Blocking transforming growth factor–beta up-regulates E-cadherin and reduces migration and invasion of hepatocellular carcinoma cells


  • Emilia Fransvea,

    1. Department of Internal Medicine, Immunology and Infectious Diseases, Section of Internal Medicine; University of Bari Medical School, Bari, Italy
    Search for more papers by this author
  • Umberto Angelotti,

    1. Department of Internal Medicine, Immunology and Infectious Diseases, Section of Internal Medicine; University of Bari Medical School, Bari, Italy
    Search for more papers by this author
  • Salvatore Antonaci,

    1. Department of Internal Medicine, Immunology and Infectious Diseases, Section of Internal Medicine; University of Bari Medical School, Bari, Italy
    Search for more papers by this author
  • Gianluigi Giannelli

    Corresponding author
    1. Department of Internal Medicine, Immunology and Infectious Diseases, Section of Internal Medicine; University of Bari Medical School, Bari, Italy
    • Dipartimento di Clinica Medica, Immunologia e Malattie Infettive, Sezione di Medicina Interna, Policlinico, piazza G. Cesare 11, 70124 Bari, Italy
    Search for more papers by this author
    • fax: ++39 (080) 5478-126.

  • Potential conflict of interest: Nothing to report.


Hepatocellular carcinoma (HCC) treatment is challenging because the mechanisms underlying tumor progression are still largely unknown. Transforming growth factor (TGF)–β1 is considered a crucial molecule in HCC tumorigenesis because increased levels of patients' serum and urine are associated with disease progression. The aim of the present study was to investigate the inhibition of TGF-β signaling and its impact on HCC progression. Human HCC cell lines were treated with a TGF-β receptor kinase inhibitor (LY2109761) whose selectivity was determined in a kinase assay. Exogenous TGF-β1 phosphorylates the TGF-β receptor, consequently activating Smad-2, whereas the drug selectively blocks this effect and dephosphorylates autocrine p-Smad-2 at concentrations ranging from 0.001 to 0.1 μM. A cytotoxic effect documented by 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT), trypan blue, and propidium iodide staining assays was observed at 10μM, whereas the drug inhibits (P < 0.001) the migration of HCC cells on fibronectin, laminin-5, and vitronectin and invasion through Matrigel (P < 0.001) at concentrations up to 0.1 μM. LY2109761 up-regulates (P < 0.001) E-cadherin mRNA and protein levels. This increase was localized at the cellular membrane where E-cadherin mediates anchorage that is cell–cell dependent. Consistently, a functional monoclonal antibody that inhibits E-cadherin–dependent cell–cell contact restores the migratory and invasive activity. Finally, nonmetastatic HCC tissues from 7 patients were cultured with TGF-β1 in the presence or absence of LY2109761. E-cadherin expression was reduced by TGF-β1 and was significantly (P < 0.0001) increased by LY2109761 treatment, measured by quantitative real-time PCR on microdissected tissues and by immunohistochemistry on serial sections. In 72 patients, E-cadherin tissue expression was more weakly expressed in metastatic than in nonmetastatic HCC (P < 0.0001). Conclusion: LY2109761 blocks migration and invasion of HCC cells by up-regulating E-cadherin, suggesting that there could be a mechanistic use for this molecule in clinical trials. (HEPATOLOGY 2008.)

Hepatocellular carcinoma (HCC) is a highly malignant cancer that is the third most frequent cause of tumor-related death in the United States and Europe.1 Current therapeutic options are invasive and aim to physically remove or destroy the tumor mass. However, later recurrence and/or metastatic spread are common and negatively affect survival. The overall prognosis is still unsatisfactory, and little progress has been made in finding new treatment options. However, a recent study with sorafenib suggests that targeting the vasculature may provide additional insights into how to develop future treatment options for patients with HCC.2

Based on the recent clinical observations with sorafenib, the focus for developing new treatments in HCC has shifted from targeting the cancer to targeting the tissue microenvironment and its role in modulating the biological behavior of HCC.3 Transforming growth factor (TGF)–β1 is stored in latent form in the tissue microenvironment and becomes active as a consequence of proteolytic tissue remodeling.4, 5 The active form binds to TGF receptor II, which phosphorylates TGF-βRI, resulting in formation of a heterotetrameric complex. Downstream signaling is mediated by the proteins Smad-2 and Smad-3, assembled, and phosphorylated in the Smad2/3 complex.

