A Self‐Oxygenating Wound‐Adaptive Stem Cell Encapsulation Platform for Chronic Wound Repair

ABSTRACT

Stem cell‐based therapies hold substantial promise for the treatment of chronic wounds. However, limited clinical efficacy has frequently arisen not from inadequate cell potency, but from failure of current delivery paradigms to sustain cell survival and function within hostile host microenvironments. Here, we report a wound‐adaptive, self‐oxygenating stem cell encapsulation platform designed to overcome this delivery paradigm failure. The device embeds living stem cells within an immunoisolating hydrogel matrix conformally shaped to individual wound geometries and converts locally accumulated CO₂ into molecular O₂, thereby providing an oxygen supply that preserves cellular metabolic activity in situ. Importantly, oxygenation in this system functions primarily to sustain cell viability rather than to directly oxygenate host tissue. Compared with conventional local or systemic delivery strategies, encapsulated cells exhibit markedly enhanced survival, persistence, and therapeutic function. Evaluation in diabetic rat and porcine wound models demonstrates accelerated resolution of inflammation, improved vascular maturation, and robust tissue regeneration. By reframing cell therapy as a problem of environment‐cell compatibility, this platform establishes a clinically translatable delivery paradigm for precision stem cell therapy in chronic wound repair.

A CO₂‐responsive BOOST device enables sustained in situ oxygenation of encapsulated MSCs for chronic wound repair. By maintaining MSC viability and therapeutic activity within an immunoisolating alginate hydrogel, the platform supports persistent cell function, enhances tissue regeneration, and accelerates diabetic wound healing.

1. Introduction

Chronic nonhealing wounds represent a major and growing clinical challenge, characterized by prolonged inflammation, impaired angiogenesis, and a failure to progress through the normal stages of tissue repair. It is estimated that 1%–2% of individuals in developed countries will experience a chronic wound during their lifetime, imposing a substantial and increasing burden on healthcare systems worldwide [1, 2]. Among these conditions, diabetic ulcers are particularly severe, as persistent non‐healing markedly increases the risk of infection and lower‐limb amputation [3]. At the mechanistic level, chronic wounds are marked by a hostile and dysregulated microenvironment, including sustained inflammation, pathological hypoxia, excessive protease activity with extracellular matrix (ECM) degradation, and aberrant immune cell responses, which collectively impair re‐epithelialization, angiogenesis, and tissue remodeling [4, 5, 6, 7].

Considerable progress has been made in the development of advanced wound dressings, gas‐based therapies, and bioactive materials for chronic wound management [2, 8, 9, 10, 11, 12, 13, 14]. However, most material‐based approaches provide limited or unidimensional functions and fail to adapt to the dynamic and multifactorial nature of the wound microenvironment. Similarly, growth factor‐ or exosome‐based therapies, despite their intrinsic bioactivity, are constrained by rapid degradation and short in vivo residence times, necessitating repeated administration and limiting sustained therapeutic efficacy [10, 15, 16].

Mesenchymal stem cells (MSCs) have been shown to promote wound healing by secretion of cytokines, growth factors, and extracellular vesicles that modulate inflammation, stimulate angiogenesis, and support tissue regeneration [17, 18, 19, 20, 21]. Encouraged by these findings, multiple MSC‐based products—including adipose‐, bone marrow‐ and umbilical cord‐derived MSCs as well as MSC‐derived exosomes—have entered early‐stage clinical trials for chronic wounds and diabetic ulcers. Despite this promise, the clinical translation of MSC therapy has been hindered by fundamental challenges related to delivery, persistence, and microenvironmental compatibility. Systemic administration of MSCs is severely constrained by instant blood‐mediated inflammatory reaction (IBMIR) and the pulmonary first‐pass effect following intravenous infusion, resulting in only ∼1%–5% of infused cells reaching peripheral wound sites [22] (Figure 1a). Local delivery strategies, although bypassing these systemic barriers, remain limited by poor cell retention [23], immune‐mediated clearance [24], and the lack of precise control over therapeutic cell dose (Figure 1b). More fundamentally, the severely hypoxic microenvironment of chronic wounds—where O₂ tension can fall below ∼7.6 mmHg—induces mitochondrial dysfunction and activates apoptotic pathways in transplanted MSCs, rapidly eroding their viability, metabolic activity, and paracrine function [25]. Collectively, these challenges indicate that the primary obstacle to effective MSC therapy lies not in the intrinsic therapeutic potential of the cells, but in the incompatibility between living cell therapies and the hostile wound microenvironment.

FIGURE 1.

Conceptual overview of limitations in conventional MSC therapy and the therapeutic strategy of the BOOST device. (a) Limitations of intravenous MSCs infusion. (b) Limitations of local administration of MSCs. (c) Schematic illustration of therapeutic strategy proposed in this study using the MSC‐encapsulating BOOST device.

Here, we report a self‐oxygenating, wound‐adaptive stem cell encapsulation platform designed to actively remodel the wound microenvironment while sustaining cellular viability and function in situ (Figure 1c). The platform regulates local O₂ availability by converting CO₂ produced within the wound site into O₂, thereby providing a demand‐adaptive O₂ supply without external intervention. This bioresponsive oxygenation strategy stabilizes MSC metabolic activity and preserves regenerative paracrine signaling under otherwise prohibitive hypoxic conditions. MSCs are embedded within a wound‐adaptive alginate‐based hydrogel matrix that minimizes cell loss and provides immune isolation [26], enabling the use of allogeneic or xenogeneic cells while attenuating immune‐mediated clearance. To enable precision therapy for irregular chronic wounds, three‐dimensional scanning is employed to reconstruct wound geometry, allowing patient‐specific control over cell dosing. Using diabetic rat and porcine wound models, we demonstrate that this microenvironment‐centered strategy enhances MSC persistence, accelerates the transition from inflammation to proliferation, and promotes robust tissue regeneration. We term this system the BOOST (Bioresponsive Oxygen‐Output Sustaining Therapeutic) platform. Together, this work establishes a new framework for precision‐engineered stem cell therapy that mitigates key limitations of conventional cell delivery by actively aligning cell function with the wound microenvironment in chronic wound repair.

2. Results

2.1. Design and Fabrication of BOOST device

The clinical translation of mesenchymal stem cell (MSC)–based therapies has recently gained momentum, exemplified by the U.S. Food and Drug Administration approval of Ryoncil in 2024 as the first MSC‐based therapeutic product. Among various MSC sources, adipose‐derived mesenchymal stem cells (ADMSCs) are particularly attractive owing to their abundance, ease of isolation, and cost efficiency [27]. However, their therapeutic efficacy in chronic wounds remains limited by poor survival and functional impairment under severe hypoxia [9]. ADMSCs exhibit optimal metabolic activity and paracrine function at approximately 5% O₂ (∼38 mmHg tissue pO₂) [28], whereas O₂ tension within chronic wound cores often declines to ∼1% (∼7.6 mmHg), creating a hostile microenvironment that rapidly compromises cell viability and regenerative capacity [25, 29].

To address this mismatch between cellular O₂ demand and the wound microenvironment, we designed a self‐oxygenating encapsulation device driven by a CO₂‐responsive chemical reaction to provide sustained, localized O₂ supply to encapsulated MSCs (Figure 2a). In this system, CO₂ produced endogenously by both wound tissue and encapsulated cells serves as the trigger for O₂ generation, enabling harvesting of locally accumulated CO₂ and its conversion into molecular O₂. Notably, while tissue O₂ tension decreases markedly in ischemic wounds, CO₂ partial pressure remains relatively stable at ∼40 mmHg, and CO₂ exhibits substantially higher solubility in biological tissues than O₂ under equivalent partial pressures. This physiological characteristic provides a continuous and readily accessible substrate for in situ O₂ generation, supporting sustained MSC metabolic activity (Figure S1).

FIGURE 2.

Design, fabrication, and interfacial engineering of the BOOST device. (a) Schematic illustration of the functional modules of the BOOST device, including an O₂‐impermeable ethylene–vinyl alcohol (EVOH) barrier, a Li₂O₂‐based O₂‐generating reservoir, perfluorocarbon (PFC) for enhanced O₂ solubility, an O₂‐permeable PDMS membrane, an immunoisolating alginate hydrogel encapsulating MSCs, and an antimicrobial bioadhesive sealing interface. (b) Exploded‐view schematic illustrating the multilayer architecture of the BOOST device. (c) Schematic representation and representative photograph of the assembled BOOST device. (d) Schematic illustration of the surface modification of PDMS via O₂ plasma activation, APTES‐mediated amination, and subsequent integration of the bioadhesive layer. (e) XPS elemental analysis confirming successful surface amination of PDMS after modification. (f) Static water contact angle measurements demonstrating enhanced surface hydrophilicity of PDMS following APTES modification. (g) Schematic depiction of the molecular interactions underlying interfacial adhesion between the bioadhesive hydrogel and skin tissue. (h) Illustration of the 180° peel test setup used to evaluate interfacial adhesion between the adhesive layer and PDMS membrane or pig skin tissue. (i–j) Quantitative measurements of interfacial adhesion strength between the adhesive hydrogel and PDMS membrane (i) or pig skin tissue (j).

