Dysregulation of Local Stem/Progenitor Cells as a Common Cellular Mechanism for Heterotopic Ossification

Abstract

Heterotopic ossification (HO), the abnormal formation of true marrow-containing bone within extraskeletal soft tissues, is a serious bony disorder that may be either acquired or hereditary. We utilized an animal model of the genetic disorder fibrodysplasia ossificans progressiva to examine the cellular mechanisms underlying HO. We found that HO in these animals was triggered by soft tissue injuries and that the effects were mediated by macrophages. Spreading of HO beyond the initial injury site was mediated by an abnormal adaptive immune system. These observations suggest that dysregulation of local stem/progenitor cells could be a common cellular mechanism for typical HO irrespective of the signal initiating the bone formation.

Introduction

Typical heterotopic ossification (HO) is characterized by pathologic endochondral bone formation inside soft tissues, such as subcutaneous tissue, skeletal muscle, and fibrous tissue adjacent to joints [1–4]. The clinical spectrum of HO is wide, and it may be either an acquired or a hereditary disorder. All typical HO involves formation of fibroproliferative lesions containing cells that follow the classic endochondral ossification pathway to form HO. Acquired HO following traumatic events, such as fracture, total hip arthroplasty, muscular trauma, spinal cord injury, or central nervous system injury, is relatively frequent but normally benign and self-limited. The histologic features of acquired HO are very variable. By contrast, hereditary forms such as fibrodysplasia ossificans progressiva (FOP) are rarer, progressive, and life-threatening [5]. The etiology of common acquired HO is unclear, although multiple contributing factors have been proposed. Bone morphogenetic proteins (BMPs) may be released locally from normal bone or infiltrating inflammatory cells in response to venous stasis or inflammation. Other factors such as prostaglandin E2, hypercalcemia, hypoxia, abnormal nerve activities, immobilization, and disequilibrium of hormones may also contribute to this disorder [1–4]. Mutation of ACVR1, a type I BMP receptor [6], leads to FOP [7, 8], suggesting that disturbance of the normal homeostasis of BMP signaling is sufficient to cause the disease. However, clinical observations have implicated participation of the immune system and non-inherited triggers (or tissue-damaging events) that facilitate the disease process [9], and the precise pathophysiologic mechanisms underlying HO, either acquired or hereditary, are unknown.

We previously reported that a transgenic mouse line over-expressing BMP4 under the control of the neuron-specific enolase (Nse) promoter (Nse-BMP4) develops a phenotype that closely recapitulates FOP [10] and that also displays the histological hallmarks of typical acquired HO. We used these mice to define the nature of the events triggering HO, the type of cells that respond to the trigger by differentiating along the osteogenic lineage, and the mechanisms underlying the spread of HO. We find that macrophage responses to tissue injury stimulate local stem/progenitor cells to differentiate into bone, suggesting that dysregulation of local stem/progenitor cells could be a common cellular mechanism for typical HO.

Materials and Methods

Animals and Injury Models

The Nse-BMP4 transgenic mice used in this study have been described previously [10]. The Myf5-cre line was a kind gift from Dr. Philippe Soriano (Fred Hutchinson Cancer Research Center, Seattle, WA, http://www.fhcrc.org). All other lines were from the Jackson Laboratory (Bar Harbor, ME, http://www.jax.org) unless otherwise specified. All animal experiments in this study were approved by the Animal Care and Use Committee at Northwestern University.

Muscle Injury

One-month-old inbred Nse-BMP4 transgenic (>6 generations) and wild-type (WT) mice were injected with a single dose of 0.1 ml of 10 μM cardiotoxin (CTX; Calbiochem, La Jolla, CA, http://www.emdbiosciences.com), diluted in phosphate buffered saline (PBS; Invitrogen, Carlsbad, CA, http://www.invitrogen.com), into the thigh muscles (quadriceps deep and superficial), as described previously [11]. Inbred Nse-BMP4 mice injected with PBS, WT mice injected with PBS, and WT mice injected with CTX served as control groups (10 mice in each group). HO formation was assayed by x-ray imaging 3 weeks later, and the injected or control hind 1egs were harvested for further histology study.

