TNFα drives pulmonary arterial hypertension by suppressing the BMP type-II receptor and altering NOTCH signalling

Heterozygous germ-line mutations in the bone morphogenetic protein type-II receptor (BMPR-II) gene underlie heritable pulmonary arterial hypertension (HPAH). Although inflammation promotes PAH, the mechanisms by which inflammation and BMPR-II dysfunction conspire to cause disease remain unknown. Here we identify that tumour necrosis factor-α (TNFα) selectively reduces BMPR-II transcription and mediates post-translational BMPR-II cleavage via the sheddases, ADAM10 and ADAM17 in pulmonary artery smooth muscle cells (PASMCs). TNFα-mediated suppression of BMPR-II subverts BMP signalling, leading to BMP6-mediated PASMC proliferation via preferential activation of an ALK2/ACTR-IIA signalling axis. Furthermore, TNFα, via SRC family kinases, increases pro-proliferative NOTCH2 signalling in HPAH PASMCs with reduced BMPR-II expression. We confirm this signalling switch in rodent models of PAH and demonstrate that anti-TNFα immunotherapy reverses disease progression, restoring normal BMP/NOTCH signalling. Collectively, these findings identify mechanisms by which BMP and TNFα signalling contribute to disease, and suggest a tractable approach for therapeutic intervention in PAH.

Where they depart from the published literature is their observations on notch signalling. They observe that TNFα increases NOTCH1 and NOTCH2 but reduces NOTCH3 levels, effects enhanced by BMP6. They argue that NOTCH 1 and 2 are proproliferative and NOTCH3 is antiproliferative. This is at odds with the published literature (see reference list below) where levels of NOTCH3 protein are regarded as a sensitive molecular marker of severity of PAH in humans and pulmonary hypertension in rodents. Notch signaling is involved in vascular development and lung tissue from patients with pulmonary hypertension has been reported to show increased NOTCH3 and NOTCH3 intracellular domain expression when compared with normotensive patients. Additionally, NOTCH3 and NOTCH3 intracellular domain expression has been shown to be increased in two animal models of pulmonary hypertension-hypoxia-induced pulmonary hypertension in mice and monocrotoline-induced pulmonary hypertension in rats. The authors reference the difference and suggest that their γ-secretase inhibitor used to inhibit NOTCH cleavage is more selective than the DAPT agent used in published reports but they do not address the fact that NOTCH3 knockout mice are resistant to pulmonary hypertension.
A number of cell lines are used in the study. The eventual focus is on distal pulmonary vascular sooth muscle cells. It is understandable that the investigators focus on one cell type/model, but it does make extrapolation to the whole animal problematic. They have attempted this with mouse models but these do not reflect the human condition. The challenge in extrapolating to humans is evident in the emerging literature where antibodies to DLL4, a NOTCH ligand, in development for the treatment of a range of tumours have been reported to cause pulmonary hypertension in patients.
Overall, the data are of high quality, yet there are some concerns regarding the interpretation of the results related to Notch signaling and some additional experimental questions regarding the shedding of the BMPR2 that remain to be addressed.
As to specifics: 1) In the Western blots for ADAM10 and ADAM17 in supplementary figure 5c, it is not clear whether the pro-or mature form of ADAM10 and ADAM17 are shown (please also include molecular weight markers). Cell lysates for Western blots of ADAM17 should be prepared in the presence of a metalloprotease inhibitor such as TAPI-1 and 10 mM 1,10 phenanthroline to block the autodegradation of the mature form, which occurs rapidly following cell lysis in the absence of these inhibitors, and can thus affect the interpretation of the levels of mature ADAM17. It is also important to point out that increased expression of ADAM17 does not necessarily result in increased activity, since this is a posttranslationally regulated metalloproteinase (see, for example, PMID:23342154). Nevertheless, the overall interpretation that both ADAMs contribute to shedding of the BMPR2 under the conditions used in this study is well supported by the results. Figure 5D presumably needs to be revised so that siADAM17 and D1(A12) are shown next to ADAM17.

