Reduction of malignant phenotype of HEPG2 cell is associated with ...

Download PDF
17 downloads 0 Views 485KB Size Report
Comment
E-cadherin cell adhesion complex. E-cadherin expression following Cx26 transfection was induced. Cx26 expression simultaneously brought E-cadherin and ...
Carcinogenesis vol.22 no.10 pp.1593–1600, 2001

Reduction of malignant phenotype of HEPG2 cell is associated with the expression of connexin 26 but not connexin 32

Tomohiro Yano1, Francisco-Javier Hernandez-Blazquez1, Yasufumi Omori1 and Hiroshi Yamasaki1,2,3 1Unit

of Multistage Carcinogenesis, International Agency for Research on Cancer, Lyon, France and 2School of Science, Kwansei Gakuin University, Hyogo, Japan

3To

whom correspondence should be addressed at School of Science, Kwansei Gakuin University, 1-1-155, Nishinomiya, Hyogo 662-8501, Japan Email: [email protected]

Connexin (Cx) genes have a negative growth effect on tumour cells with certain specificity. However, it is not clear whether each Cx gene can act similarly in growth control. Hepatocytes normally express Cx26 and Cx32 as their major gap junction genes, but HepG2 cells, a hepatoma cell line, are deficient in gap junctional intercellular communication (GJIC) based on the down-regulation of Cx26 and aberrant localization of Cx32. In this study, we showed that some of the expressed Cx26 protein in HepG2 cells localized in the plasma membrane and contributed to recovery of GJIC, while the Cx32 protein remained localized in the cytoplasm. The Cx26-transfected clones showed a significantly slower growth in vivo as well as in vitro and reduced anchorage-independent growth ability compared with a mock-transfected clone. Cx26-transfected cells had more regular cell layers due to the re-establishment of the E-cadherin cell adhesion complex. E-cadherin expression following Cx26 transfection was induced. Cx26 expression simultaneously brought E-cadherin and β-catenin proteins into the plasma membrane without any change in the expression level of β-catenin protein. These results suggest that the expression of Cx26 contributes to negative growth control of HepG2 cells and the morphological change through the induction of E-cadherin and subsequent formation of cell adhesion complex. Introduction Among the different types of cell–cell interaction, gap junctional intercellular communication (GJIC) is considered to be the only route allowing direct transfer of small cytoplasmic hydrophilic metabolites (Mr ⬍ 1000) between cells to maintain cellular homeostasis (1–3). The gap junction is made up of juxtaposed transmembrane hemichannels (connexons) provided by adjacent cells, and each connexon consists of six connexin (Cx) protein subunits (4). So far, cDNAs from at least 15 different Cx species have been cloned in mammals (5). Different combinations of connexins are expressed in different tissues with temporal specificity during development or tissue differentiation (6). Such rigid regulation of connexin expression may contribute to cell differentiation and cell Abbreviations: AS-ODN, antisense phosphorothioate oligodeoxynucleotide; Cx, connexin; DMEM, Dulbecco’s modified Eagle medium; GJ, gap junction; GJIC, gap junctional intercellular communication; HCC, human hepatocellular carcinoma; RT–PCR, reverse transcriptase–polymerase chain reaction; S-ODN, sense phosphorothioate oligodeoxynucleotide. © Oxford University Press

growth in multicellular organisms by keeping important signals, such as those involved in growth control, at equilibrium among GJIC-connected cells. Several lines of evidence support the hypothesis that aberrant GJIC is involved in carcinogenesis. For instance, aberrant GJIC observed in various kinds of tumour cells and the inhibition of GJIC by some tumour-promoting agents provide evidence for the role of disrupted GJIC in carcinogenesis and therefore in cell growth control (7). More direct evidence for the role of GJIC in tumour-suppression has been obtained by transfection of connexin genes into non-communicating tumour cells. Cx43 shows a tumour-suppressive effect in rat C6 glioma cells, chemically transformed mouse fibroblasts and human mammary carcinoma cells (8–10). Similarly, we and others have demonstrated that expression of Cx26 down-regulates the tumorigenicity of HeLa cells and human mammary carcinoma cells (11,12). Further studies have suggested that not all connexin species are able to exert a tumour-suppressive effect on a given tumour, but rather that there seems to be a connexin–cell type compatibility for this effect. For example, Cx26, but not Cx40 or Cx43, exerts control over HeLa cell growth. Cervical tissue, which was the original source of HeLa cells, expresses Cx26 as a major connexin (11). Similarly, Cx43, but not Cx32, has a growth control effect on C6 glioma cells in vivo; astrocytes express Cx43, but not Cx32 (9). These results suggest that connexins may exert growth control only in tissues or cell types in which those particular connexins are naturally expressed. In the liver, Cx26 and Cx32 are considered to be the major components of gap junctions (13). Differential expression and regulation of Cx26 and Cx32 during rat liver carcinogenesis have been reported. Thus, Sakamoto et al. (14) demonstrated that the decrease in Cx32 expression occurs earlier, whereas reduction in Cx26 expression occurs later in association with promotion and progression of carcinogenesis. This means that Cx26 and Cx32 may have different roles in the processes of liver carcinogenesis. In human hepatocellular carcinoma (HCC), levels of Cx32 gene and protein expression were similar to those in surrounding non-cancerous tissue, but GJIC capacity was reduced (15). A previous immunohistochemical study found no expression of Cx26 at all in HCC (16). A cell line established from a human hepatoma, HepG2, has expression of Cx32 but not Cx26 (unpublished data). However, this cell line shows several malignant phenotypes such as anchorage-independent cell growth (17). Taken together, these results reinforce the idea that Cx32 and Cx26 have different roles in hepatocarcinogenesis. In order to test this hypothesis, we examined whether Cx26 expression can reverse the malignant phenotypes of HepG2 cells. We established some clones of HepG2 cells having stable expression of wild-type Cx26 and found that Cx26 transfection reverses the malignant phenotypes of HepG2 cells and that the restoration of Cx26 functions, but not that of Cx32, is important to control their growth. We also provide evidence 1593

