Senescent Synovial Intimal Fibroblasts Aggravate Osteoarthritis by Regulating Macrophage Polarization and Chondrocyte Phenotype Through ANGPTL4‐α5β1 Axis
Muhai Deng, Yunsheng Jiang, Zhiyu Chen, Kaiquan Chen, Na Cao, Yang Huang, Zhengxue Quan, Cheng Chen
TL;DR
Senescent synovial fibroblasts worsen osteoarthritis by influencing macrophages and cartilage through a specific signaling pathway.
Contribution
Identifies senescent synovial intimal fibroblasts as key OA drivers via the ANGPTL4–α5β1 axis regulated by EGR1 and ATF3.
Findings
Senescent synovial intimal fibroblasts promote M1 macrophage polarization.
ANGPTL4–α5β1 signaling facilitates cartilage degeneration and OA progression.
Pharmacological inhibition of this pathway reduces disease severity.
Abstract
The incidence of osteoarthritis (OA) is strongly correlated with aging. It has been shown that the accumulation of senescent cells in the synovium precedes chondrocyte senescence and cartilage degradation, suggesting that synovial cell senescence plays a key role in OA pathogenesis. This study aimed to investigate the mechanisms underlying synovial cell senescence and its influence on intercellular communication within the joint. Using multiplex immunofluorescence, gene regulatory network reconstruction, and single‐cell RNA sequencing analyses, we identified senescent cells and characterized the senescence‐associated secretory phenotype in the synovium. A series of in vivo and in vitro functional experiments is conducted to elucidate the mechanisms of fibroblast senescence and its effects on macrophages and chondrocytes. We found that synovial intimal fibroblasts (SIF) display more…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5
FIGURE 6
FIGURE 7
FIGURE 8
FIGURE 9- —Postdoctoral Research Project of Chongqing
- —Natural Science Foundation of Chongqing10.13039/501100005230
- —Future Medical Innovation Team of Chongqing Medical University
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsOsteoarthritis Treatment and Mechanisms · Lipid metabolism and disorders · Tendon Structure and Treatment
Introduction
1
Osteoarthritis (OA) is the most common joint disease, with radiographic evidence found in 80% of adults over the age of 65 [1]. Although aging is strongly associated with OA, it is not a direct cause; instead, age‐related changes at the cellular, tissue, and organ levels contribute to disease progression [2]. A key mechanism underlying these changes is cellular senescence, which has emerged as a promising therapeutic target for modifying OA development.
Cellular senescence is a state characterized by permanent cell cycle arrest, apoptosis resistance, and the sustained secretion of factors known as senescence‐associated secretory phenotype (SASP) [3]. Previous research on senescence in OA has focused largely on chondrocytes within cartilage [4], consistent with the historical view of OA as a “wear and tear” disorder of articular cartilage. Indeed, the accumulation of senescent chondrocytes correlates positively with OA progression [5, 6]. OA chondrocytes exhibit elevated expression of p16^INK4a^ and increased activity of senescence‐associated β‐galactosidase (SA‐β‐Gal), similar to other senescent cell types [7]. However, growing evidence indicates that senescent cells are also present in the synovium and subchondral bone [8, 9]. Notably, in the synovium, senescent cells increased markedly within two weeks after destabilization of the medial meniscus (DMM) surgery, preceding both chondrocyte senescence and cartilage degradation [10]. Fibroblast‐like synoviocytes (FLS), the predominant cell type in the synovium, show p16^INK4a^ positivity and secrete SASP factors such as IL‐6, CXCL8, and MMP3 in OA [11]. These senescent FLS have been shown to promote catabolic activity of chondrocytes both in vivo and in vitro [10]. Histologically, FLS reside mainly in the synovial lining and sublining layers, categorized as synovial intimal fibroblasts (SIF) and synovial subintimal fibroblasts (SSF), respectively [12]. Although seven functionally distinct FLS subpopulations have been identified [13], none have been specifically linked to senescence. Thus, identifying a senescence‐specific FLS subpopulation and clarifying its role in OA at single‐cell resolution remains critical.
Dysregulated cell–cell communication within the joint is a key driver of OA and a hallmark of cellular senescence [14]. Macrophages, the most abundant immune cells in the synovium, can polarize into pro‐inflammatory M1 phenotype or anti‐inflammatory M2 phenotype in response to local signals [15]. The M1/M2 ratio and activation state are closely associated with OA severity [16]. Macrophage polarization is influenced by other cells, particularly fibroblasts [12]. For example, exosomes derived from inflammatory FLS can promote M1 polarization [17]. In post‐traumatic OA synovium, Proteoglycan 4 (PRG4) positive SIF uniquely secrete R‐spondin 2, which influences both M1 macrophage polarization and chondrocyte hypertrophy [13]. However, the specific effects of senescent FLS on macrophages and chondrocytes in OA remain poorly understood.
In this study, we hypothesize that senescent SIF promote M1 macrophage polarization and OA‐like changes in chondrocytes via SASP secretion. First, we identified and characterized senescent cells in OA synovium by integrating public single‐cell RNA sequencing (scRNA‐seq) data with multiplex immunofluorescence staining. Second, we demonstrated that the transcription factor (TF) Early Growth Response 1 (EGR1) mediates FLS senescence in OA and subsequently regulates SASP secretion via the Toll‐like receptor (TLR) signaling pathway. Finally, we revealed that senescent SIF specifically secrete Angiopoietin‐like 4 (ANGPTL4), which modulates both macrophage polarization and chondrocyte behavior.
Results
2
scRNA‐seq Reveals Altered Cellular States of SIF and SSF in OA
2.1
To explore pathological alterations in the OA synovium and identify key cell clusters, we analyzed the public scRNA‐seq dataset GSE216651 [ 18 ], which includes three OA samples and three control samples. After quality control and batch effect correction, 55 474 cells were retained and grouped into 18 clusters, comprising 29 901 cells from controls and 25 573 from OA samples (Figure S1A–C). Using marker gene expression and t‐SNE visualization, we identified 12 cell types: SIF, SSF, endothelial cells (EC), mural cells, resident macrophages (Resid. Mφ), T cells, inflammatory macrophages (Inflamm. Mφ), dendritic cells (DC), granulocytes, monocytes, B cells, and mast cells (Figure 1A,B; Figure S1D). Differentially expressed gene (DEG) analysis for each cell type before and after disease onset further revealed the impact of OA on synovial cells (Figure 1C). Notably, FLS, including both SIF and SSF, represented the most abundant cell population, comprising the greatest cell numbers (Figure 1D) and the highest proportion (Figure 1E).
*Pronounced cellular state alterations of FLS in OA synovium. A) t‐SNE plot of synovial cells from control and OA groups. B) Top three marker genes for each cell type. C) DEGs between OA and control groups across cell types. D) Total cell numbers per cell type. E) Proportional distribution of synovial cell types. F) UMAP visualization based on RNA expression (left) and regulon activity score (RAS) matrix (right), reflecting cell type distribution and transcriptional activity states, respectively. G) Representative images of H&E and Safranin‐O/Fast Green staining of joint, with OA histopathology scoring (n = 5x3; ***p < 0.001; Mann–Whitney test; Data are shown as mean ± SD). H) H&E staining of synovial tissue with synovitis scoring (n = 5x3; **p < 0.001; Mann–Whitney test; Data are shown as mean ± SD). Red arrows indicate inflammatory cell infiltration and fibroblast proliferation. I) UMAP plots of SIF marker PRG4. J) Immunofluorescence staining of the PRG4. K) UMAP plots of SSF marker CXCL12. L) Immunofluorescence staining of the CXCL12. White arrows indicate positive cells. S: synovium, and the dashed lines outline synovial boundaries.
