Three-dimensional-printed polycaprolactone scaffolds with interconnected hollow-pipe structures for enhanced bone regeneration

Abstract Three-dimensional (3D)-printed scaffolds are widely used in tissue engineering to help regenerate critical-sized bone defects. However, conventional scaffolds possess relatively simple porous structures that limit the delivery of oxygen and nutrients to cells, leading to insufficient bone regeneration. Accordingly, in the present study, perfusable and permeable polycaprolactone scaffolds with highly interconnected hollow-pipe structures that mimic natural micro-vascular networks are prepared by an indirect one-pot 3D-printing method. In vitro experiments demonstrate that hollow-pipe-structured (HPS) scaffolds promote cell attachment, proliferation, osteogenesis and angiogenesis compared to the normal non-hollow-pipe-structured scaffolds. Furthermore, in vivo studies reveal that HPS scaffolds enhance bone regeneration and vascularization in rabbit bone defects, as observed at 8 and 12 weeks, respectively. Thus, the fabricated HPS scaffolds are promising candidates for the repair of critical-sized bone defects.


Introduction
Bone is a highly branched vascularized system that can deliver oxygen and nutrients to cells and remove metabolites in the repair of damaged tissues [1,2]. Rapid vascularization not only allows the delivery of oxygen and nutrients for cell proliferation but also supplies required ions and growth factors for bone regeneration [3,4]. Therefore, vascularization plays a pivotal role in bone regeneration. Clinically, the reconstruction of critical-sized bone defects is a worldwide problem [5]. Current therapies, which include allografting, autografting and other bone transport techniques, are limited in availability and may lead to the development of chronic pain [6,7]. Accordingly, three-dimensional (3D) bioprinting technology has recently emerged as a potential strategy for the repair of critical-sized bone defects resulting from infections, surgery, trauma and congenital malformations [8]. However, conventional scaffolds possess relatively simple porous structures and thus exhibit poor delivery of nutrients and oxygen, low porosity and poor vascularization performance, thus limiting the generation of new bone inside the scaffold [9][10][11][12]. Moreover, due to the limited diffusion of oxygen and nutrients, the inner regions of larger 3D-printed constructs are prone to necrosis, which induces an inflammatory response in vivo [13,14].
To facilitate bone regeneration, studies focusing on the fabrication of vascular-like constructs using various microfabrication techniques, including sacrificial molding, laser-piercing and wire-array templating, have been conducted [15][16][17]. Accordingly, these different groups have demonstrated that scaffolds with hollow channels can promote vascularization and bone regeneration in critical-sized bone defects [16][17][18]. However, most of these scaffolds achieved limited success due to poor fabrication efficiency, low porosity, and problems associated with biodegradation, and biocompatibility [19][20][21]. In addition, these approaches cannot be easily used to make 3D constructs and often require multiple complex steps. Several studies focused on the creation of hollow channels in scaffolds using modified coaxial 3D-printing methods to improve bone regeneration [15,[22][23][24][25]. However, the hollow channels in these scaffolds were partially interconnected in different layers. Furthermore, these channel surfaces was often smooth or the porous structure in the channel surface was not controllable. Therefore, they did not allow perfusion throughout the whole scaffold, which further limited their success in bone regeneration. Moreover, previous studies showed that microfluidic systems containing microchannels could promote the formation of rudimentary vasculatures in vitro [23,26,27]. However, such systems are constructed from hydrogels, which are difficult to anastomose to the host, degrade rapidly after implantation and have poor mechanical properties [28,29].
In this study, we attempted to combine the use of one-pot 3D-printed sacrificial caramel templates and polymer coating with integrated phase separation to prepare a highly interconnected hollow-pipe-structured (HPS) scaffold for regenerating bone defects. These HPS scaffolds mimic the perfusable and permeable channel structure of natural micro-vascular networks to provide enhanced delivery of oxygen and nutrients to the whole scaffold. More importantly, we used the ubiquitous phase separation mechanism to readily introduce extensive and controllable micropores in the channel walls, which is seldom reported previously. The as-prepared scaffolds possessed distinctly improved porosity for cell adhesion and proliferation. In particular, our in vitro studies demonstrated that HPS scaffolds were beneficial for osteogenesis and angiogenesis, while in vivo studies revealed that HPS scaffolds enhanced bone tissue formation and vascularization of rabbit bone defects. Moreover, more bone tissues were observed in the inner regions of HPS scaffolds than in corresponding regions of non-hollow-pipe-structured (NHPS) scaffolds. Thus, we have developed a smart scaffold that can effectively regenerate critical-sized bone defects and may be applied clinically in the future.

