Dual Roles of Tight Junction-associated Protein, Zonula Occludens-1, in Sphingosine 1-Phosphate-mediated Endothelial Chemotaxis and Barrier Integrity*

In this report, sphingosine-1-phosphate (S1P), a serum-borne bioactive lipid, is shown to activate tight-junction-associated protein Zonula Occludens-1 (ZO-1), which in turn plays a critical role in regulating endothelial chemotaxis and barrier integrity. After S1P stimulation, ZO-1 was redistributed to the lamellipodia and cell-cell junctions via the S1P1/Gi/Akt/Rac pathway. Similarly, both endothelial barrier integrity and cell motility were significantly enhanced in S1P-treated cells through the Gi/Akt/Rac pathway. Importantly, S1P-enhanced barrier integrity and cell migration were abrogated in ZO-1 knockdown cells, indicating ZO-1 is functionally indispensable for these processes. To investigate the underlying mechanisms, we demonstrated that cortactin plays a critical role in S1P-induced ZO-1 redistribution to the lamellipodia. In addition, S1P significantly induced the formation of endothelial tight junctions. ZO-1 and α-catenin polypeptides were colocalized in S1P-induced junctional structures; whereas, cortactin was not observed in these regions. Together, these results suggest that S1P induces the formation of two distinct ZO-1 complexes to regulate two different endothelial functions: ZO-1/cortactin complexes to regulate chemotactic response and ZO-1/α-catenin complexes to regulate endothelial barrier integrity. The concerted operation of these two ZO-1 complexes may coordinate two important S1P-mediated functions, i.e. migration and barrier integrity, in vascular endothelial cells.

Endothelial barrier integrity is an important physiological function of the endothelium in vivo. Dysregulated barrier integrity is implicated in a variety of pathological conditions, such as stroke, inflammation, various immune responses, etc. (1). To elucidate the function and regulation of endothelial barrier integrity, cultured brain microvascular endothelial cells have been widely employed as an in vitro model system to study the blood-brain barrier (BBB) 2 (2). Evidence from these studies indicates that BBB plays a critical role in regulating the homeostatic environment of the brain and the transportation of plasma constituents into brain. Furthermore, it has been shown that severely impaired blood-brain barrier integrity is attributed to the pathological states of various neurological disorders, such as multiple sclerosis (3,4), Alzheimer disease (5,6), and human immunodeficiency virus-1-associated encephalitis or dementia (7,8).
Sphingosine 1-phosphate (S1P), a serum-borne bioactive lipid mediator secreted by activated platelets (9), enhances barrier formation in cultured pulmonary endothelial cells (ECs) (10). However, the molecular details for the formation and maintenance of endothelial barrier integrity are poorly understood. It was recently reported that the association of cortactin, an F-actin cross-linking polypeptide, and myosin light chain kinase is crucial in S1P-enhanced endothelial barrier integrity (11). Furthermore, it is well documented that tight junctions are important in regulating BBB formation (12,13). In addition, S1P greatly enhances VE-cadherin (VE-cad)-based adherens junctions in endothelial cells (14,15). Also, platelet-endothelial cell adhesion molecule-1 (PECAM), E-selectin, and intercellular cell adhesion molecules (ICAM) are regulated in endothelial cells with S1P treatment (16). Together, S1P-mediated endothelial barrier function may be the concerted activation of these molecules required for junctional structures and cell-cell interaction.
Here we demonstrate that S1P treatment not only significantly increases endothelial barrier integrity, but also greatly enhances cell migration into electrically wounded areas, as determined by electrical cell-substrate impedance sensing technology (ECIS) (17). In addition, both endothelial barrier integrity and chemotactic response were found to be controlled by the same signaling cascade, i.e. S1P1/Akt/Rac pathway. Thus, it suggests that a common modulator may be present to regulate both responses. In this report, evidence is shown that Zonula Occludens-1 (ZO-1), a tight junction-associated protein, is functionally critical in regulating S1P-mediated endothelial barrier integrity and chemotactic response. Data from this study also suggest that S1P induces the formation of distinct ZO-1 functional complexes, which in turn control the distinct physiological processes of the endothelium.

EXPERIMENTAL PROCEDURES
Materials, Cell Culture, and Adenoviral Transduction-Sphingosine 1-phosphate was obtained from Biomol. Pertussis toxin (PTx) and Ly294002 were from Calbiochem. Other reagents, unless specified, were from Sigma. Human Umbilical Vein Endothelial Cells (HUVECs) and Chinese Hamster Ovary (CHO-K1) cells stably expressing the S1P1 receptor were cultured essentially as described (18). Human brain microvascular endothelial cells (HBMECs, gift of Dr. Kwang Sik Kim, Johns Hopkins University) were cultured essentially as HUVECs. For adenoviral transduction, cells were infected with adenoviral particles carrying various cDNA constructs for 12 h prior to stimulation with S1P as previously described (18,19).
