A novel endocytic pathway induced by clustering endothelial ICAM-1 or PECAM-1

Endothelial cells internalize natural ligands and artificial macromolecular ligands, that have been designed as carriers for specific drug and gene delivery (Danilov et al.,2001; Jacobson et al.,1996; McIntosh et al.,2002; Muzykantov et al.,1996; Spragg et al.,1997). We have previously shown that platelet-endothelial cell adhesion molecule (PECAM-1), an immunoglobulin superfamily cell adhesion molecule, can serve as a receptor for delivery of active enzymes and genetic materials to endothelial cells (Muzykantov et al., 1999; Scherpereel et al., 2001; Wiewrodt et al.,2002). Importantly, monomeric anti-PECAM-1 immunoglobulin G (IgG)and multivalent conjugates larger than 500 nm are not efficiently internalized; multivalent anti-PECAM-1 conjugates 100-300 nm in diameter are readily internalized, although molecular mechanisms that regulate the internalization of anti-PECAM-1 are not well understood(Wiewrodt et al., 2002).

It is becoming apparent that both PECAM-1 and ICAM-1 may serve as plasma membrane receptors to mediate internalization of natural ligands by different types of cells. For instance, coxsackieviruses and rhinoviruses bind ICAM-1 and are internalized (Shafren et al.,1997a), although other coreceptors may be involved in this process(Shafren et al., 1997b). HIV is internalized into brain endothelial cells by a pathway that is analogous to macropinocytosis into endocytic vesicles that also contain ICAM-1(Liu et al., 2002). A pathway related to macropinocytosis has also been implicated in the clearance of apoptotic cell fragments by epithelial cells(Fiorentini et al., 2001). The notion that endothelial cells might help scavenge apoptotic cells is underscored by the observation that PECAM-1 is required for the binding of malaria-infected red blood cells to human umbilical vein endothelial cells(HUVEC) in culture (Treutiger et al.,1997), although these particles were too large to be endocytosed. In a recent study, PECAM-1 expressed by macrophages was found to play an important role in cellular recognition and uptake: apoptotic cells binding to macrophage PECAM-1 were efficiently phagocytosed, whereas live cells activated a signaling cascade through macrophage PECAM-1 to weaken their engagement to macrophages and enable their release (Brown et al., 2002). Cells, apoptotic fragments and viruses binding to ICAM-1 and PECAM-1 are multivalent, complex and, in the case of live cells,active partipicants in cell–cell interactions, making it difficult to discern roles for specific plasma membrane proteins as potential receptors. By contrast, anti-ICAM-1 and anti-PECAM-1 conjugates, although multivalent, will primarily engage only the cell adhesion molecule of interest, which makes them useful probes for examining specific internalization pathways mediated by ICAM-1 or PECAM-1.

There are multiple pathways for ligand internalization involving vesicles 100-300 nm in diameter, including clathrin-mediated endocytosis and the clathrin-independent caveolae-mediated pathway(Mukherjee et al., 1997; Nichols and Lippincott-Schwartz,2001). Each of these endocytic mechanisms differs in sensitivity to pharmacological agents, which enables the mechanism of ligand internalization to be determined. Caveolae-mediated endocytosis is a particularly important pathway in endothelial cells, where ligands such as albumin (Minshall et al.,2000) and orosomucoid(Predescu et al., 1998) are internalized via receptors clustered into caveolae and subsequently transcytosed across the endothelial barrier(McIntosh et al., 2002). There are also clathrin-independent pathways distinct from caveolar endocytosis,which mediate uptake of glycosylphosphatidylinositol (GPI)-anchored proteins,such as the folate receptor (Mayor et al.,1998) and diptheria toxin receptor(Skretting et al., 1999).

The regulation of ICAM-1 and PECAM-1 internalization by endothelial cells is not well understood at present. In particular, whether ICAM-1 and PECAM-1 are internalized by similar pathways is not known. In this study, we defined some key elements regulating the internalization of anti-ICAM-1 or anti-PECAM-1 conjugates by endothelial cells. In each case, clustering of the CAM was required for efficient internalization. Given that anti-ICAM-1 and anti-PECAM-1 conjugates did not colocalize with known endocytic coat proteins and from the analysis of the signaling pathways that regulate the uptake of anti-ICAM-1 and anti-PECAM-1 conjugates, our data suggests that that endothelial cells internalize clustered ICAM-1 and PECAM-1 using a novel endocytic pathway.

