Runt-related transcription factor 3 reverses epithelial–mesenchymal transition in hepatocellular carcinoma

Authors

  • Shigetomi Tanaka,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
    • S.T. and H.S. contributed equally to this work

  • Hidenori Shiraha,

    Corresponding author
    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    • Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, 2-5-1 Shikata-cho, Okayama 700-8558, Japan
    Search for more papers by this author
    • S.T. and H.S. contributed equally to this work

    • Tel.: [+81-86-235-7219], Fax: +[81-86-225-5991]

  • Yutaka Nakanishi,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Shin-Ichi Nishina,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Minoru Matsubara,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Shigeru Horiguchi,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Nobuyuki Takaoka,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Masaya Iwamuro,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Junro Kataoka,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Kenji Kuwaki,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Hiroaki Hagihara,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Junichi Toshimori,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Hideki Ohnishi,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Akinobu Takaki,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Shinichiro Nakamura,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Kazuhiro Nouso,

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Takahito Yagi,

    1. Department of Gastroenterological Surgery, Transplant, and Surgical Oncology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author
  • Kazuhide Yamamoto

    1. Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine and Dentistry, Okayama, Japan
    Search for more papers by this author

Abstract

Loss or decreased expression of runt-related transcription factor 3 (RUNX3), a tumor suppressor gene involved in gastric and other cancers, has been frequently observed in hepatocellular carcinoma (HCC). The objective of this study was to identify the regulatory mechanism of the epithelial–mesenchymal transition (EMT) by RUNX3 in HCC. Human HCC cell lines, Hep3B, Huh7, HLF and SK-Hep1, were divided into low- and high-EMT lines, based on their expression of TWIST1 and SNAI2, and were used in this in vitro study. Ectopic RUNX3 expression had an anti-EMT effect in low-EMT HCC cell lines characterized by increased E-cadherin expression and decreased N-cadherin and vimentin expression. RUNX3 expression has previously been reported to reduce jagged-1 (JAG1) expression; therefore, JAG1 ligand peptide was used to reinduce EMT in RUNX3-expressing low-EMT HCC cells. Immunohistochemical analyses were performed for RUNX3, E-cadherin, N-cadherin and TWIST1 in 33 human HCC tissues, also divided into low- and high-EMT HCC, based on TWIST1 expression. E-cadherin expression was correlated positively and N-cadherin expression was correlated negatively with RUNX3 expression in low-EMT HCC tissues. Correlations between EMT markers and RUNX3 mRNA expression were analyzed using Oncomine datasets. Similarly, mRNA expression of E-cadherin was also significantly correlated with that of RUNX3 in low-EMT HCC, while mRNA expression of JAG1 was negatively correlated with that of RUNX3. These results suggest a novel mechanism by which loss or decreased expression of RUNX3 induces EMT via induction of JAG1 expression in low-EMT HCC.

Hepatocellular carcinoma (HCC) is a human cancer with a high mortality rate worldwide.1–3 Although the prognosis for HCC has improved in recent years, the prognosis for advanced disease remains poor.4–9 Tumor progression in HCC is associated with altered expression of multiple molecules, including growth factors and their receptors, oncogenes, tumor suppressor genes and transcription factors.1, 10 Gaining a better understanding of the molecular mechanisms underlying tumor progression would aid in the development of new therapies for advanced HCC.

The epithelial–mesenchymal transition (EMT) is considered a critical event in the malignant transformation of cancer, especially during metastasis.11, 12 Several models demonstrating the contribution of EMT to the progression of established tumors in transformed cells have been developed,12–14 including HCC cells.15 Because EMT is characterized by loss of cell–cell adhesion and leads to cell individualization, EMT is known to be involved in increasing cell motility. Normally, adherent junctions connect cells via homotypic interactions mediated by E-cadherin molecules, the prototypic member of the cadherin superfamily. Not surprisingly, therefore, an inverse correlation has been found between expression of E-cadherin and the invasive capacity of a cell.16 The downregulation of E-cadherin expression is also a major hallmark of EMT and is closely related to the malignant progression of HCC. Loss of E-cadherin expression has been strongly implicated in the progression and metastasis of human cancers.17–19 In addition, abnormal expression of E-cadherin is a useful prognostic factor in patients with gastric carcinoma. E-cadherin expression has also been found to be inversely correlated with the expression of certain transcription factors, including slug (SNAI2), snail (SNAI1) and twist homolog 1 (TWIST1).20–22 EMT is induced by several oncogenic pathways, including Src, Ras, Ets, integrin, transforming growth factor-β1, laminin 5, Wnt/β-catenin and Notch.23–25 The process is induced by a complex orchestration of signaling cascades comprising many signaling pathways, making the understanding of its mechanism more challenging.

