Real-Time Imaging of Vascular Endothelial-Cadherin During Leukocyte Transmigration Across Endothelium¹ Free

Vascular endothelial-cadherin (VE-cadherin)³ is expressed at the adherens junction of endothelium (1), and contributes to barrier function of the monolayer to macromolecules (2, 3, 4, 5, 6, 7). VE-cadherin has also been thought to play a role in leukocyte transmigration; however, these data have been conflicting. It was originally reported that polymorphonuclear cell (PMN) transmigration across endothelium resulted in dissociation and loss of the VE-cadherin complex as assessed by immunofluorescence microscopy and protein gel electrophoresis (8, 9). Thus, when PMN interacted with cytokine-activated endothelial cells in vitro under static conditions and monolayers were subsequently fixed and stained with mAb against junctional components, reduced levels of VE-cadherin and the associated proteins β-catenin and plakoglobin were observed. Adhesion, not transmigration, was reported as sufficient to trigger this disruption. In parallel, when such monolayers were lysed and subjected to protein gel electrophoresis, VE-cadherin and associated components were shown to be partly degraded. Other experiments showed that mAb against VE-cadherin resulted in greater PMN transmigration in vivo and in vitro (6, 7, 10, 11).

These results gave rise to a model where VE-cadherin acts as a gatekeeper for passage of macromolecules and of inflammatory leukocytes. It was proposed that leukocyte adhesion triggered a signal that led to VE-cadherin complex disassociation, allowing leukocytes to migrate through the gap so formed (8, 9). This hypothesis was brought into question by subsequent reports. The degradation of β-catenin and plakoglobin during PMN adhesion/transmigration as assessed by protein gel electrophoresis or immunofluorescence was shown to be a postpreparation artifact. However, the role of VE-cadherin was not addressed (12). PMN contain substantial amounts and varieties of proteases (13) and activated PMN or purified PMN elastase have been reported to cleave VE-cadherin (14). Thus, the protocols to stop transmigration and then observe the monolayer were incapable of completely inactivating PMN proteases. Accordingly, our group examined transendothelial migration of monocytes and a monocytic leukocyte tumor cell line, U937, both of which contain fewer proteases than PMN and lack the serine protease PMN elastase (15). In this study, it was reported that focal and transient loss of VE-cadherin was associated with transmigration, mAb that inhibited transmigration (but not leukocyte adhesion) blocked this loss, and that gaps in VE-cadherin may reseal after transmigration was complete. However, like earlier studies, postfixation degradation of junctional proteins remained an issue, and it was not possible to determine the exact sequence of events or kinetics during transmigration in this end point-type assay.

Alternative hypotheses of transmigration were put forward by others. It was proposed that endothelial cells have preexisting gaps in both their adherens and tight junctions at corners where three endothelial cells meet, and that PMN preferentially transmigrate in this region (16, 17). In this model, PMN transmigration did not require disruption or loss of junctional components. However, these studies also relied on postfixation analysis, although the authors used an improved fixation protocol. In another model it was reported that chemoattractant-induced PMN transmigration occurred predominantly through the endothelial cell body, rather than at junctions between adjacent endothelial cells, as assessed by fixation and electron microscopy (18). In this model also, no disruption of junctional components was expected.

To resolve these conflicting data, we devised a system using VE-cadherin linked to an enhanced GFP (EGFP) tag. This cDNA construct was placed in a recombinant adenoviral (AdV) expression vector, such that endothelial cells were transduced at high efficiency (19). From similar experiments with E-cadherin in epithelial cells (20), we reasoned that this method would allow direct visualization of VE-cadherin during leukocyte transmigration and thus avoid postpreparation artifacts because the entire process is observed in real time in physiological fluid shear stress conditions. Using this approach, we show that VE-cadherin fused to GFP mimics the native molecule, and we document transient and reversible changes in VE-cadherin localization during leukocyte transmigration. In this in vitro flow system, PMN and monocyte transmigration occurs at small preexisting gaps in VEcadGFP or requires de novo gap formation in this component, but is always paracellular. De novo gaps are only formed during the act of transmigration, and are not triggered by leukocyte-endothelial adhesion. Lastly, we present evidence that gap formation may be due to physical displacement of VEcadGFP in the lateral plane of the adherens junction, suggesting that proteolysis is not necessary for leukocyte transmigration.

