Taxonomy of fibroblasts and progenitors in the synovial joint at single-cell resolution

Abstract

Objectives: Fibroblasts in synovium include fibroblast-like synoviocytes (FLS) in the lining and Thy1+ connective-tissue fibroblasts in the sublining. We aimed to investigate their developmental origin and relationship with adult progenitors.

Methods: To discriminate between Gdf5-lineage cells deriving from the embryonic joint interzone and other Pdgfrα-expressing fibroblasts and progenitors, adult Gdf5-Cre;Tom;Pdgfrα-H2BGFP mice were used and cartilage injury was induced to activate progenitors. Cells were isolated from knees, fibroblasts and progenitors were sorted by fluorescence-activated cell-sorting based on developmental origin, and analysed by single-cell RNA-sequencing. Flow cytometry and immunohistochemistry were used for validation. Clonal-lineage mapping was performed using Gdf5-Cre;Confetti mice.

Results: In steady state, Thy1+ sublining fibroblasts were of mixed ontogeny. In contrast, Thy1-Prg4+ lining fibroblasts predominantly derived from the embryonic joint interzone and included Prg4-expressing progenitors distinct from molecularly defined FLS. Clonal-lineage tracing revealed compartmentalisation of Gdf5-lineage fibroblasts between lining and sublining. Following injury, lining hyperplasia resulted from proliferation and differentiation of Prg4-expressing progenitors, with additional recruitment of non-Gdf5-lineage cells, into FLS. Consistent with this, a second population of proliferating cells, enriched near blood vessels in the sublining, supplied activated multipotent cells predicted to give rise to Thy1+ fibroblasts, and to feed into the FLS differentiation trajectory. Transcriptional programmes regulating fibroblast differentiation trajectories were uncovered, identifying Sox5 and Foxo1 as key FLS transcription factors in mice and humans.

Conclusions: Our findings blueprint a cell atlas of mouse synovial fibroblasts and progenitors in healthy and injured knees, and provide novel insights into the cellular and molecular principles governing the organisation and maintenance of adult synovial joints.

Keywords: Fibroblasts; Osteoarthritis, Knee; Synovitis.

Figure 1

Single-cell transcriptomic atlas of adult mouse synovial fibroblasts. Cells were isolated from knees of 11-week-old Gdf5-Cre;Tom;Pdgfrα-H2BGFP mice, sorted by FACS into Gdf5-lineage cells (Tom+) and other cells expressing the pan-fibroblast marker Pdgfrα (Tom-GFP+), and analysed by scRNA-seq. (A) Histology showing Tom+GFP+ and Tom-GFP+ cells in synovium of 15-week-old Gdf5-Cre;Tom;Pdgfrα-H2BGFP mouse knee (n=3). Blue: DAPI nuclear counterstain. White outline on left shown at higher magnification on right. scale bars: 20 µm. S: synovium; P: patella; F: femur; SL: synovial sublining; L: synovial lining. (B) Experimental workflow with flow cytometry scatter plot showing cell populations sorted by FACS. See online supplemental figure 1 for extended data. (C) Unsupervised UMAP plot of integrated scRNA-seq data from 786 Tom+ cells (n=2 mice) and 376 Tom-GFP+ cells (n=1 matched mouse). Left: unsupervised clustering. Right: colour-coded by analysed cell population. See online supplemental figure 2 for extended data. (D) Analysis of differentially expressed genes (DEGs) to identify clusters. Heatmaps show expression of top 50 DEGs for each cluster. Key genes indicate selected DEGs that identify cell types or are dominant cluster-specific genes. See online supplemental figure 3 for UMAP plots of key genes and online supplemental table 5 for top 10 DEGs for each cluster. (E) UMAP plots showing expression of Thy1 and Cd34 by fibroblast clusters F1-F6. (F) Immunohistochemical detection of Cd34 in synovial sublining (SL) of 11-week-old mouse knee (n=7). Scale bar: 20 µm. (G) Over-representation analysis of gene ontology (GO) categories for the identified Thy1+ fibroblast clusters. (H) Inferred lineage trajectories within the Tom+ and the Tom-GFP+ cell populations based on Slingshot unbiased pseudotime analysis visualised using principal curves. Within both ontogenetic cell lineages, the F4 fibroblast cluster was predicted to represent a root state. (I) UMAP plots showing expression of the cross-tissue universal fibroblast marker Pi16 and Cd55 by fibroblast cluster F4. (J) Heatmap showing expression of top 100 DEGs of the Pi16+ cross-tissue fibroblast cluster identified by Buechler et al by the clusters identified in the adult mouse knee. UMAP, uniform manifold approximation and projection; FLS, fibroblast-like synoviocytes; C, chondrocyte-lineage cells; O, osteoblast-lineage cells; T, tenocyte-lineage cells; F, fibroblasts.

