Rho and rho kinase modulation of barrier properties: cultured endothelial cells and intact microvessels of rats and mice

Abstract

Previous experiments using cultured endothelial monolayers indicate that Rho-family small GTPases are involved in modulation of endothelial monolayer permeability by regulating assembly of the cellular actin filament scaffold, activity of myosin-based contractility and junctional distribution of the Ca²⁺-dependent endothelial cell adhesion molecule, VE-cadherin. We investigated these mechanisms using both cultured endothelial cells (from porcine pulmonary artery and mouse heart) and vascular endothelium in situ (mouse aorta, and individually perfused venular microvessels of mouse and rat mesentery). Exposure to Clostridium difficile toxin B (100 ng ml⁻¹) inactivated 50–90 % of all endothelial Rho proteins within 60–90 min. This was accompanied by considerable reduction of actin filament stress fibres and junctional F-actin in cultured endothelial monolayers and in mouse aortic endothelium in situ. Also, VE-cadherin became discontinuous along endothelial junctions. Inhibition of Rho kinase with Y-27632 (30 μm) for 90–120 min induced F-actin reduction both in vitro and in situ but did not cause redistribution or reduction of VE-cadherin staining. Perfusion of microvessels with toxin B increased basal hydraulic permeability (L_p) but did not attenuate the transient increase in L_p of microvessels exposed to bradykinin. Perfusion of microvessels with Y-27632 (30 μm) for up to 100 min reduced basal L_p but did not attenuate the permeability increase induced by platelet activating factor (PAF) or bradykinin. These results show that toxin B-mediated reduction of endothelial barrier properties is due to inactivation of small GTPases other than RhoA. Rho proteins as well as RhoA-mediated contractile mechanisms are not involved in bradykinin- or PAF-induced hyperpermeability of intact microvessels.

The aim of the experiments performed in the present study was to test whether either basal permeability or mediator-induced hyperpermeability of in vivo vascular endothelium is dependent on a Rho protein-signalling pathway.

During the past several years, the Rho family of small GTPases (Rho, Cdc42 and Rac) has been shown to play a key role in the control of the assembly of the actin-based cytoskeleton and in regulation of cadherin-based intercellular junctions (Tapon & Hall, 1997; Fukata et al. 1999). Recently published studies indicating that the Rho pathway also appears to be important for regulation of barrier properties of cultured endothelial monolayers derived from various microvascular sources are reviewed below.

Treatment of cultured porcine pulmonary arterial endothelial cells (PAECs) with Clostridium difficile toxin B (Hippenstiel et al. 1997), an agent that inactivates all members of the Rho protein family (Just et al. 1995; Busch & Aktories, 2000), resulted in dose-dependent dissociation of cell-to-cell contacts and hyperpermeability of the monolayer (Hippenstiel et al. 1997). This increased permeability was assumed to result from the breakdown of the cellular actin filament scaffold after prolonged toxin action.

In order to understand which member of the Rho-protein family is involved in endothelial barrier regulation Wójciak-Stothard et al. (1998) and Carbajal & Schaeffer (1999) inactivated RhoA in cultured endothelial monolayers with Clostridium botulinum exoenzyme C3 (C3 toxin). This caused disappearance of actin filament stress fibres and attenuated hyperpermeability stimulated by thrombin and TNFα. These results indicate that RhoA might be part of the permeability-promoting signalling pathways stimulated by both agonists. The permeability-augmenting activity of RhoA in response to thrombin appears to be mainly mediated by Rho-dependent protein kinase (ROCK) that phosphorylates and inactivates myosin light chain (MLC) phosphatase in endothelial cells (Essler et al. 1998; Carbajal et al. 2000; van Nieuw Amerongen et al. 2000a). The resulting increase in MLC phosphorylation was assumed to cause opening of intercellular junctions by a contractile actomyosin-mediated mechanism. Attenuation of thrombin-induced permeability by the ROCK-specific inhibitor Y-27632 is interpreted to result from increased dephosphorylation (and thus inhibition) of MLCs (Carbajal et al. 2000; van Nieuw Amerongen et al. 2000a). These observations are in line with reports showing that endothelial gap formation induced by thrombin and Ca²⁺ ionophores in vivo is an ATP-consuming mechanism that requires binding between actin and myosin and activation of myosin ATPase by MLC phosphorylation (Schnittler et al. 1990; Goeckeler & Wysolmerski, 1995). Thus the results from thrombin-stimulated endothelial cells in culture suggest that RhoA and ROCK affect permeability through modulation of acto-myosin contraction.

However, it is difficult to extrapolate from these studies on macrovascular endothelial monolayers in vitro to physiological mechanisms controlling permeability in vivo, which occurs predominately in the venular side of microvessels (Michel & Curry, 1999). For example, as indicated above, thrombin is broadly used to perturb monolayer integrity in cultured endothelial cells, assuming that thrombin-induced increase in permeability in vitro can be taken as a general model for understanding signalling pathways involved in permeability regulation in blood vessels in situ. However, the role of thrombin as a directly acting permeability-promoting agent in whole organs has not been documented so far. Furthermore, cultured endothelial cell monolayers that respond to thrombin do not appear to respond as strongly and consistently to classical inflammatory agonists such as histamine, bradykinin and certain neurotransmitters to increase permeability. There is also evidence that, unlike thrombin, these agents do not induce significant tensile force in cultured endothelial cells (Moy et al. 1996).

Thus, the aim of the present experiments was to test the effects of toxin B, an inhibitor of all Rho GTPases (Just et al. 1995; Busch & Aktories, 2000) and the ROCK inhibitor, Y-27632, on both basal permeability and agonist-induced increase in permeability of individually perfused venular microvessels in mouse and rat mesenteries. Toxin B was used as inhibitor, because it is a rapidly internalized Rho protein inhibitor applicable for in vivo studies. The experiments of Hippenstiel et al. (1997) showed that endothelial cells respond to toxin B within 1 h (in contrast to C3 toxin, which requires 24 h or longer to achieve sufficient cell loading and RhoA-inactivation). Y-27632 has been shown to be a potent specific inhibitor of ROCK (Ishizaki et al. 2000). To provide parallel in vivo and cell culture studies in one species, we performed experiments on mouse mesentery microvessels in situ and mouse microvascular endothelial cells in vitro. These combined studies are the most detailed to date using parallel protocols to examine inflammatory mechanisms of both cultured endothelial cells and intact microvessels.

