hepatocyte growth factor (HGF) is a well-described growth factor for various tissues. It exhibits proliferative and differentiating effects on the gastric epithelium in vivo or in vitro (19, 39). Its receptors, encoded by the protooncogene c-met, are generally found on cells from epithelial lineages and are highly expressed in the gastric mucosa (39). The existence of a marked effect of HGF on the proliferation and migration of gastric epithelial cells has been demonstrated in primary culture models (34), and in vivo experiments emphasize its major contribution during gastric ulcer healing (29). However, the molecular mechanisms mediating the effects of HGF on events such as epithelial cell dissociation and scattering have been studied mostly in tumor cell lines or cells derived from other tissues, like the renal cell line Madin-Darby canine kidney (MDCK). Such studies have demonstrated that HGF stimulates cell scattering through a pathway involving the activation of phosphatidylinositol 3-kinase (PI 3-kinase) (28) and induces adherens junction disassembly via the activation of both mitogen-activated protein (MAP) kinase and PI 3-kinase, although the activation sequence of these two transduction pathways is controversial (17, 24). HGF was also shown to promote the relocalization of ZO-1 from the tight junction to the cytoplasm in MDCK cells (10). The transduction pathways involved in such an effect of HGF remain unknown. However, there are very few data on the direct role of HGF on intercellular junctions in gastrointestinal epithelium, mostly because of the absence of an appropriate cell line model.

The need for cell lines that mimic the in vivo physiological state of gastrointestinal epithelial cells has become more apparent over the years. However, the establishment of nontumoral epithelial cell lines bearing some or all of the differentiation characteristics of their in vivo counterparts has proven a daunting task, particularly in the case of the gastrointestinal tract. Most studies on the secretory or proliferative functions of gastrointestinal cells have been performed in vivo or on primary cultured cells (1, 3). More recently, studies on tumor-derived cell lines have provided us with useful insights into some of the regulatory mechanisms of the gastrointestinal epithelium and their alteration during tumor development. In recent years, several investigators have reported the establishment of cell lines displaying partially differentiated epithelial phenotypes, such as the human intestinal cell line tsFHI (26) or the fetal rat intestinal line SLC-11 (7). In the case of gastric cells, the best models obtained so far are the mucous cell lines GMS 06, derived from transgenic mice (33), and RGM-1, isolated from the rat gastric mucosa (12).

However, there is still a lack of differentiated nonmucous cell lines. Such models are essential to understand in detail the molecular mechanisms underlying gastric epithelial cell differentiation and proliferation and particularly to assess the individual effects on those events of growth factors such as HGF.

Recently, an interesting model of transgenic mice (ImmortoMice) harboring a thermosensitive version of the large T antigen of SV40 has been developed and used to immortalize intestinal and colonic epithelial cells (40). One of the main advantages of this model is that the expression of the large T antigen can be repressed under appropriate conditions, allowing a better differentiation of the cells. We recently described (13) an autocrine growth regulation by a nonamidated form of gastrin on one of these colonic cell lines.

In this work, to assess the effects of HGF, we have developed a nontumoral gastric epithelial cell line (IMGE-5) obtained from the stomach of these transgenic mice and have used it to study the effects of HGF on the establishment and stability of tight junctions as well as on cell migration.

MATERIALS AND METHODS

Cell isolation and culture.

Cells were prepared from the fundus and corpus of H-2Kb-tsA58 transgenic mice. These mice are transgenic for a interferon-γ (IFN-γ)- and temperature-sensitive mutant of the SV40 large T antigen. The origins and detailed characteristics of this strain were described previously (15). Stomachs from three adult (6 mo old) male mice were excised, everted, and rinsed in PBS for 30 min. Then, after ligation along the fundus/antrum border, the antrum was removed. The remaining pouch was injected with 1 ml of medium B (described in Ref. 22) + 1.6 mg/ml pronase and was sequentially incubated at 37°C in medium B + 1% BSA for 30 min, medium A (22) + 1% BSA for 25 min, and medium B + 1% BSA for 25 min. It was then incubated at 37°C with gentle stirring in RPMI 1640 + 1% BSA + 1 mM dithiothreitol (DTT) for two consecutive 15-min periods. The suspension from these last two incubations was pooled and spun for 5 min at 800 rpm. The pellet was resuspended in 2 ml of RPMI + additives [2 mM glutamine, 100 U/ml penicillin-streptomycin, ITS-A complement (Life Technologies), and 10 U/ml IFN-γ (Sigma)] and then counted. The total number of cells obtained was ∼22 × 10⁶ with a survival rate of 80%.

