,000 additional genes).

Oligonucleotides were spotted onto Corning^® UltraGAPS™ amino-silane-coated slides (Acton, MA, USA), rehydrated with water vapor, snap-dried at 90°C, and then covalently fixed to the surface of the glass using 300-mJ, 254-nm wavelength UV radiation. Unbound free amines on the glass surface were blocked for 15 minutes with moderate agitation in a 143 mM solution of succinic anhydride dissolved in 1-methyl-2-pyrolidinone, 20 mM sodium borate, pH 8.0. Slides were rinsed for 2 minutes in distilled water, immersed for 1 minute in 95% ethanol, and dried with a stream of nitrogen gas.

Prior to cDNA synthesis, the RNA was resuspended in diethylpyrocarbonate-treated water. RNA integrity was assessed using capillary gel electrophoresis (Agilent 2100 BioAnalyzer; Agilent Technologies, Inc., Palo Alto, CA, USA) to determine the ratio of 28 s/18 s rRNA in each sample. A threshold of 1.0 was used to define samples of sufficient quality, and only samples above this limit were used for microarray studies. cDNA was synthesised using Omniscript reverse transcriptase (Qiagen, Valencia, CA, USA) with direct incorporation of cyanine 3-dUTP (deoxy-uridine triphosphate) from 2 μg of RNA. Labeled cDNA was purified using a Montage 96-well vacuum system (Millipore Corporation, Billerica, MA). The cDNA was added to hybridisation buffer containing CoT-1 DNA (0.5 mg/ml final concentration), yeast tRNA (0.2 mg/ml), and poly(dA)_40–60 (0.4 mg/ml). Hybridisation was performed in an automated liquid delivery, air-vortexed, hybridisation station for 9 hours at 58°C under an oil-based coverslip (Ventana Medical Systems, Inc., Tucson, AZ, USA). Microarrays were washed at a final stringency of 0.1 × SSC (saline-sodium citrate). Microarrays were scanned using a simultaneous dual-colour, 48-slide scanner (Agilent Technologies, Inc.). Fluorescent intensity was quantified using Koadarray™ software (Koada Technology, Kippen, Sterling, UK).

Data were subject to normalisation and regression steps as described in detail in our earlier work 13. Genes differentially expressed between groups of samples were selected using associative analysis 13. Genes selected to be differentially expressed in any sample combinations were used to classify patients, including active, partial and inactive, and control samples using hierarchical clustering. The analysis package is provided by Spotfire DecisionSite for Functional Genomics 8.1 (Spotfire, Inc., Somerville, MA, USA). Similarity measure was the Euclidean distance, the clustering method was Unweighted Pair Group Method with Arithmetic Mean, and input rank was the ordering function.

Forty-two of the most highly expressed up- or downregulated genes in patients with JRA were used in pathway modeling using PathwayAssist Software (Ariadne Genomics Inc., Rockville, MD, USA). Relationships of protein nodes with H₂O₂ and calcium were preserved intentionally to reveal the overall networking of calcium influx and peroxide metabolism, which are highly specific to the function of neutrophils.

Hypervariable (HV) genes are a group of genes whose expressions exhibit higher variation than biological fluctuation baseline, as we have described previously 14. After the HV genes were selected, they were clustered using an F-means clustering method to determine each gene's cluster association and its connectivity with other genes. Genes were sorted based on their cluster association and connectivity in the control group, with the gene of the highest connectivity of the first cluster ranked on the top. To reveal the intrinsic dynamic relationship between each gene in a sample group, a matrix of correlation coefficiency was displayed in a colour mosaic.

Six down randomly selected genes in the patients with polyarticular JRA and controls were selected for reverse transcription-polymerase chain reaction (PCR) confirmation.

