Anti-citrullinated peptide antibodies (ACPA) can be detected years before rheumatoid arthritis (RA) onset and confer high diagnostic specificity. However, up to 40% of patients develop RA in the absence of ACPA, leading to delayed diagnosis, limited molecular stratification, and largely empirical treatment strategies. The immunopathogenic mechanisms underlying ACPA-negative (ACPA-neg) RA remain poorly defined, hindering the development of precise diagnostic and therapeutic approaches.

We performed integrated single-cell RNA sequencing (scRNA-seq) analysis of peripheral blood mononuclear cells (PBMCs) and synovial tissue from ACPA-neg and ACPA-positive (ACPA-pos) RA patients. Cellular heterogeneity, transcriptional regulation, immunometabolic reprogramming, and intercellular communication networks were systematically and comprehensively characterized. To address diagnostic challenges, we developed an interpretable machine-learning model based on peripheral blood gene signatures and further validated key findings in an independent clinical cohort using reverse transcription-quantitative polymerase chain reaction (RT-qPCR).

ACPA-neg RA was characterized by a marked expansion of synovial cytotoxic CD4⁺ T lymphocytes (CD4⁺ CTLs), which exhibited enhanced cytotoxic programs, distinct transcription factor activity (KLF4, SPI1, ATF5), and a unique immunometabolic profile marked by upregulated lipid metabolism and suppressed glutathione pathways. Trajectory analysis suggested a continuous differentiation link between naïve/proliferating CD4⁺ T cells and CD4⁺ CTLs in ACPA-neg RA. Cell–cell communication analysis revealed streamlined yet high-intensity signaling networks preferentially targeting CD4⁺ CTLs via CD45, FN1, and ANNEXIN pathways. Leveraging a CD4⁺ CTL–associated gene module, we constructed an interpretable machine learning diagnostic model that accurately distinguished ACPA-neg RA from healthy controls (AUC = 0.88). SHAP analysis identified GZMK, GNLY, IFNG, and NKG7 as key contributors, which were further validated in patient samples.

Our study identifies cytotoxic CD4⁺ T cells as key immunopathogenic drivers of ACPA-negative RA and establishes a clinically actionable, potentially translatable diagnostic framework. These findings define ACPA-neg RA as a T cell–driven immunometabolic disease entity and provide a foundation for precision diagnosis and targeted therapeutic strategies.