The transcriptional factor TWIST1 has been shown to play a key role in EMT and metastasis in HCC,15, 20 and was originally identified as a master regulator of embryonic morphogenesis.26–28 Recent studies have also revealed that TWIST1-induced EMT enhances tumor metastasis,29, 30 and TWIST1 expression enhanced cell migration activity in HCC cell lines.15 Thus, here, HCC cell lines were divided into low- and high-EMT groups based on SNAI2 and TWIST1 expression.

Runt-related transcription factor 3 (RUNX3) has been reported to be a tumor suppressor gene for gastric cancer.31 Therefore, it is not surprising that RUNX3 gene expression decreases in HCC and other human malignancies, including those of the colon, lung, pancreas and bile duct.32–35 Several authors have reported that RUNX3 gene and protein expression is decreased in HCC and that loss of RUNX3 expression prevented apoptosis in HCC cells.36–39 We also reported that ectopic RUNX3 expression deactivated Notch signaling by decreasing jagged-1 (JAG1) expression in HCC.40 Based on the morphological changes in an HCC cell line with ectopic RUNX3 expression, we hypothesized that RUNX3 increases cell–cell adhesion in HCC cells. In this study, we demonstrate that the loss of RUNX3 protein expression results in an EMT-like change via increased expression of JAG1 in HCC.

Abbreviations

CAT: chloramphenicol acetyltransferase; DAPI: 4′,6-diamidine-2′-phenylindole dihydrochloride; DMEM: Dulbecco's modified Eagle's medium; EMT: epithelial–mesenchymal transition; FBS: fetal bovine serum; GFP: green fluorescent protein; HCC: hepatocellular carcinoma; JAG1: jagged-1; MTT: 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; PAGE: polyacrylamide gel electrophoresis; PBS: phosphate-buffered saline; PCR: polymerase chain reaction; RUNX3: runt-related transcription factor 3; SDS: sodium dodecyl sulfate; TBS-T: tris-buffered saline with Tween 20; TWIST1: twist homolog 1

Material and Methods

Cell lines and culture

The human HCC cell lines Hep3B, Huh7, HLF and SK-Hep1 were obtained from the American Type Culture Collection (Manassas, VA). The cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA), supplemented with 10% heat-inactivated fetal bovine serum (FBS; Sigma, St. Louis, MO), 1% nonessential amino acids (Sigma), 1% sodium pyruvate (Sigma), and 1% penicillin/streptomycin solution (Sigma). The cells were cultured at 37°C in an atmosphere of 5% CO2 and 95% air. The cells were quiesced at subconfluence under restricted serum conditions with 0.1% dialyzed FBS for 36 hr before the experiment, if needed.

Immunoblot analysis

Cells were plated in six-well, plastic tissue culture dishes and grown to confluence. The cells were washed twice with cold phosphate-buffered saline (PBS) and lysed in 150 μL of sample buffer (100 mM Tris-HCl [pH 6.8], 10% glycerol, 4% sodium dodecyl sulfate [SDS], 1% bromophenol blue and 10% β-mercaptoethanol). The samples were resolved by SDS-polyacrylamide gel electrophoresis, and transferred to an Immobilon-P™ polyvinylidene difluoride membrane (Millipore Corporation, Bedford, MA). The membranes were blocked using Tris-buffered saline with Tween 20 (TBS-T; Sigma) buffer containing 5% bovine serum albumin for 1 hr. The membranes were then incubated with antibodies against SNAI2 (#9589; Cell Signaling Technology, Danvers, MA), TWIST1 (sc-6269; Santa Cruz Biotechnology Inc., Santa Cruz, CA), RUNX3 (R3-G54; Abcam, Cambridge, MA), E-cadherin (#610181; BD Biosciences, San Diego, CA), N-cadherin (#05-915; Millipore), vimentin (ab8069; Abcam) and α-actin (Sigma) overnight at 4°C. The membranes were washed three times with TBS-T and probed with horseradish peroxidase-conjugated secondary antibody before being developed with an ECL western blotting detection system (Amersham Biosciences, Piscataway, NJ) using enhanced chemiluminescence.

