3D-Ink-Extruded Titanium Scaffolds with Porous Struts and Bioactive Supramolecular Polymers for Orthopedic Implants

Abstract

Porous Ti addresses the longstanding orthopedic challenges of aseptic loosening and stress shielding. This work expands on the evolution of porous Ti with the manufacturing of hierarchically porous, low stiffness, ductile Ti scaffolds via direct-ink write (DIW) extrusion and sintering of inks containing Ti and NaCl particles. Scaffold macrochannels were filled with a subtherapeutic dose of recombinant bone morphogenetic protein-2 (rhBMP-2) alone or co-delivered within a bioactive supramolecular polymer slurry (SPS) composed of peptide amphiphile nanofibrils and collagen, creating four treatment conditions (Ti struts: microporous vs. fully dense; BMP-2 alone or with SPS). The BMP-2-loaded scaffolds were implanted bilaterally across the L4 and L5 transverse processes in a rat posterolateral lumbar fusion model. In-vivo bone growth in these scaffolds is evaluated with synchrotron X-ray computed microtomography (μCT) to study the effects of strut microporosity and added biological signaling agents on the bone formation response. Optical and scanning electron microscopy confirms the −100μm space-holder micropore size, high-curvature morphology, and pore fenestrations within the struts. Uniaxial compression testing shows that the microporous strut scaffolds have low stiffness and high ductility. A significant promotion in bone formation was observed for groups utilizing the SPS, while no significant differences were found for the scaffolds with the incorporation of micropores.

Graphical Abstract

1. Introduction

Titanium (in pure and alloyed form, e.g., CP-Ti and Ti-6Al-4V) is used extensively in orthopedic and dental implants due to its excellent biocompatibility and load-bearing properties [1,2]. However, Ti has drawbacks in hosting osseous tissue, due to low bioactivity and high elastic modulus (100-110 GPa) compared to cortical bone (~20 GPa) and cancellous bone (< 4 GPa), which leads to stress shielding and aseptic loosening [3,4]. Porous Ti addresses these issues by reducing the stiffness of the implant and increasing osseointegration as bone grows into and interlocks with the implant. Elastic modulus is a material-intrinsic phenomena that is independent of porosity, while stiffness incorporates the geometrical design into the ability to resist deformation on a macroscopic level.

Powder metallurgy (PM) has been the predominant method of manufacturing porous implants and includes powder sintering/hot-pressing [5], spark plasma sintering [6], metal injection molding (MIM) [7], freeze casting [8] and additive manufacturing (AM) [9]. The latter includes beam-based methods (e.g., selective laser melting (SLM) [10], direct energy deposition (DED) [11], electron beam melting (EBM)[12]), and sintering-based methods (e.g., inkjet printing [13] and direct-ink write (DIW) ink extrusion or bound-metal deposition techniques [14-17]), allowing for the creation of Ti implants with designed open lattice architectures. DIW exhibits high versatility in alloy compositions and microstructures, which is less easily achievable with other means [18-20].

One can design high levels of porosity in PM (rather than residual porosity from partial sintering) by using a metallic powder and adding a space holder powder, also known as a poragen (e.g., sugar [21], Mg [22] or salts [23-26]), which is removed before or after sintering or densification. Recent developments in the AM field have focused on controlled multi-level porosity (i.e., hierarchical porosity) which can be achieved by introducing a microporous network within 3D-printed struts via space holders. Partial sintering is one technique able to achieve two levels of porosity [16] and is suited to Ti [21], with large and angular micropores [22,23] which serve as individual anchoring locations for osteoblasts, enabling localized inorganic carbonated apatite interlocking [24] with the Ti.

Macrochannels, defined here as elongated pores >300 μm in diameter, are expected to promote cell infiltration, migration, and vascularization via the movement of nutrients, waste, and oxygen, allowing osteoblasts to colonize the internal regions of latticed implants [25]. Micropores, defined as pores with diameter of ~100 μm, can enhance fixation of bone trabeculae and osteons formed by the osteoblasts and osteocytes, increasing the long-term stability of the implant [26]. Connections between pores, known as fenestrations, play a central role in the depth of penetration and final anchoring location of bone cells [27]. Fenestrations tend to be the limiting regions of transport, as they are smaller than the full-size pores which they connect. While there is interest in the role of < 10 μm micropores for protein adsorption and high capillary forces [28], osseointegration and angiogenesis are thought to be enhanced with pores in the range of 75-135 μm [29]. Pore shape is also important: high-curvature pores can be achieved through the use of salt space holder particles, increasing the growth rate of extracellular matrix in porous scaffolds [23] as compared to more rounded, spheroidal pores. Higher surface micro-roughness (below a few μm), commonly achieved via sandblasting [30] or chemical milling [31], is associated with improved cell adhesion in intro, in vivo, and clinically [32,33].

Utilizing a space holder in combination with DIW, as demonstrated in Fe or Ni with leachable CuSO₄ [34], helps create hierarchically-porous structures; large channels are thus formed between the struts and micropores within struts. The choice of NaCl in this work is informed by studies utilizing hot-isostatic pressing of Ti or Ti alloys with NaCl [35-37], revealing i) no temperature limit for densification, ii) low cost, iii) high-curvature pore morphology, iv) insignificant metal corrosion during space holder removal, and v) no toxicity from any space holder residues.

Clinicians have utilized osseoinductive growth factors such as recombinant bone morphogenic protein-2 (BMP-2) to achieve bony fusion and enhanced clinical outcomes in spinal operations [38], but the high doses and quick release rates in current clinical strategies, such as saturating an absorbable collagen sponge with rhBMP-2 [39] can lead to serious surgical complications [40,41]. Alternative carrier materials that reduce the required BMP-2 dose by providing a sustained, local release of the protein could enable a more robust and safer bone regenerative process. One such alternative carrier material is a biopolymer-supramolecular polymer-collagen slurry (SPS), composed of bioactive peptide amphiphile (PA) supramolecular nanofibers and porous collagen particles, which we showed significantly reduced the release rate compared to a collagen sponge in our previous study [42]. Given the well-known biocompatibility of Ti scaffolds [3] and our previous work showing excellent cell viability of the SPS in a clinically relevant large animal model [42], we decided to test the implants in an in vivo model of spinal fusion.