The BOOST device was engineered as a multilayer architecture optimized for O₂ generation, retention, and delivery (Figure 2b). A gas‐permeable polydimethylsiloxane (PDMS) shell facilitates O₂ diffusion toward the cell‐encapsulating compartment, while a backside ethylene–vinyl alcohol (EVOH) membrane serves as an O₂ barrier to minimize O₂ leakage and prevent external contamination. A PDMS structural frame extends into the hydrogel core, reinforcing mechanical stability and creating a hollow gas phase that promotes efficient O₂ transport. At the core of the device, a Li₂O₂/perfluorocarbon (PFC) composite (20% w/w Li₂O₂) functions as the O₂‐generating module: Li₂O₂ undergoes CO₂‐triggered chemical conversion to release O₂, whereas the PFC phase enhances O₂ solubility and prolongs O₂ availability within the system (Figure S2) [30].

ADMSCs were encapsulated within an alginate hydrogel matrix, which serves a dual function by retaining cells at the wound site and providing an immunoisolating microenvironment [26, 31]. This encapsulation strategy mitigates immune recognition and supports the potential use of allogeneic or xenogeneic MSCs, while preserving the cells’ capacity to secrete pro‐regenerative paracrine factors that modulate inflammation, stimulate angiogenesis, and promote tissue regeneration. The fully assembled device can be readily integrated into a polyurethane dressing for direct application to wound surfaces (Figures 2c and S3).

Stable interfacial adhesion between the BOOST device and skin tissue is essential to maintain O₂ delivery efficiency and prevent device displacement during prolonged treatment. To this end, we engineered a bioadhesive sealing interface based on a hydrogel (named as BSI) incorporating acrylic N‐hydroxysuccinimide (NHS) ester–functionalized monomers containing disulfide bonds. These disulfide linkages undergo glutathione (GSH)‐triggered cleavage, enabling redox‐responsive, on‐demand detachment of the BSI and thereby minimizing tissue damage during removal [32] (Figures S4–S6). UV‐initiated copolymerization with acrylic acid (AA) generated a covalently crosslinked network, while poly(vinyl alcohol) (PVA) and hydroxypropyltrimethyl ammonium chloride chitosan (HACC) were physically incorporated to enhance mechanical robustness and antimicrobial activity. AA units enable extensive hydrogen bonding, whereas NHS esters form covalent amide bonds with tissue proteins, together conferring strong and durable tissue adhesion. Importantly, histological analysis of skin tissues after 15 days of continuous BSI application revealed no evident tissue damage, inflammatory infiltration, or structural abnormalities, demonstrating the good biocompatibility of the adhesive interface (Figure S7a,b). Because pristine PDMS is intrinsically hydrophobic and poorly adhesive to biological tissues [33, 34], we further modified the PDMS surface to enhance interfacial compatibility. Oxygen plasma treatment generated surface hydroxyl groups, followed by silanization with (3‐aminopropyl)triethoxysilane (APTES) to introduce surface amine functionalities (Figures 2d and S8). X‐Ray photoelectron spectroscopy (XPS) confirmed successful surface amination (Figure 2e), and static water contact angle measurements demonstrated a pronounced increase in surface hydrophilicity, with contact angles decreasing from 91° to 42° after modification (Figure 2f). The resulting interface supports synergistic hydrogen bonding and covalent coupling between the PDMS surface and the adhesive hydrogel network (Figure 2g).

In addition to promoting adhesion, the incorporation of HACC imparted antibacterial activity, providing an effective barrier against microbial invasion during wound treatment (Figure S9a,b). The chemical structure of the adhesive hydrogel was confirmed by ATR–FTIR spectroscopy (Figure S10), and its mechanical stability was validated under conditions relevant to wound dressing applications (Figure S11). Interfacial adhesion strength was quantitatively evaluated using a 180° peel test (Figure 2h). The bioadhesive sealing layer exhibited high interfacial toughness against porcine skin (∼80 N m^{− 1}) and PDMS membrane (∼45 N m^{− 1}) (Figures 2i,j and S12), indicating robust and durable adhesion enabled by combined physical and chemical interactions.

2.2. BOOST Device–Enabled In Situ Oxygenation Supports ADMSC Viability Under Hypoxic Conditions

Sustained and regulated O₂ delivery is essential for maintaining MSC viability and function in the hypoxic microenvironment of chronic wounds. To quantitatively characterize the O₂‐generating performance of the BOOST device, we established a closed gas‐circuit system with switchable gas inputs and real‐time O₂ monitoring using a gas mass spectrometer (Figure 3a). Prior to measurement, the system was purged with pure N₂ to eliminate residual O₂ and establish a stable baseline. Defined O₂/N₂ mixtures were subsequently introduced to generate calibration standards, enabling construction of a quantitative relationship between O₂ concentration and ion signal intensity (Figure S13).

FIGURE 3.

O₂‐generation performance of the BOOST device and its regulatory effects on MSC viability and oxidative status. (a) Schematic illustration of the experimental setup used to quantify O₂ generation by the BOOST device using gas mass spectrometry under controlled CO₂ conditions. (b) O₂ generation rates of the BOOST device measured over different release durations (mean ± s.d., n = 10). Statistical significance was assessed by one‐way analysis of variance (ANOVA) with Tukey's multiple comparisons test. (c, d) Viability of MSCs cultured under different O₂ concentrations in vitro, including adherent MSCs (c) and MSCs encapsulated within hydrogels or devices (d) (mean ± s.d., n = 4). Statistical significance was assessed by one‐way ANOVA with Tukey's multiple comparisons test. (e) Intracellular reactive oxygen species (ROS) levels in MSCs under indicated O₂ conditions. (f) Simulated spatial distributions of O₂ concentration within the BOOST device and the control device. Statistical significance was determined using an unpaired two‐tailed Student's t‐test. (g) Schematic illustration of in situ assessment of encapsulated MSC viability following implantation in vivo. (h) Representative fluorescence images of in situ live/dead staining of encapsulated MSCs. Live cells are shown in green and dead cells in red. (i) Quantitative analysis of in vivo MSC survival in diabetic rat wounds (mean ± s.d., n = 4). Statistical significance was assessed by one‐way ANOVA with Tukey's multiple comparisons test.

The O₂‐generation kinetics of the BOOST device were evaluated under a 5% CO₂ atmosphere, approximating the CO₂‐rich conditions of chronic wound tissue. To assess temporal stability, devices were preincubated in a 5% CO₂ incubator for 0, 24, or 72 h before measurement. Across all time points, the device exhibited stable and sustained O₂ release, with generation rates of 2.02, 1.94, and 1.87 mL day^{− 1}, respectively (Figure 3b). Notably, these values substantially exceed the estimated O₂ consumption of encapsulated MSCs (∼0.05 mL day^{− 1}) [35], indicating that the BOOST device provides sufficient O₂ to meet cellular metabolic demand throughout the treatment period.

We next examined the impact of O₂ availability on ADMSC viability under both adherent and encapsulated conditions. Consistent with previous reports, an O₂ concentration of approximately 5% supported optimal ADMSC viability and metabolic activity, whereas severe hypoxia (1% O₂) markedly impaired cell survival (Figure 3c,d). Importantly, ADMSCs encapsulated within the BOOST device maintained viability and metabolic activity under 1% O₂ at levels comparable to those observed under physiological hypoxia (5% O₂), indicating effective mitigation of hypoxia‐induced cellular stress. In parallel, intracellular reactive oxygen species (ROS) levels were significantly reduced in ADMSCs cultured within the BOOST device under 1% O₂, suggesting preservation of redox homeostasis and attenuation of hypoxia‐associated oxidative damage (Figure 3e).

To further elucidate the spatial O₂ distribution generated by the device, we performed COMSOL‐based simulations of O₂ diffusion [36, 37]. The control device exhibited uniformly low O₂ levels, consistent with minimal O₂ supplementation. In contrast, the BOOST device generated a pronounced O₂ gradient, with elevated pO₂ surrounding the O₂‐generating module and efficient diffusion toward the wound‐contacting interface (Figure 3f). Quantitative analysis confirmed significantly higher O₂ partial pressures throughout the BOOST device compared with the control configuration, supporting its capacity for localized and sustained O₂ delivery. Clark oxygen electrode measurements further confirmed that the O₂ partial pressures at different positions within the device were all higher than those in the wound tissue, demonstrating effective oxygen delivery. (Figure S14).