Materials and methods
Fabrication and characterization of HPS and NHPS scaffolds HPS scaffolds were generated by the indirect one-pot 3D-printing method in a similar strategy developed in our previous study [30,31] (Fig. 1). First, a caramel-based template was printed using fused deposition modeling (FDM). Sucrose was preheated to 150 C for 30 min in a 3D printer (HTS-400; Fochif Mechatronics Technology, China), and the caramel ink was printed into 12-layer constructs at 130-135 C. The center-to-center distance between the filaments was 1.3 mm with 0 /90 lay-down patterning between two successive layers; the height of every layer was 0.5 mm; and the nozzle size was 20 G. Polycaprolactone (PCL; Mn 80 000 g mol À1 , Sigma-Aldrich) was dissolved in hexafluoroisopropanol (HFIP; analytical grade, >99%, Sigma-Aldrich) at 4% (w/v). The caramel-based template was soaked in the PCL solution for 1 min, and the HFIP solvent was evaporated for 30 min under ambient conditions. Then, the PCL-coated template was soaked in distilled water for 6 h to remove the caramel and residual solvent. The water was replaced every 1 h. Next, the PCL HPS scaffold was lyophilized. NHPS scaffolds were directly printed by FDM at 110 C with PCL using the same parameters.
To observe the detailed structures of HPS and NHPS scaffolds, they were subjected to scanning electron microscopy (SEM; JSM-5600LV, Jeol Ltd, Japan) in the side-and top-view orientations. First, scaffolds were sputter-coated with gold. Cutting the scaffold transversely revealed the filament cross-section. The porosities of the scaffolds were determined using a previously reported liquid displacement method [32].

Live/dead staining
The viabilities of the cells on the scaffolds were determined using live/dead kits (Invitrogen, UK). Briefly, 6 Â 10 5 BMSCs were seeded vertically on the surfaces of scaffolds in 24-well plates and then cultured in DMEM containing 10% (v/v) fetal calf serum and 1% (v/v) penicillin-streptomycin for 3 or 7 days. Then, 2 Â 10 À6 mol calcein AM and 4 Â 10 À6 mol EthD-1 solutions were added to 2 ml phosphate buffered saline (PBS), and 1000 lL of this solution was added to the cell-seeded scaffolds in 24-well plates. After incubating for 30 min at room temperature, the scaffolds were gently washed with PBS solution. Images were obtained by laser confocal microscopy (LSM700, Zeiss, Germany). The viable cells appeared green, whereas dead cells appeared red.

DAPI staining
BMSCs (6 Â 10 5 ) were seeded vertically on the surfaces of scaffolds in 24-well plates and then cultured in DMEM containing 10% (v/v) fetal calf serum and 1% (v/v) penicillin-streptomycin. After culturing for 3 days, the cell nuclei were stained with DAPI (Beyotime, China) and incubated for 1 h at room temperature. Fluorescence images were captured using a fluorescence microscope (ZEISS, Axio, Germany). Cell numbers were determined using Image J software (National Institutes of Health, USA).

SEM analysis of scaffolds
BMSCs cultured in scaffolds were observed by SEM. First, 6 Â 10 5 BMSCs were seeded vertically on scaffolds in 24-well plates and then cultivated in DMEM containing 10% (v/v) fetal calf serum and 1% (v/v) penicillin-streptomycin for 7 days. Next, the cellseeded scaffolds were immersed in 4% paraformaldehyde for 1 h. Before SEM observation (S4800; HITACHI, Japan), the scaffolds were sputter-coated with gold.

Alkaline phosphatase activity
The differentiation of BMSCs in the early stage of osteogenesis was evaluated by alkaline phosphatase (ALP) activity. First, 6 Â 10 5 BMSCs were seeded vertically on scaffolds in 24-well plates and then cultivated in an osteoinductive medium containing 50 mg/ml ascorbic acid, 10 mM b-glycerol phosphate and 10 À8 M dexamethasone (Cyagen Biosciences, America) for 7 or 14 days. Then, BMSCs were lysed using 100 lL of RIPA lysis buffer. The ALP activity of the BMSCs was evaluated with an Alkaline Phosphatase Assay Kit (Beyotime, China) according to the manufacturer's protocol. Briefly, BMSCs were incubated with the assay buffer and the samples were centrifuged to eliminate insoluble substances. Then, p-nitrophenylphosphatase liquor was mixed with the cell lysate, and the mixed solution was incubated at 37 C for 30 min. The concentration of p-nitrophenyl was determined by measuring the absorbance at 405 nm using a microplate reader (Synergy 2; BioTek, USA) and analyzing the result using a standard curve.

