Research ArticleBone

Mechanobiologically optimized 3D titanium-mesh scaffolds enhance bone regeneration in critical segmental defects in sheep

See allHide authors and affiliations

Science Translational Medicine  10 Jan 2018:
Vol. 10, Issue 423, eaam8828
DOI: 10.1126/scitranslmed.aam8828

For better bone, use softer scaffolds

Large segmental gaps in bone caused by trauma or disease are typically treated with bone grafts and stiff scaffolds to hold the fractured bone in place, but sometimes these defects fail to heal. To optimize bone regeneration, Pobloth and colleagues modified titanium-mesh scaffold designs to provide specific strains and stresses within the fracture environment. In sheep with critical-sized segmental defects, scaffolds that reduced stress shielding around tibial fractures enhanced bone bridging compared to stiffer scaffolds and shielding plates. Scaffolds can be tuned to evoke specific mechanical and biological responses within bone defects, which could help guide regeneration.

Abstract

Three-dimensional (3D) titanium-mesh scaffolds offer many advantages over autologous bone grafting for the regeneration of challenging large segmental bone defects. Our study supports the hypothesis that endogenous bone defect regeneration can be promoted by mechanobiologically optimized Ti-mesh scaffolds. Using finite element techniques, two mechanically distinct Ti-mesh scaffolds were designed in a honeycomb-like configuration to minimize stress shielding while ensuring resistance against mechanical failure. Scaffold stiffness was altered through small changes in the strut diameter only. Honeycombs were aligned to form three differently oriented channels (axial, perpendicular, and tilted) to guide the bone regeneration process. The soft scaffold (0.84 GPa stiffness) and a 3.5-fold stiffer scaffold (2.88 GPa) were tested in a critical size bone defect model in vivo in sheep. To verify that local scaffold stiffness could enhance healing, defects were stabilized with either a common locking compression plate that allowed dynamic loading of the 4-cm defect or a rigid custom-made plate that mechanically shielded the defect. Lower stress shielding led to earlier defect bridging, increased endochondral bone formation, and advanced bony regeneration of the critical size defect. This study demonstrates that mechanobiological optimization of 3D additive manufactured Ti-mesh scaffolds can enhance bone regeneration in a translational large animal study.

INTRODUCTION

The reconstruction of large segmental bone defects, caused by trauma, infection, or tumor resection, remains a considerable clinical concern, and the treatment options for biomechanically sufficient restoration are limited. Compared to external frame fixation for distraction osteogenesis (1, 2), total endoprosthetic replacement with a metallic implant, or the induction of bone regeneration via a biologic membrane (Masquelet technique) (3), transplanting an autologous vascularized cortical bone graft represents the “gold standard” treatment in load-bearing bones (46). However, graft harvesting requires an additional surgical approach and is frequently associated with donor site morbidity or lack of sufficient graft material (7). Alternative strategies that use scaffold-based bioengineering approaches have been evaluated to solve the clinical challenge of large segmental defect regeneration (810), but a reliable solution has yet to be identified.

One promising approach for addressing these clinical needs is the implementation of three-dimensional (3D) additive manufacturing strategies for the generation of customized titanium implants (Ti-mesh scaffolds). In the present case, additive manufacturing means the use of a 3D computed tomography (CT) reconstruction of a patient-specific bone defect for the design of a corresponding Ti-mesh scaffold, which is produced in a laser sintering process. We established a process for producing patient-specific, 3D additive manufactured Ti-mesh scaffolds and evaluated them in a small number of human patients with critical segmental bone defects. However, whether bone formation within the scaffold, and thereby bridging defect ends, is possible remained unclear. We hypothesized that mechanobiological optimization of Ti-mesh scaffolds could enhance the bone formation in such defect settings. The field of mechanobiology investigates the reaction of biological tissues to distinct mechanical signals.

The 3D scaffolds consist of Ti-strut elements that can be assembled to form distinct mesh configurations and a highly interconnected macroporous network. By changing the configuration of the strut elements, it is possible to tune the local mechanical environment (tissue strains and resulting implant stresses) induced by external loading on the strut surfaces and within the scaffold voids (11). If the known mechanobiological regulation theories of bone regeneration (12, 13) are used, Ti-mesh scaffolds could be predesigned to provide a local mechanobiological profile in vivo that supports bone ingrowth. Whether mechanobiologically optimized scaffolds can enhance bone regeneration in a large animal model remains unknown.

The ranges of mechanical strain beneficial for bone formation have been previously identified in the context of uneventful bone healing conditions (14). However, to transfer these concepts to large bone defects, which have a markedly reduced range of mechanical strains, the underlying bone regeneration processes need to be validated (15). On the basis of the reduced mechanical strains within large bone defects (15), we hypothesized that a design that minimizes the mechanical stiffness of the Ti-mesh scaffold, while ensuring resistance against mechanical failure, would lead to enhanced bone regeneration.

To experimentally validate this approach, we used up front finite element (FE) analysis techniques and a honeycomb-like geometry to design a mechanobiologically optimized scaffold providing reduced stress shielding (“soft”). By altering the strut diameter, a 3.5-fold stiffer scaffold was also created (“stiff”). Both scaffolds had the same honeycomb-like geometry, were identical in macroscale length and diameter, but presented different mechanical stiffness. Large interconnected pores forming open channels in three different orientations facilitated the incorporation of autologous cancellous bone graft (ABG) material and guiding of the bone generation process. We hypothesized that ABG material enclosed in the scaffold pores would provide a beneficial biological environment to initiate the bone healing response within the Ti-mesh scaffold and that endogenous critical size bone defect regeneration can be promoted by optimizing the Ti-mesh scaffold stiffness. Scaffolds were produced via laser sintering manufacturing. A small number of patients (n = 19) received personalized, 3D additive manufactured Ti-mesh scaffolds as a treatment alternative for the reconstruction of critical segmental bone defects. These scaffolds were patient-specific, but not mechanobiologically optimized. To investigate bone regeneration within the mechanobiologically optimized scaffolds, we used a critical size bone defect model in sheep. Bone regeneration was monitored over 24 weeks among four test groups to compare the effects of scaffold stiffness and local limb loading at the defect site.

RESULTS

Critical segmental defect treatment with 3D Ti-mesh scaffolds in patients

Radiographic results of bone formation on the scaffold surface can be found in a case example of a 61-year-old woman with nonunion of the femur 16 months after initial fracture treatment. A thick, mineralized callus formation bridging the defect on the outer scaffold surface was noted 28 months after scaffold implantation surgery (fig. S1; for patient history, see the Supplementary Materials). Although bony integration was detected at the scaffold-bone interface, there was radiographically no evidence of bone growth through the scaffold mesh. In addition, not all patient cases showed consistent bone defect bridging. A 45-year-old male patient received a Ti-mesh scaffold after femur fracture due to trauma (fig. S2, A to D; for patient history, see the Supplementary Materials). Accidentally, he fell on the operated leg resulting in refracture 8 months after surgery (fig. S2D). Histologically (fig. S2, E to I), bony integration at the scaffold-bone interface was detectable (fig. S2E). However, the inner part showed no bone formation across the entire mesh network at 8 months after surgery. Remnants of transplanted Reamer/Irrigator/Aspirator (RIA, DePuy Synthes) material (fig. S2,F to H) and well-vascularized connective tissue were detectable (fig. S2I). However, no new bone leading to defect bridging through the mesh network was identified.

