Functional matrix vesicle replacement partially restores chemo-mechanical coupling in the aged bone niche

Summary

Skeletal aging involves pyrophosphate/phosphate disequilibrium and impaired mechanotransduction, which together constrain osteogenic repair. We develop OsteoVes, a matrix-vesicle-mimetic extracellular organelle that combines tissue-nonspecific alkaline phosphatase (ALP)-mediated inorganic pyrophosphate (PPi) hydrolysis, nano-hydroxyapatite nucleation, and a mesenchymal stem cell-derived membrane interface for extracellular matrix anchoring. In aged human mesenchymal stem/stromal cells (MSCs), OsteoVes restores the Pi/PPi set point, suppresses PPARG activity, promotes RUNX2 nuclear translocation, and supports osteogenic differentiation. OsteoVes also increases actomyosin tension and engages an ITG/FAK-PIEZO1-JNK/c-JUN-RUNX2 mechanotransduction-associated pathway. In elderly osteopenic/osteoporotic patient-derived MSCs, OsteoVes supports ALP activity and matrix mineralization. Systemic administration improves trabecular microarchitecture and mechanical properties in naturally aged mice within 4 weeks, remains active in Alpl^+/− progeroid mice, and elicits osteoanabolic serum and imaging responses without evident short-term toxicity in metabolically aged rhesus macaques. These findings identify matrix vesicle (MV) dysfunction as an extracellular osteometabolic control point for short-term chemo-mechanical rescue.

Graphical abstract

Highlights

• •OsteoVes restores ALP-dependent Pi/PPi balance in aged human MSCs
• •MSC membrane coating supports ECM anchoring of matrix-vesicle-mimetic cores
• •OsteoVes engages mechanotransduction to promote RUNX2-dependent osteogenesis
• •Systemic OsteoVes improves aged and ALPL-deficient bone phenotypes in the short term

Zhang et al. develop OsteoVes, a matrix-vesicle-mimetic replacement that restores Pi/PPi balance, ECM anchoring, and mechanotransduction in aged bone niches. OsteoVes rescues osteogenic defects in elderly patient-derived MSCs and improves short-term bone outcomes in aged mice and metabolically aged rhesus macaques.

Introduction

Aging-related bone loss is not only a problem of declining osteoblast activity; it is also a disorder of osteometabolic homeostasis in which local phosphate handling, matrix mineralization, and mechanical signal propagation progressively fall out of register.^{1,2,3,4,5,6,7,8} This mismatch is particularly relevant in older individuals with metabolic comorbidities, such as hyperglycemia and dyslipidemia, where bone fragility can be disproportionate to bone mass and reflects impaired bone quality as well as altered bone turnover.^9,10 Thus, skeletal aging is best understood not simply as a cell-intrinsic loss of osteogenic signaling but as a tissue-level failure to maintain the metabolic and mechanical conditions required for regenerative competence.^{11,12,13,14,15}

A major reason this failure becomes persistent is that aging phenotypes are progressively externalized into the microenvironment. Once the aged bone niche has been remodeled, mesenchymal stem/stromal cells (MSCs) are chronically exposed to an extracellular field that disfavors osteogenesis and promotes adipogenic drift.^{12,13,14,15,16,17} This is clinically consequential because older patients, including those with osteoporosis and fragility fractures, represent the population in which bone-regenerative therapies must ultimately work, and patient-derived human MSCs (hMSCs) have been reported to display disease-associated functional and transcriptional alterations that are not fully explained by non-osteoporotic aging alone.^18,19 As a result, strategies that stimulate only a single intracellular pathway may be insufficient to reset the larger extracellular steady state that constrains lineage output.^11,12,13,14

Within this osteometabolic niche, matrix vesicles (MVs) function as extracellular organelles that couple phosphate metabolism to matrix mechanics.^1,2,5,6,7,8 By concentrating tissue-nonspecific alkaline phosphatase (TNAP/ALP) and mineralization substrates within a membrane-bounded nanoscale compartment, MVs hydrolyze inorganic pyrophosphate (PPi), elevate the local Pi/PPi ratio, and nucleate calcium phosphate deposition.^5,6,7,8 In doing so, they do not merely transport mineral cargo; they convert a local metabolic state into a mechanically perceivable extracellular matrix (ECM) context.^5,6,7,8,20 During aging, reduced MV function—particularly loss of ALP activity—raises the mineralization threshold, weakens matrix mineral accrual, and degrades adhesion- and force-transmission cues, thereby suppressing osteogenic programs and reinforcing fate locking.^6,7,8,21,22 Because MVs are acellular, anucleate organelles, restoring their coupled functions offers a tractable route to reset the aged niche without transplanting “young” cells.^5,6,21,23

Here, we develop OsteoVes, a biomimetic MV replacement designed to restore three coupled functions of native MVs: ALPL-mediated PPi hydrolysis, nano-hydroxyapatite (nHA)-based nucleation, and membrane-interface anchoring within the ECM. We further build OsteoVes with translational manufacturability in mind by combining scalable recombinant human ALPL production with a human MSC membrane interface validated in this study as a substitute for scarce osteoblast-derived MV membranes. We test OsteoVes across aged hMSCs, bone marrow MSCs isolated from elderly osteopenic/osteoporotic fracture patients, naturally aged mice, Alpl^+/− progeroid mice, and a metabolically aged rhesus macaque model driven by hyperglycemia and hyperlipidemia.^9,10,24 Across these systems, we ask whether extracellular organelle replacement can restore Pi/PPi-mechanotransduction coupling, unlock osteogenic fate, and improve aged or pathology-associated bone-niche function within the observation windows tested.

Results

OsteoVes design integrates ALPL catalysis, mineral nucleation, and membrane anchoring

The critical gap in chemo-mechanical coupling within the aged bone niche can be abstracted into a “minimal replaceable functional unit” that, within a single nanoscale entity, simultaneously (1) tunes the Pi/PPi threshold via ALP to relieve mineralization inhibition, (2) provides a stable nucleation platform to initiate calcium phosphate deposition, and (3) localizes to the ECM through a membrane interface that supports cell/ECM-perceivable interactions, thereby translating “chemically mineralizable” conditions into a “mechanically perceivable” matrix information background.^2,5 Guided by this logic, we set out to construct a biomimetic MV, OsteoVes, to replenish impaired MV function and re-establish mineralization-mechanics coupling in the aged bone niche (Figure 1A). As shown in Figure 1B, the design and manufacturing workflow comprises the following: producing catalytically active ALP protein, assembling an ALP-nHA composite mineralization core, and reconstructing an MV-like membrane architecture and adhesion/targeting features using a scalable cell-membrane interface, yielding an integrated biomimetic functional vesicle.

Figure 1.

Fabrication, physicochemical characterization, and in vitro mineralization of OsteoVes

(A and B) Schematic of the OsteoVes manufacturing workflow. Recombinant human ALPL (encoding tissue-nonspecific alkaline phosphatase; C-terminal 8× His-tag) was expressed in a prokaryotic expression system, isolated from inclusion bodies, affinity-purified via the His-tag, and refolded in vitro under controlled redox conditions to restore disulfide-bond pairing and enzymatic activity. Active ALP was combined with nano-hydroxyapatite (nHA) to form ALP-nHA composite nanoparticles, which were subsequently cloaked with cell-membrane fragments derived from healthy human mesenchymal stem cells (hMSCs) by extrusion through nanoporous membranes, yielding OsteoVes.

(C) HPLC chromatogram recorded at 220 nm showing a single dominant elution peak for refolded ALP.

(D and E) Enzymatic hydrolysis of ATP (D) and inorganic pyrophosphate (PPi) (E) by refolded active ALP compared with non-refolded, enzymatically inactive ALP (Inac-ALP).

(F) Representative scanning electron microscopy (SEM), transmission electron microscopy (TEM), and cryo-electron microscopy (cryo-EM) images of OsteoVes showing spherical morphology and a membrane shell surrounding the mineralization core. Scale bars, 200 (SEM) and 100 nm (TEM and cryo-EM).

(G) Fourier transform infrared (FTIR) spectra showing characteristic functional-group signatures of nHA, ALP, ALP-nHA, and OsteoVes.

(H and I) ATP (H) and PPi (I) hydrolysis by PBS control (Ctrl), nHA, free ALP, ALP-nHA nanoparticles, and OsteoVes.

(J–O) Serum-stability assay of OsteoVes in PBS containing 10% FBS over 0–72 h, showing ATP hydrolysis (J), PPi hydrolysis (K), zeta potential (L), hydrodynamic diameter (M), intravesicular ALP activity (N), and retention of His-tagged ALP cargo (O).

(P–R) Intravesicular mineralization in a Ca²⁺/PPi buffer system (Ca²⁺, 10 mmol/L; PPi, 3 mmol/L) for up to 7 days, visualized by cryo-EM (P) and quantified calcium-particle area and density using identical ImageJ settings (Q and R). Scale bars, 50 nm (P).

For (D), (E), and (H)–(R), n = 6 biological replicates. Data are mean ± SD. Statistical significance was assessed using an unpaired two-sided Student’s t test for indicated pairwise comparisons; p < 0.05 was considered significant.

Because ALP activity is a key bottleneck during the initiation of mineralization,⁷ we established a low-cost, scalable production pipeline for ALP to enable large-scale OsteoVes manufacturing (Figures 1B and S1A). Using a prokaryotic expression system, we engineered high-level expression of human ALPL (bone/liver/kidney-type ALP) and obtained concentrated ALP from inclusion bodies, which was purified via a C-terminal His tag to yield high-purity protein (Figure S2A). Given that ALP topology depends on three disulfide-bond pairs, we controlled the oxidized-to-reduced glutathione ratio to set the solution redox potential, promoting correct disulfide pairing, and applied a guanidine hydrochloride dilution strategy to refold secondary structure. High-performance liquid chromatography (HPLC) showed a single sharp peak, indicating largely completed refolding with a relatively uniform conformation (Figure 1C). Functional assays further confirmed robust enzymatic activity: the refolded ALP rapidly hydrolyzed ATP (Figure 1D) and PPi (Figure 1E), whereas the non-refolded ALP (inactive ALP [Inac-ALP]) showed no detectable activity (Figures 1D and 1E). Because bacterially produced ALPL lacks native eukaryotic N-glycosylation and glycosylphosphatidylinositol (GPI)-anchor modification, we next directly compared the catalytic and short-term stability profiles of the refolded prokaryotic ALP with commercially available mammalian-cell-expressed human ALP. In ATP and PPi hydrolysis assays, the refolded prokaryotic ALP showed robust substrate-consumption kinetics over the tested time course, broadly comparable to mammalian-cell-expressed ALP, whereas non-refolded ALP remained inactive (Figures 1D, 1E, and S2B). Stability assays further showed that prokaryotic ALP retained ATP/PPi hydrolytic activity and His-tag signal during the tested interval (Figure S2C). These data support the use of refolded recombinant ALP as the catalytic module of OsteoVes, while limiting our stability conclusion to the examined assay and formulation window.

With catalytically competent ALP in hand, we assembled OsteoVes as outlined in Figure 1B. ALP was first combined with nHA to form composite nanoparticles that serve as a stable nucleation/mineralization core. A cell-membrane interface was then introduced to reconstruct an MV-like membrane architecture with adhesion features, and extrusion through nanoporous membranes yielded uniform, membrane-enveloped vesicles that integrated an enzymatic module, a nucleation module, and a membrane-interface module into a single unit. Structural characterization showed that both OsteoVes and native MVs displayed spherical vesicular morphology (Figures 1F and S3A) with comparable size distribution and surface charge (Figures S3A and S3B). Cryo-electron microscopy further revealed a thin membrane encapsulating nHA particles (Figure 1F). Fourier transform infrared spectroscopy showed a characteristic phosphate peak for nHA and amide peaks for ALP; after encapsulation into OsteoVes, the corresponding peaks shifted/disappeared, consistent with successful packaging of the composite core into a vesicular structure (Figure 1G).

We next assessed the preservation of ALP activity within OsteoVes. In ATP and PPi hydrolysis challenges, free ALP, the ALP-nHA mixture, and OsteoVes showed comparable hydrolytic activity (Figures 1H and 1I), indicating that the fabrication process largely preserved catalytic function. We then tested stability in a protein-rich environment: in the presence of 10% fetal bovine serum (FBS), OsteoVes retained strong ATP and PPi hydrolysis activity (Figures 1J and 1K), and its particle size and zeta potential remained essentially unchanged over 72 h of incubation (Figures 1L, 1M, S3C, and S3D). ALP hydrolytic activity within OsteoVes (Figure 1N) and intravesicular ALP protein content (Figure 1O) were also stable over 72 h. Mineralization assays further showed that, when incubated in a Ca²⁺/PPi mineralization-mimicking system for 7 days, calcium-salt deposition within the vesicle lumen increased in both quantity and quality over time (Figures 1P–1R), demonstrating that OsteoVes can initiate mineralization even under PPi-containing, mineralization-inhibitory conditions. The intravesicular calcium-salt deposition observed in Figures 1P–1R should be interpreted as evidence that OsteoVes can function as a local mineral-nucleation microreactor under a PPi-containing mineralization-inhibited condition. We do not interpret this process as direct crystal-cell contact generating mechanical signaling. Rather, once OsteoVes localize at the ECM interface, their ALP-dependent PPi hydrolysis and nHA-assisted nucleation are expected to increase local mineralization permissiveness and progressively modify the adjacent physicochemical and adhesive ECM background, which can then be sensed by MSCs through mechano-adhesive pathways. Collectively, OsteoVes reconstructs MV-like membrane architecture while preserving and strengthening the threshold-tuning and nucleation capacities required for mineral initiation, providing a replaceable nanoscale functional unit for subsequent niche resetting.