T.Yano et al.

that E-cadherin junctions are restored by Cx26, which explains the restoration of epithelial morphology of these cells after Cx26 transfection. Materials and methods Construct, transfection and cell culture The rat Cx26 cDNA insert containing the entire coding region was excised from pGEM-blue vector with EcoRI (18) and subcloned into the expression vector pcDNA3 under the control of the CMV promoter (Invitrogen, San Diego, CA). The sequence of construct pcDNA3-Cx26 was confirmed by DNA sequencing. HepG2 cells (2⫻105) were transfected with 5 µg rat pcDNA3-Cx26 as well as empty pcDNA3 vector, as a control, using TransIT Transfection Reagents (Pan Vera, Madison, WI). HepG2 cells were grown in Dulbecco’s modified Eagle medium (DMEM) supplemented with 2 mM glutamine, 0.1 mM non-essential amino acids, 1 mM sodium pyruvate, penicillin–streptomycin (Gibco, Paisley, UK) and 10% FCS (Dutcher, Brumath, France). After transfection, cells were grown in culture medium, and then transfectants were selected by growth in the same medium supplemented with 0.5 mg/ml hygromycin (Sigma, St Louis, MO). Individual transfected clones were prepared by limiting dilution cloning in 96-well plates. All the established clones were cultured in the same medium at 37°C under a humidified 5% CO2 atmosphere and were routinely passaged by trypsinization with a change of medium twice a week. Growth rate and saturation density Cells (5⫻104) were seeded on a 60 mm culture dish with DMEM culture medium at day 0 of the experiment. Cells from three dishes were trypsinized and counted every 2 days. Dead cells, as determined by trypan blue staining, were left out of the count. The cell population doubling time was estimated by analysis of cell counts obtained in the first week of cell proliferation during the log phase of cell replication. Saturation density was determined by counting cell numbers at day 12 of the experiment. Cells were seeded in triplicate in all experiments. Dye transfer assay to measure GJIC To measure GJIC of the cells, 5% neurobiotin (Vector Laboratories, Burlingame, CA) and 0.4% rhodamine dextran (Molecular Probes, Eugene, Oregon) in phosphate-buffered saline (PBS) was microinjected into a single cell, as previously described (19). Rhodamine dextran was used to identify the injected cells. After microinjection, the cells were washed with fresh medium and incubated for 15 min. The cells were then washed with PBS, fixed with 4% paraformaldehyde and permeabilized with 2% bovine serum albumin and 0.25% Triton X-100 in PBS. After washing with PBS, the cells were incubated with streptavidin–FITC conjugate (Sigma) and 0.25% Triton X-100 in PBS. Cells were washed with PBS, and subsequent dye transfer was quantified by counting the number of fluorescent cells surrounding an injected cell under an Olympus IMT-2 phase contrast and fluorescence microscope (Olympus, Tokyo, Japan) (20). The average number of communicating cells was determined from a minimum of 20 injections. Anchorage-independent cell growth assay This test was carried out in soft agar as described previously (17). Briefly, 1⫻104 cells from each clone were seeded in 4 ml 2⫻ concentrated complete DMEM containing 0.3% agar on a solidified (0.5% agar) basal layer (5 ml) in 60 mm dishes. Two weeks after seeding, the colonies containing at least 20 cells were counted in triplicate plates. Tumorigenicity assay in nude mice Suspension of 3⫻106 cells of each clone in 0.2 ml PBS were injected subcutaneously into the backs of three athymic nude mice (IFFA CREDO, l’Arbesle, France). After injection, each mouse was observed individually and tumour growth was estimated by direct measurement with callipers. RT–PCR Total RNA was isolated from confluent cultures according to an established method with TRIzol reagent (Gibco) (21). Reverse transcriptase–polymerase chain reaction (RT–PCR) to detect the expression of Cx26 mRNA was performed with ProSTARTM HF Single-Tube RT–PCR System (Stratagene, La Jolla, CA) using sense primer, 5⬘-TCTGGCTCACTGTCCTCTTC-3⬘ (nucleotides 334–354, NCBI gene reference number 57545); anti-sense primer, 5⬘-ATGATCAGCTGCAGAGCCCA-3⬘ (nucleotide 503–523). Except for using 500 ng total RNA, other reaction mixtures were prepared according to the manufacturer’s directions. RT–PCR was carried out on a Robocycler PCR System (Stratagene). Conditions for RT–PCR were equilibration at 37°C for 15 min, followed by an initial denaturation at 95°C for 1 min, 30 cycles of 95°C for 1 min, 60°C for 1 min, 68°C for 2 min and a final extension of 68°C for 5 min. For semi-quantitative RT–PCR (22), 2 µg total RNA were

