Oestrogen inhibition reverses pulmonary arterial hypertension and associated metabolic defects.

Increased oestrogen is a strong epidemiological risk factor for development of pulmonary arterial hypertension (PAH) in patients, associated with metabolic defects. In addition, oestrogens drive penetrance in mice carrying mutations in bone morphogenetic protein receptor type II (BMPR2), the cause of most heritable PAH. The goal of the present study was to determine whether inhibition of oestrogens was effective in the treatment of PAH in these mice.The oestrogen inhibitors fulvestrant and anastrozole were used in a prevention and treatment paradigm in BMPR2 mutant mice, and tamoxifen was used for treatment. In addition, BMPR2 mutant mice were crossed onto oestrogen receptor (ESR)1 and ESR2 knockout backgrounds to assess receptor specificity. Haemodynamic and metabolic outcomes were measured.Oestrogen inhibition both prevented and treated PAH in BMPR2 mutant mice. This was associated with reduction in metabolic defects including oxidised lipid formation, insulin resistance and rescue of peroxisome proliferator-activated receptor-γ and CD36. The effect was mediated primarily through ESR2, but partially through ESR1.Our data suggest that trials of oestrogen inhibition in human PAH are warranted, and may improve pulmonary vascular disease through amelioration of metabolic defects. Although fulvestrant and anastrozole were more effective than tamoxifen, tamoxifen may be useful in premenopausal females, because of a reduced risk of induction of menopause.


Introduction
Pulmonary arterial hypertension (PAH) is a disease that includes pulmonary vascular endothelial dysfunction, occlusion and dropout of the small and medium-sized pulmonary arteries and hypertrophy and proliferation of smooth muscle and adventitial cells. These combine to give a progressively worsening elevation of pulmonary vascular resistance [1,2]. This eventually leads to right heart failure and death; no current therapy is curative.
The majority of cases of the heritable form of (H)PAH are associated with mutations in bone morphogenetic protein receptor type II (BMPR2), the type 2 receptor for the BMP pathway [3]. In addition, BMPR2 is suppressed in most other forms of PAH, even in the absence of mutation [4]. Mice with BMPR2 mutation or deletion will spontaneously develop PAH [5][6][7]. However, penetrance in both mice and humans with BMPR2 mutation is incomplete: only 20% of humans with BMPR2 mutation develop clinical PAH [8].
The strongest epidemiologic risk factor for many forms of PAH is female sex [9]. While only 20% of humans with a BMPR2 mutation develop PAH, there is a striking difference according to sex: 43% of females versus 14% of males with a BMPR2 mutation develop PAH in their lifetime [10]. Consistent with this finding, we demonstrated that oestrogen metabolism was a strong predictor of penetrance in HPAH: females who preferentially metabolised oestrogens into 16-oestrogens such as 16αOHE1 developed PAH, whereas females who preferentially metabolised oestrogen into 2-or 4-oestrogens did not [11][12][13]. Oestrogen metabolism drives penetrance in males, although not to the same degree as in females [14].
The mechanism for female preponderance of human disease remains poorly explained, in part because in classical rodent models of PAH such as hypoxia and monocrotaline, oestrogen was found to be protective [15]. This may be associated with a difference between endogenous and exogenous oestrogens; potential differences include location of production (significant oestrogens are made in the pulmonary vasculature) and the cyclic nature of natural oestrogens [16]. Our data show that, as in human patients, BMPR2 mutant mice treated with 16αOHE1 developed PAH with higher penetrance and severity [14], more closely recapitulating the human phenomenon than other models. Our prior work suggested that 16αOHE1 promotes insulin resistance and other metabolic problems [14,17]. Consistent with this finding, metabolic defects have been increasingly associated with PAH [18][19][20], and we have demonstrated that they exacerbate PAH in BMPR2 mutant mice [21,22]. Although the mechanism linking metabolic defects to PAH is not clear, it may be associated with vascular dysfunction, proliferation or production of damaging superoxides [23].
Because in classical PAH models oestrogen inhibition is harmful, the current study sought to determine whether inhibition of endogenous oestrogens was therapeutically effective at preventing or reversing established BMPR2-related PAH, and whether this was associated with improved metabolic metrics such as insulin resistance and oxidised lipids. Safety of anastrazole in postmenopausal PAH patients, with suggestion of efficacy, has already been established in a small trial [24]. Demonstration of efficacy and mechanism is thus a final step necessary in preclinical models to validate the translation of this chain of research to patients.

