Bone morphogenetic protein signalling in pulmonary arterial hypertension: revisiting the BMPRII connection Open Access

Pulmonary arterial hypertension (PAH) is a rare but debilitating condition with a high mortality rate, affecting 15–26 people per million of the population in western countries [1,2]. The pathology is characterised by the abnormal muscularisation of pre-capillary pulmonary arteries, formation of concentric and plexiform lesions and narrowing of the pulmonary vascular lumen, resulting in an increase in pulmonary vascular resistance, elevated pulmonary artery pressure, right ventricle hypertrophy, and progressive right heart failure [3]. The FDA approved PAH therapies prior to 2024 target three pathways that predominantly affect vascular tone: endothelin 1, nitric oxide and prostacyclin. Although these therapies have improved exercise capacity and delayed clinical worsening time, they do not provide a cure for most patients and survival at three years post-diagnosis remains unacceptably low [4]. Therapies directly targeting the underlying disease pathophysiology are urgently needed.

Genetic studies suggest that PAH can be caused by pathogenic germline mutations. The most prevalent disease gene is BMPR2 (bone morphogenetic protein (BMP) receptor 2) [5,6], encoding the type II receptor for the large family of BMP ligands. BMPR2 mutations are found in over 80% of familial cases and ∼17% of idiopathic PAH (IPAH) patients [7–9]. Among the 12 validated PAH genes that have been recognised by the International Consortium for Genetic Studies in PAH [10], many are related to BMP signalling. Apart from BMPR2, ACVRL1 encodes BMP type I receptor activin receptor-like kinase 1 (ALK1), ENG encodes co-receptor endoglin, GDF2 encodes ligand BMP9, and SMAD9 is a component of BMP signalling machinery [6,11]. Of note, mutations in PAH genes are predisposing factors with incomplete penetrance. BMPR2 variants penetrance is estimated to be 42% for heterozygous women and 14% for heterozygous men [6,12].

BMPs are members of the transforming growth factor-β (TGF-β) superfamily. These ligands are mostly homodimers, initiating cellular signalling by forming a complex with cell surface receptors comprising two copies of a type I receptor and two copies of a type II receptor. Both types of receptors are serine/threonine kinases. After formation of the signalling complex, the constitutively active type II receptor will phosphorylate and activate the type I receptor, which will in turn phosphorylate the receptor-regulated SMADs (R-SMADs, including SMAD1, SMAD5, SMAD9, SMAD2 and SMAD3). The phosphorylated R-SMADs then form a complex with the common mediator SMAD, SMAD4, and translocate to the nucleus to regulate gene expression. Typically, BMP signals are mediated by SMAD1, SMAD5 and SMAD9, whereas signals from TGF-βs, Nodal, activins and some growth and differentiation factors (GDFs) are mediated by SMAD2 and SMAD3. Signalling from the TGF-β family ligands can also involve non-SMAD pathways such as p38, ERK1/2 and PI3K which impact on cell proliferation, apoptosis and migration [13]. A more comprehensive review on TGF-β family signalling has been published recently [14].

TGF-β family ligands are encoded by a total of 33 genes in humans, yet there are only 7 type I and 5 type II receptors mediating their signals, hence there is a high degree of promiscuity in ligand-receptor interactions. One BMP ligand can signal via different type I and type II receptor pairs, and the same type I and type II receptor pair can mediate signals from different ligands. In addition to ligand-receptor interaction, each ligand is synthesised and secreted as a prodomain bound complex; the prodomain may modify ligand bioactivity or localisation [15,16]. The extracellular regulation of BMP signalling also involves ligand traps (inhibitors) which limit ligand availability to the receptors, and cell surface co-receptors (also called type III receptors) which can modify ligand-receptor interactions [11]. Therefore, the overall signalling outcome is highly context dependent and determined by local concentrations of different ligands, ligand traps, and cell surface receptors and co-receptors [17] (Figure 1).

