Zein‐Ceria Hybrid Microparticles Enable Long‐Term ROS‐Scavenging Oxygenation for Osteogenic Microtissues Engineering

ABSTRACT

Oxygen‐ and biological cue‐deprived microenvironments formed during tissue regeneration severely limit cell survival and differentiation, resulting in long‐term structural and functional deficits. However, conventional oxygen‐releasing biomaterials often exhibit burst releases, with the vast majority of oxygen released during the first few days, which is associated with high levels of concomitant reactive oxygen species (ROS)‐derived oxidative stress and a lack of bioactive factors. Here, we report a hierarchically engineered zein‐ceria hybrid microparticle that enables sustained ROS‐neutral oxygenation for over 40 days and supplies an osteoinductive factor. A hydrophobic zein core stabilizes the oxygen source and suppresses burst release, while a ceria nanozyme‐integrated shell continuously scavenges excess ROS via redox cycling. Biocompatible surface engineering enables the seamless integration of these microparticles within stem cell spheroids, which markedly enhances cell survival under anoxia. Their biofunctional surface supports enzymatic protein immobilization under physiological conditions, enabling spontaneous osteogenesis of engineered bone microtissues. In a severely oxygen‐deprived mouse calvarial defect model, the engineered microtissues accelerated bone regeneration. Our biomaterial design enables control of burst oxygen release, ROS modulation, and growth factor release, built on a zein‐ceria double‐layer architecture, offering a modular platform that broadens the utility of oxygenating and bioactive micromaterials in regenerative medicine.

A zein‐ceria hybrid microparticle enables sustained reactive oxygen species (ROS)‐neutral oxygenation by suppressing burst release under an oxygen‐depleted microenvironment. Integrated within stem cell spheroids, the microparticles enhance cell survival under severe hypoxia and provide a biofunctional platform for regenerative microtissues and accelerated bone repair.

1. Introduction

Engineered microtissues densely populate microenvironments that enhance cell‐cell interactions and matrix synthesis, thereby influencing cell fate and supporting native‐like cellular organization without physical barriers (i.e., scaffolds), offering great potential for treating various injuries and diseases [1, 2, 3, 4, 5]. Despite their significant promise in regenerative medicine, tissue regeneration in critical‐sized defects via the transplantation of engineered microtissues assembled into large‐scale tissues is profoundly limited by significant challenges. The major issues are insufficient vascularization, which limits nutrients and oxygen supply to large, dense tissues, and a lack of biological cues needed to ensure proper spatial organization and differentiation of the microtissues [6]. This limitation is further exacerbated in dense tissues such as bone, where vascular ingrowth is inherently slow and spatially restricted, underscoring the need for alternative engineering strategies to support microtissues until sufficient vascularization [7].

Severe hypoxia due to insufficient vascularization compromises cell survival, hinders matrix remodeling and tissue integration, and delays the restoration of homeostatic conditions [8]. Although vascularization can eventually restore oxygen levels later, the early post‐implantation phase and the cores of critical‐size defects often experience severe oxygen deficiency before vascularization, creating a bottleneck for regenerative therapies [9]. While mild hypoxia (1%–3% O₂) can stimulate tissue regeneration by promoting angiogenic responses or cell proliferation, severe hypoxia (< 0.5% O₂) leads to rapid loss of cellular function and ultimately necrosis [10, 11]. Therefore, it has been predicted that local oxygen delivery can markedly improve regenerative processes in the defect [12]. However, achieving cytocompatible, long‐term oxygen delivery using a scaffold‐free microtissue with limited biomaterials remains an unmet challenge.

Although oxygenating biomaterials have been explored to address this limitation, most suffer from uncontrolled burst release and generation of cytotoxic byproducts [13]. A key challenge arises from the intrinsic difficulty of retaining gaseous oxygen in aqueous environments, given its high diffusivity [14]. To circumvent this issue, solid‐state oxygen donors such as calcium peroxide (CPO) or liquid complexes such as poly(vinylpyrrolidone)‐hydrogen peroxide (PVP‐H₂O₂) have been employed [15]. However, these sources are highly susceptible to premature hydrolysis and increase local accumulation of byproducts, paradoxically worsening the regenerative microenvironment, thereby requiring stabilizing strategies to ensure safe and sustained oxygen delivery [16].

Zein is a hydrophobic prolamin storage protein abundantly found in the maize endosperm, where it functions as a nutrient reservoir and contributes to the stability of seeds during germination [17]. Beyond its natural role, zein exhibits several physicochemical characteristics that make it attractive as a biomaterial, including excellent film‐forming ability, controlled water‐solubility, and strong encapsulation capacity for diverse molecules [18]. Zein has also been shown to be a promising macromolecule for developing biomaterials to regenerate vessels [19], cardiac tissue [20], bone [21], and skin [22], by delivering bioactive molecules safely. Inspired by this intrinsic protective role in nature, we employed zein as a stabilizing matrix to encapsulate oxygen donors and suppress premature burst release.

Another major limitation of currently reported oxygenating biomaterials is the accumulation of reactive oxygen species (ROS)‐induced stress [23]. During tissue regeneration, ROS generated from metabolic reactions or the implanted biomaterials further aggravates oxidative damage [24]. Although antioxidants such as polyphenols and enzymes have been investigated, these compounds typically suffer from a short half‐life, thus becoming quickly less effective over time. For example, catalase is characterized by a half‐life of around 30 h at 37°C, which becomes significantly shorter in biological environments with proteolytic enzymes [25, 26]. To extend catalytic activity, catalase encapsulated in hydrogels exhibited increased decomposition of H₂O₂ by > 1 week compared with pristine catalase [27]. Nevertheless, this represents a limitation in using catalase in long‐term in vivo applications. In contrast, nanozymes such as cerium oxide (ceria) and manganese dioxide (MnO₂) have been widely applied because of their stable catalytic activity and structural robustness. Among these, ceria is particularly promising, as its Ce³⁺/Ce⁴⁺ redox cycle enables both sustained ROS scavenging and oxygen generation through oxygen vacancies [28]. Additionally, ceria shows biodegradability, biocompatibility, and anti‐inflammatory properties [29]. It may also degrade more slowly than catalase. Therefore, ceria is considered a viable alternative for the decomposition of H₂O₂ over 40 days.

A locally sustained supply of biological cues, in combination with oxygen supply, is necessary to induce cellular differentiation in the microtissues. To do this, a growth factor is incorporated into an oxygen‐generating system, such as bone morphogenetic protein‐2 (BMP‐2), which is well known for inducing osteogenesis and facilitating bone formation in vivo, and will enhance osteoinductive activity while minimizing the adverse effects of BMP‐2, including ectopic bone formation, inflammation, osteolysis, and the risk of tumorigenesis [30, 31]. By utilizing a combination system, we can anticipate synergistic effects from supplying both the osteoinductive factor and oxygen, which will simultaneously promote prolonged cell survival and enhanced osteogenesis in microtissues, ultimately leading to bone tissue formation at large‐sized bone defects.