HUVECs were isolated and cultured as previously described (15). Human PMN (>95% pure) were isolated from whole blood obtained from healthy volunteers by venous puncture as previously described (21). PBMC were isolated by centrifugation on Ficoll-Hypaque (Lymphocyte separation medium; Organon Teknika, Durham, NC). Contaminating RBC were eliminated by hypotonic lysis. Instead of purified monocytes, the PBMC fraction was used in transmigration assays for ease of preparation. Typically, this procedure yielded ∼30% monocytes. The other major population in PBMC preparations is lymphocytes, which do not transmigrate in this flow system (15).

TEA 1/31 (murine anti-human VE-cadherin, IgG1; Immunotech, Westbrook, ME) was conjugated to Alexa 538 (Molecular Probes, Eugene, OR) for immunofluorescence studies. Anti β-catenin mAb (RDI-BCATENIN, murine, IgG1) was purchased from Research Diagnostics (Flanders, NJ); anti-plakoglobin (PG5.1, murine, IgG2b) was purchased from BioDesign (Carmel, NY). An anti-junction adhesion molecule (JAM) polyclonal rabbit antiserum was commercially produced to order, to a peptide corresponding to the carboxyl-terminal 19 aa of human JAM (Primm Laboratories, Cambridge, MA).

Human VE-cadherin cDNA in a Bluescript II vector was obtained as a gift from Dr. Shintaro Suzuki (Aichi Human Service Center, Aichi, Japan). Site-directed mutagenesis (Transformer kit; Clontech Laboratories, Palo Alto, CA) was used to repair a frame shift mutation in the cytoplasmic tail, and then to replace the terminal stop codon with a single base deletion. An EagI site immediately downstream was used to fuse this construct with EGFP (pEGFP-N2 plasmid; Clontech). The resulting construct encoded full-length VE-cadherin (737 aa in the mature protein), a linker of 6 aa, and a carboxyl-terminal tag of EGFP (239 aa) attached to the cytoplasmic tail (Fig. 1,A). Constructs were confirmed by DNA sequencing. Native VE-cadherin migrates at 140 kDa in SDS-PAGE. The fusion protein was anticipated to migrate at ∼170 kDa with the added mass of the EGFP tag, as confirmed by gel electrophoresis (Fig. 1 C).

Our group has previously used AdV vectors to transduce adhesion molecules such as VCAM-1 and E-selectin into human endothelium (19, 22). Accordingly, the VEcadGFP cDNA was transferred to an AdV expression vector, and high titer virus stocks were produced (23). These stocks were titrated for expression in HUVEC as follows: HUVEC were plated at subculture 2 at a ratio of 1:4 on gelatin-coated plastic 12-well plates. Twenty-four hours later, cells were infected with varying doses of AdV vector, and cultured for another 5 days. Single cell suspensions of HUVEC were then prepared from the 12-well plates using nonenzymatic cell dissociation buffer (Life Technologies, Grand Island, NY) and analyzed by flow cytometry for the EGFP tag. A dose of virus was chosen that routinely caused 60–80% of endothelial cells to express the VEcadGFP at 5 days of infection. Flow cytometric analysis showed that VEcadGFP AdV infection resulted in an increase of expression of the VE-cadherin epitope by 50%. Monolayers were morphologically identical with sham-infected endothelial cells by phase contrast microscopy and by modified Wright-Giemsa stain (Hema 3 Stain; Fisher Scientific, Pittsburgh, PA), and AdV infection did not result in endothelial activation as evidenced by lack of induction of E-selectin, VCAM-1, or up-regulation of ICAM-1 (data not shown and Ref. 19).

