MSNs were fabricated using the modified Stöber method (Zhao et al., 2015). Briefly, 1.4 g cetyltrimethyl ammonium bromide (CTAB; Macklin, Shanghai, China) and 66 ml deionized water were added to a beaker, which was stirred in an 80°C water bath. After the solution became clear and transparent, 20 ml ethyl acetate (Macklin, Shanghai, China) was added to the beaker, and the reaction was magnetically stirred for 30 min. Then, 14 ml 1 mol/L NH₄OH was added to the system and the reaction continued under magnetic stirring for 15 min. Next, 7.2 ml tetraethyl orthosilicate (TEOS; Macklin, Shanghai, China) was added and reacted under magnetic stirring for 6 h. Finally, the white porous silica precipitate was collected via centrifugation, washed three times with deionized water and ethanol, dried at 60°C for 12 h, and finally calcined at 700°C for 4 h to obtain MSNs. The morphology and microstructure of MSNs were observed using scanning electron microscopy (SEM; Gemini 300, Zeiss, Germany) and transmission electron microscopy (TEM; Talos F200 X, FEI, Waltham, MA, United States), and the particle size of the MSNs was measured by dynamic light scattering (DLS) using a Zetasizer (Malvern, United Kingdom).

The microfluidic device is comprised of a microfluidic chip and microfluidic instrumentation. The microfluidic chip (Micronit, Enschede, the Netherlands) was equipped with one inlet and one outlet, main microchannels that were approximately 100 μm in depth and 200 μm in width, and a 50 μm diameter nozzle. Polyethylene tubes with a 0.5 mm inner diameter were inserted as the continuous and sample phase inlets. The microfluidic instrumentation was assembled with a pressure pump (PP), precision pressure controller (PPC), flow sensor (FS), pressure sensor (PS), and liquid storage tank (LST), which were used to pump the fluids into the microfluidic chip (Supplementary Figure S1).

An overview of the schedule used to fabricate the CS/MSN-compromised microspheres is shown in Figure 2A. In brief, CS/MSNs microspheres were prepared using a microfluidic approach. First, 200 mg CS (Shanghai Bio Life Science and Technology Co., Ltd, Shanghai, China) was dissolved in 10 ml of a 0.5% acetic acid solution to form a CS solution (2% w/v). Then, 0, 0.5, 1, 5, 10, and 50 mg MSNs were added to a 10 ml chitosan solution to form different CS/MSN mixtures (CS, CS/MSNs-1, CS/MSNs-2, CS/MSNs-3, CS/MSNs-4, CS/MSNs-5), respectively, and the mixtures were ultrasonically oscillated for 10 min and filtered (0.22 μm pore) to obtain a homogeneous solution (Supplementary Figure S2). The CS/MSNs solution, as the sample phase, was set to flow into the microfluidic chip from inlet 2 by the pumps. The oil solution was a 9:1 (v/v) mixture of octane/span 80 and acted as the continuous phase, which was injected into the microchannel from inlet 1 by the pumps, and made the CS aqueous solution form monodispersed droplets under the shear forces. Subsequently, the CS droplets were dripped into cross-linking solutions (2:8 v/v mixture of 2.5% glutaraldehyde/octane) for 30 min to form CS/MSN microspheres. Then, the CS/MSN microspheres were washed twice with deionized water and ethanol. Finally, the CS/MSN microspheres were dried at 37°C under a vacuum.

The microstructure and surface morphology of the CS/MSN composite microspheres were observed using a scanning electron microscope (SEM, S-4800; Hitachi, Tokyo, Japan). Confocal laser scanning microscopy (CLSM, Leica TCS-SP8; Heidelberg, Germany) was used to evaluate the distribution of MSNs (labeled with rhodamine B) in chitosan (labeled with FITC). The chemical structure of the CS/MSNs composite microspheres was characterized by Fourier transform infrared spectroscopy (FTIR) using a spectrometer (Nicolet Nexus 470; ThermoFisher Scientific, Waltham, MA, United States). Furthermore, X-ray diffraction (XRD) was also used to evaluate the CS/MSNs composite microspheres using an X-ray diffractometer (Ultima IV; Rigaku, Osaka, Japan) in the 2θ range of 5–90°.

