Membrane-associated HB-EGF modulates HGF-induced cellular responses in MDCK cells

The maintenance of the structural and functional integrity of epithelia requires highly dynamic cell-cell and cell-matrix interactions. Under steady-state conditions, the assembly and maintenance of intercellular junctions appear to be tightly regulated (Braga, 2002; Mitic and Anderson, 1998). However, during development, tissue repair and tissue metamorphosis, concerted cell-cell dissociation and cell migration occur and appear to be intimately linked, such as is observed with epithelial `scattering' (Tsukamoto and Nigam, 1999). Also, during cell division or inflammation, as well as invasion and metastasis of tumor cells, cell-cell junctions are disassembled and sometimes internalized as cells diminish their contacts and become motile (Bauer et al., 1998; Gumbiner, 1996). Alterations in cell polarity and cell-cell and cell-matrix interactions are also well described in renal epithelial cells in response to acute ischemic injury (Molitoris and Wagner, 1996; Racusen, 1997).

The Madin-Darby canine kidney II (MDCK II) cell line has been utilized as an in vitro model to help define the mechanisms underlying complex epithelial cell movements such as scattering, tubulogenesis and dedifferentiation (Balkovetz, 1998; Stuart et al., 1995), especially in response to hepatocyte growth factor/scatter factor (HGF/SF). HGF/SF is a polypeptide cytokine produced by mesenchymal cells that induces a range of responses in epithelial cells (Matsumoto and Nakamura, 2001; Stella and Comoglio, 1999). In addition to causing proliferation, HGF/SF also induces normal epithelial cells grown on plastic to `scatter' because of increased motility and reduced cell-cell interactions and results in cells growing as a dispersed culture in carcinoma cells, where invasiveness is increased by HGF/SF (Bardelli et al., 1997; Maulik et al., 2002). HGF/SF also induces epithelial cells grown in three-dimensional collagen gels to form branching tubules reminiscent of epithelial ducts in the breast and kidney (Barros et al., 1995; Pollard, 2001; Zhang and Vande Woude, 2003). Interestingly, although exogenous epidermal growth factor (EGF) induces DNA synthesis and cell proliferation in MDCK cells (Hauguel-DeMouzon et al., 1992), it does not induce scattering or branching tubulogenesis (Tanimura et al., 1998), nor does it modulate the response of HGF/SF (Liang and Chen, 2001; Tanimura et al., 1998).

Heparin-binding EGF-like growth factor (HB-EGF) was first identified in the conditioned medium of a macrophage-like cell line, U937 (Higashiyama et al., 1991) but is now known to be widely expressed in cells of epithelial, endothelial and mesenchymal origin. As with other members of the EGF family (Iwamoto and Mekada, 2000; Raab and Klagsbrun, 1997), HB-EGF is synthesized as a type-1 transmembrane protein that can be shed enzymatically to release a soluble 14-20 kDa growth factor (Goishi et al., 1995; Higashiyama et al., 1992). However, a considerable amount of proHB-EGF remains on the cell surface under normal physiological conditions (Goishi et al., 1995). The cleavage of proHB-EGF has been shown to occur through metalloproteinase activation, attributed in various cells to ADAM (adisintegrin and metalloproteinase)-9, -10, -12 and/or -17 (Asakura et al., 2002; Izumi et al., 1998; Sunnarborg et al., 2002; Yan et al., 2002), as well as MMP-3 and -7 (Suzuki et al., 1997; Yu et al., 2002), and it can be blocked by metalloproteinase inhibitors such as Batimastat (BB-94) (Chen et al., 2002; Prenzel et al., 1999). Both the transmembrane (proHB-EGF) form of the molecule, as well as the soluble form, are suggested to be active signaling molecules (Iwamoto and Mekada, 2000; Raab and Klagsbrun, 1997). HB-EGF serves as a ligand for EGFR1 (erbB1) as well as erbB4 (Elenius et al., 1997; Higashiyama et al., 1992; Riese et al., 1996). The soluble form of HB-EGF can interact with its receptors in an autocrine and/or paracrine manner, is a potent mitogen and induces proliferation and migration in a variety of cells (Faull et al., 2001; Piepkorn et al., 1998; Tokumaru et al., 2000b). The transmembrane form of HB-EGF interacts with other integral membrane proteins, especially integrins, via the tetraspanin, CD9/DRAP27 (Lagaudriere-Gesbert et al., 1997; Nakamura et al., 1995). It is noteworthy that the HB-EGF-CD9-integrin complex and EGFR are both targeted to cell-cell junctions in some epithelial cells (Fukuyama and Shimizu, 1991; Nakamura et al., 1995). HB-EGF also complexes with heparin sulfate proteoglycans (Paria et al., 1999).

