Human HNSCC SAS cells and the various derived cell lines, including SASVO3, SASVO3M-1 and SASVO3M-5, were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco-BRL). The medium was supplemented with 10% fetal bovine serum (FBS) (HyClone Laboratory), 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin sulfate. The cells were maintained at 37°C in a humidified atmosphere containing 5% CO₂.

Female nude mice, aged 6–8 weeks, were obtained from the National Laboratory Animal Breeding and Research Center, Taipei, Taiwan. All mice were maintained under standard pathogen-free laboratory conditions. Animal experiments were conducted according to the institutional guidelines established by the Animal Core Facility and IACUC of the Veterans General Hospital, Taipei, Taiwan. SASVO3 cells at a density of 5×10⁵ cells/100 μl PBS were injected into nude mice via the tail vein. The mice were euthanized 65 days after injection, and metastatic nodules were removed from the lungs. Lung tissues were subjected to pathological examination and in vitro cell culture studies. For histological analyses, lung tissues were fixed with buffered-formalin and embedded in paraffin. Sections were stained using standard hematoxylin and eosin protocol. For cell line generation, metastatic nodules were torn into tiny pieces and grown in medium consisting of DMEM and defined keratinocyte serum-free medium (KSFM) (Gibco-BRL) at a ratio of 1:2. The outgrowth epithelial tumor cells were cloned and adapted to complete DMEM. Two clones derived from the metastatic nodules were designated as SASVO3M-1 and SASVO3M-5. The abbreviations for SASVO3, SASVO3M-1 and SASVO3M-5 cells are VO3, M-1 and M-5, respectively.

Total RNA was extracted from the parental SAS cells, as well as the derived cell lines, using TRIzol reagent (Invitrogen Life Technologies). RNA purity was determined based on the OD260/280 ratio and electrophoretic analysis. Genome-wide gene expression profiling was performed using the human HT-12 v4 BeadChip array (Illumina, Inc.). This system utilizes >47,000 probes, which cover 29,000 transcripts. Data were extracted from GenomeStudio software (Illumina, Inc.). After quantile normalization, a total of 47,320 probes were included for analysis. The biological functions and pathways of the selected genes were interpreted by Ingenuity pathway analysis (Ingenuity^® Systems, Inc.)

Parental SAS cells and the derived cell lines were plated (1.8×10⁶ cells) onto 60-mm dishes and grown in DMEM containing 2% FBS for 6 days. Cells were subsequently fixed with methanol and stained with 10% Giemsa solution in PBS.

Primary antibodies against integrin ανβ6 (Millipore), uPAR (American Diagnostica, Inc.) and CD133 (Miltenyi Biotec) were used to determine the expression of proteins on the cell surface. APC goat anti-mouse (Ig) (Becton-Dickinson and Company) was used as a secondary antibody. The isotype antibody was purified mouse IgG1 (BioLegend). Primary and secondary antibodies were incubated serially at 4°C for 30 min. Analyses were performed using a FACScanto flow cytometer (BD Biosciences).

The antibodies used in this study included Snail (3895; Cell Signaling Technology, Inc.), Slug (ab27568; Abcam), Twist (sc-15393; Santa Cruz Biotechnology, Inc.), vimentin (550513) and fibronectin (610078; both from BD Biosciences), GAPDH (MAB374; Millipore), FAK (sc557), Tyr 925-phosphorylated FAK (sc-11766) and Oct4 (sc-9081) (all from Santa Cruz Biotechnology, Inc.), Nanog (ab21624; Abcam) and Bmi1 (2830; Cell Signaling Technology, Inc.). For immunoblotting, the cell pellets were homogenized in 200 ml lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 10% glycerol, 1.5 mM MgCl₂, 1% Triton X-100) and incubated on ice for 20 min. After centrifugation at 13,000 × g for 30 min at 4°C, the supernatant containing the protein extracts was collected. Protein concentrations were determined using the Bio-Rad protein assay. Immunoblot analyses were performed as described previously (12).

Parental SAS cells and the derived cell lines were plated on glass coverslips in 6-well plates. For E-cadherin immunofluorescence staining, cells were fixed with 4% paraformaldehyde/phosphate-buffered saline (PBS) and permeabilized with 0.5% (v/v) Triton X-100 in PBS, followed by incubation with blocking solution [1% BSA and 0.2% (v/v) Tween-20 in PBS]. Cells were incubated with primary antibodies against E-cadherin (BD Transduction Laboratories), and subsequently incubated with the secondary antibody conjugated with FITC (Sigma-Aldrich). Nuclei were counterstained with DAPI, and the cells were mounted using fluorescence mounting medium (Dako). Fluorescence images were captured using a Zeiss Axio Observer A1 microscope. For F-actin staining, cells were fixed with 4% paraformaldehyde, incubated in blocking solution and stained with TRITC-conjugated phalloidin (Sigma-Aldrich). The cells were subsequently counterstained with DAPI and examined using an Olympus FV1000 confocal microscope.

