MURC deficiency in smooth muscle attenuates pulmonary hypertension

Emerging evidence suggests that caveolin-1 (Cav1) is associated with pulmonary arterial hypertension. MURC (also called Cavin-4) is a member of the cavin family, which regulates caveolar formation and functions together with caveolins. Here, we show that hypoxia increased Murc mRNA expression in the mouse lung, and that Murc-null mice exhibited attenuation of hypoxia-induced pulmonary hypertension (PH) accompanied by reduced ROCK activity in the lung. Conditional knockout mice lacking Murc in smooth muscle also resist hypoxia-induced PH. MURC regulates the proliferation and migration of pulmonary artery smooth muscle cells (PASMCs) through Rho/ROCK signalling. Cav1 suppresses RhoA activity in PASMCs, which is reversed by MURC. MURC binds to Cav1 and inhibits the association of Cav1 with the active form of Gα13, resulting in the facilitated association of the active form of Gα13 with p115RhoGEF. These results reveal that MURC has a function in the development of PH through modulating Rho/ROCK signalling.

P ulmonary hypertension (PH) is a progressive disease of various origins, which results in right heart dysfunction and is associated with a poor prognosis 1 . Pulmonary arterial hypertension (PAH) is a clinical condition characterized by the presence of pre-capillary PH in the absence of other causes of pre-capillary PH such as PH due to lung diseases, chronic thromboembolism or other rare diseases 2,3 . PAH is subcategorized by its underlying causes, all of which are characterized by excessive pulmonary vasoconstriction and abnormal vascular remodelling processes 1 . Current treatments for PAH involve the use of prostanoids, endothelin receptor blockers and/or phosphodiesterase (PDE)-5 inhibitors 4 . Although these medications are effective in slowing the progression of PAH, they are insufficient to regress vascular remodelling. A clearer understanding of the mechanisms of vascular remodelling may lead to the development of novel therapeutic approaches for the prevention and/or treatment of PAH.
In heritable PAH, mutations in transforming growth factor-b/bone morphogenetic protein (TGF-b/BMP) receptors and activin receptor-like kinase type 1 (ALK-1) have been identified 1,5 . Recent research has shown that mutations in caveolin-1 (Cav1), a major component of caveolae, are also associated with PAH 3,6 . Caveolae are involved in several important cellular processes, including signal transduction, endocytosis and cholesterol homeostasis, and caveolins serve as the essential structural components of caveolae and function as scaffolds for caveolar-mediated signalling pathways [7][8][9] . Cav1 is the predominant caveolin isoform in smooth muscle cells (SMCs), and Cav1 deficiency results in the loss of caveolae in SMCs 10,11 . Cav1-null (Cav1 À / À ) mice develop PH and right ventricular (RV) hypertrophy associated with pulmonary vascular remodelling 12,13 .
Here, we show that MURC is associated with Cav1 and caveolin-3 (Cav3), and that MURC in smooth muscle is implicated in the development of PH. The association of MURC with Cav1 regulates RhoA signalling in VSMCs.

Results
Colocalization of MURC with Cav1 and Cav3. Cav1 and Cav3 are expressed in VSMCs 11,32 . To examine the subcellular localization of MURC, Cav1 and Cav3 in PASMCs, we performed immunostaining using human PASMCs (hPASMCs) with anti-MURC, anti-Cav1 and anti-Cav3 antibodies. MURC was colocalized with Cav1 and Cav3 at the plasma membrane of hPASMCs (Fig. 1a). Immunoprecipitation showed the association of MURC with Cav1 and Cav3 (Fig. 1b,c). The bimolecular fluorescent complementation (BiFC) assay in hPASMCs also revealed the association of MURC with Cav1 and Cav3 in situ, which was mainly detected at the plasma membrane (Fig. 1d).
