In mice, the healing capacity for tendon injuries is high up to 2 weeks of age, but decreases as the musculoskeletal system matures at 6 weeks (Somerville et al., 2004; Howell et al., 2017). To investigate whether this phenomenon is due to changes in the cellular population of tendon tissue, including fluctuations in progenitor cells, we performed scRNA-seq on mouse Achilles tendons at these two time points.

Achilles tendons were harvested from mice at 2 and 6 weeks of age. Following collagenase digestion and dead cell removal, we employed massively parallel, droplet-enabled scRNA-seq analysis (10x Genomics Chromium). Data processing was conducted using the Cell Ranger pipeline (10x Genomics). We analyzed 10,314 cells from 2-week-old mice (median 3167 genes/cell and 41,611 mean reads/cell) and 6513 cells from 6-week-old mice (median 732 genes/cell and 60,386 mean reads/cell). After doublet removal using DoubletFinder, we merged the datasets and performed unbiased clustering using Seurat, identifying a total of 15 clusters (Figure 1A). The top differentially expressed genes (DEGs) for each cluster, based on log2 fold change and statistical significance, are summarized in Figure 1B.

We identified two clusters (0 and 11) with high expression of Mkx and Scx, TFs involved in tendon development and growth. Cluster 0 showed higher expression of ECM-related genes such as Col1a1 and Fmod compared to cluster 11, suggesting these were tenocytes (TC). Cluster 11, with high Periostin (Postn) expression (Jacobson et al., 2020), was classified as myotendinous junction (MTJ). Other clusters were characterized as follows: cartilage (CA, high SRY-box transcription factor 9 [Sox9] expression), lymphocytes (LC1, LC2, high Lysozyme 2 [Lyz2] and Complement component 1, q subcomponent, alpha polypeptide [C1qa]), endothelial cells (EC, high Platelet and endothelial cell adhesion molecule 1 [Pecam1]), red blood cells (RBC, high Hemoglobin alpha, adult chain 1 [Hba-a1]), smooth muscle cells (SM, high Myosin light chain 9 [Myl9]), proliferating cells (PC, high Mki67 and Stathmin 1 [Stmn1]), Schwann cells (SW, high Myelin basic protein [Mbp]), vascular endothelial cells (VEC, high Multimerin 1 [Mmrn1]), and macrophages (MC, high Protein tyrosine phosphatase receptor type C [Ptprc]). In cluster 3, Col1a1 and Fmod were expressed; however, Gm42418 and AY036118 were highly expressed. These long non-coding RNAs are related to RN45s and therefore cluster 3 represents ribosomal contamination (Isola et al., 2024). It is also speculated that the cartilage cluster (CA) reflects enthesis, with expression of Scx, as well as Sox9 (Zhang et al., 2023). DEGs of each cluster were summarized in Supplementary files 1–15.

To identify progenitor populations within these clusters, we analyzed expression patterns of previously reported markers Tppp3 and Pdgfra (Takakura et al., 1997; Morikawa et al., 2009; Harvey et al., 2019; Tachibana et al., 2022), along with the known stem/progenitor cell marker Ly6a (Holmes and Stanford, 2007; Sung et al., 2008; Hittinger et al., 2013; Sidney et al., 2014; Fang et al., 2022; Figure 1—figure supplement 1B). We identified subclusters within clusters 1 and 4 showing high expression of these genes, which we defined as SP1 and SP2. SP2 exhibited the highest expression of these genes, suggesting it had the strongest progenitor characteristics. It has been reported (Harvey et al., 2019) that the Tppp3-positive population is localized to peritenon, and SP clusters might reflect peritenon as well.

SP2 also showed strong expression of genes associated with early tendon development in mouse embryos, as identified by RNA-seq of mouse limb tendon cells at E14.5 (Havis et al., 2014) and the Eurexpress mouse embryo transcriptome atlas database (Diez-Roux et al., 2011; Figure 1—figure supplement 2A). Gene Ontology (GO) analysis revealed upregulation of tendon development-related pathways in SP2, including TGFβ production and collagen-containing extracellular matrix (Figure 2—figure supplement 2B). Evaluation of cell numbers in each cluster showed a decrease in SP2, TC, and MTJ at 6 weeks (Figure 1C, Figure 1—figure supplement 1A).

