1. scRNA-seq revealed 20 different cell types in synovium and meniscus samples.
To identify the expression profiles of cells in the normal and degenerated synovium and meniscus samples, as shown in Fig. 1-Figure Supplement 2A, a total of 124,026 cells from 22 samples of 16 KOA patients were sequenced by scRNA-seq (Fig. 1A, Fig. 1-Figure Supplement 1A-B), including 36,635 synoviocytes (synovial lining fibroblasts, synovial sublining fibroblasts, synovial sublining fibroblasts, CXCL14 + Fibs, Fibroblast_1, and Fibroblast_2), 38,926 meniscus cells (Ch.1, Ch.2, Ch.3, Ch.4, and Ch.5), 23,856 immune cells (B cells, T cells, dendritic cells, mast cells, macrophages), 16,223 endothelial cells (capillary-venous and capilary-arterial), 9,376 pericyte-like cells (pericyte-like cells_1 and pericyte-like cells_2). These cells were divided into 24 separate clusters using the unsupervised clustering algorithm. The heterogeneity of cell types was significant (Fig. 1-Figure Supplement 1C and Fig. 1-Figure Supplement 2C). These cells were visualized through the uniform manifold approximation and projection (UMAP) and t-Distributed Stochastic Neighbor Embedding (tSNE) (Figs. 1B-E). The number of genes per cell and that of counts per cell, as well as the PCA of genes with large variations, are shown in Fig. 1-Figure Supplement 1A, B. Except for the H12 inner samples from the public database, other samples were of good quality.
The cell type was identified by markers in Fig. 1E, Fig. 1-Figure Supplement 3, and Fig. 1-Figure Supplement 2B. The cell clusters with high expression of cartilage-related genes were renamed chondrocytes, such as CHAD, PRG4, TGF-βI, CDF, STMN1, COL3A1, DCN, and LUM. Additionally, 5 kinds of meniscus fibroblasts were defined as per Fu2, including Ch1(CHAD+), Ch2(FNDC1+), Ch3(PRG4+), Ch4(CFD+), and Ch.5(STMN1+) (Fig. 1-Figure Supplement 5A). The cells with high expression of VWF and PECAM1 were identified as endothelial cells, including endothelial cell (capilary-venous) (ECs(CAP-V))(GJA5+) and endothelial cell(capilary-arterial) (ECs(CAP-A))(VWF+). The smooth muscle cells and mural cells with high expression of ACTA2, MYL9, and TAGLN were named pericyte-like cells, PCL.1 (ACTA2 + and FABP4+), and PCL.2 (ACTA2 + and MYH11+) (Fig. 1-Figure Supplement 5B, C). As defined in previous articles, there were 3 kinds of fibroblasts, including synovial lining fibroblasts (SIFs), synovial sublining fibroblasts (SSFs) 5,6, and CXCL14 + fibroblasts (CXCL14 + Fibs) 7, 2 newly defined subtypes, including Fibroblast_1 and Fibroblast_2, and 1 transitional subtype, namely synovial fibrochondrocytes (FCs) 8 4. Moreover, CD79A and CD27 were used to identify B cells, CD2 and CD8A to identify T cells, LYVE1, FOLR2, and MRC1 to identify macrophages (Macs), and CD1C, CLEC10A, and FCGR2B to identify dendritic cells (DCs) (Fig. 1-Figure Supplement 4).
2. scRNA-seq revealed new fibroblast subtypes.
As early as 2007, Athanasiou's laboratory first discovered a mixed cell group with the characteristics of fibroblasts and chondrocytes in the meniscus of the knee joint, which was named fibrochondrocytes15. In recent years, to prepare favorable meniscus repair materials, the focus of many studies has been placed on promoting the fibrocartilage differentiation of synovium-derived stem cells16,17. In this process, investigators from several laboratories isolated fibrochondrocytes from meniscus cells15. Among them, Tang et al. isolated fibrocartilage chondrocytes (FCs) from the cartilage with the aid of scRNA-seq for the first time8. Besides, Zhang et al. used a similar method to identify fibrochondrocytes (FCs) with high expression of COL3A1, COL6A1, and COL1A14. In short, these results indicated that the markers of this cell subtype included those of chondrocytes (such as COL3A1 and COL6A) 4 and synovial fibroblasts (such as PRG4) 6.
