Overexpressed GM1 Suppresses Nerve Growth Factor (NGF) Signals by Modulating the Intracellular Localization of NGF Receptors and Membrane Fluidity in PC12 Cells*

Ganglioside GM1 has been considered to have a neurotrophic factor-like activity. To analyze the effects of endogenously generated GM1, the rat pheochromocytoma cell line PC12 was transfected with the GM1/GD1b/ GA1 synthase gene and showed increased expression levels of GM1. To our surprise, GM1 (cid:1) -transfectant cells (GM1 (cid:1) cells) showed no neurite formation after stimulation with nerve growth factor (NGF). Autophosphoryl-ation of NGF receptor TrkA and activation of ERK1/2 after NGF treatment were scarcely detected in GM1 (cid:1) cells. Binding of 125 I-NGF to PC12 cells was almost equivalent between GM1 (cid:1) cells and controls. However, dimer formation of TrkA upon NGF treatment was markedly suppressed in GM1 (cid:1) cells in both cross-link-ing analysis with Bis(sulfosuccinimidyl)suberate 3 and 125 I-NGF binding assay. The sucrose density gradient

Gangliosides, sialic acid-containing glycosphingolipids, are thought to play important roles in the development and function of the nervous system, because they accumulate in brain tissues of vertebrates, and their profiles of carbohydrate moiety alter with development (1,2). Recently, a number of glycosyltransferase genes have been isolated, and studies of these genes have shown that the various expression patterns of gan-gliosides are determined basically by the combination of activated glycosyltransferase genes (3). Among complex gangliosides, GM1 1,2 has been most rigorously studied, because it is one of the major gangliosides in vertebrate brain (2) and shows specific binding with the cholera toxin B subunit resulting in important biological events such as cAMP response (4). Since GM1 synthase cDNA was isolated by us (5), the mRNA expression of the gene has been directly examined, and a high expression level in the rat fetal brain has been demonstrated.
Recently, membrane microdomains such as glycolipid-enriched microdomains, detergent-insoluble microdomains, or lipid rafts have been thought of as sites for signal transduction as well as for endocytosis and cholesterol turnover on the cell membrane (6). They are enriched in cholesterol, glycosylphosphatidylinositol-anchored proteins, sphingomyelin, glycosphingolipids, and various signaling molecules such as growth factor receptors, G-proteins, and Src family tyrosine kinases (7). Although ganglioside GM1 has been used as a mere marker of rafts, we demonstrated that GM1 might regulate the signal magnitude of PDGF/PDGFR by altering the intracellular localization of PDGF receptor (R) in Swiss3T3 cells (8). There have also been a number of reports indicating the effects of glycosphingolipids on the growth/differentiation signals (9). These reports suggested that GM1 might affect the structure/function of lipid rafts in neuronal cells, resulting in the regulation of differentiation/proliferation signals.
In the present study, we established stable transfectant lines of PC12 with the GM1/GD1b/GA1 synthase cDNA, which we had isolated previously (5), and analyzed the response to NGF. To our surprise, marked alterations in the cell response to the induced differentiation and in the activation of signaling molecules were found in GM1 ϩ cells. We also demonstrated dramatic changes in the intracellular localization of NGF receptors and relevant molecules, i.e. the majority of TrkA, p75 NTR , and Ras moved from the raft to the non-raft fraction in GM1 ϩ cells, whereas the raft markers such as GM1 and flotillin persistently stayed in the raft fraction. These results suggested that GM1 plays critical roles in the regulation of the physicochemical nature of lipid raft and of the bio-signals to determine the cell fates.
Recombinant DNA-Human ␤1,3-galactosyltransferase cDNA clone pM1T-9 (5) was digested with XhoI-XbaI and inserted into the XhoI-XbaI site of pMIKneo to obtain pMIKneo/M1T-9. pMIKneo is a mammalian expression vector with the SR␣ promoter and was generously presented by Dr. K. Maruyama (Tokyo Medical and Dental University).
Cell Culture and Transfection-PC12 cells were maintained in RPMI 1640 medium supplemented with 10% horse serum and 5% fetal calf serum, at 37°C in a humidified atmosphere containing 5% CO 2 as described previously (14). PC12 cells used for cDNA transfection were plated in a 60-mm plastic tissue culture plate (Falcon) at a density of 7 ϫ 10 5 cells/4 ml/plate. The plasmid pMIKneo/M1T-9 (4 g) was transfected into cells with LipofectAMINE (Invitrogen, Rockville, MD) according to the manufacturer's instructions. Stable transfectant cells were selected in the presence of 250 g/ml G418 (Invitrogen) and maintained continuously in the presence of G418 (200 g/ml).
Neurite Outgrowth Assay-PC12 cells and the transfectants were seeded at 1 ϫ 10 4 cells/well in a type I collagen-coated 48-well culture plate (Falcon) and were either treated with 100 ng/ml 2.5S NGF (Alamone Laboratories) in serum-free medium or left untreated in the serumcontaining medium. Neurite-bearing cells were counted for the following 4 days. To analyze the effect of exogenous GM1 on neurite outgrowth, PC12 cells were treated with either serum-free medium or medium containing 100 ng/ml NGF, 5 ng/ml NGF, 50 M GM1 (Sigma), or 5 ng/ml NGF with 50 M GM1 (preincubated with 50 M GM1 for 12 h at 37°C, then treated with 5 ng/ml NGF). The percentage of cells bearing neurites of various lengths was counted from days 1 to 4.
