Introduction

Heterotopic ossification (HO) is a disorder characterized by ectopic bone formation at extraskeletal sites, including skeletal muscle and connective tissues. Trauma-induced HO develops as a common post-operative complication after orthopedic surgeries (e.g., hip arthroplasty), blast injuries, deep burns and nervous system injuries (Hwang et al., 2022). In addition to trauma- induced HO, fibrodysplasia ossificans progressiva (FOP) is a rare congenital autosomal dominant disorder also involving HO (Bravenboer et al., 2015). Ectopic bones form progressively through endochondral ossification, mostly in episodic flare-ups associated with inflammation (Towler and Shore, 2022). These cumulative osteogenic events lead to reduced mobility resulting from ankylosing joints. Affected individuals have a shorter life-span, most commonly due to thoracic insufficiency syndrome (Kaplan et al., 2010; Pignolo et al., 2016).

HO reflects a shift from a normal tissue repair process to an aberrant reactivation of bone-forming developmental programs that require acute inflammation and the excessive expansion of local progenitors, followed by inappropriate differentiation of these progenitors into chondroblasts and osteoblasts which finally results in bone formation (Hwang et al., 2022). In both trauma- induced HO and FOP, pathology appears following the excessive activation of receptors sensitive to Transforming growth factor-β (TGFβ) superfamily family members (Lees-Shepard et al., 2018; Wang et al., 2018). FOP arises from gain-of-function mutations in the bone morphogenetic protein (BMP) type I receptor ACVR1, with the most common mutation being c.617G>A, R206H (Shore et al., 2006). Higher SMAD 1/5 mediated signaling of mutated ACVR1/ALK2 has been partially attributed to a loss of auto-inhibition of the receptor and mild hypersensitivity to BMP ligands (Chaikuad et al., 2012; Groppe et al., 2011; Van Dinther et al., 2010). Signaling by mutated ACVR1 still relies on ligand-induced heterotetrameric clustering but, unlike wild type ACVR1, does not require ACVR2A/B kinase activity (Agnew et al., 2021; Ramachandran et al., 2021). More importantly, mutations alter their signaling specificity, abnormally transducing SMAD 1/5 and altering the non-canonical BMP signals (e.g., p38 or PI3K) in response to activin A (Hatsell et al., 2015; Hino et al., 2015; Valer et al., 2019a). Accordingly, evidence supports that activin A is necessary and sufficient for HO in FOP (Hatsell et al., 2015; Lees-Shepard et al., 2018; Upadhyay et al., 2017). In contrast, activin A does not drive post-traumatic HO (Hwang et al., 2020), and non-genetically-driven HO mostly arises from excessive BMP and TGFβ signaling, with functional redundancy between different type I receptors of the TGFβ-superfamily (Agarwal et al., 2017; Patel et al., 2022; Sorkin et al., 2020; Wang et al., 2018).

Evidence points to mesenchymal fibroadipogenic precursors (FAPs), which are widely distributed in muscle and other connective tissues, as the key cell-of-origin that aberrantly undergo chondrogenesis and further ectopic bone formation (Dey et al., 2016; Eisner et al., 2020; Lees- Shepard et al., 2018). However, either FOP or non-genetic HO also require local tissue destruction and inflammation, which indicate that FAPs should be invariably primed for ectopic bone formation by this inflammatory microenvironment (Barruet et al., 2018; Convente et al., 2018; Hwang et al., 2022; Matsuo et al., 2019; Pignolo et al., 2013; Sorkin et al., 2020).

Inflammation in HO is characterized by an initial acute response following injury with the activation of innate immunity and the influx of neutrophils and monocytes (Convente et al., 2018; Hwang et al., 2022). Additionally, B and T cells of the adaptive system are also recruited (Chakkalakal et al., 2012; Convente et al., 2018). However, mice lacking B or T lymphocytes exhibited no delay in the development of heterotopic ossification (HO) after injury, indicating that these cells may play a subsequent role in the dissemination of bone lesions (Kan et al., 2009). During intermediate and late inflammatory stages, the recruitment of monocytes, macrophages and mast cells occurs, which in turn exerts autocrine and paracrine effects on nearby FAPs (Chakkalakal et al., 2012; Convente et al., 2018; Sorkin et al., 2020; Tu et al., 2022). Highlighting its relevance in HO, the depletion of mast cells and macrophages profoundly impairs genetic and non-genetic HO (Convente et al., 2018; Torossian et al., 2017; Tu et al., 2022). BMP receptors are robustly expressed in monocytes and macrophages and the expression of ACVR1^R206H has been shown to extend inflammatory responses in patient-derived macrophages (Barruet et al., 2018; Matsuo et al., 2021, 2019). From the plethora of cytokines secreted by monocytes, macrophages and mast cells, both activin A and TGF-β stand out as extremely relevant for HO (Alessi Wolken et al., 2017; Hatsell et al., 2015; Lees-Shepard et al., 2018; Patel et al., 2022; Sorkin et al., 2020; Upadhyay et al., 2017).

