Identification of synovial lymphatic system in the temporomandibular joint and their roles in arthritis and pain

Abstract

Temporomandibular joint (TMJ) arthritis is a craniofacial disorder characterized by joint dysfunction and orofacial pain. Lymphatic regulation and function in TMJ remain unknown. Using genetic reporter mice, human tissues, tissue clearing, 3D volume imaging, and functional studies, we identified a synovial lymphatic system in TMJ. In a mouse model of TMJ arthritis, inflammation induces extensive lymphatic remodeling and leads to synovial lymphatic dysfunctions. Functional genetics and single-cell RNA sequencing (scRNA-seq) revealed that lymphatic deficiency induces a population of fibroblast-macrophage hybrid cells and enhances inflammation, exacerbating cartilage defects, bone loss, synovitis, and pain behaviors in TMJ arthritis mice. Conversely, lymphatic function promotion via a hydrogel-mediated VEGF-C delivery prevents TMJ pain, inflammation, and arthritis-like pathogenesis. Thus, we identified synovial lymphatics in TMJ and found that lymphatic dysfunction drives TMJ arthritis and pain, suggesting its potential as a therapeutic target.

Here, the authors show that the temporomandibular joint contains a previously unrecognized synovial lymphatic network. Disrupted lymphatics in inflammatory arthritis amplify synovitis, cartilage degeneration, bone loss, and pain behaviors, whereas enhancing lymphatic activity restores drainage, reduces inflammation, and alleviates pain.

Introduction

Temporomandibular disorders (TMDs) are a subset of painful craniofacial disorders that involve the TMJ, masticatory muscles, and surrounding tissues¹. TMDs are estimated to affect approximately 5% to 10% of the U.S. population (https://www.nidcr.nih.gov/news-events/nidcr-news/2023/looking-forward-greater-impact-temporomandibular-disorders). Around 15% of TMD cases continuously progress, show resistance to treatment, and result in chronic pain, making it the second most common musculoskeletal pain condition that are recapitulated in various preclinical models^2,3. TMJ arthritis and TMJ osteoarthritis (TMJOA) represent joint-specific forms of TMDs, characterized by joint inflammation, dysfunction, and degeneration. Both conditions significantly affect quality of life due to arthralgia and restricted mobility^4,5. Patients with TMJOA usually have a history of arthralgia and a popping noise while performing jaw functions^6–8. Individuals with arthritis experience severe and usually unremitting pain that can be both physically disabling and emotionally distressing, hampering their potential recovery⁹. While pain is the major reason for patients to pursue medical treatment, the underlying mechanisms and effective management strategies of TMJ pain are still unclear^10–12.

The lymphatic system is a network of vessels that is present in virtually every organ of the body, complementary to the cardiovascular system. It is composed of blind-ended, highly branched capillaries that function to take up interstitial fluid (ISF), large molecules, and leukocytes from peripheral tissues^13–15. These capillaries serve as entrance points into lymphatic circulation, playing a crucial role in preventing the accumulation of fluids and immune cells within local tissue^13,14. Lymph is transported through collecting lymphatic vessels to draining lymph nodes, and is returned into blood circulation via major lymphatic ducts, such as the thoracic duct and the right lymphatic duct¹⁶. Lymphatic vessels are lined with a single layer of lymphatic endothelial cells (LECs), which express the homeobox transcription factor prospero-related homeobox 1 (Prox1) and receptor tyrosine kinase vascular endothelial growth factor (Vegf) receptor 3 (Vegfr3)¹³. Prox1 is a master regulator of lymphangiogenesis, and its heterozygous mice have been used as a genetic model of lymphatic vessel malfunction^17,18. The Vegf-c ligand interacts with its receptor Vegfr3 to regulate lymphatic growth and function, which are promoted or inhibited by ectopic expression of Vegf-c or Vegf-c/d trap, respectively^14,19.

In collecting lymphatic vessels, endothelial cells exhibit an elongated morphology and are connected by continuous junctions known as “zipper junctions.” These vessels are enveloped by a thick basement membrane and layers of contractile lymphatic smooth muscle cells (LSMCs), which limit permeability and support unidirectional lymph flow with the help of intraluminal valves²⁰. Capillary lymphatic vessels are composed of LECs and a thin basement membrane, lacking mural cell coverage. These cells are connected by discontinuous, button-like junctions that facilitate high permeability, allowing efficient passage of solutes and immune cells into the lymphatic system^21,22. Lymphangiogenesis and lymphatic vessel remodeling are frequently observed in inflammation-related biological processes and diseases. For example, the immune-interacting subtype of Ptx3-positive LECs recruits pro-lymphangiogenic macrophages to promote progressive lymphatic overgrowth²³. The pathophysiological functions of the lymphatic system in inflammatory disorders such as TMJ arthritis remain poorly understood.

Previous studies suggest that impaired lymphatic functions are involved in the pathogenesis of inflammatory knee joint diseases, such as osteoarthritis (OA) and rheumatoid arthritis (RA)^24–26. Targeting lymphatic vessels to restore drainage, enhance lymphangiogenesis, or resolve inflammation is proposed as a therapeutic strategy for the management of knee joint arthritis^24,27. However, TMJ differs from the knee joint in terms of developmental origin, morphology, structural composition, and function^28–30. It remains unknown if and where lymphatic vessels occur and how they are regulated and function in the TMJ under pathophysiological conditions. In this study, we identified synovial lymphatics in the TMJ. In TMJ arthritis mouse models, lymphatic vessels undergo extensive lymphangiogenesis, coupled with inflammation and drainage dysfunctions. Functional studies revealed that lymphatic dysregulation drives TMJ arthritis and pain, highlighting its potential as a therapeutic target.

Results

TMJ lymphatic vessels are identified at synovial tissues

Previous studies reported the distribution and alteration of lymphatic vessels in the knee joints of normal and osteoarthritic mice³¹. Here, we sought to identify and investigate the distribution of lymphatic vessels in the TMJ. Toward this aim, we used Prox1-mScarlet mice in combination with antibodies against lymphatic vessel marker lymphatic vessel endothelial hyaluronan receptor 1 (Lyve1). An immunolabeling-enabled iDISCO tissue clearing method was applied to adult mouse TMJs, followed by confocal imaging and 3D reconstruction to generate whole mount images of the TMJ (Fig. 1a). High magnification images showed that lymphatic vessels co-express Prox1 and Lyve1, displaying tube-like structures, confirming the lymphatic vessel identities. Moreover, lymphatic vessels were mainly found in the surrounding muscle, anterior, posterior, and superior synovial tissues, as well as the superior retro-discal tissue (RDT) of the TMJ (Fig. 1b–d). No-primary antibody, secondary-only control, and autofluorescence images were acquired to rule out background signal and channel bleed-through (Fig. S1). It has been reported that tissue-resident macrophages regulate lymphatic vessel growth and patterning in the developing heart³². To investigate the lymphatic vessel-macrophage axis, we used Cx3cr1-YFP mice together with Lyve1 or Iba1 to label resident macrophages as well as Vegfr3 to mark vascular endothelial cells, followed by iDISCO tissue clearance and confocal imaging of the whole adult mouse TMJ (Fig. 1e). There are extensive Vegfr3-positive and tube-like vessels at the synovial tissues and surrounding muscles, where Cx3cr1⁺;Lyve1⁺ or Lyve1⁺;Iba1⁺ double-positive resident macrophages are adjacent to vessel cells (Figs. 1e, S2, S3k–m and Supplementary Movie 1).

Fig. 1. Identification of lymphatic vessels in TMJ.

a Confocal imaging of cleared TMJ stained with antibodies against Lyve1 (green) from adult Prox1-mScarlet mice. Scale bar: 300 µm. b–d High magnification images of the lymphatic vessels in the representative and white boxed regions. Scale bars: 100 µm. e Confocal imaging of tissue cleared TMJ stained with Vegfr3 and YFP from adult Cx3cr1-YFP mice. Scale bar: 200 µm. f Immunofluorescence imaging of human TMJ synovial sections stained with antibodies against LYVE1 (green). DAPI stains nuclei (blue). Scale bars: 100 and 10 µm. g Confocal imaging of the 100 µm TMJ section after CUBIC clearing, stained with antibodies against Lyve1 and Vegfr3 from WT mice at 3 weeks post-CFA injection. White arrowheads point to Vegfr3⁺;Lyve1⁺ lymphatic vessels in anterior and posterior synovial tissues. Scale bar: 200 µm.

