Background/Purpose: Many patients living with rheumatoid arthritis (RA) can achieve remission with modern treat-to-target disease-modifying anti-rheumatic drugs (DMARDs), albeit with the risks associated with long-term immunosuppression. Some may choose to try reducing or stopping treatment, though this carries a risk of flare which is currently unpredictable. We explored whether high-throughput serial measurement of circulating proteomic biomarkers could predict risk of future arthritis flare after DMARD cessation.

Methods: We analyzed longitudinal expression of 250 soluble factors (NULISA technology) in serum collected from 115 patient with RA in remission at 0, 2, 5, 8, 12 and 24 weeks after stopping conventional synthetic DMARDs (BIO-FLARE study). Additionally, synovial tissue (ST) biopsies from 22 recent-onset untreated disease-active ACPA+ RA patients and 6 healthy individuals were imaged using spatial transcriptomic profiling (5,101 genes, 10x Genomics).

Results: Over 24 weeks, 57 BIOFLARE participants flared and 58 maintained drug-free remission. A linear mixed model identified 68 cytokines with significantly different expression trajectories between flare and remission groups. KEGG Pathways indicate enrichment in inflammatory pathways, including Cytokine-cytokine receptor interaction, Viral protein interaction with cytokine and cytokine receptor, and JAK-STAT signaling pathway, which are associated with RA flare pathology. Addition of a random forest algorithm achieved 88% accuracy in distinguishing flare events. The model identified a subset of cytokines that best predicted flare before it was clinically apparent. Of these, Xenium spatial imaging co-located CXCL13, TNFSF13C and CD83 RNA expression to follicles containing memory B cells, plasma cells and peripheral/follicular helper T cells (Tph/fh). In contrast, CXCL10 was expressed in small lining-adjacent foci by lining-layer fibroblasts and endothelial cells surrounded by CD8+GZMK+ T cells. Notably, these fibroblasts co-expressed type 1 interferon (IFN)-inducible genes, particularly GBP1, GBP5, IFIT3 and IFIT2 that are involved in anti-viral and anti-bacterial defence.

Conclusion: Biomarkers of flare after DMARD withdrawal, identified by integrating cytokine trajectory analysis, machine learning and spatial transcriptomics, reflect chemoattractants and soluble markers of activated memory B cells and Tph/fh to ST. Furthermore, spatial data suggest that pathogen-mediated type I IFN-induced chemoattraction of CD8+GZMK+ T cells contributes to RA disease flares. With validation, these circulating markers may improve monitoring of disease activity and provide opportunity for interception of flare during DMARD withdrawal.

« Back to ACR Convergence 2025

ACR Meeting Abstracts - https://acrabstracts.org/abstract/predicting-rheumatoid-arthritis-flare-using-longitudinal-cytokine-trajectories-machine-learning-and-spatial-transcriptomic-imaging/