Oestrogen inhibition experiments
We used the Rosa26-rtTA2 × TetO 7 -Bmpr2 R899X FVB/N mice, as previously described [25,26], called Rosa26-Bmpr2 R899X or Bmpr2 R899X for brevity. R899X is an arginine-to-termination mutation at amino acid 899 in the BMPRII tail domain found in family US33 [6]. Expression of the transgene occurs in all tissue types, but only after initiation of doxycycline.
Adult female Rosa26-only or Rosa26-Bmpr2 delx4+ mice at a starting age of 8-10 weeks had transgene activated with doxycycline at 0.2 mg·g −1 , and received either vehicle (see later) or treatment. No mice in these experiments received exogenous 16αOHE1. Mice were randomised to a treatment group, and the individual performing phenotyping was blinded to group, as were the institutional specialty labs performing, for instance, insulin counts. Fulvestrant (Selleck Chemicals, Houston, TX, USA) was dissolved in ethanol to 100 mg·mL −1 . Anastrozole (Sigma, St Louis, MO, USA) was dissolved in ethanol to 2 mg·mL −1 . Inhibitor solution was diluted in peanut oil and injected subcutaneously into mice daily. Fulvestrant dosage was 200 µg·mouse −1 ·day −1 (∼6 mg·kg −1 ) and anastrozole dosage was 10 µg·mouse −1 ·day −1 (∼0.3 mg·kg −1 ). Medroxyprogesterone 17-acetate (MPA; Sigma) was dissolved in ethanol to 50 mg·mL −1 . MPA solution was diluted in peanut oil and injected intramuscularly into mice at a dosage of 1 mg·mouse −1 every 3 weeks (30 mg·kg −1 per 3 weeks). Tamoxifen (Sigma) was dissolved in ethanol to 100 mg·mL −1 and diluted with peanut oil before injection into mice. Injection dosage was 1 mg·mouse −1 ·day −1 (∼30 mg·kg −1 ·day −1 ).
Drug dosages used were converted from human dosages according to the method of REAGAN-SHAW et al. [27], using proportional body surface areas; this results in increasing dosages from human patient to mouse per unit weight by a factor of ∼12. Doses used are well matched to those used in recent human and animal studies; for a 50 kg human female, our dose would be equivalent to 1.2 mg·day −1 , similar to the 1.0 mg·day −1 used in the recent pilot study in human PAH patients [24].
For prevention experiments, animals started received the drug at time of doxycycline induction. For reversal experiments, animals received doxycycline only for 4 weeks, followed by an additional 2 weeks of both doxycycline and treatment (or vehicle).
After 6 weeks, animals underwent echocardiography, followed on the next day by haemodynamic phenotyping and tissue collection.
Echocardiography and haemodynamic phenotyping Two-dimensional echocardiography was performed using Vivo 770 High-Resolution Image System (VisualSonics, Toronto, Canada). Echocardiograms including B-mode, M-mode and spectral Doppler images were obtained the day prior to sacrifice under isoflurane anaesthetic. Velocity time integral and heart rate were measured in the ascending aorta, and diameter measured in the same location. Stroke volume (SV) was derived using the formula SV=((π 2/4)×(aortic velocity time integral), where π=aortic diameter. Cardiac output (CO) was derived using the formula CO=SV×heart rate [28,29].
Haemodynamic phenotyping was performed as previously described [6,25]. Briefly, mice are anaesthetised with tribromoethanol, systemic pressure checked using a tail cuff and then closed-chested intrajugular right cardiac catheterisation is performed.

Plasma oestradiol measurements
Four female mice from each treatment were randomly chosen for oestradiol measurements. 25 µL of plasma was dispensed into oestradiol ELISA plates in triplicate (ES180S; Calbiotech, El Cajon, CA, USA). Oestradiol ELISA was performed according to manufacturer's instructions; the absorbance at 450 nm was read using a microplate reader (model 680; Bio-Rad, Hercules, CA, USA). The concentration of plasma oestradiol level was calculated and presented as pg·mL −1 from the standard curve.

