Hypoxia-programmed apoptotic-mimetic M2 macrophage-derived small extracellular vesicles for hydrogel-mediated immunoregenerative bone repair

Abstract

Effective regeneration of critical-sized bone defects requires the coordinated reconstruction of immune homeostasis, vascular networks, and osteogenic matrix formation. However, conventional bone-regenerative materials frequently provide structural support or single-pathway osteoinduction, while failing to actively orchestrate the multicellular repair microenvironment. Here, we develop a hydrogel-assisted immunoregenerative strategy based on hypoxia-programmed apoptotic-mimetic small extracellular vesicles derived from M2-like macrophages. Interleukin-4-polarized macrophages were subjected to hypoxic preconditioning to generate M2-HP-Apo-sEVs with increased phosphatidylserine exposure, enhanced cellular internalization, and amplified immunomodulatory potency. Compared with normoxic M2-sEVs, M2-HP-Apo-sEVs more effectively reprogrammed inflammatory macrophages toward a pro-resolving phenotype, suppressed inflammatory cytokine production, and enhanced reparative cytokine secretion. The vesicles further promoted osteogenic differentiation of bone marrow mesenchymal stem cells and stimulated endothelial migration and capillary-like network formation, demonstrating coordinated osteo–immuno–angiogenic activity. To enable sustained local delivery, M2-HP-Apo-sEVs were incorporated into photocrosslinkable GelMA hydrogels and implanted into critical-sized calvarial defects. The GelMA–M2-HP-Apo-sEV system substantially enhanced bone regeneration, vascularization, and osteogenic matrix deposition while attenuating local inflammation. This work establishes a cell-free vesicle–hydrogel platform that integrates donor-cell hypoxic programming, apoptotic-mimetic membrane signaling, and local biomaterial delivery for immunoregenerative bone repair.

Graphical abstract

Schematic illustration of GelMA-mediated local delivery of M2-HP-Apo-sEVs for immunoregenerative repair of critical-sized calvarial bone defects.

1. Introduction

The repair of large craniofacial and calvarial bone defects remains a formidable clinical challenge, particularly when the defect size exceeds the intrinsic regenerative capacity of the host tissue. Although autologous bone grafting is still regarded as the clinical benchmark, its use is constrained by donor-site morbidity, limited tissue availability, prolonged operative time, and unpredictable graft remodeling. Synthetic scaffolds and osteoconductive substitutes have partially addressed structural requirements; however, many of these materials primarily function as passive fillers and fail to actively coordinate the multicellular biological events required for complete bone regeneration [[1], [2], [3], [4]]. In recent years, bone repair has increasingly been recognized as a dynamic immunoregenerative process rather than a purely osteogenic event. Successful regeneration requires the temporally orchestrated resolution of inflammation, recruitment and differentiation of osteoprogenitor cells, vascular network formation, extracellular matrix deposition, and subsequent mineral maturation [[5], [6], [7], [8]]. Accordingly, next-generation bone-regenerative biomaterials are expected not only to provide a permissive matrix, but also to instruct the defect microenvironment toward immune homeostasis, vascularization, and osteogenesis.

Macrophages occupy a central position in this regenerative cascade because they rapidly infiltrate the injured bone microenvironment and act as both inflammatory sensors and tissue-repair coordinators [[9], [10], [11]]. Immediately after injury, pro-inflammatory macrophages participate in debris clearance and pathogen defense; however, persistent inflammatory activation can amplify cytokine release, impair osteoprogenitor survival, inhibit angiogenesis, and ultimately delay bone bridging. Conversely, timely transition toward a reparative macrophage phenotype supports inflammation resolution, extracellular matrix remodeling, vascular stabilization, and osteogenic differentiation through paracrine signaling [[12], [13], [14], [15]]. This macrophage-dependent immune plasticity has inspired the design of osteoimmunomodulatory biomaterials capable of reshaping macrophage phenotype rather than merely delivering osteogenic factors [[16], [17], [18]]. Nevertheless, many current immunomodulatory scaffolds rely on small molecules, inorganic ions, or cytokines, which may exhibit rapid diffusion, narrow signaling spectra, limited cellular specificity, or insufficient ability to reproduce the complex paracrine language of reparative immune cells.

Extracellular vesicles (EVs), particularly small extracellular vesicles (sEVs), have emerged as cell-free biological nanocarriers that recapitulate important aspects of intercellular communication. Owing to their lipid bilayer membrane, nanoscale size, intrinsic biocompatibility, and ability to transfer proteins, lipids, mRNAs, and microRNAs, sEVs can modulate recipient-cell phenotype through multi-target and multi-pathway regulation [[19], [20], [21], [22]]. In bone regeneration, sEVs derived from mesenchymal stromal cells, osteogenic cells, endothelial cells, and macrophages have been reported to regulate osteogenesis, angiogenesis, inflammation, and matrix remodeling [[23], [24], [25], [26]]. Among these, macrophage-derived sEVs are particularly attractive because their cargo composition and biological function are closely linked to macrophage polarization status. M2 macrophage-derived exosomes have been shown to re-establish an osteoimmune microenvironment and promote fracture repair, including under diabetic conditions where inflammatory dysregulation compromises bone healing [27]. Moreover, macrophage-derived EVs can directly communicate with mesenchymal stromal cells, influence osteogenic differentiation, and participate in vascularized bone formation [28,29]. These findings suggest that M2 macrophage-derived sEVs may serve as instructive immunoregenerative mediators rather than conventional nanocarriers.

Despite this promise, the therapeutic efficacy of native sEVs remains limited by suboptimal yield, heterogeneous bioactivity, inefficient uptake by target cells, and rapid clearance after local administration [[30], [31], [32]]. Therefore, donor-cell preconditioning has become an important strategy for enhancing EV function. Hypoxic preconditioning is particularly relevant because low oxygen tension is a physiological feature of many stem-cell and injury niches and can reshape cellular metabolism, secretory activity, and vesicle cargo. Recent studies have demonstrated that hypoxia-preconditioned apoptotic EVs possess enhanced biological potency, including improved regulation of stem-cell activity and macrophage polarization [33]. Importantly, apoptotic or apoptotic-mimetic vesicles can display phosphatidylserine (PS) on their membrane surface, thereby mimicking “eat-me” signals associated with efferocytosis. PS-mediated recognition is not merely a clearance mechanism; it can actively trigger anti-inflammatory and pro-resolving responses in phagocytic cells [[34], [35], [36]]. In bone repair, PS-responsive hydrogel systems have recently been shown to promote regeneration through immunomodulation, supporting the concept that apoptotic-mimetic lipid signaling can be harnessed as a regenerative cue [13]. However, whether hypoxic preconditioning can be used to endow M2 macrophage-derived sEVs with PS-associated apoptotic mimicry and whether this membrane feature may contribute to enhanced macrophage uptake and downstream inflammatory reprogramming, osteogenesis, and angiogenesis remains insufficiently explored.

Another major obstacle for EV-based bone repair is the lack of a suitable local delivery platform. Direct injection of sEVs often results in rapid diffusion, enzymatic degradation, and insufficient retention within the defect region, which limits therapeutic durability [22,37,38]. Hydrogels provide an attractive solution because their hydrated polymeric networks can protect vesicles, prolong local residence, permit sustained release, and conformally fill irregular defects. Gelatin methacryloyl (GelMA), in particular, has received extensive attention because it combines cell-adhesive gelatin motifs with photocrosslinkable processability, tunable degradation, and favorable cytocompatibility [24,39,40]. Notably, recent studies have demonstrated that hypoxia-primed EVs or bone-functional exosomes can be incorporated into injectable or GelMA-based hydrogels to improve vesicle stability, local retention, and bone-regenerative efficacy, supporting hydrogel-assisted EV delivery as a promising cell-free strategy for skeletal repair [25,41]. More broadly, hydrogel–exosome systems have been increasingly applied across tissue engineering scenarios, including skin, bone, cartilage, nerve, and tendon regeneration [42]. Nevertheless, most EV–hydrogel systems still focus on stem-cell-derived EVs or osteogenic EVs, whereas macrophage-derived immunoregulatory EVs with engineered apoptotic-mimetic properties remain comparatively underdeveloped for bone defect repair.

