Animal models of Fibrodysplasia Ossificans Progressiva

1. Introduction to FOP

Fibrodysplasia Ossificans Progressiva (FOP) is a rare human skeletal disease classically characterized by the widespread, progressive, and irreversible formation of heterotopic ossification (HO) (Kaplan et al., 1993; Kaplan et al., 2008; Kaplan et al., 2012; Kartal-Kaess et al., 2010; Pignolo et al., 2011). Heterotopic ossification (HO) is the process of bone formation outside of the normal skeleton (McCarthy and Sundaram, 2005). FOP patients develop HO, and a number of additional phenotypes, due to mutations in a single gene, ACVR1 (Kaplan et al., 2009; Shore et al., 2006). These mutations result in constitutive activation of ACVR1, a type I TGFβ/BMP receptor, and over activation of BMP signal transduction pathways (Billings et al., 2008; Culbert et al., 2014; Fukuda et al., 2009; Haupt et al., 2014; van Dinther et al., 2010).

1.1 Pathology of FOP

There are two features of FOP in humans that are considered to be classical hallmarks of the disease: malformation of the big toes and progressive HO formation (Kaplan et al., 2008; Kaplan et al., 2009; Pignolo et al., 2011; Shore et al., 2006). Nearly all patients with FOP present with big toe malformations at birth (Kaplan et al., 2009; Pignolo et al., 2011; Shore et al., 2006). As they get older, patients begin to develop HO lesions within fibrous tissues, including skeletal muscle, ligaments, and tendons (Figure 1A) (Kaplan et al., 2008; Kaplan et al., 2009; Pignolo et al., 2011). Notably, the diaphragm, tongue, and cardiac and smooth muscle remain untouched by FOP lesions (Pignolo et al., 2011).

Figure 1. FOP in animal models.

Visualization of the classic FOP phenotype, heterotopic ossification, in (A) human (Kaplan et al., 2009), (B) mouse conditional-on knock-in ACVR1^R206H model (Hatsell et al., 2015), (C) zebrafish heat-shock-inducible Acvr1l^Q204D model, (D) cat (Yabuzoe et al., 2009), and (E) whale (La Sala et al., 2012).

The majority of FOP patients form HO lesions within the first decade of life (Cohen et al., 1993; Pignolo et al., 2016; Rocke et al., 1994). Lesion formation typically begins in axial and proximal regions, such as the neck and upper back, then radiates into appendicular and distal regions during adolescence and adulthood (Pignolo et al., 2016). Nearly all patients experience progressively limited mobility into adulthood due to HO formation and joint ankylosis (Pignolo et al., 2016). The median age of survival is 40 years and the most common cause of death in patients is thoracic insufficiency syndrome (Kaplan et al., 2010). While the trend of disease manifestation is progressive in all FOP patients, the rate of progression can vary quite dramatically among patients, with a subset of individuals showing slowed or minimal disease progression and extended lifespans (Pignolo et al., 2016).

A number of other phenotypes associated with FOP are common but highly variable. These include osteochondromas (>90% of patients), cervical spine malformations (>80% of patients), chronic neurological symptoms (~50% of patients), and thumb malformations (~50% of patients) (Kaplan et al., 2009; Kitterman et al., 2012). Among individual cases, patients have been documented with atypical characteristics such as loss of digits, growth retardation, aplastic anemia, cataracts, and retinal detachment (Kaplan et al., 2009).

1.2 Genetics of FOP

FOP is estimated to affect about 3,500 people worldwide, or 1 in every 2 million live births, with 685 classically-affected patients identified as of 2016 (Pignolo et al., 2016). The disease predominantly arises through sporadic (noninherited) mutation, though it can also be inherited in an autosomal dominant fashion (Kaplan et al., 1993; Kaplan et al., 2008; Kaplan et al., 2012; Kartal-Kaess et al., 2010; Pignolo et al., 2011). All individuals with FOP carry an activating mutation in one copy of the gene encoding ACVR1 (also known as ALK2) (Kaplan et al., 2009; Shore et al., 2006). Nearly 90% of cases display the classic p.R206H mutation (Hildebrand et al., 2016; Kaplan et al., 2009). The remaining ~10% of patients carry atypical mutations that often result in a number of the rare or atypical phenotypes described above (Kaplan et al., 2009; Pignolo et al., 2011).

