Endothelial Angiogenesis and Barrier Function in Response to Thrombin Require Ca2+ Influx through the Na+/Ca2+ Exchanger*

Background: Thrombin stimulates protease-activated receptors (PAR) on endothelial cells, and this pathway becomes dysregulated in disease. Results: Ca2+ influx through the Na+/Ca2+ exchanger (NCX) regulated angiogenesis and endothelial permeability in response to thrombin via reactive oxygen species generation and ERK1/2 activation. Conclusion: NCX activity is a novel determinant of thrombin signaling in the endothelium. Significance: Inhibiting NCX could improve conditions involving unregulated thrombin signaling. Thrombin acts on the endothelium by activating protease-activated receptors (PARs). The endothelial thrombin-PAR system becomes deregulated during pathological conditions resulting in loss of barrier function and a pro-inflammatory and pro-angiogenic endothelial phenotype. We reported recently that the ion transporter Na+/Ca2+ exchanger (NCX) operating in the Ca2+-influx (reverse) mode promoted ERK1/2 activation and angiogenesis in vascular endothelial growth factor-stimulated primary human vascular endothelial cells. Here, we investigated whether Ca2+ influx through NCX was involved in ERK1/2 activation, angiogenesis, and endothelial barrier dysfunction in response to thrombin. Reverse-mode NCX inhibitors and RNAi-mediated NCX1 knockdown attenuated ERK1/2 phosphorylation in response to thrombin or an agonist of PAR-1, the main endothelial thrombin receptor. Conversely, promoting reverse-mode NCX by suppressing Na+-K+-ATPase activity enhanced ERK1/2 activation. Reverse-mode NCX inhibitors and NCX1 siRNA suppressed thrombin-induced primary human vascular endothelial cell angiogenesis, quantified as proliferation and tubular differentiation. Reverse-mode NCX inhibitors or NCX1 knockdown preserved barrier integrity upon thrombin stimulation in vitro. Moreover, the reverse-mode NCX inhibitor SEA0400 suppressed Evans' blue albumin extravasation to the lung and kidneys and attenuated edema formation and ERK1/2 activation in the lungs of mice challenged with a peptide activator of PAR-1. Mechanistically, thrombin-induced ERK1/2 activation required NADPH oxidase 2-mediated reactive oxygen species (ROS) production, and reverse-mode NCX inhibitors and NCX1 siRNA suppressed thrombin-induced ROS production. We propose that reverse-mode NCX is a novel mechanism contributing to thrombin-induced angiogenesis and hyperpermeability by mediating ERK1/2 activation in a ROS-dependent manner. Targeting reverse-mode NCX could be beneficial in pathological conditions involving unregulated thrombin signaling.

The Na ϩ /Ca 2ϩ exchanger (NCX) is a plasmalemma protein that extrudes cytosolic Ca 2ϩ in exchange for extracellular Na ϩ (forward mode) or, in the reverse mode, leads to Ca 2ϩ influx, depending on transmembrane Na ϩ and Ca 2ϩ concentrations and membrane potential (12). The mammalian NCX family comprises three genes (NCX1-3) that are expressed in a tissuespecific manner (12). Under physiological conditions, NCX operates primarily in the Ca 2ϩ -extrusion mode maintaining a low cytosolic Ca 2ϩ concentration ([Ca 2ϩ ] i ) (12). However, in conditions that promote Na ϩ overload and/or membrane depolarization (such as ischemia-reperfusion or heart disease), NCX can reverse its action leading to Ca 2ϩ influx exacerbating the severity of the disease (13). Consequently, reverse-mode NCX is considered a therapeutic target for cardiovascular diseases, and pharmacological inhibitors have been developed (14).
The role of NCX is mainly studied in excitable cells (12), although it is also expressed in nonexcitable cells such as kidney tubular cells, ␤ cells, and glial cells (12). ECs also exhibit NCX activity (15,16) and primarily express the splice variants of the cardiac isoform NCX1.3 and NCX1.7 (17). The role of endothelial NCX in physiology and pathophysiology remains largely unknown.
We recently reported that ion channel activity and in particular Ca 2ϩ influx through reverse-mode NCX were required for ERK1/2 activation and angiogenesis in human ECs stimulated with vascular endothelial growth factor (VEGF) (18,19). Thrombin also activates ERK1/2 in a Ca 2ϩ -dependent manner (20). Therefore, we investigated whether Ca 2ϩ influx through reverse-mode NCX could modulate ERK1/2 activation downstream of thrombin and its main endothelial receptor PAR-1 and whether NCX activity is required for thrombin-induced angiogenesis and endothelial barrier disruption.

Experimental Procedures
Materials-Primary HUVECs were obtained from TCS Cellworks. Thrombin from human plasma was from Sigma. The PAR-1 peptide agonist TFLLR-NH 2 was from Tocris. SEA0400 was synthesized initially by Taisho Pharmaceuticals (Saitama, Japan) and later (for in vivo studies) at the Institute for Cancer Research by O. B. SN-6, BAPTA-AM, Gö6983, DPI, U73122, and SKF96365 were from Tocris. N-Acetyl-L-cysteine, EGTA-AM, and PD98059 were from Calbiochem. The Nox2 inhibi-tory peptide gp91ds-tat and its scrambled control were from Anaspec. KB-R7943, ouabain, L-NAME, ionomycin, and human VEGF-A were from Sigma. All other chemicals used were from Sigma unless otherwise indicated.
Cell Culture and Protein Extraction-HUVECs (TCS Cellworks) were cultured in complete large vessel EC growth medium with the addition of endothelial growth supplements and antibiotics (TCS Cellworks) as we described previously (18,19). For the thrombin stimulation assays, cells, washed twice with PBS, were subsequently serum-starved in a physiological buffer containing 144 mM NaCl, 5.4 mM KCl, 2.5 mM CaCl 2 , 1 mM MgCl 2 , 5.6 mM D-glucose, and 5 mM Tris-HCl, pH 7.4, at 37°C for 1 h. Inhibitors or vehicle was added 30 min prior to stimulation with thrombin, unless otherwise indicated. HUVECs were activated by the addition of 0.5 units/ml thrombin and maintained for the indicated times at 37°C. Cells were subsequently washed once with ice-cold PBS, placed on ice, and lysed by applying an ice-cold buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% v/v Triton X-100, 0.5 mM DTT, 1 mM PMSF, 1% v/v protease inhibitor mixture (Sigma), 1 mM NaF, 5 mM bpVphen (Calbiochem), 5 M fenvalerate (Calbiochem), 1 mM Na 3 VO 4 , and 1% v/v phosphatase inhibitor mixtures I and III (Sigma). The cell homogenate was placed for 10 min on ice and then centrifuged at 5,000 ϫ g for 10 min at 4°C. The supernatant was aliquoted and stored at Ϫ80°C until further use. The bicinchoninic acid assay (BCA) (Pierce), with BSA (Sigma) as the protein standard, was used for determining protein concentration. In all the experiments described, HUVECs were used between passages 3 and 9.
