Fibrodysplasia ossificans progressiva mutant ACVR1 signals by multiple modalities in the developing zebrafish

Fibrodysplasia ossificans progressiva (FOP) is a rare human genetic disorder characterized by altered skeletal development and extraskeletal ossification. All cases of FOP are caused by activating mutations in the type I BMP/TGFβ cell surface receptor ACVR1, which over-activates signaling through phospho-Smad1/5 (pSmad1/5). To investigate the mechanism by which FOP-ACVR1 enhances pSmad1/5 activation, we used zebrafish embryonic dorsoventral (DV) patterning as an assay for BMP signaling. We determined that the FOP mutants ACVR1-R206H and -G328R do not require their ligand binding domain to over-activate BMP signaling in DV patterning. However, intact ACVR1-R206H has the ability to respond to both Bmp7 and Activin A ligands. Additionally, BMPR1, a type I BMP receptor normally required for BMP-mediated patterning of the embryo, is dispensable for both ligand-independent signaling pathway activation and ligand-responsive signaling hyperactivation by ACVR1-R206H. These results demonstrate that FOP-ACVR1 is not constrained by the same receptor/ligand partner requirements as WT-ACVR1.


Introduction
Fibrodysplasia ossificans progressiva (FOP) is a rare human genetic disorder characterized by altered skeletal development and extraskeletal bone formation. Patients with FOP have pathognomonic malformation of the great toes and progressive spontaneous and injury-induced heterotopic ossification (HO) that eventually leads to loss of mobility Cohen et al., 1993;Connor and Evans, 1982). Most cases of FOP are caused by a single amino acid substitution, R206H, in the type I BMP/TGFb cell surface receptor ACVR1 (also known as ALK2), which over-activates signaling through phospho-Smad1/5 (pSmad1/5) (Shore et al., 2006;Shen et al., 2009;van Dinther et al., 2010;Fukuda et al., 2009). A small subset of patients with variant presentation of the classical FOP phenotype have distinct activating mutations in the ACVR1 gene, including the substitution G328R Haupt et al., 2018). While the phenotypic consequences of increased ACVR1 signaling have been well characterized in both patients and animal models (Casal et al., 2019;Pignolo et al., 2011;Chakkalakal and Shore, 2019), the mechanism by which ACVR1 mutations lead to over-active signaling is less well understood.
ACVR1 and its signaling partners belong to the transforming growth factor beta (TGFb) superfamily. The activity of ACVR1 is critical to several developmental processes including embryonic patterning and skeletal formation (Derynck and Akhurst, 2007;Little and Mullins, 2006). In the presence of ligand, ACVR1 complexes with other BMP receptors to signal. In zebrafish, Acvr1l (also known as Alk8), the zebrafish paralog to human and mouse ACVR1 (Yelick et al., 1998), forms a tetrameric receptor complex with one other type I BMP receptor, Bmpr1a or Bmpr1b (also known as Alk3 and Alk6, respectively), and two type II BMP receptors (Little and Mullins, 2006;Ehrlich et al., 2011;Figure 1. ACVR1-R206H and -G328R over-activate BMP signaling in the absence of an intact ligand-binding domain. (a) Embryos expressing ACVR1 HA-tagged constructs were immunostained for HA at the shield stage (early gastrula); ACVR1 domain schematics of the constructs are above each set of corresponding images. Nuclei (yellow), bcatenin (cyan), ACVR1-HA (magenta). Scale bars = 40 mm. DAcvr1 constructs lack residues 35-100, which contain the cysteine-rich ligand binding motifs. Uninjected (N = 6) and embryos injected with: Acvr1 (N = 6), DAcvr1 (N = 5), Acvr1-R206H (N = 9), DAcvr1-R206H (N = 4), Acvr1-G328R (N = 5), and DAcvr1-G328R (N = 5). (b) BMP signaling dose-dependent 12-30 hpf zebrafish embryo phenotypes: severe dorsalization (C5-C4, dark blue), mild dorsalization (C3-C1, light blue), wild-type development (WT, violet), mild ventralization (V1-V3, light orange), and severe ventralization (V4-V5, dark orange). (c) Bubble plot circle sizes correspond to the percent of total embryos within a condition that are in a particular phenotypic category. Examples are shown for: 100%, 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20% and 10%. (d) Representative phenotypes of wild-type embryos (evaluated at 12 and 30 hpf) and embryos injected with acvr1l MO (evaluated at 12 hpf; none survive to 30 hpf). (e-g) DV phenotypes of 12-30 hpf embryos injected with Acvr1 or DAcvr1 mRNA alone or together with acvr1l MO. Data are from four pooled experiments. Figure 1 continued on next page To investigate the FOP-ACVR1 signaling mechanism, we used a zebrafish DV patterning assay. For this assay, we microinjected one-cell stage wild-type or BMP component mutant zebrafish embryos to introduce mRNAs and/or anti-sense knockdown morpholinos that replicate null mutant phenotypes. The mRNAs and morpholinos were never mixed together, but instead each mRNA and morpholino was injected using separate needles, so that controls could be done to show the efficacy of each component on its own. Within one experiment, each mRNA or morpholino was always injected through the same calibrated needle on its own injection apparatus to ensure consistent amounts were injected among all the controls and experimental conditions.
To test if the FOP-ACVR1 receptor requires ligand binding to signal in a vertebrate animal model, we first compared the signaling function of the mouse wild-type ACVR1 to mouse ACVR1 lacking 65 residues containing the cysteine-rich ligand-binding motifs of the extracellular domain (DACVR1) (Haupt et al., 2014). We injected one-cell stage embryos with Acvr1 or DAcvr1 mRNA and tested if the exogenously expressed ACVR1 protein correctly localized to the cell membrane by immunostaining for the HA-epitope tag present on these receptors. We found that all the ACVR1 protein variants, regardless of their signaling activity, were expressed and localized to the cell membrane within the developing embryo ( Figure 1a).
