Experiments were performed on two cell lines: YD15 and YD15M. YD15 cells were derived from mucoepidermoid carcinoma of the tongue and YD15M cells from a lymph node metastasis from YD15. Both cell lines were purchased from the Korean Cell Line Bank (KCLB, Seoul, Korea; cat. nos. 60504 and 60505, respectively) and maintained in the RPMI-1640 medium (Gibco, Daejeon, Korea) supplemented with 10% heat-inactivated fetal bovine serum (Biomeda Co., CA, USA) and 10% antibiotic-antimycotic solution (Welgene, Daegu, Korea) at 37°C in a humidified atmosphere containing 5% of CO₂. The medium was replaced with a fresh one, and the adherent cells were allowed to reach ~70% confluence. Then, the cells were detached by means of a trypsin-ethylenediamine tetraacetic acid solution (trypsin-EDTA: Gibco-BRL) and plated again (subcultured) in 6-well plates for each experiment.

When the cells reached ~70% confluence, the medium was removed, and the cells were washed twice with phosphate-buffered saline (PBS, pH 7.4). Then, the cell lysate was prepared in 200 ml of cold lysis buffer (1% NP-40, 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.02% sodium azide, 150 mg/ml phenylmethylsulfonyl fluoride (PMSF), 2 mg/ml aprotinin, 20 mg/ml leupeptin, and 1 mg/ml pepstatin A). Approximately 30 mg of protein of the cell lysate was separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (in a 10% gel) and transferred to a polyvinylidene difluoride membrane (Amersham, Piscataway, NJ, USA). Each membrane was blocked with a blocking solution containing 5% skim milk in Tris-buffered saline with Tween-20 (TBST; 2.42 g/l Tris-HCl pH 7.6, 8 g/l NaCl, 0.1% Tween-20) for 0.5 h and rinsed briefly in TBST. The membrane was incubated overnight at 4°C with the appropriate primary antibody: an anti-E-cadherin antibody (dilution 1:1,000, Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-β-catenin (1:1,000, Santa Cruz Biotechnology), anti-vimentin (1:1,000, Santa Cruz Biotechnology), anti-CD133 (1:500, Mybiosource, San Diego, CA, USA), anti-PCNA (1:1,000, Cell Signaling Technology, Beverley, MA, USA), anti-panCK (1:1000, Cell Signaling Technology), anti-ALDHA1 (1:1,000, Cell Signaling Technology), anti-OCT-4 (1:1,000, Cell Signaling Technology), or an anti-NANOG antibody (1:1,000, Cell Signaling Technology). A mouse monoclonal anti-β-actin antibody (1:2,500, Santa Cruz Biotechnology) was used as the control. Finally, the membrane was washed in TBST and the immunoreactivity of the proteins was analyzed with an enhanced chemiluminescence detection kit (Santa Cruz Biotechnology). The intensity of bands was quantified in densitometric software ImageJ (http://rsb.info.nih.gov/ij).

The immunofluorescence assay was performed as described previously (29). Cells were seeded onto glass Lab-Tek II Chamber Slides (Thermo Fisher Scientific Inc., Loughborough, UK) and incubated for 1 day. The growth medium was then removed, and cell monolayers were washed three times with a 1% PBS solution and fixed with 3.5% paraformaldehyde for 10 min at room temperature. The cells were washed three times with PBS and permeabilized by Triton X-100 (0.2%, 10 min at room temperature). Non-specific binding sites were blocked by incubation with PBS containing 1% BSA, for 30 min at room temperature. The cells were washed three times with PBS before incubation with the rabbit anti-vimentin antibody (1:200) followed by a fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit IgG antibody (1:1,000) and a mouse anti-β-catenin antibody (1:200; or a mouse anti-CD133 antibody, 1:200), with subsequent incubation with a Texas Red (TR)-conjugated goat anti-mouse IgG antibody (1:1,000). In some cases, the samples were incubated with the rabbit anti-PCNA antibody (1:200) followed by the FITC-conjugated goat anti-rabbit IgG antibody (1:1,000) and mouse anti-CD133 antibody (1:200) with subsequent incubation with the TR-conjugated goat anti-mouse IgG antibody (1:1,000). The cells were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) for visualization of nuclear morphology. The stained cells were analyzed under a confocal microscope (Carl Zeiss, Oberkochen, Germany), at excitation wavelength 490 nm and emission wavelength 540 nm for FITC, excitation 540 nm and emission 650 nm for TR, and excitation 360 nm and emission 450 nm for DAPI.