Western Blot-Western blotting was performed as we have described previously (18,19), using the NuPAGE electrophoresis system and buffers (Invitrogen). Protein bands were visualized using the ECL TM prime detection kit (GE Healthcare). In this study, we used the following antibodies: rabbit anti-phospho-p44/p42 ERK1/2 (Thr-202/Tyr-204); rabbit anti-p44/p42 (total ERK); rabbit anti-phospho-PLC␤3 (Ser-537); and peroxidase-conjugated secondary antibodies from Cell Signaling. The mouse anti-NCX1 was from Swant; mouse anti-␤-actin was from Sigma. Mouse anti-GAPDH, rabbit anti-gp91 Phox (Nox2), and rabbit anti-Nox5 were from Abcam Labs. Films were scanned, and the optical densities of bands of interest were determined using ImageJ 1.46r (National Institutes of Health). Phosphoprotein optical density was normalized against the corresponding protein loading controls. The ratio of phospho/ total protein of the unstimulated controls in each experiment was arbitrarily set as 1. The values of the experimental conditions are represented as fold ϫ normalized unstimulated controls from at least three independent experiments.
Immunoprecipitation-Immunoprecipitation was performed as we described previously (18,19). Briefly, protein A-agarose beads (Roche Applied Science), washed twice with ice-cold PBS, were incubated with 10 l of anti-NCX1 antibody for 1 h at 4°C. Subsequently, the antibody-conjugated beads were washed twice in ice-cold PBS and added to the corresponding cell lysates. Following overnight incubation at 4°C, the beads were washed twice with lysis buffer and once with PBS, and proteins were extracted by boiling in sample buffer at 95°C for 5 min.
Ca 2ϩ Assays-Ca 2ϩ assays were performed essentially as described previously (18,19). Briefly, HUVECs (1 ϫ 10 4 cells/ well), were seeded into a flat clear-bottomed black-walled 96-well plate (Corning Glass). The following day, cells were washed twice with PBS and loaded with the Ca 2ϩ dye indicator Fluo-4NW (Life Sciences) for 45 min at 37°C in the dark in Hanks' balanced salt solution (HBSS) in the presence of probenecid (2.5 mM) to enhance dye loading. Subsequently, equal volumes of HBSS containing inhibitors or vehicle were added, and cells were incubated for a further 15 min. The plate was then transferred to the assay chamber of a FLIPR plate reader (Molecular Devices), and HUVECs were challenged with 0.5 units/ml thrombin in HBSS for 200 s at 37°C. Fluorescence intensity (excitation 485 nm, emission 525 nm, cutoff 515 nm), as a measure of [Ca 2ϩ ] i , was monitored immediately after the addition of thrombin every 2 s. Uniform sample loading was determined by an end point fluorescence reading (excitation 485 nm and emission 525 nm) immediately before thrombin stimulation. Background fluorescence was measured for 20 s prior to the addition of thrombin, and Ca 2ϩ transients were presented as the ratio of sample fluorescence at any given time point (F) divided by background fluorescence (F 0 ). All of the experimental conditions were in triplicate and were assayed simultaneously on the same plate.
Tubulogenesis Assays-HUVEC tubular differentiation assays were carried out according to a previously published method (21) with minor modifications. HUVECs (2.5 ϫ 10 3 per sample) in phenol red-free Opti-MEM (Life Sciences) containing 2% FCS were plated in 24-well plates precoated with 250 l of growth factor-reduced, phenol red-free Matrigel TM (BD Biosciences) for 1 h at 37°C. Where indicated, 0.5 units/ml thrombin and the appropriate concentrations of reverse-mode NCX inhibitors or vehicle were added. The cells were incubated in a humidified incubator at 37°C and 5% CO 2 . After 16 h, bright field images were captured with an Olympus LX70 microscope from five random fields of view per sample at ϫ10 magnification. HUVECs transfected with NCX1 siRNA for 48 h were plated on Matrigel TM as above, but tubule length was quantified after 6 h of thrombin addition. Tubule length and number of branching points for each condition were quantified using Image Pro-Plus 5.0 software (Media Cybernetics).
Proliferation Assays-96-Well plates (Nunc) were precoated with 50 g/ml rat tail collagen-I (BD Biosciences) for 2 h at 37°C. Subsequently, 3 ϫ 10 3 HUVECs were seeded per well in complete EC medium (TCS Cellworks). The next day, the medium was discarded, and 100 l of phenol red-free Opti-MEM (Life Sciences) medium containing 0.1% (w/v) BSA and 1 g/ml hydrocortisone were added. Where appropriate, the medium contained 0.5 units/ml thrombin and the indicated concentrations of reverse-mode NCX inhibitors or vehicle. Each condition was assayed in triplicate. After 48 h in a humidified incubator at 37°C and 5% CO 2 in air, the number of cells in each well was determined by the alkaline phosphatase assay as we described previously (18,19).
In Vitro Endothelial Barrier Function Assays-Endothelial barrier function in vitro was assessed by determining the passage of FITC-labeled dextran through a confluent endothelial monolayer according to a published method (22) with minor modifications. HUVECs (1 ϫ 10 4 ) were seeded into modified Boyden transwell inserts (Costar, 0.4 m pore size) precoated with 50 g/ml rat tail collagen-I (BD Biosciences) and cultured in complete medium for 2 days until confluent. Subsequently, HUVECs monolayers were serum-starved in physiological buffer (described under "Cell Culture and Protein Extraction") for 1 h. Appropriate concentrations of drugs or vehicle were added into the top chamber 30 min prior to thrombin stimulation. Thrombin (0.5 units/ml) was applied to the top chamber simultaneously with 1 mg/ml FITC-dextran (M r 40,000) (Sigma). Fluorescence in the bottom chamber of each sample was measured 30 min after thrombin stimulation in an Infinite M200PRO (Tecan) fluorescent plate reader (excitation 485 nm and emission 525 nm). For the siRNA experiments, HUVECs transfected with NCX1 targeting or control siRNA, as described under "siRNA Transfection," were seeded into collagen-I-coated transwell inserts at a density of 2 ϫ 10 4 cells/well, 24 h post-transfection. The following day, endothelial barrier function in response to thrombin was determined as above.
Immunofluorescence-HUVECs plated on collagen-I (BD Biosciences)-coated glass coverslips and cultured until 80 -90% confluent were serum-starved in physiological buffer, preincubated for 30 min with inhibitors or vehicle, and stimulated with thrombin (0.5 units/ml) for 30 min. Cells were washed with prewarmed PBS and fixed for 20 min with 4% w/v paraformaldehyde at room temperature. After an additional wash step with PBS, fixed HUVECs were permeabilized with 0.5% v/v Triton X-100 in PBS for 5 min at room temperature. Subsequently, cells were washed in PBS, blocked in 0.5% w/v BSA in PBS, and incubated for 15 min at room temperature with FITC-conjugated phalloidin (1:40 v/v) in PBS containing 0.1% v/v BSA. Cells were then washed twice with PBS and once with PBS containing 0.025% v/v Tween 20. Finally, HUVECs were washed for 10 min in PBS containing 1:10,000 v/v DAPI (Invitrogen) to visualize nuclei, and after two wash steps in PBS, slides were air-dried in the dark and mounted on glass slides with ProLong Gold anti-fade (Invitrogen). Images were captured with a Hamamatsu digital camera (Improvision) on a Zeiss Axioplan microscope at ϫ40 magnification.