Here, we designed a multi‐layered oxygen‐releasing and ROS‐modulating microparticle composed of a CPO‐zein core and a gelatin‐ceria shell using a single emulsion process. This architecture stabilizes the oxygen source against aqueous decomposition and enables fine‐tuned, sustained oxygen release for over 40 days. Simultaneously, the ceria‐laden gelatin shell scavenges excess intrinsic and extrinsic ROS via the Ce³⁺/Ce⁴⁺ redox cycle, thereby mitigating oxidative stress. The gelatin shell‐layered oxygen‐releasing microparticles were successfully embedded within mesenchymal stem cell (MSC) spheroids via abundant cell‐binding sites, thereby markedly enhancing cell survival under anoxic conditions (< 0.1% O₂) by preventing cell death. Moreover, the outer gelatin layer provided surface biofunctionality for protein immobilization, allowing for the incorporation of osteoinductive BMP‐2. Sustained release of BMP‐2 further induced osteogenic differentiation of MSC spheroids into bone microtissues. As a proof of concept, we demonstrated that this dual‐functional platform promotes bone regeneration in critical‐sized defects, thereby overcoming hypoxia, ROS‐associated barriers, and the lack of bioactive factors in regenerative medicine. Altogether, this work establishes a structurally engineered oxygen‐ROS‐regulating platform that addresses the intrinsic limitations of conventional oxygenating biomaterials and offers a broadly applicable strategy to enhance tissue regeneration in harsh microenvironments.

2. Results and Discussion

2.1. Design Strategy and Fabrication of Zein‐CPO‐Ceria Hybrid Microparticles (ZCCs)

In our previous study, we investigated the physical locations of a ROS‐scavenging enzyme (catalase) to efficiently decompose H₂O₂, which is produced during oxygen generation by CPO. We confirmed that the ROS‐scavenging enzyme immobilized directly on or adjacent to polycaprolactone (PCL)/CPO‐based microparticles exhibited superior H₂O₂ scavenging compared to a simple physical mixture. To translate this ‘proximity‐driven enhancement’ into a stable, long‐term platform, we have adopted a core–shell configuration with a gelatin shell. In this study, we designed zein‐CPO‐ceria microparticles (ZCCs) with a core–shell structure to enable sustained oxygen release while minimizing oxidative stress (Figure 1a). The design was guided by the chemical characteristics of each component: (i) CPO was encapsulated within a hydrophobic zein core to reduce exposure to the aqueous environment, (ii) ceria nanoparticles were dispersed within the aqueous phase to scavenge ROS through Ce³⁺/Ce⁴⁺ redox cycling, and (iii) the outer gelatin shell was enzymatically crosslinked, both internally and with the zein core, to provide structural integrity and a biofunctional interface [32]. Based on these design strategies, ZCCs were fabricated by emulsifying a zein‐CPO oil phase with a gelatin‐ceria aqueous phase. Gelatin and zein can be enzymatically coupled via microbial transglutaminase (mTG)‐mediated Gln‐Lys crosslinking, yielding uniform microparticles with a CPO‐loaded zein core and a ceria‐functionalized gelatin shell [33]. A fluorescence image of ZCC fabricated with rhodamine‐B mixed gelatin and fluorescein mixed zein showed a phase‐separated structure within each microparticle (Figure 1b). The conformal core–shell structure of the microparticles was more distinctly visualized in confocal images of ZCCs composed of rhodamine‐conjugated gelatin (Figure 1c). As designed, the zein‐containing oil phase was confined within the inner compartment, physically separated from the aqueous environment, thereby forming a stable hydrophobic core. This core–shell configuration is highly favorable for drug delivery systems, as it enables the prevention of burst release and controlled diffusion of encapsulated therapeutics inside [34].

FIGURE 1.

Design and characterization of zein‐CPO‐ceria hybrid oxygenating microparticles (ZCCs). (a) Schematic illustration of the fabrication process of ZCC microparticles with a zein‐calcium peroxide (CPO) core and ceria‐gelatin shell. (b) Fluorescence images of microparticles prepared using fluorescein‐labeled zein (core) and rhodamine B‐labeled gelatin (shell). Scale bar = 10 µm. (c) Confocal images confirming the core–shell architecture of ZCCs using rhodamine B‐conjugated gelatin. Scale bar = 15 µm. (d) Scanning electron microscopy (SEM) images of zein‐only particles (Zs), zein‐calcium peroxide (CPO) particles (ZCs), and ZCCs. Upper scale bar = 10 µm, Lower scale bar = 5 µm. (e) Particle size distribution quantified from SEM images. n = 40. (f) Particle stiffness measured by nanoindentation. n = 6. (g) Fourier‐transform infrared spectroscopy (FT–IR) spectra of Zs, ZCs, and ZCCs. Red shadow: ─OH stretch, Green shadow: ─CH₂, Blue shadow: O─O, Orange shadow: Ca─O and Ce─O stretch. (h) PrestoBlue assay result of human mesenchymal stem cells treated with different concentrations of ZCCs. n = 4.

Structurally, we sought particles with a high surface‐to‐volume ratio and nanoporosity because our platform also aims to deliver bioactive molecules. Consistent with this design, the emulsion‐derived particles exhibited a spheroidal morphology with tiny surface pores observable by scanning electron microscopy (SEM) images (Figure 1d). Furthermore, microparticles intended for cell‐external delivery are typically designed to be several micrometers in size, as particles of this scale can only be internalized through phagocytosis [35]. SEM analysis confirmed that the average diameters of zein‐CPO (ZC) and ZCC microparticles were approximately 8.1 and 9.4 µm, respectively (Figure 1e). In addition, the stiffness of particles was approximately 400 kPa, suggesting they can maintain structural integrity and enable sustained biologics delivery both in vitro and in vivo (Figure 1f) [36].

Chemically, we further analyzed the elemental and molecular composition of the particles to confirm the incorporation of CPO and ceria. Energy‐dispersive X‐ray spectroscopy (EDX) revealed that Ca accounted for approximately 10% in ZC and 7% in ZCC, confirming successful incorporation of CPO in both particles. In addition, Ce was exclusively detected in ZCC with 1%, indicating effective ceria loading (Table 1). Elemental mapping of ZCC demonstrated the localization of Ca and Ce at the particle surface, while the overall presence of oxygen and nitrogen confirmed the organic matrix derived from gelatin and zein (Figure S1). Fourier‐transform infrared (FT–IR) spectroscopy provided further insight into the chemical interactions between the organic and inorganic components (Figure 1g). Since zein is the primary constituent, spectral variations were compared with the particles formed without CPO and ceria (Z) as a reference. The enhanced broad band around 3700 cm⁻¹ in ZCC corresponds to surface hydroxyl groups on the incorporated ceria and CPO [37]. These polar hydroxyl groups likely form hydrogen or electrostatic interactions with the carboxyl and amine groups of gelatin or zein, promoting rearrangement of interfacial chains and enhanced exposure of aliphatic CH₂ groups (2800–3000 cm⁻¹) [38]. The distinct increase of the peak around 870 cm⁻¹ is attributed to the O─O stretching vibration of peroxide species in CPO, while the feature near 1040 cm⁻¹ arises from the Ca─O─O coupling band, further confirming CPO incorporation [39, 40]. In contrast, the bands below 500 cm⁻¹ are ascribed to the Ce─O stretching of ceria and Ca─O lattice vibrations, validating the presence of inorganic domains within the hybrid network [41]. Overall, these spectroscopic features highlight the rationally engineered interfacial chemistry, in which the deeply embedded CPO core remained spectroscopically silent, whereas the surface‐exposed CeO₂ domains governed the dominant spectral evolution, validating the hierarchical hybrid design.

TABLE 1.

Energy‐dispersive X‐ray spectroscopy (EDX) elemental analysis of particles.

