Specific glycosphingolipids mediate epithelial-to-mesenchymal transition of human and mouse epithelial cell lines

Abstract

Epithelial-to-mesenchymal cell transition (EMT) is a basic process in embryonic development and cancer progression. The present study demonstrates involvement of glycosphingolipids (GSLs) in the EMT process by using normal murine mammary gland NMuMG, human normal bladder HCV29, and human mammary carcinoma MCF7 cells. Treatment of these cells with d-threo-1-(3′,4′-ethylenedioxy)phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol (EtDO-P4), the glucosylceramide (GlcCer) synthase inhibitor, which depletes all GSLs derived from GlcCer, (i) down-regulated expression of a major epithelial cell marker, E-cadherin; (ii) up-regulated expression of mesenchymal cell markers vimentin, fibronectin, and N-cadherin; (iii) enhanced haptotactic cell motility; and (iv) converted epithelial to fibroblastic morphology. These changes also were induced in these cell lines with TGF-β, which is a well-documented EMT inducer. A close association between specific GSL changes and EMT processes induced by EtDO-P4 or TGF-β is indicated by the following findings: (i) The enhanced cell motility of EtDO-P4-treated cells was abrogated by exogenous addition of GM2 or Gg4, but not GM1 or GM3, in all 3 cell lines. (ii) TGF-β treatment caused changes in the GSL composition of cells. Notably, Gg4 or GM2 was depleted or reduced in NMuMG, and GM2 was reduced in HCV29. (iii) Exogenous addition of Gg4 inhibited TGF-β-induced changes of morphology, motility, and levels of epithelial and mesenchymal markers. These observations indicate that specific GSLs play key roles in defining phenotypes associated with EMT and its reverse process (i.e., mesenchymal-to-epithelial transition).

Epithelial cells in tissues or cultured in vitro change their morphology, growth behavior, and motility when they encounter a different microenvironment. For example, epithelial cells that come in contact with soluble extracellular matrix components acquire characteristics similar to those of mesenchymal fibroblasts, as originally observed by Hay and colleagues (1–3). After extensive further studies, a concept emerged that the transitional process from epithelial to mesenchymal phenotype is commonly associated with embryonic development, as well as pathological conditions, such as fibrosis or cancer metastasis, and the process was termed “epithelial-to-mesenchymal transition” (EMT) (4–10).

The EMT process has been characterized, in addition to cell morphology change and increased cell motility, by a striking decline in epithelial markers, such as E-cadherin, desmoplakin, and cytokeratins, accompanied by enhanced expression of mesenchymal markers, such as vimentin, fibronectin, and N-cadherin (10, 11).

Previous studies from our group and others have demonstrated that glycosphingolipids (GSLs) mediate cell adhesion (12–15), and they modulate cell growth through their effect on growth factor receptor tyrosine kinases (16–20). In addition, we have demonstrated that some GSLs, particularly gangliosides, play an essential role in defining cell motility through their interaction with integrins and tetraspanin CD9 or CD82 (21, 22). Because changes in cell adhesion, motility, and growth are an essential phenotypic change in EMT, we considered the possible involvement of GSLs in control of EMT. To our knowledge, no studies on effect of GSLs and gangliosides on EMT have been performed previously.

For this purpose, the following approaches were undertaken: (i) effect of d-threo-1-(3′,4′-ethylenedioxy)phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol (EtDO-P4), the glucosylceramide (GlcCer) synthase inhibitor (23, 24), on EMT; (ii) effect of TGF-β, the well-known EMT inducer (25), on changes of GSL pattern, and identification of specific GSLs that are significantly reduced or depleted by TGF-β; and (iii) possible phenotypic reversion or suppression of EMT process by exogenous addition of specific GSLs identified in step ii.

Through these approaches, we have demonstrated the essential role of specific GSLs in the EMT process, and possible phenotypic reversion, i.e., mesenchymal-to-epithelial transition, caused by specific GSLs.

Results

Effect of EtDO-P4 and/or TGF-β Treatment on Expression of Epithelial Marker and Mesenchymal Marker Molecules.

