Background: Synovial tissue (ST) inflammation in Rheumatoid Arthritis (RA) shows high degree of heterogeneity which may underly variable response to treatments. Novel insights from high-throughput ST analyses showed their potential in dissecting disease heterogeneity and in identifying putative novel biomarkers of prognosis.

Objectives: To assess the power of multi-modal analysis of ST inflammation in naive to treatment RA to identify predictive biomarkers of treatment response.

Methods: 373 naive to treatment RA were enrolled and underwent Ultrasound guided ST biopsy. For each patient, synovitis degree was determined using a H&E-based semiquantitative score (Krenn Synovitis Score - KSS)[1] and synovial pathotype was determined using immunohistochemistry. Among those, n=45 ST samples were used for synovial tissue macrophage (STMs)(CD206/MerTK) FACS phenotyping and Digital Spatial Profiling (DSP)(GeoMx DSP, Nanostring) to quantitate abundance and transcriptomic profiles of CD68 ^pos cells in 103 spatially distinct ST regions of interest (ROI). After study entry, each RA patient was treated according to the treat to target strategy and was followed every 3 months to assess DAS remission achievement at 6 months follow-up.

Results: Naive RA who achieved DAS-remission at 6 months follow-up with the treat to target strategy had, at baseline, lower KSS compared to patients not achieving this outcome (p=0.0006). Moreover, naive to treatment RA with lympho-myeloid (LM) as well as diffuse-myeloid (DM) pathotype had lower response rate to cDMARDs (36.1% and 44.7%) compared to RA with pauci-immune (PI) pathotype (59.1%, p=0.001 and p=0.042 respectively). However, SHAP methods showed that baseline KSS has limited individualized prediction capacity between responder and non-responder, highlighting the need for multi-modal tissue deconvolution. At FACS analysis, patients with DM and LM pathotypes showed comparable enrichment of MerTK ^pos CD206 ^pos and MerTK ^neg CD206 ^neg STMs populations, while patients with PI pathotype showed an expansion of MerTK ^pos CD206 ^pos STMs (p=0.0106). Moreover, naive RA who achieved DAS-remission at 6 months follow-up had higher MerTK ^pos CD206 ^pos STMs rates than patients not achieving this clinical outcome (p=0.0002). In particular, baseline MerTK ^pos STMs enrichment ≥44.3% [AUC:0.80 (95% IC:0.65-0.94), p=0.001] arose as an independent factor associated with DAS-remission achievement at 6 months follow-up [OR: 24.5 (95% IC:4.3-139.2), p<0.0001]. DSP analysis of naïve to treatment ST biopsies revealed differential gene networks characterising activation of STMs in distinct tissue locations (e.g. lining layer, sublining layer and intra-follicular structure distribution respectively). Moreover, DSP analysis enabled to identify transcriptomic signatures of lining and sublining STMs associated with treatment response to cDMARDs (e.g. MERTK , KLF4 , APP ) as well as of treatment refractoriness (e.g. SPP1, S100A12, TNF ). Integration of STM scRNAseq dataset[2] with STMs spatial transcriptomic data, successfully mapped all STM clusters in ST stratified based on treatment response, showing an enrichment of S100A12 ^pos and SPP1 ^pos STMs mainly located in sublining ROIs of naive to treatment RA not responding to cDMARDs.

Conclusion: Multi-modal analysis of synovitis enables to differentiate, at first medical evaluation, naïve-to-treatment RA who subsequently responded to first line cDMARDs treatment from those who failed, strongly supporting its predictive value as putative patient-based decision test tool.

REFERENCES: [1] Alivernini S, et al. Arthritis & Rheumatology 2021.

[2] Alivernini S, et al. Nat. Medicine 2020.

Acknowledgements: NIL.

Disclosure of Interests: None declared.