Amniotic membrane attenuates heterotopic ossification following high dose bone morphogenetic protein-2 treatment of segmental bone defects

Abstract

Treatment of large bone defects with supraphysiological doses of bone morphogenetic protein-2 (BMP-2) has been associated with complications including heterotopic ossification (HO), inflammation, and pain, presumably due to poor spatiotemporal control of BMP-2. We have previously recapitulated extensive HO in our rat femoral segmental defect model by treatment with high dose BMP-2 (30 μg). Using this model and BMP-2 dose, our objective was to evaluate the utility of a clinically available human amniotic membrane (AM) around the defect space for guided bone regeneration and reduction of HO. We hypothesized that AM surrounding collagen sponge would attenuate heterotopic ossification compared to collagen sponge alone. In vitro, amniotic membrane retained more BMP-2 than a synthetic poly(ε-caprolactone) (PCL) membrane through 21 days. In vivo, as hypothesized, the collagen + AM resulted in significantly less heterotopic ossification and correspondingly, lower total bone volume, compared to collagen sponge alone. Although bone formation within the defect was delayed with AM around the defect, by 12 weeks, defect bone volumes were equivalent. Torsional stiffness was significantly reduced with AM but was equivalent to that of intact bone. Collagen + AM resulted in the formation of dense fibrous tissue and mineralized tissue, while the collagen group contained primarily mineralized tissue surrounded by marrow-like structures. Especially in conjunction with high doses of growth factor delivered via collagen sponge, these findings suggest amniotic membrane may be effective as an overlay adjacent to bone healing sites to spatially direct bone regeneration and minimize heterotopic ossification.

Graphical Abstract

Use of bone morphogenetic proteins (BMPs) at supraphysiological concentrations can lead to complications including heterotopic ossification and inflammation. Therefore, investigation of biomaterials capable of localizing the growth factor to mitigate these adverse effects is crucial. The objective of this study was to evaluate the efficacy of amniotic membrane (AM) in attenuating heterotopic ossification with high dose BMP-2 treatment of rat segmental bone defects. As hypothesized, heterotopic ossification surrounding bone defects was reduced with collagen sponge + AM compared to collagen sponge alone.

Introduction

Traumatic bone injuries, and particularly critically sized defects, necessitate surgical intervention for repair, often involving augmentation with supraphysiological amounts of BMP-2. Although BMP-2 delivered on an absorbable collagen sponge has shown success in long bone healing and spinal fusion,^1,2 the use of large doses of BMP-2 has caused concern due to complications including heterotopic ossification, inflammation, and tissue swelling.³ The consequences of supraphysiological doses of BMPs on bone healing and inflammation have been observed not only clinically but also in preclinical animal models. The inflammatory response in rats following BMP-2 and BMP-7 delivered subcutaneously or intramuscularly in collagen sponge consisted of: (i) greater volumes of soft tissue edema and granuloma-like masses at both implantation sites, and (ii) larger areas of inflammation surrounding the intramuscular implants, as BMP dose was increased (1–20 μg).⁴ Regarding orthotopic delivery of BMP-2, a canine critically sized radial defect model resulted in regenerated bone with cyst-like voids and impaired mechanical properties with higher BMP-2 doses (up to 2.4 mg) compared to bone formed with the lowest dose of BMP-2 (150 μg).⁵ Furthermore, Zara et al. determined a minimum threshold dose of BMP-2—11.25 μg in a critically sized rat bone defect model—at and above which poor bone quality and heterotopic ossification were observed.⁶ Likewise in a similar rat bone defect model, 10 μg BMP-2 resulted in improved bone healing compared to higher and lower doses.⁷

