Effects of Cytotoxic Necrotizing Factor 1 and Lethal Toxin on Actin Cytoskeleton and VE-Cadherin Localization in Human Endothelial Cell Monolayers

Abstract

Integrity of the vascular endothelium is largely dependent on endothelial cell shape and establishment of intercellular junctions. Certain pathogenic bacterial toxins alter the cytoskeletal architecture of intoxicated cells by modulating the GTPase activity of p21 Rho family proteins. In the present study we have analyzed the effect of Rho-directed toxins on the actin cytoskeleton and monolayer integrity of endothelial cells. We report here that Escherichia coli cytotoxic necrotizing factor 1 (CNF1) activates Rho in human umbilical vein endothelial cells (HUVEC). In confluent monolayers, CNF1 treatment induces prominent stress fiber formation without significantly modifying peripheral localization of VE-cadherin, a specific marker of vascular endothelial cell adherens junctions. Further, Rho activation with CNF1 blocks thrombin-induced redistribution of VE-cadherin staining and gap formation in HUVEC monolayers. Inhibition of Rho by prolonged treatment of cells with C3 exoenzyme (Clostridium botulinum) eliminates actin stress fibers without disrupting the continuity of VE-cadherin staining, indicating that Rho-dependent stress fibers are not required for maintaining this adhesion receptor at sites of intercellular contact. Lethal toxin (Clostridium sordellii), an inhibitor of Rac as well as Ras and Rap, potently disrupts the actin microfilament system and monolayer integrity in HUVEC cultures.

Endothelial cells, located at the interface between blood and tissues, form the principal permeability barrier by gating the traffic of molecules and cells across the vessel wall. They also constitute an important target for blood-borne pathogens seeking access to underlying tissues. Consequently, the endothelium plays a determinant role in various biological functions, including hemostasis, inflammation and immunity. Several lines of evidence link the actin cytoskeleton of endothelial cells to endothelial barrier function.

VE-cadherin, also referred to as cadherin 5, a specific endothelial transmembrane glycoprotein located at cell-cell contacts, links the cell membrane to the actin cytoskeleton. This cell adhesion receptor which mediates homophilic calcium-dependent intercellular contacts has been shown to control endothelial permeability (21). Catenins (α-catenin, β-catenin, and plakoglobin, also called γ-catenin) are the proteins that mediate the attachment of the cytoplasmic tail of VE-cadherin to actin filaments (see reference 9 for a review). In Chinese hamster ovary cells it has been proposed that association of cadherin-catenin complexes with the actin cytoskeleton enhances adhesion strength of homophilic interactions (6). Recently, using embryonic bodies derived from VE-cadherin-negative mouse embryonic stem cells, it was demonstrated that expression of VE-cadherin is required for the assembly of endothelial cells into vascular-like structures (39). These results suggest an important role for VE-cadherin in structural organization of vascular endothelial monolayers.

The Ras-related subfamily of Rho proteins, including Rho, Rac, and Cdc42, is known to play a crucial role in controlling the actin cytoskeleton. Clues to the specific functions of the different Rho GTPases have largely been obtained by examining the actin cytoskeleton of mammalian fibroblasts which overexpress wild-type or mutant (constitutively active and dominant negative) forms of these proteins (26, 32, 33). In addition, bacterial toxins have been extremely useful tools for probing the role of the Rho family proteins in cellular signaling processes. In particular, C3 exoenzyme (C3) from Clostridium botulinum has been used to specifically inhibit Rho function by ADP-ribosylating residue Asn41 of Rho (A, B, and C), which is located in a putative effector domain (36). Using this enzyme, Chardin et al. (8) were first to observe that Rho controls the process of stress fiber assembly and disassembly. Since then, many investigators have taken advantage of this toxin to shed light on multiple Rho functions in cells, including cell growth, cytoskeleton assembly, and intracellular trafficking.

Whereas C3 is highly specific for Rho, lethal toxin produced by Clostridium sordellii has been found to inhibit, at least in vitro, Rac, Ras, and Rap by glucosylation of threonine 35 (20, 30). In cultured cells, lethal toxin blocks Ras-dependent activation of the mitogen-activated protein kinase cascade by epidermal growth factor (20, 30) and induces cell rounding, formation of numerous cell surface filopodia and loss of actin stress fibers (15, 29).