1.3 ACVR1, a Type I TGFβ/BMP receptor

ACVR1 is one of the seven Type I TGFβ/BMP receptor family members (Derynck and Zhang, 2003; Massagué, 1998; Mueller, 2015). In TGFβ/BMP signal transduction, a dimer of one of the more than 30 TGFβ superfamily of extracellular ligands binds to a Type II receptor dimer, then recruits a type one receptor dimer to complete a tetrameric receptor complex (Mueller, 2015). The ligand-receptor complex is activated by transphosphorylation of the Type I receptor intracellular GS domains by the Type II receptors (Derynck and Zhang, 2003; Massagué, 1998; Mueller, 2015). The activated Type I receptors can then phosphorylate their downstream cytoplasmic Smad signaling partners, which include the regulatory Smads (R-Smads), Smad1, Smad2, Smad3, Smad5, Smad7, Smad8 and Smad9 (Derynck and Zhang, 2003; Massagué, 1998; Mueller, 2015). The phosphorylated R-Smads then associate with the Co-Smad, Smad4, and subsequently translocate to the nucleus where they can act as transcription factor complexes to mediate gene expression (Derynck and Zhang, 2003; Massagué, 1998; Mueller, 2015).

TGFβ/BMP signaling pathways are involved in a wide range of cellular and developmental processes, including early embryonic axis formation, germ-layer specification, gastrulation, left-right asymmetry, and organogenesis [Reviewed in (Wu and Hill, 2009)]. Endochondral ossification, one of the natural pathways for bone formation, is uniquely regulated by the BMP and TGFβ signaling pathways (Wu et al., 2016). ACVR1 associates with a specific subset of ligands, Type II receptors, and downstream Smad signaling partners to propagate BMP signaling that promotes endochondral ossification (Figure 2A). ACVR1 forms tetrameric receptor complexes with ActRII and ActRIIB (Macías-Silva et al., 1998; Massagué, 1998). BMP6 and BMP7 display the strongest ligand binding interactions with the ActRII/ACVR1 receptor complex (Macías-Silva et al., 1998; Rahman et al., 2015), while BMP2 and BMP4 can bind to the receptor complex to a lesser degree (Hatsell et al., 2015; Massagué, 1998; Rahman et al., 2015). Once formed, the ligand-receptor complex signals through Smad1/5/8 to promote the gene expression required for endochondral ossification (Chen and Massagué, 1999; Macías-Silva et al., 1998). Activin A, an important activator of TGFβ-mediated inflammation (Figure 2A) (Hedger et al., 2011), is another ligand that is capable of binding to the ActRII/ACVR1 receptor complex, but instead acts as a competitive inhibitor of ACVR1-mediated signaling (Hatsell et al., 2015; Olsen et al., 2015). This inhibition may be important for blocking undesired endochondral ossification during a normal injury response (Figure 2A).

Figure 2. Constitutive activation of ACVR1 promotes aberrant BMP signaling.

(A) Diagram of signaling through ACVR1 leading to HO development. (B) Comparison of the Glycine-Serine (GS) domain of each ACVR1 homolog, with human FOP variants p.R206H and p.Q207E noted, as well as mouse p.Q207D, p.R206H, and zebrafish p.Q204D, each used in respective FOP models.

Mutations found in ACVR1 that are associated with FOP all cause abnormal constitutive activation of BMP signaling. The classical p.R206H mutation, as well as the known variant p.Q207E and the artificial mutation p.Q207D, are all located within the GS domain of ACVR1 (Figure 2B) (Haupt et al., 2014; Kaplan et al., 2009; Yu et al., 2008). A number of other FOP-associated variant mutations are found in the both the GS domain and the kinase domain (Kaplan et al., 2009; Rahman et al., 2015). Each of these mutations confers a structural change to the receptor conformation, resulting in both altered ligand specificity (Hatsell et al., 2015; Haupt et al., 2014; Hildebrand et al., 2016; Hino et al., 2015) and constitutively active ligand-independent signaling (Bagarova et al., 2013; Groppe et al., 2007; Le and Wharton, 2012). Specifically, Activin A switches roles from a competitive inhibitor to an activator of ACVR1^R206H-mediated BMP signaling (Figure 2A) (Hatsell et al., 2015; Hino et al., 2015). This aberrant signal activation has been directly linked to the endochondral ossification seen in HO lesions in FOP mouse models (Bagarova et al., 2013; Chakkalakal et al., 2012; Hatsell et al., 2015).