Elements	Z (At%)	ZC (At%)	ZCC (At%)	ZCC, Day 28 (At%)
C	65.8	49	53.8	46.1
O	17.1	31.4	27.9	29.5
N	16.8	9.2	10.1	—
Ca	—	10.0	6.9	9.3
Ce	—	—	1.1	1.0

We further confirmed that metabolic activity was unaffected up to 125 µg/mL of ZCC, indicating that the microparticles are non‐toxic at these concentrations (Figure 1h). Above this threshold, reduced metabolic activity is likely attributable to mechanical stress caused by particle sedimentation, leading to local accumulation on the cell surface and restricted nutrient and oxygen diffusion through the dense particle layer [42]. However, the reduced metabolic activity observed in 2D culture needs clarification of whether this response reflected intrinsic chemical toxicity of the particles or arose from contact‐mediated effects associated with direct physical interaction and mechanical stress. To clarify this point, we evaluated metabolic activity under defined direct contact using a 3D‐culture model to better discern the cytotoxicity of ZCCs. Detailed results and interpretation are presented in the following section.

2.2. Organic‐Inorganic Hybrid Microparticle Structure Enables Long‐Term Release of Oxygen

As a delivery vehicle for oxygen, synthetic polymers such as PCL, which rely on ester‐bond‐based polymerization, undergo bulk degradation through hydrolysis [43]. In contrast, protein‐based polymers such as zein consist of amide bonds that are intrinsically resistant to hydrolysis and primarily degrade via enzymatic surface erosion (Figure 2a) [44]. The presence of the zein layer increased particle stability by suppressing rapid CPO degradation in both acidic and neutral aqueous environments, as confirmed by the significantly lower calcium release from ZCC compared to bare CPO particles under the same conditions. In addition, trypsin‐mediated enzymatic degradation increased calcium ion release, further supporting the protective role of the zein layer against CPO hydrolysis (Figure S3). By contrast, as shown in Figure 2b, PCL‐based oxygenating microparticles (PCL‐OMPs) almost degraded after 28 days of aqueous incubation, leaving particulate residues of PCL and CPO. In comparison, ZCCs maintained their original architecture, exhibiting surface erosion. This erosion‐based degradation behavior enabled ZCCs to sustain oxygen release for over 40 days, which drastically improves the release period compared to PCL‐OMPs (Figure 2c). Achieving such an exceptionally long oxygen release from free‐standing microparticles using our novel micromaterial design thus represents a remarkable advancement beyond previously reported oxygen‐generating systems [15].

FIGURE 2.

Oxygenating and reactive oxygen species (ROS) scavenging property of ZCC. (a) Schematic illustration of degradation and oxygen‐release mechanism of conventional polycaprolactone (PCL)‐based oxygenating microparticle (PCL‐OMP) and ZCC. (b) SEM images of PCL‐OMPs and ZCCs after 1 and 28 days of incubation in aqueous conditions, showing marked structural degradation in PCL‐OMPs. Red arrow indicates degraded particulates. Scale bar = 4 µm. (c) Long‐term oxygen‐release profiles of ZCC with different amounts of CPOs, PCL‐OMPs, and Zs. Released oxygen is normalized by day 0. n = 3. (d) Optical images of H₂O₂ solution and food dye containing Zs, ZCs, zein‐ceria microparticles (ZCes) and ZCCs for 24 h. Scale bar = 1 cm. (e) Analysis of residual H₂O₂ after incubation with ZCs, ZCs coated with catalase (ZC‐Cat) and ZCCs, indicating sustained H₂O₂ decrease in ZCCs. n = 3. (f) Schematic illustrations of the ROS scavenging mechanism of ZCC. (g) PrestoBlue assay result of hMSCs culture with ZC and ZCC for 24 h, showing improved cytocompatibility of ZCCs. n = 5.

Gaining control over the ROS generation from oxygen sources remains one of the major challenges in oxygenating biomaterials [16]. Most oxygen‐releasing systems rely on either the decomposition of H₂O₂ delivered via high molecular weight polymers such as PVP, or the hydrolysis of inorganic peroxides such as CPO to generate oxygen [16]. However, these reactions inevitably produce ROS, which can compromise cell viability and tissue regeneration [45]. First, we compared gas and H₂O₂ generation among particles without CPO and ceria (Z), those containing only CPO (ZC), only ceria (ZCe), and both (ZCC). As shown in Figure 2d, oxygen bubbles were observed in CPO‐containing particles (ZC and ZCC) due to spontaneous oxygen generation from hydrolysis of CPO, while ZCe also produced oxygen bubbles through catalytic decomposition of H₂O₂. Robust oxidative activity of the H₂O₂ generated from CPO was corroborated by pronounced (oxidative) color bleaching of food dye. To quantitatively assess residual ROS, we measured the remaining H₂O₂ concentration (Figure 2e). For comparison, catalase‐coated ZC (ZC‐Cat), a widely used ROS‐scavenging control, was included. Both ZC‐Cat and ZCC showed a marked reduction in H₂O₂ levels, although ZCC maintained ROS scavenging beyond day 14, whereas catalase lost activity over time due to its limited half‐life [46]. Additionally, a 2,2‐diphenyl‐1‐picrylhydrazyl (DPPH) assay further confirmed the ROS scavenging capability of ZCC, showing concentration‐dependent radical scavenging activity (Figure S2). This sustained ROS scavenging behavior can be attributed to the intrinsic catalytic properties of ceria nanoparticles, which exhibit repeatable Ce³⁺/Ce⁴⁺ redox cycling through oxygen vacancies, providing potent and long‐lasting ROS elimination [47]. Owing to their inorganic stability and self‐regenerative redox behavior, ceria‐based scavengers possess an extended functional lifetime, often lasting for several weeks or longer [48]. Collectively, these results demonstrate that our ceria‐integrated system not only efficiently mitigates H₂O₂ formation but also achieves significantly prolonged antioxidative performance compared to conventional organic scavenging strategies [28].

Excessive accumulation of ROS within the cellular microenvironment can trigger detrimental biological cascades such as DNA cleavage or ferroptosis, ultimately leading to cellular injury (Figure 2f) [49]. As shown by the PrestoBlue assay, cells cultured with ZC microparticles exhibited a marked reduction in metabolic activity even after 24 h, whereas those exposed to ZCC microparticles maintained without detectable cytotoxicity (Figure 2g). These results indicate that ceria‐mediated regulation of ROS effectively confines oxidative activity within a cytocompatible window, preventing ROS‐associated cell damage while sustaining cell survival.

2.3. ZCC Enables Long‐Term Survival of Microtissues in Anoxic Microenvironments

Stem cell spheroids have emerged as versatile microtissue building blocks in tissue engineering and as an advanced format of cell therapy, as they recapitulate native cell‐cell and cell‐matrix interactions that are often lost in conventional culture systems [50, 51, 52]. However, their dense cellular organization can lead to diffusion limitations once they are larger than ∼500 µm in diameter, causing oxygen deprivation and oxidative stress in the spheroid core. Such microenvironmental stress becomes even more pronounced after transplantation into a hypoxic in vivo environment [53]. To address these challenges, various biomaterial‐based strategies have been explored to facilitate diffusion or directly supply oxygen [14, 54, 55]. Nevertheless, most synthetic polymers or inorganic particles lack cell‐adhesive moieties, leading to their detachment or undesirable aggregation during spheroid assembly due to mismatched surface chemistry [56]. Here, we introduced a crosslinked gelatin outer layer that not only suppresses particle aggregation but also provides abundant RGD (Arg‐Gly‐Asp) peptide motifs for integrin‐mediated cell binding [57]. Simply mixing these specially designed ZCC microparticles with stem cells enabled successful incorporation of particles during the self‐assembly of spheroids (Figure 3a). The resulting spheroids exhibited a compact, spherical morphology with consistent size throughout culture, except when the amount of particles introduced exceeded 100 µg (Figure 3b; Figure S4). This increase in spheroid size with 100 µg of ZCC is likely due to the ability of the particles to mitigate cellular contraction within the spheroids. Previous studies have shown that incorporation of micro‐sized biomaterials can restrict cellular contraction, thereby mitigating excessive spheroid compaction [54]. To clarify this observation, we analyzed gene expression in the spheroids and found that the addition of microparticles reduced the expression of mechanotransduction‐related genes (yes‐associated protein 1, integrin alpha 5, vinculin, and protein tyrosine kinase 2), suggesting attenuation of cellular contraction (Figure S5). The incorporated microparticles, visualized using fluorescein isothiocyanate‐conjugated bovine serum albumin (FITC‐BSA)‐loaded ZCCs, were homogeneously distributed throughout the spheroids (Figure 3c). These results demonstrate that our gelatin‐crosslinked particle surface allows for a stable and uniform microtissue formation.