TGF-β and BMP signalling are also regulated at intracellular levels. Inhibitory SMADs, such as SMAD6 and SMAD7, are target genes of many ligands and can directly inhibit BMP and TGF-β signalling. Some BMPs also regulate the expression of their own receptors and co-receptors, for example, BMP9 and BMP10 induce the expression of BMPR2, ENG, SMAD9, and suppress SMAD1 expression in endothelial cells (Figure 1). Such intracellular regulation provides feedback loops, ensuring a highly dynamic yet tightly controlled BMP signalling outcome.

With such complex regulation mechanisms, it is essential to establish how mutations in different genes lead to the dysregulated BMP signalling and contribute to the pathobiology of PAH. In addition, it is essential to understand how BMPR2 mutations might affect signals from different TGF-β family ligands and in different cell types, which could contribute to the initiation or exacerbation of PAH.

Genetic findings strongly support the crucial role of dysregulated endothelial BMP signalling in the initiation of PAH. Several genes that are mutated in PAH encode proteins that are part of BMP signalling complex and highly expressed in vascular endothelial cells, such as BMPR2, ACVRL1 and ENG. Importantly, ACVRL1 is almost exclusively expressed in endothelial cells, mediating signals from two specific ligands, BMP9 and BMP10. Mutations in both GDF2 (encoding BMP9) and BMP10 have been identified in PAH patients. The clinical phenotypes of PAH patients with GDF2 and BMP10 mutations have been characterised in a recent report [18].

Endothelial dysfunction, which includes endothelial cell apoptosis, compromised barrier function, and altered vasoactive mediator release, etc, plays a central role in the initiation of PAH [19]. Reduced or loss of endothelial BMPR2 expression leads to endothelial dysfunction. In vitro, a reduction or loss of BMPR2 in human pulmonary vascular endothelial cells induces mitochondrial dysfunction and promotes a pro-inflammatory and pro-apoptotic state [20], causes endothelial-to-mesenchymal transition [21], and induces apoptosis [22,23] and excess permeability [22,24]. BMPRII deficiency impairs apoptosis via the BMPRII-ALK1-BclX (B-cell lymphoma X)-mediated pathway and the Bcl-xL isoform could be a potential biomarker for PAH [25]. In vivo, loss of Bmpr2 causes increased lung vascular permeability [22]. Conditional deletion of Bmpr2 in the pulmonary endothelium [26] or knocking-in human mutation R899X into Bmpr2 gene [22] predisposes mice to PAH. Circulating BMP ligands, mostly BMP9 and BMP10, act constitutively and potently on vascular endothelial cells [27,28], inducing BMPR2 expression [29] and have a plethora of endothelial protective functions, including anti-apoptosis, anti-migration, anti-proliferation, and anti-angiogenesis [22]. Administration of BMP9 neutralising antibody in adult mice leads to excess permeability in lung vasculature [30].

Although BMPR2 mutations in PAH were first published in 2000, structural insights into how BMPRII interacts with a BMP ligand were only reported in 2022 [31]. Crystal structures of BMPRII extracellular domain (ECD) in complex with BMP10, and in complex with both BMP10 and ALK1 ECD, revealed an unprecedented degree of plasticity in the BMPRII:BMP10 interaction [31]. This suggests that stabilising the interaction between BMPRII and BMP10 requires high concentrations of BMPRII, and that under normal physiological conditions, BMPRII-dependent signalling is most active in tissues with the highest BMPR2 expression. As lung vascular endothelial cells have the highest expression of BMPR2, along with the high expression of BMP10 in the right atrium, together they partly explain why BMPR2 mutations which cause haploinsufficiency will have the most impact on lung vasculature [31].

BMPRII is the type II receptor for all BMPs, and ubiquitously expressed in many cell types. Germline mutations in BMPR2 also affect its expression in non-endothelial cells. In pulmonary artery smooth muscle cells (PASMCs) isolated from PAH patients harbouring BMPR2 mutations, BMP4-induced SMAD1 phosphorylation and ID1 gene expression were reduced [32,33]. The growth suppressive response to BMP4 was lost in proximal PASMCs harbouring BMPR2 mutations [32].

