Strategic Targeting of Multiple BMP Receptors Prevents Trauma-Induced Heterotopic Ossification

Abstract

Trauma-induced heterotopic ossification (tHO) is a condition of pathologic wound healing, defined by the progressive formation of ectopic bone in soft tissue following severe burns or trauma. Because previous studies have shown that genetic variants of HO, such as fibrodysplasia ossificans progressiva (FOP), are caused by hyperactivating mutations of the type I bone morphogenetic protein receptor (T1-BMPR) ACVR1/ALK2, studies evaluating therapies for HO have been directed primarily toward drugs for this specific receptor. However, patients with tHO do not carry known T1-BMPR mutations. Here we show that, although BMP signaling is required for tHO, no single T1-BMPR (ACVR1/ALK2, BMPR1a/ALK3, or BMPR1b/ALK6) alone is necessary for this disease, suggesting that these receptors have functional redundancy in the setting of tHO. By utilizing two different classes of BMP signaling inhibitors, we developed a translational approach to treatment, integrating treatment choice with existing diagnostic options. Our treatment paradigm balances either immediate therapy with reduced risk for adverse effects (Alk3-Fc) or delayed therapy with improved patient selection but greater risk for adverse effects (LDN-212854).

Targeted therapy for trauma-induced heterotopic ossification (tHO) has been guided by the availability of analogous genetic models with clear causative mutations. Trauma patients rarely carry specific mutations. Here the authors describe a functional redundancy of type I bone morphogenetic protein (BMP) receptors in tHO and evaluate the efficacy of broadly specific BMP signaling inhibition.

Introduction

Musculoskeletal trauma and burns result in substantial morbidity in survivors because of extensive and prolonged wound healing, the need for post-injury rehabilitation, and the development of secondary pathology. Trauma-induced heterotopic ossification (tHO) is a form of pathologic wound healing in which extra-skeletal bone forms in soft tissues, including muscle and tendon. Almost 20% of major burn patients, amputees, and joint and spine surgery patients and 63% of wounded soldiers develop tHO following combat-related extremity trauma involving limb amputation.1, 2, 3 Patients with tHO experience intractable pain, open wounds, and loss of joint function, leading to debilitating contractures. These complications are so burdensome that the military has identified tHO as one of the most important barriers to recovery in soldiers.⁴

Current treatment strategies for tHO are limited. Surgical excision remains the only definitive option to eliminate ossified lesions. However, extirpation of ectopic bone does not restore form or function because the joint contracture and chronic pain caused by tHO cannot be reversed. After surgical excision, patients may be left with functional deficits requiring further rehabilitation and open wounds that are painful and susceptible to infection.⁵ Current therapeutic options are plagued by inconsistent results, systemic adverse effects, poorly defined therapeutic windows, and lack of specificity for the underlying mechanisms that cause tHO.6, 7, 8, 9, 10

Two major barriers have made tHO prevention/treatment an unrealized goal. First is an incomplete understanding of the mechanisms responsible for this process. Much of HO literature has focused on a related pathology known as fibrodysplasia ossificans progressiva (FOP). FOP is caused by a hyperactivating mutation in the type I bone morphogenetic protein receptor (T1-BMPR) ACVR1/ALK2 (ACVR1 R206H), leading to increased SMAD1/5 phosphorylation and expression of downstream pro-osteogenic genes.11, 12, 13 Consequently, emphasis has been placed on developing inhibitors with improved specificity for this receptor. However, patients who develop tHO do not harbor known ACVR1 mutations, and it is unclear whether emphasis on ACVR1/ALK2-specific inhibition is beneficial. Two additional T1-BMPRs, BMPR1a/ALK3 and BMPR1b/ALK6, bind BMP ligands and, subsequently, phosphorylate SMAD1/5, similar to ACVR1/ALK2.14, 15 We therefore interrogate whether any single T1-BMPR is required for tHO or whether these receptors perform overlapping roles during tHO.

Several ACVR1/ALK2 inhibitors have emerged in the literature. LDN-193189 (LDN19), a small-molecule inhibitor of ACVR1/ALK2 as well as BMPR1A/ALK3 and BMPR1B/ALK6, has been shown to reduce HO in a mouse model of constitutively active ACVR1/ALK2 (ACVR1 Q207D) resembling FOP.16, 17 We and others have demonstrated that LDN19 significantly reduces tHO after murine musculoskeletal injury.18, 19 LDN19, however, inhibits other kinases involved in cell metabolism,²⁰ and it is unclear whether LDN19 reduces tHO through its effect on ACVR1/ALK2 or alternative pathways.

Because tHO has been linked specifically to local inflammation and BMP signaling pathways, it is possible that LDN19 acts through an anti-inflammatory pathway. Prior studies have shown that non-steroidal anti-inflammatory drugs (NSAIDs) are able to reduce tHO after musculoskeletal injury.¹⁸ Wide implementation of anti-inflammatory drugs to prevent tHO is restricted by off-target effects and the need for immediate and continuous therapy.21, 22 Treatment of all trauma patients without improved patient selection risks unnecessary treatment and adverse effects.

Given the high incidence of tHO following high-risk trauma, improved treatment strategies are necessary to prevent tHO.²³ Although the small-molecule T1-BMPR inhibitor LDN19 effectively reduces tHO, we show that it is anti-inflammatory, antiproliferative, and myelosuppressive. We decided to examine the effect of knockout of each T1-BMPR (Acvr1, Bmpr1a, or Bmpr1b) on tHO in mice. All mutant animals developed tHO, indicating that no single receptor is required for this process. However, combined loss of both ACVR1/ALK2 and BMPR1a/ALK3 demonstrated a synergistic and significant reduction in tHO, implying overlap among T1-BMPRs. Based on this information, we evaluated two different BMP signal inhibitors acting by separate mechanisms of action—kinase inhibition (LDN21) or ligand blockade (Alk3-Fc [A3Fc])—to prevent tHO.

The second major barrier to tHO prevention is the inability to predict which trauma patients will progress to form ectopic bone. We have recently developed a scoring assessment to stratify burn patients at risk for tHO.²⁴ Risk assessment, however, remains limited in its predictive capacity. We have shown previously that non-radiographic imaging in mice can detect early cartilage and mineral deposition associated with tHO as early as 2 weeks after injury.25, 26 Here we show that LDN21 initiated in that diagnostic window reduces tHO, allowing us to minimize unnecessary treatment. Alternatively, a therapeutic strategy initiated immediately after injury with A3Fc may have reduced off-target effects. Our approach underscores and addresses the challenges associated with pathologies caused by overlapping receptor function and the need to improve patient selection, reduce therapeutic duration, and mitigate adverse effects.