Suspensions of YD15 and YD15M cells that were detached with trypsin-EDTA were washed with PBS and fixed with 3.5% paraformaldehyde for 10 min at room temperature. The cells were washed three times with PBS and permeabilized by Triton X-100 (0.2%, 10 min at room temperature). Non-specific binding sites were blocked by treatment with PBS containing 1% BSA, at room temperature for 30 min. The cells were washed three times with PBS before incubation with the rabbit anti-vimentin antibody (1:1,000) followed by the FITC-conjugated goat anti-rabbit IgG antibody (1:1,000) and mouse anti-CD133 antibody (1:1,000) with subsequent incubation with the TR-conjugated goat anti-mouse IgG antibody (1:1,000). In some cases, the samples were incubated with the rabbit anti-PCNA antibody (1:1,000) followed by the FITC-conjugated goat anti-rabbit IgG antibody (1:1,000) and mouse anti-CD133 antibody (1:1,000) with subsequent incubation with the TR-conjugated goat anti-mouse IgG antibody (1:1,000). After the incubation, the cells with the bound antibodies were washed twice with washing buffer and analyzed by flow cytometry (Beckman Coulter, CA, USA). Approximately 10,000 counts were accumulated for each sample. Anti-IgG (isotype) antibodies were used to assess the non-specific binding (background) in this assay.

Magnetic cell separation (MACS) of vimentin⁺ cells was performed with EasySep™magnet according to the manufacturer’s instructions (Stem Cell Technologies, Durham, NC, USA). Briefly, YD15 and YD15M cells were plated in 6-well plates. After they reached 70% confluence, the cells were detached and washed twice in cold PBS. Dissociated cells were stained with the rabbit anti-vimentin antibody (1:1,000) followed by the FITC-conjugated goat anti-rabbit IgG antibody (1:1,000) and were incubated for 30 min at 4°C. After incubation with immunomagnetic microbeads for 30 min at 4°C, the cells were washed in PBS containing 0.5% BSA and 2 mM EDTA, filtered through a 40-mm cell strainer. Then the cells were passed through the magnetic cell separation device (Auto-Macs; Miltenyi Biotec) for selection of vimentin positive cells.

Cells were trypsinized to obtain a single-cell suspension, harvested by centrifugation, and washed with PBS. After fixation in ice-cold 70% ethanol at 4°C overnight, the cells were collected and washed twice with PBS. The cells were then stained with PI (3 μg/ml) for 15 min. After that, DNA from 10⁴ of G0/S phase cells was quantified on the flow cytometer (Beckman Coulter).

This assay of tumor-derived cells was conducted using the ALDEFLUOR kit (StemCell Technologies). In particular, the cells were suspended in ALDEFLUOR assay buffer containing ALDH1 substrate (BODIPY-aminoacetaldehyde, 1 M per 10⁶ cells) and incubated for 30 min at 37°C. As a negative control, some batches of the cells were incubated with a specific ALDH1 inhibitor, diethylaminobenzaldehyde (DEAB; 50 mM). After that, the cells were washed twice with washing buffer and analyzed by flow cytometry (Beckman Coulter). Approximately 10,000 counts were accumulated for each sample.

The data were expressed as mean ± standard deviation. All experiments were repeated three times, and all calculations were performed in Microsoft Excel. For significance testing, we used Student’s t-test (at P<0.05).