Given that cellular identity is largely defined by underlying gene regulatory networks (GRNs) [19], reconstructing these networks can provide critical insights into how cell states change following OA onset. We therefore applied the pySCENIC [19] algorithm to construct GRNs in the synovium for the first time, identifying 369 key regulons and 49,402 target genes (Table S1). In contrast to conventional UMAP visualizations based on RNA expression, which reflect cell type distribution, UMAP projection derived from regulon activity scores (RAS) revealed distinct transcriptional activity states across cells (Figure 1F). Notably, FLS, including both SSF and SIF, exhibited the most substantial alterations in cellular state between OA and control conditions. To validate these findings, we constructed a mouse model of OA via DMM surgery. Histological evaluation using Hematoxylin and Eosin (H&E) and Safranin‐O/Fast Green staining showed clear degradation of the cartilage matrix in the OA group compared to controls. This was further supported by a significantly higher OA histopathology score in OA group (Figure 1G). Additionally, synovial tissues from OA mice displayed inflammatory cell infiltration, fibroblast proliferation, and an elevated synovitis score (Figure 1H). We then examined the expression of established marker genes: PRG4 for SIF and CXCL12 for SSF [12, 18]. UMAP visualization confirmed the specific localization of PRG4 and CXCL12 expression to SIF and SSF clusters, respectively. Immunofluorescence staining further demonstrated that protein levels of both PRG4 and CXCL12 were markedly increased in the OA group (Figures 1I–L), reinforcing the involvement of these fibroblast subpopulations in OA pathogenesis.
OA Progression Is Closely Associated With Senescence of SIF
2.2
We next investigated changes in the cellular states of SIF and SSF in OA. In SIF, the most significantly upregulated differentially expressed genes (DEGs) in the OA group were associated with inflammation and cartilage damage, including MMP3, PTGDS, and CXCL1 (Figure 2A). KEGG pathway analysis revealed that DEGs in SIF were enriched in pathways related to cell growth, such as the Ras signaling pathway and cellular senescence (Figure 2B). Gene Set Enrichment Analysis (GSEA) using Hallmark gene sets indicated activation of the P53 pathway, a critical regulator of cellular senescence (Figure 2C; Figure S2A). To identify TFs potentially regulating P53 pathway activity, we performed reverse enrichment analysis. This revealed that EGR1 and JUNB exhibited the highest enrichment scores among associated TFs (Figure 2D). Subsequent reconstruction of the TF‐target network confirmed that EGR1 and JUNB regulate multiple genes within the P53 pathway, including CDKN1A and KLF4 (Figure S2B), supporting their functional involvement.
Cellular senescence in OA synovium. A) DEGs in SIF. B) KEGG enrichment of SIF. C) GSEA‐Hallmark analysis of SIF. D) Reverse enrichment of SIF regulons in the Hallmark P53 pathway. E) DEGs in SSF. F) KEGG enrichment of SSF. G) GSEA‐Hallmark analysis of SSF. H) Reverse enrichment of SSF regulons in the Hallmark TNF‐α signaling via NF‐κB pathway. I) Heatmap showing the scores of senescence gene sets across different cell types. J) Violin plots showing the scores of senescence gene sets across different cell types. K) Immunofluorescence staining of PRG4 (purple) and p21Cip1 (yellow). L) Immunofluorescence staining of CXCL12 (purple) and p21Cip1 (yellow). White arrows indicate positive cells. S: synovium, and the dashed lines outline synovial boundaries.
In SSF, we observed an elevated expression of fibrosis‐related genes such as COL3A1, COL1A2, and COL1A1 (Figure 2E). Although cellular senescence also emerged as an enriched term in KEGG analysis for SSF (Figure 2F), GSEA highlighted significant activation of the TNF‐α signaling pathway via NF‐κB (Figure 2G; Figure S2C). Reverse pathway enrichment analysis identified KLF10, FOSL2, and REL as key regulons significantly enriched in this pathway, each regulating distinct target genes within the NF‐κB signaling network (Figure 2H; Figure S2D).
However, cellular senescence represents a complex stress response that cannot be comprehensively captured by a single enrichment analysis [20]. To address this, we integrated six established senescence‐related gene sets [21, 22, 23, 24, 25] and applied Gene Set Variation Analysis (GSVA) to quantify senescence activity. The results revealed elevated senescence scores in SIF, Inflamm.Mφ, and SSF (Figure 2I). Among these, SIF exhibited the most pronounced increase in senescence scores in OA compared to controls, significantly surpassing the changes observed in Inflamm.Mφ and SSF (Figure 2J). Consistent with this, enrichment analyses for DNA damage and oxidative stress‐related Gene Ontology (GO) terms also showed elevated activity in SIF under OA conditions (Figure S2E, F). To visually confirm the presence of senescent cells within synovial tissue, we performed immunofluorescence staining for the senescence marker p21^Cip1^ alongside cell‐type‐specific markers. In SIF, strong co‐localization of PRG4 and p21^Cip1^ was observed in the OA group (Figure 2K). In contrast, co‐localization of CXCL12 and p21^Cip1^ in SSF was notably weaker, indicating a less distinct senescence phenotype in this fibroblast subpopulation (Figure 2L).
Identification of Senescent Subpopulation Within SIF in OA
2.3
Cellular senescence is a dynamic and highly heterogeneous process at the single‐cell level [26]. To precisely identify senescent cells within SIF, we performed a systematic subcluster analysis. SIF from OA samples was subdivided into five distinct subpopulations (Figure 3A). Expression dot plots of the top five highly expressed genes in each cluster revealed that L1_SIF exhibited a strong SASP, characterized by high expression of genes such as CXCL1 and NFKBIA (Figure 3B). Furthermore, based on the consensus senescence marker genes proposed by the SenNet Biomarker Working Group [27], we found that the senescence marker genes were predominantly highly expressed in L1_SIF (Figure 3C; Figure S3A,B). The L1_SIF subpopulation also showed relatively higher GSVA scores for established senescence‐related gene sets (Figure S3C). To reconstruct the developmental trajectory of SIF during senescence, we used the Monocle2 algorithm. The resulting pseudotime trajectory accurately captured the state transitions among SIF subpopulations (Figure 3D). Notably, L1_SIF was positioned at the endpoint of the trajectory, whereas L3_SIF was located near the start. L2_SIF and L4_SIF were situated immediately preceding L1_SIF, and L5_SIF was distributed intermediately along the path (Figures 3D, E). Expression levels of key senescence marker genes increased progressively along the pseudotime axis (Figure 3F and Figure S3D). We next performed clustering analysis of genes along the senescence trajectory, identifying two major expression patterns: genes in Cluster 1 increased with pseudotime, and those in Cluster 2 decreased (Figure 3G). Functional enrichment analysis showed that Cluster 1 was significantly associated with DNA damage response and cell cycle regulation (e.g., response to gamma radiation, negative regulation of mitotic cell cycle). In contrast, Cluster 2 was enriched for metabolic processes such as cortisol biosynthesis and tertiary alcohol biosynthesis (Figure S3E).
