Modular living assembly of bone organoids with in situ guided vascularization Authors: Jingyi Ju, Huimin Fang, Junjin Jie, Yuezhou Wu, Jian Wang, Qianmin Gao, Yifan Zhang, Chao Luo, Houfeng Zheng, Jiaming Sun, Jiacan Su, Zhenxing Wang Published: 2026 ABSTRACT Organoids have emerged as a powerful strategy to recapitulate tissue architecture and function, opening new avenues for regenerative medicine. However, clinical translation is hindered by limited control over the assembly of heterogeneous organoid units and the resulting loss of spatial organization at the tissue scale. In addition, macroscale organoid constructs face diffusion-limited oxygen and nutrient transport, leading to central necrosis. The poor vascular integration can hinder the further development and maturation of the organoids, ultimately preventing functional integration. To address these limitations, we developed Neo-Organoid Visualization and Assembly (NOVA), a living assembly strategy that integrates bioprinting with hydrogel-based bio-adhesive assembly to enable high-throughput modular assembly of bone organoids. Using this strategy, we generate a tissue-scale tubular bone organoid graft formed with in situ unidirectional guided vascularization. This structure enables the spatiotemporal coupling of angiogenesis and osteogenesis, enhances the maturation and function of bone organoids, and holds potential for one-step surgical transplantation with rapid perfusion to repair defects. Introduction With the accelerating global aging of the population, the incidence of severe tissue and organ defects is steadily increasing [1]. Allografts remain the clinical gold standard [2], while their utility is fundamentally constrained by donor morbidity, immune rejection, and graft failure [3]. These challenges underscore the urgent need for effective alternatives [4,5]. Among them, organoids have attracted considerable interest because they partially recapitulate key structural and biological features of native tissues [6,7]. However, their translational use is still limited by poor architectural control and marked heterogeneity in morphology, size, and cellular composition, which reduce reproducibility and functional reliability [[8], [9], [10]]. To address these issues, organoid assembly strategies have advanced from spontaneous fusion to guided assembloid and formation using scaffold-assisted patterning, microfluidic confinement, acoustic or magnetic manipulation, and bioprinting-based positioning [[11], [12], [13], [14]]. These methods have improved spatial organization at the tissue level, but geometric control alone cannot address the challenges of effective biological integration. A core challenge that remains unresolved is the construction of a cell-compatible interface that can maintain structural cohesion while allowing mechanical communication, molecular transport, and subsequent host integration. Another major challenge in organoid transplantation is the establishment of rapid and functional vascular integration after implantation [15]. In the absence of perfusable vasculature, engineered constructs larger than approximately 400 μm are prone to central hypoxia and necrosis [[16], [17], [18], [19]]. Transplanted organoids often fail to connect efficiently with the host vasculature, leading to ischemic injury, fibrosis, and limited functional recovery compared with their in vitro performance [20]. This indicates that successful vascularization is essential for graft survival and maturation in vivo. Existing strategies, including pro-angiogenic factor delivery, endothelial cell co-culture, and vascular scaffold incorporation, can partially enhance vascular formation within organoid constructs [20,21]. However, they remain insufficient to ensure immediate perfusion, efficient mass transport, and stable integration with host tissues after transplantation [4,18]. Thus, enabling rapid and durable vascular connection in situ remains a key requirement for the therapeutic translation of organoid-based grafts. Catechol-functionalized gelatin hydrogels have been widely reported for use in wound dressings, tissue closure, and hemostasis, where the chemical properties of catechol groups primarily enhance wet adhesion and overall material properties [22,23]. By contrast, their potential as interfacial chemistries for modular tissue assembly has been rarely investigated. In our previous study, we demonstrated that two catechol-rich opposing interfaces can form stable interfacial crosslinks between microcage modules [24]. Building on this finding, we developed a printable catechol-functionalized low-temperature gelatin-dopamine-methacrylamide (GDMA) bioink optimized to balance the printability of organoid living modules with the adhesive stability required for bottom-up assembly, enabling it to also serve as a low-temperature bioink matrix. Beyond enabling programmable geometric organization, NOVA incorporates an in situ vascularization strategy that directs the unidirectional ingrowth of host microvessels, thereby accelerating host vascular integration, promoting coordinated bone-vascular development, and overcoming key limitations of existing vascularization approaches. This strategy enabled the rapid construction of large, shape-customizable tubular bone grafts and provided insight into the in vivo developmental dynamics of engineered bone organoids, including signaling pathways related to developmental progression, osteogenic specification, and bone-vascular niche remodeling [21,25]. Results Neo-Organoid Visualization and Assembly (NOVA) To address these persistent limitations, we developed the NOVA platform, which integrates four synergistic innovations: Neo-biomaterials: We designed a dopamine-functionalized GDMA hydrogel as a bioactive scaffold material. The DOPA-enriched interface enhances cell adhesion, promotes M2 macrophage polarization, and supports both vascular ingrowth and osteogenic differentiation, creating a pro-regenerative immune microenvironment. Organoid: We utilized digital light processing (DLP) bioprinting technology to fabricate high-throughput bone organoid modules using GDMA as the scaffold matrix, preserving both structural fidelity and cellular viability. The modules support long-term in vitro osteogenic induction and allow precise regulation of differentiation and bone matrix deposition. Vascularization: To address the issue of inadequate vascularization, we constructed a perfusable hierarchical vascular network by assembling customizable tubular GDMA modules. These luminal structures are coated with endothelial cells in vitro and rapidly anastomose with the host vascular system in vivo after subcutaneous implantation. Assembly: Modular bone organoids were assembled through catechol-mediated interfacial adhesion and oxidative cross-linking technology, enabling scalable and spatially organized integration of heterogeneous tissue units in a flexible, bottom-up manner (see Fig. 1). Synthesis of Gel-Dopa-MA The amine group of dopamine hydrochloride was conjugated to the carboxyl group of gelatin through amide formation to yield gelatin-dopamine (GD) (Fig. 2A). Subsequently, GD was mixed with 2-aminoethyl methacrylate hydrochloride (AEMA) and dissolved in 2-(N-morpholino) ethanesulfonic acid monohydrate (MES) buffer to obtain hydrogel with dual functional groups, which was lyophilized to yield GDMA hydrogel (Fig. 2B). Successful functionalization was confirmed by Fourier transform infrared (FTIR) spectroscopy and proton nuclear magnetic resonance (1H NMR) analysis, which verified the introduction of both catechol and methacrylate moieties (Fig. 2C and D). Moreover, FTIR spectra confirmed the characteristic chemical features of GDMA (Fig. S1A). 1H NMR spectra remained highly consistent across GDMA samples prepared using dialysis membranes with different molecular weight cut-offs (Fig. S1B) and across independently synthesized batches (Fig. S1C), indicating reproducible synthesis and robust batch-to-batch structural fidelity. The GDMA hydrogel exhibited a static water contact angle of 35.2 ± 1.4° (Fig. 2E), indicating a highly hydrophilic surface. This level of wettability is generally regarded as beneficial for supporting cell adhesion [26]. The GDMA hydrogel can be crosslinked using three methods: UV photo-crosslinking, oxidative crosslinking with sodium periodate, and ionic crosslinking with ferric iron solution [27]. Upon EDTA treatment, only the hydrogel structure formed by ionic crosslinking showed destroyed structure (Fig. 2F). X-ray diffraction (XRD) patterns demonstrate the homogeneity of the internal structure of the hydrogel (Fig. 2G). Small-angle X-ray scattering (SAXS) analyses demonstrated that the hydrogels formed through distinct crosslinking routes exhibit a uniform network architecture at the nanoscale. Such structural homogeneity is strongly correlated with the improved mechanical robustness of the materials (Fig. 2J).Fig. 1Schematic diagram of the study design and experimental processes. A) Synthesis of hydrogels. B) Modular printing, assembly, and vascularization of bone organoids. C) The process of module assembly and vascularization. D) Systematic workflow for the modular assembly of bone organoids via the NOVA strategy. The arrow indicates a potential application route for in situ long-bone defect repair through modular bone organoid assembly, vascularization, angiogenesis, ectopic osteogenesis, and subsequent autologous, allogeneic, or xenogeneic transplantation.Fig. 1Fig. 2Synthesis and evaluation of GDMA bioinks. A) Schematic illustration of the chemical synthesis of GDMA. B) Photograph of lyophilized GDMA after vacuum freeze-drying. C) 1H NMR spectra of the reactants and final GDMA product. D) FTIR spectra of gelatin (Gel), gelatin-dopamine conjugate (Gel-Dopa), and GDMA indicating successful methacrylation. E) Water contact angle measurements of GDMA hydrogel surface. F) Photographs showing GDMA hydrogels crosslinked via 365 nm UV irradiation, ferric chloride (FeCl3), and sodium periodate (top), followed by immersion in saturated EDTA solution for 24 h (bottom). G) XRD patterns of GDMA hydrogels after photocrosslinking, oxidative crosslinking, and ionic crosslinking. H) Compressive stress–strain curves of GDMA hydrogels. I) Remaining mass of GDMA hydrogel incubated in PBS over 1, 7, and 14 days (n = 4). J) SAXS analysis showing changes in –OH activity and internal molecular organization after crosslinking. K) Swelling ratios