In HCC patients, TGF-β1 has been reported to be overexpressed in both blood and urine, correlating with a worse prognosis and survival and thus representing a marker of this cancer.6–8 Its role is very intriguing because it acts as a tumor suppressor but also as a tumor promoter of invasion and metastasis, as reported in a recent review.9 We have shown that TGF-β1 plays a key role in modulating HCC aggressiveness by triggering the epithelial to mesenchymal transition (EMT) of the cells, thereafter completed by the extracellular matrix protein laminin-5 (Ln-5), which has been reported to be expressed at the advancing edge of invasive HCC.10, 11

Another molecule associated with EMT is E-cadherin,12 a cell–cell adhesion molecule that ensures maintenance of the tissue architecture. Down-regulation of E-cadherin is one of the earliest steps in the invasive cascade process that allows cancer cells to invade through surrounding tissues. Hence, this down-regulation is considered a hallmark of invasiveness in a number of different malignancies including HCC.

In the present study, we investigated the results of inhibiting TGF-β and its ensuing modulation of HCC tumor progression. To achieve this goal, we used a TGF-βRI kinase inhibitor (LY2109761), which may in the future offer a novel approach to the treatment of HCC.

Materials and Methods


The human HCC cell lines HLE and HLF were cultured as previously described.13 The selectivity of LY2109761, a TGF-βRI kinase inhibitor, was screened against 66 kinases (Eli Lilly, personal communication) using a protein kinase assay from Upstate Biotech Inc. (UBI), as previously reported.14 The compound was kindly provided by Eli Lilly (Indianapolis, IN). TGF-β1 was purchased from Sigma (Milan, Italy), monoclonal blocking antibody against human E-cadherin, SHE78-7, from Alexis (Lausanne, Switzerland) or from BD (Biosciences, San Jose, CA), and monoclonal blocking antibody against TGF-β from R&D Systems (Minneapolis, MN). Polyclonal antibodies against phospho-Smads were purchased from Cell Signaling Technology Inc. (Danvers, MA) and monoclonal antibody against β-actin from Sigma. Ln-5 was purified as previously described.15 Fibronectin (Fn) and vitronectin (Vn) were purchased from Calbiochem (La Jolla, CA) and growth factor–reduced (GFR) Matrigel from BD.

Patients and Tissue Specimens.

We studied 72 patients with chronic liver disease who developed a single HCC, diagnosed according to different radiological techniques and classified according to the BCLC classification (Table 1).16 Tissue samples were obtained by surgical biopsy. As previously described,11 the patients were assigned to the metastatic (37 patients) or nonmetastatic (35 patients) disease group based on intrasurgical examination and histological observation of the micronodules bordering the HCC.

Table 1. Demographic Characteristics of Patients
CharacteristicHCC without MtxHCC with Mtx
Number of patients3537
Age (years, mean value)70 ± 8.767.6 ± 9.2
 Male26 (74%)29 (78%)
 Female9 (26%)8 (22%)
Cirrhosis etiology  
 Multiple viral infection44
 Alcohol abuse + viral infection22
Stage A (BCLC staging system)  
 A111 (31%)10 (27%)
 A216 (46%)18 (49%)
 A38 (23%)9 (24%)
 0–20 IU/mL16 (46%)17 (46%)
 20–200 IU/mL9 (26%)9 (24%)
 > 200 IU/mL10 (28%)11 (30%)
 > 3 cm20 (57%)23 (62%)
 < 3 cm15 (43%)14 (38%)

Immunohistochemistry and Staining Quantification.

Indirect alkaline phosphatase staining was used to localize E-cadherin in frozen tissue as described.13 Staining was quantified in each section as the total stained area, calculated as the mean of 10 randomly chosen microscopic fields using appropriate image-analyzer software, as already reported.17

Cytotoxicity Assays.