Genetic and pharmacological studies have indicated that osteochondroprogenitor specification and maturation depend on phosphatidylinositol 3-kinase-α (PI3Kα) (Ford-Hutchinson et al., 2007; Fujita et al., 2004; Gámez et al., 2016; Ikegami et al., 2011). We found that PI3K signaling was also linked to HO, since inhibitors of PI3Kα (BYL719/Alpelisib/Piqray) prevented HO in mouse models without major side-effects (Valer et al., 2019a). Mechanistically, PI3Kα inhibitors hamper canonical and non-canonical BMP signaling, decreasing total and phosphorylated SMAD1/5 levels and reducing transcriptional responsiveness to BMPs/Activin A in mesenchymal progenitors (Gámez et al., 2016; Valer et al., 2019b). In the current study, we demonstrate that the pharmacological and genetic inhibition of PI3Kα in HO progenitors at injury sites reduces HO in vivo. Moreover, envisioning a future translation into the clinic, we have optimized the administration of BYL719 and found that the delayed administration of BYL719, up to seven days after inflammatory injury, still prevents HO in mice. In addition, we found that BYL719 blocks osteochondroprogenitor specification and effectively reduces the essential inflammatory response. Altogether, the data presented here show the potent therapeutic effect of PI3Kα inhibition on HO in mouse models.

Results

Delayed initiation of treatment with PI3Kα inhibitor effectively prevents heterotopic ossification

Heterotopic ossification takes place following a precise temporal pattern of progenitor activation and the recruitment of distinct cell types at the ossification centers which causes early divergence from normal skeletal muscle repair program. We previously found that the pharmacological administration of BYL719 prevents HO in a mouse model of HO (Valer et al., 2019a). Given that BYL719 is marketed for the treatment of cancer and overgrowth syndrome, it is a promising therapeutic molecule in pathological HO. To further determine a potential therapeutic window for BYL719 and better understand the cellular targets of BYL719, we evaluated the effect of BYL719 when administered intermittently and several days after HO induction. For this, we used a conditional mouse model (ACVR1^Q207Dfl/fl) in which we combine the overexpression of Cre recombinase through adenoviral particles with the injection of cardiotoxin intramuscularly in the hindlimb (Fukuda et al., 2006). In this model, we administered BYL719 (i.p. 25 mg/kg) or vehicle control intermittently starting one, three or seven days after HO induction. We also implemented a different regimen involving the administration of BYL719 only for the initial three days following the injury (Figure 1A). None of the treatments led to significant changes in the weight of the mice (Figure 1-Figure Supplement 1). HO volume was analyzed by micro computed tomography (µCT) twenty-three days post-injury. The early administration of the treatment with BYL719 at day one or three after the onset of HO resulted in significantly lower HO; late administration seven days after the induction of HO was also partially effective (Figure 1B and C). Discontinuation of the treatment after three days did not prevent HO at day 23. These results indicate that the inhibition of PI3Kα by BYL719 after injury prevents HO and this protection is still effective if the treatment is started several days after injury, expanding its therapeutic window.

Figure 1.

Deficiency of PI3Kα at injury sites is sufficient to partially prevent heterotopic ossification

Since other small molecules designed to target PI3Kα have reported partial off-target effects on kinases other than PI3Kα (Furet et al., 2013; Jamieson et al., 2011), we aimed to confirm that the results obtained with BYL719 were due to the specific inhibition of PI3Kα. Therefore, we developed a conditional mouse model (ACVR1^Q207Dfl/fl:p110α^fl/fl) in which the Cre recombinase drives the expression of ACVR1^Q207D and deletion of the catalytic subunit of PI3Kα (p110α) after intramuscular hindlimb combined injection of adenovirus-Cre and cardiotoxin to trigger HO. With this approach, Cre mediates the expression of ACVR1 ^Q207D and the deletion of PI3Kα in the same cells, whereas cells recruited afterwards would be wild type for both genes. As expected, µCT analysis performed 23 days after injury showed extensive HO in mice expressing ACVR1^Q207D, which was prevented by intermittent treatment with BYL719 initiated one day after injury (Figure 2A, 2B and 2D). Genetic deletion of p110α in mice expressing ACVR1^Q207D led to a reduction in HO, which was further enhanced by BYL719 (Figure 2B). Importantly, no significant changes in the weight of the mice were observed, confirming the absence of severe toxicity (Figure 2-Figure Supplement 1A). Histomorphometrically, ACVR1^Q207D mice wild type for PI3Kα developed islands and/or bone spurs with a well-organized bone structure. However, in PI3Kα deficient mice ACVR1^Q207D expression only led to minor ectopic calcifications that were already surrounded by fully regenerated muscle tissue on the 23^rd day after injury (Figure 2D, Figure 2-Figure Supplement 1B). Accordingly, analysis of the ratio bone volume/tissue volume (BV/TV) and the correlation between bone volume and BV/TV of individual mice was clearly different between p110α genotypes, irrespective of BYL719 treatment (Figure 2C, Figure 2-Figure Supplement 1C). These results demonstrate that the abrogation of PI3Kα activity (either genetic of pharmacologically) in cells at injury sites is relevant for their inhibitory effects in HO. This might be attributed to the direct effects of BYL719 on progenitor cell expansion and chondroblast specification, as well as its possible ability to reduce the inflammatory response required for HO.

Figure 2.

Genetic or pharmacological inhibition of PI3Kα prevents osteochondroprogenitor specification

Next, we aimed to assess the effects of PI3Kα inhibition in osteochondroprogenitor cells. For this, we isolated bone marrow-derived mesenchymal stem cells (BM-MSCs) from p110α^fl/fl mice and transduced them ex-vivo with retrovirus expressing Acvr1 (wild type and R206H) with or without Cre recombinase. The exogenous mRNA gene expression of both Acvr1 and Acvr1^R206H was similar (Figure 3A). As anticipated, transduction with Cre viruses significantly reduced the expression levels of p110α to approximately 40%. This is consistent with an estimated transduction efficiency of nearly 60% (Figure 3B). The overexpression of Acvr1^R206Hincreased basal and activin dependent expression of canonical (Id1 and Sp7) and non-canonical (Ptgs2) BMP target genes (Figure 3C), known markers and drivers of osteochondrogenic differentiation (Nakashima et al., 2002; Wang et al., 2013). Both pharmacological and genetic inhibition of PI3Kα reduced canonical and non-canonical ACVR1^R206H transcriptional responses induced by recombinant activin A. Of note, stimulation with activin A increased its own transcription (Inhba) in Acvr1^R206H- transduced BM-MSCs, which was also inhibited by BYL719 (Figure 3C). Therefore, BYL719 may also prevent osteochondroprogenitor differentiation indirectly, via transcriptional inhibition of INHBA/Activin A expression. These results demonstrate a direct effect of BYL719 on osteochondroprogenitor cells and suggest that the effects of BYL719 on ACVR1-downstream signaling are mainly on target, due to PI3Kα kinase inhibition.

