Animal models of typical heterotopic ossification.

Heterotopic ossification (HO) is the formation of marrow-containing bone outside of the normal skeleton. Acquired HO following traumatic events is a common and costly clinical complication. In contrast, hereditary HO is rarer, progressive, and life-threatening. Substantial effort has been directed towards understanding the mechanisms underlying HO and finding efficient treatments. However, one crucial limiting factor has been the lack of relevant animal models. This article reviews the major currently available animal models, summarizes some of the insights gained from these studies, and discusses the potential future challenges and directions in HO research.

1. Introduction

Heterotopic ossification (HO) is the formation of marrow-containing bone outside of the normal skeleton [1, 2]. Acquired HO following traumatic events, such as total joint replacements (TJR) [3-5], spinal cord injury (SCI) [6], traumatic brain injury (TBI) [7], fracture, muscular trauma, or war-wounded patients [8, 9], is a common and costly clinical complication. Hereditary HO, such as fibrodysplasia ossificans progressiva (FOP), is rare, progressive, and life threatening [10]. The first description of hereditary HO in FOP was made in 1692 by Guy Patin. Acquired HO as a complication of gunshot wounds was described by Dejerine and Ceillier in 1918 [11]. 16%–53% of SCI (11,000 annually) and TBI (1.4 million) patients and 40%–50% of TJR (1 million) patients will develop HO at some point.

About 10% of HO is symptomatic resulting in limitations in range of motion. Once acquired HO develops, surgical removal is the only effective treatment, normally followed by local radiation or nonsteroidal anti-inflammatory agents (NSAIDs) to prevent recurrence [12]. However, surgical removal is costly, the effectiveness of NSAIDs is variable, and radiation has been associated with malignancies [13, 14]. Further, there is no effective treatment for debilitating hereditary HO, FOP [15].

Substantial effort has been directed towards understanding the mechanisms underlying HO and finding efficient treatments. However, one crucial limiting factor has been the lack of relevant animal models. This paper reviews the long and arduous efforts to generate clinical relevant animal models and focuses on the features of major currently available models. It also summarizes some of the insights gained from these studies and discusses the potential future challenges and directions in HO research.

For the purposes of this paper, HO is defined as a heterogeneous disorder characterized by pathologic endochondral ossification with hematopoietic bone marrow in soft tissues, such as subcutaneous tissue, skeletal muscle, or fibrous tissue adjacent to joints. Similar pathologies lacking endochondral ossification, such as Progressive Osseous Heteroplasia [15, 16] or that containing no hematopoietic bone marrow, such as ectopic calcification/mineralization [17] (also called dystrophic calcification), are not included.

2. Animal Models of Hereditary HO

A typical example of hereditary HO is FOP, which is characterized by stereotyped patterned progressive ossification in soft tissues [18]. In this disorder, mutations in a bone morphogenetic protein (BMP) receptor gene, ACVR1 [19], result in dysregulation of BMP signaling that ultimately leads to FOP [10, 20]. This mutant ACVR1 activates BMP signaling in the absence of BMP ligand leading to BMP-independent chondrogenesis that is enhanced by BMP ligands [21]. This indicates that mechanism-based animal models that faithfully replicate the disorder virtually require genetic modifications leading to enhanced BMP signaling. Theoretically, there are at least five ways to modify and enhance BMP signaling: (1) by introducing a hyperactive BMP receptor, (2) by knocking out BMP inhibitors, (3) by introducing high level of BMPs, (4) by overexpressing specific BMP target genes, and (5) by modifying BMP signaling indirectly through other factors that can interact with components of BMP signaling pathway.

2.1. Animal Models That Introduce Hyperactive BMP Receptor

Animal models that introduce hyperactive BMP receptor, especially the recently found mutations in ACVR1, would seem to be the most relevant model of FOP. However, introducing a constitutively active ACVR1 mutation into zebrafish embryos failed to induce obvious HO even though strong ventralization due to enhanced BMP signaling was observed [20]. A genetically modified mouse model that carries the same mutation has not yet been reported but will likely be a valuable addition to the field in the future. Interestingly, a number of animals including domestic cats [22-25], shepherd dog [26], pigs [27], and the Southeast Asian mouse deer of the genus Tragulus [28] develop FOP-like conditions spontaneously. Even though the exact genetic bases are still unknown, it is reasonable to think that sporadic, spontaneous mutations in the BMP signaling pathway, especially mutations in ACVR1, were likely responsible for the observed FOP-like conditions. Further genetic studies of these affected animals hopefully will clarify this issue. However, due to the rarity of the events and ethical considerations, the practicality of these large animal models as a drug testing platform is in question.

