Cell adhesion in a dynamic flow system as compared to static system. Glycosphingolipid-glycosphingolipid interaction in the dynamic system predominates over lectin- or integrin-based mechanisms in adhesion of B16 melanoma cells to non-activated endothelial cells.

Initial adhesion of B16 melanoma variants to non-activated endothelial cells is mediated through specific interaction between GM3 (NeuAc alpha 2----3Gal beta 1----4Glc beta 1----Cer) expressed on melanoma cells and lactosylceramide (LacCer, Gal beta 1----4Glc beta 1----Cer) expressed on endothelial cells. This adhesion is predominant over integrin- or lectin-mediated adhesion in a dynamic flow experimental system employing a parallel plate laminar flow chamber (Lawrence, M. B., Smith, C. W., Eskin, S. G., and McIntire, L. V. (1990) Blood 75, 227-237). In this system, a tumor cell suspension flows over a glass plate coated with glycosphingolipid, lectin, or fibronectin, and adhesion is recorded on videotape. These conditions were designed to mimic the microvascular environment in which tumor metastatic deposition takes place. In contrast, lectin- and fibronectin-based mechanisms are predominant in previously used static adhesion systems. Under static conditions, the relative degree of adhesion of the four B16 variants to endothelial cells or to LacCer-coated plates was the same as their relative degree of GM3 expression (i.e. BL6 approximately F10 greater than F1 greater than WA4), and adhesion was inhibited in the presence of methyl-beta-lactoside, or liposomes containing LacCer or GM3. Adhesion was also inhibited by pretreatment of B16 cells with anti-GM3 antibody DH2 or sialidase and by pretreatment of endothelial cells with anti-LacCer antibody T5A7. Under dynamic flow conditions, WA4 cells did not adhere to mouse endothelial cells at high shear stress (greater than 2.5 dynes/cm2) but did adhere at lower shear stress. In contrast, BL6 and F10 cells adhered strongly at both low and high shear stress. BL6 cell adhesion to endothelial cells at both low and high shear stress was inhibited in the presence of antibody DH2, ethyl-beta-lactoside, or lactose, as well as by pretreatment of BL6 cells with sialidase. Thus, some clear differences, as well as similarities, in cell adhesion under static versus dynamic conditions are demonstrated. These findings suggest that melanoma cell adhesion to endothelial cells, based on GM3/LacCer interaction, initiates metastatic deposition, which may trigger a series of "cascade" reactions leading to activation of endothelial cells and expression of Ig family or selectin receptors, thereby promoting adhesion and migration of tumor cells.

Initial adhesion of B16 melanoma variants to nonactivated endothelial cells is mediated through specific interaction between GM3 (NeuAca2~3Ga4314Glc/?l +Cer) expressed on melanoma cells and lactosylceramide (LacCer, Gal#I1+4Glc@l~Cer) expressed on endothelial cells. This adhesion is predominant over integrin-or lectin-mediated adhesion in a dynamic flow experimental system employing a parallel plate laminar flow chamber (Lawrence, M. B., Smith, C. W., Eskin, S. G., and McIntire, L. V. (1990) Blood 75, 227-237). In this system, a tumor cell suspension flows over a glass plate coated with glycosphingolipid, lectin, or fibronectin, and adhesion is recorded on videotape. These conditions were designed to mimic the microvascular environment in which tumor metastatic deposition takes place. In contrast, lectin-and fibronectinbased mechanisms are predominant in previously used static adhesion systems. Under static conditions, the relative degree of adhesion of the four B16 variants to endothelial cells or to LacCer-coated plates was the same as their relative degree of GM3 expression (i.e. BL6 = F10 > F1 > WA4), and adhesion was inhibited in the presence of methyl-&lactoside, or liposomes containing LacCer or GM3. Adhesion was also inhibited by pretreatment of B16 cells with anti-GM3 antibody DH2 or sialidase and by pretreatment of endothelial cells with anti-LacCer antibody T5A7. Under dynamic flow conditions, WA4 cells did not adhere to mouse endothelial cells at high shear stress (>2.5 dynes/cm2) but did adhere at lower shear stress. In contrast, BL6 and F10 cells adhered strongly at both low and high shear stress. BL6 cell adhesion to endothelial cells at both low and high shear stress was inhibited in the presence of antibody DH2, ethyl-#I-lactoside, or lactose, as well as by pretreatment of BL6 cells with sialidase. Thus, some clear differences, as well as similarities, in cell adhesion under static versus dynamic conditions are demonstrated. These findings suggest that melanoma cell adhesion to endothelial cells, based on GM3/ LacCer interaction, initiates metastatic deposition, which may trigger a series of "cascade" reactions leading to activation of endothelial cells and expression of Ig family or selectin receptors, thereby promoting adhesion and migration of tumor cells.
