Single-cell transcriptome atlas of human tendon adhesion tissue

To determine the cellular composition of tendon adhesion tissue, tissue samples were obtained surgically from 12 patients with flexor tendon tenolysis including three samples of normal peritendinous tissue and three pathologic samples from the lesion at 3, 10 days post injury (dpi) and 12 weeks post injury (wpi), respectively. The samples were immediately processed for 3′-end scRNA-seq using the 10× Genomics platform (Fig. 1a and Table 1). The data were clustered and integrated by time point, and clusters that were of low quality or were doublets were removed. A total of 74 350 cell transcriptomes from the 12 patients were retained for subsequent analysis, of which 10 081 cells originated from normal peritendinous tissue, 18 866 from 3 dpi tissue, 14 953 from 10 dpi tissue, and 30 450 from 12 wpi tissue (Fig. 1b and Table 2).

Fig. 1

Single-cell atlas of human normal peritendinous tissue and tendon adhesion tissue. a Overview of this study design. Normal peritendinous tissue and tendon adhesion tissue were collected from tendon injury patients and processed for 3′-end scRNA-seq using the 10× Genomics platform. b UMAP plots of 74 350 cells from total 12 patients with tendon injury. 10 081 cells from normal peritendinous tissue of three patients. 18 866 cells from tendon adhesion tissue of three 3 dpi (day post injury) patients. 14 953 cells from tendon adhesion tissue of three 10 dpi patients. 30 450 cells from tendon adhesion tissue of three 12 wpi (week post injury) patients. c Dot plot: showing cell clusters of human tendon adhesion tissue by known markers. The dot size indicates the gene expression percent in each cluster. The color indicates mean gene expression (Red, high). d Bar plots of the proportion of eight major cell types in human tendon adhesion tissue of each time point. e Bar plots of the proportion of eight major cell types in human tendon adhesion tissue of each patient

Table 1 Clinical characteristics of 12 patients in the studyTable 2 Statistics of scRNA-seq

An atlas of cell types in normal peritendinous and injured tissue was established by first clustering nuclei at a coarse level and then annotating each cluster with cell type-specific markers (Fig. 1c). We identified eight major cell clusters including mononuclear phagocytes (MPs, n = 16 299) identified by CD14 molecule (CD14) and CD68 molecule (CD68)11,12 (Fig. S1B), endothelial cells (ECs, n = 10 175) marked by platelet and endothelial cell adhesion molecule 1 (PECAM1) and von Willebrand factor (VWF)13,14 (Fig. S1B), mesenchymal cells (MCs, n = 33 284) identified by paired related homeobox 1 (PRRX1)15 and Thy-1 cell surface antigen (THY1)16 respectively (Fig. S1C), neutrophils (n = 3 578) marked by colony stimulating factor 3 receptor (CSF3R) and Fc gamma receptor IIIb (FCGR3B)17 (Fig. S1D), B cells (n = 1 425) expressed membrane spanning 4-domains A1 (MS4A1) and CD79a molecule (CD79A)18 (Fig. S1E), T cells (n = 8 914) expressed CD3 delta subunit of T-cell receptor complex (CD3D) and CD3 epsilon subunit of T-cell receptor complex (CD3E)19 (Fig. S1F), mast cells (n = 497) marked membrane spanning 4-domains A2 (MS4A2) and KIT proto-oncogene, receptor tyrosine kinase (KIT)20 and schwann cells (n = 178) identified by cadherin 19 (CDH19) and S100 calcium binding protein B (S100B)21 (Fig. S1G, H; and Table 3).

Table 3 Cell composition of 12 samplesCharacteristics of cell populations in tendon adhesion tissue of different clinical stage

The proportion of inflammatory cells, ECs, and MCs varied according to the stage of tendon adhesion (Fig. 1d, e). In the normal peritendinous tissue, MCs were the most abundant cell type, accounting for 52.9% of the total (Fig. 1d and Table 4). At 3 dpi (inflammatory phase)3, the proportion of MCs and ECs was lower (0.3% and 0.5% of total cells, respectively), whereas the proportion of inflammatory cells including mononuclear phagocytes, T cells, B cells and neutrophils increased; the most abundant cell type at this stage was MPs (55.3%), followed by T cells (25.5%) and neutrophils (14.7%) (Fig. 1d and Table 4). At 10 dpi (proliferative phase)3, the number of inflammatory cells had declined, with macrophages accounting for just 18.8% of all cells. Meanwhile, ECs and MCs had undergone extensive proliferation and accounted for 22.4% and 41.8% of cells, respectively (Fig. 1d and Table 4). At 12 wpi (remodeling phase)3, fibroblasts had migrated to the lesion and were present in the granulation tissue, which induces tendon adhesion. MCs (71%) and ECs (19.6%) were the major cell types at this stage, with few inflammatory cells (Fig. 1d and Table 4). The profile of the major cell types at 12 wpi was similar to that of normal peritendinous tissue, indicating that at an advanced stage of tendon adhesion, the tissue state was stabilizing. Based on these dynamics of cell populations, we speculated that MPs, ECs, and MCs play an important role in the development of tendon adhesion.