Results

LDN19 Is a Potent Anti-inflammatory, Anti-proliferative, and Myelosuppressive Agent that Reduces tHO

As demonstrated previously,18, 19 daily LDN19 (6 mg/kg intraperitoneally [i.p.]) for 6 weeks significantly reduced tHO compared with vehicle-treated controls (normalized volume, 0.06 versus 1.0; p = 0.005) (Figures S1A–S1C). Although LDN19 inhibits signaling through ACVR1/ALK2 and reduces pSMAD1/5, it has also been shown to affect signaling through additional receptor kinases, prompting us to evaluate an alternative mechanism of activity.²⁰

Inflammation is ubiquitous in patients with musculoskeletal trauma,27, 28, 29, 30, 31 and we have previously confirmed that anti-inflammatory drugs such as apyrase or cyclooxygenase-2 (COX-2) inhibitors reduce tHO in a burn/tenotomy model.¹⁸ Therefore, we set out to determine whether LDN19 reduces inflammation. Fluorescence-activated cell sorting (FACS) showed a significant reduction in soft tissue neutrophil (normalized count, 0.01 versus 1.0; p = 0.0002) and macrophage (normalized count, 0.04 versus 1.0; p = 0.0003) accumulation at the tenotomy site of mice treated with daily LDN19 for 5 days after injury (Figures 1A–1C). We also found a significant reduction in the proportion of Ly6G⁺CD11b⁺ neutrophils (38% versus 63%, p = 0.017) in the bone marrow of LDN19-treated mice, suggesting a broader myelosuppressive effect (Figures S2A and S2B). The tenotomy site findings were confirmed histologically (Figures 1D–1G). Overall, a reduction in tenotomy site cellularity was observed with LDN19 treatment via FACS (normalized count, 0.04 versus 1.0; p = 0.001) (Figure 1H). However, the observed decrease in neutrophils and macrophages was insufficient to explain the drastic reduction in cellularity. We have shown previously that the majority of cells in the developing HO are mesenchymal cells marked by platelet-derived growth factor receptor alpha (PDGFRα).³² When the site was examined histologically, we noted a substantial reduction in overall cellularity and PDGFRα⁺ cells (Figures 1I and 1J) and a corresponding decrease in proliferation (Figures 1K and 1L). This reduction in proliferation was confirmed in vitro with LDN19 treatment of mesenchymal cells (Figure 1M). Given these effects, we investigated whether LDN19 alters the proliferation of mesenchymal cells independent of ACVR1/ALK2. Indeed, LDN19-treated Acvr1 knockout mesenchymal cells had significantly reduced proliferation compared with vehicle-treated Acvr1 knockout mesenchymal cells (Figure 1N).

Figure 1.

LDN19 Is a Potent Anti-inflammatory, Anti-proliferative, and Myelosuppressive Agent

(A) Flow cytometry gating strategy for tissue-derived neutrophils (Ly6G⁺CD11b⁺F4/80⁻) and macrophages (Ly6G-CD11b⁺F4/80⁺) in vehicle- and LDN19-treated mice. (B) LDN19 significantly reduces tenotomy-site neutrophils 5 days after injury (normalized count, 0.01 versus 1.0; p = 0.0002; n ≥ 3/group). (C) LDN19 significantly reduces tenotomy-site macrophages 5 days after injury (normalized count, 0.04 versus 1.0; p = 0.0003; n ≥ 3/group). (D) Representative neutrophil (Ly6G) immunostaining in a vehicle-treated hindlimb 5 days after injury. (E) Representative neutrophil (Ly6G) immunostaining in an LDN19-treated hindlimb 5 days after injury. (F) Representative macrophage (F4/80) immunostaining in a vehicle-treated hindlimb 5 days after injury. (G) Representative macrophage (F4/80) immunostaining in an LDN19-treated hindlimb 5 days after injury. (H) LDN19 significantly reduces overall cellularity at the tenotomy site 5 days after injury (normalized count, 0.04 versus 1.0; p = 0.001; n ≥ 3/group), including macrophages (dark blue), neutrophils (light blue), and non-leukocytes (white), on the basis of flow cytometry. (I) Representative mesenchymal cell (PDGFRα) immunostaining in a vehicle-treated hindlimb 5 days after injury. (J) Representative mesenchymal cell (PDGFRα) immunostaining in an LDN19-treated hindlimb 5 days after injury. (K) Representative immunostaining for proliferation (Ki67) in a vehicle-treated hindlimb 5 days after injury. (L) Representative immunostaining for proliferation (Ki67) in an LDN19-treated hindlimb 5 days after injury. (M) BrdU proliferation assay showing that LDN19 significantly reduces wild-type mesenchymal cell proliferation in vitro. (N) BrDU proliferation assay showing that LDN19 significantly reduces proliferation of Acvr1 knockout mesenchymal cells in vitro. *p < 0.05. p values are listed for all non-significant findings. All flow cytometry findings were normalized to vehicle-treated controls as indicated in Materials and Methods. Scale bars, 200 μm. Error bars represent one SD.

Loss of ACVR1/ALK2 Fails to Block tHO

Given that LDN19 has effects separate from ACVR1/ALK2 inhibition, we next examined whether ACVR1/ALK2 is necessary for tHO, as has been described for FOP.11, 12 We first evaluated tamoxifen-inducible global Acvr1 knockout mice (Acvr1 tmKO: Ub.creERT/Acvr1^fl/fl) to mimic a systemic approach to ACVR1/ALK2 inhibition, as in the case of highly specific therapy (Figure 2A). Immunostaining confirmed the absence of ACVR1/ALK2 in these mice (Figure S3). However, loss of ACVR1/ALK2 did not result in a significant reduction in tHO compared with tamoxifen-treated littermate control mice (normalized volume, 0.53 versus 1.0; p = 0.09) (Figures 2B–2D). Histologic examination demonstrated minimal differences in cartilage (Alcian blue), chondrocytes (SOX9⁺), or mesenchymal cells (PDGFRα⁺) in tamoxifen-treated Acvr1 tmKO mice compared with tamoxifen-treated littermates (Figures 2E–2J). These findings are in contrast to those observed with LDN19 (Figures S1A–S1C and S4A–S4F). Interestingly, in vitro experiments demonstrate that the osteogenic differentiation of tamoxifen-treated Acvr1 tmKO mesenchymal cells was reduced with LDN19 (Figure S5). These findings provide additional evidence that LDN19 exerts biologically relevant effects independent of ACVR1/ALK2.

Figure 2.

