Spatiotemporal dynamics and regulation of suture morphology and collagen remodeling during trans-sutural distraction osteogenesis

Abstract

Trans-sutural distraction osteogenesis (TSDO) is an effective treatment of midfacial hypoplasia, a craniofacial deformity frequently associated with cleft lip and palate. Though extracellular matrix (ECM) remodeling plays a pivotal role in craniofacial correction, the characteristics and mechanisms underlying collagen reorganization and cellular morphological adaptations during TSDO remain poorly understood. This study quantitatively delineates the spatiotemporal changes of sutural cells and ECM morphology, revealing a polarized alignment parallel to the direction of mechanical force. Multi-omics analysis demonstrates that macrophages regulate collagen remodeling in suture mesenchymal stem cells (SuSCs) via the PDGF signaling pathway. Subsequent in vitro stretch loading models confirmed PDGF pathway activation enhances SuSCs migration, collagen synthesis, and cellular morphological reorganization. Validation in macrophage-elimination murine models further corroborated this regulatory axis. Collectively, our work maps the dynamic microenvironmental changes during TSDO and elucidates cell-cell interaction-driven ECM collagen remodeling. These insights advance the understanding of TSDO-mediated osteogenesis and provide a foundation for developing optimized therapeutic strategies.

Introduction

Midfacial hypoplasia (MH) constitutes a multifactorial craniofacial disorder attributable to both congenital anomalies and acquired influences¹. Its hallmark presentation contains skeletal underdevelopment, maxillary arch constriction, and Angle Class III malocclusion, resulting in compromised facial esthetics, impaired masticatory function, and associated psychosocial sequelae². Trans-sutural distraction osteogenesis (TSDO) represents a mechanobiological innovation wherein controlled mechanical stimulation of cranial sutures stimulates osteogenesis while progressively realigning skeletal architecture³. Compared to conventional osteotomy, TSDO demonstrates enhanced biomechanical stability, achieving comparable therapeutic outcomes with significantly reduced perioperative risks.

Craniofacial sutures are specialized fibrous joints that interconnect osseous structures, functioning as biomechanical niches for diverse cellular lineages while orchestrating the transduction of mechanical forces and integration of biochemical signaling^4–6. Contemporary investigations into TSDO mechanisms have predominantly investigated discrete cellular components, including suture mesenchymal stem cells (SuSCs), vascular endothelia, osteoprogenitors, immunocytes and dynamic cytokine profiles. SuSCs emerge as pivotal regulators of sutural biomechanics, osteogenic regeneration, and injury response⁷. They actively adjust their cellular activities under mechanical stimuli, which is vital for suture homeostasis.

Nevertheless, current researches have critically overlooked the extracellular matrix (ECM), which is a fundamental constituent of the mechanochemical microenvironment that directly mediates mechanical transmission and regulates cell fate. The extracellular matrix (ECM), functioning as a three-dimensional network structure, plays a crucial role in maintaining suture homeostasis and mechanosensation during TSDO⁸. The adaptive remodeling of ECM under mechanical loading proves indispensable for suture remodeling and de novo bone formation^9–14. As an essential component of the sutural microenvironment, the structural and physicochemical properties of ECM directly regulate cellular activities within the suture, steering proliferation, differentiation, and developmental trajectories of sutural cells^15,16.

However, critical knowledge gaps persist regarding ECM dynamics during TSDO, particularly collagen remodeling. Firstly, the spatiotemporal characteristics of collagen reorganization remain undefined, and the mechano-regulatory principles governing sutural adaptation require further elucidation. Secondly, the precise regulatory mechanisms underlying sutural collagen remodeling remain obscure, particularly the bidirectional interactions between resident cells and ECM components. These theoretical gaps impede comprehensive understanding of TSDO and hinder therapeutic optimization. To address these questions, we employed second harmonic generation (SHG) imaging to map collagen dynamics throughout TSDO. Through integrated transcriptomic profiling and functional validation experiments, we further elucidated macrophage-mediated regulatory mechanisms governing suture mesenchymal stem cell (SuSCs) collagen remodeling via PDGF signaling cascades. This study aims to unveil the mechanobiological underpinnings of TSDO-induced osteogenesis, potentially revealing novel therapeutic targets for bone regeneration strategies.

Results

Morpho-dynamic features of mechano-oriented collagen remodeling post-TSDO

We successfully established a 4 week-old trans-sutural distraction osteogenesis (TSDO) mouse model using a flexible nickel-titanium “W”-shaped distraction device on the zygomatico-maxillary complex (Fig. 1a). To track structural changes, we performed Micro-CT scans at days 1/3/7 post-TSDO. Visual findings revealed progressive straightening of the zygomatico-maxillary curvature and suture widening in TSDO group (DO group), with less smooth osteogenic fronts compared to sham control group (SC group). Significant increases were observed in bone circumference, cross-sectional area, and bone volume (BV) for DO group. Although bone mineral density in the DO group showed a significant decrease on day 3, no statistically significant difference was observed between the two groups by day 7 (Fig. 1b). To study microscopic morphological changes, we analyzed zygomatico-maxillary suture (ZMS) from days 1/3/7 using HE staining. The DO group showed progressively wider sutures with increased cell growth over time, and cell arrangement changed from 3−5 loose layers to multiple elongated layers with finger-like new bone appearing at both osteogenic fronts (Fig. 1c). Ki-67 staining indicated significant cellular proliferation within the cranial suture at 3 days of distraction (Supplementary Fig. 1a). To understand the effects of mechanical force, we measured cell orientation angles. DO group cells mainly aligned within ±30 ° of the force direction while SC group showing no obvious trend in direction (Fig. 1d). As a hub for material exchange, blood vessels serve as crucial supportive structures for the growth and development of cranial sutures. Therefore, we investigated the changes in sutural blood vessels. CD31 staining revealed a greater number of blood vessels in the DO group, indicating more active vascular formation (Fig. 1e).

Fig. 1. Morphological changes of TSDO and neovascularization.

a Schematic diagram of TSDO modeling. A W-shaped titanium alloy expander was applied to deliver mechanical force across the maxilla-zygomatic complex in mice. b Comparative micro-CT images of craniofacial structures pre- and post-TSDO. Morphological alterations in the zygomaticomaxillary region were observed, with quantitative analyses demonstrating significant increases in bone perimeter, cross-sectional area, and BV after TSDO, while BMD remained unchanged (Scale bars = 2 mm and 1 mm). c H&E staining of the ZMS. Post-distraction specimens exhibited increased suture width and cellular density (Scale bar = 100 μm). d Polar plot demonstrating the angular distribution of cell long-axis orientation within the suture (0 °: parallel to mechanical loading direction). e CD31 immunohistochemical staining with quantitative analysis of vascular density pre- and post-TSDO (Scale bars = 250 μm and 50 μm). Data are presented as mean ± SEM (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

To characterize collagen fiber remodeling, Masson’s staining revealed increased neo-collagen deposition and significantly elevated collagen volume fraction (CVF) in DO group, indicating enhanced osteogenic activity at the osteogenic fronts (Fig. 2a). Through second harmonic generation (SHG) imaging, ZMS was divided into three parts: head, body and tail (Supplementary Fig. 1b). The results showed disorganized collagen fiber arrangement without specific directional preference in SC group across all time points, whereas DO group exhibited progressive alignment parallel to the horizontal distraction force (Fig. 2b). Quantitative analysis of anisotropy, entropy, inverse difference moment (IDM) and average orientation confirmed transition from disordered to force-aligned collagen organization in DO group, with mean fiber orientation angles <10 ° relative to the distraction axis (Fig. 2c). Color-coded orientation maps and coherency analysis further illustrated chaotic fiber distribution in SC group versus uniform alignment in DO group (fibers predominantly within ±40 ° of force direction, Fig. 2d). Radar plots confirmed scattered fiber orientations in SC group across all angles, contrasting with DO group’s predominant ±40 ° alignment and reduced perpendicular distribution (Fig. 2e). Comprehensive analysis demonstrated mechanical force-driven collagen reorganization along the distraction axis.

