Intercellular Adhesion Molecule-1 Enrichment near Tricellular Endothelial Junctions Is Preferentially Associated with Leukocyte Transmigration and Signals for Reorganization of These Junctions To Accommodate Leukocyte Passage

Leukocyte extravasation is a complex multistep process that requires overlapping function of both leukocytes and endothelial cells (ECs). Each step has been extensively studied and is elegantly summarized in recent reviews (1–4). To effectively exit the blood vessels, leukocytes initiate rolling contacts with endothelium (5), followed by firm adhesion (6). Leukocyte rolling is primarily mediated by interactions between members of the selectin family and the glycosylated molecules expressed on the leukocyte surface (7). Leukocyte adhesion is mainly mediated by interactions between members of cell adhesion molecule family, such as ICAM-1, and their counterreceptors, the α_Lβ₂ integrin (LFA-1) and the α_Mβ₂ integrin (MAC-1) (8). Adhered leukocytes crawl along the EC surface toward locations where transendothelial migration (TEM) occurs. Leukocyte crawling was first identified in isolated cell systems (9) and later confirmed in situ (10). Interactions between ICAM-1 and β₂ integrins are essential for leukocyte crawling (9, 10). The finding that a leukocyte’s inability to crawl decreases TEM suggests that crawling allows leukocytes to get to specific locations that accommodate TEM and further suggests that some regions of the endothelium are better equipped to support leukocyte passage than others. Although transcellular transmigration has been described in vitro (11) and in vivo (12), most agree that leukocyte TEM primarily occurs via EC junctions (13–15). Numerous proteins associated with EC junctions, including VE-cadherin, CD31, CD99, and junctional adhesion molecules-1/2/3, are implicated in mediating leukocyte TEM (2, 16, 17). Interestingly, EC surface molecules ICAM-1 and VCAM-1 cluster at regions of leukocyte TEM, suggesting a role for these molecules in TEM as well (11, 18, 19). Moreover, these molecules are localized around leukocytes that use either junctional or transcellular routes for emigration (11).

Tricellular junctions (intersection of the borders of three adjacent ECs) in particular have been suggested to serve as “portals” for leukocyte TEM, because of the discontinuity of junctional proteins, such as occludin, cadherin, and ZO-1, at these sites (20). However, this does not explain transcellular TEM or the finding that some junctional molecules, such as VE-cadherin, can undergo rearrangement, creating de novo gaps to accommodate leukocyte TEM (21). This report (21) implies that the preferred locations for leukocyte TEM need not pre-exist but could be created upon receiving an appropriate signal. Whether the signal is originated in ECs or the migrating leukocytes is also not clear.

Thus, the expression levels, distribution patterns, and the overlap of signaling induced by all these molecules can potentially create a preferred location for leukocyte TEM.

Finally, the expression of EC junctional and surface molecules varies greatly from tissue to tissue (22) and in different locations within the vascular network (23), resulting in great variations in leukocyte-EC interactions (24). Undoubtedly, the variability of junctional and surface receptors also affects the mechanisms of leukocyte recruitment. Thus, we used an intact system to explore leukocyte crawling and TEM locations in blood perfused venules in situ. Using intravital confocal microscopy combined with immunofluorescence labeling, we show that leukocytes, in their innate environment, preferentially use tricellular junctional regions as portals for emigration. Moreover, we demonstrate that ICAM-1 enrichment near EC junctions likely makes these locations preferred for leukocyte TEM. EC morphology determines the number of available portals, making venular converging regions optimally equipped to support leukocyte TEM.

Male wild-type C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME) aged 12–15 wk were used. Animals were anesthetized with sodium pentobarbital (65 mg/kg, i.p.) and euthanized by overdose (>100 mg/kg) at the completion of the protocols. When indicated, inflammation was induced by locally superfusing the tissue with fMLP (10 μM in 0.01% DMSO; Sigma-Aldrich, St. Louis, MO) or by treatment with mouse rTNF-α (0.5 μg TNF-α in 0.25 ml saline, intrascrotally; Sigma-Aldrich) 4 h prior to observations. All mice were used according to protocols approved by the University of Rochester Institutional Review Board.

