Altering spacer material affects bone regeneration in the Masquelet Technique in a Rat Femoral Defect

Abstract

The Masquelet technique depends on pre-development of a foreign-body membrane to support bone regeneration with grafts over three times larger than the traditional maximum. To date, the procedure has always used spacers made of bone cement, which is the polymer polymethyl methacrylate (PMMA), to induce the foreign-body membrane. This study sought to compare (i) morphology, factor expression, and cellularity in membranes formed by PMMA, titanium, and polyvinyl alcohol sponge (PVA) spacers in the Masquelet milieu and (ii) subsequent bone regeneration in the same groups. Ten-week-old, male Sprague-Dawley rats were given an externally stabilized, 6mm femur defect, and a pre-made spacer of PMMA, titanium, or PVA was implanted. All animals were given 4 weeks to form a membrane, and those receiving an isograft were given 10 weeks post-implantation to union. All samples were scanned with microCT to measure phase 1 and phase 2 bone formation. Membrane samples were processed for histology to measure membrane morphology, cellularity, and expression of the factors BMP2, TGFβ, VEGF, and IL6. PMMA and titanium spacers created almost identical membranes and phase 1 bone. PVA spacers were uniformly infiltrated with tissue and cells and did not form a distinct membrane. There were no quantitative differences in phase 2 bone formation. However, PMMA induced membranes supported functional union in 6 of 7 samples while a majority of titanium and PVA groups failed to achieve the same. Spacer material can alter the membrane enough to disrupt phase 2 bone formation. The membrane’s role in bone regeneration is likely more than just as a physical barrier.

Introduction

Until the past century, segmental bone loss was difficult to treat short of amputation.^1–3 Today, distraction osteogenesis, traditional grafting, vascularized fibular transplant, and membrane directed grafting (Masquelet technique) have revolutionized our ability to reconstruct defects too large to heal spontaneously.^1,3–5 However, tens of thousands of patients here and abroad still undergo amputations each year partially because traditional bone grafts have a maximal threshold of 4 to 6 cm.⁶ Distraction osteogenesis works in larger scenarios but is precluded by soft tissue damage or lack of access to trained surgeons.^5,7,8 Even for ideal candidates, the overall time to consolidation and frame removal is lengthy, requires frequent clinical visits, and is accompanied by frame-associated morbidity risks.^2,9,10 Vascularized fibular transplants are highly successful, but require access to highly trained micro-vascular surgeons, has significant donor site morbidity, and has a high refracture rate.³ Thus, it has not gained the popularity of the distraction osteogenesis.

The newer Masquelet or induced membrane technique has potential for patients ineligible for traditional options and is less complex than bone transport.^3,6,7,11,12 At the first surgery, the defect site is stabilized with external or internal fixation, and a bone cement (polymethyl methacrylate, PMMA) spacer is placed where bone regeneration is desired. Soft tissue closure is then carried out so a foreign-body membrane can form around the cement spacer. At the second surgery, the ‘induced’ membrane created in response to the spacer is carefully transected to allow spacer removal. Then the resulting “pouch” is filled with bone graft material. Clinical evidence has shown that it is feasible to delay the second surgery for weeks to months^6,7,12–14 and to use internal fixation as described.^12,13,15,16 Delaying surgery allows time for patients with soft tissue damage to heal ^3,6,7,14 or those undergoing chemotherapy to complete treatment^12,17,18 thus, hopefully, avoiding amputation. It has become the gold standard for treating pediatric tumor removal.¹² Furthermore, internal fixation eliminates frame-associated comorbidities thus providing a simpler treatment regimen for all segmental defect patients. Also, Masquelet treatment uses common orthopedic surgical procedures,^6,7,14,19 allowing its use by surgeons with limited experience or resources.

