**Lactoferrin-Hydroxyapatite Containing Spongy-Like Hydrogels for Bone Tissue Engineering**

**Ana R. Bastos 1,2 , Lucília P. da Silva 1,2 , F. Raquel Maia 1,2,3, \* , Sandra Pina 1,2 , Tânia Rodrigues 1,2 , Filipa Sousa 1,2 , Joaquim M. Oliveira 1,2,3 , Jillian Cornish 4 , Vitor M. Correlo 1,2,3 and Rui L. Reis 1,2,3**


Received: 24 May 2019; Accepted: 25 June 2019; Published: 27 June 2019

**Abstract:** The development of bioactive and cell-responsive materials has fastened the field of bone tissue engineering. Gellan gum (GG) spongy-like hydrogels present high attractive properties for the tissue engineering field, especially due to their wide microarchitecture and tunable mechanical properties, as well as their ability to entrap the responsive cells. Lactoferrin (Lf) and Hydroxyapatite (HAp) are bioactive factors that are known to potentiate faster bone regeneration. Thus, we developed an advanced three-dimensional (3D) biomaterial by integrating these bioactive factors within GG spongy-like hydrogels. Lf-HAp spongy-like hydrogels were characterized in terms of microstructure, water uptake, degradation, and concomitant release of Lf along the time. Human adipose-derived stem cells (hASCs) were seeded and the capacity of these materials to support hASCs in culture for 21 days was assessed. Lf addition within GG spongy-like hydrogels did not change the main features of GG spongy-like hydrogels in terms of porosity, pore size, degradation, and water uptake commitment. Nevertheless, HAp addition promoted an increase of the pore wall thickness (from ~13 to 28 µm) and a decrease on porosity (from ~87% to 64%) and mean pore size (from ~12 to 20 µm), as well as on the degradability and water retention capabilities. A sustained release of Lf was observed for all the formulations up to 30 days. Cell viability assays showed that hASCs were viable during the culture period regarding cell-laden spongy-like hydrogels. Altogether, we demonstrate that GG spongy-like hydrogels containing HAp and Lf in high concentrations gathered favorable 3D bone-like microenvironment with an increased hASCs viability with the presented results.

**Keywords:** gellan gum; hydroxyapatite; lactoferrin; bone biomaterials

#### **1. Introduction**

Over recent years, tissue engineering (TE) has been pushing forward bone tissue regeneration field, especially in what concerns biomaterials' improvement. Further, the main struggle of bone tissue engineering is to reach a synergistic effect through the combination of biomaterials and cells [1]. Researchers have been demonstrating the importance of biochemical and microstructure cues in the success of designed biomaterials [2]. An optimal biomaterial should not only provide an adequate structural support for bone tissue engineering, but also promote tissue re-growth [3]. The essential

prerequisites include: (i) a three dimensional (3D) porous structure, presenting (ii) optimized surface properties, which will potentiate cell attachment, migration, proliferation, and the retention of normal cell functions; (iii) sustained biodegradability; (iv) biomechanical properties that mimic the native tissue; (v) reproducibility; and, (vi) bioactivity, to replace large cortical bone defects and enable load transmission [4–7].

Lf is an iron-binding glycoprotein from the transferrin family [8]. Lf is well known by its antimicrobial, antibacterial, antiviral, antiparasitic anti-neoplastic, anti-inflammatory, and immunomodulatory activities on the immune system, which is a plus for every approach [9–11]. Nowadays, despite the molecular mechanism and action of this protein not being completely understood, Lf alone or combined with others systems has been explored for bone tissue engineering [3,12]. Lf alone has been showed to promote the proliferation of rat osteoblasts in a dose and time dependent manner [13]. The integration of recombinant human Lf and MC3T3-E1 osteoblast-like cells in an injectable hydrogel has been shown to enable MC3T3-E1 cells viability, proliferation, and differentiation supporting proteins phosphorylation/dephosphorylation [14]. In vivo systemic administration of Lf alone [15,16] or in a gelatin hydrogel [17] has also been shown to enhance the new bone formation on a bone defect site.

Hydroxyapatite (HAp) (Ca10(PO4)6(OH)2) presents a huge similarity to the inorganic component of the bone matrix and it has been widely used in bone tissue engineering as bone graft materials, coatings for implants, and bone fillers due to its ability to directly bond to the deficient apatite layer of the bone through the carbonated calcium [18–21]. HAp also presents osteoconductive and osteoinductive properties, as it facilitates the migration, adhesion, proliferation, and differentiation of progenitor cells, and the cell-mediated release of growth factors that stimulate bone formation in vivo [22]. We have previously developed Gellan Gum (GG) spongy-like hydrogels that contain HAp for bone tissue engineering [23,24]. GG spongy-like hydrogels show optimal conditions for Tissue Engineering and Regenerative Medicine (TERM) due to their porous microstructure arrangement, mechanical stability, as evidenced by their high flexibility and resilience to deformation, and high water content that altogether potentiate cell adhesion and spreading [25–27]. The addition of HAp (10 and 20% w/v) and CaCl2, as a crosslinker, to the GG spongy-like hydrogels, improved the bioactivity of the materials, as evidenced by the formation of uniformly distributed apatite-like crystal phases on materials surface in a osteogenic medium, being similar in terms of composition and structure to bone-apatite [24]. In osteogenic culture conditions, HAp-containing spongy-like hydrogels prompted hASCs adhesion and spreading [24], as well as the differentiation of bone marrow cells that were isolated from mice long bones towards pre-osteoclasts [23] in the presence of vitamin D3.

The combination of Lf and HAp has been scarcely explored for bone tissue engineering. From the few existing in vitro studies, Lf-Hap have shown promising results, as evidenced by enhanced osteoblasts proliferation [3,28] and rADSCs differentiation towards osteoblasts [29]. Hence, in this work, we combined the unique properties of Lf and HAp in spongy-like hydrogels to reinforce the bone-like microenvironment. Different Lf-HAp spongy-like hydrogels formulations were developed and characterized, envisaging a material with appropriate physical-chemical properties and Lf temporal release for future bone tissue engineering approaches. Human adipose-derived stem cells (hASCs) were seeded within the Lf-Hap spongy-like hydrogels due to their promising features, such as their regenerative potential, self-renewal properties, capacity to differentiate into the osteogenic lineage, as well as their high and easy availability from the subcutaneous liposuction from adipose/fat tissue [30,31]. The ability of these biomaterials to behave as a platform to support hASCs cell adhesion and growth was then studied.

#### **2. Materials and Methods**

#### *2.1. Preparation of Hydrogels*

Gelzan CM (Sigma, St. Louis, MO, USA) was dissolved in distilled water at 90 ◦C for 30 min. with stirring. Lactoferrin (Lf, Bovine origin, New Zealand) was dissolved in distilled water. Subsequently, to the solution of Gellan Gum (GG), Lactoferrin (Lf), and Hydroxyapatite (HAp, Plasma Biotal, Buxton, UK) were added at 50 ◦C, as described in Table 1. After complete homogenization, the crosslinking solution CaCl<sup>2</sup> (Sigma, USA) was added and it was allowed to stabilize at room temperature for 30 min. Posteriorly, the hydrogels were cut in discs of 5 mm of diameter and 2 mm of thickness and replenished with phosphate-buffered saline solution (PBS, Sigma, USA) for 30 h to enable complete crosslinking, frozen at −80 ◦C for 18–20 h, and then freeze-dried (CryoDos -80, Telstar, Terrassa, Barcelona, Spain) for at least three days. Materials were sterilized by ethylene oxide.

**Table 1.** Different formulations of Gellan Gum (GG) spongy-like hydrogels with or without Lactoferrin (Lf), and/or Hydroxyapatite (HAp).


#### *2.2. Physico-Chemical Characterization of Spongy-Like Hydrogels'*

#### 2.2.1. Fourier Transformed Infrared Spectroscopy (FTIR)

Fourier Transformed Infrared (FTIR) analysis was performed while using attenuated total reflectance (ATR) (IRPrestige-21, Shimadzu Corporation, Kyoto, Japan) in transmittance mode and in the region of 550–4000 cm−<sup>1</sup> for spectroscopic study.

#### 2.2.2. X-ray Diffraction (XRD) Analysis

The qualitative analyses of crystalline phases that were presented on the GG, GG/Lf, GG/HAp, and GG/Lf/HAp spongy-like hydrogels were obtained by XRD while using a conventional Bragg–Brentano diffractometer (Bruker D8 Advance DaVinci, Rheinstetten, Germany) that was equipped with CuKα radiation, produced at 40 kV and 40 mA. The data sets were collected in the 2θ range of 5–70◦ with a step size of 0.04◦ and 1 s for each step.

#### 2.2.3. Scanning Electron Microscopy

Scanning electron microscopy (SEM, JSM-6010 LV, JEOL, Tokyo, Japan) and micro computed tomography (micro-CT) were used to assess the microstructure of the dried polymeric networks of the different formulations. For SEM analysis, the samples were immersed in liquid nitrogen to cut the sample into cross-sections, and the internal surface of the dried polymeric networks was observed.

#### 2.2.4. Micro Computed Tomography

The microstructure of dried polymeric networks was analyzed while using a high resolution X-Ray microtomography SkyScan 1272 System (SkyScan, Kontich, Belgium). The pixel size used was 2.5 µm, the exposure time was 1s, and the source conditions used were 50 kV of energy and 200 µA of current. A binary picture was created using 150 slides while using a thresholding between 20 and 255 in a grey scale. The morphometric analysis, which includes porosity, pore size, interconnectivity, and wall thickness, were assessed by CT-analyzer program (CTAn, v1.17.0.0., SkyScan, Belgium).

#### 2.2.5. Degradation Studies: Mass Loss

Lf-HAp containing spongy-like hydrogels were immersed in PBS at 37 ◦C for 30 days, with stirring. The initial weight (w<sup>i</sup> ) of the samples was recorded and the samples were then immersed in PBS and weighed again (final weight, w<sup>f</sup> ). The percentage of mass loss along the time was calculated according to the equation:

$$\text{Mass loss} \left(\% \right) = \frac{\left(\text{W}\_{\text{f}} - \text{W}\_{\text{i}}\right)}{\text{W}\_{\text{i}}} \times 100$$

Every three days, the PBS solution was replaced and the supernatant was collected and then kept at −80 ◦C for further analysis of the amount of Lf released.

#### 2.2.6. Water Uptake

Lf-HAp containing spongy-like hydrogels were weighed in the dried state (wd) and were then immersed in PBS at 37 ◦C during three days. Along this period of time, at different time points (30 min., 1, 2, 4, 6, 24, 48, and 72 h), the samples were weighed in the wet state (ww). The percentage of water uptake along the time was calculated according to the equation:

$$\text{Water Ulp take} \left(\% \right) = \frac{\left(\text{W}\_{\text{W}} - \text{W}\_{\text{d}}\right)}{\text{W}\_{\text{d}}} \times 100$$

#### 2.2.7. Quantification of Lf Release: BCA Assay

A micro-Bicinchoninic Acid Assay (micro-BCA) was performed according to manufacturer instructions In order to evaluate the amount of Lf released from spongy-like hydrogels. Briefly, 75 µL of bicinchoninic acid was placed in each well of a 96 well-plate and 75 µL of the supernatant, previously collected during mass loss assessment, was added to each one. These well-plates were incubated for 2 h at 37 ◦C. After this time, the absorbance at 562 nm was measured while using a Microplate Reader (SYNERGY HT, BIO-TEK, USA). The supernatants from GG and GG with HAp spongy-like hydrogels were used as Blank. The corrected absorbance readings were converted into protein concentrations while using a calibration curve of Lf. The theoretical value according to the final volume and concentration of Lf into GG spongy-like hydrogels was calculated. GG/Low Lf containing spongy-like hydrogels, according to the theoretical calculations, can release a maximum of 20 µg, while G/High Lf can release a maximum of 60 µg.

#### *2.3. In Vitro Studies*

#### 2.3.1. Cell Isolation and Culture

Human adipose-derived stem cells (hASCs) were isolated from the lipoaspirate samples following a protocol previously established with the Department of Plastic Surgery of Hospital da Prelada (Porto, Portugal). All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Hospital da Prelada (P.I. Nº 005/2019) and 3B's Research Group. The isolated cells were used for subsequent studies. The cells were routinely cultured in α-MEM medium (Alfagene, Carcavelos, Portugal) that was supplemented with 10% fetal bovine serum (FBS, Australia origin, Alfagene, Portugal), 1% antibiotic/antimycotic (Alfagene, Portugal), and maintained at 37 ◦C under a humidified atmosphere of 5% v/v CO<sup>2</sup> in air. The medium was changed twice a week and cells, at maximum passage 4, were seeded into spongy-like hydrogels, as described in the next section.

#### 2.3.2. Cell Entrapment

A cell suspension (40 <sup>×</sup> <sup>10</sup><sup>3</sup> , 20 µL) was dispensed dropwise on the top of the polymeric networks, which were previously hydrated for 15 min. with culture medium. Posteriorly, these constructs were incubated at 37 ◦C during 3 h, with 5% CO<sup>2</sup> to allow for maximum cell entrapment within the structures. Fresh medium (500 µL) was added after this period of time.

#### 2.3.3. Cell Viability

Cell viability was assessed at 21 days while using 20% (v/v) of Alamar Blue reagent (ALAMAR BLUE®, AbD, Kidlington, Oxford, UK) in α-MEM culture medium, followed by 3 h of incubation at 37 ◦C with 5% CO<sup>2</sup> and protected from light. Afterwards, 100 µL of supernatant was transferred from each well in triplicate to a new 96-well cell culture plate. Fluorescence intensity was read at 530/20 nm (excitation) and 590/35 nm (emission) using a microplate reader (Synergy HT, Bio-Tek, USA). Alamar Blue in medium was used as a blank. DNA normalized the corrected absorbance readings at day 21. Three specimens of each formulation were used to assess cell viability along the time and three independent experiments were performed.

#### 2.3.4. DNA Content

DNA content was assessed after 21 days of culture and it was used to normalize the cell viability values of the same time point. Constructs were washed with PBS and transferred to 1.5 mL tubes and heated at 70 ◦C for 30 min. Next, 1 mL of ultra-pure water was added into each tube that were placed at 37 ◦C for 1 h, and then frozen at −80 ◦C until analysis. Before DNA quantification, the samples were placed in an ultrasound bath for 1 h at 37 ◦C. Finally, Quant-IT PicoGreen dsDNA Assay Kit 2000 assays (Alfagene®, Portugal) was used according to the manufacturer's instructions. The fluorescence intensity was read at 485/20 nm (excitation) and 530/20 nm (emission) while using a microplate reader (Synergy HT, Bio-Tek, Winooski, VT, USA) and the readings were converted while using a standard curve that was produced with standard dsDNA solutions at different concentrations.

#### 2.3.5. Cytoskeleton Morphology (Phalloidin/DAPI)

After three and 21 days of culture, the constructs were washed with PBS during 5 min., fixed with 10% formalin for one hour, and then washed again. Next, Phalloidin-TRITC (Sigma, USA) (1:80) was added and placed at room temperature during 1 h, protected from light. The constructs were counterstained with 4,6-Diamidino-2-phenylindole, Dilactate (DAPI, Sigma, USA) (1:5000) for 5 min. in the dark. Finally, they were visualized by Confocal Laser Scanning Microscope (TCS SP8, Leica, Mannheim, Germany).

#### *2.4. Statistics*

Statistical analysis was performed while using the GraphPad Prism 5.0 software. The data were analyzed while using the Shapiro–Wilk normality test. The results that did not present a normal distribution were analyzed using the Kruskal–Wallis test with Dunn's multiple comparison post-test for statistical analysis. The results that presented a normal distribution were analyzed while using a One-way ANOVA and Tukey's multiple comparisons T-test. The significance level between the groups were set for \* *p* < 0.05, \*\* *p* < 0.01, and \*\*\* *p* < 0.001. The data were presented as mean ± standard deviation (SD). Data in each figure are representative experiment of three experiments (n = 3).

#### **3. Results**

#### *3.1. FTIR and XRD Analysis*

The ATR-FTIR spectra of the spongy-like hydrogels that are presented in Figure 1 show the stretching vibrations of C–O–C bonds and the bending mode of mehtyl vC–H, respectively, at 1022 and 1593 cm−<sup>1</sup> [32]. Regarding the Lf addition to the GG formulations, a signal of tyrosine was registered at 1138 cm−<sup>1</sup> [33]. Additionally, an absorption peak appears at 1010 cm−<sup>1</sup> , which can be assigned to N≡C or C=C stretch and C–H deformation vibrations of tryptophan. The presence of HAp was ascertained by the detection of the peak around 600 cm−<sup>1</sup> , in the region of the v4 bending mode of PO<sup>4</sup> <sup>3</sup><sup>−</sup> and the peak around 1000 cm−<sup>1</sup> that corresponded to the region of the v1 stretching mode of PO<sup>4</sup> <sup>3</sup><sup>−</sup> [24]. All of the GG/Lf/HAp spongy-like hydrogels show the broad band between 3500 and 3700 cm−<sup>1</sup> that is attributed to the hydroxyl groups (vOH) stretching vibration due to the medium hydrogen bond of intramolecular and intermolecular type.

Figure 2 displays the XRD patterns of the spongy-like hydrogels. It can be observed that all of the compositions show the intensity peaks corresponding to the GG crystalline structure with the main peak being located at 32◦ [32]. The presence of a higher content of Lf led to a slight crystallinity decrease without additional peaks being observed (Figure 2iii). The crystallographic phases identification of the hydrogels containing HAP was accomplished by comparing the XRD patterns with the standard ICDD PDF 01-074-0565 of HAp (Figure 2iv–ix). As expected, with increasing the HAp content in the hydrogels, the crystallinity became more evident, as shown in Figure 2v,viii,ix.

**Figure 1.** Fourier Transformed Infrared (FTIR) spectra of GG/Lf/HAp spongy-like hydrogels: (i) GG, (ii) GG/Low Lf, (iii) GG/High Lf, (iv) GG/Low HAp, (v) GG/High HAp, (vi) GG/Low Lf/Low HAp, (vii) GG/High Lf/Low HAp, (viii) GG/Low Lf/High HAp, and (ix) GG/High Lf/High HAp.

**Figure 2.** X-ray Diffraction (XRD) patterns of the GG/Lf/HAp spongy-like hydrogels: (i) GG, (ii) GG/Low Lf, (iii) GG/High Lf, (iv) GG/Low HAp, (v) GG/High HAp, (vi) GG/Low Lf/Low HAp, (vii) GG/High Lf/Low HAp, (viii) GG/Low Lf/High HAp and (ix) GG/High Lf/High HAp. HAp (ICDD PDF 01-074-0565), and \* GG.

#### *3.2. SEM and Micro-CT Analysis*

Figure 3 shows the representative images of the cross section surface of the different GG/Lf/HAp dried polymeric networks (DPN) that were obtained by SEM. GG DPN containing just Lf (Low or High) showed a porous structure with smooth surfaces, which indicated that the presence of Lf had no effect in the microstructure when compared with the control (GG) (Figure 3A–C). Regarding GG DPN containing HAp (Figure 3D,E), smaller pores and rough surfaces were observed. This effect was more evident in the formulations containing higher HAp concentrations (1% in relation to 10%). The microstructure of GG DPN containing Lf/HAp was similar to the respective GG/HAp DPN (Figure 3F,G in relation to Figure 3D; Figure 3H,I in relation to Figure 3E). These results indicate that no evident effect was observed by adding Lf, while the addition of HAp affected the microstructure.

**Figure 3.** SEM representative images of GG dried polymeric networks (DPN) with or without Lactoferrin (Lf) and/or Hydroxyapatite (HAp): (**A**) GG, (**B**) GG/Low Lf, (**C**) GG/High Lf, (**D**) GG/Low HAp, (**E**) GG/High HAp, (**F**) GG/Low Lf/Low HAp, (**G**) GG/High Lf/Low HAp, (**H**) GG/Low Lf/High HAp, and (**I**) GG/High Lf/High HAp.

Porosity, pore wall thickness, and average pore size were quantified by micro-CT analysis (Table 2). The obtained results revealed that the addition of HAp has an inverse effect on the porosity of the DPN: the porosity tended to decrease with the increase of HAp concentration from ~87% to 64%, respectively. Despite this tendency, no other statistically significant differences on the average porosity were observed. Nevertheless, the same trend was verified on the mean pore size for all of the conditions: a decrease of pore size was observed when HAp is added. For these formulations, the decrease on pore size was followed by an increase of the pore wall thickness, although no significant differences were observed. In accordance to SEM results, the HAp addition had a higher impact in the microstructure of GG polymeric networks than Lf addition. In addition, the 2D microarrangement of the dried polymeric networks can be observed in Figure 4.

**Table 2.** Micro computed tomography (Micro-CT) analysis of (A) mean porosity, (B) pore size, and (C) wall thickness of GG dried polymeric networks (DPN) with or without Lf and/or HAp. Data was presented as mean ± stdev, the statistical analysis was performed using a Kruskal–Wallis test followed by Dunn's test, in which all formulations were compared between them.


**Figure 4.** Two-dimensional (2D) images of the dried polymer networks of GG/Lf/HAp spongy-like hydrogels obtained by Micro-CT.

#### *3.3. Water Uptake Analysis*

The water uptake of the different GG DNP formulations was followed up to three days of immersion into PBS (Figure 5). All of the formulations have shown a burst of water uptake in the first hours of immersion, followed by an equilibrium phase corresponding to the maximum of water content. GG showed a water content of 2385% ± 357 after three days of immersion (Figure 5B). Similar values were observed for the GG/High Lf formulation. However, GG/Low Lf formulations have shown lower water content (1780% ± 143). All of the formulations containing HAp showed lower water content values (~1500%) when comparing to GG and both GG/Lf formulations. GG spongy-like hydrogels with 1 (GG/Low HAp) and 10% (GG/High HAp) of HAp showed a water content of 886% ± 160 and 603% ± 62, respectively. GG spongy-like hydrogels with 1% of HAp and Lf, independent of the concentration, showed a water content between 1277% ± 109 and 1438% ± 438, while the formulations containing 10% of HAp and Lf, the water content was lower (~300/500%). These results have shown that the addition of HAp at higher concentration (10%) to the GG formulation significantly decreased (*p* < 0.05) their water uptake ability (from 2385% ± 357 to 603% ± 63). Likewise, a significant decrease on water content between GG/High Lf (~2385% ± 683) and GG/High Lf/High HAp (~375% ± 35) (*p* < 0.001) was also observed due to HAp addition.

**Figure 5.** (**A**) Water uptake profile of dried polymeric networks (DPN) along 72 h (**B**) water content after 72 h. Data was presented as mean ± stdev and statistical analysis was performed while using a Kruskal-Wallis test followed by Dunn's test.

#### *3.4. Degradation Tests and Lf Release Analysis*

The degradation of different formulations was followed up to 30 days by quantifying the weight loss (Figure 6A). GG/Low Lf and GG/High Lf spongy-like hydrogels have shown higher mass loss in comparison with the other formulations (Figure 6A). In accordance, the GG/High Lf spongy-like hydrogels showed a significant mass loss when comparing to GG (*p* < 0.05), GG/Low Lf (*p* < 0.01), and Lf/HAp (*p* < 0.001) containing spongy-like hydrogels.

**Figure 6.** Degradation and Lf release analysis. Final mass of dried polymeric networks after 30 days (**A**). Lf release profile of Lf-HAp containing spongy-like hydrogels along 30 days (**B**) and at day 30 (**C**). The statistical analysis of the final mass was performed using a One-way ANOVA and Tukey's multiple comparisons T-test, while the Kruskal–Wallis test followed by Dunn's test was used for Lf release and all the formulations were compared between them.

#### *3.5. Cell Viability and Cell Cytoskeleton Analysis*

hASCs' cytoskeleton within spongy-like hydrogels was analyzed by Phalloidin staining (Figure 7). Independent of the spongy-like hydrogel formulation, cells attached and showed a spread morphology within the polymeric structure from three days of culture onward 21 days. Figure 4 shows cell cytoskeleton organization within GG/High Lf/High HAp, which is representative of the other spongy-like hydrogel formulations.

Cell viability within the spongy-like hydrogels was assessed through Alamar Blue®assay that was normalized by DNA at that time point (21 days) (Figure 8A). It was possible to verify that hASCs remained metabolically active after 21 days of culture. Noteworthy, the formulations containing Lf in combination or not with HAp showed the highest metabolic activity, while GG and GG/Low HAp spongy-like hydrogels showed the lowest metabolic activity. GG/High Lf, GG/High Lf/Low HAp, GG/Low Lf/High HAp, and GG/High Lf/High HAp showed significantly higher metabolic activity when comparing to the GG spongy-like hydrogels. A significant increase in metabolic activity was also verified between GG/Low HAp and GG/High Lf/Low HAp. Regarding the DNA concentration depicted in Figure 5B, it was possible to observe that, after 21 days, the cells were present in all the spongy-like hydrogels tested. Moreover, the results showed that spongy-like hydrogels with Lf combined or not with HAp have higher amounts of DNA, while GG spongy-like hydrogels have the lowest. However, statistical differences were not observed.

**Figure 7.** Representative images of cytoskeleton morphology (Phalloidin-TRITC, red) and nuclei (DAPI,blue) of hASCs within GG/High Lf/High HAp spongy-like hydrogels after (**A**) three and (**B**) 21 days of culture.

**Figure 8.** Cell viability assessment. (**A**) Metabolic activity of hASCs' normalized by Day 1 after 21 days of culture. (**B**) DNA concentration of hASCs' within spongy-like hydrogels after 21 days. The statistical analysis was performed using a One-way ANOVA and Tukey's multiple comparisons T-test and results were presented as mean ± standard deviation (SD) and the significance level between groups was set for: \* *p* < 0.05, and \*\* *p* < 0.01.

#### **4. Discussion**

New biomaterials, cells, and growth factors are being combined to reach a synergistic effect and produce functional tissue engineered bone substitutes. With this in mind, we developed a batch of novel bioactive GG spongy-like hydrogels that contain different concentrations of Lf and/or HAp, and tested its capacity to support hASCs culture, envisioning its use as improved scaffolds for bone tissue engineering. In fact, hASCs can differentiate along the osteogenic lineage when submitted to specific growth factors, such as Lf [15,16,34].

The biomaterials microarchitecture, including pore size and porosity, are crucial parameters to match the native tissue characteristics and the required integrity [35]. This work showed that the presence of HAp significantly reduced the porosity and pore size. The HAp functioned as a nucleation agent for ice crystals during the freezing step of spongy-like hydrogels preparation, leading to the formation of pores in a higher amount, but with lower sizes [25,27]. Nevertheless, the results that were obtained regarding porosity and pore size, for all of the formulations, are in agreement with the literature that indicate that porosity between 35 and 75% and the pore size in the range of 50 to 400 µm are appropriated to be used in bone tissue engineering applications [36]. In our study, all of the developed spongy-like hydrogels presented pore sizes that were between ~40 to 80 µm, which are within the range that was described to allow for the ingrowth of capillaries and facilitate the exchange of nutrients and discharge of metabolites [1]. In fact, we could verify that hASCs that were cultured within all spongy-like hydrogels were metabolically active. Furthermore, the cells cultured within spongy-like hydrogels with Lf and/or HAp were more metabolic active than cells that were cultured within GG spongy-like hydrogels. This was an expected result, since Gellan Gum hydrogels do not present specific attachment sites for anchorage-dependent cells. When considering the pore size, it was also observed that the addition of HAp lead to a decrease in the porosity while thicker and rough pore walls were promoted. The commitment of this combined effect together with the intrinsic properties of the GG spongy-like hydrogels can overcome the problems that are associated with stress shielding [37]. It is well known that the thicker pore walls and roughness might be crucial for cells biological response since it directly influences cell migration and proliferation [38]. Moreover, it promotes structural support and adhesion sites, facilitates cell movement, regulates cell behavior, and sustains cell-to-cell recognition.

On the other hand, the microarchitecture of spongy-like hydrogels was not altered with the addition of Lf. In fact, this small protein was predicted to adsorb to the GG polymeric networks and it did not act as a nucleation agent for ice crystals in the freezing stage. GG spongy-like hydrogels Lf containing have shown higher porosity and water retention capacity, when compared with HAp containing spongy-like hydrogels. Similarly, in other studies, it has been verified that, when HAp is incorporated into a matrix, a lower water content is adsorbed/expected [23,39,40]. The achievement of the maximum water content occurs earlier due to the existence of lower microvoids between the interface of the GG matrix and the HAp crystals. Since HAp has very low water absorption, when it is added to the GG matrix, its natural/intrinsic hydrophilicity decreases. Overall, the microstructure of the developed spongy-like hydrogels was proper for the diffusion of oxygen and nutrients, since the cells showed their normal phenotype after 21 days of culture. Furthermore, it was also observed that the pore size did not negatively affect cell spreading and morphology, as demonstrated by cytoskeleton staining.

When considering the bioactivity potential of HAp and/or Lf, the release of these bioactive components to the surrounding environment is of high importance in an in vivo scenario. Hence, the release of these bioactive factors was studied in vitro along 31 days. The serum level of Lf on the human body is around 2 to 7 µg/mL, and it has been applied to achieve the desired effect from the contact of Lf with different cell types (e.g. proliferation and/or the inhibition of bone resorption) [17]. Having this in mind, we applied a protein concentration within our biomaterials that, when released, was similar to the protein concentration that is found in human body. Lf containing spongy-like hydrogels showed a significant mass loss, in contrast to the HAp containing spongy-like hydrogels. This can be explained by the higher release of non-crosslinked Lf from the semi-interpenetrating

networks, which causes the disintegration of the polymeric networks. Interestingly, when they were combined, the HAp presence masked the Lf presence in all aspects that were previously mentioned. Perhaps it happens due to the high affinity between HAp and Lf, which is mainly provided by the electrostatic interactions between Lf and HAp surface [3,41]. Montesi et al., also reported that the protein quantity is a crucial parameter, since a lower amount of Lf only interacted with the negative surface ionic groups of HAp, while higher amount of Lf probably favored the protein–protein interactions (through hydrogen or hydrophobic/hydrophilic interactions) and formation of multiple protein layers with different molecular orientations [41]. Furthermore, these interactions were sustained by Lf incorporation during hydrogel preparation, which affected the hydrogel chains formation. It was also reported that the swelling ratio was correlated with the quantity of protein released [42], since a faster desorption of the Lf molecules being non-directly bounded on the particles surface is verified, when the scaffolds are immersed in a hydrated medium. However, in spite of non-existent significant differences, the formulations with 10% of HAp and Lf in both concentrations, showed lower swelling ratio and, consequently, a lower release. As expected, when Lf was present in a high amount, it had more protein to release. Montesi et al. demonstrated that, in addition to the HAp biocompatibility, osteoconductivity, and biodegradability, this bioceramic has the ability to bind several biomolecules without affecting their biological function [41]. Overall, some gaps have been identified based on the existing studies. In fact, most of the existing studies are performed while using animal cells/models hindering the correlation of these models with future clinical outcomes. Moreover, the results that were obtained can be conditioned, depending of the type and size of the defect, the degradation process, and the protein release, which resulted in different bone regeneration rates.

Our study demonstrates the synergetic effect from the combination of HAp with Lf and hASCs and their ability for use in spongy-like hydrogels for bone tissue engineering applications.

#### **5. Conclusions**

Bone tissue engineering has been explored by combining new biomaterials and cells to reach a synergistic effect. With this in mind, we studied the effect from the combination of bioactive compounds, Lf and HAp, within GG spongy-like hydrogels to be used in the bone tissue regeneration approaches.

