Sorafenib Inhibits the Hepatocyte Growth Factor–Mediated Epithelial Mesenchymal Transition in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is the fifth most common cancer and the third largest cause of cancer-related death in the world annually (1). Recurrence, metastasis, and the development of new primary tumors are the most common causes of mortality among patients with HCC (2). Sorafenib (Nexavar; Bayer HealthCare Pharmaceuticals Inc.) is a small molecule that inhibits the kinase activities of Raf-1 and B-Raf in addition to VEGFRs, PDGFR-β (platelet-derived growth factor receptor β), Flt-3, and c-KIT (3). Two recent randomized controlled trials reported a clinical benefit of single-agent sorafenib in extending overall survival in both Western and Asian patients with advanced unresectable HCC (4, 5). The potential action mechanisms that lead to these clinical benefits are thought to include antiangiogenic effects and sorafenib's characteristic inhibitory effect on Raf-1 and B-Raf signaling.

Meanwhile, growing evidence indicates that the epithelial mesenchymal transition (EMT), a developmental process by which epithelial cells reduce intercellular adhesions and acquire fibroblastoid properties, has important roles in the development of the invasive and metastatic potentials of cancer progression (6–8). To date, numerous clinicopathologic studies have shown positive correlations between the expressions of the transcription factors SNAI1 (snail homologue 1/SNAIL) and SNAI2 (snail homologue 2/Slug), which are key inducible factors of the EMT, and poor clinical outcomes in breast, ovary, colorectal, and lung cancer; squamous cell carcinoma; melanoma, and HCC (reviewed in ref. 6).

Generally, the activation of a wide variety of ligands including FGF (fibroblast growth factor), TGF-β–BMPs (bone morphogenetic protein), Wnt, EGF (epidermal growth factor), VEGF, and HGF (hepatocyte growth factor) and its receptor can upregulate the expression of EMT-regulating transcription factors, including SNAI1, SNAI2, ZEB1, ZEB2, and TWIST (6). Among them, HGF (also known as scattering factor) activates the Met signaling pathway, thereby increasing the invasive and metastatic potentials of the cells and allowing the survival of cancer cells in the bloodstream in the absence of anchorage (9). In addition, HGF is well known as a potent angiogenic cytokine, and Met signal activation can modify the microenvironment to facilitate cancer progression (9). Therefore, the HGF-Met signaling pathway is regarded as a promising therapeutic target, and many molecular targeted drugs are under clinical development (10). In HCC, the mRNA levels of HGF and Met receptor are markedly increased compared with those in normal liver (11). A high serum HGF concentration is associated with a poor prognosis for overall survival after hepatic resection, and the serum level of HGF represents the degree of the carcinogenic state in the livers of patients with C-viral chronic hepatitis and cirrhosis (12–14). Thus, we examined the effect of sorafenib on the HGF-Met–mediated EMT in HCC.

Sorafenib was provided by Bayer HealthCare Pharmaceuticals Inc. U0126, wortmannin (Cell Signaling Technology), and human HGF (R&D Systems) were purchased from the indicated companies. The structures of compounds are shown in Supplementary Figure 1.

The human HCC cell lines HepG2 and Huh7 were maintained in Dulbecco's modified Eagle's (DMEM) medium (Sigma) supplemented with 10% FBS, penicillin, and streptomycin (Sigma) in a humidified atmosphere of 5% CO₂ at 37°C. The cell lines were obtained from the Japanese Collection of Research Bioresources and were grown in culture for less than 6 months.

The method used for the scratches assay has been previously described (15). Briefly, the cells were plated onto 24-well plates and incubated in DMEM containing 10% FBS until they reached subconfluence. Scratches were introduced to the subconfluent cell monolayer, using a plastic pipette tip. The cells were then cultured with DMEM containing 10% FBS at 37°C. After 24 hours, the scratch area was photographed using a light microscope (IX71; Olympus). The wound distance between edge to edge were measured and averaged from 5 points per 1 wound area, using DP manager software (Olympus). The 2 wound areas were evaluated in an experiment and the experiment was done in triplicate.