Mice were prepared for intravital microscopy and maintained throughout the experiment as described elsewhere (25). Observations were made using an Olympus BX61WI microscope with an Olympus PlanF1 immersion objective (×20, 0.65 numerical aperture or ×40, 0.95 numerical aperture). All observed venules ranged from 30–80 μm in diameter. Total leukocyte adhesion and TEM (transmigrated leukocytes in the tissue) were quantified using bright-field images acquired via a charge-coupled device camera (Dage MTI CD72, DageMTI, Michigan City, IN). Leukocyte adhesion, crawling, and TEM with respect to EC junctions were quantified using confocal fluorescence images that were acquired by illuminating the tissue with a 50-mW argon laser and imaging with a Nipkow disk confocal head (CSU 10; Yokogawa Yokogawa Electric, Tokyo, Japan) attached to an intensified charge-coupled device camera (XR Mega 10; Stanford Photonics Palo Alto, CA). All images were either digitally acquired or recorded to a DVD recorder (SONY DVO100MD) at 30 frames per second (fps) for offline analysis.

Selected venules were (separately) stained for PECAM-1, VE-cadherin, and ICAM-1 as described previously (23). Briefly, using microcannulation, venules were locally perfused with Abs as follows: anti–PECAM-1 (ER-MP12 mAb conjugated to Alexa 488, 10 μg/ml, 10 min; Serotec, Oxford, U.K.); anti–VE-cadherin (BV13 mAb conjugated to Alexa 488, 30 μg/ml, 10 min; eBioscience, San Diego, CA); and anti–ICAM-1 (YN/1.7.4, 50 μg/ml, 15 min; eBioscience), followed by goat anti-rat secondary fluorescent polyclonal Ab (Alexa 488 anti-rat, 50 μg/ml, 15 min; Molecular Probes, Eugene, OR) or, alternatively, anti–ICAM-1 (YN/1.7.4 conjugated to Alexa 488, 30 μg/ml, 10 min; BioLegend, San Diego, CA).

Unless stated differently, to observe leukocyte behavior with respect to EC junctions, leukocytes were stained for CD11a (anti–CD11a-Alexa 488 [M17/4], 3 μg/mouse, i.v.; eBioscience) in addition to the anti–PECAM-1-Alexa 488 Ab, as above. Because the EC junctions are stationary and leukocytes are mobile, leukocytes can be easily resolved with respect to EC junctions despite using the same fluorophore. Although M/17.4 is able to block integrin function, as shown in Supplemental Figs. 1 and 2, M/17.4 combined with ER-MP12 Abs at the concentrations used in this work did not affect the specified leukocyte behavior. Anti–VE-cadherin (BV13; 30 μg/ml) Ab left the EC junctions intact, with no visible gaps after 2 h (Supplemental Fig. 3).

The peptides used consisted of 16 aa of penetratin (RQIKIWFQNRRMKWKK), followed by 13 C-terminal amino acids of mouse ICAM-1 (QRKIRIYKLQQAQ); this peptide was modified for mouse tissue from a previous report (18). The control peptide was an irrelevant sequence from rat rodopsin (CKPMSNFRFGENH). In all cases, selected venules were locally perfused with the relevant peptide (100 μg/ml, 15 min) prior to tissue stimulation with fMLP.

A venular convergence was defined by measuring 40 μm [an average length of venular EC (23)] in each direction from the inner convergence point for each of the inflow vessels at the convergence (Fig. 2A, top panel). This 80-μm length was used to quantify leukocyte adhesion (cells/80-μm vessel wall) in convergences (regions 1, 1*, and 2, white dotted lines; Fig. 1D) and in straight regions at least 40 μm away (designated region 3, not illustrated). All leukocytes that remained stationary or did not exceed a displacement of >8 μm (one leukocyte diameter) during 30 s were considered adhered. In all cases, regions 1 and 1* surrounded the smaller of the two inflow vessels. For TEM, leukocytes were counted in the extravascular tissue regions 1, 1*, 2 (Fig. 1D, black dotted lines), and 3 (not illustrated) adjacent to vessel lengths where the adhesion was quantified. Regions 1, 1*, and 2 for TEM were defined as follows: region 1 (between the two inflow vessels) was defined by projecting two lines perpendicular to the vessel walls. Regions 1* and 2 were defined by connecting two lines that were projected perpendicular to each vessel and that were equal in length to the diameter of the vessel they were projected from (illustrated in black). Adhesion and TEM in regions 1 and 1* were averaged and presented as region 1. In straight regions, transmigrated leukocytes were counted in the tissue within 50 μm of the vessel/80-μm length vessel segments. All TEM counts were normalized to 1000 μm². To quantify leukocyte crawling and TEM routes, leukocytes and ECs were visualized as described above; each site was observed and recorded for 30–40 min at 30 fps. The original movies were time-lapsed to 0.33 fps for offline analysis. Leukocyte crawling was defined as a displacement of at least one average leukocyte diameter (8 μm) during 30 min. Only rolling leukocytes that were observed to firmly adhere and then began to crawl were analyzed. Any crawling leukocytes that crawled out of the field of view were not included in the analysis. Crawling distances were obtained by measuring the path length of each crawling leukocyte from the moment it began to move and until it either detached from the endothelium or underwent TEM. A leukocyte was considered to be migrating via a junctionally associated route if immediately prior to TEM, it was observed to intersect with the EC junction and exhibited no measurable lateral displacement as it passed through the endothelial layer. When unclear, we assigned the transmigrating pathway to “un-assg.”

