COX-2 and Akt mediate multiple growth factor-induced epithelial-mesenchymal transition in human hepatocellular carcinoma

Abstract

Background and Aim

Cancer invasion and metastasis are characterized by epithelial-mesenchymal transition (EMT). Hepatocellular carcinoma (HCC) causes metastasis and significant mortality. Elucidating factors promoting EMT in HCC is necessary to develop effective therapeutic strategies.

Methods

The LH86 cell line was developed in our laboratory from well-differentiated HCC without associated hepatitis or cirrhosis and used as a model to study EMT in HCC. Effects of transforming growth factor (TGF) β-1, epidermal growth factor (EGF), hepatocyte growth factor (HGF) and basic fibroblast growth factor (bFGF) were examined using morphology, molecular markers, effects on migration and tumorigenicity. The involvement of cyclooxygenase-2 (COX-2) and Akt were examined.

Results

LH86 cells display epithelial morphology. TGFβ-1, EGF, HGF and bFGF induced mesenchymal changes in them associated with loss of E-cadherin, albumin, α-1 anti-trypsin expression and increased expression of vimentin, collagen I and fibronectin. There was associated increased migration, tumorigenicity and increased expression of COX-2, PGE2, Akt and phosphorylated Akt. Inhibition of COX-2 and Akt pathways led to inhibition of characteristics of EMT.

Conclusions

Multiple growth factors induce EMT in HCC. COX-2 and Akt may mediate EMT associated development and progression of HCC and molecular targeting of COX-2 and Akt may be an effective therapeutic or chemopreventative strategy in advanced and metastatic HCC.

Introduction

Metastasis is the commonest cause of cancer mortality [1]. Efforts have been directed at unraveling the basis of malignant transformation, but far less attention has been directed at understanding the basis for metastasis specifically. There is increasing recognition that the development of effective management for cancers requires understanding the mechanisms underlying cancer metastasis.

An important process that characterizes tumor invasion, metastasis, resistance to treatment and tumor recurrence is epithelial-mesenchymal transition (EMT) [2]. When cancer cells undergo EMT, they lose contact with neighboring cells and their basement membrane, undergo biochemical changes, acquire mesenchymal phenotype and become migratory and invasive [3]. This is characterized by molecular changes including loss of E-cadherin [4], gain of vimentin [5], collagen I [6] and fibronectin [7] as well as expression of transcription factors of the snail family [8] and nuclear over-expression of beta-catenin [9].

Hepatocellular carcinoma (HCC) is the third commonest cause of cancer mortality worldwide [10] and a fast growing cause in the western world [11]. Due to the recognition of the roles of tumor invasion, metastasis, therapeutic resistance and tumor recurrence as clinically important aspects of cancer progression and the fact that EMT underlies these processes, we have focused on further characterizing EMT in HCC using a novel cellular model of EMT in HCC, elucidating the factors that may be promoting the process and its effects on hepatocyte function.

This is the first study of EMT using the novel LH86 HCC cell line derived from a well differentiated HCC established in our laboratory [12]. Our study also examined the role of the cyclooxygenase-2 (COX-2) and Akt pathways as possible mediating intracellular signaling mechanisms in the process of EMT in HCC. Previous studies have suggested that COX-2 is overexpressed in HCC and that it may be causally related to HCC development and progression [13–15]. Moreover, Akt is known as a pro-survival factor in cancer metastasis [16]. Consequently, we have examined the role of COX-2 and Akt as possible mediating mechanisms of growth factors inducing EMT to promote HCC progression and metastasis.

Materials and Methods

Cell culture and reagents

The LH86 human HCC cell line was established in our laboratory as previously described [12]. It was maintained in culture in DMEM supplemented with 10% fetal bovine serum (FBS), 2mM L-glutamine, nonessential amino acids, 100mg/l penicillin and 100mg/l streptomycin, but without further epidermal growth factor (EGF) treatment. For comparison, the Huh7 human HCC cell line was also cultured in DMEM supplemented with 10% fetal bovine serum (FBS), 2mM L-glutamine, nonessential amino acids, 100mg/l penicillin and 100mg/l streptomycin.