OsteoVes restores Pi/PPi balance to re-bias aged hMSCs toward osteogenesis

MVs function primarily by localizing within the ECM and adhering near hMSCs and osteolineage cells; accordingly, their outer membrane is not merely a structural barrier but an essential interface that mediates ECM targeting, adhesion, and cue presentation.⁶ However, native MV membranes are produced in low yields and at high cost, which would severely limit scalable manufacturing if used directly. Given that MSCs inherently possess strong cell-cell and cell-matrix adhesion capabilities, we evaluated the cell membrane (hMSC-CM) derived from the bone marrow mesenchymal stem cells of healthy adults as a scalable substitute interface.²⁰ We conducted quantitative proteomics to compare human osteoblast-derived MV membranes (hOB-MVMs) and hMSC-CM (Figure 2A, left) and concentrated on ECM-adhesion-related proteins. The two exhibited a highly similar number and composition of adhesion proteins (Figure 2A, right; Figures S4A and S4B), which supports hMSC-CM as a viable replacement for hOB-MVMs. Because hMSC-CM is intended as a manufacturable membrane-interface substitute rather than a complete molecular replica of native osteoblast-derived MV membranes, we expanded the proteomic comparison beyond adhesion proteins. The revised analysis included proteins associated with ECM remodeling, collagen formation, mineralization, and protein adhesion, as well as ECM organization-related gene sets (Figures S4A and S4B). These data support substantial overlap in ECM-facing interface functions between hMSC-CM and hOB-MVMs, while also indicating that hMSC-CM should be interpreted as a scalable functional substitute for membrane-interface anchoring rather than a full replacement of all native MV membrane components. Consistently, ECM-adhesion assays showed that hMSC-CM-assembled OsteoVes efficiently localized to ECM surfaces (Figure 2B). To further support ECM-facing retention, we performed wash-retention confocal imaging and Imaris 3D reconstruction using PKH67-labeled OsteoVes; a parallel ErVes assay using erythrocyte membrane vesicles as a non-MSC membrane comparator is shown in Figure S4C. OsteoVes remained detectable on ECM surfaces at 12 and 24 h after washing, and quantitative analysis showed sustained vesicle-associated fluorescence and vesicle-F-actin/ECM colocalization (Figures 2B and S4D). These results provide evidence that the hMSC membrane-coated OsteoVes can localize to and remain associated with the ECM-facing interface.

Figure 2.

OsteoVes promotes osteogenesis while suppressing adipogenesis in aged hMSCs

(A) hMSCs from a healthy donor were induced to differentiate into osteoblasts (hOBs). The osteoblast-derived matrix vesicle membranes (hOB-MVMs) were then isolated and analyzed alongside hMSC membranes (hMSC-CMs) using quantitative proteomics. Comparative analyses were performed, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses and a heatmap of extracellular matrix (ECM) adhesion-related proteins.

(B) Confocal microscopy showing localization of PKH67-labeled OsteoVes (green) on ECM surfaces and association with hMSCs stained for F-actin (SF633-phalloidin, magenta) after 12 and 24 h; 2D views and Imaris 3D reconstructions are shown. Scale bars, 20 (2D) and 10 μm (3D).

(C–H) Osteogenic induction (OS) of bone marrow-derived MSCs isolated from a 62-year-old elderly female donor (aged hMSCs). Representative ALP staining on OS day 7 with quantification of ALP-positive area (C), ALP activity in conditioned medium on OS day 7 (D), and immunoblotting of osteogenic markers (ALP, RUNX2, and SP7) on OS day 7 (G). Mineralized matrix deposition on OS day 21 was assessed by alizarin red S (ARS) staining (E) and Von Kossa staining (F), with quantification of mineralized area shown at right. (H) Inorganic phosphate (Pi) and PPi concentrations in conditioned medium on OS day 7, and the corresponding unitless Pi/PPi ratio calculated after conversion to the same molar unit.

(I–K) Adipogenic induction (AD) of aged hMSCs. Immunoblotting of PPARG on AD day 7 (I), oil red O (ORO) staining at AD day 14 (J), and quantification of lipid-positive area (K). Scale bars, 200 μm (J).

(L–S) Component contribution analysis. Aged hMSCs were treated with intact OsteoVes (OV), ALP-deficient OsteoVes (OV^ΔALP), nHA-deficient OsteoVes (OV^ΔnHA), or ALP-nHA nanoparticles. On OS day 7, ALP staining (L) with quantification (M), ALP activity in conditioned medium (N), and immunoblotting of ALP, RUNX2, and SP7 with densitometry (S) were performed. On OS day 21, mineralization was evaluated by ARS staining (O) with quantification (P) and Von Kossa staining (Q) with quantification (R).

(T–W) RNA-seq analysis of aged hMSCs after 48 h of OsteoVes treatment: volcano plot of RNA-seq data showing the log10-transformed fold change in gene expression in OsteoVes-treated cells relative to control cells [log10(OsteoVes/Ctrl)] (T), heatmap of osteogenesis- and adipogenesis-related genes (U), and gene set enrichment analysis (GSEA) of osteogenic and adipogenic programs (V) and enrichment scores (W). OS, osteogenic induction; AD, adipogenic induction.

For (A) and (T)–(W), n = 3 biologically independent samples per group; for (B)–(F), (H), and (J)–(R), n = 6 biological replicates; for (G), (I), and (S), n = 3 biological replicates. Data are mean ± SD.

Statistical significance was assessed using an unpaired two-sided Student’s t test for indicated pairwise comparisons; p < 0.05 was considered significant.

To assess biological effects, we isolated bone marrow mesenchymal stem cells (hMSCs) from a 62-year-old female donor and induced osteogenic differentiation in vitro. Compared with controls, OsteoVes markedly increased ALP staining intensity and enzymatic activity (Figures 2C and 2D). Alizarin red S and Von Kossa staining further showed that OsteoVes exerted mineralization-promoting capacity comparable to that of hOB-MVs, substantially increasing mineralized matrix deposition (Figures 2E and 2F). OsteoVes also robustly upregulated osteogenic marker proteins in aged hMSCs, including ALP, RUNX2, and SP7, and its overall effect exceeded that of hOB-MVs (Figure 2G). In the extracellular chemical milieu, addition of OsteoVes or hOB-MVs significantly increased extracellular inorganic phosphate (Pi) and decreased PPi (Figure 2H), and OsteoVes outperformed hOB-MVs in PPi-to-Pi conversion efficiency (Figure 2H). These findings indicate that by strengthening ALP-mediated Pi/PPi metabolism, OsteoVes lowers the mineralization chemical threshold and substantially promotes osteogenic differentiation and mineralization execution in aged hMSCs.

To exclude the possibility that the hMSC membrane scaffold alone accounts for the observed lineage effects, we prepared empty hMSC membrane vesicles (eCMVs) and compared them with intact OsteoVes. Under standard growth or low-induction conditions, OsteoVes promoted RUNX2 nuclear localization and early osteogenic marker activation more effectively than eCMVs, supporting osteogenic priming beyond a nonspecific membrane-scaffold effect (Figures S4E–S4G). We further included a Pi control to determine whether chemical correction alone could recapitulate OsteoVes activity. Pi supplementation partially influenced osteogenic and adipogenic readouts but did not fully reproduce the effects of intact OsteoVes (Figures S4H–S4K). These results support the conclusion that Pi/PPi correction is important but not sufficient by itself; the integrated ALP-nHA core and ECM-facing membrane interface are required for full lineage execution.

Given that age-associated bone-fat imbalance is a hallmark of skeletal aging,^16,22 we further evaluated the lineage-resetting effect of OsteoVes on aged hMSCs in an adipogenic induction system. On day 7 of adipogenic induction, both hOB-MVs and OsteoVes reduced protein expression of the master adipogenic transcription factor PPARG, with OsteoVes showing stronger suppression (Figure 2I). By day 14, oil red O staining revealed reduced lipid droplet accumulation in both treatment groups, with OsteoVes yielding the greatest reduction in lipid droplet area (Figures 2J and 2K), indicating concurrent inhibition of adipogenic differentiation in aged hMSCs.

To delineate the contribution of key components, we generated ALP-deficient (OV^ΔALP) and nHA-deficient (OV^ΔnHA) derivatives and included ALP-nHA composite nanoparticles as a control. In these comparator experiments, vesicle-based groups were normalized by total protein equivalents, whereas ALP-nHA was used as a membrane-free catalytic/mineral core control. Therefore, ALP-nHA should be interpreted as a core-function comparator rather than a vesicle-number- or membrane-mass-matched control. On day 7 of osteogenic induction, ALP staining and activity assays showed that, relative to intact OsteoVes, OV^ΔALP, OV^ΔnHA, and ALP-nHA all displayed markedly reduced osteogenic promotion, with the greatest loss observed for OV^ΔALP (Figures 2L–2N). Concordantly, protein expression of ALP, RUNX2, and SP7 decreased in the same pattern (Figure 2S). On day 21, alizarin red S and Von Kossa staining further confirmed significant between-group differences in the formation of mineralized nodules (Figures 2O–2R). Together, these results indicate that ALP is the key active component driving the pro-osteogenic/pro-mineralization effects of OsteoVes, whereas nHA and the membrane architecture provide the necessary carrier context and microenvironment to enable stable, executable mineral initiation. Consistent with the osteogenic component analysis, adipogenic assays showed that intact OsteoVes most effectively reduced lipid accumulation and PPARG expression, whereas OV^ΔALP, OV^ΔnHA, and ALP-nHA controls showed weaker or incomplete effects (Figures S4L and S4M). These data further support a cooperative requirement for the ALP catalytic module, the nHA nucleation module, and the membrane-interface context.

To corroborate lineage bias at the transcriptional level, we performed RNA sequencing on OsteoVes-treated aged hMSCs. Osteogenic genes (RUNX2, ALPL, and COL1A1) were significantly upregulated, whereas adipogenic genes (PPARG and PLIN3) were significantly downregulated (Figures 2T and 2U). Gene set enrichment analysis (GSEA) further showed significant upregulation of osteogenesis-related programs, including “stem cell bone vs. adipose up,” “MSC differentiation,” and “RUNX2-regulated osteoblast differentiation,” while adipogenesis programs such as “adipogenesis by activated PPARG” were significantly suppressed (Figures 2V and 2W). Collectively, by augmenting ALP-mediated Pi/PPi metabolism and promoting mineralization execution, OsteoVes markedly enhances osteogenic differentiation and mineralization of aged hMSCs while inhibiting adipogenesis, resetting lineage balance from an adipogenesis-biased state back toward osteogenesis.

OsteoVes supports a mechanotransduction-associated ITG/FAK-PIEZO1-JNK/c-JUN-RUNX2 working model

To dissect upstream mechanisms, we conducted pathway enrichment analysis on transcriptomes of aged hMSCs with or without OsteoVes treatment. Differential gene enrichment indicated that OsteoVes-induced gene sets were significantly enriched in pathways tightly linked to mechanosensing, adhesion, and ECM remodeling, including “ITGA/B-FAK-RHO signaling,” “mechanical load activates signaling by PIEZO1 and integrins,” “ECM organization,” and “mesenchymal stem cell differentiation” (Figure 3A). GSEA further confirmed coordinated upregulation of these programs (Figure 3B), suggesting that OsteoVes may drive osteogenesis by repairing the extracellular mechanical information background and activating mechanotransduction.

Figure 3.

OsteoVes supports osteogenic commitment through an ITG/FAK-PIEZO1-JNK-RUNX2 axis

(A and B) Pathway enrichment analyses based on RNA-seq of aged hMSCs treated with OsteoVes versus PBS control (Ctrl), presented as a Sankey/bubble plot of enriched pathways (A) and representative GSEA enrichment plots (B).

(C and D) Atomic force microscopy (AFM; PeakForce tapping mode) of aged hMSCs showing cell topography (C) and maximal cell height (D).

(E) AFM analysis of aged hMSCs treated with or without OsteoVes, alongside young hMSCs, showing viscoelastic mechanical parameters (E; Young’s modulus, total stiffness [E_sum], relaxation time [τ], and viscosity coefficient [η]).

(F and G) Immunoblotting (F) and densitometric quantification (G) of adhesion and mechanosensing pathway components, including ITGA5, ITGB1, FAK, ROCK2, active RHOA (ac-RHOA), PIEZO1, CAMK2, and c-JUN.

(H) Immunofluorescence analysis of RUNX2 nuclear translocation with quantification (nuclei counterstained with DAPI). Scale bars, 50 μm.

(I) Nuclear/cytoplasmic fractionation followed by immunoblotting to validate RUNX2 nuclear translocation (LAMIN A/C, nuclear marker; β-ACTIN, cytosolic marker).

(J) RUNX2 luciferase reporter assay to assess transcriptional activity.