HUVEC were transduced with VEcadGFP as described above, and cultured for 5 days. The monolayers were cell surface biotinylated (Biotinylation kit; Amersham, Arlington Heights, IL) and lysed as previously described (9). Equal aliquots of lysate were immunoprecipitated with mAb against VE-cadherin, β-catenin, plakoglobin, or an anti-JAM polyclonal antiserum. Samples were resolved on 8% SDS-PAGE, transferred to nitrocellulose, and probed with streptavidin-horseradish peroxidase. Bands were detected using ECL (Amersham, Piscataway, NJ) and autofluorography.

Assays were conducted using a modification of previously published techniques (24). Briefly, transwell inserts (0.4-μm pore size, 6.5-mm diameter; Costar, Cambridge, MA) were coated with 0.1% gelatin, and HUVEC were plated at a concentration of 2.5 × 10⁴ cells/well and infected the next day with a dose of virus shown to cause 60–80% expression by flow cytometry. Four days later, fluorescein-conjugated dextran (70 kDa molecular mass, anionic; Molecular Probes) was added at 500 μg/ml in HBSS without phenol red, and the bottom chamber was replaced with HBSS. After 1 h at 37°C, the insert was removed, and the amount of fluorescence in the bottom chamber was measured using a fluorescence plate reader (CytoFluor II; PerSeptive Biosystems, Cambridge, MA). A positive control of thrombin at 2 U/ml was used to confirm normal response of HUVEC to this agonist (data not shown).

Costar transwells (3-μm pore size, 6.5-mm diameter) were plated and infected as for permeability assays above. After culture for 3–4 days, medium in upper and lower chambers was replaced with TNF-α (Genzyme, Cambridge, MA) at 25 ng/ml for 4 h. PMN prepared from peripheral blood of normal volunteers were labeled with Cell Tracker Green (Molecular Probes) at 0.1 μM, and added to the upper chamber at 2.5 × 10⁵ per well in M199 medium. PMN were allowed to transmigrate at 37°C for varying periods of time, and then the upper chamber was removed to stop the assay. Transmigrated cells in the bottom chamber were counted using a fluorescence plate reader.

HUVEC (passage 2) were plated at 25% confluence on fibronectin-coated 25-mm glass coverslips and infected 24 h later with VEcadGFP AdV vector. After 5 days in culture, confluent HUVEC monolayers were activated with human rTNF-α at 25 ng/ml for 4–6 h. Freshly isolated PMN were resuspended at 0.5 × 10⁶ cells/ml and drawn across HUVEC monolayers in a parallel plate flow chamber as previously described (15). Leukocytes were allowed to interact with the monolayer, and fields were counted using a ×40 phase lens at 3, 6, 9, and 12 min. Transmigrated PMN were distinguished from those interacting with the apical surface by their phase-dark morphology. Percent transmigrated = transmigrated cells/total interacting cells × 100.