Cells in culture were treated for 7 days with human transforming growth factor (TGF) β-1 (2ng/ml), EGF (50ng/ml), human hepatocyte growth factor (HGF) (5ng/ml) and human basic fibroblast growth factor (bFGF) (20ng/ml) (R&D Systems, MN) alone or in combination. Untreated cells were always used as controls. There was one planned change of medium. For mechanistic studies, selective inhibitors of COX-2 (5μM NS398) and the Akt pathway (10μM LY294002) were used. Inhibitors were purchased from EMD Chemicals, NJ and dissolved in dimethyl sulfoxide (DMSO) and control cells were always treated with equivalent volumes of DMSO. Inhibitor concentrations used were those previously established to have specific activity [17, 18]. Moreover, dose-response experiments using the Huh7 HCC cell line confirmed that 5μM NS398 and 10μM LY294002 have specific inhibitory activity against COX-2 and the Akt pathway respectively (see Supplementary figure 2). In all experiments requiring inhibitors, cells were treated with inhibitors 60 minutes before treatment with growth factors. In these experiments, medium was changed once at least 24 hours before the experimental endpoint. After this change of medium, cells were treated with inhibitors again 1 hour before treatment with growth factors. In cases when cells became confluent (such as for 11-day and 20-day experiments), we passaged them and treated with the growth factors again as described.

Effects of growth factor-treatment on cell morphology were examined using microscopy and immunofluorescent staining for specific markers of mesenchymal morphology.

Immunoblotting

Cells were grown in a monolayer in 6-well plates and treated with growth factors as mentioned above. Immunoblotting was performed as previously described [18]. The specific primary antibodies used are listed in Table 1. Immunoreactive proteins were visualized by incubating in HRP-conjugated secondary antibodies. Chemiluminescence was detected by incubating in an equal-parts mixture of the SuperSignal West Pico stable peroxide solution and luminol/enhancer solution (Pierce, IL) and subsequently using an image processing machine.

Table 1.

List of antibodies used.

Antibody	Catalog # and Source	Concentration
anti-E-cadherin	H-108: sc-7870, Santa Cruz	1: 200 (immunoblotting) 1: 100 (immunofluorescence and immunohistochemistry)
anti-albumin	#126584, Calbiochem	1: 100 (immunofluorescence)
anti-AAT	V1080, biømeda	1: 100 (immunofluorescence)
anti-vimentin	H-84: sc-5565, Santa Cruz	1: 100 (immunofluorescence and immunohistochemistry)
anti-fibronectin	ab2413-500, abcam	1: 1000 (immunoblotting) 1: 100 (immunofluorescence)
anti-collagen I	ab292-100, abcam	1: 500 (immunoblotting) 1: 100 (immunofluorescence)
mouse anti-β-actin,	A5441, Sigma	1: 2000
anti-Akt (detects total Akt1, Akt2 and Akt3)	#9272, Cell Signaling Technology	1:2000
anti-p-Akt (detects phosphorylation at Ser 473)	#4070S, Cell Signaling Technology	1:1000

Quantitative immunofluorescence analysis

Cells were seeded onto coverslips in 6-well plates in 10% FBS-containing medium, then treated with growth factors as mentioned above. After washing in PBS, cells were fixed in 5% acetic acid in ethanol. The rest of the procedure was performed as previously described [18] using 1: 100 concentration of primary antibodies listed in Table 1 and appropriate FITC-conjugated secondary antibodies (Southern Biotech) and mounted on slides with DAPI-containing mounting medium (Vector Laboratories), then visualized with a fluorescent microscope equipped with imaging software. Images were acquired at X20 magnification. Images from 3 experiments were analyzed for corrected integrated density using the ImageJ Image Analysis software (National Institutes of Health, Maryland, USA). Briefly, this involved acquiring integrated density of a randomly selected area of the slide and subtracting from its value the value of the background integrated density. The resulting “corrected integrated density” was then normalized to cell number by dividing the value with the number of cells giving off the signal. Cell number was determined by counting the number of DAPI-stained nuclei (blue staining) in the selected area.

Quantitative real-time reverse transcriptase polymerase chain reaction (qRT-PCR)

qRT-PCR was performed as previously described [12]. The primers used for amplification are listed in Table 2. qRT-PCR was performed using SYBR Green. Reactions were conducted in a 96-well spectrofluorometric thermal cycler (Applied Biosystems, CA).

Table 2.

List of primers used.