Rabbit bone marrow mesenchymal stem cells (BMSCs) were isolated, as described in our previous study, and cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, ThermoFisher Scientific, Waltham, MA, United States) with 10% fetal bovine serum (FBS; Gibco, ThermoFisher Scientific, Waltham, MA, United States), penicillin (100 U/mL), and streptomycin (100 μg/ml) at 37°C under a 5% CO₂ atmosphere.

The in vitro cytotoxicity of the CS-MSNs composite microspheres was evaluated on BMSCs using the MTT assay. Briefly, 1 mg/ml, 10 mg/ml, and 50 mg/ml CS and CS/MSNs (CS/MSNs-1, CS/MSNs-2, CS/MSNs-3, CS/MSNs-4, CS/MSNs-5) composite microspheres were incubated in culture medium for 24 h, then the supernatant was filtered using a 0.22 μm membrane (Millex-GP, MilliporeSigma, Burlington, MA, United States). BMSCs with a density of 5 × 10³ cells per well were seeded in 96-well plates and incubated for 24 h. Then the cells were treated with the extracts of different concentrations of CS and CS/MSNs composite microspheres for another 24 h. Subsequently, the cells were stained using the MTT assay, and the OD values were measured at 570 nm using a microplate reader (Bio-Rad Model 550; Hercules, CA, United States). Cell viability (%) was calculated as [OD]_sample/[OD]_control × 100.

To prepare the CS microspheres loaded with PDGF-BB, a 10 ml CS aqueous solution with 200 mg CS (2% wt) and 1 mg PDGF-BB (100 ng/ml) was used, and the fabrication procedures were performed as described above. In the case of MSNs loaded with KGN, 10 mg MSNs were added to 1 ml of 1 μM KGN PBS solution, ultrasonicated for 5 min, and then stirred for 24 h to enable the KGN to fully impregnate the MSNs. Subsequently, MSNs-KGN was obtained by centrifugation at 15,000 rpm for 10 min. For the preparation of CS-PDGF-BB/MSNs-KGN composite microspheres, 10 mg MSNs-KGN was added to a 10 ml CS aqueous solution with 200 mg CS (2 wt%) and 1 mg PDGF-BB (100 ng/ml), and subjected to ultrasonic oscillation for 10 min to obtain a homogeneous CS/MSNs solution. The fabrication procedures of the CS-PDGF-BB/MSNs-KGN composite microspheres were performed as described above.

The in vitro release profiles of PDGF-BB and KGN from the CS-PDGF-BB/MSNs-KGN microspheres were compared with those of the CS-PDGF-BB microspheres and MSNs-KGN in PBS at 37°C. CS-PDGF-BB/MSNs-KGN microspheres (5 mg), CS-PDGF-BB microspheres (5 mg), and MSNs-KGN (5 mg) were added to 50 ml PBS, and the mixture was shaken at 100 rpm. At the set time intervals, 100 μl supernatant was extracted from each sample and then added to 100 μl of fresh PBS. The PDGF-BB concentration in the supernatant was quantified using a PDGF-BB ELISA Kit (Neobioscience, Shenzhen, China), and the KGN concentration was quantified by HPLC.

The PEEK scaffolds were fabricated using a fused filament fabrication (FFF) 3D printer, according to our previous study. The printing parameters of the scaffold, such as strut size (nozzle size), pore size, layer height, and dimensions, were set to the following: 250 μm strut size, 400 μm pore size, 1 mm layer height, and Ø4 mm x 1 mm dimensions.

The sulfonated PEEK scaffold (SPK) was fabricated as follows. The PEEK scaffolds were treated with concentrated sulfuric acid (98 wt%), subjected to ultrasonic oscillation for 30 s, and then washed with deionized water for 6 h.