Skin Injury

Inbred 1-month-old Nse-BMP4 and WT mice (10 mice in each group) were anesthetized, and the hairs of right hind limbs were shaved. A 5-mm sharp, shallow cut was made through the skin but sparing the muscles. HO formation was assayed by x-ray imaging 3 weeks later, and the injured or control hind 1egs were harvested for further histology study.

Macrophage Depletion

Injection of Liposome-Encapsulated Clodronate into Inbred Nse-BMP4 Mice

Liposome-encapsulated clodronate or PBS was purchased from ClodronateLiposomes.org (Amsterdam, The Netherlands, http://www.ClodronateLiposomes.org). Depletion of macrophages was achieved by injection of 0.2 ml of liposome-encapsulated clodronate i.p. twice a week for 10 weeks (liposome-encapsulated PBS as control), according to the provider's instructions. Two days after the first injection, one 0.2-ml injection was performed subcutaneously in the right hind limb. Four hours after the subcutaneous injection, skin injury was performed as described above close to the injection site. X-ray images were taken weekly after the injury for 14 weeks to detect HO. The same procedure was repeated three times with a total of 17 Nse-BMP4 mice (1−2 months old) (nine for the encapsulated clodronate and eight for the PBS control).

Injection of Diphtheria Toxin into CD11b-DTR/Nse-BMP4 Double Transgenic Mice

Diphtheria Toxin (DT; List Biological Laboratories, Inc., Campbell, CA, http://www.listlabs.com) was injected i.p. (10 ng/g in 0.1 ml of PBS, or 0.1 ml of PBS as control) three times weekly for 10 weeks into CD11b-DT receptor (DTR)/Nse-BMP4 double transgenic mice (F1 generation) or CD11b-DTR single transgenic mice (as a control group), as described previously [12]. The DTR-enhanced green fluorescent proteins fusion cDNA is inserted between the human CD11b promoter and the human growth hormone sequences that provides splicing and polyadenylation signals in CD11b-DTR transgenic mice [12]. CD11b (Mac-1) is a common myeloid marker often used to identify macrophages. One day after the first injection, one 10 ng/g injection was given subcutaneously into the right hind limb. Four hours after the subcutaneous injection, skin injury was performed close to the injection side. X-ray images were taken weekly after the injury for 14 weeks. The same procedure was repeated in two litters. Seventeen CD11b-DTR/Nse-BMP4 double transgenic mice (nine for DT and eight for PBS) and 12 CD11b-DTR single transgenic mice (six for DT and six for PBS) were used.

Genetic Lineage Tracing

Rosa26 conditional reporter mice were first mated with Nse-BMP4 mice. The double transgenic mice were selected and mated with different cre lines under the control of different tissue-specific promoters. Five different cre lines were used: CD19-cre [13], CD19 was used as a B-cell lineage marker; LCK-cre [14], co-receptor of CD21, leukocyte-specific protein tyrosine kinase (LCK) was used as a T-cell lineage marker; Lyz-cre [15], lysozyme was used as a monocyte/macrophage lineage marker; Myf5-cre [16], myogenic regulatory factor 5 (Myf5), was used as a myogenic lineage cell marker; and nestin-cre [17], in which cre recombinase expression is controlled by the rat nestin promoter and intron two enhancer, was used to mark somite-derived cells. Skin and/or muscle injury was performed on triple transgenic mice from the second round of mating. Legs were harvested 1−2 weeks after the skin injuries for histological studies and LacZ staining.

