A Bioactive Synthetic Membrane Improves Bone Healing in a Preclinical Nonunion Model

Abstract

Objectives:

High energy long bone fractures with critical bone loss are at risk for nonunion without strategic intervention. We hypothesize that a synthetic membrane implanted at a single stage improves bone healing in a preclinical nonunion model.

Methods

Using standard laboratory techniques, microspheres encapsulating bone morphogenic protein-2 (BMP2) or platelet derived growth factor (PDGF) were designed and coupled to a type 1 collagen sheet. Critical femoral defects were created in rats and stabilized by locked retrograde intramedullary nailing. The negative control group had an empty defect. The induced membrane group (positive control) had a polymethylmethacrylate spacer inserted into the defect for four weeks and replaced with a bare polycaprolactone/beta-tricalcium phosphate (PCL/β-TCP) scaffold at a second stage. For the experimental groups, a bioactive synthetic membrane embedded with BMP2, PDGF or both enveloped a PCL/β-TCP scaffold was implanted in a single stage. Serial radiographs were taken at 1, 4, 8, and 12 weeks postoperatively from the definitive procedure and evaluated by two blinded observers using a previously described scoring system to judge union as primary outcome.

Results

All experimental groups demonstrated better union than the negative control (p=0.01). The groups with BMP2 incorporated into the membrane demonstrated higher average union Mehta scores than the other groups (p=0.01). The induced membrane group performed similarly to the PDGF group. Complete union was only demonstrated in groups with BMP2-eluting membranes.

Conclusions

A synthetic membrane comprised of type 1 collagen embedded with controlled release BMP2 improved union of critical bone defects in a preclinical nonunion model.

Introduction

Clinical management of high energy long bone fractures remains a formidable challenge to the orthopaedic surgeon. Bone loss at the time of injury or after debridement of nonviable fragments occasionally creates a critical defect. These critical bone defects will not heal spontaneously, despite surgical stabilization, without an advanced osteogenic intervention.[1] Even in the absence of segmental bone loss, high energy fractures are often associated with significant periosteal stripping predisposing to nonunion and subsequent bone grafting surgeries. It is estimated that over 400,000 fractures require bone grafting each year in the United States, representing tremendous cost and resource allocation.[2] Complex surgeries, such as the induced membrane technique, can treat critical bone defects but necessitate multiple operations with significant morbidity to the patient and require specialized surgical expertise. [3]

Despite modern advances in tissue engineering, a structural implant for single-stage segmental bone regeneration has not translated into clinical practice. We previously established a technique to construct macroporous three-dimensional printed osteoconductive scaffolds solely comprised of Food and Drug Administration (FDA) approved materials. [4] Although implantation of these scaffolds into critical bone defects demonstrated significant bone ingrowth, it did not outperform the two-stage induced membrane technique. Therefore, we developed a synthetic membrane to envelop the scaffold and provide osteoinductive growth factors locally, similar to the induced membrane. [5–8] Microspheres (MSs) facilitate controlled release of growth factors and can be incorporated into a collagen membrane.[9, 10] In combination, an acellular synthetic implant for bone regeneration becomes feasible.

We hypothesized that a synthetic membrane comprised of a collagen membrane and osteoinductive MSs would improve radiographic union of segmental bone loss compared to an induced membrane. To evaluate this hypothesis, an in vitro study first determined the elution properties of collagen embedded MSs encapsulating bone morphogenic protein-2 (BMP2) and platelet derived growth factor (PDGF). We chose BMP2 and PDGF because of their osteoinductive properties and current use in clinical practice. Then, an in vivo study was performed in a preclinical nonunion model designed to evaluate tissue engineered implants for healing critical bone loss without autograft. [4]

