Digital light processing stereolithography of hydroxyapatite scaffolds with bone- like architecture, permeability, and mechanical properties

This work deals with the additive manufacturing and characterization of hydroxyapatite

tissue engineering approaches as the use of man-made biomaterials allows overcoming the problems of transplant tissue shortage, immune compatibility/rejection, and the need for harvesting extra-surgery in the case of autologous grafts. 2 Synthetic bone grafts are often produced in the form of three-dimensional (3D) porous structures (scaffolds) that are able to form a strong bond to bone tissue. 3 Scaffolds must exhibit high and interconnected porosity, provide structural support to host tissue and withstand the physiological loads while promoting cell adhesion and growth along with vascularization, thus leading to new bone regeneration. 4 Among bioceramics, synthetic hydroxyapatite is highly suitable to fabricate bone scaffolds due to its compositional and crystallographic similarity to bio-apatite, which constitutes more than 60% of the inorganic bone matrix. 5 Its significance in biomedicine and biomaterials science has been widely demonstrated over more than 50 years of clinical use in orthopedics. 6 Hydroxyapatite-based porous scaffolds (both full-ceramic and composite) have been manufactured by several technologies, ranging from the traditional ones, that is, foam replication, 7,8 dry process with porogen agents 9 and isostatic pressing without sintering, 10 to the latest solid free-form (SFF) fabrication strategies. 11 Among traditional methods, foam replication shows several advantages including the high versatility, which allows the production of irregularly-shaped bone like products with various pore features, no need for toxic chemicals and the very low cost of implementation, making it a technology with a high commercialization potential. However, foam replication is highly operator-dependent, poorly reproducible and standardized, thus being suitable in just a minority of clinical applications.
Despite the higher technological costs, additive manufacturing technologies have recently shown great promise to produce bioceramic scaffolds, including porous hydroxyapatite, allowing an accurate control on internal pore and strut features to be achieved. 12 In this regard, lithography-based methods allow obtaining high-quality ceramic products with the best spatial resolution among the currently-available rapid-prototyping techniques (less than 50 μm). The latest evolution of stereolithographic methods is based on digital light processing (DLP), which relies on a dynamic mask to promote the polymerisation of a complete layer of photocurable resin gluing the ceramic particles per irradiation cycle. Compared to other stereolithographic methods, this bottom-up approach actually carries some advantages. First of all, photo-cured layers are no longer immersed in the slurry, and this reduces not only the amount of material required but also the risk of introducing defects during the layer-wise building of the product. 13 Moreover, regardless of the shape, size and complexity of the final products, production times are drastically reduced, thereby carrying obvious advantages from economic and technological viewpoints, which are key when mass production is required and/or just partially stable resins are used that need fast processing times. 14 Another strength of this approach is the great geometrical fidelity that comes with the additive layer-by-layer approach allowing design-oriented production instead of productionoriented design. This advantage is also highlighted by the geometrically extremely-complex test design produced that was used in the course of the present study.
Like most additive manufacturing techniques applied to bone scaffolds, DLP-based stereolithography aims at layerwise reproducing a virtual model which typically derives from (i) computer-built geometries based on repetitive geometrical elements such as unit cells (eg, grid-like structures) that, however, are unable to properly replicate the trabecular architecture of cancellous bone, or (ii) medical reconstructions of patient's bone, which may be more difficult to acquire and/or use as input files (problems of compatibility between medical file type and printing apparatus). Moreover, another issue with the latter approach is the large file size that comes with precise approximations of naturally occurring stochastic cellular/tissue designs.
In order to overcome these limitations, the present study reports, for the first time, the use of a tomographic reconstruction of an open-cell polymeric foam as an input file to the DLP-based stereolithographic system. Polymeric sponges are indeed highly suitable templates for scaffolding since they have been used for many years in the production of bonelike glass and ceramic scaffolds by the conventional sponge replica method due to their similarity with the trabecular architecture of cancellous bone. 15

| Slurry
A photocurable slurry named LithaBone HA 480 E (Lithoz, Austria) was used for the present study. The slurry, designed and produced by Lithoz, comprises hydroxyapatite powder and a photocurable binder matrix containing solvent, reactive monomers based on acrylates and methacrylates, and photoinitiator.