(K) Working model in which OsteoVes provides an ECM-targeting membrane interface that co-activates ITG/FAK adhesion signaling and PIEZO1/Ca²⁺-dependent mechanotransduction, converging on JNK/c-JUN-associated RUNX2 activation and osteogenic transcription.

For (A) and (B), n = 3 biologically independent samples per group; for (C)–(E), (H), and (J), n = 6 biological replicates; for (F), (G), and (I), n = 3 biological replicates. Data are mean ± SD. Statistical significance was assessed using an unpaired two-sided Student’s t test for indicated pairwise comparisons; p < 0.05 was considered significant.

We next quantified the mechanical phenotype of aged hMSCs using atomic force microscopy (AFM), with young hMSCs included as a reference. PeakForce tapping topographical scans revealed increased cell height after OsteoVes treatment (Figures 3C and 3D). Viscoelastic measurements further showed that OsteoVes shifted the mechanical profile of aged hMSCs toward the young hMSC range, including increases in Young’s modulus, total stiffness, relaxation time, and viscosity coefficient (Figures 3E and S5A). Because these changes approached a young-like mechanical range rather than exceeding it in an uncontrolled manner, we interpret them as restorative cytoskeletal and mechano-adhesive remodeling associated with osteogenic priming, rather than pathological rigidity or accelerated aging.

At the molecular level, immunoblotting showed increased cell-associated abundance of ITGA5 and ITGB1 after OsteoVes treatment, together with increased abundance or activation of mechanosensing-associated nodes including PIEZO1, CAMK2, FAK, ROCK2, active RHOA, and c-JUN (Figures 3F and 3G). Because OsteoVes membranes themselves may carry membrane-associated adhesion proteins, we do not interpret the ITGA5/ITGB1 immunoblot signal as unequivocal evidence of endogenous integrin synthesis. The empty membrane vesicle control further helped define the contribution of the membrane scaffold relative to intact OsteoVes (Figure S5B). Immunofluorescence and nuclear/cytoplasmic fractionation immunoblotting both showed that OsteoVes markedly promoted RUNX2 translocation from the cytoplasm to the nucleus (Figures 3H, 3I, and S5C), and a RUNX2-promoter-driven luciferase reporter assay confirmed significantly increased RUNX2 transcriptional activity (Figure 3J).

To test a key mechanosensitive node in this model, we inhibited PIEZO1 using GsMTx4 during OsteoVes treatment. PIEZO1 inhibition attenuated OsteoVes-induced ALP staining, reduced osteogenic marker induction, and decreased RUNX2 nuclear translocation as assessed by both immunofluorescence and nuclear/cytoplasmic fractionation (Figures S5D–S5H). In addition, chromatin immunoprecipitation (ChIP-qPCR) showed increased c-JUN enrichment at the RUNX2 promoter after OsteoVes treatment (Figure S5I), supporting a direct regulatory link between the JNK/c-JUN node and RUNX2 transcriptional activation. Together, these perturbation and promoter-occupancy data strengthen key node validation of the proposed mechanotransduction framework.

We emphasize that the measured evidence includes pathway enrichment, AFM-defined mechanical changes, increased cell-associated mechanosensing proteins, PIEZO1 inhibition sensitivity, c-JUN occupancy at the RUNX2 promoter, RUNX2 nuclear translocation, and RUNX2 reporter activation. The precise temporal ordering among ITG/FAK, PIEZO1, JNK/c-JUN, and RUNX2 remains a proposed convergence model rather than a fully perturbation-mapped linear cascade. Integrating these observations, we connect chemical threshold repair to mechanically executable transduction as a concrete fate-unlocking cascade. By catalytically hydrolyzing PPi and elevating the Pi/PPi ratio, OsteoVes relieves mineralization inhibition and sustains local mineral initiation, enabling the externally provided mineralizable chemical condition to be persistently written into the ECM. In parallel, the OsteoVes membrane interface promotes ECM localization and strengthens cell-matrix adhesion, reinforcing ITGA5/ITGB1-mediated focal adhesion signaling and triggering FAK activation. FAK then drives the RHOA-ROCK2 axis to increase cytoskeletal tension and focal adhesion maturation. Concurrently, PIEZO1 senses changes in membrane tension/mechanical environment and mediates Ca²⁺ influx, activating Ca²⁺-dependent signaling such as CAMK2. These pathways converge at the focal adhesion-cytoskeleton layer to promote RUNX2 nuclear translocation and transcriptional activation, thereby upregulating osteogenic programs (ALPL, SP7, and osteocalcin [OCN]) and suppressing adipogenic programs, achieving osteogenic fate unlocking that is mechanically perceivable and transcriptionally executable (Figure 3K).

Systemic OsteoVes improves microarchitecture and mechanics in naturally aged mouse bone within a short-term observation window

To validate in vivo efficacy, we performed systemic administration in naturally aged mice. Twenty-two-month-old mice received tail-vein injections of OsteoVes (100 μg per 20 g body weight) at week 0 and week 1 (two doses total), and femora were harvested at week 4 for multidimensional analyses (Figure 4A). To clarify systemic distribution after intravenous administration, we tracked DiD-labeled OsteoVes in vivo and ex vivo across major organs and skeletal tissues. Fluorescence imaging showed early distribution to clearance-related organs, together with detectable bone-associated signal in femur and vertebra at selected time points (Figures S6A and S6B). We therefore interpret systemic OsteoVes delivery as producing a biodistribution profile with bone-associated retention, rather than absolute bone-specific targeting. OsteoVes-derived His-tag signal was readily detectable in the bone marrow cavity after treatment, whereas the corresponding signal in control samples was near background (Figure S6C), indicating local marrow-associated deposition of the His-tagged ALP cargo within the tested window.

Figure 4.

Systemic administration of OsteoVes improves bone microarchitecture and mechanical properties in naturally aged mice

(A) Treatment schedule for 22-month-old naturally aged mice. OsteoVes was administered by tail-vein injection (100 μg per 20 g body weight) at weeks 0 and 1 (two injections total), and femora were harvested at week 4.

(B–D) Integrated proteomic and transcriptomic analyses of femora 4 weeks after OsteoVes or saline control (Ctrl) treatment, including a proteome-transcriptome concordance plot (B), pathway category enrichment (C), and enrichment score comparisons for osteogenesis- and adipogenesis-related gene sets (D).

(E and F) AFM PeakForce tapping analysis of frozen femoral sections to assess bone-surface topography and maximal height (E) and quantify viscoelastic mechanical parameters of bone tissue (F).

(G and H) Proteomics-based GSEA showing changes in pathways related to bone mineralization, bone morphogenesis, bone development, and extracellular matrix (ECM) organization (Sankey/bubble plot in G and enrichment plots in H).

(I and J) RNA-seq-based GSEA showing analogous pathway changes (Sankey/bubble plot in I and enrichment plots in J).

(K) Representative micro-CT reconstructions and quantification of trabecular parameters, including bone mineral density (BMD), bone volume fraction (BV/TV), trabecular number (Tb.N), and trabecular separation (Tb.Sp), in young mice, aged mice treated with saline (Ctrl), and aged mice treated with OsteoVes.

(L) Masson’s trichrome staining of femora and histomorphometric quantification of trabecular area.

(M) Oil red O staining and quantification of lipid-positive area within the marrow cavity.

(N) Immunofluorescence staining of osteocalcin (OCN) and FABP4 with quantification (nuclei counterstained with DAPI). OBs/BS denotes the density of OCN⁺ osteoblasts on trabecular bone surfaces; N.AdCs/Ar denotes the density of FABP4⁺ adipocytes within the marrow area.

(O) Three-point bending tests to evaluate femoral mechanical properties with quantification. Scale bars are indicated in the images, including 1 mm for low-magnification micro-CT/histology images, 400 μm for histological insets, 500 μm for oil red O staining, and 50 μm for immunofluorescence.

For (B)–(D) and (G)–(J), n = 3 mice per group; for (E), (F), and (K)–(O), n = 6 mice per group.

Data are mean ± SD. Statistical significance was assessed using an unpaired two-sided Student’s t test (saline versus OsteoVes, as indicated); p < 0.05 was considered significant.

Integrated femoral proteomic and transcriptomic analyses showed broad upregulation of ECM/focal-adhesion and osteogenesis-associated factors after OsteoVes treatment, including ITGA5/ITGB1, COL1A1/COL1A2, and SPARC, whereas adipogenesis-associated genes such as PPARG, FABP4, and LPL were significantly downregulated (Figure 4B). Mechanotransduction-osteogenesis programs, including integrin/FAK, RHOA-ROCK, JNK-c-JUN, and the RUNX2 program, were enriched, while adipogenesis gene sets were depleted (Figures 4C and 4D). GSEA of both proteomic and transcriptomic datasets indicated systematic activation of osteogenic processes, including “bone development,” “bone morphogenesis,” “ossification,” and “bone mineralization” (Figures 4G–4J).

AFM topography scanning revealed a ∼91% increase in relative surface height peaks in the OsteoVes group (Figure 4E). Viscoelastic parameter analysis showed substantial enhancement: Young’s modulus increased by ∼2.6-fold, E_sum by ∼2.9-fold, τ by ∼3.8-fold, η by ∼7.0-fold, and E₁ and E₂ by ∼4.2-fold and ∼5.1-fold, respectively (Figure 4F). AFM-derived local viscoelastic parameters and whole-femur three-point bending results were directionally concordant but differed in magnitude, as expected. AFM is sensitive to local mineralized surface features and viscoelastic relaxation, whereas three-point bending integrates whole-bone geometry, cortical-trabecular organization, and pre-existing age-related structural deficits. Therefore, these mechanical readouts should not be interpreted as linearly interchangeable. Three-point bending tests showed significant increases in maximal load (∼1.4-fold), bending strength (∼95% increase), and elastic modulus (∼98% increase) (Figure 4O), validating improved bone quality at the macroscopic mechanical level.

Micro-computed tomography (CT) showed improved trabecular microarchitecture within the predefined trabecular region of interest (ROI), including higher bone mineral density (BMD), bone volume fraction (BV/TV), and trabecular number (Tb.N) and lower trabecular separation (Tb.Sp) in OsteoVes-treated aged mice than in saline-treated aged mice (Figure 4K). Because the aged saline group showed a severely compromised low-baseline trabecular phenotype, these changes are presented and interpreted primarily using absolute parameter values rather than fold changes alone. Masson’s trichrome and H&E staining showed a marked increase in trabecular/collagen-associated area (∼70% increase) (Figures 4L and S6D), whereas oil red O staining revealed a strong reduction in marrow lipid droplet area (∼79% decrease) (Figure 4M). Immunofluorescence showed increased OCN⁺ osteoblast numbers (∼2.0-fold) and decreased FABP4⁺ adipocyte numbers (∼61% decrease) (Figure 4N), accompanied by enhanced ALP⁺ and RUNX2⁺ signals (∼2.0-fold and ∼5.0-fold, respectively) (Figures S6E and S6F). Because adipocyte fate tracing and apoptosis assays were not performed, we interpret the reduction in marrow lipid area and FABP4-positive adipocyte density as consistent with reduced adipogenic allocation under a pro-osteogenic niche context, rather than as direct evidence that OsteoVes eliminates pre-existing mature adipocytes. These results support that OsteoVes synchronously repairs chemo-mechanical coupling and drives niche resetting at the tissue scale.

For short-term safety evaluation, we assessed hematology, serum biochemistry, major-organ histology, and ectopic mineral deposition. One month after treatment, complete blood counts showed no significant abnormalities (Figure S7A); serum biochemistry did not indicate overt hepatic, renal, or myocardial injury (Figure S7B); and H&E staining of major organs did not reveal overt pathological changes (Figure S7C). In addition, alizarin red S and Von Kossa staining of heart, liver, and kidney did not show obvious ectopic mineral deposition within the 4-week observation window (Figures S7D–S7G). These results support preliminary short-term systemic safety in mice, while longer term safety and repeated-dose evaluation remain necessary.

OsteoVes rescues chemo-mechanical defects in Alpl^+/− progeroid bone aging

Bone aging is not driven solely by chronological time; numerous pathological states can also induce accelerated or aberrant skeletal aging, which superimposes on aging to create more complex niche disequilibrium. Deficiency or dysregulated expression of ALP is a representative molecular feature of pathological skeletal aging and mineralization disorders, directly elevating the mineralization chemical threshold and precipitating degradation of bone matrix and mechanical cue presentation.^8,24 We therefore used Alpl heterozygous (Alpl^+/−) mice to establish a pathological progeroid bone aging model,²⁴ challenging the universality and robustness of OsteoVes.

