Background: Despite major advances in the treatment of rheumatoid arthritis (RA), approximately 20-40% of patients fail to achieve an adequate response to biologic or targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs). Consequently, the choice between targeted therapies such as Janus kinase inhibitors (JAKi) and tumor necrosis factor inhibitors (TNFi) remains largely driven by clinical judgment rather than molecular stratification. A deeper understanding of the molecular mechanisms modulated by JAKi and TNFi is therefore required to identify predictive biomarkers of therapeutic response. Such insights may enable precision medicine approaches, support early treatment selection, and maximize clinical benefit before irreversible joint damage occurs.

Objectives: (1) To characterize the ex vivo molecular and cellular effects of JAKi and TNFi on RA immune cells. (2) To identify, in vivo , RA endotypes defined by distinct baseline levels of drug-specific proteomic signatures and to assess their association with clinical response. (3) To perform longitudinal validation of ex vivo -derived protein signatures in serum samples from RA patients achieving in vivo. (4) To validate the functional relevance of drug-specific protein signatures using 3D synovial models linking changes in these signatures to synovial tissue alterations. (5) To evaluate the potential clinical utility of drug-specific molecular signatures for guiding first-line treatment selection in RA.

Methods: Peripheral blood mononuclear cells (PBMCs) and neutrophils from 48 b/tsDMARD-naïve RA patients were cultured with autologous serum and treated ex vivo with baricitinib (JAKi) or etanercept (TNFi) (10 µM) for 24 and 12 hours, respectively. Immune cell proliferation, adhesion and NETosis were quantified using standardized functional assays and expression of regulatory genes controlling these processes was measured by RT-PCR. Protein secretion in culture supernatants was profiled using proximity extension assay technology, assessing a 92-protein inflammation-related panel. To evaluate the clinical relevance of the identified proteomic signatures, baseline serum samples from 223 RA patients were analyzed, including 36 b/tsDMARD-naïve and 32 non-naïve patients initiating JAKi, and 125 b/tsDMARD-naïve and 30 non-naïve patients starting TNFi. Clinical response was evaluated longitudinally using DAS28-CRP at 3, 6 and 12 months after JAKi or TNFi initiation. For longitudinal validation, paired baseline and 6-month serum samples from an independent cohort of 46 b/tsDMARD-naïve RA patients achieving remission were analyzed and used to stimulate human synovial fibroblasts in 3D culture, assessing synovial lining thickness and proteomic changes in supernatants. Finally, baseline serum proteomic data from naïve patients were integrated into machine learning models to predict response to JAKi and non-response to TNFi, and vice versa, using DAS28-CRP remission (DAS28-CRP < 2.6) as the outcome.

Results: Ex vivo treatment with JAKi and TNFi significantly modulated key immune functions in PBMCs and neutrophils from b/tsDMARD-naïve RA patients. Both therapies reduced proliferation, while TNFi exerted a stronger inhibitory effect on cell adhesion and NETosis, in parallel with modulation of corresponding regulatory genes. Autologous serum markedly enhanced inflammatory protein secretion in both cell types, including cytokines, chemokines and growth factors. Ex vivo exposure to JAKi or TNFi generated distinct inflammatory protein signatures that correlated with drug-induced functional changes.

When evaluated in baseline serum simples from a new RA cohort, these drug-specific signatures stratified b/tsDMARD-naïve RA patients into molecular subgroups, in which higher protein signature expression was associated with higher baseline disease activity and a more favorable clinical response to the corresponding therapy. This stratification was not observed in non-naïve RA patients. Cross-therapy analyses showed that high baseline expression of one drug-specific signature was associated with non-response to the alternative treatment, indicating treatment specificity. Longitudinal analysis of serum samples from an independent cohort of 46 b/tsDMARD-naïve RA patients achieving remission revealed significant modulation between baseline and 6 months of several proteins included in the ex vivo –derived signatures, confirming their in vivo relevance. Functionally, 6-month serum from TNFi-treated patients reduced synovial lining thickness and induced broader proteomic changes in 3D synovial fibroblast cultures compared with baseline, whereas serum from JAKi-treated patients produced no histological changes and less extensive proteomic modulation. Based on the relevance of these signatures as potential predictors of clinical response, a machine learning-based neural network model incorporating seven key proteins (PD-L1, LAP TGF-β-1, EN-RAGE, IL-17C, IL-24, 4E-BP1, and CDCP1) was developed and discriminated treatment-specific remission outcomes, achieving an AUC of 0.75.

Conclusions: This study supports the clinical utility of ex vivo -derived proteomic signatures to guide personalized first-line therapy in RA. These drug-specific signatures stratify b/tsDMARD-naïve patients according to their likelihood of achieving remission, are validated longitudinally in vivo , and exert functional effects on the synovial microenvironment. Their integration into predictive models may help maximize remission rates while reducing exposure to ineffective therapy, supporting a precision medicine approach for therapeutic decision-making in RA.

REFERENCES: NIL.

Acknowledgments: NIL.

Disclosure of Interests: None declared.