1.4 FOP Lesion Progression

In FOP, lesions are formed in what are referred to as ‘flare-ups’ (Shore and Kaplan, 2010). A flare-up is described as a two-stage process of explosive inflammation leading to fibrous tissue degradation and the subsequent endochondral ossification that occurs in the wake of the tissue destruction (Shore and Kaplan, 2010). Physical injury, viral infection, surgical procedures, and intramuscular injections, including immunizations, can all prompt the initial inflammation of a flare-up (Lanchoney et al., 1995; Pignolo et al., 2011; Scarlett et al., 2004). During this initial phase, inflammatory cells, including lymphocytes, macrophages, and mast cells, rush to the site of the lesion (Gannon et al., 1998; Gannon et al., 2001; Kaplan et al., 2005). While this occurs in normal tissues as well, the cellular response is amplified in FOP. The inflammatory cells drive the degradation of the damaged muscle tissue, and appear to also play a critical role in producing angiogenic factors and recruiting fibroproliferative cells (Gannon et al., 1998; Gannon et al., 2001; Kaplan et al., 2005). Interestingly, immunosuppression in mouse models of FOP and in a human FOP patient resulted in abolition of HO formation, suggesting an indispensible role for the inflammatory phase in lesion development (Kan et al., 2009; Kaplan et al., 2007).

During the second phase of HO formation, a resident pool of mesenchymal stem cells at the site of inflammation and degradation undergoes the BMP-mediated process of endochondral ossification to generate heterotopic bone (Shore and Kaplan, 2010). The mesenchymal stem cell population is believed to have numerous origins, including but not limited to Scx⁺ tendon progenitor cells (Dey et al., 2016), bone-marrow-derived muscle-resident Mx1⁺ cells (Dey et al., 2016), Glast-expressing cells (Kan et al., 2013), Tie2⁺ endothelial cells (Lounev et al., 2009; Medici et al., 2010; Wosczyna et al., 2012), and circulating osteogenic precursor (COP) cells (Suda et al., 2009).

1.5 Current Treatments

There is no cure for FOP. Prenatal testing for the disease is not currently available and, given the rarity of the disease, is not likely to become routine (Pignolo et al., 2011). In addition, the disease rarity and variability in the severity of phenotypes and the rate of disease progression often result in delayed or misdiagnosis and contraindicated exacerbating medical procedures, such as surgical excisions (Kaissi et al., 2016; Pignolo et al., 2011). Once a diagnosis of FOP has been made, historically, treatment has been palliative, involving drug therapies such as glucocorticoids to decrease inflammation and NSAIDs to alleviate pain (Kaplan et al., 2008; Pignolo et al., 2011). Surgical procedures must be avoided as they provoke additional HO formation (Pignolo et al., 2011).

Nearly 40 years ago, Pacifici and colleagues demonstrated that retinoic acid signaling acts as a strong and specific inhibitor of chondrogenesis (Pacifici et al., 1980). More recently, researchers have demonstrated that activation of retinoic acid signaling by treatment with the retinoic acid receptor gamma (RARγ) agonist palovarotene can prevent HO formation in FOP mouse models (Chakkalakal et al., 2016; Shimono et al., 2011). Based on these results, palovarotene has been moved into an NIH-supported clinical trial to assess its ability to prevent HO formation in human FOP patients during and following flare-ups. Preliminary data from Phase 2 of the clinical trial show promising results: palovarotene treatment reduced new HO incidence in FOP patients by 50% and new HO volume by 70% (http://www.prnewswire.com/news-releases/phase-2-part-a-open-label-extension-trial-of-palovarotene-for-treatment-of-patients-with-fibrodysplasia-ossificans-progressiva-continues-positive-trends-300429975.html).