Collagen-based biomaterials are one of the most common classes of natural biomaterials used in tissue engineering applications. In the clinic, absorbable collagen sponge (ACS) scaffolds of bovine origin are approved for use in spinal fusions, open tibial fractures, and non-unions as carriers of rhBMP-2 and rhBMP-7.^5,8 One of the benefits of collagen scaffolds is their ability to be enzymatically degraded; additionally, the degradation can be prolonged by material modification, such as physical or chemical crosslinking of the scaffold.^9,10 Nonetheless, a major limitation of growth factor delivery via collagen sponge is the rapid release of cargo, which has been extensively characterized in vitro, in vivo, and clinically, with 40–90% of the loaded growth factor released in an initial burst pattern.^3,5,11–13 Thus, the delivery of high dose BMP-2 within collagen sponge may exacerbate adverse effects such as heterotopic ossification and inflammation. In our recent study in which an 8-mm critically sized rat femoral defect model was treated with high dose BMP-2 (30 μg) in alginate hydrogel or collagen sponge, each surrounded by a macroporous poly(ε-caprolactone) (PCL) mesh, early exposure to high dose BMP-2 was sufficient to trigger equivalent heterotopic ossification, despite differences in in vitro BMP-2 release kinetics and amount of bone within the defect region.¹² In contrast, historical data from this rat model demonstrated that a low dose (2–5 μg) of BMP-2 in these same alginate hydrogel and collagen sponge biomaterials successfully restored bone function with little to no mineral formed outside the defect space.^14–16 Similarly to our findings, 5-mm rat femoral defects treated with 50 μg of BMP-2 in collagen sponge exhibited a large degree of heterotopic ossification compared to defects administered 5 μg, while bony bridging of the defects was not affected by dose.¹⁷ The consistent observations of bridging with low doses and adverse events with high doses notwithstanding, both ACS and high dose BMP-2 continue to be commonly used for growth factor delivery in the clinic, thus motivating the development of strategies to limit excessive BMP exposure outside of bone defects.

Due to the efficacy of BMP-2 for bone regeneration, investigations into biomaterial strategies designed to reduce the incidence of adverse events such as heterotopic ossification are ongoing. While controlled release strategies constitute the bulk of this approach,^18,19 we have demonstrated the utility of the macroporous PCL mesh around the rat segmental bone defect space treated with alginate hydrogel + low dose BMP-2 for spatially guiding and enhancing bone regeneration compared to BMP-2 delivered via hydrogel alone¹⁴ or within a nonperforated membrane²⁰. More recently, we have incorporated heparin methacrylamide microparticles, which strongly bind BMP-2, into the PCL mesh to reduce the diffusion of BMP-2 away from the delivery site. However, no reduction in heterotopic ossification was observed with this delivery system, attributable to the interference of serum-borne, heparin-binding proteins in vivo.²¹ In a similar approach, human amniotic membrane may serve as a natural alternative to synthetic materials such as PCL. The amniotic membrane contains large amounts of collagens type I and IV and hyaluronan,^22–24 as well as many growth factors, inflammatory mediators,^25,26 angiogenic cytokines,²⁷ and proteases and their inhibitors²⁸. With intrinsic structural properties (proteins, glycosaminoglycans (GAGs), adhesive ligands, etc.), immunomodulatory properties, and ability to bind growth factors, the amniotic membrane provides an environment beneficial for resident and recruited cells^26,29 and may function effectively as a sink for exogenous BMP-2.

Amniotic membrane has been used successfully in regenerative applications for skin,^30,31 cornea,^32,33 and ligament³⁴ in the clinic, as well as cartilage in preclinical animal models^35,36. A recent 80-patient, prospective, randomized controlled trial in which patients received either cryopreserved amniotic tissue or no tissue in the annular defect following lumbar microdiscectomy surgery revealed less pain and lower incidence of herniation in the group treated with amniotic tissue.³⁷ Additionally, amniotic membrane has functioned as an effective barrier membrane in the repair of orthopedic tissues,³⁸ minimizing soft tissue adhesions in tendon repair in a chicken model^39,40 and in spinal fusion procedures in dogs⁴¹ and humans⁴². In a case study of five patients who underwent transforaminal lumbar interbody fusion (TLIF) followed by implantation of dehydrated human amnion/chorion membrane (dHACM) in the epidural space, revision surgery post-fusion demonstrated 80% (4/5) patients had easily detachable tissue in the epidural space, suggesting the amniotic membrane minimized epidural fibrosis.⁴² More recently, a retrospective review of 14 patients who had received human amniotic membrane to prevent retethering subsequent to an intradural spinal surgery found retethering (evaluated at 6 months or later) in only one patient.⁴³ Thus, amniotic membrane, as a natural, bioactive ECM material with structural integrity, may provide necessary spatial control over bone formation when high doses of BMP-2 are used. Nonetheless, additional preclinical and clinical evidence is needed regarding the barrier functionality of amniotic membrane around biomaterials such as collagen sponge loaded with high doses of growth factor.