Murine monoclonal antibodies to human ICAM-1 (R6.5) and PECAM-1 (mAb 62)were provided by Robert Rothlein (Boehringer-Ingelheim, Ridgefield, CN) and M. Nakada (Centocor, Malvern, PA), respectively. Control murine IgG was from Calbiochem (San Diego, CA). Anti-caveolin-1, anti-clathrin, anti-cholera toxin B and anti-transferrin were from Calbiochem (La Jolla, CA). Polyclonal rabbit anti-6-His-Tag was a gift from MBL (Nagoya, Japan). Secondary fluorescent antibodies were from Jackson ImmunoResearch (West Grove, PA) and Molecular Probes (Eugene, OR). Fluorescent transferrin and cholera toxin B were from Molecular Probes. Polystyrene-latex beads 100 nm in diameter, and loaded with a fluorochrome compatible with FITC fluorescence (Fluoresbrite YG microspheres), were purchased from Polysciences (Warrington, PA). Unless otherwise stated, all other reagents were from Sigma (St Louis, MO).

Pooled human umbilical vein endothelial cells (HUVEC) from Clonetics (San Diego, CA) were maintained in M199 medium (GibcoBRL, Grand Island, NY)supplemented with 15% fetal bovine serum (FBS), 2 mM glutamine, 15 μg/ml endothelial cell growth supplement (ECGS), 100 μg /ml heparin, 100 U/ml penicillin and 100 μg/ml streptomycin. EAhy926 cells from an endothelial-like hybrid cell line generated from HUVEC and A549 cells(Edgell et al., 1983) were cultured in DMEM medium (GibcoBRL) supplemented with 10% FBS, glutamine and antibiotics. Cultures were maintained at 37°C, 5% CO₂ and 95%relative humidity in 1% gelatin-coated tissue culture plastic. HUVEC were used between passage 4 and 5. When seeded for experiments, the cells were cultured onto 12 mm² gelatin-coated coverslips in 24-well plates in the absence of antibiotics and then treated with tumor necrosis factor-α(TNF-α) for at least 16 hours.

Fluorescent microspheres were coated with either anti-ICAM-1, anti-PECAM-1 or control murine IgG by incubation at room temperature (RT) for 1 hour as previously described (Wiewrodt et al.,2002). The coated microspheres (immunobeads) were centrifuged to remove unbound antibodies, then resuspended in 1% bovine serum albumin-PBS and microsonicated for 20 seconds at low power. The effective immunobead diameter was determined by dynamic light scattering (DLS) using BI-90Plus particle size analyzer with BI-9000AT Digital auto-correlator (Brookhaven Instruments,Brookhaven, NY) as previously described(Wiewrodt et al., 2002). This immunobead protocol yielded uniform preparations with particle diameters ranging from 180 to 250 nm. For anti-ICAM-1 immunoconjugates, anti-ICAM-1 was biotinylated and complexed to streptavidin (90% unlabeled, 10% rhodamine labeled) in a manner equivalent to anti-PECAM-1 immunoconjugates as previously described (Wiewrodt et al.,2002). The ratio of biotinylated-ICAM-1 to streptavidin was varied to generate immunoconjugates either smaller than 500 nm or larger than 1000 nm, as determined by DLS.

Confluent HUVEC or EAhy926 cells were pre-incubated overnight with 250 units of TNF-α. By flow cytometry, TNF-α treatment increased ICAM-1 expression by HUVEC and EAhy926 cells ∼10-fold and had little effect on PECAM-1 expression, which is consistent with previously published results (Delisser and Albeda, 1998). The cells were then washed in serum-free medium and incubated in 1% BSA-medium containing a 1:10 dilution of either uncoated microspheres, or immunobeads coated with control murine IgG,anti-ICAM-1 or anti-PECAM-1. The cells were incubated with immunobead preparations for different time periods at 4°C or 37°C, washed in medium and fixed with 2% paraformaldehyde at RT. To distinguish between surface-bound or internalized immunobeads, nonpermeabilized fixed cells were counterstained for 30 minutes at RT with Texas Red (TxR)-conjugated goat anti-mouse IgG to produce double-labeled, yellow particles. The cells were washed in PBS, mounted onto slides with Mowiol and analyzed by fluorescence microscopy. Alexa Fluor 594-labeled cholera toxin B (FL-cholera toxin)counterstained with goat anti-cholera toxin + fluorescein rabbit anti-goat IgG was used as a control for caveolae-mediated uptake. TxR-labeled transferrin counterstained with goat anti-transferrin + fluorescein rabbit anti-goat IgG was used as a control for clathrin-mediated endocytosis.