2. Animal Models of FOP

There is a fundamental need for animal models of FOP in which to study mechanisms of disease progression given that investigative surgical procedures in human FOP patients exacerbates HO formation (Pignolo et al., 2011). Spontaneous cases of FOP have been described in many animals, including numerous instances in cats (Figure 1D) (Asano et al., 2006; Klang et al., 2013; Valentine et al., 1992; Warren and Carpenter, 1984; Yabuzoe et al., 2009), dogs (Guilliard, 2001), pigs (Seibold and Davis, 1967), and even a southern right whale (Figure 1E) (La Sala et al., 2012). In each of these cases, specimens presented with progressive and substantial HO development, and in some cases osteochondromas and vertebral fusions. Nearly all of these cases were described prior to the identification of ACVR1 as the causative gene for human FOP, so none have been corroborated with molecular data to confirm the presence of activating mutations in respective ACVR1 orthologs in these animals.

ACVR1 protein orthologs can be found as early in evolutionary history as the invertebrate class Insecta, including the D. melanogaster (fruit fly) protein Saxophone (Brummel et al., 1994; Twombly et al., 2009). Among vertebrates, human ACVR1 and mouse ACVR1 are 98% identical (Kaplan et al., 2009), while human ACVR1 and zebrafish Acvr1l are 69% identical (Yelick et al., 1998). Of note, the intracellular domains of ACVR1, the glycine-serine (GS) domain and the kinase domain, are 85% identical between human ACVR1 and zebrafish Acvr1l (Yelick et al., 1998). This conservation suggests preservation of protein function and provides validation for the study of ACVR1 function and dysfunction in animal models.

2.1 Fruit fly model

The fruit fly ortholog of human ACVR1, Saxophone (Sax), was first characterized in 1994 in a series of studies to identify receptors responsive to the fruit fly BMP-2/BMP-4 ortholog Decapentaplegic (Dpp) (Brummel et al., 1994; Nellen et al., 1994; Penton et al., 1994; Xie et al., 1994). Signaling through Sax is critical for dorsal ectodermal patterning (Brummel et al., 1994; Ferguson and Anderson, 1992; Nellen et al., 1994; Neul and Ferguson, 1998; Penton et al., 1994; Wharton et al., 1993; Xie et al., 1994) and for imaginal disk development leading to adult appendage formation (Bangi and Wharton, 2006; Brummel et al., 1994; Nellen et al., 1994; Penton et al., 1994; Spencer et al., 1982). The TGFβ/BMP signaling pathway is so evolutionarily well-conserved that human BMP ligands can serve as functional substitutes for their fruit fly counterparts (Brummel et al., 1994; Padgett et al., 1993). Thus researchers have reasoned that studying the basic mechanisms behind TGFβ/BMP signaling in development in the fruit fly could have important implications in vertebrate systems.

Since the discovery of ACVR1 as the causative gene for human FOP, researchers have focused their work in fruit flies on the importance of the formation of tetrameric BMP receptor complexes. Fruit flies carrying the gain of function allele p.G412E in Sax are maternal effect lethal in a dosage dependent manner that suggests Sax^G412E participates in signaling complexes that promote over active BMP signaling during development (Twombly et al., 2009). Mutations in this same residue have been found in FOP patients (Kaplan et al., 2009). Indeed, when fruit flies carry the classical p.R206H FOP mutation, over activation of BMP signaling is dependent on type II receptor association, and is ligand-independent (Le and Wharton, 2012).

