Defining subcellular synovial responses in TMJ osteoarthritis onset via mechanical stress and articular disk derangement models
Kazuhiro Shibusaka, Soichiro Negishi, Asuka Terashima, Miki Maemura, Hiroshi Yoshida, Masahiro Hosonuma, Nobuhiro Sakai, Young Kwan Kim, Yutaka Suzuki, Hiroyuki Okada, Fumiko Yano
TL;DR
This study explores how mechanical stress and joint issues in mice lead to changes in the synovium, offering insights into the early stages of temporomandibular joint osteoarthritis.
Contribution
The study introduces a novel integrated transcriptomic approach combining subcellular spatial transcriptomics and single-cell RNA sequencing to investigate TMJ-OA onset.
Findings
Molecular alterations in the synovium of the articular disk were identified in response to mechanical and inflammatory stimuli.
Cell type– and cluster–specific catabolic changes were observed, suggesting roles in TMJ-OA onset.
The study provides a methodology-oriented resource for understanding TMJ disorders at the molecular level.
Abstract
Temporomandibular joint osteoarthritis (TMJ-OA), the most common degenerative disease of the TMJ, is influenced by various adaptive, inflammatory, and mechanical stressors. In this study, we describe molecular alterations of the synovium of the articular disk in response to mechanical and inflammatory stimuli. Using an integrated transcriptomic approach combining subcellular spatial transcriptomics and single-cell RNA sequencing in murine models of mechanical stress and articular disk derangement, we characterized synovial changes associated with adipogenesis, fibrosis, and macrophage activation. In addition, cell type–and cluster–specific catabolic changes were observed under these stress conditions, suggesting potential contributions to TMJ-OA onset. These results provide a methodology-oriented resource for investigating the molecular pathology of TMJ disorders and may help guide…
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 10
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9- —This work was supported by JSPS KAKENHI Grant Numbers 23K27796 (to F.Y.), 21KK0155 (to F.Y.), and JP22H04925(PAGS) (to F.Y.)
- —Young Investigator Award from the American Society for Bone and Mineral Research The Nakatomi Foundation
- —JSPS DC Research Fellowship
- —JSPS KAKENHI Grant Numbers JP22H04925(PAGS)
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
TopicsTemporomandibular Joint Disorders · Osteoarthritis Treatment and Mechanisms · Total Knee Arthroplasty Outcomes
Introduction
The temporomandibular joint (TMJ) is a specialized synovial joint essential for the mobility and function of the mammalian craniofacial system. Like the disorders of other synovial joints, TMJ disorders (TMDs) are influenced by various inflammatory, traumatic, and developmental factors.^1^ In the United States, the reported prevalence of TMDs with at least one symptom among young to middle-aged adults (20–50 years of age) is between 40 and 75%, with a female-to-male ratio ranging from 3:1 to as high as 9:1.^1,2^ The American Academy of Orofacial Pain classifies TMDs into several subtypes, including congenital or developmental disorders, disk-derangement disorders, degenerative joint disorders, trauma, and TMJ hypermobility or hypomobility.^1^
Osteoarthritis (OA) is the most prevalent form of joint disease characterized by cartilage degeneration. Evidence suggests that synovitis and related pro-inflammatory mediators play a crucial role in the pathogenesis of OA, particularly in the observed effects on articular cartilage.^3,4^ Furthermore, baseline MRI data indicate that synovitis is an independent risk factor for the onset of knee OA even after adjusting for other structural pathologies linked to synovitis.^5^
Human knee OA samples from patients who underwent knee arthroplasty have recently been assessed using single-cell RNA sequencing (scRNA-seq).^6^ However, the number of patients with TMJ-OA who undergo TMJ arthroplasty is limited—despite similar incidence rates of OA, knee joint replacements are 2000 times more common than TMJ replacements.^7^ Additionally, the number of research articles on knee joint is six times the number of articles on TMJ.^7^ Due to this discrepancy, basic and translational research using knee orthopedics as the template is expected to drive the development and implementation of clinical products for the TMJ^7^; however, the underlying molecular mechanisms in TMJ-OA differ from those in knee joint OA due to the differences in the structure, tissue components, and cellular origins of both joints.^8^ Moreover, despite the morphological differences in the TMJ of humans and rodents, mice and humans share many developmental and molecular pathway,^8^ due to which mice models are now increasingly being used for studying human TMDs.
Given the potential roles of articular disk derangement (ADD) and mechanical stress (MS) in the etiology of temporomandibular joint osteoarthritis (TMJ-OA), we independently established two distinct mouse models—one for ADD and one for MS—to investigate molecular alterations in the synovium of the articular disk that underlie TMJ homeostasis and degeneration. These models allowed us to examine how specific mechanical or structural perturbations contribute to the onset of TMJ-OA.
To characterize the cellular and spatial landscape of the articular disk and surrounding synovium during the early stages of TMJ-OA, we adopted a comprehensive, multi-modal framework that integrates histological analysis, bulk RNA sequencing, single-cell RNA sequencing (scRNA-seq), and high-resolution spatial transcriptomics using the Xenium platform. Although prior studies have implicated synovial inflammation and mechanical loading in TMJ-OA progression, detailed spatial and single-cell molecular insights during the onset phase remain limited.
This study is intended as a methodological resource that introduces and validates two TMJ-OA mouse models while systematically applying advanced transcriptomic approaches. Through this strategy, the study provides a technical and conceptual foundation for future investigations into the mechanistic and therapeutic aspects of TMJ disorders. An overview of the experimental design—including TMJ-OA models, time points, and analytical methods—is presented in Fig. 1.Fig. 1. Overview of the experimental design and workflow. Schematic representation of the TMJ-OA mouse model experiment, including the different groups (Ctrl, MS, and ADD) and their respective sampling timelines. Mice underwent either a sham operation, MS setup, or ADD surgery at 9 weeks of age, followed by tissue sampling at 12 weeks. The collected samples were analyzed using micro-CT (n = 5) and histological assessments (n = 5). Molecular analyses included bulk RNA-seq (n = 3), real-time RT-PCR (n = 6), single-cell RNA-seq (n = 4–7), and spatial transcriptomics (Xenium, n = 1). The study focused on condylar cartilage and synovium, integrating comprehensive transcriptomic approaches to explore the molecular and cellular changes associated with mechanical stress and disk derangement in TMJ-OA pathogenesis
Results
Development of TMJ-OA mouse models
To replicate the pathological conditions of human TMJ-OA caused by MS and inflammation due to malocclusion^9–11^ and ADD,^12–14^ we used a modified MS mouse model^15^ and developed an original surgical ADD mouse model, respectively (Supplementary Fig. S1).