Attenuation of hypoxia-induced PH in Murc-knockout mice. No animal model completely recapitulates human PAH 4 . However, since chronic normobaric hypoxia is commonly used as a model of PH in mice 33 , we employed this animal model in the present study. We exposed wild-type (WT) mice to normobaric hypoxia and examined the expression level of Murc mRNA in the lung. Murc mRNA expression levels were higher in the lung exposed to hypoxia than in that under normoxia (Fig. 2a). This finding raises the possibility that Murc has a role in the development of PH. Therefore, we examined this hypothesis using Murc-knockout (Murc À / À ) mice 34 . We firstly examined the protein expression of Murc in WT and Murc À / À VSMCs. Most of the Murc protein was detected in the membrane fraction of VSMCs isolated from WT mice, while it was undetectable in VSMCs isolated from Murc À / À mice (Fig. 2b), confirming the deletion of Murc in VSMCs. Cav1 expression was observed in the membrane fraction of both WT and Murc À / À VSMCs. Moreover, Cav1 protein expression in Murc À / À lungs was not significantly different from that in WT lungs ( Supplementary Fig. 1a).
Cav1 deficiency has been shown to cause hypertensive pulmonary phenotypes 12,13 . We compared RV systolic pressure (RVSP) and the ratio of the RV to left ventricle (LV) and septum weights (RV/LV þ S) among WT, Murc À / À and Cav1 À / À mice. Under normoxic conditions, no significant differences were observed in systolic blood pressure, diastolic BP, heart rate, left ventricular systolic function, RVSP, or RV/LV þ S between WT and Murc À / À mice, while Cav1 À / À mice showed significant elevations in RVSP with RV hypertrophy, as assessed by RV/LV þ S (Supplementary Table 1; Fig. 2c,d).
To explore the functional significance of Murc in the development of PH, WT and Murc À / À mice were exposed to normobaric hypoxia for 4 weeks. After hypoxia, WT mice displayed significant elevations in RVSP with RV hypertrophy, whereas Murc À / À mice showed the attenuation of RVSP elevations and RV hypertrophy (Fig. 2e,f). Vascular remodelling, which is characterized by an increase in the wall thickness of pulmonary arterioles, as evaluated by the a-smooth muscle actin (aSMA)-positive area, was observed in WT mice after hypoxia, while it was significantly attenuated in Murc À / À mice (Fig. 2g), suggesting that VSMC proliferation is suppressed in the latter group. In accordance with these findings, PASMC proliferation evaluated by Ki67 staining was suppressed in the pulmonary vessels of Murc À / À mice exposed to hypoxia for 2 weeks (Fig. 2h). Caveolae in PASMCs were retained in Murc À / À mice, and hypoxia did not affect the caveolar morphology in WT or Murc À / À mice, while caveolae disappeared in the lung of Cav1 À / À mice (Fig. 2i).
We previously showed that Murc regulates Rho/ROCK signalling in cardiomyocytes 20 . The Rho/ROCK signalling pathway modulates the 'Ca 2 þ sensitivity' of smooth muscle mainly by suppressing myosin light chain (MLC) phosphatase activity, thereby regulating vascular smooth muscle tone 35 ; therefore, we evaluated the activity of ROCK to examine its downstream targets, myosin phosphatase targeting subunit 1 (MYPT1) of MLC phosphatase and MLC2 (refs 36,37) in the lung of WT and Murc À / À mice. ROCK has been shown to phosphorylate MYPT1 at Thr853 and Thr696, which results in a decrease in MLC phosphatase activity and increase in MLC phosphorylation 35 . ROCK also phosphorylates MLC at Ser19 to increase myosin ATPase activity 38 . Under normoxic conditions, the phosphorylation of MYPT1 at Thr853, which is considered to be a specific target of ROCK 36,39 , did not significantly differ between WT and Murc À / À lungs, while the phosphorylation of MYPT1 at Thr853 was greater in Cav1 À / À lungs than in WT and Murc À / À lungs (Fig. 3a). In addition, the greater phosphorylation of MLC2 at Ser19 in the PASMC of Cav1 À / À mice was observed by immunostaining (Fig. 3b). After hypoxia, MYPT1 phosphorylation at Thr853 was significantly greater in the lung of WT mice than in the lung of Murc À / À mice (Fig. 3c). These findings suggest that hypoxia-induced ROCK activation in the lung is reduced by Murc deficiency, which is supported by the immunostaining analysis showing that the phosphorylation of MLC2 at Ser19 was less in the PASMCs of Murc À / À mice exposed to hypoxia than in those of WT mice exposed to hypoxia (Fig. 3d). Furthermore, a treatment with Y-27632, a selective ROCK inhibitor, inhibited the phosphorylation of MYPT1 and MLC2 in the lung of WT mice exposed to hypoxia, but not in the lung of Murc À / À mice exposed to hypoxia (Fig. 3e,f).