Since heterotopic ossification was observed with a high dose of BMP-2 in our rat segmental bone defect model,¹² the objectives of this work were to investigate the ability of amniotic membrane to attenuate both BMP-2 release in vitro and heterotopic ossification in critically sized bone defects treated with high dose BMP-2 delivered in collagen sponge. We hypothesized that amniotic membrane wrapped around the bone ends and the collagen sponge containing BMP-2 (30 µg) positioned within this segmental defect would result in less heterotopic ossification compared to collagen sponge alone. Notably, one major advantage of the materials used in this study—collagen sponge and amniotic membrane—is their availability for use clinically, which provides the opportunity to accelerate the implementation of amniotic membrane for guided bone regeneration.

Materials and Methods

rhBMP-2 binding on and release from membranes

As the release kinetics of BMP-2 from collagen sponge are well-characterized,^3,5,11–13 the purpose of this experiment was to explore the functionality of poly(ε-caprolactone) (PCL) mesh and amniotic membrane alone to bind and retain BMP-2. To evaluate BMP-2 binding and release in vitro, 8-mm diameter disk-shaped samples (n=6) of dehydrated human amnion/chorion membrane (dHACM) allograft (AmnioFix^®, MiMedx Group, Inc., Marietta, GA) and PCL nanofiber mesh fabricated as described previously²⁰ were prepared using a biopsy punch. The PCL mesh, used previously in this segmental defect model to contain alginate hydrogels,^{12,14–16,20} has demonstrated retention of up to 35% of loaded BMP-2 at 26 days in vitro, most of which could be removed by vigorous washing, and only 5% of the total was bound non-specifically to the polymer mesh.¹⁶ Here, the PCL mesh was used a negative control to determine non-specific binding.

For BMP-2 binding, AM and PCL disks were incubated in 1 mL PBS containing 0.1% (w/v) bovine serum albumin (BSA, Millipore Corporation) and 350 ng recombinant human BMP-2 (rhBMP-2, Pfizer, Inc., New York, NY) for 18 h at 37°C in an ultra-low attachment 24-well plate (Corning, Corning, NY). Construct-free wells served as controls. For release assays, samples were transferred to new wells containing fresh 0.1% BSA in PBS, and the BSA/PBS solution was collected and replaced at 0.25, 1, 2, 4, 7, 14, and 21 days. Initial BMP-2 remaining in solution and BMP-2 released from constructs over time were measured via an enzyme-linked immunosorbent assay (ELISA, R&D Systems, Minneapolis, MN). BMP-2 remaining in solution in construct-containing wells was subtracted from BMP-2 remaining in solution in construct-free (control) wells to determine the amount of BMP-2 initially retained by the sample disks.⁴⁴

Surgical procedure

Unilateral mid-diaphyseal 8-mm (critically sized) femoral segmental defects, internally stabilized by radiolucent polysulfone plates, were created in 13-week-old female SASCO Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA) as described previously.⁴⁵ The defects were treated with 30 μg recombinant human BMP-2 (rhBMP-2, Pfizer Inc.) in bovine type I collagen sponge (Kensey Nash Corp., West Whiteland Township, PA) alone as described earlier¹² or collagen sponge surrounded by amniotic membrane (AmnioFix^®, MiMedx Group, Inc.) (n=11). Sample size was chosen based on historical data generated using this model.^{11,14–16,45,46} Collagen sponge alone (no BMP-2) was not included in this study because prior data from this rat model (8-mm segmental defect) treated with 0.1–1 µg BMP-2 in collagen sponge,^11,47 and from a sub-critically sized 2.5-mm segmental defect treated with blank collagen sponge,⁴⁸ demonstrated incomplete bone healing. 150 µL of 0.1% (w/v) rat serum albumin (RSA, Sigma-Aldrich, St. Louis, MO) in 4 mM HCl containing 30 μg of rhBMP-2 was added drop wise to the collagen sponge and allowed to adsorb for 10 minutes before implantation. For the amniotic membrane group, the tissue was cut into 1.5 × 2.0 cm pieces and wrapped around the hydrated collagen sponge with the amnion epithelial layer side facing outwards. The collagen sponge was press-fit into the bone defect between the cut ends. Amniotic membrane constructs were similarly press-fit into the bone defect, taking care to ensure that the amniotic membrane wrapped around the cut bone ends.