* The study was supported by National Cancer Institute Outstanding Investigator Grant CA42505 (to S. H.) and funds from The Biomembrane Institute, in part under a research contract with Otsuka Pharmaceutical Co. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The molecular basis of specific cell recognition is of central importance in current cell biology. Studies to date suggest that four combinations of molecular "families" provide the basis for most recognition events: (i) various adhesive proteins are recognized by members of the integrin family (1-3); (ii) members of the immunoglobulin family interact with each other or with integrins (3,4); (iii) carbohydrates (CHOs)' are recognized by members of the lectin family (5-7), particularly selectins (8,9); (iv) CHOs interact with other CHOs (10). These studies were all based on static adhesion systems; i.e. adhesion molecules are coated on solid-phase and incubated under static conditions with cells expressing specific receptors or CHO epitopes. Using such a static system, we previously reported the specific adhesion, spreading, and enhanced motility of GM3-expressing cells on Gg3-or LacCer-coated solidphase, based on GM3/Gg3 or GM3/LacCer interaction (11,12).
Adhesion of neutrophils and monocytes to endothelial cells has been observed under dynamic flow conditions, designed to mimic the microvascular environment (13,14). Selectindependent adhesion of neutrophils under dynamic flow conditions appears to occur preferentially over integrin-dependent adhesion (15). Tumor cell adhesion to activated endothelial cells, as mediated by integrin receptors expressed on tumor cells, and by ICAMs or selectins expressed on activated endothelial cells, has received considerable attention recently (16,17). However, there is still no firm evidence for involvement of these adhesion molecules in the very initial interaction between circulating tumor cells and non-activated endothelial cells, particularly under dynamic flow conditions. We now present evidence that initial adhesion of B16 melanoma cells to non-activated mouse and human endothelial cells is mediated by interaction between GM3 and LacCer, which are highly expressed on B16 cells and endothelial cells, respectively. This GM3/LacCer interaction apparently occurs prior to activation of endothelial cells, and prior to involvement of ICAMs or selectins, and predominates over integrin-or lectinmediated cell adhesion in a dynamic flow system.  (18); BL6 produces lung metastasis from subcutaneously grown tumors (19). Wheat germ agglutininresistant B16 clone WA4, which shows no lung colonization potential (20), was donated by Dr. Max Burger, BioCenter, University of Basel, Switzerland. All these cell lines were characterized by surface expression of GM3 as defined by mAb DH2 (see "Results"). Human umbilical vein endothelial cells (referred to hereafter simply as "human endothelial cells") were purchased from Cell Systems, Kirkland, WA. Mouse splenic endothelial cells were established by one of the authors (Y. S.) at the Department of Pathology, Kawasaki Medical School, Kurashiki, Okayama. The cell line was characterized by typical endothelial cell morphology distinct from that of fibroblasts and macrophages (i.e. contact-inhibited, extremely flattened, polygonal appearance at confluence), active uptake of fluorescent-labeled acetylated low density lipoprotein, and staining by fluorescent-labeled Griffonia simplicifolia lectin known to stain murine endothelial cells (21,22). While murine endothelial cells used in this experiment were not susceptible to stimulation by human IL-10, this cell line was activated by TNFa, and expression of VCAM-1 was greatly enhanced upon TNFa stimulation (data not shown). Cell biological properties of this cell line will be presented elsewhere.* GM3, LacCer, Gg3, Gb4, PG, and other GSLs were prepared in pure state from organ extracts (23). Methyl-p-lactoside and methyl-P-N-acetyllactosaminide were synthesized in this laboratory.