1594

transcribed to cDNA with an RNA PCR kit (TakaraBiomedicals, Verviers, Belgium) using an anti-sense primer for the E-cadherin gene together with the anti-sense primer for the control gene encoding the β-actin gene. Each gene was then amplified together with the internal control β-actin gene in a single tube. Amplification of cDNA was performed on a Perkin-Elmer (Foster City, CA) Gene Amp PCR System, and the cycling program for E-cadherin and β-actin gene included a 1 min denaturation step at 95°C and a 2 min annealing step at 57°C, followed by a 3 min elongation step at 72°C for 25 cycles. Anti-sense primers were 5⬘-CTTGGCTGAGGATGGTGTA-3⬘ (nucleotides 1016–998, NCBI 31072) for E-cadherin and 5⬘-GACGATGCCGTGCTCGATG-3⬘ (nucleotides 270–251, NCBI 28251) for β-actin, and sense primers were 5⬘-GGTTATTCCTCCCATCAGCT-3⬘ (nucleotide 562–580) for E-cadherin and 5⬘-GATCCGCCGCCCGTCCACA-3⬘ (nucleotide 6–24) for β-actin. The RT–PCR products were electrophoresed on a 1.5% agarose gel and visualized after ethidium bromide staining of the gel. A 100 base pair ladder (Amersham, Little Chalfont, UK) was used as a marker for sizing RT–PCR products. Immunoblotting analysis As described previously (23), 20 µg total protein extract from each clone was loaded onto an 8% SDS–polyacrylamide gel for detection of E-cadherin and β-catenin, or a 15% gel for Cx26 and Cx32. After electrophoresis, proteins were transferred to PVDF membranes (Bio-Rad, Hercules, CA). The blots were incubated with mouse monoclonal anti-Cx26, anti-Cx32 (1/1000; Zymed, San Francisco, CA), E-cadherin (1/1000; Transduction Laboratory, Lexington, KY) and β-catenin (1/500; Transduction Laboratory) antibodies. Each immunoreactive band was detected using the ECL system (Amersham). A horseradish peroxidase-conjugated rabbit anti-mouse IgG (Amersham) was diluted 1:5000 to detect the above antibodies. Molecular sizing was done using the Rainbow molecular weight marker (Amersham). Indirect immunofluorescence staining Cells were seeded on a tissue culture chamber (Lab-Tek; Nunc, Naperville, IL), washed with PBS and fixed in pure acetone for 5 min at –20°C. The fixed cells were incubated with 5% hydrogen peroxide in PBS for 30 min at room temperature. All subsequent procedures were carried out using Tyramide Signal Amplification kit (NEN, Boston, MA). The cells were then washed, soaked in 0.5% blocking agent in PBS and treated with a first set of monoclonal antibodies (Cx26, 1/1000; Cx32, 1/500; E-cadherin, 1/20000; β-catenin, 1/10000; Transduction Laboratory) at 4°C overnight. Incubations with a secondary biotinylated antibody against mouse IgG (1/1000) for 1 h and then with a streptavidin–peroxidase conjugate (1/100) were performed for 30 min at room temperature. Subsequently, the cells were reacted with fluorescein tyramide for 5 min at room temperature. After the cells were washed with PBS three times and mounted using a mounting medium Vectashield (Vector Laboratories), they were examined using a fluorescence/phase contrast Olympus Vannox T microscope (Olympus, Tokyo, Japan). AS-ODN treatment Cx26 antisense phosphorothioate oligodeoxynucleotide (AS-ODN) (5⬘-TGCAGTGTGCCCCAATCC AT-3⬘) was purchased from Biognostik (Gottingen, Germany). The sequence encompassed the start codon of rat Cx26 mRNA. As a control for the non-specific effects of oligonucleotide treatment, the corresponding sense phosphorothioate ODN (S-ODN) (5⬘-ATGGATTGGGGCACACTACA-3⬘) was also obtained from BIOGNOSTICK. We exposed cells to 5 µM AS-ODN or S-ODN and they were added to the culture medium at 2 day intervals. When treatment for ⬎5 days was necessary, the culture medium containing AS-ODN or S-ODN was changed at 4 day intervals. Under these conditions, the expressed Cx26 protein was almost diminished, but the S-ODN treatment at the equivalent dose did not affect the protein level. Statistical analysis Statistical analysis was performed by one-way analysis of variance followed by Duncan’s multiple-range test. Values of P 艋 0.05 were considered significant.

Results Isolation of Cx26-transfected clones After 3 weeks of selection with 0.5 mg/ml hygromycin, we obtained 10 independent clones from Cx26-transfected cell cultures. Of the transfectants, two clones of Cx26 HepG2 transfectants (Cx26-2 and Cx26-3) expressed a similar level of Cx26 protein as well as mRNA (Figure 1A and B). In another clone of the transfectants (Cx26-1), we could not detect expression of Cx26 protein but there was a weak band

Growth suppression of HepG2 cells by Cx26

Fig. 1. (A) RT–PCR analysis of Cx26 expression, western blot analysis of (B) Cx26 and (C) Cx32 levels in HepG2 cells transfected with pcDNA vector only (Vector) and pcDNA vector containing wild-type Cx26 gene (Cx26-1, Cx26-2 and Cx26-3). The results shown are representative of three independent experiments.

Fig. 3. GJIC capacity of mock- and Cx26-transfected clones. GJIC capacity was estimated by the dye (neurobiotin) transfer assay. Values are expressed as means from 20 determinations and each bar indicates SEM. *Significantly different from mock transfectant and Cx26-1 clones. **Significantly different from mock transfectant.

Fig. 2. Cell morphology of (A) mock-transfected, (B) Cx26-1, (C) Cx26-2 and (D) Cx26-3 clones. Original magnification, ⫻200.

for an RT–PCR product equal in size to Cx26 mRNA (Figure 1A and B). However, the intensity of this band in Cx26-1 was much lower than those in the other two clones. On the other hand, no band was observed for Cx26 in seven other transfectant clones nor in five mock-transfected clones. A typical representative clone of these mock-transfectants is shown in Figure 1A and B. With respect to Cx32, the levels of protein in HepG2 cell clones were almost the same before and after Cx26 transfection and not related to the level of Cx26 expression (Figure 1C). Morphologically, the cells of five independent mock-transfected clones grew as clusters that formed multilayered colonies (Figure 2A), whereas the Cx26transfected cells (Cx26-2 and 3) grew as a monolayer of polygonal cells (Figure 2C and D). However, the morphological shape of Cx26-1, which expressed a very low level of Cx26, was similar to that of the mock-transfected clone (Figures 2B). A previous report has suggested that a similar morphological change in HepG2 cells to that caused by Cx26 transfection is an indicator of reversal of malignant phenotypes (16). Thus, the present results suggest that Cx expression may reverse the malignant phenotypes of HepG2 cells. GJIC capacity of transfectants In order to examine the GJIC capacity of each clone, the dye transfer assay was performed by microinjecting neurobiotin

Fig. 4. Localization of Cx26 and Cx32 in HepG2 cells before and after Cx26 transfection. Mock-transfected clone was incubated with anti-Cx26 and anti-Cx32, and Cx26-transfected clone (Cx26-3) was incubated with anti-Cx26 and anti-Cx32. Subsequently, fluorescence microscopic analysis was carried out. Original magnification, ⫻1000.

into the clones (Figure 3). The GJIC capacity of clones Cx262 and Cx26-3 was much higher (statistically significant) than that of clone Cx26-1 and the mock transfectant. The GJIC capacity in clone Cx26-1 was moderately elevated compared with that of the mock transfectant, and the difference was statistically significant. These results suggest that the level of Cx26 protein expression in the HepG2 cell clones is related to their GJIC capacity. Localization of Cx26 and Cx32 In order to check whether the expression of Cx26 and/or Cx32 proteins contributes to the recovery of GJIC in HepG2 cells after Cx26 transfection, the localization of Cx26 and 32 proteins was examined immunohistochemically. As shown in Figure 4A and B, immunohistochemical data showed a very 1595

T.Yano et al.