LY2109761 cytotoxicity was determined by 3 methods: the 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazoliumbromide (MTT) assay, manual counting of viable cells, and propidium iodide staining. MTT (Sigma) yields a purple formazan product that is detected using a 96-well plate reader at 570 nm. Cells were plated and cultured for 2 days in a 1% fetal bovine serum medium supplemented with LY2109761 at the following concentrations: 0.001, 0.01, 0.1, 1, 10, and 20 μM. Each experimental condition was reproduced in 8 wells, and each experiment was repeated 3 times.

To confirm the cytotoxic data, cells were incubated under the described conditions and stained with the vital dye trypan blue (Sigma), which does not react with the cell membrane because of its negative charge. All the unstained cells were counted using a hemocytometer. Four squares were counted for each condition, and each condition was repeated in triplicate in the same experiment. Each experiment was repeated 3 times for each cell line. Bars represent the average and standard deviation of all experiments. Under the same experimental conditions, nonpermeabilized cells were stained with propidium iodide (Sigma) and analyzed with a flow cytometer (Beckman Coulter, Fullerton, CA).

Flow Cytometry.

Cultured cells were detached with Versene (Sigma). Then, nonpermeabilized cells were incubated with monoclonal antibodies against E-cadherin, washed, and incubated with a fluorescein isothiocyanate–labeled goat antimouse secondary antibody. Cells were suspended in 0.5 mL of PBS, stained with propidium iodide, and analyzed on a FACScan flow cytometer from Beckman Coulter (Fullerton, CA). Staining with the isotype control was run in parallel.

Cell Adhesion and Migration.

Migration and adhesion assays were performed as previously described.18 Cells were serum-starved overnight and then cultured in the absence or presence of LY2109761 at concentrations ranging from 0.001 to 0.1 μM, for 48 hours. In some experiments E-cadherin-blocking antibody, SHE78-7 (1 μg/mL), or isotype-specific control antibody was added to the migration medium of drug-treated cells.

Chemoinvasion through a Reconstituted Basement Membrane.

The invasion assay was performed as previously described.13 HLE and HLF cells were treated with LY2109761 as described for the migration assay, and Ln-5 (1 μg/mL) was used as a chemoattractant. In some experiments the E-cadherin-blocking antibody SHE78-7 (1 μg/mL) was added to cells during the experiment. The total number of invasive cells was quantified by image-analysis software as already described.10 For each condition, 6 filters were used for every experiment.

3-D Culture.

The first layer was obtained by adding GFR Matrigel (300 μL) diluted 1:2 and incubating in a 24-well plate. Then, HLE and HLF diluted in 400 μL of GFR Matrigel were seeded. Finally, 0.5 mL of RPMI supplemented with 2% fetal bovine serum was added on the top, and cells were cultured for 4 days. At this time Ln-5 (1 μg/mL) with or without LY2109761 (0.1 μM) was added to the medium for 48 hours. Cell morphology was photographed using a phase-contrast microscope (original magnification ×200).

HCC Ex Vivo Culture.

HCC tissue specimens were obtained from patients undergoing liver surgery for HCC without metastasis. Tissue samples were washed in serum-free RPMI and sliced into 2-mm cubes; some specimens were immediately snap-frozen, and others were cultured for 48 hours in serum-free RMPI supplemented with L-glutamine and antibiotics and submerged in 24-transwell filters as previously reported.19 Tissue samples were cultured in the presence of TGF-β1 (3 ng/mL) with or without LY2109761 (0.1 μM) and then snap-frozen in liquid nitrogen and processed for immunohistochemistry. Serial sections were microdissected by a laser capture microscope (Nikon) and processed by quantitative real-time RT-PCR. Experiments were performed using tissue samples from 7 patients.

Quantitative Real-Time RT-PCR Analysis.