Postnatal Loss of ACVR1/ALK2 Does Not Eliminate tHO

(A) Experimental design for tamoxifen-inducible Acvr1 knockout mice (Acvr1 tmKO: Ub.creERT/Acvr1^fl/fl). (B) Representative 3D microCT reconstructions obtained 9 weeks after injury, showing tHO (blue) at the tenotomy site of Acvr1 littermate control and Acvr1 tmKO mice 9 weeks after injury. (C) Representative serial cross-sections obtained 9 weeks after injury showing tHO (red arrow) at the tenotomy site of Acvr1 littermate control and Acvr1 tmKO mice 9 weeks after injury. (D) Genetic loss of Acvr1 does not eliminate tHO 9 weeks after injury (normalized volume, 0.53 versus 1.0; p = 0.09; n ≥ 3/group). (E) Representative pentachrome image confirming the presence of cartilage marked by Alcian blue in a littermate control mouse 3 weeks after injury. (F) Representative pentachrome image confirming the presence of cartilage marked by Alcian blue in an Acvr1 tmKO mouse 3 weeks after injury. (G) Representative chondrocyte (SOX9) immunostaining confirming the presence of chondrocytes in a littermate control mouse 3 weeks after injury. (H) Representative chondrocyte (SOX9) immunostaining confirming the presence of chondrocytes in an Acvr1 tmKO mouse 3 weeks after injury. (I) Representative mesenchymal cell (PDGFRα) immunostaining confirming the presence of mesenchymal cells in a littermate control mouse 5 days after injury. (J) Representative mesenchymal cell (PDGFRα) immunostaining confirming the presence of mesenchymal cells in an Acvr1 tmKO mouse 5 days after injury. *p < 0.05. p values are listed for all non-significant findings. All volumes were normalized to tamoxifen-treated littermate controls as indicated in Materials and Methods. Scale bars, 200 μm. Error bars represent one SD.

Although immunostaining of Acvr1 tmKO mice confirmed the absence of ACVR1/ALK2 (Figure S3), we evaluated a second model of Acvr1 conditional knockout in mesenchymal cells (Acvr1 cKO: Prx-cre/Acvr1^fl/fl) to confirm our findings (Figure S6A). Our previous studies have shown that Prxcre comprise over 90% of cells in the early HO lesion and include PDGFRα⁺ mesenchymal cells, chondrocytes, and osteoblasts.³² However, we did not detect a significant or substantial reduction in tHO volume in these mice compared with littermate control mice (normalized volume, 0.71 versus 1.0; p = 0.66) (Figures S6B–S6D). Again, consistent with previously published literature, Acvr1 cKO mice had thicker tibial sections compared with littermate controls (Figures S6E–S6G).

Loss of BMPR1a/ALK3 or BMPR1b/ALK6 Fails to Prevent tHO

We next sought to determine whether an alternative T1-BMPR, BMPR1a/ALK3, is responsible for tHO. Genetic loss of BMPR1a has recently been shown to reduce periosteal bone growth.³³ We used a similar strategy as above to knock out Bmpr1a globally upon tamoxifen administration (Bmpr1a tmKO: Ub.creERT/Bmpr1a^fl/fl) (Figure 3A). Knockout was confirmed histologically by immunostaining for BMPR1a/ALK3 (Figure S7) and by the presence of an identifiable phenotype evidenced by hair loss and cachexia. Systemic postnatal loss of BMPR1a/ALK3 was lethal in the majority of these mice prior to study completion 9 weeks after injury (Figure S8). These mice continued to form tHO despite the striking deleterious effects of systemic BMPR1a/ALK3 loss (normalized volume, 1.43 versus 1.0; p = 0.27) (Figures 3B–3D). Histologic examination confirmed the presence of cartilage 3 weeks after injury in tamoxifen-treated Bmpr1a tmKO mice (Figures 3E and 3F). These results were confirmed in a genetic model of conditional Bmpr1a loss (Figures S9A and S9B). Overall, these findings indicate that highly efficient systemic inhibition of BMPR1a/ALK3 is not a viable therapeutic strategy.

Figure 3.

Postnatal Loss of BMPR1a/ALK3 or BMPR1b/ALK6 Knockout Alone Is Unable to Eliminate tHO

(A) Experimental design for tamoxifen-inducible Bmpr1a knockout mice (Bmpr1a tmKO: Ub.creERT/Bmpr1a^fl/fl) and for Bmpr1b knockout (Bmpr1b^−/−) mice. (B) Representative 3D microCT reconstructions showing HO (blue) at the tenotomy site of littermate control and Bmpr1a tmKO mice 9 weeks after injury. (C) Representative serial cross-sections showing HO (red arrow) at the tenotomy site of littermate control and Bmpr1a tmKO mice 9 weeks after injury. (D) Genetic loss of Bmpr1a does not substantially or significantly reduce tHO 9 weeks after injury (normalized volume, 1.43 versus 1.0; p = 0.27; n ≥ 3/group). (E) Representative pentachrome image confirming the presence of cartilage marked by Alcian blue in a littermate control mouse 3 weeks after injury. (F) Representative pentachrome image confirming the presence of cartilage marked by Alcian blue in a Bmpr1a tmKO mouse 3 weeks after injury. (G) Representative 3D microCT reconstructions showing HO (blue) at the tenotomy site of littermate control and Bmpr1b^−/− mice 9 weeks after injury. (H) Representative serial cross-sections showing HO (red arrow) at the tenotomy site of littermate control and Bmpr1b^−/− mice 9 weeks after injury. (I) Genetic loss of Bmpr1b does not substantially or significantly reduce tHO 9 weeks after injury (normalized volume, 0.81 versus 1.0; p = 0.72; n ≥ 3/group). (J) Representative pentachrome image confirming the presence of cartilage marked by Alcian blue in a Bmpr1b littermate control mouse 3 weeks after injury. (K) Representative pentachrome image confirming the presence of cartilage marked by Alcian blue in a Bmpr1b^−/− mouse 3 weeks after injury. *p < 0.05. p values are listed for all non-significant findings. All volumes were normalized to respective littermate controls as indicated in Materials and Methods. Scale bars, 200 μm. Error bars represent one SD.

Finally, we evaluated a third T1-BMPR, BMPR1b/ALK6, using global knockout mice from birth (Bmpr1b^−/−). These mice are viable despite embryologic loss of Bmpr1b.³⁴ These mice had slightly reduced but persistent tHO 9 weeks after injury (normalized volume, 0.81 versus 1.0; p = 0.72) (Figures 3G and 3I) and ectopic cartilage 3 weeks after injury (Figures 3J and 3K).

Loss of ACVR1/ALK2 and BMPR1a/ALK3 Significantly Reduces tHO

Taken together, our findings indicate overlap in the contributions of T1-BMPRs (ACVR1/ALK2, BMPR1a/ALK3, and BMPR1b/ALK6) in tHO. Given these findings, we next tested whether loss of multiple T1-BMPRs can eliminate tHO. Tamoxifen-inducible double knockout mice were generated to eliminate ACVR1/ALK2 and BMPR1a/ALK3 (Acvr1;Bmpr1a tmKO:Ub.creERT/Acvr1^fl/fl;Bmpr1a^fl/fl) (Figure 4A). As might be expected, these mice demonstrated reduced histologic evidence of BMP signaling (pSMAD1/5) at the wound site (Figure S10). Similar to Bmpr1a tmKO mice, Acvr1;Bmpr1a tmKO mice had a striking phenotype including cachexia, hair loss, and lethality (Figures 4B and 4C). Mice that survived produced minimal tHO based on micro-computed tomography (microCT) (normalized volume, 0.06 versus 1.0; p = 0.025) (Figures 4D–4F) or histologic evaluation (Figures 4G and 4H). In contrast to littermate controls, histologic evaluation of Acvr1;Bmpr1a tmKO mice from 3–6 weeks after injury showed minimal evidence of cartilage (Figures 4I and 4J). These findings indicate that targeting multiple T1-BMPRs is required to eliminate tHO.

