Heterozygous germline defects in a gene encoding a type II receptor for bone morphogenetic proteins (BMPR-II) underlie the majority of inherited cases of the vascular disorder known as pulmonary arterial hypertension (PAH). However, the precise molecular consequences of PAH causing mutations on the function of the receptor complex remain unclear. We employed novel enzymatic and fluorescence activity based techniques to assess the impact of PAH mutations on pre-mRNA splicing, nonsense-mediated decay (NMD) and receptor complex interactions. We demonstrate that nonsense and frameshift mutations trigger NMD, providing further evidence that haplo-insufficiency is a major molecular consequence of disease-related BMPR2 mutations. We identified heterogeneous functional defects in BMPR-II activity, including impaired type I receptor phosphorylation, receptor interactions and altered receptor complex stoichiometry leading to perturbation of downstream signalling pathways. Importantly, these studies demonstrate that the intracellular domain of BMPR-II is both necessary and sufficient for receptor complex interaction. Finally and to address the potential for resolution of stoichiometric balance, we investigated an agent that promotes translational readthrough of a BMPR2 nonsense reporter construct without interfering with the NMD pathway. We propose that stoichiometric imbalance, due to either haplo-insufficiency or loss of optimal receptor–receptor interactions impairs BMPR-II mediated signalling in PAH. Taken together, these studies have identified an important target for early therapeutic intervention in familial PAH.
Bone morphogenetic proteins (BMPs) are required for cellular differentiation and mammalian development (1). BMP signalling is transduced by one of two pathways, each with distinct cellular outcomes. In the first, ligand stimulation triggers assembly of the receptor complex through rapid association and diffusion. In the second pathway, the receptor and its associated components exist in a so-called ‘pre-formed complex’. Activation of the receptor complex generates a phosphorylation relay of cytoplasmic signalling proteins, including the Smad family, which translocate to the nucleus to directly regulate gene transcription (2). BMPs and transforming growth factors (TGF-β) have also been reported to activate Smad independent signalling, including ERK, JNK and p38MAPK (3–6).
The BMPR2 gene, comprising 13 exons, encodes a type II receptor for bone morphogenetic proteins (BMPR-II), a member of the TGF-β receptor family. Alternative splicing of exon 12 may give rise to a long isoform (LF), generating a protein product of 1038 amino acids, or a shorter transcript of poorly defined functional significance. The mature receptor consists of four functional regions, namely the ligand-binding, transmembrane, kinase and C-terminal domains.
Heterozygous mutations of BMPR2 have been identified in subjects with the progressive vascular disorder pulmonary arterial hypertension (PAH) (7). PAH is characterized by the abnormal proliferation of endothelial and smooth muscle cells. The resultant vascular remodelling of the small pulmonary arteries leads to elevated pulmonary artery pressure and, in the absence of effective therapy, eventual right-heart failure (8,9).
To date, over 200 independent heterozygous mutations have been detected in PAH subjects of which ∼70% introduce a premature termination codon (PTC). Typically, PTC harbouring transcripts are degraded by a process termed nonsense-mediated decay (NMD) (10). However, these mutations can also lead to splicing anomalies including exon skipping or alteration of exonic and intronic splicing enhancer (ESE and ISE) and suppressors (ESSs and ISSs) (11–14).
Missense mutations are dispersed across all four domains of the receptor and have heterogeneous functional consequences. The predominant type of missense mutations in the ligand-binding domain is cysteine substitutions that impair signalling due to mutant receptor mislocalization in the cytosol (15). Interestingly, a small number of non-cysteine mutations have been reported in PAH subjects with associated congenital heart disease (CHD) and in patients previously exposed to amphetamine like agents, a known risk factor for disease (16–18). Mutant receptors bearing non-cysteine substitutions localize to the cell surface but also exhibit defects in signalling activity (4,15,19). By contrast, mutations in the cytoplasmic C-terminal domain only moderately inhibit Smad-signalling (4,15,19).
In this report, we investigated the functional consequences of a wide range of BMPR2 mutations, identified in subjects with familial, spontaneous or ‘idiopathic’ and associated forms of PAH. In addition, we assessed an agent for its capability to promote translational read-through in constructs harbouring BMPR2 nonsense mutations. Together, these studies suggest that dysregulated stoichiometry of the multi-protein receptor complex contributes to aberrant BMPR-II signalling in PAH and that restoration of the stoichiometric balance by stimulating translational readthrough of mutant alleles might provide an effective therapeutic intervention prior to or following the onset of disease.