ConA, laminin, G. simplicifolia lectin, and Erythrina corralodendron lectin (24)  Cell Surface Labeling and Characterization of GSLs in Endothelial Cells (Including Mouse Lung Microvascular Endothelial Cells)-Human and mouse endothelial cells were cultured in 25-cm2 Corning flasks until confluence, and cell surface CHOs were labeled as previously described (32). Cell layers were washed with PBS (pH 7.0), treated with Gal oxidase (10 units/ml) in PBS (pH 7.0) for 1 h at 37 "C, washed with 20 mM Tris-HC1 (pH 8.4) containing 150 mM NaC1, and reduced with 0.5 mCi of NaB3H4 in 1 ml of 20 mM Tris-HCI (pH 8.4) for 30 min at room temperature. Next, 1 mM NaBH4 in 20 mM Tris-HC1 (pH 8.4) was added and incubated for 30 min at room temperature (32). Cells were collected by rubber scraper and washed with PBS. GSLs were extracted two times from labeled cells with 3 ml of isopropyl alcohol/hexane/water 55:25:20. The extract was dried, acetylated in pyridine/acetic anhydride (l:l), and GSLs were separated from other neutral lipids and phospholipids by Florisil column (33). The separated GSLs were applied to TLC plates and developed in chloroform/methanol/water (504010 v/v/v). Plates were dried, treated with enhancer, and autoradiographed. The pattern of 3H-labeled GSLs separated on TLC was compared with those developed with orcinol-HaS04 or from immunostaining with various anti-GSL mAbs.
In order to study surface expression of GSLs on mouse lung microvascular endothelial cells, lungs were perfused with PBS (pH 7.0) followed by 10 units of Gal oxidase solution in PBS (pH 7.0), via pulmonary artery using a syringe. After 30 min, the lung was perfused with 5 mCi of NaB3H4 in PBS (pH 7.4) and then perfused with cold NaBH4. This was followed by extraction of lung, preparation of GSL fraction, TLC separation, and autoradiography as described above.
Binding of B16 Cell Variants to GSL-, Lectin-, Fibronectin-, and Laminin-coated Surfaces, and to Endothelial Cells, in a Static Adhesion System-For preparation of GSL-coated plastic wells, 100 p1 of GSL bottom assay plates (Falcon Probind) and dried at 37 'C. 100 p1 of (1 pg) in absolute ethanol was placed on each well of 96-well flatthe appropriate concentration of fibronectin or laminin was placed on each well and incubated at 4 "C for 16 h. Wells were washed extensively with PBS, and nonspecific binding sites of each well were * Y. Sadahira, N. Kojima, T. Kimoto T, unpublished results. coated with 1% bovine serum albumin in PBS at 37 "C for at least 1 h. Wells were then washed with PBS and used as fibronectin-or laminin-coated surfaces.
B16 melanoma variants BL6, F10, F1, and WA4 were cultured in 10 ml of RPMI 1640 containing 10% fetal calf serum in 25-cm* Corning flasks and labeled with [3H]thymidine (2 pCi/ml) for 16 h. For measurement of cell adhesion, labeled cells were detached by treatment in 0.02% EDTA at 37 "C. Detached cells were collected and washed two times with PBS. Cells were suspended in PBS at a density of 2 X lo5 cells/ml, and 100-p1 aliquots of suspension were added to each well of 96-well plates co-coated with GSL and fibronectin or laminin as described above. After addition of cells, plates were centrifuged at 100 X g for 1 min and incubated for 30 min at 37 "C. Plates were washed with PBS, and remaining adherent cells were collected by cell harvester and counted by scintillation counter.