Methods

In vitro study

Microspheres of gelatin/PGDF and poly (lactic-co-glycolic acid) (PLGA)/BMP2 were created using the following protocol. For the gelatin MSs, 200 mg of gelatin was dissolved in 20 ml of deionized distilled water. The pH of the resulting solution was adjusted to 7.00 with 0.2 M sodium hydroxide. The MSs were formed by gradual displacement of the water with ethanol under controlled stirring conditions. The prepared gelatin MSs were crosslinked by glyoxal for 10 h. The unreacted aldehyde groups of glyoxal were quenched with sodium metabisulfite aqueous solution. The MSs were separated using centrifugation at 12,000g for 90 min, washed twice, and lyophilized. For PLGA MSs, a mixture of 50 mg PLGA (PLGA chain extended with poly (ethylene glycol)-b-carboxyl acid) dissolved in 8 ml DMSO was loaded in a dialysis tube and dialyzed against DI water for 24 h, with a change of water every 2–4 h. After self-assembly, the MSs suspension was freeze-dried to obtain a free-flowing powder. The PLGA MSs were then activated in MES buffer supplemented with N-Ethyl-N’-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC) and N-Hydroxysulfosuccinimide sodium salt (NHS) for 1 h at room temperature. 10 mg of the activated MSs were dialyzed against DI water at 4 °C overnight, added to the BMP2 solution (1500 ng/ml) or PDGF (300ng/ml) respectively, and allowed to react overnight at 4 °C. The growth factor grafted MSs were centrifuged and lyophilized. The lyophilized MSs were imaged using scanning electron microscopy. (Figure 1) The MSs were sterilized by E-beam radiation at a dosage of 25 kGy. Eight synthetic collagen membranes were created for each experimental group by combining 6.6 mL Dulbecco’s Modified Eagle Medium, 3.3 mL type 1 rat tail collagen, and 78 μL of 1 M sodium hydroxide. Depending on the experimental group, either 10 mg of BMP2 MSs or 15 mg of PLGA MSs or both was then added to the solution. This solution was incubated at 37 °C and sonicated to allow MSs to dissolve within the solution. 1.2 mL of solution was added to each 8mm by 45mm by 4mm rectangular Teflon™ mold. Each mold was incubated for 45 minutes at 37 °C to allow for solution gelation. Finally, each rectangular membrane was lightly compressed between two sheets of filter paper for 3 minutes to obtain the final impregnated synthetic membrane ready for surgical implantation.

Figure 1:

Electron microscopy demonstrates morphology of hydrogel microspheres with and without growth factor.

Porous cylindrical scaffolds composed of 80% polycaprolactone (PLC) and 20% tricalcium phosphate (TCP) were created with a custom three-dimensional printer. Scaffolds were designed to be 8mm in length, 5.5 mm in diameter and hollow with a 2 mm inner diameter to facilitate insertion over an intramedullary rod.

To determine in vitro elution properties, 10 mg BMP2 MSs and PGDF MSs with and without coupling to the collagen membrane were incubated in 1.0 mL PBS solution and shaken at 100rpm at 37°C. At predetermined time points (day 1, 3, 5, 7, 9, 14, and 21), 0.5 mL of PBS solution containing released growth factors was collected, and the amount of released growth factors was determined by an enzyme-linked immunosorbent assay.

In vivo study

Administrative Panel on Laboratory Animal Care approval was obtained prior to commencement of the in vivo study and standard analgesia and husbandry protocols were followed. Eight 10-week old Sprague-Dawley rats were used for each treatment group, which was a sufficient number to establish significant differences between groups in preliminary study establishing our preclinical model.[4] There were five groups for a total of 40 animals:

• Negative control: Empty 8mm femoral defect

• Induced membrane: 4 weeks of polymethylmethacrylate (PMMA) spacer followed by replacement by a bare PCL/TCP scaffold

• BMP2: PCL/TCP scaffold enveloped by a BMP2 MSc / collagen type 1 membrane

• PDGF: PCL/TCP scaffold enveloped by a PDGF MSc/ collagen type 1 membrane

• BMP2/PDGF: PCL/TCP scaffold enveloped by BMP2 and PDFG MSs / collagen type 1 membrane

For the surgical procedure, a lateral parapatellar approach to the femur was utilized. Full-thickness muscle flaps continuous with a lateral arthrotomy were created to expose the articular surface of the knee and lateral aspect of the femur. The intercondylar notch was opened and the femoral canal was prepared for insertion of a retrograde intramedullary nail (RISystem AG, Switzerland) using a manual opening drill and reamer, respectively. An 8mm diaphyseal osteotomy was performed with an electric rotating saw to create a critical sized defect. This defect size was chosen because it was above the published critical defect threshold for a diaphyseal femoral rat model and it was technically advantageous for our construct. [11] In addition, saline lavage was not utilized during the osteotomy to devitalize the bone ends through thermal necrosis, a previously cited mechanism generating atrophic nonunion.[12] The nail coupled to an extramedullary jig was inserted into the distal segment. The appropriate implant was threaded on the nail. For experimental groups, the synthetic membrane was manually wrapped around the scaffold prior to insertion. Then the nail was inserted into the proximal segment and the external jig guided manual drilling and insertion of three interlocking bolts (two distal in the metaphysis and one proximal in the diaphyseal bone). Restoration of length, alignment and rotation were confirmed visually, then the external jig was removed. The wound was irrigated and closed in layers.

The induced membrane group involved two stages. [4] In the first stage, two longitudinal halves of hollow PMMA cylinders were placed around the nail in the bone defect and approximated with a nonabsorbable cerclage stitch. Four weeks later the surgical site was reopened and the resultant induced membrane was incised longitudinally along the axis of the limb. After the circumferential suture was cut and the PMMA was removed, a bare PCL/TCP incised along one side of its longitudinal axis was clam-shelled open and placed around the nail in the cavity formed by the induce membrane. The membrane and wound were closed in layers.

