Phosphorylation of vascular endothelial cadherin controls lymphocyte emigration

Lymphocytes emigrate from the circulation to target tissues through the microvascular endothelial cell (EC) barrier. During paracellular transmigration cell-cell junctions have been proposed to disengage and provide homophilic and heterophilic interaction surfaces in a zip-like process. However, it is not known whether ECs modulate junction proteins during this process. Here we show that tyrosine phosphorylation of adherens junction vascular endothelial cadherin (VEC) is required for successful transendothelial lymphocyte migration. We found that adhesion of lymphocytes or activation of the endothelial intercellular adhesion molecule 1 (ICAM1) led to tyrosine phosphorylation of VEC. Substitution of tyrosine for phenylalanine in VEC at positions 645, 731 or 733 produced ECs that were significantly less permissive to lymphocyte migration. We also found that these same tyrosine residues are involved in ICAM1-dependent changes of VEC phosphorylation. ICAM1 activation enhanced transendothelial permeability, suggesting the occurrence of junction disassembly. In agreement, the expression of VEC mutated at Y645F, Y731F or Y733F predominantly affected lymphocyte transmigration in paracellular areas. Taken together, these results demonstrate that phosphorylation of adherens junctions constitutes a molecular endpoint of lymphocyte-induced vascular EC signaling and may be exploited as a new target of anti-inflammatory therapies.


Introduction
Intercellular adhesion molecule 1 (ICAM1) on the luminal surface membrane of vascular endothelial cells (ECs) plays a crucial role during the egress of lymphocytes from the vascular compartment. It binds to ␤2 integrins, such as leucocyte-function-associated antigen 1 (LFA-1), a heteromer of ␣L (ITGAL, also known as CD11a) and ITGB2 (CD18), on lymphocytes and promotes their stable arrest (Butcher, 1991). However, endothelial ICAM1 is not only a docking receptor for lymphocyte adhesion and migration but its engagement also triggers endothelial signaling cascades that contribute fundamentally to vascular compliance to lymphocyte diapedesis and the endothelial inflammatory response (Turowski et al., 2005). ICAM1-induced signaling comprises changes in levels of intracellular Ca 2+ and dynamic actin as well as activation of the small GTPase Rho, and Src and C protein kinases. Despite the identification of an increasing number of key players involved in ICAM1-mediated endothelial signaling, downstream effectors and molecular endpoints that ultimately modulate lymphocyte migration remain elusive.
Both tight and adherens junction (AJ) proteins form attachments between ECs, and are major sites of regulation in the transport of molecules and cells across vascular barriers (Bazzoni and Dejana, 2004). At least in the case of paracellular leukocyte extravasation, various junction molecules disengage and, in turn, form homo-and heterophilic interactions with leukocyte adhesion proteins (Imhof and Aurrand-Lions, 2004;Muller, 2003;Rao et al., 2007). In this process the EC has been ascribed a rather passive role, where junctions are forced open by the migrating leukocyte (Imhof and Aurrand-Lions, 2004;Shaw et al., 2001). We investigated whether intercellular junction modulation is part of the endothelial response to lymphocyte migration.

