A modular self-assembling peptide platform targeting TGF-β1 for tendon regeneration

Chao Li; Zehao Chen; Wentao Li; Ronghui Deng; Lingan Huang; Yifan Song; Yiqun Wang; Guoqing Cui; Huawei Liu; Jiakuo Yu

doi:10.1038/s41467-026-76487-3

. 2026 Aug 11;17:9622. doi: 10.1038/s41467-026-76487-3

A modular self-assembling peptide platform targeting TGF-β1 for tendon regeneration

Chao Li ^1,^2,^#, Zehao Chen ^3,^#, Wentao Li ^1,², Ronghui Deng ¹, Lingan Huang ¹, Yifan Song ¹, Yiqun Wang ¹, Guoqing Cui ^2,^✉, Huawei Liu ^1,^✉, Jiakuo Yu ^1,^✉

PMCID: PMC13554100 PMID: 42711313

Abstract

Excessive transforming growth factor-β1 (TGF-β1) signaling is a key driver of the pathological processes of fibrosis and matrix disorganization in tendinopathy, yet current treatments do not directly target this fibrotic microenvironment. Here we show that a self-assembling peptide composed entirely of natural amino acids, AsPep-FTSQ, can locally capture TGF-β1 and remodel the pathological tendon niche. Injecting AsPep-FTSQ into rat and beagle models of tendinopathy resulted in the formation of an interconnected nanofibrous network in situ that selectively sequesters excess TGF-β1. This network suppresses pro-fibrotic signaling, limits pathological tendon cell (TC) state transitions, and promotes collagen realignment. These effects improve tendon structure and functional repair across species. This fully natural, self-assembling peptide composed of amino acids provides a modular strategy for local cytokine sequestration and suggests a therapeutic approach for tendinopathy and potentially other fibrotic soft-tissue disorders.

Subject terms: Biomaterials - proteins, Nanostructures, Regenerative medicine

Excessive TGF-β1 activity contributes to tendon degeneration and fibrosis. Here, authors develop a self-assembling peptide that locally sequesters TGF-β1, restores tendon structure and function in rat and beagle models, and offers a strategy for treating fibrotic disorders.

Introduction

Tendinopathy is a prevalent musculoskeletal disorder with a high rate of recurrence among physically active individuals, manual workers, and the elderly. A multitude of epidemiological studies have indicated that tendinopathy has the potential to profoundly impair activities of daily living and physical performance. Furthermore, it has been determined that tendinopathy is a major cause of chronic pain and loss of work capacity^1–3. Tendinopathy is characterized by a multifaceted array of symptoms, including pain, swelling, and functional impairment. It is also marked by disorganized collagen fiber alignment, pathological neovascularization, abnormally increased cell density, and dysregulated extracellular matrix (ECM) remodeling. These factors ultimately result in tendon fibrosis. Collectively, these alterations markedly impair the mechanical integrity of the tendon and substantially increase the risk of spontaneous rupture^4,5. The therapeutic options available for addressing tendinopathy at present, encompassing a range of pharmaceutical agents such as non-steroidal anti-inflammatory drugs (NSAIDs), corticosteroid injections, extracorporeal shockwave therapy, and surgical interventions, are chiefly palliative and symptom-oriented^6,7. The prevailing treatment strategies currently in use do not specifically address the imbalanced molecular microenvironment within diseased tendons. Furthermore, these strategies are unable to effectively reprogram the key signaling networks that drive fibrosis and aberrant cellular phenotypes. Consequently, there is an urgent need to develop therapeutic approaches that directly target pathological fibrosis in tendinopathy.

Transforming growth factor-β1 (TGF-β1) is one of the key pathological mediators implicated in the onset and progression of tendinopathy. It drives late-stage tendon fibrosis, ECM remodeling, and collagen disorganization^8,9. Aberrant TGF-β1 signaling, which acts through SMAD-dependent pathways, persistently drives tendon cells toward a myofibroblast-like phenotype¹⁰. Concurrently, TGF-β1 can shift tendon cells (TCs) toward osteogenic and chondrogenic lineages. This leads to heterotopic ossification, fibrocartilaginous metaplasia, and calcific deposit formation. These processes accelerate the transition of tendon pathology from a reversible injury to irreversible structural remodeling^11,12. TGF-β1-targeted agents have yet to enter the therapeutic landscape for tendinopathy. Conventional intravenous administration of TGF-β1 readily disrupts its fundamental roles in immune tolerance and tissue homeostasis, resulting in a narrow therapeutic window, pronounced off-target effects, and substantial safety concerns¹³. Unlike the knee joint cavity, however, the anatomical structure of tendons does not provide a confined space that allows for the long-term retention of sustained-release formulations¹⁴. Taken together, these considerations greatly reduce the clinical feasibility of tendinopathy treatment strategies based on TGF-β1 blockade.

Programmable, supramolecular, self-assembling peptides are a highly attractive platform for constructing interfacially friendly, nanomaterial systems with functions that can be designed rationally^15,16. Peptides have several inherent advantages, including low molecular weight, programmable sequences, and well-established synthetic chemistries. They can also spontaneously assemble into nanofibrous or nanoparticulate architectures under near-physiological conditions^17–19. However, the stability and interactions of many existing functional self-assembling peptide systems with target molecules are enhanced by the incorporation of substantial amounts of non-natural amino acids, D-amino acids, or bulky synthetic moieties^20,21. Such modifications can improve resistance to degradation or enhance binding affinity to some extent. However, they may also increase immunogenicity, disrupt in vivo metabolism and clearance, and introduce uncertainties in biosafety evaluation and regulatory review during clinical translation²². In contrast, L-form canonical amino acids maintain sequence programmability and self-assembly capacity, and offer great biocompatibility, biodegradability, and predictable metabolic fates. These characteristics better meet the safety and scalable manufacturing requirements necessary for clinical translation^23,24. Importantly, self-assembling peptide systems are “modular” by design. Through simple sequence substitution or recombination within a shared backbone scaffold, these modules can be quickly adapted to different cytokines, target tissues, or disease settings^25,26. This modular, self-assembly strategy, which is based on canonical amino acids, makes it possible to establish unified structure-function design principles across diverse pathogenic cytokines and tissue microenvironments. This provides a highly flexible engineering pathway for the stepwise development and clinical translation of these principles.

Based on this conceptual framework, we propose and implement a microenvironment-targeted strategy for tendinopathy. We design AsPep-FTSQ, a self-assembling, TGF-β1-sequestering peptide composed entirely of canonical amino acids (Fig. 1a). AsPep-FTSQ is intended to modulate the fibrotic tendon niche in situ. AsPep-FTSQ can spontaneously form a stable nanofibrous network under physiological conditions through the concerted action of a precisely embedded TGF-β1-binding motif and a self-assembly domain (Fig. 1b). This network selectively enriches free TGF-β1 with high affinity, thereby achieving on-site interception of the pathogenic signaling pool at the molecular level. In cross-species tendinopathy models involving rats and beagles, in situ self-assembly following local injection demonstrated that the AsPep-FTSQ network can stably intercalate into the tendon ECM. Concomitantly, it promotes realignment of collagen fibers along the principal loading direction (Fig. 1c). These processes enhance tendon regeneration at both the histological and functional levels. This study introduces a microenvironment-targeted therapeutic paradigm of “natural amino acid-based self-assembling peptide-mediated TGF-β1 capture” for tendinopathy. It also establishes a modular self-assembling peptide platform with generalizable design principles that can be extended to fibrotic soft-tissue disorders of the myocardium, lung, liver, and beyond.

Fig. 1 — a Schematic illustration of the therapeutic cascade. b Network of AsPep-FTSQ nanofibers interspersed between TCs and collagen bundles. c AsPep-FTSQ helps ordered regenerative repair of tendinopathy. Created in BioRender. Chao, L. (2026) https://BioRender.com/y9z8hip.

Results

Single-cell transcriptomic analysis reveals lineage remodeling of TCs in tendinopathy

To analyze TC heterogeneity and clarify the molecular mechanisms of tendinopathy, we performed single-cell RNA sequencing on a Sprague-Dawley (SD) rat model of the condition. We established a local surgical injury model to recapitulate the pathological state of tendinopathy. Four biological replicates were collected for both the normal tendon group (Nor-TC) and the tendinopathy group (SQ-TC), with each replicate consisting of ten Achilles tendons pooled from ten individual SD rats. Cell-type annotation revealed seven primary cell populations: smooth muscle cell (SMC), TC, immune cell (IC), endothelial cell (EC), neuronal cell (NC) and tendon stem cell (TSC). Ambiguous clusters are characterized by red blood cell markers and tendon cell markers. The SQ-TC group contained higher proportions of EC and TSC than the Nor-TC group (Supplementary Fig. 1a, b). Since TC and TSC are the main cell types involved in the pathological process of tendinopathy, we performed a more detailed analysis of these two subsets in both the SQ-TC and Nor-TC groups.

After annotating the cell types within the TC compartment, we resolved the TCs into nine transcriptionally distinct subclusters (Fig. 2a). The relative abundance of TC1, TC3, TC5, TC7, TC9, and TSC increased in tendinopathic tissue compared with normal tendons (Fig. 2b and Supplementary Fig. 1c). We then performed marker gene analysis on each TC subcluster to infer the biological functions of the respective TC subsets (Fig. 2c). TC1 was primarily associated with gene programs linked to proliferation, differentiation, apoptosis, and cell-cycle regulation. Therefore, it was designated a “protective TC subset”. TC2 was mainly involved in regulating collagen and ECM components, earning it the name “secretory TC subset.” TC3 was an “ECM-mineralizing TC subset” associated with ECM remodeling, immune modulation, and bone remodeling. TC4 was defined as a “bridging TC subset” linked to tissue repair and fibrosis regulation. TC5 was classified as a “stress-responsive TC subset” enriched for pathways related to cytoskeletal organization and cell adhesion. TC6 was characterized as a “migratory TC subset” associated with cell migration and signal transduction. TC7 was defined as an “osteo-/chondrogenic TC subset” related to bone and cartilage formation. TC8 was a “signaling/immune-regulatory TC subset” implicated in tissue repair and immune modulation. TC9 was identified as an “ECM-modulatory TC subset” associated with connective tissue remodeling and TGF-β signaling regulation.

To identify the key genes governing cell fate transitions during the progression from normal to pathological TCs, we arranged TCs and TSCs along a single pseudotime trajectory using Monocle 2, thereby reconstructing their developmental differentiation relationships. Pseudotime analysis revealed that cells in both the SQ-TC and Nor-TC groups originated from the left end of the trajectory. They passed through distinct intermediate differentiation states and eventually converged on two terminal states. The color scale indicates pseudotime values, as shown in the legend; higher values correspond to later developmental stages (Supplementary Fig. 1d).

Subcluster analysis was performed with the TSC-enriched state assigned as the root state for Monocle 2 pseudotime ordering, based on canonical TSC marker expression and tendon lineage biology. This analysis suggested a transcriptional continuum from TSC state toward multiple TC states in both the Nor-TC and SQ-TC groups; however, the subsequent trajectory-associated cell-state distributions diverged. In the Nor-TC group, cells primarily ended up as TC4 and TC6. Consistent with the functional annotations described above, this suggests that, under physiological conditions, TCs primarily maintain equilibrium among subsets by preserving ECM homeostasis. In contrast, the SQ-TC group predominantly converged on TC7 and TC5. This suggests that, during tendinopathy development, TCs differentiate into osteo-/chondrogenic or fibroblastic/stress-responsive TCs (Fig. 2d). Consistently, these two distinct cell fate branches in the SQ-TC group correspond to calcification and fibrosis, respectively, in tendinopathic tendons.

Analysis of differentiation states revealed seven distinct developmental stages along the tendinopathy pathological trajectory. The seven colors shown in the legends of both the SQ-TC and Nor-TC groups correspond to these distinct differentiation states (Fig. 2e and Supplementary Fig. 1e). Monocle 2 uses gene expression signals and pseudotime ordering to evaluate the expression of genes over time along an inferred trajectory. This provides a framework for examining candidate genes associated with disease-related cell state transitions. Because TGF-β1 is well-established to play a role in tendon fibrosis and extracellular matrix remodeling, we examined the dynamic expression of the Tgfb1 gene, which encodes TGF-β1, along pseudotime (Fig. 2f). The results showed that TGF-β1 exhibited two distinct expression peaks along pseudotime at stages 2 and 6. Overlaid with the differentiation trajectory, stages 2 and 6 in the SQ-TC group represented two terminal developmental states that converged on the TC7 subcluster in the subset trajectory map. These results suggest that peak Tgfb1 expression coincides with the temporal window during which the TC7 subset emerges and persists. Consistent with these observations, bubble plots of differentially expressed genes revealed that, in the SQ-TC group, Tgfb1 had the highest proportion of expressing cells and the highest mean expression level within the TC7 subcluster. In contrast, in the Nor-TC group, Tgfb1 expression peaked in the TC4 subcluster (Fig. 2g). Gene Ontology (GO) enrichment analysis revealed that the dominant TC4 subcluster in the Nor-TC group was enriched in cell-matrix adhesion and integrin activation pathways that support tendon homeostasis. In contrast, the dominant TC7 subcluster in the SQ-TC group was enriched in cartilage/bone development, ECM organization, and skeletal system morphogenesis pathways (Fig. 2h). Thus, rather than serving as an unbiased, de novo approach to discovering targets, this analysis was used to contextualize the relevance of the candidate Tgfb1/TGF-β1 axis to the disease and to support prioritizing it for subsequent functional intervention.

Key phenotypic alterations in tendinopathy

To further substantiate the characteristic pathological alterations of tendinopathy, we refined and expanded our histopathological assessments. Hematoxylin and eosin (H&E) staining revealed significant cellular hyperplasia and disruption of the bundled ECM architecture in SQ-TC tendons. In contrast, Nor-TC tendons exhibited regular, fascicular collagen alignment. Bonar scores were significantly higher in the SQ-TC group than in the Nor-TC group (9.00 ± 0.93 vs. 0.75 ± 0.71), confirming the presence of typical tendinopathic histopathology (Fig. 2i). Masson’s trichrome staining revealed uneven collagen staining and disorganized fiber alignment in SQ-TC tendons. There was a significant increase in fiber angle compared to Nor-TC tendons (4.75 ± 1.49 vs. 38.25 ± 8.94). This indicates a loss of alignment along the principal loading axis (Fig. 2j).