The current study assesses SPS-backfilled Ti scaffolds for bone formation, bone ingrowth, and bone anchoring to microporous Ti. We utilize 3D extrusion of inks containing a suspension of Ti and NaCl powders, followed by Ti powder sintering and thermal removal of the NaCl space holder to fabricate microlattices with two levels of porosity: macrochannels (~1 mm wide and spanning the length of the scaffolds) between the struts for nutrient transport and vascularization, and micropores (~100 μm wide) within the struts for direct colonization of osteoblasts and cells. Compression testing verifies a low stiffness and high ductility of the high-porosity scaffolds. Four scaffold conditions were created for deployment in the 8-week posterolateral lumbar fusion model used to assess total bone volume (TBV) and total bone ingrowth (TBI) in rats via synchrotron μCT: dense-strut and porous-strut Ti scaffolds, each backfilled with either subtherapeutic BMP-2 alone or the BMP-2-loaded SPS.

2. Methods

2.1. Ink Preparation and Extrusion

The inks (reported below in g/cm³ of ink) were prepared according to methods developed by Jakus et al. [43]. Starting materials were commercially-pure Ti powder (CP-Ti, grade 2, <25 μm, from AP&C, Canada, 3.18 g), sodium chloride powder (99.0% min, ACS, 1.36 g), polystyrene binder (PS, Sigma-Aldrich, average M_w =350,000, 0.22 g), dibutyl phthalate plasticizer (DBP, Sigma-Aldrich, 0.1 g), ethylene glycol butyl ether surfactant (EGBE, Sigma-Aldrich, 0.4 g), and dichloromethane solvent (DCM, Sigma-Aldrich, 9 mL). The NaCl powder was high-energy ball milled (Spex 8000M) for 30 minutes and sieved to a 90-106 μm particle size distribution. The resulting volume ratios were 15 vol% PS: 85 vol% powder, with the powder comprising 70 vol% CP Ti and 30 vol% NaCl. Control specimens, containing fully dense struts, were printed from a NaCl-free ink. The low-viscosity ink is sonicated then thickened via solvent evaporation in a heated water bath at 55°C until a printable, shear-thinning state is achieved, with a viscosity of ~ 40 Pa s, as described in more detail in ref. [34].

The 3D-extrusion was performed on an Envisiontec 3D-Bioplotter using straight, stainless-steel nozzles (580 μm inner diameter, Nordson EFD) at pressures of 4-5 bar, speeds of 6-10mm/s, with layer height (11 layers) equal to 70% of the inner nozzle diameter (~410 μm), 2.25 mm strut spacing, and a 0/90° strut layer arrangement, leading to a nominal channel porosity of 61% for a 25 x 25 x 5 mm³ scaffold. Evaporation of DCM occurs immediately upon extrusion, creating a load bearing, dry shell around each filament. Scaffolds were dried via DCM evaporation over ~ 12 h in air. A razor blade oriented perpendicular to the surface of the scaffold was used to create two types of green specimens: (i) rectangular scaffolds (3.5x15x5 mm³) with 2 full-length longitudinal struts per implant, 6 macrochannels and 11 vertical layers, as shown in Fig. S1a and (ii) structural lattices (12x12x25 mm³) surrounding the lattice, produced for compressive testing, as shown schematically in Fig. 1b. Energy dispersive spectroscopy in Fig. S2 of a Ti + NaCl strut illustrates the surface morphology and relative size of Ti and NaCl powders.

Fig. 1:

Summary of processing steps used to produce hierarchically porous Ti via direction-ink write extrusion (DIW) with NaCl space holders. (a) Ink was prepared by combining the following in a dichloromethane (DCM) solvent: atomized commercially-pure Ti powder, NaCl particles, polystyrene (PS) binder, ethyl butyl glycol ether (EGBE) surfactant, and dibutyl phthalate (DBP) plasticizer. (b) The ink was 3D-extruded into green-body scaffolds; DCM was evaporated, and the scaffolds were cut into 3x20x5mm³ green-body implants. (c) The green bodies were debound and densified under Ar, during which NaCl was evaporated and Ti powders were sintered. (d) The sintered specimens were chemically milled to open fenestrations at strut surfaces. (e) The specimens were filled with rhBMP-2 or rhBMP-2-SPS (green). (f) The filled specimens were implanted into the spine of rats and harvested at 8 weeks post-operative for synchrotron μCT assessment.

2.2. Debinding, Sintering, and Chemical Milling

Green-body specimens were placed into an alumina crucible coated with boron nitride (ZYP Coatings, USA) in preparation for sintering. Grade-1 Ti foil (McMaster-Carr, USA) served as the oxygen getter and was folded over the alumina crucible, with a small triangular opening at the top to allow gas ingress and egress. With 99.999% pure Ar flowing at 28 l/h, the specimens were subjected to three isothermal heat treatments: (i) solvent evaporation (150 °C / 30 min), (ii) binder removal (450 °C / 60 min), as verified via thermogravimetric analysis [14], and (iii) Ti sintering (1200°C/ 240 min). Heating and cooling rates were 10 and 5 °C / min, respectively. NaCl evaporated during heating above its melting point of 801°C. A sintered scaffold is shown in Fig. S1b. The scaffolds were ground flat on their sides for smooth implantation, achieving a final dimension of 11 x 2.5 x 4mm³, with six macrochannels, each with a 1.2 x 1.2 mm² profile.

The specimens were then chemically milled for 1 minute in a solution composed of 4% (v/v) stock hydrofluoric acid (Sigma-Aldrich), 23% (v/v) stock nitric acid (Sigma-Aldrich), and 73% (v/v) deionized water to uniformly remove a thin Ti layer and open microporosity at the strut surface. The implants were rinsed thoroughly, ultrasonicated in water, then rinsed in DI water.

2.3. Microstructure and mechanical characterization

Some of the sintered 3DP scaffolds were mounted, ground, and polished (Epothin2, Buehler, SiC P600 – P2500, diamond 3–1 μm, lapping with SiO₂). Optical (Eclipse MA200) and scanning electron microscopy (JEOL JSM-7900FLV, 15keV) were used to image these cross-sections and surfaces. Porosity was measured using ImageJ on optical cross-sections with auto thresholding.