For analgesia, slow-release buprenorphine (ZooPharm, Windsor, CO) was administered prior to surgery. Animals were euthanized by CO₂ inhalation at 12 weeks post-surgery. All procedures were approved by the Georgia Institute of Technology Institutional Animal Care and Use Committee (IACUC). No criteria were set for exclusion of animals, and no animals were excluded from analyses. No blinding to group allocation was performed. Group assignment was alternated throughout the surgical procedure, and animals were housed two per cage with random assignment based on group.

Radiography and micro-computed tomography

Bone regeneration was assessed via longitudinal radiography and micro-computed tomography (micro-CT) through 12 weeks. To quantify bone volume (BV) using micro-CT, two volumes of interest (VOIs) were used: a 6-mm diameter to characterize mineralization inside and bordering the defect (defect BV), and a large diameter including all bone formation within the thigh (total BV). Heterotopic bone volume was defined as the difference between these two VOIs.^12,21

Biomechanical testing

Ex vivo torsional testing of femora (n=8) at 12 weeks provided a functional measure of the regenerated bone.⁴⁵ After removal of the fixation plate and soft tissue, native bone ends were potted in Wood’s metal (Alfa Aesar, Haverhill, MA). Torsional testing to failure was performed at 3° per second (ELF 3200; Bose ElectroForce Systems Group, Framingham, MA). Maximum torque to failure and torsional stiffness were calculated from the resultant torque-rotation curves.

Histology

Histology was used to further characterize the distribution and maturation of newly formed bone within and surrounding the bone defect space (n=3). Samples were harvested at 12 weeks, fixed in 10% neutral buffered formalin for 48 h, and decalcified in a mild formic acid solution (Immunocal, Decal Chemical Corp., Suffern, NY), which was changed three times a week for two weeks. After paraffin embedding, samples were sectioned by a tape transfer technique¹² (Section Labs, Hiroshima, Japan) to obtain 5 µm mid-sagittal sections. Sections were stained with hematoxylin and eosin (H&E) to visualize the global cellular response, residual collagen sponge and amniotic membrane, and spatial extent of mineralization. Safranin O/Fast Green staining allowed for identification of chondrogenic tissue, while Mallory’s modified aniline blue stain was used to distinguish mature and immature osteoid as described previously.¹⁵

Statistical Analyses

All data are reported as mean ± standard error of the mean (SEM). Data were analyzed using GraphPad Prism 5 (GraphPad Software, Inc., San Diego, CA). Cumulative BMP-2 release and bone volumes from micro-CT were analyzed by two-way repeated measures analysis of variance (ANOVA), with Bonferroni post hoc tests for pairwise comparisons. Biomechanical testing data were evaluated by t-tests. A p-value less than 0.05 was considered statistically significant.

Results

rhBMP-2 binding on and release from membranes

The objective of the in vitro binding and release experiment was to evaluate the efficacy of amniotic membrane alone (i.e., without collagen sponge), compared to the polymeric barrier membrane (PCL) typically used in our rat bone defect model, to bind and retain BMP-2, prior to the use of amniotic membrane as a barrier membrane in vivo. After 18 hours of incubation in 350 ng BMP-2 solution, similar amounts of BMP-2 were bound to both PCL mesh (79.7±22.2 ng) and amniotic membrane (95.0±19.9 ng) disk-shaped samples (Fig. 1A). The BMP-2 associated with the disk samples was considered as the bound or membrane-retained fraction, irrespective of any differences in membrane hydration (of AM) or nano-porosity (of PCL). As a percentage of initially bound BMP-2, more BMP-2 was released over 21 days from PCL mesh than from amniotic membrane (p<0.001, Fig. 1B).

Figure 1.