Whether or not lymphatic vessels occur in the cortical bone or bone marrow is an unsettled issue. Vegfr3 could label blood or lymphatic vessels in a tissue-dependent manner. To achieve higher resolution imaging and investigate whether Vegfr3-positive vessels in condyle bone are blood or lymphatic vessels, we performed CUBIC (clear, unobstructed brain/body imaging cocktails and computational analysis) tissue clearance of 100 µm-thick TMJ sections from mice exposed to Complete Freund’s Adjuvant (CFA), described later as our TMJ arthritis models. Lyve1- and Vegfr3-double positive lymphatic vessels are mainly located in the anterior and posterior synovial tissues (Fig. 1g, white arrowheads), where Vegfr3⁺;Lyve1⁻ blood vessels occur less frequently than lymphatic vessels (Supplementary Movie 2). It is clear that Vegfr3-positive vessels in condyle bones are Lyve1 negative, suggesting the absence of lymphatic vessels in condyle bone or bone marrow. This is consistent with the absence of lymphatic vessels inside the skull bone marrow³³. Lastly, human TMJ synovial tissues from the arthroscopic surgery of TMJ pain patients confirmed LYVE1-positive tube-like lymphatic vessels (Fig. 1f). Together, Lyve1⁺;Vegfr3⁺ double-positive and tube-like structures are identified as lymphatic vessels, which are mainly located in synovial tissues under TMJ healthy and arthritis pathological conditions.

Lymphangiogenesis and lymphatic vessel remodeling are coupled with inflammation in mouse arthritic TMJ

To investigate lymphatic vessels under TMJ pathological conditions, we turned our attention to animal models of TMDs. Complete Freund’s Adjuvant (CFA) is an inflammatory agent that causes joint inflammation, a pathological trigger for painful TMJ dysfunction. We utilized our previously established mouse models of CFA-induced inflammatory TMJ arthritis, which recapitulate  TMD pathologies and orofacial pain behavior in patients². CFA intra-articular injection was performed to induce TMJ arthritis. We conducted spatial transcriptomic analysis of control and CFA-treated TMJs at single-cell levels using Sequential Fluorescence In Situ Hybridization (seqFISH)³⁴, as described in a separate study. Using our seqFISH data, we identified the synovial tissue-enriched location of lymphatic vessels labeled by Vegfr3 and Lyve1 double-positive RNAscope signals, which were increased in the CFA compared to the control TMJs (Fig. 2a). Ptx3 was used to mark a subpopulation of inflammatory LECs, characterized by their ability of recruiting pro-lymphangiogenic macrophages to promote lymphatic overgrowth^23,35. Therefore, we examined our seqFISH data for Ptx3 and found its upregulation in conjunction with Vegfr3 and Lyve1 in the synovial tissue regions in CFA-treated TMJs (Fig. 2b). Next, we examined protein expression in TMJ with joint inflammation, as evidenced by increased macrophage activation marker CD68 and pan-macrophage marker Iba1 in arthritic TMJ (Fig. 2c, e). IHC staining confirmed Ptx3 protein upregulation in tube-like lymphatic vessels in synovial tissues of arthritic TMJ, whereas no Ptx3 signals were detected in control synovial tissues (Fig. 2d, f).

Fig. 2. Lymphangiogenesis and lymphatic vessel remodeling are coupled with inflammation in arthritic TMJ.

a, b seqFISH images of TMJs stained with RNA probes Vegfr3 (red), Lyve1 (green), and Ptx3 (blue or gray). Scale bar: 300 µm. c, d Immunofluorescence imaging of the superior regions of TMJ sagittal sections stained with antibodies against Iba1 (white) and CD68 (red), as well as Lyve1 (green) and Ptx3 (red). DAPI stains nuclei (blue). White arrowheads denote Iba1 and CD68 double-positive regions. Scale bars: 50 µm. e, f Quantification of the area fraction of Iba1⁺;CD68⁺ double-positive macrophages or Lyve1⁺;Ptx3⁺ double-positive inflammatory lymphatic vessels in the superior area. n = 3–4 mice per group, and P-values were calculated by two-tailed unpaired t-test in (e); n = 4 mice per group, and P-values were calculated by two-tailed Mann-Whitney test in (f). g–i Confocal imaging of TMJ sagittal sections stained with antibodies against Lyve1 (green) and Vegfr3 (red). DAPI stains nuclei (blue). Scale bars: 100 µm. Images in (i) are enlargements of dashed box regions of TMJ in (g, h) at the anterior, posterior, and superior regions. White arrowheads denote elongated and tube-like lymphatic vessels. Scale bars: 50 µm. j–l Quantification of the area fraction of Lyve1⁺ and Vegfr3⁺ lymphatic vessels in the different TMJ areas. n = 3–4 biological replicates per group in (j); n = 3–5 biological replicates per group in (k); n = 3–4 biological replicates per group in (l). Each circle dot represents one mouse. P-values were calculated by a two-tailed unpaired t-test in (j–l). P-values are indicated in the figure. All data are represented as mean ± SEM. Source data are provided as a Source Data file.

To confirm the lymphangiogenesis phenotype, we performed side-by-side IHC staining using antibodies against Vegfr3 and Lyve1 on control and CFA TMJ sections (Fig. 2g, h). H&E (hematoxylin and eosin) staining was used to define TMJ architecture, and IHC staining on adjacent sections confirmed lymphatic vessels in the synovial tissues of TMJ (Fig. S4). There was a substantial increase in lymphatic vessels at the anterior, posterior, and superior synovial regions of the TMJ in CFA-treated TMJs compared to controls (Figs. 2i–l, S4). Compared to the control, CFA-treated TMJs showed increased lymphatic vessels at 1 week, peaking at 2 weeks and declining at 3 weeks post-CFA injection (Fig. S3a–f), suggesting dynamic lymphatic vessel remodeling. In addition to increased immune-reactive areas, CFA-treated TMJ synovial tissues have more and longer tube-like vessels stained by Vegfr3 and Lyve1 (Fig. S3g–j), which is consistent with our CUBIC cleared 100 µm-thick TMJ sections (Fig. 1g). Together, these studies revealed inflammatory lymphangiogenesis and lymphatic vessel remodeling in mouse models of TMJ arthritis.

Lymphatic functions are impaired in mouse arthritic TMJ

Inflammation-associated lymphangiogenesis is a double-edged sword and is actively involved in the pathophysiology of various inflammatory disorders³⁶. To investigate how lymphatic vessel remodeling is translated into lymphatic function, we examined the lymphatic drainage functions of TMJ synovial tissues. About one week following CFA administration, 10 uL of Alexa Fluor 647-conjugated ovalbumin (OVA-647) protein was injected into the superior area of the TMJ, followed by 1 h of waiting before sample collection (Fig. 3a), which allows dye drainage into the accessory mandibular lymph nodes. To confirm that OVA-647 dye can be absorbed by lymphatic vessels, we performed IHC staining on 100 µm-thick TMJ sections using the CUBIC tissue clearing method. Confocal imaging showed the co-localization of OVA-647 dye with lymphatic vessels labeled by Vegfr3 and Lyve1 in the synovial tissues of the TMJ (Fig. 3b). Next, we evaluated how the pathological condition affects lymphatic drainage capacity. Whole accessory mandibular lymph nodes were harvested and subjected to CUBIC tissue clearing. Confocal imaging and reconstruction results showed that CFA-treated mice exhibited reduced dye drainage to the mandibular lymph nodes compared to control mice (Figs. 3d, e, S3n–p, Supplementary Movie 3). Longitudinal analyses showed the reduced OVA-647 drainage in TMJ arthritis mice at different time points after CFA injection (Fig. 3k–m), suggesting impaired TMJ lymphatic functions.

Fig. 3. Lymphatic functions are impaired in mouse arthritic TMJ.

a Diagram of TMJ lymphatic drainage experiment. b, c Confocal imaging of the sagittal section (Z-stack of 100 µm) of adult mouse TMJ stained with antibodies against Lyve1 (green) and Vegfr3 (red) after intra-articular injection of OVA-647 (Cyan). Scale bars: 100 and 50 µm. High-magnification images in c represent the lymphatic absorption of OVA-647 proteins. d Immunofluorescence imaging of sections of the accessory mandibular lymph nodes with the drainage of OVA-647. Scale bar: 100 µm. e Quantification of the volume fraction of OVA-647 dyes in the lymph nodes. n = 4–7 biological replicates per group in (e). P-values were calculated by a two-tailed unpaired t-test in (e). Each circle dot represents one mouse. f–i Immunofluorescence imaging of sagittal TMJ sections stained with antibodies against Lyve1 (green) and Vegfr3 (red) after OVA-647 (white) dye injection into the superior region of TMJ. DAPI stains nuclei (blue). Scale bars: 50 µm. j Quantification of the area fraction of OVA-647 dyes in the superior region of TMJ. n = 3–5 biological replicates per group in (j). P-values were calculated by two-sided ordinary one-way ANOVA with Tukey post hoc tests in (j). Each circle dot represents one mouse. k, l Confocal 3D volume imaging of OVA-647 in lymph nodes over time. Scale bar: 200 µm. n, o Longitudinal in vivo imaging of ICG (Indocyanine Green) in TMJ at 0, 5, 30, 60, and 120 min post-injection. m Quantification of volume fraction of OVA-647 signals within lymph nodes over time. n = 4 mice (Control), n = 4 mice (CFA). p Quantification of relative ICG signal reduction in the TMJ region over time. n = 6 mice (Control), n = 7 mice (CFA). P-values were calculated by a two-tailed unpaired t-test in (m, p). P-values are indicated in the figure. All data are represented as mean ± SEM. Source data are provided as a Source Data file.