2.2 Mouse models

Mouse ACVR1 (previously known as Tsk 7L, Alk2, and ActR-I) was first cloned in 1993, and characterized for its ability to associate with Type II TGFβ receptors and bind a number of TGFβ family ligands, including TGFβ itself and activin, in vitro (Ebner et al., 1993). Studies of ACVR1-mediated BMP signaling in the mouse have revealed roles in embryonic gastrulation (Beppu et al., 2000; Gu et al., 1999), neural crest cell differentiation (Dudas et al., 2004; Kaartinen et al., 2004), germ cell development (de Sousa Lopes et al., 2004), and lens development (de Iongh et al., 2004; Rajagopal et al., 2008). Given that ACVR1 is indispensible for so many aspects of mouse embryonic development, and that loss or over activation of ACVR1 causes embryonic or perinatal lethality (Beppu et al., 2000; Chakkalakal et al., 2012; Gu et al., 1999), subsequent work on adult mouse models of FOP have focused on the use of chimeric/mosaic animals, or have taken advantage of various methods of conditional gene expression, most commonly using Cre-Lox recombination.

2.2.1 ACVR1^Q207D Models

The first mouse model expressing constitutively active ACVR1, through Cre-Lox-inducible overexpression ACVR1^Q207D, was generated and studied prior to the discovery of ACVR1 as the causative gene for FOP (Fukuda et al., 2006). Although it is now known that the p.Q207D mutation in ACVR1 is not a naturally occurring mutation found in human FOP patients, this mutation does confer constitutive activation of the ACVR1 receptor (Wieser et al., 1995), similar to, though more severe than, the FOP-associated mutations p.R206H and p.Q207E (Haupt et al., 2014). Through these experiments, it was determined that even mild, ubiquitous overexpression of ACVR1^Q207D promotes excessive BMP signaling and compromises embryonic development (Fukuda et al., 2006). The authors did not investigate the role of ACVR1^Q207D in later development at that time.

Since the discovery that all FOP patients carry constitutively activating mutations in ACVR1, a number of groups have used the ACVR1^Q207D-expressing mouse to model FOP. Yu and colleagues bypassed the embryonic lethality of the original model through the use of two methods: (1) directed injection of adenoviral Cre recombinase into the left hindlimb to induce localized ACVR1^Q207D expression or (2) mating with animals carrying tamoxifen-inducible Cre-recombinase to induce global ACVR1^Q207D expression (Yu et al., 2008). In the localized hindlimb model, ACVR1^Q207D-expressing mice developed decreased mobility and bony calluses at the site of injection within 30 days. In contrast, mice globally expressing ACVR1^Q207D did not develop detectable HO within 60 days of tamoxifen induction. Interestingly, these animals did develop bony calluses as a result of control adenoviral injection after tamoxifen induction, indicating that inflammation or injury is required to induce heterotopic ossification in the ACVR1^Q207D background (Yu et al., 2008).

Continued work with the ACVR1^Q207D mouse model confirmed previous results in the fruit fly model (Le and Wharton, 2012; Twombly et al., 2009), demonstrating that ACVR1^Q207D requires complex formation with Type II receptors in order to promote ligand-independent BMP signal transduction (Bagarova et al., 2013). Furthermore, the first in vivo assessments of the HO-inhibiting potential of the RARγ agonist, Palovarotene, were conducted in the ACVR1^Q207D mouse model (Shimono et al., 2011). Researchers found that delivery of Palovarotene protected against the formation of HO by reducing the overactivation of BMP signaling caused by ACVR1 constitutive activation (Shimono et al., 2011). The ACVR1^Q207D mouse model has also recently been used to establish an indispensible role for HIF-1α-mediated hypoxia in the early stages of HO lesion formation, identifying an additional therapeutic target for FOP (Wang et al., 2016).

2.2.2 ACVR1^R206H Models

In recent years, researchers have endeavored to generate a mouse model harboring the classical FOP mutation, p.R206H. Chakkalakal and colleagues were the first to generate such a model, using homologous recombination to introduce the R206H mutation at the endogenous murine Acvr1 locus (Chakkalakal et al., 2012). As endogenous expression of ACVR1^R206H in mice causes perinatal lethality, chimeric animals with 70–90% mutant cells were used to study ACVR1^R206H expression in adult mice. The chimeric mice develop classic FOP features, including hind limb digit malformation, joint fusions, and extensive HO development. In addition, the chimeric mice form substantial HO in response to muscle injury (Chakkalakal et al., 2012).

