A Specific Microdomain (“Glycosynapse 3”) Controls Phenotypic Conversion and Reversion of Bladder Cancer Cells through GM3-mediated Interaction of (cid:1) 3 (cid:2) 1 Integrin with CD9 *

Cell motility dependent on the organization and function of Human bladder cancer cell lines KK47 (noninvasive and nonmeta-static)andYTS1(highlyinvasiveandmetastatic),bothderivedfrom transitional bladder epithelia, are very similar in terms of integrin composition and levels of tetraspanin CD9. Tetraspanin CD82 is 24 at room temperature, added to cells in serum-free medium at concentrations of 10, 20, and 50 (cid:4) M , incubated 37 exogenously added same way in a separate experiment. Cells analyzed co-IP and confocalmicroscopyforinteractionbetweenCD9and 3andbyphago-kinetic gold sol assay for motility.

The interaction of tumor cells with their microenvironment may define the direction of tumor development and the degree of malignancy (1). Such an interaction is likely based on the structure and func-tion of the microdomain at the tumor cell surface interfacing with the normal cell microdomain and with extracellular matrix components, particularly at the basement membrane. A crucial event for the progression of many types of tumor cells of epithelial origin is their adhesion/ motility/invasion on the basement membrane underlying epithelial cells. In this process, specific microdomains of tumor cells are considered to interact with laminin-5 ("epiligrin") (2) or laminin-10/11 (3), which are major components of epithelial basement membrane and ligands of integrin ␣3 or ␣3␤1 (4). Such microdomains, having proteolipid/tetraspanin (PLP/TSP) 2 CD9, integrin ␣3␤1, and ganglioside GM3 (5)(6)(7), are capable of controlling cell adhesion and motility (6, 8 -10), in contrast to "caveolar membrane" or "raft," which is cholesterol-dependent (11), has no integrins (12), and is not involved in cell adhesion and motility. GM3/TSP/integrin microdomains have therefore been termed "glycosynapse" (8,13), in analogy to the microdomain involved in immunocyte adhesion/antigen presentation with concurrent signaling, termed immunosynapse (14). Among the glycosynapses, GM3-CD9integrin complex termed "glycosynapse 3" (8) was suggested previously to play a role in the regulation of cell motility (see "Discussion").
The goal of the present study was to clarify contrasting composition of GM3, ␣3, and CD9 in glycosynapse 3 and their interaction and to define motility/invasive properties of two closely related human bladder cancer cell lines: KK47 with noninvasive, low motility phenotype, and YTS1 with invasive, high motility phenotype. Our results make clear the dual functional role of GM3 in glycosynapse 3 as follows: (i) high or low GM3 level promotes or inhibits the interaction of ␣3 with CD9 in order to stabilize or de-stabilize the ␣3-CD9-GM3 complex within the microdomain; (ii) high or low GM3 level activates or inhibits c-Src through association or dissociation with Csk within the same microdomain. Through either process, the GM3 level regulates tumor cell motility/invasiveness. Oncogenic conversion or phenotypic reversion from invasive to noninvasive cells is associated with change of the GM3 level, which mediates the functional alteration of the ␣3 interaction with CD9 and of the degree of c-Src activation with Csk interaction.
The present study may help explain the mechanism by which glycosynapse organization and composition define oncogenic conversion or reversion to normal cell phenotypes, as suggested in previous studies of bladder cancer (7,19,20), colorectal cancer (5,6,21), and Jun-transformed cells (22).

Cells, Antibodies, and Reagents
Cells-The YTS1 cell line was established from invasive human urinary bladder cancer (23) and was donated by H. Kakizaki (Department of Urology, Yamagata University, Yamagata, Japan). The KK47 cell line was established from noninvasive, superficial bladder cancer (24) and was donated by T. Masuko (Department of Hygienic Chemistry, Faculty of Pharmaceutical Science, Tohoku University, Sendai, Japan). Both cell lines were grown in RPMI 1640 containing 10% fetal bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin at 37°C, 5% CO 2.