To determine if the ligand-binding domain mutant DACVR1 receptor was sufficient to pattern the zebrafish embryo, we knocked down endogenous zebrafish Acvr1l with morpholinos (MOs), which replicate the phenotype of acvr1l null maternal-zygotic mutant embryos (Mintzer et al., 2001). Although we have previously generated maternal-zygotic acvr1l mutant embryos that lack both maternally and zygotically expressed acvr1l, are deficient in all BMP signaling, and exhibit a C5 severely dorsalized phenotype (Mintzer et al., 2001), generating these maternal-zygotic acvr1l mutant embryos required producing acvr1l homozygous mutant adults, an arduous accomplishment due to other requirements for Acvr1l later in development (Mintzer et al., 2001). Therefore, we used MO knockdown in the current study to deplete Acvr1l on a large scale and in combination with other mutant genotypes. As previously demonstrated and re-confirmed here, knockdown of Acvr1l generates the identical, severely dorsalized C5 phenotype as the maternal-zygotic acvr1l null mutant, consistent with the loss of all BMP pathway activity (Figure 1d Mintzer et al., 2001;Little and Mullins, 2009;Bauer et al., 2001). At 12 to 14 hr post fertilization (hpf), the C5 phenotype is characterized by an elongated body axis and by expansion of the normally dorsally-located somites around the circumference of the embryo. By 30 hpf, these embryos lyse due to presumptive pressure from the radialized somites. Importantly, injection of wild-type mouse Acvr1 mRNA can fully rescue this phenotype (Figure 1e,e', compare to d), demonstrating the specificity of the knockdown for Acvr1l. All these control experiments adhere to the recently published guidelines by leaders in the zebrafish field for the use of MOs and their validity to substitute for a mutant allele (Stainier et al., 2017). We next expressed Acvr1 or DAcvr1 mRNAs in acvr1l-knockdown (acvr1l-KD) embryos to test their function. We evaluated rescue of BMP signaling activity in these embryos by assaying for DV patterning phenotypes ( Figure 1b) and quantifying the proportion of embryos in each phenotypic category ( Figure 1c). Neither Acvr1 nor DAcvr1 mRNA perturbed normal development in the presence of endogenous Acvr1l (Figure 1e Columns: 1,N = 240;2,N = 121;3,N = 140;4,N = 153;5,N = 96;6,N = 118. (e') Representative 12 or 30 hpf phenotypes. (f) Acvr1-R206H or DAcvr1-R206H injected embryos. Columns: 1,N = 240;2,N = 70;3,N = 84;4,N = 153;5,N = 73;6,N = 91. (f') Representative 30 hpf phenotypes. (g) Acvr1-G328R or DAcvr1-G328R injected embryos. Columns: 1,N = 240;2,N = 88;3,N = 79;4,N = 153;5,N = 78;6,N = 71. (g') Representative 30 hpf phenotypes. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Injected embryo phenotype raw numbers for Figure 1e, f and g. Figure supplement 1. ACVR1-R206H and -G328R over-activate BMP signaling causing ventralization in the absence of an intact ligand-binding domain. Figure supplement 2. Ventralizing amounts of Acvr1-R206H and Acvr1-G328R are comparable to rescuing amounts of Acvr1.  These results in an in vivo vertebrate model support that ACVR1-R206H and ACVR1-G328R, unlike wild type ACVR1, exhibit ligand-independent signaling activity.
To test if our results could be influenced by differences in expression levels of injected mRNA, we determined the relative amounts of receptor in our injected embryos, using immunofluorescence to measure the cell surface expression of our Acvr1-HA-tagged mutant receptors in confocal sections of mid-gastrula embryos. Mid-gastrula embryos were collected for immunofluorescence and the remaining embryos were allowed to develop to 30 hpf for phenotyping. As described above, Acvr1, but not DAcvr1 could rescue Acvr1l-KD embryos, and all FOP mutant receptors ventralized embryos regardless of the presence of a ligand-binding domain (Figure 1-figure supplement 2). The amount of WT-Acvr1 that partially to fully rescued Acvr1l KD was determined per mm 2 of b-catenin fluorescence surface area (Figure 1-figure supplement 2). This receptor amount should reflect the amount of receptor expression that is required to pattern the zebrafish embryo. The amounts of Acvr1-R206H and Acvr1-G328R fluorescence that ventralized the embryo were similar to or significantly lower than WT-Acvr1, respectively. The levels of DAcvr1-G328R and DAcvr1-R206H fluorescence were also similar to or significantly lower than DAcvr1, respectively. These results show that the amount of Acvr1-R206H or -G328R required to ventralize the zebrafish embryo is similar to or lower than the amount of Acvr1 required to pattern the wild-type zebrafish, regardless of the presence of a ligand-binding domain. Altogether, the results indicate that the ventralization is not due to overexpression of the FOP mutant receptors.
DAcvr1 also did not perturb normal gradient formation in WT embryos, but could not rescue loss of pSmad1/5 in Acvr1l-KD embryos (Figure 2c,c'). There was no significant difference in mean pSmad1/5 fluorescence between Acvr1l-KD embryos and Acvr1l-KD embryos injected with DAcvr1, showing that WT-Acvr1 cannot signal through pSmad1/5 without a ligand-binding domain ( Figure 2h). In contrast, both Acvr1-R206H and DAcvr1-R206H restored pSmad1/5 signaling in Acvr1l-KD embryos and greatly expanded the signaling gradient dorsally (Figure 2d,d',e,e'). Likewise, Acvr1-G328R and DAcvr1-G328R rescued and expanded the pSmad1/5 gradient even when endogenous Acvr1l was absent (Figure 2f,f',g,g'). Both ACVR1-R206H and ACVR1-G328R significantly increased the mean pSmad1/5 signaling intensity in Acvr1l-KD embryos regardless of the presence of a ligand-binding domain ( Figure 2h). These results confirm that ACVR1-R206H and ACVR1-G328R have acquired the capacity to over-activate Smad1/5 phosphorylation even in the absence of an intact ligand-binding domain.
ACVR1-R206H and -G328R, but not DAcvr1-R206H, are responsive to Bmp7 ligand We next tested if FOP-ACVR1 with an intact ligand-binding domain retained the ability to respond to ligand. We used Bmp7 in this experiment since previous studies showed that ACVR1 binds and signals in response to Bmp7 (Yadin et al., 2016;Little and Mullins, 2009;Heinecke et al., 2009;Allendorph et al., 2007). We injected human WT-or FOP-ACVR1 mRNAs into one-cell stage bmp7 sb1aub null mutant zebrafish embryos that also had Acvr1l KD. We then determined whether Bmp7 ligand expression enhanced signaling by these ACVR1 receptors.
Homozygous bmp7-/-embryos exhibit a severely dorsalized C5 phenotype (Figure 3a,b column 1; Figure 3-figure supplement 1; Nguyen et al., 1998;Dick et al., 2000;Schmid et al., 2000). This dorsalization can be rescued by injected bmp7 mRNA (Figure 3a  . ACVR1-R206H and -G328R over-activate pSmad1/5 signaling with or without the presence of a ligand-binding domain. (a,a'-g,g') Animal pole view of relative pSmad1/5 intensities (using arbitrary units (AU)) within each nucleus of the embryo in representative WT early-gastrula embryos (shield-65% epiboly stage) with endogenous Acvr1l present (a-g) or knocked down (a'-g'). Injected mRNAs are noted above for b-g and b'-g'. (a-c, a'-c') Embryos are oriented with ventral (V) side to the left and the dorsal (D) side to the right. (d-g, d'-g') Dorsal side of the embryo could not be identified due to loss of the shield structure with severe ventralization. Scale bars = 100 mm. (a, a') Wild-type and Acvr1l-KD embryos, respectively (N = 18 and 15).   , supporting that loss of the ligand-binding domain prevents ligand response. These data show that ACVR1-R206H and ACVR1-G328R are responsive to BMP ligand, and together with the data in Figures 1 and 2, demonstrate that these FOP-ACVR1 mutant receptors have both ligand-independent and ligand-responsive activity.

