3D-printed biomimetic scaffolds with precisely controlled and tunable structures guide cell migration and promote regeneration of osteochondral defect

Untreated osteochondral defects will develop into osteoarthritis, affecting patients’ quality of life. Since articular cartilage and subchondral bone exhibit distinct biological characteristics, repairing osteochondral defects remains a major challenge. Previous studies have tried to fabricate multilayer scaffolds with traditional methods or 3D printing technology. However, the efficacy is unsatisfactory because of poor control over internal structures or a lack of integrity between adjacent layers, severely compromising repair outcomes. Therefore, there is a need for a biomimetic scaffold that can simultaneously boost osteochondral defect regeneration in both structure and function. Herein, an integrated bilayer scaffold with precisely controlled structures is successfully 3D-printed in one step via digital light processing (DLP) technology. The upper layer has both ‘lotus- and radial-’ distribution pores, and the bottom layer has ‘lotus-’ pores to guide and facilitate the migration of chondrocytes and bone marrow mesenchymal stem cells, respectively, to the defect area. Tuning pore sizes could modulate the mechanical properties of scaffolds easily. Results show that 3D-printed porous structures allow significantly more cells to infiltrate into the area of ‘lotus- and radial-’ distribution pores during cell migration assay, subcutaneous implantation, and in situ transplantation, which are essential for osteochondral repair. Transplantation of this 3D-printed bilayer scaffold exhibits a promising osteochondral repair effect in rabbits. Incorporation of Kartogenin into the upper layer of scaffolds further induces better cartilage formation. Combining small molecules/drugs and precisely size-controlled and layer-specific porous structure via DLP technology, this 3D-printed bilayer scaffold is expected to be a potential strategy for osteochondral regeneration.


Introduction
Cartilage defects in the knee joint are mainly caused by trauma, disease, or old age, which will cause pain and affect the patient's quality of life.Articular cartilage is a smooth elastic tissue without blood vessels, lymphatics, or nerves.Thus, it has limited self-healing ability after damage [1].Without proper treatment, cartilage damage might extend into the subchondral bone, further aggravate articular cartilage defects, and lead to osteoarthritis, affecting the normal life of the patient [2,3].Over 500 million people worldwide, approximately 7% of the world's population, are affected by osteoarthritis [4].In 2019, osteoarthritis was responsible for 2% of total years lived with disability (YLD) worldwide, making it the 15th leading cause of YLD [4,5].Accordingly, it is of great significance to repair osteochondral tissue to restore physiological structure and function.Nevertheless, traditional treatment and surgical methods can only temporarily relieve the patient's pain.It remains challenging to tackle the difficulty of osteochondral defects and realize the regeneration of hyaline cartilage with complete functions [6].The major problem for osteochondral repair is to simultaneously restore the distinct biological features of cartilage and subchondral bone [7].Therefore, new techniques are demanded for osteochondral treatment.
In recent years, tissue engineering has emerged as a more effective cutting-edge approach for repairing osteochondral defects [8].As one of the key components, the scaffold has been studied for years to facilitate osteochondral repair and regeneration.A monophasic scaffold is one of the earliest explored strategies, but it is gradually eliminated due to its poor structural similarity to the naturally layered osteochondral tissue and its inability to simulate the biological environment [9,10].Highly specialized scaffolds engineered to reflect the anatomical and physiological hierarchy of osteochondral tissue have recently been prevalent for osteochondral repair [11].Conventionally, different layers of these scaffolds are fabricated separately and integrated afterward, which are complicated to construct, and two adjacent layers may delaminate and eventually fail due to insufficient bonding between them [12][13][14].On the other hand, 3D scaffolds with high porosity and channel structure are indispensable for nutrient and waste transport, providing an optimal reticular skeleton for cell attachment, proliferation, and growth, even promoting angiogenesis and osteogenesis [15].Techniques such as freeze-drying can be used for integral fabrication of scaffold, but it does not allow precise microstructure control.Digital light processing (DLP)based 3D printing has many advantages, such as fast printing speed, high printing accuracy, and high cell viability.As a result, it has been applied in many research directions in tissue engineering.Therefore, DLP-based 3D printing is adopted in our study to rapidly construct an integrated scaffold in one step with precisely controlled structures to mimic the twolayered structure of osteochondral tissue.Inspired by microfracture surgery [16], which allows mesenchymal stem cells (MSCs) from the bone marrow to migrate into the defect area and repair the tissue [8], we design longitudinal lotus-pores to guide bone marrow MSCs (BMSCs) from the bone marrow to infiltrate upward along the pore channels, thus promoting osteochondral regeneration after implantation.However, according to the previous study, BMSCs present limited regenerative effect on cartilage because they prefer to differentiate into fibrocartilage, which expresses type I collagen, rather than hyaline cartilage, which contains type II collagen [17,18].On the other hand, it was observed that cartilage regeneration tends to proceed from both sides of the defect border in our previous study [19,20], indicating the role of chondrocytes or progenitor cells from the surrounding normal tissue in cartilage repair.Thus, the radial-pore structures of the upper layer are also designed to guide the migration of the original cartilage progenitor cells in cartilage to participate in cartilage regeneration.
Gelatin methacrylate (GelMA) hydrogel is a DLP-printable biomaterial with good biocompatibility, and thus has been widely studied in tissue regeneration research, including osteochondral tissues [21,22].In our previous studies, bilayer scaffolds for osteochondral repair were constructed using GelMA and polycaprolactone-hydroxyapatite (HA) [22].As reported, Liu and Zhang et al fabricated graded triple-layered scaffolds with GelMA and nanohydroxyapatite [23,24].Gao et al designed hydrogel scaffolds with extreme mechanical properties by combining GelMA and poly(N-acryloyl 2glycine) [25].In addition, growth factors, another element in tissue engineering, also play a key role.Kartogenin (KGN), a small molecule growth factor, has been reported to have the effect of recruiting endogenous cells in the host and can induce chondrogenesis of BMSCs and synovial MSCs [26,27].However, studies concerning the synergic effects of KGN and migration-guiding structure on cartilage regeneration are still limited.
In this study, we propose a new strategy of integrated printing of biomimetic scaffolds with precisely controlled structures in different layers in one-step via DLP technology to facilitate osteochondral defect repair.This bilayer scaffold is designed with interconnected, precisely controlled microstructures: vertical pores are printed into the thick bottom layer with lotus distribution, and the same size pores are printed into the upper thin layer with both 'lotus-and radial-' distributions.The pore size can be easily controlled from 100 µm, 200 µm to 400 µm, resulting in different surface areas and mechanical properties of the scaffold.The mechanical properties of the bottom and the upper layers can be modulated based on pore distributions and diameter to match the osteochondral tissue, respectively.The pores in the bottom scaffold facilitate the migration of BMSCs to the upper layer, and the 'lotus-' scaffold has stronger mechanical properties than the upper layer, which is beneficial for bone repair.Besides, the upper hydrogel scaffold facilitates not only the migration of BMSCs but also the migration of surrounding cells from normal tissue to the defected area, including chondrocytes or Y Gu et al chondrogenic progenitor cells, which is beneficial for the repair of the cartilage layer.By further incorporating KGN into the upper layer of this biomimetic scaffold, we anticipate that better repair of osteochondral defects could be achieved.
To prepare the hydrogel ink for the bilayer scaffold, the lyophilized GelMA was dissolved in phosphate-buffered saline (PBS) at 10 and 15% concentrations at 42 • C. KGN (200 µM) (Selleck Chemicals, Houston, TX, USA), photo-initiator LAP (0.1%), and UV-absorber phenol red (0.04%) were then added to prepare the ink for the upper layer.Solutions containing GelMA and 0.1% LAP were used as the precursor for the bottom layer.
The integrated bilayer scaffold was manufactured by a custom-made DLP-based 3D printer [32,33].The scaffold diameter was 5 mm with a 1 mm high upper scaffold and a 3 mm high bottom scaffold.The printed scaffold was immersed in PBS and stored at 4 • C for 14 d to evaluate its stability.The nonstructural GelMA scaffold was prepared for comparison.For the test group, the bottom 3 mm scaffold was printed with GelMA solution, followed by printing the upper 1 mm scaffold with ink containing GelMA and KGN small molecules.Scaffolds with 100, 200, and 400 µm longitudinal apertures, respectively, were printed for mechanical tests.Scaffolds with different structures were printed for cell infiltration and migration tests.

