Influence of Geometry and Architecture on the In Vivo Success of 3D-Printed Scaffolds for Spinal Fusion

Abstract

We previously developed a recombinant growth factor-free, three-dimensional (3D)-printed material comprising hydroxyapatite (HA) and demineralized bone matrix (DBM) for bone regeneration. This material has demonstrated the capacity to promote re-mineralization of the DBM particles within the scaffold struts and shows potential to promote successful spine fusion. Here, we investigate the role of geometry and architecture in osteointegration, vascularization, and facilitation of spine fusion in a preclinical model. Inks containing HA and DBM particles in a poly(lactide-co-glycolide) elastomer were 3D-printed into scaffolds with varying relative strut angles (90° vs. 45° advancing angle), macropore size (0 μm vs. 500 μm vs. 1000 μm), and strut alignment (aligned vs. offset). The following configurations were compared with scaffolds containing no macropores: 90°/500 μm/aligned, 45°/500 μm/aligned, 90°/1000 μm/aligned, 45°/1000 μm/aligned, 90°/1000 μm/offset, and 45°/1000 μm/offset. Eighty-four female Sprague-Dawley rats underwent spine fusion with bilateral placement of the various scaffold configurations (n = 12/configuration). Osteointegration and vascularization were assessed by using microComputed Tomography and histology, and spine fusion was assessed via blinded manual palpation. The 45°/1000 μm scaffolds with aligned struts achieved the highest average fusion score (1.61/2) as well as the highest osteointegration score. Both the 45°/1000 μm/aligned and 90°/1000 μm/aligned scaffolds elicited fusion rates of 100%, which was significantly greater than the 45°/500 μm/aligned iteration (p < 0.05). All porous scaffolds were fully vascularized, with blood vessels present in every macropore. Vessels were also observed extending from the native transverse process bone, through the protrusions of new bone, and into the macropores of the scaffolds. When viewed independently, scaffolds printed with relative strut angles of 45° and 90° each allowed for osteointegration sufficient to stabilize the spine at L4-L5. Within those parameters, a pore size of 500 μm or greater was generally sufficient to achieve unilateral fusion. However, our results suggest that scaffolds printed with the larger pore size and with aligned struts at an advancing angle of 45° may represent the optimal configuration to maximize osteointegration and fusion capacity. Overall, this work suggests that the HA/DBM composite scaffolds provide a conducive environment for bone regeneration as well as vascular infiltration. This technology, therefore, represents a novel, growth-factor-free biomaterial with significant potential as a bone graft substitute for use in spinal surgery.

Impact statement

We previously developed a recombinant growth factor-free, three-dimensional (3D)-printed composite material comprising hydroxyapatite and demineralized bone matrix for bone regeneration. Here, we identify a range of 3D geometric and architectural parameters that support the preclinical success of the scaffold, including efficient vascularization, osteointegration, and, ultimately, spinal fusion. Our results suggest that this material holds great promise as a clinically translatable biomaterial for use as a bone graft substitute in orthopedic procedures requiring bone regeneration.

Introduction

Orthopedic procedures are among the most common surgeries performed annually in the United States.¹ Spinal fusions represent a significant portion of these procedures and address various disorders, including degenerative disk disease, disk herniation, segmental instability, trauma, and spinal deformities. Host bone taken from the iliac crest for implantation at the defect site (iliac crest bone autograft) generally contains the inherent osteogenic, osteoinductive, and osteoconductive capacity necessary for successful healing.^2,3 Despite this, clinical use of autograft is limited due to associated peri- and postoperative complications, that is, donor site morbidities, limited graft availability, and limited access to graft during minimally invasive procedures.^4–8

Bone graft “extender” products, such as ceramics or demineralized bone matrix (DBM), can be combined with autograft bone to reduce the volume of bone graft harvested intraoperatively, but these products do not promote high rates of fusion on their own.⁹ Bone graft substitutes, such as recombinant human bone morphogenetic protein-2 (rhBMP-2), can elicit high rates of fusion without the need for autograft or bone graft extenders. However, the supraphysiologic doses required for high fusion rates are associated with numerous complications, leading to a significant decline in use.^9–12 Developing a growth factor-free, bone graft substitute for spine fusion, therefore, remains a major goal.