Current information has led to many hypotheses as to what makes the Masquelet technique successful. It is well established that the membrane or morselized graft alone are not sufficient to regenerate bone; both must be intact.^20–23 Also, the induced membrane formed in response to a PMMA spacer forms over weeks. Vascularity, cellularity, and regenerative factors such as BMP2, VEGF, IL6, and TGF-β peak around 4–8 weeks after spacer implantation.^23–27 Although the induced membrane is a fibrous capsule formed as part of the foreign-body response cascade, it has many similarities to the periosteum, bone’s progenitor-cell rich bounding membrane.^26–28 These findings have led orthopaedic surgeons and researchers to speculate that the three important functions of the membrane are (i) to act as a barrier to protect the graft from soft tissue invasion, (ii) to provide a large vascular network that brings in nutrients and oxygen, and (iii) to supply localized, regenerative biochemical signals and cells.^{6,7,11,12,14,19,29}

While investigation of the Masquelet technique is in its infancy, foreign-body membranes, on the other hand, have been studied for decades. The membranes form in response to any large, non-biodegradable, foreign object.^20,21,30,31 This could occur by accident like a splinter or bullet or as part of a therapeutic intervention like a pacemaker, glucose sensor, or hip implant. In all of the latter scenarios the foreign-body membrane is generally detrimental to the implant’s intended purpose. Thus many efforts have been made to control foreign-body membrane thickness and vascularity in order to preserve functionality as long as possible.^24,26,32,33 In general, an implant’s material, surface topography, surface hydrophobicity, porosity, dimensions and location can all greatly affect foreign-body membrane formation.^25,26,32

For these studies we chose to focus on material effects and wanted to choose a few materials well known for their divergent foreign-body membranes not necessarily clinical feasibility. If divergent bone regeneration occurred, then, in the future, the membrane environments could be contrasted to determine what characteristics are critical to the procedure’s success. Polymethyl methacrylate (PMMA), a plastic also known as bone cement, is the current standard. Plastics are known to create relatively thick membranes.³⁰ Metals, especially titanium (TI),³⁴ typically form thinner membranes.^35,36 Some evidence suggests that TI also enhances osteoblast differentiation and the expression of osteogenic factors.^34,37 For these reasons, TI is typically the material of choice for orthopaedic implants like plates, intramedullary rods, and screws meant to remain for long time periods (i.e. years). The thinner interstitial tissue allows for better osseointegration thus increased mechanical stability between the bone and implant.³⁸ More hydrophilic surfaces are known to enhance the osteogenic environment and promote osteogenesis.^34,39 Polyvinyl alcohol sponge (PVA) is highly hydrophilic and has been used for various types of implants.⁴⁰ However, PVA typically failed due to excessive membrane formation and undesired mineralization, which could be advantageous for the Masquelet procedure.^40,41

This study sought to compare (i) morphology, factor expression, and cellularity in membranes formed by PMMA, titanium, and PVA spacers in the Masquelet milieu and (ii) subsequent bone regeneration in the same groups. We hypothesized that TI and PVA should create membranes more favorable for phase 2 bone regeneration than the current standard, PMMA.

Methods.

Animal Model.

Ten-week-old, male Sprague Dawley rats (Charles River, Wilmington, MA) were used for all experiments. A rat model was selected because rodents are common pre-clinical models for human skeletal regeneration studies.⁴² The basic biology of repair is conserved. Thus they can be used as a screening tool to prioritize experimental questions or variables before scaling up to more clinically relevant large animal models like rabbits, goats or sheep, which require greatly increased financial and technical resources. Also, rats are currently the only published rodent model for the Masquelet Technique.^{23,26,27,39,43–45} All procedures were approved by our Institutional Animal Care and Use Committee (IACUC) (protocol #2451) and conducted in accordance with national guidelines. Gradual carbon dioxide asphyxiation was used for euthanasia according to American Veterinary Medical Association guidelines.

All experimental animals underwent the initial surgery (Phase 1, n = 9/group, N=54 rats) and were monitored bi-weekly via x-ray to assure implant fixation. Significant k-wire loosening or spacer migration was cause for exclusion (N=5, detailed in results). Briefly, a custom external fixator was implanted into the right femora using fully threaded K-wires to attach it to the bone. Then a 6mm defect was created at approximately the bone mid-point (Figure 1A). A pre-fabricated spacer was placed in the defect (Figure 1B). Animals were randomly assigned to one of the three spacer groups: PMMA, TI, or PVA.