Characterization of the scaffold
To evaluate the macroscopic pore structure of the hydrogel scaffolds, the samples were stained with 4 ′ ,6-diamidino-2-phenylindole (DAPI) and observed using an optical microscope.Briefly, DAPI was diluted 1:1000 and used as a dye.The printed upper scaffold was soaked in DAPI for 15 min, washed with PBS three times before they were observed and imaged under an Olympus upright two-photon confocal microscope (S6D, Olympus, Germany).
To evaluate the microstructure of the hydrogel scaffold, the sample was examined using a field emission scanning electron microscope (Nova Nano 450, Thermo FEI, Czech Republic) after gold-palladium coating.

Mechanical test of the scaffold
Uniaxially unconfined compression tests were performed to evaluate the effect of the pores' size and distribution on the scaffolds' mechanical properties.A universal material testing machine (Zwick/Roell with a 5 kN sensor) was used to measure compressive force.The bottom lotus scaffolds were printed with 10% and 15% GelMA.The pore diameters of the scaffolds are 0, 100, 200, and 400 µm.During all experiments, the probe was pressed down at a constant speed of 1 mm min −1 until the height of the scaffold was 65% of the initial height.The compressive modulus was calculated by linear fitting the stress-strain curve in the range of 40%-60% strain.The upper-and bottom-layer scaffolds with 200 µm pore size were printed separately and tested for compressive stress-strain measurements.The experiment process is the same as above.The compression modulus was calculated by linear fitting according to the strain range of 20%-40%.
In order to further prove that the scaffold could function normally under mechanical loading, consecutive cycle compression test was performed on the scaffold with 200 µm pore size.The scaffold was compressed until the strain reached 50% at a rate of 2 mm min −1 , and the specimen was relaxed at the same deformation rate.The above two steps were repeated as one whole cycle for ten times with no waiting time between cycles.

Cell proliferation and viability assay
The cell counting kit-8 (CCK-8) assay was done to evaluate cell proliferation.First, to obtain the extracted liquid, 1 ml of cross-linked GelMA hydrogel was soaked in 10 ml of culture medium for 24 h.L929 cells (mouse fibroblast cells, Cell Bank of the Chinese Academy of Sciences) were seeded at a density of 1000 cells per well and incubated for 24 h at 37 • C. Afterward, the cell medium was renewed, and cells were further incubated with extract liquid for 1, 3, and 5 d.At every test point, the sample solution was removed, and cells were incubated with CCK-8 agent (1:10 diluted with the culture medium) at 37 • C for 2 h.The absorbance values were read at 450 nm using SynergyMx M5 (iD5, Molecular Devices, USA).Each test group has six replicate samples.
To assess cell viability on the hydrogel, a live/dead assay was performed.Briefly, 500 µl GelMA was added to a 3.5 cm dish and photo-crosslinked.C3H cells (C3H/10T1/2, mouse MSCs, Cell Bank of the Chinese Academy of Sciences) were seeded at a density of 5 × 10 5 cells ml −1 and cultured in a humidified atmosphere of 5% CO 2 at 37 • C for 1, 3 and 5 d.At each test point, a live/dead assay kit was administrated to the cells on hydrogel.Samples were observed under an Olympus upright two-photon confocal microscope.

