Influence of Retinoic Acid on Adhesion Complexes in Human Hepatoma Cells: A Clue to Its Antiproliferative Effects

Abstract

Retinoic acid exerts antiproliferative and differentiative effects in normal and transformed in vitro hepatocytes. In order to verify whether these effects are related to a modulation of adhesion molecules, we used Western blot analysis and immunofluorescence microscopy to investigate the E-cadherin/β-catenin complex, the main system of adherens junctions, and the occludin/ZO-1 complex present in the tight junctions in HepG2 cells cultured in the presence or absence of retinoic acid. Results showed that retinoic acid treatment increases the amount of β-catenin bound to E-cadherin by decreasing its tyrosine-phosphorylation level. Similar results were obtained with the tight junction system, in which the amount of occludin/ZO-1 complex is increased by a similar mechanism that reduced the level of ZO-1 phosphorylation on tyrosine. Immunofluorescence images also confirm these results, showing the localization on the cell surface of both adhesion complexes. Their insertion into the plasma membrane could be suggestive of an optimal reassembly and function of adherens and tight junctions in hepatoma cells, indicating that retinoic acid, besides inhibiting cell proliferation, improves cell-cell adhesion, sustaining or inducing the expression of a more differentiated phenotype.

Keywords:

• β-catenin
• E-cadherin
• HepG2
• occludin
• ZO-1

INTRODUCTION

Cell-cell adhesion plays a pivotal role in many biological processes, such as morphogenesis, cell motility, growth control and differentiation (24, 47). Cellular adhesion is also involved in the regulation of neoplastic transformation, since the onset of primary tumors, their invasive capacity, and their metastatic potential often depend on deficiencies in adhesion systems (8). Accordingly, many of the most important adhesion proteins are considered as tumor suppressors (6, 49).

Multiproteic complexes that mediate cell-cell adhesion consist of transmembrane glycoproteins that can bind specifically to correspondent molecules present on neighboring cells; other proteins are located in the intracellular junctional plate and are involved in structural and functional linking with cytoskeletal components, such as actin microfilaments (54). The relationships between the plasma membrane and the cytoskeleton, both inside the cell and at the level of intercellular contacts, are considered fundamental in the determination of cell polarity as well as in maintaining cell shape and the integrity of the tissue to which each cell type belongs (11). This is particularly crucial for hepatocytes, in which the presence of properly functioning adhesion complexes, localized at the level of adherens and tight junctions, is responsible for their polarity, which is an essential feature for the expression and maintenance of their specialized differentiated functions (27, 45).

Recent reports have shown that such adhesion systems should not be considered as static elements, but rather as dynamic components able to receive and integrate signals coming from other cells and/or from the extracellular environment. They participate in the signal cascades that, by also reaching the nucleus, can modulate gene expression, cell cycle, and apoptosis (4). Many of the above functions of adhesion molecules are modulated by retinoids and, particularly, by retinoic acid (RA), the most metabolically active member of the vitamin A family (44). RA is also able to inhibit proliferation and to induce a more differentiated phenotype in various cell types. Retinoids are also effective in reducing, delaying or preventing neoplastic transformation and in promoting the regression of some in vivo tumors (33).

In previous reports our group provided evidence of the antiproliferative and differentiative effects of all-trans RA in normal and transformed in vitro hepatocytes (16–18). Taking into consideration that the expression of a differentiated phenotype is closely dependent on cell-cell adhesion, a critical question is whether RA exerts its function by directly modulating the expression, localization, and function of hepatocyte adhesion molecules. We addressed this issue by investigating the possible modifications in adhesion systems induced in vitro by treatment with all-trans RA, using the human hepatoma cell line HepG2, which represents a proliferating transformed system, less differentiated than normal adult hepatocytes. In particular, we focused our attention on the E-cadherin/β-catenin complex, the most important system in the stabilization and functionality of adherens junctions. The expression of E-cadherin is generally positively correlated with the stage of differentiation and with the maintenance of cell polarity; it is considered an oncosuppressor because its increase may reduce the onset of many carcinomas (22, 49, 52). In contrast, β-catenin has been associated with oncogenesis, and its overexpression is considered a useful parameter for the prognosis of several tumors (34).

