Heterotopic ossification in mice overexpressing Bmp2 in Tie2+ lineages

Bone morphogenetic protein (Bmp) signaling is critical for organismal development and homeostasis. To elucidate Bmp2 function in the vascular/hematopoietic lineages we generated a new transgenic mouse line in which ectopic Bmp2 expression is controlled by the Tie2 promoter. Tie2CRE/+;Bmp2tg/tg mice develop aortic valve dysfunction postnatally, accompanied by pre-calcific lesion formation in valve leaflets. Remarkably, Tie2CRE/+;Bmp2tg/tg mice develop extensive soft tissue bone formation typical of acquired forms of heterotopic ossification (HO) and genetic bone disorders, such as Fibrodysplasia Ossificans Progressiva (FOP). Ectopic ossification in Tie2CRE/+;Bmp2tg/tg transgenic animals is accompanied by increased bone marrow hematopoietic, fibroblast and osteoblast precursors and circulating pro-inflammatory cells. Transplanting wild-type bone marrow hematopoietic stem cells into lethally irradiated Tie2CRE/+;Bmp2tg/tg mice significantly delays HO onset but does not prevent it. Moreover, transplanting Bmp2-transgenic bone marrow into wild-type recipients does not result in HO, but hematopoietic progenitors contribute to inflammation and ectopic bone marrow colonization rather than to endochondral ossification. Conversely, aberrant Bmp2 signaling activity is associated with fibroblast accumulation, skeletal muscle fiber damage, and expansion of a Tie2+ fibro-adipogenic precursor cell population, suggesting that ectopic bone derives from a skeletal muscle resident osteoprogenitor cell origin. Thus, Tie2CRE/+;Bmp2tg/tg mice recapitulate HO pathophysiology, and might represent a useful model to investigate therapies seeking to mitigate disorders associated with aberrant extra-skeletal bone formation.


INTRODUCTION
Bone morphogenetic proteins (BMPs) regulate fundamental processes in development and organismal homeostasis [1]. During canonical Bmp signaling, BMPs ligands bind to BMP type I receptors (BMPRIs), or activin-like kinase (ALK) 2,3, or 6. This complex binds to BMP type II receptor (BMPRII), which phosphorylates BMPRI, which in turn phosphorylates regulatory-Smads (Smad1/5/8). Phosphorylated Smad1/5/8 binds to nuclear Smad4, forming a nuclear complex that accumulates in the nucleus, where it is recruited to transcriptional complexes to mediate BMP-driven gene expression [2]. BMPs were discovered owing to their fundamental role in bone formation and homeostasis [1], and BMP2 is critical for chondrocyte proliferation and endochondral bone maturation, and necessary for bone fracture healing [3,4].
Heterotopic ossification (HO) is bone formation at extra-skeletal sites, including muscle, tendon, ligament, and other connective tissues, and a complication of injury and surgery [5,6]. HO occurs through intramembranous and endochondral bone formation, resembling fracture repair processes. Lesions are characterized by immune infiltration in damaged connective tissue, which is eventually replaced by endochondral bone through fibroblast proliferation, mesenchymal condensation, and chondroosteogenic differentiation [6,7]. Subsequently, the woven bone gives way to the lamellar bone and marrow stroma, with hematopoietic progenitors, adipocytes, osteoblasts and osteoclasts, while capillary-like vessels give rise to bone marrow (BM) sinusoid-type vessels. Acquired HO is relatively common but its etiology is poorly understood [6]. In contrast, genetic forms, like Fibrodysplasia ossificans progressiva (FOP; OMIM #135100, ORPHA337) are rare, but provide mechanistic insight [8][9][10]. FOP patients have progressive spontaneous and injury-induced HO resulting in complete mobility loss. FOP is caused by a mutation in the gene encoding the type I ACVR1/ALK2 BMP receptor [11]. The ACVR1-R206H mutant receptor acquires the ability to respond to the TGFß family ligand Activin A [12,13], and becomes sensitive to other BMPs [14][15][16]. ACVR1 mutations alone cannot explain the recurrent "flare-ups" resulting in extra skeletal ossification following trauma, muscular fatigue, or other inflammatory insults, which also trigger acquired forms of HO. The innate immune system [17,18] and local "niche" soft tissue microenvironment [19,20] need to be further characterized to help clarify this issue. Moreover, Activin A seems to play a significant role in the initial steps of FOP during immune infiltration subsequent to injury. However, once ectopic bone is fused to the normal bone skeleton, additional canonical BMP ligands may be required to sustain ectopic bone development.
The identification of bone osteoprogenitors has generated considerable interest [21,22]. Skeletal muscle-resident cells including myoblasts [23,24], satellite cells [25] or fibroadipogenic progenitors (FAPs) have osteogenic differentiation ability [26,27]. Hematopoietic progenitors participate in bone formation at sites of tissue inflammation, but are insufficient to initiate this process [28,29]. Endothelial cells in mice constitutively expressing ACVR1-R206H, transform into mesenchymal cells with progenitor properties, that give rise to ectopic bone [30]. However, lineage tracing using a Tie2 CRE driver line and local transplantation of Tie2 CRE -derived endothelial cells into skeletal muscle have excluded the endothelium as a source of ectopic bone formation [31]. Rather, non-endothelial Tie2 + resident skeletal muscle stem cells including FAP osteoprogenitors appear to be the principal FOP cells-of-origin [31,32].
In view of ACVR1 receptor activation complexity, the uncertainty of target osteo-progenitors, and modulating "niche" factors, disease modeling is necessary to better understand both acquired and genetic HO. Bmp2 plays fundamental roles in cardiac valves formation and heart chamber patterning [33,34], but its cardiac overexpression causes lethality [35,36]. To study Bmp2 vascular gain-of-function postnatally, we generated Tie2 CRE/+ ;Bmp2 tg/tg mice, which overexpress Bmp2 in hematopoietic/endothelial lineages. These mice survive birth, develop pre-calcific valve disease and a systemic bone disorder in skeletal muscle and other connective tissues, resulting in severe skeletal deformities whose nature we have investigated.