HUVEC (passage 2) were plated at 25% confluence on fibronectin-coated 25-mm glass coverslips and infected 24 h later with VEcadGFP AdV vector. After 5 days in culture, confluent HUVEC monolayers were activated with human rTNF-α at 25 ng/ml for 4–6 h. Freshly isolated PMN or PBMC were labeled for 10 min at 37°C in M199 medium, using Cell Tracker Orange at a 0.1 μM concentration (Molecular Probes). Labeled leukocytes were resuspended at 0.5 × 10⁶ cells/ml and were drawn across HUVEC monolayers in a parallel plate flow chamber as previously described (15). Leukocytes were allowed to roll and adhere to the monolayer at 20–25°C, which is permissive for leukocyte rolling and arrest but not transmigration. After sufficient leukocytes had accumulated on the apical surface, the apparatus was warmed to 37°C to permit transmigration. Images were collected using a Bio-Rad (Hercules, CA) confocal apparatus (MRC 1024, Kr/Ar), because its laser illumination and wide aperture objectives provided good sensitivity along with computer-aided image acquisition. To gather information from a broad focal plane, the laser iris was opened to maximum. Using this technique, red fluorescent leukocytes were visible both above the monolayer as well as below, indicating that the focal plane was broad enough to visualize VEcadGFP in the entire thickness of the HUVEC monolayer. Every 15 s, sequential red channel, green channel, and differential interference contrast (DIC) images were recorded digitally, typically for 20 min. Images were analyzed and processed using Confocal Assistant 4.02 shareware, Todd Clark Brelje, Adobe Photoshop 5.5, and Adobe ImageReady 2.0 (Adobe Systems, San Jose, CA). Images are representative of multiple experiments on different days for both PMN and monocytes.

Two-color fluorescence microscopy showed that VEcadGFP localized to endothelial intercellular junctions (Fig. 1 B, A–C), similar to endogenous VE-cadherin in control monolayers (D). Note that VEcadGFP was also detected intracellularly (A), whereas Ab staining of these nonpermeabilized monolayers detected only surface Ag (B). Distribution of other junctional proteins including platelet endothelial cell adhesion molecule, JAM, β-catenin, and F-actin was similar in VEcadGFP AdV vector-transduced and control monolayers (data not shown).

The cytoplasmic domain of wild-type VE-cadherin forms a complex with cytosolic proteins α- and β-catenin, and plakoglobin in confluent vascular endothelial cell monolayers (25). To determine whether attachment of a cytoplasmic EGFP tag alters this complex, confluent HUVEC monolayers were cell surface biotinylated, lysed, and then subjected to immunoprecipitation and SDS-PAGE. A mAb against the extracellular domain of VE-cadherin (TEA 1/31) immunoprecipitated a biotinylated species of 140 kDa in uninfected HUVEC (Fig. 1,C, white arrowhead). VEcadGFP transduction resulted in a new species of ∼170 kDa, due to the 27-kDa EGFP tag (Fig. 1 C, black arrowhead). Immunoprecipitation with β-catenin or plakoglobin caused coprecipitation of both VE-cadherin and VEcadGFP as detected by surface biotinylation, indicating that both proteins associated with these catenins. A negative control shows that a polyclonal antiserum to another junctional protein, JAM, did not associate with β-catenin, plakoglobin, VE-cadherin, or VEcadGFP.

To determine whether VEcadGFP contributed to paracellular permeability similar to wild-type VE-cadherin (2), HUVEC monolayers were transduced with different viral constructs. Fig. 1 D shows that overexpression of either VEcadGFP or wild-type VE-cadherin in AdV vector constructs resulted in a drop in monolayer permeability to fluorescein-dextran, as compared with control uninfected HUVEC or HUVEC transduced with control EGFP AdV vector. Thus, infection with VEcadGFP AdV vector significantly enhanced HUVEC barrier function.

HUVEC were grown on transwell inserts (for the static assay) or glass coverslips (for the flow assay) and infected with VEcadGFP. They were then induced with TNF-α, and PMN transmigration was measured as described in Materials and Methods. Fig. 1, E and F, shows that VEcadGFP infection did not significantly alter transmigration over a range of time points as compared with uninfected HUVEC.