Fig. 2. Collagen morphological changes after TSDO in mice.

a Masson’s staining showed increased collagen deposition and enhanced new bone formation at the bilateral osteogenic fronts in the DO group, with a significantly elevated collagen volume fraction (CVF) (Scale bar = 100 μm). b Second harmonic generation (SHG) imaging revealed that the ZMS gradually widened after TSDO, and the collagen fibers became increasingly aligned parallel to the direction of the mechanical force (Scale bar = 70 μm and 25 μm). c Quantitative analysis showed that the anisotropy, entropy, and mean orientation angle of collagen fibers within the ZMS significantly decreased after TSDO, while the inverse difference moment (IDM) significantly increased. d Color-coded orientation maps and coherency analysis indicated that the distribution direction of collagen fibers became more uniform after TSDO. e Radar plots showed that the orientation of collagen fibers in the DO group was mainly distributed within ±40 °, while the SC group showed a disordered arrangement. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Multi-omics analysis investigates ZMS-mediated regulation of cellular subpopulations and potential mechanisms

To explore TSDO-induced transcriptional dynamics, bulk RNA-seq was performed on ZMS samples from different time points (days 1, 3, 7) and groups (DO vs SC). Gene expression clustering heatmaps revealed distinct profiles between the two groups (Supplementary Fig. 1c). Principal component analysis (PCA) showed close clustering of SC group samples across time points, whereas DO group samples exhibited progressive divergence (Supplementary Fig. 1d). Differential gene analysis and volcano plot results indicated that the highest number of differentially expressed genes (DEGs) and the most pronounced changes were observed within the ZMS at 3 days of distraction (Fig. 3a, Supplementary Table 2).

Fig. 3. Bulk RNA-sequencing analysis of the ZMS in mice.

a Volcano plots of differentially expressed genes (DEGs) between the SC and DO groups at different time points. b Functional enrichment analysis of DEGs between the SC and DO groups at different time points, showing significant changes in pathways related to extracellular matrix remodeling and collagen metabolism. (BP: blue, CC: green, MF: pink). c Weighted gene co-expression network analysis (WGCNA), analyzing the correlation between gene modules and phenotypes, with CVF and stretch status as the investigated phenotypes. d Functional enrichment analysis of the Floralwhite and Cyan modules.

Gene ontology (GO) enrichment analysis of differentially expressed genes demonstrated time-dependent pathway activation: ECM remodeling and bone morphogenesis at day 1, immune-related pathways with fibroconnective tissue development at day 3, and stabilized cell adhesion with persistent ECM reorganization by day 7 (Fig. 3b). Weighted gene co-expression network analysis (WGCNA) identified 29 modules (Supplementary Fig. 1e), with cyan and floralwhite modules showing strongest correlation to CVF (Fig. 3c). GO enrichment revealed ECM remodeling/osteogenesis pathways in the floralwhite module and secretion-related mechanisms in the cyan module (Fig. 3d).

Single-cell RNA sequencing of ZMS from DO and SC groups at days 1 and 3 captured 29,295 cells, with UMAP clustering resolving 11 distinct cellular populations including neutrophils, monocytes, B cells, fibroblast-like cells, and macrophages (Fig. 4a). Cell proportion analysis revealed changes in the relative abundance of different cell types within the mouse cranial suture before and after force application (Fig. 4b, c). Notably, the proportions of macrophages and fibroblast-like cells showed the most significant alterations (From 0.21% to 0.83% and 0.48% to 1.17%, Supplementary Table 3). Differential gene expression analysis demonstrated that fibroblast-like cells and macrophage exhibited the highest total number of both up-regulated and down-regulated genes, reaching 573 and 501 genes (Supplementary Table 3).

Fig. 4. Single-cell RNA sequencing (scRNA-seq) analysis of the ZMS in mice.

a UMAP clustering analysis of 29,295 cells captured from the SC and DO groups at days 1 and 3, identifying 11 distinct cell populations. b Changes in the proportions of different cell types within the ZMS suture after TSDO, showing an increase in the proportion of macrophages and fibroblast-like cells in the DO group. c Functional enrichment analysis of marker genes for each cell type revealed distinct functional profiles, with fibroblast-like cells strongly associated with extracellular matrix (ECM) remodeling and mineralization. d Transcriptional changes in each cell type after TSDO, with yellow indicating the number of upregulated genes and cyan indicating the number of downregulated genes. e Distribution of collagen-related genes (Col1a1, Col1a2, Col3a1, Col4a1, Col5a1, Col11a1, and Col22a1) across different cell types, showing that fibroblast-like cells were the main expressors of these genes, with increased Col1a1, Col1a2, Col5a1, Col11a1 and Col22a1 expression in the DO group compared to the SC group. f Further subpopulation analysis divided fibroblast-like cells into two clusters, marked by Dcn⁺ and Adgrf5⁺, respectively. g Comparison of the proportions of fibroblast-like cell subpopulations between the DO and SC groups, showing an increased proportion of Dcn⁺ fibroblast-like cells in the DO group. h Functional enrichment analysis of highly expressed genes in the two fibroblast-like cell subpopulations, indicating that Dcn⁺ fibroblast-like cells are primarily associated with mineralization and ECM remodeling. i Expression distribution of collagen-related genes in the two fibroblast-like cell subpopulations, showing that collagen-related genes are predominantly expressed by Dcn⁺ cells. j Cell trajectory analysis revealing two differentiation branches, with darker blue indicating earlier pseudotime.

GO enrichment revealed ECM remodeling and ossification in fibroblast-like cells (Fig. 4d). Collagen genes (Col1a1, Col1a2, Col3a1, Col4a1, Col5a1, Col11a1 and Col22a1) were predominantly expressed in fibroblast-like cells. Among these, Col1a1, Col1a2, Col5a1, Col11a1 and Col22a1 were significantly upregulated following suture expansion in DO versus SC group (Fig. 4e, Supplementary Table 4). Sub-clustering identified Dcn⁺ and Adgrf5⁺ fibroblast subsets, with Dcn⁺ proportion increasing by 10% post-TSDO (Fig. 4f, g). Immunofluorescence staining of tissue sections further confirmed the existence of these two cell subpopulations (Supplementary Fig. 2a). GO enrichment analysis showed calcification, ECM remodeling and collagen assembly were enriched in Dcn⁺ cell, while migration and behavior regulation were enriched in Adgrf5⁺ cell (Fig. 4h). Critical collagen genes were primarily expressed in Dcn⁺ subset (Fig. 4i). Pseudotime trajectory analysis delineated three branching points guiding differentiation from naive fibroblasts to Adgrf5⁺ or Dcn⁺ lineages (Fig. 4j).