Endothelial Bmp2 overexpression results in valve dysfunction
We previously generated a transgenic mouse line in which CAGdriven Bmp2 expression is activated upon Cre-mediated removal of a β-Geo-stop cassette [35] (Supplementary Fig. 1A). We crossed the CAG-Bmp2 allele with Tie2 CRE line, which is active in hematopoietic/ endothelial lineages from E7.5 onwards [37]. Vascular GFP reporter expression was observed at E9.5, confirming Cre-mediated recombination ( Supplementary Fig. 1B).
Ectopic Bmp2 signaling leads to osteogenic differentiation of valve interstitial cells [38]. To determine the effect of increased endothelial Bmp2 expression on valve function, we generated Tie2 CRE/+ ;Bmp2 tg/tg mice. At 16 weeks, circulating Bmp2 levels were almost six-fold higher than in WT animals (Fig. 1A). Tie2 CRE/+ ; Bmp2 tg/tg mice showed shortened pulmonary acceleration time and acceleration to ejection time ratio by ultrasound (Fig. 1B), indicating pulmonary hypertension potentially leading to respiratory insufficiency. Tie2 CRE/+ ;Bmp2 tg/tg mice displayed significantly increased aortic valve mean, peak velocity and pressure gradient (Fig. 1C, D). Three of seven animals displayed chondrogenic and lipid droplet islands at the leaflet base (Fig. 1E), indicative of precalcific disease. These results indicate that ectopic endothelial and/or hematopoietic Bmp2 expression leads to aortic valve dysfunction compatible with a pre-calcific valve stage.
Hematopoietic/endothelial Bmp2 overexpression causes a HO During these studies, we found that Tie2 CRE/+ ;Bmp2 tg/tg mice develop severe scoliosis and ankylosis with depressed locomotor behavior and respiratory insufficiency. Extensive HO was diagnosed by PET-CT at 16 weeks of age and confirmed at autopsy ( Fig. 2A; Supplementary Fig. 1C Table 1 and Supplementary Fig. 1D). Circulating Bmp2 levels for heterozygous transgenic mice at 24 weeks was 1.5-fold compared with WT ( Supplementary Fig. 1E), while Bmp2 levels of homozygous transgenic Tie2 CRE/+ ;Bmp2 tg/tg animals were 5-6 fold, suggesting that HO is highly sensitive to Bmp2 levels.
HO was associated with typical histopathological changes, including mononuclear cell infiltration ( . Chondrogenic tissue organization in the epiphysis and head of the femur and tibia was similar in Tie2 CRE/+ ;Bmp2 tg/tg and WT animals ( Fig. 2C, a, a', b). However, additional cartilage was present in joints of Tie2 CRE/+ ; Bmp2 tg/tg animals, including osteochondral patches surrounding the tibial head and neck (Fig. 2C, b and b'). Ectopic bone was present adjacent to the normal bone in transgenic animals (Fig. 2C), and bone mass was increased in Tie2 CRE/+ ;Bmp2 tg/tg compared to WT animals (Fig. 2C, c, c', d and d'). We measured bone mass density (BMD) in tibia and femur bones in WT and Tie2 CRE/+ ;Bmp2 tg/tg animals [39]. BMD was increased about 20% in Tie2 CRE/+ ;Bmp2 tg/tg mice compared to WT (9226 vs. 7720 Hounsfield Units; Supplementary Fig. 2A Hematopoietic progenitors contribution to HO in Tie2 CRE/+ ; Bmp2 tg/tg mice Tie2 + cells are potential osteogenic progenitors in HO [21,30,31]. To characterize the hematopoietic contribution to HO in Tie2 CRE/+ ; Bmp2 tg/tg mice, bone marrow (BM) was isolated from normal, and ectopic bone of fore-and hindlimbs, scapulae, hips, and sternum of Tie2 CRE/+ ;Bmp2 tg/tg mice, and processed separately. The combined total BM (normal and ectopic bone) cell number was increased in Tie2 CRE/+ ;Bmp2 tg/tg mice (Fig. 3A, left panel), although there were important variations in the total hematopoietic cell number in ectopic bone (Fig. 3A, right panel) due to uneven HO in individual animals.