Effect of mutations introducing PTC on BMPR2 mRNA expression
As 70% of mutations are predicted to introduce PTC in the BMPR2 reading frame, we first investigated the impact of typical mutations upon BMPR2 mRNA expression. Using computer simulation, such as GENSCAN, we first investigated the consequence of nonsense mutations on BMPR2 coding structure. GENSCAN predicted that insertion (c.2292insA), deletion (c.2386delG) and nonsense mutations (c.2620G>T and c.2695C>T) within exon 12 of BMPR2, each might activate cryptic splice sites (see supplementary Material, Table S2). If correct, this would lead to the skipping of the nonsense codon and maintenance of a reading frame (Fig. 1A).
To further investigate the consequence of mutations predicted by GENSCAN, we utilized a novel dual-fluorescence-based assay system (Nasim et al., 2008, submitted for publication). In this system, successful splicing would lead to the production of DsRed-Express-GFP fusion protein (Fig. 1B). Expression of only the upstream reporter would occur should a pre-mature termination codon (PTC) be recognized as a translation stop signal. By way of contrast, should PTC trigger aberrant splicing, a shorter form of the fusion protein would be produced. Fluorescence microscopy revealed that the wild-type sequence underwent splicing leading to the expression of both fluorescence proteins (Fig. 1C). However, each of the BMPR2 mutations tested namely, c.2292insA, c.2386delG, c.2620G>T and c.2695C>T failed to produce the dual-fluorescence reporter indicating that mutation had led to the incorporation of a translation stop signal within exon 12. Direct RNA analysis confirmed that the dual-fluorescence was consequent upon efficient splicing of wild-type sequence (Supplementary Material).
Mutations introducing PTCs are subject to nonsense-mediated RNA decay
We next sought to investigate whether PTC bearing mutations of BMPR2 triggered the NMD machinery. To study this, we employed a novel enzyme-based NMD assay (Nasim et al., 2008, submitted for publication). The system was developed in such a way that the incorporation of the PTC would lead to the truncation of a full-length protein (Fig. 2A). Thus, if the mRNA is subjected to NMD, inhibition of translation using cycloheximide or puromycin would lead to the increased level of PTC bearing mRNA with a consequence of increased expression of the truncated protein. Transfection of the reporter constructs into HEK293 cells led to a significant increase in the reporter protein following treatment with cycloheximide (5-fold) and puromycin (20-fold) (Fig. 2B). To confirm that these mutations were subject to NMD, plasmids encoding siRNA that inhibit expression of hUpf1 (20) were transfected together with the reporter plasmid. Compared with cells transfected with the reporter alone, the gal-luc ratio increased between 3 and 6-fold when hUpf1 RNA was knocked down by RNAi (Fig. 2C). Finally, we demonstrated NMD in pulmonary artery smooth muscle cell lines (21) harbouring a nonsense mutation (exons 4 and 5 deleted) derived from bmpr2+/− mice (22) (Fig. 2D).
Mutation in the exon or intron modulates BMPR2 pre-mRNA splicing
Point mutations may dysregulate splicing through the creation of ESSs or by disruption of ESEs (23,24). To investigate splicing efficiency, we examined PAH causing mutations of BMPR2 exon 9 by constructing a splicing reporter adapted from a recently reported method (25). The efficiency of splicing of the wild-type BMPR2 sequence was established as 100%. These studies revealed that mutation at c.1270T>C had no significant impact yet and in contrast, c.1259G>A substantially reduced splicing efficiency (Fig. 3B).
We next investigated whether BMPR2 mutations may trigger alternative splicing using an exon-trapping reporter derived from a previously described system (26). Mutations were selected as follows; an exonic mutation (c1241G>A) predicted to introduce a PTC (p.W414X) and an intronic (c.1276+4A>G) mutation located in a conserved (70–75%) consensus 5′ splice site. RNA analysis revealed that the wild-type sequence led to significant incorporation of exon 9. In contrast, mutation in the intron at position c.1276+4A>G triggered exon skipping whereas the c.1241G>A mutation had no discernible effect on exon incorporation (Fig. 3D).