Results and Discussion
Migration of antigen-activated lymphocytes across brain microvascular ECs in vitro does not require stimulation of inflammatory cytokines, occurs in the absence of chemokine gradients and is predominantly dependent on ICAM1 but not vascular cell adhesion molecule 1 (VCAM1) (Greenwood et al., 1995). ICAM1-mediated signaling pathways involve Rho GTPases and MAP kinase cascades and have generally been studied by antibody-mediated crosslinking of serum-starved cells (Turowski et al., 2005). In the rat brain microvascular EC cell line (GPNT) ICAM1 activation led to a rapid and transient increase of tyrosine phosphorylation of a number of proteins Lymphocytes emigrate from the circulation to target tissues through the microvascular endothelial cell (EC) barrier. During paracellular transmigration cell-cell junctions have been proposed to disengage and provide homophilic and heterophilic interaction surfaces in a zip-like process. However, it is not known whether ECs modulate junction proteins during this process. Here we show that tyrosine phosphorylation of adherens junction vascular endothelial cadherin (VEC) is required for successful transendothelial lymphocyte migration. We found that adhesion of lymphocytes or activation of the endothelial intercellular adhesion molecule 1 (ICAM1) led to tyrosine phosphorylation of VEC. Substitution of tyrosine for phenylalanine in VEC at positions 645, 731 or 733 produced ECs that were significantly less permissive to lymphocyte migration. We also found that these same tyrosine residues are involved in ICAM1-dependent changes of VEC phosphorylation. ICAM1 activation enhanced transendothelial permeability, suggesting the occurrence of junction disassembly. In agreement, the expression of VEC mutated at Y645F, Y731F or Y733F predominantly affected lymphocyte transmigration in paracellular areas. Taken together, these results demonstrate that phosphorylation of adherens junctions constitutes a molecular endpoint of lymphocyteinduced vascular EC signaling and may be exploited as a new target of anti-inflammatory therapies. Fig. 1. Tyrosine phosphorylation of VEC following ICAM1 crosslinking or adhesion of lymphocytes. (A) GPNT cells were grown to confluence, serum starved and ICAM1 crosslinked (XL) for the indicated times. Total protein extracts (~50 g) were analyzed by immunoblotting with antibodies against phosphorylated tyrosine. Blots were subsequently stripped and probed for ␤-catenin as loading control. Four proteins with apparent molecular masses of 220 kDa, 140 kDa, 94 kDa and 83 kDa [previously identified as cortactin (Durieu-Trautmann et al., 1994)] displayed clearly enhanced tyrosine phosphorylation and are indicated by filled arrowheads. Open arrowhead indicates the position of the IgG heavy chains of the crosslinking antibody. (B) Confluent GPNT cells were serum starved and either (a,m) left untreated or ICAM1 crosslinked for (b) 15 minutes or (c-l) the different times indicated. Cells were fixed, extracted and stained for (a,b) surface ICAM1, (c-g) phosphorylated tyrosine, (h-l) F-actin or (m) VEC. Bar, 10 m. (C-F) Confluent (C,D,F) GPNT cells or (E) mouse brain endothelioma EC, bEND5, were serum starved and subjected to crosslinking of ICAM1 (XL) or unrelated surface molecules (MHC class I; endomucin, EMCN). At the indicated times cells were washed and lysed. VEC immunoprecipitates were then analyzed by immunoblotting using either antibodies against phosphorylated tyrosine or VEC. (D) The amount of tyrosine-phosphorylated VEC (see C) was quantified by densitometry from five independent experiments and expressed as fold-increase of untreated controls (mean ± s.e.m.). (F) Prior to ICAM1 crosslinking (15 minutes), and where indicated, cells were pre-treated using PP2 (10 M, 30 minutes), C3 transferase (2 g/ml, 16 hours), cytochalasin D (CD, 2 M, 30 minutes) or BAPTA (BA, 20 M, 30 minutes). (G,H) Confluent GPNT cells were co-cultured with rat peripheral lymph node (PLN) lymphocytes (approximately five lymphocytes per EC). At the indicated times cells were lysed and VEC immunoprecipitates prepared and analyzed as described above. (H) Data from four independent experiments were quantified by densitometry, normalized and expressed as foldincrease of untreated controls (mean ± s.e.m.). Significant differences were determined by Student's t-test (*P<0.003, **P<0.002). In all blots the position of size markers (in kDa) is indicated on the left. which was maximal within 10-15 minutes of ICAM1 stimulation (Fig. 1A). When analyzed by immunocytochemistry (Fig. 1B) the increase in phosphorylated tyrosine was most marked at the intercellular junction area and preceded major actin rearrangements typically seen following ICAM1 crosslinking (Adamson et al., 1999). Immunoprecipitation and subsequent analysis of phosphorylated tyrosine of junction proteins revealed that VEC (Fig. 1C,E), but not the tight junction protein 1 (ZO-1), occludin, or any of the catenins (Fig. 2), displayed altered tyrosine phosphorylation following ICAM1 crosslinking. Phosphorylation of VEC was rapid and transient, reaching a maximum after 15 minutes of ICAM1 crosslinking (Fig. 1D). It depended on upstream signaling by Rho GTPase, dynamic actin and also intracellular Ca 2+ (Fig. 1F), all of which are essential for successful lymphocyte migration (Turowski et al., 2005). No involvement of Src family protein kinases was observed in our experimental set up. This is in contrast to a recent report by Allingham et al. (Allingham et al., 2007) and may reflect that different signaling modules are activated and used by different subsets of ECs and/or leukocytes. Indeed, T-cell adhesion to GPNT cells also induced a significant rise in VEC tyrosine phosphorylation (Fig. 1G,H), albeit with a slightly delayed time course (presumably reflecting settling and adhesion times). Taken together these data suggest that interendothelial junctions can be molecular endpoints of ICAM1-mediated signaling during leukocyte adhesion and migration.