Hence, it can be maintained that this is a transitional subtype and may exist in synovium and meniscus samples. Similar to previous studies, it was found that this cell subtype can highly express the markers of chondrocytes and synovial fibroblasts, among which BGN, COL6A1, and COL3A1 were the markers of Ch.22, and HTRA1 and PRG4 were the markers of SIFs56. FCs were found in meniscus and synovium samples (Fig. 1F), with a larger number in meniscus samples. The pseudotime analysis was performed in synovium and meniscus samples, respectively. It can be observed that FCs were mainly distributed around the starting point of the trajectory in synovium samples and the trajectory's root of the pseudotime trajectory in meniscus samples (Fig. 5).
In the clustering process, two newly defined subtypes, Fibroblast_1 and Fibroblast_2, were observed. Both subtypes can express the marker (PRG4) of synovial fibroblasts, mainly in synovium samples (Fig. 1F). Fib_1 was mainly observed in the normal synovium samples and distributed in cell fate 2 (Fig. 5), while Fib_2 was mainly observed in degenerated synovium samples (Fig. 1F) and distributed in the starting point of the trajectory (Fig. 5).The TNFA SIGNALING VIA NFKB and EMT pathways of Fib_1 and Fib_2 were highly expressed in degenerated samples, while the INTERFERON GAMMA RESPONSE pathway was low expressed in degenerated samples (Fig. 3-Figure Supplement 3A). Moreover, compared with HALLMARK_TNFA_SIGNALING_VIA_NFKB in cartilage samples, the expression of Fib_2 was up-regulated in synovium samples under different states. The EMT pathway, which was up-regulated in normal synovium samples, was down-regulated in degenerated synovium samples (Fig. 3-Figure Supplement 3B). It can be speculated that the mechanism related to the progression of OA was limited to the synovium and other single tissues to a certain extent. Furthermore, EMT and other mechanisms can act on different tissues in the progression of OA at different stages.
3. The CellChat-based analysis revealed the changes in synovium/meniscus samples caused by the progression of OA.
The crosstalk was observed in normal and degenerated meniscus and synovium samples based on CellChat. Of note, the interaction between over-expressed ligands and receptors in meniscus chondrocytes and synovial fibroblasts can be observed in different tissue samples. A variety of cells were found to have extensive correlations in the internal environment of KOA. It was observed that there were several different expression patterns in normal and degenerated meniscus and synovium cells. The inflammation-related pathways critical to the progression of OA were up-regulated in degenerated samples compared with normal samples, including the IGF18, PTN19, and FGF pathways in OA and the MK pathway in RA20. Besides, the cartilage regulatory pathways (such as ANGPTL, PROS, and PDGF pathways) and vascular regulatory pathways (such as VISFATIN21, VEGF22, and SEMA323 pathways) were up-regulated in degenerated samples. On the contrary, the GALECTIN pathway24,25, CXCL pathway, TGF-b pathway26, and other pathways related to macrophage polarization were down-regulated in degenerated samples.
Based on that, it can be assumed that the meniscus and synovium may exhibit similar CellChat trends during the development of OA. The CellChat-based analysis of normal and degenerated meniscus and synovium samples may provide a new method for understanding the synergistic effects from the mechanisms related to OA progression between tissues (meniscus and synovium) and cells (chondrocytes and fibroblasts). Subsequently, a CellChat-based analysis was performed to explore the interaction between synovial fibroblasts and chondrocytes.
3.1. The crosstalk between synovium and meniscus in key pathways in OA
The IGF pathway, PTN pathway, MK pathway, and FGF pathway were highly expressed in OA samples18 19 20. In the progression of OA, 4 pathways presented extremely similar trends in meniscus and synovium samples. Significant crosstalk between synovial fibroblasts and chondrocytes can be observed in normal and degenerated samples (Fig. 2). In particular, there were more and stronger interactions between SSF (senders) and Ch.1–5 (receivers) in degenerated samples than in normal samples. This was consistent with previous findings, namely that insulin-like growth factor 1 (IGF-1) and pleiotrophin (PTN) were highly expressed in the OA cartilage27–29. Besides, fibroblast growth factor-18 (FGF-18), a key molecule in the FGF signaling pathway, was highly expressed in the OA synovium30. This may be related to the expression of MMP-13 in chondrocytes induced by bFGFs31. Of note, there is no research on the role of the MK pathway (a pathogenic pathway related to RA) in OA, and there is no report with a focus on the expression patterns of the above 4 pathways in the OA synovium. These results suggested that synovial fibroblasts (senders) were involved in the changes of multiple pathways during the degeneration of tissues.The whole Cellchat pattern in different samples was further observed. It was found that synovium samples presented more active communication patterns, and normal synovial tissues also exhibited complex communication patterns. This also verified that the occurrence of OA may be earlier than that of cartilage inflammation, suggesting the significance of early intervention in the management of synovitis (Fig. 2-Figure Supplement1B).
