Heterogeneity of active mast cells, endothelial cells, and fibroblasts in hemophilic arthritis defined by synovial single-cell sequencing Authors: Kun Lin, Shuai Fan, Xiaoqiang Yang, Wenyuan Hou, Junjiao Zhang, Jiaru Liao, Longfei Han, Peng Peng, Fangjun Xiao, Weihua Fang, Xianshun He, Jiaqing Tian, Shun Lu, Fan Yang, Qiushi Wei, Mincong He Published: 2025 ABSTRACT Recurrent synovial bleeding is the major cause of Hemophilic Arthritis (HA). Up to now, the transcriptomic profiles of bleeding syvonium in HA patient have largely remained unknown. Exploring the transcriptomic characteristics of synovium obtained from HA patients undergoing total knee arthroplasty (TKA), and uncovering potential pathological mechanisms of HA synovium through bioinformatics analysis and differential analysis. Single-cell RNA sequencing (scRNA-seq) technology was utilized to identify distinct cellular subsets within HA synovium. Comparative analysis was conducted with scRNA-seq data from osteoarthritis (OA) and rheumatoid arthritis (RA) synovium to assess transcriptional differences. Histological evaluation, immunofluorescence (IF), immunohistochemistry (IHC) and in vitro cell testing were performed for validation. We observed that the single-cell transcriptomic characteristics of HA synovium differ significantly from those of OA and RA patients. Mast cells, identified as unique immune cells in HA synovium, are actively involved and may be the initiating factor for changes in vascular permeability or bleeding in HA synovium. The transcriptomic features of endothelial cells and fibroblasts, which are non-immune stromal cells, reveal the synovial microenvironment: repeated bleeding leads to iron deposition, which causes substantial ferroptosis within the synovium. Based on the pseudotime analysis of endothelial cells, along with IHC staining and in vitro cell assays, it was demonstrated that ferroptosis in endothelial cells induces vascular damage and triggers a significant migration of endothelial cells. The stress of ferroptosis leads to the differentiation of HMOX1+ endothelial cells. Cell interaction analysis shows that HMOX1+ endothelial cells can induce the SCARA5+ fibroblast subtypes via FTL/SCARA5. This may be a critical target for alleviating cellular ferroptosis in HA synovium. This study presents the first single-cell transcriptomic atlas of hemophilic arthropathy (HA) synovium, identifying activated perivascular mast cells and a novel SCARA5 + fibroblast subpopulation that regulates ferroptosis. These findings reveal a unique cellular interactome in HA and propose new potential targets for gene therapy aimed at recurrent bleeding.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-27368-0. Purposes Hemophilia is a group of X-linked recessive hereditary diseases caused by deficiency of coagulation Factor VIII (FVIII) or IX, resulting in a hereditary coagulation disorder. Hemophilia A is a bleeding disorder caused by deficiency of FVIII. Joint bleeding is one of the complications of haemophilia A, and recurrent joint bleeding can lead to HA, which is the most severe complication of haemophilia1. To improve the quality of life for HA patients, joint replacement surgery like TKA, is preferred for those for whom conservative treatment is ineffective2. Research indicates that recurrent intra-articular bleeding in hemophilia patients leads to changes in the intra-articular environment, resulting in cartilage degeneration, synovial inflammation, vascular neogenesis, and other pathological conditions. In severe cases of haemophilia3, the gross state of the joint synovium during surgery differs significantly from that of OA or RA. The inflammatory proliferation of synovium and large hemorrhages may be related to the pathogenesis of HA. Therefore, this study employs scRNA-seq4 and bioinformatics analysis to investigate the articular synovium of HA patients in comparison with that of OA and RA patients. The aim is to identify potential therapeutic targets for HA synovium from the perspective of single-cell gene expression. Methods Patient specimen collection and ethics All experimental procedures involving human tissues were conducted in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The specimens used in this study were obtained from the Third Affiliated Hospital of Guangzhou University of Chinese Medicine (PJ-XS-20241029-001). Patients were diagnosed with hemophilia A and confirmed with HA through MRI. Fresh synovial membranes were collected from their knees after TKA, immersed in tissue storage solution (130-100-008, MACS). Additionally, specimens from patients with OA and RA were similarly diagnosed through medical history and imaging, and fresh synovial membranes was obtained during TKA. This study adhered to all relevant ethical regulations concerning human subjects. The Ethics Review Committee of the Third Affiliated Hospital of Guangzhou University of Chinese Medicine approved the tissue acquisition procedures and experimental protocols. Informed consent was obtained from all participants. For patients, knee joint MRI imaging was conducted during the clinical diagnostic process. Fresh tissue samples were immediately dissected into portions for (1) rapid freezing, (2) fixation in 4% paraformaldehyde solution followed by paraffin embedding, (3) enzymatic digestion into single cells. The scRNA-seq process is described as follows. Syvonium dissociation and scRNA-seq Adjust the cell concentration to the ideal range of 700–1200 cells/µL. Gel beads containing barcode information are mixed with cells and enzymes. Then encapsulated in oil surfactant droplets within a microfluidic system, forming Gel Beads-In-Emulsions (GEMs). GEMs flow into a collection reservoir where gel beads dissolve, releasing barcode sequences, reverse transcribing cDNA fragments, and labeling the samples. Gel beads are ruptured and oil droplets shattered to PCR amplify cDNA using it as a template. The products from all GEMs are mixed to construct a standard sequencing library. Initially, cDNA