of GDMA hydrogels with different molecular weights and concentrations in 0.9% NaCl solution (n = 4). L) Swelling ratios of GDMA hydrogels in urea solution (n = 4). M) X-ray photoelectron spectroscopy (XPS) O1s core-level spectra, indicating catechol group participation in crosslinking reactions.Fig. 2 Schematic diagram of the study design and experimental processes. A) Synthesis of hydrogels. B) Modular printing, assembly, and vascularization of bone organoids. C) The process of module assembly and vascularization. D) Systematic workflow for the modular assembly of bone organoids via the NOVA strategy. The arrow indicates a potential application route for in situ long-bone defect repair through modular bone organoid assembly, vascularization, angiogenesis, ectopic osteogenesis, and subsequent autologous, allogeneic, or xenogeneic transplantation. Synthesis and evaluation of GDMA bioinks. A) Schematic illustration of the chemical synthesis of GDMA. B) Photograph of lyophilized GDMA after vacuum freeze-drying. C) 1H NMR spectra of the reactants and final GDMA product. D) FTIR spectra of gelatin (Gel), gelatin-dopamine conjugate (Gel-Dopa), and GDMA indicating successful methacrylation. E) Water contact angle measurements of GDMA hydrogel surface. F) Photographs showing GDMA hydrogels crosslinked via 365 nm UV irradiation, ferric chloride (FeCl3), and sodium periodate (top), followed by immersion in saturated EDTA solution for 24 h (bottom). G) XRD patterns of GDMA hydrogels after photocrosslinking, oxidative crosslinking, and ionic crosslinking. H) Compressive stress–strain curves of GDMA hydrogels. I) Remaining mass of GDMA hydrogel incubated in PBS over 1, 7, and 14 days (n = 4). J) SAXS analysis showing changes in –OH activity and internal molecular organization after crosslinking. K) Swelling ratios of GDMA hydrogels with different molecular weights and concentrations in 0.9% NaCl solution (n = 4). L) Swelling ratios of GDMA hydrogels in urea solution (n = 4). M) X-ray photoelectron spectroscopy (XPS) O1s core-level spectra, indicating catechol group participation in crosslinking reactions. The Young's modulus, degradation behavior, and swelling properties were further evaluated because these parameters are directly related to printability and co-culture compatibility [28]. Increasing concentration and molecular weight significantly enhanced stiffness and compressive strength, with 1000 kDa GDMA showing superior mechanical support for high-fidelity printing, whereas 3500 Da GDMA exhibited a softer network that may provide a more permissive microenvironment for cell proliferation and infiltration (Fig. 2H). All hydrogels retained more than 50% of their initial mass after 14 days in PBS, indicating stable degradation behavior [29] (Fig. 2I). Swelling analysis showed limited expansion in saline but significantly higher swelling in urea (Fig. 2K, L, Fig. S1G), indicating that hydrogen bonding is also stabilizing interaction within the network, which can be disrupted by urea [30]. Hydrogels constructed through different crosslinking methods exhibit distinct internal bonding energies. The UV photo-crosslinking gels exhibit the most stable structure, followed by oxidation-crosslinked gels. (Fig. 2M). Cyclic compression testing over 50 cycles showed that GDMA hydrogel blocks with different concentrations maintained good mechanical resilience, indicating robust fatigue resistance(Fig. S1D–F). Porosity and adhesion-assembly properties of GDMA The pore size distribution of biomaterials modulates cellular behavior and critically influences tissue regeneration in implantable scaffolds [31]. To define a practically relevant molecular-weight window for biofabrication, we selected 3500 Da and 1,000,000 Da as the lower and upper boundary values of the chain-length-dependent network regime, using cold-water fish gelatin, which is commonly used for GelMA preparation, as the reference background. This design was intended to capture weakly entangled and highly entangled network states, rather than to represent arbitrary molecular-weight extremes. Scanning electron microscopy (SEM) was employed to examine the surface morphology of the samples in order to evaluate the structural characteristics of the GDMA hydrogel (Fig. 3A). Quantitative analysis of the internal pore size distribution (Fig. 3B) revealed that the 3500 Da GDMA hydrogels exhibited a more fragile internal structure at 10%-12% concentration, with collapse observed at 15% concentration. In contrast, the 1,000,000 Da hydrogels formed a uniform and stable internal structure at 10-15% concentration, which was attributed to its higher molecular weight and enhanced mechanical strength. The pore sizes for all hydrogel variants predominantly fell within the 30-50 μm. This scale has been demonstrated to promote mesenchymal stem cell growth and infiltration [32]. Notably, both molecular-weight groups exhibited the largest median pore size at 12%, suggesting that this concentration provides the most balanced porous architecture for bioink design. Importantly, the differences in porosity, mechanics, and printability should not be interpreted as a simple linear function of molecular weight alone, but rather as the integrated outcome of chain length, precursor concentration, molecular-weight distribution, physical entanglement, and effective crosslinking density. This tunable material window therefore provides a structural basis for gradient printing and heterogeneous assembly of multicellular systems (MSCs, iPSCs, ESCs, and EBs) across different size scales, expanding the utility of GDMA for biomimetic tissue construction and regenerative biofabrication [33].Fig. 3Porosity and adhesion-assembly properties of GDMA hydrogels. A) SEM images of GDMA. B) Porosity quantification by ImageJ analysis of SEM images. C) Dry-wet adhesion of hydrogel on porcine skin. D) Oscillatory step rheology test of GDMA hydrogels. E-G) Interfacial adhesion strength evaluated by peeling, lap-shear, and pull-off tests. H) Assembly and perfusion of GDMA tubular modules. I) Macroscale bioprinting and assembly of human organ models (ear, kidney, liver, spine). J) Microscale bioprinting of human-inspired structures (cochlea, glomerulus–tubule complex, liver lobule, single vertebra). K) Construction of a hierarchical, circular, and three-dimensional vascular perfusion system. L) Assembling of large-scale, heterogeneous, tubular bone models. M) Quantitative evaluation of assembly fidelity by structural similarity index measurement (SSIM).Fig. 3 Porosity and adhesion-assembly properties of GDMA hydrogels. A) SEM images of GDMA. B) Porosity quantification by ImageJ analysis of SEM images. C) Dry-wet adhesion of hydrogel on porcine skin. D) Oscillatory step rheology test of GDMA hydrogels. E-G) Interfacial adhesion strength evaluated by peeling, lap-shear, and pull-off tests. H) Assembly and perfusion of GDMA tubular modules. I) Macroscale bioprinting and assembly of human organ models (ear, kidney, liver, spine). J) Microscale bioprinting of human-inspired structures (cochlea, glomerulus–tubule complex, liver lobule, single vertebra). K) Construction of a hierarchical, circular, and three-dimensional vascular perfusion system. L) Assembling of large-scale, heterogeneous, tubular bone models. M) Quantitative evaluation of assembly fidelity by structural similarity index measurement (SSIM). We evaluated the dry and wet adhesion of GDMA hydrogels on porcine skin [22]. The material adhered firmly under dry conditions and remained stable during twisting. The hydrogels were applied either directly or after secondary crosslinking with NaIO4 or FeCl3 and soaked in saline for 72 h. Directly coated hydrogels detached after immersion, whereas crosslinked samples maintained strong adhesion and tolerated 180° bending and repeated twisting without damage (Fig. 3C), demonstrating that secondary crosslinking is essential for interfacial stability in moist environments. In addition, step-strain rheological testing showed that GDMA hydrogels rapidly restored their viscoelastic network after high-strain disruption, indicating strong structural recovery after shear and supporting their suitability for bioprinting applications (Fig. 3D). Inspired by the robust underwater adhesion of marine mussels [34], the catechol groups enable hydrogen bonding and metal chelation in iron-rich environments, generating strong surface adsorption [34,35]. Besides, oxidation of catechols to o-quinone enhances cross-linking and mechanical stabilization at the adhesive interface, resulting in durable and irreversible cohesion [27]. The interfacial adhesion performance of GDMA hydrogels was quantitatively assessed using lap-shear, peel, and pull-off tests. Adhesion strength was determined from the separation points recorded during tensile testing. The lap-shear test showed no significant difference in adhesion with increasing concentration, but the peel test indicated a parallel increase. The pull-off test revealed the strongest adhesion at 12% concentration, might result from the unreacted catechol groups at the interface (Fig. 3E, F, 3G). In the burst pressure test, GDMA exhibited a significantly higher burst pressure than GelMA, indicating improved wet interfacial retention under biomimetic irrigation conditions. Furthermore, GDMA groups treated with oxidants and metal chelators were also included to further evaluate the effect of secondary crosslinking on catecholamine groups. The results confirmed that oxidative crosslinking did indeed slightly reduce interfacial adhesion, which may be related to the conversion of catechol groups to o-quinone species after oxidation (Fig. S1H). These results indicate that GDMA hydrogels have the potential to provide adequate mechanical stability and adhesive performance to support structural integrity under humid biological conditions. 