Inhibition of Endothelial Cell Migration, Intercellular Communication, and Vascular Tube Formation by Thromboxane A2 *

The eicosanoid thromboxane A2 (TXA2) is released by activated platelets, monocytes, and the vessel wall and interacts with high affinity receptors expressed in several tissues including endothelium. Whether TXA2 might alter endothelial migration and tube formation, two determinants of angiogenesis, is unknown. Thus, we investigated the effect of the TXA2 mimetic [1S-(1α,2β(5Z),3α(1E,3R),4α]-7-[3-(3-hydroxy-4-(4′-iodophenoxy)-1-butenyl)-7-oxabicyclo-[2.2.1]heptan-2-yl]-5′-heptenoic acid (IBOP) on human endothelial cell (HEC) migration and angiogenesis in vitro. IBOP stimulation inhibited HEC migration by 50% and in vitro capillary formation by 75%. These effects of IBOP were time- and concentration-dependent with an IC50 of 25 nm. IBOP did not affect integrin expression or cytoskeletal morphology of HEC. Since gap junction-mediated intercellular communication increases in migrating HEC, we determined whether IBOP might inhibit coupling or connexin expression in HEC. IBOP reduced the passage of microinjected dyes between HEC by 50%, and the effects of IBOP on migration and tube formation were mimicked by the gap junction inhibitor 18β-glycyrrhetinic acid (1 μm) with a similar time course and efficacy. IBOP (24 h) did not affect the expression or phosphorylation of connexin 43 in whole HEC lysates. Immunohistologic examination of HEC suggested that IBOP may impair functional coupling by altering the cellular distribution of gap junctions, leading to increased connexin 43 internalization. Thus, this finding that TXA2 mimetics can prevent HEC migration and tube formation, possibly by impairing intercellular communication, suggests that antagonizing TXA2 signaling might enhance vascularization of ischemic tissue.

Thromboxane A 2 (TXA 2 ) 1 is a biologically active eicosanoid primarily released from activated platelets, monocytes, and damaged vessel wall (1). The biological half-life of TXA 2 is approximately 30 s in in vitro models. Thus, TXA 2 acts as a paracrine/autocrine factor with potent effects on the vasculature including the initiation of platelet degranulation and aggregation, vasoconstriction, smooth muscle cell proliferation, and perturbation of endothelial cells (EC) (1,2). All of these effects of TXA 2 are mediated through interaction with cell surface receptors that are members of the G-protein-linked, seven-transmembrane domain receptor family (1). Thus far two TXA 2 receptor (TP) isoforms have been cloned (TP-␣ and -␤) (3,4). These isoforms vary in the length of their cytoplasmic tails and arise via alternate splicing of a single gene, with TP-␣ the result of a retained intron (4). The downstream signaling pathways of the two TP isoforms differ in their modulation of adenylate cyclase, with TP-␣ increasing cAMP accumulation and TP-␤ inhibiting cAMP accumulation (5,6). Expression and desensitization following prolonged agonist stimulation of the two receptor isoforms are also regulated differently; prolonged stimulation induces the expression of the TP-␤, but not the TP-␣, isoform, and the signaling pathway responsible for TP-␤ desensitization is much more sensitive to protein kinase C activation than is that of TP-␣ (7).
While expression of TP in the vasculature is widespread, little is known of the physiological effects of TXA 2 on endothelial cell growth or migration. Such effects might be expected to be of considerable clinical relevance; conditions in which TXA 2 synthesis is elevated, such as myocardial ischemia, unstable angina pectoris, and diabetes (1,8,9), are also those in which angiogenesis or vessel growth and generation of collateral circulation are important.
Thus, we have investigated the effects of TP stimulation on the migration and tube forming capabilities of endothelial cells. We report here that activation of the TXA 2 receptor by a potent and stable ligand (IBOP) inhibits the formation of vascular networks by HEC in vitro in a time-and dose-dependent fashion. A TXA 2 mimetic also inhibits endothelial cell migration in a denudation-injury model. Based on experiments measuring gap junction function and expression, we suggest that at least one mechanism by which both the antimigratory and antiangiogenic effects of TP stimulation occur is via the inhibition of intercellular communication.

EXPERIMENTAL PROCEDURES
Materials-Sterile plasticware was from Costar (Cambridge, MA). Tissue culture materials and reagents, excluding pooled human serum (Gemini Bio-Products Inc., Calabasas, CA), were from Life Technologies, Inc. The thromboxane A 2 mimetic IBOP and the TP blocker SQ29548 were from Cayman Chemical (Ann Arbor, MI). Type 2 collagenase was obtained from Worthington, and dispase was from Becton Dickinson (Bedford, MA). All other chemicals and reagents were obtained from Sigma unless otherwise stated.
Isolation of Endothelial Cells from Human Umbilical Veins-Human * This work was supported by National Institutes of Health Grants HL47032-05 and HL38449 and Fundacao de Amparo a Pesquisa do Estado de Sao Paulo, Brazil, Grant J997/2379-2. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  endothelial cells (HEC) were isolated from umbilical veins as previously reported (10). Culture medium consisted of M199, containing 20% (v/v) newborn calf serum and 5% (v/v) pooled human serum with 2 mM L-glutamine. Cells were plated onto plates precoated with 0.02% (w/v) gelatin with a medium change after 24 h and every 2 days thereafter until confluent. Primary cultures of HEC were harvested by incubation with 0.05% trypsin, 0.02% EDTA, and the cells were collected and cell number was determined using a dual threshold cell counter (Coulter Electronics, Luton, UK). Cells were plated at the concentrations described.