METHODS

Mouse myocardial endothelial cell line (MyEnd)

Mice used to generate cultured cells were cared for under a protocol that conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and was approved by the Bioethical Committee of the District of Darmstadt, Germany. An immortalized cell line of endothelial cells from myocardial vessels (MyEnd) was generated by transformation with polyomavirus middle T oncogene as described previously for the generation of microvascular endothelial cells from brain (Aumailley et al. 1991). In brief, myocardial tissue of newborn mice (killed by decapitation) was minced, digested with 0.05 % trypsin (Biochrom, Berlin, Germany), 0.02 % collagenase (Boehringer, Mannheim, Germany), and seeded onto gelatin-coated culture dishes. Cells were grown in Dulbecco's modified Eagle's medium (DMEM; Life Technologies, Eggenstein, Germany) supplemented with 50 u ml⁻¹ penicillin-G, 50 μg streptomycin and 10 % fetal calf serum (Biochrom) in a humidified atmosphere (95 % O₂-5 % CO₂) at 37 °C. One day after cell isolation, adherent cells (mainly fibroblast-like cells and endothelial cells) were transfected with polyomavirus middle T oncogene (PymT) using the retrovirus packaging cell line GP+E-86 (Markowitz et al. 1988) that was kindly provided by B. Engelhardt (Bad Nauheim, Germany). PymT transfection causes a growth advantage of endothelial cells over non-endothelial cells. After 4 weeks the transfected cells had formed a homogeneous monolayer of cells with endothelial morphology. The culture was split once a week and used for experiments between passages 5 and 20. MyEnd cells formed monolayers of highly elongated cells frequently organized into whirl-like formations. The overall cell shape and growth pattern resembled primary cultures of microvascular endothelial cells from brain and skin (Karasek, 1989; Mischeck et al. 1989; Rubin et al. 1991) and differed significantly from the typical cobblestone pattern formed by macrovascular endothelial cells from various sources. MyEnd cells were immunopositive for three endothelial marker proteins tested, i.e. von Willebrand factor, VE-cadherin, and PECAM-1, and displayed continuous junctional staining for the two tight junction-associated proteins, occludin and ZO-1 (not shown).

Porcine arterial endothelial cells (PAECs)

Pulmonary arterial endothelial cells were obtained from freshly slaughtered pigs by exposure of the excised pulmonary artery to 0.1 % collagenase for 12–15 min. Cells were cultured in medium 199 containing 10 % fetal calf serum (GIBCO, Karlsruhe, Germany), passaged and characterized as described by Suttorp et al. (1988). Only cultures of passages 2–5 were used.

Cell culture studies

Cells were grown on coverslips, coated with gelatin cross-linked with glutaraldehyde (Schnittler et al. 1993). Toxin B was added to the culture medium at a concentration of 100 ng ml⁻¹. At different time intervals (30, 60, 90, 120 min), the culture medium was removed and the monolayer fixed for 5 min at 21 °C (room temperature, RT) with 2 % formaldehyde (freshly prepared from paraformaldehyde) in phosphate-buffered saline (PBS; consisting of (mm): 137 NaCl, 2.7 KCl, 8.1 Na₂HPO₄ and 1.5 KH₂PO₄, pH 7.4). Afterwards, monolayers were rinsed at RT with PBS (3 × 5 min) and treated with 0.1 % Triton X-100 in PBS for 5 min. After brief rinsing with PBS, MyEnd cells were incubated for 16 h at 4 °C with rat monoclonal antibody (undiluted hybridoma supernatant) directed to the external domain of mouse VE-cadherin (Gotsch et al. 1997), and PAECs with rabbit polyclonal antibody to human VE-cadherin cytoplasmic domain (obtained from D. Vestweber, Münster, Germany, diluted 1 : 150 with PBS), respectively. After several rinses with PBS (3 × 10 min), monolayers were incubated for 60 min at RT with Cy3-labelled rabbit anti-rat Ig (Dianova, Hamburg, Germany, diluted 1 : 150 with PBS), and Cy3-labelled goat anti-rabbit Ig (Dianova, diluted 1 : 150 with PBS). For visualization of filamentous actin (F-actin), monolayers were fixed and permeabilized as described above and incubated for 60 min at RT with ALEXA-labelled phalloidin (Mobitec, Göttingen, Germany). Monolayers stained with antibodies and phalloidin were rinsed with PBS (15 min). Coverslips were subsequently mounted on glass slides with 60 % glycerol in PBS, containing 1.5 % n-propyl gallate as antifading substance.

Test reagents

Toxin B was prepared as described previously (von Eichel-Streiber et al. 1987). Bradykinin (B-3259) and platelet-activating factor (PAF, P-7568) were bought from Sigma Chemical Co. Y-27632 was a gift from Welfide Corporation (Osaka, Japan).

Glucosylation assay

MyEnd cells were incubated with 100 ng ml⁻¹ toxin B in culture medium for 30, 60 and 90 min. After a brief rinse with PBS, cells were scraped in the presence of lysis buffer (Just et al. 1995). Lysates were incubated with toxin B (1 μg ml⁻¹) in the presence of 10 μm [¹⁴C]UDP-glucose for 60 min. The reaction was stopped by the addition of 1 ml of ice-cold trichloroacetic acid (20 % w/v). Precipitated proteins were subjected to sodium dodecyl sulfate-polyacrylamide (12.5 %) gel electrophoresis (SDS-PAGE), and the gels analysed by phosphorimager system (Molecular Dynamics).