The impaired lymphatic drainage function in CFA mice could be due to the disruption of fluid influx (entering synovial capillary lymphatic vessels), fluid efflux (flowing out of capillary synovial lymphatic vessels), or fluid clearance (collecting vessel-mediated flow out from synovial tissues and into lymph nodes). To provide more insights into impaired lymphatic drainage in CFA-treated TMJ, we performed IHC staining of TMJ sections at 5 min and 1 h after dye injection into synovial tissues. There was no significant difference in OVA-647 dye intensity at lymphatic vessel-enriched regions between control and CFA groups at 5 min post-injection (Fig. 3f, g, j). In contrast, there was a significant increase of dye retention in the CFA group compared to controls at 1 h post-injection (Fig. 3h–j). This OVA-647 dye retention could be due to impaired lymphatic efflux or immune retention, given its relatively large size ( ~ 46 kDa). To further assess lymphatic drainage kinetics, we used Indocyanine Green (ICG, 0.77 kDa), a small organic dye, which diffuses rapidly through the interstitial space and is cleared quickly by lymphatic vessels. Longitudinal analyses of In Vivo Imaging System-ICG (IVIS-ICG) showed the increased ICG dye retention in the arthritic TMJ at different time points after CFA injection (Fig. 3n–p). Together with LEC inflammation, these studies suggest that lymphatic vessel remodeling is coupled with defective lymphatic drainage functions in TMJ arthritis mouse models.

Lymphatic deficiency associates with synovial inflammation and pain behavior in TMJ arthritis mice

CFA-induced TMJ arthritis mice exhibited orofacial pain, synovitis, cartilage remodeling, and bone loss², which recapitulate TMD patient phenotypes. To investigate the biological significance of lymphatic remodeling and drainage disruption in TMJ arthritis, we performed functional perturbation of the lymphatic system using Prox1^+/− heterozygous mice. Prox1 is a key transcriptional factor regulating lymphangiogenesis. Despite the embryonic lethality of its homozygous knockout mice, Prox1^+/− mice grow normally and have been used as a genetic tool for lymphatic vessel malfunctions^17,18. Decreased LECs were confirmed in TMJ synovial tissues of Prox1^+/− mice after CFA treatment (Fig. S5a–d). Bite force and von Frey filament assays were used to measure the nociceptive pain behavior in mouse models of TMDs³⁷ (Fig. 4a, b). There was no difference in bite force or head withdrawal thresholds between wild-type (WT) and Prox1^+/− mutant mice (Fig. 4c, d), suggesting that lymphatic vessel malfunction in Prox1^+/− mutant mice per se did not cause orofacial pain. Under CFA-induced inflammatory conditions, longitudinal quantification showed that mutant mice had a more significant reduction in bite force compared to control mice over several days after CFA injection (Fig. 4a, c), suggesting that lymphatic deficiency exacerbates TMJOA pain. Consistent with the bite force assay, von Frey filament longitudinal quantification showed that mutant mice had a more significant reduction in head withdrawal threshold compared to control mice under joint arthritis conditions (Fig. 4b, d).

Fig. 4. Lymphatic deficiency exacerbates synovial inflammation and pain behavior in TMJ arthritis mice.

a, b Diagram of bite force and von Frey filament measurement, created in BioRender. Jariyasakulroj, S. (2026) https://BioRender.com/m52v381; https://BioRender.com/k99e389. c Quantification of relative bite force values. n = 6 mice (WT), n = 7 mice (Prox1^+/− Mutant), n = 12 (WT-CFA), n = 8 (Prox1^+/− Mutant-CFA). d Quantification of head withdrawal threshold measurement of the same cohort of mice. n = 6 mice (WT), n = 7 mice (Prox1^+/− Mutant), n = 12 (WT-CFA), n = 8 (Prox1^+/− Mutant-CFA). P-values were calculated by two-sided ordinary one-way ANOVA with Tukey post hoc tests in (c, d). e, f Combined RNAscope of IL-1β (red) and immunofluorescence staining of Ly6b (white) and Iba1 (green) in different regions surrounding TMJ. DAPI stains nuclei (blue). Scale bars: 100 µm. Images in (f) are enlargements of dashed box regions in e. Scale bars: 50 µm. g–i Quantification of the area fraction of IL-1β⁺;Iba1⁺ macrophages as well as IL-1β⁺;Ly6b⁺ neutrophils at anterior, posterior, and superior regions of TMJ. n = 4 biological replicates per group in (g–i). P-values were calculated by two-tailed unpaired t-test and two-tailed Welch’s t-test in (g, h); P-values were calculated by two-tailed unpaired t-test in (i). Each circle dot represents one mouse. P-values are indicated in the figure. All data are represented as mean ± SEM. Source data are provided as a Source Data file.

To investigate mechanisms underlying pain behavior, we examined inflammation of the TMJ using anterior, posterior, and superior regions as representative areas. We performed RNAscope analysis of inflammatory cytokine IL-1β, in conjunction with IHC staining using antibodies against Iba1 to label macrophages and Ly6b to label neutrophils, respectively. There were more IL-1β⁺;Iba1⁺ double-positive inflammatory macrophages in mutant TMJs compared to WT controls under the same CFA treatment (Figs. 4e–i, S5e–j). Perivascular macrophages were detected and increased in CFA-treated TMJs (Fig. S6). Similarly, IL-1β⁺;Ly6b⁺ neutrophils were increased in mutant TMJs compared to WT controls after CFA treatment (Figs. 4e–i). Our scRNA-seq (described below) analyses also showed increased IL-1β⁺;Iba1⁺ double-positive inflammatory macrophages and IL-1β⁺;Ly6b⁺ neutrophils in CFA TMJ compared to controls (Fig. S5k–o), which is consistent with our immunostaining results. These results suggest that lymphatic deficiency is associated with more inflammation and orofacial pain in TMJ arthritis mouse models.

scRNA-Seq analysis reveals fibroblast-macrophage hybrid cell expansion in lymphatics-deficient TMJ arthritis

To investigate cellular and molecular mechanisms underlying the severe inflammation of lymphatic-deficient TMJs, we performed scRNA-seq analysis to systematically examine major cell types in the TMJ. The TMJ samples with post-intra-articular inoculation of CFA were taken from 2-month-old female WT or Prox1^+/− mutant mice (Fig. 5a). Using Seurat 3 R-Package, we obtained ~7000–8000 cells with a median of ~3000–4000 genes per cell. After unbiased clustering of gene profiles, cell types were identified based on the top differentially expressed genes (DEGs) and known cell-type marker genes. These analyses resulted in 13 primary cell types (Fig. S7a), including macrophages (C1qa, C1qb, C1qc), endothelial cells (Flt1, Kdr, Pecam1), mesenchymal cells (Dcn, Col3a1, Igfbp6), erythrocytes (Hba-a1, Car2, Slc4a1), smooth muscle cells (MSC, Myh11, Acta2, Tpm2), neutrophils (S100a8, Retnlg, Mmp8), Schwann cells (Plp1, Mbp, Mpz), B cells (Cd79a, Vpreb3, Ighm), T cells (Icos, Skap1, Itk), adipocytes (Adipoq, Car3, Cfd), muscle cells (Pax7, Dmd, Lama2), chondrocytes (Col2a1, Acan, Col1a1), and osteocytes (Bglap, Runx2). These cell clusters were relatively well-separated (Fig. 5b), indicative of the high integrity of our scRNA-seq data.