To study the functional significance of ICAM1-mediated VEC phosphorylation, the structure of the cytoplasmic domain of VEC was further analyzed in silico and by mutagenesis. The cytoplasmic tail of VEC contains eight conserved tyrosine residues (Fig. 3A), three of which are predicted to constitute phospho-acceptor sites for insulin receptor family (Y645, Y731) or Src family (Y685) protein kinases (http://kinasephos.mbc.nctu.edu.tw, data not shown). A structural computer model of the VEC intracellular domain bound to ␤-catenin suggested that Y725, Y731 and Y733 are exposed and accessible to phosphorylation, whereas Y685, Y757 and Y774 are oriented towards the bound ␤-catenin and, possibly, inaccessible (Fig. 3B). Y645 and Y658 are located outside of this area in a domain predicted to mediate binding to p120 (Lampugnani et al., 2002).
Recently, Y658, Y685 and Y731 have been described as target phosphorylation sites (Potter et al., 2005;Wallez et al., 2006), and we therefore asked whether phosphorylation of any of these three tyrosines residues is important for lymphocyte migration. For this we derived EC lines from VEC-null mouse endotheliomas that reexpressed wild-type (wt) VEC, Y658F-VEC, Y685F-VEC or Y731F-VEC. The VEC expression levels and localization were similar in all four cell lines (Fig. 4A,B), suggesting that Y to F mutations do not interfere with junctional targeting. Migration of activated T cells across VEC-null ECs was very low but significantly enhanced when wt VEC was re-expressed ( Fig. 4C). A similar, twofold increase was observed when either a clonal cell line that stably re-expressed wt VEC or a pool of cells transiently expressing EGFP-VEC was analyzed. Absolute migration rates measured after 4 hours of lymphocyte co-culture increased from 10% to 20-25% when VEC was reconstituted in VEC-null endothelioma cells and were then in the same range as those usually observed for GPNT cells (ca. 30%) indicating that the expression of VEC is vital to the functional integrity of the vascular endothelium. This was somewhat unexpected because VEC disruption using antibodies has been reported to enhance neutrophil migration in vivo (Gotsch et al., 1997). However, VEC elimination Immunoprecipitates were then analyzed by immunoblotting using antibodies against phosphorylated tyrosine, ZO-1, VEC or catenin. Black and white arrowheads in A-E indicate the position of migration of VEC and relevant catenins, respectively, as determined by stripping and re-probing of the immunoblots. Immunoprecipitates of p120 did not contain detectable VEC, whether phosphorylated (black arrowhead) or not (data not shown). (F) The chicken occludinexpressing GPNT cell line (see supplementary material Fig. S1) was grown to confluence, serum starved and ICAM1 crosslinked (XL). At the indicated times chicken occludin was immunoprecipitated and analyzed by immunoblotting for phosphorylated tyrosine or occludin. In all blots the position of size markers (in kDa) is indicated on the left. leads to significant changes in the composition of AJs (Zanetta et al., 2005) and might, in this way, inhibit leukocyte migration indirectly. Alternatively, VEC might have opposing effects on the migration of lymphocytes and neutrophils. When cells that express VEC mutants were analyzed we found that the re-expression of VECs carrying single mutations in Y658 or Y685 did not have any effect on lymphocyte transmigration (Fig. 4D). However, mutation of Y731 led to an approximately twofold reduction in lymphocyte migration. This indicated that VEC phosphorylation may be important for leukocyte migration.
