AUTHOR=Hur Benjamin , Gupta Vinod K. , Oh Minsik , Zeng Hu , Crowson Cynthia S. , Warrington Kenneth J. , Myasoedova Elena , Kronzer Vanessa L. , Davis John M. , Sung Jaeyun 

TITLE=Integrative multi-omic profiling in blood reveals distinct immune and metabolic signatures between ACPA-negative and ACPA-positive rheumatoid arthritis

JOURNAL=Frontiers in Immunology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1667662

DOI=10.3389/fimmu.2025.1667662

ISSN=1664-3224

ABSTRACT=ObjectiveTo investigate whether patients with ACPA-negative (ACPA–) and ACPA-positive (ACPA+) rheumatoid arthritis (RA) exhibit distinct immune and metabolic profiles in blood, using integrative proteomic and metabolomic analyses. By uncovering subgroup-specific molecular signatures, we aim to improve the biological understanding of RA heterogeneity and support the development of more precise diagnostic and stratification strategies.MethodsWe performed high-throughput proteomic and metabolomic profiling on plasma from a well-characterized cohort comprising 40 patients with ACPA– RA, 40 patients with ACPA+ RA, and 40 healthy controls. To identify key immune and metabolic differences, we applied statistical comparisons, pathway enrichment analyses, and network inference methods. Additionally, an integrative network-based machine learning framework was used to distinguish RA subgroups from controls based on plasma molecular profiles.ResultsACPA– and ACPA+ RA exhibited distinct plasma proteomic and metabolomic biomolecular signatures. Complement proteins (CFB, CFHR5, and F9) and the anti-inflammatory cytokine IL1RN were exclusively elevated in ACPA– RA and remained distinct in a treatment-naïve sub-cohort. Metabolomic analysis revealed subgroup-specific differences in lipid and pyrimidine metabolism, including contrasting patterns in bilirubin-derived metabolites. Correlation analyses identified differential associations between molecular features and clinical inflammatory markers across RA subgroups. An integrative machine learning framework incorporating multi-omic features achieved high classification performance in cross-validation (AUC ≥ 0.90), outperforming models based on single-omic data.ConclusionThis study suggests that ACPA status may not fully capture the biological heterogeneity between ACPA– and ACPA+ RA subgroups, indicating additional immune and metabolic distinctions that warrant further investigation. Our findings highlight the potential of multi-omic profiling to enhance RA diagnostics, refine disease stratification, and inform subgroup-specific disease management strategies.