The parameters studied, such as porosity, pore size, pore wall thickness, weight loss, water uptake, and Lf release, were tailored through the addition and/or combination of each compound. Lf containing spongy-like hydrogels showed similar features with GG spongy-like hydrogels. In contrast, differences in terms of all parameters studied were observed in GG/HAp and Lf-HAp containing spongy-like hydrogels. The HAp addition promoted an increase of the wall thickness and a decrease on porosity and pore size as well as, on degradability and water retention capabilities. Related to the Lf release, along the 30 days all formulations released at least half of the existent lactoferrin in both (0.05 and 0.15%) of the formulations containing Lf. Moreover, the capacity of these biomaterials to support hASCs in culture for 21 days was assessed. The microstructure of Lf/HAp-containing spongy-like hydrogels seemed to be proper for the diffusion of oxygen and nutrients, since the cells spread and showed their normal phenotype after 21 days. In conclusion, Lf and HAp in high concentrations assembled the required conditions to conduce these biomaterials to promote bone regeneration.

**Author Contributions:** All authors above mentioned have contributed to the preparation of the proposed manuscript, attested the authenticity of the data and agreed to its submission to the previous referred interest journal. For this work, no other persons have in any way contributed.

**Funding:** This research was funded by FROnTHERA (NORTE-01-0145-FEDER-0000232).

**Acknowledgments:** F.R.M. acknowledges FCT (SFRH/BPD/117492/2016), L.P.d.S. thanks FCT (SFRH/BD/78025/2011), J.M.O. thanks FCT Investigator program (IF/01285/2015) and A.R.B. thanks the funds provided by FCT under the doctoral program in Tissue Engineering, Regenerative Medicine and Stem Cells (PD/BD/143043/2018).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **In Vitro and In Vivo Evaluation of Starfish Bone-Derived** β**-Tricalcium Phosphate as a Bone Substitute Material**

**Haruka Ishida 1,2 , Hisao Haniu 1,2,3, \* , Akari Takeuchi 2,3 , Katsuya Ueda 1,2 , Mahoko Sano 1,3 , Manabu Tanaka 4 , Takashi Takizawa 4 , Atsushi Sobajima 4 , Takayuki Kamanaka 4 and Naoto Saito 1,2,3**


Received: 18 April 2019; Accepted: 10 June 2019; Published: 11 June 2019

**Abstract:** We evaluated starfish-derived β-tricalcium phosphate (Sf-TCP) obtained by phosphatization of starfish-bone-derived porous calcium carbonate as a potential bone substitute material. The Sf-TCP had a communicating pore structure with a pore size of approximately 10 µm. Although the porosity of Sf-TCP was similar to that of Cerasorb M (CM)—a commercially available β-TCP bone filler—the specific surface area was roughly three times larger than that of CM. Observation by scanning electron microscopy showed that pores communicated to the inside of the Sf-TCP. Cell growth tests showed that Sf-TCP improved cell proliferation compared with CM. Cells grown on Sf-TCP showed stretched filopodia and adhered; cells migrated both to the surface and into pores. In vivo, vigorous tissue invasion into pores was observed in Sf-TCP, and more fibrous tissue was observed for Sf-TCP than CM. Moreover, capillary formation into pores was observed for Sf-TCP. Thus, Sf-TCP showed excellent biocompatibility in vitro and more vigorous bone formation in vivo, indicating the possible applications of this material as a bone substitute. In addition, our findings suggested that mimicking the microstructure derived from whole organisms may facilitate the development of superior artificial bone.

**Keywords:** starfish; calcium carbonate; porous calcium phosphate; β-tricalcium phosphate; bone substitute; angiogenesis

#### **1. Introduction**

Bone is a tissue with excellent regenerative ability; however, supplementation is necessary for reconstruction of bone defects that cannot be naturally healed due to fracture or tumor resection. Autologous bone grafting can be used to compensate for defects. In this method, grafts are harvested from a healthy part of the patient and transplanted into the defective part of the bone. Because the graft itself has bone-forming ability, the process after transplantation is usually effective. However, it may not be possible to obtain bone of the necessary amount, shape, and size, and there are other

disadvantages, such as pain and deformation at the collection site. Additionally, problems such as infection and immune response arise when allogeneic bone grafting and heterogeneous bone grafting are used [1].

In order to overcome these problems, researchers have been attempting to develop artificial bone prosthetic materials showing good biocompatibility [2]. For example, a glass material that bonds to bones without showing foreign body reactions was reported by Hench et al. in 1971 [3]. Since then, the development of artificial bone has progressed due to the discovery of physiologically active materials, even inorganic materials, and the synthesis of hydroxyapatite (HAp) as a main component of bone inorganic matter and direct bonding to bone were described by Jarcho et al. [4] and Akao et al. [5]. In addition to HAp, β-tricalcium phosphate (β-TCP) has been extensively studied as a calcium phosphate-based biomaterial. β-TCP has higher solubility at neutral pH than HAp and is not only hydrolyzed in body fluids but is also biologically absorbed by osteoclasts, representing an effective bioabsorbable material that can be used as a bone substitute. Studies on β-TCP have been reported by Driskell et al. [6], and SynthoGraft, which was first developed in 1981, has been approved by the US Food and Drug Administration as an absorbable synthetic bone grafting material. Such bioceramics account for approximately 40% of total bone grafts [7].

Regardless of material, porous bodies with continuous pores are required for supporting osteoinductive factors and for binding, after implantation, with strong bone, cellular tissues, and blood vessels for invasion [8]. It is necessary to control pore size distribution and pore structure in order to achieve both secured pore structures that promote bone formation and strength when used as a filler material. Porous bodies have been produced by various methods [9,10]. However, no artificial bone with performance comparable to bone has been developed.

Roy and Linnehan have reported methods for converting coral-derived calcium carbonate into HAp as an artificial bone using biological materials [11]. Coral-derived HAp is a porous HAp with a pore size of 150 to 500 µm and has high biocompatibility [12]. This material has been commercialized as ProOsteon in the United States of America. However, coral harvesting is costly and associated with environmental problems.

The inorganic matter forming on starfish is composed of calcite granules containing Mg2<sup>+</sup> and has a fine porous structure with a pore diameter of several tens of microns. Takeuchi et al. reported that this porous structure is converted into β-TCP containing Mg2<sup>+</sup> by phosphatization of starfish [13].

Therefore, in this study, we evaluated the physical properties of starfish bone-derived β-TCP as a potential bone substitute material using MC3T3-E1 mouse calvaria-derived osteoblast-like cells and in vivo implantation into rat calvaria bone defects.

#### **2. Materials and Methods**

#### *2.1. Ethics Statement*

All experiments were carried out according to institutional guidelines for animal experimentation at Shinshu University School of Medicine. All protocols used in this study were reviewed and approved by the Division of Laboratory Animal Research (#280054). All surgery was performed under general anesthesia (intraperitoneal injection of sodium pentobarbital), and all efforts were made to minimize suffering. All rats were euthanized using isoflurane inhalation at the end of the study.

#### *2.2. Preparation of Porous Calcium Phosphate Derived from Starfish Bone*

Collection of starfish-bone-derived calcite and its phosphatization to β-TCP were carried out as previously described [13]. Organic substances were dissolved by immersing starfish (*Patiria pectinifera*) in a commercial bleach (sodium hypochlorite aqueous solution with a volume ratio of about 6%; Hiter®; Kao Corp., Tokyo, Japan). After dissolving organic matter, starfish bone was obtained by filtering, washing with ion-exchanged water, and drying at 60 ◦C for 24 h. Briefly, 1.5 g of starfish bone granules were phosphatized in a reaction vessel for hydrothermal treatment (inner volume 25 mL) with 20 mL of

a 0.5 mol/L diammonium hydrogen phosphate aqueous solution [(NH4)2HPO4, FUJIFILM Wako Pure Chemical Corp., Osaka, Japan], at 200 ◦C for 72 h. Phosphatization treatment was performed under more severe conditions than autoclave sterilization, so samples were considered sterile and treated accordingly. The product was then washed with ion-exchanged water and dried at 60 ◦C for 24 h.

The obtained phosphate-treated starfish-bone-derived β-TCP (Sf-TCP) was sieved at 150–500 µm to obtain materials with the same size as the β-TCP used in the clinical setting (Cerasorb M [CM]; curasan AG, Kleinostheim, Germany).

#### *2.3. Observation of Surface Structures by Scanning Electron Microscopy (SEM)*

The Sf-TCP was affixed on a brass sample holder using conductive carbon paste (DOTITE XC-12, JEOL Ltd., Tokyo, Japan), coated with osmium oxide (FUJIFILM Wako Pure Chemical Corp.) by an osmium coater (Neoc-AN; Meiwafosis, Tokyo, Japan) at 10 mA for 20 s, and then observed under a field emission-scanning electron microscope (FE-SEM, JSM-7600F; JEOL Ltd., Tokyo, Japan) at an accelerating voltage of 2.00 kV.

#### *2.4. Measurement of Specific Surface Area*

The specific surface area of Sf-TCP was measured in triplicate with a high-precision multisample gas adsorption amount measuring device (Autosorb®-iQ; Quantachrome Instruments, Kanagawa, Japan) using the nitrogen adsorption method. Before measuring, vacuum degassing was performed at 120 ◦C for 3 h. Specific surface area was calculated from the range of relative pressures of adsorption-desorption isotherms by 0.05 to 0.35 according to the BET theory formula, as follows:

Monomolecular adsorption:

$$\upsilon\_{\rm m} = \upsilon \left( 1 - \frac{\rm P}{\rm P\_0} \right) \tag{1}$$

ν : Adsorption amount p p0 : Relative pressure Surface area:

$$\mathbf{A}\_{\rm s} = \left(\frac{\mathbf{v}\_{\rm m} \mathbf{N} \mathbf{a}\_{\rm m}}{\mathbf{m}}\right) \times 10^{-18} \tag{2}$$

N : Avogadro constant a<sup>m</sup> : Molecular occupied cross section m : Molecular weight of adsorbate

#### *2.5. Measurement of Porosity and the Most Frequent Pore Diameter*

We then used MicrotracBEL (Osaka, Japan) to measure porosity and the most frequent pore diameter of Sf-TCP by the mercury intrusion method using an automatic mercury porosimeter (Pascal Model 140; Thermo Scientific, Tokyo, Japan). Density was measured by the helium gas replacement method using a true density measuring apparatus (BELPycno, MicrotracBEL, Osaka, Japan). Before measuring, vacuum degassing was performed at room temperature for 15 min. Each item was measured once. The measured samples were used for subsequent experiments.

#### *2.6. Cell Growth Tests*

Mouse calvaria-derived osteoblast-like cells, MC3T3-E1 (RIKEN Cell Bank, Tsukuba, Japan) were used as cultured cells. MC3T3-E1 cells were cultured in an incubator at 37 ◦C in an atmosphere containing 5% CO2. αMEM medium supplemented with 10% fetal bovine serum and 1% antibiotic-antimycotic mixed solution was used for cell culture. For passaging of cells, cells were seeded into 10 cm dishes at a density of 4.0 <sup>×</sup> <sup>10</sup><sup>4</sup> cells/mL and subcultured twice a week.

For cell proliferation tests, Alamar Blue assay (Alamar Blue Cell Viability Reagent, Remel, Lenexa, KS, USA) was used. MC3T3-E1 cells were seeded into 96-well plates at 6.0 <sup>×</sup> <sup>10</sup><sup>3</sup> cells/well. After culturing for 24 h, medium was exchanged and used for experiments.

Sf-TCP and CM were added at 10 mg/well. The control wells contained cells only. As a sample blank, evaluation targets were added to medium and wells without cells were used. Those groups consisted of eight wells, each with two sample blanks.

After addition of evaluation targets, the cells were cultured for 24 h. For the Alamar Blue assay, Alamar Blue Cell Viability Reagent was added to 10% of the total volume and reacted for 1 h. The Alamar Blue assay was performed by reading the fluorescence intensity at an excitation wavelength of 535 nm and an emission wavelength of 590 nm using a Plate Reader (AF2200, Eppendorf, Hamburg, Germany).

#### *2.7. Cell Observation on the Sf-TCP Surface*

The Sf-TCP cultured with MC3T3-E1 cells for 24 h was removed with tweezers, fixed by freeze-drying in 2.5% glutaraldehyde (used after dilution of 70% glutaraldehyde, TAAB Laboratories Equipment, Berks, UK), 1% osmium solution (aqueous solution of osmic acid, Nisshin EM, Tokyo, Japan), and t-butyl alcohol (FUJIFILM Wako Pure Chemical Corp.) and observed under an FE-SEM at an accelerating voltage of 15.0 kV.

#### *2.8. Implantation in a Rat Calvaria Defect Model*

Experimental animals were male Wistar rats (8 weeks old, weighing 150–200 g, SLC, Hamamatsu, Japan). According to animal experiment guidelines, rats were housed at five rats per cage in a breeding room with controlled room temperature (25 ◦C ± 2 ◦C) and humidity (50% ± 10%). Food and water were available ad libitum.

In accordance with the methods described by Tanaka et al. [14], a rat calvaria defect model was established. Briefly, after induction of anesthesia by isoflurane (Forane, ABBOTT JAPAN, Tokyo, Japan) inhalation, pentobarbital (Somnopentyl, Kyoritsu Seiyaku, Tokyo, Japan) was subcutaneously injected at 40 mg/kg, and the operation was performed. Bone defects of 5 mm in diameter were made in the rat calvaria using a trephine bar. No implant was placed in the sham (i.e., control) group, whereas the experimental groups were implanted with 10 mg Sf-TCP or CM. Rats were then housed for 4 or 8 weeks. There were eight rats in each group. Rats that died during the 4–8 week period were not used for evaluation.

#### *2.9. Histological Examination*

After euthanasia under anesthesia, the heads of rats were dissected and fixed for 1 week with 10% formalin (FUJIFILM Wako Pure Chemical Corp.). Rat skulls were then decalcified for 3 days inquick dehydrating liquid (K-CX, Pharma, Tokyo, Japan). Samples after demineralization were processed to an appropriate size and embedded in paraffin. After embedding in paraffin, samples were sliced into sections (4 µm thick) using a microtome. Sections were then subjected to hematoxylin (Muto Pure Chemicals, Tokyo, Japan) and eosin (FUJIFILM Wako Pure Chemical Corp.) staining (HE staining) and Masson's Trichrome (MT) staining (Muto Pure Chemicals) and observed with an optical microscope (BX50, Olympus, Tokyo, Japan).

MT-stained specimens were observed with a multispectral automatic tissue section quantitative analysis system (Vectra3, PerkinElmer, Waltham, MA, USA) and photographed as multispectral images. Upon quantification, dye used for MT staining was incorporated and analyzed with inForm (PerkinElmer). For each specimen, we selected and photographed the inForm analysis part manually such that the area occupied by CM and Sf-TCP was maximized within the photograph range of 250 µm × 334 µm defined by Vectra3. The threshold for quantitative analysis was set automatically. Analysis results quantified by pixel number were averaged for each group and calculated for the area of fibrous tissue and cells in the bone defect area.

#### *2.10. Statistical Analysis*

For statistical analysis, Student's t-tests were used, and for multiple comparisons, Bonferroni corrections were performed. Statistical results were expressed as means ± standard errors (SEs). The significance level was set at *p* < 0.05.

#### **3. Results**

#### *3.1. Size of CM and Sf-TCP*

Figure 1 shows photographs of CM, as shown in Figure 1a, and Sf-TCP, as shown in Figure 1b, after sieving according to the catalog size of CM. The catalog CM size was 150–500 µm, but some particles of less than 150 µm were present.

**Figure 1.** Photographs of the analyzed materials. (**a**) Cerasorb M (CM). (**b**) starfish-derived β-tricalcium phosphate (Sf-TCP) after sieving.

#### *3.2. Surface Structure*

Figure 2 shows ×1000 SEM images of CM, as shown in Figure 2a, and Sf-TCP, as shown in Figure 2b. In CM, pores of various shapes were observed on the surface, as shown by the arrows in Figure 2a. In Sf-TCP, pores were found all over the surface, and the structures were connected with other pores inside.

**Figure 2.** Surface images obtained by scanning electron microscopy. Magnification: ×1000. (**a**) CM. Arrows indicate pores. (**b**) Sf-TCP. Scale bar: 10 µm.

#### *3.3. Density, Most Frequent Pore Diameter, and Porosity*

Table 1 shows measurement results of density and porosity. The density of the material was approximately 3 g/cm<sup>3</sup> . The total pore volume and total pore surface area for Sf-TCP were 445.42 mm<sup>3</sup> /g and 0.130 m<sup>2</sup> /g, respectively, whereas those for CM were 415.31 mm<sup>3</sup> /g and 0.062 m<sup>2</sup> /g, respectively. The Sf-TCP pore surface area was approximately twice that of CM. CM had a median pore diameter of about 57 µm and a most frequent pore diameter of about 100 µm, representing a 2-fold difference. For Sf-TCP, the median diameter and most frequent pore diameter were about 10–12 µm, which showed almost no difference. Additionally, pores that excluded mercury were observed in Sf-TCP in porosity measurements by mercury intrusion.


**Table 1.** Density, most frequent pore diameter, and porosity.

#### *3.4. Specific Surface Area*

Figure 3 shows the adsorption-desorption isotherm resulting from measurement of CM, as shown in Figure 3a, and Sf-TCP, as shown in Figure 3b, using the nitrogen adsorption method.

The calculated specific surface area of CM was 3.693 m<sup>2</sup> /g, and that of Sf-TCP was 9.676 m<sup>2</sup> /g; thus, Sf-TCP was about three times larger than CM.

**Figure 3.** Adsorption-desorption isotherms. The blue line indicates adsorption, and the red line indicates desorption. (**a**) CM; (**b**) Sf-TCP.

From the shape and hysteresis of the materials with a relative pressure range of adsorption-desorption isotherms between 0.45 and 1.00, the pore structure of the surface could be identified by IUPAC classifications [15]. CM was found to be a type II material, having macropores with a diameter of 50 nm or more or no pores, whereas Sf-TCP was a type V material having mesopores with a diameter of 2–50 nm.

### *3.5. Cell Growth Test*

Figure 4 shows the results of the Alamar Blue assay as a ratio of the control when 10 mg Sf-TCP or CM was added. The positive control CM showed no significant differences compared with the control group. In contrast, Sf-TCP-treated cells showed more than a 1.5-fold increase compared with the control and CM groups.

**Figure 4.** Effects of CM and Sf-TCP on cell proliferation. Values are the means ± standard errors (SEs) (*n* = 6). \*\*\**P* < 0.001.

#### *3.6. Observation of Cells on the Sf-TCP Surface*

Figure 5 shows ×3000 SEM images of Sf-TCP after culture with MC3T3-E1 cells. A scaly microstructure was observed on the Sf-TCP surface. MC3T3-E1 cells adhered not only to the Sf-TCP skeleton but also to pores with stretched filopodia.

**Figure 5.** Surface SEM image of Sf-TCP cultured with MC3T3-E1 cells. Magnification: ×3000. The arrows show the scaly structure on the surface, and the triangles show cell filopodia.

### *3.7. Histological Examination of HE-Stained Specimens*

Figure 6 shows photographs of HE-stained specimens. In the sham group, there was a thin tissue layer covering the bone defect. In contrast, in the CM group, the amount of tissue present in CM was nearly unchanged from 4 to 8 weeks. In the Sf-TCP group, tissue was already present in the porous structure of Sf-TCP at 4 weeks.

**Figure 6.** Hematoxylin and eosin (HE) staining. Hematoxylin stains cell nuclei blue, while eosin stains the cytoplasm, connective tissue, and other extracellular substances pink or red. Brain tissues are shown in the lower portion of the images, and the outer side of the skull is shown in the upper portion of the images. Boxed parts are prosthetic materials. Scale bar: 200 µm.

Notably, vascular structures were observed at 8 weeks after implantation in the Sf-TCP group, as shown in Figure 7. Capillary structures were observed only in the Sf-TCP group.

**Figure 7.** Blood vessel structures (red boxed area) were identified by hematoxylin and eosin (HE) staining.

#### *3.8. Observation and Quantification of MT-Stained Specimens*

Figure 8 shows a multispectral image of MT-stained specimens in the CM and Sf-TCP groups. Aniline blue stains fibrous tissue blue, Masson liquid stains cells red, and Orange G stains blood cells orange. For quantitative analysis with inForm, CM and Sf-TCP existed in the bone defects.

**Figure 8.** *Cont*.

**Figure 8.** Masson's trichrome (MT)-stained images taken with Vectra3. Aniline blue stains fibrous tissue blue, Masson liquid stains cells red, and Orange G stains blood cells orange. (**a**) CM implant at 4 weeks. (**b**) Sf-TCP implant at 4 weeks. The blue boxes indicate the area quantified in Figure 10.

Figure 9 shows an enlarged image of the region selected in Figure 8 and an image obtained by analyzing the corresponding portion with inForm. Cells and fibrous substances existed mainly in the periphery of CM in the CM group at 4 and 8 weeks, whereas in the Sf-TCP group, cells and fibers entered into the pores of Sf-TCP.

**Figure 9.** *Cont*.

**Figure 9.** Images of Masson's trichrome (MT)-stained sections and analysis images for quantification. Representative images showing cells and fibrous tissue existing mainly around CM or Sf-TCP pores. Cells appear more uniformly scattered at 8 weeks than at 4 weeks. (**a**,**b**) CM images with MT staining. (**c**,**d**) CM images for analysis. (**e**,**f**) Sf-TCP images with MT staining. (**g**,**h**) Sf-TCP images for analysis. Aniline blue staining (blue) shows fibrous tissue, and Masson liquid and Orange G staining (red) show cells. Scale bar: 50 µm.

Figure 10 shows the results of quantifying areas of fibrous tissues and cells. Fiber and cell areas in the CM and Sf-TCP groups were compared at 4 and 8 weeks. The results showed that the fiber mass was significantly larger in the Sf-TCP group than in the CM group at both 4 and 8 weeks. However, actual cell amount did not change significantly.

**Figure 10.** Quantitative analysis using inForm. The numbers of samples used for analysis were five in the Sf-TCP group at 4 weeks and six in other groups. Mean ± SEs. \* *P* < 0.05.

Figure 11 shows images of erythrocytes stained with Orange G inside the implant at the time of observation, and Table 2 shows the number of individuals exhibiting erythrocytes in the prosthetic material. These results were observed only in the Sf-TCP group (one of five animals at 4 weeks and three of six animals at 8 weeks).

**Figure 11.** Image of erythrocytes observed inside Sf-TCP. (**a**) 4 weeks. (**b**) 8 weeks. Red circles indicate erythrocytes observed in Sf-TCP. Scale bar: 20 µm.

**Table 2.** Number of individuals having erythrocytes in prosthetic material (number/total).


#### **4. Discussion**

In this study, we evaluated the physical properties of Sf-TCP and performed in vitro and in vivo experiments to examine the potential applications of Sf-TCP as a bone substitute material. Our results showed that St-TCP did not induce cytotoxicity, which is important for the development of biomaterials, and facilitated the healing of wounds in vivo, supporting the potential clinical applications of this material.

Our results showed that Sf-TCP induced significant improvement in cell proliferation compared with the control or CM treatments. These findings were thought to be related to the excellent cell adhesion and proliferation properties of Sf-TCP. Indeed, the specific surface area of Sf-TCP was about three times larger than that of CM, consistent with studies demonstrating that cell proliferation increases as the specific surface area increases [16,17]. Because cell proliferation is a main factor indicating biocompatibility in MC3T3-E1 cells [18], our results supported that Sf-TCP was highly biocompatible. Additionally, in morphological observations, MC3T3-E1 cells adhering to Sf-TCP showed well-stretched filopodia, which has been reported to occur during cell migration [19], and the structure of Sf-TCP indicated that the cells could migrate.

In our in vivo experiments, cells and fibrous tissues invaded into the communicating porous structures of Sf-TCP in MT-stained specimens, indicating that cellular tissues could penetrate Sf-TCP, even when the pores were 10–12 µm in size. In addition, increased fibrous tissue area was observed in the Sf-TCP group compared with that in the CM group. The amount of fibrous material has been reported to be related to the abundance of collagen fibers and calcification [20,21]. Thus, the large amount of fibers present in Sf-TCP may indicate progression of bone formation. In porous artificial bone prosthetic materials, new bone invades into the pores and binds to surrounding bone [11]; accordingly, our findings supported the performance of Sf-TCP as a bone substitute material.

In the Sf-TCP group, vascular structures were observed in HE-stained specimens, and erythrocytes were present inside Sf-TCP in MT-stained specimens, suggesting the formation of capillary vessels. The pore size of Sf-TCP (10–12 µm) was smaller than that of previously reported biomaterials (100 µm), which allows invasion by cells and tissues, or that (300 µm) required for angiogenesis, e.g., invasion

of capillaries [22]. However, the pore sizes of HAp and β-TCP are typically within the range of 100 to 600 µm (or 150 to 200 µm when coral is used as a biological material) [23–26]. Therefore, it is possible that few studies have evaluated such small pores or that angiogenesis was not observed in CMs, which have pores of 50–100 µm because of differences between artificial and biologically-derived porous structures. Various factors, such as disconnection of communicability by closed pores, variations in pore size, and sharp connecting structures with respect to the size of cells in artificial communication pores, are thought to hinder invasion of cellular tissues. However, in Sf-TCP, there were few of these structures and few elements that inhibited the invasion of cellular tissues; thus, vigorous tissue invasion and formation of capillary vessels would have occurred. In the rat calvaria defect model, capillary vessels are approximately 4–10 µm in diameter in the cerebrum around the defect, and capillaries can invade pores with a diameter of 10–12 µm. This result is important in considering the three-dimensional structure of the potential prosthetic bone filling material. Moreover, artificial preparation of starfish bone-like structures adapted to the capillary diameters of humans may yield artificial bone prosthetic materials capable of invading capillaries and tissues while improving the strength of the material by decreasing the pore size. Additionally, pore sizes increase when β-TCP is dissolved, absorbed, and replaced with autologous bone, and tissues larger than the pore may therefore invade the material.

#### **5. Conclusions**

In this study, we evaluated whether Sf-TCP derived from starfish could be used as a bone substitute material. Compared with CM, which is already used clinically, Sf-TCP was found to have comparable porosity but a larger specific surface area. Cell proliferation tests showed that Sf-TCP promoted cell proliferation. Moreover, in animal experiments, Sf-TCP showed high regeneration ability, vigorous invasion of cellular tissue into the pore structure, and introduction of capillary vessels into the prosthetic material, even when the pore size was only approximately 10 µm. However, this study did not examine the pore size required for angiogenesis, so there may be a more appropriate pore size. In the future, we hope to optimize regeneration accompanied by angiogenesis by artificially creating materials with structures similar to that of Sf-TCP but with different pore diameters.

From the above results, we concluded that Sf-TCP obtained by phosphatization of starfish bone may be effective for bone regeneration applications, such as in the treatment of fractures and bone loss. In addition, mimicking the structure of the starfish bone may lead to the development of superior artificial bone substitute materials.

**Author Contributions:** Conceptualization, H.H. and A.T.; Methodology, H.I., H.H., A.T., K.U., M.S. and M.T.; Formal Analysis, H.I.; Investigation, H.I.; Resources, H.H., A.T. and N.S.; Data Curation, H.I., K.U., M.S., M.T., T.T., A.S. and T.K.; Writing-Original Draft Preparation, H.I.; Writing-Review & Editing, H.H., A.T. and M.S.; Visualization, H.I.; Supervision, H.H., A.T. and N.S.; Project Administration, H.H.

**Funding:** This research received no external funding.

**Acknowledgments:** We thank MicrotracBEL Corp. for measurement of the density, most frequent pore diameter, and porosity of Sf-TCP. We thank Ms. Kayo Suzuki (Research Center for Human and Environmental Sciences, Shinshu University) for technical assistance in SEM analysis. We thank Takeuchi laboratory members for their assistance in the preparation of Sf-TCP. We would like to thank Editage (www.editage.jp) for English language editing.

**Conflicts of Interest:** The authors state that there are no conflicts of interest with regard to the publication of this paper.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Mechanobiological Approach to Design and Optimize Bone Tissue Sca**ff**olds 3D Printed with Fused Deposition Modeling: A Feasibility Study**

**Gianluca Percoco, Antonio Emmanuele Uva, Michele Fiorentino , Michele Gattullo , Vito Modesto Manghisi and Antonio Boccaccio \***

Dipartimento di Meccanica, Matematica e Management, Politecnico di Bari, Via E. Orabona 4, 70126 Bari, Italy; gianluca.percoco@poliba.it (G.P.); antonio.uva@poliba.it (A.E.U.); michele.fiorentino@poliba.it (M.F.);

michele.gattullo@poliba.it (M.G.); vitomodesto.manghisi@poliba.it (V.M.M.) **\*** Correspondence: antonio.boccaccio@poliba.it; Tel.: +39-080-5963393

Received: 21 December 2019; Accepted: 28 January 2020; Published: 1 February 2020

**Abstract:** In spite of the rather large use of the fused deposition modeling (FDM) technique for the fabrication of scaffolds, no studies are reported in the literature that optimize the geometry of such scaffold types based on mechanobiological criteria. We implemented a mechanobiology-based optimization algorithm to determine the optimal distance between the strands in cylindrical scaffolds subjected to compression. The optimized scaffolds were then 3D printed with the FDM technique and successively measured. We found that the difference between the optimized distances and the average measured ones never exceeded 8.27% of the optimized distance. However, we found that large fabrication errors are made on the filament diameter when the filament diameter to be realized differs significantly with respect to the diameter of the nozzle utilized for the extrusion. This feasibility study demonstrated that the FDM technique is suitable to build accurate scaffold samples only in the cases where the strand diameter is close to the nozzle diameter. Conversely, when a large difference exists, large fabrication errors can be committed on the diameter of the filaments. In general, the scaffolds realized with the FDM technique were predicted to stimulate the formation of amounts of bone smaller than those that can be obtained with other regular beam-based scaffolds.

**Keywords:** tissue engineering; biomaterials; mechanobiology; scaffold design; geometry optimization

#### **1. Introduction**

The development of the recent additive manufacturing techniques and, consequently, the possibility of building constructs with very sophisticated geometries, led many researchers to investigate the scaffold geometries that mostly favor the formation of bone in the shortest time. To this purpose, both regular and irregular scaffold geometries were proposed and investigated. The regular scaffolds include unit cell configurations all with the same shape and dimensions that are regularly replicated in the scaffold volume [1,2]. Analytical solutions were recently developed that put in relationship the equivalent material properties of the entire scaffold with the dimensions and the material properties of the single unit cell [3–6]. The irregular ones include pores differently shaped and dimensioned and present a geometry that can be described with statistical parameters [7–9] but not in a precise form [10]. Although irregular geometry scaffolds are commonly utilized in experiments for bone tissue engineering, regular scaffolds have gained importance in recent decades as they allow precise control of the actual geometry and are, hence, suited to create repeatable physical environments and easier to investigate [5]. Among the possible geometries of regular scaffolds, the geometry realized by means of the fused deposition modeling technique certainly represents an important solution.