We queried whether inflammation was required for HO in Tie2 CRE/+ ;Bmp2 tg/tg mice. FACS analysis revealed that the proinflammatory CD11b + Gr1 + cell population was increased in combined (normal and ectopic) Tie2 CRE/+ ;Bmp2 tg/tg BM ( We tested fibroblast (Fb) and osteoblast (Ob) colony-forming potential in Tie2 CRE/+ ;Bmp2 tg/tg normal and ectopic BM. Fb (Fig. 3F) and Ob (Fig. 3G) CFUs readily formed from WT and transgenic mice BM, from ectopic BM of transgenic animals ( Supplementary  Fig. 1F). Both CFU-Fbs and CFU-Obs were increased in combined tg BM (normal and ectopic) and tg BM (Fig. 3F, G), suggesting that forced Bmp2 expression in Tie2 + cells expands the fibroosteoblastic BM progenitor population.
We analyzed multi-lineage reconstitution in PB (CD11b + , B220 + and CD3 + ) by FACS every month for 12 months. Whilst myeloid reconstitution was unaffected, B and T lymphoid cells were decreased at all time points ( Supplementary Fig. 3D). BMP2/4 has been shown to antagonize T-cell lineage differentiation [41,42]. None of the transplanted animals developed HO by Nano-PET-CT (data not shown), even 12 months after the BM transplant assay. Thus, Bmp2 secreted by hematopoietic Tie2 CRE/+ ;Bmp2 tg/+ cells in WT mice is not sufficient to drive HO. Alternatively, local expression of other Tie2-Cre-targeted cells is crucial to initiate flare-ups and HO.
Tie2 CRE/+ ;Bmp2 tg/tg mice transplanted with WT hematopoietic cells eventually develop HO, likely because HO was already taking place at the time of transplantation (8 weeks). Five animals presented typical HO with variable severity, including complete hindlimb immobilization by 16 weeks (n = 1), dorsal vertebrae at 20 weeks (n = 1), and hindlimb and dorsal vertebrae at 22 weeks (n = 2) (Fig. 4E). One mouse presented a fat cyst dorsally but no HO, and another developed HO in dorsal vertebrae at 26 weeks (not shown). One mouse remained asymptomatic 37 weeks after transplant. Thus, 5 out of 7 Tie2 CRE/+ ;Bmp2 tg/tg mice reconstituted with WT hematopoietic cells developed HO, with delayed onset compared to non-transplanted animals. Disease onset in Tie2 CRE/+ ; Bmp2 tg/tg mice occurred at 16-28 weeks in transplanted mice, versus 8-18 weeks in non-transplanted ones (Fig. 4F, left panel). Tie2 CRE/+ ;Bmp2 tg/tg mice transplanted with WT HSCs survive 10 weeks longer than non-transplanted transgenics (Fig. 4F, right  panel). Therefore, Bmp2 expression in non-hematopoietic cells is essential for HO development, and concomitant Bmp2 expression by hematopoietic cells accelerates this process.
Chondro-osteogenic differentiation is associated with BMP signaling activation in FAP cells Tie2 marks a subset of resident skeletal muscle cells [31], which potentially contribute to HO. Tie2 CRE/+ ;Bmp2 tg/tg cartilage and bone lesions showed GFP + co-staining with IB4 + (Fig. 5A), colocalizing with Tie2 + fibroblastic/adipocytic skeletal-muscle cells Fig. 2 Constitutive Tie2-driven Bmp2 expression causes HO lesions in mice. A Nano PET-computed tomography (CT) images of 16-week-old WT and Tie2 CRE/+ ;Bmp2 tg/tg mice showing ectopic bone lesions close to ribs, scapulae, and neck (red arrows, ribs, rb; nk, neck; dorsal vertebrae, dv). B H&E staining on sections of skeletal muscle of the hindlimbs of Tie2 CRE/+ ;Bmp2 tg/tg mice showing histological features typical of HO lesions. a. Evidence of inflammation in HO lesion. a' Damaged skeletal muscle fibers with central nuclei (arrows), mononuclear infiltration (black arrowheads), and fat cells (white arrowheads). b, b' Area of massive fibroblast accumulation. c Ectopic bone in skeletal muscle (arrows) next to tibia. c' Chondro-osteogenic areas with chondrocytes (white arrowhead), and osteoblasts (black arrowheads). d, d' Mature ectopic bone with colonizing