Herein, we report a GelMA-based local delivery system incorporating hypoxia-preconditioned, apoptotic-mimetic, M2 macrophage-derived small extracellular vesicles for immunoregenerative bone repair. RAW264.7 macrophages were first polarized by IL-4 to generate reparative M2-like macrophages and then subjected to hypoxic preconditioning to produce M2-HP-Apo-sEVs. We hypothesized that hypoxic programming would enrich an apoptotic-mimetic membrane phenotype, particularly PS exposure, which might contribute to improved vesicle recognition by macrophages and stronger immunomodulatory efficacy. The resulting M2-HP-Apo-sEVs were systematically evaluated in terms of vesicle morphology, size distribution, surface charge, EV markers, PS exposure, cellular uptake, macrophage reprogramming, osteogenic differentiation of bone marrow mesenchymal stem cells, and angiogenic behavior of endothelial cells. Subsequently, M2-HP-Apo-sEVs were incorporated into photocrosslinkable GelMA hydrogels to establish a local vesicle depot for critical-sized calvarial defect repair. This study demonstrates that M2-HP-Apo-sEVs coordinate inflammation resolution, vascular activation, and osteogenic matrix formation, thereby offering a cell-free strategy for recapitulating macrophage-guided bone regeneration. By integrating donor-cell hypoxic programming, apoptotic-mimetic vesicle engineering, and hydrogel-mediated local delivery, this work provides a rational osteoimmunomodulatory platform for the regeneration of critical-sized bone defects.

2. Experimental section

2.1. Cell culture and generation of M2-like reparative macrophages

RAW264.7 murine macrophage-like cells were obtained from ATCC (USA) and cultured in high-glucose Dulbecco's modified Eagle's medium (DMEM; HyClone, Cytiva, USA) supplemented with 10% fetal bovine serum (FBS; HyClone, Cytiva, USA) and 1% penicillin/streptomycin solution (HyClone, Cytiva, USA). Cells were maintained at 37 °C in a humidified atmosphere containing 5% CO₂, and the culture medium was refreshed every 2 d.

To establish a reparative macrophage phenotype, RAW264.7 cells were seeded at an appropriate density and allowed to adhere overnight. The cells were then stimulated with recombinant murine interleukin-4 (IL-4; BioLegend, USA) at 20 ng mL⁻¹. After 24 h of stimulation, cells were collected for quantitative transcriptional analysis of M2-associated genes. After 48 h, cells and culture supernatants were harvested for flow-cytometric phenotyping and cytokine analysis. RAW264.7 cells maintained in complete culture medium without IL-4 served as the non-polarized control.

2.2. Hypoxic preconditioning of M2-like macrophages and small extracellular vesicle isolation

For hypoxic programming, IL-4-polarized M2-like macrophages were washed three times with PBS and transferred to high-glucose DMEM supplemented with 10% extracellular-vesicle-depleted FBS for the 24 h conditioning period. EV-depleted FBS was prepared from HyClone FBS by ultracentrifugation at 100 000 × g for 18 h at 4 °C, followed by collection of the supernatant and filtration through a 0.22 μm membrane. Cells were then placed in a sealed low-oxygen culture system containing oxygen-scavenging sachets (AnaeroPack; Mitsubishi Gas Chemical, Japan) and incubated at 37 °C for 24 h. Normoxic M2-like macrophages maintained in the same EV-depleted collection medium under standard conditions were processed in parallel. HIF-1α accumulation was used as a biological validation of hypoxia-responsive signaling. After the 24 h conditioning period, conditioned media were collected immediately for vesicle purification.

Small extracellular vesicles (sEVs) were isolated by differential centrifugation followed by ultracentrifugation. Briefly, conditioned media were centrifuged at 300 × g for 10 min to remove suspended cells and then at 2000 × g for 10 min to eliminate cellular debris. The resulting supernatant was centrifuged at 10 000 × g for 30 min at 4 °C, and the pellet containing larger vesicles and apoptotic bodies was discarded. The cleared supernatant was subsequently ultracentrifuged at 100 000 × g for 70 min at 4 °C using an Optima XPN ultracentrifuge (Beckman Coulter Life Sciences, USA). The vesicle pellet was washed with sterile PBS and subjected to a second ultracentrifugation under the same conditions. Purified vesicles were resuspended in sterile PBS and stored at −80 °C until use. This workflow yielded an operationally defined sEV-enriched fraction; it was not interpreted as an absolutely pure exosome population, and small apoptosis-associated EVs could not be completely excluded. sEVs derived from normoxic M2-like macrophages were denoted M2-sEVs, whereas those derived from hypoxia-preconditioned M2-like macrophages were denoted M2-HP-Apo-sEVs.

Hypoxic activation of donor macrophages was verified by western blot analysis of HIF-1α. Phosphatidylserine externalization on hypoxia-preconditioned macrophages was examined using a FITC-Annexin V/propidium iodide apoptosis detection kit (BioLegend, USA), followed by fluorescence imaging. Annexin V staining was also used to assess phosphatidylserine exposure on the membrane surface of M2-HP-Apo-sEVs.

2.3. Physicochemical and protein characterization of sEVs

The ultrastructure of isolated sEVs was visualized by transmission electron microscopy (TEM). Vesicle suspensions were dropped onto carbon-coated copper grids, allowed to adsorb, negatively stained, air-dried, and imaged using a JEM-1400Flash transmission electron microscope (JEOL, Japan). Hydrodynamic size distribution, polydispersity index, and ζ-potential were measured using a Zetasizer Ultra system (Malvern Panalytical, UK). Before measurement, samples were diluted in particle-free PBS, and each preparation was analyzed in at least three independent acquisitions.

EV-associated proteins and a negative marker were examined by western blotting. CD81, a canonical tetraspanin transmembrane protein, and ALIX, an ESCRT-associated cytosolic protein commonly enriched in EV preparations, were selected as positive EV-associated markers. CALNEXIN, an endoplasmic-reticulum protein, was used as a negative marker to assess cellular-organelle contamination. In accordance with MISEV2023, marker expression was interpreted together with TEM morphology, particle sizing, particle concentration, ζ-potential, and the particle-to-protein ratio rather than as stand-alone evidence of EV identity [43]. Protein concentration was determined using a bicinchoninic acid assay before equal loading.

To compare vesicle production between normoxic and hypoxia-preconditioned M2 macrophages, conditioned media were collected from parallel cultures seeded at the same initial density and maintained for an identical 24 h collection period. Before medium collection, cell viability and viable cell numbers were recorded to normalize vesicle output to donor-cell number. After removal of cells, debris, and large vesicular contaminants by sequential centrifugation, sEVs were purified by ultracentrifugation as described above and resuspended in an equal volume of sterile PBS.

Particle concentrations were determined by nanoparticle tracking analysis using a NanoSight NS300 system (Malvern Panalytical, Malvern, UK). Samples were diluted in particle-free PBS to achieve an optimal particle concentration for acquisition. Vesicle yield was expressed as particles per milliliter and further normalized to the number of viable donor cells as particles per 10⁶ cells. In parallel, total vesicular protein was quantified using a bicinchoninic acid assay according to the manufacturer's instructions. Absorbance was measured at 562 nm using a multimode microplate reader, and protein concentration was calculated from a bovine serum albumin standard curve. The protein yield was normalized to viable donor-cell number and expressed as μg protein per 10⁶ cells. To further evaluate relative vesicle purity, the particle-to-protein ratio was calculated by dividing the NTA-derived particle number by the corresponding BCA-derived protein amount.

2.4. Fluorescent labeling of sEVs and cellular uptake assay

For cellular internalization studies, purified sEVs were labeled with the lipophilic membrane dye DiO (Biotium, USA). Briefly, vesicles were incubated with DiO under light-protected conditions, and unincorporated dye was removed by ultracentrifugation at 100 000 × g for 70 min at 4 °C, followed by washing with PBS.

RAW264.7 cells, bone marrow mesenchymal stem cells (BMSCs), or human umbilical vein endothelial cells (HUVECs) were incubated with DiO-labeled M2-sEVs or M2-HP-Apo-sEVs at an equivalent vesicle protein concentration of 20 μg mL⁻¹ for 4 h. After incubation, cells were extensively washed with PBS, fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and stained with rhodamine-conjugated phalloidin (Biotium, USA) to visualize F-actin. Nuclei were counterstained using DAPI-containing antifade mounting medium (Vector Laboratories, USA). Fluorescence images were acquired using a STELLARIS confocal microscope platform (Leica Microsystems, Germany). Quantitative uptake was further assessed using a CytoFLEX flow cytometer (Beckman Coulter Life Sciences, USA).