Targeting of the HB-EGF complex and EGFR to immediately juxtaposed cellular regions has been proposed to facilitate juxtacrine signaling (Higashiyama et al., 1995). During the process of organ development, juxtacrine signaling may be an efficient method to deliver selective signals to adjacent cells. Ectodomain shedding of membrane proteins changes their fate, localization and mode of action, thus affecting the biological activities of membrane proteins (Davis et al., 1994; Massague and Pandiella, 1993; Werb and Yan, 1998).

ProHB-EGF may serve a variety of functions depending on its cellular context and has been reported to be mitogenic and cytoprotective in some cells while growth inhibitory and pro-apoptotic in others (Iwamoto et al., 1999; Nishida et al., 2000; Takemura et al., 1997). ProHB-EGF has also been suggested to promote cell-cell adhesion (Kimber, 2000; Takemura et al., 1997). The fact that a considerable amount of proHB-EGF is uncleaved and remains membrane-bound under normal physiological conditions further complicates the analysis of juxtacrine signaling versus tightly coupled autocrine/paracrine signaling due to soluble HB-EGF. The present study was undertaken to investigate the respective roles of soluble and membrane-associated HB-EGF in epithelial cell morphogenesis. We utilized MDCK II cells stably expressing proHB-EGF, a noncleavable membrane-anchored mutant form of proHB-EGF or soluble HB-EGF. We show that the expression of soluble or membrane-anchored HB-EGF molecules results in distinct cell phenotypes, with major differences in both cell-extracellular matrix and cell-cell interactions. Furthermore, we show that expression of membrane-bound HB-EGF inhibits HGF/SF-induced cell scattering and formation of branched tubular structures. This is the first report to describe distinct roles of soluble and membrane-anchored HB-EGF and show that an EGFR ligand can modulate HGF/SF-induced scattering and tubulogenesis in MDCK II cells.

MDCK type II cells were grown in modified Dulbecco Eagle”s medium (DMEM) containing Earle's balanced salt solution supplemented with 5% fetal bovine serum, 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.25 mg/ml amphotericin B in 5% CO₂/95% air at 37°C. To develop MDCK cells stably expressing HB-EGF mutant constructs, cells (1×10⁵ cells/35 mm dish) were transfected with 1 μg of pcDNA3-neo containing the HB-EGF mutant cDNAs, using LipofectAMINE Reagent (Gibco BRL, Life Technologies, Grand Island, NY). The cells were cultured for five passages in DMEM containing G418 (1 mg/ml; Sigma Chemical Co., St Louis, MO), and colonies derived from single cells were picked to generate clones stably expressing the molecule of interest. Cell lines expressing the HB-EGF mutants were identified by immunoblotting with the anti-FLAG (M2) and anti-HB-EGF antibodies. At least two independent stably transfected clones with comparable protein expression of the transfected molecule for each mutant were maintained and used for experiments. The data presented throughout the manuscript are from clone# 1 of the cells transfected with HB-EGF mutants and are representative of the findings from two separate clones unless stated otherwise.