2) Supplementary
3) Regarding the interpretation of data related to Notch signaling, it is important to point out that increased Notch expression does not necessarily lead to increased Notch signaling. Notch signaling is regulated through binding of a membrane-anchored ligand such as Jagged1 or 2, or Dll1, 3 or 4, on a signal-sending cell to a Notch receptor on a signal-receiving cell. The resulting endocytosis of Notch and its ligand is thought to pull open the membrane-proximal negative regulatory domain, allowing access of ADAM10 to the Notch cleavage site. So increased expression of Notch or of ADAM10 will not lead to increased signaling in the absence of this ligand-induced activation of Notch. The two references implicating ADAM17 in Notch signaling are not representative of the majority of papers in this field, in which targeted deletion of ADAM10 in mice typically results in Notch-dependent defects, whereas targeted deletion of ADAM17 does not. Moreover, the Western blots of Notch2 and Notch3 do not show the Notch intracellular domain, but instead show the typical S1-cleaved Notch fragments of around 100 -120 kD (please include molecular weight markers on all gels). This fragment is constitutively generated by cleavage through furin (or related pro-protein convertases), and is not indicative of signaling (PMID:19379690 provides one of several excellent reviews on this topic). The S2 fragment, generated by ADAM10 under physiological conditions, and the S3 fragment, generated by gamma-secretase, are indicative of ligand-induced Notch processing and signaling, but these are most likely not shown on the various Notch blots presented here. So the authors must make it very clear that these blots only allow conclusions to be drawn about Notch expression levels, not about Notch signaling (the same is true for histochemical data).
Results obtained with DAPT, on the other hand, are indicative of a block of Notch signaling, and can be interpreted in this manner. However, the authors should also discuss that Notch2 and Notch3 might not be active in the same cell types, and in general, offer a more careful interpretation of their data related to Notch expression levels. It is very important to distinguish between differences in Notch expression and Notch signaling, which do not have to be directly related, and to consider different effects of Notch signaling depending on the cell types where it occurs.
Reviewer #3 (expert in PAH and BMP receptors) Remarks to the Author: General Comments: The authors have extensively studied the interaction between BMP and Notch signaling that is dysregulated in response to a specific inflammatory stimulus, TNF-alpha. The cell biology studies have been carefully carried out and the main conclusions indicated in the summary are certainly substantiated and are novel. There are however, other studies in the literature that clearly indicate important points of interaction between the BMPR2 pathway and inflammation not cited, including TNF alpha beginning with Song et al (Circulation, 2005:112;552-62) but also Hagen et al (Am J Physiol 2007: 292;1473-9), Lawrie et al, Am J Pathol 2008: 172: 256-64) Sawada et al, 2014263-80) Diebold et al, (Cell Metab 2015: 21;598-608) and many others. Also there are papers on Notch and BMP crosstalk in blood vessels (Rostama et al, ATVB 2015: 35;2626-37). The authors also do not reconcile their results with their recent studies related to rescue of PAH by the ligand BMP9, in fact surprisingly, work with BMP9 as a ligand seems to be completely excluded. The SUGEN/hypoxia model treated with the TNF soluble receptor reverses or retards progression of the pressure and right ventricular hypertrophy in concert with elevating Notch signaling but the impact on remodeling is less severe and the role of TNF alpha (despite its elevation) vis a vis other cytokines is not clear.
Specific Comments: Introduction: Page 3 last paragraph, see above comments. The impact of TNFalpha and loss of BMPR2 was studied in Sawada et al cited above, although the paper did not show that TNF specifically reduces BMPR2 levels in endothelial cells. Page 5 first paragraph: The mechanism by which TNF reduces BMPR2 in endothelial cells is never addressed. Page 6, second paragraph. It is hard to visualize why two disintegrins are necessary to cleave the same site. This needs further explanation. Page 6, third paragraph. In view of the authors' recent work showing a protective effect of BMP9, it would be interesting to determine whether BMP9 signaling is compromised by TNF alpha. (Also true for data presented on Page 7). Page 10: The controversy with the Li et al paper cited is still problematic. Notch3 and Notch3 ICD are elevated in PAH PA SMC and in human PAs in tissue sections. The blots in Figure 4 should definitely be quantified because it appears that Notch 3 mRNA is higher in the PAH vs. control PA SMC. It is hard to make much of differences in immunoblots and immunohistochemistry because they rely on the affinity of the relative antibodies. The authors should be sure that the siRNAs they use are specific, ie. Do Notch 3 and Notch 2 reduce with Notch 2 siRNA? Does Notch 3 but not Notch 2 decrease with siRNA for Notch 3. Those data should be included. Finally it does appear from the mRNA data in the supplement that the effect of Smad6 induced by TNFα is different, increasing Notch 2 and reducing Notch 3. Thus the TNF effect may be selective to Notch 3 particularly in the setting of loss of BMPR2. In the mice, knockout of Notch 3 is protective, so there is more to try to reconcile either in the text or in the discussion. It should also be noted that the PASMC come from familial PAH with a BMPR2 mutation so here again the populations are different.
Page 11: The selective effects of the FYN and YES siRNAs should be documented in view of the results.
Interestingly the TNF overexpression does not result in more severe RVSP or RVH when BMPR2 is heterozygous. On the other hand, in the SUGEN/Hypoxia model there is lcear suppression of RVSP and RVH by the TNF soluble receptor but less impact on the muscularization of vessels. The authors should discuss these differences.