Histology and Immunohistochemistry and Western Blotting

Alizarin Red and Alcian Blue staining were done according to published protocols [10]. Hematoxylin and eosin staining was performed on fixed or frozen tissue sections using Harris Modified Hematoxylin and Eosin Y Solution (Sigma, St. Louis, MO, http://www.sigmaaldrich.com), according to the manufacturer's instructions. LacZ staining was performed as previously described [18]. Immunohistochemistry: Immunostaining for different markers was done according to standard protocols. Briefly, sections and cultured cells were fixed with 4% paraformaldehyde in PBS. Nonspecific binding was blocked with 10% normal serum diluted in 1% bovine serum albumin (BSA; Jackson ImmunoResearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com) and 0.25% Triton X-100 (Sigma) for 1 hour in room temperature. The sections were then incubated with primary antibodies diluted with 1% BSA + 0.25% Triton X-100 at 4°C overnight. The sections were then incubated with appropriate secondary antibodies (Cy3 or Cy2 conjugated antibodies (Jackson ImmunoResearch Laboratories) diluted with 1% BSA + 0.25% Triton X-100 or Alexa Fluor 488, Alexa Fluor 594, and Alexa 647 (1:1000, Invitrogen) in the dark at room temperature for 2 hours. Counterstaining was then performed with 4,6-diamidino-2-phenylindole (1:5000). Mouse anti-col II (CIIC1) was obtained from the Developmental Studies Hybridoma Bank (Iowa City, IA, http://dshb.biology.uiowa.edu. Mouse anti-BMP4 was purchased from Chemicon (Temecula, CA, http://www.chemicon.com), and rat anti-CD45 was purchased from BD (BD PharMingen, San Diego, http://www.bdbiosciences.com/index_us.shtml). Rabbit anti-ACVR1 (#AP7101c), rabbit anti-BMPR1A (#AP2004a), and rabbit anti-BMPR1B (#AP2005b) were purchased from Abgent (Abgent, Inc., San Diego, http://www.abgent.com) for Western blotting.

Soft X-Ray Assessment

The images of radio-opaque HO were acquired by whole body x-ray examination at 42 kV, 25 mA, 0.05 seconds (General Electric, Model 225) as previously reported [10] or with use of the TruDR Digital Radiography System (Sound Technologies, Carlsbad, CA, http://www.soundvet.com).

Results

Muscle or Skin Injuries Facilitate HO Formation in Nse-BMP4 Mice

Nse-BMP4 mice never develop HO before 2 months of age and there is large variability among individual littermates in terms of onset time of HO. Since the mice have the same genetic backgrounds and live in similar environments, this suggested that the triggering event must be random in nature. We focused on two such possible triggers that are clinically relevant—muscle or skin injury—in young (∼1 month old) Nse-BMP4 inbred mice. We first injured muscle using cardiotoxin injection [11] and found that all cardiotoxin-injected limbs of Nse-BMP4 mice developed HO within three weeks. By contrast, no limbs in control groups developed HO. (Fig. 1 A–1D, 1M). Histological studies at different time points after the cardiotoxin injections found sequential pathological changes of muscle degeneration followed by profound fibroproliferative lesions, and then cells in the lesions that followed the classic endochondral ossification pathway to form HO with marrow (Fig. 1E, 1F, 1H). Consistent with previous reports [11], we observed robust muscle regeneration in WT mice 7 days after cardiotoxin injection (Fig. 1G). In contrast, there were profound fibroproliferative lesions and chondrocyte formation at this time point in cardiotoxin-injected Nse-BMP4 muscles (Fig. 1H). These findings indicated that muscle injury is an efficient trigger of HO in Nse-BMP4 mice.

Figure 1.

Muscle or skin injuries facilitate HO formation. (A–D): Typical x-ray images taken 3 weeks after cardiotoxin injection. (A): WT mouse injected with PBS. (B): WT mouse injected with CTX. (C): Nse-BMP4 mouse injected with PBS. (D): Nse-BMP4 mouse injected with CTX. White arrow in (D) points to the HO. (E, F): Hematoxylin and eosin (HE) staining of HO sections from CTX-injected mice (3 weeks after injection). (E): Low-power image demonstrates HO located within muscle and surrounded by fibroproliferative lesions. (F): Higher power image shows the prominent muscle degeneration, profound fibroproliferative lesions, and HO with marrow. (G, H): Seven days after CTX injection, Nse-BMP4 and WT mice showed drastically different histological features. (G): WT mice show significant muscle regeneration. Note the newly generated muscle cells that have centrally localized nuclei. (H): In contrast, Nse-BMP4 mice have profound fibroproliferative lesions with chondrocyte formation at this time point. (I, J): Typical x-ray images of WT and Nse-BMP4 mice taken 3 weeks after skin injuries. White arrow in (J) indicates HO at the injury site. (K, L): HE staining of the injury site from an Nse-BMP4 mouse (3 weeks after injury). (K): Low-power image shows the general morphology and the features of HO generated by skin injury. Black arrow indicates the injury site. Notice the fibroproliferative lesions, the chondrocyte formation, and the nascent trabeculae-like structure at one edge. (L): High-power image of the nascent trabeculae-like structure. (M): Quantitative summary of muscle and skin injuries experiments. Bar = 500 μm in (E) and (K). Bar = 100 μm in (F)–(H) and (L). Abbreviations: BMP, bone morphogenetic proteins; CTX, cardiotoxin; HO, heterotopic ossification; Nse, neuron-specific enolase; PBS, phosphate buffered saline; WT, wild type.