To test the role of intracellular tyrosines of VEC more systematically and in an EC system more relevant to inflammation, VEC mutants containing single Y to F substitutions were transiently expressed as N-terminal fusion proteins of EGFP in GPNT cells. Expression levels and overall localization of the individual VEC mutants were similar and mostly restricted to the periphery of the cell (Fig. 5A). Some cells expressed the protein at very high levels throughout the cell but these appeared non-viable and were never integrated into endothelial monolayers. In all other expressing cells, EGFP was almost exclusively found at cell-cell junctions once EC monolayers were formed (Fig. 5B). Again we observed that the introduction of Y731F-VEC significantly reduced lymphocyte migration. In addition we also observed impaired lymphocyte migration across GPNT cells that expressed Y645F-VEC or Y733F-VEC (Fig. 5C). By contrast, the introduction of wt VEC or any other Y to F mutant did not affect migration. The inhibitory effect of Y645F-VEC and Y733F-VEC was also corroborated in a VEC-null background. Transient expression of Y645F-VEC-GFP or Y733F-VEC-GFP but not any other VEC-GFP mutants, produced endothelioma cells with a similar propensity to reduced lymphocyte migration (Fig. 5D). In all cases the mutant VEC clearly interfered on the level of lymphocyte diapedesis because adhesion rates were comparable.
Commercially available antibodies against VEC phosphorylated at Y658 or Y731 (Potter et al., 2005) were not reactive with VEC immunoprecipitates from ICAM1-crosslinked cells (data not shown). They also appeared to recognize a protein of the same size whether full-length or truncated VEC (Lampugnani et al., 2002) was expressed (data not shown). To examine whether ICAM1induced tyrosine phosphorylation involved the tyrosine at positions 645, 731 or 733 the phosphorylation of wt and mutant GFP-VEC was analyzed in CHO expressing ICAM1 cells (CHO-ICAM1). Crosslinking of ectopic human ICAM1 on CHO cells induced tyrosine phosphorylation of co-expressed GFP-VEC in a time dependent manner (Fig. 6A), similar to that observed in GPNT cells. When the phosphorylation status of the Y to F mutants of GPF-VEC was tested, only mutant Y733F-VEC displayed significantly reduced tyrosine phosphorylation in response to Journal of Cell Science 121 (1) Fig. 3. Conserved tyrosines in the cytoplasmic domain of VEC. (A) Sequence alignment of cytoplasmic domains of VEC from mouse, rat, human and chicken, and mouse E-cadherin. Tyrosines conserved in VEC and their relative position (mouse versus human) are indicated in green. Identical residues, conserved and semiconserved substitutions are indicated by asterisks, colons and dots, respectively. Grey boxes correspond to those parts of E-cadherin that interact with ␤-catenin when crystallized together (Huber and Weis, 2001). (B) Mouse VEC was computer-modeled on the crystal structure of mouse E-cadherin in the E-cadherin-␤catenin complex (i.e. boxed in A) (pdb: 1i7x, 1i7w). Shown in the left panel is a ribbon representation of this model with the position and orientation of six tyrosines highlighted in green. The right panels represent enlarged views of areas surrounding these tyrosines. The ␤-catenin chain is in a space filling representation. Significantly, in this model, Y685, the residue predominantly phosphorylated following VEGF stimulation (Wallez et al., 2006), is not accessible when ␤-catenin is bound. Fig. 4. Y731 within the intracellular domain of VEC is important for lymphocyte migration. Mouse endothelioma cell lines, null for VEC and stably re-expressing wt or Y to F mutants of VEC were grown to confluence. (A) Equal amounts of proteins were analyzed by immunoblotting using anti-VEC and anti-ERK antibodies. The position of size markers (in kDa) is indicated on the left. (B) Immunocytochemical analysis of the VEC distribution. Bar, 10 m. (C) Mouse endothelioma cell lines, null for VEC, stably re-expressing wt VEC or transiently nucleofected with the VEC-GFP-expressing plasmid pEGFP-NЈ-VEC were grown to confluence. They were then incubated with antigen-specific T cells, which were allowed to adhere and migrate for 4 hours. Adhesion (white) and migration (black) across these EC populations were then determined as described in the Materials and Methods section. Results are expressed as the percent increase of VEC-null EC (mean ± s.e.m. of six replicates from five independent experiments). (D) Lymphocyte migration across the indicated stable mouse endothelioma cell lines. Adhesion (white) and migration (black) across individual transfected EC populations were then measured as above. Results are expressed as % of control cells re-expressing wt VEC (mean ± s.e.m. of six replicates from at least three independent experiments). Significant differences were determined by Student's t-test (*P<0.005, **P<0.0001). (B) Three days after transfection, GPNT cells were fixed and VEC-GFP was expression analyzed using confocal microscopy. Bar, 10 m. (C) Nucleofected GPNT cells were grown to confluence for 24-48 hours at which point equal expression was verified by fluorescent microscopy (see A). Lymphocyte adhesion (white) and migration (black) were then measured as described in Fig. 4. (D) Mouse VEC-null endothelioma cells (see Fig. 4) were nucleofected with wt or the indicated Y to F mutants of pEGFP-NЈ-VEC before T cell adhesion and migration was assessed. Significant