Suspensions of human and mouse endothelial cells (8 X 10' cells/ ml) were placed in 96-or 48-well flat-bottom plates (purchased respectively from Falcon, Lincoln Park, NJ, and Costar, Cambridge, MA) pretreated with 0.5% gelatin at 37 "C for 1 h. Endothelial cells were cultured until confluence. Medium was removed, endothelial cells were washed with PBS, and labeled B16 cells (1 X 10' cells/well (volume 100 pl) for 96-well plates, 3 X lo5 cells/well (volume 300 pl) for 48-well plates) in PBS were placed in each well and incubated at 37 "C for 30 min. In IL-1 stimulation experiments, endothelial cells were treated with 5 units/ml of human recombinant IL-10 (Boehringer Mannheim, 1000 units/ml) for at least 4 h at 37 "C before addition of B16 cells. After incubation, cells were washed two times with PBS, and adherent cells were detached by trypsinization, collected, and counted by scintillation counter.
Assay for Inhibition of Cell Adhesion by Oligosaccharides, GSLliposomes, mAbs, and Enzymatic Pretreatment of Cells-BL6 cells were harvested with 0.02% EDTA, washed with PBS, and suspended in PBS at a concentration of 1 X lo6 cells/ml. Methyl-0-D-lactoside and lactose were dissolved in PBS at a concentration of 200 mM. Liposomes (1 ml) were made from 500 nmol of cholesterol, 500 nmol of dipalmitoyl PC, and 200 nmol of GSL in PBS as described previously (11). In this case, GSL concentration was 200 pM. 100 p1 of 200 mM oligosaccharide or liposomes containing 200 p M of GSL was diluted 2-fold with 100 pl of PBS. 100 pl of BL6 cell suspension (2 X lo5 cells) was added to 100 pl of oligosaccharide solution or liposome suspension and incubated for 30 min at 37 "C. After incubation, mixtures of cells with oligosaccharide or liposome were placed on plates coated with LacCer (1 pg/well), human endothelial cells, or mouse endothelial cells, as described above. After incubation at 37 "C for 30 min, wells were washed and adherent cells collected and counted as described above.
BL6 cells (1 X 106/ml) were treated with 10 pg/ml of mAbs directed against various GSLs, lectins, or adhesion molecules for l h a t 4 "C, then washed two times with PBS. mAbs used were DH2 (anti-GM3, IgGs), T5A7 (anti-LacCer, IgM), 2D4 (anti-Gg3, IgM), BE2 (anti-H, IgM), SNH3 (anti-SLe", IgM), 3B7 (anti-ELAM-1, IgG1), and 5D7 (anti-Gal-binding lectin, IgG3). In separate experiments, BL6 cells were treated with sialidase (0.1 unit/ml) in PBS at 4 "C for 30 min. After treatment, cell viability was tested, cells counted by using Trypan Blue, and cell number adjusted to 1 X 106/ml. Endothelial cells were cultured in 48-well plates, and treated BL6 cells (3 X lo'/ well) were added. After 30 min, wells were washed with PBS and remaining cells were counted. treated with 10 pg/ml of mAbs at 37 "C for 30 min, washed with PBS, In other experiments, endothelial cells in 48-well plates were and untreated BL6 cells (3 X 10'/well) were added to treated endothelial cells. As controls, BL6 cells or endothelial cells were treated with a mixture of 10 pg/ml mouse IgG and IgM. Adhesion Assay in a Dynamic Flow System-A parallel plate laminar flow chamber connected to an infusion pump (model 935, Harvard Apparatus, Cambridge, MA) was used to simulate the flow shear stresses present in physiological microvascular environments. The flow chamber consists of a glass plate on which a parallel, transparent plastic surface is attached with a Silastic rubber gasket; there is a 114-pm gap between the two surfaces, and this gap is connected to an inlet and outlet. A laminar flow with defined rate and wall shear stress is achieved by manipulation of the infusion pump, which is connected to the inlet of the flow chamber. Endothelial cells are grown as a monolayer, or adhesion molecules are coated, on the glass plate, and a laminar flow of cell suspension is passed through the chamber. Cell movements are observed under inverted phase-contrast microscope (Diaphot-TMD Nikon) and recorded by time-lapse videocassette recorder. Adhesion is observed as rolling followed by stopping of cells. This assembly is essentially the same as that described by Lawrence et al. (13,15). Number of cells bound during 3 min a t different shear stresses from 0.4 to 4.8 dynes/cm2 were counted from several fields recorded on videotape. Wall shear stress (T) was calculated by the equation of Lawrence et al. (13,14): where p = coefficient of viscosity (1.0 cP), Q = volumetric flow rate (cm:'/s), a = half channel height (in this case, 5.7 X lo-" cm), and b = channel width (1.3 cm).