More recently, activating toxins specific for Rho, including cytotoxic necrotizing factor 1 (CNF1) produced by certain Escherichia coli strains that cause infections in humans, have been described. Activation of Rho by CNF1 occurs via a novel mechanism involving the deamidation of Gln63 on Rho (13, 34). In vivo, CNF1 toxicity has been causally linked to pathological states, including diarrhea and urinary infections (for a review, see reference 35). In epithelial cell systems, the toxin has been shown to induce multinucleation, the accumulation of thick stress fibers, and spreading (13), as well as membrane ruffling (11) and modification of microvillus structure (17). Morphologic effects of CNF1 have also been described in human monocytic cells (7).

In the present study, we have analyzed the effect of CNF1, C3, and lethal toxin on human endothelial cells, using confluent HUVEC monolayers as a model. In particular, we have examined the functional consequences of Rho activation or inhibition on the actin cytoskeleton and on monolayer integrity using VE-cadherin as a marker of endothelial adherens junctions.

MATERIALS AND METHODS

Materials.

Tissue culture plasticware was from Life Technologies (Cergy Pontoise, France). Human α-thrombin (3,209 NIH units/mg) was a generous gift of J. W. Fenton II (New York State Department of Health, Albany). Monoclonal anti-VE-cadherin antibody BV9 was kindly supplied by E. Dejana (Mario Negri Institute, Milan, Italy). Triton X-100, saponin, gelatin, fibronectin, and fluorescein isothiocyanate (FITC) were from Sigma (L’Isle d’Abeau, France). UDP-[¹⁴C]glucose and ³²P-NAD were from NEN Research Products.

Cells and culture conditions.

HUVECs were isolated from umbilical cord veins by collagenase perfusion as previously described (2). Cells were grown on 1% gelatin-coated 10-mm-diameter dishes in EBM-2 medium supplied by BioWhittaker (Walkersville, Md.). Cells were used between the first and the fourth passages and maintained at 37°C in 5% CO₂.

Bacterial toxins.

Highly purified CNF1 from a uropathogenic E. coli strain was obtained as described by Falzano et al. (11). Recombinant C3 was expressed as a His-Tag fusion protein in the E. coli strain BL21DE3 and prepared as described in reference 8. Lethal toxin from C. sordellii was obtained from culture supernatant of the pathogenic IP82 strain and purified to homogeneity as described previously (29).

Localization of actin filaments and VE-cadherin.

HUVECs (2 × 10⁵) were plated on 17-mm-diameter Lab-Tek chamber slides (Poly Labo, Strasbourg, France) coated with fibronectin (7 μg/ml) and grown to confluence. Five days later, cells were treated with the bacterial toxins and fixed in a 3% paraformaldehyde–2% sucrose solution for 15 min at room temperature. After three washes in phosphate-buffered saline containing Ca²⁺ and Mg²⁺ (PBS²⁺), cells were permeabilized with 0.2% Triton X-100 for 3 min. Triton was removed by extensive washing in PBS²⁺ and nonspecific binding sites were saturated with PBS²⁺ containing 10% fetal calf serum for 15 min. Then, slides were incubated in the same solution containing anti-VE-cadherin antibody from Transduction Laboratory (clone C26120) or the BV9 monoclonal antibody (21) for 1 h at room temperature in a humid chamber. The second antibody conjugated to Texas red (dilution, 1/200) and FITC-phalloidin (90 ng/slide) were added for an additional 45 to 60 min at room temperature. Slides were then washed 3 times in PBS²⁺, dried, mounted in Mowiol (Calbiochem), and photographed with a fluorescence-equipped photomicroscope (Nikon Diaphot).

Mobility shift of Rho upon treatment with CNF1.

HUVEC cultures grown to confluence in six-well plates were incubated for 16 h with 10⁻⁹ M CNF1 and washed once in PBS before detachment of cells with a rubber policeman. After low-speed centrifugation, cells were resuspended in 50 μl of ADP-ribosylation buffer (20 mM Tris [pH 7.8], 2 mM MgCl₂, 5 mM dithiothreitol) and lysed by four freeze-thaw cycles. After quantification of the proteins, ADP-ribosylation was performed on 50 μg of total proteins by adding 10 × 10⁻⁹ M purified C. botulinum C3 as described in reference 8 and 2 μCi of ³²P-NAD to each sample for 2 h at 37°C. Ten microliters of sample buffer was added to stop the reaction and each sample was boiled and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis on 15% gels. After staining with Coomassie blue, radioactivity was analyzed by autoradiography.

Measurement of F-actin content.