is enzymatically fragmented into segments around 200–300 bp, followed by traditional second-generation sequencing library construction processes like adding sequencing adapters and primers, and finally PCR amplification to obtain a DNA library. High-throughput sequencing of the prepared library is performed using Illumina sequencing platform’s paired-end sequencing mode. After obtaining raw data, the official analysis software Cell Ranger from 10x Genomics (https://support.10xgenomics.com/single-cell-gene-expression/software/overview/welcome) is used for data filtering, alignment, quantification, cell identification, and obtaining gene expression matrices for each cell. Raw scRNA-seq files of OA and RA synovial membranes are obtained from Gene expression omnibus(GEO): GSE1528055,6 and GSE2008157. Subsequently, individual or integrated analyses of the datasets are performed using Seurat(v4.1.1) R-package (https://satijalab.org/seurat/). Standard Seurat pipeline is employed for cell filtering, normalization, dimensionality reduction analysis, clustering, differential gene expression analysis for each subgroup, and marker gene selection. For the dataset containing multiple samples, Harmony (v1.2.0)8 is employed to remove batch effects from multiple sample datasets, followed by the utilization of Clustree (v0.5.1)9to select the optimal resolution parameter for clustering. GO and KEGG pathway enrichment analysis To interrogate the biological functions and signaling pathways associated with each cell subpopulation, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed on the sets of significantly highly expressed genes (marker genes) specific to each cluster10,11. The marker genes for each cluster were identified using the FindAllMarkers function in Seurat (v4.3.0) with a threshold of |avg_log2FC| > 0.25 and an adjusted p-value (Bonferroni correction) < 0.05. Enrichment analysis was conducted using the enrichGO and enrichKEGG functions from the clusterProfiler R package (v4.6.2)12. The universe of genes was set to all genes detected in the dataset. Terms with an adjusted p-value (Benjamini-Hochberg method) < 0.05 were considered statistically significant. The results were visualized using the dotplot function to display the most significantly enriched terms. Pseudotime trajectory analysis To delineate the dynamic differentiation process of synovial endothelial cells, we performed unsupervised pseudotime analysis using the Monocle2 algorithm (v2.22.0) based on the scRNA-seq gene expression matrix. Significantly varying genes (q-value < 0.01) identified by the differentialGeneTest() function were selected as ordering features. The DDRTree dimensionality reduction method was applied to construct cell developmental trajectories. Cells were projected onto a continuous pseudotime axis via the orderCells() function, generating cell differentiation trajectory plots (depicting branch structures and state transitions) and core gene expression heatmaps (visualizing key gene modules driving differentiation along pseudotime, e.g., regulators associated with endothelial proliferation and ferroptosis). Cell-Cell communication analysis CellPhoneDB (v2.1.4) was employed to systematically decode intercellular interaction networks using annotated cell subtypes and their quantified gene expression data. This algorithm integrates subunit composition of ligand-receptor complexes, spatial co-localization, and expression abundance, with interaction significance evaluated by permutation testing (1,000 random shuffles; P < 0.05). Outputs include: intercellular interaction heatmaps (quantifying communication strength between cell types via interaction weight scores) and ligand-receptor pair dot plots (highlighting significant interacting pairs and their expression patterns across specific cell types, e.g., key signaling pathways between mast cells and endothelial cells). Histology analysis Fresh synovial membranes from patients with HA, OA, and RA undergoing TKA surgery were fixed overnight in 4% paraformaldehyde and then embedded in paraffin. Subsequently, 5 μm-thick maximum surface paraffin sections were cut from the diseased synovial membranes and stained according to the manufacturer’s instructions using hematoxylin and eosin (H&E, PN0004), Perl’s blue (S191029), and Masson solution (PN0004). IF double staining was performed using rabbit anti-SCARA5 (1:250, DF14221, Affinity Biosciences), mouse anti-COL1A1 (1:400, 67288-1-Ig, Proteintech), rabbit anti-HMOX1 (1:200, 10701-1-AP, Proteintech), and mouse anti-CD31(PECAM1) (1:200, 66065-2-Ig, Proteintech). IHC staining was carried out using rabbit anti-TPSB2(1:50, A19801, ABclonal), rabbit anti-vWF (1:50, 27186-1-AP, Proteintech) and GPX4 (1:1000, 67763-1-Ig, Proteintech), as described above. Cell culture Human Umbilical Vein Endothelial Cells (HUVECs) are cultured at 37 °C in a 5% CO2 environment. Once the cells have adhered to the culture surface, fresh medium is replaced. When the HUVECs reach 70–80% confluence, they are treated with a 0.25% Trypsin-EDTA solution to detach them from the culture surface. The cell suspension is then transferred to new culture flasks for subculturing. CCK-8 assay Cells are passaged to the 3rd-5th generation and seeded at a density of 6000 cells per well in a 96-well plate. After 24 h of culture, the cells are grouped for drug intervention. Complete medium serves as the blank control group, and pure cell suspension serves as the control group. Cells are treated with RAS-Selective Lethal compound 3 (RSL3) - a potent and direct inhibitor of the antioxidant enzyme Glutathione Peroxidase 4 (GPX4) - at concentrations of 100, 150, 200, 250, 300, 350, 400, and 450 nM as the model group, with six replicate wells for each concentration. The cells are cultured for 12 h, 24 h, and 48 h. After the intervention, 10 µl of CCK-8 working solution is added to each well and the cells are cultured further. Then, the absorbance is measured hourly at 450 nm using a microplate reader, and cell viability is calculated. Cellular