ESR1/ESR2 × Bmpr2 R899X experiments
Oestrogen receptor (ESR)1 or ESR2 knockout mice were purchased from the Jackson Laboratory (Sacramento, CA, USA). They were firstly crossed with Rosa26 mice with C57BL/6 background. Then they were crossed with Rosa-R899X mice with FVB/N background. Controls (Rosa26-Bmpr2R899X only) were phenotyped on the same days as the ESR1 and ESR2 knockouts, and are a different group of animals than those used in any other experiment (in no experiment were control values "re-used"). Otherwise, experiments were conducted exactly as for oestrogen inhibition experiments.

Vessel muscularisation
Paraffin-embedded mouse lung sections were prepared as described previously. Slides were incubated with mouse monoclonal α-smooth muscle actin (SMA) antibody (1:200; Dako, Carpinteria, CA, USA) overnight at 4°C. After three washes with PBS, the slides were treated with Alexa 488 fluorescent secondary antibody (1:1000; Invitrogen, Carlsbad, CA, USA) for 1 h at room temperature. After three washes with PBS, the slides were air dried and mounting medium was added (Vector Laboratories, Burlingame, CA, USA) and covered with cover glass. SMA-positive vessels were counted under 10× field using a Nikon Eclipse 90i upright fluorescent microscope (Belmont, CA, USA) for 10 randomly chosen fields per mouse.

Detection of mitochondrial superoxide
Mitochondrial superoxide was detected using MitoSox Red (Molecular Probes, Eugene, OR, USA). Mouse pulmonary microvascular endothelial cells (PMVECs) (wild-type orBMPR2R899X) and transgene induction were conducted as previously reported. The cells were grown to 50-80% confluence on chamber slides. MitoSOX was added to a final concentration of 0.5 µM, Hoechst stain was added to a final concentration of 0.1 µg·mL −1 . Cells were placed in an incubator for 20 min at 37°C and then washed three times with Hanks' balanced salt solution containing calcium and magnesium. Mounting medium (Vector Laboratories) was added to the slides and covered with coverslips. The digital images were taken using a Nikon Eclipse 90i upright fluorescent microscope with an 100× oil immersion objective lens.

Measurement of mitochondrial oxygen uptake
Wild-type and BMPR2 R899X murine PMVECs were harvested and cultured from the Immortomouse background, as previously described [30]. 72 h prior to experiments, cells were transitioned to 37°C, and doxycycline (300 ng·mL −1 ) was added to induce transgene expression. For the 24 h prior to analysis, cells were treated with 1 µM 16αOHE1 or vehicle. Cells were washed in basal assay medium (Seahorse Biosciences, North Billerica, MA, USA). 1 h prior to analysis, cells were equilibrated in a carbon dioxide-free incubator in basal assay medium supplemented with glucose (1 g·L −1 ), L-glutamine (2 mM) and sodium pyruvate (1 mM) at pH 7.4. Cells were then analysed in the Seahorse XFe96 Extracellular Flux Analyzer using the mito stress test protocol with the following reagent concentrations (optimised for these analyses in separate experiments): oligomycin (2.5 µM final well), FCCP (0.5 µm final well), FCCP (0.5 µm final well, total 1 µM) and rotenone and antimycin A (0.5 µM final well) at the manufacturer's recommended concentrations. Data are reported as the average of five to eight wells per individual condition in each experiment, with experiments performed in duplicate with 70 K PMVECs per well.

Measurement of Glut4 mobilisation in response to insulin
Glut4-GFP expression vector was purchased from OriGene (Rockville, MD, USA). Cells were transfected with Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's instructions. Transfection solutions were replaced with fresh medium 4 h after transfection, and cells were treated with 1 µM 16αOHE1 or vehicle for 18 h. The cells were then serum-starved for 6 h, stimulated with 10 nM insulin for 1 h, fixed, washed and stained for 4′,6-diamidino-2-phenylindole (DAPI) and DY-554 phalloidin (red). Coverslips were mounted in Vectashield mounting medium (Vector Laboratories) and observed under a microscope.
Immunohistochemical analysis of isoketal protein adducts and ceramides in murine lung Immunolocalisation of isoketal protein adducts and ceramides was performed on murine lung sections. Tissue sections were deparaffinised, rehydrated and treated with 0.3% hydrogen peroxide as described previously. For isoketal staining, sections were incubated in 0.1 M PBS ( pH 7.4) containing 5% normal mouse serum and 5% bovine serum albumin (BSA) for 30 min at room temperature to block nonspecific binding of the secondary antibody. Sections were incubated with 5 µg·mL −1 D11 ScFv (isoketal antibody) for 2 h at room temperature and then incubated with horseradish peroxidase (HRP)-labelled anti-E tag (1:500 dilution; GE Healthcare, Pittsburgh, PA, USA) for 2 h at room temperature. For ceramide staining, antigen retrieval was performed in 10 mM citrate buffer. The sections were blocked with 5% BSA, followed by incubation with ceramide antibody (Enzo Life Sciences, Farmingdale, NY, USA) overnight at 4°C. The following day, the sections were incubated with biotinylated secondary antibody followed by incubation with HRP-conjugated streptavidin. Diaminobenzidine was used as a substrate for HRP (Vector Laboratories). The sections were dehydrated and mounted in Cytoseal XYL (Richard-Allan Scientific, MI, USA) for light microscopic examination.

