Atrial natriuretic peptide (ANP) is a cardiac hormone released by the atrium in response to stretching forces. Via its receptor, guanylyl cyclase-A (GC-A), ANP maintains cardiovascular homeostasis by exerting diuretic, natriuretic, and hypotensive effects mediated, in part, by endothelial cells. Both in vivo and in vitro, ANP enhances endothelial barrier function by reducing RhoA activity and reorganizing the actin cytoskeleton. We established mouse endothelial cells that stably express GC-A and used them to analyze the molecular mechanisms responsible for actin reorganization. Stimulation by ANP resulted in phosphorylation of myosin light chain (MLC) and promotion of cell spreading. p21-activated kinase 4 (PAK4) and cerebral cavernous malformations 2 (CCM2), a scaffold protein involved in a cerebrovascular disease, were required for the phosphorylation of MLC and promotion of cell spreading by ANP. Finally, in addition to the GC domain, the kinase homology domain of GC-A was also required for ANP/GC-A signaling. Our results indicate that CCM2 and PAK4 are important downstream mediators of ANP/GC-A signaling involved in cell spreading, an important initial step in the enhancement of endothelial barrier function.
Guanylyl cyclase-A (GC-A) is a transmembrane receptor for atrial natriuretic peptide (ANP) and B-type natriuretic peptide [1,2]. ANP, which is primarily secreted from the atrium in response to stretching, exerts pleiotropic effects on the cardiovascular system via GC-A. These effects include reduction of blood volume and blood pressure and induction of vasodilation, resulting in the maintenance of cardiac integrity. GC-A is highly expressed in the endothelium. Endothelial-specific deletion of the GC-A gene in mice causes arterial hypertension and cardiac hypertrophy, although the vasodilating effects of ANP are unaffected . Thus, the endothelium is thought to be one of the major sites of ANP function.
Both in vivo and in vitro, ANP enhances endothelial barrier function via regulation of the actin cytoskeleton. Using pulmonary endothelial cells, Tian et al.  demonstrated that ANP regulates the actin cytoskeleton by inactivating RhoA and non-muscle myosin-II (NM-II), resulting in protection against endothelial barrier dysfunction by thrombin. Furthermore, they showed that RhoA is inactivated by a serine/threonine kinase, p21 protein-activated kinase 1 (PAK1), through the phosphorylation and inactivation of GEF-H1, a guanine nucleotide exchange factor (GEF) specific for RhoA. RhoA activates NM-II by phosphorylation of the myosin light chain (MLC) via Rho-kinase. Although most studies have focused on phosphorylation of MLC by the Rho/Rho-kinase pathway, several other kinases, including PAK1 and PAK4, also enhance MLC phosphorylation, indicating that different upstream signals activate NM-II and regulate rearrangement of the actin cytoskeleton [5–8]. Ando et al.  demonstrated that myotonic dystrophy kinase-related CDC42-binding kinase (MRCK) promotes phosphorylation of MLC at cell–cell junctions and enhances endothelial barrier function in cells stimulated with forskolin. Thus, NM-II plays opposing roles in the regulation of endothelial barrier function under different conditions.
GC-A consists of an extracellular ligand-binding domain, a transmembrane domain, a kinase homology domain (KHD), a hinge region, and a guanylyl cyclase domain (GCD) . Upon ligand binding, GCD produces cGMP from GTP. Deletion of the KHD causes constitutive activation of GC activity; thus, the KHD acts as a repressor of GC activity in the absence of ligand . The cGMP produced by GC-A mediates ANP signaling by activating downstream molecules. Protein kinase G I (PKGI) is thought to be the main target of cGMP in endothelial cells; however, it remains unclear whether PKGI is involved in enhancement of endothelial barrier function by ANP .
Loss-of-function mutations in CCM1, CCM2, or CCM3 cause the neurovascular disease cerebral cavernous malformations (CCMs). CCMs are vascular lesions characterized by abnormally dilated and hyperpermeable blood vessels lacking smooth muscle support [12–14]. Endothelial cells lacking CCM1, -2, or -3 exhibit aberrant activation of RhoA, increased formation of stress fibers, and loss of barrier function [15–17]. Inhibition of Rho-kinase, a major effector of RhoA, rescues loss of barrier function in CCM1- or CCM2-deficient mice [17,18]. Thus, in normal endothelial cells, CCM1, -2, and -3 are thought to suppress RhoA activity to maintain endothelial barrier function. The roles of CCM1, -2, and -3 are similar to those of ANP, suggesting that there is a close functional relationship between the CCM and ANP/GC-A pathways. To date, however, no study has tested this possibility.
Cell spreading is the initial step in formation of the endothelial barrier; therefore, analyses of the molecular mechanism of actin rearrangement followed by cell spreading contribute to our understanding of endothelial barrier function . In general, cell biological analysis of the action of ANP has been hampered by low expression of endogenous GC-A in cell culture [1,20]. To address this issue, several studies have used HEK293 cells stably expressing exogenous GC-A [21,22]. In this study, we established an endothelial cell line that stably expresses GC-A. Specifically, we used a retroviral vector expressing FLAG-tagged GC-A to infect SV40-transformed mouse endothelial cells (SVECs), yielding the cell line SVEC/GC-A. We used these cells to investigate the molecular mechanism responsible for reorganization of the actin cytoskeleton and promotion of cell spreading by ANP. We found that ANP dynamically rearranges the actin cytoskeleton via regulation of NM-II, PAK4, and CCM2, resulting in cell spreading. Moreover, we found that the KHD of GC-A is important for activation of the downstream targets of ANP/GC-A signaling.