RUNX3-expressing cell lines

The human RUNX3 construct was obtained by reverse transcriptase polymerase chain reaction (PCR)-based cloning from normal human hepatocytes.38 The human RUNX3 and/or chloramphenicol acetyltransferase (CAT) (mock) constructs were transfected into Hep3B, Huh7, HLF or SK-Hep1 cells using the FuGene™6 transfection reagent (Roche Diagnostics, Basel, Switzerland), per the manufacturer's instructions. Forty-eight hours after transfection, cells were grown in complete medium, containing 250 μg/mL hygromycin (Roche Diagnostics). Polyclonal lines consisting of >20 colonies were established. At least two independent transfections were performed for each cell line, with a transfection efficiency of more than 80%, as determined by immunocytochemistry using anti-CAT antibody and/or anti-RUNX3 antibody. After selection, all experiments were performed within five passages to avoid natural transformation and natural selection. The cells were cultured under serum-starved conditions for 24–36 hr, if needed, and utilized in the following experiments.

Morphological analysis

Cells were plated onto six-well, plastic tissue culture dishes with quiescent medium at a density of 105 cells/mL. After 24 hr, morphological observations were made using phase contrast microscopy (IMT-2; Olympus, Tokyo, Japan).

Immunocytochemistry analysis

Cells were cultured in a glass chamber (Lab-Tek II; Nalgene Nunc, Roskilde, Denmark) to a density of approximately 2 × 105 cells/mL for 24 hr, prior to being fixed for 20 min in 3% freshly hydrolyzed paraformaldehyde in PBS, at room temperature. After washing with PBS, the cells were incubated with anti-E-cadherin antibody (M3613, Dako Japan, Tokyo, Japan) for 30 min at room temperature, and visualized using Oregon Green-conjugated secondary antibody (Invitrogen). Photographs were taken using a fluorescence microscope (BX51, Olympus) equipped with a digital image capturing system (DP50, Olympus).

MTT assay

Cell proliferation activity was assessed using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Sigma) assay. In brief, cells were grown in 96-well tissue culture dishes. After 24 hr of quiescence, the cells were cultured for the indicated period with or without 10% FBS. At the end of the treatment, 10 μL of MTT (5 mg/mL in PBS) was added to each well and incubated for an additional 4 hr at 37°C. The purple-blue MTT formazan precipitate was dissolved in 100 μL of isopropanol/HCl (0.04 N). The activity of the mitochondria, reflecting cellular growth and viability, was evaluated by measuring the optical density at 570 nm with a microplate reader (Bio-Rad, Hercules, CA).

Cell migration assay

Cell migration was assessed using an in vitro wound healing assay. The cells were cultured in six-well tissue culture dishes. After 24 hr of quiescence, experimental wounds were made by dragging a rubber Policeman™ (Fisher Scientific, Hampton, NH) across the cell culture. The cultures were rinsed with PBS and placed in fresh quiescence medium. The cells were treated with 10% FBS and/or caspase inhibitor (Caspase Inhibitor IV, 100 μM; Calbiochem, Gibbstown, NJ) for 24 hr at 37°C. Three wounds were created for each specimen. Photographs were taken at 0 and 24 hr, and the relative distance traveled by the cells at the acellular front was determined.

DAPI staining

Cells were grown to 50% confluence in a glass chamber (Lab-Tek II; Nalgene Nunc) and quiesced for 24 hr. The cells were then treated with 10% FBS and/or caspase inhibitor IV (100 μM) for 24 hr. Following exposure, chromosomal DNA was stained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI; Dojindo, Kumamoto, Japan), at a concentration of 1 μg/mL in PBS for 2 min, according to the manufacturer's protocol. The stained cells were then observed by fluorescence microscopy (IX-70; Olympus).

