Tumor necrosis factor inhibitors (TNFi) are first-line biologic agents for rheumatoid arthritis (RA), yet a substantial proportion of patients—approximately 30% to 40%—do not respond adequately. This clinical challenge is particularly pronounced among individuals who test positive for anti-citrullinated protein antibodies (ACPA). ACPA-positive RA patients often present with more aggressive disease, a higher risk of joint damage, and greater variability in treatment outcomes compared to their ACPA-negative counterparts. Despite extensive research over the past two decades, reliable biomarkers for predicting TNFi response remain elusive. One key reason is that the critical pathological processes driving treatment resistance are thought to be localized within the synovial tissue and cannot be adequately captured in peripheral blood. Moreover, the cellular origins of these tissue-specific mechanisms have not been clearly defined, limiting the development of targeted therapeutic strategies. The present study aims to identify tissue-specific molecular features associated with TNFi resistance in ACPA-positive RA and to determine their cellular origins using a multi-level transcriptomic approach that integrates bulk tissue, protein, and single-cell data.

This study conducted a multi-stage transcriptomic analysis based on publicly available datasets from the Gene Expression Omnibus and ArrayExpress repositories. First, we characterized baseline transcriptomic differences between ACPA-positive and ACPA-negative RA synovial tissues using the GSE198520 dataset, which contains synovial biopsy samples from 91 RA patients with detailed treatment response annotations. This step confirmed that the two ACPA subtypes exhibit distinct molecular profiles at the local disease site, justifying separate analyses. Second, we focused exclusively on ACPA-positive patients and compared gene expression profiles between TNFi non-responders (n=20) and responders (n=41) to identify resistance-associated genes using stringent statistical thresholds. Third, to evaluate tissue specificity, we constructed a resistance signature based on the differentially expressed genes identified in the synovium and applied it to two independent peripheral blood cohorts (GSE15258 with 75 samples and E-GEOD-58795 with 59 samples). This let us test whether the synovium-derived signature could distinguish responders from non-responders in the circulation, a critical test of its potential as a blood-based biomarker. Fourth, we validated the protein-level expression of the core resistance genes in whole blood using the Human Protein Atlas (HPA) database, which provides quantitative protein expression data from RA patient samples. Finally, we examined the cellular origins of these resistance-associated genes by analyzing an independent RA synovium single-cell RNA-seq dataset (GSE299518), which offers high-resolution cell-type annotation for synovial fibroblasts, macrophages, T cells, B cells, and other immune cell populations.

 In ACPA-positive synovial tissue, genes involved in oxidative phosphorylation, ribosome biogenesis, and proteasome function were significantly upregulated in non-responders compared to responders. A signature score constructed from these pathway-related genes effectively distinguished responders from non-responders in the synovial discovery cohort (p < 0.001). However, when this same signature was applied to two independent peripheral blood cohorts, it showed no significant difference in enrichment scores between responders and non-responders (both p > 0.05). Consistent with these transcriptomic findings, protein-level analysis using the Human Protein Atlas revealed that the core resistance genes, including ATP5MC1, COX6C, RPS27, and PSMB4, were undetectable or had very low expression in whole blood. This provides independent evidence that these resistance-associated molecules are not present in the circulation. Single-cell RNA-seq validation using an independent RA synovium dataset (GSE299518) further showed that these genes are predominantly expressed in synovial fibroblasts and macrophages, offering direct evidence for their cellular origins within the synovial tissue. Notably, this single-cell analysis showed that the resistance signature is rooted in specific cell populations rather than being a general inflammatory response.

TNFi resistance in ACPA-positive RA is characterized by transcriptomic features that are strictly confined to the synovial tissue, with the core resistance genes originating from synovial fibroblasts and macrophages. The inability to reliably detect these resistance-associated molecules in the bloodstream helps explain why previous efforts to develop blood-based predictive biomarkers have yielded inconsistent and often non-reproducible results. By integrating multi-level transcriptomic, protein, and single-cell evidence, this study provides testable hypotheses for future investigations aimed at developing synovium-targeted therapeutic strategies for ACPA-positive RA patients.