Reagents and antibodies
Reagents were acquired from the following suppliers: blebbistatin and 3-isobutyl-1-methylxanthine, Santa Cruz Biotechnology (Dallas, TX, U.S.A.); Fasudil (HA-1077), Abcam (Cambridge, MA, U.S.A.); bovine myelin basic protein, Sigma–Aldrich (St. Louis, MO, U.S.A.); and rhodamine–phalloidin, Invitrogen (Carlsbad, CA, U.S.A.). Pulldown assay kits for RhoA and Rac1/Cdc42 were obtained from Merck Millipore (Darmstadt, Germany). Antibodies were purchased from the following suppliers: anti-FLAG M2, Sigma–Aldrich; anti-GC-A (ab14356), Abcam; anti-HA (TANA2), MBL (Nagoya, Japan); anti-phospho-MLC2 (S19), anti-phospho-vasodilator-stimulated phosphoprotein (VASP) (S157), anti-phospho-VASP (S239), anti-phospho-Zyxin (S142/S143) (D1E8), anti-VASP (9A2), anti-Zyxin, and anti-Actin (13E5), Cell Signaling Technology (Beverly, MA, U.S.A.); anti-CCM2 and anti-PAK4, Proteintech (Rosemont, IL, U.S.A.); and anti-GAPDH (6C5) and anti-RhoA (26C4), Santa Cruz Biotechnology.
siRNAs targeting the indicated genes and control siRNA duplexes were purchased from Sigma–Aldrich. Nucleotide sequences for siRNAs used were as follows:
for mouse RhoA, 5′-CCGGAAGAAACUGGUGAUUdTdT-3′ (siRhoA#1) and 5′-CCAGUUCCCAGAGGUCUAUdTdT-3′ (siRhoA#2);
for mouse PAK4, 5′-GCUGCCCAUCACUGGAAAUdTdT-3′ (siPAK4#1) and 5′-GGGACAGAGACGACACUAUdTdT-3′ (siPAK4#2);
for bovine PAK4, 5′-GGGACUACCAGCAUGAGAAdTdT-3′ (siPAK4#3) and 5′-GCAUGAGCAGAAGUUCACUdTdT-3′ (siPAK4#4);
for mouse CCM2, 5′-GCAGAAACCUCUGUGCCUAdTdT-3′ (siCCM2#1) and 5′-GGACCGAGCAAUAUUUGAUdTdT-3′ (siCCM2#2);
for bovine CCM2, 5′-CCGAAAUCCUGCAUUUCAUdTdT-3′ (siCCM2#3) and 5′-GGACAGAGCCAUAUUUGAUdTdT-3′ (siCCM2#4); and
for mouse PKGI, 5′-GCAUUCGCCUAUCUGCAUUdTdT-3′ (siPKGI#1).
Human GC-A expression vector (pcDNA3 GC-A-HA) was constructed by introducing the GC-A fragment [amplified by PCR from vector pME18SFL3, which encodes GC-A, obtained from the Biological Resource Center (NBRC), National Institute of Technology and Evaluation (http://flj.lifesciencedb.jp/top/, Chiba, Japan)] into the NotI and XhoI sites of pcDNA3 HA. The plasmid expressing Lifeact-mCherry (pmCherry-Lifeact) was obtained from N. Mochizuki (National Cerebral and Cardiovascular Center Research Institute, Japan). The plasmids expressing the EGFP or mCherry-tagged CAAX motif of human K-Ras (pEGFP-CAAX) were constructed by introducing the CAAX motif (amino acids 169–188) of human K-Ras, prepared by annealing synthetic oligonucleotides, into the BglII and EcoRI sites of pEGFP-C1 or pmCherry-N1. The plasmid used for preparing retrovirus vector expressing GC-A-FLAG (pCX4-puro GC-A-FLAG) was constructed by introducing the GC-A-Flag fragment (amplified by PCR from pcDNA3 GC-A-HA) into the NotI site of pCX4-puro. The plasmid used for preparing retrovirus vector expressing untagged GC-A (pCX4-puro GC-A) was constructed as follows: a GC-A fragment was prepared by digesting pcDNA3 GC-A-HA with XhoI and blunting with T4 polymerase, followed by digestion with NotI. The fragment was inserted into the NotI and HpaI sites of pCX4-puro. The plasmid used for preparing retrovirus vector expressing FLAG-Rac1 (pCX4 hygro FLAG-Rac1) was prepared by introducing the FLAG-Rac1 fragment (amplified by PCR from pCEFL AU5-Rac1) into the BamHI and NotI sites of pCX4 hygro. The plasmid used for preparing retrovirus vector expressing FLAG-Cdc42 (pCX4 hygro Flag-Cdc42) was prepared by introducing the FLAG-Rac1 fragment (amplified by PCR from pCEFL AU5-Cdc42) into the BamHI and NotI sites of pCX4 hygro. The plasmid used for preparing retrovirus vector expressing HA-CCM2 (pCX4 hygro HA-CCM2) was prepared by introducing the HA-CCM2 fragment [amplified by PCR from vector pDONR223, which encodes CCM2; Promega (Madison, WI, U.S.A.)] into the NotI site of pCX4 hygro. The plasmid used for preparing retroviral vector expressing FLAG-PAK4 (pCX4 bsr FLAG-PAK4) was prepared by introducing the FLAG-PAK4 fragment (amplified by PCR from pcDNA3 FLAG-PAK4) into the HpaI site of pCX4 bsr using the In-Fusion HD cloning kit (TaKaRa Bio, Otsu, Japan). Point and deletion mutations were introduced into pcDNA3 GC-A-HA or pCX4-puro GC-A-FLAG using the PrimeSTAR Mutagenesis Basal Kit (TaKaRa Bio). FlincG3 expression vector (pTriEx4-H6-FGAm) was obtained from J. Garthwaite (University College London, U.K.). The primers used are listed in Supplementary Table S1.
