Identification of Distinct Inflammatory Programs and Biomarkers in Systemic Juvenile Idiopathic Arthritis and Related Lung Disease by Serum Proteome Analysis

Abstract

Objective: Recent observations in systemic juvenile idiopathic arthritis (JIA) suggest an increasing incidence of high-mortality interstitial lung disease often characterized by a variant of pulmonary alveolar proteinosis (PAP). Co-occurrence of macrophage activation syndrome (MAS) and PAP in systemic JIA suggests a shared pathology, but patients with lung disease associated with systemic JIA (designated SJIA-LD) also commonly experience features of drug reaction such as atypical rashes and eosinophilia. This study was undertaken to investigate immunopathology and identify biomarkers in systemic JIA, MAS, and SJIA-LD.

Methods: We used SOMAscan to measure ~1,300 analytes in sera from healthy controls and patients with systemic JIA, MAS, SJIA-LD, or other related diseases. We verified selected findings by enzyme-linked immunosorbent assay and lung immunostaining. Because the proteome of a sample may reflect multiple states (systemic JIA, MAS, or SJIA-LD), we used regression modeling to identify subsets of altered proteins associated with each state. We tested key findings in a validation cohort.

Results: Proteome alterations in active systemic JIA and MAS overlapped substantially, including known systemic JIA biomarkers such as serum amyloid A and S100A9, and novel elevations in the levels of heat-shock proteins and glycolytic enzymes. Interleukin-18 levels were elevated in all systemic JIA groups, particularly MAS and SJIA-LD. We also identified an MAS-independent SJIA-LD signature notable for elevated levels of intercellular adhesion molecule 5 (ICAM-5), matrix metalloproteinase 7 (MMP-7), and allergic/eosinophilic chemokines, which have been previously associated with lung damage. Immunohistochemistry localized ICAM-5 and MMP-7 in the lungs of patients with SJIA-LD. The ability of ICAM-5 to distinguish SJIA-LD from systemic JIA/MAS was independently validated.

Conclusion: Serum proteins support a systemic JIA-to-MAS continuum; help distinguish systemic JIA, systemic JIA/MAS, and SJIA-LD; and suggest etiologic hypotheses. Select biomarkers, such as ICAM-5, could aid in early detection and management of SJIA-LD.

Figure 1:. Serum proteome profile associated with active sJIA/MAS/sJIA-LD

A, B, C) Volcano plots highlight the proteins (shown by gene names) with significantly changed abundance in the sJIA (A) or MAS (B) or sJIA-LD disease component (C). Significance thresholds are represented by the dashed lines: false discovery rate [FDR (adjusted p-value)] < 20% and fold change > 1.5. D) The MAS serum activity score (calculated without CRP or ferritin, see methods) across different groups is shown. sJIA-LD is sub-grouped by ferritin and CRP values (see Table 1). Dashed horizontal line indicates median value of healthy controls. Between-group comparisons were performed using Wilcoxon signed-rank test without the assumption of normal distribution and reflect comparisons to the control group unless otherwise indicated. We used the Benjamini-Hochberg procedure to adjust p-values. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p< 0.0001. E) Correlation between the coefficients (Log₂FC) assigned by the LIMMA model to the sJIA and MAS disease components (see Supplemental Methods) for the 174 proteins identified as significantly altered in at least one of the three disease components (sJIA, MAS, or sJIA-LD). F and G are similar to D and E, but they plot the sJIA-LD disease component.

Figure 2:. Protein functional groups for different disease components.

A) Examples of significantly elevated heat-shock proteins (HSP1A, HSP90AA1) and proteins involved in glycolytic process (GAPDH, ENO1) were plotted for different patient groups. Comparisons between each indicated group and the healthy control group were performed using Wilcoxon signed-rank test with p-values adjusted by Benjamini-Hochberg Procedure. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p< 0.0001. Dashed horizontal line indicates median value of healthy controls. C-D) Proteins present in SOMAscan for two different functional groups defined by Gene Ontology (GO) terms (C: Leukocyte-mediated immunity, D: Monocarboxylic acid metabolic process. GO term enrichment results are provided in Supplementary Table 5). The heatmap shows the coefficients of each protein assigned by the LIMMA model to each disease component, with disease components and genes clustered by Euclidean distances. To facilitate analysis, gene names are presented but represent protein targets, see Supplemental Methods.

Figure 3:. Cytokine/chemokine serum abundance differs between sJIA-LD and sJIA/MAS

A) The heatmap presents the coefficients of concentration change from the LIMMA analysis for cytokines/chemokines associated with each disease component in the linear regression model. Cytokines/chemokines shown are those that reached significance in at least one disease component and were associated with the respective GO term (see Supplemental Methods). B) Eight sigificantly altered cytokines/chemokines in the sJIA-LD disease component were plotted by different patient groups (see Supplementary Figure 9 for groups beyond the main disease groups). C) CCL2, which approaches significance for sJIA-LD disease component is also shown. D) CXCL9 and CXCL10, two interferon-inducible chemokines are shown. Dashed horizontal lines indicate median values of healthy controls. The False Discovery Rate (FDR) for the comparison between sJIA-LD and active sJIA or MAS patients, controlling for MAS activity scores, is provided (see methods and Supplementary Table 4). To facilitate analysis, gene names are presented but represent protein targets, see Supplemental Methods.

Figure 4:. Pulmonary localization of ICAM5 and MMP7 and their performance in the discovery cohort.

A) The 30 organs with highest-expression of ICAM5 mRNA from an autopsy cohort (Genotype-Tissue Expression, GTEx); expression in lung samples is depicted in red. B) The correlation of MMP7 and ICAM5 protein levels among all sJIA serum samples. C) ROC curves using ICAM5 to classify different comparisons in the discovery cohort. D) Staining of ICAM5 and MMP7 in control and sJIA lung. Inset shows characteristic PAP/ELP histology in an sJIA-LD patient. ICAM5 protein expression in lung interstitial fibroblast cells in contrast with epithelial cells (marked by cytokeratin, CK7) and macrophages (CD163). MMP7 is expressed by the indicated epithelial cells (arrows) and macrophages and hematopoietic cells (arrowheads) in control and sJIA-LD lung. * denotes alveolar lumen. Nuclei counterstained with DAPI (blue). All immunofluorescence images are from sJIA-LD. See also Supplementary Figures 12 and 13.

Figure 5:. Validation of ICAM5 as an MAS-independent marker of lung disease in sJIA

Serum (circles) and plasma (triangles) samples from an independent cohort were assayed for IL-18 (A) and CXCL9 (B) by Luminex and for MMP7 (C) and ICAM5 (D) by ELISA. “Other LD” indicates sJIA patients with intercurrent non-PAP lung disease (lobar pneumonia or pulmonary hypertension), whereas “sJIA-LD resolved” indicates a patient with a distant history of radiographic abnormalities resolved at time of sampling. Further details in Supplemental Tables 1 & 7. Between-group comparisons were performed using Wilcoxon signed-rank test without the assumption of normal distribution and reflect comparisons to the sJIA-LD group unless otherwise indicated. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p< 0.0001. (E) Receiver Operating Characteristic (ROC) analyses of ICAM5 distinguishing sJIA-LD from other forms of sJIA. # the combined included inactive/active sJIA and MAS. In each patient group, data from a given patient only appeared only once. Exception is made for the “other-LD” group, where samples from the same patient in different disease stages were used to represent variability (such as pneumonia with inactive sJIA vs. MAS, noting the large variation in sJIA/MAS markers (Figure 5A-B), but small variations among lung disease markers MMP7(C) and ICAM5 (D)).