For the MS model, a metal plate was set on the posterior surface of the maxillary incisors of mice, positioned at a 45-degree angle to the palatal plane, to induce malocclusion using a custom-made mouth opener (Supplementary Fig. S1b–e). For the ADD model, the articular disk of the TMJ was surgically displaced anteriorly and double-tied to the zygomatic arch using 7-0 sutures (Supplementary Fig. S1f, g) to induce ADD and inflammation. There was no significant difference in the body weights of the model mice during the experimental period (at 8, 9, and 12 weeks of age) (Supplementary Fig. S1a, h, i). Micro-computed tomography (micro-CT) images of the mandibular condyle in ADD and MS model mice (Fig. 2a, b) revealed degenerative changes, with ADD model mice developing deformities comparable to those observed in patients with TMJ-OA,^12–14^ and MS model mice exhibiting structural and morphological changes characteristic of Angle’s Class II Division 2 malocclusion.^11^Fig. 2. Subchondral bone loss in the MS and ADD models. a Representative 3D images of a mouse head, with the inset box indicating the TMJ showing the region of interests (ROI) in the condyle. The color-scale shows bone mineral density values. Scale bars, 1 mm. b Representative micro-CT images of the condyle at 12 weeks of age in the Ctrl and MS and ADD model groups, taken 3 weeks after induction. The black arrow indicates subchondral bone loss, the yellow arrow indicates osteophyte formation, and the white arrow indicates bone erosion. Temporal: temporal bone. Scale bars, 1 mm. c TRAP staining of the subchondral bone area of the condyle in the Ctrl, MS, and ADD model groups 3 weeks after induction. Lower panels show a higher magnification of the upper panels. Scale bars, 100 μm. d–j Quantitative analysis of the condyles in the Ctrl and MS and ADD model groups 3 weeks after induction. d BV/TV (%): bone volume fraction; e Tb.Th (µm): trabecular thickness**; f** TMD (mg/cm^3^): tissue mineral density; g Tb.N (1/mm): trabecular number; h Tb.Sp (µm): trabecular separation; i Oc. S/BS (%): osteoclast surface/bone surface; j Oc. N/BS (cell/mm^2^): osteoclast number/bone surface. All data are presented as mean ± SEM values. **P < 0.01, ***P < 0.001, ****P < 0.000 1 were compared between groups (n = 5 mice per group). Symbols represents individual mice
Condylar deformation and subchondral bone degradation in the TMJ-OA models
As the condyle of TMJ-OA patients shows subchondral bone loss^16,17^, we further investigated subchondral bone formation in the MS model, which may affect mechanical regulation of bone remodeling^18,19^ and the ADD model, which causes condyle growth and degradation (Fig. 2a, b). Tartrate-resistant acid phosphatase (TRAP) staining showed that osteoclast activity was significantly increased in the subchondral bone of the condyle in both MS and ADD models compared to the Ctrl (Fig. 2c). micro-CT analyses revealed that bone volume over total volume (BV/TV) were decreased the subchondral bone of the condyle in both the MS and ADD models compared to the Ctrl (Fig. 2d), trabecular thickness (Tb.Th) and tissue mineral density (TMD) were decreased the subchondral bone of the condyle in the ADD models compared to the Ctrl (Fig. 2e, f). Additionally, there were increases in trabecular number (Tb.N), trabecular separation/spacing (Tb.Sp), osteoclast surface/bone surface (Oc. S/BS), and osteoclast number/bone surface (Oc. N/BS) in the subchondral bone of the condyle in both the MS and ADD models compared to the Ctrl (Fig. 2g–j). These results indicate that MS and ADD-induced TMJ-OA lead to abnormally activated trabecular bone turnover and osteoclast activity.
Pathological changes in condylar and synovial tissues in TMJ-OA model mice
The histological changes in the model mice were investigated after 3 weeks using hematoxylin and eosin (H&E) and safranin O staining (Fig. 3a, b). Compared to the whole TMJ of control (Ctrl) mice, which had normal synovium and a round condyle, the MS model mice showed cartilage fibrillation in the superficial layer and adipogenic and fibrous changes in the posterior synovium of the articular disk. The ADD model mice displayed significant deformity of the condyle, with alterations in the extracellular matrix composition, including loss of proteoglycan in the superficial cartilage layer, and synovial hyperplasia throughout the entire disk (Fig. 3a, b).Fig. 3. Pathological changes in the condyle and synovium of the TMJ-OA model mice. a Representative H&E staining image of the condyle at 3 weeks after induction surgery (scale bars = 1 mm). Inset boxes show the regions in the lower panels (scale bars = 100 μm). Black arrowheads indicate fibrillations, and blue arrowheads indicate areas of superficial bone erosion at the outer surface of the cortical bone. b Safranin O staining of the condyle at 3 weeks after induction surgery. Inset boxes show the regions in the lower panels. Red arrowheads indicate irregularities in the surface lamina. Black arrowheads indicate fibrillations in the superficial zone of cartilage in the ADD models. Yellow arrowheads indicate pre-hypertrophic chondrocytes in the MS model and hypertrophic chondrocytes in the ADD model. White arrowheads indicate hypocellularity in both the MS and ADD models. Orange arrowheads indicate the clustering of chondrocyte in the ADD models. Scale bars, 100 μm. c Measurement of the modified Mankin score for assessing cartilaginous degradation. d H&E staining of the synovium at 3 weeks after induction surgery. Inset boxes show the regions in the lower panels. Black arrowheads indicate increased thickening of synovial lining cells and enhanced infiltration of inflammatory cells. Red arrowheads indicate vascular invasion into the synovial lining cell layer. Yellow arrowheads indicate fibroblastic cells in the synovial mesenchyme. Scale bars, 100 μm. e Synovitis scores for assessing the severity of synovitis. f Representative H&E staining images of the MS model posterior synovium. The inset box indicates adipocytes, as shown in the enlarged H&E staining image. g mRNA levels of Pparg, Adipoq, Cebpa, Fabp3, and Fabp4 in the articular disk synovium in Ctrl and MS model mice 3 weeks after induction. Symbols represent individual mice; error bars show the mean ± SEM (n = 5 mice per group for c, e). **P < 0.01, ***P < 0.001, ****P < 0.000 1; one-way ANOVA followed by Dunnett’s post hoc test (for c, e) and Student’s unpaired two-tailed t-test (for g)
Safranin O staining revealed that the surface of the condyle in MS model mice was irregular and that the condyle contained hypertrophic cartilage layer chondrocytes. Degradative cartilaginous matrix and unlayered columnar chondrocytes and hypertrophic chondrocytes were observed in the condyles of ADD model mice (Fig. 3b). Furthermore, the Modified Mankin scores indicated that TMJ-OA development in both MS and ADD model condyles was significantly accelerated compared to that in Ctrl condyles (Fig. 3c).
We assessed extracellular matrix degradation and evaluated cartilage thickness in the condyle by dividing it into three regions: Posterior, Middle, and Anterior (Supplementary Fig. S2a, d). Cartilage thickness varied among the three regions, with the Middle and Posterior regions being thicker than the Anterior region. Notably, cartilage degeneration was most pronounced in the ADD model, particularly in the Middle and Posterior regions (Supplementary Fig. S2b, c, e, f). Additionally, hypertrophic chondrocytes were also increased in the Middle and Posterior regions (Supplementary Fig. S2c). In the MS model, significant differences were observed only in the Posterior region for the Safranin O-positive area and hypertrophic zone thickness (Supplementary Fig. S2b, f). Compared to the Ctrl synovium, adipose tissue was observed in the synovium of the TMJ articular disk in MS model mice, and significantly enlarged and hyperplastic fibrotic synovial lining cell layers were observed in the ADD synovium (Fig. 3d). The synovitis scores indicated that the severity of synovial hypertrophy was significantly increased in both the MS and ADD models (Fig. 3e). To confirm adipogenic differentiation in the articular disk synovium in the MS model, we examined the expression of adipogenic markers and observed that adipogenesis-associated genes were upregulated in the MS model compared to that in the Ctrl synovium (Fig. 3f, g).
These data indicate that mechanical and ADD-induced inflammatory stimuli to the TMJ lead to synovial hyperplasia and condylar cartilage degradation, contributing to the onset and progression of TMJ-OA. Furthermore, the histological changes observed throughout the TMJ suggest that the posterior synovium of the articular disk may play a pivotal role in disrupting TMJ homeostasis and initiating pathogenic degeneration during the early stages of disease.