Quantitative real-time polymerase chain reaction analysis
To evaluate the mRNA expression levels of osteogenic genes (COL-I, OCN, OPN and RUNX2), BMSCs were processed for total RNA extraction using an RNA prep Micro Kit (TaKaRa, Japan) at 7 or 14 days. First, 6 Â 10 5 BMSCs were seeded vertically on the surfaces of scaffolds and then cultivated in an osteoinductive medium. Then, the mRNA expression levels of osteogenic genes were assessed by quantitative real-time polymerase chain reaction (RT-PCR) analysis. Specifically, total RNA was extracted from the BMSCs using the Trizol reagent (Ambion, CA) following the manufacturer's instructions. Thereafter, complementary DNA was synthesized using All-In-One RT MasterMix (ABM, CA). After reverse transcription, RT-PCR was performed using EvaGreen 2 Â qPCR MasterMix (ABM, CA) on a Light Cycler 480 II (Roche, CHE).
To evaluate the expression of angiogenesis genes (HIF1-a and VEGF), human umbilical vein endothelial cells (HUVECs) were purchased from AllCells. First, 6 Â 10 5 HUVECs were seeded vertically on the scaffolds and cultured in an endothelial basal medium. Then, the same procedure used to evaluate osteogenic genes was also used for angiogenesis-specific genes. All primer sequences were designed using the software Primer 5.0 from the NCBI database. Primer sequences: rabbit, GAPDH 5 0 -TCA CCATCTTCCAGGAGCGA-3 0 and 5 0 -CACAATGCCGAAGTGGTCGT-

Bone defect model
All animal experiments were conducted following the relevant laws and guidelines. After receiving the approval of the ethics committee of Nanjing Medical University, New Zealand, rabbits were purchased from Nanjing First Hospital Animal Center for in vivo experiments. Both the posterior limbs of rabbits were used for defects. Defects were treated with HPS scaffolds, NHPS scaffolds or left untreated (no scaffold) as a control (CTR). First, rabbits were anesthetized through intramuscular injection. After exposing the femoral epicondyle by lateral incision, critical size defects (6 mm diameter, 6 mm height) were transversally created at the femoral epicondyle. HPS or NHPS scaffolds with a diameter of 6 mm and a height of 6 mm were implanted into the defects. The rabbits were sacrificed 8 or 12 weeks after implantation for in vivo evaluation.

Micro-CT analysis
The samples were removed from the rabbits for Micro-CT analysis (Siemens Inveon, Siemens Medical Solution, Germany) immediately after sacrifice. New bone regeneration was distinguished from other tissues by gray value with a threshold of 1500-2800. The new bone volume fractions in the defects (BV/TV, new bone area/total area) and the bone mineral densities (BMDs) of the samples were determined.

Histological analysis
All samples were immersed in 4% paraformaldehyde for 48 h and then decalcified in 5% formic acid for 3 days. Then, these samples were embedded in paraffin and cut into 5-lm-thick sections using a microtome. For histological analysis, hematoxylin and eosin staining was performed on the sections. For the immunofluorescence staining analysis of in vivo samples, bone sections were incubated overnight with primary antibodies of HIF-1a antibody (Beyotime, China), CD31antibody (Beyotime, China) and DAPI (Beyotime, China). The slides were analyzed using a fluorescent stereomicroscope and digital camera (Olympus BX-53, Tokyo, Japan). Signal intensities were determined using Image J software (National Institutes of Health, USA).

Statistical analysis
All data were expressed as mean 6 standard deviations and analyzed using one-way analysis of variance. P < 0.05 was considered statistically significant.