RhoA Activity Assay-RhoA activity was determined with the ELISA-based kit G-LISA RhoA (Cytoskeleton) according to the manufacturer's instructions. Serum-starved HUVECs (1 ϫ 10 5 /sample) were preincubated with inhibitors or vehicle for 30 min and then stimulated with thrombin (0.5 units/ml) for the indicated times. Cells were then lysed with the lysis buffer provided in the kit. Cell lysates were precleared by centrifugation (5000 ϫ g for 1 min at 4°C). A small sample was used for protein concentration determination with the BCA assay, and the remaining supernatant was snap-frozen in liquid N 2 and stored at Ϫ80°C until used for the assay. Absorbance was corrected for protein concentration.
ROS Assays-For DCF, HUVECs were plated at a density of 10 4 cells per well in clear bottom black 96-well plates (Corning). The following day, the medium was aspirated, and cells were serum-starved in physiological buffer (described under "Cell Culture and Protein Extraction") in a humidified incubator at 37°C, 5% CO 2 . After 30 min, the appropriate inhibitors or vehicle was added, and HUVECs were serum-starved for a further 30 min at 37°C. Next, the medium was removed, and fresh medium, containing 10 M DCF, was added for 10 min at 37°C in the dark. Finally, the DCF solution was removed; cells were washed twice with prewarmed buffer and stimulated with 0.5 units/ml thrombin in the presence of inhibitors or vehicle for 5 min at 37°C. Fluorescence was measured immediately in an Infinite M200PRO (Tecan) fluorescent plate reader (excitation 485 nm and emission 525 nm). All the buffers contained 0.2 mM L-NAME to prevent peroxynitrite formation, which is reported to interfere with DCF fluorescence (23).
For DHE fluorescence, HUVECs, plated at a density of 10 4 cells per well in clear bottom black 96-well plates (Corning Glass) were serum-starved the following day in physiological buffer for 1 h. Appropriate inhibitors or vehicle was added 30 min prior to stimulation with a solution containing thrombin and DHE. The final concentrations of thrombin and DHE were 0.5 units/ml and 5 M, respectively. After 10 min, fluorescence was measured in an Infinite M200PRO (Tecan) fluorescent plate reader (excitation 520 nm and emission 590 nm).
MCLA chemiluminescence measurements were performed as described previously (24). Briefly, HUVECs (10 4 cells/well) were plated in clear bottom black 96-well plates (Corning Glass). The following day, cells were serum-starved in physiological buffer for 1 h. The appropriate inhibitors or vehicle were added for the last 30 min of the starvation step. HUVECs were then stimulated with a solution of 0.5 units/ml thrombin containing 1 M MCLA, and the plate was immediately transferred to the measuring chamber of an Infinite M200PRO plate reader (Tecan). The photons emitted from each sample, including a blank sample containing 1 M MCLA, were measured simultaneously, for 2 s/min for a total of 10 min. The value of the blank sample was subtracted from each sample, and emitted photons were expressed as percentage of the unstimulated control (arbitrarily set to 100%) corrected for protein concentration. In the case of siRNA-transfected HUVECs, cells were plated 24-h post-transfections and assayed the following day as above.
In Vivo Assays-All animal experiments were conducted in accordance with the United Kingdom Home Office Animals (1986 Scientific Procedures) with local ethical committee approval. Vascular permeability in response to a specific peptide agonist of PAR-1 (TFLLR-NH 2 ) was evaluated according to previously published methods (25,26), with minor modifications. Male C57BL/6J mice (8 -10 weeks old) received 10 mg/kg SEA0400 dissolved in 4% Arabic gum (Sigma) or vehicle by oral gavage, as described previously (27). After 2 h, the conscious animals were placed in a restrainer, and 20 mg/kg Evans blue albumin (EBA) with or without 1 mg/kg PAR-1 peptide agonist, as indicated, in a 100-l final volume, was injected through the tail vein. After 30 min, animals were terminally anesthetized and sacrificed by cardiac puncture after thoracotomy. Blood was collected in heparinized syringes, and EBA in the plasma was evaluated against an EBA standard curve to ensure comparable EBA loading. Subsequently, mice were perfused with 12 ml of ice-cold sterile PBS containing 5 mM EDTA (4 ml/min) to remove red blood cells. The lungs and the kidneys were removed, washed in cold PBS, dried with sterile blotting paper, weighed, snap-frozen in liquid nitrogen, and stored at Ϫ80°C until analyzed as described below.
EBA Tissue Leakage-Right superior lung lobes or kidneys of mice treated as described above were homogenized in 1 ml of PBS per 0.1 mg of tissue using a Polytron homogenizer. Two volumes of formamide were subsequently added to the lysates, and EB dye was extracted from the tissues by incubation at 60°C for 18 h (25,26). The lysates were centrifuged for 30 min at 5,000 ϫ g in a benchtop centrifuge. The EBA content of the tissues was determined by measuring the absorbance of the supernatant at 620 nm corrected at 740 nm for hemoglobin content in a plate reader (Infinite M200PRO, Tecan) against standards of known EBA content.
Lung Edema Formation-Left lungs from the animals used for EBA determination were dried by incubation in an oven for 16 h at 60°C. The ratio of wet tissue weight (obtained immediately after the harvest of tissues) divided by the dry tissue weight was used as an indication of the water content (and hence edema formation) of the lungs (25,26).
Lung Western Blot-The middle and inferior right lobes of experimental animals were homogenized in mammalian protein extraction buffer (GE Healthcare) containing protease and phosphatase inhibitors. Protein homogenization of tissues and Western blotting were performed as we described previously (28).
Statistical Analysis-The data are expressed as the means Ϯ S.E. Statistical significance was determined by one-or two-way analysis of variance and Tukey's post hoc test using GraphPad Prism 5 software, as appropriate. Values of p Ͻ 0.05 were deemed statistically significant.
HUVEC stimulation by thrombin in a nominal Ca 2ϩ -free physiological medium (Ca 2ϩ was omitted) resulted in a substantial reduction in phospho-ERK1/2 levels compared with controls ( Fig. 1C). Thrombin-induced ERK1/2 activation was also inhibited by the nonselective inhibitor of Ca 2ϩ influx SKF96365 (Fig. 1D). Neither of these treatments visibly altered pPLC␤3 levels, suggesting that Ca 2ϩ influx does not affect PAR activity. Thus, extracellular Ca 2ϩ was required for ERK1/2 phosphorylation in response to thrombin, in agreement with a previous study (20).

FIGURE 1. Ca 2؉ influx through reverse-mode NCX is required for thrombin-induced ERK1/2 activation.