Since severe muscle injury is rarely observed in mice housed in the vivarium, we examined the effects of skin injury, which is more commonly observed. Three weeks after skin injury, the majority of Nse-BMP4 mice (8/10) developed subdermal HO (Fig. 1I–1M), whereas neither WT mice with the same injury nor uninjured Nse-BMP4 mice developed HO. Detailed histological studies found that the HO after skin injury closely mimicked the cascade of events of classic endochondral ossification [10] (data not shown). However, in contrast to muscle injury, not every mouse developed HO after skin injury. Since the muscle injury was more severe than the skin injury, this suggested a dose-dependent effect depending upon the degree of injury/inflammation. To test the injury dosage hypothesis, mice that did not develop HO at 3 weeks after the first round of skin injury were subjected to a second round of skin injuries, and all mice developed HO after the second injury (data not shown).

Macrophages Mediate the Injury Response and Trigger HO

These observations suggested that an abnormal injury/immune response, either adaptive or innate, mediates HO. In view of the importance of BMP signaling to HO, we sought to determine whether inflammatory macrophages might be a source of BMPs. We examined thioglycollate-elicited peritoneal macrophages from WT, Nse-BMP4 (inbred), and outbred (CD11b-DTR/Nse-BMP4) mice and found that peritoneal macrophages from all mice express BMP4. Macrophages from CD11b-DTR/Nse-BMP4 (outbred) mice expressed higher levels of BMP4 than WT macrophages and extensively inbred Nse-BMP4 mice have even higher levels of BMP4 expression, suggesting that both the transgene and the endogenous BMP4 gene are expressed by activated macrophages (Fig. 2A, 2B).

Figure 2.

Macrophage depletion inhibits HO formation. (A): Western blotting shows that BMP4 is expressed by thioglycollate-elicited peritoneal macrophages and that the transgene expression level is higher in inbred Nse-BMP4 mice than in Nse-BMP4/CD11b-DTR double transgenic mice. Each lane represents an individual mouse. β-Actin acts as the loading control. (B): BMP4 protein expression normalized to actin expression in WT, Nse-BMP4/CD11b-DTR, and inbred Nse-BMP4 mice. (C): Effect of macrophage depletion and transgene dosage. The percentages of mice that developed HO at different time points are depicted. Nse-BMP4 mice with clodronate (CLO, heavy green line) or without (phosphate buffered saline (PBS), heavy black line), Nse-BMP4/CD11b-DTR with DT (light blue line) or without (PBS, light black line) are presented. All other control groups, including WT mice with CLO or without (PBS), CD11b-DTR single transgenic mice with DT or without (PBS), never generate HO and are depicted. The thickness of the lines is proportional to the level of BMP4 transgene expression. Comparing the heavy green line with the heavy black line reveals the inhibitory effect of dlodronate treatment. Comparing the light blue line with the light black line reveals the inhibitory effect of DT treatment. Comparing the heavy black line with the light black line reveals the transgene dosage effect. The horizontal line at the bottom indicates the period of the injection following the injuries. Abbreviations: *, differs from WT by ANOVA at p < .05; **, differs from WT by ANOVA at p < .01; BMP, bone morphogenetic proteins; CLO, clodronate; DT, diphtheria toxin; DTR, DT receptor; HO, heterotopic ossification; Nse, neuron-specific enolase; WT, wild type.