The number of junctions was quantified in convergences, and the average area of all convergences (6000 ± 187 μm², n = 11) was used as a region of interest (ROI) for comparative analysis of straight regions. Vessel diameter in each straight region was used as the width of the ROI, and the length was adjusted as needed to achieve an area of 6000 μm² (Fig. 2A, bottom panel). All dimensions were calculated from lengths measured as two-dimensional projections of the three-dimensional length; this measurement has <13% projection error (10). All images were acquired as “continuous stacks” and presented as summed Z-stack projections for measurement of EC junctional length.

To determine the level of ICAM-1 enrichment, three independent measurements on the EC surface were taken. ICAM-1 expression near tricellular junctions was measured by placing a 5-μm-diameter ROI at the edges of each EC comprising the junction as close as possible to each outlined junction. The mean intensity of the three ROIs represents the ICAM-1 intensity near that junction. These measurements were taken for all tricellular junctions in the field of view. Similarly, two measurements of ICAM-1 intensity were taken on each side of all bicellular junctions and averaged. A third measurement was taken randomly in the middle of each EC, away from all junctions. Each ROI measurement was normalized to the mean intensity of the whole EC on which it was located. Intensities that were 2 SDs higher than the mean intensity of the ROIs on the EC surface away from junctions were defined as enriched regions. Similar analyses were performed on ECs that were labeled using only the primary anti–ICAM-1 Ab (YN-1) conjugated to Alexa 488 as controls for possible clustering induced by secondary Ab. No differences were found (data not shown). For an additional control, similar analyses were performed on ECs stained for the unrelated adhesion molecule P-selectin, using primary followed by secondary Abs (similarly to principal approach for labeling of ICAM-1). This showed no P-selectin enrichment in the tricellular junctional regions, thus further supporting our conclusion that the observed enrichment of ICAM-1 was not an artifact (Fig. 6A).

Statistical significance was assessed by one-way ANOVA with Newman-Keuls Multiple Comparison Test using GraphPad Prism (version 4.0; GraphPad, San Diego, CA). Statistical significance was set at p < 0.05.

We have previously demonstrated increased leukocyte adhesion and TEM in converging regions of TNF-α–activated venules (26). Similarly, in the current study, fMLP-induced leukocyte adhesion and TEM were significantly increased in venular convergences (region 1) compared with the neighboring straight regions (region 3) (Fig. 1A).

The number of adhered leukocytes in region 1 of the convergence was significantly higher compared with regions 2 and 3 (7.0 ± 0.4 versus 5.1 ± 0.4 and 4.3 ± 0.5 leukocytes/80-μm vessel length, respectively; p < 0.01) (Fig. 1B). Likewise, leukocyte TEM in region 1 was significantly higher compared with regions 2 and 3 (29.75 ± 3.1 versus 12.6 ± 1.3 and 11.3 ± 1.3 leukocytes/10,000 μm² tissue, respectively; p < 0.001) (Fig. 1C). Leukocyte TEM localized to the convergence but not the straight venular region is shown in Fig. 1E, Supplemental Movie 1, and Fig. 1F, which displays the origins and paths of transmigrating leukocytes. In this particular vessel, leukocyte TEM is mostly localized to one side of the vessel wall (the upper part of the field of view). In different vessels, TEM occurs at specific locations and can primarily occur from either side of the vessel wall or both sides simultaneously and is likely a function of the distribution of the “active portals” for TEM. In control (unstimulated) tissue, the number of adhered leukocytes was one to two leukocytes per 80-μm vessel length and was not significantly different in all regions (data not shown). Similarly, the number of extravasated leukocytes ranged between one and four leukocytes per 10,000 μm² and was not different in all regions (data not shown). In comparison with the findings for adhesion and TEM, leukocyte rolling fluxes (6.6 ± 0.5 versus 5.2 ± 0.7 and 5.8 ± 0.6 leukocytes/40 s for regions 1, 2, and 3, respectively) (Fig. 1A) and rolling velocities (7.5 ± 0.8 versus 8.2 ± 0.6 and 9.1 ± 0.8 μm/s) were not different in converging and straight venular regions. These data confirm that proinflammatory stimuli dramatically increase the number of firmly adhered and transmigrating leukocytes, and importantly, these data suggest that vessel architecture might play a role in leukocyte recruitment in situ.