ACVR1-R206H signaling in response to activin A ligand depends on its ligand binding domain
Studies have shown that ACVR1-R206H has acquired the ability to respond to Activin A ligand, in addition to its normal BMP ligands (Hatsell et al., 2015;Lees-Shepard et al., 2018;Hino et al., 2015). Previous studies reported that DAcvr1-R206H does not respond to Activin A or BMP ligand in cell culture (Hildebrand et al., 2017). To test if ACVR1-R206H requires a ligand-binding domain to activate pSmad1/5 in response to Activin A or BMP7 ligand in vivo in the zebrafish vertebrate model, we injected Acvr1l-KD, bmp7-/-embryos with mouse Acvr1 or DAcvr1 mRNA with or without Activin A or bmp7 mRNA and assayed pSmad1/5 activation in six hpf embryos and DV phenotypes in 30 hpf embryos.
Expression of Activin A ligand in the early embryo enhances Nodal signaling through pSmad2 ( Figure 4p) and dorsal organizer mesoderm formation (Gritsman et al., 1999;Thisse et al., 2000;Green et al., 1992;Green et al., 1994), precluding our ability to assay DV patterning defects. In    Therefore, to assess the ability of Acvr1-R206H to respond to Activin A with or without a ligandbinding domain, we used immunostaining to quantify pSmad1/5 activity in Acvr1l-KD, bmp7-/embryos injected with Acvr1 or DAcvr1 with or without bmp7 or Activin A. For these experiments, a lower amount of Acvr1-R206H and DAcvr1-R206H mRNA was injected than in the DV phenotyping experiments to decrease the initial pSmad1/5 intensity prior to the addition of ligand and permit detection of Activin A ligand-induced effects. Embryos were collected at shield stage before Activin A expressing embryos halt their development.
A Bmp2/7 heterodimer is the obligatory patterning ligand in the gastrula zebrafish embryo , and as a result, bmp7-/-early gastrula embryos lack detectable pSmad1/ 5 signaling ( Figure 4c). Injection of bmp7 mRNA rescues pSmad1/5 signaling to wild-type levels that peak ventrally and decrease dorsally ( Figure 4d). Addition of bmp7 mRNA did not rescue signaling in the absence of endogenous Acvr1l ( Figure 4f). Addition of Activin A did not rescue pSmad1/5 signaling regardless of the presence of Acvr1l (Figure 4e,g). To quantitatively compare the pSmad1/5 signaling levels of the embryos in these conditions, we determined the mean pSmad1/5 intensities for each condition (Figure 4n). We found that Bmp7 significantly increased the mean pSmad1/5 signaling intensity compared to bmp7-/-embryos ( Figure 4n). Expression of Bmp7 or Activin A in Acvr1l-KD embryos did not significantly alter the mean pSmad1/5 signaling level (Figure 4n), showing that Bmp7 requires Acvr1 to induce pSmad1/5 signaling and Activin A normally has no significant effect on pSmad1/5 signaling.
The low level of injected Acvr1-R206H mRNA activated low levels of pSmad1/5 signaling in bmp7-/-embryos ( Figure 4h). Importantly, co-injection of Acvr1-R206H with Activin A or bmp7 mRNA significantly increased the mean pSmad1/5 signaling intensity (Figure 4i,j,n). Our data suggest that ACVR1-R206H had a more intense response to Bmp7 than Activin A; however, it is important to note that the amount of translated ligand in an embryo was not determined in these experiments. In addition, the dorsal organizer expansion in response to Activin A in our experiments may mask pSmad1/5 intensity enhancement by ACVR1-R206H in response to Activin A. These data demonstrate that Acvr1-R206H activates pSmad1/5 signaling in response to both Activin A and Bmp7.
We next evaluated the ability of Acvr1-R206H to activate pSmad2 signaling. A nuclear pSmad2 gradient forms along the margin of the zebrafish embryo during early gastrula stages in response to Nodal-Vg1 signaling and specifies endodermal and mesodermal cell fates (Hill, 2018;Pelliccia et al., 2017;Montague and Schier, 2017;van Boxtel et al., 2015;Figure 4o). Activin A can rescue loss of Nodal signaling by activating pSmad2 through Nodal receptors (Gritsman et al., 1999). Acvr1-R206H has been shown to signal in response to either BMP or Activin A ligand in cell culture and over-activate pSmad1/5, but not pSmad2, signaling (Hatsell et al., 2015;Hino et al., 2015;Hildebrand et al., 2017). Expression of Activin A alone in the developing zebrafish over-activates pSmad2 throughout the entire embryo (Figure 4p).
To test if Acvr1-R206H enhances signaling through the pSmad2 pathway, we injected Acvr1-R206H mRNA into bmp7-/-, Acvr1l-KD embryos with or without Activin A mRNA and measured pSmad2 immunofluorescence intensity. Embryos mutant for bmp7 were used to eliminate possible competition between Activin A and Bmp7 for Acvr1 and type II BMP receptors and enhance any potential results. Expression of Activin A significantly increased mean pSmad2 signaling regardless of the presence of endogenous Acvr1l (Figure 4p,q,t). Acvr1-R206H alone did not increase the mean pSmad2 intensity compared to uninjected embryos and had significantly lower mean pSmad2 intensity compared to embryos injected with Activin A (Figure 4r,t). Activin A significantly increased mean pSmad2 intensity in Acvr1-R206H injected embryos, but no more than in the absence of Acvr1-R206H (Figure 4s,t). These data show that Activin A enhances pSmad2 signaling in the embryo, but Acvr1-R206H does not, and confirm that Acvr1-R206H over-activates pSmad1/5, but not pSmad2 signaling.