Identification of senescent subpopulation within SIF in OA. A) UMAP plot of SIF subpopulations in OA. B) Dot plot of the top five genes expressed in each SIF subpopulation. C) The expression density plots of senescence marker genes. D) Monocle pseudotime trajectory showing the progression of senescence in SIF colored by pseudotime (upper) or SIF subpopulations (lower). E) Density plots of SIF subpopulations along the Monocle pseudotime trajectory. F) Expression of senescence markers along the Monocle pseudotime trajectory. Each point represents a cell, colored by subpopulations, with the solid line representing the loess regression. G) Hierarchical clustering‐based dynamic expression heatmap of genes across SIF subpopulations along the senescence trajectory. H) UMAP visualization of senescence signature and scoreK from SIT algorithm evaluation metrics. I) UMAP plot of the SIT algorithm. (YES: senescent; NO: non‐senescent) J) Proportion of each SIF subpopulation identified as senescent by the SIT algorithm. K) DEGs in L1_SIF. L) KEGG analysis revealing the different enrichment patterns of upregulated and downregulated genes in L1_SIF.
Given that cell cycle arrest is a hallmark of senescence, we further employed the Senescence Index Tool (SIT) [28], an algorithm incorporating cell cycle gene sets, to validate senescent cell identification. SIT evaluation metrics (Senescence_signature, scoreK, scoreR, and scoreWP) were consistently elevated in L1_SIF (Figure 3H; Figure S3F). Consistent with this, the majority of cells classified as senescent by SIT were located within L1_SIF (Figure 3I), accounting for 59.87% of all senescent cells identified (Figure 3J). KEGG pathway analysis indicated that upregulated genes in L1_SIF were enriched in cellular senescence and TNF signaling pathways, while downregulated genes were linked to nucleotide excision repair and ATP‐dependent chromatin remodeling (Figures 3K,L). Together, these multifaceted results robustly demonstrate that L1_SIF represents the predominant senescent subpopulation within SIF in the OA synovium.
Activating Transcription Factor 3 (ATF3) and EGR1 Regulate Senescence in SIF
2.4
TFs are pivotal in defining cell identity and modulating cellular functions during aging [29]. To elucidate the regulatory mechanisms underlying SIF senescence in OA, we reconstructed GRNs at the subpopulation level, identifying 341 key regulons targeting 13,390 genes (Table S2). Using AUCell based on these GRNs, we assessed TF activities and employed Jensen–Shannon divergence combined with regulon activity scores (RAS) to identify subpopulation‐specific regulons. This analysis revealed that NFIL3, JUNB, and ATF3 exhibited higher TF activity specificity in L1_SIF (Figure 4A). A heatmap of the top three regulons for each subpopulation further highlighted these associations (Figure 4B).
*Screening and validation of key TFs in SIF senescence. A) L1_SIF specificity regulons screening. Regulons in L1_SIF were ranked based on RSS (left). L1_SIF location in UMAP plot (red dots, middle). RAS of ATF3 in UMAP plot (green dots, right). B) Heatmap of the top three RSS regulons for each SIF subpopulation. C) GSEA result of KEGG cellular senescence pathway in L1_SIF. D) Reverse enrichment results of regulons in L1_SIF for the cellular senescence pathway. E) TF‐target network diagram of the cellular senescence pathway in L1_SIF. F) Violin plots of TF activities of EGR1 and ATF3. G) SA‐β‐gal staining of FLS at 0, 24, 48, and 72 h following a 24 h H2O2 stimulation (n = 5). H,I) Immunofluorescence images and quantitative analysis (n = 5) of the senescence markers γH2AX (H) and p16INK4a (I) in FLS with or without H2O2 stimulation. J) Western blot and quantification of EGR1 and ATF3 in nuclei of FLS (n = 3). K) Immunofluorescence images of p21Cip1 (red) and EGR1 (green) in FLS with or without H2O2 stimulation. L) Fluorescence intensities of EGR1 and p21Cip1 in the nuclei of FLS. A total of 60–70 FLS from three independent biological replicates were analyzed. M) Quantitative assessment of the percentage of EGR1/p21Cip1 double‐positive FLS in control and H2O2‐treated groups (n = 3). N) Immunofluorescence images of p21Cip1 (red) and ATF3 (green) in FLS with or without H2O2 stimulation. O) Fluorescence intensities of ATF3 and p21Cip1 in the nuclei of FLS. A total of 60–70 cells from three biological replicates were analyzed. P) Quantitative assessment of the percentage of ATF3/ p21Cip1 double‐positive FLS in control and H2O2‐treated groups (n = 3). For comparison among multiple groups, one‐way ANOVA was used; for comparison between two groups, Student's t‐test was used. Data are shown as mean ± SD; ns, no significance; *p < 0.05, **p < 0.01, and **p < 0.001.
We systematically examined the combinatorial regulation of TFs and their target genes using connection specificity index analysis, grouping the 341 regulons into six distinct modules (Figure S4A). Among these, modules M3 and M6 showed specific activity within L1_SIF (Figure S4B). GSEA‐KEGG pathway analysis confirmed significant enrichment of cellular senescence and TNF signaling pathways in L1_SIF (Figure 4C; Figure S4C). Notably, the TF activities of EGR1 and KLF4 were markedly enriched in the cellular senescence pathway, implicating their critical roles in synovial aging (Figure 4D). Further regulatory network analysis indicated that while both EGR1 and KLF4 contribute to senescence pathway regulation, EGR1 also engages in mutual regulatory interactions with ATF3—a key regulon in L1_SIF (Figure 4E). Violin plots confirmed elevated TF activities for both EGR1 and ATF3 in L1_SIF (Figure 4F). Pseudotime trajectory analysis from Monocle2, combined with RNA expression data, demonstrated that expression levels of EGR1 and ATF3 increased progressively along the senescence trajectory and were predominantly enriched in L1_SIF (Figure S4D, E).
Then we used an H_2_O_2_‐induced senescence model of mouse FLS to experimentally validate these findings. To confirm that senescence was primarily induced by oxidative stress rather than a general inflammatory response in OA synovial aging [30], we compared H_2_O_2_ with IL‐1β. Treatment with 200 µm H_2_O_2_ for 24 h resulted in sustained elevation of SA‐β‐gal activity at 24, 48, and 72 h compared to IL‐1β (Figure S5), with no significant differences among these time points (Figure 4G). Based on these kinetics, the 24 h time point was selected for subsequent experiments to maximize efficiency. In H_2_O_2_‐stimulated FLS, fluorescence intensities of the senescence markers γH2AX (a marker of DNA damage) and p16^INK4a^ (a key mediator of stable cell cycle arrest) were significantly elevated (Figures 4H,I). Western blot analysis further confirmed the increased nuclear expression of ATF3 and EGR1 in these senescent cells, consistent with transcriptomic findings (Figure 4J). Moreover, co‐staining of p21^Cip1^ (a marker of stress‐induced senescence) and EGR1 (Figure 4K) showed an increased intensity of both proteins (Figure 4L) and an enhanced co‐localization of EGR1 with p21^Cip1^(Figure 4M) in H_2_O_2_‐treated FLS compared to controls. A similar pattern of increased intensity and co‐localization was observed between p21^Cip1^ and ATF3 (Figures 4N–P). Collectively, these results suggest that EGR1 and ATF3 form a key regulatory node linking transcriptional regulation, subpopulation identity, and cellular senescence in OA FLS.
