Development and optimisation of hydroxyapatite-polyethylene glycol diacrylate hydrogel inks for 3D printing of bone tissue engineered scaffolds

In the event of excessive damage to bone tissue, the self-healing process alone is not sufficient to restore bone integrity. Three-dimensional (3D) printing, as an advanced additive manufacturing technology, can create implantable bone scaffolds with accurate geometry and internal architecture, facilitating bone regeneration. This study aims to develop and optimise hydroxyapatite-polyethylene glycol diacrylate (HA-PEGDA) hydrogel inks for extrusion 3D printing of bone tissue scaffolds. Different concentrations of HA were mixed with PEGDA, and further incorporated with pluronic F127 (PF127) as a sacrificial carrier. PF127 provided good distribution of HA nanoparticle within the scaffolds and improved the rheological requirements of HA-PEGDA inks for extrusion 3D printing without significant reduction in the HA content after its removal. Higher printing pressures and printing rates were needed to generate the same strand diameter when using a higher HA content compared to a lower HA content. Scaffolds with excellent shape fidelity up to 75-layers and high resolution (∼200 µm) with uniform strands were fabricated. Increasing the HA content enhanced the compression strength and decreased the swelling degree and degradation rate of 3D printed HA-PEGDA scaffolds. In addition, the incorporation of HA improved the adhesion and proliferation of human bone mesenchymal stem cells (hBMSCs) onto the scaffolds. 3D printed scaffolds with 2 wt% HA promoted osteogenic differentiation of hBMSCs as confirmed by the expression of alkaline phosphatase activity and calcium deposition. Altogether, the developed HA-PEGDA hydrogel ink has promising potential as a scaffold material for bone tissue regeneration, with excellent shape fidelity and the ability to promote osteogenic differentiation of hBMSCs.


Introduction
The limitations of existing bone grafting procedures and the rising incidences of bone and joint disorders have necessitated the development of scaffold-based tissue engineering strategies as an alternative treatment for bone regeneration.For bone tissue engineering, the characteristics of the three-dimensional (3D) design along with the material properties of a scaffold are critical.In this regard, additive manufacturing (AM), and particularly 3D printing technologies, have shown promising advantages over conventional methods (e.g.solvent casting and electrospinning) for fabricating complex 3D tissue engineered scaffolds.The advantages of this technology are the precise control over the architectural features of the scaffolds, fabricating customised and patient-specific constructs with interconnected pores to allow cell migration and waste and nutrient transportation, as well as direct printing of various materials such as metals, ceramics, cells and biomolecules [1,2].3D printing technologies use computer aided design (CAD) to generate a 3D model, convert it to a Standard Triangulate Language format, and eventually print the component through the layer-by-layer automated deposition of inks onto a substrate [1,3].Various biomaterial inks based on polymers, ceramics and their composites have been developed for 3D printing of bone tissue scaffolds [4][5][6].However, low shape-fidelity of natural-based inks and compromised biological properties of synthetic-based inks remarkably limit the number of potential inks for 3D printing of bone tissue scaffolds [1,7,8].Therefore, the development of suitable 3D printing inks, which fulfils both biological requirements and print fidelity satisfying the biofabrication window, is still very challenging [9].
Hydroxyapatite (HA, Ca 10 (PO 4 ) 6 (OH) 2 ) has the closest composition to natural bone of mammalians among the various calcium phosphate ceramics (CPCs).HA is thermodynamically the most stable and least soluble CPC in physiological conditions after implantation, making it beneficial when long term healing is expected [3,10,11].In addition, HA can provide nucleating sites on its surface to precipitate apatite crystals [12].HA has been shown to promote adhesion and proliferation of osteoblasts, as well as inducing osteogenesis differentiation of MSCs when studied in vitro and in vivo [13,14].However, like other CPCs, HA is brittle and has low toughness, which restricts its use mainly to filling teeth and bone, or coating orthopaedic and dental implants [15].Therefore, the incorporation of HA in a polymeric system is of high interest to produce composite materials with good bioactivity and compressive strength resulting from HA, and good toughness and biodegradability granted by the polymer matrix [3,16].Polyethylene glycol (PEG), a water soluble and biocompatible polymer, is one of the most used biomaterials in various biomedical applications [17].Polyethylene glycol diacrylate (PEGDA) is a nonionic hydrophilic derivative of PEG with two acrylate groups showing low toxicity, non-immunogenicity, and antifouling properties [18].The acrylates allow free radical photopolymerisation of PEGDA in the presence of a photoinitiator [19,20], converting from a low-viscosity monomer (liquid) to a stable polymer (solid), thereby forming high-precision and complex geometries.These advantages make PEGDA a promising candidate in 3D printing for biomedical applications.
Although a variety of nanocomposites have been fabricated using HA as the filler and PEGDA as the polymer matrix, most were prepared by conventional fabrication methods, exhibiting disadvantages such as poor control over geometrical design and long processing time.A few studies have focused on using 3D printing techniques to fabricate hydroxyapatite-PEGDA (HA-PEGDA) nanocomposites for bone regeneration.For instance, Zhou et al [21] incorporated HA into PEGDA ink for fabricating nanocomposite bone scaffolds using a stereolithography (SLA)-based 3D bioprinter.They demonstrated that HA containing 3D printed scaffolds under low intensity pulsed ultrasound (LIPUS) treatment promoted mesenchymal stem cells (MSCs) proliferation, alkaline phosphatase (ALP) activity, and calcium deposition.Mondal et al [22] used PEGDA to reduce the overall ink viscosity of acrylated epoxidized soybean oil (AESO)/HA ink required for masked stereolithography-based 3D printing.At 10 vol% HA, they reported improved tensile strength, apatite formation on the nanocomposite surfaces within 7 d of incubation in simulated body fluid (SBF), as well as good viability and proliferation of differentiated mouse pre-osteoblastic cells (MC3T3-E1) on the nanocomposites.Deng et al [23] applied continuous liquid interface production (CLIP) for 3D printing of HA/PEGDA nanocomposites.The incorporation of HA resulted in significant improvement of both compression strength and biocompatibility of the final 3D prints.However, in these studies often PEGDA is mixed with HA powder to form a slurry, and there is a risk of the sedimentation of ceramic particles, or there is a need for additional support during printing.
Herein, we explored a new processing method, i.e., extrusion-based 3D printing, to prepare HA-PEGDA scaffolds with high resolution.To the best of our knowledge, there is no study in the literature that utilises extrusion-based 3D printing for fabricating bone tissue scaffolds based on HA-PEGDA inks.In order to provide adequate rheological behaviour and gelation mechanism for HA-PEGDA inks required for extrusion-based 3D printing, pluronic F127 (PF127) was added to the inks as a sacrificial carrier.PF127 is an amphiphilic and water-soluble triblock copolymer with hydrophilic block polyethylene oxide (PEO) and hydrophobic block polypropylene oxide (PPO) [24].Above the critical micelle concentration, PF127 can go through thermo-reversible gelation when the temperature increases.We hypothesised that PF127 can be a suitable carrier for HA-PEGDA inks providing proper distribution of HA within the PEGDA network before photopolymerisation, and sufficient shear thinning properties during processing to 3D print high resolution complex scaffolds.For this purpose, a comprehensive investigation was carried out on the effect of HA content on rheological properties and printability of HA-PEGDA hydrogel inks, and the optimal printing parameters were evaluated.The swelling behaviour, compressive strength, and in vitro degradation of these 3D printed scaffolds were investigated.Furthermore, the cytocompatibility, cell attachment, osteogenic differentiation of human bone mesenchymal stem cells (hBMSCs), and bone-bioactivity in the presence of 3D printed HA-PEGDA scaffolds were evaluated.