mating SM22Cre transgenic mice with Murc fl/fl mice. In cKO mice, the Murc protein was deleted in VSMCs, but was retained in the skeletal muscle ( Fig. 4a; Supplementary Fig. 1b). The Murc protein was detected in the cKO heart by western blotting, but was significantly decreased ( Supplementary Fig. 1b). Murc mRNA and protein were detected by RT-quantitative PCR and immunostaining, respectively (Supplementary Fig. 1c,d). In addition, we performed hematoxylin and eosin (H&E) staining on the aorta of WT, Murc À / À , Murc fl/fl and cKO mice. H&E staining revealed no significant differences in vascular walls among WT, Murc À / À , Murc fl/fl and cKO mice ( Supplementary Fig. 1e).
After exposure to normobaric hypoxia for 4 weeks, hypoxiainduced RVSP elevations, RV hypertrophy and vascular remodelling were less in cKO mice than in Murc fl/fl mice ( Fig. 4b-d). In addition, we performed morphometric and echocardiographic analyses to assess cardiac function in Murc fl/fl and cKO mice. Under normoxia, no significant differences were noted in morphometric or echocardiographic parameters between Murc fl/fl and cKO mice (Supplementary Table 2). Furthermore, after exposure to normobaric hypoxia for 4 weeks, LV systolic function in cKO mice was not significantly different from that in Murc fl/fl mice (Supplementary Table 3). These results suggest that LV function does not affect pulmonary arterial pressure in cKO mice under hypoxia, and that Murc in PASMCs plays crucial roles in the development of PH and RV remodelling.
MURC regulates p115RhoGEF/Rho/ROCK signalling in PASMCs. Hypoxia and hypoxia-associated stimuli induce the production of various chemokines/cytokines and growth factors in the pulmonary artery 4 . Among them, TGF-b1 mRNA and protein expression is induced by hypoxia in PASMCs and the lung 40,41 , and its signalling has been shown to mediate hypoxia-induced pulmonary arterial remodelling 42 . Since MURC mRNA expression in the lung is induced by hypoxia, we determined whether TGF-b1, IL-1b and endothelin-1 (ET-1) induce MURC mRNA expression in hPASMCs. TGF-b1 induced MURC mRNA expression in hPASMCs, whereas IL-1b and ET-1 did not ( Supplementary Fig. 2), suggesting that TGF-b1 is one of the upstream regulators involved in hypoxia-induced MURC expression in PASMCs.
To elucidate the mechanisms by which MURC modulates pulmonary vascular remodelling, we knocked down MURC expression in hPASMCs using human MURC-specific siRNA hPASMCs was less than that in hPASMCs transfected with control siRNA (Fig. 5a). Wound healing and Boyden chamber assays revealed that migration was also less in MURC-knockdown hPASMCs than in control hPASMCs ( Fig. 5b; Supplementary  Fig. 3b). We then established MURC-overexpressing hPASMCs using a recombinant retrovirus expressing MURC ( Supplementary  Fig. 3c). hPASMCs overexpressing MURC showed increase in proliferation and migration (Fig. 5c,d). In accordance with these findings, FBS-induced proliferation and migration in VSMCs isolated from Murc À / À mice were significantly less than those in WT VSMCs ( Supplementary Fig. 3d,e). In SMCs, the Rho/ROCK pathway has been demonstrated to regulate proliferation and migration 43,44 . We examined whether MURC regulates the proliferation and migration of PASMCs through the Rho/ROCK pathway. RhoA activity in MURCknockdown hPASMCs was significantly weaker than that in control hPASMCs (Fig. 5e). Correspondingly, RhoA activity was weaker in Murc À / À VSMCs than in WT VSMCs ( Supplementary Fig. 3f). MURC overexpression increased RhoA activity in hPASMCs (Fig. 5f). Furthermore, MURC knockdown attenuated FBS-induced MYPT1 and MLC2 phosphorylation in hPASMCs, while MURC overexpression induced MYPT1 and MLC2 phosphorylation in hPASMCs ( Supplementary Fig. 3g,h). We then used Y-27632 to assess the involvement of Rho/ROCK signalling in the MURC-induced proliferation and migration of hPASMCs. Y-27632 suppressed the MURC-induced proliferation of hPASMCs in a dose-dependent manner (Fig. 5g). Y-27632 also inhibited the MURC-induced migration of hPASMCs (Fig. 5h). Moreover, hydroxyfasudil, another ROCK inhibitor, suppressed MURC-induced proliferation and migration in rat VSMCs ( Supplementary Fig. 3i,j). Collectively, these results indicate that MURC-induced proliferation and migration are mediated by the Rho/ROCK pathway in hPASMCs.