Statistical methods
Statistical tests were performed using the JMP programme (SAS, Cary, NC, USA). One-way or multiple ANOVA was used to determine effects of interacting variables, with post hoc Fisher's least significant difference (LSD) used to determine differences between individual groups. Significance of correlations were established using correlation z-test.

Animal use approval
The institutional animal care and use committee at Vanderbilt University (Nashville, TN, USA) approved all animal studies.

Oestrogen inhibition prevents pulmonary hypertension in BMPR2 mutant mice
Having previously shown that 16αOHE1 amplifies PAH, we first sought to determine whether oestrogen inhibition, or androgen exposure were protective, and whether they worked synergistically. For oestrogen inhibition, we used the combination of anastrozole and fulvestrant, as commonly used for oestrogen suppression in human breast cancer. We used them in combination only, rather than independently, because the goal of this initial study was to determine without the strongest possible pharmacological signal if inhibition of oestrogen signalling was protective in this model. Control (Rosa26 only) or Rosa26-Bmpr2 R899X mice were fed doxycycline at 0.2 g·kg −1 for 6 weeks (a time in which Bmpr2 R899X mice will typically develop elevated right ventricular systolic pressure (RVSP) [25]), and then each genotype was split into four groups: plus or minus the androgen MPA (1 mg·mouse −1 every 3 weeks) and plus or minus a combination of the aromatase inhibitor anastrozole (10 µg·mouse −1 ·day −1 ) and the oestrogen receptor inhibitor fulvestrant (200 µg·mouse −1 ·day −1 ) (anastrozole/fulvestrant). By multiple-factor ANOVA, use of MPA had no significant effect on any variable measured, and so for the remainder of the study these results were grouped with the non-MPA animals. Furthermore, the results are not changed if MPA-exposed animals are excluded.
Use of anastrozole/fulvestrant was effective in suppressing oestrogen signalling, as demonstrated by a strong reduction in uterus weight (figure 1a). As expected, Bmpr2 mutant mice treated with vehicle had elevated RVSP, averaging 35 mmHg, and ∼50% penetrance (figure 1b). RVSP is variable in Bmpr2 R899X mice because pulmonary vascular resistance is driven by loss and narrowing of vessels, rather than constriction [25]. Use of anastrozole/fulvestrant prevented most of the increase in RVSP in Bmpr2 R899X mice, with RVSP averaging 27 mmHg in the treatment group, and a penetrance of ∼10% ( p<0.0001 compared to vehicle). There was no effect on cardiac output in any group (figure 1c). Fulton index (online supplementary figure S1) is not informative in this model; we have previously published that the right ventricle in Bmpr2 mutant mice (and patients) has an impaired hypertrophic response [19].
Oestrogen inhibition with anastrozole/fulvestrant nearly normalised muscularised vessel counts (figure 1d). Because we have previously shown increased oxidative stress as a marker of the metabolic defects in Bmpr2 mutant mice, we performed immunohistochemistry for oxidised lipids, and found increased oxidised lipid in the vessel wall in Bmpr2 mutant mice; however, this was normalised by oestrogen inhibition (red arrows in figure 1e). Oestrogen inhibition does not appear to reduce oxidised lipid in infiltrating inflammatory cells (figure 1e). We have previously shown an increase in accumulation of the toxic lipid ceramide with Bmpr2 mutation [31] as part of metabolic dysfunction. Ceramide was once again increased in the vascular walls of Bmpr2 mutant mice, and this was reduced by anastrozole/fulvestrant (figure 1f).