Costar High Binding 96-well plates (Corning Life Sciences, Acton, MA, USA) were coated with 100 μl/well of S100A8/A9-specific monoclonal antibody 5.5 (kindly provided by Dr. Nancy Hogg, Cancer Research UK, London, UK) diluted to a concentration of 1 μg/ml in 0.1 M carbonate buffer (pH 9.6) and left overnight at 4°C. After incubation, the plates were washed with phosphate-buffered saline (PBS)/0.1% Tween-20 and blocked with PBS/0.1% Tween-20/2% bovine serum albumin (BSA) (100 μl/well) for 30 minutes at room temperature. The samples (plasma from children with polyarticular JRA and healthy controls) and standards (100 μl) were added and incubated for 40 minutes at room temperature. After three washes with PBS/0.1% Tween-20, the plates were incubated with 100 μl/well of S100A9 polyclonal antibodydiluted 1:10,000 in PBS/0.1% Tween-20/2% BSA for 40 minutes at room temperature. After incubation, the plates were washed three times and incubated with 100 μl/well of peroxidase-conjugated donkey anti-rabbit immunoglobulin G (IgG) at a dilution of 1:7,500 in PBS/0.1% Tween-20/2% BSA for 40 minutes at room temperature. After three washes, the presence of IgG was detected with 100 μl of a peroxidase substrate solution (3,3',5,5'-tetramethylbenzidine; RDI Division of Fitzgerald Industries Intl, Concord, MA, USA, formerly Research Diagnostics Inc.) according to the manufacturer's instructions; the reaction was stopped by adding 100 μl of 0.36 mM H₂SO₄, and the optical density was read at 500 nm. Results from patient samples were compared against standards of known S100A8/A9 concentration. The detection limit for this assay is 1 ng/ml A8/A9 dimer. The antibodies used in this assay have been tested against murine S100A8 and S100A9, bovine S100A and S100B, and human S100A12 and found to be specific.

Results were tabulated in a commercially available statistics and graphics software program (GraphPad Prism; GraphPad Software, Inc., San Diego, CA, USA), and comparisons of children with active and inactive polyarticular JRA and controls were accomplished using a two-tailed independent t test. Results ≤ 0.05 were considered statistically significant.

Neutrophils were placed on glass coverslips, incubated with 1 μg fluorescein isothiocyanate (FITC)-conjugated anti-myeloperoxidase (MPO) at 4°C for 30 minutes, and then washed again with Hanks' balanced salt solution at room temperature. Cells were observed using an axiovert fluorescence microscope (Carl Zeiss, Inc., Thornwood, NY, USA) with mercury illumination interfaced to a computer using Scion image processing software (Scion Corporation, Frederick, MD, USA). A narrow band-pass discriminating filter set (Omega Optical, Inc., Brattleboro, VT, USA) was used with excitation at 485/22 nm and emission at 530/30 nm for FITC. A long-pass dichroic mirror of 510 nm was used. The fluorescence images were collected with an intensified charge-coupled device camera (Princeton Instruments Inc., Trenton, NJ, USA).

NAD(P)H autofluorescence oscillations were detected as described 1516. An iris diaphragm was adjusted to exclude light from neighboring cells. A cooled photomultiplier tube held in a model D104 detection system (Photon Technology International, Inc., Birmingham, NJ, USA) attached to a microscope (Carl Zeiss, Inc.) was used.

A total of 712 genes were shown to have differential levels of expression between the patients with polyarticular JRA and the control subjects. For simplicity, the 84 genes showing the highest levels of differential expression expression are shown in Table 1 (genes over-expressed in polyarticular JRA neutrophils) and Table 2 (genes under-expressed in polyarticular JRA neutrophils). The full data sets are available online 17. Genes over-expressed in patients with polyarticular JRA included principally mediators and regulators of oxidative response, neutrophil activation, and inflammation control (Table 1) (Figure 1), suggesting that peripheral neutrophils are active in patients with polyarticular JRA and contribute to the systemic inflammatory nature of this disorder. These results provide a catalog of neutrophil-mediated aspects of disease pathology, with both well-characterised and putative pathogenic pathways identified, suggesting that inhibition of neutrophil activity may provide a useful means of limiting key aspects of the pathology of polyarticular JRA. Genes downregulated in JRA neutrophils relative to healthy controls (Table 2) included the immune and inflammatory mediators CCL3, CCL4, CCL5, IL-1B, COX-2, MHC-II DR-α, granzyme A, galectin 1, V-Fos, and inhibitor of nuclear factor-κB-α.

Validity of the array data was then tested using quantitative real-time PCR on the six randomly selected genes (Figure 2). In each case, real-time PCR data corroborated the array findings, as shown in Table 3.

To determine the functional relationship among these genes, computer modeling based on the differentially expressed genes was used. These studies indicated links to both innate and adaptive immunity (Figure 1), with clusters of both interleukin (IL)-8- and interferon-γ (IFN-γ)-regulated genes differentially expressed in children with polyarticular JRA and control subjects. Furthermore, multiple genes in the computer model were linked to both calcium influx (Figure 1, top left) and superoxide ion production (green circles, 'H₂O₂'). These findings were of considerable interest given that IL-8 and IFN-γ independently regulate oscillations of key metabolites in neutrophils, which in turn regulate both calcium ion influx and superoxide ion release 18. This model was tested directly using single-cell autofluorescence, as described below.