Kobayashi et al. [51] reported that Col2-caBmpr1a transgenic mice that express constitutive active Bmpr1a (caBmpr1a) under the control of rat type II collagen promoter created enhanced BMP signaling. E17.5 transgenic embryos showed severe skeletal abnormalities; the femur, tibia, and patella were fused together, eliminating joint tissues [51]. This study demonstrated that overactive BMP signaling through caBMPR1A in chondrocytes stimulates chondrocyte maturation toward hypertrophic differentiation, but an HO phenotype was not reported in this model. Fukuda et al. reported using a Cre-loxP system to conditionally express a constitutively active ALK2 receptor (caALK2) to activate BMP signaling, but this produced embryonic lethality [29]. Their data indicate that low levels of caAlk2 expression are sufficient to transduce a sufficient amount of BMP signaling to compromise normal development of embryos. We speculate that conditionally activating the caAlk2 expression with late tissue-specific Cre in future studies might generate a mouse model that is useful for study of HO.

2.2. Animal Models That  Knock out BMP Inhibitors

Genetic deletion of a BMP inhibitor is another strategy for enhancing BMP signaling and potentially producing an animal model with an FOP-like phenotype. In fact, Noggin-/- mice have some congenital skeletal defects, including congenital HO, but Noggin-/- mice die soon after birth [30]. Mice with targeted disruption of Chordin, another BMP inhibitor, also die at birth, and they develop defects in inner and outer ear development and show abnormalities in pharyngeal and cardiovascular organization [31]. Mutation of another mouse competitive BMP inhibitor gene, Gremlin, resulted also in a severe abnormal skeletal pattern [52]. Interestingly, conditional deletion of Gremlin by crossing the floxed mice with osteocalcin promoter-driven cre (Oc-Cre) caused only a transient increase in bone formation and bone mass, but not HO [53]. Null mutation of less specific inhibitors of BMP signaling, such as Dan [54] or Cerberus-like [55], did not generate gross defects; these two mutant lines are born alive and fertile without a postnatal HO phenotype. Thus, although null mutation of genes encoding BMP inhibitors provided insights into how enhanced BMP signaling affects embryonic development, especially skeletal development, none of the mutant mouse lines generated by this strategy are useful for postnatal HO studies. This likely reflects the pleiotropic roles of BMP signaling in various tissues.

2.3. Animal Models That  Overexpress BMP Ligand

The rationale for enhancing BMP signaling by overexpressing BMP ligand is straightforward, but this approach has met a number of unexpected complications. Overexpression of BMP4 under the control of many different promoters does not lead to postnatal HO. For example, HO does not develop after overexpression of BMP4 under control of either the keratin promoter (K14) [32] or the bovine cytokeratin IV promoter [56]. Transgenic mice overexpressing BMP4 under control of surfactant protein-C gene promoter die from abnormally formed lungs [57]. Transgenic mice expressing BMP4 in cartilage under the control of the Col11a2 promoter/enhancer sequences die at birth due to respiratory failure [58] while mice overexpressing human BMP4 under control of mouse Msx1 minimal promoter develop no visible abnormalities [33]. Overexpression of BMP2 under the human αSM-actin promoter in an ApoE-deficient background accelerates atherosclerotic intimal calcification in transgenic lines but does not produce typical HO [59]. Mice that overexpress BMP4 under the Nephrin promoter have interesting defects in glomerular capillary formation but not the HO phenotype [60].

The only exception has been mice that overexpress BMP4 under control of the neuron specific enolase promoter (Nse-BMP4). These mice develop a phenotype that closely recapitulates the FOP phenotype and that also displays the histological hallmarks of typical acquired HO [34]. These findings suggest that overexpression of BMP itself may be necessary but is not sufficient to generate the HO phenotype and that the correct expression patterns or contexts are crucial. We have extensively characterized the phenotype in this transgenic mouse line and have used these mice, in collaboration with other labs [39], to study different aspects of HO, including definition of the events that trigger HO, the type of cells that respond to the trigger by differentiating along the osteogenic lineage, and the mechanisms underlying the spread of HO [61].