The fabrication of scaffolds with additive manufacturing techniques is an issue that recently received recognition and attention from the research community [11,12]. Different rapid prototyping techniques were proposed to fabricate biomaterials scaffolds, based on liquid polymerization [13], material deposition processes [14], powder-based processes [15], sheet lamination [16], binder jetting [17], and material jetting [18]. Among the other rapid prototyping techniques, fused deposition modeling (FDM) is one of the most common 3D printing technologies available on the market, thanks to its low cost, ease of use, and the variety of usable materials. It consists in depositing layers of a polymeric, ceramic, or metallic material; each layer includes cylindrical strands all oriented in a given direction and equally spaced. To guarantee an adequate structural response, the orientation of filaments changes layer by layer. Furthermore, between the cylindrical filaments of adjacent layers, an overlap region exists where the single strand is melted with the adjacent one during the deposition process. Recently, this technique has been successfully utilized to fabricate scaffolds for bone tissue engineering [19–23]. Process parameters of the FDM technique were also optimized to improve the dimensional accuracy of the manufactured components [24]. The typical/traditional approach consists in using the FDM technique by implementing standard process parameters that allow obtaining accurate structures. However, adjusting the strand diameter, and modifying the process parameters is a challenging task and still remains a research topic. In fact, when small modifications on the scaffold geometry must be achieved, acting on process parameters of the traditional FDM technique can be a valid option instead of looking for expensive, more accurate technologies. A recent study proposed a numerical model to simulate the extrusion of a strand of semi-molten material to investigate how the strand cross-section changes for variable process parameters [25]. The porosity and the micro-architecture of parts built with the FDM technique appear very suited to stimulate the colonization and subsequent differentiation of mesenchymal stem cells [26].

Optimization algorithms were implemented to improve the scaffold performance and to minimize the negative effects related to the implantation of the scaffold in the fracture site [27]. Many objective functions were investigated and different optimization strategies were implemented [28–31]. Most of the optimization studies reported in the literature aim to minimize the difference between the equivalent mechanical properties of the scaffold and those of the tissue within which it is implanted. In one word, such studies that consider scaffold stiffness as the design variable aim to minimize the effects of stress shielding at the bone/scaffold interface [32,33]. Optimization strategies based on the compressive modulus expressed as a ratio between third principal stress and the prescribed compressive strain were also utilized to optimize the geometry of scaffolds fabricated with the FDM technique [34]. Singh et al. have proposed a multifactor optimization for the development of biocompatible and biodegradable composite material-based feedstock filament of fused deposition modeling [35]. Only recently, optimization algorithms based on mechanobiological criteria were proposed to design and optimize small volumes of scaffolds with both, regular [36–40] and irregular [41] micro-geometry [42]. In these mechanobiology-based optimization algorithms, the scaffold geometry is perturbed until the micro-architecture that allows maximizing the formation of new bone, is identified. However, to the knowledge of the present authors, no studies are reported in the literature that optimize the geometry of scaffolds fabricated with the FDM technique and that are based on mechanobiological criteria. In this study, we want to bridge this gap.

The objective of this feasibility study is to optimize the geometry of scaffolds fabricated with the FDM technique and to investigate whether this technique can be utilized to fabricate scaffolds designed and optimized to undergo a compression load. The objective function that was utilized is based on the computational mechano-regulation algorithm by Prendergast and Huiskes [43,44]. Such an algorithm models the fracture domain as a biphasic poroelastic material, and hypothesizes the biophysical stimulus that triggers the tissue differentiation process to be a function of the octahedral shear strain and of the fluid velocity. This mechanobiological algorithm was successfully utilized in previous studies investigating the healing process in fractured bones [45,46], in osteochondral

defects [47], at the implant/bone interfaces [48] and the regeneration process in scaffolds for bone tissue engineering [49,50].

Other mechano-regulation computational models are reported in the literature, investigating the role of the mechanical environment on the biophysical stimulus that triggers the tissue differentiation process [51–54]. However, the patterns of tissue differentiation predicted by the model of Prendergast and Huiskes were shown to be closest to experimental results compared to other mechanobiological algorithms [55].

We determined, for different values of the filament diameter, the optimal distance between the strands that, for the specific load acting on the scaffold, can maximize the formation of bony tissue. The optimized scaffolds were physically fabricated and successively measured. By utilizing a CMM machine, the distance between the filaments and the filament diameter were measured with high accuracy and compared with the corresponding nominal values.

#### **2. Materials and Methods**

#### *2.1. Parametric Finite Element Model*

The parametric poroelastic finite element model of a cylindrical scaffold with radius *R* = 20 mm and *h* = 5 mm high was built in Abaqus® (version 6.12, Dassault Systèmes, Vélizy-Villacoublay, France) (Figure 1). Scaffolds with the same dimensions were utilized by Teng et al. [56]. The model consists of layers including aligned cylindrical strands with diameter *D* equally spaced. The filaments of two adjacent layers form an angle of 90◦ (Figure 1a). Five different values were considered for the diameter *D*: 400, 500, 600, 700, and 800 µm while the distance between the filaments *dfil* was optimized via the optimization algorithm described below (Figure 2). Following Somireddy and Czekanski [57], between two adjacent layers, an overlap region of 0.1 × *D* was hypothesized. The model of the tissue occupying the scaffold pores (highlighted in red in Figure 1) was built by means of Boolean operations of subtraction, from the entire model volume *VTOT* = π × *R* <sup>2</sup> <sup>×</sup> *<sup>h</sup>*, the volume of the scaffold.

According to previous studies [38,58], the volume inside the pores was hypothesized to be occupied by granulation tissue. Exploiting the symmetry of the system, to reduce the computational cost, a one-quarter model was developed. The lower base of the scaffold was clamped while a compression load was applied on the upper surface by means of a rigid plate (highlighted in blue in Figure 1). Different values of the compression load *F* (Figure 1) were applied to the model, corresponding to the following values of force per unit surface *p* = 0.2, 0.5, 1.0, and 1.5 MPa. Such values are consistent with those hypothesized in previous studies [36,58]. Symmetry constraints were applied on the lateral surfaces to simulate the continuity of the entire model (Figure 1). Poroelastic four-node tetrahedral finite elements C3D4P available in Abaqus were adopted to mesh the scaffold model (Figure 3). The model of the scaffold and of the granulation tissue included about 5M elements with 1M nodes.

The modeled scaffolds were physically fabricated via the FDM technique by utilizing the polylactic acid (PLA), a biodegradable thermoplastic polyester considered a bioplastic, possessing a Young's modulus of 2300 MPa [59]. This same value of Young's modulus was implemented in the finite element model of the scaffold and utilized in the optimization algorithm described below. The other material properties implemented for scaffold and granulation tissue are the same as those utilized in previous studies [36,40,60] and are listed in Table 1.

**Figure 1.** Parametric model of the scaffold (**a**,**c**) and of the granulation tissue (**b**) occupying the scaffold pores. (**d**) Exploiting the symmetry of the system, the one-quarter model was investigated.

**Figure 2.** Models of scaffold investigated in the study including strands with diameter *D* = 400 µm (**a**), 500 µm (**b**), 600 µm (**c**), 700 µm (**d**), and 800 µm (**e**). The distance between the filaments was computed by means of the proposed optimization algorithm.

**Figure 3.** Finite element mesh utilized to model the scaffold (**a**) and the granulation tissue (**b**) including poroelastic four-node tetrahedral elements (C3D4P).

 = + = 3 μm/s are two empirical constants determined in a previous study

*ε*

Young's modulus (MPa) Poisson's ratio

**issue** 

−14 −14

*ε ε*


**Table 1.** Material properties utilized in the model of scaffold and granulation tissue

*2.2. Mechanobiological Model by Prendergast and Huiskes to Describe the Bone Regeneration Process inside the Sca*ff*old*

Once the scaffold is implanted in the anatomic region with bone deficiency, the mesenchymal stem cells (MSCs) migrate from the surrounding tissues and invade the scaffold pores. Based on the mechanical stimulus acting on them, MSCs will start to differentiate into different phenotypes. The mechanobiological model by Prendergast and Huiskes hypothesizes that the biophysical stimulus *S* that triggers the tissue differentiation process is a function of the octahedral shear strain γ and of the interstitial fluid flow *v*, i.e., the velocity with which the fluid flows through the solid phase, according to the relationship

$$S = \frac{\mathcal{V}}{a} + \frac{v}{b} \tag{1}$$

where *a* = 3.75 % and *b* = 3 µm/s are two empirical constants determined in a previous study [43]. In particular, the octahedral shear strain γ can be expressed in function of the principal strains ε*<sup>I</sup>* , ε*II* and ε*III* as

$$\gamma = \frac{2}{3}\sqrt{(\varepsilon\_I - \varepsilon\_{\text{II}})^2 + (\varepsilon\_{\text{II}} - \varepsilon\_{\text{III}})^2 + (\varepsilon\_I - \varepsilon\_{\text{III}})^2} \tag{2}$$

Depending on the value of *S*, stem cells will be differentiated into the following phenotypes

$$\begin{cases} \text{ } & \text{if } S > 3 \rightarrow \text{ } & \text{formation of } \text{ } \text{ } \text{ } & \text{Fibrosis } (\text{Fibrosis } \text{} \text{ } \text{ } \text{ } & \text{else})\\ \text{ } & \text{else} \text{ } \text{ if } 1 < S < 3 \rightarrow \text{ } & \text{formation of } \text{ } \text{ } \text{ } \text{ } & \text{(Craftage tissue)}\\ \text{ } & \text{else} \text{ } if \, 0.53 < S < 1 \rightarrow \text{ } & \text{formation of } \text{ } \text{ } & \text{(Immuture } \text{ } \text{ } & \text{(\$ } \text{)}\\ \text{ } & \text{else} \text{ } if \, 0.01 < S < 0.53 \rightarrow \text{ } & \text{formation of } \text{ } \text{ } \text{ } & \text{(Mature bone)}\\ & \text{else} \text{ } if \, 0 < S < 0.01 \rightarrow \text{ } & \text{bone resolution} \end{cases} \tag{3}$$

The threshold limits reported in the inequalities (3) are the same as those utilized in a previous study [61].

#### *2.3. Mechanobiology-Based Optimization Algorithm*

The task of determining the optimal distance *dfil* between the filaments was accomplished by implementing an optimization algorithm, a schematic of which is illustrated in Figure 4. The choice of utilizing *dfil* as a design variable and the strand diameter *D* as input parameter entered by the user derives from the fact that in the FDM technique, the distance between the strands can be changed with continuity while the strand diameter cannot be controlled with precision as it depends on the nozzle diameter. The algorithm implements the optimization tool available in MATLAB® (Version R2016b, MathWorks, Natick, USA) *fmincon* devoted to finding the minimum of a multivariable scalar function starting at an initial estimate. The objective of the optimization algorithm is to identify the optimal value of the filaments distance *dfil\_optim* that allows maximizing the amounts of mature bone that are predicted to generate within the scaffold pores.

The user has, first, to select the value of the diameter *D* and of the load per unit area *p* acting on the scaffold model (Blocks [A] and [B], Figure 4). Then, the user is prompted to specify a tentative initial value of the distance between the filaments *dfil* (Block [C], Figure 4). At his point, the algorithm writes a python script (Block [D]) and enters into it the tentative value provided by the user (Block [E], Figure 4). The python script is then given in input to Abaqus (Block [F], Figure 4) that, executing the instructions of the script: (i) builds the CAD model of both, the scaffold and the granulation tissue and applies the boundary and loading conditions above described (Figure 1d) (Block [G], Figure 4); (ii) generates the poroelastic tetrahedral finite element mesh (Block [H], Figure 4); and (iii) runs the finite element (FE) analysis (Block [L], Figure 4). Once the FE analysis has terminated, Abaqus prints, for all the elements inside the scaffold pores (highlighted in red in Figure 1), the volume of the element and the values of strain and interstitial fluid velocity computed in the analysis. At this point, the algorithm reads the document printed by Abaqus and computes, according to the Equation (1) the value of the biophysical stimulus *S* acting on the single element (Block [M], Figure 4). Then, the algorithm compares all the obtained values of *S* with the threshold limit reported in the inequalities (3). For the elements where the formation of mature bone is predicted (i.e., the inequality 0.01 < *S* < 0.53 is satisfied) the algorithm stores the value of the element volume (Block [N], Figure 4).

Once for all the elements, *S* was computed, the algorithm determines the total volume of the elements that were predicted to differentiate into mature bone, by summing up all the element volumes previously stored. Then, if *VBONE* is the total volume of the elements 'mature bone' and *VTOT* = π × *R* <sup>2</sup> <sup>×</sup> *<sup>h</sup>* the total volume of the model, the algorithm computes the percentage *BO%* of the scaffold volume that is predicted to be occupied by mature bone as the ratio between *VBONE* and *VTOT* multiplied by 100 (Block [P], Figure 4). As it is clear, the task pursued by the algorithm is to increase as much as possible the percentage *BO%*. However, considering that the optimization tool utilized *fmincon* is designed to determine the minimum value of functions, the objective function Ω(*dfil*) was defined as the opposite value of *BO%* (Block [Q], Figure 4). At this point, the algorithm perturbs the scaffold geometry many times (Block [R], Figure 4), i.e., it proposes different values of *dfil* as possible candidate solutions (Block [T], Figure 4) and for each of the proposed value, it stores the value of Ω(*dfil*). Figure 5 shows, for instance, the typical values of *BO%* computed by the algorithm for different values of hypothesized *dfil*. Once the algorithm has enough points to identify the minimum of Ω(*dfil*) or, equivalently, the maximum of *BO%*, (denoted as *BO%\_max* in Figure 5) its stopping criteria are satisfied and hence stops, giving in output, for the above-selected *D* and *p*, the optimal value of the strand distance *dfil\_opt* (Block [S], Figure 4).

The domain of variability of *dfil* was hypothesized to range between the following lower and upper bounds: *dfil\_min* = *D* (which corresponds to have the strands in reciprocal contact) and *dfil\_max* = 1100 µm which is approximately the average value of the distance of strands utilized by Neufurth et al. [62] and Bartolo et al. [63].

Each finite element analysis had an average duration of 4 hours on an HP XW6600-Intel®Xeon®DualProcessor E5-5450 3 GHz–32 Gb RAM workstation. Considering that each optimization cycle required about 50 finite element analyses to identify *dfil\_opt* and considering that five values of *D* (*D* = 400, 500, 600, 700, and 800 µm) and four values of *p* (0.2, 0.5, 1.0, and 1.5 MPa) were hypothesized, one can conclude that the total time to carry out all the analyses conducted in this study is: 4 × 50 × 5 × 4 = 4000 hours. In summary, 5(no. of values of *D*) × 4(no. of values of *p*) = 20 values of *dfil\_opt* were computed, i.e., 20 scaffold geometries were optimized.

**Figure 4.** Schematic of the optimization algorithm written in MATLAB environment.

**Figure 5.** Typical values of *BO%* obtained in an optimization process for different values of *dfil* . In detail, the diagram refers to the case of *D* = 600 µm and *p* = 0.5 MPa.

#### *2.4. Fabrication of the Optimized Sca*ff*olds*

– For each of the 20 optimized geometries, three samples of scaffold were fabricated via the FDM technique. With the aim of keeping approximately a cylindrical shape of the deposited filament, an Ultimaker 3 was utilized, adjusting the slice height and the flux of extruded filament into the nozzle. If *D<sup>n</sup>* is the nozzle diameter, *v<sup>n</sup>* is the travel speed of the extrusion head, and *v<sup>h</sup>* is the speed of the filament inside the nozzle, slicing software allows modifying the flow rate of extruded strand according to a flow rate coefficient *f*, usually expressed in terms of percentage

$$D\_n^2 v\_n = D^2 v\_h \tag{4}$$

with

$$v\_n = f v\_h \tag{5}$$

The deposited filaments are nearly cylindrical if the slice height is equal to *D<sup>n</sup>* , while it is possible to decrease *D* using *f* values lower than 100% and consequently reducing the slice height, according to the Comminal's model [25], under the hypothesis of negligible effects on the section circular shape. Considering that, commercial extruders are available for Ultimaker 3D printers with a diameter equal to 400 and 800 µm, intermediate values of the strand diameter have been obtained using nozzles 0.4 and 0.8, lowering *f* according to Equations (4) and (5), respectively (Table 2).

 <sup>2</sup> = <sup>2</sup><sup>ℎ</sup> **Table 2.** Values of the flow rate *f* utilized to obtain different strand diameters


As regards the remaining process parameters default values have been exploited: *v<sup>h</sup>* = 40 mm/s, extrusion temperature 180 ◦ and build plate temperature 50 ◦ to lower shrinkage. <sup>ℎ</sup> = 40 /

#### *2.5. Measurement of the Dimensions of the Fabricated Sca*ff*olds*

A De Meet 400 Coordinate Measuring Machine CMM, by Schut Geometrical Metrology, Germany, was utilized to measure the diameter of strands and their reciprocal distance *dfil* . The objective of these measurements was to compare: (i) the distance of filaments actually fabricated and that, denoted as *dfil\_opt*, optimized with the algorithm above described; (ii) the dimension of the strand diameter actually realized with the nominal one. The CMM is equipped with lenses that can be moved and focused by the user. In detail, *dfil* was measured as the center to center distance between the filaments (Figure 6). For each fabricated sample, 10 measurements were taken of *dfil* and 10 measurements of *D*.

**Figure 6.** For each scaffold fabricated, the center to center distance between the filaments *dfil* and the dimension of the diameter *D* were measured.

In order to evaluate the correctness of the hypothesis that an overlap region of 0.1 × *D* exists [57] between two adjacent strands of two consecutive layers, three measurements of this region were carried out for each sample fabricated.

#### **3. Results**

By implementing the optimization algorithm above described, the optimal geometry of the scaffold for different levels of *D* and *p* was computed (Figure 7). As expected, for increasing values of load, decreasing values of the optimal distance *dfil\_opt* were predicted (Figure 8a). In fact, as the load increases, the biophysical stimulus *S* increases too thus stimulating the formation of 'softer' tissues like the cartilage and the fibrous tissue (see Equations (1) and (3)). To prevent this, the algorithm predicts smaller distances between the strands. This leads to an increase in the global stiffness of the scaffold and hence to a decrease in the value of the biophysical stimulus thus stimulating the formation of

'harder' tissues such as the bone. Interestingly, for an assigned value of load *p* and for decreasing dimensions of diameter *D*, the algorithm predicts increasing amounts of bone (Figure 8b).

**Figure 7.** Optimal scaffold geometries predicted by the optimization algorithm for different values of the filament diameter *D* and the load per unit area *p*.

**Figure 8.** Optimal dimensions of *dfil* (**a**) and percentage of the scaffold volume occupied by mature bone (**b**) predicted for different values of the load.

The measurements carried out on the overlap region of adjacent filaments revealed an average value of this dimension of (0.12 ± 0.04) × *D*, which demonstrates the reasonable appropriateness of the hypothesis followed [57]. The values of the distance *dfil\_opt* optimized with the proposed algorithm fell, for almost all the hypothesized values of *p*, within the range [average ± standard\_deviation] of the measured dimensions (Figure 9). In general, it appears that the difference between the optimized distances *dfil\_opt* and the average measured distances never exceeded 8.27% of the optimized distance. Regarding the measured values of the diameter, we noticed that, for *D* = 400 µm and *D* ≥ 600 µm, the measured dimensions are very close to the nominal ones (Figure 10a,c–e)), while for *D* = 500 µm, large differences can be seen (Figure 10b). In fact, for *D* = 400 µm and *D* ≥ 600 µm, the nominal dimension of the diameter fell, almost in all the values of *p* investigated, in the range [average ± standard\_deviation] of the measured dimensions. In the case of *D* = 500 µm, instead, the nominal value of the diameter is abundantly out of the above-mentioned range, which indicates that significant fabrication errors are committed.

**Figure 9.** Comparison between measured and optimized distances (between the filaments), for different diameters: (**a**) *D* = 400 µm, (**b**) *D* = 500 µm, (**c**) *D* = 600 µm, (**d**) *D* = 700 µm, (**e**) *D* = 800 µm.

**Figure 10.** Comparison between measured and nominal dimension of diameter: (**a**) *D* = 400 µm; (**b**) *D* = 500 µm; (**c**) *D* = 600 µm; (**d**) *D* = 700 µm; (**e**) *D* = 800 µm.

#### **4. Discussion**

A feasibility study was conducted aimed to investigate whether the FDM technique can be utilized to fabricate scaffolds designed and optimized to undergo a compression load. A mechanobiology-based optimization algorithm was developed and implemented to determine the optimal distance between the filaments of cylindrical scaffolds for bone tissue engineering. The optimal distance was predicted for different hypothesized values of the load acting on the scaffold and diameter of strands. The scaffolds with the optimized dimensions were hence physically fabricated with the FDM technique and successively measured. The precision guaranteed by the FDM technique was finally evaluated by comparing the measured dimensions with the nominal ones.

This study presents some limitations in the model, the computational mechanobiological algorithm, and the experimental measurements. Regarding the model, first, we hypothesized that the strands of a given layer are aligned and form an angle of 90 ◦ with those of the adjacent one. Ideally, the model of the scaffold should include the angle formed between the layers as a design parameter that should be optimized via the mechanobiology-based optimization algorithm above described.

Second, we hypothesized that an overlap region of 0.1 × *D* exists between two adjacent filaments of consecutive layers [57]. In an ideal model, also this dimension should be included as a design parameter to be optimized by means of fmincon. However, including these two additional design variables would make the computational costs tremendously larger than those spent in this study. Third, it would be interesting to investigate how the proposed optimization algorithm works in the case of more complex loading conditions acting on the scaffold model.

However, the hypothesis of more complex loading conditions would lead to losing the symmetry conditions and hence to the impossibility of using the simplified one-quarter model. Regarding the mechanobiological model, we identified the optimal scaffold geometry based on the values of the biophysical stimulus acting on the granulation tissue, i.e., the tissue that was hypothesized to occupy the volume inside the scaffold pores. In reality, this biophysical stimulus changes in time as the granulation tissue is replaced by the other tissues forming during the regeneration process. In other words, in the mechanobiological model, we did not take into account the variable time. However, the inclusion of the time would increase by at least two orders of magnitude the computational cost required to carry out the analyses. Furthermore, at different compression force p, the optimized value dfil\_opt is different. In physiological conditions, the compression force acting on a bone may not be constant but variable. A possible strategy that can be adopted to optimize a scaffold subject to a variable compression load consists in designing functionally graded scaffolds, i.e., scaffolds where the geometric parameters change depending on the specific load value acting in the specific scaffold region. For instance, in the regions where the load acting is higher, a functionally graded scaffold may include strands at a shorter distance, in the regions where, instead, the load is smaller the distance between the filaments may be increased. Such a strategy requires including as many design variables as required to provide an adequate structural response to the variable load and is, therefore, different orders of magnitude more expensive than the approach adopted in this study. Increases in computational power will ultimately allow the investigation of the effect of additional geometric parameters on the optimal scaffold geometry and include different design variables and variable time in the optimization analyses. Regarding the experimental measurements, only two dimensions—the strand diameter and the distance between the filaments—were measured and compared with the corresponding quantities obtained via the mechanobiology-based optimization algorithm. The choice of measuring only these two geometrical parameters is due to the fact that by adjusting the diameter and distance between extruded strands, it is possible to design various topologies with variable values of porosity and therefore to have a wide control of the scaffold micro-architecture [34]. All the other geometric parameters involved in the scaffold designing will certainly play a role less relevant than that played by the distance between the strands and the diameter of the filaments. Furthermore, the proposed optimization algorithm was not validated experimentally. The validation requires a large number of experiments as well as an experimental set-up properly studied and organized to make the experimental conditions equivalent to those hypothesized in the numerical model, which goes beyond the scope of this feasibility study. However, should be clarified that the optimization carried out in this study takes into account only mechanobiological aspects and neglects many other chemical and genetic aspects that certainly affect the differentiation process. Therefore, by 'optimized design' we should intend only a design optimized from the mechanobiological point of view. More sophisticated optimizations taking into account the large number of aspects and variables involved in fracture healing should be the object of future studies.

In spite of these limitations, the predictions of the proposed optimization algorithms are consistent with the results of other studies reported in the literature.

(i) Barba et al. [64] implanted cylindrical scaffolds fabricated with FDM into bone defects generated in the limb of adult beagle dogs. In detail, two scaffold types were implanted, one with filaments of 250 µm and the other with filaments of 450 µm. Interestingly, they found that the amounts of bone formed in scaffolds with filaments of 250 µm are larger than those observed in scaffolds with filaments of 450 µm. This result is consistent with the predictions of the proposed optimization algorithm that found increasing amounts of bone in filaments with decreasing values of the diameter *D* (Figure 8b).

(ii) The typical distribution of the von Mises stresses within the scaffold model displays the presence of stress peaks in the proximity of the point where the generic strand enters in contact with the

125

strand of the adjacent layer (Figure 11). This same mechanical behavior was reported by Uth et al. [34] who observed peaks of stress in alignment with the filaments of the previous layer.

**Figure 11.** Normalized von Mises stress distribution in a section of the scaffold model. Stress peaks are in alignment with the filaments of the previous layer.

(iii) The amounts of bone predicted with the proposed algorithm are comparable with those predicted with other scaffolds based on different unit cell geometries [36,38,40,41]. However, it appears that scaffolds fabricated with the FDM technique allow the formation of amounts of bone significantly smaller than those obtained with other regular, beam-based scaffolds. In general, in scaffolds fabricated with the FDM technique, the amounts of bone predicted to create are approximately 20% lesser than those generated in other scaffolds [41].

'soft' tissues such as the car The results obtained are consistent with the physics of the problem. As the load increases, the biophysical stimulus *S* increases too and with it, the percentage of the scaffold volume occupied by 'soft' tissues such as the cartilage and the fibrous tissue. To counterbalance this tendency, the optimization algorithm tends to decrease the distance between the filaments hence making the scaffold stiffer. The increase in stiffness leads to decreasing levels of *S* and hence to the formation of harder tissues such as the immature and the mature bone. For very high values of load, the optimal distance *dfil\_opt* tends, asymptotically, towards the dimension of the filament diameter *D* (Figure 8a). The optimal distance between the filaments *dfil\_opt* was determined for different values of load and for different assigned values of diameter *D* (Figure 8b). In general, it appears that using smaller diameters, for a fixed value of load *p*, leads to the formation of larger amounts of bone (Figure 8b). Therefore, if one can choose the nozzle diameter (i.e., if the FDM machine is equipped with nozzles of different diameter), to which the strand diameter is strictly related, should prefer the smaller diameters.

 The dimensions measured on the samples were compared with the nominal one, in order to assess the accuracy guaranteed by FDM in the fabrication of scaffolds for bone tissue engineering. For almost all the hypothesized values of load per unit area p, the optimal dimensions dfil\_opt fell within the range (average ± standard\_deviation) of the measured distances. This leads us to conclude that the FDM is suited to reproduce with high accuracy the designed and optimized distance between the filaments dfil\_opt. Furthermore, the values of the standard deviation of the measured distances (between the filaments) never exceeded 6.02% of the optimized dimension, which indicates a reasonably small dispersal of data and hence a rather high reproducibility of the fabrication process, in terms of the distance between the filaments. Regarding the measured dimensions of the filament diameter, we found that the diameters are well reproduced in the case of D = 400 µm and D ≥ 600 µm, but large reproduction errors are made when D = 500 µm. This behavior can be justified with the argument that the strategy of using different flow percentages of PLA to have filaments with different diameters presents a lower limit below which the quality of the filament diameter decreases significantly. Indeed, 500 µm is the diameter that mostly differs with respect to that of the nozzles utilized to extrude the filaments. We can hypothesize that when the material flow is significantly smaller than 100%, the nozzle does not fill correctly and hence the filament cannot form properly. This leads to the deposition of filaments with dimensions significantly different with respect to the nominal ones. However, this limitation can be easily overcome by equipping the FDM machine with a greater number of nozzles. In general, in ideal conditions, to minimize the fabrication errors, nozzles with the same diameter of the strands to be fabricated should be utilized.

#### **5. Conclusions**

We conducted a feasibility study which aimed to investigate the potentialities of the FDM technique to be used for the fabrication of scaffolds designed and optimized with mechanobiological algorithms. The present article is the first study ever reported in the literature where the geometry of scaffolds fabricated with the FDM technique is optimized via a mechanobiology-based optimization algorithm. In detail, the optimal distance between the filaments was predicted in function of the filament diameter and of the load acting on the scaffold. The designed and optimized scaffolds have been fabricated and measurements on the dimensions of the samples realized were carried out. We found that the difference between the average dimensions of the fabricated scaffolds and the nominal ones never exceeded 8.27% of the nominal dimension, which demonstrates the rather good accuracy of the FDM technique in reproducing the distance between the filaments. Furthermore, the values of the standard deviation of the acquired distances (between the filaments) never exceeded 6.02% of the optimized dimension, which indicates a reasonably small dispersal of data and hence a rather high reproducibility of the fabrication process in terms of distance between the filaments. However, we found that large reproduction errors are made on the filament diameter when the filament diameter to be realized differs significantly with respect to the nozzle diameter.

In conclusion, we can state that the FDM technique is suitable to build accurate scaffold samples only in the cases where the filament diameter is close to the nozzle diameter. Conversely, when a large difference exists, large fabrication errors can be made on the diameter of the filaments. In general, the scaffolds realized with the FDM technique were predicted to stimulate the formation of amounts of bone smaller than those that can be obtained with other regular beam-based scaffolds.

**Author Contributions:** Conceptualization, A.B. and G.P.; Methodology, A.B.; Experimentation, G.P., Computations, A.B.; Validation, A.B. and G.P.; Investigation, A.B., G.P., and A.E.U.; Data curation, A.B. and G.P.; Writing—original draft preparation, A.B.; Writing—review and editing, G.P., A.E.U., M.F., M.G., and V.M.M.; Visualization, A.B., G.P., A.E.U., M.F., M.G., and V.M.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Review*

### **Exploring Macroporosity of Additively Manufactured Titanium Metamaterials for Bone Regeneration with Quality by Design: A Systematic Literature Review**

### **Daniel Martinez-Marquez , Ylva Delmar, Shoujin Sun and Rodney A. Stewart \***

School of Engineering and Built Environment, Griffith University, Gold Coast, QLD 4222, Australia; daniel.martinezmarquez@griffithuni.edu.au (D.M.-M.); ylvadelmar@hotmail.com (Y.D.); shoujin.sun@griffith.edu.au (S.S.)

**\*** Correspondence: r.stewart@griffith.edu.au

Received: 23 September 2020; Accepted: 23 October 2020; Published: 27 October 2020

**Abstract:** Additive manufacturing facilitates the design of porous metal implants with detailed internal architecture. A rationally designed porous structure can provide to biocompatible titanium alloys biomimetic mechanical and biological properties for bone regeneration. However, increased porosity results in decreased material strength. The porosity and pore sizes that are ideal for porous implants are still controversial in the literature, complicating the justification of a design decision. Recently, metallic porous biomaterials have been proposed for load-bearing applications beyond surface coatings. This recent science lacks standards, but the Quality by Design (QbD) system can assist the design process in a systematic way. This study used the QbD system to explore the Quality Target Product Profile and Ideal Quality Attributes of additively manufactured titanium porous scaffolds for bone regeneration with a biomimetic approach. For this purpose, a total of 807 experimental results extracted from 50 different studies were benchmarked against proposed target values based on bone properties, governmental regulations, and scientific research relevant to bone implants. The scaffold properties such as unit cell geometry, pore size, porosity, compressive strength, and fatigue strength were studied. The results of this study may help future research to effectively direct the design process under the QbD system.

**Keywords:** porous implants; bone implants; metamaterials; titanium; mechanical properties; pore size; unit cell; porosity; elastic modulus; compressive strength; additive manufacturing

#### **1. Introduction**

#### *1.1. Current Issues with Traditional Bone Implants and Sca*ff*olds*

Many physical conditions necessitate bone tissue replacements and joint implants. Some of these conditions are caused by degenerative diseases, birth defects, and orthopaedic traumas [1]. However, despite the tremendous progress in biomedical engineering, 20% of patients subjected to joint reconstructive surgery experience significant problems [2]. This situation is reflected in the fact that orthopaedic products, such as knee and hip prostheses, are the fifth most recalled medical products; of these recalls, 48% are due to manufacturing issues and 34% to design flaws [3,4]. Some of the main flaws with orthopaedic implants are associated with their longevity, material properties, and mismatch with patient size and shape requirements [5,6]. Stress shielding is one of the main design flaws of load-bearing prostheses. This phenomenon occurs because bone is a self-healing material that requires load application to remodel itself, but a material with a higher modulus of elasticity (*E*) absorbs all the stress generated, leading to bone reabsorption and subsequent loosening of the implant [7].