Cell-cell communication analysis reveals potential collagen regulatory mechanisms between macrophages and fibroblast-like cells

To investigate cellular interactions during TSDO, cell-cell interaction analysis revealed extensive bidirectional communication between fibroblast-like cells and macrophages (Fig. 5a). Systematic evaluation of ligand-receptor pairs demonstrated enhanced collagen receptor activation in fibroblast-like cells concurrent with upregulation of macrophage-to-fibroblast PDGF signaling axis (Fig. 5b). Relative information flow corroborated these interaction patterns (Fig. 5c). Cell type-specific expression profiling identified elevated PDGF-related ligands in macrophages and corresponding receptors in fibroblast-like cells as well (Fig. 5d). Concomitant upregulation of collagen-associated pathways was observed in fibroblast-like cell (Fig. 5e). These findings suggest macrophage-derived PDGF signaling may orchestrate fibroblast-like cell activation to potentiate collagen remodeling during TSDO.

Fig. 5. Cell-cell communication analysis of scRNA-seq and immunofluorescence staining of macrophages in the ZMS.

a Cell-cell interaction analysis revealed extensive bidirectional communication between fibroblast-like cells and macrophages. b (An overview of craniosynostosis craniofacial syndromes for combined orthodontic and surgical management) Ligand-receptor pair analysis showed increased interactions between collagens and their receptors, as well as enhanced PDGF signaling between macrophages and fibroblast-like cells, in the DO group. c Analysis of relative information flow further indicated an increase in PDGF signaling in the DO group. d, e Ligand-receptor expression plots showing upregulation of PDGF signaling and collagen-related signaling pathways in the DO group. f Immunofluorescence staining and quantification of F4/80, CD86, and CD206 in ZMS tissue sections. Positive signals for all markers were significantly increased in the DO group compared to the SC group. F4/80 was mainly distributed around blood vessels, CD86 was sparsely distributed along the midline of the ZMS, and CD206 was diffusely distributed within the suture (Scale bars = 200 μm). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

To detect changes in macrophage subtypes within ZMS after TSDO, immunofluorescence staining for F4/80, CD86, and CD206 was performed on tissue sections. Results showed positive signals for F4/80, CD86, and CD206 were primarily distributed within the suture and surrounding zygomatic bone marrow cavities, mainly concentrated along the sutural midline. Compared to the SC group, the DO group exhibited significantly increased positive signals at all time points. At day 3 and 7, F4/80-positive signals were observed around vascular lumens, CD86-positive signals were scattered along the sutural midline extending toward osteogenic fronts, while CD206-positive signals diffusely distributed within ZMS. Quantitative analysis of relative fluorescence intensity for F4/80, CD86, and CD206 confirmed significantly enhanced signals in DO group, indicating increased immune cell subtypes post-TSDO (Fig. 5f). Notably, CD206 fluorescence intensity markedly increased in DO group at day 7, suggesting substantial M2 macrophage expansion.

The PDGF signaling pathway mediates macrophage-regulated collagen expression in SuSCs

To investigate stretch-induced collagen synthesis, we cultured SuSCs in vitro and validated their stemness and other properties (Fig. S1e–g). SuSCs were subjected to in vitro cyclic stretching. SuSCs were seeded into elastic stretching chambers pre-coated with rat-tail type I collagen and randomized into control (static culture) and stretch groups (8% strain amplitude, 0.5 Hz sinusoidal waveform) (Fig. 6a). RNA samples collected at 2 h, 4 h, 6 h, and 3 d post-stretching underwent RT-qPCR analysis. Mechanical stimulation significantly upregulated collagen gene expression (Col1A1, Col1A2, Col3A1, Col5A1 and Col11A1), with expression levels increasing progressively with prolonged stretching duration (Fig. 6b).

Fig. 6. Mechanistic insights into mechanically induced collagen remodeling.

a Schematic diagram of the in vitro cyclic stretching experiments, showing primary cultures of SuSCs and macrophages, followed by cyclic stretching and co-culture experiments. b Expression levels of various collagen genes in SuSCs after different durations of cyclic stretching. c Changes in PDGF ligand expression in macrophages after different durations of cyclic stretching. d Immunohistochemical (IHC) staining of ZMS sections showing a significant increase in PDGF-B after cyclic stretching (Scale bars = 50 μm). e Expression levels of various collagen genes in SuSCs after different durations of cyclic stretching in the Stretch, Stretch + Co-culture, and Stretch + PDGF-BB groups, showing that collagen-related gene expression was markedly increased after PDGF-BB treatment. f Expression levels of PDGFRa in SuSCs across different groups. g Immunofluorescence staining of PDGFRa in SuSCs from different groups. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. In b and c, the same letter indicates no significant difference, while different letters represent statistically significant differences.

To validate the PDGF signaling axis between macrophages and SuSCs in collagen regulation identified by transcriptomic analysis, peritoneal macrophages from mice were subjected to mechanical stretching in vitro. RT-qPCR analysis of macrophages after 2 h continuous or 12 h intermittent stretching revealed mechanical stretching-induced upregulation of PDGF ligands: both Pdgfa and Pdgfb showed significant increases versus control group, peaking at 12 h (Fig. 6c). Immunohistochemical staining of PDGF-B in ZMS confirmed these findings in vivo. DO group exhibited intensified PDGF-B staining at day 3 compared to SC group, though no intergroup difference persisted by day 7 (Fig. 6d). To simulate and investigate whether the presence of macrophages affects collagen expression in SuSCs, and to clarify the macrophage-SuSCs crosstalk, SuSCs were co-cultured with stretched macrophage-conditioned medium. RT-qPCR demonstrated time-dependent collagen gene (Col1A1, Col1A2, Col3A1 and Col11A1) upregulation in co-cultured SuSCs, with maximal induction at 3 day stretching (Fig. 6e).

To isolate PDGF signaling effects on collagen regulation, SuSCs were treated with PDGF-BB (20 ng/mL) and subjected to stretching. RT-qPCR revealed significant upregulation of Col1A1, Col1A2, Col3A1 and Col11A1 at 3 d post-stretching compared to the stretch group and co-culture group (Fig. 6e). To investigate PDGFR dynamics in SuSCs under macrophage-conditioned medium co-culture, RNA samples from stretched cells (2 h, 4 h, 6 h and 3 d) with/without co-culture were analyzed via RT-qPCR. Co-cultured SuSCs exhibited markedly elevated PDGFR levels after 3 d of stretching compared to the stretch group (Fig. 6f). Immunofluorescence confirmed enhanced PDGFRα membrane localization in both stretch and PDGF-BB-treated groups versus controls (Fig. 6g), with PDGF-BB further promoting SuSC migration (Supplementary Fig. 2b).