β-catenin is certainly a key molecule in regulating the transition from quiescence to neoplastic proliferation (32). In particular, in normal cells this protein is associated mainly with the intracellular tail of E-cadherin, playing an important role in strengthening cell-cell adhesion, as it links E-cadherin to the actin cytoskeleton through α-catenin. Intercellular adhesion can be negatively regulated via the phosphorylation of β-catenin on tyrosine residues and this protein can accumulate in the cytoplasm (7, 37). The membrane associated and cytoplasmic pools of β-catenin have adhesion and signaling activities, respectively. The accumulation of cytoplasmic β-catenin drives its interaction with members of the LEF/TCF family of nuclear transcription factors that results in altered gene expression, which is the transduction of the Wnt/Wg signal (9, 25, 51). This accumulation of cytoplasmic β-catenin is regulated at the level of its degradation. In the absence of the Wnt/Wg signal, phosphorylation of specific serine residues by GSK-3β kinase on β-catenin induces its binding with the APC protein and leads to its ubiquitination and degradation by the proteasome. Mutations of serine residues inhibit the ubiquitination of β-catenin, which accumulates and transduces the growth signal constitutively, indicating a clear involvement of this protein in oncogenesis (32, 51).

We also addressed our attention to the occludin/ZO-1 system, present in tight junctions (46). In this system, while occludin represents one of the most important transmembrane elements of this kind of junction, ZO-1 shows similarities with β-catenin: localization in the intracytoplasmic junctional plate, connection with the actin of the cytoskeleton, and its possible involvement in intracellular signaling cascades (2, 3, 10, 36, 39). In hepatocytes, tight junctions are essential in defining the membrane domains that are the basis of hepatocyte polarity.

MATERIALS AND METHODS

Cell Cultures

The human HepG2 cells (ATCC, Rockville, Maryland) were routinely maintained in RPMI-1640 medium, supplemented with 10% fetal calf serum (Euroclone Ltd., UK), 2 mM L-glutamine, 2.5 μg/ml amphotericin-B, 10⁻⁷M dexametasone, 50 U/ml penicillin, 0.05 mg/ml streptomycin and 100 μg/ml gentamicin (Sigma, St Louis, MO, USA), at 37°C in a humidified atmosphere of 5% CO₂. Cells were plated at a density of 5 × 10³ cells/cm² and cultured for 2, 7, and 12 days in the absence or presence of 5 μM RA (Sigma). RA was used from a 5 mM stock dissolved in absolute ethanol; possible ethanol effects were preliminarily excluded. Dishes without RA were considered as controls.

Western Blotting

For Western blot analysis, HepG2 cells were solubilized in 20 mM Tris-HCl buffer containing 50 mM NaCl, 10% glycerol, 5 mM EDTA, 5 mM EGTA,1% Triton X-100, 2% SDS, 2 mM sodium orthovanadate, 16 μg/ml aprotinin, 10 μg/ml leupeptin, 16 μg/ml pepstatin and 1 mM phenylmethylsulfonyl fluoride (Sigma). The suspension was centrifuged at 12000 g for 30 minutes at 4°C to remove DNA and the proteins from the supernatant were submitted to 5% or 10% SDS-polyacrylamide gel (70 μg/lane), according to Laemmli (31). Proteins were electrophoretically transferred to nitrocellulose sheets, according to Towbin et al. (50).

The nitrocellulose sheets were incubated (90 min at room temperature) alternatively with a rabbit anti-ZO-1, a mouse anti-β-catenin, a mouse anti-E-cadherin, a mouse anti-occludin, a mouse anti-ubiquitin, a mouse anti-phosphotyrosine or a rabbit anti-phosphoserine antibodies (Zymed Laboratories, CA, USA) diluted 1:100, 1:300, 1:250, 1:300, 1:300, 1:100, or 1:700, respectively in TBS (50 mM pH 7.5 Tris-HCl, 150 mM NaCl), containing 1% nonfat milk powder. Non specific binding of the membranes was previously blocked with 5% nonfat milk powder in TBS 0.1% Tween-20 at 4°C overnight. Following incubation with an alkaline phosphatase-conjugate anti-rabbit or anti-mouse secondary antibody, the protein-antibody complexes were visualized with the BCIP/NBT (5-bromo-4-chloro-3-indolyl-phosphate/4-nitro blue tetrazolium chloride) system (BioRad). Control blots were obtained by omitting the primary antibody. Positive bands were quantified on digitized images using the Phoretix software. Results are expressed as mean ±S.D. and statistical analysis was performed by using the Student's test (P<0.05 was considered significant).