Table 4 Cell composition of total samples at different stages ADAM12 expression defines multipotent profibrogenic mesenchymal stromal cells

In human tendon adhesion tissue, mesenchymal cells (MCs) were categorized into four subsets based on distinct markers: ADAM metallopeptidase domain 12 (ADAM12), C-X-C motif chemokine ligand 14 (CXCL14), regulator of G protein signaling 5 (RGS5) and myosin heavy chain 11 (MYH11) (Fig. S2A-D). Two samples at 3 dpi had no MCs, as they were scarce at this time point (Fig. S2F and Table 3). ADAM12 and periostin (POSTN) was the marker of MC0 and identified as fibroblasts (Fig. S2B, C). MC1 was distinguished by CXCL14 expression, was identified as fibroblasts (Fig. S2B, C). MC2 expressed high levels of RGS5, STEAP4 metalloreductase (STEAP4) and collagen type IV alpha 1 chain (COL4A1) was identified as pericytes22 (Fig. S2B). MC3 was marked by MYH11 and actin alpha 2, smooth muscle (ACTA2) and identified as myofibroblasts23 (Fig. S2B, C). The proportions of the four subsets revealed that MC1 primarily constituted MCs in normal peritendinous tissue (Fig. S2E). MC0 and MC2 proliferated in 10 dpi tissue, while MC1 and MC3 proliferated in 12 wpi tissue (Fig. S2E). After tendon injury, a total of three stages compared with normal peritendinous tissue, the upregulated signaling pathways of MCs in tendon adhesion tissue including ECM-receptor interaction and focal adhesion (Fig. S2G). Visualization of the pseudotemporal trajectory and RNA velocity suggested that MC0 differentiated into MC1, followed by MC1 differentiating into MC2, and ultimately MC2 differentiating into MC3 after tendon injury, 3 dpi to 10 dpi then to 12 wpi (Fig. S3A and Fig. S4A–C), and the origin of MC0 was bone-marrow. This observation implies that tendon injury may induce pericyte-myofibroblast transition (PMT) during human tendon adhesion. MCs in human normal peritendinous tissue were primarily CXCL14+ fibroblasts (Fig. S2E).

ADAM12+ cells are known progenitors of a large proportion of collagen-overproducing cells generated during scarring that are progressively eliminated during healing.24 MC0 expressed high levels of profibrotic genes including COL1A1, collagen type III alpha 1 chain (COL3A1) and POSTN25 (Fig. S2C). The Gene Ontology enrichment analysis suggested MC0 participated in collagen fibril organization and ECM organization (Fig. S3B), and the QuSAGE analysis of enriched pathways showed that genes related to ECM–receptor interaction and platelet activation that are associated with fibrosis were upregulated in MC026,27,28 (Fig. S3C). The gene enrichment of MC0 ECM-receptor interaction hallmark gene set suggested the profibrotic genes including collagen type I alpha 2 chain (COL1A2) and FN1 upregulated (Fig. S4D), which was supported by KEGG pathway analysis (Fig. S3D). The heatmap of TFs revealed that MC0 were enriched in TFs that promote fibrosis including hypoxia-inducible factor 1 subunit alpha (HIF1A)29 (Fig. S4E). To further investigate MC0, we isolated ADAM12+ MCs from 10 dpi human adhesion tissue and induced them to undergo adipogenic, osteogenic, and chondrogenic differentiation (Fig. S4F). The results indicated that ADAM12+ fibroblasts were multipotent stromal cells. Immunofluorescence demonstrated colocalization of ADAM12 and CXCL14 around collagen, as well as colocalization of COL4A1+ pericytes with MYH11+ myofibroblasts in 10 dpi tissue (Figs. S3E and S4G). These findings indicate that in the progression of human tendon adhesion, ADAM12+ multipotent stromal cells might transform into CXCL14+ fibroblasts, while COL4A1+ pericytes could convert into MYH11+ myofibroblasts. PMT could play a role in human tendon adhesion.