In the case of bone defects, they can be caused by tumour resection, infections, complex fractures, and non-unions [8]. The most common treatment for bone defects is surgical intervention, where an autograft (bone taken from the patient's body) is used to fill bone defect spaces [9]. However, due to their restricted availability, allografts (bone tissue from a deceased donor) are frequently used to treat critical-size defects [9]. Bone grafting is a common surgical procedure; it has been estimated that 2.2 million grafting procedures are performed worldwide each year [8]. However, late graft rupture has been reported to be as high as 60% 10 years after the grafting procedure [10]. Allograft transplantation has a success rate of approximately 70%. The low success rate of allografts is caused by the prevalence of infection, rejection by the host's immune system, fatigue fractures, delayed union, non-union, and incomplete graft resorption [11,12]. In the case of autografts, the disadvantages are increased post-operative morbidity, lack of available tissue, chronic pain, infection, nerve injury, and weakened bone donor graft sites [12,13].

To solve these grafting problems, several scaffold traditional techniques have been used without much success: solvent-casting particulate-leaching, gas foaming, fibre meshes (fibre bonding), phase separation, melt moulding, emulsion freeze drying, solution casting, and freeze drying [14]. Some of the disadvantages of traditional scaffold fabrication techniques are their lack of control over porosity characteristics, such as pore size, pore distribution, and interconnectivity; the toxic by-products of scaffold degradation; and their lack of consistent mechanical properties [14]. Hence, traditional techniques for bone reconstruction including grafting and prostheses are not sufficiently effective, which represent a medical challenge that comes with several limitations and risks [9]. Moreover, no material yet exists with the ideal properties for bone tissue replacement [15–17]. To overcome these issues, tissue engineering has focused on additive manufacturing technologies to produce the next generation of bone implants and scaffolds.

#### *1.2. Additive Manufacturing*

Additive manufacturing (AM) technologies, supported by computer-aided design (CAD) software, progressively build 3D physical objects from a series of cross-sections, which are joined together to create a final shape [18]. With AM, it is possible to create complex interconnected and porous structures with controlled pore size, shape, and distribution and properties resembling bone mechanical properties, such as a modulus of elasticity to induce bone ingrowth [19,20]. This capability permits the fabrication of hierarchical structures at the microscale and the manipulation of material properties to create metamaterials. In terms of implant design, this advance means that products can be designed with a biomimetic approach according to the patient's anatomy and the bone tissue's mechanical properties [21]. The design freedom of AM allows its use in difficult clinical scenarios in which bone diseases, deformities, and trauma usually necessitate the reconstruction of bone defects with complex anatomical shapes, which is extremely difficult even for the most skilled surgeon [22]. The complex reconstruction of bone defects is possible through combining the advantages of AM with CAD and medical imaging technologies, such as computed tomography and magnetic resonance, to fabricate implants according to the patient's specific anatomy, thus achieving an exact adaptation to the region of implantation [23]. In the search of suitable materials for AM, bone regeneration, and implant application tissue engineering has focused on developing a variety of different types of synthetic and natural materials.

#### *1.3. Materials for Bone Regeneration and Implant Applications*

Materials appropriate for implantation within the human body require distinct biocompatible properties. Therefore, in the selection of appropriate materials for implant applications, several factors must be considered. First, the intended implant location must be considered to predict host response, which is governed by the biochemical and physical environments in contact with the medical device [24,25]. Second, the material should possess appropriate biological and mechanical properties for its specific purpose to prevent physical damage to the body. Third, from the perspective of tissue engineering, materials should mimic one or multiple characteristics of the natural region of repair. In the case of bone repair, the desired characteristics are osteoconductivity, osteoinductivity, and osseointegration. As a result, for an optimum scaffold and prosthesis design, material science may combine several technologies to create suitable materials that fulfil these needs.

#### 1.3.1. Polymers

Polymers for AM and tissue engineering applications are biocompatible materials that offer several advantages over other materials, including biodegradability, cytocompatibility, easy processability, and flexibility in the tailoring of their properties [26]. Polymers can be classified as natural or synthetic and some of them already have regulatory approval [27].

Natural polymers are made from proteins such as alginate, gelatine, collagen, silk, chitosan, cellulose, and hyaluronic acid [28]. The advantages of natural polymers are their excellent biodegradability, low production costs, and superior chemical versatility, as well as their improved biological performance that allow better interactions with cells than other biomaterials, improving their attachment and differentiation [29]. However, natural polymers can be expensive to produce, due to the difficulty in controlling their mechanical properties, biodegradation rate, and quality consistency [30].

Due to the disadvantages of natural polymers, different synthetic polymers, such as polycaprolactone (PCL), polylactic acid (PLA), and poly Lactic-co-Glycolic Acid (PLGA), have been developed. Their advantages include low immunogenic potential, large scale low production cost, and good quality consistency [31]. Moreover, their mechanical properties, microstructure, and degradation rate can be tuned according to needs [27]. Despite the advantages of natural and synthetic polymers, they are unsuitable for load-bearing applications due to their lower modulus of elasticity compared to bone, unstable mechanical strength, and tendency to creep [32,33]. Hence, in recent years, a variety of polymers have been combined with different materials to such as bioceramics (e.g., bioglasses, tri-calcium phosphates, and carbon nanotubes) and metals to create composite materials with tunable mechanical properties as well as with the capacity to deliver drugs, exosomes, and growth factors, to name a few [34–37].

#### 1.3.2. Bioceramics

Bioceramics are a large group of materials used for bone substitution and regeneration. Calcium phosphate (CaP) ceramics is one of the main groups of bioceramics. Calcium phosphate ceramics are abundant in bone, constituting between 80% and 90% of bone's anorganic matter. This group of bioceramics is widely used as implant coating, bone grafting, and more recently have been fabricated for bone scaffolding applications with AM [38,39]. Hydroxyapatite HAP and β-tricalcium phosphate (β-TCP), are the most-studied CaP bioceramics. The main advantages of calcium phosphate materials are their osteoinductive and osteoconductive properties, as well as their dissolution in body fluids [40]. For load-bearing applications, the major disadvantage of CaPs is their poor mechanical properties. Despite their good compressive strength, CaPs lack plastic deformation, making them brittle and prone to cracking. Consequently, these materials are not yet suitable for load-bearing applications [41]. Nevertheless, the lower wear rate of CaPs makes these materials the preferred choice for surface coating to reduce wear in joint prostheses [42]. They are also commonly used for spinal fusion, maxillofacial and cranio-maxillofacial reconstruction, as well as bone filler and bone cement due to their excellent biocompatibility and osteoconductivity [43].

Discovered in 1969 by Larry Hench, bioglasses are ceramic materials composed of calcium, phosphorus, and silicon dioxide [16]. Bioglasses are bioactive ceramic materials with strong osteointegrative and osteoconductive properties, as well as higher mechanical strength than calcium phosphate ceramics [44]. Hence, bioglasses have been intensely investigated with AM for bone tissue engineering applications [45,46]. The advantage of these materials is that by changing the proportions of their basic components, different forms with different properties can be obtained; for example, non-resorbable bioglasses can be transformed into resorbable bioglasses [44]. Moreover, they

can be designed with controlled biodegradability and drug and cell delivery capabilities [47,48]. Their applications also include bioglass scaffolds produced using AM with controlled porosity architecture and improved mechanical properties for bone regeneration [49]. However, bioglasses are limited for use in practical load-bearing applications due to their low resistance to cyclic loading and their brittleness [50].

#### 1.3.3. Metals and Titanium as a Bio-Metamaterial

Metals have been the common choice to replace hard tissue in load-bearing applications due to their mechanical properties, corrosion resistance, and biocompatibility. Most of these materials are alloys, such as 316L stainless steel (316LSS), cobalt chromium (Co–Cr), and titanium (Ti) alloys [5]. Among all metallic materials, the titanium alloy Ti–6Al–4V is the gold standard for orthopaedic applications [51,52] because of its high biocompatibility [53], high corrosion resistance, low modulus of elasticity [5], and high strength-to-weight ratio [19]. Furthermore, Ti is a reactive metal that naturally forms a thin layer of oxide, which blocks metal ions from reaching its surface, increasing its biocompatibility [54]. The biomedical applications of Ti–6Al–4V are quite broad and encompass dental implants; hip, shoulder, knee, spine, elbow, and wrist replacements; bone fixation components; and cardiovascular applications [5].

Nevertheless, the most common problems of metallic materials are wear and the stress-shielding effect caused by their high modulus of elasticity compared with bone [52,55–57]. Moreover, despite the excellent biocompatibility and mechanical properties of Ti and Ti alloys, they usually require long healing periods to create a stable interface with the surrounding bone [58], with insufficient implant osseointegration as a potential outcome [59]. Hence, to further augment the biological, mass transport, and mechanical performance of Ti and Ti alloys different metamaterials have been developed. For example, metallic bone implants with a modulus of elasticity similar to that of bone can drastically reduce wear, shear stress, and bone resorption and consequently prevent implant loosening and revision surgery [60]. This may translate into enhanced quality of life for the patient, reductions in hospital expenses and recovery time, and improvement in joint dynamic performance [61]. With porous Ti and Ti alloy bio-metamaterials, osseointegration is also improved, and superior results have been accomplished in relation to mechanical properties. Nonetheless, pores act as stress concentrators, reducing the material load capacity [23]. As a result, for the design of load-bearing prostheses, it is crucial to balance mechanical properties with biological stimulation. Consequently, there have been several efforts to find the optimal balance between pore size and porosity percentage in different materials. For example, Zaharin et al. [62] investigated the effect of pore variation on the porosity and mechanical properties of several Ti–6A–l4V porous scaffolds. According to their results, scaffolds based on cube and gyroid unit cells with a pore size of 300 µm provided similar properties to bone. Moreover, they concluded that increments in porosity decreased the scaffolds' elastic modulus and yield strength. In an earlier study Bobyn at al. [63] investigated the effects of pore size variation of cobalt-base alloy implants on the rate of bone growth. For this purpose, casted cobalt-base alloy implants were coated with powder particles and implanted into canine femurs for several weeks. The results indicated that pore sizes between 50 and 400 µm provided the maximum bone ingrowth and fixation strength.

Despite the excellent biocompatibility and mechanical properties of Ti and Ti alloys, they usually require long healing periods to create a stable interface with the surrounding bone, frequently resulting in insufficient osseointegration [64]. Hence, to further augment Ti's bioactivity, corrosion resistance, and mechanical properties different mechanical, chemical, and physical surface modification methods have been developed [65–67]. Depending on the surface treatment used to modify Ti substrate, different topographic features can be achieved at the macroscale, microscale, and nanoscale. There is a large amount of evidence that rough Ti surfaces with topographic microfeatures better protein adsorption and provide higher osteoblasts attachment growth proliferation and activity than surface smooth surfaces [68]. Nonetheless, it has been demonstrated that nanoscale topography outperforms macro

and micro-scale surface features towards augmenting cellular functions [69]. More recently, at has been proposed that a combination of different topographic features at the macro, micro, and nanoscale with local drug delivery functions can further enhance the biological, chemical, tribological, and mechanical performance of Ti bone implants [70–73].

#### *1.4. Purpose and Objectives*

The purpose of this research is to provide researchers and industry with an in-depth adaptation of the Quality by Design (QbD) system for the fabrication of additively manufactured porous Ti implants considering the QbD guidelines for 3D printed bone implants and scaffolds [74]. The QbD system is composed by eight main steps that need to be systematically followed to acquire a complete comprehension of the product and its manufacturing process, including the identification and control of all variables to achieve the desired quality. Specifically, the scope of this present study was limited to the first step of the QbD framework (Figure 1). Thus, the objectives of this study are:


**Figure 1.** Schematic showing the focus of this study within the QbD system.

#### **2. Materials and Methods**

This systematic research study is part of the implementation of the QbD approach for porous metal implants. Therefore, a constructive research approach was used to further extend the QbD system for patient-specific bone implants and scaffolds produced by AM [75,76]. For an in-depth interpretation and synthesis the researchers immersed themselves in the contextual literature [77]. This was an exploratory qualitative study which requires the collection of secondary data from various datasets of peer-reviewed publications following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [78].

#### *2.1. Data Collection*

A systematic search was conducted on 20 January 2020 in the Science Direct and Google Scholar databases according to objectives and the PRISMA statement. Relevant keywords were connected with the Boolean operators "OR" and "AND". Terms relevant to this research included the following: Titanium, Ti, additive manufacturing, 3D printing, rapid prototyping, bone tissue engineering, bone implant, implant, scaffold, prostheses, porous, porosity, and mechanical properties. To specify the search further, the terms were connected with Boolean operators (AND, OR): Implant(s); scaffold(s); prosthes(is, es); titanium; Ti; additive manufacturing; additively manufactured; 3D printing; and 3D printed.

The full phrase used was: (Implant\* OR scaffold\* OR prosthes?s) AND (titanium OR Ti) AND ("additive\* manufactur\*" OR "3D print\*") AND "mechanical properties" AND (porous OR porosity) AND "pore size" AND "elastic modulus" AND "fatigue".

#### *2.2. Study Selection*

Selected studies in the systematic literature search were limited to the following inclusion criteria: (1) peer-reviewed papers with full-text published within the last 20 years (2000 ± 2020); (2) empirical studies reporting the mechanical properties of Ti and Ti alloy porous scaffolds produced by AM for bone repair; (3) published in the English language; (4) the first 10 pages of the search results were assessed; and (5) the search results were sorted by relevance. From the systematic literature search in both the Science Direct and Google Scholar databases, a total of 5941 results were generated from which 83 articles were fully assessed, as presented in Table 1 and Figure 2.


**Table 1.** Search strategy, studies between 2000 and 2019.

#### *2.3. Data Extraction and Analysis*

The systematic literature search conducted in this study aimed to gather results of different studies regarding the characteristics of natural bone tissue, as well as the mechanical, geometrical, and dimensional properties of additively manufactured Ti porous implants with controlled porosity and/or pore size. The classification topics used in this study were pore size, pore shape, porosity, interconnectivity, multi-scaled, elastic modulus, compressive yield strength (σy), ultimate compressive strength (σtu), and fatigue strength. The references from the collected articles were systematically reviewed to identify further articles relevant to the subject. A full-text screening was performed by Y.D and D.M to avoid potential bias. A consensus meeting resolved any discrepancies between the reviewers.

Once all applicable literature had been identified, the extracted data were used to further extend the first step of the QbD system for fully porous Ti bone implants. Moreover, the data were also used to compare the different characteristics of Ti porous bone implants with natural bone tissue, and also to identify the most relevant characteristics that need to be imitated in the development of fully porous Ti bone implants.

**Figure 2.** Search strategy and selection of studies in the Google Scholar and ScienceDirect databases.

#### **3. Results**

The systematic search identified a total of 64 different studies with data relevant to additively manufactured (AMd) porous Ti implants and scaffolds, as presented in Tables 2 and 3. From the 64 studies identified, 14 studies were used to extract information related to bone structure and mechanical properties (Table 2). The remaining 50 studies provided detailed information in relation to different characteristics of porous Ti scaffolds fabricated by AM for bone implant purposes. A total of 807 experimental data from these studies was extracted, analysed, and categorised in the following categories: pore size, pore shape, porosity, multi-scaled porosity, elastic modulus, interconnectivity, yield strength, ultimate compressive strength, and fatigue strength, as shown in Table 3. However, from these nine categories multi-scaled porosity was excluded from the analysis due to insufficient data for further analysis. Therefore, a total of eight different implant features were selected due to availability of recorded data in scientific research articles. The selected eight features are: unit cell, porosity, pore size, interconnectivity, elastic modulus, compressive yield strength, ultimate compressive strength, and fatigue strength.


#### **Table 2.** Studies selected in the systematic search according to information related to bone properties.


**Table 3.** Studies selected in the systematic search according to information related to additively manufactured (AMd) porous Ti structures' mechanical, geometrical, and dimensional properties.

Through the systematic research performed in this study and by reviewing the medical device regulations from the Food and Drug Administration (FDA), a Quality Target Product Profile has been established including proposed values for the Ideal Quality Attributes for mechanical and dimensional properties of porous bone implants. These target properties are aimed for porous metal implant structures designed for load bearing implant applications. Following this, the results of the selected studies were compared and discussed, from a biomimetic point of view, with the characteristics of natural human bone to identify implants with properties similar or superior to human bone and current medical standards.

#### *3.1. QbD Step 1: Ideal Quality Target Product Profile*

The Quality Target Product Profile (QTPP) is critical for formulating the ideal features of a product considering both performance and safety. To direct the product development process, it is vital to understand user needs. Using QTPP, design failures can be identified early in the product development process to reduce costs and time. According to Martinez-Marquez et al. [36], the quality of bone implants should be defined from three perspectives: product-based, manufacturing-based, and user-based. In the context of fully porous Ti microstructures, the dimensions of quality considered most relevant are performance, features, reliability, conformance, durability, and perceived quality. The requirements relating to bone implants corresponding to these quality dimensions have been identified through studies of existing scientific research results.

#### *3.2. Bone Implant Requirements*

An additively manufactured bone implant must act as a stable scaffold that is biocompatible without causing inflammation or leaching material toxins into surrounding tissue. It requires a suitable surface that promotes cell adhesion and differentiation as well as provides a constant flow of cell nutrients and metabolic waste. This allows for bone tissue formation [81]. The scaffold material must have mechanical properties matching those of the surrounding tissues to avoid stress shielding and mechanical failure [81,82]. The bone pore size, geometry, interconnectivity, and porosity are microscopic features that make for the foundation of bone regeneration, cell growth, osteoconduction, and cell proliferation [74]. Designing implants with adequate pore dimensions allows for a constant flow of cell nutrients and waste. It also allows for sufficient connections to establish between the local bone area and the scaffold [74]. If the right attributes are chosen for the implant microstructure, it can mimic human bone's natural characteristics, which is the end goal of biomimetic implant design [79,80]. Vasireddi and Basu [140] completed a list of general requirements for 3D printed implants:


In similar research, Jabir et al. [141] described the main fundamental requirements for implants as biocompatibility, good manufacturability, geometric precision, appropriate design, biomechanical stability, resistance to implant wear, corrosion and aseptic loosening, bioactivity, and osteoconduction. Since the environment in the human body is highly corrosive and biomaterials are usually bioactive, the implant will interact with its environment after implantation [140]. The implant must therefore be designed with both useful functions and biological safety, providing utmost biocompatibility. The implant design must have a high degree of reproducibility, which will ensure faster and cheaper manufacturing as well as predictability and reliability.

Furthermore, the implant must be durable and of initial strength for safe handling during sterilisation, transport, and surgery, and to survive physical forces in vivo after implantation [142]. The implant will be subjected to constant load in the body, from walking and further strenuous

movements [127]. Its mechanical strength is vital for its viability, where the implant must last the entirety of its expected lifetime without defects or failure [127].

Taking all the above into consideration in conjunction with the ideal eight quality dimensions of AMd bone implants proposed by Martinez-Marquez at al. [74], we proposed seven ideal quality dimensions of porous internal architecture of Ti bone implants. These quality dimensions are based on three quality perspectives, namely product, manufacturing, and user-based, as presented in Table 4.


**Table 4.** The ideal quality dimensions of porous internal architecture of Ti bone implants.

#### *3.3. QbD Step 1.1: Ideal Quality Attributes*

The Ideal Quality Attributes (IQA) are the tissue or biological construct characteristics that must be mimicked to imitate the desired tissue biological architecture and functions. The IQA can be dimensional, physicochemical, mechanical, biological, or functional. However, if any technological and regulatory limitations exist it is important to consider that these IQA serve just as 'ideal models' to imitate even if it is not possible to achieve them. Therefore, the IQA can provide an ideal goal for the product development process in any tissue engineering project.

By studying the properties of natural human bone, we can find the different IQAs for bone implants [79,80]. As mentioned in Section 3.1, the implant must be of adequate strength while the elastic modulus must be similar to that of surrounding tissue to avoid damage [81]. The structure must be porous, using unit cells that allow for fluid movement and bone ingrowth, and the right porosity will reduce the implant stiffness. Interconnectivity completes the flow within the structure and facilitates bone ingrowth [82]. A combination of cancellous and cortical bone properties applied on bone implants can allow the implant to function like a natural bone, yet stronger without impairing surrounding native tissue. Another specific property of bone is its ability to heal itself when fractured [90,143]. The process involves cell migration, differentiation, and cellular proliferation [125]. The ability of natural bone to selfheal must be considered when designing a Ti implant, since the structure will not be able to do the same.

#### 3.3.1. Structure and Composition of Bone

To fully understand the microstructure requirements of fully porous Ti implants, one must first understand the properties of human bone. Bone is essentially an open-cell composite material of fibrous protein, collagen, and calcium phosphate crystals, with an intricate vascular system forming various structures and systems in a five-level hierarchically organised structure [84,144]. According to Rho, et al. [145], these five hierarchical levels are dimensional scales ranging from the macro to the sub-nano levels (Figure 3).

**Figure 3.** Hierarchical structural organization of bone: (**a**) cortical and cancellous bone; (**b**) osteons with Haversian systems; (**c**) lamellae; (**d**) collagen fiber assemblies of collagen fibrils; (**e**) bone mineral crystals, collagen molecules, and non-collagenous proteins. Image reproduced from Ref [81].

Bone's macrostructure is composed of cancellous and cortical bone, which are two different regions of bone with different density. Cortical bone forms the outside layer of the bone providing a strong, compact structure, leaving only 3–10% of the volume for its biological elements, such as blood vessels, osteocytes, erosion cavities, and canaliculi [81]. Cancellous bone forms the inside of the bone and is spacious and highly porous [81]. The pores are filled with bone marrow and the spacious architecture allows space for metabolic waste and nutrients to flow. Cancellous bone has an active metabolism and regenerates quicker compared with cortical bone [146]. By changing its density, cancellous bone can reorganise its structure depending on the stress direction [107]. These features cause cancellous bone mechanical properties to vary from bone to bone and change longitudinally [144].

The microstructure of bone ranges from 10 to 500 µm and it contains three major cavities. These are Haversian canals, osteocyte, lacunae, and canaliculi [88]. Cortical bone microstructure is composed of cylindrical structures called osteons with diameters ranging between 70 to 140 µm [81,87,92]. Along osteons' central axis are pores called haversian systems, with diameters ranging between 20 to 50 µm, containing blood vessels and nerves [92]. Cancellous bone micro-architecture is composed of irregular units called trabeculae which create its porous structure. The pore geometry and pore size of cancellous bone porous structure is critical for cell distribution and cell migration [81]. Cancellous bone is naturally stochastic, with random pore distribution of pores of different size.

The pores in cancellous bone are ellipsoidal in the natural direction of loading and are usually 300–600 µm wide [147]. The interconnectivity between the pores of cancellous bone is essential for nutrient and waste diffusion.

At the micro level are also located the three types of differentiated bone cells osteoclasts, osteoblasts, and osteocytes, as shown in Figure 4. Osteoclasts and osteoblasts are vital for the functions of developing and healing bone tissue [86]. Osteoclasts (Figure 4a) are the main cells responsible for resorption of old bone tissue. Osteoblasts (Figure 4b) are bone cells responsible for bone formation, remodelling, fracture healing (for which they are critical), and bone development [148]. Osteocytes (Figure 4c) are osteoblasts cells present inside mature bone and serve as mechanosensory cells to control the activity of osteoclasts and osteoblasts [149].

Bone's sub-microstructure, from 1 to 10 µm, is composed of lamellae which in cortical bone compromise the concentric layers of osteons and in cancellous bone lamellae forms the trabeculae volume [150,151]. The sub-nanostructure of bone, below a few hundred nanometres, is composed molecular constituent elements, such as collagen, non-collagenous organic proteins, and mineral. From a few hundred nanometres to 1 µm is bone's nanostructure comprising fibrillar collagen with embedded hydroxyapatite nanocrystals [152].

#### 3.3.2. Bone Mechanical Properties

Most bones in the body are load bearing and require high mechanical strength. Bone tissue is anisotropic and stronger in compression than in tension. The mechanical properties must be measured in two orthogonal directions: longitudinal, which is the natural loading direction, and transverse [86]. The mechanical strength of bone is complex to measure as it varies with age, health, activity, and position in the body [86]. Bone becomes stiffer and less ductile with age and its ability to heal decreases [155]. It is also likely to weaken with immobilisation, such as for a person with movement disabilities or limited physical activity.

The two different types of bone, cortical and cancellous, have completely different mechanical properties; therefore to specify the properties of bone as one material, both bone types must be considered. Cortical bone is highly dense an act like a shell that provides the greatest stiffness and resistance to bending. In contrast, the mechanical properties of cancellous bone are determined by its apparent density and trabecular architecture. The trabecular structure of cancellous bone is arranged accordingly to the stress distribution of load, as shown in Figure 5. As a result, the least material is used in the most strategic locations to carry the greatest loads with the least strain [156].

**Figure 4.** Coloured scanning electron micrographs of bone cells: (**a**) Activated osteoclast and resorption pit by kind permission of Timothy Arnett Ref [153]; (**b**) Osteoblast growing on a bone scaffold made of calcium oxide and silicon dioxide with added strontium and zinc by kind permission of Guocheng Wang from [154]; (**c**) Osteocytes embedded in the bone matrix with long cytoplasmic extensions reaching into the bone tissue, by kind permission of Kevin Mackenzie. Here, the minerals in the bone have been removed by embedding in resin and etching with perchloric acid. This reveals the spaces in the bone and the shape of the osteocyte cells.

**Figure 5.** Cortical and cancellous bone; image adapted from [157,158] by kind permission of Alan Boyde.

The mechanical properties of cortical and cancellous bone are difficult to measure, and tend to vary depending on bone orientation, test methods, mathematical formulas, and assumptions [146,151,159]. According to Chen and Thouas [57], the elastic modulus of cortical bone is approximately 11–21 GPa in the longitudinal direction and 5–13 GPa in the transverse direction. Another study suggested

18–22 GPa [147], whereas Wang et al. [81] suggested a range of 3–30 GPa. Cortical bone has a porosity of less than 10% [85]. The elastic modulus of cancellous bone is estimated to be 0.02–6 GPA and it has a high porosity of 50–90% [85,86].

To achieve adequate strength in bone implant design, one must understand the strength requirement of bone. Especially cortical bone must be considered, because the strength of such tissue is the minimum strength required by the implant. The compressive yield point of bone represents the threshold from where the structure accumulates irreversible deformation. Unlike metals such as steel, the yield point cannot be clearly distinguished, and it is rather associated with a continuous transition zone [160]. Furthermore, it has been proven that the compressive yield strength of bone varies depending on the anatomic site [161]. The challenges in determining the yield strength of cortical bone results in varying values in the literature. Researchers have estimated cortical bone to have a compressive yield strength of 133.6 ± 34.1 MPa [162]. The same characteristic was estimated to be 108–117 MPa by Yeni and Fyhrie [163]. Further studies have tested compressive yield strain using uniaxial compression and achieved 141.0 ± 5.0 [164], 111.0 ± 18.6 [165], 112.5 ± 9.5 [166], and 115.1 ± 16.4 MPa [167].

If the loading surpasses the yield point for bone, it will eventually reach the ultimate point. This point represents the ultimate compressive strength the bone can withstand until irreversible strains and damage occur. Past this point, macrocracks are formed and fracture occurs [160]. Unlike yield strength, the ultimate compressive strength of cortical bone can be exactly determined, using a stress– strain experiment [160]. However, due to bone properties differentiating, values vary in the literature. For example, Wang et al. [84] suggested that cortical bone has an ultimate compressive strength of 180–200 MPa, whereas Calori et al. [8] suggested a wider range of 130–290 MPa and Henkel et al. suggested 100–230 MPa [168]. Tables 5 and 6 summarise the mechanical and dimensional properties of natural bone considered in this study.



**Table 6.** Summary of mechanical properties of natural human bone.


*3.4. Comparison of Properties of Porous Ti Sca*ff*olds Fabricated by AM and Ideal Quality Attributes*

#### 3.4.1. Unit Cell Geometry

Metamaterials can be rationally designed by changing their geometry at the microscale of the constituting unit cells of the porous structure. In this systematic search, a total of 169 porous scaffolds were identified as a rationally designed and fabricated for bone implant applications. It was found that there are three preferred strategies for fabricating bio-metamaterials: beam-based, sheet-based, and including irregular porous structures [172].

According to Figure 6, the preferred design approach was beam-based, which represent 74.6% of the total scaffolds produced in the selected studies. The beam-based bio-metamaterials' micro-architecture is composed of a lattice structure created using unit cells based on platonic solids, Archimedean solids, prisms and anti-prisms, and Archimedean duals [173–177] to mimic bone porous macro structure and mechanical properties such as modulus of elasticity. Nevertheless, the biological performance of bio-metamaterials created with beam-based geometries is limited by their inaccurate description of complex natural shapes due to their straight edges and sharp turns [109].

**Figure 6.** Distribution of experimental studies on beam-based, triply periodic minimal surfaces (TPMS), and random based geometries for bone implant applications.

The unit cells of sheet-based geometries, on the other hand, are based on triply periodic minimal surfaces (TPMS), which are present in different organisms and cellular structures [178]. Therefore, it is no surprising that the second most used design strategy identified in this systematic search was sheet-based representing 20.1% of the total scaffolds produced, from which all used a TPMS as unit cell. Bio-metamaterials based on TPMS can mimic the various properties of bone to an unprecedented level of multi-physics detail in terms of mechanical properties and transport properties [79,172]. Moreover, the bone-mimicking mean surface curvature of zero of TPMS eliminates the effect of stress concentrators at nodal points [120].

In the case of irregular porous structures, are created in a random way to generate irregular porous structures that mimic trabecular bone geometry and mechanical properties [179]. These irregular structures are generated using the Voronoi and Delaunay tessellation methods. Irregular structures have been found to further enhance scaffold's permeability and bone ingrowth compared with porous structures designed with regular unit cells [180]. According to our results only 5.3% of the studies used a randomised design approach to create irregular porous structures to mimic trabecular bone. This result was surprising considering that irregular porous structures were the first additively manufactured coatings used for orthopaedic implants in the medical industry. However, unlike random porous scaffolds, the great advantage of using regular repeating arrays of unit cells made of beam or sheet-based geometries is that they allow the creation of metamaterials with properties that can accurately be predicted [181]. This explain why these design strategies are preferred in research.

From all the different possible unit cells that can be used to produce metamaterials, a total of 17 types of unit cell were used by the selected studies, as presented in Figure 7. According to Figure 7, the beam-based diamond unit cell (59 studies) was the most used, followed by the cubic (18 studies), and the gyroid TPMS (17 studies). These results correlate with the opinion of different experts who have stated that the diamond unit cell is the most studied for the development of metamaterials due to its biomimetic mechanical properties [121,182]. The high mechanical properties and self-supporting properties of the diamond unit cell are due to its unique geometrical arrangement, where one node is tetrahedrally surrounded by four other nodes coming from the crystal structure of the diamond

crystal [113], as shown in Figure 8a. Moreover, this arrangement gives 48 symmetry elements to the diamond structure, making this unit cell invariant to different symmetry operations such as translations, reflections, rotations, and inversion [183]. Similarly, extensive research for bone regeneration has been performed to study the cubic unit cell (Figure 8b). The research interest in the cubic unit cell is because it is based on one of the simplest and easy–to–manufacture platonic solids thanks to its struts at an angle of 90◦ [62]. In the case of porous metamaterials based on the gyroid TPMS (Figure 8c), they have been found to exhibit similar topology to human trabecular bone, and also superior mechanical properties compared with metamaterials based on other types of TPMS [120]. For example, according to Yang et al. [184] metamaterials based on the gyroid TPMS have a more homogeneous stress distribution, which can provide equal mechanical stimulation to bone cells.