The top and bottom surfaces of sintered compression lattices were ground flat and parallel using concentric grinding platens, achieving specimens with ~1:1.5 aspect ratio. Lattices were placed into a piston-in-sleeve steel compression cage and deformed under uniaxial compression at constant strain rate (20%/min) using the MTS Criterion C45.105. Crosshead displacement after machine compliance correction was used to determine strain, using the direct technique developed by Kalidindi et al. [44]. The yield strength was calculated as the 0.2% offset yield stress using the elastic slope.

2.4. Backfilling 3DP Ti scaffolds with Bioactive Agents

SPS was prepared as described previously [42]. Briefly, PA molecules were synthesized using standard solid-phase peptide synthesis and purified by reverse-phase high-performance liquid chromatography. PAs were synthesized with the following sequences: C₁₂-(K)VVVAAAEEESGGGYPVHPST-NH₂ (BMP-2-binding PA) and C₁₆-VVVAAAEEE-NH₂ (diluent PA). All PAs were at least 95% pure by liquid chromatography-mass spectrometry. PAs were reconstituted at a total concentration of 10 mg/mL (5 mg/mL BMP-2-binding PA, 5 mg/mL diluent PA) in sterile water and adjusted to pH 7.5 using dropwise addition of 1M NaOH. PA solutions were bath sonicated for 30 minutes, thermally annealed at 80 °C for 30 minutes, and then slow cooled to room temperature overnight. A 1.5 mg/mL stock solution of rhBMP-2 (Medtronic) was diluted into the PA solution to achieve the desired BMP-2 concentration. Collagen particles for the SPS were prepared by homogenizing collagen sponge sheets (Integra) in a blender (Magic Bullet, Homeland Housewares), flash freezing in liquid nitrogen, and lyophilizing to obtain dried collagen particles. The BMP-2-loaded SPS was then prepared by mixing 5 wt% collagen particles with 95 wt% PA/BMP-2 solution and stirring with a spatula.

Each Ti scaffold was loaded with 50 ng rhBMP-2 (100 ng rhBMP-2/rat). This dose demonstrated robust fusion in the rat PLF model when delivered using the SPS, but 0% fusion when delivered using commercially available collagen scaffolds [42]. Four experimental groups were evaluated: 1) dense-strut scaffolds loaded with BMP-2-solution, 2) dense-strut scaffolds backfilled with BMP-2-loaded SPS, 3) porous-strut scaffolds loaded with BMP-2-solution, and 4) porous-strut scaffolds backfilled with BMP-2-loaded SPS. Based on gravimetric analysis, the macrochannel volume was determined to be ~90 μL per scaffold. Therefore, for groups with BMP-2-solution as the bioactive agent, 90 μL of BMP-2-solution in milli-Q water (50 ng/90 μL) was pipetted onto the scaffold following placement in the intertransverse space. For backfilling Ti scaffolds with SPS, sterile scaffolds were placed in 1 cc syringes, and 300 μL of BMP-2-loaded-SPS (50 ng BMP-2 per 90 μL of SPS) was packed on top of the scaffold. The syringe plunger was then inserted and used to mechanically push the slurry through the macrochannels of the Ti scaffold, as shown in Fig. S1c and Fig. S5. After backfilling, the tip of the syringe barrel was cut off which allowed SPS-filled scaffold to be retrieved. Any excess SPS on the surface of the scaffold was removed using a spatula before implantation so that the volume and dose of the bioactive agent was comparable between the BMP-2-solution and SPS groups.

2.5. Surgical Procedure

This study was approved by the Northwestern University IACUC with Animal Welfare Assurance from the Office of Laboratory Animal Welfare. Twenty-four 12- to 16-week-old female and male Sprague-Dawley rats underwent fusion at L4–L5 with implantation, following a well-established surgical procedure [45].

Under isoflurane anesthesia, a posterior midline incision was made over the lumbar spine, with ensuing bilateral fascial incisions around the spinous processes. The L4 and L5 transverse processes (TPs) were exposed by blunt dissection and decorticated with a high-speed burr. Implants were then placed bilaterally to bridge the TPs. The fascia and skin were closed with absorbable sutures and wound clips, respectively. Analgesics (0.3 mg/kg buprenorphine SR and 5 mg/kg meloxicam) were administered for three days postoperatively. After euthanasia at eight weeks post-operative (via bilateral thoracotomy under anesthesia), spines were harvested and fixed in formalin [45]. Fig. 2 shows the anterior, posterior, and lateral orientational schematics with BI direction (SOLIDWORKS software, Dassault Systèmes, Vélizy-Villacoublay, France) of a human male cadaver (Negar An, GrabCAD, Eden Prairie, MN). Fig. S1d depicts an explant that is ready to undergo μCT.

Fig. 2:

Schematic of the various 3D views of the scaffolds after implantation into Sprague-Dawley rats, where the transverse processes (TPs), numbered macrochannels (MCs), and bone ingrowth (BI) in various anatomical directions are labeled. BI is divided into three directions: posteroanterior (PA, red), mediolateral oblique (MLO, green), and craniocaudal (CC, blue).

2.6. Synchrotron μCT

Synchrotron μCT was performed at Argonne National Laboratory, Advanced Photon Source, beamline 5BM-C (DND-CAT), with 3 scans stitched together to cover the entire scaffold and the adjacent volumes of the transverse processes. A total of 1440 projections were collected over 180° (0.125° increment) using 25keV x-rays, and 900ms exposure per projection provided usable x-ray transmissivity through the scaffold with enough contrast to distinguish Ti and bone as well as include the transverse processes in the projection. Reconstruction utilized a Shepp-Logan filtered back-projection algorithm on a six-node Linux cluster at the beamline; these reconstructions consisted of isotropic volume elements (voxels) 12 μm in size. A preliminary synchrotron μCT experiment evaluated one laterally bisected scaffold from each group under the same conditions as above with isotropic volume elements 6 μm in size; the reconstructed field of view, however, could not cover the TPs and scaffold (Figs. S6-S9). Limitations in synchrotron μCT time restricted imaging to 12 scaffolds (N=3/group) to quantify bone growth. A total of 48 scans (each ~2hrs) were performed.