Gene	Primer sequences	Reference
GAPDH	f: 5′-TCACCAGGG CTGCTTTTA-3′ r: 5′-TTCACACCCATGACGAACA-3′	12
Vimentin	f: 5′-TCTGGATTCACTCCCTCTGG-3′ r: 5′-TGCACTGAGTGTGTGCAATTT-3′	25
E-cadherin	f: 5′-TAACCGATCAGAATGAC-3′ r: 5′-TTTGTCAGGGAGCTCAGGAT-3′	26
Albumin	f: 5′-AGAACAGGACAATGGGCAAC r: 5′-ACCAGCACCGACCACTATTC	Designed by Chen Liu’s Lab, University of Florida, FL.
AAT	f: 5′-GGGGATAGACATGGGTATGG r: 5′-GGGTCAACTGGGCATCACTA	Designed by Chen Liu’s Lab, University of Florida, FL.
COX-2	f: TTTGGTGAAACCATGGTAGA r: CTCTGGATCTGGAACACTGA	Designed by Chen Liu’s Lab, University of Florida, FL.
Akt1	f: TAAGTACTTGGGGCATTTCC r: GATGTACTCCCCTCGTTTGT	Designed by Chen Liu’s Lab, University of Florida, FL.

Migration assays

1 × 10⁴ cells were seeded into 6-well plates and treated with growth factors. At 70% confluency, the cell monolayer was wounded with a 200μl-pipette, washed with PBS and medium replaced. Images were taken 24 hours later. Images from 3 experiments were analyzed for percentage of cell-covered area using the Wimasis Image Analysis software (Wimasis GmbH, Munich, Germany).

Tumorigenicity studies

Cells were treated with TGFβ-1 for 11 days. Untreated cells were used as controls. There was planned change of medium once a week and cells were treated with TGFβ-1 again with each change of medium. 1×10⁶ cells were transplanted into the flank of immune-deficient mice following University of Florida IACUC approved protocols. Briefly, cells were harvested in serum-containing medium. After spinning down, the harvested cells were washed with PBS and resuspended in PBS at a concentration of 1×10⁶ cells/100μl and 100 μl was injected. Growth of the tumors was measured weekly in 3 dimensions. Animals were humanely sacrificed at the same time point for each experiment after the development of tumors. Tumor volume was compared between tumors from control cells and those from growth factor-treated cells.

Immunohistochemistry

Mouse xenograft tumors were formalin-fixed and paraffin-embedded. Sections were cut, immobilized on slides, deparaffinized in xylenes, rehydrated in increasing dilutions of ethanol and then distilled water. After incubating in 3% hydrogen peroxide for 10 minutes to block endogenous peroxidase, antigen retrieval was performed by incubating in pH 6.1 citrate buffer (Dako) at 95°C for 30 minutes. Sections were then rinsed in PBS, blocked in goat serum for 30 minutes at room temperature, incubated in avidin and biotin blocking reagents (SP-2001, Vector Laboratories) each for 15 minutes and then incubated in 1: 100 dilution of rabbit anti-E-cadherin antibody or mouse anti-vimentin antibody for 1 hour at room temperature. After washing in PBS, sections were then incubated for 30 minutes in 1: 100 dilution of appropriate biotinylated secondary antibody. 15 minutes incubation with the Vectastain ABC reagent (PK-4000, Vector Laboratories) was followed by washing in PBS. Protein expression was then visualized by treating sections with diaminobenzidine (DAB) substrate (Vector Laboratories). Sections were dehydrated in increasing concentration of ethanol and then in xylenes. Images were acquired at X20 magnification.

PGE2 ELISA

1×10⁵ LH86 cells per well were seeded into 24-well plates and treated with TGF β-1 (2ng/ml), EGF (50ng/ml), human hepatocyte growth factor (HGF) (5ng/ml) and human basic fibroblast growth factor (bFGF) (20ng/ml) (R&D Systems, MN) alone or with TGF β-1 (2ng/ml) and EGF (50ng/ml) for 7 days. Untreated cells were used as control. After change of medium to serum-free medium, cells were treated for 30 minutes with the calcium ionophore A23187 (1 μM). Analysis of PGE2 secreted into serum-free medium was performed as previously described [18].