Scaffold design strain optimization

In the present large animal study, we wanted to investigate whether bone bridging through the scaffold is possible, and whether mechanobiological optimization of Ti-mesh scaffolds would enhance the segmental bone defect regeneration process. A honeycomb-like arrangement was selected for the scaffold design because of its favorable elastic properties and high failure resistance under compression and bending loads (11). For the sheep study, the aligned struts of the honeycombs were designed to form open channels oriented in an axial, perpendicular, and tilted orientation relative to the scaffold axis. The chosen strut length of 7 mm ensured an easy incorporation of the cancellous bone graft material. On the basis of the specific honeycomb-like configuration, a Ti-mesh scaffold was designed to minimize stress shielding, while simultaneously withstanding failure under physiological loading. A second 3.5-fold stiffer scaffold was designed by increasing the strut diameter from 1.2 to 1.6 mm. This resulted in two scaffolds with a bulk stiffness of either 0.84 GPa (soft) or 2.88 GPa (stiff) (table S1). In comparison to the scaffolds under clinical evaluation, even the stiff scaffold was markedly softer. The combination with plate fixation allowed us to further modulate the stiffness at the defect site: The stiffness of the 4-cm empty tibia defect stabilized with the half-shell steel plate (shielding plate) was five times higher than the stiffness of the defect stabilized with a locking compression plate (LCP: 555 N/mm, shielding plate: 2857 N/mm; table S2).

FE calculations showed that under a combined compressive and bending load, the maximum von Mises stresses in both scaffolds under both fixation systems were lower than the material’s yield strength (soft + LCP: 350 MPa, stiff + LCP: 200 MPa; soft + shielding plate: 100 MPa, stiff + shielding plate: 80 MPa, and material yield strength: 600 MPa). In both scaffolds and with both fixation systems, the largest stresses were located at the lateral side, opposite to the fixation plate (Fig. 1, A to C). Similarly, within the callus tissue, the largest strains were predicted in the lateral region, with higher values in the soft than in the stiff scaffold for both plate fixation systems (Fig. 1D). The stresses in the scaffold and within the callus region were highest in the soft scaffold group stabilized with an LCP. In addition, scaffold stresses and callus strains were higher in the stiff scaffold group stabilized with an LCP than in the soft scaffold group stabilized with the shielding plate. Lowest scaffold stresses and callus strains were found in the stiff scaffold group stabilized with the shielding plate (Fig. 1D).

Fig. 1 Finite element model and calculations.

(A) Three-dimensional (3D) finite element (FE) model of the 4-cm tibia defect in sheep stabilized with a locking compression plate (LCP) and augmented with a Ti-mesh scaffold. (B) 3D FE model of the 4-cm bone defect in sheep stabilized with a custom-made shielding plate (half-shell steel plate) and augmented with a Ti-mesh scaffold. (C) Honeycomb scaffolds formed by titanium struts of the same length (7 mm) but different diameters (1.2 and 1.6 mm). Scaffolds consist of a cylinder (2 cm in diameter and 4 cm in length) and a hole that follows the medullary canal (1 cm in diameter). Top: Von Mises stresses in soft and stiff scaffolds stabilized with an LCP. Bottom: Von Mises stresses in soft and stiff scaffolds stabilized with a custom-made shielding plate. (D) Top: Absolute maximum (Abs. max.) principal strains within the soft and stiff scaffolds stabilized with an LCP. Bottom: Abs. max. principal strains within the soft and stiff scaffolds stabilized with a custom-made shielding plate.

In vivo verification of bone filling in the sheep model

Twenty-seven adult merino-mix sheep received a critically sized mid-diaphyseal tibia defect of 4 cm in length (2 to 2.5 times the diameter of the affected bone) that was created with a double osteotomy. The soft or stiff scaffold, filled with the same amount of ABG (7.5 ml), was applied within the defect and stabilized with either a common 4.5-mm steel LCP or a rigid, custom-made shielding plate (half-shell steel, AO Research Institute Davos) (fig. S3, A to J) (16). We studied four test groups with successively reduced stiffness at the defect site by a combination of scaffold and plate fixation (soft + LCP < stiff + LCP < soft + shielding plate < stiff + shielding plate). All sheep showed full weight bearing of the operated hindlimb after surgery. To avoid overloading in the soft + LCP and stiff + LCP groups, an external half-shell plaster bandage helped to minimize the risk of very high shear and torsional moments that would otherwise occur during running and flight reactions. In total, three animals had to be excluded from the study because of an anesthesia problem or the development of a tibia fracture after an accidental fall (table S3). Additional in vivo observations and details are noted in the Supplementary Materials.

Radiographic follow-up animal study to verify bone bridging

Standardized anterior-posterior and lateral-medial radiographic follow-up images of all groups were performed monthly to document and score the healing process. Comparing the soft + LCP group and the stiff + LCP group, the soft + LCP group showed an earlier complete bridging of the large segmental defect (Fig. 2). In both groups, bone growth was predominantly seen on the contralateral side of the plate fixation system. The mineral density of the callus formation progressed until week 24. Eight weeks after surgery, two animals in the soft + LCP group were scored as A (Fig. 2C; score A = complete bony bridging, score B = incomplete bony bridging), due to a complete bony bridging along the outer surface of the scaffold. Twelve weeks after surgery, three animals in the soft + LCP group showed bony bridging, compared to zero animals in the stiff + LCP group (Fig. 2, D and K). The soft + LCP group tended to experience earlier complete bridging of the large segmental defect (Fisher’s exact test, P = 0.091). In the soft + LCP group, four of six animals received a score of A at week 24. The stiff + LCP group showed the early formation of a slight callus along the outer surface of the scaffold at 4 weeks (Fig. 2I), which continued to extend over the course of 24 weeks; however, complete bridging still developed later than in the soft + LCP group. It was not until 24 weeks after surgery that three of six animals in the stiff + LCP group received an A score for a complete bony bridging of the segmental defect on the scaffold surface (Fig. 2N).

Fig. 2 Radiographic analysis.

Radiographic follow-up images (anterior-posterior plane) of one representative animal in the soft + LCP group and the stiff + LCP group at 0, 4, 8, 12, 16, 20, and 24 weeks after surgery. (A to G) Soft + LCP group. (H to N) Stiff + LCP group. All images were scored as complete (score A) or incomplete (score B) bony bridging of the critical size defect. A score of A was given when both tibia segments were connected by bone along the outer surface of the scaffold [(C), (D), (E), (F), (G), and (N)]. A score of B was given for an incomplete bridging of the critical size defect [(B), (I), (J), (K), (L), and (M)]. Star indicates P = 0.091, Fisher’s exact test.

Because of the radiodensity of the large shielding plate used in the soft + shielding plate and stiff + shielding plate groups, the progression of callus formation and a scoring could not be reliably investigated on radiographs. The radiographs confirmed that there was no scaffold or plate failure or displacement over the course of time in any of the four groups. At 8 weeks after surgery, two animals in the soft + LCP group showed radiolucency around one of the small fragment screws used to secure the scaffold, which resulted in a radiologic loosening of this screw over time; however, neither sheep showed signs of instability in the construct as a whole. Three animals in the soft + shielding plate and stiff + shielding plate groups developed radiolucency around the first proximal screw of the plate 8 weeks after surgery, slightly progressing over the course of time but without a loosening of this screw.

Histomorphological evaluation 24 weeks after surgery

Histomorphological evaluation was performed on a thin section of each sample in the mid-sagittal plane through the center of the Ti-mesh network stained with Safranin Orange/von Kossa (Fig. 3 and fig. S4). The scaffolds of all four groups were intact and centrally located between the proximal and distal tibia segments at 24 weeks after surgery. The bone healing result was most advanced in the soft scaffold + LCP group with a gradual decrease in bone formation from the stiff + LCP group to the soft + shielding plate and stiff + shielding plate groups (Fig. 3). The segmental defects in all four groups showed a clear spatial difference in healing, with advanced bone formation on the lateral side and less mineralization on the medial side (Fig. 3, A to D). The lateral dominating mineralized callus formation originated from the adjacent proximal and distal parts of the tibia (Fig. 3, A to D). In the soft + LCP and stiff + LCP groups, the tibia showed greater periosteal activation with mineralized callus formation, whereas in the soft + shielding plate and stiff + shielding plate groups, the origin of bone formation seemed to stem from the endosteum predominantly. In the soft + LCP and stiff + LCP groups, bone grew simultaneously on the outer surface of the scaffold, through the pores, and filled the central hole of the scaffold, whereas in the soft + shielding plate and stiff + shielding plate groups, bone grew predominantly through the scaffold pores. Five of six samples in the soft + LCP and stiff + LCP groups showed a unilateral bony defect bridging through the honeycomb-like pores and on the outer surface of the scaffold. Only two animals in the soft + shielding plate and stiff + shielding plate groups achieved a continuous bony bridging through the lateral scaffold’s pore network. In addition, in the soft + LCP group, the central cavity of the scaffold was filled with large amounts of woven bone that closed the medullary cavity and connected both tibia ends (Fig. 3A), whereas the stiff + LCP, soft + shielding plate, and stiff + shielding plate groups showed gradually less bone growth within the central channel (Fig. 3, B to D).