(A) PCL mesh and amniotic membrane (AM) bound similar amounts of BMP-2 after 18 hours. (B) Significantly more of this bound BMP-2 was released over 21 days from PCL mesh compared to amniotic membrane (*p<0.001, n=6).

Radiography and micro-computed tomography

Representative radiographs of defects treated with high dose BMP-2 in collagen sponge alone (top) or with amniotic membrane (bottom) at 2, 4, 8, and 12 weeks post-operatively allowed for qualitative comparison of new bone formation (Fig. 2). The collagen group showed both defect and heterotopic ossification as early as 2 weeks, while the collagen + AM group showed comparatively less ossification. The total amount of ossification, especially the heterotopic ossification (located outside the bone defect region), appeared attenuated in the amniotic membrane group.

Figure 2.

Representative radiographs of defects treated with high dose BMP-2 in collagen sponge (top) or collagen sponge surrounded by amniotic membrane (AM, bottom) at 2, 4, 8, and 12 weeks post-operatively. Total ossification—and in particular, heterotopic ossification—appeared to be attenuated in the collagen + AM group. Arrows indicate heterotopic ossification.

Micro-CT analyses provided spatially distinct measurements of bone volume and distribution. Bone density maps of a sample each of the collagen and collagen + AM group generated from ex vivo micro-CT data (Fig. 3A) showed sparse trabecular bone structure in the defect region and an extensive heterotopic ossification shell in the collagen sponge group. In contrast, the majority of ossification in the collagen + AM group was located within the defect region as seen from the cross-section images. Total bone volume (BV), inclusive of both the bone within the defect as well as the surrounding soft tissue, increased over time and was significantly higher in the collagen group at 4, 8, and 12 weeks compared to collagen + AM (p<0.05, Fig. 3B). Bone volume within the defect region was also significantly greater in the collagen group at 4 and 8 weeks (p<0.05, Fig. 3C). Although early mineralization in the defect region was delayed in the collagen + AM group, it recovered by 12 weeks, as the defect BVs were equivalent between groups at 12 weeks. As hypothesized, heterotopic BV was significantly reduced at 8 and 12 weeks in the presence of amniotic membrane (p<0.05, Fig. 3D).

Figure 3.

(A) Representative bone density maps of total BV. Left: entire volume of interest (VOI) analyzed, Right: sagittal cross-section through VOI. Dashed lines delineate the defect space. A heterotopic ossification shell and an inner trabecular structure were visible in the collagen group, while minimal heterotopic ossification was present in collagen + AM samples. (B) Total bone volume (BV) increased over time and was significantly higher in the collagen group at 4, 8, and 12 weeks. (C) Although defect BV was delayed with collagen + AM, no differences between the treatment groups were observed by 12 weeks. (D) Heterotopic BV was significantly reduced at 8 and 12 weeks in the presence of amniotic membrane (*p<0.05, n=11).

Biomechanical testing

Biomechanical properties of the regenerated bone at 12 weeks were used as a measure of the functional recovery of the mineralized tissue. Maximum torque to failure was similar between groups (Fig. 4A). However, torsional stiffness was significantly attenuated in the collagen + AM group (Fig. 4B, p<0.05), likely due to the minimal heterotopic ossification present. Nonetheless, stiffness for collagen + AM was similar to that of intact bone.²⁰

Figure 4.

Functional assessment of the regenerated bone at 12 weeks. For reference, dashed lines indicate mean values for historical intact control bone.²⁰ (A) Maximum torque to failure was similar between groups. (B) However, torsional stiffness was significantly attenuated in collagen + AM, likely due to the minimal heterotopic ossification present (*p<0.05, n = 8). Nonetheless, stiffness for collagen + AM was similar to that of intact bone.

Histology

Mid-sagittal histological sections demonstrate regenerating bone tissue at 12 weeks. In the center of defects treated with high dose BMP-2 in collagen sponge alone, the lamellar structure (Fig. 5A-C) and orange-red staining of the bony spicules (Fig. 5C) were indicative of highly mineralized tissue, while the collagen sponge appeared to have been resorbed. This bone tissue had a highly trabecular-like structure and was surrounded by abundant loose connective tissue and marrow-like tissue. In contrast, collagen + AM treated defects contained heterogeneous cell infiltrate throughout the dense, fibrous tissue (Fig. 5D-F). Residual amniotic membrane tissue was also observed. Cells stained dark red with Safranin O, and light blue in Mallory’s aniline blue, were likely chondrocytes (Fig. 5E), suggesting endochondral ossification may have occurred. Despite differences in morphology between groups, dense red staining in the collagen + AM defect suggests the presence of mineralized tissue (Fig. 5F).