Ectopic TWIST1 protein expression

Human TWIST1 cDNA was obtained by PCR-based cloning from a normal human placenta cDNA library (Takara Bio Otsu, Japan) and subcloned into the pcDNA 3.1 expression vector (Invitrogen).15 TWIST1 and green fluorescent protein (GFP; control) constructs were transfected into RUNX3-expressing Hep3B cells using Fugene™6 (Roche Diagnostics). Cells were incubated for 48 hr after transfection, after which immunoblot analyses and the cell migration assays were performed.

JAG1 ligand treatment

Because RUNX3 expression deactivates Notch signaling by downregulating JAG1 expression,40 the JAG1 ligand peptide was used to activate Notch signaling.41 RUNX3- or control CAT-expressing Hep3B cells were treated with or without 50 μM JAG1 ligand peptide (AnaSpec, Fremont, CA) and/or control-scrambled peptide for 48 hr. Cells were then lysed and an immunoblot analysis was performed.

HCC tissues and immunohistochemistry

A group of 33 patients, 24 men (age range, 39–77 years; average age, 62.2 years) and nine women (age range, 54–82 years; average age, 64.9 years) were included in this study. Resected HCC tissues were obtained after receiving written informed consent that adhered to the stringent ethical criteria of the Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences. Immunohistochemistry was performed on formalin-fixed paraffin sections prepared from the resected tissue. Sections were dewaxed and dehydrated; after rehydration, endogenous peroxidase activity was blocked for 30 min in a methanol solution containing 0.3% hydrogen peroxide. Following antigen retrieval in citrate buffer, the sections were blocked, again, overnight at 4°C. The sections were probed with anti-RUNX3 rabbit polyclonal antibody (ab49117; Abcam), anti-E-cadherin monoclonal antibody (Dako Japan), anti-N-cadherin monoclonal antibody (M3613; Dako Japan), anti-SNAI2 rabbit polyclonal antibody (#9585; Cell Signaling Technology) and anti-TWIST1 goat polyclonal antibody (Santa Cruz Biotechnology). The primary antibody was detected using a biotinylated anti-rabbit antibody (Dako Japan) or a biotinylated anti-goat antibody (Dako Japan). The signal was amplified by avidin-biotin complex formation and was developed with diaminobenzidine, followed by dehydration in alcohol and xylene, and the sections were mounted. The sections were scored for RUNX3, E-cadherin, N-cadherin and TWIST1 expression using a 4-point scale: 0, negative; 1, weak signal; 2, intermediate signal; and 3, strong signal.42 All sections were scored independently by two observers, without prior knowledge of the clinical background. All discrepancies in scoring were reviewed and a consensus was reached. Because TWIST1 is a major EMT inducer and an indicator,43 TWIST1-positive (score 2 or 3) HCCs were designated as high-EMT HCCs, while TWIST1-negative (score 0 or 1) HCCs were designated as low-EMT HCCs. The clinicopathological features, TWIST1, and RUNX3 expression levels are shown in the Table 1, Supporting Information. Statistical analyses were performed using JMP software (SAS Institute, Cary, NC).

Gene expression profiling analysis in human HCC

To further investigate correlations between RUNX3 expression and transcriptional regulation of E-cadherin, N-cadherin and vimentin, JAG1 expression profiles in low-EMT HCCs from publicly available HCC data sets were evaluated using Oncomine (http://www.oncomine.org). In brief, after excluding TWIST1-positive (high-EMT) (TWIST1/213943_at > −0.20) HCCs from the Chiang Liver data sets, mRNA expression profiles for RUNX3, E-cadherin, N-cadherin, vimentin and JAG1 were evaluated using RUNX3/234928_x_at, CDH1/201130_s_at, CDH2/203440_at, VIM/155938_x_at, and JAG1/231183_s, respectively. Statistical analyses were performed using JMP software (SAS Institute).

Results

EMT-inducer expression in HCC cell lines

HCC cell lines were divided into two groups according to SNAI2 and TWIST1 expression: Hep3B and Huh7 cells were designated as low-EMT cells, and HLF and SK-Hep1 cells as high-EMT cells (Fig. 1). Low-EMT cells expressed E-cadherin, while no detectable level of E-cadherin was found in high-EMT cells.