Guanine nucleotide exchange factors (GEFs) stimulate the exchange of GDP for GTP to generate the activated form of Rho proteins 43,45 . Among RhoGEFs, p115RhoGEF (also referred to as Arhgef1) and leukemia-associated Rho guanine nucleotide exchange factor (LARG) have been shown to regulate vascular tone 46,47 . To examine the functional significance of p115RhoGEF in MURC-induced RhoA activation, we used human p115RhoGEFspecific siRNA. The knockdown of p115RhoGEF suppressed MURC-induced RhoA activation in hPASMCs (Fig. 5i). We also used a mutant of p115RhoGEF [p115(2A)] that contains alanine point substitutions of two residues (E423 and N603) in the Dbl homology (DH) domain, which is responsible for the catalytic exchange reaction. RhoA activity in hPASMCs was reduced more by p115(2A) overexpression than by LacZ overexpression, indicating that p115(2A) acts in a dominant negative fashion. MURC-induced RhoA activation was significantly attenuated by p115(2A) (Supplementary Fig. 3k). Collectively, these findings indicate that p115RhoGEF mediates MURC-induced RhoA activation.
Dissociation between Cav1 and activated Ga13 by MURC. G-protein-coupled receptors (GPCRs) for various vasoconstrictors, such as ET-1 and thrombin, couple to Ga12/13, and activated Ga13 binds to p115RhoGEF, which enhances the activity of p115RhoGEF, leading to Rho activation 45,48,49 . The absence of Cav1 has been shown to increase RhoA activity in VSMCs 11 . To examine the relationship between Cav1 and MURC and its effects on RhoA activity in hPASMCs, Cav1 and MURC were transduced with recombinant retroviruses expressing Cav1 and MURC into hPASMCs. Cav1 overexpression suppressed RhoA activity in hPASMCs, which was reversed by MURC overexpression (Fig. 6a), suggesting that Cav1 negatively regulates RhoA activity, and that MURC counteracts the effects of Cav1 on the activation of RhoA. Cav1 has been identified as a potent inhibitor of heteromeric G-proteins 9 . Therefore, we assumed that Cav1 inhibits RhoA activation through its association with Ga13, which is modulated by MURC. To test this, we performed glutathione S-transferase (GST)-pulldown assays using GST-Cav1 and GST-Ga13. Cav1 exhibited higher affinity for the GTPgS (a non-hydrolyzable GTP analogue)-bound form of Ga13 (Ga13-GTPgS) than for the GDPbound form of Ga13 (Ga13-GDP), and MURC reduced the association of Cav1 with Ga13-GTPgS (Fig. 6b). Cav1 also had higher affinity for a constitutively active mutant of Ga13, Ga13(Q226L), than for Ga13 (Supplementary Fig. 4a). In addition, Cav1 had higher affinity for Ga13 incubated with GTPgS than for Ga13 incubated with GDP ( Supplementary  Fig. 4b). These findings suggest the preferential binding of Cav1 to active Ga13 over inactive Ga13. Furthermore, immunoprecipitation revealed that Cav1-associated Ga13 was more prominent in MURC-knockdown hPASMCs than in hPASMCs transfected with control siRNA (Fig. 6c).
