Universal Solvent Escape Strategies for Efficient Curing of Hydrogen‐Bond‐Rich 3D Printing Inks

ABSTRACT

Hydrogen‐bonded polymers have attracted significant interest in biomedical applications due to their excellent biocompatibility, adjustable mechanical properties, and responsiveness to environmental cues. However, these materials face substantial challenges in direct ink writing, primarily arising from persistent solvent entrapment within dense hydrogen‐bonded networks. This trapped solvent severely impairs the printability, drying efficiency, and structural fidelity of scaffolds. To overcome these limitations, this study introduced innovative universal solvent escape strategies integrating three key mechanisms: first, solvent replacement disrupted existing hydrogen‐bonded polymer complexes; second, nanoparticle‐induced microturbulence significantly enhanced solvent evaporation rates; third, computationally optimized printing paths facilitated efficient solvent evacuation. Molecular dynamics simulations provided quantitative insights into how various ways can effectively destroy hydrogen bond solvent networks, allowing rapid solvent removal. Finite element analysis accurately visualized the curing behavior to maximize solvent extraction. This integrated strategy enabled exceptionally rapid (< 3 min) and structurally precise scaffold fabrication across diverse hydrogen‐bonded polymers, including chitosan, collagen, and cellulose. Furthermore, these scaffolds exhibited multifunctional capabilities, utilizing hydrogen‐bonding networks for both structural integrity and pH‐responsive drug delivery. Functional scaffold confirmed significantly improved osteogenic and angiogenic performance via enhanced calcium signaling and activation of HIF‐1α pathways, thereby advancing the fields of tissue engineering and controlled therapeutic delivery.

This study developed a new 3D printing method for hydrogen‐bonded polymers by combining solvent replacement, nanoparticles, and optimized printing paths. This allows fast, precise scaffold fabrication. The scaffolds can be easily customized and release therapeutic agents slowly through protonation, enabling personalized bone, blood vessel, and nerve repair for advanced tissue engineering applications.

1. Introduction

Tissue engineering has emerged as a pivotal interdisciplinary field addressing the growing demand for tissue regeneration and organ repair solutions. By constructing three‐dimensional (3D) biocompatible scaffolds integrated with bioactive factors or cells, this technology demonstrates exceptional potential in restoring damaged physiological structures [1, 2]. However, conventional scaffold fabrication techniques, constrained by limited geometric design flexibility and inadequate control over microstructural characteristics, pose significant barriers to clinical translation [3]. These limitations underscore the critical need for innovative manufacturing approaches that can achieve both structural precision and biological functionality.

In this scenario, 3D printing technology, recognized as a remarkably flexible additive manufacturing approach, offers a novel avenue for the advancement of tissue engineering scaffolds. Through its layer‐by‐layer construction methodology, 3D printing not only enables the fabrication of intricate geometric forms but also allows for the precise regulation of critical parameters including porosity, morphology, and mechanical characteristics. This precision leads to a substantial enhancement in the scaffold's functionality [4]. Direct Ink Writing (DIW) technology, in particular, has emerged as a focal point of research. It boasts numerous merits, such as uncomplicated equipment requirements, cost‐effectiveness, and a broad spectrum of applicable materials [5, 6]. In contrast to high‐temperature melt printing, DIW can be carried out at room or low temperatures. This characteristic makes it an ideal choice for heat‐sensitive materials, especially polymers rich in hydrogen bonds, thereby paving new ways for the development of tissue engineering scaffolds [7, 8]. Nevertheless, the central hurdle faced by DIW lies in achieving rapid ink solidification [9]. This process must maintain the integrity of the printed shape while simultaneously guaranteeing the mechanical and functional attributes of the final product [10]. The existing DIW technology predominantly depends on solidification strategies, including light curing, thermal curing, and cooling curing. However, these approaches come with certain limitations. Photocuring technology, for instance, has a fast curing speed (1–3 min), high precision (<100 µm), and high compressive strength (>100 MPa). Nevertheless, it mandates the use of specialized photosensitive materials. Moreover, its heavy reliance on UV light might result in a surge in equipment costs [8, 11]. Thermal curing is applicable to some thermoset materials. Curing times are typically 5–15 min, with resolutions around 200 µm and compressive strengths between 0.5 and 2 MPa [8, 12]. But the high‐temperature environment involved can potentially damage biologically active molecules like proteins and growth factors [12, 13]. Cooling solidification suits low melting point or phase change materials. Its curing time is usually 2–10 min, the resolution is about 150 µm, and the compressive strength is low (<1 MPa) [14]. However, the cooling rate and uniformity have a direct bearing on the printing effect and the internal microstructure of the material [15, 16]. In addition, there were also strategies that rely on the phase transition of the second phase for solidification, such as adding wax to silicon‐based inks to achieve rapid solidification of the ink through its solid‐liquid solid phase transition [14]. However, this was also based on temperature changes and might introduce second phase impurities.

Polymer materials abundant in hydrogen bonds find wide application in tissue engineering, thanks to their dynamic cross‐linking characteristics and outstanding biocompatibility. Through reversible hydrogen bonding cross‐linking, these materials can also offer excellent mechanical properties and self‐healing capabilities [17, 18]. Nonetheless, when it comes to 3D printing, its solidification speed is sluggish, which easily leads to structural collapse. As a result, its application in DIW 3D printing is restricted [19, 20].

To overcome the aforementioned bottleneck issues in DIW technology, this paper puts forward three optimization strategies. The first strategy is the solvent replacement method. This approach involves utilizing an alternative solvent that is immiscible with the solute yet miscible with the current solvent. Then the solute in the ink is precipitated, leading to the rapid solidification of the ink. The second strategy is to introduce inorganic particles as channels for solvent evaporation, thereby accelerating the curing process. Moreover, the interaction between inorganic particles and the polymer matrix endows tissue scaffolds with better biomechanical characteristics. The third strategy is path optimization. This aims to increase the interval time between contact points as much as possible by adjusting the printing path, thereby leaving more sufficient time for the slurry to solidify.

Through the systematic optimization of the solidification strategy for DIW slurry, this paper offers innovative concepts for the advancement of tissue engineering scaffolds. These optimization techniques not only enrich the theoretical research regarding DIW technology but also open up new avenues for the development and application of multifunctional biomaterials.

2. Results

2.1. Solvent Replacement Solidification Strategy

DIW 3D printing of materials primarily depended on two factors. First, the material required a suitable solvent for uniform dissolution to ensure continuous extrusion under a specific pressure. Second, after extrusion by the printer, the solvent in the material was supposed to volatilize rapidly to enable the material to solidify and form [12]. A polyurethane (PU) soluble in ethanol was synthesized (Figure S1), which was attributed to the strong hydrogen bonding that could form between the carbamate bonds (─NHCOO─) in the PU molecular chain and the hydroxyl groups (─OH) in ethanol. However, these alcohol molecules were entrapped by the PU macromolecular chains and were challenging to remove completely through simple drying (Figure 1A), which impeded the solidification process. After drying and solidification the dissolved PU, infrared testing indicated the presence of a distinct ─OH peak compared to the undissolved PU (Figure 1B). Given that hydrogen bonds broke upon heating, the infrared spectra of PU molecules were measured at room temperature (25°C) and 100°C respectively (Figure 1C). It was observed that the peak intensity of ‐NH‐ in the range of 3500–3200 cm⁻¹ decreased after heating. This phenomenon suggested the existence of hydrogen bonds between or within the PU macromolecules, and the hydrogen bonding effect weakened as the temperature rose [21]. As the hydrogen bonding effect diminished, the electron cloud density around the electron donating group ─NH─ increased, leading to an increase in the vibration frequency. The ─NH─ peak underwent a blue shift from 3324 cm⁻¹ at 25°C to 3333 cm⁻¹ at 100°C. Since PU was a polar polymer, the secondary amine (─NH─) groups on its molecular backbone carried a positive charge and could act as proton donors, while the carbamate bonds were negatively charged and could function as proton acceptors (Figure 1D). Consequently, the main chain of the PU could form a three‐dimensional “cage” structure due to hydrogen bonding between these groups, which trapped the solvent molecules and hindered their escape and volatilization.

FIGURE 1.

(A) The state of PU‐ethanol solution in air; (B) Fourier Transform Infrared (FTIR) spectrum of PU before and after drying; (C) Displacement of hydroxyl FTIR spectrum of PU before and after heating; (D) Schematic diagram of hydrogen bonding between PU molecular chains, and between PU molecules and ethanol molecules. (E) PU rapidly solidifies and forms in water; (F) The model of PU solidifies in water/air; (G) The radial distribution function of ethanol molecules around PU molecules; (H) The interaction energy between ethanol and PU; (I) Schematic diagram of hydrogen bonding involved in PU solidification process.

In order to make PU materials solidification and mold quickly during the DIW 3D printing, the solvent replacement strategy was first adopted. That was to immerse the printed scaffold in water, taking advantage of the fact that ethanol solvents can easily dissolve in water while PU is insoluble in it. It was found that the PU scaffolds printed in the air were difficult to cure and form quickly, which led to the scaffold's collapse and inter‐layer fusion. In contrast, the PU scaffolds printed in the water could cure quickly and had clear shapes (Figure 1E). The idea of solvent replacement was also similar to the recently published paper that supports gel to replace the solvent in the fiber to prevent fiber fracture [22]. Then a model of the PU/ethanol complex in water was constructed. In the initial state, PUs was randomly and uniformly distributed in ethanol molecules. After structural optimization, annealing, Molecular Dynamics (MD) under the NPT ensemble, and MD under the NVT ensemble, finally, it was found that under anhydrous conditions (M1), PU molecules remained randomly and uniformly distributed in ethanol molecules. However, when water was present, PU molecules precipitated and agglomerated from ethanol molecules (Figure 1F, Videos S1 and S2). Figure 1G showed the radial distribution function of ethanol molecules around PU molecules. Compared with the situation without water, the probability of aqueous ethanol appearing around PU molecules decreased. The interaction trend between the two was weakened, which was consistent with the precipitation and reunion of PU from ethanol when water was contained as shown in Figure 1F. Afterwards, we measured the interaction energy between ethanol and PU (Figure 1H). When water was present, the interaction between PU and ethanol was weakened. This was consistent with the agglomeration of PU in water and also in line with the radial distribution function results of Figure 1G. These results were represented by Figure 1I. That is, when there was no water, the ethanol molecule formed hydrogen bonding with groups such as carbonyl group (─C═O), oxygen atom (─O─), ─NH─ in the PU molecular chain, dissolving the PU. When cured in water, the hydrogen bonding between ethanol molecules and water molecules was stronger. Since PU molecules were insoluble in water, the PU molecules separated and solidified from the water. Afterwards, the solidification process of polyurethane in air environment and water was simulated by finite element method. In less than 3 min, the solidification degree of PU in water could reach 10%, while in the air, it took more than 10 min (Figure S2). In addition, since ethanol evaporated slowly in the air, a dense separator was easily formed on the surface of the slurry, further preventing the evaporation and diffusion of ethanol. But in the water, a phenomenon of solidification from the surface to the center was observed (Figure S3).