Fig. 3 Histomorphological evaluation of the segmental defect healing.

(A to D) Safranin Orange/von Kossa–stained thin section of a representative tibia sample in the mid-sagittal plane from each group; additional images are given in fig. S4. Scale bars, 10 mm. La, lateral; Me, medial; Pr, proximal; Di, distal, contralateral to the plate. (E) Sample harvested within the first weeks after surgery used as a control. The augmentation of the honeycombs with autologous cancellous bone graft (ABG) (remnants stained in black; marked with a blue star) and a prominent osteotomy hematoma in the central hole of the scaffold (white star) are still visible.

Bone grew through the three designed open channels (axial, perpendicular, and tilted orientation) of the scaffold relative to the tibia axis (Fig. 4, A to H). A thin, unmineralized layer remained around each strut in the soft + LCP and stiff + LCP groups (Figs. 4, A and B, and 3, A and B), whereas in the soft + shielding plate and stiff + shielding plate groups, bone was detected in closer contact with the titanium (Fig. 3, C and D). A well-aligned collagen fiber network was visible on the surface of each single honeycomb strut (Fig. 5, A to C, small yellow arrows). In all groups, bone formation via direct intramembranous ossification was still active at the lateral and medial defect sides 24 weeks after surgery. Bone surface osteoblasts synthesized a new layer of osteoid on the woven bone (Fig. 5D, white triangles). Indirect endochondral bone formation via hyaline cartilage was prominently detectable in the soft + LCP group [Fig. 5, E and F (blue triangles), and figs. S4 and S5]. The cartilage was located on the surface of the woven bone as a layer, in which the chondrocytes showed signs of hypertrophy and the surrounding matrix had started to mineralize. Hyaline cartilage was also detectable close to the Ti struts within the connective tissue. Three animals in the soft + LCP group showed a band of hyaline cartilage proceeding horizontally through the defect (fig. S4, A to C, yellow arrowheads). Advanced bone remodeling was apparent in the soft + LCP group compared to the other three groups, because the bone was, in large areas, already lamellar structured in one animal on the lateral outer surface of the scaffold (fig. S4D, white star).

Fig. 4 Bone tissue formation within the Ti-mesh channels.

Thin sections in the mid-sagittal plane stained in Safranin Orange/von Kossa of one animal in the soft + LCP group and the stiff + LCP group, 24 weeks after surgery (A and B); scale bars, 10 mm. (C to H) Magnifications of (A) and (B) show the bone growth through different scaffold channels (Ti struts). The bone growth on the scaffold’s lateral outer surface is marked with a dotted blue arrow [(C), (E), and (F)]. The dashed yellow arrow follows the axial scaffold channel (D). Dashed red arrows follow the direction of the perpendicular channel [(C), (D), (F), and (G)]. Dashed green arrows mark the tilted channels of the scaffold [(C), (D), (E), and (H)]. LaDiCt, lateral, distal, corticalis; MeDiCt, medial distal corticalis (G and H). Scale bars, 500 μm.

Fig. 5 Connective tissue and cartilage formation within the Ti-mesh channels.

(A to C) Connective tissue formation around the Ti struts guiding the collagen fiber orientation and alignment (yellow arrows). Bone growth followed this secondary network via direct ossification [(D) surface osteoblasts synthesizing osteoid, white triangles; osteoid, red surface layer] or endochondral ossification via cartilage formation and secondary mineralization (E and F, blue triangles). (A to D) Scale bars, 100 μm. (E and F) Scale bars, 200 μm.

As a reaction to the osteotomy and the internal fixation, a reduction in density of the adjacent cortical bone was visible in three samples in the soft + LCP and stiff + LCP groups and two samples each in the soft + shielding plate and stiff + shielding plate groups (fig. S4, C, E, F, I to K, N, Q, V, and X). A slight lateral muscle prolapse into the defect side as a histological sign of a nonunion was noted in one animal each in the soft + shielding plate and stiff + shielding plate groups (fig. S4, N and T, blue arrowheads); the soft + LCP and stiff + LCP groups showed no evidence of nonunion. In one sample in the soft + shielding plate group, remnants of the ABG filling were still detectable 24 weeks after surgery (fig. S4Q, blue star). No signs of an implant-related inflammatory reaction could be observed in any group. All samples showed a limited number of osteoclasts. Osteoclast density was quantified exemplarily within the 4-cm defect [total region of interest (ROI)] in one sample in the stiff + LCP and stiff + shielding plate groups [0.51 osteoclasts per mineralized bone area (mm2) and 1.34 osteoclasts per mineralized bone area (mm2); fig. S5, L to N, and table S4].

Histomorphometrical evaluation at 24 weeks after surgery and comparison to mechanical tissue straining

Quantitative measurements of mineralized bone, soft tissue, and hyaline cartilage in the ROIs (lateral and medial total ROIs) confirmed the histomorphological result of an asymmetrical mineralized callus formation to the lateral defect side, corresponding to the higher lateral strains predicted by FE analyses (Fig. 6, A to D). At the lateral side, interquartile absolute maximum principal strains were between 0.65 and 0.35% in the soft scaffold and between 0.35 and 0.19% in the stiff scaffold (Fig. 6C). At the medial side, ranges of interquartile absolute maximum principal strains were 0.24 to 0.1% and 0.12 to 0.064% in the soft and stiff scaffold, respectively. In the central callus region, a gradual increase in mineralized bone area was measured, coinciding with higher strains (lower stress shielding) in the FE analyses (soft + LCP group < stiff + LCP group < soft + shielding plate group < stiff + shielding plate group; Fig. 6, E to G). In the central ROI, ranges of interquartile absolute maximum principal strains were 0.6 to 0.23% and 0.32 to 0.1% for the soft and stiff scaffolds with LCP plate and 0.17 to 0.03% and 0.09 to 0.02% for the soft and stiff scaffolds with the shielding plate, respectively (Fig. 6E). The endochondral bone formation process, quantified as the amount of hyaline cartilage area in the total medial and lateral ROI, was significantly higher in the soft scaffold + LCP group in comparison to the stiff scaffold + LCP group (Mann-Whitney test, P = 0.026; Fig. 6H), quantitatively underlining the histomorphological findings.

Fig. 6 Histomorphometrical evaluation of the segmental defect healing 24 weeks after surgery and quantification of mechanical strains.

(A and B) The regions of interest (ROIs) within the FE model and the histological section were determined as a total 40-mm defect ROI (light blue rectangle), which was divided into a lateral (La) and a medial (Me) part, and a central ROI with a height of 10 mm (dark blue rectangle). Mid-sagittal thin section of a sample in the soft scaffold + LCP group 24 weeks after surgery (Safranin Orange/von Kossa staining). (C) Interquartile absolute maximum principal strains in the lateral and medial ROI of the soft and stiff scaffolds stabilized with an LCP. (D) The mineralized bone area in the medial and lateral ROI (in %; a dashed gray reference line marks the area filled with ABG during surgery = 7% ABG filling of the total ROI). (E) Interquartile absolute maximum principal strains in the central ROI for all groups. (F) The mineralized bone area in the central ROI. (G) Diagram showing the effect of stiffness (scaffold + plate fixation) and mechanical stimulation at the defect side on the bone regeneration response. (H) The cartilage area (as a measurement of endochondral ossification) in the total ROI (Mann-Whitney test, n = 6, P = 0.026).