Distinct mRNA expression profiles in Ctrl and MS and ADD model condyles
Next, we used RNA-seq analysis to comprehensively analyze the gene alterations in the articular disks, including the synovium, and mandibular condyles in our three TMJ models (Ctrl, MS, and ADD), three weeks after model establishment (Fig. 4a). Heatmaps based on genes, Gene Ontology terms, and pathways revealed that the gene expression patterns in the articular disk synovium and condyles (each group, n = 3) could clearly be divided into two tissue-specific groups, termed Condyle and Synovium (Supplementary Fig. S3). A Venn diagram of the number of differentially expressed genes (DEGs) between the Ctrl vs. MS and Ctrl vs. ADD group Condyle groups is shown in Fig. 4b.Fig. 4. Comprehensive mRNA analysis of Ctrl, MS, and ADD group mandibular condyles. a Schematic representative image of the condyle and the synovium of the articular disk in the TMJ for bulk RNA-seq samples. b A Venn diagram showing the comparisons of DEGs among the Ctrl, MS, and ADD group condyles. c Heatmap of DEGs for the five common genes among the Ctrl, MS, and ADD group condyles. d Expression levels (TPM) of DEGs for the significantly downregulated common genes in the condyles of MS and ADD mice compared to the Ctrl group. Symbols represent individual mice; error bars show the mean ± SEM (n = 3 mice per group). **P < 0.01 and ***P < 0.001, one-way ANOVA followed by Dunnett’s post hoc test. e Volcano plots of DEGs between the Ctrl and MS group condyles. f Volcano plots of DEGs between the Ctrl and ADD group condyles. g IPA of the RNA-seq data from the Ctrl and ADD group condyles. The top 30 upregulated pathways are shown. P < 0.05 was considered as the significance threshold
Notably, comparing the Ctrl vs. MS and Ctrl vs. ADD group condyles showed that several genes related to cartilage homeostasis and OA were downregulated in the models (Fig. 4c, d). Normalized gene expression levels (in transcripts per million [TPM]) from the RNA-seq data showed that four of these genes—Cytl1,^20–23^ Ankrd37,^24^ Plec,^25,26^ and Mylk^27^—were significantly downregulated in both the MS and ADD condyles (Fig. 4d), as indicated by the volcano plots of the DEGs (Fig. 4e, f). Furthermore, the expression of Fetub,^28^ which mediates the NF-κB signaling pathway, and Ccn5^29^, Wnt1 inducible signaling pathway protein 2 were significantly elevated in the ADD condyle group (Fig. 4f); in contrast, Pappa2,^30^ a regulator of insulin-like growth factor, was significantly downregulated (Supplementary Fig. S4a). Ingenuity pathway analysis (IPA) with terms and pathways based on the DEGs identified through each group comparison indicated upregulation of “Degradation of the extracellular matrix”, “Role of osteoclasts in rheumatoid arthritis signaling pathway”, and “Osteoarthritis pathway” in the ADD group as compared to the Ctrl group (Fig. 4g). In addition, the expression of degradation markers Ccl5, Ccn5, Mmp13, and Mmp9 was validated in the condyle tissues of the Ctrl, MS, and ADD groups using real-time RT-PCR (Supplementary Fig. S5).
Thus, RNA-seq analyses of the condyles suggest that the onset of TMJ-OA involves changes in gene expression profiles, characterized by the downregulation of cartilage homeostasis-related genes and the upregulation of factors associated with OA progression.
Distinct mRNA expression profiles in the articular disk synovium in the Ctrl and MS and ADD model groups
Comparisons of Ctrl vs. MS and Ctrl vs. ADD group articular disk synovium RNA-seq data revealed commonly upregulated genes related to pathogenesis-related cytokines and catabolic markers (Fig. 5a), as indicated by volcano plots of DEGs (Fig. 5b, c). IPA indicated upregulation of the “Pathogen induced cytokine storm signaling pathway” in the MS synovium in (Fig. 5d) and “Pulmonary fibrosis idiopathic signaling pathway” and “Role of osteoclasts in rheumatoid arthritis signaling pathway” in the ADD synovium (Fig. 5e) compared to the Ctrl group. To verify pathway alterations in the ADD synovium, we also conducted Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis, another method commonly used for pathway analyses and found that “Rheumatoid arthritism”, “HIF-1 signaling pathway”, and “NF-kappa B signaling pathway” were more enriched in the ADD synovium (Fig. 6a).Fig. 5. Comprehensive mRNA analysis of the Ctrl, MS, and ADD group synovium. a A Venn diagram of the comparisons of DEGs among the Ctrl, MS, and ADD group synovium. b Volcano plots of DEGs between the Ctrl and MS group synovium. c Volcano plots of DEGs between the Ctrl and ADD group synovium. d IPA of RNA-seq data from the Ctrl and MS group synovium, showing the top 30 upregulated pathways. P < 0.05 was considered as the significance threshold. e IPA of the RNA-seq data from the Ctrl and ADD group synovium, showing the top 20 upregulated and downregulated pathways. P < 0.05 was considered as the significance thresholdFig. 6KEGG pathway enrichment and expression profiles of Mmps, volcano plot–derived DEGs, and fibroblastic marker DEGs in the Ctrl, MS, and ADD group synovium. a Histogram of KEGG pathway enrichment analysis from the Ctrl and ADD group synovium. b Heatmap of Mmps DEGs (left panel) and expression levels (TPM) of the DEGs (right panels) in the Ctrl, MS, and ADD group synovium. c Heatmap of DEGs (left panel) from the volcano plot in (Fig. 5c) and the expression levels (TPM) of the DEGs (right panels) in the Ctrl, MS, and ADD group synovium. d Heatmap of fibroblastic marker DEGs (left panel) and the expression levels (TPM) of the DEGs (right panels) in the Ctrl, MS, and ADD group synovium. Symbols represent individual mice; error bars show the mean ± SEM (n = 3 mice per group) *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.000 1; one-way ANOVA followed by Dunnett’s post hoc test
Since volcano plots of DEGs showed that Mmp3, 9, and 13 were upregulated in the ADD group and Mmp13 was commonly upregulated in the MS and ADD groups compared to the Ctrl group (Fig. 5b, c), we focused on matrix metalloproteinase expression in the synovium. TPM values from RNA-seq data for six genes—Mmp2, Mmp3, Mmp9, Mmp13, Mmp16, and Mmp19—were significantly increased in the ADD model synovium (Fig. 6b). Moreover, the expression of genes related to macrophages, inflammation, and bone remodeling, such as Ctsk,^31,32^ Ccl21a,^33^ Ccn5,^29,34,35^ Cthrc1,^36,37^ Cdh11,^38^ and Mdk,^39^ were significantly elevated in the synovium of ADD model mice (Fig. 6c). To further investigate both the similarities and differences in the spatial context of inflammatory and catabolic gene expression between the condylar cartilage and synovial tissue, we performed wide-scale spatial gene expression analysis using the Visium HD platform (10× Genomics), which enables transcriptome-wide profiling of approximately 20 000 genes without limitation to predefined target genes. The whole-transcriptome spatial map of our three TMJ models (Ctrl, MS, and ADD) revealed that M1 macrophage markers were expressed predominantly in the posterior synovial tissue of the articular disk, particularly in the superior lamina, in both the MS and ADD models. In addition, compared to the control, M1 macrophages extended deeper into the cartilage layers. This pattern was accompanied by a similar spatial distribution of Mmps markers, which were also observed spreading into the posterior-superior region of the articular disk and into the articular cartilage layers (Supplementary Fig. S6).
As synovitis promotes the production of pain neurotransmitters as well as synovial angiogenesis, which in turn accelerates inflammation and directly leads to synovial fibrosis at the later stage of OA^40^ we then focused on fibroblastic markers in the synovium. TPM values from RNA-seq data for six genes—Col1a1, Col3a1, Fn1, Lum, Dpt, and Fap—were significantly increased in the synovium of ADD model mice (Fig. 6d). Furthermore, the expression of the above markers (Fig. 6b–d) was validated in the synovial tissues of the Ctrl, MS, and ADD groups using real-time RT-PCR (Supplementary Fig. S7), supporting the gene expression trends observed in the RNA-seq analysis. In addition, TPM values and heatmaps from RNA-seq data for other catabolism-related gene families are shown in Supplementary Fig. S8.