Immunoprecipitation

For immunoprecipitation, HepG2 cells were solubilized in PBS containing 1 mM calcium chloride, 1 mM magnesium chloride, 1 mM phenylmethylsulfonyl fluoride, 1% Triton X-100, 2 mM sodium orthovanadate, 20 μg/ml pepstatin, 20 μg/ml leupeptin, 15 μg/ml aprotinin (Sigma).

After centrifugation at 12000 g for 30 min at 4°C, cell lysate containing 0.5 mg of proteins was precleared for 2 h at 4°C with 25 μl of protein G-sepharose “4 fast flow” (Pharmacia Biotech, Uppsala, Sweden), and centrifuged again briefly to remove the sepharose beads. Lysate was incubated overnight at 4°C with 4 μg anti-β-catenin or ZO-1 antibody and 25 μl of protein G-sepharose. The immune complexes were collected by centrifugation, washed with lysate buffer and analyzed by Western blot analysis, as above described.

Immunofluorescence

Cells grown on coverslips, fixed in 3.7% paraformaldeyde for 10 min at 4°C, permeabilized with 0.2% Triton X-100 for 15 min, were incubated, for 1 h at 37°C, with either mouse anti E-cadherin or mouse anti-occludin (Zymed Laboratories) diluted 1:40 in PBS 1% BSA. For ZO-1 and β-catenin, cells were fixed in methanol for 10 min at -20°C and incubated for 1 h at 37°C, with either rabbit anti-ZO-1 or mouse anti-β-catenin (Zymed Laboratories) diluted 1:50 in PBS 1% BSA. They were then incubated for 1 h in FITC-conjugate goat anti-mouse or Cy3-conjugate goat anti-rabbit IgG (Zymed Laboratories) diluted 1:50 in PBS containing 10% normal goat serum. After rinsing, the coverslips were mounted on slides in an aqueous medium and examined under an epifluorescence microscope (Axioplan 2, Zeiss). Negative controls were performed by exposing slides under similar conditions, while omitting the primary antibody.

RESULTS

A quantitative evaluation of the expression of E-cadherin and β-catenin was performed by means of Western blot and densitometry. As shown in Figures 1A and 1B, densitometric analysis of both proteins, expressed as percentage of control at two days of culture, shows that the only increase is related to the time of culture, there being no significant differences between controls and RA-treated samples. It should be noted that of the two bands that can be identified on the blot for β-catenin, the lowest one corresponds to the deleted form of β-catenin and is characteristic of hepatoma cells, as reported by De La Coste et al. (12). Because both forms belong to the pool available for the cell, they were analyzed and considered as a whole.

1 Western blot analysis of E-cadherin (A) and β-catenin (B) in HepG2 cells cultured for 2, 7, and 12 days in the absence (C) or presence of retinoic acid (RA). Densitometric evaluation of the bands, expressed as percentage of control at two days of culture, shows that the only increase is related to the time of culture, there being no significant differences between controls and RA-treated samples. Results are the average of at least four different experiments ±S.D.

By immunoprecipitation, we attempted to evaluate the presence and importance of the E-cadherin/β-catenin complex; it being well known that the efficiency of cell-cell contacts is strongly dependent on this kind of interaction. The results obtained after immunoprecipitation of cell extracts with anti-β-catenin and revelation with anti-E-cadherin antibodies are shown in Figure 2A. The amount of E-cadherin bound to β -catenin is strongly increased after 7 and after 12 days of treatment with RA (40%), while there are no evident differences between treated and control cells after short treatment (48 h). Figure 2B also shows a net decrease in the level of β-catenin phosphorylation on tyrosine (35% and 45% after 7 and 12 days of treatment, respectively), confirming that the stability of the E-cadherin/β-catenin complex is negatively regulated by the grade of tyrosine-phosphorylation of the β-catenin, as reported in the literature (37). These data allow us to speculate that treatment with RA improves the stability of the cadherin-catenin system and thus improves cell-cell adhesion.

2 Western blot analysis of E-cadherin (A) and phosphotyrosine (B) after immunoprecipitation with anti β-catenin in HepG2 cells cultured for 2, 7, and 12 days in the absence (C) or presence of retinoic acid (RA). Densitometric evaluation of the bands, expressed as percentage of control after two days of culture, shows an increase of about 40% after 7 and 12 days of treatment in the amount of E-cadherin bound to β-catenin (A) and a decrease of about 35%, after 7 and 45% after 12 days of RA treatment in the amount of tyrosine-phosphorylated β-catenin (B). Results are the average of at least four different experiments ±S.D. ^*P < 0.05.