Figure 1. Phase 1 Surgery.

(A) All experimental animals underwent an initial surgery to implant the external fixator and spacer. (B) The three different types of spacers: PMMA, TI, PVA (top to bottom).

To study bone regeneration, some animals underwent the second surgery (Phase 2, n=9/group, N=27) 28 days following the first. Four weeks was chosen for this study since many characteristics peak around this time³⁴ and it is the current clinical recommendation.^34,46 During this surgery, the spacer was removed, and the membranous chamber between the two bone ends was filled with a morselized bone isograft. Morselized bone grafts were harvested from unoperated cohort animals (1 donor animal for 3–4 experimental animals). The graft was harvested and morselized just prior to beginning the first of the surgeries, and then kept in a covered container with additional moisture. Implantation order was randomized between groups to mitigate any effects of decline in graft viability. The graft volume was approximately 48mm³ but was altered to avoid overfilling if needed (i.e. situations with significant bone overgrowth). For additional details about spacer fabrication and the surgical procedures, please see supplemental data.

Phase 1 Bone Formation.

During our pilot studies we noted significant bone growth over the spacer ends in line with a previous rodent study,²⁷ so the amount of bone formation during the first surgical phase was determined using microCT. Four weeks after external fixator and spacer implantation, animals for these studies were euthanized and the operated femur was harvested (n=7–9/group). Samples were fixed in 10% neutral buffered formalin (VWR, Radnor, PA) and the external fixator and Kirschner wires were carefully removed. The bone between the inner pins was scanned with microCT (MicroCT 35, ScanCo Medical, Brüttisellen, Switzerland; X-ray tube potential 70 kVp, integration time 300 ms, X-ray intensity 145 µA, isotropic voxel size 10 um, frame averaging 1, projections 1000, high resolution scan). Any bone extending from the original bone ends to cover the spacer was contoured. Because the spacers were still in place, the titanium resulted in a level of beam hardening in all samples resulting in some unavoidable artifact that would skew or obstruct measurement of finer bone architecture and mineral density. Thus extension length and total volume for the proximal and distal protrusions were the only variables quantified.

Membrane Morphology & Cellularity.

After the samples were microCT scanned, they were processed for histology. Briefly, the limb was decalcified in formic acid and cut longitudinally in half following the original surgical approach to remove the spacer. One half was embedded in OCT media and cryosectioned longitudinally. All measurements were taken from the defect mid-point on the medial side to avoid any possible effects from the surgical approach. To measure membrane thickness and collagen alignment one section per animal was stained with weigert’s hematoxylin, picrosirius red, and alcian blue to differentiate cell nuclei, fibrillar proteins, and cartilage, respectively, as well as enhance collagen birefringence (n=7–8/group, ). Both brightfield and polarized light images were used to measure membrane thickness (ImageJ, NIH, Bethesda, MD); measurements were split into two distinct regions: non-birefringent (NB) and birefringent (BR), which were not distinguishable from each other under brightfield imaging. The midpoint polarized/compensated images were also processed using a custom MATLAB program (MathWorks, Natick, MA) to determine relative fiber alignment within the birefringent membrane region.

To measure matrix protein composition, collagen type 1 (n=5/group) and elastin (n=7–8/group) in each region was assayed with immunohistochemistry (IHC) (method details in supplemental data). Briefly, after antigen retrieval, the sections were incubated with blocking solutions containing normal goat serum (10% v/v). Then the experimental sections were incubated over night at 4C with a primary antibody (1:50 dilution in blocking serum, Table 1, all rabbit polyclonal, Abcam, Cambridge, MA). The negative control slide was incubated under identical conditions with another round of blocking serum. Then all slides were incubated with the secondary antibody (1:100 dilution in 1X PBS, Goat polyclonal Secondary Antibody to Rabbit IgG - H&L w/ Alexa Fluor 488, Abcam) for 60min at room temperature. All sections were mounted with DAPI containing media (Fluoro-Gel II with DAPI, Electron Microscopy Sciences, Hatfield, PA) and imaged at the medial defect mid-point at 20X (DMB 4000B, Leica Microsystems, Wetzlar, Germany). To control for non-specific staining and tissue auto-fluorescence, the average fluorescent intensity in each region (non-birefringent and birefringent) for each animal’s negative control was subtracted from the corresponding region in each experimental section. To measure membrane cellularity the segmented images of the IHC negative control slide from each animal was used (n=7–8/group). The blue channel (DAPI) of each region was processed in a custom MATLAB program to measure the number of cell nuclei per unit area.