EGR1 Inhibition Attenuates SASP Secretion in FLS via the TLR Signaling Pathway
2.5
Given the identification of EGR1 as a key TF mediating SIF senescence, we further investigated its functional role in cellular senescence in vitro. Capitalizing on the DNA‐binding property of EGR1 [31], we employed EGR‐1‐IN‐1, an inhibitor that selectively targets its DNA‐binding domain and disrupts transcriptional activity (Figure 5A) [32]. Bulk RNA sequencing was performed on control, H_2_O_2_‐induced senescent (SEN), and EGR1‐inhibited (EGR1‐inhibit) FLS. Compared to controls, the SEN group exhibited 322 DEGs, with 201 downregulated and 121 upregulated (Figure 5B). Inhibition of EGR1 in senescent FLS resulted in 305 DEGs relative to the SEN group, including 220 downregulated and 85 upregulated genes (Figure 5C). KEGG enrichment analysis indicated that DEGs between SEN and control groups were predominantly involved in SASP‐related pathways, such as cytokine–cytokine receptor interaction and IL‐17 signaling (Figure 5D). GSEA further confirmed that H_2_O_2_ stimulation activated both the ROS and P53 pathways, supporting a model wherein ROS promotes FLS senescence and SASP secretion via p53 signaling (Figure 5E).
*Transcriptomic profiling reveals EGR1 modulates senescence and SASP via TLR signaling. A) Experimental scheme: FLS were stimulated with 200 µm H2O2 for 24 h and incubated for another 24 h to establish senescence (SEN), or co‐treated with 10 µm EGR‐1‐IN‐1 and H2O2 for 24 h (EGR1‐inhibit). B) Volcano plot of DEGs between SEN and control groups. C) Volcano plot of DEGs between EGR1‐inhibit group and SEN groups. D) KEGG pathway enrichment of DEGs from SEN vs control comparison. E) GSEA enrichment plots for the ROS and p53 pathways. F) Functional terms enriched among overlapping DEGs from both comparative sets. G) GSEA plot showing suppression of the Toll‐like receptor signaling pathway upon EGR1 inhibition, with a heatmap of key gene expression (Cxcl9, Cxcl10, Irf7, Il6, Tlr2). H) Western blot and quantification of CXCL9, CXCL10, IRF7, IL‐6, and TLR2 protein levels across groups. (n = 3; *p < 0.05, *p < 0.01; one‐way ANOVA; Data are shown as mean ± SD).
Intersection of the DEGs from both comparisons yielded 29 common genes (Figure S6A). Functional annotation revealed that these shared genes were primarily associated with SASP‐related biological processes, including positive regulation of TNF production, TNF superfamily cytokine production, and cytokine–cytokine receptor interaction (Figure 5F). Although EGR1 inhibition did not exhibit obvious regulatory trends in the ROS pathway or the P53 pathway (Figure S6B), these findings suggest its potential role in modulating SASP secretion.
GSEA revealed that EGR1 inhibition led to significant suppression of the TLR signaling pathway in senescent FLS. Consistent with this finding, a heatmap displayed five core genes within this pathway that were downregulated following EGR1 inhibition (Figure 5G). Western blot analysis further confirmed that protein expression of these candidate genes was correspondingly reduced upon EGR1 inhibition, with the exception of CXCL9, which showed no significant change (Figure 5H). Together, these results demonstrate that targeting EGR1 alleviates SASP secretion in senescent FLS, likely through repression of the TLR signaling pathway.
Senescent FLS Promote Macrophage M1 Polarization via ANGPTL and MIF Signaling Pathways
2.6
We next utilized CellChat to analyze cell–cell communication networks. Both the number and interaction strength of ligand–receptor pairs were significantly elevated in OA synovium (Figure S7A), with SIF exhibiting the most pronounced increase (Figure S7B). Given the central role of SASP in OA progression and our prior finding that SIF show the most marked senescence score increase (Figures 2K, L), we focused on communication pathways originating from SIF (Figure S7C). Interaction strengths from SIF to Inflamm.Mφ, Resid.Mφ, and monocytes were all enhanced in OA, with the strongest signals directed toward Inflamm.Mφ.
We therefore investigated communication between SIF subpopulations and macrophage subtypes in OA. A complex interaction network was observed between these cell types (Figure 6A; Figure S7D). Among all signaling pathways, ANGPTL and MIF signaling were strongly implicated in the crosstalk between L1_SIF and Inflamm.Mφ, where L1_SIF primarily functioned as signal senders and Inflamm.Mφ as receivers (Figure 6B and Figure S7E). Density UMAP plots confirmed that MIF and ANGPTL4 expression was specifically enriched in L1_SIF (Figure 6C).
*Senescent SIF promote macrophage M1 polarization through ANGPTL and MIF signaling. A) Circle plot of cell–cell communications weights between SIF and macrophage subpopulations. B) Heatmaps of ANGPTL and MIF signaling pathways between SIF and macrophage subpopulations. C) Density plots of ANGPTL4 and MIF RNA expression. D) Experimental design diagram of indirect co‐culture of FLS with RAW264.7 using a transwell system (0.4 µm membrane). FLS were stimulated with 200 µm H2O2 for 24 h and incubated for another 24 h to establish senescence. E–H) Immunofluorescence images of CD86 (E), CD206 (F), IL‐6 (G), and IL‐10 (H) in RAW264.7 after co‐culture (n = 5). I) Western blot analysis of CD86, CD206, IL‐6, and IL‐10 in RAW264.7 under the same co‐culture conditions (n = 3). J) Confocal images of PRG4 (green) and ANGPTL4 (red) in FLS, and α5β1 (red) in macrophages (n = 3). K) Confocal images of PRG4 (green) and MIF (red) in FLS, and CXCR4 (orange) in macrophages (n = 3). L) Multiple immunofluorescence staining of synovial tissue from control and OA mice for PRG4 (green), ANGPTL4 (purple), F4/80 (yellow), and α5β1 (red). Dashed lines outline synovium; boxed area highlights co‐localization. M) Multiple immunofluorescence staining of synovial tissue from control and OA mice for PRG4 (green), MIF (red), F4/80 (yellow), and CXCR4 (purple). Dashed lines outline synovium; boxed area highlights co‐localization. Student's t‐test was used to determine the statistical significance of differences between two groups. Data are shown as mean ± SD; **p < 0.01; **p < 0.001.
To functionally validate these interactions, we established a co‐culture system to assess the effect of senescent FLS on macrophages. A transwell migration assay showed that H_2_O_2_‐induced senescent FLS attracted significantly more RAW264.7 macrophages than control FLS (Figure S7F). Using a separate transwell setup to study polarization (Figure 6D), we found that co‐culture with senescent FLS markedly increased protein levels of the M1 markers CD86 and IL‐6 in RAW264.7 cells, while decreasing expression of the M2 markers CD206 and IL‐10, as shown by immunofluorescence (Figures 6E–H) and western blot analysis (Figure 6I; Figure S7G). Further immunofluorescence confocal imaging revealed increased expression of ANGPTL4 in senescent FLS and its receptor α5β1 in macrophages (Figure 6J; Figure S7H). Similarly, expression of MIF and its receptor CXCR4 was elevated in the senescence group (Figure 6K; Figure S7I). These findings were corroborated in a mouse OA model, where synovial tissues from OA mice showed increased co‐localization of ANGPTL4^+^ SIF and α5β1^+^ macrophages (Figure 6L), as well as MIF^+^ SIF and CXCR4^+^ macrophages, with closer cellular proximity compared to controls (Figure 6M). These results indicate that senescent L1_SIF promotes M1 macrophage polarization through ANGPTL4–α5β1 and MIF–CXCR4 signaling axes.