An epithelial cell marker molecule, E-cadherin, which is highly expressed in epithelial cell lines NMuMG and MCF7, was clearly reduced when the cells were treated with 1 μM EtDO-P4 or with 2 ng/mL TGF-β (Fig. 1A, B, E, and F). EtDO-P4 inhibits GlcCer synthase and depletes all GSLs derived from GlcCer (23, 24), and TGF-β induces a series of EMT processes (25). Associated with down-regulation of E-cadherin in NMuMG, the mesenchymal markers vimentin and N-cadherin were strongly enhanced by EtDO-P4 or TGF-β, whereby the enhancing effect of TGF-β was stronger than that of EtDO-P4 (Fig. 1B). Levels of vimentin, fibronectin, and N-cadherin expression in MCF7 were very low. Therefore, only changes of E-cadherin expression induced by EtDO-P4 or TGF-β are shown (Fig. 1 E and F).

Fig. 1.

Expression of E-cadherin as epithelial cell marker vs. vimentin, fibronectin, and N-cadherin as mesenchymal markers, induced by EtDO-P4 and/or TGF-β. NMuMG (A and B), HCV29 (C and D), and MCF7 (E and F) cells (2 × 10⁵ per well in 6-well plates) were cultured overnight as described in SI Materials and Methods. Medium was changed to regular culture medium alone (lane 1 or column 1; control), medium containing 1 μM EtDO-P4 only (lane 2 or column 2), medium containing 2 ng/mL TGF-β (lane 3 or column 3), or medium containing 1 μM EtDO-P4 plus 2 ng/mL TGF-β (lane 4 or column 4). Cells were cultured for 72 h for 1 μM EtDO-P4 (lane 2 or column 2) or 48 h for 2 ng/mL TGF-β (lanes 3 and 4 or columns 3 and 4). Cells were harvested and lysed in RIPA buffer (1% Triton X-100; 150 mM NaCl; 25 mM Tris, pH 7.4; 5 mM EDTA; 0.5% sodium deoxycholate; 0.1% SDS; 5 mM tetrasodium pyrophosphate; 50 mM sodium fluoride; 1 mM Na₃VO₄; 2 mM phenylmethanesulfonyl fluoride; and 0.076 unit/mL aprotinin). Protein content was determined by using a microBCA protein assay reagent kit (Pierce), with BSA as standard. Aliquots (10 μg of protein per slot) were subjected to SDS/PAGE, followed by transfer to PVDF membrane (Millipore), as described previously (40, 43). Membranes were incubated with antibodies against E-cadherin, vimentin, N-cadherin, desmoplakin, and fibronectin, then incubated with the appropriate secondary antibody conjugated with HRP, and proteins were revealed with a Supersignal Chemiluminescence substrate kit (Pierce). Experiments were performed in triplicate, and representative Western blot results are shown (A, C, and E). After stripping with Re-blot Plus Mild Solution (Millipore), each membrane was reblotted with β-actin or GAPDH as loading control. Intensity of each band was quantified by densitometry using the Scion Image program. Normalized values with each loading control are shown as relative intensity on ordinate (B, D, and F). Mean ± SD is shown in each figure. Two other independent experiments gave similar results. NS indicates not significant. P > 0.05; *, P = 0.01–0.05; **, P = 0.001 to 0.005; ***, P < 0.001. For cell staining for E-cadherin expression (G), 2 × 10⁴ NMuMG cells were seeded onto 12-mm diameter glass coverslips in 24-well tissue culture plates. After 24 h, cells were treated with EtDO-P4 and/or TGF-β for 72 h, then fixed with 2% paraformaldehyde in PBS, i.e., D-PBS(−) (2.7 mM KCl, 1.5 mM KH₂PO₄, 8.1 mM Na₂HPO₄, and 138 mM NaCl), for 15 min, washed with PBS, and blocked with 1% BSA/0.1% NaN₃/ PBS for 30 min. Cells were stained with anti-E-cadherin for 2 h at room temperature, incubated with secondary FITC-labeled antibody for 1 h at room temperature, washed, and mounted with Glycergel (Dakocytomation). Stained cells were subjected to Fluoro-View (Olympus) and observed at 400× magnification.