HCC cells were cultured in serum-free conditions for 24 or 48 hours in the absence or presence of LY2109761 at concentrations ranging from 0.001 to 0.1 μM. Total RNA was extracted from whole cells using an RNeasy minikit (Qiagen, Valencia, CA) and reversed-transcribed using a Retro-script kit (Ambion, Austin, TX). Real-time PCR analysis was performed according to the manufacturer's instructions as previously described.10 E-cadherin gene expression was measured and compared with GAPDH expression. Sequences of the oligonucleotides were 5′-GGAGTCAACGGATTTGGT-3′ and 5′-CAG-TCAAAAGGCCTCTACGG-3′ for GAPDH, with a product of 206 bp, and 5′-CAGTCAAAAGGCCTCTACGG-3′ and 5′-GTGTATGTGGCAATGCGTTC-3′ for E-cadherin, with a product of 436 bp.

Western Blot Analysis.

Cell lysates were prepared directly in the cell extraction buffer (Biosource International Inc., CA), which enabled us to investigate both cytoplasmic and cell membrane expression of E-cadherin, supplemented with a protease inhibitor cocktail (Sigma). Protein concentration was measured by the bicinchoninic acid method (Pierce Chemical Co., Rockford, IL).

Statistical Analysis.

Quantitative variables have been summarized as mean and standard deviation. Differences in E-cadherin tissue levels for normally distributed data were assessed with the Student t test.


The activation status of the TGF-β pathway was determined by p-Smad-2 expression using Western blotting in the 2 HCC cell lines, HLE and HLF, featuring strong migratory and invasive properties.13

In serum-free conditions HLE cells expressed detectable levels of autocrine phosphorylated Smad-2 that progressively increased in a time-dependent manner following incubation with TGF-β1 (3 ng/mL) and peaked after 1 hour (Fig. 1A). Therefore, to study the inhibitory effect of LY2109761 on the TGF-β1-mediated pathway, we challenged HLE cells, stimulated with TGF-β1 (3 ng/mL) for 1 hour, with different concentrations of the compound. The TGF-β1-induced phosphorylation of Smad-2 in HLE cells was completely blocked by LY210976 in a dose-dependent manner (Fig. 1B), ranging from 0.001 to 3 μM, whereas no effect was observed on the total Smad forms, suggesting a selective action of the drug on the TGF-β1–stimulated pathway via dephosphorylation of Smad-2. Thanks to this selective effect, LY2109761 was used to inhibit the TGF-β1 autocrine phosphorylated pathway to investigate the ensuing biological effects. As shown in Fig. 1C, strong inhibition of p-Smad-2 was observed with drug concentrations up to 0.001 μM. The selectivity of the drug was further supported by the data in Supplementary Fig. 1. Similar results were observed with HLF cells (data not shown). In conclusion, LY2109761 selectively blocks the exogenous and endogenous TGF-β1-mediated pathway, as demonstrated by the dephosphorylation of Smad-2.

Figure 1.

LY2109761 selectively inhibits the exogenous and endogenous TGF-β1-stimulated pathway. HLE cells stimulated with exogenous TGF-β1 were lysed and processed for Western blotting. (A) The phosphorylation of Smad-2 increased following TGF-β1 incubation. (B) LY2109761 induced a dose-dependent reduction in phosphorylation of Smad-2. (C) HLE endogenous phosphorylation of Smad-2 was inhibited by LY2109761.

To exclude the possibility that the biological effects of LY2109761 could be related to cytotoxicity of the molecule, both HLE and HLF cells were incubated for 48 hours at strong drug concentrations ranging from 0.001 to 20 μM. As reported in Fig. 2A, LY2109761 showed some cytotoxic effect from 10 μM, whereas no effect was observed at lower concentrations. Consistent with these data, a similar cytotoxic effect was observed when counting the number of cells refractory to trypan blue staining with light microscopy and numbering the nonpermeabilized cells stained with propidium iodide with flow cytometry (Fig. 2B,C).

Figure 2.

LY2109761 toxicity is dose dependent. (A) LY2109761 toxicity appeared at the 10 μM concentration, measured by the MTT assay. Similar results were obtained with trypan blue (B) and propidium iodide staining (C) of nonpermeabilized cells. LY2109761 was not cytotoxic at 0.001 (black line), 0.01 (gray line), 0.1 (yellow line), or 1 (green line) μM, whereas it was toxic at 10 (orange line) and 20 (blue line) μM.