A, HUVECs, serum-starved for 40 min, were treated with BAPTA-AM (10 M) or vehicle for a further 20 min prior to challenge with thrombin (0.5 units/ml) for 5 min. ERK1/2 and PLC␤3 activation were demonstrated by Western blot. Membranes were subsequently stripped and re-probed for total ERK1/2 and GAPDH proteins. B, serum-starved HUVECs were preincubated with the PLC inhibitor U73122 (1 M) for 30 min prior to challenge with thrombin (0.5 units/ml) for 5 min. C, HUVECs were serum-starved for 45 min and then incubated for a further 15 min in Ca 2ϩ -free medium, prior to challenge with 0.5 units/ml thrombin (5 min). ERK1/2 and PLC␤3 activation was determined as in A. D, HUVECs were incubated with the broad-spectrum inhibitor of nonselective cation channels SKF96365 (30 M) and treated as in B. E, serum-starved HUVECs were preincubated with the reverse-mode NCX inhibitor SEA0400 (1 M) or vehicle for 30 min prior to challenge with thrombin (0.5 units/ml) for the times indicated. F, SEA0400 was also applied over a range of concentrations, and ERK1/2 and PLC␤3 activation was assayed after 5 min. G, optical densities of the phospho-ERK1/2 and phospho-PLC␤3 bands in F were normalized against the ERK1/2 total protein of the corresponding sample. The normalized density of the unstimulated control (bar C) was set to 1. The mean value of the ratio for each condition is expressed as fold ϫ unstimulated control value. Bars represent the means Ϯ S.E. from n ϭ 3 experiments. *, p Ͻ 0.05 versus the thrombin-stimulated control. A second reverse-mode NCX inhibitor, SN-6, also suppressed thrombin-induced ERK1/2 activation in a dose-dependent (H) and time-dependent manner (I). HUVECs were stimulated with ionomycin (1 M for 5 min) in the presence of SEA0400 (1 M), SN-6 (10 M), KB-R7943 (10 M), or vehicle (J), and ERK1/2 activation was assessed by Western blot. n ϭ 3.

Effect of Reverse-mode NCX Inhibitors on Thrombin-induced Ca 2ϩ
Transients-Next, we investigated the effect of reversemode NCX inhibitors on thrombin-induced Ca 2ϩ transients. HUVECs were serum-starved for 1 h in physiological buffer in the presence of probenecid (2.5 mM). With the exception of probenecid (included for improved dye loading) conditions were similar to the ERK1/2 stimulation experiments to allow for direct comparisons (see under "Experimental Procedures").
Thrombin stimulation of HUVECs loaded with the Ca 2ϩ dye indicator Fluo-4NW resulted in a rapid increase in emitted fluorescence. The Ca 2ϩ response peaked at ϳ20 -30 s and then declined toward the baseline within 120 s. Steady-state bulk [Ca 2ϩ ] i remained slightly higher than baseline for the duration of the experiment (200 s). Preincubation of HUVECs with the reverse-mode NCX inhibitors KB-R7943 (10 M), SN-6 (10 M), or SEA0400 (1 M) for 15 min prior to stimulation resulted in a modest decrease to the tonic phase of the Ca 2ϩ response (Fig. 2, A-C). In contrast, there was a sharper decline and a lower steady state of the Ca 2ϩ transient in comparison with controls. The Ca 2ϩ response (quantified as area under the curve) was modestly, but significantly, diminished by pretreatment with KB-R7943 (10 M), SN-6 (10 M), and SEA0400 (1 M) by 17.8 Ϯ 2.96, 12.4 Ϯ 1.21, and 10.9 Ϯ 0.6%, respectively, compared with thrombin-stimulated control (arbitrarily set to 100%) (Fig. 2D).
The potential involvement of Ca 2ϩ microdomains was investigated using EGTA-AM, which does not affect local Ca 2ϩ transients within 100 nm of the source of Ca 2ϩ entry (30). Loading HUVECs with EGTA-AM had no apparent effect on ERK1/2 phosphorylation in response to thrombin (Fig. 2E). Conversely, when HUVECs were loaded, under similar conditions, with the "fast" Ca 2ϩ chelator BAPTA-AM, which suppresses localized Ca 2ϩ signals (22), ERK1/2 activation was inhibited (Fig. 1A).
Knockdown of NCX1 by siRNA Suppresses Thrombin-induced ERK1/2 Activation-Next, we genetically targeted expression of NCX1, the predominant NCX isoform in HUVECs (17,18). Transfection of HUVECs with NCX1-target-  Fig. 1  (n ϭ 3). G, optical densities of the phospho-ERK1/2 and phospho-PLC␤3 bands in F were quantified as in Fig. 1G. *, p Ͻ 0.05 versus the thrombin-stimulated control (n ϭ 3). H, NCX1 protein was immunoprecipitated from control-or NCX1-siRNA-treated cells in experiments run in parallel to those shown in F. NCX1 protein levels were determined by Western blot. Equal protein in the input was ensured by probing for ␤-actin. I, optical densities of the NCX1 bands in H (n ϭ 3). Control optical density was set to 100%. The optical density of the NCX1 protein bands from NCX1-siRNA treated cells is expressed as % change relative to the corresponding control value. Bars represent the means Ϯ S.E. from n ϭ 3 experiments. *, p Ͻ 0.05 versus the control.
ing siRNA significantly attenuated ERK1/2 activation in response to thrombin compared with control siRNA (15.71 Ϯ 2.14-fold down to 8.31 Ϯ 2.46-fold of unstimulated control; Fig.  2, F and G). Phospho-PLC␤3 levels were unaffected, indicating that knockdown of NCX1 does not affect PAR activity. Knockdown of NCX1 by ϳ50% at the protein level, as reported previously (18), in parallel transfection experiments (Fig. 2, H and I) was confirmed by Western blot.
Ouabain Augments ERK1/2 Activation by Thrombin-HUVECs were then treated with the Na ϩ -K ϩ -ATPase inhibitor ouabain, reported to rapidly increase [Na ϩ ] i in HUVECs and to promote reverse-mode NCX (31). Ouabain alone did not affect ERK1/2 phosphorylation for the duration of the experiment (3 min), as we have reported previously (18). However, preincubation of HUVECs with ouabain (100 M for 1min) prior to short term stimulation with thrombin (0.5 units/ml for 2 min) augmented ERK1/2 phosphorylation (13.3 Ϯ 2.15-to 20.10 Ϯ 2.54fold of unstimulated control). Preincubation of the thrombinstimulated, ouabain-treated cells with SN-6 suppressed ERK1/ 2 phosphorylation (2.4 Ϯ 0.71-fold versus the control) indicating that the effect of ouabain was probably due to the induction of reverse-mode NCX (Fig. 3, A and B).
Similarly, in HUVECs transfected with control siRNA (100 nM for 48 h) thrombin application (0.5 units/ml) for 6 h resulted in a significant increase in the length of EC tubules to 1.34 Ϯ 0.04-fold of the control value, set at 1 (n ϭ 4 in duplicate) (Fig.  5, F and G). Conversely, when HUVECs transfected with NCX1 targeting siRNA were stimulated with thrombin, we did not observe any significant increase in tubule length in comparison with the unstimulated control (0.98 Ϯ 0.09; n ϭ 4 in duplicate), whereas NCX1 knockdown did not have any discernible effects on tubule length at baseline condition (1.01 Ϯ 0.07-fold of unstimulated control; n ϭ 4 in duplicate).
Effect of Reverse-mode NCX on Thrombin-induced Endothelial Permeability-Because ERK1/2 activation has been implicated in EC barrier disruption in response to thrombin (35), we investigated the effect of reverse-mode NCX inhibitors on barrier function. Stimulation of confluent HUVEC monolayers on transwell inserts with thrombin resulted in an ϳ2-fold increase in permeability after 30 min (179.26 Ϯ 21.74% in comparison with control values arbitrarily set to 100%) (Fig. 6A), quantified as the fluorescence intensity of FITC-labeled dextran (M r 40,000) in the lower chamber. Preincubation of the HUVEC monolayer with SN-6 (10 M) or SEA0400 (1 M) resulted in a significant decrease in fluorescence compared with controls (98.49 Ϯ 1.22 and 125.46 Ϯ 12.24% fluorescence of unstimulated control, respectively) (Fig. 6A). The ERK1/2 pathway inhibitor PD98059 at a concentration (10 M) that abolished thrombin-induced ERK1/2 activation (Fig. 6B) also significantly suppressed EC permeability to 108.19 Ϯ 29.7 of control (Fig. 6A), in agreement with a previous report (35). SEA0400 (1 M) and SN-6 (10 M) did not significantly affect unstimulated HUVEC monolayer barrier function (Fig. 6C). Thus, our results suggest that reverse-mode NCX activity is required for thrombin-induced endothelial barrier disruption, potentially upstream of ERK1/2 activation. This was confirmed using NCX1 siRNA.