COX-2 siRNA transfection

1 × 10⁴ LH86 cells were cultured in 6-well plates until 70% confluent. 50nM of a specific COX-2 siRNA (sc-29279, Santa Cruz) was transfected into the cells as previously described [19]. Control cells were transfected with equivalent concentration of a non-specific control siRNA (sc-37007, Santa Cruz). After 6 hours incubation, medium was changed to complete culture medium and cells were cultured for a further 24 hours before RNA extraction and qRT-PCR were performed.

Statistical analysis

Immunofluorescent staining (normalized to cell number), qRT-PCR (normalized to GAPDH), migration and tumorigenicity experiments were performed at least 3 times. Results are expressed as mean ± SEM. Effects were compared with controls. Paired t tests were used to analyze the effect of growth-factor treatment as compared to controls. P < 0.05 was considered significant.

Results

Effects on morphology

LH86 cells have epithelial morphology. After treating LH86 cells with TGFβ-1 for 5 days, it was noticed that some cells were floating in culture medium. Many of the surviving cells had, however, undergone morphological change to mesenchymal. The addition of EGF led to reduction in proportion of floating cells and surviving cells had undergone more mesenchymal morphological change and about 48 hours earlier (Figure 1). HGF and bFGF treatment also led to mesenchymal change in both cell lines in 5 days (Figures 2 – 3).

Figure 1. Growth factors induce mesenchymal change in HCC cells.

LH86 HCC cells were treated for 7 days with 2ng/ml TGFβ-1 or a combination of 2ng/ml TGFβ-1 and 50ng/ml EGF. Image magnification was X10 objective.

Figure 2. Multiple growth factors induced increased vimentin expression in HCC cells.

A. LH86 HCC cells were treated for 7 days with 2ng/ml TGFβ-1 or 50ng/ml EGF or 5ng/ml HGF or 20ng/ml bFGF or a combination of 2ng/ml TGFβ-1 and 50ng/ml EGF. Vimentin expression was assessed with immunofluorescent staining. Image magnification was X20 objective. Green coloration represents vimentin expression and blue coloration represents DAPI-stained nuclei. B. Quantitative analysis of immunofluorescent staining (corrected integrated density normalized to cell number). C. Real-time quantitative RT-PCR analysis of vimentin gene expression (normalized to GAPDH). Results expressed as mean ± SEM. * P < 0.05; ** P < 0.01, control cells versus treated cells; NS, not significant; N = 3.

Figure 3. Multiple growth factors induced increased fibronectin expression in HCC cells.

A. LH86 HCC cells were treated for 7 days with 2ng/ml TGFβ-1 or 50ng/ml EGF or 5ng/ml HGF or 20ng/ml bFGF or a combination of 2ng/ml TGFβ-1 and 50ng/ml EGF. Fibronectin expression was assessed with immunofluorescent staining. Image magnification was X20 objective. Green coloration represents fibronectin expression and blue coloration represents DAPI-stained nuclei. B. Quantitative analysis of immunofluorescent staining (corrected integrated density normalized to cell number). C. Immunoblotting analysis. Results expressed as mean ± SEM. * P < 0.05; ** P < 0.01, control cells versus treated cells, N = 3.

Effects on mesenchymal marker expression

To further confirm the mesenchymal state of the cells that were treated with growth factors, we examined the effect of growth factor-treatment on fibronectin, collagen I and vimentin expression. These proteins are characteristically expressed by mesenchymal cells [7] [6] [5]. Treatment of LH86 cells with TGFβ-1, EGF, HGF, bFGF and a combination of TGFβ-1 and EGF led to increased fibronectin, collagen I and vimentin expression by LH86 cells as determined by immunofluorescence (Figures 2 – 3 and Supplementary figure 1) and qRT-PCR (Figure 2). Quantitative analysis of the corrected integrated density of immunofluorescent staining from 3 experiments revealed no significant differences between the corrected integrated density of cells treated with combination of TGFβ-1 and EGF and cells treated with either alone. It was also observed that cells with significant expression of vimentin, fibronectin and collagen I had acquired mesenchymal phenotype. Untreated cells showed little expression and remained epithelial in phenotype as observed from the close proximity of the DAPI-stained nuclei (blue coloration). This effect of growth factor treatment on increased vimentin protein and gene expression was confirmed in Huh7 cells with quantitative immunofluorescence analysis and qRT-PCR respectively (Supplementary Figure 2). Further, we compared the effects of growth factor treatment for 7 days with effect of treatment for 20 days on vimentin gene expression and found that 20 days treatment leads to further increase in vimentin gene expression which is statistically significant for bFGF treatment alone.