Three-dimensional (3D)-printed scaffolds of the appropriate material composition can act as biocompatible and bioactive structures that provide a template for cell attachment, vascularization, and stimulate the formation of bone.¹³ In designing regenerative implants, the interaction between cells and the foreign biomaterial is critically important; manipulating pore size and geometry can have a direct impact on progenitor cell adhesion, viability, and differentiation, ultimately influencing the regenerative capacity and clinical success.^14,15 In the case of porous titanium implants, for example, pore size influences the proliferation and differentiation of bone marrow stromal cells and bone-healing capacity.^16–19 Macropores >300 μm have been shown to enhance bone formation and vascularization, and micropores (<10 μm) can promote osteoinduction via increased surface contact and protein adsorption.^20,21 The capillary force provided by the micropores can also improve the attachment of osteogenic cells to the scaffold surface.^18,19

Pore geometry—as determined by relative strut angle and layer alignment—is similarly believed to influence regenerative potential. For instance, polycaprolactone-tricalcium phosphate scaffolds printed with a 90° relative strut angle produced significantly greater and stiffer new bone matrix relative to those printed with 60° angles, resulting in improved mechanical strength.²² A layered strut alignment results in “through-pores,” an architectural modification that results in distinct differences in mechanical properties²³ and could impact cell seeding/attachment efficiency.²⁴ For radiopaque materials, strut alignment or lack thereof can also impact postoperative imaging of the defect site; aligned struts may allow for areas of radiolucency, whereas scaffolds printed with offset struts might be radiopaque on imaging, even if the struts were widely spaced (i.e., even with large macropores).

We previously developed a 3D-printable elastomer-hydroxyapatite (HA) scaffold system for spine fusion,²⁰ now commercially termed Hyperelastic Bone^®. We recently extended the material system to an HA-DBM composite material for bone regeneration,^20,21,25 where the DBM serves as an osteoinductive component, and the HA provides inorganic mineral ions and imparts mechanical strength and osteoconductivity. We found that when both components were present, de novo mineralization of the near-surface volumes of DBM particles occurred within the scaffold struts.²⁶ The formation of these bone-like “spicules” did not occur in scaffolds printed from inks that contained only DBM or only HA, suggesting that in this setting, the osteoinductive nature of the DBM is sufficient to induce de novo mineralization only when sufficiently high concentrations of Ca²⁺ and PO₄⁻ are present in the immediate vicinity.

One iteration of this HA/DBM scaffold, made from ink containing a 3:1 ratio of HA:DBM, elicited an impressively high fusion rate (92% success) in the rat preclinical model of spine fusion. In this study, we investigate the influence of scaffold geometry and architecture of the 3:1 HA:DBM material on osteointegration of the scaffold with native tissue in the rat spine fusion model. As cells may respond differently to different local scaffold microenvironments, we compare relative strut angles of 90° versus 45°, strut spacings of 1000 μm versus 500 μm versus 0 μm, and aligned versus offset strut layers. The capacity for implant vascularization was also assessed, with a goal to determine whether certain architectural parameters provided a better bone regenerative environment in support of successful spine fusion.

Materials and Methods

Hydroxyapatite/demineralized bone matrix ink synthesis and scaffold fabrication

Medical-grade synthetic HA powder (particle diameters nominally 1–25 μm) was provided by Merz North America. DBM particles (∼100–1000 μm) were provided by Xtant Medical (Belgrade, MT). DBM particles were milled in-house by using a cutting-mill (Thomas Scientific) to reduce their size and were subsequently sieved to obtain DBM powder (particle diameters <80 μm). HA/DBM ink for 3D printing was synthesized based on previously developed and described protocols.²⁰ Briefly, polylactide-co-glycolide (PLG; 82:18 glycolide to lactide) copolymer (Evonik Cyro) was dissolved in dichloromethane (DCM; Sigma), under ambient conditions (∼10 mL DCM per gram PLG). Separately, a 10:2:1 mass mixture of DCM, 2-butoxyethanol (2-Bu; Sigma) and dibutylphthalate (Sigma) was prepared. The HA and sieved DBM powders were added to the solvent mixture and vortexed for ∼2 min. The PLG-DCM and the powder/solvent suspension were then combined and vortexed for 5 min.

The HA/DBM ink was prepared by controlling the relative volumetric amount (based on weight and density; ρ_HA = 3.15 g/cm³; ρ_DBM = 1.05 g/cm³) of each powder within the ink. The PLG content was 30 vol.% (ρ_PLG = 1.15 g/cm³), and the solid loading of 70 vol.% of HA+DBM comprised three parts of HA: one part of DBM. The mixture was then allowed to thicken via ambient temperature pressure evaporation of the excess DCM and was periodically stirred by hand until a viscosity of ∼30 Pa·s was achieved.²⁰

A 3D-Bioplotter (EnvisionTEC, GmbH, Germany) was used to print the scaffolds via room-temperature, pneumatic extrusion by using a 410 μm diameter (22 Ga) stainless steel nozzle. Scaffolds 5 layers thick were printed into one of seven different architectural configurations (Table 1). One scaffold was solid and contained no macropores, and the remaining designs were printed with a combination of the following parameters: Each successive layer of struts was oriented at either 90° or 45° relative to the underlying layer; the edge-to-edge strut spacing was either 500 or 1000 μm, with this distance defined as the macropore size.