Animal preparation

Mice and rats used for aortic perfusion and hydraulic permeability experiments were cared for under a protocol that conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and was approved by the Institutional Animal Care and Use Committee of the University of California, Davis. Male mice (C57BL6J, Jackson Laboratory, Bar Harbor, ME, USA), 8–10 weeks old, were anaesthetized briefly with isofluorane and then by subcutaneous injection with pentobarbitol (100 mg (kg body weight)⁻¹). Anaesthesia was maintained by giving additional pentobarbital (subcutaneous 0.5 mg per dose) as needed. Rats (male, Sprague-Dawley, 350–450 g, Hilltop Laboratory Animals, Inc.) were anaesthetized with pentobarbital sodium (65 mg (kg body weight)⁻¹) given subcutaneously. Anaesthesia was maintained by giving additional pentobarbital (subcutaneous 3 mg per dose) as needed. At the end of the procedures described, both mice and rats were killed by a pentobarbitol overdose.

Preparation of mouse aorta for perfusion and immuno-labelling

Mice were anaesthetized as above. Each animal was positioned on its back, the thorax opened and the heart held gently near the apex to enable left ventricular puncture with a needle connected to the perfusion device. The device was composed of syringes of appropriate size driven by a constant flow pump to deliver perfusate through polyethylene tubing to a 23 gauge needle. The tubing was passed through a water bath to enable delivery of perfusate at 38 °C and 2 ml min⁻¹ for 15, 30, 60, or 120 min. Mammalian Ringer solution consisted of (mm): 132 NaCl, 4.6 KCl, 2 CaCl₂, 1.2 MgSO₄, 5.5 glucose, 5.0 NaHCO₃, and 20 N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (Hepes) and Na-Hepes, with the ratio of acid-Hepes to Na-Hepes adjusted to achieve pH 7.40–7.45. All perfusates were mammalian Ringer solution additionally containing bovine serum albumin at 10 mg ml⁻¹ (Sigma A4378). Immediately after beginning perfusion, the right atrium was cut to provide an outflow from the vasculature. At the end of the experiments, a valve in the tubing line was switched to deliver from a second constant flow pump ice-cold methanol to the aorta for 3–5 min. The thoracic aorta was immediately removed from the animal and placed in −20 °C methanol for 30 min, changed to −20 °C acetone for 1–2 min, and then stored in PBS. Sections of aorta were incubated with rat monoclonal antibody (2 h, room temperature, undiluted) raised against the external domain of mouse VE-cadherin (Gotsch et al. 1997). After several rinses with PBS the tissue was incubated (1 h, room temperature) with secondary antibody labelled with TRITC (goat-anti-rat IgG, Jackson ImmunoResearch Laboratories, West Grove, PA, USA). In some experiments, the perfusion was switched to deliver formaldehyde (freshly prepared from paraformaldehyde, 4 % in PBS) for 5 min before removal of the aorta and storage in fixative. Sections of aorta were rinsed in PBS, then labelled to reveal F-actin with Oregon Green phalloidin (20 min, RT, 5 units ml⁻¹, Molecular Probes) and VE-cadherin as above.

Measurement of loss of VE-cadherin continuity

To provide an index of the degree to which toxin B induced removal of VE-cadherin from the borders of mouse aortic endothelium, confocal images of aortic endothelial cells were examined for apparent gaps in the peripheral VE-cadherin immunolabel. For this purpose, a gap was defined as a region of cell margin at least 1 μm long in which the fluorescent intensity was no more than twice the background level after adjusting contrast and brightness so that background was no more than 40 and nearby immunolabel was above 200. For a single gap between two cells, each cell was given a value of 0.5 gap per cell; a cell having two gaps in its perimeter was given a value of 1 gap per cell, etc. A minimum of 138 cells was examined for each condition. Gap lengths and cell perimeters were also measured.

Preparation of mice for hydraulic conductivity (L_p) experiments

Each mouse was anaesthetized as above and placed on a heating pad to maintain normal body temperature. A midline surgical incision (1 to 1.5 cm) was made in the abdominal wall, and the mesentery was gently taken out from the abdominal cavity and spread over a pillar for L_p measurements. The upper surface of the mesentery was continuously superfused with mammalian Ringer solution during preparation and experimentation. The temperature of the superfusate was maintained at 35–37 °C. All of the experiments were carried out in venular microvessels, which were classified as segments where there is convergent flow, one or two branches distal to true capillaries. All of the vessels selected for cannulation had brisk blood flow and were free of white cells sticking or rolling along the vessel wall.

Preparation of rats for L_p experiments

The preparation of the rat mesentery for microperfusion and measurement of hydraulic conductivity was essentially the same as described above for the mouse mesentery.

Measurement of L_p of the microvessel wall

All measurements were based on the modified Landis technique, which measures the volume flux of water across the wall of a microvessel perfused via a glass micropipette following occlusion of the vessel. The assumptions and limitations of the measurement have been evaluated in detail (Michel & Curry, 1999). The initial transcapillary water flow per unit area of the capillary wall (J_v/S)₀ was measured at predetermined capillary pressures of 30 to 60 cmH₂O. Microvessel L_p was calculated as the slope of the relation between (J_v/S)₀ and pressure. All perfusates were mammalian Ringer solution additionally containing bovine serum albumin at 10 mg ml⁻¹ (Sigma A4378). Experimental reagents (toxin B, Y-27632, bradykinin and PAF) were added to the perfusate in the respective experiments and delivered via the micropipette continuously during L_p measurement. Changes in perfusate were accomplished by withdrawing the initial pipette and replacing it with a second pipette filled with new perfusate solution of the appropriate composition. Measurements of (J_v/S)₀ were made at approximately 10-min intervals for up to 140 min both in absence and in presence of inhibitors. Measurements of (J_v/S)₀ were made at approximately 1-min intervals during challenge with the inflammatory mediators PAF and bradykinin. Average values are reported as mean ± s.e.m. unless noted otherwise.