Figure 4.

Combined Postnatal Loss of ACVR1/ALK2 and BMPR1a/ALK3 Significantly Reduces tHO but Is Lethal

(A) Experimental design for tamoxifen-inducible Acvr1 and Bmpr1a double knockout mice (Acvr1;Bmpr1a tmKO: Ub.creERT/Acvr1^fll/fl/Bmpr1a^fl/fl). (B) Representative whole-body photos of Acvr1/Bmpr1a littermate control and mutant mice showing hair loss and cachexia in the Acvr1;Bmpr1a tmKO mouse. (C) A significant proportion (7 of 9) of Acvr1/Bmpr1a tmKO mice died within the first 6 weeks after tamoxifen induction. (D) Representative 3D microCT reconstructions showing HO (blue) at the tenotomy site of an Acvr1;Bmpr1a littermate control mouse and absent HO in the Acvr1;Bmpr1a tmKO mouse 9 weeks after injury. (E) Representative serial cross-sections showing HO (red arrow) at the tenotomy site of Acvr1;Bmpr1a littermate control and Acvr1;Bmpr1a tmKO mice 9 weeks after injury. (F) Genetic loss of Acvr1 and Bmpr1a significantly reduces tHO 9 weeks after injury (normalized volume, 0.06 versus 1.0; p = 0.025; n ≥ 3 for littermate control and n = 2 for double knockout). (G) Representative pentachrome image showing the histologic presence of bone and cartilage at the tenotomy site of an Acvr1;Bmpr1a littermate control mouse 9 weeks after injury. (H) Representative pentachrome image showing a markedly reduced presence of cartilage at the tenotomy site of an Acvr1;Bmpr1a tmKO mouse 9 weeks after injury. (I) Representative pentachrome image confirming the presence of cartilage (Alcian blue) at the tenotomy site of an Acvr1;Bmpr1a littermate control mouse 3 weeks after injury. (J) Representative pentachrome image confirming the relative absence of cartilage at the tenotomy site of an Acvr1;Bmpr1a tmKO mouse 3 weeks after injury. *p < 0.05. p values are listed for all non-significant findings. All volumes were normalized to tamoxifen-treated littermate controls as indicated in Materials and Methods. Scale bars, 200 μm. Error bars represent 1 SD.

BMP Ligand Blockade Reduces tHO without Reducing the Cortical Thickness of Normal Anatomic Bone

These findings indicate that knockout of multiple T1-BMPRs is sufficient to prevent tHO in a mouse model. We then examined whether pharmaceutical inhibition of T1-BMPRs can be achieved while limiting deleterious effects. A major drawback of LDN19 is its off-target inhibition of other receptor tyrosine kinases,²⁰ leading to profound immunosuppression, reduced cellular proliferation at the wound site, and loss of bone thickness—all consequences that cannot be tolerated in trauma or burn patients at risk for tHO.29, 35, 36, 37, 38 Therefore, we turned our attention to an alternative strategy to inhibit BMP signaling using the BMP ligand trap A3Fc (Figure 5A).39, 40

Figure 5.

Long-Term Treatment with A3Fc Significantly Reduces tHO

(A) A3Fc functions as a BMP ligand trap. (B) Representative 3D microCT reconstructions showing HO (blue) at the tenotomy site of vehicle- and A3Fc-treated mice 9 weeks after injury. (C) Representative serial cross-sections showing HO (red arrow) at the tenotomy site of vehicle- and A3Fc-treated mice 9 weeks after injury. (D) Long-term treatment with A3Fc (daily 2 mg/kg i.p. for 6 weeks) significantly reduces tHO (normalized volume, 0.35 versus 1.0; p = 0.001; n ≥ 3/group). (E) Representative pentachrome image confirming the presence of cartilage marked by Alcian blue in a vehicle-treated mouse 3 weeks after injury. (F) Representative pentachrome image showing a relative decrease in cartilage marked by Alcian blue in an A3Fc-treated mouse 3 weeks after injury. (G) Representative chondrocyte (SOX9) immunostaining confirming the presence of chondrocytes in a vehicle-treated mouse 3 weeks after injury. (H) Representative chondrocyte (SOX9) immunostaining confirming the near absence of chondrocytes in an A3Fc-treated mouse 3 weeks after injury. (I) Representative mesenchymal cell (PDGFRα) immunostaining confirming the presence of mesenchymal cells in a vehicle-treated mouse 5 days after injury. (J) Representative mesenchymal cell (PDGFRα) immunostaining confirming the presence of mesenchymal cells in an A3Fc-treated mouse 5 days after injury. (K) Representative microCT cross-section showing the tibia-fibula confluence of the injured hindlimb in vehicle-treated mice 9 weeks after injury. (L) Representative microCT cross-section showing the tibia-fibula confluence of the injured hindlimb in A3Fc-treated mice 9 weeks after injury. (M) A3Fc does not reduce the cortical thickness of the injured hindlimb. *p < 0.05. p values are listed for all non-significant findings. All volumes were normalized to vehicle-treated controls as indicated in Materials and Methods. Scale bars, 200 μm. Error bars represent one SD.

First we confirmed that A3Fc treatment reduces histologic evidence of BMP signaling (pSMAD1/5) at the tenotomy site (Figure S11). Correspondingly, we found that mice treated daily with A3Fc (2 mg/kg i.p.) for 6 weeks after injury demonstrated a significant reduction in tHO volume (normalized volume, 0.35 versus 1.0, p = 0.001) (Figures 5B–5D). This corresponded with a histologic decrease in cartilage and chondrocytes (SOX9⁺) (Figures 5E–5H). However, A3Fc preserved the presence of mesenchymal cells (PDGFRα⁺) at the tenotomy site (Figures 5I and 5J). In vitro and in vivo proliferation remained intact, although minimally reduced, with A3Fc treatment as well (Figure S12A–S12D). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assays for mesenchymal cells (PDGFRα⁺TUNEL⁺) in vivo demonstrate no substantial difference at 1 or 3 weeks post-injury (Figure S13). Macrophage and neutrophil counts at the wound site, overall cellularity at the wound site, and neutrophil counts in the bone marrow remained unsubstantially altered by A3Fc treatment (Figures S14A–S14D). The cortical thickness of tibial bone was not reduced with A3Fc (Figures 5K–5M).

Abbreviated A3Fc Treatment Is Sufficient to Reduce tHO when Administered Early after Injury

Reduced treatment efficacy, in part because of poor patient compliance with long-term therapy, has been reported for other conditions affecting trauma patients.41, 42, 43 Therefore, we sought to establish a translational protocol for abbreviated treatment using A3Fc (Figure 6A). We found that immediate daily treatment with A3Fc for 2 weeks was sufficient to significantly and substantially reduce tHO (normalized volume, 0.43 versus 1.0; p = 0.021) (Figures 6B and 6C). Furthermore, this reduction with short-term treatment was not significantly different compared with long-term treatment (normalized volume, 1.21 versus 1.0, p = 0.47) (Figures 5B–5D and 6B and 6C). Taken together, these findings suggest that immediate short-term treatment with A3Fc safely prevents tHO, providing a therapeutic option for patients.