HUVECs were cultured to confluence in 60-mm-diameter dishes and incubated with the toxin for the indicated times. Cells were then washed twice in PBS and collected with a rubber policeman in 100 μl of F-actin buffer {20 mM KPO₄, 10 mM PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid)], 5 mM EGTA, 2 mM MgCl₂, 0.1% Triton X-100, 3.7% formaldehyde, and 2 μM FITC-phalloidin}. Remaining cells were scraped off in an additional 100 μl of F-actin buffer. Cells were pooled in Eppendorf tubes, incubated for 1 h on a rotating platform at room temperature, and centrifuged for 2 min at 10,000 rpm in an Eppendorf centrifuge. Triton-insoluble pellets were washed twice (0.1% saponin, 20 mM KPO₄, 10 mM PIPES, 5 mM EGTA, 2 mM MgCl₂) and then resuspended in 1.5 ml of ice-cold methanol. Methanol extraction resulted in the removal of phalloidin bound to F-actin. After 1 h on a rotating platform at room temperature, methanol suspensions were collected and the amount of FITC was determined in each sample by using a fluorimeter (fluorescence emission at 520 nm and excitation at 492 nm) and expressed as a percentage of that of the control. Protein determinations were carried out by the bicinchoninic acid protein assay (Pierce Chemical Co., Rockford, Ill.), and identical amounts of protein in each sample were analyzed.

RESULTS

Effect of Rho-directed bacterial toxins on the actin cytoskeleton of HUVECs.

Confluent HUVEC monolayers were treated with CNF1 or C3, and then polymerized actin was examined by fluorescence microscopy after the cells were stained with FITC-phalloidin. As shown in Fig. 1A (middle), overnight treatment with CNF1 increases the overall intensity of staining and induces the appearance of dense actin stress fibers across cells, a hallmark of Rho activation. We also observed enhanced labeling of actin between adjacent cells. This effect was maximal at 10⁻⁹ M CNF1, the highest concentration of the toxin tested. It has previously been observed that deamidation of Rho Glu63 by CNF1 is accompanied by a decrease in the electrophoretic mobility of the protein during sodium dodecyl sulfate-polyacrylamide gel electrophoresis (13, 34). Therefore, we examined whether the treatment of HUVECs with CNF1 resulted in a similar decrease in the electrophoretic mobility of Rho. To do so, Rho present in lysates from toxin-treated cells was labeled with recombinant C3 in the presence of ³²P-NAD. As shown in Fig. 1B (top, lane 2), CNF1 pretreatment of HUVEC cultures slightly retards the migration of Rho. A similar shift in electrophoretic mobility was observed with purified RhoA following treatment in vitro with CNF1 (lane 4). Note that recombinant RhoA migrates more slowly than the endothelial cell-derived Rho, due to the fact that the protein purified from bacteria is not geranylated (not shown). Interestingly, we found that CNF1 does not significantly modify cellular F-actin content in treated cells as compared to nontreated monolayers, suggesting that stress fiber formation induced by Rho activation may not require actin polymerization per se (Fig. 1B, bottom). Rather, it may activate bundling of preexisting stress fibers as suggested by the experiments of Machesky et al. (24). In this assay, cytochalasin D treatment of cells, under conditions which completely abolish F-actin staining, decreases cellular F-actin content by approximately 40% (data not shown).

FIG. 1.

Effect of CNF1 and C3 on the actin cytoskeleton of HUVECs. (A) CNF1 induces stress fiber formation. Confluent HUVEC monolayers were treated with medium alone (control), 10⁻⁹ M CNF1 for 16 h, or 4 × 10⁻⁷ M C3 for 48 h, and F-actin was monitored as described in Materials and Methods. Results (magnification, ×830) are representative of four independent experiments. (B) (Top) CNF1 pretreatment modifies Rho. Shown are the results of electrophoretic mobility shift of Rho from nontreated cells (lane 1), cells treated with 10⁻⁹ M CNF1 (lane 2), purified RhoA (lane 3), or CNF1-treated purified RhoA (lane 4). Low-molecular-weight standards (in thousands) from Bio-Rad are indicated on the left. (Bottom) Assay of F-actin content in cells treated with 2 × 10⁻⁹ M CNF1 for 3 or 6 h was performed as described in Materials and Methods. Mean values (+ error) from two independent experiments are shown.

