Background/Purpose: Extensive evidence has correlated epigenetic alterations in articular tissues with both the presence and progression of human osteoarthritis, but data regarding extraarticular tissues are lacking. Given recent findings implicating systemic inflammation in OA, we sought to characterize the DNA methylome of OA patients’ peripheral blood to ascertain whether PBMC methylation patterns were associated with rapid progression of human knee OA.

Methods: Peripheral blood mononuclear cell (PBMC) DNA was obtained from baseline blood draws of 72 OA patients enrolled in the Osteoarthritis Initiative (OAI) longitudinal study. All patients had baseline symptomatic and radiographic OA. 36 rapidly-progressive OA patients, defined as >= 1.0mm radiographic joint space loss or joint replacement within the first 24 months of follow-up were compared to 36 non-progressive OA patients defined as <= 0.5mm radiographic joint space loss over 48 months of follow-up. Sets were frequency matched for age, sex, race, BMI, and baseline K/L grade among the larger OAI cohort.

DNA methylation was quantified with Illumina HumanMethylation 450k arrays. Preprocessing was done in ChAMP, SNP-dependent and sex chromosome CpG sites were excluded from analysis, batch effects were corrected with frozen surrogate variable analysis. The estimateCellCounts function of minfi was utilized to ensure differential methylation results were not skewed by PBMC subset differences. GLMnet was used to derive machine learning-based algorithms correlating methylation of CpG sites with rapid progression, and correlations were validated on a random split of 30% of cases. This machine learning strategy was repeated 40 times and results compared. Genes associated with differential methylated sites and progression predictors were then submitted to Ingenuity (IPA) for ontological analysis.

Results: Our analysis identified 44 CpG sites as meeting our criteria for differential methylation, 35 of which were hypomethylated. Most were located within CpG islands or nearby shores or shelves, or upstream enhancers. The 14 associated genes were involved in canonical pathways including B cell development, antigen presentation, tRNA splicing, T helper cell differentiation, and IL4 signaling, among others. Many CpG sites were selected by our machine learning algorithms, the majority of predictive capability being provided by 29 CpG sites selected in at least 5 iterations of the model. Over multiple splits of data into training and testing sets, the predictive modeling strategy achieved a mean error rate of 33.3% on previously unseen validation data. The average receiver operating characteristic (ROC) curve of the model demonstrated an AUC of 0.75 for prediction of rapid progression. Ontologic analysis of genes selected by this algorithm were enriched in various pathways including the role of osteoblasts, osteoclasts, and chondrocytes in RA, NFAT signaling, EGF signaling, IL-10 signaling, TLR signaling, IL-6 signaling, and neuropathic pain signaling, among others.

Conclusion: Our data suggest that differential DNA methylation may be a readily-accessible biomarker for prediction of future radiographic progression in symptomatic knee OA patients. Further work will need to be done to confirm this pattern, and to define the specific cell populations which may be driving this differential methylation. These data further support the association of epigenetic modifications with human osteoarthritis.

« Back to 2016 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/differential-peripheral-blood-dna-methylation-patterns-are-predictive-of-radiographic-oa-progression/