Figure 6.

A Translational Approach to Preventing tHO by Targeting BMP Signaling

(A) Experimental design for short-term daily treatment initiated immediately after injury, during weeks 2–4 after injury, when spectroscopic evidence of cartilage and mineral deposition is observed, or during weeks 4–6 after injury, when radiographic evidence of the ossified lesion is observed (the red dashed arrow indicates no treatment; the gray dashed arrow indicates post-treatment; the solid black arrow indicates active treatment). (B) Representative 3D microCT reconstructions obtained 9 weeks after injury showing HO (blue) at the tenotomy site of mice treated with daily A3Fc during weeks 0–2, weeks 2–4, or weeks 4–6 after injury. (C) Short-term treatment with A3Fc (daily 2 mg/kg i.p. for 2 weeks) during weeks 0–2 significantly reduces tHO (normalized volume, 0.43 versus 1.0; p = 0.021; n ≥ 3/group) but not during weeks 2–4 (normalized volume, 0.63 versus 1.0; p = 0.25; n ≥ 3/group) or weeks 4–6 (normalized volume, 0.87 versus 1.0; p = 0.90; n ≥ 3/group). (D) Representative 3D microCT reconstructions obtained 9 weeks after injury showing HO (blue) at the tenotomy site of mice treated with daily LDN21 (6 mg/kg i.p.) for 6 weeks initiated immediately after injury. (E) Long-term treatment with LDN21 initiated immediately after injury significantly decreases tHO (normalized volume, 0.34 versus 1.0; p = 0.005; n ≥ 3/group). (F) Representative 3D microCT reconstructions obtained 9 weeks after injury showing HO (blue) at the tenotomy site of mice treated with daily LDN21 during weeks 0–2, weeks 2–4, or weeks 4–6 after injury. (G) Short-term treatment with LDN21 (daily 2 mg/kg i.p. for 2 weeks) during weeks 0–2 (normalized volume, 0.51 versus 1.0; p = 0.043; n ≥ 3/group) or weeks 2–4 (normalized volume, 0.38 versus 1.0; p = 0.032; n ≥ 3/group) but not during weeks 4–6 (normalized volume, 0.57 versus 1.0; p = 0.18; n ≥ 3/group) significantly reduces tHO. *p < 0.05. p values are listed for all non-significant findings. All volumes were normalized to the control (vehicle-treated) as indicated in Materials and Methods. Scale bars, 200 μm. Error bars represent one SD.

Although these findings are encouraging, only a small proportion of trauma patients will go on to develop tHO.²⁴ Therefore, short-term treatment guided by improved patient selection would substantially reduce the number of patients needed to treat (NNT). We sought to identify whether A3Fc treatment could be delayed until either spectroscopic (2 weeks after injury) or radiographic evidence of tHO (4 weeks after injury) is present to improve patient selection.25, 26 However, daily A3Fc did not significantly reduce tHO when administered from weeks 2–4 (normalized volume, 0.63 versus 1.0, p = 0.25) (Figure 6C) or from weeks 4–6 (normalized volume, 0.87 versus 1.0; p = 0.90) (Figure 6C), indicating that delayed, short-term treatment with A3Fc is not an effective strategy.

A “Second-Generation” Receptor Tyrosine Kinase Inhibitor Reduces tHO with Abbreviated Delayed Treatment

To validate that delayed treatment to prevent tHO is feasible, we performed a similar treatment series with LDN19. Short-term daily treatment with LDN19 was effective both when initiated immediately after injury for 2 weeks (normalized volume, 0.14 versus 1.0, p = 0.004) (Figures S15A and S15B) or when initiated 2 weeks after injury for 2 weeks (normalized volume, 0.53 versus 1.0 p = 0.026) (Figures S15A and S15B). To address the adverse effects observed with LDN19, we used a second small-molecule T1-BMPR inhibitor, LDN-212854 (LDN21). Similar to LDN19 and A3Fc, LDN21 demonstrated reduced BMP signaling (pSMAD1/5) after trauma (Figure S16). Daily treatment with LDN21 (6 mg/kg i.p.) for 6 weeks significantly reduced tHO (normalized volume, 0.34 versus 1.0; p = 0.005) (Figures 6D and 6E). Although LDN21 had an anti-inflammatory effect (Figures S17A and 17B), it did not cause appreciable myelosuppression based on marrow neutrophil content versus an untreated control (61% versus 63%; p = 0.07) (Figure S17C). Cellularity at the wound site was reduced with LDN21 (normalized count, 0.75 versus 1.0; p = 0.05) (Figure S18A), as were in vivo and in vitro proliferation (Figure S18B–S18D) and normal tibial bone thickness (Figures S18E–S18G). Similar to A3Fc, short-term daily treatment with LDN21 from weeks 0–2 (normalized volume, 0.51 versus 1.0; p = 0.043) (Figures 6F and 6G) significantly reduced tHO, allowing for the possibility of reduced treatment duration. In contrast to A3Fc, however, treatment with LDN21 during weeks 2–4 also significantly reduced tHO (normalized volume, 0.38 versus 1.0; p = 0.032) (Figures 6F and 6G). Taken together, these findings suggest that LDN21 may be better tolerated than LDN19 because of the absence of myelosuppression and, unlike A3Fc, is amenable to delayed treatment, allowing for improved patient selection with nonradiographic diagnostic tools.25, 26

Discussion

The T1-BMPR ACVR1/ALK2 has received considerable attention in recent studies of HO because of its central role in FOP. Patients with FOP have a hyperactivating ACVR1 mutation (ACVR1 R206H) and a striking phenotype with the development of ectopic bone lesions after only minor trauma.11, 12, 13, 44 However, patients with tHO are not known to harbor mutations of ACVR1, suggesting that tHO is a related disease with a different contribution from BMP ligands and their receptors. These ligands activate multiple T1-BMPRs, including BMPR1a/ALK3 and BMPR1b/ALK6. Therefore, it is unclear whether the therapeutic emphasis placed on ACVR1/ALK2 confers a clinically substantial benefit to patients at risk for tHO after musculoskeletal trauma.

In this study, we used a mouse model of tHO18, 32, 45, 46 that reliably generates heterotopic bone at the site of direct trauma (Achilles tenotomy). However, tHO may not always occur in sites of obvious injury, such as in patients with HO following spinal cord injury (SCI) or severe burns. In these patients, HO may develop at sites distant from the obvious injury. Although it is possible that the burn/tenotomy model used in our study is not entirely representative of these clinical scenarios, we believe that these instances of HO are likely mediated by similar differentiation processes governed by BMP activity. In addition, these scenarios all share a heightened systemic inflammatory state because of injury. Furthermore, a common factor in each of these is the absence of a definitive causative mutation, allowing for a contribution through functional redundancy of receptor activity. Our model of tHO allows us to thus examine the contribution of multiple different T1-BMPRs to this pathology and generate a translational approach to preventing this challenging disease process.