Analysis of collagen fiber orientation

Second-harmonic generation (SHG) imaging was performed to analyze the collagen fiber orientation within the soft and stiff Ti-mesh scaffolds. The aim was to evaluate whether the scaffold functioned as a guiding structure for soft and mineralized tissue during bone regeneration (soft + LCP group and stiff + LCP group) (Fig. 7, A to X).

Fig. 7 Collagen fiber orientation analysis.

(A to D) Illustration of the three orientations of the continuous open pores (channels). The dashed yellow arrow denotes the axial oriented channel; the dashed red arrow shows the perpendicular channel relative to the scaffold axis; the dashed green arrow shows the tilted channel relative to the scaffold axis; the dashed blue arrow shows the collagen fiber orientation on the scaffold surface. (E, J, O, and T) Bone tissue illustrated as a schematic in gray area within two PMMA-embedded samples in the soft + LCP) and stiff + LCP groups used for second-harmonic generation (SHG) imaging. (G, I, L, N, Q, S, V, and X) SHG imaging performed within the central blue squares in (E), (J), (O), and (T). Primary fiber orientation is visualized as a green line in each sub-ROI (F, H, K, M, P, R, U, and W). The collagen fiber orientation follows the orientation of the different continuous open pores [green and yellow arrows in (G), (L), (X)]. Young woven bone shows unorganized collagen fibers [(P) to (S), (U) and (V) denoted with a blue star]. Ti, titanium scaffold; white scale bars, 1 mm.

Qualitatively, collagen fibers on the outer surface of the scaffold, within the connective tissue (medial side) and within the bone tissue (lateral side), were well aligned and oriented along the tibia long bone axis (Fig. 7, F to I and K to N, blue arrow). In addition, along the open channels in the tilted (Fig. 7, F to I, K, L, W, and X, green arrow), perpendicular (Fig. 7, K and L, red arrow), and axial (Fig. 7, W and X, yellow arrow) orientation relative to the scaffold axis, a long-range alignment was observed. The analyzed lateral ROI at the mid-height of the defect showed brighter SHG signals and advanced mineralization compared to the medial side, where the tissue was primarily composed of fibrous tissue (Fig. 7, compare F and G to H and I, K and L to M and N, P and Q to R and S, and U and V to W and X). Higher SHG signal intensities were associated with denser collagen fibers in the mineralized versus nonmineralized tissue.

Quantification of fiber orientation revealed that in both scaffold types, collagen fibers were oriented parallel to the Ti surface on all protruding surface elements (surfaces confined by two corners with internal angles <180° as illustrated in fig. S6, A1, A2, and C). Accordingly, the angles ΔΦ between the primary collagen fiber orientation and the orientation of the Ti surface were low for these regions (fig. S6, B and C). In contrast, surfaces confined by at least one corner with an internal angle >180° (fig. S6, A, B1, and B2) showed variable fiber alignment. At these surfaces, occasionally, a strong mismatch between surface and collagen fiber orientation with ΔΦ of up to 90° was found (fig. S6, B and C). When fiber orientation in the center of the scaffold pores was compared to the orientation at the two opposing, parallel Ti surfaces framing the pore, deviations were predominantly small (|ΔΦS1,S2 − ΔΦC| < 20°) (fig. S7A). However, at individual regions, the fiber orientation in the channel differed markedly from the orientation at the scaffold surface (fig. S7B). This was the case when long-range tissue guidance in scaffold channels competed with local surface guidance (fig. S7C). For example, the center between the four scaffold struts in Fig. 7K shows a diagonal orientation (top left to bottom right) even though the next surfaces are oriented vertically (sketched in fig. S7C, right). This location also shows an exemplar of the crossing of two optional pore directions (both diagonal), one of which becomes the dominating one for tissue patterning. These results indicate that collagen fibers inside the scaffold are locally guided by the Ti surface, especially by surface A1, A2, and C representing convex regions of the Ti-scaffold architecture. In addition, the pore architecture with open channels through the entire scaffold supports long-range tissue patterning in distinct directions (see Fig. 7) that might compete with local surface guidance.

Connective tissue formation seemed to be guided by the scaffold pore architecture, illustrated by the above-described collagen fiber alignment in both scaffold designs. Subsequently, collagen fibers functioned as a templating structure for tissue mineralization and guided bone into the scaffold pore channels and along the outer scaffold surface. The soft and the stiff scaffold showed the same principles of collagen fiber alignment within the scaffold pore channels and on the outer scaffold surface. In contrast to the collagen fibers of the connective tissue, younger woven bone showed a more disoriented local collagen fiber network in both scaffold groups (Fig. 7, P, Q, R, S, U, and V, blue stars).

Analysis of mineralized bone structure

Backscattered scanning electron microscopy (B-SEM) imaging was performed to analyze the mineralized bone structure and relative mineral density during bone regeneration within the soft + shielding plate and stiff + shielding plate groups (Fig. 8, A to I). Corresponding to the histological analysis, asymmetrical bone formation dominating the lateral defect side was evident in the soft and stiff scaffolds (Fig. 8, A and B). The overview images of both scaffold designs showed bone growth from the adjacent tibia through the scaffold pore network, which guided the regeneration process (Fig. 8, A and B). Areas of newly formed bone were characterized by a woven bone structure with a lower degree of mineralization (darker gray scale) and less dense organization (Fig. 8, C to E) as compared to areas of older bone, located close to the adjacent tibia (Fig. 8F), which showed a dense plexiform structure and were characterized by a higher degree of mineralization. Both scaffold designs were integrated by newly formed bone at the interface to the adjacent tibia ends (Fig. 8G). The lamellar structured adjacent tibia exhibited higher mineral content than the plexiform bone, visualized by a brighter gray scale signal intensity (Fig. 8G). However, the corticalis showed a reduction of the bone density at the interface close to the scaffold with Haversian remodeling (Fig. 8H). A direct contact between Ti struts and bone tissue was more prominent in the stiff scaffold group (Fig. 8, A, B, C, and I).

Fig. 8 Analysis of bone structure within the soft and stiff scaffold.

(A and B) Backscattered scanning electron microscopy overview images of one sample each from the soft + shielding plate group (A) and the stiff + shielding plate group (B). (C to E) Woven bone formation as the “first wave” of critical size defect regeneration. Newly formed bone was characterized by a lower mineral content (darker gray scale). (F) Dense plexiform bone formation at the bone-scaffold interface characterized the “second wave” of critical size defect regeneration. Brighter (gray) areas show a higher degree of mineralization. (G) Bony integration of the Ti-mesh scaffold with plexiform bone, which had a lower mineral content in comparison to the adjacent cortical bone. (H) The higher magnifications of the cortical bone adjacent to the scaffold showed a reduction of the cortical bone density with Haversian remodeling. (I) Direct bone-scaffold contact. (A and B) Scale bars, 2 mm; (C to F) scale bars, 500 μm; (G) scale bar, 1 mm; and (H and I) scale bars, 500 μm.

DISCUSSION

We investigated whether Ti-mesh scaffolds could enhance large bone defect healing in vivo by tuning the scaffolds’ mechanical properties. Design principles and FE techniques were used to design a relatively soft Ti scaffold to minimize stress shielding effects that enabled the incorporation of bone graft material, while avoiding material failure. We hypothesized that soft, mechanobiologically optimized Ti-mesh scaffolds would improve the regeneration of large segmental bone defects compared to stiffer scaffolds. Shielding of elastic deformation in the soft scaffold, using a stiffer scaffold or a rigid plate fixation, reduced the amount of bone formation in a critically sized, 4-cm large bone tibia defect model in sheep at 24 weeks. Using two scaffold stiffnesses and two different plate fixation systems in combination, we gradually reduced the elasticity within the defect area. Bone defect regeneration was enhanced in the soft scaffold with the common LCP (soft + LCP group), which enabled increased load transmission to the scaffold. The stiff scaffold and the groups with the custom-made half-shell plate shielding the defect side from mechanical loading showed decreasing amounts of bone regeneration (stiff + LCP, soft+ shielding plate, and stiff + shielding plate groups; Fig. 6, F and G).