Determination of Total Levels of TSPs and Integrins by SDS-PAGE and Western Blot
Total cell lysate was prepared as described previously (6). Briefly, ϳ1 ϫ 10 7 cells were collected, and the cell pellet was suspended in 1 ml of RIPA buffer (1% Triton X-100, 150 mM NaCl, 25 mM Tris, pH 7.5, 0.5% sodium deoxycholate, 0.1% SDS, 5 mM pyrophosphate, 50 mM NaF) containing 75 units of aprotinin and 2 mM phenylmethanesulfonyl fluoride. The suspension was kept on ice for 30 min and Dounce homogenized (10 -15 strokes). The lysate was centrifuged at 15,000 rpm for 15 min at 4°C, and the supernatant was subjected to SDS-PAGE and Western blot after determination of protein concentration. The membranes were reblotted with anti-␤-actin antibody after stripping with Re-Blot Plus solution (Chemicon). The intensity of Western blot was determined by densitometry using Scion image program.

Determination of Total GSL and Ganglioside Levels by Thin Layer Chromatography and Immunostaining
GSLs were extracted and analyzed as described previously (27,28). Briefly, YTS1 and KK47 cells were grown until ϳ90% confluence in 15-cm dishes and washed three times with PBS (137 mM NaCl, 8.1 mM Na 2 HPO 4 , 2.68 mM KCl, 1.47 mM KH 2 PO 4 , pH 7.4). ϳ2ϫ10 7 cells were collected and extracted twice with C/M (2:1). The extracts were dried under a N 2 stream. To remove phospholipids, the dried residue was dissolved in 2 ml of 0.1 M NaOH in methanol and incubated at 40°C for 2 h. After neutralization with 1 N HCl, fatty acids were extracted twice with 2 ml of hexane. GSLs in the lower phase were evaporated, dissolved in water, and applied to 3 ml of BondElut C18 columns (Varian, Harbor City, CA) to remove salt. Columns were washed with water, and GSLs were eluted with C/M (2:1). Solvents were evaporated, and equal aliquots of GSLs dissolved in C/M (1:1) were subjected to HPTLC plate (Merck), developed with C/M, 0.2% CaCl 2 in H 2 O (50:40:10), and stained with 0.5% orcinol in 2 N sulfuric acid to visualize GSLs, or immunostained with DH2 using Vecstain ABC kit and Immunostain horseradish peroxidase-1000 kit according to the manufacturer's instructions. For analysis of GSLs in GEM, after fractionation of postnuclear fractions (PNF) through sucrose density gradient ultracentrifugation, low density fractions and high density fractions were applied to C18 column and processed as described above.

PNF and GEM Preparations for Determination of GEM Components
PNF was prepared as described previously (6). Briefly, ϳ4 ϫ 10 7 cells were collected, and the pellet was suspended in 1 ml of Brij 98 lysis buffer (1% Brij 98, 25 mM HEPES buffer, pH 7.5, 150 mM NaCl, 5 mM EDTA) containing 75 units of aprotinin and 2 mM phenylmethanesulfonyl fluoride. The suspension was kept on ice for 30 min and Dounce homogenized (10ϳ15 strokes). The lysate was centrifuged at 2,500 rpm for 7 min at 4°C to remove nuclei and debris. PNF was subjected to sucrose density gradient ultracentrifugation to separate low density membrane fractions as described previously (6). Fractions were separated and numbered 1-12 from top to bottom. Aliquots of each fraction, containing equal protein content (ϳ2.5 g), were analyzed by SDS-PAGE and Western blot.

Phagokinetic Gold Sol Assay for Cell Motility
Cell motility was studied by an improved method (29) based on Ref . 30. Briefly, 24-well plates were incubated with 1% bovine serum albumin for 24 h at 37°C, washed with 100% ethanol, and dried. Gold sol suspension (gold colloidal particles) was prepared by adding 11 ml of H 2 O and 6 ml of 36.5 mM Na 2 CO 3 to 1.8 ml of 14.5 mM AuHCl 4 . The mixture was boiled, and 1.8 ml of freshly prepared 0.1% formaldehyde solution was added. Gold sol suspension was put in each well and incubated for 40 min at room temperature, and the plate was washed with culture medium. Cells in culture medium (1 ϫ 10 4 /well) were plated onto gold sol-coated wells and incubated for 16 h at 37°C. Migratory cells were observed and photographed under light microscope (Nikon). Migratory areas of 20 cells of each well were measured by Scion image program and expressed as square pixels.