Oestrogen inhibition reverses established PAH in BMPR2 mutant mice
Because reversal of established disease is particularly relevant to the PAH patient population, we allowed Bmpr2 mutant mice to develop PAH for 6 weeks, and then treated them with either anastrozole/ fulvestrant or with the somewhat weaker oestrogen receptor antagonist tamoxifen for an additional 4 weeks. Tamoxifen was added because it can be used in premenopausal females safely without inducing menopause.
As expected, tamoxifen did not impact uterine weights, but anastrozole/fulvestrant reduced uterus weights (figure 2a). Use of anastrozole/fulvestrant reduced plasma oestrogen levels ∼3.5-fold (online supplementary figure S2a). Although, as expected, tamoxifen did not impact circulating oestrogen, it did reduce expression of known oestrogen targets FHL1 and SNAT2 [32] in lung by 50% and 35%, respectively (online supplementary figure S2b).
Vehicle-treated Bmpr2 R899X mice had penetrance and pressures comparable to previous experiments, which were largely reversed by treatment with anastrozole/fulvestrant, but with less effect by tamoxifen ( figure 2b). Counts of muscularised small vessels showed a strong increase in Bmpr2 R899X mice, with modest reduction with tamoxifen, and normalisation with anastrozole/fulvestrant (figure 2c). Ceramide was increased in the vascular walls of Bmpr2 mutant mice, and this was reduced by anastrozole/ fulvestrant, but was not apparent in tamoxifen (figure 2d).

Knockout of either ESR1 or ESR2 partially protects BMPR2 mutant mice
Because pharmacological methods to inhibit oestrogen receptors may lack specificity and vary by cell type, we used genetic knockouts to attempt to determine whether the oestrogen effect was mediated through FIGURE 1 a) Anastrozole and fulvestrant (A+F) reduce uterine weights in mice ( p<0.0001) as expected; by multiple ANOVA, R899X mutation did not have an effect on uterine weights. Each symbol represents the measurement from one mouse. b) A+F delivered in osmotic pumps through the final 4 weeks of 6 weeks' BMPR2 R899X transgene induction prevents development of elevated right ventricular systolic pressure (RVSP) ( p<0.0001 by multiple ANOVA, using transgene and drug as variables). Each symbol represents a value from one mouse. Mice represented by triangles received the androgen medroxyprogesterone acetate (MPA) in addition, which did not impact RVSP, so these data have been not been separated. Mice were age-matched and controls were tested contemporaneously for each group. c) Cardiac outputs from echocardiography were not significantly changed by either A+F treatment or Bmpr2 mutation. d) A+F nearly normalised the increase in muscularised vessels normally seen in BMPR2 R899X mice. By multiple ANOVA considering BMPR2 mutation and treatment as factors, BMPR2 mutation increases muscularised vessels ( p=0.0004), while A+F decrease them, specifically in BMPR2 mutant mice ( p=0.003). e) Staining for oxidised lipids demonstrates increased oxidative stress in BMPR2 R899X mice, normalised with A+F (arrows). f) Ceramide (toxic lipid) staining (brown) is increased in lungs from BMPR2 R899X mice and reduced by A+F. NS ESR1 or ESR2. Thus, these mice each had two transgenes and two knockout alleles (Rosa26-rtTA2M2, TetO7-Bmpr2 R899X , and two alleles knocked out of either ESR1 or ESR2). Adult animals were fed doxycycline for 6 weeks to induce Bmpr2 R899X transgene expression.
As expected, ESR1 but not ESR2 knockout resulted in reduction in uterine weight (figure 3a) [33]. Western blots for ESR1, ESR2, and the noncanonical G protein coupled estrogen receptor (GPER) in ESR1 and ESR2 knockout and heterozygous mice show the expected and specific loss of protein (online supplementary figure S3). There appears to be a compensatory increase in GPER, but not the reciprocal ESR, with either ESR1 or ESR2 knockout.
Closed-chest measurement of RVSP in spontaneously breathing mice found 20 (54%) out of 37 Bmpr2 R899X mice with pressures >32 mmHg (figure 3b), while this number was lower in Bmpr2 R899X mice with ESR1 knockout (seven (33%) out of 21), and three (14%) out of 22 in Bmpr2 R899X mice with ESR2 knockout. The trend was similar for muscularisation, with numbers of muscularised small vessels in Bmpr2 R899X mice nearly double that of controls (figure 3c and d), and these numbers reduced somewhat in ESR1 and more in ESR2 knockouts. The occluded vessels that can occasionally be seen in Bmpr2 R899X mouse lungs were still present in ESR1 knockouts, but not apparent in ESR2 knockouts (figure 3d). Taken  p=0.0004 p=0.02 p=0.0004 FIGURE 2 a) Anastrozole and fulvestrant (A+F), but not tamoxifen used as treatment reduces uterine weights in mice ( p=0.002 for comparison to control; p=0.007 for comparison to vehicle). Each symbol represents the measurement from one mouse; data are presented as median±SEM. b) A +F delivered in osmotic pumps through the final 4 weeks of 10 weeks' BMPR2 R899X transgene induction reverses development of elevated right ventricular systolic pressure (RVSP) ( p=0.003 by multiple ANOVA, using transgene and drug as variables). Tamoxifen used the same way did not have a significant effect on RVSP, but RVSP trended downwards. Each symbol represents a value from one mouse. c) A+F used as treatment nearly normalised the increase in muscularised vessels normally seen in BMPR2 R899X mice; tamoxifen had intermediate effect. By ANOVA, BMPR2 mutation increases muscularised vessels ( p=0.0004) while A+F (p=0.0004) and tamoxifen ( p=0.02) both decrease them. d) Ceramide (toxic lipid) staining (brown) is increased in lungs from BMPR2 R899X mice, and reduced by A+F, but not by tamoxifen. Scale bar=25 µm. together, these data suggest that the effects of oestrogen in PAH are mediated partially through both ESR1 and ESR2, with ESR2 predominating.