Western Blot Analysis-Proteins were separated on 10% SDS-PAGE and transferred to nitrocellulose membranes. After blocking with 5% nonfat milk, the membranes were blotted FIGURE 1. Real-time measurement of endothelial migration with ECIS assay. a, HUVECs were plated into each well of 8W1E arrays and incubated at 37°C to allow the TEER to reach equilibrium. Twenty-two hours later, the cells attached to the center microelectrodes were killed by applying 5 volts electrical current for 30 s (arrowhead), which resulted in an abrupt impedance drop. Subsequently, endothelial migratory responses were determined in real-time by measuring the recovery of TEER, an indicator of endothelial migration into the wounded area. b, micrographs were taken from ECIS arrays before (panel I), immediately after (panel II), 2 h after (panel III), and 4 h after (panel IV) wounding (arrows in a show the corresponding time points). Note that the high electrical current completely killed endothelial cells attached to the microelectrode (panel II), and the rise of TEER was a result of the surrounding viable cells migrating into the wounded electrodes (panels III and IV). The micrographs are representative of four ECIS wells at each time point. Scale bar, 125 M. with 1 g/ml of ZO-1 (Zymed Laboratories Inc.) or cortactin antibody (4F11, Upstate). Subsequently, the nitrocellulose membranes were incubated with their corresponding horseradish peroxidase-conjugated secondary antibody (Pierce) and visualized using the ECL method (Amersham Biosciences).
Transmission Electron Microscopy-HUVECs (1 ϫ 10 5 cells) were plated in tissue culture flasks (12.5 cm 2 , Falcon). Three days later, cells were starved in plain M199 for 2 h, followed by stimulating without or with S1P (500 nM, 30 min). Cells were fixed in 2.5% (v/v) glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4), postfixed in 1% osmium tetroxide-0.8% potassium ferricyanide in cacodylate buffer, treated with 0.5% aqueous uranyl acetate and embedded in epoxy resin. Thin sections were cut with a diamond knife, stained with uranyl acetate and lead citrate and observed in a Philips CM10 electron microscope at 60 kV.
Electrical Cell Substrate Impedance Sensing (ECIS) Assay-ECIS TM Model 1600R (Applied BioPhysics) was used to measure the transendothelial electrical resistance (TEER) in confluent or electrically wounded endothelial cells to study the S1Pregulated barrier integrity and migratory response, respectively FIGURE 2. S1P accelerates endothelial migration in ECIS wounding assay. a, endothelial cells on the microelectrodes of the ECIS wells were killed with a high electrical voltage as described. Subsequently, the migration of viable cells into the wounded microelectrodes was measured in real-time by electrical impedance in the presence or absence of S1P (500 nM). b, micrographs of ECIS arrays, taken at 3 h after electrical injury. ϪS1P and ϩS1P, without or with S1P addition, respectively. Scale bar, 125 M. c, cell migration in ECIS wounding assays was performed in endothelial cells stably transduced with si-RNA for S1P1 (si-S1P1) or Luciferase (si-Luc) with or without S1P stimulation. d, endothelial cells were transduced with adenoviral particles carrying ␤-galactosidase (␤-gal), dominant negative ϪAkt (dnAkt), and ϪRac (dnRac) cDNAs as described (200 moi each) (18). After killing the cells on the microelectrodes, the migratory responses were measured with or without S1P treatment. e, 50 g of extracts from endothelial cells transduced with si-Luc and two different si-S1P1 constructs (23 and 24) were directly immunoblotted with anti-S1P1 (upper panel). Lower panel, 500 g of extracts were immunoprecipitated with anti-S1P1 followed by immunoblotting with anti-S1P1. Densitometric quantitation showed that ϳ80 and 60% of S1P1 were knocked-down in si-S1P1 23 and 24 transduced endothelial cells. Middle panel, Western blot with anti-actin showed the protein equal loading and the absence of off-target effect by si-S1P1 silencing. f , in a parallel experiment, endothelial cells were transduced with 200 moi of adenoviral particles as described in panel d. 500 g of cell extracts were immunoprecipitated with anti-S1P1, followed by immunoblotting with anti-S1P1 (top panel). Alternatively, 50 g of extracts were immunoblotted with anti-Akt or anti-Myc to show the expression of transduced polypeptides (middle and lower panels) (18,19). Left lane, without endothelial extracts. The arrows in panels a, c, and d are when cells on the microelectrodes were killed by elevating the voltage, followed by stimulation with or without S1P. Panels a, c, and d are mean Ϯ S.E. of two determinants from a representative experiment, which has been repeated at least three times with identical results. (17). Briefly, to study the endothelial integrity function, 200 l of cell suspension (5 ϫ 10 5 cells per ml) were seeded to each well of either a 8W1E or 8W10 ECIS array (one electrode or 10 electrodes per well, respectively), which was pre-equilibrated with 200 l of medium at 37°C for 30 min. 2-3-days later, endothelial cells were serum-starved in plain M199 medium for 2 h at 37°C. After stimulation with S1P, the endothelial integrity function was measured in real-time as described (17). For migratory response, the cell-covered electrode was connected to an AC source. By applying 5 volts for 30 s, a relatively high current was delivered to the cell-covered electrode and killed the cells on the electrode (250 m diameter) (17). Thereafter, endothelial migratory responses were determined in real-time by measuring the recovery of electrical impedance, an indicator of the surrounding viable endothelial cells migrating into the wounded area, in the presence or absence of S1P treatment.