Immunohistochemistry for TGF-β1 revealed a significantly higher proportion of TGF-β1-positive cells in the SQ-TC group than in the Nor-TC group (10.56 ± 5.51 vs. 34.54 ± 7.53), which is consistent with elevated TGF-β1 expression at the single-cell level (Fig. 2k). Type I collagen (Col I) immunofluorescence revealed an increased Col I signal intensity in tendinopathic tendons, suggesting excessive local fibrotic matrix deposition (Fig. 2l). The Col I signal in Fig. 2l represents the normalized immunofluorescence intensity within the selected region of interest. The Nor-TC group is set as the reference value of 1. Thus, the increased Col I signal in the SQ-TC group does not mean that normal tendon contains little Col I. Rather, it reflects a higher local Col I signal compared to the normal baseline. Additionally, microcomputed tomography (micro-CT) analysis revealed bone-like, high-density lesions at the Achilles tendon insertion in the SQ-TC group. There was a significant increase in bone volume (0.00 ± 0.00 vs. 16.75 ± 6.01), suggesting a propensity for ectopic ossification during tendinopathy (Fig. 2m). Similarly, an in vitro cell migration assay revealed that tendon cells exhibited increased migratory capacity after 24 hours (8.90 ± 3.61 vs. 48.96 ± 9.97), suggesting the potential for enhanced collagen matrix remodeling within the pathological microenvironment (Supplementary Fig. 1f).

Design and self-assembly characterization of AsPep-FTSQ

After identifying TGF-β1 as a key pathological driver, we sought to design a structurally stable and translationally tractable local peptide system capable of sequestering this cytokine within the tendon niche. Caveolin-1 is a membrane protein that suppresses TGF-β1-mediated SMAD phosphorylation^27,28. The caveolin scaffolding domain (CSD) is a functional segment within the caveolin-1 sequence that serves as an active structural motif capable of binding TGF-β1²⁹. We selected FTTFTVT, the minimal sequence within this domain that retains TGF-β1-inhibitory activity, as the TGF-β1-capturing motif of AsPep-FTSQ³⁰. AsPep-FTSQ is composed of four structural units: a Flag unit (DYKDDDDK), a spacer unit (GSG/SGSG), a functional unit (FTTFTVT), and an assembly unit (QQKFQFQFEQQ) (Fig. 3a and Supplementary Fig. 2a). All four structural units are composed of natural amino acids with evenly distributed hydrophilic and hydrophobic residues, which give the peptide an overall hydrophilic character (Fig. 3b).

Fig. 3 — a The amino acid sequence and chemical structure formula of AsPep-FTSQ. Flag Unit, DYKDDDDK. Function Unit, FTTFTVT. Assembly Unit, QQKFQFQFEQQ. b Distribution of hydrophilic and hydrophobic amino acids in the AsPep-FTSQ sequence. c (i) TEM image of 75 μM AsPep-FTSQ network in pure water and (ii) localized enlarged view. d TIRF imaging reveals that AsPep-FTSQ assembles on and around the surface of the TC membrane. White arrow, self-assembled aggregate. e SEM observation of the in situ self-assembled network on the TC surface following co-incubation with AsPep-FTSQ. Yellow outline, TC. White arrow, AsPep-FTSQ network. f The CAC of AsPep-FTSQ. (n = 3 per group). g The TEM image of AsPep-FTSQ was taken at a concentration of 28.7 μM. (n = 3 per group). h The TEM image of AsPep-FTSQ was taken at a concentration of 15 μM. i 3D confocal reconstruction images of self-assembled aggregates around TCs, taken 48 h after AsPep-FTSQ injection. j Quantitative analysis of the diameter, volume, and surface area of self-assembled aggregates. (n = 4 per group). k Molecular docking model of AsPep-FTSQ with TGF-β1, with a right-side enlarged view showing hydrogen bonds and electrostatic interactions between key amino acid residues. l Results from 100 ns molecular dynamics simulations, including (i) RMSD, (ii) RMSF, (iii) Rg and (iv) SASA. m Bond energy decomposition analysis of MMGBSA. Evdw, van der Waals energy. Eele, electrostatic energy. EGB, polar contribution to solvation. ESA, non-polar contribution to solvation. (n = 3 per group). n SPR sensor response curve and fitting results. o The results of the competitive ELISA assay for TGF-β1. (n = 3 per group). Data in f, j, m and o are presented as mean ± s.d.

Due to the presence of the self-assembly domain, AsPep-FTSQ spontaneously assembles under appropriate conditions. In vitro, at a concentration of 75 μM, AsPep-FTSQ forms linear nanostructures with a diameter of approximately 15 nm (Supplementary Fig. 2b). AsPep-FTSQ filaments intertwine to form a network-like architecture (Fig. 3c and Supplementary Fig. 2c). To elucidate the assembly mechanism further, we used AlphaFold to predict the self-assembled structure of AsPep-FTSQ. This analysis indicated that the extended polypeptide chain adopts β-sheet-rich conformations at the monomer level (Supplementary Fig. 2d). Circular dichroism (CD) spectroscopy confirmed that AsPep-FTSQ exhibits β-sheet secondary structures (Supplementary Fig. 2e).

To more closely mimic the assembly behavior of the peptide in human biofluid environments, we replaced the pure water with a cell culture medium. AsPep-FTSQ no longer formed extended linear filaments under these simulated physiological conditions, but it still assembled into a network-like architecture (Supplementary Fig. 2f). Next, we visualized the DYKDDDDK-containing AsPep-FTSQ construct using anti-FLAG antibodies conjugated to fluorophores that were excited at two different wavelengths. Using total internal reflection fluorescence (TIRF) microscopy, we observed the colocalization of the two fluorescent signals on the surface of TCs and in the pericellular space (Fig. 3d). No fluorescent signal was detected around TCs in the sham group that did not receive fluorescent antibody labeling (Supplementary Fig. 3a). Since two-dimensional (2D) cultures do not fully replicate the structural complexity of human three-dimensional (3D) tissues, we mimicked the 3D tendon environment in vitro by embedding TCs within a collagen matrix (Supplementary Fig. 3b). In addition to labeling AsPep-FTSQ as described above, we introduced a fluorescently tagged antibody against TGF-β1. Confocal fluorescence microscopy revealed extensive colocalization of all three fluorescent signals. This indicates that AsPep-FTSQ underwent de novo structural rearrangements in vitro to form higher-order assemblies composed of at least two peptide monomers. The assembled AsPep-FTSQ successfully engaged TGF-β1. Furthermore, the AsPep-FTSQ network expanded progressively over time, which is consistent with an increasing capacity to capture TGF-β1 (Supplementary Fig. 3c). Next, we examined TCs co-incubated with AsPep-FTSQ by scanning electron microscopy (SEM). As outlined by the yellow dashed line in Fig. 3e, TCs were clearly surrounded by a prominent AsPep-FTSQ network on their surface and in the pericellular region. This provides a physical basis for subsequent in situ sequestration of TGF-β1 within tendon lesions.

We used a Thioflavin T (ThT) fluorescence assay to quantitatively define the assembly threshold of AsPep-FTSQ and determine its critical assembly concentration (CAC). As the peptide concentration increased, the ThT fluorescence intensity exhibited a clear inflection at approximately 28.7 μM (Fig. 3f). This concentration was defined as the CAC of AsPep-FTSQ in aqueous solution. Consistently, transmission electron microscopy (TEM) imaging at concentrations near the CAC revealed small, network-like structures, as well as partially aggregated species (Fig. 3g). In contrast, at 15 μM, below the CAC, only short, discrete fibrils and incomplete aggregates were observed (Fig. 3h). Through 3D confocal reconstruction and morphometric analysis, we found that AsPep-FTSQ assemblies surrounding TCs gradually thickened and increased in volume over time (Fig. 3i and Supplementary Fig. 3d). Accordingly, the mean diameter, volume, and surface area of these aggregates increased over the 12-48 h period (Fig. 3j).

Specific binding of AsPep-FTSQ to TGF-β1 and its binding kinetics

At the molecular level, we used molecular docking and molecular dynamics simulations to determine the interaction pattern between AsPep-FTSQ and TGF-β1. The docking results revealed that AsPep-FTSQ inserts into a specific groove on the surface of TGF-β1. Multiple residues within the functional motif of TGF-β1 (ARG-278, LYS-56, CYS-355, CYS-33, THR-32, THR-282, LYS-44, and THR-35) form hydrogen bonds and electrostatic interactions with the positively charged or polar residues of AsPep-FTSQ (Fig. 3k). We used the natural receptor TGF-β1R as a reference control and found that TGF-β1 engages TGF-β1R via TYR-104, ASP-246, GLU-75, ASN-117, ALA-60, LYS-44, and GLU-100 (Supplementary Fig. 4a). We performed docking analyses of AsPep-FTSQ, TGF-β1, and TGF-β1R. Interestingly, LYS-44, a critical residue in the interaction between TGF-β1 and TGF-β1R, served as an additional binding site for AsPep-FTSQ. This finding suggests that AsPep-FTSQ interacts with TGF-β1 within the same binding domain as its native receptor, thereby exerting a competitive blocking effect on TGF-β1 (Supplementary Fig. 4b).

Root-mean-square deviation (RMSD) analysis was used to monitor the conformational evolution of the complexes throughout the simulation. This analysis provides insight into structural stability and molecular flexibility. If the protein RMSD continues to increase or decrease toward the end of the simulation, it indicates that the system has not yet equilibrated. The RMSD profiles show that the mean RMSD values of all complexes remained below 0.8 nm. The AsPep-FTSQ-TGF-β1 complex reached dynamic equilibrium at approximately 20 ns, indicating that AsPep-FTSQ forms a structurally stable complex with TGF-β1 after dynamic relaxation (Fig. 3li).

Root-mean-square fluctuation (RMSF) was then used to quantify conformational flexibility of individual residues along the protein chain during the simulation. Peak values indicate regions with the greatest flexibility. Higher RMSF values reflect greater conformational mobility and flexibility of the corresponding residues. The RMSF profiles showed that most residues within the AsPep-FTSQ-TGF-β1 complex fluctuated within an acceptable range (Fig. 3lii). Additionally, AsPep-FTSQ alone displayed relatively low residue-level fluctuations, consistent with its intrinsic structural stability (Supplementary Fig. 4c).

We used the radius of gyration (Rg) to evaluate the compactness of the protein structures. Lower Rg values indicate a more compact conformation, while higher Rg values reflect increased conformational entropy and structural disorder. As shown in the Rg profiles, both AsPep-FTSQ and TGF-β1 exhibited a pronounced decrease in Rg upon complex formation. This suggests that their interaction promotes additional hydrophobic contacts and intramolecular packing within TGF-β1 while improving structural complementarity between the two proteins. Together, these effects are consistent with an enhancement of the complex’s overall stability (Fig. 3liii).

Solvent-accessible surface area (SASA) represents the area of a biomolecule accessible to solvent. As shown by the SASA profiles, the SASA of the AsPep-FTSQ-TGF-β1 complex decreased markedly upon binding. This indicates that complex formation does not compromise the intrinsic structural stability of TGF-β1. It also indirectly supports favorable peptide-protein interactions (Fig. 3liv). To better understand the interaction between AsPep-FTSQ and TGF-β1, we quantified the number of hydrogen bonds between the two molecules over the course of the simulation. Analysis of the hydrogen-bonding network revealed that AsPep-FTSQ consistently formed six or more hydrogen bonds with pocket-forming residues of TGF-β1. This finding further corroborates the robustness of the complex interface (Supplementary Fig. 4d).

A negative binding free energy (ΔG binding) indicates a thermodynamically stable complex, while a positive value suggests instability. The calculated ΔG binding for the AsPep-FTSQ-TGF-β1 complex was −162.4 ± 6.52 kcal/mol. Decomposition of the binding energy revealed that van der Waals interactions (E_vdw) (−141.3 ± 3.50 kcal/mol) were the predominant stabilizing force, followed by electrostatic interactions (E_ele) (−75.87 ± 4.15 kcal/mol). However, the polar solvation term (E_GB) (75.37 ± 3.60 kcal/mol) partially offset these favorable interactions. Together, these data suggest a strong affinity between AsPep-FTSQ and TGF-β1, which supports the formation of a stable peptide-protein complex underlying the biological activity of AsPep-FTSQ (Fig. 3m).

Surface plasmon resonance (SPR) analysis confirmed a substantial affinity between AsPep-FTSQ and TGF-β1. As the concentration of AsPep-FTSQ increased, the SPR sensorgrams displayed a concentration-dependent response. Global fitting yielded an equilibrium dissociation constant (Kd) of approximately 24.2 μM (Fig. 3n), which is similar to its CAC. This indicates that AsPep-FTSQ can efficiently bind TGF-β1 within the concentration range at which it self-assembles into a network. A competitive enzyme-linked immunosorbent assay (ELISA) showed that AsPep-FTSQ dose-dependently suppressed the detection signal of soluble TGF-β1, with a calculated IC₅₀ of approximately 37.5 μM (Fig. 3o). This result corroborates the capacity of AsPep-FTSQ to capture free TGF-β1.

TGF-β1 combined with cyclic mechanical stretch induces a tendinopathy-like fibrotic phenotype in TCs in vitro

Since tendinopathy arises from a combination of biochemical and mechanical cues, we created an in vitro model that incorporates both factors to test peptide function. To induce a tendinopathy-like fibrotic phenotype in TCs and establish an in vitro disease model, we stimulated TCs with exogenous TGF-β1. Initial testing of three different concentrations revealed that 10 ng/mL elicited the highest expression of the fibrotic marker alpha smooth muscle actin (α-SMA). This dose was selected for all subsequent experiments (Supplementary Fig. 5a). Phase-contrast microscopy revealed that confluent control TCs exhibited a typical cobblestone morphology. In contrast, TGF-β1-treated TCs became elongated and adopted an overall spindle-shaped appearance at high density (Supplementary Fig. 5b).