We next examined the association of Ga13 with p115RhoGEF. In the GST-pulldown assay using GST-Ga13, Ga13-GTPgS had higher affinity for p115RhoGEF than Ga13-GDP, and Cav1 reduced the association of p115RhoGEF with Ga13-GTPgS, which was reversed by MURC (Fig. 6d). Immunoprecipitation also demonstrated that p115RhoGEF had higher affinity for Ga13 incubated with GTPgS than for Ga13 incubated with GDP, and that Cav1 reduced the association of p115RhoGEF with Ga13 incubated with GTPgS, which was reversed by MURC ( Supplementary Fig. 4c). We then attempted to identify the domain of Cav1 responsible for the associations of both Ga13 and MURC. Cav1(61-101), which is the oligomerization domain containing a scaffolding domain in the distal half, associated with MURC and Ga13 (Fig. 6e). These findings suggest that MURC and the active form of Ga13 are competitively associated with Cav1, and that the association of MURC with Cav1 decreases that between Cav1 and the active form of Ga13, thereby facilitating the active form of Ga13 to interact with p115RhoGEF.
Association of MURC with p115RhoGEF and RhoA. Since caveolae have been recognized to serve as signalling platforms for complexes of receptors, signal components and their targets 8,9,50 , we investigated the localization of MURC and p115RhoGEF. Immunostaining showed that MURC was colocalized with p115RhoGEF in hPASMCs (Fig. 7a). MURC was immunoprecipitated with p115RhoGEF, and the BiFC assay revealed the association of MURC with p115RhoGEF in hPASMCs (Fig. 7b,c). The association of MURC with p115RhoGEF was not inhibited by Cav1 (Supplementary Fig. 5). Furthermore, MURC was immunoprecipitated with RhoA (Fig. 7d). Thus, MURC has the ability to induce clustering molecules involved in RhoA signalling.

Discussion
In SMCs, Cav1 is predominantly expressed among caveolins and is the predominant determinant of caveolar morphology 10,11 . We showed here that MURC is primarily distributed at the plasma membrane of PASMCs, in which MURC is associated with Cav1 and Cav3. Even in Murc À / À VSMCs, Cav1 is expressed and located at the plasma membrane, and caveolae are retained. These observations indicate that although MURC is located together with Cav1 and Cav3 at the caveolae of VSMCs, MURC is not essential for the formation of caveolae in VSMCs. However, our observations also indicate that MURC regulates caveolar-mediated RhoA signalling through modulation of the function of Cav1. PASMC proliferation is a prominent feature of PAH 1,4 . Hypoxia-induced PH has been shown to increase the activity of Rho/ROCK signalling and be attenuated by ROCK inhibitors 18 . We demonstrated that ROCK activity evaluated by MYPT1 phosphorylation in the lung of hypoxia-exposed Murc À / À mice was weaker than that in the lung of hypoxia-exposed WT mice. cKO mice showed the alleviation of the development of hypoxiainduced PH and vascular remodelling. In addition, RhoA activity, proliferation and migration in MURC knockdown PASMCs and in Murc À / À VSMCs were less than those in control cells. These findings suggest that the attenuation of PH in Murc À / À and cKO mice is attributable to suppressed RhoA signalling in PASMCs. Previous studies demonstrated that Cav1 À / À mice developed PH 12,13 , and that proliferation, migration and RhoA activation were increased in Cav1 À / À VSMCs 11,51 . Taken together with MURC reversing the effects of Cav1 on RhoA activation in PASMCs, these observations suggest that Cav1 in PASMCs has a protective role in the development of PH, which contributes to the alleviation of PH in Murc À / À and cKO mice. The scaffolding domain of Cav1 is capable of switching off the activity of Ga-subunits 52 . We showed that both Cav1 and p115RhoGEF bind more efficiently to active Ga13 than to inactive Ga13, and that Cav1 inhibits the association of active Ga13 with p115RhoGEF, implying that the interaction of Cav1 with active Ga13 prevents further binding of active Ga13 to p115RhoGEF, which may account for the inhibitory effects of Cav1 on RhoA activity through the effects of Ga13. We also showed that the oligomerization and scaffolding domains of Cav1 bind to both Ga13 and MURC, suggesting that the competitive binding of MURC to Cav1 facilitates the association of Ga13 with p115RhoGEF, which leads to the activation of Rho/ROCK signalling. Various cellular functions ascribed to the small GTPase Rho are dependent on the spatial control of activation 45 . Inactive Rho is located within the cytoplasm and, following the activation of G-protein-coupled receptors, translocates to the plasma membrane, in which Rho is activated 43,49 . The subcellular localization of GEFs is a key aspect of Rho activity, and GEF activation is intimately linked with its localization 45,49 . Because MURC is associated with both p115RhoGEF and RhoA, our findings suggest that MURC functions as a scaffold to facilitate the compartmentation of p115RhoGEF and RhoA within the caveolae of PASMCs, which is likely to be involved in their activation and the subsequent activation of downstream signalling, which regulates the proliferation and migration of PASMCs. Collectively, our results suggest the novel concept that MURC functions as a component of caveolae to regulate Cav1-mediated Rho/ROCK signalling (Fig. 7e).