Preparation of hydrogel inks
After some preliminary studies (Supplementary data), the hydrogel inks were prepared by dispersing three different concentrations of HA (1, 2 and 5 wt%) and 20 w% PEGDA (700 kDa) in distilled water.The maximum concentration of HA was set at 5 wt% because it did not block the dispensing nozzle.Then, 0.1 wt% Irgacure 2959 was added, and stirred to make homogenous solutions.Afterwards, 25 wt% PF127 was added, and the hydrogel inks were stored in the fridge for two days for complete dissolution of PF127.Finally, all hydrogel inks were loaded into 10 ml cartridges (Polypropylene, UV/light block amber barrels, Nordson EFD, USA), and allowed to equilibrate to room temperature for one hour before printing to initiate the physical gelation of PF127.

Rheological characterisation of the hydrogel inks
The effect of adding HA on the rheological properties of hydrogel inks was investigated using a HR-3 rheometer (Discovery Hybrid rheometer, TA Instrument) with a 40 mm diameter parallel plate geometry, a plate-to-plate gap of 150 µm, and a solvent trap to prevent hydrogel drying.To characterise their shearthinning properties, G ′ and G ′′ were recorded as a function of oscillatory strain sweeps (0.01%-1000%) at a constant frequency of 10 rad s −1 .An oscillatory temperature sweep was used to study the gelling behaviour of the inks from −5 • C to 10 • C at a constant temperature step of 1 • C. Samples were equilibrated for 5 min before testing and for 1 min at each subsequent temperature to minimise thermal gradients within the sample.The temperature at which the elastic modulus made a drastic jump towards a higher value was recorded as the gel point of the hydrogel.All measurements were performed within the linear viscoelastic region at 25 • C and at least by duplicate.

Optimising and evaluating the shape fidelity of 3D printed scaffolds
In order to optimise the printing conditions, printability of each hydrogel ink was tested by extruding the ink using a pneumatic bioplotter (BioScaffolder 3.1, GeSiM, Germany) in a 2-layered construct with grid-like patterns (0/90 • ) with adjacent filaments of 210 µm in diameter at inter-spacing of 300 µm using a 27 gauge dispensing nozzle (Polyethylene Smoothflow tapered tips, Nordson EFD, USA).The strand diameter was measured at the second layer.The z value of the printing point was specified to be the printer's substrate with z value of 0, and nozzle offset value of 210 µm (strand diameter) + 160 µm (height of the petri dish) + 30 µm (error) = 0.4 mm was specified in the 3D printing process.Printing was performed at a room temperature of 25 • C, pressure 60-100 kPa, printing speed 1-15 mm s −1 , and UV curing for 8 s after printing each layer (OmniCure Series1500, 365 nm, 25 mW cm −2 ).The Pr values of hydrogel inks under different conditions were determined based on equation (1), which compares printed area versus those in digital design [25] Pr = A e A t (1) where A e represents the area of printed mesh in experiments determined by images, while A t stands for the area of mesh according to the design, which is the area of a square hole of 300 mm on each side.An ideal pore exhibited a square (or rectangular) shape resulting in Pr = 1, while Pr < 1 and Pr > 1 correspond to a more round or irregular shaped, respectively.The Pr values in the range of 0.9-1.1 were considered acceptable for the 3D printed hydrogel construct based on the literature [25].Immediately after fabrication, 3D printed structures were imaged using an inverted optical microscope (Olympus IX71 inverted microscope, Japan) using a 4× objective.To measure the Pr value of each set of printing parameters, optical images of the scaffolds were analysed using ImageJ software (ImageJ v.1.51r,National Institutes of Health, USA).At least 20 squares were measured to determine A mean values.After optimising the conditions, a cylindrical geometry (8 × 1 mm 2 ) embedded with strands in a 0/90 • pattern in five layers was designed using BioCAD software and fabricated using a 27 gauge nozzle.

Post-printing rinsing process
The post-printing rinsing process was optimised and performed to remove PF127 from the 3D printed scaffolds, thereby improving the cell viability on scaffolds (supplementary data).Briefly, the scaffolds were washed twice with cold phosphate buffered saline (PBS; pH = 7.4), and left in PBS for 48 h at 4 • C with daily changes of PBS to equilibrate.In addition, prior to cell studies, scaffolds were immersed in cell culture media overnight to exchange the PBS with media.

Thermogravimetric analysis (TGA)
The effect of the post-printing rinsing process on the HA content of the scaffolds was investigated using a Q50 TGA analyser (TA Instruments, New Castle, DE, USA).Approximately 5 mg of air-dried scaffolds (asprepared or washed) were loaded in a platinum pan and heated from 25 • C to 900 • C at a constant heating rate of 10 • C min −1 under N 2 supply with a flow rate of 10 ml min −1 .

Fourier-transform infrared spectroscopy (FTIR)
Fourier transform infrared spectroscopy (Varian 610-IR FTIR, Bruker, USA) was used for the identification of the chemical structure of 3D printed scaffolds, HA, and PEGDA monomer.Scans (n = 24) for each FTIR spectrum were obtained in ATR mode over the region 400-4000 cm −1 with a 4 cm −1 spectral resolution.

Swelling behaviour
After post-printing rinsing and drying the scaffolds in the oven at 37 • C for 48 h, the initial dry mass, m dry , was obtained.Then, the scaffolds (N = 5) were incubated in PBS (pH = 7.4) for 48 h at 37 (2)

Dynamic mechanical properties
Dynamic mechanical analysis (DMA) was performed using a DMA 242 E Artemis (Netzsch, Germany) in compressive mode for the 3D printed scaffolds with the aim of evaluating their mechanical properties.After post-printing rinsing, the scaffolds (N = 3) remained submerged in PBS to stay hydrated until the time of testing, which occurred within 1-2 d.The scaffolds (8 × 5 mm 2 in 25-layers) were placed on a compression flat surface fixture, and a force with a frequency of 1 Hz and a displacement of 10 mm was applied by an oscillating plate at the isothermal temperature of 37 • C and humidity of 80%.