Cell migration assays were performed using an Oris™ cell migration assembly kit (Platypus Technologies) according to the manufacturer’s instructions. In brief, the assay utilizes Oris™ cell seeding stoppers to restrict cell seeding to the outer annular region of the wells. SAS cells and the derived cell lines were plated onto each well at the indicated density in serum-free DMEM and allowed to attach for 6 h at 37°C. The stopper was subsequently removed to form an unseeded region at the center of each well. An image of the clear detection zone was captured as a control for cell migration. The cells were then incubated at 37°C for 16 h to permit cell migration into the detection zone. The cells were stained with Calcein AM fluorescent dye (Sigma-Aldrich) for 1 h before images were captured using a Zeiss Axio Observer A1 microscope.

Cell invasion assays were performed using an Oris™ 96-well cell migration assay kit following the manufacturer’s instructions. In brief, a thin layer of basal membrane was coated by rinsing each well with 3.5 mg/ml BME coating solution, after which the seeding stopper was pressed down into each well. The cells were starved in medium containing 1% FBS for 18 h prior to plating. The cells were incubated for 6 h to permit cell attachment. The stopper was then removed, and 40 μl BME coating solution was added to create a 3-D BME layer. After incubating for 48 h, the cells were stained for the cell migration assay.

SAS cells and the derived cell lines were cultured in tumor sphere medium consisting of serum-free DMEM/F12 medium, N2 supplement (both from Gibco-BRL), 10 ng/ml human recombinant basic fibroblast growth factor-basic (FGF) and 10 ng/ml epidermal growth factor (EGF) (R&D Systems). Cells were plated at a density of 7.5×10⁴ to 1×10⁵ live cells/100-mm dish, and the medium was replaced every other day until tumor sphere formation was observed (~4 weeks) (7).

Nude mice were randomly grouped into sets of 5 mice each. Cells at the indicated cell number were individually injected into mice via the tail vein. The date of animal death or euthanasia was recorded, and the lung tissue from these mice was fixed with Bouin’s solution.

Kaplan-Meier survival was analyzed with the Statistical Package for the Social Sciences (SPSS) statistical program version 17.0 (SPSS, Inc.).

To establish a cellular model for investigating the progression of metastasis, we attempted to generate metastatic cell lines from the highly tumorigenic SASVO3 (VO3) cell line, which was previously developed from parental SAS cells (12). Empirically, we induced pulmonary metastatic tumors by injecting VO3 cells into the tail vein of nude mice. Pathological analysis revealed that the tumor cells subsequently found in the lung were capable of invading into the trachea (Fig. 1A). Two metastatic cell line clones were generated and designated as M-1 and M-5. The in vitro malignancy of VO3, M-1 and M5 were further examined by a focus formation assay. As shown in Fig. 1B, these metastatic cell lines exhibited significantly enhanced focus formation when compared with the original VO3 cells, even when the cells were not confluent. In contrast, the SAS cells with reduced tumorigenicity rarely formed foci, even at full confluence (Fig. 1B).

To compare the lung colonization activity of cell lines with different malignant potential, SAS cells and cells from the 3 derived cell lines were collected and injected into nude mice via the tail vein. The results showed that the metastatic M-1 and M-5 cells, even when injected at a 10-fold lower cell number than SAS or VO3 cells, were capable of inducing more pulmonary metastatic colonies and shortening the overall survival of the injected mice (Fig. 1C). In contrast, non-metastatic SAS cells rarely formed pulmonary metastatic colonies. The data thus revealed that the newly established metastatic cells possessed an enhanced ability to extravasate and proliferate in the lung parenchyma as compared to the ancestor VO3 cells.

To understand the transcriptional changes in the cell lines with different malignant potential, genome-wide gene expression profiling of SAS, VO3, M-1 and M-5 cells was carried out. Scatter plot analysis revealed a high correlation of metastatic M-1 and M-5 clones (R²=0.9941). After quantile normalization, 47,320 probes were subjected to 2 systemic analyses to measure the correlation of cells with differing malignant potential (heatmap) and to select the genes that displayed a marked alteration in parallel with malignancy (fold-change). A schematic of the transcriptome data analysis is shown in Fig. 2. The criteria of selection were: i) to limit false results arising from low-expression genes. Probes with detection of a p-value ≥0.05 in any of the 4 test samples (heatmap), or in either of the paired samples (fold-change) were excluded. ii) The ratio of probe intensity of VO3, M-1 or M-5 vs. that of SAS was calculated. iii) Cutoff values of the ratio were as indicated in Fig. 2.