In addition to the Rho/ROCK pathway, Ga13 has been shown to be involved in several pathways, such as cadherin, PYK2 and apoptosis signal-regulating kinase-1 (ASK-1) signalling pathways 53 . Since MURC inhibits the association of Cav1 with Ga13, MURC may also influence signalling pathways mediated by Ga13. Cav1 deficiency and mutations cause PAH 6,12,13 . Therefore, alterations in the expression and function of Cav1 in PASMCs may generally contribute to the pathology of PAH. MURC modulates the effects of Cav1 in PASMCs; therefore, the inhibition of MURC may serve as a therapeutic target in PAH.
In conclusion, MURC deficiency in SMCs alleviates the development of PH. The association of MURC with Cav1 promotes the Ga13-mediated activation of p115RhoGEF, which leads to the subsequent activation of Rho/ROCK signalling, resulting in the enhanced proliferation and migration of PASMCs. Our findings provide a previously undescribed function for MURC in caveolar-mediated Rho/ROCK signalling and novel insights into the underlying pathogenetic mechanisms of PH. When MURC is present, it associates with Cav1, leading to reductions in the association between Cav1 and Ga13, which facilitates the interaction of Ga13 with p115RhoGEF, thereby activating the Rho/ROCK pathway. MURC also serves as a platform to compartmentalize p115RhoGEF and RhoA at caveolae. In the absence of MURC, Cav1 interacts with Ga13 and inhibits the association of Ga13 with p115RhoGEF, leading to the suppression of the Rho/ROCK pathway. Uncropped images of blots are shown in Supplementary Fig. 6.
the rat monoclonal antibody to hemagglutinin (HA, 1:500) was from Roche Applied Science; the mouse monoclonal antibody to T7 (1:5,000) was from Novus Biologicals; the horseradish peroxidase-conjugated monoclonal antibody to GST (1:5,000) was from Wako Pure Chemical Industries, Ltd. (Osaka, Japan); the mouse monoclonal antibody to GAPDH (1:2,000) was from Millipore; the mouse monoclonal antibody to FLAG (1:1,000), Cy3-conjugated monoclonal antibody to aSMA (1:200), GDP and collagenase type 1 from Clostridium histolyticum were from Sigma-Aldrich; GTPgS was from Cytoskeleton, Inc.; the mouse monoclonal antibodies to Cav1 (1:200) and Cav3 (1:200) were from BD Biosciences; the rabbit polyclonal antibodies to MYPT1 (1:500), phospho-MYPT1 at Thr853 (pMYPT1 Thr853 , 1:500), and phospho-MLC2 at Ser19 (pMLC2 Ser19 , 1:500 for Western blotting, 1:50 for immunostaining) were from Cell Signaling Technology, Inc.; and the rabbit polyclonal antibody to Ki67 (1:100) and FITC-conjugated monoclonal antibody to aSMA (1:50) were from Abcam Biochemicals. The human Cav1-and Cav3-expressing vectors, pcDNA3.1-T7-hCav1 and pcDNA3.1-T7-hCav3, were gifts from Yukiko K. Hayashi (Tokyo Medical University, Tokyo, Japan) 27  Isolation and culture of SMCs. Rat VSMCs isolated from the thoracic aorta of Sprague-Dawley rats by the collagenase digestion method were maintained in DMEM with 10% FBS. In brief, the rat thoracic aorta was isolated under the thoracotomy. After the resection of connective tissue and fat tissue, the aorta was incubated with Hank's balanced salt solutions (HBSS) containing collagenase at 37°C for 25 min. After the adventitia and intima were removed, the aorta was cut into 1 mm in the HBSS containing collagenase and elastase, and incubated at 37°C for 3 h. The sample was suspended and centrifuged. The pellet was resuspended in 10 ml of fresh DMEM containing 10% FBS and incubated at 37°C. Mouse VSMCs