Because HCV29 cells do not express a detectable level of E-cadherin, change in expression of this component by EtDO-P4 or TGF-β could not be determined. Treatment of the cells with EtDO-P4 or TGF-β, however, caused significant increase in expression of the mesenchymal markers vimentin and fibronectin, similarly to NMuMG, and a slight increase in N-cadherin (Fig. 1D). In all cases, EtDO-P4 and TGF-β had no additive effect over either TGF-β or EtDO-P4 alone.

The reduction of E-cadherin expression by EtDO-P4 treatment of NMuMG cells was confirmed by cell staining observed by fluorescence microscopy. E-cadherin expression, clearly detected in control cells, was reduced significantly by incubation with EtDO-P4 or TGF-β or EtDO-P4 plus TGF-β (Fig. 1G), similarly to Western blot analysis as described above.

Enhanced Cell Motility Induced by Treatment with EtDO-P4 and/or TGF-β.

Because enhancement of cell motility is a well-known phenotypic change associated with EMT, cell motility change was analyzed by haptotactic/phagokinetic assay using gold sol. Treatment of cells with 1 μM EtDO-P4 highly increased cell motility to a level similar to that induced by TGF-β treatment, compared with control cells. EtDO-P4 plus TGF-β had no additional enhancing effect on cell motility compared with EtDO-P4 alone or TGF-β alone. Changes of cell motility induced by EtDO-P4, TGF-β, or EtDO-P4 plus TGF-β were essentially the same in NMuMG (Fig. 2A), MCF7 (Fig. 2B), and HCV29 (Fig. 2C).

Fig. 2.

Haptotactic motility of NMuMG, MCF7, and HCV29 cells induced by EtDO-P4 and/or TGF-β. NMuMG (A), MCF7 (B), and HCV29 (C) cells were grown (2 × 10⁵ cells per well) in 6-well plates for 24 h. Medium was changed to complete culture medium alone as control (column 1), to medium containing 1 μM EtDO-P4 only (column 2) and incubated 72 h, to medium containing 2 ng/mL TGF-β only (column 3) and incubated 48 h, or to medium containing 1 μM EtDO-P4 plus 2 ng/mL TGF-β (column 4) and incubated 72 h. After incubated cells were detached with trypsin/EDTA, 5 × 10³ cells in complete culture medium were added onto each gold sol-coated well, and haptotactic motility during 18-h incubation was determined as described in SI Materials and Methods. Photos of track areas of 30 cells were taken, and cleared areas on gold sol were measured by using the Scion Image program. Two independent experiments gave similar results. Mean ± SD of cleared area (square pixels) is shown on the ordinate in each panel. ***, P < 0.001.

Cell Morphology Change Induced by EtDO-P4.

Flat, epithelial shape of NMuMG cells grown in normal medium was not changed by treatment with 1 μM EtDO-P4. Cell shape was converted to bipolar, well-oriented, slanted, fibroblastic appearance by treatment with 2 ng/mL TGF-β. Fibroblastic appearance was unchanged for cells treated with EtDO-P4 plus TGF-β (Fig. 3Upper). Flat, epithelial shape of HCV29 cells was converted to fibroblastic appearance by treatment with either 1 μM EtDO-P4, 2 ng/mL TGF-β, or EtDO-P4 plus TGF-β (Fig. 3 Lower).

Fig. 3.

Morphological change of NMuMG and HCV29 cells treated with EtDO-P4 and/or TGF-β. NMuMG (Upper) and HCV29 (Lower) cells (2 × 10⁵ cells per well) were grown in 6-well plates for 24 h. Medium was changed to complete culture medium alone as control (column 1), medium containing 1 μM EtDO-P4 only (column 2), medium containing 2 ng/mL TGF-β (column 3), or medium containing 1 μM EtDO-P4 plus 2 ng/mL TGF-β (column 4); then, cells were cultured for 72 h. Cell monolayers were fixed with 2% paraformaldehyde in PBS and stained with 1% toluidine in water, and photos were taken by phase-contrast microscopy (Nikon) at 80× magnification.

These results indicate that both NMuMG and HCV29 epithelial cells were converted to fibroblast-like shape by TGF-β. EtDO-P4 had a strong effect on morphology of HCV29 but no clear effect on morphology of NMuMG. On the other hand, NMuMG was highly susceptible to both EtDO-P4 and TGF-β in terms of reduced expression of E-cadherin and enhanced expression of mesenchymal markers. Thus, morphology change does not always reflect EMT process (see Discussion).