To more accurately recreate the pathological microenvironment of tendinopathy in vitro, we combined TGF-β1 stimulation with cyclic mechanical stretch. We also established a tendon cell stretching model using BioFlex flexible-bottom culture plates (Fig. 4a). This system effectively mimics the mechanical loading conditions experienced by tendons in vivo. Phase-contrast microscopy revealed that, consistently, mechanical stretch alone induced a more ordered alignment of TCs, whereas the presence of additional TGF-β1 caused spindle-shaped TCs to orient their long axis parallel to the direction of stretch (Supplementary Fig. 5c).

Fig. 4 — a Schematic of a periodic stretching system constructed using BioFlex flexible bottom culture plates and vacuum-driven loading columns, illustrating the deformation direction of TCs under resting (Rest) and stressed (Strain) conditions. b (i) Col I immunofluorescence staining reveals differences in TCs alignment under resting versus mechanically stretched conditions. (ii) The panel shows the distribution of fiber orientation angles. c (i) Western blot analysis of α-SMA protein expression following treatment with increasing concentrations of AsPep-FTSQ under constant TGF-β1 stimulation (10 ng/mL), and (ii) grayscale quantification. (n = 5 per group). d Western blot analysis of α-SMA, Col I, and FN 1 in different treatment groups and e grayscale quantification. (n = 4 per group). f GO enrichment analysis of TC7 fibrotic TC subpopulations in single-cell sequencing. g (i) Immunofluorescence staining of Col I and Col III under different stimulation conditions and (ii) quantitative analysis. (n = 4 per group). h (i) Immunofluorescence staining of FN 1 and SCX under different stimulation conditions and (ii) quantitative analysis. (n = 4 per group). i (i) Immunofluorescence staining of α-SMA and Postn under different stimulation conditions and (ii) quantitative analysis. (n = 4 per group). j qPCR analysis of expression heatmaps for tenogenic differentiation and fibrosis-related genes under different stimulation conditions. (n = 4 per group). Data in c, e, g, h and i are presented as mean ± s.d., and were compared using one-way ANOVA with post hoc Dunnett′s test. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Created in BioRender. Chao, L. (2026) https://BioRender.com/fnsxtb3.

This arrangement closely resembles the anatomical organization of TCs within the Achilles tendon microenvironment in vivo. Immunofluorescence staining corroborated the TC alignment pattern. Col I is a structural component associated with the TC cytoskeleton that outlines the cells’ overall morphology. When cyclic stretch was applied, the cells adopted a markedly parallel orientation. The Col I fibers were clearly aligned along the loading axis with minimal angular deviation. In contrast, the cells in the non-stretched group exhibited a disordered arrangement with substantially greater angular dispersion (Fig. 4b).

AsPep-FTSQ attenuates fibrotic activation and promotes tenogenic differentiation of TCs in vitro

Building on this model, we next evaluated the modulatory effects of AsPep-FTSQ on TC fibrotic activation. First, we examined the cytocompatibility of AsPep-FTSQ in vitro. Over an eight-day observation period, there were no statistically significant differences in cytocompatibility among the three peptide-treated groups compared with the sham group (Supplementary Fig. 5d). Then, we treated TCs with increasing concentrations of AsPep-FTSQ at a fixed TGF-β1 level. Western blot analysis revealed a gradual decrease in α-SMA protein expression as the AsPep-FTSQ concentration increased. Densitometric quantification demonstrated a significant, dose-dependent inhibitory effect of AsPep-FTSQ on the TGF-β1-induced fibrotic phenotype (Fig. 4c).

Next, we introduced three peptide variants into the established TGF-β1 plus stretch model: AsPep-FTSQ, which contains both domains; AsPep-FT, which lacks the self-assembly domain; and AsPep-SQ, which lacks the TGF-β1-binding motif. In this in vitro model, the sham group served as the baseline control group. The AsPep-SQ group served as the fibrotic model control group. The AsPep-FT and AsPep-FTSQ groups served as the treatment groups. AsPep-FTSQ markedly reduced the levels of the proteins α-SMA, Col I, and fibronectin (FN 1), while the two control peptides only modestly affected these readouts (Fig. 4d). Compared with the fibrotic model control group (AsPep-SQ), the AsPep-FTSQ treatment group showed lower levels of α-SMA, Col I, and FN 1, with α-SMA decreasing from 8.70 ± 0.27 to 2.60 ± 0.29, Col I from 1.90 ± 0.26 to 1.50 ± 0.21, and FN 1 from 2.40 ± 0.33 to 1.40 ± 0.08 (Fig. 4e). We subsequently performed GO functional annotation of the TC7 subset, which is enriched in pathological tendons. TC7 was significantly enriched for pathways related to the ECM and collagen remodeling (Fig. 4f), which closely matched the upregulation of ECM remodeling- and fibrosis-associated genes observed in our in vitro model.

Under combined TGF-β1 stimulation and cyclic mechanical stretch, TCs displayed a typical fibrotic ECM remodeling profile. Compared with the baseline control group (sham), the fibrotic model control group (AsPep-SQ) exhibited an increase in total Col I (1.00 ± 0.06 vs. 3.24 ± 0.20) and a marked increase in collagen III (Col III) signal (1.00 ± 0.51 vs. 110.40 ± 6.02). This imbalance in the Col I/Col III ratio (Fig. 4g) is consistent with the over-deposition of Col III observed in tendinopathic tissue. Relative to the sham group, the AsPep-SQ group showed a marked reduction in scleraxis (SCX) expression (11.07 ± 0.52 vs. 1.00 ± 0.12) and a substantial increase in FN 1 levels (1.00 ± 0.06 vs. 5.36 ± 0.20) (Fig. 4h). This indicates a shift from a homeostatic tenogenic phenotype to a fibrotic phenotype.

Further immunofluorescence staining revealed that, with combined TGF-β1 stimulation and mechanical stretch, the number of α-SMA-positive stress fibers in TCs increased significantly (16.09 ± 1.03 vs. 1.00 ± 0.03), as did periostin (Postn) deposition within the ECM (24.26 ± 1.56 vs. 1.00 ± 0.14) (Fig. 4i). These results indicate that TCs acquired a myofibroblast-like activated phenotype. Quantitative polymerase chain reaction (qPCR) analysis corroborated these changes at the transcriptional level. In the AsPep-SQ group, genes related to fibrosis and ECM remodeling, including Col1a1, Col3a1, Fn1, alpha smooth muscle actin (Acta2), and Tenascin-C (Tnc), were markedly upregulated, while tendon homeostatic markers, such as Scx, Tnmd, and fibromodulin (Fmod), were downregulated. Conversely, AsPep-FTSQ treatment decreased fibrotic marker expression while increasing tenogenic marker expression to varying degrees (Fig. 4j). Finally, ELISA revealed that AsPep-FTSQ significantly suppressed soluble Col I secretion, reducing its level by approximately 3.9-fold (3,447.0 ± 329.9 vs. 893.1 ± 225.8) (Supplementary Fig. 5e).

Imaging and histological assessment of AsPep-FTSQ repair in an SD rat Achilles tendinopathy model

To evaluate the capacity of AsPep-FTSQ to remodel the tendinopathic microenvironment and resolve tendon fibrosis in vivo, we surgically created an Achilles tendinopathy model in SD rats. Then, we administered five local postoperative injections of AsPep-FTSQ, AsPep-FT, or AsPep-SQ (Fig. 5a). The Achilles tendon surgery involved localized disruption of tendon fibers using surgical instruments, followed by suture stabilization and closure of the skin incision (Supplementary Fig. 6a). Eight weeks postoperatively, the rats were euthanized, and the Achilles tendon region was re-exposed. In the sham group, the tendons appeared slender, smooth, and glistening white. In contrast, the tendons of the AsPep-SQ group were markedly thickened and showed abundant neovessels on their surfaces. They were dull in appearance and exhibited adhesions to the surrounding tissues. Tendons in the AsPep-FT group showed partial recovery but remained darkened and exhibited evident proliferative changes. Notably, the tendons in the AsPep-FTSQ group more closely resembled those in the sham group. They had relatively smooth surfaces, and their color and morphology approached normal (Fig. 5b).

Fig. 5 — a Experimental flowchart. Following establishment of the SD rat Achilles tendon disease model, different peptide formulations (AsPep-FTSQ, AsPep-FT, AsPep-SQ) were administered via local injection at specific postoperative time points. During follow-up, gait analysis, MRI, micro-CT, and ultrasound examinations were conducted. At the endpoint, animals were euthanized for tissue collection and subjected to molecular and histological analysis. b Macroscopic morphology of the Achilles tendon at 8 weeks in each group. Black arrow, small vessels. c (i) MRI T2-weighted image and (ii) Achilles tendon thickness quantification. (n = 4 per group). d (i) MRI heatmap pseudocolor display of local signal intensity in the Achilles tendon, (ii) quantitative analysis of average signal intensity. (n = 4 per group). e (i) Ultrasound longitudinal section image, (ii) modified Ohberg score and Achilles tendon thickness quantification. M, muscle. AT, Achilles tendon. (n = 4 per group). f H&E staining reveals the overall morphology and local tissue structure of the Achilles tendons of each group after eight weeks. Black arrows: tendon cells. Red arrows: blood vessels. Green arrow: disorganized collagen. g Histological quantitative analysis: total Bonar score. (n = 4 per group). h (i) Immunofluorescence staining of α-SMA and Col I in the Achilles tendon at 8 weeks. Quantitative analysis of immunofluorescence: (ii) Col I fluorescence intensity and (iii) α-SMA fluorescence intensity. (n = 4 per group). i Heatmap of qPCR results for Achilles tendon tissue at the endpoint, showing relative expression changes of tendon and fibrosis-related genes such as *Col1a1*, *Scx*, *Tnmd*, and *Acta2* across different treatment groups. (n = 3 per group). Data in c, d, e, g and h are presented as mean ± s.d., data in c, d, e, g and h were compared using one-way ANOVA with post hoc Dunnett′s test, data in i were compared using two-way ANOVA with post hoc Tukey′s test. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Created in BioRender. Chao, L. (2026) https://BioRender.com/fnsxtb3.

Eight weeks after surgery, the rats in each group underwent magnetic resonance imaging (MRI) (Supplementary Fig. 6b). On T2-weighted imaging, the Achilles tendons of the sham group appeared thin with homogeneous low signal intensity. The tendon in the AsPep-SQ group displayed continuous bands of high signal intensity and marked thickening. In contrast, the tendon in the AsPep-FTSQ group showed substantially reduced thickness and signal intensity, resulting in a 1.5-fold decrease in thickness compared to the AsPep-SQ group (2.13 ± 0.13 vs. 1.38 ± 0.05). In the AsPep-FT group, the tendon caliber was irregular with focal adhesions to the surrounding tissues at its narrowest portion. Although the degree of thickening was partially alleviated, it remained greater than in the sham and AsPep-FTSQ groups (Fig. 5c). To more intuitively visualize the spatial distribution and heterogeneity of lesion-related T2 signal changes, we created pseudocolor heat maps using the original T2-weighted MRI images. This post-processing step enhanced visualization of focal high-signal regions within the Achilles tendon and enabled cross-group comparison of lesion extent while remaining consistent with conventional gray-scale MRI findings. AsPep-FTSQ reduced Achilles tendon signal intensity by approximately 2.7-fold compared with the AsPep-SQ group (660.5 ± 217.1 vs. 242.6 ± 5.38) (Fig. 5d).

We then used ultrasonography to evaluate the Achilles tendon repair in each group. The musculature was located above the tendon, with the calcaneus located below it (Supplementary Fig. 6c). In the AsPep-SQ group, the tendon margins appeared irregular, exhibiting heterogeneous echotexture, focal hyperechoic signals, and an overall thickened sonographic profile. In contrast, the AsPep-FTSQ group exhibited parallel, homogeneous fibrillar echoes, a normal-appearing fiber structure, and sharply demarcated interfaces. The Ohberg score was significantly lower in the AsPep-FTSQ group than in the AsPep-SQ group (0.50 ± 0.58 vs. 3.00 ± 0.00) and was not significantly different from that of the sham group (0.00 ± 0.00 vs. 0.50 ± 0.58). Similarly, tendon thickness decreased significantly in the AsPep-FTSQ group compared with the AsPep-SQ group (0.66 ± 0.14 vs. 3.15 ± 0.32) and was comparable to that of the sham group (0.55 ± 0.08 vs. 0.66 ± 0.14) (Fig. 5e).

At the histological level, H&E staining revealed increased cell density and wavy, crisscrossed, disorganized collagen fibers in the Achilles tendons of the AsPep-SQ group. These hallmark features of tendinopathy were markedly alleviated following AsPep-FTSQ treatment (Fig. 5f). Quantitative Bonar scoring confirmed that the total score was substantially lower in the AsPep-FTSQ group than in the AsPep-SQ group (9.00 ± 0.82 vs. 2.00 ± 0.82) and was similar to that of the sham group (1.00 ± 0.82 vs. 2.00 ± 0.82) (Fig. 5g). Immunofluorescence analysis revealed an increase in α-SMA-positive cells and fibers, as well as an elevated Col I signal in the AsPep-SQ group. In contrast, AsPep-FTSQ treatment reduced α-SMA (182.60 ± 21.77 vs. 1.10 ± 0.77) and Col I (17.52 ± 1.50 vs. 1.16 ± 0.22) expression (Fig. 5h).

Consistently, qPCR analysis at the transcriptional level corroborated these findings. Compared to the AsPep-SQ group, AsPep-FTSQ treatment increased the expression of tendon-related genes, such as Scx (0.20 ± 0.04 vs. 2.25 ± 0.33) and Tnmd (0.24 ± 0.04 vs. 1.65 ± 0.33), while decreasing the expression of Acta2 (13.29 ± 0.87 vs. 1.39 ± 0.30) and Col1a1 (10.30 ± 0.44 vs. 1.23 ± 0.32) (Fig. 5i).