Discussion
Page 13: The discussion with respect to Notch seems to short-change the previous observations, particularly related to the prevention of pulmonary hypertension in the Notch 3 knockout mice. There is no question, however, that the mechanistic studies in this paper are comprehensive and well thought through.
Reviewer #4 (expert in PAH and BMP receptors) Remarks to the Author: αThe manuscript describes an interesting and complex system involving a switch in TGFβ receptor super-family usage in PASMCs from an ALK3 / BMPR-II-driven system to an ACTR-IIA / ALK2 systemdriven primarily by TNFα degradation of BMPR-II and the induction in expression of various genes including BMP6 and ACTR-IIA. In the background of compromised BMPR-II expression / function as might be found in PASMCs from patients with heritable forms of Pulmonary Arterial Hypertension (PAH) and genetic mouse models of PAH, this alteration in receptor usage results in a switch from an anti-proliferative response of these cells to a pro-proliferative response following stimulation with BMP6. To add to the complexity, TNFα-mediated modulation in the expression of components of the NOTCH pathway with induction and repression of NOTCH2 and NOTCH3 respectively, directly contributes to the pro-proliferative effects of PASMCs in a Src-kinase family-dependent manner. Although the mechanism described in the manuscript is novel and provides several new therapeutic options for the treatment of heritable PAH, I do have some questions that need to be addressed and suggestions for improvement of the manuscript outlined below before it should be considered for publication. 1. The authors present data showing that TNF induces degradation of BMPR-II via an ADAM10/17mediated mechanism that appears to be operative in PASMCs but not in PAECs. The expression of BMPR-II is clearly suppressed in PAECs in response to TNFαstimulation, however the authors do not provide any explanation as to the mechanism in this cell type. Do the authors have any mechanistic insights into how BMPR-II protein levels are regulated in PAECs via TNFαstimulation that could be included in the manuscript? 2. The authors have used the SP-C/TNF transgenic mouse model to demonstrate that BMPR-II expression in the whole lung of these animals also appears to be under the control of the TNFαpathway. The data presented by the authors seems to imply that the expression of BMPR-II in whole lung homogenates is almost completely suppressed in this model. This finding seems incongruent with the alveolar epithelial cell-restricted expression of TNFαpreviously described in this model (Miyazaki et al. 1995J Clin Invest. 199596(1):250-9). Immunohistological analysis of BMPR-II and TNFαexpression in the lungs of these animals would provide additional spatial information confirming that loss of BMPR-II expression is co-localised to areas of TNFαexpression. 3. The legend for Figure 1e needs to have a better description of what stains / proteins the colours in the images refer to. From the images presented, it also appears that αSMA positive cells appear to express TNFαwhich might suggest that PASMCs isolated from patients with heritable PAH may express higher baseline levels of TNFαin culture. Have the authors determined whether this is the case? 4. The authors have cited information showing that the BMPR-II cytoplasmic tail can act as a scaffolding element that can bind to and modulate the activity of a number of kinases including c-Src (Wong et al. Am J Respir Cell Mol Biol. 2005 Nov;33(5):438-46). Can I presume that the generation of the BMPR-II ICD might still be capable of binding to kinases like c-Src and modulating their activity? If so, it would then be important to know how long the BMPR-II ICD remains within the cytoplasm. 5. Previously published information by the authors has shown that PASMCs isolated from patients with PAH lose their growth suppressive responses to BMP ligands including BMP2 and BMP4 (Morrell et al. Circulation. 2001 Aug 14;104(7):790-5), which is in contrast to the data depicted in Figure 2D. Could the authors provide an explanation for the differences observed in this study compared to the previously published report? 'BMPR2 silencing also promoted the TNFα-dependent reduction of NOTCH3-ICD generation and NOTCH3 transcription'. On inspection of Figure 4a, TNFα does induce a reduction in NOTCH3 ICD levels; however, BMPR2 siRNA does not enhance the effect of TNFαon NOTCH3 ICD levels compared to the siCP or DH1 control. Similarly, in Supplementary Figure 14c, TNFαdoes reduce the transcription of NOTCH3 mRNA; however, this effect is not enhanced by BMPR2 siRNA compared to the siCP or DH1 controls. c) Also on page 9 of the manuscript the authors state in Figure 4b 'In BMPR2 heterozygous HPAH PASMCs treated with TNFα, siACVR2A reduced NOTCH2-ICD generation and abrogated NOTCH3-ICD'. On inspection of Figure 4b, treatment of HPAH PASMCs with siACVR2A seems to be without effect on TNFα-mediated NOTCH2-ICD generation. d) In Figure 4b, treatment of HPAH PASMCs stimulated with 0.1% serum with DH1 alone seems to have had a significant suppressive effect on NOTCH1 protein expression. This does not appear to be due to a loading issue as the authors state that the gels were re-probed for α-tubulin which appears to be uniformly expressed in all lanes on the gel. As this gel is representative of the other replicates, is the effect of DH1 on NOTCH1 expression also observed in the other replicates? 7. The authors have utilised the γ-secretase inhibitor, DAPT in experiments to support the notion that NOTCH ICD generation is involved in the TNFαand BMP6-induced proliferative responses of PASMCs. The concentration of DAPT used in this study was 5mM. Are the authors confident that other proteases important in their proposed mechanism are unaffected by this concentration of DAPT? How does 5mM DAPT affect TNFα-mediated generation of BMPR-II ICD? 8. In Figure 4e, the authors state that expression of NOTCH2 in the medial layer of the pulmonary arteriolar lesion depicted in the lung section taken from an HPAH patient is enhanced compared to control vessels. The intensity of the staining looks somewhat similar between the normal vessel and HPAH vessel in the sections depicted. 9. In Supplementary Figure 17b, Etanercept did not change the % wall thickness for all vessels yet in Figure 6c there appears to be a modest effect of Etanercept on the % of muscularized vessels with Etanercept causing a small increase in non-muscularized vessels compared to the S/H control. Are these endpoints essentially measuring the same thing i.e. pulmonary vascular wall remodelling and, if so, is the data inconsistent? 10. In the body of the text, the authors state repeatedly that the transient induction of Smad1/5 responses by BMP6 / ACTR-IIA / ALK2 signalling is unlikely to promote heightened PASMC proliferation. There are no data shown in the manuscript or supplementary materials that indicate whether the Smad1/5 responses mediated by BMP6 / ACTR-IIA / ALK2 are transient or otherwise. If the authors cannot qualify this statement with data or a citation, then the authors should desist from describing the responses as transient.
Thank you for the thorough and thoughtful reviews of our manuscript. The reviewers were consistent in their recognition of the high quality and internal consistency of our findings and the novelty in several areas that provide new insights into the pathobiology of PAH and offer tractable targets for therapeutic intervention, including repurposing of anti-TNFα approaches.
Our specific responses to the reviewers' comments are below:

Reviewer #1 (expert in PAH)
Remarks to the Author: Comment 1: This study seeks to address the interaction between TNFα and BMP signalling in the pulmonary vasculature. Using a combination of cell and animal models, the authors report that TNFα promotes ADAM10/17 dependent cleavage of the BMP receptor, BMPR2 (the most commonly mutated in human pulmonary arterial hypertension), in pulmonary vascular smooth muscle cells, leaving the ectodomain to act as a ligand trap. The loss of BMPR2 reduces BMP4 signalling but promotes BMP6 signalling via alternative receptor-mediated pathways, and is associated with alterations in NOTCH signalling and c-SRC. Treatment of a rodent model with an anti-TNFα antibody restored NOTCH signalling and reversed the pulmonary hypertension phenotype. The studies are well described and the data look internally consistent. The authors thus identify 3 mechanisms -increased BMP6 signalling, ligand trapping and NOTCH signalling -to explain an interaction between TNFα and BMP.

Response 1:
We thank the reviewer for recognising the high quality of these data and the new mechanisms revealed by our studies.
Comment 2: Increased BMP6 signalling in the presence of reduced BMPR2 expression has been reported before. That TNFα augments BMP6 expression is interesting but as the authors note, it is difficult to reconcile this finding (which would increase Smad signalling) with a proproliferative state.

Response 2:
The reviewer is correct that increased BMP6 expression, coupled with a reduction in smooth muscle cells BMPR-II, has been reported by our lab and others to lead to an increase in p-Smad1/5 signalling via ActR-IIa. However, our lab and others have consistently shown that the Smad1/5 signalling pathway is anti-proliferative in PASMCs (Yang et al Circ Res 2005, etc). In addition, the increased Smad signalling via ActR-IIa is reported to be transient and therefore unlikely to be responsible for the sustained proliferative response to BMP6 that we have observed. To confirm the transient nature of this Smad response we have included new experiments demonstrating that BMPR-II knockdown leads to increased p-Smad1/5 activity in response to BMP6 (Supplementary Fig. 7b,c)

but that this response is very limited in duration (<4 hours). We have added text referring to this observation on page 7.
In contrast to this transient response, the sustained signalling pathway identified by our studies as responsible for hyper-proliferation of PASMCs is robustly shown to be aberrant Notch signalling downstream of ALK2/ActRIIa. Comment 3. TNFα-induced cleavage of BMPR2, leading to release of the ectodomain for ligand trapping, is new. This shows cell specificity, in that it is only seen in vascular smooth muscle cells, not endothelial cells, and this is related to the cell specific expression of ADAM10/17. But it is seen in aortic as well as pulmonary vascular smooth muscle cells, raising the question of how this might explain why it contributes to pulmonary but not systemic vascular remodelling in vivo.