bone marrow cells (arrows), chondrocytes (white arrowhead), and osteoblasts (black arrowhead). C Top panels: Alcian blue staining of sections of WT and Tie2 CRE/+ ;Bmp2 tg/tg knee joint. a, a' In WT, chondrocyte tissue (in blue) is located in the epiphysis region, tip of the bone (black arrowheads), and head of the fibula (arrow). b, b' In Tie2 CRE/+ ;Bmp2 tg/tg , chondrocyte tissue is located in the epiphysis region at the tip of the bone (black arrowheads), accumulated in connective tissue (arrow) of the meniscus and head of the fibula (white arrow) and chondrogenic areas (white arrowhead) inside the skeletal muscle. Bottom panels: Alizarin red staining of WT and Tie2 CRE/+ ; Bmp2 tg/tg knee joint. c, c' In WT, osteogenic tissue (in red) is located in the epiphysis region and tip of the bone complementary to chondrogenic areas (white arrowheads). d, d' Intense staining in the head of Tie2 CRE/+ ;Bmp2 tg/tg joints (white arrowheads in (d')). Extra ossification inside the skeletal muscle and head of the fibula (arrows in (d)), and the meniscus (black arrowhead). Scale bar 200 µm.
The expression of chondro-osteogenic markers in damaged skeletal muscle fibers adjacent to ectopic bone, suggests an active repair process. The number of satellite cells labeled by CD45 + Sca1 -CD34 + α7int + was not significantly different between Tie2 CRE/+ ;Bmp2 tg/tg and WT hindlimb skeletal muscle (Fig. 5D, left  panel). There were no GFP + satellite cells among Tie2 CRE/+ ;Bmp2 tg/tg skeletal muscle cells (Fig. 5D, right panel), suggesting that satellite cells are not overexpressing Bmp2 and probably not directly implicated in HO in Tie2 CRE/+ ;Bmp2 tg/tg mice.
Our transplant studies indicate that while Bmp2 transgenic progenitors contribute BM stem cell components, the HO chondro-osteogenic stem/progenitor cell is not BM-derived, consistent with previous studies [28,29]. Otherwise, hematopoietic/endothelial Bmp2 overexpression profoundly affects hematopoiesis. LSK stem cells, myeloid CFU-GM and Cd11b/Gr11 cells, and fibroblast and osteoblast progenitors are all expanded in transgenic mice. Tie2 CRE/+ ;Bmp2 tg/tg mice display increased Cd11b/ Gr11 cells in BM and PB, and increased erythroid lineage differentiation, consistent with in vitro findings [40]. Overall increased BM cellularity is consistent with HSC niche enlargement, enhancement of HSC self-renewal and pool size. Within the BM niche, non-canonical BMP signaling regulates intrinsic HSC maintenance in vivo [48][49][50], and is implicated in determining the HSC fate, by promoting a pro-lymphoid transcriptional program and sustaining lymphoid-biased HSC commitment [48,49].
Previous BMP overexpression studies using a variety of promoters failed to cause HO (reviewed in [53]), because the relevant progenitor cell type had not been targeted. One exception is the transgenic mouse overexpressing BMP4 under control of neuron-specific enolase (Nse) promoter [53,54]. Our Tie2 CRE/+ ;Bmp2 tg/tg mice resembles this model, matching a stereotyped spreading pattern of HO formation. Neither the NSE-BMP4 model, nor our Tie2 CRE/+ ;Bmp2 tg/tg model recapitulate any of the congenital phenotypes associated with FOP. Moreover, direct versus indirect Bmp effects on target stem/progenitor cell populations cannot be assessed using these models. Nevertheless, Tie2 CRE/+ ;Bmp2 tg/+ mice are viable, and develop HO within weeks. This HO model does not require surgical procedures involving the implantation of a BMP-loaded matrix and Bmp2 dosage can be modulated via copy number gene expression. Tie2 CRE/+ ;Bmp2 tg/tg mice are maintained on a heterozygote background, so that a single cross allows for the generation of experimental animals.