Cell culture, transfection, infection, immunoprecipitation, and Western blotting
SVECs were obtained from Y. Takuwa (Kanazawa University). SVEC and HEK293T cells were cultured in DMEM supplemented with 10% FCS. Bovine pulmonary artery endothelial cells (BPAECs) were obtained from Cell Applications (San Diego, CA, U.S.A.) and cultured in bovine endothelial growth medium (Cell Applications). Human pulmonary artery endothelial cells (HPAECs) were obtained from Lonza (Walkersville, MD, U.S.A.) and cultured in EGM-2 medium (Lonza). For immunofluorescence analysis or electric cell-substrate impedance sensing (ECIS) measurements, SVECs were seeded at low (5 × 103/cm2) or high density (5 × 104/cm2). For the other experiments, SVECs were seeded at moderate density (1 × 104/cm2). HEK293T cells were seeded at high density (5 × 104/cm2). For transient transfection of HEK293T, cells were plated onto collagen-coated dishes and transfected with plasmid DNA using Lipofectamine 2000 (Invitrogen). Retroviral infection of SVEC was performed using ecotropic retroviruses prepared from pE-eco, pGp (Takara Bio), and pCX4 as described previously . Retroviral infection of HEK293T cells was performed using ecotropic retroviruses and Ecotropic Receptor Booster (Takara Bio). After infection, cells were selected with puromycin, blasticidin, or hygromycin. Transfection of siRNA into SVEC was performed using Lipofectamine RNAiMAX (Invitrogen). SVEC and HEK293T cells were treated with 100 nM ANP for the indicated times. Immunoprecipitation assays were performed using antibodies coupled with Protein A-Sepharose, as described previously ; cells were lysed in NP-40 buffer [1% Nonidet P-40, 20 mM Tris–HCl (pH 7.4), 150 mM NaCl, and 5 mM EDTA] supplemented with protease and phosphatase inhibitor cocktails. Western blotting was performed as described previously . To enhance the signal from phospho-MLC, PVDF membrane was fixed with 0.25% glutaraldehyde as described previously .
Live imaging of SVEC expressing GFP-CAAX
To visualize the plasma membrane, SVEC or SVEC/GC-A were transiently transfected with pEGFP-CAAX and seeded on glass-bottom dishes. Images were acquired at 37°C on a DeltaVision Elite system (GE Healthcare, Waukesha, WI, U.S.A.) equipped with an inverted microscope (IX-70; Olympus, Tokyo, Japan), a UPlan-Apochromat 100×/NA 1.4 oil immersion objective lens, and a 12-bit CCD camera (CoolSNAP HQ; Photometrics, Tucson, AZ, U.S.A.). Time-lapse image stacks were taken in 2 µm z-increments at 10 s intervals, deconvoluted, and projected using the softWorx software (GE Healthcare).
Live imaging of SVEC/GC-A expressing Lifeact-mCherry
To visualize F-actin, SVEC/GC-A cells were infected with the retrovirus vector expressing Lifeact-mCherry, and cells stably expressing Lifeact-mCherry were selected using blasticidin. Phase-contrast and mCherry images were collected at 37°C in a 5% CO2 humidified atmosphere using an inverted microscope (IX81; Olympus), a 40×/NA 1.3 oil immersion objective lens, and an iXon3 897 CMOS camera (Andor Technology, Belfast, U.K.). Time-lapse images were taken at 1 min intervals. Acquired images were processed with MetaMorph (Molecular Devices, Sunnyvale, CA, U.S.A.).
Live imaging of cGMP in SVEC/GC-A
To measure cGMP production in live cells, SVEC/GC-A cells were seeded on collagen-coated glass-bottom dishes and transiently transfected with the FlincG3 expression vector . GFP images were collected at 37°C in a 5% CO2 humidified atmosphere using an inverted microscope (IX81; Olympus), a 40×/NA 1.3 oil immersion objective lens, and an iXon3 897 CMOS camera (Andor Technology, Belfast, U.K.). Time-lapse images were taken at 3 min intervals. Acquired images were processed with MetaMorph (Molecular Devices, Sunnyvale, CA, U.S.A.). For spatiotemporal analysis of cGMP production, SVEC/GC-A cells were transiently transfected with FlincG3 and mCherry-CAAX vectors. Live images of GFP and mCherry were obtained using a confocal microscope (FV1000; Olympus) with a 60×/NA 1.5 oil immersion objective lens.
