Activated Neutrophils Induce Hyperpermeability and Phosphorylation of Adherens Junction Proteins in Coronary Venular Endothelial Cells*

The endothelial adherens junction is formed by complexes of transmembrane adhesive proteins, of which β-catenin is known to connect the junctional protein vascular endothelial (VE)-cadherin to the cytoskeleton and to play a signaling role in the regulation of junction-cytoskeleton interaction. In this study, we investigated the effect of neutrophil activation on endothelial monolayer integrity and on β-catenin and VE-cadherin modification. Treatment of cultured bovine coronary endothelial monolayers with C5a-activated neutrophils resulted in an increase in permeability as measured by albumin clearance across the monolayer. Furthermore, large scale intercellular gap formation was observed in coincidence with the hyperpermeability response. Immunofluorescence analysis showed that β-catenin and VE-cadherin staining changed from a uniform distribution along the membrane of control cells to a diffuse pattern for both proteins and finger-like projections for β-catenin in neutrophil-exposed monolayers. Correlatively, there was an increase in actin stress fiber formation in treated cells. Finally, β-catenin and VE-cadherin from neutrophil-treated endothelial cells showed a significant increase in tyrosine phosphorylation. Our results are the first to link neutrophil-mediated changes in adherens junctions with intercellular gap formation and hyperpermeability in microvascular endothelial cells. These data suggest that neutrophils may regulate endothelial barrier function through a process conferring conformational changes to β-catenin and VE-cadherin.

The endothelial adherens junction is formed by complexes of transmembrane adhesive proteins, of which ␤-catenin is known to connect the junctional protein vascular endothelial (VE)-cadherin to the cytoskeleton and to play a signaling role in the regulation of junctioncytoskeleton interaction. In this study, we investigated the effect of neutrophil activation on endothelial monolayer integrity and on ␤-catenin and VE-cadherin modification. Treatment of cultured bovine coronary endothelial monolayers with C5a-activated neutrophils resulted in an increase in permeability as measured by albumin clearance across the monolayer. Furthermore, large scale intercellular gap formation was observed in coincidence with the hyperpermeability response. Immunofluorescence analysis showed that ␤-catenin and VE-cadherin staining changed from a uniform distribution along the membrane of control cells to a diffuse pattern for both proteins and finger-like projections for ␤-catenin in neutrophil-exposed monolayers. Correlatively, there was an increase in actin stress fiber formation in treated cells. Finally, ␤-catenin and VE-cadherin from neutrophil-treated endothelial cells showed a significant increase in tyrosine phosphorylation. Our results are the first to link neutrophil-mediated changes in adherens junctions with intercellular gap formation and hyperpermeability in microvascular endothelial cells. These data suggest that neutrophils may regulate endothelial barrier function through a process conferring conformational changes to ␤-catenin and VE-cadherin.
The wall of exchange vessels consists of a layer of endothelial cells that connect to each other with closely opposed intercellular junctions. A major function of the junctional connection is to maintain the semi-permeable property of the endothelial barrier and to control the transvascular passage of solutes, fluid, and blood cells. Four types of junctions associated with endothelial cells have been identified: adherens junctions (AJ), 1 tight junctions, gap junctions, and complexus adherentes (1,2). AJ, formed by transmembrane adhesive proteins called cadherins, appear to be the main complex regulating macromolecular permeability in microvascular endothelium. Cadherins, specifically vascular endothelial (VE)-cadherin, are associated with the actin cytoskeleton through a family of proteins called catenins, including ␣-catenin, ␤-catenin, and plakoglobin (3,4).
The endothelial permeability is affected by many agonists including ␣-thrombin, histamine, and phorbol esters (5-9) as well as by a group of inflammatory cells, namely polymorphonuclear leukocytes (PMNs) (10 -14). At the site of injury or inflammation, circulating PMNs often adhere to and subsequently migrate through the endothelium and enter surrounding tissues (15). It has long been documented that the process of PMN adherence and migration is associated with an increase in endothelial permeability (10,11). Although much work has been dedicated to identify PMN-derived hyperpermeability factors (11)(12)(13)(14), little is known about the molecular targets of activated PMNs and the derived factors. Within this context, whether PMNs affect microvascular barrier function by altering the structural and functional integrity of the endothelium remains elusive.