FIGURE 3.

Self‐assembly of ZCC composite spheroids and cytoprotective performance of ZCC under anoxic conditions. (a) Schematic illustration of composite spheroid fabrication. (b) Spheroid size after 14 days of culture with increasing ZCC content. n = 4. (c) Fluorescence images of spheroids with different amounts of ZCCs, showing stable self‐assembly of composite spheroids. Scale bar = 250 µm. (d) PrestoBlue assay results of composite spheroids cultured under anoxic conditions (< 0.1% O₂) with varying ZCC content, demonstrating ZCC‐dependent enhancement of cell survival in anoxia. n = 3. (e) Schematic illustration of cytoprotective mechanisms of ZCCs under anoxic microenvironment. (f) Relative DNA contents of spheroids with ZC or ZCC cultured for 14 days under anoxic conditions, indicating reduced apoptosis‐associated DNA loss with ZCCs. n = 4. Relative gene expression of (g) caspase 9 (CASP9), (h) BCL2‐associated X (BAX), (i) hypoxia‐inducible factor‐1α (HIF‐1α) and (j) carbonic anhydrase 9 (CA9) in the composite spheroids with different microparticles showing reduced apoptotic signaling and hypoxic stress in ZCC‐treated spheroids (cell‐only spheroids as control). n = 3.

Having established the structural uniformity and compactness of the ZCC spheroids, we next evaluated the cytoprotective effect of ZCC under anoxic conditions, which mimics a large non‐perfused in vivo microenvironment. Although mild hypoxia has been reported to promote stem cell proliferation and differentiation, the anoxia (< 0.1% O₂) that typically develops within large microtissues can instead compromise cell survival [58]. Notably, composite spheroids formed with ZCC exhibited markedly higher metabolic activity than cell‐only spheroids even after 4 days of anoxic culture (Figure 3d). As shown earlier, while ZC microparticles supply oxygen, they simultaneously generate excessive CPO‐derived ROS, resulting in long‐term DNA damage (Figure 3e). Consistent with this notion, DNA quantification revealed only a modest increase in total DNA due to the limited proliferative capacity of MSC spheroids. However, a significant increase in DNA content was observed on day 14 in the ZCC group, suggesting that latent oxidative damage was substantially reduced (Figure 3f). Improved cell survival in the ZCC group is consistent with the sustained oxygen release from ZCC particles (Figure 2c), which likely mitigates cumulative anoxic stress and preserves long‐term cell viability within the spheroids. Although ZC microparticles can also generate oxygen, the superior cytoprotective effect of ZCC is attributed to ceria's Ce³⁺/Ce⁴⁺ redox cycling, which dynamically scavenges ROS generated during oxygen release [37]. Apoptosis‐related genes such as caspase 9 (CASP9) and BCL2‐associated X (BAX) were significantly downregulated in ZCC spheroids, confirming reduced cell death signaling induced by oxygen depletion and ROS (Figure 3g,h). Interestingly, despite similar DNA content at day 7, both ZC and ZCC spheroids exhibited substantially higher ATP activity, indicative of a transient metabolic rescue during the early phase, likely triggered by CPO‐mediated oxygen supplementation (Figure S6). Furthermore, gene expression analysis supported this mechanism. Interestingly, the expression of hypoxia‐inducible factor‐1α (HIF‐1α) and carbonic anhydrase 9 (CA9), which is typically upregulated under hypoxic conditions, was markedly reduced (Figure 3i,j). This trend aligns with previous reports showing that oxygen‐releasing biomaterials effectively suppress hypoxia‐inducible signaling by restoring intracellular oxygen tension to near‐normoxic levels [59]. Such downregulation of HIF‐1α and its downstream targets further confirms that ZCC‐mediated oxygen release alleviated cellular hypoxia and stabilized the microenvironment toward a physiological oxygen state. Collectively, these findings demonstrate that ZCC incorporation not only mitigates anoxia‐induced cytotoxicity but also maintains redox homeostasis within dense stem cell aggregates. This dual oxygen‐supplying and ROS‐regulating capacity provides a powerful strategy to overcome one of the fundamental limitations of 3D stem cell spheroid systems, thereby enhancing their therapeutic potential in hypoxia‐prone tissue environments.

2.4. Surface Engineering of ZCC Using Functional Proteins

Beyond maintaining cellular activity under hypoxic stress, incorporating ZCCs into spheroids also confers remarkable versatility as a bioactive carrier for the localized and sustained delivery of therapeutic proteins. The gelatin‐based outer layer, enriched in glutamine residues, enables enzymatic conjugation to lysine‐containing proteins via mTG‐mediated covalent crosslinking [60]. This reaction forms a stable and highly specific protein–gelatin network without affecting the inherent biocompatibility of the system, allowing efficient immobilization of diverse biomolecules directly onto the particle surface (Figure 4a). As a proof of concept, FITC‐labeled albumin (FITC‐Albumin) was immobilized onto the gelatin shell of ZCCs, and the signal was distinctly observed on the outer gelatin shell (Figure 4b). Even in the absence of mTG, weak fluorescence signals were detected on the particle due to physical interaction between albumin and gelatin; however, the fluorescence intensity was markedly lower than that of the mTG‐treated group (Figure 4c). Quantitative fluorescence analysis revealed that the signal intensity of ZCCs prepared with mTG reached nearly that of the original FITC‐Albumin solution, whereas particles treated without mTG exhibited fluorescence comparable to bare ZCCs (Figure 4d). These results confirm that mTG catalyzes highly efficient, stable protein immobilization at the gelatin interface.

FIGURE 4.

Engineering bone‐forming microtissues via mTG‐mediated immobilization of functional proteins on ZCCs. (a) Schematic illustration of microbial transglutaminase (mTG)‐mediated protein immobilization on the ZCC using mTG. (b) Confocal microscopy 3D image of FITC‐albumin‐immobilized ZCC fabricated using rhodamine B conjugated gelatin. Scale bar = 15 µm. (c) Fluorescence images of FITC‐albumin‐immobilized ZCCs with or without mTG, demonstrating mTG‐dependent immobilization. Scale bar = 100 µm. (d) Loading efficiency of FITC‐albumin on ZCCs. The fluorescence intensity of FITC‐albumin solution was used as a positive control. n = 4. (e) Loading efficiency of BMP‐2 on ZCCs. n = 4. (f) Accumulated BMP‐2 release profile of BMP‐2 immobilized ZCCs in PBS for 14 days. n = 4. Immunofluorescence images of bone microtissues with BMP‐2‐immobilized particles (ZC‐B and ZCC‐B) stained for (g) osteopontin (OPN) and (h) type I collagen (COL1), together with their corresponding fluorescence intensity showing enhanced osteogenic matrix deposition. Scale bar = 200 µm. n = 5. (i) Alizarin Red S staining images of bone microtissues. Scale bar = 200 µm. (j) Relative gene expression of OPN and osteocalcin (OCN) in the bone microtissues indicating stronger osteogenic differentiation with ZCC‐B compared to ZC‐B and control. n = 3.