We have previously explored the differences in flow profiles in straight and converging regions (25–27) and concluded that flow is unlikely to be the main cause for increased leukocyte TEM in venular convergences. EC morphology varies greatly in different vessels (23); thus, we hypothesized that increased leukocyte TEM in venular convergences might result from local differences in EC morphology (e.g., size and shape), leading to different EC alignment and junctional arrangement compared with straight regions. Confirming our hypothesis, we found that in 30- to 60-μm-diameter venules, ECs in the convergence regions were significantly shorter (38.2 ± 0.8 versus 52.3 ± 1.1 μm; p < 0.01) and had significantly smaller surface area (593.8 ± 23.6 versus 879.9 ± 18.2 μm²; p < 0.01) compared with straight regions (Table I). The aspect ratio (EC width/length, 1 = circle) in convergences was also significantly higher (0.46 ± 0.02 versus 0.34 ± 0.01; p < 0.001) compared with straight regions, indicating that ECs in convergences are less elongated. The alignment of ECs in these regions was significantly less ordered (17.8 ± 1.8 versus 10.6 ± 1.8, degrees, major axis relative to axial direction, p < 0.01, n = 219 cells). This irregular alignment of smaller and variously shaped ECs has the potential to change the local junctional morphology, because in convergences two, three, or four adjacent ECs can create an EC junction. Thus, we asked whether the number of tricellular EC junctions varies in straight versus converging venular regions. The number of tricellular junctions in convergences was indeed significantly higher compared with straight regions (20.7 ± 1.2 versus 12.43 ± 1.1 junctions/6000 μm², respectively), but the total junctional length per area of vessel wall was not different (623.8 ± 54.8 versus 609.0 ± 28.3 μm/6000 μm², respectively) (Fig. 2B). In contrast, arteriolar ECs, which differ from venular ECs in being significantly longer (23), maintain similar morphology in straight and converging regions (data not shown). Consequently, the number of tricellular junctions is not significantly different in straight versus converging regions (8.2 ± 0.9 versus 7.9 ± 0.7 junctions/6000 μm², respectively) (Fig. 2C). Moreover, although the total junctional length per area of vessel wall in arterioles is similar to that in venules (601.8 ± 25.2 μm in arterioles versus 623.8 ± 54.8 μm in venules at the diverging/converging regions and 616.6 ± 49.9 versus 609.0 ± 28.3 μm in straight regions, respectively), the number of tricellular junctions in arterioles was significantly lower (∼50% less) compared with venules (Fig. 2). This suggests that EC morphology can indeed affect leukocyte TEM by dictating the incidence and assembly of tricellular junctions.

An important question arising from these observed morphological differences in straight and converging regions is whether they impact leukocyte recruitment and whether they could explain increased leukocyte TEM in convergences. We first asked whether the different EC junctional arrangement in straight versus converging regions produced different distributions of adhered leukocytes with respect to EC junctions. Representative images of leukocytes adhered at the different locations are presented in Fig. 3C. Indeed, 55 ± 1.8% of firmly adhered leukocytes in convergences were associated with tricellular junctions, leaving 24 ± 3.6 and 20 ± 3.0% at bicellular and nonjunctional regions, respectively (Fig. 3A). In contrast, in straight regions, most adhered leukocytes (58 ± 2.3%) were located at bicellular junctions, and only 21 ± 2.8% were located at tricellular EC junctional regions (Fig. 3A).

To determine whether locations of adhered leukocytes were a direct consequence of EC morphology (different ratio of bi- to tricellular junctions), we randomly generated x-y position coordinates for test “leukocytes” on the mapped EC morphology. The distribution of randomly adherent “virtual” cells was not different from that for actual leukocytes, as shown in Fig. 3B. This implies that leukocyte adhesion on bi- versus tricellular junctions is a direct consequence of EC morphology.