2.5. MerTK knockdown and loss-of-function analysis

To investigate whether MerTK-mediated recognition contributed to the enhanced uptake and immunomodulatory activity of M2-HP-Apo-sEVs, Mertk expression in RAW264.7 macrophages was silenced using small interfering RNA. RAW264.7 cells were seeded in culture plates and transfected at approximately 60–70% confluence with Silencer Select mouse Mertk-targeting siRNA (siMerTK; Thermo Fisher Scientific, USA) or Silencer Select Negative Control No. 1 siRNA (siNC; Thermo Fisher Scientific, USA) at a final concentration of 10 nM using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific, USA) according to the manufacturer's instructions. The siRNA and RNAiMAX reagent were separately diluted in Opti-MEM reduced-serum medium, combined at a 1:1 vol ratio, and incubated for 10 min at room temperature to allow formation of siRNA–lipid complexes. The complexes were subsequently added to the cells. After 6 h, the transfection medium was replaced with fresh complete culture medium. MerTK knockdown efficiency was evaluated by measuring Mertk mRNA expression using quantitative real-time PCR 48 h after transfection.

For functional analysis, RAW264.7 macrophages were transfected with siMerTK or siNC for 48 h, stimulated with LPS at 100 ng mL⁻¹ for 24 h, and subsequently treated with M2-sEVs or M2-HP-Apo-sEVs at 20 μg mL⁻¹ for an additional 48 h. The mRNA expression levels of the pro-inflammatory marker Nos2 and the reparative marker Arg1 were determined by quantitative real-time PCR. The concentrations of IL-1β and TGF-β in the culture supernatants were quantified by ELISA. Non-polarized macrophages and LPS-stimulated macrophages were included as the M0 and M1 controls, respectively. Three independent biological experiments were performed.

2.6. In vitro cytocompatibility evaluation

The cytocompatibility of sEV formulations was assessed in RAW264.7 cells, BMSCs, and HUVECs. Cells were seeded in 96-well plates at 5 × 10³ cells per well and treated with M2-sEVs or M2-HP-Apo-sEVs at 20 μg mL⁻¹ for the indicated time points. Metabolic activity was quantified using a WST-8-based cell proliferation assay kit (Cayman Chemical, USA), and absorbance at 450 nm was recorded using a Spark multimode microplate reader (Tecan, Switzerland).

Apoptotic cell death was evaluated by dual Annexin V-FITC/PI staining. Briefly, cells were incubated with M2-sEVs or M2-HP-Apo-sEVs at a vesicle protein concentration of 20 μg mL⁻¹ for 72 h. After treatment, cells were gently detached, collected by centrifugation, and rinsed with PBS before fluorescent labeling using an Annexin V-FITC/PI apoptosis detection kit in accordance with the manufacturer's protocol. The stained cell suspensions were analyzed using a FACSCanto II flow cytometer (BD Biosciences, USA), and the proportions of viable, early apoptotic, late apoptotic, and PI-positive cells were quantified with FlowJo software.

2.7. Reprogramming of inflammatory macrophages in vitro

To generate an inflammatory macrophage model, RAW264.7 cells were stimulated with lipopolysaccharide (LPS; List Biological Laboratories, USA) at 100 ng mL⁻¹ for 24 h. After LPS priming, cells were treated with M2-sEVs or M2-HP-Apo-sEVs at a vesicle protein concentration of 20 μg mL⁻¹ for an additional 48 h. Untreated and LPS-only groups were included as controls.

Macrophage phenotype switching was evaluated at the mRNA, secreted-protein, and immunofluorescence levels. The expression of pro-inflammatory genes, including Il1b and Nos2, and reparative genes, including Arg1 and Mrc1, was analyzed by quantitative real-time PCR. Secreted cytokines in the culture supernatants were quantified by enzyme-linked immunosorbent assay (ELISA). For immunofluorescence staining, cells were fixed, permeabilized, blocked, and incubated with primary antibodies against CD86 and CD206, followed by fluorophore-conjugated secondary antibodies. Fluorescence intensity was quantified using ImageJ.

2.8. RNA extraction and quantitative real-time PCR

Total RNA was isolated from cultured cells or harvested calvarial defect tissues using a Quick-RNA Miniprep Kit (Zymo Research, USA). RNA quantity and purity were determined by measuring absorbance at 260 and 280 nm. Reverse transcription was performed using a PrimeScript RT reagent system (Takara Bio, Japan). Quantitative PCR was conducted using TB Green-based chemistry (Takara Bio, Japan) on a LightCycler 480 Instrument II real-time PCR system (Roche, Switzerland). Relative gene expression was calculated using the 2⁻ΔΔCt method, with Gapdh used as the endogenous reference gene. Primer sequences are listed in Table S1.

2.9. ELISA analysis of cytokines

Culture supernatants were collected and centrifuged at 1000 × g for 20 min to remove residual cells and debris. Clarified samples were stored at −80 °C until analysis. The concentrations of mouse IL-6, IL-1β, IL-10, and TGF-β were measured using commercially available ELISA kits (BioLegend, USA) according to the manufacturer's instructions. Absorbance was measured using a Spark multimode microplate reader (Tecan, Switzerland), and cytokine concentrations were calculated from standard curves.

2.10. Immunofluorescence staining

Cells cultured on coverslips and paraffin-embedded tissue sections were fixed with 4% paraformaldehyde. Samples were permeabilized with 0.1% Triton X-100 when required and blocked with bovine serum albumin-containing blocking buffer. Primary antibodies against CD86, CD206, and CD31 were obtained from BioLegend (USA), and the primary antibody against BMP2 was obtained from Proteintech (USA). Samples were incubated with primary antibodies overnight at 4 °C, followed by incubation with species-matched fluorophore-conjugated secondary antibodies. Nuclei were counterstained with DAPI-containing mounting medium. Images were captured using a STELLARIS confocal microscope platform (Leica Microsystems, Germany), and fluorescence intensity was analyzed with ImageJ under identical acquisition and thresholding parameters.

2.11. BMSC culture, phenotypic characterization, and osteogenic differentiation

Primary murine BMSCs were obtained from Cell Biologics (USA) and expanded in α-minimum essential medium supplemented with 10% FBS and 1% penicillin/streptomycin. Cells were cultured at 37 °C under 5% CO₂, and the medium was replaced every 2 d. Cells from passages 3–5 were used for subsequent experiments.

The mesenchymal phenotype of BMSCs was confirmed by flow cytometry. Cells were stained with antibodies against CD29 and CD44 as positive mesenchymal markers and against CD31 and CD45 as endothelial/hematopoietic lineage markers. Data were acquired using a CytoFLEX flow cytometer.

For osteogenic induction, BMSCs were seeded in 24-well plates and cultured in osteogenic medium containing ascorbic acid, β-glycerophosphate, and dexamethasone. Where indicated, M2-sEVs or M2-HP-Apo-sEVs were added at 20 μg mL⁻¹. Early osteogenic differentiation was evaluated after 7 d by alkaline phosphatase (ALP) staining and ALP activity measurement using a colorimetric ALP assay kit (Abcam, UK). Matrix mineralization was assessed after 14 d by Alizarin Red S staining (Merck, Germany). Bound Alizarin Red S was extracted using cetylpyridinium chloride and quantified spectrophotometrically. The expression of osteogenesis-related genes, including Alpl, Runx2, Opn, Ocn, and Bmp2, was quantified by qRT-PCR. BMP2 protein expression was visualized by immunofluorescence staining.

2.12. EdU proliferation assay

HUVEC proliferation was evaluated using an EdU incorporation assay. HUVECs were seeded in 96-well plates at 5 × 10³ cells per well, allowed to adhere overnight, and treated with PBS, M2-sEVs, or M2-HP-Apo-sEVs at 20 μg mL⁻¹ for 24 h. Cells were then incubated with 10 μM EdU for 2 h, fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton X-100 for 15 min, and subjected to the click-chemistry detection reaction according to the kit instructions. Nuclei were counterstained with DAPI. Images were acquired using an inverted fluorescence microscope under identical exposure settings. Five randomly selected fields per well were analyzed with ImageJ, and proliferation was expressed as the percentage of EdU-positive nuclei among total DAPI-positive nuclei. Three independent experiments were performed.

2.13. Tube formation assay

The pro-angiogenic activity of sEVs was evaluated using a basement membrane extract-based tube formation assay. Growth factor-reduced basement membrane extract (Cultrex RGF BME; R&D Systems, USA) was thawed on ice, added to pre-chilled 48-well plates, and allowed to polymerize at 37 °C for 30 min. HUVECs obtained from PromoCell (Germany) were pretreated with PBS, M2-sEVs, or M2-HP-Apo-sEVs at 20 μg mL⁻¹ for 4 h and then seeded onto the polymerized matrix. After 4 h of incubation, capillary-like structures were imaged using an inverted microscope. Tube network parameters, including total nodes, junctions, and branches, were quantified using the Angiogenesis Analyzer plugin in ImageJ.