The images concerning immunolocalization experiments also agree with the above hypothesis. As observed in Figure 3, after prolonged treatment with RA (12 days), E-cadherin (b) and β-catenin (d) both localize on plasma membranes and the scattered immunoreactivity distributed on whole cells, well visible in the controls (a, c), is completely absent, indicating total insertion of the complexes at the level of cell contacts, probably the adherens junctions, where cell adherence would be improved. Moreover, β-catenin nuclear immunoreactivity, well evident in the controls, is also absent in treated cells, suggesting an RA-mediated inhibition of nuclear translocation. These effects are quite evident after 7 days of treatment, whereas they are completely absent after short treatment (48 h) (data not shown). Control experiments performed omitting primary antibody gave negative results (data not shown).

3 Immunostaining of E-cadherin (a, b) and β-catenin (c, d) in HepG2 cells after 12 days of culture in the absence (a, c) or presence (b, d) of RA. After long-term RA treatment E-cadherin and β-catenin are both localized on the cell membrane and the scattered cytoplasmic staining of controls disappears. Note the lack of β-catenin nuclear immunoreactivity in the treated cells (d). There was no labeling when the first antibody was omitted (not shown). Bars =10 μm. (See Color Plate III).

III CELL COMMUNICATION & ADHESION VOLUME 11, NUMBER 1. COLOR PLATE III. See C. Ara et al., Figure 3.

Concerning the lack of scattered cytoplasmic immunoreactivity of β-catenin in treated cells, we wondered whether treatment with RA could also promote the degradation of this protein through the ubiquitin-proteasome system. To verify this hypothesis, we performed immunoprecipitation and Western blot and analyzed the ubiquitination level of β-catenin and its grade of phosphorylation on serine. The assay was performed after the longest interval of RA treatment (12 days), when the effects of RA were much more evident. It can easily be seen that the treatment does not affect either the ubiquitination level of β-catenin (Figure 4A), or its grade of phosphorylation on serine (Figure 4B). Figure 4C shows the β -catenin load that was used to determine ubiquitination and phosphorylation levels.

4 Western blot of ubiquitin (A) and phosphoserine (B) after immunoprecipitation with anti-β-catenin in HepG2 cells cultured for 12 days in the absence (C) or presence of retinoic acid (RA). Figure C shows the β-catenin load. Densitometric evaluation of the bands, expressed as percentages of controls, does not show any change. Results are the average of at least four different experiments ±S.D.

Concerning the tight junction system, the amounts of occludin and ZO-1 proteins were also evaluated by Western blot. In the case of occludin, the level remains unchanged after RA treatment (Figure 5A), whereas the amount of ZO-1 appears decreased, reaching 40% of the control value after 12 days of treatment (Figure 5B). When cellular extracts were immunoprecipitated with anti-ZO-1 antibody and visualized with anti-occludin, we found that a greater amount of protein complex was present in the treated cells (Figure 6A) and that ZO-1 showed a decreased level of tyrosine phosphorylation (Figure 6B). It is well established that increased phosphorylation on tyrosine destabilizes the tight junctions, changes the localization of junctional components and negatively affects paracellular permeability (10). We, therefore, hypothesize that the effect of RA on the phosphorylation of ZO-1 may contribute to the correct localization of the junctional proteins occludin and ZO-1, leading to optimal reconstitution of the tight junctions. This is also suggested by the immunofluorescence images. Figure 7 shows that both proteins localize on the plasma membrane of treated cells (b, d), which also appear closely joined, while in the controls the immunoreactivity is not localized only on the membrane, but is also scattered on the cytoplasm (a, c). No specific labeling was present when the first antibody was omitted (data not shown).

5 Western blot analysis of occludin (A) and ZO-1 (B) in HepG2 cells cultured for 2, 7, and 12 days in the absence (C) or presence of retinoic acid (RA). Densitometric evaluation of the bands, expressed as percentages of controls at two days of culture, shows a decrease in ZO-1 after RA treatment, particularly evident after 12 days (60%). Results are the average of at least four different experiments ±S.D. ^*P < 0.01.

6 Western blot analysis of occludin (A) and phosphotyrosine (B) after immunoprecipitation with anti-ZO-1 in HepG2 cells cultured for 2, 7, and 12 days in the absence (C) or presence of retinoic acid (RA). Densitometric evaluation of the bands, expressed as percentage of control after two days of culture, shows an increase of about 45% after 7 and 12 days of treatment in the amount of ZO-1 bound to occludin (A) and a decrease of about 35% after 7 days and of about 50% after 12 days of RA treatment in the amount of tyrosine-phosphorylated ZO-1 (B). Results are the average of at least four different experiments ±SD. ^*P < 0.01.