Eight-month-old Alpl^+/− mice received tail-vein injections of OsteoVes (100 μg per 20 g body weight) at week 0 and week 1 (two doses total), and tissues were collected at week 4 (Figure S8A). AFM analyses showed that OsteoVes increased bone surface height (∼1.1-fold) and markedly improved viscoelastic parameters: Young’s modulus increased by ∼2.1-fold, E_sum by ∼2.6-fold, τ by ∼2.8-fold, η by ∼6.8-fold, and E₁ and E₂ by ∼6.0-fold and ∼3.3-fold, respectively (Figures 5A and 5B). Micro-CT revealed substantial microstructural rescue: BMD increased by ∼1.5-fold, BV/TV by ∼5.1-fold, Tb.N by ∼4.2-fold, and Tb.Sp decreased by ∼83% (Figure 5C). Histology showed increased trabecular area (∼1.2-fold; Figures 5D and S8B) and reduced marrow lipid droplet area (∼77% decrease; Figure 5E). Immunofluorescence demonstrated that OsteoVes-derived His-tag signal was readily detectable in treated Alpl^+/− marrow, whereas control sections showed near-background signal (Figure 5G). ALP and RUNX2 signals were also increased after treatment (Figures 5H and 5I), supporting local delivery of the ALP-containing cargo and activation of osteogenic programs within the tested window. Three-point bending tests also showed significant improvement of mechanical performance (Figure 5J). Short-term safety assessments 1 month after treatment showed no significant abnormalities in complete blood counts or serum biochemistry (Figures S8C and S8D), and H&E staining of major organs did not reveal overt inflammatory infiltration, necrosis, fibrosis, or architectural disruption (Figure S8E). Alizarin red S and Von Kossa staining of heart, liver, and kidney further showed no obvious ectopic mineral deposition within the tested window (Figures S8F–S8I).

Figure 5.

OsteoVes alleviates bone regeneration defects in Alpl^+/− mice

(A and B) AFM PeakForce tapping analysis of frozen femoral sections from Alpl^+/− mice after systemic OsteoVes or saline control (Ctrl) treatment for 4 weeks, including bone-surface topography/maximal height (A) and viscoelastic mechanical parameters (B).

(C) Representative micro-CT reconstructions and quantification of trabecular parameters (BMD, BV/TV, Tb.N, and Tb.Sp) in Alpl^+/+ mice, Alpl^+/− mice treated with saline (Ctrl), and Alpl^+/− mice treated with OsteoVes.

(D) Masson’s trichrome staining and histomorphometric quantification of trabecular area.

(E) Oil red O staining and quantification of lipid-positive area.

(F) Immunofluorescence staining of OCN and FABP4 with quantification (nuclei counterstained with DAPI; OBs/BS and N.AdCs/Ar are defined as in Figure 4).

(G–I) Immunofluorescence staining and quantification of OsteoVes-derived His-tag (G), ALP (H), and RUNX2 (I) in femoral sections (nuclei counterstained with DAPI).

(J) Three-point bending tests to evaluate femoral mechanical properties with quantification. Scale bars are indicated in the images, including 1 mm for low-magnification micro-CT/histology images, 400 μm for Masson’s trichrome insets, 500 μm for oil red O staining, 50 μm for OCN/FABP4 immunofluorescence, and 100 μm for high-magnification His-tag/ALP/RUNX2 immunofluorescence.

For (A)–(J), n = 6 mice per group. Data are mean ± SD. Statistical significance was assessed using an unpaired two-sided Student’s t test (saline versus OsteoVes, as indicated); p < 0.05 was considered significant.

Thus, systemic OsteoVes administration not only improves natural aging-associated skeletal phenotypes but also synchronously enhances microstructure, reduces marrow adiposity, and enhances mechanical competence in an ALP-deficient pathological progeroid model, supporting cross-etiology osteo-regenerative potential with favorable short-term safety.

OsteoVes elicits short-term osteoanabolic signals in a metabolic-abnormality-associated rhesus macaque bone deterioration model

To directly test translational relevance in a clinically authentic human substrate, we isolated bone marrow MSCs from five elderly osteopenic/osteoporotic patients undergoing fracture surgery and assessed whether OsteoVes could support osteogenic output in this clinical substrate. The 62-year-old aged human MSC preparation used earlier served as a defined mechanistic model, whereas this independent patient-derived hMSC cohort was used to evaluate reproducibility across heterogeneous clinical samples. The young donor-derived hMSC preparation was included as a reference comparator, and patient-derived MSC readouts were analyzed as measured values for each donor-derived biological sample rather than being normalized to the young donor. Under osteogenic induction, OsteoVes-treated patient-derived MSCs showed day-7 ALP staining and conditioned-medium ALP activity across all five patient samples, with values presented alongside the young donor-derived reference preparation (Figures 6A–6C). By day 21, alizarin red S staining showed substantial mineralized matrix deposition, with OsteoVes-treated aged/osteoporotic MSCs approaching the osteogenic output range of young donor MSCs (Figures 6D and 6E). These data indicate that OsteoVes can support osteogenic competence in clinically relevant elderly hMSCs, thereby bridging our mechanistic findings in aged hMSCs to a patient-derived disease context.

Figure 6.

OsteoVes supports osteogenic output in patient-derived elderly osteopenic/osteoporotic hMSCs and elicits osteoanabolic responses in metabolically aged adult female rhesus macaques

(A–E) Patient-derived human model. Bone marrow MSCs isolated from five elderly osteopenic/osteoporotic patients undergoing surgery for fractures were treated with OsteoVes and cultured under osteogenic induction, with one young healthy donor-derived MSC preparation included as a reference comparator. Representative ALP staining on OS day 7 (A), quantification of ALP-positive area (B), ALP activity in conditioned medium on OS day 7 (C), representative alizarin red S (ARS) staining on OS day 21 (D), and quantification of mineralized area (E).

(F–I) Non-human primate (NHP) model. Adult female rhesus macaques (13 years old) were rendered metabolically aged by a high-fat diet, high-sugar drinking water, and two cephalic-vein injections of streptozotocin (STZ; 30 mg/kg) administered 1 month apart and were then evaluated in a self-controlled longitudinal design before and after intravenous OsteoVes administration (5 mg/kg at weeks 0 and 1). Experimental scheme and sampling schedule (F), serum calcium (Ca²⁺) and inorganic phosphate (Pi) concentrations over time (G), longitudinal ELISA quantification of bone-specific ALP (bALP), osteocalcin (OCN), TGF-β1, and C-terminal telopeptide of type I collagen (CTX-1) (H), representative in vivo micro-CT reconstructions of bilateral femora at pretreatment, 4 and 12 weeks after OsteoVes administration (I). Scale bars, 10 mm.

For (A)–(E), analyses were performed using six donor-derived biological samples (one reference sample from a young donor and five independent samples from elderly osteopenic/osteoporotic patients). Statistical significance was determined by an unpaired two-sided Student’s t test. For (G)–(I), n = 3 rhesus macaques were included. Statistical significance was evaluated as described in the STAR Methods section, with longitudinal analyses in rhesus macaques conducted using paired two-sided Student’s t test; p < 0.05 was considered statistically significant.

We next extended validation from patient-derived human cells to a non-human primate (NHP) model of metabolic-abnormality-associated bone deterioration. Because hyperglycemia and dyslipidemia commonly coexist with aging and can aggravate bone fragility and bone quality deterioration, we generated a metabolic comorbidity-associated bone deterioration model in adult female rhesus macaques using high-sugar/high-fat dietary intervention combined with streptozotocin (STZ) administration.^9,10 Following two STZ injections and 6 months of dietary intervention, fasting glucose and circulating lipid levels remained persistently elevated. By month 6, fasting glucose had increased 1.6-fold, total cholesterol by 82%, and triglycerides by 9%, accompanied by an ∼89% rise in serum C-terminal telopeptide of type I collagen (CTX-1), an established marker of osteoclast activity, indicating successful establishment of the model (Figure S10A).

We used a self-controlled study design (Figure 6F). A saline phase served as baseline monitoring, and initial serum parameters (Ca²⁺, Pi, bone-specific ALP [bALP], OCN, TGF-β1, and CTX-1) were collected at 4 weeks. When serial monitoring showed no appreciable fluctuations, indicating relative equilibrium of bone metabolism, this time point was designated as the pretreatment control (0 weeks), at which live micro-CT was performed to obtain baseline images and bone parameters. OsteoVes was then administered via cephalic vein injection (5 mg/kg) at week 0 and week 1 (two doses total). Blood samples were collected every 2 weeks, and micro-CT evaluations were conducted 4 and 12 weeks post-treatment (Figure 6F). The 5 mg/kg dose was used as an exploratory proof-of-concept dose to evaluate biological activity and preliminary safety in a large-animal setting, rather than as a clinically optimized dose derived from formal cross-species pharmacokinetic or dose-ranging studies.

At 4 weeks post-treatment, serum Ca²⁺ and Pi transiently increased, with Ca²⁺ reaching approximately 67% above the pretreatment level and Pi reaching approximately 2.3-fold of the pretreatment level (Figure 6G). Importantly, extended longitudinal monitoring to 3 months showed that both mineral ions declined after the early post-treatment peak: serum Ca²⁺ decreased to approximately 24% above the pretreatment level, and serum Pi declined to approximately 1.1-fold of the pretreatment level by month 3 (Figure 6G). These data indicate that the early Pi increase was not maintained as a continuously rising or persistently high-phosphate state within the 3-month observation window. We therefore interpret the Pi elevation as a transient systemic mineral-metabolism response associated with OsteoVes-induced osteoanabolic remodeling, rather than sustained phosphate-homeostasis disruption. Osteogenesis-associated markers rose substantially: bALP increased by ∼1.9-fold, OCN by ∼40%, and TGF-β1 by ∼79% (Figure 6H), whereas the bone resorption marker CTX-1 decreased by ∼83% (Figure 6H), suggesting a net shift toward formation. Imaging analyses of bilateral femora from three rhesus macaques showed overall improvement in trabecular parameters: BMD increased by ∼4%, BV/TV by 10%, bone surface area (BS)/bone volume (BV) decreased by 21%, Tb.N increased by 24%, Tb.Th increased by 36%, and Tb.Sp decreased by 18% (Figures 6I and S9A). To extend the imaging observation window beyond the initial 4-week endpoint, we further performed paired micro-CT analysis at 12 weeks after OsteoVes administration. The paired longitudinal analysis showed that several trabecular parameters observed at 4 weeks were maintained or further improved at 12 weeks, including BV/TV, Tb.N, Tb.Th, and Tb.Sp trends, although the small n and absence of a parallel saline-control arm require cautious interpretation (Figures 6I and S9B). These data provide extended imaging follow-up but do not replace formal long-term efficacy, repeated-dose safety, or controlled NHP studies.

For safety, complete blood counts and serum biochemistry at 4 weeks showed no significant differences from baseline (Figures S10B and S10C), and ultrasonography detected no abnormalities in major organs (Figure S10D). Together, these human and NHP data extend OsteoVes validation beyond rodent aging models: OsteoVes supports osteogenic output in patient-derived elderly osteopenic/osteoporotic hMSCs and, under systemic dosing, induces osteoanabolic biochemical and structural responses without evident short-term toxicity in metabolically aged rhesus macaques. This establishes a translational bridge from mechanism to clinically relevant human substrate and large-animal physiology.

Discussion

In this study, we frame skeletal aging as an operational steady-state defect in which mineralization chemistry and matrix mechanical information deteriorate together. The present data support short-term functional rescue of this coupled niche defect after OsteoVes treatment, rather than definitive long-term rejuvenation of the entire skeletal system. We therefore interpret OsteoVes as a niche-directed functional-unit replacement strategy that partially restores extracellular Pi/PPi handling, mineral nucleation, and mechanosensitive osteogenic execution within the observation windows tested. We further identify MVs as a replaceable extracellular control point within this loop: decline of MV ALP-related function elevates the Pi/PPi mineralization threshold, weakens mineralization execution, remodels mechanical cue presentation, and stabilizes fate locking.^5,6,7,8,24 OsteoVes was engineered to intervene at this spatial scale by restoring threshold tuning, nucleation, and interface anchoring within a single functional unit.

Notably, we obtained convergent evidence across complementary models, strengthening both etiological coverage and translational relevance. Naturally aged mice captured canonical chronological skeletal aging, Alpl^+/− mice modeled ALP-defect-centric mineralization failure, and the diet/STZ rhesus macaque model represented pathology-age composite bone aging driven by metabolic comorbidity.^9,10,15,24 Data from elderly osteopenic/osteoporotic fracture patient-derived MSCs further bridge mechanism to a clinically authentic human substrate. In the patient-derived human cells, OsteoVes-treated samples showed ALP activity and matrix mineralization approaching the young reference range, whereas in vivo, it improved trabecular architecture, marrow adiposity, and mechanical or osteoanabolic readouts across species. The reduction in marrow adiposity is most consistently interpreted as reduced adipogenic allocation from MSC-lineage cells under a pro-osteogenic niche context. Because we did not perform adipocyte fate tracing, dedifferentiation assays, or apoptosis assays in mature marrow adipocytes, we do not conclude that OsteoVes directly eliminates pre-existing adipocytes. Together, these findings support the robustness of the functional-unit replacement-niche resetting logic across different etiologies and biological scales.

Mechanistically, our data support a chain model for how aging phenotypes become self-maintaining. First, as local reactors for mineral initiation, MVs set the mineralization threshold through their membrane-associated enzyme system, especially TNAP/ALP, by regulating Pi/PPi.^5,6,7,8 When aging or ALPL insufficiency reduces ALP-related activity, PPi-mediated inhibition strengthens, mineral initiation is blocked, and newly formed matrix fails to mineralize efficiently.^7,8,24 Second, insufficient mineralization alters matrix physicochemistry and mechanical cue presentation, chronically exposing MSCs to an adhesion and mechanotransduction context unfavorable for osteogenesis.^1,2,3,4 Third, fate-constrained MSCs fail to generate high-quality osteoblasts and MV output, thereby amplifying the functional-unit deficit and closing a positive-feedback loop.^12,15,16,17 Thus, intravesicular mineral formation is best viewed as an upstream extracellular mineralization permissiveness event rather than a direct mechanical stimulus. The mechanotransduction response likely arises after OsteoVes remodel the ECM-facing chemical and adhesive context.