Selective inhibition of Rho in HUVEC cultures with C3 leads to actin depolymerization. Breakdown of the cytoskeleton is striking after 24 h of treatment with 4 × 10⁻⁷ M C3 and by 48 h we could not detect any stress fibers; only a faint ring of polymerized actin remained visible at the periphery of cells (Fig. 1A, bottom).

Lethal toxin from C. sordellii, which has no effect on Rho but inhibits the Rho-related protein Rac as well as Ras and Rap, was also found to exert a profound effect on HUVEC cultures (Fig. 2). Overnight treatment with low doses of the toxin (4 × 10⁻¹² M) leads to a modest disorganization of the cytoskeleton accompanied by the appearance of relatively small gaps between neighboring cells and an increase in the number of stress fibers, as compared to nontreated cells (Fig. 1). Stress fiber formation is enhanced further with 40 × 10⁻¹² M lethal toxin. At this concentration toxin-treated cells become elongated and large gaps appear in the monolayer. During a 16-h treatment with 400 × 10⁻¹² M lethal toxin, most cells become round and detach from the plate. Those cells that remain attached following the incubation (Fig. 2, bottom) are highly fluorescent and possess prominent stress fibers.

FIG. 2.

Lethal toxin disrupts actin cytoskeleton. Confluent HUVEC monolayers were treated with the indicated concentrations of lethal toxin for 16 h and actin filaments were visualized as described in Materials and Methods. Results (magnification, ×840) are representative of two independent experiments.

Effect of Rho activation or inhibition on localization of VE-cadherin.

To determine the consequences of endogenous Rho activation or inhibition by the toxins on integrity of the endothelial monolayer, we examined the staining of VE-cadherin, a specific marker of vascular endothelial adherens junctions. As previously described (21) and shown in Fig. 3, a characteristic honeycomb pattern is obtained upon VE-cadherin staining of nontreated confluent HUVEC monolayers. Although staining between cells is noninterrupted, we observed short perpendicular extensions and (at a higher magnification) lacelike fringes in some areas. When cells are treated with CNF1, the pattern of VE-cadherin staining is not significantly modified, indicating that Rho activation does not trigger redistribution of VE-cadherin.

FIG. 3.

Rho activation by CNF1 does not perturb VE-cadherin localization. Confluent monolayers were pretreated or not pretreated with 10⁻⁹ M CNF1 for 16 h and fixed in paraformaldehyde solution as described previously. Actin cytoskeleton and VE-cadherin staining are shown. Typical results (magnification, ×520) from three independent experiments are presented.

We next examined whether CNF1 treatment would modify disruption of the monolayer by vascular permeabilizing agents. To do so, we analyzed the effect of CNF1 pretreatment on thrombin-stimulated cells. Thrombin has previously been shown to compromise endothelial barrier function in vivo and increase monolayer permeability in cultured endothelial cells (for a review, see reference 23). In our HUVEC system, as we and others have observed (31, 40), thrombin promotes stress fiber formation followed by cell retraction and disruption of intercellular junctions. Under these conditions, VE-cadherin staining becomes discontinuous and acquires a zigzag pattern, with label remaining only at sites of residual cell-cell contact (arrow in Fig. 4). Interestingly, we observed that pretreatment of HUVECs with CNF1 completely blocked the effect of thrombin on VE-cadherin redistribution and formation of gaps in the monolayer. Constitutive activation of Rho with CNF1 was also found to inhibit thrombin-stimulated hyperpermeability of the monolayer to macromolecules such as FITC-dextran (data not shown).

FIG. 4.

CNF1 pretreatment inhibits thrombin-induced disruption of monolayer integrity. Confluent HUVECs were treated or not treated with 10⁻⁹ M CNF1 for 16 h prior to the addition of thrombin (1 U/ml) during the last 15 min of the incubation. Reorganization of the actin cytoskeleton and VE-cadherin localization were analyzed as previously described. The arrow highlights the thrombin-induced disruption of VE-cadherin staining. Results (magnification, ×830) are representative of three independent experiments.

We then analyzed the consequence of Rho inhibition on the localization of VE-cadherin. Following a 48-h treatment with recombinant C3, in the absence of stress fibers, VE-cadherin staining remains qualitatively similar to that observed in nontreated cultures. Importantly, staining around cells is noninterrupted. We did note in toxin-treated cells that staining is finer and smoother than in control cells, as illustrated in Fig. 5 (top right). In addition, the intracellular space of C3-treated cells is strictly devoid of staining and cells appear larger and more spread than control cells.

