Circulating microRNA signatures identified in early versus late knee osteoarthritis

Abstract

Purpose: There continues to be a lack of specific osteoarthritis (OA) biomarkers that can be used in a therapeutic, prognostic, or diagnostic manner. Since microRNAs are promising biomarkers for several diseases, we previously used microRNA PCR-arrays and identified a panel of microRNAs in the synovial fluid that differentiate between early- and late-stage radiographic knee OA. We also performed microRNA microarrays and identified 2 microRNAs as mediators of cartilage degeneration. However, the gold standard approach to identifying biomarkers is next generation sequencing because it offers the sensitivity and specificity to detect novel and low abundance microRNAs that are unique to various cohorts (early versus late OA, men versus women, young versus old, normal weight versus overweight). Here, we use next generation sequencing to identify signatures of circulating microRNAs as biomarkers for knee OA. We hypothesize that identification and characterization of known and novel circulating microRNAs can be used as biomarker signatures to define cohorts of patients with OA. Methods: Plasma samples from knee OA patients are obtained from the Knee Osteoarthritis BioBank at the University Health Network in Toronto, Canada. Samples include healthy donors with no history of musculoskeletal disease (N=100), Kellgren-Lawrence grades 0 and 1 for early OA (N=100), and Kellgren-Lawrence grades 3 and 4 for late OA (N=750). Patients with comorbidities that might affect microRNA signatures are excluded. In a pilot experiment, plasma samples from 5 healthy donors, 5 early OA patients, and 5 late OA patients were subjected to next generation sequencing of microRNAs. Differentially expressed microRNAs were identified using the negative binomial exact test. To validate our findings, we are currently sequencing plasma from larger cohorts of patients. Further statistical analysis will be used to identify relationships between microRNA signatures and known risk factors such as sex, age, and body mass index. Bioinformatic methods will be used to predict the targets and biological function of candidate microRNAs in OA. Results: After mapping sequencing reads to mirbase20, 164 microRNAs with greater than 10 counts per million were identified in all samples. Principal component analysis was performed using the 50 microRNAs with the largest coefficient of variation based on normalized counts. This revealed a clear separation of early OA samples from both late OA and healthy samples. The most differentially expressed microRNAs were identified based on false discovery rate less than 0.05, log counts per million greater than 2, and log fold change greater than 1.5. Hierarchical clustering of these microRNAs revealed a distinct pattern where 60 microRNAs were upregulated only in early OA samples (Figure 1). Among these are 4 novel putative microRNAs that have not previously been identified. We are currently validating these findings with additional sequencing experiments in each group (N=100 healthy donors, N=100 early OA, N=750 late OA) to establish a clear microRNA signature for early knee OA. Conclusions: Data from our pilot experiment demonstrate that next generation sequencing is a useful approach for identifying known and novel circulating microRNAs in OA. Sequencing of early OA samples will allow identification of microRNA signatures that can be used to distinguish these patients from late OA patients. Circulating microRNAs may represent valid and reliable biomarkers with potential applications for improving OA detection and treatment.