Figure 2

Ontogenetic compartmentalisation of synovial lining and sublining fibroblasts. (A) Detection of Thy1-Itga6+ FLS and Thy1+Cd55+ universal fibroblasts within Tom+GFP+ and Tom-GFP+ cell populations in indicated joints of 21–23 week old Gdf5-Cre;Tom;Pdgfrα-H2BGFP mice (n=8) by flow cytometry. Flow plots show gating to identify cell populations. See online supplemental figure 6 for full gating strategy and controls. Graphs show percentage of Thy1-Itga6+ FLS and Thy1+Cd55+ fibroblasts within Tom+GFP+ and Tom-GFP+ cell populations in each of the joints. P values: two-way repeated-measures ANOVA with Holm-Sidak post-test after log-transformation. (B) Tom+cells expressing the FLS marker Clic5 (arrowheads) and adjacent Tom+Clic5- cells (arrows) in synovial lining of 10-week-old Gdf5-Cre;Tom;Pdgfrα-H2BGFP mice detected by immunofluorescence staining. Scale bars: 10 µm. Graph shows the percentage of Clic5+ FLS that are Tom+ (n=3). See online supplemental figure 7 for low magnification image and isotype negative control staining. (C) Clonal lineage analysis in 13–16 week old Gdf5-Cre;Confetti mouse knees. Clonal cell clusters in synovium, marked by expression of cerulean fluorescent protein (CFP), yellow fluorescent protein (YFP) or red fluorescent protein (RFP), show alignment along the proximal-distal axis parallel to synovial lining. Scale bar: 20 µm. S: synovium; P: patella; F: femur. Graph shows percentage of labelled cells with at least one neighbouring cell expressing the same fluorescent protein in medial or lateral (M/L), or proximal or distal (P/D), direction (n=13; seven unoperated mice and six unoperated contralateral knees from injured mice). P value: paired two-tailed t-test. (D) Expression of selected marker genes to identify subclusters in the steady-state scRNA-seq data, as indicated on the UMAP plot on the right. Cells within a Prg4+ progenitor subcluster (P) expressed Prg4 and Creb5, but were largely negative for other FLS and chondrocyte-lineage markers. Tom+ cells in the chondrocyte-lineage cluster (C) displayed a superficial zone chondrocyte (SZC) phenotype, while Tom- cells displayed a mature/hypertrophic chondrocyte phenotype and likely derived from growth plate. An additional Tom+ subcluster was identified as vascular smooth muscle cells (V). See online supplemental figure 8 for extended data. (E) UMAP plots showing expression of indicated genes. Dotted outline indicates enlarged region shown on right to highlight expression in FLS and the Prg4+ progenitor subcluster (P). All graphs show mean ± 95% CI. ANOVA, analysis of variance; UMAP, uniform manifold approximation and projection; FLS, fibroblast-like synoviocyte; SZC, superficial zone chondrocytes; VSMC, vascular smooth muscle cells.