Table 1.

Hydroxyapatite/Demineralized Bone Matrix Scaffold Architectural Iterations Evaluated in the Rat Spine Fusion Model

Scaffold ID/Tx group	Relative strut angle	Macropore size	Strut alignment	n per group
90°/500 μm/aligned	90°	500 μm	Aligned	12
45°/500 μm/aligned	45°	500 μm	Aligned	12
90°/1000 μm/aligned	90°	1000 μm	Aligned	12
45°/1000 μm/aligned	45°	1000 μm	Aligned	12
90°/1000 μm/offset	90°	1000 μm	Offset	12
45°/1000 μm/offset	45°	1000 μm	Offset	12
No-macropores	N/A	0 μm	N/A	12

NA, not applicable.

To assess any impact of tortuosity, two additional scaffold designs were printed by using an offset strut configuration for each progressive layer, creating scaffolds without through pores (Fig. 1A and Table 1). After printing, the scaffolds were washed for 4–5 h in 70% ethanol (exchanging the ethanol solution half-way through) followed by three 5-min rinses in sterile phosphate buffered saline (PBS). Subsequently, the scaffold sheets were refrigerated in sterile PBS until implantation. The scaffolds were cut to size (12 × 4 mm) by using a digital caliper before implantation.

FIG. 1.

Representation of scaffolds. (A–D) Schematic representations depicting top-down and cross-sectional side views of (A) 90°/aligned, (B) 90°/offset, (C) 45°/aligned, and (D) 45°/offset scaffolds. Note that the 45° scaffold cross-sections will appear different depending on where the “cut” is made; only one cross-sectional view of each is depicted here. The lighter gray color represents struts that are farther away relative to the darker struts. ω—relative strut angle (90° or 45°). w—strut-to-strut distance, that is, “macropore size.” (E) Representative 12-week postoperative radiographs from each treatment group depicting the scaffolds implanted between the L4 and L5 transverse processes.

Animal groups and surgical procedure

Eighty-four female Sprague-Dawley rats, ages 12–16 weeks, were randomly assigned to one of seven treatment groups (n = 12 per group) (Table 1). Animals received bilateral implants of one of the HA/DBM scaffold designs described in the previous subsection, in Figure 1 and in Table 1. Approval was obtained from the Institutional Animal Care and Use Committee under protocol IS00002610, and all procedures were conducted in compliance with the NIH Guide for Care and Use of Laboratory Animals.

Rats were maintained under anesthesia using an isoflurane inhalational anesthetic delivery system and were monitored by an assistant for cardiac or respiratory anomalies while maintaining a temperature range of 35.9–37.5°C throughout the procedure. The posterolateral spine fusion procedure was performed as previously described.^27–31 A posterior midline incision was made over the lumbar spinous processes, followed by two separate fascial incisions 4 mm from the midline. The L4 and L5 transverse processes were exposed via blunt dissection to create space for graft implantation. The fusion bed was irrigated with gentamicin/saline solution, a high-speed burr was used to decorticate the posterior surfaces of the L4-L5 transverse processes, and grafts were placed bilaterally to bridge the L4-L5 intertransverse space.

Fascial incisions were closed with 3–0 monocryl absorbable sutures by using a simple interrupted pattern, and skin incisions were closed by using wound clips. Prewarmed lactated ringer solution was administered for fluid replacement via intraperitoneal injection, and buprenex and meloxicam were administered for 3 days postoperatively. Animals were housed in separate cages and allowed to eat, drink, and bear weight ad libitum.

Radiography and fusion scoring

All animals underwent plain anteroposterior radiography at 10 days postoperatively (after the removal of wound clips) and before euthanasia at 12 weeks postoperatively (Fig. 1B). After euthanasia, lumbar spines were harvested en bloc and fixed in 10% neutral-buffered formalin for 10 days. Specimens were then placed in 50% ethanol (EtOH) for 2 h and were subsequently stored in 70% EtOH. All spines were scored for fusion via manual palpation by three independent, blinded observers using a scoring system in which “0” indicates motion between segments bilaterally, “1” indicates no motion between segments unilaterally, and “2” indicates no motion between segments bilaterally.²⁰ An average fusion score of ≥1 (unilateral fusion) was considered a successfully fused spine.