AsPep-FTSQ improves gait and functional recovery in a rat model of tendinopathy

We used a fluorescent paw print-based gait analysis system to evaluate the impact of AsPep-FTSQ on Achilles tendon repair (Fig. 6a). The gait cycle in the sham group followed a regular sequence, initiating with the left forepaw and continuing with the right hind paw, then the right forepaw, and finally the left hind paw. In contrast, the AsPep-SQ and AsPep-FT groups exhibited disordered gait patterns with frequent and irregular switching between paws. However, the AsPep-FTSQ group largely restored a near-normal stepping sequence (Fig. 6b).

Next, we focused on the right hind paw of the operated limb to evaluate plantar pressure. 3D pressure maps revealed that the right hind paw of the sham group had a broad contact area with the substrate and homogeneous toe pressure distribution. In contrast, the AsPep-SQ group exhibited uneven pressure distribution concentrated in the mid-paw region (Fig. 6c). To further compare hindlimb loading patterns, we analyzed representative pawprint traces. In the sham and AsPep-FTSQ groups, the ipsilateral fore- and hind paw prints overlapped, indicating longer stride lengths and smooth progression. In the AsPep-SQ group, however, the ipsilateral prints were scattered, reflecting insufficient hindlimb loading and shortened strides (Fig. 6d). Quantitative analysis at week 8 showed that AsPep-FTSQ treatment significantly increased injured-side pawprint area (Fig. 6e) and mean pawprint fluorescence intensity (Fig. 6f), indicating improved weight-bearing capacity. In contrast, swing phase duration did not differ among the groups (Fig. 6g). Meanwhile, AsPep-FTSQ treatment significantly increased stride length (Fig. 6h), supporting an overall improvement in gait pattern.

The Achilles functional index (AFI) is a gait-derived score that evaluates Achilles tendon function in rats. It is based on footprint parameters from the injured and contralateral limbs. An AFI closer to 0 indicates better, more normal tendon function³¹. More negative values reflect greater functional impairment. Using the ink-based footprint method, we measured several key parameters—paw length (PL), toe spread (TS), and intermediary toe spread (IT)—and calculated the AFI for each group according to the established formula³²(Fig. 6i). We determined AFI values at three postoperative time points. The results showed that the AFI remained persistently within a markedly negative range throughout follow-up in the AsPep-SQ and AsPep-FT groups (Fig. 6j). Conversely, the AFI in the AsPep-FTSQ group increased progressively over time. By week eight, the AFI in the AsPep-FTSQ group was significantly higher than in the AsPep-SQ group (−7.10 ± 18.64 vs. −70.72 ± 8.78). This value approached that observed in the sham group (−3.22 ± 13.69 vs. −7.10 ± 18.64) (Supplementary Fig. 6d).

AsPep-FTSQ restores collagen fiber alignment and attenuates heterotopic ossification

Using SEM, we examined the alignment of collagen fibers on the surface of the Achilles tendon and quantified it using orientation analysis. In the sham group, the collagen fibers were highly aligned in parallel along the bone-muscle axis. In contrast, in the AsPep-SQ and AsPep-FT groups, the fibers appeared wavy and intersecting. In some regions, they formed tangled, network-like structures. In contrast, the AsPep-FTSQ group exhibited a collagen fiber orientation pattern that was largely restored to that of the sham group, with only focal adhesions remaining (Fig. 6k). As in our previous approach, deviations in fiber angle were visualized as shifts in pseudocolor, enabling a quantitative assessment of orientation dispersion³³. Orientation pseudocolor maps and angular distribution curves revealed a broad, flattened distribution of fiber orientations within the AsPep-SQ group, suggesting significant dispersion of collagen alignment. In contrast, the AsPep-FTSQ group exhibited a narrow, nearly unimodal distribution with a tightly clustered principal peak around 0°, demonstrating re-established orderly collagen fiber alignment along the primary loading axis (Fig. 6l).

Heterotopic ossification is an important late complication of refractory tendinopathy. Therefore, we evaluated heterotopic mineralization in each group using micro-CT³⁴(Supplementary Fig. 6e). 3D micro-CT reconstructions revealed prominent, high-density, bone-like lesions around the Achilles tendon insertion in the AsPep-SQ and AsPep-FT groups. Focal regions of high mineral density were evident on axial views and corresponding pseudocolor maps (Fig. 6m). In contrast, no obvious newly formed bone was detected in the AsPep-FTSQ group, whose mineral density and 3D structural distribution more closely resembled those of the sham group. Quantitatively, bone mineral density was reduced by approximately 2.3-fold in the AsPep-FTSQ group compared to the AsPep-SQ group (262.4 ± 48.11 vs. 115.0 ± 91.82), and the volume of ectopic bone decreased by about 12.5-fold (18.2 ± 10.10 vs. 1.46 ± 1.10) (Supplementary Fig. 6f and Supplementary Fig. 6g). Finally, we monitored body weight as an indicator of overall health. All four groups exhibited only minor fluctuations, with growth trajectories remaining within the normal physiological range for rats (Supplementary Fig. 6h).

In situ self-assembly and biosafety evaluation of AsPep-FTSQ in a Beagle tendinopathy model

To further evaluate the applicability and translational potential of AsPep-FTSQ across species, we developed a large-animal model of Achilles tendinopathy in beagles (Fig. 7a). We used a standardized surgical protocol to expose, transect, and suture the Achilles tendon (Supplementary Fig. 7a). Then, we performed multi-site intralesional injections of AsPep-FTSQ or the control peptides AsPep-FT and AsPep-SQ (Supplementary Fig. 7b). At the 16-week endpoint, gross inspection revealed that the tendons in the AsPep-FT group were markedly thickened with a dull appearance and surface scar adhesions. In contrast, the tendons in the AsPep-FTSQ group closely resembled those in the sham group, exhibiting near-normal morphology and tension (Fig. 7b).

Fig. 7 — a Schematic of the experimental protocol in Beagles. b Gross appearance of Achilles tendons at 16 weeks. Black arrow, disorganized collagen. c Transmission electron microscopy images of tendon ultrastructure at 16 weeks. Black dashed box, self-assembled peptide network. d Time course of serum IL-1β and TNF-α concentrations during follow-up. (n = 4 per group). e Sagittal T2-weighted MRI images. White arrow, Achilles tendon. f MRI heatmaps of T2 signal distribution. g Quantification of (i) relative mean signal intensity and (ii) tendon thickness. (n = 4 per group). h 3D CT reconstructions and coronal/sagittal slices of the distal Achilles region. White dashed box, Achilles tendon. i H&E staining of Achilles tendons at 16 weeks. Black arrows: tendon cells. Green arrow: disorganized collagen. j Quantification of total Bonar score. (n = 4 per group). k Masson staining showing collagen fiber organization and staining patterns in each group. l Quantification of collagen fiber angle distribution. (n = 4 per group). m Immunofluorescence staining of Col I and α-SMA in the Achilles tendon at 16 weeks. n Quantitative analysis of immunofluorescence: (i) Col I fluorescence intensity and (ii) α-SMA fluorescence intensity. (n = 4 per group). Data in d, g, j, l and n are presented as mean ± s.d., data in d were compared using two-way ANOVA with post hoc Tukey′s test, and data in g, j, l and n were compared using one-way ANOVA with post hoc Dunnett′s test. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Created in BioRender. Chao, L. (2026) https://BioRender.com/fnsxtb3.

Further analysis using TEM highlighted differences at the tissue microenvironment level. In the sham group, collagen fibers were organized into parallel bundles, and tenocytes exhibited regular morphology. In contrast, tendons in the AsPep-SQ and AsPep-FT groups exhibited pericellular ECM compaction, and the collagen fibers appeared wavy and disorganized. Interestingly, the AsPep-FTSQ group exhibited a continuous, electron-dense fibrous network between TCs and collagen fibers. This network was morphologically similar to the AsPep-FTSQ nanofibers observed in vitro (black dashed boxes) and was interwoven throughout the ECM (Fig. 7c). These observations suggest that following injection, the peptide persists in a self-assembled state within the tendon ECM of large animals. It integrates well with the host tissue without disrupting the native collagen ultrastructure.

Throughout the follow-up period, all animals remained in good general condition. Serum ELISA revealed a slight decline in Interleukin-1 beta (IL-1β) and tumor necrosis factor alpha (TNF-α) levels over time in all groups, with no significant differences among groups (Fig. 7d). To systematically assess the biosafety of this delivery strategy, we monitored hematological and biochemical parameters longitudinally at 0, 1, 2, 3, and 4 months. Complete blood counts revealed minor physiological fluctuations in leukocyte and erythrocyte counts, hemoglobin, hematocrit, platelet count, and mean corpuscular volume. There were no sustained upward or downward trends, and no significant differences were observed between groups (Supplementary Fig. 7c). Indices of liver, kidney, and myocardial injury were likewise stable overall. Alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), and alkaline phosphatase (ALP) remained within reference ranges at all time points, with no AsPep-FTSQ-specific elevations, indicating that hepatic function was not impaired (Supplementary Fig. 7d). Blood urea nitrogen and creatinine showed no persistent abnormalities, suggesting that renal function was not significantly affected (Supplementary Fig. 7e). Finally, cardiac enzymes creatine kinase-MB (CK-MB) and lactate dehydrogenase (LDH) exhibited no abnormal fluctuations during follow-up, indicating that the local injection of AsPep-FTSQ at the tested dose and frequency did not induce cardiotoxicity (Supplementary Fig. 7f).

At the four-month endpoint, we performed H&E staining of the kidneys, lungs, spleens, hearts, and livers of all Beagles. The histological architecture of all organs remained intact across the sham, AsPep-SQ, AsPep-FTSQ, and AsPep-FT groups, with preserved glomeruli and renal tubules, alveolar structures, splenic lymphoid follicles, myocardial fibers, and hepatic lobules. There was no evident necrosis, hemorrhage, extensive inflammatory cell infiltration, or marked fibrotic changes (Supplementary Fig. 7g).

AsPep-FTSQ attenuates tendon fibrosis on imaging and histology in Beagles

T2-weighted MRI revealed focal signal elevation and persistent Achilles tendon thickening in the AsPep-SQ group. In contrast, the sham group displayed homogeneous signal intensity and sharply defined tendon contours. AsPep-FTSQ treatment markedly reduced the extent of high-signal regions and decreased tendon thickness. The corresponding pseudocolor heatmaps facilitated visualization of the spatial contraction of high-signal regions and improved comparison of signal heterogeneity among groups (Fig. 7e, f). Quantitative analysis revealed that the AsPep-FTSQ group exhibited lower mean signal intensity (0.85 ± 0.23 vs. Sham 1.00 ± 0.37) and reduced tendon thickness (6.15 ± 0.24 vs. Sham 5.75 ± 0.26) compared with the AsPep-FT group. These values approached those observed in the sham group and were indicative of alleviated local edema and chronic inflammatory changes (Fig. 7g). Consistent with this, CT 3D reconstructions and tomographic images showed tendon thickening and focal narrowing due to adhesions in the AsPep-SQ group. No abnormal osseous overgrowth or obvious ectopic ossification was detected in the AsPep-FTSQ group (Fig. 7h).

At the histological level, H&E staining revealed that the Achilles tendons of the AsPep-SQ and AsPep-FT groups exhibited hallmark tendinopathic changes, including increased cell density, nuclear irregularity, and disrupted, interwoven collagen bundles. These abnormalities were markedly attenuated after AsPep-FTSQ treatment, which resulted in improved continuity of tendon bundles and clearer tissue stratification (Fig. 7i). Quantitative Bonar scoring confirmed that the total score in the AsPep-FTSQ group was significantly lower than in the AsPep-SQ group (10.00 ± 0.82 vs. 2.00 ± 1.41) and approached the sham level (Fig. 7j). Masson’s staining combined with collagen fiber angle analysis showed that fiber orientation in the AsPep-SQ group followed a broad-peaked distribution with impaired anisotropy along the loading axis. In contrast, collagen fibers in the AsPep-FTSQ group realigned along the tendon’s longitudinal axis, accompanied by a reduction in fiber angle (57.50 ± 6.86 vs. 8.25 ± 2.63) (Fig. 7k, l). Immunofluorescence analysis revealed an increase in α-SMA-positive cells and fibers, as well as an elevated Col I signal in the AsPep-SQ group. In contrast, AsPep-FTSQ treatment reduced α-SMA (876.30 ± 93.85 vs. 14.59 ± 2.95) and Col I (12.16 ± 2.13 vs. 1.16 ± 0.08) expression (Fig. 7m, n).

Discussion

Achilles tendinopathy is the most common type of tendinopathy in the lower limbs³⁵. The Achilles tendon is the strongest in the human body. Its structural and compositional features make it particularly susceptible to tendinopathy and subsequent rupture. Once Achilles tendinopathy develops, the tendon’s architecture becomes disorganized, accompanied by increased cellularity and neovascularization^1,36. To deepen our understanding of the mechanisms underlying tendinopathy, we combined single-cell transcriptomics with histological and imaging analyses. This allowed us to characterize disease-associated tendon cell states and contextualize potential molecular pathways contributing to pathological tendon remodeling. We discovered that the TC lineage originates from TSC in both normal and diseased tendons. Under tendinopathic conditions, however, the differentiation trajectory shifts from TC4 and TC6—primarily involved in ECM maintenance and repair—toward TC7 and TC5, dominated by osteo-/chondrogenic and stress-responsive programs. This “fate drift” closely mirrors the exacerbated fibrosis, collagen bundle disorganization, and heterotopic ossification observed in tendinopathic tissue. Notably, TC7 is the most abundant TC subset in diseased tendons and represents the terminal differentiation state of the TC lineage. After SoupX correction, the global clustering structure and major cell-type annotations remained largely unchanged, and the marker overlap among TC subclusters persisted, supporting the interpretation that this overlap mainly reflected biological similarity within the tendon cell lineage rather than ambient RNA contamination. Although the SoupX correction did not change our clustering or key biological conclusions, we acknowledge that ambient RNA contamination at low levels cannot be completely eliminated in droplet-based single-cell RNA sequencing.