Fig. 5. scRNA-Seq analysis reveals fibroblast-macrophage (FM) cell expansion in lymphatic deficient TMJ arthritis.

a Schematic diagram of single-cell analysis, created in BioRender. Jariyasakulroj, S. (2026) https://BioRender.com/k014qgq. b UMAP visualization of major cell types highlighted with different colors in TMJ. The black arrowhead indicates an increased macrophage population in the mutant TMJ. c Proportion of each cell cluster in the mutant group vs. control. d UMAP plot of macrophage-like immune cells reveals four subclusters, including M1 macrophages, Fibro-macrophages, M2 macrophages, and monocytes. e UMAP plot of macrophage-like immune cells for both control (red) and mutant (blue). f Volcano plot analysis of up- and down-regulated genes in fibro-macrophages compared to M1 macrophages. Differentially expressed genes between clusters were identified using Seurat FindMarkers with a two-sided bimodal likelihood-ratio test. P-values were adjusted for multiple comparisons using Bonferroni correction. g Heatmap of signature genes in different macrophage cell clusters. h Feature plots showing the expression of C1qa (Macrophages), Il-1β (Inflammatory cytokines), and Col1a1 (ECM), whose co-expression indicates the inflammatory and fibrosis-related fibro-macrophage expansion in lymphatic-deficient TMJ. i Brightfield imaging of sagittal sections stained with Masson’s Trichrome Staining in the anterior and posterior regions of TMJ. Scale bar: 50 µm. j, k Quantification of the area fraction of collagen deposits in the superior area of TMJ. n = 5 biological replicates per group in (j), n = 4–5 biological replicates per group in (k). P-values were calculated by a two-tailed unpaired t-test in (j); P-values were calculated by a two-tailed Mann-Whitney test in (k). Each circle dot represents one mouse. P-values are indicated in the figure. All data are represented as mean ± SEM. Source data are provided as a Source Data file.

Compared to WT controls, mutant TMJs appear to have an increased proportion of macrophages (Fig. 5b, c), which is consistent with enhanced inflammation in mutant synovial tissues (Fig. 4e–i). To have a more in-depth analysis, we re-clustered the macrophage population and identified four sub-clusters, including monocytes (Ccr2, Ly6c2, Ctsg), M1-like inflammatory macrophages (Nos2/iNOS, Mmp12, Spp1), M2-like anti-inflammatory and tissue-repairing macrophages (Mrc1/CD206, Retnlg, Ccl8), as well as a new population of fibroblast-macrophage hybrid (FM) cells with both macrophage and fibroblast identities (Fig. 5d, g), which have been described in other pathological conditions^38,39. These FMs account for most of the increased macrophage population in mutant compared to control TMJs (Fig. 5e) and exhibit a gene program of pro-fibrosis status, extracellular matrix (ECM) generation, and tissue remodeling, as evidenced by the upregulation of Col1a1, Col1a2, Sparc, and Acta2 (Fig. 5f, g). This FM cluster expresses typical macrophage marker C1qa, is inflammatory (Il-1b), and has an increased expression of fibroblast genes Dcn, Igfbp6, Col3a1 as well as ECM gene Col1a1 (Fig. 5h). Fibrosis, characterized by excessive deposition of ECM components such as collagen, leads to tissue stiffening and impaired joint function, ultimately resulting in ankylosis, one of the end-stage phenotypes of TMJOA⁴⁰. To confirm the pro-fibrosis status of the increased fibroblast-macrophage cluster, we performed Masson’s Trichrome Staining to characterize collagen deposition and reorganization. Under the same CFA-induced joint arthritis condition, mutant synovial tissues contained high levels of cross-linked collagen along with tissue reorganization when compared to controls (Fig. 5i–k). Our scRNA-seq analysis revealed the expansion of a new fibroblast-macrophage hybrid subpopulation, which might contribute to synovial tissue inflammation, stiffness, and lymphatic vessel remodeling in the mutant TMJ.

Lymphatic deficiency exacerbates TMJ arthritis defects in mice

Increased pain and inflammation prompted us to examine TMJ integrity in Prox1^+/− lymphatic-deficient mice. To this end, we focused on previously characterized CFA-induced inflammatory TMD phenotypes, including cartilage remodeling, bone loss, synovitis, and increased OA score². H&E staining was performed on joint paraffin sections at 2 weeks after CFA injection into WT control or Prox1^+/− mutant TMJs. As shown in the TMJ anterior regions, mutant mice had more severe synovial membrane hyperplasia (marked by red arrows) compared to control mice (Fig. 6a, c). Focusing on the posterior region, we found that mutant mice exhibited dense inflammatory infiltrates, with some areas forming large follicle-like aggregates (Fig. 6b). In contrast, control TMJs displayed less immune cell infiltration, as well as fewer and smaller follicle-like structures in the synovial tissues (Fig. 6b, d).

Fig. 6. Lymphatic deficiency exacerbates TMJ arthritis defects.

a, b H&E staining of sagittal TMJ sections after the intra-articular injection of CFA into WT and Prox1^+/− mutant mice. Red arrows in image a denote hyperplastic epithelial lining in the anterior area of TMJ. Red arrowheads in image b denote immune cell infiltration in the posterior area of TMJ. Scale bars: 50 µm. c, d Quantification of TMJ synovitis evaluated by Synovitis Scoring System with two assessment criteria, including synovial hyperplasia (c) and immune cell infiltration (d). n = 11–13 biological replicates per group in (c, d). P-values were calculated by a two-tailed Mann-Whitney test in (c, d). Each circle dot represents one mouse. e, h Immunofluorescence imaging of sagittal TMJ sections stained with antibodies against Runx2 or Cathepsin K. DAPI stains nuclei (blue). Scale bars: 50 µm. f, j Quantification of the area fraction of Runx2⁺ cells or Cathepsin K⁺ cells in the TMJ condyle. n = 3 biological replicates per group in (f), n = 3–5 biological replicates per group in (j). P-values were calculated by a two-tailed unpaired t-test (f) and a two-tailed Mann-Whitney test in (j). Each circle dot represents one mouse. g H&E staining and Safranin-O staining of sagittal TMJ sections. Scale bars: 50 µm. i Quantification of Osteoarthritis Research Society International (OARSI) score in TMJs. n = 10-13 biological replicates per group in (i). P-values were calculated by two-sided ordinary one-way ANOVA with Tukey post hoc tests in (i). Each circle dot represents one mouse. P-values are indicated in the figure. All data are represented as mean ± SEM. Source data are provided as a Source Data file.

To examine the cartilage integrity, we performed H&E staining and found a relatively thinner and disorganized layer of fibrocartilage in mutant mice compared to controls (Fig. 6g). Consistently, Safranin-O staining showed more severe cartilage degeneration-like defects in mutant than control TMJs (Fig. 6g). We employed the Osteoarthritis Research Society International (OARSI) scoring system and found an increased OARSI score in mutant TMJs (Fig. 6i). These results suggest that lymphatic deficiency exacerbates the cartilage defects in arthritic TMJs. Next, we examined osteoblast and osteoclast phenotypes in the TMJ. Phenotypically unstable osteoarthritic chondrocytes pathologically express osteoblast genes such as Runx2, which indicates the aberrant cartilage remodeling⁴¹. There is a significant increase in Runx2-positive cells across the fibrocartilage and bone marrow regions in mutant compared to control TMJs (Fig. 6e, f). In addition, Cathepsin K-positive osteoclasts were robustly increased in mutant bone marrow compared to controls (Fig. 6h, j), suggesting the increased osteoclast activity as the potential underlying mechanism of bone loss in lymphatic-deficient TMJs. Therefore, lymphatic deficiency leads to more severe synovitis and TMJ degeneration under arthritic conditions, establishing the causative relationship between lymphatic dysfunction and TMJ arthritis.

Lymphangiogenesis prevents immune activation and orofacial pain in TMJ arthritis mouse models

The VEGF-C ligand activates its LEC-enriched receptor VEGFR3 to enhance lymphatic growth and functions^13,42. To promote lymphatic functions, we used the VEGF-C156S recombinant protein (refer to VEGF-C) with a Cys156 to Ser mutation, which yields a lymphangiogenesis-specific form of VEGF-C that specifically binds to VEGFR3⁴³. We utilized a degradable, injectable, and sustainable hydrogel as the delivery vehicle of the VEGF-C protein (refer to DishGel) (Fig. 7a). This hydrogel is composed of A Gel and B Gel, both of which are liquids that can be mixed with VEGF-C protein thoroughly, followed by the intra-articular injection into the TMJ. Our in vitro experiments revealed that A gel and B gel solidify at ~5 min after their combination, allowing for the sustained TMJ retention and VEGF-C release to occur at a slow pace (Fig. 7b). CFA injection causes joint inflammation and swelling, which prevents us from doing a second injection of DishGel. Therefore, we injected VEGF-C-containing DishGel first, followed by intra-articular CFA injection 2 days later, which is technically feasible (Fig. 7a). IHC staining confirmed the lymphangiogenesis after VEGF-C DishGel treatment, as evidenced by increased Lyve1⁺;Vegfr3⁺ double-positive and tube-like lymphatic vessels (Fig. 7e, f).