differences were determined by Student's t-test (*P<0.005; **P<0.0001). ICAM1 ligation (Fig. 6B), suggesting that this mutation interfered with tyrosine phosphorylation of VEC. Next, GFP-VEC was also isolated and analyzed from 32 P-labelled cells. Significantly, GFP-VEC was immunoprecipitated as a strongly phosphorylated protein, with the majority of phosphate present as phosphorylated serine and only very little phosphorylated tyrosine (data not shown). This observation is also in agreement with the recently reported phosphorylation of S665 by PAK (Gavard and Gutkind, 2006). Tryptic phosphopeptide mapping revealed at least five groups of peptides with apparently variable phosphorylation levels (Fig. 6C,D; supplementary material Fig. S2). When ICAM1 was crosslinked most phosphopeptides remained unchanged but two peptides displayed significantly reduced chromatographic mobility, indicative of additional phosphorylation (arrows in Fig. 6D). The Y645F mutation suppressed phosphorylation of one peptide but enhanced that of the other, indicating that this mutation interferes with the recognition of the phosphorylation site. Mutation Y731F induced a perfect reversal to VEC from non-crosslinked cells suggesting that this site was directly phosphorylated. Overall phosphorylation was low in Y733F-VEC and the migration of most phosphopeptides was altered, which suggests that this mutation renders most sites unphosphorylatable. More refined analysis of phosphopeptides by mass spectroscopy will be required to delineate the exact phosphorylation changes that occur on tyrosine and serine residues when ECs are stimulated. In summary, VEC mutations of Y645, Y731 or Y733, which dominantly inhibit lymphocyte migration, also affect ICAM1-mediated VEC phosphorylation, suggesting that these events are mechanistically linked.
VEC has been reported to be phosphorylated on tyrosine under various conditions (Andriopoulou et al., 1999;Esser et al., 1998) and, indeed, we found that tyrosine phosphorylation was also increased when GPNT cells were treated with a variety of vasoactive compounds (Fig. 7A). Generally it is thought that such phosphorylation induces the disruption of AJs, which can be measured by concomitant rises in transendothelial permeability. We found that ICAM1 crosslinking is likely to affect AJ organization because it also induced an increase in paracellular permeability (Fig. 7B) and monolayer impedance (data not shown). However, in contrast to growth factor stimulation, where this is frequently associated with altered catenin binding (Lampugnani et al., 1997;Potter et al., 2005;Weis et al., 2004), ICAM1 ligation did not affect the levels of VEC-associated ␣-, ␤or ␥catenin (Fig. 7C,D), suggesting that the molecular mechanism may be different. Intuitively, ICAM1-mediated VEC phosphorylation might lead to the disengagement of AJs and thus facilitate the path Journal of Cell Science 121 (1) Fig. 6. ICAM1 induced VEC phosphorylation in wt and mutant VEC. (A) CHO-ICAM1 cells were transfected with wt pEGFP-NЈ-VEC or not, grown to confluence and then starved. Cells were then subjected to ICAM1 crosslinking and VEC-GFP immunoprecipitated and analyzed by immunoblotting for phosphorylated tyrosine and VEC. C, untransfected controls; PV, sample from cells pretreated with pervandate (100 M). (B) As described for A, except that the CHO-ICAM1 cells were transfected with wt or Y to F mutants of VEC as indicated. ICAM1 crosslinking was 10 minutes. (C) The sequence of the cytoplasmic domain of mouse VEC (as shown in Fig. 3A) has been used to predict tryptic peptides. Amino acids in small letters in peptide 11 are from the linker sequence to EGFP (which is not shown). Five out of the eleven peptides (bold) contain many phosphorylatable serine and tyrosine residues in line with our observation that VEC is strongly phosphorylated on serine and less so on tyrosine (data not shown). Note, in contrast to the report by Wallez et al. (Wallez et al., 2006) we have assumed that trypsin digestion does not occur when a proline is found at the carboxylic side of lysine or arginine. (D) CHO-ICAM1 cells were transfected with pEGFP-NЈ-VEC as described above. Cells were labeled with 32 P and then subjected to ICAM1 crosslinking or not. VEC-GFP was immunoprecipitated and processed for tryptic peptide mapping. Arrows denote the position of crosslinking-specific phosphopeptides. The three maps displayed in a single row were chromatographed in the same tank and Rf values were directly comparable. Enlarged sections of the phosphopeptide maps showing ICAM1 crosslinking specific phosphopeptides are shown in supplementary material Fig. S2. during paracellular migration. In support of such a mechanism we found that the expression of Y645F-VEC, Y731F-VEC or Y733F-VEC led to a significant reduction of migration in paracellular areas of the EC monolayer (Fig. 7E) and this correlated well with the general rate of inhibition (see Fig. 5C). Therefore, paracellular but not transcellular migration might be affected by dysfunctional VEC. However, more careful analyses using dynamic confocal microscopy of stable transfectants are required to draw a definite conclusion, in particular because transcellular migration also relies on juxta-junctional positioning of the lymphocyte and, subsequently, on a transcytotic pathway (Carman et al., 2007;Millan et al., 2006) -in which VEC may also be involved (Gavard and Gutkind, 2006;Xiao et al., 2005).