We then examined DEGs in SP2 compared to SP1 (Figure 1D). The whole list of DEGs comparing SP2 and SP1 was summarized in Supplementary file 16. Among these, we identified Cd34, Cd55, and Cd248 as genes encoding surface antigens. Cd34 is known to be highly expressed in mouse embryonic limb bud E14.5 compared to E11.5 (Havis et al., 2014), suggesting that Cd34-positive cluster might reflect a progenitor cell that consists of the limb bud, including tendons. However, Cd55 and Cd248 have not been discussed in this context. Expression of Cd55 and Cd248 was localized to areas corresponding to SP1 and SP2 (Figure 1E). We evaluated the correlation coefficients between the expression of these genes and previously suggested TSPC markers (Cd73, Cd90, Cd105, Cd44, Cd146), as well as Tppp3, Pdgfra, and Ly6a at 2 weeks of age. Cd34, Cd55, and Cd248 showed high correlation, suggesting these genes as new candidate surface antigens for TSPCs (Figure 1F).

Comparing gene expression between 2 and 6 weeks, we observed a decrease in Mkx and Scx expression, consistent with previous reports (Grinstein et al., 2019). Furthermore, the expression of our candidate tendon progenitor cell markers Cd34, Cd55, and Cd248 all decreased at 6 weeks. Inflammatory cells, including macrophages, increased at 6 weeks (Figure 1G). Gene expression analysis revealed a shift from M2 macrophage-related genes (Mannose receptor C-type 1 [Mrc1] and Cathepsin C [Ctsc]) at 2 weeks to M1 macrophage-related genes (Chemokine [C-C motif] ligand 6 [Ccl6] and Chemokine (C-C motif) ligand 9 [Ccl9]) at 6 weeks. Comparative analysis between 2-week and 6-week tendons was summarized as a heatmap in Figure 1—figure supplement 3.

Analysis of cell-cell communication changes in postnatal tendon and surrounding tissues (Lui and Wong, 2019) using CellChat (Jin et al., 2025) showed a significant decrease in overall interaction at 6 weeks compared to 2 weeks. These results demonstrate that the cellular profile of tendon tissue changes dramatically at 6 weeks, with decreased expression of stem progenitor markers and reduced interaction with surrounding tissues.

Gene regulation is mediated by the binding of TFs to cis-regulatory elements proximal to the gene. Consequently, epigenetic changes such as chromatin accessibility play a crucial role in gene expression (Miller et al., 2014). Moreover, TFs are typically expressed at low levels, potentially leading to false negatives in scRNA-seq due to detection limits. Chromatin accessibility changes often precede gene expression changes, potentially allowing us to predict transcriptional changes (Ranzoni et al., 2021). Therefore, we performed simultaneous ATAC-seq and RNA-seq at the single-cell level to identify gene expression changes and their associated epigenetic alterations, aiming to elucidate the postnatal growth mechanisms of tendon tissue and evaluate the validity of Cd55 and Cd248 as markers.

We extracted nuclei from cells isolated from 2-week-old mouse Achilles tendons and conducted droplet-enabled multi-omics analysis (10x Genomics Chromium), performing simultaneous snATAC-seq and snRNA-seq as a reference. We first analyzed the snATAC-seq data. After mapping reads to the genome and calling peaks, we annotated the location of peaks in terms of genomic features. The peaks were associated with promoters, introns, exons, and intergenic regions.

We evaluated 6571 cells (median high-quality fragments per cell: 17,408) and clustered them using latent semantic indexing and uniform manifold approximation and projection (UMAP) with the R package Signac (Stuart et al., 2021; Figure 2A). Given the limited knowledge of cell-specific chromatin accessibility, we assessed gene activity by computing counts per cell within the gene body and promoter of protein-coding genes. Using this gene activity data, we performed unsupervised clustering and identified 17 clusters (ground-truth annotation). Clusters expressing tendon-related genes such as Mkx and Scx were identified as clusters 1 and 6, defined as A1: TC1 and A6: TC2, respectively. Clusters 0, 14, and 15 were identified as expressing stem progenitor markers, including Cd34, Ly6a, Pdgfra, Cd55, and Cd248, and were defined as A2-0: SP1, A2-14: SP2, and A2-15: SP3, respectively. The characteristic gene activities and annotations for each cluster are summarized in Figure 2B.