Figure 1.

Expression of EMT inducers in HCC cell lines. Immunoblot analyses were performed using antibodies against SNAI2, TWIST1 and E-cadherin. Immunoblotting for α-actin levels was used to verify equal loading of cellular proteins. HCC cell lines were divided into two groups according to EMT-inducer SNAI2 and TWIST1expression. Representative blots of more than three independent experiments are shown.

Ectopic RUNX3 expression in HCC cell lines

Ectopic RUNX3 expression was evaluated in endogenous RUNX3-negative HCC cell lines,38 and the morphological changes were analyzed under phase contrast microscopy. Ectopic RUNX3 expression changed the morphology of the cells from a fibroblast-like spindle shape to a paving stone, sheet-like structure in Hep3B (Fig. 2a) and Huh7 cells (Fig. 1, Supporting Information). These cells showed well-organized cell–cell adhesion, characteristic of epithelial cells. E-cadherin expression was enhanced in the RUNX3-expressing Hep3B cells (Fig. 2b) or Huh7 cells (data not shown), especially at the cell–cell junction. In contrast, RUNX3 expression did not induce significant morphological changes in HLF (Fig. 2a) or SK-Hep1 cells (Fig. 1, Supporting Information), neither was E-cadherin expression significantly different in these cells (Fig. 2b). RUNX3 protein expression was also confirmed in RUNX3 cDNA-transfected cells (Fig. 2c).

Figure 2.

Ectopic RUNX3 protein expression in HCC cell lines. Eukaryotic expression constructs for CAT (mock) and RUNX3 were introduced into Hep3B, Huh7, HLF and SK-Hep1 cell lines. After a 24-hr quiescence, cells were used for the experiments in the presence of 10% FBS. (a) Micrographs were taken by phase contrast microscopy. The micrographs are representative of three independent experiments. Bar = 100 μm. (b) Immunocytochemical analysis was performed using antibody against E-cadherin. The micrographs are representative of three independent experiments. Bar = 40 μm. (c) Immunoblot analyses were performed using antibodies against RUNX3, E-cadherin, N-cadherin and vimentin. Immunoblotting for α-actin levels was used to verify equal loading of cellular proteins. Representative blots of more than three independent studies are shown. (d) Cell proliferation activity was measured by MTT assay (mock, white bars; RUNX3, black bars). (e) Cell migration activity was measured as described in the Material and Methods section (mock, white bars; RUNX3, black bars). (d, e) All results were normalized to cell proliferation and migration activity for CAT-expressing Hep3B cells. Data represent the mean ± SE of more than three independent experiments, each performed in triplicate. n.s., p > 0.05; *, p < 0.05; **, p < 0.01 (vs. data for mock-transfected Hep3B); Student's t test.

Ectopic RUNX3 protein expression inverted EMT in HCC cell lines

E-cadherin, N-cadherin and vimentin expression were assessed by immunoblot analysis in HCC cell lines (Fig. 2c). E-cadherin expression was induced by ectopic RUNX3 protein expression in low-EMT cells. High-EMT HCC cells did not express E-cadherin, and ectopic RUNX3 protein expression did not induce E-cadherin expression and it had little effect on N-cadherin and vimentin expression in high-EMT HCC cells. However, ectopic RUNX3 protein expression suppressed N-cadherin and vimentin expression in low-EMT HCC cells.

Ectopic RUNX3 protein expression suppressed cell growth under serum starvation

To confirm the biological effects of ectopic RUNX3 protein expression, MTT assays were performed to determine the effect on cell proliferation. In all four HCC cell lines, RUNX3 expression suppressed cell growth by 31–38% (Fig. 2d).

Ectopic RUNX3 protein expression suppressed cell migration in HCC cell lines

The effect of ectopic RUNX3 expression on HCC cell lines was examined in the in vitro wound healing assay, which is commonly used to assess the effects of promigratory and antimigratory agents on cultured cells. RUNX3 expression significantly suppressed cell migration by 70 ± 20% and 63 ± 5% in Hep3B and Huh7 cells, respectively, but not in HLF and SK-Hep1 cells (Fig. 2e).