Leukocyte adhesion and subsequent ICAM1 engagement on EC induces an endothelial response and a complex network of signaling pathways regulating the inflammatory response of the endothelium (Turowski et al., 2005). From the results presented here, tyrosine phosphorylation of VEC meets all the criteria for a molecular endpoint of an endothelial signaling cascade that regulates lymphocyte migration: it is induced by surface engagement of ICAM1, it can be placed downstream of pathways that have previously been identified to affect lymphocyte migration and it appears to regulate paracellular migration directly. The importance of this mechanism and its general validity is underscored by recent findings by Allingham et al. (Allingham et al., 2007) who reported similar VEC phosphorylation to operate when neutrophils migrate across human umbilical vein endothelial cells. Collectively, these results demonstrate that VEC plays a more active role during leukocyte migration than previously anticipated. Single amino acid mutations, namely at Y645, Y731 or Y733, suppressed this function and, surprisingly, did so in a dominant fashion. Since dominant effects are rarely observed with mutations introducing non-phosphorylatable amino acids (as opposed to phospho-mimetic substitutions), it will be interesting to elucidate how these mutations interfere with the function of endogenous VEC during leukocyte migration. Currently, anti-adhesion therapies are amongst the most specific and advanced strategies to treat inflammation. The identification of a dominant molecular endpoint of adhesion-induced signaling may constitute a valid target for novel anti-inflammatory therapies.

Cell culture and treatment
The rat brain microvascular EC line GPNT (Romero et al., 2003) and mouse brain endothelioma cells bEND5 (Lyck et al., 2003) were grown as previously described. Mouse ECs with a homozygous null mutation of the VEC gene and the cell line reexpressing wild-type VEC derived by retroviral gene transfer have been described in detail (Lampugnani et al., 2002). Additionally, cell lines that stably express Y658F-VEC, Y685F-VEC, or Y731F-VEC have been generated in the same manner. Fig. 7. ICAM1-mediated VEC phosphorylation affects paracellular migration and coincides with increased EC permeability. (A) GPNT cells were grown to confluence, serum starved and then either left untreated (NT) or subjected to ICAM1 crosslinking (XL), 50 ng/ml VEGF, 10 M lysophosphatidic acid (LPA), 10 M bradykinin (BK), 100 M histamine (HST) or 1 U/ml thrombin (TBN) for 15 minutes. Subsequently, cells were lysed and VEC immunoprecipitates analyzed by immunoblots using anti-phosphorylated tyrosine or anti-VEC antibodies. (B) The flux of 4 kDa or 140 kDa FITC-dextran across confluent GPNT cell monolayers was measured when ICAM1 was crosslinked (XL) or not (NT). In each case, the FITC-dextran flux was linear over 120 minutes. The values shown are mean permeability changes that occurred over the initial linear 50-minute period following crosslinking in three independent experiments. (C,D) Confluent GPNT cells were serum starved and ICAM1 crosslinked (XL). At the indicated times cells were lysed and subjected to immunoprecipitation of VEC (C) or ␥-catenin (D). Immunoprecipitates were then analyzed by immunoblotting using antibodies against phosphorylated tyrosine, ␣-, ␤-, ␥-catenins or VEC. Similar results were achieved when the order of the proteins for immunoprecipitates and immunoblots was inverted (data not shown). In all blots the position of size markers (in kDa) is indicated on the left. (E) GPNT cells were nucleofected with wt or Y to F mutants of pEGFP-NЈ-VEC as described in Fig. 3. They were then incubated with antigen-specific T cells, which were allowed to adhere and migrate for 1-4 h. Subsequently time-lapsed microscopy was performed over a period of 5-10 minutes to determine the fraction of T cells migrating in the paracellular area of the EC. Results are the mean ± s.e.m. of six replicates from at least three independent experiments. Significant differences were determined by Student's t-test (*P<0.05, **P=0.005, ***P<0.0001).