The efficacy of OsteoVes arises from refilling this deficit at the correct scale: ALP-mediated threshold repair and nHA nucleation restore mineralization execution, while membrane-interface anchoring enhances ECM localization and cue presentation, activates ITG/FAK-PIEZO1-JNK/c-JUN signaling, and promotes RUNX2 nuclear translocation and transcriptional activation. The sustained osteogenic tissue readouts observed after the dosing period are best interpreted through a trigger-reshaping model. OsteoVes does not need to remain indefinitely within the tissue to account for later RUNX2/ALP and matrix-remodeling signals. Instead, early extracellular correction of the Pi/PPi threshold, local mineral-nucleation permissiveness, and ECM-facing adhesion cues may initiate endogenous MSC/osteoblast-lineage programs that continue after most vesicles have been cleared. This interpretation is consistent with the biodistribution data and avoids assuming long-term particle residence.

From a therapeutic standpoint, this strategy is better viewed as niche-directed functional-unit replacement than as a stronger pro-mineralization drug or a miniaturized cell therapy. Compared with simple protein or enzyme supplementation, MV-mimetic vesicles couple catalysis, nucleation, and matrix interaction within a single nanoscale architecture, more closely resembling an in vivo-operable mineralization microreactor.^{5,6,7,8,21,23} Compared with cell transplantation, acellular vesicle-like formulations can, in principle, provide clearer compositional boundaries and more flexible dosing, although rigorous manufacturing and characterization frameworks remain essential for clinical translation.^21,23,25 Our multi-scale mechanical readouts—including AFM of cells and bone tissue, together with three-point bending—also provide quantitative functional metrics for future release criteria centered on ALP activity, mineral initiation capacity, and ECM anchoring/adhesion.

Several aspects of OsteoVes strengthen its translational potential. First, we demonstrate direct activity in a clinically relevant human substrate: bone marrow MSCs isolated from elderly osteopenic/osteoporotic fracture patients showed OsteoVes-supported ALP activity and matrix mineralization. We deliberately frame this human validation in terms of functional rescue, rather than assuming that all osteopenic/osteoporotic patient-derived MSCs are uniformly defective, because the clinical literature supports disease-associated heterogeneity, with altered transcriptomic states and impaired mechanosensitive or matrix-forming behavior reported in different cohorts.^18,19 Second, OsteoVes is designed to rehabilitate the endogenous aged MSC pool—arguably the most clinically relevant target in older patients—rather than depending on the availability and performance of ex vivo-expanded MSC therapies.^16,17,26 Third, we assembled a preclinical evidence ladder spanning natural aging, ALP-insufficient progeroid bone aging, and metabolically aged rhesus macaques, thereby providing cross-etiology and cross-species signals of efficacy and short-term safety.^9,10,15,24 Finally, the key structural elements of OsteoVes were rebuilt for scalable manufacturing: ALPL is produced by prokaryotic expression followed by controlled refolding, and an MSC membrane interface substitutes for low-yield osteoblast-derived MV membranes. Together, these features lower barriers to standardization, batch consistency, and cost-conscious scale-up.

From a translational perspective, OsteoVes may be most relevant for bone microenvironments in which extracellular mineralization chemistry and mechanosensitive matrix signaling are jointly impaired. Candidate contexts include aged or osteoporotic marrow niches with reduced ALP activity, elevated PPi or disturbed Pi/PPi balance, low mineralization competence, increased marrow adiposity, and impaired mechanosensing. Pathology-age composite settings, such as metabolic bone deterioration associated with hyperglycemia or dyslipidemia, may also be relevant, although these should not be considered equivalent to natural chronological aging. Future patient stratification could incorporate baseline ALP activity; Pi/PPi status; marrow adiposity; mechanosensitivity markers such as ITGA5/ITGB1, PIEZO1, and RUNX2 nuclear responsiveness; and ex vivo OsteoVes responsiveness of patient-derived bone marrow MSCs.

Overall, by linking extracellular phosphate metabolism to matrix mechanics and validating functional responses in patient-derived hMSCs and metabolically aged rhesus macaques, this work shifts replacement therapy from substituting a single molecule to restoring a spatially executable coupled function. Functional MV replacement therefore offers a pragmatic osteometabolic route toward clinically translatable rejuvenation of skeletal aging-associated niche defects, particularly in pathology-age composite bone deterioration associated with metabolic disease.

Limitations of the study

This study has several limitations. First, the patient-derived MSC cohort and the rhesus macaque cohort are small, and the NHP study uses a self-controlled longitudinal design without an independent saline-treated parallel arm. Second, the mouse studies mainly evaluate a 4-week post-treatment window, and the NHP biochemical and safety analyses remain short term, although paired imaging follow-up extends to 12 weeks. Third, the NHP dose is exploratory and was not selected through formal pharmacokinetic, biodistribution, or dose-escalation studies. Fourth, although short-term hematology, serum biochemistry, histology, ultrasonography, and ectopic mineralization staining do not reveal overt toxicity, these analyses do not exclude delayed toxicity, immune cytokine responses, anti-vesicle antibody formation, ectopic mineralization after repeated dosing, or bone-overgrowth risk. Fifth, the proposed ITG/FAK-PIEZO1-JNK/c-JUN-RUNX2 mechanism is supported by pathway enrichment, mechanical measurements, PIEZO1 inhibition, c-JUN promoter occupancy, RUNX2 nuclear translocation, and reporter assays, but the precise temporal order of these nodes remains to be fully mapped. Finally, applicability across patients with severe metabolic disease, polypharmacy, or diverse osteoporosis etiologies requires larger, controlled, dose-ranging, and longer term studies.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Wenjia Liu (wenjialiu@xjtu.edu.cn).

Materials availability

OsteoVes and related reagents generated in this study will be made available by the lead contact upon reasonable request, subject to institutional approvals and completion of a materials transfer agreement if applicable. This study did not generate stable cell lines or mouse lines intended for unrestricted distribution.

Data and code availability

• •The RNA sequencing (RNA-seq) datasets reported in this study have been deposited at the Sequence Read Archive under accession number BioProject: PRJNA1469690 (https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1469690). The proteomics datasets have been deposited in iProX under accession number IPX0017448000 (https://www.iprox.cn/page/project.html?id=IPX0017448000). All datasets are publicly available as of the date of publication. Accession numbers are listed in the key resources table.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Acknowledgments

This work was supported by the National High Level Talents Special Support Plan (W.L.), the “Young Talent Support Plan” of Xi’an Jiao Tong University (W.H. and W.L.), Innovation Team of Shaanxi Province Health Commission (2024SF2-GJHX-41), and Shaanxi Key R&D Program (2024TD-13). We thank the Instrument Analysis Center of Xi’an Jiaotong University for their assistance with SEM, TEM, cryo-EM, and LC-MS. We thank the Laboratory Animal Center of Xi’an Jiaotong University for their assistance in blood glucose-related testing and animal feeding.

Author contributions

L.Z., X.W., and K.Z. contributed equally to this work. L.Z. contributed to the conception of the OsteoVes design, recombinant ALP preparation, OsteoVes fabrication, physicochemical characterization, and in vitro functional assays. X.W. contributed to animal experiments, histological analyses, micro-CT analyses, AFM measurements, and systemic safety evaluation. K.Z. contributed to human specimen processing, patient-derived hMSC validation, omics analyses, statistical analysis, and figure preparation. F.D., Z.G., and B.Z. assisted with cell culture, biochemical assays, imaging, and data analysis. W.H. and W.L. conceived and supervised the study, interpreted the data, and revised the manuscript. All authors reviewed and approved the final manuscript.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-osteocalcin (OCN)	ABclonal	Cat# A14636; RRID:AB 2761511
Anti-FABP4	Proteintech	Cat# 12802-1-AP; RRID:AB 2102442
Anti-His-tag	Proteintech	Cat# 66005-1-Ig; RRID:AB 11232599
Anti-ALP	HUABIO	Cat# ET1601-21; RRID:AB 3069604
Anti-RUNX2	Cell Signaling Technology	Cat# 12556S; RRID:AB 2732805
Anti-SP7	Abcam	Cat# ab209484; RRID:AB 2892207
Anti-PPARG	Proteintech	Cat# 16643-1-AP; RRID:AB 10596794
Anti-ITGA5	Proteintech	Cat# 10569; RRID:AB 2130060
Anti-ITGB1	Proteintech	Cat# 26918; RRID:AB 2880685
Anti-FAK	Proteintech	Cat# 66258: RRID:AB 2881646
Anti-active RHOA	NewEast Biosciences	Cat# 26904: RRID:AB 1961799
Anti-ROCK2	Proteintech	Cat# 21645: RRID:AB 10858624
Anti-c-JUN (Immunoblotting); Anti-c-JUN (ChIP-qPCR)	Cell Signaling Technology;Proteintech	Cat# 9165T; RRID:AB 2130165; Cat# 24909; RRID:AB 2130165
Anti-PIEZO1	Beyotime	Cat# AF7743: RRID:AB 3752112
Anti-CAMK2	Beyotime	Cat# AF1639; RRID:AB 3711723
Anti-LAMIN A/C	Beyotime	Cat# AG2517; RRID:AB 3752111
Anti-β-ACTIN	PTM BIO	Cat# PTM-5018; RRID:AB 3086678
Fluorophore-conjugated secondary antibodies	Yeasen	Cat# 33107ES60; RRID:AB 3752113; Cat# 33112ES60; RRID:AB 3752114
HRP-conjugated secondary antibodies	Yeasen	Cat# 771485ES; RRID:AB 3752115; Cat# 771526ES60; RRID:AB 3752116
Control IgG for ChIP	Beyotime	Cat# A7016; RRID:AB 2905533
Biological samples
Young donor-derived hMSCs	This paper	See STAR Methods: Human subjects
Aged/elderly osteopenic/osteoporotic patient-derived hMSCs	This paper	See STAR Methods: Human subjects
Chemicals, peptides, and recombinant proteins
Glycosylated recombinant eukaryotic ALP	Sino-Bio	Cat# 10440-H08H
Isopropyl β-D-1-thiogalactopyranoside (IPTG)	Aladdin	Cat# I274316
LB medium	Aladdin	Cat# L113084
Dithiothreitol (DTT)	Aladdin	Cat# D104859
Oxidized/reduced glutathione (GSSG/GSH)	Aladdin	Cat# G105426; Cat# L274260
Ni-nitrilotriacetic acid (Ni-NTA) beads	Smart Lifescience	Cat# SA004100
Nano-hydroxyapatite (nHA)	Macklin	Cat# H875578
Protease inhibitor cocktail	Sigma-Aldrich	Cat# 04693159001
Fetal bovine serum (FBS)	Gibco	Cat# 26010074
Penicillin-streptomycin	Gibco	Cat# 15140122
L-glutamine	Gibco	Cat# 25030081
α-MEM	Gibco	Cat# 32561037
Collagenase	Solarbio	Cat# C8140
Uranyl acetate	Rhawn	Cat# R032929
ATP	Beyotime	Cat# ST1092
Inorganic pyrophosphate (PPi)	Solarbio	Cat# IYT5130
Streptozotocin (STZ)	Solarbio	Cat# S8050
DiD lipophilic fluorescent dye	Beyotime	Cat# C1423S
TRIzol reagent	Invitrogen	Cat# 15596026CN
RIPA lysis buffer	Beyotime	Cat# P0013B
Hematoxylin and Eosin staining reagents	Baso	Cat# BA4097; Cat# BA4098
Masson’s trichrome staining kit	Solarbio	Cat# G1340
Goat serum	Boster Biological Technology	Cat# 20K13B09
DAPI	Solarbio	Cat# C0060
Oil Red O solution	Beyotime	Cat# C0158S
Alizarin Red S solution	Beyotime	Cat# C0140
Von Kossa silver solution	Solarbio	Cat# G3282
Dexamethasone	Sigma-Aldrich	Cat# D4902
Ascorbic acid	Beyotime	Cat# ST1434
β-glycerophosphate	Sigma-Aldrich	Cat# G9422
3-isobutyl-1-methylxanthine (IBMX)	Solarbio	Cat# II0010
Hydrocortisone	Sigma-Aldrich	Cat# H0396
Indomethacin	Solarbio	Cat# II0100
GsMTx4	Solarbio	Cat# IG6360
PKH67 dye	Beyotime	Cat# C2073M
SF633-labeled phalloidin	Solarbio	Cat# CA1670
Paraformaldehyde (PFA)	Solarbio	Cat# P1110
Cetylpyridinium chloride	Solarbio	Cat# IC0280
PMSF	Beyotime	Cat# ST505
Critical commercial assays
ALP activity assay kit	Nanjing Jiancheng Bioengineering Institute/NJJCBIO	Cat# A059-2-2
ATP assay kit	Beyotime	Cat# S0027
Pyrophosphate assay kit	Yeasen	Cat# 50115ES70
Amplex Red Phosphate Assay Kit	Beyotime	Cat# S0235S
His-tag ELISA kit	Cayman	Cat# 10012445
Bone-specific alkaline phosphatase ELISA kit	ELK Biotechnology	Cat# ELK11640
CTX-1 ELISA kit	ELK Biotechnology	Cat# ELK9628
Osteocalcin (OCN) ELISA kit	ELK Biotechnology	Cat# ELK9627
TGF-β1 ELISA kit	Boster Biological Technology	Cat# EK0513-PR
Total cholesterol assay kit	NJJCBIO	Cat# A111-1-1
Triglyceride assay kit	NJJCBIO	Cat# A110-1-1
ALP staining kit	Beyotime	Cat# C3250S
BCA Protein Assay Kit	Beyotime	Cat# P0010S
Dual Luciferase Assay System	Beyotime	Cat# RG088M
NEBNext Ultra RNA Library Prep Kit	NEB	Cat# E7530L
Experimental models: Organisms/strains
C57BL/6 mice	Experimental Animal Center of Xi’an Jiaotong University; Jiangsu Huachuang Xinnuo Pharmaceutical Technology Co., Ltd.	License SYXK 2020-001; license SCXK 2020-0009
Alpl± C57BL/6 mice	Cyagen Biosciences Inc	N/A
Adult female rhesus macaques	Shaanxi Longteng Macaque Breeding Co., Ltd.	N/A
Escherichia coli BL21	Beyotime	Cat# D1009S
Recombinant DNA
pET-28a-ALPL-8xHis expression plasmid	GenScript	Custom design
pRUNX2-TA-Luc	Beyotime	Cat# D4313
pRL-TK-Renilla	Beyotime	Cat# D2760
Software and algorithms
GraphPad Prism	GraphPad Software	v10
SPSS	IBM	v22
Qlucore Omics Explorer	Qlucore	v3.2
Proteome Discoverer	Thermo Fisher Scientific	v3.2
Imaris	Bitplane	v9.8
ImageJ	NIH	v1.54q
Deposited data
RNA-seq datasets	Generated by the Authors	PRJNA1469690
Proteomics datasets	Generated by the Authors	IPX0017448000
Other
PVDF membranes	Millipore	Cat# IPFL00010
Customized high-fat diet (40% fat)	Beijing Keao Xieli Feed Co., Ltd.	Custom design
Standard chow	Xietong Bio	Cat# XTI01SL-002