Macrophage elimination impairs PDGF signaling and collagen remodeling in ZMS

To further validate the role of macrophage PDGF signaling in collagen remodeling of ZMS, clodronate liposomes were used to eliminate macrophages in mice. Administration of clodronate liposomes resulted in the near-complete elimination of macrophages within the cranial sutures, indicating the successful depletion (Supplementary Fig. 2c). These mice received the TSDO surgical procedure following macrophage depletion. HE staining showed that macrophage-eliminated (ME) mice exhibited significant suture widening after 3 and 7 days of mechanical distraction, with no marked reduction in cell proliferation compared to DO group. By day 7, the ME group exhibited discontinuous osteogenic fronts and incomplete ossification at the suture edges, forming island-like bone fragments embedded within fibrous connective tissue. This aberrant ossification pattern progressed by day 14, manifesting as fragmented bone margins, scattered bilateral mineralization, and bone channels bridging the suture to the marrow space, collectively suggesting impaired osteogenic activity. (As shown by the arrow in the Fig. 7a). Masson’s staining revealed significantly reduced blue staining at osteogenic fronts in ME group, suggesting limited collagen formation under mechanical stimulation (Fig. 7b). SHG imaging demonstrated disordered collagen fiber arrangement in ME group, resembling the SC group. Quantitative analysis showed random collagen orientation, loss of force-directional alignment, increased anisotropy and entropy values, and decreased IDM, indicating disrupted collagen remodeling (Fig. 7c-d). Immunohistochemistry confirmed downregulated expression of PDGF-B and PDGFRα in ME group, while PDGFRβ showed no significant change among different groups (Fig. 7e, f, Supplementary Fig. 2d).

Fig. 7. Changes in collagen fibers in the ZMS after macrophage elimination.

a H&E staining showed that the ZMS remained wide after stretching in the macrophage elimination (ME) group, but the newly formed bone appeared fragmented and disorganized, lacking a continuous and smooth osteogenic front (Scale bars = 250 μm). b Masson’s staining showed reduced blue staining at both osteogenic fronts and within the suture in the ME group, indicating impaired collagen deposition (Scale bars = 100 μm). c Second harmonic generation (SHG) imaging showed that collagen fiber remodeling was impaired in the ME group, and the characteristic changes observed in collagen fibers in the DO group were not seen (Scale bars = 70 μm and 25 μm). d Quantitative analysis of SHG images revealed that the average angle orientation, anisotropy, and entropy were significantly increased, while the inverse difference moment (IDM) value was significantly decreased, in the ME group compared to the DO group. e Immunohistochemical staining in the ME group showed reduced PDGF-B expression compared to the DO group (Scale bars = 50 μm). f Immunohistochemical staining in the ME group showed significantly reduced PDGFRa expression (Scale bars = 100 μm). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Discussion

Though current evidence suggests mechanical forces play a pivotal role in ECM remodeling, existing quantitative approaches predominantly characterize collagen fibers through bulk measurements, lacking precise spatial resolution of fiber orientation and physicochemical properties^17–22. The dynamic reorganization of ECM collagen architecture during TSDO remains poorly delineated, with no systematic investigations reported. Deciphering the mechano-regulatory effects on ECM structure is imperative for elucidating the mechano-transduction pathways underlying TSDO-driven osteogenesis. We note that second-harmonic generation (SHG) imaging via two-photon microscopy, as a non-invasive technique with high resolution and specificity, is particularly suitable for quantifying collagen fiber abundance and three-dimensional structure in various tissues^23,24. In this study, we employed multimodal approach combining SHG-based morphometrics with histological staining for multidimensional characterization of collagen reorganization during TSDO. Our findings demonstrate that mechanical forces not only induce sutural expansion but also trigger adaptive morphological changes in collagen fibers, transitioning from isotropic disorder to anisotropic alignment parallel to force vectors. These results indicate mechanical force is the master regulator orchestrating both sutural adaptation and ECM remodeling.

Existing studies have demonstrated that mechanical loading enhances the stability of collagen fibers, a phenomenon termed the “use it or lose it” effect^25–28. We observed a similar phenomenon of collagen remodeling in ZMS during TSDO, which is critical for new bone formation. As a dynamic component of mechano-biochemical responses, mechanical forces directly drive the remodeling and reorganization of collagen fibers in sutural ECM²⁹. The density and architecture of collagen fibers significantly influence the stiffness and elasticity of ECM, providing a material foundation, structural support, and cellular scaffolds for mechano-sensing^30–33. In the context of TSDO, the interaction between collagen and cells is essential for perceiving and adapting to mechanical forces^34–36. We hypothesize that collagen fibers transmit mechanical forces within the ZMS, and these mechanical stimuli are relayed to cells via ECM-cell adhesions^37–39. To clarify the impact of collagen remodeling on cellular behavior, we analyzed morphological changes in cells pre- and post-TSDO. Post-distraction, the long axis of cells aligned with the orientation of collagen fibers, exhibiting polarization parallel to the direction of force. This indicates that adaptive morphological changes in cells regulated by ECM transformations signify the initiation of diverse processes, including proliferation, migration, differentiation, and osteogenesis. Collagen fibers act as regulatory switches that modulate the microenvironmental transition within the suture, thereby influencing cell fate^40–42. Furthermore, the coordinated remodeling and reorganization of collagen and cells in mechanical environments reflect the suture’s ability to sense and respond to mechanical forces, which is crucial for the success of TSDO.

To further investigate the underlying mechanisms of collagen remodeling during TSDO, we employed transcriptomic sequencing which showed dynamic changes in functional pathways within the ZMS during TSDO, revealing temporal shifts in transcriptional priorities. And the regulatory circuit of “ECM sensing-cellular response-collagen remodeling” was delineated. Cell-cell communication analysis, in vivo and in vitro experiments confirmed that mechanical stretch induced macrophage secretion of PDGF ligands, which modulated SuSCs’ morphological changes and collagen expression, ultimately altering ECM architecture. The PDGF signaling pathway has diverse biological roles in various tissues, including promoting mitosis, enhancing motility, and inducing chemotaxis^43,44. In the musculoskeletal system, activation of the PDGF pathway enhances biomechanical properties through multifaceted mechanisms including amplified inflammatory responses, accelerated angiogenesis, and stimulated cellular proliferation and differentiation, while concurrently promoting bone neoformation through coordinated osteogenic processes^45–47.