Figure 3.

It is known that 2-aminothiazole-derived PI3Kα inhibitors could also inhibit ACVR1 kinase activity (Jamieson et al., 2011). Therefore, we investigated whether BYL719 effects could be explained by the reduction in ACVR1 kinase activity. For this, we made use of a target kinase engagement approach. Upon transient transfection of Nanoluciferase-tagged TGFβRs, we incubated the cells with BYL719 and/or an ATP-like tracer analogue. In the absence of an ATP competitor molecule, the tracer analogue and Nanoluciferase enzyme stay in close proximity, thereby allowing for bioluminescence resonance energy transfer (BRET) between the Nanoluciferase (energy donor, emitting at 460 nm) and the tracer molecule (energy acceptor) Once excited, the acceptor molecule emits fluorescence (610 nm) (Figure 4A and B). BYL719 incubated at 1 and 10 µM failed to significantly inhibit the emission of the BMP type I receptors ALK1, ALK2 (ACVR1), ALK3, ALK4, ALK5 and the mutant receptor ACVR1/ALK2^R206H (Figure 4C). In addition, we demonstrated that BYL719 does not target the activity of the type II kinase receptors ACVR2A, ACVR2B, BMPR2 and TGFBR2. We confirmed these results using in vitro kinase activity assays with recombinant ACVR1^R206H. BYL719 exhibited no major effects on ACVR1^R206H activity up to 10µM (Figure 4D).

Figure 4.

Given that BYL719 does not directly target the activity of ACVR1 or related receptors, we established an in vitro model using human MSCs (hMSCs) stably overexpressing ACVR1 or ACVR1^R206H to profile further the main targeted pathways. In this model, recombinant Activin A was able to induce SMAD1/5 activation and the expression of ACVR1 downstream targets ID1 and ID3 in cells over-expressing ACVR1^R206H (Figure 5-Figure Supplement 1A and B). In these cells, coincubation with BYL719 (10 μM) was able to repress Activin A-induced Alkaline phosphatase expression and activity upon 7 and 11 days of culture, respectively (Figure 5A, B and Figure 5-Figure Supplement 1C). Furthermore, we performed 3D chondrogenic differentiation assays. Coincubation with BYL719 (10 μM) almost completely inhibited the formation of a chondrogenic glycosaminoglycans-rich matrix in response to Activin A after 3 weeks of culture (Figure 5C and D) and partially abrogated the expression of the osteochondrogenic genes COL1A1 and ALPL in early chondrogenic differentiation (Figure 5E). These results confirmed that BYL719 administration blocks osteochondrogenic specification irrespective of the stage of differentiation.

Figure 5.

To identify the immediate targets of BYL719, we performed bulk RNA sequencing (GSE237512) where we preincubated hMSCs-ACVR1^R206H with BYL719 for 30 min, which was followed by stimulation with recombinant Activin A (50ng/ml) for 1 hour (Figure 6A). Consistent with our results on osteogenic progenitor cells, preincubation with BYL719 (1 μM) resulted in the robust regulation of genes involved in ossification and osteoblast differentiation (including PTGS2, MSX2, SOX9, and BMP2), as identified through enrichment analysis of the GO terms GO:0001503 (padj=3.83e^-5) and GO:0001649 (padj=3.10e^-5), respectively (Figure 6B and C), indicating that BYL719 mainly targets osteochondral specification processes in these cells. Moreover, a number of inflammatory cytokine genes (e.g. IL6 and IL15) and inflammatory signaling pathways (TNF and NFκB), were downregulated by BYL719 (Figure 6-Figure Supplement 1A and B). Therefore, we hypothesized that, apart from preventing the activation of osteogenic and chondrogenic differentiation pathways in progenitor cells, part of the mechanism by which BYL719 prevents HO in vivo may be due to the modulation of inflammatory responses.

Figure 6.

PI3Kα inhibition reduces proliferation, migration and inflammatory cytokine expression in monocytes, macrophages and mast cells

Next, we used several models of disease-relevant immune cells to study the effect of pharmacological PI3Kα inhibition. Activation of ACVR1 signaling with BMP6 did not significantly modify the proliferation rate of any of the cell lines tested, that is, monocytes (THP1), macrophages (RAW264.7) and mast cells (HMC1) (Figure 7A, B and C). However, incubation with BYL719 in vitro strongly reduced their growth rate at 2µM and almost completely arrested their proliferation at 10µM. It is well known that monocytes are actively recruited from circulation to the ossification sites early after injury (Sorkin et al., 2020). Therefore, we also analyzed chemotactic migration of monocytes in transwell assays. Whereas monocytes actively migrated towards FBS as a chemotactic agent, the incubation of monocytes with BYL719 deeply diminished their migratory ability (Figure 7D). These results suggest that BYL719 is able to diminish monocyte recruitment, while reducing also the expansion of inflammatory cells at injury sites.

Figure 7.