Finally, we evaluated the ability of the BOOST device to support MSC survival in vivo using a diabetic rat wound model (Figure 3g). After 3 days of treatment, MSC‐encapsulating hydrogels were retrieved from wound sites and analyzed by live/dead fluorescence staining. Cells cultured under 5% O₂ served as a positive control and exhibited high viability, whereas MSCs recovered from the control device showed extensive cell death, reflecting the harsh hypoxic wound environment (Figures 3h and S15). In contrast, MSCs encapsulated within the BOOST device retained markedly higher viability. Quantitative CCK‐8 analysis following alginate lyase‐mediated cell release further confirmed significantly enhanced MSC survival in the BOOST group (Figure 3i).

2.3. MSCs‐Encapsulating BOOST Device Accelerates Chronic Wound Healing in Diabetic Rats

Having established that the BOOST device sustains MSC viability under hypoxic conditions in vivo, we next examined whether this enhanced cellular persistence translates into functional therapeutic benefit in a diabetic wound model. The therapeutic efficacy of the MSCs‐encapsulating BOOST device was evaluated using a streptozotocin (STZ)‐induced diabetic male Sprague–Dawley rat full‐thickness wound model (Figure 4a). Throughout the treatment period, all diabetic animals exhibited sustained hyperglycaemia and progressive body‐weight loss (Figures S16 and S17), confirming the stability of the diabetic phenotype during wound healing assessment. Circular full‐thickness excisional wounds (15 mm in diameter) were created on the dorsal skin, and animals were randomly assigned to five treatment groups: untreated diabetic control, acellular control device (alginate hydrogel only; device core comprises Li₂CO₃/PFC), MSCs‐encapsulating control device, acellular BOOST device, and MSCs‐encapsulating BOOST device. An additional nondiabetic, untreated wound group served as a physiological reference (normal control). Treatments were maintained for 15 days, with dressing replacement on days 3, 7, and 11.

FIGURE 4.

Therapeutic efficacy of the MSCs‐encapsulating BOOST device in a diabetic rat full‐thickness wound model. (a) Schematic illustration of the experimental timeline and treatment procedure in the streptozotocin‐induced diabetic rat wound model. (b) Representative photographs showing wound closure progression in different treatment groups at days 0, 3, 7, 11, and 15 post‐injury. Statistical significance was assessed using two‐way ANOVA, followed by Tukey's multiple comparisons test. (c) Quantitative analysis of wound closure over time, expressed as the percentage of remaining wound area relative to the initial wound size (mean ± SD, n = 6). (d) Representative histological and immunohistochemical analyses of wound tissues. H&E staining of wound sections collected on days 7 and 15 post‐injury, with red arrows indicating newly formed hair follicles. CD31 immunohistochemical staining of wound tissues on day 7 showing neovascularization. Masson's trichrome staining of wound tissues on day 15 illustrates collagen deposition and extracellular matrix remodeling. (e–h) Quantitative histological analyses of (e) blood vessel density on day 7, (f) epidermal thickness on day 15, (g) granulation tissue thickness on day 15, and (h) collagen volume fraction on day 15 (mean ± SD, n = 6). Statistical significance was determined by one‐way ANOVA with Tukey's multiple comparisons test.

Macroscopic wound evaluation revealed distinct differences in healing dynamics among treatment groups (Figure 4b,c). Neither the acellular control device nor the acellular BOOST device significantly improved wound closure compared with diabetic controls, whereas the MSCs‐encapsulating control device produced only a modest enhancement. In contrast, wounds treated with the MSCs‐encapsulating BOOST device exhibited markedly accelerated closure and consistently outperformed all other diabetic treatment groups. This effect was particularly evident during the early healing phase. By day 3 post‐injury, wound areas in the diabetic control, acellular control device, MSCs‐encapsulating control device, and acellular BOOST groups remained above 90% of their initial size, whereas wounds treated with the MSCs‐encapsulating BOOST device had contracted to 57.8% of baseline. These results indicate that early‐stage O₂ supplementation, when coupled with viable MSC delivery, enables rapid cellular engagement and functional activity immediately following wound formation.

Histological analyses further corroborated the superior regenerative outcome achieved by the MSCs‐encapsulating BOOST device (Figure 4d). On day 7, haematoxylin and eosin (H&E) staining revealed a large epithelial gap in diabetic control wounds (10.52 mm), whereas non‐diabetic wounds exhibited substantially reduced gaps (5.45 mm). Among diabetic treatment groups, the acellular control device showed minimal improvement (9.55 mm), while the MSCs‐encapsulating control device moderately enhanced re‐epithelialization (7.66 mm). The acellular BOOST device also partially reduced the epithelial gap (8.32 mm), suggesting a limited benefit from O₂ delivery alone. Notably, the MSCs‐encapsulating BOOST device achieved the greatest improvement, reducing the epithelial gap to 5.60 mm—comparable to non‐diabetic healing. This group displayed continuous epithelial coverage and well‐organized granulation tissue, indicative of synergistic effects between sustained oxygenation and MSC‐mediated repair during the inflammatory‐to‐proliferative transition. Consistently, wounds treated with the MSCs‐encapsulating BOOST device exhibited reduced expression of the pro‐inflammatory markers iNOS and IL‐1β on day 7, confirming an enhanced anti‐inflammatory response [38] (Figures S18 and S19).

As healing progressed, regeneration of skin appendages became apparent. By day 15, both MSCs‐encapsulating control device and MSCs‐encapsulating BOOST device groups showed markedly increased formation of hair follicles compared with other diabetic treatments, highlighting the intrinsic regenerative capacity of MSCs in skin reconstruction. At this later stage, as wound hypoxia was alleviated and tissue remodeling dominated, differences between the two MSC‐treated groups became less pronounced, indicating that the principal contribution of the BOOST device occurs during the early hypoxic phase of wound repair. Quantitative immunohistochemical analysis of CD31 staining on day 7 demonstrated significantly enhanced neovascularization in wounds treated with the MSCs‐encapsulating BOOST device compared with diabetic controls and all other treatment groups (Figure 4e). Measurements of epidermal thickness further confirmed superior re‐epithelialization in the MSCs‐encapsulating BOOST group (Figure 4f). In addition, Masson's trichrome staining on day 15 revealed substantially increased collagen deposition and more organized extracellular matrix architecture in the MSCs‐encapsulating BOOST device group relative to diabetic controls and other device‐based treatments (Figures 4g,h, and S20), indicating enhanced matrix remodeling and structural maturation of regenerated tissue. To benchmark the BOOST platform against a clinically used wound dressing, we compared its efficacy with a commercial hydrogel in a diabetic rat wound model. The MSCs‐encapsulating BOOST device resulted in smaller wound areas on days 7 and 11 and showed greater epidermal thickness and collagen deposition on day 15, indicating superior regenerative efficacy compared with the commercial hydrogel dressing (Figure S21).

2.4. Proteomic Profiling Reveals Metabolic Reprogramming and Regenerative Signaling Induced by the MSCs‐Encapsulating BOOST Device

To elucidate the molecular mechanisms underlying the accelerated wound repair observed with the MSCs‐encapsulating BOOST device, we performed quantitative proteomic profiling of wound tissues harvested at day 15 post‐treatment. Proteins were extracted from wound tissue samples and analyzed using data‐independent acquisition (DIA) mass spectrometry to enable unbiased and reproducible quantification across experimental groups.

Principal component analysis (PCA) revealed clear segregation of proteomic profiles among treatment groups (Figure 5a). Samples from the MSCs‐encapsulating BOOST device group formed a distinct cluster, separate from both the diabetic control and the MSCs‐encapsulating control device groups, indicating substantial molecular reprogramming induced by BOOST‐mediated therapy. Biological replicates within each group clustered tightly, confirming high analytical reproducibility. Differential expression analysis identified 54 upregulated and 88 downregulated proteins in the MSCs‐encapsulating BOOST device group relative to diabetic controls, and 45 upregulated and 94 downregulated proteins relative to the MSCs‐encapsulating control device group (Figure 5b,c). Hierarchical clustering of differentially expressed proteins further demonstrated a pronounced divergence in proteomic landscapes between diabetic control and MSCs‐encapsulating BOOST device–treated wounds (Figure 5d). Whereas diabetic control wounds were characterized by enrichment of proteins associated with persistent inflammation and impaired remodeling, BOOST‐treated wounds exhibited increased abundance of proteins linked to extracellular matrix organization, angiogenesis, and metabolic activity, collectively indicating a shift toward a reparative molecular state.

FIGURE 5.