Fig. 6. In vitro anchorage-independent growth capacity of mock- and Cx26transfected clones. Values are expressed as means from three determinations, and each bar indicates SEM. *Significantly different from mock transfectant and Cx26-1 clones.

Fig. 5. (A) Population doubling time and (B) saturation density of mockand Cx26-transfected clones. Values are expressed as means from three determinations, and each bar indicates SEM. *Significantly different from mock transfectant and Cx26-1 clones.

similar localization of Cx32 protein before and after Cx26 transfection, mostly in the cytoplasm. On the other hand, while the signal for Cx26 protein was not detected in the mock transfectant (Figure 4C), some of the transfected Cx26 was localized in the plasma membrane (Figure 4D). These results indicate that the recovery of GJIC in Cx26-expressing HepG2 cell clones is based on the formation of Cx26-dependent gap junction (GJ). This conclusion is further supported by the fact that we could not detect any difference in the permeability to lucifer yellow between the mock-transfected and Cx26transfected clones (data not shown). Since it has been reported that Cx26-coupled HeLa cells showed about 9-fold lower permeability to lucifer yellow than those coupled by Cx32 (24), it seems that the recovery of GJIC, as detected by neurobiotin in Cx26 transfected-HepG2 cells, is mainly due to the formation of Cx26-dependent GJ. 1596

Cell growth in vitro The cell growth rate and saturation density in vitro of each clone were examined. As shown in Figure 5A, the population doubling time in clones Cx26-2 and -3 increased with respect to the mock-transfected clone, whereas for clone Cx26-1 it remained almost the same. In addition, the saturation density showed the same tendency (Figure 5B). There was a statistically significant difference between the mock transfectant and Cx262/Cx26-3 clones, and between Cx26-1 and Cx26-2/Cx26-3 clones in the cell growth rate and saturation density. Anchorage-independent growth capacity of transfectants In order to examine whether the Cx26 gene alters the anchorage-independent growth capacity of HepG2 cells, we compared their ability to grow in soft agar (Figure 6). Clones Cx26-2 and -3 had much lower (statistically significant) ability to grow in soft agar compared with the mock-transfected clone. On the other hand, the ability of clone Cx26-1 was similar to that of the mock-transfected clone. This anchorage-independent growth capacity was also correlated with the level of Cx26 protein in each clone. Overall, the level of Cx26 expression was closely linked with the negative growth control of HepG2 cells by the Cx transfection, indicating that the effect is exerted by Cx26 and not due to simple clonal variations. Tumorigenicity in vivo Cells of each established clone were injected into the backs of nude mice after trypsinization and the appearance and growth of tumours were checked. The mice injected with mock-transfected and Cx26-1 clones developed detectable tumours on day 11 after injection and the growth rate of tumours were almost similar (Figure 7). On the other hand, the first appearance of tumours in Cx26-2 and Cx26-3 transfected clones was delayed to day 30 after injection and the growth rate in the clones was slower than that in mock-transfected and Cx26-1 clones (Figure 7). These results indicate that Cx26

Growth suppression of HepG2 cells by Cx26

Fig. 7. In vivo tumorigenicity assay of transfected clones in nude mice. Tumour size was estimated at the indicated days after subcutaneous injection of 3⫻106 cells. Each clone was injected into three mice.

Fig. 9. Localization of E-cadherin and β-catenin in HepG2 cells before and after Cx26 transfection. Mock-transfected clone was incubated with (A) anti-E-cadherin and (C) anti-β-catenin, and Cx26-transfected clone (Cx26-3) was incubated with (B) anti-E-cadherin and (D) anti-β-catenin. Subsequently, fluorescence microscopic analysis was carried out. Original magnification, ⫻1000.

Table I. Effect of antisense phosphorothioate oligodeoxynucleotide (AS-ODN) treatment on population doubling time and saturation density of the Cx26-transfected clone, Cx26-3 Group

Population doubling time (h)

A B C D

25.13 23.30 19.98 13.41

⫾ ⫾ ⫾ ⫾

0.28 0.22 0.30 0.15*

Saturation density (⫻105 cells/cm2) 0.450 0.486 0.713 0.781

⫾ ⫾ ⫾ ⫾

0.048 0.052 0.092* 0.075*

Cx26-3 cells were seeded in 60 mm Petri dishes with or without phosphorothioate oligodeoxynucleotide and the cell growth was measured as described in Materials and methods. A, Cx26-transfected cells; B, antisense phosphorothioate oligodeoxynucleotide (S-ODN)-treated cells; C, AS-ODN-treated cells; D, mock-transfected cells. Values are expressed as means ⫾ SEM from three determinations. *Significantly different from A and B.