immunofluorescence Cells are passaged to the 3rd generation and seeded at a density of 3000 cells per well on 24-well plates with glass coverslips. The blank control group (CON) is replaced with fresh medium once. The ferroptosis induction group (RSL3) is cultured with medium containing 150 nM RSL3 for 24 h. After discarding the medium, the cells are rinsed three times with PBS, fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton, rinsed with PBS, blocked with goat serum, and incubated with primary antibodies against CD31, HO-1, and GPX4 overnight at 4 °C. The next day, the primary antibodies are recovered, and the cells are rinsed three times with PBS, then incubated with the corresponding fluorescently labeled secondary antibodies for 1 h at room temperature. After three more rinses with PBS, the coverslips are mounted with an anti-fluorescence quenching mounting medium containing DAPI. The expression of CD31, HO-1, and GPX4 is observed and photographed under a fluorescence microscope. Statistical analysis The analysis was conducted within the R framework, known for its statistical computation and visualization capabilities. To ensure robustness, P-values underwent adjustment for multiple hypothesis testing employing the Benjamini–Hochberg method, particularly in the realms of disparate proportions, expression differentials, and enrichment analyses. Results Cell annotation of synovial membranes from the scRNA-seq data preliminary revealed the heterogeneity among HA, OA, and RA synovium To elucidate the environmental and intracellular regulatory shifts within the synovial membranes of HA joints, and to assess their heterogeneity relative to OA and RA, we procured synovial membrane samples during TKA. Synovial tissue samples from HA patients were obtained following standardized protocols and subjected to scRNA-seq for subsequent downstream analysis (Supplement Figure S1A). Synovial tissues from HA, OA, and RA joints were collected using magnetic resonance imaging (MRI)-guided techniques for subsequent pathological examination (Supplement Figure S1B). Concurrently, we integrated our data with scRNA-seq datasets from OA and RA synovial membranes, sourced from the Gene Expression Omnibus (GEO) database. Following meticulous quality control measures and principal component analysis (Supplement Figure S2), we derived transcriptomic profiles from 37,456 cells. Utilizing Clustree (Supplement Figure S3), we determined the optimal clustering resolution at 0.2 for subsequent annotation. The transcriptomic expression characteristics of these 14 clusters are presented in Table S1. Employing the t-distributed stochastic neighbor embedding (t-SNE) algorithm for dimensionality reduction and visualization, we delineated eight predominant cell types (Fig. 1A). This categorization was corroborated with existing literature and the CellMarker 2.013 database (http://117.50.127.228/CellMarker/CellMarkerSearch.jsp). Fig. 1Dimensionality reduction, visualization, marker definition for each cell type, and cell statistics with inter-group comparisons from scRNA-seq: (A) T-SNE algorithm was used for dimensionality reduction and visualization of scRNA-seq annotations for HA, OA, and RA synovial tissues. (B) Visualization of marker genes for identifying cell types on the reduced dimensionality plot. (C) Statistics of various cell types and comparison of differences between groups, with significance levels indicated as ***: P < 0.001, **: P < 0.01, *: P < 0.1. Dimensionality reduction, visualization, marker definition for each cell type, and cell statistics with inter-group comparisons from scRNA-seq: (A) T-SNE algorithm was used for dimensionality reduction and visualization of scRNA-seq annotations for HA, OA, and RA synovial tissues. (B) Visualization of marker genes for identifying cell types on the reduced dimensionality plot. (C) Statistics of various cell types and comparison of differences between groups, with significance levels indicated as ***: P < 0.001, **: P < 0.01, *: P < 0.1. We specifically identified endothelial cells by their expression of PECAM1 and TM4SF1, fibroblasts by COL1A1 and COL1A2, smooth muscle cells (SMC) by ACTA2 and TAGLN, mesenchymal stem cells (MSC) by CENPF, monocytes and macrophages by LYZ and CD163, T cells by CCL5, B cells by MS4A1, and mast cells by TPSAB1 (Fig. 1B). The GO (Gene Ontology) biological function information for these 8 cell types can be found in Table S2. These cell types were then stratified into HA, OA, and RA groups for comparative cell proportion analysis (Fig. 1C). Our findings indicated a higher prevalence of mast cells within the immune cell population of HA synovium (P < 0.1). Endothelial cells were the most abundant in HA synovium (P < 0.01), contrasting with the predominance of fibroblasts in OA and RA synovium (P < 0.001). Furthermore, a higher incidence of smooth muscle cells was noted in HA synovium. This observation was corroborated by Masson’s trichrome staining of the specimens (Fig. 2A), which revealed pronounced fibrosis and an increased number of tubular structures in HA synovium. Fig. 2Combined results of synovial Masson staining, vWF immunohistochemistry, mast cell differential analysis, and marker validation demonstrate disease-specific differences in fibrosis, vascular features, and mast cell signatures.: (A) Overview of Masson staining for HA, OA, and RA synovial tissues, Masson staining at 4× magnification, and vWF immunohistochemistry at 40× magnification. (B) Quantitative analysis of vWF IHC staining among the three groups. ****: P < 0.0001; *: P < 0.1. (C) Gene expression volcano plot showing the inter-group differences between mast cells from HA and OA/RA synovium. (D) GO enrichment analysis represented as bidirectional bar charts for upregulated (red) and downregulated (blue) genes based on the differential genes from plot (A). (C) IHC staining validation and statistical analysis of TPSB2 across various synovial specimens: ****: P < 0.0001. Combined results of synovial Masson