Effect of RUNX3 expression on cell migration in the presence of an apoptosis inhibitor

An apoptosis inhibitor was used to determine if apoptosis-inducing factors affected cell migration activity in ectopic RUNX3 protein-expressing HCC cell lines. Caspase inhibitor IV was observed to inhibit serum starvation-induced apoptosis by 74% in RUNX3-expressing Hep3B cells (Fig. 3a). However, caspase inhibitor IV did not abrogate the RUNX3-induced reduction in cell migration activity in these cells (Fig. 3b).

Figure 3.

Effect of an apoptosis inhibitor. Hep3B cells were cultured on glass chamber slides and quiesced, then treated with or without caspase inhibitor IV (100 μM) and 10% FBS. (a) Apoptosis was quantified using the DAPI apoptosis detection assay (mock, white bars; RUNX3, black bars). Results are expressed as percent apoptotic cells. Data represent the mean ± SE of more than three independent experiments, each performed in triplicate. n.s., p > 0.05; **, p < 0.01 (vs, data for mock-transfected cells); Student's t test. (b) Cell migration activities were measured as described in the Material and Methods section (mock, white bars; RUNX3, black bars). All results are normalized to data for untreated control cells. Data represent the mean ± SE of more than three independent experiments, each performed in triplicate. n.s., p > 0.05; *, p < 0.05; **, p < 0.01 (vs. data for mock-transfected cells); Student's t test.

TWIST1 expression abrogated the anti-EMT effect of ectopic RUNX3

RUNX3 appeared to have a weak effect on EMT in TWIST1-positive HCC cell lines. To determine if ectopic TWIST1 expression affects the anti-EMT effect of RUNX3 expression, TWIST1 cDNA was introduced into the TWIST1-negative HCC cell line Hep3B, causing the cells to transiently express TWIST1 (Fig. 4a). RUNX3 expression did not have a significant effect on the EMT markers, E-cadherin, N-cadherin or vimentin in TWIST1-expressing Hep3B cells (Fig. 4a). Likewise, RUNX3 expression did not have the antimigratory effect on TWIST1-expressing Hep3B cells (Fig. 4b).

Figure 4.

Effect of TWIST1 expression on EMT markers. CMV-driven eukaryotic expression constructs for GFP (control) and TWIST1 were introduced into CAT- and RUNX3- expressing Hep3B. (a) Cell lysates were collected 48 hr after transfection. Immunoblot analyses were performed using antibodies against RUNX3, TWIST1, E-cadherin, N-cadherin and vimentin. Immunoblotting for α-actin levels was used to verify equal loading of cellular proteins. Representative blots of more than three independent experiments are shown. (b) Cell migration activities were measured as described in the Material and Methods section (mock, white bars; RUNX3, black bars). Results are normalized to data for GFP (control) transfected CAT-expressing Hep3B cells. Data represent the mean ± SE of more than three independent experiments, each performed in triplicate. n.s., p > 0.05; **, p < 0.01 (vs. data for mock-transfected cells); Student's t test.

JAG1 expression

The JAG1 ligand peptide was used to determine if RUNX3 expression is regulated via JAG1. The JAG1 ligand peptide decreased E-cadherin expression in RUNX3-expressing Hep3B cells (Fig. 5) and induced re-expression of N-cadherin and vimentin in RUNX3-expressing Hep3B cells (Fig. 5).

Figure 5.

Effect of JAG1 stimulation on EMT markers. CMV-driven eukaryotic expression constructs for GFP (control) and TWIST1 were transfected into CAT- and/or RUNX3- expressing Hep3B. Cell lysates were collected 48 hr after transfection. Immunoblot analyses were performed using antibodies against RUNX3, E-cadherin, N-cadherin and vimentin. Immunoblotting for α-actin levels was used to verify equal loading of cellular proteins. Representative blots of more than three independent experiments are shown.