Thus, our RNA-seq analyses suggested that cartilage homeostasis in MS and ADD model condyles was disrupted due to imbalances in collagen catabolism in the synovium in the MS model and severe synovial fibrosis in the ADD model. partly due to fibroblast proliferation and the disturbance of collagen synthesis and degradation by matrix metalloproteinases, ultimately leading to excessive collagen deposition in the extracellular matrix.
scRNA-seq identified key cell populations and crosstalk mechanisms in the mechanical and inflammatory stress TMJ models
To investigate the MS and ADD-induced inflammatory changes in the respective models of TMJ, we performed scRNA-seq on cells obtained from the articular disk, including the synovium, 3 weeks after model establishment (Fig. 7a). Following unsupervised graph clustering of the combined datasets of the Ctrl, MS, and ADD groups, uniform manifold approximation and projection (UMAP) of the scRNA-seq data analyzed using the scanpy package^41^ identified 17 cell types and 24 distinct cell clusters (labeled as clusters 0–23) (Fig. 7b, c) (Supplementary Fig. S9), taking into account the percentages of mitochondrial reads, ribosomal reads, and cell cycle phase (Supplementary Fig. S9a–c).Fig. 7scRNA-seq analysis of the Ctrl, MS, and ADD group articular disks. a Schematic overview of the scRNA-seq workflow. b UMAP plot with cell type annotation using CellAssign. c. UMAP plot showing the 0–23 clusters from a total of 9302 cells from articular disks from the Ctrl, MS, and ADD groups. d Streamline presentation based on the spliced/unspliced ratio of RNA using scVelo. e Feature plot on UMAP, showing the number of cells in the articular disks from the Ctrl, MS, and ADD groups. f Cumulative bar plot categorized by cell types. g Cumulative bar plot of the 0–23 clusters by CellAssign (upper panel) and by Ctrl, MS, and ADD groups (lower panel). h Dot plot of synovium-, fibroblast-, catabolism-, and inflammation-related gene expression across the 0–23 clusters
The differentiation pathway was examined based on the RNA-spliced/unspliced ratio using scVelo,^42^ which revealed that the clusters in the articular disk synovium predominantly comprised endothelial, fibroblast, and macrophage clusters (Fig. 7d). The 24 distinct cell clusters exhibited relatively conserved cell proportions, with all principal cell types distributed across the Ctrl, MS, and ADD-induced inflammatory conditions (Fig. 7e). A total of 9302 cells were captured, and the number of cells in the MS and ADD models was higher than that in the Ctrl condition; these differences might be due to the presence of hypertrophic tissues in the articular disks in the MS and ADD model mice (Fig. 7e).
Since adipogenesis-associated genes were upregulated in the synovium of the MS model compared to that in the Ctrl synovium (Fig. 3f, g), we also performed a dimensional reduction analysis (via UMAP) that included the adipocyte cell type (Supplementary Fig. S10). However, adipocytes represented only a small number of plots, labeled near the keratinocyte clusters (Supplementary Fig. S10a). We also verified the expression of the adipogenic marker Ppargc1a using in situ RNA expression analysis (Supplementary Fig. S11), and found that it was upregulated in the MS model synovium, which predominantly contained endothelial, pericyte, and osteoblastic cell populations, while the ADD model synovium predominantly contained fibroblast, keratinocyte, and macrophage cell populations (Fig. 7f). The 24 cell clusters were annotated and linked to the 17 cell types for comparison between the Ctrl, MS, and ADD model populations (Fig. 7g). Among the 24 distinct clusters, 11 clusters (1, 3, 4, 7, 8, 11, 15, 16, 19, 20 and 22) expressed fibroblastic markers such as Col1a1, Fos, Prg4, and Thy1, while 5 clusters (0, 2, 6, 14, and 18) expressed endothelial markers such as Clic5 and Ccl21a (Fig. 7h). Cluster 5 was primarily annotated as a macrophage cluster, while cluster 17 was annotated as a keratinocyte cluster (Fig. 7g). Indeed, matrix plots of gene expression revealed that cluster 2 was enriched for endothelial cell types, while cluster 17 was enriched for keratinocytes (Supplementary Fig. S9d). In addition, fibroblast, endothelial, and macrophage markers at the single-cell level were validated by gene expression in synovial tissues of the Ctrl, MS, and ADD groups using real-time RT-PCR, with Lum, Prox1, and Cd68 serving as respective markers (Supplementary Fig. S12).
Finally, we investigated cell-cell interactions among clusters 0–23 based on ligand-receptor expression levels using the CellChat platform.^43^ Regarding mechanosignal transduction, CellChat analysis of the scRNA-seq data showed that fibroblast clusters (1, 3, 4, 7, 8, 11, 15, 16, 19, 20, and 22) were predicted to interact with the endothelial clusters (0 and 2), macrophage cluster (5), and keratinocyte cluster (17), which was dominant in the ADD model (Supplementary Fig. S13a). Notably, Notch signaling in the endothelial clusters 0, 2, 6, 14, and 18 was conveyed from various clusters within the fibroblastic group (Supplementary Fig. S13a). Additionally, the macrophage-related clusters 5 and 12 were found to receive signals from 16 clusters, including endothelial, keratinocyte, and fibroblast clusters, with Cd36, Cd44, and Cd74 identified as a few of the predicted receptor molecules (Supplementary Fig. S14). In addition, the predictions derived from CellChat analysis of the scRNA-seq data were consistent with the spatial transcriptomic findings from the whole-transcriptome spatial maps of our three TMJ models (Ctrl, MS, and ADD) generated using the Visium HD platform. Notch signaling components (Supplementary Fig. S15a) and endothelial cell markers (Supplementary Fig. S15b) were predominantly expressed in the posterior synovial tissue of the articular disk—particularly in the superior lamina—in both the MS and ADD models, similar to the expression pattern observed for macrophage markers (Supplementary Fig. S6a). Interestingly, keratinocyte markers also exhibited a similar distribution, with marked expression in the posterior synovial tissue in the MS model, and deeper extension into the deep layers of the mandibular condylar cartilage in the ADD model, resembling the spatial patterns seen with M1 macrophages, MMPs, Notch signaling, and endothelial markers (Supplementary Fig. S15).
Collectively, these scRNA-seq data suggest that the identified endothelial, keratinocyte, and macrophage clusters in articular disk synovium might represent potential targets for modulating fibroblast crosstalk in the TMJ disease control.
Spatial transcriptomics revealed gene expression patterns corresponding to pathological changes in TMJ models
To further spatially validate the transcriptomic events, we first performed spatial transcriptome sequencing analysis using the Xenium platform (10x Genomics) (Fig. 8a). Specifically, we examined the spatial expression patterns in the posterior synovium, which exhibited severe pathological changes, across three representative images from the Ctrl, MS, and ADD groups (one image each) (Fig. 8b). To explore the transcriptome features in the synovium, we prepared a custom in situ gene expression profiling panel of 100 genes (Supplementary Fig. S16) (Xenium Custom Gene Expression panel ID: KEPGAC) based on the scRNA-seq data for the articular disk, including the posterior synovium, and performed in situ RNA expression analysis at a single-cell level using a pre-designed panel (ID: mMulti_v1) and the aforementioned custom panel. We obtained expression levels of 28 816, 37 503, and 31 187 cells from the Ctrl, MS, ADD group tissues, respectively. The expression of the signature genes was spatially localized in the regions corresponding to pathological features (Fig. 8c). Moreover, these histological and transcript features were consistent with the gene expression patterns and were further validated by single-cell, high-resolution spatial analysis.Fig. 8. Spatial transcriptomics analysis of the Ctrl, MS, and ADD group posterior synovium. a Schematic overview of the spatial transcriptomics analysis (Xenium) workflow. b H&E staining of the whole TMJ (upper panel), with the inset box showing the H&E staining and the region of interest (ROI) in the Xenium spatial expression pattern images. c In situ gene expression profiling using Xenium for representative genes in the ROI from Ctrl, MS, and ADD group images. d Number of transcript counts for representative genes in the ROI from the Ctrl, MS, and ADD group images. Scale bars, 100 μm
To systematically validate the transcriptomic changes in the posterior synovium, we counted the number of detected transcripts in the region-of-interests (ROIs) of the posterior synovium. The expression levels of the signature gene transcripts in an ROI corresponded to the population of cell types annotated by the scRNA-seq analysis; for example, in the posterior synovium ROI in the ADD model, the number of transcripts for osteogenic markers, such as Spp1, Fos, and Grem1, and catabolic markers, such as Mmp3, Mmp13 and Cxcl14, were notably increased (Fig. 8d) as shown in the cell-type populations in Fig. 7f. These high resolution subcellular spatial transcriptomics data confirmed that the pathological features in the posterior synovium corresponded to the catabolic gene expression patterns.