Specimen preparation for study of vascularization

Blood vessel infiltration into the macropore spaces of the scaffolds was examined by perfusing animals with Microfil^® contrast agent (lead chromate-containing polymer) following the manufacturer's protocol (Flow Tech, Inc., Carver, MA) at the study endpoint (12 weeks postoperatively). Briefly, heparinized 0.9% saline was perfused through a needle inserted into the left cardiac ventricle, followed by pressure fixation of the blood vessels with 10% neutral buffered formalin and flushing with heparinized saline. Microfil was then perfused through the vasculature and allowed to polymerize overnight. For each of the treatment groups receiving implants printed with aligned struts, n = 2 animals were perfused. For the offset groups (1000 μm strut spacing), n = 3 for 90° architecture and n = 4 for 45° architecture were perfused.

Higher resolution synchrotron microComputed Tomography (microCT) (see below) required sample cross-sections smaller than that of the rat spine. Therefore, spines were bisected longitudinally to fit the field of view (FOV) of the instrument, and each fusion bed (hemispine) was imaged individually. Two hemispines were imaged for each scaffold with aligned struts, and n = 7 and n = 6 hemispines were imaged for 45° and 90° offset scaffolds, respectively.

microComputed Tomography

microCT was used to image the explanted spines and their scaffolds in 3D and to investigate the following: (1) bone ingrowth into scaffolds (osteointegration); (2) integrity of the scaffold struts, as well as the presence and morphology of bone-like spicules within the scaffolds; and (3) infiltration of blood vessels into the scaffold macropore spaces. Different microCT instruments were used, depending on the spatial resolution and width of the FOV required for each of outcome measures, but the entire length of the scaffold was imaged in all cases.

Laboratory microCT using a Scanco MicroCT-40 system and reconstructions with 30 μm isotropic volume elements (voxels) were used to quantify osteointegration; these measurements were supplemented by synchrotron microCT imaging with 6 μm isotropic voxels at beamline 5-BM, the Advanced Photon Source (APS). Scaffold integrity, bone-like spicule formation, and blood vessel growth within the scaffolds were all studied by using synchrotron microCT with 1.3 or 3.25 μm isotropic voxels at beamline 2-BM, APS. Even after bisection, the specimens' transverse width was still larger than the width of the x-ray beam (FOV), so a local (i.e., region of interest) tomography approach was employed,³² with each FOV centered on the scaffold.

Bone ingrowth into each scaffold was quantified by using the osteointegration scoring system developed earlier by the authors²⁶; six hemisected spines were imaged from each group. A score of 0 denoted no bone ingrowth beyond the proximal (ventral) surface of scaffold; 1 signified bone growth beyond the proximal surface, but not beyond the first scaffold strut; 2 denoted a scaffold with bony growth extending beyond the first scaffold strut, but not the second; and 3 represented scaffolds with bone growth beyond the second scaffold strut.

Three regions within each scaffold were scored separately: the medial end of the scaffold (the “medial endcap,” adjacent to the vertebral body), the scaffold body, and the lateral end of the scaffold (the “lateral endcap”). Scores from each region were summed to generate a total ingrowth score per scaffold, all of the slices covering a scaffold were examined, and the highest score observed for each region was used in computing the total score.

The presence and morphology of bone-like spicules²⁶ were visualized in reconstructed data sets from 2-BM of the APS as previously done. All scaffold architectures were examined in the analysis. Hemi-spines were perfused with Microfil and were imaged at 2-BM, APS. The presence of blood vessels within macropore spaces was determined for at least two Microfil-perfused hemi-spines for each group. Each slice of each FOV was examined to identify any macropores that did not contain a blood vessel. Because macropores are 3D structures, this determination required scrolling through the groups of slices containing each macropore. In representative hemi-spines, the 3D network of blood vessels was visualized by using Amira (Thermo Fisher Scientific), in which bone, scaffold, and vessels were each segmented separately to generate the rendering.

Histology

After microCT imaging, tissue types within scaffold macropores were visualized in representative fused spines with 90°/1000 μm/aligned and the 45°/1000 μm/aligned implants (n = 4 hemi-spines per group). Tissue processing was performed by the Northwestern Mouse Histology and Phenotyping Laboratory core facility, where spines were placed in Immunocal 12% formic acid solution (Statlab) for 72 h and processed by using an ASP300S automated tissue processor (Leica Biosystems) for tissue dehydration, clearing, and paraffin infiltration. Specimens were then embedded in paraffin in a sagittal orientation and sectioned onto glass slides at a thickness of 5 μm. Staining was performed by using a solution of Gill modified hematoxylin (Millipore), Alcian blue, Orange G, and Eosin Y (Sigma-Aldrich).