Upon tissue injury, a multitude of pro-inflammatory and anti-inflammatory cytokines are released at the site of injury. This triggers the recruitment and infiltration of various cell types, including immune cells and osteochondrogenic progenitors. We examined the expression of distinct cytokines upon ACVR1 stimulation with BMP6 in the presence of BYL719 in monocytic, macrophagic and mast cell lines. In monocytes, BYL719 inhibited the expression of the pro-inflammatory cytokine TNF-α and the expression of activin A (INHBA), whereas the expression of transforming growth factor TGF-β1 was increased by BMP6 and not modified by the addition BYL719 (Figure 7E). CCL2 (MCP-1), a chemokine inducing the recruitment of monocytes and macrophages into inflamed tissues as well as the M2 polarization of macrophages (Sierra-Filardi et al., 2014) was significantly induced by BMP6 and partially reduced by BYL719 (Figure 7E). In macrophages, BYL719 reduced the expression of pan-macrophage markers Adgre1 (F4/80), Cd68 and the M1 macrophage marker Cd80 (Figure 7F). Conversely, BYL719 increased the expression of the M2 macrophage marker Il10, regardless of the presence or absence of BMP6. In the human mast cell line HMC1, BYL719 inhibited the expression of the pro-inflammatory cytokine IL6, while the expression of TGFβ1, TNFα, and activin A (INHBA) remained unchanged upon the addition of either BMP6 or BYL719 (Figure 7G). Altogether, these data suggest that PI3Kα inhibition reduces the expression of certain pro-inflammatory cytokines and might skew macrophages towards M2. Moreover, the reduction on activin A expression by BYL719 in inflammatory cells and mesenchymal progenitors could be relevant for HO in FOP.

PI3Kα inhibition reduces the hyperinflammatory response in FOP

To confirm whether BYL719 can modulate excessive inflammation at the injured sites, we examined its impact on monocytes, macrophages and mast cells in vivo. These cell populations are known to increase in number two to four days after injury (Convente et al., 2018; Sorkin et al., 2020), which coincides with the effects observed following early administration of BYL719 (Convente et al., 2018; Sorkin et al., 2020). Hindlimb muscle samples were collected at two, four, nine, sixteen and twenty-three days post-induction of HO with Cre viruses and cardiotoxin in ACVR1^Q207Dfl/flmice (Figure 8A). The group of treated animals received BYL719 intermittent treatment starting one day after injury. After sixteen days, structures of mineralized bone became detectable in vehicle-treated mice (Figure 8-Figure Supplement 1A and B). The number of F4/80- positive monocytes/macrophages was high two days after injury and remained elevated in the HO lesions throughout the ossification process (Figure 8B and C). BYL719-treated animals showed an almost 50% reduction in the number of F4/80 positive cells at day two post-injury and maintained the relative difference with vehicle-treated mice afterwards. Consistent with previous studies, we observed an increased number of mast cells at injury sites. Their numbers peaked at four days and only dropped twenty-three days after injury in vehicle-treated mice (Figure 8D and E). In BYL719-treated mice, mast cells were increased up to day 9, but their number was completely normalized to preinjury levels already at day 16 (Figure 8D and E). This observation correlates with the complete muscle regeneration observed in BYL719-treated mice compared to untreated mice, which are actively developing ossifications at that time (Figure 8-Figure Supplement 2). These results confirm a functionally relevant dual role of PI3Kα inhibition, by preventing the differentiation of osteochondrogenic progenitor cells and also by reducing the local inflammatory response.

Figure 8.

Discussion

We previously showed that pharmacological PI3Kα inhibition using BYL719 has the potential to suppress HO by increasing SMAD1/5 degradation, reducing transcriptional responsiveness to BMPs, and blocking non-canonical responses such as the activation of AKT/mTOR (Valer et al., 2019a). In this manuscript, we confirmed these results and further expanded the insights into the molecular and cellular mechanisms responsible for the therapeutic effect of PI3K inhibition. Here, we optimized the therapeutic window for BYL719 administration and found that the delayed administration of BYL719 effectively prevents HO. Moreover, in vivo simultaneous genetic inhibition of p110α and the activation of mutated ACVR1 at injury sites partially prevents HO, and we demonstrate that genetic or pharmacological inhibition of PI3Kα inhibits osteochondroprogenitor specification. We determined that BYL719 does not inhibit recombinant ACVR1^R206H kinase activity in vitro and does not behave as an ATP competitor inhibitor for none of the TGF-β kinase receptors. Finally, the administration of BYL719 prevented an exacerbated inflammatory response in vivo, possibly due to specific effects observed on immune cell populations.