Integration of scRNA-seq and subcellular spatial transcriptomics analysis results revealed dynamic cellular changes in TMJ pathology
As the subcellular spatial transcriptome analysis provided novel information regarding the in situ gene expression profiles in ROIs during pathological changes within the posterior synovium, spatial subcellular transcriptomics using the Xenium platform allowed single-cell resolution, even though the number of hybridization probes in the pre-designed and custom panels was limited. We then integrated the scRNA-seq data, consisting of 17 cell types and 24 cell clusters (Fig. 7b, c), with the Xenium data using Transferring Annotations to Cells and their Combinations (TACCO), which unifies annotation transfer and decomposition of cell identities^44^ (Fig. 9a). The integration of the transcriptomic profiles between the scRNA-seq and Xenium data led to the prediction of the spatial distribution of 479 transcripts within the designed gene panel and/or the subcellular deconvolution of cell types and clusters in histological sections^45^ (Fig. 9b). We first demonstrated that the subcellular cell type population rates and the counted the number of cells corresponding to the annotated cell types in the entire articular disk including synovium based on the scRNA-seq data (Fig. 9c, e). Since we could successfully obtain cell segmentation and subcellular deconvolution of the 17 cell types integrated by scRNA-seq (Fig. 9c, d), we then focused on the spatial distribution of the annotated cell types, especially the representative clusters of macrophages, endothelial cells, fibroblasts, and keratinocytes, in the posterior synovium (Fig. 9d). The articular disk synovium in the Ctrl, MS, and ADD conditions were predominantly composed of fibroblast cell types; however, differentiated cell types, such as myoblasts, osteocytes, chondrocytes, and osteoblasts, were observed only in the MS and ADD models (Fig. 9e). Specifically, in the macrophage cluster, the pie charts for the Ctrl group showed a higher proportion of macrophages compared to those in the MS and ADD groups. In contrast, the macrophage cell counts in the MS and ADD groups were higher than that in the Ctrl group (Fig. 9e). Next, we obtained cell segmentation and subcellular deconvolution of the 24 cell clusters integrated by scRNA-seq (Fig. 10a), we then focused on the spatial distribution of the annotated cell types, especially the representative clusters of macrophages, fibroblasts, endothelial cells, and keratinocytes, in the posterior synovium. The spatial distribution of the 24 cell clusters exactly corresponded to the annotated cell types by scRNA-seq analysis (Fig. 10b).Fig. 9. Integration of scRNA-seq and spatial transcriptomics in the articular disk. a Schematic overview of the integration analysis workflow. b Conceptual diagram illustrating the integration of scRNA-seq data (17 cell types annotation and 0–23 clusters) with spatial transcriptomics data. The area of interest is manually marked (light blue) onto the Xenium spatial expression pattern images. c Inset boxes in H&E staining indicate the Xenium spatial expression pattern with annotated 17 cell types. The inset boxes in the Xenium spatial expression pattern images (middle panels) indicate enlarged images with cell-segmentations. Scale bars: 1 mm (right panels), 100 μm (middle panels), and 20 μm (right panels). d Xenium spatial expression pattern images showing macrophage-, fibroblast-, endothelial-, and keratinocyte-type cell segmentation. Scale bars: 100 μm for fibroblast-, endothelial-, and keratinocyte-type images, and 50 μm for macrophage-type images. e Integrated data showing 17 different cell types mapped onto the Xenium spatial expression pattern. The upper pie charts display the percentages of cell types in the marked area shown in (b). The 17 cell type column chart shows cell counts in the marked area across the Ctrl, MS, and ADD groups (lower panel)Fig. 10. Spatial distribution of 0–23 clusters with annotated cell-type composition in the articular disk. a Xenium spatial expression pattern images with cell segmentation assigned to the 0–23 clusters. b Xenium spatial expression pattern images with cell segmentation assigned to the 0–23 cluster and annotated cell types. Scale bars: 100 μm for fibroblast-, endothelial-, and keratinocyte-type images, and 50 μm for macrophage-type images. c Upper pie charts show the percentages of annotated 17 cell types in the marked area shown in (Fig. 9b). The 0–23 clusters column chart shows cell counts in the marked are across the Ctrl, MS, and ADD groups (lower panel)
Finally, the population rate and the number of cells within the subcellular segmentation corresponded to the classified cell clusters (0–23) in the entire articular disk, including the synovium (Fig. 10c). We found that highly differentiated cell clusters were identified and spatially distributed in the posterior synovium of the MS and ADD model mice. These integration data findings in the histological sections suggest that dynamically catabolic changes in the cell types and cell clusters occurred in the posterior synovium, consistent with the observed single-cell cluster changes.
Discussion
In the present study, we employed an integrated transcriptomic approach, combining subcellular spatial transcriptomics and single-cell RNA sequencing in the MS and ADD model mice. This approach allowed us to characterize molecular alterations in the synovium of the articular disk in response to mechanical and inflammatory stimuli. Rather than demonstrating causative roles, our data provide a spatial and cellular landscape that may help elucidate how stress-related changes in the synovial microenvironment are associated with the early stages of TMJ osteoarthritis. Through the integration of bulk RNA-seq, scRNA-seq, and spatial transcriptomic data, we delineated region-specific transcriptomic profiles, particularly in the posterior synovium, prior to cartilage degeneration. These profiles were marked by pathological features such as adipogenesis, fibrosis, and macrophage activation. Notably, Notch signaling in endothelial cell clusters was implicated as a potential mediator of stress-induced synovial responses.
We employed two mouse TMJ-OA models of mechanical stress and inflammation induced by surgical ADD to replicate the pathological conditions in human TMJ-OA. The human TMJ experiences forces equivalent to the body’s weight (i.e., 770–900 N) during biting.^7^ Our two mouse TMJ-OA models successfully replicated human TMJ-OA, showing not only morphological changes^9–14^ but also the upregulation of the Ccl5-Ccr5 chemokine-chemokine receptor axis (Supplementary Fig. S4b), consistent with elevated serum levels of CCL5 in whole TMJ disease patients^46^ and increased chemokine levels in the synovial fluid of TMJ disease patients.^47^ Consistently, Visium HD images revealed that chemokine-related molecular markers and pain-related markers were predominantly expressed in retrodiscal fat tissue (MS) and the posterior-superior thickening region of the articular disk (ADD) (Supplementary Fig. S17). To validate this phenomenon in vitro, we used human fibroblasts, which reflect the cellular composition of articular disk tissue. Upon stimulation with IL-1β to mimic inflammatory conditions, treatment with the selective CCR5 antagonist Maraviroc effectively suppressed the expression of Mmp3 marker (Supplementary Fig. S18).
Our MS model demonstrated that Angle’s Class II Division 2 malocclusion^11^ is characterized by retroclined maxillary incisors and distal positioning of the mandibular arch relative to the maxillary arch. This malocclusion causes excessive incisor guidance during mandibular closure, leading to posterior derangement of the mandible. This, in turn, results in structural and morphological changes in both soft and hard tissues of the TMJ,^9–11^ which is consistent with the synovium and condyle changes observed in our MS model. Specifically, adipogenic differentiation was observed in the posterior synovium of the MS model (Fig. 3f, g). These data suggest that excessive mechanical stress promotes adipogenesis in mesenchymal stem cells.^48,49^ Furthermore, adiponectin, an adipogenic factor, has been identified as a potential catabolic mediator that increases the production of cartilage-degrading matrix metalloproteinases and cytokines in synovial fibroblasts.^50^
Patients with TMJ disease with ADD show stronger associations with degenerative bone changes, and internal TMJ disorders significantly increase the risk of OA.^12–14^ Our mouse ADD model exhibited more severe structural and morphological changes than the MS model, as demonstrated by histological analysis results and the expression of degenerative markers in bulk RNA-seq and drastic cluster changes in scRNA-seq. Moreover, this successfully replicated the clinical conditions in human ADD. It is thought that the synovium and retrodiscal tissue, rich in blood vessels and nerves,^51^ undergo excessive stretching leading to inflammation due to forced anterior displacement of the articular disk, which eventually causes synovial degeneration.