2.7. Bone Growth Assessment

The scaffolds were oriented with the macrochannels approximately perpendicular to the cranial-caudal axis (and the tomography rotation axis). Fig. 3. quantification of bone growth into scaffolds employed two metrics: total bone volume within each scaffold macrochannel (TBV/MC) and the maximum distance that bone grew into each macrochannel. Analysis was performed on each individual macrochannel, and each metric combined the results from the six macrochannels.

Fig. 3:

Summary of tomographic measurement analysis. (a) Radiograph at 8-weeks postoperative informs scaffold orientation relative to transverse processes (TPs). (b) Schematic demonstration of the measurement process with a hatched bone area and bone ingrowth (BI) distance measurements labeled; BI in all three directions occurs simultaneously, but only the mediolateral oblique (MLO) and craniocaudal (CC) directions are measured. BIMLO is measured relative to the scaffolds’ vertical struts, labeled MLO Struts 1-6. Total Bone Volume (TBV) is calculated from the measured area and voxel thickness. (c) A typical μCT rendering with 4 color-coded, overlayed planes in separate macrochannels (MCs) illustrates the integration of bone areas to produce the TBV for a representative scaffold. (d-g) Tomographs of the planes indicated in (c) with overlayed area in orange and TPs labeled.

Bone was visible within the macrochannels but the long paths through Ti produced strong and highly variable absorption contrast artifacts. As a result, simple binary segmentation could not accurately separate bone and soft tissue/void; instead, the envelope of bone was outlined (hatched area in Fig. 3b) within 10 of the approximately 100 numerical sections covering each macrochannel. The area within each envelope was calculated and converted to volume by multiplying by the section thickness, i.e., the voxel size. After the 10 sections were measured, the volumes between measured sections were estimated by interpolation. The measured volumes for all of the sections were summed to obtain TBV/MC. Finally, TBV/MC were summed for the six macrochannels to obtain the total bone volume within the scaffold, TBV.

The second metric, BI_MLO, defined as the maximum distance that bone grew into each macrochannel, was scored based on how many of the six mediolateral oblique (MLO) struts the bone had grown past at its maximum extent. If no bone was observed in a macrochannel, the score was zero; if bone had grown as far as strut five but not as far as strut six, that macrochannel would be scored as 5, e.g. the schematic of Fig. 3b. The scores for the six macrochannels were summed (maximum of 36) to give a numerical value for how extensively the bone was integrated into that scaffold, defined as total bone ingrowth (TBI). Note that this approach is a modification of methods of Driscoll et al. [25].

Fig. 3c illustrates tomogram bone area measurements overlayed along the craniocaudal (CC) direction of a μCT rendering of a porous-strut scaffold with BMP-2-SPS (Dragonfly^® 2022.2 software, Object Research Systems, Montréal, Québec). Figs. 3d-g illustrate the associated tomograms with outlined area measurements.

The strength of the anchoring of the scaffold to the TP is given by BIcc values for macrochannels 1 and 2 for the first TP and 5 and 6 for the second TP, where the bone initially enters the scaffold via contact with the TP. TP bridging is the BIcc across macrochannels 3 and 4, forming a continuous section of bone spanning the middle of the scaffold. Figs. S10 illustrates an alternative orientation of the scaffold. S10f - S10g illustrates null anchoring and bridging. Typical clinical fusion metrics (manual palpation) are uninformative because the stiffness of the Ti scaffold dominates the palpation response and therefore are not used in this study.

2.8. Statistical Analysis

Statistics were performed using GraphPad Prism v.10 software. Microporosity and BMP-2-SPS were treated as continuous, independent, and potentially interacting variables. Two-way analysis of variance (ANOVA) of normally distributed data was used to assess Tukey’s test for post hoc analyses. Fisher’s exact test was used to compare the categorical data. A p-value <0.05 was considered statistically significant.

3. Results

3.1. Debinding and Sintering

The as-received atomized CP Ti powder displayed an oxygen concentration of 0.26 wt% and carbon concentration of 0.03 wt%, as verified by combustion analysis (performed by Westmoreland, Youngstown, PA). This was consistent with its grade 2 label. Debound and sintered scaffolds showed a bulk oxygen and carbon concentration of 0.47 and 0.13 wt%, respectively, placing the resulting Ti just outside the range of grade 4 (ASTM 265). Scaffolds experienced an average shrinkage rate of 9.0 ± 2.9% for the dense-strut and 10.6 ± 2.4% for the porous-strut groups (ImageJ).

3.2. Metallography

Fig. 4a illustrates a cross-section of the control scaffold Ti sintered without NaCl space holders, which demonstrate high levels (> 97%) of strut densification with small residual pores that are <15 μm in size. Grains range in size from 50-150 μm.

Fig. 4:

Optical micrograph of longitudinal cross-sections of pre-chemically milled struts, with Ti grains made visible via polarized light for: (a) dense-strut scaffold printed with NaCl-free ink, showing very few residual sintering micropores and (b) porous-strut scaffold with space holder pores, a surface fenestration (SF), and inter-pore fenestrations (IPF). SEM micrographs of the surface of porous-strut scaffolds (c) before chemical milling, showing struts with only occasional surface fenestrations (SF) and surface pores. (d) Cross-section of a porous strut, subjected to chemical milling after sectioning, showing more pores and inter-pore fenestrations (IPF), with an orange outline marking the original strut diameter. Small pits are visible on the cross-section and the horizontal struts above and below the cross-sectioned strut, which were formed during chemical milling.

In scaffolds printed with NaCl-containing inks, as shown in Fig. 4b, strut microporosity was 30-35% (ImageJ), and Ti grains were comparable in size to space holder-free struts. With a macroporosity of 55-60% (ImageJ), the microporous scaffolds achieve a total porosity of 68-73% (ϱ= 0.27-0.32), excluding vertical strut spacing. Pore size was consistent with initial NaCl size, taking into account that strut sintering, after NaCl removal, leads to partial shrinkage and spheroidization. Pores were connected to one another via inter-pore fenestrations and to the scaffold surface via surface fenestrations, as labeled in the pre-chemically-milled micrograph (Fig. 4b). Before chemical milling, few surface fenestrations were observed.