Next, we analyzed the snRNA-seq data. We evaluated 10,314 cells (median 3167 genes/cell and 41,611 mean reads/cell). After doublet removal, we performed unbiased clustering using Seurat and identified 17 clusters (Figure 2—figure supplement 1A). Clusters expressing tendon-related genes Mkx and Scx were clusters 1 and 13, defined as R2-1: TC_1 and R2-13: TC_2, respectively. Clusters 2, 3, and 7 were identified as expressing stem/progenitor-related genes and defined as R2-2: SP_1, R2-3: SP_2, and R2-7: SP_3, respectively. To validate the consistency of our approaches, we conducted a comprehensive comparison between the cell clusters identified in our scRNA-seq analysis (which included both 2-week and 6-week samples) and our snRNA-seq analysis. We confirmed that all clusters identified in the scRNA-seq data could be similarly annotated in the snRNA-seq data, demonstrating strong concordance between these complementary approaches (Figure 2—figure supplement 2). This validation supports the reliability of our cell-type identification across different single-cell methodologies. Interestingly, R2-5: The MTJ cluster expressed not only *Tnc* but also *Sox9*, consistent with previous reports (Nagakura et al., 2020). Using the Signac package’s ‘FindTransferAnchors’ function, we calculated predicted IDs for snATAC-seq clusters based on snRNA-seq annotations (predicted annotation) (Figure 2C). Evaluation of the two annotation methods for snATAC-seq revealed that both methods allowed annotation of major cell types, and correlation between the two was maintained. Thus, the validity of predicted annotations based on snRNA-seq was confirmed (Figure 2D).

R2-7: SP3, considered the most undifferentiated stem/progenitor fraction in snRNA-seq, could be divided into A2-0: SP1 and A2-14: SP2 in snATAC-seq.

Comparing gene activity of tendon-related and stem/progenitor genes in each cluster of the ground-truth annotation, A2-14: SP2 showed high expression of Ly6a, Pdgfra, and Tppp3, and inverse correlation with Mkx, Scx, and Tnmd. Furthermore, the newly identified surface antigens Cd55 and Cd248 also showed high expression in cluster 14. In contrast, Cd34 did not show a high expression pattern in A2-14: SP2. These results suggest that A2-14: SP2 is the most immature cluster, with Cd55 and Cd248 showing characteristic expression (Figure 2E). Subcluster analysis of R7, considered the most immature fraction in snRNA-seq, revealed two clusters based on Cd55/Cd248 and Cd34 expression, similar to the snATAC-seq analysis. The former (R7-1) showed high expression of Ly6a, Tppp3, and Pdgfra (Figure 2F). Trajectory analysis using R2-7 and A2-14 as roots in snRNA-seq and snATAC-seq, respectively, showed increased expression of tendon-related genes Mkx and Scx along the trajectory, while expression of Tppp3, Cd55, and Cd248 decreased (Figure 3A). Evaluation of peaks in each cluster based on gene expression levels in snRNA-seq revealed multiple peaks with heights correlated to gene expression for Mkx and Scx. A similar evaluation of Cd55 and Cd248 showed multiple peaks upstream of the transcription start site (TSS) for Cd248 (Figure 3B). These results indicate correlation between snRNA-seq and snATAC-seq, and consistent with scRNA-seq results, suggest that A2-14 represents the most undifferentiated TSPCs based on Cd55 and Cd248 expression.

To confirm the presence of the identified Cd55/Cd248 fraction at 6 weeks, we performed snRNA-seq and snATAC-seq on 6-week-old mouse Achilles tendon cells. Using the 2-week snATAC-seq data as a reference, we annotated the 6-week snATAC-seq clusters. Clusters A2-14 and A2-0 corresponded to clusters A6-12 and A6-0, respectively (Figure 3—figure supplement 1). We also annotated the 6-week snATAC-seq clusters using 2-week snRNA-seq with the same results (Figure 3—figure supplement 2). Gene activity evaluation revealed that, similar to the 2-week data, the 6-week A6-12 and A6-0 clusters showed Cd55/Cd248 high and Cd34 dim, and Cd55/Cd248 dim and Cd34 high patterns, respectively. Furthermore, annotation of the 6-week snRNA-seq data using the 2-week snRNA-seq as a reference revealed clusters with expression patterns similar to those observed in snATAC-seq. In both cases, clusters with high Cd55 and Cd248 expression showed high gene activity (snATAC-seq) and gene expression (snRNA-seq) of Tppp3, Pdgfra, and Ly6a. These results confirm that the Cd55 and Cd248 high clusters identified in the 2-week snRNA-seq+snATAC-seq analysis are similarly detected at 6 weeks.

We also compared the snATAC-seq data between 2 and 6 weeks. No significant differences were observed in genomic annotations between the two time points (Figure 3—figure supplement 3). After merging the datasets (Figure 3—figure supplement 3B), we compared gene activity of tendon/stem-related genes to evaluate changes in gene activity. We found that the activity of Tppp3, Pdgfra, Ly6a, Cd55, Cd248, and Cd34 all decreased at 6 weeks (Figure 3—figure supplement 3C). This was consistent with the decreased gene expression observed when comparing 2-week and 6-week data.