Response 3.
As the reviewer points out, the smooth muscle cell specific cleavage of BMPR-II we report is novel. Indeed, both pulmonary artery smooth muscle cells and human aortic smooth muscle cells exhibit sBMPR-II shedding in response to TNFα. It is entirely reasonable for the reviewer to propose that TNFα-mediated suppression of BMPR-II in the systemic circulation might also drive vascular remodelling. We did not address this in systemic vascular injury models. However, the models employed in our manuscript are specific to pulmonary arterial hypertension, where there is local production of TNFα by inflammatory cells and, as shown in Figure 1e, in the medial layer of the pulmonary artery. Since pulmonary hypertension is a disease state specific to the lung vasculature, there is no major impact on systemic vessels. Comment 4. Where they depart from the published literature is their observations on notch signalling. They observe that TNFα increases NOTCH1 and NOTCH2 but reduces NOTCH3 levels, effects enhanced by BMP6. They argue that NOTCH 1 and 2 are proproliferative and NOTCH3 is antiproliferative. This is at odds with the published literature (see reference list below) where levels of NOTCH3 protein are regarded as a sensitive molecular marker of severity of PAH in humans and pulmonary hypertension in rodents. Notch signaling is involved in vascular development and lung tissue from patients with pulmonary hypertension has been reported to show increased NOTCH3 and NOTCH3 intracellular domain expression when compared with normotensive patients. Additionally, NOTCH3 and NOTCH3 intracellular domain expression has been shown to be increased in two animal models of pulmonary hypertension-hypoxia-induced pulmonary hypertension in mice and monocrotoline-induced pulmonary hypertension in rats. The authors reference the difference and suggest that their γ-secretase inhibitor used to inhibit NOTCH cleavage is more selective than the DAPT agent used in published reports but they do not address the fact that NOTCH3 knockout mice are resistant to pulmonary hypertension.
Response 4. The reviewer highlights that our observations on NOTCH appear to be at odds with some of the published literature in the pulmonary hypertension field. Indeed, when we embarked on these experiments we expected to see the previously reported increase in NOTCH3 and Hes signalling, originally demonstrated in the manuscript by Li et al (Nat Med 2009). Unexpectedly, we consistently observed suppression of NOTCH3 in our model systems and increased expression and signalling via NOTCH2 and HEY1/2, both in vivo and in vitro. We went on to confirm that the NOTCH2 pathway was involved in PASMC proliferation and that NOTCH3 appeared anti-proliferative. In fact, we used the same γ-secretase inhibitor, DAPT, as used in previous reports. However, since DAPT has no selectivity for individual NOTCH pathways we also employed siRNA knockdown of NOTCH2 and HEY1/2 to confirm the proproliferative function of this pathway in PASMCs. Our data are internally consistent in multiple experiments in vitro and in vivo.
The reviewer provides a list of previous publications that at first sight appear at odds with our findings. Here we critically appraise these manuscripts in order to, where possible, provide the best explanation for apparent differences. The totality of previously published papers is that there are significant inconsistencies and/or deficiencies in many of these manuscripts that make direct comparison with our data difficult. Most failed to assess NOTCH2 levels. Our studies mainly concern PAH in which BMPR-II expression is reduced markedly and TNFα is activated. This is of less relevance to the study of pure hypoxia-induced PAH in which the downregulation of BMPR-II is minimal. Nevertheless, our data both in vitro and in vivo are entirely consistent between models.
The analysis of NOTCH signalling in this manuscript was confined to NOTCH3 and did not assess contributions from other NOTCHs. The authors used an antibody from Santa Cruz for detection of the NOTCH3 ICD (catalog number: sc-7424). We tested this antibody in cell lysates, and despite extensive experience with immunoblotting, we found that this antibody performed very poorly compared with the antibodies we sourced from Cell Signaling Technologies. We confirmed the identity of the bands detected by NOTCH antibodies used in our experiments by siRNA knockdown (Supplementary Fig.  13g). One of the central observations in the Li et al. manuscript, as mentioned by the reviewer, was that NOTCH3 knockout mice are protected from hypoxia-induced pulmonary hypertension. We would strongly contest whether this observation supports a specific role for NOTCH3 in the development of pulmonary hypertension, since Notch3 knockout mice demonstrate grossly abnormal arterial maturation in all vascular beds with altered myogenic responses and structural defects (Domenga et al. Genes Dev 2004). In view of this, it is not surprising that a mouse deficient in Notch3 during development exhibits a deficient response to chronic hypoxic exposure, as well as any other contractile or remodelling stimulus. As mentioned above, the therapeutic benefits of DAPT used in this manuscript could equally well implicate a role for any of the NOTCHs in vascular remodelling. This manuscript attempted to more fully characterise the changes in NOTCH1-4 expression in the rat lung during development of hypoxia-induced pulmonary hypertension. These authors did not use immunoblotting for NOTCHs, which is concerning, but instead relied on changes in mRNA expression over 4 weeks. These authors did not detect any changes in NOTCH2 transcripts, but did demonstrate This manuscript did not examine changes in NOTCH2 expression. The authors did not directly assess the expression of NOTCH2 and NOTCH3 in lung tissue from monocrotaline rats and hypoxic mice. Instead they isolated PASMCs from these animals and grew them in culture for an unspecified time before isolating and measuring NOTCH1 and NOTCH3 mRNA expression. Jagged1 levels were measured in whole lung and were correlated with mean pulmonary arterial pressure. The ascertainment of NOTCH2 and 3 levels in isolated PASMCs are very different to the approach used in our experiments (which used whole lung) and it is not clear why the authors used a different approach since NOTCH levels will be greatly influenced by culture in serum and multiple passages. The main emphasis of this manuscript by Xiao et al. is that Jagged1 is increased in the PAH models and that soluble Jagged1 inhibits PAH. These results are consistent with our findings since Jagged1 is likely the important NOTCH ligand and can signal via NOTCH2 and/or NOTCH3. This Phase I study, and others, have reported a significant incidence of pulmonary hypertension in patients exposed to anti-DLL4 antibody treatments for cancer. The pulmonary hypertension is usually reversible. This side effect is thought to be due to inhibition of NOTCH signalling in endothelial cells, particularly in the tip cells of angiogenic sprouts, which enhances VEGF signalling, so most likely functions through a different mechanism to the smooth muscle responses we examine in this manuscript. The effect of anti-DLL4 also raises concerns regarding therapeutic approaches that broadly inhibit NOTCH signalling such as DAPT. More specific approaches targeting smooth muscle cell proliferation, such as anti-jagged1, may be less likely to adversely promote the development of PAH.
Comment 5. A number of cell lines are used in the study. The eventual focus is on distal pulmonary vascular smooth muscle cells. It is understandable that the investigators focus on one cell type/model, but it does make extrapolation to the whole animal problematic. They have attempted this with mouse models but these do not reflect the human condition. The challenge in extrapolating to humans is evident in the emerging literature where antibodies to DLL4, a NOTCH ligand, in development for the treatment of a range of tumours have been reported to cause pulmonary hypertension in patients.