Pseudotime analysis revealed two peaks of TGF-β1 expression along the disease trajectory that coincided precisely with the terminally differentiated TC7 subset. GO enrichment analysis showed that the predominant normal subset, TC4, is mainly involved in cell-matrix adhesion and integrin signaling. In contrast, the disease-dominant subset, TC7, is associated with bone/cartilage development and ECM organization. These findings provide functional support for the hypothesis that TGF-β1-driven osteo-/chondrogenic TCs orchestrate late-stage structural remodeling and offer mechanistic corroboration of the clinical observation that advanced tendinopathy is frequently accompanied by heterotopic ossification.

Consistent with these findings, the SQ-TC group exhibited a markedly higher proportion of TGF-β1-positive cells, intensified yet disorganized collagen staining, and significantly increased Bonar scores, collagen fiber angles, and quantifiable heterotopic ossification at the Achilles tendon insertion. These features collectively suggest that the tendon microenvironment transitioned from a reversible reparative phase to an “irreversible remodeling” stage, which is characterized by scar formation and bone-like reconstruction. Concurrently, the migratory capacity of TCs was significantly enhanced in the disease model, suggesting that activated TC subsets continuously participate in ECM rearrangement and filling. This provides both cellular sources and matrix support for scar tissue and calcific lesions. In summary, our data suggest that disease progression is associated with stage-specific elevation of Tgfb1/TGF-β1 signaling, expansion of osteo-/chondrogenic TC states such as TC7, and enrichment of stress-responsive TC5. These findings delineate clear cellular targets for TGF-β1-centered interventions and provide a conceptual basis for treating tendinopathy by modulating the lesional microenvironment rather than merely repairing structural defects.

When designing and constructing AsPep-FTSQ, we used the TGF-β1-binding segment of caveolin-1 as a template. Then, we grafted the minimal functional motif, FTTFTVT, into a canonical amino acid peptide scaffold. This scaffold comprises a Flag tag, flexible linker segments, and a self-assembly domain. Previous studies have shown that FTTFTVT suppresses TGF-β1 expression³⁰. However, this motif is a short peptide that cannot remain stable on its own, and previous studies did not modify or optimize it. Due to its multiple hydrophobic residues, including phenylalanine and valine, FTTFTVT exhibits poor aqueous solubility and cannot be developed directly as a small-molecule-like drug. To counterbalance this pronounced hydrophobicity, we selected amino acids with higher hydrophilicity when designing the other modules. The DYKDDDDK module consists mainly of hydrophilic lysine and aspartic acid residues. These residues increase the overall hydrophilicity of the peptide chain and facilitate experimental tracking of the construct. In the present design, this module functions as an auxiliary hydrophilic segment rather than as the therapeutic core of the peptide. The Q11 peptide (QQKFQFQFEQQ) is a well-established backbone sequence in self-assembling peptide systems, so it was chosen as the core assembly domain³⁷. The Q11 peptide has a balanced hydrophilic-hydrophobic profile and aromatic side chains. These properties enable the peptide to adopt β-sheet conformations and assemble into fibrils through hydrogen bonding, hydrophobic interactions, and other noncovalent forces³⁸. These three modules are linked by a serine-glycine-based linker peptide, resulting in a self-assembling construct with an integrated “capture-and-assembly” architecture.

Unlike many peptide systems, which require D-amino acids or non-natural building blocks for stable assembly, AsPep-FTSQ is composed entirely of natural L-amino acids. This design greatly enhances its prospects for future clinical translation and reduces the risk of immunogenic and inflammatory responses³⁹. AsPep-FTSQ spontaneously assembles into nanofibers and network-like architectures in aqueous and biofluid-mimicking environments. Together with its low CAC and the progressive expansion of the network over time, this indicates that AsPep-FTSQ can transition from discrete molecules to a spatially continuous network within near-physiological concentration ranges. This provides a tunable scaffold for the subsequent enrichment of TGF-β1. At the molecular level, AsPep-FTSQ interacts with TGF-β1 through multiple hydrogen bonds, as well as through pronounced van der Waals and electrostatic forces. Several of these contact sites overlap with those of the TGF-β1 receptor (TGF-β1R). This suggests that the peptide can sequester TGF-β1, reducing its accessibility to the receptor by competitively occupying part of its binding interface.

The Achilles tendon often bears loads exceeding seven times an individual’s body weight during running and other forms of physical activity⁴⁰. To more accurately mimic the in vivo state of tendinopathic TCs, we developed a system that applies cyclic mechanical stretch. Combining TGF-β1 stimulation with mechanical loading caused the TCs to transition from a high-density, cobblestone-like morphology to an elongated, spindle-shaped phenotype aligned with the direction of strain. This model reproduced key histopathological features of tendinopathy at the levels of cell morphology and ECM architecture and successfully induced a pathological drift toward a myofibroblast-like phenotype at the molecular level. Based on these findings, we introduced AsPep-FTSQ, which was shown to not exert nonspecific cytotoxic effects. Instead, it was found to selectively and dose-dependently downregulate fibrotic markers while restoring tenogenic gene expression. AsPep-FTSQ establishes a high-density TGF-β1-sequestering network at the cell- ECM interface. This network attenuates persistent TGF-β1-driven profibrotic signaling in TCs and partially restores a more tenogenic phenotypic state. The present data support sustained local sequestration as an effective strategy against chronic, pathological TGF-β1 overactivity in tendinopathy. However, TGF-β1 likely plays stage-dependent roles during tendon repair. Therefore, future studies should determine if different dosing intensities and treatment durations are necessary in the acute versus chronic phases to preserve early reparative signaling and suppress maladaptive fibrotic remodeling.

Taken together, the results from the rat and beagle in vivo models demonstrate that AsPep-FTSQ durably remodels the fibrotic microenvironment of tendinopathic lesions over the long term, providing concomitant structural and functional benefits through microenvironmental reprogramming. In the rat model, AsPep-FTSQ reduced local edema and chronic inflammation and suppressed myofibroblast-like activation at the cell-matrix interface. Notably, this microenvironmental and histological remodeling was not merely a structural improvement; it also led to tangible gains in tendon mechanics and functional performance. In the Beagle large-animal model, AsPep-FTSQ exhibited a high degree of cross-species consistency and translational potential. TEM revealed a persistent, electron-dense AsPep-FTSQ network between tenocytes and collagen bundles. This network closely resembled the nanofibrous morphology observed in vitro and preserved the native collagen ultrastructure. This provides strong morphological evidence for in situ self-assembly within the tendon ECM. Serial evaluation over four months, including complete blood counts, liver and kidney function tests, cardiac enzymes, and multi-organ H&E staining, showed no toxicity-related abnormalities. This indicates that, at the current dose and dosing schedule, the strategy confers favorable systemic biosafety at the large-animal level. Taken together, these findings introduce a safe, extensible, self-assembling peptide strategy for microenvironmental intervention and offer a highly translatable approach to precision therapy for tendinopathy and other fibrotic soft tissue disorders.

Methods

Ethics statement

All of the animals used in this study were purchased from the Experimental Animal Department at Peking University Health Science Center (SCXK2022-0009). This research was approved in advance by the Institutional Animal Care and Use Committee of Peking University Health Science Center (DLASBE0567 for SD rats, DLASBE0473 for beagles).

SD rats

SD rats weighing between 180 and 200 g, with an equal number of males and females, were randomly divided into four groups. Histological scoring and imaging-based quantitative analyses were performed by investigators blinded to group allocation. Food was withheld, and the surgical area was shaved 12 h before anesthesia was administered. General anesthesia was induced with 2% isoflurane (RWD, China) followed by an intraperitoneal injection of 1% sodium pentobarbital for maintenance. A midline incision was made over the Achilles tendon of the hindlimb. The tendon was exposed and partial bundles of collagen fibers were carefully transected using ophthalmic scissors to induce a focal injury and establish an Achilles tendinopathy model. In the sham group, only the skin over the Achilles tendon was incised and closed without injuring the tendon. One, two, three, four and six weeks after surgery, rats in the AsPep-SQ group received local injections (30 μL, 80 μM) of a peptide solution with the amino acid sequence DYKDDDDK-GSG-SGSG-QQKFQFQFEQQ into the Achilles tendon. Rats in the AsPep-FT group received a peptide with the sequence DYKDDDDK-GSG-FTTFTVT-SGSG, and those in the AsPep-FTSQ group received a peptide with the sequence DYKDDDDK-GSG-FTTFTVT-SGSG-QQKFQFQFEQQ.

Eight weeks after surgery, all rats were anesthetized and underwent magnetic resonance imaging (MRI; Philips, Netherlands), ultrasound examination (FUJIFILM VisualSonics, Canada), and microcomputed tomography (micro-CT; Bruker, Belgium). To visualize lesion-related signal heterogeneity, the original sagittal T2-weighted MRI images were rendered as pseudocolor heatmaps. The same intensity scale was used across groups. Quantitative gait analysis was performed in each group at 4 and 8 weeks postoperatively using an automated gait analysis system (CatWalk XT, Noldus, Netherlands). Simultaneously, the hind paws were coated with non-toxic ink, and the rats walked along an enclosed corridor, leaving black footprints on paper. The Achilles functional index (AFI) was then calculated from the footprint parameters, according to a previously described protocol for rat Achilles tendon injury(5). At the eight-week endpoint, all rats were euthanized, and their Achilles tendons were harvested. One portion of each tendon was fixed, embedded in paraffin, and sectioned for hematoxylin and eosin (H&E) staining, Masson’s trichrome staining, immunofluorescence staining, and immunohistochemical staining. A second portion was processed for scanning electron microscopy (SEM; Hitachi, Japan). The remaining tissue was used to extract total RNA with TRIzol (Sigma-Aldrich, USA), followed by quantitative polymerase chain reaction (qPCR) analysis of gene expression.

Beagles

Beagles weighing between 10 and 11 kg, with an equal number of males and females, were randomly assigned to four groups. Histological scoring and imaging-based quantitative analyses were performed by investigators blinded to group allocation. Food was withheld, and the surgical field was shaved 12 h before anesthesia was administered. Thirty min prior to surgery, atropine (0.03 mg/kg, subcutaneously; Handu, China) and butorphanol (0.2 mg/kg, intramuscularly; Hengrui, China) were administered. A 22-gauge intravenous catheter was placed in the cephalic vein, and anesthesia was induced with a slow intravenous injection of propofol (4 mg/kg; AstraZeneca, China) until the eyelid reflex was lost. Thereafter, 2% isoflurane (RWD, China) in oxygen was delivered via an endotracheal tube to maintain anesthesia. The Achilles tendinopathy model was established as described above. In the sham group, only a skin incision over the Achilles tendon was made, which was then closed without causing injury to the tendon. One, two, and three months after surgery, dogs in the AsPep-SQ group received local injections of a peptide solution (100 μL, 100 μM; DYKDDDDK-GSG-SGSG-QQKFQFQFEQQ) into the Achilles tendon. The AsPep-FT group received injections of a peptide with the sequence DYKDDDDK-GSG-FTTFTVT-SGSG (100 μL, 100 μM), while the AsPep-FTSQ group received injections of a peptide with the sequence DYKDDDDK-GSG-FTTFTVT-SGSG-QQKFQFQFEQQ (100 μL, 100 μM).

Three, five, seven, ten and thirteen weeks after surgery, venous blood was collected for a complete blood count, a comprehensive serum biochemistry test, and an enzyme-linked immunosorbent assay (ELISA). Four months after surgery, all dogs were anesthetized for MRI (Siemens, Germany) and computed tomography (CT; Philips, Netherlands). To visualize lesion-related signal heterogeneity, the original sagittal T2-weighted MRI images were rendered as pseudocolor heatmaps. The same intensity scale was used across groups. After imaging, the animals were euthanized and their Achilles tendons and major organs were harvested. Portions of each tissue were fixed, embedded in paraffin, and sectioned for H&E and Masson’s trichrome staining. Additional tendon samples were processed for transmission electron microscopy (TEM; Hitachi, Japan).

Peptide synthesis

The peptides used in this study were synthesized via standard Fmoc-based solid-phase peptide synthesis. Rink amide resin (0.6 mmol/g; GL Biochem, China) was swollen in dimethylformamide (DMF; Merck, Germany), and the Fmoc groups were then removed using a solution of 20% piperidine in DMF (Sigma-Aldrich, USA). Then, Fmoc-protected amino acids (GL Biochem, China) were sequentially coupled from the C-terminus to the N-terminus, according to the desired sequence. Oxyma Pure (Aladdin, China) was used as the coupling reagent in DMF, and the mixture was shaken for 30–45 min. Coupling completeness was monitored using the Kaiser test (Sigma-Aldrich, USA).

After chain assembly, the peptides were cleaved from the resin, and the side-chain protecting groups were simultaneously removed by treatment with a trifluoroacetic acid (TFA)/triisopropylsilane (TIPS)/water (95:2.5:2.5, v/v/v) mixture at room temperature for 1–2 h. The cleavage mixture was then filtered and precipitated with ice-cold, anhydrous diethyl ether (Aladdin, China). After centrifugation and drying, crude peptides were obtained. The crude products were then purified via reverse-phase high-performance liquid chromatography (HPLC) using a linear gradient of 0.1% TFA in water and 0.1% TFA in acetonitrile (5-60% over 30 min at a flow rate of 1.0 mL/min) on an HPLC system (Agilent, USA). The main peak was collected and lyophilized to yield final peptides with a purity greater than 97%. The molecular weights of the purified peptides were confirmed by electrospray ionization mass spectrometry (ESI-MS; Bruker, Belgium) and were consistent with the theoretical values.