Statistics

Statistics were performed by using SPSS. Continuous variables were compared by using one-way analysis of variance (ANOVA) for normally distributed data. Tukey's test was used for post hoc analyses. Fisher's exact test was used to compare the categorical data. A p-value <0.05 was considered statistically significant.

Results

Assessment of spine fusion

One rat from the 45°/1000 μm/aligned treatment group died in the immediate postoperative period, which was attributed to anesthesia-related complications. In the remaining 83 rats, radiographic imaging showed bilateral placement of the scaffolds at the L4-L5 transverse processes (Fig. 1E). All groups with scaffolds containing macropores demonstrated significantly higher mean fusion scores than the no-macropore group (p < 0.05) (Fig. 2A) (Table 2). The highest fusion score was found in the 45°/1000 μm/aligned group (1.61 ± 0.40), which was significantly higher than that of the 45°/500 μm/aligned group (0.92 ± 0.65) (p < 0.05). The no-macropore group had the lowest mean fusion score (0.33 ± 0.40).

FIG. 2.

Assessment of fusion. (A) Fusion scores as determined via manual palpation by three blinded examiners. Scores were based on detectable motion or lack thereof between the L4-L5 segments bilaterally. ω: relative strut angle; w: macropore size; A/O: A = aligned struts, O = offset struts. (B) Fusion rates were calculated based on fusion scores, where unilateral lack of motion was considered successful fusion (i.e., a mean score of ≥1.0). *Denotes a significant difference relative to all treatment groups other than the 45°/500 μm/aligned group. ^ denotes significant difference relative to both the 90°/500 μm/aligned and the 45°/1000 μm/aligned treatment groups. (C–E) Subgroup analysis of fusion scores by strut angle and strut spacing. (C) Fusion scores by strut angle (90° vs. 45°) including both aligned and offset scaffolds. (D) Fusion scores by strut angle (90° vs. 45°) excluding offset scaffolds in the analysis. (E) Fusion scores by strut spacing (500 μm vs. 1000 μm macropores), depicting analyses both including (A/O) and excluding (A only) scaffolds with offset struts. (F–H) Subgroup analysis of fusion rates, with categorization by strut angle and strut spacing as shown in C-E (*p < 0.05).

Table 2.

Fusion Scores and Rates

Treatment group	n	No. of fused	Fusion rate	Fusion score
90°/500 μm/aligned	12	12	100%	1.42 ± 0.38
45°/500 μm/aligned	12	7	58%	0.92 ± 0.65
90°/1000 μm/aligned	12	10	83%	1.42 ± 0.62
45°/1000 μm/aligned	12	12	100%	1.61 ± 0.40
90°/1000 μm/offset	11a	9	82%	1.30 ± 0.43
45°/1000 μm/offset	12	8	67%	1.15 ± 0.43
No-macropores	12	2	17%	0.33 ± 0.40

^aOne animal from this treatment group died in the immediate postoperative period.

When categorized by strut angle alone, no significant difference in fusion scores was observed between scaffolds with 45° or 90° struts or between scaffolds with 500 or 1000 μm macropores (Fig. 2C–E). This same result was noted whether the analysis was performed on all implants (i.e., the analysis included both aligned and offset groups) as well as when the analysis was performed only on those treatment groups receiving implants with aligned struts. Therefore, strut alignment appears not to influence the impact of macropore size on fusion score. Similarly, when the influence of relative strut angle on fusion score was investigated, no significant differences were found; again, this result was independent of strut alignment or lack thereof.

All groups except for the 45°/500 μm/aligned group had significantly higher fusion rates than the no-macropore group (p < 0.05). The no-macropore group had the lowest fusion rate (17%). Both the 90°/500 μm/aligned and 45°/1000 μm/aligned had fusion rates of 100%, which was significantly higher than that for the 45°/500 μm/aligned treatment group (p < 0.05). As with fusion score analyses, categorization according to only macropore size or only strut angle found no significant impact of either on fusion rates (Fig. 2F–H).

New bone formation and scaffold osteointegration

Figure 3 shows representative synchrotron microCT slices recorded at 2-BM, APS for each group. In 12-week explants, strut organization remains mostly intact (Fig. 3A, B). As previously reported,²⁶ new bone-like spicules formed within the near-surface volume of the DBM particles within scaffold struts. The spicules had microCT contrast indistinguishable from that of native bone (Fig. 3B–D). Figure 3C shows enlarged images of the cross-sections of struts from each group, with fine bone spicules that generally follow DBM particle surfaces. Manual segmentation shows these spicules in magenta (Fig. 3D). These spicules extend above and below the slice plane. Qualitatively, the morphology of the spicules in this study (i.e., in 12-week explants) did not differ from that in the earlier study (8-week explants).²⁶

FIG. 3.

