Endothelial permeability for macromolecules. Mechanistic aspects of pathophysiological modulation.

Abstract

HomeArteriosclerosis, Thrombosis, and Vascular BiologyVol. 17, No. 6Endothelial Permeability for Macromolecules Free AccessResearch ArticleDownload EPUBAboutView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toFree AccessResearch ArticleDownload EPUBEndothelial Permeability for MacromoleculesMechanistic Aspects of Pathophysiological Modulation Victor W.M. van Hinsbergh Victor W.M. van HinsberghVictor W.M. van Hinsbergh From the Gaubius Laboratory TNO-PG, Leiden, and the Institute for Cardiovascular Research, Free University, Amsterdam, the Netherlands. Search for more papers by this author Originally published1 Jun 1997https://doi.org/10.1161/01.ATV.17.6.1018Arteriosclerosis, Thrombosis, and Vascular Biology. 1997;17:1018–1023The endothelium separates blood from tissue and actively regulates the exchange between these compartments. Throughout the body, this one-cell-thick lining of blood vessels extends over several thousand square meters. It harbors many receptors and functions that actively regulate the extravasation of nutrients, solutes, hormones, macromolecules, and leukocytes. If the barrier function of the endothelium does not perform appropriately, leakage of macromolecules occurs, causing edema formation and exposure of the interstitium to high concentrations of plasma constituents. In severe cases of vascular leakage, platelets will adhere to the exposed subendothelium and red blood cells may even accumulate in the interstitium. Vascular leakage may be desirable for recruiting plasma proteins, such as complement factors during infection, but it can become life-threatening, particularly when it occurs in the lungs or pericardium or when it causes hypovolemic shock. Furthermore, frequent or prolonged leakage may impair proper functioning of the affected tissue. In recent years, considerable progress has been made in understanding the molecular processes that contribute to the regulation of endothelial permeability.Nature of the Endothelial BarrierWith the exception of the liver, adrenal, and bone marrow sinusoids, in which the endothelium has rather large pores, the endothelium constitutes a selective barrier between blood and tissue. In principle, this barrier consists of the EC monolayer and its matrix. Macromolecules can cross the endothelial barrier in three ways: (1) between the cells, through cell junctions (paracellular); (2) through the EC, via pores (diaphragms or fused vesicles); and (3) transcellularly, via shuttling vesicles and specific receptors. It is generally believed that the charge and compactness of the endothelial matrix (both glycocalyx and basal membrane) contribute additionally to the selectivity of the endothelial barrier toward molecules of different size and charge.Electron microscopic evidence strongly suggests that in tissue capillaries, hormones and macromolecules are shuttled across the endothelial barrier via vesicles, in which these biomolecules may accumulate owing to receptor binding.1 By such mechanisms can an adequate supply of hormones and albumin be provided to the different tissues. However, according to Starling's law, the extravasation of fluid and macromolecules is directly related to hydrostatic and osmotic pressure. Only when vesicles fuse into cell-spanning pores2 is it understandable how an increase in hydrostatic pressure gradient increases vesicle-mediated exchange. However, it is plausible that the flux of macromolecules from the blood to the interstitium that contributes to the lymphatic fluid also involves other structures, probably located in another segment of the vascular tree. The postcapillary venules are good candidates for this because their ECs have rather simple intercellular junctions, and minute gaps within them that cause vascular leakage are formed on exposure of a postcapillary venule to vasoactive agents.In the early 1960s, Majno and Palade3 showed that after exposure of tissue to histamine, carbon particles injected into the blood compartment left it selectively via postcapillary venule ECs, between which gaps were occasionally seen. In a subsequent study, Majno et al4 observed that the nucleus of these ECs had a wrinkled appearance and postulated that EC contraction was the basis of the increased extravasation of macromolecules. This concept has now been generally accepted, and subsequent biochemical and cell studies have indicated that both actin-myosin interaction56789 and reversible disintegration of intercellular tight and adherens junctions1011 are involved. The formation of small intercellular gaps is shared by many other vasoactive agents and is particularly important in edema formation during acute inflammation and the capillary leakage syndrome.In addition, other forms of vascular leakage may occur. After prolonged inflammation, capillaries may become leaky, a phenomenon that is probably