Table 1.

Protein Targets for IHC

Protein Name	Protein Abbreviation	Abcam Catalogue Number
Collagen Type 1	Col1	ab34710
Elastin	Elastin	ab21610
Bone Morphogenetic Protein 2	BMP2	ab14933
Transforming Growth Factor Beta	TGFβ	ab66043
Vascular Endothelial Growth Factor	VEGF	ab46154
Interleukin 6	IL6	ab6672

Membrane Factor Expression.

To quantify factor expression in the membrane, cryosections were processed for IHC for BMP2, TGFβ, VEGF, IL6 (n=7–8/group, Table 1). These factors were selected because they have been used in many previous Masquelet studies.^{14,20,25,26,30,39,47} The methods used were identical to those used to detect collagen type 1 and elastin.

Phase 2 Bone Formation.

Animals that underwent the second surgery to remove the spacer and implant the isograft were given 10 weeks to regenerate bone in the defect (n=7–8/group). After euthanasia, both femora were harvested keeping the overlying muscle and soft tissue intact but removing the external fixator from the operated limb. Since the original intention was to perform mechanical testing, the limbs were wrapped in saline soaked gauze and stored at −20C until further analysis. First the operated limbs were thawed overnight at 4C and 1–2 hours at room temperature. Under blinded conditions all bones were screened by gentle manipulation to determine if basic mechanical stability was achieved, which would indicate functional union and allow for meaningful mechanical testing. Then the operated limbs were scanned with microCT using the same scanning parameters used for phase 1 bone formation. Scans encompassed the length between the two inner most pins. The defect volume of interest was defined as either the bone length between the original cortical ends or, when there was no distinct border between new trabecular regenerate and original cortical bone, a 6mm length centered at the apparent defect location (threshold: 260 per milles, determined from pilot testing). Composites of each bone’s three dimensional reconstruction and final x-ray (i.e. microCT scoutview) (Figure 2A) were used to blindly qualify each sample’s bridging as either united, unclear (obvious void) or empty (Figure 2B).

Figure 2. Example Phase 2 MicroCT.

(A) Final 10 week x-ray and composite of limb’s three-dimensional reconstruction which was used to categorize defect union. White box on left panel indicates region scanned with microCT. White box on right panels indicates contoured VOI on the microCT scan. (B) Examples of the three union classifications.

Statistics.

All data is presented as mean ± standard deviation. A priori power analyses suggested that a minimum of 6 animals per group were sufficient to detect our desired differences (α=0.05., β = 0.80, G*Power, Heinrich Heine Universitat, Dusseldorf,Germany). Originally 9 animals were randomly assigned to each experimental group so that with attrition more than enough animals would remain to detect significant differences. Phase 1 bone formation outcomes were compared between groups with ANOVA (factor: spacer material – PMMA vs TI vs PVA) with repeated measures (factor: location - proximal vs distal). Membrane morphology and factor expression results were compared with ANOVA (factor: spacer material - PMMA vs TI) with repeated measures when applicable (factor: location – non-birefringent layer vs birefringent layer). All microCT outcomes were compared between groups with ANOVA (factor: spacer material – PMMA vs TI vs PVA). Because the outcomes for basic mechanical stability and bone union were nominal, the outcomes were analyzed with non-parametric Kruscal-Wallis testing followed by pair-wise Mann-Whitney comparisons for post-hoc analysis. The significance threshold was set at p ≤ 0.05.

Results.

Animals.

All rats tolerated the procedures well and consistently gained weight until euthanasia. A total of 5 animals were excluded from the final analysis due to hardware failure (i.e. loose pins causing bone lysis) (1 - Phase 1 PMMA, 2 - Phase 2 PMMA, 1 - Phase 2 TI, 1 – Phase 2 PVA). Two additional animals were excluded from the IHC studies due to sample mishandling which precluded analysis (1- PMMA, 1 – TI).