Thrombin stimulation of HUVECs transfected with nontargeting siRNA duplexes resulted in a 177.37 Ϯ 4.60% fluorescence increase compared with unstimulated controls (set to 100%). The permeability of the thrombin-stimulated NCX knockdown cells was reduced to 129.28 Ϯ 2.72% (Fig. 6D). Knocking down NCX1 in unstimulated HUVECs reduced permeability to 83.88 Ϯ 5.12% of control, but this was not statistically significant.
Effect of Reverse-mode NCX on Stress Fiber Formation and RhoA Activity in Response to Thrombin-In ECs, thrombin promotes stress fiber formation, resulting in contraction and the formation of inter-endothelial gaps (1, 2). Visualization of polymerized actin with FITC-phalloidin revealed a primarily peripheral distribution (Fig. 6E). Thrombin (0.5 units/ml, 30 min) induced marked morphological changes characterized by stress fiber formation, an apparent EC contraction, and the appearance of inter-endothelial gaps (Fig. 6E). Preincubation with SEA0400 or SN-6, did not apparently inhibit the formation of stress fibers; however, they appeared to remain at the cell periphery in association with more spread EC phenotypes and preserved cell-cell contacts (Fig. 6E).
The small Rho-GTPase RhoA is activated by thrombin and required for stress fiber formation (1, 2); moreover, Ca 2ϩ influx regulates RhoA activity (36). Exposure to thrombin for 5 min resulted in a 2-fold increase in RhoA activity (2.03 Ϯ 0.10) compared with controls (set to 1), gradually declining over time (Fig.  6F), similar to a previous study (37). SN-6 or SEA0400 had no significant effect on RhoA activity (Fig. 6, F and G, respectively). Thus, the differences in permeability following reverse-mode NCX inhibition were probably independent of RhoA activity.
Involvement of Reverse-mode NCX in ROS Generation-We have shown that Ca 2ϩ influx through reverse-mode NCX reg- . Reverse-mode NCX activity is required for ERK1/2 activation by a peptide agonist of PAR-1. A, HUVECs loaded with BAPTA-AM were challenged with the PAR-1 peptide agonist TFLLR-NH 2 (25 M for 5 min). ERK1/2 and PLC␤3 activation was determined as in Fig. 1A (n ϭ 3). B, HUVECs, preincubated with SEA0400, were stimulated with TFLLR-NH 2 (25 M for 5 min). ERK1/2 and PLC␤3 phosphorylation was determined as before (n ϭ 3). C, HUVECs were preincubated with SN-6 and stimulated with TFLLR-NH 2 for the indicated times. D, HUVECs transfected with NCX1 or control siRNA as in Fig. 2F were stimulated with TFLLR-NH 2 (25 M for 5 min); ERK1/2 and PLC␤3 activation was evaluated as in Fig. 1A. NCX1 protein knockdown for this experiment is shown in Fig. 2H, in cells transfected in parallel. E, HUVECs were treated with ouabain (100 M for 1 min) prior to stimulation with TFLLR-NH 2 (5 M for 2 min). Where indicated, samples were preincubated with SN-6 (10 M) for 30min. For A-F, n ϭ 3.
ulates VEGF-induced ERK1/2 activation by the targeting and subsequent activation of PKC␣ to the plasma membrane (18), and thrombin-induced PKC␣ activation has also been reported to require Ca 2ϩ influx and regulate barrier integrity (38). The broad spectrum PKC inhibitor Gö6983 had no discernible effect on thrombin-mediated ERK1/2 activation (Fig. 7A) but did inhibit VEGF-induced ERK1/2 activation in parallel experiments (data not shown), as we recently described (18). Thus, we infer that reverse-mode NCX possibly suppressed ERK1/2 activation independently of PKCs.
NADPH oxidase 2 (Nox2) activation and subsequent ROS production has been implicated in thrombin-induced angiogenesis (33). Thus, we investigated the effect of Nox inhibition on ERK1/2 phosphorylation by thrombin. The inhibitor of ROS formation by flavoenzyme DPI dose-dependently suppressed ERK1/2 activation by thrombin (Fig. 7B). Because DPI would also inhibit the endothelial nitric oxide synthase, L-NAME, a NOS inhibitor had no effect on phospho-ERK1/2 levels (Fig.  7C). Moreover, the ROS scavenger N-acetyl-L-cysteine (NAC), dose-dependently attenuated ERK1/2 phosphorylation (Fig.  7D), and gp91 ds-tat, a specific peptide inhibitor of Nox2 assembly, suppressed thrombin-induced ERK1/2 activation, although a scrambled peptide had no effect (Fig. 7E). To further corroborate the involvement of Nox2-generated ROS in thrombin-induced ERK1/2 activation in HUVECs, we knocked down the catalytic subunit of Nox2 gp91 Phox (Nox2 thereafter) with siRNA. Targeting Nox2 expression resulted in an ϳ50% decrease at the protein level (Fig. 7, F and G). Thrombin stimulation of HUVECs transfected with nontargeting control siRNA increased ERK1/2 phosphorylation by 10.05 Ϯ 2.9 when compared with the unstimulated control (set to 1), and this was reduced to 3.50 Ϯ 1.7 of control in cells transfected with Nox2 siRNA (Fig. 7, F and G).
The Ca 2ϩ -sensitive NADPH oxidase Nox5 has been implicated in ERK1/2 phosphorylation downstream of angiotensin-II (39). We attempted to investigate the role of Nox5 in thrombin-induced ERK1/2 activation by knocking down Nox5 with siRNA. However, although Nox5 siRNA had no effect on ERK1/2 phosphorylation (data not shown), we could not confirm Nox5 protein knockdown using a commercially available antibody; hence, the involvement of Nox5 in this pathway cannot be categorically excluded.
In the study of Montezano et al. (39), Ca 2ϩ influx through L-type voltage-gated Ca 2ϩ channels was required for ERK1/2 activation by angiotensin-II upstream of Nox5. Preincubation of HUVECs with the L-type voltage-gate Ca 2ϩ channel inhibi-tor nifedipine had no apparent effect on thrombin-induced ERK1/2 phosphorylation (Fig. 7H).