Oestrogen inhibition reverses BMPR2 mutation-induced insulin resistance
We and others have previously shown that PPAR-γ and CD36 expression are suppressed in BMPR2 mutant mice and human PAH patients and insulin resistance is increased [22,34].
Here we found that these changes were prevented (figure 4a) and reversed (figure 4b) by oestrogen inhibition. Suppression of PPAR-γ and CD36 in the vehicle BMPR2 mutant groups is more impressive in the reversal study, presumably because the animals have had longer to develop disease (10 weeks rather than 6 weeks). Measurement of insulin resistance by measurement of blood glucose and insulin showed, as before, a strong increase in homeostatic model assessment (HOMA) insulin resistance in BMPR2 mutants, substantially reduced by anastrozole/fulvestrant, and trending down with tamoxifen (figure 4c). Plotting HOMA insulin resistance against RVSP shows that the two are strongly correlated (r=0.71) (figure 4d). These data suggest that the effect of oestrogen inhibition on RVSP is closely correlated with its metabolic effects.
In addition, we measured PPAR-γ and CD36 in ESR1 and ESR2 knockout mice with and without BMPR2 mutation. ESR1 knockout did not have a consistent effect (figure 5a), while ESR2 knockout led to increased PPAR-γ and CD36 in both control and BMPR2 mutant mice (figure 5b). This supports the haemodynamic measurements in figure 3, which suggested that ESR2 is more responsible for the negative effects of oestrogen in PAH than ESR1.

Oestrogen drives insulin resistance on a cellular level
Although glucose metabolism is regulated on a whole-organism level by the interaction of multiple organs, on a cellular level, insulin resistance is characterised by reduced mobilisation of the glucose transporter Glut4 in response to insulin. We have previously shown a basic defect at this level in BMPR2 mutant human endothelial cells; specifically, insulin signalling is intact but Glut4 fails to mobilise [35]. Here we replicate that result, and show that oestrogenic 16αOHE1 significantly worsens the insulin response in PMVECs cultured from Bmpr2 R899X mice ( figure 6a and b). 16αOHE1 was used instead of E2, because it is more oestrogenic than E2, and E2 in patients is preferentially metabolised to 16αOHE1 [11].
Comparing maximum oxygen consumption rate (OCR) in Bmpr2 R899X PMVEC compared to wild-type PMVEC using Seahorse extracellular flux analysis, we found that PMVECs from Bmpr2 R899X mice have a significantly lower maximum OCR (figure 6c and d) in standard media. However, the addition of 16αOHE1 increases maximum OCR in Bmpr2 R899X mutants; in contrast, 16αOHE1 lowers maximum OCR in wild-type cells ( figure 6c and d). These data suggest that oestrogens modify the already perturbed cellular metabolic activity in the Bmpr2 mutant setting.
Although we do not know in detail how oestrogen impacts metabolism in these cells, staining live cells with the superoxide detector Mitosox Red suggests that increased oxygen consumption with 16αOHE1 is caused by increased superoxide production (figure 6e).