7 Immunostaining of occludin (a, b) and ZO-1 (c, d) in HepG2 cells after 12 days of culture in the absence (a, c) or presence (b, d) of RA. After long-term RA treatment occludin and ZO-1 are both localized in the cell membrane and the scattered cytoplasmic staining of controls disappears. There was no labeling when the first antibody was omitted (not shown). Bars = 10 μm. (See Color Plate IV).

IV CELL COMMUNICATION & ADHESION VOLUME 11, NUMBER 1. COLOR PLATE IV. See C. Ara et al., Figure 7.

DISCUSSION

This report provides evidence that treatment with RA is able to modulate intercellular adhesion systems, that is, adherens and tight junctions, in the human hepatoma cell line HepG2.

In studying the adherens junctions, the most significant result we obtained is undoubtedly the increase and stabilization in the cell membrane of the β-catenin/E-cadherin complex, induced by RA, that seems to exert its effect essentially by reducing tyrosine-phosphorylated β-catenin. This mechanism appears to be in accordance with the literature, in which for several experimental models it is reported that the phosphorylation grade on tyrosine of β-catenin is inversely correlated with its grade of association with E-cadherin and with its proper localization at the level of the junctional complexes (10).

A different, very relevant result of the present study consists of the disappearance of β-catenin immunoreactivity from both the nucleus and the cytoplasm of treated cells, as seen in our immunofluorescence images. In HepG2 control cells, which are transformed cells, it can be assumed that the presence of nuclear and free scattered cytoplasmic β-catenin contributes to the mechanisms responsible for neoplastic transformation. A pivotal role in hepatocarcinogenesis seems to be played by modifications of the β-catenin molecule, as described in many in vivo systems as well as in various cell lines (13). In HepG2 cells, the modification consists of a large deletion that removes the potential GSK-3β regulatory site and probably the binding site for α-catenin, impairing both the degradative ubiquitin-proteasome pathway and the binding of the plasma membrane to the cytoskeleton (12). These alterations are thought to represent a major event in the development of hepatocarcinoma, and are similar to those observed in colon cancer and in melanoma cell lines (29, 38). The lack of degradation could explain the maintenance of cytoplasmic immunoreactivity in control cells, while the decreased binding to α-catenin could justify the reduced immunoreactivity at the level of the plasma membrane observed in our experiments. However, the main mechanism responsible for cell transformation is probably represented by the signaling activity exerted by β-catenin after its migration to the nucleus and its association with transcription factors of the LEF/TCF family, which in turn gives rise to the transcription of mitogenic, antiapoptotic and migratory factors (35, 53). A redistribution of β-catenin from the membrane to the cytoplasm, together with an altered interaction of this protein with the cytoskeleton, is often described in tumors of the stomach, pancreas, esophagus, and bladder. Retinoids are able to counteract these events, carrying on its antitumoral effects (14, 41). In fact, after RA treatment, the nuclear and cytoplasmic immunoreactivity disappears, while a strong β-catenin/E-cadherin signal is observed at the plasma membrane level. Because neither the total amount of β-catenin nor its ubiquitination level show variations after treatment, it is plausible that the cytoplasmic disappearance of β-catenin and its different localization could be the result of a decreased level of phosphorylation on tyrosine induced by RA, which, by promoting its association with E-cadherin at the level of cell membrane, may sustain the accumulation of β-catenin in this cell compartment.

Hyperphosphorylation of β-catenin and, consequently, a destabilization of the β-catenin/E-cadherin complex in the cell membrane, is frequently observed in vivo in hepatic carcinomas and chronic hepatitis or cirrhosis (21). There is ample evidence that the activation of the LEF/TCF pathway by nuclear β-catenin can control the expression of cyclin D1; in particular, a direct relationship between the amount of β-catenin and an over expression of cyclin D1 has been demonstrated in cells of mammary carcinoma (43). A positive relationship with the expression of c-myc has also been observed in several kinds of tumors (26). Data from our laboratory show that treatment of HepG2 cells with RA does not affect the expression of cyclin D1, but is instead responsible for a marked decrease in c-myc (1).