3D printing properties and scalability potential of GDMA hydrogels An oscillatory strain-sweep test was carried out to determine the linear viscoelastic region (LVR) and the strain-dependent viscoelastic response. The storage modulus (G′) and loss modulus (G″) were recorded throughout the test (Fig. S2A). In addition, multiscale pore models with diameters of 400, 600, 800, and 1000 μm were designed to evaluate DLP print fidelity (Fig. S2D). Printed structures were imaged under identical illumination conditions, and geometric fidelity was quantified using the structural similarity index (SSIM) (Fig. 3M) [36]. Thus, this hydrogel enables rapid, Lego-like assembly of complex structures through photopolymerized modules, ensuring leak-free liquid infusion (Fig. 3H) [37]. Supplementary data showed that treatment with 0.1 mM NaIO4 triggered rapid oxidative interfacial crosslinking between adjacent modules, with visible bonding at ∼30 s and stable adhesion achieved within ∼90 s (Fig. S2C). The resulting multilevel branched vascular construct enabled syringe-driven perfusion of a biomimetic fluid without observable leakage. Reconstructing the structural complexity of native human tissues remains a major challenge in tissue engineering, particularly for architectures spanning organ-scale geometries and microscale functional units. Using a bottom-up adhesive assembly strategy, we fabricated centimeter-scale constructs with anatomically relevant forms, including external ears, kidneys, livers, and spinal segments (Fig. 3I), while also achieving precise reconstruction of micrometer-scale architectures (Fig. 3J). A glomerular complex surrounded by a dense capillary-like network was fabricated, with each ∼50 μm glomerulus produced by extrusion printing of GDMA hydrogel and visualized by red aqueous dye infusion (Fig. S3), demonstrating faithful replication of key spatial features. Similarly, interconnected hepatic lobule-like units were assembled while preserving the central arteriovenous portal, interlobular arteriovenous vessels, and bile ducts, and a half-section vertebral construct was engineered for seamless in situ integration with preserved neural and vascular continuity. As a representative demonstration, we constructed a human-scale biomimetic inner ear model in which the external, superior, and posterior semicircular canals were organized in an orthogonal spatial configuration and stably integrated with both the cochlea and vestibule. Such hierarchical organization is difficult to achieve using conventional single-process 3D printing alone, which is often constrained by printing path complexity, limited shape fidelity in overhanging or curved regions, and the difficulty of integrating geometrically discrete substructures into a unified construct. In contrast, the modular adhesive assembly strategy enables independent fabrication of structurally distinct building blocks followed by their precise spatial integration, thereby improving geometric freedom, structural fidelity, and scalability. These results highlight its potential as a versatile platform for complex organoid assembly and programmable biofabrication and demonstrate a practical route for constructing anatomically sophisticated, multicomponent tissue substitutes [38]. Modular assembly enables the fabrication of large-scale grafts while interstitial gaps facilitate nutrient diffusion and reduce central necrosis [39]. A perfusable, biomimetic vascular system was constructed by assembling GDMA modules with lumens, incorporating a ring-mounted A-V loop and a multilayered 3D vascular network (Fig. 3K). A cylindrical bone construct was assembled from heterogeneous modules labeled with distinct fluorescent particles (Fig. 3L). This structure exhibited mechanical stability and resisted moderate compressive forces in a wet PBS environment (Fig. S2B). These examples highlight the potential of modular bottom-up adhesive assembly for constructing complex and spatially ordered architectures [2]. More importantly, from a bioprinting perspective, it addresses persistent challenges including interlayer delamination induced by cell deposition and the uneven distribution of printed components. By decoupling module fabrication from final assembly, this approach also markedly reduces the printing time required for cell-laden constructs, thereby minimizing cell stress and loss of viability during fabrication [40]. Biocompatibility of materials Live/dead staining and confocal imaging confirmed HUVECs viability and spreading within hydrogels on Days 1, 3, and 5 (Fig. 4A). Initial proliferation was comparable on Day 1. By Day 7, cell numbers in both the GelMA and GDMA groups remained similar but were lower than those on tissue culture plastic substrates (Fig. 4B). HUVECs were seeded onto 12% GDMA hydrogel slices to assess the adhesive capacity of GDMA hydrogels. HUVECs exhibited early pseudopodial extension and initial spreading on Day 1. On Day 5, a continuous, three-dimensional tubular network had formed across the hydrogel surface (Fig. 4C). Cell proliferation and viability were quantified using a CCK-8 assay, showing no significant differences among the three cultivation conditions (Fig. 4D). Zhou et al. demonstrated that NaIO4 has minimal cytotoxicity, maintaining ≥83% L929 cell viability [41]. In this study, Live/dead staining of HUVECs cultured on GDMA hydrogels treated with sodium periodate solution confirmed negligible cytotoxicity at low concentration. (Fig. S5A). The ROS activity experiment can also preliminarily confirm that the sodium periodate oxidation cross-linking method exhibits controllable low-level oxidative cytotoxicity (Fig. S6A).Fig. 4Biocompatibility and cell-cell interactions of GDMA hydrogels. A) Viability of HUVECs encapsulated in GDMA hydrogels at concentrations of 10%, 12%, 15%, and 18%, assessed on Days 1, 3, and 5 using Calcein-AM/PI staining (n = 5). B) Quantification of HUVECs proliferation and viability within the hydrogels on Days 1, 3, 5, and 7 using a CCK-8 assay (n = 4, two-way ANOVA, p < 0.001). C) HUVECs were seeded on the surface of 12% GDMA hydrogels and stained with Phalloidin/DAPI to visualize cytoskeletal organization and nuclei. D) HUVECs were seeded on hydrogels of different concentrations, and proliferation and viability were assessed on Days 1, 3, 5, and 7 using a CCK-8 assay (n = 4, two-way ANOVA, p > 0.05). E) HUVECs were seeded into the lumens of 12% GDMA hydrogels and stained for VE-cadherin/DAPI; images were acquired by confocal laser scanning microscopy. F) Quantitative analysis of VE-cadherin expression and cell coverage within the lumens, corresponding to panel (n = 3, Student's t-test). G) Transwell assay showing co-culture of BMSCs and HUVECs. H) Quantitative analysis of scratch closure in wound healing assays (n = 4). I) Quantitative analysis of migrated cells in the lower chambers of transwell assays (n = 5, Student's t-test).Fig. 4 Biocompatibility and cell-cell interactions of GDMA hydrogels. A) Viability of HUVECs encapsulated in GDMA hydrogels at concentrations of 10%, 12%, 15%, and 18%, assessed on Days 1, 3, and 5 using Calcein-AM/PI staining (n = 5). B) Quantification of HUVECs proliferation and viability within the hydrogels on Days 1, 3, 5, and 7 using a CCK-8 assay (n = 4, two-way ANOVA, p < 0.001). C) HUVECs were seeded on the surface of 12% GDMA hydrogels and stained with Phalloidin/DAPI to visualize cytoskeletal organization and nuclei. D) HUVECs were seeded on hydrogels of different concentrations, and proliferation and viability were assessed on Days 1, 3, 5, and 7 using a CCK-8 assay (n = 4, two-way ANOVA, p > 0.05). E) HUVECs were seeded into the lumens of 12% GDMA hydrogels and stained for VE-cadherin/DAPI; images were acquired by confocal laser scanning microscopy. F) Quantitative analysis of VE-cadherin expression and cell coverage within the lumens, corresponding to panel (n = 3, Student's t-test). G) Transwell assay showing co-culture of BMSCs and HUVECs. H) Quantitative analysis of scratch closure in wound healing assays (n = 4). I) Quantitative analysis of migrated cells in the lower chambers of transwell assays (n = 5, Student's t-test). Endothelialization refers to the formation of a continuous monolayer of endothelial cells lining the vascular lumen. This structure protects blood vessels from harmful plasma components and inflammatory cells while preventing vascular occlusion due to thrombus formation [42,43]. We fabricated lumen modules (1 mm) using GelMA and GDMA hydrogels, followed by HUVECs seeding. The modules were cultured under 3D conditions for one week. Then the modules were carefully harvested for immunofluorescence staining for VE-cadherin, a adhesion junction protein essential for endothelial integrity [44,45]. Immunofluorescence staining demonstrated HUVECs in GDMA-printed modules formed a uniform endothelial monolayer with elevated VE-cadherin expression (2.53-fold higher), whereas GelMA modules showed irregular multilayered arrangements (Fig. 4E and F). Beyond that, endothelial cells (ECs) are often the source of paracrine acting molecular signals, which regulate and balance the self-renewal and differentiation of stem cells in surrounding tissues and promote tissue regeneration after damage [46,47]. In the skeletal system, ECs communicate with osteoprogenitor cells during bone development and bone repair, and provide a protective and supportive microenvironment for hematopoietic stem cells [48]. To investigate cellular crosstalk, we examined interactions between BMSCs and HUVECs encapsulated in GDMA hydrogels [49,50]. Transwell assays demonstrated that both BMSC-loaded GelMA and GDMA hydrogels significantly enhanced HUVECs migration. In addition, GDMA alone also promoted migration in the absence of BMSCs, an effect not observed with GelMA (Fig. 4G–I). Consistently, scratch assays confirmed that BMSCs encapsulated in GDMA hydrogels markedly accelerated HUVECs migration (Fig. 4H, Fig. S5B). This effect may result from the slow release of low-dose dopamine within GDMA hydrogels [51]. As a catecholamine neurotransmitter, dopamine can activate endothelial D1 receptors, stimulate the PKA signaling pathway, and upregulate VEGF-A secretion. These combined effects promote endothelial migration and angiogenesis, consistent with findings reported by Chakroborty et al. [52]. In situ unidirectional guided vascularization strategies for modules To evaluate the potential of this strategy in promoting unidirectional vascular ingrowth, we designed cell-free modular constructs that form a J-shaped hollow tube, thereby directing the extension of host vessels in a single direction (Fig. 5A). We prepared microcryogels (100 μm), which serve to maintain a moist environment, prevent reverse vascular growth, and act as a carrier for the gradient delivery of bioactive factors (Fig. 5B). We designed the animal experiments with four groups: (1) Control group: the hydrogel lumen was seeded with HUVECs but was not connected to the subcutaneous arterioles on the dorsum of nude mice; (2) Blank group (GDMA/MC): the lumen