| Printing system
Scaffolds were fabricated by using the CeraFab 7500 system that employs the DLP-based stereolithographic method. This system is highly effective for the manufacturing of complex 3D ceramic structures and uses a LED light source in the blue range of the electromagnetic spectrum (wavelength 460 nm). The light engine utilizes powerful LEDs as a light source and a digital mirror device (DMD) chip as a dynamic mask with a resolution of 1920 × 1080 pixels and a pixel size of 40 × 40 μm 2 . 16 The DMD chip of CeraFab 7500 system enables a spatially selective photopolymerization of each layer; the resolution in the x-y plane is 40 µm and the thickness of a single layer can in the range of 25 to 100 µm, depending on the user's choice; specifically, the layer thickness used in this work was 25 μm. During printing, the light source radiates from below the rotating vat that is uniformly filled with slurry (bottom-up approach). When the photosensitive suspension is irradiated by external blue light, free radicals are formed and polymerization can take place. As a result, the ceramic particles are bonded in the desired pattern by the polymer, acting as glue between them.

| Scaffold design and manufacturing
In order to obtain 3D porous scaffolds closely resambling the architecture of cancellous bone, a computer-aided design (CAD) file created from micro-computed tomography (micro-CT) of a 45-ppi commercial polyurethane sponge was used as the input to the printing system after being converted to STL extension. The average pore size of the porous template, assessed in a previous study by micro-CT imaging, 17 was 580 μm, which is compatible with the typical reference range reported for trabecular bone.
DLP stereolithography was used to carry out the printing process, thus allowing to faithfully reproduce the selected design, while InVesalius software, widely appreciated in various fields of engineering, archaeology and industrial applications as well as in medicine to assist physicians for diagnosis and surgical planning, 18 was used for the processing of the CAD file describing the structural features of the polymeric template. The InVesalius software allowed reconstructing the 3D profile of the polymeric sponge starting from 771 2D image (.bmp) files, thus obtaining an almost cubic 3D structure for scaffold printing (4.72 × 5.15 × 5.00 mm; total porosity of 81.8 vol%).
In order to obtain scaffolds with similar size and geometry to commercial hydroxyapatite products, like Endobon ® , 19 it was decided to fabricate samples with cylindrical shape and length-to-diameter ratio (aspect ratio) of 2:1. This was possible by joining an exact and overturned copy of the sponge structure to the original cuboid from micro-CT imaging, thus obtaining a parallelepiped; then, the circle inscribed on the squared base was considered in order to extrapolate a cylinder with diameter of 5 mm and height of 10 mm. This process is displayed in Figure 1.
In order to ensure successful printing and mechanical integrity of the 3D scaffolds, both scaling (+25%) and contouring operations (contouring offset = +25 μm) were performed on the original STL file during the creation and tuning of the CAD model. This was useful to increase the thickness of the struts in the greens and limit their breakage upon printing and/or cleaning procedures.
The STL data corresponded to file sizes between 80 and 500 MB for an individual part, or multiples if more than one part was printed simultaneously in one run.
Before printing, the surface of the building platform was coated with a special LithaFoil sheet (Lithoz, Austria) which is a thin layer of thermoplastic polymer foil to increase adhesion between the building platform and the first layer created. Eighteen cylindrical scaffolds were produced simultaneously over a complete printing cycle.
In order to obtain scaffolds with final dimensions close to the nominal ones, shrinkage compensation factors equal to 1.206 along the x, y axes and 1.250 along the z axis were applied (oversizing of green samples).