In Vitro Tube Forming Assay-The spontaneous formation of capillary-like structures by HEC on a basement membrane matrix preparation (Matrigel, Becton Dickinson, Bedford, MA) was used to assess angiogenic potential. Twelve-well plates were coated with matrix gel (10 mg/ml) according to the manufacturer's instructions. HEC were seeded on coated plates at 1.5 ϫ 10 5 cells/well and incubated at 37°C for 60 min. The medium was supplemented with the agents outlined, and the cultures were incubated at 37°C for 24 h. Tube formation was observed and photographed over the 24-h period using a solid state TV camera (COHU Electronics, CA) attached to an inverted phase contrast microscope. Images were captured using a video graphics system (Sony Electronics; monitor model PVM97, printer model UP890MD). The degree of tube formation was assessed by counting the number of tubes contained in two random fields from each well.
Migration Assay-The migration assay used was a monolayer denudation assay as described by Tang et al. (11). Confluent endothelial cells were wounded by scraping with a 2-200-l pipette tip, denuding a strip of the monolayer 300 m in diameter. Variation in the wound diameter within experiments was approximately 5%. Cultures were washed twice with PBS and incubated with serum-containing medium supplemented with the agents as indicated. Control cultures were exposed to medium alone. The rate of wound closure was observed and photographed over a 24-h period as for the tube forming assay. The progression of cell migration was quantitated by calculation of the denuded area using the Scion image program (version 1.61, Scion Corp., Frederick, MD).
Determination of Cell Proliferation and Cell Cycle Analysis-For cell proliferation experiments, HEC were plated into gelatin-coated 12-well plates at 1.3 ϫ 10 4 cells/cm 2 and allowed to attach for 48 h. Cells were washed twice with PBS and incubated in medium with or without 50 nM IBOP. Cells were counted every day for 3 days, and the amount of proliferation was compared. For cell cycle analysis, cells were plated into gelatin-coated 12-well plates at 1.3 ϫ 10 4 cells/cm 2 incubated at 37°C for 48 h. Cells were washed with PBS twice and synchronized by incubation in low serum for 24 h. Cells were either incubated in fresh medium alone or supplemented with 50 nM IBOP for 24 h. Cell cycle analysis was performed using the method of Giaretti and Nusse (48).
Flow Cytometric Analysis of Integrin Expression-Confluent IBOPtreated and untreated HEC were washed twice with PBS, and the cells were detached using 5 mM EDTA (pH 7.4). Cells on matrix gel were isolated by incubation with dispase. Cells were suspended in 1% bovine serum albumin in PBS (PBS-B) for 30 min at 4°C. At the end of the incubation, antibodies to the integrins ␣ v ␤ 3 or ␣ v ␤ 5 (5 g/ml, clone LM609 or P1F6; Chemicon, Temecula, CA) were added with negative controls stained with preimmune mouse serum. Cells were incubated with antibodies for 90 min at 4°C and washed three times with PBS-B before incubation with a FITC-conjugated secondary antibody (1:100 in PBS-B) for 60 min. Analysis of integrin surface expression was performed on a flow cytometer using an argon ion laser ( ex 488 nm).
Immunofluorescence and F-actin Staining-For staining of cellular F-actin, confluent HEC were treated with 50 nM IBOP for 24 h, washed twice with PBS, and fixed with 70% ethanol in PBS for 6 h at Ϫ20°C. Cells were washed twice again and incubated with 0.1 mg/ml rhodamine-conjugated phalloidin (Molecular Probes, Inc., Eugene, OR) for 20 min. For tubulin staining, cells were fixed with 70% ethanol in PEM buffer (100 mM PIPES, 2 mM EGTA, 2 mM MgCl 2 , pH 6.8), and nonspecific binding was blocked by incubation with 5% bovine serum albumin in PBS for 30 min at 37°C. Staining of ␣-tubulin used a monoclonal primary antibody (1:100; clone B-5-1-2) and a Cy3-conjugated antimouse secondary antibody (1:1000; Amersham Pharmacia Biotech). Both incubations were for 1 h at room temperature in a humid chamber. Stained cells were washed with PBS and mounted in PBS/glycerol (1:1). HEC stained for F-actin and ␣-tubulin were viewed using ex 510 -560-nm and em 590-nm filters.
For staining of Cx43, treated HEC were fixed with 4% formaldehyde in PBS for 10 min and rendered permeable by incubation with 0.1% TX-100 in PBS for 20 min. Cells were washed three times with PBS-B and blocked with 3% preimmune goat serum in PBS-B for 90 min at room temperature. Cells were stained with the Chemicon monoclonal anti-Cx43 antibody used for Western blotting (1:1000 in PBS-B) for 1 h. Staining was detected with a FITC-conjugated secondary antibody. Images were collected with a Bio-Rad MRC 600 krypton/argon laser scanning confocal microscope linked to a Nikon Eclipse 200 microscope, using a 60ϫ N.A. 1.4 planapo infinity-corrected objective lens. Controls were imaged to assure no background fluorescence and no cross-fluorescence between the FITC channel and the Cy3 channel.