Next, we investigated the effect of SEA0400 (1 M) and SN-6 (10 M) on thrombin-induced ROS generation. Thrombin stimulation resulted in an increase in the emitted fluorescence of the ROS indicator DCF to 132 Ϯ 3.9% of the unstimulated control (set to 100%) per mg of protein (Fig. 8A). SEA0400 and SN-6 significantly attenuated ROS production in response to thrombin to 106.7 Ϯ 8.8 and 99.3 Ϯ 6.3% of control, respectively. The Nox inhibitor DPI (10 M) and the ROS scavenger NAC (1 mM) also suppressed DCF fluorescence to 87.0 Ϯ 8.5 and 77.0 Ϯ 19.5% of control value, respectively. Conversely, SEA0400 and SN-6 did not have a significant effect on DCF fluorescence under baseline conditions (Fig. 8B). Similar results were obtained with a second ROS indicator, DHE. Thrombin stimulation increased DHE fluorescence to 143.4 Ϯ 6.8% of . D, HUVECs transfected with control or NCX1 targeting siRNA were seeded on transwells, and after 24 h changes in permeability in response to thrombin were assessed as in A. E, subconfluent HUVEC monolayers on collagen-I coated coverslips were serum-starved, preincubated for 30 min with SEA0400 (1 M), SN-6 (10 M), or vehicle and stimulated with thrombin (0.5 units/ml) for 30 min. Cells were fixed with 4% paraformaldehyde, permeabilized, and incubated with FITC-conjugated phalloidin. Nuclei were stained with DAPI. Images were captured at ϫ40 magnification. HUVECs preincubated with SN-6 (10 M) (F) or SEA0400 (1 M) (G) were stimulated with thrombin (0.5 units/ml) as indicated. RhoA activity was determined with an ELISA-based kit. Absorbance was corrected for protein concentration and expressed as -fold control levels (set to 1). Bars are the means ϮS.E. *, p Ͻ 0.05 versus control (n ϭ 3). JULY 24, 2015 • VOLUME 290 • NUMBER 30 control (set to 100%) corrected for protein, and SEA0400, SN-6, DPI, or NAC preincubation suppressed DHE fluorescence to 96.9 Ϯ 11.6, 96.4 Ϯ 14.5, 105.8 Ϯ 3.3, and 106.1 Ϯ 9.4% of control, respectively (Fig. 8C). The involvement of reversemode NCX in thrombin-induced ROS production was further corroborated by the use of the lucigenin derivative MCLA that emits light when reacting with superoxide. Stimulation of HUVECs with thrombin resulted in an increase in the emitted chemiluminescence to 160.62 Ϯ 2.65% of control (arbitrarily set to 100%) corrected for protein and measured over 10 min. Preincubation of HUVECs for 30 min prior to thrombin stimulation with SEA0400 (1 M), SN-6 (10 M), or DPI (10 M) attenuated chemiluminescence to 89.41 Ϯ 23.76, 79.95 Ϯ 8.54, and 62.36 Ϯ 16.27% of control, respectively (Fig. 8D). As in the case of DCF, SEA0400 or SN-6 did not significantly affect MCLA chemiluminescence of nonstimulated serum-starved HUVECs (Fig. 8E). The involvement of NCX1 and Nox2 activities in thrombin-induced superoxide generation was further supported by the use of RNAi. Thrombin stimulation of HUVECs transfected with control siRNA resulted in an increase in the emitted chemiluminescence of 152.0 Ϯ 3.5% when compared with unstimulated controls (set to 100%).

Role of Reverse-mode NCX in Thrombin EC Signaling
Stimulation of HUVECs transfected with NCX1-or Nox2-targeting siRNA significantly reduced chemiluminescence to 107.8 Ϯ 6.2 and 105.2 Ϯ 8.3% of controls, respectively (Fig. 8D). Knockdown of NCX1 or Nox2 protein did not appear to have a significant effect on HUVEC baseline superoxide production (Fig. 8F). Taken together, our data suggest that Ca 2ϩ influx through reverse-mode NCX could be facilitating thrombin-induced ERK1/2 phosphorylation by modulating Nox2 activity and subsequent ROS production.
Reverse-mode NCX Inhibitor Preserves Endothelial Barrier Function in Vivo-Finally, we investigated the in vivo relevance of our cell-based data. C57BL/6 mice were treated with vehicle or 10 mg/kg SEA0400 by oral gavage for 2 h (27) prior to administration of the PAR-1-activating peptide (1 mg/kg) together with EBA (20 mg/kg) via the tail vein to measure vascular permeability according to published methods (38) with some modifications (see "Experimental Procedures"). PAR-1 activation significantly increased EBA extravasation from 1.46 Ϯ 0.20 and 3.54 Ϯ 0.14 g of EBA/g of wet tissue in the lungs and kidneys of control animals to 3.78 Ϯ 0.43 and 6.67 Ϯ 0.27 g as determined by formamide extraction. Conversely, SEA0400 decreased EBA to 2.12 Ϯ 0.27 and 4.47 Ϯ 0.21 g of EBA/g of FIGURE 7. ROS generation by NADPH oxidase 2 is required for ERK1/2 activation downstream of thrombin. A, HUVECs preincubated with the PKC inhibitor Gö6983 were stimulated with thrombin (0.5 units/ml) for 5 min. Serum-starved HUVECs were preincubated the inhibitor of NADPH oxidases DPI (B), the inhibitor of NOS L-NAME (1 mM) (C), the ROS scavenger NAC (D), or the peptide inhibitor of Nox2 assembly gp91 ds-tat (50 M), or its scrambled analogue (sc; E), prior to thrombin stimulation. F, HUVECs transfected with nontargeting or Nox2-targeting siRNA were serum-starved and then stimulated with thrombin. ERK1/2 activation and Nox2 protein levels were determined by Western blot (n ϭ 3). G, optical densities of the phospho-ERK1/2 and gp91 Phox (Nox2) bands in F were normalized against the ERK1/2 total protein of the corresponding sample. The normalized density of the unstimulated control (bar C) was set to 1. The mean value of the ratio for each condition is expressed as fold ϫ unstimulated control value. Bars represent the means Ϯ S.E. from n ϭ 3 experiments. #, p Ͻ 0.05 when compared with control Nox2 levels; *, p Ͻ 0.05 versus the thrombin-stimulated control. H, HUVECs were incubated with 1 M nifedipine for 30 min prior to stimulation with thrombin (0.5 units/ml) for 5 min (n ϭ 3).
Our findings suggest, in agreement with our in vitro studies, that reverse-mode NCX activity is required for enhanced vascular permeability in response to PAR-1 activation, possibly upstream of ERK1/2 activation.

Discussion
In this study, we report that Ca 2ϩ influx through reversemode NCX can modulate EC responses to thrombin downstream of PAR-1 by influencing ERK1/2 activation, as we described previously for VEGF (18).
The inhibitors used here are proposed to inhibit reversemode NCX by accelerating and stabilizing the Na ϩ -dependent inactivated state of the exchanger (14,40,41), and at the concentrations used, they are not expected to affect the forward mode of NCX. Nonetheless, although the more recent inhibitors SN-6 and SEA0400 are more selective for the reverse mode NCX (14), there is evidence that they could have off-target effects, because in mouse embryos where the NCX1 gene was deleted, SEA0400 suppressed to some extent Ca 2ϩ transients in cardiomyocytes (42).
To control for any off-target effects, we employed at least two chemically distinct inhibitors in each assay and used siRNA to knock down NCX1 in HUVECs. Nonetheless, knocking down NCX1 protein also abolishes the forward mode of the exchanger, and this could ultimately result in Ca 2ϩ overload and/or compensatory changes in the gene expression of other Ca 2ϩ -handling proteins. Consequently, we opted for the siRNA transfection conditions (100 nM siRNA for 48 h) that produced the clearest result in the functional assays (ERK1/2 activation, permeability, and tubulogenesis) with the minimum knockdown of NCX1, as we have optimized previously (18).