The heatmap (by cellChat-based analysis) revealed the interaction numbers and communication probabilities in cellular communication networks among the Normal_meniscus, Degenerated _meniscus, Normal_synovium, and Degenerated_synovium groups.
3.2 The crosstalk between synovium and meniscus in cartilage regulatory pathways
It can be observed that the ANGPTL pathway, PDGF pathway, and PROS pathway presented similar trends in the CellChat-based analysis during the progression of OA. Ch1-5, SSF, and SIF (senders) had more interaction and stronger interactions in degenerated samples. This phenomenon was more prominent in meniscus samples. This may be related to 3 pathways that can lead to disease progression by affecting the growth of articular cartilage or cartilage vessels32 33 34. From the observation of 3 pathogenic pathways in degenerated samples: SIF and SSF (senders) exhibited a larger communication probability than Ch1-5 (senders). This suggested that synovial fibroblasts may play an important role during the degeneration of tissues. In normal meniscus samples, it can only be observed that PRG4 + Ch.3 (senders) ligands acted on various cell receptors. Of note, this special chondrocyte subtype was distributed on the meniscus surface and can highly express PRG4, an HTRA1 marker gene, which was common in the synovium2. These results verified that the synovium may play a role in the progression of OA35.
3.3 The crosstalk between synovium and meniscus in neovascularization pathways
ECs (CAP-A) and ECs (CAP-V) (receivers) exhibited a larger communication probability in the VISFATIN pathway, VEGF pathway, and SEMA3 pathway, presenting similar trends in the CellChat-based analysis. Besides, there were more interactions between ECs (receivers) in the degenerated group than in the normal group in meniscus samples. These results were consistent with previous findings, namely that the VISFATIN pathway21,36, VEGF pathway22,37, and SEMA3 pathway23 promoted the progression of OA by regulating neovascularization during cartilage development. VISFATIN up-regulated the expression of vascular endothelial growth factors (VEGFs) in human OA synovial fibroblasts and promoted tube formation and migration of endothelial progenitor cells36. VEGF was highly expressed in the OA cartilage38 and involved in angiogenesis and chondrocyte metabolism37, thus promoting the expression of MMPs in chondrocytes39.
Besides, significant communication between macrophages (senders) and ECs (receivers) can be observed in synovium samples. This was consistent with previous findings, namely that VEGF secreted by macrophages can promote angiogenesis and aggravate inflammation in the OA synovium40. It was also found that ECs (receivers) in degenerated meniscus samples exhibited a larger communication probability than those in normal meniscus samples. This may be attributed to the binding of Sema3A to NRP-1, which served as an antagonist of VEGF signaling transduction in ECs23. These results were similar to those of Enomoto et al. in that the VEGF receptor neuropilin 1 [NRP-1] was up-regulated in the OA cartilage38. In addition, SEMA3A was highly expressed in chondrocytes of the OA meniscus. This may be related to the inhibitory effect of semaphorin 3A (Sema3A) on the protective migration of VEGF165 in chondrocytes reported by Okubo et al23. In this study, it was also found that the communication probability of meniscus chondrocytes (receivers) in degenerated meniscus samples was larger than that in normal meniscus samples. Moreover, the above 3 pathways presented similar trends in the interaction number and strength in degenerated meniscus, normal synovium, and degenerated synovium samples. This may be related to the angiogenesis of synovial tissues and cartilage tissues during the progression of OA41,42. Additionally, the early high expression of pathway ligand molecules in the synovium may act on meniscus tissues through the synovial fluid43. In summary, these results validated the views of other researchers44. Vascular regulatory signaling pathways may construct the communication between synovium and meniscus in the process of degeneration.
3.4 The crosstalk between synovium and meniscus in macrophage polarization pathways
The GALECTIN pathway45, CXCL pathway46, and TGF-β pathway26 (some pathways related to macrophage polarization) presented similar trends in the CellChat-based analysis. Fernandes et al. also found that M2 polarization of macrophages can promote the repair of cartilage26 and synovium44. The GALECTIN pathway24,25 of chondrocytes (receivers), a harmful upstream mediator of the cartilage, presented significant differences in the interaction of meniscus samples. This may be related to the role of GALECTIN-1 as an upstream mediator of inflammation and matrix decomposition in OA. The secretion of GALECTIN-1 was not induced by proinflammatory conditions24,25. This may be attributed to the high GALECTIN pathway activity in normal meniscus samples. GALECTIN-3 can drive M2 polarization of macrophages in mice45. This also suggested that the GALECTIN pathway was involved in macrophage-related mechanisms. In particular, the strength of interactions between the antigen-presenting cells (APCs) (senders) of macrophages/DCs and the GALECTIN pathway of DCs (receivers) in degenerated synovium and meniscus samples was weakened compared with normal samples. Moreover, the CXCL pathway between APCs (senders) and ECs (CAP-V) (receivers) decreased in degenerated samples of different tissues.