RESULTS

Toxin B-induced glucosylation of Rho proteins

Toxin B contains glucosyl transferase activity that inactivates Rho proteins by UDP-glucose-dependent glucosylation at threonine 37. The time course of glucosylation was determined in cultured MyEnd cells and cultured PAECs treated for 30, 60 and 90 min with 100 ng ml⁻¹ toxin B. As shown in Fig. 1, glucosylation occurred more rapidly in MyEnd cells as compared to PAECs. After 90 min of exposure, 90 % of Rho proteins in MyEnd cells were glucosylated as compared to 65 % in PAECs. The data observed with PAECs correspond to previously published data in which about 50 % of Rho proteins were glucosylated by 100 ng ml⁻¹ toxin B after 90 min of exposure (Hippenstiel et al. 1997).

Figure 1. Toxin B-induced glucosylation of Rho proteins in MyEnd cells and PAECs.

Cultured MyEnd cells and PAECs were treated for 30, 60 and 90 min with toxin B. Degree of in vivo glucosylation was determined by [¹⁴C]glucose incorporation in cell lysates incubated with 1 μg ml⁻¹ toxin B and [¹⁴C]UDP-glucose and subjected to SDS-PAGE and phosphorimaging. Value of control cultures was set to 1. Data are means ±s.d. of two separate experiments.

Effects of toxin B on cultured endothelial cells

Both MyEnd cells and PAECs responded in a similar way to toxin B. However, the response of PAECs was more pronounced and occurred earlier during toxin B treatment as compared to MyEnd cells. Untreated monolayers displayed numerous actin filament stress fibres brightly labelled with fluorescent phalloidin (Fig. 2A and C). After 60 min of toxin B exposure, MyEnd cells did not show significant changes in the stress fibre pattern, whereas in PAEC stress fibres were already strongly reduced (not shown). After 90 min, stress fibres had further decreased in PAECs and had diminished in MyEnd cells (Fig. 2B and D). VE-cadherin displayed strong junctional immunostaining in both MyEnd cells (Fig. 3D) and PAECs (Fig. 3A). In wedge-shaped overlaps at junctions, VE-cadherin label was seen to consist of numerous streaks, dots and reticular structures that have been described previously (Drenckhahn & Ness, 1997). PAECs already displayed local disruptions of VE-cadherin label 60 min after toxin B exposure (Fig. 3B) and had undergone dramatic dissociation from junctions (ca 70–90 % of the cell periphery) after 90 min (Fig. 3C). VE-cadherin was mainly confined to narrow cytoplasmic connections that resisted disruption after 90 min of toxin B treatment (Fig. 3C). In MyEnd cells exposed to toxin B for 90 min, the VE-cadherin pattern was still largely continuous, but displayed local discontinuities that were often associated with the formation of intercellular gaps (Fig. 3E). After 120 min of toxin B treatment, MyEnd cells displayed similar changes as described above for PAECs at 90 min of treatment, i.e. local clustering of VE-immunoreactivity at focal sites of cellular adhesion and formation of numerous intercellular gaps, the margins of which were mainly devoid of VE-cadherin immunoreactivity (Fig. 3F). The portion of the cell periphery devoid of VE-cadherin staining increased from ca 5–15 % after 60 min to ca 30–50 % after 90 min toxin B treatment. Almost identical cellular changes occurred when cultures were treated for only 60 min with toxin B and then cultured for a further 90 min with fresh medium not containing toxin B. This indicates that the amount of toxin B taken up during 60 min is sufficient to irreversibly poison the cells and promote the cytoskeletal and junctional changes which occur during the following 90 min.

Figure 2. Toxin B reduces F-actin in PAECs and MyEnd cells.

F-actin visualized by ALEXA-labelled phalloidin in cultured PAECs (A and B) and MyEnd cells (C and D) before (A and C) and after (B and D) 90 min exposure to 0.1 μg ml⁻¹ toxin B. Note in B, toxin B-induced reduction of actin filament stress fibres which is more prominent in PAECs, as compared to MyEnd cells (D). Images shown are representative of eight separate experiments. Scale bar, 20 μm.

Figure 3. Effect of toxin B on distribution of VE-cadherin in cultured PAECs and MyEnd cells.

PAECs (A-C) and MyEnd cells (D-F) were exposed to 100 ng ml⁻¹ toxin B for the time intervals indicated. Note toxin B-induced redistribution and disruption of the junction-associated VE-cadherin immunostain accompanied by formation of intercellular gaps. PAECs respond more rapidly to toxin B with junctional dissociation than MyEnd cells. Images shown are representative of eight separate experiments. Scale bar, 20 μm.

Effect of ROCK inhibition on cultured endothelial cells

We next tested whether specific inhibition of the RhoA-ROCK pathway by the rho kinase inhibitor Y-27632 would modify the actin cytoskeleton or VE-cadherin distribution. Y-27632 pretreatment (10 μm, 60 min) has been shown to rearrange peripheral band actin and to disassemble stress fibres in cultured endothelial cells and to block formation of stress fibres that are otherwise induced by lysophosphatidic acid (LPA) or histamine (Hirase et al. 2001) and by thrombin (Carbajal et al. 2000). Also, Y-27632 (10 μm, 30 min) induced a complete loss of stress fibres from Swiss 3T3 cells and prevented induction of stress fibres by LPA (Ishizaki et al. 2000). In the present experiments treatment of MyEnd cells and PAECs with ROCK inhibitor Y-27632 (30 μm) resulted within 120 min in significant reduction of stress fibres, which was more pronounced in PAECs (not shown) than in MyEnd cells (Fig. 4A and B) or in mouse aortic endothelium in situ (see Fig. 6 below). The junction-associated actin filament system remained intact. Intercellular gaps were not observed in either MyEnd cells or PAECs and the VE-cadherin distribution remained continuous in both cell types (Fig. 4C and D) indicating that inhibition of the RhoA-ROCK pathway is not critically involved in junctional dissociation observed above after toxin B treatment.

Figure 4. Effect of ROCK inhibition on F-actin and VE-cadherin in cultured MyEnd cells.

Distributions of F-actin (A and B) and VE-cadherin (C and D) are shown from cultured MyEnd cells treated for 120 min with the ROCK inhibitor Y-27632 (30 mm). Note significant reduction of stress fibres but well preserved junction-associated actin filament system. The VE-cadherin distribution remains unchanged and no intercellular gaps are seen. Images shown are representative of eight separate experiments. Scale bar, 20 μm.