To estimate TF activity in each cluster, we used the Single-Cell Regulatory Network Inference and Clustering (SCENIC) package (Aibar et al., 2017) to calculate gene regulatory network activity from scRNA-seq gene expression data. SCENIC constructs gene expression networks centered on TFs and infers TF activity in each cluster. We summarized the predicted TF activities in tendon and stem/progenitor-related clusters R2-7, R2-2, R2-3, R2-1, and R2-13 identified by snRNA-seq (Figure 4).

Next, we evaluated the gene activity of these identified TFs in each snATAC-seq cluster, along with motif activity calculated by chromVAR (Schep et al., 2017). We also assessed the expression levels of TFs in each snRNA-seq cluster. In the most immature fractions, represented by cluster A2-14 in snATAC-seq and cluster R2-7 in snRNA-seq, KLF transcription factor 3 (Klf3), KLF transcription factor 4 (Klf4), and cAMP responsive element binding protein 5 (Creb5) showed consistent behavior in gene activity, motif activity, and gene expression. Additionally, Signal transducer and activator of transcription 2 (Stat2) and cAMP responsive element binding protein 3 like 1 (Creb3l1) showed high values in A2-0/A2-15 and A2-1, respectively. Stat2 gene expression was not observed, likely due to false positives resulting from low expression levels. The gene activity and motif activity of each TF estimated by snATAC-seq correlated with the transcription factor activity calculated by SCENIC. Therefore, simultaneous analysis of snRNA-seq and snATAC-seq could be used to more accurately evaluate the function of TFs in each cluster.

Furthermore, SCENIC can predict candidate TFs regulating each gene (Table 1). Cd55 and Cd248 were predicted to be regulated by Klf3 and Klf4, while Mkx and Scx were predicted to be under the control of Creb3l1. These predictions were consistent with the data calculated from gene activity and motif activity in snATAC-seq.

Immunostaining evaluation of CD55 and CD248, the identified candidate stem/progenitor markers, revealed expression in the peritenon (Figure 5C and A). This was consistent with previous reports of localized expression of Tppp3/Pdgfra-positive cells in the peritenon and our analysis showing co-expression of Tppp3/Pdgfra and Cd55/Cd248. The results support the possibility that SP clusters reflect peritenon. Given that Cd55 and Cd248 expression appeared to reflect Tppp3, Pdgfra, and Ly6a expression more sensitively than Cd34, we extracted tendon cells from 2-week-old mice and sorted them by fluorescence-activated cell sorting (FACS) to determine the biological phenotype of Cd55- and Cd248-positive cells (Figure 5B).

Gene expression analysis of sorted cells showed that CD55/CD248-positive cells, compared to -negative cells, had lower expression of Mkx, Scx, Col1a1, and Creb3l1, but higher expression of Ly6a, Tppp3, Pdgfra, Creb5, and Klf3, consistent with snRNA-seq analysis (Figure 5—figure supplement 1). To assess stem/progenitor capacity, we performed colony formation assays. CD55/CD248 sorted cells showed significantly increased colony formation compared to CD55/CD248-negative cells (control) (Figure 5C).

We then evaluated the differentiation tendencies of these cells toward tendon, cartilage, and bone. Tendon differentiation resulted in clusters of spindle-shaped cells from CD55/CD248-positive cells, suggesting tenogenic differentiation (Figure 5D). Gene expression analysis revealed that while CD55/CD248-positive cells initially had lower Mkx and Scx expression than negative cells, this pattern reversed after differentiation (Figure 5E). Increased expression of other tendon-related genes such as Col1a1 and Bgn was also observed. In contrast, no significant increases in Sox9 and Collagen type II alpha 1 chain (Col2a1) (chondrogenic) Craft et al., 2013 or RUNX family transcription factor 2 (Runx2), and Alkaline phosphatase, biomineralization associated (Alpl) (osteogenic) expression were observed compared to negative cells during cartilage and bone differentiation assays (Figure 5—figure supplement 2). Cells negative for CD55/CD248 could be mixed cell populations, including hematopoietic lineages, cells from tendon mid-substance, immune cells, and/or endothelial cells. However, given that CD55/CD248-negative cells have increased expression of genes characteristic of tendons such as Mkx, Scx, and Col1a1, we could speculate that this popularity might reflect terminally differentiated tenocytes and CD55/CD248-positive cells possess tenogenic differentiation capacity.