related to an increase in angiogenesis.12 The angiogenic vascular endothelial growth factor causes serious edema when applied to the skin of animals and can induce clusters of connected vesicular structures, the so-called vesicular vacuolar organelles, and pores within EC (ie, VEGF). It makes the endothelial monolayer leaky1314 and alters protein phosphorylation in adherens junctions.15 Impairment of endothelial barrier function can also occur in arterioles after prolonged exposure to histamine.16 It is not yet known whether the loss of tight junction integrity recently demonstrated in cultured ECs after prolonged exposure to histamine11 contributes to this phenomenon. Furthermore, the endothelial permeability of arteries and veins can occasionally increase. Massive endothelial injury and desquamation occur only in exceptional cases, eg, after introduction of a bacterial toxin, such as in pig edema disease. More frequently, focal leaky spots are encountered in the endothelium of arteries and veins. They are found after exposure of the vessel to injurious conditions, such as hypercholesterolemia17 and atherosclerosis.18 These focal leaky spots are frequently associated with EC division19 or occasionally with leaky junctions.20Actin-Myosin Interaction and Endothelial Gap FormationStudies on permeabilized ECs in vitro have shown that they have a contractile system that requires actin, nonmuscle myosin II, ATP, calcium ions, and calmodulin.56 This permeabilized cell system permitted Schnittler et al6 to inhibit endothelial contraction by myosin head fragments, thus providing firm evidence of a role for actin–nonmuscle myosin interaction in endothelial contraction. It probably does not extend the length of the whole cell but occurs in small areas at the cell perimeter. This contraction causes small gaps at cell junctions without complete cell separation. These small gaps, which are induced immediately after exposure to vasoactive agents, have recently been visualized in vivo by silver staining techniques.21The mechanism of contraction in ECs is comparable to that in SMCs, although many of the proteins involved in the contraction process and its regulation are different in the two cell types (see Reference 2222 for a review). As in SMC contraction, the interaction between actin and nonmuscle myosin in ECs is regulated by the phosphorylation status of MLC78923 (Fig 1). Two MLCs, one structural and one regulatory, are bound to the head part of myosin heavy chain. By the action of MLCKs, the regulatory MLC is monophosphorylated or diphosphorylated. The phosphorylated MLC activates the myosin molecule, after which it interacts with F-actin filaments and can then cause contraction by sliding along the actin filament.When ECs in vitro are stimulated by histamine or thrombin, which induces a transient or prolonged increase in endothelial permeability, respectively, monophosphorylated and diphosphorylated MLCs are generated. Within 1 to 2 minutes phosphorylation reaches a maximum, which follows an increase in the cytoplasmic Ca2+ concentration. The induced increase in permeability can be partially inhibited by intracellular Ca2+ chelators or inhibitors of calmodulin and MLCK.6792324 Therefore, it is generally believed that Ca2+/calmodulin–dependent PK I (the classic MLCK) plays a central role in the onset of endothelial gap formation after cell exposure to vasoactive agents. In close agreement with these data, it has been observed in vivo that a transient increase in cytoplasmic Ca2+ concentration parallels an increase in endothelial permeability after exposure of frog mesenteric microvessels to histamine.25 In this case the endothelial barrier recovers within several minutes.MLCKs and PhosphatasesThe nature of endothelial MLCK has not been fully resolved. While searching for MLCK isoforms, Gallagher et al26 found a 208-kD protein, called embryonic MLCK, in embryonic tissues. This protein was also found to be abundant in human umbilical vein ECs, in addition to small amounts of the classic 130- to 155-kD form of MLCK. Embryonic MLCK differs markedly from classic MLCK in SMCs in terms of its primary (ie, amino acid) sequence.27 In addition, Garcia and colleagues have cloned an endothelial MLCK, which also has a molecular mass of ≈210 kD but represents an alternatively spliced product of the same gene as that of smooth muscle MLCK (Genbank access No. U 48959, 1996). In other types of mesenchymal cells, it was recently demonstrated that Rho kinase can also act as an MLCK28 and thus can contribute to stress fiber contraction29 and tethering of actin fibers to the cell membrane.30 It is likely that Rho kinase is also present in ECs. This brings the number of MLCK (iso)forms to four. It has been demonstrated that the classic MLCK and its endothelial splice variant are activated by Ca2+/calmodulin, whereas activation of Rho kinase depends on RhoA and tyrosine phosphorylation. No information is yet available