Figure 6. Effect of ROCK inhibition on F-actin and VE-cadherin on mouse endothelium in situ.

Mouse aortas were perfused for 120 min in the absence (A and C) or presence (B and D) of Y-27632 (30 mm). VE-cadherin distribution is not affected by ROCK inhibition. Stress fibres were reduced by treatment with Y-27632 while peripheral band F-actin remained as in controls.

Toxin B effect on endothelial cells in blood vessels

In mesenteric microvessels we were not able to combine microperfusion of single capillaries with subsequent immunostaining for VE-cadherin of the same vessels. In addition, in mesenteric microvessels background fluorescence created by phalloidin binding to pericytes and to other perivascular cells rich in actin, did not allow selective visualization of the endothelial actin filament cytoskeleton. Therefore, we perfused mouse aorta by ventricular puncture to test the effects of toxin B on mouse endothelium in situ. In the endothelium of the thoracic aorta perfused with control solution for up to 60 min, VE-cadherin displayed a continuous junction-associated distribution (Fig. 5A). In places, the typical streak- and dot-like distribution of VE-cadherin could be seen as described above for cultured cells. F-actin was seen in association with both junctions and stress fibres (Fig. 5C). Stress fibres were less numerous as compared to cultured endothelial cells. Perfusion of the aorta with 100 ng ml⁻¹ toxin B caused gradual loss of stress fibres and discontinuities (interruptions) of the junctional VE-cadherin label. These changes were observed at 30 min and were increased after 60 min of treatment (Fig. 5B and D). The patterns reported were observed in both dorsal and ventral sections of mid-thoracic aortas from two mice in each experimental condition. Qualitatively, they resembled the changes described above for cultured MyEnd cells.

Figure 5. Effect of toxin B on mouse endothelium in situ.

Distributions of VE-cadherin (A and B) and F-actin (C and D) are shown in mouse aortic endothelium after 60 min perfusion with control solution (A-C) or 60 min perfusion with toxin B (100 ng ml⁻¹) (B-D). After exposure to toxin B VE-cadherin reveals numerous discontinuities in the perimeter label (single arrows in B) and frequent broad regions composed of numerous short spurs, some disconnected and some in intricate networks (double arrows in B). In control conditions the F-actin pattern, labelled with Oregon Green phalloidin, shows peripheral band (pb) and some stress fibres (sf) oriented parallel to the long axis of the cells (C). The intensity of F-actin labelling was diminished in both the peripheral band and the stress fibres in vessels perfused with toxin B (D). Images shown are representative of two in vivo experiments for each condition. Scale bar, 20 μm.

In contrast to the extreme rearrangement of VE-cadherin seen in cultured cells, the changes observed in the in situ endothelium, although obvious to visual inspection, were much less dramatic. At 30 min of perfusion the number of interruptions in VE-cadherin label, counted as gaps per cell, increased significantly from 0.18 ± 0.02 in the absence of toxin B to 0.47 ± 0.04 in the presence of the toxin (P < 0.05). At 60 min of perfusion control aortas had 0.25 ± 0.03 gaps per cell, while in vessels perfused with toxin B the number increased significantly to 1.22 ± 0.07 (P < 0.05). When multiplied by respective mean gap lengths these numbers were converted to apparent proportions of cell periphery not labelled by antibody to VE-cadherin. For the 30-min group this proportion increased from 0.5 % in the control to 4.2 % in the presence of toxin B. In the 60-min group the proportion changed from 0.9 % in the absence to 8.5 % in the presence of toxin B. These measurements do not represent absolute changes in the expression or assembly of VE-cadherin at the cell junctions, but provide a rough index of the toxin B-dependent change in VE-cadherin distribution. Therefore, toxin B induced a statistically significant rearrangement of VE-cadherin within endothelium in situ.

Effects of ROCK inhibition on aortic endothelium in situ

To determine the effects of specific ROCK inhibition on endothelium in situ, we perfused mouse aorta for 60 and 120 min in both the presence and the absence of Y-27632 (30 μm). Aortas from two mice were examined in each protocol. While no change in peripheral band actin was noted, stress fibres were reduced in Y-27632-treated endothelium relative to non-treated. In contrast to the toxin B exposure, VE-cadherin was not affected by exposure to Y-27632. The labelling pattern appeared as intense and continuous in treated aortas as in controls (Fig. 6A and B). Thus, while toxin B treatment is associated with redistribution of VE-cadherin, specific inhibition of ROCK for up to 2 h has no effect on the distribution of VE-cadherin in endothelium in situ.

Toxin B effect on mouse microvessel L_p

Biochemical and immunolabelling studies with cultured endothelial cells and with mouse aortic endothelium in situ showed that the toxin B concentration used (100 ng ml⁻¹) had pronounced effects on F-actin and VE-cadherin both in situ and in vivo. The same concentration of toxin B was used in permeability studies in mouse and rat microvessels. This concentration has previously been shown to increase the hydraulic conductivity of monolayers of cultured PAECs between 45 and 60 min after initial exposure to the toxin (Hippenstiel et al. 1997). In the present study, exposure of mouse microvessels to 100 ng ml⁻¹ toxin B increased the hydraulic conductivity of the microvessels after a delay of about 60 min. At that time, cultured MyEnd cells displayed about 50 % glucosylation (inactivation) of Rho proteins. A representative response from one of four similar experiments is shown (Fig. 7A). The mouse microvessels had a mean L_p of (4.7 ± 1.0) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹ during the first 10 to 60 min of exposure. Values of L_p attained near 90 min (time to glucosylate 65–90 % of Rho proteins) ranged from 13.9 × 10⁻⁷ to 29.0 × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹ and had a mean value of (19.8 ± 3.3) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹ at 82 ± 5 min of treatment. Thus, L_p of all vessels increased well above baseline in response to toxin B (P < 0.05, paired t test). The increased L_p was associated with a tendency of marker red cells, used to measure L_p in the Landis technique, to track towards localized sites on the wall, indicating that the increase in permeability along the wall may have been spatially non-uniform. This was reflected in variability of the calculated L_p from one measurement to the next after permeability began to increase.