Cell migration test 2.5.1. Cell migration test in vitro
To observe the cell migration, the DiI (1,1 ′ -di octadecyl-3,3,3 ′ ,3 ′ -tetramethylindocarbocyanine perchlorate) (Invitrogen D282)-stained C3H cells were seeded on a 3.5 cm dish at 5 × 10 5 cells ml −1 .After 24h, the culture medium was renewed and the printed scaffolds with/without pores were immersed in the medium at the dish center.On the first day, scaffolds without pores were collected, while the poriferous scaffolds were incubated and collected at D1, D3, and D5.Samples were observed under an Olympus upright two-photon confocal microscope.3D image reconstruction and data analysis were processed using Imaris.

Cell migration test in vivo
A subcutaneous migration test was performed to show the cell migration in vivo.Briefly, sodium pentobarbital was injected into the abdominal cavity of the rat at a dose of 50 mg kg −1 .Then, 2 cm long incisions were made on the rat's back, in which the GelMA hydrogel scaffolds with the different structures were placed and sutured.Samples were collected on the 3rd and 5th day after the operation and then fixed with 4% formaldehyde for more than 24 h.After tissue fixation, the samples were processed with 20% and 30% sucrose gradient solutions at 4 • C for dehydration until the tissue pieces sank to the bottom and then frozen sectioned.The slices were stained with 1:1000 DAPI for 15 min and washed with PBS three times.Samples were imaged using a fluorescent microscope (BX61, Olympus, Japan) at the same magnification and exposure time.The number of cells and the migration depth were quantified using Image J software.Hematoxylin and eosin (H&E) staining was also carried out to assess cell migration.

Rabbit model of osteochondral defect
Adult male New Zealand white rabbits (2.8-3.2 kg) were used to evaluate the function of the 3D-printed integrated bilayer scaffold in vivo.The treatment was conducted under the standard guidelines approved by the Ethics Committee of Zhejiang University (ZJU20210289).The surgery was performed under general anesthesia with an abdominal cavity and ear veins injection of 3% sodium pentobarbital (40 mg kg −1 ).The skin was treated with iodophor in a sterile environment.A 1 cm long incision was made along the inside of the patellar ligament to expose the knee joint.A cylindrical shallow osteochondral defect (4 mm diameter and 2 mm depth) was made at the femoral trochlear by a dental drill.Forty defects were generated and randomly divided into five groups: blank group (B), non-structural scaffold group (N, the 3D-printed scaffold without any pore channels), 'lotus-' scaffold group (L, the 3D-printed scaffold with only 'lotus-' pores), 3D-printed structural scaffold group (P, the 3D-printed scaffold with both 'radial' and 'lotus-' pores), and P + KGN (the 3D-printed structural scaffold combined with KGN) group.The scaffolds were implanted into the defect, and then the tissue was disinfected with iodophor and sutured immediately.For the blank group, the defect was sterilized and sutured directly.After the operation, the experimental animals were intramuscularly injected with penicillin at 400 000 U d −1 for 3 d.Samples were collected on days 7 and 42 after surgery.
For the full-thickness osteochondral defect model, the operation process was the same as above, while the defect size was 5 mm in diameter and 4 mm deep.Joints were randomly divided into four groups: blank group (B), non-structural group (N), 3D-printed structural group (P), and 3D-printed bilayer scaffold with KGN in the upper layer group (P + KGN).There are 12 joints in each group.Animals were sacrificed at 8 and 16 weeks postoperatively, and joint samples from each group were collected for further analysis.

Evaluation of osteochondral regeneration
The harvested joint samples were immediately photographed and then assessed according to the general evaluation standards of the International Cartilage Repair Society (ICRS), specifically including the degree of defect repair, integration to the border zone, and macroscopic appearance [34].The samples were fixed in 4% paraformaldehyde (PFA) for 2 days and decalcified with 10% ethylene diamine tetraacetic acid (EDTA) for 8 weeks.Then, the samples were dehydrated in gradient ethanol, hyalinized in xylene, embedded in paraffin, and sectioned to slices with a thickness of 10 µm using a microtome.Repair of the defect was assessed by H&E and safranin-O/fastgreen (SO/FG) staining.

Biomechanical test
To test the compressive mechanical properties of the repaired cartilage at 16 weeks post-operation (n = 3 per group), the universal material testing machine (Instron-5543 with a 1 kN sensor) was used.

Micro-computed tomography analysis
After the rabbit joint samples were harvested at 16 weeks post-operation, the joints were investigated by micro-computed tomography (micro-CT) scanning imaging system (SCANCO Medical AG, µCT 100, Bassersdorf, Switzerland) provided by Hangzhou Yuebo Biological Company.A 3D 180 • rotation scan was performed at a rotating angular velocity of 0.8 • s −1 to reconstruct the cylindrical regeneration area with a layer thickness of 18 µm using 60 kV scanning voltage.Quantitative analysis, including bone volume/tissue volume (BV/TV), trabecular number (Tb.N), and trabecular separation/spacing (Tb.Sp) was obtained.

Statistical analysis
GraphPad Prism software was used for statistical analysis.The difference between the two groups was analyzed by t-test.One-way analysis of variance with Tukey post-test was used when there were more than two variables.All results were presented as mean ± standard deviation.Statistical significance was defined as p values < 0.05.