GSL Composition in Epithelial Cells and Its Changes in EMT Process Induced by TGF-β Treatment.

Because TGF-β treatment induces EMT, we examined the possibility that it may also induce change of GSL composition. GSLs of NMuMG and HCV29 cells, with or without pretreatment with TGF-β, were extracted; class-separated as nonacidic, monosialogangliosides, disialogangliosides, and trisialogangliosides and compared by high-performance thin-layer chromatography (HPTLC) with orcinol/sulfuric acid staining; and confirmed by immunostaining with specific mAbs or with cholera toxin subunit B (CTB) for GM1. TGF-β treatment of NMuMG cells caused: (i) depletion of 2 bands of Gg4 (asialo-GM1), staining by mAb TKH7 (Fig. 4A), and staining by orcinol/sulfuric acid (Fig. 4C, lane 2 vs. 1, indicated by ←*); and (ii) clear reduction of ganglioside GM2 (Fig. 4D, lane 2 vs. 1, indicated by ←**). GM2 was also reduced in NMuMG cells immunostained with anti-GM2 mAb MK1-8 (Fig. 4A).

Fig. 4.

Effect of TGF-β treatment on GSL composition of NMuMG and HCV29 cells. NMuMG (A, C, and D) and HCV29 (B, E, and F) cells (≈2 × 10⁷) were cultured in culture medium containing 2 ng/mL TGF-β for 48 h. Cells were harvested and extracted with isopropanol/hexane/water (55:25:20). Nonacidic fractions (C and E) and monosialoganglioside fractions (D and F) were prepared as described in SI Materials and Methods. After dialysis against distilled water and lyophilization, the residue was dissolved in 50 μL of chloroform/methanol (2:1), and 5-μL aliquot was spotted on an HPTLC plate. After developing with chloroform/methanol/aqueous 0.2% CaCl₂ (55:40:10), GSLs were visualized by spraying with orcinol/sulfuric acid (C–F). The HPTLC band from NMuMG (A) and that from HCV29 (B), respectively, immunostained by TKH7 for Gg4 (Top), by CTB for GM1 (Second), by MK1–8 for GM2 (Third), or by DH2 for GM3 (Bottom), and their patterns are shown. Arrow with asterisk (←*) in C indicates position of reference Gg4 (from Matreya). Arrow with 2 asterisks (←**) in D indicates position of reference GM2 (from Matreya).

In HCV29 cells, Gg4 is absent, and there was no clear change of nonacidic GSLs in these cells upon TGF-β treatment (Fig. 4E). However, GM2 level in HCV29 was reduced significantly upon TGF-β treatment (Fig. 4B), similarly to the GM2 reduction in NMuMG cells caused by TGF-β. In both NMuMG and HCV29, levels of GM1 and GM3 were not strongly affected by TGF-β treatment, as revealed by TLC staining with CTB for GM1 and by TLC immunostaining with anti-GM3 mAb DH2 (Fig. 4 A and B).

Depletion of Gg4 expression in NMuMG cells treated with TGF-β was also observed by cell surface staining and flow cytometry analysis.

Enhanced Motility of EtDO-P4-Treated Cells Is Abrogated by Exogenous Addition of GM2 but Not GM3 or GM1.

The strong enhancement of cell motility induced by EtDO-P4 in MCF7, HCV29, and NMuMG cells (Fig. 2) may be due to elimination of motility-suppressing GSLs. This possibility was tested by determining motility of EtDO-P4-treated cells with or without exogenous addition of GSLs. Motility of cells enhanced by EtDO-P4 treatment, compared with nontreated control cells (column 2 vs. 1 in Fig. 5 A, B, and C), in MCF7 (Fig. 5A), HCV29 (Fig. 5B), and NMuMG cells (Fig. 5C) was clearly abrogated by exogenously added GM2 in a dose-dependent manner, although the same doses of GM3 or GM1 had no motility-inhibitory effect (Fig. 5 A and B). Similarly, Gg4 addition decreased motility of NMuMG cells (Fig. 5C).

Fig. 5.