Dye Microinjection and Electrophysiology in HEC-The degree of coupling in HEC monolayers was determined by quantification of the spread of the fluorescent dye Lucifer yellow following the microinjection into one cell in a microscope field. Confluent HEC monolayers, untreated or treated for 24 h with 50 nM IBOP, were rinsed three times with PBS and injected with a solution of Lucifer yellow (5% (w/v) in 150 mM LiCl) through drawn glass capillaries, using brief overcompensation of the capacitance neutralization circuit of a WP Instruments electrometer (New Haven, CT). Cells were photographed 10 min after injection using FITC epi-illumination and excitation filters. Visual inspection of images and quantitation of the cells containing dye was used to assess coupling. Recordings of junctional conductance were obtained from freshly dissociated, paired HEC on glass coverslips using the dual whole cell voltage clamp technique (12). Experiments were performed at room temperature in bathing solution containing 140 mM NaCl, 2 mM CsCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, 4 mM KCl, 5 mM dextrose, 2 mM pyruvate, and 1 mM BaCl 2 , pH 7.2. Each cell in the pair was voltageclamped with heat-polished patch pipettes (3-7 M) filled with internal solution 130 mM CsCl, 0.5 mM CaCl 2 , 2 mM Na 2 ATP, 3 mM MgATP, 10 mM HEPES, 10 mM EGTA (pH 7.2)). Seals to cell surfaces were achieved with suction and were monitored by recording currents while simultaneously applying, through both electrodes, 20-ms, 4-mV pulses at 10 Hz. During suction, the current from the pipette was measured in the track mode of the voltage clamp system (model 1-C; Axon Instruments, Inc.).
Ca 2ϩ Imaging in Stimulated HEC-Primary cultures of HEC were plated at 5 ϫ 10 5 cells/plate onto confocal imaging dishes (Glass Bottom Microwells; MatTek Corp.) coated with gelatin. After achieving confluence, HEC were wounded and incubated for up to 24 h with the agents of interest. For the calcium imaging experiments, HEC were incubated for 45 min at 37°C with 10 M of the Ca 2ϩ indicator Indo-1/AM (Molecular Probes) and rinsed three times with Tyrode's solution (137.0 mM NaCl, 4.0 mM KCl, 0.5 mM MgCl 2 , 2.0 mM CaCl 2 , 24.0 mM NaHCO 3 , 1.8 mM NaH 2 PO 4 , 5.5 mM glucose, 5.0 mM HEPES, pH 7.2). The alteration in intracellular Ca 2ϩ and the propagation of the Ca 2ϩ waves were induced by focal mechanical stimulation of a single cell with a glass pipette (1-2-m outer diameter) and measured as the ratio of Indo-1 fluorescence intensity emitted at two wavelengths (390 -440 and 440 nm) imaged with a Nikon RCM8000 real time confocal microscope equipped with UV laser, excitation at 351 nm, small pinhole, and Nikon ϫ 40 UV-corrected water immersion objective (numerical appeture 1.15, working distance 0.2 mm). Ratio images were continuously acquired at 1 Hz after background and shading correction and were saved as the average of 32 image pairs. The images were further analyzed during playback using Polygon-Star Software (Nikon) that averages the number of pixels (gray level) within the region of interest (circular spots placed on top of each cell, radii of 4.2 m containing ϳ130 pixels) as function of elapsed time. The efficacy of Ca 2ϩ wave communication was measured as the number of responding cells per total number of cells in the confocal field. The velocity of Ca 2ϩ wave propagation between HEC was calculated from the distance between the mechanically stimulated cell and neighboring cells divided by the latencies to half-maximal response (peak fluorescence) determined from sigmoidal curve fitting using Microcal Origin 4.1 software. The wave amplitude was measured as the ratio between the peak maximal increase and basal Indo-1 fluorescence, and the efficacy of Ca 2ϩ wave communication was measured as the number of responding cells per total number of cells in the confocal field. All determinations of Ca 2ϩ wave propagation were measured from HEC cultures untreated and treated for 2, 8, 16, and 24 h with IBOP (50 nM), 24 h with SQ29548 (5 M), and both agents together (24 h).
Analysis of Cx43 Expression (Western Blotting) and Phosphorylation in HEC-Cells were grown to confluence and treated for up to 24 h with 50 nM IBOP. For the last 2 h of the incubation period, the phosphatase inhibitors sodium vanadate (1 mM) and okadaic acid (1 M) were included. For Western blotting, monolayers were washed twice in Hanks' balanced salt solution, lysed in 200 ml of hot (85°C) SDS gel-loading buffer (2% SDS, 10% glycerol, 50 mM Tris-HCl pH 6.8), and transferred to microcentrifuge tubes. Samples were boiled for 10 min, sonicated briefly, and centrifuged at 10,000 ϫ g for 10 min. The protein concen-tration of the retained supernatants was determined using the BCA assay system (Pierce). Aliquots of samples (10 g) were separated by SDS-polyacrylamide gel electrophoresis under reducing conditions (13), using 12% acrylamide gels, transferred onto polyvinylidene difluoride membrane (Bio-Rad), and analyzed by Western blotting. Membranes were preincubated with 5% nonfat dry milk, 0.05% Tween 20 in Trisbuffered saline for 1 h and incubated for 16 h at 4°C with monoclonal anti-human connexin 43 antibody (1:1000; Chemicon). Detection utilized a horseradish peroxidase-conjugated goat anti-mouse IgG antibody (Bio-Rad) and chemiluminescent detection (Pierce).