was connected to the dorsal subcutaneous arterioles of nude mice; (3) Experimental group (GDMA/Endo/MC): the lumen was endothelialized with HUVECs, connected to the subcutaneous arterioles and supplemented with MC soaked in ECM medium at the distal end; (4) Positive control group (GDMA/Endo/MC-VEGF): the configuration was identical to the experimental group, except that the microcryogels were pre-soaked in VEGF solution (20 ng mL−1) instead of ECM medium (Fig. 5B). BALB/c nude mice were selected as the in vivo model. Fourteen days after implantation, we observed substantial vascular outgrowth in the constructs (Fig. 5C). Tissue samples were obtained on Days 5, 7, 14, and 21. Chronologically arranged images from the experimental group (GDMA/Endo/MC) demonstrated progressive infiltration of host vasculature into the hydrogel lumen (Fig. 5D). However, we noted variability in both the extent and density across individual samples. In the following text, we define “in situ vascularization” as the process involving robust endothelial cell seeding within the tubular lumens of modules, followed by stable connection with host vasculature.Fig. 5Effects of vascularization strategy based on the GDMA modules. A) Schematic diagram of the animal experiment for pre-vascular surgery. B) Schematic diagram of the vascularization experimental groups and effects after implantation. C) in vivo studies demonstrated functional anastomosis between host vasculature and implanted hydrogel modules in the dorsal region of nude mice. D) Results of vascular sprouting in vivo, taken with a stereo microscope. (E) Immunofluorescence staining of vWF and α-SMA confirmed endothelial and smooth muscle marker expression within the constructs. Scale bars: 100 μm (left), 20 μm (right). (F) H&E, Masson's trichrome, and CD31 immunohistochemistry at 21 days revealed tissue integration, ECM deposition, and vascular remodeling.Fig. 5 Effects of vascularization strategy based on the GDMA modules. A) Schematic diagram of the animal experiment for pre-vascular surgery. B) Schematic diagram of the vascularization experimental groups and effects after implantation. C) in vivo studies demonstrated functional anastomosis between host vasculature and implanted hydrogel modules in the dorsal region of nude mice. D) Results of vascular sprouting in vivo, taken with a stereo microscope. (E) Immunofluorescence staining of vWF and α-SMA confirmed endothelial and smooth muscle marker expression within the constructs. Scale bars: 100 μm (left), 20 μm (right). (F) H&E, Masson's trichrome, and CD31 immunohistochemistry at 21 days revealed tissue integration, ECM deposition, and vascular remodeling. H&E, Masson trichrome staining, and CD31 immunohistochemistry of Day 14 revealed distinct outcomes: 1) In the control group, lacking connection to host arterioles, HUVECs remained but formed an underdeveloped capillary network. This likely resulted from the absence of guidance cues and supporting cells such as pericytes, which hindered further blood vessel development. (2) In the blank group, the lumen connected to host vessels was primarily occupied by fibrous connective tissue, containing only sparse capillary lumens. (3) The experimental group displayed a dense vascular network with a functionally mature, hierarchical lumen structure, confirmed by vWF/CD31 immunostaining. H&E and Masson staining also revealed that erythrocyte flow through the capillaries, and there is a large accumulation of collagen, mainly concentrated in the central area. (4) In the positive control group, the VEGF gradient induced excessive sprouting and ingrowth of small vessels, with a structure dominated by collagen bundles (Fig. 5E and F). Quantitative assessment of vascularization showed that the GDMA/Endo/MC group exhibited significantly greater average vessel diameter and microvessel density than the Control and GDMA/MC groups. The GDMA/Endo/MC-VEGF group displayed comparable values, with no significant difference relative to the GDMA/Endo/MC group, suggesting a similar extent of vascular enhancement between these two endothelialized conditions (Fig. S6B). The system promotes vascular network formation by carefully endothelializing the lumen. Nevertheless, we do not exclude the possibility that incorporating various factors could offer effective and feasible enhancements. We selected the GDMA/Endo/MC system for further experiments due to challenges associated with growth factors, including uncontrolled dosing, batch variability, safety concerns, and high costs. Our preliminary findings suggest that vascularizing the GDMA hydrogel promotes the growth of host blood vessels into the transplanted tissue. This vascularization process provides the necessary blood components, oxygen, and nutrients for long-term survival, and creates a suitable microenvironment to support further maturation and differentiation of the graft. Furthermore, the preserved vascular pedicle facilitates the secondary integration of the transplanted tissue with the host vascular network, which may contribute to the success of future transplantation procedures. 3D printing and osteogenic induction of cell-laden modules We further constructed bone organoid modules in vitro by printing BMSC-laden hydrogels following the protocol reported by Wang et al. [53,54], and evaluated their osteogenic differentiation using Alizarin Red S and alkaline phosphatase (ALP) staining. The hydrogel modules printed without BMSCs exhibited faint Alizarin Red staining, while BMSC-laden modules without osteogenic induction (OI) exhibited sparse surface deposits. Following osteogenic induction, pronounced mineral deposition was observed after 14 days, which intensified by Day 21, rendering the entire hydrogel opaque reddish-brown (Fig. 6A). Consistently, BCIP/NBT staining demonstrated progressive ALP activity, with the intensity of dark-purple coloration increasing over time (Fig. 6B). Quantitative analysis with cetylpyridinium chloride confirmed that calcium deposition in osteoinduced BMSC-laden modules was 3.8-fold higher than in uninduced controls after 14 days (Fig. 6C). The BCIP/NBT kit confirmed that alkaline phosphatase activity in BMSCs was 2.29 times higher after 14 days of osteoinduction compared to the uninduced group (Fig. 6D).Fig. 6GDMA hydrogel for the induction and culture of bone organoids. A, B) Alizarin Red S (ARS) and alkaline phosphatase (ALP) staining of printed bone organoid modules following in vitro osteogenic induction. C, D) Quantitative analysis using ARS and Alp activity Assay Kit (n = 3, Student's t-test). E) qPCR analysis of osteogenesis-related genes at day 21 of induction (n = 3, Student's t-test). F, H) Live/dead staining of the cell-laden modules at day 7 and day 14 in vitro. G) Live/dead staining of cell-laden modules after 1, 3, 5, and 7 printing cycles. I) Cryo-SEM imaging showing single-cell spreading and progressive calcium deposition during osteogenic induction. J) Assembly of GDMA hydrogel modules loaded with sCherry-HUVECs and sGFP-HUVECs. Scale bar: 200 μm. K) Cell aggregation and differentiation can be observed in osteogenic bone organoids after osteogenic induction (red arrow). Scale bar: 100 μm. L) Quantification of Live/Dead ratios in panel G (n = 4, one-way ANOVA). Data are presented as mean ± SD; n represents biological replicates unless otherwise stated.Fig. 6 GDMA hydrogel for the induction and culture of bone organoids. A, B) Alizarin Red S (ARS) and alkaline phosphatase (ALP) staining of printed bone organoid modules following in vitro osteogenic induction. C, D) Quantitative analysis using ARS and Alp activity Assay Kit (n = 3, Student's t-test). E) qPCR analysis of osteogenesis-related genes at day 21 of induction (n = 3, Student's t-test). F, H) Live/dead staining of the cell-laden modules at day 7 and day 14 in vitro. G) Live/dead staining of cell-laden modules after 1, 3, 5, and 7 printing cycles. I) Cryo-SEM imaging showing single-cell spreading and progressive calcium deposition during osteogenic induction. J) Assembly of GDMA hydrogel modules loaded with sCherry-HUVECs and sGFP-HUVECs. Scale bar: 200 μm. K) Cell aggregation and differentiation can be observed in osteogenic bone organoids after osteogenic induction (red arrow). Scale bar: 100 μm. L) Quantification of Live/Dead ratios in panel G (n = 4, one-way ANOVA). Data are presented as mean ± SD; n represents biological replicates unless otherwise stated. Previous studies confirmed that GDMA hydrogels possess intrinsic hydrophilicity, interconnected porosity, excellent cytocompatibility, and robust cell-adhesive capacity, rendering them promising candidates for biofabrication [55]. Building on these attributes, we next evaluated their printability and osteoinductive potential in bone organoid-laden constructs. BMSCs and printed GDMA constructs were cultured for 21 days under osteogenic induction (OI) and non-induced conditions. After 21 days, RNA was extracted and osteogenic marker genes (Alp, Runx2, Ocn, Col1a) were quantified by qPCR (Table S1). In GDMA-BMSCs-OI, Alp and Ocn expression increased 9.8- and 4.9-fold, respectively, compared with BMSCs-OI, whereas Col1a expression was markedly higher in BMSCs-OI. This suppression of Col1a may derive from the gelatin-based scaffold of GDMA, which reduces endogenous collagen secretion. Notably, Runx2 expression was elevated even in the uninduced GDMA-BMSCs group, suggesting that the GDMA hydrogel itself imparts osteo-inductive cues capable of driving early osteoblast differentiation without exogenous induction condition (Fig. 6E). In vivo implantation of bone organoids The successful integration and functional restoration of bone tissue in vivo require bone tissue engineering (BTE) grafts with strong vascularization and osteogenic potential [56]. Kusumbe, A. P. previously demonstrated that angiogenesis and osteogenesis are intrinsically linked both spatially and functionally within bone [57]. Specifically, the arrangement of arteries and arteriole-capillaries dictates regional variation, while endothelial cell-driven angiogenesis delivers niche signals, such as those from the HIF and Notch pathways, to regulate nearby osteoprogenitor cells [57,58]. As shown in Fig. 7A, bone organoids were assembled for in vivo transplantation using NOVA strategy. Guided by a biomimetic bone-vascular interface design (Fig. 7B), BMSC-derived bone organoid modules were assembled into columnar constructs with a preformed central lumen. For