To identify compartments containing internalized immunobeads, HUVEC monolayers were incubated with immunobeads for 1 hour at 4°C to allow surface binding, washed, then incubated at 37°C for different time periods to permit endocytosis. The cells were fixed, permeabilized and incubated with rabbit polyclonal anti-human caveolin-1, followed by incubation with goat anti-rabbit IgG conjugated to Alexa Fluor 350. Colocalization with clathrin heavy chain was done in a comparable manner, using TRITC-conjugates anti-clathrin.

For microscopy, samples mounted onto glass slides were observed using an Olympus IX70 inverted fluorescence microscope, 40× or 60× PlanApo objectives and filters optimized for fluorescent immunobeads (excitation BP460-490 nm, dichroic DM505 nm, emission BA515-550 nm), TxR fluorescence(excitation BP530-550 nm, dichroic DM570 nm, emission BA590-800+ nm) and Alexa Fluor 450 (excitation BP360-370 nm, dichroicDM400 nm, emission BA420-460 nm)(Chroma Technology, Brattleboro, VT). Separate images for each fluorescence channel were acquired using a Hamamatsu Orca-1 CCD camera. The images were then merged and analyzed with ImagePro 3.0 imaging software (Media Cybernetics, Silver Spring, MD) as previously described(Wiewrodt et al., 2002). For quantitation, merged images of cells labeled with immunobeads were scored automatically for total green fluorescent particles and noninternalized immunobeads (double-labeled yellow particles). Uptake was calculated as the percentage of internalized immunobeads with respect to the total number of cell-associated immunobeads. Statistical significance was determined by Student's t test.

Mammalian pcDNA3 expression vectors encoding for 6-His-tagged versions of human dynamin-2 [wild-type and dominant-negative forms (K44A), (PH*)] were gifts from Drs S. Schmid (Scripps Research Institute, La Jolla, CA)(Altschuler et al., 1998) and M. Lemmon (U Penn School of Medicine, Philadelphia, PA)(Lee et al., 1999). EAhy926 endothelial cells were transfected using Lipofectin (GibcoBRL) complexed to 1.5 μg DNA/dish encoding either dynamin-2, dynamin-2(K44A) or dynamin-2(PH*). Each construct includes a 6-His amino terminus tag to distinguish it from endogenous dynamin-2. Twelve hours after transfection, the cells were stimulated with TNF-α, incubated for 36 hours and then anti-ICAM-1 or anti-PECAM-1 uptake was determined by double labeling as described above. Following labeling of surface-bound material, the cells were permeabilized with 0.2% Triton X-100 and then immunostained using 5 μg/ml rabbit anti-6-His-Tag and goat anti-rabbit IgG conjugated to Alexa Fluor 350 to identify cells expressing recombinant dynamin-2.

For studies using pharmacological inhibitors, TNF-α-stimulated HUVEC or EAhy926 were pre-incubated for 30 minutes at 37°C in the presence of one of the following agents: 50 μM monodansyl-cadaverine (MDC), 1 μg/ml filipin, 50 μM genistein, 3 mM amiloride, 25 μM monensin, 0.5 mM cytochalasin D, 0.1 μM latrunculin A, 20 μM nocodazole, 5 μM bisindolyl-maleimide-1 (BIM-1), 10 μM 1-(5-isoquiniline sulphonyl)-2-methylpiperazine (H7), 0.1 μM phorbol 12-myristate 13-acetate(PMA), 10 μM radicicol, 10 μM Y-27346 or 0.5 μM wortmannin(Barreiro et al., 2002; Fujimoto et al., 2000; Parton et al., 1994; Racoosin and Swanson, 1989; Sahai and Marshall, 2002; Schlegel et al., 1982; Schnitzer et al., 1994; Swanson, 1989; Torgersen et al., 2001; Watanabe et al., 2001; West et al., 1989). Molecular targets for selected inhibitors are shown in Fig. 10. The concentration of each agent was selected using literature values and was optimized qualitatively by fluorescence microscopy (not shown). Also, we examined the effectiveness of each agent using suitable controls (e.g. Fig. 4). Potassium depletion was done by pre-incubating the cells for 15 minutes in potassium depletion buffer (0.14 M NaCl, 2 mM CaCl₂, 1 mg/ml glucose, 20 mM HEPES, pH 7.4) diluted 1:1 with water to make it hypotonic(Koval et al., 1998). After treatment, the cells were incubated with immunobeads, cholera toxin or transferrin at 37°C, in the presence of K⁺-depletion buffer or the given inhibitors, and then fixed and double labeled for surface-bound material as described above.

Monomeric antibodies to PECAM-1 are poorly internalized by endothelial cells (Muzykantov et al.,1999; Wiewrodt et al.,2002). Multivalent anti-PECAM-1 conjugates with a diameter of 100-300 nm are readily internalized, but the efficiency of internalization decreases with increasing conjugate size(Wiewrodt et al., 2002). To test whether this is the case for anti-ICAM-1 conjugates, we examined the internalization of anti-ICAM-1 conjugates in two different size ranges by TNF-α-stimulated HUVEC (Fig. 1A,B). Consistent with our previous results using anti-PECAM-1 conjugates, we found that anti-ICAM-1 conjugates with a diameter less than 500 nm were internalized by HUVEC, whereas conjugates with diameter greater than 1000 nm showed little, if any, internalization. Because antibody conjugates show a broad distribution of particle sizes(Wiewrodt et al., 2002), we performed subsequent experiments using anti-ICAM-1 and anti-PECAM-1 immunobeads, which have a more uniform diameter in the size range that allows internalization (Wiewrodt et al.,2002). Anti-ICAM-1 immunobeads were internalized by HUVEC when incubated at 37 °C but not at 4°C(Fig. 1C,D). Similar results were obtained using anti-PECAM-1 beads (S.M., R.W. and A.T. et al.,unpublished). Following internalization, there was further clustering of conjugates and immunobeads, which was probably due to endosome fusion events that occurred as the particles were transported along the endocytic pathway.

On average, ∼125 immunobeads/cell were internalized after a 1 hour incubation at 37°C. Fig. 1Gshows that uptake of anti-ICAM-1 immunobeads was more rapid (∼10 minutes half time for uptake) than that of anti-PECAM-1 immunobeads (∼20 minutes half time for uptake). Similar results were obtained for anti-ICAM-1 and anti-PECAM-1 immunobeads internalized by naïve HUVEC, indicating that TNF-α had little effect on the mechanism of internalization. This also suggests that the mechanism for internalization is not sensitive to the surface density of ICAM-1, as virtually all of the anti-ICAM-1 immunobeads were internalized after 30 minutes at 37°C, despite the difference in binding in the absence (42±15 beads/cell) or presence of TNF-α(165±54 beads/cell). Also, as found for anti-PECAM-1, monomeric anti-ICAM-1 was not internalized (S.M., R.W. and A.T. et al.,unpublished).

Because dynamin-2 is frequently involved in vesicle-mediated internalization and phagocytosis(Altschuler et al., 1998; Gold et al., 1999; Henley et al., 1998; Lee et al., 1999), we examined the role for dynamin-2 in uptake of anti-ICAM-1 immunobeads. Given the low transfection rate of HUVEC, we used the endothelial-like cell line, EAhy926,which showed a 40% transfection efficiency using Lipofectin. EAhy926 cells internalized anti-ICAM-1 immunobeads in a manner comparable to HUVEC(Fig. 2). These cells were transiently transfected to express amino-terminal 6-His-tagged forms of either wild-type or dominant-negative dynamin-2 (K44A or PH*). Expression of recombinant proteins was identified by immunofluorescence, using an antibody that recognizes the 6-His epitope (Fig. 2). Cells transfected with wild-type dynamin-2 showed no effect on the uptake of anti-ICAM-1 immunobeads compared with control cells. By contrast, cells expressing ether dynamin-2(K44A) or dynamin-2(PH*) showed less anti-ICAM-1 immunobead uptake than control cells, suggesting that uptake of these immunobeads required dynamin-2 (Fig. 2). This was not due to a net decrease in immunobead binding,which was equivalent for nontransfected EAhy926 cells (23±3 beads/cell)and EAhy926 cells expressing wild-type (24±16 beads/cell) or mutant dynamin-2 [25±4 beads/cell (PH*), 14±5 beads/cell (K44A)].