Biocompatibility assessment
The scaffolds were placed into 96-well plates and 20 µl of human immortalised bone-marrow derived mesenchymal stem cells (T0523-hTERT, resolving IMAGES, Australia) were seeded on the scaffolds at a density of 1.5 × 10 4 cells/well.Same density of cells was directly seeded onto the wells for both the positive control group, where cells were cultured only with media, and the negative control group, where cells were treated with 10% SDS.Media with no cells served as blank.The cells were incubated for 2 h at 37 • C to allow the cells to adhere to the scaffold before the addition of 80 µl of culture media (DMEM supplemented with 10% FCS, 1% penicillin/streptomycin).Scaffolds were not washed after 2 h of cell adhesion to maintain the absolute number of cells across all the samples so direct comparisons with controls could be made.After 48 h, the media was discarded and subsequently, 100 µl of MTT (0.5 mg ml −1 ) was added to each well.After 4 h of incubation, the supernatant of the culture was gently discarded, and 100 µl of DMSO was added to end the reaction by dissolving the formazan crystals formed in living cells.Three replicates were used in three independent experiments, and absorption was measured at 570 nm using a plate reader (Varioskan Lux, Thermo Fisher Scientific).The relative cell viability (%) compared to positive control was calculated according to equation (4):

Live/dead assay
The scaffolds were placed into 96-well plates and hBMSCs were seeded on the scaffolds at a density of 1.5 × 10 4 cells/well.After 48 h of cell seeding, the live/dead reagent containing calcein AM (1.5 µM) and PI (3 µM) was added to the scaffolds.Data was then acquired using a fluorescence microscopy (Olympus IX71 inverted microscope, Olympus Corporation, Japan).

Osteogenic differentiation of hBMSCs
The scaffolds were placed in 48-well plates and 50 µl of the hBMSCs were seeded on the scaffolds at a density of 4.5 × 10 4 cells/well.The cells were incubated at 37 • C for 2 h to allow the cells to adhere to the scaffold before the addition of 250 µl of culture medium.The normal culture media was changed to osteogenic media after cells reached 90% confluence (after 2 d).The osteogenic medium was prepared by adding 100 nM dexamethasone, 50 µg ml −1 ascorbic acid, and 10 mM glycerophosphate disodium salt into the normal culture medium, and it was changed every two days.After 7, 14, and 21 d, the osteogenic differentiation of the cells was analysed by ALP activity and calcium deposition.

ALP activity
ALP activity of the cells was evaluated as an early marker of osteoblast differentiation.An ALP substrate kit (SensoLyte pNPP ALP assay kit) was used according to the manufacturer's protocol.Firstly, cells were gently washed twice with 1× assay buffer, and incubated in lysis buffer (0.2% Triton X-100 in 1× assay buffer) for 10 min.Then, the cells were sonicated on iced water for 1 min in cycles of 1.5 s (on) followed by 1 s (off) at 50% amplitude (SFX150, Branson Ultrasonic Corporation, USA).Afterward, the cell lysates were transferred into a 96-well plate (50 µl/well).Subsequently, 50 µl of pNPP substrate solution was added, gently mixed for 30 s on a shaker, and incubated in the dark at room temperature for 60 min.The reaction was then terminated by adding 50 µl of the stop solution.The production of P-nitrophenol was determined by measuring the absorbance at 405 nm using a plate reader.ALP activity was expressed as the amount of p-nitrophenol released, and calculated using equation ( 5): where, B is the amount of pNP in the sample well calculated from the standard curve (µmol), D is the sample dilution factor, ∆T is the reaction time (min), and V is the original sample volume added into the reaction well (ml).Finally, all values were normalised to the corresponding total protein concentration measured by using a bicinchoninic acid (BCA) assay kit (Thermo fisher).

Alizarin red S (ARS) staining
ARS staining was used to detect calcium deposition on the scaffolds after 7, 14 and 21 d of cell culture.Briefly, the scaffolds were gently washed three times with PBS, and then the attached cells were fixed with 4% paraformaldehyde and incubated for 20 min at room temperature.Afterwards, cells were washed twice with deionised water, and stained with 40 mM Alizarin Red S solution (pH 4.1-4.3)for 30 min at room temperature with gentle shaking in the dark.Finally, the scaffolds were washed several times with distilled water, air-dried and observed under a light microscope (Olympus CKX41 inverted microscope, Olympus, Japan).

Evaluation of the mineralisation capacity of the 3D printed scaffolds using SBF
The in vitro mineralisation capacity of the 3D printed scaffolds was investigated after immersion in SBF at 37 • C. The SBF was prepared according to Kokubo's protocol [26].Briefly, the scaffolds were immersed in plastic vials containing SBF for 21 and 35 d.Afterward, the 3D printed scaffolds were taken out from SBF and carefully rinsed with deionised water to remove the soluble inorganic ions from the SBF.The results of the mineralisation test were assessed by scanning electron microscopy (SEM) and x-ray diffraction (XRD) analysis.

Evaluation of the mineralisation capacity of SBF conditioned scaffolds using SEM
Morphological characterisation of the surface was carried out using SEM (JEOL 6700F FE-SEM, JEOL Ltd, Japan).All the samples were coated with palladium/platinum using an Emitech K575X Peltier-cooled high-resolution sputter coater (EM Technologies Ltd, Kent, England) prior to scanning.The elemental analysis of the mineral deposits was evaluated by using energy dispersive x-ray spectroscopy (EDS) in conjugation with SEM.Four spots on each sample were used for the EDS analysis.Ca/P ratio of each sample was calculated as an average of these four spots.

Evaluation of the mineralisation capacity of SBF conditioned scaffolds using XRD
To investigate the components of the apatite layer, the SBF conditioned 3D printed scaffolds were analysed using an XRD (Agilent Technologies Supernova system) with monochromatic Cu Kα radiation.
The spectra for all the scaffolds were recorded from 10 • to 80 • 2θ and treated using CrysAlisPro software.

Effect of HA on shear thinning behaviour of the hydrogel ink
A suitable ink for extrusion-based 3D printing must have high extrudability and good shape-fidelity, which in turn defines its printing accuracy.Both characteristics are directly related to the rheological properties of the ink.It is known that shear thinning behaviour enables decreasing proportional stress alongside increasing flow, facilitating uniform extrusion of the ink from the nozzle [27].The effect of HA content on the viscoelastic behaviour of the hydrogel inks is shown in figure 1.At the beginning, all the hydrogel inks formed a consistent 3D network and were in gel-state (G ′ > G ′′ ), critical for extrusion-based 3D printing.Afterwards, the breakdown of the structure started with some micro cracks, although the elastic portion of the viscoelastic behaviour still prevailed.
As the shear strain increased above the critical yield strain of ∼1%, all the hydrogel inks underwent structural breakdown.The crossover points of the G ′ and G ′′ curves in all four hydrogels were found within a comparable shear strain range of 4%-6%.By additional increase of the strain and passing the crossover point (G ′ = G ′′ ), the individual micro cracks grew further and formed a macro crack that eventually ruptured the entire sample, resulting in the flow of the ink (G ′′ > G ′ ).It is worth mentioning that the viscoelastic plateau did not change remarkably with increasing HA content.This means that the addition of HA, within the applied range, increased the stiffness, while not significantly affecting the physical stability and viscoelastic behaviour of the inks.As expected, by increasing the HA concentration from 0 to 1, 2 and 5 wt%, the G ′ value of the inks increased from 15.1 kPa to 17.3 kPa, 19.3 kPa and 27.5 kPa, respectively.The HA nanoparticles acted as physical crosslinks or reinforcements of the amorphous phase, thereby the free movement of polymer chains was restricted, and the relaxation of the polymer chains became difficult, increasing the moduli [28].