Determination of Interaction between CD9 and Integrins by Co-IP
Interaction between CD9 and integrins was analyzed by co-IP as described previously (6). Briefly, PNF (400 l, containing 400 g protein) was prepared as described above and mixed with protein A/Gagarose beads (ϳ30-l bed volume). The mixture was placed on rotator at 4°C for 3 h and centrifuged at 1,000 rpm to collect supernatant. ϳ3 g of antibody were added to the supernatant and rotated at 4°C over-night; the antibodies used were rabbit anti-␣3 and -␣5 and mouse anti-␤1. Protein A/G-agarose beads was added, rotated at 4°C for 3 h, and centrifuged to collect the beads. After washing twice with lysis buffer, the immunoprecipitates were resolved in SDS-PAGE sample buffer and subjected to SDS-PAGE and Western blot with mouse anti-CD9 IgG 1 antibody.

Interaction of CD9 with ␣3 Analyzed by Laser-scanning Confocal Microscopy
The interaction between CD9 and ␣3 was also analyzed by confocal microscopy as described previously (6). Briefly, YTS1 and KK47 cells were grown for 24 h on cover glass (12 mm diameter) placed in 24-well plates. Cells on cover glass were washed three times with PBS and fixed with 3.7% paraformaldehyde/PBS for 15 min. Fixed cells were washed three times with PBS, incubated with 1% bovine serum albumin, 0.1% NaN 3 , PBS for 30 min and incubated with mouse anti-CD9 IgG 1 mAb for 1 h at room temperature. After washing and permeabilization with 0.05% Triton X-100 in PBS for 5 min, rabbit anti-␣3 was added and incubated for 1 h at room temperature. Cells were incubated with a mixture of goat anti-mouse IgG 1 -labeled with Texas Red and goat antirabbit IgG labeled with FITC for 1 h at room temperature, washed, mounted with a drop of Glycergel mounting medium (Dako, Carpinteria, CA), and observed by laser-scanning confocal microscopy (Fluo-View TM , Olympus, Tokyo, Japan) using an appropriate filter set.

Effect of Exogenous GM3 on Interaction of CD9 with ␣3 and Cell Motility in YTS1 Cells
YTS1 cells were grown in a 10-cm dish for 24 h, and growth medium was changed to serum-free medium. GM3 was dissolved in serum-free medium by sonication and standing for 24 h at room temperature, added to cells in serum-free medium at concentrations of 10, 20, and 50 M, and incubated at 37°C for 24 h. GM1 was exogenously added in the same way in a separate experiment. Cells were analyzed by co-IP and confocal microscopy for interaction between CD9 and ␣3 and by phagokinetic gold sol assay for cell motility.

Effect of P4 on GM3 and Ceramide Levels, Interaction of CD9 with ␣3, and Cell Motility in KK47 Cells
KK47 cells were grown in a 10-cm dish for 24 h at 37°C. The medium was changed to fresh growth medium with or without P4 (1 M) and further incubated for 72 h. The cells were analyzed by co-IP and confocal microscopy for interaction between CD9 and ␣3 and by phagokinetic gold sol assay for cell motility. Ceramide in P4-treated cells was analyzed by HPTLC with C/M/water (65:25:4) and charring in 3% cupric acetate and 10% phosphoric acid for 20 min at 130°C, on 400 g of cellular protein basis, as described previously (31,32).
Flow Cytometry-Cells were detached by trypsin/EDTA and washed with PBS. Aliquots of cells (1 ϫ 10 5 ) were incubated with mouse anti-CD9 IgG 1 for 1 h on ice, washed with PBS, incubated with goat antimouse IgM ϩ G-labeled with FITC for 40 min on ice, fixed in 2% paraformaldehyde/PBS, and analyzed using a Coulter EPICS XL flow cytometer (Beckman Coulter, Fullerton, CA).