We also found that the dominant-negative dynamin-2 constructs inhibited the uptake of Alexa Fluor 594-conjugated cholera toxin (FL-cholera toxin, S.M.,R.W. and A.T. et al., unpublished) that is internalized by caveolae-mediated endocytosis (Schnitzer et al.,1994). However, few, if any, anti-ICAM-1 immunobeads colocalized with caveolin-1-positive structures, regardless of whether the immunobeads were bound to the plasma membrane or internalized by HUVEC(Fig. 3). By contrast,FL-cholera toxin showed extensive colocalization with caveolin, which is consistent with previously published reports(Puri et al., 2001). Newly internalized anti-ICAM-1 immunobeads also did not colocalize with clathrin. In fact, there was a low level of clathrin immunofluorescence shown by HUVEC,consistent with a less-dominant role for clathrin-coated pits in endocytosis than caveolae-mediated pathways in endothelial cells(Schubert et al., 2001). Thus,despite being a dynamin-dependent process, anti-ICAM-1 conjugate uptake by endothelial cells was unlikely to be through caveolae- or clathrin-coated vesicles.

We therefore used a series of pharmacological inhibitors to further characterize internalization of anti-ICAM-1 and anti-PECAM-1 conjugates by HUVEC. The specificity of different inhibitors was confirmed using fluorescent transferrin and cholera toxin as controls for clathrin-mediated and caveolar endocytosis, respectively (Fig. 4). As shown in Fig. 5, inhibitors of clathrin-mediated transferrin endocytosis (MDC,potassium depletion) did not inhibit the uptake of anti-ICAM-1 or anti-PECAM-1 immunobeads. Furthermore, inhibitors of caveolae-dependent cholera toxin uptake (filipin, genestein) were not effective at inhibiting anti-ICAM-1 or anti-PECAM-1 immunobead internalization. Because uptake of anti-ICAM-1 and anti-PECAM-1 immunobeads appeared to be through a unique internalization pathway, we examined the effect of other inhibitors on immunobead endocytosis. Previous work has indicated that amiloride, an inhibitor of the sodium/proton pump, can inhibit macropinocytosis by dendritic cells(West et al., 1989). Amiloride had little effect on internalization of FL-cholera toxin or transferrin,suggesting that it did not inhibit caveolae- or clathrin-mediated endocytosis(Fig. 4). However, amiloride inhibited uptake of anti-ICAM-1 and anti-PECAM-1 immunobeads by HUVEC(55±15% and 60±9%, respectively) and by EAhy926 cells(34±9% and 24±3% inhibition, respectively. Anti-ICAM-1 immunobead binding was equivalent for control (125±21 beads/cell) and amiloride-treated HUVEC (154±37 beads/cell), suggesting that amiloride did not decrease ICAM-1 surface expression. TNF-α stimulation was not required, given that amiloride inhibited anti-PECAM-1 immunobead uptake by naïve HUVEC by 50.5±6.2%. Also, anti-ICAM-1 immunobead uptake by amiloride-treated HUVEC remained inhibited during a 3 hour (43.8±2.9%)and 5 hour incubation (50.6±5.2%), suggesting that amiloride altered the extent of immunobead uptake, rather than uptake kinetics. Furthermore,this was not likely to be due to an effect on ion homeostasis, given that the ionophore monensin had little, if any, effect on anti-ICAM-1 and anti-PECAM-1 immunobead uptake (Fig. 5).