Fabrication of the 3D-printed integrated bilayer scaffold
In this study, as shown in scheme 1, the bilayerintegrated hydrogel scaffolds were successfully fabricated via DLP 3D printing to accurately mimic the osteochondral architecture.To boost the efficacy of the cartilage repair, the growth factor KGN was added to the upper layer of the bilayer scaffolds.
Prior to the scaffold fabrication, as first reported by Van Den Bulcke et al [35], a portion of gelatin was modified with MA to produce GelMA with a 34% degree of substitution (figure 1(A)). 1 H-NMR analysis was used to observe lysine conjugation with MA.The lysine methylene proton (2H) peaks around 2.9 ppm (figure 1(A-b)) decreased significantly in the spectrum of GelMA, exhibiting complete interlocking of the lysine moiety with MA.Further, the appearance and enhancement of the peaks corresponding to the acrylic protons (2H, around 5.3-5.7 ppm, figure 1(Aa)) and the methyl protons (3H, around 1.8-2.1 ppm, figure 1(A-c)) of grafted methacrylamide groups were observed in GelMA samples [36].This confirmed the successful conjugation to MA.
Then, the bilayer scaffold was 3D-printed by DLP technology.Figure 1(D-a, b) showed the top and side views of the 3D-printed hydrogel scaffold, respectively.It exhibited rounded edges and clear boundaries.The scaffold immersed in PBS can maintain its overall shape for at least 14 d (supplementary figure 1).Meanwhile, the scaffold exhibited overlap of reloading and unloading curves across ten cycles, suggesting highly reversible compressibility and mechanical stability (supplementary figure 2).The surface morphology of the 'lotus-and radial-' scaffold (figure 1(E)) was further characterized by a two-photon confocal laser scanning microscopy, and the lotus-distribution pores with a diameter of 200 µm can be observed.Furthermore, as shown in figure 1(F), the radial-distribution pore structure of the upper scaffold and the abundant micropores inside the scaffold could be observed by scanning electron microscopy (SEM).

Structure and mechanical characterization of the 3D-printed integrated bilayer scaffold
To explore the structural flexibility of the scaffold, we prepared bilayer scaffolds with different longitudinal pore sizes: diameters of 100 µm, 200 µm, and 400 µm.The microscopy results (figure 2(A)) and the SEM images (supplementary figure 3) showed that the size of the channels can be precisely controlled.
Using this printing strategy, we prepared bilayer scaffolds with different concentrations of GelMA hydrogels.The SEM morphology (figures 2(B-a) and (C-a)) showed that both 15% and 10% GelMA hydrogels had high porosity and precisely defined interconnected porous structures.We analyzed the relationship between the load and compression deformation of two GelMA concentrations (10% and 15%) with different pore sizes (figures 2(B-b, c) and (C-b, c)).The compression modulus of 15% GelMA hydrogel scaffolds with pore sizes of 0, 100, 200, and 400 µm were 388.0 ± 17.52, 344.3 ± 16.5, 302.0 ± 15.00, and 203.3 ± 10.41 kPa, respectively; while the compression modulus of 10% GelMA hydrogel scaffolds with pore sizes of 0, 100, 200, and 400 µm were 178.0 ± 7.211, 157.3 ± 6.429, 139.0 ± 5.568, and 107.3 ± 3.055 kPa, respectively.3D-printed integrated bilayer scaffolds fabricated with 15% GelMA had higher mechanical properties compared to those printed with 10% GelMA.The scaffold with a pore size of 400 µm had the highest specific surface area up to 4.338 m 2 m −3 .The non-porous scaffold had a higher compressive strength (388.0 ± 17.5 kPa), but its specific surface area (1.895 m 2 m −3 ) was significantly lower than the other scaffolds.Considering the balance between mechanical property and surface area, a 15% GelMA hydrogel with a pore size of 200 µm (302.0 ± 15.0 kPa) was selected for further study.This hydrogel exhibited good mechanical properties and a relatively large specific surface area (2.861 m 2 m −3 ) (supplementary table 1), providing mechanical support and enough surface for cell attachment.
After determining the composition and pore size of the scaffold, the compressive mechanical properties of the upper and bottom layers of scaffolds were further investigated (figure 2(D)).The bottom layer of scaffold (558.4 ± 22.9 kPa) exhibited significantly higher compressive mechanical properties compared with the upper layer of scaffold (339.3 ± 9.1 kPa) ( * * * p < 0.001).Therefore, our strategy provided a functional bilayer scaffold with layer-specific strength, contributing to mimicking the anisotropy of natural osteochondral tissue.
Next, bovine serum albumin (BSA) was chosen as the model drug to test its release profile from scaffolds.As shown in figure 2(E), BSA was released slowly and steadily, reaching 66.23 ± 3.31% on day 7.This suggested that our scaffold fabricated with GelMA hydrogel can slowly release small molecules or drugs for synergetic effect.

In vitro biocompatibility of 3D-printed hydrogel scaffold
The CCK8 assay was performed to assess the cytotoxicity of GelMA hydrogel when cells were cultured in hydrogel extract liquid (figure 3(A)).The absorbance values of cell viability increased over 5 d in both groups.The results affirmed a significant rise in cell numbers over time, comparable to the control group.The result indicated that the GelMA hydrogel did not inhibit cell proliferation.
To further evaluate the cell compatibility, the live/dead staining test was performed (figures 3(B) and (C)).During the 7 day culture, almost no C3H cells were stained with red fluorescence, which indicated that GelMA hydrogel had no obvious toxicity.On the 7th day, the cell survival rate reached 99.63 ± 0.09%.It proved that the GelMA hydrogel had great biocompatibility and did not affect cell proliferation.As shown in figure 3(D), the cytoskeleton of C3H cells adhering to the hydrogel was stained in green with fluorescein isothiocyanate after culturing for 1, 4, 7 d, which showed that the cells spread well on the hydrogel.Taken together, the GelMA hydrogel was suitable for cell adhesion, spreading, and proliferation.