Effect on E-cadherin, albumin and AAT expression

In contrast, treatment of LH86 cells with TGFβ-1, HGF, bFGF and a combination of TGFβ-1 and EGF led to decreased E-cadherin expression as determined by quantitative immunofluorescence analysis and qRT-PCR (Figure 4). It was, however, observed that EGF treatment had no significant effect on E-cadherin gene expression. E-cadherin is an established marker of epithelial cell phenotype [4]. Similarly, treatment of LH86 cells with TGFβ-1, EGF, HGF, bFGF and a combination of TGFβ-1 and EGF led to decreased albumin and alpha-1 antitrypsin (AAT) expression by LH86 cells as determined by quantitative immunofluorescence analysis (Figure 5). Untreated cells showed good expression and remained epithelial in phenotype as observed from the close proximity of the cells and cell-cell adhesion. The effects of TGFβ-1 treatment on decreased albumin and AAT expression were confirmed with qRT-PCR (Figure 5). Further, treatment of Huh7 cells with TGFβ-1, EGF, HGF, bFGF and a combination of TGFβ-1 and EGF led to decreased E-cadherin protein expression as determined by quantitative immunofluorescence analysis (Supplementary figure 2). qRT-PCR analysis revealed that TGFβ-1 treatment of Huh7 cells led to decreased E-cadherin and albumin gene expression (Supplementary figure 2). Albumin is an epithelial marker and a marker of hepatocyte function [20]. AAT is also a marker of hepatocyte function [21].

Figure 4. Multiple growth factors induced decreased E-cadherin expression in HCC cells.

A. LH86 HCC cells were treated for 7 days with 2ng/ml TGFβ-1 or 50ng/ml EGF or 5ng/ml HGF or 20ng/ml bFGF or a combination of 2ng/ml TGFβ-1 and 50ng/ml EGF. E-cadherin expression was assessed with immunofluorescent staining. Image magnification was X20 objective. Green coloration represents E-cadherin expression and blue coloration represents DAPI-stained nuclei. B. Quantitative immunofluorescent analysis of E-cadherin protein expression (corrected integrated density normalized to cell number). C. Real-time quantitative RT-PCR analysis of E-cadherin gene expression (normalized to GAPDH). Results expressed as mean ± SEM. * P < 0.05; ** P < 0.01, control cells versus treated cells; NS, not significant; N = 3.

Figure 5. EMT in HCC cells is associated with dysregulation of hepatocyte function.

LH86 HCC cells were treated for 7 days with 2ng/ml TGFβ-1 or 50ng/ml EGF or 5ng/ml HGF or 20ng/ml bFGF or a combination of 2ng/ml TGFβ-1 and 50ng/ml EGF. Protein of expression of albumin (A) and AAT (B) was assessed with immunofluorescent staining. Image magnification was X20 objective. Green coloration represents albumin and AAT expression respectively and blue coloration represents DAPI-stained nuclei. C, D. Quantitative immunofluorescent analysis of albumin and AAT protein expression respectively (corrected integrated density normalized to cell number). E. Real-time quantitative RT-PCR analysis of albumin and AAT gene expression (normalized to GAPDH). Results expressed as mean ± SEM. * P < 0.05; ** P < 0.01, control cells versus treated cells; NS, not significant; N = 3.

The role of COX-2 and Akt

We examined the role of COX-2 and Akt as possible intracellular signaling mechanisms underlying the process of growth factor-induced EMT. The effects of growth factor treatment of LH86 cells on expression of COX-2 and Akt were analyzed with qRT-PCR. TGFβ-1, EGF, HGF, bFGF all stimulated significantly increased expression of both COX-2 mRNA and Akt1 mRNA (Figure 6). Treatment of cells with a combination of TGFβ-1 and EGF, however, did not lead to any significant further increase over either alone. Further, we found that growth factor treatment also led to increased PGE2 secretion and increased total Akt (Akt1, Akt2 and Akt3) protein expression. Increased p-Akt (at Ser 473) was detected in only HGF- and bFGF-treated cells. To determine if COX-2 or Akt had a direct role in mediating growth factor-induced EMT in HCC, we used TGFβ-1 as the representative growth factor and vimentin and E-cadherin as representative markers of the EMT process. Our experiments revealed that inhibition of COX-2 with NS398 or the Akt pathway with LY294002 significantly inhibited TGFβ-1-induced increased expression of vimentin mRNA and reversed TGFβ-1-induced loss of E-cadherin protein expression (Figure 6). It is also noteworthy that NS398 on its own significantly reversed the TGFβ-1-induced increased expression of vimentin and loss of E-cadherin. To determine if the NS398 effect is a specific COX-2 effect, we examined the effect of transfection of a specific COX-2 siRNA on vimentin gene expression. We found that COX-2 gene knock-down significantly decreased basal vimentin gene expression. The effects of LY294002 on its own were, however, not significant.