CS/MSNs composite microspheres (containing PDGF-BB and KGN) were immobilized in the SPK scaffold using the PDA coating method. The SPK scaffolds were immersed in Tris-HCl buffered solution (pH 8.5) with 2 mg/ml dopamine and 20 mg/ml CS/MSNs microspheres (containing PDGF-BB and KGN) under magnetic stirring for 4 h (termed SPK@PDA-CS/MSNs), and the samples without CS/MSNs microspheres were denoted as SPK@PDA.

We used the transwell chemotactic migration system (Corning, Corning, NY, United States) to evaluate the recruitment ability of PDGF-BB released from the SPK@PDA-CS/MSNs scaffold. In brief, 5 × 10⁴ BMSCs were seeded in the upper chamber of the transwell plate, and the SPK@PDA and SPK@PDA-CS/MSNs scaffolds were placed in the lower chamber. After 24 h of culture, the transwell membrane was scraped on the upper surface to remove the cells, and the migrated cells on the lower side were fixed in 4% paraformaldehyde, stained with crystal violet, imaged using an optical microscope (Olympus Corporation, Japan), and quantitatively evaluated using the Imaris software.

The femoral condyles were harvested at 1 and 3 months postoperatively, fixed in 4% paraformaldehyde for 2 days, decalcified with ethylenediaminetetraacetic acid (EDTA, Servicebio, Wuhan, Hubei, China), dehydrated, embedded, and sectioned into 7 μm slices. The sections were stained with HE and toluidine blue (TB). The sections were also subjected to IHC staining for collagen II as described above.

Statistical analysis was performed with Tukey’s multiple comparison tests, as well as one-way analysis of variance (ANOVA), using GraphPad Prism 9 (GraphPad Software, Inc, La Jolla, CA, United States). All data were expressed as the mean ± standard deviation (SD), and the significance of the difference was set at p < .05.

As shown in Figure 2B, the morphology of the MSNs was observed using SEM. It could be seen that the MSNs were nanoscale particles. The internal pore structure of the MSNs was observed using TEM. We confirmed that the MSNs were monodisperse and possessed a highly porous structure and sufficient space for the adsorption of biomolecules, such as KGN (Figure 2C). The particle size distribution of the MSNs was measured using dynamic light scattering (DLS), as shown in Figure 2D. The particle size was approximately 74 nm, and it was mainly distributed between 68 and 92 nm.

Moreover, different amounts of MSNs (0.5, 1, 5, 10, and 50 mg) were added to 10 ml chitosan solution, and then ultrasonically oscillated for 10 min to form a homogeneous CS/MSNs (CS/MSNs-1, CS/MSNs-2, CS/MSNs-3, CS/MSNs-4, and CS/MSNs-5) mixture solution. As shown in Supplementary Figure S2, CS/MSNs-1, CS/MSNs-2, CS/MSNs-3, and CS/MSNs-4 mixture solutions were homogeneous, while there were some sediments in the CS/MSNs-5 mixture solution, indicating that 50 mg MSNs cannot disperse completely in a 10 ml chitosan solution. Next, different CS/MSN composite microspheres were prepared from a chitosan mixed solution by the microfluidic method.

Subsequently, the CS/MSN composite microspheres were characterized by SEM, fluorescence labeling, FTIR, and XRD. As shown in Figure 2E, SEM was used to observe the surface morphology of the CS/MSN composite microspheres. It can be seen that the structures of the CS/MSNs composite microspheres were significantly different, and the distribution of the MSNs in the CS/MSNs composite microspheres increased as the MSNs concentration increased. Furthermore, we utilized fluorescence labeling (MSNs labeled with rhodamine B, and CS marked with FITC) to observe the internal structure of CS/MSNs composite microspheres, as shown in Figure 2F. The CLSM results showed that the MSN content in the CS/MSNs composite microspheres increased with the addition of MSNs, and MSNs aggregation was evident in the CS/MSNs-5 group, which was consistent with the results of SEM.