Effect of HA on gelation temperature of the HA-based inks
To investigate if the physical gelation of PF127 was hampered by adding HA, the temperature sweeps were conducted, and the results are shown in figure 2. At temperatures lower than the gelation temperature, all hydrogel inks displayed a viscoelastic behaviour (G ′ > G ′′ ) because the interaction of water with HA was weak, and the formation of more hydrogen bonds with the relatively hydrophilic PEG in PEGDA and PEO block in PF127 was more favourable [29].However, above the gelation temperature, the interaction of PPO block in PF127 and HA increased because the overall hydrophobicity of PPO block in PF127 increased, and the adsorption of water on these segments became energetically unfavourable.This phenomenon resulted in more PPO molecules accumulating on HA through hydrogen bonding, exerting the bridging effect of HA and increasing the modulus significantly.Therefore, the elastic response of the hydrogel inks dominated the viscous response (G ′ > G ′′ ).By increasing the HA content, the gelation temperature decreased from 8 • C for the ink without HA to 7 • C, 5 • C, and 1 • C for the inks with 1 wt%, 2 wt%, and 5 wt% HA, respectively.This result might be attributed to the structuremaking properties of the ions at the surface of HA nanoparticles.Structure-making effect is referred to the orientation of water molecules around charged ions due to the electrostatic interaction [30].Based on Jones-Dole viscosity B-coefficients, the degree of water structuring depends on the ion-solvent interactions.Generally, positive values of the B-coefficient indicate kosmotropes (small ions of high charge density, which are strongly hydrated), while negative values represent chaotropes (large ions of low charge density, which are weakly hydrated) [31].The ions at the surface of HA nanoparticles (Ca 2+ , PO 4 3− and OH − ) possess B-coefficients of 0.284, 0.590, and 0.122, respectively [31].Therefore, HA preferentially bound the water molecules, taking up water before it is energetically favourable to be released.As a result, the degree of hydrogen bonding between water and the OH groups of the PPO units decreased and the hydrophobic interactions between the PPO residues increased, which consequently decreased the gelation temperature [32].As the HA concentration increased, the gelation process was accelerated due to the higher ion content in the ink that binds adjacent water molecules tightly, thus immobilising them, and allowing the PPO groups to hydrophobically associate at lower temperatures [29].It is worth mentioning that although the thermos-sensitivity was not directly exploited in the ambient printing process, it facilitated HA dispersion and homogenisation with PF127, as well as loading cartridges with PF127 solutions at low temperature (0 • C).

Evaluation of printability and shape fidelity of HA-based hydrogel inks
The printability of HA-based hydrogel inks was evaluated using a combination of different printing pressures (60-100 kPa), and printing rates (1-15 mm s −1 ).Generally, a lower printing pressure and higher printing rate resulted in a thinner strand diameter.The extent of decrease in the printed strand diameter was more significant on the less viscous hydrogel (0 wt% HA) compared to the most viscous hydrogel (5 wt% HA).By increasing the printing pressure, higher printing rates were required to fabricate parallel strands using the inks with 0 wt% and 1 wt% HA, whereas lower printing rates were needed to extrude uniform strands using the inks with 2 wt%  and 5 wt% HA.For instance, at a constant pressure of 90 kPa, the minimum printing rate required to fabricate consistent parallel strands was 10, 8, 3, and 2 mm s −1 for 0, 1, 2, and 5 wt%, respectively.It is likely attributed to the same fact that the higher the printing rate the lower the ink viscosity during extrusion.Additionally, higher pressures were required to extrude more viscous hydrogels.It was observed that the strand diameter of ink with 0 wt% HA increased exponentially with increasing printing pressures.For example, at a constant printing ratio of 10 mm s −1 , by increasing the printing pressure from 70 to 80 and 90 kPa, the strand diameter increased from 176 to 266 and 531 µm, respectively (figure 3(a)).This is probably due to the intrinsic lower viscosity of 0 wt% HA ink, which caused a higher extent of strand spreading when a larger printing pressure was used.By increasing the viscosity, a linear relationship between printing pressures and strand diameter was observed in 1 wt% HA ink; where at a constant printing ratio of 9 mm s −1 , by increasing the printing pressure from 70 to 80 and 90 kPa, the strand diameter increased from 188 to 336 and 442 µm, respectively (figure 3(b)).For 5 wt% ink, at a constant printing ratio of 2 mm s −1 , the strand diameter was 128, 143, 305, 412, and 498 µm, for the printing pressure of 60, 70, 80, 90, and 100 kPa, respectively (figure 3(d)).The high viscosity of these hydrogels likely reduced the extent of filament spreading at higher printing pressures.It was also observed that the SD of printed strand diameter decreased with hydrogel inks of higher viscosity.Hence, a more viscous hydrogel offered higher printing consistency and better control over the printed strand diameter at increasing printing pressure.The shaded orange areas in figure 3 show the optimum combination of printing pressure and printing rate to obtain a strand diameter of 210 ± 10.5 µm for different hydrogel inks.A five percent error was selected as the tolerance threshold for the strand diameter based on the literature [33].As it can be seen, lower printing pressures of 60 and 70 kPa were suitable for inks with 0 wt% and 1 wt% HA, whereas a printing pressure of >80 kPa was required for inks with 2 wt% and 5 wt% HA.
Figure 4 shows the Pr value of HA-based inks with different concentrations of HA under different printing pressures and printing rates.A very low printing rate resulted in excessive ink deposition and fusion of the printed strands on the cross site, where mesh areas were nearly zero, making it unfeasible to manufacture a 3D scaffold (Pr < 1).On the other hand, a very high printing rate showed irregular or fractured morphology, resulting in incomplete patterning (Pr > 1).Under optimum conditions, smooth and uniform strands were extruded continuously, resulting in a grid-like pattern close to a square with regular edges.Therefore, the optimal combination of both printing pressure and printing rate that enable the fabrication of complete grid-like patterns at the highest printing resolution were selected for further experiments.[23,[34][35][36].The mass loss of 2.5%-3.5% in the third region corresponded to the dehydration of HA.It is known that HA is thermally stable up to 700 • C, and then by increasing the temperature between 850 • C and 1100 • C, gradual weight loss occurred due to the partial desorption of water [37,38].The TGA curves of washed scaffolds were also divided into three regions.In the first region, the initial weight loss of less than 2% occurred below 200 • C. In the second region, between 250 • C and 450 • C a loss of 88%-97% of the initial weight was  By increasing the HA content, in both as-prepared and washed scaffolds, the weight loss decreased, and the thermal stability of the scaffolds increased.This thermal stability enhancement can be ascribed to (I) the larger amount of physical crosslinking that HA made within the network through the formation  of strong hydrogen interactions and van der Waals forces, which restricted the segmental mobility of the macromolecular chains, thereby decreasing the weight loss [39,40]; and (II) the barrier effect of the HA nanoparticles that effectively obstructed the diffusion of volatile products from the bulk of the polymer to the gas phase, thereby slowing down the decomposition process [41].
The onset temperature corresponding to 5% weight loss (T 5 ), the temperature corresponding to the 50% weight loss (T 50 ), as well as the total weight loss measured at 890 • C, for all the scaffolds are shown in table 1 The remarkable difference between the as-prepared and washed scaffolds was that washed scaffolds had lower rapid thermal degradation onset temperatures compared with as-prepared scaffolds, which can be attributed to the less compact network and increased porosity after removal of PF127.It can be observed that the residual mass at 890 • C increased gradually with increasing the HA content due to the high thermal stability of the mineral nanoparticles [40].The residual mineral masses for the washed scaffolds were measured to be 9.09%, 4.91%, and 2.35% for scaffolds with 5 wt% HA, 2 wt% HA, and 1 wt% HA, respectively.These values were close to the residual mineral masses for the as-prepared scaffolds, which are 9.17%, 5.23%, and 2.82% for scaffolds with 5 wt% HA, 2 wt% HA, and 1 wt% HA, respectively.Therefore, HA was incorporated efficiently throughout the scaffolds without significant loss of HA during the washing process.