ACVR1-R206H signals in the absence of the type I Bmpr1 receptor
The ability of FOP-ACVR1 to signal independently of ligand suggests the possibility that the mutant receptor could signal in the absence of receptor complex partners as well. In the developing zebrafish, both type I BMP receptors, Acvr1l and Bmpr1, are required for signaling and to pattern the embryo (Mintzer et al., 2001;Little and Mullins, 2009). BMP ligand is required for Acvr1l to associate with Bmpr1 in the zebrafish gastrula embryo . Previous studies showed that ACVR1-R206H retained the ability to over-activate BMP signaling when either BMPR1A or BMPR1B was knocked down in cell culture (Hino et al., 2015). However, given that these two genes have largely redundant activity (Yoon et al., 2005;Wine-Lee et al., 2004), one copy of BMPR1 (A or B) may be sufficient to allow ACVR1-R206H to signal.
Zebrafish have two bmpr1a genes (aa and ab) and two bmpr1b genes (ba and bb). To test if ACVR1-R206H can signal in the absence of all other type I BMP receptors, we first intercrossed bmpr1aa+/-; bmpr1ab-/-zebrafish ( Figure 5a) to deplete bmpr1a gene function. We next injected the bmpr1a-deficient embryos with MOs against bmpr1ba, bmpr1bb, and acvr1l (designated as Type I KD fish henceforth). We then injected these Type I KD fish with human ACVR1-R206H mRNA, collected embryos at an early gastrula stage for pSmad1/5 immunostaining or at 12 to 30 hpf for phenotyping, analyzed them blindly, then genotyped for the bmpr1aa mutation. Using immunostaining, we confirmed that Flag-tagged ACVR1-R206H is expressed and localized to the cell membrane in bmpr1a+/-embryos (Figure 5b).
We examined how the pSmad1/5 gradient was affected in the injected embryos. ACVR1-R206H significantly increased the mean pSmad1/5 intensity compared to uninjected bmpr1a+/-sibling embryos (Figure 5g,h,k). Type I KD embryos lacked a pSmad1/5 gradient and had a significantly lower mean pSmad1/5 fluorescence than uninjected siblings (Figure 5i,k), demonstrating that loss of type I receptors results in loss of pSmad1/5 signaling. However, ACVR1-R206H significantly increased pSmad1/5 intensity in Type I KD embryos (Figure 5j,k), consistent with the observed ventralized phenotype. These data suggest that FOP-ACVR1 does not require BMPR1a or BMPR1b to phosphorylate Smad1/5 and that this mutant receptor does not require the presence of wild-type ACVR1-BMPR1 signaling complexes.

DACVR1-R206H signals in the absence of the type I Bmpr1 receptor
To test if ligand-independent FOP-ACVR1 signaling requires Bmpr1 or endogenous Acvr1l, we injected Type I KD embryos with mouse DAcvr1-R206H or Acvr1-R206H and evaluated DV patterning phenotypes at 30 hpf and dorsal marker expression in 5-to 9-somite stage embryos. Interestingly, like Acvr1-R206H ( Figure  We additionally evaluated expression of two dorsal markers, pax2.1 and krox20, by whole-mount in situ hybridization. In wild-type 5-to 9-somite stage embryos, pax2.1 is expressed in the midbrainhindbrain (MHB) boundary (Figure 6c, white arrowhead at anterior) and krox20 is expressed in rhombomeres 3 and 5 (Figure 6c, black arrowheads) (Thisse et al., 2001;Strähle et al., 1993;Krauss et al., 1991;Hashiguchi and Mullins, 2013). In bmpr1a-/-embryos, which develop to a C4 dorsalized phenotype, the MHB and rhombomere expression of pax2.1 and krox20, respectively, were expanded laterally (Figure 6d). In Type I KD embryos, which display a C5 dorsalized phenotype, pax2.1 and krox20 became radially expressed in the MHB and rhombomere 3 and 5 ( Figure 6e). Acvr1-R206H mRNA injection rescued Type I KD embryos to a ventralized phenotype, characterized by dorsally restricted neural expression of pax2.1 and krox20 (Figure 6f) compared to Type I KD embryos. Injection of Type I KD embryos with DAcvr1-R206H mRNA, which lacks critical regions of the ligand-binding domain, similarly restricted expression of pax2.1 and krox20 to dorsal regions ( Figure 6g). These data provide additional support that ligand-independent signaling by ACVR1-R206H also does not require Bmpr1a, Bmpr1b, or endogenous Acvr1l.