Furthermore, macrophage-derived PDGF executes specific biological functions encompassing tissue repair promotion, fibrotic response modulation, and optimization of bone remodeling microenvironment^48–50. Our previous study demonstrated that under mechanical distraction, macrophages were recruited by various chemokines and tended to polarize toward M2 phenotype, regulating SuSCs activity and promoting bone regeneration¹⁶. These findings collectively indicate the PDGF signaling pathway plays a pivotal role in mediating macrophage-derived regulatory effects on mesenchymal cells during musculoskeletal homeostasis. What’s more, ECM remodeling, which is essential for osteogenesis within suture, begins in the early stages of TSDO and persists throughout the whole osteogenic process. This study further confirms that the PDGF signaling pathway acts as a crucial link between macrophages and SuSCs, playing a key role in the mechanical sensing–immune response–ECM remodeling cascade. Meanwhile, as indicated by single-cell transcriptome analysis, we have also noted that SuSCs themselves might regulate through autocrine PDGF signaling. However, current research on SuSCs’ autocrine PDGF regulation is relatively scarce. SuSCs primarily function by responding to external PDGF signals for subsequent regulation^51–53. Therefore, we believe that although SuSCs have the capacity to secrete PDGF ligands, the main focus should remain on their receptors.

The present findings in this study significantly expand our understanding of the underlying mechanisms governing bone suture remodeling, thereby paving the way for more profound investigations into the mechanical microenvironment of sutural tissues. To the best of our knowledge, this study is also the first to focus on the spatiotemporal changes in suture collagen remodeling and partially reveals the biological mechanisms involved during TSDO. However, there are still several limitations in this study. For example, we didn’t fully analyze the expression changes of various subtypes due to the diversity of macrophage types and PDGF family members. Therefore, there is much room for improvement in functional validation.

Collectively, our investigation elucidates the reorganization of sutural collagen architecture and force-dependent cellular dynamics during TSDO, thereby establishing the PDGF signaling cascade as a pivotal molecular nexus orchestrating macrophage-SuSCs mechano-transduction crosstalk. This mechanistic deciphering of ECM remodeling with cellular activities not only provides new perspectives of craniofacial mechanobiology, but also unveils the potential strategies for the clinical treatment.

Methods

Ethical approval was obtained from the Peking University Biomedical Ethics Committee (No. SA2022341). Healthy 4 week-old male C57BL/6 J mice (sourced from Peking University Health Science Center Department of Laboratory Animal Science, Beijing, China) were used in this study. All germ-free C57BL/6 J mice were bred in sterile vinyl isolators and housed at the gnotobiotic mouse facility within the Department of Laboratory Animal Science, Peking University Health Science Center.

Surgical operation

Trans-sutural distraction was implemented on the ZMS in mice, establishing both TSDO group (DO group) and sham control group (SC group). Our operational methodology was adapted from established protocols developed by previous researchers, with modifications to the distraction device dimensions according to the mouse zygomatic arch length^54–57. Following induction of general anesthesia with a 1% pentobarbital sodium solution (30-90 mg/kg, IP), the bilateral infraorbital skin incisions were made, exposing and separating the masticatory and temporal muscles until the maxilla and temporal bone were visible. In DO group, two holes were drilled on each side of the maxilla and temporal bone, positioned ~7 mm apart. Subsequently, a “W”-shaped expander, composed of nickel-titanium alloy with an original length of 12 mm and a diameter of 0.25 mm, was inserted. The expander was calibrated to exert an initial force of ~30 g for trans-sutural distraction. The SC group underwent an identical surgical procedure, but without the placement of an expander device. On days 1, 3, and 7, all mice were euthanized using excess carbon dioxide, and ZMS tissues were collected, excluding muscles, fat, or subcutaneous connective tissues.

Establishment of macrophage elimination TSDO mouse model

TSDO was performed on 4 week-old C57BL6/J mice. A macrophage elimination (ME) group was established by administering 0.05 mg/mL clodronate liposomes (m-Clodrosome®, Encapsula NanoSciences). The drug was injected intraperitoneally 2 days prior to modeling, with subsequent maintenance injections every other day until tissue collection, at a dose of 10 μL per 10 g body weight. The TSDO surgical procedure and subsequent tissue collection were performed identically to the control group, which did not receive the drug injection.

Micro-computed tomography (micro-CT) imaging and analysis

The entire mouse skull was fixed and stored in a 4% paraformaldehyde solution. The skull samples were then scanned using a Quantum FX Micro-CT scanner, and 3D models were reconstructed. The scanner’s built-in software was used to analyze craniofacial bone parameters, including bone volume (BV) and bone mineral density (BMD). The scan files were saved in DICOM format and imported into Mimics Research 23.0 software. After setting the bone window threshold, the 3D models were reconstructed, and relevant data, such as the circumference and length of the mouse maxilla-zygomatic arch, were measured and recorded in Excel.

Bulk RNA library preparation and sequencing

The total RNA quantity and integrity were evaluated using the RNA Nano 6000 Assay Kit on the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). Total RNA served as the starting material for RNA sample preparations. In summary, mRNA isolation from total RNA involved the use of poly-T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature in First Strand Synthesis Reaction Buffer(5X). The first strand cDNA synthesis utilized a random hexamer primer and M-MuLV Reverse Transcriptase, followed by RNA degradation using RNaseH. Subsequently, second strand cDNA synthesis was executed using DNA Polymerase I and dNTP, and remaining overhangs were transformed into blunt ends through exonuclease/polymerase activities. After adenylation of the 3’ ends of DNA fragments, adapters with a hairpin loop structure were ligated for hybridization preparation. In order to select cDNA fragments of preferentially 370 ~ 420 bp in length, the library fragments were purified with AMPure XP system (Beckman Coulter, Beverly, USA). Following PCR amplification, the PCR product underwent purification with AMPure XP beads, resulting in the final library. Quality control of the library was imperative. Post-library construction, initial quantification occurred using the Qubit 2.0 Fluorometer. The library was then diluted to 1.5 ng/μl, and the insert size was determined using the Agilent 2100 bioanalyzer. Upon confirmation that the insert size met expectations, qRT-PCR was employed for accurate quantification of the library’s effective concentration (ensuring it exceeded 2 nM) to guarantee its quality. After the qualification of the library, the different libraries were pooling according to the effective concentration and the target amount of data off the machine, then being sequenced by the Illumina NovaSeq 6000. The end reading of 150 bp pairing is generated. The sequencing methodology employed was known as Sequencing by Synthesis, where synthesis and sequencing occurred concurrently. Four fluorescent labeled dNTP, DNA polymerase and splice primers were added to the sequenced flowcell and amplified. As the sequence cluster extended the complementary chain, each fluorescently labeled dNTP released corresponding fluorescence. The sequencer captured these fluorescence signals and, through computer software, converted the optical signals into sequencing peaks. This process facilitated the acquisition of sequence information for the tested fragment.

Bulk RNA sequencing data analysis

The image data obtained from the high-throughput sequencer underwent conversion into sequence data (reads) through CASAVA base recognition. Raw data, initially in fastq format, underwent preliminary processing via in-house perl scripts. During this phase, clean data (clean reads) were derived by excluding reads containing adapters, those with N bases, and low-quality reads from the raw dataset. Simultaneously, metrics such as Q20, Q30, and GC content for the clean data were calculated. All downstream analyses were conducted exclusively on the high-quality clean data. The reference genome and gene model annotation files were acquired directly from the genome website. Utilizing Hisat2 (v2.0.5), an index for the reference genome was established, and paired-end clean reads were aligned. Hisat2 was also chosen as the mapping tool due to its capability to generate a splice junctions database based on the gene model annotation file. For read quantification, FeatureCounts (v1.5.0-p3) was employed to count the number of reads mapped to each gene. Subsequently, the Fragments Per Kilobase of transcript sequence per Millions base pairs sequenced (FPKM) for each gene was calculated, taking into account both the gene’s length and the reads count mapped to it.