MTT Assay-3 ϫ 10 3 cells were seeded with serum-containing medium in 98-well plates. At 0 -5 day of culture, MTT assay was performed as described (8). Growth of cells was quantified by assessing the reduction of MTT to formazan, measured as the absorbance at 590 nm using a plate reader ImmunoMini NJ-2300 (System Instrument, Tokyo, Japan).
Electrophoresis and Immunoblotting-PC12 cells and transfectants were plated in 60-mm tissue culture plates in serum-containing medium, then cells were treated as described below. For the assay of mitogen-activated protein kinase (MAPK) and TrkA phosphorylation, the medium was replaced with serum-free medium and incubated for 1 h. Then cells were incubated with 100 ng/ml 2.5S NGF for 0, 5, 15, 30, 60, and 120 min at 37°C. After each treatment, the medium was removed, and cells were washed three times with 3 ml of serum-free medium, then solubilized in 200 l of lysis buffer (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM MgCl 2 , 0.5% Nonidet P-40, 1 mM NaVO 4 , 1 mM phenylmethylsulfonyl fluoride, 20 units/ml aprotinin). Lysed cells were transferred into microcentrifuge tubes, centrifuged at 3,000 rpm for 10 min, and the supernatant was collected to remove nuclei.
Lysates were separated with SDS-PAGE using 7-12% gels. The separated proteins were transferred onto an Immobilon-P (polyvinylidene difluoride) membrane (Millipore). Blots were blocked with 10% skim milk in phosphate-buffered saline (PBS) for 1 h. The membrane was first probed for 1 h with primary antibodies at the dilution suggested by the suppliers. After washing three times with PBST (0.05% Tween 20 in PBS), the blots were then incubated for 1 h with goat anti-rabbit IgGs or goat anti-mouse IgGs conjugated with horseradish peroxidase (1:4000). After the membranes were washed three times with PBST, bound conjugates were visualized with an ECL detection system (PerkinElmer Life Science). For detection of GM1, the membranes were probed with CTB-biotin (1:500) and detected with an ABC kit (Vector Laboratories). To analyze the effects of the reduced GT1b expression, the transfectant cells were cultured with exogenous GT1b (50 M) and used for immunoblotting as described above after confirmation of GT1b incorporation on the cell surface with flow cytometry.
TrkA Receptor Cross-linking-Cells were plated at a density of 6 ϫ 10 6 /10-cm dish, and serum-starved for 1 h before NGF treatment. Then, cells were washed with plain medium without serum and treated with NGF (50 ng/ml) for 5 min at 37°C. The medium was removed, and cells were washed twice with ice-cold PBS, and then cross-linked using 1 mM BS3 in a cross-linking buffer (25 mM HEPES, pH 8.5, 120 mM NaCl, 6 mM KCl, 1 mM MgCl 2 , 10 mM EGTA) at 4°C for 1 h. The cross-linking reaction was terminated by adding 1 M Tris-HCl (pH 7.4) to a final concentration of 0.1 M. Then, cells were washed twice in Tris-buffered saline and lysed in the lysis buffer as described above. TrkA was immunoprecipitated, and immunoprecipitates were separated on a 7% gel with an acrylamide/bisacrylamide ratio of 200:1. Samples were transferred onto membranes and immunoblotted with an anti-Trk antibody (B-3).
Isolation of Raft Fraction-Microdomain rafts were prepared using a non-detergent extraction method essentially as described by Song et al. (15). Cells were plated at a density of 2 ϫ 10 7 /15-cm dish and cultured up to 90% confluency, and five dishes of cells were used for each preparation. After being washed twice with ice-cold PBS, the cells were scraped in 1 ml of 0.5 M sodium carbonate buffer, pH 11.0. The cells were homogenized sequentially using a loose fitting Dounce homogenizer (10 strokes), a Polytron tissue grinder (three 10-s bursts), and a sonicator (three 20-s bursts). All procedures were carried out at 4°C. The homogenate (1 ml) was then adjusted to 45% (w/v) sucrose by adding 1 ml of 90% (w/v) sucrose prepared in 2ϫ MNE buffer (25 mM Mes, pH 6.5, 150 mM NaCl, 5 mM EDTA). The final pH of the mixture was 10.2. A discontinuous sucrose gradient was formed by overlaying 2 ml of 35% (w/v) sucrose onto the mixture, and then 1 ml of 5% (w/v) sucrose was overlaid. Both of these layers were prepared with MNE containing 0.25 M sodium carbonate. The samples were centrifuged at 20,000 ϫ g in an SW50.1 rotor for 16 h at 4°C. From the top of the gradient, 0.5 ml of each fraction was collected to yield 10 fractions. The components in each fraction were concentrated by centrifugation at 100,000 ϫ g for 2 h at 4°C in MNE buffer, and precipitates were resolved in the lysis buffer and used for immunoprecipitation or Western immunoblotting.