Experimental model and study participant details

Human specimens

Clinical sample collection was approved by the Ethics Committee of the Second Affiliated Hospital of Xi’an Jiaotong University (approval no. MR-61-23-051601), and written informed consent was obtained from all participants or their legal representatives, where applicable. Participants of either gender were eligible and were stratified by age into a young group (19–25 years) and an aged group (61–68 years).

Inclusion criteria were: (a) fracture or bone injury caused by acute accidental trauma requiring orthopedic surgical treatment (for example, open reduction and internal fixation, debridement, or bone trimming); and (b) intraoperative availability of bone specimens obtained exclusively as surplus/discarded bone tissue or fragments generated during routine surgical procedures, without altering clinical management and without additional risk to the participant.

Exclusion criteria were: (a) non-traumatic fractures or suspected pathological fractures unrelated to the index traumatic event (for example, those related to malignancy, infection, or metabolic bone disease); (b) systemic conditions that could affect bone metabolism or fracture repair, including but not limited to, diabetes mellitus, chronic hepatic or renal dysfunction, endocrine/metabolic disorders, autoimmune or chronic inflammatory diseases, and hematological disorders; (c) a history of or current malignancy, or evidence of infection or tumor at the sampling site; (d) recent exposure to medications or treatments known to influence bone metabolism or immune status, for example, long-term glucocorticoids, antiresorptive or anabolic osteoporosis therapies, immunosuppressants, chemotherapy, or radiotherapy; and (e) absence of informed consent or incomplete key clinical information. Participant confidentiality and data security were strictly maintained throughout the study.

For the patient-derived bone marrow MSC validation in Figures 6A–6E, bone marrow MSCs isolated from five elderly osteopenic/osteoporotic fracture patients were analyzed, with one young healthy donor-derived hMSC preparation included as a reference comparator. Available clinical information, including age, gender, T-score and Z score, is summarized below. These clinical variables were used to document donor heterogeneity and were not used as exclusionary normalization factors in the in vitro functional assays. Gender was recorded from clinical records. Gender identity was not collected or analyzed. Because of the limited donor number, gender-specific effects were not statistically analyzed.

No.	Group	Age	Gender	T-score	Z score
1	Young	22	male	0.52	1.75
2	Young	19	male	0.78	1.88
3	Young	25	female	0.52	1.28
4	Aged	64	male	−1.67	−0.96
5	Aged	68	male	−2.24	−1.46
6	Aged	62	female	−2.80	−0.80
7	Aged	61	female	−2.02	−0.02
8	Aged	66	male	−1.74	−1.09
9	Aged	65	female	−2.62	−0.46

Animals and ethical approval

Eight-week-old male C57BL/6 mice were obtained from the Experimental Animal Center of Xi’an Jiaotong University (license SYXK 2020-001). Twenty-two-month-old male C57BL/6 mice were purchased from Jiangsu Huachuang Xinnuo Pharmaceutical Technology Co., Ltd. (license SCXK 2020-0009). Two pairs of Alpl ^± C57BL/6 mice were purchased from Cyagen Bioscience Inc. Mice were maintained under specific pathogen-free conditions on a 12 h light/12 h dark cycle and fed standard chow (Xietong Bio, XTI01SL-002) with water ad libitum. All mouse experiments were performed in accordance with institutional and national guidelines and were approved under protocol no. 20211746.

Adult female rhesus macaques (Macaca mulatta; n = 3; 13 years old) were obtained from Shaanxi Longteng Macaque Breeding Co., Ltd. and maintained under standard housing conditions on a 12 h light/12 h dark cycle with a laboratory primate maintenance diet. Macaque experiments were performed in accordance with institutional and national guidelines and were approved under protocol no. XJTUAE202608.

Bacterial expression system

Escherichia coli BL21 was used as the bacterial expression host for prokaryotic production of recombinant human ALPL. The pET-28a expression plasmid encoding C-terminally 8 × His-tagged human ALPL was transformed into E. coli BL21, and a high-expression clone was selected for protein production.

Method details

Unless otherwise stated, centrifugation was performed at 4°C and speeds are reported as relative centrifugal force (×g). Cells were cultured at 37°C in a humidified incubator with 5% CO₂.

Recombinant ALP expression, purification and refolding

To generate a recombinant form of human ALPL (encoding tissue-nonspecific alkaline phosphatase, ALP) suitable for bacterial expression, the N-terminal signal peptide and C-terminal glycosylphosphatidylinositol (GPI) anchor sequence were removed, and an eight-histidine (8 × His-tag) was appended to the C terminus. The resulting coding sequence was cloned into the pET-28a expression plasmid and transformed into Escherichia coli BL21. A high-expression clone was selected for protein production.

An overnight starter culture in LB medium (Aladdin, L113084) was used to inoculate fresh LB. At an optical density at 600 nm (OD600) of 0.8, protein expression was induced with isopropyl β-D-1-thiogalactopyranoside (IPTG; Aladdin, I274316) and cultures were incubated for an additional 24 h. Cells were harvested (4°C) and resuspended in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 2 mM MgCl₂ and 10 mM dithiothreitol [DTT; Aladdin, D104859], pH 8.0). Cells were lysed by sonication and insoluble material was collected by centrifugation.

The insoluble fraction was washed three times in wash buffer (0.1 mM EDTA and 0.5% Triton X-100), with brief sonication during each wash. To further deplete non-specifically associated proteins, the washed material was sequentially extracted twice using buffers containing 2 M, 4 M and 6 M urea. The target protein was then fully solubilized in denaturing buffer containing 6 M urea.

Denatured ALP was purified using Ni-nitrilotriacetic acid (Ni-NTA) beads (Smart Lifescience, SA004100). Purity was assessed by high-performance liquid chromatography (HPLC), and fractions corresponding to the dominant peak at 220 nm were collected. Fractions were flash-frozen in liquid nitrogen, lyophilized and stored at −80°C. For refolding, Ni-NTA-purified ALP was dialyzed at 4°C using a stepwise DTT gradient (0, 20 and 70 mM) together with an oxidized/reduced glutathione (GSSG/GSH; Aladdin, G105426 and L274260) redox system to obtain catalytically active ALP.

ALP enzyme kinetics and stability analysis

ALP enzyme kinetics assay: Refolded or non-refolded prokaryotic ALP was added to reaction buffers containing 10 μmol/L ATP (Beyotime, ST1092) or 20 μmol/L PPi (Solarbio, IYT5130) and incubated at 37°C for 0, 3, 6, 9, 12, 30, 60, 180, and 360 min. Following incubation, the residual ATP or PPi concentrations were measured, and substrate consumption curves were fitted to calculate enzymatic kinetic parameters. Glycosylated recombinant eukaryotic ALP (Sino-Bio, 10440-H08H) was used as a positive control.

ALP time-course activity assay: Using the same experimental setup, refolded human alkaline phosphatase expressed in a prokaryotic system (Prok-ALP), its unfolded form (Inac-ALP), and the commercial ALP (Sino-Bio, 10440-H08H) were incubated in reaction buffer for 0, 0.2, 0.5, 1, 3, 6, 12, 24, 48, and 72 h. Residual ATP or PPi was quantified at each time point to assess enzymatic activity.

ALP stability assay: Commercial recombinant ALP (Sino-Bio, 10440-H08H) or refolded prokaryotic ALP at 0.25 mg/mL was incubated in phosphate-buffered saline (PBS) over the same time course and His-tag levels were monitored using an ELISA kit (Cayman, 10012445). Changes in His-tag content were used as an indicator of ALP structural integrity.

Primary hMSC culture

Bone samples were immersed in PBS containing 2% penicillin-streptomycin (Pen-Strep) for 5 min to remove blood clots and soft tissue. The bone tissue was then cut into approximately 3 mm³ fragments and digested with collagenase at 37°C for 30 min. The fragments were washed multiple times with α-MEM complete medium containing 10% fetal bovine serum (FBS; Gibco, 26010074), 2 mM L-glutamine (Gibco, 25030081), 100 U/mL penicillin, and 100 μg/mL streptomycin (Gibco, 15140122). The bone fragments were seeded into culture flasks and cultured in α-MEM complete medium at 37°C with 5% CO₂. Once the cells reached 70%–80% confluence, they were passaged at a 1:3 ratio.

Cell membrane isolation

Cell membranes were isolated from healthy bone marrow mesenchymal stem cells (hMSCs) as indicated. Cell pellets were resuspended in double-distilled water (ddH₂O) containing protease inhibitor cocktail (Sigma-Aldrich, 04693159001) and allowed to swell for 5 min, followed by sonication. Lysates were centrifuged at 3,000×g for 5 min to remove nuclei and unbroken cells. The supernatant was centrifuged at 10,000×g for 20 min, and the resulting supernatant was ultracentrifuged at 100,000×g for 1 h using an Optima L-100 XP ultracentrifuge (Beckman Coulter). Membrane pellets were resuspended in RNase-free water and stored at −80°C.

Preparation of biomimetic osteogenic vesicles (OsteoVes)

Purified ALP (1 mg/mL) was mixed with nano-hydroxyapatite (nHA; 0.01 mg/mL; Macklin, H875578) at a 1:1 (v/v) ratio and incubated on ice for 30 min to facilitate adsorption. Healthy hMSC membranes (5 mg/mL) were then added, mixed thoroughly and incubated on ice for an additional 30 min.

The suspension was processed by membrane extrusion using a liposome extruder. Samples were sequentially passed through polycarbonate membranes with pore sizes of 400 nm, 200 nm and 100 nm (Whatman), performing 80–100 extrusion passes for each pore size to obtain a uniform vesicle population. The preparation was centrifuged at 15,000×g for 30 min at 4°C to remove unassembled/free ALP and nHA. The final product was resuspended in PBS and stored at 4°C for subsequent experiments.

Preparation of empty hMSC membrane vesicles

Empty hMSC membrane vesicles (eCMV) were prepared using the same hMSC membrane source and extrusion procedure as OsteoVes, but without ALP and nHA loading. Briefly, isolated hMSC membranes were resuspended in PBS and sequentially extruded through polycarbonate membranes with pore sizes of 400 nm, 200 nm and 100 nm. The resulting vesicles were centrifuged to remove unassembled material, resuspended in PBS and used as membrane-scaffold controls. eCMV dosing was matched to OsteoVes by membrane-protein equivalent.

Scanning electron microscopy

For scanning electron microscopy (SEM), 10 μL OsteoVes suspension was deposited onto a carbon-coated copper grid. After 5 min, excess liquid was removed and samples were air-dried. Grids were sputter-coated with gold for 90 s and imaged using a field-emission SEM (GeminiSEM 500, Zeiss).

Transmission electron microscopy

For transmission electron microscopy (TEM), 10 μL OsteoVes was applied to a carbon-coated copper grid for 5 min and excess liquid was blotted off. Grids were negatively stained with 2% (w/v) uranyl acetate (Rhawn, R032929) for 1 min, air-dried and imaged using a 120 kV TEM (Talos L 120C G2, Thermo Fisher Scientific).