The analysis of differential expression between two groups (each with three biological replicates) was conducted using the DESeq2 R package (version 1.40.2). DESeq2 offers statistical procedures for identifying differential expression in digital gene expression data, employing a model based on the negative binomial distribution. The P-values obtained were subjected to adjustment using the Benjamini and Hochberg approach to control the false discovery rate. Significantly differential expression was defined by a threshold of p-value < 0.05 and |log2(fold change)| > 1.

The enrichment analysis for Gene Ontology (GO) of differentially expressed genes was conducted using the ClusterProfiler R package (version 4.8.3), incorporating correction for gene length bias. GO terms exhibiting a P-value below 0.05 were deemed significantly enriched by differentially expressed genes.

Weighted correlation network analysis (WGCNA) is a systematic biological approach employed to characterize gene association patterns across different samples. The WGCNA methodology outlined in the literature was adhered to for this analysis⁵⁸. WGCNA was applied to investigate co-expression modules and identify key genes associated with Stretch Force and Collagen Volume Fraction. A soft-threshold (β) of 9 was selected with a resulting R² of 0.9. Subsequently, the adjacency matrix underwent transformation into a topological overlap matrix (TOM). Modules were delineated through hierarchical clustering (minModuleSize = 20). The eigengene was calculated, and modules were hierarchically clustered. The correlation between phenotypic data (Stretch Force and Collagen Volume Fraction) and the modules was calculated to identify crucial modules. Module eigengene (ME) represents the first principal component within a module, encapsulating its expression pattern. Module membership (MM) reflects the association between genes and module eigengenes, signifying the reliability of genes within modules. Genes within the key module were subjected to further functional enrichment analysis, utilizing the GO, through the ClusterProfiler R package (version 4.8.3).

Single-cell RNA library preparation and sequencing

Following harvest, tissues were rinsed in ice-cold RPMI1640 and enzymatically dissociated using Collagenase Ⅱ (Sigma, V900892-100MG) and DNase Ⅰ (Sigma, DN25-1G). Cell counts and viability was estimated using fluorescence Cell Analyzer (Countstar® Rigel S2) with AO/PI reagent after removal erythrocytes (Solarbio R1010) and then debris and dead cells removal was decided to be performed or not (Miltenyi 130-109-398/130-090-101). Finally, the freshly isolated cells were subjected to two washes in RPMI1640 and resuspended at a concentration of 1 × 10⁶ cells per ml in 1×PBS with 0.04% bovine serum albumin. In each experimental group, two dissected sutures were utilized for sequencing analysis.

Single-cell RNA-Seq libraries were crafted using the SeekOne® MM Single Cell 3’ library preparation kit (SeekGene Catalog No. SO01V3.1). In a concise overview, the appropriate number of cells was introduced into the SeekOne® MM chip’s flow channel, housing 170,000 microwells, where they settled by gravity. After the removal of unsettled cells, an ample quantity of Cell Barcoded Magnetic Beads (CBBs) was carefully added to the flow channel, settling in microwells under the influence of a magnetic field. Following rinsing to eliminate excess CBBs, the cells within the MM chip were lysed to release RNA, subsequently captured by the CBBs within the same microwell. The collected CBBs underwent reverse transcription at 37 °C for 30 min, labeling cDNA with cell barcodes affixed to the beads. Further Exonuclease I treatment were performed to remove unused primer on CBBs. Subsequently, the barcoded cDNA on the CBBs was hybridized with a random primer featuring the SeqPrimer sequence on the 5’ end, extending to form the second strand DNA with a cell barcode on the 3’ end. The resulting second strand DNA was denatured off the CBBs, purified, and subjected to amplification in PCR reaction. The amplified cDNA product underwent cleaning to eliminate unwanted fragments, followed by the addition of a full-length sequencing adapter and a sample index through indexed PCR. The indexed sequencing libraries were purified using SPRI beads, quantified via quantitative PCR (KAPA Biosystems KK4824), and subsequently sequenced on the Illumina NovaSeq 6000 with a paired-end read length of 150 base pairs.

Cell clustering and marker recognition

The Seurat package “NormalizeData” function was employed to normalize the matrix data of single-cell counts. In this process, gene expression for each cell was normalized by dividing the total expression counts, multiplied by a scale factor of 10000, and subsequently log-transformed. After normalization, Principal Component Analysis (PCA) was conducted on the normalized expression matrix, focusing on highly variable genes identified by the Seurat R package (version 4.3.0)⁵⁹. From the PCA results, the top 20 Principal Components (PCs) were selected, with a resolution parameter set to 0.8. Visualization of cell clusters was achieved using T-distributed Stochastic Neighbor Embedding (TSNE). The “FindAllMarkers” function was then applied to pinpoint differentially expressed marker genes for each cell type. Marker genes were defined as those with adjusted p-values below 0.05, an average log2FoldChange exceeding 0.25, and a percentage of cells exhibiting expression >0.25. The determination of cell type identities relied on the expression patterns of known markers sourced from the CellMarker 2.0 database⁶⁰.

scRNA-seq differential expression analysis and enrichment analysis

The differentially expressed genes (DEGs) were identified using the edgeR package and filtered by |log2FoldChange | ≥ 1 and FDR < 0.05. After that, GO enrichment analysis of DEGs was implemented by the ClusterProfiler R package (4.8.3).

Pseudotime analysis

The Monocle R package (version 2.28.0) was employed to assess lineage differentiation between cell populations using default parameters⁶¹. Initially, a CellDataSet object was established with the expressionFamily parameter set to negbinomial. Subsequently, genes for pseudotime ordering were chosen through the “dispersionTable” function. The “DDRTree” method was applied for dimension reduction and cell ordering along the pseudotime trajectory. Branch analysis was executed using the branched expression analysis modeling (BEAM) function. Genes significantly altered at the branch point were clustered via the “plot_genes_branched_pseudotime” functions based on distinct patterns of gene expression changes.

Acquisition of second harmonic generation (SHG) imaging of ZMS

The exposed zygomatic-maxillary tissue was properly fixed in normal saline environment, and the positive two-photon microscope (Nikon A1R MP, NIKON, Japan) with 40XW/ 0.8 (WD 3.0 mm, Water) was used to capture the images in Institute of Biophysics, Chinese Academy of Sciences. Excitation was provided by a 920 nm laser, while emitted photons with wavelengths between 400 and 492 nm were collected via photomultiplier tubes. To obtain a complete image of the slide, we acquired tile scans (512 × 512) that were stitched in larger mosaics. The observation depth was 200 μm, and the distance between each layer was 4 μm.

Hematoxylin-Eosin (HE) staining

Mouse zygomaticomaxillary bone tissues were sectioned and then dewaxed in water. They were stained with hematoxylin dye, followed by rinsing off the dye under running water, and differentiation with 0.1% hydrochloric acid ethanol. Subsequently, the sections were stained with eosin dye. Afterward, they were rapidly dehydrated in 95% anhydrous ethanol for 3 s, followed by three rounds of dehydration with anhydrous ethanol, each lasting 5 s. The sections were cleared in xylene for 3 times, each for 1 min, and finally mounted with neutral resin.