Fig. 7. Lymphangiogenesis prevents immune activation and orofacial pain in TMJ arthritis mouse models.

a Diagram of the VEGF-C treatment experiment. b Cumulative amount of VEGF-C release measured by ELISA over time (days). n = 3 independent replicates. c Quantification of relative bite force values at different time points. n = 10 mice (PBS), n = 10 mice (CFA), n = 8 mice (CFA + Hydrogel), n = 14 mice (CFA + Hydrogel + VEGF-C). d Quantification of head withdrawal threshold measurement at different time points. n = 10 mice (PBS), n = 10 mice (CFA), n = 8 mice (CFA + Hydrogel), n = 14 mice (CFA + Hydrogel + VEGF-C). P-values were calculated by a two-sided ordinary one-way ANOVA with Tukey post hoc tests in (c, d). e, g, h Immunofluorescence staining of TMJ sagittal sections using antibodies against Lyve1 (green) and Vegfr3 (red), Cathepsin K (green), or CD68 (white). DAPI stains nuclei (blue). Scale bars: 50 µm. f, i, j Quantification of the area fraction of Lyve1⁺;Vegfr3⁺ lymphatic vessels, CD68⁺ macrophages, or Cathepsin K⁺ osteoclasts. n = 4–6 biological replicates per group in (f), n = 3–4 biological replicates per group in (i, j). P-values were calculated by two-sided ordinary one-way ANOVA with Tukey post hoc tests in (f, i, j). Each circle dot represents one mouse. k Tartrate-Resistant Acid Phosphatase (TRAP) staining of TMJ condylar sections. Scale bar: 50 um. l Quantification of the area fraction of TRAP-positive signals. n = 3 biological replicates per group. P-values were calculated by a two-tailed unpaired t-test (l). Each circle dot represents one mouse. Note that hydrogel alone cannot, and VEGF-C-loaded hydrogel can significantly reduce pain, macrophage activation, and osteoclast number in the condylar bone. P-values are indicated in the figure. All data are represented as mean ± SEM. Source data are provided as a Source Data file.

Longitudinal measurement showed that VEGF-C treatment significantly enhanced bite force in CFA-treated mice (Fig. 7c), suggestive of pain mitigation and improved TMJ functions. Similarly, the head withdrawal threshold was enhanced after the VEGF-C-containing DishGel treatment (Fig. 7d). These results suggest that VEGF-C-mediated lymphangiogenesis prevents pain behavior in TMJ arthritis mice. To examine how lymphangiogenesis affects immune cells, we performed IHC staining of TMJ sections using antibodies against Cathepsin K to label osteoclasts and CD68 to label activated macrophages, respectively. VEGF-C treatment had a significant reduction of macrophage activation compared to controls (Fig. 7g, i). Similarly, the TRAP (tartrate-resistant acid phosphatase) assay and Cathepsin K staining revealed a reduction of osteoclasts in the VEGF-C-treated group (Fig. 7h, j–l). Together, these results suggest that lymphangiogenesis by VEGF-C-containing DishGel prevents immune activation and pain behavior in TMJ arthritic mice.

Lymphangiogenesis prevents joint degeneration in TMJ arthritis mouse models

To examine whether lymphangiogenesis by VEGF-C could prevent degenerative processes in our mouse model, micro-CT imaging was used to analyze bone architecture at 2 weeks after CFA intra-articular injection. There was a clear morphological change in the subchondral bone in VEGF-C groups compared to controls (Fig. 8a). VEGF-C-treated TMJs exhibited a significant increase in bone volume (bone volume over total volume, BV/TV) and in trabecular thickness (Fig. 8b, e), as well as a robust decrease in trabecular space and trabecular number (Fig. 8c, d). Therefore, VEGF-C-treated TMJ prevent bone loss in CFA-induced TMJ arthritis mice, possibly due to decreased bone-absorbing osteoclasts as revealed in our previous analyses (Fig. 7h–l). Together, these results suggest that the prevention of subchondral bone loss in arthritic TMJs after lymphatic activation is likely due to decreased osteoclasts.

Fig. 8. Lymphangiogenesis prevents joint degeneration in TMJ arthritis mouse models.

a Micro-CT imaging of the mandibular condyle. Scale bars: 500 µm. b–e Quantification analysis of microarchitecture parameters of the subchondral bone. BV/TV: bone volume/total volume; 1/mm: 1 trabecular number per mm region. n = 4 biological replicates per group. P-values were calculated by a two-tailed unpaired t-test (c–e), and a two-tailed Mann-Whitney test in (b). Each circle dot represents one mouse. f, g H&E staining of TMJ sagittal sections. Note that VEGF-C-treated mice presented less synovitis, including less hyperplastic epithelial lining and less immune cell infiltration. Scale bars: 50 µm. h, i Quantification of TMJ synovitis evaluated by the Synovitis Scoring System with two assessment criteria: synovial hyperplasia (h) and immune cell infiltration (i). n = 11–12 biological replicates per group in (h), n = 11 biological replicates per group in (i). P-values were calculated by a two-tailed Mann-Whitney test in (h, i). Each circle dot represents one mouse. j, k H&E staining and Safranin-O staining of TMJ sagittal sections. Scale bars: 50 µm. l Quantification of Osteoarthritis Research Society International (OARSI) score in TMJs. n = 11–12 biological replicates per group in (l). P-values were calculated by a two-tailed unpaired t-test (l). Each circle dot represents one mouse. P-values are indicated in the figure. All data are represented as mean ± SEM. Source data are provided as a Source Data file.

Next, we performed histopathology analyses to characterize synovitis and cartilage integrity. As exampled in the anterior region, VEGF-C-treated mice had less synovial membrane hyperplasia compared to controls (Fig. 8f, h). We focused on the posterior region and found that VEGF-C-treated mice exhibited less immune cell infiltration compared to controls (Fig. 8g, i). H&E staining and Safranin-O staining revealed that VEGF-C-treated TMJs have improved cartilage integrity compared to controls (Fig. 8j, k), which is consistent with the OARSI score reduction after VEGF-C-mediated lymphangiogenesis in TMJ arthritic mice (Fig. 8l). Together, these findings suggest that lymphangiogenesis prevents synovitis and joint degeneration in TMJ arthritic mice.

Discussion

Here, we identified a synovial lymphatic system in the TMJ. Using TMJ arthritis mouse models, we found that lymphatic vessels undergo extensive lymphangiogenesis and remodeling coupled with lymphatic drainage disruption during the progression of joint arthritis with pain. Lymphatic dysfunction exacerbates TMJ arthritis and pain, likely due to induced interstitial fluid retention and fibroblast-macrophage hybrid cell expansion, which collectively contribute to worsen inflammation, leading to nociceptive activation, pain sensation, and joint degeneration.

We identified lymphatic vessels, mapped their anatomical locations, and characterized their structure and function in TMJs under pathophysiological conditions. We utilized Prox1-mScarlet genetic mice, coupled with immunolabeling with Lyve1 to mark lymphatic vessels, followed by tissue clearance, 3D volume imaging, and imaging reconstruction. Lymphatic vessels are primarily localized at TMJ synovial tissues, where they are adjacent to Cx3cr1-labeled resident macrophages. Consistently, arthritic TMJ displayed robust lymphangiogenesis and lymphatic vessel remodeling, resulting in elongated and tube-like vessels, which is likely attributed to macrophage expansion and their secretion of VEGF-C (Fig. S7b–i). In turn, VEGF-C can activate VEGFR3 in LECs and lead to an increase in lymphatic vessel growth. These findings are consistent with previous reports describing lymphatic remodeling and increased lymphangiogenesis under inflammatory conditions^15,36. Despite their growth, these lymphatic vessels in arthritic TMJs are compromised, as evidenced by the aberrant expression of inflammatory LEC marker Ptx3 and impaired lymphatic drainage. Future studies should examine lymphatic permeability, contractility, and the transition between lymphatic button and zipper junctions in arthritic TMJs. One alternative interpretation of OVA-647 dye accumulation in CFA-treated TMJs is the antigen retention from immune activation in synovial tissue. Therefore, it is important to investigate to what extent lymphatic influx, efflux, clearance, and immune retention distinctly contribute to lymphatic drainage defects in CFA-induced TMJ arthritis mice.