about the presence of embryonic MLCK in other types of ECs than those from the umbilical cord or about its regulation. Also unknown is the distribution of the various potential MLCKs in ECs. Future studies have yet to reveal where these MLCKs are located and which of them are involved in the organization of the actin cytoskeleton, formation of filopodia and membrane ruffles, stress fiber tension, and induction of gaps in EC junctions.The phosphorylation state of MLCs depends not only on its phosphorylation by MLCKs but also on its dephosphorylation by MLC phosphatase. Indeed, ample evidence exists that in SMCs, inactivation of MLC phosphatase also can enhance the phosphorylation state of MLC and the contractile state of the cell.31 The involvement of a type 1 myosin–associated protein phosphatase in the regulation of endothelial barrier function was recently demonstrated in bovine pulmonary artery ECs.32 In this context, it is also of interest that RhoA not only activates MLCK but also binds to the regulatory subunit of MLC phosphatase and that this binding reduces dephosphorylation of MLC.33Experiments with calcium ionophore7 have shown that elevation of cytoplasmic Ca2+ concentration per se is insufficient to induce endothelial contraction. However, these data should be interpreted with caution, because it is known from studies in SMCs that the high concentration of Ca2+ generated by the ionophore activates Ca2+/calmodulin–dependent PK II, which interferes with MLC phosphorylation (see Reference 3131 ).Additional Mechanisms Contributing to Prolonged PermeabilityA number of in vitro experiments suggests that regulation of endothelial barrier function is more complicated. In most of these experiments, thrombin has been used as a stimulus to increase permeability. In contrast to the transient increase in permeability induced by histamine, thrombin induces a prolonged increase in permeability in vitro.9 The maximal increase in permeability induced by thrombin extends far beyond the increase in cytoplasmic Ca2+ concentration.7 On the basis of this observation, it has been suggested that in addition to Ca2+/calmodulin–dependent phosphorylation of MLC, other factors contribute to the increase in permeability or at least its prolongation.Recent measurements of isometric tension in ECs demonstrated that the thrombin-increased isometric tension was accompanied by MLC phosphorylation to a considerable extent, whereas histamine had a much smaller effect89 The effects of thrombin were inhibited by cytochalasin D and an MLCK inhibitor, suggesting that actin and MLC phosphorylation play a role in thrombin-induced isometric tension. These data strongly suggest that an increase in actin-myosin–dependent isometric tension contributes to the prolonged increase in endothelial permeability. However, two aspects remain to be clarified.89 First, it is possible that thrombin acts on another MLCK in addition to the Ca2+/calmodulin–dependent MLCK that is also expected to be activated by histamine. This may explain the striking difference between the effects of thrombin and histamine. Alternatively, histamine may activate MLCK in only selected areas of the cell. Second, the involvement of isometric contraction in thrombin-induced permeability does not exclude the simultaneous involvement of other mechanisms. In particular, signal transduction via PKC and protein tyrosine phosphorylation has been implicated in the regulation of endothelial permeability.Several authors have reported that activation of PKC contributes to the thrombin-induced increase in permeability of bovine ECs.3435 In human EC monolayers, platelet-activating factor–induced permeability was increased by PKC activation,36 but direct activation of PKC by phorbol ester reduced endothelial permeability.37 Overexpression of PKC β-I isotype in ECs by transfection of cDNA increased the permeability in these cells, further supporting the role of PKC in the regulation of endothelial permeability.38 However, in contrast to direct MLC phosphorylation by PKC in SMCs, activation of PKC by phorbol ester did not directly cause phosphorylation of MLC in ECs.7 Equally well, the pattern of phosphorylation of individual amino acids in MLC by thrombin indicated phosphorylation by MLCK but not by PKC.8 This finding suggests that PKC either indirectly facilitates the increase in permeability induced by vasoactive agents or acts on other factors that contribute to the maintenance of endothelial barrier function, such as the integrity of cell-cell junctions and cell-matrix contacts. In line with this latter suggestion, Rabiet et al10 reported that thrombin disrupted the VE-cadherin-catenin complex in adherens junctions and that this effect could be prevented by inhibition of PKC or the tyrosine PK inhibitor herbimycin A. This suggests that in addition to the contraction mechanism that involves actin–nonmuscle myosin interaction, disintegration of adherens junctions also