TEM

Peptide samples: Peptide solutions were allowed to self-assemble in ultrapure water at the predetermined concentrations for 24 h before sampling. A 10 μL aliquot of each solution was dropped onto a carbon-coated copper grid (Zhongjingkeyi, China) and incubated at room temperature for 3 min. Excess liquid was carefully blotted, and the grid was negatively stained with 1% uranyl acetate (Sigma-Aldrich, USA) for 30 s, followed by gentle blotting and air-drying. Samples were imaged using a TEM (Hitachi, Japan) operated at an acceleration voltage of 120 kV.

Achilles tendon samples: Immediately after harvest, the specimens were immersed in pre-chilled 2.5% glutaraldehyde fixative (Servicebio, China) for 24 h. After fixation, the samples were rinsed three times with PBS (Beyotime, China) and post-fixed in 1% osmium tetroxide (Electron Microscopy Sciences, USA) for 2 h. The tissues were then dehydrated through a graded ethanol series (30%, 50%, 70%, 90%, and 100%) and treated with acetone as a transitional solvent. The tissues were embedded in epoxy resin (SPI Supplies, USA) and cured at 60 °C for 24 h. Ultrathin sections, approximately 70 nm thick, were cut using an ultramicrotome (Leica, Germany) and collected on copper grids (Zhongjingkeyi, China). The sections were stained sequentially with 2% uranyl acetate (Sigma-Aldrich, USA) and 2.6% lead citrate (Sigma-Aldrich, USA) for 10 min each. The samples were imaged using a TEM (Hitachi, Japan) operated at an acceleration voltage of 120 kV.

SEM

Peptide-cell and tendon samples: Tendon cells (TCs) were seeded onto pre-sterilized silicon wafers and cultured to approximately 70% confluence. Then, peptide solutions at predetermined concentrations were added, and the samples were incubated for defined time periods to allow the peptides to self-assemble on the cell surface or within the extracellular matrix (ECM). After treatment, the culture medium was discarded and the cells were rinsed three times with PBS. Then, the cells were fixed immediately in 2.5% glutaraldehyde (Servicebio, China) at 4 °C for 2 h. Achilles tendon specimens were harvested and immersed promptly in pre-chilled 2.5% glutaraldehyde fixative (Servicebio, China) for 24 h. Both the peptide-cell samples and the tendon tissues were dehydrated using a graded ethanol series (30%, 50%, 70%, 90%, 95%, and 100%). Then, they were transitioned from absolute ethanol (Aladdin, China) to a 1:1 ethanol/tert-butanol mixture and finally to pure tert-butanol. Critical point drying was performed using a critical point dryer (Leica, Germany) to preserve the morphology of the cell surfaces and the nanostructures of the peptides. After drying, the samples were sputter-coated with gold using an ion sputter coater (Quorum Technologies, UK) to enhance conductivity. The samples were then examined by SEM (Hitachi, Japan) at an accelerating voltage of 5 kV.

Circular dichroism (CD) spectra

CD spectra of peptide solutions at various concentrations were recorded after 12 h of pre-incubation using a CD spectropolarimeter (JASCO-1500, Japan). The spectra were acquired with a resolution of 1 nm, a scanning speed of 300 nm/min, and a wavelength range of 200–250 nm.

Critical assembly concentration (CAC)

First, a 1 mM peptide stock solution was prepared in dimethyl sulfoxide (DMSO; Sigma-Aldrich, USA). Prior to the assay, the peptide was diluted in ultrapure water (Millipore, USA) to create a series of working solutions at various concentrations. These solutions were then allowed to self-assemble at 4 °C for 12 h. A Thioflavin T (ThT; MedChemExpress, USA) stock solution was diluted to a final concentration of 1 mM to create the ThT working solution. ThT is a fluorescent dye with an excitation wavelength of 385 nm. Upon binding to β-sheet-rich structures, its excitation maximum shifts to 450 nm, and its emission maximum shifts to 490 nm. ThT was added to each peptide working solution at a final concentration of 10 μM. The solutions were gently mixed and incubated in the dark at room temperature for 30 min. Fluorescence intensity was measured using a spectrofluorometer (Hitachi F-7000, Japan) with an excitation wavelength of 450 nm; the emission peak was recorded at 490 nm. A plot of ThT fluorescence intensity versus peptide concentration was generated. The inflection point at which the fluorescence increased from slow to rapid was taken as the CAC of the peptide in pure water.

Molecular docking

Peptide sequences were submitted to AlphaFold 3 (https://golgi.sandbox.google.com/) to predict their three-dimensional structures. The crystal structure of TGF-β1 was obtained from the Protein Data Bank (PDB). The receptor structure was processed using PyMOL 2.1.0 to remove water molecules and small-molecule ligands, and further prepared in AutoDock Tools 1.5.6 by adding hydrogens and assigning charges, after which it was saved in pdbqt format. Molecular docking was then performed with TGF-β1 as the receptor and the peptides as ligands using the Vina 2.0 engine embedded in PyRx, which provided binding energies and output files. Docking results were finally visualized and analyzed using PyMOL.

Molecular dynamics simulation

To investigate the molecular mechanisms underlying peptide binding to TGF-β1 further, we performed molecular dynamics (MD) simulations of selected receptor-peptide complexes using the GROMACS 2020 software package. The AMBER99SB-ILDN force field was applied to the protein and the peptide, and the TIP3P explicit water model was used. Each complex was solvated in a cubic water box, ensuring a minimum distance of 1.0 nm between any protein atom and the box edge. Na⁺ or Cl⁻ ions were added as needed to neutralize the system according to the docking results. The MD workflow consisted of four steps: energy minimization, heating, equilibration, and production MD. First, the heavy atoms of the protein (and peptide) were restrained while the water molecules were minimized for 10,000 steps (5000 steps of steepest descent followed by 5000 steps of conjugate gradient). Then, the restraints were released and the entire system underwent an additional 10,000 steps of energy minimization (5000 steps of steepest descent followed by 5000 steps of conjugate gradient). The system was subsequently heated from 0 to 300 K over 50 ps and equilibrated under an NPT ensemble for 50 ps. Finally, a 100 ns production MD run was carried out under NPT conditions. Trajectory coordinates were saved every 10 ps and analyzed using trjconv and associated analysis tools. We calculated the binding free energies between the peptide and the protein using the gmx-MMPBSA approach implemented in GROMACS 2020.

Surface plasmon resonance (SPR)

SPR experiments were performed using a CM5 sensor chip (Cytiva, Sweden) to which recombinant rat TGF-β1 (R&D Systems, USA) was coupled via standard amine chemistry. Briefly, the flow cells were first activated with a 1:1 mixture of 0.1 M N-hydroxysuccinimide (NHS; Aladdin, China) and 0.4 M N-ethyl-N′-(3-dimethylaminopropyl) carbodiimide (EDC; Aladdin, China). Then, TGF-β1 diluted in 10 mM sodium acetate buffer (pH 4.5; Aladdin, China) was injected to achieve covalent immobilization with a target level of 1500–3000 resonance units (RU). Any remaining active groups on the chip surface were then blocked with 1 M ethanolamine (pH 8.5; Aladdin, China). The running buffer consisted of 10 mM PBS (pH 7.4) containing 0.05% Tween-20 and 0.1% bovine serum albumin (BSA) (Sigma-Aldrich, USA). Peptide samples were diluted in the same buffer, filtered, and degassed before use. Peptides at graded concentrations were injected over the TGF-β1-immobilized flow cell at a flow rate of 30 μL/min with an association phase of 120 s and a dissociation phase of 300 s.

A reference flow cell that was only subjected to ethanolamine blocking was used for background subtraction. Blank buffer injections were also included for double referencing. Each concentration was tested at least in duplicate. Sensorgrams were fitted using Biacore T200 evaluation software (Cytiva, Sweden), and equilibrium dissociation constants (Kd) were obtained using a 1:1 Langmuir binding model.

ELISA neutralization

Peptide solutions at various concentrations were pre-incubated for 12 h, after which they were added to 96-well plates. Then, an equal volume of TGF-β1 (1000 pg/mL; R&D Systems, USA) was added to each well, and the plates were incubated at 37 °C for 4 h. Following centrifugation, the resulting supernatants were collected, and the residual TGF-β1 concentration was quantified using a TGF-β1 ELISA kit (NeoBioScience, China). Absorbance at 450 nm (OD₄₅₀) was converted to TGF-β1 concentration using a standard curve to determine the amount of free (non-neutralized) TGF-β1. Then, inhibition curves were fitted using four-parameter logistic regression to calculate the Inhibitory concentration 50 (IC₅₀) values.

TC

Primary TCs were isolated from four-week-old SD rats. Briefly, the Achilles tendons were transferred to culture dishes containing 0.1% type I collagenase (Corning, USA) and digested overnight. The next day, an equal volume of complete medium (10% FBS, 89% MEM medium, and 1% penicillin/streptomycin) was added to stop the digestion process. Then, the cell suspension was centrifuged at 150 × g for 5 min to collect the TCs. The cell pellet was then resuspended in complete medium and cultured at 37 °C in a humidified incubator with 5% CO₂. Primary TCs from passage 2 (P2) to passage 4 (P4) were used for all in vitro experiments.

Extracellular assembly

Primary rat TCs were seeded onto pre-sterilized, glass-bottom culture dishes (NEST, China) and cultured at 37 °C with 5% CO₂ until they reached 70% confluence. The peptides were diluted in serum-free medium to a final concentration of 35 μM. After removing the original medium, the peptide-containing medium was added. The cells were then co-incubated with the peptides in the dark with gentle agitation for 48 h to prevent uneven peptide deposition. After incubation, the cells were gently washed twice with prewarmed PBS to remove unbound peptide, then fixed with 4% paraformaldehyde (Solarbio, China). The cells were then washed three times with PBS. Non-specific binding was blocked using a rapid blocking buffer (Beyotime, China) at room temperature for 15 min. The cells were then incubated with an anti-Flag primary antibody (CST, USA) overnight at 4 °C. After three PBS washes, two fluorescently labeled secondary antibodies with different excitation wavelengths (Abcam, USA) were applied, and the cells were incubated for one hour at room temperature. Then, Dil (ApexBio, USA) was added, and the cells were incubated at 37 °C for 20 min. Images of the cell membrane, cytoplasm, and ECM were acquired using a confocal laser scanning microscope (CLSM; Nikon, Japan), including z-stack images for 3D reconstruction. Total internal reflection fluorescence microscopy (TIRF, Nikon, Japan) was used to selectively excite fluorophores at the cell membrane-substrate interface to obtain high-resolution images of peptide aggregates and cell-peptide interactions at the adhesion plane. Image processing and analysis were performed using NIS-Elements software (Nikon, Japan). Detailed antibody information is provided in Supplementary Table 1.

Type I collagen-coated 3D culture

Primary rat TCs were cultured under standard conditions until they reached the logarithmic growth phase. Then, the cells were detached using 0.25% trypsin (Gibco, USA), centrifuged, and resuspended in pre-chilled serum-free medium. A solution of type I collagen (Corning, USA) and NaOH was mixed with the serum-free cell suspension at a volume ratio of 105:15:150 (v/v/v) on ice. The pH was then rapidly adjusted to 7.2–7.4, yielding a final collagen concentration of 3 mg/mL. Peptide working solutions were added simultaneously to achieve a final peptide concentration of 35 μM within the hydrogel. Recombinant TGF-β1 (R&D Systems, USA) was then supplemented to achieve a final concentration of 10 ng/mL. The mixture was transferred to a 37 °C, 5% CO₂ incubator and allowed to gel for 2 h to form 3D collagen hydrogels. After gelation, complete medium containing 10% FBS was gently added to the gels. Cultures were then maintained for an additional 48 h. Detailed antibody information is provided in Supplementary Table 1.

Cellular stress model

Primary rat TCs were seeded onto pre-sterilized, flexible-bottom culture plates (Flexcell International, USA) and cultured at 37 °C with 5% CO₂ until they reached 70% confluence. Then, the medium was replaced with a low-serum medium containing 2% FBS to minimize interference from the serum. Recombinant TGF-β1 (R&D Systems, USA) was added to a final concentration of 10 ng/mL together with pre-prepared peptide working solutions. The cells were then gently mixed and incubated for an additional 4 h to allow for initial peptide-cell interactions. Subsequently, the plates were mounted on a Flexcell FX-5000 tension system and subjected to cyclic stretch (5% equibiaxial strain at 1.0 Hz for 8 h per day) for 48 h. This mimics an in vitro tendinopathy-like microenvironment, which is characterized by repetitive mechanical loading and TGF-β1 enrichment.

Immunofluorescence

After completion of the preconditioning phase in the cellular stress model, subsequent procedures were performed as described for the “Extracellular assembly” protocol. Detailed antibody information is provided in Supplementary Table 1.

Western blot (WB)

The cells were lysed in pre-chilled RIPA buffer containing protease inhibitors (Beyotime, China) and incubated on ice for 30 min. Then, the cells were centrifuged at 12,000 g for 15 min at 4 °C to collect the supernatant. Protein concentrations were determined using a bicinchoninic acid (BCA) protein assay kit (Beyotime, China). Samples were adjusted to equal protein amounts before the addition of 5× SDS loading buffer. The samples were denatured at 95 °C for 5 minutes and separated on 4–20% gradient SDS-PAGE gels (GenScript, China), with 20–50 μg of protein per lane, at a constant voltage of 120 V, until the bromophenol blue dye front reached the bottom of the gel. The proteins were then transferred onto PVDF membranes, which were blocked with rapid blocking buffer (Beyotime, China) at room temperature for 20 min. The membranes were then incubated overnight at 4 °C with primary antibodies against collagen I (Col I), collagen III (Col III), fibronectin (FN 1), α-smooth muscle actin (α-SMA), periostin (Postn) (Abcam, USA), and SCX (Santa Cruz Biotechnology, USA). The next day, the membranes were washed 3 times with TBST (Beyotime, China) for 10 min each. Then, the membranes were incubated with HRP-conjugated secondary antibodies (CST, USA) for 1 h at room temperature. After an additional series of TBST washes, the protein bands were visualized using an enhanced chemiluminescence (ECL) kit (Millipore, USA) and imaged using a chemiluminescence detection system. Band intensities were quantified using Image Lab software (Bio-Rad, USA). Detailed antibody information is provided in Supplementary Table 1.

qPCR

Total RNA was extracted from cultured cells and tendon tissues using a standard phenol-chloroform-based procedure with TRIzol reagent (Invitrogen, USA), according to the manufacturer’s protocol. The yield and purity of the ribonucleic acid (RNA) were evaluated by spectrophotometry. Equal amounts of RNA from each sample were then reverse-transcribed into complementary deoxyribonucleic acid (cDNA) using a commercial kit (TaKaRa, Japan). qPCR was performed on a Thermo 7500 platform (Applied Biosystems, USA) using SYBR Green chemistry. Primer sets were validated to ensure comparable amplification efficiency. Relative gene expression levels were calculated using the 2^(−ΔΔ)Ct method with β-actin as the internal control. Detailed primer information is provided in Supplementary Table 2.