Fig. 8. (A) Semi-quantitative RT–PCR analysis of E-cadherin expression, western blot analysis of (B) E-cadherin and (C) β-catenin levels of mockand Cx26-transfected clones. The results shown are representative of three independent experiments.

expression is required for negative growth control of HepG2 cells in vivo. Localization and level of E-cadherin and β-catenin It has been reported that intercellular adhesion mediated by the E-cadherin–catenin system plays a role in morphological changes in non-malignant and malignant hepatic diseases (25). We therefore considered the formation of normal adherens junctions as a possible mechanism by which the morphological changes occurred in HepG2 cells after Cx26 transfection. In this context, we examined whether Cx26 transfection could induce changes in E-cadherin and β-catenin, which are components of adherens junctions. RT–PCR and western blot analysis for E-cadherin showed an increase in E-cadherin protein after Cx26 transfection, and it appeared that this increase was due to induction at the transcriptional level (Figure 8A and B). In addition, several mock-transfected clones having no expression

of Cx26 showed quite low levels of E-cadherin compared with that of Cx26-tranfected clones (data not shown). Furthermore, immunostaining results indicated that E-cadherin was expressed in the Cx26-transfected cells and localized in the membrane (Figure 9B), while no expression of E-cadherin was seen in the mock-transfected cells (Figure 9A). While HepG2 cells express both wild-type and deletion mutant β-catenins, the level of expression of both proteins was almost the same in mock and Cx26 transfectants (Figure 9C). However, the localization of β-catenin changed from the plasma membrane in the Cx26-transfected clones (Figure 9D) to the cytoplasm in the mock-transfected clone (Figure 9C). These results suggest that the induction of E-cadherin by Cx26 expression mobilizes pre-existing β-catenin protein to form E-cadherin junctions and that this may be associated with the morphological change observed after Cx26 transfection. Antisense oligonucleotide treatment In order to see that the observed phenotypic changes are causally related to Cx26 expression, we investigated the effect of AS-ODN treatment on some phenotypes of the Cx26 transfectant, Cx26-3. As shown in Table I, AS-ODN treatment 1597

T.Yano et al.

reduced population doubling time and increased saturation density, whereas they were unaffected by S-ODN treatment. Morphologically, AS-ODN treatment reversed the phenotype of the Cx26 transfectant to that of a mock transfectant, whereas S-ODN treatment had no influence on the phenotype (data not shown). Discussion While normal hepatocytes express Cx26 and Cx32 (13), HepG2 cells are tumorigenic, GJIC-deficient and express Cx32 but not Cx26. Therefore, we hypothesized that recovery of Cx26 expression might reverse the malignant phenotypes of HepG2 cells. As reported above, the expression of Cx26 indeed exerted in vivo as well as in vitro negative growth control in HepG2 cells, that was associated with the recovery of GJIC based on the formation of Cx26-dependent gap junctions. In mammals, 15 different connexins are known to exist and various combinations of them are expressed in different tissues and/or cell types, suggesting that each connexin has a cell type-specific role to maintain the normal growth of the cells (26). We have already reported that expression of Cx26, but not of other Cx species, contributes to negative growth control of HeLa cells and that the Cx26 transcript is strongly expressed in cervix, the tissue from which HeLa cells originated (11). This notion was reinforced by the finding that Cx32 and Cx43 partially acquired a growth-suppressive function in HeLa cells as a result of deletion of their C-terminal tails; such deleted mutants are structurally similar to Cx26 (27). Similarly, glioma cell growth in vivo was suppressed by Cx43, but not by Cx32; astrocytes express Cx43, but not Cx32, again emphasizing cell/connexin compatibility in growth control. With regard to connexin/cell-type compatibility, it might be expected that both Cx26 and Cx32 could exert growth control of hepatomas, since both of these connexins are normally expressed in hepatocytes. However, our results suggest that only Cx26 is able to do so, indicating that there is a preference in connexin species in growth control when multiple connexins are expressed in a given type of cells. Recently, it has been shown that there is a high prevalence of spontaneous and chemically induced liver tumours in Cx32 gene-deficient mice (28,29). We have recently established a transgenic mouse line in which a dominant-negative mutant Cx32 gene is specifically expressed under the albumin promoter (30). In these transgenic mice, there was no increase in spontaneous liver tumours (unpublished results), in contrast with the Cx32 gene-deficient mice. It is important to emphasize that the Cx32 gene-deficient mice had much reduced Cx26 expression (28), whereas our transgenic mice also had no such reduction (31). These results are therefore consistent with the idea that Cx26 has a stronger control growth capacity than Cx32 in hepatocytes even if both of them are expressed. GJ channels composed of different connexins exhibit distinct channel permeability properties (32), indicating that there may be qualitative differences in GJIC controlled by different Cx species. A previous study has shown the existence of three different types of GJ channels in hepatocytes: two homomeric channels of Cx26 and Cx32, and one heteromeric channel composed of both Cx types (33). The existence of heteromeric GJ channels in other cells has also been demonstrated (34,35). In addition, a recent study using a rat hepatocyte culture system has revealed that Cx26 forms heteromeric, rather than homomeric, GJ channels with Cx32 under certain culture 1598