contributes to the increased permeability observed in vitro. This disintegration is reversible after several hours.10 The time course of its induction and reversal is on par with that of thrombin-induced increase in permeability. These data suggest that disintegration of adherens junctions between cells and also the possible loss of focal contacts between the cell and its matrix may contribute largely to the prolonged leakage induced by thrombin in vitro.Further studies are needed to verify whether these mechanisms also act in vivo or are overactive in vitro. Published studies on the microcirculation have demonstrated a short-lasting transient leakage in postcapillary venules after stimulation with a vasoactive agent. These observations seem to give little support to the concept of prolonged leakage sites. However, these observations were made in healthy tissues and with a single stimulus. It is indeed conceivable that disintegration of cell-cell contacts and prolonged leakage occurs, particularly in areas of inflammation where leukocytes adhere to ECs and act on them by producing platelet-activating and other factors.39 If this is true, then it is important to elucidate its mechanism and regulation, because a different approach may be required to treat such cases of vascular leakage.Stabilization of the Endothelial Barrier In VivoWhen an animal is injected intravenously with histamine, its effect on endothelial permeability is much less than when it is administrated to the tissue itself, even though the plasma histamine concentration exceeds that achieved by local application. In a series of experiments, Grega and colleagues demonstrated that intravenous infusion of histamine caused hypotension and release of catecholamines. Subsequently, they identified catecholamines and β-adrenergic stimuli as agents that could reduce the increase in endothelial permeability induced by histamine and other vasoactive agents (see Grega et al40 for a review). Thus, while catecholamines increase blood pressure, they simultaneously improve endothelial barrier function and thus keep blood volume uncompromised. Recent studies have demonstrated that stimulation of β2-adrenergic receptors prevents the formation of small interendothelial gaps induced by vasoactive agents.21Other experiments have indicated that platelet-release products are important for maintaining endothelial barrier function.41 Vascular leakage is associated with gray platelet syndrome, a disease in which platelets do not release their contents. Thrombocytopenic animals suffer from vascular leakage and microvascular bleeding, but endothelial barrier function can be restored in these animals by exogenous serotonin. Factors that stabilized endothelial barrier function were also demonstrated in the blood of hemorrhaged animals40 ; such factors included catecholamines, vasopressin, and serotonin.Improvement of Endothelial Barrier Function by cAMPBoth β-adrenergic agents and serotonin increase the cellular cAMP concentration in ECs. Studies on ECs cultured on porous filters showed that a rise in cAMP level indeed reduced endothelial permeability, irrespective of the mechanism whereby such elevations were achieved (ie, induced by activation of adenylate cyclase or inhibition of phosphodiesterase).424344 This action involves cAMP-dependent PK42 and reduced phosphorylation of MLC,45 despite the fact that this action does not affect the Ca2+ concentration in ECs.4346 Furthermore, elevations of cAMP cause a redistribution of F-actin and a loss of interaction between F-actin and nonmuscle myosin.44 The effect of cAMP elevation on endothelial barrier function is rapid and continues so long as the cAMP-elevating stimulus is active. Because cAMP elevation also prevents and even reverses the prolonged increase in permeability induced by thrombin, it is likely that cAMP exerts its barrier-improving effect not only by reducing the degree of phosphorylation of MLC but also by improving or restoring the junctional complexes between cells.It should be noted that elevations of cAMP affect endothelial barrier function not only directly but also indirectly by reducing the interaction between leukocytes and the endothelium.47 A number of compounds, such as complement factors 3a and 5a, increase endothelial permeability not directly but via interaction with polymorphonuclear leukocytes, which then bind to and activate the endothelium of postcapillary venules in particular. Indeed, interference of leukocyte-endothelium interaction by anti–ICAM-1 and anti-CD18 integrin antibodies reduced experimental lung edema in animals.48 This suggests that selective elevation of cAMP in ECs and leukocytes may be a target for reducing vascular leakage. Cell type–selective elevation of cAMP is essential to avoid the unwanted effects of treatment on the heart and kidney. Furthermore, it should be realized that cAMP also acts on SMCs and induces smooth muscle relaxation. This may