SVECs grown on glass-bottom dishes were fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton X-100 in PBS for 3 min, blocked with 2% BSA in PBS for 1 h, incubated with primary antibodies in 2% BSA in PBS for 1 h, and then incubated with Alexa Fluor-conjugated secondary antibodies for 1 h, followed by incubation with rhodamine–phalloidin. Phase-contrast and fluorescent images were acquired using an inverted microscope (IX81; Olympus), a 40×/NA 1.3 oil immersion objective lens, and a CoolSNAP HQ 12-bit charge-coupled device camera (Photometrics). Acquired images were processed using MetaMorph (Molecular Devices). Confocal images were acquired using FV1000. Confocal images were processed using the FluoView software (Olympus). Statistical analysis of co-localization and reconstruction of three-dimensional images were performed using the Volocity software (PerkinElmer, Waltham, MA, U.S.A.).
Measurement of transendothelial resistance by ECIS
For ECIS measurements, SVECs were seeded on 8W10E+ gold-coated electrodes (Applied Biophysics, Troy, NY, U.S.A.) at high density (5 × 104/cm2). Cells were allowed to attach and spread for 24 h. Capacitance and resistance were measured using an ECIS model 1600 (Applied Biophysics).
Real-time PCR analysis
Total RNA was purified using the CellAmp Direct RNA Prep kit for RT-PCR (Takara Bio). Quantitative real-time PCR was carried out using the One Step SYBR PrimeScript PLUS RT-PCR kit (Takara Bio) on a LightCycler 480II (Roche, Indianapolis, IN, U.S.A.). For normalization, hypoxanthine phosphoribosyltransferase 1 was used as an internal control. Primer sequences are provided in Supplementary Table S1.
Measurement of GC activity by competitive ELISA
Cells seeded at moderate density were treated with 0.2 mM 3-isobutyl-1-methylxanthine for 10 min, and then with 100 nM ANP for 15 min. Production of cGMP was measured in whole-cell lysates using the cyclic GMP XP™ assay kit (Cell Signaling Technology).
Measurement of PAK4 kinase activity
HEK293T cells stably expressing untagged GC-A and FLAG-PAK4 were stimulated with ANP (100 nM) or Fasudil (20 µM) for 30 min. FLAG-PAK4 was immunoprecipitated with anti-FLAG antibody coupled to Protein A-Sepharose. Immobilized FLAG-PAK4 was eluted with FLAG peptide (150 µg/ml) at 4°C for 30 min. Purified FLAG-PAK4 was incubated at 37°C for 30 min in 50 µl of kinase buffer [40 mM Tris–HCl (pH 7.4), 20 mM MgCl2, 1 µM ATP, 0.1 mg/ml BSA, and 1 mM dithiothreitol] containing 50 µM myelin basic protein as a substrate. After the kinase reaction, ATP concentrations were determined using the Kinase-Glo assay kit (Promega).
We analyzed quantitative differences among multiple groups by one-way ANOVA followed by Bonferroni's post-test, and differences between two groups by Student's unpaired t-test. Values are expressed as means ± SD. P-values of <0.05 were considered statistically significant.
ANP induces robust phosphorylation of VASP and Zyxin and cell spreading of mouse immortalized endothelial cells stably expressing GC-A.
To understand the molecular basis of reorganization of the actin cytoskeleton by ANP/GC-A signaling in endothelial cells, we established a mouse immortalized endothelial cell line stably expressing GC-A (SVEC/GC-A). First, we confirmed high-level expression of exogenous GC-A and ANP-dependent enhanced production of cGMP in SVEC/GC-A (Figure 1A,B). Live imaging of the cGMP biosensor revealed sustained production of cGMP for at least 5 h (Figure 1C). Spatiotemporal analysis revealed that cGMP is initially produced at the plasma membrane and then spreads into the cyctoplasmic region (Figure 1D). In human endothelial cells, VASP and Zyxin are phosphorylated by ANP stimulation [27,28]. As shown in Figure 1E, when SVEC/GC-A cells were stimulated with ANP, we observed robust phosphorylation of Zyxin at Ser142 and VASP at Ser157 and Ser239, with band shifts as reported , indicating that stable expression of GC-A confers hyper-responsiveness to ANP.
ANP suppresses membrane blebbing and induces membrane ruffling in SVEC/GC-A.