Our findings indicate that LDN19, a small-molecule kinase inhibitor traditionally used in the literature to study ACVR1/ALK2 signaling,16, 17, 20, is a potent anti-inflammatory, anti-proliferative, and myelosuppressive agent. We confirmed our previous findings that LDN19 reduces tHO when administered daily for 6 weeks after injury.¹⁸ Upon further examination, we found that LDN19 profoundly reduced wound-site neutrophils and macrophages, marrow neutrophils, and mesenchymal proliferation in vitro and in vivo. Furthermore, LDN19 reduced the proliferation and osteogenic differentiation of Acvr1 knockout mesenchymal cells, confirming that LDN19 imparts substantial effects independent of ACVR1/ALK2. Additionally, LDN19 reduced normal cortical thickness, in contrast to mice with loss of ACVR1/ALK2. Taken together, these findings suggest that LDN19 may prevent tHO through mechanisms outside of the ACVR1/ALK2 pathway, possibly through a reduction in inflammation or proliferation. These findings are consistent with studies showing that LDN19 affects non-BMP receptor kinases.¹⁸ Although potentially valuable in other proliferative disease processes, in the context of trauma, loss of the inflammatory response, cellular proliferation, or bone volume can be detrimental to wound healing and recovery.47, 48, 49, 50, 51 In the context of the severe burns and/or musculoskeletal polytrauma that are commonly associated with ectopic bone formation, these anti-inflammatory properties may limit the use of drugs such as LDN19 as treatment or prevention for tHO. This ambiguity in the mechanism through which LDN19 reduces tHO led us to examine the role of BMP signaling in tHO.

We found that isolated loss of ACVR1/ALK2, BMPR1a/ALK3, or BMPR1b/ALK6 is unable to eliminate tHO. Although combined loss of ACVR1/ALK2 and BMPR1a/ALK3 nearly eliminates tHO, it leads to substantial toxicity and is lethal in a majority of mice within 6 weeks after induced knockout. Even with combined loss of ACVR1/ALK2 and BMPR1a/ALK3, a small amount of cartilage can be observed, suggesting that this process may be rescued even by the presence of BMPR1b/ALK6. Other studies have also shown overlap between BMP receptors in developmental processes.34, 52 Although hyperactivating mutations in ACVR1/ALK2 are responsible for FOP, this is likely caused by an overall increase in BMP signaling and pSMAD 1/5/8 levels. In the setting of trauma, BMP transcripts are upregulated and may lead to broad upregulation of pSMAD 1/5 through multiple T1-BMPRs and not only through ACVR1/ALK2. Overlapping contributions of T1-BMPRs to pathology highlight the challenge of targeting BMP signaling to treat diseases in which BMP ligands are implicated, such as trauma-induced HO and atherosclerosis.⁵³ However, our findings suggest that prevention of tHO through inhibition of BMP signaling requires broad T1-BMPR inhibition beyond a single receptor. Novel techniques to mitigate adverse effects, including reduced duration of treatment and improved patient selection, are required to translate the relationship between BMP signaling and pathology into therapeutic strategies.

Given these findings, we chose to evaluate two inhibitors of BMP signaling with differing mechanisms of action—A3Fc and LDN21. A3Fc is a ligand trap composed of the ligand-binding region of BMPR1a/ALK3. Because of similarities in BMP ligands bound by ACVR1/ALK2, BMPR1a/ALK3, and BMPR1b/ALK6,14, 15 we reasoned that this drug would broadly affect these three T1-BMPRs because of shared ligands among the three receptors. We also reasoned that this drug would remain specific to the BMP pathway because it is composed specifically of the ligand-binding region of BMPR1a/ALK3.39, 54, 55 A3Fc reduced tHO in vivo but did not cause myelosuppression or immunosuppression as observed with LDN19. Additionally, A3Fc did not substantially reduce cellular proliferation or normal bone cortical thickness, consistent with previous literature.³⁹ These findings are critical in trauma patients who are at risk for bone loss and extensive wounds36, 37, 38 and suggest that this drug may have fewer off-target effects compared with LDN19, therefore improving its translational potential.

LDN21 is a tyrosine receptor kinase similar to LDN19, with improved specificity for ACVR1/ALK2.²⁰ Because of this improved specificity for ACVR1/ALK2, we did not expect LDN21 to substantially reduce tHO. Although also anti-inflammatory, LDN21 did not have the same degree of anti-inflammatory or myelosuppressive effects as LDN19. It did, however lead to a reduction in normal cortical bone thickness, an undesirable effect in trauma and burn patients who are already susceptible to bone loss. LDN21 also reduced proliferation of Acvr1 knockout mesenchymal cells in vitro and mesenchymal cells at the injury site in vivo. These findings and the striking reduction in tHO suggest that LDN21, similar to LDN19, may exert its effects through a mechanism independent of ACVR1/ALK2. These findings led us to further examine the translational potential of A3Fc and LDN21 to prevent tHO.

Based on our clinical experience with trauma patients, we recognize that patient compliance with long-term treatment is challenging.41, 42, 43 We found that short-term A3Fc initiated immediately after injury also significantly and substantially reduced tHO. We recognize that it is not possible to predict which trauma patients will develop tHO—although tHO causes substantial morbidity in thousands of patients, its incidence among all-comer trauma patients remains low. Therefore, a therapeutic approach that improves patient selection for treatment would present a substantial advance. However, patients who develop tHO often present late when symptomatic ossified lesions can be diagnosed radiographically. We have recently validated Raman spectroscopy as a non-invasive, point-of-care technique that can identify cartilage and mineral changes associated with tHO within 2 weeks after injury.²⁶ Therefore, we tested delayed, short-term daily treatment with A3Fc during weeks 2–4 after injury to simulate a treatment paradigm initiated upon spectroscopic diagnosis or during weeks 4–6 after injury to simulate radiographic diagnosis. However, A3Fc did not reduce tHO with either delayed treatment option. Although these findings could be attributed to dosing or frequency, the striking effects observed with early treatment (weeks 0–2) suggest that our dosing regimen (2 mg/kg daily) is reasonable. Taken together, these findings indicate that A3Fc is a potentially effective and safe therapeutic option when initiated immediately at the time of injury. The trade-off, however, is that treatment must be implemented broadly across trauma patients, which may lead to unnecessary treatment or a high NNT.