2.14. Transwell migration assay

Endothelial migration was assessed using 24-well Transwell chambers with 8 μm pore-size polycarbonate membranes (Greiner Bio-One, Germany). HUVECs suspended in serum-free medium were seeded into the upper chamber, whereas complete endothelial medium containing 10% FBS was added to the lower chamber as a chemoattractant. After 8 h of incubation at 37 °C, cells remaining on the upper surface of the membrane were carefully removed with a cotton swab. Migrated cells on the lower membrane surface were fixed with 4% paraformaldehyde and stained with 0.2% crystal violet (Merck, Germany). Images were captured using an inverted microscope, and migrated cells were counted in randomly selected fields using ImageJ.

2.15. Fabrication of GelMA-Based sEV delivery depots

Sterile methacrylated gelatin (GelMA; PhotoGel, Advanced BioMatrix, USA) was dissolved in sterile PBS at 5% w/v under aseptic conditions. Lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP; CELLINK, Sweden) was added at a final concentration of 0.25% w/v. M2-sEVs or M2-HP-Apo-sEVs were gently dispersed into the precursor immediately before gelation to minimize aggregation and loss at the liquid–air interface. Each construct contained 20 μL precursor and was formed in a cylindrical mold measuring 5 mm in diameter and 1 mm in height. For in vivo implantation, the vesicle loading amount was normalized to 5 × 10¹¹ particles per construct according to nanoparticle tracking analysis.

The precursor solution was photocrosslinked under sterile conditions using a 365 nm LED curing system (Excelitas Technologies, Canada) at 10 mW cm⁻² for 60 s. For in situ implantation, 20 μL precursor was introduced into the 5 mm calvarial defect and crosslinked using the same irradiation parameters. All preparation steps were performed in a class II biosafety cabinet using sterile commercial GelMA, sterile PBS, sterile single-use consumables, and sterilized molds and instruments; no terminal sterilization was applied after sEV incorporation. Preformed hydrogels used for in vitro experiments were washed with sterile PBS before use.

For microstructural analysis, freeze-dried hydrogel samples were sputter-coated with a conductive layer and examined using a GeminiSEM 360 field-emission scanning electron microscope (ZEISS, Germany). Hydrogel cytocompatibility was assessed by incubating RAW264.7 cells, HUVECs, and BMSCs with hydrogel-conditioned media, followed by WST-8 analysis.

2.16. In vitro release of sEVs from GelMA hydrogels

Because GelMA is a proteinaceous matrix that may interfere with protein-based quantification of vesicle release, sEV release was evaluated using a fluorescence-tracking method. M2-sEVs and M2-HP-Apo-sEVs were labeled with DiO, purified by ultracentrifugation to remove unincorporated dye, and incorporated into GelMA hydrogels as described above.

To evaluate vesicle release under enzymatic degradation conditions, each sEV-loaded hydrogel was immersed in 1 mL of PBS containing collagenase type I at 0.01 U mL⁻¹ and incubated at 37 °C under gentle shaking. This collagenase-containing medium was used as a simplified proteolytic degradation condition to evaluate sEV release during enzyme-mediated degradation of the GelMA network.

At predetermined time points of 1, 3, 5, and 7 d, the entire release medium was collected and replaced with an equal volume of freshly prepared collagenase-containing PBS. Fluorescence intensity in the collected supernatants was measured using a Spark multimode microplate reader at excitation and emission wavelengths appropriate for DiO. A standard curve generated from serial dilutions of the same DiO-labeled sEV preparation in collagenase-containing PBS was used to convert fluorescence intensity into the amount of released vesicles. Cumulative release was calculated as follows:

$$\text{Cumulative}\text{release}(\%)=\text{cumulative}\text{released}\text{sEV}\text{amount}/\text{total}\text{loaded}\text{sEV}\text{amount}\times 100\%$$

Blank GelMA hydrogels without labeled sEVs, dye-only controls, and collagenase-containing medium controls were processed under identical conditions to correct for hydrogel autofluorescence, medium-associated background, and potential free-dye contamination. Three independently prepared hydrogels were analyzed for each sEV formulation.

2.17. Murine critical-sized calvarial defect model

This project was supervised and approved by the Ethics Committee for Laboratory Animal Welfare of Wuhan Lobin Life Science and Technology Company Limited (Approval No.: LBSM2025093) and was reported with reference to ARRIVE 2.0 [44]. Eighteen male BALB/c mice (8 weeks old; Charles River Laboratories, USA) were acclimatized for 7 d and allocated by a computer-generated random sequence to GelMA, GelMA–M2-sEVs, or GelMA–M2-HP-Apo-sEVs groups (n = 6 mice per group; one animal constituted one experimental unit). Anesthesia was induced with 3% isoflurane and maintained with 1.5–2.0% isoflurane in oxygen. A midline scalp incision was made under sterile conditions, and a unilateral full-thickness 5 mm circular calvarial defect was created using a sterile trephine bur under continuous irrigation while preserving the dura. The assigned 20 μL hydrogel precursor was introduced into the defect and photocrosslinked in situ as described above, after which the periosteum and skin were closed in layers. The surgeon was aware of group allocation because the formulations were prepared separately; however, micro-CT reconstruction, histological and immunofluorescence quantification, and statistical analysis were performed by investigators blinded to coded group identities. No animals were excluded from analysis. At 12 weeks, mice were euthanized by CO₂ inhalation followed by cervical dislocation to confirm death. All six specimens per group underwent micro-CT. Three specimens per group were randomly assigned to histology and tissue immunofluorescence, and the remaining three were used for tissue RNA extraction and qRT-PCR.

2.18. Micro-computed tomography analysis

Regenerated bone within the calvarial defect region was evaluated using a SkyScan 1276 micro-computed tomography system (Bruker, Belgium). Specimens were scanned under standardized acquisition parameters. Three-dimensional reconstruction was performed using manufacturer-provided software, and a cylindrical region of interest corresponding to the original defect area was defined for quantitative analysis. Bone volume fraction (BV/TV), bone mineral density (BMD), trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular separation (Tb.Sp) were calculated to assess bone regeneration.

2.19. Histology and tissue immunofluorescence

After micro-CT scanning, calvarial samples were fixed in 4% paraformaldehyde, decalcified in EDTA solution at 4 °C, dehydrated, embedded in paraffin, and sectioned at 5 μm thickness. Hematoxylin and eosin staining was performed to evaluate tissue bridging and defect closure. Masson's trichrome staining was used to visualize collagen deposition and newly formed bone matrix.

For tissue-level immunofluorescence, sections were deparaffinized, rehydrated, subjected to antigen retrieval when required, permeabilized, and blocked. Sections were incubated with primary antibodies against CD86, CD206, CD31, and BMP2 overnight at 4 °C, followed by fluorophore-conjugated secondary antibodies. Nuclei were counterstained with DAPI. Fluorescence images were acquired using a confocal microscope, and signal intensity within the defect region was quantified using ImageJ.

For systemic histological evaluation, the heart, liver, spleen, lung, and kidney were harvested at 12 weeks after implantation and fixed in 4% paraformaldehyde. The tissues were dehydrated, embedded in paraffin, sectioned at 5 μm, and stained with hematoxylin and eosin. Representative sections were examined by bright-field microscopy for overt histopathological abnormalities.

2.20. Statistical analysis

All quantitative data are presented as mean ± standard deviation (SD). In vitro analyses included three independent biological experiments unless otherwise stated; the independent experiment, rather than individual microscopic fields or cells, was treated as the statistical unit. For in vivo analyses, the individual mouse was the experimental unit. Statistical analyses were performed using GraphPad Prism software. Two-group comparisons, including the siNC-versus-siMerTK comparison of Mertk mRNA expression, were analyzed using an unpaired two-tailed Student's t-test. Comparisons among three or more groups were conducted by one-way analysis of variance followed by Tukey's post hoc test. A p value < 0.05 was considered statistically significant.

3. Results and discussion

3.1. Hypoxic preconditioning endows M2 macrophage-derived sEVs with an apoptotic-mimetic and uptake-favorable membrane phenotype

To construct a pro-reparative vesicle system with enhanced biological recognition, RAW264.7 macrophages were first polarized toward an M2-like phenotype and subsequently exposed to hypoxic preconditioning. IL-4 stimulation efficiently drove macrophages toward a reparative state, as demonstrated by the pronounced upregulation of Mrc1 and Arg1, together with increased secretion of the anti-inflammatory and pro-repair cytokines IL-10 and TGF-β (Fig. 1A and B). Consistently, flow-cytometric analysis revealed elevated CD206 expression after IL-4 treatment, confirming successful phenotypic polarization at the protein level (Fig. 1C).

Fig. 1.