Discussion
Over the past decade, the PAH research community has made tremendous strides towards developing molecularly targeted therapies. There is broad consensus in the field that in addition to, or perhaps because of, suppressed BMP signalling, progression of PAH requires dedifferentiation and proliferation, inflammation and metabolic changes. Trials targeting each of these have either been completed [36][37][38], are in progress [39] or are in development [40]; at this point the debate is how to effectively intervene, rather than where we need to intervene. The field is thus in a state in which our goal is effectively translating our molecular findings into treatments.
Female sex is the strongest epidemiological risk factor for PAH, with a 3:1 female:male ratio. However, until recently, the reasons for this were unclear: oestrogen and female sex were protective in classical PAH models such as hypoxia and monocrotaline [15]. However, a chain of research has contributed new information regarding sex and PAH. PAH patients have increased level of oestradiol [41], while increased risk of PAH in patients is strongly associated with preferential metabolism of oestrogens into 16-oestrogens such as 16αOHE1 [11,12,42]. There are underlying genetic variations in both aromatases, which produce parent oestrogens [41] and in CYP1B1, which modulates oestrogen metabolism [11,43] associated with outcomes in both portopulmonary and idiopathic pulmonary hypertension. Increase in 16αOHE1 is important because it covalently binds the receptor and is thus much more oestrogenic than 2-and 4-oestrogens.
Treating BMPR2 mutant mice with 16αOHE1 increases penetrance and severity of disease [14], associated with markers of increased metabolic problems [17]. Markers of metabolic derangements, including the metabolic syndrome, are strongly associated with disease in patients [18][19][20] and are causal in BMPR2 mutant mice [21,22]. Metabolic problems may be important on their own, for instance, by facilitating proliferation, but they may also be important because they result in increased production of reactive oxygen and nitrogen species. These phenomena are heavily interlinked, and both have been extensively studied in PAH (reviewed by FESSEL and WEST [44]). Oestrogens, and in particular 16αOHE1, have recently been shown to induce reactive oxygen species production in an NADPH-dependent fashion [45], supporting the link between oestrogens, metabolism and reactive oxygen species.
In the present study, we finalise preparations for translation by addressing key remaining issues. We demonstrate that oestrogen inhibition both prevents and reverses PAH in the context of BMPR2 mutation, β-Act CD36 Treatment FIGURE 4 a) Anastrozole and fulvestrant (A+F) used preventatively result in increased peroxisome proliferator-activated receptor (PPAR)-γ ( p=0.001) and CD36 ( p=0.02) in whole lung, which are otherwise suppressed by R899X mutation (p=0.003 for PPAR-γ and p=0.007 for CD36) by two-way ANOVA. Numbers represent densitometry for the bands above; each lane is whole lung from a different animal. b) A+F used as treatment result in increased PPAR-γ (p=0.019) and CD36 (p=0.027) in whole lung, which are otherwise suppressed by R899X mutation ( p=0.047 for PPAR-γ and p=0.043 for CD36) by two-way ANOVA. Numbers represent densitometry for the bands above; each lane is whole lung from a different animal. c) A+F used as a treatment significantly reduce insulin resistance in R899X mice ( p=0.012), while tamoxifen (Tam) trends towards improvement ( p=0.06). Tests are Fisher's least significant difference after ANOVA ( p=0.01 for rejection of null). Note that insulin resistance even in control mice is quite high, as they are FVB/N mice on a Western diet. d) Insulin resistance is highly correlated with right ventricular systolic pressure (RVSP) (correlation=0.71, p<0.0001). β-Act: β-actin; HOMA-IR: homeostatic model assessment insulin resistance ctl: control; veh: vehicle.  probably through canonical oestrogen receptor signalling. We show that oestrogen inhibition reverses markers of metabolic defects including HOMA-insulin resistance, and that oestrogen impacts insulin resistance on a cellular level. The combination of these data provides a possible mechanism of action, and demonstrates efficacy in mice, paving the way for clinical trials in patients using oestrogen inhibition to impact the metabolic axis.