Masson’s staining

Mouse zygomaticomaxillary bone tissues were sectioned and subsequently dewaxed with water. After staining with a ready-made hematoxylin-iron solution for 5 min, the sections were thoroughly rinsed. Subsequently, they were washed with distilled water for 1 min and then treated with Masson’s blue solution for 3 min. Ponceau-magenta staining was applied for 5 min. A weak acid working solution was prepared by mixing weak acid and distilled water in a 2:1 ratio, and the sections were rinsed with this solution for 1 min. After a 1 min rinse with 1% phosphomolybdic acid, the sections were further washed with the previously prepared weak acid working solution for 1 min. Subsequently, the sections were stained with a solution of Aniline Blue for 2 min, followed by a 1 min rinse with the weak acid working solution. Then, the sections were rapidly dehydrated with 95% anhydrous ethanol for 3 s, followed by three rounds of dehydration with anhydrous ethanol, each lasting 5 s. They were subjected to xylene clearing for 3 times, each for 1 min, and finally mounted with neutral resin.

Image analysis

All SHG scanning files were opened in NIS-Elements (NIKON, Japan) and underwent subsequent processing, with the files saved in ND format. These images were rendered as 3D images using maximum projection. A region parallel to the direction of distraction force at the head of the zygomaticomaxillary suture was selected for the study, and this portion of the image was exported in TIF format. All HE staining and Masson staining slices were scanned using NanoZoomer S210 (Hamamatsu, Japan) and saved as NanoZoomer Digital Pathology Image (ndpi) files. These files were previewed using NDP.view 2 (Hamamatsu, Japan) software and output in TIF format for further analysis.

ImageJ software (NIH, Bethesda, MD, USA, available for free download at https://imagej.nih.gov/ij/download.html) was used to open the saved TIF files. Above all, use the selection tool to choose the region of interest (ROI) in the ZMS area, and then proceed with subsequent analyses for different parameters.

To quantify the optical density of collagen fiber in SHG imaging, the “Integrated density” option in the “analyze” menu of ImageJ was checked. And the “Measurement” was used. Then, the measurement results including the average optical density values in the “results” window were recorded in an Excel spreadsheet for further analysis.

To measure the anisotropy and orientation of collagen fiber in SHG imaging, the “Threshold” function in ImageJ was used to select collagen fibers in the suture. Next, the FibrilTool plugin was adopted for analysis⁶². In the “results” window, the anisotropy and orientation values were recorded in Excel for further analysis. 0 ° was defined as parallel to the direction of mechanical distraction force, and absolute values were employed for plotting.

For entropy analysis of collagen fiber in SHG imaging, the “Threshold” function in ImageJ was used to select collagen fibers in the suture. Then, gray level co-occurrence matrix (GLCM) texture plugin was used for analysis. We selected Entropy (a measure of textural disorder within the analyzed structure) and Inverse Difference Moment (IDM, an indirect parameter of textural homogeneity) to characterize the characteristics of collagen fiber. The Entropy and IDM values in the “results” window were recorded in Excel for further analysis.

To further analyze the orientation of collagen fibers in SHG imaging in the suture, the “Threshold” function in ImageJ was used to select collagen fibers in the suture. Then, the Riesz Filters method was utilized within the OrientationJ plugin for orientation analysis, and orientation distribution map⁶³. The orientation information of different collagen fibers in the “results” window was recorded in Excel. Finally, R was used for visualization. To quantify the changes in ZMS width in HE staining before and after distraction, we defined the line connecting the most concave point of the maxilla to the most prominent point of the zygomatic arch as the width of ZMS. The width of ZMS were measured for different samples and were recorded in Excel for further analysis.

To measure cell density and the number of cells per unit length within ZMS, we used StarDist plugin to segment and count cell nuclei within the suture⁶⁴. The results in the “results” window were recorded in Excel. Then, the values were divided by the corresponding suture area and suture length to obtain cell density and the number of cells per unit length within ZMS.

To quantify the characteristics of cell distribution direction in HE staining, we defined 0 ° as parallel to the direction of mechanical distraction force. And the “Threshold” function in ImageJ was used to select collagen fibers in the suture. Then, we employed the “Marking shape directionality” feature of the PAT-GEOM plugin to analyze the long and short axes of the cells and obtained the angles of cell distribution⁶⁵. The results in the “results” window were recorded in Excel, and finally R was used for visualization.

To quantify the collagen volume fraction (CVF) in Masson’s staining, we used ImageJ’s “Threshold” function to select the collagen within ZMS. And the collagen distribution area and the total suture area were recorded from the “results” window. The collagen volume fraction was obtained through dividing the former by the latter.

All biological samples analyzed in the above procedures were n = 3, and measurements were conducted by two independent researchers. The average of their measurements was used as the final result.

Quantitative real‑time PCR

Total RNA was extracted using Trizol reagent (TIANGEN, Beijing, China). Complementary DNA was reverse-transcribed using a FastKing RT Kit (TIANGEN, Beijing, China). Quantitative real-time PCR (RT-qPCR) was conducted using the QuantiNova SYBR Green PCR Kit (QIAGEN, Germany) for the following genes: Col1A1, Col1A2, Col3A1, Col5A1, Col11A1. The relative gene expression was normalized to the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (Gapdh) using the ΔΔCt method. The primer sequences utilized in our study are detailed in Supplementary Table 1.

Immunohistochemical staining

The immunohistochemical staining procedure involves several steps. First, the sections are deparaffinized by immersing them in Xylene substitutes I and II for 15 min each. Next, the sections are hydrated by sequentially placing them in anhydrous ethanol, 95% ethanol, 90% ethanol, 70% ethanol, and distilled water, each for 5 min. For antigen retrieval, the sections are immersed in sodium citrate solution and subjected to microwave treatment: 6 min on high power, followed by 15 min on low power, then cooled naturally. The slides are then transferred to PBS and washed three times on a decolorizing shaker for 3 min each. Endogenous peroxidase activity is blocked by incubating the sections in 3% hydrogen peroxide at room temperature for 25 min, after which they are washed in PBS three times. The sections are then incubated with 3% BSA at room temperature for 30 min to block nonspecific binding. After gently wiping off the blocking solution, a diluted primary antibody is applied to the sections, which are incubated overnight at 4 °C in a humidified chamber. After washing the slides in PBS three times, a species-specific HRP-labeled secondary antibody is added and incubated at room temperature for 50 min. Following another round of washing, DAB chromogenic solution is applied, and the sections are monitored under a microscope to control the color development, which should appear brownish-yellow for positive staining. The reaction is stopped by rinsing with tap water. The sections are then counterstained with hematoxylin for about 3 min, washed with tap water, and differentiated with hematoxylin differentiating solution for a few seconds. After rinsing with tap water, the sections are blued with hematoxylin bluing solution, and washed under running water. To dehydrate, the sections are sequentially immersed in 70%, 90%, 95%, and anhydrous ethanol for 5 min each, then cleared in Xylene substitute for 5 min. Finally, the slides are mounted with neutral balsam.