Preparation and physicochemical characterization of M2-sEVs and M2-HP-Apo-sEVs. (A) Relative expression of the M2-associated transcripts Mrc1 and Arg1 after 24 h of IL-4 stimulation. (B) ELISA quantification of IL-10 and TGF-β secretion after 48 h of polarization. (C) Flow-cytometric analysis of CD206 expression in IL-4-treated macrophages; the orange filled curves indicate M0-RAW264.7 cells, and the purple filled curves indicate M2-RAW264.7 cells. (D) Western blot analysis of HIF-1α in M2 macrophages after hypoxic preconditioning. (E) Representative flow-cytometric analysis of Annexin V-FITC/PI staining in IL-4-polarized M2 macrophages exposed to hypoxia for 0, 6, 12, 24, and 48 h. (F) Representative fluorescence images of hypoxia-preconditioned M2 macrophages stained with Annexin V-FITC, PI, and DiD. Annexin V-FITC signals indicate phosphatidylserine externalization, PI staining reflects membrane-compromised cells, and DiD was used to delineate the plasma membrane. At 24 h, Annexin V-FITC⁺/PI⁻ cells predominated over Annexin V-FITC⁺/PI⁺ cells, indicating substantial apoptosis-associated phosphatidylserine externalization while most Annexin V-positive cells retained membrane integrity. (G) Confocal visualization of phosphatidylserine exposure on M2-HP-Apo-sEVs using FITC-Annexin V staining. (H) Representative transmission electron microscopy images of M2-sEVs and M2-HP-Apo-sEVs. (I) NTA-based quantification of vesicle particle yield normalized to viable donor-cell number; BCA-based quantification of total vesicular protein yield normalized to donor-cell number; and particle-to-protein ratio of the indicated sEV formulations. (J) Polydispersity index of M2-sEVs and M2-HP-Apo-sEVs. (K) ζ-potential analysis of the indicated vesicles. (L) Western blot detection of the EV-associated markers CD81 and ALIX and the endoplasmic-reticulum-associated negative marker CALNEXIN. Data are presented as mean ± SD (n = 3 independent preparations). *p < 0.05, **p < 0.01. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

After M2 polarization, low-oxygen preconditioning was introduced to remodel the donor macrophage state before sEV collection. Western blotting showed marked HIF-1α accumulation, supporting activation of a hypoxia-responsive program (Fig. 1D). Annexin V-FITC/PI analysis showed a time-dependent increase in apoptosis-associated phosphatidylserine externalization. At 24 h, 50.6% of cells were Annexin V⁺/PI⁻ and 11.3% were Annexin V⁺/PI⁺, demonstrating substantial apoptotic involvement rather than a purely non-apoptotic mimicry state, although the Annexin V⁺/PI⁻ population remained predominant (Fig. 1E). Fluorescence imaging similarly showed strong Annexin V labeling with comparatively limited PI uptake (Fig. 1F). Because phosphatidylserine is an “eat-me” lipid signal recognized through efferocytic pathways, its externalization provides a plausible interface for macrophage recognition and immunoregulatory signaling. The contribution of a phosphatidylserine-associated MerTK pathway was subsequently examined using siMerTK, although direct phosphatidylserine masking or competition was not performed. This interpretation is consistent with studies showing that apoptotic or hypoxia-preconditioned EVs can acquire altered regenerative and immunomodulatory properties [27,33].

Confocal imaging further demonstrated Annexin V-positive signals on purified M2-HP-Apo-sEVs, confirming that the phosphatidylserine-enriched membrane feature was retained at the vesicle level (Fig. 1G). TEM analysis showed that both M2-sEVs and M2-HP-Apo-sEVs exhibited rounded or cup-shaped nanoscale vesicular structures with clearly discernible membrane boundaries (Fig. 1H). Hydrodynamic size-distribution analysis showed average diameters of approximately 148 and 151 nm for M2-sEVs and M2-HP-Apo-sEVs, respectively (Fig. S1). Together with the sequential 2000 × g and 10 000 × g pre-clearing steps, these findings support enrichment of a predominantly nanoscale vesicle fraction and depletion of many detectable larger particles before the final 100 000 × g collection. Nevertheless, TEM sampling and differential ultracentrifugation cannot completely exclude residual larger particles or smaller apoptosis-associated EV subpopulations with overlapping physical characteristics. Hypoxic preconditioning not only endowed M2 macrophage-derived sEVs with an apoptotic-mimetic membrane phenotype but also increased their production yield. NTA revealed a significantly higher particle output from hypoxia-preconditioned M2 macrophages than from normoxic M2 macrophages (Fig. 1I). BCA-based protein quantification further confirmed an increased vesicular protein recovery in the M2-HP-Apo-sEV group (Fig. 1I). Notably, the particle-to-protein ratio remained comparable between the two vesicle formulations (Fig. 1I), indicating that the enhanced yield was primarily attributable to increased vesicle release rather than nonspecific protein co-isolation. This observation is mechanistically plausible because hypoxia is increasingly recognized as a potent regulator of EV biogenesis and secretion. Hypoxic signaling can influence endosomal trafficking, multivesicular body turnover, lysosomal degradation, and HIF-associated secretory programs, thereby shifting the balance toward increased extracellular vesicle release [45,46]. In the context of tissue repair, hypoxia-preconditioned EVs have been incorporated into bioactive hydrogels to enhance bone or osteochondral regeneration, highlighting hypoxic donor-cell programming as an effective approach for improving EV-based regenerative therapies [33,41]. Therefore, the increased vesicle yield observed here provides an additional translational advantage for M2-HP-Apo-sEVs, enabling greater vesicle recovery while preserving the functional membrane features required for macrophage-targeted immunomodulation.

Both vesicle formulations exhibited low PDI values, indicating favorable colloidal homogeneity (Fig. 1J). Notably, hypoxic preconditioning further increased the negative surface potential of the vesicles (Fig. 1K), which may reflect altered membrane lipid composition and surface-associated biomolecules. Western blotting confirmed the presence of the EV-associated proteins CD81 and ALIX and the absence of CALNEXIN, supporting enrichment of an sEV-containing fraction with limited detectable endoplasmic-reticulum contamination (Fig. 1L).

Collectively, these data demonstrate that hypoxic preconditioning does not merely alter the donor macrophage state, but also imprints a distinct membrane phenotype onto the released sEVs. The resulting M2-HP-Apo-sEVs integrate three design features that are particularly relevant to bone immunoregeneration: a reparative macrophage origin, hypoxia-responsive donor-cell programming, and phosphatidylserine-associated apoptotic mimicry. This combination provides a plausible biological basis for their subsequent preferential internalization and amplified immunomodulatory activity.

3.2. M2-HP-Apo-sEVs Potently reprogram inflammatory macrophages toward a pro-resolving phenotype

Efficient entry into target cells is a prerequisite for sEV-mediated regulation. Confocal microscopy revealed that both vesicle formulations could be internalized by RAW264.7 macrophages; however, M2-HP-Apo-sEVs produced a visibly stronger intracellular DiO signal than normoxic M2-sEVs (Fig. 2A). Flow-cytometric quantification further confirmed this enhanced uptake profile (Fig. 2B). The preferential internalization of M2-HP-Apo-sEVs may plausibly relate, at least in part, to phosphatidylserine-rich membrane remodeling and recognition through macrophage-associated phagocytic or efferocytic pathways. This possibility was further examined using partial MerTK knockdown (Fig. S3). Because MerTK silencing was incomplete and direct phosphatidylserine blockade was not performed, the data support the contribution of a phosphatidylserine-associated MerTK pathway but do not establish phosphatidylserine as the sole or indispensable determinant of vesicle uptake. Such an uptake advantage is particularly relevant for immunomodulatory vesicles, because macrophages are both the intended target cell and a central regulator of the osteoimmune microenvironment.

Fig. 2.

Preferential macrophage uptake and immunoregulatory activity of M2-HP-Apo-sEVs in vitro. (A) Representative confocal images of RAW264.7 macrophages after incubation with DiO-labeled M2-sEVs or M2-HP-Apo-sEVs. DiO-labeled vesicles are shown in green, F-actin is stained with rhodamine-phalloidin, and nuclei are counterstained with DAPI. (B) Flow-cytometric quantification of vesicle internalization by RAW264.7 macrophages; the red filled curves mean M2-sEVs and the blue filled curves represent M2-HP-Apo-sEVs. (C) WST-8-based metabolic activity assay of macrophages after exposure to the indicated vesicle formulations. (D) Apoptosis assay using Annexin V-FITC/PI staining. (E) qRT-PCR analysis of pro-inflammatory genes (Il1b, Nos2) and reparative genes (Arg1, Mrc1) in LPS-primed macrophages after vesicle treatment. (F) ELISA quantification of IL-6, IL-1β, IL-10, and TGF-β in macrophage culture supernatants. (G) Immunofluorescence staining of CD86 and CD206 in macrophages subjected to the indicated treatments. Data are presented as mean ± SD (n = 3); *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Before evaluating immunological function, the cytocompatibility of both vesicle formulations was assessed. Neither M2-sEVs nor M2-HP-Apo-sEVs impaired macrophage metabolic activity at the tested dose (Fig. 2C). Annexin V-FITC/PI staining revealed that the majority of cells remained viable, with only minimal PI-positive signals, indicating that vesicle treatment did not trigger obvious apoptosis or membrane-compromised cell death (Fig. 2D). These findings support that the subsequent phenotypic shifts were not secondary to nonspecific cell toxicity.

