Background We have recently demonstrated the feasibility of developing prognostic biomarkers from whole blood gene expression profiles in children with newly diagnosed juvenile idiopathic arthritis (JIA). However, determining disease status at one year was more challenging.
Objectives To determine whether prognostic biomarkers for disease status at 12 months can be developed from whole blood gene expression profiles in children with JIA after the initiation of therapy.
Methods We studied whole blood expression profiles from children enrolled in the TREAT study, an NIH-funded clinical trial comparing methotrexate (MTX) with MTX + etanercept (Et) in children with newly-diagnosed JIA. We extracted RNA from whole blood (PAXgene) samples using conventional methods and removed globin genes using Ambion GlobinClear reagents. Labeling, hybridization reactions, and scanning were performed using conventional methods for Illumina WG-6_V3 whole genome microarrays. Gene expression profiles were then examined to determine those genes whose expression levels best predicted outcome (active vs inactive disease) at 12 months.
Results Predictive models could be developed at 6 months from baseline data, as we have previously reported. Baseline expression profiles could not predict status at 12 months. Using 4 month data (the earliest point at which samples were collected from children on treatment), we were able to construct receiver operator curves (ROC) with strong predictive properties for disease status at 12 months (AUC =0.89 for RF-negative patients). Thus, longer term outcome is predictable only after children have initiated therapy. When examining the transcriptional changes that occur between baseline and 4 months, we found numerous genes showing significant differential expression (absolute fold change >1.2, p<0.05) in the RF negative (n=1,376) and RF positive (n=680) groups. Functional analysis of genes that showed differential expression between 0 and 4 months showed interesting differences between the RF negative and RF positive patients. In RF negative patients,DAVID analysis showed enrichment for immune response genes (p=2.89 E-.05) and genes associated with the respiratory burst (6.62 E-.05). It was also interesting to note that translational elongation (p=0.0013) and RNA splicing (p=0.0029) among the gene ontology (GO) terms enriched in the differentially expressed genes. For RF positive patients, the top 3 GO terms included macromolecular complex assembly (p=7.78E-.04), macromolecular complex subunit organization (p=0.002) and intracellular transport (p=0.004).
Conclusions These studies show the feasibility of developing whole blood prognostic biomarkers in children with JIA. Longer term prognoses can only be made after therapy has been initiated. A curious additional finding in this study was the appearance of different mechanisms of response in RF positive and RF negative patients after 4 months of therapy, a finding that may explain the relative refractoriness of RF positive patients to otherwise effective therapies.
Acknowledgements This work was supported by R01-AR-060604, R01-AI-084200, and R01-AR-049762 from the National Insitutes of Health.
Disclosure of Interest None declared