Phase 1 Bone Formation.

After 4 weeks of spacer implantation, all PMMA and TI induced membranes had bone extending from the original bone ends to partially cover the spacer surface (Figure 3A). Comparable volumes originated from the proximal and distal ends that reached similar maximal lengths (Figures 3B-C). There were no differences between the two materials. PVA, on the other hand, did not permit the same phase 1 bone growth.

Figure 3. Phase 1 Bone Results.

(A) After 4 weeks of spacer implantation, all PMMA and TI induced membranes had bone extending from the original bone ends to partially cover the spacer surface (white arrows). PVA did not permit the same phase 1 bone growth. (B-C) in the PMMA and TI groups, the length and volume of the new bone within the defect region was similar between groups and originating location (proximal vs distal).

Membrane Morphology & Cellularity.

All PMMA and TI spacers created similar membranes while PVA had a very different response. At the spacer midpoint between the spacer surface and muscle, all PMMA and TI membranes had two regions that were only distinct under polarized imaging. The inner, NB layer was significantly thinner than the outer, well aligned BR layer in both groups (p<0.001, Figures 4 and 5A-B ). In a few samples (n=1–2/PMMA and TI groups) there was a third, disorganized region between the BR layer and muscle. However, since it was not prevalent in most samples it was not analyzed further. Collagen type 1 distribution was similar in both regions, but elastin expression was 25–29% higher in the NB layer than the BR layer (Figures 5C-D). Structures resembling blood vessels on the Collagen type 1 stained IHC sections were not noted in either the NB or BR layer. They were often seen at the outer border of the BR region, but not extending into it. There were no differences in the cell nuclei density between the two regions(Figure 5E) . In membrane regions between the spacer surface and bone overgrowth, there were differences in membrane thickness between PMMA and TI (Figure 6). The BR region of TI membranes was closer in thickness to that of the NB region resulting in an overall significantly thinner membrane. PVA, on the other hand, did not create a similarly structured membrane (Figure 4). There was often little to no tissue between the spacer material and muscle. Furthermore, fibrous tissues and cells had infiltrated uniformly through the construct. Since there was not a clear membrane region, further membrane analysis was not performed for this group.

Figure 4. Midpoint Membrane Development.

At the spacer’s medial midpoint between the spacer surface and muscle, all PMMA and TI membranes had two regions that were not distinct under brightfield imaging. It became clear that there was a non-birefringent (NB) layer directly apposed to the spacer that was not visible under polarized light. The second, birefringent (BR) layer was illuminated by polarized light and had well aligned fibers. PVA did not create a similarly structured membrane. There was often little to no tissue between the spacer material and muscle. Furthermore, fibrous tissues and cells had infiltrated uniformly through the construct.

Figure 5. Midpoint Membrane Structure & Cellularity.

(A) The NB layer directly apposed to the spacer was significantly thinner for both PMMA and TI, but the over all thickness was similar for both groups. (B) The fibers of the BR layer were well aligned in both groups. (C) Although the NB layer was much thinner, the number of cell nuclei per unit area was similar in the two regions and between material groups. (D-E) IHC showed that Collagen type 1 was similar between layers and groups while elastin was more highly expressed in the NB layer.

Figure 6. Membrane Development at Bone Extensions.

(A) In membrane regions between the spacer surface and bone overgrowth (outlined by white dashed lines), there were differences in membrane thickness between PMMA and TI. (B) The BR region of TI membranes was closer in thickness to that of the NB region resulting in an overall significantly thinner membrane.

Membrane Factor Expression.

For all examined factors (BMP2, TGFβ, VEGF, and IL6) there were no significant differences between groups (Figure 7). However, all proteins had significantly elevated expression in the NB region. For PMMA induced membranes the expression of all factors in the NB layer was approximately double that of the BR layer. For TI induced membranes, the differences between the two regions were less (30–85% increase in NB).

Figure 7. IHC Results.