staining, vWF immunohistochemistry, mast cell differential analysis, and marker validation demonstrate disease-specific differences in fibrosis, vascular features, and mast cell signatures.: (A) Overview of Masson staining for HA, OA, and RA synovial tissues, Masson staining at 4× magnification, and vWF immunohistochemistry at 40× magnification. (B) Quantitative analysis of vWF IHC staining among the three groups. ****: P < 0.0001; *: P < 0.1. (C) Gene expression volcano plot showing the inter-group differences between mast cells from HA and OA/RA synovium. (D) GO enrichment analysis represented as bidirectional bar charts for upregulated (red) and downregulated (blue) genes based on the differential genes from plot (A). (C) IHC staining validation and statistical analysis of TPSB2 across various synovial specimens: ****: P < 0.0001. Endothelial cells, one of the two cell types capable of secreting the essential clotting factor von Willebrand factor (vWF), also serve as a marker for vascular injury29. Functioning as the primary carrier for FVIII, vWF contributes substantially to vascular repair while concurrently concealing antigenic determinants on exogenous FVIII14,15. This immunomodulatory action diminishes immune recognition and mitigates inhibitor formation risk. Quantitative analysis reveals significantly elevated vWF expression in HA synovium versus OA and RA tissues (p < 0.05) (Fig. 2A&B). The vascular allergic reactions caused by the degranulation of active mast cells in the synovium May be an immune environmental factor contributing to bleeding in HA synovium In our analysis of scRNA-seq data derived from synovial samples across three distinct pathological conditions, we meticulously annotated mast cells, which are recognized for their longevity within the immune system. Our cell count statistics revealed that the majority of mast cells identified by scRNA-seq are predominantly sourced from HA and OA synovium. We categorized these mast cells and conducted a comparative differential expression analysis between the groups. This analysis yielded 561 significantly upregulated genes (log2 Fold Change ≥ 0.5) within mast cells from HA synovium, notably featuring TPSB2, RACK1, MS4A2, ATP5IF1, CTSG, FOSB, YWHAZ, TXNIP, VWF, IL18, MBD2, CITED2, and POSTN. A Gene Ontology (GO) functional enrichment analysis of these upregulated genes highlighted terms associated with cellular responses, angiogenesis, mechanosensitivity, mast cell degranulation regulation, and wound healing responses. Conversely, we identified 385 downregulated genes (log2 Fold Change ≤ -0.5), which included SAT1, PRG4, GEM, ANXA2, FN1, TNFAIP6, CST7, PDCD6, DEFB1, GLUL, CCL4, BSG, TGFBI, and HEY1. The GO functional enrichment analysis for these downregulated genes uncovered terms related to cell adhesion, immune response, inflammation, complement activation, and the organization of collagen fibrils (Fig. 3A&B). Fig. 3The results of endothelial cell re-clustering analysis and dimensionality reduction visualization demonstrate the heterogeneity of endothelial cells in HA, OA, and RA synovium. The gene expression heatmap of endothelial cell subpopulations further reveals the gene expression characteristics of each endothelial cell subpopulation: (A) Re-clustering analysis and dimensionality reduction visualization of endothelial cells in synovium from the three diseases. (B) Gene expression heatmap for each endothelial cell subpopulation. The results of endothelial cell re-clustering analysis and dimensionality reduction visualization demonstrate the heterogeneity of endothelial cells in HA, OA, and RA synovium. The gene expression heatmap of endothelial cell subpopulations further reveals the gene expression characteristics of each endothelial cell subpopulation: (A) Re-clustering analysis and dimensionality reduction visualization of endothelial cells in synovium from the three diseases. (B) Gene expression heatmap for each endothelial cell subpopulation. TPSB2, a quintessential mast cell marker, exhibited a pronounced upregulation in HA synovium. The GO enrichment analysis for this gene pointed towards the regulation of degranulation and angiogenesis, suggesting heightened activity of mast cells within the HA synovium microenvironment. The upregulation of POSTN, CITED2, MBD2, FOSB, and TXNIP further implies a robust response to mechanical stimuli. To substantiate our findings, we performed immunohistochemical (IHC) staining for TPSB2 on synovial samples from the three diseases and conducted a statistical analysis of the results (Fig. 2C&D). This analysis confirmed and underscored the increased prevalence of mast cells in HA synovium relative to the other synovial types. The extensive degranulation of mast cells, potentially leading to vascular allergic reactions, may induce alterations in vascular permeability and bleeding episodes. This phenomenon could represent a significant immunological environmental factor contributing to the bleeding observed in HA synovium. The synovium of HA exhibits a new subset of endothelial cells with high expression of HMOX1 in response to ferroptosis In our comprehensive analysis of scRNA-seq data, we conducted a sub-clustering analysis of endothelial cells utilizing the re-clustering capabilities within the Seurat package. This analysis encompassed a total of 11,401 endothelial cells, which were re-clustered into five distinct clusters. Employing the t-SNE algorithm for dimensionality reduction and categorizing by disease, we discerned significant heterogeneity among endothelial cells across HA, OA, and RA synovium (Fig. 3A). We delineated the gene expression profiles of the five endothelial subgroups (EC0-4) and synthesized a heatmap for visual representation (Fig. 3B). Notably, EC0, predominantly identified in HA synovium, exhibited a distinct gene set with significant GO and KEGG enrichment analysis results (Fig. 4A&B). The GO enrichment analysis indicated that this subgroup is enriched with terms associated with active secretion, including “translation,” “protein folding,” “protein refolding,” “chaperone cofactor-dependent protein refolding,” “RNA processing,” and “cellular response to heat.” Specific genes such as HSPA6, HSP90AB1, HSPA1A, HSPA1B, HSP90AA1, and HMOX1 were prominently enriched in the categories of “chaperone cofactor-dependent protein refolding” and “cellular response to heat.” Literature suggests that these heat shock proteins are linked to mechanical stress and may be implicated in ferroptosis16,17. The complete GO enrichment results for the marker genes of each subgroup are summarized in Table S3. Fig. 4The GO and KEGG enrichment analysis results for the EC0, predominantly derived from HA synovium, suggest the presence of ferroptosis and stress responses to ferroptosis. Staining validations for the three disease synovial specimens are as follows: (A) GO enrichment analysis results for EC0. (B) KEGG enrichment analysis results for EC0. (C) Perl’s blue staining, HE staining, and GPX4 IHC staining for the three types of synovial specimens. (D) Quantitative analysis of Perl’s blue and GPX4 IHC staining results: ****: P < 0.0001. The KEGG pathway maps in (B) were generated from the KEGG database10,11. The GO and KEGG enrichment analysis results for the EC0, predominantly derived from HA synovium, suggest the presence of ferroptosis and stress responses to ferroptosis. Staining validations for the three disease synovial specimens are as follows: (A) GO enrichment analysis results for EC0. (B) KEGG enrichment analysis results for EC0. (C) Perl’s blue staining, HE staining, and GPX4 IHC staining for the three types of synovial specimens. (D) Quantitative analysis of Perl’s blue and GPX4 IHC staining results: ****: P < 0.0001. The KEGG pathway maps in (B) were generated from the KEGG database10,11. HA joints are known to exhibit substantial iron deposition due to recurrent bleeding, a phenomenon confirmed by Perl’s Blue staining. The accumulation of iron triggers cellular ferroptosis, with an upregulation of GPX4 serving as a compensatory mechanism, indicating the presence of ferroptosis in the synovium (Fig. 4C&D). The presence of ferric iron or the lipoxygenase-catalyzed peroxidation of unsaturated fatty acids on cell membranes can precipitate ferroptosis18. In HA synovium, the significant iron deposition suggests that, post-bleeding, a considerable amount of ferrous iron from hemoglobin is oxidized and deposited in the synovial environment, contributing to ferroptosis. The processes of iron transport and ferroptosis are intricately associated with HMOX119,20, and the KEGG enrichment results underscore the ferroptosis pathway. Pseudotime analysis revealed (Fig. 5A&B) that endothelial cells identified in HA synovium exhibit a lower degree of differentiation compared to those in OA and RA synovium. The pseudotime distribution heatmap of highly variable genes and the trajectory of each subgroup on the pseudotime axis (Fig. 5C) indicate that the marker genes of EC0 emerge at earlier pseudotime points, with HMOX1 showing an early appearance and an increased expression as pseudotime progresses. Immunofluorescence (IF) double staining results on HA synovial pathological specimens confirmed the presence of HMOX1+ endothelial cells (Fig. 5D). After intervention with RSL3 to induce ferroptosis in vitro cultured HUVECs by inhibiting the activity of Glutathione Peroxidase 4 (GPX4) (Fig. 6A&C), the migration rate of HUVECs was accelerated within 6 h (Fig. 6B&D), and the expression intensity of Heme Oxygenase-1 (HO-1) protein increased in proportion to CD31 (Fig. 6E&F). Upon ferroptosis stimulation, a high-expression HMOX1 endothelial cell subgroup emerges, characterized by robust expression of secretion function genes and low differentiation. Based on the function of this gene and protein, we infer that the differentiation of HMOX1+ endothelial cell subgroups represents the synovial endothelial response to ferroptosis, leading to neovascularization and significantly influencing the internal environment of HA synovium. Fig. 5Pseudotime analysis of identified endothelial cells to determine differentiation trajectories of each subtype and group, with specimen validation for the feature gene HMOX1 in EC0: (A) Visualization of pseudotime trajectory distribution for endothelial cells, where darker blue indicates proximity to the pseudotime start point and lower differentiation, while lighter blue indicates greater distance from the pseudotime start point and higher differentiation. (B) Visualization of pseudotime trajectory distribution for endothelial cells grouped by HA, OA, and RA. (C) Heatmap of highly variable genes along pseudotime progression in endothelial cells, with arrows indicating the pseudotime trajectory from bottom to top. (D) Visualization of pseudotime trajectory distribution for subpopulations EC0-5. (E) IF dual staining results for CD31 (PECAM1) and HO-1 (HMOX1) in HA synovial specimens. Pseudotime analysis of identified endothelial cells to determine differentiation trajectories of each subtype and group, with specimen validation for the feature gene HMOX1 in EC0: (A) Visualization of pseudotime trajectory distribution for endothelial cells, where darker blue indicates proximity to the pseudotime start point and lower differentiation, while lighter blue indicates greater distance from the pseudotime start point and higher differentiation. (B) Visualization of pseudotime trajectory distribution for endothelial cells grouped by HA, OA, and RA. (C) Heatmap of highly variable genes along pseudotime progression in endothelial cells, with arrows indicating the pseudotime trajectory from bottom to top. (D) Visualization of pseudotime trajectory distribution for subpopulations EC0-5. (E) IF dual staining results for CD31 (PECAM1) and HO-1 (HMOX1) in HA synovial specimens. Fig. 6After in vitro intervention of HUVECs with the GPX4 inhibitor RSL3, the migration area of HUVECs and the protein fluorescence intensity ratio of HO-1 and CD31 were upregulated: (A & C) After RSL3 intervention, the