Background/Purpose: Psoriatic arthritis (PsA) poses an immense clinical burden with significantly increased morbidity and mortality risk compared to psoriasis alone. PsA may develop in 30% of psoriasis patients, however, a large proportion of individuals with PsA remain undiagnosed due to several factors, including a lack of patient and clinical awareness, and a lack of means of predicting which psoriasis patients will develop PsA.

Regulation of gene expression through DNA methylation can be altered by stochastic events or environmental factors and can potentially trigger and maintain PsA pathophysiological processes. With this research, we hope to identify DNA methylation changes that can predict which psoriasis patients will develop PsA at an early stage of the disease, helping prevent permanent joint damage and disability.

Methods: We obtained blood samples from 60 psoriasis patients that developed arthritis (converters) and 60 psoriasis patients that did not (biologic naive, matched for age, sex, psoriasis duration, and duration of follow up). Genome-wide DNA methylation was assessed using Infinium Methylation EPIC BeadChips (Illumina, San Diego, CA, USA). Array data preprocessing, normalization, and correction for technical sources of variation were performed in the R programming environment as recommended in the ChAMP package pipeline. Methylation differences between converters and non-converters were identified by a multivariate linear regression model including clinical covariates (age, sex, BMI, smoking) and conversion status using the Limma package.

The predictive performance of methylation markers was assessed by developing machine learning classification models. Support vector machine models were trained using 75% of samples, keeping the other 25% as an independent set for evaluating the prediction of conversion or non-conversion. Prediction performance of differentially methylated markers was compared to that achieved by an unbiased model trained on the complete set of methylation markers.

Results: We identified 36 significantly differentially methylated positions (with FDR-adjusted p-values lower than 0.05 and a minimum change in methylation of 0.05). This set of 36 highly relevant methylation markers were found across 15 genes and several intergenic regions. Enrichment analysis of the 15 genes with highly relevant methylation markers showed no significantly enriched functional pathways.

The support vector machine classification model for the set of 36 significantly methylated markers achieved an accuracy of 93%, outperforming an unbiased classification model based on the complete set of methylation markers that showed 90% accuracy.

Conclusion: We identified a set of 36 highly significant methylation markers associated with the development of PsA in psoriasis patients. This work shows that DNA methylation patterns at an early stage of psoriatic disease can distinguish between psoriasis patients that will develop PsA from those that will not.

« Back to ACR Convergence 2021

ACR Meeting Abstracts - https://acrabstracts.org/abstract/prediction-of-psoriatic-arthritis-in-patients-with-psoriasis-using-dna-methylation-profiles/