The migration assays were done using the Boyden chamber methods and polycarbonate membranes with an 8-μm pore size (Chemotaxicell), as previously described (15). The membranes were coated with fibronectin on the outer side and dried for 2 hours at room temperature. The cells to be analyzed (2 × 10⁴ cells/well) were then seeded onto the upper chambers with 200 μL of migrating medium (DMEM containing 0.5% FBS), and the upper chambers were placed into the lower chambers of 24-well culture dishes containing 600 μL of DMEM containing 10% FBS or with 10 ng/mL of HGF or with HGF and 10 μmol/L of sorafenib. After incubation for 36 hours (HepG2) and 24 hours (Huh7), the media in the upper chambers were aspirated and the nonmigrated cells on the inner sides of the membranes were removed using a cotton swab. The cells that had migrated to the outer side of the membranes were fixed with 4% paraformaldehyde for 10 minutes, stained with 0.1% Giemsa stain solution for 15 minutes, and then counted using a light microscope. Migrated cells were averaged from 5 fields per 1 chamber and 3 chambers were used on 1 experiment. The experiment was done in triplicate.

HepG2 and Huh7 cells (2 × 10⁴ and 1 × 10⁴ cells/well, respectively) were seeded in 6-well tissue culture dishes. After 24 hours of incubation, the cells were stimulated with 10 ng/mL of HGF or control PBS. When the inhibitors were used, the cells were exposed to each inhibitor for 3 hours before the addition of HGF. After 48 hours, the cells were analyzed using a light microscope. The experiment was done in triplicate.

The following antibodies were used in this study: phospho-Met (Y1349), Met, phospho-AKT (S473), AKT, phospho-p44/42 mitogen-activated protein kinase (MAPK), SNAI1/Snail, E-cadherin, N-cadherin, vimentin, β-actin antibody horseradish peroxidase–conjugated secondary antibody (Cell Signaling Technology), and fibronectin (Santa Cruz Biotechnology). All the experiments were done at least in duplicate. The Western blot analysis was done as described previously (16). The data were quantified by automated densitometry using Multigauge Ver. 3.0 (Fujifilm). Densitometric data were normalized by β-actin in triplicate and the average was shown above the Western blot as a ratio of control sample.

The real-time reverse transcription PCR (RT-PCR) method has been previously described (17). Briefly, 1 μg of total RNA from the cultured cells was converted to cDNA using a GeneAmp RNA-PCR kit (Applied Biosystems). Real-time RT-PCR amplification was done using a Thermal Cycler Dice (Takara) in accordance with the manufacturer's instructions under the following conditions: 95°C for 6 minutes, 40 cycles of 95°C for 15 seconds, and 60°C for 1 minute. Glyceraldehyde 3-phosphate dehydrogenase (GAPD) was used to normalize the expression levels in the subsequent quantitative analyses. To amplify the target genes, the following primers were purchased from TaKaRa: CDH1, forward 5′-TTA AAC TCC TGG CCT CAA GCA ATC-3′ and reverse 5′-TCC TAT CTT GGG CAA AGC AAC TG-3′; CDH2, forward 5′-CGA ATG GAT GAA AGA CCC ATC C-3′ and reverse 5′-GGA GCC ACT GCC TTC ATA GTC AA-3′; SNAI1, forward 5′-TCT AGG CCC TGG CTG CTA CAA-3′ and reverse 5′-ACA TCT GAG TGG GTC TGG AGG TG-3′; SNAI2, forward 5′-ATG CAT ATT CGG ACC CAC ACA TTA C-3′ and reverse 5′-AGA TTT GAC CTG TCT GCA AAT GCT C-3′; VIM, forward 5′-TGA GTA CCG GAG ACA GGT GCA G-3′ and reverse 5′-TAG CAG CTT CAA CGG CAA AGT TC-3′; FN1, forward 5′-GGA GCA AAT GGC ACC GAG ATA-3′ and reverse 5′-GAG CTG CAC ATG TCT TGG GAA C-3′; and GAPD, forward 5′-GCA CCG TCA AGG CTG AGA AC-3′ and reverse 5′-ATG GTG GTG AAG ACG CCA GT-3′.