For the analysis of the phosphorylation state of the Cx43 protein, cells were grown to confluence and treated for up to 24 h with 50 nM IBOP. For the last 2 h of the incubation period, the phosphatase inhibitors sodium vanadate (1 mM) and okadaic acid (1 M) were included. Cells were washed three times with Tris-buffered saline (25 mM Tris, pH 7.4) and incubated in Tris-buffered saline supplemented with 1% bovine serum albumin, 0.5 mCi of H 3 PO 4 , 1 mM sodium vanadate, and 1 M okadaic acid for 2 h at 37°C. Cells were washed three times with cold Tris-buffered saline and lysed with radioimmune precipitation buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) on ice for 20 min. Lysates were clarified by centrifugation at 14,000 rpm for 20 min. Labeled Cx43 was immunoprecipitated by incubation with protein G-agarose beads, coated with saturating amounts of Cx43 monoclonal antibody, overnight at 4°C. The resulting immune complexes were boiled in SDSloading buffer and separated on 12% acrylamide gels. Gels were dried down and exposed to autoradiographic film for the appropriate amount of time.
Statistical Analysis-Statistical analysis of pooled data was performed using the Mann-Whitney U test.

TXA 2 Mimetic Prevents HEC Differentiation in Three-dimensional
Cultures-HEC on matrix gel undergo alignment into cords within 2 h, which establishes the pattern for further tube formation. This is shown by the stability in the number of tubes per field in control cultures (Fig. 1E). By 6 h, tube formation had begun, and by 12 h virtually all cells had fused into continuous tubes, with stabilization and refinement progressing up until 24 h (Fig. 1B). Generation of tubes by HEC was reduced by the TXA 2 mimetic IBOP (Fig. 1C). IBOP (50 nM) reduced to 43 Ϯ 14% of controls the alignment of HEC into cordlike structures at 2 h (p Ͻ 0.01) (Fig. 1E). By 16 h, IBOPtreated cultures contained only 15 Ϯ 12% of the number of tubes in untreated wells (p Յ 0.05) (Fig. 1, C and E). HEC incubated with the TP blocker SQ29548 alone formed a number of tubes per field similar to that formed by untreated cells (Fig.  1E), and the simultaneous addition of SQ29548 prevented the effect of the TXA 2 mimetic at each time point (p Յ 0.05 compared with IBOP alone) (Fig. 1, D and E). The inactive hydrolytic product of TXA 2 , TXB 2 , did not influence tube formation by HEC, nor did the prostaglandin PGE 2 (data not shown).
IBOP was a concentration-and time-dependent inhibitor of tube formation by HEC (Fig. 2, A and B, respectively). Vascular tube formation was significantly depressed by IBOP concentrations higher than 25 nM (p Յ 0.01), with maximal inhibition (40 Ϯ 20% of control) at IBOP concentrations of 50 nM or more (p Յ 0.01). Another TXA 2 mimetic, U46619, also inhibited tube formation in a concentration-dependent manner. U46619 inhibited HEC tube formation with an IC 50 of 100 nM (p Յ 0.05), with maximal inhibition by U46619 at concentrations above 400 nM, at which tube formation was reduced to 17 Ϯ 5.4% of control (p Յ 0.01) ( Fig. 2A). IBOP also prevented formation of tubes if added to cultures already in the process of tube formation (Fig. 2B). Inclusion of IBOP (50 nM) in the medium as late as 18 h after the initiation of tube formation significantly reduced the eventual number of tubes at 24 h (70 Ϯ 21%, p Յ 0.001). IBOP was most effective at periods greater than 18 h, with the number of tubes suppressed to 40 Ϯ 21% of control (p Յ 0.0001). Thus, TP activation was found to exert a potent inhibitory effect on in vitro tube formation.

TP Stimulation Also Diminishes the Rate of EC Wound Heal-
ing-Tube formation in three-dimensional cultures is dependent on the migration of cells after plating. The inhibition of tube formation by IBOP and U46619 suggested that TXA 2 mimetics might also inhibit HEC migration. We used a denudation injury model to assess the impact of IBOP on endothelial cell migration. Confluent, scrape-wounded HEC monolayers were incubated with either IBOP (50 nM) or U46619 (400 nM), and the rate of closure was observed over the following 24 h. Untreated cultures migrated into the denuded area, recovering the exposed surface and reducing the uncovered area to 27 Ϯ 16.5% of the original area (Fig. 3, B and E). IBOP (50 nM) significantly inhibited migration of HEC into the denuded area (Fig. 3C), with inhibition detectable 6 h after initial wounding (p Յ 0.05). Wound closure was inhibited by 58.4 Ϯ 8.1% at 24 h (p Յ 0.05). The TP agonist U46619 (400 nM) also inhibited endothelial cell migration with kinetics similar to IBOP with closure of the wound inhibited by 71.7 Ϯ 14.1% after 24 h of treatment (p Յ 0.005) (data not shown). TP blockade by SQ29548 did not affect HEC migration into the wounded area; however, the migration of IBOP-treated HEC was restored by inclusion of 5 M SQ29548 (p Յ 0.05 versus IBOP alone) for longer than 12 h (Fig. 3, D and E). The inactive metabolite of TXA 2 , TXB 2 , did not affect HEC migration.