Using these cells, we investigated the effects of ANP on cell morphology. As shown in Figure 1F, when SVECs were cultured on non-coated glass-bottom dishes at low density, most cells exhibited membrane blebbing, as also observed in bovine aortic endothelial cells after trypsinization [30,31]. Live imaging of SVEC/GC-A transfected with GFP-CAAX, a plasma membrane marker, revealed rapid formation and retraction of blebs (Supplementary Movie S1), a process that is tightly coupled with reorganization of cortical F-actin [32,33]. We next investigated how ANP regulates the dynamics of the plasma membrane and actin cytoskeleton in SVEC/GC-A. As shown in Figure 1F and Supplementary Movie S2, stimulation with ANP caused loss of membrane blebs and promoted cell spreading in SVEC/GC-A, but not in parental SVEC. Analysis of ANP-mediated morphological changes in SVEC/GC-A by confocal microscopy revealed that ANP induced loss of membrane blebs and actin stress fibers, formation of membrane ruffles, and cortical accumulation of F-actin (Figure 1G,H). Three-dimensional images and live imaging of SVEC/GC-A also revealed loss of membrane blebs and formation of membrane ruffles (Figure 1I and Supplementary Movie S3). These results indicate that SVEC/GC-A respond to ANP and dynamically change their morphology concomitant with rearrangement of the actin cytoskeleton.
ANP promotes cell spreading and cell–cell adhesion of SVEC/GC-A
We next investigated the localization of F-actin in cells cultured at moderate density. Under these conditions, formation of membrane blebs was suppressed, and F-actin accumulated weakly at sites of cell–cell interaction in the absence of ANP (Figure 2A). When cells were stimulated with ANP, we observed clear accumulation of F-actin (Figure 2A), suggesting that cell–cell adhesion was enhanced by ANP.
ANP-mediated increase in cell spreading and cell–cell interactions of SVEC/GC-A cultured at moderate density.
Measurement of transendothelial resistance using ECIS is a convenient method for monitoring changes in cell morphology and barrier function of endothelial cells in real time . Decreased capacitance measured at 16 000 Hz reflects an increase in cell spreading, whereas increased resistance measured at 4000 Hz in general reflects enhancement of barrier function, although it also reflects an increase in cell adhesion to the electrode. As shown in Figure 2B,C, stimulation of ANP swiftly decreased capacitance, indicating that ANP promotes cell spreading. Resistance was increased by treatment with ANP (Figure 2C). Together with the accumulation of F-actin at sites of cell–cell interactions (Figure 2A), our results suggest that ANP enhances barrier function following cell spreading. In contrast with SVEC/GC-A, parental SVEC did not exhibit clear changes in capacitance or resistance following treatment with ANP (Figure 2D,E).
NM-II activity is required for the enhancement of cell spreading by ANP
Next, we tried to clarify the molecular mechanism of reorganization of the actin cytoskeleton and cell morphology by ANP/GC-A signaling. Rearrangement of the actin cytoskeleton is often coupled to regulation of NM-II activity. Hence, we first tested the effects of blebbistatin, an NM-II inhibitor, on the actin cytoskeleton and cell morphology. Treatment with blebbistatin caused rapid loss of membrane blebs (Figure 3A). When cells were stimulated with ANP in the presence of blebbistatin, we observed cell shrinkage instead of spreading (Figure 3A), indicating that NM-II activity is required for spreading of cells stimulated with ANP. Consistent with the morphological changes, ECIS analysis revealed that blebbistatin completely reversed the effects of ANP on capacitance and resistance (Figure 3C).
NM-II activity is required for the promotion of cell spreading by ANP.
NM-II activity is regulated by phosphorylation of MLC. As shown in Figure 3D,E, ANP promoted phosphorylation of MLC at Ser19. Before ANP stimulation, phosphorylated MLC co-localized at stress fibers and blebs (Figure 3F). After ANP stimulation, phosphorylated MLC mostly co-localized with the cortical actin cytoskeleton (Figure 3F), indicating that ANP not only increases phosphorylation of MLC, but also changes the site of MLC phosphorylation.
Inhibition of RhoA signaling by ANP increases phosphorylation of MLC
Rho family GTPases, including RhoA, Rac1, and Cdc42, are master regulators of the actin cytoskeleton and membrane dynamics. For example, membrane blebbing and stress fiber formation require RhoA activity, whereas membrane ruffling and lamellipodia require Rac1 activity [32,33,35]. Therefore, we investigated whether ANP regulates the activities of Rho GTPases in SVEC/GC-A. We found that ANP suppressed RhoA, but activated Rac1 (Figure 4A,B). These results correlated well with the morphological changes caused by ANP (Figure 1D–G). The activity of Cdc42 was not altered by ANP (Figure 4C).
Inactivation of the Rho/Rho-kinase pathway increases phosphorylation of MLC.
We then examined whether inhibition of RhoA activity is sufficient for the phosphorylation of MLC. As shown in Figure 4D,E, the phosphorylation of MLC was elevated in cells depleted of RhoA. One of the important effector molecules of RhoA signaling is Rho-kinase (also known as ROCK) . To determine whether Rho-kinase is involved in the regulation of MLC phosphorylation, we inhibited Rho-kinase activity using Fasudil, which rapidly increased the phosphorylation level of MLC (Figure 4F,G). Thus, inhibition of the Rho/Rho-kinase pathway promotes the phosphorylation of MLC. We next investigated the effect of knockdown of RhoA and inhibition of Rho-kinase activity on cell morphology and the actin cytoskeleton. Both knockdown of RhoA and inhibition of Rho-kinase activity suppressed membrane blebbing (Figure 4H,I), although there was no obvious increase in cell spreading relative to stimulation with ANP. These results indicate that inhibition of Rho/Rho-kinase pathway, at least in part, mediates ANP signaling.