This supported the hypothesis that injury-recruited macrophages might mediate the injury response by secreting osteogenic factors, including BMP4, which in turn trigger HO. We tested this hypothesis with two independent methods. First, we injected liposome-encapsulated clodronate into Nse-BMP4 inbred mice to deplete macrophages, a depletion strategy that has been used successfully in multiple applications [19–22]. Clodronate has poor cell membrane permeability, but liposome-encapsulated clodronate can be readily taken up by cells of the reticuloendothelial system, especially macrophages. After phagocytosis, clodronate is metabolically incorporated into nonhydrolyzable analogs of ATP that accumulate in phagocytic cells, resulting in induction of apoptosis. Liposome-encapsulated clodronate is not toxic to nonphagocytic cells [23]. The efficiency of macrophage depletion in the Nse-BMP4 injected mice was confirmed by F4/80 antibody staining, which demonstrated a marked reduction of macrophages in all tissues examined (supporting information Fig. 1 and data not shown). Weekly x-ray imaging after injury found that injection of clodronate significantly reduced HO formation in Nse-BMP4 mice (Fig. 2C).

Since liposome-encapsulated clodronate is potentially toxic to other cells with endocytic activity, such as dendritic cells, osteoclasts, and local tissue stem/progenitor cells, the protection against HO could not be ascribed with certainty to depletion of macrophages. We therefore used another independent strategy for depleting macrophages, injection of diphtheria toxin to CD11b-DTR/Nse-BMP4 double transgenic mice. The CD11b-DTR transgenic mouse expresses the human DT receptor under the control of the CD11b promoter [12], and macrophages can be conditionally depleted by injection of DT [24]. Injection of DT to the double transgenic mice markedly reduced HO formation (Fig. 2C), further strengthening the conclusion that macrophages are involved in triggering HO. Additionally, inbred mice form HO much faster after injury than outbred CD11b-DTR/Nse-BMP4 mice, suggesting that the gene dosage of BMP4 played a key role in triggering the HO.

Adaptive Immune System Plays a Role in HO Spreading

To address whether the adaptive immune system also plays a role in HO in this model, we crossed the Nse-BMP4 mice with recombination activating gene 1 (RAG1) null mice, which have no mature B and T lymphocytes, [25] to eliminate the adaptive immune response. Interestingly, Nse-BMP4/RAG(−/−) mice developed HO after injury without delay (Fig. 3 and data not shown), indicating that the adaptive immune system was not necessary for the initial formation of HO. However, the rate of spreading and overall amount of HO were much smaller in Nse-BMP4/RAG(−/−) than in Nse-BMP4/RAG(−/+) mice (Fig. 3 and data not shown), suggesting that the adaptive immune system plays a role in spreading of HO. The findings were the same at other time points (1 month, 2 months, and 4 months) after the injury (data not shown). These observations indicate that the initiation/triggering stages and the later spreading stages of HO are separable and controlled by different mechanisms; macrophages in the innate immune system are an efficient trigger for HO, whereas the adaptive immune system may play a key role in spreading it. These findings may help to explain the major phenotypic difference between FOP and common acquired HO.

Figure 3.

The adaptive immune system plays a key role in the spread of heterotopic ossification (HO). X-ray images (3 months after skin injury of right hind limb) demonstrate that the size of HO in Nse-BMP4/RAG(±) mice (A, B) is larger and spreads to a far remote location (left side) compared to that of Nse-BMP4;RAG(−/−) mice (C, D). White arrows indicate the HO. Abbreviations: BMP, bone morphogenetic proteins; Nse, neuron-specific enolase; RAG, recombination activating gene 1.