Building upon this strategy, we next incorporated BMP‐2 as a representative growth factor to induce osteogenesis [61]. Owing to its smaller molecular size (∼26 kDa) and positively charged surface compared with albumin (∼66 kDa), BMP‐2 exhibited stronger electrostatic and hydrogen‐bonding interactions with the gelatin network, leading to approximately 70% loading efficiency (Figure 4e) [62]. In contrast, FITC labeling of albumin partially masks lysine residues and alters the surface properties of the protein, which may weaken its interaction with the gelatin matrix and consequently reduce loading efficiency. Nonetheless, mTG‐mediated conjugation significantly enhanced its retention through covalent immobilization. Remarkably, BMP‐2‐immobilized ZCCs exhibited exceptional stability, releasing less than 0.1% of the total loaded protein over 14 days, thereby ensuring long‐term local presentation of bioactive cues (Figure 4f). Such strong binding interactions are particularly favorable in composite spheroids, where microparticles are in direct contact with surrounding cells, enabling sustained and localized presentation of BMP‐2 at the cell‐material interface, thereby promoting osteogenic differentiation, while minimizing the adverse effects of BMP‐2 [63]. From an engineering perspective, this represents a key advantage of the present system. Importantly, BMP‐2 induces osteogenic signaling primarily through binding to cell surface bone morphogenetic protein receptors (BMPRs), which activate downstream pathways such as SMAD phosphorylation and subsequent osteogenic gene expression [64]. Therefore, many biomaterial systems employing covalent immobilization strategies utilize this mechanism to induce osteogenesis through matrix‐bound BMP‐2 presentation [65, 66]. However, in enzyme‐rich microenvironments, partial release of BMP‐2 may also occur and potentially influence the differentiation and regenerative responses of adjacent cells. To further investigate this possibility, we performed additional experiments using collagenase type I. Collagenase treatment significantly increased BMP‐2 release, likely by promoting proteolytic erosion of the zein‐based matrix and improving BMP‐2 accessibility. Although this condition does not fully recapitulate the in vivo biological environment, these results provide insight into the potential release behavior of BMP‐2 under biologically relevant proteolytic conditions (Figure S7).

To evaluate whether the immobilized BMP‐2 retained its osteoinductive activity within the composite spheroids, we cultured the spheroids for 14 days in growth medium (LG‐DMEM) without any osteogenic supplements. Because the osteogenic role of BMP‐2 is well established, we directly examined its efficacy under anoxic conditions, where stem cell differentiation toward the osteogenic lineage is typically suppressed [67]. As expected, ZCC spheroids loaded with BMP‐2 (ZCC‐B) successfully induced osteogenic differentiation even under oxygen‐depleted conditions, showing significant upregulation of the representative bone marker osteopontin (OPN) compared with all other groups (Figure 4g). Consistently, the deposition of type I collagen (COL1), a major extracellular matrix component of bone tissue, was markedly enhanced in ZCC‐B spheroids (Figure 4h), accompanied by robust calcium deposition (Figure 4i). Although the histological analyses further confirmed partial osteogenic differentiation within ZC‐B spheroids, these aggregates exhibited pronounced structural collapse, whereas ZCC‐B spheroids maintained their architectural integrity and showed sustained upregulation of multiple osteogenic genes on day 14 (Figure 4j). These results suggest that the mineral deposition found in ZC‐B primarily represents residual signatures of early osteogenic initiation, whereas ZCC‐B supports sustained osteogenic progression and maturation beyond day 14. Furthermore, as osteogenesis increases oxygen consumption [67], the limited oxygen supplement from ZC likely intensified this metabolic burden to cells over time. Collectively, these results highlight that ZCC‐B spheroids uniquely integrate oxygen regulation and controlled BMP‐2 presentation, enabling robust bone‐like tissue formation even under anoxic conditions. This dual functionality underscores the potential of ZCC‐based microtissues as a universal, self‐oxygenating, and instructive platform for regenerative applications in hypoxia‐prone environments.

2.5. Enhanced In Vivo Bone Tissue Regeneration Using ZCC‐Laden Spheroids

Building upon the potent osteogenic potential of ZCC‐B spheroids observed under anoxic conditions in vitro, we next evaluated their regenerative efficacy in vivo. While most previous studies have explored oxygen regulation in large bone defects with trabecular bone, such models are relatively well‐vascularized and do not accurately recapitulate the oxygen‐limited conditions [68, 69, 70]. In contrast, the calvarial defect represents a far more stringent and hypoxic environment characterized by limited vascular ingrowth and poor oxygen diffusion. Indeed, the oxygen level in mouse calvarial bone has been reported to fall to as low as 0.6% O₂ in certain regions, highlighting the severity of hypoxia in this model [71]. Thus, we selected a critical‐sized calvarial defect as an ideal platform to assess the therapeutic performance of the synergistic effects of ceria‐mediated oxygen‐ and osteoinductive factor‐regulating ZCC‐B‐laden MSC spheroids under harsh, diffusion‐limited conditions (Figure 5a). To further dissect the contribution of ceria‐mediated effects, spheroids containing ceria but lacking the oxygen‐generating CPO core (ZCe‐B) were used as a control group to evaluate the intrinsic biological effects of ceria‐based materials, including their antioxidant and osteogenic activities, independent of oxygen generation in the presence of BMP‐2 [72]. 20 spheroids were implanted into a mouse calvarial defect model with a 4‐mm diameter circular defect and covered with bioresorbable surgical sheets.

FIGURE 5.

In vivo bone tissue regeneration promoted by engineered bone microtissues. (a) Schematic of the implantation procedure of engineered bone microtissues in a mouse calvarial defect model. (b) Representative micro‐computed tomography (µCT) images of calvarial defects treated with different microtissues at the study endpoint. Scale bar = 2 mm. (c) Quantification of bone volume/total volume (BV/TV) and (d) new bone area from µCT analysis, showing enhanced bone formation in the ZCC‐based microtissue group. n = 3. (e) H&E and (f) Goldner's trichrome staining of the defect region, demonstrating more mature bone matrix deposition with ZCC‐based microtissues. Scale bar = 500 µm (scale bar for magnified images = 250 µm). (g) Immunofluorescence staining of implanted sites for OCN, osteopontin (OPN), and HIF‐1α, showing increased osteogenic marker expression and HIF‐1α with ZCC‐B‐containing microtissues. Scale bar = 500 µm.