To study the behavior of adhered leukocyte prior to TEM, leukocytes and EC junctions were fluorescently labeled with anti-CD11a and anti–PECAM-1 Abs, respectively. Confirming previous findings (12), we show that most leukocytes (92.2 ± 1.6%) (Fig. 4A) in straight regions undergo intraluminal crawling. The majority (89.6 ± 1.7%) (Fig. 4A) of adhered leukocytes in convergences also crawled, but average crawling distance was significantly less compared with straight regions (20.1 ± 1.1 versus 32.1 ± 1.5 μm) (Fig. 4B). In both regions a small subset (≤10%) of leukocytes crawled distances of up to 100 μm. Leukocyte crawling velocity in convergences was 5.4 ± 0.3 μm/min, which was significantly slower than 11.2 ± 0.5 μm/min in straight regions (velocity distribution) (Fig. 4C). Interestingly, in both straight and converging regions, leukocyte crawling distances (real-time trajectories) were not significantly different from their straight-line displacement from origin (20.1 ± 1.1 versus 16.7 ± 1.8 μm in convergences and 32.1 ± 1.5 versus 26.0 ± 2.6 μm in straight regions, respectively), suggesting that there was some directionality in the observed crawling. Furthermore, following fMLP application, most leukocytes crawled in the direction of, or perpendicular to, blood flow but not against it (Fig. 4D, 4E), unlike in control venules, where leukocyte crawling is independent of the flow direction (10, 28). Taken together, these data strongly suggest that the locations where leukocytes initially adhere are merely a station where leukocytes pause, subsequently crawling to locations where the environment is right for TEM.

Leukocytes crawl finite distances of up to ∼100 μm (Fig. 4) prior to either undergoing TEM or detaching from the vessel wall. Although only a small fraction of crawling leukocytes eventually undergo TEM, this process is significantly more efficient in convergences compared with straight regions (32.8 ± 6.1 versus 14.6 ± 6.2% TEM/adhered) (Fig. 5B). The rest of the crawling leukocytes become more rounded and detach from the endothelium. The trajectory of a representative leukocyte that crawled toward the EC junctional region and transmigrated is shown in Fig. 5D and Supplemental Movie 2.

Because it has been suggested that leukocyte TEM occurs at tricellular junctions (20), we hypothesized that the higher TEM in convergences was due to the higher number of tricellular junctions in these regions. We show that in situ, leukocyte TEM occurs primarily via a junctionally associated route in both straight and converging regions (91.3 ± 0.6 and 87.7 ± 1.7%, respectively) (Fig. 5A). In convergences, most (59.3 ± 2.1%) leukocytes used tricellular junctions to undergo TEM versus 45.8 ± 4.1% using bicellular junctions. In straight regions, equal fractions of leukocytes underwent TEM via tricellular and bicellular junctions (45.1 ± 4.1 and 45.0 ± 5.0%, respectively) (Fig. 5C), despite that significantly more adhered leukocytes were originally adhered at bicellular compared with tricellular junctions (58.2 ± 2.2 versus 28.4 ± 2.1%) (Fig. 3A). This again implies a clear preference toward tricellular junctional regions for TEM.

Consistent with the idea that TEM preferentially occurs at tricellular junctional regions, we also found that in arterioles (where both leukocyte adhesion and transmigration are rare events) the number of tricellular junctions was not significantly different in straight versus converging regions (Fig. 2B) but was significantly lower (∼50% less) compared with venules. Taken together, these data suggest that EC morphology and alignment, and in particular the number of tricellular junctions, play a key role in determining preferred regions for leukocyte transmigration.

The distribution of ICAM-1 on ECs is not uniform and under appropriate conditions undergoes redistribution to more localized regions within individual ECs (23). Furthermore, ICAM-1 clustering occurs around transmigrating leukocytes (11, 19, 29), suggesting that regions surrounding tricellular junctions could also be enriched in ICAM-1. If true, this would explain why these junctional regions are preferred locations for TEM.