Finally, we evaluated the same series of treatment strategies with LDN21. Short-term daily treatment during weeks 0–2 or weeks 2–4 both significantly and substantially reduced tHO. Although LDN21 reduces cellular proliferation and reduces tibial cortical thickness, both undesirable effects in trauma patients, it can be administered in a directed manner to patients who demonstrate spectroscopic evidence of tHO (weeks 2–4). LDN21 offers a translational paradigm with improved patient selection through early diagnosis and reduced treatment duration. The trade-off, however, is that LDN21 can affect cellular proliferation and reduce cortical bone thickness, but it is unclear if this is an off-target effect. The efficacy of LDN-212854 suggests that a T1-BMPR kinase inhibition strategy might be effective when administered in a delayed manner to patients who demonstrate spectroscopic evidence of tHO (weeks 2–4) and offers a translational paradigm with improved patient selection through early diagnosis and reduced treatment duration; it will need to be investigated whether newer compounds inhibiting ACVR1/ALK2, BMPR1A/ALK3, and BMPR1B/ALK6, but with fewer off-target effects outside of the BMP pathway, will retain this efficacy without impacting cellular proliferation and orthotopic bone cortical thickness. Similar to A3Fc, our findings should be interpreted in the context of the dose used (6 mg/kg i.p.). However, the striking inhibitory effect on tHO at this dose with long-term daily treatment and short-term daily treatment suggests that the selected dose is appropriate for our studies.

In this study, we demonstrate a related but unique signaling pathway compared with FOP with overlap among T1-BMPRs in the development of tHO. These findings are critical to the development of therapies that prevent this challenging disease process by broadly targeting BMP signaling. We have developed a translational paradigm that underscores the challenges of treating diseases that cause substantial morbidity but have low prevalence among a broadly defined population of at-risk patients. The challenges associated with early diagnosis, therapeutic morbidity, and efficacy affect our management of other disease processes with evolving risk profiles, such as deep vein thrombosis (DVT), prostate cancer, and atherosclerosis. A translational strategy should have high therapeutic efficacy, allow for improved patient selection, and have reduced adverse or off-target effects. We anticipate that strategies in which all patients receive a short-term treatment with A3Fc or in which patients with early diagnostic evidence of tHO receive delayed short-term treatment with LDN21 provide a framework for approaching this condition.

Materials and Methods

Animal Use

All animal procedures were carried out in accordance with the guidelines provided in the Guide for the Use and Care of Laboratory Animals from the Institute for Laboratory Animal Research (ILAR, 2011) and were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Michigan (PRO0005909). All animals were housed in IACUC-supervised facilities and did not to exceed four mice housed per cage at 18°C–22°C, 12-hr light-dark cycle with ad libitum access to food and water. For all in vitro and in vivo studies requiring wild-type mice, young adult male (6–8 weeks old) C57BL/6J mice were purchased from Charles River Laboratories.

Transgenic Strains

C57BL/6 background Prx-cre/Acvr1^fl/fl (and littermate control) and Scx-cre/Bmpr1a^fl/fl (and littermate control) mice were bred by crossing Prx-cre x Acvr1^fl/fl (or Scx-cre x Bmpr1a^fl/fl), selecting heterozygous floxed-mice (Prx-cre/Acvr1^fl/wt or Scx-cre/Bmpr1a^fl/wt) and crossing again with the respective knockouts (Acvr1^fl/fl or Bmpr1a^fl/fl). Similarly, tamoxifen-inducible Acvr1 and Bmpr1a knockout mice (Acvr1 tmKO: Ub.creERT/Acvr1^fl/fl, Bmpr1a tmKO: Ub.creERT/Bmpr1a^fl/fl, and Acvr1;Bmpr1a tmKO: Ub.creERT/Acvr1^fl/fl/Bmpr1a^fl/fl) and littermate controls were produced. Littermate control mice were Ub.creERT(−) or floxed heterozygotes. Tamoxifen (7 mg /40 g) was administered intraperitoneally to all tamoxifen-inducible knockout mice and littermate controls 7 and 3 days prior to burn/tenotomy; Acvr1 tmKO and littermate control mice received additional tamoxifen 7 days after injury. Acvr1;Bmpr1a tmKO and Bmpr1a tmKO and respective littermate control mice were not injected with tamoxifen 7 days after injury because of frailty and risk of death. Bmpr1b^−/− mice were obtained by crossing Bmpr1b^+/−mice. PCR was used to confirm all genotypes.

Burn/Tenotomy tHO Model

A partial-thickness scald burn injury was administered to animals according to a previously described protocol.18, 45 Briefly, mice were anesthetized with inhaled isoflurane. Dorsal hair was closely clipped, and an aluminum block heated to 60°C was exposed to the dorsal region over 30% of the total body surface area for 18 s to achieve a partial-thickness burn injury. Each mouse then received a concurrent sterile dorsal hindlimb tendon transection at the midpoint of the Achilles tendon (Achilles tenotomy) with placement of a single 5-0 vicryl suture to close the skin. Pain management was achieved with subcutaneous injections of buprenorphine (Buprenex, Reckitt Benckiser Pharmaceuticals) every 12 hr for 3 days.

Drug Treatment

Wild-type, adult male C57BL/6 mice (age 8–10 weeks) underwent burn/tenotomy injury as described above, followed by daily intraperitoneal injection of LDN19 (6 mg/kg in 500 μL of PBS), LDN21 (6 mg/kg in 500 μL of PBS), A3Fc (2 mg/kg in 500 μL of PBS; Acceleron Pharma), or saline (500 μL). A separate group of mice received daily injection of LDN19 (6 mg/kg), LND21 (6 mg/kg), or A3Fc (2 mg/kg) during weeks 0–2, 2–4, or 4–6 after burn/tenotomy. Each experimental group had n ≥ 3 animals. Mice receiving continuous drug therapy were sacrificed at either 1, 3, or 9 weeks for further analysis. Animals in the week 0–2, week 2–4, or week 4–6 treatment groups were carried to 9 weeks before sacrifice and analysis.

Isolation and Culture of Primary Cells

Mouse adipose-derived derived MSCs (AdMSCs) and osteoblasts were harvested from the inguinal fat pads and femora and tibiae, respectively, of Acvr1^fl/fl littermate control or Ub.creERT/Acvr1^fl/fl mice. Cells were passaged three times before being used for proliferation and osteogenic assays. Tamoxifen induction was performed in vitro with 4-hydroxytamoxifen (1 μM final concentration) over the course of 6 days.

Cellular Proliferation

Mouse AdMSCs from wild-type and Acvr1 tmKO animals were seeded in 12-well plates at a density of 5 × 10³ cells/well performed in triplicate with or without LDN19 (750 nM), LDN21 (750 nM), or A3Fc. Cells were grown in standard growth medium (DMEM with 10% fetal bovine serum). At 12, 24, 48, 72, and 96 hr, cell proliferation was assessed by bromodeoxyuridine (BrdU) incorporation as described previously.⁵⁶

In Vitro Osteogenic Differentiation

Mouse AdMSCs from wild-type and Acvr1 tmKO animals were seeded in triplicate onto a 12-well plate at a density of 35,000 cells/well in triplicate.57, 58, 59 After 24 hr, the medium was changed to osteogenic differentiation medium (ODM) containing DMEM, 10% fetal bovine serum, 100 g/mL ascorbic acid, 10 mM glycerophosphate, and 100 IU/mL penicillin and 100 μg/ml streptomycin.26, 57 AdMSCs were treated with LDN19 (750 nM) or vehicle control supplemented to ODM. Early osteogenic differentiation was assessed by alkaline phosphatase (ALP) stain and quantification of ALP enzymatic activity after 7 days in ODM as described previously.⁵⁶

MicroCT analysis

In vivo development of HO was assessed with longitudinal microCT scans 9 weeks post-injury (GE Healthcare Biosciences, using 80 peak kilovoltage (kVp), 80 mA, and 1,100-ms exposure). Images were reconstructed, and HO volume formation was analyzed using a calibrated imaging protocol as described previously with the MicroView microCT viewer (Parallax Innovations).²⁶ Calculation of cortical thickness was performed in MicroView. Briefly the fusepoint of the tibia and fibula in the uninjured right leg was selected as a reference landmark. A region encompassing only the proximal tibia was defined as the region of interest. The mean cortical thickness of the given region was then determined automatically with a 1,800-Hounsfield unit threshold cutoff.