BMP6 and BMP7 also signal in PASMCs. In one study, it was shown that the induction of ID1 and ID3 gene expression by BMP6 treatment was reduced in BMPR2 mutant PASMCs [33], In another report, the result was more complicated [34]. In this latter study using mouse Bmpr2^−/− and Bmpr2^+/− PASMCs, it was shown that while BMP7 signalling was reduced in Bmpr2^+/− PASMCs, there was a gain of BMP6 and BMP7 signalling in Bmpr2^−/− PASMCs, even when BMP2 and BMP4 signalling remained reduced upon complete knockout of Bmpr2. This suggests that after a threshold change of cell surface BMPRII to somewhere below 50%, BMP6 and BMP7 gain of signal appears. More interestingly, in these Bmpr2^−/− PASMCs, ActRIIA took over to mediate BMP4 and BMP6 signalling, and the type I receptor preference changed. ActRIIA can pair up with both ALK2 and ALK3 to mediate BMP4 signalling in Bmpr2^−/− cells, whereas BMP6 (or BMP7) employs ALK2 only when BMPRII is absent [34]. Such gain of BMP6 signalling was also observed in PASMCs when BMPRII was shed from cell surface and BMPR2 mRNA levels were reduced by more than 50% upon tumour necrosis factor-α (TNF-α) treatment [35].

The penetrance of gene mutations causing PAH is low, and a second hit is often present in the pathogenesis of PAH. Inflammation is widely accepted as a major second hit for PAH, and multiple studies have shown inflammation further contributes to dysregulated BMP signalling in pulmonary vascular cells. BMPRII deficiency promotes an exaggerated inflammatory response in human and mouse SMCs, producing higher levels of IL-6 and IL-8 after LPS-stimulation compared with controls [36]. TNF-α causes reduced mRNA expression of BMPR2 in both human pulmonary artery endothelial cells (hPAECs) and PASMCs [35], thus exacerbating the loss of BMPRII protein function in these vascular cells. BMPR2 deficiency in PASMCs conferred insensitivity to TGF-β induced growth inhibition, and this process is associated with enhanced IL-6 and IL-8 induction by TGF-β [37]. IL1-β drives an exaggerated inflammatory response when BMPR2 is deficient in PASMCs [38]. On the endothelial cell front, loss of BMPR2 leads to increased permeability in PAEC monolayers [24], and mice with an endothelial-specific knockout of Bmpr2 showed increased leukocyte recruitment and reduced barrier function [39]. In humans, aberrant immune regulation is a key feature in a significant proportion of patients with IPAH and associated with clinical outcomes, with a small subset of patients showing immunoglobin reactivity to BMPRII [40].

There are many ongoing efforts targeting BMP signalling for PAH treatment which have been reviewed recently [41–44]. Some focus on directly enhancing the cell surface BMPRII expression and such efforts include: (1) ataluren/PTC124 which promotes the read-through of stop-gain BMPR2 mutations [45,46]; (2) small chemical chaperons such as 4-phenylbutyrate (4BA) which promote the secretion of misfolded BMPRII mutant proteins trapped in the endoplasmic reticulum [47]; (3) chloroquine and hydroxychloroquine which prevent lysosomal degradation of wild type BMPRII proteins [48]. However, these approaches are not specific to BMPRII, and the efficacy and potential side effects for treating PAH are yet to be seen in humans. Another approach employs low-dose FK506 which enhanced BMP signalling and reversed PAH in rodent models [49]. FK506 binds to FKBP12 and releases it from BMP type I receptors thereby enhancing BMP signalling. In a Phase IIa randomised placebo-controlled trial, FK506 was shown to be safe, increased BMPR2 expression and improved 6-min walk distance in a subset of patients, but the overall efficacy is yet to be evaluated in a larger, multicentre trial [50]. Two approaches will be discussed further here: (1) targeting BMP9 signalling, and (2) Sotatercept, which has been approved by the FDA in March 2024 for treating PAH. Both approaches target the extracellular regulation of the TGF-β family signalling complex.