The main chemical groups of CS microspheres, CS/MSNs composite microspheres, and MSNs could be evaluated from the FTIR spectra, as shown in Figure 2G. For the CS microspheres, the characteristic peaks were at 2,872 cm⁻¹ (C-H stretching vibration), 1,634 cm⁻¹ (-C=O stretching vibration), 1,550 cm⁻¹ (N-H stretching vibration), and 1,024 cm⁻¹ (C-O stretching vibration). The characteristic absorption peak of MSNs at 806 cm⁻¹ (Si-O-Si stretching vibration) was weakened in the spectra of the CS/MSNs composite microspheres, which may mean that the MSNs were dispersed into the CS matrix. For the CS/MSNs composite microspheres, the main characteristic peaks were also observed at 2,872 cm⁻¹, 1,634 cm⁻¹, 1,550 cm⁻¹, and 1,024 cm⁻¹, which were similar to those of the CS microspheres. Furthermore, the CS microspheres, CS/MSNs composite microspheres, and MSNs were characterized by XRD, as shown in Supplementary Figure S3. The characteristic diffraction peak of the CS microspheres was at 2θ = 19.94° (hydrated crystalline). MSNs appeared as an intense reflection peak at 2θ = 21.6°. With the addition of the MSNs, the characteristic reflection peak of the CS/MSNs composite microspheres was similar to that of the CS microspheres, indicating that the MSNs were dispersed in the CS matrix.

Next, we assessed the cytotoxicity of the CS microspheres and CS/MSNs composite microspheres at 1 mg/ml, 10 mg/ml, and 50 mg/ml using BMSCs by MTT assay. Figure 2H shows the cell viability of BMSCs treated with the extracts of CS microspheres and CS/MSNs composite microspheres at 1 mg/ml, 10 mg/ml, and 50 mg/ml, respectively. The cell viability of CS/MSNs composite microspheres was approximately the same as that of the CS microspheres and the control group; the cell viability of the different samples was above 90%, indicating that CS and CS/MSNs would be safe as drug delivery carriers. As for the five different CS/MSNs composite microspheres (CS/MSNs-1, CS/MSNs-2, CS/MSNs-3, CS/MSNs-4, and CS/MSNs-5), the SEM and CLSM results also confirmed that, with the addition of the MSNs, the distribution of the MSNs in the CS/MSNs composite microspheres increased. In contrast, in CS/MSNs-5, the MSNs aggregation could be observed. Similarly, the CS/MSNs-5 mixture solution contained some sediments, indicating that 50 mg MSNs could not disperse completely in a 10 ml chitosan solution. Therefore, we chose CS/MSNS-4 as a drug delivery carrier.

The PDGF-BB and KGN release profiles of the MSNs loaded with KGN, CS microspheres loaded with PDGF-BB, and CS/MSNs composite microspheres loaded with PDGF-BB and KGN in vitro were investigated using ELISA kits and HPLC (Figure 3). For the KGN release from the MSNs, which was showed in Figure 3A, there was a rapid release of 74% of the total KGN released from the MSNs on the first day, after which the release rate slowed and reached approximately 98% of the total KGN released after 10 days. For the PDGF-BB release from the CS microspheres loaded with PDGF-BB, there was still an initial burst release, with nearly 55% of the total PDGF-BB released on the first day; subsequently, the release slowed and approximately 86% was released after 10 days. In contrast, the drug release behavior of the composite microspheres was completely different from that of the pure MSNs and CS microspheres, as shown in Figure 3B. The release of KGN from CS/MSNs composite microspheres (loaded with PDGF-BB and KGN) was sustained and controlled, with a comparatively slow release of 30% on the first day, and then sustainably released for as long as 3 weeks. The PDGF-BB release from CS/MSNs composite microspheres was sharp compared with KGN, with 51% released on the first day, and nearly 84% released at 10 days, whereas approximately 40% of the total KGN was retained after 10 days. The results indicated that PDGF-BB and KGN were sequentially released from the CS/MSNs composite microspheres. According to the literatures (Charnay et al., 2004; Zhao et al., 2015), drugs (KGN in this study) were loaded in the pore walls of the MSNs and were usually located in the center of the MSNs. As for the CS/MSNs composite microspheres, the KGN first traveled to the surface of the MSNs and entered the CS matrix, then passed through the CS matrix, lengthening the diffusion route, whereas the PDGF-BB diffused directly out of the CS matrix. Therefore, the sequential release behavior of PDGF-BB and KGN from the CS/MSNs composite microspheres could be attributed to the characteristic composite structure of the CS/MSNs.