Figure 3

Single-cell transcriptomic atlas of synovial fibroblasts from 12-week-old Gdf5-Cre;Tom;Pdgfrα-H2BGFP mice 6 days after joint surface injury, integrated with steady-state data shown in figure 1. (A) Histology showing Tom+GFP+ and Tom-GFP+ cells in synovium 6 days after injury (n=3). Blue: DAPI nuclear counterstain. White outline on left shown at higher magnification on right. Scale bars: 20 µm. S: synovium; P: patella; F: femur; SL: synovial sublining; L: synovial lining. (B) UMAP plot of integrated scRNA-seq data from 786 steady-state Tom+ cells (n=2 mice), 376 steady-state Tom-GFP+ cells (n=1 matched mouse), 2383 injured-state Tom+ cells (n=4 mice) and 651 injured-state Tom-GFP+ cells (n=2 matched mice). Left: unsupervised clustering. Right: colour-coded by analysed cell population and state. Injury-induced fibroblast (IF) clusters are in bold; clusters with steady-state analogues in italics. See online supplemental figure 10–12 for extended data. (C) Expression of key genes that identify clusters. See online supplemental figure 13 for extended data. (D) Relative abundance of cells across identified cell clusters. *FDR<0.05 vs steady-state, negative binomial generalised linear model with Benjamini-Hochberg post-test. (E) Detection by flow cytometry of Thy1-Itga6+ FLS and Thy1+Cd55+ universal fibroblasts within the Tom+GFP+ and Tom-GFP+ cell populations in control (Ctl) and injured (Inj) knees of 15–19 week old Gdf5-Cre;Tom;Pdgfrα-H2BGFP mice 6 days post-injury (n=7). See online supplemental figure 14 for gating strategy and controls. Graphs show percentage of cells within the respective cell populations that are Thy1-Itga6+ FLS or Thy1+Cd55+ fibroblasts. Lines and error bars: mean ± 95% CI. P values: two-way repeated-measures ANOVA with Holm-Sidak post-test. (F) Regulons active in injury-induced fibroblast clusters IF1-3. (G) Expression of lineage-specifying transcription factors. Venn diagram shows percentage of cells in injury-induced clusters IF1-IF3 expressing Sox9, Runx2 and Scx. See online supplemental figure 17 A, B for other cell clusters. (H) Immunohistochemical detection of Sox9 and Runx2 in synovium (S) and periosteum (P) in consecutive tissue sections of control and injured knees of 11–13 week old mice 7 days post injury (n=5). Scale bars: 20 µm. ANOVA, analysis of variance; UMAP, uniform manifold approximation and projection; FDR, false discovery rate; C, chondrocyte-lineage cells; F, fibroblasts; FLS, fibroblast-like synoviocytes; IF, injury-induced fibroblasts; OC, osteochondral-lineage cells; T, tenocyte-lineage cells; VSMC, vascular smooth muscle cells.

Figure 4

Integrated single-cell transcriptomic analysis of synovial fibroblasts from mice 6 days after joint surface injury (data shown in figure 3) and day 9 serum-transfer induced arthritis (STIA) mice. (A) UMAP plot of integrated scRNA-seq data from 2955 Tom+ and Tom-GFP+ injured-state cells and 3549 STIA cells. Left: unsupervised clustering. Right: colour-coded by state. See online supplemental figures 19, 20 for cluster annotation. Perturbed-state fibroblast (PF) clusters are in bold; clusters with steady-state analogues in italics. (B) Expression of key genes that identify steady-state analogous clusters (top) and perturbed-state clusters (bottom). (C) Relative abundance of cells across identified clusters. *FDR<0.05 vs injured-state, negative binomial generalised linear model with Benjamini-Hochberg post-test. (D) Expression of lineage-specifying transcription factors. Venn diagrams show percentage of STIA cells and injured-state cells in perturbed-state clusters (PF1-PF5) expressing Sox9, Runx2 and Scx. (E) Heatmaps showing expression of differentially expressed genes (DEGs) involved in inflammation and catabolism in non-proliferating perturbed-state fibroblast clusters (PF2-PF5). Left: integrated injured-state and STIA cells. Middle: Injured-state cells only. Right: STIA cells only. UMAP, uniform manifold approximation and projection; FDR, false discovery rate.