3.3. Chemical milling

Fig. 4c demonstrates that the sintered microporous scaffolds have significantly fewer surface pores and surface fenestrations than expected from the high-volume fraction (30 vol%) of space holders. Fig. 4d depicts the strut surfaces of a Ti lattice after chemical milling, showing both types of fenestrations and marked increases to the surface roughness of the struts through the formation of small pits. An average of 8.9 ± 1.7% and 10.7 ± 1.7% mass was removed during the chemical milling for dense-strut and porous-strut scaffolds, respectively. Detailed reconstructions of the dense-strut and porous-strut scaffolds are shown in Figs. S6-S9.

3.4. Compressive Properties

Fig. 5a shows characteristic compressive stress-strain curves for non-contoured (open-cell), chemically-milled dense-strut and porous-strut lattices. The elastic regime of the stress-strain curve provides a stiffness of 0.4-5 GPa for the dense-strut lattices and 0.9-2.2 Gpa for the porous-strut lattices, with relative densities ranging from 0.11 to 0.29, calculated from the mass and overall volume measurements. A 0.2% offset of the elastic region demonstrates a yield ranging from 12 to 50Mpa for the dense-strut lattices and from 7 to 31Mpa for the porous-strut lattices. Figure S4 provides the full stress-strain curves for all lattices tested.

Fig. 5:

(a) Compressive stress-strain curves for etched, dense-strut ($\rho =0.27$) and porous-strut lattices ($\rho =0.20$). (b) Relative elastic modulus and (c) relative yield stress vs. relative density with present data (red and black circles) demonstrating a lower relative stiffness and higher relative yield stress than predictions for open cell foams (Gibson-Ashby equation [46]). Included for comparison are published data for various porous Ti and Ti64 structures created via direct ink writing DIW (Chen et al. 2019 [47], Coffigniez et al. 2021 [16]), selective laser melting SLM (Ataee et al. 2018 [48]) and powder pressing (Niu et al. 2009 [49], Jha et al. 2013 [50], Ye and Dunand 2010 [36]). The table (d) lists the densities and mechanical properties for the dense and porous-strut specimens in this study.

Fig. 5b and 5c illustrate the relative yield stress and relative elastic modulus, respectively, to density for the eight compressive lattices in this study. These mechanical properties are compared to the Gibson-Ashby model for the elastic modulus and plastic yield stress of an open-cell foam (equation in insert in Fig 5b and 5c) and literature data for porous CP Ti and Ti64 made via various processes including DIW, SLM, and powder pressing.

3.5. Bone Ingrowth and Bone Area / Volume

The different quantitative measures of bone growth are shown in Fig. 6, with each column representing the treatment listed above the column. Note that the results for macrochannels 1–6 are plotted separately, with macrochannel 1 adjacent to the TP of L4 and 6 adjacent to the TP of L5. The scaffolds treated with BMP-2-solution served as controls. One data point represents each of the ten sampling planes plotted in Fig. 6a-d. For example, in macrochannel 1 of scaffold 2 of the dense-strut, BMP-2-solution group, the maximum area for one of the ten sampling planes approached 4 mm², and scaffolds 1 and 3 of that group had no bone growing into macrochannel 4.

Fig. 6:

Plots for quantitative bone growth metrics vs. macrochannels. Schematics for each of the 3 metrics are illustrated at the top; each of the schematic illustrates a single macrochannel. In-growing bone is colored yellow; the envelope of bone is outlined in dark orange in the left schematic; the middle schematic shows the bone envelope integrated across the macrochannel distance to produce a volume. The dark orange outline in the right schematic shows bone ingrowth corresponding to BI=5. The three scaffold replicates in each group are differentiated by color (red, blue, orange). (top-row: a, b, c, d) Line plots describe the measured bone area (BA) inside the scaffold, where growth direction from the transverse processes, indicated by BI_CC, is towards the middle of the plot. (middle-row: e, f, g, h) Point plots describe the bone volume (BV) inside each macrochannel, determined from the previous line plot. (bottom-row: I, j, k, l) Point plots describe the extent of bone ingrowth (BI_MLO) for each scaffold. The average values in the point plots, for each macrochannel, are indicated by black horizontal lines.

Curves of Fig. 6a-d are integrated to compute the volume of the bone envelope in each macrochannel; note that the volume of the bone envelope contains significant soft tissue because the envelope consists of a thin cortical shell and internal trabeculae (Fig. S11). Fig. 7i-7l show the maximum distance that bone grows into the scaffold lattice.

Fig. 7:

Comparison of treatment groups with two-way analysis of variance (ANOVA) for (a) total bone volume (TBV) and the (b) total bone ingrowth score (TBI). The average of each group is indicated with a black horizontal line. The addition of SPS demonstrates a trend (p = 0.161) towards enhanced TBV, in plot a, and a significant (p<0.001) increase in TBI, in plot b, relative to the BMP-2-Solution control.

In the dense-strut, BMP-2-solution group, bone did not bridge the TPs for scaffolds 1 and 3, but the ends of these two scaffolds were both anchored to the TPs. Scaffold 2 was the only specimen in the solution-based backfilling groups to achieve full TP bridging. None of the scaffolds in the porous-strut, BMP-2-solution group exhibited TP bridging, and all demonstrated poor anchoring to the TPs (Fig. 6c, g,k). All scaffolds in the slurry-based groups achieved TP bridging with bone growth into every macrochannel (Fig. 6b, d), substantial bone envelope volumes in every macrochannel (Fig. 6f, h) and high BI for each macrochannel (Fig. 6j, l). The positions of scaffolds relative to the TPs varied between groups but this did not dominate the bone growth measurements.

The stacks of slices were inspected to see whether the mineralized bone within the macrochannels contacted the Ti metal. Bone-Ti contact occurred in some portions of some macrochannels, but the extent of contact was less than one might expect given the considerable bone ingrowth observed. Examples of apparent bone-Ti contact (i.e., bone and Ti filling adjacent voxels) are given in Fig. S12.