Proteomic profiling reveals molecular programs underlying the therapeutic efficacy of the MSCs‐encapsulating BOOST device. (a) Principal component analysis (PCA) of wound‐tissue proteomes from diabetic control, control device + MSCs, and BOOST device + MSCs groups, showing distinct clustering of BOOST‐treated wounds. (b) Volcano plot of differentially expressed proteins (DEPs) comparing BOOST device + MSCs versus diabetic control wounds. (c) Volcano plot of DEPs comparing BOOST device + MSCs versus control device + MSCs, highlighting BOOST‐specific molecular effects beyond cell delivery alone. (d) Heatmap of representative DEPs between diabetic control and BOOST device + MSCs groups. Each row represents a protein and each column represents an individual biological replicate; colors indicate z‐score–normalized relative expression levels (red, upregulation; blue, downregulation). (e) Gene Ontology biological process (GO BP) enrichment analysis of proteins upregulated in BOOST device + MSCs compared with diabetic control wounds. (f) GO BP enrichment analysis of proteins upregulated in BOOST device + MSCs compared with control device + MSCs, highlighting BOOST‐specific functional programs. (g) GO BP enrichment analysis of proteins downregulated in BOOST device + MSCs compared with diabetic control wounds. (h) KEGG pathway enrichment analysis of upregulated proteins in BOOST device + MSCs versus diabetic control wounds. (i) KEGG pathway enrichment analysis of upregulated proteins in BOOST device + MSCs versus control device + MSCs. Enrichment analyses were performed using raw p values, with significance thresholds indicated in each panel.

Functional enrichment analyses revealed that metabolic reprogramming represents a central feature of BOOST device–mediated wound repair. Gene Ontology (GO) biological process enrichment of proteins upregulated in the MSCs‐encapsulating BOOST device group versus diabetic controls highlighted coordinated activation of lipid and energy metabolic pathways, including monocarboxylic acid metabolism, fatty acid biosynthesis, acylglycerol metabolism, and arachidonic acid metabolism (Figure 5e). These pathways are closely associated with membrane synthesis, inflammatory resolution, and energy supply during tissue regeneration. A similar enrichment pattern was observed when comparing the MSCs‐encapsulating BOOST device group with the MSCs‐encapsulating control device group (Figure 5f), indicating that sustained oxygenation further amplifies MSC‐driven metabolic remodeling beyond cell delivery alone. In parallel, GO enrichment analysis of downregulated proteins revealed suppression of pathways related to programmed cell death and cell death (Figure 5g), suggesting improved cellular survival and stabilization of the wound microenvironment.

KEGG pathway enrichment analysis further supported the emergence of a pro‐regenerative signaling landscape in BOOST‐treated wounds. Compared with diabetic controls, wounds treated with the MSCs‐encapsulating BOOST device exhibited significant enrichment of angiogenesis‐ and tissue repair–associated signaling pathways, including VEGF, Ras, ErbB, PI3K–Akt, and mTOR signaling, alongside pathways involved in lipid and sphingolipid metabolism [39, 40] (Figure 5h). Enrichment of vesicular transport–related pathways, such as SNARE‐mediated trafficking, suggests enhanced intracellular communication and coordinated secretion of regenerative factors. Consistently, comparison between the MSCs‐encapsulating BOOST device and MSCs‐encapsulating control device groups revealed enrichment of pathways governing cell proliferation, survival, and stress adaptation, including Ras, MAPK, PI3K–Akt, and p53 signaling [41] (Figure 5i). Concurrent upregulation of peroxisomal function and redox‐related metabolic pathways, such as ascorbate and aldarate metabolism as well as nicotinate and nicotinamide metabolism, further indicates enhanced metabolic homeostasis and oxidative stress resilience in BOOST‐treated wounds.

Together, these proteomic data demonstrate that the MSCs‐encapsulating BOOST device orchestrates a coordinated molecular program characterized by suppression of chronic inflammatory cues, reinforcement of metabolic and redox capacity, and activation of angiogenic and regenerative signaling pathways, thereby creating a tissue microenvironment conducive to sustained wound repair. While proteomic profiling in the rat model revealed global metabolic and regenerative reprogramming, we next sought to determine whether similar cellular remodeling occurs in a clinically relevant large‐animal context.

2.5. MSCs‐Encapsulating BOOST Device Accelerates Chronic Wound Healing in a Diabetic Bama Miniature Pig Model

Because porcine skin architecture and wound‐healing kinetics closely resemble those of humans [42, 43], we next evaluated the therapeutic efficacy of the MSCs‐encapsulating BOOST device in a clinically relevant large‐animal model using one‐year‐old male Bama miniature pigs. Diabetes was induced by intravenous streptozotocin (STZ) administration, and blood glucose levels were continuously monitored using a continuous glucose monitoring (CGM) system. Only animals exhibiting sustained hyperglycaemia were included for subsequent experiments (Figure S22).

To assess the feasibility of allogeneic MSC therapy in a large‐animal setting, adipose‐derived mesenchymal stem cells (ADMSCs) isolated from Duroc pigs were immortalized and used for wound treatment in diabetic Bama miniature pigs (Figure 6a). Full‐thickness dermal wounds measuring 2 × 2 cm with an approximate depth of 5 mm were created on the dorsal skin, followed by application of the designated dressings (Figure S23). To enable precision control of cell dosing, three‐dimensional (3D) scanning was performed immediately after wound creation, and wound‐bed volumes were reconstructed to calculate the required number of encapsulated MSCs for each wound (Figure 6b). Wounds corresponding to different treatment groups were spatially interleaved across the dorsal region of each animal and separated by at least 4 cm to minimize site‐dependent bias (Figure 6c,d).

FIGURE 6.

Therapeutic evaluation of the MSCs‐encapsulating BOOST device in a Bama miniature pig full‐thickness dermal wound model. (a) Schematic overview of the experimental timeline and therapeutic workflow in the diabetic Bama miniature pig wound model. (b) Workflow for personalized determination of MSC dosage based on three‐dimensional (3D) wound scanning and volumetric reconstruction. (c) Schematic depiction of the spatial distribution of different treatment groups on the dorsal region of Bama miniature pigs. (d) Representative photographs illustrating key surgical steps, including wound creation, device placement, and wound coverage. (e) Representative gross images showing wound healing progression over time in different treatment groups. (f) Quantification of wound closure over time, expressed as the percentage of remaining wound area relative to the initial wound size (mean ± SD, n = 4 wounds per group). Statistical significance was assessed using two‐way ANOVA, followed by Tukey's multiple comparisons test. (g) Representative immunofluorescence images of wound sections stained for CD31 to visualize vascular endothelium. Nuclei were counterstained with DAPI (blue), and CD31‐positive endothelial structures are shown in green. (h) Quantification of vascular density based on CD31‐positive area (mean ± SD, n = 4). Statistical significance was assessed using one‐way ANOVA with Tukey's multiple comparisons test. (i) Representative histological analysis of porcine wound tissues harvested on day 20 post‐treatment, including H&E (top row) and Masson's trichrome staining (bottom row). Enlarged views highlight collagen organization within the wound bed. (j) Quantification of collagen deposition in wound tissues based on Masson's trichrome staining (mean ± SD, n = 5 histological sections per group). Statistical significance was assessed using one‐way ANOVA with Tukey's multiple comparisons test.

Macroscopic assessment revealed a clear treatment‐dependent improvement in wound‐healing dynamics (Figure 6e). Quantitative analysis of wound area over time demonstrated that, whereas the gel‐only and MSCs‐encapsulating control device groups exhibited gradual wound contraction, wounds treated with the MSCs‐encapsulating BOOST device showed a more pronounced and sustained reduction in wound area (Figure 6f). Statistically significant differences in wound closure became apparent during the mid‐to‐late stages of healing, highlighting the benefit of sustained oxygenation in supporting prolonged MSC activity in the porcine wound microenvironment.

Consistent with accelerated wound closure, immunofluorescence staining for the endothelial marker CD31 revealed denser and more continuous microvascular networks in wounds treated with the MSCs‐encapsulating BOOST device (Figure 6g). Quantitative analysis confirmed a significantly higher vascular density in this group compared with all other treatments (Figure 6h), indicating enhanced angiogenesis. Histological evaluation further supported these findings. H&E staining revealed more advanced tissue regeneration and improved granulation tissue architecture in BOOST‐treated wounds, while Masson's trichrome staining demonstrated more extensive and organized collagen deposition (Figure 6i). Quantification of collagen volume fraction confirmed significantly enhanced extracellular matrix accumulation in the MSCs‐encapsulating BOOST device group (Figure 6j), indicative of more robust matrix remodeling. In addition, immunofluorescence staining for CD3 and CD68 around the explanted devices showed minimal T‐cell and macrophage infiltration, suggesting that the MSCs‐encapsulating BOOST device did not induce obvious host immune rejection (Figure S24).

Collectively, these results demonstrate that the MSCs‐encapsulating BOOST device effectively accelerates chronic wound healing in a clinically relevant large‐animal model. Beyond promoting wound closure, the device supports functional tissue regeneration by coupling enhanced angiogenesis with improved extracellular matrix reconstruction, underscoring its translational potential for precision‐engineered cell therapy in chronic wound management.