MATERIALS AND METHODS Mouse strains and genotyping
The following mouse strains were used: male and female mixed background C56BL/6-CD1 R26CAGBmp2 tg [35] and Tie2 CRE [37]. For simplicity, R26CAGBmp2 tg/+ and R26CAGBmp2 tg/tg are abbreviated in the text and figures as Bmp2 tg/+ and Bmp2 tg/tg , respectively. Details of genotyping will be provided upon request. Recipient transplanted animals were WT C56BL/6, CD1 or Tie2 Cre/+ ;Bmp2 tg/+ animals.

Ultrasound
Mice were anaesthetized by inhalation of isoflurane and oxygen (1.25% and 98.75% respectively) and examined by a 30 MHz transthoracic echocardiography probe. Images were obtained with VEVO 2100 (VisualSonics, Toronto, Canada) from Tie2 Cre/+ ;Bmp2 tg/tg (n = 10) and WT (n = 9) littermates. Short axis and long axis, B Mode, and 2D M-Mode views were obtained from the M mode by an expert in ultrasound in a blind fashion as described previously [55]. From these images, left ventricle (LV) function was estimated by fractional shortening (FS) and ejection fraction (EF). For FS measurements a long or short-axis view of the heart was selected to obtain an M mode registration in a line perpendicular to the LV septum and posterior wall at the level of the mitral chordae tendinea. Pulmonary acceleration time (PAT) and ejection time (PET) were measured in the parasternal short-axis view by pulsed-wave Doppler of pulmonary artery flow [56]. B-mode and color-Doppler guided pulsed-wave Doppler was used to record the maximal transvalvular jet velocity. Specifically, to avoid Doppler misalignment, coaxial interrogation of the aortic flow was ensured by the operator, and all the measurements were obtained using an angle of interrogation <30°. To correct for flow dependence, we computed an EF velocity ratio (EFVR = EF (%)/maximal aortic velocity [m/s]) as an additional indicator of disease severity [57].