Analysis of chemical structure by using FTIR
The chemical structures of the 3D printed scaffolds with varying HA concentrations as well as the PEGDA monomer and HA were identified using FTIR analysis.As shown in figure 6, for PEGDA monomer, the peak at 2867 cm −1 and 1721 cm −1 are assigned to the stretching vibrations of C-H in the alkyl and stretching vibrations of CO in the carbonyl group, respectively.Also, the peaks at 1452 cm −1 , 1349 cm −1 , and 1093 cm −1 correspond with the bending of C-H and the stretching vibrations of C-O and C-O-C, respectively.As for pure HA, the peak at 882 cm −1 is for HPO 4 2− , and the others at 1000-1100 cm −1 (1087 cm −1 , 1033 cm −1 ), 600 cm −1 , and 561 cm −1 are for PO 4 3− [42][43][44].Also, peaks between 1420 and 1455 cm −1 correspond to Figure 7. Equilibrium degree of swelling of the 3D printed HA-PEGDA scaffolds; bar = mean ± SD.Statistical analysis was carried out using a one-way ANOVA followed by Tukey's multiple comparisons test, where p < 0.002 ( * * ).
carbonates (CO 3 2+ ) [45].For the 3D printed scaffolds, FTIR spectra demonstrated successful curing as there were no peaks associated with unreacted acrylate double bond (810 cm −1 , 1191 cm −1 and 1410 cm −1 ) remaining after curing.In addition, the carbonyl (C=O) peak of acrylate group slightly shifted towards a higher wavenumber, from 1721 cm −1 to 1730 cm −1 , implying an increased carbonyl bond strength due to the loss of conjugation between the C=C and carbonyl groups during the photo polymerisation reaction [46].With increasing the HA content in the scaffolds, the characteristic absorption peak at 600 cm −1 and 561 cm −1 gradually increased.The peaks at approx.1455 cm −1 , 1093 cm −1 , and 1033 cm −1 correspond and overlap with the bands of carbonate and phosphate groups of HA and characteristic bands of PEGDA [47].In addition, the broad band at 3500 cm −1 is attributed to the O-H stretching vibration due to the absorption of water molecules on PEGDA and HA.Nevertheless, there was no remarkable difference between the spectra of pure PEGDA and HA-PEGDA scaffolds likely because the peaks for the functional groups of HA overlapped by the peaks of PEGDA due to its relatively low concentration of HA [23].FTIR analysis confirmed the presence of both the polymer phase and the mineral phase in the fabricated scaffolds.

Analysis of swelling behaviour influenced by HA concentration
Swelling is based on occupying the free volume within a sample until it reaches a state of equilibrium and is important in the fabrication of bone scaffolds because a decrease in swelling may increase the mechanical properties [48].The effect of HA addition on the swelling of the 3D printed HA-PEGDA scaffolds is shown in figure 7.For scaffolds without HA (0 wt%), the equilibrium degree of swelling was found to be 244.8± 12, while this value decreased to 237.0 ± 10 and 226.3 ± 7.0 for scaffolds with 1 wt% and 2 wt%, respectively.A significant decrease in swelling to 215.4 ± 6 was also observed with the incorporation of 5 wt% HA compared to 0 wt% (p = 0.0095).This reduction may be attributed to the physical interaction of HA nanoparticles with PEGDA, which acted as a filler that occupy free space within the polymer network, thereby impeding the penetration of liquid into the interior network of the hydrogel.Furthermore, interaction with HA restricted the PEGDA polymeric chains motion, which promoted the densification of the hydrogel [49].Similar studies showed that increasing the amount of HA in PEGDA composites decreased the swelling ability of the tested materials, confirming the contribution of HA in the formation of a more crosslinked network [50,51].

Analysis of compressive modulus influenced by HA concentration
The addition of HA nanoparticles to the PEGDA matrix slightly increased the compressive modulus of the 3D printed scaffolds as shown in figure 8.The scaffolds without HA (0 wt% HA) had a compressive modulus of 176.77 ± 22 kPa.Upon the addition of 1 wt% HA, the compressive modulus slightly increased to 179 ± 35 kPa.As the HA concentration further increased to 2 wt% and 5 wt%, the compressive modulus reached 189 ± 38 kPa and 219 ± 17 kPa, respectively, with no significant difference between them.The increase in the compressive modulus of 3D printed scaffolds can be attributed to three factors: (I) the effective dispersion of HA nanoparticles and the physical crosslinking between HA and the PEGDA matrix, (II) a reduction in PEGDA mobility near the HA nanoparticles and a decrease in the swelling percentage, and (III) the ability of HA nanoparticles to absorb and distribute compressive loads [48,52].Consequently, the incorporation of HA nanoparticles within the polymer matrix, in an appropriate loading range, enhances the compressive modulus of the composite and facilitates load transfer between the components.Our results fall within the range reported in the literature for HA-PEGDA composites [53,54].However, it should be noted that studies utilizing continuous 3D printing techniques such as SLA and CLIP have reported higher compressive moduli for printed scaffolds.For example, Huang et al [55] reported a compressive modulus of 459.1 kPa for SLA-printed PEGDA scaffolds.Deng et al [23] fabricated 3D printed PEGDA-HA scaffolds with 1 wt% HA using CLIP technology resulting in a compression modulus of approximately 40 MPa, while pure PEGDA scaffolds exhibited a value of around 20 MPa.However, an increase in HA loading to 2 wt% decreased the compressive modulus due to the agglomeration of HA nanoparticles.In another study, Kumar et al [56] demonstrated that SLA-printed PEGDA-HAP scaffolds with 1 wt% HA reached a compressive modulus of approximately 50 MPa, which was not significantly different from unfilled PEGDA.
Therefore, it can be inferred that not only the molecular weights and concentrations of the polymer and filler, as well as the crosslinking method, affect the mechanical properties of the scaffolds, but also the specific 3D printing technique employed significantly impacts the final mechanical characteristics of the printed scaffolds.Continuous printing processes like SLA and CLIP ensure solidification of the innermost sections of the print, contributing to the overall robustness of the structure.Furthermore, the conditions under which mechanical testing is conducted also play a role.In our study, mechanical tests were performed in a hydrated state at 37 • C, which is closer to the physiological environment, whereas other studies conducted testing in a dry state or did not provide information on the testing conditions.It is well known that hydration influences the mechanical properties of scaffolds.For instance, a study by Suchý et al [57] fabricated scaffolds with various compositions and demonstrated that the elastic modulus and compressive strength decreased by approximately 95% in the hydrated state compared to the dry state.This highlights the importance of analysing scaffolds in the hydrated state, as it more accurately simulates the real in vivo environment for which these scaffolds are designed.Overall, although the compressive modulus of the 3D printed HA-PEGDA scaffolds in our study remains lower than that of native human bone tissue, it is postulated that in a biologically functional implant, the surrounding native bone would provide the initial mechanical support while scaffold mineralisation and the healing process occur [58].