All animals were euthanized at 12 weeks after the definitive surgery for analysis.

Serial radiographs were performed in all groups at 1, 4, 8, and 12 weeks postoperatively. Radiographic review was performed by two blinded orthopaedic surgeons using a modified Mehta score as described previously (see appendix): [4, 13]

• 0 points: pseudarthrosis (or implant failure/shortening on the nail)

• 1 point: no bridging, some callus (outline of the scaffold is noticeable on radiograph)

• 2 points: incomplete bridging

• 3 points: complete bridging

The sample with the highest radiographic healing score per group was analyzed using micro-CT-analysis. Specimens were imaged at a custom isotropic resolution of 20 μm isometric voxel size with a voltage of 55 kV and a current of 200 μA. Volumetric reconstruction was performed about the region of interest (ROI), defined as the 8-mm defect. Cross-sectional slices of the defect were used for bone tissue volume fraction (Bone volume/Total volume, BV/TV) measurement, excluding the density threshold > 250 (metallic implant). 3D bone structure was made from the segmented dataset for visual inspection. For histology, specimens were fixed in 10 % formalin for 48 hrs, then transferred to 70% ethanol. The specimens were decalcified in 10 % EDTA solution for 5 weeks. Intramedullary nails were removed after decalcification. Then the specimens were embedded in paraffin after dehydration with ethanol. Thin sections (5 μm) were cut by a microtome along the long axis of each femur in the sagittal plane. The slides were stained with hematoxylin and eosin (H&E; Sigma-Aldrich, St Louis, MA) or Masson Trichrome staining (Abcam, Cambridge, MA) following standard protocol.

Statistical analysis

Statistical analysis was performed using Student’s T-tests to determine differences in elution rates for BMP2 and PDGF MSs. For the in vivo study, the primary outcome was the final average radiographic score at 12 weeks. Analysis of variance was used to determine if there was overall difference in radiographic healing between groups, and a post-hoc Tukey test was used to identify which groups performed better. Significance was set at alpha <0.05.

Results

In vitro

The in vitro elution assay showed the release of PDGF in gelatin MSs (61.3%, t=21 days) to be faster than the release of BMP2 in PLGA MSs (42.5%, t=21 days, p= 0.007), leading to a controlled biphasic release of PDGF followed by BMP2. When embedded in the collagen membrane, the release assay demonstrated continued consistent release of growth factor in both MSs but at a 24.3% and 21.2% slower rate for PDGF and BMP2, respectively, as represented in Figure 2 (p<.01).

Figure 2:

In vitro growth factor release assay after electron beam sterilization and type 1 collagen membrane embedment.

In vivo

A total of 39/40 animals survived the duration of the study period without significant complication (one rat in the BMP2/PDGF group was euthanized secondary to a deep wound infection). There was a significant difference in final radiographic healing between groups (Figure 3, p<0.01). None of the animals in the negative control group healed (avg 0.1±0.2 at 12 weeks). All experimental groups had higher average union Mehta scores than the negative control (p<0.01). Healing between the PDGF (avg 1.1±0.3 at 12 weeks) and two-stage induced membrane (avg 0.5±0.2 at 12 weeks) groups was not different (p= 0.12). The BMP2 alone (avg 2.0±0.3 at 12 weeks) and BMP2/PDGF (average 2.0±0.8 at 12 weeks) groups demonstrated superior healing compared to other groups (p<0.01). There was no difference in final healing for groups that contained BMP2 only or BMP2 and PDGF (p=0.9). A summary of statistical comparisons for the in vivo study is listed in Table 1. Three rats in the groups that contained BMP2 demonstrated complete radiographic healing by 12 weeks (Figure 4). None of the blinded Mehta scores between the two observers were discordant by more than one point.

Figure 3).

Radiographic healing of critical defects demonstrates significant differences between groups at 12 weeks (p<0.01).

Table 1:

Summary of group comparisons of radiographic Mehta scores at final healing.

Groups	Significance	Improved Healing
Negative vs Induced Membrane	p=0.34
Negative vs BMP2	p=0.01	BMP2
Negative vs PDGF	p=0.01	PDGF
Negative vs BMP2/PDGF	p=0.01	BMP2/PDGF
Induced Membrane vs BMP2	p=0.01	BMP2
Induced Membrane vs PDGF	p=0.12
Induced Membrane vs BMP2/PDGF	p=0.01	BMP2/PDGF
BMP2 vs PDGF	p=0.01	BMP2
BMP2 vs BMP2/PDGF	p=0.89
PDGF vs BMP2/PDGF	p=0.01	BMP2/PDGF

Figure 4).

