DNA-hypomethylation in the TNF-alpha gene predicts rheumatoid arthritis classification in patients with early inflammatory symptoms

Abstract

AbstractBiomarkers for the classification of rheumatoid arthritis (RA), and particularly of anti-citrullinated peptide antibody (ACPA) negative patients, remains an important hurdle for the early initiation of treatment despite the use of 2010 classification criteria. Taking advantage of DNA-methylation patterns specific to early RA, quantitative methylation-specific qPCR (qMSP) offers a robust technology for the development of biomarker. We developed assays and established their value as RA classification biomarkers.Methods: DNA-methylation data were screened to select candidate to design for qMSP assays. 8 assays were developed and tested on 2 early inflammatory arthritis cohorts. Logistic regression and bootstrapping were used to demonstrate added value.Result: Differentially methylated CpG data were screened for candidate-CpG meeting the qMSP assay requirements. The top CpG candidate was in theTNFgene, for which we successfully developed a qMSP-assay. Significant lower DNA-methylation levels were observed in RA (p < 4x10− 9), with high predictive value (OR < 0.54/AUC < 0.198) in 2 cohorts (n = 127/n = 157). Regression using both cohorts showed improved accuracy = 87.7% and AUC = 0.944 over the model using only clinical variables (accuracy = 85.2%, AUC = 0.917). Similar data were obtained in ACPA-negative patients (n = 167, accuracy = 82.6%, AUC = 0.930) compared to clinical variable model accuracy = 79.5%, AUC = 0.892. Bootstrapping using 2000 datasets confirmed that the AUCs for the clinical + TNF-qMSP model had significant added value in both analysis.Conclusion: The qMSP technology is robust and can successfully be developed with high specificity of the TNF qMSP-assay for RA in patients with early inflammatory arthritis. It should assist classification in ACPA-negative patients providing a means of reducing delay to diagnosis and treatment.