Bone mass spectrometry
Nano-PET-CT acquired and 3D-reconstructed images from 16-week-old WT and Tie2 Cre/+ ;Bmp2 tg/tg animals (n = 3 of each group) were further analyzed for bone mass densitometry. Hind limbs mean attenuation coefficient of X rays expressed in Hounsfield Units was measured after segmenting bones using the Multimodality Workstation MMWKS [39].

Histology
Skeletal muscles and bone tissue (and ectopic bone in transgenic animals) were fixed in 4% PFA for 24 h at 4°C, and after decalcification with ImmunoCal (StatLab, Fisher scientific) embedded in paraffin or sucrose treated to be cryopreserved in OCT. Hematoxylin/eosin (H&E), Masson´s trichromic, Alizarin red, and Alcian blue stainings were performed according to standard protocols on paraffin-embedded 7 μm sections. Oil Red, staining was performed according to standard protocols on 5 μm cryosections.

Immunohistochemistry
Paraffin-embedded 7 μm sections of hind limb tissues were citrateunmasked and stained with the following primary antibodies: polyclonal P-

Confocal imaging
Confocal images of tissue sections were acquired with a Nikon A1R laser scanning confocal microscope and NIS-Element SD Image Software. Images of stained explants were collected as z-stacks. Z-projections and lateral sections were assembled using ImageJ. Images were processed in Adobe Photoshop Creative Suite 5.1.

CFU assays
16-20-week-old WT (n = 16) and Tie2 Cre/+ ;Bmp2 tg/tg animals (n = 20) were sacrificed and blood samples were obtained by cardiac puncture. Spleen and bones were dissected from posterior and anterior limbs, hips, and sternum. Extra bone formations from transgenic animals were dissected and processed separately. BM cells were obtained by crushing bones with a mortar in PBS. The solution containing BM cells was separated, and the remaining bone was treated with Collagenase I for 45 min at 37°C in a shaking bath to obtain stromal cells. All samples were 70 μm-filtered. Red blood cells were removed from BM and spleen samples using lysis buffer (0.15 M NH4Cl for 10 min at 4°C) and cell number was determined. Peripheral blood mononuclear cells were isolated from diluted-blood (300 μl of PB mixed with 1.5 ml phenol-red-free RPMI). These samples were added carefully over 2 ml of Lympholite-M (Cedarlane) and were carefully layered on top and centrifuged without brake for 25 min at RT. The halo phase with mononuclear cells was centrifuged at 1500 rpms for 10 min. Cells were resuspended in 200 μl RPMI w/o phenol red. For CFU-GM, cells were mixed with methyl-cellulose and seeded on low adherence p35mm (BM 1 × 10 4 cells; Spleen 4 × 10 5 cells, PB 200 μl) by duplicate. The colonies were scored after 7 days in culture. For CFU-Fb and CFU-Ob (WT, n = 13 and 14; Tg n = 19), cells were plated by triplicate in three wells of six-well plates (1 × 10 6 /well) in α-MEM + 15% FBS. CFUs were stained and counted after a week in culture.

BM transplantation assays
For BM transplantation assays we used 8-week-old male WT C57BL6 or heterozygous Tie2 Cre/+ ;Bmp2 tg/+/− (mixed background) as donors, (n = 2 per transplantation assay) and two groups of male 8-week-old WT as recipients (n = 7 and 8 per transplantation assay). Briefly, lethality irradiated receptor mice (11 grays) were transplanted with 2 × 10 6 bone marrow cells (BMCs) by tail vein injection. BMCs were obtained as mentioned above in FACS analysis section. The homogenized samples were filtered through a 40-μm mesh to obtain single-cell suspensions, and depleted of red blood cells by lysis. For reciprocal transplantation assays we used 8-week-old WT CD1 or Tie2 Cre/+ ;Bmp2 tg/tg as donors, (n = 2 per transplantation assay) and two groups of 8-week-old WT CD1 (n = 10) or Tie2 Cre/+ ;Bmp2 tg/tg (n = 7), as recipients without HO symptoms. In order to check the engraftment of hematopoietic cells in each group of transplants, we bled the animals every month after the transplant, until 4 (WT BMC into Tie2 Cre/+ ;Bmp2 tg/tg ) or 5 (Tie2 Cre/+ ;Bmp2 tg/tg BMC into WT) months posttransplant.

Statistics
Due to the high variability in HO onset and severity, the analysis was made in groups n > 10 in some cases. Statistical assessment is indicated in the figure legends. For each experiment comparing two groups, a mean ± SD is represented and a two-tailed t test was performed. For experiments comparing two groups at different time points, a mean ± SD is represented at each timepoint and a two-way ANOVA followed by Sidak's correction was performed. *P < 0.05; **P < 0.01; ***P < 0.001; **** P < 0.0001.

DATA AVAILABILITY
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.