Huh7 and Hep3B HCC cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 µg/ml penicillin and 0.25 µg/ml streptomycin (all from Invitrogen, Carlsbad, CA, USA). SKHep1, HepG2 and PLC/PRF/5 cells were grown in minimum essential medium (MEM) supplemented with 10% FBS, 100 µg/ml penicillin and 0.25 µg/ml streptomycin (all from Invitrogen). All of the cell lines were maintained at 37°C in an atmosphere of 5% CO₂.

The small interfering RNA (siRNA) method was used to knock down the expression of ELK3. Three different siRNAs targeting ELK3 (5′-CCAAGAUCUCCUCUUUAAUtt-3′; 5′-GGACUCAUUAACUGAUGAAtt-3′; and 5′-GGUCUCUAGUAGAAUUUCAtt-3′) (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were pooled together and used at a final concentration of 30 nM. Control cells were subjected to mock transfection with scrambled siRNA. The cells were transfected using G-fectin (Genolution Pharmaceuticals, Seoul, Korea) according to the manufacturer's protocol.

The cells were harvested using 0.5 mM trypsin/EDTA (Invitrogen) and subsequently incubated with phycoerythrin (PE)-conjugated anti-CD133/1 and allophycocyanin (APC)-conjugated anti-CD44 (Miltenyi Biotec, Auburn, CA, USA) at 4°C for 10 min. The LCSCs were sorted from the Huh7 cells using flow cytometry (MoFlo XDP; Beckman Coulter, Miami, FL, USA) with antibodies against CD133/1 and CD44. An isotype-matched mouse IgG was used as the negative control.

The cells were seeded in multiple ultra-low attachment 24-well plates (Corning Costar Corp., Cambridge, MA, USA) at a density of 2×10² cells/well in serum-free DMEM/F12K with B27 supplement (both from Invitrogen), basis fibroblast growth factor (bFGF) (20 ng/ml) and epidermal growth factor (EGF) (20 ng/ml) (both from PeproTech, Rocky Hill, NJ, USA). The cells were incubated at 37°C in an atmosphere of 5% CO₂ for 5 days. Then, the spheres (diameter >50 µm) in each well were counted using an inverted microscope (Olympus, Tokyo, Japan). The average number of spheres was calculated from three independent experiments.

Isolated Huh7 cells were grown for 3 days in 10% FBS complete medium. Then, the cells were harvested using 0.5 mM trypsin/EDTA (Invitrogen) and resuspended in serum-free medium. Cell migration was evaluated using 6.5-mm Transwell plates with 8-µm pore size filters (Corning Costar Corp.). The resuspended cells were seeded into the upper chamber (3×10⁴ cells) of the Transwell, and 600 µl of 10% FBS complete medium was added to the bottom chamber and used as a chemoattractant. The cells were incubated for 48 h at 37°C in an atmosphere of 5% CO₂. The cells that had not migrated were removed using a cotton swab. The cells that had migrated were stained using a Diff-Quick™ Three-Step Stain kit (Sysmex, Kobe, Japan) and mounted on glass slides. The cells in 4 randomly selected fields (magnification, ×200) from each slide were quantified using a slide scanner (Pannoramic MIDI; 3DHISTECH Ltd., Budapest, Hungary).

The Transwell invasion assays were conducted using a Transwell plate (Corning Costar Corp.) coated with Matrigel (BD Biosciences, San Jose, CA, USA) according to the manufacturer's instructions. Briefly, the Matrigel (BD Biosciences) was diluted with serum-free medium at a dilution of 1:5, added to the upper chamber, and incubated at 37°C until it had completely solidified. The unbound Matrigel was removed by washing. Cells in serum-free medium were seeded into the upper chamber (5×10⁴ cells), and 600 µl of complete medium was added to the bottom chamber and used as a chemoattractant. The cells were incubated for 48 h at 37°C in an atmosphere of 5% CO₂, and the non-invasive cells were removed using a cotton swab. The cells that had invaded the membrane were stained using a Diff-Quick™ Three-Step Stain kit and mounted on glass slides. The stained cells were counted using a slide scanner (Pannoramic MIDI) (magnification, ×200).

The protein extracts were separated using 8% SDS-PAGE and transferred to nitrocellulose membranes (Schleicher & Schuell, Dassel, Germany). The membranes were blocked in 5% skim milk and incubated with the indicated primary antibodies according to the manufacturer's protocol. The following primary antibodies were used in the western blot assays: anti-ELK3, anti-VEGF (both from Abcam, Cambridge, MA, USA), anti-HIF-1α (Santa Cruz Biotechnology) and anti-β-actin (Sigma, St. Louis, MO, USA). The membranes were subsequently incubated with horseradish peroxidase (HRP)-conjugated anti-mouse or anti-rabbit secondary antibodies (Santa Cruz Biotechnology). The target proteins were visualized using an enhanced chemiluminescence reagent (Amersham Biosciences, Cardiff, UK). The density of each band was measured using Multi Gauge imaging software (Raytest, Straubenhardt, Germany).