the vascularized group (experimental), the luminal surface was endothelialized with HUVECs in vitro before implantation, whereas the osteogenic component was provided by the bone organoid modules surrounding the lumen. After implantation, a host microartery was inserted into the lumen to direct in situ vascular ingrowth. Thus, endothelialization was used as a vascularization step for the lumen, while functional vascular integration occurred primarily after implantation through host-vessel guidance. In the nonvascularized group (control), the GDMA-assembled constructs contained BMSC-derived bone organoid modules, but were implanted without luminal endothelialization or microarterial insertion.Fig. 7In vivo evaluation of osteogenesis following transplantation of GDMA-based bone organoids. A) Schematic illustration of bone organoids construction, in vitro osteogenesis induction, and in vivo transplantation. B) Spatiotemporal coupling of endochondral ossification and vascularization in endochondral ossification and the biomimetic design of bone organoids assembly. C) Tibia of age-matched nude mice from the same batch, used as control. D) In vivo implantation of blank GDMA hydrogel for 4 weeks. E) In vivo implantation of GDMA-BMSCs for 12 weeks. F) In vivo implantation of vascularized bone organoids for 12 weeks. G-J) Representative compression testing of native tibia, blank GDMA hydrogel, GDMA-BMSCs constructs, and vascularized constructs. K-L) Histology and immunofluorescence: Elastica–Van Gieson (EVG) staining, and von Willebrand factor (vWF) and α-smooth muscle actin (α-SMA) immunofluorescence to assess small arterial structures. M) micro-CT scan and 3D reconstruction of bone organoids graft at 4-12 weeks in vivo. N) micro-CT imaging of in vitro osteogenesis-induced hydrogel modules, showing a cross section of the middle layer. O) Quantification of BV/TV ratio, BMD, Tb.Th, and Tb.N. P) The overall appearance of GDMA-BMSCs and vascularized modules after 12 weeks of in vitro osteogenesis induction.Fig. 7 In vivo evaluation of osteogenesis following transplantation of GDMA-based bone organoids. A) Schematic illustration of bone organoids construction, in vitro osteogenesis induction, and in vivo transplantation. B) Spatiotemporal coupling of endochondral ossification and vascularization in endochondral ossification and the biomimetic design of bone organoids assembly. C) Tibia of age-matched nude mice from the same batch, used as control. D) In vivo implantation of blank GDMA hydrogel for 4 weeks. E) In vivo implantation of GDMA-BMSCs for 12 weeks. F) In vivo implantation of vascularized bone organoids for 12 weeks. G-J) Representative compression testing of native tibia, blank GDMA hydrogel, GDMA-BMSCs constructs, and vascularized constructs. K-L) Histology and immunofluorescence: Elastica–Van Gieson (EVG) staining, and von Willebrand factor (vWF) and α-smooth muscle actin (α-SMA) immunofluorescence to assess small arterial structures. M) micro-CT scan and 3D reconstruction of bone organoids graft at 4-12 weeks in vivo. N) micro-CT imaging of in vitro osteogenesis-induced hydrogel modules, showing a cross section of the middle layer. O) Quantification of BV/TV ratio, BMD, Tb.Th, and Tb.N. P) The overall appearance of GDMA-BMSCs and vascularized modules after 12 weeks of in vitro osteogenesis induction. In the in vivo nude mouse implantation study, the constructs were monitored by stereomicroscopy at predefined intervals, and representative explants retrieved at 12 weeks are presented in Fig. 7C–F. On micro-CT, the blank GDMA gel was close to the normal soft tissue density (Fig. S6D) and failed to retain its structural stability by 4 weeks, as shown in Fig. 7D. The GDMA-BMSCs group exhibited a pale sequestrum-like appearance at 12 weeks (Fig. 7E). In contrast, the pre-vascularized group preserved its predefined cylindrical architecture and underwent full cortical bone ossification on the surface after 12 weeks. These bone organoids showed a visible, bone-like morphology, resembling a functional bone-vascular membrane (Fig. 7F). Mechanical testing revealed that the tibia deformed by 80% under 400 kPa, leading to irreversible internal collapse (Fig. 7G). The GDMA-BMSCs group displayed intermediate mechanical performance (Fig. 7I), falling between the blank GDMA hydrogel (Fig. 7H) and natural tibia. In contrast, the vascularized group withstood 500 kPa of pressure with 50% deformation (Fig. 7J), demonstrating mechanical strength comparable to that of natural nude mouse tibia. The micro-CT scans and 3D reconstructions at 4, 8, and 12 weeks demonstrated that the vascularized group significantly enhanced osteogenesis. At each time point, this group showed greater formation of cortical bone (green) and cancellous bone (gray) than the GDMA-BMSCs group, and both groups had undergone osteogenic induction before implantation (Fig. 7M). Quantitative analysis revealed that the experimental groups achieved higher Bone Volume Fraction (BV/TV) and Bone Mineral Density (BMD). The vascularized group also exhibited increased Trabecular Thickness (Tb.Th) and Trabecular Number (Tb.N) at 8 weeks. However, Tb.Th declined significantly by 12 weeks, possibly due to cortical bone fusion into continuous sheets, which may not be fully resolved by micro-CT with a minimum resolution of 4 μm (Fig. 7O). To monitor the structural evolution of the GDMA-BMSCs during in vitro osteogenic induction, micro-CT scans were performed on hydrogel modules at selected time points. The scans revealed scattered high-density spots and linear regions within the constructs, resembling the early ossification centers typically observed during endochondral ossification (Fig. 7N). By 12 weeks of osteogenic induction in vitro, bone-like matrix had accumulated on the hydrogel surfaces (Fig. 7P). To further characterize the implanted subcutaneous microvessels, which arose from perforating branches of the gluteal musculature in nude mice, we performed immunostaining and histology α-SMA and CD31 confirmed their microvascular identity. The EVG staining and Masson's trichrome staining methods complement each other, clearly demonstrating the structural characteristics of small arteries, including the thick vessel wall and the intima rich in elastic fibers, which are consistent with the structural features of arterioles (Fig. 7K and L). Osteogenic processes in vivo At the indicated time points, tissues were harvested for subsequent immune-fluorescence and histological analyses, as outlined in Fig. 8A. At 12 weeks post-implantation, we performed double immunofluorescence staining for CD31 and Osteocalcin, followed by tissue clearing (Fig. S6C). Three-dimensional imaging and digital reconstruction were performed using the Light Film Illuminating Micro Imaging System, and fluorescence intensity together with three-dimensional spatial metrics were quantified and rebuild (Fig. 8B). The reconstruction shows a rich microvascular network growing into the bone organoid graft from the implanted arterioles and dermis. Additionally, the high overlap between the Osteocalcin immunofluorescence and the vascular signals suggests that osteoblasts may gradually migrate from the blood vessels to the bone organoid region (Fig. 8C).Fig. 8Histological evaluation of bone organoid maturation and vascularization after implantation (2, 4, 8, and 12 weeks). A) Schematic of in vivo sampling workflow and staining timeline. B) Analysis of fluorescence intensity and 3D spatial quantification from cleared specimens. C) Whole-mount clearing of bone organoid-skin composites, dual immunofluorescence for CD31/OCN, and 3D imaging by light-sheet microscopy. D) H&E and Masson's trichrome staining across time points. (Black arrows, osteoblasts with adjacent lacunae; red arrows, host vessels infiltrating the grafts). E) Immunohistochemistry for osteocalcin at indicated time points. (Green arrows indicate positive areas, representing deposition of periosteocalcin secreted by osteoblasts). F) Volcano plot of DEGs between the GDMA-BMSCs and vascularized groups. G) GO enrichment analysis (top 20 terms); blue boxes denote bone-related functional categories. H) KEGG pathway enrichment (top 20 pathways); blue boxes denote osteogenesis-related pathways.I) Network map of key signaling pathways. J) Correlation heatmap of the global gene-expression matrix. K) Sox9-centered protein–protein interaction network.Fig. 8 Histological evaluation of bone organoid maturation and vascularization after implantation (2, 4, 8, and 12 weeks). A) Schematic of in vivo sampling workflow and staining timeline. B) Analysis of fluorescence intensity and 3D spatial quantification from cleared specimens. C) Whole-mount clearing of bone organoid-skin composites, dual immunofluorescence for CD31/OCN, and 3D imaging by light-sheet microscopy. D) H&E and Masson's trichrome staining across time points. (Black arrows, osteoblasts with adjacent lacunae; red arrows, host vessels infiltrating the grafts). E) Immunohistochemistry for osteocalcin at indicated time points. (Green arrows indicate positive areas, representing deposition of periosteocalcin secreted by osteoblasts). F) Volcano plot of DEGs between the GDMA-BMSCs and vascularized groups. G) GO enrichment analysis (top 20 terms); blue boxes denote bone-related functional categories. H) KEGG pathway enrichment (top 20 pathways); blue boxes denote osteogenesis-related pathways.I) Network map of key signaling pathways. J) Correlation heatmap of the global gene-expression matrix. K) Sox9-centered protein–protein interaction network. H&E and Masson's trichrome staining revealed the long-term survival of BMSCs within the hydrogel (black arrows), along with bone matrix deposition surrounding cellular niches and extensive microvascular infiltration (Fig. 8D). The results of osteocalcin immunohistochemical staining indicate that as the implantation time increases, bone matrix gradually forms, and the vascularized group exhibits superior morphological fidelity, with more extensive deposition of bone matrix and better integration with adjacent tissues (Fig. 8, Fig. 9E). It should be noted that the samples from the vascularized group at 8 weeks required a 2-week decalcification process for paraffin sectioning due to excessive hardness, resulting in light brown sections that were extensively weakly positive.Fig. 9Immune microenvironment, inflammatory response, and transcriptomic validation of implanted bone organoids. A) qPCR