Senescent SIF Aggravate OA Progression via Paracrine ANGPTL4 Signaling
2.7
Given that cartilage degeneration is a central pathological hallmark of OA [33], and that senescent FLS secrete SASP factors capable of inducing OA‐like phenotypes in chondrocytes, we investigated whether senescent SIF influence chondrocyte behavior in OA. We integrated a public human articular cartilage scRNA‐seq dataset (GSE169454) [34] to analyze communication between SIF and chondrocytes. CellChat predicted a complex interaction network between SIF subpopulations and chondrocytes (Figure 7A; Figure S8A). Among all pathways analyzed (Figure S8B), the ANGPTL signaling pathway was strongly implicated in communication from L1_SIF to chondrocytes, with L1_SIF as the primary sender and chondrocytes as the main receiver (Figures 7B, C). The interaction strength between L1_SIF and chondrocytes exceeded that of other SIF subpopulations (Figure 7D). Based on these findings, we selected ANGPTL4 for experimental validation.
*Senescent SIF induces chondrocyte degeneration via ANGPTL4–α5β1 signaling. A) Circle plot of cell–cell communications weights between SIF and chondrocytes. B) Heatmap of ANGPTL signaling pathway between SIF subpopulations and chondrocytes. C) Scatter plot of outgoing and incoming interaction strengths of ANGPTL signaling. D) Bubble plot of ligand–receptor pairs in ANGPTL signaling between SIF and chondrocytes. E) Experimental design diagram of indirect co‐culture of FLS with ATDC5 using a transwell system (0.4 µm membrane). FLS were stimulated with 200 µm H2O2 for 24 h and incubated for another 24 h to establish senescence. F–I) Immunofluorescence images of COL X (F), COL I (G), ACAN (H), and COL II (I) in ATDC5 (n = 5). J) Western blot analysis of COL X, COL I, ACAN, and COL II in ATDC5 (n = 3). K) Confocal images of PRG4 (green) and ANGPTL4 (red) in FLS, and F4/80 (green) and α5β1 (red) in chondrocytes (n = 3). L) Multiple immunofluorescence staining of cartilage tissue from control and OA mice for COL II (green) and α5β1 (red). Arrows indicate positive cells; boxed area highlights co‐localization. Student's t‐test was used to determine the statistical significance of differences between two groups. Data are shown as mean ± SD; *p < 0.05, **p < 0.001.
Using a transwell co‐culture system (Figure 7E), we found that chondrocytes co‐cultured with senescent FLS exhibited increased protein levels of calcification and fibrosis markers (COL X and COL I) and decreased expression of anabolic markers (ACAN and COL II), as shown by immunofluorescence (Figures 7F–I) and western blot (Figure 7J; Figure S8C). These results indicate that senescent FLS promote a catabolic and fibrotic phenotype in chondrocytes. To determine whether this effect is mediated specifically through ANGPTL4 signaling, we performed immunofluorescence staining for ligand–receptor pairs. Senescent FLS showed higher ANGPTL4 expression, and co‐cultured chondrocytes exhibited increased levels of its receptor α5β1 (Figure 7K; Figure S8D). However, given the physical constraints between cartilage and synovium, along with our in vivo results showing that SIF secrete ANGPTL4, we focused on the expression of the α5β1 receptor in cartilage. In OA group, the expression of COL II was decreased, while the expression of α5β1 was significantly increased in the superficial zone of cartilage. (Figure 7L). Together, these results demonstrate that senescent SIF exacerbate OA progression by promoting chondrocyte degeneration through the ANGPTL4–α5β1 paracrine signaling axis.
Inhibition of α5β1 With ATN‐161 Alleviates Cartilage Damage and Synovitis Induced by Senescent SIF
2.8
To further evaluate the role of the ANGPTL4–α5β1 axis in OA cartilage and synovium, we intra‐articularly injected the α5β1 inhibitor ATN‐161 into OA mice (Figure 8A). This approach was chosen as ATN‐161 has proven efficacy in preclinical models of fibrosis and connective tissue remodeling, which is highly relevant to the pathological processes in OA synovium [35]. We assessed structural and compositional changes in knee joint sections using Safranin‐O/Fast Green staining, H&E, and Masson's trichrome staining (Figures 8B,C; Figure S9A–C). Both the control and control+ATN groups exhibited well‐organized chondrocyte arrangement, uniform matrix thickness, and smooth cartilage surface, indicating the safety of ATN‐161. In contrast, the OA group showed pronounced cartilage fibrosis, matrix degradation, inflammatory cell infiltration, and synovial hyperplasia over time. These degenerative changes were partially ameliorated in the ATN‐161 treatment group, with reduced OA histopathology score and synovitis score (Figure 8D). Furthermore, we observed a decrease in Collagen I‐positive regions (indicating reduced fibrotic degradation), and an increase in Collagen II‐positive areas (suggesting enhanced cartilage synthesis) in ATN‐161‐treated OA mice (Figures 8E,F).
*Inhibition of α5β1 with ATN‐161 alleviates cartilage damage and synovitis. A) Experimental scheme: intra‐articular injection of ATN‐161. B) Safranin O/Fast green staining of the mouse joint sections. C) H&E staining of the mouse synovium. D) OA histopathology scoring and synovitis scoring among groups at 4 weeks and 8 weeks after DMM surgery. (n = 5x3; *p < 0.05, **p < 0.001, **p < 0.001; Kruskal–Wallis test, followed by the Benjamini, Krieger, and Yekutieli post‐hoc test; Data are shown as mean ± SD). E, F) Immunohistochemical staining of COL I and COL II in the mouse joint sections. G) Multiple immunofluorescence staining of cartilage tissue for COL II (red) and α5β1 (green). H) Multiple immunofluorescence staining of synovial tissue for F4/80 (red) and α5β1 (green).
To further investigate the effect of α5β1 inhibition on cellular crosstalk in OA joints, we performed multiplex immunofluorescence staining. Interestingly, compared to the OA group, ATN‐161 treatment significantly downregulated α5β1 expression in both macrophages and chondrocytes (Figures 8G,H). Notably, the administration of ATN‐161 was associated with a concomitant reduction in the expression of its potential ligand, ANGPTL4, in the synovium (Figure S9D). These results indicate that the therapeutic benefits of ATN‐161 are associated with a coordinated downregulation of both the α5β1 integrin and its potential ligand, ANGPTL4, within the joint.
Finally, to explore the potential mechanisms through which ATN‐161 exerts its effects, we assessed its impact on chondrocyte differentiation and macrophage polarization by Western blot analysis. Stimulation with recombinant ANGPTL4 led to reduced expression of anabolic markers (ACAN, COL II) in chondrocytes, alongside increased protein levels of markers associated with calcification and fibrosis (COL X, COL I). Treatment with ATN‐161 attenuated these ANGPTL4‐induced alterations in chondrocytes (Figure S9E). Similarly, in macrophages, ATN‐161 suppressed the ANGPTL4‐induced shift toward M1 polarization (Figure S9F). Together, these findings suggest that ATN‐161 may alleviate OA progression by mitigating ANGPTL4‐driven pathological changes in chondrocytes and modulating macrophage polarization.
Discussion
3
Cellular senescence is a major contributor to chronic diseases such as OA and represents a promising therapeutic target [36]. While current research has largely focused on chondrocyte senescence, emerging evidence indicates that synovial cell senescence may occur earlier than that of chondrocytes and play a critical role in OA initiation [37]. For instance, transplantation of senescent synovial cells into mouse knee joints has been shown to induce cartilage degradation and osteophyte formation [38]. However, the specific senescent subpopulations within the synovium and their dynamic interactions with the local microenvironment during OA progression remain poorly characterized. In this study, we identified a distinct SIF subpopulation, L1_SIF, that exhibits strong associations with cellular senescence. Using GRN analysis, we predicted underlying mechanisms driving SIF senescence and experimentally validated the pivotal role of senescent SIF in OA pathogenesis. Our findings provide new insights into synovial aging and senescence‐associated alterations in cellular communication.