Genetic and clinical evidence strongly supports that loss of BMP9 signalling contributes to the pathogenesis of PAH. Rare heterozygous detrimental mutations in GDF2 (encoding BMP9) have been found in several large cohort genomic studies [51–53]. Patients with pathogenic BMP9 mutations have lower plasma levels of BMP9 and BMP10 [52,54]. Several homozygous null mutations in GDF2 have also been identified in paediatric PAH patients and circulating BMP9 is unmeasurable in these patients [54–57].

BMP9 is secreted from the liver and circulates at active concentrations, acting constitutively on vascular endothelium as a vascular quiescence factor [27]. BMP9 and BMP10 are the only two known high affinity ligands for ALK1. They form a signalling complex with ALK1 and BMPRII in endothelial cells and signal potently with an EC₅₀ below 0.1 ng/ml [30,58]. While BMPRII protein levels in endothelial cells reduce rapidly after protein synthesis inhibition [59], BMP9 induces BMPR2 mRNA expression in endothelial cells [29], thus forming a dynamic balance. Based on the hypothesis that the loss of endothelial BMPRII and circulating BMP9 could be rescued by supplementation of BMP9, it was reported in 2015 that administration of recombinant BMP9 reversed PAH in three different rodent models: a genetic mouse knock-in model containing a human BMPR2 mutation, the monocrotaline (MCT) — induced rat model, and a rat model induced by Sugen alongside chronic hypoxia (Sugen-Hypoxia) [22]. Here, BMP9 was also shown to confer protection against endothelial dysfunction. For example, in vitro, treatment of hPAECs with BMP9 offered protection against apoptosis induced by TNF-α and cycloheximide co-treatment, and BMP9 prevented excessive permeability in PAEC monolayer induced by TNF-α or LPS [22]. Of note, a potential beneficial role of BMP9 has also been reported in sepsis. Human patients with sepsis have lower BMP9 concentrations at admission, and lower BMP9 concentrations are associated with higher risk of death. BMP9 treatment improved the outcome in mice with experimental sepsis [60].

However, Tu et al. [61] reported in 2019 that BMP9 knockout mice, or mice administrated with a neutralising anti-BMP9 antibody, were significantly protected against chronic hypoxia-induced pulmonary hypertension. Furthermore, they showed that ALK1-Fc treatment rescued rat PAH models induced either by MCT or Sugen-Hypoxia. Such results are intriguing as they are different from hypotheses derived from human genetics. Further studies using BMP9 and BMP10 double knockout mice revealed even more complex picture where the double knockout mice developed high-output heart failure [62]. Here they also showed that BMP9 contributed to the hypoxia-induced pulmonary vascular remodelling, whereas BMP10 played a role in hypoxia-induced cardiac remodelling. In a separate study, BMP9 and BMP10 were shown to directly act on vascular smooth muscle cells and affect the contractility state of the SMCs [63]. Table 1 summarises in vivo studies supporting BMP9 agonist or antagonist approaches in the context of PAH.

The controversial observations on BMP9 extend to in vitro cell biological studies. A recent report suggests that loss of endothelial BMPR2 expression reverses the endothelial response to BMP9, causing enhanced proliferation [65]. It is difficult to compare this study with previous published data as it uses very different treatment conditions, i.e. 1 ng/ml of BMP9 treatment which is well above the concentrations measured in human plasma [54], and the experiments were performed in full growth media which already contain high concentrations of BMP9. Of note, most of the reports use serum-restricted conditions when evaluating BMP9 signalling. Altered BMP9 response in endothelial cells derived from PAH patients was also observed in another study. Here BMP9 induced unfavourable endothelial-to-mesenchymal transition only in endothelial cells isolated from PAH patients but not healthy controls [66].

With the current understanding of BMP9 and BMP10 signalling [67,68], it is difficult to reconcile these controversial observations. The functions of BMP9 and BMP10 may have more complexity than hitherto recognised, i.e. more cell types might be involved when investigating BMP9 and BMP10 effects in vivo. One might interpret such controversial findings as BMP9 exerts different roles at different stages of PAH; a beneficial role during the initial stage of PAH (where genetic studies are very powerful in identifying the underlying cause of the disease) and a more complicated role during the late stage of the disease which can often be demonstrated in cells isolated from patients who are at advanced stages of PAH. Such a scenario was seen for TGF-β, where it could act as either a tumour suppressor or a tumour promotor depending on different stages of the tumour development [69].