Figure 5.

Regenerated tissue in the center of bone defects at 12 weeks, as indicated by the red box in the density map (bottom). (A, D) H&E, (B, E) Safranin O, and (C, F) Mallory’s staining. (A-C) In defects treated with high dose BMP-2 in collagen sponge without amniotic membrane, the collagen sponge appeared entirely resorbed. The resultant lamellar structure (arrows) and orange-red staining (C) of the bony spicules was indicative of highly mineralized trabecular tissue (T). Marrow-like tissue (M) was observed between trabeculae. (D-F) Collagen + AM treated defects appeared as much denser tissue, with fibrous (#) and bone (*) tissue interspersed, along with residual AM (arrows) and numerous cell nuclei adjacent (D). (E) Cells stained dark red with Safranin O were likely chondrocytes, suggesting endochondral ossification may have occurred. (F) Red staining with Mallory’s suggests the presence of highly mineralized tissue. Scale bar=50 μm.

At 12 weeks, tissue outside of the defect region in the collagen group (Fig. 6A-B) and tissue bordering to the defect region in the collagen + AM group (Fig. 6C-D) had a similar appearance to tissue within the bone defect area of each treatment group (Fig. 5), respectively. Heterotopic tissue in the collagen group was highly mineralized, as indicated by the organized, lamellar structure (Fig. 6A-B). In contrast, collagen + AM treated limbs contained little to no heterotopic ossification, so tissue at the defect edge is shown (Fig. 6C-D). Residual amniotic membrane was dispersed throughout the defect edges. Robust cell infiltration was observed in some areas adjacent to AM, while other residual AM particulate appeared more isolated and devoid of cell nuclei.

Figure 6.

Tissue outside (A-B) and bordering (C-D) the bone defect space at 12 weeks, as indicated by the red box and white asterisk, respectively, in the density map (right). (A, C) H&E and (B, D) Safranin O staining. (A-B) Similarly to mineral within the defect space, heterotopic tissue in the collagen group was highly mineralized, as indicated by the lamellar architecture (arrows) of the trabeculae (T), and surrounded by marrow-like infiltrate (M). (C-D) In contrast, minimal heterotopic ossification was observed in collagen + AM treated limbs, so tissue at the defect edge is shown. Residual amniotic membrane (all arrows) was observed throughout the defect border. The density of cell nuclei adjacent to AM was varied, with some AM surrounded by many cells (white arrows), while other AM areas appeared more isolated, lacking infiltrating cells (black arrows). Scale bar=50 μm.

Discussion

The use of BMPs at supraphysiological concentrations can lead to complications including heterotopic ossification and inflammation. Therefore, the investigation of biomaterials capable of localizing the growth factor to the site of injury and thus mitigating associated adverse effects is crucial. Native ECM materials such as amniotic membrane represent a class of naturally derived biomaterials already employed in the clinic for tissue healing applications. Effectively harnessing BMP-2 to promote a more localized pattern of bone formation, particularly via a clinically available material such as AmnioFix^® amniotic membrane (AM), could improve the current treatment paradigm for large bone defects. Following upon recent work in which negative effects of high dose BMP-2 (e.g., heterotopic ossification, tissue swelling) were recapitulated in this rat segmental bone defect model,^6,12 the objective of this study was to evaluate the efficacy of amniotic membrane in attenuating heterotopic ossification with high dose BMP-2 treatment.

As hypothesized, heterotopic ossification surrounding bone defects was reduced with collagen sponge + AM compared to collagen sponge alone. In vitro, though PCL mesh and amniotic membrane bound similar amounts of BMP-2, the percentage (and actual amount, data not shown) of BMP-2 detected in the subsequent release assay was lower with amniotic membrane. Amniotic membrane, which contains GAGs such as hyaluronic acid and other growth factor binding molecules,^{25–28,49,50} may indeed have sequestered the BMP-2 that otherwise could diffuse into the surrounding soft tissue and initiate heterotopic ossification. Further, the processing technique used for the amniotic membrane in this study is known to maintain its bioactivity,²⁶ so the presence of bioactive, immunomodulatory factors may have also contributed to limit heterotopic ossification.