Experiments have also been conducted using mesenchymal progenitor cell lines established from chimeric ACVR1^R206H/+ mice (Culbert et al., 2014). Using these cells, researchers have demonstrated that ACVR1^R206H is necessary, though not sufficient on its own, to promote chondrogenesis in vitro, and HO formation in vivo (Culbert et al., 2014). BMP stimulation is required to provide the initial activation of ACVR1^R206H, suggesting that ACVR1^R206H/+ cells exhibit an initial ligand dependence to promote BMP signaling, that subsequently switches to ligand-independent constitutive activation after stimulation (Culbert et al., 2014).

Another FOP mouse model, this time utilizing a Cre-dependent conditional-on knock-in system to express the Acvr1^R206H allele at the endogenous locus, was created to bypass the perinatal lethality of the first ACVR1^R206H mouse model (Hatsell et al., 2015). Using the conditional-on ACVR1^R206H model, Hatsell and colleagues demonstrated the progressive development of HO at many of the same sites commonly affected in human FOP patients (Figure 1B). Importantly, this model was used to identify a role for activin A in the aberrant activation of ACVR1^R206H. While activin A normally acts as an inhibitor of ACVR1, the p.R206H mutation confers a gain-of-function activity that permits activin A-mediated activation of ACVR1. Activin A delivery in the conditional-on ACVR1^R206H mouse leads to HO development, an effect that can be abolished by activin A-blocking antibodies (Hatsell et al., 2015). This novel role for activin a in the activation of ACVR1^R206H was also observed in human FOP patient-derived induced mesenchymal stromal cells (Hino et al., 2015).

Further studies utilizing the conditional-on ACVR1^R206H mouse model have investigated therapeutic approaches for FOP (Chakkalakal et al., 2016), and have generated the first natural history of developing HO lesions (Upadhyay et al., 2017). Following the finding that the RARγ agonist, Palovarotene, could inhibit HO development in the ACVR1^Q207D mouse model (Shimono et al., 2011), Chakkalakal and colleagues sought to demonstrate the same potent inhibition in mice carrying the FOP-associated mutation, p.R206H. They showed that prenatal delivery of Palovarotene to ACVR1^R206H-expressing mice not only reduced HO lesion formation, but also rescued malformations of the skeleton (Chakkalakal et al., 2016). Very recently, Upadhyay and colleagues used a combination of MRI, Micro-CT, and PET/CT imaging techniques to track the development and progression of HO lesions in the conditional-on ACVR1^R206H mouse model (Upadhyay et al., 2017). These tracking modalities demonstrated that activin A, the aberrant activator of ACVR1^R206H, is required not only for the formation of new HO lesions, but also for the sustained growth and expansion of existing HO lesions (Upadhyay et al., 2017).

2.2.3 Progenitor cell populations

An ongoing goal in the FOP research field has been to identify the progenitor cell populations contributing to the development of heterotopic ossification. Mouse models have been indispensible for progress towards this goal. Several studies have used non-ACVR1-mediated mouse models of HO development, including transgenic BMP4 overexpression (Kan et al., 2009; Kan et al., 2013; Lounev et al., 2009; Suda et al., 2009), Matrigel delivery of BMP2 (Lounev et al., 2009; Wosczyna et al., 2012), and BMP2-loaded collagen scaffold delivery coupled with cardiotoxin injection injury (Agarwal et al., 2017). Many candidate cell populations were excluded as HO-contributing cells through these experiments, such as FoxD1+ mesenchymal cells (Kan et al., 2013), CD19+ B-cells, LCK+ T-cells, Lyz+ neutrophils, Myf5+ myoblasts, and nestin+ nerve progenitor cells (Kan et al., 2009). The studies did identify a number of contributing progenitor cell populations however, namely Tie2+ endothelial cells (Lounev et al., 2009; Wosczyna et al., 2012), circulating osteogenic progenitor (COP) cells (Suda et al., 2009), Glast-expressing cells (Kan et al., 2013), and tendon-derived Scx+ cells (Agarwal et al., 2017).