PAK4 is required for phosphorylation of MLC and enhancement of cell spreading by ANP
We next sought to identify the kinase involved in the phosphorylation of MLC by ANP. Several kinases, including Aurora kinase, Citron kinase, Rho-kinase, MLC kinase, MRCK, PAK1, and PAK4, phosphorylate MLC either directly or indirectly [5,7,36–41]. Among these, we were primarily interested in PAK4, because knockdown of PAK4 induces membrane blebbing during mitosis in HeLa cells . Intriguingly, knockdown of PAK4 completely suppressed the phosphorylation of MLC (Figure 5A,B) and also inhibited cell spreading (Figure 5C). Consistent with this, the increase in resistance and decrease in capacitance were also suppressed by knockdown of PAK4 (Figure 5D,E), indicating that PAK4 is an important mediator of ANP/GC-A signaling. We then investigated whether ANP signaling activates PAK4. For this purpose, FLAG-tagged PAK4 and untagged GC-A were stably expressed in HEK293T cells. After stimulation with ANP, FLAG-PAK4 was purified and subjected to an in vitro kinase assay. The results revealed that ANP activated PAK4 (Figure 5F). Treatment with Fasudil also increased the kinase activity of PAK4 (Figure 5F), suggesting that PAK4 is activated upon inhibition of the Rho/Rho-kinase pathway by ANP. Members of the PAK family directly interact with the active forms of Rac1 and CDC42 , and Rac1 and RhoA engage in negative cross-talk with each other . Accordingly, we examined whether Fasudil activates Rac1. As shown in Figure 5G, the activity of Rac1 was not altered by Fasudil, suggesting that Rac1 does not directly activate PAK4 in SVECs.
PAK4 is required for phosphorylation of MLC and promotion of cell spreading by ANP.
CCM2 is required for phosphorylation of MLC and enhancement of cell spreading by ANP
In endothelial cells, RhoA activity is suppressed, in order to maintain barrier function, by the scaffold proteins CCM1, -2, and -3 [15–17]. To determine whether CCM proteins are involved in ANP signaling, we examined the effect of CCM2 depletion on the ANP-mediated phosphorylation of PAK4 and MLC. As shown in Figure 6A,B, knockdown of CCM2 inhibited the phosphorylation of MLC in response to ANP. ANP-induced cell spreading was also inhibited by knockdown of CCM2 (Figure 6C). Consistent with this, the decrease in capacitance and increase in resistance caused by ANP were suppressed (Figure 6D,E). These results indicate that CCM2 mediates ANP/GC-A signaling.
CCM2 is required for phosphorylation of MLC and promotion of cell spreading by ANP.
Both the GC and KHDs are required for proper ANP/GC-A signaling
Next, we sought to determine which domain of GC-A is important for ANP/GC-A signaling. To this end, we introduced point or deletion mutations into GC-A to obtain gain- or loss-of-function mutants. The resultant mutants were transiently expressed in HEK293T cells to examine production of cGMP and phosphorylation of Zyxin and VASP (Figure 7B,C). SVECs were used to examine the phosphorylation of MLC, because phosphorylation of MLC by ANP can be clearly observed in these cells (Figure 7D,E). The domain structure of GC-A and the sites of point mutations are shown in Figure 7A. The D925A mutation in GCD abolishes GC activity without affecting homodimerization . As shown in Figure 7B–E, the D925A mutation suppressed ANP-mediated cGMP production and phosphorylation of Zyxin and VASP in HEK293T cells and MLC in SVECs, indicating that cGMP production is required for ANP signaling. Deletion of the KHD increases basal GC activity, and the deletion mutant produces cGMP in the presence or absence of ANP (Figure 7B) . As expected, deletion of KHD resulted in constitutive phosphorylation of VASP at Ser239 (Figure 7C). In contrast, deletion of KHD suppressed phosphorylation of Zyxin and VASP at Ser157 (Figure 7C), suggesting that the KHD is required for the phosphorylation of Zyxin and VASP at Ser157. Recent studies described gain-of-function mutations in GC-B, a receptor for C-type natriuretic peptide. Both the V883M mutation in the GC-B GCD, corresponding to V898M in GC-A, and the R655C mutation in the GC-B KHD, corresponding to R671C in GC-A, increase ligand-dependent GC activity, resulting in skeletal overgrowth, although a patient with the R655C mutation exhibits milder phenotypes [46,47]. As expected, the V898M mutation of GC-A increased ANP-mediated cGMP production (Figure 7B) and phosphorylation of Zyxin and VASP (Figure 7C). In contrast, the R671C mutation suppressed ANP-mediated cGMP production and phosphorylation of Zyxin and VASP at Ser157, as in the case of the ΔKHD mutation (Figure 7B,C). Similar to the D925A mutation, the R671C mutation suppressed the phosphorylation of MLC in response to ANP (Figure 7D,E). These results indicate that both the GCD and KHD domains of GC-A are required for full activation of downstream molecules. In SVEC, the D925A mutation inhibited ANP-mediated promotion of cell spreading (Figure 7F), decrease in capacitance (Figure 7G), and increase in resistance (Figure 7H). The R671C mutation suppressed membrane blebbing in the absence of ANP and substantially suppressed the ANP-mediated promotion of cell spreading (Figure 7F), decrease in capacitance (Figure 7G), and increase in resistance (Figure 7H). Collectively, our results demonstrate the importance of the KHD in ANP/GC-A signaling.