Analysis of in vitro degradation influenced by HA concentration
The in vitro degradation behaviour of 3D printed HA-PEGDA scaffolds in PBS at 37 • C was evaluated, as shown in figure 9.During the experimental period, all scaffolds maintained their structural integrity.However, an increase in the HA content of the scaffolds led to a decrease in mass loss, indicating a slower degradation rate.The mass loss percentages for scaffolds with 0 wt%, 1 wt%, 2 wt%, and 5 wt% HA on day 21 were 18.74% ± 1.16, 16.09% ± 0.86, 14.00% ± 0.80, and 11.18% ± 2.19, respectively.Significantly lower mass loss was observed in scaffolds with 5 wt% HA compared to those with 0 wt% HA (p = 0.0002).
Several factors may contribute to the observed differences in degradation behaviour.HA possesses higher stability and resistance to degradation compared to PEGDA.Therefore, the incorporation of HA within the scaffolds enhanced their stability and hindered their degradation to some extent.Moreover, this reduction in degradation rate can be attributed to the fact that HA nanoparticles acted as physical crosslinking centres within the polymer network, resulting in decreased hydrophilicity.Consequently, the water penetration became more difficult within the more organised polymer chain network, leading to slower hydrolytic cleavage of the ester bonds in PEGDA and subsequently reducing the degradation rate [59].Cell viability results of hBMSCs on 3D printed HA-PEGDA scaffolds for 48 h measured by MTT assay.Results are represented as the mean ± SD of three independent experiments with three technical replicates.One-way ANOVA followed by Tukey's comparisons test was used for statistical analysis, p < 0.05 ( * ).
Therefore, the ratio of PEGDA and HA in the scaffolds can be tailored to control the degradation rate to match the rate of bone healing, which varies depending on bone type, size, location, as well as individual factors such as age, comorbidities, lifestyle.
It is important to note that all scaffolds exhibited an increase in mass loss over the 21 d period.For scaffolds with 2 wt% HA, the degradation rate appeared to reach a plateau from day 7, while scaffolds with 5 wt% HA showed a slowing down or reaching a steady state of degradation by day 14 after an initial period of mass loss.These time points correspond to the equilibrium point of degradation rate, where minimal or no changes in mass loss occur until the end of the assay [60].This behaviour may be attributed to the release of HA into the media, where the higher HA content and more compact structure of scaffolds with 5 wt% HA prolonged the time needed to reach the equilibrium point compared to scaffolds with 2 wt% HA.Nevertheless, further investigation and characterization are necessary to gain a comprehensive understanding of the underlying mechanisms responsible for these observations.

Biocompatibility assessment of 3D printed scaffolds by using MTT assay
To investigate whether cell viability is compromised by the presence of HA in the fabricated 3D printed scaffolds, the viability of hBMSCs after 48 h culture was analysed.Figure 10 demonstrates that the addition of HA (1 and 2 wt%) into the scaffolds did not significantly decrease cell viability (72.5% for 1 wt% and 72.7% for 2 wt%).However, by increasing HA to 5 wt%, a significant reduction in cell viability (67.8%) was observed compared to scaffolds with 0 wt% HA (84.3%) (p = 0.0444).This result could be correlated to the presence of a higher amount of HA on the surface of the scaffold with 5 wt% HA, and hence the release of higher amounts of HA into the static culture medium within the first 48 h of cell culture.There are three different scenarios on the cytotoxicity of nanoparticles [61].(I) Nanoparticles can react with the surrounding fluids, causing the release or depletion of ions and proteins crucial for cell function; (II) nanoparticles can interact with cell membrane receptors, potentially inducing apoptotic signalling cascades; and (III) nanoparticles can be internalised by cells, exerting their effects inside the cell.The third assumption is the most accepted hypothesis, which relies on the degradability of the nanoparticles upon internalisation.Based on this hypothesis, after HA nanoparticles uptake by endocytosis, they degrade under the acidic conditions in the lysosome, yielding an increase of calcium and phosphorus ions.A slow and sustained dissolution of HA in the lysosomes would benefit transfection, while a faster internalisation of nanoparticles would release a high concentration of calcium ions that would lead to cell death.For example, Huang et al [62] reported that HA crystals were endocytosed by A7R5 cells, where a high degree of crystal endocytosis corresponded to a high intracellular calcium concentration, leading to cell membrane rupture.Nonetheless, the size, shape, concentration, and exposure duration of HA nanoparticles as well as the cell type may influence the cellular responses and degree of potential damage.In our study, it was observed that by increasing the culture time and changing the cell culture media, the cell viability increased as evidenced by livedead assay.The cellular response to HA nanoparticles is an active area of research, and further studies are needed to understand the detailed mechanisms and concentration-dependent patterns of cellular damage associated with HA nanoparticles.

Cell attachment and growth on the 3D printed scaffolds
Due to the complete inertness of scaffolds with 0 wt% HA, hBMSCs showed a preference for cell-cell and cell-plate contact rather than cell-scaffold interaction, thereby no cell attachment was observed on the scaffolds with 0 wt% HA.The attachment and viability of hBMSCs at different layers of the 3D printed HA-PEGDA scaffolds after 48 h of culture are shown in figure 11.High cell viability (>95%) of attached hBMSCs was observed on all three scaffolds with no significant difference (figure 11(a)).However, cell attachment on scaffolds with 5 wt% HA was significantly higher compared to 1 wt% (p = 0.0372), which can be attributed to the rougher surface of scaffolds with 5 wt% HA.Moreover, HA has the ability to adsorb protein from the culture media [63], and the absorption of protein likely increased with increasing the HA content, resulting in higher adhesion of hBM-SCs to the scaffolds with 5 wt% HA.It can also be seen the cells on scaffolds with 5 wt% HA already started to spread out along the pores after 2 d.This indicated that the addition of HA nanoparticles into the inert 3D printed PEGDA scaffolds significantly improved the adhesion of hBMSCs.After 14 d, the cells filled up the pores and covered the entire scaffolds, exhibiting a vigorous proliferation of hBMSCs, and the biocompatibility of all scaffolds for further use (figure 11(b)).Comparison between alkaline phosphatase (ALP) activity of hBMSCs in various 3D printed scaffolds cultured in either osteogenic or normal media at day 7, 14, and 21.'OS' and 'N' stand for scaffolds cultured in osteogenic and normal media, respectively.Results are expressed as mean ± SD from three independent experiments with three technical replicates.Statistical analysis was carried out using two-way ANOVA followed by Tukey's multiple comparisons test, p < 0.0001 ( * * * * ) and p < 0.002 ( * * ).
Finally, these images suggested that lower cell viability of scaffolds with 5 wt% HA obtained by MTT assay (figure 10), was a result of a higher rate of HA leach out within the first 48 h, which significantly improved by changing the cell culture media.Therefore, the 3D printed scaffolds with 2 wt% and 5 wt% HA were selected for further experiments as they showed the lowest swelling ratio, highest compression strength, and improved cell attachment.