The 3D-printed hydrogel scaffold with precisely controlled and tunable structure facilitates cell infiltration and migration in vitro and in vivo
We further evaluated the effect of the structure of 3D-printed hydrogel scaffolds on cell migration and infiltration in vitro and in vivo (figure 4(A)).In the in vitro experiment, C3H cells stained with DiI were allowed to adhere to the surface of culture dishes for 24 h before the 3D-printed structural (P) and non-structural (N) hydrogel scaffolds were separately placed in the dishes.Confocal microscopy was used to observe cell migration into the scaffold after 1, 3, and 5 d.3D reconstructions of the confocal images (figure 4(B)) showed a significant increase of cells in the P group over time.Figure 4(C) showed cell distribution in hydrogel scaffolds after 1 d culture.Quantitative analysis (figure 4(D)) showed that the number of migrating cells in the P group (29.2 ± 6.6) was approximately four times greater than the N group (7.6 ± 0.1) ( * * p < 0.01).
We further implanted the 3D-printed structural and non-structural hydrogel scaffolds into rats subcutaneously to evaluate the cell infiltration in vivo (figures 4(E)-(G)).3D reconstructed images showed that 3 d after implantation in vivo, the number of cells in the P group (276.8 ± 12.1) was approximately 1.4 times higher than that in the N group (225.0 ± 11.1) ( * * p < 0.01).5 d after implantation, the number of cells in the P group (554.3 ± 22.2) was still significantly higher than that in the N group (457.8 ± 20.1) ( * * * p < 0.001).The above results confirmed that the 3D-printed structural hydrogel scaffold effectively promoted cell migration or infiltration into the scaffold.
In addition, we subcutaneously implanted different structural scaffolds in rats to compare their effects on cell infiltration.The hydrogel scaffolds were harvested 5 d after implantation.DAPI staining showed that both the number of cells and depth of migration were significantly enhanced when both 'lotus-' and 'radial-' pores were present (figures 5(A)-(C)).Quantitative analysis (figure 5(B)) showed that the maximum vertical infiltration distances of cells were 279.0 ± 56.9, 222.3 ± 63.4, 141.3 ± 68.0, and 25.6 ± 14.6 µm in the 3D-printed structural (P), only 'lotus-' (L), only 'radial-' (R), and non-structural (N, without any pores) groups, respectively.The maximum vertical migration distance of cells in the P group was significantly higher than in groups R and N.However, it was not significantly different from that in the L group, indicating the critical role of the 'lotus-' pores on the longitudinal migration of cells.Figure 5(C) shows that the number of cells in the P group (354.5 ± 12.2) was significantly higher than that in the L (223.5 ± 64.0), R (120.5 ± 43.9) and N (65.6 ± 26.9) groups.It demonstrated that the scaffold with both 'lotus-and radial-' distribution pore structures provided the best support for cell infiltration and migration into the scaffold.Next, we further observed that the cells infiltrated the scaffold mainly along the 3D-printed pore structures instead of elsewhere.Therefore, the scaffolds were sliced horizontally and vertically in two different directions (figure 5(D) and supplementary figure 4).The results of H&E staining of the P group (figure 5(D)) showed that, in both the horizontal and vertical directions, there were indeed more cells and greater migration depth in the pore structure (marked by the black dashed line) than that outside the pore area.Overall, the 3D-printed scaffold with 'lotus-and radial-' distribution pore structures facilitated cell migration to the interior of the scaffold along the established pores.

The 3D-printed hydrogel scaffold promotes cartilage regeneration
Following the results of the in vitro experiments, we evaluated the feasibility of cartilage repair using a 'lotus-and radial-' 3D-printed structural scaffold in a rabbit cartilage defect model (diameter = 4.0 mm, depth = 2 mm).Five different groups were investigated: blank group, non-structural group (N, the 3D-printed scaffold without any pore channels), 'lotus-' scaffold group (L, the 3D-printed scaffold with only 'lotus-' pores), 3D-printed structural scaffold group (P, the 3D-printed scaffold with both 'radial' and 'lotus-' pores), and P + KGN (the 3Dprinted structural scaffold combined with KGN) group.After 7 d and 6 weeks, the rabbits were euthanized to harvest the repaired knee joints for gross morphological observation and histological analysis.There was no mortality or infection in any animal after the operation.
From the representative histological evaluation of the repaired regions (figure 6(A)), we observed that none of the hydrogel scaffolds in the cartilage defect were degraded 7 d after surgery.Even so, there was better integration of the implanted hydrogel scaffolds in the P group with 'radial-' pores compared to the blank and N group, possibly because the 'radial-' pores facilitated the horizontal migration of cells.
In order to evaluate the therapeutic effects of different scaffolds, we analyzed the gross specimens at 6 weeks after surgery according to the ICRS macroscopic assessment scoring criteria (figures 6(B) and (C)).We observed that the defect site in the P + KGN group was relatively flat and well-repaired.Although the defect was still visible, its boundary with the surrounding area was blurred.The ICRS scores of the P + KGN group (10.67 ± 0.58) were significantly higher than the blank group (1.44 ± 0.29) and the N group (2.44 ± 0.91) ( * * p < 0.01).The P group, with a score of 8.56 ± 1.47, still had partial defects.On the contrary, there was no significant difference in the repair effect between the N and L group (6.22 ± 1.79) compared with the blank group (p > 0.05), where a clear boundary with the surrounding healthy cartilage could be seen.Figure 6(B) also showed images of H&E and SO/FG staining of the repaired cartilage after 6 weeks.Cluttered fibrous tissue was regenerated in the blank group, and the defect was not filled.In the N and L group, most of the hydrogel remained in the defect area, preventing the growth of new tissue.In contrast, in the P group, the defective cartilage site not only formed neo-tissue but also appeared SO staining positive area, indicating that the regeneration of the P group was started yet semimature and incomplete.Notably, the P + KGN group showed the best repair performance, with a large number of SO staining positive areas and typical round articular chondrocytes.These results showed that 3D-printed scaffolds with 'lotus-and radial-' distribution pores promoted cartilage regeneration, and the addition of KGN accelerated the regeneration process.