The radiographic follow-up analysis of the soft and stiff scaffolds in combination with the LCP showed similarities in bone formation patterns to those observed in our clinical cases (fig. S1 and Fig. 1, soft + LCP group and stiff + LCP group). In both soft and stiff 3D Ti-mesh scaffolds, early bone formation was only visible on the implant surface. Bone growth on the outer surface led to a complete bony bridging of the critical size segmental defect that dominated the lateral defect side and was first visible in two animals in the soft + LCP group just 8 weeks after surgery. On radiographs, the soft + LCP group showed faster bone regeneration with earlier bridging of the critical size defect as compared to the stiff + LCP group, which can be explained by a higher mechanical stimulation of the healing region induced by the soft scaffold. The bony bridging of the critical size segmental defect was apparent much later in the stiff + LCP group. As a consequence of the medial plate fixation system in all groups, highest strains were determined via FE analysis at the lateral side, matching with the asymmetrical healing progression. Smaller strains were predicted close to the fixation plate (medial side), corresponding to less bone formation on this side, confirming that bone formation can be enhanced by mechanical stimulation of the appropriate magnitude.

Ti-mesh cage implants were approved for the reinforcement of deficient bone in 1970 by the Food and Drug Administration but have not been optimized to supply a local mechanical stimulus at the defect site. These cage implants have been used in spinal fusion procedures (17) and in diaphyseal and metadiaphyseal segmental tibia defects or foot and ankle reconstructions (18, 19), packed with various autologous and/or allogene bone grafts (as in the present study). The design was rather simple, as only one outer mesh layer in a rhombic design formed the cage (Harms design). Some studies showed an early functional recovery, but due to limited analysis methods in patients and metal artifacts in CT scans, patterning of bone within the cage was not previously thoroughly analyzed (1719). Unlike the 3D Ti-mesh scaffold of the present study, rhombic designed cages (20) were adjusted to avoid implant failure rather than to augment mechanical stimulation within the defect or to induce specific biological processes.

Histological analysis of the inner part of the Ti-mesh scaffold 24 weeks after surgery confirmed the radiographically observed asymmetrical lateral bone formation in all four test groups. Higher mechanical stimulation led to increased bone formation. Corresponding to the predicted higher mechanical stimulation on the lateral defect side, bone ingrowth dominated the lateral defect side. In addition to the bone growth on the outer lateral scaffold surface, histological evaluation unveiled that bone grew also through the network of scaffold struts. The soft and stiff scaffolds stabilized with the LCP showed an advanced healing result with unilateral bridging of the critical size segmental defect in five of six samples, whereas only two samples in the groups stabilized with the half-shell shielding plate fixation exhibited continuous bony bridging through the honeycomb-like scaffold pores. The group with the highest mechanical stimulation (soft + LCP group) led to larger bone formation in the central scaffold pore and further remodeling of the lamellar bone structure, as compared to the three stiffer groups. Compared to the lateral side, the medial defect side reflected a premature healing stage in all groups. The pores at this scaffold side were predominantly filled with connective tissue, but bone ingrowth from the defect ends had already taken place. It is likely that the larger soft tissue coverage of the lateral tibia aspect, which is known to improve bone defect healing by ingrowth of cells and vessels, supported the asymmetrical callus formation (21). At 24 weeks, active direct and indirect bone formation was ongoing, predominantly in the soft + LCP and stiff + LCP groups. Thus, we expect that a complete recovery of the critically sized defect would have been reached after several additional weeks in the soft + LCP and stiff + LCP groups, whereas the groups with the shielding plate fixation (soft + shielding plate and stiff + shielding plate groups) would have likely resulted in a nonunion.

SHG imaging analysis of the soft and stiff scaffolds (soft + LCP and stiff + LCP groups) revealed highly aligned collagen fibers that were oriented parallel to the outer surface of the scaffold and that corresponded to the bone axis within the soft tissue (medial) and the bone tissue (lateral). Within the scaffolds, collagen fibers were oriented parallel to the surface of the individual struts, highlighting the strong guiding character of the Ti surface. A long-range alignment of collagen fibers was observed in the direction of open channels within the scaffold pore network. This indicated that the collagen fiber network was predominantly guided by the pore geometry of the scaffold. The collagen fiber network functioned as a secondary, cell-deposited template, supporting the ingrowth of bone into both scaffold types. This mechanism has previously been demonstrated for another porous, resorbable, composite polycaprolactone/β-tricalcium phosphate scaffold in a smaller segmental defect model in sheep (9, 21, 22) and supports the concept of the extracellular matrix working as a natural scaffold for all cellular events during successful bone defect regeneration (23).

B-SEM imaging analysis of the mineralized bone structure within the soft and stiff scaffolds (soft + shielding plate and stiff shielding plate groups) supported the SHG imaging analysis, as bone formation originating from the adjacent tibia was guided through the open porous scaffold channels. Bone regeneration began with the formation of woven bone, which showed a less organized and less mineralized structure, whereas older bone at the interface between adjacent tibia and scaffold had already been augmented and replaced by plexiform bone. From these analyses, the postulated “two waves theory” of cortical bone growth in an unimpaired defect healing situation (24, 25) was recapitulated within the Ti-mesh scaffolds in a critically sized defect model. The initially highly porous primary bone structure serves as an endogenous scaffold to direct the following formation of lamellar bone with improved mechanical properties (24, 25).

We could show that Ti-mesh scaffolds can be mechanically optimized, proving the hypothesis that a soft scaffold minimizing stress shielding promotes the bone healing response. Mechanical strains at a certain magnitude are known to support tissue mineralization and bony callus formation (26, 27). Here, we showed that the mechanical strains within the soft Ti-mesh scaffold group stabilized with a clinical plate were favorable for segmental bone defect regeneration, because endochondral bone formation was enhanced. The soft + LCP group showed a larger area of hyaline cartilage with hypertrophic chondrocytes. Hypertrophic chondrocytes can secrete angiogenic and osteogenic factors that play pivotal roles in the vascularization of the defect and the deposition of mineralized extracellular matrix that result in bone formation (28). Low mechanical strain has been shown to promote intramembranous bone formation, whereas intermediate mechanical strains lead to endochondral ossification. This has been shown using theoretical models based on FE techniques (29, 30) and corroborated using experimental measurements of callus tissue strain using digital image correlation (27). In agreement with this, in our study, the endochondral bone formation process was stimulated by the higher mechanical strains acting within the soft titanium scaffold.

The mechanical strains within the scaffold pores in our large bone defect were in the range of 0.6 to 0.02%. Previous studies have suggested that strain ranges between 5 and 0.04% are beneficial for bone formation, whereas below 0.04%, bone resorption takes place (31, 32). The strains within the scaffold pores in our large defect are one order of magnitude lower than those found in an uncritically sized osteotomy gap in sheep leading to successful healing (30). We have previously shown that large bone defects lead to a reduction in the mechanical strains, under which bone healing takes place (15). In this study, the incorporation of a Ti-mesh scaffold further reduced the mechanical environment within the regenerating region (fig. S8). Ti-mesh scaffolds are stiff structures that give excellent mechanical stability for large bone defects; however, they do not transmit high amounts of mechanical stimuli to the fracture site. Despite this, we could demonstrate that optimizing their mechanical properties can favorably influence the bone healing response. The ABG filling provided a sufficient biological stimulus to initiate the bone healing response; however, it is known to be insufficient as a stand-alone therapy for critical size defects (3336), emphasizing the importance of additional mechanical stimulation for bone defect regeneration.

Our results suggest that Ti-mesh scaffolds for large segmental defect regeneration should be designed (i) to minimize stress shielding in the regenerating region, where strains in the range 0.23 to 0.6% seem beneficial for endochondral and intramembranous ossification; (ii) to facilitate resistance against implant failure; (iii) to guide the fibrous and subsequent bony tissue organization on the scaffold surface and through oriented pores; and (iv) to provide large interconnected pores for incorporation of graft materials as biological stimuli to initiate bone regeneration and subsequent bony restoration.

