VEGF transiently disrupts gap junctional communication in endothelial cells

Cell-to-cell junctions determine the morphology of epithelial and endothelial cell layers lining organ cavities and blood vessels. The initial contact sites responsible for maintaining cell junctions in epithelial and endothelial cell monolayers are formed by homotypic cadherin/cadherin interactions. The formation of these so called adherens junctions is a prerequisite for tight junction assembly. Tight junctions link adjacent cells and allow only regulated transfer of molecules between the apical and the basal side of epithelial cell layers. Similarly, tight sealing of endothelial cell monolayers ensures that blood vessels are sealed towards the interstitial side of vascularized tissues. Cells in contact with each other form also another type of junctions, called gap junctions. These latter structures are formed by a family of proteins called connexins (Cx) (Goodenough et al., 1996). Six connexin molecules form either homo-or heteromeric connexons that, upon apposition in adjacent cells, form a pore spanning the membranes of the cells in contact with eachother and bridge the extracellular space between the cells. GJs are a means through which cells rapidly exchange information in the form of low molecular mass molecules such as second messengers, e.g. Ca²⁺, or cellular metabolites. In endothelial cells connexins Cx37, Cx40 and Cx43 are concomitantly expressed giving rise to connexons of various composition and specificity (Beyer et al., 2000).

Coupling of cells via GJs is essential for normal cell function and is reduced under many pathological conditions, e.g. in tumor tissue (Simon, 1999; Sulkowski et al., 1999; Simon and Goodenough, 1998). The exact role of these junctions in physiological processes such as the regulation of cell morphology, migration, growth and differentiation is, however, only poorly understood.

We are interested in the role of vascular endothelial growth factors (VEGFs), a subfamily of the PDGF family of polypeptide growth factors, in regulating cell-cell communication. Addition of either of two members of this ligand family, VEGF-A or placenta growth factor (PlGF), causes changes in vessel homeostasis (Bates et al., 1999; Veikkola and Alitalo, 1999). The role of VEGF family proteins has been studied in tissue culture models, in ex vivo vessel preparations and in intact vessels in animals (Ferrara and Alitalo, 1999; Dvorak, 2000; Tallquist et al., 1999; Chaytor et al., 1998; Schnittler, 1998; Brink et al., 2000). Here we use endothelial cell monolayers generated from umbilical vein-derived cells to study GJ communication. We found that VEGF-A, but not PlGF, rapidly and reversibly disrupts GJs. The signalling cascade stimulated by VEGF-A activates VEGF receptor 2 (VEGFR-2) followed by activaton of the c-Src tyrosine kinase and MAP kinases (MAPK) of the Erk subfamily. The activation of these kinases may be required directly or indirectly for phosphorylation of Cx43 giving rise to altered connexon function.

The 164 amino acid form of VEGF-A, VEGF₁₆₄, and VEGF-E were produced in our laboratory in Pichia pastoris using the pPICZαA expression system (Invitrogen, San Diego CA, USA; Scheidegger et al., 1999, and unpublished), PlGF was a kind gift from Dr C. Failla (IDI-IRCCS, Instituto Dermopatico dell’Immacolata, Istituto di Ricovero e Cura a Carattere Scientifico Rome, Italy). Myelin basic protein (MBP) and enolase were purchased from Sigma (St Louis MO, USA), CGP77675 was a generous gift from Dr F. Doriano (Novartis, Basel, CH), Lucifer yellow was from Molecular Probes (Eugene OR, USA), the MEK inhibitor U0126 from BioMol research lab Inc. (Plymouth MA, USA), the PI 3-kinase inhibitor LY294002 from Calbiochem (La Jolla CA, USA), Cx43 antibodies were from Transduction Laboratories (Franklin Lakes NJ, USA) and Chemicon (Tenecula CA, USA). Rabbit polyclonal anti-ERK1 (F15P) and anti-ERK2 (F13S) antibodies were kindly provided by Dr D. Fabbro (Novartis, Basel, CH). Protein A-Sepharose CL-4B was obtained from Amersham Pharmacia Biotech AB (Uppsala, Sweden) and [γ-³²P]ATP from ICN (Costa Mesa CA, USA).

Ea.hy926 cells were kindly supplied by Dr C. J. Edgell (Edgell et al., 1983) and were maintained in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FCS, GIBCO-BRL). HUVEC (Clonetics, Walkersville MD, USA) were cultured in EGM containing 8% FCS and were starved in EGM minus EGF plus 0.5% FCS for 3 hours before growth factor treatment. Both cell types express VEGF receptor 1 and 2 (Waltenberger et al., 1994; our unpublished work).