Genetic Lineage Tracing Excluded Major Candidate Populations as Receptive Cells

What are the receptive cells that respond to the macrophages by proliferating and undergoing osteogenic differentiation? Numerous candidates have been proposed including vascular cells [26], muscle stem cells (see “Eleventh Annual Report” at http://www.ifopa.org/reports.html), inflammatory mast cells [27], and others [28], but the origin of the bone-forming cells in pathologic bone has not been unequivocally determined. We preformed genetic lineage tracing experiments to identify the receptive cell populations. Conditional reporter mice, Rosa26, were first mated with Nse-BMP4, and then the double transgenic mice were mated with different cre lines to label different candidate lineages. CD19-cre [13], LCK-cre [14], and Lyz-cre [15] were used to label the B, T, and mononuclear cells of hematopoietic origin, respectively, Myf5-cre [16] was used to label the myogenic population, and nestin-cre was used to label the somitogenic populations [17]. LacZ staining and histological studies of triple transgenic sections focused primarily on the early stages after injury, when fibroproliferative lesions, chondrocytes, and hypertrophic chondrocytes are the predominate feature. All cre lines labeled the appropriate target subpopulations properly, but these labeled subpopulations failed to contribute significantly to early fibroproliferative lesions, chondrocytes, or later bone formation (Fig. 4).

Figure 4.

Five tested cre lines worked properly, but none of these cre-labeled cell populations made a significant contribution to heterotopic ossification. (A): Schematic mating and tracing strategy. (B, C): Typical image of X-gal staining of section from injured triple transgenic mice carrying CD19-cre with fast nuclear red counterstaining (A). Note that none of the cells in fibroproliferative lesions are LacZ positive, although in the same animal the CD19-cre mediated recombination worked properly in the spleen, which contains many LacZ positive cells in the germinal center (C). (D): Typical image of X-gal staining of section from injured triple transgenic mice carry Myf5-cre. Notice that none of the cells in fibroproliferative lesions was LacZ positive. Although the Myf5-cre mediated recombination worked efficiently in neighboring muscle cells. (E, F): Typical image of X-gal staining of section from injured triple transgenic mice carrying Lyz-cre (E). Note that no cells in fibroproliferative lesions are LacZ positive, although in the same animal the Lyz-cre mediated recombination worked in the spleen in which many cells in red pulp are LacZ positive. (F). (G): Typical image of X-gal staining of section from injured triple transgenic mice carrying Nestin-cre. Many subcutaneous muscle cells are LacZ positive (black), whereas there are no cells in the HO that are LacZ positive. Green or yellow florescence is autoflorescence, which is present in ossificated tissues and the surface of the skin. (H, I): Typical image of X-gal staining of section from injured triple transgenic mice carrying LCK-cre. (H): Note that no cells in fibroproliferative lesions are LacZ positive, although in the same animal the LCK-cre mediated recombination worked properly in the thymus in which many medullary cells are LacZ positive (I). Abbreviations: BMP, bone morphogenetic proteins; Nse, neuron-specific enolase.

Discussion

Heterotopic ossification is defined as the formation of lamellar bone inside soft-tissue structures where bone normally does not exist, but there is no consensus on the subclassification of different types of HO [1, 2, 5]. In this report, we arbitrarily divided HO into two major categories, typical and atypical based on pathohistological features. All typical HO (both acquired and hereditary HO) involves formation of fibroproliferative lesions containing cells that follow the classic endochondral ossification pathway to form HO, whereas atypical HOs have variable features. We hypothesized that there are common underlying mechanisms for all types of typical HO. More specifically, we hypothesized that osteogenic factors, such as gain-of-function of BMP receptors (ACVR1 in FOP patients) or high BMP activity (BMP4 transgene in Nse-BMP4 mice), can trigger the pathogenic endochondral ossification process of typical HO in a permissive microenvironment, and the local receptive cells respond to the trigger and turn into HO.

We tested this hypothesis in the Nse-BMP4 transgenic mouse model. There are several currently available animal models that can produce typical HO, such as the immobilization-manipulation model (also called the Michelsson model) [29], hip arthroplasty model [28], Achilles tenotomy model [30], heterotopic implantation of demineralized bone matrix or BMPs, or injection of alcohol or calcium chloride [31]. These models have variable repeatability and clinical relevance but none of them have well-defined molecular or cellular mechanisms. furthermore, none of these animal models recapitulates the phenotype of hereditary HO. The current study not only highlighted the similarities between the Nse-BMP4 mouse model and FOP but also provided insights that support a mechanistic model that can explain the differences as well as the shared cellular mechanisms between FOP and common acquired HO (Fig. 5).

Figure 5.

Current working model. See Discussion section for the details.