Synchrotron microCT. (A) Representative synchrotron microCT images acquired at 2-BM, APS, depicting an axial cross-sectional view of scaffold architecture 12 weeks after implantation. Each row represents a different treatment group. Osteointegration scores were determined by the extent of bone growth from the transverse process into scaffold macropores. For example, the 45°/1000 μm/offset representative image would produce an osteointegration score of 2 (0 for the lateral endcap +1 for the scaffold body +1 for the medial endcap), since the bony fingers are seen extending beyond the proximal surface of the scaffold but not beyond the first scaffold strut. Scale bars = 1 mm. (B) Representative synchrotron microCT images of scaffolds from each treatment group, acquired at 5-BM, APS. These scaffolds differ from those shown in (A), and there is a narrower region of interest. Scale bars = 500 μm. (C) Zoomed ROI of yellow box from (B), showing individual struts. Bone spicules can be seen around spaces originally containing DBM particles within the struts. HA particles are seen as bright white. The yellow arrow in the 45°/1000 μm/offset image highlights new bone that has grown completely around a scaffold strut. Scale bars = 250 μm. (D) The same zoomed ROI as depicted in (C), with new bone spicules segmented and false colored in magenta. APS, Advanced Photon Source; DBM, demineralized bone matrix; HA, hydroxyapatite; microCT, microComputed Tomography.

Bony fingers extending from the transverse processes can be seen growing into scaffold macropores in Figure 3A. Because of the limited FOV of the 2-BM, APS reconstructions, laboratory microCT (data not shown) was required to image enough of the scaffold and adjacent transverse process bone to determine the osteointegration score. With the exception of the spicules, all bone was contiguous with the transverse processes or nearby lamina.

The bone growing from the transverse processes into the macropores generally did not make direct contact with the HA particles, even when entire struts were partially or entirely encapsulated (e.g., Fig. 3A, row four, lower left of scaffold). As expected, all groups with porous scaffolds had significantly higher average osteointegration scores relative to the group receiving non-macroporous scaffolds (Fig. 4). The 45°/1000 μm/aligned treatment group had the highest average osteointegration score (9.5). This score was significantly higher than that of the 45°/500 μm/aligned group (4.0), but it was not statistically different from the other porous scaffold groups.

FIG. 4.

Osteointegration scoring. Osteointegration scores of scaffolds were summed and averaged for each group (maximum score of 18 per spine). All scaffold iterations with macropores demonstrated a significantly higher average osteointegration score relative to the no macropore group. The 45°/1000 μm/aligned scaffold group demonstrated the highest average osteointegration score, which was significantly greater than that of the 45°/500 μm/aligned scaffold (*p < 0.05), but it was statistically equivalent to the average scores for the other scaffold iterations. Osteointegration scoring was performed for n = 6 spines (n = 12 scaffolds) per group.

Assessment of vascularization

Representative 2-BM slices for Microfil-perfused samples are shown in Figure 5, and the structures are labeled. In each of the Microfil-perfused hemi-spines imaged, every macropore was occupied by a least one blood vessel. In contrast, none of the nonporous scaffolds was vascularized. Representative 3D renderings (Fig. 5) illustrate the complex 3D network of blood vessels in the porous scaffolds; the bone and the HA particles of the scaffold are shown as semi-transparent to allow for clearer visualization of the blood vessels.

FIG. 5.

Vascular ingrowth. Representative microCT reconstructions of segments from a 45°/1000 μm/aligned scaffold (A–C) and a 90°/1000 μm/aligned scaffold (D–F) harvested from Microfil^®-perfused rats. Bone (yellow) and scaffold struts (white) are shown with progressively greater transparency from A to C, to allow for better visualization of the blood vessels (red). Multiple vessels can be seen within every macropore in both scaffolds here, as well as in every other porous scaffold evaluated. In (A–C), a bony finger can be seen growing out from the transverse process and between scaffold struts. A vessel protruding from the native transverse process bone is extending into the macropore space. Scale bars = 1 mm.

Histological analysis

Figure 6 shows stained thin sections of a representative 90°/1000 μm/aligned scaffold (Fig. 6A–C) harvested at 12 weeks, which is compared with one harvested at 8 weeks postoperatively (Fig. 6D). At the 12-week time point, the degree of strut integrity—that is, maintenance of the circular shape of the cross-section—qualitatively appears somewhat lower that of 8-week explants (see Fig. 6 and²⁶), suggesting that some early degradation of the polymer may be occurring.

FIG. 6.