Cells were seeded into 24-well plates (2.0×10⁵ cells per well) on glass coverslips and grown for 2 days post confluency in DMEM containing 10% FCS. Three hours before stimulation coverslips were transferred to 35 mm dishes containing fresh medium. Cells were treated with 50 ng/ml growth factor for the indicated time with or without inhibitors. Ten cells per coverslip were microinjected with an Eppendorf 5242/5171 microinjecting device, the injected solution was 2% w/v Lucifer yellow (Lithium salt) in water. 15 minutes after injection, cells were fixed for 20 minutes in 3.7% formaldehyde, stained with DAPI and mounted on slides. The number of dye-containing cells was counted using a fluorescence microscope (Zeiss-Axiophot).

MAPK and c-Src activity were determined in immune complex kinase assays as described (Urich et al., 1995; Kaech et al., 1991). Connexin 43 phosphorylation in vitro was performed using the same protocol as for the determination of c-Src activity. To determine the type of amino acids at which Cx43 was modified, the gels were treated with 1 N NaOH for 30 minutes at 65°C. All the label was lost from the band representing Cx43 indicative of phosphorylation at serine and threonine residues. Quantification of electrophoresis gels was done with ImageQuant 5.0 (Molecular Dynamics).

Cells were grown and stimulated as described above. Cells were fixed in methanol at −20°C for 10 minutes. After blocking with 5% milk/PBS for 1 hour at room temperature the coverslips were incubated for 3 hours at room temperature with an anti-Cx43 antibody (Chemicon, MAB3068, diluted 1:250 in 0.1% BSA in PBS), washed with PBS, followed by incubation with a secondary antibody solution (anti-mouse-FITC, diluted 1:100 in 0.1% BSA in PBS) for 2 hours at room temperature. After washing, the coverslips were mounted on slides using Gelvatol as mounting medium, fluorescence microscopy was performed with a Zeiss-Axiophot microscope equipped with a kappa CF 8/1 camera.

Western blot analysis was performed on the same PVDF membrane used for in vitro phosphorylated material. After exposure to Fuji X-Ray film, the membrane was rehydrated and blocked for 30 minutes at room temperature in a solution of 3% gelatin in Tris buffered saline (TBS). After washing in TBS the membrane was incubated for 2 hours at room temperature with the anti-Cx43 antibody (Transduction laboratories, C13720, diluted 1:250 in 1% gelatin in TBS supplemented with 0.05% Tween, TTBS), followed by 3 washes in TTBS and incubation for 1 hour in the secondary antibody solution (rabbit anti-mouse, Southern biotechnology associates, 6170-01, diluted 1:500 in 1% gelatin in TTBS). After 3 washes in TTBS, the membrane was incubated with Protein A conjugated with alkaline phosphatase (Calbiochem, 539251, diluted 1:500 in 1% gelatin in TTBS) and developed with NBT/BCIP reagent (Boehringer Mannheim).

Statistical analysis by the student’s t-test was performed using the Sigma plot statistical analysis software. Quantitative data are presented as mean ± s.d. Experimental data interpreted as representing statistically different series showed P values ± 0.05.

We first investigated whether VEGF-stimulated cells still communicated with each other through GJs. Single cells in a monolayer were microinjected with a solution of lucifer yellow (LY) and the number of cells adjacent to the injected cell that became LY positive within 15 minutes was quantified using a fluorescence microscope. Fig. 1 shows that VEGF treatment maximally reduced GJC within 15 minutes in human umbilical vein endothelial cells (HUVECs) and in 30 minutes in Ea.hy926 cells (Edgell et al., 1983) and reversed to normal level within one to two hours. Both cell types show many of the hallmarks attributed to endothelial cells in vivo, such as uptake of low density lipoprotein, expression of endothelial cell surface specific markers (e.g. PECAM), formation of tubular structures in collagen or fibrin gels upon treatment with growth and differentiation factors, and chemotactic attraction by VEGF or fibroblast growth factor (FGF). These cell types express both receptors for VEGF.

We next investigated which of the two angiogenic receptors for VEGF family growth factors, VEGFR-1 or VEGFR-2, mediates disruption of GJC. This is possible by using receptor-specific VEGF isoforms. VEGF-A activates both VEGFR-1 and -2 while a pox family virus-encoded variant, VEGF-E, exclusively activates VEGFR-2 (Wise et al., 1999). A more distantly related family member, PlGF, is specific for VEGFR-1 (Petrova et al., 1999). Fig. 2 shows that the 164 amino acid isoform of VEGF-A and VEGF-E, but not PlGF, cause disruption of GJC indicating that receptor 2 activates the biochemical signals impinging on GJs. These findings were further corroborated by the fact that a recently described VEGFR-2-specific inhibitor, CGP41251 (Fabbro et al., 1999) completely blocked disruption of GJC by VEGF-A.