FIG. 5.

Effect of C3 and lethal toxin on localization of VE-cadherin. Confluent HUVEC monolayers were treated with 4 × 10⁻⁷ M C3 for 48 h or 4 × 10⁻¹² M lethal toxin for 16 h. F-actin and VE-cadherin staining were analyzed. Arrowheads indicate the breakdown of VE-cadherin staining. Results (magnification, ×530) are representative of three independent experiments.

In contrast, pretreatment of cells with a low dose of lethal toxin (4 × 10⁻¹² M) results in a pronounced disorganization of VE-cadherin staining. As shown in Fig. 5 (bottom right) the overall intensity of fluorescence is higher than in control cultures, yet VE-cadherin staining is interrupted by gaps which correspond to areas of cell retraction. Since lethal toxin does not target Rho, these results suggest that the other p21 GTPases inhibited by the toxin, such as Rac or Ras, play an important role in the maintenance of junctional contacts in endothelial cells.

DISCUSSION

Using bacterial toxins directed against endogenous Rho proteins we have examined the role of p21 Rho GTPases in controlling the actin cytoskeleton and monolayer integrity of human endothelial cells. To our knowledge, this is the first report describing the effects of CNF1 and lethal toxin on primary cultured human endothelial cells. The HUVEC system constitutes a well-characterized in vitro model to study endothelial barrier function. It is noteworthy that we have observed that mouse pulmonary artery endothelial cells are equally sensitive to the morphological effects of these toxins (41).

Functional activation of Rho has been described for E. coli toxins CNF1 and -2 as well as for dermonecrotizing toxin from Bordetella bronchiseptica (12, 19, 28). The molecular mechanism of activation has been elucidated, in the case of both CNF1 (13, 34) and dermonecrotizing toxin (18). Deamidation of Gln63 on Rho by the toxin converts it into a glutamic acid, leading to inhibition of both intrinsic GTP hydrolysis and GTP hydrolysis stimulated by its GTPase-activating protein. Thus, CNF1-induced deamidation of Gln63 renders Rho constitutively active in intoxicated cells and allows the persistent formation of stress fibers. Indeed, we have demonstrated here that CNF1 promotes prominent stress fiber bundling as well as F-actin accumulation at junctional borders in endothelial cells. Further, this toxin protects resting HUVEC monolayers from cytoskeletal remodeling by barrier-disruptive agents such as thrombin. A protective effect of CNF1 against agonist-induced barrier dysfunction was also observed in a thrombin-responsive HUVEC line, EA.hy926 cells (results not shown). In terms of pathogenesis, the paralyzing action of CNF1 on blood vessels may account in part for the tissue-damaging effect of this toxin in vivo. In rabbits, CNF1 induces a strong necrotic reaction after intradermal injection and high lethality in mice after systemic inoculation.

It has previously been established that Rho regulates tight-junction permeability and perijunctional actin reorganization in intestinal epithelial cells, with C3 used to inhibit the GTPase (27). More recently, the consequences of Rho activation with CNF1 on epithelial barrier function have been examined. In polarized T84 epithelial intestinal cell monolayers, the toxin did not influence transepithelial resistance whereas it attenuated transepithelial migration of polymorphonuclear leukocytes (17). Using 10-fold-higher concentrations of CNF1, Gerhard et al. observed an increase in transepithelial resistance in Caco-2 cell monolayers (14). These authors have suggested that Rho activation, like Rho inhibition, can alter barrier function of intestinal tight junctions, although they also reported that in their hands CNF1 can modify Rac and Cdc42 in addition to Rho in intact cells (14).

Concerning junctional integrity, we show here that modulation of Rho activity in HUVEC monolayers (either constitutive activation or inhibition) does not perturb intercellular labeling of VE-cadherin, suggesting that Rho activation is not essential for restriction of this endothelial cell cohesive receptor to sites of cell-cell contact. This observation is intriguing given the fact that it has previously been reported by Braga et al. that Rho and Rac are required for the establishment of cadherin-mediated cell-cell adhesion in epithelial cells (5). During the preparation of our manuscript, this same group has extended their studies to the role of Rho and Rac in control of cell-cell adhesion in endothelial cells (4). Consistent with our findings, they report that VE-cadherin localization at endothelial junctions is independent of Rho and Rac and that the insensitivity of VE-cadherin to inhibition of these GTPases is not due to the maturation status of the junctions. Rather, they present evidence showing that the cellular context plays an important role in regulation of cell-cell adhesion by cadherins (4).