The cells were next layered onto Nycodenz gradients [according to Prinz et al. (25), except that 4 ml of both 1:1 and 1:2 (vol/vol) mixtures of Nycodenz and medium B + 1% BSA + 0.5 mM DTT stocks were used] and centrifuged for 8 min at 4°C and 150g. After centrifugation, cells from the 2-ml bottom layer of the gradient were collected (density ∼1.080 g/l), washed in PBS, and resuspended in RPMI + additives + 10% FCS.

The cells were plated on 24-well plates (2 × 10⁵cells/well) and grown in RPMI + additives + 5% FCS at 33°C. The medium was renewed every 3 days until cells became confluent. Cells were removed from each well using dispase (collagenase 0.35 mg/ml + pronase 0.3 mg/ml) and replated in three new wells. After four successive passages, cells were cloned in a 96-well plate by limiting dilution. A few colonies were amplified and split with trypsin-EDTA. Cells were kept in RPMI + additives + 5% FCS during the first 10 passages, after which ITS-A was omitted and the IFN-γ concentration was reduced to 1 U/ml.

Because RPMI only contains 0.4 mM calcium, all experiments concerning tight and adherens junctions were performed in DMEM, which contains 2 mM calcium. In this article, the expression “permissive conditions” (proliferative conditions) will refer to a culture at 33°C in the presence of 1 U/ml of IFN-γ and “nonpermissive conditions” (nonproliferative conditions) will be used for cells kept at 39°C without IFN-γ.

Immunofluorescence and confocal microscopy analysis.

Cells (1 × 10⁵) were plated in DMEM + 5% FCS + 1 U/ml IFN-γ on 14-mm glass coverslips in 12-well plates and allowed to grow until confluent at 33°C in 5% CO₂. Half of the cells were then kept in these permissive conditions, and the other half were transferred to nonpermissive conditions for 48 h. Cells were then washed once with PBS and fixed either for 3 min at 4°C with ice-cold methanol (for cytokeratin immunochemistry) or for 15 min in PBS + 1% paraformaldehyde at room temperature (for occludin immunochemistry we used a gentle 0.2% Triton X100 preextraction before paraformaldehyde fixation). Immunofluorescence analysis was then performed with fluorescein isothiocyanate- or rhodamine-coupled secondary antibodies, and coverslips were mounted on slides in Cytifluor (Oxford Instruments). Monoclonal mucin M1 antibodies were kindly provided by Dr. J. Bara, Hôpital St. Antoine, Paris, France. Anti H⁺-K⁺- ATPase antibody 2H6 was a gift of Dr. T. Matsuda, Institute for Immunobiology, Kyoto University, Japan. Cytokeratin 18, ZO-1, and occludin antibodies were from Zymed; E-cadherin and β-catenin antibodies were from Transduction Laboratories; SV40 and Met antibodies were from Santa Cruz Biotech.

Confocal microscopy was performed using a Biorad 1024 CLMS System, including a Nikon Optiphot II upright microscope and an argon-krypton ion laser with two emission lines (488 nm for fluorescein isothiocyanate and 568 nm for rhodamine); ×40 (1.3 NA) and ×60 (1.4 NA) Nikon lenses were used. When a double staining was performed, images were collected sequentially to avoid cross-contamination. Series of horizontal optical sections were collected for each fluorochrome, and each section was assessed for colocalization.

Detection of HGF receptors (c-met) by RT-PCR.