VEGFR-2 activates many signalling pathways, the most prominent targets being PLCγ, PI-3 kinase, MAP kinases, PKC, and Src family tyrosine kinases (reviewed by Petrova et al., 1999). The exact role of these signalling intermediates in the various responses of endothelial cells to VEGF remains unclear. Fig. 3A shows that a specific inhibitor of c-Src, CGP77675 (Missbach et al., 1999), blocks VEGF-induced disruption of GJC. Similarly, U0126, an inhibitor of MAP kinase kinase (MEK) that is specific for the Erk branch of MAPK signalling cascades, also blocks disruption of GJC by VEGF (Fig. 3B). Therefore, both enzymes are necessary for eliciting the response of VEGF on cell communication.

These conclusions were further supported by the experiments shown in Fig. 4. The time course of c-Src activation in vitro in response to VEGF-A correlated with disruption of GJs (Fig. 4A,B). Activation of the MAPK cascade was biphasic with an early and a late peak of activity (Fig. 4C,D). Only the early activity correlated with disruption of GJC. This suggests that late stimulation of this signalling cascade did not play a role in disruption of cell communication, it may, however, be required for restoring GJC. In conclusion, we show that VEGF-A rapidly stimulates c-Src and MAPK activity coincident with VEGFR-2 autophosphorylation (data not shown). This, together with the data shown in Fig. 3, suggests that the early peak of MAPK activity together with c-Src activity are required for disruption of GJs by VEGF family growth factors.

Connexins, the transmembrane proteins that form gap junctions, are susceptible to phosphorylation by a variety of kinases (Cooper et al., 2000; Saez et al., 1998). Phosphorylation at serine, threonine and tyrosine residues is responsible, at least in part, for the regulation of GJs (Lau et al., 1992; Lau et al., 1996; Warn-Cramer et al., 1998; Xie et al., 1997; Yao et al., 2000). Earlier experiments performed with EGF and PDGF showed transient phosphorylation of Cx43 upon growth factor stimulation. We investigated Cx43 phosphorylation in VEGF-stimulated cells. Fig. 5A shows that Cx43 is transiently phosphorylated in immunoprecipitates made from extracts of VEGF-stimulated cells. The steady state expression level of Cx43 was not affected by VEGF treatment as shown in the western blot given in Fig. 5B. This rules out the possibility that VEGF caused increased degradation of Cx43. Phosphorylation of Cx43 is alkali-labile indicative of serine or threonine modification and the time course of Cx43 phosphorylation coincides with the effect of VEGF on GJC as seen in Fig. 5C. This suggests, but does not formally prove, that modification of Cx43 is responsible for the disruption of GJs.

We finally investigated GJs microscopically using Cx43-specific antibodies (Fig. 6). All antibodies gave high cytoplasmic fluoresecence that may result from the presence of a large pool of soluble material not involved in junction formation. A small fraction of Cx43 was found in discrete membrane-associated patches as described before (Yao et al., 2000). VEGF did not change the morphology or number of these structures indicating that Cx43 phosphorylation may directly affect connexon function, rather than turnover.

The data presented here show that VEGF disrupts GJC in endothelial cells. These findings depict a very rapid and fully reversible biological readout through VEGFR-2. The characterization of the signalling pathways responsible for disruption of GJs indicates that both c-Src and MAPK are involved. These kinases are rapidly and transiently activated upon treatment with VEGF-A and reach maximal activity after about 15 minutes in agreement with earlier published work (Kroll and Waltenberger, 1997; Guo et al., 1995). Transient reduction of GJC therefore correlates with the time course of c-Src and the early phase of MAPK activation by VEGF-A. The late peak in MAPK activity that was observed upon VEGF stimulation apparently serves a purpose distinct from disruption of GJs. It may, for instance, be involved in restoring functional GJs late after VEGF treatment. A series of specific inhibitory drugs that block either PI 3-kinase, PKCs or eNOS had no effect on VEGF-induced GJ disruption ruling out a direct function of these enzymes in disruption of GJs (data not shown).