Cryo-electron microscopy

For cryo-electron microscopy (cryo-EM), 100 μL OsteoVes was mixed with an equal volume of calcium/pyrophosphate buffer (10 mmol/L CaCl₂ and 3 mmol/L PPi, pH 7.4) and incubated for 0, 1, 3, 6, 12 or 24 h, or for 3, 5 or 7 days. After incubation, 2 μL sample was applied to a carbon-coated copper grid and blotted to form a thin film. Grids were plunge-frozen in liquid ethane to vitrify the sample and imaged using a cryo-EM (Talos F200C, Thermo Fisher Scientific).

Dynamic light scattering

To assess serum stability, 100 μL FBS was added to 900 μL OsteoVes and samples were incubated for 0, 3, 6, 12, 24, 48 or 72 h. Hydrodynamic diameter and zeta potential were measured using dynamic light scattering (Zetasizer Nano ZSE, Malvern).

Micro-infrared spectroscopy

ALP solution, nHA, an ALP-nHA mixture and OsteoVes samples (100 μL each) were flash-frozen in liquid nitrogen and lyophilized under vacuum. Infrared spectra were recorded from 400 to 4,000 cm⁻¹ using a micro-infrared spectrometer (VERTEX70, Bruker).

ALP activity and substrate consumption assays

ALP activity assay: OsteoVes (1 mg/mL) was incubated in PBS (pH 7.4) supplemented with 10% FBS at room temperature for 0, 0.2, 0.5, 1, 3, 6, 12, 24, 48 or 72 h. ALP activity was quantified using a commercial kit (Nanjing Jiancheng Bioengineering Institute, A059-2-2) following the manufacturer’s instructions, and absorbance was read at 520 nm. The calculation of ALP activity is as follows:

$$ALPactivity\left(U/L\right)=\frac{A_T-A_0}{A_S-A_0}\times 100\times Cs\times 7.14$$ (Equation 1)

where A_T is the absorbance of the test sample, A₀ is the absorbance of the blank, A_S is the absorbance of the standard at 520 nm, and Cs is the concentration of the standard solution.

ATP assay: For ATP consumption assays, ALP (1 mg/mL) was mixed 1:1 (v/v) with ATP (10 μmol/L) in PBS (pH 7.4) and incubated for 10 min at room temperature. Alternatively, OsteoVes (1 mg/mL) was mixed 1:1 (v/v) with ATP (10 μmol/L) in PBS containing 10% FBS and incubated for 0, 0.2, 0.5, 1, 3, 6, 12, 24, 48 or 72 h. Residual ATP was quantified using an ATP assay kit (Beyotime, S0027) following the manufacturer’s instructions.

PPi assay: For inorganic pyrophosphate (PPi) consumption assays, ALP (1 mg/mL) was mixed 1:1 (v/v) with PPi (20 μmol/L) in PBS (pH 7.4) and incubated for 10 min at room temperature. Alternatively, OsteoVes (1 mg/mL) was mixed 1:1 (v/v) with PPi (20 μmol/L) in PBS containing 10% FBS and incubated for 0, 0.2, 0.5, 1, 3, 6, 12, 24, 48 or 72 h. Residual PPi was quantified using a PPi assay kit (Yeasen, 50115ES70) according to the manufacturer’s instructions.

Mouse dosing regimen

Twenty-two-month-old naturally aged C57BL/6 mice and 8-month-old Alpl ^± mice (with their cage-mate controls) received OsteoVes by tail vein injection at 100 μg per 20 g body weight (5 mg/kg). A second dose was administered 7 days after the first injection (two total injections). Control mice received an equivalent volume of saline. Bone phenotyping was performed 4 weeks after the initial injection. All mice in both the naturally aged C57BL/6 and Alpl ^± cohorts, including cage-mate controls, were male.

Rhesus macaque hyperglycemia/hyperlipidemia model and treatment

To establish a combined hyperglycemia/hyperlipidemia model, rhesus macaques received streptozotocin (STZ; 30 mg/kg; Solarbio, S8050) by intravenous injection into the forearm cephalic vein, followed by a second identical injection 1 month later. Animals were fasted for 8 h before STZ administration with free access to water. After STZ administration, macaques were maintained on a customized high-fat diet (40% fat; Beijing Keao Xieli Feed Co., Ltd.) and provided 10% sucrose drinking water for 6 months, with periodic blood glucose monitoring.

Before OsteoVes treatment, macaques received two intravenous saline injections 2 weeks apart; the time point after the second saline injection was defined as baseline (week 0), at which blood sampling and in vivo micro-CT scanning were performed. OsteoVes was then administered intravenously at 5 mg/kg, and a second dose was given 7 days later. Blood samples were collected every 2 weeks after treatment initiation; serum Ca²⁺ and Pi were monitored longitudinally for up to 3 months where indicated. In vivo micro-CT scans were performed at weeks 4 and 12 after treatment initiation for paired longitudinal assessment of trabecular parameters.

In vivo tracking of OsteoVes distribution

OsteoVes were labeled with the lipophilic fluorescent dye DiD (Beyotime, C1423S), yielding DiD-labeled vesicles (OsteoVes-DiD). Mice were administered 300 μL OsteoVes-DiD via tail vein injection. At 0, 0.5, 2, 6, 24, 48, and 72 h post-injection, mice were euthanized, and major organs, including heart, lung, spleen, liver, kidney, vertebrae, and long bones, were harvested. Fluorescence imaging was carried out using an IVIS Lumina XRMS Series III system (Caliper Life Sciences/PerkinElmer). DiD signal intensity was quantified to assess the biodistribution and accumulation of OsteoVes across organs and bone tissues.

Micro-computed tomography scanning and analysis

Mouse femora were scanned using a high-resolution micro-CT system (Xradia 610 Versa, Zeiss) at 100 kV, 110 μA and 500 ms integration time. In vivo femoral scans in rhesus macaques were acquired using a high-resolution micro-CT system (AX3000-D, Always Imaging) with tube voltage 160–180 kV, tube current 0.1–0.5 mA and an integration time of 1 s. Bone morphometric parameters, including bone mineral density (BMD), bone volume fraction (BV/TV), trabecular number (Tb.N), trabecular separation (Tb.Sp) and trabecular thickness (Tb.Th), were calculated using the manufacturer’s software.

Three-point bending test

Fresh mouse femora were cleaned of adherent soft tissues and tested using a computer-controlled universal testing machine (Jinan Chendal Testing Machine Manufacturing Co., Ltd.). Specimens were mounted on custom fixtures with a span length of 10 mm and loaded at a crosshead speed of 2 mm/min. Load-displacement curves were recorded automatically and used to derive maximum load, bending strength and elastic modulus.

Atomic force microscopy measurements

Biomechanical properties of bone cryosections or hMSCs were assessed using an atomic force microscope (Nano-Observer AFM, CSI, France) equipped with a conical-tip probe (ACTA, APPNANO, USA; nominal resonance frequency 300.0 kHz; spring constant 37 N/m; half-opening angle 18°). The spring constant was calibrated by the thermal noise method before each measurement, and deflection sensitivity was determined by indenting a clean glass substrate.

For quasi-static measurements, the indentation speed was 200 nm/s and the peak force was 25.5 nN. For dynamic creep testing, the indentation speed was 20 μm/s, the force was held at 25.5 nN for 10 s and data were acquired at 500 points/s. For PeakForce tapping imaging, the scan rate was 1 line/s, resolution was 512 pixels, and the scan area was 100 × 100 μm².

The static elastic modulus was calculated using the Sneddon model, which is applicable to soft materials indented by a conical tip:

$$F=\frac{2}{\pi }{E}^{\ast }{\delta }^2\tan \phi$$ (Equation 2)

where F is the applied force, δ is the indentation depth, φ is the probe half-opening angle, E is Young’s modulus and ν is Poisson’s ratio. Poisson’s ratio was set to 0.5, assuming incompressible behavior.

Thus, the Young’s modulus of the samples can be calculated as:

$$E^{\ast }=\frac{E}{1-v^2}$$ (Equation 3)

where E is the Young’s modulus and ν represents the Poisson ratio, assumed to be 0.5 for an incompressible gel.

RNA sequencing and analysis

Mice were euthanized 4 weeks after OsteoVes treatment. Femora were harvested, flash-frozen in liquid nitrogen and pulverized. Total RNA was extracted using TRIzol reagent (Invitrogen, 15596026CN). RNA-seq libraries were prepared using the NEBNext Ultra RNA Library Prep Kit (NEB, E7530L) and sequenced as 2 × 150 bp paired-end reads on an Illumina NovaSeq 6000 platform. Downstream analyses, including differential expression analysis, heatmap visualization and gene set enrichment analysis (GSEA), and Gene Ontology (GO) enrichment, were performed using Qlucore Omics Explorer v3.2.

Proteomics and pathway analysis

Mice were euthanized 4 weeks after OsteoVes treatment and total protein was extracted from femora using RIPA lysis buffer supplemented with protease and phosphatase inhibitors. Protein concentration was determined by BCA assay. Equal amounts of protein were separated by SDS-PAGE, subjected to in-gel tryptic digestion, and peptides were desalted using mixed-mode MCX-C18 solid-phase extraction cartridges. Peptides were analyzed by LC-MS/MS on a Thermo Fisher Scientific system. RAW files were processed, searched and quantified using Proteome Discoverer (Thermo Fisher Scientific). Quantified protein intensities were imported into Qlucore Omics Explorer v3.2 for principal component analysis (PCA), and pathway analysis was conducted using Ingenuity Pathway Analysis (IPA).

Bone histology and immunofluorescence

Histological staining: Decalcified bone sections were stained with Hematoxylin and Eosin (H&E; Baso, BA4097 and BA4098) and Masson’s trichrome (Solarbio, G1340) according to the manufacturers’ protocols. Images were acquired using an automated whole-slide scanner (Pannoramic DESK, 3DHISTECH).

Immunofluorescence staining: Bone cryosections were air-dried for 2 h at room temperature, washed with PBS and permeabilized with 0.3% Triton X-100 for 10 min. Sections were blocked with goat serum (Boster Biological Technology, 20K13B09) for 1 h and incubated overnight at 4°C with primary antibodies (1:200): Osteocalcin (OCN; ABclonal, A14636), FABP4 (Proteintech, 12802-1-AP), His-tag (Proteintech, 66005-1-Ig), ALP (HUABIO, ET1601) and RUNX2 (Cell Signaling Technology, 12556S). After PBS washes, sections were incubated with fluorophore-conjugated secondary antibodies (Yeasen) for 1 h at room temperature and counterstained with DAPI (Solarbio, C0060) for 15 min. Images were captured using a confocal microscope (FV3000, Olympus) or an automated scanning system (Axio Scan Z, Zeiss).

Oil Red O staining of bone sections: To evaluate adipocyte content in trabecular bone regions, bone cryosections were stained with Oil Red O (Beyotime, C0158S) for 10–20 min at room temperature, rinsed with 60% isopropanol, washed twice with PBS and mounted. Images were acquired using an automated whole-slide scanner (Pannoramic DESK, 3DHISTECH).

Blood glucose measurement

Capillary blood was collected from the fingertip of rhesus macaques and glucose concentration was measured using a glucometer (OneTouch UltraEasy; Johnson & Johnson) based on the glucose oxidase method, following the manufacturer’s instructions. Results are reported as mmol/L.

Hematology and serum biochemistry

Peripheral blood from mice and rhesus macaques was analyzed for complete blood counts (CBC), including red blood cell (RBC), white blood cell (WBC), neutrophil (NEUT), lymphocyte (LYMPH) and platelet (PLT) counts, as well as hemoglobin (HGB). Serum was separated and analyzed for aspartate aminotransferase (AST), alanine aminotransferase (ALT), creatine kinase (CK), lactate dehydrogenase (LDH), urea (UREA) and creatinine (CREA). Total cholesterol and triglycerides were quantified using commercial kits (Nanjing Jiancheng Bioengineering Institute, A111-1-1 and A110-1-1).

Ectopic mineral deposition

Organs (including heart, liver, and kidney) were collected from control and OsteoVes-treated mice, fixed, dehydrated, and embedded in paraffin. Paraffin sections were subsequently stained with Alizarin Red S and Von Kossa to assess ectopic calcification. Whole-slide images were acquired using an automated slide scanner (Pannoramic DESK, 3DHISTECH).

Alizarin Red S Staining: Sections were deparaffinized and rehydrated to water, then incubated with Alizarin Red S solution (Beyotime, C0140) at room temperature for 5 min, followed by rapid rinsing with distilled water for 5 min. After dehydration and clearing, sections were mounted with neutral resin.

Von Kossa Staining: Following deparaffinization and rehydration, sections were covered with Von Kossa silver solution (Solarbio, G3282) and exposed to ultraviolet light for 10 min. They were then rinsed with distilled water for 1 min, treated with sodium thiosulfate for 2 min, and rinsed again with distilled water for another 2 min. Eosin counterstaining was performed before dehydration, clearing, and mounting with neutral resin.