Numerous hydrogen bond‐rich polymeric materials, including chitosan, PU, nylon 6, collagen, and cellulose, demonstrated extensive applications in biomedical fields. However, their three‐dimensional fabrication predominantly relied on high‐temperature melt‐based 3D printing techniques, which imposed significant limitations on their utility in tissue engineering applications. Even when DIW 3D printing could be processed, the resulting gel‐like structures exhibited compromised structural resolution and insufficient mechanical performance. To address this issue, we promoted the DIW printing method of PU in water to more soluble polymers. Several polymer materials commonly used in biomedicine were selected: chitosan, nylon 6, collagen, and cellulose (Figure 2A). Analysis of their molecular structures showed that these substances contained ─OH, amino group (─NH₂), ─C═O, and ─NH─ groups, which were prone to hydrogen‐bonding (Figure 2B). In their molecular chains, hydrogen bonds could form between the hydrogen atoms in ─NH₂ and ─OH, between ─OH groups, between the oxygen atoms in ─NH₂ and ─OH, between ─OH and ─O─, between ─NH₂ and ─C═O, and between ─OH and ─C═O (Figure 2C). This hydrogen bonding network was the reason why they had difficulties in DIW 3D printing. Subsequently, these materials were dissolved in different solutions: 2% (v/v) acetic acid, trifluoroacetic acid, 1% (v/v) acetic acid, and 1‐butyl‐3‐methylimidazole chloride. It was found that some of these solutions could dissolve the materials well without significant agglomeration (Figure 2D), indicating potential for DIW 3D printing. However, after extrusion, similar to the aforementioned PU, these materials were difficult to solidify and mold quickly in an air environment. The curing times of chitosan and cellulose in the air both exceeded 30 min (Table S1). Furthermore, due to their inability to cure promptly, the scaffolds of nylon 6 and collagen collapsed. Then, they were separately placed in 1% (v/v) sodium hydroxide (NaOH) solution, distilled water, ethanol, and distilled water, and their insolubility in these solvents was used to carry out displacement solidification. It was discovered that the materials could be solidified and molded rapidly (curing time: chitosan ∼90 s; Nylon 6–30 s; Collagen ∼120 s; Cellulose ∼90 s.) (Table S1), and the scaffolds could maintain an excellent shape and structure (Figure 2E).

FIGURE 2.

(A) The photos, (B) structural formulas, (C) hydrogen bonding interactions between molecules, (D) dissolution, and (E) solvent displacement solidification experiments of several common biomedical polymer materials (Chitosan, Nylon 6, Collagen, Cellulose).

2.2. Nanoparticle‐Induced Evaporative Microturbulence

Based on empirical observations from polymer/solvent system studies, the incorporation of inorganic particles was identified as an effective approach to facilitate solvent evaporation. Consequently, this finding informed our second strategy to accelerate slurry solidification through deliberate particle addition. To enhance the biological performance of the scaffold, carbonated hydroxyapatite (HA) was employed as a functional additive. Before implementation, binding energy calculations were conducted to investigate HA's potential influence on ethanol solvent evaporation kinetics. Comparative analysis revealed that the PU‐ethanol binding energy exhibited a fourfold magnitude over the HA‐ethanol interaction energy (Figure 3A). This energy differential provided theoretical justification for incorporating HA to modulate the solidification kinetics of 3D printing paste through optimized solvent evaporation pathways.

FIGURE 3.

(A) The binding energy between PU and ethanol, HA and ethanol. (B) SEM image of HA and PU composite. (C) TEM image of HA and PU composite. (D) Molecular dynamics simulation of different proportions of slurry composite systems, (E) diffusion coefficient of ethanol solvent. (F) FTIR spectrum of carbonyl and peak separation of free carbonyl and hydrogen‐bonded carbonyl, (G) abundance of free carbonyl and hydrogen‐bonded carbonyl, and ratio of hydrogen‐bonded carbonyl. (H) Schematic diagram of hydrogen bonding between HA and PU.

When preparing HA/PU 3D printed slurry, we found that the maximum addition ratio of HA was 45wt.%. This is because the PUH45 slurry exhibited a greater storage modulus (Figure S4), which was directly related to the shape retention of 3D printed parts. This demonstrated that the PUH45 slurry exhibited superior printing fidelity compared to the PUH50 slurry. To investigate this, we observed the internal cross‐sectional morphology of scaffolds printed with the two slurries and found that the HA distribution in the PUH45 scaffold was more uniform, while the HA in the PUH50 scaffold was more prone to agglomeration (Figure S5). HA agglomeration directly impacted the scaffold fidelity. Furthermore, the shear rate‐stress results showed that when the shear rate exceeded 10 s⁻¹, the PUH50 slurry exhibited a rapid increase. By around 50 s⁻¹, the stress of the PUH50 slurry was approximately twice that of the PUH45 slurry. Higher stress meant higher extrusion pressure during printing, which could increase the risk of nozzle clogging or equipment damage. In fact, when the proportion of HA exceeds 50 wt.%, the slurry still retains its printability. However, at this point, the surface of the 3D‐printed scaffold already exhibited the phenomenon of inorganic HA particles falling off (Figure 3B), indicating that the PU was no longer able to completely bind the excessive HA. Then, we immersed the 50 wt.% HA scaffold in PBS for 24 h and dried it. SEM observation revealed swelling of the scaffold and the presence of significant pores on its surface. This is due to the inability of the PU matrix to fully bind the HA particles, causing the HA to swell and fall off in the PBS. In contrast, the 45 wt.% HA scaffold showed no pores (Figure S6). At this ratio, the apatite powder was semi‐wrapped by PU. It did not separate from the PU matrix and could directly contact the tissue. When the apatite content reached 50wt.%, the PU matrix could no longer completely coat the HA particles. Thus, the maximum amount of HA to 45wt.% was set for subsequent past experiments (Figure 3B). From the Transmission Electron Microscope (TEM) of the composite with HA added to the PU/ethanol system (Figure 3C), it could be found that when the proportion of HA was less than 45wt.%, HA was wrapped by PU and could not be connected, forming “islands”. While when the HA content reached 45wt.%, HA particles connected to each other and extended to the surface, providing channels for solvent diffusion, allowing the solvent to form micro‐turbulence and evaporate rapidly. Computational Fluid Dynamics (CFD) modeling was performed to simulate solvent displacement in a representative microchannel containing dispersed nanoparticles. The results demonstrated that the presence of nanoparticles induces significant local velocity gradients and vortical flow patterns within the solvent boundary layer, consistent with “micro turbulence” (Figures S7 and S8).

From the kinetic model of adding HA to the slurry in the PU/ethanol system (Figure 3D), the binding sites between HA and PU were increased. This might have made ethanol more likely to escape from the PU‐ethanol system. Subsequently, the diffusion coefficient of ethanol at room temperature and pressure after adding different contents of HA to PU/ethanol were measured (Figure 3E). As the HA content gradually increased, the diffusion coefficient of ethanol also gradually increased, proving that ethanol in the system was more likely to evaporate.

Owing to the small increase in the infrared extinction coefficient caused by hydrogen bonding, the concentration of the group was proportional to the absorbance of both [23]. In Figure 3F, the interval between 1754–1602 cm^{− 1} was a stretching vibration peak with ─C═O. The high‐wavenumber peak was the free carbonyl absorption peak, and the low‐wavenumber peak was the carbonyl absorption peak forming hydrogen bonds. After area integral calculation (Figure 3G), as the HA content increased, the abundance and proportion of hydrogen‐bonded carbonyl groups decreased. This also proved that adding HA weakened the hydrogen bonding between PU and the ethanol solvent. The reason was that HA formed a hydrogen bonding effect with ─NH─ and ester group (─COO─) in the molecular chain of PU, which destroyed the hydrogen bonding effect between ethanol and ─NH─ and ─COO─ (Figure 3H), reducing the binding force between them. As a result, ethanol molecules were more likely to evaporate, making the slurry easier to solidify and mold.

The negative correlation commonly observed between thermal stability and volatility [24]. So the thermodynamic properties of slurry systems with different HA ratios were subsequently measured. According to the Differential Scanning Calorimetry (DSC) results (Figure 4A), an endothermic peak emerged from approximately 40°C to over 100°C, accompanied by a mass reduction (Figure 4B). This was attributed to the volatilization of free ethanol in the slurry. Moreover, as the HA content increased, both the ethanol volatilization temperature and the final volatilization temperature shifted to lower values. This finding confirmed that HA had an impact on the diffusion coefficient of the slurry system, which was consistent with the result shown in Figure 3F. At 400°C–450°C, an endothermic peak appeared in the DSC curve (Figure 4A) and was accompanied by a mass decrease in the Thermal Gravimetric (TG) curve (Figure 4B). This corresponded to the decomposition of soft long chains, chain extender polyols, and isocyanate in the PU molecular chain. The addition of HA caused the decomposition peak of the composite slurry to shift toward higher temperatures compared to that of pure PU, indicating an increase in the thermal stability of the system. This was because the incorporation of HA particles in the PU matrix affected the arrangement of polymer chains, further leading to changes in the polymer's behavior during heating. Specifically, the HA particles acted as nucleation centers and significantly influenced the rigid phase of the polymer.

FIGURE 4.

(A) DSC, (B) TG, (C) Freidman method kinetic analysis results and (D) volatilization rate of slurry composite systems with different proportions. (E) Schematic diagram of the volatilization of ethanol solvent in the composite system. (F) Changes in viscosity of slurries with different proportions over time, (G) curves of viscosity and stress changing with shear rate. (H) Water absorption of composite materials with different proportions after drying. (I) Schematic diagram of simulation of slurry volatilization with different proportions.