Total RNA was purified from CD133⁺/CD44⁺ and CD133⁻/CD44⁻ cells using TRIzol reagent (Invitrogen). The cDNA was synthesized from 2 µg of total RNA, and ELK3 was amplified using PCR with ELK3-specific primers (forward, 5-ACCCAAAGGCTTGGAAATCT-3 and reverse, 5-TGTATGCTGGAGAGCAGTGG-3′). The expression level of the target gene was normalized to the expression of the β-actin internal control. β-actin was amplified using PCR with β-actin-specific primers (forward, 5′-GGCACCCACCCTTCTACATAGA-3′ and reverse, 5′-CCCTCGTAGATGGGCACACT-3′). The PCR products were separated using electrophoresis on a 1% agarose gel with 0.5 µg/ml ethidium bromide (Research Genetics, Huntington, AL, USA) and visualized using a Gel-Doc system (Bio-Rad, Vienna, Austria).

Human VEGF Quantikine ELISA kits (R&D Systems, Minneapolis, MN, USA) were used to measure VEGF levels in cell supernatants according to the manufacturer's instructions. Briefly, 200 µl of the harvested cell supernatants was dispensed into wells coated with coating buffer and incubated for 2 h at room temperature (RT). After the wells were completely washed, the VEGF conjugate was dispensed into each well, and the wells were incubated at RT. Then, the plates were developed using substrate solution for 20 min. The absorbance at 450 nm was measured using a microplate reader.

The cell supernatants were collected and mixed with non-denatured sample buffer [1 M Tris-Cl (pH 6.8), 1% bromophenol blue and 20% SDS]. The resulting cell supernatant solution was separated on a polyacrylamide gel containing 1 mg/ml gelatin (Sigma). The gels were renatured with 2.5% Triton X-100 and incubated in activation buffer [50 mM Tris buffer (pH 7.4), 5 mM CaCl₂ and 1 µM ZnCl₂) at 37°C for 12 h. Then, the gels were stained with Coomassie Brilliant Blue R-250 for 20 min and destained with destaining buffer (45% methanol and 10% acetic acid). MMP-2 appeared as a clear area and the band density was measured using Multi Gauge imaging software.

The data are presented as the mean ± SD. of at least 3 separate experiments. The statistical analyses were conducted using a two-tailed Student t-test. p<0.05 and p<0.01 indicate statistical significance.

To identify a putative marker of LCSCs, we evaluated the expression of the cell surface markers CD133 and CD44 in 5 HCC cell lines using flow cytometry. The expression of CD133 and CD44 was significantly different in each cell line. As shown in Table I, the proportion of CD133⁺ cells was highest in the PLC/PRF/5 cells (74.6%) and lowest in the Hep3B cells (0.6%), whereas the proportion of CD44⁺ cells was highest in the Hep3B cells (99.6%) and lowest in the PLC/PRF/5 cells (2.1%). The proportion of CD133⁺/CD44⁺ cells was highest in the Huh7 cells (11.1%) and lowest in the Hep3B cells (0.6%). As the co-expression of 2 or more CSCs markers is more specific compared with the presence of a single LCSCs marker, we selected the Huh7 cell line for further experiments. To explore the biological properties of the CD133⁺/CD44⁺ LCSC subpopulation of Huh7 cells, we compared the clonogenic potential of CD133⁺/CD44⁺ Huh7 cells and non-CD133⁺/CD44⁺ Huh7 cells. Huh7 cells were sorted into 4 subgroups according to the expression of the CD133 and CD44 surface markers using flow cytometry (Fig. 1A), and the clonogenic potential of the subgroups was examined using the sphere formation assay. The CD133⁺/CD44⁺ subgroup formed the largest spheres (Fig. 1B). In addition, the number of spheres was significantly higher in the CD133⁺/CD44⁺ subgroup (15 spheres/well) compared with the CD133⁻/CD44⁻ subgroup (9 spheres/well) (Fig. 1C). These results demonstrated that CD133⁺/CD44⁺ LCSCs have greater clonogenic potential compared with non-CD133⁺/CD44⁺ cells.

Recent studies have revealed that high expression levels of CD133 or CD44 contribute to CSC-associated cancer metastasis and invasion (10,14,39). Therefore, we performed in vitro Transwell migration and invasion assays to evaluate the metastatic potential of the Huh7 cell subpopulations. The migratory cells in each well were stained using a Diff-Quick Three-Step stain (Fig. 2A, upper panel), and the number of migratory cells in each group were compared. The number of migratory cells in the CD133⁺/CD44⁺ subgroup increased ~4-fold compared with the number in the CD133⁻/CD44⁻ subgroup (Fig. 2B; p<0.01). The results obtained from the Matrigel-coated Transwell invasion assay were consistent with these findings. The number of CD133⁺/CD44⁺ cells that invaded the Matrigel was significantly greater compared with this number in the non-CD133⁺/CD44⁺ subgroups (Fig. 2A, lower panel). The number of invasive cells in the CD133⁺/CD44⁺ subgroup was ~4-fold greater when compared with the CD133⁻/CD44⁻ subgroup (Fig. 2B; p<0.05). Together, these results indicated that the cell migration and invasion activities were significantly elevated in the CD133⁺/CD44⁺ LCSCs.