Effect of GM3 Level on c-Src and Csk in Glycosynapse 3
c-Src and its phosphorylation at Tyr-416 (for activation) and at Tyr-527 (for inhibition) were determined by Western blot analysis in KK47 versus YTS1 cells, using phosphorylation site-specific antibodies. PNF, low density membrane fractions (fractions 4 -6), and high density membrane fractions (fractions 10 -12) separated through sucrose density gradient ultracentrifugation were analyzed. Csk was determined simultaneously in these fractions. To test the GM3 effect on c-Src activity, KK47 cells were pretreated with P4 for 72 h (see "Effect of P4 on GM3 and Ceramide Levels, Interaction of CD9 with ␣3, and Cell Motility in KK47 Cells"), followed by Western blot analysis of c-Src, Tyr-416, Tyr-527, and Csk. The intensity of ␥-tubulin in each fraction was used as loading control. YTS1 cells were incubated with 50 M GM3 in serumfree medium for various durations (1, 3, and 16 h) (see "Effect of Exogenous GM3 on Interaction of CD9 with ␣3 and Cell Motility in YTS1 Cells"), followed by sucrose density gradient ultracentrifugation to separate low density and high density fractions, and Western blot analysis of each fraction for determination of c-Src phosphorylation status and Csk translocation into low density fractions (GEM).
␤1; PLP/TSP CD9 and CD82; gangliosides) in total cell lysate prepared in RIPA buffer from KK47 and YTS1 as shown in Fig. 1, A-C. Levels of the integrins were essentially the same in the two cell lines (Fig. 1A). The CD9 level was similar, and CD82 was absent in both cell lines (Fig. 1B). The ganglioside expressed in YTS1 was mainly GM2, whereas that in KK47 was mainly GM3, as revealed by TLC with orcinol/ sulfuric acid staining. GM3 level in KK47 was 4 -5 times higher than in YTS1 (Fig. 1C).
The amounts of ␣3 and CD9 in low density insoluble membrane fraction and high density soluble fraction, separated by sucrose density gradient ultracentrifugation from PNF of cell lysate prepared with lysis buffer containing 1% Brij 98, were analyzed by SDS-PAGE followed by Western blot (Fig. 1D, top). ␣3 was enriched in fractions 4 -6, whereas CD9 was enriched in fractions 4 and 5. The degree of enrichment of both components was higher in YTS1 than in KK47. Neither ␣3 nor CD9 was detectable in fractions 1-3 or fractions 7-9 from YTS1 or KK47 (data not shown). Gangliosides were concentrated in fractions 4 -6 from both cell lines (Fig. 1D, bottom). The GM3 concentration in combined low density fractions (fractions 4 -6) from KK47 and YTS1 was ϳ15 ng/30 g of protein and Ͻ2 ng/30 g of protein, respectively. GM2 concentration in the same fractions from YTS1 and KK47 was ϳ12 ng/30 g of protein and Ͻ2 ng/30 g of protein, respectively. Neither GM3 nor GM2 was detectable in fractions 1-3 or fractions 7-9 from YTS1 or KK47 (data not shown). Association of ␣3, ␣5, and ␤1 with CD9 in YTS1 versus KK47 cells. A, different levels of ␣3/CD9, ␣5/CD9, and ␤1/CD9 in association with YTS1 versus KK47. PNF of YTS1 or KK47 was immunoprecipitated (IP) with ␣3, ␣5, or ␤1. Level of CD9 immunoprecipitated with ␣3, ␣5, or ␤1 was determined by Western blot (WB) with anti-CD9 antibody. As control, PNF was immunoprecipitated with CD9 and Western-blotted with anti-CD9. B, different level of ␣3/CD9 association in YTS1 versus KK47 studied by confocal microscopy. The cells were double-stained with anti-␣3 (green) and anti-CD9 (red). In KK47, ␣3 was co-localized with CD9 (yellow) more extensively than in YTS1. Interactions of Components in KK47 Versus YTS1 Microdomain-Interactions were determined by co-IP and by confocal microscopy. Co-IP of ␣3 with CD9 and of ␤1 with CD9 was performed by the addition of anti-␣3 or anti-␤1 antibody to PNF, followed by Western blot with anti-CD9, as described under "Materials and Methods." Co-IP of ␣3 with CD9 was much higher in KK47 than in YTS1, whereas co-IP of ␤1 with CD9 was weak (and similar) in the two cell lines. ␣5 was not co-immunoprecipitated with CD9 from either cell line ( Fig. 2A). Co-localization of ␣3 with CD9 was significantly higher in KK47 than in YTS1, as indicated by the enhanced merge image in confocal microscopy (Fig. 2B).
Difference in Motility of KK47 Versus YTS1 by Phagokinetic Gold Sol Assay-Transwell membrane motility through thick Matrigel was much higher for YTS1 than for KK47 (7). This difference in cell motility was confirmed in the present study by phagokinetic gold sol assay (Fig.  3, A and B).