As hypothesized, the amniotic membrane wrapped around the bone defect region reduced heterotopic ossification in the surrounding soft tissue. This reduction in heterotopic ossification was also associated with lower total bone volume present in this group compared to collagen alone. Moreover, a delay in bone formation within the defect center was seen with use of amniotic membrane. A similar observation of attenuated early mineralization was noted previously using this same segmental defect model, when defects were treated with BMP-2 in alginate hydrogel surrounded by a PCL mesh lacking macro-perforations, compared to PCL mesh with macro-perforations.²⁰ Further, in a critically sized sheep tibial defect model, membrane perforations were necessary for successful bone regeneration.^51,52 The surrounding tissue provides an excellent source of osteoprogenitor cells, growth factors, and vasculature, which each play crucial roles throughout the bone healing process.⁵³ Notably, in a rat composite injury model in which volumetric muscle loss in the adjacent quadriceps was performed in addition to the femoral defect, a 2 μg dose of BMP-2 in alginate hydrogel surrounded by a perforated PCL mesh—which was sufficient to consistently bridge the bone defect without muscle injury^11,16,54—was insufficient to heal bone defects with concomitant muscle injury.⁵⁴ During heterotopic ossification, connective tissues are presumably replaced by bone tissue through a complex process involving inflammation, muscle cell death, fibrous tissue proliferation, angiogenesis, and ossification (often endochondral).⁵⁵ Expectedly, local tissue environmental factors (e.g., source of BMP stimulus, ability to recruit osteoprogenitor cells and induce their differentiation into bone tissue) play a major role in the extent to which heterotopic ossification occurs.⁵⁶ For example, osteogenic genes were upregulated in wounds with heterotopic mineralization compared to wounds without.⁵⁷ Despite progress in understanding the role of BMP signaling pathways in heterotopic mineralization, the cells that contribute to the pathology remain under investigation. Due to the proximity of muscle progenitors/myoblasts to bone and their ability to differentiate down the osteogenic lineage when exposed to BMP-2,⁵⁸ these cells may contribute directly to heterotopic bone formation. Muscle-derived progenitor cells from wounds that subsequently developed heterotopic bone were present in greater numbers and exhibited increased osteogenic differentiation compared to cells from wounds that did not form heterotopic bone.⁵⁹ However, Lounev and colleagues, using two mouse models involving intramuscular BMP-2 injection or muscle injury via cardiotoxin injection, found that muscle satellite cells comprised a very minimal portion of the heterotopic bone, while endothelial precursor cells contributed significantly to all three stages of heterotopic ossification (fibroproliferative, chondrogenic, and osteogenic).⁶⁰ Furthermore, their findings suggest both a dysregulation of BMP signaling and an inflammatory environment are necessary for heterotopic ossification to occur. The amniotic membrane composed of dense connective tissue layers including collagens and proteoglycans^22–24 could be a barrier to cell invasion⁶¹ due to an effective pore size on the order of nanometers. The membrane used in the present study consists of the amnion and chorion layers and is ~100 μm thick,²⁵ portions of which remained even at 12 weeks as noted in histology. The slow degradation of amniotic membrane and its dense matrix may have thus hindered early cell infiltration into the defect space and contributed to the delayed bone formation in the defect. However, by 12 weeks, the defect bone volumes were equivalent between groups. If time points beyond the 12-week endpoint of this study would have been assessed, it is thought that defect mineralization may have continually progressed. In the future, the use of perforated amniotic membrane may alleviate the early attenuation in defect mineralization by promoting cell infiltration into the defect space while maintaining the growth factor sequestration and other biological functions of the devitalized amniotic membrane.