Recent evidence suggests that the AJ is involved in PMNpromoted endothelial barrier dysfunction. Upon PMN adhesion, VE-cadherin proteolysis is accompanied by the disappearance of both VE-cadherin and catenins from AJ (14,16,17). Another hyperpermeability agonist, vascular endothelial growth factor, has been shown to increase the transendothelial flux of albumin concomitant with a loss of VE-cadherin (18). These studies have provided a possible linkage between PMN activation and VE-cadherin disorganization in the regulation of endothelial barrier function. Furthermore, evidence is accumulating that the VE-cadherin-mediated cell-cell adhesion is controlled by a dynamic balance between phosphorylation and dephosphorylation of the junctional proteins including cadherins and catenins. Increased tyrosine phosphorylation of ␤-catenin resulted in a dissociation of the catenin from cadherin and from the cytoskeleton, leading to a weak AJ (19,20). Similarly, tyrosine phosphorylation of VE-cadherin and ␤-catenin occurred in loose AJ and was notably reduced in tightly confluent monolayers (21).
This study focuses on the effect of activated PMNs on both endothelial permeability and phosphorylation of AJ proteins. The results showed that tyrosine phosphorylation of VE-cadherin and ␤-catenin was increased when cultured coronary venular endothelial cells (CVECs) were exposed to activated PMNs. ␤-Catenin and VE-cadherin staining revealed a marked decrease in the amount of these proteins at the cell periphery upon stimulation by PMNs. Correspondingly, PMN-treated monolayers showed a significant increase in permeability as measured by albumin clearance. This work is the first to correlate hyperpermeability and tyrosine phosphorylation of junc-tional proteins in response to PMNs in microvascular endothelial cells.
Isolation of Porcine PMNs-Platelet-poor plasma was obtained by centrifuging 40 ml of plasma at 2500 ϫ g for 10 min. The plasma was added to Hanks' buffer solution to a concentration of 10% and used as diluent or washing solution for the following procedures. To isolate PMNs, 20 ml of whole blood were diluted with 40 ml of Hanks' buffer solution, carefully layered on top of a 59% isotonic Percoll column, and centrifuged at 400 ϫ g for 20 min. The top band on the centrifuged Percoll column (containing lymphocytes) was discarded. The pellet containing red blood cells and PMNs was collected and mixed with two parts (v/v) of 2.5% gelatin in Hanks' buffer solution for incubation at 37°C for 45 min. The supernatant was collected and centrifuged at 300 ϫ g for 10 min. The remaining red cell pellet was then removed by hypotonic hemolysis. PMNs were harvested and washed twice by centrifuging at 400 ϫ g for 5 min. From this technique, 20 ml of blood yielded 10 6 -10 7 cells, of which 90 -95% were PMNs. In vitro analyses demonstrated that the isolated PMNs were viable and displayed normal chemotaxis function and metabolic oxygenation activity (10).