Sotatercept, a fusion protein comprising the extracellular ligand binding domain of ActRIIA fused to the Immunoglobin Fc domain, is approved by the FDA for treating Group 1 PAH and is the first PAH therapy targeting the TGF-β superfamily. In a Phase III clinical trial for PAH [70], Sotatercept improved the 6-min walk distance primary endpoint as well as in eight out of nine secondary efficacy endpoints compared with placebo controls, including time to death and clinical worsening [70,71]. Adverse events include epistaxis, telangiectasia, increased haemoglobin levels, thrombocytopenia and increased blood pressure, some of which may be related to the known affinity of ActRII-A with BMP and GDF ligands.

The mechanism of action of Sotatercept is still not fully understood. An early study suggests that it restores the balance between SMAD1/5/9 and SMAD2/3 signalling in PAH [72]. Here the authors showed that treatment with ActRIIA-Fc reversed elevated phospho-SMAD2/3 in a rat MCT model, but no restoration of reduced phospho-SMAD1/5/9 was observed in either rat MCT or rat Sugen-Hypoxia models [72]. Of note, Sotatercept is a ligand trap for activins and potentially also BMP9 and BMP10, so increased phospho-SMAD1/5/9 is not a direct outcome expected from Sotatercept treatment. Another study showed that treatment of ActRIIA-Fc in a Sugen-Hypoxia rat model normalised inflammatory response in the lungs, and importantly, the treatment suppressed the elevation of Inhba (encoding activin A, or ActA) and Inhbb (encoding ActB) expression in the right ventricle of Sugen-Hypoxia rats [73]. However, Sotatercept is an extracellular ligand trap and SMAD proteins are the intracellular mediators of signalling; these data still do not reveal the mode of action of Sotatercept at direct protein-protein interaction levels. We still do not know which target ligand or ligands are trapped by Sotatercept for its efficacy in PAH, nor do we know the major cell type that is responsible for the efficacy of Sotatercept.

Ligands with high affinities for ActRIIA are more likely to be bound and inhibited by Sotatercept. ActRIIA and ActRIIB are the major type II receptors for Activins, and ActRIIA has been shown to mediate signals from multiple BMP ligands using siRNA approaches [29,34,74]. In Biacore direct binding assays, ActRIIA-Fc has been shown to bind multiple TGF-β family ligands with high affinities (Table 2). For activin ligands, ActA and ActB bind to ActRIIA-Fc with the highest affinity [77,79], whereas ActC only binds to ActRIIA transiently and no reported data on ActE. ActRIIA-Fc binds tightly to several GDF and BMP ligands, with K_D in the sub-nanomolar range for GDF11 and BMP10, and in the nanomolar range for GDF8, BMP7, BMP4, BMP9 and BMP6 (Table 2, Figure 2). Interestingly, many of these ligands also bind BMPRII with high affinity (Table 3, Figure 2); for example, ActB binds to BMPRII-Fc with comparable affinity to BMP10 and stronger than many other BMP ligands [77,79]. ActA also binds BMPRII-Fc with very high affinity, but weaker than ActB or BMP10. Serum levels of both ActA and ActB are significantly elevated in PAH patients [84]. INHBA (encoding ActA) is highly expressed in lung microvascular endothelial cells [85]. PAECs isolated from the lungs of patients with IPAH synthesised more INHBA mRNA and released more ActA protein into the culture medium [85]. ActA has been shown to be capable of inhibiting BMP9 but not BMP2 and BMP4 signalling in two multiple myeloma cell lines [86], but such inhibition was not observed in endothelial cells [87]. Interestingly, it was suggested that BMPRII inhibits activin signalling via ALK2 because knocking down BMPR2 by siRNA lead to enhanced ActA-phospho-SMAD1/5 signalling via ALK2 in multiple myeloma cells [88]. This observation agrees with another report that ActA forms a non-signalling complex with ALK2 and type II Activin/BMP receptors [89]. Taken together, these reports could potentially point to a hypothesis that elevated ActA and ActB contribute to the disease progression of PAH, partly by competitive binding to BMPRII thereby further reducing the availability of BMPRII for BMP signalling and exacerbating BMPRII loss. Another potential mechanism suggested by a recent report is that binding of ActA to BMPRII leads to endocytosis of BMPRII protein, hence further reducing the cell surface BMPRII [85]. However, increased BMPRII levels or BMPRII-mediated signalling after ActRIIA-Fc or Sotatercept treatment has not been reported in either preclinical or clinical data. Nevertheless, both hypotheses predict that Sotatercept should have a beneficial effect in PAH by directly sequestering the elevated ActA and ActB. Of interest, treatment of PAECs with either BMP9 or BMP10, both in the physiologically relevant prodomain-bound forms and at physiologically relevant concentrations, can supress the expression of the INHBB gene which encodes ActB (Figure 3) [90], in agreement with a beneficial effect of a BMP9 agonist approach for treating PAH.