We show that 50 min following fMPL application, the expression of ICAM-1 on the EC surface significantly increased (Supplemental Fig. 4A). Likewise, ICAM-1 was enriched near EC junctions (Fig. 6A, 6B). Representative images in Fig. 6A demonstrate enrichment of ICAM-1 (second panel from the top), near tricellular junctions in venules in fMLP-treated tissue (middle panel), but not under control conditions (upper panel). ICAM-1 enrichment was also observed with TNF-α treatment, which is known to directly activate ECs (third panel from the top), but not P-selectin (bottom panel). P-selectin (green) exhibits a typical punctate distribution on EC surface [as previously determined (27, 30)] but no preferential localization to the EC junctions (red). Importantly, ∼43% of tricellular junctions were associated with enriched ICAM-1 versus only 23% of bicellular junctions, directly supporting our hypothesis that ICAM-1 at least in part contributes to making the tricellular endothelial junctions a preferred location for leukocyte TEM. This enrichment occurred in both converging and straight regions (data not shown). The finding that not all tricellular junctions are enriched in ICAM-1 suggests that not all of them will equally support TEM. Supporting this, we determined that ∼40% of all crawling leukocytes that transmigrated, crossed at least one tricellular junction on their way to a transmigratory portal. The trajectory of a typical crawling leukocyte crossing two tricellular junctions is shown in Fig. 6E and Supplemental Movie 3. Furthermore, we found that leukocyte crawling was slowest across tricellular junctions (4.2 ± 0.3, straight and 4.1 ± 0.3 μm/min, converging regions) (Fig. 6C); velocity across bicellular junctions was also significantly slower than at nonjunctional regions (7.6 ± 0.6 versus 11.1 ± 0.7 μm/min in straight and 7.1 ± 0.4 versus 10.4 ± 0.6 μm/min in converging regions, respectively) (Fig. 6C). Moreover, individual leukocytes that crawled for longer distances (>40μm) and crossed at least two tricellular junctions predictably slowed down near the EC junction and sped up on the EC surface (Fig. 6D). A representative velocity trace of a leukocyte in relation to EC junctions and its crawling trajectory are shown in Fig. 6D (bottom panel). The leukocyte slowed down near tricellular junctions, consistent with evidence for ICAM-1 enrichment in these regions (Fig. 6A, 6B). Crawling velocity across EC cellular junctions was not different in straight versus converging regions (Fig. 6C), suggesting that the differences in leukocyte crawling behavior in both regions are indeed relatable to local ICAM-1 distribution.

ICAM-1 engagement leads to phosphorylation of junctional proteins, such as VE-cadherin (31), likely contributing to its redistribution away from junctions during TEM (21). We observed redistribution of VE-cadherin and formation of gaps at the locations of leukocyte TEM in situ [as described in monolayers (21)] following fMLP treatment of the tissue but not in untreated venules (Fig. 7A and representative images Fig. 7C, middle and upper panels). Importantly, gap formation was significantly attenuated by treatment with the ICAM-1 tail peptide that blocks ICAM-1 signaling (18) (Fig. 7A and representative image Fig. 7C, bottom panel), consistent with the hypothesis that high ICAM-1 density around EC junctions is important for VE-cadherin rearrangement. Approximately 70% of all observed gaps were formed in association with tricellular junctions (Fig. 7B), confirming that these are preferred TEM locations. Despite the overall lower number of gaps formed following treatment with the ICAM-1 tail peptide, the same ∼70% were still seen in association with the tricellular junctions (Fig. 7B).

We also tested the effect of ICAM-1 tail peptide on leukocyte behavior and compared this with the effect of a mAb that blocks luminal interactions of leukocytes with ICAM-1. As predicted from isolated cell systems (18), treatment with the ICAM-1 tail peptide did not significantly alter the ability of leukocytes to adhere or crawl on the endothelium (thus presumably did not affect their ability to reach the right location for TEM) but reduced leukocyte TEM by 71% (Fig. 7D), indicating that the inability to open EC junctions is the cause of impaired leukocyte TEM. The direction of crawling and leukocyte crawling velocity also remained unchanged (Fig. 7E, 7G). Interestingly, treatment with the ICAM-1 tail peptide significantly increased leukocyte crawling distances (41.4 ± 2.4 versus 28.9 ± 2.7 μm in untreated venules) (Fig. 7F), implying that reduced TEM was due to the inability of leukocytes to find the portal locations. Furthermore, one of every four leukocytes that were observed undergoing TEM following blockade of the ICAM-1 tail had difficulty in disengaging from the endothelium as it traversed the vessel wall. Ab blockade of ICAM-1, as expected (32), significantly decreased leukocyte adhesion (42%), abrogated leukocyte crawling (84%), and resulted in a 68% decrease in leukocyte TEM (Fig. 7D). Interestingly, the residual crawling was significantly slower than that in untreated venules (6.5 ± 0.9 versus 10.5 ± 0.8 μm/min) (Fig. 7G), and the direction of crawling was more random (same fraction in all directions) (Fig. 7E), confirming a role for ICAM-1 in directing leukocyte crawling. Treatment with either a nonspecific Ab or a control peptide had no significant effect on leukocyte adhesion and TEM and did not prevent the formation of gaps in VE-cadherin (data not shown). Taken together, these findings suggest a dual role for ICAM-1 in mediating leukocyte TEM: signaling to leukocytes the location of a portal for TEM, and preparation of EC junctions to accommodate leukocyte passage.