2.4. Animal Models That Overexpress a Specific BMP Target Gene

If overexpression of BMP ligand can produce HO, it is reasonable to think that expressing specific BMP target genes might also be capable of copying the phenotype of BMP overexpression. In fact, overexpression of MSX2, a BMP target gene, can induce an HO-like phenotype. MSX2 overexpression elicited the phenotype under control of either an ubiquitous promoter, such as CMV, a tissue specific promoter, such as TIMPl (tissue inhibitor of metalloproteinase 1), or the endogenous MSX2 promoter [35, 62]. Overexpression of another BMP target gene, Runx2, under the type II collagen promoter also caused an HO-like phenotype and ectopic expression of hypertrophic chondrocyte markers [36]. These two models show that both MSX2 and Runx2 can partially mediate the osteogenic effects of BMPs in vivo. However, since the phenotypes in these two lines do not closely mimic that of FOP, the relevance of these models to the human disease is still unclear. Moreover, multiple transgenic lines that overexpress other BMP target genes, especially the Id family genes, that is, Id1-Id4 [63-65], have failed to produce an HO phenotype. This could be partially explained by the inadequate tissue specific promoters used in generating these transgenic lines. However, the failure more likely indicates that not all BMP target genes are important in mediating the HO phenotype, even though Id1 and Id3 are positive factors in promotion of bone formation in vivo [65].

2.5. Animal Models That Overexpress Other Factors That Indirectly Modify the BMP Signaling Pathway

Theoretically, it is also possible to enhance the BMP signaling indirectly through factors that can interact with components of the BMP signaling pathway. For example, overexpression of Fos in bone cells under control of an FBJ long terminal repeat element (H2-FosLTR) resulted in the development of calcified tumors similar to HO, and Fos-ES cell chimeras developed chondrosarcomas with high efficiency at all skeletal sites containing cartilage [37]. However, transgenic mice that overexpress other related AP1 members (e.g., JUN and FOSB) do not exhibit abnormalities, despite high expression in bone tissue. Not surprisingly, further studies provided evidence of specific interactions between the BMP-signaling pathway and c-Fos, but not the other related AP1 members in FOP-like lesions [38].

Overall, even though there are multiple ways to enhance BMP signaling in vivo, only a few genetic modified animal lines showed typical HO, or a phenotype resembling FOP. Further, only one line, Nse-BMP4 transgenic mice, closely recapitulated the major aspects of the FOP phenotype.

3. Animal Models for Acquired HO

Acquired HO usually follows traumatic events, such as fracture, total hip arthroplasty, muscular trauma, spinal cord injury, or central nervous system injury. It is a relatively frequent clinical complication with a wide clinical spectrum but normally it has a relatively benign course [12]. The etiology of common acquired HO is still unclear, and multiple contributing factors have been proposed including BMPs, inflammation, prostaglandin E2, hypercalcemia, hypoxia, abnormal nerve activities, immobilization, and disequilibrium of hormones [66, 67]. Lack of deep understanding of underlying molecular mechanisms has directly hindered the validation of existing animal models, and this also has limited the development of new mechanism-based animal models. Currently, there are several available animal models that can produce typical HO: (1) heterotopic implantation models, (2) hip arthroplasty model, (3) the immobilization manipulation model (also called the Michelsson model), (4) Achilles tenotomy model, (5) trauma-induced model, and (6) models generated by injection of irritants and other materials to muscle.

3.1. Heterotopic Implantation Model

Currently the most commonly used animal model for HO involves the surgical implantation of BMP containing matrix at heterotopic sites. Implantation of demineralized bone matrix was first used by Urist in 1965 [68]; then Wozney et al. were able to repeat the experiment using partially purified BMP proteins [69]. Currently, the most widely used approach is BMP matrigel implantation [39]; an advantage of this method is that a chilled mix can be injected into heterotopic sites as a liquid which gels on site at body temperature and thereafter releases BMP4 continuously at the site. Many modifications/variations of this method have been used in different species under different conditions, including introduction of a DNA construct that produces BMPs [40], microbubble-enhanced transcutaneous sonoporation of human BMP2 [70], nanogel-cross-linking hydrogel as a scaffold [41], implantation of a slow-release system of polylactic acid and rhBMP-2, or sintered porous-surfaced Ti-6Al-4V implants coated with native BMPs [71].

One interesting variation on this theme involves direct injection into the heterotopic site of cells that have osteogenic, and/or osteogenic factor producing potential, such as bone marrow cells [42], or implantation of a diffusion chamber containing such cells. Tested cell types have included urinary tract epithelia [72], certain transformed cells such as transformed human amnion cells (FL cells) [73], Moloney sarcoma [74], and epithelial-like cells [75]. In a similar system, these cells are impregnated into ceramic blocks to test their osteogenic activity [76] in the presence or absence of an osteogenic inducer.