While FAPs play a key role in bone formation in different types of HO, inflammation is also closely associated with all types of HO (Barruet et al., 2018; de Ruiter et al., 2021; Matsuo et al., 2019). For instance, the depletion of mast cells and macrophages has been shown to reduce bone volume in FOP mouse models (Convente et al., 2018). Moreover, Activin A, TGF-β and other cytokines secreted by FAPs, monocytes, macrophages and mast cells are essential for FOP and non-genetic HO (Alessi Wolken et al., 2017; Hatsell et al., 2015; Lees-Shepard et al., 2018; Patel et al., 2022; Sorkin et al., 2020; Upadhyay et al., 2017). In addition, ACVR1 antibodies can activate ACVR1^R206H even in the absence of Activin A, but muscular trauma is still required to induce HO (Aykul et al., 2022; Lees-Shepard et al., 2022), indicating the involvement of additional immune factors. Our study finds that BYL719, a PI3Kα inhibitor, reduces the expression of certain pro-inflammatory cytokines, effectively blocks the proliferation of monocytes, macrophages and mast cells and reduces the migratory potential of monocytes. This likely contributes to a decreased number of monocytes and macrophages at injury sites and throughout the in vivo ossification process. It is known that excessive pro-inflammatory cytokine expression, including activin A, by monocytes and macrophages is observed in all types of HO and is induced by ACVR1^R206H (Matsuo et al., 2021). Non-canonical signaling pathways (mTOR, p38, TAK1, NF-κB) are also affected by ACVR1^R206H in immune cell types (Barruet et al., 2018; Hwang et al., 2022). M1-type macrophages initiate acute inflammatory responses but later transition to M2-type macrophages associated with anti-inflammatory and reparative functions. Therefore, modulating macrophage phenotype towards regeneration may dampen HO (Sorkin et al., 2020). BYL719 reduces monocyte/macrophage numbers, migration, and pro-inflammatory cytokine expression, potentially altering their polarization. It also reduces activin A expression and promotes a shift towards the M2 phenotype. BYL719 can help to reduce the hyper-inflammatory state in HO, inhibit FAP expansion, and favor myogenic regeneration of muscle tissue (Stanley et al., 2022)..

Our observation that the delayed administration of BYL719 is still effective has pathophysiological and therapeutic derivatives. Muscle injury induces sequential changes in HO progression: early immune cell infiltration (days 1-3), a fibroproliferative and late inflammatory phase, with an expansion of FAPs and a spectrum of monocytes, macrophages and mast cells as highly secretory cells (days 3-7), followed by chondrogenesis (day 7-14) and osteogenesis (days 14-23) (Hwang et al., 2022). Starting BYL719 administration 3 days post-injury is still fully therapeutically effective and even partially at 7 days post-injury. In addition, whereas untreated mice developed cartilage and bony lesions by days 9 and 16 post-injury, respectively, BYL719- treated animals fully regenerated muscle by day 16 post-injury. Altogether, this evidence suggests that the major effects of BYL719 occur between 3 and 16 days post-injury. We found that BYL719 completely blocks chondrogenesis in cultured hMSCs irrespective of the differentiation duration. In addition, in the histological analyses of these BYL719-treated mice that do not develop HO, we cannot find any sign of cartilage formation on either the 9^th, 16^th or 23^rd days post-injury, which confirms that BYL719 is also able to block chondrogenesis in mice in vivo. This suggests that intervention with BYL719 over a temporal window after a HO flare might be sufficient to inhibit endochondral ossification. In addition, although surgical resection is not recommended in FOP patients and has a risk of recurrence in resected non-genetic HO, it could be envisaged using as preventive therapy in subjects undergoing surgeries to remove ectopic bone.

Treatment options for HO and FOP are currently limited, primarily consisting of anti- inflammatories such as corticosteroids and NSAIDs. Promising pharmacological treatments have progressed to II/III clinical trials. Palovarotene, a retinoic acid receptor-γ agonist, has been approved in Canada for patients over 10 years old (NCT03312634). However, a previous clinical trial with palovarotene was paused for children due to concerns about premature growth plate closure, which is debated in FOP mice (Chakkalakal et al., 2016; Rosen et al., 2018). Antibodies targeting activin A (Garetosmab) showed effectiveness in a phase II trial, but serious adverse effects led to a temporary hold on the trial (NCT03188666). The outcome of a clinical trial with Rapamycin has not been publicly disclosed and a case report indicated limited benefits in classical FOP patients (UMIN000028429) (Kaplan et al., 2018). BYL719 (Alpelisib/Piqray®) was recently licensed in combination with hormone therapy for a PIK3CA-mutated breast cancer (Cardoso et al., 2020). BYL719 has been shown to be also clinically effective in adult patients with PIK3CA- related overgrowth syndrome (PROS) and children under 1 year of age (Madsen and Semple, 2022; Morin et al., 2022). In both scenarios, BYL719 had no major adverse effect when patients were treated for more than one year (in adults at a dose of 300mg daily for oncological or PROS purposes and at a dose of 25 mg daily in infants with PROS) (Morin et al., 2022). Thus, implementing BYL719 for treatment of HO, at least during a restricted temporal window (i.e. surgery to remove ectopic bone in HO, or during flare-ups in FOP individuals), may be a valid therapeutic option for FOP patients.

Materials and Methods

Murine bone marrow mesenchymal stem cells isolation

Murine bone marrow-derived mesenchymal stem cells (BM-MSCs) were isolated from 6-8 weeks-old C57BL6/J mice for the floxed p110α allele (Beatriz Gámez et al., 2016)as previously described (Soleimani and Nadri, 2009). Mice were euthanized and femurs were dissected and stored in DMEM with 100 U/ml penicillin/streptomycin (P/S). Soft tissues were cleaned and femur ends were cut under sterile conditions. Bone marrow was flushed with media using a 27-gauge needle. The resulting cell suspension was filtered through a 70-μm cell strainer and seeded in a 100-mm cell culture dish. Non-adherent cells were discarded after 3 hours. Media was slowly replaced every 12 hours for up to 72 hours. Then, media was replaced every 2 days until the culture reached 70% confluence. Then, cells were lifted by incubation with 0.25% trypsin/0.02% EDTA for 5 minutes at room temperature. Lifted cells were cultured and expanded.

Production of human fetal mesenchymal stem cells (hMSCs) expressing ACVR1

Bone marrow-derived hMSCs were transduced by lentiviral delivery encoding ACVR1^WT- nanoluc or ACVR1^R206H-nanoluc. The nanoluc fusion inserts (Promega, NV2341/NV2381) were subcloned into a PLV-CMV-IRES vector. First, an eGFP was amplified with PCR containing PstI, SalI and XbaI restriction sites and subcloned in the PLV using PstI and XbaI. Next, the ACVR1- nanoluc inserts were subcloned in the PLV by restriction of PstI and SalI, exchanging the eGFP. Subsequently, lentiviral particles were produced using HEK293t cells and cells were transduced.