A key question is whether the oestrogen is modulating metabolism directly, or indirectly through modulation of the BMP pathway. Oestrogen suppresses BMPR2 signalling in human pulmonary artery smooth muscle cells [46], possibly by direct binding of the oestrogen receptor to the BMPR2 promoter [47], which can lead to increased proliferation in cell culture. Since the BMP pathway regulates metabolism, at least some of the effect may be through ESR1 (ERα)-mediated suppression of BMPR2. ESR1 single-nucleotide polymorphisms have previously been associated with pulmonary hypertension risk in portopulmonary hypertension [41]. However, some of the effect may be direct; ESR2 (ERβ) can localise to the mitochondria, where it directly forces a metabolic shift supportive of proliferation [48]. Since we have previously a shown a strong impact of oestrogen on mitochondrial morphology [17], it is likely that at least part of the oestrogen effect on metabolism bypasses BMPR2. Because both ESR1 and ESR2 have plausible and distinct modes of action, the dominant effects is probably exquisitely dependent on the precise conditions and organ under investigation. For instance, in the present study, since BMPR2 is genetically suppressed, the role of ESR1 in suppressing BMPR2 may be less important. The increase in GPER with ESR knockout may be relevant to outcomes as well: GPER has been shown to increase with pharmacological oestrogen blockade [49]. Increased signalling through GPER has potential benefits in PAH [50].
Another interesting point is the question of whether the source of the oestrogens is important. Most previous studies showing a protective effect of oestrogen used male rodents administered exogenous oestrogen; by contrast, anastrozole is protective in female but not male rodents using either hypoxia or Sugen/hypoxia models [16]. Moreover, only female mice develop pulmonary hypertension in the dexfenfluramine model [51], and ovariectomy prevents pulmonary hypertension in mice overexpressing the serotonin transporter [52]. The theory is that the site of oestrogen production is important, and that extragonadal oestrogen production may be of particular importance. In support of this, we have previously shown that mode of oestrogen metabolism correlates with disease penetrance in male PAH patients as well [17]. This suggests that in male and postmenopausal patients, peripheral oestrogen production, for instance in adipose tissue [53], may still be a pathological source of oestrogen driving disease.
These and other scientific questions remain to be addressed. In the long term, oestrogens may have a protective impact upon the function of the right ventricle in humans with PAH, which remains an opportunity for further study [43,[54][55][56]. Although clearly neither oestrogen nor insulin resistance alone is enough to cause PAH, as the vast majority of females and diabetics do not develop PAH, in the context of BMPR2 mutation these are strong drivers of progression to clinical disease. However, we still don't understand the mechanisms in detail. Exposure to oestrogens and exposure to a BMPR2 mutation are associated with fragmented mitochondria [17], increased oxidative stress [57], alterations in protein, sugar and lipid handling [18] and increased insulin resistance [14,22]. Although all of these phenomena are related, we don't as yet have a top-to-bottom mechanism linking these to either oestrogen or BMPR2, and we don't know which of these are pathogenic and which are bystanders. In addition, we don't know the precise contribution, if at all, of noncanonical oestrogen signalling.
While scientifically interesting, it is not mandatory to answer all of these questions before proceeding to translation. Particularly in males and in postmenopausal females, strong oestrogen inhibition is safe, addresses many molecular pathologies causally linked to development of PAH and reverses established disease in mice carrying the same BMPR2 mutation found in many patients. In fact, a short phase II human PAH trial using a form of oestrogen modification was recently completed (clinicaltrials.gov identifier NCT01545336); it was safe, well tolerated, and showed hints of efficacy [24]. One of the central questions in clinical translation is which drug to use. Our data suggest that while stronger inhibition, such as that achieved by anastrozole and fulvestrant, is more effective, and thus probably preferable for postmenopausal patients and males, tamoxifen has some efficacy, and as it is easier to use in premenopausal females [58], it is worth determining whether it can be substituted in these younger patients in order to avoid induction of menopause.
The present study contributes critical background work to further demonstrate the efficacy of oestrogen inhibition and to elucidate mechanisms including improvement in some metabolic irregularities present in PAH. By adding to the groundwork needed to proceed to human trials, we look forward to proceeding with additional testing and early translational studies.