Histology. Decalcified spines were sectioned to visualize the constituents of the 12-week postoperative scaffolds. (A) A 90°/1000 μm/aligned scaffold from a 12-week explant. Scale bar = 2 mm. (B, C) Magnified regions of interest from black boxes in (A) showing cells (purple) infiltrating partially degraded struts and making contact with a DBM particle (stained in red, highlighted by black circle). HA particles appear as small black dots. Scale bar = 300 μm. (D) Similar scaffold explanted from an 8-week time point from a previous study. Note that the strut appears less degraded and has a more rounded shape. Scale bar = 300 μm.

Discussion

3D printed materials—including ceramic composites—have shown significant promise for promoting bone regeneration.^33–35 Among other advantages, the tunable architecture of 3D-printed scaffolds affords the flexibility to optimize functions such as vascularization and osteointegration. In earlier work, we reported the development of 3D-printed HA/DBM composite scaffolds, which were printed from HA-PLG inks containing varying amounts of DBM.²⁶ In that work, scaffolds were printed with 500 μm macropores and with aligned struts laid at 90° relative angles. We found that the material showed potential as a bone graft substitute for spine fusion, with fusion scores directly correlating with osteointegration score.

This study utilized the best performing HA/DBM composition from that original work to examine whether 3D geometry and architecture influence osteointegration and vascularization in the same rat posterolateral spine fusion model. Quantitative measures were manual palpation-based fusion scoring and microCT-based osteointegration scoring. Other observations included evaluation of the extent of vascularization, observation of bone-like spicules, and histological identification of tissue within the scaffold macropores.

Progenitor cell adhesion, infiltration, and differentiation—as well as a robust vasculature—are believed to be important factors in establishing an environment conducive to bone formation. With the understanding that the physical structure of a scaffold can impact these factors, and based on previous published work, we anticipated that scaffolds bearing struts printed at 90° relative angles and with through pores (i.e., aligned instead of offset struts) would provide the most efficient path for cell infiltration, mineralization, and vascularization, and would produce maximal regenerative capacity. Instead, all of the macroporous HA/DBM scaffold architectures examined produced statistically equivalent fusion scores. Overall, the fusion rates with the macropore-containing scaffolds are remarkable, given that these scaffolds do not contain recombinant growth factor nor do they carry other osteoinductive pharmaceuticals.

Notably, fusion rates of 100% were achieved in both the 90°/500 μm/aligned and the 45°/1000 μm/aligned treatment groups. The latter may represent the best architecture for promoting bone regeneration and spine fusion, as it was the only scaffold design whose fusion rate and osteointegration score were significantly higher than those of the 45°/500 μm/aligned scaffold (which was the poorest performing macropore-containing scaffold). On balance however, the differences in osteointegration and fusion success were marginal, and all of the macropore-containing scaffolds generally performed well.

The finding that scaffolds without macropores produced the worst performance was unsurprising, since other studies have shown that the macroporosity (as opposed to microporosity alone) afforded by 3D-printed scaffolds provides superior osteointegration. For example, a 3D-printed porous titanium interbody cage outperformed a plasma sprayed (micro) porous titanium-coated polyether ether ketone cage in a sheep spine fusion model, with superior biomechanical stability.³⁶ Moreover, side-wall pore architecture has been shown to be required for optimal repair of calvarial defects in rabbits.³⁷

In our study, all macroporous scaffolds contained blood vessels within every macropore. Further, excepting the formation of bone-like spicules within the scaffold struts, new bone was observed growing only from the pre-existing bone. At the scale of synchrotron microCT, little contact was observed between native bone and HA or DBM particles. Despite this, the high fusion rates suggest that within certain ranges, macropore size, strut angle, and strut alignment do not play a major role in the HA/DBM scaffold success in the setting of bone regeneration and spine fusion. Rather, the osteoconductive and osteoinductive properties of the scaffold composite material may be the most important factors for promoting bone growth.

The poor performance of the no-macropore group demonstrates that both osteoconductive/osteoinductive material and a porous structure are required for successful stabilization of adjacent vertebrae when employing the HA/DBM material. Apparently, new bone tends to incorporate into rather than simply around the scaffold. Thus, osteointegration of the scaffold appears to be both necessary and sufficient for stabilization of the fused segment.

The importance of architectural design in 3D-printed bone regenerative composites has been investigated elsewhere, where consistent with our results, superior bone ingrowth into scaffold macropores is generally seen with dimensions greater than 250–300 μm.^18,38,39 For example, Lee et al. found that 500 μm macroporous scaffolds produced greater osteointegration and bone regeneration relative to scaffolds with 250 μm macropores in rats.³⁹ However, Entezari et al. has shown that in a rabbit calvarial defect model, bone ingrowth was highest in scaffolds with pore sizes greater than 390 μm, but less than 590 μm in diameter.³⁸ Therefore, the surgical setting, animal mode, and/or mode of bone formation (endochondral vs. intramembraneous) may also impact the “ideal” window for pore size.