Measuring osteogenic differentiation of hBMSCs by using ALP activity
The ALP activity of the hBMSCs was measured for the comparison of their osteogenic activities at day 7, 14 and 21, and results are shown in figure 12.All data were normalised with regard to the total protein content to confirm that the higher ALP activity was not due to the increased cell number.In the stimulated group, ALP activity of hBMSCs in all three scaffolds showed a similar pattern with a peak during osteogenic differentiation.In detail, ALP activity significantly increased with increasing culture time from day 7 to day 14 in scaffolds with 0 wt% HA (p < 0.0001), 2 wt% HA (p < 0.0001), and 5 wt% HA (p < 0.0001).This was followed by a significant decrease in ALP activity at day 21 in scaffolds with 0 wt% HA (p < 0.0001), 2 wt% HA (p < 0.0001), and 5 wt% HA (p < 0.0001).The lower ALP activities of cells at day 21 in all groups indicated that the cells on these scaffolds switched to the next differentiation state with the down regulation of the ALP gene [64].Furthermore, 3D printed HA-PEGDA scaffolds showed an increase in ALP activity in a dose-dependent pattern, and scaffolds with 5 wt% HA showed a lower ALP activity compared to 0 wt% HA at all time points.Similarly, Wang, Hu [65] reported that HA affects ALP activity of MC3T3E1 cells in a dose-dependent manner.In their study, the ALP activity at day 7 increased when the concentration of HA was ⩽100 mg ml −1 , while it decreased when the concentration reached 200 mg ml −1 .When comparing the influence of HA in normal media, the ALP expression on all the scaffolds was dependent on the culture time.The ALP activity on scaffolds with 2 wt% HA and 5 wt% HA continued to increase up to day 21, with a significant increase in ALP activity of scaffolds with 2 wt% HA from day 7 to day 14 (p = 0.0194).This indicated that the cells cultured in normal media on scaffolds with 2 wt% HA and 5 wt% HA were still at an earlier differentiation stage.Moreover, all the scaffolds showed significantly higher ALP activity of cells cultured in osteogenic media compared to that of normal media at day 14 (p < 0.0001).Significantly higher ALP activity was also observed at day 21 in cells cultured in osteogenic media compared to that of normal media on scaffolds with 0 wt% HA (p = 0.0100).This indicated the synergic effect of the scaffold composition and the osteogenic media on cell response.Overall, these results demonstrated that hBMSCs undergo similar changes of ALP activity during osteogenic differentiation on 3D printed HA-PEGDA scaffolds cultured in either normal or osteogenic media, and an appropriate HA incorporation can promote hBMSCs differentiation into mature osteoblasts phenotypes in normal media.

Measuring calcium deposition by using ARS staining
Matrix mineralisation is a controlled biological process whereby differentiating osteoblasts accumulate environmental calcium and phosphate to form calcium apatite.In bone tissue engineering, extracellular mineralisation serves as a strong indicator of osteogenic differentiation [66,67].Figure 13 presents optical microscopic images of 3D printed scaffolds stained with ARS after 7, 14, and 21 d of culture in either normal or osteogenic media to qualitatively assess the mineralisation of hBMSCs.Scaffolds without HA (0 wt% HA) exhibited no calcium deposition on days 7 and 14, regardless of the culture medium used.However, on day 21, hBMSCs cultured on scaffolds with 0 wt% HA in osteogenic media displayed orange-red stains indicative of mineralisation, while hBMSCs in normal media did not show mineralisation.On the other hand, mineralisation continuously improved on 3D printed HA-PEGDA scaffolds during osteogenic differentiation.hBMSCs on scaffolds with 2 wt% HA and 5 wt% HA showed greater amounts of mineral deposits and bone nodules compared to cells cultured on the scaffolds with 0 wt% HA at day 7, 14, and 21.
Recent evidence has demonstrated that HA nanoparticles possess remarkable osteoinductive properties on hMSCs by up-regulating osteogenic genes such as ALP, COL1, and RUNX2 when hMSCs are grown on HA-based biomaterials [68].Moreover, multiple studies have suggested that the presence of extracellular HA can influence the commitment and function of osteoblasts by releasing Ca 2+ continuously.
For example, Sattary et al [69] showed that the addition of 15 wt% HA nanoparticles in PCL/Gel scaffolds promoted the rate of mineral deposition by MG-63 cells compared to scaffolds without HA at day 7 and 21.In another study, Gleeson et al [70] demonstrated that the addition of different concentrations of HA (50, 100, and 200 wt%) to a collagen-based scaffold enhanced the mineralisation of MC3T3-E1 preosteoblast cells in a concentration-dependent manner as early as day 7. Similarly, Kim et al [71] showed enhanced mineralisation of BMSCs on poly (lacticco-glycolic acid) scaffold by increasing the concentration of HA from 0% to 10%, 20%, 40% and 60%, in a timely manner from day 1 to day 7, 14, 21 and 28.Therefore, our findings aligned with previous studies that have shown enhanced mineralisation on HAbased scaffolds, with HA acting as a chelating agent for mineral deposition on MSCs as early as day 7 [72].Moreover, it is plausible that the partial dissolution of HA led to elevated levels of calcium and phosphate ions in the immediate vicinity, thereby promoting the activity and mineralisation of osteoblast-like cells [73].
Even in the absence of osteogenic media, the addition of HA can induce osteogenic differentiation.For example, in a study conducted by Jamshidi et al [74], the addition of HA in gellan gum beads induced nodule formation in MC3T3-E1 cells.It also significantly enhanced the osteogenic differentiation of bone marrow stromal cells, even in the absence of osteogenic media, when compared to tissue culture plastic under similar conditions.These effects were observed within a timeframe of 5 d.Similar study by Calabrese et al [75] investigated the effects of collagen-HA scaffolds loaded with human adipose-derived stem cells (hADSCs) in both normal and osteogenic media.The results revealed that in the presence of normal media, calcium deposits began to appear after 2 weeks of culture.However, when the samples were grown in the presence of osteogenic factors, mineralisation of the extracellular matrix initiated as early as the first week, with a statistically significant increase over time.Furthermore, the presence of osteogenic factors resulted in a greater amount of calcium deposition at each assessed time point compared to samples cultured in normal media.These results indicated that the scaffold itself induced osteogenic differentiation of hADSCs in vitro, while the presence of osteogenic factors accelerated this process.Another study examined the osteogenic properties of collagen-HA scaffolds compared to collagen scaffolds.Enhanced ARS staining was observed in MSCs cultured on collagen-HA scaffolds, both in normal and osteogenic media, suggesting that the addition of HA in the scaffold improved its osteoinductive and osteoconductive properties.Furthermore, the presence of HA in the collagen-based scaffold significantly increased calcium deposition, reducing the need for high levels of BMP2 to induce the osteogenic ability of MSCs [76].In addition, the amount of calcium deposition in hAD-MSCs cultured in cell culture media containing HA was observed to increase as early as day 7, compared to hAD-MSCs cultured in normal media alone [77].Similarly, Fu et al [78] reported an increasing number of stained bone nodules recovered from HA scaffolds as the culture time progressed from day 3 to day 12.Therefore, consistent with previous reports, our findings indicated that the presence of HA in the 3D printed HA-PEGDA scaffolds can enhance mineralisation, even in the absence of osteogenic supplements.