**Figure 7.** Comparison of the number of experimental studies of each unit used for bone regeneration identified in the reviewed articles.

**Figure 8.** Some examples of unit cells with their corresponding meta-biomaterial scaffold below: (**a**) diamond beam-based unit cell; (**b**) cubic beam-based unit cell; (**c**) Gyroid triply periodic minimal surface based (TPMS-based) unit cell; (**d**) Diamond TPMS-based unit cell; and (**e**) Voronoi (top) and Delaunay (bottom) irregular porous structures.

In the case of random or stochastic structures, Kou et al. [185] suggested that scaffolds based on this structures are more realistic; that is, they look more like natural bone with random non-uniform pore distribution and pore size [185]. Such structures are believed to provide benefits such as improved mechanical properties, including strength, fluid dynamics, surface area, and surface-to-weight ratio. They combine advantages of small and large pores without necessarily decreasing the mechanical strength or reducing the bone in-growth to levels that are inappropriate in application [186]. Figure 8 presents the five most representative unit cells identified in this study.

#### 3.4.2. Porosity

It is known that the degree of micro-porosity in bone implants directly affects their biological and mechanical properties. The porosity of natural bone is crucial for vascularisation, diffusion of cell nutrients and metabolic waste, and cell migration [187], and in a similar way it is important for metal bone implants. Moreover, several studies have considered porosity as the main parameter affecting stiffness and strength of porous biomaterials. Increased porosity reduces the strength of the implant [187,188]. As a result, porous metallic biomaterials are used as coatings in many medical applications, but more recently porous biomaterials have been proposed for load-bearing applications beyond surface coatings [138,142]. Ti and Ti alloys are commonly used for load-bearing implant applications due to their relatively low elastic modulus fatigue resistance, high strength to weight ratio, and corrosion resistance [82,94]. However, bulk Ti and Ti alloys do not completely match all the mechanical properties of natural bone such as modulus of elasticity. Therefore, it is a need of the hour to accomplish specific mechanical properties for Ti or Ti-based alloys by controlling the porosity and pore characteristics for customised implants [131]. However, the ideal porosity for medical implants seems controversial in the literature [181].

In this systematic review a total of 49 articles out of 50 recorded porosity of various degrees, as shown in Figure 9. For example, Stamp et al. [186] recommend using a porosity above 65% in medical implants whereas Ghanaati et al. [189] found that vascularisation increased in vivo when reducing the porosity from 80 to 40%. Sarhadi et al. [190] and Schiefer et al. [191] have recommended using a porosity of approximately 50%. According to Will et al. [192], the porosity that best promotes vascularization in porous scaffolds is 40–60%. Pattanayak et al. [193] manufactured porous Ti implants and reported an increase in compressive strength from 35 MPa to 120 MPa when reducing the porosity from 75 to 55%. Murr et al. reduced the porosity from 88 to 59% with an increase in stiffness from 0.58 GPa to 1.03 GPa [131]. As mentioned, natural cancellous bone has a porosity of 50–90% [85,86].

Zou et al. [102] designed three implants with similar pore size and shape. By reducing the porosity from 72 to 53% they achieved a compressive strength of 200 MPa and Young's modulus of 4.3 GPa, instead of a compressive strength of 60 MPa and Young's modulus of 2 GPa. Even though both implants achieved a stiffness close to that of human bone, only the implant with 53% porosity achieved a compressive strength greater than human bone. Hence, the implant with 72% porosity would not

qualify as a load-bearing bone implant. These results, along with previous research by Pattanyak et al. and Murr et al., confirm the influence of porosity on mechanical properties [193]. They also show that the implant porosity may be adjusted within limits to increase strength and adjust bone stiffness, which varies for each patient's characteristics.

From a medical regulatory perspective, the U.S Food and Drug Administration (FDA) only approves implant porosities of 30–70% for porous coatings on solid Ti implants [194]. The range is relatively large but can be used to one's advantage since both the elastic modulus and strength of the implant can be adjusted by adjusting the porosity. Implants with porosity outside of this range do not comply with FDA regulations and cannot enter the market. It is known that medical regulations, especially for implantable medical devices, are based on strong scientific evidence. Therefore, it is vital to adhere to these regulations when designing bone implants.

Taking into consideration the medical regulations for porous implants, we selected a porosity range of 30–70% as the IQA for porous metal implants to identify studies in the systematic search that fabricated Ti scaffolds with porosity values within this range. From the 49 articles that recorded porosity of various degrees, a total of 167 results were extracted and compared with the selected IQA porosity range. The results of this comparison are presented in Figure 9. According to our results, 56.6% of the porous scaffolds studied in the 49 selected articles had porosity values within the acceptable porosity range (30–70%) required to satisfy medical regulations such as the FDA. By contrast, a total of 26 studies explored the properties of porous scaffolds with porosity values above the acceptable porosity range representing 37.3% of the total results extracted in this systematic literature review. There were several reasons for these studies to explore porosity levels higher than 70%. For example, Zhang et al. [100] and Amin Yavari et al. [106] fabricated different porous structures with various porosities to explore their mechanical properties and deformation mechanisms. Moreover, porous Ti scaffolds with high levels of porosity can serve as storage for mesenchymal stem cells to facilitate bone tissue regrowth, and also to improve cell oxygenation and nutrition [100]. On the other hand, Ti porous scaffolds with porosity levels lower than 30% can provide similar mechanical properties to cortical bone [98,100].

#### 3.4.3. Macropore Size

Since macro pore size is directly related to the strength, porosity, and stiffness of the implant, it is an important property for implant design [195]. Pore size has a profound effect on the behaviour of osteogenic cells even in an organ culture system [196]. The implant's macro porosity determines whether bone cells can successfully penetrate and grow within the structure, and many studies have discussed the influence of pore size on the biological properties of implants [197]. Furthermore, several studies have shown that a minimum pore size of 100 µm is required for vascularization and bone ingrowth, but pores larger than 100 µm increase bone in-growth by allowing improved vascularization and oxygenation [86,91,193]. A minimum macropore size limit of 100 µm is supported by further research as vascular penetration has been found to be restricted in smaller pore interconnections [85,168,188,192].

Studies have found that pores greater than 300 µm are required for vascularisation and bone ingrowth [86,168]. Tang et al. [188] found that 200–350 µm is the optimal macropore size, and various studies have found that bone ingrowth is less likely to occur beyond 400 µm [90,195,198]. However, research that used pore sizes of 300, 600, and 900 µm in porous Ti scaffolds found that those with macropores sizes of 600 and 900 µm had much higher bone ingrowth compared with the scaffolds with 300 µm pores [101]. Bose et al. [85] suggested that all macropore sizes between 100 and 600 µm are osteoconductive. Fukuda et al. [199] experienced greater results in 500 and 600 µm pores compared with 900 and 1200 µm pores. Xue et al.'s [98] results showed that macropore sizes in the range of 100–600 µm possess the optimum ability for cell growth into the pore structure of porous titanium.

According to the FDA regulations, macropore sizes of 100–1000 um are approved for coatings for Ti implants [194]. Large macropores have a smaller surface area than do small pores, decreasing the cell attachment on the implant [86]. However, large macropores increase scaffold vascularisation, which is vital for supplying oxygen and nutrients to the tissue as well as osteoblast proliferation and migration [80], but they decrease the mechanical strength of the material. The limit of how much the macropore size can be increased while maintaining sufficient mechanical strength depends on both the material and the processing conditions. Therefore, regulatory guidelines for surface coating may be misleading for fully macroporous implants, since the strength-to-weight ratio differs between a porous and solid structure. Since the porosity decreases the strength of the implant, and large pore sizes decrease the strength of the internal architecture, large pore sizes must be avoided to increase the structure's strength. A more defined pore size range is therefore sought [86].

In this study, the macroporosity used in different studies was explored. It was found that no consensus currently exists on what upper limit to macropore size that is ideal, but somewhat of a consensus on the lower limit exists (100 µm). According to FDA regulations, porous implants should have macropore sizes between 100 and 1000 µm [194]. It has further been found that macropores start to lose their osteogenic functionality when larger than 500–600 µm [8,16,85,98,199]. Considering that a fully porous structure is weaker than a solid structure, and that high strength is vital for implants, it can be assumed that there is no need to design a structure with pores larger than what is needed to cater for all functions within the implant.

These findings made us choose a macropore size range of 100–600 µm as the IQA for porous metal implants to identify the studies in the systematic search that fabricated Ti scaffolds with pore size values within this range. From the 42 articles that recorded pore size of various degrees, a total of 144 results were extracted and compared with the selected IQA pore size range, as seen in Figure 10. According to our results, 51.4% of the results of all studies had a macropore size within 100–600 µm. It was further noted that 86.8% of the experimental results of all studies had a macropore size within the FDA recommended range of 100–1000 µm. From these results, we could infer that most of the research studies identified through the systematic search somewhat considered the macroporosity range required to satisfy medical regulations.

By contrast, a total of 68 results out 144 showed pore size values above the acceptable macropore size range representing 48.6% of the total results extracted in this systematic literature review. There were several reasons for these studies to explore pore sizes above 600 µm. For example, the FDA approves macropore sizes between 100 and 1000 µm [194]. Hara et al. [96] tested four porous structures with different macropore size to explore their mechanical properties. Taniguchi tested 300, 600, and 900 µm pore sizes and found that the structures with 600 and 900 µm pore size exhibited higher bone ingrowth. Large macropores increase scaffold vascularisation, which is vital for supplying oxygen and nutrients to the tissue as well as osteoblast proliferation and migration [80]; however, larger pores also have a

smaller surface area compared with small pores, decreasing the cell attachment on the implant [86]. Larger macropores also result in higher porosity, and this reduces the strength of the implant [187,188].

The ideal macropore size for bone implants is controversial and undefined, and according to Otsuki et al., a reason for the varied pore size data may be that the interconnectivity of pores was not considered [198]. Macropore size regulations have been developed for the first generation of porous implants, which use a single-scaled porous network with repeated, equally sized, and shaped pores. However, bone grows in a naturally random structure with pores of different sizes, shapes, and directions similar to the structure of a sponge [130,155]. The use of a multiscale porous scaffold that combines smaller and larger pore sizes within the same structure is a recent strategy to optimise the internal architecture of implants [142,188]. This method combines advantages of both small and large macropores without decreasing the strength or reducing the bone in-growth to levels that are inappropriate in application [188]. According to our results, a total of 8 out of 50 studies used some sort of multiscale pore approach, but it was observed that the design method varied, and that the researchers failed to provide the percentage of the total structure that used each pore size. It was further observed that a multiscale porous structure occurred in some implants where the manufacturing of a single-scaled structure resulted in varying pore sizes due to unprecise manufacturing tolerances.

#### 3.4.4. Pore Inter Connectivity

According to our results, 46% of the collected studies registered pore interconnectivity. Interestingly, all of these studies designed their porous scaffolds with an interconnectivity of 100%. The pores in a porous bone implant must be interconnected to ensure movement and the supply of necessary nutrients through ingrowth of tissue and bone [200]. Interconnected pores tend to facilitate the flow of fluids and biological cells through the structure which is essential for bone tissue formation [185]. According to Nyberg et al. [201] the integration of artificial material tissue with native tissue can be improved by interconnected pores. Tang et al. [188] suggested that an increased pore interconnectivity increases the number and size of blood vessels formed in scaffolds. The interconnectivity is also a critical factor for ensuring that all cells within the structure are within a 200 µm range from a blood supply to provide transfer of nutrients and oxygen [202]. According to the FDA's recommendations for porous metal coatings, pores in such structures must be interconnected [194]. Although this requirement is for surface coatings, it also indicates the importance of an interconnected porosity for fully porous implants. In the systematic review, it appears as though a vast majority of studies had indicated the importance of an interconnected porosity. Therefore, to guarantee all processes and fluid movements necessary for tissue and bone ingrowth, the selected IQA for pore interconnectivity would ideally be 100%.

#### 3.4.5. Elastic Modulus

It was observed in this systematic review that an elastic modulus is a property commonly reported in AM porous scaffolds studies (by 89% of all studies). A controlled modulus of elasticity has proved to be critical in prostheses and scaffolds to avoid stress shielding [81,82]. Stress shielding occurs when there is a stiffness mismatch between the implant and surrounding bone, and it can cause inflammation and the need for revision surgery [197]. Ti and common implant Ti alloys have an elastic modulus of roughly 100–120 GPa [81,84,138,197]. A reduced modulus is necessary to avoid stress shielding and can be achieved by designing implants with a porous structure [90].

Defining an ideal specific modulus of elasticity for porous bone implants is not practical because Since the mechanical properties of human bone, especially the elastic modulus, change drastically with factors, such as age, physical activity, and health. For example, femoral bone specimens from patients aged 3, 5, and 35 years had an elastic modulus of 7, 12.8, and 16.7 GPa, respectively, indicating a dramatic change with age [90]. As previously shown in Table 6, the elastic modulus of human bone varies in the literature. Chen and Thouas [57] estimated the elastic modulus of cortical bone to be approximately 11–21 GPa in the longitudinal direction, whereas Lee et al. [147] suggested 18–22 GPa. Wang et al. [81] suggested a wider range of 3–30 GPa. These findings indicate that the stiffness of an implant may need to be adjusted specifically for the person it is intended for, and that the target value for the elastic modulus may be specific to each patient. Therefore, it is more practical to think that for the design of porous scaffolds, an ideal target area exists for the modulus of elasticity. Based on this, the IQA for elastic modulus is proposed to be 3–30 GPa for fully porous Ti implants.

Figures 11 and 12 show the elastic modulus that was reported in the reviewed articles and these values were compared with the proposed IQA. From the extracted of elastic modulus results, 55.5% reached the target area of 3–30 GPa. These implants achieved an elastic modulus within the range of natural bone and would therefore eliminate risk of stress shielding. By contrast, 40% of the results exhibited an elastic modulus below 3 GPa, and only 3.6% of the results reported an elastic modulus higher than 30 GPa. These results clearly demonstrated that most studies are aiming towards a modulus of elasticity closer to the bone modulus.

**Figure 11.** Elastic modulus of porous metamaterials based on triply periodic minimal surfaces (TPMS) compared with the Ideal Quality Attribute (IQA) target zone.

The elastic modulus of metals such as titanium and its alloys naturally have a much higher elastic modulus compared with bone [81]. However, research shows that the elastic modulus of metals can be readily adjusted by modifying their porosity. Porous metals with a low modulus of elasticity correspond to high levels of porosity. For example, Wang et al. [84] explored five types of porous

structures using the same material (TiNbZr) and pore size (550 µm) but ranging porosity. His results revealed that four of the implants with porosities ranging from 42% to 69% all achieved an elastic modulus within the approved range of 3–30 GPa; however, the implant with the highest porosity (74%) achieved the lowest elastic modulus of 1.6 GPa. Similarly, Li et al. [203] used a porosity of 91% resulting in a low elastic modulus of 0.8 GPa. Furthermore, Chen et al. [204] received an elastic modulus of 44.4 GPa for a porous titanium structure using 30% porosity but by increasing the porosity to 40% the elastic modulus was reduced to 24.7 GPa.

Using the data obtained through the systematic literature search we calculated two multiple linear regressions to predict the modulus of elasticity of beam and TPMS-based AMd Ti scaffolds. The regression model used the independent variables of pore size, relative density (porosity), and the interaction of pore size–porosity. According to our results, a regression equation was found for beam-based AMd Ti scaffolds (F(3,75) = 54.139, *p* < 0.0001), with an R<sup>2</sup> adj of 0.671, as shown in Figure 13a. The residuals of the multiple linear regression are randomly scattered around the centre line of zero with no obvious pattern. The predicted compressive yield strength of beam-based scaffolds is equal to 27.738 – 0.078 (pore size) – 27.417 (porosity) + 0.0689 (pore size\*porosity), where pore size is coded or measured in µm, and relative density expressed as porosity as a percentage. The beam-based scaffolds' modulus of elasticity decreased 0.078 MPa for each µm, 27.417 MPa per 1% of porosity increment, and increased 0.0689 MPa for the interaction pore size\*porosity. Both pore size (*p* < 0.0005) and porosity (*p* < 0.0001) were significant predictors of beam-based scaffolds' modulus of elasticity, including the interaction between pore size and porosity (*p* < 0.0001).

**Figure 13.** Overall predicted model of elastic modulus actual versus an elastic modulus predicted (**a**) beam-based scaffolds, and (**b**) TPMS-based scaffolds.

In the case of the multiple linear regression of TPMS based AMd Ti scaffolds a significant regression equation was also found (F(3,28) = 4.897, *p* < 0.0073), with an R<sup>2</sup> adj of 0.273, as shown in Figure 13b. The residuals of the multiple linear regression are randomly scattered around the centre line of zero with no obvious pattern. The scaffolds' predicted modulus of elasticity is equal to 0.008 − 0.002 (pore size)–2.342 (porosity) + 0.002 (pore size\*porosity). The TPMS based scaffolds' modulus of elasticity decreased 0.002 MPa for each µm, 2.342 MPa per 1% of porosity increment, and 0.002 MPa for the interaction pore size\*porosity. Pore size was a significant predictor of TPMS-based scaffolds' modulus of elasticity with a *p*-values < 0.0563. However, porosity was not a significant predictor with *p*-values < 0.553 and 0.843, respectively. Moreover, no interaction between pore size and porosity was found regarding to modulus of the elasticity.

#### 3.4.6. Compressive Yield Strength

For an adequate functioning of any load-bearing implant, it is vital that its design withstand the required forces and loading cycles. Mechanical strength is one of the implant's most crucial features for avoiding implant failure. To withstand the loads of daily activities, load-bearing implants must have at least the same yield strength as the bone that they replace [81]. The yield point of bone represents the threshold from where the structure accumulates irreversible deformation. Unlike bulk metals such as steel, the yield point of bones cannot be clearly distinguished; it is rather associated with a continuous transition zone [160]. Strain beyond the yield point will deform the structure beyond its point of resilience causing material damage, usually occurring as micro-cracks [205]. Bone tissue has evolved to mainly support compressive stress [206,207]. Bone is 30% weaker under tensile stress, and 65% weaker under shear stress [208]. Therefore, load-bearing implant scaffolds require a high compressive strength to prevent fractures and improve functional stability [209]. The compressive yield strength of cortical bone varies in the literature. As previously shown in Table 6, the compressive yield strength of cortical bone varies approximately between 90 and 170 MPa. To replace like with like, using a biomimetic approach for comparison purposes, a minimum and a maximum compressive yield strength of 90 MPa and 170 MPa were selected as the IQA for fully porous Ti implants.

The systematic search identified that 37 out of 50 studies recorded compressive yield strength, from which a total of 133 experimental results were extracted and compared, as shown in Figures 14 and 15. Figure 14 presents the results of studies using TPMS structures and Figure 15 presents the results of studies using porous beam-based metamaterials. Both comparisons show high numbers of studies resulting in a compressive yield strength below 90 MPa. A total of 55.7% of all results had a compressive yield strength below the defined IQA target and 25% of the studies achieved a compressive yield strength within the bone region. On the other hand, only 19% of the extracted experimental results had a strength above the bone region. Such implants would have strengths similar to or higher than cortical bone and are expected to not experience permanent deformation caused by the expected bone compressive loading conditions in the human body.

**Figure 14.** Compressive yield strength of porous metamaterials based on triply periodic minimal surfaces (TPMS) compared with the Ideal Quality Attribute (IQA) target zone.

*Materials* **2020**, *13*, 4794

**Figure 15.** Compressive yield strength of porous beam-based metamaterials compared with the Ideal Quality Attribute (IQA) target zone.

Decreased strength of a porous implant can result from high porosity and large pore sizes, [81,188]. For example, Zhang et al. [113] fabricated porous scaffolds based on the TPMS diamond unit cell with a wide range of compressive yield strengths from 36 MPa to 140 MPa just by varying the scaffolds' porosity and maintaining the pore size constant. The type of unit cell used to design porous scaffolds can also drastically change it mechanical properties. For example, Zhao et al. [116] fabricated four porous scaffolds with the same pore size and similar porosities using two different unit cells (tetrahedron and octahedron). However, the scaffolds based on the octahedron unit cell registered almost double the compressive strength compared with those based on the tetrahedron unit cell [116]. The compressive yield strength of porous metals can also be enhanced by gradually changing the porosity level along the radial direction of the scaffold. This was demonstrated by Zhang et al. [100], who reported functionally graded porous scaffolds based on the diamond unit cell with superior comprehensive mechanical properties to the biomaterials with uniform porous structures.

Using the data obtained through the systematic literature search, we calculated two multiple linear regressions to predict compressive yield strength based on pore size, porosity, and the interaction of size–porosity for beam-based and TPMS-based AMd Ti scaffolds, respectively.

According to our results, a regression equation was found for beam-based AMd Ti scaffolds (F(3,75) = 31.452, *p* < 0.0001), with an R<sup>2</sup> adj of 0.539, as shown in Figure 16a. The residuals of the multiple linear regression are randomly scattered around the centre line of zero, with no obvious pattern. The scaffolds' predicted compressive yield strength is equal to 380.557 − 0.075(pore size)–350.828 (porosity) + 0.557 (pore size\*porosity), where pore size is coded or measured in µm, and porosity is measured as a percentage. The beam-based scaffolds' compressive yield strength decreased 0.075 MPa for each µm, 350.828 MPa per 1% of porosity increment, and increased 0.557 MPa for the interaction of pore size\*porosity. Both pore size (*p* < 0.0378), porosity (*p* < 0.001), and the interaction between pore size and porosity (*p* < 0.0048) were significant predictors of beam-based scaffolds' compressive yield strength. The interaction between pore size and porosity was found to be significant with a *p*-value < 0.0048.

b b

**Figure 16.** Overall predicted model of compressive yield strength Actual versus compressive yield strength Predicted of (**a**) beam-based scaffolds, and (**b**) TPMS-based scaffolds.

Regarding the multiple linear regression of TPMS-based AMd Ti scaffolds, a significant regression equation was also found (F(3,27) = 65.547, *p* < 0.0001), with an R <sup>2</sup> adj of 0.872, as shown in Figure 16b. The residuals of the multiple linear regression are randomly scattered around the centre line of zero, with no obvious pattern. The scaffolds' predicted compressive yield strength is equal to 524.780 − 0.008(pore size)–625.266 (porosity) + 0.183 (pore size\*porosity). The TPMS-based scaffolds' compressive yield strength decreased 0.008 MPa for each µm, 625.266 MPa per each 1% of porosity increment, and increased 0.183 MPa for the interaction of pore size\*porosity. Porosity was a significant predictor of TPMS based scaffolds' compressive yield strength with *p*-values < 0.0001. However, pore size and the interaction between pore size and porosity were non-significant, with *p*-values < 0.7018 and 0.3260, respectively.

#### 3.4.7. Ultimate Compressive Strength

If loading surpasses the yield point of bone, it will eventually reach the ultimate point. This point represents the maximum compressive strength that a material can withstand without irreversible strains and damage occurring. Past this point, macrocracks are formed and fracture occurs [160]. Bone implants in load-bearing applications must withstand high stress within the body, to a degree where no permanent deformation occurs during the load that the implant is expected to be exposed to. Hence, controlled ultimate compressive strength is a crucial property to study in bone implant research. Natural bone is estimated to have an ultimate compressive strength of 180–200 MPa [84]. However, results vary in the literature. For example, Calori et al. [8] suggested a more widespread range of 130–290 MPa, whereas Henkel et al. suggested 100–230 MPa [168]. To replace like with like using a biometric approach, bone implants should have an ultimate compressive strength similar to that of bone [8]. Taking into consideration the compressive yield strength suggested previously and the three results presented in Table 5, the proposed IQA region for the ultimate compressive strength is between 180 MPa and 290 MPa.

In this systematic search a total of 60 experimental results of ultimate compressive strength from 19 different studies were extracted. Figures 17 and 18 show the ultimate compressive strength of the different studies compared with the defined IQA target of between 180 MPa and 290MPa. Figure 17 corresponds to experimental results of porous scaffolds composed of TPMS unit cells compared with the IQA target. According to Figure 17, only one study with three different experimental results measured the ultimate strength of porous scaffolds based on TPMS unit cells. In this study by Yanez et al. [120], three ultimate compressive strengths of 17, 47.5, and 83.5 MPa were achieved using the gyroid unit cell. Dramatic improvement in the ultimate compressive strength of the scaffold was achieved. This improvement in mechanical properties was possible by slightly changing the gyroid unit cell into an elongated gyroid. However, none of the experimental results obtained by Yanez

et al. [120] were able to reach the minimum IQA ultimate compressive strength proposed in this study. The low strength of Yanez at al.'s [120] samples can be attributed to their high porosity values which ranged between 75% and 90%. This increased the stress concentration, reducing the ultimate compressive strength.

**Figure 17.** Ultimate compressive strength of porous metamaterials based on triply periodic minimal surfaces (TPMS) compared with the Ideal Quality Attribute (IQA) target zone.

**Figure 18.** Ultimate compressive strength of porous beam-based metamaterials compared with the Ideal Quality Attribute (IQA) target zone.

Figure 18 is composed of 19 different studies on beam-based unit cells where 57 experimental results are compared with the IQA target. According to Figure 18, 10 experimental results reached higher ultimate compressive strengths than bone. The highest ultimate compressive strength (830 MPa) recorded was achieved with a scaffold based on the diamond unit cell by Zhang et al. [100]. From all results recorded, a total of 16.7% achieved an ultimate compressive strength above the proposed IQA ultimate compressive strength; 10% had similar ultimate compressive strengths to bone; and 73.3% had lower ultimate compressive strengths than bone. The scaffolds that did not fulfil the required IQA ultimate compressive strength would risk fracturing due to macrocracks occurring during high loads. Porosity within a structure has been proven to decrease the strength of a structure [81,188], which may explain the high number of scaffolds with low ultimate strength.

The goal of designing a porous structure with enough porosity and pore size without diminishing strength is a difficult task, and as researchers aim to create highly porous structures with low elastic modulus, many structures experienced low ultimate strength. Attar et al. [111] manufactured a porous titanium structure by SLM with rectangular pores and 17% porosity and achieved an ultimate compressive strength of 747 MPa. Using the same material, manufacturing method, and unit cell, a different structure with 37% porosity achieved an ultimate strength of 235 MPa. In a similar manner, Chen et al. [204] designed three structures of the same material and manufacturing method. Using porosities of 30%, 40%, and 50%, the ultimate compressive strengths recorded were 524, 301.7, and 120.3 MPa, respectively, indicating that increased porosity reduces the strength of the structure.

#### 3.4.8. Fatigue Strength

During normal daily activities, load-bearing implants experience just a fraction of the material's ultimate stress [209–211]. However, after years of use, the high cyclic loading to which load-bearing implants are subjected eventually leads to the accumulation of small stresses, causing progressive and localised material damage that results in implant failure [212]. For instance, one of the most critical mechanical properties for load-bearing implants is fatigue strength. However, fatigue strength is the most difficult mechanical property to determine [213].

The required fatigue resistance of a load bearing implant and its components mainly depends on their cyclic loading conditions and the required life span. For example, it is estimated that lower limb prostheses are subjected to up to 2 million gait cycles per year [214], and in the case of orthodontic prostheses these can reach up to 300,000 loading cycles per year [213]. Therefore, a large variety of medical standards exist for testing fatigue strength. Some of these fatigue tests differ depending on the type of load applied such as tension–tension, compression–compression, and tension–compression. In the case of load-bearing bones, their loading conditions in real-life activities are complex [135]. However, bone is mainly loaded in compression [206,207]. Therefore, to test the fatigue life of metamaterials for bone implant applications, compression fatigue tests are preferred due to the simplicity of the test setups [215].

Regarding the number of cycles that load bearing implants and their components need to have tested for fatigue strength, all the different medical standards agree that such products need to have a fatigue life within the high-cycle fatigue region (N > 10<sup>4</sup> cycles). For example, the ASTM standard F2777 – 16 recommends testing tibial inserts' endurance and deformation under high flexion with a minimum number of cycles of 2.2 <sup>×</sup> <sup>10</sup><sup>5</sup> , and in the case of dental implants they are typically tested up to 5 million cycles [213,216]. Nevertheless, for a component of a load-bearing implant to have at least 25 years of life span [217], the highest number of cycles that must be tested is 10<sup>7</sup> cycles [213,216]. Taking into consideration current medical standards for load-bearing implants, the high-cycle fatigue region between 10<sup>4</sup> cycles and 10<sup>7</sup> cycles was selected as the IQA fatigue life for porous titanium metamaterials for bone regeneration.

Using the selected high-cycle fatigue region, this systematic literature search identified a total of 13 different studies on fatigue resistance, among which 11 studies performed compression–compression fatigue tests. Then, for comparison purposes, a total of 51 experimental results were extracted and compared. Moreover, to facilitate the comparison of the results of the studies, they were classified according to the type of unit cell used to produce porous structures as beam and TPMS-based as resented in Figures 19–21. According to our results, the TPMS porous structures that withstood the highest stresses at the high-cycle fatigue region were achieved by Bobbert et al. [79]. The primitive TPMS structure presented the highest stress within the high-cycle fatigue region, with 232 MPa at <sup>3</sup> <sup>×</sup> <sup>10</sup><sup>4</sup> cycles, as shown in Figure 19. The TPMS porous structures that were able to withstand the second and third highest stresses within the high-cycle fatigue region were the I-WP and diamond structures, with 227 MPa at 3 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles and 204 MPa at 3 <sup>×</sup> <sup>10</sup><sup>6</sup> cycles as shown in Figures 19 and 20. Remarkably, the primitive TPMS structure was the only one to pass the 10<sup>7</sup> threshold with 80 MPa at <sup>3</sup> <sup>×</sup> <sup>10</sup><sup>7</sup> cycles, as presented in Figure 20.

**Figure 19.** Comparison of the results of different studies on the fatigue strength of porous metamaterials based on triply periodic minimal surfaces (TPMS) within the high-cycle fatigue region between 10<sup>4</sup> and 10<sup>5</sup> cycles.

**Figure 20.** Comparison of the results of different studies on the fatigue strength of porous metamaterials based on triply periodic minimal surfaces (TPMS) within the high-cycle fatigue region between 10<sup>6</sup> and 10<sup>7</sup> cycles.

**Figure 21.** Comparison of the results of different studies on the fatigue strength of porous beam-based metamaterials within the high-cycle fatigue region between 10<sup>6</sup> and 10<sup>7</sup> cycles.

In the case of beam-based metamaterials the study performed by Zhao et al. [116] achieved the highest fatigue strength with 130 MPa at 10<sup>6</sup> cycles using a lattice structure based on the octahedron unit cell, as shown in Figure 21. In second and third place with the highest fatigue strength are the tetrahedron and the cubic porous lattice structures with 90 MPa at 1 <sup>×</sup> <sup>10</sup><sup>6</sup> cycles and 90 MPa at 1 <sup>×</sup> <sup>10</sup><sup>6</sup> cycles fabricated by Amin Yavari et al. [106] and Zhao et al. [116], as presented in Figure 21.