Zang et al. conducted scRNA-seq on human TMJ-OA condylar cartilage and found that M1-like macrophages contribute to inflammation, while CD31^+^ endothelial cells promote bone mineralization.^52^ However, their study did not focus on the synovium of the articular disk. Bi et al. established an ADD model by drilling a 1 mm diameter hole anterior to the front junction of the zygomatic arch for suturing^53^ and examined condyle cartilage degeneration and subchondral bone homeostasis, but did not investigate synovial cellular changes. Our study extended beyond the condyle by analyzing gene expression profiles in both the condyle and the articular disks, including the synovium, providing a more comprehensive understanding of cellular changes.
Our spatial in situ RNA expression and scRNA-seq analyses revealed the expression of cytokine- and protease-related genes, such as Mmp3 and Mmp13, which are involved in bone resorption and cartilage degeneration,^54^ and Spp1, a driver of fibrosis and inflammation,^55^ in the posterior synovium of articular disk (Fig. 8d). This suggests that cartilage degeneration may be driven by catabolic and inflammatory factors secreted from the synovium. Bi et al ^53^. also noted that repositioning the articular disk in their ADD model alleviated cartilage degeneration, indicating that the articular disk, including the synovium, may possess mechanical sensor properties that regulate TMJ homeostasis.^56^
Demerau et al. highlighted that different arthritic joint synovial fibroblast subpopulations exhibit distinct markers and profiles. They also noted that fibroblasts transitioning to a myofibroblast-like phenotype drive fibrosis through mechanotransduction.^57^ However, the pathological mechanisms linking mechanotransduction and the TMJ synovium remain unclear. Our data indicate that the synovium and retrodiscal tissue play a crucial role in the onset of TMJ-OA by responding to mechanical and inflammatory stress, underscoring their clinical significance in the development of novel therapeutic strategies.
Considering the lining layer of human synovium, Smith et al. reported that scRNA-seq analysis of human rheumatoid arthritis synovium identified PRG4 and CLIC5 in the same cluster using UMAP.^58^ These genes were expressed in the resting lining layer, which lacked inflammatory responses and cytokine signaling based on the ATAC-seq and pathway analyses. In line with these findings, our spatial in situ RNA expression analysis demonstrated that Prg4 and Clic5 were spatially distributed in the surface layers of the posterior synovium, along with Col1a1 and Thy1 (Fig. 8c), suggesting a role in maintaining TMJ homeostasis.
Hill et al. found that Prg4 knockout mice exhibited age-related TMJ degeneration and synovial hypertrophy.^59^ We previously reported that Prg4^+^ and Rspo2^+^, which were induced by inflammatory stimulation and mechanical loading via NF-κB in the tendon stem/progenitor cell cluster, suppressed ectopic ossification and contribute to tendon/ligament homeostasis under pathogenic conditions.^60^ Notably, our scRNA-seq data indicated that Prg4^+^, Rspo2^+^, and Tspn15^+^ cells which is also known stem cell and progenitor marker^61^ were co-localized in the UMAP cluster 16 of the (Supplementary Fig. S19), suggesting that they may serve as a progenitor properties and suppressor of ectopic ossification. Furthermore, the spatial location of cluster 16 in the UMAP was identified in the superficial layer of the synovium under normal conditions, as revealed by the integrated scRNA-seq and Xenium analysis (Fig. 10a, b). However, the stress induced by MS and ADD disrupted the uniformity of these superficial synovial cells, leading to a breakdown of the environment responsible for maintaining TMJ homeostasis. Recent studies have identified a high degree of heterogeneity in synovial fibroblasts, and the concept that changes in the proportions of fibroblast subsets may underlie pathological changes in joint tissue has gained traction.^62–64^ Our data also show a diversification of fibroblasts, suggesting that functionally distinct fibroblast subsets contribute to more severe and persistent inflammatory arthritis.
In our previous report,^65^ we discussed the regulation of Mmp9 expression by lubricin, encoded by Prg4 in the superficial zone of cartilage, potentially through the Toll like receptors- or CD44-NF-κB axis. Al-Sharif et al. reported that recombinant lubricin inhibits the proliferation of fibroblast-like synoviocytes enhanced by interleukin-1β (IL-1β) or tumor necrosis factor α (TNF-α) via a CD44-mediated mechanism.^66^ Similarly, Alquraini et al. demonstrated that recombinant lubricin inhibits NF-κB activation in OA synoviocytes in a CD44-dependent manner.^67^ In the present study, the CD44-NF-κB axis in macrophage clusters (5 and 12 in Supplementary Fig. S14) may have been activated in response to mechanical stress or inflammatory stimuli.
Bi et al. performed scRNA-seq on mouse TMJ disk tissues at different postnatal developmental stages and found that the resident mural cell population serves as the source of disk progenitors.^68^ They also discovered that Myh11^+^ mural cells coordinate angiogenesis during disk injury but gradually lose their NOTCH3 and THY1 expression, as well as their progenitor characteristics, ultimately becoming fibroblasts. In our analysis, UMAP of the MS and ADD models showed that the endothelial population clusters (0, 2, 6, 14, and 18) were dominantly increased in (Fig. 7b, c, e), as also observed in the integrated analysis with Xenium (Figs. 9d, e and 10b, c). Sellam et al. reviewed that endothelial cell proliferation, macrophage infiltration and inflammation occur to a greater extent in OA patients compared to healthy controls.^4^ They also noted that angiogenesis and inflammation are closely integrated processes and may affect disease progression and pain. The Notch pathway is a key endothelial signaling pathway, and Notch1 within these endothelial clusters (0, 2, 6, 14, and 18) may play a key role as a receptor-ligand, as indicated by our scRNA-seq data (Supplementary Fig. S13a), which utilized the CellChat platform to focus on mechanoreceptor molecules. Focusing on the ligands of Notch 1 (Supplementary Fig. S13a), Mfap5 and Cxcl12 were identified as upstream factors and were expressed in the posterior synovium underlying fibroblast cluster (Supplementary Fig. S13c). NOTCH signaling in joint cartilage plays a dual role in maintenance and OA pathogenesis.^69^ Sustained NOTCH signaling in joint cartilage leads to early and progressive OA pathology,^70^ whereas transient NOTCH activation promotes cartilage matrix synthesis and joint maintenance under normal physiological conditions.^69^ To investigate how NOTCH signaling affects the TMJ disk, we conducted in vitro experiments using DAPT, a γ-secretase inhibitor that blocks NOTCH pathway activation. We used human fibroblasts, which reflect the cellular composition of articular disk tissue. The cells were stimulated with IL-1β to mimic inflammatory conditions, and treatment with DAPT successfully suppressed the expression of inflammatory markers (Supplementary Fig. S18). Furthermore, our synovial scRNA-seq data suggested that NOTCH1 may play a key role in regulating TMJ homeostasis (Supplementary Fig. S13a) and that its downstream effectors could serve as potential therapeutic targets. These findings indicate that appropriate inhibition of NOTCH signaling may represent a viable strategy for disease-modifying therapy in TMJ-OA.
This study has several limitations. First, all analyses were conducted at an early time point (3 weeks after model induction), and long-term disease progression was not evaluated. While significant TMJ degeneration in these mouse models was evident within 3 weeks, further analysis was limited due to excessive condylar hypertrophy and potential feeding difficulties. Second, the rapid progression in our mouse models differs from the gradual onset of human TMJ-OA, which develops over years with features like cartilage thinning and osteophyte formation. One possible explanation for this difference is the disparity in bite force relative to body weight. In humans, the TMJ experiences forces approximately equivalent to body weight (770–900 N),^7^ whereas in mice, we measured a bite force of approximately 1 kg (10 N),^71^ which is about 30 times their body weight (30 g). This relatively higher mechanical load on the mouse TMJ may contribute to the accelerated tissue degeneration observed within the short experimental period.