ZO-1 in S1P-regulated Endothelial Chemotaxis and Barrier Integrity
Migration Assay-Endothelial chemotaxis was measured by using the transwell migration assay (8 m pore size, Costar) as described (18). Briefly, polycarbonate membranes were coated with human fibronectin (1 g/ml) for overnight at 4°C. Following serum starvation, HUVECs were trypsinized, counted, and resuspended in M199 media with 0.5% fatty-acid free bovine serum albumin (Sigma). Cells (7.5 ϫ 10 4 cells in 100 l) were placed in the upper chamber and media (600 l) containing various concentrations of S1P were added to the lower chamber. Endothelial cells were allowed to migrate for 4 h at 37°C in a humidified chamber with 5% CO 2 . Subsequently, the nonmigrated cells on the upper side of the filter were removed with a cotton swab. The filters were fixed with 4% formaldehyde and stained with 0.1% crystal violet solution. Following eluting the stained dye with 10% acetic acid, cell chemotaxis was quantitated with BMG FluoStar Galaxy microplate reader (BMG Labtechnologies) at A 590 nm .
Statistic Analysis-Student's t test was used for statistic analysis. p-value Ͻ 0.001 was considered as statistically significant.

RESULTS
To study the dynamic behavior of endothelial cells in culture, the ECIS was used to measure the S1P-regulated TEER in real time (17). In this technology, as endothelial cells attach and spread on the small gold film electrode of the ECIS array, cellular membranes restrict the electrical current, forcing it to flow beneath and between the cells, resulting in a dramatic increase in electrical impedance (Fig. 1a). Furthermore, the endothelial cells attached to the circular microelectrode (250-m diameter) can be selectively killed by applying a high voltage electrical current (5 volts for 30 s, arrowhead in Fig. 1a). The efficiency of this electrical current to completely kill endothelial cells attached to the microelectrode was shown by an abrupt drop of FIGURE 3. S1P-mediated signaling regulates endothelial migration. a, chemotactic responses were measured by transwell migration assay as described under "Experimental Procedures." Note that S1P-induced chemotactic responses were significantly inhibited in both construct 23 and 24 si-S1P1 endothelial cells. b, Transwell migration assays were performed with ECs transduced with Adenoviral particles carrying ␤-galactosidase, dominant-negative ϪAkt and ϪRac vectors in the presence or absence of S1P stimulation. The data in a and b are the mean Ϯ S.E. of three determinants from a representative experiment, which was repeated two times with similar results. c, phase contrast micrographs show the S1P-enhanced repair of endothelial injuries. Wounds were introduced in confluent monolayers of wild-type (upper panels), ␤-galactosidase (middle panels), dominant-negative Akt or Rac (lower panels) transduced cells with sterile microtips. After stimulation without or with S1P for 16 h, cells were fixed. The distances (numbers in upper right, m) between two edges of wounds were quantitated with a Zeiss microscope equipped with Axiovert image software (Carl Zeiss). The data are the mean Ϯ S.E. of 6 microscopic fields from two determinants, which were repeated three times with similar results. SEPTEMBER 29, 2006 • VOLUME 281 • NUMBER 39 electrical impedance (Fig. 1a) and was verified by microscopic observation (panels I and II, Fig. 1b). Subsequently, the viable cells surrounding the wounded microelectrode were allowed to migrate into the electrically wounded area, and reestablished the cell-cell interaction over the electrode (panels III and IV, Fig. 1b). The kinetics for this sequence of events can be monitored in real-time by the increase of electrical impedance (Fig. 1a).