Using the automated microscopy system described in Materials and Methods, we followed PMN as they transmigrated through a 4- to 6-h TNF-α-activated HUVEC monolayer, collecting images every 15 s, both in two-color fluorescence and in DIC. Analysis of these paired images from multiple experiments showed that in many cases, adherent PMN approached a continuous junction of VEcadGFP, a gap was formed in the wall, and then the PMN migrated through the resulting gap (Fig. 2).⁴ In other cases, PMN migrated through small preexisting gaps in VEcadGFP (Fig. 4,B). Monocytes were observed to behave similarly (Fig. 3). From these experiments it was not possible to determine whether these gaps in VEcadGFP were populated with endogenous wild-type VE-cadherin. However, even for migration through preexisting gaps, leukocyte passage typically coincided with a significant widening of the gap. Typical gap dimensions during transmigration events were 4–6 μm in size. In a few cases, two different leukocytes consecutively transmigrated, one behind the other, through the same gap, where the second leukocyte migrated before the pore created by the first had completely closed. Of a total of 52 transmigration events viewed in entirety, PMN transmigration was associated with de novo gap formation in 19 instances; in the remaining 33 events, PMN migrated through a preexisting gap in VEcadGFP. Of 32 monocyte transmigration events, 18 were associated with de novo gap formation; the rest migrated through preexisting VEcadGFP gaps. It may be that overexpression of VEcadGFP alters this ratio; however, it is reasonable to conclude that at least part of the time leukocyte transmigration is associated with new gap formation.

Others have reported that PMN transmigration occurs primarily at corners of three or more endothelial cells, where they observed small gaps in cadherin staining (16). In the current experimental system, tricellular gaps were not consistently observed in either endogenous wild-type VE-cadherin or in VEcadGFP. Instead, there were occasional small gaps in junctional staining, which occurred both at multicellular corners, and at bicellular junctions. Furthermore, both PMN and monocytes migrated through multicellular corners (53 and 26%, respectively; Fig. 4,A) and also through bicellular junctions (47 and 74%; Figs. 2, 3, and 5). When migration occurred at multicellular corners, it occurred at preexisting gaps as well as through fresh gap formation. In the current in vitro system, no evidence of transcellular transmigration was observed. At the resolution provided by the experimental apparatus, all transmigration events occurred at lateral junctions, through gaps in VEcadGFP.

After leukocyte transmigration through a gap, VEcadGFP was found to fill in and thus reseal the gap formed. This resealing process took 5 min on the average for both PMN and monocytes, and ranged from 1:45 to 11 min. Interestingly, even gaps that were present before approach of a leukocyte were capable of sealing after transmigration (Fig. 4 B), suggesting that endothelial junctional proteins are dynamic and not fixed.

In several cases of transmigration, it was possible to monitor VE-cadherin gap formation and closing to a higher degree of clarity (Fig. 5). Here it was observed that as a gap was formed in VEcadGFP, there appeared a simultaneous clustering of green fluorescent label in the junction proximal to the gap. After transmigration, the bunched up VEcadGFP gradually diffused laterally back into the gap, analogous to the opening and closing of a curtain. Gap formation was not due to vertical displacement of VEcadGFP, because the focal plane was broad enough to visualize the entire monolayer.

We generated a VEcadGFP fusion protein, and expressed it in HUVEC using an AdV expression vector. This fusion protein localized at lateral junctions, along with endogenous wild-type VE-cadherin, and associated with β-catenin and plakoglobin. Overexpression of VEcadGFP, like overexpression of VE-cadherin, caused a decrease in endothelial permeability and, therefore, enhanced barrier function. VEcadGFP did not alter the rate or extent of PMN transmigration assessed under flow or static conditions. Thus by biochemical and functional analyses, VEcadGFP mimicked the behavior of native endothelial VE-cadherin. This construct was used in real-time fluorescent imaging to extend our previous studies and to analyze whether leukocyte transmigration led to alterations in VE-cadherin distribution under flow conditions.

Direct real-time observation of VEcadGFP during leukocyte transmigration under flow yielded several interesting findings. We report exclusively paracellular leukocyte transmigration, migration through tricellular corners and bicellular junctions, and migration through both preexisting gaps and through de novo gap formation. Because the current experimental system uses real-time observation of leukocytes and endothelial cells under flow, it avoids the limitations of previous studies such as static assays and postfixation artifacts, and yet is sensitive to small localized and temporary changes in protein localization that may not be appreciated by bulk biochemical analysis.