Fig. 7 condenses the results of Fig. 6e-6h and 6i-6l into two panels for comparison of the total bone volume (TBV) and total bone ingrowth (TBI). The average TBVs, proceeding from left to right, are illustrated by the black horizontal line in Fig. 7a: 10mm³, 23mm³, 4mm³, and 14mm^3,, respectively. The TBIs for the four groups, proceeding left to right, are illustrated by the black horizontal line in Fig. 7b: 14, 31, 13, and 30, respectively.

The specimens backfilled with BMP-2-SPS demonstrate significantly (p<0.001) increased TBI and moderately increased (p = 0.161) TBV relative to their BMP-2-solution control. The groups utilizing porous-strut scaffolds demonstrate no significant change in TBV and TBI compared to the dense-strut control.

4. Discussion

4.1. Printing, Debinding and Sintering

The extrusion characteristics of the Ti + NaCl ink follows trends observed by Kenel et al. [34] but differs in the space holder particle segregation during extrusion. As the ink moves through the extrusion nozzle, some segregation of the NaCl particles occurs towards the center of the nozzle due to their larger size (90-106 μm across) compared to the atomized CP Ti (<25 μm). The smaller Ti powders fill the interstices between the NaCl particles and the nozzle wall. Liu et al. observed a similar flow-induced particle migration of coarse aggregates toward the middle of concrete extrusion vessels, which is consistent with our observation, though at a larger scale [51].

The crosshead speed differs between the Ti-only and Ti + NaCl groups with 10mm/s and 6mm/s, respectively, caused by a slight difference in viscosity between the two inks. We optimized our extrusion to produce scaffolds with the highest reproducibility and minimum defects to reduce extraneous effects during bone growth and measurement process; this difference in extrusion speed caused an expected reduction in the average strut diameter between the microporous-strut and dense-strut groups, reducing the stiffness of the dense-strut lattices to a value similar to that of the porous-strut lattices. While coincidental, these similar stiffnesses between the dense and porous-strut groups allows us to probe the functionality of the microporosity more directly.

Debinding and sintering under Ar yields a useful combination of strength and ductility for the DIW Ti scaffolds. The modest uptake of 0.1wt% C and 0.21wt% O during ink fabrication, debinding, and sintering are challenging to achieve for DIW Ti microlattices. Carbon contamination is doubled utilizing this Ar-only method, as compared to the 0.054wt% C and 0.21wt% O reported in Song, et al. [14], who utilized 100% H2.

The use of low-oxygen, atomized Ti rather than milled hydride dehydride Ti significantly decreases O contamination; this enables maintenance of ductility after densification, which is an essential mechanical characteristic for orthopedic use. However, utilizing spherical Ti limits processing to inert gas (Ar) sintering methods. Extensive testing utilizing 100% and 4% H2, as performed by Song et al.[14] to reduce the formation of TiC during thermochemical treatment, results in the collapse of our lattices during and shortly after the debinding process. After the debinding isothermal at 450 °C but before sintering occurs at 700 °C [52], a particle-jamming phenomenon occurs during which there is no polymeric binder to hold particles and metallic powders together; they instead interlock strongly enough to resist gravity, but otherwise collapse with movement. If an H2 environment is used in place of Ar to reduce the carbon contamination from the polystyrene binder, the volumetric expansion from the hydrogenation of the Ti will cause our spherical-Ti green bodies to collapse. However, a hydride-dehydride Ti green body (as used by Song et al [14]) remains intact, likely due to enhanced particle jamming. This phenomenon of static mechanical equilibrium for dry, granular materials interacting via contact points has been investigated by Behringer and Chakraborty in detail [53].

4.2. Chemical Milling

Our use of chemical milling removes enough material from the surface of struts to open pores located just beneath the surface, while also increasing the size of existing surface fenestrations and inter-pore fenestrations, as depicted in Fig. 5b. This chemical milling exchanges mechanical performance for biological performance. By exposing and enlarging the pores, the fatigue resistance, yield stress, and stiffness were reduced due to the reduced load bearing area and increased number of stress concentration points. However, additional cell anchoring locations were formed and surface roughness was also increased, which should promote adhesion of osteoblasts [31].

4.3. Compressive Properties

The mechanical properties of scaffolds are strongly dependent on their relative density, as expected from the scaling laws for cellular solids [46]. Based on printing design, sintering treatment, space holder volume fraction, and chemical milling extent, the resulting porosity controls the mechanical properties over a wide range of values.

The stress-strain plots in Fig.6a display three regimes typical of metallic cellular foams when compressed: (i) a small elastic segment at low strains (below 1% in our study); (ii) an extended flat region indicating the occurrence of plastic deformation within the material (from 1 to ~20%); and (iii) a significant rise in stress as the material undergoes densification (above ~20%), often with significant localized failure of struts.

Deformation mechanisms varied among samples even with similar relative densities. Fig. S4 illustrates that the onset of plastic buckling, indicated by the local maximum in the stress-strain plot, occurred at a strain between 10-20% for lattices P1, P3, P4 and D1. By contrast, lattices P2, D3, and D4 deformed without buckling to 37, 38, and 36%, respectively, allowing each layer of struts to provide maximal compressive support, leading to a significantly higher ultimate compressive strength compared to the other lattices. Samples P1, P3, P4, and D1 experienced their second buckling events at strains of 30, 34, 31 and 38%, respectively. Lattice D2 experienced several small fractures visible as serrations throughout the stress-strain curve with no clear local maximums, indicating that fracture dominates over plastic deformation [54].

The Gibson-Ashby model for the Young’s modulus $E^{\ast }$, of an open-cell foam [46], is:

$$\frac{E^{\ast }}{E_s}=C_1{\left(\frac{{\rho }^{\ast }}{{\rho }_s}\right)}^2$$ (1)

where $E_s$ is the Young’s modulus of the bulk material (105 GPa for CP Ti and 110 GPa for Ti64)[55], ${\rho }^{\ast }$is the density of the sample, ${\rho }_s$ is the density of the bulk material (4.50 g/cc for CP Ti and 4.43 g/cc for Ti64)[55], and $\mathrm{C}1$ is a geometric constant equal to unity. The experimentally determined Young’s moduli in Fig. 5b are approximately ½ to 1/3 of the values predicted by equation 1. This deviation from the model is in agreement with literature utilizing both DIW and other manufacturing techniques. Variances in the strut diameters, such as reduced strut thickness for the middle of the lattice relative to the hairpin turn regions that occur during the extrusion process, may contribute to this reduced elastic modulus. Lattice D1 demonstrates an unusually low elastic modulus followed by an early onset of plastic deformation, indicating that either significant non-visible strut defects were present or the lattice became askew during the compression test.