validation in independent biological samples for representative genes identified from the RNA-seq dataset, including vegfa, Mef2d, Sox5, and Sox9 (n = 3, Student's t-test). B) In vivo quantification of inflammatory cytokines in hydrogel-associated tissue exudates and hydrogels (n = 3, Student's t-test). C) CD83/CD163 double-label immuno-fluorescence staining to identify macrophage-associated phenotypes. Scale bar: 100 μm D) Quantification of CD83/CD163 fluorescence intensities (n = 4, two-way ANOVA). E) Percentage quantitative analysis of osteocalcin immunohistochemical area (n = 3, two-way ANOVA). Data are presented as mean ± SD with individual data points overlaid.Fig. 9 Immune microenvironment, inflammatory response, and transcriptomic validation of implanted bone organoids. A) qPCR validation in independent biological samples for representative genes identified from the RNA-seq dataset, including vegfa, Mef2d, Sox5, and Sox9 (n = 3, Student's t-test). B) In vivo quantification of inflammatory cytokines in hydrogel-associated tissue exudates and hydrogels (n = 3, Student's t-test). C) CD83/CD163 double-label immuno-fluorescence staining to identify macrophage-associated phenotypes. Scale bar: 100 μm D) Quantification of CD83/CD163 fluorescence intensities (n = 4, two-way ANOVA). E) Percentage quantitative analysis of osteocalcin immunohistochemical area (n = 3, two-way ANOVA). Data are presented as mean ± SD with individual data points overlaid. To investigate the gene expression underlying bone regeneration in bioprinted bone organoids, bulk RNA sequencing was conducted on biological replicates derived from three independent animal cohorts. The volcano plot highlights the differentially expressed genes (DEGs) between the GDMA-BMSCs group and the vascularized group, DEGs defined as |log2FC| ≥ 1, FDR <0.05 (Fig. 8F). Gene Ontology (GO) analysis identified the top 20 enriched biological processes, prominently including immune response, glycolysis, lipid metabolism, and carbohydrate metabolism (Fig. 8G). These results suggest that vascularization reshapes the osteoimmune micro-environment and reprograms metabolic/endocrine pathways. KEGG pathway enrichment analysis revealed strong activation of signaling pathways including Axon guidance, ECM-receptor interaction, Focal adhesion, and PI3K/Akt signaling. We also observed enrichment in additional osteogenesis-related pathways such as Hippo signaling, Wnt signaling, and MAPK signaling (Fig. 8H), indicating their potential involvement in promoting bone formation. Gene interaction network analysis (Fig. 8I) showed the vascularized bone organoids were highly enriched in pathways associated with hematopoietic cell lineage, osteoclast differentiation, chemokine signaling, and NF-κB signaling. To further investigate transcriptional regulation during osteogenesis, we analyzed differentially expressed genes associated with endochondral bone development and angiogenesis. Sox9, Sox5, Zfp521, Fgfr1, and Foxa2 genes involved in endochondral ossification were significantly upregulated in the vascularized group. In parallel, angiogenesis-related genes including Vegfa, Col4a5, Mef2d, and Fgfr1 also showed marked upregulation (Fig. 8J). Functional annotations and relevant biological roles of these genes are summarized in Table S2. A Sox9-centered protein-protein interaction network was constructed using representative differentially expressed genes (Fig. 8K). Sox9 occupied a central position and was connected with multiple genes related to developmental regulation, inflammatory signaling, extracellular matrix remodeling, and vascular-associated processes, including TNF, IL1B, CXCR4, CD44, MAPK3, FGF2, and VEGFA. In addition, qPCR validation was performed in independent biological samples for representative genes identified from the RNA-seq dataset, including Vegfa, Mef2d, Sox5, and Sox9 (Fig. 9A). The qPCR results showed expression patterns consistent with the RNA-seq data, further supporting the reliability of the differential expression analysis and the associated pathway enrichment results presented in Fig. 8J. Ecological niche and physiological function of bone organoids in vivo Monocytes and macrophages serve as precursors of multinucleated osteoclasts, playing a crucial role in immune regulation during bone development and remodeling. During bone development, they not only mediate bone resorption but also contribute to skeletal remodeling by coordinating with osteoblasts to maintain bone homeostasis. We collected bone organoids samples and performed immunofluorescence staining for CD163/CD83 to assess macrophage infiltration (Fig. 9C). Quantitative analysis revealed significant differences in macrophage infiltration ratios among the three groups at both 2 and 8 weeks post-implantation. In the GDMA hydrogel group, the M1/M2 macrophage ratio remained relatively stable over time, with a slight increase in M1 macrophages. The GDMA-BMSCs group exhibited a dynamic shift in the macrophage ratio: M1 macrophages dominated the early stages, while M2 macrophages became more prevalent in the later stages. The vascularized group maintained a relatively stable M1/M2 ratio throughout the period, but consistently showed a significantly higher ratio of M2 macrophages (Fig. 9D). We evaluated local inflammatory responses at the implant site and in the surrounding tissue exudate in vivo at 2 weeks after subcutaneous implantation. Concentrations of IL-1β, TNF-α, and IL-6 in tissue exudates collected from the GDMA-BMSCs and vascularized groups were measured by ELISA (Fig. 9B). The vascularized group exhibited a significantly lower IL-1β level than the GDMA-BMSC group, whereas no significant differences were observed in TNF-α or IL-6. These results indicate that vascularization did not aggravate the early local inflammatory response and may help attenuate implant-associated inflammation at this stage. Discussion The main advance of this work is the establishment of NOVA as an integrated strategy for tissue-scale bone organoid construction. Its significance does not lie in optimizing a single material parameter or merely enlarging bone organoids, but in linking module fabrication, interfacial assembly, in situ vascularization, and in vivo maturation within one continuous framework. Existing organoid engineering approaches can reproduce selected developmental features and spatial patterns at the microscale. However, once extended to the tissue scale, these systems often lose architectural control because heterogeneous modules cannot be integrated in a controllable manner and rapid internal perfusion remains difficult to establish. As a result, large constructs are prone to structural disorganization, restricted mass transport in the core, and insufficient host integration after transplantation. In this context, NOVA extends catechol-based interfacial chemistry beyond its conventional role in wet-tissue adhesion and converts it into a practical strategy for living modular assembly. Within this framework, catechol does not simply improve adhesion, but enables modular organoid bioprinting, adhesion-driven spatial assembly, and subsequent host integration after implantation. A second key finding is that GDMA provides a favorable material window for living modular assembly because its major properties correspond directly to the functional requirements of the system. Its interfacial adhesiveness supports stable module-to-module bonding, which is essential for bottom-up assembly. Its low swelling behavior helps preserve interfacial geometry after implantation and reduces deformation-induced mismatch and local shear stress. Its suitable pore architecture provides sufficient space for cell accommodation, infiltration, and tissue remodeling while avoiding an overly dense network that would hinder mass transport. Its endothelial cell-adhesive surface facilitates luminal endothelialization and thereby improves the efficiency of subsequent host-guided vascular integration. Finally, its low-temperature printability is advantageous for cell-laden fabrication by preserving structural fidelity while reducing thermal and processing-related stress on encapsulated cells. Taken together, these features indicate that GDMA should be regarded not simply as a printable adhesive hydrogel, but as an interfacial bioink that couples assembly, stabilization, and biologically permissive cell support within the same platform. In addition, GDMA behaves as a fully degradable scaffold, and its degradation is not merely a passive loss of mass, but is temporally associated with host tissue infiltration, collagen deposition, and progressive scaffold replacement in vivo (Fig. S4A–F). The effect of vascularization in this system also extends beyond a survival benefit. Bone development and regeneration are governed by a well-recognized bone-vascular axis, in which angiogenesis and osteogenesis are spatially and functionally coupled[59]. Within this framework, the preformed lumen in NOVA should be viewed as a developmental conduit rather than merely a perfusion space. By combining luminal endothelialization with host-guided vascular ingrowth after implantation, the construct gains earlier access not only to oxygen and nutrients, but also to vascular-derived signals, recruited progenitor stromal cells, and remodeling cues that can support further organoid maturation. This interpretation helps explain why the vascularized group showed not only improved viability and organization, but also more advanced bone-like maturation and stronger bone-vascular coupling signatures at the transcriptomic level. Thus, the principal role of the preformed lumen is to impose a directional vascular entry path that promotes organized maturation rather than relying on delayed and stochastic ingrowth from the periphery alone. Several limitations should nevertheless be acknowledged. The present in vivo study is best interpreted as a mechanistically focused ectopic vascularized maturation model, not as a definitive test of functional load-bearing bone repair. This design is appropriate for evaluating modular assembly, host integration, and guided vascularization under relatively controlled conditions, but it does not reproduce the full biomechanical and anatomical demands of orthotopic bone healing. In addition, although the current work demonstrates progressive maturation and ectopic bone formation, it does not yet establish regenerative efficacy in a critical-size defect or