ACVR1-R206H responds to ligand in the absence of Bmpr1
Since ligand facilitates association of Acvr1 and Bmpr1 in the developing zebrafish ), we next examined if FOP-ACVR1 requires other type I BMP receptors for its enhanced ventralizing activity in response to ligand. We expressed human ACVR1-R206H in Type I KD embryos that were either deficient in BMP (by overexpressing the BMP ligand-binding inhibitor Chordin) or overexpressing Bmp7. Embryos were evaluated for DV patterning and dorsal marker gene expression.

Discussion
Using an in vivo zebrafish model, a highly informative and sensitive vertebrate system for BMP signaling activity during DV patterning, we investigated the signaling mechanism of the ACVR1 type I BMP receptor and the effects of ACVR1 activating mutations that cause a rare genetic disorder of ectopic bone formation. We show that although wild-type ACVR1 requires its ligand-binding domain for signaling, the FOP-ACVR1 mutants, ACVR1-R206H and ACVR1-G328R, do not require a ligandbinding domain to over-activate pSmad1/5 signaling. However, these mutant receptors do retain the ability to respond to BMP and Activin A ligand when their ligand-binding domain is intact. We further show for the first time that Bmpr1, which is a required partner for signaling by WT Acvr1l, is dispensable for both ligand-independent and ligand-responsive signaling by mutant FOP-ACVR1. Our data support that the causative mutations of FOP allow the mutant receptor to bypass normal ligand-receptor complex assembly. This aberrant signaling highlights the importance of ligandreceptor complexes to facilitate and regulate signaling by ACVR1. Elucidation of FOP-ACVR1 signaling mechanisms not only provides insight into therapeutic targets for treating FOP, but also gives us unique insight into how BMP signaling is regulated.
Recent studies have shown that FOP-ACVR1 has acquired the ability to signal through pSmad1/5 in response to the TGFb family ligand Activin A, which normally binds ACVR1b to signal through pSmad2/3 (Hatsell et al., 2015;Lees-Shepard et al., 2018;Hino et al., 2015). Our study confirms that ACVR1-R206H is able to respond to both its normal activating ligand, Bmp7, and to Activin A, in an in vivo model, demonstrating that FOP-mutant ACVR1 has an expanded repertoire of ligands to which it can respond. These data support a mechanism of over-active signaling induced by availability of multiple ligands for ACVR1-R206H. Interestingly, the magnitude of Bmp7 response by ACVR1-G328R appeared to be greater than that of ACVR1-R206H (Figure 3c,d). While multiple factors could be contributing to this observation, other studies have shown similar trends (Haupt et al., 2018), suggesting that the signaling mechanisms of the two mutants are not identical. An interesting next experiment would be to compare Activin A responses of ACVR1-R206H and -G328R in vivo.
In this study, we also demonstrate in an in vivo vertebrate system that neither Acvr1-R206H nor Acvr1-G328R require a ligand-binding domain to signal. Further supporting ligand-independent pathway activation, we showed that Acvr1-R206H and Acvr1-G328R are able to pattern the zebrafish embryo in the absence of their normal obligatory ligand, Bmp2/7 (Figure 3). This does not preclude the possibility that other components of an FOP-ACVR1 signaling complex play a role in binding ligands other than BMP. We show that Acvr1-R206H pSmad1/5 signaling also is inappropriately activated by Activin A in the zebrafish. By RNAseq and wholemount in situ hybridization, low levels of activin A expression are detectable in the prechordal plate during late gastrulation (75% epiboly) (Thisse et al., 2004;White et al., 2017), a stage later than our analysis of pSmad1/5 at early gastrulation (shield to 65% epiboly) (Figures 2 and 4d,e). However, another report detected low activin A expression during early gastrulation by RT-PCR (Hashiguchi et al., 2008), thus we cannot exclude Activin A as a factor in FOP-ACVR1 signaling in Figure 3. We show, however, that ligand-binding deficient ACVR1 does not respond to Bmp7 or Activin A ligand (Figure 4k-n), demonstrating that the ligand-binding domain of ACVR1-R206H is required for ligand response and FOP-ACVR1 does not require ligand to signal (Hildebrand et al., 2017).
Studies suggest that not only is ligand critical for normal type I receptor complex assembly, but ligand also mediates which receptors assemble together through differential ligand-receptor affinity (Yadin et al., 2016;Little and Mullins, 2009;Heinecke et al., 2009;Allendorph et al., 2007;Antebi et al., 2017). In zebrafish, Acvr1l and its type I BMP receptor partner Bmpr1 will not associate with each other and signal in the absence of appropriate ligands . Therefore, the ability of ACVR1-R206H and -G328R to signal in response to novel ligands suggests an acquired ability to signal in novel receptor complex combinations. Conversely, the ability of the mutant receptors to signal in the absence of an intact ligand-binding domain suggests an acquired ability to signal in the absence of normally-regulated complex formation. In the context of the developing zebrafish, these mutations decouple the activity of ACVR1 from its morphogen signal, Bmp2/ 7. Indeed, in this study, we also determined that ACVR1-R206H does not require its normal type I BMP receptor partner, Bmpr1, for its signaling activity.
We note that in our experimental system, residual maternal bmpr1aa mRNA remains in our bmpr1aa-/-, ab-/-mutants; however, the inability of Bmp7 overexpression to even partially rescue the C4 dorsalized phenotype of bmpr1aa-/-; ab-/-mutants or the C5 dorsalization of Type I KD mutants indicates that residual bmpr1aa+/-maternal transcript has little or no effect on DV patterning (Figure 5f). Further, the inability of Bmp7 overexpression to rescue Type I KD embryos suggests that any remaining receptor from our knockdown was not sufficient to contribute to DV patterning, indicating a severe loss of Type I receptor function.