BMPR-II mutations exhibit differences in transcriptional activity of a BMP-responsive reporter
Having established the consequence of mutation on receptor expression, we next sought to investigate the functional activities of wild-type and mutant BMPR-II. Smad-induced transcriptional activity mediated by wild-type BMPR-II in the presence and absence of BMP4 ligand were examined using 3Gc2wt-Lux in HEK 293 cells. The activity of 3Gc2wt-Lux reporter, a BMP responsive promoter–reporter construct (27) was enhanced between 3 and 6-fold following the overexpression of BMPR-II and BMP4 stimulation, respectively (Fig. 4A), and was further increased to 20-fold when cells transfected with BMPR-II were stimulated by BMP4. We next tested a range of mutations (Fig. 4B) including atypical non-cysteine substitutions within the ligand-binding domain (p.Q42R, p.G47N, p.Q82H), the linker region between transmembrane and the kinase domain (p.G182D, p.M186V), the kinase domain (p.D485G, p.E503D) and the C-terminal domain (p.R899X, p.R899P) for their ability to stimulate the reporter. Of interest, the transcription activities stimulated by the mutants at position p.Q42R, p.G47N, p.G182D, p.M186V and p.E503D were comparable with that stimulated by the wild-type receptor (Fig. 4C). In contrast, p.Q82H, p.D485G and p.R899X mutants were unable to activate reporter expression, while p.R899P construct generated a slight increase in activity. These data indicate that the basal activity of these three constructs as measured by stimulation of the reporter was abrogated. We next investigated whether stimulation with BMP4 ligand was capable of inducing a response. BMP4 upregulated the activity of the mutant p.R899X receptor albeit to a level lower than that of wild type (Fig. 4D), whereas p.Q82H, p.D485G and p.R899P mutants remained severely functionally impaired. In contrast, the remaining mutations displayed similar signalling characteristics to the wild-type receptor.
As type I receptors are involved in BMP mediated signalling events, we next investigated their impact on reporter activation. The effects of type I receptors, namely ALK3 (also known as BMPR1A) and ALK6 (BMPR1B) on the ability of BMPR-II to stimulate transcription were examined by overexpressing plasmids encoding ALK3 and ALK6 along with BMPR-II (Fig. 4E). The activities of the BMPR-II stimulated reporter were further increased between 5 and 6-fold by over-expression of ALK3 and ALK6, respectively. Both ALK3 and ALK6 promoted the activity of p.Q82H mutant to a level comparable to the wild type. In the presence of the type I receptors, the p.R899P and p.R899X mutants exhibited increased signalling activity. However, the level of augmentation was lower than that of wild type. Neither ALK3 nor ALK6 stimulated the p.D485G mutant. We next tested the impact of BMP4 on ALK3+BMPR-II and ALK6+BMPR-II stimulated reporter. With the exception of p.D485G mutant, BMP4+ALK3 increased observed activity of all mutants, to a level comparable to wild type. BMP4+ALK6 mediated increase for mutants p.R899X and p.R899P was significantly lower than that of wild type (Fig. 4F). The p.D485G mutant was not simulated at all.
Mutation in the kinase domain of BMPR-II affects type I receptor binding and phosphorylation
As concomitant overexpression of type I and type II receptors enhanced transcriptional activation of a reporter construct, we next examined whether these two receptors physically interacted and sought the consequence of mutation upon the interaction. To investigate whether BMPR-II binds ALK3, we performed co-immunoprecipitation experiments in Hela cells using plamids encoding myc-tagged wild-type BMPR-II and mutant receptors (p.C118W, p.C483R, p.D485G, p.N519K) along with HA-tagged ALK3. As shown in Figure 5A, the wild-type BMPR-II binds to ALK3. Mutation in the ligand-binding domain has no discernible effect on the interaction but the ability of the p.D485G, p.C483R and p.N519K mutants to bind ALK3 is markedly reduced. We then tested the ability of wild-type and mutant BMPR-II constructs to phosphorylate the type I receptor (Fig. 5B). By comparison to wild type, both the p.R899X and p.N519K mutants demonstrated a reduced ability to phosphorylate ALK6, whereas both p.D485G and p.C483R were effectively unable to mediate phosphorylation.
The efficiency of type I and type II receptor interactions
To identify the functional domains required for receptor interactions and determine the efficiency of interactions, we developed a novel method based on a dual-light reporter system (28). In brief, the red fluorescence protein signals regardless of receptor–receptor interactions (Fig. 6A) and acts as a reference standard. In the event of a protein–protein interaction both red and green fluorescence proteins are produced. Expression of the GFP protein was not activated when reporter was transfected alone (6Bi) or co-transfected with LF (6Bii) and p.D485G (6Biii). The expression of GFP was activated when ALK3 was transfected with BMPR-II (6Bv), which demonstrated that the intracellular domains of both receptors are necessary and sufficient for interactions. The expression of GFP was not detected following the overexpression of ALK3 together with the p.D485G mutant form of a BMPR-II receptor.