As a further control we showed that by using ouabain to inhibit the Na ϩ -K ϩ -ATPase and thus load HUVECs with Na ϩ , thrombin-induced ERK1/2 activation was enhanced, probably by promoting reverse-mode NCX (Fig. 3). Ouabain treatment in the absence of thrombin did not result in ERK1/2 activation (Fig. 3A), as we reported previously for VEGF-stimulated HUVECs (18). This is not surprising because human NCX1 requires an initial increase in [Ca 2ϩ ] i for activation (12). Consequently, simply loading the cells with Na ϩ in isolation is not expected to activate reverse-mode NCX. Ouabain cannot be used for longer term functional assays (permeability and angiogenesis) because Na ϩ overload would also affect other ionic mechanisms. Taken together, our findings suggest that reverse-mode NCX is implicated in thrombin-induced ERK1/2 activation, downstream of PAR-1, as we reported for VEGF-induced ERK1/2 activation and angiogenesis downstream of the receptor-tyrosine kinase VEGFR2 (18). In addition to our work in ECs, others have independently reported that reverse-mode NCX activity is required for ERK1/2 phosphorylation in cardiac fibroblasts (43) and neuroblastoma cells (44). Thus, reversemode NCX could modulate ERK1/2 activation downstream of multiple stimuli activating receptor-tyrosine kinases and G-protein-coupled receptors in a variety of cell types. Further work is required to support this exciting hypothesis and to elucidate the signaling pathways influenced by reverse-mode NCX in these diverse systems.
Role of Ca 2ϩ Microdomains in ERK1/2 Activation-We demonstrated that Ca 2ϩ influx through reverse-mode NCX was required for ERK1/2 phosphorylation in response to thrombin. However, although reverse-mode NCX inhibitors suppressed maximal ERK1/2 phosphorylation by almost 80% (Fig. 1, F and  G), their impact on bulk [Ca 2ϩ ] i was less pronounced (between 10 and 20%; Fig. 2D). This discrepancy could be explained by the contribution of Ca 2ϩ microdomains in thrombin signaling in ECs, as our differential Ca 2ϩ chelation experiments indicate (Figs. 1A and 2E). EGTA-AM is a "slow" Ca 2ϩ chelator allowing Ca 2ϩ to diffuse up to 100 nm from the source, although BAPTA-AM is a fast chelator binding Ca 2ϩ (with similar affinity to EGTA-AM) within 2 nm from the source (30). The existence of Ca 2ϩ microdomains regulating ERK1/2 activation in HUVECs is also supported by our previous work. Thapsigargin, an inhibitor of the sarcoendoplasmic reticulum Ca 2ϩ -ATPase that substantially increased [Ca 2ϩ ] i due to Ca 2ϩ leakage from the internal stores and subsequent store-operated Ca 2ϩ entry, failed to induce ERK1/2 phosphorylation (18). Conversely, the Ca 2ϩ ionophore ionomycin, which is expected to raise [Ca 2ϩ ] i uniformly throughout the plasma membrane, induced ERK1/2 phosphorylation (19). Thus, in agreement with our work with VEGF (18,19), local, as opposed to global, Ca 2ϩ signals appear to affect ERK1/2 phosphorylation in response to thrombin. Consequently, because NCX is a high capacity Ca 2ϩ transporter (12), inhibition of reverse-mode NCX could significantly modify Ca 2ϩ signals in microdomains close to the plasma membrane, while modestly affecting global Ca 2ϩ signals. Thus, reverse-mode NCX activity could conceivably have a major impact on signaling events requiring Ca 2ϩ at the locality of the plasma membrane, providing spatial resolution to such a signal.
Interestingly, in the [Ca 2ϩ ] i experiments shown in Fig. 2, thrombin stimulation appeared to result in a higher steadystate [Ca 2ϩ ] i , after the initial peak of the Ca 2ϩ signal, when compared with unstimulated controls. Moreover, reversemode NCX inhibitors suppressed this increase in steady-state [Ca 2ϩ ] i , suggesting that reverse-mode NCX activity could be involved. Nonetheless, further experimental work with thrombin-stimulated ECs in nominal Ca 2ϩ -free buffers (to estimate the relative contribution of Ca 2ϩ influx on bulk [Ca 2ϩ ] i ) and with the use of inhibitors or siRNA of other Ca 2ϩ -influx mechanisms known to be activated by thrombin in ECs, would be FIGURE 9. SEA0400 attenuates endothelial barrier dysfunction and edema formation in response to PAR-1 activation in vivo. C57BL/6 mice received 10 mg/kg SEA0400 by oral gavage. Two hours later, animals were injected i.p. with EBA (20 mg/kg) solution and, where appropriate, 1 mg/kg PAR-1 peptide agonist. After 30 min, animals were culled, and blood was collected for EBA concentration levels. Mice were then perfused with PBS containing 5 mM EDTA, and tissues were collected. A, EBA content of the lung tissue was determined with formamide extraction (data not shown). Bars represent means Ϯ S.E. g of EBA corrected for tissue weight (n ϭ 4 -7). *, p Ͻ 0.05 versus the PAR-stimulated control. B, pulmonary edema determination was assessed by the ratio of wet/dry lung weight (C). ERK1/2 activation was assessed by Western blot of lung tissue as described under "Experimental Procedures." D, densitometric analysis of ERK1/2 phosphorylation of Western blots in C was performed as in Fig. 1E, n ϭ 4. E, EBA extravasation in the kidneys of experimental animals was determined as in A, n ϭ 4 -7.
required to conclude that this steady-state increase in [Ca 2ϩ ] i occurs primarily via NCX activation, as the pharmacology indicates.
Reverse-mode NCX in Thrombin-induced Angiogenesis-Thrombin promotes in vitro and in vivo angiogenesis (11,21,34). The complex angiogenic response can be broken down to elementary processes, including EC adhesion to substrate, proliferation, motility, tubular differentiation, and permeability (45). Reverse-mode NCX inhibitors and siRNA attenuated thrombin-induced tubular differentiation and the same inhibitors suppressed EC proliferation in response to thrombin (Fig.  5). ERK1/2 phosphorylation has been reported to be crucial for EC tubulogenesis (46) and mitogenesis (47) in response to thrombin. Moreover, in cancers, endothelial ERK1/2 activation by PAR signaling has been implicated in resistance to angiogenic therapy (34), and activation of PAR-1 expressed on ECs by tumor-secreted matrix metalloproteinases promotes tumor progression (48). Consequently, it is probable that reversemode NCX activity regulates the angiogenic response to thrombin by promoting ERK1/2 activation.
Interestingly, the more selective reverse-mode NCX inhibitors SEA0400 and SN-6 significantly reduced tubulogenesis below unstimulated control levels only in thrombin-stimulated HUVECs (Fig. 5B), although their effect on tubulogenesis in unstimulated controls was minimal (Fig. 5E). A crucial difference of the tubulogenesis experiments shown in Fig. 5 compared with all the other assays (where the effect of SEA0400 or SN-6 under basal conditions is minimal) is that inhibitors and stimulant were applied when cells were in suspension. It is feasible that thrombin has an additional effect on EC adhesion and/or spreading on extracellular matrix components that is also inhibited by SEA0400 and SN-6, thus resulting in more pronounced suppression of tubule formation in the thrombinstimulated samples. More experimental work is needed to investigate this intriguing possibility.