ELISA

Serum levels of bone-related factors in rhesus macaques were quantified by ELISA, including bone-specific alkaline phosphatase (bALP; ELK Biotechnology, ELK11640), CTX-1 (ELK Biotechnology, ELK9628), Osteocalcin (OCN; ELK Biotechnology, ELK9627) and TGF-β1 (Boster Biological Technology, EK0513-PR). Assays were performed following manufacturers’ instructions. Briefly, 100 μL standards or samples were added to each well (blank wells received no sample), followed by HRP-labeled detection antibodies. Plates were sealed and incubated at 37°C for 60 min, washed five times, and developed by adding 50 μL each of chromogenic reagents A and B. Plates were incubated in the dark at 37°C for 30 min, the reaction was stopped, and absorbance was measured at 450 nm.

hOB-MV extraction

Osteoblasts used in this study were induced from the hMSCs obtained from healthy young volunteers. Cells were cultured for 7 days in osteogenic induction medium containing exosome-depleted FBS, dexamethasone (10 nM), ascorbic acid (100 μg/mL) and β-glycerophosphate (1 mM). Conditioned medium was collected and osteoblast-derived matrix vesicles (hOB-MVs) were isolated by differential ultracentrifugation. Conditioned medium was centrifuged at 20,000 × g for 30 min to remove apoptotic bodies and cell debris, and the supernatant was ultracentrifuged at 100,000 × g for 60 min. The pellet was resuspended in ice-cold PBS, aliquoted and stored at −80°C.

hMSC differentiation and treatments

Passage-1 aged hMSCs were seeded at 2 × 10⁵ cells per well (unless otherwise specified). When cultures reached >90% confluence, cells were induced toward osteogenic or adipogenic lineages. OV^ΔALP and OV^ΔnHA refer to OsteoVes prepared in the absence of ALP or nHA, respectively. For vesicle-based treatments, concentrations are reported as total protein equivalents.

For osteogenic differentiation, the medium was replaced with osteogenic induction medium containing dexamethasone (10 nM), ascorbic acid (100 μg/mL) and β-glycerophosphate (1 mM). Cells were treated with hOB-MVs, OsteoVes, OV^ΔALP, OV^ΔnHA or an ALP-nHA mixture (each at 20 μg/mL). PBS was used as vehicle control. Induction medium was refreshed every 2–3 days. At day 7, ALP activity and ALP staining were performed, Pi and PPi levels in conditioned medium were measured, and osteogenic protein expression was analyzed by immunoblotting. At day 21, mineral deposition was assessed using Alizarin Red S staining and Von Kossa staining.

For adipogenic differentiation, the medium was replaced with adipogenic induction medium containing 3-isobutyl-1-methylxanthine (IBMX; 0.5 mM), hydrocortisone (0.5 μM) and indomethacin (60 μM). Cells were treated with hOB-MVs or OsteoVes (20 μg/mL) or PBS (control). Induction medium was refreshed every 2–3 days for 14 days. Lipid accumulation was assessed by Oil Red O staining. PPARG expression was evaluated by immunoblotting at day 7.

For empty membrane control experiments, aged hMSCs were treated with eCMV at membrane-protein equivalents matched to intact OsteoVes. For chemical phosphate-control experiments, inorganic phosphate was supplied as Na₂HPO₄/NaH₂PO₄ to achieve a final total phosphate concentration of 5 mM, unless otherwise stated. For PIEZO1 inhibition experiments, aged hMSCs were treated with OsteoVes in the presence or absence of GsMTx4 (10 μM) during osteogenic induction. ALP staining, immunoblotting and RUNX2 nuclear-translocation assays were performed at the indicated time points.

Comparator normalization and treatment logic

Unless otherwise stated, vesicle-based in vitro treatments were normalized by total protein equivalents. hOB-MV, OsteoVes, OV^ΔALP and OV^ΔnHA were used at 20 μg/mL total protein equivalents. Empty hMSC membrane vesicles (eCMV) were matched to OsteoVes by membrane-protein equivalent. The ALP-nHA group was included as a membrane-free catalytic/mineral core control and was matched to the corresponding ALP/nHA input where applicable; because this formulation lacks a membrane shell, it was not interpreted as a vesicle-number- or membrane-mass-matched control.

For patient-derived MSC assays, ALP-positive area, conditioned-medium ALP activity and Alizarin Red S mineralized area were analyzed as measured values for each donor-derived biological sample. The young donor preparation was included as a reference comparator and was not used as a normalization denominator unless explicitly stated.

ALP activity in conditioned medium

Conditioned media from osteogenic cultures at day 7 were collected and ALP activity was quantified using an ALP assay kit (NJJCBIO, A059-2-2) following the manufacturer’s instructions. ALP activity was calculated as described above.

Phosphate (Pi) assay

Pi levels in conditioned media from osteogenic cultures at day 7 were quantified using the Amplex Red Phosphate Assay Kit (Beyotime, S0235S). In black 96-well plates, 10 μL standards, blank (assay buffer only) or samples were mixed with 90 μL Amplex Red working solution per well. After incubation at 37°C for 30 min, fluorescence was measured (Ex/Em = 560/590 nm).

Pyrophosphate (PPi) assay

PPi levels in conditioned media from osteogenic cultures at day 7 were measured using a fluorimetric pyrophosphate assay kit (Yeasen, 50115ES70). In black 96-well plates, 50 μL PPi standards, blank (assay buffer only) or samples were mixed with 50 μL PPi sensor working solution (total volume 100 μL per well). After incubation at room temperature for 10–30 min, fluorescence was measured (Ex/Em = 316/456 nm).

ALP staining

hMSCs were stained for ALP after 7 days of osteogenic induction using an ALP staining kit (Beyotime, C3250S) according to the manufacturer’s protocol.

Alizarin Red S staining and Von Kossa staining

After 21 days of osteogenic induction, hMSCs were fixed with 4% paraformaldehyde (PFA) and washed with PBS before mineralization staining.

For Alizarin Red S staining, cells were incubated with 1% Alizarin Red S working solution for 1–3 min at room temperature, washed thoroughly with ddH₂O and imaged. For quantification, mineral deposits were solubilized with 10% cetylpyridinium chloride and absorbance was measured at 570 nm.

For Von Kossa staining, cells were incubated with Von Kossa silver solution under UV light for 20 min, rinsed with distilled water for 1 min, incubated with sodium thiosulfate solution for 2 min, rinsed, air-dried and imaged.

Oil Red O staining

After 14 days of adipogenic induction, hMSCs were fixed with 4% PFA and stained with Oil Red O solution (Beyotime, C0158S) for 10–20 min at room temperature. Cells were rinsed with 60% isopropanol and PBS, and images were acquired by light microscopy. For quantification, Oil Red O was eluted with isopropanol and absorbance was measured at 520 nm.

RNA sequencing of hMSCs

hMSCs were treated with OsteoVes (20 μg/mL) or an equal volume of PBS (control) for 48 h. Total RNA was extracted using TRIzol reagent (Invitrogen, 15596026CN). RNA-seq library preparation, sequencing (Illumina NovaSeq 6000; 2 × 150 bp paired-end) and downstream analyses (including differential expression, GSEA and GO enrichment) were performed as described for mouse femora.

Proteomics of hOB-MV and hMSC membranes

Total protein was extracted from matrix vesicles derived from induced osteoblasts and hMSC membranes using RIPA lysis buffer supplemented with protease and phosphatase inhibitors and quantified by BCA assay. Equal amounts of protein were processed for LC-MS/MS and analyzed as described above for bone proteomics, including Proteome Discoverer-based processing, Qlucore Omics Explorer v3.2-based PCA and Ingenuity Pathway Analysis (IPA).

Immunofluorescence imaging

For OsteoVes association imaging, hMSCs were seeded onto glass-bottom dishes (2 × 10⁵ cells per dish) and incubated with PKH67-labeled OsteoVes (20 μg/mL) for 12 h or 24 h. Cells were washed with PBS, fixed with 4% PFA and permeabilized with 0.3% Triton X-100 for 10 min. F-ACTIN was stained with SF633-labeled phalloidin (Solarbio, CA1670; 1:150) for 1 h and nuclei were counterstained with DAPI (Solarbio, C0060) for 15 min. Images were acquired using a super-resolution confocal microscope (TCS SP8 STED 3×, Leica) and 3D reconstruction was performed using Imaris v9.8 (Bitplane).

For RUNX2 immunostaining, hMSCs were seeded onto glass-bottom dishes (3,000 cells per dish). The next day, cells were treated with OsteoVes (20 μg/mL) for 24 h (PBS as control), fixed and permeabilized as above, and blocked with goat serum (Boster Biological Technology, 20K13B09) for 1 h. Cells were incubated overnight at 4°C with RUNX2 antibody (Cell Signaling Technology, 12556S; 1:200), followed by fluorophore-conjugated secondary antibodies (Yeasen) for 1 h at room temperature and DAPI counterstaining. Images were acquired on a confocal microscope (FV3000, Olympus).

For ECM retention assays, PKH67-labeled OsteoVes were incubated with ECM-bearing hMSC cultures for 12 or 24 h. After gentle washing to remove unbound particles, cells were fixed, stained with SF633-phalloidin and DAPI, and imaged by confocal microscopy. Three-dimensional reconstruction was performed using Imaris. Vesicle-associated fluorescence intensity and vesicle–F-actin/ECM colocalization were quantified using identical acquisition and analysis settings across groups.

Immunoblotting

hMSCs were seeded in 6-well plates at 2 × 10⁵ cells per well and treated as indicated. Cells were lysed on ice for 30 min in RIPA buffer containing PMSF. Protein concentration was determined using a BCA Protein Assay Kit (Beyotime, P0010S). Proteins were denatured in SDS sample buffer containing 50 mM DTT at 100°C for 10 min.

Equal amounts of protein were resolved on 10% SDS-PAGE gels and transferred to 0.45 μm PVDF membranes (Millipore, IPFL00010). Membranes were blocked with 5% BSA for 1 h at room temperature and incubated overnight at 4°C with primary antibodies: RUNX2 (Cell Signaling Technology, 12556S), ALP (HUABIO, ET1601), SP7 (Abcam, ab209484), PPARG (Proteintech, 16643-1-AP), ITGA5 (Proteintech, 10569), ITGB1 (Proteintech, 26918), FAK (Proteintech, 66258), active RHOA (NewEast Biosciences, 26904), ROCK2 (Proteintech, 21645), c-JUN (Cell Signaling Technology, 9165T), PIEZO1 (Beyotime, AF7743), CAMK2 (Beyotime, AF1639), LAMIN A/C (Beyotime, AG2517) and β-ACTIN (PTM BIO, PTM-5018). After washing with TBST, membranes were incubated with HRP-conjugated secondary antibodies (Yeasen) for 1 h at room temperature and visualized using enhanced chemiluminescence (ECL; Millipore).

Dual-luciferase reporter assay

RUNX2 transcriptional activity was measured using a dual-luciferase reporter assay. Cells were seeded in 24-well plates at 1 × 10⁵ cells per well and co-transfected with pRUNX2-TA-Luc (Beyotime, D4313) and pRL-TK-Renilla (Beyotime, D2760). After transfection, OsteoVes (20 μg/mL) was added and cells were incubated for 24 h. Luciferase activities were quantified using the Dual Luciferase Assay System (Beyotime, RG088M) on a SpectraMax iD5 plate reader (Molecular Devices). Relative luciferase activity was calculated as the ratio of firefly to Renilla luminescence.

Chromatin immunoprecipitation (ChIP)

Human mesenchymal stem cells were subjected to osteogenic induction and treated with OsteoVes or PBS control. After 48 h, cells were crosslinked with 1% formaldehyde, quenched with glycine, lysed and sonicated to shear chromatin. A fraction of chromatin was reserved as input. The remaining chromatin was immunoprecipitated overnight at 4°C using an anti-c-JUN antibody (Proteintech, 24909) or control IgG (Beyotime, A7016). Immune complexes were captured using Protein A/G magnetic beads, washed sequentially, and crosslinks were reversed. DNA was purified and analyzed by qPCR to assess c-JUN enrichment at the RUNX2 promoter. The primers used were:

Forward primer: 5′-AGGAGTTTGGGCTCCTTCAG-3′; Reverse primer: 5′-CTTAAGTAAAGTGGGACTGCCT-3′.

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism 10.0 and SPSS 22.0. Data are presented as mean ± s.d. unless otherwise stated. Exact n values, the biological unit represented by n, statistical tests, and comparison structures are provided in the corresponding figure legends. For in vitro experiments, n represents independently prepared biological replicates, independent vesicle preparations or donor-derived samples, as indicated. For mouse experiments, n represents individual animals. For rhesus macaque analyses, n represents individual macaques. For imaging-based histology and immunofluorescence quantification, multiple non-overlapping technical fields of view or sections were analyzed using identical acquisition and thresholding settings; technical measurements were averaged to generate one value per biological replicate before group-level statistical analysis.

For comparisons between two independent groups, unpaired two-sided Student’s t-tests were used when data met the assumptions for parametric testing. For longitudinal rhesus macaque analyses, paired two-sided Student’s t-tests were used for prespecified pre-versus post-treatment comparisons; when more than two longitudinal time points were compared, paired or repeated-measures analyses were used as appropriate and specified in the corresponding legend. For omics and GSEA analyses, normalized enrichment scores, nominal p values, and FDR q values are reported where applicable. Targeted biological endpoint panels were interpreted as prespecified endpoint analyses and were not adjusted across unrelated endpoints unless explicitly indicated. No data were excluded. Sample size was not predetermined by formal power calculation. Experiments were not randomized or blinded unless otherwise stated. A p value <0.05 was considered statistically significant.

Published: July 16, 2026