We investigated the potential to foster bone healing in large bone defects by optimizing the Ti-mesh scaffold stiffness combining FE analyses and additive manufacturing technique. So far, only a few studies have addressed defect healing by cages, using commercially available cylindrical titanium mesh cages (in Harms design). Lindsey et al. (37) filled a mesh cage with morselized canine cancellous allograft and canine mineralized bone matrix for the regeneration of a mid-diaphyseal segmental defect of 3 cm length in dogs, stabilized with an intramedullary nail. Similar to our study, bone growth was evident on the outer surface and integrated through the pores but no endochondral bone formation was evident. Fujibayashi et al. (38) investigated a chemical and thermal surface treatment of a stiff titanium cylindrical Ti-mesh cage in a rabbit with a 10-mm mid-diaphyseal femur defect (38). They also observed new bone formation on the outer surface of the bioactive mesh cage. Penetration of the mesh structure by bone with mechanical interlocking between cage and bone was visible in microradiographs, but they did not observe scaffold-guided tissue patterning (38). These simple designed cages were not mechanically optimized to promote bone regeneration.

The use of highly biocompatible 3D Ti-mesh scaffold for the reconstruction of large segmental bone defects is valuable for complex 3D anatomic defects (pelvic and peri-acetabular region). Concerning limb reconstruction, this surgical augmentation technique could be superior to distraction osteogenesis or vascularized cortical bone graft transfer. The immediate limb stability and original limb length restoration lead to early limb and joint mobilization and is comparable to the situation after endoprosthetic solutions, while additionally offering all the advantages of biological restoration (18, 20). The easy filling of the construct with ABG or other bone graft substitutes allows for additive biological stimulation. Although the technique has been used in patients, little was known about the potential to tune the mechanical properties of these scaffolds to promote bone regeneration. Our study supports the hypothesis that endogenous bone defect regeneration within 3D Ti-mesh scaffolds can be promoted by minimizing the scaffold mechanical stiffness.

As a limitation of the study, the half-shell cast bandage that was used in the soft + LCP and stiff + LCP groups to avoid high shear and torsional forces that occur during running and jumping (39) was not modeled in the FE analysis. Nevertheless, the cast did not hinder the regular limb loading during normal slow walking but mainly avoided torsional overloading. Thus, the internal loads at the scaffold could be assumed to be comparable (39, 40) to the clinical situation in patients (41) that avoid extreme loading after surgery as well. As another limitation, the collagen fiber orientation could only be analyzed in 2D; thus, the complex 3D arrangement of fibers within the scaffold was not assessed. Except for the radiological evaluation, only healing after 6 months after surgery was analyzed because of the limited the number of animals permitted to answer our research question. Although additional investigation time points and a destructive biomechanical testing would have been useful to add as information, we are convinced that the current data show the key message of the study.

In summary, we showed that a relatively soft, mechanically optimized Ti-mesh scaffold filled with ABG enhanced regeneration in a large segmental bone defect in sheep, promoting endochondral and intramembranous bone formation. Using bone healing as an example system, we showed that 3D printing can be used to influence the biological cascades of tissue regeneration. Ti-mesh scaffolds could be mechanobiologically optimized by design to reduce stress shielding within the given defect, while the geometrical design of the interconnected pores enabled easy incorporation of graft material for an initial biological stimulus and guiding of the bone regeneration process.

MATERIALS AND METHODS

Study design

The objectives of the study were to establish a process for producing patient-specific, 3D additive manufactured Ti-mesh scaffolds in patients and to investigate whether critical segmental bone defect healing can be enhanced by mechanobiological optimization of 3D Ti-mesh scaffolds. FE techniques were used to design two mechanically distinct Ti-mesh scaffolds in a honeycomb-like configuration to minimize stress shielding, while ensuring resistance against mechanical failure. To demonstrate efficacy, we established a preclinical large segmental tibia defect model in sheep. Four test groups with a gradually reduced stiffness at the defect site were investigated in vivo. Experimental groups included defect reconstruction (i) with soft Ti-mesh scaffold and LCP fixation (n = 6), (ii) with stiff Ti-mesh scaffold and LCP fixation (n = 6), (iii) with soft Ti-mesh scaffold and shielding plate fixation, or (iv) with stiff Ti-mesh scaffold and shielding plate fixation. Bone regeneration was monitored radiologically monthly over the course of 24 weeks. By histological, SHG imaging, and B-SEM analysis, bone regeneration within the scaffold was investigated 24 weeks after surgery. A total of 27 animals were operated on, with 24 (n = 6 for each group) reaching end point. Three animals were excluded earlier from the study because of anesthesia complication or accidental fracture. No outliers were eliminated.

Large bone defect treatment with 3D Ti-mesh scaffolds in patients

A selected set of 19 patients received Ti-mesh scaffolds as a treatment for large bone defects. The patient population was heterogeneous with respect to the underlying cause of the defect (tumorous destruction or nonunion after trauma). Patients differed by age, gender, and medical histories. Different defect sites and sizes were treated (large segmental defect of the proximal and distal part of femur and humerus, as well as extended maxillofacial defects). Each patient received an individualized Ti-mesh scaffold manufactured using a rapid prototyping procedure that was based on the 3D CT data of the defect, in combination with an additive stabilization. The scaffolds were custom-made by DePuy Synthes in a laser sintering process with a honeycomb-like structure, but without mechanobiological optimization for constraints of load-bearing segmental defects. Because of the availability, defect localization, and defect size, some scaffolds were augmented with RIA material harvested from the femur or stuffed with a fibula transfer. Patients received radiographic follow-up images and CT imaging at different time points, depending on their clinical condition. The results of two representative patients [a 61-year-old woman (fig. S1) and a 45-year-old man (fig. S2)] who received a Ti-mesh scaffold treatment are shown (for patient history, see the Supplementary Materials).

Scaffold design and strain analysis

On the basis of our previous study on the influence of scaffold architecture on load transfer behavior (11), we designed two scaffold prototypes (soft and stiff) that have distinct mechanical stiffness yet retain the same architecture. The scaffolds consist of titanium struts of different diameters arranged in a honeycomb-like configuration. This configuration was previously shown to have a good resistance against mechanical failure and a small change in beam diameter resulted in a considerable change in the apparent stiffness of the scaffold (11). The soft and stiff scaffolds were selected out of a large parametric analysis in which the beam length (seven length sizes) and diameter (three diameters) of honeycomb-like structures were modulated. For the final designs, the same beam length of 7 mm was chosen for easy filling with ABG. The soft scaffold was designed to minimize stress-shielding effects and to avoid mechanical overloading and failure. The scaffolds consisted of a cylinder (2 cm in diameter and 4 cm in length) with a central hole that followed the medullary canal (1 cm in diameter; Fig. 7, A to C).

To determine the mechanical behavior of the scaffolds when implanted at the defect site, we developed FE models of the two configurations. The models included the cortical bone, the osteotomy gap, the Ti-mesh scaffold, and the fracture fixation system (a nine-hole LCP plate with six interlocking screws or a custom-made half-shell steel plate; AO Research Institute Davos) (16). In addition, to determine the mechanical strains induced within the scaffold pores and to investigate the ranges of mechanical strain under which bone formation occurs, the scaffold pores and the medullary canal were filled with a soft callus material resembling early stages of healing.

The mechanical behavior of the scaffolds and the strains induced within the healing region were investigated under a combined static load, including axial compression and anterior-posterior bending. Static compression and bending loads were applied as axial ramps (1 Hz) and strain magnitudes were determined at the peak load. Loads were applied to the bone’s proximal end, while the distal end was constrained in translation and rotation. Loading magnitudes were set as follows:

(1) Proximal-distal axial load (compression): 1372 N, equivalent to two times the average animal body weight (70 kg) under normal walking conditions (40).

(2) Anterior-posterior load (bending): 86 N. Applying the Euler-Bernoulli beam theory, this load would result in a maximum bending moment of 0.025 BWm at the fixed side in an intact bone of an equivalent size (about 20 cm long) (40).