To further evaluate the tenogenic potential of CD55/CD248-positive cells, we created tendon-like tissue (bio-tendon) using our previously reported 3D stretch stimulation culture system. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) analysis to assess collagen fiber density and thickness showed that bio-tendons derived from CD55/CD248-positive cells had an increased proportion of collagen fibers parallel to the stretch direction and increased collagen fiber diameter. TEM also revealed characteristic banding patterns and triple helix structures in these bio-tendons, indicating mature collagen organization (Wieczorek et al., 2015; Figure 5F–H). Stretch tests to evaluate the mechanical capacity of the bio-tendons showed high tensile strength (Figure 5I).

These results demonstrate that CD55/CD248-positive cells have a tendency to differentiate into tendon cells.

Mice were euthanized by exposure to CO₂ followed by cervical dislocation. To remove Achilles tendons, a longitudinal incision through the skin was made down the midline of the posterior aspect of the lower limb. A sharp transverse incision was made just distal to the myotendinous junction and again just proximal to the enthesis, and the Achilles tendon was carefully removed. The procedure was performed bilaterally. All animals were purchased from Sankyo Lab Service (Tokyo, Japan). Achilles tendons were digested to obtain a single-cell suspension. In total, 40 or 60 (2 weeks and 6 weeks, respectively) mouse Achilles tendons were processed together as a single sample. After washing with 1× phosphate-buffered saline (PBS) several times and cutting 1-mm-wide sections using scissors, tendons were digested for 1 hr at 37°C with continuous shaking at 1200 rpm and 37°C in a dissociation solution consisting of 30 mg/mL collagenase (Wako, Osaka, Japan). After digestion, the single-cell suspension was filtered for debris using a 40 μm cell strainer and washed twice with 1× PBS. The samples were centrifuged for 15 min at 1500 rpm and 4°C. Dead cells were removed using a Dead Cell Removal Kit (Miltenyi Biotec, Bergisch Gladbach, Germany). Cell viability was 76% (2 week) and 73% (6 week), respectively. The experiment involving mice was approved by the Animal Experimental Committee of the Institute of Science Tokyo.

Cells isolated from Achilles tendons of both 2-week-old and 6-week-old mice were processed separately. Cells were resuspended in PBS with 1% fetal bovine serum (FBS) at a concentration of 10,000 cells per μL. Then, 10,000 cells per sample were loaded on a Chromium Controller (10x Genomics, Pleasanton, CA, USA) for single-cell capture. Libraries were prepared using Single Cell 3′ Library & Gel Bead Kit v3 (10x Genomics) following the manufacturer’s instructions. A single-cell emulsion (Gel Bead-In-EMulsions [GEMs]) was created by making barcoded cDNA unique to each individual emulsion. A recovery agent was added to break GEM, and cDNA was then amplified. A library was produced via end repair, dA-tailing, adapter ligation, post-ligation cleanup with SPRIselect, and sample index PCR. The quality and concentration of the amplified cDNA were evaluated using the Bioanalyzer (Agilent 2100) on a High Sensitivity DNA chip (Agilent, #5065-4401; Santa Clara, CA, USA). Sequencing was performed using the HiSeq X system (Illumina, San Diego, CA, USA) to generate 28/90 bp paired-end reads.

After obtaining cells from the mouse Achilles tendons of 2-week-old and 6-week-old mice (processed separately for each age group), nuclei were isolated following the Chromium Next GEM Single Cell Multiome ATAC+Gene Expression Reagent Bundle (10x Genomics), and all of the buffers were made according to the manufacturer’s instructions. Briefly, after centrifugation, the supernatant was discarded, and the cells were resuspended in 1 mL of 1% FBS in PBS. Approximately 500,000 cells were transferred to a new tube for further lysis. To remove the supernatant, cells were centrifuged again for 5 min at 300 rcf and 4°C and resuspended in 100 mL of chilled lysis buffer. Cells were lysed for 9 min on ice, and 1 mL of wash buffer was added to stop the reaction. Then, the suspension was filtered through a 40 μm cell strainer, cells were centrifuged and resuspended in 66.2 mL of chilled Diluted Nuclei Buffer aiming to target 7000 nuclei. The final concentration of nuclei was determined, followed by transposition, GEM generation, barcoding, and library construction according to the Chromium Next GEM Single Cell Multiome ATAC+Gene Expression (10x Genomics). Libraries were sequenced with the parameters recommended by the manufacturer, using the NovaSeq 6000 (Illumina) to generate 28/90 bp paired-end reads for gene expression and 50/49 bp paired-end reads for ATAC sequencing.