The metastatic cell line, YD15M, which was derived from mucoepidermoid carcinoma, showed morphological features of cells undergoing EMT (Fig. 1A). The observed morphological changes in YD15M cells include switching from a cuboid epithelial shape to a fibroblastic elongated appearance, with colonies of irregular spherical shape. The protein expression analysis of YD15M cells revealed a loss of the epithelial markers E-cadherin and β-catenin; this was not the case in YD15 cells (Fig. 1B). Simultaneously, an increased level of the mesenchymal marker vimentin was observed in YD15M cells. To verify these findings, we used laser scanning confocal fluorescence microscopy to analyze the expression of these epithelial and mesenchymal markers (Fig. 1C). As shown in the figure, YD15M cells showed decreased protein expression of β-catenin (red) and increased expression of vimentin (green). Moreover, a small population of cells with nuclear localization of β-catenin was observed among YD15M cells, but not among YD15 cells, indicating that the metastatic YD15M cells were derived from YD15 cells in association with EMT.

The protein expression levels of vimentin and CD133 were significantly higher in YD15M cells than in YD15 cells according to immunofluorescence assays (Fig. 2A). The percentage of cells expressing CSC-specific protein CD133 was greater among YD15M cells than among YD15 cells. The CD133 protein was observed mainly in the cytoplasm and plasma membrane of the cells. To evaluate the correlation of vimentin and CD133 expression in cells of the primary and secondary tumors, we conducted dual-color flow cytometric analysis. As evident in Fig. 2B, the subpopulation of double-stained vimentin⁺CD133⁺ cells (gated in sector B2 in the graphs) constituted 1.8% of YD15 cells. In contrast, the percentage of cells in sector B2 among YD15M cells was much greater: 16.9%; this result indicated that the metastasis associated with the double-stained CD133⁺vimentin⁺ cells showing features of EMT. To test whether the metastatic cells gained CSC-like properties, CD133 expression was assessed in vimentin⁺ cells during MACS (Fig. 2C). As expected, the percentage of vimentin⁺CD133⁺ cells (gated in sector B2) was significantly greater (20.0%) among YD15M cells than among YD15 cells (5.7%). This finding means that the population of vimentin⁺ cells acquired some properties of CSCs during metastasis.

We then tested whether the metastatic cells have the proliferative properties corresponding to acquisition of CSC characteristics. To this end, we quantified the expression of CD133, PCNA, and CK by western blotting (Fig. 3A). Downregulation of PCNA, but not CD133, was observed in YD15M cells, in comparison with YD15 cells. In addition, there was no significant difference in CK expression between the two cell lines. In the dual-color flow cytometric analysis for both PCNA and CD133, simultaneous strong expression of PCNA and CD133 (gated in sector B2) could not detected among either YD15 or YD15M cells (Fig. 3B). As shown in the figure, the decreased expression of cell proliferation marker PCNA (green) in YD15M cells was not accompanied by downregulation of CD133 (red); this change corresponds to cells with increased CSC-like properties (Fig. 3C). To confirm proliferation indicators in YD15M cells (according to the PCNA western blotting), we performed cell cycle analysis (Fig. 3D). The population of YD15M cells contained a higher percentage of cells in the G0/G1 phase (91.9%) than did the population of YD15 cells (61.15%). In contrast, the total percentage of cells in the S phase was 1.1% among YD15M cells; this figure is lower than that among YD15 cells (23.9%), confirming that YD15M cells had weak proliferative characteristics. Thus, we hypothesized that the weaker proliferation of the metastatic cell line YD15M may correspond to CSC-like properties.

We utilized the ALDEFLUOR assay to assess the presence and size of the subpopulation with ALDH enzymatic activity among primary cancer cells and metastatic cells (Fig. 4A). The ALDEFLUOR-positive subpopulation constituted 30.5% of YD15 cells. In contrast, it constituted 39.3% of YD15M cells. Furthermore, protein levels of ALDHA1, OCT4, and NANOG were higher in YD15M cells than in YD15 cells; in particular, OCT4 expression in YD15M cells was >5 times that in YD15 cells according to the western blot analysis (Fig. 4B). The increased ALDH activity and protein expression of ALDHA1, OCT4, and NANOG were suggestive of acquisition of stem cell-like properties during metastasis.