NGF-TrkA Cross-linking-Cells were resuspended in a binding buffer (Hanks' containing 1 mg/ml each of glucose and bovine serum albumin) at 5 ϫ 10 6 cells/ml and incubated at 4°C for 1 h with 1 nM 125 I-NGF (2750 dpm). To correct the nonspecific binding, unlabeled NGF (1 M) was added during the binding. To cross-link NGF with TrkA, BS3 was added to a final concentration of 1 mM to the reaction mixture, and the mixture was incubated at 4°C for 1 h. After the cross-linking reaction was terminated, cells were washed and lysed in the lysis buffer. The lysates were used for immunoprecipitation with an anti-Trk antibody, and the immunoprecipitates were separated using a 7% gel as described above. 125 I-NGF-bound proteins were detected with autoradiography using an imaging plate and BAS2000 Bioimage Analyzer TM (Fujifilm, Tokyo).
NGF Binding Assay-Cells were resuspended in a binding buffer at 1 ϫ 10 5 cells/0.2 ml and incubated at 37°C for 45 min with various concentrations (5 pM to 10 nM) of 125 I-NGF. The reaction mixture was overlaid onto 5% bovine serum albumin in 0.32 M sucrose and was centrifuged at 10,000 ϫ g for 1 min at 4°C. The supernatants were collected, and the tips of tubes containing the cell pellets were cut off. A 1,000-fold excess of unlabeled NGF was used to assess the nonspecific binding. The radioactivity was counted in a ␥-counter, and the results were analyzed on the basis of Scatchard plot.
In Vitro Kinase Assay-The PC12 lysate was used for immunoprecipitation with an anti-TrkA antibody as described above. Immunoprecipitates were washed twice with the lysis buffer and twice with the Tris-buffered saline containing a phosphatase inhibitor (100 M Na 3 VO 4 ). Immunoprecipitates were resuspended in a tyrosine kinase assay buffer (50 mM HEPES, pH 7.4, 20 mM MnCl 2 , 5 mM MgCl 2 , 1 mM dithiothreitol, 100 M Na 3 VO 4 , 5 Ci/sample [␥-32 P]ATP) and then incubated for 10 min at 30°C in the presence of various concentrations of GM1. The kinase reaction was terminated by adding Laemmli sample buffer followed by boiling. Relative kinase activity was measured by scanning the bands in the autoradiogram as described above. The effects of GM3 or GT1b were also examined for comparison.
Immunofluorescence Study-Cells were cultured on a poly-L-lysinecoated coverslip, and cells were fixed with 4% paraformaldehyde in PBS. Then, cells were treated with CTB-Alexa555 (10 g/ml) in PBS for 1 h at 4°C. The fixed cells were analyzed with a confocal laser microscope Fluoview FV500 TM (Olympus, Tokyo).
Rho Kinase Assay-To determine the alteration in the activation levels of Rho, activated Rho was isolated and analyzed with the method The expression levels of gangliosides were indicated in the abscissa with logarithm. V1 and V2 are vector control cells transfected with a vector alone, and M3 and M6 are GM1 ϩ cells. C, to compare the distribution pattern of GM1 in the cells, we used CTB-Alexa555 for staining cell surface GM1. Whereas GM1 showed the patched distribution in vector control cells, it was distributed throughout the plasma membrane with much stronger intensity in GM1 ϩ cells. The original magnification is ϫ400, and two panels at the bottom are 2.5-fold more magnified.
of Ren et al. (16) using a Rho Activation Assay kit (Cytoskeleton, Denver, CO) according to the manufacturer's instruction. Cells were plated at a density of 6 ϫ 10 6 /10 cm dish. After being washed, cells were lysed with lysis buffer, and the cell lysates were centrifuged at 10,000 ϫ g for 5 min at 4°C, and the supernatants were incubated with Rhotekin-RBD beads under rotation for 1 h at 4°C. The beads were washed three times with the lysis buffer, and the precipitates were solubilized in a Laemmli sample buffer, heated for 3 min at 98°C, and then separated in SDS-PAGE. Bound Rho proteins were detected by Western immunoblotting using an anti-Rho polyclonal antibody in the kit.
Measurement of the Membrane Fluidity by Fluorescence Recovery after Photobleaching-Cells were cultured on a glass bottom dish coated with poly-D-lysine, and the medium was replaced with a standard external solution (10 mM Hepes, 150 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose) prewarmed at 37°C. A fluorescent probe, DiI (DiIC 18 ; D-282, Molecular probes, Eugene OR) was added to the standard external solution at 2.5 g/ml and incubated for 10 min, to be incorporated into the cell membranes. FRAP analyses were performed with a confocal laser scanning microscope (Olympus) equipped with a stage heater at 37°C. A small area of the labeled membrane was photobleached by full laser power (100%), resulting in the immediate reduction of the intensity of fluorescence. A selected area (3-m square) on the labeled cell membrane was photobleached for 8 -15 s by a 488-nm laser beam. Then, the fluorescence recovery in the photobleached area was immediately recorded using a time-lapse option of the system until the fluorescence was recovered at a plateau level. Relative recovery in fluorescence intensity was calculated for 60 s after photobleaching and analyzed by Microsoft Excel. FRAP was calculated by a ratio of postbleach intensity/pre-bleach intensity, and the membrane fluidity was evaluated by comparing the percentage of FRAP to a plateau as follows. Fluorescence intensities at a pre-photobleach state and the post-photobleach intensities at a plateau were expressed as average values of 10 scans. In each experiment, the value obtained at pre-photobleach was taken as 100% and that at just after photobleach was taken as 0% to compensate for the differences among experiments. The half-maximal recovery of fluorescence (t1 ⁄2 ) was also calculated on the basis of the duration (t) of fluorescence recovery reaching a plateau.