3D-printed bilayer hydrogel scaffold promotes osteochondral defect regeneration
Next, we evaluated the osteochondral regenerative capacity of the 3D-printed integrated bilayer scaffolds.Four different groups were investigated: blank group, non-structural group (N, the 3D-printed scaffold without any pore channels), 3D-printed structural group (P, the 3D-printed bilayer scaffold with both 'radial' and 'lotus-' pores), and P + KGN (the 3D-printed bilayer scaffold combined with KGN in the upper layer) group.Full-thickness osteochondral defects (diameter = 5.0 mm, depth = 4.0 mm) were made on the bilateral knee joints of rabbits (figure 7(A)).At sacrifice (8, 16 weeks postimplantation), no obvious infection or implantation response was found in the repaired tissue of the four groups of rabbits.Then, the regenerative capability of osteochondral tissues was assessed by gross morphological observations, histologic analyses, micro-CT, and mechanical tests.
As shown in figure 7(B), the gross morphological images showed that the osteochondral defects in the blank group exhibited a heavily cavitated surface at both 8 and 16 weeks.Compared to 8 weeks, the N group showed some regeneration at the defect site at 16 weeks but still exhibited a poor repair outcome.Blood infiltration could be observed as red areas in the P group at 8 weeks.At 16 weeks, the P group showed improved regeneration and integration, but the irregular surface still differed from natural cartilage.The P + KGN group exhibited optimal osteochondral repair as early as week 8, with the defects at 16 weeks being replaced with typical cartilage-like tissues.At 8 weeks, the ICRS scores of the blank, N, P, and P + KGN groups were 2.222 ± 0.7698, 2.889 ± 0.3849, 5.556 ± 1.018 and 6.556 ± 1.018, respectively.At 16 weeks, the ICRS scores were 6.778 ± 1.836, 8.778 ± 0.7698, 9.556 ± 0.3849, and 9.778 ± 0.5092 in the blank, N, P, and P + KGN groups, respectively.Based on macroscopic observations and the ICRS scoring system, the P + KGN group showed better repair than the N and P groups, especially at 8 weeks (figures 7(B)-(D)).
Micro-CT was used to assess the subchondral bone regeneration at 16 weeks after surgery.Figure 7(E) shows representative images of each group after 3D reconstruction.The formation of new subchondral bone tissue was most pronounced in the P + KGN group, where the defect area was almost completely filled.The P group also showed some improvement, but there was a significant cavity in the defect area.Admittedly, the cavity in the N group was larger, while the blank group only had very little new subchondral bone formation along the edges of the defect area.Quantitative analysis of the newly formed subchondral bone in the defect area confirmed the micro-CT reconstruction results (supplementary figure 5).The BV/TV value verified that the bone repair effect of the P + KGN group was significantly better than that of the blank group ( * p < 0.05).Compared to the other groups, the subchondral bone at the defect site was well repaired with or without KGN.These results indicated that the 3D-printed structural scaffolds had the potential to promote the regeneration of the subchondral bone.
We further evaluated the mechanical properties of the articular cartilage samples at 16 weeks postoperatively (figures 7(F) and (G)).The results showed that the reduced modulus of the cartilage treated with P + KGN (6.32 ± 0.45 GPa) was significantly greater than both the N group (1.30 ± 0.18 GPa) and the blank group (0.37 ± 0.02 GPa) ( * * * * p < 0.0001).There was no significant difference between the P + KGN group and the P group (5.59 ± 0.34 GPa) (p > 0.05).
Histological analysis of regenerated osteochondral tissue after 8 and 16 weeks was performed by H&E and SO/FG staining in figure 8.At 8 weeks, the defect area in the blank group was only filled with fibrous tissue, and a small amount of matrix staining was observed.In other groups, the implanted hydrogels were still visible at 8 weeks.Regeneration of cartilage and bone was delayed due to the slow degradation of the hydrogel.All groups had more tissue fillings and reduced defect area at 16 weeks compared to 8 weeks.In the blank group, the regenerated surface was rough and unsmooth, with massive fibrous tissue formation.In the N group, despite most of the defect sites being filled, we observed many fat-like structures.The surface cartilage layer of the P group was thin, and the regeneration of cartilage was discontinuous.Notably, the thickness and surface smoothness of regenerated cartilage in the P + KGN group were better than those in the P group.A large number of newly formed, regularly arranged hyaline chondrocytes can be observed.In addition, the cartilage-bone interfaces of the P + KGN group were more regular than those of the P group.The results showed that the bilayer printed scaffolds releasing KGN could further promote the repair of the cartilage layer in osteochondral defect repair.