Results and discussion
Fabrication and characterization of HPS scaffolds HPS scaffolds have controllable multi-stage bionic structures, including an interconnected branch network and permeable channel walls with microporous structures (Fig. 2). The SEM images of HPS scaffolds showed that the frameworks had three-level hierarchical structures (Fig. 3). The primary structure of the framework was formed by vertically stacked hollow channels ( Fig. 3a  and b). The crisscross pattern of the HPS scaffold was closely aligned with that of the caramel template. The secondary microchannel structure was continuous and connected seamlessly to adjacent perpendicular layers at numerous cross-points (Fig. 3e). Removal of the caramel template gave rise to integrated and interconnected microchannels that support physiologically relevant perfusion to mimic microvasculature for mass transfer in three dimensions (Fig. 3c). Furthermore, the micropore-rich tertiary structure, which was generated by phase separation during solvent evaporation, was distributed throughout the thin channel walls (Fig. 3d and f). These micropores showed relatively uniform pore sizes of 5.48 6 1.80 lm, and the thickness of hollow channel wall was 13.76 6 3.18 lm. The phase separation process has been widely used to build porous structures. In a previous study, we investigated the coating process via the phase separation mechanism in detail including different molecular weight, concentrations and solvents of PCL. These factors had significant effects on the morphology of the resultant scaffolds [30]. The permeable porous channel walls enable the osmotic exchange of nutrients and waste between the inside and pores of the channels.
A liquid displacement method was used to measure the porosities of HPS and NHPS scaffolds, revealing porosities of 91% and 55%, respectively (Fig. 2e). Thus, the porosity of HPS scaffolds was significantly enhanced compared to that of NHPS scaffolds. HPS and NHPS scaffolds had similar structures in terms of frame structure and stacking layers (Fig. 3h), therefore, the porosities resulting from the interstitial space between the channels of HPS scaffold and between the filaments of NHPS scaffold were similar. However, unlike the solid filaments and smooth surface of NHPS scaffold (Fig. 3i and j), the channels of HPS scaffold were hollow with porous and thin walls. Such structural differences led to a significant difference in porosity.

Adhesion and proliferation on HPS and NHPS scaffolds
3D-printed PCL scaffolds are nontoxic, slowly degrading and biocompatible, so they are widely used in bone tissue engineering [34,35]. The CCK-8 results for BMSCs (Fig. 5a) showed that cell proliferation of HPS scaffold was significantly greater than that of NHPS scaffolds on each day of measurement. We speculate that the perfusable and permeable features of HPS scaffolds, as well as their higher porosity, contribute to the higher cell proliferation levels. Previous studies have demonstrated that improved porosity has a positive impact on cell adhesion and proliferation, and thus in vivo bone regeneration [36,37]. In addition, HPS scaffolds had channels that are open and interconnected from different directions, which may also be beneficial for cell attachment and proliferation.
The SEM images of BMSCs adhered to HPS scaffolds (Fig. 3d) and NHPS scaffolds (Fig. 3k) after incubation for 7 days revealed that the cell numbers and densities were much higher for HPS scaffolds. Furthermore, the SEM images revealed the close cellcell contact of BMSCs in HPS scaffolds. In particular, the BMSCs adhered to the inner surfaces of the hollow struts and proliferate. In contrast, few cells were observed on the surfaces of the NHPS scaffolds.
To further assess cell viability, live/dead staining was performed. As shown in Fig. 4, after culturing for 3 days, dense and numerous BMSCs remained alive along the HPS scaffold struts.
Almost no dead cells were observed, and more living cells were observed on Day 7 on HPS scaffolds. By comparison, much less cells were observed on NHPS scaffolds on both Day 3 and Day 7. Consistent with live/dead staining, the result of DAPI also confirmed more cells on HPS scaffolds.  SEM and confocal laser scanning microscopy images clearly revealed the improved attachment and proliferation of BMSCs in HPS scaffolds compared to NHPS scaffolds. Thus, we can infer that HPS scaffolds provide a better microenvironment for BMSCs to adhere and proliferate. These results suggest that the integrated and interconnected hollow channels and porous channel walls in HPS scaffolds are beneficial for oxygen and nutrient distribution through the scaffolds and thus cell adhesion, spread and proliferation are improved.