Expression of Gangliosides in GM1 Synthase
Gene Transfectant Cells (GM1 ϩ Cells)-After the transfection of PC12 cells with a GM1 synthase gene expression vector (pMIKneo/⌴1⌻-9), or pMIKneo, two transfectant lines (M1 and M3) and two vector controls (V1 and V2) were established, respectively. Expression profiles of gangliosides were examined by flow cytometry (Fig. 1A). The mean fluorescence values of individual gangliosides obtained by subtracting those of corresponding controls are shown in Fig. 1B. Among the gangliosides examined, expression levels of GM1 and fucosyl-GM1 markedly increased in the transfectants (M3 and M6). Expression levels of other gangliosides showed no apparent change compared with those in the vector control lines (V1 and V2) except that GT1b expression level reduced after the transfection. To compare the distribution pattern of GM1 on the cell membrane, we stained the cultured cells with CTB-Alexa555. Whereas GM1 in vector control cells showed a weak and patched distribution, those in the GM1 ϩ cells showed very strong staining and a uniform and thick distribution throughout the plasma membrane (Fig. 1C).
No Neurite Extension with NGF in PC12 Cells Transfected with GM1 Synthase cDNA-To examine the effects of GM1 expressed endogenously on NGF-induced differentiation, the GM1 ϩ cells and vector controls were cultured with or without 50 ng/ml NGF for 5 days. Fig. 2 shows the morphological change in controls and GM1 ϩ cells after 5 days of NGF treatment. As reported previously, cell proliferation was reduced and neurite formation was observed in vector control cells ( Fig.  2A). In contrast, the GM1 ϩ cells showed no neurite extension, although cell proliferation was not affected (Fig. 2, A and B). Quantitative analysis of the neurite formation revealed that only 10 -20% of GM1 ϩ cells showed neurites longer than twice the cell diameter even on day 4 of culture (Fig. 2C). On the other hand, 80% of control cells showed neurites longer than twice the diameter. Although the vector controls showed little neurite outgrowth when treated with low density of NGF (5 ng/ml), the pretreatment with GM1 (50 M) enhanced the outgrowth as previously reported (12, 13) (data not shown). In contrast, GM1 ϩ cells showed no neurite outgrowth even when stimulated with a high concentration of NGF (100 ng/ml) (data not shown).
Reduced Phosphorylation of TrkA and ERKs with NGF Treatment-It has been known that exogenously added GM1 enhanced NGF-induced neurite extension. To determine the effects of the overexpressed GM1 on NGF-derived signals in the GM1 ϩ cells, the time course of tyrosine phosphorylation of Trk after NGF treatment was investigated. Tyrosine phosphorylation of TrkA was detected at 5 min after NGF treatment and gradually reduced in the control cells (Fig. 3A). On the other hand, tyrosine phosphorylation of TrkA was scarcely detected in the GM1 ϩ cells. As for the downstream ERK1/2 phosphorylation, only weak phosphorylation was found at 5 min after the NGF stimulation in the GM1 ϩ cells correspondingly with the reduced phosphorylation of TrkA (Fig. 3B). In the control cells, the phosphorylation of ERK1/2 was clearly detected at 5 min after NGF treatment with subsequent slow reduction. Immu-noblots with anti-ERK1/2 antibody demonstrated that applied lysates contained equivalent amounts of ERK proteins (Fig. 3B, bottom panels). To examine the effects of the reduced GT1b expression level in the GM1 ϩ cells, the transfectant cells were cultured in the presence of exogenous GT1b and were analyzed for the phosphorylation of ERK1/2. GT1b was well incorporated on the membrane (Fig. 3C). The phosphorylation patterns of ERK1/2 after NGF stimulation showed no change after GT1b incorporation, suggesting that the reduced NGF signals in the GM1 ϩ cells were not due to the loss of GT1b expression.