Figure 6. Growth factor-induced EMT in HCC is dependent on COX-2 and Akt.

LH86 HCC cells were treated for 7 days with 2ng/ml TGFβ-1 or 50ng/ml EGF or 5ng/ml HGF or 20ng/ml bFGF or a combination of 2ng/ml TGFβ-1 and 50ng/ml EGF. All treatments resulted in increased COX-2 and Akt (Akt1) mRNA expression (A, C) and PGE2 secretion (B) as determined by qRT-PCR (normalized to GAPDH) and specific ELISA respectively. 7 days treatment with 2ng/ml TGFβ-1 or 50ng/ml EGF or 5ng/ml HGF or 20ng/ml bFGF also stimulated increased total Akt (Akt1, Akt2 and Akt3) protein expression. Increased p-Akt (at Ser 473) expression was, however, detected in only HGF- and bFGF-treated samples. D (Top) shows immunoblotting of 20μg total protein. D (Bottom) shows immunoblotting of 60μg total protein. Inhibition of COX-2 (with 5μM NS398) and the Akt pathway (with 10μM LY294002) inhibited TGFβ-1-induced increased vimentin gene expression (E) and loss of E-cadherin protein expression (H) as determined by qRT-PCR (normalized to GAPDH) and immunoblotting respectively. F and G show effect of specific COX-2 siRNA gene knock-down on COX-2 and vimentin mRNA expression respectively. Results expressed as mean ± SEM. * P < 0.05; ** P < 0.01, control versus growth factor or versus inhibitor; ^# P < 0.05; ^## P < 0.01, TGFβ-1 versus TGFβ-1+inhibitor; NS, not significant; N = 3

Effects on migration and tumorigenicity

To clarify the functional effects of EMT on tumorigenicity of the LH86 HCC cell line, we examined the effects of EMT in these cells on migration using wound healing assays. LH86 cells showed increased migration after treatment with TGFβ-1 or EGF (Figure 7a, b). Inhibition of COX-2 with NS398 or the Akt pathway with LY294002 significantly inhibited this effect (Figure 7a, b). To further confirm the functional effects of EMT on our novel HCC cell line, we performed some tumorigenicity studies. LH86 cells that were treated with TGFβ-1 for 11 days were transplanted into the flanks of immune-deficient mice. These cells had undergone mesenchymal change as observed by microscopy. We observed that tumors had developed by 4 weeks and that tumors from TGFβ-1-treated LH86 cells were 135.67±36% more voluminous than tumors from untreated LH86 cells (Figure 7c).

Figure 7. Growth factor-induced EMT is tumorigenic and promotes migration in HCC cells. A.

LH86 HCC cells that were treated for 7 days with 2ng/ml TGFβ-1 or 50ng/ml EGF or 5μM NS398 (COX-2 inhibitor) or 10μM LY294002 (Akt pathway inhibitor) were used in wound healing migration assays. B. Quantitative analysis of migration assays (% of cell-covered area). In (C) LH86 HCC cells that were treated for 11 days with 2ng/ml TGFβ-1 and had demonstrated morphological evidence of EMT were transplanted into immune deficient mice for 4 weeks. Untreated cells were also transplanted for comparison. Assessment of tumor volume was made by multiplying length by width by depth. Results expressed as mean ± SEM. * P < 0.05; ** P < 0.01, N = 3.

Discussion

Cancer progression is characterized by invasion, metastasis, resistance to therapy and recurrence after therapy. This progression, if unstopped, usually leads to mortality. It is an established fact that metastasis is the commonest cause of mortality due to cancer [1]. Consequently, it is now recognized that to find effective management strategies for cancers we need to extend our understanding of the basis of cancer progression and metastasis. We are, therefore, reporting the study of mechanisms promoting and mediating epithelial-mesenchymal transition (EMT) (a process known to play a key role in cancer progression [2]) in hepatocellular carcinoma (HCC).