As regards the disappearance of nuclear immunoreactivity observed after RA treatment, a possible mechanism could consist of a transrepression action exerted on the nuclear complex β-catenin /LEF-TCF by RA receptors, following activation by the ligand binding. In fact, very recently a transrepression activity of mitogenic signaling by nuclear receptors, including those for retinoids as well as for vitamin D and androgen hormones, has been demonstrated in several transformed cell lines (42). The effect does not influence the general amount of β-catenin and is not due to a direct interaction, but it is a consequence of the competition of the system with transcriptional coactivators; sometimes, it is also accompanied by an improved expression of E-cadherin. This mechanism could also explain the decreased proliferative activity previously observed after RA treatment (18), which would depend on the inhibition of the mitogenic signals activated by β-catenin/ LEF-TCF.

Similar results were obtained with the tight junction molecules. The level of phosphorylation on tyrosine of ZO-1 decreases following RA treatment, and the disappearance of ZO-1 cytoplasmic staining is paralleled by a coincident localization of this protein with occludin in the cell-cell contact areas. A possible explanation of this effect can be consistent with an induction of phosphatase activity and/or a down regulation of protein kinases mediated by RA, acting therefore at posttranslational level. This is not surprising, as in previous experiments we observed that the treatment of both normal and transformed cultured hepatocytes with RA regulates the three-dimensional organization of the cytoskeleton and induces a recovery of the cells' polyhedric shape and polarity by modulating the level of phosporylation of cytokeratin intermediate filaments (18, 48).

However, in contrast with the observations made on β-catenin, the total amount of ZO-1, measured by Western blot, appears clearly decreased. It is not known, at our knowledge, whether the cytoplasmic pool of free ZO-1, to which a signaling function was ascribed (23), could be modulated by the ubiquitin-proteasome system. As concerns our data on RA treated cells, it can be supposed that the decreased amount of ZO-1 and its disappearance from the cytoplasm could also be dependent on proteolytic degradation. To examine this possibility, experiments are in progress to evaluate the effect of RA in the presence of proteasome inhibitors, such as the lactacystine.

The colocalization of occludin and ZO-1 in the tight junctions is particularly relevant and it could correlate with the grade of cell differentiation, as already shown in other cell systems (15, 23). These findings indicate that RA also supports the reconstitution of the tight junctions and is of fundamental relevance in the reestablishment of the polarity of the fully differentiated quiescent hepatocyte. They are consistent with data from the literature that show a clear relationship between altered cell growth and the expression of proteins homologous to ZO-1 (46) and are thus in line with an anti-tumoral effect of RA. In addition, the specific targets of RA appear to be the constituents of the cytoplasmic plaque (i.e., β-catenin and ZO-1) rather than the transmembrane components (i.e., E-cadherin and occludin) of the tight junctions. This could suggest that the proper assembly of the transmembrane components is directed by β-catenin and ZO-1 and not vice versa. As for the catenin-cadherin system, the action of RA is exerted also in this case by maintaining low levels of ZO-1 phosphorylation, which is instead increased in all proliferating systems with consequent alteration in tight junction functionality (10).

RA can positively modulate other junctional systems, including gap junctions. It has been demonstrated that treatment of HepG2 cells with RA leads to an increase in the amount and phosphorylation of connexins, which also appear better localized on cell membranes and, hence, to enhanced gap junctional intercellular communication (5). It must be borne in mind that there is a certain interdependence in hepatocytes among the elements of gap and tight junctions, especially during their formation; connexins, for example, may play a role in modulating occludin expression (20). The protein ZO-1, with its well characterized cytoskeleton binding domain (19), can also be considered as a cross linker between the β-catenin/E-cadherin complex and the actin of cytoskeleton (28, 30).

All these considerations are important in evaluating the pleiotropic effects of RA on HepG2 cells and allow us to postulate that they may all work together towards the same final objective: the reversion of a transformed phenotype and the expression of a higher state of differentiation. Our results with hepatoma cells are in accordance with others from the literature obtained with different cell models. It has been demonstrated that RA exerts its differentiative effect in carcinoma cells (41) and in particular in thyroid carcinoma cell lines, allowing the reexpression of thyroidal markers as well as the improvement of cell-cell and cell-matrix adhesion. These last findings in vitro have made possible to conceive new therapeutic treatments that have already proved effective in pilot clinical trials (40).

We believe that the results presented in this work may make a contribution to the general knowledge of RA effects and help in the designing of therapeutic or preventive strategies based on retinoids also for the treatment of hepatic tumors.

ACKNOWLEDGEMENTS

This work was supported by a grant from the University of L'Aquila for young investigators.