NGF Binding Was Equivalent between the Controls and GM1 ϩ Cells-Using gradually diluted 125 I-labeled NGF, NGF binding to the controls and to the GM1 ϩ cells was investigated as described under "Experimental Procedures." Nonspecific binding of 125 I-labeled NGF was determined by adding excess amounts (ϫ1000) of cold NGF, and the values were used to subtract from the individual counts. The 125 I-labeled NGF specifically bound to PC12 cells in a dose-dependent manner. The Scatchard plots indicated that PC12 cells express two kinds of NGF receptors, i.e. a high affinity receptor and a low affinity receptor, as previously reported. B max values calculated from the high affinity binding curves were 2.52 ϫ 10 2 pmol/10 5 cells and 2.62 ϫ 10 2 pmol/10 5 cells for V1 and V2, respectively. Those for M3 and M6 were 2.08 ϫ 10 2 pmol/10 5 cells and 2.44 ϫ 10 2 pmol/10 5 cells, respectively (Fig. 4). In addition, the dissociation constants (K d values) of NGF were 0.55 and 0.53 nM for V1 and V2, respectively, and K d values for M3 and M6 were 0.46 and 0.55 nM, respectively (Fig. 4). There were no significant differences in the binding levels and kinetics between the

FIG. 3. Reduced phosphorylation of TrkA and ERK1/2 after NGF treatment in the transfectant cells.
A, time courses of the phosphorylation levels of TrkA in the vector controls (V1 and V2) and the GM1 ϩ cells (M3 and M6) were analyzed. Cells were treated with NGF (50 ng/ml) for the times indicated. TrkA was immunoprecipitated with an anti-Trk antibody (C-14), and immunoblotting was carried out using PY-20 (upper) or anti-TrkA antibody (bottom). B, activation of ERK1/2. Cells were treated as described in A and were then lysed with Nonidet P-40 lysis buffer, and total lysates were separated by 12% SDS-PAGE. Western blotting was carried out using an anti-phosphorylated ERK1/2 antibody (upper) or an anti-ERK1/2 (lower). A similar experiment was repeated, and essentially similar results were obtained. C, GM1 synthase gene transfectants were treated with GT1b (50 M) for 4 h, and the cell surface GT1b was detected with flow cytometry (left). Then, phosphorylation of ERK1/2 after NGF treatment was examined as described above (right). Note that there are no differences between M3 and M6 in B and those in C.

FIG. 4. NGF binding was equivalent between the controls and GM1 ؉ cells.
Using gradually diluted 125 I-labeled NGF, NGF binding to the vector controls and GM1 ϩ cells were investigated. Two vector control cells and two GM1 ϩ cells were analyzed. Nonspecific binding was obtained by adding 1000-fold excess amounts of unlabeled NGF, and the values were used for the subtraction. Scatchard plot analysis was performed, and the results of individual samples are presented. Insets are the binding curves, Note that there is no significant difference in the binding kinetics between the controls and GM1 ϩ cells. B max values calculated from the high affinity binding curves are expressed with the unit (ϫ 10 2 pmol/10 5 cells) in the individual figures. control cells and the GM1 ϩ cells, suggesting that overexpressed GM1 did not disturb the NGF binding itself.
Reduced Dimerization of TrkA with NGF Treatment in the GM1 ϩ Cells-The GM1 ϩ cells showed neither neurite extension nor signal activation after NGF treatment. Then, TrkA dimerization after NGF treatment was investigated. The cell surface proteins were cross-linked with BS3 under the NGF treatment as described under "Experimental Procedures." After cross-linking, TrkA was immunoprecipitated from the cell lysates, and separated in a 7% gel followed by immunoblotting using anti-TrkA antibody (B-3). The immunoblotting showed high molecular mass bands (about 300 kDa) corresponding to TrkA dimer at 5 min after NGF treatment in control cells. On the other hand, the dimer bands of TrkA were scarcely found in the GM1 ϩ cells (Fig. 5).
Reduced Complex Formation between TrkA Dimer and NGF in the Transfectants-Based on the reduction in the phosphorylation and dimerization of TrkA in the GM1 ϩ cells, we questioned whether NGF binding to TrkA is attenuated by overexpression of GM1. The simple binding assay showed no differences as shown in Fig. 4. Then, the dimer formation was markedly reduced in GM1 ϩ cells (Fig. 5). Therefore, we examined NGF binding to TrkA monomer and dimer with crosslinking experiment. Cells were incubated with 125 I-NGF (1 nM) at 4°C for 1 h, and then with a cross-linker BS3 for 1 h. TrkA was immunoprecipitated and analyzed by SDS-PAGE and autoradiography. Radioactive bands, corresponding to the 125 I-NGF-bound TrkA monomer (ϳ150 kDa) and 125 I-NGF-bound dimer (ϳ300 kDa), were found in vector control cells (Fig. 6A).  1 and 3), whereas mainly monomer form of TrkA was detected in GM1 ϩ cells. B, an evidence for the binding specificity. When binding was carried out in the presence of a 1000-fold excess of unlabeled NGF (indicated with "ϩ"), the bands completely disappeared, suggesting that NGF specifically bound to TrkA.

FIG. 7. Alteration in the flotation in a sucrose gradient of NGF receptors in the GM1 ؉ cells.
Cells were lysed under detergent-free conditions, and the extracts were fractionated. Fractions were subjected to immunoblotting analysis using the antibodies against the proteins indicated. GM1 was detected with CTB. A, the results of vector control cells. B, the results of GM1 ϩ cells. Note an increased ratio of TrkA and P75 NTR in non-raft fractions (6 -10) in the GM1 ϩ cells.