Activation energy is defined as the energy required for a molecule to transition from a stable state to an activated state. The curve in Figure 4C was fitted based on the TG curves at different heating rates. The activation energies of several slurry systems were further calculated according to the slopes of the curve. The values obtained were E_PU = 43.13 kJ mol⁻¹, E_{15wt.% HA} = 59.33 kJ mol⁻¹, E_{30wt.% HA} = 76.08 kJ mol⁻¹, and E_{45wt.% ‍HA} = 108.19 kJ mol⁻¹ respectively. It was found that the addition of HA increased the thermal degradation activation energy of the slurry, and the thermal stability of the system gradually improved, which was consistent with the thermogravimetric results.

After that, the accumulated solvent volatilization at different time points in different slurry systems was measured. By analyzing the fitting curve of the results, it was observed that the addition of HA accelerated the volatilization of ethanol in the system. Notably, the volatilization rate of ethanol in the slurry with a 45wt.% HA content was significantly enhanced (Figure 4D). In the slurry with 45wt.% HA, HA was in a semi‐exposed state, which not only increased surface roughness, but also accelerated the solvent evaporation rate due to the local micro turbulence caused by the mediated diffusion dynamics. The principle was to increase the evaporation surface area of the solvent and provided more evaporation channels (Figure 4E).

DIW allows printing of practically any material, as long as the precursor ink can be engineered to demonstrate appropriate rheological behavior [16]. Based on the result of the change in material viscosity over time (Figure 4F), as the HA content increased, the viscosity of the slurry increased rapidly. This was also due to the rapid evaporation of ethanol. From the rheological properties of the slurry (Figure 4G), it was found that in the initial stage, the viscosity of several slurries decreased sharply with the increase in the shear rate, and all exhibited good “shear thinning” performance. This characteristic facilitated the extrusion molding process in 3D printing. As the shear rate continued to increase, the slurry viscosity tended to stabilize, suggesting that several slurries had good printability. According to the Scanning Electron Microscope (SEM) results in Figure 3B,D, the large‐scale composition of HA not only increased the roughness of the slurry system but also expanded the diffusion and volatilization channels of the solvent. Therefore, a water absorption test was conducted on the solvent‐volatile solid slurry (Figure 4H). The results showed that the slurry with a 45wt.% HA content had the highest water absorption rate, which was twice that of the other three slurries. This result was consistent with the previous experimental findings. All the above results demonstrated that the addition of inorganic HA particles affected the viscosity and rheological properties of the polymer/solvent system. It increased the movement paths of solvent molecules and made them more likely to diffuse to the surface for volatilization. finally, finite element simulations of slurry systems with different HA contents were carried out (Figure 4I). It was found that when 50% of the surface solvent evaporated, it took less than 1 min for the slurry with a 45wt.% HA content, about 3 min for the slurry with a 30wt.% HA content, about 7 min for the slurry with a 15wt.% HA content, and more than 10 min for the pure PU slurry. In the DIW 3D printing process, rapid solidification of the slurry was always necessary to provide mechanical support to the scaffold.

2.3. Topology Path Optimization

Although the addition of HA significantly improved the solidification time of the slurry, when 45wt.% HA/PU slurry was printed, numerous printing defects were still found to occur due to the slurry was not instantaneously solidified, particularly at the corner positions (Figure 5A). There was research on the rapid formation of an oxide layer on the surface of gallery indium compound during printing, which solves the problem of uncontrollable printing corner shape [25]. However, the root cause lay in the untimely solidification of the slurry. During 3D printing, once the first layer was printed and the printer needle was lifted to start the second layer printing, if the first layer slurry at the contact area between the second and the first layer was not fully solidified, the first layer slurry would be displaced because of the traction force of the printer on the slurry wire (Figure 5B). Alternatively, due to the delayed solidification of the slurry, it was difficult to support the upper layer slurry, leading to deformation of the lower layer slurry (Figure 5C) and ultimately resulting in 3D printing defects.

FIGURE 5.

(A) Photos of 3D printed HA/PU composite scaffolds, (B,C) schematic diagrams of printing defects, (D) 3D printing path planning, and (E) volatilization simulation. (F) List of adjacent layer paths and optimal second layer path planning. (G) Model analysis for irregular objects and situations where the angles of adjacent layer printing paths are not perpendicular.

When planning paths for DIW 3D printing, the most common paths were vertical (Figure 5D). Based on the results in Figure 4I, the evaporation of the solvent took a certain period. Therefore, after a layer was printed, in the final part of the printing, the solvent had not completely evaporated and the slurry had not solidified (Figure 5E). Consequently, when choosing the starting point for the second layer printing, locations that were not fully solidified (i.e., the end point) should be avoided. A representative rectangular 3D model was selected for analysis. For the convenience of calculation, the starting point of each layer was set at the limit position (i.e., the vertex of the rectangle). By simplifying the 3D printing path, it was discovered that after the starting point of the first layer printing was fixed at point A (in fact, fixing the starting point at any vertex of the rectangle would yield the same result), there were 4 possible printing path outcomes: vertical printing with an even path; vertical printing with an odd path; horizontal printing with an odd path; and horizontal printing with an even path. The path of the second layer was perpendicular to the first layer, and the wire diameter could be either odd or even. The starting point of the second layer could be any one of vertex 1, vertex 2, vertex 3, or vertex 4 (Figure 5F). By calculating the time differences between the second and the first layers starting from different vertices, the optimal starting point of the second layer could be determined. After calculations (Tables S2 and S3), when the first layer followed Scheme I, the situations where the contact time differences between each point in the second layer and the first layer were optimized were Scheme d and Scheme h. When the first layer followed solution II, the best choices for the second layer were a and e. When the first layer was solution III, the optimal second layer solutions were also a and e. When the first layer was solution IV, the best second layer solutions were b and f. It was evident that regardless of whether the number of paths was odd or even, the starting point of the second layer should be selected at the end of the first layer to guarantee the longest contact interval between the corresponding points of the upper and lower adjacent layers, meaning that the lower layer had sufficient time to solidify. Based on this logic (Figure S9), we developed Python code (Python Code in Supporting Information) to optimize the printing path, maximizing the time interval between each contact point between the upper and lower layers. We input six different parameter combinations (Table S4, Figures S10–S15) and found that these intervals increased by at least 29.8 s, significantly improving slurry curing during 3D printing. For models of other sizes, simply inputting these parameters will result in extended intervals, allowing us to determine the optimal printing path.

For irregularly shaped (such as various organs), they could be expanded into an external rectangle (Figure 5G). According to the results of running the Python code (Supporting Python code), when the second layer path was perpendicular to the first layer, the optimal starting point of the second layer was the diagonal point of the rectangle's vertex closest to the end point of the first layer. When the second layer path made an angle (θ) with the first layer instead of being perpendicular, the optimal starting point of the second layer was the position closest to the diagonal point of the rectangle's vertex nearest to the end point of the first layer.

2.4. Preparation of Programmable Scaffolds

Based on the above three experiences (solvent replacement, inorganic particle addition, and 3D printing path optimization), a series of proportional composites was prepared and 3D printed using synthesized PU and HA. Some of their basic physical properties were characterized (Figure 6A). The intention was to make them play different roles in the field of biomedical engineering according to their characteristics. A series of composites was created by compounding 15wt.% (PUH15), 30wt.% (PUH30), and 45wt.% (PUH45) HA in PU for DIW 3D printing (Figure 6B). It was found that during printing, all slurries undergo solvent displacement and path optimization to achieve rapid curing. By using the above‐mentioned method, the printability of each slurry was significantly enhanced, and they could be molded into various shapes without printing defects. After printing, the pore sizes of the four types of scaffolds were well‐defined. The surfaces of the PU, PUH15, and PUH30 scaffolds were smooth, with no exposed HA. Energy dispersive X‐ray spectroscopy (EDS) results showed that the element content was consistent with the trend of HA addition, and HA was evenly distributed in the PU matrix, without obvious phase separation. TEM images also revealed that HA was in the nanoscale and evenly dispersed in the PU matrix without obvious agglomeration.

FIGURE 6.

(A) Schematic diagram of the preparation process of HA/PU scaffolds; (B) SEM, EDS, cross‐sectional TEM and mineralization on the surface of HA/PU composite scaffolds with different ratios. (C) Compressive strength (n = 3), (D) elastic modulus (n = 3), (E) hardness (n = 5), and (F) surface contact angle with water of the scaffolds. (n = 5). (***) p < 0.001, (**) p < 0.01, (*) p < 0.05.

Due to the differences in HA content, the four types of scaffolds exhibited different mineralization properties. Coarse and well‐crystallized crystals formed on the surface of the PUH45 scaffold, while the other three scaffolds had fewer mineral crystals and smaller grains. Subsequently, the compression performance of each scaffold was tested (Figure 6C,D). Since the soft segments and chain extenders used to synthesize the PU were both diols, the obtained PU was a soft elastomer. When the amount of HA added was small, the main stress‐bearing matrix of the composite was PU, resulting in lower compression strength of the scaffold. With the increase in HA addition, the compression strength and elastic modulus of the scaffold gradually increased. The compression strength and elastic modulus of the PUH45 scaffold reached 1.34 and 25.2 MPa, respectively, which fell within the strength range of cancellous bone [26]. The hardness of the composite scaffold is presented in Figure 6E. When the HA content was low, the soft nature of PU determined the relatively soft hardness of the scaffold. As the HA content increased, the hardness of the scaffold increased significantly (p < 0.001).

By combining SEM images of scaffolds with different proportions, it was found that as the amount of HA added increased, the roughness of the material surface increased significantly, while the contact angle decreased gradually with the addition of HA (Figure 6F). This was because the surface energy of PU was significantly smaller than that of HA, and HA was highly hydrophilic. There was little difference in the surface contact angles between PUH0 and PUH15 scaffolds. At this time, the HA content was very low and completely wrapped by PU, with little exposure on the surface of the trabecular scaffold. When the addition amount reached 45wt.%, a large amount of HA was exposed on the scaffold surface, which significantly affected the wettability of the sample surface (p < 0.001). Scaffolds with different HA contents had different physical properties and good printability, which provided more options when designing tissue reconstruction scaffolds.