The rabbit anti-rat HB-EGF polyclonal antibody was a generous gift from Li Feng (Department of Immunology, The Scripps Research Institute, La Jolla, CA). The mouse anti-FLAG (M2) antibody, heparinase I and the HGF/SF were purchased from Sigma Chemical Co. The anti-EGFR and anti-P-EGFR antibodies were purchased from Upstate Biotechnology (Lake Placid, NY), the anti-phospho-c-Met antibody was purchased from Biosource International (Camarillo, CA), and the anti-ZO-1 and anti-E-cadherin antibodies were purchased from Zymed Laboratories (South San Francisco, CA) and BD Biosciences Transduction Laboratories, respectively. The anti-CD9 antibody was purchased from abcam.com. The P-ERK 1/2, ERK (extracellular signal-regulated kinase), P-JNK and JNK (Jun N-terminal kinase) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and the P-Akt and Akt antibodies were obtained from Cell Signaling (Beverly, MA).

MDCK II cells were plated on transwell filters (12 mm, 0.4 μm pore size; Corning Costar, Cambridge, MA) at confluence (∼2×10⁵ cells) and allowed to attach overnight to form a confluent monolayer in normal Ca²⁺-containing medium (NC). Transepithelial resistance (TER) was measured after 12 hours of plating (day 1) and every 24 hours thereafter using a Millipore (Bedford, MA) electrical resistance measurement system, and the results are expressed in Ω/cm².

Confluent vector and HB-EGF transfected cells were washed twice with PBS and rendered quiescent by exposure to serum-free culture medium for 48 hours. Conditioned media (CM) were centrifuged and soluble HB-EGF was purified as described by Chen et al. (Chen et al., 2002). Equal amount of eluants were subjected to SDS-PAGE and immunoblotted with anti-HB-EGF antibody.

Rat-tail collagen I (Collaborative Research, Bedford, MA) was diluted in PBS [(–) without Ca⁺⁺ and Mg⁺⁺] according to the manufacturer's directions, and plates were coated as described by Kanai et al. (Kanai et al., 1999). In each assay, six wells were used for each experimental condition. Three wells received a suspension of 1×10⁵ viable cells, while the other three were used for background measurements. The plated cells were then incubated covered at 37°C/5% CO₂ for 60 minutes, the time necessary for MDCK to adhere maximally to collagen I (data not shown). Plates were then gently washed twice with 200 μl/well of PBS (+), and fixed with 100 μl/well of 3.7% formaldehyde for 30 minutes at room temperature. The plates were then stained for 20 minutes with 100 μl/well of 0.1% crystal violet filtered through a 0.22 μm filter unit (Millipore), washed four times with 200 μl/well of PBS (+), and air-dried before the dye was dissolved in 20% acetic acid (100 μl/well) for 10-20 minutes. The amount of staining, which is proportional to the number of attached cells, was measured with a 96-well plate reader (Molecular Devices Corporation, Sunnyvale, CA) at 570 nm. The average absorbance of wells without cells was measured and subtracted from the average of wells with cells to yield the absorbance due to the adherent cells only. Each graph represents the data from four independent experiments.

Rat collagen I was prepared as above and 35 mm tissue culture dishes were coated with 2.5, 5 and 10 μg/ml of collagen I overnight at 4°C and blocked with 1% BSA+PBS as above. Dishes were washed with 1×PBS, and 5×10⁴ cells were plated for 4 hours at 37°C/5% CO₂. Thereafter, cells were washed with PBS, fixed with 3.7% formaldehyde and photographed under phase contrast using a Nikon Diaphot TMD inverted microscope.

3× 10³cells were plated in 12-well tissue culture dishes and cells were allowed to grow as discrete colonies. After 48 hours cells were transferred into medium containing 0.2% fetal calf serum and treated with 20 ng/ml HGF/SF. 12-16 hours later, cells were fixed in 3.7% formaldehyde and photographed under phase contrast using a Nikon Diaphot TMD inverted microscope.