conditions (36). These results suggest that GJIC mediated through the heteromeric GJ channels is important to maintain the normal phenotype of hepatocytes. Thus, it can be expected that heteromeric GJ channels composed of Cx26 and Cx32 are formed in HepG2 cells after Cx26 transfection and contribute to the reversal of the malignant phenotypes of the cells. However, our present data indicate that Cx32 stayed in the cytoplasm and did not form heteromeric GJ channels with Cx26. This suggests that GJIC through heteromeric GJ channels is unnecessary for reversion of the malignant phenotypes of HepG2 cells. In order to rule out the possibility that Cx32 fails to form heteromers with Cx26 or to locate in the plasma membrane after Cx26 transfection due to the presence of mutations in the Cx32 gene, we sequenced the gene in HepG2 cells, but did not detect any mutation (unpublished data). It has been suggested that Cx26 and Cx32 can be transported to the plasma membrane via different pathways in hepatopcytes (37). We thus cannot rule out the possibility that a specific system for transporting Cx32 to the plasma membrane does not work well in HepG2 cells. An interesting finding of this study was that Cx26 transfection reversed the morphological phenotype of HepG2 cells from a multilayered form to a monolayered form. This report is the first to show such a clear morphological change after connexin gene transfection, as far as we know. Previous studies have shown that such a morphological change in HCC depends on the expression level and localization of E-cadherin in the plasma membrane (38,39). Thus, we propose that alterations in the expression level and localization of E-cadherin after Cx26 transfection into HepG2 cells contributes to the induction of the observed morphological change. In fact, we found that Cx26 transfection caused induction of E-cadherin mRNA and localization of the protein in the plasma membrane. In addition, the AS-ODN for Cx26 reversed the Cx26 transfection-induced phenotypic changes. These results support the idea that Cx26 transfection reverses some malignant phenotypes of HepG2 cells partly through induction of E-cadherin. Recent studies have shown that the full adhesive activity of E-cadherin is expressed when it is complexed via a cytoplasmic binding domain with three cytosolic proteins, α-catenin and β-catenin or γ-catenin (40,41). Of these catenins, in addition to a role as a component of adherence junctions, β-catenin has been shown to play an important role in the signal transduction pathways mediated by the proto-oncogenes src and wnt-1 (42). In this wnt signalling pathway, the activation is initiated by binding of wnt factors to members of the frizzled family of transmembrane receptors, leading to activation of dishevelled which blocks the phosphorylation of β-catenin by glycogen synthase-3β. As a result, cytoplasmic β-catenin is stabilized due to resistance to degradation by the ubiquitin/proteasome system and finally induces abnormal growth and transformation of cells through binding of β-catenin to a transcription factor of the T cell factor/lymphoid enhancer factor family (43). In HepG2 cells, β-catenin is mutated so that the potential glycogen synthase-3β regulatory site is removed, and thus is resistant to degradation by the ubiquitin/proteasome system irrespective of the initial activation of the wnt-signalling pathway, finally leading to an increase in the cytoplasmic pool of catenin in the cells (44). Therefore, previous reports concluded that the activation of the signalling pathway in HepG2 cells was caused by mutation of β-catenin (44,45). However, our results suggest another possibility to explain the aberrant localization of β-catenin. In our study, the expression of E-cadherin seems to

Growth suppression of HepG2 cells by Cx26

be linked with the localization of β-catenin in the plasma membrane of HepG2 cells. The mutant of β-catenin still has an intact binding site for E-cadherin, so it can bind and colocalize with E-cadherin in the plasma membrane. Therefore, the localization of E-cadherin in the plasma membrane after Cx26 transfection may induce the localization of wild- and mutant-types of β-catenin in the plasma membrane, leading to a decrease in the cytoplasmic pool of catenin and suppression of β-catenin-dependent signal-induced cell growth. Thus, the expression level of E-cadherin in HepG2 cells is important in regulation of the cytoplasmic pool of β-catenin. Furthermore, overexpression of β-catenin and an increase in the cytoplasmic pool of this protein in normal epithelial cells stimulate anchorage-independent growth (46), so the recovery of anchorage-dependence in HepG2 cells following expression of Cx26 may partly be due to reduction of the free pool of β-catenin through induction of E-cadherin. In a more recent study, it has been demonstrated that, in response to wnt-signalling, the accumulated Cx43 is co-localized with β-catenin in the junctional membrane in heart; moreover, forced expression of Cx43 in cardiomyocytes reduced the transactivation potential of β-catenin (47). This report further supports the possibility that Cx26 expression regulates β-catenin-dependent signalling in HepG2 cells. We and others have previously demonstrated that E-cadherin is required for the establishment of GJIC (48,49). We have recently obtained evidence that E-cadherin is essential for intracellular transport of Cx26 and Cx43 and the formation of GJ in mouse epidermal cells (unpublished data). Our present data demonstrate that expression of Cx26 in HepG2 cells induces the expression of E-cadherin and that part of the expressed Cx26 is localized with E-cadherin in the plasma membrane. In summary, we propose that the formation of Cx26-dependent GJ and cell adhesion driven by E-cadherin in some cells are reciprocally coordinated. It has been suggested that in order to restore the role of GJIC in a malignant phenotype, it is necessary to have co-expression of both Cx26 and E-cadherin in communication-defective cell populations derived from undifferentiated tumours (50). Thus, coordination between Cx26 and E-cadherin may similarly be required to reverse some malignant phenotypes of HepG2 cells. Recent studies have suggested that connexins are also associated with tight junction components (51,52). Further investigation may shed light on the existence of such a coordinated intercellular junction network and on how it may participate in cell growth control. Acknowledgements We are grateful to Mrs Chantal De´ chaux for her secretarial help and Dr John Cheney for editing the manuscript. We are grateful to Dr Marc Mesnil for helpful discussions, and to Miss Nicole Martel and Mrs Colette Piccoli for their technical assistance. This work was partially supported by a grant from the US National Institutes of Health (R01-CA-40534).

References 1. Mesnil,M. and Yamasaki,H. (1993) Cell–cell communication and growth control of normal and cancer cells: evidence and hypothesis. Mol. Carcinog., 7, 14–17. 2. Saez,J.C., Connor,J.A., Spray,D.C. and Bennett,M.V.L. (1989) Hepatocyte gap junctions are permeable to the second messenger, inositol 1,4,5triphosphate and calcium ions. Proc. Natl Acad. Sci. USA, 86, 2708–2712. 3. Simpson,I., Rose,B. and Lowenstein,W.R. (1977) Size limit of molecule permeating the junctional membrane channels. Science, 195, 294–296.