2.5. Application Paradigm for Drug Loading and Release in Scaffold

Hydrogen‐rich polymers were found to form stable non‐covalent bonds with drugs. This process increased the solubility and loading of drugs, making them particularly suitable for the delivery of insoluble drugs. The dynamic reversibility of hydrogen bonds also allowed the drug to achieve intelligent response release under different pH, temperature, or ionic strength environments. The molecular structure of desferrioxamine (DFO) contained a large amount of ─CO─, ─OH, ─NH─, and ─NH₂. These groups could form hydrogen bonds with ─CO─, ─O─, and ─NH─ in PU, thus improving the load capacity and bond strength to the matrix. However, in acidic environments (such as inflammatory environments and bone defect areas), ─NH₂, ─NH─, and ─OH were all protonated, leading to the release of DFO and achieving an intelligent response. DFO could activate hypoxia inducible factor‐1α (HIF‐1α) [27], upregulate the cascade of hypoxia responsive genes, resulting in high induction of vascular endothelial growth factor [28]. This further enabled the cells to continue differentiating, promoting angiogenesis and osteogenesis (Figure 7A). DFO was sensitive to temperature and light. It gradually decomposed under light (especially ultraviolet light), which made it unsuitable for high‐temperature and photocuring 3D printing and forming. Therefore, DFO loaded on the PUH45 scaffold (DFO‐PUH45) was chosen for DIW 3D printing and molding.

FIGURE 7.

(A) After DFO was loaded on the PUH45 scaffold, the hydrogen bonding between the drug and the scaffold, the drug release in the acidic environment, and the angiogenesis and osteogenesis mechanism. (B) Surface SEM, (C) cross‐sectional SEM and EDS, (D) TEM (the embedded image is a photo) of DFO‐PUH45 scaffold. (E) FTIR spectrum, high‐resolution (F) N 1s and (G) O 1s spectra of the scaffold. (H) Drug color reaction of the drug‐loaded scaffold and (I) XPS high‐resolution spectrum of S 2p. (J) Micro‐CT reconstruction image of DFO‐PUH45 scaffold. (K) The binding energy of DFO with the polymer. (L) DSC and TG spectrum, and (M) drug release curve of the DFO‐PUH45 scaffold.

After loading DFO, the surface of the PUH45 scaffold did not change significantly and still maintained a high roughness (Figure 7B). The scaffold had good printability, a high degree of molding, and could maintain good pores in the longitudinal section. The elements C, O, Ca, P, and S at the fractures cut by the scaffold were evenly distributed (Figure 7C). The internal porosity of the scaffold was clear, and the porosity was about 64.9% (Figure 7J). It could also be found from the TEM images that HA was nanoscale and uniformly distributed in the PU matrix (Figure 7D), without obvious agglomeration, which proved that DFO was uniformly distributed in the PUH45 matrix. After DFO was compounded to the PUH45 matrix, both the C═O stretching vibration peak (1619 cm^{− 1}) and the N‐H stretching vibration peak (1561 cm^{− 1}) in the composite scaffold were present. This indicated that DFO was successfully grafted to the HA/PU scaffold. Since the electron‐withdrawing group C═O and the electron‐donating groups N─H and ─O─ formed a hydrogen bond, the electron density of C═O and N─H was affected. The electron withdrawing group C═O underwent a blue shift, while the electron‐donating group N─H underwent a red shift (Figure 7E). This also proved that bonding formed between DFO and the PUH45 matrix rather than simple physical mixing. By analyzing the N 1s XPS peak of the scaffold after DFO loading, it was found that the ─NH₂ peak position disappeared after DFO was loaded (Figure 7F). It was speculated that ─NH₂ chemically reacted with the remaining ‐NCO in PU or had a strong hydrogen‐bonding interaction with ─C═O in PUH45, causing the ─NH₂ peak position to disappear. After that, the O 1s peaks of the drug and scaffolds were analyzed (Figure 7G). The O 1s peak abundance of the DFO‐PUH45 scaffold increased, and a slight blue shift occurred compared with the O 1s peak of DFO. This also proved that a bond was formed between DFO and the PUH45 scaffold rather than simple physical mixing.

DFO could complex with Fe^{3 +} to form an orange‐red complex. Therefore, the presence of DFO in the scaffold could be detected using Fe^{3 +}. As shown in Figure 7H, the FBS solution loaded with DFO showed an orange‐red color under the action of FeCl₃, which proved that DFO was successfully compounded into the scaffold and released in FBS. Figure 7I was the S 2p fraction of the DFO‐PUH45 scaffold XPS. Sulfur was a specific element in DFO. The presence of the sulfur element was detected in the scaffold, proving that DFO was successfully compounded into the HA/PU scaffold. However, due to the small amount of added DFO, the abundance was not high. From the DSC‐TG curve (Figure 7L) of the scaffold, it was found that the endothermic peak occurred only at around 280°C, accompanied by a mass reduction. Compared with the PUH45 scaffold, the temperature increased by about 60°C. This was because the ─NH₂, ─OH, and C═O in DFO caused the cross‐linking of PU molecules, thus increasing the melting temperature of the scaffold. Between 280 and 350°C, the endothermic peak was accompanied by a sudden decrease in mass, which was due to the decomposition of PU, the decomposition of carbonate in hydroxyapatite, and the decomposition of DFO. By the end of 500°C, about 45wt.% of the mass of hydroxyapatite crystals remained.

Molecular simulations calculated that the binding energy between DFO and the polymer matrix was as high as 39.6 Kcal mol⁻¹ (Figure 7K), which proved that there were multiple hydrogen bonds between DFO and the polymer matrix [29], which was also confirmed by the FTIR (Figure 7E) and XPS (Figure 7G). The sustained release curve of the DFO‐PUH45 scaffold in PBS is shown in Figure 7M. Within 2 h, the drug showed explosive release, with a release of approximately 35% of the total amount. Subsequently, the DFO in the scaffold showed a slow‐release state until day 28. Judging from the release amount, from day 2 to day 7, the DFO in the scaffold showed a stable release state, with a release of approximately 0.25 µmol per day. The early explosive release was because the incorporated DFO was only physically mixed with the HA/PU matrix and had no bonding, while the scaffold was penetrating and porous, with a large specific surface area, so the drug was easily released. By the later stage, the physically mixed drugs had been basically released, and the drugs with a weak chemical bond to the matrix began to be released, and continued to be released as the scaffold swelled. The large‐scale early drug release and the continuous later‐stage release were conducive to the rapid activation of HIF‐1α in cells around the scaffold, thereby promoting vascularization and assisting in bone reconstruction [30].

Both scaffolds were immersed in serum, and a large amount of protein was adsorbed on their surfaces (Figure 8A). Cell proliferation results also showed that cells on both scaffolds had good proliferation, and neither scaffold was cytotoxic (Figure 8B). The F‐actin fluorescence staining results of bone marrow mesenchymal stem cells (BMSCs) on the scaffolds are presented in Figure 8C. The blue nuclei did not exhibit abnormalities such as malformations. On the first day, the green F‐actin was less prominent because the cells were not fully spread. By day 3, fully spread cells were observed in both groups of scaffolds. Due to the effect of DFO, the green fluorescence in the DFO‐PUH45 group was slightly more intense. Subsequently, the vascular endothelial growth factor (VEGF) and osteocalcin (OCN) were observed via immunofluorescence (Figure 8D). It was found that DFO facilitated the expression of VEGF and OCN. Moreover, since the PUH45 scaffold contained a large amount of HA, its osteogenic ability was also satisfactory. Reverse Transcription‐Polymerase Chain Reaction (RT‐PCR) was used to detect osteogenesis related genes (Runx2, Col‐I, OCN) and vascular‐associated genes (VEGF) (Figure 8E). It was revealed that the loading of DFO had a significant impact on the expression of vascular and osteogenic proteins in the scaffold cells.

FIGURE 8.

(A) Protein adsorption of the scaffold in serum. (B) Cell proliferation after co‐culture of the scaffold with cells (n = 5). (C) Fluorescent staining of F‐actin (green) and nucleus (blue). (D) CLSM images, (E) PCR result of protein expression (n = 3), and (F) Alizarin red staining image. (G) Differential metabolites after co‐culture of the drug‐loaded scaffold with cells, and (H) the affected metabolic pathways. (***) p < 0.001, (**) p < 0.01, (*) p < 0.05.

The generation of mineralized nodules in each group was detected (Figure 8F). It was apparent that the DFO‐PUH45 group showed the darkest color, followed by the PUH45 group, and the control group was the weakest. Hydroxyapatite, an important component of bone tissue with a certain osteogenic inducibility, showed a darker red color than the control group. DFO promoted osteogenesis and differentiation by facilitating vascularization. Additionally, HA promoted osteogenesis, and the combined effect of the two strongly promoted the regeneration of new bone tissue. Later, the relative content of two metabolites, glucose and lactate, was measured. It was found that the amount of these two metabolites in the DFO‐PUH45 group was significantly higher than that in the PUH45 group (Figure 8G). This was because DFO activated the HIF‐1α signaling pathway, which allowed the cells to absorb and metabolize glucose in large quantities. Through the TCA cycle, ATP was produced to provide energy for the vascular and osteogenic differentiation of cells, accompanied by the production of lactic acid by‐products (Figure 8H). Changes in genes and proteins directly led to alterations in cell metabolic pathways. Therefore, cell metabolomics became another important method for evaluating cell phenotypes [31]. According to the results of cell immunofluorescence staining and PCR transcription and expression, BMSCs exhibited different phenotypes on DFO‐PUH45 and PUH45 scaffolds. That is, DFO had a strong promoting effect on the vascular and osteogenic differentiation of BMSCs. Thus, metabolomics was further employed to study cell metabolites and related pathways.

After analyzing the experimental samples of DFO‐PUH45 and PUH45, in the PCA score plots and multivariate models (PLS‐DA and OPLS‐DA) (Figure S16), both the control (PUH45) and experimental (DFO‐PUH45) groups exhibited distinct separation trends, suggesting significant differences in their metabolic profiles. A total of 180 metabolites were detected, among which 2 metabolites were down‐regulated, and 178 metabolites were up‐regulated (Figure 9A). A clustering heat map of differential metabolites (Figure 9B) indicated that Resveratrol, Formononetin, Hypoxanthine, Inosine‐5'‐monophosphate, Uridine, and L‐tryptophan were all metabolites associated with vascularization. 3‐hydroxybutyric acid, Glycerol 3‐phosphate, Ethanolamine, Citric acid, L‐valine, L‐isoleucine, L‐proline, Glycine, Serine, Vitamin D3, Cholesterol, and Pyrophosphate were all products related to osteogenic metabolism. Next, all differential metabolites were classified (Figure 9C), and it was found that these differential metabolites were related to immunity, mineralization, glycolysis, energy metabolism, cell membrane synthesis, nucleic acid synthesis, signaling, anti‐inflammation, antioxidant, protein synthesis, and collagen synthesis. After that, the metabolic pathways affected by differential metabolites were analyzed (Figures 9D, E). It was discovered that the loading of DFO ultimately had a significant influence on protein digestion and absorption and other metabolic pathways. Finally, the affected metabolites and metabolic pathways were sorted out (Figure 9F). It was found that changes in multiple metabolic pathways ultimately interfered with the calcium ion signaling pathway, which played a key regulatory role in biological processes such as cell metabolism, proliferation, differentiation, and apoptosis. Its role in bone metabolism and angiogenesis was particularly important.