Response 5. We have included data derived from human pulmonary artery smooth muscle cells and endothelial cells, in addition to cell lines. In addition, we have included relevant and informative animal models to show how this pathway is regulated in vivo. Taken together, the data are entirely internally consistent from human cells to animal models. The emerging literature on a causal role for DLL4 inhibition in PAH is not at odds with our findings since the likely pathway implicated in these anti-DLL4 studies is inhibition of NOTCH signalling in endothelial cells. As pointed out by the reviewer, the anti-TNFα approach, central to our study, leads to inhibition of dysregulated NOTCH signalling in pulmonary artery smooth muscle cells.
Minor comments Comment 1. The last sentence in discussion needs rethinking. It is not clear how anti-TNFα can be used "in the prevention of at-risk cohorts".

Response 1: This comment has been removed from the Discussion and we have clarified our patient target group on page 15.
Comment 2. References 35 and 36 are incomplete.

Remarks to the Author:
This manuscript explores the contribution of a pathway involving TNFα-induced processing of the BMPR2 in the pathogenesis of pulmonary arterial hypertension (PAH). This very thorough study employs a combination of genetic and cell based approaches together with pharmacological inhibitors and siRNA-mediated knockdown of various signaling components to build a quite comprehensive picture of the pathway resulting in pulmonary arterial hypertension. In addition, it provides a compelling explanation for how mutations in the BMPR2 could contribute to the pathogenesis of PAH. Overall, the data are of high quality, yet there are some concerns regarding the interpretation of the results related to Notch signaling and some additional experimental questions regarding the shedding of the BMPR2 that remain to be addressed.

As to specifics:
Comment 1: In the Western blots for ADAM10 and ADAM17 in supplementary figure 5c, it is not clear whether the pro-or mature form of ADAM10 and ADAM17 are shown (please also include molecular weight markers). Cell lysates for Western blots of ADAM17 should be prepared in the presence of a metalloprotease inhibitor such as TAPI-1 and 10 mM 1,10 phenanthroline to block the autodegradation of the mature form, which occurs rapidly following cell lysis in the absence of these inhibitors, and can thus affect the interpretation of the levels of mature ADAM17. It is also important to point out that increased expression of ADAM17 does not necessarily result in increased activity, since this is a post-translationally regulated metalloproteinase (see, for example, PMID:23342154). Nevertheless, the overall interpretation that both ADAMs contribute to shedding of the BMPR2 under the conditions used in this study is well supported by the results.

Response 1: We are thankful for the reviewer's insightful comments regarding the autodegradation of ADAM10 and ADAM17 and the requirement for the addition of inhibitors to block this. The study detailing this autodegradation (Schlöndorff J, Becherer JD, Blobel CP. 2000. Intracellular maturation and localization of the tumour necrosis factor alpha convertase (TACE). Biochem. J. 347(Part 1):131-138) examined the process in lysis buffers containing EDTA and Triton.
We homogenised our frozen tissue with a more aggressive lysis buffer containing 2% SDS, which we might expect to denature enzyme activity and reduce the potential for autodegradation. We acknowledge that this may not be entirely sufficient, but we did not observe any 60kDa band observed representing the autodegraded form.
For confirmation of the siRNA knockdown in vascular cells in Fig. 1h, we observe definitive losses of ADAM proteins only with their specific siRNAs. The use of appropriate inhibitors is an important detail, but as the reviewer points out, the overall interpretation that both ADAMs contribute to BMPR2 shedding is well supported by the experimental data provided. We have included a sentence on page 6 of the results section stating that the ADAM levels alone may not be reflective of altered activity. As requested by the reviewer we have now included molecular weight markers for ADAM10 and 17 on the immunoblots (Fig. 1h, Supplementary Fig. 5c). Figure 5D presumably needs to be revised so that siADAM17 and D1(A12) are shown next to ADAM17. Supplementary Fig. 5d. We apologise for this oversight and have corrected the figure.

Response 2: The reviewer is correct in their observation that siADAM17 and D1(A12) should be shown next to ADAM17 in
Comment 3: Regarding the interpretation of data related to Notch signaling, it is important to point out that increased Notch expression does not necessarily lead to increased Notch signaling. Notch signaling is regulated through binding of a membrane-anchored ligand such as Jagged1 or 2, or Dll1, 3 or 4, on a signal-sending cell to a Notch receptor on a signal-receiving cell. The resulting endocytosis of Notch and its ligand is thought to pull open the membrane-proximal negative regulatory domain, allowing access of ADAM10 to the Notch cleavage site. So increased expression of Notch or of ADAM10 will not lead to increased signaling in the absence of this ligand-induced activation of Notch. The two references implicating ADAM17 in Notch signaling are not representative of the majority of papers in this field, in which targeted deletion of ADAM10 in mice typically results in Notch-dependent defects, whereas targeted deletion of ADAM17 does not. Moreover, the Western blots of Notch2 and Notch3 do not show the Notch intracellular domain, but instead show the typical S1-cleaved Notch fragments of around 100 -120 kD (please include molecular weight markers on all gels). This fragment is constitutively generated by cleavage through furin (or related pro-protein convertases), and is not indicative of signaling (PMID:19379690 provides one of several excellent reviews on this topic). The S2 fragment, generated by ADAM10 under physiological conditions, and the S3 fragment, generated by gamma-secretase, are indicative of ligand-induced Notch processing and signaling, but these are most likely not shown on the various Notch blots presented here. So the authors must make it very clear that these blots only allow conclusions to be drawn about Notch expression levels, not about Notch signaling (the same is true for histochemical data). Fig. 4a, Fig. 4b, Supplementary Fig. 11c, Supplementary  Fig. 13g, Supplementary Fig. 14f and the newly added Supplementary Fig. 11a, 14d and 14e. In summary, NOTCH1 was just over 120kDa, NOTCH2 was 100kDa and NOTCH3 was a doublet between 90-95kDa. We do not claim any specific selectivity of ADAM10 or ADAM17 for the cleavage but, as the reviewer suggests, we have now included more representative citations implicating ADAM10 in this process. We now clarify on page 9 that Notch expression levels do not reflect signalling, but that the data regarding HES/HEY responses indicate signalling.