Tissue Isolation and Digestion

Tissue for flow cytometry analysis was isolated from the marrow and tendon transection site using standardized techniques. For marrow studies, the femoral medullary cavity was flushed with Hank’s balanced salt solution (HBSS) three times. To isolate the tenotomy site, tissue from the calcaneus (excluding bone) to the convergence of the tendon with calf musculature was excised, minced at 4°C, and digested. The tissue was digested for 120 min in 2 mg/mL Collagenase 3 (Worthington Biochemical) in HBSS at 37°C under constant agitation. Following tissue digestion, the samples were filtered using a 70-μm sterile nylon mesh cell strainer, and the cell suspension was centrifuged at 800 rpm for 5 min before removing the supernatant and resuspending the cells in HBSS. This process was repeated three times before incubation with fluorescently labeled antibodies.

Flow Cytometry

Flow cytometry was performed to quantify the number of neutrophils (Ly6G⁺CD11b⁺), macrophages (Ly6G-CD11b⁺F4/80⁺), and all cells (DAPI⁺) in each sample. The antibodies used included Ly6G-APC (BD Biosciences), CD11b-V450 (BD Biosciences), and F4/80-PECy7 (eBioscience). Following 1 hr of incubation at 4°C, the samples were washed and filtered through a 45-μm nylon mesh filter before being run on a FACSAria II cell sorter (BD Biosciences) at the University of Michigan Flow Cytometry Core in the Biomedical Science Research Center. Samples were gated to separate debris and autofluorescent signals from the cell population. Data were then analyzed using FlowJo software (Tree Star).

Histologic Processing and Analyses

1, 3, or 9 weeks post-injury, animals were euthanized for histology. The distal hindlimb was removed by sharp dissection at the hip. The skin was removed carefully to leave the injury site undisturbed, and the toes were removed to facilitate rapid decalcification. Decalcification of the sample was completed with 19% EDTA solution for 28 days at 4°C. Decalcified tissues were dehydrated through graded ethanol and paraffin-embedded. Transverse sections from all samples were completed with a thickness of 5 μm and mounted on charged microscope slides (Globe Scientific). Paraffin samples were dried overnight at 37°C. A representative subset of samples was stained with Movat’s pentachrome stain to confirm the anatomy and identify gross areas of tHO.

Immunohistochemistry

Immunofluorescent staining was performed as described previously for the following primary antibodies: anti-phospho-Smad1/5/8 (Santa Cruz Biotechnology), anti-ACVR1 (Abcam), anti-BMPR1a (EMD Millipore), anti-SOX9 (Santa Cruz), anti-PDGFRα (Santa Cruz), TUNEL (Roche), and anti-Ki67 (EMD Millipore). Briefly, sections were de-paraffinized and rehydrated in xylenes and graded ethanol. Antigen retrieval was performed with citrate solution (pH 6.0). Samples were then quenched for autofluorescence in 3% glycine before blocking and permeabilization. Primary antibodies were applied overnight at 4°C. Appropriate dilutions were determined prior to achieving final images. After washing, fluorescently conjugated secondary antibodies were tagged with donkey anti-rabbit immunoglobulin G (IgG) and Alexa Fluor 594 (Thermo Scientific). The nuclear counterstain was DAPI (Thermo Scientific), and samples were mounted with aqueous mounting medium (Sigma-Aldrich). Primary antibody, secondary antibody, and autofluorescent controls were run simultaneously with each tested sample.

Microscopy

All tissue sections and fluorescently stained samples were imaged using an Olympus BX-51 upright light microscope equipped with standard DAPI, 488, and tetramethylrhodamine (TRITC) reflector cubes attached to an Olympus DP-70 high-resolution digital camera. Pentachrome-stained sections were imaged at 10× magnification. Immunofluorescent images were taken at either 20× or 40× magnification. Scale bars were placed for all images with a standard 200-μm diameter. Each site was imaged in all channels and overlaid in DPViewer before examination in Adobe Photoshop.

Photography

Images of mutant mice and littermate controls were acquired via gross photography using a Nikon digital camera at a standardized distance against a white background. Images were digitally edited using Adobe Photoshop to remove background and shadow artifacts around each mouse.

Statistical Analysis

A power analysis was performed to determine the number of mice required for tHO studies. For the power analysis, the primary outcome of interest is differences in HO volume with treatment. To confirm a 50% decrease in HO volume with a power of 0.8, assuming an SD of 1.5 mm³ and mean HO volume of 7.5 mm³ in untreated mice, we required 3 mice/group. All in vitro experiments were performed in technical triplicates. For all microCT analyses of treatment and vehicle groups, values were normalized to the untreated population. For all microCT analyses of mutant and littermate groups, values were normalized to the littermate population. Means and SD were calculated from numerical data and statistical analysis as presented in the text, figures, and figure legends. Equivariance was assessed with Levene’s test, and a Welch correction was applied when indicated. Outlier analysis was performed using Grubbs’ test for outliers with an alpha of 0.05. In figures, bar graphs represent means, whereas error bars represent 1 SD. For all assays, significance was defined as p < 0.05. Asterisks are representative of statistical significance.

Author Contributions

S.A., S.J.L., and B.L. designed the experiments. S.A., S.J.L., D.C., and J.P. performed all murine traumas. S.A., S.J.L., C.B., J.L., D.C., C.B., J.P., H.H.S., J.D., K.R., Y.S.N., W.X., and S.L. collected data. Y.M. provided the transgenic mice. R.K. provided the T1-BMPR inhibitors. S.A., S.J.L., J.P., S.L., R.K., R.T., P.Y., Y.M., and B.L. helped with study design and analysis. S.A., S.J.L., C.B., D.C., C.B., J.P., H.H.S., K.R., Y.S.N., S.L., R.K., R.T., P.Y., Y.M., and B.L. reviewed the data. S.A., S.J.L., and D.C. designed the figures. S.A., S.J.L., and B.L. drafted the manuscript. S.A., S.J.L., R.K., R.T., T.D., P.Y., Y.M., and B.L. provided critical reviews and edited the manuscript.

Conflicts of Interest

B.L. began a collaboration with Boehringer Ingleheim after data collection on this manuscript was complete.