In summary, ADAM12+ cells, as profibrotic multipotent stromal cells, proliferate during the intermediate stages of tendon adhesion after injury and differentiate into CXCL14+ fibroblasts, possibly playing a vital part in this process. Moreover, PMT might be crucial for human tendon adhesion.

Unique subgroups of endothelial cells reside within the fibrotic microenvironment

Endothelial cells (ECs) in tendon adhesion tissue were identified by the high expression levels of PECAM1 and VWF13,14 (Fig. 1c). Then the ECs were further subclustered into four subpopulations based on different markers including gap junction protein alpha 4 (GJA4), regulator of cell cycle (RGCC), atypical chemokine receptor 1 (ACKR1) and C-C motif chemokine ligand 21 (CCL21) (Fig. 2a–d). ECs were scarcely present in human tendon adhesion tissue at 3 dpi, with two samples lacking ECs (Fig. 2e, f and Table 3). GJA4 and semaphorin 3G (SEMA3G) were specially expressed in human arteries, suggesting ENDO0 as arterial ECs14,30 (Fig. 2b, c). RGCC was reported as the marker of human capillary which suggested ENDO1 as capillary ECs31 (Fig. 2b, c). ENDO2 expressed a high level of ACKR1 which was specially expressed in the human vein,32 suggesting ENDO2 as venous ECs (Fig. 2b, c). ENDO3, marked by CCL21 and podoplanin (PDPN), was identified as lymphatic ECs14 (Fig. 2b, c). Recent research has reported that RGCC and ACKR1 play roles in ECM organization.8,33 ECs primarily proliferated at 10 dpi and 12 wpi (Fig. 2f, Tables 3 and 4), with ENDO1 expanding at 10 dpi (Fig. 2e). After tendon injury, total three stages comparing with normal peritendinous tissue, the upregulated signaling pathways of ECs in tendon adhesion tissue including ECM–receptor interaction and focal adhesion (Fig. 2g).

Fig. 2

Distinct EC clusters present in human tendon adhesion tissue. a Clustering 10 175 endothelial cells from total 10 patients. ECs, endothelial cells. b Dot plot: showing cell clusters of endothelial cells by known markers. The dot size indicates the gene expression percent in each cluster. The color indicates mean gene expression (Red, high). c The violin plot of selected gene expression of each cluster in ECs. d Heatmap of marker genes in each EC cluster. Top, clusters. Left, marker genes. e Bar plots of proportion of each cluster in ECs from 10 patients. f UMAP plots of ECs of each time point: normal, 3 dpi, 10 dpi and 12 wpi. g Pathway analysis of upregulated signaling pathway of all three stages after tendon injury

Further analysis of pseudotemporal trajectory suggested that following injury, a differentiation trajectory from ENDO1 into ENDO0 into ENDO2 and the origin of ENDO1 were endothelial progenitor cells, we constituted 3 distinct gene expression modules (Fig. 3a and Fig. S5A, B). The heatmap of differential genes module across ENDO1 to ENDO2 pseudotemporal trajectory suggested fibrotic genes including SPP1, FN1, COL1A1, and MMP9 upregulated in the process8,34 (Fig. 3b). GO enrichment analysis of Module 1 revealed that genes involved in ECM organization were upregulated during this process (Fig. 3c). To better understand ENDO1 and ENDO2 phenotypes, Gene Ontology enrichment analysis revealed angiogenesis and ECM organization in ENDO1 and ECM organization and collagen fibril organization in ENDO2 (Fig. 3d). The marker pathway analysis suggested the ECM-receptor interaction in both ENDO1 and ENDO2 (Fig. S5C). The Qusage analysis of enriched pathways of ECs indicated the fibrosis-associated pathway including NOTCH signaling pathway was upregulating in ENDO18,35 (Fig. 3e). The KEEG analysis revealed the enriched pathway of ECM-receptor interaction in ENDO1 and ENDO2 (Fig. 3f). A heatmap of TFs shows ENDO1 enriched profibrotic TFs including Jun proto-oncogene (JUN)36 and ENDO2 enriched profibrotic TFs including Fos proto-oncogene (FOS)37 (Fig. S5D).