According to the results of this systematic search it could be seen that TPMS structures provide superior fatigue strength to porous bio-metamaterials compared with beam-based unit cells. Moreover, in terms of fatigue resistance, it was identified that the primitive, I-WP, and diamond TPMS provided the best performance, whereas the octahedron, cubic, and tetrahedron are the best-performing lattice unit cells. However, it is crucial to note that several factors can affect the fatigue life of additively manufactured Ti metamaterials. Some of these factors are residual stresses and stress concentrators caused by high surface roughness and manufacturing defects [218]. Moreover, the fatigue strength of bulk materials significantly degrades when they are in porous form or when voids and pores are developed during fabrication [219]. In the case of additively manufactured components, it has been proven that their fatigue strength is extremely sensitive to localised and nonuniform heat and uncontrolled cooling cycles during fabrication [220,221].

#### *3.5. Discussion and Summary of the Findings*

Table 7 presents a summary of proposed IQA target values for porous Ti and Ti alloy bone implants aimed at load-bearing applications. These values are based on scientifically supported values found in human bone research, federal regulations (FDA), and research articles on porous Ti implants manufactured using AM between 2000 and 2020. These properties are part of the first step of the QbD framework of "Define the Quality Target Product Profile." The IQAs are necessary for a systematic and qualitative design approach. These properties will provide benchmark guidance to facilitate future research on porous bone implant design.


**Table 7.** Summary of Ideal Quality Attributes proposed in this study.

An IQA target zone for the porosity of porous Ti implants has been proposed as 30–70% based on results found in research articles as well as in current FDA regulations. We found that 56.6% of all studies in this review achieved a porosity within this range. AM was found to produce porous structures with highly controlled porosity. Numerous studies as well as FDA regulations have discussed the importance of using porous structures for tissue ingrowth, which have numerous advantages to non-porous implants. Whereas porous coatings are used in some instances, it was observed in this study that many researchers believe in using a fully porous structure to achieve a biomimetic structure mimicking natural bone. All data used extracted in this systematic literature review came from studies that used scaffolds with constant porosity. However whether a repeated lattice structure throughout the entire structure is sufficient or a biomimetic "sponge-like" structure with irregular, elongated pores should be used has not yet been confirmed in research; nevertheless implants using a biomimetic design approach replicating natural bone received attention in recent research [86]. Since bone has a random and stochastic structure with pores of different sizes, a multiscale porous structure with pores of different sizes deserves further research. Multiscale porous scaffold structures have been proven to perform better than one-dimensional structures [142]. Nonetheless, to properly compare implants with a multiscale structure, pore distribution and size must be recorded.

In the case of scaffolds' pore size, the IQA was proposed to be 100–600 µm based on numerous performed research trials. Whereas the FDA requirement is a pore size between 100–1000 µm, a more defined pore size range of 100–600 µm is supported by research due to numerous research trials having experienced reduced bone ingrowth in pores larger than 600 µm. Strut thickness is the thickness of the pore walls within a porous structure. It may have a substantial impact on implant mechanical properties [90] and can serve as a unit cell's characteristic to compare different mechanical properties in relation to it. However, it was found that struct thickness is not commonly reported in the literature.

Research has proven that the elastic modulus of bone changes drastically with age and that patient customisation is a necessity in specific cases. As a result, in this study, an IQA for elastic modulus was proposed to be 3–30 GPa based on values from research studies. According to data analysed in this study a porous structure has a significant influence on mechanical properties in Ti-based implants. Therefore, a porous structure can be altered to provide an elastic modulus comparable to that of natural bone and be adjusted to modify the elastic modulus according to the patient's age and health condition. This will further reduce the risk of stress shielding between the implant and surrounding tissue.

As reported by most studies it was found in the regression analysis that the elastic modulus of Ti and its alloys is directly influenced by the implant porosity, and that by increasing the porosity the elastic modulus is increased and vice versa. A total of 55.5% of the studies in this review recorded an elastic modulus within this range. Regarding the effect of unit cells on the modulus of elasticity of porous structures, it was found that the studies that used beam-based unit cells covered a wider range of modulus of elasticity values than did those than used TPMS-based unit cells (Figure 22a). Moreover, the studies that used beam-based unit cells also obtained modulus of elasticity values within the whole range of human bone Young's modulus, as shown in Figure 22a.

**Figure 22.** Comparison in logarithmic scale between mechanical properties of bone, beam-based, and triply periodic minimal surfaces (TPMS): (**a**) modulus of elasticity; (**b**) compressive yield strength; and (**c**) ultimate compressive strength.

Compressive yield and ultimate strength were studied. The compressive strength values for cortical bone were found to vary in the literature, which is why a range of values was selected to represent this bone characteristic. The IQA for compressive yield strength was proposed to be a minimum of 90 MPa and the IQA target for ultimate compressive strength a minimum of 180 MPa. A total of 44.1% of the studies achieved a compressive yield strength above 90 MPa, whereas only

26.7% achieved an ultimate compressive strength above 180 MPa. Both TPMS-based and beam-based scaffolds provided a wide range of compressive yield strengths covering the whole range of human bone compressive yield strength, as presented in Figure 22b. However, according to Figure 22c only beam-based scaffolds provided an adequate range of ultimate compressive yield strength values that include bone properties. The narrow range of ultimate compressive yield strength values of TPMS-based scaffolds can be explained by the lower number of studies that have addressed this mechanical with this type of unit cell. This indicates that a need exists for TPMS-based scaffold research to report the ultimate compressive yield strength.

Just as with the elastic modulus, the strength of implants was found to be directly affected by scaffold porosity. Increased porosity was found to rather dramatically decrease the strength of the implant as demonstrated by the regression model performed with the results of several different studies. Moreover, it was found that large pores are directly related to lowered strength and elastic modulus, and that several studies have struggled to produce highly porous structures with strength higher than or close to natural bone. Therefore, to ensure the adequate strength of implants, pores larger than required (>600 µm) and porosity values higher than necessary (>70%) are recommended to be avoided as these have been found to likely result in low-strength structures.

#### **4. Conclusions**

This systematic literature review presented an overview of the reported properties in research studies of fully porous Ti bone implants manufactured with AM received in the last two decades. The study focused on implants made of Ti and Ti alloy since they are considered ideal biomaterials for load-bearing applications. This study followed a QbD approach and includes the first step of the QbD system that defines the QTPP for properties relating to the porous internal architecture of fully porous Ti implants designed for load bearing applications. IQA, part of the QbD system, have been proposed supported by properties of natural human bone, governmental regulations, and scientific research relevant to bone implants. Unit cell geometry, porosity, elastic modulus, compressive yield strength, ultimate compressive yield strength, and compressive fatigue strength were systematically reviewed and benchmarked against the proposed IQA.

This study found that many implant geometrical, mechanical, and dimensional characteristics were directly related to each other. Scaffolds' pore size influences the porosity of the structure, and the porosity alters the elastic modulus as well as the strength of the implant. The unit cell geometry was also found to directly affect the Young's modulus and strength of porous scaffolds, and naturally would also impact the structures' interconnectivity. Moreover, by using ranges rather than set values where possible, such as for elastic modulus, porosity, and pore size, there is a flexibility to the design that allows the implant to be adjusted to its purpose and patient. Such design flexibility is necessary as bone properties vary with patients' age and anatomic site.

Despite the variety of scaffold characteristics reviewed in this study, future systematic literature searches should also focus on other properties, such as fluid dynamics, surface finish (topography), creep, and hardness, as well as surface coatings. Moreover, it should be considered that bone ingrowth may modify implants' mechanical properties; hence, they may be dependent on the level of bone ingrowth. Therefore, it is important to measure the changes in strength and stiffness in metallic bone scaffolds after bone ingrowth has occurred. Since implant location may determine the importance of each implant property, future studies would ideally find IQA for different implant locations within the human body. For example, mechanical features are important to study in load bearing implants, whereas fluid dynamics and vascularisation might be more important features for facial implants that require good aesthetic results and that are placed close to fragile tissue and nerves.

This study was possible due to the abundance of data available in research articles. However, to further develop an effective QbD engineering strategy for the development of specific bone implants it is important to identify the degree of importance of each bone scaffold characteristic according to the implant's future location within the human body. This can be identified by following

the second step in the QbD system, namely Critical Quality Attributes (CQA), as detailed described in previous work [222]. CQA are product characteristics that must fall within specific limits to comply with the quality standards defined in the QTPP. They can be identified through prior knowledge and experimental data from systematic research based on scientific and risk management rationale that considers regulatory and business requirements [223]. Each step of the QbD system must be further developed to facilitate the strategic, qualitative development of bone implants and increase the rate of successful research studies.

Overall, Ti and Ti alloy porous bone implants are well underway to achieving the ideal properties that will fully allow them to replace natural bone. With the help of the QbD system, consistent and qualitative design of medical implant devices can be achieved.

**Author Contributions:** Conceptualisation: D.M.-M., Y.D., and R.A.S.; Formal analysis: D.M.-M. and Y.D.; Funding acquisition: R.A.S.; Investigation: D.M.-M. and Y.D.; Methodology: D.M.-M., Y.D., S.S., and R.A.S.; Project administration: D.M.-M. and R.A.S.; Resources: D.M.-M. and Y.D.; Supervision: D.M.-M., S.S., and R.A.S.; Visualisation: D.M.-M. and Y.D.; Writing—original draft: D.M.-M. and Y.D.; Writing—review and editing: D.M.-M., S.S., and R.A.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** This work was supported by Griffith School of Engineering and Built Environment.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Bone Morphogenetic Proteins, Carriers, and Animal Models in the Development of Novel Bone Regenerative Therapies**

**Nikola Stokovic 1 , Natalia Ivanjko 1 , Drazen Maticic 2 , Frank P. Luyten <sup>3</sup> and Slobodan Vukicevic 1, \***


**Abstract:** Bone morphogenetic proteins (BMPs) possess a unique ability to induce new bone formation. Numerous preclinical studies have been conducted to develop novel, BMP-based osteoinductive devices for the management of segmental bone defects and posterolateral spinal fusion (PLF). In these studies, BMPs were combined with a broad range of carriers (natural and synthetic polymers, inorganic materials, and their combinations) and tested in various models in mice, rats, rabbits, dogs, sheep, and non-human primates. In this review, we summarized bone regeneration strategies and animal models used for the initial, intermediate, and advanced evaluation of promising therapeutical solutions for new bone formation and repair. Moreover, in this review, we discuss basic aspects to be considered when planning animal experiments, including anatomical characteristics of the species used, appropriate BMP dosing, duration of the observation period, and sample size.

**Keywords:** animal model; bone fracture; bone healing; posterolateral spinal fusion; regenerative medicine; bone morphogenetic proteins

#### **1. Introduction**

Bone tissue possesses unique regenerative properties, and bone fractures regularly heal under physiological conditions. However, large segmental bone defects resulting from severe trauma or extensive tumor resection cannot be restored by endogenous selfrepair mechanisms, decrease quality of life, and may sometimes lead to limb amputation. Indeed, the management of large segmental defects is one of the most challenging issues in orthopedic medicine, typically due to the biologically hampered microenvironment [1,2]. The standard of care for the healing of large bone defects requires the use of an autologous bone graft (ABG), which is usually harvested from the iliac crest. ABG is also used as a gold standard to achieve spinal fusions, including posterolateral spinal fusion (PLF). PLF is a commonly performed surgical procedure used for the treatment of degenerative diseases of the spine, including degenerative disc disease, spondylolisthesis, spinal instability, and symptomatic scoliosis [3–6].

However, ABG possesses several disadvantages, including a limited amount of bone that might be harvested, the potential transfer of contaminating agents, acute and chronic pain, skin scarring, and deformity at the donor site [4,7]. In addition, the use of ABG increases the blood loss, duration, and cost of surgical procedures. Therefore, there remains an imminent need for the development of novel bone regeneration strategies to enrich or replace ABG. Among these, osteoinductive devices are under investigation for clinical use in PLF and healing of large segmental long bone defects.

In the last few decades, numerous preclinical studies using animal models have been conducted to test novel bone bridging or fusion strategies [8,9]. The principles of

**Citation:** Stokovic, N.; Ivanjko, N.; Maticic, D.; Luyten, F.P.; Vukicevic, S. Bone Morphogenetic Proteins, Carriers, and Animal Models in the Development of Novel Bone Regenerative Therapies. *Materials* **2021**, *14*, 3513. https://doi.org/ 10.3390/ma14133513

Academic Editor: Antonio Boccaccio

Received: 18 May 2021 Accepted: 21 June 2021 Published: 24 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

the rational use of animal models in the evaluation of novel bone regenerative therapies have been previously described [8]. Hence, we further investigated the use of animal models in the development of osteoinductive therapies of large segmental bone defects and PLF procedures, in particular the selection of a proper anatomical model, treatment dose, observation period, and sample size. We specifically analyzed published in vivo studies looking into the development of bone morphogenetic protein (BMP)-based bone inducing implants.

#### **2. Bone Regeneration by Bone Morphogenetic Protein Devices**

BMPs are well-known osteoinductive molecules, required and sufficient for ectopic bone induction, and powerful agents for the restoration of large orthotopic bone defects [10]. BMP2 is the most widely used osteoinductive BMP in preclinical testing, and it is a part of an osteoinductive device (Infuse™, Medtronic, Dublin, Ireland), currently approved for anterior lumbar interbody fusion (ALIF), acute tibial fractures, and maxillofacial reconstructions [11–15]. However, BMPs have been used off- label in various spinal indications, including cervical spine fusion, posterior lumbar interbody fusion (PLIF), transforaminal lumbar interbody fusion (TLIF), posterolateral spinal fusion (PLF), and thoracolumbar fusions [16,17]. Reported side effects in patients included implant displacement, infection, swelling of the adjacent tissue and dysphagia, formations of seroma, radiculitis and nerve root compression, ectopic bone formation, osteolysis, and retrograde ejaculation [11,12,16–21]. These side effects eventually resulted from the use of supraphysiological doses as registered BMP2-based devices contain 4–12 mg recombinant protein, while the human body contains only a total of 2 mg of BMPs [22].

Other commonly used osteoinductive BMPs are BMP7, which is no longer in clinical use, and more recently, BMP6 [23]. We demonstrated that BMP6 appears to be superior to BMP2 and BMP7 in promoting osteoblast differentiation in vitro and inducing bone formation in vivo [23,24]. The superiority of BMP6 may arise from its resistance to noggin inhibition and affinity across the BMP type I receptors. Therefore, BMP6-based devices are expected to be more efficacious at lower doses compared to BMP2 and BMP7.

BMPs require a carrier/delivery system that will sustain the BMP concentration and allow prolonged BMP release [25–28]. Moreover, the ideal BMP carrier should be biocompatible, enable vascular and cellular infiltration, resist compression, and define the contours of the resulting bone [25,26,29]. BMP carriers can be divided into four major groups: natural polymers, synthetic polymers, inorganic materials, and combinations between these groups [25,26,30].

Natural polymers include collagen, hyaluronic acid, gelatin, fibrin, chitosan, alginate, and silk and have been extensively evaluated in preclinical studies [6,31–38]. The advantages of natural polymers are biocompatibility, biodegradability, and resorbability in the physiological environment [25–27]. The most commonly used is bovine tendon collagen which delivers BMP2 in the clinically approved Infuse™ device. However, collagen has significant disadvantages, including a low affinity for BMPs, immunogenicity due to its animal origin, and weak biomechanical properties resulting in compression by surrounding tissues [25–27].

Biocompatible and biodegradable synthetic polymers, such as polylactic acid (PLA), polyglycolic acid (PGA), poly(D, L-lactide-co-glycolide) (PLGA), polyethylene glycol (PEG), poly-E-caprolactone (PCL), and polypropylene fumarate (PPF), as well as their block polymers have been evaluated as potential BMP carriers to overcome the disadvantages of natural polymers, including immunogenicity and disease transmission risk [39–42]. They are also moldable into highly porous three-dimensional scaffolds, linearly oriented scaffolds, fibers, sheets, blocks, or microspheres [26]. Apart from these advantages, synthetic polymers decrease local pH as a result of acidic breakdown byproducts, have poor clearance, cause bulk degradation, and cause chronic inflammation associated with high-molecular weight polymers, resulting in substantial disadvantages [26]. They have also been tested with other potentially osteogenic molecules, such as PGE2 and PGE4 prostaglandin receptor analogs [43], and materials such as calcium silicate (CaSi) and dicalcium phosphate dihydrate (DCPD) [44–47].

Inorganic materials as potential BMP carriers include calcium phosphate (CaP) ceramics, calcium phosphate and calcium sulfate cement, and bioglass [2,5,29,32,38,42,48–72]. The most commonly used inorganic preclinical materials are CaP ceramics, further subdivided into hydroxyapatite (HA), tricalcium phosphate (TCP), and biphasic calcium phosphate (BCP) containing both HA and TCP at various ratios. We have recently shown that the chemical composition of ceramics does not affect the amount of newly formed bone induced by the osteoinductive device [73,74]. However, HA and TCP significantly differ in resorbability (HA is very stable, while TCP is more resorbable), which would eventually result in different residual ceramic volumes. The resorbability might be adjusted by varying HA/TCP ratios in BCP ceramics [75]. Moreover, CaP ceramics might be formulated into particles or blocks in a broad range of sizes and geometrical shapes while porosity, pore size, and interconnectivity are adjusted during the sintering process [73,75,76]. We demonstrated that particle size affects the volume of newly formed bone; smaller particles (74–420 µm) combined with rhBMP6 resulted in higher bone volume than larger particles (1000–4000 µm) [73]. Another important determinant of ceramics is the pore size since pores from 300 to 400 µm promoted the formation of the largest bone volume [51].

The fourth group of BMP carriers are composites of the aforementioned materials which have been introduced to overcome the encountered limitations of a single component. The most typical combinations are composites containing either natural or synthetic polymers with CaP ceramics [39,77–85]. In these combinations, ceramics increase the biomechanical properties of the implants and are used to address compressibility issues. Less frequent, natural, and synthetic polymers might be combined.

We have recently developed an autologous bone graft substitute (ABGS) comprised of BMP6 delivered within an autologous blood coagulum to which a compression-resistant matrix, such as allograft or synthetic ceramics, can be added [22,73,74,76,86–92]. Moreover, the volume of newly induced bone increased with the elevation of the CRM amount, which might be attributed to the enlargement in an overall surface area [73].

#### **3. Animal Models**

Animal models are routinely used in the development of novel bone regenerative therapies [8]. Models might be categorized according to the species (mouse, rat, rabbit, sheep, non-human primate) and tested indication (ectopic model, critical-size defect, PLF). In this review, we suggested classification based on the stage of preclinical development, namely as initial, intermediate, and advanced testing of osteoinductive devices (Figure 1). Initial testing includes rodent ectopic and rodent critical-size defect models for rapid comparison of different osteoinductive responses. Intermediate evaluation includes adequate rabbit models (segmental defect and PLF model), while advanced testing uses canine, sheep, and non-human primates as a final step before clinical trials.

#### *3.1. Initial Evaluation in Rodents*

#### 3.1.1. Ectopic Models

Rodent ectopic models have been extensively used for the initial evaluation of novel osteoinductive therapies. They might be also used for investigating the biology of ectopic bone induction and the formation of a bone organ or ossicle, including bone and bone marrow [31,32,39,48–57,71,73,76,86,87,93–105]. Rodent ectopic models (Tables 1 and 2) are further subdivided according to the species (mouse, rat) and the implantation site (subcutaneous or intramuscular). Implantation under the skin (Figure 2A–D) or into the muscle does not affect the bone formation outcome, and the bone formation occurs in the first two weeks following implantation of an osteoinductive device [76,86,87]. The later time points are needed for the evaluation of the bone longevity and maintenance of the ectopic bone structure. Molecular and cellular events during the cascade of bone formation can be evaluated using microCT/nanoCT and histological analyses. Immunohistochemistry,

flow cytometry, gene profile microarrays, and single-cell RNA sequencing are among other analytical techniques used for unraveling the mechanism of ectopic bone formation.

**Figure 1.** Classification of animal models from mouse to NHP based on the stage of preclinical development, namely as initial, intermediate, and advanced evaluation of osteoinductive devices. E—ectopic, SD—segmental defect, PLF posterolateral spinal fusion models, and L—level of lumbar transverse processes.


#### **Table 1.** Mouse ectopic and bone defect models.


#### **Table 2.** Rat ectopic and bone defect models.

**Figure 2.** Histological sections and gross anatomy of newly formed bone induced by rhBMP6 in a rat subcutaneous assay (**A**–**D**); rabbit (**E**,**F**) and sheep (**G**,**H**) posterolateral spinal fusion model. (**A**,**B**) On day 7 following implantation, endochondral bone formation occurs at the peripheral site of ABGS, while 14 days (**C**,**D**) after implantation, newly formed bone is present throughout the implant containing rhBMP6 lyophilized on allograft mixed with ABC. Gross anatomy of newly formed bone between transverse processes in rabbit (**E**) and sheep (**G**) PLF model. Histological sections through fusion mass in rabbit (**F**) and sheep (**H**) PLF model. Histological sections were processed undecalcified and stained by Von Kossa (**A**,**B**), Goldner (**C**,**D**,**F**), and hematoxylin and eosin (**H**). (Modified from [74,88], respectively.)

Murine models are initially used to unravel the potential mechanism of action of various signaling pathways and genes or proteins in bone induction or enhancement. However, due to limited translation of mouse to human bone regeneration and disease outcome, the rat is therefore a more suitable model when testing functional outcomes [106–108].

#### 3.1.2. Bone Defect Models

Mouse or rat bone defects are the initial orthotopic models to evaluate the osteoinductive properties of novel therapies and the osseointegration of newly formed bone with adjacent native bone. There are two main bone defect models in rodents: a calvarial critical-

size defect and segmental femoral defect. In the calvarial critical-size defect, circular bone defects are created in the mouse (3–5 mm) [109–118] and rat (4–8 mm) [29,55,58–60,66,77] calvaria (Figure 3A; Tables 1 and 2), while segmental defects of the long bone are typically created in the femur, both in mice (2–3 mm) [119–121] and rats (6–10 mm) [5,33–35,67,78,79] (Figure 3B; Tables 1 and 2) and filled with tested osteoinductive material. The development of a reproducible non-union model in the mouse is demanding, and, in contrast to rat non-union models, mouse non-union models are sparse [122]. The main shortcoming of this model is a relatively small defect size compared to clinically relevant proportions. Moreover, it is difficult to obtain a full stabilization of the fracture, therefore resulting in increased callus formation. Methods of evaluation include analyses of radiological images (CT/microCT), histological and histomorphometric analyses, and biomechanical testing, which might be conducted at the end of the observation period (Figure 4) [5,29,33–35,55,58–60,66,67,77–79,109–121,123].

**Figure 3.** First row: segmental defect (SD) models performed on various bones depending on the species; (**A**) calvarial, (**B**) femoral, (**C**) ulnar (or radial), and (**D**) tibial defect. Second row: posterolateral lumbar fusion (PLF) is conducted between adjacent transverse processes; the figure shows differences in the anatomy of rabbit (**E**), sheep (**F**), and human (**G**) lumbar spine.

**Figure 4.** BMP doses (µg), observation period (weeks), and sample size (*n*) used in the initial phase —rodent models. Data is presented through box and whisker plots in which dots represent all individual values from studies listed in Tables 1 and 2. E—ectopic model, C—calvarial, and F—femoral defect.

#### *3.2. Intermediate Evaluation in Rabbits*

#### 3.2.1. Segmental Defect Model

Potential therapeutical solutions for segmental bone defect restoration have been extensively evaluated in rabbits [36,41,42,61,68–70,80–84,87,124–126] (Table 3), and a defect can be created in the femur, radius, or ulna (Figure 3B,C). Regardless of the chosen anatomical site, the typical defect size is 15–20 mm. In published work, the defect was bridged with a broad range of delivery systems containing up to 150 µg of BMPs. The observation

period was typically 6–12 weeks. Few studies evaluated bone formation at earlier time points (2 and 4 weeks) or for a prolonged period (24 weeks) (Figure 5, 1st column).

**Figure 5.** BMP doses (µg), observation period (weeks), and sample size (*n*) used in the intermediate and advanced phases, which include rabbit, dog, sheep, and non-human primate models. Data are presented through box and whisker plots in which dots represent all individual values from studies listed in Tables 3–6. SEG—segmental defect models; PLF—posterolateral spinal fusion model.


#### **Table 3.** Rabbit segmental defect model.

#### **Table 4.** Rabbit PLF model.



#### **Table 4.** *Cont*.

**Table 5.** Dog and sheep segmental defect model.


**Table 6.** Sheep and non-humane primate PLF models.



**Table 6.** *Cont*.

#### 3.2.2. Posterolateral Spinal Fusion (PLF) Model

Rabbit is the most commonly used species for the evaluation of the efficacy and safety of promising therapeutical solutions for achieving PLF [5,6,37,38,40,62–64,74,85,86,127–129]. The transverse processes of lumbar vertebrae are exposed, and an osteoinductive device is implanted bilaterally between adjacent transverse processes (L4-L5 or L5-L6) [127]. Transverse processes should be decorticated before the implantation to promote osseointegration of newly formed bone with native bone [86]. In the majority of previous studies, the BMP dose was up to 1000 µg and was delivered on various carriers (Table 4). The spinal fusion outcome was evaluated 6 weeks following surgery, and the majority of rabbit PLF studies had an observation period of fewer than 10 weeks. Few studies had a prolonged observation period (>10 weeks), but later time points might be important to determine the survival and long-term maintenance of newly induced bone [6,86], which is clinically relevant in patients undergoing PLF surgery (Figure 5, 1st column).

Methods of evaluation in rabbit segmental defect and PLF models are similar and include segmental mobility testing, radiological methods (x-ray and CT/microCT), histological (Figure 2F), histomorphometric analyses, and biomechanical testing [5,6,37,38,40,62– 64,74,85,86,127–129].

#### *3.3. Advanced Evaluation of Bone Regeneration Therapies*

#### 3.3.1. Dog and Sheep Segmental Defect Model

Dog and sheep segmental defect models are used for advanced evaluation of bone regeneration therapies. In dogs, the defect (20–25 mm) is created in the radius or ulna (Figure 3C) [130–135]. Applied doses of BMPs were in the range between 100 and 650 µg, which is higher compared to the rabbit model. Moreover, the typical observation period (12–24 weeks) was also prolonged compared to the rabbit model (Figure 5, 2nd and 3rd columns).

Tibial segmental bone defects in sheep (Figure 3D) were recently developed to evaluate novel bone regeneration therapies in conditions mimicking the size and biology of segmental bone defects in the clinics [2,136–140]. Moreover, there are two subtypes of this model: a fresh defect (FD) and biologically exhausted defect (BED), the latter mimicking a patient with a non-union [2]. Following the creation of a large defect (30 or 45 mm) in the sheep tibia in the FD model, a polymethyl-methacrylate spacer is inserted to induce the formation of the Masquelet membrane. Six weeks following the creation of the defect, an osteoinductive device was inserted after the removal of the spacer (FD model). In the BED model, the defect is in the first instance left untreated leading to a non-union. Subsequently, debridement of the non-union or fibrotic tissue ingrowth (BED model) is performed, followed by implantation of a spacer for 6 weeks, and then, finally, after removal of the spacer, an implant is inserted. BMP doses applied in this model ranged from 344 to 3800 µg, while the typical observation period was up to 16 weeks [2]. Although the osteoinductive device containing BMP6 on a carrier achieved bridging in FD (30 mm), it was found that larger, biologically exhausted defects appear to require a cell-based implant together with BMP to achieve proper clinically relevant bridging (Table 5) [2]. Importantly, defects were mechanically well stabilized with a circular external fixator according to the Ilizarov technique.

#### 3.3.2. Sheep PLF Model

The sheep PLF model is highly translatable to clinics because the size of the lumbar vertebrae of the sheep is comparable to humans. However, only a few preclinical studies have been conducted on this model [3,65,88] (Table 6). Sheep PLF may be conducted at a single level or as a multisegmental procedure. Moreover, it may be performed with or without instrumentation [88]. The observation period and applied BMP doses in this model were typically significantly longer/higher than in studies on small animals: the follow-up period was up to 6 months with a BMP amount up to 10 mg (Figure 5, 3rd column). Methods of evaluation included X-ray monitoring, microCT evaluation, histological analyses (Figure 2H), and biomechanical testing [3,65,88].

#### 3.3.3. Non-human Primate (NHP) PLF Model

Non-human primates are the most similar animal species to humans, both anatomically and genetically. However, only a few studies were conducted using NHP PLF [14,63,141] (Table 6), primarily due to ethical and economic reasons. In these studies, the goal was to achieve a single-level fusion between adjacent lumbar transverse processes, which are anatomically similar to humans. The applied BMP2 doses (3–12 mg), as well as observation period (24 weeks), were comparable to the sheep PLF model (Figure 5, 4th column).

#### *3.4. Anatomical Characteristics of the Species*

#### 3.4.1. Segmental Bone Defect

The general anatomy of long bones (Figure 3B–D) of species discussed in this review is similar, and obviously, the greatest difference among them is size. Differences in the bone size reflect the segmental defect created in each species. The length of the long bone segmental defect in mice (3 mm) or rats (5–8 mm) is small and created in the femur, the largest bone in rodents [5,33–35,67,78,79,119–121,123,142]. Long bones are significantly larger in rabbits/dogs, and segmental defects (15–20 mm in rabbits and 20–25 mm in dogs) are created in the femur, radius, or ulna [36,41,42,61,68–70,80–84,87,124–126,130–135]. Sheep segmental bone defects are usually created in the tibia and, due to their size (30–45 mm), are considered clinically relevant because the defect size compares to those in patients [2].

#### 3.4.2. PLF

Posterolateral spinal fusion (PLF) in preclinical studies is conducted in the lumbar portion of the spine. Although the basic anatomical features of lumbar vertebrae are similar among species discussed in this review, they differ in size and proportions of the different parts of the vertebrae. Rabbits (Figure 3E) and sheep (Figure 3F) have long transverse processes compared to the size of the vertebral body, while humans (Figure 3G), as an adaptation to erect posture and bipedal locomotion, have large bodies and short transverse processes. Importantly, transverse processes in rabbits are slanted and oriented anteriorly (Figure 2E or Figure 3E). On the other hand, the transverse processes in sheep (Figure 2G or Figure 3F) and humans (Figure 3G) are horizontal. The distances between the transverse processes are relatively short in rabbits (20–30 mm), while they are comparable in sheep and humans (40–50 mm).

#### **4. Appropriate Bone Morphogenetic Proteins Dosing**

The selection of the appropriate dose for each indication is one of the most challenging steps in the design of experiments. In this review, we compared doses used in various models in mice, rats, rabbits, sheep, and NHP. In the mouse and rat ectopic models, BMP doses were typically up to 25 µg [31,32,39,48–57,71,73,76,86,87] (Figure 4, 1st column, 1st row) per implant, while they were increased in the rat bone defects (up to 50 µg) [5,29,33–35,55,58–60,66,67,77–79,123] (Figure 4, 2nd column, 1st row). BMP doses used in rabbits were up to 150 µg in segmental bone defects [36,41,42,61,68–70,80– 84,87,124–126,143] (Figure 5, 1st column, 1st row) and up to 300 µg in rabbit PLF procedures [5,6,37,38,40,62–64,74,85,86,127–129] (Figure 5, 1st column, 1st row). Tested BMP

doses in dogs were higher and were typically in the range between 100 and 650 µg [130–135] (Figure 5, 2nd column, 1st row). Moreover, in the sheep and NHP models, BMP doses were significantly higher: in the sheep, the doses were between 500 µg and up to 4 mg [3,65,88] (Figure 5, 3rd column, 1st row), while in the NHP, the doses were up to 12 mg [14,63,141] (Figure 5, 4th column, 1st row).