Despite the early delay in mineralized healing within the defect space with amniotic membrane treatment, torsional testing at 12 weeks indicated biomechanical properties of regenerated bone from the collagen + AM group were comparable to those of intact femora. Although torsional stiffness was lower in the collagen + AM group compared to collagen alone, it was equivalent to the stiffness of intact bone. In contrast, the average torsional stiffness for the collagen group was approximately double that of intact bone. These results are similar to prior findings with this model and the same high dose of BMP-2.¹² Specifically, both alginate hydrogel and collagen sponge loaded with BMP-2, surrounded by PCL mesh, resulted in torsional stiffness values double that of intact bone, and both had a significant amount of heterotopic ossification present. Based on the equation for torsional stiffness:

$$k_t=\frac{G\hspace{0.17em}\text{x}\hspace{0.17em}{J}_{avg}}{L}$$

where G is the elastic shear modulus, J_avg is the average polar moment of inertia (pMOI), and L is the gauge length, an increase in average pMOI resulting from a wider distribution of mineralization about the long axis of the femur (i.e., heterotopic ossification) would cause a proportionate increase in torsional stiffness. Here, heterotopic ossification was believed to be the primary contributor to the increased torsional stiffness observed in the collagen group.

Histology at 12 weeks revealed a large amount of heterogeneous cell infiltrate in the collagen + AM group, while the collagen group had mainly trabecular-like mineralized tissue present, as was observed recently with this model and dose of BMP-2.¹² It is important to note this study evaluated human amniotic membrane (xenograft tissue) in immunocompetent rats. Few studies have examined the host response to ECM-derived biological materials. Allogeneic and xenogeneic ECM materials, even after decellularization, are capable of eliciting an immune response due to the presence of ECM proteins, which have been shown to stimulate the migration of neutrophils and macrophages.^62–64 In this rat model, the amniotic membrane may have elicited a heightened cellular response, although our conclusions are limited because histology prior to 12 weeks was not performed, and the sample size for histological analyses was limited. The devitalization process that the tissue undergoes typically results in a portion of the cellular structures persisting, and the immunogenicity of this cellular material, particularly in an immunocompetent rat, remains unknown.⁶⁵ We observed moderate inflammatory infiltrate mostly in the area of residual AM tissue, and this response may have exacerbated the delay in bone healing. In contrast, historically, this hypercellularity has not been observed adjacent to polymeric nanofiber mesh (surrounding alginate or collagen sponge) with this model at 12 weeks with high dose BMP-2¹² or low dose BMP-2^11,14. Further exploration is needed to determine specific phenotypes of cells that invaded in response to amniotic membrane implantation. Nonetheless, a similar hypercellular response to human micronized AM was observed in the joint space in a rat osteoarthritis model, persisting throughout the 21-day study, and was associated with attenuated cartilage degradation compared to saline controls.³⁶ In our case, the delivery of amniotic membrane surrounding the femoral defects could be expected to influence the inflammatory response beyond the effects seen in an immunoprivileged location such as the joint space. Recently, a multilayered human amniotic membrane served as an effective barrier to reduce fibrous tissue invasion, while guiding and enhancing bone ingrowth and maturation in a rat tibial defect model.⁶⁶ Similarly, cross-linked human amniotic membrane limited scar tissue formation in the epidural space in a canine laminectomy model.⁴¹ In both of these studies, little to no inflammatory reaction was observed. Importantly, the clinical use of these allogeneic grafts may result in attenuated inflammatory/immune responses compared to that observed in the current rat model.

In summary, amniotic membrane retained significantly more BMP-2 in vitro than PCL mesh, and amniotic membrane surrounding collagen sponge loaded with high dose BMP-2 reduced heterotopic ossification surrounding segmental bone defects. A major advantage of the materials used in this study is their commercial availability for use clinically, which would dramatically accelerate the translation of amniotic membrane-based strategies for bone regenerative applications. Although heterotopic ossification was minimized, likely as a result of both the structural and biological properties of amniotic membrane, further studies are needed to elucidate mechanisms behind the initial delay in mineralization, and to optimize spatiotemporal control strategies for bone regeneration while limiting heterotopic ossification. Rather than completely enveloping the collagen sponge, amniotic membrane may be more effective in guiding bone regeneration as an overlay, spatially limiting the diffusion of (high doses of) BMP-2 from collagen sponge, while slowing bone healing.