Effect of exogenous addition of gangliosides on cell motility enhanced by EtDO-P4 treatment. GM1, GM2, GM3, and Gg4 dissolved in chloroform/methanol (2:1) were evaporated to dryness under nitrogen stream and dissolved in serum-free medium to obtain a 50 μM concentration of each GSL. Each suspension or solution was sonicated for 3 h in sonication bath, then added with FBS to 10% concentration. MCF7 (A) and HCV29 (B) in RPMI, and NMuMG (C) in DMEM, each with 10% FBS, were cultured overnight as described in SI Materials and Methods. Medium was changed to that containing 1 μM EtDO-P4 and further cultured for 72 h. Then, medium was changed to that containing various concentrations (0.1, 1, 5, or 10 μM) of GM3, GM2, GM1, or Gg4 prepared as above, and culture was continued for 18 h. Control indicates no treatment; EtDO-P4 indicates EtDO-P4 treatment alone. After detachment, cell motility was analyzed as described in SI Materials and Methods and in Fig. 2 legend, and is shown as gray columns in A–C. Cell surface expression of exogenously added GM2 or Gg4 was assessed by cell surface staining with anti-GM2 mAb MK1-8 or anti-Gg4 mAb TKH7, followed by flow cytometry (Beckman Coulter) in MCF7 (D), HCV29 (E), or NMuMG (F) cells. Graph 1: cultured as above with no additional treatment. Graph 2: EtDO-P4 treatment but no GSL addition. Graph 3: EtDO-P4 treatment plus 50 μM GM2 (D and E) or 50 μM Gg4 (F). Shaded histograms show staining with mAb MK1-8 for GM2 (D and E) and mAb TKH7 for Gg4 (F). Open histograms show staining with isotype control.

Cell surface expression of GM2 was greatly reduced by EtDO-P4 treatment and restored after incubation in GM2-containing medium for MCF7 (Fig. 5D), HCV29 (Fig. 5E), and NMuMG cells. Similarly, cell surface Gg4 expression in NMuMG was depleted after EtDO-P4 treatment and restored to the level of original control cells by exogenous addition of Gg4 (Fig. 5F).

These results suggest that exogenously added GM2 or Gg4 was adsorbed and retained in cell membrane, and it inhibited cell motility that was originally stimulated by EtDO-P4 treatment.

Suppression of TGF-β-Induced EMT Process by Exogenous Addition of Gg4 in NMuMG Cells.

Because GM2 and Gg4 are reduced or depleted in TGF-β-treated NMuMG cells (Fig. 4), and addition of GM2 or Gg4 abrogated enhanced cell motility induced by EtDO-P4 treatment (Fig. 5), we examined the possibility that Gg4 or GM2, when exogenously added, may suppress the TGF-β-induced EMT process. Addition of exogenous Gg4 to NMuMG cells causes inhibition of TGF-β-induced EMT process, based on several findings. (i) Motility of cells treated with TGF-β was strongly enhanced compared with nontreated control cells. The enhanced motility was not inhibited by GM2 but was significantly inhibited by Gg4 (Fig. 6A). (ii) Many cells treated with TGF-β alone showed a fibroblastic appearance (Fig. 6B, 2), whereas treatment with Gg4 caused the majority of cells to have flattened, cubical shapes (Fig. 6B, 4) similar to control cells (Fig. 6B, 1). Cells treated with GM2 showed a fibroblastic appearance similar to that of TGF-β-treated cells (Fig. 6B, 3). Cells treated with only Gg4 without TGF-β did not undergo a morphology change. (iii) Relative expression of epithelial marker E-cadherin was greatly reduced by TGF-β treatment. E-cadherin level was clearly enhanced by pretreating cells with 50 μM Gg4 in 2%-FBS-DMEM for 18 h, followed by culture with 1 ng/mL TGF-β for 36 h (Fig. 6 C and D). (iv) In contrast, expression of mesenchymal markers N-cadherin, vimentin, and fibronectin was greatly enhanced by TGF-β under this condition and was significantly reduced by pretreatment with Gg4 to a level similar to that of controls. However, exogenous addition of GM2 had no effect on enhanced mesenchymal markers (Fig. 6 C and D).

Fig. 6.