In addition, lineage-tracing studies have been conducted in the constitutively active ACVR1 mouse models to identify progenitor cells contributing to HO. Tie2+ cell contributions to HO lesions were confirmed in both the Acvr1^Q207D mouse model (Medici et al., 2010) and in the Acvr1^R206H/+ chimeric mouse model (Chakkalakal et al., 2012). Scx+ cells and bone-marrow-derived muscle-resident Mx1+ cells were identified as HO contributing populations in the conditional-on knock-in Acvr1^R206H mouse model (Dey et al., 2016).

2.3 Embryonic chicken model

After identification of an FOP patient carrying the variant FOP mutation, p.Q207E, who displayed only classical FOP phenotypes, researchers chose to study the role of this mutation, as well as the classical p.R206H mutation and the engineered p.Q207D mutation, in early limb development (Haupt et al., 2014). All three variants are closely positioned in the GS domain of ACVR1, suggesting that they might cause similar constitutive activation of ACVR1. Overexpression of ACVR1^R206H and ACVR1^Q207E in chick limbs caused FOP-like phenotypes, including joint fusions and ectopic cartilage formation (Haupt et al., 2014). Overexpression of ACVR1^Q207D strongly exacerbated the FOP-like phenotypes, causing severe skeletal malformations and massive ectopic cartilage lesion formation (Haupt et al., 2014). This work provided crucial evidence that the engineered p.Q207D is biochemically distinct from the FOP-associated variants and only marginally appropriate for modeling FOP in animals.

2.4.1 Embryonic zebrafish models

Since its discovery in 1998, zebrafish Acvr1l (previously known as Alk8) has largely been studied for its important role in establishing the dorsoventral axis in the developing embryo (Bauer et al., 2001; Mintzer et al., 2001; Payne et al., 2001; Yelick et al., 1998). Expression of kinase-mutated or truncated loss-of-function forms of Acvr1l in the zebrafish embryo result in dorsalization phenotypes, such as loss of ventral tail structures and tail shortening (Bauer et al., 2001; Mintzer et al., 2001; Payne et al., 2001). Alternatively, embryonic expression of constitutively active Acvr1l, in the form of Acvr1l^Q204D, leads to ventralized features that include loss of anterior structures and expansion of ventral mesodermal tissues (Bauer et al., 2001; Payne et al., 2001). Both dorsalization and ventralization ultimately result in embryonic death (Bauer et al., 2001; Mintzer et al., 2001; Payne et al., 2001).

Following the identification of the FOP-associated p.R206H mutation, experiments were conducted in zebrafish embryos to determine whether overexpression of human ACVR1^R206H caused developmental defects similar to those of Acvr1l^Q204D (Shen et al., 2009). Indeed, overexpression of human ACVR1^R206H resulted in strong ventralization of zebrafish embryos, an effect that required functional downstream BMP/Smad signal transduction, but not BMP2 or BMP7 ligand stimulation (Shen et al., 2009).

2.4.2 Adult zebrafish model

Recently, researchers have endeavored to generate the first adult zebrafish model of FOP. Given that embryonic expression of both zebrafish Acvr1l^Q204D and human ACVR1^R206H cause lethality in zebrafish (Bauer et al., 2001; Payne et al., 2001; Shen et al., 2009), the use of a conditional gene expression system was imperative to establish such a model. Transgenic zebrafish were created that express heat-shock-inducible Acvr1l^Q204D, and were exposed to heat-shock to induce Acvr1l^Q204D expression at 14 days post fertilization, allowing them to exhibit normal embryonic patterning (LaBonty et al., 2017). Adult zebrafish expressing Acvr1l^Q204D develop a number of FOP-like features, including HO (Figure 1C), spinal lordosis, vertebral fusions, and fin malformations, validating this new animal model for human FOP (LaBonty et al., 2017).

3. Advantages of Adult Zebrafish as a Model for Human FOP

Recent comprehensive natural history studies in human FOP patients have indicated that clinical progression of the disease does not necessarily require a classical flare-up (Pignolo et al., 2016). This finding strengthens the need to continue studies of basic developmental pathways of FOP in animal models in order to elucidate all of the molecular pathways mediating, and contributing to, the progression of FOP. Many of these basic research questions can be answered by taking advantage of the tractability of the zebrafish model. Numerous advantages of the zebrafish model system are described below.