Evaluation of bone bioactivity of the 3D printed scaffolds by using SEM and XRD
The morphology and composition of mineral deposits on the 3D printed scaffolds after 21 d and 35 d incubation in SBF are shown in figures 14(a) and (c).There was an increased apatite deposition on the surface of the scaffolds over time.After soaking in SBF for 21 d, a few small granules were visible on the surface of scaffold with 0 wt% HA.After 35 d, the surface of these scaffolds formed more heterogeneous mineral deposits with small aggregations.A higher magnification examination showed that the granules were composed of crystals with the cauliflower morphology of HA, the same morphology observed by Ni et al [79].Despite the absence of HA in the original composition of the scaffold with 0 wt% HA, a few mineral deposits were observed on the scaffolds due to ions sedimentation.Similarly, Zhang and Ma [80] reported the formation of an apatite layer on the polyl-lactic acid (PLLA) surface in SBF, where PLLA was slowly hydrolysed in SBF, leading to the formation of new -OH hydroxyl and -COOH carboxyl groups on its surface.The presence of these groups facilitated the accumulation of Ca 2+ and PO 4 3− through electrostatic forces and hydrogen bonding on the surface.
The surface of scaffolds with 2 wt% HA and 5 wt% HA formed many more apatite clusters and mineral deposits compared to the pristine scaffolds.The surface of scaffolds with 2 wt% HA exhibited rough mineral deposits on day 21.The surface of scaffolds with 5 wt% HA showed a deposited layer with typical globular cauliflower morphology of HA.After 35 d, the surface of scaffold with 2 wt% HA formed uniform spherical petal-like particles (∼5 µm), the same morphology seen by Gao, Wei [81].The surface of scaffolds with 5 wt% HA exhibited larger mineral deposits after 35 d.Previous studies have shown that mineralisation was higher on scaffolds containing HA than on pristine scaffolds [72].The major reason for the enhancement of apatite formation on the 3D printed HA-PEGDA scaffolds might be that the HA particles acted as nucleation initiation sites [82].Nucleation is affected by the surface charge of the HA scaffolds that absorb Ca 2+ and PO 4 3− from the metastable SBF solution and form an apatite layer [83].The more HA content in the composite scaffolds, the more nucleation initiation sites existed, as a result the apatite formation was faster.Once the apatite nuclei were formed, they grew spontaneously by consuming the Ca 2+ and PO 4 3− present in the surrounding fluid [82].Moreover, the rougher and bumpy surface of scaffolds with 2 wt% HA and 5 wt% HA compared to the pristine scaffolds provided a higher surface area and induced higher deposition of minerals.The main elements on the surface of the scaffolds detected by EDS were calcium, phosphorus, oxygen, and carbon, indicating that calcium (Ca) and phosphate (P) were present in the minerals and was not contaminated by other ions in the SBF.The Ca/P ratio was not significantly different between various scaffolds and remained constant at ∼1.9-2.This is slightly higher than the Ca/P ratio of natural HA (1.67).However, studies have shown that the Ca/P ratio between 1.3 and 2.0 can be classified as HA [84,85], which confirms the formation of HA on the surface of the 3D printed HA-PEGDA scaffolds.

Conclusion
In conclusion, printing pressures and printing rates of a pneumatic extrusion-based 3D bioplotter were optimised to fabricate HA-PEGDA scaffolds with high shape fidelity and resolution (200 µm).HA incorporation had no negative impact on the rheological properties of the developed HA-based hydrogel inks, showing excellent printability for extrusionbased 3D printing.PF127 was an excellent sacrificial carrier, which provided good distribution of HA nanoparticles within the scaffolds.The HA content in the 3D printed scaffolds did not significantly decrease after the post-printing rinsing process was applied to remove PF127.Among all the 3D printed scaffolds, scaffolds with 5 wt% HA showed the least swelling ratio and degradation rate, and the highest compression strength and hBMSCs attachment onto the scaffolds.On the other hand, scaffolds with 2 wt% HA showed slightly higher cell viability and higher ALP activity, while also improving the calcium deposition, compared to scaffolds containing 5 wt% HA.Collectively, these results show that the fabricated 3D printed HA-PEGDA scaffolds have promising applications for bone regeneration by exhibiting excellent shape fidelity and promotion of hBMSCs adhesion and osteogenic differentiation.

Figure 1 .
Figure 1.Strain-sweep graph for HA-based hydrogel inks.The closed and open symbols represent the storage modulus (G ′ ) and loss modulus (G ′′ ), respectively.

Figure 2 .
Figure 2. Oscillatory temperature sweeps of the HA-based hydrogel inks.

4. 4 .
Evaluating the HA content of 3D printed scaffolds by using TGA TGA analysis was performed to determine the amount of HA after the scaffolds were subjected to the washing process, and the mass loss versus temperature curves are shown in figure5.Three stages of mass loss were observed for all the as-prepared scaffolds (before washing), two minor and one major mass loss.The initial stage corresponded to the evaporation of residual water in the samples, at a temperature between 50 • C-200 • C (mass loss 1%-2%).The great and rapid mass loss occurred within the narrow range of temperature (350 • C-450 • C) during the thermal decomposition of PEGDA and PF127.Based on the published literature, PF127 shows a single step degradation between 300 • C-452 • C via a random scission mechanism, and the pyrolysis of PEGDA occurs around 410 • C

Figure 5 .
Figure 5. Mass loss of the 3D printed HA-PEGDA scaffolds.The solid lines represent as-prepared and the dashed lines represent washed scaffolds.

Figure 8 .
Figure 8. Compressive modulus of the 3D printed HA-PEGDA scaffolds in hydrated state at 37 • C.

Figure 9 .
Figure 9. Degradation rate of the 3D printed HA-PEGDA scaffolds in PBS for different times, bar = mean ± SD.

Figure 10 .
Figure10.Cell viability results of hBMSCs on 3D printed HA-PEGDA scaffolds for 48 h measured by MTT assay.Results are represented as the mean ± SD of three independent experiments with three technical replicates.One-way ANOVA followed by Tukey's comparisons test was used for statistical analysis, p < 0.05 ( * ).

Figure 12 .
Figure 12.Comparison between alkaline phosphatase (ALP) activity of hBMSCs in various 3D printed scaffolds cultured in either osteogenic or normal media at day 7, 14, and 21.'OS' and 'N' stand for scaffolds cultured in osteogenic and normal media, respectively.Results are expressed as mean ± SD from three independent experiments with three technical replicates.Statistical analysis was carried out using two-way ANOVA followed by Tukey's multiple comparisons test, p < 0.0001 ( * * * * ) and p < 0.002 ( * * ).

Figure 14 .
Figure 14.SEM images, EDS results, and XRD patterns of mineral deposits on the 3D printed scaffolds after (a), (b) 21 d, and (c), (d) 35 d of incubation in SBF.

Table 1 .
TGA critical points of 3D printed HA-based scaffolds with different mass fraction of HA nanoparticles.T5 and T50 represent the decomposition temperature of the scaffolds when the mass loss was 5% and 50%, respectively.