Cell culture

Murine BM-MSCs were cultured in DMEM supplemented with 10% FBS, 2 mM glutamine, 1 mM sodium pyruvate, and 100 U/ml P/S and incubated at 37°C with 5% CO₂. The human monocyte cell line THP1 was cultured in RPMI-1640 supplemented with 10% FBS, 2 mM glutamine, 1 mM sodium pyruvate, 100 U/ml P/S, 0.1mM NEAA, 10mM HEPES and 50µM 2-mercaptoethanol. Mouse macrophage cell line RAW264.7 was cultured in DMEM supplemented with 10% FBS, 2 mM glutamine, 1 mM sodium pyruvate and 100 U/ml P/S. The human mast cell line HMC1 was cultured in IMDM supplemented with 10% FBS, 2 mM glutamine, 1 mM sodium pyruvate and 100 U/ml P/S. The hMSCs were cultured in αMEM supplemented with 10% FBS, 1 ng/ml human bFGF and 100 U/ml P/S (Growth Medium; GM).

Proliferation and chemotactic migration assays

For proliferation assays, 3000 cells/well were seeded in 12-well culture plates in correspondent culture medium with 2 nM BMP6 and/or 2 μM or 10 μM BYL719 (Chemietek). Photos of several fields of each well were taken at 24, 48, 72 and 144 hours using a Leica DM-IRB inverted microscope. Cell counting was performed through ImageJ software.

The chemotactic migration assay of THP1 was performed through the InnoCyte Monocyte Migration Assay (Merck Millipore). Cells were serum-starved for 3h prior to the experiment and, when needed, treated with 2 μM BYL719 for 30 minutes before the experiment started. Then, 10⁵ cells/100 μL were added to each upper chamber. Lower chambers were charged with 10% FBS- supplemented RPMI-1640 (positive control), FBS-depleted RPMI-1640 (negative control) or 2 nM BMP6 (R&D), respectively. Cells migrated for 2h and were then stained with calcein-AM and pipetted into a 96-well conical bottom plate (Sigma-Aldrich). A plate reader was used to measure florescence from the top of the plate at an excitation wavelength of 485 nm and an emission wavelength of 520 nm. Unless otherwise stated, BYL719 at 2 or 10 μM (Chemietek) and 2 nM BMP6 (R&D) was used for 6 days (proliferation) or 4 hours (migration) in HMC1, THP1 and RAW264.7 cells.

Retroviral transduction and gene expression analysis

BM-MSCs were infected with mock virus (pMSCV-GFP) or with viruses expressing wild type Acvr1 (WT), Acvr1^R206H (RH) and/or pMSCV-Cre. Plasmids with WT and RH Acvr1 forms were kindly provided by Dr. Petra Seemann and were subcloned into pMSCV vector. Acvr1 and p110α expression levels were analyzed by qRT-PCR. Cells were treated with BYL719 (2µM) and/or activin A (2nM) for 48 hours in complete media without FBS. THP1, RAW264.7, and HMC-1 cells lines, cells were treated in the media indicated for each cell line with BYL719 (2µM) and/or BMP6 (2nM) for 48 hours.

Total RNA for RT-PCR was extracted from all cellular models using TRIsure reagent. At least 2 μg of purified RNA was reverse-transcribed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative PCRs were carried out on ABI Prism 7900 HT Fast Real-Time PCR System with Taqman 5′-nuclease probe method and SensiFAST Probe Hi-ROX Mix. All transcripts were normalized using Tbp as an endogenous control.

NanoBRET target engagement assays

COS-1 cells were cultured in DMEM (Gibco, 11965092) supplemented with 10% FBS and 100 U/ml P/S. These cells were transfected in 70% confluent 6-wells TC-treated plates with 2 µg TGFβ receptor-Nanoluciferase constructs (constructs were provided by Promega and Promega R&D) using a 1 µg DNA: 2 µl PEI (1 mg/ml) ratio. The transfected cells were reseeded in white 384-wells TC-treated assay plates one day before the nanoBRET readout in DMEM supplemented with 1% FBS in a quantity of 15*10^3 cells in 40 µl per well. Two hours before measurement, the dedicated tracer (Promega) and the test compounds were incubated to the cells at 37°C. Immediately after intracellular TE Nano-Glo Substrate/Inhibitor (Promega, N2160) addition, the wells were measured by 450-80BP and 620-10BP filters using the ClarioSTAR (BMG Labtech). Subsequent nanoBRET ratios were measured by the formula: acceptor emission (620-10 nm)/donor emission (450-80 nm) * 1000 (milliBRET units, mBU).

Chondrogenic progenitor differentiation

The hMSCs were cultured in αMEM supplemented with 10% FBS, 1 ng/ml human bFGF and 100 U/ml P/S (Growth Medium; GM). Chondrogenic differentiation medium (DM) consisted of DMEM supplemented with 10 µl/ml Insulin-Transferrin-Selenium 200 µM Ascorbic Acid, 0.1 µM Dexamethasone, 350 µM L-proline, 1 mM Sodium Pyruvate, 1.25 µg/ml BSA and 100 U/ml P/S. During differentiation, the medium was refreshed twice a week for the duration of the experiment. For the ALP staining and ALP activity experiments, the hMSC-ACVR1 lines were seeded at 2×10⁴ cells per well in 48-wells plates and differentiated for 7 and 11 days, respectively. For the Alcian Blue experiments, 3D micromass cultures were made by carefully seeding 3×10⁵ cells per 10 µl droplets in a 24-wells plate and incubated for 2 hours at 37°C prior to GM addition. Chondrogenic progenitor differentiation was started 1 day after micromass seeding and continued for 3 weeks. For the analysis of early chondroprogenitor specification, the hMSC-ACVR1 lines were differentiated for only 3 days where after RT-qPCR analysis was performed.