This is the first study of the process of EMT in the novel cell line LH86. This cell line usually has typical epithelial morphology as shown in Figure 1. We are, however, reporting for the first time that this novel cell line is capable of undergoing EMT. Hence, it is a good model for studying the process of EMT.

Prior to our study, virtually all studies of EMT in HCC have relied exclusively on TGFβ-1 as an inducer of EMT. The role of TGFβ-1 as an inducer of EMT has long been known. It was originally described in early embryogenesis [22]. Since then, however, its role in carcinogenesis has been reported in a variety of cancers [23] [24]. Reports have shown that other factors may also induce EMT. It was recently reported that EMT was inducible in pancreatic adenocarcinoma cells with defective TGFβ signaling, thus suggesting that TGFβ is unlikely to be the only inducer of EMT [23]. Consequently, we examined the effects of TGFβ-1 and other growth factors (EGF, HGF and bFGF) on EMT in the LH86 HCC cell line. We observed that all four growth factors induced mesenchymal changes in this HCC cell line after 5 days of treatment. Mesenchymal morphology was maintained beyond 11 days. Cells were subsequently passaged due to confluency. For consistency, in our in vitro experiments we used the 7-day time point. Observations of the effects of EMT on morphology show that EMT induction leads to fibroblastoid phenotype. Furthermore, we examined the effects of combined action of growth factors by comparing the effects of TGFβ-1 treatment alone with the effects of combined treatment with TGFβ-1 and EGF. Using bright field microscopy, we observed that induction of EMT appeared to be quicker and cells appeared to develop more profound mesenchymal phenotype morphologically when treated with TGFβ-1 and EGF together as opposed to when treated with TGFβ-1 alone.

EMT is additionally characterized using established molecular markers: loss of E-cadherin expression, increased expression of vimentin, fibronectin and collagen I. It has long been established that loss of E-cadherin is an important characteristic of EMT [4]. The factors that drive these molecular changes and consequent EMT in cancer cells, however, remain largely unclear. We, therefore, examined the hypothesis that a number of growth factors alone or in combination may be promoting these EMT-associated molecular changes in HCC cells. Our data shows that TGFβ-1, EGF, HGF and bFGF promote EMT-associated molecular changes in the LH86 HCC cell line. Although our microscopic study of morphology initially suggested that the combination of TGFβ-1 and EGF may be more potent than either alone in inducing EMT, analysis of molecular characteristics however suggests that the combination of growth factors may not necessarily be more important than any of them on its own in inducing EMT in HCC. Furthermore, although EGF treatment resulted in increased vimentin, fibronectin and collagen I expression, it had no significant effect on E-cadherin expression. The reason for this is currently unclear. Interestingly too, we compared the effect of treatment with growth factors for 7 days with the effect of treatment for 20 days on vimentin gene expression and found that there was further increase with 20 days treatment. This was, however, statistically significant for only bFGF. Again, the specific reasons for this are currently unclear but from the current data we believe that continued exposure to these EMT-inducing growth factors results in further EMT induction and that bFGF may be a more important long-term inducer of EMT in HCC than TGFβ-1, EGF and HGF. Further studies in the future will be required to clarify the role of individual molecular markers of EMT with regards to the chronological sequence of changes and the details of their regulation.

Immunostaining for mesenchymal markers was very instructive: it made it obvious that EMT induction under the influence of TGFβ-1, EGF, HGF and bFGF leads to increased production of extracellular matrix material. Not only are enlarged cells with large nuclei seen, but these cells are often observed to be surrounded with a mesh of material positive for vimentin, collagen I and fibronectin and obviously devoid of E-cadherin, albumin and AAT. Interestingly, the observed effects of multiple growth factor – induced EMT in LH86 cells were comparable to those in the well-established human Huh7 HCC cell line.

Furthermore, we were interested in the effect of growth factor-induced EMT on HCC cell tumorigenicity. We examined tumorigenicity in immune-deficient mice and found that EMT induction led to increased tumorigenicity. Immunohistochemical and qRT-PCR analysis of tumor cells from mouse xenograft tumors revealed that sustained induction of EMT was associated with increased tumorigenicity (see Supplementary figure 3). This confirms previous thinking that EMT is a process that promotes tumor progression. Macrometastasis was not detected in our system, although it is possible that micrometastasis may have taken place. EMT induction was further associated with increased migratory capacity. This also further confirms current thinking that EMT is an important process in tumor metastasis.