ZO-1 in S1P-regulated Endothelial Chemotaxis and Barrier Integrity
We utilized the ECIS technique to monitor the spatial and temporal behavior of S1P-regulated migration in repairing the endothelial injury. Endothelial cells attached to the circular microelec-trodes were killed by elevating electrical current and the surrounding viable cells were allowed to migrate into the wounded area in the presence or absence of S1P. As shown in Fig. 2a, S1P treatment significantly increased the TEER in the electrically injured endothelial monolayers. Microscopic examination showed that the raised TEER correlated well with the numbers of cells migrated into the wounded area (Fig. 2b). Moreover, this S1P effect on cell motility was markedly diminished in S1P1 knockdown ECs (si-S1P1) (Fig. 2c), indicating S1P1 receptor play a critical role in this process. The reduced but detectable S1P-responsiveness in si-S1P1 cells may be caused by the residual S1P1 receptor in knockdown cells (Fig. 2e). It was shown that S1P is able to activate Akt and Rac GTPase in endothelial cells (18), and S1P-regulated cytoskeletal remodeling and chemotactic response are dependent on Akt and Rac activities (14,18). As shown in Fig. 2d, the S1P-induced cell migration in ECIS wound-healing assay required Akt and Rac activities, because transduction with adenoviral particles carrying dominant-negative -Akt and -Rac polypeptides completely inhibited endothelial migration to repair the wounds; whereas, transduction with control ␤-galactosidase particles had no effect (Fig. 2, c and d).
Two alternative approaches were employed to confirm the capability of the S1P1/Akt/Rac pathway in regulating cell motility to repair the endothelial injury. As shown in Fig.  3a, S1P significantly enhanced endothelial migration in the transwell migration assay, and S1P-induced chemotaxis was markedly reduced in S1P1 knockdown endothelial cells. Moreover, transduction with dominant-negative Akt and Rac completely abrogated S1P-induced chemotactic response (Fig. 3b). In a control, transduction with ␤-galactosidase had no effect on S1P-enhanced endothelial chemotaxis. In addition, a small scrape wound was made across a confluent monolayer of HUVEC and cell migration into the denuded area was measured (Fig. 3c). S1P treatment significantly accelerated cell migration into the wounded area; whereas, wound closure in the absence of S1P was retarded (Fig. 3c, upper panels). Furthermore, the denuded area remained unpopulated in dominant-negative Akt and Rac . S1P signaling enhances transendothelial electrical resistance and induces tight junction formation. a, the TEER, an indicator of endothelial integrity function (10,17), was measured in real-time in confluent endothelial cultures treated without or with S1P (500 nM). b, TEER was determined in si-S1P1 and si-Luc cells in the presence or absence of S1P. c, HUVEC cells were pretreated with or without PTx (100 ng/ml, 1 h). After stimulation with S1P, TEER was measured in real-time. d, cells were pretreated with Ly294002 (10 M, 30 min), ␤-gal, dnAkt, and dnRac adenoviral particles (200 moi each). After stimulation without or with S1P, the TEER was measured. Panels a, b, c, and d are mean Ϯ S.E. of two determinants from a representative experiment, which has been repeated at least three times with identical results. e, electron microscopic analysis shows S1P induced tight junction formation in HUVECs (arrows). Scale bar, 0.226 m. Lower panel, random sections of control or S1P-treated samples were quantitated for tight junctions under electron microscope. Data represent the mean Ϯ S.E. from two samples of control or S1P-treated HUVECs, and ϳ15-40 sections were examined in each sample. **, p Ͻ 0.001 (Student's t test). transduced cells after S1P stimulation (Fig. 3c, lower panels). In a control, the denuded area was repopulated by ␤-galactosidase transduced endothelial cells in the presence of S1P treatment (Fig. 3c, middle panel). These results together suggest that the S1P-activated S1P1/Akt/Rac pathway is capable of enhancing cell motility to repair the injured endothelial monolayer.