#### **5. Duration of the Observation Period**

The bone induction process is significantly faster in small animals compared to higher species animals and humans. Therefore, observation periods were significantly shorter in rodent ectopic and bone defect models than in studies on sheep and non-human primates (Figures 4 and 5, 2nd row). The average observation period in mouse and rat ectopic models was 3–4 weeks [31,32,39,48–57,71,73,76,86,87] (Figure 4, 1st and 2nd columns, 2nd row). However, depending on the purpose of the study, observation periods in these studies might vary from a few days (studies on the mechanism of bone induction) to several months (bone longevity). Typical observation periods in the rat calvarial defect and femoral segmental defect models are slightly prolonged and last 5 and 10 weeks, respectively [5,29,33–35,55,58–60,66,77–79,123] (Figure 4, 2nd column, 2nd row). The observation period in segmental defect studies was up to 12 weeks in rabbits. [36,41,42,61,68–70,80–84,87,124–126,130–135,143] (Figure 5, 1st column. 2nd row). On the other hand, spinal fusion outcome in the rabbit PLF model was typically evaluated after 6 weeks [5,6,37,38,40,62–64,74,85,86,127–129] (Figure 5, 1st column, 2nd row). However, to study longevity or resorbability of compression-resistant matrices, the follow-up period might be prolonged. As expected, a longer observation period in dogs and sheep was up to 12 months [3,65,88,130–135] (Figure 5, 2nd and 3rd columns, 2nd row), while in NHP studies, it was 6 months [14,63,141] (Figure 5, 4th column, 2nd row).

#### **6. Sample Size**

Defining an appropriate sample size is a prerequisite for obtaining valid conclusions from each study. Moreover, the appropriate size of the sample is affected by several parameters, including experimental design and purpose of the study as well as expected differences among experimental groups. The sample size in the majority of reviewed studies here was 5–10 per group regardless of the animal species or model (Figures 4 and 5, 3rd row). Moreover, there is a consensus in published work that the minimal number of animals per experimental group is four. However, a few animals might die during surgery or follow-up periods due to reasons non-related to the tested osteoinductive therapy; therefore, at least five animals per experimental group should be included.

#### **7. Study Outcomes**

In Tables 1–6, it was not possible to describe the study outcomes due to non-comparable scoring grades for healing or spinal fusion experiments. The prerequisite in reporting the outcome of bone defect and spinal fusion studies is a clearly described success rate as the percentage of successfully rebridged defects or fused spine segments, respectively. Moreover, the method (radiological images, mobility testing) used to determine rebridgment/fusion should be clearly described. Surprisingly, in a large number of published studies, the success rate was not explicitly described. Several authors used their own scoring grades instead of standardized binary outcomes (successful or unsuccessful rebridgment/fusion). However, even when the binary outcome was used, the determination of successful rebridgment/fusion differed among authors. For example, a few authors determined success rate only on X-ray images without microCT, histology, and biomechanical testing. We suggest that successful rebridgment/fusion should also be determined with microCT, histological sectioning, and biomechanical testing.

The experimental outcome of osteoinductive therapies using rodent ectopic models should be determined by microCT and histology. MicroCT analyses provide information on newly formed bone volume expressed as bone volume (BV) or bone volume/tissue volume ratio (BV/TV). Additionally, if the tested osteoinductive device contains ceramics, microCT analyses might be used to determine the amount of residual ceramic matrix. Moreover, microCT analyses allow the determination of structural properties of newly formed bone by calculating trabecular parameters (trabecular number, trabecular thickness, trabecular separation). The structural properties of newly induced bone should also be analysed by histology and histomorphometry to determine the volume of the bone and remaining carrier/matrix.

#### **8. Conclusions**

Due to the large socioeconomic burden of degenerative diseases of the spine and segmental defects of long bones, there is an imminent need for the development of novel osteoinductive therapeutic solutions [1,22]. However, until now, none of the osteoinductive devices have been approved for use in PLF and large segmental defects in patients. A broad range of bone regeneration strategies have been proposed and tested in different animal models. A vast majority of these studies have been conducted in rats and rabbits, leading only to the initial and intermediate steps of preclinical testing, and despite claiming positive results, only a few have been further tested in sheep and NHP models. Infuse™, a BMP2-containing osteoinductive device, has been approved for use in ALIF and acute tibial fractures but has also been used in various off-label indications. However, numerous side effects related to high BMP dose and a large release from the bovine collagen as a carrier have been reported. Therefore, there is a need for an osteoinductive device that would be efficacious at lower doses of BMP delivered on a carrier with a prolonged BMP release. There is some hope that novel engineered BMPs or innovative delivery systems for BMPs may reduce the required therapeutic doses. A novel ABGS containing rhBMP6 within autologous blood coagulum was evaluated in preclinical studies, and in exploratory clinical trials (high tibial osteotomy, distal radial fracture, and posterolateral interbody fusion), it was proven safe and efficacious at relatively low BMP6 doses [73,74,76,86–88,91,92].

**Author Contributions:** Conceptualization, N.S. and S.V.; Writing—Original Draft Preparation, N.S. and S.V.; Writing—Review and Editing, N.S., N.I., D.M., F.P.L., and S.V.; Visualization, N.I. and N.S.; Supervision, S.V., F.P.L., and D.M.; Funding Acquisition, S.V., F.P.L., and D.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This program was funded by the FP7 Health Program (FP7/2007-2013) under grant agreement HEALTH-F4-2011-279239 (Osteogrow), H2020 Health GA 779340 (OSTEOproSPINE), and European Regional Development Fund—Scientific Center of Excellence for Reproductive and Regenerative Medicine (project "Reproductive and regenerative medicine—exploration of new platforms and potentials," GA KK.01.1.1.01.0008 funded by the EU through the ERDF).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** Special thanks to Jack Ratliff (Ratliff Histology Consultants LLC) for the excellent preparation of the undecalcified histology sections presented in this paper.

**Conflicts of Interest:** S.V. is a cofounder of perForm Biologics and a coordinator of the EU HORI-ZON 2020 grant OSTEOproSPINE funding clinical studies of the new drug for bone repair (patent WO2019076484A1).

#### **References**


#### *Case Report*

### **Design Techniques to Optimize the Sca**ff**old Performance: Freeze-dried Bone Custom-Made Allografts for Maxillary Alveolar Horizontal Ridge Augmentation**

#### **Felice Roberto Grassi 1 , Roberta Grassi 2 , Leonardo Vivarelli 3 , Dante Dallari 3 , Marco Govoni 3 , Gianna Maria Nardi 4 , Zamira Kalemaj <sup>5</sup> and Andrea Ballini 1,6,7, \***


Received: 29 January 2020; Accepted: 16 March 2020; Published: 19 March 2020

**Abstract:** The purpose of the current investigation was to evaluate the clinical success of horizontal ridge augmentation in severely atrophic maxilla (Cawood and Howell class IV) using freeze-dried custom made bone harvested from the tibial hemiplateau of cadaver donors, and to analyze the marginal bone level gain prior to dental implant placement at nine months subsequent to bone grafting and before prosthetic rehabilitation. A 52-year-old woman received custom made bone grafts. The patient underwent CT scans two weeks prior and nine months after surgery for graft volume and density analysis. The clinical and radiographic bone observations showed a very low rate of resorption after bone graft and implant placement. The custom-made allograft material was a highly effective modality for restoring the alveolar horizontal ridge, resulting in a reduction of the need to obtain autogenous bone from a secondary site with predictable procedure. Further studies are needed to investigate its behavior at longer time periods.

**Keywords:** geometry optimization of scaffolds; allograft; block bone grafts; custom made bone; design techniques for scaffold; precision and translational medicine

#### **1. Introduction**

Implant-supported rehabilitation of the edentulous ridge requires adequate volume and integrity of the alveolar bone [1].

Bone resorption in the maxillary ridge, due to trauma, pathology or tooth loss, frequently results in a knife-edged deformity, which complicates implant placement and stabilization, particularly in the posterior jaw [2–4]. Grafting with allograft bone has been documented to be a useful tool in reconstructing jaw anatomy, [5] restoring esthetics [6], and providing biomechanical support for the placement of dental implants [7].

Clinically, the most suitable banked bone allografts are fresh-frozen (FFBAs), freeze-dried bone allografts (FDBAs), and demineralized freeze-dried (DFDBAs) [8], although in oro-maxillofacial surgical interventions FDBAs and DFDBAs are the most used.

Frozen bone is accessible for human receivers after at least 6 months of quarantine at −80 ◦C [9] and no additional preparation is required. Moreover, the osteoinductive proteins are preserved [10].

Strict guidelines for tissue harvesting and storing at −80 ◦C make the risk of primary infections and antigenicity reasonably low [11]. In addition, the reduction of water content in frozen bone by the freeze-drying process further decreases potential microbial contaminations [12].

Frozen bone and relative freeze-dried/demineralized products are accessible as cancellous granules/blocks, corticocancellous granules/blocks, and cortical granules or chips. Once defrosted, frozen bone has handling qualities similar to fresh bone [8] while freeze-dried allografts have the additional advantage of storage at room temperature.

Bone density can be measured with high reproducibility by means of Cone-Beam-Computed-Tomography (CBCT) scans [13]. Further techniques, such as orthopantomography, do not assure fitting precision in density determination [13].

The aim of the present translational study was to evaluate the clinical success of horizontal ridge augmentation in severely atrophic maxilla (Cawood and Howell class IV) using custom made bone harvested from the tibial hemiplateau of cadaver donors.

#### **2. Materials and Methods**

A 52-year-old woman presented compromised anterior maxillary ridges, who presented for the placement of dental implants, was included in this pilot study. Written and verbal information was given to the patient before enrollment, and written informed consent was obtained. The study was conducted in full accordance with the World Medical Association Declaration of Helsinki on experimentation involving human subjects, as revised in 2008.

Inclusion criteria was a horizontal severely atrophic maxilla (Cawood and Howell class IV), needing a bone grafting procedure prior to implant placement. Exclusion criteria were established according to Venet et al. (2017) [12]. Plaque index score was maintained ≤25% throughout the study [14].

#### *Graft Sample Blocks*

The processing was performed on corticocancellous bone blocks obtained from a proximal tibial epiphysis, in the anatomical region between the articular surface of the tibial plateau and tuberosity.

Human allogeneic bone blocks were collected from cadaveric donors, stored at −80 ◦C and processed in the accredited public non-profit Musculoskeletal Tissue Bank of IRCCS Istituto Ortopedico Rizzoli (Bologna, Italy), authorized by the Italian National Transplant Center for the collection, processing, and distribution of human musculoskeletal tissue [15], and registered in the European Tissue Establishment list (code IT000096). The choice of the block to be machined involved considering dimensions slightly greater than the machining area defined with the graft design.

After thawing, each block was fixed with clamps on a special stainless-steel table in a GMP-Class A (Good Manufacturing Practice) Clean Room environment. Then, the table was fixed inside the CNC (Computer Numerical Control) milling machine (model Bright, Delta Macchine, Rieti, Italy). After tool fixing, a 3 mm diameter ball mill, the execution of the machining trajectory was started and accurately monitored by the operator.

At the end of the processing, the block was removed and finished, the supports were broken by hand and sharp edges, and rough corners were hand-refined using sterile rasps. Then, grafts were cleaned with organic solvents, washed with sterile water, and freeze-dried (VirTis Genesis 25, SP Scientific, Warminster, PA, USA). Finally, grafts were individually wrapped in triple pack. Microbiological

sampling was performed during the processing of grafts and after the lyophilization protocol in order to exclude microbial contamination and declare tissues suitable for implantation.

According to the Guidelines of the Italian National Transplant Center, lyophilized bone grafts are preserved at room temperature for a maximum of 5 years.

*Materials* **2020**, *13*, x FOR PEER REVIEW 3 of 11

#### *Graft Design*

[16–18].

Digital Imaging and Communications in Medicine (DICOM) data of the maxilla (Figure 1) were acquired by a Cone Beam Computerized Tomography (CBCT) scanner (Orthophos XG 3D Ceph, Dentsply Sirona Italia Srl, Roma, Italy) and imported into the 3DSlicer software (www.slicer.org) [16–18]. *Graft Design*  Digital Imaging and Communications in Medicine (DICOM) data of the maxilla (Figure 1) were acquired by a Cone Beam Computerized Tomography (CBCT) scanner (Orthophos XG 3D Ceph, Dentsply Sirona Italia Srl, Roma, Italy) and imported into the 3DSlicer software (www.slicer.org)

**Figure 1.** Cone-Beam-Computed-Tomography (CBCT) axial (a) and cross-sectional (b1,b2) images before surgical procedure. **Figure 1.** Cone-Beam-Computed-Tomography (CBCT) axial (**a**) and cross-sectional (**b1,b2**) images before surgical procedure.

After setting a threshold for the automatic selection of the areas delimiting the cortical bone, a

manual analysis for each slice was performed to correct potential errors. Starting from the selected areas, an automatic procedure generated the 3D model as a stereolithography (STL) file. The graft design was performed by Rhinoceros ver. 4 (www.rhino3d.com) following these steps: (a) placement of a parallelepiped in the position where the volume increase was required; (b) Boolean subtraction between the parallelepiped and the STL model of the patient's anatomy; (c) After setting a threshold for the automatic selection of the areas delimiting the cortical bone, a manual analysis for each slice was performed to correct potential errors. Starting from the selected areas, an automatic procedure generated the 3D model as a stereolithography (STL) file.

revision of the model for the manufacture by a 3-axis milling machine to obtain a L-shape section. This procedure was performed for each graft. The design of the grafts has been repeatedly validated and subsequently corrected following the surgeon's indications, up to the final model (Figure 2). The graft design was performed by Rhinoceros ver. 4 (www.rhino3d.com) following these steps: (a) placement of a parallelepiped in the position where the volume increase was required; (b) Boolean subtraction between the parallelepiped and the STL model of the patient's anatomy; (c) revision of the model for the manufacture by a 3-axis milling machine to obtain a L-shape section. This procedure was performed for each graft.

The design of the grafts has been repeatedly validated and subsequently corrected following the surgeon's indications, up to the final model (Figure 2).

#### *Trajectory Planning and Graft Manufacturing*

The planning of machining trajectories was performed using Rhinoceros by the Rhino-CAM 2 plug-in (MecSoft Corporation, Irvine, CA, USA). Each graft was positioned in the center of the machining CAM area with the larger flat area facing down to keep the cortical portion of the tissue intact, in order to enhance the graft resistance during processing and avoid breaking during implantation. Bone bridges have been added to the CAD design to maintain fixation during processing and removed by hand at the end of the graft manufacturing procedure. The working area was defined according to the design's dimensions and the diameter of the milling tool.

The tool-path trajectories have been programmed, subsequently simulated and modified to obtain the best result (Figure 3). Then, the definitive trajectories have been exported as G-Code coding specific for the milling machine used.

*Materials* **2020**, *13*, x FOR PEER REVIEW 4 of 11

**Figure 2.** (a) Render of 3D bone donor site harvested from a cadaver tibial hemiplateau. (b1–b3) Render of 3D reconstruction of patient's anatomy (yellow) and designed bone grafts (white). (c) Final block for clinical use. **Figure 2.** (**a**) Render of 3D bone donor site harvested from a cadaver tibial hemiplateau. (**b1**–**b3**) Render of 3D reconstruction of patient's anatomy (yellow) and designed bone grafts (white). (**c**) Final block for clinical use.

*Materials* **2020**, *13*, x FOR PEER REVIEW 5 of 11

*Trajectory Planning and Graft Manufacturing* 

custom allograft was freeze-dried.

*Surgical Procedure* 

**Figure 3.** Tool-path trajectory planning: (a) added bone bridges; (b) work area; (c) graft design; (d) tool-path. The machining operation was composed by two different phases: horizontal roughing and **Figure 3.** Tool-path trajectory planning: (**a**) added bone bridges; (**b**) work area; (**c**) graft design; (**d**) tool-path.

parallel finishing. The tool used for each phase was the same, a fluted 3 mm ball mill with two teeth, hard metal k10 material (tungsten carbide). In each phase rotational velocity of the tool was 5300 RPM and feed rate was 200 mm/min, that corresponds to a feed per tooth of 19 μm and a tangential velocity of 50 m/min (0.833 m/s). Horizontal roughing was performed as a full immersion vertical milling, with a machining depth of 3.5 mm, leaving a stock margin of 0.3 mm. Parallel finishing was a vertical milling, with a machining depth depending on stock margins and parallel steps of 0.7 mm (Figure 3d) leaving a variable stock margin between 0 mm and 0.15 mm (a sufficient dimensional accuracy for oral implant surgery). This machining strategy, performed by few passages, reduced the machining time and the bone heating during the process. Although temperature was not monitored during the fabrication of these two custom grafts, our machining parameters, especially the small feed per tooth, allow a local temperature between 40 °C and 60 °C for cortical bone and lower values for cancellous bone. According to Krause [19], these parameters do not cause the denaturation of proteins and enzymes that occurs for temperature above 70 °C. As reported by Van Isacker et al. [20], frozen bone can be handled and reshaped like normal bone and is fully workable, but freeze-dried bone, unless rehydrated in saline, is brittle like ceramic, and is not fully workable. The machining operation was composed by two different phases: horizontal roughing and parallel finishing. The tool used for each phase was the same, a fluted 3 mm ball mill with two teeth, hard metal k10 material (tungsten carbide). In each phase rotational velocity of the tool was 5300 RPM and feed rate was 200 mm/min, that corresponds to a feed per tooth of 19 µm and a tangential velocity of 50 m/min (0.833 m/s). Horizontal roughing was performed as a full immersion vertical milling, with a machining depth of 3.5 mm, leaving a stock margin of 0.3 mm. Parallel finishing was a vertical milling, with a machining depth depending on stock margins and parallel steps of 0.7 mm (Figure 3d) leaving a variable stock margin between 0 mm and 0.15 mm (a sufficient dimensional accuracy for oral implant surgery). This machining strategy, performed by few passages, reduced the machining time and the bone heating during the process. Although temperature was not monitored during the fabrication of these two custom grafts, our machining parameters, especially the small feed per

full-thickness crestal incision with two vertical releasing incisions (Figure 4a).

Local anesthesia was obtained by infiltrating articaine (4% containing 1:100,000 adrenaline). The exposure of the three-dimensional aspect of the bone defects was achieved through a

The clinically sized, anatomically shaped, custom-made bone block was placed in position strictly overlapping the underlying alveolar crest and fitted securely to the residual bone (Figure

Thus, to execute a better processing procedure and to obtain a further reduction in temperature, machining was performed on a frozen bone block. At the end of the machining operation, the tooth, allow a local temperature between 40 ◦C and 60 ◦C for cortical bone and lower values for cancellous bone. According to Krause [19], these parameters do not cause the denaturation of proteins and enzymes that occurs for temperature above 70 ◦C.

As reported by Van Isacker et al. [20], frozen bone can be handled and reshaped like normal bone and is fully workable, but freeze-dried bone, unless rehydrated in saline, is brittle like ceramic, and is not fully workable.

Thus, to execute a better processing procedure and to obtain a further reduction in temperature, machining was performed on a frozen bone block. At the end of the machining operation, the custom allograft was freeze-dried. *Materials* **2020**, *13*, x FOR PEER REVIEW 6 of 11

#### *Surgical Procedure* 4b1). The recipient site was weakened with multiple micro-holes to enhance bleeding from the trabecular bone [12,21,22]. Rigid fixation of the scaffold to the residual crest was obtained by means

Local anesthesia was obtained by infiltrating articaine (4% containing 1:100,000 adrenaline). The exposure of the three-dimensional aspect of the bone defects was achieved through a full-thickness crestal incision with two vertical releasing incisions (Figure 4a). of a titanium mini-screw (1.5 mm width, 8 mm length) (Tekka by Global-D, Lyon, France) (Figure 4b2) [12]. The grafted area was closed with a pulley suture for proper flap adaptation and to avoid any tissue strangulation by an absorbable 4.0/5.0 suture material. Sutures were removed 14 days postoperatively (Figure 4c) and no prosthetic device was admitted in the following 90 days, with

particular care in domiciliary oral hygiene procedures.

**Figure 4.** The custom-made freeze-dried bone allograft (FDBA) was positioned in the deficit area and stabilized with a screw. (a) Anatomical site prior surgical procedure and clinical view of the defect. (b1,b2) Screw positioning on a custom-made FDBA. (c) Sutures. Preceding to the second phase, supplementary CBCT scans were performed to evaluate grafts **Figure 4.** The custom-made freeze-dried bone allograft (FDBA) was positioned in the deficit area and stabilized with a screw. (**a**) Anatomical site prior surgical procedure and clinical view of the defect. (**b1,b2**) Screw positioning on a custom-made FDBA. (**c**) Sutures.

The clinically sized, anatomically shaped, custom-made bone block was placed in position strictly overlapping the underlying alveolar crest and fitted securely to the residual bone (Figure 4b1). The recipient site was weakened with multiple micro-holes to enhance bleeding from the trabecular bone [12,21,22]. Rigid fixation of the scaffold to the residual crest was obtained by means of a titanium mini-screw (1.5 mm width, 8 mm length) (Tekka by Global-D, Lyon, France) (Figure 4b2) [12].

The grafted area was closed with a pulley suture for proper flap adaptation and to avoid any tissue strangulation by an absorbable 4.0/5.0 suture material. Sutures were removed 14 days postoperatively (Figure 4c) and no prosthetic device was admitted in the following 90 days, with particular care in domiciliary oral hygiene procedures.

Preceding to the second phase, supplementary CBCT scans were performed to evaluate grafts gain (Figure 5).

**Figure 5.** CBCT scans were done after 9 months to define implant placement width and length, on (a) axial and (b1,b2) cross-sectional images. After a 9 month healing period, micro-screws were removed (Figure 6a,b), and 3.7 mm in diameter for a 10 mm in length dental implants (iRES SAGL, Mendrisio, Switzerland) were placed, and sutures performed (Figure 6c). No supplementary grafting procedure was required in the current investigation [22].

#### *Statistical Evaluation*

gain (Figure 5).

The outcome values were analyzed using the *t-test* for paired samples for pre–post differences with time as the factor and IBM Statistical Package for the Social Sciences (SPSS Inc. Version 21.0, 4b2) [12].

gain (Figure 5).

particular care in domiciliary oral hygiene procedures.

Chicago, IL, USA) software to detect significant differences between pre-test and post-test scores obtained from bone volume analysis and measurements of variation in ridge thickness from CBCT images at 9 months after allograft insertion. **Figure 4.** The custom-made freeze-dried bone allograft (FDBA) was positioned in the deficit area and stabilized with a screw. (a) Anatomical site prior surgical procedure and clinical view of the defect. (b1,b2) Screw positioning on a custom-made FDBA. (c) Sutures. Preceding to the second phase, supplementary CBCT scans were performed to evaluate grafts

*Materials* **2020**, *13*, x FOR PEER REVIEW 6 of 11 4b1). The recipient site was weakened with multiple micro-holes to enhance bleeding from the trabecular bone [12,21,22]. Rigid fixation of the scaffold to the residual crest was obtained by means of a titanium mini-screw (1.5 mm width, 8 mm length) (Tekka by Global-D, Lyon, France) (Figure

The grafted area was closed with a pulley suture for proper flap adaptation and to avoid any tissue strangulation by an absorbable 4.0/5.0 suture material. Sutures were removed 14 days postoperatively (Figure 4c) and no prosthetic device was admitted in the following 90 days, with

**Figure 5.** CBCT scans were done after 9 months to define implant placement width and length, on (a) axial and (b1,b2) cross-sectional images. **Figure 5.** CBCT scans were done after 9 months to define implant placement width and length, on (**a**) axial and (**b1,b2**) cross-sectional images. *Materials* **2020**, *13*, x FOR PEER REVIEW 7 of 11 After a 9 month healing period, micro-screws were removed (Figure 6a,b), and 3.7 mm in diameter for a 10 mm in length dental implants (iRES SAGL, Mendrisio, Switzerland) were placed, and sutures performed (Figure 6c). No supplementary grafting procedure was required in the

**Figure 6.** Nine months after the surgical procedure, the implant was positioned. (a) Soft tissues before surgical procedure. (b) Surgical flap is elevated. (c1) Implant was positioned and (c2) sutures performed. (d) Frontal view of the final rehabilitation. *Statistical Evaluation*  **Figure 6.** Nine months after the surgical procedure, the implant was positioned. (**a**) Soft tissues before surgical procedure. (**b**) Surgical flap is elevated. (**c1**) Implant was positioned and (**c2**) sutures performed. (**d**) Frontal view of the final rehabilitation.

#### **3. Results**

with time as the factor and IBM Statistical Package for the Social Sciences (SPSS Inc. Version 21.0, Chicago, IL, USA) software to detect significant differences between pre-test and post-test scores obtained from bone volume analysis and measurements of variation in ridge thickness from CBCT images at 9 months after allograft insertion. A patient presenting atrophic anterior maxillae was selected to participate in the study. The patient received 2 custom-made allografts. The healing period was uneventful.

The outcome values were analyzed using the *t-test* for paired samples for pre–post differences

**3. Results**  A patient presenting atrophic anterior maxillae was selected to participate in the study. The patient received 2 custom-made allografts. The healing period was uneventful. At the crestal level, the patient met the inclusion criteria of ridge width around 2 mm [22]. The titanium fixation mini-screws detached and bleeding from the bone graft were detected, demonstrating revascularization of the site.

> At the crestal level, the patient met the inclusion criteria of ridge width around 2 mm [22]. The titanium fixation mini-screws detached and bleeding from the bone graft were detected,

demonstrating revascularization of the site.

Moreover, the regenerative surgical procedure went well. In fact, the custom-made allograft scaffolds perfectly fitted in the bone anatomy and were therefore easily adapted to the bone defects during surgery, secured by titanium mini-screws (Figures 4–6). This excellent matching of the size/shape helped the surgeon to reduce the operation time [23]. Moreover, all implants were inserted with fitting primary stability (Figure 6d).

Paired comparisons showed a significant mean increase in ridge thickness of 5.0 ± 0.55 mm from the preoperative measurement to the postoperative measurement (p < 0.01).

The bone resorption that occurred during the incorporation period corresponds to 7.8% (95% confidence interval (CI): (6.4%, 9.2%)) of the measured postoperative ridge thickness.

#### **4. Discussion**

Maxillary ridge augmentation relates to procedures designed to correct a thin alveolar ridge. For dental implant placement, adequate bone volume is necessary, unfortunately, bone volume is not always adequate, particularly in elderly patients.

Through the current improvements in planning and manufacturing software and hardware, graft customization can be combined with the implant dentistry digital workflow with the potential patients to become part of daily clinical practice [24].

The reduction of risk factors such as infections caused by contamination of surgical site, or contamination of the graft during preoperative preparation, can be obtained with the manufacturing of a custom sterile graft [23]. The correct design of a custom graft, with also a reduction of surgical time, can be obtained with a proper planning of surgical procedure.

Nevertheless, the sterile manufacturing of bone in a GMP-Class A environment, without the need of terminal sterilization via gamma radiation, guarantees the safety of graft in terms of sterility and the quality of tissue in terms of biological properties [25,26].

Among bone replacement grafts, autograft is considered the gold standard [27]. However, owing to the low level of patient acceptance of autografts, allografts (grafts from the same species) have become the most popular bone grafts in oral surgery. Among allografts available for the periodontal regeneration, freeze-dried bone allograft (FDBA) and the demineralized freeze-dried allograft (DFDBA) are the allograft-forms most used [28].

Although FDBAs and DFDBAs were first used in periodontal therapy in early 1970s, to date they continue to be the most requested allografts in clinical practice because of their osteoconductive and osteoinductive properties, respectively [29]. Compared to autogenous bone, allografts feature several advantages such as adequate availability and the elimination of additional donor site surgery and morbidity [30]. Besides, the freeze-drying technique removes more than 95% of water content from bone, allowing further storage of tissues at room temperature and reducing antigenicity [31]. In addition, the application of stringent tissue banking guidelines together with bone processing and liophilization protocols performed in a GMP-Class A environment, minimize the risk of microbial contaminations [32], and reduce the need for terminal sterilization by gamma rays and consequent loss of mechanical strength of bone tissues [33]. As a matter of fact, although in freeze-dried bone tissues most of the water content has been removed, ionization has a direct effect on the collagen fiber bundles causing the breakage of protein chains and hence alters mechanical resistance [34].

Since only the demineralized bone matrix retains substantial amounts of endogenous osteoinductive proteins such as bone morphogenetic proteins (BMPs), known to promote bone growth and regeneration (i.e., BMP-2, BMP-4, BMP-7) [35,36], FDBAs are considered only as osteoconductive allografts. However, their structure can serve as a scaffold and support for the host cells that can migrate into the grafts and then differentiate into new osteogenic cells. In addition, in respect to DFDBAs, the superior mechanical properties of FDBAs make these allografts more suitable for restoring the alveolar horizontal ridge in athrophic maxillae.

Hence, the findings of our study show, with a high success rate, that FDBAs represent a reliable treatment option for extensive rehabilitation of atrophic maxillae, consistent with findings reported with the use of autologous bone.

The success rate of the block grafts was very effective and comparable with those reported by other authors [37–39]. Besides, the augmentation procedure permitted the insertion of implants in the grafted area 9 months after surgery. The clinical and radiographic observations showed a very low rate of bone resorption of the graft material, improving the ability to place endosseous implants.

A potential scaffold frontier for bone tissue engineering protocols seems to be the mesenchymal stem cells seeded in nanocomposite materials with antibacterial properties, but further studies in this direction are necessary prior to be consider it as a safe and useful tool for human implantation purposes [40].

Our data showed that custom-made FDBA can be used as successful graft material for the treatment of bone maxillary ridge defects. If adequate surgical techniques are adopted, this type of bone graft can be safely used in regions of implant placement as a suitable alternative to autogenous grafts. Furthermore, the personalized design of custom grafts and sterile machining in a GMP-Class A environment, as described above, can be applied to different surgical specialties, such as orthopedics, spinal surgery, and maxillofacial surgery.

#### **5. Conclusions**

This pilot study demonstrated the possibility of fabricating customized CAD/CAM grafts from the tibial hemiplateau of cadaver donors. The grafts were digitally designed based on CBCT scans of partially dentate patients using a set of 3D software, showing that grafts dimensions were correlated to the defect type. The sterile manufacturing of grafts resulted in the correct reproduction of the designed graft, leading to a correct matching with the bone surface of patient. Additional in vivo studies with custom-made bone allografts are required for the validation of this digital workflow for maxillary bone augmentation.

**Author Contributions:** Conception and design: F.R.G., R.G. and G.M.N.; data acquisition: Z.K.; writing and manuscript editing, analysis and interpretation of data: A.B., L.V., M.G.; 3D reconstructions and volume renderings: L.V., D.D., M.G.; supervised the manuscript final approval of the version to be published: F.R.G. and A.B. All authors read and approved the final manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Evaluation of the Use of an Inorganic Bone Matrix in the Repair of Bone Defects in Rats Submitted to Experimental Alcoholism**

**Iris Jasmin Santos German 1,2,3 , Karina Torres Pomini 1 , Ana Carolina Cestari Bighetti 1 , Jesus Carlos Andreo 1 , Carlos Henrique Bertoni Reis 4 , André Luis Shinohara 1 , Geraldo Marco Rosa Júnior 5,6 , Daniel de Bortoli Teixeira 4 , Marcelie Priscila de Oliveira Rosso 1 , Daniela Vieira Buchaim 4,7 and Rogério Leone Buchaim 1,4, \***


Received: 13 December 2019; Accepted: 27 January 2020; Published: 4 February 2020

**Abstract:** To assess the effects of chronic alcoholism on the repair of bone defects associated with xenograft. Forty male rats were distributed in: control group (CG, *n* = 20) and experimental group (EG, *n* = 20), which received 25% ethanol ad libitum after a period of adaptation. After 90 days of liquid diet, the rats were submitted to 5.0-mm bilateral craniotomy on the parietal bones, subdividing into groups: CCG (control group that received only water with liquid diet and the defect was filled with blood clot), BCG (control group that received only water with liquid diet and the defect was filled with biomaterial), CEG (alcoholic group that received only ethanol solution 25% *v*/*v* with liquid diet and the defect was filled with blood clot), and BEG (alcoholic group that received only ethanol solution 25% *v*/*v* with liquid diet and the defect was filled with biomaterial). In the analysis of body mass, the drunk animals presented the lowest averages in relation to non-drunk animals during the experimental period. Histomorphologically all groups presented bone formation restricted to the defect margins at 60 days, with bone islets adjacent to the BCG biomaterial particles. CEG showed significant difference compared to BEG only at 40 days (17.42 ± 2.78 vs. 9.59 ± 4.59, respectively). In the birefringence analysis, in early periods all groups showed red-orange birefringence turning greenish-yellow at the end of the experiment. The results provided that, regardless of clinical condition, i.e., alcoholic or non-alcoholic, in the final period of the experiment, the process of bone defect recomposition was similar with the use of xenograft or only clot.