For the collagen gel culture, cells were suspended at 5× 10³ cells/ml in a neutralized collagen solution (Collaborative Biomedicals), dispensed into 96-well plates and incubated at 37°C. After the collagen solution had gelled, the complete culture medium with HGF/SF (20 ng/ml) was added and renewed every day. One group of cells did not receive HGF/SF and served as control. The cultures were photographed after 7-9 days of culture under phase contrast using a Nikon Diaphot TMD inverted microscope.

For migration assays, cells were prepared as for the cell-adhesion assay and resuspended in serum-free media. The migration assay was performed using transwell chambers (12 mm, 0.8 μm pore size; Corning Costar, Cambridge, MA) with the lower chambers filled with serum-free media containing rat-tail collagen (10 μg/ml). Each well was plated with 1×10⁴ cells in 500 μl serum-free media. The transwell chamber was placed at 37°C in a humidified 5% CO₂ atmosphere. After 4 hours, cells on the upper surface of the membrane were removed using a cotton swab. The membrane was then fixed with 3.7% formaldehyde for 30 minutes at room temperature and stained with crystal violet (0.1%, Sigma Chemical Co.) for 1 hour. Cells were washed extensively after staining and the cells on the lower surface were counted under a light microscope at high-power magnification (40×). Multiple fields (four to eight) were counted for each well and the mean values were obtained.

Cells were washed with 2 M NaCl before washing with PBS to remove any soluble HB-EGF bound to the membrane. Preparation of total cell lysate and immunoprecipitation and immunoblotting was performed as described by Dhawan and Richmond (Dhawan and Richmond, 2002). Immunoprecipitation samples for HB-EGF were resolved on 10% Tricine gels (Invitrogen, Carlsbad, CA) while other western blot analyses were performed using regular SDS-PAGE. Signals were detected using an ECL detection kit (Amersham-Pharmacia Biotech, Piscataway, NJ).

Cells plated on transwell filters or grown on glass coverslips were rinsed in PBS (with Ca⁺⁺, Mg⁺⁺ and 0.2% BSA) and fixed with 3.7% formaldehyde in PBS for 30 minutes on ice. The fixed cells were treated with 0.2%Triton X-100 in PBS for 10 minutes on ice and washed three times with PBS. After soaking in PBS containing 5% normal goat serum and 1% BSA in PBS, the samples were incubated with primary antibodies for 1 hour in a moist chamber at 37°C. Samples were then washed three times with PBS, followed by incubation for 30 minutes with the respective conjugated second antibody (Jackson ImmunoResearch Laboratories, West Grove, PA). The samples were washed with PBS three times, embedded in glycerol-PBS-based mounting medium and examined using a Zeiss-410 confocal microscope at the Vanderbilt Imaging Core facility. The image analysis was performed using Metamorph cell-imaging program (Universal Imaging Corporation, Downingtown, PA).

Graphic data are presented as mean±s.e.m. Statistical analysis was performed, where appropriate, using the Student t-test, or one-way analysis of variance (ANOVA), followed by the Tukey's Multiple Comparison Test. Differences with P<0.05 were considered statistically significant.

The cDNA encoding the open reading frame for the native proHB-EGF molecule was amplified from rat kidney total RNA using PCR primers based on the rat HB-EGF sequence (Acc. # L05489.1). The resultant PCR product was subcloned into PCR II and was used as a template for generation of HB-EGF mutant cDNAs. To generate a cDNA corresponding to soluble HB-EGF, PCR primers were designed to amplify amino acids 1-148 of the rat proHB-EGF, while amino acids PVENP (aa 149-153), the putative cleavage site or the complete juxtamembrane domain (aa 149-160) (Hirata et al., 2001; Suzuki et al., 1997) were deleted to generate the 5 aa and 12 aa deletion mutant constructs (Fig. 1). To differentiate between effects mediated by the membrane-anchored and soluble forms of HB-EGF, we developed stably transfected renal epithelial cells (MDCK II cells) expressing soluble HB-EGF (aa 1-148; MDCK^sol), proHB-EGF (aa 1-208; MDCK^pro) and deletion mutant constructs of proHB-EGF lacking the amino terminal 5 (MDCK^5aa) and 12 amino acids (MDCK^12aa) of the juxtamembrane extracellular domain (Fig. 1). One group of cells was transfected with empty vector (MDCK^con) and served as control. To the HB-EGF constructs we added a COOH terminus FLAG epitope, which allowed immunoprecipitation of membrane-anchored HB-EGF with anti-FLAG antibody M2 and immunolocalization with anti-HB-EGF antibodies. Immunoblotting using a polyclonal anti-rat HB-EGF antibody showed similar expression levels of transfected molecules, whereas sHB-EGF was not detected in the total cell lysate (Fig. 2A) but instead was secreted in the medium (Fig. 2B).