2.6. Single‐Cell Transcriptomic Landscape Reveals Vascular Maturation and Immune Remodeling in BOOST‐Treated Porcine Wounds

To gain mechanistic insight into the cellular programs underlying enhanced wound repair in the large‐animal model, we performed single‐cell RNA sequencing (scRNA‐seq) of diabetic Bama miniature pig wound tissues harvested on day 20 post‐treatment. Unsupervised clustering and UMAP (uniform manifold approximation and projection) visualization resolved the wound microenvironment into major epidermal, stromal, vascular, and immune cell populations, including adipocytes, Schwann cells, plasma cells, T cells, macrophages (MPs), smooth muscle cells (SMCs), pericytes, endothelial cells (ECs), fibroblasts, and keratinocytes (Figure 7a,b). Cell identities were assigned based on canonical marker gene expression, yielding a comprehensive cellular atlas suitable for treatment‐associated comparisons.

FIGURE 7.

Single‐cell transcriptomic analysis reveals cell‐type–specific molecular remodeling in diabetic Bama miniature pig wounds following BOOST device–mediated MSC therapy. (a) Dot plot showing the expression of canonical marker genes used for cell‐type annotation of major cell populations, including T cells, macrophages (MPs), plasma cells, keratinocytes, fibroblasts, endothelial cells (ECs), pericytes, smooth muscle cells (SMCs), Schwann cells, and adipocytes. (b) Uniform manifold approximation and projection (UMAP) visualization of all profiled cells colored by annotated cell type, together with stacked bar plots showing the relative proportions of major cell populations across different treatment groups. (c) Stacked bar plots depicting the relative proportions of functionally distinct SMC subclusters among different treatment groups. (d, e) Relative mRNA expression levels of selected genes in wound tissues across groups (mean ± SD, n = 3). Statistical significance was assessed by one‐way ANOVA with Tukey's multiple comparisons test. (f) Volcano plot illustrating treatment‐induced gene expression changes in wound tissues following BOOST device + MSCs therapy compared with diabetic control. (g) KEGG pathway enrichment analysis of downregulated genes in mGDTh17 cells comparing BOOST device + MSCs with diabetic control. (h) KEGG pathway enrichment analysis of downregulated genes in macrophages comparing BOOST device + MSCs with diabetic control. Statistical significance for gene expression comparisons was assessed using one‐way ANOVA. All p values shown in the enrichment analyses are adjusted p values (P‐adjust) after multiple‐testing correction.

Among these populations, wounds treated with the MSCs‐encapsulating BOOST device exhibited a notable increase in the relative abundance of SMCs compared with diabetic controls (Figure 7c), suggesting enhanced vascular maturation and tissue remodeling. This observation is consistent with the elevated expression of pro‐angiogenic and vessel‐stabilizing factors, including VEGF‐A and PDGF–BB [44], detected in BOOST‐treated wounds (Figure S25). To further interrogate vascular remodeling at higher resolution, SMCs were subclustered based on transcriptional heterogeneity, revealing distinct functional states characterized by marker genes such as MGP [45], HOXC9 [46], and FN1 [47] (Figure S26). Compositional analysis demonstrated a pronounced shift in SMC subtype distribution following BOOST‐mediated therapy, with a relative enrichment of HOXC9‐expressing SMCs and a concomitant reduction in MGP‐expressing SMCs compared with diabetic controls. Given that these SMC states are associated with differential roles in vascular structure and stability, the observed redistribution indicates phenotypic reprogramming toward a more mature and stabilized vascular architecture in the chronic wound microenvironment.

Feature‐level quantification extracted from the single‐cell dataset further revealed significantly increased expression of key regenerative mediators in BOOST‐treated wounds. In particular, the relative expression levels of nerve growth factor (NGF) and vascular endothelial growth factor A (VEGFA) were elevated compared with diabetic controls (Figure 7d,e). Increased NGF expression suggests enhanced neurotrophic support that may facilitate re‐innervation, whereas elevated VEGFA is consistent with strengthened angiogenic signaling to promote neovascularization and improve local perfusion.

Beyond vascular remodeling, the MSCs‐encapsulating BOOST device induced broad transcriptional reprogramming of the immune microenvironment (Figure 7f). In mGDTh17 cells, gene set enrichment analysis revealed significant downregulation of pathways associated with cytoskeletal organization, focal adhesion, and extracellular matrix–receptor interactions (Figure 7g), indicative of reduced migratory and tissue‐infiltrative potential. In parallel, inflammatory and metabolism‐related pathways—including PI3K–Akt signaling, AGE–RAGE signaling in diabetic complications, and ether lipid metabolism—were significantly suppressed [48]. Downregulation of pathways linked to aberrant activation states, such as amoebiasis and human papillomavirus infection signatures, further suggests attenuation of pathological immune activation. Collectively, these changes indicate normalization of pro‐inflammatory transcriptional programs in mGDTh17 cells following BOOST‐mediated therapy.

Macrophages also exhibited pronounced transcriptional remodeling in response to BOOST‐mediated therapy (Figure 7h). Pathway enrichment analysis revealed downregulation of gene programs associated with lysosomal activity, endoplasmic reticulum stress responses, autophagy, phagosome formation, and proteasome‐mediated protein turnover. Rather than indicating loss of macrophage functionality, these changes are consistent with a reduction in stress‐ and damage‐associated transcriptional programs that are typically elevated in chronically inflamed, nonhealing wounds. Together, these findings suggest that the MSCs‐encapsulating BOOST device promotes a shift of macrophages toward a more homeostatic and inflammation‐resolving state, thereby contributing to improved immune regulation within diabetic‐chronic wounds. Notably, immune cell remodeling in the BOOST‐treated group was more pronounced than that observed with MSC encapsulation alone (Figure S27), highlighting the role of sustained microenvironmental modulation in shaping immune cell behavior. Together, these single‐cell analyses indicate that sustained, microenvironment‐responsive MSC delivery reshapes both vascular and immune compartments of diabetic wounds, providing mechanistic support for the therapeutic paradigm established by the BOOST platform.

3. Discussion

The MSC‐based therapy is widely recognized as a biologically effective approach for chronic wound repair [49]. A large body of preclinical and clinical evidence demonstrates that MSCs can promote tissue regeneration through dynamic paracrine modulation of inflammation, angiogenesis, and extracellular matrix remodeling [50]. Importantly, the limited and inconsistent efficacy observed in many clinical studies does not reflect a failure of MSC biology itself, but instead points to a fundamental mismatch between living cell therapies and the delivery paradigms currently used in clinical practice. Our findings support the view that the principal barrier to effective MSC therapy lies not in the intrinsic therapeutic potential of the cells, but in the inability of existing delivery strategies to sustain cell viability and function within hostile host microenvironments.

Both major clinical routes of MSC administration—intravenous infusion, as exemplified by multiple ongoing trials (for example, NCT05158127), and local implantation approaches (for example, NCT06103409 and NCT06122532)—expose transplanted cells to conditions that are largely incompatible with long‐term survival and functional persistence. Systemic intravenous delivery is fundamentally constrained by blood‐cell incompatibility and unfavorable biodistribution. Upon infusion, MSCs are immediately recognized as hemoincompatible cells, triggering IBMIR characterized by activation of coagulation and complement cascades, platelet aggregation, and innate immune responses [51]. In parallel, the large size and limited deformability of MSCs result in a pronounced pulmonary first‐pass effect, with approximately 50%–80% of cells mechanically trapped within the lung microvasculature during the first circulation pass [23, 52, 53]. As a consequence, only a small fraction (∼1%–5%) of administered MSCs reaches peripheral target tissues, while most cells are rapidly inactivated or eliminated. Local delivery circumvents these systemic barriers but places MSCs directly into ischemic wound beds characterized by severe vascular damage, profound hypoxia, and persistent inflammation, leading to rapid cell death and loss of paracrine function. Compounding these challenges, MSCs are increasingly recognized as immune evasive rather than immune privileged, rendering them vulnerable to immune‐mediated clearance in both local and systemic settings [24]. Collectively, these observations indicate that therapeutic failure of MSCs in chronic wounds is predominantly a consequence of delivery paradigm failure.

In this work, we introduce a fundamentally different paradigm for cell‐based wound therapy by establishing a local, in situ, long‐term, and immune‐compatible MSC delivery platform. The BOOST system integrates sustained cell encapsulation with bio‐responsive O₂ generation to actively align living cell function with the wound microenvironment. A critical conceptual distinction of this platform is that O₂ generation is not intended to act as a conventional wound O₂ therapy. Instead, O₂ is supplied specifically to maintain high cellular viability, metabolic activity, and paracrine function of encapsulated MSCs under otherwise prohibitive hypoxic conditions. This distinction is directly supported by the observation that acellular BOOST devices provide limited therapeutic benefit compared with MSC‐loaded devices, underscoring that oxygenation in this system functions as an enabling factor for sustained cell activity rather than as an independent therapeutic modality.