**Keywords:** bone repair; biomaterial; alcoholism; alcohol

#### **1. Introduction**

The concept of alcoholism peaked in the eighteenth century, shortly after the growing production and marketing of distilled alcohol, resulting from the industrial revolution. Ethanol is a water-soluble organic solvent with the ability to penetrate all compartments of the human body, and its systemic effects produce changes in the central nervous system, muscular system, liver disease, chronic pancreatitis, cardiovascular disease, lung disease, among others [1].

Ethanol impairs bone formation by inhibiting osteoblast proliferation. In addition, alcohol induces oxidative stress and participates in the regulation of osteoclast differentiation, resulting in increased signaling of RANKL-RANK, kB nuclear factor activating receptor ligand in bone cells, and increased osteoclastogenesis [2].

In addition, ethanol changes the levels of cytokines responsible for regulating bone metabolism, such as increased tumor necrosis factor alpha (TNF-alpha) and interleukin-6 (IL-6). These, in turn, cause the suppression of osteoblast synthesis and consequently the deposition of osteoid matrix [3].

Stimulation of ethanol on osteocytes causes increased secretion of sclerostin protein, which binds to cellular receptors in order to antagonize the action of BMP in the Wnt signaling cascade. The morphology and osteocyte apoptosis number are also altered with alcohol consumption, which sends signs of osteoclast recruitment, increasing bone resorption [4].

This change in bone remodeling may be expressed by bone loss and serum osteocalcin levels, a marker of bone formation. Moreover, it compromises collagen gene expression, also non-collagenous matrix proteins, and significantly reduces the levels of carboxy-terminal procollagen I propeptide [5]. These changes are also related to the duration and amount of alcohol exposure [6].

Alcohol consumption causes harmful effects on bone density [7] and is related to osteoporosis, due to imbalance in bone repair [8], altering bone microarchitecture, decreasing trabecular and cortical bone thickness, and consequently increasing the risk of fracture [9].

There are many scientific reports about xenografts in the bone repair process, whose results evidence that it is a viable alternative to autologous grafts [10]. There is currently a social and commercial incentive for alcohol consumption worldwide, which may lead to a larger number of patients using ethyl derivatives in clinical practice. Given this context, associated with the lack of experimental studies in the literature that relate alcoholism to biomaterial used in this innovative research and translational projection, it was decided to analyze a low-cost bone substitute used in the areas of dentistry and medicine with alcoholism.

Therefore, the aim of this study was to evaluate the effects of chronic alcoholism on the repair of bone defects in rats filled or not with a bovine bone matrix.

#### **2. Materials and Methods**

#### *2.1. Biomaterial—Bone Graft Substitutes*

The Bonefill biomaterial (Bionnovation Biomedical A.B., Bauru, São Paulo, Brazil) evaluated in this study is produced by decalcification of the cortical portion of bovine femur, totally denatured and sterilized by 25 kGy gamma radiation. It is commercially available in average particle size between 0.6–1.5 mm in diameter and 0.5 gr package. This dental product is registered in Brazilian Ministry of Health by The National Sanitary Surveillance Agency (ANVISA10392710012).

#### *2.2. Experimental Design*

Forty adult 60-day-old male Wistar rats (*Rattus norvegicus*) were used, weighing approximately 250 g. The animals were kept in conventional cages containing 4 animals/box. The macroenvironment presented artificial timer-managed lighting, which controlled the 12/12-h light/dark cycle, with 220 lux brightness, 55% humidity, exhaust fan and air conditioning, maintaining an average temperature of 22 ◦C. All animals had free access to standard rat chow (NuvilabTM, Nuvital, Colombo, Paraná, Brazil) and their daily consumption was not measured (Table 1). All experimental procedures on the animals

were conducted after approval by the Institutional Review Board on Animal Studies of Bauru School of Dentistry, University of São Paulo (Protocol: CEEPA-023/2012).

**Table 1.** Nutritional composition of the chow used in the experiment for all animals. Ad libitum feeding of chow.


\* Nuvilab CR1. Nuvital, Colombo, Paraná, Brazil.

The animals were randomly separated into two large groups according to the type of liquid diet (drinking water or ethanol): CG, *n* = 20 (control group - non-alcoholic), which received a standard rat chow and water ad libitum; and EG, *n* = 20 (experimental group–alcoholic), which received a standard rat chow and ethanol ad libitum. After the surgical procedure, the groups were subdivided according to the type of treatment (clot or biomaterial) and the clinical condition (alcoholic or non-alcoholic), as follows: CCG (control group that received only water with liquid diet and the defect was filled with blood clot), BCG (control group that received only water with liquid diet and the defect was filled with biomaterial), CEG (alcoholic group that received only ethanol solution 25% *v*/*v* with liquid diet and the defect was filled with blood clot), and BEG (alcoholic group that received only ethanol solution 25% *v*/*v* with liquid diet and the defect was filled with biomaterial) (Figure 1A1).

2.30 21.96 4.61 8.36 4.04 48.73 1.32 0.82 3913

**Figure 1.** (**A**) Experimental design. (**A1**) Random Allocation—Forty adult male Wistar rats, (*Rattus norvegicus*), aged 60 days old, weighing around 250 g were divided into two broad groups: CG—Control group (*n* = 20)—that received only water with liquid diet and EG—Experimental group (*n* = 20)—that received ethanol solution 25% (*v*/*v*) with liquid diet after adaptation period. (**A2**) Ethanol Adaptation and Dependence Induced—Animals were gradually drunk at progressive concentrations of ethanol solution (8–16–25% *v*/*v*). After 21 days of alcohol adaptation, the animals remained at 25% (*v*/*v*) until the surgical procedure for 90 days. (**A3**) Treatments—After surgical procedures—bilateral bone defect model in the parietals, four subgroups were preformatted according to treatment (blood clot vs. biomaterial) and clinical conditions (alcoholic vs. non-alcoholic): Animals that received only water with liquid diet: CCG (right parietal bone defect was filled with blood clot) and BCG (left parietal bone defect was filled with biomaterial); Animals that received ethanol solution 25% (*v*/*v*): CEG (left parietal bone defect was filled with blood clot); BEG (right parietal bone defect was filled with biomaterial). B) Experimental Periods—at 10, 20, 40 and 60 days the skulls of 5 animals/group were collected, totalizing 5 defects/period of each subgroup CCG, BCG, CEG and BEG. (C) Surgical Procedures–Bilateral Bone Defect Model–(**C1**) 5-mm left parietal osteotomy; (**C2**) Two bone defects in parietal bones; (**C3**) One defect filled with biomaterial and the contralateral only with blood clot; (**C4**) Periosteum suture with nylon 5-0.

#### *2.3. Ethanol Adaptation and Induced Dependence—Semi-Voluntary Ethanol Administration—Alcohol-Liquid Diet*

This study followed the determined chronic alcoholism model of "semi-voluntary", where alcohol was the only liquid food available. Animals in the experimental group (EG) were submitted to an alcohol adaptation model, where the only available liquid source was ethanol ad libitum. In the first 7 days, the animals received 8% (*v*/*v*) ethanol solution, in the second week 16% (*v*/*v*) and third week 25% (*v*/*v*). After this period of gradual adaptation, the animals remained on the 25% (*v*/*v*) liquid ethanol diet for further 90 days, when they underwent experimental surgery and remained on the same diet until the corresponding euthanasia period. Animal health was monitored daily (Figure 1A2,B).

#### *2.4. Surgical Procedures*

After 111 days of ethanol intake and induced dependence, all rats were weighed and subjected to intramuscular general anesthesia with ketamine at a dose of 50 mg/kg i.m. (Dopalen™, Ceva, Paulinia, São Paulo, Brazil) plus xylazine at the dose of 10 mg/kg i.m. (Anasedan™, Ceva, Paulinia, São Paulo, Brazil), with strict monitoring of anesthesia mainly in alcoholic animals [11].

After frontoparietal trichotomy and antisepsis with 2% chlorhexidine, a cranio-caudal longitudinal incision of approximately 20 mm in length was made for tissue exposure and divulsion. Circular bilateral osteotomy was performed on the parietal bones with a 5 mm diameter trephine drill (Neodent, Curitiba, Paraná, Brazil) at low speed (1500 rpm) under constant saline irrigation (Figure 1C1,C2).

The defect on the left parietal was filled with 14 mg of biomaterial (previously established in a pilot study) and the right parietal was filled only with blood Clot (Figure 1A3,C3). The periosteum was repositioned and sutured with VicrylTM polyglactin suture (Ethicon J&J, São Paulo, Brazil) 5-0 and the integument with 4-0 silk suture (EthiconTM, J&J, São Paulo, Brazil) (Figure 1C4).

After the surgical procedure, the animals were placed under incandescent light for complete anesthetic recovery and submitted to a single intramuscular injection of enrofloxacin 2.5 mg/kg (Flotril™; Schering-Plough SA, Rio de Janeiro, Brazil) and intramuscular injections of 0.06 mg/kg dipyrone (Analgex™; Agener União, São Paulo, Brazil) for 3 days.

#### *2.5. Collection of Specimens and Histological Procedures*

Five animals from each group were euthanized using the aforementioned anesthetic overdose at the respective periods of 10, 20, 40 and 60 days (Figure 1B). The specimens were removed and fixed in 10% buffered formalin for 48 h and then demineralized in EDTA, a solution containing 4.13% Titriplex™ III (Merck KGaA, Darmstadt, Germany) and 0.44% sodium hydroxide, for a period of approximately 40 days. Then, the specimens were subjected to standard histological procedures and included in Histosec ™ (Merck KGaA, Darmstadt, Germany). Histological sections were obtained with 5 µm thickness prioritizing the defect centers for hematoxylin and eosin, Masson's trichrome and picrosirius-red staining.

#### *2.6. Body Mass Analysis*

Body mass was determined by simple weighing on a Bel Mark 3500 precision scale (BELTM Analytical Equipment Ltda, São Paulo, Brazil), with maximum capacity of 3500 g and minimum of 200 g with the aid of a Styrofoam box for animal containment. Measurements occurred on the day of the surgical procedure (initial mass) and on the corresponding days of euthanasia (10, 20, 40, and 60 days).

#### *2.7. Histomorphological and Histomorphometric Evaluation*

The histologic sections were analyzed in the Histology Laboratory of Bauru School of Dentistry, University of São Paulo (São Paulo, Brazil) by light microscopy (Olympus model BX50, Tokyo, Japan) at approximate magnifications of ×4, ×10, ×40 and ×100. To establish a standard criteria for judgment, there was a training session with an experienced pathologist.

For histomorphological description of the bone defect area, the central region (edge-to-edge measurement) was considered with the aid of free-scale image capture system (DP Controller software (3.2.1.276 version, Olympus, Tokyo, Japan) to analyze tissue formation granulation, inflammatory infiltrate, formation of primary bone tissue and bone maturation.

A virtual overall image of the defect (Masson's trichrome, ×10) was generated to quantify the volume density (%) of newly formed bone tissue and biomaterial by AxioVision Rel. 4.8 Ink (Carl Zeiss MicroImaging GmbH, Jena, Germany). For determination of volumetric density (%), the equation Vvi = AAi = Ai/A × 100 was considered, considering Vvi (volume density), AAi (area density), Ai (area filled with newly formed bone tissue or particle of biomaterial), A (total area examined) (Figure S1) [12].

Images from picrosirius-red stained sections were captured using a higher resolution digital camera Leica DFC 310FX (LeicaTM, Microsystems, Wetzlar, Germany) connected to a confocal laser microscope Leica DM IRBE and capture system LAS 4.0.0 (LeicaTM, Microsystems, Heerbrugg, Switzerland). The quality of newly formed bone in the defects was evaluated by the orientation pattern and width of the collagen fibers detected by the birefringence of polarization colors ranging from red-orange (primary disorganized bone tissue) to green-yellow (organized bone-lamellar bone tissue).

#### *2.8. Statistical Evaluation of Data*

Data on body mass, volume density (%) of newly formed bone tissue and biomaterial particle were expressed as mean ± standard deviation of the mean (SEM). All tests were performed using Statistica 10.0 software (StatSoft Inc., Tulsa, OK, USA) and the significance level was set at *p* < 0.05. The independent "t" test was used to compare the initial body mass in alcoholic and non-alcoholic groups, and the paired "t" test was applied to compare the initial and final body mass within the same group. The percentage of bone formation and biomaterial in groups in different periods was assessed by one-way ANOVA variance test (time) for independent samples, and Tukey's post hoc test, at a significance level of *p* < 0.05. To compare the percentage of bone formation in drunk vs. non-alcoholic animals (CCG vs. CEG and BCG vs. BEG) in the different periods, the t-test was applied for independent samples, and post hoc Tukey test, at a significance level of *p* < 0.05. To compare the percentage of bone formation in animals of the same group for different treatments (CCG vs. BCG and CEG vs. BEG), at different periods, the paired t-test and post hoc Tukey test were applied, at a significance level of *p* < 0.05.

#### **3. Results**

#### *3.1. E*ff*ects of Ethanol Administration on Behavioral and Clinical Profiles of Rats*

Regarding the general clinical profile, all animals presented good physical condition throughout the experiment, with no signs of morbidity and no mortality rate. There was no infection in either group, nor in the surgical area. However, some animals in the experimental group-alcohol group showed changes in their behavioral profile, especially regarding the parameters of agitation, aggressiveness and exploratory activity.

In the evaluation of body mass, after the period of alcoholic induction (111 days), the animals of the experimental group showed lower initial mass gain (day of surgery) compared to the control group (343.82 ± 41.93 vs. 444.59 ± 45.20, weight in grams, respectively). At 20 days, the mean mass of EG group showed no significant difference, but between 40–60 days there was a slight increase in the means from 5.9% to 6.4%, relative to day 0. In non-alcoholic animals, CG, the increase between 20–60 days was 7.9% to 14.1%, relative to day 0 (Figure 2).

**Figure 2.** (**A**) Graphic representation of body mass (g) during ethanol adaptation-induced dependence (111 days) and experimental periods of control (CG-water ad libitum) and experimental (EG-ethanol solution ad libitum) groups. (**B**) Comparison between body masses of CG vs. EG showing the negative effect of alcohol. Different letters *p* < 0.05 (independent t-test and paired t-test showed interaction between group and period).

#### *3.2. Histological Evaluation*

At 10 days, all experimental groups presented bone formation at the defect margins (Figure 3A–D). In groups treated with blood clot, CCG and CEG, there was predominance of richly vascularized granulation tissue filling the entire surgical area. However, BCG and BEG were shown to be reaction tissue surrounding the biomaterial particles (Figure 4A,B and Figure 5).

**Figure 3.** Panoramic histological views in skull defects created in the animals (**A–D**). Non-alcoholic (**A,B**) and alcoholic (**C**,**D**) treated with blood clot or biomaterial at different experimental periods, 10, 20, 40 and 60 days. (**A**,**B**) Non-alcoholic Groups: showed new bone formation (blue arrows) from the defect border (b), with partial closure by fibrous connective tissue in GCC (black arrows) or by particles of biomaterial surrounded by fibrous connective tissue in BCG (asterisk). (**C**,**D**) Alcoholic groups: new bone formation was observed from the border and on the dura-mater surface, with bone islets on the defect center in CEG and bone islets adjacent to the biomaterial particles in BEG. In both groups, at 60 days, lamellar tissue transition to compact tissue was observed (Masson's Trichrome; original magnification ×4; bar = 2 mm).

**Figure 4.** Details of evolution of bone healing of cranial defects created in the Non-Alcoholic (**A**) and Alcoholic (**B**) animals treated with blood clot or biomaterial. (**A1**) CCG (Non-alcoholic Group; defects filled with blood clot): at 10 days, defects showed trabecular bone formation (asterisk), with the presence of osteoclastic cells (Oc) on the edge of the remaining bone tissue. 20–40 days, the new bone formed showed bone maturation phase surrounded by blood capillaries (black arrow). At 60 days, collagen fibers were arranged in a more regular manner (Cf). (**A2**) BCG (Non-alcoholic Group: defects filled with biomaterial): at 10 days, defects showed tissue reaction (RT) surrounding the particles of the biomaterial (B); artifacts in histologic sections (double arrow–gap between biomaterial and tissue). Between 20–60 days, connective tissue (CT) presented scarce inflammatory cells with thin and thick collagen fibers, which were parallelly arranged at the end of the experimental period. (**B1**) CEG (Alcoholic Group; defects filled with blood clot): in the early periods, the defects presented inflammatory cells, decreasing at 40 days. The bone tissue formed at 60 days was predominantly compact and mature. (**B2**) BEG (Alcoholic Group; defects filled with biomaterial): 10–20 days, sections showed discrete bone formation, and biomaterial particles permeated by reaction tissue. In the later periods, collagen fibers were organized in parallel, and osteoblastic cells (Ob) forming a single cell line adjacent to the matrix. Masson's Trichrome; original magnification ×40; bar = 100 µm; and Insets, magnified images ×100; bar = 50 µm.

**Figure 5.** Histological details with HE staining of evolution of cranial defect bone healing created in the Non-Alcoholic and Alcoholic animals treated with blood clot or biomaterial. At 10-20 days, CCG and CEG showed the presence of richly vascularized connective tissue (CT) (black arrow) and new bone tissue (asterisk) at the defect margin (b) with trabecular arrangement. All groups presented inflammatory cells (inside the black lined area), more evident in BCG and BEG, permeating the particles (B). At 40 days there was typical lamellar arrangement, interspersed with osteocytes (green arrow), and in BCG and BEG inflammatory cells and multinucleated giant cells (blue arrow). At 60 days, bone tissue was mature and compact in defects filled with blood clot, there was decreased inflammatory reaction in BCG and BEG groups and regularly organized collagen fibers. Artifacts in histologic sections (double arrow–gap between biomaterial and tissue). HE; original magnification x 40; bar = 100 µm.

At 20 days, in the CCG and CEG groups, immature trabecular bone formation was in the transition phase of bone maturation, obtaining a denser lamellar arrangement at 40 days. In BCG, the reaction tissue was in resolution phase, with sparse inflammatory infiltrate, unlike that observed in the alcohol-treated / biomaterial-treated animals, BEG (Figure 4A,B and Figure 5).

At the end of the experimental period, at 60 days, all groups presented complete closure of the surgical area by fibrous connective tissue and / or particles of biomaterial. In defects treated with blood clot, in CCG and CEG, the new bone formation remained restricted to the defect margins and over the dura mater with a more evident bone maturation pattern, but with smaller thickness than the remaining bone. In the BCG and BEG groups, the particles were encased in evenly arranged, thicker collagen fibers (Figure 3A–D, Figure 4A,B and Figure 5).

#### *3.3. Histomorphometric Evaluation*

In relation to biomaterial volume density (%) in the BCG and BEG groups, a tendency of decrease between 10 (mean of 30.4%) to 60 days (mean of 18.98%) was observed (compare data in the Table 2). However, no statistical differences between periods in each group (ANOVA, *p* > 0.2) or groups per period ("t" test, *p* > 0.6) were observed.


**Table 2.** Mean ± standard deviation of volume density of comparison of biomaterial in the different experimental groups.

Same lowercase letters indicate that there was no statistically significant difference. Significant differences *p* < 0.05.

Regarding bone formation volume density (Table 3), a significant increase was observed only in the defects created in alcoholic rats filled with blood clot, CEG group (*p* = 0.007), between 10 (mean of 6.67%) and 40 days (mean of 17.42%), as well as between 10 and 60 days (mean of 18.29%). In the same animals, the contralateral defects filled with biomaterials, BEG group, the bone formation volume density was 0.43 times smaller than CEG. Regardless of clinical condition (Table 4), no statistical differences in the bone formation were observed between CCG vs. CEG (*p* > 0.1) and BCG vs. BEG (*p* > 0.3).

**Table 3.** Mean ± standard deviation of volume density of new bone formation. Comparison among periods within the same group was evaluated by one-way ANOVA (column, 10 vs. 20 vs. 40 vs. 60 days). Comparison between defects treated with biomaterial vs. clot per condition (non-alcoholic and alcoholic) (line, CCG vs. BCG and CEG vs. BEG).


Different letters (<sup>a</sup> , b ) indicate a statistically significant difference (*p* < 0.05), except for asterisks (\*) that do not show significant difference (CEG vs. BEG; *p* = 0.122).

**Table 4.** Mean ± standard deviation of volume density of new bone formation Comparison between same treatment per animal condition (non-alcoholic and alcoholic). Unpaired t-test (line, CCG vs. CEG and BCG vs. BEG).


Same lowercase letters indicate that there was no statistically significant difference. Significant differences *p* < 0.05.

#### *3.4. Influence of Clinical Condition (Alcoholic Versus non-Alcoholic) and*/*or Type of Treatment (Clot vs. Biomaterial) on Collagen Content During the Bone Repair Process*

Between 10–20 days, in the analysis of collagen matrix birefringence in polarized microscopy, all groups showed predominance of red-orange birefringence in the defect margins related to immature bone formation. In the BCG and BEG groups, the biomaterial presented red-orange birefringence surrounded by thin and disorganized collagen fibers (Figure 6).

**Figure 6.** Photomicrographies of birefringent fibers stained with Picrosirius red under polarized light at 10, 20, 40 and 60 days of repair. In the initial periods, non-alcoholic groups, CCG and BCG showed red-orange collagen fibers becoming yellow-green at the end of the experiment. Alcoholic groups demonstrated evidently disorganized bone collagenous matrix, with red-orange birefringence around the grafted biomaterial (asterisk) and adjacent to the defect edges, CCG; BCG, at 60 days. The xenogeneic graft (BCG; BEG) presented red-orange birefringence. Picrosirius red staining, original magnification ×5. Scale bar = 50 µm.

In the final periods, 40–60 days, in non-alcoholic groups (CCG and BCG), the bone collagenous matrix was more organized with green-yellow birefringence than the alcoholic animals (Figure 6).

#### **4. Discussion**

The growing increase in chronic alcohol consumption in the last decades has encouraged the development of numerous researches in the medical and dental areas to alleviate the deleterious effects of ethanol on bone loss due to osteoporotic conditions or with difficulty in consolidating extensive bone defects [13,14].

However, there are still few studies in the scientific literature evaluating its impact on the repair and osseointegration process of biomaterials. Thus, the results of this in vivo study showed that the biomaterial served as a scaffold for bone cells, a biological event that can attenuate the harmful effects of ethanol on the bone repair process.

Experimentally several bone defect sites are tested, but in rats the most commonly used is the critical defect in calvaria. The experimental animal model used in this study involving 5-mm bilateral craniotomies in the parietal bones has been used in scientific research because of its ability to produce paired analyzes in a standardized manner and allow the evaluation of bone substitute materials in the reconstruction of critical size bone defects without involvement of the sagittal suture [15].

In addition, the bones of the calvaria and face have intramembranous ossification, a typical location of dental defects [16]. It is also possible to take into consideration the constitution of the biomaterial tested, being particulate and of medium size (0.6–1.5 mm), more commonly indicated in smaller defects because, in large orthopedic defects, typical of long bones, it is preferable to the use of block grafts.

Initial studies have graphically described a U-shaped curve relating ethanol consumption to various chronic diseases, but the literature still remains controversial regarding bone tissue effects [17]. Thus, we adopted in this experiment the 25% (*v*/*v*) alcohol dosage, as we based on preliminary analyzes by De Souza et al. [18], Buchaim et al. [19] and [20] who reported destructive effects on bone with the use of 20% and 25% (*v*/*v*) ethanol without inducing animal death, contrary to the 5% dosage that had a protective effect [21].

In addition, previous studies by our research group evaluated the plasma concentrations of three ethanol dosages and their effects on bone repair, which showed that 25% alcoholization had pharmacologically relevant plasma concentration (540 mg/dL) to cause alterations bones that compromised the morphofunctional restoration of lost tissue [19].

In this study, we evaluated the effects of alcohol on body mass and new bone formation by descriptive and histomorphometric analysis by Masson's trichrome, and collagen fiber birefringence analysis by picrosirius-red staining in order to observe the alignment of the bone collagen and fiber structure at 10, 20, 40, and 60 days after injury.

The results of the body mass analysis showed lower averages in alcoholic than in non-alcoholic animals during the whole experimental period [20]. According to the literature consulted, the extensive use of alcohol can lead to dysfunctions in nutrient metabolism, causing changes such as decreased digestion and absorption, as alcohol has an influence on the stomach [22] and intestines [23] that can increase nutrient excretion, and consequently the risk of malnutrition [24,25]. Therefore, prolonged alcohol consumption may result in the lower body mass gained in animal experiments compared to non-alcoholic animals [26].

In the behavioral evaluation, some alcoholic rats initially presented aggressiveness and irritability, which are associated with hyperexcitability of the central nervous system, triggered by physiological dependence of alcohol [27,28].

All panoramic histological images of this experiment showed no integumentary invasion in the surgical bed, and the formation of fibrous connective tissue observed on the defect center and/or adjacent to the particles originated from the injured margins. Proper repositioning of the periosteum acted as a mechanical barrier preventing the collapse of extracranial tissues and consequently the migration of competing cells to osteoblasts [29,30].

Histomorphologically, defects filled with blood clot, CCG and CEG, presented the same pattern of bone repair, forming new bone tissue at the defect margins and extending centripetally, but with complete closure by fibrous connective tissue. This finding agrees with previous authors who suggest that it is a critical bone defect when spontaneous bone regeneration does not occur during the experimental period, requiring reconstruction of these defects by grafting [31].

In the initial periods, all experimental groups presented inflammatory infiltrate, a local defensive process against tissue aggression [32,33]. Between 20–40 days, the newly formed bone tissue was in transition from fine to compact trabeculae as observed by the study of Rocha et al. [34].

In the same period, BCG and BEG showed persistence of material implanted in the receptor bed with some dispersed inflammatory cells, without formation of macrophage aggregates and multinucleated giant cells, characteristic of chronic granuloma inflammatory process [35]. Thus, it is believed that the biomaterial proved to be biocompatible, i.e., the organism recognized the particles as part of its structure and not as an aggressor to its microenvironment [36,37].

At 60 days, the new bone formation remained restricted to the defect margins in all groups, with compact lamellar arrangement [38]. In BCG, bone islets adjacent to the particles were observed, which is consistent with studies that reported to be characteristic of osteoconductive biomaterial by providing a scaffold for osteoblastic cells, facilitating the deposition of new bone on its surface [39].

At the end of the experimental period, the inflammatory process present in the BCG and BEG groups was in the resolution phase, being more evident in non-alcoholic animals (BCG). These results lead to the perception of increase in serum levels of proinflammatory cytokines, derived from ethanol-induced liver disorders, which may have contributed to the persistence of the reaction tissue [40].

Histomorphometrically, in the analysis of the influence of time on bone formation, CEG showed a gradual increase in averages up to 40 and 60 days with statistically significant difference (Table 3, column). All other groups presented higher average in the final period of the experiment (60 days) in relation to the initial period (10 days), but without significant difference. Previous investigations have reported the physiological events that occur after the accommodation of biomaterial particles in the surgical bed, which alters the microenvironment and consequently delays the new bone formation [41].

Regarding the percentage of biomaterial in the BCG and BEG groups (Table 2, line), there was a decrease in final periods, but without significant difference. The delayed degradation of particles, as evidenced by their presence at 60 days, even in non-alcoholic animals, may be related to the intrinsic characteristics of the biomaterial [42]. This finding is corroborated by the study of Desterro et al. [43], who stated that non-sintered bovine apatites (<1000 ◦C) with organic matrix have lower dissolution rates, which directly impacts the biodegradation time.

Regarding the analysis of interference of the clinical condition, alcoholic and non-alcoholic, for each treatment, there was no statistical difference in all experimental groups (Table 4). In animals that used the biomaterial filling the bone defect, the results show that the formation of new bone was similar between the groups (BCG and BEG), regardless of the clinical condition (alcoholic or non-alcoholic), possibly for its osteoconductive property [39].

However, in comparing the percentage of bone formation according to treatment, clot or biomaterial (Table 3), CEG showed significant difference compared to BEG at 40 days (17.42 ± 2.78 vs. 9.59 ± 4.59, respectively). This fact may be correlated to the impact of ethanol on activities of the immune system, leading to changes in the phagocytic activity of polymorphonuclear cells. Thus, the persistence of particles may also have contributed to the delayed new formation in later periods compared to blood clot filling [44].

In the histochemical analysis of collagen fibers by picrosirius-red, all experimental groups initially presented red-orange birefringence, characteristic staining of formation of thin and disorganized collagen fibers turning greenish-yellow over the periods [45].

The biomaterial particles showed birefringence close to the newly formed bone tissue, precluding the measurement of specific fibers from bone repair. The study by Desterro [43] proved the presence of residual organic material in the particles of this biomaterial by X-ray diffraction analysis (XRD diffractogram of BonefillTM), justifying the markup in this analysis.

Based on the experimental model employed, it can be concluded that, regardless of the clinical condition, alcoholic or non-alcoholic, in the final period of the experiment, the process of bone defect recomposition was similar with the use of xenograft or only clot. The use of biomaterial can provide a scaffold that guides bone growth, especially in larger defects.

For prospective research in the field of tissue engineering, it is suggested to associate with biomaterials, plasma-derived biodegradable polymers such as PRP, PRF and fibrin sealants in order to make graft material moldable in the surgical bed, facilitate insertion and agglutination, and prevent its dispersal [46]. In addition, adjuvant and noninvasive methods are also recommended to accelerate and improve the regeneration process such as hyperbaric chamber, pulsed ultrasound (LIPUS), and laser photobiomodulation therapy [46–49].

#### **5. Limitations**

Knowing that ethanol causes β-catenin signaling pathway dysregulation by increasing/decreasing the activity or expression of its protein constituents, suggests that future studies employ molecular and biochemical analyses in order to detail the effects of ethanol on filled bone defects with biomaterial.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1996-1944/13/3/695/s1, Figure S1: Representative image of the methodology used to quantify the area of newly formed bone, graft particles, by Axio Vision software.

**Author Contributions:** Conceptualization—I.J.S.G. and R.L.B.; Data collection—I.J.S.G., K.T.P.; Formal analysis—I.J.S.G., K.T.P., D.d.B.T., A.C.C.B. and A.C.C.B.; Methodology—I.J.S.G., K.T.P., D.V.B., A.L.S., G.M.R.J., M.P.d.O.R.; C.H.B.R.; Resources—J.C.A. and R.L.B.; Supervision—R.L.B.; Visualization—I.J.S.G., K.T.P., D.V.B., A.C.C.B. and R.L.B.; Writing of original draft—I.J.S.G., K.T.P. and R.L.B.; Writing, review, and editing—I.J.S.G. and R.L.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brasil (CAPES)—Finance Code 001.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Materials* Editorial Office E-mail: materials@mdpi.com www.mdpi.com/journal/materials

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-3953-9