FIGURE 9.

(A) Volcano plot of cell differential metabolites between DFO‐PUH45 and PUH45 groups, and (B) relative expression comparison of differential metabolites. (Red indicates upregulation, blue indicates downregulation). (C) Classification of differential metabolites. (D) chord and (E) lollipop plots of differential metabolite enrichment analysis. (F) Analysis of pathways affecting osteogenesis.

Micro‐CT images for 3D reconstruction were taken 4 and 8 weeks postoperatively (Figure 10A). The results indicated that in the PUH45 group and the DFO‐PUH45 group, new bones grew along the scaffold and into the defect area. However, the DFO‐PUH45 group exhibited more abundant and denser new bone tissue. A semi‐quantitative analysis of the new bone amount also revealed that the loading of DFO promoted new bone formation. After the Micro‐CT scan, the femoral condyle defect area (Figure 10A) was reconstructed. Each group demonstrated a favorable growth state of new bone tissue. At 4 weeks, there was no obvious gap between the scaffold of the DFO‐PUH45 group and the original bone tissue, while in the bone defect area, a distinct gap still existed between the scaffold and the tissue. From the cross‐sectional view and the overall three‐dimensional image of the new bone and trabecular bone thickness in the XY, XZ, and YZ directions of the defect area (Figure 10A, Figure S17), it was found that the new bone tissues of all groups grew inside the defect area. The bone mass and trabecular thickness of the DFO‐PUH45 group were superior to those of the PUH45 group, which proved that DFO had a significant osteogenic effect. Based on the semi‐quantitative results of the bone trabecula number, trabecular thickness, and trabecula separation degree (Figure 10B–D), the trabecular thickness in the DFO‐PUH45 scaffold group was significantly better than that in the PUH45 group, and the trabecular bone separation in the DFO‐PUH45 group was smaller. This indicated that the DFO‐PUH45 scaffold had a better osteogenic effect.

FIGURE 10.

In vivo biological evaluation of scaffolds. (A) Micro‐CT reconstruction images of the defect area after implantation into the femoral condyle of SD rats. Semi‐quantitative analysis of (B) new bone mass (n = 3), (C) trabecular thickness (n = 3), and (D) trabecular separation (n = 3). (E) HE staining images of the defect area (black circle is the defect area, "M" marks the material area, green arrow marks the new bone, blue arrow marks the osteoid, and yellow arrow marks the osteoblast). (F) Immunohistochemical staining and (G) semi‐quantitative analysis of proteins related to osteogenesis and angiogenesis in the defect area. (***) p < 0.001, (**) p < 0.01, (*) p < 0.05.

Figure 10E shows the histological section staining images of the femoral condyle defects in rats from each experimental group. The sections were stained with methylene blue and alkaline magenta. Inflammatory cells (such as neutrophils, lymphocytes, macrophages, and plasma cells) were almost absent from the tissues, demonstrating the scaffold's good in vivo biocompatibility. Because the raw materials and synthetic polyurethane were not biodegradable, intact scaffold morphology could still be seen in tissue sections. New bone tissue was dyed dark red, bone‐like substances were stained purple‐gray, osteoblast nuclei were dark blue, and black circles marked the defect area. At 4 weeks, tissues and cells of each experimental group grew into the scaffold, and new bone was mainly distributed around the defect edge and the scaffold, which demonstrated the good osteogenic induction of the two scaffolds. Nevertheless, the new bone trabeculae in the DFO‐PUH45 group were more uniform compared to those in the PUH45 group and the control group. By magnifying the local area, it was observed that in the DFO‐PUH45 group, not only were osteoblasts densely spread on the scaffold, but neovascularization also occurred in the pores of the porous scaffold, which was crucial for later bone reconstruction. A large number of osteoblasts were also found in the PUH45 group, which was attributed to the osteogenic induction effect of hydroxyapatite. The blank control group still had a large defect. By 8 weeks, the tissue had fully grown into the scaffold. There were more dark‐blue osteoblasts in the DFO‐PUH45 scaffold, and more and more uniform new bones had formed. At 8 weeks, neovascularization started to occur inside the PUH45 scaffold. However, the control group still had large regional defects.

The immunohistochemical staining results of tissue sections in the defect area two weeks after surgery are shown in Figure 10F, and the quantitative analysis is presented in Figure 10G. Most of the tissues in the DFO‐PUH45 group had grown into the scaffold, while the PUH45 group and the blank control group still had large defect areas. HIF‐1α was positively expressed in the DFO‐PUH45 group with a high expression level. This was caused by angiogenesis after DFO was released from the scaffold. In contrast, the HIF‐1α expression was lower in the PUH45 group and the control group, which might be due to the short implantation time and the lack of vascular tissue formation around the defect. The expression pattern of the vascular‐associated factor VEGF was similar to that of HIF‐1α. In addition to the high expression of osteogenesis‐related factors OCN, OPN, and Runx2 in the DFO‐PUH45 group, a small amount of these factors was also expressed in the PUH45 group, which might be the outcome of hydroxyapatite‐induced osteogenesis. To assess potential organ damage induced by HA nanoparticles, histological analyses of liver and kidney tissues were conducted across all stent material implantation groups. The liver tissue sections from rat experimental groups (Figure S18) demonstrated relatively intact endothelial architecture with normal cellular arrangement and preserved hepatic capsules, showing no significant pathological alterations. The renal tissue sections from each experimental rat group were analyzed in Figure S19. Histological examination revealed that the renal capsule was intact across all groups, with no observable tissue proliferation or inflammatory reactions.

3. Discussion

The ability to rapidly solidify hydrogen‐bond‐rich polymer inks is critical to achieving structural fidelity in DIW systems. Polymers like PU, chitosan, and collagen rely heavily on intermolecular hydrogen bonds, which—while advantageous for bio‐functionality—also trap solvents within their molecular network, inhibiting rapid phase transition. The solvent displacement strategy exploits the differential hydrogen bonding affinities: water, being more polar and a better hydrogen bond donor/acceptor, preferentially associates with ethanol, displacing it from the polymer‐solvent complex. Simulation and experimental results confirm that water facilitates ethanol removal while inducing polymer precipitation. A similar approach was also used to 3D bio‐print collagen in a precipitation bath containing the recombinant spider silk protein eADF4, thereby obtaining composite materials with excellent mechanical properties [32]. Unfortunately, the dissolution and solvent replacement solidification of hydrogen‐bonded polymers is a very complex process. There is currently no reliable guidance method for reference, and it can only be implemented based on experiments and experience. According to our experience, this process is closely related to the strength of hydrogen bonds of polymer molecules, the polarity of the solvent, the solubility with the replacement solvent, and the binding energy between molecules. If the relevant research continues to deepen and the relevant empirical formula can be obtained, it will surely be a breakthrough for the development of DIW 3D printing.

Nanoparticle addition modifies the ink's microstructure, increasing the diffusion pathways and reducing the energy barrier for solvent evaporation. Endo et al. [33] demonstrated that HA nanoparticles can disrupt polymer‐solvent interactions by forming secondary hydrogen bonds, effectively competing with the solvent molecules and promoting evaporation. FTIR results confirm a decrease in hydrogen‐bonded carbonyl peaks in the presence of HA, indicating reduced solvent retention. Moreover, HA served as a structural modifier. Incorporation of 30–45 wt.% HA increased compressive strength and elastic modulus of the scaffold, consistent with studies using HA/PU composites for bone scaffolds [34]. Enhanced surface roughness and hydrophilicity were also observed, both known to improve cellular adhesion and proliferation [35]. HA is used only for osteogenesis. If the DIW‐printed polymer scaffold is to be used in other tissues, such as skin, heart, etc., then silica [36], zinc oxide [37], nano‐silver [38], etc., can be considered to achieve different functionalities such as antibacterial and immunomodulatory.

Poorly optimized paths can result in shear‐induced collapse, incomplete layer adhesion, and geometrical inaccuracies—particularly when inks require time to stabilize. In our study, we used finite element simulations to model drying time and stress distribution, optimizing layer‐by‐layer deposition by maximizing the time between adjacent layer contacts. Some peers also claim that extrusion‐based print path optimization can reduce deformation and warpage in soft material printing by adjusting extrusion order and delay times [39]. In our experiments, modifying the starting point of each layer and alternating deposition directions minimized contact between freshly extruded and unsolidified regions, significantly improving dimensional fidelity.

Beyond mechanical considerations, functional bioactivity is crucial for scaffold performance in regenerative medicine. Here, DFO, a known hypoxia‐mimicking agent, was incorporated into HA/PU scaffolds via non‐covalent hydrogen bonding. The PU network provides reversible binding sites (─C═O, ─NH─), while HA offers a porous microenvironment conducive to sustained drug release. DFO stabilizes HIF‐1α, leading to the upregulation of VEGF and other angiogenesis‐promoting genes. In our assays, DFO‐loaded scaffolds significantly enhanced VEGF and OCN expression.

For polymer‐inorganic particle composite scaffolds, sterilization should maintain their structural integrity and avoid high temperatures to preserve the bioactivity of the incorporated components. Suitable sterilization methods include ethylene oxide sterilization, gamma irradiation, filtration sterilization, and low‐temperature plasma sterilization. However, gamma irradiation can damage the microstructure and mechanical properties of polyurethane; [40] the viscosity of our 3D‐printed composite slurry is very high, making filtration sterilization difficult; and low‐temperature plasma has limited permeability, and reactive free radicals may oxidize the polyurethane surface layer or alter the surface functional groups of the drug [41]. Therefore, in our study, ethylene oxide sterilization was the preferred method for maintaining the mechanical properties of the scaffold and the functionality of DFO. DFO is known to be sensitive to hydrolysis and oxidation in aqueous environments. Encapsulation within the PU/HA matrix and hydrogen bonding to the polymer backbone can protect DFO from rapid degradation, thereby supporting sustained release. Under physiological conditions (37°C, pH ≈ 7.4), DFO release from the PU/HA scaffold is governed by both diffusion from surface‐accessible pores and gradual polymer relaxation/degradation. Our in vitro release studies in PBS showed a biphasic profile—an initial burst within the first 24 h due to loosely adsorbed drug, followed by a sustained release over ≥28 days, attributable to weak coordination between DFO and surface hydroxyl/calcium sites on HA. This coordinated binding is relatively stable under physiological ionic strength, limiting rapid desorption. The PU matrix exhibited minimal hydrolytic degradation within this period, supporting structural integrity and consistent release kinetics. Previous reports have similarly demonstrated that polyurethane–ceramic composite carriers can maintain small‐molecule release stability for several weeks in simulated body fluids without significant loss of bioactivity of chelating agents such as DFO [42].

It is worth mentioning that the DFO, which has amino, hydroxyl and carbonyl groups in its molecular structure, is only a representative. In theory, any substance that has a hydrogen bond with the matrix in its molecular structure can be grafted to increase its loading rate, delay release, and even achieve pH and temperature responsive release.

The integration of solvent displacement, nanoparticle‐enhanced evaporation, path optimization, and therapeutic loading represents a robust strategy for advancing 3D printed biomaterials. Unlike traditional photopolymer systems, this platform operates at ambient conditions and is inherently biocompatible, making it suitable for cell‐laden constructs and drug‐eluting implants. This strategy also avoids the thermal curing steps commonly used in other DIW processes, making it compatible with bioactive compounds. Our system enables tunable scaffold design: HA content modulates mechanical performance, surface energy, and drug interaction, while path optimization ensures reproducibility. The dual functionality—structural support and biological activation—positions this approach as a promising candidate for personalized bone regeneration therapies.

4. Conclusion

This study resolved the critical challenge of solvent entrapment in hydrogen bond‐mediated DIW 3D printing through an integrated strategy combining solvent displacement, nanoparticle‐enhanced evaporation, and path optimization. The resulting scaffolds achieved exceptional structural precision while utilizing hydrogen‐bond networks as dynamic platforms for controlled drug release. The experimental results demonstrated that after loading DFO with hydrogen bond rich scaffolds, the osteogenesis and angiogenesis activities were significantly enhanced through calcium signaling and HIF‐1α pathways. By establishing fundamental design principles for hydrogen‐bonded polymer processing, this work advances both precision tissue engineering and smart therapeutic delivery. The demonstrated dual functionality of structural control and biological activity opens new possibilities for complex tissue regeneration.

5. Experimental Section

5.1. Material Preparation

5.1.1. Preparation of Polyurethane (PU)

PTMEG and IPDI were weighed in a molar ratio of 1:1 into a 250 mL plastic beaker, which was then sealed with cling film. Subsequently, the plastic beaker was placed in an oil—bath maintained at 70°C. The mixture was slowly stirred and pre—polymerized under nitrogen protection for 2 h. After that, a certain amount of tin salt was added dropwise, and the pre—polymerization reaction was allowed to continue for another 2 h. Next, the chain extender BDO was added, and the chain extension reaction was carried out for 2 h. Following the chain extension reaction, a small amount of ultrapure water, used as a foaming agent, was added, and the mixture was rapidly stirred for 0.5 h. Finally, the beaker was placed in an oven at 90°C for the foaming reaction, and a PU foam elastomer was obtained after molding.

1. Pre‐polymerization:

1. Chain extension:

1. Foaming reaction:

5.1.2. Preparation of 3D Printed HA/PU Scaffold

The prepared HA and PU were composited to prepare the 3D printing paste. After the HA powder was dried, ground, and sieved, it was ultrasonically dispersed in an alcohol solvent. Immediately, the synthesized PU was cut into fragments, dried at 60°C to remove water, and dissolved in an alcohol solvent. Then, under rapid stirring, the HA alcohol suspension was slowly added to the PU alcohol solution. After the drop—wise addition was completed, stirring was continued for 2–3 h, during which the HA powder and PU molecules were thoroughly and uniformly mixed. During the mixing process, the conditions were controlled to achieve a certain viscosity of the slurry and meet the printing conditions (Figure S20).

The prepared slurry was transferred to the syringe equipped in the 3D printer, the air in the syringe was squeezed out, the appropriate needle was selected, the printer program was followed, the required model was selected, and the size was adjusted. The model parameters selected for the PU/HA stent prepared in this experiment were: needle diameter: 210 µm; wire spacing: 500 µm; floor height: 100 µm. The abbreviations for each scaffold were shown in Table S5.

5.2. Material Characterization

5.2.1. Nuclear Magnetic Resonance Spectroscopy

The prepared sample was dissolved in deuterated chloroform (CDCl₃) and subsequently analyzed via ¹H nuclear magnetic resonance (NMR) spectroscopy using a 600 MHz spectrometer.

5.2.2. Chemical Bond Analysis

A qualitative analysis of the chemical bonds in the sample was carried out using a total reflection Fourier Transform Infrared Spectrometer (FTIR). The resolution was set to 4 cm^{− 1}, and the sample was scanned 64 times to acquire the average spectral value. The scanning range was from 4000 to 650 cm^{− 1}. Origin software was utilized to fit the curve, and multiple iterations were performed to obtain the best—fitting Gaussian peak. The maximum error was less than 5%.

5.2.3. Molecular Weight

The number‐average molecular weight (Mn), weight‐average molecular weight (Mw), and molecular weight distribution of the PU samples were measured by gel permeation chromatography (GPC) at 40°C. N, N‐Dimethylformamide (DMF) was used as the mobile phase, and the PU concentration was set at 2 mg mL⁻¹. Monodisperse polymethyl methacrylate was used as a reference sample to obtain a standard curve.

5.2.4. Slurry Solidification Experiment

The polyurethane (PU) was completely dissolved in ethanol and subsequently loaded into a 3D printer cartridge. The material was then extruded through the printer using a pressure‐driven deposition process. After solvent volatilization, the mass of the printed scaffold was quantitatively assessed with an electronic balance at predetermined time intervals.

5.2.5. Molecular Dynamic (MD) Simulations

Molecular dynamic (MD) simulations were applied to calculate the radial distribution function (RDF) and interaction energy between PU and EA molecules with H₂O molecules presence or not, denoted as M1 and M2 systems. For M1 system, the model was comprised of 10 PU chains and 500 EA molecules. For M2 system, the model contained 10 PU chains, 500 EA and 2000 H₂O molecules. All solution components were randomly packed into cubic simulation boxes. All MD simulations were carried out by Forcite module with COMPASS III force field in Materials Studio (MS) 2020. Van der Waals and Coulomb interactions were respectively considered by atom based and Ewald methods with a cut‐off value of 12.5 Å. Equations of motion were integrated with a time step of 1 fs. After energy minimization and anneal process, the system was fully relaxed under periodic boundary conditions for 500 ps in the NPT (p = 1 atmosphere, T = 298.0 K) ensemble using the Nose thermostat and Berendsen barostat, which was long enough for system temperature, potential, and total energy to get stable. After reaching equilibrium state, another 500 ps simulation under NVT ensemble was performed to extract trajectory and data for RDF and interaction energy calculation. The dynamic trajectory for each system was outputted at an interval of 5 ps. The interaction energies Ei were calculated according to the following equation:

$$E_i=E_{total}-E_{PU}-E_{EA}$$ (1)

where E_i is the total energy of the complex structure, E_total is the total energy of the complex structure, E_PU and E_EA are the energies of the PU and EA molecules in the system.

5.2.6. Surface Morphology

The morphology, particle shape, and size of HA were analyzed using Scanning Electron Microscopy (SEM). Prior to observation, the samples underwent a standardized preparation protocol: complete drying followed by ultrasonic dispersion in anhydrous ethanol. A precisely measured HA suspension aliquot was transferred onto a silicon wafer substrate, air‐dried under ambient conditions, and subsequently coated with gold sputtering for 120 s to ensure optimal conductivity. Backscattered electron imaging was performed at an acceleration voltage of 15 kV to capture high‐contrast microstructural details.

For the PU/HA composite scaffolds with varying composition ratios, surface morphological characterization was conducted via SEM under modified parameters. After thorough desiccation of the scaffolds, a multi‐stage gold sputtering process was implemented, comprising four discrete coating intervals totaling 20–30 s to prevent thermal damage to the polymer matrix. Secondary electron detection mode was selected with the acceleration voltage adjusted between 15–20 kV to accommodate the hybrid material's differential electron emission characteristics.

Further structural elucidation of HA crystals was achieved through Transmission Electron Microscopy (TEM). The powder samples were ultrasonically homogenized in anhydrous ethanol to form colloidal suspensions, which were then deposited onto copper TEM grids. Following natural evaporation of the solvent, high‐resolution imaging was executed at 200 kV. Digital Micrograph software facilitated precise interplanar spacing calibration and diffraction pattern analysis, enabling crystallographic verification of the hydroxyapatite phase.

5.2.7. Elemental and Surface Chemical Energy Analysis

Elemental characterization was performed using energy‐dispersive spectroscopy (EDS) for qualitative and semi‐quantitative analysis of sample composition. Surface chemical analysis was subsequently conducted through X‐ray photoelectron spectroscopy (XPS) to determine elemental species and their relative concentrations. The XPS full‐scan parameters were configured as follows: 80 eV pass energy, 110‐s acquisition duration, aluminum anode (150 W X‐ray source), 1000.0 meV energy step size, and 100 ms dwell time per data point. Spectral processing involved XPS Peak Fit 4.1 software implementation, where individual peaks were modeled using Gaussian‐Lorentzian line shapes. Shirley background subtraction was systematically applied prior to calculating each component's percentage contribution through peak area normalization.

5.2.8. Thermal Stability and Composition

The crystallization behavior of HA was investigated through a combined analytical approach utilizing Differential Scanning Calorimetry (DSC) and Thermogravimetric Analysis (TG). Experimental measurements were conducted under nitrogen protection atmosphere with a controlled nitrogen flow rate of 20 mL/min and a heating rate of 10°C/min. During the thermal scanning process spanning from 25°C to 1000°C, the instrumentation simultaneously recorded both thermal transition profiles and mass variation patterns of the specimen.

5.2.9. Slurry Viscosity

After preparing the PU/HA slurry, the viscosity of the slurry was tested using the rotation method. First, estimate the viscosity range from large to small and select the appropriate rotary drum. Subsequently, the rotary drum is placed in the slurry, and the viscous torque of the slurry acts on the rotary drum. The dynamic viscosity can be calculated by measuring the torque:

$$\eta =\frac{AM}{n}$$ (2)

η: Dynamic viscosity of slurry, Pa·s; A: Constant; M: The viscous torque of the slurry acting on the drum, N·m; n: Rotation speed of the drum, rad/s.

Start the viscometer, wait for the reading to stabilize, repeat 3 times, and ensure that the maximum deviation of the average value does not exceed 1.5%.

5.2.10. Wettability

Prior to the printing process, slurries with varying mass ratios were uniformly spread on a smooth glass substrate and dried to form continuous composite films; subsequently, the static contact angle at the film‐water interface was quantitatively determined using a standardized contact angle goniometer with ultrapure water as the test liquid (measurement temperature: 25 ± 1°C)

5.2.11. Compression Performance

Rectangular prism specimens measuring 10 × 10 × 30 mm³ were fabricated using additive manufacturing technology. Uniaxial compressive loading was subsequently applied along the longitudinal axis of each specimen through a servo‐hydraulic universal testing machine (UTM), with displacement control set at 2 mm min⁻¹. Continuous data acquisition was implemented during mechanical testing to record complete stress‐strain curves. Compressive strength values were determined from the peak stress points in these curves, while elastic modulus calculations were performed using linear regression analysis on the initial linear portion (typically within 0.1%–0.3% strain range) of the stress‐strain profiles.

5.2.12. Hardness

The surface hardness of the sample was rapidly assessed using a standardized Shore D durometer. First, the slurry was cast into a customized steel mold to fabricate rectangular prism specimens with standardized dimensions of 50 × 50 × 6 mm. Following complete curing at ambient conditions (25 ± 2°C, 60 ± 5% RH), the specimens were subjected to hardness measurement through the following procedure: The durometer's pressor foot was vertically applied to the flat surface with 1 kg test force, maintaining contact for 15 s until the indicator stabilized. To ensure statistical reliability, this testing protocol was systematically implemented across five replicates per material group, with each measurement conducted at three geometrically equidistant positions on specimen surfaces.

5.2.13. Water Absorption Rate

The sample with dimensions of 8 mm in diameter and 1 mm in thickness was oven‐dried at 50°C until reaching constant weight. After cooling to ambient temperature in a desiccator, the sample was weighed and recorded as M₀. Subsequently, the sample was fully immersed in distilled water and subjected to vacuum degassing to ensure complete infiltration of water into the internal structure. Following 24‐h incubation at 37°C under controlled conditions, the sample was carefully removed, with residual surface moisture thoroughly blotted using filter paper. The saturated mass (M₁) was immediately measured through gravimetric analysis. The water absorption rate was calculated according to the following formula:

$$\eta =\frac{M_1-M_0}{M_0}\times 100\%$$ (3)

η: Water absorption rate, %. M₁: Mass after water absorption, g. M₀: Quality of drying scaffold, g.

5.2.14. Rheological Properties

The rheological characterization of the printing slurry was conducted using a rotational rheometer equipped with parallel plate geometry by first configuring the measurement system with matched parallel plates (40 mm diameter) maintaining a fixed gap distance of 1.0 mm, then carefully loading the test specimen onto the rheometer's lower plate using a precision syringe under ambient temperature conditions (25 ± 1°C), fixing the frequency domain parameters at 1 Hz for dynamic testing, and executing a controlled shear rate ramp protocol, systematically increasing the shear rate from 0.1 s^{− 1} to 200 s^{− 1} to establish the complete flow behavior profile.

5.2.15. Internal Structure and Porosity

The internal structure of the stent (Φ13 × 1 mm) was reconstructed by Micro‐CT scanning, and the porosity was calculated based on the reconstructed images obtained by Micro‐CT.

5.2.16. Mineralization Performance

The scaffolds were immersed in a 5‐fold concentrated static simulated body fluid (5× SBF) for 24 h to investigate their mineralization potential. Following the incubation period, the specimens were carefully retrieved from the solution and underwent sequential post‐treatment procedures: [1] thorough rinsing with ultrapure water to remove residual ions and loosely adsorbed precipitates [2], lyophilization under vacuum conditions to preserve the mineralized microstructure, and [3] morphological characterization using scanning electron microscopy (SEM) to evaluate the surface mineralization patterns.

5.2.17. Protein Adsorption

A 5% (v/v) calf serum‐PBS solution was prepared and its baseline absorbance (A₀) was measured at 562 nm using ultraviolet‐visible spectroscopy. The scaffolds were subsequently immersed in the prepared serum solution and incubated for 24 h under physiological temperature conditions (37°C). Following incubation, the residual solution was collected through sterile filtration, and its absorbance (A_x ) was determined using the same spectroscopic parameters. The adsorption rate of the scaffold on the protein is:

$$\eta =\frac{A_0-A_x}{A_0}\times 100\%$$ (4)

For morphological characterization, the protein‐adsorbed scaffolds were subjected to sequential processing: (1) rinsed with physiological saline (0.9% NaCl) for 2 min under gentle agitation (50 rpm) to remove non‐adsorbed proteins; (2) fixed in 2.5% glutaraldehyde solution for 30 min at 4°C to stabilize protein conformation; (3) dehydrated through graded ethanol series prior to freeze‐drying. The processed specimens were then sputter‐coated with gold‐palladium alloy and imaged using scanning electron microscopy (SEM) at 15 kV accelerating voltage.

5.2.18. Drug Release

The printed scaffold (specification: Φ10 × 1 mm) was immersed in a sealed polyethylene (PE) tube containing 2.5 mL phosphate‐buffered saline (PBS), followed by uniform shaking in a 37°C constant temperature incubator. Sample collection was performed at predetermined time intervals: 0.5, 1, 2, 4, 6, 8, 12, 24, 48, 96, 168, 336, and 960 h.

At each time point, 100 µL of supernatant was carefully aspirated using calibrated micropipettes and immediately replaced with an equivalent volume of fresh PBS solution to maintain constant immersion conditions. All collected samples were stored at −20°C for subsequent analysis.

Prior to measurement, 5 µL of 0.67% (w/v) ferric chloride solution prepared in 1% (v/v) hydrochloric acid (FeCl₃/HCl) was added to each thawed sample. The absorbance values were quantitatively determined at 485 nm wavelength using a microplate reader (ELISA‐based detection system). The concentration of deferoxamine (DFO) in each sample was calculated through interpolation against a pre‐established standard curve generated from DFO solutions of known concentrations.

$$C_x=\frac{A_x\ast C_s}{A_{\mathrm{s}}}$$ (5)

C_x : The concentration of the drug to be tested; C_s : Standard sample concentration; A_x : The absorbance of the drug to be tested; A_s: Standard sample absorbance.

5.2.19. Drug Loading Rate

The 3D‐printed scaffold possessing geometric parameters of Φ10 × 1 mm was gravimetrically measured. Subsequently, the specimen was quantitatively transferred into 10 mL of anhydrous ethanol to achieve complete dissolution. The DFO concentration was determined through extrapolation on the pre‐established standard curve constructed using DFO solutions with certified concentrations, with the resultant concentration designated as “C”. The drug loading rate is:

$$Drugloadingcapacity=C\times 1010mL/M$$ (6)

$$\begin{array}{ccc}\hfill Encapsulationrate& =& drugloading/theoreticaldrugloading\hfill \\ & & \times 100\%\hfill \end{array}$$ (7)

5.3. In Vitro Biological Evaluation

5.3.1. Preparation of Scaffold Extraction Solution

The PUH45 and DFO‐PUH45 scaffolds were subjected to immersion in α‐MEM complete culture medium under controlled conditions (37 ± 2°C) for a 72‐h incubation period to generate scaffold extraction solutions.

5.3.2. Cell Culture

The cylindrical scaffold (Φ10 × 1 mm) underwent sterilization via hydrogen peroxide low‐temperature plasma treatment prior to being immersed in α‐minimum essential medium (α‐MEM) for 24‐hour equilibration. Third‐passage bone marrow‐derived mesenchymal stem cells (BMSCs) were subjected to enzymatic digestion using 0.25% trypsin‐EDTA solution. Upon reaching 80%–90% confluence, digestion was terminated by adding an equal volume of complete culture medium containing 10% fetal bovine serum. The cell suspension was subsequently adjusted to a density of 1 × 10⁶ cells mL⁻¹ through centrifugation and resuspension protocols. Finally, the conditioned scaffold was co‐cultured with 1 mL of prepared cell suspension in ultra‐low attachment 6‐well plates maintained at 37°C with 5% CO₂ humidified atmosphere.

5.3.3. Cell Morphology and Proliferation

CCK‐8 proliferation assay: Cells were plated in 24‐well culture plates at 2 × 10⁴ cells mL⁻¹ (1 mL well⁻¹). At designated time points (days 1, 3, 7 post‐seeding), the metabolic activity was quantified by: a) Adding 100 µL CCK‐8 reagent per well b) Incubating at 37°C (5% CO₂) for 2 h) Transferring 200 µL aliquots to 96‐well plates d) Measuring absorbance at 450 nm using a microplate reader.

5.3.4. Fluorescence Staining Protocol

The fluorescence staining protocol involved culturing BMSCs in the extraction medium for 3 days, after which the culture medium was carefully aspirated and the cells were rinsed three times with phosphate‐buffered saline (PBS, pH 7.4) using 1 mL per wash; subsequent cellular fixation was performed by incubating the cells with 4% paraformaldehyde (w/v in PBS) at room temperature for 30 min, followed by three additional PBS washes; actin cytoskeleton staining was then conducted by adding a 5 µg/mL phalloidin conjugate solution (sufficient to cover the well bottom) and incubating in darkness for 60 min, with post‐staining washing comprising three PBS cycles (2 mL/wash, 5 min each) under gentle orbital shaking; nuclear counterstaining was achieved through the application of DAPI solution (0.1 µg mL⁻¹) via bottom‐loading of the well plate, followed by a 15 min dark incubation; finally, after a concluding PBS washing step (three cycles of 5 min each), the specimens were mounted with anti‐fade medium and immediately imaged using laser scanning confocal microscopy with an oil objective and appropriate filter sets.

5.3.5. Calcified Nodule Staining

Following a 14‐day incubation period of bone marrow mesenchymal stem cells (BMSCs) in osteogenic induction medium, the culture medium was carefully aspirated and discarded. Subsequently, the adherent cells underwent three sequential washing cycles with sterile distilled water (5 min per wash) to remove residual medium components. For cellular fixation, freshly prepared 4% paraformaldehyde (PFA) in phosphate‐buffered saline (PBS) was applied to the cultures under 4°C conditions for 30 min, followed by another triple washing procedure with distilled water. Mineralized matrix deposition was histochemically detected through incubation with 0.2% Alizarin Red S solution for 5 min at room temperature. After final removal of unbound dye via three distilled water rinses, calcium‐rich nodules were visualized under phase‐contrast microscopy and digitally documented using standardized imaging parameters.

5.3.6. Osteogenesis‐Related Gene Testing

The expression of osteogenic‐related genes (Type I collagen, Col‐I; Osteocalcin, OCN; Runt‐related transcription factor 2, Runx2; Vascular endothelial growth factor, VEGF) in cells was analyzed through quantitative real‐time polymerase chain reaction (qPCR) according to the following protocol: cells were seeded onto the sample surface at a density of 2 × 10⁴ cells mL⁻¹ and cultured in a 5% CO₂ humidified incubator at 37°C for 3 and 7 days, respectively, followed by total RNA isolation using TRIzol reagent and quality assessment by measuring optical density (OD) values at 260/280 nm using a spectrophotometer; first‐strand cDNA was then synthesized using the Geneseed Reverse Transcription System in a 20 µL reaction mixture containing 2 × RT Mix (10 µL), Reverse Transcriptase (1 µL), Reverse Transcription Primer (2 µL), and 1 µg of total RNA, with thermal cycling conditions set as 25°C for 10 min (primer annealing), 42°C for 15 min (reverse transcription), and 85°C for 5 min (enzyme inactivation); qPCR amplification was performed using Geneseed SYBR Green Master Mix in a 20 µL system containing SYBR Green Mix (10 µL), Forward/Reverse primers (0.5 µL each, 10 µM), 50 × ROX Reference Dye (0.4 µL), cDNA template (2 µL), and sterile distilled water (6.6 µL) to final volume, with amplification parameters including an initial denaturation step at 95°C for 10 sec and 40 cycles of denaturation at 95°C for 10 s and annealing/extension at 60°C for 34 s; post‐amplification dissociation (melt curve analysis) was conducted with the following thermal cycling: 95°C for 15 s, 60°C for 60 s, and 95°C for 15 s, with fluorescence signals collected during the final annealing step at 60°C. The primer sequences used in the experiment are shown in Table S6.

5.3.7. Metabolomics

The metabolic profiling of BMSCs co‐cultured with PUH45 and DFO‐PUH45 scaffolds for 2 weeks was initiated by rapidly quenching the outer wall of the culture plate with liquid nitrogen, followed by the addition of a 20 µL aliquot of internal standard solution (L‐2‐chlorophenylalanine, 0.06 mg mL⁻¹ in methanol) and 1 mL of ice‐cold methanol: ultrapure water (4: 1, v/v) to each sample, after which the cell suspension was transferred to a glass vial; chloroform (200 µL) was then added to each vial, thoroughly mixed, and the cells were lysed using an ultrasonic homogenizer for 6 min, followed by transferring the homogenized mixture to a centrifuge tube and subjecting it to ultrasound‐assisted extraction in an ice‐water bath for 20 min, after which the extract was centrifuged at 13 000 rpm (4°C, 10 min), a 400 µL aliquot of the supernatant was transferred to a glass vial, and the solvent was lyophilized using a vacuum centrifugal concentrator; to the dried sample, 80 µL of methoxyamine hydrochloride pyridine solution (15 mg mL⁻¹) was added, vortexed for 2 min, and incubated at 37°C for 90 min to complete oximation, followed by the addition of 50 µL of BSTFA (with 1% TMC) and 20 µL of n‐hexane along with 10 µL of a mixed internal standard solution (C8‐C24 fatty acid methyl esters in chloroform), vortexing for 2 min, and derivatization at 70°C for 60 min; the derivatized samples were equilibrated at room temperature for 30 min prior to GC‐MS analysis, which was performed using helium carrier gas (99.999% purity) at a constant flow rate of 1 mL min⁻¹, with the injector temperature maintained at 260°C and a 1 µL splitless injection, and the mass spectrometer operating under the following parameters: ion source temperature at 230°C, quadrupole temperature at 150°C, electron impact energy at 70 eV, and full‐scan mode (m/z 50–500); quality control (QC) samples were analyzed every 9 injections to monitor system stability, followed by importing the raw data into MS‐DIAL for preprocessing, with the resulting data matrix analyzed in RStudio using Principal Component Analysis (PCA) to assess sample clustering and process stability, and Orthogonal Partial Least Squares Discriminant Analysis (OPLS‐DA) employed to identify group‐specific metabolites, where variables with VIP >1.0, p < 0.05 (two‐tailed Student's t‐test), and fold change ≥1.2 or ≤0.67 were selected as significantly differential metabolites, and metabolite identification was performed using the Lumingbio GC‐MS LUTarget database, NIST database, and KEGG pathway references.

5.4. In Vivo Biological Evaluation

All the experiments were approved by the Animal Experimental Ethics Review Committee of Kunming Medical University (approval number: KMMUX 202412011).

5.4.1. Surgical Procedure

The PUH45 and DFO‐PUH45 cylindrical scaffolds (Φ4×4 mm) underwent low‐temperature plasma sterilization prior to implantation. A cohort of specific pathogen‐free (SPF) Sprague‐Dawley rats (approximate body weight 300 g; sex‐unrestricted) was acclimatized under controlled environmental conditions (25 ± 1°C) with ad libitum access to food and water. Following a 24‐h preoperative fasting period, the animals were anesthetized through sequential administration of diethyl ether (induction) and sodium pentobarbital (30 mg kg⁻¹, maintenance).

The surgical procedure commenced with preoperative preparation, in which the hindlimb surgical site was depilated and followed by triple‐cycle iodophor disinfection; subsequent surgical exposure involved making a longitudinal skin incision over the distal femur and blunt dissection of muscular layers to expose the femoral condylar surface, after which osteochondral defects (4 mm diameter × 4 mm depth) were created using a trephine drill under saline irrigation, followed by scaffold implantation where sterilized cylindrical scaffolds were press‐fit into the defects according to a cross‐over grouping design (material‐implanted vs. blank control groups), and finally, the wound was closed through layered suturing of the muscle fascia and cutaneous layers, with postoperative iodophor disinfection and ear tagging for identification.

Intramuscular penicillin G (50 000 IU kg⁻¹) was administered daily for 3 days to prevent infection, with continuous monitoring of vital parameters. Animals were stratified into two experimental cohorts based on implant status, including an ultrasound intervention subgroup.

At predetermined endpoints (4 and 8 weeks post‐implantation), euthanasia was induced via sodium pentobarbital overdose (150 mg kg⁻¹). Target tissues (liver, kidney, femoral specimens) were immediately dissected and immersion‐fixed in 4% paraformaldehyde (pH 7.4, 4°C) for subsequent histomorphometric and biochemical analyses.

5.4.2. Microcomputer Tomography Analysis

Femoral specimens were scanned using an X‐ray micro‐computed tomography (µCT) system (SkyScan 1272, Bruker) at 70 kV/142 µA. A cylindrical region of interest (ROI, 4 mm diameter × 4 mm height) corresponding to the defect area was isolated using NRecon software (v1.7.4.2), followed by 170‐layer tomographic reconstruction. Three‐dimensional renderings generated through CTVox (v3.3.0) enabled structural evaluation of the scaffold architecture and bone regeneration patterns. Semi‐quantitative morphometric analysis quantified mineralized tissue volume (mg HA/cm³), trabecular thickness (sphere‐fitting algorithm) and separation (3D distance transformation), with additional assessment of scaffold‐bone interface connectivity.

5.4.3. Histological Analysis

The femoral condyle, liver, and kidney tissues underwent standardized histological processing: sequential dehydration through graded ethanol series (70%–100%), xylene clearing, paraffin embedding, and sectioning at 5 µm thickness. Sections were deparaffinized with xylene, stained with hematoxylin and eosin (HE), re‐dehydrated through ethanol series, cleared with xylene, and mounted with neutral balsam. Microscopic observation was conducted, with subsequent semi‐quantitative analysis of relative bone area performed using ImageJ software through standardized grayscale thresholding protocols.

For specialized staining: After standard sample pretreatment, tissues were successively stained with 0.1% toluidine blue for 30 min and methylene blue/acid fuchsin solution for 15 min. Slides were rinsed with phosphate‐buffered saline and mounted with aqueous mounting medium prior to microscopic examination.

Immunohistochemical procedures included: Tissue fixation in 4% paraformaldehyde, turpentine oil‐based clearing, and paraffin embedding. Antigen retrieval was performed using citrate buffer at 95°C for 20 min, followed by blocking with 10% normal goat serum for 1 h at room temperature. Primary antibodies against HIF‐1α, VEGF, Runx2, OCN, and OPN were applied and incubated at 4°C for 12 h. Secondary antibodies were subsequently incubated for 1 h at 37°C. DAB chromogenic development was monitored microscopically, followed by hematoxylin counterstaining, final ethanol dehydration series, and xylene‐based mounting.

5.5. Statistical Analysis

The experimental data were cleaned, missing values were handled, and outliers were assessed using Z‐scores and boxplots. Data were expressed as mean ± standard deviation (SD). The sample size (n) was explicitly stated in the figure caption for each analysis. Statistical analysis was performed using SPSS software, with one‐way ANOVA followed by Tukey's post hoc test for multiple comparisons. Statistical significance thresholds were defined as follows: p values < 0.05 (*), < 0.01 (**), and < 0.001 (***) were considered to indicate statistically significant, highly significant, and extremely significant differences, respectively.

Conflict of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: advs72398‐sup‐0001‐SuppMat.docx.

Supporting File 2: advs72398‐sup‐0002‐VideoS1.mp4.

Supporting File 3: advs72398‐sup‐0003‐VideoS2.mp4.

Data Availability Statement

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