Hierarchical clustering of genes that were differentially expressed in patients with polyarticular JRA was used to group individuals who have similar expression profiles in their peripheral blood neutrophils. Patients with polyarticular JRA and controls formed distinct clusters, confirming the validity of the differential gene expression analysis on a global scale.

Figure 3 shows a hierarchal cluster analysis of neutrophil mRNA expression in children with polyarticular JRA and a panel of eight healthy control subjects. Children were grouped according to disease activity as described in Materials and methods. Of note is that healthy control subjects cluster together at the left side of the graph. Children with polyarticular JRA, however, scatter across the graph regardless of disease activity. That is, children with polyarticular JRA showed persistent abnormalities in neutrophil gene expression when their disease was well controlled. This finding was similarly demonstrated using the connectivity analysis procedure (Figure 4) described in Materials and methods and in our previously published work 1314. The contingency analysis for these selected genes demonstrated disruption of normal gene relationships in neutrophils of children with polyarticular JRA when those relationships were compared with healthy controls. These findings strongly suggest that neutrophils are chronically dysregulated in polyarticular JRA and that therapy only minimally ameliorates the disordered pattern.

To further support a role for chronic neutrophil activation in polyarticular JRA, we examined S100A8/A9 and S100A12 plasma levels. Both S100A8/A9 and S100A12 (data not shown) were identified as over-expressed in patients with polyarticular JRA (relative to controls; Figure 1) in array experiments and confirmed on real-time PCR analysis. These proteins are highly expressed in neutrophils and monocytes (up to 40% of cytosolic proteins), are released upon cell activation, and contribute to the migration of neutrophils to inflammatory sites 1920. As predicted from the array data, S100 proteins were markedly elevated in children with polyarticular JRA (662 ± 40 ng/ml) compared with controls (40 ± 9 ng/ml; p > 0.001; Figure 5a). Children with inactive disease (198 ± 60 ng/ml) had lower levels of S100 proteins compared with children with active disease (p = 0.007; Figure 5b), but levels were still significantly higher (p = 0.047) than those seen in healthy controls (Figure 5c). These findings suggest that neutrophils in children with polyarticular JRA remain in an activated state during disease quiescence.

The computer model generated through analysis of differentially expressed genes (Figure 1) suggested pathologically relevant links between IL-8- and IFN-γ-regulated genes in polyarticular JRA neutrophils and that gene expression was functionally linked to calcium influx and superoxide ion production. IL-8 and IFN-γ independently regulate oscillatory phenomena in neutrophils, with IFN-γ regulating amplitude and IL-8 oscillatory frequency. We proceeded to test that model by monitoring the autofluoresence of NAD(P)H in living neutrophils, which reflects various stages and mechanisms of neutrophil activation 21. Metabolic oscillations of neutrophils from six children with polyarticular JRA and five healthy control subjects were monitored. Figure 6 provides representative tracings of NAD(P)H oscillations in resting and stimulated cells from patients. Because metabolic frequencies and amplitudes have been linked with the hexose monophosphate shunt activity and the peroxidase cycle, respectively, we assessed MPO surface expression on living neutrophils. In contrast to controls that show no MPO surface expression, all patients with polyarticular JRA demonstrated a subpopulation of neutrophils (10% to 23% of the cells) that expressed surface-associated myeloperoxidase. Neutrophils staining MPO-negative from patients responded to lipopolysaccharide (LPS) stimulation by increasing the frequency of NAD(P)H oscillations, reducing the period from 20 to 10 seconds, as previously described in activated neutrophils 18. This behaviour is identical to that observed for control neutrophils. However, MPO-positive neutrophils from patients with polyarticular JRA (including two with inactive disease) demonstrated increases in both frequency and amplitude in NAD(P)H oscillation after LPS stimulation. In contrast, activated control cells show no changes in metabolic amplitude. This novel finding suggests a fundamental breakdown in the regulation of neutrophil metabolism, as will be discussed below. We are now preparing to determine whether the number of aberrantly functioning, MPO-positive cells changes with disease severity or during the course of therapy.