To determine the efficiency of interaction of BMPR-II and ALK3, we employed a gal-luc-based dual-reporter assay system. The dual-reporter (28) (pTN114) along with the respective plasmids was transfected into HEK293 cells and their luciferase and β-galactosidase activities were determined. Cells overexpressing recombinant wild type, p.D485G and ALK3 alone failed to activate luciferase expression (Fig. 6C). However, that the expression of luciferase was activated by overexpression of plasmids containing ALK3 and BMPR-II further confirmed that both proteins interacted with each other via the intracellular domains. We next investigated the intensity of interaction of BMPR-II with ALK3 or ALK6. The gal-luc ratio indicated that ALK3 interacts more efficiently with BMPR-II compared to ALK6 (Fig. 6D). Since the reporter assays suggest that BMPR-II interacts with ALK3 and ALK6, we next sought to investigate the consequences of a series of missense mutations located in the kinase domain of BMPRII on receptor complex interactions. For the mutations at position p.A313P, p.C347Y, p.D485G and p.R491Q, interactions with ALK3 and ALK6 were impaired (Fig. 6E). In contrast, p.L401S and p.E503D had no discernible effect upon these receptor interactions.
Translational readthrough of nonsense codons by aminoglycosides
Having demonstrated that PAH causing nonsense mutation of BMPR2 triggers the NMD pathway, we next progressed to investigate an approach by which a significant amount functional protein can be generated from a nonsense transcript. As the NMD pathway depends on a set of factors that modulate both RNA stability and translation termination efficiency, inactivation of any of these stabilizes nonsense transcripts and promotes translation readthrough (10). Aminoglycosides such as gentamicin have been shown to promote translational readthrough of PTC in mammalian cells and animal models of nonsense-associated diseases permitting expression of full-length functional protein (29–31). We, therefore, sought to determine whether such agents might protect BMPR2 nonsense transcripts from NMD. To investigate the effect of aminoglycosides on transcript stability, pulmonary artery smooth muscle cells derived from a PAH subject harbouring a nonsense mutation (p.W9X) were grown in the presence and absence of gentamicin. QPCR data indicated that the level of BMPR2 transcript was significantly reduced in p.W9X cell line, providing evidence of NMD. We observed no significant increase in the level of BMPR2 transcript following gentamicin treatment (Fig. 7A). In the absence of an antibody for reliable detection of endogenous full-length BMPR-II protein in a quantitative assay, we adapted a cell-based approach to measure the impact of gentamicin exposure. We employed a double reporter to determine the efficiency of stop codon suppression (Fig. 7B). We found that the treatment enhanced translational readthrough of the reporter construct (Fig. 7C). Finally, a reporter construct containing PTC introduced by a BMPR2 insertion mutation (Fig. 7D) showed an enhanced level of beta-galactosidase activity following gentamicin exposure (Fig. 7E).
The demonstration that heterozygous mutations of the BMPR2 gene underlie inherited PAH has provided a significant opportunity to advance our understanding of the molecular basis of this typically fatal disease. However, it remains a significant challenge to resolve the consequence of BMPR2 defects in predisposing to vascular disease. In an earlier study, we hypothesized that PAH causing nonsense mutations of BMPR2 are likely to follow the NMD pathway and thus generate a state of haploinsufficiency for the BMPR-II protein (32). We have used a number of independent approaches to generate a body of data to support this proposal. First, utilizing novel fluorescence-based assays, we found that mutations that introduce a PTC fail to stimulate aberrant splicing (Supplementary Material, Figs S1 and S2), a possible consequence of nonsense mutation and one predicted by computer simulation (see Supplementary Material). In this study, we investigated a range a disease causing defects that reflects the mutation diversity identified in PAH cohorts (33). Secondly, we wished to perform quantitative assay of BMPR2 transcripts, but were hampered by the limited availability of samples of relevant cell type from subjects with PAH. We, therefore, developed an enzyme-based reporter assay that provided evidence of transcripts loss which was shown by use of inhibitors to be due to activation of NMD pathway (Fig. 2A–B). We confirmed these data by analysis of ex-vivo mouse (22) and human cells (34) (Figs 2D–E and 7A). Taken together, we conclude that the nonsense mutations reduce the abundance of mRNA through the NMD pathway and thereby generate haplo-insufficiency for BMPR-II protein, a situation that contributes to stoichiometric imbalance of the receptor complex (Fig. 8).