Total RNA was prepared from confluent cells cultured in DMEM + 5% FCS under permissive or nonpermissive conditions according to the method of Chomczynski and Sacchi (4). For reverse transcription experiments, 10 μg of total RNA were mixed with 0.5 μg oligo(dT) primers for 10 min at 70°C and left on ice for 2 min. After addition of 5× first-strand buffer, 10 mM DTT, and 0.5 mM dNTPs, samples were prewarmed for 2 min at 42°C, and 1 ml of Superscript II reverse transcriptase (Life Technologies) was added for an incubation at 42°C for 50 min followed by 15 min at 70°C. PCR was performed in a final volume of 25 μl using dNTPs (0.27 mM each), 1.5 mM MgCl₂, 1× PCR buffer, 0.3 U of Taq polymerase (Promega), c-met primers (1 μM each; sense CTCCTATGTGGATCCTGTAATAAC, antisense CCTGGACCAGC TCTGGATTTAG) and 2 μl of template cDNA. The standard PCR procedure involved 39 cycles with denaturation of samples at 94°C, elongation of the DNA strand at 72°C, and annealing at 54°C, each step being 1 min long. Finally, 10 μl of the samples were run on a 1.2% agarose gel containing ethidium bromide and examined under ultraviolet light.

Proliferation assay.

Three thousand cells were plated in 96-well plates in RPMI + 5% FCS and grown under permissive conditions overnight. Cells were synchronized in Go in FCS-free RPMI for 24 h under nonpermissive conditions and then transferred to RPMI + 1% FCS with or without agents to be tested for a further 24 h, and their proliferation was measured using an XTT assay (Boehringer-Mannheim) according to the manufacturer's instructions.

Soft agar growth assay.

Fifty thousand cells were resuspended in 0.3% low-melting-point (LMP) agarose (Difco) in RPMI supplemented with 20% FCS and plated on 6-cm dishes precoated with 3% LMP agarose in RPMI with 20% FCS. Cultures were either kept permanently at 33°C or transferred after 24 h from 33°C to 39°C and then supplemented every 3 days with 2 ml of RPMI containing 20% FCS and allowed to grow for 3 wk with daily checks for the appearance of colonies.

In vivo tumorigenicity assays.

Cells (5 × 10⁶) were injected subcutaneously into five athymic nude mice (BALB/c) at two sites, and the animals were checked at weekly intervals for a period of at least 12 mo.

Western blotting.

Cells were grown in 100-mm petri dishes under permissive conditions until they reached 90% confluence. They were then transferred to 39°C and serum starved for 24 h, stimulated with 50 ng/ml human recombinant HGF (Sigma-Aldrich, St. Quentin Fallavier, France) for 30 min with or without 30-min preincubation with the PI-3K inhibitor LY-294002 (10 μM; Sigma-Aldrich), and lysed using the standard procedure previously described (27). In the case of ZO-1-occludin association, we immunoprecipitated 100 μg of protein lysate per sample in a Tris-NaCl (pH 7.5) buffer containing 1% Nonidet P40, 100 μM sodium orthovanadate, and 1 mM DTT (WLB buffer), using 2 μg of anti-ZO-1 or anti-occludin antibody for 2 h at 4°C and then adding 100 μl of 20% protein A-Sepharose CL-4B (Pharmacia) overnight. Samples were washed three times in WLB buffer and spun for 10 s at 10,000 g; the pellet was directly resuspended in loading buffer, denatured for 3 min at 95°C, and spun for 30 s at 10,000 g, and proteins in the supernatant (20 μl) were separated onto a 9% SDS-polyacrylamide gel. For detection of Met, occludin, and ZO-1 as well as measurement of phosphorylation, 20 μg of total protein lysates were directly mixed with loading buffer and then denatured and separated onto an 8% SDS polyacrylamide gel. Proteins were then transferred onto a nitrocellulose membrane using a semi-dry blotting system (Bioblock). The membranes were then incubated with the appropriate primary antibodies [polyclonal anti-ZO-1 and anti-occludin (Zymed, San Francisco, CA); PY20 monoclonal anti-phosphotyrosine (Transduction Laboratories, Lexington, KY)], and detection was performed with horseradish peroxidase-coupled protein A (1:3,000) or anti-mouse IgG (1:5,000) followed by enhanced chemiluminescence (Amersham).

Measurement of transepithelial resistance.

Cells were grown to confluence in DMEM + 5% FCS in 24-well plates with a polyester filter (Transwell, 0.4-mm pore size; Costar, Cambridge, MA). They were then shifted to nonpermissive conditions, and the transepithelial resistance (TER) of the IMGE-5 monolayers was assessed every day with a Millicell electrical resistance system from Millipore (Bedford, MA) connected to dual Ag-AgCl electrodes. For HGF experiments, cells were serum starved for 24 h under nonpermissive conditions. Medium was then changed in all wells to DMEM without serum in the top compartment (apical side) and either DMEM without serum alone (control cells) or with 50 ng/ml HGF (stimulated cells) in the bottom compartment (basolateral side). TER was regularly measured for 6 h thereafter and expressed as ohms per square centimeter.