Comment 4:
Results obtained with DAPT, on the other hand, are indicative of a block of Notch signaling, and can be interpreted in this manner. However, the authors should also discuss that Notch2 and Notch3 might not be active in the same cell types, and in general, offer a more careful interpretation of their data related to Notch expression levels. It is very important to distinguish between differences in Notch expression and Notch signaling, which do not have to be directly related, and to consider different effects of Notch signaling depending on the cell types where it occurs.

Response 4:
We have gone through the original description of our findings and made sure that we give due consideration to these points. We have specified that we are examining the NOTCH cleaved/transmembrane intracellular (NTM) regions in the results section (page 9) and have only referred specifically to NOTCH signalling in the context of the HEY/HES transcriptional responses and the impact of HEY1 and HEY2 siRNAs on PASMC proliferation. In particular we now include consideration in the Discussion (pages15/16) that the Notch signalling needs to be considered in the cellular context.

Reviewer #3 (expert in PAH and BMP receptors).
Remarks to the Author: General Comments: Comment 1: The authors have extensively studied the interaction between BMP and Notch signaling that is dysregulated in response to a specific inflammatory stimulus, TNF-alpha. The cell biology studies have been carefully carried out and the main conclusions indicated in the summary are certainly substantiated and are novel. There are however, other studies in the literature that clearly indicate important points of interaction between the BMPR2 pathway and inflammation not cited, including TNF alpha beginning with Song et al (Circulation, 2005:112;552-62)  ; 598-608) and many others. Also there are papers on Notch and BMP crosstalk in blood vessels (Rostama et al, ATVB 2015: 35;2626-37). The authors also do not reconcile their results with their recent studies related to rescue of PAH by the ligand BMP9, in fact surprisingly, work with BMP9 as a ligand seems to be completely excluded. The SUGEN/hypoxia model treated with the TNF soluble receptor reverses or retards progression of the pressure and right ventricular hypertrophy in concert with elevating Notch signaling but the impact on remodeling is less severe and the role of TNF alpha (despite its elevation) vis a vis other cytokines is not clear.

Response 1:
We thank the reviewer for their very positive comments regarding our manuscript including the novelty. We recognise the contribution of others in the field who have shown that inflammation is an important aspect of PAH and BMPR2 signalling, although none have dissected the mechanistic pathways to the extent presented in our manuscript. We thank the reviewer for the suggestion of further references that could be included. We have included the reference relating to the promotion of PAH in Bmpr2+/-mice (Song et al. 2005) in the introduction (page 3 and Reference 10). The references by Hagen et al (IL6 and PAH in mice) and Lawrie et al (osteoprotegerin /TNFRSF11B increased in PAH) are not directly related to the theme of this manuscript. We have not included the reference by Rostama et al (DLL4 and BMP9 relating to EC survival and Thromobospondin 1) as it is not directly related to the theme, but have included a paragraph in the discussion (pages15/16) to include a reference to endothelial cells.
We have undertaken extensive studies of BMP9 signalling in endothelial cells and the impact of TNFα on this response. As expected, TNFα inhibits BMP9 signalling in a BMPR2-specific manner. However, these studies are central to another manuscript that is in preparation that specifically characterises the mechanisms of BMPR2 signalling by BMP9 in ECs. Given the length and complexity of the current manuscript it would be difficult to combine these data. We have now included a sentence that such studies are warranted and underway.

Introduction:
Comment 2: Page 3 last paragraph, see above comments. The impact of TNFalpha and loss of BMPR2 was studied in Sawada et al cited above, although the paper did not show that TNF specifically reduces BMPR2 levels in endothelial cells.

Response 2: The reviewer is correct that the impact of TNFα in the setting of reduced BMPR2 levels has been studied in the manuscript by Sawada et al, 2014. Indeed it is interesting that these authors identified other mechanisms by which TNFα exerts exaggerated effects when BMPR2 levels are reduced (in that manuscript via increased GM-CSF). We have now cited this manuscript in the revised Discussion (pages 15/16 and reference 53).
Comment 3: Page 5 first paragraph: The mechanism by which TNF reduces BMPR2 in endothelial cells is never addressed. Supplementary  Figure 1e. We have shown that the TNFα-dependent repression of BMPR-II is mediated by p65(RelA) and refer to this on page 4. Comment 4: Page 6, second paragraph. It is hard to visualize why two disintegrins are necessary to cleave the same site. This needs further explanation.

Response 4: Our knockdown results clearly show that each of these ADAMs can compensate for the loss of the other. Thus only dual knockdown or inhibition inhibits TNFα-mediated cleavage of BMPR-II
but we are unsure of the reason for the redundancy. It is possible that ADAM10 and ADAM17 may be required to be incorporated into a complex that is necessary for BMPR-II cleavage, to ensure that this event can only take place in cells where both enzymes are expressed. This has now been clarified in the first paragraph of the discussion (page 13).
Comment 5: Page 6, third paragraph. In view of the authors' recent work showing a protective effect of BMP9, it would be interesting to determine whether BMP9 signaling is compromised by TNF alpha. (Also true for data presented on Page 7).
Response 5: As mentioned above, we have now undertaken extensive studies of BMP9 signalling in endothelial cells and the impact of TNFα on this response. As expected, TNFα inhibits BMP9 signalling in a BMPR2-specific manner. However, these studies are central to another manuscript that is in preparation that specifically characterises the mechanisms of BMPR2 signalling by BMP9 in ECs. Given the length and complexity of the current manuscript it would be difficult to combine these data. We have now included a sentence that such studies are warranted and underway (page 16).
Comment 6: Page 10: The controversy with the Li et al paper cited is still problematic. Notch3 and Notch3 ICD are elevated in PAH PA SMC and in human PAs in tissue sections. The blots in Figure 4 should definitely be quantified because it appears that Notch 3 mRNA is higher in the PAH vs. control PA SMC. It is hard to make much of differences in immunoblots and immunohistochemistry because they rely on the affinity of the relative antibodies. The authors should be sure that the siRNAs they use are specific, ie. Do Notch 3 and Notch 2 reduce with Notch 2 siRNA? Does Notch 3 but not Notch 2 decrease with siRNA for Notch 3. Those data should be included. Finally it does appear from the mRNA data in the supplement that the effect of Smad6 induced by TNFα is different, increasing Notch 2 and reducing Notch 3. Thus the TNFα effect may be selective to Notch 3 particularly in the setting of loss of BMPR2. In the mice, knockout of Notch 3 is protective, so there is more to try to reconcile either in the text or in the discussion. It should also be noted that the PASMC come from familial PAH with a BMPR2 mutation so here again the populations are different.