Complete radiographic healing of a critical defect shown at 12 weeks with BMP2 microspheres embedded into type 1 collagen membrane.

Micro-CT and histological analysis of the best specimens from each group are shown in Figure 5 and Figure 6. Results from micro-CT showed that very small amount of bone could be found in the defect sites in blank control group and PDGF group. Higher bone volume could be found in the BMP-2 group and the BMP-2/PDGF group. Results from histology showed that only fibrous or fat tissue could be found in the bone defect sites in the negative control group. New regenerated bone tissue could be found in other groups, with highest amount in BMP-2 group and BMP-2/PDGF group.

Figure 5).

Micro-CT analysis of best specimens in each group at 12 weeks of healing.

Figure 6).

Histological samples of best specimens in each group at 12 weeks of healing.

Discussion

High energy diaphyseal fractures with significant periosteal stripping or bone defects predispose to nonunion unless surgical adjuncts improve biologic healing capacity. This study compared an engineered membrane to the induced membrane technique described by Masquelet in a preclinical nonunion model.[6] In vivo radiographic assessment supports the hypothesis that a type 1 collagen membrane eluting osteoinductive MSs improves union of segmental defects compared to an induced membrane. Synthetic membranes embedded with BMP2 outperformed PDGF, which demonstrated negligible healing benefits in the presence of both factors. Specimens with PDGF MSs healed comparably to those with induced membranes. Both of these groups, however, did not demonstrate meaningful bridging callus at final analysis. In addition, we did not appreciate a difference in healing potential when PDGF was added to implants with BMP2 compared to BMP2 alone.

Complete healing of critical bone defects occurred in some specimens with controlled release of BMP2. The in vitro results suggest BMP2 elutes locally over the duration of a few weeks when encapsulated in hydrogel MSs. This mimics the secretion of various growth factors by induced membranes, including BMP2, which peaks at four weeks.[14] Other studies have similarly shown in situ release of BMP2 improves segmental bone healing using a variety of delivery methods.[15–21] Compared to the induced membrane group, BMP2-laden membranes improved final healing by 400%. Although this treatment effect may not fully translate into clinical practice, it represents a substantial opportunity for effective single-stage bone repair without the patient morbidity and economic burden associated with multiple procedures.

Microspheres of PDGF were not as effective as BMP2 MSs to stimulate segmental bone regeneration. These findings mirror conclusions from another study where BMP2 and not PDGF delivered in a fibrin matrix supported bone healing in a delayed-union rat model. [22] In the presence of skeletal stem cells, PDGF may improve healing, but this may be most effective for intramembranous bone regeneration of flat bones. [23, 24]

Our preclinical nonunion model was originally designed to evaluate structural bone implants spanning a critical defect.[4] In that study, a PCL/TCP scaffold implanted into the defect without a membrane did not demonstrate meaningful callus formation. This models the clinical scenario where devitalized bone fragments span a fracture segment with limited capacity to heal. In surgical management of open fractures, bone devoid of soft tissue attachments or periosteum is typically debrided and discarded. Potentially these structural fragments could be retained if overlaid with synthetic membrane that improves biologic healing potential.

Mechanical stability is critical when managing diaphyseal fractures with significant bone loss. The Masquelet technique utilizes PMMA cement, which is strong in compression and can be molded around an intramedullary nail to span the defect. In a rat model, this strategy is also effective in maintaining mechanical stability.[25] As such, the PCL/TCP scaffold utilized in this study is hollowed for sliding over a locked intramedullary nail, improving compressive and angular stability of the overall construct. Anecdotally, there were vastly fewer mechanical failures in the experimental groups compared to the group with an empty defect.

This study has several limitations. We assessed healing as our primary outcome through blinded review of serial radiographs conducted by two orthopaedic surgeons. When tabulated, the Mehta scores did not differ by more than one point, so interobserver reliability was not reported. Also, we did not perform more advanced analyses to comprehensively evaluate healing of the specimens. Although future investigations may incorporate these tests, we emphasized radiographic evaluation, which more closely resembles serial evaluation of fracture healing in clinical practice. Lastly, our methods do not address potential infection associated with high energy open fractures. Incorporating antimicrobial properties into the synthetic membrane may be beneficial.

In summary, an engineered membrane comprised of type 1 collagen and BMP2 eluting MSs improves bone healing in a preclinical nonunion model. This advancement may restore regenerative potential to nonviable bone fragments. Coupled to an osteoconductive scaffold, the synthetic membrane has capacity to heal critical bone defects in a single stage. This implant warrants preclinical investigation in larger models to determine safety and efficacy before clinical translation.