There are many experiments with a focus on the angiogenesis of the CXCL pathway47,48. The serum level of CXCL8, CXCL9, CCL2, and CXCL10 in patients with KOA increased with age49. CXCL7 can promote M2 polarization of macrophages in tumors50. M1 and M2 polarization of macrophages can secrete different CXCL molecules51. It can be assumed that the crosstalk between APCs (senders) and ECs (receivers) may be involved in the regulation of macrophage polarization through the CXCL pathway. The communication probability between macrophages (senders) and Ch2 and Ch5 (receivers) decreased in synovium samples. Changes in the anti-inflammatory and pro-chondrogenic cytokine transforming growth factor-β (TGF-β) secreted by macrophages with disease progression can also be observed. The increased secretion of TGF-β in M2 macrophages contributed to the repair of cartilage by MSCs26. The decreased secretion of TGF-β may be the common initiating factor for meniscus and synovial metamorphosis. Macrophage (senders) presented similar trends in the GALECTIN pathway, CXCL pathway, and TGF-β pathway between different samples, which was consistent with previous findings. Macrophages played an important role in the whole progression of OA. However, it remains undefined that macrophage infiltration mainly affects the early stage52 or the end stage53. In summary, these results suggested that the pathways related to macrophage polarization had similar action patterns in synovium and meniscus samples.
4. Analysis of the dynamic relationships between synovial fibroblast and meniscus chondrocyte subtypes
A total of 11 subtypes were selected to construct a new trajectory, including 6 synovial fibroblasts (SIFs, SSFs, CXCL14 + Fibs, Fib_1, Fib_2, and FCs) and 5 meniscus chondrocytes (Ch.1, Ch.2, Ch.3, Ch.4, and Ch.5). In synovium samples, the trajectory root was mainly populated by FCs, while the two primary termini of the tree were populated by Ch.2 and Fib_2 of cell fate 1 and Ch.3, Ch.5, and Fib_1 of cell fate 2. In meniscus samples, cell fate 1 was populated by FCs and Fib_2, and cell fate 2 was populated by SSFs. Several genes that changed significantly during the transformation were identified in the pseudotime differential expression analysis. Figure 5BE presents the genes with the most significant changes in the pseudotime trajectory in the heatmap. In the synovium pseudotime heatmap, cluster 2 was populated by Ch4 (EFEMP1 and CFH), Ch3 (PRELP), and Ch1 (CHI3L2 and CHI3L1), and its expression was down-regulated with the progression of OA. In cluster 4, marker genes (CD55 and PRG4) of SIFs were highly expressed with the progression of OA, and the number of SIFs in degenerated synovium samples was significantly higher compared with normal synovium samples (Fig. 1E). It can be speculated that SIFs in degenerated synovium samples played an important role in the progression of OA.
In meniscus samples, the expression of genes of clusters 1 and 2 was up-regulated, and that of cluster 4 was down-regulated with the progression of OA. Of note, the marker genes of Ch.2 (COL6A3), Ch.3 (COL3A1, FN1, and TNFAIP6), and Ch.4 (PLA2G2A, EFEMP1, and CYP1B1) in clusters 1 and 2 were highly expressed with the progression of OA; while, the marker genes of Ch.1 (FGFBP2, APOD, CYR61, C2orf40, CHI3L2, CHI3L1, and CP) in cluster 4 exhibited a reverse expression trend. Overall, Ch.1–5 chondrocytes mainly populated the pseudotime trajectory (Fig. 5C) in meniscus samples, the trajectory root in normal meniscus samples, and the termini of two cell fates in degenerated meniscus samples. With the progression of OA, the expression of marker genes of Ch.2–5 was up-regulated significantly; while the expression of marker genes of Ch.1 was down-regulated. This result verified the finding of Zhang et al., namely that Ch.1 may be a normal subtype and can maintain the homeostasis of meniscus tissues2. Except for Ch.4, few chondrocytes could populate the pseudotime trajectory of synovium samples, which may be explained that Ch.4 had the markers of SSFs such as CXCL122. The fibroblasts/chondrocyte intermediate subtype FCs populated the root of synovium samples and cell fate 1 of meniscus samples. The marker gene COL3A1 of FCs was down-regulated in cluster 2 of synovium samples and up-regulated in cluster 1 of meniscus samples. This validated the hypothesis that the synergistic effect of synovium and meniscus was the initial factor in the progression of OA. FCs may play a role in this process.