Because protein kinase C (PKC) has been reported to play a pivotal role in macropinocytosis and phagocytosis by macrophages(Araki et al., 1996; Larsen et al., 2000; Swanson, 1989), we tested the effect of PKC inhibitors on immunobead internalization by HUVEC. As shown in Fig. 6, the PKC inhibitors BIM-1 and H-7 inhibited immunobead uptake by ∼30% and ∼60%,respectively. H-7 treatment also inhibited the uptake of anti-ICAM-1 and anti-PECAM-1 by EAhy926 cells by 55±11% and 48±7%, respectively. HUVEC pretreated with 0.1 μM PMA for 30 minutes (conditions that stimulate PKC activity) showed a high level of anti-ICAM-1 and anti-PECAM-1 internalization (>95%), and the total level of immunobead uptake by HUVEC was stimulated nearly twofold. Anti-ICAM-1 immunobead binding was equivalent for control (125±21 beads/cell) and BIM-1-treated HUVEC (140±7 beads/cell), suggesting that BIM-1 did not decrease ICAM-1 surface expression. Stimulation by TNF-α was not required, since inhibiting PK-C activity also inhibited uptake of anti-PECAM-1 immunobeads by naïve HUVEC(47.5±9.8%) and uptake of anti-PECAM-1 immunobeads by naïve HUVEC was enhanced 1.3-fold by PMA. Also, anti-ICAM-1 immunobead uptake by BIM-1-treated HUVEC remained inhibited during a 3 hour (39.6±4.2%) and 5 hour incubation (42.9±5.2%), suggesting that BIM-1 altered the extent of immunobead uptake, rather than uptake kinetics. Taken together, these results are consistent with internalization of anti-ICAM-1 and anti-PECAM-1 immunobeads by a PKC-dependent pathway. However, since H-7 may also interfere with actin-based contractility (Volberg et al., 1994), this effect may also contribute to the inhibition of uptake.

In fact, the formation of F actin stress fibers is frequently associated with ICAM-1 crosslinking (Thompson et al.,2002; Wang and Doerschuk,2002). Therefore, we examined the effect of anti-ICAM-1 immunobeads on the formation of actin stress fibers by HUVEC. As shown in Fig. 7, stress fibers were rapidly induced by anti-ICAM-1 immunobead binding. Immunobeads appeared to align along actin stress fibers before internalization; this is shown most prominently by the blue labeled immunobeads in Fig. 7C,D. Vesicles containing internalized immunobeads continued to be associated with stress fibers after internalization and remained associated with actin during a 3 hour incubation. Few, if any, anti-ICAM-1 immunobeads induced formation of an actin coat(phagocytic cup) at the site of internalization, which is a hallmark of phagocytosis and macropinocytosis (Grimmer et al., 2002; Lee and Knecht,2002). Note that this is probably not a problem with the detection of actin coats, as a previous study found that ∼20% of 0.2 μm beads internalized by Fc receptors in macrophages were associated with actin coats(Koval et al., 1998).

Given the dramatic association of anti-ICAM-1 immunobeads with actin in HUVEC, we also examined the cytoskeletal requirements for the uptake of anti-ICAM-1 and anti-PECAM-1 immunobeads. In contrast to macropinocytosis by macrophages (Racoosin and Swanson,1992), microtubules were not required for the internalization of anti-ICAM-1 immunobeads by HUVEC, since internalization was not inhibited by nocodazole (Fig. 8). Also,nocodazole did not significantly inhibit the uptake of anti-PECAM-1 immunobeads (12±9% inhibition). Cytochalasin D, which caps short actin filaments, had little effect on the uptake of anti-ICAM-1 immunobeads(18±7% inhibition) or anti-PECAM-1 immunobeads (8±8%inhibition). However, the more effective actin depolymerizing agent,latrunculin, inhibited anti-ICAM-1 immunobead uptake(Fig. 8) and uptake of anti-PECAM-1 immunobeads (69±15% inhibition).

We also examined inhibitors that affect kinases known to play a role in regulating actin organization. In contrast to macropinocytosis by phagocytes(Araki et al., 1996; West et al., 2000), wortmanin had little, if any, effect on uptake of anti-ICAM-1 immunobeads by HUVEC,suggesting that PI3-kinases were not involved in immunobead uptake. However,as shown in Fig. 8, uptake of anti-ICAM-1 immunobeads was inhibited by the Src kinase inhibitor radicicol and the ROCK inhibitor Y27632, consistent with the notion that both the PKC pathway and Rho pathway regulate actin cytoskeletal rearrangements required for internalizaton of clustered ICAM-1. Four inhibitors of anti-ICAM-1 immunobead internalization – latrunculin, amiloride, radicicol and Y27632 – all inhibit the formation of actin stress fibers induced by anti-ICAM-1 immunobead binding (Fig. 9), underscoring the correlation of uptake by HUVEC with actin mobilization.