Discussion
Osteochondral damage caused by sports or lesions can lead to mechanical instability and loss of joint function, and severe osteochondral damage can develop into osteoarthritis if left untreated.Bilayer scaffolds provide two different microenvironments for cartilage and subchondral bone, which are preferable for osteochondral tissue engineering [37].Meanwhile, a strong bond between cartilage and bone is particularly challenging.Scaffolds composed of two separate phases present a potential risk of delamination [38].Recently, Browe et al prepared anisotropic scaffolds with larger pores by a modified freeze-drying method to improve tissue repair outcomes in a goat model of osteochondral defects [39].However, techniques such as freeze-drying do not allow precise control of the pore size within the scaffold, which makes it impossible to prepare the scaffold with consistency [40].Light-assisted DLP 3D printing technology is an excellent candidate to provide superior resolution and speed for the precise construction of a wide range of complex structures [41].In this study, we used DLP and GelMA hydrogels to print a 3D integrated bilayer scaffold and realized the integrated construction of a 'lotus-and radial-' structure containing KGN in the upper layer and a 'lotus-' structure in the bottom layer.This strategy has yielded satisfactory results in the regeneration of rabbit osteochondral defects.
This strategy offers some advantages over scaffolds constructed by conventional techniques.Firstly, we established a method for 3D printing cell-free templated osteochondral bilayer scaffolds containing precisely controlled internal structures.It has been demonstrated by previous studies that porous scaffolds mimicking native tissue structures could greatly enhance the regular deposition of extracellular matrix (ECM) and tissue regeneration in situ [42][43][44].
As expected, cell migration was well guided in our 3D-printed porous scaffold to facilitate osteochondral regeneration.The longitudinal 'lotus-' pores of this scaffold promote the upward migration of BMSCs from the bone marrow.However, it has been showed that BMSCs are less inclined to differentiate into cartilage tissue [17,18].The transverse 'radial-' pores were designed in order to further allow chondrocytes or cartilage stem/progenitor cells (CSPCs) to participate in the regeneration process of the cartilage layer.It has been reported that when cartilage was damaged, CSPCs emerged and migrated to the site of injury [45][46][47].CSPCs migration is believed to be a major factor in the healing process of osteochondral defects.Therefore, CSPCs recruitment capacity should be considered in the scaffold design of cartilage tissue engineering [2,20,48].In addition, the integrated printing of the scaffolds by DLP technology can solve the problem of layer separation that may occur in traditional twolayer scaffolds.By virtue of the integrity of our scaffold, the composition of the hydrogels can be changed in the future to achieve better restorative results.
Appropriate mechanical support is essential for biomimetic scaffolds, which should be delicately tuned to promote osteochondral regeneration while avoiding the collapse and failure of newly formed tissue [11].The porous structure of the scaffolds has been proven to facilitate nutrient and oxygen transport and metabolic waste removal [49].Inevitably, the existence of interconnected pores inside the scaffolds will compromise the mechanical properties [50].To obtain the best combination of ink concentration and printing structure, we compared the parameters such as mechanical strength and relative surface area.Zhang et al proposed an optimal pore size of 150-250 µm for scaffolds to enhance collagen II deposition and mechanical properties of the generated cartilages [51].Therefore, in this study, 200 µm was chosen as the pore size of the 'radial-' pores of the upper cartilage scaffold.Then, scaffolds were 3Dprinted with different 'lotus-' pore sizes (0, 100, 200, and 400 µm) using GelMA at 10% and 15% concentrations, respectively.The 3D-printed scaffolds were all well-defined at the edges and with clear pore diameters, providing the possibility that the structure of our scaffolds is finely tunable.Based on the mechanical analysis, we found that scaffolds with different pore sizes exhibited different mechanical properties: as the pore size becomes larger, the mechanical property of the scaffold decreases.In addition, we found that the pore size affected the compressive strength of the scaffolds, and the specific surface area was also closely related.Larger pore size also causes higher specific surface area.Scaffolds with larger specific surface areas can increase cell interactions and provide more binding sites for cell attachment [52][53][54].Previous reports have shown that the Young's modulus of GelMA increases with increasing GelMA concentration [29,55], which is consistent with our results.Based on the above considerations, we finally chose to use 15% GelMA to print an integrated double-layer scaffold with both 'lotus-' and 'radial-' pores of 200 µm.Mechanical tests were performed on the upper and bottom layers of the scaffold, respectively, and the results showed that the integrated double-layer scaffold we designed had layer-specific mechanical properties, providing the possibility of mechanical tuning to better match the natural osteochondral tissue.
In addition to providing favorable mechanical support, the 3D-printed integrated bilayer scaffolds are also able to provide a favorable microenvironment for cell growth.Biocompatibility is the primary criterion for tissue-engineered scaffolds [56].GelMA, a biomaterial with superior biocompatibility, has been broadly used in tissue engineering [57][58][59].The porosity of GelMA facilitates cell penetration, vascularization, and nutrient delivery, as well as increasing the metabolic activity of osteoblasts [60,61].In our current study, GelMA hydrogels provide a matrix for cell attachment and proliferation.GelMA hydrogels have been proven effective in delivering drugs and releasing growth factors in tissue engineering [62][63][64].We then used KGN in the upper layer scaffold to promote cell differentiation into chondrocytes [26,27].Compounding different growth factors in the upper and bottom layers provides a suitable microenvironment for the regeneration of different tissues.
The 3D integrated bilayer scaffolds provide an efficient way for cells to infiltrate and migrate into the scaffold.In vitro, we stained the cells with DiI and quantified the infiltration of cells.It was found that the number of cells infiltrating in the 3D-printed group was almost four times of that in the nonstructural group (figures 4(B) and (C)).Moreover, the number of cells in the 3D-printed scaffolds increased with time (figure 4(A)).Possible reasons for this phenomenon are the increased number of cells migrating inward and the proliferation of cells in the hydrogel.And in vivo, we confirmed a similar finding that the 3D-printed group facilitated cell migration into the scaffold.These cells that migrated into the scaffolds may differentiate into bone or cartilage tissue, respectively, in the in-situ microenvironment [65].Further, we comparatively explored the role of channels in different directions.The results showed that longitudinally, the 'lotus-' pores played a decisive role in the depth of cell infiltration, considering that there was no significant difference in the depth of cells between the P and L groups.While in the R and N groups, it was difficult for cells to penetrate the scaffolds because of the absence of pore channels.This is consistent with the findings in the previously reported article [66].Radially oriented scaffolds have been showed to be more effective for cell and tissue ingrowth, nutrient and waste exchange, ECM deposition, and cellular interactions [20,67,68].These suggested that our 3D-printed integrated scaffolds provide strong support for osteochondral regeneration.
Animal experiment results demonstrated an improved repair for both the subchondral bone and articular cartilage in 3D-printed integrated bilayer scaffolds combined with KGN, compared to the blank group and N group.Although 3D-printed integrated bilayer scaffolds with KGN resulted in superior histological structure, no significant difference in biomechanical property was found between 3Dprinted scaffolds with or without KGN 16 w after surgery.The 3D-printed integrated bilayer scaffold reconstructed the function of the osteochondral bone with a reduced elastic modulus similar to that reported for natural osteochondral tissue in our previous study [21].On this basis, the addition of KGN further enhances cartilage repair.Based on our results and the available literature [69], we propose the following possible reasons.First, the stem cells, such as BMSCs or CSPCs, are recruited from their native niches to the injury site due to the 'lotus and radial' pores in the printed scaffolds.The biomimetic bilayer scaffolds provide support for the attachment, proliferation, and differentiation of these cells into osteoblasts or chondrocytes.Then the bioactive factor KGN is released from the upper layer of P + KGN scaffolds, promoting chondrogenic differentiation and accelerating cartilage regeneration.This finding was consistent with the in vitro study results, highlighting the important role of the bilayer 3D-printed scaffold with porous structure and growth factor release in the process of cartilage regeneration.
However, we also found that the complex internal layered structure within the osteochondral tissue has not been perfectly mimicked yet.More structurally complex multi-layer scaffolds can be further designed and applied.Furthermore, although the mechanical properties of the upper and bottom layers of the bilayer scaffold are significantly different and the reduced modulus of repaired cartilage tissues in vivo is close to that of the normal cartilage, considering the weak mechanical properties of GelMA itself, they are still not comparable to natural tissues, especially the mechanical property of bottom layer is orders of magnitude lower than that of subchondral bone.The mechanical properties can be further improved in the future by compounding with other materials to better match the stiffness of cartilage and subchondral bone.For example, incorporating nHAs or nano-oligomers into the bottom layer is under investigation in our group.Meanwhile, more types of growth factors, such as BMP2 (bone morphogenetic protein-2) and VEGF (vascular endothelial growth factor), can also be compounded into the inks.