The signaling activity that we observed in our ligand-binding mutant ACVR1 receptors may represent a baseline constitutive activity by ACVR1-R206H and -G328R in the absence of receptor complex assembly. All confirmed ACVR1 mutations in FOP are located within the GS or protein kinase domains and are theorized to alter receptor phosphorylation and, as a result, the activation energy required for ACVR1 to signal (Chaikuad et al., 2012;Shore and Kaplan, 2010). Previous studies have suggested that the FOP mutations weaken binding by the type I receptor inhibitor FKBP12, allowing for constitutive or 'leaky' signaling (Chen et al., 1997;Huse et al., 1999;Groppe et al., 2011). However, more recent evidence suggests that this loss of FKBP12 inhibition does not fully account for the level of increased activity displayed by the FOP mutant receptors (Machiya et al., 2018). In addition, ACVR1-R206H has been shown to require type II BMP receptors and GS domain activation to signal (Hino et al., 2015;Bagarova et al., 2013;Le and Wharton, 2012). In the absence of ligand-induced complex assembly with other receptors, GS domain activation of FOP-ACVR1 may occur by stochastic association with free type II BMP receptors. These data highlight that even more remains to be elucidated about the mechanism by which BMP receptor components assemble and coordinately regulate signaling.
Conversely, the Bmp7 and Activin A ligand-responsive activity that we observed by ACVR1-R206H and -G328R may reflect mutation-induced stabilization of signaling activity by ligand-facilitated receptor complex assembly. Under normal circumstances, WT Acvr1l is an obligate type I receptor partner with either Bmpr1a or Bmpr1b . Our results show that FOP-ACVR1 can respond to Bmp7 ligand without its normal type I BMP receptor partner, Bmpr1. These data suggest that ACVR1-R206H is either able to respond to ligand on its own or with novel receptor partners. An important future experiment would be to test if ACVR1-R206H requires Bmpr1 or other type I receptors to respond to Activin A.
In the absence of Bmpr1, ACVR1-R206H may homodimerize with itself to signal. BMPR1, although not ACVR1, has been shown to form inert pre-formed homodimer complexes that are poised to respond to ligand . ACVR1, however, normally binds ligand poorly on its own, indicating that this type I receptor may not easily associate or form stable complexes with ligand in the absence of other receptor partners (Heinecke et al., 2009). We also note that the type I BMP receptor, ACVRL1 (ALK1), was not targeted for knockdown in this study; however, expression studies show that ACVRL1 is not detectable by RNAseq until 75% epiboly, after the stage in which we see expanded pSmad1/5 signaling, and it is not detectable by in situ during the first 24 hr of development (Thisse et al., 2004;White et al., 2017). Our study, however, cannot exclude that low levels of ACVRL1 play a role in the ability of ACVR1-R206H to respond to Bmp7 in the absence of BMPR1. In our study, the type II BMP receptors are also present in the zebrafish embryo and may be sufficient to form a working signaling complex with FOP-ACVR1 as the only type I receptor and still respond to ligand. Previous reports have shown that both BMPR2, a type II receptor with high affinity for BMP ligand, and ACVR2a, a type II receptor with high affinity for Activin A, are required for signaling by ACVR1-R206H (Yadin et al., 2016;Hino et al., 2015;Walton et al., 2012), and an important next step would be to test this requirement in our in vivo system.
Recent studies have shown that TGFbR1 can associate with ACVR1 and initiate pSmad1/5 signaling (Ramachandran et al., 2018). TGFbR1 normally signals with TGFbR2 in response to TGFb ligand, components with which ACVR1 is not known to associate. ACVR1b, the Nodal/Activin receptor, is also present in the zebrafish embryo. Notably, ACVR1b uses the same type II receptors, ACVR2a and ACVR2b, as ACVR1 to signal, although these complexes normally signal through pSmad2/3 rather than pSmad1/5. The ability of FOP-ACVR1 to signal in response to Activin A, may occur by signaling through components of the normal Activin signaling complex: ACVR1b and ACVR2. Our data show that if ACVR1 signals by complexing with Acvr1b, this was not sufficient to sequester Acvr1b and reduce pSmad2 signaling in our system (Figure 4t). We show that there is no significant difference in pSmad2 intensity in embryos injected with Activin A alone or with Acvr1-R206H. Much remains to be investigated to elucidate the mechanism of an ActivinA/ACVR1-R206H response.
The pathological significance of the ACVR1-R206H signaling modalities elucidated in this study remain to be further investigated. One possibility is that they represent two proposed modes of HO initiation in FOP patients: spontaneous and injury-induced. Ligand-independent basal activity of FOP-ACVR1 may lead to increasing levels of signaling within tissues and the initiation of spontaneous HO when a threshold is reached with no obvious acute inciting factor (Shore and Kaplan, 2010). Ligand-responsive hyperactivity may account for injury-induced HO, in which cascades of signaling in response to injury-induced ligands, including Activin A (Hatsell et al., 2015;Lees-Shepard et al., 2018;Hino et al., 2015), contribute to increased activity by FOP-ACVR1. It is important to note, however, that these two modes of HO initiation may not represent separate mechanisms, and ligand independent and responsive activation may play more nuanced roles in the pathogenesis of disease. Research to identify FOP treatment strategies has examined multiple drug targets in ACVR1 signaling including ligand binding and ACVR1 kinase activity (Cappato et al., 2018). Our data suggest that directly targeting the kinase activity of FOP-ACVR1 to inhibit both ligand independent and responsive signaling by the receptor would be a highly promising strategy for treatment.