Another interesting approach takes advantage of the osteoinductive ability of certain biomaterials, such as microporous calcium phosphate ceramic particles [43], that do not release BMP or other known osteogenic factors. The mechanism of osteoinduction by such biomaterials is not currently clear, although the geometry of the material is thought to be important [77].

Generally, heterotopic implantation models are straightforward, repeatable, and mechanistically relevant to human HO. However, certain limitations do exist: (1) they are artificial systems that may create unphysiologically high local concentrations of osteogenic factors in implanted sites leading to effects not relevant to the human disorder, (2) the implantation is a local event and thus has limited ability to mimic the potential effects of the involvement of multiple systems, (3) different variations of this method have variable reliabilities and relevance to human conditions, (4) the incidence of implantation-induced bone formation varies depending upon the material or animal species. Normally rabbits are the most, and mice the least, susceptible [78], and experimental conditions that produce ectopic bone do not always coincide with clinical observations in humans.

3.2. Hip Arthroplasty Model

HO is commonly observed after hip arthroplasty in humans for unknown reasons. To develop a model relevant to the human condition [44], Schneider et al. subjected rabbits to surgery analogous to human hip arthroplasty either with or without muscle and bone injury on each hip. This led to HO, and the effectiveness of postoperative radiation in prophylaxis of HO was then analyzed using this model. The rationale behind this model is straightforward, and it can produce HO with certain reliability; however, despite being a phenocopy of the human condition, it is not a mechanism-based model. This method has not been widely adapted by other investigators, probably due to the relatively complicated surgical procedure.

3.3. The Michelsson Model (Also Called Immobilization Manipulation Model)

Michelsson et al. [45] found that repeated forced mobilization of an immobilized knee joint caused HO in the quadriceps muscle in rabbits, and similar procedures can induce HO around other joints in the rabbit as well. The precise inductive stimulus has not been identified in this model, but an interaction between the periosteum and the necrotic muscle seems necessary since the introduction of a plastic membrane between bone and muscle prevents bone formation [79]. The first sign of osteoblastic activity was seen in the periosteum, and the new bone was often formed in continuity with the periosteum. Interestingly, early changes in prostaglandins preceded bone formation [80], consistent with the hypothesis that inflammation is the basis of the heterotopic bone formation in that process. Several authors have used this model to study the development and prevention of HO in animals [81-85]. However, since HO in this model is not affected by denervation, in contradistinction to clinical findings in patients with neurologic injuries, the relevance of this model to human HO is unclear.

3.4. Achilles Tenotomy Model

The Achilles tenotomy model was first described in rats by Buck in 1953 [46], and in 1983, McClure applied the model to mice and found that ectopic bone developed in 60% of animals by 5 weeks and in 100% by 10 weeks after Achilles tenotomy [86]. The advantages of this model are its relative simplicity and excellent predictability. However, the molecular mechanisms of HO induced by Achilles tenotomy are poorly understood, and the relevance of this model to clinical conditions is also unclear since ectopic bone formation in Achilles tendon is a rare condition in humans. Further, in humans HO is not only associated with prior surgery or trauma to the tendon but is also an important manifestation of rheumatoid arthritis and ankylosing spondylitis [86].

3.5. Trauma-Induced Models

Traumatic muscle or CNS injury often leads to HO in humans, but the underlying causative factor(s) remains unknown. Efforts to establish trauma induced models have had only limited success. Zaccalini and Urist failed to induce HO in rabbit thigh by blunt force [47]. Walton et al. reported limited success in inducing HO in sheep thigh by repeated blunt force (7 out of 42 sheep) [48]. Further, intramembranous and not endochondral ossification was the histological feature within scar tissue. Based on these reports, these models do not seem to be sufficiently reliable to be used routinely. Further, the failure of this strategy has forced us to rethink why trauma, which clearly plays a role in human HO, does not routinely induce it in such models. Fortunately recent studies using Nse-BMP4 mice have demonstrated that mild trauma leads to HO with high frequency irrespective of which limb is injured. In turn this suggests that trauma-induced HO depends upon susceptibility determined by other factors—in this case elevated levels of BMP4. The high frequency and reproducibility of trauma-induced HO in this model may provide a means of exploring the underlying mechanisms.