Immunoprecipitation and Western Analysis-CVECs were treated as described above for 30 min at 37°C, washed with phosphate-buffered saline, and lysed on ice for 30 min with 300 l of radioimmune precipitation buffer (1ϫ phosphate-buffered saline, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS) containing the protease inhibitors phenylmethylsulfonyl fluoride (0.1 mg/ml), aprotinin (1 g/ml), and sodium orthovanadate (1 mM) (Sigma). After centrifugation for 3 min at 14,000 rpm, cell lysates were removed, and protein concentrations were determined using the Bradford assay. For immunoprecipitation (IP), 100 g of cell lysate was used in 1 ml of radioimmune precipitation buffer. Five g of antibody against phosphotyrosine PY20 (Santa Cruz Biotechnology, Inc., Santa Cruz, California) was added, and the sam-ples were rotated for 1 h at room temperature. After this, 20 l of protein A/G PLUS-agarose (Santa Cruz) was added, and the IP was continued overnight at 4°C. After washing five times with radioimmune precipitation buffer, the samples were electrophoresed on 6% SDS-polyacrylamide electrophoresis gels and blotted to nitrocellulose membrane. Primary antibodies to ␤-catenin and VE-cadherin (Santa Cruz) were used at 0.5 g/ml for 1 h followed by secondary antibodies conjugated to horseradish peroxidase (Santa Cruz). Protein bands were detected using LumiGLO chemiluminescent substrate (New England BioLabs, Beverly, MA). Images of the bands were scanned by reflectance-scanning densitometry, and the intensity of the bands was quantified using NIH image software. For non-IP Western blotting, 20 g of protein was run/lane.
Measurement of Endothelial Permeability-CVECs were seeded at a density of 10 5 cells/cm 2 on gelatin-coated Costar Transwell membranes (VWR, Houston, Texas) and allowed to grow 4 -5 days until confluent. Cells were washed with phosphate-buffered saline and treated with drugs or PMNs for 30 min at 37°C. Fluorescein isothiocyanate-labeled bovine serum albumin (Sigma) was then added to the top (luminal) chamber at a concentration of 3.5 mg/ml, and the plates were returned to incubation at 37°C. After 45 min, samples were removed from the bottom (abluminal) and luminal chambers and analyzed by fluorometry. Sample readings were converted with the use of a standard curve to albumin concentration. These concentrations were then used in the following equation to determine the permeability coefficient of albumin Immunocytochemistry-CVECs were grown to confluence on gelatincoated coverslips and treated with C5a-activated PMNs for 30 min at 37°C. They were fixed with 2% paraformaldehyde for 15 min and permeabilized with 100% acetone for 2 min for ␤-catenin staining only. Primary antibodies to ␤-catenin and VE-cadherin were applied for 1.5 h followed by secondary antibodies conjugated to fluorescein isothiocyanate (Santa Cruz) for 45 min. Actin was visualized using rhodamine phalloidin (Molecular Probes, Eugene, OR) following permeabilization. Coverslips were then mounted on slides, and cells were examined under a Zeiss Axiovert 135 inverted microscope equipped with fluorescence filter sets and photographed using Kodak GoldMax 400 film.
Data Analysis-In the immunoblot studies, representative images of Western blots were selected to present. To compare the difference in optical densities of protein bands, the optical density values obtained from treated samples were expressed as a percentage of the controls. Analysis of variance was used to evaluate the significance of intergroup differences in the immunoblot analyses and permeability studies. A value of p Ͻ 0.05 was considered significant for the comparisons. respect to microvascular permeability, it was necessary to adapt a model previously used with large vascular endothelial cells for use with our venular endothelial cells. Our preliminary studies revealed that coating the polycarbonate membranes with gelatin was superior to fibronectin, and the gelatin imposed no major restriction to albumin clearance. Membranes with a gelatin coating and no cells had a 14-fold higher P a than membranes with gelatin and a confluent CVEC monolayer. The model was tested with drugs that would increase tyrosine and serine phosphorylation, two events that are known to be involved in the regulation of endothelial permeability in intact venules (9, 23). Fig. 1 shows that PAO (a tyrosine phosphatase inhibitor), phorbol 12-myristate 13-acetate (a serine kinase activator), and calyculin A (a serine phosphatase inhibitor) increase albumin permeability across the monolayer by 35, 75, and 74%, respectively. Hyperpermeability was also observed with CVEC monolayers when the typical inflammatory agonists ␣-thrombin and histamine were applied (data not shown). We therefore established a monolayer model using a two-chambered system that would allow us to detect increases in permeability. Fig. 2 shows that activated PMNs increase albumin permeability across CVEC monolayers to ϳ145% that of control cells. It is also clear from the data that nonactivated PMNs and C5a alone do not affect monolayer permeability (Fig. 2). We decided then to determine if activated PMNs increase permeability through processes involving phosphorylation. As Fig. 2 shows, we were able to attenuate the increases in permeability provoked by activated PMNs with both DAM and genistein, tyrosine kinase inhibitors, and BIM and Go6976, serine kinase inhibitors. Additionally, BIM seemed to cause a decrease in the basal permeability, but the significance of this effect has not been determined.