Although widespread proteolysis of VE-cadherin as a result of PMN adhesion has been reported (8, 9), we have no evidence that this occurred with real-time imaging. Instead, the current system clearly documents formation of gaps (or widening of preexisting gaps) at a distal step, immediately before transmigration. Gap formation or widening is frequently accompanied by a clustering effect of VEcadGFP at adjacent regions of lateral junctions. After transmigration, this clustered material gradually diffuses back to refill the junction. These observations lead us to hypothesize that gap formation or widening allows the endothelium to accommodate a transmigrating leukocyte. Leukocyte-endothelial interaction and cross-linking of endothelial surface proteins have been shown to induce intracellular signals in endothelium (26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38). Gap formation or widening may be a result of signaling from such an interaction, or in response to a soluble factor released by leukocytes. Alternatively, it is possible that gap formation may occur as a passive process. From studies of a related protein, E-cadherin, lateral diffusion of cadherin molecules is thought to occur in a time scale of minutes (20). It is possible that as a leukocyte approaches a lateral junction it may extend a pseudopod, which simply forces aside dimers of VE-cadherin that bridge the junction. If there is already a small gap in VE-cadherin, the leukocyte may use this means to widen it sufficiently to permit its passage. The leukocyte then transmigrates through the resulting gap, maintaining it by physically interrupting the junction. Throughout leukocyte passage, we observed that VEcadGFP was in close apposition to the leukocyte perimeter. After transmigration is completed, once more there is no obstruction to the lateral movement of VE-cadherin, and the displaced material returns to close the gap. Like other members of the cadherin family, VE-cadherin is likely to derive adhesive strength by cooperative binding and tethering to the actin cytoskeleton as recently suggested by atomic force microscopy experiments (39). To allow free lateral movement to accommodate leukocyte transmigration, VE-cadherin may need to undergo decoupling from the cytoskeleton. This may occur as an endothelial response to leukocyte signaling, or alternatively, leukocytes may be able to recognize portions of the endothelial junction that are already undergoing remodeling in response to migration, shear stress, or other stimuli. Alternatively, it is possible that VE-cadherin movement is directed by rearrangement of the cytoskeleton. These hypotheses will be tested in future studies.

In the current study, only the EGFP tag at the carboxyl-terminus of VEcadGFP is detected, thus it remains possible that partial proteolysis of other portions of the construct may have occurred while the fluorescent marker is pushed aside. However, physical displacement is at least partly responsible for gap formation. Gap formation and closing were also observed where a transduced endothelial cell lay next to another expressing no detectable VEcadGFP, and where two endothelial cells both expressed much lower levels (data not shown). Although these examples display less well, they suggest along with the results shown in Fig. 1, E and F, that the effects reported here are not dependent on overexpression of VEcadGFP.

Our group previously reported gaps in VE-cadherin staining at sites where leukocytes were adherent, presumably about to transmigrate using fixation and subsequent immunofluorescent microscopy techniques (15). We concluded that a correlation existed between the position and frequency of gaps and leukocyte transmigration. Thus, leukocytes capable of robust transmigration caused large numbers of gaps in VE-cadherin staining, whereas leukocytes incapable of transmigration did not. However, it was not possible to definitively conclude that a leukocyte was actually going to transmigrate through a particular gap, or even that the gap occurred before fixation and staining. In the current study we confirm and extend our previous findings. First, we document changes in VEcadGFP distribution as a leukocyte is actively transmigrating under physiologic flow conditions. Second, these gaps are in close apposition to the migrating leukocyte and imply a tight fit of the leukocyte within the endothelial junction. Third, the current study indicates a specific order of events as the leukocyte prepares to transmigrate, and kinetics for gap formation and subsequent closure. Finally, leukocyte movement to the lateral junctions is not in itself sufficient to trigger gap formation. The corollary is that without exception, leukocyte migration required either de novo gap formation or widening of a preexisting gap in VE-cadherin.