Knockdown of CD9 in KK47 Cells through RNAi Causes Increased
Motility-Preliminary experiments with transient transfectants of KK47 using four different CD9-specific sequences (A-D) showed that sequence B was the most effective. Therefore, permanent transfectants with this sequence were cloned as described under "Materials and Methods." Four clones (B201, B201, B254, and B255) were analyzed with control transfectant (S17), which was transfected with control scrambled sequence based on CD9 sequence B. All four clones showed significant reduction of CD9 expression, particularly B255, in which CD9 expression was depleted (Fig. 4, A-C), compared with the control, whereas expression of ␣3 and ␤1 and GM3 was essentially the same in the four clones and control clone (Fig. 4D). Phagokinetic motility was assayed using gold sol-coated plates as described under "Materials and Methods" and compared among the four clones and control. Motility of B255 and B254 was significantly higher than that of control (p Ͻ 0.001 and p Ͻ 0.01 respectively). Motility of the other two clones (B201 and B202) was almost the same as control (Fig. 4E).
GM3 Depletion by P4 in KK47 Cells Causes Reduced ␣3-to-CD9 Interaction and Increased Motility-The high GM3 level characteristic of KK47 was nearly depleted by the incubation of cells with 1 M P4 (Fig.  5A), whereby significant change of ceramide level was not detected (data not shown). GM3-depleted KK47 showed clear reduction of ␣3-to-CD9 interaction, as indicated by reduced co-IP between ␣3 and CD9 (Fig. 5B), and by the reduced merge image in confocal microscopy  KK47 cells were transfected with CD9-specific oligonucleotide or a scrambled control oligonucleotide in pSUPER, together with pPUR. Transfectants were cloned after puromycin selection and analyzed as described under "Materials and Methods." Four knockdown clones (B201, B202, B254, and B255) were compared with control clone (S17). A, CD9 mRNA level measured by RT-PCR. Reduced level of CD9 mRNA in four knockdown clones compared with control clone. As control, mRNA level of actin was measured by RT-PCR. B, flow cytometry analysis of CD9 expression at the cell surface of each transfectant. Expression was reduced in four knockdown clones (shaded lines) compared with control clone (solid lines). C, left, Western blot analysis of CD9 level in four knockdown clones compared with control clone. ␤-actin was reblotted as control after stripping. Right, scion image densitometry analysis of each band, normalized with actin level. D, ␣3, ␤1, and GM3 expression in four CD9 knockdown clones compared with control clone. ␣3 and ␤1 levels were measured by Western blot. GM3 level was revealed by orcinol staining and DH2 immunostaining. E, phagokinetic cell motility of four CD9 knockdown clones compared with control clone. Cleared areas on gold sol-coated plate produced by cell movement were calculated by Scion image program and indicated as square pixels as described under "Materials and Methods." (Fig. 5C). Phagokinetic motility of KK47 was greatly enhanced following depletion of GM3 with P4 (Fig. 5D).

