Reprogramming of synovial macrophage metabolism by synovial fibroblasts under inflammatory conditions

The positional relationships between synovial macrophages (SM) and synovial fibroblasts (SF) have not been well characterized in mouse models of arthritis. To reveal localizations of both synovial cell types, we first established collagen antibody induced arthritis (CAIA) model mice and conducted histological analysis of joint tissues from these mice. The knee joints of CAIA mice exhibited remarkable swelling (Fig. 1a). Besides, H.E. staining of knee sections confirmed the presence of hyperplastic synovium that is characteristic of the arthritis phenotype (Fig. 1b). Immunofluorescent staining of the hyperplastic synovium from CAIA mouse knees revealed F4/80high- and F4/80low-positive SM were localized in regions adjacent to the inner surface of the synovium and articular cavity; in addition, inflammatory macrophage markers such as Tnf and Nos2 were expressed in F4/80high-positive SM (Fig. 1c). Cells that were positive for the inflammatory SF marker Podoplanin (Pdpn) were also located in a similar region (Fig. 1d). Synovial cells from CAIA mice were discontinuously and randomly distributed, but closely localized with each other in the inner region of the hyperplastic synovium while some SM were detached from SF (Fig. 1d). On the other hand, normal synovial membranes had few Tnf-positive SM and no detectable Nos2 (Additional file 2: Fig. S1A). Moreover, F4/80-positive SM continuously lined on ER-TR7-positive fibroblasts and no Pdpn was detected (Additional file 2: Fig. S1B). These data indicate that the normal spatial relationships between SM and SF were disrupted by arthritis and a de novo pathological microenvironment is associated with activation of SM in CAIA.

Histological analysis of knee joint tissue from CAIA mice. a Appearance of wild type mice with or without experimental arthritis (CAIA). b Sections of normal (left) and CAIA (right) knee joints stained with Hematoxylin and Eosin showing hyperplastic synovium (HS) in the joint (arrow head). Scale bar represents 500 μm. In immunofluorescent staining c F4/80-positive (green) cells were observed in the HS and articular cavity (AC). The inflammatory macrophage markers Nos2 and Tnf (red) co-localized in F4/80high-positive cells (arrow, yellow). Scale bar represents 20 μm. d Synovial cells were localized randomly in the HS. Panels on the right are high magnification images of boxed areas in the left panels. Podoplanin (Pdpn)-positive (red) cells were also observed in the HS. Scale bar represents 50 μm (left panel) and 20 μm (right panel). Histological data were technically replicated more than 2 times

The microenvironment in tissue niches affects cellular interactions to maintain or determine cell fate [2, 3, 5, 13, 26, 27]. To investigate cellular interactions between SM and SF, we generated primary cultures of bulk synovial cells from normal and CAIA ankle. To enrich the macrophage-like and fibroblast-like cells, isolation of SF and tissue resident macrophages were performed as previously reported with slight modification [18, 19]. We confirmed the enrichment of primary culture of macrophage-like cells from CAIA ankle and fibroblast-like cells from normal and CAIA ankle, even though we could not isolate enough amount of macrophage-like cells from normal ankle (Fig. 2a). Then, the expression of pan-macrophage and SF markers in respective synovial cells were analyzed by RT-qPCR (Fig. 2b). Pan-macrophage markers (CD68, Emr1, ItgαM, Csf1r) were expressed to significantly higher levels in macrophage-like cells than in fibroblast-like cells derived from normal and CAIA ankle. Meanwhile, the expression levels of SF markers (Vcam1, Cdh11, Col6a1) was significantly higher in both fibroblast-like cells compared to macrophage-like cells, and Csf1 expression levels was significantly elevated in fibroblast-like cells derived from CAIA ankle compared to that derived from normal ankle (Fig. 2b). Flow cytometry analysis confirmed few other leukocytes were included in macrophage-like cells (Fig. 2c). These data verified the successful separation of arthritis tissue-derived SM (ADSM), normal tissue-derived SF (NDSF) and arthritis tissue-derived SF (ADSF). Next, we analyzed the gene expression profiles of ADSM, NDSF and ADSF by RNA-seq. Hierarchical cluster analysis of selected transcriptomes of pan-macrophage and SF markers revealed distinct gene expression patterns for ADSM and both SFs (Fig. 2d). In addition, principal-component analysis (PCA) also revealed separate gene clusters in ADSM, NDSF and ADSF (Fig. 2e). Recently, heterogeneous subsets of both SM and SF in synovium tissue of K/BxN serum transfer arthritis model mice has been reported using single-cell RNA-seq analysis [28, 29]. In these studies, SM and SF were classified 6 and 5 subsets, respectively. Thus, mRNA expression of major marker genes of each subset in isolated ADSM and ADSF were examined. RNA-seq data revealed that expression levels of Fxyd2 (SM-subset 1 marker), Cx3cr1 (SM-subset 4 marker), Stmn1, Ube2c and Birc5 (SM-subset 5 marker) were relatively higher than other genes in ADSM (Additional file 2: Fig. S2A) and expression levels of Sfrp2, Col11a1 (SF-subset 1 marker), Inhba, Prg4 (SF-subset 2 marker), Top2a, Hmgb2 and Cdk1(SF-subset 4 marker) were relatively higher than other genes in ADSF (Additional file 2: Fig. S2B). These data indicate that heterogeneity of isolated ADSM and ADSF may be lost and each cell population was changed into homogeneous with multiple subset properties under cultured condition. Comparing RNA-seq data sets after TMM normalization, the expression levels of 2175 genes in NDSF and 2007 genes in ADSF were more than 16-fold higher than in ADSM (value > MEDIAN and Log2 Fold-Change > 4). In addition, integrative analysis of the two data sets allowed extraction of 212 genes that were specifically expressed in ADSF (Fig. 2f). To characterize these genes, we performed Gene Ontology (GO) analyses using Database for DAVID bioinformatics resources. Top of functional annotation clustering was associated with secreted-related gene (Fig. 2g). Besides, GO biological process and KEGG pathway analysis revealed that immune response- and cytokine-related genes were enriched among ADSF specific genes (Additional file 2: Fig. S2C). These data suggest that secretory stimulation from ADSF is a key cellular interaction between SF and SM in the pathological synovial microenvironment.

Gene expression profile in arthritis tissue-derived synovial macrophages (ADSM), normal tissue-derived synovial fibroblasts (NDSF) and arthritis tissue-derived synovial fibroblasts (ADSF). a ADSM, NDSF and ADSF were isolated from 4 independent ankles with or without CAIA. Representative phase-contrast images of ADSM, NDSF and ADSF. Scale bar represents 100 μm. b Gene expression of pan-macrophage and SF markers in ADSM, NDSF and ADSF were analyzed by RT-qPCR (n = 4). ** indicates P