Coating of Adhesion Molecules or Endothelial Cells on Glass Plates in the Dynamic Flow System-For lectins, fibronectin, laminin, and GSLs used in this study, 10 pl of solution having a concentration of 20-200 pg/ml was placed on a marked area (0.5-cm diameter) on a glass plate (38 X 75 mm, Corning Glassworks, Corning, VA) and dried in a refrigerator a t 4 "C. Dried plates were immersed in PBS at 37 "C for 1 h and washed extensively with several changes of PBS. For GSL coating, GSL-liposomes were prepared from 200 pg of GSL, 200 pg of cholesterol, and 400 pg of PC in 1 ml of PBS as described previously (11). 10 pl of GSL-liposome solution was placed on glass plate and dried at 4 "C, and plates were washed with PBS as described above. Quantity of adsorbed molecules was determined using 1 2 ' 1 labeling for lectins, fibronectin, or laminin, or [3H]cholesterol labeling for GSLliposomes. Under these conditions, almost the entire quantity of protein, regardless of whether fibronectin, laminin, or lectin, was adsorbed on the glass plate. For example, when 100 pg/ml fibronectin was applied, 12.5 ? 1.8 ng/mm' was adsorbed. Likewise, almost all GSL-liposome dried on the glass plate was adsorbed; e.g. when 200 pg/ml GSL-liposome was applied, 31.3 k 5.2 ng GSL/mm2 was adsorbed.
Endothelial cells were coated by placing 100 pl of a suspension containing 2 X lo5 mouse or human endothelial cells on glass plates and culturing in a COZ incubator a t 37 "C until confluence.
Plates coated with adhesion molecules or endothelial cells were affixed in a flow chamber, and a suspension of B16 melanoma cells was passed through the chamber as described in the preceding section. B16 cells were harvested from culture by 0.02% EDTA in PBS, and suspended in PBS at a concentration of 1 X 105/ml.

Patterns of GSL Surface Expression on Mouse Melanoma B16 Variants, Human and Mouse Endothelial Cells, and
Mouse Lung Microvascular Endothelial Cells-GSL compositions of BL6, F10, and F1 cells were essentially the same, consisting of GlcCer, GM3, LacCer, and a minor quantity of SPG. WA4 was characterized by a much lower chemical quantity of GM3, but higher quantities of SPG, PG, and IV3FucnLc4. These B16 variants differed markedly in surface reactivity with anti-GM3 mAb DH2 (BL6 = F10 > F1 >> WA4) (Fig. IA). Mouse endothelial cells contained GlcCer, LacCer, and Gb4 or PG as neutral GSLs, a major ganglioside with the same mobility as GDla, and relatively small quantities of GM3 (Fig. 1B). GD3 was absent from both mouse and human endothelial cells, as indicated by lack of immunostaining with anti-GD3 mAb (data not shown). Based on labeling experiments, LacCer was the major GSL exposed at the cell surface in both mouse and human endothelial cells. In the Gal oxidase/NaB3H4 experiment with mouse lung microvascular endothelial cells, three GSL bands were observed LacCer, Gb3 (or Gg3), and Gb4 (Fig. IC). Human endothelial cells contained GlcCer, LacCer, Gb4, and PG as neutral GSLs, and GM3 and SPG as major gangliosides (data not shown), in agreement with a previous report (34).