Our finding that Rho activation protects monolayers from thrombin-induced cell shape changes and disruption of cell-cell junctions is in apparent contradiction with a recent study suggesting that Rho participates in the thrombin-induced cell contraction and monolayer permeability in HUVEC cultures (10). Indeed, thrombin stimulates the rapid formation of stress fibers in endothelial cells, indicative of Rho activation. However, Rho activation alone may not be sufficient for the disruptive effect of thrombin on endothelial cell cultures. In addition to stress fiber formation, we observe a slower accumulation of actin at cell peripheries followed by retraction and rounding, in some cells. The latter effect can be blocked by overexpression of a dominant interfering mutant of Rac in a HUVEC line, suggesting that Rac also participates in cytoskeletal remodeling by thrombin (40). Consistent with this hypothesis, we observe a potent disruptive effect of lethal toxin on HUVEC monolayers, as discussed below, pointing to a role for other p21 GTPases in the control of monolayer integrity. Therefore, although Rho inhibition may compromise thrombin’s effect by blocking stress fiber formation, results shown here using CNF1 indicate that Rho activation alone is not sufficient to mimic the effect of thrombin. It is not surprising that CNF1 protects barrier function in endothelial monolayers, since the assembly of focal contacts and stress fibers is associated with cell spreading and inhibition of cell motility.

Similar to Essler et al. (10), we observed that a pretreatment of HUVECs with C3 breaks down the actin cytoskeleton and prevents stress fiber formation by thrombin. Nonetheless, we do observe some round cells in C3-treated cultures following thrombin addition (data not shown). At least two explanations could account for these findings. First, we and others have shown that C3 treatment inhibits endogenous Rho in endothelial cells by only 80%. It could be argued that the remaining 20% is sufficient for limited cell rounding to occur. Second, as detailed above, it is possible that additional, Rho-independent signaling pathways are involved in this effect. Unfortunately, bacterial toxins that selectively inhibit Rac or Cdc42 GTPases have not been identified to date.

Lethal toxin, a glucosyltransferase causally associated with severe edemas, was found to be a potent disruptive agent in human endothelial monolayers. Overnight treatment with 400 × 10⁻¹² M toxin resulted in extensive cell rounding and detachment, similar to the previously described effect of the toxin on different cell types (1, 3, 15). Interestingly, 10- to 100-fold-lower doses of lethal toxin induced cell elongation with noticeable increases in stress fiber formation rather than a decrease in actin microfilaments, which has been previously documented. Although the primary target(s) of this toxin in endothelial cells has not been established, it has been demonstrated in vitro that lethal toxin glucosylates and inactivates Ras, Rap, and Rac while having no effect on Ral, Rho, Cdc42, Arf1, or Rab (20, 30). Therefore, the link (if any) between Ras, Rac, or Rap inhibition and the observed cellular modifications remains to be defined.

It has been proposed that the Rho GTPases control the activation of the ezrin, radixin, and moesin (ERM) family of proteins, which cross-link actin filaments with plasma membrane (reviewed in reference 38). Moesin has been identified as a key component of Rho and Rac signaling pathways in Swiss 3T3 cells (25). In a permeabilized cell system, the addition of moesin allows stress fiber assembly, cortical actin polymerization, and focal complex formation in response to activated Rho and Rac. ERM proteins also appear to be involved in cell adhesion processes since expression of antisense oligonucleotides for ERM proteins perturbs cell adhesion (37). Further, it has been shown that ERM proteins interact with intercellular adhesion molecules 1 and 2 (16). It remains to be determined whether these proteins bind to other adhesion molecules involved in the maintenance of the endothelial lining of blood vessels, such as CD31 (PECAM) or cadherins.

Both molecular (yeast two-hybrid system) and biochemical (affinity purification) techniques have been successfully used to identify proteins that interact with activated Rho family GTPases (for a review, see reference 22). It remains to be established which of these potential effector molecules are present in endothelial cells and what their precise roles in governing endothelial cell morphology may be.

Increased vascular permeability accompanies inflammatory responses and normal wound healing processes and is a hallmark of inflammatory disorders. The use of bacterial toxins directed against various members of the p21 Rho family has permitted us to highlight the importance of Rho-dependent stress fiber formation in the maintenance of vascular barrier function. These findings should contribute to our understanding of the pathological consequences of infection caused by pathogenic bacteria producing these Rho-directed toxins.