Wound healing experiments.

Wound healing experiments were performed on cells grown on glass coverslips in 12-well plates under permissive conditions until they reached 90% confluence and then shifted to 39°C and serum starved for 24 h. The confluent monolayer was then wounded linearly using a pipette tip, washed three times with PBS, pretreated or not for 30 min with the PI-3K inhibitor LY-294002 (10 μM), and treated with or without HGF 50 ng/ml for the indicated length of time in the presence of 10 μg/ml mitomycin C to block DNA synthesis. For each time point, cells were fixed in PBS + 2% paraformaldehyde for 15 min at room temperature, photographed under a phase-contrast microscope, labeled with an anti-ZO-1 antibody and stained with a fluorescence-coupled secondary antibody as described in Immunofluorescence and confocal microscopy analysis, and photographed again using a fluorescence microscope.

RESULTS

Preparation and culture of IMGE 5 cells.

Mouse gastric epithelial cells are difficult to isolate under conditions suitable for a high rate of survival in culture. The technique of incubating an everted stomach in solutions containing pronase was the most successful in our hands. The isolated cells were separated by Nycodenz gradient centrifugation. The cell line presented here was derived from the bottom layer of these gradients (density of ∼1.080 g/l). These cells were passaged four times with a pronase-collagenase mixture and then cloned by limiting dilution. The three clones showing the clearest epithelium-like morphology were subcloned using the same technique. Of the 16 subclones obtained,clone 5 showed the most epithelium-like morphology and the best expression of cytokeratin 18. This clone, named IMGE-5, has been further characterized and has retained both features after 25 passages. The appearance of IMGE-5 cells under light microscopy (Fig.1, A and B) is influenced by culture conditions; the cells displayed a much more homogenous epithelium-like morphology when kept under nonpermissive conditions (39°C; Fig. 1 A) than when kept under permissive conditions (33°C; Fig. 1 B).

Staining with periodic acid-Schiff reagent and with anti-M1 mucin antibodies suggested an absence of mucus production by IMGE-5 under any culture conditions. Staining of IMGE-5 cells using Gomori's aldehyde fuchsin did not allow detection of gastric chief cells; IMGE-5 cell staining with an anti-H⁺-K⁺- ATPase antibody to detect some parietal cell characteristics also proved negative. Finally, SV40 large T antigen was consistently expressed in the nucleus of IMGE-5 cells under permissive conditions, whereas the expression was decreased from 36 h (not shown) and strongly repressed after 6 days under nonpermissive conditions (Fig. 1, C andD).

Absence of tumorigenicity.

The extent of transformation and potential tumorigenicity of IMGE-5 cells were evaluated by assessing their capacity to form colonies in soft agar and by injecting them into BALB/c nude mice, respectively. No colonies were formed by IMGE-5 cells after 4 wk of culture in soft agar under nonpermissive conditions, and, in fact, most cells were dead within 2 wk under these conditions. Under permissive conditions, some cells grown in soft agar began to proliferate very slowly over 3 wk to form very small colonies, after which they died within 2 wk. Furthermore, IMGE-5 cells did not produce any tumors for up to 12 mo after injection into nude mice.

Expression of epithelial markers.

The expression by IMGE-5 cells of several proteins generally considered as markers for epithelial cells was checked using selective antibodies. Cells were found to strongly express cytokeratin 18 in both permissive and nonpermissive conditions; however, expression of the differentiation markers tested was increased at 39°C in the absence of IFN-γ. After 48 h of confluence, expression of E-cadherin, β-catenin (adherens junction), and occludin (tight junction) was barely detectable under permissive conditions, whereas cells displayed a slight membrane expression of ZO-1. In contrast, all these junction proteins were strongly expressed, almost exclusively at the membrane, 48 h after confluent cells were incubated at 39°C in the absence of IFN-γ (Fig. 2 A). Western blotting experiments confirmed that cytokeratin 18 was highly expressed in IMGE-5 cells irrespective of the culture conditions and showed that junction proteins such as β-catenin and ZO-1 were detectable under permissive conditions but that their expression was increased when shifted to nonpermissive conditions, particularly in the case of ZO-1. This increase was prevented by incubating cells with actinomycin D and cycloheximide during the temperature shift, indicating that it was largely caused by the stimulation of mRNA expression (Fig.2 B).