Leukocyte TEM occurs primarily at EC junctions (12, 21, 33), but recent studies suggest that changes in EC morphology, and consequently junctional arrangement, or disruption of β₂ integrin-ICAM-1 interactions can lead to transcellular migration (12, 18). Thus, in this study we asked whether differences in EC morphology in straight and converging venular regions affect leukocyte trafficking and TEM in situ. We showed that leukocytes, in their innate environment, preferentially use tricellular junctional regions as “portals” for TEM. Our data suggest that the “portals” are junctions that are surrounded by enriched regions of ICAM-1. These enriched ICAM-1 regions mark the location of a nearby portal, as well as preparing the EC junctions to accommodate leukocyte passage. Thus, discontinuities in junctional molecules need not pre-exist but can be initiated by approaching leukocytes via ICAM-1–mediated signaling. Supporting our findings, others have found no evidence for pre-existing discontinuities of the endothelial barrier at tricellular junctions in retinal wholemounts but observed loss of tight junction proteins during leukocyte TEM (34). Although we cannot unequivocally demonstrate that all leukocytes transmigrating via these portals were taking a junctional route (versus a paracellular route in very close apposition to the actual junction), our findings that the ICAM-1 tail peptide blocked leukocyte TEM and VE-cadherin rearrangement at the EC junctions, together with the findings that ICAM-1–mediated signaling directly affects VE-cadherin phosphorylation and mobility (30), strongly argue that the route taken is indeed junctional.

Rearrangement of junctional molecules, such as VE-cadherin, and formation of transient gaps during leukocyte TEM have been demonstrated in vitro (21, 35, 36). We confirmed these findings in situ and, further, showed that signaling mediated via the ICAM-1 cytosolic tail is essential for the formation of observed gaps in VE-cadherin. As the ICAM-1 tail is a potential binding target for Src (37), the likely signaling mechanism for the disassembly of the homologous VE-cadherin interactions is the activation of proteins Src and Pyk2 upon engagement of ICAM-1, leading to phosphorylation of VE-cadherin (31). However, the ability of leukocytes to exert force on EC contacts during TEM (38) could also contribute to the formation of these gaps and cannot be ruled out.

We add another dimension to the complexity of recruitment by showing that the arrangement of EC junctions in different venular regions greatly contributes to leukocyte recruitment and TEM. ECs in venular convergences are significantly smaller, more rounded, and aligned in a more random fashion than ECs in straight regions (Table I). This arrangement produces more tricellular junctions in convergences, apparently making convergences optimally equipped to support TEM. In previous work, we also showed that in venular convergences the two inlet vessels are predicted to create a region of low velocity, increasing leukocyte adhesion probability (26), as was indeed demonstrated in Fig. 1B. Increased leukocyte adhesion together with a higher number of tricellular junctions in venular convergences, resulted in 2.6-fold higher leukocyte TEM compared with straight regions. Thus, although fluid shear likely contributes to leukocyte accumulation in convergences, it is unlikely to affect leukocyte TEM directly. This is because not only the absolute number (higher as a result of more adhered cells in the region) but also TEM efficiency (%TEM/all adhered cells) was significantly higher in convergences (Fig. 5B). Similarly, we ruled out the differential expression of adhesion molecules as key contributors for increased TEM, because the expression of ICAM-1 (Supplemental Fig. 4B) and E-selectin (unpublished data) is not different in straight versus converging regions, confirming local EC morphology as a prime candidate.

Previous analysis of leukocyte adhesion in vivo (10), and work in monolayers (39), showed that as a result of the EC morphology, most adhered leukocytes overlap EC junctions. We have extended these findings to show that differences in leukocyte adhesion distribution with respect to junctional type in convergences versus straight regions are solely due to differences in EC size, shape, and alignment, producing a higher ratio of tricellular to bicellular junctions in venular convergences.