3.6. Irritant and Other Miscellaneous Material-Induced Models

Injection of various irritant materials into muscle sometimes leads to HO. For example, Heinen et al. reported the induction of HO in rabbit by injection of 40% ethanol [49]. Selle and Urist also reported that acid-alcohol could induce HO in a small percent of animals, while injections of calcium chloride produced only amorphous calcified plaques, not new bone or cartilage [50]. In addition, Arai et al. [87] and Caselli et al. [88] reported a controversial finding that colchicine induced intramedullary bone formation. This finding could not be repeated by K. H. Wlodarski and P. Wlodarski [89], and later Dudkiewicz et al. found that colchicine actually inhibits HO in a rabbit model [90]. The issues of repeatability and relevance of these models to human HO limits their potential utility.

Overall, due to limited understanding of molecular mechanisms, most animal models for acquired HO can only mimic some aspects of the human conditions. Further, the reliability and questionable clinical relevance hinder their use as drug test platforms. Thus caution must be taken in choosing one of these models to be appropriate for the specific question being asked.

4. Summary and Future Directions

Multiple animal models have been generated for studies of HO (see Table 1). For the simplicity of description in this review, we divided these models into two major groups, acquired or hereditary. However, to some extent, this division is arbitrary since injury and inflammation facilitates and triggers HO in FOP as well as in animal models of hereditary HO, and the high variability in susceptibility of different individuals to acquired HO suggests a genetic basis for individual predisposition. In fact, accumulating clinical and experimental evidence suggests that similar cellular and molecular mechanisms underlie the pathophysiology of all typical HO which involves formation of fibroproliferative lesions containing cells that follow the classic endochondral ossification pathway. Thus, in hereditary HO, a specific genetic mutation plays the central role, while in acquired HO the environmental factors play more important roles. For this reason, some animal models such as Nse-BMP4 mice can be used to study both hereditary and acquired HO.

Table 1

Summary of commonly used animal models.

	General strategy	examples	HO related phenotypes?	references
Animal models of hereditary HO	Hyperactive BMP receptor	ACVR1(R206H) mutation into zebrafish	No HO	[20]
		Conditionally expresses caALK2 in mice	Embryonic lethal, no HO	[29]
	Knocking out BMP inhibitors	Noggin-/- mice	Postnatal lethal, skeletal defects, HO?	[30]
		Chordin-/- mice	Postnatal lethal, skeletal defects	[31]
	Overexpressing BMPs	K14-BMP4	No HO	[32]
		Msx1-BMP4	No observable defects	[33]
		Nse-BMP4	Progressive HO, FOP-like	[34]
	Overexpressing BMP target genes	CMV-MSX2, TIMPl-MSX2 or MSX2-MSX2	HO-like	[35]
		col2-Runx2	HO-like	[36]
	Modifying BMP signaling indirectly	H2FosLTR-Fos	Calcified tumors similar to HO	[37]
		Fos-ES cell chimeras	Chondrosarcomas, similar to HO	[38]
Animal models of acquired HO	Heterotopic implantation	BMP matrigel implantation	HO	[39]
		DNA construct that produces BMPs	HO	[40]
		Nanogel-cross-linking hydrogel as a scaffold	HO	[41]
		Bone marrow cells	HO	[42]
		Microporous calcium phosphate ceramic	HO	[43]
	Hip arthroplasty model	Hip arthroplasty in rabbits	HO	[44]
	Michelsson model	Immobilization manipulation in rabbits	HO	[45]
	Achilles tenotomy model	Achilles tenotomy in rats	HO	[46]
	Trauma-induced model	Blunt force in rabbit thigh	No HO	[47]
		Repeated blunt force in sheep thigh	HO in small % of treated animals	[48]
	Inject irritants and other materials	Injection of 40% ethanol	HO	[49]
		Injection of acid-alcohol	HO in small % of treated animals	[50]

Understanding the fundamental pathophysiology underlying HO is the key to development of mechanism-based animal models. Just as determination of the genetic basis of FOP opened up a whole new avenue for generating models for hereditary HO, deeper understanding of the molecular mechanisms underlying acquired HO will lead to more fruitful approaches in generating new animal models for the disorder. Multiple contributing factors are necessary for acquired HO including a trigger (trauma, injury), osteogenic progenitor cells, and a permissive microenvironment. However, thus far there is no single hypothesis that integrates most clinical and experimental findings, and current data strongly suggests the involvement of multiple organ systems in this disorder. For this reason, future multidisciplinary studies of neuroimmunological interactions and osteoneuroimmunology using currently available animal models, such as Nse-BMP4 mice, will be necessary to provide the new insights which in turn could lay the foundation for new mechanism-based animal models.