Three different sequences of small interfering RNA (siRNA) targeting human SNAI1 (Hs_SNAI1_9785, 9786, and 9787) and those of 2 scramble control siRNAs were purchased from Sigma Aldrich Japan. The transfection methods have been previously described (17).

The statistical analyses were done using Microsoft Excel (Microsoft) both to calculate the SD and to test for statistically significant differences between the samples using a Student t test. A value P < 0.05 was considered statistically significant.

To examine the activity of HGF-Met signaling in HCC cells, we examined the expressions of phospho-Met, Met, phospho-AKT, AKT, phospho-MAPK, and MAPK in the HepG2 and Huh7 cell lines, using Western blotting. The phosphorylation levels of Met, AKT, and MAPK were dose-dependently increased by HGF stimulation (Fig. 1A). A time-course analysis showed that the phosphorylation levels of Met, AKT, and MAPK peaked at 1 to 2 hours after HGF stimulation and gradually recovered to the baseline values at 4 hours later (Fig. 1B). These results indicated that Met signaling is actually capable of being activated in response to HGF in HCC cells.

From a morphologic aspect, the EMT is characterized by an increase in cell scattering and an elongation of the cell shape (18). To evaluate whether HGF mediates the morphologic change that is characteristic of the EMT in HCC cells, cellular morphology was examined after HGF stimulation. HGF clearly mediated both cell scattering and the elongation of the cell shape in HepG2 and Huh7 cell lines (Fig. 1C). These data indicate that HGF mediates the morphologic changes that are compatible with the induction of the EMT in HCC cell lines.

Because SNAI1 and SNAI2 are considered to be master regulators of the EMT, changes in the mRNA expression levels of EMT-related genes in response to HGF stimulation were evaluated using real-time RT-PCR (Fig. 2A). HGF stimulation upregulated SNAI1 mRNA expression by more than 2-fold, whereas the baseline expression of SNAI2 was very low compared with that of SNAI1 and did not respond to HGF in either of the HCC cell lines that were examined. Cadherin switching, which is characterized by the downregulation of E-cadherin and the upregulation of N-cadherin, is known as one of the most pivotal cellular events in the EMT (19). Cadherin switching was clearly observed on the basis of mRNA levels after HGF stimulation. The mesenchymal marker fibronectin was also upregulated (Fig. 2A).

Consistent with the mRNA changes, HGF stimulation dose-dependently upregulated the protein expression of SNAI1, N-cadherin, fibronectin, and vimentin and downregulated the expression of E-cadherin in both cell lines (Fig. 2B). Vimentin expression of HepG2 was not detected (baseline vimentin mRNA was also extremely low; data not shown). A time-course analysis showed that HGF upregulated the SNAI1 expression at 2 hours after stimulation and that the expression level recovered to the baseline value at 24 hours thereafter (Fig. 2C). Cadherin switching after HGF stimulation was observed at 8 hours later in HepG2 cells and 48 hours later in Huh7 cells (Fig. 2C and D). Generally, upregulation of SNAI1 is observed within few hours, but cadherin switching occurs around 24 hours later after stimulation (20, 21), consistent with our result. These results indicate that HGF meditates the induction of SNAI1, cadherin switching, and the EMT in HCC cells.

Besides SNAI1 and SNAI2, other transcription factors of several genes also have the potential to repress E-cadherin and to induce the EMT; these factors include ZEB1/TCF8, ZEB2/SMAD interacting protein 1, TWIST, E47/TCF3, and TCF4/E2-2 (6). Therefore, we examined whether SNAI1, among several EMT-inducible genes, has a central role in the HGF-mediated EMT in HCC cells. Three sequences of SNAI1-siRNA (1, 2, and 3) were used. A Western blot showed that both sequences 2 and 3 of SNAI1-siRNA completely suppressed the HGF-mediated upregulation of SNAI1 in the HepG2 and Huh7 cells (Fig. 3A); thus, the #2 SNAI1-siRNA was used in the following experiments: The siRNA knockdown of SNAI1 canceled the morphologic changes observed in HepG2 cells undergoing HGF-mediated EMT, whereas the control-siRNA did not (Fig. 3B). Similar results were obtained in Huh7 cells, indicating that SNAI1 is required for the morphologic changes observed in HGF-mediated EMT. Similarly, the siRNA knockdown of SNAI1 strongly canceled the HGF-mediated downregulation of E-cadherin and the upregulation of N-cadherin in both HepG2 and Huh7 cells (Fig. 3C). Those of mRNA expression changes were relatively correlated with the results of Western blot, except for N-cadherin in Huh7 cells (Supplementary Fig. 2A). Regarding the cellular migration, the siRNA knockdown of SNAI1 inhibited the HGF-mediated cellular migration (Fig. 3D). Collectively, these results indicate that SNAI1 is required to induce the HGF-mediated EMT in HCC cells.