The Gibson-Ashby model for the plastic yield stress, ${\sigma }_{pl}^{\ast }$, of an open-cell foam is as follows [46]:

$$\frac{{\sigma }_{pl}^{\ast }}{{\sigma }_s}=C_2{\left(\frac{{\rho }^{\ast }}{{\rho }_s}\right)}^{3∕2}\left(1+{\left(\frac{{\rho }^{\ast }}{{\rho }_s}\right)}^{1∕2}\right)$$ (2)

where ${\sigma }_s$ is the compressive yield strength of the bulk material (480 MPa for CP Ti grade 4 and 836 MPa for Ti64)[1] and $\mathrm{C}2$ is a constant equal to 0.3 for open-cell foams. The measured yield strengths of the lattices were greater than the predictions from equation 2, perhaps due to the surface oxidation and solid-solution strengthening caused by the chemical milling process or the greater amount of TiC precipitate strengthening formed from residual binder pyrolysis under an Ar environment. Coffigniez et al. and Chen et al. investigated DIW CP Ti [16] and Ti64 [47] lattices, respectively, and likewise report greater compressive yield strength than equation 2.

As reported earlier, similar stiffnesses and yield strengths between the dense and porous-strut groups enable a more direct comparison of the osseointegration-promoting response of microporosity, discussed below.

4.4. Bone Ingrowth and Bone Area / Volume

Assessment of the promotion of bone growth for Ti substrates via μCT is a fundamentally challenging measurement due to the high x-ray attenuation for Ti relative to bone [56-58]. The use of additively manufactured Ti scaffolds designed for this quantitative measurement, where the bone is enclosed by the Ti scaffold, makes this an even more challenging study. Histology is not utilized in this study due to the very limited volume sampling. Additionally, sufficient resolution is provided by μCT; all mineralized tissue with dimensions larger than one-half the volume of one voxel is detected [59]. The design of our scaffolds enables implantation spanning adjacent level transverse processes in the posterolateral spine fusion model and quantification of BI, as defined in section 2.7.

Significant artifacts exist in the tomographic reconstructions due to the presence of Ti; these artifacts cannot be removed with automated programs and thus require manual measurements. Artifacts included rings and various expressions of photon starvation, which prevented thresholding inside the bone architecture, such as trabecular morphometric measurements. Thus, traditional metrics such as bone volume / total volume (BV / TV) or mean trabecular thickness (Tb.Th) were impractical, and only TBV and TBI were quantified. Maximizing both TBV and TBI is desired for strong mechanical interlocking of the bone with the scaffold architecture, thereby enhancing mechanical stability for spinal fusion. Greater BI represents a greater bone formation response and stronger mechanical connection between the native bone and scaffold. Larger values of TBV represent larger cross-sectional areas of interlocking bone and increased robustness of the mechanical connection. One might expect a positive correlation between the TBV and BI, as a longer bony ingrowth into the lattice would likely have a greater volume. Fig. 7 summarizes the effects of microporosity and SPS in this study; we found that the bone formation promoting effects of the BMP-2-SPS were superior to BMP-2 loaded into the lattice as a solution, while the microporosity of the Ti scaffold had a nonsignificant effect.

The bioactive component of BMP-2-SPS lies in the PA molecules, which self-assemble to form nanofibers that can be customized for specific applications in regenerative medicine by displaying bioactive signals on their surfaces [60]. The capacity of the bioactive PA system to promote BMP-2-induced bone regeneration and spine fusion in composite with a DIW Ti scaffold was investigated in the rat posterolateral lumbar intertransverse spinal fusion model. We previously demonstrated that the PA system biodegrades over several weeks in vivo, leading to enhanced retention of the growth factor and consequently, enhanced bone growth [61,62]. This study did not independently evaluate BMP-2 delivered with and without collagen particles; thus, collagen may have contributed to bone formation.

Deployment of the SPS to backfill the Ti scaffold macrochannels likely promotes osseogenesis via at least two mechanisms. As described in McClendon, et al. [42], SPS potentiates BMP-2 signaling by binding and prolonging the bioactive half-life of the BMP-2 protein and enabling nucleation of apatite crystals on the nanofiber surface. Collagen microparticles alone, without peptide amphiphile nanofibers, do not induce osseogenesis. The nanofibers both extend the exogenous BMP-2 half-life, and recruit endogenous BMP-2, and extend its half-life, potentiating signaling. In contrast, the scaffolds loaded only with a solution of BMP-2 prior to implantation lack the ability to retain the growth factor for an extended period of time, likely resulting in an initial burst release of the protein [40]. The choice of 50 ng rhBMP-2 per scaffold provided the threshold BMP-2 to observe improved bioactivity enabled by the PA [42].

We hypothesize that incorporation of micropores into the Ti scaffold struts did not improve bone growth into the BMP-2-solution treatment groups due to a lack of improved retention of BMP-2. Without BMP-2 retention, ingrowth via progenitor cells would only be due to non-chemotactic mechanisms, resulting in fibrous tissue instead of osteoblast precursors. Additionally, relative to the SPS group, cells in the solution group may have experienced less steric hindrance due to the presence of the peptide amphiphiles and collagen particles, which could impede direct and immediate access to the Ti surface. Fig. 3d shows an example of the lack of interaction between the newly formed bone and the Ti. The ingrown bone finger makes very little direct contact with the Ti as it extends into the scaffold macrochannel; thus, introducing micropores appears to do little to impact ongoing bone formation and ingrowth, as osteoblasts may not adhere to the Ti directly. The microenvironment created by the microporosity may have contributed to improved bone growth by locally enhancing vascularization and cellular infiltration, though no statistical significance is demonstrated in this study.