| Post-processing and sintering
After being printed, green samples were removed from the building platform by using razor blades and carefully cleaned by a cleaning agent and pressurized air in order to eliminate the uncured slurry from the micrometric cavities. Cleaned green samples were then stored at 42°C for no longer than 5 days and thermally treated in an electrical furnace (Nabertherm P330) to completely remove the organic binder and sinter the ceramic particles. As shown in Figure 2, a multistep and slow thermal treament was applied, thus avoiding the formation of internal stresses resulting from gases formed during the polymer burn-off. The thermal post-processing protocol was developed on the experience of slurries which were based on the same F I G U R E 1 STL files visualized in Autodesk Netfabb Premium 2019. (A) cubic sponge; (B) "doubled" cubic sponge; (C) final cylindrical structure with aspect ratio of 2:1 basic binder system in Lithoz. 20 The total treatment time was 99 hours (about 4 days); a maximum temperature of 1300°C was reached and mantained for about 2 hours. An auxiliary paste (IPS Object Fix Flow) was used to stabilize and fix the samples to the supporting platform in the furnace; the paste was then easily removed once the sintering process was completed.

| Microstructural and morphological analysis
After being crushed in powder, the scaffolds underwent wide-angle (2θ within 10-70°) X-ray diffraction (XRD) by using a X'Pert Pro PW3040/60 diffractometer (PANalytical, Eindhoven, The Netherlands) in order to identify the crystalline phase(s). Working parameters were voltage 40 kV and current 30 mA; setup included Bragg-Brentano camera geometry with Cu Kα incident radiation (wavelength λ = 0.15405 nm), step size of 0.02° and counting time of 1 second per step. Phase identification was carried out by using X'Pert HighScore software (2.2b) equipped with the PCPDFWIN database.
After being sputter-coated with a thin layer of chromium (10 nm), intact scaffolds were inspected by scanning electron microscopy (SEM; JCM 6000Plus Versatile Benchtop SEM, JEOL).

| Assessment of permeability
At the macroscale level for laminar flows, the proportional relationship between the pressure drop gradient and the fluid volumetric flow rate is described by the Darcy's law [21][22][23] in its empirical form (Equation (1)): where ΔP is the pressure differential upstream and downstream the porous material along the length L of the sample, μ is the dynamic viscosity of the fluid, k is the intrinsic Darcian permeability, Q v is the volumetric flow rate, and A is the cross-sectional area of the sample perpendicular to the flow direction.
In this work, the intrinsic Darcian permeability k was experimentally determined by measuring the pressure wave drop of a slow alternating airflow, 24-26 with a known volumetric airflow rate through the scaffold, according to Equation 2: where q v,rms was the root-mean-square (r.m.s.) value of alternating volumetric airflow rate, that is, , depending on the imposed angular frequency of oscillation ω and the corresponding volume variation ∂V within a given volume of air V 0 ; and p rms was the r.m.s. value of the occurring sinusoidal pressure component, ie p rms = γp 0 ∂V , depending on the atmospheric static pressure p 0 (with heat capacity ratio γ = 1.4) and the volume variation ∂V, with respect to the volume of the air V 0 . The amplitude of the sinusoidal pressure component was measured by a proper capacitive microphone, with a given sensitivity in mV/Pa; the measurement device and the experimental procedure were previously described in detail elsewhere. 24 Therefore, by combining the quantities above described, the experimental intrinsic Darcian permeability k was obtained by the following physical model (Equation (3)): where is the experimental ratio between the microphone sensitivity (in mV/Pa) and the corresponding microphone indication (in mV/Pa) of the pressure wave drop amplitude during the compression and rarefaction ∂V of the air volume V 0 through the scaffold under investigation.
According to Darcy's law, 24,26,27 a very low frequency of oscillation (ω ≈ 1 rad/s) was imposed to keep the laminar regime of the volumetric airflow rate. Once the quantities in Equation (3) have been assessed, the intrinsic Darcian permeability k of the scaffolds was determined from experimental measurements performed on six different samples (diameter 4.99 ± 0.13 mm, height 9.88 ± 0.09 mm).