Adhesion of Mouse Melanoma B16 Variants to Non-activated Human and Mouse Endothelial Cells Is Based on GM3/ LacCer Interaction, and Occurs Prior to Integrin-mediated Adhesion"B16 melanoma variants BL6, F10, F1, and WA4, which show declining metastatic potential and reactivity with anti-GM3 mAb DH2 in that order (see "Experimental Procedures"), also showed relative adhesion to LacCer-or Gg3-  WA4) at longer incubation times (Fig. 2, C and D). BL6, F10, and F1 cells showed similar integrin-dependent adhesion to fibronectin-coated plates (Fig. 2 A ) , whereas adhesion of WA4 cells was much lower (data not shown; see "Discussion"). It should be noted that BL6 adhesion to LacCer-or Gg3-coated plates occurred earlier than integrin-dependent adhesion to fibronectin or laminin (10-20 min uersus 30-50 min under comparable experimental conditions) (Fig. 2B). This finding suggests that cell adhesion based on GSL/GSL interaction is a faster process than integrin-dependent adhesion, which may explain the differences observed for static adhesion systems uersus dynamic flow systems (see following section).
Since high quantities of cell surface LacCer are expressed in human and mouse endothelial cells, including mouse lung microvascular endothelial cells3, we compared adhesion (and inhibitability of adhesion) of BL6 cells to human and mouse endothelial cells, and to LacCer-coated plates. All three types of adhesion were inhibited by LacCer or Gg3 at 50-100 pM concentration, and by methyl-or ethyl-P-lactoside at 50-100 mM concentration, but not by free lactose or methyl-p-Nacetyllactosaminide at the latter concentration (Fig. 3). The difference in adhesion of the four B16 variants to human and mouse endothelial cells was particularly obvious when endothelial cells were in the non-activated state (Fig. 4, A and B ) , and adhesion was =lo times higher to mouse endothelial cells than to human endothelial cells (maximum binding 1-1.5 x 10' versus 1-2.5 X lo4 cells/well, respectively). For mouse endothelial cells, differences in adhesion of the B16 variants were obvious even after activation (Fig. 40). Since the degree of LacCer expression on mouse endothelial cells is similar to that on human endothelial cells, the much greater adhesion of melanoma cells to mouse endothelial cells is assumed to be due to expression of unidentified adhesion molecules on nonactivated mouse endothelial cells, and to a synergistic effect of GMB/LacCer and integrin-dependent adhesion, as previously reported (35). In human endothelial cells, differences in adhesion of B16 variants were less clearly observed after ILl p stimulation (Fig. 4C), suggesting that adhesion following activation is based mainly on integrin receptors, ICAMs, or selectins.
BL6 adhesion to non-activated human or mouse endothelial cells was blocked by pretreatment of BL6 cells with mAb DH2 or sialidase (Fig. 5, A and B ) , but no such effect was observed for activated human endothelial cells (Fig. 5C). Conversely, BL6 adhesion to non-activated endothelial cells was blocked by pretreatment of endothelial cells with anti-LacCer mAb T5A7 (Fig. 5, D and E ) , but this effect was not observed for activated human endothelial cells (Fig. 5F). These findings suggest that B16 adhesion to non-activated human and mouse endothelial cells is based on GM3/LacCer interaction.