Alkaline phosphatase (ALP) staining, ALP activity assay and Alcian Blue staining

For the ALP staining, the progenitors were washed with PBS, fixated using 3.7% formalin for 5 minutes at RT, washed twice with PBS and stained in ALP solution containing 2 mg Naphthol AS-MX, 6 mg Fast Blue, 5 ml 0.2M Tris (pH 8.9), 100 µl MgSO₄ and dH₂O up to 10 ml. Images were acquired using the Leica DMi8 with a 10x magnification. Densitometry analysis of the whole well (from total plate scan) was performed using ImageJ (v1.53t).

ALP lysates were obtained by washing twice with PBS, freezing at −80°C for 1 hour, and lysing on ice for 1 hour using 100 µl ALP buffer (100 µM MgCl₂, 10 µM ZnCl₂, 10 mM Glycine (pH 10.5)) plus 0.1% Triton X-100 per well. ALP activity was quantified by adding 20 µl lysate and 80 µl 6 mM PNPP in ALP buffer to a clear 96-well plate incubated at RT until the samples turned yellow. Next, the samples were measured at an absorbance of 405 nm.

The micromass pellets were stained with Alcian Blue as described before (Sánchez-Duffhues et al., 2019). Images were acquired using the Leica DMi8 with 5x magnification. To quantify the chondroprogenitor specification, the Alcian Blue was solubilized by 250 µl of 6M guanidine hydrochloride and absorbance was measured at 595 nm.

Bulk RNA sequencing and analyses

The hMSCs-ACVR1^R206H were seeded at 100% confluency and serum starved overnight with DMEM without supplements. Afterwards, the cells were pre-treated with 1 µM of BYL719 in starvation medium for 30 minutes and stimulated with 50 ng/ml Activin A for 1 hour. RNA was isolated using the ReliaPrep RNA Miniprep Systems (Promega) and sequenced using the Illumina NovaSeq 6000 (Illumina) platform. The reads were mapped using the Hisat2 (v2.0.5) package and the total and normalized counts were measured using featureCounts (v1.5.0-p3). Differential expression analysis was performed using the DESeq2 R package (v1.20.0), in which the p-values were adjusted following the Benjamini and Hochberg approach. Genes with an adjusted p-value < 0,05 were considered differentially expressed. Subsequent Gene Ontology analysis was performed using the clusterProfiler R package and the GO terms were considered significant if the p-values < 0,05. Gene Set Enrichment Analysis was performed using the GSEA analysis tool (gsea-msigdb.org). Heatmaps were generated using heatmap2 function within the R gplots package in Galaxy (usegalaxy.eu). The datasets are publicly available (see GSE237512).

Western blot

Cells were lysed using RIPA lysis buffer and quantified using the Pierce BCA Protein assay (Thermo) following standard protocols. The protein samples were loaded using β-ME containing sample buffer and run using SDS-PAGE and transferred to PVDF membranes. Primary antibodies against SMAD1 (Cell Signaling, 6944), p-SMAD1/5 (Cell Signaling, 9516), and vinculin (Sigma, V9131) were used at 1:1000 dilution in 5% BSA in TBST. Binding was detected with HRP-conjugated secondary antibodies and visualized by Brightfield ECL (Thomas Scientific) on the ChemiDoc (BioRad).

In vitro recombinant kinase activity assay

Fifty ng of recombinant ACVR1^R206H (ab167922, Abcam) was diluted in 30 µl of 25mM Tris/HCl (pH7.5), 10mM MgCl₂, 10mM MnCl₂ and 2mM DTT. The reaction was started by the addition of 5µM ATP-Mg, 0.05 µCi/µl ATP [γ-³²P] and 0.5mg/ml of dephosphorylated casein (Sigma) and performed for 30 min at 30°C. The reaction was stopped and samples were processed by SDS-PAGE. After drying the gel, phosphorylation was determined by autoradiography.

Heterotopic ossification mouse model

To study heterotopic ossification in vivo, we used the Cre-inducible constitutively active ACVR1^Q207D (CAG-Z-EGFP-caALK2) mouse model as previously described (Fukuda et al., 2006; Shimono et al., 2011; Yu et al., 2008). Mice ACVR1Q207D ^fl/fl p110α ^fl/fl were obtained by crossing the detailed homozygous ACVR1Q207D ^fl/fl with the p110α ^fl/fl homozygous mutant mice carrying loxP sites flanking exons 18 and 19 of the p110α alleles (Beatriz Gámez et al., 2016). Heterozygous mice were crossed until ACVR1Q207D ^fl/fl p110α ^fl/fl was obtained.

To induce heterotopic ossification in P7 ACVR1^Q207D mice, 1×10⁸ pfu of adenovirus-Cre (Ad-CMV-Cre, Viral Vector Production Unit, UAB) and 0.3 μg of cardiotoxin in 10 μl 0.9% NaCl volume were injected into the left hindlimb. Control groups had the same procedure without Ad-Cre in the injection. On P8, mice started receiving either placebo (intraperitoneal administration (i.p.) of DMSO) or BYL719 (i.p. of 25 mg/kg), both diluted in 0.5% carboxymethylcellulose sodium. Mice were housed under controlled conditions (12-hour light/12-hour dark cycle, 21°C, 55% humidity) and fed ad libitum with water and a 14% protein diet (Teklad2014, Envigo). Mice were regularly weighted over the whole period. Heterotopic ossification induction was performed blinded to mouse and group identity. All procedures were approved by the Ethics Committee for Animal Experimentation of the Generalitat de Catalunya.