Single-cell transcriptome sequencing

The tendon samples included a normal group (Nor-TC) and a tendinopathy model group (SQ-TC). The SQ-TC model was established as described above. For each sample, the Achilles tendons of ten SD rats were pooled for processing and analysis. After euthanizing the rats under sterile conditions, the Achilles tendons were rapidly exposed, and the surrounding fat, muscle, and connective tissue were carefully removed. The clean tendon tissue was minced into small fragments, rinsed in pre-chilled phosphate-buffered saline (PBS; Beyotime, China), and transferred to a digestion buffer containing 3 mg/mL of type I collagenase (Corning, USA) and 4 mg/mL of neutral protease Dispase II (Roche, Switzerland). The mixture was then incubated at 37 °C in a 5% CO₂ cell incubator for 2 h. Undigested tissue was filtered out using a 40-μm sterile cell strainer (Millipore, USA). The filtrate was then centrifuged at 150 × g for 5 min. The supernatant was discarded, and the cell pellet was resuspended in sterile PBS. An aliquot of the single-cell suspension was mixed 1:1 with a 0.4% trypan blue solution (Beyotime, China), and the number of viable cells was counted using an automated cell counter (Countstar, China). The concentration of live cells was adjusted to 1000–2000 cells/μL.

To construct the single-cell libraries, qualified cell suspensions were loaded onto a 10x Genomics Chromium Controller using Chromium Next GEM Single Cell 3′ Reagent Kits v3.1, following the manufacturer’s instructions. The libraries were sequenced on an Illumina NovaSeq 6000 platform using 150-bp paired-end reads. The raw sequencing data were processed using Cell Ranger (10x Genomics) for demultiplexing the samples, aligning the reads, identifying the cells, identifying the barcodes, and counting the unique molecular identifiers (UMIs) to generate the gene-barcode expression matrix. Reads were aligned to the rat reference genome annotated according to [Ensembl_release111] for read alignment and gene quantification. Downstream analyses were performed in R using the Seurat package. Cells with fewer than 200 or more than 6200 detected genes, more than 36,000 UMIs, or more than 10% mitochondrial gene content were excluded. After quality control, normalization, and scaling, the top 2000 highly variable genes were selected using Seurat and used for principal component analysis (PCA) and subsequent dimensional reduction. To reduce sample-associated batch effects, data integration was performed using Harmony, with sample identity treated as the batch variable during Harmony correction. The first 50 principal components were used for downstream graph-based clustering and uniform manifold approximation and projection (UMAP) visualization. Graph-based clustering was performed using the Louvain algorithm. Cell types and subclusters were annotated based on canonical marker genes. Differentially expressed genes were identified using the hurdle model implemented in MAST (Model-Based Analysis of Single-Cell Transcriptomics). Genes with |log2FC | > 0.36, P < 0.05, and min.pct > 0.1 were considered differentially expressed. The min.pct value represents the proportion of cells expressing a given gene within the analyzed cell population; therefore, min.pct > 0.1 indicates that differentially expressed genes were required to be expressed in at least 10% of cells. For marker gene analysis, the Wilcoxon rank-sum test was used. Upregulated genes were considered potential marker genes when they met the following criteria: log2FC > 0.36, P < 0.01, and min.pct > 0.25. Marker gene analysis was performed using each cell as an individual input rather than a pseudobulk approach. For the TC and TSC subset analysis, TC and TSC populations were extracted after global cell-type annotation and analyzed independently. The same quality control, normalization, highly variable gene selection, Harmony-based data integration, PCA, Louvain clustering, UMAP visualization, differential expression analysis, and marker gene analysis workflow described above was applied to this subset analysis. For the pseudotime analysis, the tendon stem cell (TSC) and tendon cell (TC) populations that passed the Seurat filtering process were imported into Monocle 2. There, they were used to infer the trajectory using the DDRTree method. The reconstructed trajectory was then used to infer trajectory-associated transcriptional relationships between TSCs and TCs, which were visualized using the plot_cell_trajectory function. Monocle 2 assigned cell states along pseudotime, and the dynamic expression pattern of Tgfb1, which encodes TGF-β1, was plotted against pseudotime along the inferred trajectory. Gene ontology enrichment analysis was performed on the differentially expressed genes, and terms with a false discovery rate (FDR) of less than 0.05 were considered significantly enriched. All bioinformatic analyses were performed on the Omicsmart platform (https://www.omicsmart.com/home.html#/). The detailed number of cell information is provided in Supplementary Table 3-4.

Statistical analysis

All statistical data were processed with GraphPad Prism 9.0 software and expressed as mean ± s.d. (standard deviation). Two groups were compared using two-tailed, unpaired Student′s t test, and multiple groups were compared using one-way ANOVA with post hoc Dunnett’s test or two-way ANOVA with post hoc Tukey′s test. P < 0.05 was set as the statistically significant standard. Statistical significance was represented as *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. The representative data reported were performed at least three independent experiments.

Reporting summary

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

Supplementary information

Supplementary information. pdf^{(2.4MB, pdf)}

Reporting Summary^{(88.8KB, pdf)}

Transparent Peer Review file^{(1.8MB, pdf)}

Source data

Source Data^{(1.8MB, xlsx)}

Acknowledgements

The authors appreciate the help from Jingjing Ning at Guangzhou Gene Denovo Biotechnology Co., Ltd. with bioinformatics analysis. We thank Bingbing Xu (Peking University Third Hospital) for her contributions to the preparation and approval of the animal ethics applications. We thank Jiying Zhang (Peking University Third Hospital) for insightful assistance with pathological evaluation and data interpretation.

Author contributions

C.L. and Z.C. contributed equally to this work. C.L., Z.C. and W.L. conceived the study and designed the experiments with input from G.C., H.L., and J.Y. C.L., W.L., R.D., L.H., Y.S., and Y.W. performed experiments, collected data, and contributed to data analysis and interpretation. G.C. supervised the establishment and optimization of animal models. Z.C. provided guidance on study design, statistical analysis, and data interpretation. C.L. drafted the manuscript with contributions from W.L., R.D., L.H., Y.S., Y.W., and Z.C. H.L. and J.Y. critically revised the manuscript, supervised the overall project, and secured funding. All authors discussed the results, commented on the manuscript, and approved the final version.

Peer review

Peer review information

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

Funding

This work was supported by the National Natural Science Foundation of China (82172423 G.C.), the Young Scientists Fund of the National Natural Science Foundation of China (82302656 Y.S., 82502959 R.D.), the Beijing Municipal Natural Science Foundation (7254454 Y.W.), the China Postdoctoral Science Foundation (2023M730123 Y.W., 2025M782923 R.D.), the Postdoctoral Fellowship Program of CPSF (GZB20250663 R.D.), the Joint Fund for Regional Innovation and Development of the National Natural Science Foundation of China and Beijing Municipal Natural Science Foundation (U22A20283 J.Y.), the Capital Clinical Characteristic Diagnosis and Treatment Technology Research and Translational Application Project (Z221100007422013 J.Y.), the Researcher-initiated research projects (12025Z01037 H.L., 12025Z01036 H.L.) and the National Key Research and Development Program of China (2024YFF0618203 R.D.).

Data availability

The data supporting the results in this study are available within the paper and its Supplementary Information. The raw sequence data of Single-cell RNA sequencing reported in this paper have been deposited in the Genome Sequence Archive in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences with accessions CRA040031. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Chao Li, Zehao Chen.

Contributor Information

Guoqing Cui, Email: drcuiguoqing1964@126.com.

Huawei Liu, Email: lhwleo@163.com.

Jiakuo Yu, Email: yujiakuo@126.com, Email: yujiakuo@tsinghua.edu.cn.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76487-3.

References

1.Traweger, A. et al. Achilles tendinopathy. Nat. Rev. Dis. Prim.11, 20 (2025). [DOI] [PubMed] [Google Scholar]
2.Magnusson, S. P., Langberg, H. & Kjaer, M. The pathogenesis of tendinopathy: balancing the response to loading. Nat. Rev. Rheumatol.6, 262–268 (2010). [DOI] [PubMed] [Google Scholar]
3.Cardoso, T. B., Pizzari, T., Kinsella, R., Hope, D. & Cook, J. L. Current trends in tendinopathy management. Best. Prat. Res. Clin. Rheumatol.33, 122–140 (2019). [DOI] [PubMed] [Google Scholar]
4.Jin, S. et al. Young exosome bio-nanoparticles restore aging-impaired tendon stem/progenitor cell function and reparative capacity. Adv. Mater.35, e2211602 (2023). [DOI] [PubMed] [Google Scholar]
5.Wang, Y. et al. Functional regeneration and repair of tendons using biomimetic scaffolds loaded with recombinant periostin. Nat. Commun.12, 1293 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Pringels, L. et al. Exploring the role of intratendinous pressure in the pathogenesis of tendon pathology: a narrative review and conceptual framework. Br. J. Sports Med57, 1042–1048 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Yang, R. et al. Gradient bimetallic ion-based hydrogels for tissue microstructure reconstruction of tendon-to-bone insertion. Sci. Adv.7, eabg3816 (2021). [DOI] [PMC free article] [PubMed]
8.Massagué, J. & Sheppard, D. TGF-β signaling in health and disease. Cell186, 4007–4037 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Deng, Z. et al. TGF-β signaling in health, disease, and therapeutics. Signal Transduct. Target Ther.9, 61 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Fu, W., Yang, R. & Li, J. Single-cell and spatial transcriptomics reveal changes in cell heterogeneity during progression of human tendinopathy. BMC Biol.21, 132 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Zhang, D. et al. Targeting local lymphatics to ameliorate heterotopic ossification via FGFR3-BMPR1a pathway. Nat. Commun.12, 4391 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Lin, J. et al. Single cell analysis reveals inhibition of angiogenesis attenuates the progression of heterotopic ossification in Mkx-/- mice. Bone Res. 10, 4 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Solomon, J. J. et al. Safety, tolerability, and efficacy of pirfenidone in patients with rheumatoid arthritis-associated interstitial lung disease: a randomised, double-blind, placebo-controlled, phase 2 study. Lancet Respir. Med.11, 87–96 (2022). [DOI] [PubMed] [Google Scholar]
14.Tam, K. T. & Baar, K. Using load to improve tendon/ligament tissue engineering and develop novel treatments for tendinopathy. Matrix Biol.135, 39–54 (2024). [DOI] [PubMed] [Google Scholar]
15.Wang, M.-D., Lv, G.-T., An, H.-W., Zhang, N.-Y. & Wang, H. In situ self-assembly of bispecific peptide for cancer immunotherapy. Angew. Chem. Int Ed. Engl.61, e202113649 (2022). [DOI] [PubMed] [Google Scholar]
16.Gao, Y. et al. Advances in self-assembled peptides as drug carriers. Pharmaceutics15, 482 (2023). [DOI] [PMC free article] [PubMed]
17.Mao, C. et al. In situ editing of tumour cell membranes induces aggregation and capture of PD-L1 membrane proteins for enhanced cancer immunotherapy. Nat. Commun.15, 9723 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Wang, M.-D. et al. Homologous peptide foldamer promotes FUS aggregation and triggers cancer cell death. J. Am. Chem. Soc.146, 28669–28676 (2024). [DOI] [PubMed] [Google Scholar]
19.Yorke, S. K. et al. Design and sustainability of polypeptide material systems. Nat. Rev. Mater.10, 750–768 (2025). [Google Scholar]
20.Liang, Q.-L. et al. Binding-induced fibrillogenesis peptide inhibits RANKL-mediated osteoclast activation against osteoporosis. Biomaterials302, 122331 (2023). [DOI] [PubMed] [Google Scholar]
21.Zou, R. et al. Peptide self-assembly triggered by metal ions. Chem. Soc. Rev.44, 5200–5219 (2015). [DOI] [PubMed] [Google Scholar]
22.Zhou, Y. et al. Molecularly stimuli-responsive self-assembled peptide nanoparticles for targeted imaging and therapy. ACS Nano17, 8004–8025 (2023). [DOI] [PubMed] [Google Scholar]
23.Ramirez-Labrada, A. et al. Multiparametric in vitro and in vivo analysis of the safety profile of self-assembling peptides. Sci. Rep.14, 4395 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Gregorio, N. E., Haas, C. M., King, N. P. & DeForest, C. A. Engineering complexity into protein-based biomaterials for biomedical applications. Nat. Rev. Mater.11, 328–342 (2025).
25.Wang, W. et al. Bio-orthogonal engineered peptide: a multi-functional strategy for the gene therapy of osteoporotic bone loss. Biomaterials302, 122352 (2023). [DOI] [PubMed] [Google Scholar]
26.Hou, D.-Y. et al. In vivo assembly enhanced binding effect augments tumor specific ferroptosis therapy. Nat. Commun.15, 454 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Chintalaramulu, N. et al. Caveolin-1: an ambiguous entity in breast cancer. Mol. Cancer24, 109 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Razani, B. et al. Caveolin-1 regulates transforming growth factor (TGF)-beta/SMAD signaling through an interaction with the TGF-beta type I receptor. J. Biol. Chem.276, 6727–6738 (2000). [DOI] [PubMed] [Google Scholar]
29.Bucci, M. et al. In vivo delivery of the caveolin-1 scaffolding domain inhibits nitric oxide synthesis and reduces inflammation. Nat. Med6, 1362–1367 (2000). [DOI] [PubMed] [Google Scholar]
30.Marudamuthu, A. S. et al. Caveolin-1-derived peptide limits development of pulmonary fibrosis. Sci. Transl. Med.11, eaat2848 (2019). [DOI] [PubMed]
31.Dong, Z. et al. Tendon repair and regeneration using bioinspired fibrillation engineering that mimicked the structure and mechanics of natural tissue. ACS Nano17, 17858–17872 (2023). [DOI] [PubMed] [Google Scholar]
32.Lu, J. et al. Rejuvenation of tendon stem/progenitor cells for functional tendon regeneration through platelet-derived exosomes loaded with recombinant Yap1. Acta Biomater.161, 80–99 (2023). [DOI] [PubMed] [Google Scholar]
33.Chen, Z. et al. Mechanical signal-tailored hydrogel microspheres recruit and train stem cells for precise differentiation. Adv. Mater.35, e2300180 (2023). [DOI] [PubMed] [Google Scholar]
34.Wang, X. et al. Inhibition of overactive TGF-β attenuates progression of heterotopic ossification in mice. Nat. Commun.9, 551 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Traweger, A. et al. Achilles tendinopathy. Nat. Rev. Dis. Primers11, 21 (2025). [DOI] [PubMed]
36.Mimpen, J. Y. et al. Single nucleus and spatial transcriptomic profiling of healthy human hamstring tendon. FASEB J.38, e23629 (2024). [DOI] [PubMed] [Google Scholar]
37.Rudra, J. S., Tian, Y. F., Jung, J. P. & Collier, J. H. A self-assembling peptide acting as an immune adjuvant. Proc. Natl. Acad. Sci. USA107, 622–627 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Yang, Y. et al. A personalized vaccine combining immunogenic cell death-induced cells and nanosized antigens for enhanced antitumor immunity. J. Control Release376, 1271–1287 (2024). [DOI] [PubMed] [Google Scholar]
39.Wang, C., Qi, R., Wang, R. & Xu, Z. Photoinduced C(sp3)-H functionalization of glycine derivatives: preparation of unnatural α-amino acids and late-stage modification of peptides. Acc. Chem. Res.56, 2110–2125 (2023). [DOI] [PubMed] [Google Scholar]
40.Demangeot, Y., Whiteley, R., Gremeaux, V. & Degache, F. The load borne by the Achilles tendon during exercise: a systematic review of normative values. Scand. J. Med Sci. Sports33, 110–126 (2022). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary information. pdf^{(2.4MB, pdf)}