4. Stauffer,K.A., Kumar,N.M., Gilula,N.B. and Unwin,N. (1991) Isolation and purification of gap junction channels. J. Cell Biol., 115, 141–150. 5. Bruzzone,R., White,T.W. and Paul,D.L. (1996) Connections with connexins; the molecular basis of direct intercellular signaling. Eur. J. Biochem., 238, 1–27. 6. Kumar,N.M. and Gilula,N.B. (1996) The gap junction communication channel. Cell, 84, 381–386. 7. Yamasaki,H. and Naus,C.C.G. (1996) Role of connexin genes in growth control. Carcinogenesis, 17, 1199–1213. 8. Zhu,D., Caveney,S., Kidder,G.M. and Naus,C.C.G. (1991) Growth retardation in glioma cells overexpressing a gap junction protein. Proc. Natl Acad. Sci. USA, 88, 1883–1887. 9. Naus,C.C.G., Elisevitch,K., Zhu,D., Belliveau,D.J. and DelMaestro,R.F. (1992) In vivo growth of C6 glioma cells transfected with connexin 43 cDNA. Cancer Res., 52, 4208–4213. 10. Rose,B., Mehta,P.P. and Loewenstein,W.R. (1996) Gap-junction protein gene suppresses tumorigenicity. Carcinogenesis, 14, 1073–1075. 11. Mesnil,M., Krutovskikh,C., Piccoli,C., Elfgang,C., Traub,O., Willecke,K. and Yamaski,H. (1995) Negative growth control of HeLa cells by connexin genes: connexin species specificity. Cancer Res., 55, 629–639. 12. Hirschi,K.K., Xu,C., Tsukamoto,T. and Sanger,R. (1996) Gap junction genes Cx26 and Cx43 individually suppress the cancer phenotype of human mammary carcinoma cells and restore differential potential. Cell Growth Differ., 7, 861–870. 13. Zhang,J.T. and Nicholson,B.J. (1989) Sequence and tissue distribution of a second protein of hepatic gap junctions, Cx26, as deduced from its cDNA. J. Cell Biol., 109, 3391–3401. 14. Sakamoto,H., Oyamada,M., Enomoto,K. and Mori,M. (1992) Differential changes in expression of gap junction proteins connexin 26 and 32 during hepatocarcinogenesis in rats. Jpn. J. Cancer Res., 83, 1210–1215. 15. Krutovskikh,V., Mazzoleni,G., Mironov,N., Omori,Y., Aguelon,A.-M., Mesnil,M., Berger,F., Partensky,C. and Yamasaki,H. (1994) Altered homologous and heterologous gap-junctional intercellular communication in primary human liver tumors associated with aberrant protein localization but not gene mutation of connexin 32. Int. J. Cancer, 56, 87–94. 16. Wilgenbus,K.K., Kirkpatrick,C.J., Knuechel,R., Willecke,K. and Traub,O. (1992) Expression of Cx26, Cx32 and Cx43 gap junction proteins in normal and neoplastic human tissues. Int. J. Cancer, 51, 522–529. 17. Carloni,V., Romanelli,R.G., Mercuri,A.M., Pinzani,M., Laffi,G., Cotrozzi,G. and Gentilini,P. (1998) Knockout of α6β1-integrin expression reverses the transformed phenotype of hepatocarcinoma cells. Gastroenterology, 115, 4330–4342. 18. Zhang,J.T. and Nicholson,B.J. (1989) Sequence and tissue distribution of a second protein of hepatic gap junctions, Cx26, as deduced from rats cDNA J. Cell Biol., 109, 3391–3401. 19. Brissette,J.L., Kumar,N.M., Gilula,N.B., Hall,J.E. and Dotto,G.P. (1994) Switch in gap junction protein expression is associated with selective changes in junctional permeability during keratinocyte differentiation. Proc. Natl Acad. Sci. USA, 91, 6453–6457. 20. Duflot-Dancer,A., Mesnil,M. and Yamasaki,H. (1997) Dominant-negative abrogation of connexin-mediated cell growth control by mutant connexin genes. Oncogene, 15, 2151–2158. 21. Chomczynski,P. and Sacchi,N. (1987) Single-step method of RNA isolation by acid guanidinium thiocyanate–phenol–chloroform extraction. Anal. Biochem., 162, 156–159. 22. Jani,B., Melchoir,C., Gouon,V., Vallar,L. and Kieffer,N. (1999) Autocrine TGFβ-regulated expression of adhesion receptors and integrin-linked kinase in HT-144 melanoma cells correlates with their metastatic phenotype. Int. J. Cancer, 83, 255–262. 23. Omori,Y., Krutovskikh,V., Mironov,N., Tsuda,H. and Yamasaki,H. (1996) Cx32 gene mutation in a chemically induced rat liver tumour. Carcinogenesis, 14, 2077–2080. 24. Cao,F., Eckert,R., Elfgang,C., Nitsche,J.M., Snyder,S.A., Hulser,D.F., Willecke,K. and Nicholson,B.J. (1998) A quantitative of connexin-specific permeability differences of gap junctions expressed in HeLa transfectants and Xenopus oocytes. J. Cell Sci., 111, 31–43. 25. Ihara,A., Koizumi,H., Hashizume,R. and Uchikoshi,T. (1996) Expression of epithelial cadherin and α- and β-catenins in non-tumoral livers and hepatocellular carcinomas. Hepatology, 23, 1441–1447. 26. Dermietzel,R. and Spray,D.C. (1993) Gap junctions in the brain: where, what type, how many and why? Trends Neurosci., 16, 186–192. 27. Omori,Y. and Yamasaki,H. (1999) Gap junction proteins connexin32 and connexin43 partially acquire growth-suppressive function in Hela cells by deletion of their C-terminal tails. Carcinogenesis, 20, 1913–1918. 28. Temme,A., Buchmann,A., Gabriel,H.D., Nelles,E., Schwarz,M. and Willecke,K. (1997) High incidence of spontaneous and chemically induced liver tumors in mice deficient for connexin32. Curr. Biol., 7, 713–716.