quantify functional recovery under load. For true translational validation, the next step should move to orthotopic large-animal segmental bone defect models, in which defect size, fixation, biomechanics, and functional restoration can be evaluated in a clinically relevant manner. Accordingly, the present study should be viewed as establishing the biological and engineering basis of modular living assembly with in situ guided vascularization, while future work should determine whether this strategy can produce functional regenerative benefit in orthotopic defect repair. Conclusion This study establishes NOVA as a modular living assembly strategy for bone organoid construction with in situ guided vascularization. Using a GDMA bioink that combines printability, interfacial adhesion, and post-assembly stabilization, we achieved the bottom-up construction of lumenized bone organoid grafts with customizable geometry and structural integrity. Following implantation, host-guided vascular ingrowth through the preformed lumen promoted coordinated angiogenesis, osteogenesis, and tissue remodeling, ultimately supporting the maturation of tubular bone-like constructs with functional vascular integration. Together, these results identify modular living assembly and guided vascularization as two complementary design principles for building larger, more organized, and functionally maturing organoid-based grafts. NOVA therefore provides a useful framework for the development of vascularized tissue substitutes in regenerative medicine. Experimental section Preparation of GDMA: Low-temperature soluble gelatin from cold-water fish skin (CAS 9000-70-8) was used as the gelatin precursor. Briefly, gelatin (4.0 g) and dopamine hydrochloride (2.0 g, CAS 62-31-7) were dissolved in 200 mL phosphate-buffered saline under magnetic stirring. Subsequently, N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC·HCl, 1.0 g, CAS 25952-53-8) and N-hydroxysuccinimide (NHS, 0.6 g, CAS 6066-82-6) were added as carboxyl-activating coupling reagents to mediate the amidation reaction between the carboxyl groups of gelatin and the primary amine of dopamine. The reaction was allowed to proceed overnight on a shaker at 40 rpm. The resulting solution was transferred into dialysis tubing and dialyzed against deionized water for 48 h, with the external water refreshed every 4 h. After dialysis, the solution was frozen at −20 °C and lyophilized to obtain Gel-Dopa. For methacrylation, Gel-Dopa was dissolved in 100 mL of 100 mM MES buffer (pH 5.0) under stirring. EDC and NHS were then added again to activate the remaining carboxyl groups, followed by the dropwise addition of 2-aminoethyl methacrylate hydrochloride (AEMA, 0.4 g, CAS 2420-94-2). The reaction mixture was stirred overnight at room temperature at 40 rpm under light-protected conditions. The product was purified by dialysis against deionized water for 48 h, followed by lyophilization to obtain Gel-Dopa-MA. The lyophilized product was stored a −20 °C before further use. Catechol quantification: The catechol content of GDMA was quantified by UV–vis analysis using the Arnow assay with dopamine hydrochloride as the standard. A calibration curve was established using dopamine solutions of known concentrations, and the catechol concentration of GDMA samples was calculated accordingly. Catechol content was expressed as μmol per gram of polymer using the equationCatecholcontent(μmolg−1)=Ccatechol×Vmpolymerwhere Ccatechol is the catechol concentration in the tested solution (μmolmL−1), V is the solution volume (mL), and mpolymer is the mass of the polymer sample (g). 3D bioprinting protocol: 3D printing were conducted using a DLP printer (EFL-BP-8601 Pro). Three-dimensional models of the ear, kidney, liver, and spine were created using Shining 3D scanning. The kidney was sectioned along the coronal plane, the liver along the coronary ligament, and the spine along the vertebral level using Magics 24.0 software. Parameters of 3D printing: The GDMA hydrogel precursor solution was prepared with 12% (w/v) GDMA, 0.1% (w/v) tartrazine, and 0.25% (w/v) LAP. DLP printing was performed using a layer thickness of 30 μm, a light intensity of 16 W, and an exposure time of 18 s per layer. Tubular bone models were generated in UG NX 12.0.0.27, consisting of a 4-mm-diameter cubic construct containing internal channels ranging from 1 to 1.5 mm. Assembly of modules: The hydrogel modules were subsequently assembled using 0.1 mM sodium periodate solution at the interfaces. The time for cross-linking between module interfaces to stabilize is approximately 90 s (Fig. S2C). Printability evaluation: Models with pore diameters of 400, 600, 800 and 1000 μm were designed to assess structural support and concave-pore features. Printed constructs were imaged under identical illumination conditions using a stereomicroscope. Structural fidelity was quantified using the structural similarity index (SSIM), and the resulting values were subjected to statistical analysis and visualization. Mechanical Properties: Cylindrical hydrogel specimens were tested using a ZQ-990A universal testing machine. Uniaxial compression was performed at 1 mm min−1 until failure. The compressive Young's modulus was determined from the linear region of the stress-strain curve (10-20% strain). Cyclic compression was conducted to 50% strain for 50 cycles. Pore Morphology and Size Distribution: Pore morphology was characterized by SEM (Sigma 300, Zeiss, Oberkochen, Germany) at an accelerating voltage of 4.0 kV. Representative images were recorded at 160 × , while pore-size quantification was performed on SEM images acquired at a uniform magnification of 400 × . Five randomly selected images from the surface of each hydrogel group were analyzed in ImageJ. All images were calibrated to the same scale and segmented using an identical manually defined threshold range (85-255). Pore size was measured from the binarized images and expressed as equivalent circular diameter. Swelling Experiment: GDMA hydrogels with molecular weights of 1,000,000 Da and 3500 Da and concentrations of 10%, 12%, and 15% were prepared into round modules (1 cm diameter). After 24 h of immersion in 0.9% saline and 8% urea solutions, the hydrogel modules were placed on the grid ruler and the modules were weighed and recorded. Rheological characterization: Rheological measurements were performed using a Mars 40 rheometer (Haake Mars40, Germany) equipped with a parallel-plate geometry. All tests were conducted at 25 °C using a plate diameter of 8 mm. Before testing, samples were equilibrated for 1 min under pre-shear condition. An oscillatory strain-sweep test was carried out at a fixed frequency of 1 Hz with strain ranging from 0.1% to 1000%. A step-strain test was further performed at a fixed frequency of 1 Hz by alternating the strain between 1% and 500%. Cell-Laden 3D Bioprinting: Primary bone marrow–derived mesenchymal stem cells (BMSCs) were isolated using established protocols [53]. BMSCs were first seeded at an initial density of 2.0 × 107 cells mL-1 to generate bone organoid spheres, and the resulting spheroids were subsequently collected and resuspended in GDMA hydrogel ink at a density of approximately 1 × 105 spheroids mL-1 for printing. Following printing, constructs were rinsed with PBS and cultured in low-glucose DMEM (Hyclone) for subsequent studies. Cell Viability: Cells were stained with a live/dead staining assay kit. After rinsing the cells three times with PBS, Calcein-AM was diluted to 2 μM with serum-free medium and PI was diluted to 2 μg mL−1 with PBS, and the staining was performed for 30 min and 10 min respectively. After rinsing with PBS, the cells were photographed with a scanning confocal microscope and the live and dead cells were counted using ImageJ. Cell Morphology: After fixation with 4% paraformaldehyde solution at room temperature for 30 min, the sections were stained with 100 nM TRITC-labeled phalloidin stain for 30 min, rinsed three times with PBS, and counterstained with DAPI for 5 min. Vascularization: HUVECs were digested with 0.25% trypsin–EDTA and resuspended at 5 × 106 cells mL−1 in complete medium. The suspension was injected into pre-printed vascular channels using a sterile 1 mL syringe. After 15 min incubation on one side, the construct was rotated 90°, and the process was repeated for four cycles to ensure uniform seeding. Constructs were then transferred to 24-well plates, and fresh medium was added for further culture. Osteogenic Induction: Bone organoids were encapsulated in GDMA hydrogels and cultured in osteogenic induction medium (low-glucose DMEM supplemented with 10 nM dexamethasone, 10 mM β-glycerophosphate, and 250 μM L-ascorbic acid or ascorbic acid-2-phosphate). Cultures were maintained for 2-3 weeks with medium changes every 2-3 days. Basic fibroblast growth factor (bFGF, 10 ng mL−1) was included to enhance osteogenic differentiation. Alkaline Phosphatase Activity: To evaluate Alp activity in modules after 21 days of osteogenic induction, we employed the BCIP/NBT Alkaline Color Development Kit (Beyotime) according to the manufacturer's guidelines for color development. Quantitative analysis was conducted using the Alkaline Phosphatase Detection Kit (Beyotime). The modules were incubated with p-nitrophenyl phosphate (pNPP) at 37 °C for 30 min. Absorbance of yellow p-nitrophenol (pNP) was measured at 405 nm using a multifunctional microplate reader (BioTek). Alizarin Red Staining: The modules were stained using the Alizarin Red S Staining Kit (Beyotime) to assess osteogenesis. After 21 days of osteogenic induction, the modules were rinsed with PBS and stained with ARS. Calcium nodules were visualized using a microscope and camera. Subsequently, ARS was dissolved in an alkaline hexadecylpyridinium chloride solution, and the optical density (OD) at 562 nm was measured using a microplate reader (BioTek). qPCR Analysis: Total RNA was isolated from BMSC-laden GDMA or GelMA modules using TRIzol (Servicebio) with on-column DNase (Servicebio). cDNA was synthesized from 500 ng RNA (PrimeScript RT, Takara). qPCR (QuantStudio 5) employed PowerUp SYBR Green (Thermo) in 10 μL reactions (0.4 μM primers). Cycling: 95 °C 2 min; 40 × (95 °C 15 s, 60 °C 30 s); melt curve 65-95 °C. Primers for osteocalcin (Ocn), collagen I (COL1A1), Runt-related transcription factor 2 (Runx2), Alkaline Phosphatase (Alp), are listed in Supplementary Table S1. Expression was normalized to the