Anti-angiogenicmonotherapyoftenfailsduetoadaptiveresistance (e.g. ECs switching to a different pro-angiogenic pathway) (4). Taken together, our findings with thrombin described here and our previous work with VEGF (18,19) suggest that reversemode NCX activity could be required for ERK1/2 activation and the angiogenic response downstream of diverse pathways. Thus, inhibitors of reverse-mode NCX could potentially be advantageous for the treatment of cancers resistant to conventional anti-angiogenic agents. The use of reverse-mode NCX inhibitors in in vivo tumor models merit further investigation.
Thrombin-induced ERK1/2 Activation and Vascular Permeability-Reverse-mode NCX inhibitors and siRNA targeting NCX1 preserved EC barrier function in response to thrombin stimulation in vitro (Fig. 6) and in vivo (Fig. 9), probably due to the inhibition of ERK1/2 activation. Unlike the role of ERK1/2 in thrombin-induced angiogenesis (34,46,47), its function as a regulator of endothelial permeability is less clear. Suppression of ERK1/2 phosphorylation prevented endothelial barrier disruption and maintained cortical actin levels in response to thrombin in HUVECs (35), in agreement with our data (Fig.  6A). This was further supported by the finding that ERK1/2 pathway inhibitors alleviated endothelial hyperpermeability in response to endotoxin insult in vivo (49), a process proposed to involve PAR-1 transactivation by matrix metalloproteinases (50). Nonetheless, the issue is still controversial, and the role of ERK1/2 activity in thrombin-induced HUVEC barrier disruption has been challenged (37). More- Our work is consistent with the following simplified mechanistic model. Activation of PLC␤3 downstream of protease-activated receptor 1 (PAR-1) leads to the generation of diacylglycerol (DAG) and inositol 1,4,5-triphosphate (IP 3 ). IP 3 binds to and activates the endoplasmic reticulum (ER) IP 3 receptors (IP 3 Rs), leading to Ca 2ϩ release from the internal stores, a rise in cytosolic Ca 2ϩ concentration ([Ca 2ϩ ] i ), and the activation of the plasma membrane NCX. Concurrently, associated Na ϩ -influx pathways that could be diacylglycerol-sensitive and/or plasma membrane depolarization (V m ) could lead to induction of reverse-mode NCX and Ca 2ϩ influx. This Ca 2ϩ influx promotes indirectly, via a yet unidentified Ca 2ϩ -sensitive mechanism, the assembly and subsequent activation of the NADPH oxidase 2 complex, resulting in ROS generation, ERK1/2 activation, and ultimately diminished endothelial barrier function and a pro-angiogenic phenotype. Ca 2ϩ uptake via the sarcoendoplasmic reticulum Ca 2ϩ -ATPase (SERCA) could switch off the Ca 2ϩ response. Other associated ionic mechanisms not depicted here for the sake of simplicity, such as mitochondrial Ca 2ϩ -uptake and/or store operated Ca 2ϩ entry could further fine-tune indirectly the degree of NCX activation/reversal impacting on the functional output. over, there is no conclusive mechanistic evidence linking ERK1/2 activity and permeability in ECs. Recently, ERK1/2 was shown to directly regulate barrier function by modulating the association of the tight junction protein zona occludens-1 (ZO-1) with occludin in ECs stimulated with the arachidonic acid metabolite 15(S)-hydroxyeicosatetraenoic acid (51). Whether a similar mechanism operates in thrombin-stimulated ECs remains to be determined.
Ca 2ϩ Influx through Reverse-mode NCX as a Mediator of Vascular Permeability-A role for Ca 2ϩ in endothelial barrier disruption by thrombin has been demonstrated in numerous in vitro and in vivo models. The most well documented is the role of Ca 2ϩ influx through the nonselective cation channel canonical transient receptor potential (TRPC)-1 and -6. In particular, TRPC6 was required for RhoA and PKC␣ activation and subsequent EC barrier dysfunction, and TRPC1 was required for NF-B activation downstream of thrombin (1,2,36,38). In endotoxin-induced acute lung injury, a model characterized by extensive endothelial hyperpermeability attributed, at least partially, to PAR-1 activation (50), inhibition of Ca 2ϩ signaling through TRPC6 (52) or stromal interaction molecule-1 (STIM1) (53) enhanced EC barrier function and improved survival.
In a previous study investigating the role of ERK1/2 activation in EC thrombin signaling (35), ERK1/2 inhibition did not affect myosin light chain kinase activity, which is downstream of RhoA activation and regulates stress fiber formation (1,2). Moreover, the investigators reported that inhibiting the ERK1/2 pathway activity resulted in the appearance of stress fibers at the EC periphery (35), similar to our findings with reverse-mode NCX inhibitors (Fig. 6E). Given that Ca 2ϩ influx through TRPC6 activates RhoA by thrombin (36), and reversemode NCX inhibitors did not affect RhoA activation (Fig. 6, F and G), reverse-mode NCX and TRPC6 possibly regulate distinct Ca 2ϩ influx pathways. Alternatively, given the increasing evidence suggesting that TRPC3-and/or TRPC6-mediated Na ϩ entry promotes reverse-mode NCX in smooth muscle cells (54), cardiomyocytes (55), and recently platelets (56), a proportion of TRPCs could be functionally coupled with NCX linking PLC activation to reverse-mode NCX in ECs. The role of TRPC3 and/or TRPC6 in ERK1/2 activation in ECs is unclear and merits investigation. Additionally, reverse-mode NCX could also be functionally associated with other ionic mechanisms, such as voltage-gated Na ϩ channels, via the membrane potential, which could also influence the direction of the exchanger (12,19), or the Na ϩ -K ϩ -ATPase, by controlling [Na ϩ ] (Fig. 3A).
Our research suggests that reverse-mode NCX is a novel, complementary Ca 2ϩ influx mechanism contributing to thrombin-induced angiogenesis and permeability, at least partially by mediating ERK1/2 activation. Thus, reverse-mode NCX, in cooperation with other associated ionic mechanism(s) would further fine-tune the complex EC response to PAR-1 activation, providing spatiotemporal resolution to the Ca 2ϩ signal and subsequent ERK1/2 activation and also linking the signal intensity to transmembrane [Na ϩ ] and [Ca 2ϩ ], the membrane potential, and the energy status of the cell. (51) or hydrogen peroxide (60), suggesting that ROS can modulate ERK1/2 activation in ECs and impact on tight junction stability and endothelial permeability, as we propose here. Second, although the cross-talk between Nox2 activity, Ca 2ϩ signaling, and vascular permeability has been reported in models of endothelial barrier dysfunction (53,58), the exact molecular mechanism, especially in the case of thrombin, is not clear. Thus, further experimental work is required to firmly establish the role of ERK1/2 and determine the Ca 2ϩ -sensitive mechanism(s) regulating Nox2 activity/assembly in thrombin signaling in ECs.
Nonetheless, our work consistently suggests that Ca 2ϩ influx through reverse-mode NCX is a novel determinant of thrombin-induced EC angiogenesis and barrier dysfunction. Collectively, given our previous work on the role of reverse-mode NCX in VEGF signaling (18,19), inhibition of this pathway could be beneficial in conditions characterized by deregulated PAR-1 and VEGF activity, such as tumor angiogenesis and sepsis.