Our functional studies showed that the TMJ lymphatic network plays critical roles in regulating fluid homeostasis and immune response. This observation is consistent with findings in knee joints, where lymphatic vessels function to drain cavity space fluid, interstitial fluid, macromolecules, and immune cells, thus preventing inflammation and joint degeneration^31,44,45. Beyond joint degeneration, our studies found that lymphatic dysfunction exacerbates pain behavior, which is barely examined in previous lymphatic studies of joints. In the Prox1^+/− lymphatic malfunction mice, lymphatic deficiency leads to increased expression of IL-1β⁺Iba1⁺ macrophages and IL-1β⁺Ly6b⁺ neutrophils around the TMJ, which might yield more pro-inflammatory mediators or immune cell infiltration^46,47. In addition, Prox1^+/− mice displayed more severe arthritic pathological features than control mice under the same CFA treatment. In our lymphatic gain-of-function studies, hydrogel-mediated delivery of VEGF-C to the TMJ prevented pain behaviors and reduced immune cell activation. Hydrogel, together with VEGF-C, but not hydrogel itself, can significantly prevent pain behaviors in CFA-induced arthritis mice. Micro-CT and histological analyses further revealed less subchondral bone loss, decreased synovitis, and improved cartilage integrity. Together, these studies established the causative relationship between lymphatic dysfunction and painful TMJ degeneration in arthritic TMD. In our preclinical CFA-induced TMD mouse model, inflammation is initiated by heat-killed Mycobacterium tuberculosis bacilli, which elicit a strong immune response leading to synovitis, pannus formation, and cartilage destruction. In clinically relevant TMD arising from trauma, excessive mechanical loading, or low-grade infection, joint damage is primarily driven by mechanical stress, chondrocyte death, and secondary mild inflammation. In these settings, the lymphatic system primarily facilitates clearance of damage-associated molecules and supports tissue repair. Thus, whereas CFA models provide insight into immune mechanisms of TMJ inflammation, they may not fully recapitulate the lymphatic dynamics in mechanical TMJOA. It is important to investigate how lymphatic vessels function in different types of TMJ disorders, such as mechanical and injury models of TMD.

Our studies identified impaired lymphatic drainage and induced fibroblast-macrophage hybrid (FM cells) as the potential mechanisms underlying TMD in mice. Our findings suggest that in response to inflammation, lymphatic vessels remodel and grow to enhance the drainage functions for removing inflammatory mediators from the interstitial tissues. After reaching a certain threshold in the transition from physiological to pathological, lymphatic vessels are impaired in both structure (Ptx3 inflammation) and function (lymph node drainage). As a result, inflammatory mediators cannot be effectively removed and are trapped in the interstitial synovial tissues, which in turn impair lymphatic vessels while also activating nociceptors. Our study on lymphatics in TMJ provides experimental evidence and validates the previous hypothesis of impaired lymphatic drainage and interstitial inflammatory stasis in pain⁴⁸. Not only are inflammatory mediators ineffectively removed, but their generation increases due to the rise in fibroblast-macrophage hybrid cells. Our scRNA-seq used post-perfusion tissues to avoid circulating immune cells and identified a cluster of fibroblast-macrophage hybrid cells (FM cells). These FM cells are highly enriched with ECM proteins (Col1a1, Col1a2) and pro-inflammatory cytokines (IL-1β). These fibroblast-macrophage dual identity cells may be protective early on, but progress towards fibrosis and stiffness if chronically activated, as suggested by their gene expression profile. Using Masson’s Trichrome Staining, we observed severe collagen deposition and fibrotic tissue accumulation in lymphatic-deficient synovial tissues. Our finding is consistent with previous reports of fibrosis as a critical feature in joint degeneration, leading to stiffness, pain, and impaired function^49,50. The excessive and abnormal deposition of ECM, characterized by increased dense and cross-linked collagen fibrils, disrupts the normal biomechanical environment necessary for joint homeostasis, leading to loss of tissue elasticity and joint mobility^40,51. We acknowledge that our studies have not established the direct link between fibrotic tissue and nociceptive activation in TMD, which will be a future direction.

Our studies provide therapeutic implications for targeting lymphatics to prevent inflammatory TMJ degeneration and pain. Our hydrogel-based DishGel delivery of VEGF-C activates lymphatic growth and prevents arthritis pathologies and orofacial pain, which provides a feasible and reliable method of functional perturbation of the lymphatic system in the TMJ. It should be noted that our VEGF-C treatment occurred prior to the CFA lesion; therefore, lymphangiogenesis happened before the onset of TMJ arthritis. We demonstrated that promoting healthy lymphangiogenesis is beneficial in TMJs. The decision of whether the lymphatic system should be activated or inhibited for the treatment of TMJ arthritis depends on its pathophysiological stage. For example, it has been reported that lymphangiogenesis mediates renal inflammation and fibrosis⁵², which suggests the detrimental effects of pathological lymphangiogenesis. Future studies should investigate the effects of lymphatic perturbation on specific stages and types of TMD after the phenotype manifestation.

Study limitations

We cannot definitively conclude whether lymphatic influx is altered under inflammatory conditions. Although our short-term imaging data show no appreciable difference in ICG accumulation within the TMJ between control and CFA-treated mice during the 0–5 min window after injection (Fig. 3n–p), this readout represents a tissue-level surrogate and cannot distinguish true lymphatic capillary uptake from nonspecific interstitial diffusion or changes in vascular permeability. Moreover, immune activation induced by CFA may promote antigen retention, which could contribute to increased tracer signal independent of lymphatic influx or efflux. Therefore, while our data argue against a major increase in early tracer influx in CFA-treated TMJ, they cannot fully disentangle impaired efflux from immune-mediated antigen retention. Future studies using direct lymphatic imaging, real-time flow measurements, or selective labeling of lymphatic capillary entry will be required to rigorously resolve these mechanisms.

Methods

Materials and Mouse models

The C57BL/6 J (JAX#000664) mice were obtained from Jackson Laboratory. Prox1-EGFP and Prox1^+/− were from Dr. Young-Kwon Hong's laboratory. Prox1-mScarlet mice were obtained from Dr. Hu Zhao’s laboratory at the Chinese Institute for Brain Research, Beijing. Animals were housed with regulated humidity, temperature, and a 12-h alternating light-dark environment. Animals were euthanized by carbon dioxide inhalation, with cervical dislocation performed immediately afterward to confirm death. Pain behavioral studies used female mice due to inherent differences between sexes in nociceptive sensitivity, with higher TMJ pain incidence in females. Female mice were used in all additional experiments in this study. All animal studies were performed with the approval of the Institutional Animal Care and Use Committee of the University of Southern California. Reporting of animal experiments in this study complies with the ARRIVE guidelines. Synovial tissue was obtained as a biopsy specimen during TMJ arthroscopy, performed as a part of routine clinical care. The sample was collected from the anteromedial synovium of a 20-year-old female patient with bilateral TMJ internal derangement and arthralgia. The tissue was fixed immediately in 10% formalin after collection. Human TMJ synovial tissue specimens were de-identified and obtained under the guidelines of the Institutional Review Board of the University of Southern California.

CFA intra-articular injection

The TMJ injection area was identified by palpating the zygomatic arch. A depression located approximately 2 mm anterior to the external auditory canal, beneath the posterior part of the zygomatic arch, served as the anatomical landmark. The needle was inserted until its tip contacted the bone, roughly 2 mm below the surface. Subsequently, 10 µL of CFA (Chondrex, Inc, concentration 5 mg/mL) or fluorescently labeled Ovalbumin (OVA-647; Thermo Fisher; concentration 0.5 mg/mL) were slowly injected into each TMJ capsule. Following the injection, the needle was held in place for at least 5 s before careful withdrawal. The control group mice underwent bilateral injections using sterile PBS (10 µL per capsule).

Nociceptive behavior assessment

Before behavioral testing, mice were habituated in the behavior room for at least 1 h. Baseline measurements of bite force and head withdrawal thresholds were recorded during a 3-day pre-training phase, prior to intra-articular injections of PBS or CFA. After training, mice typically initiate biting behavior within 10 s after exposure to the bite force sensor. Mice that failed to bite spontaneously within 10 s after 3 days of training were excluded from the experiment. To measure bite force, each mouse was gently restrained in a modified 50 mL plastic tube, allowing free and comfortable head movement of testing mice. A bite force sensor (YFM-1-100, measurement range 0–100 N) was connected to the NBIT RSD-V2.6.3 software via an NST2000 data acquisition system (Nanjing Shen-yuan-sheng Intelligent Technology Co.) for measuring bite force. The voluntary bite force was recorded over a period of 2 min per session, and the five strongest bite amplitudes were averaged to determine the final bite force value.

Head withdrawal thresholds were measured using an electronic von Frey analgesiometer. Mice were habituated for at least 1 h in wire mesh cages placed within the behavior room prior to testing. The von Frey filament was applied perpendicularly to the TMJ region, and the head withdrawal threshold was defined as the minimum force required to elicit a head withdrawal response. Measurements were conducted for at least five individual trials per animal, separated by 10-s intervals, during approximately 2 min of restrained positioning. Threshold values from each trial were averaged to get the final head withdrawal threshold. The detailed bite force and Von Frey assays for TMJ pain were described in our recent video publication³⁷.