Fig. 3

Identifying the profibrotic ECs. a The pseudotemporal trajectory analysis of ENDO0, ENDO1 and ENDO2. Arrows indicated the direction of pseudotemporal trajectory. b Heatmap of differential gene modules across ENDO1 to ENDO2 (right arrow) and ENDO1 to ENDO0 (left arrow) pseudotemporal trajectories. Grouped by hierarchical clustering (n = 3). The genes of module 1 were labeled at left. c The TOP 15 Gene Ontology enrichment of all genes in module 1, along ENDO0 to ENDO2 pseudotemporal trajectory. d The Gene Ontology enrichment analysis of ENDO1 (left) and ENDO2 (right). e The Qusage analysis of enriched pathways of each cluster of ECs. The color indicates mean pathway intensity (Red, high. Blue, low). Right, pathways. Bottom, clusters. f The KEGG analysis of enriched pathways of each cluster of ECs. The color indicates mean pathway intensity (Red, high. Blue, low). Bottom, clusters. Right, pathways. g Primary human fibroblasts treated with conditioned media from ENDO2 (ACKR1+ ECs) (n = 3) or ACKR1- ECs (n = 3), qPCR of stated genes, expression relative to COL1A1 mean expression of control primary human fibroblasts (n = 3), Mean ± SEM. h The polychromatic immunofluorescence for CD31 (marker of ECs), ACKR1, RGCC, and COL1A1 showed RGCC+ACKR1+ ECs (white arrows) existed around collagen at 10 dpi. RGCC and ACKR1 had part colocalization. White arrows CD31+RGCC+ACKR1+ cells. CD31(green), RGCC (orange), ACKR1 (yellow), COL1A1 (red), DAPI (blue), scale bars 10 μm

Our cellular experiments demonstrated that ACKR1+ endothelial cells promote fibrosis. We isolated and cultured ACKR1+ ECs from human 10 dpi tissue, and in vitro experiments showed that these cells increased the mRNA level of COL1A1 in human fibroblasts (Fig. 3g and Fig. S9A). The presence of ACKR1+ ECs was adjacent to COL1A1, involving in fibrotic niche of human tendon adhesion tissue at 10 dpi (Fig. 3h). Additionally, the colocalization of CD31, RGCC, ACKR1, and GJA4 at 10 dpi suggested that GJA4+ ECs and RGCC+ ECs might transform into ACKR1+ ECs during tendon adhesion progression (Fig. 3h and Fig. S5E). Collectively, we identified ECs subsets in human tendon adhesion tissue, and our experiments indicated that ACKR1+ ECs contribute to fibrosis in human tendon adhesion.

Distinct MP clusters are presented in tendon adhesion tissue

In our study, we characterized five MPs subsets that included dendritic cells (DCs), monocytes and macrophages (Fig. 4a). MP4 was defined as DCs marked CD1c molecule (CD1C) and Fc epsilon receptor Ia (FCER1A)38 (Fig. 4b, e). MP0 was defined as monocytes enriched in S100 calcium binding protein A8 (S100A8) and S100 calcium binding protein A12 (S100A12)38,39,40 (Fig. 4b, e). MP1, MP2 and MP3 were defined as macrophages marked CD68, apolipoprotein E (APOE) and complement C1q A chain (C1QA)41 (Fig. 4b). MP1 enriched by interleukin 1 beta (IL1B) and epiregulin (EREG) and the GO Ontology (GO) analysis revealed the MP1 participated in inflammatory response (Fig. 5b). Therefore, MP1 were defined as proinflammatory macrophages. MP2, characterized as profibrotic macrophages, was marked by secreted phosphoprotein 1 (SPP1) and matrix metallopeptidase 9 (MMP9)8 (Fig. 4b, c). MP3 was defined as antifibrotic macrophages expressing folate receptor beta (FLOR2) and lymphatic vessel endothelial hyaluronan receptor 1 (LYVE1) by our in vitro experiment42,43 (Fig. 4b, c). Normal peritendinous tissue primarily consisted of MP3, making up 70.9% of total MPs (Fig. 4d and Table 5). After the tendon injury, MP1 and MP2 increased significantly at 3 dpi, accounting for 48.9% and 15.9%, respectively, while FOLR2 proportion decreased to 30.2% (Fig. 4d and Table 5). Tendon injury led to a reduction in MP3 abundance within adhesion tissue, and this stage exhibited the highest macrophage count during adhesion progression (Fig. 4f). The MPs’ expression levels of pro- and anti-inflammatory cytokines at four distinct stages showed the cytokines had the highest levels at 3 dpi (Fig. 4g), indicating the inflammatory follow tendon injury was the most active at 3 dpi. As adhesion progressed, MP1 and MP2 proportions decreased, while MP3 increased at 10 dpi (Fig. 4d and Table 5). Macrophage numbers gradually declined, and by 12 weeks, granulation tissue formed with fewer macrophages. MP3 proportion among macrophages neared that of 10 dpi but remained lower than in normal peritendinous tissue (Fig. 4d and Table 5). After tendon injury, total three stages comparing with normal peritendinous tissue, the GO analysis of MPs revealed the inflammation relative response had upregulated (Fig. 4h), the upregulated signaling pathways of MPs in tendon adhesion tissue including HIF-1 signaling pathway, TNF signaling pathway and IL-17 signaling pathway (Fig. 4i), which were reported in tendon injury.44 By comparing with normal peritendinous tissue, the downregulated signaling pathways of MPs including MAPK signaling pathway after the tendon injury (Fig. 4j).