FLS are the predominant cell type in the synovium, located mainly in the synovial lining and sublining layers [39]. Beyond their recognized roles in inflammation and proliferation post‐injury, functionally distinct FLS subpopulations contribute diversely to joint pathology [13]. SIF, as a subset of FLS, traditionally known for secreting lubricating molecules to maintain joint homeostasis [39], has recently been implicated in OA progression through aberrant secretion of factors such as R‐spondin 2, which activates Wnt/β‐catenin signaling [13]. Consistent with these findings, our gene set scoring analysis revealed that SIF display significantly elevated senescence, DNA damage, and oxidative stress scores compared to other synovial cells. In OA mouse models, we observed strong co‐localization of the SIF marker PRG4 with the senescence marker p21^Cip1^, whereas the SSF marker CXCL12 showed weaker association with p21^Cip1^. These results align with recent reports of senescent fibroblast accumulation in the synovial lining of OA patients, where elevated RCNA1 expression induces mitochondrial dysfunction and premature senescence in SIF [40]. Our data further support a role for oxidative stress, as indicated by elevated pathway activity in SIF (Figure S2E), suggesting that targeting SIF senescence may mitigate OA‐associated inflammation.
Accurate identification of senescent cells is critical for their targeted removal [41]. However, no single marker has been found to be unique to senescent cells, including commonly used indicators such as SA‐β‐gal, p21^Cip1^, and p16 [INK4a [42]]. In this study, we employed a multifaceted approach to identify senescent cells within SIF and identified L1_SIF as a subpopulation with prominent senescent features. We further demonstrated that the TF EGR1 regulates senescence pathways in SIF. EGR1 is involved in diverse cellular processes, including proliferation, apoptosis, senescence, and differentiation [43]. Its role appears context‐dependent: while EGR1 promotes senescence in cardiac fibroblasts, facilitating ECM degradation and cardiac repair [44], its downregulation alleviates senescence and mitochondrial dysfunction in nucleus pulposus cells, attenuating intervertebral disc degeneration [45]. In OA, EGR1 is upregulated in aged mouse cartilage and influences chondrocyte metabolism by suppressing COL2A1 and enhancing MMP9/MMP13 expression [45]. Our results show that EGR1 inhibition suppresses the TLR2 signaling pathway and downregulates SASP factors, including CXCL9, CXCL10, and IL‐6. Given that TLR2 activation promotes cell cycle arrest and SASP secretion [46, 47] and induces inflammatory responses and mitochondrial impairment in chondrocytes [48], our findings suggest that EGR1 modulates FLS senescence and SASP secretion via TLR signaling.
Senescent cells accumulate in joints over time and secrete pro‐inflammatory and matrix‐degrading factors that perpetuate inflammation and tissue damage [2]. Through autocrine and paracrine mechanisms, SASP components such as IL‐6 and IL‐8 can reinforce senescence in a feedback loop and induce senescence in neighboring cells [49]. In this study, we demonstrated that senescent SIF promote M1 macrophage polarization and OA‐like changes in chondrocytes via secretion of ANGPTL4. Although ANGPTL4 has been reported to exhibit both anti‐ and pro‐inflammatory properties [50], such as mitigating saturated fat‐induced inflammation and promoting anti‐inflammatory macrophage accumulation [51], our results highlight its catabolic role in the OA joint. To simultaneously target the downstream signaling of ANGPTL4 and other pathogenic ligands that engage the α5β1 integrin receptor, we employed ATN‐161, a selective inhibitor of α5β1. We showed that ATN‐161 delays cartilage degeneration, reduces synovitis, and partially attenuates OA progression. Previous studies indicate that ATN‐161 only partially rescues ITGBL1 depletion‐induced OA damage, which may be attributed to the activation of other integrin subtypes not inhibited by ATN‐161 [52]. Nevertheless, regarding the ANGPTL4–α5β1 axis, ATN‐161 effectively reduces α5β1 expression in both chondrocytes and macrophages and inversely suppresses ANGPTL4 secretion from SIF—an effect likely resulting from the amelioration of the inflammatory joint microenvironment following ATN‐161 treatment. Taken together, these data reveal a potential mechanism through which senescent SIF influence macrophages and chondrocytes via the ANGPTL4–α5β1 axis (Figure 9).
A schematic diagram illustrating senescent SIF exacerbate OA development by promoting M1 macrophage polarization and OA‐like changes in chondrocytes.
Conclusion
4
This study reveals the crucial involvement of SIF senescence in OA progression. From a cellular senescence perspective, we identified that senescent cells in the OA synovium are predominantly located within SIF, and that the TFs EGR1 and ATF3 mediate the induction of senescence in this population. Targeting EGR1 effectively reduced the expression of SASP factors in senescent SIF. Regarding intercellular communication, we demonstrated that senescent SIF promote M1 macrophage polarization and induce OA‐like changes in chondrocytes through the ANGPTL4–α5β1 axis. Importantly, pharmacological inhibition of this signaling pathway attenuated OA progression in vivo. Together, these findings elucidate the mechanistic contributions of senescent SIF to OA pathogenesis and provide a foundation for developing novel therapeutic strategies aimed at mitigating OA through targeting synovial senescence.
Experimental Section
5
Processing of scRNA‐seq Data
5.1
Single‐cell datasets for human synovial tissue (GSE216651) [18] and human cartilage (GSE169454) [34] were obtained from GEO. Synovial data included three OA and three control samples, and cartilage data included three normal and four OA samples. The Seurat package (V4.4.0) [53] was used for data integration, quality control, dimensionality reduction, clustering, annotation, and identification of DEGs. The quality control criteria are as follows: nFeature_RNA 200–4000, percent_mt < 20, nCount_RNA 1000–20000. The MarkDoublets function of scutilsR (V0.1.0) was used to filter out double cells. We employed the harmony integration method to correct batch effects. UMAP and t‐SNE functions were used for visualizing cell populations. The classic markers for each cell type in synovial tissue were shown as follows: SSF: PDGFRA, COL1A2, MEDAG, NAV1, MMP2, and CXCL12; SIF: PRG4, CLIC5, HTRA1, HAS1 and PDPN; EC: PECAM1, EMCN, CDH5, VWF and FLT1; Resid.Mφ: F13A1, SELENOP, LYVE1, MAN1A1, and SLC40A1; Inflamm.Mφ: CCL3, TNF, IL1B, OLR1, and CCL3L1; Monocytes: FCGR3A, TIMD4, MARCO, and FN1; Mast: TPSB2, TPSAB1, CPA3, MS4A2 and CD69; DC: HLA‐DPB1, HLA‐DQA1, HLA‐DQB1, FCER1A and CD1C; mural cell: TAGLN, MYL9, TPM2, NOTCH3 and ACTA2; T cell: IL32 and CXCR4; B cell: CD79A, HLA‐DRA, MS4A1, CD74 and CD83; granulocytes: FCN1, S100A9, S100A12, EREG and S100A8 [12, 18]. The DEGs and marker genes for each cell type were identified using FindAllMarkers and FindMarkers functions.
PyScenic Analysis
5.2
We used pySCENIC [19] to calculate gene co‐expression based on the gene expression matrix. This process primarily involves three steps: co‐expression analysis, target gene motif enrichment analysis, and regulator activity assessment. In brief, we first inferred co‐expression modules between TFs and target genes using GRNboost. We then used RcisTarget to analyze each co‐expression module to identify enriched motifs, retaining only those modules and targets for which the TFs motif is enriched. Each TFs and its direct target genes form a regulon. We used AUCell (v1.24.0) to calculate the RAS matrix, which can be used for clustering cells. This allows us to visualize regulon networks via UMAP to identify cell types and states.
