Identifying molecular signatures underlying knee osteoarthritis through transcriptomics

Purpose: Osteoarthritis (OA) is a pervasive, lifelong disorder targeting joints, and is a leading cause of mobility-disability worldwide. Yet, there are no successful pharmacological approaches that can stop its progression. This substantial unmet need can be attributed to the complex and poorly understood molecular mechanisms which initiate and drive this disease. Total RNA sequencing is a powerful technique which enables the identification of active genes driving disease progression, as well as non-coding epigenetic changes that influence whether gene changes result in functional changes. Due to the heterogeneity in OA, it is increasingly viewed as a disease with different sub-types or phenotypes which may influence approaches to diagnoses and therapy. Methods: We have used our large-scale knee OA patient biobank (n=750) to perform total RNA sequencing (Illumina Truseq-Stranded Total RNA, NextSeq550) on synovium samples from 50 late stage (Kellgren-Lawrence Grade 3/4) radiographic knee OA patients. We have developed separate bioinformatic pipelines for mRNA, long non-coding RNA and circular RNA. We have also collected substantive anthropometric, pain and functional data in order to elucidate relationships between clinical variables and molecular signatures. Results: RNA extracted from synovium had an average RIN of 8.6±0.72, and RNA sequencing metrics indicate an average Q30 of over 90% indicating 99.9% base-call accuracy. After filtering, 19,857 genes were expressed in synovium. Sex, BMI (normal, overweight, obese), synovial inflammation, 1 year response to surgical intervention (WOMAC MCID, patient response), as well as baseline pain (WOMAC pain, stiffness, function) distinguished a number of differentially expressed genes. Multivariate analysis further indicated that subsets of genes can be distinguished by synovial inflammation, and responder status after adjusting for other factors. Unsupervised cluster analysis indicates that molecular signatures between patients cannot be distinguished by clinical variables alone, and these variables themselves do not self-segregate patient clusters. Conclusions: Transcriptomics of late-stage OA synovial tissue indicates a number of differentially regulated genes by demographic and clinical variables. Phenotyping of these patients indicates that clinical variables alone cannot distinguish sub-groups, but combinations of these variables with molecular signatures or molecular signature independent of clinical phenotype may be able to distinguish patient sub-groups. Future directions include incorporation of long non-coding and circular RNA molecular signatures to form a comprehensive molecular profile of each patient, as well as RNA sequencing of other joint tissues within the same cohort.