In this study, we found that endothelial cells internalize clustered ICAM-1 and clustered PECAM-1 by a novel endocytic pathway(Table 1). CAM-mediated endocytosis was distinct from caveolae-mediated uptake(McIntosh et al., 2002; Minshall et al., 2000; Predescu et al., 1998) as it was not inhibited by genestein or filipin and the bound conjugates did not colocalize with caveolin. Also, uptake of anti-ICAM-1 and anti-PECAM-1 conjugates was distinct from clathrin-mediated endocytosis, because it was not inhibited by potassium depletion or MDC and the conjugates did not colocalize with clathrin (Hansen et al.,1993; Schlegel et al.,1982). Whether CAM-mediated endocytosis is specific for endothelial cells remains to be determined, although PECAM-1 transfected REN mesothelioma cells show a similar endocytic pathway(Wiewrodt et al., 2002).

On the basis of amiloride sensitivity and PKC dependence, internalization of clustered anti-ICAM-1 and anti-PECAM-1 seemed to be related to macropinocytosis (Lamaze and Schmid,1995; Nichols and Lippincott-Schwartz, 2001; Orth et al., 2002; Swanson and Watts, 1995), a pathway that is not typically associated with endothelial cells. Nonetheless,CAM-mediated endocytosis was distinct from `classical' macropinocytosis, on the basis of several criteria (Table 1). For example, in contrast to the dynamin-2 requirement we observed for uptake of anti-ICAM-1 immunobeads by EAhy926 endothelial cells,the K44A dominant-negative dynamin-2 did not inhibit macropinocytosis by fibroblasts (Orth et al.,2002). Although dynamin is required for endocytosis(Altschuler et al., 1998; Henley et al., 1998; Lee et al., 1999) and phagocytosis (Gold et al.,1999), CAM-mediated endocytosis differed in other ways from these processes. For instance, anti-ICAM-1 immunobeads did not colocalize with either clathrin or caveolin (Fig. 4). Also, uptake of anti-ICAM-1 immunobeads did not require PI3K activity, which is needed for phagocytosis(Araki et al., 1996; Cox et al., 1999), as well as macropinocytosis (West et al.,2000).

The uptake of anti-ICAM-1 and anti-PECAM-1 conjugates required clustering of cell adhesion molecules (Muzykantov et al., 1999; Wiewrodt et al.,2002). ICAM-1 clustering has been found to stimulate multiple intracellular signaling pathways (Adamson et al., 1999; Etienne et al.,1998), including a PKC signaling pathway that results in the phosphorylation of cytoskeletal and focal adhesion proteins, thereby enabling actin filament rearrangement(Etienne-Manneville et al.,2000). Consistent with this, anti-ICAM immunobeads induced actin stress fiber formation and were associated with stress fibers before internalization (Fig. 7). Both Src kinase and ROCK activity are required for CAM-mediated endocytosis and these inhibitors also inhibited stress fiber formation induced by anti-ICAM-1 immunobeads (Fig. 9). ROCK activity also enables remodeling of F-actin attached to adherence junctions and controls their stability (Sahai and Marshall, 2002). This also provides a potential link between ICAM-1 or PECAM-1 clustering, Src kinase, PKC and dynamin-2(Fig. 10), as dynamins are downstream targets for Src kinase (Ahn et al., 2002) and PKC (Powell et al., 2000).

Dynamin, via interactions with endophilin and profilin, helps recruit actin to sites of endocytic activity (Farsad et al., 2001; Witke et al.,1998). Proteins in the ezrin-radixin-moesin (ERM) family are also good candidates to link internalization of ICAM-1 and PECAM-1 to the actin cytoskeleton (Bretscher et al.,1997; Cao et al.,1999). For instance, ezrin binds directly to the C-terminus of ICAM-1 (Heiska et al., 1998). Another actin binding protein, α-actinin, has also been shown to bind to the C-terminus of ICAM-1 (Carpen et al.,1992). Perhaps signaling induced by ICAM-1 or PECAM-1 clustering can indirectly recruit ERM proteins to the plasma membrane. For instance,phosphorylation of ERM proteins by a ROCK-dependent pathway can help recruit them to the plasma membrane (Hirao et al.,1996). Intriguingly, ROCK activity has been associated with complement receptor-mediated phagocytosis, but is not required for Fc receptor-mediated endocytosis (Olazabal et al., 2002). ROCK has also been shown to phosphorylate sodium proton exchangers (NHE) to enhance binding of ERM proteins(Denker et al., 2000). Both processes might correspond to the ROCK requirement for uptake of clustered ICAM-1 or PECAM-1 (Fig. 10). This is further suggested by the ability of amiloride to inhibit uptake of clustered ICAM-1 and PECAM-1, since amiloride can disrupt the association of ERM proteins with NHE, an effect that is independent of ion channel activity(Denker et al., 2000; Putney et al., 2002).