Micro-computed tomography analysis

The caudal half of mice was collected and fixed in 4% paraformaldehyde (PFA) for 48 hours at 4°C. Samples were conserved in PBS and high-resolution images were acquired using a computerized microtomography imaging system (Skyscan 1076, Bruker microCT), in accordance with the recommendations of the American Society of Bone and Mineral Research (ASBMR). Samples were scanned in air at 50 kV and 200 µA with an exposure time of 800 ms, using a 1-mm aluminum filter and an isotropic voxel size of 9 µm. Two-dimensional images were acquired every 1° for 180° rotation and subsequently reconstructed, analyzed for bone parameters, and visualized by NRecon v1.6, CT-Analyser v1.13 and CTVox v3.3 programs (Bruker), respectively. For heterotopic ossification, manual VOIs comprising heterotopic ossifications were employed and a binary threshold was established at 25-255.

Histology

Whole legs were fixed in 4% PFA for 48 hours at 4°C, decalcified in 16% EDTA pH 7.4 for 6 weeks, and embedded in paraffin. Five µm sections were cut and stained with Fast Green/Safranin O, or Masson’s Trichrome. Images were obtained with brightfield Eclipse E800 (Nikon). To detect monocyte/macrophage cells, paraffined slides were immunostained with Anti-F4/80 antibody [SP115] (ab111101). Briefly, after deparaffination and rehydration, an antigen retrieval step with citrate buffer at pH6 in a decloaking chamber was applied to the slides. Slides were cooled down and rinsed with 1x TBS and endogenous peroxidases were blocked with 70% methanol, 28% distilled water, and 2% hydrogen peroxide for 5 minutes at room temperature. After a blocking step, F4/80 primary antibody incubation at concentration 1:100 overnight at 4 °C was applied. Envision Dual Link conjugated with HRP was applied for 40 minutes at room temperature, and detection was performed with diaminobenzidine (DAB) for 3 minutes. Nuclear staining with Lilly’s hematoxylin wasapplied to the slides for 30 seconds. Images of stained slides were obtained with brightfield Eclipse E800 (Nikon). Images were randomly acquired with 20x magnification from the region of interest (injected muscle, with or without visible heterotopic ossification). Five images were obtained per mice, from 2 mice per group, for a total of 10 quantified images per group. F4/80 positive cells were detected as cells with dark brown staining. To detect mast cells, paraffined slides were stained with C.E.M. staining using the Eosinophil-Mast cell staining kit (ab150665, Abcam). The recommended staining procedure was applied to the deparaffinized sections. Images of stained slides were obtained with brightfield Eclipse E800 (Nikon). Images were randomly acquired with 20x magnification from the region of interest (injected muscle, with or without visible heterotopic ossification). Three images were obtained per mice, from 4 mice per group, for a total of 12 quantified images per group. Mast cells were detected as cells with bright blue staining.

Statistical analysis

Unless stated otherwise in each figure legend, the results were expressed as mean ± SD. The median was shown in heterotopic ossification datasets where non-parametric tests were applied. Each figure legend has detailed information explaining the statistical test used to compare between groups, the compared groups, the explanation of the symbols used to show significance, usually *P<0.05, **P<0.01, ***P<0.001, and the sample size of the experiment. Heterotopic ossification, evaluation, and quantification was performed blinded to mouse and group identity. Microscopy and histology representative images were selected from the total quantified images. Microscopy and histology quantified fields were randomly acquired from the plate or the region of interest. Statistical tests were performed on GraphPad Prism 9.5.

Acknowledgements

We thank E. Adanero, E. Castaño for technical assistance. We acknowledge the support from Promega R&D. Alexandre Deber is a recipient of a F.P.I. fellowship from the Ministry of Science and Innovation (MCIN/AEI/ 10.13039/501100011033). Carolina Pimenta-Lopes is the recipient of a F.P.U. fellowship from the Spanish Ministry of Education. This research was supported by grants PID2022-141212OA-I00, PDC2021-121776-I00 and PID2020-117278GB-I00 from MCIN/AEI/10.13039/501100011033, co-funded by FEDER ‘Una manera de hacer Europa’ or ‘NextGenerationEU’/PRTR; a grant 202038-30 from La Marató de TV3 and grants from IFOPA (ACT for FOP) and FOPItalia. MW and MJG are sponsored by the Netherlands Cardiovascular Research Initiative (the Dutch Heart Foundation, Dutch Federation of University Medical Centers, the Netherlands Organization for Health Research and Development, and the Royal Netherlands Academy of Sciences), PHAEDRA-IMPACT (DCVA) and DOLPHIN-GENESIS (CVON). GSD is also sponsored by the Spanish Ministry of Science through the Ramón y Cajal grant RYC2021-030866-I and the BHF-DZHK-DHF, 2022/23 award PROMETHEUS.

Author contributions

J.A.V., A.D., M.W., M.J.G., J.L.R., G.S-D. and F.V. designed the study, and analyzed and interpreted the results. J.A.V, A.D., M.W. and C.P-L. performed the experiments. J.A.V., A.D., G.S-D. and F.V. wrote the manuscript. All authors contributed to the manuscript. G.S-D. and F.V. supervised the project.

Competing interests

The authors declare no conflicts of interest.