At this point, the biological activities of LY2109761 were tested in cell motility functional assays at different doses shown to be noncytotoxic but still able to selectively dephosphorylate the autocrine Smad-2-mediated pathway. In these assays, cells were treated under the same experimental conditions used to demonstrate the autocrine inhibition of p-Smad-2. As shown in Fig. 3A, both the HLE and HLF cell lines efficiently migrated on Ln-5, Fn, and Vn, but migration was strongly (P < 0.0001) inhibited by LY2109761, whereas cell adhesion was not affected by LY2109761 treatment (Fig. 3B). This suggests that the drug does not have any effect on the cell–matrix interaction mediated by integrins

Figure 3.

LY2109761 inhibits migration but not adhesion of HCC cells A: LY2109671 significantly inhibits HLE and HLF migration, as measured in a Boyden modified chamber in a dose range shown to be noncytotoxic [*P < 0.0001 with respect to controls (CTR)]. Magnification ×400. B: The same drug treatments do not affect cell adhesion.

Both HLE and HLF, embedded in a thick 3-dimensional gel made of Matrigel, grew as round, nodule-like structures with well-defined borders in control conditions. However, in the presence of Ln-5, the morphological aspect of the nodules underwent dramatic changes. The borders were no longer defined because of the spread of cells at the periphery that took on an elongated shape, penetrating through the surrounding gel and causing the whole nodule to take on a “crab”-like morphology. This effect was completely reversed by the presence of LY2109761 (Fig. 4A). This anti-invasive property was further investigated by a Matrigel invasion assay using Ln-5 as the chemoattractant (Fig. 4B). HLE and HLF strongly invaded through Matrigel, but LY2109761 blocked the invasive activity of both cell lines. In conclusion, LY2109761 blocks migration on different ECM proteins and invasion of both HLE and HLF through a 3-dimensional structure.

Figure 4.

LY2109761 inhibits HCC cell invasion. HLE cells grow as round nodule-like structures in a thick 3-dimensional Matrigel, whereas in the presence of Ln-5, the nodule is disaggregated, and peripheral cells penetrate the surrounding gel. A: The presence of LY2109761 completely blocks this effect (magnification ×100). B: LY2109671 significantly (P < 0.001) inhibits HLE and HLF invasion through Matrigel, using Ln-5 as a chemoattractant in a dose range shown to be noncytotoxic (magnification ×20).

To gain insight into the molecular mechanism underlying these biological effects, the expression of E-cadherin was investigated on HLE and HLF cells incubated with LY2109761 (Fig. 5A). After 24 hours of incubation with LY2109761, Expression of E-cadherin mRNA was strongly (P < 0.0001) up-regulated, albeit with some differences, in HLE and HLF cells compared with that in the controls, as measured by quantitative real-time PCR, whereas the level of E-cadherin protein was not increased. By 48 hours of LY2109761 treatment, expression of E-cadherin mRNA was still up-regulated, but the protein level was also strongly (P < 0.0001) increased in both cell lines. In short, LY2109761 increased E-cadherin mRNA expression after 24 hours and protein levels after 48 hours.

Figure 5.

LY2109761 up-regulates E-cadherin in HCC cells. HLE and HLF were treated with different doses of LY2109761 for 24 and 48 hours. RNA and protein were analyzed by quantitative real-time PCR and western blotting. LY2109761 significantly (P < 0.001) up-regulated E-cadherin mRNA after 24 hours and protein levels after 48 hours. A: Graphs indicate the ratio between E-cadherin protein levels and β-actin used for normalization. Cells were treated for 48 hours with LY2109761 and analyzed by flow cytometry to investigate the E-cadherin present on the cellular membrane. B: All concentrations of LY2109761 increased the expression of E-cadherin on the cell surface.

To better characterize the increase in E-cadherin, both HLE and HLF cells were treated with LY2109761 for 48 hours under the same experimental conditions described and analyzed by flow cytometry. Cells were not permeabilized in order to detect E-cadherin localized at the cellular surface. Consistent with Western blot data, HLE and HLF cells expressed very low levels of E-cadherin at baseline, but increased expression on the cell membrane was observed with all concentrations of LY2109761 (Fig. 5B).