Figure 7. Toxin B increases L_p of venules in mouse mesentery.

A, data from a representative experiment demonstrate that toxin B (100 ng ml⁻¹) caused a large increase in L_p after about 1 h of perfusion. B, data from a control vessel, perfused with vehicle solution containing serum albumin (10 mg ml⁻¹), show that mouse mesentery vessels are stable for over 2 h of perfusion.

In control experiments (two vessels) we perfused mouse venular microvessels with Ringer solution containing bovine serum albumin as in the test vessels. There was no increase in L_p in these control vessels perfused for at least 120 min (Fig. 7B). We did not test lower concentrations of toxin B because the results of Hippenstiel et al. (1997) indicated that the increase in L_p occurred after 120 min (at 50 ng ml⁻¹) and after 150 min (at 25 ng ml⁻¹). These times are longer than those where control baseline has been demonstrated to remain constant.

Investigations in rat microvessels

Mouse mesenteries have few microvessels suitable for L_p measurement by the present methods. Because rat mesenteries have many more microvessels than mouse mesenteries, we extended our investigations to include rats. The results in rat venular microvessels were similar to those in the mouse mesentery. Data from a representative experiment illustrate that L_p was unaffected for at least an hour in each case and then responded by increasing severalfold during the second hour of perfusion (Fig. 8A). In four venular vessels perfused with toxin B, the mean baseline L_p was (1.5 ± 0.1) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹ (100 ng ml⁻¹). The average L_p in these vessels measured near 90 min (86 ± 6 min) of toxin B perfusion increased to (8.6 ± 1.7) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹. L_p of these vessels increased above baseline in response to the toxin B treatment (P < 0.05, paired t test). In four control microvessels perfused for 120 min with mammalian Ringer solution containing bovine serum albumin there was no increase in L_p (Fig. 8B).

Figure 8. Effect of toxin B on rat mesentery venules.

A, data from a representative experiment show that L_p is stable for about 1 h before increasing by severalfold during exposure to toxin B. B, data from a control experiment demonstrate that the L_p of rat mesentery venules is stable during perfusion with vehicle solution containing serum albumin (10 mg ml⁻¹) over a comparable 2 h period.

Toxin B fails to inhibit bradykinin response of rat venular microvessels

We tested the effect of pretreatment of microvessels with toxin B on the acute inflammatory response to bradykinin. In four vessels the mean baseline L_p (L_p,control) was (2.0 ± 0.2) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹. The vessels were treated with toxin B (100 ng ml⁻¹) for up to 40 min followed by 1 nm bradykinin. The mean peak response to bradykinin after pretreatment with toxin B relative to paired L_p during the initial control period (L_p,peak/L_p,control) was 3.5 ± 0.6. This response to bradykinin was significantly different from 1, the value expected if toxin B had blocked the responsiveness to bradykinin (P < 0.05, t test). Moreover, the response to bradykinin in the presence of toxin B was not different (P > 0.05, unpaired t test) from the mean value of the peak L_p after exposure to bradykinin alone in a separate group (4.8 ± 0.4, n = 11) for which the baseline L_p was (1.8 ± 0.2) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹. Representative responses in a vessel pretreated with toxin B and a control are shown in Fig. 9A and B. As shown above, 40 min is sufficient to detect changes in the organization of VE-cadherin in the junctions. In one additional vessel we waited until toxin B (100 ng ml⁻¹) had caused a significant increase in permeability after 120 min, and retested with bradykinin. Again, the bradykinin response was not blocked (Fig. 9C).

Figure 9. Toxin B does not block the acute inflammatory response to bradykinin in rat venules.

A, L_p of a representative vessel is plotted during 35 min perfusion with toxin B (100 ng ml⁻¹). When stimulated with bradykinin (1 nm) the vessel responds with a characteristic transient threefold increase in L_p. B, a representative response to bradykinin in the absence of toxin B is shown from a second vessel. C, a third vessel was perfused for nearly 2 h with toxin B (100 ng ml⁻¹) at which time it showed an increased L_p, characteristic of the toxin effect. When exposed to bradykinin (1 nm) this vessel also responded with a transient increase in L_p.

We further examined these data to determine whether toxin B treatment altered the time course of the inflammatory response. If the action of ROCK, stimulated by RhoA, is needed to mediate the acute inflammatory response, then inhibition of RhoA by toxin B might be expected to slow such a response. We determined the time between initiating perfusion of bradykinin and the peak value of the L_p as an index of the time course. The time to L_p,peak in the bradykinin-only group was 8.3 ± 0.5 min while the time to L_p,peak in the vessels pretreated with toxin B was 5.0 ± 0.3 min (P < 0.05, unpaired t test). Contrary to the hypothesis, these data indicate that RhoA inhibition quickens the response of in situ venules to bradykinin rather than slowing it.

ROCK inhibition reduces basal hydraulic permeability of rat vessels

As a further test of the dependence of the in vivo permeability barrier on the RhoA-ROCK pathway, we used the ROCK inhibitor Y-27632 (Ishizaki et al. 2000). In a previous study Y-27632 (10 μm, 60 min) was found to completely inhibit increased endothelial monolayer permeability to horseradish peroxidase and to fully block the associated increase in myosin light chain phosphorylation that was otherwise induced by lysophosphatidic acid (Hirase et al. 2001). Similarly, Y-27632 (10 μm, 60 min) partially inhibited thrombin-induced monolayer permeability and blocked phosphorylation of MLC (Carbajal et al. 2000; van Nieuw Amerongen et al. 2000a,b). Endothelial cell pretreatment with Y-27632 (5 and 10 μm, respectively, 30 min) also prevented MLC phosphorylation induced by TNF-α (Petrache et al. 2001) and by mildly oxidized LDL (Essler et al. 1999). In the present studies, after an initial control period, rat mesentery vessels were perfused with a solution containing Y-27632 (30 μm) and the L_p was monitored for up to 100 min. The mean value of L_p expressed as a ratio to that of the control period (L_p,Y-27632/L_p,control) decreased to 0.68 ± 0.06 (n = 5) at 60 min and to 0.62 ± 0.05 (n = 4) at 90 min of perfusion (Fig. 10). The decrease was highly significantly different from a slight fall in L_p found in a separate vehicle control group (P < 0.0001, two-way ANOVA). This effect of ROCK inhibition to decrease L_p is in strong contrast to our results above in which toxin B inactivation of Rho family proteins caused an increase in permeability.