Micropores likely have a greater impact on the efficiency of the cell adhesion rate when the cells come into direct contact with the surface of a porous (vs dense) Ti scaffold in a relatively short period of time. Since presumably cells first encountered the SPS material, it is much more likely that local progenitor cells would adhere strongly to that surface rather than to the Ti, especially since the solution group did not contain peptide amphiphiles or collagen. Another consideration is the possibility that the loading and backfilling process may not have generated sufficient pressure to fully infiltrate the BMP-2-solution and SPS, respectively, into the scaffold micropores, leading to some degree of air-filed micropores.

TP bridging and TP anchoring are bone fusion metrics (defined in section 2.7) that characterize the extent of BI. The migration and proliferation of osteoblasts and mesenchymal cells near and on the Ti can lead to a focal adhesion of cells if protein adsorption is sufficient. This can enable nutrient/waste pathways that further bolster bone growth into the scaffold [63], leading to TP anchoring. The extent of TP anchoring is likely correlated with the position of the scaffold relative to the TP, as shorter distances enable shorter times for cells to mobilize to the surface.

Imperfect orientation of scaffolds with respect to the vertebral column may have occurred due to a variety of factors: animal movement post-operative, size of the scaffold relative of the size of the rat, precise placement over the transverse processes, and lack of hardware in the rat-based PLF model. In cases where misorientation and/or suboptimal scaffold placement occurred, shown in more detail in Fig. S3, lack of intraoperative hardware placement, variance in ambulation in the immediate post-operative period and suboptimal tightness of the gutter closures may have been contributing factors. Scaffold fixation is not employed in this rat model because the decorticated transverse processes are thin and hardware fixation would fracture the processes. Suturing in place would add considerably more disruption to the surrounding soft tissue. Scaffolds 1 and 2 in the porous-strut + BMP-2-solution group are examples in which the TP anchoring onto one of the transverse processes was absent, potentially due to misorientation of the scaffolds. TP anchoring potential for these scaffolds would therefore have been reduced. A time-resolved cell infiltration study from the edge of the scaffold to the center of the scaffold may provide insight on the effect of TP anchoring for TP bridging and clarify the optimal placement and dimensions of the scaffolds in the fusion model, but this is beyond the scope of this study.

A range of BI directions were observed in this study, with a majority occurring in the 1) MLO direction through the scaffold, but also in the 2) posteroanterior oblique (PAO) direction and 3) simultaneously MLO and PAO directions. The measurement procedure of BI and BV remain consistent despite the changes in orientation. The values of TBI and TBV did not vary significantly amongst the different orientations. Scaffolds were chosen randomly for BI and BV measurements, and not on the basis of scaffold alignment.

One limitation of this study is the saturation of the BI scaffold metric. Both SPS-based groups achieved TBI very near the maximum allowed ingrowth values, which saturates and thus limits the differentiating characteristics. When the bone fully penetrates the scaffold, there is no other region for the bone to colonize, and only the enlargement of the bone in girth facilitates increased anchoring. Thus, after full ingrowth has occurred, as in the slurry-based groups, only differences in TBV are additionally observable. A thicker scaffold with more struts, and correspondingly, additional space for bone colonization may enable the assessments of BI and bone girth to be studied with less interdependency.

A longer study could provide greater insight into the bone-porous Ti scaffold interaction, since more time might enable the surface of the bone finger to interact with the strut microporosity. Another useful follow-up study could investigate bone growth change due to the inflammatory and immune modulation caused by the response of the host tissue to the microporous scaffold. The bone finger in Fig. 3g is an example of the potential for additional bone growth into and around the scaffold.

5. Conclusions

We manufactured macroporous (via geometrical design) and hierarchically porous (via ~100μm NaCl space holders) Ti scaffolds and compression lattices with the direct-ink write method of a Ti and Ti + NaCl ink. Chemical milling exposed surface fenestrations in the porous struts. Comparable stiffness of the dense-strut and microporous-strut scaffolds (due to variation in the printhead speed) enabled direct evaluation of the micropore functionality for bone formation performance. To verify the effect of microporosity and BMP-2-SPS on the TBI, dense-strut and porous-strut scaffolds were backfilled with BMP-2-solutions or BMP-2-SP slurries, then implanted in Sprague-Dawley rats for assessment of spine fusion capacity in an established rat model. μCT of the explanted Ti scaffolds demonstrates successful quantitative imaging of bone growth surrounded by a Ti medium, enabled by the macrochannel design of the scaffold. BI and TBV of four treatment groups were measured: (1) dense-strut and (2) porous-strut scaffolds loaded with a subtherapeutic dose of BMP-2 alone, (3) dense-strut and (4) porous-strut scaffolds with BMP-2 co-delivered within the SPS.

We found that low oxygen and carbon contamination is possible through DIW and Arenvironment debinding and sintering, producing high ductility, low stiffness Ti scaffolds with struts that are either dense or microporous (with ~58% and ~73% total porosity, respectively). Chemical milling of the porous-strut Ti opens up surface fenestrations blocked during the DIW process. No significant bone formation enhancing response is found for the scaffolds with ~100 μm micropore functionality. Scaffolds with and without micropore functionality in their struts display significant bone formation increases when backfilled with BMP-2-SPS compared to BMP-2 alone. This work demonstrates the viability and versatility of DIW in orthopedic scaffold applications and the effectiveness of bioactive peptide amphiphile (PA) supramolecular nanofibers and porous collagen particles for enhancing osseointegration in vivo. Future work includes modification of the design of the scaffold to reduce μCT artifacts, utilization of a space holder with bioactive utilization, and scaffold surface modifications to encourage direct progenitor cell colonization.

Statement of Significance:

By 2050, the anticipated number of people aged 60 years and older worldwide is anticipated to double to 2.1 billion. This rapid increase in the geriatric population will require a corresponding increase in orthopedic surgeries and more effective materials for longer indwelling times. Titanium alloys have been the gold standard of bone fusion and fixation, but their use has longstanding limitations in bone-implant stiffness mismatch and insufficient osseointegration. We utilize 3D-printing of titanium with NaCl space holders for large- and small-scale porosity and incorporate bioactive supramolecular polymers into the scaffolds to increase bone growth. This work finds no significant change in bone ingrowth via microporosity but significant increases in bone ingrowth via the bioactive supramolecular polymers in a rat posterolateral fusion model.