Figure 5

Stromal cell proliferation and differentiation trajectories after injury. (A) RNA velocity analysis using the dynamical model projected onto the integrated steady-state and injured-state uMAP plot as shown in figure 3B. Arrows show the local average velocity and direction. (B) Monocle3 lineage inference showing differentiation trajectories originating from injury-induced fibroblast clusters with a branchpoint towards Thy1+ fibroblasts or FLS. Colouring of plot represents pseudotime. (C) UMAP plot of cell cycle gene module score based on expression of Mki67, Ccna2, Ccnb1, Ccnb2 and Cdk1. Arrows indicate proliferating cells. See online supplemental figure 21 for separate analysis of injured-state Tom+, injured state Tom-GFP+, and steady-state populations. (D, E) Immunofluorescence staining in 12–14 week old Gdf5-Cre;Tom;Pdgfrα-H2BGFP mouse synovium 6 days after injury (n=6) to locate proliferating fibroblasts. (D) Ki67+GFP+ proliferating fibroblasts in synovial lining (arrows), and near Cd31+blood vessels in synovial sublining (arrowheads). See online supplemental figure 22A for split channel images and isotype negative control staining. Graph shows the median distance to the nearest Cd31+ blood vessel for Ki67+GFP+ proliferating fibroblasts and Ki67-GFP+ non-proliferating fibroblasts in synovial sublining. Lines and error bars indicate geometric mean ± 95% CI. P value: Wilcoxon signed rank test. (E) Ki67+Tom+ proliferating Gdf5-lineage cells in synovial lining (arrows), adjacent to Clic5+Tom+ FLS. see online supplemental figure 22B for split channel images and isotype negative control staining. (F) Clonal lineage analysis in 13–14 week old Gdf5-Cre;Confetti mouse knees 6 days after injury (n=6). Ctl: unoperated contralateral control knees. Cerulean fluorescent protein (CFP), yellow fluorescent protein (YFP), red fluorescent protein (RFP) and TO-PRO-3 nuclear counterstain were detected by confocal fluorescence microscopy. Graph shows percentage of labelled cells with at least one neighbouring cell expressing the same fluorescent protein in the medial or lateral (M/L), or the proximal or distal (P/D), direction. Lines and error bars indicate mean ± 95% CI. P value: two-way repeated-measures ANOVA with Holm-Sidak post-test. ANOVA, analysis of variance; UMAP, uniform manifold approximation and projection.

Figure 6

Transcriptional programmes regulating fibroblast differentiation trajectories. (A–C) Transcription factors and regulons associated with the two fibroblast differentiation trajectories (Thy1+ fibroblasts and FLS) identified in figure 5. (A) Heatmaps showing expression across pseudotime of transcription factors whose expression changed significantly across either trajectory. Transcription factors are clustered by pseudotemporal expression pattern. Top differentially expressed transcription factors are indicated. (B) Activity of selected regulons associated with transcription factors identified by pseudotime analysis, projected onto the integrated steady-state and injured-state UMAP plot as shown in figure 3B. See online supplemental figure 23 for regulon activity analysed independently in steady state and injured state. (C) Expression of selected FLS-associated transcription factors identified from the pseudotime analysis projected onto the integrated steady-state and injured-state UMAP plot. Expression of Sox5 and Foxo1 is mostly restricted to the FLS, while Creb5 shows a wider expression pattern. (D) Immunohistochemical detection of SOX5 and FOXO1 in near-consecutive tissue sections of 11–13 week old mice 7 days post-injury (n=5) showing expression in synovial lining of control (unoperated contralateral control knee) and injured knees. Scale bars: 20 µm. See online supplemental figure 24A for lower magnification images and isotype negative control stainings. (E) Predicted FLS-associated gene regulatory network based on the mouse scRNA-seq data, constructed using GENIE3 and Cytoscape, indicating that the transcription factors Sox5 and Foxo1 drive key FLS-associated genes. (F) UMAP plots of scRNA-seq data from human OA synovium obtained at arthroplasty (n=3; 2 female and one male, mean age 68) showing from left to right expression of indicated marker genes to identify FLS (THY1-PRG4+CLIC5+), expression of transcription factors SOX5, FOXO1 and CREB5, and activity of associated regulons. See online supplemental figure 25 for analysis of two additional human OA synovial cell datasets. (G) Immunohistochemical detection of SOX5 and FOXO1 in near-consecutive tissue sections of human OA synovium obtained at arthroplasty (n=6; see online supplemental table 1 for donor information), showing expression in quiescent and active areas of synovial lining. Scale bars: 20 µm. See online supplemental figure 24B for lower magnification images and isotype negative controls. UMAP, uniform manifold approximation and projection; FLS, fibroblast-like synoviocytes; OA, osteoarthritis.