Immunofluorescence staining

To obtain the mouse complete maxilla-zygomatic arch tissue, the specimens were fixed in 4% paraformaldehyde solution at room temperature for 24 h, followed by decalcification in 0.5 M EDTA solution for 14 days, with the decalcification solution replaced every 2–3 days until the tissue became cartilaginous. After decalcification, the tissue samples were dehydrated overnight in 30% sucrose solution at 4 °C in the dark, then embedded in OCT compound. The samples were sectioned on a cryostat at 8 μm thickness along the sagittal plane of the maxilla-zygomatic arch, adjusting the blade angle to ensure full presentation of the zygomaticomaxillary suture. The frozen sections were stored at -80 °C and subjected to immunofluorescence staining under dark conditions. The sections were warmed at room temperature for 15 min, then immersed in PBS three times for 3 min each to remove the OCT embedding medium. The sections were blocked with 3% BSA at 37 °C for 30 min, followed by overnight incubation at 4 °C with diluted primary antibody solution. After three washes with PBS, each for 3 min, the sections were incubated with the secondary antibody at room temperature for 40 min, washed again with PBS, and stained with DAPI for 10 min. Following three additional washes with PBS, the sections were mounted and stored at -20 °C for immunofluorescence imaging within 48 h. Fluorescence images were captured using Leica imaging software and saved as TIF format files, and subsequent quantitative analysis was performed using ImageJ software.

For all groups undergoing cell immunofluorescence staining, SuSCs were processed directly in cell elasticity stretch dishes. After removing the supernatant, the cells were washed with PBS three times, each for 3 min. The cells were then fixed with 4% paraformaldehyde at room temperature for 15 min, followed by three PBS washes, each for 3 min. Next, the cells were blocked with sheep serum blocking solution at room temperature for 1 h, after which they were washed with PBS three times. The appropriately diluted primary antibody solution was then applied to the cells, and the cells were incubated overnight at 4 °C. After incubation, the primary antibody solution was removed, and the cells were washed with PBS three times. The diluted secondary antibody solution was applied, and the cells were incubated at room temperature in the dark for 1 h. Following this, the secondary antibody solution was discarded, and the cells were washed again with PBS three times. The cells were then incubated with ready-to-use DAPI stain in the dark for 10 min, washed with PBS three times, and maintained in a small amount of PBS to keep them moist. The cells were stored at 4 °C and fluorescence imaging was performed as soon as possible.

Suture mesenchymal stem cells (SuSCs) culture

The complete frontal and maxillary sutures, along with ~0.5 mm of adjacent bone, were excised from six C57BL/6 J mice and sectioned into small pieces. The tissue blocks were uniformly distributed in a 6 cm diameter culture dish and cultured in a medium composed of 0.5 mL of DMEM high-glucose basal medium, 20% fetal bovine serum, and 100 IU/mL penicillin/streptomycin under conditions of 37 °C and 5% CO_2. After 6 h, an additional 3 mL of culture medium was introduced into the dish. Following cell migration and achieving 80% confluence, primary cells underwent digestion with 0.25% trypsin for subsequent passaging. SuSCs at passage 4 to 6 were utilized for subsequent experiments. Following 48 h adherent culture of fragmented murine ZMS, radially arranged spindle-shaped cell outgrowths were observed surrounding the explants. Third-passage cells maintained stable morphology and viability (Supplementary Fig. 1f). Trilineage differentiation assays confirmed SuSCs’ multipotency (Supplementary Fig. 1g). Flow cytometry of SuSCs obtained through explant culture revealed high expression of mesenchymal stem cell markers CD90 (98.3%), CD44 (95.7%), and Sca-1 (92.4%), with partial CD29 expression (14.61%), along with low expression of negative markers CD31 (2.1%), CD34 (1.8%), and CD117 (3.5%) (Supplementary Fig. 1h).

Mechanical stretch of SuSCs

The SuSCs were seeded at a density of 2 × 10⁵ cells per well in cell-stretching chambers coated with Type I collagen (4 cm × 4 cm). After overnight adherence of SuSCs, the time points of 2 h, 4 h, and 3 days (with 6 h of continuous stretching each day) were established for the DO group. And three cell-stretching chambers for each group were placed into a cell stretching device (Cell Tank, Dongdi (Beijing) Technology Company,Beijing, China). The stretching parameters were set to 0.5 Hz sinusoidal wave, subjecting the cells to 8% uniaxial cyclic. The SC group was placed in the same cell-stretching chambers but without stretching manipulation. After stretching, the cells were harvest for subsequent experiments.

Flow Cytometry (FCM) analysis

When the SuSCs culture reaches over 80% confluence, cells are dissociated into single cells using 0.25% trypsin and resuspended in flow cytometry buffer (PBS containing 0.1% BSA). The cell density is then adjusted to 3 × 10⁶/mL. A 1.5 mL EP tube is labeled with the name of the primary antibody. To each tube, 100 µL of cell suspension (~3 × 10⁵ cells) is added, followed by 2 µL of the corresponding primary antibody. The mixture is gently mixed and incubated at 4 °C for 30 min. After incubation, the samples are washed twice with 200 µL of flow cytometry buffer. The cells are centrifuged at 250 g for 5 min, and the supernatant is discarded. Next, 100 µL of flow cytometry buffer and 2 µL of the corresponding fluorescent secondary antibody are added to each tube, and the cells are resuspended and incubated at 4 °C for 30 min. After the second incubation, the samples are washed twice with 200 µL of flow cytometry buffer, centrifuged at 250 g for 5 min, and the supernatant is discarded. Finally, the cells are resuspended in 300 µL of flow cytometry buffer and immediately analyzed by flow cytometry.

Data analysis

Data were analyzed using Statistical Product and Service Solutions software (SPSS, version R26.0.0.2 for Windows, IBM, Armonk, America) and GraphPad Prism (version 10.0.0 for Windows, GraphPad Software, Boston, America). Continuous variables are expressed as the mean ± standard deviation, and classified variables are expressed as frequencies and/or percentages. Paired sample t-tests were performed for data subject to a normal distribution, and Wilcoxon tests were performed for data not subject to a normal distribution. P < 0.05 was regarded as statistically significant. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Author contributions

Z. L. and Y. C. contributed equally to this article and should be regarded as co-first author. Z. L. and Z. Z. conceived and designed the study. Z. L., Y. C., P. Z., and M. J. were involved in data collection. Y. Z. performed the statistical analysis and generated the initial statistical charts. Z. L. and Y. C. were responsible for the creation of all statistical graphs and tables. Z. L. drafted the manuscript. All authors critically reviewed, edited, and approved the final version of the manuscript, including all figures and tables.

Data availability

The Data and Code that support the findings of this study are available from Peking University Third Hospital, but restrictions apply to the availability of these data, which were used under licence for the current study and so are not publicly available. The data are, however, available upon request and with the permission of Peking University Third Hospital.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41536-025-00448-5.