A central advance of the BOOST platform is the recognition that cell encapsulation is not merely protective, but essential for clinical translation of cell‐based wound therapies. Encapsulation within an immunoisolating alginate hydrogel matrix prevents rapid cell loss, shields MSCs from immune attack, and decouples therapeutic efficacy [9] from immediate host immune recognition [54]. This feature enables the use of allogeneic or even xenogeneic MSCs, as validated in the porcine wound model, and addresses one of the most persistent bottlenecks in regenerative medicine. Importantly, the alginate hydrogel employed in this study undergoes gelation under mild calcium‐ion conditions, minimizing cellular stress. The combined use of D‐glucono‐δ‐lactone (GDL) and CaCO₃ enables controlled and gradual crosslinking, allowing rapid filling and conformal gelation within wounds of diverse and irregular geometries. Integration of three‐dimensional wound scanning further enables patient‐specific control of device geometry and cell dosage, aligning this delivery paradigm with emerging precision‐medicine frameworks.

From a translational perspective, the BOOST delivery paradigm aligns naturally with the rapidly advancing infrastructure of MSC manufacturing and banking. Large‐scale MSC production, cryopreservation, and quality control have matured substantially, with multiple MSC‐based products already approved for clinical use [55, 56]. It is therefore increasingly plausible that, in the near future, cell banking may become as standardized and accessible as blood banking, enabling on‐demand availability of therapeutic cells. In this context, the dominant translational bottleneck is no longer cell supply, but the lack of delivery systems capable of preserving cell potency after administration. By providing a scalable, wound‐adaptive, and immune‐compatible encapsulation platform, the BOOST device directly addresses this gap and enables standardized cell products to function effectively within complex wound environments. Beyond chronic wounds, the conceptual framework established here has broader implications for regenerative medicine. Many cell therapies fail in ischemic, inflamed, or mechanically hostile tissues, not because the cells lack therapeutic efficacy, but because delivery strategies do not account for environmental constraints imposed by the host. By reframing cell therapy as a problem of environment–cell compatibility, rather than cell potency alone, the BOOST platform offers a generalizable blueprint for cell delivery systems in hypoxic and ischemic disease contexts.

In summary, this study redefines chronic wound cell therapy by addressing the failure of existing delivery paradigms to sustain viable and functional cells within hostile tissue environments. Through self‐oxygenating, wound‐adaptive, and immune‐compatible encapsulation, the BOOST platform preserves MSC survival and functional persistence in situ, thereby transforming cell therapy from a transient intervention into a controllable and durable treatment strategy. More broadly, this work shifts the focus of regenerative medicine from enhancing intrinsic cell potency to engineering environment‐cell compatibility as a controllable therapeutic variable, establishing a translational framework for cell‐delivery systems.

4. Experimental Section

4.1. Materials

All reagents were used as received unless otherwise stated. Polydimethylsiloxane (PDMS, Sylgard 184) was obtained from Dow Corning (Midland, MI, USA). Lithium peroxide (Li₂O₂) and fetal bovine serum were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Perfluorocarbon (PFC, model GPL103) was purchased from DuPont (Wilmington, DE, USA). Ethylene‐vinyl alcohol (EVOH) copolymer film was purchased from Wenxi Packaging Products Co., Ltd. (China). Streptozotocin (STZ) was obtained from Macklin Biochemical Co., Ltd. (Shanghai, China). Sodium citrate buffer was purchased from J&K Scientific Ltd. (Shanghai, China). The cell counting kit‐8 (CCK‐8) assay kit and reactive oxygen species (ROS) detection kit were purchased from Beyotime Biotechnology (Shanghai, China). PRONOVA UP MVG medium‐viscosity sodium alginate (G‐content ≥60%) and alginate lyase were purchased from Sigma‐Aldrich (St. Louis, MO, USA). Dulbecco's Modified Eagle Medium (DMEM) was obtained from Biochannel (Beijing, China). Penicillin‒streptomycin solution, phosphate buffer solution (PBS), and trypsin were purchased from Biosharp (Hefei, China). Dichloromethane, acetic anhydride, calcium chloride dihydrate (CaCl₂·2H₂O), Sodium chloride (NaCl), calcium carbonate (CaCO₃), and sodium bicarbonate (NaHCO₃) were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Lithium carbonate (Li₂CO₃), 3‐aminopropyltriethoxysilane (APTES), dithiodiglycolic acid, acrylic acid (AA), N‐Hydroxysuccinimide (NHS), dicyclohexylcarbodiimide (DCC), polyvinyl alcohol (PVA, 1788 grade), 4‐dimethylaminopyridine (DMAP), α‐Ketoglutaric acid (α‐KG), glutathione (GSH), D‐Gluconic acid δ‐lactone (GDL), and D‐glucose were purchased from Aladdin (Shanghai, China). Chitosan quaternary ammonium salt was purchased from Macklin (Shanghai, China). 2‐Hydroxyethyl methacrylate (HEMA) was purchased from Bide Pharmatech (Shanghai, China). Sodium citrate buffer (0.1 mol/L, pH 4.5) was purchased from Acmec Biochemical Co., Ltd (Shanghai, China). Anti‐iNOS antibody (Cat. No. 18985‐1‐AP) was purchased from Proteintech (used at a 1:200 dilution). Anti‐IL‐1β antibody (Cat. No. HA723965) was obtained from HUABIO (dilution 1:500). Anti‐CD31 antibody (ab182981) was sourced from Abcam (dilution 1:2000). Fresh porcine skin was obtained from a local market and used immediately upon arrival.

4.2. Cell Lines and Animals

Rat adipose‐derived mesenchymal stem cells (rADMSCs) were obtained from Procell Life Science & Technology Co., Ltd. (Wuhan, China). Porcine adipose‐derived mesenchymal stem cells (pADMSCs) were purchased from Shanghai Enzyme‐linked Biotechnology Co., Ltd. (Shanghai, China). The cells were cultured according to the supplier's protocol. Male Sprague‐Dawley rats (200 g) were purchased from Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). Male Bama miniature pigs (30 kg) were purchased from Jiangsu W&Z Bio‐Pharmaceutical Co., Ltd. (Jiangsu, China). The rats had access to food and water ad libitum and were hosted in ambient temperature (21°C–26°C), humidity at 50%–60%, under 12 h dark/light cycles. Miniature pigs were housed in single cages. The ambient temperature of the animal room was 18°C–22°C, the daily temperature difference was not more than 4°C, the relative humidity was 40%–70%, and 12 h of lighting/12 h of darkness were alternated between light and darkness. All animal procedures were approved by the Institutional Animal Care and Use Committee of the First Affiliated Hospital of University of Science and Technology of China [Approval No. 2025‐N(A)‐0147].

4.3. Fabrication of BOOST Device

The BOOST device was constructed as a multilayer encapsulation system comprising a CO₂‐responsive O₂‐generating module, a gas‐permeable structural shell, an MSC‐encapsulating alginate hydrogel core, and a bioadhesive sealing interface.

The structural shell was fabricated from polydimethylsiloxane (PDMS; Sylgard 184). A stainless‐steel mesh was embedded during curing to enhance mechanical stability. Individual PDMS components were assembled and bonded using a thin layer of uncured Sylgard 184 prepolymer, followed by thermal curing to ensure airtight sealing of the device structure. To minimize O₂ loss to the external environment, the PDMS surface (0.3 mm thick) facing away from the wound was laminated with an EVOH membrane (0.1 mm). An annular PDMS tube was incorporated within the device to reinforce the hydrogel compartment (2 mm thick) and facilitate gas diffusion from the O₂ reservoir to the cell‐laden core. The O₂‐generating compartment was loaded with a Li₂O₂/PFC composite (20% w/w). Li₂O₂ served as a CO₂‐responsive oxygen source, while PFC enhanced oxygen solubility and retention. The composite was confined within the PDMS reservoir to prevent direct contact with cells or wound tissue. Prior to integration with the adhesive interface, PDMS surfaces were rinsed with isopropanol, dried at 60 °C, and treated with oxygen plasma (150 W, 3 min; TS‐SY05, Tonson Tech, China) to generate surface hydroxyl groups. The activated PDMS was immersed in 10% (v/v) APTES in deionized water for 20 min, rinsed thoroughly, and dried at 60°C for 24 h to introduce surface amine functionalities.