expression of GPX4 in HUVECs decreased; (B & D) The rate of increase in the migration area of HUVECs within 0–6 h post-RSL3 intervention was greater; (E & F) After 24 h of RSL3 intervention, the fluorescence intensity ratio of HO-1 and CD31 proteins in HUVECs was upregulated. (*: P < 0.1). After in vitro intervention of HUVECs with the GPX4 inhibitor RSL3, the migration area of HUVECs and the protein fluorescence intensity ratio of HO-1 and CD31 were upregulated: (A & C) After RSL3 intervention, the expression of GPX4 in HUVECs decreased; (B & D) The rate of increase in the migration area of HUVECs within 0–6 h post-RSL3 intervention was greater; (E & F) After 24 h of RSL3 intervention, the fluorescence intensity ratio of HO-1 and CD31 proteins in HUVECs was upregulated. (*: P < 0.1). Cellular communication reveals that the HMOX1+ endothelial cell induces the proliferation and differentiation of the SCARA5 fibroblast subpopulation through FTL/SCARA5 The detection of HMOX1+ endothelial cells in HA synovium signifies the presence of ferroptosis stress processes within this tissue. Building on this finding, we proceeded to perform a sub-clustering analysis of fibroblasts (Fig. 7A). Leveraging database information, we categorized the THY1-positive expression group as Sublining_FB and the PRG4-positive expression group as Lining_FB (Fig. 7B). To further delineate the characteristics of other fibroblast subgroups, we scrutinized their gene expression profiles and conducted GO enrichment analysis. Fig. 7Re-clustering of fibroblasts from synovium samples of the three diseases and analysis of transcriptomic features of each subpopulation: (A) Dimensionality reduction and visualization of fibroblast subpopulations using the t-SNE algorithm. (B) Expression levels of PRG4 and THY1 in fibroblasts. Higher expression of PRG4 is associated with lining fibroblasts, while higher expression of THY1 is associated with sublining fibroblasts. (C) & (D) Gene expression heatmaps and GO enrichment matrix plots for regulatory fibroblasts, angiogenic fibroblasts, and inflammatory fibroblasts as shown in panel (A). Re-clustering of fibroblasts from synovium samples of the three diseases and analysis of transcriptomic features of each subpopulation: (A) Dimensionality reduction and visualization of fibroblast subpopulations using the t-SNE algorithm. (B) Expression levels of PRG4 and THY1 in fibroblasts. Higher expression of PRG4 is associated with lining fibroblasts, while higher expression of THY1 is associated with sublining fibroblasts. (C) & (D) Gene expression heatmaps and GO enrichment matrix plots for regulatory fibroblasts, angiogenic fibroblasts, and inflammatory fibroblasts as shown in panel (A). The Regulatory FB subgroup demonstrated elevated expression of genes such as ANK2, ID4, HSPA1A, DDX3X, FTX, ACTA2, and PLCG2. The GO enrichment analysis for this subgroup emphasized terms associated with the positive regulation of transcription by RNA polymerase II, negative regulation of transcription by RNA polymerase II, RNA splicing, mRNA processing, regulation of DNA-templated transcription, chromatin organization, protein phosphorylation, and other transcription-related processes. The Angiogenesis FB subgroup exhibited high expression of genes including PECAM1, TM4SF1, VWF, ACKR1, SELE, FABP5, CALCRL, PLVAP, SYNE2, and A2M. GO enrichment for this subgroup encompassed terms related to angiogenesis, positive regulation of transcription by RNA polymerase II, signal transduction, cell differentiation, cell adhesion, negative regulation of DNA-templated transcription, protein phosphorylation, and positive regulation of cell migration. The Inflammatory FB subgroup showed high levels of expression for IGKC, IL7R, CXCR4, SRGN, PTPRC, CCL5, TNFAIP3, CCL4, CD52, and CD69. GO enrichment for this subgroup revealed terms associated with immune response, adaptive immune response, antigen processing and presentation, inflammatory response, and T cell activation (Fig. 7C&D). Utilizing the CellphoneDB R package, we analyzed cell-cell interactions among various cell types in the synovium across the three diseases. The results indicated that endothelial cell subgroups in HA synovium interact intimately with various fibroblast subgroups (Fig. 8A), with FTL and FTH1, expressed by endothelial cells, serving as ligands for the SCARA5 on fibroblasts (Fig. 8B). Ferritin Light Chain (FTL) and Ferritin Heavy Chain 1 (FTH1) collectively store iron in a non-toxic form and regulate its release. While FTL stabilizes the ferritin complex without enzymatic function, FTH1 oxidizes Fe²⁺ via ferroxidase activity. Scavenger Receptor Class A Member 5 (SCARA5) is a cell-surface receptor within the scavenger receptor superfamily, characterized by a collagen-like domain and an N-terminal scavenger receptor cysteine-rich domain. Functionally, it binds serum ferritin and mediates its endocytosis, playing a critical role in non-transferrin iron delivery and cellular iron homeostasis21. Based on gene expression characteristics, SCARA5+ fibroblasts were identified in all fibroblast subgroups, with a higher proportion or greater expression of SCARA5+ fibroblasts in HA synovium (Fig. 8C). FTL and FTH1 are integral components of ferritin and are essential proteins in the ferroptosis process18. SCARA5 is a crucial surface receptor for ferritin recycling22, and related studies suggest that SCARA5 is a key regulator of ferroptosis23. We confirmed the presence of SCARA5+ fibroblast subgroups in HA synovium through immunofluorescence (IF) staining (Fig. 8D). Fig. 8Cell communication analysis results and identification of SCARA5 + fibroblasts in synovium samples from the three diseases: (A) Heatmaps of cell communication between different cell types identified in HA, OA, and RA synovium. (B) 2D scatter plots showing the intensity of receptor-ligand interactions in cell communication within HA, OA, and RA synovium. (C) Expression levels of SCARA5 across different fibroblast subpopulations in each disease group. (D) IF dual