Membrane-associated HB-EGF (proHB-EGF) is partially cleaved in response to metalloproteinase activation under normal physiological conditions (Suzuki et al., 1997). To determine whether deletion of the putative cleavage site inhibited the normal metalloproteinase-dependent cleavage of the membrane-anchored HB-EGF mutants, conditioned media from transfected and control cells were purified using heparin-sepharose columns. Equal volumes of eluants were subjected to SDS-PAGE and immunoblotted with an anti-HB-EGF antibody. sHB-EGF expression was detected in the conditioned media from MDCK^sol and MDCK^pro cells, but no sHB-EGF was detected in media from MDCK^5aa or MDCK^12aa cells, suggesting lack of cleavage (Fig. 2B).

To determine whether the secreted HB-EGF in the conditioned medium was biologically active, we examined EGF receptor activation in response to the addition of equal amounts of eluants from the conditioned media to quiescent A431 cells, a cell line known to overexpress EGF receptor (Yoshimoto and Imoto, 2002). After 5 minutes, cells were washed and lysed, and equal amounts of protein were subjected to immunoblotting using a phosphospecific anti-EGFR antibody. As indicated in Fig. 2C, conditioned media from MDCK^sol cells induced significant EGFR tyrosine phosphorylation, and detectable EGFR activation was also seen with the conditioned media from MDCK^pro cells, whereas media from MDCK^5aa, MDCK^12aa and control cells did not induce detectable tyrosine phosphorylation of the EGF receptor, although equal concentrations of total EGFR were detected (Fig. 2C).

Protein kinase C activation with phorbol myristate acetate ester (PMA) induces cleavage of proHB-EGF, and 1 μM of PMA has been shown to cleave the majority of proHB-EGF in 30 minutes (Goishi et al., 1995; Izumi et al., 1998). To ascertain whether deletion of the putative cleavage site would block cleavage of proHB-EGF, we exposed quiescent control and transfected cells to PMA (1 μM) for 30 minutes. Thereafter, cell lysates were subjected to immunoprecipitation using anti-FLAG M2 antibody and immunoblotted with the same antibody. As shown in Fig. 2D, in MDCK^pro cells, transfected proHB-EGF was recognized as 24-28 kDa bands; in addition, there was a less prominent smaller (approximately 12-15 kDa) FLAG-labeled band, representing residual transmembrane (TM) and cytoplasmic tail regions of HB-EGF after cleavage of the proHB-EGF molecule. Phorbol ester (PMA) treatment induced cleavage of HB-EGF protein in the MDCK^pro cells, as evidenced by the marked decrease in the ∼28 kDa band, representing the full-length HB-EGF protein and a corresponding increase in the 12-15 kDa FLAG-labeled fragment. By contrast, phorbol ester treatment of the cells did not alter the expression of the ∼28 kDa band in the MDCK^5aa or MDCK^12aa cells, and essentially none of the 12-15 kDa FLAG-labeled band could be detected, suggesting lack of cleavage. As expected, in cells expressing soluble HB-EGF, no FLAG-labeled cell-associated product could be detected, as the protein is secreted into the medium (Fig. 2D). We could not detect any endogenous expression of HB-EGF in nontransfected MDCK II cells with our antibodies.