Micro-CT (µCT) analysis

Micro-CT live imaging was performed on a Scanco Medical µCT 50 scanner (Scanco Medical, Switzerland) at the University of Southern California Molecular Imaging Center (90 kVp, 78 µA, 10 µm pixel size). AVIZO 9.4.0 (Thermo Fisher Scientific) was used to perform a 3D reconstruction of the TMJ. The microarchitecture parameters of the subchondral bone, including bone volume over total volume (BV/TV), trabecular spacing (TbSp), trabecular thickness (TbTh), and trabecular number (TbN), were analyzed using VGStudio Max3.3 (Volume Graphics, Inc., USA). For each sample, three spherical regions of interest (ROI) with a radius of 0.1 mm were selected at the midpoints of the anterior, middle, and posterior condyles for all measurements. Each dot in the graph quantification represents one sample (n). Student’s t-test was used for statistical analysis. A significance level was set at a p-value of 0.05.

Histology analysis

The TMJs from mice were dissected, fixed, and processed through decalcification and washing steps. Subsequently, tissues were dehydrated through a graded ethanol series, cleared with xylene, and embedded in paraffin for sectioning. Histological sagittal sections of the TMJs in 5 µm were used to detect histological changes in joints. The severity of the cartilage degeneration of TMJ samples was assessed using Hematoxylin and Eosin (H&E) and Safranin-O staining, according to a modified Osteoarthritis Research Society International (OARSI) scoring system^2,53. Scores ranged from 0 to 6 points, with 0 representing intact and continuous cartilage surface, evenly distributed matrix, properly oriented chondrocytes with clear boundaries, and no signs of proliferation or cell death. A higher OARSI score is calculated by the degree of unclear borders between cartilage and subchondral bone, uneven cartilage surfaces, and decreased hypertrophic layer thickness. Synovitis was quantified based on the severity of synovial lining hyperplasia and immune cell infiltration. Synovial lining hyperplasia was graded on a scale from 0 to 2. Synovial membrane staining was graded as follows: 0 for 1–3 layers, 1 for 4–6 layers, and 2 for more than 7 layers. Inflammation was graded based on the severity of inflammatory cell infiltration: 0 for no infiltrates, 1 for a few perivascular lymphocytes or plasma cells, 2 for numerous lymphocytes or plasma cells occasionally forming follicle-like aggregates, and 3 for a dense band-like infiltrate or multiple large follicle-like aggregates. Collagen deposits were evaluated using Masson’s Trichrome Staining (Trichrome Stain Kit - Masson’s, Statlab, #204183) on 5 µm paraffin sections. Cytoplasm, muscle fibers, and keratin were stained in shades of pink to red, while collagen fibers appeared blue. Histological scoring was performed in a blinded manner by independent researchers who were not involved in the experimental procedures or data analysis. To assess scoring reliability, inter- and intra-observer variability were evaluated using the intraclass correlation coefficient (ICC). A subset of 20% of the samples was re-scored by each observer after a two-week interval. Both inter- and intra-observer ICC values exceeded 0.85, indicating good to excellent reproducibility of the histological scoring.

Immunostaining

TMJ and lymph node sections were prepared for immunofluorescence staining following standard protocols. Lymph nodes were harvested from the neck region and fixed in 4% PFA at 4 °C overnight. The samples were then dehydrated in 30% sucrose at 4 °C overnight, followed by a 1:1 mixture of 60% sucrose and OCT (Tissue-Tek, Sakura) at 4 °C overnight. Tissues were then embedded in OCT on dry ice and sectioned at a thickness of 8 µm. For TMJ cryosections, samples were decalcified in 14% EDTA for at least 10 days, followed by gradual dehydration in 30% sucrose overnight and in a 1:1 mixture of 60% sucrose and OCT at 4 °C overnight. The samples were then embedded in OCT, frozen on dry ice, and sectioned at a thickness of 14 µm using a cryostat (Leica CM1850).

Antigen unmasking solution (Vector, H-3300) was used for antigen retrieval. The primary antibodies included Lyve1 (Invitrogen, 14-0443-82), Vegfr3 (R&D, AF743), Cathepsin K (CTSK; 11239-1-AP), Iba1 (FujiFilm, 019-19741), CD68 (Bio-Rad, MCA1957GA), Runx2 (Cell Signaling, 12556), Ptx3 (Invitrogen, PA5-36156), Ly6b (Bio-Rad, MCA771GT), Anti-GFP (Abcam, ab13970), and Endomucin (Santa Cruz, sc65495). Alexa Fluor 488/568/647 (Invitrogen) were used as secondary antibodies. DAPI (Invitrogen; Cat# 62248) was used for nuclear staining. The percentage of positive immunofluorescence signals and area fractions were analyzed using Image J software/Adobe Photoshop. To calculate lymphatic vessel area, Lyve1⁺;Vegfr3⁺ double-positive areas were selected from comparable anatomical positions between different groups of TMJ sections and then divided by DAPI area. Only >50 um lymphatic vessel tubes were counted for the lymphatic vessel length calculation. Quantification was performed using three to five sections per mouse. Each dot in the graph represents the mean value of an individual mouse within the group. At least three mice were used for each group or genotype. Fluorescence imaging was performed on a Keyence BZ-X810 microscope using the standard DAPI, GFP, Texas Red, and Cy5 filter sets. The excitation/emission ranges were: DAPI, 360/40 and 460/50 nm; GFP, 470/40 and 525/50 nm; Texas Red, 545/25–560/40 and 605/70–630/75 nm; and Cy5 (far-red), 620/60 and 700/75 nm.

RNAscope staining

Sample preparation and RNAscope staining were conducted following the standard ACD protocol using the RNAscope® Multiplex Fluorescent Reagent Kit v2 (Cat. No. 323100). Mm-Il1β (316891-C1), Mm-Vegfc (430661-C3), and Mm-Csflr (428191-C2) probes were used in this study. A negative control probe (Cat. No. 320871) was used for staining and imaging to minimize background signal. RNAscope staining was quantified using ImageJ at 20x magnification and analyzed according to ACD scoring guidelines. At least three mice from independent litters were used for this study.

CUBIC clearing

Animals were administered an intra-articular injection of OVA-647. One hour later, samples were collected following the steps below. Following anesthesia, the animals underwent transcardiac perfusion with cold PBS, followed by 4% PFA. The dissected TMJs were further fixed with 4% PFA at 4 °C overnight. After thorough rinsing with cold PBS, the samples were decalcified in 20% EDTA (pH 7.4) for 7 days. The decalcified TMJ was subjected to CUBIC-L (TCI, Cat# T3740) for permeabilization. After permeabilization, the decalcified TMJ was stained with VEGFR3 (Bio-techne, Cat# AF743) at a 1:200 dilution for 2 days. The samples were then washed with PBST before incubation in CUBIC-HL for 2 days. CUBIC-HL solution for RI matching was made of 50% w/w iohexol (Aladdin, Cat# I134719), 23% w/w urea (Sigma, U5378), 11% w/w trithanolamine (Aladdin, Cat# T58300), and 16% w/w dH2O. Finally, the cleared TMJ was mounted in CUBIC-HL solution on a confocal imaging dish for imaging.

Lymph node clearing was performed using the same CUBIC protocol as described above. Briefly, animals received OVA-647 injections and were sacrificed at different time points post-injection for tissue collection. Accessory mandibular lymph nodes were isolated and fixed in 4% PFA overnight. Following washes with PBS, the samples were incubated in CUBIC-L solution at 37 °C for a minimum of 3 days. After PBS washing, the attached membrane was carefully removed under the microscope. Lymph node nuclei were stained with PI at a 1:1000 dilution, resulting in a pink-to-red coloration after 48 h. The samples were then washed with PBS before incubation in CUBIC-R for 2 days. Finally, lymph nodes were mounted in CUBIC-R solution on a confocal imaging plate.

iDISCO tissue clearing

TMJ samples were post-fixed with 4% PFA/PBS at 4 °C overnight and washed twice with PBS for 1 h each. After rinsing with PBS, the samples were decalcified in 20% EDTA for 7 days. Then, samples were gradually dehydrated through two sequential incubations in 30% ethanol (prepared in ddH₂O) for 1 h each, followed by two changes of 50%, 80%, and 100% ethanol, with each step lasting 1 h. Samples were then treated overnight at 4 °C with 5% H₂O₂ (Sigma, H1009), followed by progressive rehydration through a graded ethanol-to-water series. Next, samples were incubated in a cold permeabilization solution composed of 25% (w/w) urea, 15% (w/w) glycerol, 15% (w/w) Triton X-100, and 45% (w/v) double-distilled water at 4 °C for 5 h. This was followed by enzyme-mediated digestion using 0.2% (w/v) collagenase (Merck, 10,103,578,001) in PBS at 37 °C for 30 min under continuous shaking. Samples were then washed twice for 5 min each using a solution containing 2% (v/v) FBS (Sigma-Aldrich, F7524) in PBS on a rocking shaker. For the staining, samples were transferred to a fresh blocking solution composed of 10% (v/v) donkey serum (Abcam, ab7475), 10% (v/v) DMSO (Sigma-Aldrich, D5879), and 0.5% (v/v) Triton X-100 in PBS and incubated at 37 °C for 20 min. After the blocking, tissues were treated with VEGFR3 and GFP primary antibodies (Rockland, 48776) prepared in an antibody dilution buffer containing 2% (v/v) donkey serum, 10% (v/v) DMSO, and 0.5% (v/v) Triton X-100 in PBS. The samples were incubated overnight at 37 °C with gentle agitation at 120 rpm, PBS wash, and 2^nd antibody treatments. Lastly, tissues were dehydrated using a gradient of ethanol concentrations (30%, 50%, and 80%) for 30 min each, transferred to pure methanol for 1 h, and washed twice with ethyl cinnamate (ECi) (Sigma-Aldrich, 112372) for 5 min per wash. Subsequently, tissues were incubated in a clearing solution composed of 80% (v/v) ECi and 20% (v/v) polyethylene glycol (PEG) (Sigma, 447943) under gentle rotation at room temperature for 30–60 min before imaging.