Wild-type Chinese hamster ovary (CHO) cells and stable transfectants expressing human ICAM1 (CHO-ICAM1; a kind gift from Jeremy Pearson, King's College London, UK) were cultured in DMEM/F12 supplemented with 10% FCS.
Lymphocyte-EC co-cultures were carried out with peripheral lymph node cells (PLNCs) isolated from Lewis rats (Harlan Olac) as previously described (Adamson et al., 1999). For biochemical analysis of the ECs, PLNCs were washed and fixed using 3.7% formaldehyde before being added to the ECs, thus avoiding analysis of lymphocyte proteins after cell lysis. For adhesion assays, PLNCs were fluorescently labeled with 1 M calcein-AM (Molecular Probes) before addition to EC monolayers (approximately ten labeled PLNCs per EC). Co-cultures were left at 37°C for 60 minutes and adherent T cells quantified in a fluorescent plate reader. Migration assays were performed using a myelin basic protein (MBP) Ag-specific T-cell line and timelapse video microscopy as previously reported (Adamson et al., 1999). Migration was considered paracellular when it clearly occurred in defined areas of the periphery of the EC that appeared bright by phase-contrast microscopy. Only T cells that had completed transmigration were counted as migrated cells.

Immunofluorescence
GPNT cells were fixed in 3.7% formaldehyde and extracted with acetone (-20°C). For the detection of phosphorylated tyrosine the monoclonal antibody 4G10 (Upstate) was biotinylated and used at 2 g/ml. Cells were then stained using streptavidin-coupled Texas Red (GE Healthcare) and phalloidin conjugated to Oregon Green (Molecular Probes; 1:50). VEC was revealed using a polyclonal goat antibody (sc-6458, Santa Cruz). Cells were mounted using Moviol 4-88 and analyzed by standard epifluorescent or confocal microscopy (Turowski et al., 1999).

Plasmids, mutagenesis and transfections
Mouse VEC was mutagenized using the Quickchange mutagenesis (Stratagene). Subsequently, wt and mutant VEC were cloned into pEGFP-N1 (BD Biosciences) to result in expression of the open reading frame of mouse VEC (aa 1-784), followed by the linker sequence LVDPPVAT and the entire sequence of EGFP. Resulting plasmids (pEGFP-NЈ-VEC) were verified by DNA sequencing and purified using endotoxin-free preparation methods (Qiagen). Subsequently, GPNT cells or mouse VEC-null endothelioma cells were nucleofected using 20 g plasmid per 2ϫ10 6 cells according to the manufacturer's instructions (Amaxa). CHO cells were transfected using Lipofectamine 2000 (Invitrogen).

P labeling and phosphoamino acid and phosphopeptide analysis
Transfected CHO-ICAM1 cells were labeled with 32 P (PBS 13; GE Healthcare; 20 MBq/ml) in phosphate-free DMEM for 3 hours, ICAM1 crosslinked, lysed and VEC-GFP was precipitated using 5 g of a polyclonal anti-GFP antibody (Abcam 290). Immunoprecipitates were resolved by SDS-PAGE and transferred to nitrocellulose or PVDF. Radiolabelled VEC was excised and subjected to acid hydrolysis for phosphoamino acid analysis or to digestion with trypsin using standard procedures (Boyle et al., 1991). Amino acids or phosphopeptides were subsequently resolved on cellulose TLC plates by electrophoresis and chromatography and detected by autoradiography.

Computer model of the VEC cytoplasmic domain
A homology-based model was obtained using the coordinates of E-cadherin-␤catenin complex (pdb: 1i7x, 1i7w). Cytoplamic domains of VEC from various species and E-cadherin were aligned using ClustalW. The optimal alignment was used by PyMol to build the VEC model.