Although it was shown that EMT in rat primary hepatocytes led to reduced albumin expression [6], literature search revealed that this phenomenon has not been tested in the context of EMT in human HCC. Consequently, this is the first study of the effect of EMT on human HCC cell function. This was examined by examining the effect of EMT on expression of albumin and AAT. Albumin is known to be an epithelial marker in the liver and an important determinant of optimal hepatocyte function. Our study has shown that EMT in HCC leads to decreased expression of albumin. This confirms that the process actually leads to loss of the epithelial phenotype. It also suggests that the process impacts on normal hepatocyte function and may explain the associated hepatic dysfunction and failure that is often characteristic of advanced liver cancers. This study is also the first to show that EMT leads to significant down-regulation of the hepatocyte-specific protein AAT [21] in human hepatocytes. This may further explain the hepatic dysfunction characteristic of advanced liver cancers. Taken together, our data on the effects of EMT in HCC on albumin and AAT expression suggest that EMT in human liver cancers has significant functional implications for hepatocytes: it ultimately makes the hepatocytes less effective in performing their functions which ultimately leads to their failure.

Prior to this study, the intracellular signaling mechanisms underlying the process of EMT in human HCC had not been elucidated. Our current data show that TGFβ-1, EGF, HGF and bFGF all stimulate increased expression of COX-2 mRNA and PGE2 secretion and Akt mRNA and protein. Although not often reported, growth factor-induced increased Akt mRNA expression was shown in a recent study of murine hepatocellular carcinoma [25]. Also, it is noteworthy that increased p-Akt was detected in only HGF- and bFGF-treated cells when only 20μg of total protein was analyzed. It is possible that with significant increase in the quantity of total protein we may detect increased p-Akt in TGFβ-1 and EGF-treated cells too. This is because when only 20μg of total protein was analyzed even increased total Akt was not detectable in TGFβ-1-treated cells. However, when the total protein sample was increased to 60μg increased total Akt protein expression was now detectable. The increased p-Akt expression observed in HGF-treated cells in comparison to TGFβ-1 and EGF-treated cells is consistent with the Akt mRNA, PGE2 secretion and COX-2 mRNA data. In every case, HGF appears to have the most profound effect. Interestingly, in our studies using a murine hepatocellular carcinoma model HGF also had profound effect in EMT induction via COX-2 and Akt pathways [25].

Further, we found that inhibition of either COX-2 or the Akt pathway can inhibit the process of EMT in HCC. More importantly, we show for the first time that inhibition of COX-2 reversed molecular characteristics of EMT. This finding is particularly very important because specific COX-2 inhibitors are already in widespread clinical use and they have a much better safety profile than many chemotherapeutic agents. Consequently, the current data provide a very strong basis for exploring the use of specific COX-2 inhibitors as adjuvant chemotherapeutic or chemopreventative agents in advanced and metastatic HCC.

In conclusion, using a novel human HCC cell line we have shown that EMT in HCC is induced by multiple growth factors (TGFβ-1, EGF, HGF and bFGF). It may be that a combination of these growth factors may not necessarily be more important than any one of them alone, although further studies are required to clarify this. Growth factor-induced EMT makes HCC more tumorigenic and alters basal functional capacities of HCC cells. It is also associated with increased expression of COX-2 and Akt. Moreover, inhibition of COX-2 and the Akt pathway inhibited the process of EMT and inhibition of COX-2 completely reversed the process. Consequently, targeting COX-2 and Akt pathways may be a useful adjuvant or neo-adjuvant treatment strategy in patients with advanced HCC and a possible role for COX-2 inhibitors as chemopreventative agents in HCC should be investigated. Furthermore, the novel cell line LH86 is a good model for studying the role of EMT in HCC progression.

Abbreviations used

AAT

alpha-1 antitrypsin

EMT

epithelial-mesenchymal transition

bFGF

basic fibroblast growth factor

EGF

epidermal growth factor

TGFβ-1

transforming growth factor β-1

HGF

hepatocyte growth factor

COX-2

cyclooxygenase-2