The following material properties were assigned: Ti alloy (scaffolds): Young’s modulus: 104 GPa, Poisson’s ratio: 0.30; according to ASME B31.1-1995, steel (plate and screws): Young’s modulus: 200 GPa, Poisson’s ratio: 0.305; according to ASME B31.1-1995, bone tissue: Young’s modulus: 17 GPa, Poisson’s ratio: 0.30 (42); soft callus: Young’s modulus: 0.2 MPa, Poisson’s ratio: 0.167 (42).

Fixation plate stiffness

FE models were used to determine differences in the stiffness of the two fixation plates (a nine-hole LCP plate with six interlocking screws and a custom-made half-shell steel plate) mounted onto the 4-cm segmental tibia defect. The two fixation systems were tested under unconfined compression. FE models included the bones, the plate, the screws, and the 4-cm bone defect. Material properties for the different components were set as described in the “Scaffold design and strain analysis” section.

Scaffold fabrication

3D Ti-mesh scaffolds were fabricated by DePuy Synthes in a laser sintering process. Before the preclinical in vivo model, the soft scaffold underwent cyclic mechanical compression testing, which led to failure after 160,000 cycles with a load of 1100 N. Hence, the scaffolds showed the resistance required to avoid breakage in application in a preclinical large animal model.

Experimental animal handling

Twenty-seven adult, female, merino-mix sheep (mean ± SD weight, 62 ± 5 kg; age ≥ 2.5 years) were included in this study. All experiments were carried out according to the policies and principles established by the German Animal Welfare Act, the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals, and the National Animal Welfare Guidelines and were approved by the local legal representative (G0172/12, LAGeSo). The experiment was performed in the animal facility of the Charité–Universitätsmedizin Berlin. All animals received training in the requisite postoperative bandage application and were acclimated to the supportive sling system to reduce stress and facilitate easy handling during the subsequent 24 weeks (the duration of the experiment). The custom-made sling system was combined with a commercially available patient lift (Birdie, Invacare) to hold the animals. This device prevented the sheep from accidentally falling, which would have resulted in a fracture in the operated limb during the critical recovery phase (immediately after anesthesia). An earlier established anesthesia procedure was performed (43).

Scaffold preparation

All sheep were positioned ventrally on their thorax and abdomen for the harvesting of ABG from both iliac crests following an earlier established surgical procedure (44). The harvested ABG was coarsely milled in a micro bone mill (Ergoplant, Aesculap AG). The honeycomb-structured pores in both scaffold types were filled with the same amount of 7.5-ml ABG in all groups.

Surgical procedure of critical size defect osteotomy, stabilization, and scaffold implantation

For the critical size osteotomy, a medial approach to the right tibia was chosen. The sheep was positioned in right recumbency. The skin was prepared sterile and incised 15 cm longitudinally along the tibia with a scalpel. The underlying tibia was freed from the attaching muscles blunt or with a raspatorium. Bleeding was stopped with an electric cautery. In the soft + LCP and stiff + LCP groups, the clinically approved LCP (steel LCP, nine hole, DePuy Synthes) with dynamic locking screws (DLS, DePuy Synthes) was used for stabilization, simulating the surgical conditions in clinical patients. The plates were anatomically precontoured with a bending press according to the curvature of the tibia. The osteotomy gap was positioned between the fourth and sixth hole of the plate, which were left empty. Initially, the distal part of the plate was fixed to the tibia with DLS positioned in the seventh and ninth hole. The proximal part of the plate was only fixed with a 2.4-mm Kirschner wire that was placed in the middle of the third hole. This facilitated the reduction of the bone axis after the creation of the osteotomy gap. The critical size osteotomy defect was created using a custom-made sawing template, which was initially fixed unicortical to the tibia by a DLS (DePuy Synthes). The two slots within the template ensured an osteotomy gap of 4 cm, when an oscillating saw with a particular blade of 0.6 mm thickness (DePuy Synthes) was used. Both saw cuts were performed down to the plate, but for the last millimeters of the cut, the plate needed to be removed from the tibia. The cylindrical bone fragment of 4 cm in length was explanted, and the wound area was carefully cleaned and rinsed with saline solution. Afterward, the plate was again placed over the k-wire on the proximal bone fragment and fixed distally with the DLS in the predrilled holes. The scaffold was positioned in the 4-cm osteotomy defect with the small proximal and distal anchor plates, posterior to the bone axis. For a press-fit implantation of the scaffold, a slight compression of the scaffold was achieved using a tension device. The proximal part of the plate was fixed to the tibia with DLS, after removing the k-wire. The scaffold was fixed with a small fragment screw (DePuy Synthes) to the tibia through the proximal and distal anchor plate. The subcutaneous tissue and skin were closed with continuous sutures (ETHICON Vicryl 3-0 and ETHICON Prolene 3-0, Johnson & Johnson Medical GmbH).

In the soft + shielding plate and stiff + shielding plate groups, a custom-made half-shell steel plate (AO Research Institute Davos) (16) was used for critical sized defect stabilization. The plate ends were fixed to the tibia with four self-tapping locking screws each (DLS, DePuy Synthes). Two rods connected the plate ends and facilitated the view on the segmental defect underneath, which was created using the custom-made sawing template as described for the soft + LCP and stiff + LCP groups without removing the plate. The augmented scaffold was positioned into the osteotomy gap, with the small anchor plates oriented posterior to the bone axis without additional small fragment screw fixation. A screw within the anterior rod of the plate was used for press-fit fixation of the scaffold.

Colloidal aluminum spray, sterile gauze, and synthetic cotton were used for wound covering in all four groups. In the soft + LCP and stiff + LCP groups, an external half-shell plaster bandage (gypsum) was shaped around the ankle joint to reduce extreme high shear and torsional moments during running or flight reactions that might cause a fixation failure. The cast was changed weekly during the 24-week observation time. In the soft + shielding plate and stiff + shielding plate groups, sheep wore a bandage within the first week postoperatively until soft tissue recovery.

Radiographic follow-up images

Standardized anterior-posterior and lateral-medial radiographic images were performed immediately after surgery and 4, 8, 12, 16, 20, and 24 weeks after surgery in all four groups. Additional microradiographs were taken 24 weeks after surgery ex vivo (Faxitron, Bioptics). The position of the plate, screws, Ti-mesh scaffold, failures, and the occurrence of radiolucencies at the screws were assessed. Mineralized callus formation and a bridging of the critical size osteotomy gap were evaluated 4, 8, 12, 16, 20, and 24 weeks after surgery. Bridging of the critical size osteotomy gap was scored. A score of A was given for complete bony bridging along the surface of the scaffolds, connecting both tibia ends with bone grown. The radiographic score B was given for incomplete bony bridging of the osteotomy gap.

Histological preparation and evaluation

The samples were harvested 24 weeks after surgery for undecalcified bone histology of thin sections [polymethyl methacrylate (PMMA); Technovit 9100 new, Heraeus Kulzer GmbH]. All samples were sawed in the mid-sagittal plane to obtain the view on the central part of the scaffold and the adjacent bone marrow of the proximal and distal part of the tibia. A 100-μm-thick section of the central part of the defect was prepared in a sawing and grinding technique (45) and stained in Safranin-Orange/von Kossa. In this stain, mineralized tissue appears black, whereas cartilage and connective tissue is red. Osteoid, as a sign of active direct bone formation, was stained as a red layer on mineralized bone covered with osteoblasts. Hyaline cartilage, developed during endochondral bone formation, was identified by the typical morphology of chondroblasts, chondrocytes forming a chondron, within the matrix that was already partially mineralized (fig. S5, A and B). Tissue types were verified exemplarily by 10-μm-thick sections of the tissue within the 4-cm defect by the use of a laser microtome (TissueSurgeon, LLS ROWIAK LaserLabSolutions GmbH) stained in Movat’s pentachrome, Safranin-Orange/Fast green, and Giemsa to verify the different tissue types (fig. S5, C to K). Tartrate-resistant acid phosphatase (TRAP) staining was performed in one sample from the stiff + LCP group and the stiff + shielding plate group to quantify the amount of osteoclast exemplarily (fig. S5, L to N). Osteoclasts (stained TRAP-positive with at least two nuclei and located on the bone surface) were measured quantitatively in the bony callus of the whole 40-mm defect ROI. Osteoclast density was calculated in relation to the mineralized bone tissue in the ROIs (osteoclasts per mineralized bone tissue in mm2).