The subcellular localization of these proteins was then examined by confocal microscopy in confluent IMGE-5 cells grown under nonpermissive conditions. Cells were double-stained using anti-ZO-1 and anti-β-catenin antibodies. Staining by different fluorochrome-coupled secondary antibodies confirmed that both ZO-1 and β-catenin were located along the cellular membrane of IMGE-5 cells. However, ZO-1 was expressed toward the apical part, whereas β-catenin was found along a broad lower part of the lateral membrane (Fig.3) without any colocalization of these two proteins. Similar results were obtained when cells were double-stained with anti-occludin and anti E-cadherin antibodies (not shown). These results are in agreement with the physiological involvement of occludin/ZO-1 and E-cadherin/β-catenin in the formation of tight and adherens junctions, respectively, in epithelial IMGE-5 cells.

Expression of HGF receptors by IMGE-5 cells.

RT-PCR and Western blotting were used to assess the expression of HGF receptor (Met) by IMGE-5 cells. Our results show that Met mRNA (Fig.4 A) and protein (Fig.4 B) were detected in these cells under both permissive and nonpermissive culture conditions.

Regulation of IMGE-5 cell proliferation by growth factors and gastrin.

Under permissive conditions, 45% of IMGE-5 cells incorporated bromodeoxyuridine (BrdU; reflecting entry into the S phase of the cell cycle) even after 24 h in the absence of serum. However, under nonpermissive conditions, only 4% of the cells incorporated BrdU after 24 h in the absence of serum and most of them died within 48 h.

Further proliferation experiments were therefore performed under nonpermissive conditions in the presence of 1% FCS after a synchronization period of 24 h without FCS. Under these conditions, ∼15% of IMGE-5 cells incorporated BrdU and their proliferation (measured with an XTT proliferation assay) was stimulated by a high concentration of FCS (10%), growth factors such as HGF (20 ng/ml), and epidermal growth factor (EGF; 100 ng/ml) (both ∼70% of the maximal FCS stimulation), and one trophic hormone, gastrin (1 nM; 55% of the FCS stimulation). Two other growth factors, acidic (aFGF; 50 ng/ml) and basic (bFGF; 100 ng/ml) fibroblast growth factor, had opposite but not significant effects on the proliferation of IMGE-5 cells (Fig. 5).

Effect of HGF on TER.

The presence of functional tight junctions in confluent IMGE-5 cells was confirmed by measurement of a TER across a monolayer grown on polyester-coated Transwell filters. TER gives an evaluation of the ionic paracellular permeability of the monolayer cell culture. TER was regularly monitored from the moment when cells reached confluence, and it reached a value of 150 ± 15 Ω/cm²(n = 4) after 48 h.

When confluent cells were treated with 50 ng/ml HGF, TER was found to decrease significantly from 2 to 6 h after treatment, stabilizing at 75% of the value measured on control cells (Fig.6). TER was back to control after 20 h even after HGF treatment.

Effect of HGF on ZO-1 localization.

We further investigated the effect of HGF on the localization of the tight junction protein ZO-1 in IMGE-5 cells under nonpermissive conditions. When nearly confluent IMGE-5 cells were treated with 50 ng/ml HGF, the appearance of ZO-1 staining at the plasma membrane was partly prevented, with a slight concomitant increase of cytoplasmic staining. After 24 h in the presence of HGF (50 ng/ml), membrane staining for ZO-1 was decreased and discontinuous, whereas ZO-1 was exclusively expressed along the membrane of untreated cells (Fig.7).

Effect of HGF on phosphorylation of ZO-1 and its association with occludin.

To determine whether the disappearance of ZO-1 from the membrane on HGF treatment was caused by reduced expression or by a specific relocalization of the protein, we next investigated the effect of HGF on ZO-1 at the molecular level. Our results show that a 3-h treatment with HGF (50 ng/ml) did not significantly affect the amount of extractable ZO-1 in IMGE-5 cells (Fig.8 A) but strongly stimulated the tyrosine phosphorylation of ZO-1 (Fig. 8, B andE). This stimulation was abolished by a 30-min preincubation with the PI 3-kinase antagonist LY-294002 (10 μM; Fig. 8,B and E).