Besides, it was noticed that SSFs, SIFs, CXCL14 Fib, and Fib_2 only populated the pseudotime trajectory of degenerated synovium samples. However, Fib1 mainly populated the pseudotime trajectory of normal synovium samples, which had the same expression pattern as cluster 4 in the synovium pseudotime heatmap. As a normal synovial subtype, Fib1 mainly populated cell fate 2, which was different from SSFs, SIFs, and CXCL14 Fib which populated cell fate 1. These results suggested that cell fate 2 may be a normal cell state during meniscus degeneration. The development of cells was conservative, which played a role in inhibiting inflammatory progression. Cell fate 1 may be an abnormal cell state. The inner meniscus samples and the degenerated meniscus samples exhibited similar trends(Fig. 2—figure supplement 5).
5. Multiple mechanisms triggered the progression of OA in synovial tissues.
In many studies on the initiation factors of OA, the initiation factors inducing the early progression of OA were attributed to the synovium or cartilage, or synergistic effects from multiple tissues. To simulate the OA environment more realistically, it is required to consider the core factors of degeneration. New observation methods were provided in this study: The communication between synovial fibroblasts and meniscus chondrocytes in synovium/meniscus samples was compared through the CellChat-based analysis. The results demonstrated that EPITHELIAL MESENCHYMAL TRANSITION (EMT), TNFA SIGNALING VIA NFKB, and other pathways exhibited common expression characteristics in various cells (Ch.3, Ch.4, Ch.5, Fib_2, DCs, ECs, PCL_1, and PCL_2). These gene sets related to inflammatory injury were highly expressed in normal synovium samples than in normal meniscus samples. In contrast, the expression of these gene sets was higher in degenerated meniscus samples than in degenerated synovium samples. Synovium and meniscus samples can be considered as a whole, and hence the influence of the progression of OA on major joint tissues was analyzed as a whole. In most cells (Ch.1, Ch.3, Ch.4, Ch.5, Fib_1, Fib_2, FCs, ECs, mast cells, T cells, and DCs), the EMT and TNFA pathways were highly expressed in degenerated samples compared with normal samples (Figs. 3A and B, Fig. 3-Figure Supplement 1A). Based on that, it was speculated that there was an inflammatory mechanism in the synovial and meniscus tissues that transferred from synovial tissues to meniscus ones, which could affect the degeneration of various tissues regarding bones and joints. Meanwhile, it was noticed that similar phenomena widely appeared in hypoxia and apoptosis (Fig. 3-Figure Supplement 2).
SIFs constituted an important cell subtype in the degenerated synovium. Different from the initial effects of degeneration of synovial tissues for most cells, SIFs had low expression in TNF, hypoxia, and other injury mechanisms in degenerated samples. Besides, these cells may participate in the mechanism initially inducing cartilage injury (Fig. 4A). In meniscus tissues, SIFs were distributed at the cell fate termini of pseudotime trajectory (Fig. 5C and Fig. 4A). In addition, Ch.2 and CXCL14 Fib had low expression of EMT, APOPTOSIS, TNFA, YPOXIA, and other inflammatory injury functions in degenerated samples; while they had high expression of potential anti-inflammatory mechanisms such as MTORC (Fig. 4B, Fig. 3-Figure Supplement 3C). It can be speculated that SIFs, Ch.2, and CXCL14 + Fib may play a protective role in meniscus and synovium samples from OA patients, which was consistent with previous studies. Ch.2 (FNDC1) is a chondrocyte population associated with abnormal ECM degradation and remodeling. This chondrocyte population may contribute to the progression of meniscus degeneration in humans.
Finally, the newly identified synovial fibroblast subtypes Fib_1 and Fib_2 were analyzed, respectively. In degenerated samples, both of them had high expression of the TNFA and EMT pathways and low expression of HALLMARK INTERFERON GAMMA RESPONSE (Fig. 3-Figure Supplement 3A). Compared with meniscus samples, the TNFA pathway in Fib_2 was highly expressed in normal and degenerated synovium samples (Fig. 3-Figure Supplement 3B). This may be explained that Fib_2 was the main fibroblast subtype in degenerated synovium samples and Fib_1 was the main fibroblast subtype in normal synovium samples (Fig. 5C).