E-cadherin expression correlation with RUNX3 expression in HCC tissues

Thirty-three HCC tissue samples were available for comparing E-cadherin and RUNX3 protein expression by immunohistochemistry. Representative images of E-cadherin-negative (Score 0), E-cadherin-positive (Score 3), N-cadherin-negative (Score 0) and N-cadherin-positive (Score 3) tissues are shown in Figures. 6a, b′, c′ and d′, respectively. To determine if there was a relationship between E-cadherin and RUNX3 expression in HCC tissues, the scores for E-cadherin immunoreactivity were plotted against those for RUNX3 (Fig. 6b). In general, low E-cadherin expression scores were observed in tissues with low RUNX3 expression. Statistical analysis revealed that RUNX3 expression and E-cadherin expression were significantly correlated (r = 0.58, p = 0.0025) in low-EMT HCC tissues. Scores for N-cadherin immunoreactivity were also plotted against those for RUNX3 (Fig. 6c), where high N-cadherin expression scores were observed in tissues with low RUNX3 expression. A significant inverse correlation was found between RUNX3 and N-cadherin expression (r = −0.74, p < 0.0001). Thus, high-EMT HCC tissues showed low expression of E-cadherin and high expression of N-cadherin.

Figure 6.

Correlations between RUNX3 expression and EMT markers in human HCC (a) The images show the immunohistochemical staining of E-cadherin (a, protein score of 0; b, protein score of 3) and N-cadherin (c, protein score of 0; d, protein score of 3). Bar = 100 μm. (b, c) RUNX3, E-cadherin, N-cadherin and TWIST1 protein expression were assessed by immunohistochemical analysis in human HCC tissues and corresponding tumor-free sections. Plots of the E-cadherin (b) and N-cadherin (c) expression score are shown compared with the RUNX3 expression score. (d) Correlations between RUNX3 expression and EMT markers were analyzed using publicly available microarray data sets (www.oncomine.org). RUNX3, E-cadherin, N-cadherin, vimentin and JAG1 expression are plotted. The data are shown as arbitrary expression values. A 95% tolerance ellipse for each pair of variables was calculated and plotted.

Gene expression profiling analysis in human HCC

A further examination of the correlation between expression of RUNX3 and of EMT markers was conducted by analyzing the Oncomine datasets. These analyses revealed that JAG1 mRNA expression was negatively correlated with RUNX3 mRNA expression in low-EMT HCC tissues (r = 0.28, p < 0.05) (Fig. 6d). E-cadherin mRNA expression was also found to be significantly correlated with RUNX3 mRNA expression in low-EMT HCC tissues (r = 0.33, p < 0.05) (Fig. 6d). mRNA expressions of N-cadherin and vimentin tended to be negatively correlated with RUNX3 mRNA expression, but the correlations were not significant (r = −0.24 and −0.07, respectively) (Fig. 6d).

Discussion

RUNX3 expression is generally reduced in human HCC tissues.36, 38 However, the role of decreased RUNX3 expression in the development and progression of HCC has not yet been fully elucidated. The results obtained here suggest that decreased expression of the RUNX3 protein contributes to induction of EMT in HCC cell lines.

Ectopic RUNX3 expression was evaluated in the endogenous RUNX3-negative HCC cell lines Hep3B, Huh7, HLF, and SK-Hep1 (Fig. 2c). The establishment of RUNX3-expressing cell lines was advantageous in order to elucidate the RUNX3-related molecular mechanism, as they helped to provide additional clarity to the results obtained. RUNX3 expression is known to cause apoptosis, especially in the absence of FBS,38 whereas RUNX3-expressing cells can grow in the presence of FBS. Because RUNX3 is a strong apoptosis inducer,38, 44 the anti-cell migratory effect of RUNX3 could be due to the apoptosis enhanced by serum starvation. Therefore, a caspase inhibitor was used to eliminate the potential effects of apoptosis on cell migration. The caspase inhibitor successfully abrogated RUNX3-induced apoptosis (Fig. 3a), but RUNX3-expressing cells did not recover their cell migration activity (Fig. 3b). Morphological studies also showed that ectopic RUNX3 expression induced phenotypic changes, resulting in the cells adopting a more plate-like shape instead of their normal, spindle-like shape, in the low-EMT HCC cell lines, but not in the high-EMT HCC cell lines. These results are consistent with the cell shapes generally associated with migratory and nonmigratory cells. Collectively, these results suggest that RUNX3 is directly connected to the inhibition of cellular migration associated with metastatic cells.