We next utilized the ECIS assay to examine the effects of S1P on cell-cell interaction in confluent endothelial monolayers. As shown in Fig. 4a, a robust increase of TEER was observed immediately after S1P treatment, indicating that S1P was able to enhance endothelial cell-cell interaction. The S1P-induced cell-cell interaction was sustained for more than 6 -10 h after S1P treatment and thus was not a transient event. S1P-enhanced TEER was markedly abrogated in S1P1 knockdown endothelial cells (si-S1P1), and not in control cells stably expressing si-RNA for luciferase (si-Luc) (Fig. 4b). This indicates that S1P-enhanced TEER is primarily controlled by the endothelial S1P1 receptor. This conclusion was supported by the observation that S1P-enhanced endothelial barrier integrity was abrogated in the presence of pertussis toxin (PTx, Fig. 4c), as S1P1 signaling is dependent on the G i heterotrimeric G protein. Previously it was shown that S1P1 signaling activates the PI 3-kinase (PI3K)/Akt/Rac pathway to regulate endothelial chemotaxis and morphogenesis (18). Therefore, we examined whether PI3K/Akt/Rac signaling controls S1P-mediated TEER. As shown in Fig. 4d, treatment with Ly294002, an inhibitor of PI3K, or adenoviral particles carrying dominant-negative Akt or Rac constructs, significantly inhibited the S1P-enhanced TEER. Together, these results indicate that S1P-enhanced intercellular interaction is controlled by the S1P1-mediated G i /PI3K/Akt/Rac signaling pathway. S1P-induced increase in TEER is a consequence of enhanced barrier integrity in pulmonary endothelial cells (10,11). Because tight junctions are important in regulating barrier integrity, we examined whether S1P stimulates the formation of tight junctions in endothelial cells. Using electron microscopy analysis, we determined that S1P markedly enhanced the formation of tight junctions in neighboring endothelial cells (arrows, Fig. 4e). Tight junctions were quantitated in endothelial cells treated without or with S1P. Approximately 15-40 electron microscopic sections were examined in each sample. In S1P-treated endothelial cells, tight junction structures were identified in 72.06 Ϯ 11% of electron microscopic sections; whereas, tight junctions were identified in 50.51 Ϯ 4.5% of the examined sections in control endothelial cells (lower panel, Fig.  4e). This result indicates that the observed S1P-dependent increase in monolayer TEER is due at least in part to a stimulation of tight junction formation.
To investigate the molecular details of S1P-regulated integrity function and migratory response, endothelial cells treated without or with S1P were immunostained with antibodies for proteins functionally involved in cell-cell interaction and chemotaxis. These included ZO-1, claudin-5, Junctional adhesion molecule (JAM), PECAM, vascular/endothelial-cadherin, and cortactin. As shown in Fig. 5, PECAM, JAM, and VE-cad were significantly and exclusively re-distributed to cell-cell junction areas after S1P stimulation (arrows, Fig. 5), suggesting that these molecules may be involved in regulation of endothelial cell-cell interaction, but not functions at other sites, and thus contribute to S1P-enhanced endothelial integrity function. In contrast, ZO-1 and claudin-5 were translocated to both cell-cell contact areas and migratory fronts in S1P-treated ECs (Figs. 5 and 6a). Cortactin polypeptide was re-distributed to lamellipodia after S1P stimulation (Fig. 6b). This observation implies that ZO-1 and claudin-5 may be functionally important in regulating both endothelial barrier integrity and migratory response.
Interestingly, ZO-1 polypeptides were shown to be colocalized with cortical actin in the leading edges of migrating cells after S1P stimulation (arrowheads, Fig. 6a). It should be noted that the colocalization of ZO-1 and cortical actin was only observed at the cell migratory fronts and not in the cell junctional regions (arrows, lower panels, Fig. 6a). Because cortactin is an essential component in the formation of cellular cortical actin (20,21) and also because ZO-1 was shown to interact with cortactin (22), we examined whether S1P-mediated ZO-1 redistribution to the lamellipodia is dependent on cortactin FIGURE 5. S1P induces translocation of junctional or adhesion molecules. HUVEC cells were treated without or with S1P (500 nM, 30 min) and the cellular localization of the junctional or adhesion molecules were visualized by immunostaining. Note that PECAM, JAM, and VE-cad were markedly and exclusively relocated to cell-cell junctional areas after S1P treatment. In contrast, claudin-5, a tight junction molecule, was observed in both the lamellipodia and junctional regions. Arrows, cell-cell contacts; arrowheads, migratory fronts. Scale bar, 18 m. SEPTEMBER 29, 2006 • VOLUME 281 • NUMBER 39 polypeptide. This notion was supported by the observation that cortactin and ZO-1 polypeptides were co-localized in the lamellipodia (arrows, Fig. 6b), but not in the cellular junctions (arrowheads, Fig. 6b), in S1P-treated endothelial cells.

ZO-1 in S1P-regulated Endothelial Chemotaxis and Barrier Integrity
Previously, it was shown that the discs large (dlg)-like domain in the N-terminal half of ZO-1 polypeptide interacts with ␣-catenin (23). Therefore, we determined whether the S1P-mediated ZO-1 translocation to endothelial cell-cell contact regions is mediated by ␣-catenin association. As shown in Fig. 6c, S1P strongly stimulated the formation of zigzag-like junctional structures in endothelial cells. ZO-1 and ␣-catenin polypeptides were exactly colocalized in the newly formed junctional structures (arrows in Fig. 6c). This result suggests that the S1P-induced ZO-1 redistribution to endothelial junctional regions might be mediated at least in part by a ␣-catenindependent mechanism.