distally cause an increase in surface area of the capillary bed that is being perfused. As a consequence, cAMP may increase vascular leakage in selected areas of the vascular bed, despite the fact that overall endothelial barrier function improves.49Effects of cGMP and NO on Endothelial PermeabilityIn smooth muscle both cAMP and cGMP induce cell relaxation. Given the strong resemblance in contraction systems between SMCs and ECs, one might assume that cGMP also improves endothelial barrier function. In vivo data suggest that cGMP can have both positive and negative effects on endothelial permeability. In the lung circulation, cGMP-elevating agents caused a decrease of vascular leakage induced by oxidation damage, whereas in the peripheral circulation, cGMP-enhancing agents tend to worsen vascular leakage.505152In cultured aortic ECs, an elevation in cGMP reduced thrombin-enhanced permeability but did not affect basal permeability.53 cGMP also counteracted the H2O2-induced increase in hydraulic conductivity through monolayers of pulmonary artery ECs.54 Subsequent studies have shown that cGMP affects endothelial permeability both directly by activating cGMP-dependent PK2455 and indirectly by inhibiting and activating cAMP-degrading phosphodiesterases (Fig 2). The presence of active cGMP-dependent PK I was demonstrated in ECs of large arteries and veins5556 but was absent in those of the umbilical vein and kidney glomeruli.5557 Selective activation of cGMP-dependent PK reduced thrombin-induced accumulation of cytoplasmic Ca2+ ions as well as the passage of macromolecules through human aortic EC monolayers, whereas it had no effect on umbilical vein ECs.24The fact that cGMP counteracts increases in cytoplasmic Ca2+ concentration has an interesting consequence. The increase in cytoplasmic Ca2+ concentration generated by vasoactive agents not only affects endothelial permeability but also activates endothelial NO synthase. NO increases the cGMP concentration in the EC, which then again counteracts Ca2+ accumulation and the increase in endothelial permeability. Studies with NO synthase inhibitors have demonstrated that thrombin-induced NO indeed counteracts endothelial permeability increases in aorta and pulmonary artery ECs. Limitation of the extent of Ca2+-activated processes in these cells by NO also explains the observation that NO modulates its own Ca2+-dependent formation.58cGMP also affects endothelial permeability indirectly by influencing the cAMP concentration via cAMP-degrading phosphodiesterases. In umbilical vein ECs, cGMP reduced thrombin-enhanced endothelial permeability by inhibiting a cGMP-inhibited phosphodiesterase (PDE III). This effect on the permeability of umbilical vein cells was mimicked by synthetic PDE III inhibitors, whereas these inhibitors had little effect in human aorta and pulmonary artery ECs.24 In addition to PDE III, a cGMP-stimulated phosphodiesterase (PDE II) can be involved, which reduces the concentration of both cAMP and cGMP. Inhibition of PDE II prevents the loss of barrier function induced by H2O2, as shown in pulmonary artery ECs.5459 This suggests that PDE II can play a significant role in the regulation of endothelial permeability. This may be important, because NO and cGMP frequently increase the permeability of mesenteric and peripheral microvessels in vivo.505260 Unfortunately, little information is available regarding the presence of PDE II and PDE III in different types of ECs in vivo. Therefore, one can only speculate at present whether a difference in the amount of PDE II in ECs of different vascular beds contributes to the difference in the effect of cGMP-elevating agents on vascular leakage in the lung and peripheral vascular beds.PerspectiveFrom the aforementioned issues, it is clear that much has been learned in recent years about the regulation of endothelial permeability of macromolecules. Most of the recent data have concentrated on endothelial contraction and the formation of minute gaps at junctional areas of postcapillary venules. Challenging new information has become available about the structure and involvement of adherens and tight junctions. Other mechanisms of increased vascular exchange have also been recognized, such as pore formation through the bodies of ECs themselves. Therefore, the mechanisms contributing to vascular leakage may be different in various pathological conditions, not least because of the varying contribution of leukocytes and angiogenic growth factors. Furthermore, it should be realized that vascular leakage and edema are not identical. EC-cell and cell-matrix contacts are important for vascular leakage, but in edema the contacts between interstitial cells and between these cells and their matrix also contribute to an interstitial tension that counteracts interstitial swelling. Once this integrity of the interstitium becomes impaired, swelling may originate even without impairment of endothelial barrier