LPS stimulation generated a robust inflammatory macrophage state, characterized by elevated expression of Il1b and Nos2. Treatment with M2-sEVs significantly attenuated this inflammatory signature, whereas M2-HP-Apo-sEVs produced a stronger suppression of Il1b and Nos2 while concomitantly increasing Arg1 and Mrc1 expression (Fig. 2E). At the secreted-protein level, vesicle treatment reduced the release of IL-6 and IL-1β and increased IL-10 and TGF-β, with the most pronounced cytokine rebalance observed in the M2-HP-Apo-sEV group (Fig. 2F). Immunofluorescence staining further supported this phenotypic conversion, showing reduced CD86 and enhanced CD206 expression following vesicle treatment (Fig. 2G); the corresponding fluorescence-intensity quantification is provided in Fig. S2.

These results indicate that M2-HP-Apo-sEVs do not simply suppress inflammatory activation; rather, they actively reshape macrophages toward a pro-resolving, tissue-reparative state. This distinction is important because successful bone healing requires a timely transition from inflammatory debris clearance to regenerative matrix deposition, vascularization, and osteoprogenitor recruitment. Recent studies have similarly shown that M2 macrophage-derived exosomes can restore osteoimmune balance and accelerate fracture healing by reducing inflammatory macrophage polarization and promoting reparative immune signaling [27]. In the present system, hypoxic apoptotic-mimetic programming further amplifies this M2-sEV-mediated immunoregulatory effect, suggesting that donor-cell preconditioning and vesicle membrane engineering can be integrated to enhance macrophage-targeted regenerative communication.

To determine whether MerTK contributed to the preferential macrophage recognition of M2-HP-Apo-sEVs, a loss-of-function experiment was performed using Mertk-targeting siRNA. Transfection with siMerTK reduced Mertk mRNA expression to approximately 50% of the siNC level, confirming partial silencing in RAW264.7 macrophages (Fig. S3A). Immunofluorescence results further confirmed decreased MERTK protein expression in siMerTK group (Fig. S3B and S3C). Following partial MerTK silencing, the DiO fluorescence profiles of macrophages treated with M2-sEVs and M2-HP-Apo-sEVs largely overlapped, indicating attenuation of the preferential uptake observed for M2-HP-Apo-sEVs (Fig. S3D). In siMerTK-transfected, LPS-stimulated macrophages, the two vesicle formulations produced comparable effects on Nos2 and Arg1 expression, with no significant differences between the M2-sEV and M2-HP-Apo-sEV groups (Fig. S3E). Similarly, IL-1β and TGF-β concentrations did not differ significantly between the two vesicle-treated groups after MerTK silencing (Fig. S3F). Collectively, these loss-of-function findings indicate that partial MerTK silencing attenuated the relative uptake and immunomodulatory advantage of M2-HP-Apo-sEVs, supporting the contribution of phosphatidylserine-associated MerTK recognition. The residual activity of both vesicle formulations suggests that additional uptake pathways and intravesicular cargo may also contribute.

3.3. M2-HP-Apo-sEVs enhance osteogenic commitment and matrix mineralization of BMSCs

The osteogenic response of BMSCs was next examined to determine whether the immunoregulatory vesicles could directly influence osteoprogenitor behavior. Flow-cytometric analysis confirmed that the BMSCs displayed a typical mesenchymal phenotype, with positive expression of CD29 and CD44 and negligible expression of CD31 and CD45 (Fig. 3A). Confocal imaging showed that both M2-sEVs and M2-HP-Apo-sEVs were efficiently internalized by BMSCs (Fig. 3B), and flow cytometry further confirmed measurable vesicle uptake in recipient cells (Fig. 3C). Importantly, vesicle treatment did not compromise BMSC viability, as demonstrated by metabolic activity assays and apoptosis assay using Annexin V-FITC/PI staining (Fig. 3D and E).

Fig. 3.

M2-HP-Apo-sEVs promote BMSC osteogenic differentiation in vitro. (A) Flow-cytometric characterization of BMSC surface markers, including CD29, CD44, CD31, and CD45. (B) Representative confocal images showing internalization of DiO-labeled M2-sEVs or M2-HP-Apo-sEVs by BMSCs. Vesicles are shown in green, F-actin is stained with rhodamine-phalloidin, and nuclei are counterstained with DAPI. (C) Flow-cytometric quantification of vesicle uptake by BMSCs; the red filled curves mean M2-sEVs and the blue filled curves represent M2-HP-Apo-sEVs. (D) Cell metabolic activity of BMSCs after treatment with the indicated vesicle formulations. (E) Apoptosis assay using Annexin V-FITC/PI staining. (F) ALP staining of BMSCs after 7 d of osteogenic induction. (G) Quantitative analysis of ALP activity. (H) Alizarin Red S staining of mineralized extracellular matrix after 14 d of osteogenic induction. (I) Spectrophotometric quantification of extracted Alizarin Red S staining. (J) qRT-PCR analysis of osteogenesis-related genes, including Alpl, Runx2, Opn, Ocn, and Bmp2. (K) Immunofluorescence staining of BMP2 in differentiated BMSCs. (L) Quantification of BMP2 fluorescence intensity. Data are expressed as mean ± SD (n = 3); *p < 0.05, **p < 0.01, ***p < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

During early osteogenic induction, M2-sEVs enhanced ALP staining intensity relative to the untreated osteogenic control, indicating accelerated early lineage commitment. This effect was further strengthened by M2-HP-Apo-sEVs, which produced the highest ALP staining and enzymatic activity among all groups (Fig. 3F and G). At a later stage of differentiation, Alizarin Red S staining revealed increased mineralized nodule formation after vesicle treatment, with M2-HP-Apo-sEVs again showing the strongest matrix mineralization capacity (Fig. 3H and I). These results suggest that M2-HP-Apo-sEVs promote not only early osteoblastic commitment but also subsequent extracellular matrix mineral deposition.

At the transcriptional level, vesicle treatment upregulated multiple osteogenesis-associated genes, including Alpl, Runx2, Opn, Ocn, and Bmp2 (Fig. 3J). The coordinated induction of Runx2 and Alpl supports enhanced osteogenic specification, whereas the increase in Opn and Ocn indicates progression toward matrix maturation and mineralization. BMP2 immunofluorescence further confirmed elevated osteoinductive protein expression, with the strongest BMP2 signal detected in the M2-HP-Apo-sEV group (Fig. 3K and L).

These findings support a direct pro-osteogenic role of macrophage-derived sEVs, particularly after hypoxic apoptotic-mimetic programming. Recent high-impact studies have shown that EVs can function as cell-free osteoinductive modules when integrated with biomaterial delivery systems, either by transferring osteogenic cargo or by remodeling the local immune niche that indirectly supports osteogenesis [24,25]. In the present study, M2-HP-Apo-sEVs appear to act through both routes: they directly enhance BMSC osteogenic differentiation in vitro and, as shown above, reprogram inflammatory macrophages into a pro-repair phenotype. This dual activity is especially valuable for bone defect repair, where osteogenesis is highly dependent on the surrounding immune milieu rather than being governed solely by osteoprogenitor-intrinsic differentiation programs.

3.4. M2-HP-Apo-sEVs promote endothelial migration and capillary-like network formation

Vascular ingrowth is indispensable for the repair of critical-sized bone defects because newly formed blood vessels provide oxygen, nutrients, osteoprogenitor recruitment routes, and paracrine cues required for bone matrix deposition. Therefore, the effects of the vesicle formulations on endothelial cell behavior were investigated. Confocal microscopy demonstrated efficient internalization of DiO-labeled M2-sEVs and M2-HP-Apo-sEVs by HUVECs, with vesicle-derived fluorescence distributed throughout the cytoplasmic region (Fig. 4A). Flow cytometry further confirmed vesicle uptake at the population level (Fig. 4B). Neither vesicle formulation elicited detectable cytotoxicity or apoptotic responses, as evidenced by preserved metabolic activity and negligible Annexin V-FITC/PI-positive apoptotic populations (Fig. 4C and D). EdU staining further demonstrated a pronounced increase in proliferating HUVECs following M2-HP-Apo-sEVs treatment, in sharp contrast to the weaker proliferative responses observed in the control and M2-sEV groups (Fig. 4E and F). These findings suggest that M2-HP-Apo-sEVs reinforce the angiogenic competence of endothelial cells by promoting cell-cycle entry and expanding the available endothelial cell pool, thereby creating a favorable cellular basis for subsequent vascular assembly.