In a previous study, we demonstrated that ELK3 expression plays an important role in the process of EMT during liver fibrogenesis (26). Acquiring cell motility is necessary for the transition from an epithelial to a mesenchymal phenotype (27,28). We used western blot assays to determine whether ELK3 protein levels differ between CD133⁺/CD44⁺ and CD133⁻/CD44⁻ cells. As shown in Fig. 3A, the ELK3 protein level was increased ~2.5-fold in the CD133⁺/CD44⁺ cells compared with the level in the CD133⁻/CD44⁻ cells (p<0.01). Next, we conducted RT-PCR to measure mRNA expression levels of ELK3 in both subpopulations of cells. Similar to the ELK3 protein level, the ELK3 mRNA expression level was increased nearly 1.5-fold in the CD133⁺/CD44⁺ cells compared with the level in the CD133⁻/CD44⁻ cells (Fig. 3B; p<0.05). These results indicated that ELK3 expression was upregulated in the CD133⁺/CD44⁺ cells and that the upregulation of ELK3 enhanced HCC cell migration and invasion.

To further evaluate the role of ELK3 in CD133⁺/CD44⁺ cell migration and invasion, we transfected CD133⁺/CD44⁺ cells with an siRNA targeting ELK3. Western blot analysis of the siRNA silencing efficiency demonstrated that ELK3 expression decreased by ~60% in the siELK3-transfected cells when compared with the negative control cells (Fig. 4A; p<0.01). The clonogenic potential of ELK3-knockdown CD133⁺/CD44⁺ cells was measured using a sphere formation assay. As shown in Fig. 4B, the size of the spheres formed by the ELK3-knockdown CD133⁺/CD44⁺ cells was significantly decreased compared with the size of the spheres formed by the negative control cells. The control CD133⁺/CD44⁺ cells formed an average of 15 spheres, whereas the ELK3-knockdown CD133⁺/CD44⁺ cells formed less than half of that number (Fig. 4C; p<0.05). Next, we conducted a Transwell cell migration assay in the ELK3-knockdown CD133⁺/CD44⁺ cells (Fig. 4D, upper panel). Notably, the number of migratory cells was decreased by ~70% in the ELK3-knockdown CD133⁺/CD44⁺ cells compared with this number noted in the negative control cells (Fig. 4E; p<0.05). In addition, similar results were obtained using the Matrigel-coated Transwell invasion assay (Fig. 4D, lower panel). The number of cells that invaded the Matrigel was decreased by ~90% in the ELK3-knockdown CD133⁺/CD44⁺ cells compared with this number in the negative control group (Fig. 4E; p<0.05). Collectively, these results indicate that ELK3 expression is associated with the migration and invasion of CD133⁺/CD44⁺ cells.

To investigate the molecular mechanism underlying the ELK3-mediated properties of CD133⁺/CD44⁺ cells, we focused on the relationship between ELK3 and expression of its target gene HIF-1α. Previous studies revealed that several genes including Egr-1, PAI-1 and HIF-1α are ELK3 target genes by forming ternary complexes. Among these target genes, HIF-1α is known to stimulate angiogenic and metastatic responses by activating transcription of the genes encoding several growth factors, including VEGF and MMP-2 (19–23,35,36).

To this end, we evaluated the expression of HIF-1α in the ELK3-knockdown CD133⁺/CD44⁺ cells. The HIF-1α protein level was decreased by ~70% in the ELK3-knockdown CD133⁺/CD44⁺ cells compared with this level noted in the negative control cells (Fig. 5A; p<0.01). The VEGF protein level was significantly decreased by ~60% in the ELK3-knockdown CD133⁺/CD44⁺ cells compared with this level in the negative control cells (Fig. 5A; p<0.05), and a VEGF-specific ELISA demonstrated that VEGF secretion was decreased by ~50% in the ELK3-knockdown CD133⁺/CD44⁺ cells compared with the secretion observed in the negative control group (Fig. 5B; p<0.05). These result suggested that ELK3 knockdown has an inhibitory effect on VEGF-induced angiogenesis.

Therefore, we performed a gelatin zymography assay to determine the enzymatic activity of MMP-2 in the ELK3-knockdown CD133⁺/CD44⁺ cells since recent studies have demonstrated that MMP-2 participates in the cancer metastasis process and is also known to correlate with HIF-1α expression (36,38). As shown in Fig. 5C, the enzymatic activity of MMP-2 was decreased by ~50% in the ELK3-knockdown CD133⁺/CD44⁺ cells compared with that in the negative control group (p<0.05). Together, these results indicate that silencing of ELK3 expression in the CD133⁺/CD44⁺ cells can decrease the activity of metastasis-associated molecules by inhibiting HIF-1α expression.

Collectively, these data demonstrated that ELK3 overexpression promotes metastasis in CD133⁺/CD44⁺ cells and that inhibiting ELK3 expression attenuates the metastatic potential of CD133⁺/CD44⁺ LCSCs (Fig. 6).