The Antimigratory Effects of TXA 2 Mimetics Are Not Mediated through Alterations to Endothelial Cell Proliferation, Cytoskeletal Morphology, or Expression of ␣ v ␤ 3 and ␣ v ␤ 5 Integrin Receptors-Among the factors important in the regulation of migration and angiogenesis are the expression of cell surface integrins, cellular proliferation, and architecture of the cytoskeleton (14 -17). We examined the effects of IBOP stimulation on endothelial cell proliferation using growth curves and cell cycle kinetics. Cells grown on gelatin-coated 12-well plates in medium alone increased cell number over 3 days with cell number 172 Ϯ 9.5% of the plating density by day 3. IBOP treatment (50 nM) increased cell number to 174 Ϯ 5.7% of the original plating density in the same period, which was not significantly different from control (data not shown). When examining cell cycle kinetics, we found that unstimulated and IBOP-treated HEC progressed through the cell cycle at comparable rates (data not shown). IBOP did not retard cell cycle progression, nor did it inhibit HEC proliferation; thus, the inhibition of HEC tube formation and migration was unlikely to be a result of inhibition of cell proliferation.
We examined the expression of the integrin complexes ␣ v ␤ 3 and ␣ v ␤ 5 on cells from two-and three-dimensional cultures to determine if the effect of IBOP on migration and tube forma-tion was mediated through decreased expression of those integrin receptors (Fig. 4, A and B). Untreated cells from twodimensional cultures showed a 3-and 5-fold increase in their mean fluorescence intensity for ␣ v ␤ 3 and ␣ v ␤ 5 , respectively, over cells stained with preimmune serum. Cells from threedimensional matrix gel cultures expressed both ␣ v ␤ 3 and ␣ v ␤ 5 integrin complexes similarly to HEC in two-dimensional cultures (data not shown). IBOP (50 nM) did not influence the surface expression of either ␣ v ␤ 3 or ␣ v ␤ 5 in HEC from either culture model. Thus, the effect of IBOP on HEC migration and tube formation was not mediated through decreased integrin expression of those two integrin receptors.
To determine the effects of IBOP on the HEC cytoskeleton, immunofluorescent staining of the cytoskeletal components F- . Initial wound area differed by less than 5% between wells. #, significant difference (p Յ 0.05) from control; *, significant difference (p Յ 0.05) from IBOP treatment. actin and ␣-tubulin was performed. Staining for cytoskeletal architecture showed extensive stress fiber formation in HEC that was in a random orientation throughout the monolayer (Fig. 4C). Stress fibers in migrating cells at the wound edge were ordered with the orientation parallel to the direction of cell migration (data not shown). Incubation with the TXA 2 mimetic IBOP did not decrease stress fiber formation or alter the arrangement of F-actin within the monolayer (Fig. 4D), nor did it prevent the reorganization of the actin cytoskeleton at the wound edge. The pattern of ␣-tubulin staining in control cells was cytoplasmic and radial ( Fig. 4E) with no apparent alterations in the morphology in migrating cells. Cells exposed to IBOP (50 nM) maintained the same morphology of ␣-tubulin as did the control cultures (Fig. 4F). Thus, gross alterations to the cytoskeletal morphology were not responsible for mediating the inhibitory effects of IBOP on HEC migration.

TXA 2 Mimetic Potently Inhibits Gap Junction-mediated
Intercellular Communication-The degree of cell-cell communication is an important factor affecting the migration of endothelial cells and other cell types (18 -21). Migrating endothelial cells up-regulate gap junction-mediated intercellular communication maximally at 24 h (19), the time point at which the antimigratory activity of IBOP is most evident. We advanced the hypothesis that ligand binding to the TXA 2 receptor inhibits migration and tube formation by interfering with intercellular communication. Junctional communication in HEC monolayers was studied by microinjection of the fluorescent dye Lucifer yellow into cells, and the extent of dye diffusion was determined. After injection into untreated cells, Lucifer yellow diffused to an average of 5.5 Ϯ 0.7 cells (Fig. 5, A and B). Treatment for 24 h with the TXA 2 mimetic IBOP (50 nM) reduced the diffusion of Lucifer yellow to 2.4 Ϯ 0.5 cells (p Յ 0.05), a reduction in coupling of 64% (Fig. 5, A and C). To examine the electrical properties of IBOP-treated cells directly, pairs of freshly dissociated cells were voltage-clamped, and recordings of junctional conductance were made (Fig. 5D). Treatment of HEC with 50 nM IBOP (24 h) reduced the conductance between cell pairs to 1.9 Ϯ 1 microsiemens, an 85% reduction of conductance in comparison with controls (13 Ϯ 5.8 microsiemens, p Ͻ 0.05).