We demonstrated that injuries, including relatively minor ones, can trigger and synchronize the formation of HO. The occurrence of minor injuries should be random and increase with age, which is consistent with our observation that there is large variability among individual littermates in terms of time of onset. This suggests that the “spontaneous HO” in older NSE-BMP4 animals or patients with FOP is actually caused by unnoticed minor injuries. Standardization of minor injuries to young animals greatly reduced the variations among individual animals. This not only facilitated examination of the underlying mechanism but also laid a solid foundation for future drug tests on this animal model.

Our studies demonstrate that macrophages can mediate the injury response and trigger the HO cascade in this model. Consistent with this observation, macrophages accumulation is observed in response to implants [32], venous stasis [33], and inflammation, [34, 35] and macrophages are capable of secreting BMP and other osteogenic cytokines in vitro and in vivo [36–38]. We further demonstrated that depletion of macrophages substantially reduces the occurrence of HO. In turn, this suggests that depletion of macrophages can potentially be used therapeutically to prevent HO.

Although all typical HO involves the creation of fibroproliferative lesions followed by classic endochondral ossification, one prominent difference exists: acquired HO is normally self-limited whereas the HO in FOP is always progressive [5]. This study demonstrates that the triggering of HO and its later spreading are separable and controlled by different mechanisms. Specifically, our study suggests that macrophages trigger formation of HO by local inducible stem/progenitor cells, but that the adaptive immune system plays a key role in spreading it. There is no evidence to suggest participation of the adaptive immune system in acquired HO, whereas numerous clinical features implicate the immune system in HO formation in FOP patients. This can explain not only the key difference between common acquired HO and FOP but also why immune suppression treatment can temporarily control the symptoms of FOP but cannot prevent new HO formation and disease progression in the long term.

There has been substantial debate about the responsive cell type that forms HO. Theoretically, any proliferative stem/progenitor cell population that has the potential to differentiate into the osteogenic lineage could participate in HO formation. However, the terminology and the hierarchical relationship among subpopulations of mesenchymal progenitor cells are still not well defined, and currently there is no practical way to comprehensively isolate or label all such subpopulations. Multiple potential candidate populations have been proposed. Our genetic lineage tracing experiments virtually excluded most major proposed candidate populations, including hematopoietic, myogenic, and somitogenic populations. Other potential populations, such as inducible osteogenic precursors proposed by Friedenstein and co-worker [39] or pericytes by Diaz-Flores et al. [40] could also be receptive populations that turn into bone in response to high BMP signaling. In addition, we speculate that multipotent mesenchymal stromal cells/mesenchymal stem cells (MSC) could be the receptive cells that form HO. Several characteristics of MSC, including their multipotentiality and expression of BMPs and BMP receptors (especially ACVR1), make MSC-like cells an attractive candidate as the pivotal component in the HO cascade. Our unpublished data indicated that bone marrow-derived MSC are capable of contributing to HO formation from early to later stages (data not shown), which suggests that local populations of MSC-like cells could be the receptive cell population in this model. Consistent with this hypothesis, radiation therapy, which would reduce or eliminate local MSC, has been successfully used to prevent HO after traumatic events [41]. Unfortunately, radiation therapy is not commonly used in FOP patients due to potentially severe side effects [42]. However, the possibility that MSC from different tissues may differ in their phenotype and functional properties [43] complicates the interpretation of these data. Overall, the emerging picture from all available data suggests that receptive cells are local inducible stem/progenitor cells, consistent with Friedenstein's proposal [39].

In summary, our current study supports the following working model. Tissue injury leads to local inflammation and macrophage accumulation, which in turn leads to accumulation of osteogenic factors including BMPs. This likely dysregulates local stem/progenitor cells and stimulates them to follow osteogenic pathways with HO formation. In acquired HO the process stops at this point. Interestingly, acquired HO can sometimes spontaneously resolve, most likely due to resolution of the inflammatory response and subsequent decrease in BMP activity. However, in patients with FOP, paracrine effects and abnormal crosstalk between the adaptive immune system and the HO is sufficient to keep the pathological process spreading (Fig. 5). Overall, our observations indicate that macrophage responses to tissue injury stimulate local inducible stem/progenitor cells to differentiate into bone, suggesting that dysregulation of these cells could be a common cellular mechanism for typical HO.