We evaluated the microstructure and hydrophilicity of the PEEK, SPK, SPK@PDA, and SPK@PDA-CS/MSNs scaffolds using SEM and the water contact angle (WCA). As shown in Figure 4A, Supplementary Figure S4, the morphology observation of the different scaffolds was conducted using general observation and SEM. The PEEK scaffold had a smooth surface, while the SPK and SPK@PDA scaffolds were rougher, which was mainly attributed to the sulfonation treatment and PDA coating. The SEM results also confirmed that the CS/MSNs composite microspheres could be immobilized in the SPK@PDA-CS/MSNs scaffold by the PDA coating method. Figure 4B shows the water angles of the prepared scaffolds and confirmed that these surface modification methods altered the surface wettability of the scaffolds. The sulfonation treatment changed the hydrophilia of the PEEK scaffold from 68.2 ± 1.5°(PEEK) to 77.9 ± 1.1°(SPK), whereas the PDA coating enhanced the hydrophilicity of SPK from 77.9 ± 1.1° to 58.6 ± 1.9°. After the deposition of CS/MSNs composite microspheres in the scaffold, the contact water angles further decreased to 44.7 ± 1.4°.

The cytocompatibility of the prepared specimens, including the PEEK, SPK@PDA, and SPK@PDA-CS/MSNs scaffolds, was evaluated via a cell adhesion assay, live/dead staining, and cytoskeletal staining. The cell adhesion of the scaffolds was evaluated using the CCK-8 kit, and the results in Figure 4C demonstrate that the cell attachment capacity of the SPK@PDA-CS/MSNs scaffold was superior to that of the PEEK and SPK@PDA scaffolds. We then assessed the cell growth of the BMSCs on the PEEK, SPK@PDA, and SPK@PDA-CS/MSNs scaffolds through live/dead staining after culturing them for 3 and 7 days. As shown in Figure 4D, most of the BMSCs on the three different scaffolds were alive after 3 and 7-days culture, and there was an increase in cells on the scaffolds as the culture time increased. In addition, there were more cells on the SPK@PDA-CS/MSNs scaffolds than that on the PEEK and SPK@PDA scaffolds. Furthermore, we observed the morphology of the cells on the three scaffolds after 3-days culture through cytoskeletal staining, which is shown in Figure 4E, Supplementary Figure S5. The BMSCs were spread well on all three scaffolds, while there were more well-spread cells on the SPK@PDA-CS/MSNs scaffold compared to the PEEK and SPK@PDA scaffolds. Therefore, all the experiments demonstrated that the SPK@PDA-CS/MSNs scaffold had superior cytocompatibility and biocompatibility.

The chemotactic capacity of the SPK@PDA-CS/MSNs scaffold (containing PDGF-BB and KGN) was evaluated via a transwell system, as shown in Figure 5A. The BMSCs traversing the transwell membrane were identified and quantified by crystal violet staining, and the results are shown in Figure 5B. The number of migrating cells in the SPK@PDA-CS/MSNs group was significantly higher than that in the SPK@PDA group, confirming the recruitment capacity of PDGF-BB released from the SPK@PDA-CS/MSNs scaffold.