Confocal imaging and process

Cleared TMJ and lymph node samples were imaged using a Leica STELLARIS 5 confocal microscope. Imaging was performed with a 10x/NA0.3 objective lens (2 mm working distance) and a 63x/NA1.40 objective lens (200 μm working distance). LAS X Life Science Microscope Software (version 1.4.6) was used for image acquisition. The microscope was equipped with fixed laser wavelengths at 405 nm, 488 nm, 561 nm, and 633 nm. Scans were conducted at a zoom factor of 0.75x, utilizing either the 10x/NA0.3 or 63x/NA1.40 objective lens, with a step size of 1 µm under continuous scanning in the 488 nm, 561 nm, and 633 nm channels. To improve visual representation, gamma correction was applied to raw data from the confocal microscope. ImageJ (NIH, http://imagej.nih.gov/ij) was used for file format conversion, while Bitplane Imaris (http://www.bitplane.com/imaris/imaris, version 10.1.1) facilitated 3D reconstructions, manual annotations, quantification, and video generation. For Fig. 1a–d, the excitation/emission settings were: Autofluorescence (AF): Excitation: 405 nm. Emission (detector window): 420–480 nm. Lyve1: Excitation: 488 nm. Emission (detector window): 500–550 nm. Prox1: Excitation: 561 nm. Emission (detector window): 570–620 nm. Filters: No filter. For Fig. 1g, the excitation/emission settings were as follows: Autofluorescence (AF): Excitation: 405 nm. Emission (detector window): 420–480 nm. Lyve1: Excitation: 488 nm. Emission (detector window): 500–550 nm. VEGFR3: Excitation: 633 nm. Emission (detector window): 650–720 nm. Filters: Median Filter size 5 × 5 × 5.

Hydrogel preparation and VEGF-C release assay

We employed an in situ crosslinking strategy using chemically modified hyaluronic acid (HA) derivatives. Specifically, HA-adipic acid dihydrazide (HA-ADH) was synthesized by conjugating adipic dihydrazide to the carboxyl groups of HA (150 kDa), designated as A hydrogel⁵⁴. HA-aldehyde (HA-CHO), referred to as B hydrogel⁵⁵, was prepared by reacting HA with equimolar sodium periodate. Upon mixing, A and B hydrogels undergo a Schiff base reaction, forming an injectable hydrogel network suitable for in vivo implantation.

A hydrogel-based controlled release system was prepared using a combination of 100 ul Gel A (2%) and 100 ul Gel B (2%). 25 ug of recombinant VEGF-C (R&D, Cat# 752-VC-025/CF) was incorporated into Gel B, and the mixture was transferred into 1.5 mL Eppendorf tubes. Following gelation, each hydrogel sample was overlaid with 200 μL PBS supplemented with a protease inhibitor cocktail (Thermo Fisher, Cat#A32955) and 0.02% sodium azide to prevent protein degradation and bacterial contamination, respectively. The release medium was collected at 24-h intervals. This procedure was repeated daily for 14 consecutive days. VEGF-C concentrations in the collected samples were quantified using a commercially available human VEGF-C ELISA kit (Novus, DVEC00) according to the manufacturer’s instructions. All release assays were independently repeated three times. The cumulative release profile of VEGF-C over the 14-day period was generated by plotting the mean concentration of released VEGF-C against time.

IVIS-ICG (In Vivo Imaging System-Indocyanine Green) lymphatic imaging

Mice were anesthetized with 2.5% isoflurane and administered with intra-articular injections of Indocyanine Green (ICG; Sigma, Cat# I2633-25MG) into TMJs. To assess joint clearance, serial imaging was performed using an IVIS Live Animal Imaging System (Perkin Elmer Inc.) at baseline (0 min) and at 5, 30, 60, and 120 min following ICG injection. ICG signals were detected in the TMJ region and quantified as epifluorescence total radiant efficiency. To minimize inter-animal variability, clearance was expressed as the percentage change from baseline for each mouse.

Library preparation and Single-Cell RNA sequencing (scRNA-Seq)

After cold PBS transcardiac perfusion, TMJs were carefully dissected from 3 WT or 3 Prox1^+/− mice after CFA treatment, finely minced in a tissue suspension medium consisting of Minimum Essential Medium (MEM) with 2% FBS, placed in a fresh centrifuge tube containing 10 mL of digestion medium (Collagenase P, 1 mg/mL; Dispase II, 2 mg/mL), and incubated at 37 °C under continuous rotation for 25 min. Following digestion, tissues were resuspended in suspension medium and centrifuged at 500 g for 10 min at 4 °C, after which the supernatant was discarded. The pellet was then resuspended in 4 mL of DNase I solution (2 U/mL in MEM) and incubated at 37 °C for 10 min. To facilitate tissue dissociation, suspension medium was added, and samples were gently pipetted on ice. After thorough dissociation, the suspension was filtered through a 70-µm nylon mesh, followed by centrifugation at 4 °C for 10 min. Cells were loaded into the 10x Chromium system, aiming for a recovery of 10,000 cells. Library construction was carried out according to the Chromium Next GEM Single Cell 3’ Reagent Kits v3.1 (Dual Index) protocol provided by the manufacturer. Sequencing was performed using the NovaSeq X Plus platform.

scRNA-seq analysis

Raw read quality control was conducted using 10x Genomics Cell Ranger 7.0.1. Sample alignment to the reference genome GRCm38 (mm10), read quantification, and barcode filtering were performed with Cell Ranger Count. Data preprocessing and analysis were carried out in Seurat version 4.9.9. Cells were retained if they contained at least 500 detected genes and exhibited ≤10% mitochondrial content. Normalization was conducted using the SCTransform method. PBS and CFA datasets were integrated via the FindIntegrationAnchors function. Principal component analysis (PCA) was performed, selecting the top 200 principal components for dimensionality reduction using the Uniform Manifold Approximation and Projection (UMAP) algorithm. Clustering was executed with the FindClusters function at a resolution of 1.5. Marker gene identification was performed by comparing each cluster against all others using the FindAllMarkers function, with a log-fold change threshold of 0.25 and inclusion criteria of >25% cells expressing the gene.

Statistical analysis and reproducibility

Statistical analyses were performed using GraphPad Prism (version 9.0.0) software. Data are presented as mean values ± SEM. To compare more than two experimental groups, a one-way ANOVA with Tukey post hoc test was performed, and for comparison between two groups, a two-tailed or one-tailed (as indicated in figure legends) unpaired Student’s t-test was used to calculate P values. Each experiment was repeated independently at least three times. Graphs showing mean ± SEM for each group represent at least three biological replicates (as indicated in figure legends). Images of immunofluorescence staining were representative of at least three biological replicates. Data were tested for normality using the Shapiro–Wilk test and for homogeneity of variance using Levene’s test. When assumptions for parametric testing were not met, appropriate non-parametric tests were applied. For Figs. 1a, g, 4 independent experiments were repeated. For Fig. 1e, Figs. 2a, b, 3b, 3 independent experiments were repeated.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y.S., Q.C., Z.L., P.F.K. and J.Y.C. performed all experiments. F.X.C. prepared the hydrogel. Z.L. analyzed all bioinformatic data. J.F.C. and Z.Z. designed the experiments and supervised the research. D.A. and Y.Q. prepared human biospecimens. G.T.C., A.V.M., Z.Y.L. and J.X. helped with manuscript writing. J.F.C. and Y.S. co-wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Alejandro Almarza and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The scRNA-seq data generated in this study have been deposited in GEO under accession number GSE295404. The seqFISH data are available in the FaceBase database under the 10.25550/8Q-3KAR. Source data are provided with this paper. The raw data generated in this study was deposited in the FaceBase repository under the 10.25550/A2-0MH4. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72400-0.