Abstract
Background/Aims
Neutrophils contribute to disease pathology in inflammatory diseases including rheumatoid arthritis (RA). Activated RA neutrophils release ROS and proteases which damage joints, and produce neutrophil extracellular traps (NETs) that expose citrullinated nuclear proteins leading to the development of ACPA auto-antibodies. We previously described an altered gene expression signature in RA patients compared to healthy controls. This aim of this work was to use computational modelling of neutrophil transcriptomes to provide novel insight into the physiological factors controlling neutrophil phenotype in RA.
Methods
RNA from peripheral blood neutrophils (RA patients (DAS28>5.1, cohort 1 n = 23, cohort 2 n = 53), healthy controls (n = 11)) was sequenced using RNAseq. Reads were mapped to the human genome (hg38) using Tophat2 and read counts generated using featureCounts. Partial least squares discriminant analysis (PLS-DA) was carried out using mixOmics (with random sampling and ‘leave one out’ cross-validation). Gene expression network analysis was carried out using tmod, ARACNE2 and GALGO. Gene networks were visualised using Cytoscape. Functional annotation was carried out using Ingenuity Pathway Analysis (IPA) and DAVID.
Results
PLS-DA modelling discriminated RA and HC neutrophil transcriptomes with an F1 score of 98.2% +/- 1.6% over 5 repetitions using a model with 1 component. Blood transcriptional modular (tmod) enrichment analysis of gene expression in RA and HC neutrophils from cohort 1 identified the gene networks activated in RA and absent in HC as: myeloid cells activated via NFκB, innate antiviral response, type I interferon response, inflammasome receptors and cell signalling (FDR<0.05, AUC >0.8). Cell cycle and growth arrest was also considerably more enriched in RA compared to HC (RA AUC=0.82, HC AUC=0.57). Activation of type I interferon and inflammasome signalling pathways in cohort 1 was confirmed by IPA and correlated closely with data from a second cohort of RA patients (cohort 2). ARACNE2 identified five major gene modules activated in RA neutrophils (MI threshold 0.5, p < 10-20). IPA and DAVID predicted that Module 1 gene networks regulated amino acid, nucleic acid, carbohydrate and lipid metabolism as well as initiation of gene expression. Module M2 contained a network of genes regulated by integrins and cytokine receptors (e.g. IL-8, JAK/STAT, TNF/NFκB). Module 3 also contained genes activated by NFκB as well as by AMP-protein kinase. Module M4 genes were regulated by activation of type I interferon receptors and pattern recognition receptors (e.g. IRFs, STATs). Module M5 genes regulated amino acid metabolism. Multivariate modelling using GALGO identified genes that predict clinical characteristics, including genes involved in NFκB signalling, apoptosis and kinase activation which were associated with disease activity (DAS28), and cellular stress response genes and chromatin modification that was associated with raised inflammatory markers (ESR, CRP).
Conclusion
The results of the computational analyses are currently being validated experimentally.
Disclosure
M. Fresneda Alarcon: None. E. Caamano-Gutierrez: None. P. Antczak: None. R.J. Moots: None. H.L. Wright: Grants/research support; H.W. has received funding from Versus Arthritis and the Masonic Charitable Fund.
Type I interferon response gene expression in established rheumatoid arthritis is not associated with clinical parameters