Furthermore, the amount of the integral tight junction protein occludin coimmunoprecipitating with ZO-1 was decreased markedly in HGF-treated compared with untreated cells (Fig. 8, C and E). Interestingly, this inhibitory effect of HGF on the association between the two tight junction proteins was also blocked by the PI 3-kinase inhibitor LY-294002 (Fig. 8, C and E). Direct immunoprecipitation of occludin revealed that HGF did not affect the amount of occludin (Fig. 8, D and E).

Effect of HGF on IMGE-5 cell migration.

Finally, we investigated the effects of HGF on the migration of IMGE-5 cells in a wound repair model as described in materials and methods. HGF strongly stimulated the migration of IMGE-5 cells inside the wound area (Fig. 9). Depending on the experiment, the migratory effect was detectable from 5 to 8 h after treatment and maximal at 20–28 h (24 h in Fig.9 A), after which no major change was noted for a further 24 h. Furthermore, the variation of ZO-1 localization in IMGE-5 cells during the course of wound repair was controlled using immunofluorescence. Our results showed that HGF strongly accelerated the disappearance of ZO-1 staining from the cell membrane and induced its progressive relocalization to the cytoplasm and then the nuclear area in a time course similar to that of HGF-induced migration (Fig.9 A). The effect was detected as early as 2–3 h after treatment, and ZO-1 was mainly located in the perinuclear area by 20–24 h in HGF-treated cells located in the wound area, whereas untreated cells still displayed a significant amount of ZO-1 membrane expression at that time.

We also investigated whether a PI 3-kinase-dependent pathway was involved in the effect of HGF on IMGE-5 cell migration. The migrating effect of HGF was reversed by preincubating the cells for 30 min with 10 μM LY-294002 before the addition of HGF, suggesting the implication of a PI 3-kinase-dependent pathway in the HGF wound repair effect (Fig. 9 B). When the experiments were performed in the absence of mitomycin C (not shown), migration was not detected earlier but the wound was virtually invisible by 30–40 h, indicating that the effect of HGF on the repair mechanism involved a proliferative as well as a migratory component (not shown).

DISCUSSION

This paper reports for the first time the establishment of a nontumorigenic gastric cell line (IMGE-5) displaying a regulated expression of epithelial differentiation markers such as adherens and tight junctions and demonstrates the role of HGF in delaying the establishment and promoting the dissociation of tight junctions in gastric epithelial cells.

The phenotype of IMGE-5 cells was in part affected by the culture conditions. The cells homogeneously displayed a typical epithelial appearance and barely expressed the SV40 large T antigen when grown under nonpermissive conditions, whereas under permissive conditions they displayed this phenotype only in rare patches and all cells expressed the large T antigen. A similar pattern of T antigen expression was found in a colon cell line obtained from the same mouse model (40).

Thus growth under nonpermissive conditions generated cells whose phenotype was potentially much closer to that found under physiological conditions than the phenotype of transformed and tumorigenic cell lines from gastric or colorectal carcinomas such as HGT-1 (21) or HT-29 (18). This was corroborated by the lack of colony formation from dispersed IMGE-5 cells in soft agar as well as by the absence of tumor formation up to 12 mo after injection into BALB/c nude mice. Furthermore, the proliferation of these cells in vitro was greatly reduced under nonpermissive conditions, even in the presence of 5% FCS. However, IMGE-5 cells did not express specific markers for mucous, parietal, or chief cells under either permissive or nonpermissive conditions and could therefore originate from less specialized epithelial precursors, commonly present in gastric pits.

Because the lack of serum dependence is generally considered an important cell transformation step, we assessed the ability of IMGE-5 cells to proliferate in the absence of serum. Interestingly, <10% of IMGE-5 cells incorporated BrdU after a 24-h serum starvation under nonpermissive conditions. This result showed that IMGE-5 cells have a low basal growth activity, which permits accurate measurement of the stimulatory effects of growth factors on proliferation and probably indicates that they express little or no autocrine proliferative signal. They could therefore be valuable to study the effect of individual exogenous agents on proliferation because many of the epithelial cell lines commonly used are either embryonic and partly differentiated (6) or come from tumor tissues (31) and hence display some level of autocrine growth, with the apparent exception of the widely used MDCK cell line (9).