Zebrafish
Procedures involving animals were approved by the University of Pennsylvania IACUC. Adult zebrafish were kept at 28˚C in a 13 hr light/11 hr dark cycle. All zebrafish husbandry was performed in accordance with institutional and national ethical and animal welfare guidelines. The characterized mutant bmp7a sb1aub (bmp7-/-) (Schmid et al., 2000) and bmpr1aa p3/+ ; bmpr1ab sa0028 (bmpr1aa+/-; bmpr1ab-/-) were used in this study. bmp7 sb1aub fish were maintained as homozygous mutant stocks by rescuing the C5 dorsalized embryonic phenotype with bmp7 mRNA injection. bmpr1aa+/-; bmpr1ab-/-fish were generated by intercrossing and genotyping for the bmpr1aa allele (described below). Embryos used in these experiments were 0-48 hpf. Embryos were maintained at 28-32˚C in E3 solution. Sex/gender was not accounted for as zebrafish sex determination takes place later, during juvenile stages of development (Santos et al., 2017).

CRISPR generation and identification of bmpr1aa mutant allele
A mutant allele of bmpr1aa (bmpr1aa p3 ) in the zebrafish was generated using CRISPR-Cas9 mutagenesis. The target site GGTATAAGTGGCAGACAGAG in exon 8 (out of 13) in the kinase domain, was selected with the assistance of the web tool CHOPCHOP (https://chopchop.cbu.uib.no). Single guide RNAs (sgRNAs) were designed to utilize the T7 Promoter. sgRNAs were synthesized in-vitro from PCR amplified templates using the cloning-free method of Gagnon et al., 2014 with the following changes: sgRNA templates were amplified using Phusion Polymerase 40 uL reaction (Thermo-Fischer F553S). sgRNA templates were purified after amplification using the MinElute purification kit (QIAGEN 28004). sgRNAs were synthesized from the templates in vitro using the MEGAshortscript T7 kit (ThermoFischer AM1354). Megashortscript reactions were run overnight rather than the recommended 2-4 hr, as this was found to increase yield. sgRNA was purified post synthesis using the ethanol precipitation protocol from Gagnon et al., 2014. Purified sgRNAs were assessed and quantified visually by running a dilution series on a glyoxal/sodium phosphate buffer RNA gel and comparing to RiboRuler (ThermoFischer SM1821), as the Nanodrop was found to be an unreliable to quantify these guide RNAs. Immediately prior to injection, 2.5 mL of undiluted sgRNA were mixed with 1 ml of 5 mg/mL Cas9-NLS protein (PNA Bio CP01-50), 1 ml of Phenol Red, and 0.5 ml 1M KCl (total volume 5 ml), this solution was kept at room temperature for the duration of the injection. This mixture was injected in volumes of 2 nl into wild type and bmpr1ab sa0028 embryos.
CRISPR efficiency was assessed using High-Resolution Melt Analysis (HRMA) (Dahlem et al., 2012). At 48 hpf, a subset of the injected embryos was sacrificed for HRMA. HRMA was performed using the MeltDoctor HRM master mix (ThermoFischer 4415440). Embryos from injections with greater than 50% efficiency were raised to adulthood, and outcrossed to WT or bmpr1ab-/-fish. 12 embryos from each outcross were sacrificed for Sanger Sequencing. DNA was isolated from lysed embryos and the target sequence was amplified with PCR. Products from these PCR reactions were purified using the Qiaquick PCR purification kit (Qiagen 28106). Purified PCR products were quantified via Nanodrop and sequenced by the University of Pennsylvania Sequencing core by Sanger sequencing. Mutant alleles were identified by directly reading trace data using Lasergene SeqMan Pro.
A mutant allele with a 53 base pair deletion and a 29 base pair insertion, which includes the exon eight splice acceptor site was identified. Siblings of allele-carrying embryos were raised to adulthood, genotyped, and in-crossed. In the bmpr1ab-/-background, homozygous mutants display a C4 phenotype. Mutant embryos were re-sequenced to confirm the allele sequence. RNA was later isolated from homozygous mutant embryos at 12 hpf, a stage at which they could be identified phenotypically, using Trizol (ThermoFischer 15596026), and reverse transcribed to create cDNA using the SuperScript II kit (ThermoFischer 18064014). The bmpr1aa transcript was amplified by PCR, sequenced and compared to WT transcript. Sequence analysis revealed that the mutation lead to the inclusion of the intron between exons 7 and 8, which contains a premature stop codon.

Genotyping
Adult and embryonic genomic DNA was obtained using HotShot DNA isolation. Genotyping of adults and embryos for bmpr1aa was performed using KASPar genotyping (Smith and Maughan, 2015). Primers were designed and synthesized by LGC Bioscience Technologies.

mRNA synthesis
Human, mouse, or zebrafish ACVR1 cDNAs were cloned into the pCS2+ expression vector. Mouse Acvr1 constructs contained an HA tag inserted after amino acid M34. DAcvr1 constructs contained a 64 amino acid deletion (C35-C99) within the extracellular domain (Haupt et al., 2014). Human ACVR1 constructs all contained a C-terminal Flag tag. mRNA was synthesized using SP6 mMessage machine kit (Sigma Aldrich) and purified using phenol:chloroform extraction. The mRNA was stored in nuclease-free water at À80˚C.

Microinjection of one-cell stage zebrafish embryos
Eggs at 0-15 min post fertilization were collected in E3 media and injected at 22˚C. For each experiment, each mRNA was injected into eggs with the same calibrated needle. For serial injections, eggs were loaded onto the same plate and injected with the first mRNA or morpholino, a subset were then set aside for controls and the remainder were injected with the next mRNA or morpholino. This process was repeated multiple times in different orders to ensure consistency between experimental and control conditions. Injection concentrations of each mRNA were determined based on phenotypic evaluation. In cases in which multiple mRNA syntheses were used, different mRNA concentrations were used to achieve the same phenotype due to presumed inconsistent mRNA 5' capping.

