Synovial Inflammation of Osteoarthritis and Rheumatoid Arthritis Revealed by Single-Cell and in Silicodeconvolution Bulk RNA Sequencing

Z. Huang,Z. Luo,Y. Cai,C-H Chou,F. Pei,V. Kraus,Z. Zhou

DOI: https://doi.org/10.1016/j.joca.2021.02.462

IF: 7.507

2021-01-01

Osteoarthritis and Cartilage

Abstract:Purpose: Synovitis plays an important role in the pathogenesis of arthritis. Our objective was to differentiate the inflammatory phenotypes of osteoarthritis (OA) and rheumatoid arthritis (RA) synovium through single-cell RNA sequencing (scRNA-seq) and to test and validate the methodology by using bulk-RNA-seq data to create in silico scRNA-seq data for synovial tissue. Methods: OA data were provided by 17 patients undergoing total knee arthroplasty (TKA) for OA: 14 with scRNA-seq of 93,208 OA synoviocytes and 9 matched bulk RNA-seq data; and 3 with scRNA-seq of 8,869 OA synoviocytes (NCBI GEO GSE152805). RA data were provided by 5 TKA patients with scRNA-seq data of 20,172 RA synoviocytes (phs001529.v1.p1). Transcriptome-wide analyses were performed between these two groups. Nine paired OA scRNA and bulk RNA datasets were used to evaluate six in silico gene deconvolution tools. Predicted and observed cell types and proportions were compared to identify the best deconvolution tool for synovium. Results: Significantly higher proportions of T (p=0.001) and B (p=0.007) cells were found in RA while significantly higher proportions of smooth muscle cells (p=0.006) were identified in OA (Figure 1). Through gene set enrichment analyses, OA and RA were distinguished on the basis of five patterns (complement system, inflammation response, chemokine and cytokine, leukocyte trans-endothelial migration, and heat shock protein family). Gene deconvolution analyses identified three (of six) platforms (CIBERSORT, QUANTICEQ and MCPCOUNTER) as suitable for extrapolating cellular gene expression from bulk RNA-seq data. In addition, using our paired scRNA-seq and bulk RNA-seq data, an "arthritis" specific signature matrix and was validated to have significantly better prediction performances for synoviocytes (p=0.049) (Figure 2). Conclusions: Differences of cell proportions, gene expression patterns and differentially enriched pathways in seven distinct cell types identified by scRNA-seq revealed different inflammation phenotypes of OA and RA. We established a road-map for evaluating bulk RNA-seq deconvolution methods and provided guidelines to researchers working on synovium to select appropriate methods for characterization of the cellular context of the synovial tissues.Figure 2Significant improvements in terms of predictive accuracy after using the generated "Arthritis" specific signature compared to the default one. The correlation analysis was performed by Pearson correlation.View Large Image Figure ViewerDownload Hi-res image Download (PPT)