Indeed, IMGE-5 cells retained the ability to proliferate in response to some growth factors and hormones when grown under nonpermissive conditions. HGF, one of the most potent stimulants of gastric epithelial proliferation (34), clearly stimulated IMGE-5 cell proliferation to a similar extent as EGF, a known trophic factor in the gastrointestinal epithelium as well as in many other tissues (36, 37). In addition, the gastrointestinal hormone gastrin also increased IMGE-5 cell proliferation. The sensitivity of the IMGE-5 cell line to gastrin is of great interest in light of the multiple roles of this hormone in gastric functions (1). This cell line could therefore prove to be a very useful tool to complete our understanding of the signal transduction coupled to gastrin receptors, which traditionally has been studied in primary cultures (25) or tumoral cell lines (20). The lack of a significant effect of aFGF is consistent with the cells' epithelial phenotype, because this growth factor is generally considered as a stimulator of proliferation for cells of the mesenchymal lineage. aFGF has been reported to stimulate the proliferation of epithelial cells but only at a fetal (5) or cancerous (2) stage. IMGE-5 cell proliferation was inhibited slightly but not significantly by bFGF, which usually displays a stimulatory effect on cell growth, although inhibitory effects have been described on epithelial cells, including MCF-7 human breast cancer cells (8).

We then attempted to evaluate the extent of the differentiation process induced by shifting IMGE-5 cells to nonpermissive conditions (39°C). To do so, we measured under both culture conditions the expression of a panel of proteins generally expressed in differentiated epithelial cells, such as cytokeratin 18, adherens junction proteins (E-cadherin and β-catenin), and tight junction proteins (occludin, ZO-1). We found that IMGE-5 cells constitutively expressed cytokeratin 18, indicating that they are naturally committed toward an epithelial phenotype. Adherens and tight junction proteins were also expressed in small amounts after 48 h of confluence under permissive conditions, but only ZO-1 could be detected in small amounts at the plasma membrane. However, 48 h after confluence was reached under nonpermissive conditions, these proteins were highly expressed along the membrane of IMGE-5 cells. This process was prevented by preincubating the cells with both actinomycin D and cycloheximide, indicating that the expression of junction proteins is regulated at the mRNA level. These results clearly show that shifting the cells to nonpermissive conditions induced (or at least strongly accelerated) a further differentiation process in IMGE-5 cells. Furthermore, a differential expression of adherens versus tight junction proteins was demonstrated by confocal microscopy, strongly suggesting the establishment of functional junctions in IMGE-5 cells.

The stability of tight junctions has also been assessed by measurement of TER, which is generally considered to reflect the ionic paracellular permeability of confluent epithelial cells. When cultured under nonpermissive conditions, IMGE-5 cells were shown to display a stable TER, albeit lower than that measured in other cell lines (11,16). This could be explained either by the relative flatness of IMGE-5 cells or by a putative simpler strand structure of their tight junctions. Moreover, most TER values reported elsewhere are measured on cell lines after 5 days at confluence, whereas this study was performed after only 48 h.

To further improve our understanding of the effect of HGF on gastric epithelial cell differentiation, we investigated its effect on the establishment and stability of intercellular junctions. HGF is a potent activator of growth and differentiation of gastrointestinal epithelial cells in vivo (39). In the gastrointestinal tract, HGF is normally produced by cells from the mesenchymal lineage, whereas its receptors are expressed on epithelial cells (39). Therefore, we first assessed the expression of membrane receptors for HGF in IMGE-5 cells under both culture conditions. Our results showed that IMGE-5 cells constitutively expressed HGF receptors, confirming that they seem already committed to the epithelial lineage under permissive conditions.

We then studied the effect of this growth factor on the TER across a confluent IMGE-5 monolayer. Our results clearly demonstrated that HGF significantly decreased TER when incubated for >2 h on the basolateral side of these cells, reflecting a partial loss of tight junction stability leading to a selective ionic paracellular permeability. Such an effect of HGF has been described in T84 intestinal cells (23), whereas HGF does not seem to affect TER in tracheal epithelial cells (30). Decrease of the TER induced by various factors has been correlated with the subcellular localization and the tyrosine phosphorylation of ZO-1 (32, 35), although TER and ZO-1 membrane localization have been found to be regulated in opposite directions by RhoA (11). Furthermore, HGF has been shown to induce relocalization of ZO-1 from the membrane to the cytoplasm in MDCK cells (10). In this study, we demonstrated that HGF is not only important in regulating mature tight junctions but can also be critical during the establishment of tight junctions because it greatly delayed the targeting of ZO-1 to the cell membrane after shifting confluent IMGE-5 cells to nonpermissive conditions.