Phenotypic evaluation
Embryos between 12 to 48 hpf were categorized into dorsoventral patterning phenotypes ( Figure 1g). All images of embryos were photographed in E3 media with a Leica IC80HD. Injection results and controls from multiple experiments were then pooled. Phenotype pictures were corrected with background subtraction using ImageJ, and white balanced using Adobe Photoshop. Embryo bubble graphs were generated by pooling total numbers of embryos within each phenotypic category (C5-V5, Figure 1g) in each condition. These numbers were converted into percent of total embryos to generate the bubble graphs using excel bubble graph. (Figure 1-source data 1, Figure 1-figure supplement 2-source data 1, Figure 3-source data 1, Figure 5-source data 1, Figure 5-figure supplement 2-source data 1, Figure 6-source data 1, Figure 7-source data 1).
Immunofluorescence P-Smad1/5 immunostaining and imaging were performed as previously described (Zinski et al., 2017;Zinski et al., 2019). For all immunostaining, embryos were fixed in 4% formaldehyde in PBST between shield stage and 60% epiboly (approximately 6-7 hpf), blocked with 10% FBS in PBST, and probed overnight at 4˚C with a 1:200 dilution of anti-PSmad1/5/9 (Cell Signaling 13820). Embryos were then treated overnight at 4˚C with a 1:500 dilution of antibody Alexa 647 (Invitrogen A-21245) and 1:2000 of Sytox green (Fisher S7020) diluted in blocking solution. Stained embryos were stored in the dark at 4˚C in PBST for up to 2 months. P-Smad2 immunostaining was performed as described for P-Smad1/5 immunostaining with the following changes: Primary antibody used was anti-PSmad2/3 (1:800; Cell Signaling 8828), embryos were dehydrated in MeOH after fixing, rehydrated embryos were incubated in Acetone at À20C prior to blocking.

Immunofluorescence imaging and analysis
Prior to imaging, immunostained embryos were gradually dehydrated in MeOH and then cleared using BABB: a 1:2 ratio of benzyl alcohol (Sigma B-104) and benzyl benzoate (Sigma B-6630). Whole embryos were mounted with the DV axis parallel to the cover slip (either animal pole up or down). Imaging was performed using a Ziess LSM880 confocal microscope with an LD LCI Plan-Achromat 25X/0.8 lmm Corr DIC M27 multi-immersion lens in the oil-immersion setting. A single bead from a calibration slide (ThermoFisher Scientific Cat#F369009, Well A1) was imaged between each slide of embryos to account for fluctuations in power of the 633 nm laser over time (This laser was used for detecting pSmad1/5, pSmad2, or HA/Flag). Immunoflourescence was performed as described in Zinski et al., 2017 for pSmad2 immunostained embryos in Figure 2 and pSmad1/5 Immunostained embryos in Figure 4. pSmad1/5 immunostained embryos in Figures 2 and 5 were imaged with the following changes: embryos were imaged in a single~567Â567 mm frame and pixel dwell time was reduced to 0.77msec. Receptor HA and Flag tagged embryos were imaged at 63x. Images of receptor stains were taken approximately 18 mm from the outer yolk syncytial layer of the embryo in a~225Â225 mm frame (Figure 1a). Quantitative receptor tag images were taken by imaging 16 2.2 mm stacks starting at the apical end of the embryo in a~225Â225 mm frame (Figure 1-figure  supplement 2).
Embryo immunofluorescent intensity for pSmad1/5, pSmad2 or HA/Flag was normalized to a calibration bead imaged at the same time. Mean fluorescence of each fluorescent bead stacked image was measured using Imaris statistical analysis. Fold change in mean fluorescence was then calculated for each bead compared to a reference bead that was imaged immediately after an uninjected embryo was imaged. The fold change in mean fluorescence was then applied to the total fluorescence of each experimental-imaged embryo using the ImageJ multiply function.
For pSmad1/5 and pSmad2 immunostained embryos, calibrated Images were analyzed using the Imaris software spots function which identifies each nucleus within the immunostained embryo using the Sytox green nuclear stain and refined based on quality. Spot quality threshold is set using a control image to allow the program to detect all nuclei without detecting fluorescent points that are not nuclei, and then refined by eye (Figure 2-source data 2, Figure 4n-source data 2, Figure 4tsource data 2, Figure 5-source data 3). The Imaris program was then used to color each spot based on mean relative pSmad1/5 or pSmad2 (Alexa 647) intensity (AU) within the spot.
For HA immunostained embryos, calibrated embryos were analyzed using the Imaris software surfaces function to detect the area stained by b-catenin. The total HA flourescence was then measured within the area of b-catenin and the fluorescence per mm 2 was calculated (Figure 1-figure supplement 2-source data 2).

Statistical analysis
For pSmad1/5 and pSmad2 immunostained embryos, mean fluorescence was calculated for all individual nuclei within each embryo using the Imaris spots function, as described above. For each embryo, nuclei that fell outside 1.8 times the interquartile range were removed. The mean of the remaining nuclei was calculated to generate a mean nuclear pSmad1/5 or pSmad2 fluorescence for each embryo. Mean nuclear fluorescence was then compared using a two-tailed T-test assuming unequal variance (Figure 2-source data 1, Figure 4n-source data 2, Figure 4t-source data 2, Figure 5-source data 2).
For receptor tag immunostained embryos, HA fluorescence per mm 2 calculated values that fell outside 1.8 times the interquartile range were excluded. HA fluorescence per mm 2 was also compared by a two-tailed T-test assuming unequal variance (Figure 1-figure supplement 2-source data 2).
Representative embryos in all figures were selected from the three embryos closest to the mean fluorescence of the group, using embryos without any significant tears or defects.

In situ hybridization
Whole mount in situ hybridizations were performed on fixed 5-9 somite stage embryos using DIGlabeled anti-sense RNA probes (made with labeling kit: Roche 11277073910) to pax2.1, krox20, and myod. Probes were visualized with anti-DIG-Alkaline Phosphatase (Roche11093274910) and developed in BM Purple (Roche 11442074001). Embryos were mounted on agarose in methanol and photographed with a Leica IC80HD. Images were processed using image J. After imaging, embryos were collected in MeOH for genotyping. pax2.1 and krox20 expression patterns were categorized as completely radialized (as in a C5 embryo), expanded compared to wild-type, similar to wild-type, or restricted compared to wild-type. A representative embryo was chosen from the category most represented in each condition.