Micro‐computed tomography (µCT) analysis revealed that only the ZCC‐B group exhibited significant bone tissue regeneration after 8 weeks of implantation (Figure 5b). Quantitatively, ZCC‐B achieved approximately 15% bone volume fraction (BV/TV), indicating a markedly greater extent of mineralized tissue formation compared with all other groups. Consistently, the regenerated area reached nearly 50% of the defects, showing a statistically significant improvement (Figure 5c,d). Interestingly, unlike most cell‐therapy‐based bone regeneration that proceeds through the formation of isolated pre‐bone islands, the ZCC‐B group exhibited a continuous bone tissue ingrowth from the defect edges toward the center. Trabecular bone factor analysis further confirmed a more homogeneous and less segregated architecture in ZCC‐B (Figure S8), suggesting a more mature and structurally integrated regeneration pattern. In calvarial defects, new bone formation initiates as woven bone throughout the defect. However, bone extending centripetally from the defect edge typically undergoes more advanced remodeling toward lamellar or cortical bone, whereas island‐like bone formed within the defect center often remains less remodeled [73]. Histological evaluation further supported these findings. Hematoxylin and eosin (H&E) staining revealed dense and well‐organized bone matrix in ZCC‐B, whereas other groups displayed sparse or immature tissue formation (Figure 5e). Goldner's trichrome staining further demonstrated an intense green signal corresponding to dense collagen deposition, indicative of highly mature bone tissue (Figure 5f). Although potential tissue damage could theoretically arise from byproducts generated during ZCC degradation (CPO decomposition products or ceria nanoparticles), the in vivo results demonstrated normal tissue regeneration without abnormal tissue morphology or excessive inflammatory responses.

To further assess biological maturation, immunofluorescence staining for osteogenic markers, including osteocalcin (OCN), OPN, and runt‐related transcription factor 2 (RUNX2), was performed together with HIF‐1α. The ZCC‐B group exhibited abundant OCN depositions at the regenerating front, along with strong expression of OPN and RUNX2, implying sustained and active osteogenesis (Figure 5g; Figure S9). Interestingly, HIF‐1α was also distinctly observed in the ZCC‐B group, suggesting that under the complex and deteriorated in vivo microenvironment, a mild hypoxic stimulus persisted and acted synergistically to promote bone formation through a rescue‐type response (Figure 5g; Figure S10) [74]. Consistently, the additional immunofluorescence staining results for CD31 and VE‐cadherin (CDH5) further demonstrated enhanced vascular marker expression in the ZCC‐B group, indicating that HIF‐1α‐mediated signaling may promote angiogenic responses and thereby facilitate bone tissue regeneration (Figure S11).

Collectively, these results demonstrate that ZCC‐B spheroids orchestrate a unique regenerative microenvironment that couples oxygen regulation with localized osteoinductive signaling, thereby enabling robust and spatially integrated bone regeneration even under severe hypoxic conditions in vivo.

Compared to recent oxygen‐delivering biomaterial systems that rely on permanent scaffolds or bulk matrices such as hydrogels, the proposed ZCC‐B platform offers distinct advantages as a scaffold‐free strategy [15]. Although minimizing the amount of exogenous biomaterial can impose functional limitations, including challenges in ROS regulation, sustained oxygen release, and biological molecule delivery, these constraints were effectively addressed through the ZCC‐based system. While the scaffold‐free design inherently limits initial mechanical strength, this limitation may be mitigated through hybrid integration with complementary scaffold‐based or transient support frameworks, without compromising the core functional advantages of the platform [75]. In addition to osteogenic growth factor delivery demonstrated in this study, the modular design of the ZCC‐B system enables the incorporation of diverse bioactive molecules, allowing application to any organs or tissues that experience delayed vascularization. Collectively, these features position the ZCC‐B system as a versatile and scalable platform for tissue regeneration in poorly vascularized or avascular environments where conventional vascular‐dependent strategies remain insufficient.

3. Conclusion

In this study, we developed a hierarchically engineered oxygen‐ROS and stem cell fate‐regulating microparticle that achieves long‐term and balanced oxygen release through a CPO‐zein core while minimizing oxidative stress via a gelatin‐ceria bioactive shell. This hybrid architecture enables prolonged oxygen generation over 40 days, during which ceria‐mediated redox cycling efficiently scavenges the ROS produced from the oxygen source. As a result, the ZCC system establishes a finely tuned oxygen microenvironment that is both sustainable and cytocompatible. When incorporated into hMSC spheroids, ZCC particles achieved structural integrity and preserved viability even under anoxic stress by sustaining oxygen supply while mitigating ROS accumulation. Moreover, the gelatin interface endowed the system with modular biofunctionality, enabling enzymatic conjugation of osteoinductive proteins, such as BMP‐2, to drive spontaneous bone‐like microtissue formation. In a critically hypoxic calvarial defect model, ZCC‐B facilitated robust and mature bone tissue regeneration. Histological and molecular analyses further revealed mature collagen‐rich tissue and strong expression of osteogenic and hypoxia‐responsive markers, indicating an adaptive but well‐oxygenated regenerative niche. Collectively, this work presents a rationally designed, self‐oxygenating, and biofunctional microtissue platform capable of overcoming oxygen‐diffusion limitations in harsh microenvironments. The integration of oxygen regulation with tunable biochemical functionality provides a versatile framework for engineering metabolically favorable tissues, offering broad potential across regenerative and transplant medicine.

4. Experimental Section

4.1. Fabrication of Microparticles

To fabricate ZCC, porcine gelatin type A (Sigma–Aldrich, MO, USA) was first dissolved in distilled water (8% w/v) at 37°C and filtered through a 0.45 µm pore syringe filter. For fluorescent labeling, rhodamine‐gelatin (Sigma–Aldrich, MO, USA) was mixed with gelatin at a 1:6 ratio. Zein (Sigma–Aldrich, MO, USA) was dissolved in 90% ethanol (5% w/v) at 70°C and filtered through a 0.22 µm pore syringe filter. Then, CPO (75%, ∼200 mesh; Sigma–Aldrich, MO, USA) was dispersed in the zein solution under sonication (30% amplitude, 5 min; Branson, CT, USA). Microbial transglutaminase (mTG; Activa TI; Ajinomoto, Tokyo, Japan) was dissolved in distilled water (0.8% w/v) and filtered (0.22 µm). Then, cerium (IV) oxide nanoparticles (ceria, < 25 nm; Sigma–Aldrich, MO, USA) were added to the solution and dispersed by sonication (30% amplitude, 5 min). Equal volumes of zein solution were added to the gelatin solutions slowly, followed by mixing with the mTG‐ceria dispersion. The mixture was gently inverted and crosslinked on a shaker (80 rpm, 30 min). After incubation, aggregates were removed by a 40 µm‐pore strainer, and the filtrate was centrifuged (500 rpm, 5 min). Pellets were washed three times with distilled water and frozen at −80°C. The particles were lyophilized for 3 days to obtain dry microparticles. Control particles lacking specific components were fabricated using the same protocol, with ceria omitted (ZC), CPO omitted (ZCe), or both omitted (Z). To prepare ZC‐Cat, catalase from bovine liver (Sigma–Aldrich, MO, USA) was dissolved in DMEM at a concentration of 1 mg/mL. 2 mg of ZC particles were then incubated with 1 mL of the catalase solution at 4°C for 1 h under gentle shaking. After incubation, the suspension was centrifuged at 3750 rpm for 15 min, then washed 4 times with PBS to remove unbound catalase.

PCL‐OMPs were fabricated via the double emulsion method as previously described [76]. Briefly, a polymer phase was prepared by dissolving PCL (10% w/v; Sigma–Aldrich, MO, USA) in dichloromethane (Sigma–Aldrich, MO, USA), while the inner aqueous phase contained CPO (3% w/v in distilled water). Ultrasonic homogenization (2 min; QSonica, CT, USA) yielded the primary W/O emulsion. This was immediately dispersed into a secondary aqueous solution of gelatin (1% w/v, kept on ice), where an additional 4 min of sonication promoted stable droplet formation. Dichloromethane was removed under gentle stirring at room temperature overnight, resulting in solidified particles. The product was isolated by centrifugation (3750 rpm, 15 min) and repeatedly washed with water. The frozen products were finally lyophilized for 5 days to obtain dry PCL‐OMP powder.