Fig. 4.

Pro-angiogenic activity of M2-sEVs and M2-HP-Apo-sEVs in HUVECs. (A) Representative confocal images showing internalization of DiO-labeled M2-sEVs and M2-HP-Apo-sEVs by HUVECs. Vesicles are shown in green, F-actin is stained with rhodamine-phalloidin, and nuclei are stained with DAPI. (B) Flow-cytometric quantification of vesicle uptake by HUVECs; the red filled curves mean M2-sEVs and the blue filled curves represent M2-HP-Apo-sEVs. (C) Cell metabolic activity of HUVECs after treatment with the indicated vesicle formulations. (D) Apoptosis assay using Annexin V-FITC/PI staining. (E) EdU proliferation assay. (F) Relative proportion of EdU-positive cells in Fig. 4E. (G) Representative images of Transwell migration assays. (H) Quantitative analysis of migrated HUVECs. (I) Representative images of capillary-like network formation on Cultrex RGF BME. (J) Quantitative analysis of angiogenic network parameters, including nodes, junctions, and branches. Data are presented as mean ± SD (n = 3 independent biological experiments); **p < 0.01, ***p < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

In Transwell migration assays, HUVECs treated with M2-sEVs displayed enhanced motility compared with the untreated control group. M2-HP-Apo-sEVs further increased the number of migrated cells, suggesting a stronger capacity to activate endothelial migratory behavior (Fig. 4G and H). In the Cultrex RGF BME tube formation assay, untreated HUVECs generated relatively sparse and discontinuous capillary-like structures, whereas vesicle-treated cells formed denser and more interconnected networks (Fig. 4I). Quantitative analysis confirmed significant increases in nodes, junctions, and branches after vesicle treatment, with M2-HP-Apo-sEVs exhibiting superior pro-angiogenic activity (Fig. 4J).

The stronger angiogenic response induced by M2-HP-Apo-sEVs aligns with the broader concept of immuno–angiogenic–osteogenic coupling in bone repair. Macrophages can regulate vascular morphogenesis through both soluble factors and EV-mediated communication, while vascular endothelial cells reciprocally support osteogenesis by establishing a permissive regenerative niche. Recent biomaterial studies have emphasized that effective bone-regenerative platforms should not treat osteogenesis, angiogenesis, and immune regulation as isolated modules; rather, these processes must be orchestrated as a temporally coordinated repair axis [7,18]. In this context, M2-HP-Apo-sEVs are particularly attractive because they simultaneously attenuate inflammatory macrophage activation, promote BMSC osteogenesis, and enhance endothelial migration and network formation. This macrophage–angiogenesis coupling is in line with recent evidence showing that bioactive glass-elicited stem cell-derived EVs promote tissue repair by enhancing M2 macrophage polarization and macrophage-mediated angiogenesis [26].

3.5. GelMA-mediated local delivery of M2-HP-Apo-sEVs promotes immunoregenerative bone repair in vivo

To translate the in vitro bioactivity of M2-HP-Apo-sEVs into a local regenerative therapy, vesicles were incorporated into a photocrosslinkable GelMA hydrogel. SEM imaging showed that blank GelMA and vesicle-loaded GelMA constructs possessed an interconnected porous architecture (Fig. 5A), which is favorable for nutrient diffusion, cellular infiltration, and tissue ingrowth. Importantly, vesicle incorporation did not visibly disrupt the porous hydrogel network, suggesting that the sEV payload was compatible with GelMA photocrosslinking. In vitro release analysis in collagenase-containing medium demonstrated progressive release of M2-sEVs and M2-HP-Apo-sEVs from the GelMA hydrogel over 7 d (Fig. 5B). The two vesicle formulations exhibited generally comparable release profiles, indicating that release behavior was predominantly governed by enzymatic degradation and remodeling of the GelMA network rather than by the type of encapsulated sEVs. These findings demonstrate that GelMA retained a sustained vesicle-delivery capacity under protease-mediated degradation conditions relevant to the tissue microenvironment. Hydrogel-conditioned media showed good cytocompatibility toward macrophages, endothelial cells, and BMSCs, confirming that neither the GelMA matrix nor the vesicle-loaded constructs introduced overt cytotoxicity (Fig. 5C).

Fig. 5.

GelMA-based local delivery of M2-HP-Apo-sEVs enhances calvarial bone regeneration. (A) Scanning electron microscopy images of freeze-dried GelMA, GelMA–M2-sEVs, and GelMA–M2-HP-Apo-sEVs, showing interconnected porous microstructures. (B) Cumulative release profiles of DiO-labeled M2-sEVs and M2-HP-Apo-sEVs from GelMA hydrogels incubated in collagenase-containing PBS over 7 d. (C) Metabolic activity of RAW264.7 cells, BMSCs, and HUVECs after exposure to hydrogel-conditioned media, indicating favorable cytocompatibility of the vesicle-loaded hydrogel constructs. (D) Representative three-dimensional micro-CT reconstructions of 5 mm critical-sized calvarial defects at 12 weeks after implantation of the indicated formulations; six independent animals per group underwent micro-CT (n = 6). (E) Quantitative micro-CT analysis of bone volume fraction (BV/TV), bone mineral density (BMD), trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular separation (Tb.Sp) within the defect region. (F) qRT-PCR analysis of osteogenesis-associated genes (Runx2, Opn, Bmp2, and Ocn). (G) qRT-PCR analysis of angiogenesis-related genes (Pecam1 and Vegfa). (H) qRT-PCR analysis of inflammatory and reparative macrophage-associated genes (Il1b, Nos2, Mrc1, and Arg1). qRT-PCR data are presented as mean ± SD (n = 3 independent animals per group); *p < 0.05, **p < 0.01, ***p < 0.001.

In a murine critical-sized calvarial defect model, the therapeutic performance of the hydrogel constructs was assessed 12 weeks after implantation. Micro-CT reconstruction revealed limited spontaneous bone regeneration in the GelMA-only group, with incomplete mineral bridging across the defect. Incorporation of M2-sEVs substantially increased new bone formation, whereas GelMA–M2-HP-Apo-sEVs produced the most extensive defect filling and radiographic bridging (Fig. 5D). Quantitative micro-CT analysis further confirmed that GelMA–M2-HP-Apo-sEVs significantly improved BV/TV, BMD, Tb.N, and Tb.Th, while reducing Tb.Sp (Fig. 5E). These structural indices collectively indicate that M2-HP-Apo-sEV delivery not only increased the amount of regenerated bone but also improved the microarchitectural organization of the newly formed tissue.

Molecular analysis of regenerated calvarial tissues provided further evidence that M2-HP-Apo-sEVs remodeled the defect microenvironment toward a pro-regenerative state. The expression of inflammatory genes Il1b and Nos2 was reduced after vesicle-loaded hydrogel treatment, whereas reparative macrophage-associated genes Arg1 and Mrc1 were increased, with the strongest shift observed in the GelMA–M2-HP-Apo-sEV group (Fig. 5H). Angiogenesis-related transcripts, including Pecam1 and Vegfa, were also upregulated (Fig. 5G), consistent with the in vitro endothelial migration and tube formation results. In parallel, osteogenesis-associated genes Runx2, Opn, Ocn, and Bmp2 were markedly enhanced in the regenerated tissue (Fig. 5F).

These in vivo results suggest that GelMA–M2-HP-Apo-sEVs establish an immunoregenerative niche rather than acting as a simple osteogenic supplement. The superior in vivo performance of GelMA–M2-HP-Apo-sEVs also echoes the emerging principle that inflammation-responsive biomaterial interfaces can sequentially regulate the immune-to-osteogenic transition and thereby enhance bone regeneration and osseointegration [18]. GelMA provides a permissive local retention platform, whereas M2-HP-Apo-sEVs deliver instructive biological cues that coordinate macrophage phenotype resolution, vascular activation, and osteogenic matrix formation. This interpretation is consistent with recent EV–hydrogel strategies showing that hydrogel-mediated vesicle retention can improve local regenerative efficacy by extending vesicle bioavailability and restricting activity to the defect microenvironment [24,25]. Compared with conventional growth factor delivery, this vesicle-based approach may offer a broader bioactive spectrum, because sEVs carry membrane ligands, proteins, lipids, and nucleic acids capable of acting on multiple cell populations in parallel. Recent bone-regeneration studies further show that immune-microenvironment remodeling can be coupled with cytoskeletal and mitochondrial regulation, redox and senescence control, topological mechanotransduction, and programmed hydrogel cues to improve osteogenesis and angiogenesis [[47], [48], [49], [50], [51], [52], [53]].