In general, SNAI1 expression is regulated by ligand-receptor signal transduction through a downstream signal pathway that includes the Smad, MAPK, AKT, and GSK3 pathways (6, 22, 23). Sorafenib has been shown to inhibit RAF-MAPK signaling in HCC cells (24). Accordingly, we hypothesized that sorafenib might downregulate SNAI1 expression by inhibiting RAF-MAPK signaling, which is a unique activity of sorafenib. As expected, sorafenib and the MEK inhibitor U0126 (10 μmol/L) markedly inhibited the HGF-induced phosphorylation of MAPK, but the PI3K inhibitor wortmannin (10 μmol/L) did not. In contrast, only wortmannin inhibited the phosphorylation of AKT (Fig. 4A). Notably, SNAI1 expression was strongly downregulated by sorafenib and U0126 but not by wortmannin (Fig. 4A). These results showed that sorafenib downregulated SNAI1 expression via MAPK signaling. Meanwhile, we examined the HGF- and sorafenib-mediated expression changes of SNAI2, ZEB1, ZEB2, and TWIST using real-time RT-PCR and Western blot (Supplementary Fig. 3). Baseline and expression changes of SNAI2 and TWIST were very low compared with SNAI1, and the expression changes of ZEB1 and ZEB2 seemed not to be significant. Collectively, we considered that SNAI2, TWIST, ZEB1, and ZEB2 are not likely to be involved in the effect of HGF and sorafenib on EMT in this cell lines. Then, we examined the activity of sorafenib on HGF-mediated morphologic changes in HCC cells. HGF stimulation mediated the cell scattering and spindle-shaped changes, and these effects were clearly canceled by sorafenib and U0126, but not by wortmannin, in both HepG2 and Huh7 cells (Fig. 4B). These results were consistent with the results of Western blotting. To show whether sorafenib cancels the effect of HGF-mediated downregulation of E-cadherin, we examined the Western blot in dose–response analysis. Downregulation of E-cadherin was clearly canceled by sorafenib in a dose-dependent manner (Fig. 4C). Time-course analysis showed that HGF-mediated downregulation of E-cadherin was also canceled by sorafenib (Supplementary Fig. 4). HGF stimulation downregulated E-cadherin expression and upregulated N-cadherin, vimentin, and fibronectin in HCC cells; however, these effects were canceled by sorafenib in both HCC cell lines (Fig. 4D and Supplementary Fig. 2B). The mRNA data of N-cadherin in Huh7 cells were not correlated with protein level. These results show that sorafenib inhibits the RAF-MAPK pathway, thereby downregulating SNAI1 and inhibiting the EMT in HCC.

Because sorafenib inhibits the HGF-mediated EMT in HCC cells, we next examined whether the inhibitory effect of sorafenib on the EMT leads to an inhibition of cellular migration in HCC cells. A scratch assay revealed that HGF stimulation increased cellular migration by about 2- fold in both HCC cell lines; however, sorafenib significantly inhibited this effect to the baseline levels (Fig. 5A). Similarly, a migration assay using the Boyden chamber method revealed that sorafenib canceled HGF-mediated cellular migration in both cell lines (Fig. 5B). These results suggest that sorafenib actually inhibits the cellular migrating phenotype of the EMT in HCC cells. The combination of migration data with siRNA and sorafenib (Fig. 3D and Fig. 5B) suggests that inhibitory effects of sorafenib on migration may be mediated by Snail downregulation in some tumors (e.g., Huh7) but not in others (e.g., HepG2). It is assumed that the inhibitory activity of sorafenib on the cellular migrating phenotype is due to its inhibitory effect of Raf-MAPK signaling pathway (Fig. 4A and B). Regarding HGF-dependent PI3K-AKT signaling pathway, wortmannin weakly inhibited the wound closure in Huh7 cells and to the same extent by sorafenib in HepG2 cells (Supplementary Fig. 5). In contrast, wortmannin has no effect on Snail levels or on HCC morphology changes (Fig. 4A and B). Collectively, we speculate that activation of HGF-dependent PI3K-AKT pathway may not be involved in SNAI1 induction or morphologic change but at least partially involved in cell migration independent of Raf-MAPK-SNAI1 signaling.