With autocrine/paracrine activation of EGF receptors, ligand-receptor binding initiates receptor internalization and degradation, which results in the downregulation of receptor expression (Carpenter, 2000). In MDCK^sol cells, EGF receptor expression was found to be markedly downregulated, and in MDCK^pro cells EGFR expression was detectably lower than in control cells, whereas in MDCK^5aa and MDCK^12aa cells, EGFR expression appeared to be similar to vector transfected cells (Fig. 2E). Basal EGF receptor tyrosine phosphorylation in MDCK^pro, MDCK^5aa and MDCK^12aa cells was similar to that seen in control cells, whereas lower basal EGFR tyrosine phosphorylation was observed in MDCK^sol cells (Fig. 2E).

Membrane-anchored HB-EGF has been suggested to signal in a juxtacrine manner (Iwamoto and Mekada, 2000; Raab and Klagsbrun, 1997). To verify that the deletion of the amino acids from the juxtamembrane extracellular domain did not affect the ability of deletion mutants to initiate juxtacrine signalling, we performed a juxtacrine assay (Higashiyama et al., 1995). Subconfluent A431 cells were overlain with formalin-fixed MDCK cells transiently transfected with different HB-EGF mutant constructs. After 10 minutes, the MDCK cells were washed away, membranes were prepared from A431 cells, EGF receptor was immunoprecipitated with an anti-EGFR antibody and an in vitro kinase assay was performed. All mutant constructs induced EGFR phosphorylation, indicating their ability to signal in a juxtacrine manner (Fig. 3A). The cells expressing the 5 amino acid deletion mutant were chosen for further studies.

To determine whether the deletion of amino acids from the juxtamembrane extracellular domain of pro-HB-EGF altered membrane localization, MDCK^con, MDCK^5aa and MDCK^pro were grown on transwell filters and immunofluorescent staining was performed using an anti-rabbit polyclonal anti-FLAG (M2) antibody followed by confocal microscopy. Results from this study confirmed a predominantly lateral membrane localization of both transfected proHB-EGF and 5 amino acid deletion mutant molecules (Fig. 3B,C).

In vitro expression of proHB-EGF promoted cell adhesion in renal tubular epithelial cells (Takemura et al., 1999; Takemura et al., 1997); however, it has not been determined whether the effect is due to membrane-bound HB-EGF or to sHB-EGF released following cleavage of proHB-EGF. Because our stably expressed juxtamembrane deletion mutants were not cleaved, we tested the potential role of juxtacrine signaling in cell-matrix interactions. We determined integrin-mediated collagen I-dependent cell adhesion in control and HB-EGF transfected cells using a standard cell-adhesion assay (described in Materials and Methods). The cells showed differences in their attachment at all concentrations of collagen tested; at the concentration of 10 μg/ml of collagen, MDCK^5aa cells showed 25% greater cell adhesion compared with control cells and 37% increased adhesion compared with MDCK^sol, which had lower cell adhesion compared with all other cell types tested. MDCK^pro cells had similar collagen I-dependent cell adhesion to MDCK^con cells (Fig. 4A). Pretreating cells with β₁-integrin blocking antibody completely abolished cell-adhesion in all cells, showing that the cell-adhesion was integrin dependent. There was no difference in β₁-integrin expression among the cell types, as determined by FACS and western blot analysis using anti-β₁-integrin antibody (data not shown), confirming that the differences in adhesion were not due to changes in β₁-integrin expression.

We next determined whether these cells also exhibited differences in their ability to spread on a matrix after attachment. MDCK^5aa cells expressing membrane-anchored HB-EGF showed elaborate cell spreading with extensive cellular extensions at all collagen concentrations; however, the differences were most pronounced at 10 μg/ml of collagen I. By contrast, MDCK^sol cells remained rounded and did not spread within the time period of the assay, whereas MDCK^pro cells exhibited an intermediate phenotype, with a minority of the cells spreading with rudimentary cellular extensions (Fig. 4B).