(A) Example IHC of VEGF to highlight the two regions. (B) For all examined regenerative factors (BMP2, TGFβ, VEGF, and IL6) there were no significant differences between groups. However, all proteins had significantly elevated expression in the NB region.

Phase 2 Bone Formation.

Ten-weeks post-engraftment, bone regeneration within the defect was evaluated. Basic mechanical stability was achieved in almost all PMMA animals (6 of 7). However, a majority of the TI and PVA animals did not pass this initial screening (6 of 8 and 7 of 8, respectively, p=0.07 Kruskall-Wallis) indicating that meaningful information could not be gained from mechanical testing. Of the quantitative microCT outcomes, there were trends for increased BV/TV, Tb.N, and BMD as well as decreased Tb.S in PMMA animals, especially in comparison to PVA animals (Table 2). However, these failed to reach statistical significance. Qualitative categorization determined significantly more PMMA samples to be united than TI and PVA samples (n = 4 of 7, 2 of 8, and 1 of 7, respectively, Figure 8, p=0.038 Kruskal-Wallis). All groups had 2 samples where bone filled the defect but there was either a clear continuous void between trabecular bone areas or the bone was globular and disconnected (Figure 8 TI). The remaining samples had little to no bone within the defect (Figure 8 PVA).

Table 2.

Phase 2 Bone Regeneration. There were no differences between groups for any microCT outcome.

Group	n	TV (mm3)	BV (mm3)	BV/TV	Tb.N (1/mm)	Tb.Th (mm)	Tb.Sp (mm)	Tb.Conn (1/mm3)	BMD (mgHA/cm3)	TMD (mgHA/cm3)
PMMA	7	122.9 ± 49.5	50.0 ± 23.4	0.39 ± 0.05	3.4 ± 0.7	0.20 ± 0.03	0.39 ± 0.07	82.6 ± 56.9	484 ± 62	1076 ± 22
TI	8	149.3 ± 45.3	56.4 ± 19.9	0.37 ± 0.06	2.7 ± 0.7	0.18 ± 0.04	0.54 ± 0.17	114.0 ± 71.1	446 ± 76	1076 ± 23
PVA	8	137.9 ± 67.7	43.5 ± 24.4	0.31 ± 0.09	2.6 ± 1.2	0.18 ± 0.04	0.55 ± 0.21	92.6 ± 69.8	389 ± 110	1065 ± 38

^ap<0.05 for spacer material (PMMA vs TI vs PVA)

TV = Total Volume, BV = Bone Volume, Tb.N = Trabecular Number, Tb.Th = Trabecular Thickness, Tb.Conn = Connection density,

BMD = Bone Mineral Density, TMD = Tissue Mineral Density

Figure 8. Phase 2 Bone MicroCT.

Representative samples of each spacer group. PMMA spacers resulted in more consistent bone repair while TI or PVA spacers typically did not bridge.

Discussion.

Many factors can affect the body’s response to implants and foreign-body membrane encapsulation such as location, porosity, and surface topology.^25,26,32 One of the easiest to control is material. To our knowledge this is the first study to report phase 1 bone volume, to quantify membrane collagen alignment, to evaluate factor expression in the two distinct membrane regions, and to demonstrate that spacer material can affect subsequent bone formation even though the spacer itself has been removed.

The two solid spacer materials, PMMA and TI, formed similar membranes after 4 weeks of implantation, while PVA spacers did not form a distinct membrane. PMMA and TI induced membranes were composed of two layers: one that did not illuminate under polarized light (NB) and a well-aligned one that did illuminate (BR). The NB layer was significantly thinner than BR layer, but they had equal cellular density and collagen type 1 expression. The NB layer was more enriched than the BR layer for elastin in addition to the osteogenic, angiogenic, and inflammatory factors measured (BMP2/TGFβ, VEGF, and IL6 respectively). PMMA and TI membranes also provoked bone growth from the proximal and distal ends to partially cover the spacer while PVA did not.