Next, we investigated the mechanisms of S1P-mediated lamellipodia localization of ZO-1 polypeptides. It has been demonstrated that ZO-1 interacts with the SH3 domain of cortactin via its C-terminal proline-rich region (22). Therefore, we predicted that ZO-1 lacking this region should be deficient in re-distribution to lamellipodia. We utilized CHO cells as a working model, because HUVECs are very difficult to transfect with plasmid DNA. As shown in Fig. 7a, the Myc-tagged fulllength ZO-1 polypeptides, when expressed in CHO cells stably expressing the S1P1 receptor (18), were translocated to the lamellipodia following S1P stimulation (arrows in top panels). Similarly, the S1P stimulated Myc-tagged C-terminal-half of ZO-1 (ZO-1(CH)) redistributed to the leading edges of the migrating cells (arrows in middle panels, Fig. 7a). In contrast, the N-terminal-half of the ZO-1 was unable to relocate to cell cortical actin areas in S1P-treated CHO/S1P1 cells (lower panels, Fig. 7a), despite the fact that lamellipodia were clearly formed (arrows in lower right panel, Fig. 7a). In addition, we knocked-down endothelial cortactin with cortactin antisense PTO, and examined the spatial behavior of ZO-1 in cortactin knockdown endothelial cells. As shown in Fig. 7b, transfection of cortactin antisense PTO markedly diminished endogenous cortactin (41 Ϯ 16% reduction, n ϭ 2) in human brain microvascular endothelial cells (HBMEC). Immunofluorescent analysis showed that S1P was unable to induce lamellipodia formation and cortactin translocation in antisense PTOtransfected cells (arrowhead in lower left panel, Fig. 7c). Furthermore, ZO-1 polypeptides were undetected in the migratory fronts (arrowhead, lower right panel) but were able to relocate to cell-cell junctional regions (black arrow, lower FIGURE 6. S1P induces colocalization of ZO-1/cortactin in the lamellipodia and ZO-1/␣-catenin in junctional areas. a, after stimulation without or with S1P (500 nM, 30 min), HUVECs were doubly stained with anti-ZO-1 and Phalloidin. Note that S1P significantly induced the formation of cellular junctions (arrows), and markedly relocated ZO-1 to cortical (arrowheads) and junctional (arrows) regions. b, HUVEC cells were doubly stained with rabbit anti-ZO-1 and mouse anti-cortactin in the absence or presence of S1P stimulation. Note that cortactin and ZO-1 polypeptides were colocalized in the S1P-induced lamellipodia, whereas cortactin was not observed in S1P-induced junctional regions. c, endothelial cells were stimulated without or with S1P, the subcellular localization of ZO-1 and ␣-catenin was detected by immunostaining with rabbit anti-ZO-1 and goat anti-␣-catenin. Note that ZO-1 and ␣-catenin were completely colocalized in S1P-induced endothelial junctional regions (arrows) . Scale bars in a, b, and c are 26, 32, and 36 m, respectively.

ZO-1 in S1P-regulated Endothelial Chemotaxis and Barrier Integrity
right panel) in antisense PTO-transfected cells after S1P stimulation. Together, it suggests that ZO-1, a polypeptide well documented in regulating cell-cell junctions, may also play a critical role in controlling endothelial migration via a cortactin-dependent mechanism.
We next examined whether S1Pinduced ZO-1 redistribution was governed by the S1P1/G i /Akt/Rac signaling pathway, which was demonstrated to be crucial in regulating S1P-mediated endothelial migratory response and barrier integrity (Figs. 2-4). The S1P-mediated redistribution of ZO-1 to the junctional regions and lamellipodia was markedly reduced by PTx treatment and dominant-negative Akt expression in HUVECs (Fig. 8). Akt-mediated phosphorylation at Thr-236 residue of S1P1 receptor is indispensable for S1P-induced Rac activation, endothelial migration, and morphogenesis (18). Here, we observed that Akt-mediated S1P1 receptor phosphorylation was a prerequisite for S1P-induced ZO-1 translocation, because the expression of the Akt phosphorylation-deficient S1P1 mutant (S1P1 T236A ) (18) inhibited ZO-1 redistribution to endothelial junctions and lamellipodia. Furthermore, the redistribution of ZO-1 was significantly abrogated by the expression of dominant negative -Cdc42 and -Rac GTPases. Therefore, the translocation of ZO-1 to both the endothelial junctional regions and migratory edges in S1P-stimulated endothelial cells is mediated by the G i /Akt/S1P1 phosphorylation/Cdc42 and Rac signaling cascade.