To test the hypothesis that the increased expression of E-cadherin is functionally active, LY2109761-treated HLE and HLF cells were challenged to migrate on different ECM proteins, in the presence of a functional antibody that inhibits E-cadherin-dependent cell–cell contact. The migratory activity of HLE and HLF cells on Ln-5, Fn, and Vn, inhibited by LY2109761, was completely restored by the presence of the blocking antibody against E-cadherin, whereas the addition of the antibody to the cells did not show any effect (Fig. 6A). Consistently, the same antibody also restored the invasive ability of HLE and HLF cells, otherwise blocked by LY2109761, through Matrigel (Fig. 6B) without affecting their invasive ability when used alone. Similar results were also observed with a monoclonal blocking antibody against TGF-β (Supplementary Fig. 2). In short, LY2109761 blocks migration and invasion, increasing the expression of E-cadherin along the cell surface.

Figure 6.

Anti–E-cadherin functional antibody restores migration and invasion of LY2109761-treated HCC cells. A blocking antibody against E-cadherin (SHE78-7) that inhibits the E-cadherin–dependent cell–cell contact restores the (A) migratory and (B) invasive activity.

Finally, the modulation of E-cadherin by LY2109761 (0.1 μM) was investigated in ex vivo experiments using human HCC specimens from 7 patients cultured for 48 hours with TGF-β1 (3 nm/mL) in the presence or absence of LY2109761 or with medium alone; the results were consistent in all the experiments. To exclude the possibility of an experimental condition modulating E-cadherin expression, a noncultured specimen of each tissue was included as a control. Tissues were microdissected by laser capture microscopy, and mRNA was analyzed by quantitative real-time PCR as previously described. Tissues cultured in the presence of TGF-β1 displayed significantly (P < 0.001) lower E-cadherin mRNA levels than controls, whereas tissues cultured with TGF-β1 and LY2109761 had higher (P < 0.001) expression of E-cadherin mRNA compared with the controls but also with TGF-β1-treated tissues (Fig. 7). No differences were observed between cultured and noncultured tissue. Previously microdissected serial sections were also processed by immunohistochemistry and, consistently with the results already described, the expression of E-cadherin was similar in the cultured and noncultured tissue samples, being decreased in tissues treated with TGF-β1 as compared with controls but strongly increased and distributed around the cell surface at cell–cell contacts in tissue treated with TGF-β1 plus LY2109761.

Figure 7.

LY2109761 up-regulates E-cadherin in HCC tissues. Tissue specimens were cultured for 48 hours in the presence of TGF-β1 with or without LY2109761. E-cadherin was measured by real-time PCR on tissues microdissected by a laser capture microscope and by immunohistochemistry in serial sections. As control, some noncultured tissue was included. As shown, there was no difference between cultured and noncultured control tissues. TGF-β1 significantly down-regulated E-cadherin mRNA and protein levels while drug-treated tissues had higher levels of E-cadherin compared with that in the controls and TGF-β1 cultured tissues (P < 0.0001). (A) Noncultured. (B) Cultured. (C) TGF-β1 treated. (D) TGF-β1-plus LY2109761-treated tissues. *P < 0.001. Scale bar = 60 μm.

In the light of these findings, we also evaluated E-cadherin in the tissues of patients with metastatic or nonmetastatic HCC. E-cadherin was localized irregularly at cell–cell contacts but showed strong staining in the nonmetastatic HCC patient group, whereas in the metastatic patient group, E-cadherin was occasionally distributed at the cell surface of only a few cells, with very weak staining (Fig. 8). This difference, quantified in 10 microscopic fields of each section by appropriate software for imaging analysis, was statistically significant (P < 0.0001). In conclusion, E-cadherin was decreased in metastatic HCC tissue compared within nonmetastatic tissue.

Figure 8.