Human genetics and pre-clinical studies both support a fundamental role of BMP signalling in the pathogenesis of PAH. BMP and activin signalling complexes are intertwined at multiple levels, involving the competitive binding of all three BMP type II receptors, BMPRII, ActRIIA and ActRIIB. The positive outcome from the Sotatercept Phase III trial strongly suggests that dysregulated BMP signalling in PAH also involves activin signalling. A more in-depth investigation of the effect of ActA and ActB in PAH models and patient samples is warranted. Another area of BMP signalling in PAH that is not covered in this review is the shared mutations of endothelial BMP signalling components in PAH and hereditary haemorrhagic telangiectasia, which will be another important and intriguing topic to review.

• The importance of the field: pathogenic mutations in multiple components of BMP signalling pathways have been identified in human genetic studies on PAH, supporting a crucial role of dysregulated BMP signalling in the pathogenesis of PAH. Sotatercept is the first FDA-approved therapeutic modality that directly targets extracellular regulation of TGF-β/BMP signalling.

• Summary of the current thinking: compromised endothelial BMP signalling involving ALK1 and BMPRII is likely an initial trigger for PAH. Aberrant BMP signalling in other cell types and other TGF-β family ligands and receptors may also contribute to the pathogenesis of PAH.

• Future directions: a deeper mechanistic insight into the extracellular regulation of signalling from BMPs, activins and GDFs may provide novel therapeutic opportunities. This could be achieved by in vitro cell signalling assays in a physiologically relevant context, such as using human primary cells, patient cells and co-culture models, and coupled with biochemical and structural studies to address the direct protein–protein interactions amongst ligands and receptors.

The authors declare that there are no competing interests associated with the manuscript.

W.L. is funded by a British Heart Foundation Senior Basic Science Research Fellowship (FS/SBSRF/20/31005). K.Q. is funded by British Heart Foundation 4-year PhD programme (FS/4yPhD/F/22/34170C). This research received support by the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.

Open access for this article was enabled by the participation of University of Cambridge in an all-inclusive Read & Publish agreement with Portland Press and the Biochemical Society under a transformative agreement with JISC.

W.L. wrote the manuscript. K.Q. generated Figures 1 and 2 and reviewed the manuscript.

We thank Prof. Nick Morrell, Dr Mark Toshner and Dr Jingxu Guo for critically reviewing the manuscript and providing constructive comments.

ALK1

activin receptor-like kinase 1

BMP

bone morphogenetic protein

ECD

extracellular domain

GDF

growth and differentiation factor

hPAEC

human pulmonary artery endothelial cell

IPAH

idiopathic PAH

MCT

monocrotaline

PAH

pulmonary arterial hypertension

PASMC

pulmonary artery smooth muscle cells

TGF-β

transforming growth factor-β

TNF-α

tumour necrosis factor-α