We also assessed whether the expression and phosphorylation of ZO-1, as well as its capacity to associate in a protein complex with occludin, were correlated with the effect of HGF on tight junctions. The role of ZO-1 phosphorylation in tight junction regulation was documented previously (32, 35). Furthermore, Jiang et al. (16) showed recently that the HGF-induced reduction of the transendothelial resistance in human vascular endothelial cells was correlated with a decrease in the expression of occludin. Results obtained in this study showed that stimulation with HGF did not induce significant changes in the amount of extractable ZO-1 but stimulated the tyrosine phosphorylation of this protein. In light of the fact that HGF also induced a decrease in TER and a reduction in the membrane localization of ZO-1 in IMGE-5 cells, these results seem to point toward an inhibitory role of ZO-1 tyrosine phosphorylation on tight junction stability. Our results, however, differ markedly from those of van Itallie et al. (14), who reported that EGF-induced stimulation of the tyrosine phosphorylation of ZO-1 was associated with the movement of ZO-1 from a diffuse cytoplasmic location toward the plasma membrane in subconfluent A431 cells (15). This discrepancy could be caused by the differences in the degree of confluence or by the tumoral nature of A431 cells.

Furthermore, we also found that the association of occludin with ZO-1 was decreased markedly by HGF. Interestingly, in our model this inhibition did not seem to be caused by the presence of smaller amounts of occludin, differing from the marked decrease in occludin expression induced by HGF in endothelial cells (16). This might indicate that the regulation of tight junctions by HGF could be, to a certain degree, cell type specific.

Finally, we showed that HGF stimulated the migration of IMGE-5 cells in a wound repair model at the same time as it induced relocalization of ZO-1 from the membrane to the cytoplasm and the nucleus. Scattering effects of HGF have been reported in various models, including in gastrointestinal cells (23, 39). However, we report here for the first time that, as for the increased phosphorylation of ZO-1 and the decrease of its association with occludin, the migratory effect of HGF on gastric epithelial cells was abolished by preincubating the cells with a PI 3-kinase inhibitor. The role of PI 3-kinase also seems pivotal in the effect of HGF on dissociation and induction of scattering in MDCK cells (28), but this pathway does not seem to be the only one mediating the effect of HGF on adherens junction disassembly in the same cells (24). Our results could mean that the same signal transduction pathway could mediate the effects of HGF on both tight junction regulation and epithelial cell migration or that the dissociation of tight junctions is a limiting step in the processes leading to cell scattering, even if migration itself necessitates activation of other pathways. These results are consistent with HGF acting at early stages during the process of intercellular dissociation, which is a prerequisite for 1) epithelium-mesenchyme transition and cell migration during morphogenesis (38), 2) regenerative processes such as wound healing in the gastric mucosa, 3) cell transformation, and 4) tumor invasion (39).

In summary, using a newly developed gastric epithelial cell line, we have shown that both the establishment and the stability of tight junctions are regulated by HGF. These effects were partly correlated with an increase of the tyrosine phosphorylation of ZO-1 and a decrease of its association with occludin and were mediated by activation of PI 3-kinase. This novel cell line, which is responsive to several hormones and growth factors and displays a IFN-γ- and temperature-sensitive functional expression of adherens and tight junctions, should be of continuing value in further studies of the proliferation and differentiation of gastric epithelial cells.

FOOTNOTES

• Present address of R. H. Whitehead: GI Cancer Program, CC-2218 MCN, Vanderbilt Medical Center, Nashville TN 37232-2583.

• Address for reprint requests and other correspondence: F. Hollande, Laboratoire de Biochimie des Membranes, Faculté de Pharmacie, 15 av. C. Flahault, 34060 Montpellier cedex, France. (E-mail: fhollande@ww3.pharma.univ-montp1.fr).

• The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.