Sequencing reads were processed with the Cellranger_arc (10x Genomics, v2.0.0) using the mouse reference mm10. We performed separate analyses for scRNA-seq data (collected from both 2-week and 6-week samples) and snRNA-seq data (collected from both 2-week and 6-week samples) to enable comprehensive characterization of transcriptional landscapes. From the gene expression matrix, downstream analyses were carried out using R. Quality control, filtering, data clustering and visualization, and a differential expression analysis were carried out using Seurat (Butler et al., 2018). For each dataset, cells with unique feature counts over 2500 or less than 200 and >5% mitochondrial counts were filtered. Then, heterotypic doublets (assuming 5% of barcodes represent doublets) were removed using DoubletFinder (McGinnis et al., 2019).

Unsupervised shared nearest neighbor clustering was performed with varying resolution, and the results were visualized using UMAP (Becht et al., 2019). DEGs among each cell cluster were determined using the FindAllMarkers function in Seurat. The criteria of DEGs as a marker gene for the cluster was logFC >0.25, adjusted p<0.05, expression in >25% of cells. We then annotate each cluster based on DEGs (Supplementary files 1–15). To ensure consistency between our different sequencing approaches, we compared cell clusters identified in the scRNA-seq analysis with those from the snRNA-seq analysis (Figure 2—figure supplement 2).

Monocle3 (Trapnell et al., 2014) was used to convert the snRNA-seq dataset into a cell dataset object, preprocess data, correct for batch effects, embed with dimensional reduction, and perform pseudo-temporal ordering. Cicero (Pliner et al., 2018) was used to generate pseudo-temporal trajectories for the snATAC-seq dataset.

Intercellular communication networks were quantitatively inferred and analyzed using scRNA-seq data. The R package CellChat (Jin et al., 2025) was used to visualize the interactions among different cell groups. Two hundred twenty-nine signaling pathway families were grouped as a library to analyze cell-cell communication.

The R package ClusterProfiler (Yu et al., 2012) was used to perform a gene enrichment analysis. The p-values were corrected by the Benjamini-Hochberg method.

Genomic regions containing snATAC-seq peaks were annotated using ChIPSeeker (Yu et al., 2015) and the UCSC database on mouse (mm10).

The Cell Ranger ATAC pipeline (1.2.0) (Satpathy et al., 2019) was used to preprocess the data resulting from sequencing. First, Tn5 sites were mapped to the mouse reference transcriptome mm10, and duplicate reads and background cells were removed. This returned barcoded fragment files, which were loaded into Signac (Stuart et al., 2021) for downstream analyses using the standard Signac/Seurat pipeline. Macs2 (Wang et al., 2008) was run on the fragment files to call peaks using the Signac ‘CallPeaks’ function. Fragments were mapped to the Macs2 called peaks and assigned to cells using the Signac ‘FeatureMatrix’ function. Nucleosome signal strength and TSS enrichment for each cell were calculated using the Signac ‘NucleosomeSignal’ and ‘TSSEnrichment’ functions, respectively. Outliers in the QC metric categories were removed as per Signac’s standard processing guidelines. Latent semantic indexing (LSI), a form of dimensional reduction, was performed using the Signac ‘RunTFIDF’ and ‘RunSVD’ functions. The UMAP hyperparameters were varied to produce consistent object shapes (using R). Once hyperparameters were chosen, the Signac/Seurat’s ‘RunUMAP’ function was run on the LSI dimensions chosen earlier for UMAP embedding. The Signac/Seurat ‘FindNeighbors’ function was run using the same LSI dimensions used for UMAP to compute the nearest neighbor graph. Signac/Seurat ‘FindClusters’ was then run at varying resolutions. A gene activity matrix was constructed by counting ATAC peaks within the gene body and 2 kb upstream of the transcriptional start site for protein-coding genes annotated in the Ensembl database. The gene activity matrix was log-normalized prior to label transfer with the aggregated snRNA-seq Seurat object using a canonical correlation analysis (CCA). The aggregated snATAC-seq object was filtered using a 97% confidence threshold for cell-type assignment following label transfer to remove heterotypic doublets. The filtered snATAC-seq object was reprocessed with TFIDF, SVD, and batch effect correction followed by clustering and annotation based on lineage-specific gene activity. Differential chromatin accessibility between cell types was assessed with the Signac FindMarkers function for peaks detected in at least 20% of cells using a likelihood ratio test and a log-fold-change threshold of 0.25. Bonferroni-adjusted p-values were used to determine significance at an FDR<0.05.

snRNA-seq and snATAC-seq data were integrated using the cluster-label transfer procedure as implemented in Signac and Seurat. Each snRNA-seq sample was clustered individually, and its cluster labels were projected onto the matching, individually clustered snATAC-seq sample or vice versa. Anchors were identified for condition-matched snRNA- and snATAC-seq samples using the FindTransferAnchors function, and a CCA was performed using the snRNA expression values and the snATAC-imputed gene expression values. The anchors were used to transfer cluster-label identifiers between the two data types using the TransferData function. Each cell in the query was assigned the cluster label with the highest prediction score, and label transfer was considered successful for query cells with prediction scores above 0.3.