Within the osteotomy gap, tissue constitution and distribution and scaffold position were assessed. Newly formed mineralized callus formation and the form and structure of the corticalis adjacent to the osteotomy gap were evaluated. Signs of a nonunion (muscle prolapse into the osteotomy gap, closing of the medullar canal, and atrophic tibia ends) were assessed. An accumulation of inflammatory cells and vessels were identified as a sign of an inflammatory reaction. The sample of the animal in the soft + shielding plate group that was excluded from the study as a result of an accidental tibia fracture 1 week after surgery (Fig. 3E) was used as a control. Scaffold filling with autologous bone graft was clearly detectable as small bone chips.

Histomorphometrical analysis

Safranin-Orange/von Kossa slides were digitalized with the AxioVision software (Carl Zeiss MicroImaging GmbH). The ROIs were determined with ImageJ (W.S. Rasband, U.S. NIH). The total ROI had a standard length of 40 mm, corresponding to the osteotomy defect size, and was divided into a lateral and medial part through the middle of the scaffold (Fig. 6B, light blue rectangle). The width of this ROI was individually adapted to the thickness of the mineralized callus. The central area of the osteotomy gap (height of 10 mm) was defined as the central ROI (Fig. 6B, dark blue shaded rectangle). Because of the individual size of these three ROIs per animal, the measured mineralized bone tissue, hyaline cartilage tissue, and connective tissue in square millimeters were expressed as a percentage of the total area of the ROI. For normalization, the area of the Ti-mesh scaffold was excluded from the total area for all groups.

SHG imaging

SHG imaging was performed on two PMMA-embedded sections of the mid-sagittal plane in the soft + LCP and stiff + LCP groups to visualize collagen fiber orientation within both scaffold designs. The central part of the Ti-mesh scaffold within the lateral and medial strut network was chosen as an ROI (Fig. 7, E, J, O, and T). Images were recorded using a Leica SP5 II microscope (Leica Mikrosysteme Vertrieb GmbH). The SHG signal was generated by a Spectra Physics Ti:sapphire laser (Mai Tai HP) with a 100-fs pulse width at 80 MHz and a wavelength of 910 nm. The SHG signal was detected in the range of 450 to 460 nm (that is, half excitation wavelength). Images were recorded using a 25× water immersion objective with a numerical aperture of 0.95. Overview images of the ROI were recorded by stitching the image stacks (20- to 28-μm stack depth, 4-μm z-plane spacing). For quantification of collagen fiber orientation, single z planes at a depth of 4 or 8 μm were selected.

Local anisotropy of collagen fibrils was analyzed from the SHG images using the freely accessible ImageJ plug-in FibrilTool (46). FibrilTool was plugged into a macro that performed orientation analysis within multiple sub-ROIs of 360 × 360 μm size, which were in turn situated within a grid-like pattern covering the overview image. The primary fiber orientation was visualized as a green line in each sub-ROI. The length of the line indicates the degree of anisotropy (Fig. 7, F, H, K, M, P, R, U, and W). Information about the tissue composition in the ROIs was gained from the histological thin sections of the segmental defect. FibrilTool was also used to analyze the orientation of collagen fibers relative to the Ti-scaffold surface and inside scaffold channels. Fiber orientation was analyzed in ROIs of 310 μm × 310 μm or 620 μm × 310 μm, depending on the length of the respective surface element. The longer side of the ROI was positioned to match the Ti surface. The deviation of the angle of primary collagen fiber orientation and the orientation of the Ti-scaffold surface element was calculated as ΔΦ = |Φcollagen fibers − ΦTi surface|.

Backscattered scanning electron microscopy

B-SEM imaging was performed on all PMMA-embedded sections of the mid-sagittal plane in the soft + shielding plate and stiff + shielding plate groups using a Tescan Vega3 GMU scanning electron microscope (TESCAN). The mineralized bone structure was analyzed within the critically sized defect in order to investigate the bone regeneration process within the soft and stiff Ti-mesh scaffolds. Overview large-area images and ROIs for higher-magnification studies were selected in the adjacent cortical tibia bone, and within the newly formed bone of the proximal, middle, and distal part of the scaffold in one sample of both groups (Fig. 8, A to I).

Statistical analysis

As normal distribution of data could not be confirmed, only nonparametrical tests were performed. Data from histomorphometrical analysis and strains from FE analysis were presented in box plots showing the median, 25th and 75th percentile, and min-max (whiskers). The radiographic score of the soft + LCP and stiff + LCP test groups was compared with the Fisher’s exact test. Comparison of central ROI minimum bone area and total cartilage area between the soft + LCP and stiff + LCP groups and between the soft + shielding plate and stiff + shielding plate groups was performed with the Mann-Whitney test. For all tests, P < 0.05 was considered significant. All statistical analyses were performed with SPSS (SPSS version 22, SPSS GmbH Software). Individual subject-level data are shown in table S5.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/10/423/eaam8828/DC1

Materials and Methods

Fig. S1. Large segmental femur defect treated with a Ti-mesh scaffold in a female patient.

Fig. S2. Large segmental femur defect treated with a Ti-mesh scaffold in a male patient.

Fig. S3. Surgical procedure in sheep.

Fig. S4. Histological overview images of all tibia samples.

Fig. S5. Endochondral ossification and verification of tissue types.

Fig. S6. Analysis of collagen fiber orientation on the Ti-scaffold surface.

Fig. S7. Quantification of collagen fiber orientation across Ti-scaffold pores.

Fig. S8. Absolute maximum principal strains occurring in an empty defect.

Table S1. Bulk stiffness of both scaffold types used in the sheep study.

Table S2. Stiffness of both plate systems used in the sheep study.

Table S3. Summary of clinical observations in the sheep study.

Table S4. Osteoclast density in segmental defects in sheep.

Table S5. Individual subject-level data for the sheep study (Excel file).

REFERENCES AND NOTES

Acknowledgments: We thank M. Princ, G. Korus, and M. Thiele at the Julius Wolff Institute for their assistance with sample preparation, staining, and evaluation. We acknowledge the support of A. Niederberger and J. Fierlbeck during the study. M. Windolf is the inventor on patent application WO2010017649 A1 submitted by AO Technology AG that covers the design of a dynamic bone plate. Funding: This study was funded by the Berlin-Brandenburg Center for Regenerative Therapies, Charité–Universitätsmedizin Berlin, and DePuy Synthes. Scaffold manufacturing was provided by DePuy Synthes. Author contributions: A.-M.P. wrote animal ethical approval and the manuscript; performed animal care, anesthesia, surgeries, histological analysis and interpretation, radiographic analysis and interpretation, and image preparation; and raised external funds. S.C. supervised the development of in silico models, interpreted and analyzed computer model results, and contributed to manuscript writing. H.R. designed and performed in silico modulation of scaffolds and contributed to manuscript writing. A.P. performed SHG imaging, analyzed collagen fiber orientation in Ti scaffolds, and contributed to manuscript writing. J.C.W. performed SEM imaging, prepared figures, and contributed to manuscript writing. K.S.-B. supported animal study planning and performance, and revised the manuscript. M.W. was responsible for the invention of the half-shell shielding plate, supply of the study with plates and parts of the supportive sling system, and revision of the manuscript. A.Á.T. provided support in animal care, surgeries, anesthesia, performance of radiographs, histological preparation, and data collection. C.P.R. provided support in animal care, surgeries, anesthesia, performance of radiographs, histological preparation, and data collection. K.-D.S. performed patient surgeries; supervised study performance, evaluation, and manuscript writing; and raised external funds. G.N.D. was responsible for the study design and concept; supervised study performance, evaluation, and manuscript writing; raised external funds; and revised the manuscript. P.S. participated in writing the animal ethical approval, performed patient surgeries and radiographic follow-up images, performed animal surgeries, raised external funds, supervised data interpretation, and revised the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the corresponding author.
View Abstract

Navigate This Article