| Mechanical properties
The compressive strength of scaffolds (σ) was determined as the ratio F∕A by crushing tests, where F was the maximum load registered during the test and A was the initial crosssectional area. Mechanical tests were performed by using (1) Thermal treatment used to consolidate (binder burnoff and sintering) bioceramic scaffolds an MTS System Corp. apparatus (5-kN cell load, cross-head speed 1 mm/min). The elastic modulus was determined from the linear region of the stress-strain curve.
The total porosity of each scaffold was assessed by massvolume measurements as p = (1 − ∕ 0 ), where ρ is the apparent density of the sample (mass-to-volume ratio) and ρ 0 is the bulk density of hydroxyapatite.
Values of porosity, compressive strength, and elastic modulus were expressed as average ± standard deviation calculated on 23 scaffolds. Dimensions of samples were assessed by a digital calliper: diameter 5.01 ± 0.10 mm, length 10.29 ± 0.13 mm.
The Weibull modulus m of the scaffolds was determined following the procedures described in ASTM C1239-07, 28 which also recommends the testing of at least 20 specimens to obtain reliable results. Specifically, the data from compressive tests were fitted through the Equation (4): where P f was the probability of failure at a stress σ and σ 0 was the Weibull scale parameter.
Thus, a plot of ln ln 1 1 − P f vs. ln( ) followed by linear regression gives a straight line of slope m.
The probability of failure P f was estimated as (j − 0.5)∕N , where N is the total number of scaffolds tested and j is the specimen rank in ascending order of failure stresses.
Compressive tests were also performed on scaffolds after immersion in simulated body fluid (SBF), which was prepared according to the Kokubo's protocol, 29 for 14 days (five scaffolds) and 1 month (five scaffolds). This solution mimics the composition of human plasma and is commonly used for in vitro testing of biomaterials. 30 A mass-to-volume ratio (scaffold/solution) of 1.5 mg/ml was used, according to common testing protocols. 31 The samples were kept at 37°C in an orbital shaker incubator (IKA 3510001 KS 4000 I Control Incubator Shaker, IKA-Werke GmbH & Co. KG, Staufen, Germany) with a constant shaking speed of 100 rpm over the whole immersion period.