4.2. Characterization of Microparticles

4.2.1. SEM Analysis

Before analysis, microparticles were cast onto a small silicon substrate and allowed to air‐dry. Then, SEM was performed on a JSM 7610 F‐Plus microscope (JEOL) at an acceleration voltage of 1.5 kV and a working distance of 8 mm. For elemental composition analysis. EDX was performed using an EDX detector (Oxford Instruments Aztec Live Standard Ultim 40 mm²) equipped on the electron microscope, at an acceleration voltage of 8 kV.

4.2.2. Stiffness Analysis

Nanoindentation measurements were performed in PBS using a cantilever (spring constant 0.26 N/m) with a glass colloidal probe (radius = 3 µm) attached to the tip. The probe was brought into close contact with the surface of the particle, and indentation was performed using the following indentation procedure: an indentation of 100–1000 nm (depending on the size of the particle) for 2 s, followed by a 1 s holding and 2 s retraction time. To minimize the movement of particles during nanoindentation, they were seeded into a MaxiSorp 96‐well plate. The obtained indentation curves were fitted between 0 and 500 nm indentation using the Hertzian model, from which the effective elastic modulus was obtained, according to P = 4/3 E_effR^1/2h^2/3, where P is the applied load, E_eff is the effective Young's Modulus, R is the radius of the indentation tip, and h is the indentation depth. The elastic modulus (herein also referred to as stiffness) was calculated by E = E_eff(1−ν²), assuming a Poisson's ratio of ν = 0.5.

4.2.3. FT–IR Analysis

FT–IR spectroscopy was recorded on S‐21, a Bruker Vertex 70 spectrometer in absorbance mode with particle samples using a diamond attenuated total reflectance (ATR) accessory.

4.2.4. Morphological Analysis

A hierarchical structure was confirmed in ZCC particles fabricated from fluorescein‐dissolved zein solution and rhodamine B‐dissolved gelatin solution using fluorescence microscopy (Nikon Eclipse Ti‐5; Nikon, Tokyo, Japan). To further visualize the core–shell architecture, particles prepared with rhodamine‐conjugated gelatin were examined by confocal microscopy (Leica Stellaris X5; Wetzlar, Germany).

4.2.5. Oxygen Release

Oxygen release from microparticles was assessed by dispersing the particles in PBS (3 mg/mL) and incubating the suspensions in an anoxic chamber (ProOx 21 with C‐Chamber; BioSpherix, NY, USA). To establish oxygen‐depleted conditions, PBS was purged with nitrogen inside the chamber for 30 min. Oxygen concentration in the particle suspensions was then monitored at predetermined time intervals using a fiber‐optic oxygen microsensor (Opto‐F1 UniAmp; Unisense, Aarhus, Denmark) at room temperature. Signals were recorded for approximately 30 s until a stable plateau was reached. All measurements were carried out in triplicate.

4.2.6. Calcium Release Analysis

To evaluate CPO dissociation, released calcium ions were quantified using a calcium assay reagent (Pointe Scientific Inc., MI, USA). CPO and ZCC particles were dispersed in Ca²⁺/Mg²⁺‐free DPBS adjusted to different pH conditions (pH 4.0, 7.4, and 12.0) and incubated for 12 h at room temperature. For the experiment with trypsin, the particles were incubated for 24 h at 37°C. After incubation, the samples were centrifuged (16 000 rpm, 10 min), and the supernatant was collected. Calcium ion concentrations in the supernatant were quantified using the calcium assay according to the manufacturer's instructions. The measured calcium concentration was used to estimate the extent of CPO dissociation, normalized to the initial calcium content of the particles.

4.2.7. Hydrogen Peroxide Analysis and DPPH Assay

Residual hydrogen peroxide was assessed by incubating the particles in 30% H₂O₂ solution (Supelco, MO, USA) overnight in 1.5 mL microtubes. Bubble formation was qualitatively observed, and the supernatant was collected for analysis. Hydrogen peroxide concentration was determined using a fluorometric hydrogen peroxide assay kit (Sigma–Aldrich, MO, USA) according to the manufacturer's protocol. In addition, the dye bleaching capacity of the particles was evaluated by incubating them with blue food dye overnight and monitoring the extent of color loss.

A DPPH assay (Cayman Chemical, MI, USA) was performed using different concentrations of ZCCs. Briefly, serial dilutions of ZCCs (starting from 10 mg/mL) were mixed with 0.6 mm DPPH solution and incubated for 30 min in the dark at room temperature. The absorbance was then measured at 517 nm using a plate reader.

4.3. Cell Culture

hMSCs (Lonza Bioscience, MD, USA) were cultured in low‐glucose Dulbecco's Modified Eagle Medium (L‐DMEM) supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin (Thermo Fisher Scientific, MA, USA). Cultures were maintained at 37°C with media renewal every 2 days, and cells were expanded until reaching 70%–80% confluency. For passaging, cells were detached using 0.025% trypsin/EDTA in PBS for 3 min at 37°C. Cells at passages 4–6 were used in subsequent experiments.

4.4. Cellular Activity Analysis

Cellular activity was analyzed using the PrestoBlue assay, 3D Cell Viability Assay, and DNA assay. For the PrestoBlue assay (Thermo Fisher Scientific, MA, USA), the reagent was diluted 1:9 with culture medium and incubated with cells at 37°C for 1.5 h. Following incubation, the supernatant was collected, and the fluorescence was measured at 560 nm (excitation) and 590 nm (emission) using a microplate reader (SpectraMax i3; Molecular Devices, CA, USA). To further assess the ATP activity of spheroids, the CellTiter‐Glo 3D Cell Viability Assay kit (Promega, WI, USA) was employed according to the manufacturer's protocol. Briefly, an equal volume of reagent was added to the culture medium and shaken for 5 min. Then, the solution was incubated at room temperature for 25 min before luminescence was measured using a microplate reader. For DNA content evaluation, spheroids were lysed by vigorous pipetting in RIPA buffer (Thermo Fisher Scientific, MA, USA), and total DNA was quantified using the Quant‐iT PicoGreen dsDNA Assay Kit (Thermo Fisher Scientific, MA, USA). Fluorescence intensity was then measured at 480 nm excitation and 520 nm emission with a microplate reader.

4.5. Fabrication and Characterization of Composite Spheroids

4.5.1. Spheroid Fabrication

For spheroid formation, round‐bottom 96‐well plates were coated with a thin agarose layer to prevent cell attachment. A 1.5% agarose solution was prepared and sterilized by autoclaving, and 200 µL was dispensed into each well. The excess solution was immediately aspirated to leave a thin coating. Plates were then placed on ice for 5 min to solidify. hMSCs were detached using 0.025% trypsin/EDTA at 37°C for 3 min and seeded at 4 × 10⁴ cells per well. For composite spheroids, microparticles were sterilized in 70% ethanol, followed by 1 h of UV irradiation. After three washes with PBS, the particles were resuspended in culture medium and added together with the cells at the time of seeding. Unless otherwise specified, 40 µg of microparticles were used per spheroid. The spheroids were examined by phase‐contrast microscopy on days 1, 4, 7, and 14 after formation. The projected area was measured using Photoshop software (Adobe, CA, USA). For anoxic spheroid culture, the spheroids were incubated under anoxic conditions (O₂ < 0.1%).