Effect of exogenous addition of gangliosides on cell motility, morphology, and expression of epithelial vs. mesenchymal cell markers. NMuMG cells (2 × 10⁵ per well in 6-well plates) were cultured in 10% FBS/DMEM overnight. Cells were preincubated with 50 μM GM2 or Gg4 prepared as described in the Fig. 5 legend for 8 h, then incubated for 48 h in 10% FBS/DMEM containing 1 ng/mL TGF-β plus 50 μM GM2 or Gg4. (A) Cell motility was determined as described in SI Materials and Methods and the Fig. 2 legend. For analysis of cell morphology and expression of EMT markers, NMuMG cells (5 × 10⁴ per well in 12-well plates) were cultured overnight as above. Cells were preincubated with 50 μM GM2 or Gg4 for 18 h in 2% FBS/DMEM, then incubated for 36 h in 10% FBS/DMEM containing 1 ng/mL TGF-β plus 50 μM GM2 or Gg4. Cells were fixed and stained, and photos were taken (B). Expression of marker proteins was analyzed by Western blot. Representative results (C) and relative level of expression normalized with loading control (D) are presented as mean ± SD from triplicate experiments. 1 indicates no treatment; 2, TGF-β treatment alone; 3, GM2 plus TGF-β treatment; and 4, Gg4 plus TGF-β treatment. P > 0.05. *, P = 0.01 to 0.05. **, P = 0.001 to 0.005. ***, P < 0.001.

Expression of desmoplakin, another epithelial marker, was similar to that of E-cadherin, was clearly reduced by treatment with TGF-β, and was enhanced by preincubation with Gg4. In this case, relatively short preincubation (8 h) with 50 μM Gg4 was necessary for enhanced desmoplakin expression, whereas Gg4 effect on E-cadherin expression required 18 h of preincubation, as described above.

Discussion

EMT process, the phenotypic conversion from less mobile epithelial cells to fibroblastic mesenchymal cells with higher motility, is accompanied by declining expression of epithelial cell molecules and enhanced expression of mesenchymal molecules. This process is essential for normal developmental process during embryogenesis and is implicated in promotion of tumor invasion and metastasis (4–10). There is increasing evidence that GSLs and glycosylation status of proteins play key roles in defining ontogenic development (12–15), oncogenic transformation (26, 27), and their phenotypic reversion (28, 29). GSLs and protein glycosylation are therefore expected to play some role in EMT process.

To clarify the functional effect of GSLs in EMT process, 3 approaches were used. (i) We measured the effect of GSL depletion by EtDO-P4 on expression of epithelial vs. mesenchymal molecules and the associated changes of EMT process. EtDO-P4 (23, 24) causes efficient depletion of all GSLs derived from GlcCer without enhanced expression of ceramide (28, 30), an inducer of various types of signal transduction (31). Thus, EtDO-P4 effect on cellular phenotype is based exclusively on GSL depletion and does not involve ceramide-induced signaling. (ii) We measured the effect of TGF-β (25) on GSL changes, particularly reduction or depletion of specific GSLs that are considered closely associated with EMT process. (iii) We measured the reversion or suppression of EMT process by exogenous addition of specific GSLs. These approaches were applied to 3 epithelial cell lines: mouse NMuMG, human HCV29, and human MCF7 cells.

All these cell lines displayed clearly enhanced cell motility and enhanced expression of mesenchymal markers by treatment with EtDO-P4 and/or TGF-β; i.e., these cells are highly susceptible to EMT process. However, EtDO-P4-induced cell morphology change was different for NMuMG vs. HCV29, in contrast to same-morphology change caused by TGF-β (Fig. 3). These findings suggest that the mechanism of morphology change is complex and is not simply controlled by up- or down-regulation of epithelial or mesenchymal molecules. TGF-β treatment induced complete depletion of Gg4 and partial reduction of GM2 in NMuMG cells. This may explain the results shown in Figs. 5 and 6; i.e., both GM2 and Gg4 reversed the enhanced motility of EtDO-P4-treated cells (Fig. 5), although Gg4 but not GM2 had a clear suppressive effect on TGF-β-induced EMT process (Fig. 6).