Previous reports of PMMA induced membrane thickness has varied widely – anywhere from 50–100um,⁴⁷ which is close to this study’s observations, to well over 1000um.^26,39 Since striking differences were not found between the two solid spacer groups in this study or between bone cement brands in another study,³⁹ it is unlikely that PMMA lot or brand differences can account for the study to study discrepancies. Fixation is a more likely contributing factor. Previous Masquelet studies have utilized casting,³⁷ external fixation,^18,30,47 internal plates,^39,48 and intramedullary rods.^14,26,39,49 All internal fixation devices will cause an independent foreign-body reaction which could confound results if measurements are taken too close to the affected region (i.e. in the immediate proximity of the fixation device). Furthermore, each type of fixation can result in a different mechanical environment, which can have profound effects on bone repair.^36,44,50 For example, external fixation stability is highly dependent on the distance between the bone and fixator as well as the diameter of the connecting pins. Longer or thinner connecting pins provide less stability than shorter or thicker ones. Also, plates shift the force distribution from roughly symmetrical about a bone’s cross section to highly asymmetrical. However, there is currently no data as to whether mechanical environment would affect Masquelet membrane formation or bone repair.

Few other studies have commented on the membrane’s multi-layered nature^26,39 although it can be observed from the available histology in many papers.^{43–45,47,49,51} Those that have differentiated the two to three regions often reported the inner layer to be thin and composed of cuboidal cells while the outer layer was more fibrous and the cells more elongated.^26,39 Also, vessels were not observed directly apposed to the spacer.²⁶ While we did not analyze cell shape, our results are mostly in line with previous reports. We found the outer, BR layer to be more fibrous. This was especially evident under polarized light where the inner, NB region failed to illuminate. Although, collagen type 1 was similar between regions. So there are likely other fibrous matrix proteins that make up a majority of the BR region. Also similar to previous studies,³⁹ vessel-like structures were only observed at the periphery of the BR region. In contrast, the number of cell nuclei per unit area was not significantly different, which suggests that the regions have similar cellularity. Although, foreign-body membranes are predominately created by macrophages that have fused together to become multi-nucleated foreign-body giant cells^31,36,52 and contain osteoclasts²³ which are also multi-nucleated. With our methods each one of these cells would have been counted multiple times, once for each nucleus. So we could have overestimated the relative number of cells. Future work is necessary to accurately segregate the multiple cell lineages and types.

It is difficult to put our factor expression and phase 1 bone formation data into context with previous studies. No other Masquelet studies have compared PMMA to another material in a bone defect. For factor expression, most did not differentiate between membrane regions, and any histological characterizations were qualitative rather than quantitative.²⁶ It is striking how well the expression pattern of all factors, especially VEGF, aligned with the NB region in all of our samples. This suggests that the NB layer, which would be in direct contact with the graft during phase 2, is the major repository for regenerative factors. For phase 1 bone formation, many previous pre-clinical studies have reported bone ‘lips’ or ‘islands’ within the membrane.^45,47 To our knowledge, this is the first quantification of these features. Similar features have not been alluded to in the clinical research literature. This may be due to procedural changes that were necessary for study implementation. For our studies TI and PVA could not be molded in situ to cap the bone ends, which is the clinical recommendation. Thus, all of our spacers, and those of other pre-clinical studies, were pre-fabricated to be approximately flush with the bone surface. Having this contiguous, solid surface may have facilitated periosteal cell expansion along the spacer and the observed bone formation. Capping the bone ends allows the developing membrane to overgrow the defect region and encapsulate the defect as well as the junctional region, which is reported more favorable. ^14,19 This overlap of the membranes has been shown clinically to decrease the rate of nonunion that occurs at the junctional regions when the membrane is not overlapped.

Our results for TI membranes and phase 2 bone formation were surprising and encourage further study. Titanium is considered one of the most ‘biocompatible’ metals,³⁷ and TI generally forms a thinner membrane than plastics and integrates well with bone.^34,37 This is what led us to hypothesize that it would be superior Masquelet spacer material than PMMA despite its clinical infeasibility. However, this was not the case. In line with previous studies the membrane formed between the implant and bone was thinner for TI than PMMA; but the membrane formed between implant and muscle, which is the majority of the induced membrane, was similar in size, structure, and factor expression. It is possible that the cells responding to this region are not differentially affected like osteogenic cells, but more research work would be needed to support this theory.