Third, we acknowledge that these mouse models do not fully replicate the complex and multifactorial nature of TMJ-OA pathogenesis in humans. TMJ-OA in humans develops over a prolonged period and is influenced by multiple interacting factors, including biomechanical stress, inflammation, genetic predisposition, and habitual or systemic conditions. In contrast, our models primarily focus on mechanical stress as a key driver of TMJ changes, which may not fully capture the progressive and multifactorial nature of the disease. The MS model mimics posterior condylar displacement due to malocclusion, leading to compressive stress on the condyle, while the ADD model replicates anterior disk displacement, which alters joint loading patterns and likely contributes to inflammation and cartilage degeneration. While these models provide insights into distinct mechanical stress conditions that may contribute to TMJ-OA, they do not account for the full spectrum of contributing factors observed in human cases.
Given these limitations, our findings should be interpreted with caution when extrapolating to human TMJ-OA. Future studies should incorporate more complex models, including longitudinal studies that better reflect the chronic nature of the disease and the interplay between mechanical, inflammatory, and degenerative processes. Expanding the scope of investigation to include additional factors such as inflammatory mediators and systemic influences will be crucial for a more comprehensive understanding of TMJ-OA pathogenesis.
Despite these limitations, our findings provide valuable insights into early TMJ degeneration in MS and ADD models, laying the groundwork for future research.
In conclusion, this study establishes a comprehensive and integrated methodological framework combining histology, bulk RNA-seq, scRNA-seq, and high-resolution spatial transcriptomics using the Xenium and Visium HD platforms to investigate the early pathogenesis of TMJ osteoarthritis. To our knowledge, this is the first study to generate a spatial and cellular map of the TMJ articular disk and surrounding synovium at single-cell and subcellular resolution, specifically during the onset of TMJ-OA. This multi-modal approach enables detailed characterization of synovial responses, cell–cell interactions, and inflammatory microenvironments in the earliest phases of disease, and can be broadly applied to other joint tissues and inflammation-related disorders. Our methodological strategy offers a valuable resource for researchers seeking to elucidate complex tissue dynamics and supports future efforts in developing early-stage interventions for TMJ-OA.
Materials and methods
Animals
All animal experiments were conducted in accordance with the guidelines described in the Guide for the Use and Care of Laboratory Animals of the Institute for Laboratory Animal Research (ILAR, 2011). Furthermore, animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Showa Medical University (224038). All animals were housed in a facility under the supervision of the IACUC at 18–22 °C, with a 12-h light/dark cycle and free access to food and water. For all in vivo studies requiring wild-type mice, 8-week-old female C57BL/6J mice were purchased from Sankyo Lab Services (Tokyo, Japan). As considering sexually mature,^72^ and the onset of severe TMJ-OA, which mainly occurs in young female.^73^
Mouse MS and ADD models
All surgeries were performed using an operating microscope under general anesthesia on 9-week-old female mice after a 1-week acclimatization period. Anesthesia was induced by intraperitoneal injection of a mixture of medetomidine (0.3 mg/kg; #0002949, Nippon Zenyaku Kogyo, Koriyama, Japan), midazolam (4.0 mg/kg; #169738, Astellas, Tokyo, Japan), and butorphanol (5.0 mg/kg; #016440, Meiji Seika Pharma, Tokyo, Japan).
For the MS model, a metal plate was attached to the posterior surface of the maxillary incisors of the mice using resin (#859-0365, B.S.A. Sakurai, Aichi, Japan), positioned at a 45° angle to the palatal plane, to induce malocclusion utilizing custom-made mouth opener fabricated using 0.02-in. stainless steel wire. The dimensions of the metal plate (#131-00, TOMY INTERNATIONAL, Tokyo, Japan) were standardized to 2.8 × 1.0 × 0.3 mm^3^.
The surgical method for the ADD model was adapted from partial discectomy of the articular disk^74^ and rat ADD model.^75^ The pre-auricular region on both sides was shaved and sterilized with 70% ethanol. A linear incision at approximately 5.5 mm was made along the zygomatic arch, exposing the joint using a blunt dissection approach. The incision extended from the temporal root of the zygomatic arch to near the lateral canthus of the eye. The frontal segment of the articular disk was identified and secured using an ophthalmic needle (#JMDN70417001, TOMITZ, Gunma, Japan) and 7-0 nylon thread (#229AFBZX00015000, Kono Seisakusho Co., Ltd, Chiba, Japan), taking care to avoid damaging the cartilage surface. The suture was tied with two knots in front of the maxillary process of the zygomatic bone, resulting in anterior derangement of the articular disk by approximately 0.5 mm from its original position.
Finally, the surgical site was rinsed with sterile saline (0.9% NaCl), and the incision was sutured with 7-0 nylon thread. Buprenorphine (NISSIN, Yamagata, Japan) was administered subcutaneously twice daily for three days postoperatively to alleviate pain. All animals were provided with minced pellets for the first week after surgery, followed by the usual hard pellets.
micro-CT analysis
TMJ specimens collected from mice were fixed overnight in 4% paraformaldehyde, and then in phosphate-buffered saline (PBS), and stored at 4 °C for 3D micro-CT analysis. The specimens were scanned using a ScanXmate-L090H instrument (Comscantecno, Co.). Three-dimensional images were reconstructed using the TRI/3D-Bon-FCS system (RATOC System Engineering Corporation, Tokyo, Japan). To examine the condylar structure, scanning was conducted at 80 kV and 81 μA, with images reconstructed at an isotropic voxel size of 8.0 μm/pixel. The ROI was selected to include the entire mandibular head within a volume of 2 500 × 1 000 × 1 000 μm³. The orientation of the mandibular head was set as shown in Fig. 2a. The skull was analyzed at 80 kV and 83 μA, with images reconstructed at an isotropic voxel size of 25.3 μm. The following parameters—BV/TV (bone volume per tissue volume), Tb.Th (trabecular bone thickness), Tb.N (trabecular number), Tb.Sp (trabecular separation), and TMD (tissue mineral density)—were measured and compared among the three groups.
Histological analyses
Mice were euthanized, and the mandibular heads were fixed in 4% paraformaldehyde in PBS (pH 7.4) at 4 °C for 24 h. Specimens were decalcified in 10% ethylenediaminetetraacetic acid (EDTA, pH 7.4) (NACALAI TESQUE, Kyoto, Japan) at 4 °C for 4 weeks, embedded in paraffin, and sectioned into 4-μm-thick sagittal slices.
For analysis, sagittal sections were selected from the central region of the mandibular condyle. Hematoxylin and eosin (H&E), Safranin O, and tartrate-resistant acid phosphatase (TRAP) staining were performed according to standard protocols. Image analysis of Safranin O and TRAP-stained sections was conducted using the BZ-X Analyzer (KEYENCE, Osaka, Japan).
To evaluate the severity of TMJ-OA, all samples were assessed using a modified Mankin score^76^ for the TMJ condyle and a synovitis score^77^ for synovial inflammation. Four independent observers, blinded to the experimental groups, evaluated the samples (n = 5), and the mean values were used for analysis.
For osteoclast differentiation assessment, TRAP-positive cells with three or more nuclei were counted as osteoclasts (n = 5). The region of interest (ROI) was defined such that its upper boundary aligned with the surface of the mandibular condyle, with a standardized area of 1 500 × 600 μm². Any regions outside the condyle were excluded from the analysis.
Safranin O-positive area, number of hypertrophic chondrocytes, fibrocartilage thickness, and hypertrophic zone thickness were quantified using Safranin O-stained sections. For Safranin O-positive area and hypertrophic chondrocyte counts, three ROIs were randomly selected in the posterior, middle, and anterior regions of each sample. The ROI size was defined as 200 × 300 μm², with the upper boundary aligned with the surface of the mandibular condyle. Within each ROI, the number of hypertrophic chondrocytes and the Safranin O-positive area were measured (Supplementary Fig. S2a).