To further confirm the role of ZO-1 polypeptides in endothelial barrier integrity and chemotactic response, ZO-1 stably knockeddown endothelial cells were established by the Lentiviral-mediated si-RNA oligonucleotides technique. As shown in Fig. 9a, ZO-1 expression was significantly diminished in endothelial cells stably expressing si-ZO-1 constructs 1 and 2; whereas, si-Luciferase (si-Luc) had no effect. ZO-1 knockdown appears to be specific as the cortactin protein remained intact in si-ZO-1 cells. S1P-mediated endothelial chemotaxis was significantly abrogated in ZO-1 knockdown cells (Fig. 9b). Also, both S1P-mediated endothelial barrier integrity and migratory response in ECIS assays were markedly diminished in ZO-1 knockdown endothelial cells, compared with the control si-Luc cells (Fig. 9, c and d). Together, these FIGURE 7. S1P induces ZO-1 relocation to the lamellipodia by a cortactin-dependent mechanism. a, CHO cells stably expressing the S1P1 receptor were transfected with the Myc-tagged full-length (FL), C-terminal-half containing cortactin-interacting domain (CH), or N-terminal half without cortactin-interacting domain (NH) of ZO-1 cDNAs (23). After S1P stimulation, cells were immunostained with anti-Myc (left panels) and phalloidin (right panels) to show the ectopically expressed ZO-1 polypeptides and actin structures, respectively. Note that the Myc-tagged full-length and C-terminal half of ZO-1 polypeptides were redistributed to the lamellipodia (arrows in upper and middle panels), whereas the N-terminal-half of ZO-1 was unable to relocate to the lamellipodia. Scale bar, 21 m. b, HBMECs were co-transfected with sense (S-Cort.) or antisense (␣S-Cort.) PTO (100 nM), along with GFP cDNA by Lipofectamine reagent. Western blotting analysis shows that antisense PTO treatment results in a 41 Ϯ 16% reduction of cortactin (mean Ϯ S.E. of two determinants); whereas it does not have nonspecific effect on ZO-1 polypeptides. c, HBMECs were cotransfected with cortactin antisense PTO and GFP cDNA vector. After S1P stimulation, cells were immunostained with cortactin or ZO-1 antibody (lower panels). Upper panels, the transfected cells were identified by the expression of GFP polypeptides. Note that S1P treatment resulted in lamellipodia formation in the untransfected cells (white arrows). Also note that cortactin and ZO-1 were relocated to lamellipodia in the untransfected cells (white arrows). In contrast, neither lamellipodia formation nor translocation of cortactin and ZO-1 to lamellipodia was observed in antisense PTO-transfected cell (arrowheads, lower left, and right panels). However, S1P was able to redistribute ZO-1 to zigzag-like junctional structures in antisense PTO-transfected cells (black arrow, lower right panel). Scale bar, 18 m. SEPTEMBER 29, 2006 • VOLUME 281 • NUMBER 39 results strongly suggest that ZO-1 polypeptide plays a critical role in controlling endothelial barrier integrity and chemotactic responses.

DISCUSSION
By utilizing ECIS analysis on electrically injured or confluent endothelial cultures, S1P treatment was shown to significantly enhance the endothelial migration and barrier integrity (Figs. 2  and 4). These two S1P-mediated activities are regulated by the S1P1/Akt/Rac small GTPase pathway, which was previously demonstrated to play a critical role in regulating S1P-induced cytoskeletal remodeling and morphogenesis (14,18). It suggests that the regulatory molecule(s) downstream of S1P/Akt/ Rac signaling may be present to control these two endothelial activities. This regulatory molecule should be activated at both cell-cell junctions and migratory fronts in S1P-stimulated endothelial cells. To identify this molecule, the cellular localizations of several candidate molecules were examined in endothelial cells with or without S1P treatment. Among them, VE-cadherin, PECAM, and JAM were observed only in endothelial junctional areas (Fig. 5), suggesting they may be only functionally involved in S1P-induced cell-cell interactions and barrier integrity. In contrast, ZO-1 and claudin-5, well characterized molecules in the formation of tight junctions, were redistributed to both endothelial junctional regions and the lamellipodia. Furthermore, the S1P-stimulated ZO-1 redistribution to both endothelial junctions and lamellipodia was mediated by the S1P1/G i /Akt/Rac pathway (Fig. 8). Together, these results suggest tight junction-associated protein ZO-1 is one of the missing links in S1P-regulated endothelial chemotaxis and barrier integrity. This notion was supported by the fact that S1P-mediated endothelial migration and barrier integrity were markedly inhibited in ZO-1 knockdown endothelial cells (Fig. 9).
ZO-1 was first described as a component of epithelial tight junctions (24,25); it was also identified in nonepithelial cells, where it interacts with adherens junctions (23,26). Moreover, ZO-1 was shown to associate with connexin43 to regulate the plaque size and organization of gap junctions (27)(28)(29). Gap junctions are abundant in endothelial cells. However, we did not observe any detectable redistribution of connexin32 and connexin43 in endothelial cells after S1P stimulation (data not shown), suggesting that gap junctions may not contribute to S1P-induced TEER. Furthermore, we demonstrate that in addition to ZO-1, S1P also induced redistribution of Claudin-5, a well-characterized tight junctional polypeptide, to endothelial junctional areas (Fig. 5). Moreover, examinations with electron microscope directly demonstrated that S1P treatment stimulated tight junction formation (Fig. 4e). Because barrier integrity function is primarily mediated by intercellular tight junctions; the result of this study demonstrates for the first time that S1P/ S1P1 signaling is capable of regulating endothelial barrier integrity by stimulating the formation of tight junctions in endothelial cells.