staining results using COL1A1 and SCARA5 antibodies in HA synovium. Blue fluorescence indicates DAPI staining of cell nuclei, green fluorescence indicates COL1A1 protein expression, red fluorescence indicates SCARA5 protein expression, and yellow fluorescence represents co-expression of both proteins. Cell communication analysis results and identification of SCARA5 + fibroblasts in synovium samples from the three diseases: (A) Heatmaps of cell communication between different cell types identified in HA, OA, and RA synovium. (B) 2D scatter plots showing the intensity of receptor-ligand interactions in cell communication within HA, OA, and RA synovium. (C) Expression levels of SCARA5 across different fibroblast subpopulations in each disease group. (D) IF dual staining results using COL1A1 and SCARA5 antibodies in HA synovium. Blue fluorescence indicates DAPI staining of cell nuclei, green fluorescence indicates COL1A1 protein expression, red fluorescence indicates SCARA5 protein expression, and yellow fluorescence represents co-expression of both proteins. Thus, we infer that the ferroptosis stress-induced high expression of HMOX1 in endothelial cells can promote the recycling and metabolism of ferritin by fibroblasts through the FTL/SCARA5 pathway. This mechanism has significant implications for regulating ferroptosis in HA synovium and potentially improving recurrent bleeding within the synovium. Discussion Hemophilia A (HA) is characterized by deficient or dysfunctional coagulation factor VIII (FVIII), leading to spontaneous or trauma-induced hemorrhages in joints, muscles, and deep tissues. Recurrent intra-articular bleeding under low coagulability drives progressive joint damage—distinct clinically from osteoarthritis (OA) and rheumatoid arthritis (RA). HA joints exhibit rigidity, muscle contractures, and deformities from chronic disuse, whereas OA and RA present contrasting pathology. MRI reveals similar joint space narrowing between the distal femur and tibial plateau in HA and RA, alongside synovial proliferation. However, HA synovium demonstrates hemorrhagic proliferation, contrasting with RA’s inflammatory pannus-covered synovium. Although daily high-dose FVIII replacement remains central to HA management, it carries inhibitor development risks. Alternative therapies under investigation fail to address acute symptoms like joint swelling, mobility loss, and intractable pain. Therefore, we performed the first transcriptomic analysis of full synovial single-cell suspensions in HA. This delineates synovial cellular subtypes and transcriptomic profiles, establishing at single-cell resolution that HA synovium is fundamentally distinct from OA and RA synovium. In this study, we have unveiled, for the first time, distinct cell type subgroups and transcriptional profiles within the synovial membranes of HA patients at the single-cell level. Our findings reveal a stark contrast in the cellular landscape and molecular signatures when compared to those observed in OA and RA patients, shedding new light on the pathophysiological underpinnings of HA and potentially informing future therapeutic strategies. The synovial membrane, a specialized form of loose connective tissue, plays a pivotal role in the joint by secreting essential nutrients that sustain the health and function of the articular cartilage and surrounding structures. ScRNA-seq has emerged as a powerful tool for dissecting the cellular heterogeneity within the synovial membrane, having previously identified a multitude of cell populations in RA and OA5–7, encompassing fibroblasts, endothelial cells, and a spectrum of immune cells. In the present study, we conducted a comparative analysis of synovial membranes from patients with HA, obtained during surgical interventions. Our findings revealed a notable increase in the abundance and proportional representation of endothelial cells within HA synovial membranes, setting them apart from those observed in RA and OA. This distinction suggests a unique vascular phenotype in HA, which may have profound implications for understanding the disease pathogenesis and could inform targeted therapeutic interventions. Mast cells24, as long-lived, tissue-resident secretory immune cells, are equipped with numerous cytoplasmic granules that can number in the hundreds. Employing TPSAB1 and TPSB2 as definitive markers, we meticulously annotated mast cells from the synovial scRNA-seq data, drawing upon comprehensive databases and scholarly literature. In the context of HA synovium, mast cells exhibit heightened expression of TPSB2, which encodes tryptase β-2, a predominant neutral protease found within mast cells and released during activation and degranulation processes. Mast cell activation can be precipitated by a variety of stimuli, including cytokines, toxins, thermal changes, or mechanical trauma, leading to a cascade of allergic reactions. As persistent immune sentinels within the perivascular tissue, mast cells are instrumental in modulating vascular tone, regulating permeability, promoting angiogenesis, and contributing to fibrotic processes. Their involvement in these functions is crucial for understanding their link to hemorrhagic disorders. Our study reveals that, in comparison to OA and RA synovium, HA synovium harbors a significantly elevated mast cell population. The attendant alterations in vascular permeability and the exacerbation of fibrosis heighten the susceptibility and propensity for vascular bleeding episodes. This not only elucidates the etiology of intra-articular bleeding but also suggests that the elevated mast cell count in HA synovium contributes to increased vascular fragility and a heightened risk of recurrent bleeding. Conversely, in OA synovium, mast cells are predominantly quiescent and less numerous, while in RA synovium, the unique immunological milieu redirects research focus away from mast cells. Thus, the robust transcriptomic activity observed in mast cells within HA synovium positions them as potential cellular targets for