Leukocyte crawling is presumably a tool to get adhered leukocytes to the nearest EC junction, where TEM occurs. Confirming previous observations (10, 12), we showed that ∼90% of leukocytes undergo intraluminal crawling (Fig. 4). This suggests that the initial location of leukocyte adhesion is simply a way station where leukocytes pause before crawling to a “portal” to undergo TEM. Leukocyte crawling, in control venules and monolayers (28, 39) or in MIP-2–stimulated venules (12), is random (independent of blood flow). However, in our study, leukocyte crawling in the presence of fMLP was parallel or perpendicular to blood flow but very rarely against it, suggesting some degree of directionality. The finding that blockade of the luminal portion of the ICAM-1 molecule, but not blockade of the cytoplasmic tail, resulted in loss of the observed directionality (Fig. 7E) argues that ICAM-1 might play a role in directing leukocyte crawling.

We showed that 50 min after fMLP application ICAM-1 expression was significantly increased (Supplemental Fig. 4A) compared with control conditions, indicating endothelial activation. The fMLP-evoked response in this time frame is surprising as fMLP is considered to be a leukocyte activator, and there is very little evidence for its ability to affect ECs. Whether the increase in ICAM-1 expression was directly induced by fMLP treatment or whether it was a consequence of fMLP-induced leukocyte activation and adhesion is not clear and will require new studies. Importantly, upon exposure of the tissue to fMLP, otherwise homogenously expressed ICAM-1 on individual ECs became enriched near EC junctions (primarily tricellular) (Fig. 6A, 6B). The presence of ICAM-1–enriched regions near tricellular junctions could explain the shorter crawling distances that leukocytes exhibit in venular convergences, as well as their lower crawling velocities (Fig. 4D).

In our preparation, fMLP-induced TEM occurred mainly at or very close to EC junctions. Our results differ from those previously reported (40), possibly because of the route of fMLP administration or the different tissue that was investigated. Intriguingly, leukocytes that originally adhered at bi- or tricellular junctions were often observed to crawl away and either detach from the endothelium or transmigrate elsewhere. This argues that initial adhesion doesn’t determine whether the leukocyte will undergo TEM or affect the route it will use, although this requires future studies. It also suggests that not all EC junctions are the same; some EC junctions can act as “portals” for leukocyte TEM and some apparently cannot. Our study implicates ICAM-1 in making some EC junctions act as portals for TEM.

Many molecules make up EC junctions, and most have been implicated in leukocyte TEM (14, 16, 41). Importantly, the nature of endothelial cell-cell contacts varies with the need to regulate vessel barrier function. For example, arterioles have a significantly higher number of tight junctions compared with venules (42) and are significantly less permeable (43). Likewise, there are differences in expression and subcellular localization of junctional adhesion molecules (22). In the current work, we suggest that the junctions that could act as portals for TEM are those junctions that are surrounded by high ICAM-1 expression, but an alternative/additional explanation could be that active portals have a different assembly or different array of junctional proteins compared with elsewhere. The local microanatomy of these junctions might also be different.

As mentioned above, not all junctions are used by leukocytes for TEM. In fact, only a small portion of the junctions act as active portals, and often the same junctional region (bi- or tricellular) is used by multiple leukocytes. Supporting the idea that portals are EC junctions that are enriched in ICAM-1, only ∼43% of all tri- and 23% of all bicellular junctions became enriched in ICAM-1, correlating with the percentages of leukocytes undergoing TEM in these regions. Moreover, we showed that gaps in VE-cadherin staining were primarily formed at tricellular junctions (Fig. 7) and were mediated via ICAM-1 signaling, again suggesting that the observed leukocyte behavior is directly connected to ICAM-1–enriched regions.

Some regions within the venular wall express lower levels of key extracellular matrix proteins that make up the basal lamina than other regions in the same vessel (44). These regions are also associated with gaps between the pericytes and are preferentially used by migrating leukocytes (44). We speculate that localization of these regions with EC junctions could also make these locations optimal for leukocyte TEM.

In summary, we show that leukocytes in their innate environment preferentially target tricellular EC junctions to traverse the vessel wall. We show that EC morphology plays an important role in determining these portals, making venular converging regions optimally equipped to support leukocyte TEM. Moreover, in exploring the nature of the active portals, we suggest that enrichment of ICAM-1 surrounding some EC junctions makes these locations preferred for TEM by signaling to leukocytes the location of the portal and further by mediating rearrangement of junctional molecules at these locations.

We thank J.M. Kuebel for expert technical assistance and Hen Drori for thoughtful discussion.

Disclosures The authors have no financial conflicts of interest.