Cortactin, an actin-binding polypeptide, redistributes to the leading edges of migratory endothelial cells and plays a critical role in S1P-mediated chemotaxis (11,30,31). Evidence presented in this study show that S1P induces ZO-1 and cortactin colocalization at endothelial lamellipodia (Fig. 6b), lamellipodia translocation of ZO-1 requires C-terminal half of ZO-1, which contains the cortactin-interacting domain (Fig. 7a), and cortactin knockdown inhibits S1P-induced lamellipodia relocation of ZO-1 (Fig. 7c). Furthermore, ZO-1 knockdown significantly abrogates S1Pstimulated endothelial chemotactic response (Fig. 9, b and d). Collectively, these data suggest that the lamellipodia localization of ZO-1 may be mediated by a cortactin-dependent mechanism and play a critical role in S1P-induced endothelial migratory response. However, the molecular mechanism underlying ZO-1 lamellipodia localization in regulating endothelial migration is not understood. Previously, it was shown that cortactin and Arp 2/3 complexes translocate to endothelial migratory fronts after S1P stimulation (31). Therefore, the lamellipodia localization of ZO-1 may modulate cortactin interacting partners and thus regulate the cytoskeletal architecture at the leading edges of migratory endothelial cells.
The cortactin polypeptide was not colocalized with ZO-1 in endothelial junctional regions (Fig. 6b), suggesting that S1Pmediated ZO-1 redistribution to endothelial cell-cell contact areas and the formation of endothelial junctions are mediated by a cortactin-independent mechanism. The molecular details of S1P-induced ZO-1 localization in the cellular junctions are currently unknown. ZO-1 and ␣-catenin interact directly (23,32,33) and the discs large (dlg)-like domain in the N-terminal half of the ZO-1 polypeptide is required for ␣-catenin association (23). In this study, we observed that S1P treatment resulted in the ZO-1 and ␣-catenin polypeptides being markedly redistributed to endothelial cell-cell junctional areas, where these two polypeptides were found to be colocalized (Fig. 6c). Therefore, this suggests that the redistribution of ZO-1 to endothelial junctional regions may be, at least in part, mediated by a mechanism involving ␣-catenin, and is independent of cortactin.
ZO-1 and ␣-catenin are two important components in the formation of tight junctions and adherens junctions, respec-  constructs 1 and 2). Cell extracts were Western-blotted with anti-ZO-1 (upper panel) and anti-cortactin (lower panel). Note that the expression of ZO-1 polypeptides in HUVECs stably transduced with si-ZO-1 constructs was significantly inhibited, whereas the expression of cortactin was not affected. Data are mean Ϯ S.E. of four determinants. b, chemotactic responses were performed with transwell migration assay as described under "Experimental Procedures." Note that the S1P-induced chemotactic responses were significantly inhibited in both constructs 1 and 2 si-ZO-1 endothelial cells. The figure is the mean Ϯ S.E. of three determinants from a representative experiment, which was repeated two times with similar results. S1P-induced endothelial integrity (c) and chemotactic response (d), measured by ECIS assays, were markedly reduced in si-ZO-1 knockdown endothelial cells. The arrow in panel c indicates the addition of S1P (500 nM), and in panel d indicates when cells were killed by elevating the voltage and treated with S1P. Panels c and d are mean Ϯ S.E. of two determinants from a representative experiment, which has been repeated at least three times with identical results. SEPTEMBER 29, 2006 • VOLUME 281 • NUMBER 39 tively (23,32,33). Previously, it was shown that the S1P-activated Rho family small GTPases enhanced VE-cadherin based adherens junction formation in endothelial cells (14). In this study, we demonstrated that S1P stimulation markedly enhanced the formation of endothelial tight junctions (Fig. 4e). The observed ZO-1 and ␣-catenin colocalization in junctional regions implies that tight junctions and adherens junctions may be cross-regulated and may concertedly mediate S1P functions in vasculature. Furthermore, PECAM and JAM polypeptides were significantly enriched in cell-cell contact areas in S1Ptreated endothelial cells (Fig. 5). Thus, S1P-regulated endothelial barrier integrity function may be a biological manifestation of concertedly regulated events by various cell-cell interaction machineries, e.g. tight junctions, adherens junctions, PECAM-, and JAM-mediated intercellular interaction.