Figure 10. Inhibition of ROCK reduces baseline L_p.

L_p values, expressed as ratios to initial control period L_p (L_p,Y-27632/L_p,control), are shown versus time of exposure to Y-27632 (30 μm). The L_p of the treated group was significantly different from that of a second group perfused with vehicle control solution (P < 0.0001, two-way ANOVA).

Effect of ROCK inhibition on response to inflammatory mediators

Finally, we tested whether acute permeability responses due to the mediators bradykinin or PAF could be prevented by pre-treatment with Y-27632. Basal L_p in the first group of eight venules was (1.7 ± 0.2) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹. The vessels were then treated by perfusion with a solution containing 30 μm Y-27632 for 30 min. When the perfusion solution was changed to deliver bradykinin (1 nm) in addition to Y-27632, we observed a characteristic increase in L_p that reached a peak in about 10 min then slowly decreased in a manner similar to those vessels challenged with bradykinin alone. A representative experiment is shown in Fig. 11A (see Fig. 9B for bradykinin control). The mean value of the peak L_p response relative to control (L_p,peak/L_p,control) for eight venules was 4.1 ± 0.8. This mean peak response is significantly different from 1, the value expected if ROCK inhibition by Y-27632 had fully blocked the inflammatory response to bradykinin (P < 0.01, t test). In addition, the mean peak response with Y-27632 present is not significantly different from the mean peak response seen in vessels treated with bradykinin alone (L_p,peak/L_p,control = 4.8 ± 0.4, n = 11; P > 0.05, unpaired t test). Basal L_p in that control group was (1.8 ± 0.2) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹.

Figure 11. ROCK inhibition does not affect venule response to bradykinin or PAF.

Data from representative experiments are shown as L_pversus time. A, after approximately 25 min pretreatment with Y-27632 (30 μm) a vessel responded characteristically to 1 nm bradykinin. A representative response to bradykinin in the absence of Y-27632 is shown in Fig. 9B. B, a vessel perfused for 100 min with Y-27632 (30 μm) responded typically to challenge with 1 nm PAF. C, a representative response to PAF in the absence of Y-27632 from a separate vessel.

To further test whether venular responsiveness to inflammatory mediators depends on ROCK, we examined the L_p responses resulting from exposure to platelet activating factor (PAF) in vessels pretreated with Y-27632. Vessels were perfused for approximately 90 min with a solution containing Y-27632 (30 μm) before changing to solution additionally containing PAF (1 nm). In response to PAF the L_p of each vessel increased to a peak and then began to decrease (representative experiment in Fig. 11B). The mean basal L_p in five venules was (1.3 ± 0.2) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹). Mean L_p,peak/L_p,control was 9.2 ± 2.5. Similar to the results above with bradykinin, the response to PAF in the presence of Y-27632 was significantly higher than 1, the value expected if ROCK inhibition blocked the acute inflammatory response (P < 0.05, t test). Data from a separate experiment to illustrate the response to PAF in the absence of Y-27632 is also shown (Fig. 11C). The mean basal L_p for this group of nine vessels was (1.1 ± 0.1) × 10⁻⁷ cm s⁻¹ cmH₂O⁻¹). For PAF alone the mean L_p,peak/L_p,control was 14.2 ± 2.8 (significantly different from 1, P < 0.01). While the mean L_p,peak/L_p,control in the Y-27632-treated group was somewhat lower (9.2 vs. 14.2), the difference did not reach statistical significance (P > 0.05, unpaired t test). Under the conditions of these experiments, ROCK inhibition with Y-27632 did not block venule responsiveness to either bradykinin or PAF. Also, note that L_p returned toward basal values over the subsequent 20–40 min of perfusion in both PAF and bradykinin treatments in the continued presence of Y-27632.

We also examined these data to see if ROCK inhibition with Y-27632 altered the time course of the inflammatory response. We determined the time between initiating perfusion of inflammatory mediator and the peak value of the L_p as an index. The time to L_p,peak in the bradykinin-only group was 8.3 ± 0.5 min and the time to L_p,peak in the bradykinin with Y-72632 group was 7.4 ± 0.7 min (P > 0.05, unpaired t test). The time to L_p,peak in the PAF-only group was 9.4 ± 1.6 min and for the PAF with Y-27632 group L_p,peak was 11.6 ± 1.7 min (P > 0.05, unpaired t test). These data further emphasize that ROCK inhibition does not alter the response of in situ venules to bradykinin or PAF.

DISCUSSION

A major limitation to the detailed interpretation of experimental results from cultured endothelial cells in terms of the mechanism regulating increased permeability of intact microvessels has been the significant differences between the results from cultured endothelial cells stimulated to increase permeability by inflammatory agents compared to the response of intact microvessels (Lum & Malik, 1994; Michel & Curry, 1999). As far as we know, the present results are the most detailed to date from parallel investigations by the same investigators from experiments in cultured endothelial cells and in intact vessels using common modulation of signalling pathways in inflammation and common analysis of the data. Specifically, using an inhibitor of the small GTPases and an inhibitor of ROCK we have begun to evaluate similarities and differences in the contribution of small GTPases and the RhoA-ROCK pathway to modulate microvessel permeability in cultured endothelial cell monolayers and in intact microvessels.