geometric mean of Gapdh (geNorm-validated). Relative expression used 2∧-ΔΔCt with Day 0 GDMA as calibrator. ΔCt values were analyzed using an unpaired two-tailed Student's t-test on biological replicates (n = 3). Fold changes were calculated using the 2∧−ΔΔCt method and are reported with 95% confidence intervals. Animal Model and Surgical Procedure: Four-week-old male BALB/c nude mice (Hubei Baiente Biotechnology Co., Ltd., China) were housed under specific pathogen-free (SPF) conditions and acclimated for 5 days before surgery.1)In situ unidirectional guided vascularization strategies: Mice were anesthetized by intraperitoneal injection of sodium pentobarbital (50 mg kg−1). After routine surgical preparation, a 2 cm dorsal midline incision was made and bilateral subcutaneous pockets were created by blunt dissection. Under a surgical microscope, subcutaneous microvessels were carefully exposed, and approximately 5 mm of distal free vessel length was obtained by gentle adventitial dissection. The free end of the vessel was then inserted into the preformed lumen of the printed GDMA-based construct. A microcryogel (∼100 μm) soaked in low-glucose DMEM was placed at the distal end of the lumen to maintain local hydration. The incision was closed with 5-0 absorbable sutures, and postoperative wound care was provided for 3 days. Samples were harvested at 2, 4, and 8 weeks after implantation for subsequent analyses.2)In vivo implantation of bone organoids: Mice were randomly assigned to three experimental groups: GDMA, consisting of 3D-printed and assembled acellular GDMA hydrogel blocks; GDMA-BMSCs, consisting of GDMA-assembled modules containing BMSC-derived bone organoids; and prevascularized GDMA-BMSCs, in which the GDMA-BMSC constructs were further seeded with endothelial cells in the lumen and cultured for prevascularization before implantation. The implanted constructs measured 3 × 3 × 6 mm3. For each group, 8 mice were included at each time point. Sample size was determined based on our prior experience with this implantation model and pilot studies. Bilateral subcutaneous pockets were created in each mouse, and implants from the same experimental group were placed on both sides. The two implants from the same mouse were allocated to different downstream analyses. Accordingly, the mouse, rather than the individual implant, was defined as the statistical unit. 1)In situ unidirectional guided vascularization strategies: Mice were anesthetized by intraperitoneal injection of sodium pentobarbital (50 mg kg−1). After routine surgical preparation, a 2 cm dorsal midline incision was made and bilateral subcutaneous pockets were created by blunt dissection. Under a surgical microscope, subcutaneous microvessels were carefully exposed, and approximately 5 mm of distal free vessel length was obtained by gentle adventitial dissection. The free end of the vessel was then inserted into the preformed lumen of the printed GDMA-based construct. A microcryogel (∼100 μm) soaked in low-glucose DMEM was placed at the distal end of the lumen to maintain local hydration. The incision was closed with 5-0 absorbable sutures, and postoperative wound care was provided for 3 days. Samples were harvested at 2, 4, and 8 weeks after implantation for subsequent analyses. In situ unidirectional guided vascularization strategies: Mice were anesthetized by intraperitoneal injection of sodium pentobarbital (50 mg kg−1). After routine surgical preparation, a 2 cm dorsal midline incision was made and bilateral subcutaneous pockets were created by blunt dissection. Under a surgical microscope, subcutaneous microvessels were carefully exposed, and approximately 5 mm of distal free vessel length was obtained by gentle adventitial dissection. The free end of the vessel was then inserted into the preformed lumen of the printed GDMA-based construct. A microcryogel (∼100 μm) soaked in low-glucose DMEM was placed at the distal end of the lumen to maintain local hydration. The incision was closed with 5-0 absorbable sutures, and postoperative wound care was provided for 3 days. Samples were harvested at 2, 4, and 8 weeks after implantation for subsequent analyses. 2)In vivo implantation of bone organoids: Mice were randomly assigned to three experimental groups: GDMA, consisting of 3D-printed and assembled acellular GDMA hydrogel blocks; GDMA-BMSCs, consisting of GDMA-assembled modules containing BMSC-derived bone organoids; and prevascularized GDMA-BMSCs, in which the GDMA-BMSC constructs were further seeded with endothelial cells in the lumen and cultured for prevascularization before implantation. The implanted constructs measured 3 × 3 × 6 mm3. For each group, 8 mice were included at each time point. Sample size was determined based on our prior experience with this implantation model and pilot studies. Bilateral subcutaneous pockets were created in each mouse, and implants from the same experimental group were placed on both sides. The two implants from the same mouse were allocated to different downstream analyses. Accordingly, the mouse, rather than the individual implant, was defined as the statistical unit. In vivo implantation of bone organoids: Mice were randomly assigned to three experimental groups: GDMA, consisting of 3D-printed and assembled acellular GDMA hydrogel blocks; GDMA-BMSCs, consisting of GDMA-assembled modules containing BMSC-derived bone organoids; and prevascularized GDMA-BMSCs, in which the GDMA-BMSC constructs were further seeded with endothelial cells in the lumen and cultured for prevascularization before implantation. The implanted constructs measured 3 × 3 × 6 mm3. For each group, 8 mice were included at each time point. Sample size was determined based on our prior experience with this implantation model and pilot studies. Bilateral subcutaneous pockets were created in each mouse, and implants from the same experimental group were placed on both sides. The two implants from the same mouse were allocated to different downstream analyses. Accordingly, the mouse, rather than the individual implant, was defined as the statistical unit. Histological Analysis: Histological analysis was performed on samples collected at different time points. Tissue sections were dewaxed, rehydrated, and processed using standard H&E and Masson's trichrome staining. Immunohistochemical and immunofluorescent staining, including CD31 and OCN, were used to assess vascularization and bone formation, while α-SMA/CD31 co-staining was employed to visualize subcutaneous blood vessels. Additionally, Verhoeff–Van Gieson (EVG) staining was performed to evaluate elastic fibers in microvessels. Stained sections were imaged using a Nikon H600L optical microscope and Nikon laser confocal microscope. 3D Immunolabeling: The 3D immunolabeling of the whole mouse heart was conducted by Jarvis (Wuhan) Biological Pharmaceutical Co., Ltd. The tissue was initially treated with organic reagents at room temperature for 5 days and decolorized using a 5% H2O2 solution overnight at 4 °C. Subsequently, the sample was incubated with Triton X-100 solution for 4 days at room temperature with gentle shaking. The sample was then subjected to a 10% serum blocking solution for 4 days at room temperature, followed by incubation in the primary antibody solution for 7 days at room temperature. The secondary antibody solution (donkey anti-rabbit 647 nm, 1:500 dilution) was applied for 7 days at room temperature. After antibody incubation, the sample was washed three times with a 0.1% Tween 20 solution for 2 days at room temperature to remove excess antibodies. Clearing Method: Tissue clearing was performed using the FDISCO clearing kit (JA11012, Jarvis, Wuhan, China) according to the manufacturer's protocol. Fluorescence Imaging: The cleared samples were imaged using a light sheet microscope (LiToneXL, Light Innovation Technology, China) equipped with a 4 × objective lens. Thin light sheets were illuminated from four sides of the samples, and the merged images were captured. For image acquisition, the cleared samples were manually secured onto the sample holder adapter. The samples were then immersed in the imaging reagent within a 3D printing sample chamber and excited with light sheets of varying wavelengths. Data processing: The animal experimental workflow was created using BioRender. Basic image processing was performed using Fiji (ImageJ). Three-dimensional reconstruction and quantitative image analysis of micro-CT data were conducted using CTAn software. Statistical analyses and graph generation were performed using GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA). Statistical analysis: Data are presented as mean ± standard deviation (SD), unless otherwise stated. Before parametric analysis, normality and homogeneity of variance were assessed. For comparisons between two groups, an unpaired two-tailed Student's t-test was used. For comparisons among multiple groups, one-way ANOVA or two-way ANOVA followed by appropriate post hoc multiple-comparison tests was applied, depending on the experimental design. Unless otherwise specified, n represents independent biological replicates, and technical replicates were averaged before statistical analysis. Statistical significance was defined as ns, p > 0.05; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; and ∗∗∗∗p < 0.0001. CRediT authorship contribution statement Jingyi Ju: Investigation, Data curation, Conceptualization. Huimin Fang: Investigation, Conceptualization. Junjin Jie: Methodology, Investigation. Yuezhou Wu: Investigation, Data curation. Jian Wang: Investigation, Formal analysis, Data curation. Qianmin Gao: Visualization, Validation, Formal analysis. Yifan Zhang: Visualization, Validation, Software. Chao Luo: Supervision. Houfeng Zheng: Methodology, Conceptualization. Jiaming Sun: Writing – review & editing, Resources, Funding acquisition. Jiacan Su: Writing – review & editing, Funding acquisition, Conceptualization. Zhenxing Wang: Writing – review & editing, Funding acquisition, Conceptualization. Ethics approval and consent to participate This study did not involve human participants; therefore, informed consent was not applicable. All animal procedures were approved by the Institutional Animal Care and Use Committee of Huazhong University of Science and Technology and were performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals (NIH Publication No. 8023, revised 1978). Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.