E-cadherin expression in metastatic and nonmetastatic HCC tissues. E-cadherin expression was evaluated by immunohistochemistry at lower (upper) and higher (bottom) magnification. E-cadherin was expressed, as expected, at cell–cell contact but was more weakly expressed in metastatic HCC than in nonmetastatic HCC. To better quantify the difference, appropriate software for imaging analysis was used. Ten randomly chosen microscopic fields from each section were captured, and E-cadherin expression was measured as the mean stained area. The box plot shows the statistically significant difference (< 0.0001). Scale bar = 60 μm.


New therapies for patients with HCC are urgently needed, even if a first such approach, namely, the use of sorafenib, has very recently been approved as a potential therapy.2 The identification of molecular targets responsible for the biological behavior of HCC certainly represents a key step toward development of such therapies.

In this regard, increased TGF-β1 in both serum and urine of HCC patients is considered a hallmark of the disease.5–7, 9, 20 For this reason, in the present study we targeted the TGF-β1 pathway with a selective TGF-βRI kinase inhibitor, namely, LY2109761, in order to block HCC tumor progression in a preclinical model.

Our data show that LY2109761 blocks migration on different ECM proteins and invasion of HCC invasive cells through a reconstituted 3-dimensional Matrigel structure. These findings can be explained by up-regulation via a transcriptional mechanism of E-cadherin localized at the cellular membrane, where it exerted an adhesive function. In fact, after abrogating the E-cadherin-mediated cell–cell contact mechanism engagement by a blocking antibody against E-cadherin, HCC cells reacquire migratory and invasive activity, further supporting the effect of LY2109761 activity on E-cadherin. This conclusion is also supported by integrin expression not being affected by drug treatment, so the ability of the cells to interact with ECM proteins is maintained This result is in agreement with a previous study showing that inhibition of E-cadherin with a blocking antibody leads to proteinase-dependent Matrigel invasion.21

Thus, increased expression of E-cadherin was able to switch the invasive phenotype of constitutively invasive HCC cells, inducing the mesenchymal-to-epithelial transition. Similar results have recently been reported by others, who showed that LY2109761 induces the expression of the Coxachie and adenovirus receptor (CAR), a part of the tight-junction protein complexes involved in the EMT process.22 This is also consistent with our previous work, in which TGF-β1 stimulated a more invasive and aggressive phenotype of constitutively noninvasive HCC cells, promoting the EMT via down-regulation of E-cadherin.10

E-cadherin is a cell–cell adhesion molecule strongly implicated in the invasiveness of different malignancies including HCC, and down-regulation of this molecule has been proposed to have a role in facilitating tumor recurrence and progression.23, 24 In agreement with these data, we report that E-cadherin was more strongly expressed in the tissue of patients without metastasis than in those with metastasis. This inverse correlation between E-cadherin expression and metastasis occurrence has also been widely reported in cancer and further underlines the role of this molecule in the metastatic process.12 Can this model be translated into a human application? To address this important question, we cultured human specimens of HCC with TGF-β1, which is known to down-regulate E-cadherin in the presence or absence of LY2109761. Our results showed that, as expected, TGF-β1 reduced the expression of E-cadherin at the mRNA and protein levels, whereas LY2109761 up-regulated E-cadherin localization on the cell surface. The modulation of E-cadherin induced by LY2109761 could play an important role in prognosis because low expression of this molecule is predictive of a high risk (up to 88%) of tumor recurrence after surgical treatment of HCC and even after liver transplantation.25, 26

This is the first study suggesting that inhibition of the TGF-β1 pathway with LY2109761 may be a promising therapy in HCC patients, but more studies are needed to better define patients whose biological characteristics make them candidates for such a therapy.

In conclusion, LY2109761 efficiently blocks migration and invasion via the up-regulation of E-cadherin at the cellular membrane, thus affording the scientific rationale for further developing this therapy as a potential new weapon against HCC. Furthermore, in this scenario, E-cadherin is an important molecule not only because of its pathogenetic role in HCC progression but also because, being detectable in tissue and serum, it could be used as a potential marker for investigating drug effectiveness in patients.


We thank Dr. Michael Lahn (medical adviser) for his support and review of the manuscript and are grateful to Mary V. Pragnell, B.A., for language revisions.