To identify TFs and characterize cell states, a cis-regulatory analysis was performed using the R package SCENIC (Aibar et al., 2017), which infers gene regulatory networks based on co-expression and DNA motif analyses. The network activity was then analyzed for each cell to identify recurrent cellular states. In short, TFs were identified using GENIE3 and compiled into modules (regulons), which were subsequently subjected to cis-regulatory motif analysis using RcisTarget with two gene-motif rankings: 10 kb around the TSS and 500 bp upstream. Regulon activity in every cell was then scored using AUCell.

Immunohistochemistry was performed as previously described (Tsutsumi et al., 2022). Briefly, tissue samples were fixed in 4% paraformaldehyde (PFA) overnight at 4°C and embedded in paraffin. Sections of 10 μm in thickness were deparaffinized and activated with citric acid buffer (10 mM sodium citrate, 1  mM EDTA; pH 6.0) at 80°C for 60 min in a decloaking chamber NxGen (BIOCARE Medical, CA, USA) or with 0.1% trypsin/PBS at 37°C for 30 min. After they were blocked with Blocking One solution (Nacalai Tesque, Kyoto, Japan) for 60 min, they were incubated with rabbit anti-CD55 antibody (1:100; A13918, RRID:AB_2760771, ABclonal, Woburn, MA, USA) or rat anti-CD248 (1:100; MAB7535, RRID:AB_3658829, R&D Systems, Minneapolis, MN, USA) overnight at 4°C. Following this step, they were incubated with Alexa Fluor plus 488 goat anti-rabbit antibody (1:1000; A32731, RRID:AB_2633280, Thermo Fisher Scientific, Waltham, MA, USA) or Alexa Fluor plus 488 donkey anti-rat antibody (1:1000; A21208, RRID:AB_2535794, Thermo Fisher Scientific) for 60 min. Hoechst 33342 (100 ng/mL, Thermo Fisher Scientific) was added during this incubation. The sections were then mounted with ProLong Glass Antifade Mountant (P36980, Thermo Fisher Scientific).

Cells harvested from the Achilles tendon were incubated for 60 min on ice with APC-CD55 (#131812, RRID:AB_2800632, BioLegend, San Diego, CA, USA) and a primary antibody against CD248 (#LS-B2712, RRID:AB_1664704, LSBio) using a FACS buffer (PBS containing 1% [vol/vol] FBS). After washing, cells were reacted with fluorescent-conjugated rabbit secondary antibody. Cell sorting was performed using MoFlo XDP FACS (Beckman Coulter, Brea, CA, USA). DAPI was used as the live/dead discriminator to the gate. CD55/CD248 dual positive and negative cell clusters were sorted.

Primary TSPCs were isolated from the Achilles tendons of 2-week-old mice with collagenase digestion described above. Single-cell suspensions were cultured in the culture medium (MEMα+20% FBS+1% penicillin-streptomycin+1% [vol/vol] 100× non-essential amino acid solution [Gibco], 1% [vol/vol] 100× GlutaMAX [Gibco]). At 80–90% confluence, cells were trypsinized, centrifuged, resuspended in culture medium as passage 1 cells, and incubated in 5% CO₂ at 37°C, with fresh medium every 2–3 days.

For the colony formation assay, single-cell suspensions of TSPCs (1000 cells/well) were seeded and incubated in six-well plates for 12–14 days in the growth medium and fixed with 4% PFA (Sigma-Aldrich, St. Louis, MO, USA). Then, 0.1% crystal violet solution (Wako) was used to stain the cells. Colonies of >30–50 cells were defined as a single colony unit (Franken et al., 2006), and the number of clusters was counted using the ImageJ package ColonyArea (Guzmán et al., 2014).

Total RNA was purified using TRIzol reagent (Invitrogen, Grand Island, NY, USA). Reverse transcription of mRNA was carried out using the PrimeScript RT Reagent Kit (Takara, Shiga, Japan). Quantitative PCR (qRT-PCR) was performed on cDNA with the Thunderbird SYBR mix (Toyobo, Osaka, Japan). B2m was used as a reference gene, and relative gene expression levels were calculated through the ΔCT method.