Fig. 4

Distinct MP clusters present in human tendon adhesion tissue. a Clustering 16 299 mononuclear phagocytes (MP) from total 12 patients. DCs, dendritic cells. b Dot plot: showing cell clusters of MPs by known markers. The dot size indicates the gene expression percent in each cluster. The color indicates mean gene expression (Red, high). c The violin plot of selected gene expression of each cluster in MPs. d Bar plots of proportion of each cluster in MPs from 12 patients. e Heatmap of marker genes in each MP cluster. Top, clusters. Left, marker genes. f UMAP plots of MPs of each time point: normal, 3 dpi,10 dpi, and 12 wpi. g The violin plot of expression levels of pro- and anti-inflammatory cytokines in MPs, including pro-inflammatory cytokines such as IL1, IL8, MIF, TNF, and anti-inflammatory cytokines such as IL1RN, IL4, IL10. h The GO analysis of up-regulated genes of all three stages after tendon injury. i Pathway analysis of up-regulated signaling pathway of all three stages after tendon injury. j Pathway analysis of down-regulated signaling pathway of all three stages after tendon injury

Fig. 5

Identifying the profibrotic macrophages and antifibrotic macrophages. a The pseudotemporal trajectory analysis of MP0, MP1, MP2, and MP3. Arrows indicated the direction of pseudotemporal trajectory. b The Gene Ontology enrichment analysis of MP1, MP2 and MP3. c The Qusage analysis of enriched pathways of each cluster of MPs. The color indicates the mean pathway intensity (Red, high. Blue, low). Right, pathways. Bottom, clusters. d The KEGG analysis of enriched pathways of each cluster of MPs. The color indicates the mean pathway intensity (Red, high. Blue, low). Bottom, clusters. Right, pathways. e Primary human fibroblasts treated with conditioned media from MP2 (SPP1+macrphages) (n = 3) or SPP1- macrophages (n = 3), qPCR of stated genes, expression relative to COL1A1 mean expression of control primary human fibroblasts (n = 3), Mean ± SEM. f Primary human fibroblasts treated with TGFβ1 and conditioned media from MP3 (FOLR2+macrophages) (n = 3) or FOLR2- macrophages (n = 3), qPCR of stated genes, expression relative to COL1A1 mean expression of control primary human fibroblasts (n = 3), Mean ± SEM. g The pathway analysis of MP3

Table 5 MP clusters composition of total samples at different stagesProfibrotic phenotype of SPP1+ macrophages

In a recent investigation, TREM2+SPP1+ macrophages in human fibrotic liver tissue were identified as scar-associated macrophages that could promote hepatic fibrosis.8 In the present study, we had identified MP2 marked SPP1, triggering receptor expressed on myeloid cells 2 (TREM2) and MM9 in human tendon adhesion tissue (Fig. 4b, c). However, the ontogeny of human peritendinous macrophage subpopulations remains unclear. To further investigate MP2 origin, we visualized the pseudotemporal trajectory. These analyses indicated that upon injury, a differentiation trajectory progressed from MP0 to MP1, then MP1 to MP2 (Fig. 5a and Fig. S6A, B), the origin of MP0 was circulating monocytes in blood, suggesting that MP2 was monocyte-derived, consistent with previous research.

To further characterize the MP2 phenotype, we analyzed the heatmap of genes expressed by MP2 and observed upregulation of the profibrotic genes SPP1, TREM2, and CD9 molecule (CD9)8,45 (Fig. 4c). The results of GO analysis indicated the MP2 participated in extracellular matrix (ECM) organization (Fig. 5b), and the KEEG analysis suggested that the cells were involved in ECM–receptor interaction (Fig. 5d). The marker pathway analysis suggested the ECM-receptor interaction in MP2 (Fig. S6C). The Qusage analysis of pathways enriched in each MP subset showed that fibrosis-associated pathways including ECM–receptor interaction and glycol