Phenotypic Reversion of High Motility YTS1 Cells to Low Motility Variant by Exogenous GM3 Addition
GM3 synthase gene transfection to YTS1 did not sufficiently increase the GM3 level, because GM3 could be converted to GM2 in these cells (data not shown). We therefore tried exogenous addition of GM3 in the culture medium to observe possible effects on ␣3-to-CD9 interaction and cell motility. Incubation of YTS1 with GM3 caused a significant increase of GM3 level, as determined by flow cytometry (Fig. 6A, left). Increased GM3 levels caused enhanced co-IP with ␣3 and CD9, i.e. more CD9 was co-immunoprecipitated with ␣3 in PNF from GM3preincubated YTS1 than from control nontreated cells, as detected by Western blot analysis (Fig. 6B, left) and densitometry of co-immunoprecipitated bands (Fig. 6B, right). Exogenous addition of GM3 also enhanced co-localization of ␣3 and CD9 in YTS1, as indicated by the enhanced merge image observed by confocal microscopy (Fig. 6C), whereby significant reduction of phagokinetic cell motility was observed, particularly when cells were incubated with 20 or 50 M GM3 (Fig. 6D). GM2 is present in significant amount in YTS1, so exogenous GM2 addition experiment was not relevant.
GM1 is absent in YTS1, and exogenous GM1 is incorporated efficiently at the cell surface, as revealed with FITC-labeled cholera toxin subunit B (Fig. 6A, right). However, GM1 addition did not significantly affect ␣3/CD9 interaction as determined by co-IP (Fig. 6B, left and right) or by confocal microscopy (Fig. 6C). Motility of 50 M GM1-added cells was much higher than that of 50 M GM3-added cells but was not significantly different from that of control cells (Fig. 6E).

Expression Level of c-Src, Activated c-Src, and Csk in YTS1 Versus KK47 Cells
c-Src level in PNF was lower in KK47 than in YTS1, and this difference was greater in low density fractions 4 and 5. The activated form of  c-Src, with phosphorylation at Tyr-416, was clearly present in PNF from YTS1 and nearly absent in PNF from KK47. In contrast, the inactive form of c-Src, with phosphorylation at Tyr-527, showed similar levels in PNF of YTS1 and KK47. These comparisons were based on the same level of loading control ␥-tubulin (Fig. 7A). The distribution pattern of c-Src and Csk in low density fractions 4 -6 and high density fractions 10 -12 from PNF, separated by sucrose density gradient centrifugation, was determined by Western blot. c-Src level was much higher in fractions 4 -6 of YTS1 compared with KK47. Csk was virtually absent in fractions 4 -6 of YTS1 but was clearly present in fraction 5 of KK47 (Fig.  7B). c-Src and Csk were not detected in fractions 1-3 or fractions 7-9 from either YTS1 or KK47 (data not shown).

Effect of GM3 Level on c-Src Activation Status, Csk Translocation, and Tumor Cell Motility
The results shown in Fig. 7 suggest that the low GM3 level activates c-Src in YTS1 cells, whereas the high GM3 level inactivates c-Src in KK47 cells. We examined this possibility by depletion of GM3 with P4 in KK47 cells and by exogenous GM3 addition to YTS1 cells.
Low GM3 Level Activates c-Src-KK47 cells were incubated with 1 M P4 for 72 h, followed by Western blot for detection of c-Src phosphorylation status and Csk expression in PNF. P4 treatment caused increased Tyr-416 phosphorylation and decreased Tyr-527 phosphorylation, whereas total c-Src and Csk levels did not change (Fig. 8A, left). The ratio of phosphorylation level (Tyr-416 divided by Tyr-527) was significantly higher in P4-treated cells (Fig. 8A, right). Enhanced GM3 Level in YTS1 Inhibits c-Src, in Association with Translocation of Csk to GEM (Fractions 4 -6)-Exogenous addition of GM3 to YTS1 did not cause significant change of c-Src phosphorylation status or Csk in PNF (Fig. 8B, left). However, c-Src with Tyr-527 phosphorylation increased in GEM (fractions 4 -6) with GM3 incubation (maximum at 3 h), with clear translocation of Csk to GEM (fractions 4 -6). This was closely associated with a decrease of activated c-Src with Tyr-416 phosphorylation after prolonged incubation with GM3 (Fig.  8B, middle). Phosphorylation pattern in c-Src and the Csk status were essentially unchanged in high density fractions 10 -12 of YTS1 cells (Fig.  8B, right).
Effect of Src Inhibitor "PP2" on YTS1 Cell Motility-The possibility that c-Src activity is correlated with GM3-mediated changes of cell motility in YTS1 was further assessed by using Src tyrosine kinase inhibitor PP2. Motility of YTS1, associated with high c-Src activity, was decreased significantly in a dose-dependent manner in PP2-incubated cells (Fig. 8C).