The KHD of GC-A is required for ANP/GC-A signaling.
Interactions of GC-A, CCM2, and PAK4
Because CCM2 acts as an adaptor protein, we investigated whether CCM2, GC-A, and PAK4 interact. In SVECs stably expressing GC-A and CCM2, CCM2 interacted with GC-A and endogenous PAK4 in an ANP-dependent manner (Figure 8A). The R671C mutation in GC-A abolished the ANP-mediated interaction of GC-A and CCM2 (Figure 8B). Knockdown of PAK4 slightly suppressed the interaction of GC-A and CCM2 (Figure 8C), whereas knockdown of CCM2 almost completely suppressed the interaction of GC-A and PAK4 (Figure 8D), suggesting that CCM2 acts as a linker between GC-A and PAK4.
ANP-dependent interactions of GC-A, CCM2, and PAK4 in SVECs.
Co-localization of GC-A, CCM2, and PAK4 at membrane ruffles
We next investigated the intracellular localizations of GC-A, CCM2, and PAK4. In the absence of ANP, GC-A localized at the surface of membrane blebs (Figure 9A), whereas in the presence of ANP, GC-A localized at membrane ruffles (Figure 9A). Consistent with this, ANP enhanced co-localization of GC-A with CCM2 at membrane ruffles (Figure 9B and Supplementary Movies S4 and S5). As reported in ref. , PAK4 localizes both at the cytoplasm and the nucleus (Figure 9C). In the presence of ANP, PAK4 also co-localized with GC-A at membrane ruffles (Figure 9C). Co-localization analysis revealed that CCM2, rather than PAK4, co-localized with GC-A at membrane ruffles (Figure 9E). The R671C mutation caused aggregation of GC-A in the perinuclear region and inhibited co-localization of CCM2 and GC-A in response to ANP (Figure 9D,E).
ANP-dependent co-localization of GC-A, CCM2, and PAK4 at membrane ruffles in SVECs.
Involvement of PKGI in ANP-induced phosphorylation of MLC
PKGI and II are direct effectors of cGMP. PKGI is predominantly expressed in endothelial cells . Knockdown of PKGI in SVEC/GC-A resulted in suppression of MLC phosphorylation and cell spreading (Supplementary Figure S1), indicating that PKGI is involved in the ANP-induced phosphorylation of MLC.
Effects of ANP on cell morphology and actin cytoskeleton in primary bovine endothelial cells
Finally, we investigated whether ANP exerts similar effects on cell morphology and the actin cytoskeleton in primary endothelial cells. For this purpose, we chose BPAECs because they produce a higher level of cGMP in response to ANP than HPAECs or SVECs (Figures 1A and 10A). ANP treatment induced phosphorylation of Zyxin, VASP, and MLC in BPAECs (Figure 10B) and also promoted cell spreading (Figure 10C). Knockdown of PAK4 or CCM2 inhibited the ANP-dependent phosphorylation of MLC (Figure 8D,E,G,H,I). Consistent with this, morphological changes in response to ANP were also inhibited in PAK4- or CCM2-knockdown cells (Figure 8F,J). These results indicate that ANP regulates cell morphology and the actin cytoskeleton in BPAECs as it does in SVEC/GC-A, although the effects of ANP are weaker in BPAEC than in SVEC/GC-A.
Effects of ANP on GC activity, phosphorylation of downstream effectors of GC-A, cell morphology, and the actin cytoskeleton in primary bovine endothelial cells.
ANP exerts profound effects on cardiovascular systems, primarily via activation of GC-A expressed on endothelial and smooth muscle cells. In the present study, we established immortalized mouse endothelial cells stably expressing GC-A, which exhibit robust responses to ANP, and used them to study GC-A signaling. We found that PAK4 and CCM2 act downstream of ANP/GC-A signaling and are required for promotion of cell spreading by activation of NM-II. Furthermore, we obtained similar results using bovine primary endothelial cells.
CCM1, -2, and -3 form a ternary complex and are thought to act as scaffolds in the same signaling pathway , suggesting that CCM1 and -3 are also involved in ANP/GC-A signaling. CCM1 interacts with Rap1, a small GTPase, which inactivates RhoA and enhances endothelial barrier function when activated by cAMP signaling [49–51]. Because ANP/GC-A requires the CCM pathway for signal transduction, activation of Rap1 might be an upstream event in the formation of the CCM complex and inactivation of RhoA. Although cAMP activates Rap1 through binding and activation of Epac, a Rap1-specific GEF, the affinity of Epac for cGMP is very low . Therefore, if Rap1 is activated by ANP/GC-A signaling, other GEFs may be involved.