C5a-activated PMNs Induce Protein Flux Across Microvascular Endothelial Monolayers-
Activated PMNs Induce Tyrosine Phosphorylation of Endothelial Junctional Proteins-Knowing that PMNs are involved in hyperpermeability processes involving tyrosine and serine phosphorylation events, we looked for changes in phosphorylation levels of proteins found at intercellular junctions and, therefore, suspected of regulating endothelial integrity. Fig. 3 shows that activated PMNs increase tyrosine phosphorylation of ␤-catenin ϳ225% above control levels. As expected, this increase could be attenuated by the addition of a tyrosine kinase inhibitor (DAM); however, we also somewhat unexpectedly observed an attenuation of this increase with the addition of a serine kinase inhibitor (BIM) (Fig. 3). Another AJ protein, VE-cadherin, showed an ϳ100% increase in tyrosine phosphorylation content after exposing CVECs to activated PMNs (Fig.  4). As with ␤-catenin, the increase in tyrosine phosphorylation of VE-cadherin brought on by activated PMNs was attenuated with DAM and BIM (Fig. 4). These results suggest an interaction between tyrosine and serine phosphorylation events as they relate to AJ proteins and monolayer permeability. Additionally, C5a-activated PMNs appear to be a more potent stimulator of AJ protein tyrosine phosphorylation than the potent tyrosine phosphatase inhibitor PAO (Figs. 3 and 4).
Distribution of ␤-Catenin, VE-cadherin, and Actin in Response to Activated PMNs-In view of the result that activated PMNs induced phosphorylation either directly or indirectly of two major AJ proteins, we wanted to determine if changes in cellular distribution accompanied this hyperphosphorylation. Confluent CVEC monolayers were exposed to activated PMNs and then subjected to immunocytochemistry to detect changes in ␤-catenin and/or VE-cadherin localization. In addition, because VE-cadherin is associated with the actin cytoskeleton through the catenins, we examined actin distribution in PMNtreated cells. In control cells, ␤-catenin is uniformly dispersed around the periphery of a tightly confluent monolayer (Fig. 5A). When the cells are treated with activated PMNs, we see widespread gap formation as the cells retract and a loss of ␤-catenin staining in areas where individual cells have lost contact with neighboring cells (Fig. 5B). Note also that ␤-catenin assumes a finger-like configuration between these minimal cell contacts (Fig. 5B). VE-cadherin shows a peripheral staining pattern in control CVEC monolayers (Fig. 5C). However, PMN-treated cells show widespread gap formation and a diffuse staining pattern for VE-cadherin, leaving little if any VE-cadherin at the cell periphery (Fig. 5D). We surmised that for cells to retract in response to PMNs, there are probably accompanying changes in the actin cytoskeleton. Actin staining of control cells revealed that most of the F-actin was located at the cell periphery (Fig. 5E). However, in PMN-treated cells, we consistently observed an increase in stress fiber formation (Fig. 5F), which indicates cell contraction and shape change. In all of the PMNtreated endothelial monolayers, adherent PMNs were often observed where intercellular gaps were seen (Fig. 5B, arrow). DISCUSSION PMN-induced vascular leakage has long been implicated in the development of coronary microvascular dysfunction during ischemia-reperfusion injury and other types of cardiovascular disorders (12, 13, 24 -27). Our study is the first systematic examination of the direct effects of PMN activation on the function and structure of coronary microvascular endothelial cells. We had previously demonstrated the interaction of PMNs with the endothelium using intact, perfused coronary venules and arterioles (10). We modified existing monolayer models (12,13,18) to accommodate CVECs to study processes occurring at the molecular level in these cells. Compared with the intact, perfused coronary venules, the CVEC monolayer exhibited a higher basal permeability but a similar responsiveness to the hyperpermeability stimulators.