For in‐device cell encapsulation, ADMSCs were suspended in sterile sodium alginate (final concentration: 1% w/v; cell density: 1.5 × 10⁶ cells/mL). For rat‐implantation experiments, gelation was achieved by using calcium sulfate (CaSO₄) deposited on the device to provide Ca^{2 +} for crosslinking. The resulting construct was then immersed in 100 mM CaCl₂ for 5 min to complete curing, yielding a conformal cell‐encapsulated hydrogel core. The bioadhesive sealing interface was prepared using an NHS ester–functionalized acrylic monomer containing a disulfide bond (synthesized according to Zhao et al. [32]). PVA (7 wt%), acrylic acid (35 wt%), hydroxypropyltrimethyl ammonium chloride chitosan (1 wt%), and α‐ketoglutaric acid (0.2 wt%) were dissolved in deionized water. The functional monomer (100 mg in 1 mL DMSO) was added to 10 mL of the precursor solution, cast into a PTFE mold (500 µm thickness), and polymerized under ultraviolet irradiation (284 nm, 10 W) for 30 min. The resulting hydrogel (0.1 mm thick) was integrated with the APTES‐modified PDMS surface to form a stable sealing interface prior to wound application.

4.4. Compressive and Tensile Modulus Measurement of Bioadhesive Sealing Interface

Mechanical tests were performed on a texture analyzer (BosinTech Co., Ltd., China) in compression and tension modes. For compression, cubic bioadhesive sealing interface samples were subjected to unconfined compression between parallel plates at 0.5 mm/s; the nominal contact area was 19.625 mm², which was used to calculate engineering stress (σ = F/A). The compressive modulus was obtained from the slope of the initial linear region of the stress–strain curve (0%–10% strain). For tension, bioadhesive sealing interface strips were clamped and stretched at 50 mm/min until failure. The tensile modulus was determined by linear fitting of the elastic region of the stress‐strain curve (0%–5% strain).

4.5. Surface Elemental Analysis and Hydrophilicity Test of PDMS Membrane

X‐Ray photoelectron spectroscopy (XPS, Thermo Fisher ESCALAB 250Xi, Al Kα = 1486.6 eV, 150 W) was employed to analyze the surface elemental composition of PDMS and PDMS‐APTES. Before measurement, all samples were cleaned with ethanol and deionized water and dried under a nitrogen flow. The wettability of the PDMS membrane surface was evaluated at room temperature using a contact angle goniometer (SL200B, Solon Tech. Co., Ltd., China). Deionized water (5 µL) was dispensed onto the sample surface using a microsyringe, and the contact angle was recorded by the sessile drop method after equilibrium was reached.

4.6. In Vitro Antibacterial Zone of Inhibition Test

The antibacterial activity of the samples was evaluated using the agar diffusion method against Staphylococcus aureus (S. aureus) and methicillin‐resistant S. aureus (MRSA). Briefly, bacterial strains were cultured overnight in nutrient broth at 37°C, and the bacterial suspensions were adjusted to approximately 1 × 10⁸ CFU mL^{− 1}. Sterile nutrient agar plates were uniformly inoculated with 100 µL of each bacterial suspension using sterile cotton swabs. Discs (10 mm in diameter) of the bioadhesive sealing interface without quaternary ammonium chitosan and the bioadhesive sealing interface containing quaternary ammonium chitosan were aseptically placed onto the inoculated agar surfaces. The plates were then incubated at 37°C for 24 h, after which the diameter of the clear inhibition zones around each sample was measured using a digital caliper. All experiments were performed in triplicate, and the average inhibition zone diameter was recorded as the antibacterial activity.

4.7. Adhesion Test Between Bioadhesive Sealing Interface and Porcine Skin Tissue/PDMS Membrane

The adhesion strength between the bioadhesive sealing interface and skin tissue/PDMS membrane was evaluated using a 180° peel test. Bioadhesive sealing interface samples were adhered to the surface of skin tissue/PDMS membrane with a contact width of 3 cm, and the peeling force was measured at a crosshead speed of 50 mm/min using a Texture Analyzer (TA.XTC20, BosinTech Co., Ltd., China).

4.8. In Vitro Cell Viability of ADMSCs

Both encapsulated and adherent ADMSCs (1 × 10⁶ cells per sample) were cultured under O₂ concentrations of 1%, 5%, 10%, and 21% for 48 h. For comparison, cells (1 × 10⁶ cells per sample) encapsulated in the BOOST device and ctrl device were also cultured under 1% O₂ for 48 h. After incubation, cells were retrieved from cross‐linked hydrogels by digesting with alginate lyase (3.4 mg/mL) for 15 min at 37°C. At last, cell viability in each group was determined using the CCK‐8 assay, and the absorbance was recorded at 450 nm with a microplate reader (Infinite M PELX, Tecan, Switzerland).

4.9. Intracellular ROS Detection

ADMSCs encapsulated in alginate hydrogels were incubated for 48 h under 1%, 5%, and 21% O₂ conditions. Meanwhile, ADMSCs encapsulated in the device were incubated under 1% O₂ conditions for 48 h. Intracellular ROS levels were measured using a ROS detection kit. Cells were incubated with 2′,7′‐dichlorodihydrofluorescein diacetate (DCFH‐DA) at 37°C for 30 min in the dark. After incubation, the cells were washed with PBS to remove excess dye. The oxidation of DCFH–DA by ROS generates a green fluorescent signal, which was detected using an inverted fluorescence microscope (IX73, Olympus, Japan).

4.10. Computational Modeling

A computational framework was developed to simulate O₂ transport within the device. To reduce computational complexity and simulation time while preserving the essential mass‐transport behavior, a dispersed cell encapsulation model was employed to represent the hydrogel, in which MSCs were assumed to be uniformly distributed throughout the entire hydrogel domain. The experimentally measured O₂ generation rate of the perfluorocarbon (PFC) phase was applied uniformly within the corresponding domain, and the MSC density was set to 1 × 10¹² cells m⁻³. Owing to the concentric configuration of the device, 2D axisymmetric geometries were constructed to represent the 3D structure.

Physiologically relevant boundary conditions were applied. A constant O₂ partial pressure (pO₂) of 5 mmHg was prescribed at the device‐host interface to represent the hypoxic in vivo environment, while a constant pO₂ of 160 mmHg was imposed at the device‐air interface to represent ambient air conditions. O₂ transport in all domains was assumed to occur under steady‐state conditions with negligible convection.

$$D_{O_2,i}\left(\frac{1}{r}\frac{\partial }{\partial r}\left(r\frac{\partial c_{O_2}}{\partial r}\right)+\frac{1}{r^2}\frac{{\partial }^2{c}_{O_2}}{\partial {\theta }^2}+\frac{{\partial }^2{c}_{O_2}}{\partial z^2}\right)=R_{O_2,i}$$

where, O₂ consumption or generation was included only in the hydrogel and PFC domains and set to zero elsewhere. O₂ generation in the PFC domain followed the experimentally determined rate 5 × 10⁻¹¹ mol m⁻³ s⁻¹), while O₂ consumption in the hydrogel was described using Michaelis–Menten kinetics.

$$R_{O_2,i}=\left\{\begin{array}{c}0,c_{O_2}<c_{necrosis}\\ -\frac{c_{O_2}{R}_{max}}{c_{O_2}+K_m},c_{O_2}\ge c_{necrosis}\end{array}\right.$$

O₂ concentration was converted to partial pressure using Henry's law based on domain‐specific O₂ solubility coefficients.

$$p{O}_2=\frac{c_{O_2}}{{\alpha }_{O_2,i}}$$

At all internal material interfaces, partition coefficients were applied to account for discontinuities in O₂ solubility between adjacent materials.

$$K_{i|j}=\frac{{\alpha }_{O_2,i}}{{\alpha }_{O_2,j}}$$

All simulations were performed using COMSOL Multiphysics. Computational meshes were generated using the “extremely fine” setting to ensure numerical accuracy. Material properties, O₂ transport parameters, and cellular consumption constants used in the simulations are summarized in the table.

	Description	Symbol	Value(units)	Refs.
O2 diffusion coefficients	Hydrogel	$D_{O_2,h}$	2.7 × 10−9 m2 s−1	[57]
	PDMS	$D_{O_2,PMDS}$	3.25 × 10−9 m2 s−1	[58]
	PFC	$D_{O_2,PFC}$	5.6 × 10−9 m2 s−1	[57]
	EVOH	$D_{O_2,EVOH}$	1 × 10−14 m2 s−1	[59]
	Air	$D_{O_2,air}$	1.8 × 10−5 m2 s−1	[60]
O2 solubility coefficients	hydrogel	${\alpha }_{O_2,h}$	9.3 × 10−6 mol m−3 Pa−1	[57]
	PDMS	${\alpha }_{O_2,PMDS}$	7.3 × 10−5 mol m−3 Pa−1	[61]
	PFC	${\alpha }_{O_2,PFC}$	1.9 × 10−4 mol m−3 Pa−1	[57]
	EVOH	${\alpha }_{O_2,EVOH}$	4 × 10−8 mol m−3 Pa−1	[59]
	air	${\alpha }_{O_2,air}$	3.9 × 10−4 mol m−3 Pa−1	[60]
O2 consumption parameter	Maximum consumption rate	Rmax	6.565 × 10−3 mol m−3 Pa−1	[35]
	Half‐maximal coefficient	Km	0.212 mol m−3	[35]
	Viability threshold	cnecrosis	1 × 10−4 mol m−3	[62]

4.11. In Vivo Diabetic Rat Skin Wound Healing Study

Thirty‐six male Sprague–Dawley rats (6 weeks old, 200 ± 5 g; Shanghai Slack Laboratory Animal Co., Ltd.) were randomly assigned to six groups (n = 6 per group). Diabetes was induced in five groups by intraperitoneal injection of streptozotocin (STZ, 60 mg kg^{− 1}). Ten days post‐injection, rats exhibiting persistent fasting blood glucose levels above 16.7 mmol L^{− 1} were considered diabetic and included in subsequent experiments.