We evaluated the chondrogenic differentiation capacity of the SPK@PDA and SPK@PDA-CS/MSNs scaffolds in the BMSC pellet culture using the transwell system (Figure 5C). As shown in Figures 5D,E, the pellet treated with SPK@PDA-CS/MSNs after a 21-days culture was more transparent than that in the SPK@PDA group. The HE results showed that there were plenty of chondrocyte-like cells and cartilage lacuna in the pellet treated with SPK@PDA-CS/MSNs, whereas the SPK@PDA-treated pellet showed abundant spindle-shaped cells and a loose structure. The AB and SO/FG staining showed greater intensity in the pellets treated with SPK@PDA-CS/MSNs than in those treated with SPK@PDA. Moreover, the GAG per DNA content of the pellet treated with SPK@PDA-CS/MSNs increased significantly compared with that of the SPK@PDA-treated pellet, which was consistent with the histological results (Figure 5F). We then evaluated the expression of chondrogenic differentiation-related genes, including SOX 9 (Figure 5G), aggrecan (Figure 5H), collagen I (Figure 5I), and collagen II (Figure 5J), in the pellets treated with SPK@PDA and SPK@PDA-CS/MSNs by RT-PCR. The expression of SOX 9, aggrecan, and collagen II increased significantly in the pellet treated with SPK@PDA-CS/MSNs compared to those treated with SPK@PDA, while only slight changes were observed in the collagen I expression of the pellets treated with SPK@PDA and SPK@PDA-CS/MSNs, with no statistical significance. These results demonstrated that the chondrogenic differentiation capacity of the BMSC pellets was enhanced by treatment with SPK@PDA-CS/MSNs.

To evaluate the effect of the SPK@PDA and SPK@PDA-CS/MSNs scaffolds on cartilage regeneration, we implanted them into a cartilage defect model of rabbits and observed cartilage regeneration at 1 and 3 months after surgery, using gross observation, histology, biochemical assays, and RT-PCR. Macroscopically, there was no obvious inflammatory reaction in any of the SPK@PDA, SPK@PDA-CS/MSNs, and control groups, and some neo-tissue could be seen in the SPK@PDA and SPK@PDA-CS/MSNs groups, while the smoothness of the repair area, graft level, and adjacent cartilage degeneration varied in the SPK@PDA, SPK@PDA-CS/MSNs, and control groups (Figure 6A). We then semi-quantitatively evaluated the repair effect of the cartilage defect treated with the SPK@PDA and SPK@PDA-CS/MSNs scaffolds via macroscopic cartilage repair scoring (Figure 6B). As shown in Figure 6 B1-7, the scoring of macroscopic cartilage repair showed a significantly improved repair effect in the SPK@PDA-CS/MSNs group compared to that of the SPK@PDA and control groups. The individual parameters of blood vessel presence, surface, graft level, and adjacent cartilage degeneration of the repair tissue of defects treated with SPK@PDA-CS/MSNs were superior to those of the SPK@PDA and control groups.

The histological findings according to HE and TB staining are showed in Figure 6C, which revealed that application of the SPK@PDA-CS/MSNs scaffold significantly improved the repair effect of the cartilage defect relative to the SPK@PDA and control groups. We then evaluated the total collagen and GAG content of the repair tissue using a biochemical assay, as shown in Figures 6D,E, which revealed that application of the SPK@PDA-CS/MSNs scaffold significantly increased the collagen and GAG contents of the repair tissue relative to the SPK@PDA and control groups, which was consistent with the histological findings. Furthermore, the immunohistochemistry analysis of collagen II also showed that application of the SPK@PDA-CS/MSNs significantly increased collagen II staining intensity compared to the SPK@PDA and control groups at 1 and 3 months after surgery (Figure 7A).

RT-PCR with mRNA isolated from the repair tissue was used to evaluate the expression of cartilage-related genes, including SOX 9 (Figure 7B), aggrecan (Figure 7C), collagen II (Figure 7D), and collagen I (Figure 7E). The gene expression of SOX 9, aggrecan, and collagen II in the SPK@PDA-CS/MSNs group were higher than those in the SPK@PDA and control groups at 1 and 3 months after surgery, respectively, while the collagen I expression of the repair tissue in the SPK@PDA-CS/MSNs group was only higher than that in the control group at 1 month after surgery. Other groups only showed slight changes, which were not statistically significant.