Several studies examined the downregulation of E-cadherin in HCCs by immunohistochemical analyses.44–46 In accordance with these reports, E-cadherin expression was generally reduced in human HCC tissues and was correlated with RUNX3 expression (Fig. 6b). The strong correlation between RUNX3 and E-cadherin expression suggests that loss of RUNX3 expression may cause reduced E-cadherin expression. Similarly, RUNX3 was only able to induce E-cadherin expression in low-EMT HCC cell lines (Figs. 2b and 2c). As TWIST1 is a strong inducer of EMT and reduces E-cadherin expression, RUNX3 had no effect on E-cadherin expression and very little effect on N-cadherin and vimentin expression in high-EMT cells (Fig. 2c). But, N-cadherin and vimentin expression were reduced by ectopic RUNX3 expression in low-EMT HCC cells. Expression of N-cadherin was also analyzed in human HCC tissues. N-cadherin expression was inversely related to RUNX3 expression in low-EMT HCC tissues (Fig. 6c).

The effects of TWIST1 expression on RUNX3-expressing HCC cell lines were examined by transiently expressing the TWIST1 protein in TWIST1-negative Hep3B cells by transfecting TWIST1 cDNA into CAT- and RUNX3-expressing Hep3B cells (Fig. 4a). Ectopic RUNX3 protein expression did not have a significant effect on the EMT markers and cell migration activity in TWIST1-expressing cells (Fig. 4b). Although there is a possibility that TWIST1 and RUNX3 independently act on E-cadherin, N-cadherin and vimentin expression, the present results suggest that TWIST1 is the more important regulator of these EMT markers. Therefore, a sequence of genetic alterations in the progression of EMT in HCC may be hypothesized based on the frequency of gene alteration in HCC. The loss of RUNX3 protein expression was found in ∼90% of HCC tissue samples, while TWIST1 expression was found in ∼30% of HCC tissue samples.15, 38 Recently, Aleksic et al.47 also reported that loss of RUNX3 was a frequent, early event in chemical liver carcinogenesis. Various authors have reported that a loss of or decreased expression of RUNX3 is generally found in HCC,37–39 suggesting that decreased expression of RUNX3 could be an early event in human HCC. Since TWIST1 expression reduced RUNX3 expression (Fig. 4a), TWIST1 may interact with RUNX3 to progress EMT in HCC. Further studies are needed to elucidate the detailed mechanism of interaction between RUNX3 and TWIST1 expression.

RUNX3 has also been previously reported to regulate JAG1 expression in HCC cells by suppressing transcription through direct binding to the transcriptional regulatory region of JAG1.40 To clarify whether RUNX3 regulates EMT in HCC cells through JAG1 expression, the JAG1 ligand peptide was used. The JAG1 ligand peptide successfully reinduced EMT in RUNX3-expressing Hep3B cells (Fig. 5).

Because a relatively small number of HCC tissues were analyzed for RUNX3 and EMT marker expression using immunohistochemistry, Oncomine datasets were also analyzed for RUNX3 mRNA expression and the mRNA of EMT markers in low-EMT HCCs (Fig. 6d). In accordance with both a previous report40 and the present results, JAG1 mRNA expression was significantly, negatively correlated with RUNX3 mRNA expression (Fig. 6d) and E-cadherin mRNA expression was significantly correlated with RUNX3 mRNA expression (Fig. 6d). Although significant correlations were not found between RUNX3 mRNA expression and the other EMT markers, N-cadherin mRNA expression and vimentin mRNA expression tended to be inversely related to RUNX3 mRNA expression.

A repressive function of EMT has been found in a study of tumor suppressor genes in which depletion of a retinoblastoma protein induced EMT in a breast cancer cell line.48 This study suggests that other tumor suppressor genes may be similarly involved in EMT. A relationship between RUNX3 expression and EMT was reported by Chang et al.,49 who demonstrated that RUNX3 expression induced claudin expression. The regulatory function of EMT by RUNX3 was also predicted in a prospective review.50 Herein, we demonstrated that RUNX3 expression regulates EMT through JAG1 expression.

Conclusions

We uncovered a novel function of RUNX3 in regulating EMT in HCC. Our results suggest that loss or decreased expression of RUNX3 induced EMT via inducing JAG1 expression.

Acknowledgements

The authors wish to thank Tatsuya Fujikawa for valuable suggestions.

Ancillary