In contrast, dimer bands were hardly detected in GM1 ϩ cells, whereas the band intensities of the monomer were equivalent with those in the controls. These findings indicated that the monomer form of TrkA in the GM1 ϩ cells binds with NGF at the similar level with that in the control cells, although the transfectant cells could not form TrkA dimer or could not bind with NGF, if present. Taken together with the results in Fig. 5, the former case appeared more likely. When binding was carried out in the presence of a 1000-fold excess of unlabeled NGF (Fig. 6B), all bands completely disappeared, suggesting that NGF specifically bound to TrkA.

Alteration in the Flotation in a Sucrose Gradient of NGF Receptors in GM1 Synthase Gene Transfectant Cells-Because
GM1 has been considered to be localized in lipid rafts, overexpression of GM1 may affect the nature of rafts. Then, we examined the effects of GM1 overexpression on the intracellular localization of NGF receptors (TrkA and p75 NTR ) by isolating lipid rafts with a detergent-free method. Ten fractions from the discontinuous sucrose gradient were prepared and analyzed for distribution of p75 NTR , TrkA, and raft markers, such as flotillin. The distribution of Ras as well as GM1 was also analyzed. Immunoblot analysis showed that most of flotillin, Ras, and GM1 were found in fractions 3 and 4 containing the raft fraction (Fig. 7A). The majority of TrkA and p75 NTR were also detected in this raft fraction. These results indicated that PC12 cells contained the lipid raft in buoyant density, GM1 content, and protein constituents. To our surprise, the distribution pattern of TrkA and P75 NTR dramatically changed in the GM1 ϩ cells (Fig. 7B). The main portions of TrkA and p75 NTR were detected in the non-raft fraction comprising fractions 6 -10, suggesting that the distribution of NGF receptors are strongly affected with the expression levels of GM1, probably based on the modification of the properties of GEM/rafts. The Ras protein also moved in part from the raft to the non-raft fraction, whereas GM1 and flotillin consistently existed in the raft fraction.
GM1 Differentially Regulates TrkA Tyrosine Kinase-To analyze direct effects of GM1 on the TrkA kinase activity, immunoprecipitated TrkA was used for the in vitro kinase assay in the presence of various concentrations of GM1. As shown in Fig. 8A, low concentrations of GM1 rather enhanced the kinase activity, whereas relatively high concentrations of GM1 (Ͼ500 M) suppressed the kinase in a dose-dependent manner. To examine the specificity of GM1 in the enhancing effect on the TrkA kinase activity, GM3 or GT1b was also added to the kinase assay system at various concentrations (Fig. 8B). Consequently, as shown in Fig. 8B, clearly neither GM3 nor GT1b enhanced TrkA activity at any concentration as GM1 did.
Rho Is Not Modulated in the GM1 Synthase Gene Transfectants-Because RhoA has been shown to be involved in the regulation of neuronal differentiation and neurite outgrowth (17), overexpression of GM1 might suppress the neurite extension by modulating the Rho activity. Then, we examined the activation level of Rho in the GM1 synthase gene-transfectant cells. Activated Rho was not detected in either the vector controls or the GM1 ϩ cells, although LPA could induce an activated Rho band (data not shown). Therefore, it was not likely that GM1 affected the Rho kinase activity in the transfectant cells.
High Expression of GM1 Resulted in the Reduction of Membrane Fluidity-Because the GM1 ϩ cells showed reduced dimerization of the TrkA receptor (Figs. 5 and 6), and an altered intracellular localization of TrkA receptors (Fig. 7), we thought that overexpression of GM1 might affect the membrane fluidity. Then, we examined the effects of overexpression of GM1 on the plasma membrane fluidity in live cells using a FRAP system as described under "Experimental Procedures." To examine the membrane fluidity in living cells with FRAP experiments, we stained cell surface GM1 by using CTB-Al-exa555. As shown in Fig. 1C, GM1 ϩ cells were stained uniformly and strongly, and vector control cells showed patched distribution with low intensity. Furthermore, CTB-Alexa555 rapidly underwent endocytosis within 5-10 min, causing troubles to compare FRAP between these two cell types. In contrast, cell surface membranes were uniformly and moderately stained with DiI in both the vector controls and GM1 ϩ cells. Therefore, we used DiI instead of CTB-Alexa555 for cell surface labeling in FRAP experiments. Cells were loaded with 2.5 g/ml DiI for 10 min, then photobleach was applied to plasma membrane and the fluorescence recovery in these bleached regions was monitored. Fluorescence of DiI was markedly lost after photobleaching, and recovery was observed for ϳ60 s after photobleaching. Although the fluorescence recovery of membrane in vector control cells was 41.37%, the recovery in the GM1 ϩ cells was 24.8% (Fig. 9B). The t1 ⁄2 values of fluorescence recovery were 7.27 Ϯ 1.9 s and 9.36 Ϯ 2.8 s for vector control and GM1 synthase gene transfectants, respectively (Fig. 9C). FIG. 8. GM1 differently affects TrkA tyrosine kinase activity in vitro. Immunoprecipitation with anti-Trk antibody was performed. Immunoprecipitated TrkA was mixed with various concentrations of GM1 (A), GM3 or GT1b (B), and its kinase activity was measured by adding [␥-32 P]ATP and incubating for 10 min at 30°C, followed by SDS-PAGE and autoradiography. Relative intensity of each band was measured by National Institutes of Health Image and presented as the percentage of the sample with no GM1 (100%). The data shown are the mean Ϯ S.D. from three experiments.