Despite the seemingly identical membranes that were formed by PMMA and TI, TI induced membranes did not promote graft incorporation and bone regeneration as effectively as PMMA. This suggests that critical parameters not assessed in this study are altered by using a TI spacer. Studies into the barrier properties, vascularity, cellular populations, and wider ranging factors for each material are warranted and could be contrasted to determine essential membrane environment characteristics.

In contrast, our results do not encourage further study of PVA as a spacer material either for clinical use or mechanistic study. PVA was originally chosen because it reportedly formed a very fibrotic membrane and is commercially available in a wide range of porosities, which is another parameter known to affect membrane formation. The PVA sponge selected for this study had relatively small pores (~80um) in an attempt to limit tissue integration.⁴⁹ However, integration still occurred as evidenced by fibrotic tissue and cell infiltration without a distinct membrane. Removing the spacer during phase 2 surgery was somewhat difficult. The spacer circumference easily separated from the muscle, but it was almost impossible to disconnect the ends from the bone. As alluded to in the methods, often small PVA pieces were left attached to the bone ends because they were too interconnected to remove. It’s possible that the failure to union in phase two was due to two mechanisms. First the remnant PVA on the bone ends could have inhibited osteogenic cell migration. Second, any protective tissue that formed along the circumference was likely removed with the spacer. This would essentially leave the graft without a membrane. Both mechanisms would likely be exacerbated with increasing porosity and not yield any informative data. So, we do not feel there is reason to pursue PVA going forward.

There are a few limitations that should be considered when interpreting our results. First, relatively young animals were used. At 10 weeks old the animals are physiologically similar to young adult humans (~18 years old).⁵³ Many processes alter as an organism ages, and this especially true for bone repair.²⁵ It’s unclear if similar results would be obtained using differently aged animals. Although, while the Masquelet technique was originally created for adults, it is currently standard of care for pediatric tumor removal.^12,18 Thus aging is an important factor that should be studied in the future. Second, only one spacer implantation duration was examined. Many studies demonstrate that PMMA induced membranes are dynamic.^24,26,39 Four weeks was chosen for this study since many characteristics peak around this time³⁴ and it is the current clinical recommendation.^34,46 However, spacers are often left in place much longer depending on a patient’s situation.^54,55 Different results could have been obtained if more time points were assayed. Determining if phase 1 duration affects phase 2 bone regeneration is greatly desired by clinicians and should be researched going forward.⁵⁶ Finally, a small animal model was used rather than a large animal model. While length-scale can certainly affect some important healing parameters like cell migration, diffusion, and solute transport, many aspects of biological repair are conserved between species. Many discoveries that went on to greatly impact clinical practice were first observed and studied in small animal models.⁴² Previous pre-clinical Masquelet studies have been almost evenly split between small^{23,26,27,39,43–45} and large^{20–22,24,33,57} animals. There are remarkable similarities between the small/large animal findings and those made in human samples.^25,28,47 For example, many membrane characteristics peaks in the 4–8 week time frame in both humans and rats. So, while any major procedural change or adjuvant therapy should be vetted in a large animal model prior to clinical implementation, rodent use is a good way to screen these procedures and explore basic biology. Thus, given our study focus, we do not feel this is a major limitation to our findings’ utility.

In conclusion, we demonstrated that after 4 weeks of implantation PMMA and TI spacers created bi-layered membranes that encourage bone growth from the original cortex. The inner layer is non-birefringent and more enriched for BMP2, TGFβ, VEGF, and IL6; the outer layer is birefringent and thicker. Despite the similarities, PMMA induced membranes provide an environment more conducive to phase 2 bone regeneration than TI induced membranes. PVA spacers do not form a distinct membrane and do not encourage phase 2 bone regeneration. Together, this suggests that initial host environment has a significant effect on bone regeneration. It is likely that the membrane’s role is more than just as a barrier. If the membrane’s critical factors could be identified, then the technique’s clinical application could be refined and a tissue engineered product could possibly be developed to eliminate the first surgery.