Adhesion of BL6 Cells to GSL-coated Glass Plates Is Predominant Over That to Lectin-or Fibronectin-coated Surfaces in a Dynamic Flow System-A dynamic flow system as described by Lawrence et al. (13)(14)(15), in which a cell suspension flows over a GSL-, lectin-, or fibronectin-coated glass plate and is recorded on videotape (see "Experimental Procedures"), is designed to mimic conditions in the microvascular environment in which tumor cell metastatic deposition takes place. In such a system, adhesion of BL6 cells to glass plates coated with Gg3 or LacCer was high, but no adhesion to surfaces coated with PG or GM3 was observed (Fig. 6A). Adhesion to surfaces coated with ConA or an E. corralodendron lectin that 'Chemical quantity of LacCer in human endothelial cells was %30% of total neutral GSL, and =lo% of total neutral GSL was surface-labeled. In metabolic labeling, 10.8% of total neutral GSL was labeled as LacCer in human endothelial cells (34). In contrast, 80-90% of total neutral GSL was surface labeled in mouse endothelial cells, while the chemical quantity of LacCer was <2% of total GSL. Details of chemical quantities and characterization will be presented elsewhere (see footnote 2). recognizes N-acetyllactosamine residue (24) was not pronounced unless very high concentrations (~2 0 0 pg/ml) of these lectins were applied (Fig. 6B), despite the fact that much lower concentrations (10-20 pg/ml) of these lectins were sufficient to cause strong adhesion and spreading of BL6 cells in a static system (Fig. 6C). Fibronectin or laminin, even applied at very high concentrations (~1 0 0 pg/ml), had no significant effect on BL6 adhesion in a dynamic system (Fig.  6B), whereas fibronectin or laminin concentrations as low as 20 pg/ml produced strong adhesion and spreading in a static system (Fig. 6C). Fibronectin-dependent adhesion, even with 200 pg/ml concentration, required much longer incubation (20-30 min) as compared to ConA-dependent adhesion. Adhesion based on GM3/Gg3 interaction was much weaker than ConA-or fibronectin-dependent adhesion in a static system (Fig. 6C).
BL6 adhesion to LacCer-coated glass plates in a dynamic system was obvious even at high shear stress (1.5-3.0 dynes/ cm2) and remarkable at low shear stress ( d . 0 dynes/cmz).

Cell Adhesion through
Glycosphingolipids in Dynamic Flow System Such adhesion was diminished by inclusion of ethyl-p-lactoside in the medium or by pretreatment of BL6 cells with sialidase or anti-GM3 mAb DH2 (Fig. 7).
Adhesion of B16 Variants to Mouse Endothelial Cells in a Dynamic Flow System-Adhesion of F10, BL6, F1, and WA4 cells to mouse endothelial cells in a dynamic flow system under different shear stress conditions is shown in Fig. 8A. BL6 cells adhered much more strongly than F1 cells under both low and high shear stress. WA4 cells did not adhere at all, whereas BL6 and F1 cells showed clear adhesion, under high shear stress (>2.0 dynes/cm2). Under low shear stress, WA4 cells adhered more strongly than F1 cells. In the static system, WA4 (compared to F1) also showed a greater number of cells adhering to mouse endothelial cells with a 5-min incubation period. In the dynamic system, BL6 adhesion to mouse endothelial cells was inhibited in the presence of 0.1 M lactose, 50 mM ethyl-p-lactoside, or by pretreatment of BL6 cells with anti-GM3 mAb DH2 (Fig. 8B). DISCUSSION Cell adhesion mediated by various mechanisms (see Introduction) has been repeatedly studied under static conditions. Physiologically, however, adhesion of blood cells (or tumor cells) among themselves (aggregation) or to endothelial cells takes place under dynamic flow conditions, i.e. in a moving bloodstream. While the role of CHOs in cell adhesion based on lectins (5-7), selectins (8,9), or GSL/GSL interaction (10) has been clearly documented, there have been no comparative studies of adhesion under static versus dynamic flow conditions. In this study, we have (i) demonstrated that adhesion of melanoma cells to non-activated endothelial cells is based on GMB/LacCer interaction, and (ii) compared this adhesion system to lectin-and fibronectin-dependent adhesion under dynamic flow conditions.
There are several lines of evidence that adhesion of B16 melanoma cells to non-activated endothelial cells depends on GM3/LacCer interaction: (i) GM3 is the major GSL expressed cytofluorometry, was minimal. Therefore, the role of Galbinding lectin in adhesion of melanoma cells to endothelial cells appears to be minor compared to the role of GM3/ LacCer interaction.