As shown in Fig. 3, hierarchical clustering revealed a differential signature in cells with varying malignant potential. The signatures of the 2 metastatic cell lines were the most closely related to each other, but were most distinct from the non-metastatic SAS cells. However, the signature of highly tumorigenic VO3 cells was in-between the aforementioned cell lines and contained certain uniquely expressed genes. To transcriptionally profile these changes between the highly tumorigenic cells and metastatic cells, genes with a statistically significant difference in expression in the VO3, M-1 and M-5 cells were selected and subjected to pathway analysis. There were 825 and 972 ingenuity pathway analysis (IPA)-identified genes displaying significant changes in VO3 cells and with common preservation in M-1 and M-5 cells (designate M), respectively. Furthermore, a total of 535 overlapping genes were observed in both tumorigenic and metastatic cells, while 209 and 431 genes were uniquely found in VO3 cells and commonly reserved in M cells, respectively (Fig. 2, right). Biological pathways involved in VO3 (825) and common M (972) gene sets were then analyzed. Several canonical pathways were identified in both the VO3 and the common M groups, including the integrin signaling pathway, VDR/RXR activation pathway and IL-17A, which has a role in psoriasis. Of note, Notch signaling was found exclusively in the highly tumorigenic cells, while interferon signaling, RhoA signaling and Oct4, which plays a role in mammalian embryonic stem cell pluripotency, were significantly altered in the metastatic cells. Table I depicts the top 10 canonical pathways identified by IPA. Table II shows the top 5 upregulated and downregulated genes uniquely identified in the VO3 or metastatic cells. Collectively, the transcriptome analysis suggests that metastatic HNSCC cells may have profound changes associated with cell movement and stemness activity.

To verify the differential expression of cell movement-related genes identified from transcriptome analysis, the protein levels in the distinct cell lines of the above-mentioned genes were determined. Highly tumorigenic VO3 cells displayed upregulation of vimentin and fibronectin (Fig. 4A) and downregulation of the epithelial marker E-cadherin (Fig. 4C-a), suggesting that a gain of EMT may have occurred in VO3 cells. In the 2 metastatic clones, the expression patterns of the 3 genes listed above remained similar, without any significant change. Moreover, the protein levels of genes related to the EMT transcription factor Slug, focal adhesion (α_vβ₆ and p-FAK), and proteolytic degradation, urokinase-type plasminogen activator receptor (uPAR) were increased in the metastatic cells (Fig. 4A and B). However, no obvious changes in Snail or Twist proteins (EMT-related transcription factors) in the 4 tested cell lines were observed (Fig. 4A).

IPA analysis identified the upregulation of small GTPase RhoA signaling in the metastatic cell lines (Table I). To further verify downstream activation of RhoA signaling, the polymerization of F-actin was characterized using phalloidin staining. The metastatic M-5 clone displayed significant actin reorganization in lamellipodia, as well as filopodium formation, when compared to the parental SAS cells (Fig. 4C-b). The highly tumorigenic VO3 cells also showed augmentation of lamellipodia, but less of filopodia (Fig. 4C-b). The mobility assays revealed that the migratory ability of VO3 and M-5 cells was higher than that of the parental SAS cells (Fig. 4C-c). Additionally, the invasive ability of M-5 cells was moderately higher than that of VO3 cells (Fig. 4C-d). Taken together, our data revealed that actin-polymerization, cell migration activity and partial EMT changes occurred in the highly tumorigenic VO3 cells, while focal adhesion, filopodium formation and invasive activity were enhanced in the metastatic cells.

Transcriptome analysis revealed activation of the canonical Oct4 pathway, which plays a role in mammalian embryonic stem cell pluripotency, in the metastatic M-1 and M-5 cells (Table I). Hence, we examined the acquisition of stem cell properties in the metastatic cells by sphere formation which is one of the characteristics of stemness (7). The metastatic cells showed increased sphere formation ability in regards to both sphere size and number as compared to the parental SAS cells that rarely form spheres and VO3 cells that displayed moderate sphere formation ability (Fig. 5A and B). Furthermore, the expression of the stem cell markers, CD133, Oct4 and Bmi1, was also augmented in the metastatic cells, along with a moderate increase in Nanog protein levels (Fig. 5C and D).