Another major distinction from macropinocytosis and phagocytosis is that uptake of anti-ICAM-1 and anti-PECAM-1 conjugates larger than 500 nm in diameter was poor (Fig. 1)(Wiewrodt et al., 2002). Also,anti-ICAM-1 immunobeads smaller than 500 nm diameter did not induce the formation of an actin cup or coat (Fig. 7), which is typically induced by larger particles internalized by phagocytosis (Koval et al.,1998), suggesting that formation of an actin coat is crucial for internalization of larger particles. Although the mechanisms that cells use to control the size threshold for internalization is not known at present, given that ICAM-1 and PECAM-1 primarily regulate cell–cell contacts, a small size threshold for internalization may be a means by which endothelial cells avoid engulfing other cells.

Consistent with a size threshold for ICAM-1-mediated internalization, an ICAM-1 enriched structure is formed at the contact site between lymphocytes and HUVEC, where the endothelial cell appears to partially engulf the lymphocyte (Barreiro et al.,2002). Furthermore, the lymphocyte-endothelial cell docking structure requires ROCK activity, but not phosphatidylinositol 3-kinase (PI-3 kinase) (Barreiro et al.,2002), comparable to our observations for anti-ICAM-1 and anti-PECAM-1 immunobeads (Fig. 9). Whether plasma membrane internalization is part of the mechanism required to maintain this docking structure is not known at present. One possibility is that CAM-mediated endocytosis might help to remodel cell–cell junctions as leukocytes migrate along endothelial cells. If so, this might be analogous to the turnover of gap junctions, which is mediated by the engulfment of cell–cell junctions sites to create endocytic vesicles in the 200-500 nm diameter size range(Gaietta et al., 2002; Jordan et al., 2001).

The size threshold for internalization of clustered ICAM-1 and PECAM-1 might also enable endothelial cells to distinguish small apoptotic fragments from intact cells bound to the endothelial cell surface, such as other endothelial or blood cells (Barreiro et al., 2002; DeLisser and Albelda, 1998; Johnson-Leger et al., 2000; Worthylake and Burridge, 2001). The notion that endothelial cells could also scavenge apoptotic fragments via a pathway comparable to CAM-mediated endocytosis is appealing (Brown et al.,2002; Treutiger et al.,1997); however, whether this is the case remains to be determined.

Understanding the mechanisms that regulate uptake of anti-ICAM-1 and anti-PECAM-1 conjugates will probably help to extend the utility of these agents as the basis for endothelium-specific drug-targeting vehicles(Li et al., 2000; Muzykantov et al., 1999; Scherpereel et al., 2002; Scherpereel et al., 2001; Wiewrodt et al., 2002). For instance, inhibitors of conjugate uptake might help to increase their stability by reducing the extent of delivery to lysosomes and other degradative compartments. Animal studies combining the agents used in this work with the administration of pharmacologically active, enzyme-carrying anti-ICAM or anti-PECAM conjugates will be used to determine the feasibility of this approach.

We thank B. Daugherty for a critical reading of the manuscript. S.M. is supported by a fellowship from Fundación Ramón Areces (Spain). R.W. is a postdoctoral fellow of the Mildred Scheel Stiftung fur Krebsforschung der Deutschen Krebshilfe e.V. (D/98/02288). Supported by an American Heart Association grant-in-aid 9950389N (M.K.) National Institutes of Health SCOR in Acute Lung Injury, HL60290, Project 4, (V.R.M., S.M.A.); grants HL/GM 71175-01 (V.R.M.), GM61012 (M.K.) and P01 HL019737-26, Project 3 (M.K.)and Department of Defense Grant PR 012262 (V.R.M.).