3.6. Histological and immunofluorescence evidence confirms coordinated immunomodulation, vascularization, and osteogenesis

To further validate the micro-CT and gene-expression findings, tissue-level immunofluorescence and histological analyses were performed. CD86 staining remained relatively pronounced in the GelMA-only group, indicating persistence of an inflammatory macrophage-associated phenotype within the defect region. In contrast, vesicle-loaded hydrogels reduced CD86 signal and increased CD206 expression, with GelMA–M2-HP-Apo-sEVs producing the most evident macrophage phenotype shift (Fig. 6A and B). This result is consistent with the in vitro macrophage reprogramming assays and confirms that the vesicles retained immunomodulatory bioactivity after incorporation into GelMA and implantation in vivo.

Fig. 6.

Tissue-level validation of the immunoregenerative microenvironment induced by GelMA–M2-HP-Apo-sEVs. (A) Representative immunofluorescence staining images of CD86, CD206, BMP2, and CD31 in regenerated calvarial tissues after implantation of GelMA, GelMA–M2-sEVs, or GelMA–M2-HP-Apo-sEVs. (B) Quantitative analysis of fluorescence intensity corresponding to panel (A). (C) Representative H&E and Masson's trichrome staining images showing tissue bridging, collagen deposition, and new bone formation within the defect area. (D) Quantitative analysis of new bone area and collagen deposition area. Data are presented as mean ± SD (n = 3 independent animals per group); *p < 0.05, **p < 0.01, ***p < 0.001.

BMP2 immunofluorescence was substantially enhanced in the vesicle-treated groups, especially in the GelMA–M2-HP-Apo-sEV group, indicating activation of an osteoinductive tissue microenvironment (Fig. 6A and B). CD31 staining was also increased, supporting enhanced vascular formation in the regenerating defect. Since vascular invasion and osteogenic matrix deposition are mutually reinforcing during bone repair, the concurrent increase in CD31 and BMP2 provides tissue-level evidence that M2-HP-Apo-sEVs promoted coordinated angiogenesis and osteogenesis rather than isolated mineral deposition.

Histological staining further confirmed improved tissue regeneration. H&E staining showed limited defect bridging in the GelMA-only group, whereas GelMA–M2-sEVs promoted partial tissue reconstruction. In contrast, GelMA–M2-HP-Apo-sEVs generated more continuous tissue bridging across the defect area, and Masson's trichrome staining revealed more abundant collagen-rich matrix and better organized newly formed tissue in the GelMA–M2-HP-Apo-sEV group (Fig. 6C and D), consistent with advanced bone matrix maturation. To further assess the systemic biocompatibility of the implanted hydrogel formulations, major organs, including the heart, liver, spleen, lung, and kidney, were examined by H&E staining at 12 weeks after implantation. No obvious histopathological abnormalities were observed in the examined organ sections from the hydrogel-treated groups (Fig. S4), supporting the absence of overt organ-level toxicity at the evaluated endpoint.

Beyond exogenous delivery, bone matrix-derived EVs have been identified as endogenous messengers capable of transmitting mineralization-associated signals during skeletal aging, further supporting the concept that vesicular communication is a powerful regulatory layer in bone tissue remodeling [54]. Overall, these findings establish a coherent regenerative mechanism: hypoxic preconditioning confers apoptotic-mimetic and uptake-favorable features on M2 macrophage-derived sEVs; these vesicles reprogram inflammatory macrophages toward a pro-resolving phenotype, directly enhance BMSC osteogenesis, and stimulate endothelial angiogenic behavior; GelMA then localizes and sustains these biological cues within the defect site, enabling integrated immunomodulation, vascularization, and osteogenesis. This osteo–immuno–angiogenic coupling is increasingly recognized as a central design principle for advanced bone-regenerative biomaterials, particularly in large or critical-sized defects where single-lineage osteoinduction is insufficient to restore functional tissue architecture [7,18].

Several limitations should be acknowledged. First, the in vivo evidence was obtained using a single murine calvarial-defect model at a single 12-week endpoint; additional time points, orthotopic load-bearing models, disease-associated models, and large-animal studies are needed to establish the generalizability of the therapeutic effects. Second, the donor vesicles were generated exclusively from RAW264.7 cells, and validation using primary bone-marrow-derived macrophages and human monocyte-derived macrophages will be required to strengthen their translational relevance. Third, the AnaeroPack system provided qualitative low-oxygen preconditioning but did not continuously control or record the oxygen concentration. Fourth, although CD81, ALIX, CALNEXIN, TEM, particle sizing, and particle-to-protein analysis were used to characterize the vesicle preparations, differential ultracentrifugation cannot completely resolve apoptosis-associated EV subpopulations with overlapping sizes and biophysical properties. Fifth, the key protein, lipid, or RNA cargo responsible for the observed biological effects remains unidentified. The involvement of a PS-associated MerTK recognition pathway was supported by partial Mertk silencing; however, the incomplete knockdown efficiency and the absence of direct PS competition or complete receptor-ablation experiments prevent attribution of the enhanced uptake and immunomodulatory effects exclusively to this pathway. Sixth, although vesicle release was evaluated under collagenase type I-mediated degradation conditions, this simplified in vitro system does not fully reproduce the complex protease composition, fluid exchange, cellular remodeling, and degradation kinetics of the bone-defect microenvironment. Finally, representative H&E staining of the heart, liver, spleen, lung, and kidney revealed no obvious histopathological abnormalities at the 12-week endpoint; nevertheless, this preliminary organ-level evaluation does not constitute a comprehensive systemic toxicology assessment. Longitudinal in vivo GelMA degradation, biodegradation-product safety, sEV biodistribution and clearance, immunogenicity, hematological and serum biochemical parameters, quantitative organ pathology, and safety at extended follow-up periods remain to be investigated.

Clinical translation will also require standardized and scalable manufacturing. GelMA and the photoinitiator may undergo validated sterilization before vesicle loading, whereas the sEV-containing final product should be manufactured and filled aseptically because terminal sterilization may compromise vesicle membrane integrity and biological activity. Future production should therefore incorporate GMP-compatible donor-cell banks, closed or controlled processing systems, and lot-to-lot release criteria covering particle number, size distribution, identity, purity, potency, sterility, mycoplasma, endotoxin, and residual host-cell components, together with long-term safety and dose-ranging studies.

4. Conclusion

In summary, this study presents a hydrogel-enabled extracellular vesicle strategy for orchestrating immune resolution, angiogenesis, and osteogenesis during critical-sized bone defect repair. Hypoxic preconditioning endowed M2 macrophage-derived sEVs with an apoptosis-associated, phosphatidylserine-enriched membrane phenotype that was accompanied by enhanced vesicle internalization and a strengthened capacity to reprogram inflammatory macrophages toward a pro-resolving state. Beyond immunomodulation, M2-HP-Apo-sEVs directly promoted BMSC osteogenic differentiation and endothelial angiogenic activity, indicating a coordinated osteo–immuno–angiogenic mode of action. When incorporated into photocrosslinkable GelMA hydrogels, M2-HP-Apo-sEVs were locally retained and released within the defect region, resulting in superior calvarial bone regeneration, enhanced vascular formation, increased osteogenic marker expression, and reduced inflammatory activation in vivo. These findings demonstrate that hypoxia-programmed apoptotic-mimetic macrophage-derived sEVs can serve as instructive biological nanomodulators for reshaping the regenerative microenvironment. The proposed GelMA–M2-HP-Apo-sEV platform provides a cell-free and biomaterial-integrated approach for immunoregenerative bone repair and may offer a broadly applicable design principle for vesicle-based tissue regeneration.

CRediT authorship contribution statement

Hao Pan: Writing – review & editing, Writing – original draft. Jiefeng Huang: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision. Haoze Zhu: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources. Likai Chen: Formal analysis, Data curation, Conceptualization. Siman Huang: Project administration, Methodology, Investigation. Yueyue Huang: Software, Resources, Project administration, Methodology. Xiangyu Zhu: Visualization, Validation, Supervision, Software. Xiaokun Li: Writing – review & editing, Writing – original draft, Funding acquisition. Cailong Liu: Writing – review & editing, Writing – original draft, Investigation, Funding acquisition.

Declaration of competing interest

The authors declare that they have no competing interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.