At 12 weeks postoperatively, we observed cells beginning to permeate the scaffold material (Fig. 6), but the polymer elastomer is still mostly present. microCT imaging of the entirety of the scaffolds shows they are still relatively intact for all architectures, indicating that scaffold breakdown is only in the very early stages.

Ideally, a scaffold would immobilize adjacent vertebral segments until the scaffold could be replaced by bone bridging the segments, and it may be that persistence of the elastomer binder and the in vivo insolubility of HA may be hindering replacement of the scaffold with bone. Further, osteointegration (bone fingers growing into the macropores, as well as bone encapsulating the medial and lateral surfaces of the scaffolds) produces a rigid transverse process-scaffold system, and one speculates that the resulting local environment for bone growth might be improved by increasing the rate of scaffold resorption and concomitantly decreasing its stiffness. This might be done by replacing the polymer binder and/or calcium phosphate phase with more soluble alternatives, increasing material/fiber porosity,²⁵ and/or by printing with a finer nozzle, thereby increasing the scaffold surface area-to-volume.

The bone-like spicule morphology and overall content of the 12-week explants does not appear to differ from that seen in the 8-week explants in our previous work.²⁶ This suggests that the rate of spicule formation has greatly decreased, perhaps because the HA and phosphate ions are blocked (by the spicule) from reaching the unreacted interior of the DBM particles. The volume fraction of spicules might be increased, therefore, by decreasing the diameter of the DBM particles or otherwise increasing surface area, while keeping the DBM volume fraction constant.

Increasing the spicule volume fraction may be important to rapid integration of adjacent transverse processes, because the (presumed ectopically formed) calcium phosphate mineral that comprises the spicules could become a target for bone remodeling processes. This newly remodeled bone could then act as a template scaffold for further bone outgrowth at locations distinct from the transverse processes. Formation of such islands of bone (containing osteocytes and canaliculi) may be a way of greatly speeding the bony bridging of segments intended for fusion.

The HA/DBM composite described here represents a novel, recombinant growth factor-free biomaterial for promoting bone regeneration and spinal fusion. Within the range of geometries and architectures tested, the scaffolds demonstrated consistent and successful spine fusion in a preclinical rat model, suggesting that the material has significant potential as a bone graft substitute for use in spinal surgery. A preclinical comparative efficacy study is currently underway to evaluate the biomechanical performance of the 3D-printed HA/DBM composite alongside the established positive control and industry standard bone graft substitute (rhBMP-2/ACS; INFUSE™). Mechanistic assessments are also ongoing, as it will be important to establish the safety of this material for use in the setting of spine fusion.

Disclosure Statement

M.H., A.D., R.L., K.C., M.H., R.P., S.J., C.Y., W.H., S.R.S., and E.H. have no commercial associations that might create a conflict of interest in connection with this work. A.J. and R.S. are co-founders of and shareholders in Dimension Inx, LLC, which develops and manufactures new advanced manufacturing compatible materials and devices for medical and nonmedical applications, including some of the materials on which this work is based (i.e., Hyperelastic Bone). As of August 2017, A.J. is currently full-time Chief Technology Officer of Dimension Inx, LLC, and R.S. serves part time as Chief Science Officer of Dimension Inx LLC. A.J. and R.S. are inventors on relevant patents that are licensed to Dimension Inx LLC. Dimension Inx owns the trademark for Hyperelastic Bone. Dimension Inx LLC did not influence the conduct, description, or interpretation of the findings in this article. No other authors have commercial interests in the materials described in this work.

Funding Information

This study was supported by the National Institute of Arthritis, Musculoskeletal, and Skin Diseases, grant R01AR069580. This research used resources of the Northwestern University Center for Advanced Microscopy Core Facility (National Cancer Institute Cancer Center Support Grant P30 CA060553 to the Robert H. Lurie Comprehensive Cancer Center); Histology and Phenotyping Core Facility (Robert H. Lurie Comprehensive Cancer Center support grant NCI CA060553); and the Rush University MicroCT and Histology Core Facility. Portions of this work were performed at the DuPont-Northwestern-Dow Collaborative Access Team (DND-CAT) located at Sector 5 of the Advanced Photon Source (APS). DND-CAT is supported by Northwestern University, The Dow Chemical Company, and DuPont de Nemours, Inc. This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357.