Gene Set Enrichment Analysis
5.3
GSEA, GO, and KEGG enrichment were performed using clusterProfiler (v4.10.0). Senescence scores were computed using GSVA (v1.50.0) with six senescence‐related gene sets: FRIDMAN_SENESCENCE_UP [21], Cellage [22], Geneage [23], Global Senescence Literature Curated 2020 [24], SASP_Literature Curated UP [24], and SenMayo [25].
Trajectory Inference Analysis
5.4
We used Monocle2 (V.2.30.0) to construct the cell pseudotime trajectory for the OA SIF. After filtering out genes with low expression (num_cells_expressed < 10), we identified 16,157 DEGs using the differentialGeneTest function. We selected the top 1,000 DEGs ranked by q‐value for dimensionality reduction analysis using the DDRTree function. After pseudotime cell ordering, the cell trajectory was visualized using the plot_cell_trajectory function with default parameter settings.
Senescence Index Tool
5.5
The senescence index tool (SIT) [28] algorithm is primarily used to identify senescent cells. First, the average expression of known senescence markers (P16, P15, P19, P21, P27, and PAI1) was defined using the AddModuleScore function from the Seurat package. Second, GSVA scores for each cell were calculated using three different gene sets (KEGG CELL CYCLE, REACTOME CELL CYCLE, and WP CELL CYCLE) with the GSVA package, and the direction of the scores was inverted. Third, the scores obtained from the first and second steps were normalized separately and then summed. Finally, the results from the third step were divided into quantiles, and cells in the fourth quantile were classified as senescent cells.
Regulon Module Analysis
5.6
The Context Specificity Index (CSI) [54] was used to identify regulatory submodules and served as a context‐dependent measure for detecting specific associated partners. First, the activity ratings of each pair of regulators were calculated using Pearson correlation coefficients (PCC). Then, the proportion of regulator pairs was determined by comparing their PCC values. A higher CSI indicates stronger correlations between regulators. Hierarchical clustering was performed on the fully connected CSI matrix to identify different regulatory submodules. Additionally, a threshold of five was selected for the regulator association network to study the connections between different regulators. The activity score associated with each cell type for a given regulatory submodule was defined as the average activity score of its regulator members across all cells within that cell type. Subsequently, the top‐ranked cell types were identified for each module, and the results were displayed on a UMAP plot.
Cell–Cell Communication Analysis
5.7
The CellChat package (V1.6.1) [55] was used to infer and quantify cell–cell communication networks. For the analysis of cell communication between SIF and chondrocytes, we first merged the single‐cell data from synovium and cartilage, and then conducted the cell–cell communication analysis. We performed all analyses using the standard pipeline (https://github.com/sqjin/CellChat), with default parameters and the human ligand–receptor database.
OA Animal Model
5.8
All procedures followed the National Institute of Health Guide for the Care and Use of Laboratory Animals and were approved by the Animal Care and Use Committee of Chongqing Medical University (approved number: IACUC‐CQMU‐2024‐0833). To induce mechanical instability‐associated OA, we surgically performed DMM in the right knee joints of 10‐week‐old male mice [10]. Sham operations were also performed by opening and exposing the structures of the right knee and then closing the skin incision without disturbing the joint tissue. As for α5β1 inhibitor ATN‐161 interaction, one week after the DMM surgery, mice received intra‐articular injections of the α5β1 inhibitor ATN‐161 (1 mg/kg, 10 µL) weekly. At the fourth and eighth week, mice were euthanized by anesthetic overdose. Knee joints were harvested for subsequent histological analysis.
Histology, Immunohistochemistry, and Immunofluorescence
5.9
For histological analysis, the knee joint samples were fixed with 4% neutral paraformaldehyde, decalcified with 0.5 m EDTA (pH 7.4) for 4 weeks, and then embedded in paraffin. The samples were then sectioned into 4.5 µm pieces. Subsequently, H&E staining, Masson's Trichrome Stain, and Safranin O‐Fast Green staining were conducted to investigate the cartilage and synovium, respectively. To evaluate histopathological differences in joint tissues among the experimental groups, articular cartilage degeneration was assessed using a combined grade and stage OA histopathology scoring system [56], while synovitis was evaluated using the established synovitis scoring system [57]. The scoring was performed independently by three authors who were blinded to the group assignments. The resultant scores were then pooled for statistical analysis. For comparisons between two groups, the Mann–Whitney U test was applied. For comparisons across multiple groups, the Kruskal–Wallis test was used, followed by the Benjamini, Krieger, and Yekutieli post‐hoc test. For immunohistochemistry analysis, antigen retrieval was performed by microwave heating in citrate buffer (pH 7.2). Endogenous peroxidase activity was quenched with 3% H_2_O_2_, followed by blocking with 10% normal goat serum. Sections were incubated overnight at 4°C with primary antibodies against Collagen type II and Collagen type I (1:200). After washing, HRP‐conjugated secondary antibodies were applied for 1 h at room temperature. Immunoreactivity was visualized using 3,3'‐diaminobenzidine (DAB) with hematoxylin counterstaining. Finally, sections were dehydrated, cleared in xylene, and coverslipped. Protein expression was quantified using ImageJ software from images captured under an optical microscope. The sections were analyzed with an inverted optical microscope (Leica, DMI‐8). For immunofluorescence analysis, knee joint samples were retrieved from the mice and snap‐frozen in an optimal cutting temperature medium. Immunofluorescence staining and analysis were performed as described previously. Knee joint sections were cut using a cryotome, mounted on slides, and stained with different primary antibodies overnight at 4°C. Primary antibodies were then visualized with species‐appropriate secondary antibodies. The sections were mounted by antifade reagent with DAPI. The sections were analyzed with a confocal laser microscope (ZEISS, Germany).
Cell Culture
5.10
Healthy 4‐week‐old BALB/c mice were denecked and sacrificed, then soaked in 70% ethanol for 3 min for disinfection. The knee joint of the mouse was opened, and the synovial membrane was dissected carefully. The trimmed synovial tissue was transferred to the mixed culture medium containing Dulbecco's Modified Eagle Medium (DMEM; Gibco) and collagenase type IV (1 mg/mL; Solarbio) for 2 h. The supernatant was filtered through a cell strainer (40 µm pore size) and centrifuged to collect the cell precipitate [58]. The FLS were washed twice with PBS and cultured with DMEM supplemented with 10% fetal bovine serum (FBS; Biochannel Biological Technology) and 1% penicillin−streptomycin solution at 37°C in 5% CO_2_ incubator. The RAW267.4 cells and the chondrogenic cell line ATDC5 cells were cultured in DMEM 10% FBS and 1% penicillin−streptomycin solution. FLS at P2, ATDC5, and RAW264.7 within P10 were used for the cell experiment.
SA‐β‐Gal Staining
5.11
SA‐β‐gal staining was performed by using the Cell senescence β‐galactosidase staining kit (Bestbio, China) according to the manufacturer's protocol. Briefly, FLS was pretreated with H_2_O_2_ for 24 h. At 0, 24, 48, and 72 h post H_2_O_2_ stimulation, cells were washed with PBS and fixed with 4% paraformaldehyde for 5 min. Then, the cells were washed and incubated with SA‐β‐gal staining solution at 37°C for 16 h. After incubation, the cells were washed and imaged by an upright fluorescence microscope (ZEISS, Germany). Total cells and SA‐β‐Gal–positive cells were calculated in three random fields per culture dish via ImageJ.