Further characterization of the effects of TXA 2 receptor agonists on gap junction-mediated intercellular communication employed real time confocal microscopy to examine the propagation of Ca 2ϩ waves through endothelial cell monolayers. Confluent HEC monolayers were treated with 50 nM IBOP for 0 -24 h and loaded with the Ca 2ϩ indicator dye, and the efficacy of Ca 2ϩ wave propagation, defined as the number of cells reached by the Ca 2ϩ signal per total number of cells in the confocal field, was analyzed (Fig. 6). In untreated cells, the intracellular Ca 2ϩ waves spread to 63 Ϯ 4% of the cells in the field (n ϭ 16, Fig. 6). In similarly confluent cultures, exposure to the TXA 2 mimetic IBOP (50 nM) for periods longer than 2 h significantly reduced the spread of Ca 2ϩ waves over the same period of study (Fig. 6). After 8-and 16-h exposure, IBOP inhibited the communication of the Ca 2ϩ signal by 35% (p Յ 0.05) with inhibition maximal after 24 h with Ca 2ϩ signal spreading to only 52% of FIG. 4. Expression of integrins and cytoskeletal morphology following IBOP. A-B, surface expression of the integrins ␣ v ␤ 3 and ␣ v ␤ 5 was examined using complex-specific monoclonal antibodies. Untreated and IBOP-treated HEC (24 h) were harvested with EDTA and stained for surface expression of integrins. A FITC-conjugated anti-mouse antibody was used for detection, and expression was analyzed by flow cytometry. Histograms are composites with the expression of integrins on untreated cultures in black, IBOP-treated cultures in white, and negative controls in gray. C-F, confluent HEC were treated for 24 h with 50 nM IBOP, stained with either rhodamineconjugated phalloidin (F-actin) or a monoclonal antibody for ␣-tubulin, and examined for changes in cytoskeletal morphology. The morphology of F-actin and ␣-tubulin is shown in untreated (C and E, respectively) and IBOP-treated cells (D and F, respectively). Photomicrographs were taken at ϫ 400 magnification using excitation and emission filters for rhodamine. Results are representative of three individual experiments. cells in the field (p Յ 0.05) (Fig. 6). The inhibition of Ca 2ϩ waves was reversed by the TP antagonist SQ29548 (5 M, Fig.  6). Other parameters measured (velocity of wave propagation and amplitude of the response) were similar in both control and IBOP-treated groups (data not shown).

Pharmacological Manipulation of Gap Junction Function Mimics the Effects of the TXA 2 Receptor on Endothelial Cell
Migration and Tube Formation-As the antimigratory and antiangiogenic effects of TXA 2 mimetics were correlated with decreased strength of coupling in treated HEC, we investigated the effects of the gap junction inhibitor 18␤-glycyrrhetinic acid (18␤-GA) on HEC tube formation and migration in vitro. The effects of 18␤-GA on HEC migration in the denudation injury assay were assessed (Fig. 7A). Concentrations of 18␤-GA as low as 1 M reduced HEC migration by 64% (p Յ 0.05) with 57 Ϯ 11.8% of the wound area uncovered after 24 h. 18␤-GA inhibited HEC migration in a concentration-dependent fashion with inhibition greatest at 18␤-GA concentrations above 5 M and migration abrogated at 20 M. Since IBOP inhibited the migration of endothelial cells 50% compared with controls (Fig. 3E), we chose to use 1 M 18␤-GA in further experiments, since it was equivalent in efficacy to IBOP. In the three-dimensional model of vascular tube formation, 18␤-GA (1 M) reduced the formation of tubular networks to 39 Ϯ 20% of control (p Յ 0.001), thereby implicating gap junction-mediated intercellular communication as a vital part in the angiogenic process (Fig.  7B). The morphology of 18␤-GA-treated cultures was identical to that of IBOP-treated cultures (Fig. 1C), which is consistent with the concept that inhibition of intercellular communication through gap junctions by TXA 2 is at least one mechanism by which TXA 2 can inhibit EC migration and tube formation in vitro.