Response 6:
The analysis of NOTCH signalling in the manuscript by Li et al was confined to NOTCH3 and did not assess contributions from other NOTCHs. The authors used an antibody from Santa Cruz for detection of the NOTCH3 ICD (catalog number: sc-7424). We tested this antibody in cell lysates, and despite extensive experience with immunoblotting, we found that this antibody performed very poorly compared with the antibodies we sourced from Cell Signaling Technologies. We confirmed the identity of the bands detected by NOTCH antibodies used in our experiments by siRNA knockdown. These data can be found in Supplementary Fig. 13g.
One of the central observations in the Li et al. manuscript, as mentioned by the reviewer, was that NOTCH3 knockout mice are protected from hypoxia-induced pulmonary hypertension. We would strongly contest whether this observation supports a specific role for NOTCH3 in the development of pulmonary hypertension, since Notch3 knockout mice demonstrate grossly abnormal arterial maturation in all vascular beds with altered myogenic responses and structural defects (Domenga et al. Genes Dev 2004). In view of this, it is not surprising that a mouse deficient in Notch3 during development exhibits a deficient response to chronic hypoxic exposure, as well as any other contractile or remodelling stimulus. In addition, the therapeutic benefits of DAPT used in this manuscript could equally well implicate a role for any of the NOTCHs in vascular remodelling. We have reviewed the evidence in the PAH/Notch literature extensively in response to Reviewer 1 (see above) and do not think that our findings are at odds with the general conclusion that cell specific inhibition of Notch may have protective or deleterious effects in PAH depending on the ligand and on the Notch signalling pathway activated. Furthermore, we have highlighted the fact that our observations are related to the disease-relevant scenario of reduced BMPR-II, where non-BMP pathways may be perturbed compared to a normal BMP status. This is particularly pertinent when considering pathways such as NOTCH and src, both of which interact directly with BMP signalling. We stress the importance of BMPR-II deficiency throughout the revised Discussion.
As the immunoblots in Figure 4 would not have been exposed to equal extents, we felt that comparing these blots for the expression levels of NOTCH proteins was not appropriate. To address this question by the reviewer, we conducted Western blots for the NOTCH proteins using lysates from 3 different control and HPAH PASMC lines all serum-restricted and lysed in parallel. The Western blots and densitometry graph have been included as Supplementary Fig. 11a and 11b, respectively. We have also added these data into the Results section on page 9. We did not observe any significant differences between the protein levels of NOTCH1, NOTCH2 and NOTCH3 in control and HPAH PASMCs. This is consistent with the BMPR2 siRNA data already presented in Supplementary Fig. 14 showing that BMPR2 siRNA does not alter basal expression of NOTCH1, NOTCH2 or NOTCH3.
Comment 7: Page 11: The selective effects of the FYN and YES siRNAs should be documented in view of the results.

Response 7:
We thank the author for this suggestion. We confirmed in our original experiments that the siRNAs for SRC, FYN and YES only reduced the mRNA expression of their dedicated targets without affecting the expression of the other two genes. We have incorporated these data as Supplementary Fig. 16d.
Comment 8: Interestingly the TNF overexpression does not result in more severe RVSP or RVH when BMPR2 is heterozygous. On the other hand, in the SUGEN/Hypoxia model there is clear suppression of RVSP and RVH by the TNF soluble receptor but less impact on the muscularization of vessels. The authors should discuss these differences.

Response 8:
The reviewer is correct that there is no statistically significant difference in RVSP or RVH between SPC/Tnf +/-:Bmpr2 +/+ and SP-C/Tnf +/-:Bmpr2 +/-, although there is between Bmpr2 +/+ and SP-C/Tnf +/-:Bmpr2 +/animals. We have changed the text in the results section on page 8 to better reflect these results. This may be because the high levels of TNFα produced by the SP-C/TNF +/background is causing sufficient reduction in BMPR2 levels to cause the changes observed in RVSP and RVH. However, there is a significant difference in RVSP between Bmpr2 +/and SP-C/Tnf +/-:Bmpr2 +/-, demonstrating that TNFα has a much more severe affect on RVSP in animals with a BMPR2 heterozygous background.
The reviewer is correct in that etanercept appeared to have a greater effect on RVSP and RVH than on peripheral lung vascular muscularization. However, small measured changes in lung precapillary vascular muscularization in this severe reversal model can be expected to have profound effects on haemodynamics based on the fact that vascular resistance is inversely proportional to the radius to the fourth power (r4).
Comment 9: Discussion, Page 13: The discussion with respect to Notch seems to short-change the previous observations, particularly related to the prevention of pulmonary hypertension in the Notch 3 knockout mice. There is no question, however, that the mechanistic studies in this paper are comprehensive and well thought through.

Response 9:
We do not intend to short change the previous observation with respect to Notch and indeed, having further reviewed the literature, provide some further discussion of the field in our revised Discussion. Importantly our studies have been performed in models where BMPR-II expression is critically reduced.

Reviewer #4 (expert in PAH and BMP receptors).
Remarks to the Author: The manuscript describes an interesting and complex system involving a switch in TGFβ receptor super-family usage in PASMCs from an ALK3 / BMPR-II-driven system to an ACTR-IIA / ALK2 systemdriven primarily by TNFα degradation of BMPR-II and the induction in expression of various genes including BMP6 and ACTR-IIA. In the background of compromised BMPR-II expression / function as might be found in PASMCs from patients with heritable forms of Pulmonary Arterial Hypertension (PAH) and genetic mouse models of PAH, this alteration in receptor usage results in a switch from an anti-proliferative response of these cells to a pro-proliferative response following stimulation with BMP6. To add to the complexity, TNFα-mediated modulation in the expression of components of the NOTCH pathway with induction and repression of NOTCH2 and NOTCH3 respectively, directly contributes to the pro-proliferative effects of PASMCs in a Src-kinase family-dependent manner. Although the mechanism described in the manuscript is novel and provides several new therapeutic options for the treatment of heritable PAH, I do have some questions that need to be addressed and suggestions for improvement of the manuscript outlined below before it should be considered for publication.