Conclusion
In summary, a 3D-printed integrated bilayer scaffold consisting of a 'lotus-and radial-' porous cartilage layer combined with KGN and a 'lotus-' porous subchondral layer was successfully constructed for osteochondral repair via DLP printing.Our results showed that the printed integrated bilayer scaffold had tunable mechanical properties, good biocompatibility and could promote cell infiltration and migration into the scaffold.In vivo animal evaluation further validated that this 3D-printed integrated bilayer scaffold compounded with KGN can successfully induce osteochondral repair and facilitate regeneration of new articular cartilage and subchondral bone.This composite design will be a potential osteochondral repair scaffold and deserves to be further investigated.

Scheme 1 .
Scheme 1. Schematic illustration of 3D-printed biomimetic scaffolds for the regeneration of osteochondral defect.(A) The compositions of ink-A and ink-B.(B) Schematic of the scaffold prepared by digital light processing (DLP) technique.(C) 3D-printed model images of the bilayer scaffold, including the upper and bottom layers of the scaffold.(D) The regeneration of osteochondral defects was treated with the integrated bilayer scaffold with precisely controlled structures in the animal experiment.(E) Chemical structure of Kartogenin (KGN) and gelatin methacrylate (GelMA).

Figure 3 .
Figure 3. Evaluation of cytotoxicity and biocompatibility of the 15% GelMA hydrogel.(A) Comparison of cell proliferation in the control and GelMA groups for 5 d.(B) Live/dead staining and (C) quantification analysis of C3H cells cultured on the GelMA hydrogel for 7 d (scale bar = 200 µm).(D) F-actin and nuclear staining of C3H cells on the GelMA hydrogel for 7 d (scale bar = 50 µm).(n = 3).

Figure 5 .
Figure 5. Evaluation of the pore distributions of the bilayer scaffolds.(A) Comparison of cell infiltration in different scaffolds in vivo at 5 d (scale bar = 200 µm).(B) The maximum vertical distance and (C) number of infiltrating cells were quantified by counting.(D) Images of illustration and H&E staining of the P group's scaffold at 5 d in the (a) horizontal and (b) vertical directions (scale bar = 100 µm; the pore structures in the magnified images are marked by black dashed lines).(n = 5, * p < 0.05, * * p < 0.01, * * * p < 0.001, * * * * p < 0.0001).

Figure 8 .
Figure 8. Histological analysis of the regenerated osteochondral tissues.(A) H&E and (B) SO/FG staining of the repaired cartilages after different treatments for 8 and 16 weeks.(Scale bar in the large image = 2 mm; scale bar in the small image = 400 µm).