| RESULTS AND DISCUSSION
In the present study, trabecular-like hydroxyapatite scaffolds for bone repair were obtained by DLP technology using a CAD file created from micro-CT of a polyurethane sponge. The idea behind this study was directly inspired by one of the most famous manufacturing strategies currently available for the production of highly porous, trabecular-like scaffolds: the foam replica technique. Open-cell polymeric foams, in fact, have been widely used as sacrificial templates in the production of bioactive glass and ceramic scaffolds by implementing this versatile and simple strategy, essentially based on the coating of the polymer struts with a layer of glass or ceramic particles. 17,[32][33][34][35] However, this technique suffers from important shortcomings which limit the industrial scale-up of the process and the spread of foam-replicated structure in the clinical practice.
To the best of the authors' knowledge, the structure of a polyurethane foam has never been reproduced so far by an additive manufacturing system applied to bioceramics. This was also the first time that micro-CT was coupled with DLP to produce a "2.0 version" of foam-replicated bone-like bioceramic scaffolds.
Despite their delicate structure and high porosity, the green samples were able to withstand the layer-by-layer printing process and no samples crumbled or detached from the building platform before the end of the process ( Figure  3A). The size, geometry and porous architecture of the green bodies were not apparently damaged by the cleaning process ( Figure 3B), which was carried out very cautiously due to the thinness of the struts ( Figure 3C). The as-printed samples (the so-called green bodies) initially exhibited a gray-white appearance and then changed colour to light blue after sintering ( Figure 3D). This is a normal chromatic effect in sintered hydroxyapatite, which was also observed in some commercial products (eg, Bio-Eye ® porous hydroxyapatite orbital implant): negligible traces of manganese can change the hydroxyapatite color to blue after sintering at high temperature in oxidizing atmosphere as a result of the oxidation of manganese ions within the crystalline lattice of hydroxyapatite. 36 Preliminary tests revealed a shrinkage of ~17% in xy direction and ~19% in z direction after sintering. Overall, the sample shrinkage was low and almost uniform in all the spatial directions as compared to the results reported in other studies on DLP-based grid-like hydroxyapatite scaffolds, which exhibited heterogeneous and higher shrinkage (~36% in xy direction vs. ~14% in z direction) in spite of being sintered at lower temperature (1200°C for 2 hours). 37 This suggests that there was no direction undergoing more stress than another one in the samples sintered in the present work. The application of shrinkage compensation factors yielded final dimensions of the scaffolds very close to the nominal ones, as alredy reported in the Section 2.2.
XRD analysis performed on sintered scaffolds ( Figure  4) confirmed the presence of hydroxyapatite as the unique crystalline phase; the JCPDS code for such phase (00-009-0432) is the same found by other authors who fabricated grid-like scaffolds by DLP-based stereolithography. 37 Neither impurities (eg, CaO) nor secondary calcium phosphate phases were detected unlike in the article by Zeng et al., 37 who reported the conversion of small fractions of hydroxyapatite to both β-and α-tricalcium phosphate (TCP) upon sintering above 1200°C. The stability of (4) ln ln hydroxyapatite phase in our scaffolds could be justified by the high-quality of the starting materials (Ca-to-P weight ratio of 2.35 with 99.9% purity according to the test certificate, as assessed by Rietveld calculation), which was shown to play a role in inhibiting the phase transformation of hydroxyapatite during sintering. 38 Morphological investigations shown in Figure 5A,B confirm that the 3D pore-strut architecture of the scaffolds closely reproduces the cellular structure of the starting polymeric template (virtual model) and mimics the trabecular organization of cancellous bone. The macropore size is in the range of 100 to 800 µm, which is optimal to allow diffusion of nutrients/removal of waste byproducts, bone cell colonization, blood vessel infiltration and new bone regeneration. [39][40][41] Struts are approximately round-shaped with diameter of about 259 ± 30 μm, well-densified ( Figure 5C) and without any macroscopic defects, suggesting a successful sintering of the scaffold. Surface microporosity deriving from interparticle voids ( Figure 5D) can be beneficial to promote the adhesion of osteoblasts to the scaffold struts. 42 The total porosity of sintered hydroxyapatite scaffolds was 0.80 ± 0.020, which is comparable to the typical range of healthy human cancellous bone (0.50-0.85 [43][44][45] ) and, therefore, suggests a potential suitability for bone substitution. The low value of standard deviation for scaffold porosity also indicates a good reproducibility of the manufacturing process.
Dealing with scaffolds for bone tissue engineering applications, an accurate determination of intrinsic permeability is crucial to predict scaffold performances in the physiological environment. Intrinsic permeability, in fact, characterizes both scaffold architecture and mass-transport properties, and hence quantifies the ability of a porous material to conduct fluid flow through its whole volume. When a 3D bone scaffold is implanted, proper mass transport has to be guaranteed to promote cell migration from the periphery to the internal core and favor nutrient/oxygen supply and catabolite elimination, thus improving the quality of the final tissue. 46 The intrinsic permeability of the hydroxyapatite scaffolds is in the range of 0.75 to 1.74 × 10 −9 m 2 , which is comparable to the lower ranges determined for human cancellous bone harvested from calcaneus (0.4-11.0 × 10 −9 m 2 47 ) and vertebral body (1.5-12.1 × 10 −9 m 2 46 ); a closer similarity was observed with bone from proximal femur (0.01-4.7 × 10 −9 m 2 46 ). Scaffold permeability could be further optimized by refining the 3D model in the CAD file used for printing (e.g., adjustment of strut diameter/pore size). For the sake of comparison, it is interesting to mention that the intrinsic permeability of commercial clinically used calcium carbonate scaffolds, analysed under the same experimental setup (acoustic method), was two orders of magnitude lower (around 3.12 × 10 −11 m 2 ). 24 This further demonstrates that the overall strategy used to fabricate hydroxyapatite scaffolds (micro-CT combined with DLP-based stereolithography) has the potential to lead to an improvement over the scientific, commercial and clinical state of the art and to developing porous biomaterials with bone-like architecture and mass transport properties.
