Supplementary Figure 11: Bulk RNA-seq data analysis.

a, Quality control of bulk RNA-seq samples. Common genes are defined as the set of genes detected with at least 1 mapped fragment in 95% of the samples (13,041 genes). X axis is the number of cells for each bulk RNA-seq sample. Y axis is the percentage of detected common genes for each sample. We discarded 25 low quality samples that have less than 99% (dashed line) of common genes detected, resulting 167 post–QC samples in all. b, PCA analysis on all the post–QC samples shows that most of the variance in the bulk RNA-seq data is due to cell type. c, Cell type marker genes show that there is no obvious contamination in the bulk RNA-seq data. d–g, PCA analysis on samples from each cell type. The samples from leukocyte-rich RA appear distinct from leukocyte-poor RA and OA samples. This difference in transcriptional signatures in inflamed tissue is largely determined by altered cellular composition. h,i, Distribution of significantly enriched GO terms in leukocyte-rich RA by GSEA. Leukocyte-rich fibroblasts and monocytes share the common pathways of Type I interferon and inflammatory response. j, Correlation between bulk RNA-seq genes and immune cell type abundances in RA synovial fibroblasts. Integrating bulk RNA-seq samples from fibroblasts with multiple cell type flow gates reveals that T cells, B cells, and monocytes that are abundant in RA synovial tissue directly influence the expression of fibroblasts in the RA synovium.