In human primary endothelial cells, inhibition of the Rho/Rho-kinase pathway results in inhibition of MLC phosphorylation . In contrast, we observed enhancement of MLC phosphorylation by inhibition of Rho-kinase in SVEC, suggesting that other kinase(s) are activated by the inhibition of Rho-kinase and enhanced phosphorylation of MLC. We identified PAK4, a member of the PAK family, as the kinase responsible for phosphorylation of MLC by ANP. PAKs are further categorized into Group I (PAK1, -2, and -3) and Group II (PAK4, -5, and -6) based on their primary structures. The kinase activities of Group I PAKs are increased by binding to active Rac1 or Cdc42. Although Group II PAKs have high affinity for active Cdc42, but not for Rac1, their kinase activity is not increased by binding to Cdc42, suggesting the existence of other mechanisms for activation of PAK4 . In light of our observation that PAK4 is activated by a Rho-kinase inhibitor, it will be interesting to determine whether Rho-kinase directly phosphorylates unknown sites of PAK4 in order to inactivate it. Although the molecular mechanism of PAK4 activation is currently obscure, it is known that PAK4 is highly expressed in endothelium and required for normal formation of blood vessels during development . Characterization of the roles of PAK4 in endothelial cells will contribute to our understanding of the mechanisms of blood-vessel formation during development.
We found that deletion of the KHD or introduction of R671C mutation in the KHD inhibited phosphorylation of Zyxin and VASP at Ser157 by ANP, whereas VASP was still phosphorylated at Ser239, suggesting the importance of the KHD in ANP/GC-A signaling. Paradoxically, the KHD of GC-A acts as a repressor of GC activity, and its deletion results in cGMP production in the presence or absence of ANP , suggesting that cGMP production is not sufficient for activation of downstream molecules, and that KHD has another function in addition to regulation of GC activity. We found that GC-A interacts with CCM2 and PAK4 in response to ANP, and that the R671C mutation in the KHD suppressed the interaction of GC-A and CCM2 (Figure 8), indicating that KHD is important for formation of the signaling complex. Identification of other GC-A-interacting proteins will improve our understanding of ANP/GC-A signaling.
The R655C mutation in the GC-B KHD, corresponding to R671C in GC-A, increases ligand-dependent GC activity . In contrast, the R671C mutation in GC-A decreased ligand-dependent GC activity (Figure 7). SVECs expressing the R671C mutant exhibited slight spreading morphology in the absence of ANP (Figure 7). Thus, the R671C mutant may have an unidentified activity that regulates cell morphology, independently of GC activity.
Using FlincG3, we were able to show that cGMP is initially produced at the plasma membrane (Figure 1D), although it is reported that cGMP is produced at endosomes in HEK293T cells . Compared with FRET-based biosensors which contain two different fluorescent proteins, FlincG3 contains only one fluorescent protein. Therefore, FlincG3 is supposed to be less toxic to cells and useful for long-time live imaging. However, as long as tested, cells expressing FlincG3 did not spread as well as non-transfected cells by ANP (Figure 1D). FlincG3 may inhibit signals downstream of cGMP.
In conclusion, we identified PAK4 and CCM2 as new components of ANP/GC-A signaling, which regulates the actin cytoskeleton and promotes cell spreading via regulation of NM-II activity (Figure 11). Furthermore, our results suggest that dysfunction of the endothelial barrier and hemorrhage in patients with CCMs are caused by inhibition of proper ANP/GC-A signaling due to loss of CCM proteins. Future work aimed at elucidating the signal transduction mechanism of ANP/GC-A signaling may contribute to our understanding of the mechanisms that cause CCMs and provide new therapeutic targets in addition to Rho-kinase.
A proposed model for the enhancement of cell spreading and barrier function by the ANP/GC-A system.
atrial natriuretic peptide
bovine pulmonary artery endothelial cell
cerebral cavernous malformations
electric cell-substrate impedance sensing
guanylyl cyclase A
guanylyl cyclase domain
guanine nucleotide exchange factor
human pulmonary artery endothelial cell
kinase homology domain
myosin light chain
myotonic dystrophy kinase-related CDC42-binding kinase
p21 protein-activated kinase
protein kinase G I
SV40-transformed mouse endothelial cell
K.M., T.N., K.A., M.M., H.H., and K.K. designed the experiments. K.M., Y.A., M.Z., J.H., and T.K. performed the experiments. S.F. provided unpublished materials. K.M. wrote the paper with input from K.A. and S.F.
This work was supported in part by grants from the Ministry of Education, Science, Sports, and Culture of Japan (to H.H.); the Ministry of Health, Labour, and Welfare of Japan (to K.K.); and the Takeda Science Foundation (to K.K.).
We thank N. Mochizuki, M. Masuda and M. Kumazoe (National Cerebral and Cardiovascular Center Research Institute), J. Garthwaite (University College London), K. Takeya (Asahikawa Medical University), and Y. Takuwa (Kanazawa University) for reagents and helpful suggestions; and H. Shimamoto, N. Mochizuki, M. Nakamura, M. Fukui, and C. Nakaya for technical assistance.
The Authors declare that there are no competing interests associated with the manuscript.