Taken together, these results indicate that sorafenib inhibits the HGF-mediated EMT, which is characterized by cadherin switching, morphologic changes, and an increase in the cellular migrating phenotype, by inhibiting Raf-MAPK signaling, resulting in the downregulation of SNAI1 in HCC cells (Fig. 6).

Recent accumulating evidence has shown that the EMT is involved in drug sensitivity to several anticancer agents (25). Within this topic, the most intensively investigated drugs have been endothelial growth factor receptor (EGFR)–targeting drugs for the treatment of lung cancer. A clinical trial has revealed that lung cancer cells with strong E-cadherin expression exhibit a significantly longer time to progression after EGFR-TKI (tyrosine kinase inhibitor) treatment (26). Other studies on EGFR-targeting drugs have shown that mesenchymal type lung cancer cells exhibit an EMT-dependent acquisition of PDGFR, FGF receptor, and TGF-β receptor signaling pathways (27), and integrin-linked kinase is a novel target for overcoming HCC resistance to EGFR inhibition (28). Regarding sensitivity to gemcitabine, mesenchymal type cancer cells are reportedly associated with gemcitabine resistance in pancreatic cancer cells (29). The mechanism of resistance to gemcitabine has been shown to involve the activation of Notch signaling, which is mechanistically linked with the mesenchymal chemoresistance phenotype of pancreatic cancer cells (30). Thus, baseline cellular characteristics based on the EMT phenotype might be useful not only as prognostic biomarkers for a malignant phenotype but also as predictive markers of sensitivity to anticancer agents. In this study, we focused on the signaling pathway responsible for inducing the EMT and showed that the multitarget TKI sorafenib downregulates SNAI1 by inhibiting Raf-MAPK signaling, thereby inhibiting the HGF-mediated EMT in HCC cells. Our findings may provide a novel insight into the actions of TKIs and their anti-EMT effects.

The mechanisms underlying the SNAI1-induced metastatic and aggressive phenotypes of cancer cells have recently been intensively investigated in both basic and clinical research studies. A novel aspect of the activity of SNAI1 is its involvement in immunosuppression. The SNAI1-induced EMT mediates regulatory T cells and impairs dendritic cells, accelerating cancer metastasis not only by enhancing invasion but also by inducing immunosuppression (31). A complex of histone deacetylase (HDAC) and SNAI1 plays an essential role in silencing E-cadherin (32), suggesting that the use of HDAC inhibitors to inhibit SNAI1 function might represent a promising therapeutic approach. On the other hand, large-scale clinical data on SNAI1 expression and the prognosis of patients with HCC were recently reported (33) and the overexpression of SNAI2 and/or TWIST was correlated with a worse prognosis. In contrast, no such significant differences were observed in samples that overexpressed SNAI2. The coexpression of Snail and TWIST was correlated with the worst prognosis for HCC (33). This evidence suggests that SNAI1 might be a useful therapeutic target for oncology. Our findings showed that sorafenib completely canceled the HGF-mediated SNAI1 induction in HepG2 and Huh7 cells. This activity of sorafenib, in addition to sorafenib's antiangiogenic effects, might contribute to a clinical benefit against metastatic and aggressive phenotypes in patients with HCC.

No potential conflicts of interest were disclosed.

We thank Tomoko Kitayama and the staff of the Life Science Research Institute for their technical assistance.

This work was supported in part by the Third-Term Comprehensive 10-Year Strategy for Cancer Control and a Grant-in-Aid for Cancer Research (H20-20-9) from the Ministry of Health and Labor Scientific Research Grants.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.