Inhibition of Intercellular Communication by IBOP Coincides with Altered Distribution of Cx43 without Changes in
Cx43 Abundance-Intercellular communication is regulated by a number of factors, including the phosphorylation state and half-life of the connexin protein, the rate of shuttling of connexin in and out of the membrane, and changes in the rate of transition between the open/closed state (22). Low passage HEC express both Cx40 and Cx43, although Cx43 is the dominant gap junction protein in these cells (23,24). Thus, we examined the expression levels and intracellular distribution of Cx43 in HEC stimulated with the TXA 2 mimetic IBOP. To determine the effect of IBOP on Cx43 expression, lysates of treated cells were subjected to electrophoresis and blotted with an anti-Cx43 antibody. Connexin 43 was found to exist in two different species in HEC, one of which migrated at approximately 44 kDa and another at 45 kDa (Fig. 8A). Exposure to IBOP (50 nM) for up to 24 h did not change the overall expression of Cx43, nor did it alter the ratio of the two species (Fig.  8A). To assess the effect of IBOP treatment on Cx43 phosphorylation, we loaded cells with [ 32 P]H 3 PO 4 and analyzed the phosphorylation of immunoprecipitated Cx43 via autoradiog-FIG. 5. Gap junction-mediated intercellular communication following IBOP. A, cells were washed and microinjected with a 10% (w/v) solution of Lucifer yellow using a micromanipulator and pulled glass pipettes. The dye was allowed to diffuse through the cell monolayer for 10 min, and control (B) and IBOP-treated (C) cells were photographed using FITC-filters under ϫ 400 magnification. D, junctional conductance (g i ) in freshly dissociated cell pairs was measured by passing 20-ms, 2-mV pulses in one cell and recording junctional current (I j ) in the other cell; g i was calculated as I j divided by the transjunctional driving force. The data represent the mean conductance Ϯ S.D. from 10 recordings. #, significant difference (p Յ 0.05) from control.
FIG. 6. Ca 2؉ imaging in confluent HEC following IBOP. HEC were grown to confluence on mat-tek plates and were scrape-wounded and incubated with 50 nM IBOP for 2, 8, 16, or 24 h. Cells were loaded with Indo-1/AM, and the efficacy of Ca 2ϩ wave propagation induced by focal mechanical propagation was measured as the number of cells that displayed an increase in the Indo-1 fluorescence ratio (increase in intracellular Ca 2ϩ level) per total number of cells in the microscope field. Results are the combination of five experiments. Data are represented as mean Ϯ S.E. #, significant difference (p Յ 0.05) from control. raphy subsequent to SDS-polyacrylamide gel electrophoresis. Analysis of immunoprecipitated Cx43 protein from 32 P-loaded cells showed two bands migrating at approximately the same molecular weight as those described for Western blot analysis (Fig. 8B). IBOP was again shown not to influence the amount or ratio of the two 32 P-labeled Cx43 species in HEC. Thus, TP stimulation did not control junctional communication through alterations in the amount of Cx43 expressed or the phosphorylation state of Cx43.
To assess the intracellular distribution of Cx43 within HEC, immunofluorescence analysis was performed on cells rendered permeable and then incubated with either medium alone or 50 nM IBOP for 24 h (Fig. 8, C and D). In untreated cells, Cx43 immunoreactivity was observed both in the membrane and cytoplasm (Fig. 8C). Diffuse perinuclear staining presumably represents Cx43 protein being synthesized or on its way to the plasma membrane; particulate cytoplasmic staining may represent internalized Cx43 complexes previously resident in the membrane (25). Membrane staining in control cultures was particulate and extensive along areas of cell-cell contact. Incu-bation with IBOP (50 nM) did not significantly alter the amount of diffuse perinuclear Cx43 immunoreactivity in the cytoplasm of HEC. However, IBOP dramatically altered the size of the junctional plaques, causing aggregation of channels into larger plaques, which were fewer in number compared with controls (Fig. 8D). IBOP stimulation also increased particulate cytoplasmic Cx43 staining by 2.3-fold from 3.8 Ϯ 1.8 to 8.7 Ϯ 2.2 particles/cell (p Յ 0.05). Thus, the inhibition of coupling by TXA 2 may reflect the change in the shape or clustering of gap junctions in the membrane or redistribution of Cx43 from membrane to cytoplasm. DISCUSSION In this study, we found that the TXA 2 mimetic IBOP was a potent time-and concentration-dependent inhibitor of in vitro tube formation, with an IC 50 of 25 nM. At an IBOP concentration of 50 nM, endothelial cell migration was inhibited by 50%. These effects contrast sharply with those that would be expected from the known effects of TXA 2 on signal transduction. TP stimulation causes the activation of phospholipase C, protein kinase C, and extracellular signal-regulated kinase (26 -29), which have been associated with enhanced migration and angiogenesis in different systems (30 -32). This disparity is unlikely to result from the fact that one of the TP isoforms in low passage HEC, TP-␤, couples to the G-protein G i as well as G q (4). Activation of G i has been shown to be necessary for the generation of tubes on matrix gel (6,33), and indeed, other ligands for G-protein-coupled receptors that activate G i , such as PF4, substance P, endothelin, and PGE 1 , stimulate angiogenesis and enhance migration in similar EC assays (34 -36). Thus, despite the similarity that these ligands share concerning other aspects of vascular function (e.g. vasoconstriction), TXA 2 receptor stimulation causes effects that are qualitatively different from those of many other G-protein-linked receptor ligands in endothelial cells, in that it inhibits cell migration and tube formation.