function.61It is anticipated that in the coming years, the structures and biochemical pathways will be elucidated that underlie differences in the regulation of barrier function of the endothelium in different vascular beds, not only between different tissues such as the brain, heart, and lung, but also within tissues at the level of capillaries, arterioles, and postcapillary venules. More important for clinical practice, the new insights obtained will provide new leads for developing drugs to combat vascular leakage.Selected Abbreviations and AcronymsEC=endothelial cellMLC=myosin light chainMLCK=myosin light-chain kinasePK=protein kinaseSMC=smooth muscle cellDownload figureDownload PowerPoint Figure 1. Vasoactive agents can increase the permeability of endothelial monolayers by increasing actin–nonmuscle myosin interactions regulated by the phosphorylation of MLC (right) and by inducing the transient disintegration of cell-to-cell adherens junctions (left). CaM indicates calmodulin.Download figureDownload PowerPoint Figure 2. cGMP can affect endothelial permeability by interaction with cGMP-dependent PK I, cGMP-stimulated phosphodiesterase (PD-II), and cGMP-inhibited phosphodiesterase (PD-III). The contribution of the three pathways differs markedly in various types of ECs. cGMP-dependent PK I is present in ECs from large arteries but absent in those of the umbilical vein (see text). CaM indicates calmodulin.This study was financially supported by the Praeventiefonds (28-2622-2) and the Netherlands Heart Foundation.FootnotesCorrespondence to Prof V. van Hinsbergh, Gaubius Laboratory TNO-PG, Zernikedreef 9, PO Box 2215, 2301 CE Leiden, Netherlands. References 1 Palade GE. The microvascular endothelium revisited. In: Simionescu N, Simionescu M, eds. Endothelial Cell Biology in Health and Disease. New York, NY: Plenum Publishing Corp; 1988:3-21.Google Scholar2 Bundgaard M. The three-dimensional organization of tight junctions in a capillary endothelium revealed by serial-section electron microscopy. J Ultrastruct Res.1984; 88:1-17.CrossrefMedlineGoogle Scholar3 Majno G, Palade GE. Studies on inflammation, I: the effect of histamine and serotonin on vascular permeability—an electron microscopic study. J Biophys Biochem Cytol.1961; 11:571-605.CrossrefMedlineGoogle Scholar4 Majno G, Shea SM, Leventhal M. Endothelial contraction induced by histamine-type mediators: an electron microscopic study. J Cell Biol.1969; 42:647-672.CrossrefMedlineGoogle Scholar5 Wysolmerski RB, Lagunoff D. Involvement of myosin light-chain kinase in endothelial cell retraction. Proc Natl Acad Sci U S A. 1990; 87:16-20.Google Scholar6 Schnittler H-J, Wilke A, Gress T, Suttorp N, Drenckhahn D. Role of actin and myosin in the control of paracellular permeability in pig, rat and human vascular endothelium. J Physiol.1990; 431:379-401.CrossrefMedlineGoogle Scholar7 Garcia JGN, Davis HW, Patterson CE. Regulation of endothelial cell gap formation and barrier dysfunction: role of myosin light chain phosphorylation. J Cell Physiol.1995; 163:510-522.CrossrefMedlineGoogle Scholar8 Goeckeler ZM, Wysolmerski RB. Myosin light chain kinase-regulated endothelial cell concentration: the relationship between isometric tension, actin polymerization, and myosin phosphorylation. J Cell Biol.1995; 130:613-627.CrossrefMedlineGoogle Scholar9 Moy AB, Van Engelenhoven J, Bodmer J, Kamath J, Keese C, Giaever I, Shasby S, Shasby DM. Histamine and thrombin modulate endothelial focal adhesion through centripetal and centrifugal forces. J Clin Invest.1996; 97:1020-1027.CrossrefMedlineGoogle Scholar10 Rabiet M-J, Plantier J-L, Rival Y, Genoux Y, Lampugnani M-G, Dejana E. Thrombin-induced increase in endothelial permeability is associated with changes in cell-to-cell junction organization. Arterioscler Thromb Vasc Biol.1996; 16:488-496.CrossrefMedlineGoogle Scholar11 Gardner TW, Lesher T, Khin S, Vu C, Barber AJ, Brennan WA. Histamine reduces ZO-1 tight-junction protein expression in cultured retinal microvascular endothelial cells. Biochem J.1996; 320:717-721.CrossrefMedlineGoogle Scholar12 Joris I, Cuénoud HF, Doern GV, Underwood JM, Majno G. Capillary leakage in inflammation: a study by vascular labeling. Am J Pathol.1990; 137:1353-1363.MedlineGoogle Scholar13 Dvorak HF, Brown LF, Detmar M, Dvorak AM. Vascular permeability factor/vascular endothelial growth factor, microvascular hyperpermeability, and angiogenesis. Am J Pathol.1995; 146:1029-1039.MedlineGoogle Scholar14 Roberts WG, Palade GE. Increased microvascular permeability and endothelial fenestrations induced by vascular endothelial growth factor. J Cell Sci.1995; 108:2369-2379.CrossrefMedlineGoogle Scholar15 Dejana E. Endothelial adherens junctions: implications in the control of vascular permeability and angiogenesis. J Clin Invest.1996; 98:1949-1953.CrossrefMedlineGoogle Scholar16 Cuénoud HF, Joris I, Langer RS, Majno G. Focal arteriolar insudation: a response of arterioles to chronic nonsp