were isolated from the mouse aorta. In brief, the mouse aorta was isolated from the iliac bifurcation to the aortic arch by longitudinal incision. The aorta was incubated with DMEM containing 745 units per ml collagenase type 1 at 37°C for 20 min. After washing the aorta with DMEM, the tissue was incubated with DMEM containing 280 units per ml collagenase and 11.7 units per ml elastase at 37°C for 1 h. After filtration through a 40-mm filter, the filtrated sample was transferred to a 50-ml conical tube and centrifuged. The pellet was resuspended in 4 ml of fresh DMEM containing 10% FBS and incubated at 37°C. Only early-passage (passage 3 or 4) cells were used in the experiments described below. hPASMCs were cultured in SmGM-2 medium (Lonza, Walkersville, MD, USA) supplemented with SmGM-2 SingleQuots (Lonza, Walkersville, MD, USA) according to the manufacturer's instructions.
Production of a polyclonal antibody to MURC. Rabbit immunization was conducted by UNITECH (Chiba, Japan) using synthetic peptides corresponding to the N-terminal residues of mouse Murc (MEHNGSASNAGKIHQNRC). In the Western blot analysis, IgG was purified from antisera.
Protein-protein interaction analysis. In the in situ protein association analysis, we performed BiFC assays using a CoralHue Fluo-chase Kit (Medical & Biological Laboratories) according to the manufacturer's protocol 34,54,55 . In each protein association assay, the appropriate pairs of plasmids in which each target protein gene was fused to the divided CoralHue Kusabira-Green gene N-terminal fragment or C-terminal fragment were cotransfected into the cultured hPASMCs. After a 24-h incubation, the fluorescent signal was detected using a Zeiss LSM510 META Confocal Imaging System.
Hypoxic mouse model of PH. WT and Murc À / À mice, or Murc fl/fl and cKO mice were exposed to normobaric hypoxia (10% O 2 ) in a chamber in which oxygen was tightly regulated by the oxygen controller ProOx110 (KYODO International, Kanagawa, Japan) for 4 weeks 57 . Nitrogen was automatically introduced as required to maintain the proper fraction of inspired oxygen (FiO 2 ). Age-and sex-matched littermates were exposed to identical conditions in normoxia and served as controls.
Measurement of RV pressure and histological analyses. Mice were intubated with a 22-gauge Teflon tube and placed in a supine position. To measure RV hemodynamics, open-chest RV catheterization using a 1.2-F pressure catheter (Scisense Inc.) was performed during anaesthesia with 1.5% isoflurane. Pulmonary vascular remodelling was assessed by measuring the medial thickness of alveolar/ distal pulmonary vessels of 25-100 mm in diameter, which are not associated with bronchi, from lung sections immunostained with aSMA. Per cent wall thickness is expressed as the medial wall area (the area between the internal and external lamina) divided by the area of the vessel (the area between the external lamina). Multiple lung sections were made for each mouse and 45 vessels were analysed in each lung section.
Insertion of an osmotic pump. Mice were anaesthetized with 1.5% isoflurane, the skin was shaved, and mini-osmotic pumps (model 2004; Alzet, Cupertino, CA, USA) containing either PBS or Y-27632 (CALBIOCHEM; 125 mg ml À 1 of saline) were inserted into a subcutaneous pocket through a small incision made in the skin between the scapulae. These pumps delivered a volume of 0.25 ml h À 1 , which was equivalent to 30 mg kg À 1 Á day À 1 of Y-27632 for a 25-g mouse 58 .