Reporting Summary^{(88.8KB, pdf)}

Transparent Peer Review file^{(1.8MB, pdf)}

Source Data^{(1.8MB, xlsx)}

Data Availability Statement

[CR1] 1.Traweger, A. et al. Achilles tendinopathy. Nat. Rev. Dis. Prim.11, 20 (2025). [DOI] [PubMed] [Google Scholar]

[CR2] 2.Magnusson, S. P., Langberg, H. & Kjaer, M. The pathogenesis of tendinopathy: balancing the response to loading. Nat. Rev. Rheumatol.6, 262–268 (2010). [DOI] [PubMed] [Google Scholar]

[CR3] 3.Cardoso, T. B., Pizzari, T., Kinsella, R., Hope, D. & Cook, J. L. Current trends in tendinopathy management. Best. Prat. Res. Clin. Rheumatol.33, 122–140 (2019). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Jin, S. et al. Young exosome bio-nanoparticles restore aging-impaired tendon stem/progenitor cell function and reparative capacity. Adv. Mater.35, e2211602 (2023). [DOI] [PubMed] [Google Scholar]

[CR5] 5.Wang, Y. et al. Functional regeneration and repair of tendons using biomimetic scaffolds loaded with recombinant periostin. Nat. Commun.12, 1293 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR6] 6.Pringels, L. et al. Exploring the role of intratendinous pressure in the pathogenesis of tendon pathology: a narrative review and conceptual framework. Br. J. Sports Med57, 1042–1048 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Yang, R. et al. Gradient bimetallic ion-based hydrogels for tissue microstructure reconstruction of tendon-to-bone insertion. Sci. Adv.7, eabg3816 (2021). [DOI] [PMC free article] [PubMed]

[CR8] 8.Massagué, J. & Sheppard, D. TGF-β signaling in health and disease. Cell186, 4007–4037 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Deng, Z. et al. TGF-β signaling in health, disease, and therapeutics. Signal Transduct. Target Ther.9, 61 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Fu, W., Yang, R. & Li, J. Single-cell and spatial transcriptomics reveal changes in cell heterogeneity during progression of human tendinopathy. BMC Biol.21, 132 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Zhang, D. et al. Targeting local lymphatics to ameliorate heterotopic ossification via FGFR3-BMPR1a pathway. Nat. Commun.12, 4391 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Lin, J. et al. Single cell analysis reveals inhibition of angiogenesis attenuates the progression of heterotopic ossification in Mkx-/- mice. Bone Res. 10, 4 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Solomon, J. J. et al. Safety, tolerability, and efficacy of pirfenidone in patients with rheumatoid arthritis-associated interstitial lung disease: a randomised, double-blind, placebo-controlled, phase 2 study. Lancet Respir. Med.11, 87–96 (2022). [DOI] [PubMed] [Google Scholar]

[CR14] 14.Tam, K. T. & Baar, K. Using load to improve tendon/ligament tissue engineering and develop novel treatments for tendinopathy. Matrix Biol.135, 39–54 (2024). [DOI] [PubMed] [Google Scholar]

[CR15] 15.Wang, M.-D., Lv, G.-T., An, H.-W., Zhang, N.-Y. & Wang, H. In situ self-assembly of bispecific peptide for cancer immunotherapy. Angew. Chem. Int Ed. Engl.61, e202113649 (2022). [DOI] [PubMed] [Google Scholar]

[CR16] 16.Gao, Y. et al. Advances in self-assembled peptides as drug carriers. Pharmaceutics15, 482 (2023). [DOI] [PMC free article] [PubMed]

[CR17] 17.Mao, C. et al. In situ editing of tumour cell membranes induces aggregation and capture of PD-L1 membrane proteins for enhanced cancer immunotherapy. Nat. Commun.15, 9723 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Wang, M.-D. et al. Homologous peptide foldamer promotes FUS aggregation and triggers cancer cell death. J. Am. Chem. Soc.146, 28669–28676 (2024). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Yorke, S. K. et al. Design and sustainability of polypeptide material systems. Nat. Rev. Mater.10, 750–768 (2025). [Google Scholar]

[CR20] 20.Liang, Q.-L. et al. Binding-induced fibrillogenesis peptide inhibits RANKL-mediated osteoclast activation against osteoporosis. Biomaterials302, 122331 (2023). [DOI] [PubMed] [Google Scholar]

[CR21] 21.Zou, R. et al. Peptide self-assembly triggered by metal ions. Chem. Soc. Rev.44, 5200–5219 (2015). [DOI] [PubMed] [Google Scholar]

[CR22] 22.Zhou, Y. et al. Molecularly stimuli-responsive self-assembled peptide nanoparticles for targeted imaging and therapy. ACS Nano17, 8004–8025 (2023). [DOI] [PubMed] [Google Scholar]

[CR23] 23.Ramirez-Labrada, A. et al. Multiparametric in vitro and in vivo analysis of the safety profile of self-assembling peptides. Sci. Rep.14, 4395 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Gregorio, N. E., Haas, C. M., King, N. P. & DeForest, C. A. Engineering complexity into protein-based biomaterials for biomedical applications. Nat. Rev. Mater.11, 328–342 (2025).

[CR25] 25.Wang, W. et al. Bio-orthogonal engineered peptide: a multi-functional strategy for the gene therapy of osteoporotic bone loss. Biomaterials302, 122352 (2023). [DOI] [PubMed] [Google Scholar]

[CR26] 26.Hou, D.-Y. et al. In vivo assembly enhanced binding effect augments tumor specific ferroptosis therapy. Nat. Commun.15, 454 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Chintalaramulu, N. et al. Caveolin-1: an ambiguous entity in breast cancer. Mol. Cancer24, 109 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Razani, B. et al. Caveolin-1 regulates transforming growth factor (TGF)-beta/SMAD signaling through an interaction with the TGF-beta type I receptor. J. Biol. Chem.276, 6727–6738 (2000). [DOI] [PubMed] [Google Scholar]

[CR29] 29.Bucci, M. et al. In vivo delivery of the caveolin-1 scaffolding domain inhibits nitric oxide synthesis and reduces inflammation. Nat. Med6, 1362–1367 (2000). [DOI] [PubMed] [Google Scholar]

[CR30] 30.Marudamuthu, A. S. et al. Caveolin-1-derived peptide limits development of pulmonary fibrosis. Sci. Transl. Med.11, eaat2848 (2019). [DOI] [PubMed]

[CR31] 31.Dong, Z. et al. Tendon repair and regeneration using bioinspired fibrillation engineering that mimicked the structure and mechanics of natural tissue. ACS Nano17, 17858–17872 (2023). [DOI] [PubMed] [Google Scholar]

[CR32] 32.Lu, J. et al. Rejuvenation of tendon stem/progenitor cells for functional tendon regeneration through platelet-derived exosomes loaded with recombinant Yap1. Acta Biomater.161, 80–99 (2023). [DOI] [PubMed] [Google Scholar]

[CR33] 33.Chen, Z. et al. Mechanical signal-tailored hydrogel microspheres recruit and train stem cells for precise differentiation. Adv. Mater.35, e2300180 (2023). [DOI] [PubMed] [Google Scholar]

[CR34] 34.Wang, X. et al. Inhibition of overactive TGF-β attenuates progression of heterotopic ossification in mice. Nat. Commun.9, 551 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Traweger, A. et al. Achilles tendinopathy. Nat. Rev. Dis. Primers11, 21 (2025). [DOI] [PubMed]

[CR36] 36.Mimpen, J. Y. et al. Single nucleus and spatial transcriptomic profiling of healthy human hamstring tendon. FASEB J.38, e23629 (2024). [DOI] [PubMed] [Google Scholar]

[CR37] 37.Rudra, J. S., Tian, Y. F., Jung, J. P. & Collier, J. H. A self-assembling peptide acting as an immune adjuvant. Proc. Natl. Acad. Sci. USA107, 622–627 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Yang, Y. et al. A personalized vaccine combining immunogenic cell death-induced cells and nanosized antigens for enhanced antitumor immunity. J. Control Release376, 1271–1287 (2024). [DOI] [PubMed] [Google Scholar]

[CR39] 39.Wang, C., Qi, R., Wang, R. & Xu, Z. Photoinduced C(sp3)-H functionalization of glycine derivatives: preparation of unnatural α-amino acids and late-stage modification of peptides. Acc. Chem. Res.56, 2110–2125 (2023). [DOI] [PubMed] [Google Scholar]

[CR40] 40.Demangeot, Y., Whiteley, R., Gremeaux, V. & Degache, F. The load borne by the Achilles tendon during exercise: a systematic review of normative values. Scand. J. Med Sci. Sports33, 110–126 (2022). [DOI] [PubMed] [Google Scholar]

PERMALINK

A modular self-assembling peptide platform targeting TGF-β1 for tendon regeneration

Chao Li

Zehao Chen

Wentao Li

Ronghui Deng

Lingan Huang

Yifan Song

Yiqun Wang

Guoqing Cui

Huawei Liu

Jiakuo Yu

Abstract

Introduction

Fig. 1. Design and mechanism of AsPep-FTSQ.

Results

Single-cell transcriptomic analysis reveals lineage remodeling of TCs in tendinopathy

Fig. 2. Lineage drift of tendon cells and phenotypic alterations in tendinopathy.

Key phenotypic alterations in tendinopathy

Design and self-assembly characterization of AsPep-FTSQ

Fig. 3. Design, self-assembly behavior, and TGF-β1-binding properties of AsPep-FTSQ.

Specific binding of AsPep-FTSQ to TGF-β1 and its binding kinetics

TGF-β1 combined with cyclic mechanical stretch induces a tendinopathy-like fibrotic phenotype in TCs in vitro

Fig. 4. Establishment of a TGF-β1 plus cyclic stretch-induced in vitro tendinopathy-like model and characterization of fibrotic and tenogenic phenotypes.

AsPep-FTSQ attenuates fibrotic activation and promotes tenogenic differentiation of TCs in vitro

Imaging and histological assessment of AsPep-FTSQ repair in an SD rat Achilles tendinopathy model

Fig. 5. Imaging and histological evaluation of AsPep-FTSQ-mediated repair in an SD rat Achilles tendinopathy model.

AsPep-FTSQ improves gait and functional recovery in a rat model of tendinopathy

Fig. 6. AsPep-FTSQ improves gait function, restores collagen fiber alignment, and attenuates heterotopic ossification.

AsPep-FTSQ restores collagen fiber alignment and attenuates heterotopic ossification

In situ self-assembly and biosafety evaluation of AsPep-FTSQ in a Beagle tendinopathy model

Fig. 7. AsPep-FTSQ undergoes in situ self-assembly and alleviates tendon fibrosis in a Beagle tendinopathy model.

AsPep-FTSQ attenuates tendon fibrosis on imaging and histology in Beagles

Discussion

Methods

Ethics statement

SD rats

Beagles

Peptide synthesis

TEM

SEM

Circular dichroism (CD) spectra

Critical assembly concentration (CAC)

Molecular docking

Molecular dynamics simulation

Surface plasmon resonance (SPR)

ELISA neutralization

TC

Extracellular assembly

Type I collagen-coated 3D culture

Cellular stress model

Immunofluorescence

Western blot (WB)

qPCR

Single-cell transcriptome sequencing

Statistical analysis

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review

Peer review information

Funding

Data availability

Competing interests

Footnotes

Contributor Information

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases