A programmable shape memory bioscaffold incorporating doxorubicin for postoperative chemotherapy and enhanced bone regeneration

Abstract

The clinical management of bone cancers remains challenging due to the risk of residual tumour cells following surgical resection and the associated complications of systemic chemotherapy. To address this, we developed a shape memory scaffold (SMS) with ion-responsive functionality, designed to enable localised drug delivery and promote bone regeneration postoperatively. Constructed from silk fibroin, carboxymethyl chitosan, alginate, and montmorillonite, the optimised rSCAM scaffold exhibited a porous microarchitecture (93.0 ± 1.2%) and toughness (10.6 ± 0.5 MJ m⁻³), along with rapid shape recovery triggered by bicarbonate ions. rSCAM sustained doxorubicin release over 14 days and showed high drug loading efficiency. In vitro, the scaffold was highly biocompatible, with over 80% cell viability, low ROS induction, and potent anti-osteosarcoma activity, achieving complete tumour cell inhibition within 12 hours. It also supported MC3T3-E1 proliferation, migration, and osteogenic differentiation, as evidenced by elevated MMP2/MMP9 expression, TEER enhancement, and Alizarin Red S staining. CAM assay confirmed neovascularisation and scaffold–host tissue integration. These results demonstrate the multifunctionality of SMS rSCAM as a promising strategy for minimally invasive treatment of postoperative bone cancer, integrating chemotherapeutic delivery and tissue regeneration within a single therapeutic platform.

---

1. Introduction

Cancer treatment has seen considerable advances over the past decades, with an increasingly multimodal approach being employed to manage various malignancies.¹ Among these, bone cancer, particularly osteosarcoma, presents unique clinical challenges due to its aggressive nature and anatomical location.² Surgical intervention remains the cornerstone of primary tumour management, especially for solid tumours. In this context, minimally invasive laparoscopic surgery has emerged as a highly effective and increasingly preferred technique, offering numerous advantages over traditional open surgery.³ These include reduced postoperative pain, shortened recovery time, decreased risk of infection, and improved cosmetic outcomes. Moreover, laparoscopic approaches facilitate greater surgical precision, enabling effective excision of tumours while preserving surrounding healthy tissues. Such benefits have made it particularly attractive in managing bone tumours located in accessible anatomical regions.³

Despite its clinical benefits, the surgical resection of bone tumours often falls short in achieving complete oncological clearance.⁴ Microscopic remnants of malignant cells may persist at the tumour margin, posing a substantial risk of local recurrence or even distant metastasis.⁴ As such, postoperative chemotherapy is routinely administered to eradicate residual cancer cells and prevent disease progression.⁵ However, systemic chemotherapy is frequently associated with a wide array of adverse effects, ranging from nausea and immunosuppression to cardiotoxicity and bone marrow suppression. These side effects are often debilitating and diminish the patient's quality of life, necessitating the development of strategies that can maintain therapeutic efficacy while minimising systemic toxicity.⁶

One promising solution to this challenge lies in the use of localised or targeted drug delivery systems.^7,8 By concentrating chemotherapeutic agents directly at the site of tumour resection, such systems can mitigate systemic exposure and reduce adverse effects, thereby improving patient compliance and treatment outcomes. Among various drug delivery platforms, biocompatible scaffolds have garnered significant interest owing to their ability to simultaneously function as structural implants and local drug depots.⁹ These scaffolds can be engineered to release anticancer agents such as doxorubicin (DOX) in a controlled and sustained manner, ensuring effective cytotoxicity against residual tumour cells at the surgical site without adversely affecting surrounding healthy tissues.

Nevertheless, the clinical translation of drug-loaded scaffolds for bone cancer therapy remains constrained by a key technical challenge: the need to implant the scaffold with minimal invasiveness while preserving its structural integrity. Implantation through narrow surgical access pathways often necessitates mechanical compression of the scaffold, which may compromise its architecture or functionality.^10,11 A feasible strategy to overcome this limitation is the development of shape memory scaffolds (SMSs) based on polymeric networks capable of undergoing reversible deformation.¹² These materials can be compressed into a temporary shape for insertion and subsequently recover their original form upon exposure to physiological stimuli. Alginate-based hydrogels have been extensively studied as ion-responsive systems due to their ability to form reversible “egg-box” crosslinking structures with divalent calcium ions, providing an ideal basis for shape memory behaviour.¹³ Recent studies have demonstrated the feasibility of alginate-based SMSs for tissue engineering applications, confirming their responsiveness and recoverability within biologically relevant environments.^14,15 The use of alginate-based shape memory systems has proven highly suitable for biomedical applications, not only due to the inherently biocompatible nature of this polysaccharide but also owing to the physiological relevance of the ionic triggers employed in their actuation. Divalent cations such as calcium, magnesium, and zinc were commonly utilised to stabilise the temporary conformation, while recovery of the original shape was typically initiated by the introduction of competing anions such as bicarbonate and phosphate.¹⁶ These ions at the concentrations employed are integral to maintaining homeostasis and electrolyte balance in biological systems, thereby minimising the risk of adverse cellular or systemic responses.^16,17

Beyond localised chemotherapy, regenerating bone tissue following tumour resection is critical to restoring both function and structural integrity. An ideal scaffold must therefore not only deliver therapeutic agents effectively but also support bone healing. To this end, composite scaffolds incorporating naturally derived biopolymers and functional nanomaterials have shown considerable promise. Silk fibroin (SF) is renowned for its mechanical robustness, biocompatibility, and ability to support osteogenic differentiation.¹⁸ Carboxymethyl chitosan (CMCs), a water-soluble chitosan derivative, has demonstrated immunomodulatory and wound healing properties, making it highly suitable for tissue regeneration.¹⁹ Meanwhile, montmorillonite (MMT), a layered silicate nanoclay, has been employed to enhance mechanical strength, increase the surface area, and facilitate cellular migration and proliferation.²⁰ To stabilise the polymer network and reinforce its shape memory function, a dual crosslinking strategy employing glutaraldehyde (GTA) can be adopted to create a double-network hydrogel matrix, wherein reversible ionic crosslinking is supported by covalent bonds.

Taken together, this study aimed to synthesise and characterise a novel shape memory scaffold system capable of multi-functionality: shape memory effect upon ion-triggered activation, localised doxorubicin delivery for postoperative chemotherapy, and osteoinductive potential for enhanced bone regeneration. The SMSs were engineered using SF, CMCs, ALG, and MMT and crosslinked via both ionic (Ca²⁺-mediated) and covalent (GTA-mediated) mechanisms. This study was built upon the previous work by our group, both published and unpublished, to refine the composition and synthesis of scaffold materials.^10,21–23 Although the foundational optimisation of component concentrations was inherited from earlier efforts, the present research still systematically characterised and reported the specific contributions and effects of individual constituents in order to maintain scientific continuity. The primary objective was to engineer and fine-tune the microarchitecture of various scaffold formulations and to identify the optimal structure for implantation in bone tissue environments. To achieve this, a series of analytical assessments were conducted on each candidate SMS, including morphological evaluation, chemical and crystallographic characterisation, porosity measurements, swelling behaviour, and compressive strength testing. Beyond these structural and physicochemical properties, the biocompatibility of the SMSs was examined using both in vitro and in ovo models to confirm the safety of the material in biological contexts. The scaffolds were further assessed for their drug-loading capacity and ability to achieve controlled release, with a particular focus on their cytotoxic effect against osteosarcoma cells, demonstrating the scaffold's therapeutic potential for postoperative chemotherapy. Additionally, the SMSs were evaluated as platforms for supporting the growth of preosteoblast cells, including their proliferation, migration, and osteogenic differentiation. These features are critical to achieving successful bone regeneration following the resection of tumour-infiltrated bone tissue. Taken together, the findings of this research highlighted the potential of this multifunctional scaffold system for modern oncological treatment regimens. Its ion-activated shape memory behaviour allowed for minimally invasive implantation, while its sustained drug release capability and osteoinductive properties collectively made the material a compelling candidate for post-surgical interventions that require both localised cancer therapy and structural bone reconstruction, as illustrated in Scheme 1.

Scheme 1 Schematic depiction of the proposed therapeutic strategy for post-operative bone cancer management, involving the use of a minimally invasive endoscopic approach to alleviate pain and accelerate recovery. (A) The shape memory scaffold is inserted into the post-resection bone cavity via keyhole surgery, where it can recover its original shape and locally release doxorubicin to ensure complete eradication of residual tumour cells. Subsequently, the bioscaffold functioned as a regenerative platform for bone tissue reconstruction, enabled by the biocompatibility of the silk fibroin-based matrix. (B) Overview of the multifunctional characteristics of the shape memory scaffold fabricated from natural materials, including silk fibroin, carboxymethyl chitosan, alginate, and montmorillonite, as developed in the present study.

2. Materials and methods

2.1. Materials

Cocoons of Bombyx mori, harvested on the fourth day following silkworm pupae removal, were sourced from Lam Dong, Vietnam. Carboxymethyl chitosan (CMCs; degree of carboxymethylation ≥90%, molecular weight 240 kDa) was acquired from Shanghai Macklin Biochemical Co., Ltd (Shanghai, China). Sodium alginate (ALG; viscosity 398 cps) was procured from Duchefa Biochemie B. V. (Haarlem, The Netherlands). Montmorillonite (MMT) was generously provided by Kunimine Industries Co., Ltd (Tokyo, Japan). Glutaraldehyde (GTA, 50 wt%) was supplied by Shanghai Zhanyun Chemical Co., Ltd (Shanghai, China). Doxorubicin hydrochloride (DOX, 2000 μg mL⁻¹) was obtained from Bidiphar (Binh Dinh, Vietnam). Additional reagents, including calcium chloride (CaCl₂, ≥99.7%), sodium bicarbonate (NaHCO₃, ≥99.7%), and phosphate-buffered saline (PBS, pH 7.4), were purchased from Sigma-Aldrich Co. LLC (Missouri, United States).

2.2. Preparation of SMSs

SF was extracted and purified from Bombyx mori cocoons following a previously established protocol, yielding an aqueous SF solution.^10,24 Separately, CMCs and ALG solutions were prepared by dissolving the respective polymers in distilled water. MMT powder was subsequently dispersed in distilled water using an ultrasonic homogeniser for 30 minutes. These solutions were then mixed and stirred thoroughly at ambient temperature for 30 minutes to ensure homogeneity. Following this, GTA solution was introduced into the blend, and the stirring process was continued. After a further 5 minutes, the resultant mixture was transferred into moulds and allowed to stand at room temperature overnight to facilitate gelation. The formed hydrogels were subsequently frozen and subjected to lyophilisation at −40 °C to obtain double-network cSCAM shape memory scaffolds (SMSs).

In another investigation, we observed that the inherent instability of imine linkages significantly compromised scaffold integrity due to susceptibility to biodegradation, thereby diminishing both the material's applicability and biocompatibility.²² To address this limitation, imine-based Schiff bases were chemically reduced in a 1 wt% NaBH₄ solution for 24 hours prior to scaffold fabrication; the resulting structure was designated as the rSCAM scaffold. Additionally, control scaffolds were prepared without either MMT or GTA in order to facilitate comparative analysis of their respective properties, namely, cSCA and nSCAM, respectively. The concentrations of all scaffold constituents had been pre-optimised based on the previous work conducted by our group, with full details provided in Table S1.^10,21

2.3. Characterisation of SMSs

Morphology. The porous architecture of the bioscaffolds was examined using scanning electron microscopy (SEM, FESEM S-4800, Hitachi High-Technologies Corporation, Japan). Pore dimensions were quantitatively assessed using ImageJ software (National Institutes of Health, United States).

Spectroscopy. The chemical characteristics of the samples were investigated using Fourier-transform infrared spectroscopy (FTIR, Spectrum Two, PerkinElmer Inc., United States). The proportions of secondary structural components within the protein, including α-helices, β-sheets, antiparallel β-sheets, and random coils, were quantitatively determined by deconvoluting the FTIR spectra in the amide I region (1500–1700 cm⁻¹), using Origin software. The crystalline structure was identified via X-ray diffraction (XRD, D8 Advance, Bruker Corporation, Germany) analysis. The absorbance profiles in the ultraviolet-visible (UV–Vis) range were recorded using a UV-1800 spectrophotometer (Shimadzu Corporation, Japan).

Mechanical properties. The compressive mechanical performance of the fabricated scaffolds was assessed through uniaxial compression testing, employing a texture analyser (TA.XTplus, Stable Micro Systems Ltd, United Kingdom). The stress–strain curves were fitted using a continuous hinge function, incorporating segmental regression lines with smooth transitions to accurately identify the yield point. Detailed information regarding the regression methodology and fitting parameters is provided in the SI.

2.4. Porosity of SMSs

Scaffold porosity, defined as the proportion of void volume within the material, plays a critical role in facilitating cellular infiltration and promoting effective nutrient and waste exchange. In the present study, the porosity of each scaffold sample was evaluated using the ethanol displacement technique. Briefly, each dry scaffold, with pre-determined dimensions and an initial weight (W₀), was completely immersed in absolute ethanol until saturation was achieved. Following saturation, the scaffold was gently blotted with filter paper to remove residual ethanol from the surface, and its saturated weight (W₁) was subsequently recorded. The porosity was then calculated using eqn (1), in which ρ denotes the density of ethanol (0.789 g mL⁻¹) and V corresponds to the geometric volume of the scaffold,

2.5. Swelling ratio of SMSs

The ability of a scaffold to absorb and retain water plays a crucial role in supporting cellular infiltration and proliferation, as well as facilitating the diffusion of nutrients and oxygen, thereby enhancing overall tissue regeneration. In the present study, the swelling behaviour of the scaffold samples was assessed by immersing them in PBS (pH 7.4) at 37 °C. The samples were retrieved and weighed at predetermined intervals until they reached equilibrium swelling. The swelling ratio was calculated in accordance with eqn (2),where M₀ represents the initial dry weight of the scaffold and M₁ denotes the weight of the scaffold following equilibrium swelling.

2.6. Shape recovery profiles of SMSs

Shape recovery behaviour was evaluated to investigate the ability of the scaffolds to return to their original configuration following mechanical deformation. In this assessment, SMS specimens with an initial height of 1.1 cm were subjected to a compressive load using a 100 g weight for a duration of 5 minutes under ambient conditions. Upon removal of the applied load, the extent of height recovery was measured. This process was repeated under both dry and hydrated conditions. For the wet condition, each SMS sample was pre-wetted with 2 mL of distilled water and the same procedure was conducted.

2.7. Shape memory behaviour of SMSs

An essential characteristic of SMSs lies in their ability to retain a temporarily fixed form while subsequently recovering their original geometry upon exposure to appropriate stimuli. In the present study, the shape memory behaviour of the scaffold was activated via an ion-responsive mechanism, specifically involving calcium ions (Ca²⁺) and bicarbonate ions (HCO₃⁻). Initially, scaffolds were exposed to compressive stress and momentarily fixed by immersion in a 3 wt% calcium chloride solution for 5 minutes at ambient temperature. To induce shape recovery, the scaffolds were transferred from the fixing medium into various environments, including air, deionised water (DIW), PBS, bicarbonate solution (BC), and bicarbonate in whole blood (BC/WB), where the BC solution had a concentration of 7 wt%. The shape memory performance was evaluated by monitoring changes in the scaffold height over time, from which both the recovery time and the recovery rate were calculated.

2.8. Drug encapsulation and release of SMSs

The drug loading process was carried out on the optimised scaffold formulation (rSCAM) following mechanical characterisation. Specifically, DOX was encapsulated into the scaffold by immersing it in a DOX-containing solution with a concentration of 100 μg mL⁻¹ for durations ranging from 6 to 24 hours prior to freeze-drying. The resulting samples were denoted as rSCAM/DOX, with the numerical suffix indicating the immersion time. The drug loading efficiency (DLE) and drug loading capacity (DLC) were subsequently determined by quantifying the residual DOX remaining in the immersion solution. A DOX calibration curve was constructed using solutions of varying concentrations, with a linear regression described by the equation absorbance (a.u.) = 0.0106 × concentration (μg mL⁻¹) + 0.0228. This equation was applied to evaluate the DOX content in the soaking medium.

Regarding the release study, DOX-loaded rSCAMs were fully submerged in 10 mL of PBS at 37 °C. At a consistent time each day, 3 mL of the release medium was withdrawn and stored in a refrigerator for later analysis, while the same volume of fresh PBS was replenished to maintain a constant release volume. The UV–Vis absorbance of DOX in the collected samples was measured using a spectrophotometer at a wavelength of 490 nm. The cumulative release profiles of the DOX-loaded rSCAM samples were established following methodologies previously reported in the literature.²⁵

2.9. Haemolytic activity of SMSs

For this assessment, red blood cells (RBCs) were isolated from citrated rabbit blood by centrifugation at 3000 rpm for 10 minutes. The collected RBCs were rinsed twice with PBS to remove plasma residues and then diluted to a final concentration of 2% (v/v) in PBS. To evaluate haemocompatibility, 360 μL of the diluted RBC suspension was combined with 40 μL of various test solutions in Eppendorf tubes. These included 1% Triton X-100, which served as the positive control, PBS as the negative control, and a series of scaffold suspensions at concentrations ranging from 0 to 1000 μg mL⁻¹.

Following preparation, all mixtures were incubated at 37 °C for 24 hours to allow potential haemolytic interactions to occur. Subsequently, the tubes were centrifuged at 3000 rpm for 5 minutes to pellet intact RBCs. The absorbance of the resulting supernatant was then measured at 540 nm using a spectrophotometer to determine the extent of haemolysis, as calculated using eqn (3),

2.10. Cytocompatibility of SMSs

Cell culture. The following cell lines were cultured under standard conditions in Dulbecco's Modified Eagle's Medium (DMEM, Merck KGaA, Hesse, Germany) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. All cell lines were maintained in a humidified incubator set at 37 °C with a controlled atmosphere of 5% CO₂.

ROS induction. The production of reactive oxygen species (ROS) by macrophages was evaluated by seeding macrophages (RAW 264.7, ATCC TIB-71) at a density of 1 × 10⁴ cells per well. The cells were cultured in medium containing rSCAM scaffolds at varying concentrations ranging from 100 to 1000 μg mL⁻¹. As a control group, cells were maintained in DMEM without scaffold supplementation. After 24 hours, intracellular ROS levels were quantified using the fluorescent probe 2′,7′-dichlorofluorescein diacetate (DCFDA, Thermo Fisher Scientific Inc., Massachusetts, United States) at a final concentration of 50 μM.

Cell viability. Aortic smooth muscle cells (MOVAS, ATCC CRL-2797) and endothelial cells (SVEC4-10, ATCC CRL-2181) were seeded at a density of 1 × 10⁴ cells per well and initially cultured in DMEM for 24 hours. Following this pre-incubation period, cells were treated with rSCAM scaffolds at concentrations ranging from 100 to 1000 μg mL⁻¹ and incubated for an additional 24 hours. Cells maintained in untreated DMEM served as the control group. Quantitative evaluation of cell viability was conducted using the PrestoBlue™ reagent (Thermo Fisher Scientific Inc., Massachusetts, United States), following the manufacturer's instructions. In addition, cells were qualitatively confirmed via dual staining with LIVE/DEAD™ reagents.

2.11. Chick embryo chorioallantoic membrane assay

The chick embryo chorioallantoic membrane (CAM) assay was utilised to evaluate the biocompatibility of the scaffold materials by monitoring embryonic development in the presence of the tested samples. Fertilised chicken eggs were incubated under standard conditions at 36.5 °C with 65% relative humidity. On the tenth day of embryonic development, a small window was carefully opened in the eggshell to expose the underlying CAM. Sterilised scaffold samples, including cSCAM and rSCAM, were gently positioned onto the CAM surface. To ensure a hydrated environment conducive to tissue viability, 30 μL of PBS was applied directly over each sample. Concurrently, fertilised eggs treated with sterile saline were employed as the control group. Once treatment was completed, the window was sealed using a paraffin film, and the eggs were returned to the incubator for continuous observation under controlled conditions.

2.12. Anticancer activity of SMSs

To simulate the drug release behaviour of the scaffold and evaluate its anticancer efficacy in an in vitro model, human osteosarcoma cells (Saos-2, ATCC HTB-85) were cultured in 24-well plates. Once the cells reached approximately 80–90% confluency, Transwell inserts were placed into each well. Cylindrical rSCAM/DOX samples, with a height of 1 cm and a diameter of 1.5 cm, were carefully positioned within the Transwells. The culture medium was subsequently replaced with fresh media to ensure that the scaffolds were fully submerged, thereby allowing the release process. The anticancer effect of the released DOX was quantitatively assessed using the PrestoBlue™ assay, while qualitative confirmation was obtained using the LIVE/DEAD™ viability assay, as previously described. Additionally, the generation of ROS by cancer cells exposed to DOX was evaluated using the DCFDA assay to further elucidate the mechanism of cytotoxicity.

2.13. Bone regeneration on SMSs

Cell culture. Mouse preosteoblasts (MC3T3-E1, ATCC CRL-2593) were cultured in minimum essential medium (MEM, Merck KGaA, Germany) supplemented with 2 mM L-glutamine to support cellular growth and function. In this study, cylindrical rSCAM scaffolds (0.5 cm in height and 6 mm in diameter) were placed into Nunc™ 96-Well Polystyrene Round Bottom plates. A cell suspension containing 1 × 10⁴ cells in 100 μL of medium was then carefully introduced into each scaffold. For the control group, the same cell suspension was added directly to the wells without any scaffold present, thereby simulating a scaffold-free environment.

Cell proliferation. The proliferation of MC3T3-E1 within the scaffold was monitored over a period of 7 days using the PrestoBlue™ viability assay. Simultaneously, the secretion of matrix metalloproteinase-2 and matrix metalloproteinase-9 (MMP2/MMP9) by the cells was quantified using an MMP2/MMP9-specific substrate (Echelon Biosciences Inc., Utah, United States) to assess extracellular matrix remodelling activity during cell growth.

TEER. A modified trans-epithelial/endothelial electrical resistance (TEER) measurement was employed. Both blank scaffolds and scaffolds seeded with MC3T3-E1 cells were evaluated using an ohmmeter (Fluke Corporation, Washington, United States) by applying two electrodes to either end of each scaffold.

Cell staining. On the final day of the culture period, cells within the scaffolds were directly stained with CellMask™ Plasma Membrane Stains and 4′,6-diamidino-2-phenylindole (DAPI), both supplied by Thermo Fisher Scientific Inc. (Massachusetts, United States), to enable visualisation of the cell morphology and nuclear localisation. In addition, Alizarin Red S staining (Merck KGaA, Hesse, Germany) was carried out to examine osteogenic differentiation of the preosteoblasts within the rSCAM scaffolds.

2.14. Statistical analysis

All experiments were performed in triplicate or more to ensure reproducibility, and the results were presented as mean values accompanied by the standard deviation (mean ± SD). Data analysis and graphical representation were carried out using GraphPad Prism software (GraphPad Software, Inc., Massachusetts, USA). Statistical significance was interpreted based on p-values and annotated as follows: ns (not significant) for p ≥ 0.05, *p < 0.05, **p < 0.01, ****p < 0.001, and ****p < 0.0001.

3. Results and discussion

3.1. Characterisation of SMSs

As previously mentioned, the SMS system represents one of the key scaffold optimisations developed by our group, based on both prior and ongoing studies aimed at biomedical applications.^10,21,22 In this study, scaffold formulations containing optimised component ratios were employed, as detailed in Table S1. Nevertheless, representative formulations were still included to reaffirm the rationale behind the optimisation and to demonstrate the suitability of the rSCAM system for the specific objectives of this research.

The SMS system demonstrated uniformity and excellent mould adaptability upon completion of the fabrication process, as illustrated in Fig. 1A. The matrix conformed well to the shape of the mould regardless of its geometric complexity, indicating its suitability for use as a bone bioscaffold, particularly in clinical scenarios involving irregular or patient-specific implant geometries. The SEM images in Fig. 1B show well-defined porous architectures across all crosslinked samples, namely, cSCA, cSCAM, and rSCAM. In contrast, the non-crosslinked nSCAM sample lacked a discernible pore structure, displaying instead an interwoven arrangement of polymeric layers, likely due to insufficient crosslinking. Among the tested groups, only cSCAM and rSCAM exhibited relatively uniform pore sizes, in contrast to the cSCA sample, which displayed greater heterogeneity. The surface morphology of MMT-containing scaffolds revealed characteristic textural patterns indicative of well-embedded additives, suggesting homogeneous dispersion of MMT within the polymeric matrix and consequent enhancement of the microscale structure. The pore size and distribution were further analysed using ImageJ software. As shown in Fig. 1C, the largest average pore size was observed in the nSCAM sample (666.7 ± 171.2 μm), while the cSCA group presented a broad and inconsistent distribution with a mean of 232.0 ± 279.9 μm. In comparison, the cSCAM and rSCAM samples exhibited more moderate and consistent pore sizes of 319.4 ± 191.2 μm and 477.2 ± 128.0 μm, respectively. These findings indicate that significant microstructural reorganisation occurred following the reduction of imine bonds in cSCAM, as illustrated in Fig. 1D. Importantly, pore sizes within the range of 300–500 μm are widely considered optimal for preventing hypoxic conditions and facilitating effective bone regeneration and microvascular development within the scaffold construct.^26,27

Fig. 1 Morphological and chemical characteristics of SMSs. (A) Macroscopic images of SMSs captured from different viewing perspectives. (B) Porous architecture and surface morphology of SMSs observed on the cross-sectional plane of the scaffolds at ×100 and ×2500 magnifications. (C) Pore size distribution across different SMS formulations. (D) Chemical structural differences of the crosslinking GTA in the cSCA/cSCAM system and the rSCAM system. (E) FTIR spectra of SMSs. (F) Secondary structural composition of SF in SMSs determined from deconvoluted FTIR spectra. (G) XRD pattern of MMT and rSCAM. Some components of this figure created using BioRender. Luu, C. (2025), https://BioRender.com/8onw4w8.

Subsequently, FTIR spectroscopy was employed to elucidate the structural differences among the various scaffold formulations. As shown in Fig. 1E, all samples exhibited a broad absorption band in the range of 3280–3340 cm⁻¹, attributed to the stretching vibrations of O–H and N–H bonds. This feature reflected the presence of hydroxyl and amine functional groups originating from the constituent biopolymers, including SF, CMCs, and ALG. Notably, a slight upward shift of the band above 3278 cm⁻¹ in the rSCAM sample indicated the emergence of secondary amine groups following the chemical reduction of imine linkages, thereby enhancing the intensity of this singlet absorption peak. Additionally, moderate-intensity peaks were detected around 1619 cm⁻¹ and 1020 cm⁻¹, corresponding to the vibrational modes of C=N, C=O, and C–O bonds, respectively. In the crosslinked samples, a characteristic shoulder band within the 1680–1620 cm⁻¹ region was assigned to the stretching vibration of C=N bonds, indicative of imine formation via GTA-mediated crosslinking. Conversely, this imine-associated peak was absent in the nSCAM sample, confirming the lack of covalent bonding due to the absence of the crosslinker. The chemical reduction of imine to amine groups not only resulted in the disappearance of the C=N signal but also led to an enhancement of the absorption band around 1128 cm⁻¹, which is characteristic of C–N bond formation. Collectively, these spectral changes validated the chemical transformations occurring during the synthesis of the SMS scaffolds and supported the successful progression from crosslinking to reduction stages in the fabrication process.

To enable a more detailed structural analysis, the FTIR spectra within the wavenumber region of 1500–1700 cm⁻¹ were deconvoluted to determine the proportion of secondary structural motifs present in the SMS samples, as illustrated in Fig. 1F.²⁸ Overall, highly ordered conformations such as α-helix and β-sheet were more prominent in the cSCA and nSCAM samples, whereas a greater proportion of random coil structures was observed in both cSCAM and rSCAM. These findings suggest that the presence of crosslinking agents and additives significantly influenced chain arrangement, restricting polymer backbone mobility and thereby limiting the ability of the chains to self-assemble into highly ordered structures. When comparing cSCAM and rSCAM, it was anticipated that the reduction process would allow for protein chain rearrangement and potential restoration of order. However, no notable increase in the β-sheet content was detected. Instead, there was a pronounced reduction in α-helix and antiparallel β-sheet structures, accompanied by a substantial increase in disordered random coil regions. This structural shift may be attributed to spatial disruption caused by salt ions present in the NaBH₄ solution, as previously reported in the literature.²⁹ Nevertheless, the resulting increase in random coil content contributed positively to scaffold hydrophilicity and flexibility, desirable features for applications in shape memory scaffolds aimed at bone tissue regeneration.

The XRD results presented in Fig. 1G revealed that the characteristic (001) diffraction peak of MMT, typically observed at approximately 7° 2θ, either had disappeared or was markedly reduced in intensity in the rSCAM sample.¹⁰ This disappearance of the basal reflection indicated a high degree of delamination and intercalation of the clay within the polymeric matrix, confirming that MMT was effectively dispersed throughout the scaffold. Such uniform distribution is known to enhance the mechanical strength, physicochemical stability, and structural integrity of the system. This phenomenon is primarily driven by strong electrostatic interactions between the negatively charged surface of MMT single layers and the positively charged functional groups present within the polymer matrix. In addition, the XRD pattern of rSCAM displayed a broad reflection slightly above 25° 2θ, which is characteristic of the Silk I crystalline form of the SF structure. This feature suggests that the predominant molecular conformation was helical in nature, specifically associated with β-turn structures.³⁰ These observations were consistent with the earlier FTIR deconvolution results, where the presence of dominant random coil structures was accompanied by a significant contribution from β-sheets. The β-turn itself is a secondary structure defined by a sharp 180-degree bend in the polypeptide backbone, typically connecting β-strands within β-sheets. Its presence further supported the concordance between the chemical composition and the crystalline structure, reinforcing the validity of the scaffold's hierarchical organisation as revealed through both spectroscopic and diffractometric analyses.

3.2. Mechanical and physical properties of SMSs

High porosity is a crucial design parameter in the development of artificial extracellular matrices for bone tissue engineering, as it directly affects nutrient and oxygen diffusion and, more importantly, the infiltration and proliferation of osteogenic cells within the scaffold architecture. Furthermore, a well-developed porous network supports neovascularisation, which plays a pivotal role in successful bone regeneration. As illustrated in Fig. 2A, the nSCAM scaffold exhibited the lowest porosity (55.1 ± 3.1%), likely resulting from the absence of crosslinking, which contributed to structural fragility and subsequent pore collapse. In contrast, the cSCA, cSCAM, and rSCAM scaffolds maintained more structurally stable polymeric frameworks, as previously confirmed by SEM analysis. These samples displayed porosity values of 65.2 ± 2.2%, 84.8 ± 1.9%, and 93.0 ± 1.2%, respectively. Notably, the rSCAM scaffold achieved the highest porosity, approaching 100%, which can be attributed to the post-crosslinking NaBH₄ reduction treatment. This observation aligns well with theoretical porosity requirements (typically ranging from 60% to 90%) necessary to ensure sufficient space for cell adhesion, proliferation, and vascular network formation.³¹ In support of this, previous studies have indicated that scaffolds exhibiting porosity greater than 80% are optimal for cellular infiltration, assuming that mechanical strength is preserved to avoid structural collapse.³² Within this framework, although all three crosslinked SMSs met the porosity criterion, cSCAM and rSCAM were particularly well suited for bone tissue engineering applications, owing to their exceptionally high porosity (>80%). This feature is expected to promote efficient wound healing and facilitate robust tissue regeneration in clinical settings.

Fig. 2 Mechanical and physical properties of SMSs. (A) Porosity of SMSs, with statistical comparison between samples performed against rSCAM using one-way ANOVA. (B) Stress–strain curves of SMS samples obtained using a texture analyser in compression mode; yield points were determined using a continuous hinge function. (C) Young's modulus and toughness of SMSs calculated from the corresponding stress–strain curves. (D) Swelling ratio of cSCAM and rSCAM. Statistical analysis performed against rSCAM using one-way ANOVA in each group (E) Recovery test performed by applying a 100 g load onto cSCAM and rSCAM scaffolds; dashed lines represent the scaffold height after unloading and the original height. (F) Shape recovery profile of rSCAM in dry and hydrated states under compressive stress. Some components of this figure created using BioRender. Luu, C. (2025), https://BioRender.com/8onw4w8.

In the compression test, the typical stress–strain pattern of polymeric scaffolds was clearly observed across all samples, comprising a distinct elastic region followed by a plastic region, without any indication of fracture at the end of the loading process, as shown in Fig. 2B. However, the stress–strain curve of the nSCAM sample deviated markedly from the other three, reflecting a significantly lower resistance to compressive deformation due to the absence of crosslinking. Yield points were identified using segmental regression with a continuous hinge function, and the resulting values are summarised in Table S2. Specifically, the yield stress for the cSCA, cSCAM, and rSCAM scaffolds was 180.9 ± 15.6, 176.9 ± 3.0, and 179.1 ± 7.7 kPa, respectively, indicating no statistically significant differences among the three crosslinked groups. However, when evaluating resilience, the influence of MMT became more evident. The cSCAM and rSCAM scaffolds achieved resilience values of 3859.3 ± 65.3 and 3806.7 ± 191.9 kJ m⁻³, respectively, whereas cSCA exhibited a lower value of 3210.3 ± 174.0 kJ m⁻³. A similar trend was observed in the compressive modulus and toughness results shown in Fig. 2C, where nSCAM consistently demonstrated significantly reduced values in comparison to rSCAM. Although cSCAM showed no notable differences in compressive properties, the rSCAM scaffold consistently outperformed its counterparts, displaying enhanced load-bearing capacity. These results suggest that the chemical transformation from imine to amine linkages not only improved the scaffold's chemical stability but also significantly enhanced its physical performance. Moreover, the incorporation of MMT into the polymer matrix appeared to further reinforce the mechanical properties, highlighting the beneficial role of nanoclay in optimising scaffold integrity.

Additionally, the swelling behaviour of scaffolds represents a critical parameter in bone tissue engineering, as it governs the expansion of the three-dimensional internal structure under in vivo and in vitro conditions.³³ Such expansion facilitates pore enlargement, thereby enhancing cellular infiltration and supporting bone tissue regeneration. As depicted in Fig. 2D, the swelling ratios of cSCAM and rSCAM scaffolds in PBS were monitored over time. Both scaffolds rapidly reached equilibrium swelling within approximately 10 minutes. Notably, the cSCAM scaffold exhibited a remarkably high swelling ratio, approximately 40-fold relative to its initial dry mass; whereas the rSCAM scaffold demonstrated a more moderate swelling ratio of around 20-fold. This pronounced water uptake was primarily attributed to the hydrophilic nature of the scaffold's base components which are known for their strong water-retention capacity.²³ In addition, the presence of nano-sized MMT, characterised by its high specific surface area and interlayer spacing rich in hydroxyl groups, further contributed to enhanced water absorption. While elevated swelling capacity can be beneficial for maintaining a moist environment and supporting cell viability, excessive swelling may compromise structural integrity and hinder precise control over scaffold dimensions following implantation. In contrast, the moderate swelling profile observed in the rSCAM scaffold suggested superior morphological and structural stability due to the enhancement of the crosslinking density of 8.9 ± 0.7 nmol mL⁻¹, compared to 3.2 ± 0.4 nmol mL⁻¹ of cSCAM. Moreover, this controlled expansion is likely to contribute to a more sustained and predictable drug release behaviour, rendering the rSCAM scaffold particularly well-suited for localised post-surgical chemotherapy applications.

3.3. Shape recovery profiles of SMSs

Fig. 2E and F illustrate the shape recovery performance of the rSCAM scaffold under compressive stress in both dry and wet states, with the experimental procedure schematically outlined in the accompanying inset. In the dry state, both cSCAM and rSCAM scaffolds exhibited limited shape recovery, regaining approximately 60% and 70% of their original height, respectively. These results were consistent with the earlier compression tests, reaffirming the superior elasticity of the rSCAM sample compared to cSCAM. However, neither scaffold fully recovered its original form, and their recovery capacity diminished progressively with repeated compression cycles.

In stark contrast, under hydrated conditions, the rSCAM scaffold displayed outstanding elasticity, achieving near-complete shape recovery across multiple compression cycles. This enhanced performance was attributed to water absorption, which increased the flexibility of the polymer network, allowing for effective dissipation of mechanical stress and enabling the scaffold to return to its initial shape post-deformation. This behaviour is consistent with the theoretical framework of hydrogels and water-absorbent polymer systems, in which hydrated polymer chains exhibit greater molecular mobility and energy dispersion, thereby supporting efficient elastic recovery. The exceptional shape recovery of rSCAM in the wet state represents a critical functional property for biomedical applications, particularly in replicating the mechanical environment of bone and soft tissue, where scaffolds must endure repeated mechanical loading under physiological conditions.

3.4. Shape memory behaviour of SMSs

The shape-memory performance of the scaffolds was assessed using a cylindrical rSCAM scaffold with an initial height of 10 mm. The scaffold was compressed in a 3 wt% calcium chloride solution under a 100 g load, resulting in a temporary height reduction to approximately 5 mm, as illustrated in Fig. 3A. This temporary shape fixation was achieved through the ionic complexation between calcium ions and the α-L-guluronic acid blocks in alginate, forming an “egg-box” structure that enabled the scaffold to maintain its deformed configuration.¹⁷ The underlying mechanism is depicted in Fig. 3B. To reverse the temporary shape and restore the original geometry, the scaffold was exposed to a bicarbonate ion solution, which served as a chemical stimulus. The efficacy of this reversal process was further evaluated under various environmental conditions, as presented in Fig. 3C, to determine the optimal medium for triggering the shape-memory effect.

Fig. 3 Shape-memory behaviour of SMSs. (A) Programming process of the temporary shape of the ion-responsive scaffold using calcium ions. (B) Schematic illustration of the shape-memory mechanism within the scaffold matrix, in which the presence of calcium ions promotes the formation of an “egg-box” structure that stabilises the temporary configuration. (C) Experimental setup and results showing shape recovery performance in various environments including air, DIW, PBS, BC, and BC/WB; SMSs also demonstrated shape-memory capability in more complex geometries such as flower-like structures. (D) Shape-memory profiles of SMSs under different environmental conditions. (E) Recovery time and (F) recovery rate of SMSs in the tested media. Some components of this figure created using BioRender. Luu, C. (2025), https://BioRender.com/8onw4w8.

The results presented in Fig. 3C–F demonstrated that, in ambient air, the SMS scaffold retained its deformed shape following fixation, with no observable change in the height throughout the monitoring period. When immersed in DIW, the scaffold began to recover in size, although it failed to return fully to its original shape. This partial recovery suggested that the reversal of the “egg-box” structure was limited in DIW, proceeding at a relatively slow rate of 7.0 ± 1.2% per minute. In contrast, PBS facilitated faster recovery, with full shape restoration occurring within 503.3 ± 4.7 seconds and at a rate of 12.3 ± 1.2% per minute. As anticipated, the scaffold immersed in BC solution achieved complete recovery within only 88.3 ± 6.2 seconds, while the BC/WB mixture was less efficient, requiring 293.3 ± 6.2 seconds for full restoration. These findings highlighted bicarbonate ions as effective antagonists to calcium ions within the shape-memory mechanism. The strong chemical interaction between these two ions enabled BC to displace calcium ions from the alginate “egg-box” structure, disrupting the physical crosslinks between alginate chains and thereby releasing the temporary shape and allowing the scaffold to revert to its original form.

The data confirmed that the calcium–alginate network was sufficiently robust to maintain the deformed structure under hydrated conditions without uncontrolled swelling, as previously observed in the shape recovery experiments. Nevertheless, water alone was capable of gradually disrupting the ionic crosslinks, likely due to ion diffusion driven by the concentration gradient between the scaffold interior and the surrounding medium. The presence of phosphate ions in PBS also contributed to scaffold recovery through calcium chelation, although the relatively low phosphate concentration (∼0.15 wt%) limited its effectiveness. In the case of WB (whole blood), the more complex ionic composition, potentially including other biologically relevant ions, may have further stabilised the “egg-box” structure and hindered the shape recovery process.

The shape memory mechanism of the SMSs primarily relied on the formation of an “egg-box” structure, whereby divalent cations interacted with the α-L-guluronic acid blocks of alginate. This reversible physical crosslinking temporarily restricted the mobility of the scaffold, effectively fixing it in a compressed conformation, while the covalent network crosslinked by GTA preserved the matrix integrity. Upon exposure to anions with high binding affinity, the divalent cations were withdrawn from the “egg-box” structure, thereby deliberating the alginate chains and allowing the scaffold to recover its original shape. During this recovery process, the chemically crosslinked network remained intact, preventing disintegration of the structure. In addition to anions, environments with low divalent cation concentrations or the presence of monovalent cations could also promote ion exchange and gradually displace the calcium–alginate coordination, albeit over a longer timescale. This underpinned a slower but still observable shape memory response under such conditions.

Altogether, these results underscored the tunable shape-memory behaviour of SMSs based on the alginate “egg-box” mechanism, which could be modulated by ionic species, concentration, and surrounding medium and potentially by pH or electric fields. Such responsiveness renders the system particularly attractive for minimally invasive surgical procedures and a broad range of biomedical applications requiring customisable shape-memory profiles.

3.5. Drug encapsulation and release profiles

Drug encapsulation was carried out using the immersion method, allowing the drug to diffuse into the scaffold and become embedded on the pore surfaces. Consequently, immersion time was regarded as a critical factor influencing the efficiency of the process. The results of DLE and DLC at different immersion periods of rSCAM in the drug solution are presented in Fig. 4A. Both DLE and DLC were found to increase markedly with longer immersion times. Among these, the highest values were observed in the rSCAM/DOX24 sample, with the DLE and the DLC reaching 88.9 ± 1.2% and 26.2 ± 0.3%, respectively. These findings suggested that the rSCAM scaffold possessed a strong ability to interact with DOX molecules, most likely through ionic interactions, π–π stacking, or hydrogen bonding, supported by its porous structure, its large surface area, and the residual functional groups present within the polymer matrix.

Fig. 4 DOX loading and release profile from the rSCAM system. (A) DLE and DLC of rSCAM scaffold with varying immersion durations in drug solution. (B) Illustration of sustained release of DOX at the implantation site. (C–F) DOX release profiles after adsorption periods of 6, 12, 18, and 24 hours, respectively, demonstrating sustained and stable release patterns.

As is anticipated, differences in drug loading efficiency resulted in distinct release profiles. In this study, we employed a localised anticancer drug delivery approach using scaffolds, as illustrated in Fig. 4B. The desired outcome was to achieve sustained DOX release at a relatively constant rate, thereby ensuring the eradication of any residual cancer cells and minimising the risk of recurrence. To this end, four rSCAM/DOX formulations, each corresponding to a different drug-loading duration, were investigated for their release behaviour. The release profiles of these samples, recorded under physiological conditions in PBS at body temperature, are presented in Fig. 4C–F. The rSCAM/DOX6 sample exhibited the most rapid release, with nearly all DOX released within 10 days. In contrast, longer loading durations resulted in extended release periods, with the rSCAM/DOX24 sample demonstrating sustained release over 14 days. In general, the amount of DOX released on the first day increased with longer drug-loading times. This trend is consistent with the concentration gradient between the DOX within the scaffold and the surrounding release medium, which drives the diffusion process. For scaffold-mediated drug delivery, this localised method provides targeted release at the treatment site. Consequently, an initial burst release is not necessarily undesirable, as it may contribute to immediate cytotoxicity against remaining tumour cells. Importantly, prolonged loading times appeared to modulate the release rate, leading to a more gradual and controlled drug release. This behaviour is advantageous, as it minimises damage and prolonged stress to adjacent healthy tissues, offering a safer and more effective therapeutic profile over extended treatment periods.

The release profiles of the rSCAM/DOX scaffolds were further examined by fitting the experimental data to a range of mathematical models, as summarised in Tables S4–S7. Among all formulations, the Higuchi and Korsmeyer–Peppas models consistently provided the best fit, with R² values exceeding 0.95, suggesting that DOX release was primarily governed by diffusion through the scaffold's porous matrix. Notably, the Korsmeyer–Peppas model produced the highest correlation coefficients for rSCAM/DOX12 (R² = 0.9997) and rSCAM/DOX24 (R² = 0.9972), further supporting its relevance in describing the release behaviour.

However, analysis of the release rate constants derived from these two models indicated that the release mechanism was not purely Fickian. Instead, the data pointed towards a combination of Fickian diffusion and anomalous transport. This dual mechanism implied that, in addition to drug diffusion driven by concentration gradients, matrix swelling also played a role, an effect commonly observed in polymer-based drug delivery systems. Such swelling facilitates enhanced chain mobility and contributes to the gradual release of encapsulated agents, thereby accounting for the sustained release patterns observed in the rSCAM scaffolds.

Overall, the rSCAM scaffold demonstrated not only a high drug loading capacity but also a stable and prolonged release profile. Among the evaluated formulations, rSCAM/DOX24 emerged as the most promising candidate, offering an optimal balance between loading efficiency, drug entrapment, and sustained release behaviour, features that are particularly advantageous for postoperative cancer therapy. These results highlighted the scaffold's potential for long-term, localised drug delivery applications, where consistent therapeutic dosing and reduced systemic toxicity are essential for effective treatment outcomes.

3.6. Cytocompatibility of SMSs

In the present study, haemocompatibility evaluation was carried out to examine the interaction between the fabricated scaffolds and erythrocytes, thereby providing further evidence of their overall biocompatibility. Haemocompatibility is widely recognised as a fundamental requirement for biomaterials intended for bone tissue engineering, particularly due to their likely direct contact with blood components during surgical implantation.³⁴ As illustrated in Fig. 5A, the haemolysis levels recorded for all tested scaffold concentrations remained consistently low, with values not exceeding 2.5%. According to ISO 10993-4 standards, materials that induce less than 5% erythrocyte lysis are categorised as highly haemocompatible. These results confirmed the suitability of the scaffold formulations for in vivo biomedical applications involving exposure to the circulatory system.

Fig. 5 In vitro and in ovo biocompatibility of SMSs. (A) Haemolysis assay of rSCAM at various concentrations. (B) ROS induction in RAW 264.7 macrophages following exposure to rSCAM. (C) Cell viability of MOVAS and SVEC4-10 cells cultured with rSCAM; statistical comparisons were performed using one-way ANOVA with control samples as the reference group. (D) Fluorescence images of cells stained with LIVE/DEAD™ reagents for qualitative assessment of viability. (E) Images of cSCAM and rSCAM cultured on embryonic models. White arrows indicate vascular infiltration into the scaffold.

ROS induction refers to the oxidative stress response initiated by macrophages upon encountering a foreign body. The use of biologically derived or highly biocompatible materials can significantly attenuate immune activation by pro-inflammatory cells, thereby minimising adverse reactions when biomedical scaffolds are implanted in vivo. As shown in Fig. 5B, the level of ROS generated by RAW 264.7 macrophages following exposure to the rSCAM scaffold was assessed across a range of concentrations. The results indicated no notable increase in ROS production compared to the control group, even at the highest tested concentration of 1000 μg mL⁻¹. This finding underscores the cytocompatibility of the SMS system, which is largely attributed to its composition from naturally sourced polymers such as silk fibroin, marine alginate, and carboxymethyl chitosan derived from crustacean shells. These biopolymers are known for their minimal immunogenicity, supporting the scaffold's potential for safe application in tissue engineering and regenerative medicine.

With regard to cellular viability, MOVAS cells are widely employed to assess in vitro cytocompatibility, while SVEC4-10 cells, representing vascular endothelial lineage, play a crucial role in neovascularisation, which is essential for tissue repair and regeneration.³⁵ The formation of microvessels around the regenerating region supports nutrient transport and metabolic exchange, thereby facilitating tissue development. As shown in Fig. 5C, both cell lines exhibited a slight reduction in viability as the concentration of rSCAM increased from 100 to 1000 μg mL⁻¹. Notably, SVEC4-10 cells displayed a more pronounced sensitivity to higher concentrations. Despite this decline, cell viability for both cell types remained above 80% across all tested conditions, satisfying the cytocompatibility threshold defined by ISO 10993-5. Consistently, fluorescence imaging of cells stained with LIVE/DEAD™ reagents confirmed the quantitative results from the PrestoBlue™ assay. A gradual decrease in the number of viable cells and the emergence of a small population of dead cells were observed at higher scaffold concentrations, reinforcing the overall conclusion that the SMS system remains highly biocompatible under physiological conditions.

3.7. Chick embryo chorioallantoic membrane assay

One of the most widely accepted methods for evaluating the biocompatibility of biomedical materials is the CAM assay, which utilises fertilised chicken eggs.³⁶ This approach offers distinct advantages, as the developing vascular network of the embryo can be directly visualised through the CAM, thereby enabling real-time assessment of angiogenesis and vascular responses in the presence of test materials. In the present study, scaffold samples were placed onto the CAM surface of fertilised eggs and incubated for approximately seven days to evaluate their biocompatibility in ovo. A saline solution served as the negative control throughout the experiment to establish a baseline for comparison. Two scaffold systems, cSCAM and rSCAM, were tested in parallel to assess and contrast their respective impacts on vascular development and tissue compatibility. The experimental setup and outcomes are shown in Fig. 5E.

As a result, the CAM assay findings confirmed that neither the cSCAM nor rSCAM scaffolds elicited any observable acute toxicity during the monitoring period spanning embryonic days 11 to 17. No signs of necrosis, oedema, haemorrhage, or embryo mortality were detected in any of the experimental groups. In addition, the vascular networks surrounding the implantation sites remained structurally intact and evenly distributed, with no evidence of vasoconstriction or tissue degradation.

Both scaffold types appeared to integrate well with the CAM tissue, as evidenced by seamless interface formation and enhanced vascular infiltration, denoted by the white arrows in Fig. 5E. A marked increase in the vessel diameter was observed at the scaffold–CAM interface when compared to the control group, with the rSCAM sample exhibiting a particularly pronounced effect. This suggests a superior pro-angiogenic response, likely attributed to the physicochemical properties of the scaffold material. Notably, scaffold degradation was more apparent in the cSCAM group, which began to exhibit visible morphological deformation from day 16 onwards. In contrast, the rSCAM scaffold maintained its structural integrity despite minor desiccation effects induced by the incubation environment, highlighting its superior chemical stability. Taken together, these results demonstrate that both scaffold systems were highly biocompatible under in ovo conditions. However, given its enhanced vascularisation capacity and improved durability, rSCAM remains the more promising candidate for dual applications in postoperative chemotherapy and bone tissue regeneration.

3.8. Anticancer activity of SMSs

To endow the scaffold with anticancer functionality, DOX, a widely used chemotherapeutic agent, was incorporated into the biomaterial system. Like other cytotoxic drugs, DOX is associated with numerous side effects; therefore, its local or targeted delivery through biomaterial carriers represents a promising strategy for postoperative chemotherapy. Since surgical resection typically eliminates the majority of the tumour mass, this study focused on evaluating the anticancer performance of the rSCAM/DOX scaffold using a two-dimensional osteosarcoma cell model.

To simulate in vitro drug release conditions, DOX-loaded scaffolds were placed in Transwell inserts suspended above adherent Saos-2 cells cultured in a standard well plate. This configuration, inspired by the study of Sun et al., ensured the diffusion of DOX through the medium without a direct contact between the scaffold and cells, as illustrated in the inset of Fig. 6A.³⁷ As a result, within the first 4 hours of incubation, the viability of Saos-2 cells had decreased to 67.1 ± 7.5%. Over the subsequent 4 hours, viability dropped further to 30.8 ± 18.8%, and by the 12-hour mark, the majority of cells had undergone complete cell death. The calculated half-maximal inhibitory time was approximately 6 hours. These findings were in line with expectations for assessing cytotoxic efficacy using a two-dimensional cancer model.

Fig. 6 Anticancer activity of DOX-loaded SMSs. (A) Cell toxicity assay on Saos-2 osteosarcoma cells, with the inset illustrating the experimental setup using a transwell system. (B) Fold change in ROS induction over time in Saos-2 cells treated with rSCAM/DOX. (C) Fluorescence imaging of Saos-2 cells stained with LIVE/DEAD™ reagents. Some components of this figure were created using BioRender. Luu, C. (2025), https://BioRender.com/8onw4w8.

In practice, the drug dosage embedded within the scaffold can be modulated based on the predicted density of residual cancer cells post-surgery, thereby mitigating damage to surrounding healthy tissue and promoting the timely transition to bone healing. Fluorescence imaging (Fig. 6C) further supported these observations, revealing progressive morphological alterations and an increase in non-viable cells over time, consistent with the results of the cytotoxicity assays. DOX is known not only to interfere with DNA replication but also to generate oxidative damage in both nuclear and cytoplasmic components. Accordingly, DCFDA staining was employed to confirm ROS production by Saos-2 cells cultured in the presence of rSCAM/DOX. As shown in Fig. 6B, a marked increase in ROS levels was detected within the first 4 hours, with a significant elevation observed from 8 hours onward. ROS production peaked at 12 hours, reaching 1.6 ± 0.2 times the baseline level recorded at time zero. A slight decline in ROS at 16 hours was likely due to the detachment of dead cells from the culture surface during DCFDA staining. Nevertheless, the ROS levels at this time point remained significantly higher than the initial baseline. Collectively, these findings confirmed the cytotoxic efficacy of DOX released from the rSCAM/DOX scaffold and highlighted its potential as a dual-function material for localised chemotherapy and bone regeneration following tumour resection.

3.9. Bone regeneration on SMSs

For bone regeneration studies, preosteoblasts were employed to evaluate their capacity to proliferate on the biomaterial scaffold rSCAM. The ability of these cells to grow effectively reflects the scaffold's potential to serve as a supportive platform for cellular migration, differentiation, and function, processes essential for accelerating tissue regeneration. In this context, the proliferation of MC3T3-E1 cells was compared under culture conditions with and without scaffold support, with the latter serving as a control. The results are presented in Fig. 7A. As expected for an adherent cell line, MC3T3-E1 cells continued to grow on the surface of standard tissue culture plates, even in the absence of scaffolds. However, significant growth relative to the initial seeding density was only observed by day 7, with a modest increase of 1.2 ± 0.1-fold, indicating a relatively slow proliferation rate. In contrast, cells cultured within the scaffold showed a notable increase in viability as early as day 1 post-seeding, reaching 1.3 ± 0.1-fold compared to baseline. This result suggests that the biopolymer-based scaffold provided an advantageous microenvironment for cell adhesion and early-stage proliferation. Cell attachment within the scaffold was likely facilitated by electrostatic interactions between the cell membrane and positively charged functional groups present in SF and CMCs, thereby enhancing the initial cell–material contact and supporting sustained cellular development.

Fig. 7 Osteogenic cell migration and differentiation on SMSs. (A) Cell proliferation on the scaffold compared to a tissue culture well plate. Statistical analysis was performed using one-way ANOVA to compare each timepoint with day 0. (B) MMP2/MMP9 secretion levels of preosteoblasts cultured with and without the rSCAM bioscaffold. (C) TEER values of scaffolds cultured with cells compared to acellular scaffolds incubated under identical conditions. Statistical comparison of TEER values between cell-seeded and non-seeded scaffolds at each timepoint was conducted using two-way ANOVA. (D) Fluorescence micrographs of MC3T3-E1 cells grown on the scaffold surface. (E) Alizarin Red S staining indicating calcium deposition and osteogenic differentiation in cell-seeded scaffolds, in contrast to non-seeded controls.

MMP2 and MMP9 are gelatinases secreted by a wide range of cell types, including endothelial cells, fibroblasts, macrophages, osteoblasts, and mesenchymal stem cells, typically in response to tissue injury, inflammation, growth factor signalling, or dynamic tissue remodelling.³⁸ As such, the expression of MMP2/MMP9 is widely regarded as a hallmark of active cellular migration, angiogenesis, and extracellular matrix (ECM) remodelling. As shown in Fig. 7B, the levels of MMP2/MMP9 secreted by MC3T3-E1 cells were compared between cultures grown in non-adherent well plates and those seeded onto the rSCAM scaffold. The results revealed significantly higher gelatinase activity in cells cultured on the scaffold compared to the control. This upregulation correlated with the earlier observed enhancement in cell proliferation, suggesting that the scaffold environment promoted matrix remodelling activity, which, in turn, facilitated cell migration and colonisation. The enzymatic degradation of ECM barriers by MMP2 and MMP9 enables cells to infiltrate the scaffold, thereby accelerating tissue development. Moreover, this process supports endothelial cell invasion and neovascularisation, as demonstrated in the CAM assay, while also contributing to the integration and maturation of the newly regenerated tissue.

In this study, TEER measurements were employed as an indirect yet robust method to evaluate the integrity of cellular growth on the scaffold. TEER provides a quantitative indication of cell attachment and the formation of cohesive cellular layers. Elevated TEER values reflect improved cell–cell junction formation, implying not only successful adhesion of cells to the scaffold surface but also subsequent proliferation and establishment of barrier-like structures that resemble physiological tissue function. As shown in Fig. 7C, the electrical resistance of the scaffold immersed in culture medium alone remained relatively stable, fluctuating around 5 kΩ, with a slight downward trend over time. This minor decrease may be attributed to the gradual degradation of biopolymeric components in the aqueous environment. In contrast, the scaffolds seeded with MC3T3-E1 preosteoblasts exhibited a progressive increase in TEER at each measured time point. Notably, a significant rise in TEER was observed as early as day 1, reaching 10.6 ± 0.4 kΩ, which continued to increase throughout the experimental period, peaking at 45.8 ± 1.9 kΩ by the final day. This upward trend corresponded well with the proliferation of preosteoblasts within the scaffold, as previously demonstrated, suggesting that cellular growth directly contributed to changes in scaffold resistance. These findings not only reinforced the earlier observations of cellular expansion within the scaffold but also highlighted the potential of TEER as a biosensing platform. This method offers a non-invasive means of monitoring cell migration and proliferation and may prove to be valuable in predicting and tracking the dynamics of tissue regeneration within engineered constructs.

To visualise the adhesion and subsequent development of MC3T3-E1 cells on the scaffold, fluorescence staining was performed using CellMask™ for membrane labelling and DAPI as a nuclear counterstain, enabling direct imaging of the cells within the scaffold, as depicted in Fig. 7D. The images reveal clear evidence of cellular adhesion, with cells distributed across various locations within the porous architecture, indicating effective migration into the scaffold interior. MC3T3-E1, a widely utilised preosteoblastic cell line, is frequently employed in studies investigating osteogenic differentiation. In Fig. 7E, Alizarin Red S staining, commonly used to visualise calcium deposition and confirm osteoblast differentiation, was applied to assess mineralisation. The distinct red staining observed in the rSCAM scaffold following seven days of perfusion with MC3T3-E1 cells indicated the presence of calcium deposits, suggesting partial differentiation of preosteoblasts into mature osteoblasts, a key event in bone formation and regeneration. In contrast, the control scaffold, which was not seeded with cells, showed no comparable staining, confirming the specificity of the osteogenic response. The observed differentiation is likely attributable to the intrinsic properties of the scaffold constituents. Both SF and ALG have been extensively documented to promote the proliferation and osteogenic differentiation of bone-related cell types.^39,40 These results highlighted the considerable potential of the rSCAM system not only to promote cellular adhesion and proliferation but also to facilitate osteogenic differentiation. Such a dual capability renders the scaffold particularly well suited for bone tissue regeneration, as it encourages both cellular integration and maturation. By simultaneously stimulating preosteoblastic proliferation and its progression towards mature osteoblasts, rSCAM creates a conducive microenvironment for bone repair. This feature is especially advantageous for postoperative applications following cancer resection, where the scaffold may act as both a regenerative matrix and a localised drug delivery platform, ultimately supporting structural recovery whilst mitigating the risk of tumour recurrence.

4. Conclusions

In this study, the initial objectives were successfully achieved through the fabrication and comprehensive evaluation of ion-responsive shape-memory scaffolds (SMSs), specifically engineered with microarchitectural features tailored for bone tissue applications. Among the developed scaffolds, the optimised rSCAM system exhibited exceptional performance across multiple key parameters, including porosity, compressive strength, controlled drug release, in vitro biocompatibility, anticancer efficacy, and the capacity to promote bone tissue regeneration and osteogenic differentiation. Importantly, the present work established several significant findings:

1. The chemical reduction of imine to amine functionalities within the scaffold network markedly enhanced the structural stability of the system under aqueous conditions. This reduction altered the arrangement of protein chains, leading to an increased proportion of β-turns and an overall decrease in highly ordered conformations. Nonetheless, these changes contributed to improved compressive resilience due to a more stabilised and elastic matrix.

2. The “egg-box” structure, driven by calcium ion crosslinking with alginate, served as the fundamental mechanism for shape memory functionality. This structure remained relatively stable under damp conditions unless disrupted by ionic concentration gradients capable of reversing calcium–alginate bonding. Shape recovery was found to be dependent on the presence of antagonistic ions, those exhibiting a strong binding affinity for calcium ions, and their concentration.

3. The SMS system demonstrated excellent biocompatibility and supported vascular infiltration. When cultured on rSCAM scaffolds, preosteoblastic cells exhibited enhanced proliferation, migration, and effective differentiation into osteoblasts. These biological processes were likely facilitated by the upregulated secretion of matrix metalloproteinases MMP2 and MMP9, which play pivotal roles in extracellular matrix remodelling and tissue regeneration.

Despite these promising outcomes, the study was limited by the lack of validation in in vivo models. Specifically, further investigation is required to explore shape memory recovery, anticancer efficacy, and regenerative performance in animal systems. In addition, the strategic incorporation of bioactive agents that promote tissue regeneration could be considered as a complementary design element. These aspects are planned to be addressed in future studies.

In conclusion, the results and insights obtained in this work highlight the success and potential applicability of the ion-stimuli responsive scaffold system, characterised by calcium ion-triggered shape memory behaviour, as a dual-functional platform for postoperative chemotherapeutic delivery and bone tissue regeneration.

Author contributions

Hong-Phuc Nguyen: writing – review and editing, methodology, investigation, formal analysis, data curation, and conceptualization. Quynh-Nhu Doan Nguyen: writing – original draft, methodology, investigation, formal analysis, data curation, and conceptualization. Thanh-Ngan Le Phong and Ngoc Hong T. Luu: investigation, formal analysis, and data curation. Shehzahdi S. Moonshi and Hien Ngoc Trieu: investigation and formal analysis. Vinh-Han Dac Le: investigation. Thanh-Mai Ngoc Nguyen: data curation. Diep Phan: data curation. Cuong Hung Luu and V. H. Giang Phan: supervision, methodology, conceptualization, and administration.

Conflicts of interest

The authors declare that they have no conflict of interest.

Data availability

The data supporting this article have been included as part of the SI. See DOI: https://doi.org/10.1039/d5tb01552a.

Acknowledgements

This work was supported by Ton Duc Thang University. It was also supported by the Queensland node of the Australian National Fabrication Facility, Griffith University (ANFF).

References

1. U. Anand, A. Dey, A. K. S. Chandel, R. Sanyal, A. Mishra, D. K. Pandey, V. De Falco, A. Upadhyay, R. Kandimalla, A. Chaudhary, J. K. Dhanjal, S. Dewanjee, J. Vallamkondu and J. M. Pérez de la Lastra, Genes Dis., 2023, 10, 1367–1401.
2. A. Niculescu, A. F. Grecu, A. E. Stepan, M. I. Muşat, A. E. Moroşanu, T. A. Bălşeanu, M. Hadjiargyrou and D. C. Grecu, Rom. J. Morphol. Embryol., 2024, 65, 723–736.
3. E. Angst, J. R. Hiatt, B. Gloor, H. A. Reber and O. J. Hines, J. Am. Coll. Surg., 2010, 211, 412–423.
4. H. Shoman, J. Al-Kassmy, M. Ejaz, J. Matta, S. Alakhras, K. Kahla and M. D'Acunto, J. Bone Oncol., 2023, 39, 100469.
5. G. Perri, L. Prakash, W. Qiao, G. R. Varadhachary, R. Wolff, D. Fogelman, M. Overman, S. Pant, M. Javle, E. J. Koay, J. Herman, M. Kim, N. Ikoma, C.-W. Tzeng, J. E. Lee and M. H. G. Katz, Ann. Surg., 2020, 271, 996–1002.
6. Y. Yang, Y. Lu, H. Tan, M. Bai, X. Wang, S. Ge, T. Ning, L. Zhang, J. Duan, Y. Sun, R. Liu, H. Li, Y. Ba and T. Deng, BMC Cancer, 2023, 23, 422.
7. L. Hung-Cuong, N. Cuong Quoc, N. Ngoc Hoi, T. Dieu Linh, N. Dai Hai and N. Cuu Khoa, Vietnam J. Sci. Technol. Eng., 2023, 65, 25–31.
8. S. S. Moonshi, K. X. Vazquez-Prada, H. Adelnia, N. J. Westra van Holthe, Y. Wu, J. Tang, A. C. Bulmer and H. T. Ta, Appl. Mater. Today, 2024, 37, 102150.
9. Z. Yang and X. Zhao, Int. J. Nanomed., 2011, 6, 303–310.
10. P.-K. T. Ngo, D. N. Nguyen, H.-P. Nguyen, T.-H. H. Tran, Q.-N. D. Nguyen, C. H. Luu, T.-H. Phan, P. K. Le, V. H. G. Phan, H. T. Ta and T. Thambi, Int. J. Biol. Macromol., 2024, 279, 135329.
11. Q. L. Loh and C. Choong, Tissue Eng., Part B, 2013, 19, 485–502.
12. C. Zhang, D. Cai, P. Liao, J.-W. Su, H. Deng, B. Vardhanabhuti, B. D. Ulery, S.-Y. Chen and J. Lin, Acta Biomater., 2021, 122, 101–110.
13. J. Xu, Y. He, Y. Li and Q. Ye, Mater. Lett., 2020, 260, 126968.
14. L. Wang, J. Shansky, C. Borselli, D. Mooney and H. Vandenburgh, Tissue Eng., Part A, 2012, 18, 2000–2007.
15. H. V. Almeida, B. N. Sathy, I. Dudurych, C. T. Buckley, F. J. O'Brien and D. J. Kelly, Tissue Eng., Part A, 2016, 23, 55–68.
16. M. A. Zoroddu, J. Aaseth, G. Crisponi, S. Medici, M. Peana and V. M. Nurchi, J. Inorg. Biochem., 2019, 195, 120–129.
17. V. H. G. Phan, B.-P. T. Nguyen, N. Y. Nguyen, C.-N. D. Tran, Q.-N. D. Nguyen, C. H. Luu, P. Manivasagan, E.-S. Jang, D. C. Yang, D. U. Yang, Y. Li, J. Conde and T. Thambi, Int. J. Biol. Macromol., 2025, 308, 142254.
18. Z. Wenhao, T. Zhang, J. Yan, Q. Li, P. Xiong, Y. Li, Y. Cheng and Y. Zheng, Acta Biomater., 2020, 116, 223–245.
19. J. Ruan, X. Wang, Z. Yu, Z. Wang, Q. Xie, D. Zhang, Y. Huang, H. Zhou, X. Bi, C. Xiao, P. Gu and X. Fan, Adv. Funct. Mater., 2016, 26, 1085–1097.
20. M. M. d A. C. Velo, F. G. N. Filho, T. R. de Lima Nascimento, A. T. Obeid, L. C. Castellano, R. M. Costa, N. C. M. Brondino, M. G. Fonseca, N. Silikas and R. F. L. Mondelli, Sci. Rep., 2022, 12, 10259.
21. N.-Q. T. Nguyen, C. H. Luu, N.-D. H. Nguyen, D. Q. Vo, N. H. T. Luu, N.-Y. N. Ha, T.-T. N. Vo, Q.-P. L. Huynh, T. Thambi, C. T. T. Ngo and G. V. H. Phan, J. Mater. Chem. B, 2025, 13, 6765–6783.
22. T.-H. H. Tran, C. H. Luu, K.-T. T. Nguyen, M.-A. L. Hoang, Q.-K. Pham, C. M. Phan, N.-K.-L. Thai, H. T. Nguyen, T. Thambi and V. H. G. Phan, Macromol. Biosci., 2025, 25, 2400351.
23. C. H. Luu, G. Nguyen, T.-T. Le, T.-M. N. Nguyen, V. H. Giang Phan, M. Murugesan, R. Mathiyalagan, L. Jing, G. Janarthanan, D. C. Yang, Y. Li and T. Thambi, Gels, 2022, 8, 246.
24. P.-K. T. Ngo, C. H. Luu, H.-P. Nguyen, D. N. Nguyen, K. D. Nguyen, T. Van Luu, P. K. Le, Z. Pan, V. H. G. Phan, Y. Li and T. Thambi, Surf. Interfaces, 2025, 72, 107112.
25. T.-D. Nguyen-Dinh, C. H. Luu, C. K. Nguyen, H. Q. Ly, T. H. Nguyen, D. K. Vo Nguyen, T.-K.-C. Huynh, C. Y. Chee and D. H. Nguyen, ACS Appl. Polym. Mater., 2025, 7, 9540–9554.
26. J. Jiao, Q. Hong, D. Zhang, M. Wang, H. Tang, J. Yang, X. Qu and B. Yue, Front. Bioeng. Biotechnol., 2023, 11, 1117954.
27. H.-T. Liao, M.-Y. Lee, W.-W. Tsai, H.-C. Wang and W.-C. Lu, J. Tissue Eng. Regener. Med., 2016, 10, E337–E353.
28. A. Barth, Biochim. Biophys. Acta, Bioenerg., 2007, 1767, 1073–1101.
29. R. Qi, J. Guo, Y. Liu, R. Zhang and Z. Gan, Dyes Pigm., 2019, 171, 107686.
30. J. L. Whittaker, N. R. Choudhury, N. K. Dutta and A. Zannettino, J. Mater. Chem. B, 2014, 2, 6259–6270.
31. T. M. Freyman, I. V. Yannas and L. J. Gibson, Prog. Mater. Sci., 2001, 46, 273–282.
32. F. Mukasheva, L. Adilova, A. Dyussenbinov, B. Yernaimanova, M. Abilev and D. Akilbekova, Front. Bioeng. Biotechnol., 2024, 12, 1444986.
33. M. Murugesan, R. Mathiyalagan, Z. M. Ramadhania, J. Nahar, C. H. Luu, V. H. G. Phan, D. C. Yang, Q. Zhou, S. Chan Kang and T. Thambi, Bioact. Mater., 2025, 49, 154–171.
34. C. H. Luu, N.-T. Nguyen and H. T. Ta, Adv. Healthcare Mater., 2024, 13, 2301039.
35. F. Akther, D. Sajin, S. S. Moonshi, Y. Wu, K. X. Vazquez-Prada and H. T. Ta, Adv. Biol., 2024, 8, 2300463.
36. I. Moreno-Jiménez, G. Hulsart-Billstrom, S. A. Lanham, A. A. Janeczek, N. Kontouli, J. M. Kanczler, N. D. Evans and R. O. C. Oreffo, Sci. Rep., 2016, 6, 32168.
37. R. Sun, M. Wang, T. Zeng, H. Chen, T. Yoshitomi, M. Takeguchi, N. Kawazoe, Y. Yang and G. Chen, Bioact. Mater., 2025, 44, 205–219.
38. H. Li, Z. Qiu, F. Li and C. Wang, Oncol. Lett., 2017, 14, 5865–5870.
39. D.-W. Li, F.-L. He, J. He, X. Deng, Y.-L. Liu, Y.-Y. Liu, Y.-J. Ye and D.-C. Yin, Carbohydr. Polym., 2017, 178, 69–77.
40. B.-H. Lee, B. Li and S. A. Guelcher, Acta Biomater., 2012, 8, 1693–1702.

---

Footnote

† These authors contributed equally.

---