These data suggested that overexpression of GM1 reduced the membrane fluidity and might result in the interference with the dimerization and phosphorylation of TrkA.

DISCUSSION
The characteristic expression patterns of gangliosides in the nervous tissues during development have suggested that they play important roles in the neurogenesis of vertebrates (18). Glycosyltransferase genes responsible for the synthesis of gangliosides also showed corresponding expression patterns during brain development (19). Because exogenously added gangliosides induced differentiation of neuronal cells in vitro (20), various gangliosides are thought to have neurotrophic effects, and have been administrated to experimental animals after generating artificial neurological damages or disorders by mechanical or chemical manipulation (21), by the injection of toxic reagents (22) or by ischemic treatment (23). Although various gangliosides have often had positive effects presumably due to their neurotrophic activity, it has been difficult to elucidate the molecular mechanisms for the effects of gangliosides endogenously generated in the cells or tissues.
A number of studies have been performed to investigate the effects of gangliosides on the function of nerve growth factor receptors (12, 13, 24 -27). Physical and functional association of GM1 with NGF receptor TrkA was reported (24 -26). Activation of TrkA with GM1 (24, 26), or enhancement of the TrkA dimerization due to NGF with GM1 (25) was also reported.
Ferrari et al. (13) reported that GM1 alone could at least partly replace the activity of NGF. In particular, Mutoh et al. (12) reported that exogenously added GM1 tightly binds to TrkA and enhances the phosphorylation of TrkA when stimulated with NGF. Rabin et al. (27) showed that GM1 activated Trk receptors via the induction of neutrophin release. However, it is not clear whether exogenously added GM1 actually acts in the same way as endogenously generated GM1, because the density and molecular topology of the added GM1 on the cell membrane are hard to be determined. Its molecular form and effect in the liquid on the cultured cells are also difficult to be precisely determined.
Using a cloned GM1/GD1b/GA1 synthase gene (5), we have tried to modify the ganglioside composition of PC12 cells and succeeded in obtaining clones expressing higher levels of GM1. Although the gene product had activity to produce not only GM1 but GD1b (5), the resulting transfectant cells expressed increased levels of GM1 and fucosyl-GM1, and reduced levels of GT1b. Consequently, we could analyze the effects of GM1 derived from introduced cDNA on the response to NGF.
In the response to NGF, GM1 ϩ cells showed much reduced reaction in both the neurite extension and the activation of TrkA/ERK1/2 signaling pathway. The vector control cells showed fairly good neurite formation and prompt phosphorylation of TrkA upon NGF treatment. ERK1/2 were also activated quickly with a peak at 5 min after NGF treatment. In contrast, GM1 ϩ cells showed no neurite extension and no or a very faint response of TrkA phosphorylation. Activation of ERK1/2 was also scarcely detectable. Thus, expression of GM1 showed suppressive effects on NGF-induced differentiation.
Recent progress in the analysis of membrane microdomains indicates that even neuronal cells have caveolae-like domains containing Shc, Ras, caveolin, and TrkA (28), and GEM/rafts are structurally unique components of plasma membranes, crucial for neural development and function (29). Differentiation signals from NGF could be transduced only through the NGF receptors localized in GEM/rafts showing a different behavior from that of epidermal growth factor receptor (30). Moreover, association and interaction of caveolin with TrkA and p75 NTR were demonstrated as a mechanism affecting the neurotrophin-induced signals (31,32). Then, the most intriguing issue is how the regulation of growth/differentiation factor receptors is performed by gangliosides. Among the three models presented by Miljan et al. (33), i.e. ganglioside modulation of ligand binding, ganglioside regulation of receptor dimerization, and ganglioside implication with receptor activation state and subcellular localization, the latter two seemed likely in the results of our study.
There have been a number of reports to indicate the changes in the intracellular localization of receptors and signal molecules following the modification of components in lipid rafts (34). In Swiss3T3 cells, overexpression of GM1 synthase gene induced clear reduction of cell growth. It also resulted in the reduction of phosphorylation levels of PDGFR and ERK1/2 after PDGF treatment (8). The mechanisms behind these phenotypic changes should be the dramatic changes in the intracellular localization of receptor molecules in the cells, and this was also the case in PC12 as demonstrated in this study. Originally, NGF induces neuronal differentiation through the TrkA/Ras/MEK/ERK pathway in PC12 cells (35)(36)(37)(38)(39)(40). Our results indicated that overexpression of GM1 modulated the initial events in TrkA activation just after binding of NGF, i.e. TrkA dimerization and phosphorylation. This means that TrkA outside of GEM/rafts is hard to be activated in GM1 ϩ cells, suggesting that appropriate physicochemical circumstances of GEM/rafts are needed for the early step of its activation. Not only TrkA but also p75 NTR and Ras underwent translocation from GEM/rafts to non-raft compartment. Although the roles of p75 NTR are not clear, it might be important in the full reaction of PC12 cells to NGF, and its cotransfer with TrkA may enhance the reduction in NGF signals in GM1 ϩ cells. Recently, new roles of p75 NTR have been reported not only in the apoptosis but in the myelination (41,42). Moreover, Ras translocation to the non-raft fraction might also enhance the suppressive effects. H-Ras is believed to localize in GEM/rafts (43,44). Our study, however, indicated that intracellular localization of Ras could be changed with the modification of carbohydrate moiety of expressed glycosphingolipids. Mechanisms for this regulation remain to be investigated.