The exact mechanism by which GSLs function to maintain epithelial cell status, inhibiting EMT process, remains to be elucidated. Recent studies indicate that many transcription factors, signal transducers, and microRNAs may be involved as inducers or regulators in EMT process (5, 32–34). An interesting possibility is that GSLs may affect translocation of such transcription factors as Snail (35), Slug (36), or Twist (37) into nuclei, inducing or inhibiting expression of epithelial or mesenchymal molecules. The ability of GSLs and gangliosides to interact with various signal transducers, as well as with receptors for growth factors or integrins, to define cell adhesion, motility, and growth was well established in our previous studies (21, 28, 38).

It should be noted, however, that the cells used in this study were cell lines, not primary epithelial cells. In general, epithelial cell lines may have already acquired mesenchymal characteristics, and the degree of mesenchymal phenotype expressed is different from one cell line to another. It is therefore not justified to expect that expression of the same type of epithelial molecules in various epithelial cell lines or expression of the same mesenchymal molecules in various fibroblasts will be comparable. In fact, HCV29, considered to be a typical normal human bladder epithelial cell line, expresses minimal levels of E-cadherin but expresses high levels of the mesenchymal markers N-cadherin, fibronectin, and vimentin. MCF7, human benign breast epithelial tumor cells, express E-cadherin but do not express mesenchymal markers, as described in Results. That is, not all epithelial and mesenchymal molecules change in the same way during EMT processes.

The epithelial molecules examined in this study were limited to E-cadherin and desmoplakin; however, a few others, such as Muc1 and epithelial cytokeratin, may also be down-regulated. The number of types and the variability of up-regulated mesenchymal molecules are much higher than those of down-regulated epithelial molecules (10). Both epithelial and mesenchymal molecules may be organized with different GSLs at specific membrane microdomains, particularly “glycosynaptic microdomains” (refs. 28, 39, and 40; for review, see refs. 41 and 42), which control adhesion, motility, and growth.

Further studies on organizational assembly of GSLs with epithelial and mesenchymal molecules and the genetic and epigenetic basis for expression of specific GSLs involved in EMT process are expected to clarify the detailed mechanism of EMT process.

Materials and Methods

For details of cell culture, antibodies, reagents, cell motility assay, GSL extraction, and analysis, see SI Materials and Methods.

Analysis of Expression of Epithelial and Mesenchymal Markers Induced by EtDO-P4 and/or TGF-β Treatment in 3 Epithelial Cell Lines.

Cells were cultured in complete culture medium overnight, and monolayers were cultured with medium containing 1 μM EtDO-P4 for 72 h, medium containing 2 ng/mL TGF-β for 48 h, or medium containing 1 μM EtDO-P4 plus 2 ng/mL TGF-β for 72 h. EMT process was analyzed based on: (i) cell morphology change; (ii) cell motility change based on phagokinetic gold sol assay (SI Materials and Methods); and (iii) Western blot analysis of epithelial and mesenchymal markers. Cells were harvested, lysed, and analyzed as described previously (40, 43).

Reversing Effect of GSL on Enhanced Motility Induced with EtDO-P4.

After culturing in medium containing 1 μM EtDO-P4 for 72 h, cells were cultured in medium containing various concentrations (0.1, 1, 5, or 10 μM) of GSLs for 18 h, as indicated in Fig. 5 A–C. Cells then were detached, and haptotactic motility was analyzed as described in SI Materials and Methods.

Inhibitory Effect of Preincubation with Gg4 on TGF-β-Induced EMT.

NMuMG cells were cultured overnight in 10% FBS/DMEM. For study of Gg4 or GM2 effect on TGF-β-induced motility, medium was changed to that containing 50 μM Gg4 or GM2 for 8 h, and cells were then cultured in medium with addition of 1 ng/mL TGF-β in the continued presence of 50 μM Gg4 or GM2 for 48 h. Cell motility was determined as described in SI Materials and Methods. Different conditions were required to observe the effect of Gg4 on morphology change and on expression of epithelial or mesenchymal markers. That is, different preincubation time or lower FBS concentration gave better results. See also the legend for Fig. 6.

Statistical Analysis.

Data were analyzed by t test by using the Prism 3 program (GraphPad Software). Differences in results were considered significant when P value was less than or equal to 0.05.