PMN transmigration of in vitro cultured HUVEC has been reported to occur predominantly at tricellular corners, where gaps were reported in staining of both VE-cadherin and occludin (16, 17, 38). We did not observe consistent gaps in junctional VEcadGFP (or in VE-cadherin staining) at tricellular corners in the current system. However, we did find that PMN migrated ∼50% of the time through tricellular corners through both de novo and preexisting gaps. Interestingly, even preexisting gaps were enlarged by the leukocyte during the actual process of transmigration. In contrast to PMN, monocytes preferentially migrated through bicellular junctions. We infer that different leukocyte types may exhibit site preferences for transmigration, and speculate that different vascular beds may behave likewise. The somewhat lower level of PMN transmigration at tricellular corners in the present system may reflect differences in junctional structure, as Burns et al. (16) use medium that enhance occludin expression and the number of tight junctional strands formed. Our current model system did not address tight junctional components; however, previous reports have found that tight junction “strands” can be rapidly increased in small vessels in vivo in response to certain stimuli (40), suggesting that these structures are dynamic, like adherens junctions. Venous endothelium has been reported to express less occludin than arterial endothelium (41); however, we are not aware of studies addressing the expression of tight junctional components in postcapillary venules. We speculate that tight junctional components may also dissociate or diffuse out of the way to accommodate transmigrating leukocytes, and that these gaps may need to occur in concert with VE-cadherin to coordinate formation of a passage through the entire junctional complex. These questions will require development of reagents allowing real-time visualization of multiple junctional components, rather than relying on postfixation end point-type assays.

In the present study, monocytes preferred bicellular junctions, whereas PMN migrated through both bi- and tricellular junctions. Less strikingly, monocytes showed a moderate preference for migration through de novo gaps, whereas PMN showed a moderate preference for preexisting gaps. These results suggest potential differences in mechanisms of transmigration between the two cell types, possibly due to differential use of adhesion molecules or chemokines.

VE-cadherin has been proposed to function as a gatekeeper at endothelial junctions. Thus, VE-cadherin mAbs have been well documented to enhance leukocyte transmigration while increasing monolayer permeability to macromolecules (decreased barrier function) (6, 7, 10, 11). We were surprised to find that overexpression of VEcadGFP (or VE-cadherin) in HUVEC did not alter PMN transmigration in static or flow conditions, although barrier function was increased. From these results it appears that improved barrier function to macromolecules does not necessarily equate with reduced transmigration, suggesting a more complicated mechanism beyond simply a generalized “gatekeeper” function for VE-cadherin.

In conclusion, we present data showing real-time changes in VE-cadherin as a result of paracellular transmigration of PMN and monocytes across cytokine-treated endothelial monolayers. The current system avoids postfixation artifacts and difficulties in visualization of small localized effects in VE-cadherin that may not be observed by protein chemistry techniques used in previous studies. Although it is possible that the movement of VEcadGFP does not accurately reflect that of endogenous wild-type VE-cadherin, this is unlikely because no difference in localization or function was observed. The technique described here may serve as a useful model for leukocyte transmigration studies and may be used in the future to track transmigration in different endothelial beds, and to follow trafficking of VE-cadherin in response to other stimuli, such as thrombin or shear stress. Lastly, it may become possible to track multiple components of the lateral junction simultaneously, to gain an understanding of the dynamic response of the entire complex and its role in endothelial behavior and responses.

We thank Drs. Xinzhong Wang and Mason Freeman (Massachusetts General Hospital Boston, MA) for the AdV vector expression system, Michelle Lowe for technical help with confocal microscopy, Kay Case for cell culture support, Andrew J. Connolly for helpful discussions, and Brigham & Women’s Hospital labor and delivery and South Shore Hospital Birthing Unit for umbilical cords.