Dynamic changes in the subcellular localization of receptors during the ligand binding and subsequent receptor activation/ polymerization have been observed in various systems (45). In the case of c-Ret receptor, GDNF binding to GFR␣ (glycosylphosphatidylinositol-anchored) in GEM/rafts recruit c-Ret molecule to the microdomain (46). EGF/EGFR binding in GEM/ rafts results in the translocation out of the microdomain. In the case of fibroblast growth factor receptor, the primary signaling out side of GEM/rafts causes secondary engagement of GEM/ raft signaling components. In contrast with these systems, Trk receptors usually show constant localization and activation in GEM/rafts during the signal transduction. We also confirmed no apparent changes in the subcellular localization of NGF receptors after NGF treatment (data not shown). Cell typespecific factors might determine which pattern of regulation individual receptor molecules undergo. Structures of receptors are also very important in the regulation of the intracellular localization. However, little is known about the universal principle determining the behavior of receptors inside/outside of GEM/rafts, and it remains to be analyzed.
Epidermal growth factor (EGF) induces cell proliferation through the EGF receptor/Ras/MEK/MAPK pathway, too. The reason why these two factors, NGF and EGF, can induce different cellular events via the same signaling pathway has been explained based on the different kinetics of ERK1/2 activation after their binding to the individual receptors (46). Therefore, the response of the GM1 ϩ cells to EGF might be very interesting, and remains to be analyzed.
As for the reason for the difference in the effects of exogenous GM1 and cDNA-derived GM1, four possibilities should be considered. First, exogenously added GM1 may exert effects on the cell surface molecules (such as TrkA) in quite a different way from that of cis-existing glycolipids, although some portion of the added GM1 is certainly incorporated into the membrane and expressed (data not shown). Second, the expression levels of GM1 may vary resulting in the different effects on TrkA molecular function as observed in the experiments of in vtro kinase of the precipitated TrkA (Fig. 8). The artificial control of the expression levels of the transfected gene is not so easy at this moment, although it is needed to establish multiple transfectants with various levels of GM1 expression to confirm the dose-dependent effects in the future. Third, the altered NGF signal may come not from increased GM1, but from other changes of glycolipid components in the transfectant cells, i.e. increased fucosyl-GM1 or decreased GT1b. These additional changes should not be present in the cells treated with exogenous GM1. Fourth, differences in the duration of the exposure of cells to GM1 might be crucial to determine the fates of cells as suggested by the experiments for alcohol toxicity (47).
Although an important role of glycosphingolipids in the formation of the membrane microdomains has been discussed (48 -50), the quantitative effects of glycosphingolipids have not been well understood. Little is also known about the effects of ganglioside expression on the membrane fluidity. The effects of exogenous gangliosides on the cell membrane fluidity have been examined in a few system using the technique of FRAP. HL-60 cells showed decreased fluidity after GM3 treatment (51). In the analysis of the ethanol effects on the mobility and viability of embryonal neural crest cells, added GM1 decreased membrane fluidity, resulting in the protection of the cell death (52). Furthermore, the membrane fluidity of the reconstituted liposomes increased with dioleoylphosphatidylcholine (liquid crystal phase) compared with distearoylphosphatidylcholine (gel phase), resulting in the increased kinase activity of EGFR (47). Taken together, it seems reasonable that overexpression of GM1 results in the decreased membrane fluidity and in the reduction of the differentiation signal with NGF/TrkA as demonstrated in this study. However, how the membrane fluidity associates with the nature of the GEM/raft is not well understood and remains to be investigated.
In conclusion, our findings suggest that glycosphingolipids produced and expressed in the cells play critical roles in the modulation of the quality and quantity of signals for the cell differentiation/proliferation and probably for death. Although the mechanisms by which glycosphingolipids regulate the signals have been poorly understood, the results presented here clearly indicate one example of the modes for such regulation. Namely, the microdomain is important as a place for the interaction between glycolipids and signaling molecules, and carbo-hydrate moieties in the glycosphingolipids contain much more influence than expected on the intracellular localization of various receptors and signal molecules. This means that we need to recognize glycolipids as critical factors determining the cell fates and to understand their roles in malignant tumors and neuronal degeneration when we construct strategies to apply glycosphingolipids and their synthetic machineries for the treatment of cancers and neurological degenerative diseases.