DISCUSSION
Cell adhesion and motility are controlled by types of integrin receptor expressed at the cell surface, in combination with various cytoplasmic signaling molecules, some of which connect to cytoskeletal components (35).   Cells were incubated with GM3 for 1, 3, or 16 h, harvested, and separated into PNF, low density fractions 4 -6 and high density fractions 10 -12, as described under "Materials and Methods." Aliquots of PNF and low and high density fractions were analyzed by Western blot. Note that Csk translocation to fractions 4 -6 was maximal at 3 h after GM3 addition, associated with maximal phosphorylation at Tyr-527. Thereafter, activated c-Src signal (Tyr-416 phosphorylation) declined significantly. C, inhibitory effect of Src family kinase inhibitor PP2 on YTS1 cell motility. Phagokinetic motility of cells pretreated with PP2 at 5, 10, or 20 M concentration was compared with that of nontreated cells. Motility was decreased by PP2 in a dose-dependent manner, confirming involvement of c-Src activation in enhanced YTS1 cell motility. review see Ref. 9); and (iii) ganglioside association that may mediate integrin interaction with TSP (41,42).
TSPs CD9 and CD82 were originally identified as cell motility inhibitory factors highly expressed in normal cells or nonmetastatic tumor cells but down-regulated in metastatic deposits (43)(44)(45). Decreased expression of CD9 (46) and CD82 (47) in urothelial cancer was recently reported to be associated with recurrence and increased metastasis. However, detailed study on CD9 indicates that CD9 alone does not inhibit tumor cell motility or invasiveness; rather, such inhibition requires GM3 and complete N-glycosylation, including that of integrin. This concept arose from studies (5) with mutant ldlD cells defective in UDP-Gal 4-epimerase, in which such glycosylation does not occur unless galactose is added to growth medium (48). CD9 was found to be soluble in C/M, to display properties of PLPs (49,50), and to be closely associated with GSLs (6,51). Studies with various colorectal cancer cell lines expressing different levels of GM3 and CD9, aided by application of photoactivable GM3, showed that GM3 is a co-factor of CD9 for inhibition of tumor cell motility (21).
Several lines of studies, including those above, indicate that microdomains controlling glycosylation-dependent adhesion and concurrent signal transduction are cholesterol-independent, soluble in 1% Triton X-100, and provide adhesion sites. These properties of GM3/TSP/integrin microdomains are distinguishable from those of moving, signaling platforms, "rafts," that do not contain TSP/integrin, are not soluble in Triton X-100, and are not involved in cell adhesion but display clear cholesterol-dependent signaling function. We therefore applied the term glycosynapse to such glycosyl microdomains (8,13,52), in analogy to "immunological synapse" (14), the microdomain involved in immunocyte adhesion and antigen presentation to the T-cell receptor. Among various glycosynapses, those having the GM3-TSP-integrin complex are termed glycosynapse 3 (8).
Thus, the "classic" concept of integrin-dependent motility and invasiveness of tumor cells is now re-formulated in terms of a particular interaction and organization of integrin with PLP/TSP and GM3 in glycosynapse 3. This concept has been extended to phenotypic conversion or reversion induced by deletion or addition of a single component in glycosynapse 3, as described in the present study.
Phenotypic conversion from low motility bladder cancer cells, KK47, to high motility cells was caused by knockdown of CD9 by RNAi, whereby association of ␣3 with CD9 was disrupted. Similarly, depletion of GM3 by P4 treatment of KK47 gave rise to a high motility variant characterized by dissociation of ␣3 from CD9. The level of ceramide, a well established, versatile signaling molecule (53,54), was not significantly changed in P4-treated KK47, similarly to the case of WI38 cells (17), Madin-Darby canine kidney cells (18), and 3T3 cells (32). Thus, the conversion of KK47 to high motility variant by P4 treatment was because of GM3 depletion and not to ceramide change.