Comment 1:
The authors present data showing that TNFα induces degradation of BMPR-II via an ADAM10/17-mediated mechanism that appears to be operative in PASMCs but not in PAECs. The expression of BMPR-II is clearly suppressed in PAECs in response to TNFα stimulation, however the authors do not provide any explanation as to the mechanism in this cell type. Do the authors have any mechanistic insights into how BMPR-II protein levels are regulated in PAECs via TNFα stimulation that could be included in the manuscript?

Response 1: We have now investigated the mechanism by which TNFα represses BMPR-II in PAECs,
showing that the TNFα-dependent repression of BMPR-II is mediated by p65(RelA). We have now incorporated these data into Supplementary Fig. 1e.

Comment 2:
The authors have used the SP-C/TNF transgenic mouse model to demonstrate that BMPR-II expression in the whole lung of these animals also appears to be under the control of the TNFα pathway. The data presented by the authors seems to imply that the expression of BMPR-II in whole lung homogenates is almost completely suppressed in this model. This finding seems incongruent with the alveolar epithelial cell-restricted expression of TNFα previously described in this model (Miyazaki et al. 1995J Clin Invest. 199596(1):250-9). Immunohistological analysis of BMPR-II and TNFα expression in the lungs of these animals would provide additional spatial information confirming that loss of BMPR-II expression is co-localised to areas of TNFα expression.

Response 2:
The SP-C/TNF transgenic mouse is a good model to study lung specific effects of TNFα. However, although the expression of TNFα is restricted to alveolar epithelium TNFα is a soluble ligand, which on release results in high levels of TNFα in lavage fluid and elevated serum TNFα levels (Fujita 2001 Am J. Physiol. Cell Mol.Physiol. 280:L39-L49.). These data indicate that the pulmonary vasculature will be directly exposed to TNFα and therefore, immunostaining for the cells expressing TNFα would not provide the relevant information relating to the action of TNFα on target cells.

Comment 3:
The legend for Figure 1e needs to have a better description of what stains / proteins the colours in the images refer to. From the images presented, it also appears that αSMA positive cells appear to express TNFα which might suggest that PASMCs isolated from patients with heritable PAH may express higher baseline levels of TNFα in culture. Have the authors determined whether this is the case?
Response 3: We have edited the figure legend as requested. We have examined RNA samples from control and HPAH PASMC cell lines for the expression of TNFα but did not detect any TNFα mRNA in these cells .

Comment 4:
The authors have cited information showing that the BMPR-II cytoplasmic tail can act as a scaffolding element that can bind to and modulate the activity of a number of kinases including c-Src (Wong et al. Am J Respir Cell Mol Biol. 2005 Nov;33(5):438-46). Can I presume that the generation of the BMPR-II ICD might still be capable of binding to kinases like c-Src and modulating their activity? If so, it would then be important to know how long the BMPR-II ICD remains within the cytoplasm.
Response 4: This is a very interesting question. We identified during our experiments that the BMPR-II ICD is very labile and can only be detected if the cells are snap-frozen prior to lysis to inhibit this rapid degradation. It would be interesting to establish whether accumulation of this fragment alters intracellular signalling pathways, but the labile nature of this product has rendered further investigation very difficult.
Comment 5: Previously published information by the authors has shown that PASMCs isolated from patients with PAH lose their growth suppressive responses to BMP ligands including BMP2 and BMP4 (Morrell et al. Circulation. 2001 Aug 14;104 (7):790-5), which is in contrast to the data depicted in Figure 2D. Could the authors provide an explanation for the differences observed in this study compared to the previously published report?

Response 5: The 2001 Circulation manuscript described a loss of growth inhibition to BMP ligands in
PASMCs from patients with idiopathic and heritable PAH. This loss of growth inhibition was not complete, was dependent on the concentration of BMP, and was more evident in cells from BMPR2 mutation carriers than PAH patients without mutations. Figure 2D also shows that PASMCs from BMPR2 mutation carriers are less susceptible to growth suppression compared with control cells. Thus the data are not inconsistent with our previous work. The experiments conducted in the present manuscript were in 5% serum rather than 0.1% serum used in the 2001 paper.

Response 6a:
We thank the reviewer for identifying these discrepancies. As addressed by reviewer 2, the blots represent the S1 cleaved products of NOTCH1, NOTCH2 and NOTCH 3 rather than the ICDs. We have edited all previous references to the ICD to reflect that we are examining the NOTCH cleaved/transmembrane intracellular (NTM) regions in the Results section (page 9). We have now adjusted the figure legends for Figure 4 and Supplement Figures 11, 13 and 14 to reflect this. b) On page 9 of the manuscript the authors state that in Figure 4a and Supplementary Figure 14c 'BMPR2 silencing also promoted the TNFα-dependent reduction of NOTCH3-ICD generation and NOTCH3 transcription'. On inspection of Figure 4a, TNFα does induce a reduction in NOTCH3 ICD levels; however, BMPR2 siRNA does not enhance the effect of TNFα on NOTCH3 ICD levels compared to the siCP or DH1 control. Similarly, in Supplementary Figure 14c, TNFα does reduce the transcription of NOTCH3 mRNA; however, this effect is not enhanced by BMPR2 siRNA compared to the siCP or DH1 controls.

Response 6b:
The reviewer is correct that the reduction of Notch3 after siBMPR2 knockdown was observed in the co-treatment of TNFα and BMP6 we have adjusted the manuscript to reflect this Page 10). c) Also on page 9 of the manuscript the authors state in Figure 4b 'In BMPR2 heterozygous HPAH PASMCs treated with TNFα, siACVR2A reduced NOTCH2-ICD generation and abrogated NOTCH3-ICD'. On inspection of Figure 4b, treatment of HPAH PASMCs with siACVR2A seems to be without effect on TNFα-mediated NOTCH2-ICD generation.