In the first series of experiments we have extended investigations of changes in cellular cytostructure using toxin B, an agent which inhibits Rho-family proteins by glucosylation and which increases the permeability of PAECs in culture. We found that the magnitude and time course of the increase in permeability of rat and mouse venular microvessels exposed to toxin B is similar to that described in PAECs in culture (Hippenstiel et al. 1997). We also demonstrated that exposure to toxin B caused striking reduction of F-actin staining in PAECs and MyEnd cells in culture, as well as in intact mouse aortic endothelial cells. Furthermore, the toxin B-induced interruptions of VE-cadherin labelling along the endothelial cell junctions were similar in endothelial cells in culture and in intact mouse aorta. In addition, we were able to use the cultured endothelial cells to demonstrate that between 60 and 90 min exposure to toxin B, 50–90 % of the Rho proteins were glucosylated. To distinguish effects of a general inhibition of GTPases, and the action of the Rho A, we also inhibited ROCK, a downstream target of Rho A. With the Rho-kinase inhibitor (Y-27632) PAECs and MyEnd cells responded with significant loss of stress fibres at 30 μm and changes in cell shape. Although there was a pronounced effect on the actin filament cytoskeleton, junctions appear to be closed as indicated by continuous VE-cadherin staining and the absence of any visible larger gaps. These observations at the cytostructural level suggest that a RhoA-ROCK pathway does not modify junction structure and are consistent with the result that specific inhibition of RhoA by C3-toxin in cultured endothelial cells does not perturb continuous distribution of VE-cadherin (Wójciak-Stothard et al. 1998; Vouret-Craviari et al. 1999).

Mechanism of action of Rho proteins to increase endothelial barrier permeability

Treatment with toxin B elicited gap formation in endothelial monolayers and induced increased permeability of intact mesenteric microvessels. The increase in permeability caused by toxin B after 60–90 min occurred under conditions where our biochemical analysis indicated that about 90 % of all Rho proteins in MyEnd cells were inhibited by glucosylation. Because inhibition of RhoA by C3-toxin in cultured endothelial cells does not perturb continuous distribution of VE-cadherin (Wójciak-Stothard et al. 1998; Vouret-Craviari et al. 1999), we suspected that toxin B-induced barrier dysfunction may be the result of inactivation of other Rho proteins, such as Cdc42 and Rac. Our result that the ROCK inhibitor in unstimulated rat microvessels actually reduced permeability, rather than increased permeability, provides quantitative functional data to support the cytostructural observations that GTPases other than RhoA modify the structure in the junctions that regulate permeability.

The most striking result from our experiments is that intact microvessels treated with either toxin B to inactivate up to 90 % of Rho-family proteins, or the specific ROCK inhibitor Y-27632, still responded normally to the inflammatory agents, bradykinin and PAF, with a transient increase in permeability. Moreover, after the peak response the permeability returned toward control levels indicating that the barrier could reform despite continued ROCK inhibition. It appears reasonable to conclude from these observations that the mechanisms whereby bradykinin and PAF rapidly increase permeability may not involve any Rho protein signalling pathways. This result is consistent with the study of Vouret-Craviari et al. (1999), who did not see any involvement of RhoA in thrombin-induced monolayer disintegration, but is in contrast to observations of the effect of inhibitors of RhoA and ROCK that have been shown to significantly attenuate the increase in permeability of cultured endothelial cell monolayers exposed to thrombin (Essler et al. 1998; Carbajal & Schaeffer, 1999; Carbajal et al. 2000; van Nieuw Amerongen et al. 2000a) and the action of Y-27632 to entirely block lysophosphatidic acid-induced hyperpermeability of human umbilical vein endothelial cells (HUVECs; van Nieuw Amerongen et al. 2000b). These results suggest that at least in some thrombin-stimulated cultured endothelial cell monolayers, the RhoA-ROCK pathway plays a more significant role to regulate barrier integrity than in intact microvessels.

One of the models describing the mechanism that modulates endothelial barrier function in vivo suggests that the permeability response (increase or decrease) depends on a balance between centripetal tension (leading to cell retraction) and cellular adhesion (tethering), which resists this tension (Garcia & Schaphorst, 1995; Drenckhahn & Ness, 1997; Michel & Curry, 1999). Thus, an increase in permeability might result from increased centripetal tension and/or decreased cellular adhesion, and a decrease in permeability can result from decreased tension and/or increased adhesion. Adhesion structures may be diminished and the actin/myosin contractile apparatus may be hypertrophied in cultured endothelial cells so that activation of myosin by inflammatory and other pathways will lead to formation of relatively large intercellular gaps associated with hyperpermeability of the cultured monolayers. Consequently, inhibition of contractile mechanisms by inactivation of the Rho-ROCK pathway was found to largely prevent increases in monolayer permeability of cultured cells. By contrast, in intact microvessels that are exposed to distending forces of blood pressure, tethering mechanisms may be particularly critical for barrier maintenance. The observed decrease in basal permeability of microvessels treated with the ROCK inhibitor might correspond to a decrease in centripetal tension of the actin cytoskeleton. Acute permeability increases in intact vessels may primarily reflect mechanisms that decrease cell-to-cell adhesion. This would explain how bradykinin and platelet activating factor elicited normal permeability responses in microvessels in situ despite inhibition of Rho proteins by prior treatment with toxin B, indicating that Rho-ROCK-dependent contractile mechanisms might not be involved in regulation of microvascular hyperpermeability induced by bradykinin or other physiological stimuli.

Further studies to compare cultured endothelial cells and intact microvessels

The most important conclusion from the present investigations is that the results in intact microvessels do not support the hypothesis that signalling pathways involving RhoA contribute generally to mechanisms leading to increased permeability. The dominant role for RhoA, particularly in thrombin-induced permeability in cultured endothelial cell monolayers, does not appear to be part of a general mechanism to increase permeability. Surprisingly, at this time we do not know if there are any vessels in mammalian microvascular beds that respond to thrombin with the vigour and consistency described for endothelial cells in culture. At least, in venular microvessels of mouse and rat mesenteries we could not observe any acute or subacute responsiveness to thrombin (authors' unpublished observations). One possibility is that endothelial cells in vivo which have undergone some types of dedifferentiation (corresponding to the process of isolation and culturing on artificial substrates) may have altered responses to inflammatory agents such as thrombin. This possibility remains to be investigated.