While BL6 adhesion to LacCer-or Gg3-coated plates was evident at an earlier time than adhesion to fibronectinor laminin-coated plates under static conditions, maximal strength of fibronectin-, laminin-, or ConA-mediated adhesion was greater than that of adhesion based on GSL/GSL interaction. Adhesion under dynamic flow conditions differed in many important respects. In the dynamic system, BL6 adhesion to LacCeror Gg3-coated surfaces was stronger than integrin-dependent adhesion to fibronectin-or laminincoated surfaces, and adhesion to ConA or Erythrina lectincoated surfaces was negligible at low concentration (20 pg/ ml) even at low shear stress, but became evident at high concentration (~2 0 0 pg/ml). In the static system, as little as 10-20 pg/ml of these lectin8 was sufficient to induce strong cell binding or spreading. Similarly, in the static system, BL6 cells adhered strongly to plates coated with 10 pg/ml fibronectin or laminin; however, an incubation period of >30 min was required for this adhesion process, in contrast to the rapid adhesion based on GSL/GSL interaction. In the dynamic system, B16 cell adhesion to fibronectinor laminin-coated glass surfaces was negligible even at an fibronectin or laminin concentration of 100 pg/ml, and even at low shear stress (c1.0 dynes/cm2). B16 cells did not adhere to glass plates coated with GM3-liposomes, PG-liposomes, and control PC-cholesterol liposomes (lacking GSL). In these experiments, liposomes were applied on glass and air-dried in a refrigerator (see "Experimental Procedures"). The lack of adhesion to GM3 and other liposomes indicates that B16 cell adhesion to LacCer-liposome-coated plates is highly specific.
Blood-borne tumor metastasis is generally believed to be initiated by adhesion of tumor cells to microvascular endothelial cells (36,37) and to activate platelets, leading to tumor cell-platelet or tumor cell-neutrophil aggregation, a major cause of microembolism and metastatic deposition. Expression of selectin GMP-140, which recognizes the tumor-associated antigens sialosyl-Le" and sialosyl-Le', may play an important role in tumor cell adhesion and aggregation (38,39). Results of the present study make clear the pathobiological significance of GM3/LacCer interaction as the basis for specific adhesion of melanoma cells to non-activated endothelial cells under dynamic flow conditions. The possibility that this represents the initial event in tumor cell metastasis is suggested by the apparent correlation of strength of adhesion (under high shear stress) with relative metastatic potential of the melanoma cells. Recent interest has been focused on the possible mediation of tumor cell adhesion by ICAMs or selectins which are expressed on activated endothelial cells (16,17). For this process, a sufficient local concentration of factors secreted from tumor cells (e.g. TGFP or TNFa) is required to activate endothelial cells in situ. In dynamic microvascular flow, adhesion of circulating tumor cells to nonactivated endothelial cells is a prerequisite for the process of endothelial cell activation by tumor cells, and therefore an essential step in initiation of metastasis. Tumor-associated lectins which recognize the Gal residue of N-acetyllactosamine have been assumed to play a role in such adhesion (25, 40), and Gal-binding lectin on endothelial cells has been claimed to mediate adhesion of tumor cells to endothelial cells (41).
It is highly plausible that initial adhesion of tumor cells, under physiological dynamic flow conditions, to non-activated endothelial cells is based on GSL/GSL interaction, which induces activation of endothelial cells, and is subsequently reinforced by induction of selectin or ICAM expression on activated endothelial cells. This trend is particularly pronounced in a dynamic adhesion system, where cell adhesion mediated by GSL/GSL interaction is greater than that mediated by glycoprotein CHO/lectin interaction, and integrindependent adhesion is relatively minor. It should be noted that BL6 adhesion to LacCer strongly stimulated cell migration, in agreement with our previous report (35), thereby promoting transendothelial migration. Adhesion of tumor cells to endothelial cells could induce activation of endothelial cells through TGFP or TNFa. Thus, a series of "cascade" reactions could be triggered by GSL/GSL interaction between tumor cells and non-activated endothelial cells. In vivo metastasis based on GSL/GSL interaction can be blocked by oligosaccharides, GSL derivatives, or GSL-liposomes. Studies based on this approach are in progress.