The fact that TP stimulation inhibited intercellular communication in HEC monolayers and that the pharmacological uncoupling reagent 18␤-GA also inhibited HEC migration and tube formation suggests one possible explanation for our results, namely that the inhibition of tube formation and antimigratory activity of the TXA 2 mimetic was due to inhibition of gap junction function. Indeed, endothelial cell migration and angiogenesis are sensitive to inhibition of gap junction-mediated intercellular communication, and antiangiogenic cytokines, such as interleukin-1␤, potently inhibit intercellular communication in endothelium (37)(38)(39). Inhibition of junctional communication alone is not sufficient to explain the loss of EC function, however, since many proangiogenic G-proteinlinked receptor ligands also decrease junctional communication (40,41). The difference between the effects of TXA 2 and other G-protein-linked receptor ligands may be related to the kinetics of the inhibition and the mode of action. Lysophosphatidic acid, thrombin, and endothelin (ET) all inhibit junctional communication in various cells (40,41). With lysophosphatidic acid, ET, and thrombin, however, maximal inhibition occurs within 5-10 min of ligand binding with the restoration of junctional function at 3 h mediated through receptor desensitization (40,41). In those studies, gap junction closure by lysophosphatidic acid, thrombin, and ET was not accompanied by changes to the morphology of junctional plaques. Rather, the loss of intercellular communication was associated with changes in the phosphorylation state of Cx43, mediated by either c-Src or MAP kinase activation (40,41).
Indeed, Cx43 phosphorylation has been shown to be altered by changes in Ca 2ϩ /inositol 1,4,5-trisphosphate generation; protein kinase C, protein kinase A, and mitogen-activated protein kinase activation; cAMP accumulation; and c-Src activity (42). Binding of TXA 2 to the receptor has been shown to activate all of these signal transduction pathways (29,43). 2 Yet, in low passage HEC, stimulation of TP with ligand produces no discernable change in the expression level or phosphorylation state of Cx43, as detected by immunoblotting. Along with the alterations in the morphology of junctional plaques found with IBOP, these observations may indicate alternate pathways used by TXA 2 to uncouple the cells.
The parameters examined that correlated with decreased functional coupling in IBOP-treated HEC were a change in the size of the junctional plaques, with IBOP causing clustering of Cx43 into larger gap junction plaques and redistribution of Cx43 aggregates from membrane to cytoplasm. Agents that increase cytoplasmic levels of cAMP have previously been shown to induce aggregation of Cx43 into larger sized plaques (44,45); such cAMP-induced enlargement of gap junctions has been associated with decreased intercellular communication in myocytes (46). The mechanism by which cAMP causes these effects is not certain. Cyclic AMP has effects on coupling that vary among multiple cell types. In many cases, but not all, cAMP induces alteration in junctional communication with simultaneous alteration in phosphorylation of the Cx43 protein (42,43,45). Thus, one mechanism by which TXA 2 receptor stimulation could control gap junction-mediated intercellular communication in HEC is by elevation of cAMP. Evidence in favor of additional mechanisms includes the fact that TXA 2 mimetics decrease coupling, independently of changes in Cx43 phosphorylation.
A recent report by Jordan and colleagues (25) demonstrated internalization of Cx43-GFP chimeras after insertion in the membrane. Cx43 proteins inserted in the membrane coalesced into larger aggregates and were then internalized and assumed a perinuclear location. Confocal micrographs of control and IBOP-treated HEC revealed that IBOP treatment increased the number of perinuclear Cx43-positive particles from 3.8 Ϯ 1.8 to 7.5 Ϯ 2.2 particles/cell. We believe that the increased particles are internalized gap junction aggregates, which represent a redistribution of Cx43 from membrane to cytoplasm. The coalescence of Cx43 into larger aggregates preceded the internalization of Cx43 (25). The induction of larger aggregates by IBOP may be a preliminary step in the internalization process. Thus, the redistribution of Cx43 by IBOP may be the primary cause for the decrease in intercellular communication present in TXA 2 -treated HEC.
Migration can be inhibited through mechanisms that change neither integrins nor cytoskeleton; in particular, inhibition of migration caused by alteration of gap junction function is independent of changes in the cytoskeleton (19). We found that stimulation of HEC with the TXA 2 mimetic IBOP did not induce changes to the endothelial cell cytoskeleton after 24 h of treatment, in contrast to the observation that TXA 2 decreased stress fibers in endothelial cells within 15 min, which is associated with an increase in vascular permeability (47). The effects of TXA 2 on vascular permeability are more acute than those on intercellular communication, however, and the inhibition of TXA 2 on stress fiber formation are reversed at the time when TXA 2 most profoundly inhibits migration. The lack of effect on cytoskeletal morphology may also exclude the activation of Rho family members as the mechanism of the antimigratory effects of TP stimulation. TXA 2 mimetics also did not alter the surface expression of integrin complexes ␣ v ␤ 3 or ␣ v ␤ 5 . However, we cannot exclude the possibility that TP stimulation may cause changes in the expression of other integrins important for angiogenesis or the signaling by these integrins.
Thus, we have found that TXA 2 receptor stimulation suppresses HEC migration and tube formation in association with inhibition of gap junction-mediated intercellular communication. Intercellular communication is important in these events and thus potentially for angiogenesis and vessel wall function in general. TXA 2 has effects on endothelial cell biology that are unusual among G-protein-linked receptors and inhibits cell-cell coupling through an as yet undetermined mechanism. Further understanding of the biochemical basis of these events may lead to strategies to antagonize better the inhibitory effects of TXA 2 on endothelial cell function and thus potentially enhance re-endothelialization and vascularization in ischemic or thrombotic disease.