1599

T.Yano et al. 29. Moennikes,O., Buchmann,A., Ott,T., Willecke,L. and Schwarz,M. (1999) The effect of connexin32 null mutation on hepatocarcinogenesis in different mouse strains. Carcinogenesis, 20, 1379–1382. 30. Omori,Y., Duflot-Dancer,A., Mesnil,M. and Yamasaki,H. (1998) Role of connexin (gap junction) genes in cell growth control: approach with sitedirected mutagenesis and dominant-negative effects. Toxicol. Lett., 96–97, 105–110. 31. Zaidan Dagli,M.-L., Omori,Y., Krutovskikh,V. and Yamasaki,H. (1999) Reduced GJIC capacity and delayed initiation of liver regeneration in liver-specific transgenic mice with a dominant-negative mutant of connexin32. Proc. Annu. Meet. Am. Assoc. Cancer Res., 40, 462–463. 32. Veenstra,R.D., Wang,H.-Z., Beblo,D.A., Chilton,M.G., Harris,A.L., Beyer,E.C. and Brink,P.R. (1995) Selectivity of connexin-specific gap junctions does not correlate with channel conductance. Circ. Res., 77, 1156–1165. 33. Barrio,L.C., Suchyna,T., Bargiello,T., Xu,L.X., Roginski,R.S., Bennett,M.V.L. and Nicholson,B.J. (1991) Gap junctions formed by connexins 26 and 32 alone and in combination are differently affected by applied voltage. Proc. Natl Acad. Sci. USA, 88, 8410–8414. 34. Stauffer,K.A. (1996) The gap junction proteins β1-connexin (connexin32) and β2-connexin (connexin26) can form heteromeric hemichannels. J. Biol. Chem., 270, 6768–6772. 35. Jiang,J.X. and Goodenough,D.A. (1996) Heteromeric connexons in lens gap junction channels. Proc. Natl Acad. Sci. USA, 93, 1287–1291. 36. Lee,M.-J. and Rhee,S.K. (1998) Heteromeric gap junction channels in rat hepatocytes in which the expression of connexin26 is induced. Mol. Cell, 8, 295–300. 37. Diez,J.A., Ahmad,S. and Evans,W.H. (1998) Biogenesis of liver gap junctions. In Werner,R. (ed.), Gap Junctions. IOS Press, New York, pp. 130–134. 38. Shimoyama,Y. and Hirohashi,S. (1991) Cadherin intercellular adhesion molecule in hepatocellular carcinoma: loss of E-cadherin expression in an undifferentiated carcinoma. Cancer Lett., 57, 131–135. 39. Nuruki,K., Toyoyama,H., Ueno,S., Hamanoue,M., Tanabe,G., Aikou,T. and Ozawa,M. (1998) E-cadherin expression but not N-cadherin expression is correlated with the intracellular distribution of catenins in human hepatocellular carcinomas. Oncol. Rep., 5, 1109–1114. 40. Ozawa,M., Ringwald,M. and Kemler,R. (1990) Uvomorulin–catenin complex formation is regulated by a specific domain in the cytoplasmic region of the cell adhesion molecule. Proc. Natl Acad. Sci. USA, 87, 4246–4250. 41. Kemler,R. (1993) From cadherins to catenins: cytoplasmic protein interactions and regulation of cell adhesion. Trends Genet., 9, 317–321.

1600

42. Hinck,L., Nathk,I.S., Papkoff,J. and Nelson,W.J. (1994) β-Catenin: a common target for the regulation of cell adhesion by WNT-1 and Src signaling pathway. Trends Biochem. Sci., 19, 538–542. 43. Barth,A., Nathke,I.S. and Nelson,W.J. (1997) Cadherins, catenins and APC proteins: interplay between cytoskeletal complexes and signal pathways. Curr. Opin. Cell. Biol., 9, 638–642. 44. Coste,A.D.L., Romagnolo,B., Billuart,P., Renard,C.-A., Buendia,M.-A., Soubrane,O., Fabre,M., Chelly,J., Beldjord,C., Kahn,A. and Perret,C. (1998) Somatic mutation of the β-catenin gene are frequent in mouse and human hepatocellular carcinomas. Proc. Natl Acad. Sci. USA, 95, 8847–8851. 45. Korinek,V., Barker,N., Morin,P.J., van Wichen,D., de Weger,R., Kinzler,K.W., Vogelstein,B. and Clevers,H. (1997) Constitutive transcriptional activted by a β-catenin–Tcf complex in APC–/– colon carcinomas. Science, 275, 1784–1787. 46. Orford,K., Orford,C.C. and Byers,S.W. (1999) Exogenous expression of β-catenin regulates contact inhibition, anchorage-independent growth, anoikis and radiation-induced cell cycle arrest. J. Cell Biol., 146, 855–867. 47. Ai,Z., Fischer,A., Spray,D.C., Brown,A.M.C. and Fishman,G.I. (2000) Wnt-1 regulation of connexin43 in cardiac myocytes. J. Clin. Invest., 105, 161–171. 48. Jongen,W.M.F., Fitzgerald,D.J., Asamoto,M., Piccoli,C., Slaga,J.J., Gros,D., Takeichi,M. and Yamasaki,H. (1991) Regulation of connexin 43mediated gap junctional intercellular communication by Ca2⫹ in mouse epidermal cells is controlled by E-cadherin. J. Cell Biol., 114, 545–555. 49. Meyer,R.A., Laird,D.W., Revel,J.-P. and Johnson,R.G. (1992) Inhibition of gap junction and adherens junction assembly by connexin and A-CAM antibodies. J. Cell Biol., 119, 179–189. 50. Terzaghi-Howe,M., Chang,G.-W. and Popp,D. (1997) Emergence of undifferentiated rat tracheal cell carcinomas, but not squamous cell carcinomas, is associated with a loss of expression of E-cadherin and of gap junction communication. Carcinogenesis, 18, 2043–2050. 51. Toyohuku,T., Yabuki,M., Otsu,K., Kuzuya,T., Hori,M. and Tada,M. (1998) Direct association of the gap junction protein connexin 43 with ZO-1 in cardiac myocytes. J. Biol. Chem., 273, 12725–12731. 52. Kojima,T., Sawada,N., Chiba,H., Kokai,Y., Yamamoto,M., Urban,M., Lee,G.-H., Hertzberg,E.L., Mochizuki,Y. and Spray,D.C. (1999) Induction of tight junctions in human connexin 32 (hCx32)-transfected mouse heptocytes: connexin 32 interacts with occludin. Biochem. Biophys. Res. Commun., 266, 222–229. Received January 29, 2001; revised May 2, 2001; accepted May 23, 2001