We next sought to gain insight as to the molecular consequences of less frequently observed class of defects, namely missense mutations. Direct association of receptor association as assayed by co-immunoprecipitation, unexpectedly demonstrated that type I and type II receptor interactions do not take place through the ligand-binding domain (Fig. 5A). The kinase domain appears both necessary and sufficient for receptor–receptor interactions as proteins containing amino-acid substitutions within this region are unable to precipitate the type I receptor. These findings were further supported by dynamic activity-based assays performed using living cells (Fig. 6). In these studies, we confirmed that type I and type II receptors are capable of interacting with each other through the intracellular domains, an interaction that appears independent of ligand stimulation. We suggest that these observations provide further insight as to the mechanisms of BMPR-II mediated signalling. It has previously been speculated that ligand binding to either the type I or type II receptor, brings receptor components into close proximity and thereby leads to the formation of a functional complex. We interpret our data to demonstrate that receptors form a functional complex but do so through their intracellular domains, a finding that provides a molecular basis for the existence of so-called ‘preformed receptor complex’ (35).
Having demonstrated the significance of the kinase domain for receptor–receptor interplay, we progressed to study the importance of specific residues by interrogation of BMPR2 missense mutations identified within this region. We found that a majority of such defects were severely impaired in type I receptor interactions (Figs 5 and 6), which indicated that several residues (e.g. p.A313P, p.C347Y, p.D485G and p.R491Q) are more important than others (p.L401S) to retain the interaction. Interestingly, among these mutations p.D485G, pC347Y and p.R491Q were defective in BMPR-II-mediated signalling (15).
We next wished to investigate less frequently observed sequence variation detected in subjects where the development of PAH occurred in association with other recognized risk factors for disease, namely the presence of CHD or exposure to appetite suppressants. As representative of this class of variation, we selected the p.E503D variant. Interestingly, we found that this amino-acid substitution was capable of interacting with type I receptors suggesting that such a receptor complex may be capable of conveying signalling. As predicted, p.E503D generated Smad-dependent signalling at levels more or less comparable to that seen in wild type (Fig. 4). We therefore investigated additional reported amino-acid substitutions from this class. Importantly, we found that many of these DNA variants conveyed BMPR-II-mediated signals following ligand stimulation and type I receptor overexpression (Fig. 4). We thus conclude these variants do not share functional consequences with those previously characterized in BMPR2 mutations identified in familial PAH (16–18) findings that raise the question as to whether they may more appropriately be described as variants of unknown clinical significance. We acknowledge, however, that we cannot rule out that such variants may exhibit more subtle effect on either BMPRII expression or function and specifically in relation to the recognized critical role of this pathway during mammalian development.
We therefore propose that stoichiometric imbalance due to haploinsufficiency and/or optimal receptor component interaction is the major molecular consequence of pathogenic mutations contributing to dysfunctional BMPR-II-mediated signalling in PAH (Fig. 8). Restoration of stoichiometry either by increasing the expression of full-length protein or promoting the receptor interactions might provide protection from disease development. To this end, we tested whether the aminoglycosides can promote readthrough of nonsense mutations found in PAH. Aminoglycosides such as gentamicin have been shown to enhance translational readthrough in mammalian cells (29,30) and in animal models of nonsense mutation diseases including Duchenne muscular dystrophy and cystic fibrosis (CF) (31,36). Indeed, gentamicin-dependent correction of CF transmembrane conductance regulator (CFTR) function has been reported in early clinical translational studies (37). In our work, we observed that treatment of cells, to which we had introduced a BMPR2 reporter construct harbouring a nonsense codon, promoted reporter activity without affecting NMD pathway. These observations together with increasing interest in the development of chemicals that specifically target nonsense mutations provide a strong platform for further studies that should examine the clinical potential for the treatment of nonsense-associated PAH mutations.
MATERIALS AND METHODS
The details of the plasmid construction and list of the constructs (Supplementary Material, Table S1) used in this study are available upon request.
The effect of mutation on BMPR2 transcript integrity was analysed by GENSCAN (38) (http://genes.mit.edu/GENSCAN.html) algorithm. Gene transfer, cell culture, reporter assays, fluorescence microscopy, co-immunoprecipitation, treatments with chemicals and siRNAs were performed as described elsewhere (12,20,25,26,28,29,39). The amount of plasmid DNA transfected into the HEK293 cell line varied from (from 30 ng to 1 µg) assay to assay.
Quantitative PCR for determining BMPR2 transcripts was performed using TaqMan Gene Expression Assay (Applied Biosystems) on 7900HT Fast Real-Time PCR system (Applied Biosystems) according to manufacturer’s protocol. The detailed protocols for immunoprecipitation, immunoblot and kinase assays are described in the supplemented method section.
The research is supported by a Programme Grant (RG/03/005) from the British Heart Foundation awarded to NWM and RCT. This study also received financial support from the European Commission under the 6th Framework Programme (Contract No LSHM-CT-2005-018725, PULMOTENSION).
The authors express sincere thanks to R. Machado for critically reading through the manuscript.
Conflict of Interest statement. None declared.