The compressive strength of scaffolds as such and after immersion in SBF for 2 weeks and 1 month was 1.60 ± 0.79, 1.28 ± 0.32, and 1.54 ± 0.36 MPa, respectively. All these values within the typical range of trabecular bone (0.1-16 MPa 48,49 ), although being close to the lower limit. Prolonged immersion in SBF did not lead to any significant variation in the mechanical strength of scaffolds; this can be justified by the fact that hydroxyapatite a very slow resorption rate in biological fluids, unless in the form of nanoparticles. 50 These values are from 4 to 8 times higher than those obtained for hydroxyapatite scaffolds (0.21-0.41 MPa) having analogous total porosity (0.82-0.86) and fabricated by conventional sponge replication. 51 Another study revealed that the use of nanosized hydroxyapatite particles (diameter around 150 nm) to coat the polyurethane sponge in the conventional method yielded no significant improvement in the compressive strength (0.51 MPa). 8 This suggests that, although the same template (polyurethane sponge) is used in both DLP-based stereolithography (as a virtual model in the present work) and conventional foam replication, the production method indeed plays a key role in affecting the mechanical performance of scaffolds and, thus, their clinical suitability for bone repair.
Grid-like hydroxyapatite scaffolds produced by DLPbased stereolithography and sintered under analogous conditions than the present trabecular samples exhibited comparable compressive strength (1.45-1.92 MPa) but, surprisingly, markedly lower porosity (0.49-0.52) around the minimum threshold for cancellous bone. 52 A typical stress-strain curve obtained upon compression is depicted in Figure 6.
The elastic modulus of scaffolds, 513 ± 290 MPa, is comparable to the reference range for cancellous bone (50-500 MPa 53,54 ) this is an important result to avoid stiffness mismatch between implanted biomaterial and host bone. Figure 7 shows the Weibull plot of the compressive strength data; least mean squares fitting of a straight line through the experimental points yields Weibull modulus m = 2.2 and Weibull scale parameter σ 0 = 1.81 MPa, corresponding to the stress value at which 63% of the specimens are expected to survive. The Weibull modulus determined from the strength data for a large number of theoretically identical samples (more than 20) provides a measure of the mechanical reliability of brittle materials, like ceramics. The mechanical response of brittle materials is sensitive to microstructural flaws such as pores and micro-cracks: therefore, Weibull modulus is used to evaluate the probability of failure in such materials under a given stress. This parameter was seldom determined for bioceramic scaffolds in the literature; the Weibull modulus of robocast hydroxyapatite, β-TCP and calcium polyphosphate scaffolds has been reported to be in the range of 3.2 (perpendicular to ceramic rods) to 9.9 (parallel to ceramic rods) upon testing in compression, 11,55 and an improvement from 3.0 to 7.7 could be achieved by applying a polymeric coating (poly(lactic acid) and poly(ε-caprolactone)) on β-TCP structures. 56 Furthermore, β-TCP scaffolds produced by DLP exhibited a Weibull modulus ranging from 3.2 to 7.4 depending on the architectural geometry (eg, grid-like and hexagonal). 20 However, all these calcium phosphate scaffolds had a less porous, ordered structure of parallel/perpendicular rods instead of a highly porous foam-like architecture, like that of the hydroxyapatite scaffolds produced in this work; such a difference could explain the relatively low value of Weibull modulus in the latter case.
The fabrication approach followed in this work, relying on the "ceramic replication" of a polymeric sponge by DLP-based stereolithography, allows obtaining standardized scaffolds with high control on pore/strut characteristics, good reproducibility, appealing market potential and higher mechanical strength compared to conventional sponge replication. Indeed, some degrees of freedom in scaffold design are still possible by properly processing the starting CAD file (volumetric reconstruction) from micro-CT imaging. In this work, thickening of 25 μm was applied to the sponge struts but other options are possible on F I G U R E 6 Representative stressstrain curve obtained upon compressive mechanical tests F I G U R E 7 Weibull plot with linear fitting for the compressive strength of DLPderived hydroxyapatite scaffolds demand: the proposed fabrication platform is potentially suitable for both mass production and customization of scaffolds (personalized medicine), if necessary.

| SUMMARY
It has been shown, for the first time, that porous bioceramics can be produced by DLP-based stereolithography using a micro-CT reconstruction of an open-cell polymeric foam as a virtual template (CAD model). As a result, it was possible to fabricate hydroxyapatite scaffolds with 3D trabecular architecture, pore size, intrinsic permeability, elastic modulus, and compressive strength comparable to those of human cancellous bone. This novel approach opens new horizons in the functional design and manufacturing of bone-like scaffolds that could be highly beneficial in bone replacement and bone tissue engineering strategies.