In cells of epithelial origin, EGF and PDGF have been reported to disrupt GJC. This seems to be due to activation of multiple pathways and in all cases Cx43 was phosphorylated (Saez et al., 1998; Loewenstein and Rose, 1992; Hossain et al., 1999a,b). Serine/threonine phosphorylation of Cx43 by PKC (Cooper et al., 2000; Lau et al., 1992), MAP kinases (Vikhamar et al., 1998; Hossain et al., 1999a,b; Kanemitsu and Lau, 1993) or indirectly upon stimulation by PI 3-kinase (Yao et al., 2000) have been described. However, not in all cases has phosphorylation been shown to be essential for GJC disruption (Hossain et al., 1999a; Hossain et al., 1999b; Pelletier and Boynton, 1994). The role of connexin phosphorylation in connexon and gap junction assembly remains therefore elusive. GJC is severely disrupted in most tumor cells. For instance, in cells transformed by the tyrosine kinase oncogene v-Src, disruption of communication was associated with phosphorylation of Cx43 at tyrosine residues although a functional correlation between tyrosine phosphorylation and altered GJ permeability was not always possible (Kanemitsu et al., 1997; Loo et al., 1995; Azarnia et al., 1988; Chang et al., 1985; Crow et al., 1992; Filson et al., 1990; Goldberg and Lau, 1993; Zhou et al., 1999). The role of tyrosine kinases in the regulation of cell-cell communication via GJs has also been deduced from the fact that c-Src, the proto-oncogene homologue of v-Src, modulates GJs in cells overexpressing this kinase (Azarnia et al., 1988) although it remains to be shown that this is a direct consequence of Cx43 phosphorylation by Src kinase.

Recent reports show that MAP kinase (Hill et al., 1994) and c-Src (Postma et al., 1998) also modulate GJC upon activation by G proteins. In the latter case, Src was shown to directly phosphorylate Cx43 at tyrosine. PDGF family tyrosine kinase growth factor receptors have been shown earlier to transiently recruit Src family tyrosine kinases into a membrane-bound high molecular mass complex. The ensuing activation of Src kinases is essential for entry of PDGF-treated cells into the cell cycle since blocking Src kinase abolishes the activity of the added growth factors (Erpel and Courtneidge, 1995). Disruption of GJs after treatment with growth factors is very rapid, peaks within minutes and returns to basal level within 2 hours similar to the VEGF response reported here. This apparently reflects the refractory nature of activation of signalling molecules such as MAPK and c-Src by growth factor receptors. Finally, when administered to endothelial cells, thromboxanes induce a response with the maximal effect observed many hours after ligand administration (Ashton et al., 1999). This indicates that this ligand activates a different set of signalling molecules than the growth factors mentioned above. Taken together, phosphorylation of connexins might be a means through which cells regulate GJ permeability under varying physiological conditions.

A detailed analysis of the role of the various Cx43 phosphorylation sites in protein turnover, localization and, most importantly, function in pre-assembled connexons will be of crucial importance in future experiments (Cooper et al., 2000; Saez et al., 1998). At present it is not possible to rule out the possibility that Src and MAPK play only an indirect role in GJ homeostasis. To define which phosphorylation sites of Cx43 are involved in regulating the formation and disassembly as well as the permeability of connexons requires to study mutant connexins in specially engineered cells lacking endogenous wild type molecules.

The role of this newly discovered readout from VEGFR-2 in blood vessel homeostasis is at present unclear. In a natural situation disruption of GJs by VEGF might be an acute response to growth factor released by endothelial cells or pericytes upon vessel damage. The effect of VEGF on GJC might prevent the propagation of second messenger molecules through cells coupled via GJs in the endothelial layer of the vessel wall. Ca²⁺, for instance, is known to spread among coupled endothelial cells in blood vessels. Blocking GJC will restrict the propagation of signals initiated at a site of local vessel injury into the adjacent healthy part of a vessel presumably preventing excessive vessel leakage that might otherwise cause massive edema. In support of this idea, changes in GJC between endothelial and smooth muscle cells have been recently reported to modulate endothelium-dependent relaxation of rabbit arteries in vessel explants (Chaytor et al., 1998).

We thank Dr C. J. Edgell for supplying Ea.hy926 cells, Dr D. Fabbro for Erk antibodies, Dr A. Mercer for the VEGF-E cDNA and Dr C. Failla for PlGF. We are also grateful to Drs R. Schwendener and R. Jaussi for critically reading the manuscript and Prof. J. Jiricny for helpful discussions and continuous support of our work. KBH and SS were supported in part by Bundesamt für Energiewirtschaft through Hauptabteilung für die Sicherheit der Kernanlagen (#65680).