Histological analysis. Fixation was performed using 4% PFA/PBS. Hearts, lungs and aortae were cut and paraffin sections of 3 mm thick were stained with H&E.
Transmission electron microscopy and quantitation. Twelve-week-old mouse lungs were fixed with 2% glutaraldehyde in 0.1 M cacodylate buffer, post-fixed with 2% OsO 4 , and stained with uranyl acetate and lead citrate. Microtome sections were examined under a H-7100 transmission electron microscope (HITACHI, Tokyo, Japan) and photographed at a magnification of Â 40,000 or Â 20,000. Caveolae were identified by their characteristic flask shape and location at or near the plasma membrane 34,56 .
Echocardiographic and morphometric analyses. After mice had been anaesthetized with isoflurane (1.5%, Abbott Laboratories), echocardiography was performed using a Vevo 2100 system (VisualSonics) equipped with a 30-MHz microprobe. After echocardiography, mice were sacrificed by cervical dislocation and then weighed. Hearts were excised, rinsed in PBS and weighed. Body weight and left tibial length were measured to normalize heart weight.

Recombinant retroviruses and gene transfer.
To generate recombinant retroviruses, GP2-293 cells (Clontech) were cotransfected with the helper vector pVSV-G and pMSCVpuro-LacZ, pMSCVhyg-LacZ, pMSCVpuro-Murc-FLAG or pMSCVhyg-hCav1-HA. The medium supernatant was collected and centrifuged to concentrate virus stocks according to the manufacturer's instructions. hPASMCs were infected with the retrovirus in the presence of 8 mg ml À 1 polybrene for 24 h, and medium was changed to fresh medium. Infected cells were selected with 1 mg ml À 1 puromycin or 50 mg ml À 1 hygromycin.
Migration assay. Confluent monolayer cells in a 60-mm dish were scraped in a straight line to create a 'scratch' with a p200 pipette tip. The scratched cells were washed once with DMEM to remove debris and smooth the edge of the scratch, and then replaced with 4 ml of migration medium (DMEM with 0.1% BSA). The dish was incubated at 37°C for 24 h. Wound closure was quantified by the per cent change in the wound area. A Boyden chamber assay was performed using a Transwell Permeable Support 8.0-mm polycarbonate membrane (Coster). The lower chamber contained 1% FBS as a chemoattractant. hPASMCs transfected with control siRNA or MURC siRNAs were prepared in serum-free medium, and 3 Â 10 4 cells were added to the upper chamber. After a 4-h incubation, migrated cells were stained with Hoechst 33342 and counted under a microscope [59][60][61][62] .
GST pulldown assay. Ten micrograms of GST-Cav1(FL), GST-Cav1(1-61), GST-Cav1(61-101), or GST-Cav1(135-178) was bound to glutathione-Sepharose beads. The beads were incubated with lysates from COS cells transfected with the MURC or Ga13 expression plasmid at 4°C for 2 h. GST-conjugated glutathione-Sepharose beads were used as a control. Beads were washed five times, and proteins were eluted for western blotting. In the GST pulldown assay of the Ga13 protein, 10 mg of the GST-Ga13 protein was preloaded with GDP (1 mM) or GTPgS (0.1 mM) to binding buffer (0.1 M Tris-HCl, pH 7.4, 1 mM EDTA, 2 mM DTT, 0.2 M NaCl) and incubated at 30°C for 15 min. After adding a stop buffer (50 mM MgCl 2 ), the GST-Ga13 protein converted to the inactive or active form was linked to glutathione-Sepharose beads, which were then incubated with lysates from COS cells transfected with T7-Cav1, MURC-HA, and/or p115RhoGEF-FLAG expression plasmids at 4°C for 2 h.
Statistical analyses. All experiments were performed at least three times. Data are expressed as means ± standard errors. Data were analysed by the unpaired Student's t-test for comparisons between two groups or a one-way analysis of variance with a post hoc analysis for multiple comparisons. A P value of o0.05 was considered significant.
Data availability. The authors declare that all the data supporting the findings of this study are available within the article and its Supplementary Information Files and from the authors upon request.