Cellular Immunofluorescence Staining
5.12
Cells were washed with PBS and fixed with 4% paraformaldehyde for 10 min. Next, cells were permeabilized with 0.1% Triton X‐100 (Sigma, Japan) for 15 min and blocked with bovine serum albumin (BSA; Sigma, Germany) for 1 h at room temperature, followed by the incubation of the primary antibodies overnight at 4°C (Table S3). Primary antibodies were then visualized with species‐appropriate secondary antibodies for 1 h at room temperature. Nuclei were stained with DAPI (Beyotime, China). Images were taken by a confocal laser microscope (ZEISS, Germany) using different channels, and results were analyzed using ImageJ. For the transcription factors EGR1 and ATF3, nuclear masks were generated by thresholding the DAPI channel to measure nuclear mean fluorescence intensities of EGR1, ATF3, and P21^Cip1^. The relative fluorescence expression per nucleus and the proportion of EGR1⁺ P21^Cip1⁺^/DAPI⁺ cells were then quantified and plotted. For PRG4, ANGPTL4, MIF, α5β1, and CXCR4, the average fluorescence intensity at the single‐cell level and the proportion of immunofluorescence‐positive cells were measured, quantified, and plotted. For all other immunofluorescence analyses not specifically described, relative quantification was performed by measuring the mean fluorescence intensity of the entire cell population within the corresponding channel of each image.
RNA‐Sequencing Study
5.13
Bulk RNA sequencing was performed on control, H_2_O_2_‐induced senescent (SEN), and EGR1‐inhibited (EGR1‐inhibit) FLS. After quality inspection, libraries were sequenced, and DEGs were identified using DESeq2. The following criteria were used to recognize DEGs: p < 0.05 and |log2(FoldChange)| > 0.5. We performed gene set enrichment analysis using the enrichGO, enrichKEGG, and GSEA functions from the clusterProfiler package (V4.10.0).
Transwell Co‐Culture Assay
5.14
For co‐culture assay, transwell (0.4 µm membrane, Beyotime, China) was utilized to establish the co‐culture system. Briefly, 5 × 10^4^ RAW264.7 cells were counted and seeded in a 24‐well plate 6 h prior to co‐culture. Then an equivalent number of FLS pretreated with or without H_2_O_2_ for 24 h were seeded in the insert apparatus, which was directly placed on the top of a well containing RAW264.7 cells. After co‐culture, RAW264.7 cells were harvested for further analysis. Likewise, a similar co‐culture system was established with FLS on the top chamber and ATDC5 cells in the plate well.
Transwell Migration Assay
5.15
Cell migration was measured by a transwell assay (8 µm membrane, Beyotime, China) following manufacturer's instructions. Briefly, 1 × 10^5^ cells were seeded on the matrigel, and the same number of cells of senescent or non‐senescent FLS were seeded underneath the plate well. After co‐culturing RAW264.7 cells with senescent or non‐senescent FLS for 12 h, we removed the matrigel and washed off the remaining cells. Then we used crystal violet dye solution for staining and photographed by fluorescence microscope (ZEISS, Germany).
Western Blot
5.16
The target cells were extracted from the co‐culture system, and the total protein was extracted by radioimmunoprecipitation assay lysis buffer (RIPA) containing phenylmethylsulfonyl fluoride, and quantified by BCA protein kit (Beyotime, China). The proteins were then isolated using SDS‐PAGE and transferred to PVDF membranes. After incubation overnight at 4°C with the corresponding primary antibody (Table S3), the PVDF membranes were washed with buffer, followed by incubation with the corresponding secondary antibody. Finally, hypersensitive ECL was used for imaging, and the imprinting strength was quantified by Image J software. The concentration of primary antibody protein was 1:1000, and the concentration of secondary antibody was 1:10000. For the Western blot of EGR1 and ATF3 specifically, nuclear proteins were extracted using a Nuclear and Cytoplasmic Protein Extraction Kit (Beyotime, China) according to the manufacturer's instructions, with Histone H3 (Proteintech, China) serving as the nuclear loading control.
Statistical Analysis
5.17
Statistical analysis was performed using GraphPad Prism (version 10.6.1). Data were expressed as mean ± standard deviation. For different types of data, one‐way or two‐way ANOVA and unpaired t‐test were employed for parametric testing, while the Mann–Whitney test and Kruskal–Wallis test were utilized for nonparametric testing. Differences were considered statistically significant if p < 0.05 (), p < 0.01 (), and p < 0.001 ().
Author Contributions
M.D. and Y.J. conducted the majority of the assays, acquired and analyzed data, and drafted the manuscript. K.C. and Z.C. assisted in project design and provided technique supports. Y.H. and N.C. participated in animal experiments. C.C, Z.Q., and Y.H. conceived, designed the project, supervised experiments, and provided research resources. All authors approved the final version of this manuscript.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Supporting File 1: advs73455‐sup‐0001‐SuppMat.docx.
Supporting File 2: advs73455‐sup‐0002‐TableS1.xlsx
Supporting File 3: advs73455‐sup‐0003‐TableS2.xlsx.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1D. T. Felson , Y. Zhang , M. T. Hannan , et al., “The Incidence and Natural History of Knee Osteoarthritis in the Elderly,” The Framingham Osteoarthritis Study Arthritis Rheum 38, no. 10 (1995): 1500.7575700 10.1002/art.1780381017 · doi ↗ · pubmed ↗
- 2B. O. Diekman and R. F. Loeser , “Aging and the Emerging Role of Cellular Senescence in Osteoarthritis,” Osteoarthritis and Cartilage 32, no. 4 (2024): 365–371.38049031 10.1016/j.joca.2023.11.018PMC 10984800 · doi ↗ · pubmed ↗
- 3B. G. Childs , M. Durik , D. J. Baker , and J. M. van Deursen , “Cellular Senescence in Aging and Age‐related Disease: From Mechanisms to Therapy,” Nature Medicine 21, no. 12 (2015): 1424–1435.10.1038/nm.4000 PMC 474896726646499 · doi ↗ · pubmed ↗
- 4Y. Zhang and Y. Zhou , “Advances in Targeted Therapies for Age‐related Osteoarthritis: A Comprehensive Review of Current Research,” Biomedicine & Pharmacotherapy 179 (2024): 117314.39167845 10.1016/j.biopha.2024.117314 · doi ↗ · pubmed ↗
- 5J. S. Price , J. G. Waters , C. Darrah , et al., “The Role of Chondrocyte Senescence in Osteoarthritis,” Aging Cell 1, no. 1 (2002): 57–65.12882354 10.1046/j.1474-9728.2002.00008.x · doi ↗ · pubmed ↗
- 6P. R. Coryell , B. O. Diekman , and R. F. Loeser , “Mechanisms and Therapeutic Implications of Cellular Senescence in Osteoarthritis,” Nature Reviews Rheumatology 17, no. 1 (2021): 47–57.33208917 10.1038/s 41584-020-00533-7PMC 8035495 · doi ↗ · pubmed ↗
- 7B. O. Diekman , G. A. Sessions , J. A. Collins , et al., “Expression of p 16 INK 4a is a biomarker of chondrocyte aging but does not cause osteoarthritis,” Aging Cell 17, no. 4 (2018): 12771.10.1111/acel.12771 PMC 605246429744983 · doi ↗ · pubmed ↗
- 8J. N. Farr , D. G. Fraser , H. Wang , et al., “Identification of Senescent Cells in the Bone Microenvironment,” Journal of Bone and Mineral Research 31, no. 11 (2016): 1920.27341653 10.1002/jbmr.2892 PMC 5289710 · doi ↗ · pubmed ↗