interventions aimed at mitigating bleeding and rebleeding episodes. However, single-cell transcriptomics has only revealed a weak association between the transcriptomic characteristics of activated mast cells and vascular hemorrhage in HA synovium, and the causal relationship warrants further investigation. Examining the alterations in the synovial matrix microenvironment subsequent to hemorrhagic events is pivotal for deciphering the etiology of recurrent bleeding episodes. In our analysis of the HA synovial scRNA-seq, we identified a highly active EC0 cluster that expressed stress-related genes associated with ferroptosis. This is likely related to the significant release of hemoglobin outside the vasculature, which undergoes oxidation and forms iron-containing hemichrome deposits25. Among these, the key protein HO-1 and its critical gene HMOX1 were prominently expressed in the EC0 cluster. We quantified the expression intensity of HO-1 in endothelial cells following RSL3-mediated ferroptosis through cellular fluorescence. This confirmed the high expression of HMOX1 induced by ferroptosis, with an increase in the differentiation of the HMOX1 + cluster. A pivotal protein in this context is HO-1, which is markedly upregulated in the EC0 cluster. HO-1 facilitates the breakdown of hemoglobin into biliverdin, carbon monoxide (CO), and free ferrous iron26. The profuse hemoglobin liberated in HA synovial membranes oxidizes free ferrous iron to its trivalent state, thereby triggering ferroptosis in synovial cells. This cascade of events leads to vascular damage, stimulates endothelial proliferation, prompts the release of vWF to ensnare platelets, and fosters the proliferation and migration of endothelial subgroups within the EC0. Concurrently, this process is accompanied by the expression of a suite of stress-related genes. HO-1, also known as HMOX1, is a stress protein that plays a crucial role in ferroptosis, underscoring the significant pathological implications of this endothelial cell subset in HA. The heightened expression of HO-1 in EC0 endothelial cells suggests a critical role in the pathogenesis of recurrent bleeding in HA, potentially serving as a therapeutic target to mitigate the vascular fragility and propensity for rebleeding observed in this condition. Abundant research evidence underscores the pivotal roles of synovial fibroblasts and endothelial cells in the pathogenesis of joint diseases. For instance, during the progression of RA, these cell types engage in a complex interplay, co-expressing proteins that exert both pro-inflammatory and anti-inflammatory influences, which are crucial in shaping the disease’s inflammatory landscape27. Similarly, in the context of OA, synovial fibroblasts secrete vascular endothelial growth factor (VEGF), which stimulates angiogenesis and endothelial cell-mediated vascular formation. This process contributes to pathological synovial proliferation and subsequent degradation of articular cartilage, thereby hastening the progression of OA28. Conversely, endothelial cells exert reciprocal effects on fibroblasts, as evidenced by scRNA-seq results in this study. This bidirectional communication between endothelial cells and fibroblasts is integral to the intricate cellular crosstalk that drives the pathophysiological changes observed in joint diseases, highlighting the interconnected nature of these cellular interactions in disease pathology. Ferroptosis, a form of regulated cell death characterized by iron-dependent lipid peroxidation, exerts significant stress on endothelial cells within the HA synovium. This stress triggers the differentiation of a specialized endothelial cell subset that expresses elevated levels of HO-1, also known as HMOX1. HO-1 plays a critical role in the degradation of hemoglobin, leading to the release of free ferrous iron ions. Concurrently, these cells synthesize ferritin light chain (FTL) and ferritin heavy chain (FTH1), which are key components of the ferritin complex responsible for iron storage and homeostasis. The synthesis of FTL and FTH1, in turn, induces a subset of fibroblasts that highly express SCARA5, a surface protein with a well-established role in regulating ferroptosis. Initially identified in hepatic epithelial cells, SCARA5 is primarily involved in the endocytosis and clearance of vWF, coagulation FVIII, or their complexes within the liver23. The presence of SCARA5+ fibroblasts in the HA synovium is a novel finding that underscores the active process of ferroptosis in this tissue. The synovium’s response to ferroptosis appears to be initiated by endothelial cells, highlighting the intricate interplay between these two cell types. Furthermore, the identification of SCARA5+ fibroblasts opens new avenues for therapeutic intervention. Targeting SCARA5 could potentially enhance the recycling and metabolic pathways of ferritin, thereby modulating ferroptosis. Such interventions may reduce the propensity for recurrent bleeding episodes in the synovium by stabilizing the vascular environment and mitigating the oxidative stress associated with hemoglobin degradation. This approach offers a promising strategy for managing the vascular complications observed in HA, where recurrent bleeding poses a significant challenge to joint health and function. Conclusions This study presents the first single-cell transcriptomic atlas of Hemophilic Arthropathy (HA) synovium. Comparative analysis with osteoarthritis and rheumatoid arthritis revealed disease-specific cellular heterogeneity. We identified activated perivascular mast cells and an endothelial cell-fibroblast interaction hub under ferroptosis stress in HA. A key discovery is a SCARA5 + fibroblast subpopulation that regulates ferroptosis, representing a promising target for gene-based therapy to mitigate recurrent bleeding. While limited by sample size, our findings provide a foundational resource for understanding HA pathophysiology and unveil novel therapeutic avenues. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4