For fibrocartilage thickness, in each posterior, middle, and anterior region, three randomly selected lines crossing the cartilage layer were measured, and the average value was calculated for each region. Hypertrophic zone thickness was measured using the same method in the same regions, followed by statistical analysis (Supplementary Fig. S2d).
Quantitative reverse transcription polymerase chain reaction (qRT-PCR)
Total RNA was extracted using the Direct-zol RNA Kit (#R2062, Zymo Research, Irvine, CA) in accordance with the manufacturer’s protocol. We confirmed that the A260/A280 was between 1.8 and 2.0 for all RNA samples. Total RNA was reverse-transcribed into cDNA using Rever TraAce qPCR RT Master Mix (#FSQ-201, TOYOBO, Osaka, Japan). qRT-PCR was then performed using the THUNDERBIRD Next SYBR qPCR Mix (#QPX-201, TOYOBO) and StepOnePlus (Applied Biosystems, Foster City, CA). Relative quantification using a standard curve method was used to compare gene expression levels. Target gene expression levels were normalized using glyceraldehyde-3-phosphate dehydrogenase (Gapdh) as an internal control for the ΔΔCT method.^78^ We prepared three biological and three technical replicates for each sample (n = 3). The primers used in this study are listed in Supplementary Table S1.
Bulk RNA-seq
We prepared total RNA samples as described above and submitted them to BGI (BGI, Hong Kong, China) for library preparation and sequencing. The cartilage of condyle and articular disk of TMJ were collected from three mouse per group after 3 weeks in each model. The purified total RNA was confirmed by the ratio of A260/A280 (1.8–2.0), and the RNA integrity number was >6. RNA-seq was conducted on a DNBSEQ-G400 (BGI, Shenzhen, China) with 100-bp paired-end reads. The fastp-filtered data were aligned against the mouse reference genome (GRCm39 GENCODE primary assembly) using STAR (version 2.7.10b).^79^ Mapping bam files were counted by RSEM (version 1.3.3). Data analysis was performed by BGI’s analysis solution (https://www.bgi.com/jp/dr-tom/) and ingenuity pathway analysis (QIAGEN, Dusseldorf, Germany).
Single-cell dissociation and RNA sequencing
To isolate articular disk cells, including the synovium of TMJ, the articular disks were freed from the joint space and precisely isolated from the surrounding tissue using fine forceps and scissors under a stereomicroscope. The harvested tissues were minced in ice-cold PBS and digested in RPMI 1640 medium (#30264-85, Nacalai Tesque Inc.) containing 2.5% Liberase TL (#05401020001, Roche, Germany) and 0.2% DNase I (#10-104-159-001, Roche, Germany) at 37 °C for 15 min with constant agitation. Cells were filtered twice through 70-, and 40-μm strainers, centrifuged, counted, and resuspended at a concentration of 1 000 cellsper μL. As a result, the total number of harvested cells was 4 700,000 for Ctrl, 6 000,000 for MS, and 5 600,000 for ADD. These cells were obtained from Ctrl: n = 7 (14 articular disks), MS: n = 7 (14 articular disks), and ADD: n = 4 (8 articular disks). Cell viability was assessed using the trypan blue exclusion method with a hemocytometer and a LUNA-FL automated counter (Logos Biosystems, Anyang, South Korea). Library samples that passed the quality check, as determined by a Bioanalyzer (G2939BA; Agilent Technologies, Santa Clara, CA), were processed for further sequencing. Single-cell libraries were prepared using the Chromium Controller (10× Genomics, Pleasanton, CA) according to the protocol of Chromium Next GEM Single Cell 3’ Reagent Kits (v3.1 10× Genomics). Sequencing was performed by Novogene (Tokyo, Japan) using the Novaseq X Plus (Illumina, San Diego, CA, USA) system.
Bioinformatics analysis of single-cell sequencing data
Raw fastq files were mapped using the STARsolo^80^ pipeline (ver. 2.7.10b) primarily with Python 3.10.1162 onto the GENCODE primary assembly mouse genome GRCm39 (mm39) and basic gene annotation.^81^ Downstream analysis was performed mainly using Python 3.10 and the Scanpy package^41^ as described previously.^82^ The Python package “LIANA” was used to process the ligand-receptor (L-R) database obtained using mouse consensus pairs from CellPhoneDB and CellChat. The results were analyzed on R 4.4.1 to identify the top 1000 L-R relationships by each receptor cluster, sorted by composed genes of terms on mSigDB 2023.2 edition including “Mechanoreceptor” and “Cytokine receptor”, then visualized using circlize.
Preparation of spatial transcriptomics using Xenium spatial gene expression
In situ RNA expression analysis at a single-cell level was performed using the Xenium platform (10× Genomics). Decalcified TMJ tissues embedded in paraffin blocks were sectioned at 5-µm thickness using a microtome (REM-710, YAMATO KOHKI INDUSTRIAL CO., LTD, Saitama, Japan) with a MASAMUNE blade holder (BH-220, YAMATO KOHKI INDUSTRIAL CO., LTD) for decalcified hard tissues. The sections were then prepared on Xenium slides (PN 3000941, 10× Genomics) following the manufacturer’s “Tissue Preparation Guide CG0000578”. The Probe Hybridization Mix was prepared using a pre-designed panel (Xenium Mouse Gene Expression panel, mMulti_v1) and custom panel (Xenium Custom Gene Expression panel, design ID: KEPGAC) designed with the Xenium Panel Designer (10x Genomics) based on the out scRNA-seq data. The list of 100 target genes in the custom panel and their population rates among the Ctrl, MS, and ADD groups are provided in Supplementary Fig. S14. Three samples each from the Ctrl, MS, and ADD models were processed for Xenium analysis, with support from Department of Computational Biology and Medical Sciences, the University of Tokyo (JP22H04925, PAGS).
Xenium data acquisition
Probes were hybridized following the manufacturer’s “Demonstrated_Protocol_Xenium_FFPE_Deparaffinization_Decrosslinking CG000580” and “XeniumInSitu_GeneExpression_UserGuide CG000582”. Fluorescent probe hybridization and imaging were conducted using the Xenium Analyzer (on-board analysis: version 1.8.2.1, software: version 2.0.01.7.1.0, 10× Genomics). Output images and expression profiles were evaluated using Xenium Explorer (version 2.0.0, 10× Genomics). After the Xenium run, H&E staining was performed on the Xenium slide. For quality control, the output summary HTML file was reviewed, and the ROI was selected using Xenium Explorer (Fig. 8b).
Xenium differential gene expression (DGE) analysis
We delineated the ROI in the posterior synovium of the articular disks (Fig. 8b) and performed DGE analysis by counting transcripts within the ROI using Xenium Explorer software (version 2.0.0, 10× Genomics).
Integration of scRNA-seq data with spatial transcriptomics and visualization
To integrate the scRNA-seq dataset of the articular disks with the spatial transcriptomics dataset,^45^ we used the computational framework for TACCO.^44^ Label transfer was performed using TACCO for 17 cell types annotated by CellAssign and 24 cell clusters identified by the Leiden algorithm. The cell types and cluster numbers of all segmented cells were annotated and visualized in Xenium Explorer (Fig. 9a).
Statistical analysis
Results were analyzed using GraphPad Prism version 10.2.3 (GraphPad Software, San Diego, CA, USA). The two-tailed Mann–Whitney U test was used to determine statistical significance. One-way ANOVA with Dunnett’s post hoc test was used for comparisons involving more than three groups or time-course comparisons of multiple groups. P values below 0.05 were considered statistically significant. The number of biologically independent samples is indicated in the figures and figure legends.
Supplementary information
Supplementary figures, table & methods
The reference list from the paper itself. Each links out to its DOI / PubMed record.