4.5.2. Reverse Transcription Quantitative Polymerase Chain Reaction (RT‐qPCR)

Total RNA was isolated from spheroids using TRIzol reagent (Invitrogen, CA, USA) following a freeze–thaw cycle. The lysates were mixed with chloroform, centrifuged at 13 000 rpm for 15 min at 4°C, and the aqueous phase was collected. RNA was precipitated by adding an equal volume of isopropanol, and pellets were collected after centrifugation. The pellet was washed once with 75% ethanol, air‐dried, and dissolved in nuclease‐free water. RNA concentration and purity were assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, MA, USA). cDNA was synthesized from total RNA with the iScript cDNA Synthesis Kit (Bio‐Rad, CA, USA) according to the manufacturer's protocol. Quantitative real‐time PCR was performed using SYBR Green Universal Master Mix (Applied Biosystems, CA, USA) on an iQ5 real‐time PCR system (Bio‐Rad, CA, USA). The amplification program consisted of denaturation at 95°C for 10 s and annealing at 60°C for 60 s. Relative gene expression levels were calculated using the ddCT method. Gene expression levels were normalized to GAPDH, and all reactions were carried out in triplicate. Primer sequences are listed in Table S1.

4.6. Immunostaining Staining

For spheroid staining, spheroids were fixed with 4% paraformaldehyde (PFA) for 1 h and washed three times with PBS. The fixed spheroids were embedded in optimal cutting temperature (OCT) compound with freezing and sectioned at a thickness of 10 µm. The sections were washed three times with PBS to remove residual OCT. Samples were then incubated with blocking buffer (1% bovine serum albumin; Sigma–Aldrich, MO, USA) for 1 h at room temperature. Primary antibodies, including mouse anti‐human OPN (Thermo Fisher Scientific, MA, USA) and mouse anti‐human COL1 (Thermo Fisher Scientific, MA, USA), were diluted 1:100 in blocking buffer and incubated with the samples overnight at 4°C. After washing three times with PBS, Alexa Fluor 488 goat anti‐mouse IgG (H+L) secondary antibody (Invitrogen, CA, USA) was applied at a 1:1000 dilution and incubated for 1 h at 37°C. Following three PBS washes, all samples were counterstained with DAPI (1:1000 dilution in PBS) for 15 min at room temperature. Fluorescence images were acquired using a fluorescence microscope, and signal intensity was quantified using Photoshop software (Adobe Systems, CA, USA).

4.7. Protein Loading on Microparticles

To immobilize proteins on the microparticles, 40 µg of particles were sterilized in 70% ethanol and washed three times with PBS by centrifugation at 15 000 rpm for 10 min. The particles were then dispersed in a 0.8 wt.% mTG solution containing 100 ng/mL of human recombinant BMP‐2 (PeproTech, NJ, USA) and incubated for 24 h at 4°C. After incubation, the particles were washed twice with PBS. For FITC‐albumin immobilization, 1 mg of particles was mixed with 10 µg/mL of FITC‐albumin and incubated for 1 h at room temperature, following the same immobilization protocol described above. The release of BMP‐2 was evaluated by collecting the supernatant (PBS) after centrifugation at 15 000 rpm for 10 min. At each time point, the collected supernatant was stored at −80°C until analysis, and the BMP‐2 concentration was quantified using a Human BMP‐2 DuoSet ELISA kit (R&D Systems, MN, USA) according to the manufacturer's instructions. The amount of BMP‐2 released at each time point was accumulated to calculate the cumulative release, and the percentage was normalized to the initially loaded BMP‐2 amount. To evaluate the BMP‐2 release profile under proteolytic conditions, samples were incubated with 0.5 U/mL collagenase type I in the presence of 0.1% BSA and 2 mm CaCl₂ to maintain enzymatic activity. FITC‐albumin‐immobilized particles were visualized using confocal and fluorescence microscopy.

4.8. Osteogenic Differentiation of Spheroids and Alizarin Red S Staining

BMP‐2‐loaded ZC and ZCC were used to induce osteogenic differentiation of MSC spheroids after particle incorporation. Spheroids without particles were used as controls. After spheroid formation as described above, the spheroids were cultured under anoxic conditions (O₂ < 0.1%) for 14 days in L‐DMEM. For Alizarin Red S staining, cryosectioned spheroids were prepared as described above. The slides were equilibrated to room temperature for 15 min and washed with PBS to remove residual OCT. The sections were then stained with 2% (w/v) Alizarin Red S solution (Sigma–Aldrich, MO, USA) in deionized water, adjusted to a pH of 4.2. After staining for 90 s at room temperature, the slides were rinsed thoroughly with deionized water and imaged using an optical microscope.

4.9. Animal Experiments

4.9.1. In Vivo Calvarial Defect Model

Spheroids containing 40 µg of ZCe‐B or ZCC‐B particles, as well as particle‐free spheroids, were prepared before implantation. Calvarial bone defects were created in 6‐week‐old female ICR mice anesthetized by intraperitoneal injection of Zoletil (60 mg/kg) and Rompun (20 mg/kg). Mice were housed under standard laboratory conditions with free access to food and water. Mice were randomly assigned to four groups (Defect, Spheroids, ZCe‐B spheroids, and ZCC‐B spheroids; n = 3 per group). All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Hanyang University (2024‐0042A). A circular defect with a diameter of 4 mm was generated on the mouse calvaria using a surgical trephine drill under sterile conditions. The prepared spheroids were gently positioned within the defects and covered with a 0.02 mm thickness of biodegradable surgical film (SurgiWrap, MAST Biosurgery; CA, USA). After 8 weeks, the mice were euthanized by CO₂ asphyxiation, and the calvarial specimens were harvested. Outcome assessment and data analysis were performed in a blinded manner. No animals were excluded from the analysis.

4.9.2. µCT Analysis

The tissues were fixed in 10% neutral‐buffered formalin for 3 days and subjected to µCT analysis (Skyscan 1176, Bruker, MA, USA). 3D reconstructions were generated using the accompanying 3D viewer software. The bone volume to total volume (BV/TV) ratio and trabecular bone pattern factor (Tb. Pf) were quantified using CTAn and CTVol software (Bruker, MA, USA). The regenerated bone area was further quantified using Photoshop software (Adobe Systems, CA, USA).

4.9.3. Histological Analysis

The fixed samples were embedded in paraffin after decalcification in RapidCal solution for 7 days and sectioned at a thickness of 5 µm. The sections were deparaffinized and stained with hematoxylin for 10 min, followed by eosin Y for 8 min. Goldner's trichrome staining was performed according to the manufacturer's protocol with sequential staining in hematoxylin (10 min), ponceau (30 sec), orange G (5 min), and light green (6 min). Immunofluorescence staining was conducted on deparaffinized sections using antibodies against RUNX2, OPN, and OCN following the same procedure described above. For HIF‐1α, CD31, and CDH5 staining, an antigen retrieval process was additionally performed prior to staining. Briefly, deparaffinized samples were incubated in a pressure cooker for 20 min and then cooled at room temperature for 30 min. The remaining staining procedures were subsequently performed using the following primary antibodies: HIF‐1α (GeneTex, CA, USA), CD31 (Invitrogen, MA, USA), and CDH5 (LSBio, WA, USA). The stained tissues were imaged using optical and fluorescence microscopy.

4.10. Statistical Analysis

All quantitative data are presented as mean ± standard deviation. Statistical significance was assessed using the unpaired Student's t‐test and one‐way ANOVA, after verifying normality of the datasets with Shapiro–Wilk testing, followed by Tukey's HSD post hoc analysis. All statistical analyses were performed using GraphPad Prism 7 (La Jolla, CA, USA), and differences with p < 0.05 were considered significant.

Ethics Approval Statement

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Hanyang University (2024‐0042A).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adma73849‐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.