A circular full‐thickness excisional wound (15 mm diameter) was created on the dorsal skin of each animal under anesthesia. Wound healing progression was monitored over a 15‐day period through serial digital imaging. Blood glucose levels and body weight were recorded throughout the study. For in vivo cell survival analysis, MSC‐loaded BOOST or control devices were retrieved on day 3 post‐implantation. The alginate matrix was enzymatically degraded using alginate lyase to release encapsulated cells, and cell viability was quantified using a CCK‐8 assay. Wound tissues were harvested on days 7 and 15 for histological evaluation and proteomic analysis. Wound area measurements were calculated from serial images of the same wound at predefined time points.

4.12. Data‐Independent Acquisition (DIA)‐Based Quantitative Proteomics of Wound Tissues

4.12.1. Sample Preparation

Sample preparation contains the process of protein extraction, denaturation, reduction, alkylation, as well as the tryptic digestion and peptide cleanup. Commercially available iST Sample Preparation kit (PreOmics, Germany) was used according to protocols provided. Briefly, 50 µL of Lyse buffer was added and heated at 95°C for 10 min at 1000 rpm with agitation. After cooling the sample to room temperature, trypsin digestion buffer was added, and the sample was incubated at 37°C for 2 h at 500 rpm with shaking. The digestion process was stopped with a stop buffer. Sample clean‐up and desalting were carried out in the iST cartridge using the recommended wash buffers. Peptides were eluted with elution buffer (2 × 100 µL) and then lyophilized by SpeedVac.

4.12.2. Data‐Independent Acquisition (DIA) Mass Spectrometry

Peptide samples were analyzed on a timsTOF HT ion‐mobility quadrupole time‐of‐flight mass spectrometer (Bruker Daltonik, Bremen, Germany) coupled to a Vanquish Neo UHPLC system (Thermo Fisher Scientific, MA, USA). Dried peptides were reconstituted in 0.1% formic acid (FA), and 200 ng of peptides were loaded onto an AUR3‐15075C18 analytical column (8 cm length, 75 µm inner diameter, 1.7 µm particle size, 120 Å pore size; IonOpticks). Peptides were separated using a 30‐min gradient at a flow rate of 300 nL/min. The gradient started at 2.2% buffer B (80% acetonitrile with 0.1% FA), increased to 44% over 21 min, ramped to 90% within 3 min, and was held for 3 min, followed by a decrease to 4% and re‐equilibration for 3 min. The column temperature was maintained at 50°C throughout the analysis. DIA data were acquired in the diaPASEF mode. We defined 82 × 10 Th precursor isolation windows from m/z 380.8 to 1,170.8. To adapt the MS1 cycle time, we set the repetitions to 3–6 steps in the 16‐scan diaPASEF scheme in our experiments. During PASEF MS/MS scanning, the collision energy was ramped linearly as a function of the mobility from 59 eV at 1/K0 = 1.6 Vs/cm² to 20 eV at 1/K0 = 0.6 Vs/cm².

4.12.3. Database Search

Raw DIA data were processed and analyzed using Spectronaut 19 (Biognosys AG, Switzerland) with default settings. The reference database was the Rattus norvegicus UniProt proteome (version 2025, 22,369 entries). Trypsin was specified as the digestion enzyme with fully specific cleavage. Carbamidomethylation of cysteine was set as a fixed modification, whereas methionine oxidation and protein N‐terminal acetylation were defined as variable modifications. Retention time prediction was performed using the dynamic iRT model. Data extraction was carried out automatically by Spectronaut based on extensive mass calibration, and the software dynamically determined the optimal extraction window according to iRT calibration and gradient stability. False discovery rate (FDR) thresholds were controlled at 1% at the precursor, peptide, and protein levels (Q‐value < 0.01). Decoy generation was set to the mutated method, which is similar to the scrambled approach but applies a random number of amino‐acid position swaps (minimum = 2; maximum = length/2). For pathway enrichment and data visualization, FDR‐adjusted p values were calculated using the Benjamini–Hochberg procedure, and an FDR threshold of < 0.25 was used as an exploratory criterion for identifying biologically relevant pathways.

All raw proteomics datasets have been deposited in the iProX public repository under accession number IPX0017512000.

4.13. In Vivo Diabetic Porcine Skin Wound Healing Study

Male Bama miniature pigs (approximately 30 kg, n = 2) were rendered diabetic by a single intravenous injection of streptozotocin (STZ, 120 mg kg^{− 1}) dissolved in sodium citrate buffer (35 mg mL^{− 1}, pH 4.5). Blood glucose levels were continuously monitored using a continuous glucose monitoring system (CGMS; SGBio, China). Only animals maintaining persistent hyperglycemia for more than one month were included for subsequent wound‐healing experiments.

Six full‐thickness dermal wounds (2 × 2 cm) were created on the dorsal region of each pig under general anesthesia. Immediately after wound creation, each wound was scanned using a three‐dimensional scanner (Reeyee Three, Wiiboox; accuracy, 0.2 mm) to reconstruct surface morphology and calculate wound volume for individualized cell dosing. Wounds were assigned to three treatment groups: (i) alginate hydrogel alone (control), covered with a sterile 3 M film dressing; (ii) MSCs combined with the control device; and (iii) MSCs combined with the BOOST device. For cell encapsulation, MSCs were suspended in 2% (w/v) sodium alginate solution containing 0.4% (w/v) glucose and 100 mM CaCO₃ at a final density of 1.5 × 10⁶ cells mL^{− 1}. The suspension was mixed with an equal volume of 200 mM GDL solution immediately prior to application to initiate controlled internal gelation. The resulting precursor was applied to the wound bed or loaded into the device for in situ crosslinking and treatment, and the gelation time is approximately 20 min under the conditions used. Dressings were replaced every five days. Digital photographs were acquired at each dressing change to document wound closure. Upon complete healing or at predefined endpoints, regenerated tissues were harvested for histological evaluation, including hematoxylin and eosin (H&E) staining and immunohistochemical analysis. Quantitative wound area measurements were obtained from serial images of the same wound over time.

4.14. Tissue Dissociation, Single‐Cell/Single‐Nucleus Preparation, and Library Construction

Diabetic wound tissues were processed on ice to obtain both single‐cell and single‐nucleus suspensions using a slightly modified 10x Genomics protocol. Briefly, tissues were mechanically dissociated and subjected to enzymatic digestion for single‐cell preparation, whereas nuclei were isolated by cold lysis followed by washing and resuspension. Suspensions were filtered and centrifuged to remove debris, red blood cells were lysed when necessary, and cell viability was assessed by trypan blue staining. Single‐cell/nucleus suspensions were loaded onto the 10x Chromium platform for 3′ gene expression library construction using the Chromium Single Cell 3′ kit (v3) according to the manufacturer's instructions. Libraries were sequenced on an Illumina NovaSeq 6000 by Shanghai Personal Biotechnology (Shanghai, China).

4.15. Single‐Cell RNA‐Seq Processing, Clustering, and Annotation

Raw FASTQ files were processed using Cell Ranger v7.1.0 (10x Genomics) and aligned to the Sus scrofa reference genome and gene annotation from Ensembl (specify release/version). For single‐nucleus datasets, intronic reads were included during quantification when applicable. Gene expression matrices were imported into Seurat v5 for downstream analysis. Cells/nuclei with  < 400 detected genes or  > 7,500 detected genes, as well as those with >20% mitochondrial transcripts, were excluded. Data were normalized and log‐transformed using NormalizeData, highly variable features were identified (e.g., FindVariableFeatures, nfeatures = 2000), followed by scaling, dimensionality reduction, graph‐based clustering, and visualization using UMAP. Cell types and subtypes were assigned based on canonical marker gene expression.

Funding

The National Key R&D Program of China (2024YFB3814600), the Natural Science Foundation of China (52473155), and “USTC Research Funds of the Double First‐Class Initiative” (YD2060002501).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adma74161‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.