For the differentiation experiment, TSPCs were cultured in six-well plates (50,000 cells/well). Osteogenic, chondrogenic, and tenogenic differentiation were induced using a corresponding differentiation medium. The osteogenic differentiation medium contained a culture medium supplemented with 10 nM dexamethasone (Sigma-Aldrich), 5 mM β-glycerophosphate (APEXBIO), and 0.05 mM L-ascorbic acid 2-phosphate (Sigma-Aldrich). The chondrogenic differentiation medium contained a culture medium supplemented with 100 nM dexamethasone and 10 ng/mL BMP2 (Sigma-Aldrich). The tenogenic differentiation medium contained a culture medium supplemented with 10 ng/mL TGF-β1 (Peprotech, Rocky Hill, NJ, USA), 10 ng/mL GDF-5 (R&D Systems, Minneapolis, MN, USA), and 0.05 mM L-ascorbic acid 2-phosphate. After a 2-week induction period, cells were harvested for RT-PCR.

Generation of bio-tendon has been reported previously (Kataoka et al., 2020; Tsutsumi et al., 2022) Briefly, sorted cells were embedded in a 3D-culture cocktail, which is consistent with 2 mg/mL Cellmatrix (Type I-A, Nitta Gelatin Inc, Osaka, Japan), pro-survival cocktail (final concentrations: 100 nM Bcl-Xl BH4 4-23 [Merck Millipore, Burlington, MA, USA], 100 μM carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone [Z-VAD-FMK] [Promega, Madison, WI, USA]) in culture medium. The 3D chamber was coated with 1% gelatin and incubated at 37°C and 5% CO₂ for 30 min. After washing with 1× PBS three times, 1.0×10⁶ cells were transferred to a 3D-culture cocktail mixture and incubated at 37°C and 5% CO₂ for 60 min for gelation. Following gelation, tenogenic differentiation medium was added to the chamber. Following 24 hr of incubation, the chamber was placed within a cell stretching system (Shellpa Pro, Menicon Life Science, Aichi, Japan). Cyclic mechanical stretch was applied for 1 week, with gradual increase in stretch load: 1% on day 1, 2% on day 2, 3% on day 3, 4% on day 4, and 5% from days 5 to 7. The stretching cycle was programmed at 0.25 Hz for 18 hr/day, followed by resting for 6 hr/day at 37°C and 5% CO₂. Daily medium changes were also required.

The cultured bio-tendon was placed into a plastic pack filled with saline and sealed to prevent implosion and leakage during the procedure. The pack was then pressurized at 1000 MPa at 30°C for 10 min using an HHP machine (Dr. CHEF; Kobe Steel, Hyogo, Japan). After pressurization, the bio-tendon was washed thrice with 30% ethanol (EtOH) by continuous shaking for 5 min at each step. Finally, 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC)/N-hydroxysuccinimide (NHS)-based cross-linking was performed by adding 70 mM EDC (Wako) and 70 mM NHS (Wako) with 30% EtOH for 24 hr at 4°C. The processed bio-tendon was incubated in 1× PBS at 4°C until the experiment.

Bio-tendon tissues were dissected and fixed overnight in 2.5% glutaraldehyde in 0.1 M phosphate buffer (PB). For TEM, the specimens (n=3) were rinsed with 0.1 M PB, post-fixed in 1% osmium buffered with 0.1 M PB for 2 hr, and dehydrated in a graded ethanol series. The specimens were then embedded in Epon 812, sectioned into ultrathin sections (70 nm), mounted on copper grids, and double-stained with uranyl acetate and lead citrate. TEM observation was conducted on a JEM-1400Flash (JEOL, Tokyo, Japan). For SEM, the specimens (n=2) were dried in a critical-point dryer (HCP-2; Hitachi, Tokyo, Japan) with liquid CO₂ and coated with platinum. The specimens were subsequently observed under SEM (JSM-7900F/JED-2300; JEOL). Fiber orientation within the microscopic sections was analyzed using the OrientationJ image processing tool, a plugin for ImageJ.

The mechanical properties were measured using a creep meter (RE-3305S; Yamaden, Tokyo, Japan). After measuring the initial length (mm), diameter (mm), and thickness (mm) using a micrometer, the samples were fixed with two grips, which were pulled at a constant speed of 0.05 mm/s until failure, and the tensile strength (N) and failure strain (mm) were measured. The cross-sectional area (mm²) was calculated using the initial diameter and thickness. The stiffness was manually determined from the slope in the linear region of the failure-stress curve.

The tensile strength (MPa), failure strain (%), stiffness, and Young’s modulus were calculated using the following formulas:

All values are presented as means ± SEM. Statistically significant differences were assessed using unpaired two-tailed Student’s t-tests and one-way analysis of variance (ANOVA) with Tukey’s post hoc tests. Statistical significance was set at p<0.05.