3.1 Basic biology

Many benefits of the zebrafish as a model for human disease derive from the basic biology of the animal. A pair of adult zebrafish can produce individual clutches of ~200–300 embryos per week, which undergo synchronous, ex-utero development and are optically transparent. These embryos are small enough to fit into 96-well plates, but large enough for manipulations such as microinjection. Embryonic development is rapid and has been extensively studied and documented (Kimmel et al., 1995). Zebrafish reach sexual maturity around 3 months post fertilization and many remain fertile for well over a year. Adult zebrafish develop many of the same organs, which function in nearly the same capacity, as other vertebrates, including humans. For example, zebrafish have fully functional, well conserved innate and adaptive immune systems (Meeker and Trede, 2008), which are crucial in order to study the integral role of the immune response in the early initiation of HO formation in FOP (Kaplan et al., 2015; Upadhyay et al., 2017).

3.2 Genetic manipulation for modifier screens

The fully sequenced zebrafish genome offers evidence for the existence of more than 26,000 protein-coding genes, approximately 70% of which are orthologs of human genes (Howe et al., 2013). This high degree of conservation at the nucleotide sequence level, when combined with the ease of genetic manipulation by chemical mutagenesis (Solnica-Krezel et al., 1994), transgenesis (Kawakami, 2004), and CRISPR/Cas9 editing (Albadri et al., 2017), makes the zebrafish a particularly useful model in which to conduct relatively rapid genetic modifier screens for FOP. Human FOP patients display variability in the rate and severity of disease progression (Kaplan et al., 2017; Pignolo et al., 2011) and this variability may be caused in part by genetic mutations outside of the ACVR1 locus, which may either suppress or exacerbate aberrant signal activation and alter FOP disease progression. Candidate gene mutations could be easily introduced into the zebrafish genome and assessed for their ability to alter the severity of phenotypes in the zebrafish FOP model.

3.3 Reporter lines for lineage tracing

Zebrafish are uniquely amenable to live cell lineage tracing due to their transparency, which can even be maintained in adults through the use of casper mutants (White et al., 2008). In addition, there exists a wide variety of currently available fluorescent transgenic reporter lines, including many promoter-driven cell-type specific reporter lines, to study FOP. Finally, a number of conditional gene expression systems have been adapted for use in zebrafish, such as heat shock induction (Shoji and Sato-Maeda, 2008), Cre/lox recombination (Felker and Mosimann, 2016), and photoconversion (Dempsey et al., 2012). Many if not most of these transgenic lines can be easily obtained from the NIH-sponsored Zebrafish International Resource Center (ZIRC), or from individual labs. These resources can be used to validate existing identified progenitor cell populations contributing to human FOP, as well as identify novel contributing populations using the zebrafish FOP model.

3.4 High-throughput in vivo drug screening

Given their high fecundity and rapid embryonic development, zebrafish are the ideal in vivo FOP model system to validate therapeutic compounds identified through in vitro tissue culture assays. In addition, high-throughput screening of novel small molecules and antibodies can be conducted in zebrafish – a benefit exclusive to zebrafish among the existing vertebrate models of human FOP (MacRae and Peterson, 2015). Such compound screens could help identify additional key regulators of HO initiation and progression in FOP, providing new inroads to novel and effective therapies to treat human FOP. Typically, compounds identified through in vivo screens in whole animal models perform better in human clinical trials.

4. Conclusions and Future Directions

The variety of animal models that have been generated for the study of FOP have already been used to provide an incredible wealth of information about the basic mechanisms driving FOP disease progression and potential therapeutic advances. We anticipate that future studies will elucidate the progenitor cell type(s) contributing to HO in FOP patients, and identify new therapeutic targets. Given the variability in FOP disease progression, we also expect that personalized medicine approaches will be designed to more effectively treat individual patients. These therapies may target and regulate multiple mechanisms of BMP signal activation, based on a patient’s individual genetic profile. These scientific questions can be quickly and safely probed first in animal models for FOP, to benefit all of the present and future human FOP patients.