Osteogenesis and angiogenesis in HPS scaffolds
As shown in Fig. 5, the expression of osteogenesis-related genes was significantly upregulated on HPS scaffolds on both Day 7 and Day 14 compared to that on NHPS scaffolds. Moreover, HPS scaffolds significantly improved ALP activity, which indicated early mineralization and osteogenic differentiation in the BMSCs. Immunofluorescence staining of OCN and OPN further confirmed osteogenesis was remarkably enhanced on HPS scaffolds (Fig. 6).
The upregulation of bone-related genes may be due to the fact that the dense cell areas lead to increased opportunities for cell-cell contact, which facilitates osteoblastic differentiation [38]. The SEM images of BMSCs adhered to HPS scaffolds clearly revealed higher cell density and number as well as close cell-cell contact between the BMSCs. Next, we investigated the angiogenic effect of HPS scaffolds on HUVECs. HIF1-a is essential for both angiogenesis and osteogenesis, and it activates VEGF, an important angiogenic factor that can attract blood vessels [39][40][41]. The RT-PCR results in Fig. 5 demonstrated that HIF1-a gene expression of HPS scaffolds was significantly higher than that of NHPS scaffolds on both Day 7 and Day 14. Similarly, VEGF expression was also remarkably upregulated on HPS scaffolds on both Day 7 and Day 14.
In addition, Immunofluorescence staining of VEGF and CD31 further confirmed angiogenesis was obviously enhanced on HPS scaffolds compared to NHPS scaffolds (Fig. 6). It has been reported that endothelial cells are more likely to interact with surfaces that have high porosity [42,43]. Moreover, previous studies demonstrated that interconnected pores enhance the formation of vascular networks and offer channels for the distribution of ions, nutrients and cells [25]. Thus, the improved cell adhesion and proliferation and unique structure of HPS scaffolds could contribute to angiogenesis. In summary, the 3D-printed HPS scaffolds facilitate osteogenesis and angiogenesis in vitro.

In vivo evaluation
The Micro-CT images in Fig. 7 showed that the defects in the CTR group remained mostly empty at 8 weeks and were not fully regenerated at 12 weeks, with only limited new bone formation at the edge surrounding the defect when observed in the coronal view. For the NHPS group, new bone was formed at the edge surrounding the scaffold and limited new bone formation in the inner part of the porous scaffold was observed in the coronal view at 12 weeks. However, for the HPS group, new bone ingrowth was observed throughout the whole defect. At 8 weeks, extensive bone formation was observed in the inner part of the scaffold, and more bone regeneration was observed at 12 weeks. The Micro-CT images indicated that HPS scaffolds not only stimulated new bone tissue formation around the scaffold but also facilitated bone formation in the core of the scaffolds. In addition, the BV/TV and BMD values for the HPS group were significantly higher than those for the NHPS and CTR groups at both time points. Therefore, the unique structure of the HPS scaffolds promotes bone regeneration more effectively.
The HE staining images in Fig. 8 revealed that new bone formation for HPS group was significantly more extensive than that in the NHPS and CTR group. Importantly, there were more newly formed vessels and osteoblasts in the HPS scaffold pores compared to NHPS and CTR group. Red blood cells, which facilitate bone formation, could be observed in the newly formed vessels in the HPS scaffold. In addition, some parts of the PCL struts in the HPS scaffold had degraded, while the slow degradation of NHPS scaffold restrained new bone growth in the defect. Comparatively, fibrous tissues were observed in the CTR defects. Moreover, we also observed higher expression of both HIF-1a and CD31 in the HPS scaffold than NHPS scaffold based on immunofluorescence staining images (Fig. 9). HPS scaffolds are vascularlike scaffolds, which contain plenty of interconnected channels. Vascular-like scaffolds have been demonstrated to be a successful strategy to enhance rapid vascularization both in vitro and in vivo, which further improve formation of new bone tissue [18,23,44]. Channels that resemble physiologic microvasculature support vascularization. In addition, compared with NHPS scaffolds, HPS scaffolds have a higher porosity and rougher texture, which contribute to osteogenesis [45,46]. In summary, the highly interconnected structure and permeable walls of the HPS scaffolds promote vascularization and bone regeneration in vivo.

Conclusions
In this study, we fabricated perfusable and permeable PCL scaffolds with highly interconnected hollow channels via an indirect  one-pot 3D-printing method. The unique interconnected HPS scaffolds significantly enhance cell adhesion, spread and proliferation as well as osteogenic differentiation and angiogenesis. Thus, these structures show promise for application in cell delivery and bone regeneration. In addition, this strategy can be used to prepare other polymer scaffolds, indicating its potential for tissue engineering, mechanical engineering, catalysis and environmental remediation.