In contrast, phenotypic reversion from high motility bladder cancer cells, YTS1, to low motility cells was caused by exogenous addition of GM3, whereby a significant increase of ␣3-to-CD9 interaction and a reduction of cell motility occurred. We attempted to enhance GM3 expression in YTS1 by transfection of the GM3 synthase gene but without success, presumably because GM3 produced in this way is immediately converted to GM2 or other higher gangliosides and is not accumulated. These findings indicate that GM3 mediates and stabilizes ␣3-to-CD9 interaction in glycosynapse 3. In addition, high GM3 level in KK47 and/or low GM3 level in YTS1 affect not only the interaction of ␣3 with CD9 but also the c-Src activation state in glycosynapse. High GM3 level inhibits c-Src activation because of translocation of Csk, which promotes Tyr-527 phosphorylation in c-Src (55). In contrast, low GM3 level in YTS1 causes c-Src activation, with enhanced Tyr-416 phosphorylation. This process is promoted by Csk translocation out of glycosynapse 3 in YTS1. Exogenous GM3 addition to YTS1 promotes Csk translocation into glycosynapse, resulting in inhibition of c-Src activation. GM3 effect on YTS1 cannot be replaced by GM2 or GM1, because a significant amount of GM2 pre-exists, and exogenous GM1 addition had no effect on ␣3/CD9 interaction or on cell motility.
Reversion and conversion of phenotype can be manipulated by GM3 addition or depletion via the well organized, stable framework versus the disorganized, unstable framework of glycosynapse 3. In addition, the high GM3 level in KK47 may inhibit c-Src activation because of the presence of Csk, which promotes phosphorylation of Tyr-527 in c-Src (55). In contrast, low GM3 level in YTS1 causes c-Src activation, with phosphorylation at Tyr-416. This process is promoted by the absence of Csk in glycosynapse 3 of YTS1. In our previous study of human fibroblast WI38, GM3 was implicated as an inhibitor of signal transduction, because various signaling processes initiated by c-Src activation following Akt/mitogen-activated protein kinase were promoted by P4-induced depletion of GM3 (17). Effects of GSLs on signal transduction and cellular functions, e.g. cell cycle arrest (31), and phospholipase C activation with bradykinin stimulation (56), have been studied using the GlcCer synthase inhibitors PDMP (D-threo-1-phenyl-2-decannoylamino-3-morpholino-1-propanol) (31,56) and, more recently, P4 (20).
The survival signals phosphatidylinositol 3-kinase, Akt, and Rac, occurring through activation of ␣3␤1 integrin, induce adhesion-mediated cell growth and motility (3). Such signaling may be blocked when the GM3-CD9-integrin complex is formed, as in glycosynapse 3. Reversion from malignant to benign phenotype induced by the high GM3 level is based on not only stabilized glycosynapse framework but also inhibition of c-Src activation, which causes reduced cell motility, growth, and invasiveness. Similarly, the low GM3 level not only disorganizes glycosynapse framework but also induces c-Src activation, with consequent enhancement of cell motility, growth, and invasiveness. However, the c-Src activation status may not control the GM3 level, because the low GM3 level in YTS1, which shows high c-Src activation, is not increased by treatment of cells with c-Src inhibitor PP2 (data not shown).
It is not yet clear how the GM3 level in glycosynapse 3 is controlled. A mechanism involving the GM3 synthase promoter region, which is affected by the transformation process, must be crucial. Down-regulation or depletion of GM3 is often observed associated with oncogenic transformation, e.g. in polyoma virus-transformed baby hamster kidney cells (57), Rous sarcoma virus-transformed chicken embryonic fibroblasts (58), and more recently v-Jun-transformed chicken and mouse fibroblasts, in which reversion of oncogenic to normal phenotype is caused by enhanced GM3 synthesis through GM3 synthase gene transfection (22). In the latter case, a possible role of GM3 in glycosynapse was suggested.
Current trends in cell biology and molecular oncology are based on genomic or transcriptomic analysis by microarray assay. If this approach were performed on either KK47, YTS1, or various breast cancer cell lines, activation or inhibition of many genes would be detected. However, such changes cannot easily identify a single or a few crucial molecules, such as GM3, CD9, integrin, or their organization in glycosynapse, involved in defining oncogenic transformation or its reversion. Obviously, GM3, CD9, ␣3, and ␤1 are neither oncogene nor anti-oncogene products. Therefore, genomic or transcriptomic analysis by itself has limited usefulness in molecular cell biology and oncology research. We are facing the task to look for the essential mechanism defining organization of a few crucial molecules in glycosynapse on one the hand