The complexity of the precise molecular events that lead to microvascular leakage is becoming evident. To allow large molecules and even blood cells to pass through the endothelial layer, individual endothelial cells must contract to form gaps. In this process, the intercellular junction may undergo conformational changes, facilitating the gap formation and macromolecular transflux. VE-cadherin, a transmembrane protein, and ␤-catenin, a signaling protein that links cadherins to the actin cytoskeleton, were found to exhibit increases in tyrosine phosphorylation levels concomitant with an increase in permeability. We suggest that this phosphorylation is an important signal responsible for breaking down the AJ and diminishing the ability of neighboring cells to interact. Our results and those of others showed the disappearance of AJ proteins at the cell periphery under conditions of hyperpermeability (14,16,17). Additionally, we observed an increase in the formation of actin stress fibers that coincides with the AJ breakdown and permeability increases. In PMN-treated monolayers, although not all cells show stress fiber formation, there is an overall increase in stress fiber formation across the monolayer versus control cell monolayers. Presumably these stress fibers make contact with opposite sides of the cell membrane and "pull" the cell into a spherical shape, breaking the contacts between neighboring cells. The question then becomes, to what protein or protein complex are the actin fibers anchoring during hyperpermeability situations? Our hypothesis is that AJ proteins anchor the peripheral actin filaments under resting conditions and allow for the integrity of the monolayer to remain intact. In response to agonists, actin stress fibers form and interact with structure(s) including the AJ to instigate changes in cellular morphology, leading to a weak or diminished AJ. Phosphorylation of VE-cadherin and ␤-catenin may be a critical signaling event that elicits such structural changes. In further support of this, our data show that inhibiting phosphorylation of VE-cadherin and ␤-catenin blocks the hyperpermeability response in the endothelial cells.
The precise mechanism by which PMNs increase microvascular permeability has not been established. A body of evidence supports that the leakage is mainly attributed to PMN-derived cytotoxic mediators including oxygen radicals and various proteases (28 -31). At the site of PMN adhesion, large amounts of the cytotoxic factors are released and accumulate, initiating a full cascade of signaling reactions in the endothelium. Ultimately, endothelial cell contraction occurs, and intercellular gaps are formed. Based on our previous and present studies, it appears that protein phosphorylation is one of the signaling events that is activated in response to PMN adhesion, and these kinase cascades result in the hyperpermeability response. In support of this, we found that activated PMNs induce increases in tyrosine phosphorylation of AJ proteins accompanied by an increase in permeability. Interestingly, we also discovered that serine kinase inhibitors could attenuate the permeability increase provoked by PMNs, and a serine phosphatase inhibitor increased basal permeability. However, our attempts to detect serine phosphorylation of ␤-catenin and VE-cadherin were unsuccessful. It is likely that activated PMNs initiate events that directly or indirectly up-regulate both the tyrosine and serine phosphorylation pathways, which may interact (cross-talk) with each other. Our opinion is that both types of phosphorylation are necessary to instigate hyperpermeability of CVEC monolayers.
This study is the first to correlate PMN activation with phosphorylation of AJ proteins and hyperpermeability in cultured microvascular endothelial monolayers. We had previously reported that tyrosine phosphorylation of the focal adhesion proteins paxillin and focal adhesion kinase occurs in CVEC monolayers in conjunction with a breakdown of endothelial barrier function (23). Therefore, we now have evidence that alterations in both the focal adhesion complex and the AJ can result in hyperpermeability of the microvascular endothelium and that activated PMNs can induce at least some of these events. We have established the appropriate model for further studies aimed at methodically dissecting the molecular events associated with changes in CVEC permeability.