Because expression of membrane-anchored and soluble HB-EGF affected cell adhesion in distinctly opposite ways and soluble HB-EGF has been implicated as an important mediator of wound healing and cell migration in keratinocytes (Tokumaru et al., 2000a), we performed cell migration assays (described in Materials and Methods). MDCK^sol cells showed significantly higher cell migration compared with control cells, and expression of the 5 amino acid deletion mutant inhibited cell migration compared with control cells (Fig. 4C).

Epithelial cells polarize when grown on permeable filter supports, and TER across the confluent epithelial cell layer is an established measure of the degree of cell-cell contact (Hopkins et al., 2000). Cells were plated on clear-transwells at superconfluent density and TER was measured every 24 hours. MDCK^5aa cells developed significantly higher TER than control cells. By contrast, the MDCK^sol cells developed minimal resistance, whereas MDCK^pro cells showed low resistance initially but achieved significantly higher TER than MDCK^con and MDCK^sol cells by day 4-5 after plating (Fig. 5).

To determine whether release of sHB-EGF due to cleavage of proHB-EGF was responsible for the initial low TER observed in MDCK^pro cells, we used a specific metalloproteinase inhibitor Batimastat (BB-94). BB-94 (5 μM) has been shown to inhibit the cleavage of proHB-EGF completely (Chen et al., 2002; Prenzel et al., 1999). The cells were plated on transwells as above and after 10 hours of plating (the time necessary to obtain a confluent cell layer), BB-94 (5 μM) was added. Addition of the metalloproteinase inhibitor eliminated the initial delay in the increase in TER in MDCK^pro cells, and the TER values and timing were similar to those of MDCK^5aa cells. Addition of BB-94 had no significant effect on TER in any other cell type under study, including MDCK^con cells (Table 1).

MDCK II cells normally scatter in response to HGF/SF. For scattering to occur, cells must dissociate from each other and from the substratum (Balkovetz, 1998; Stella and Comoglio, 1999). Thereafter, cells develop a fibroblast-like appearance and migrate away from each other. Although the exogenous administration of soluble EGF induces cell scattering in certain cells, it neither induces nor inhibits cell scattering in MDCK cells (Liang and Chen, 2001; Tanimura et al., 1998). At 16 hours post-treatment with HGF/SF, the scattering pattern of MDCK^sol and MDCK^pro cells was not different from MDCK^con. By contrast, MDCK^5aa did not dissociate or migrate apart from each other (Fig. 6). Exogenous administration of either recombinant EGF (20 ng/ml) or soluble HB-EGF (20 ng/ml) 30 minutes before HGF/SF administration did not alter HGF/SF-induced scattering in the control cells, nor did it alter the inhibition of scattering observed in MDCK^5aa (data not shown). Furthermore, PD 153035, a specific and stable inhibitor of EGF receptor tyrosine phosphorylation (Bancroft et al., 2002; Pai et al., 2002) (Fig. 7A) failed to inhibit or induce cell scattering in the HGF-treated MDCK^con or MDCK^5aa cells, respectively (Fig. 7B).

Western blot analysis using a specific antibody for c-Met, the HGF/SF receptor, showed identical receptor expression among HB-EGF-transfected and control cells (data not shown). Furthermore, analysis with phosphospecific antibodies for c-Met or EGFR, although confirming the possibility of cross-talk between the c-Met and the EGFR activation (Jo et al., 2000; Scheving et al., 2002), indicated no differences in HGF-induced tyrosine phosphorylation of either receptor in the MDCK^5aa cells compared with control cells (data not shown). We could not detect any ErbB4 protein expression in our cells using multiple antigen-specific antibodies (data not shown).

Because inhibition of metalloproteinase activity in MDCK^pro cells eliminated the initial delay in the increase in TER in MDCK^pro cells such that the TER values and timing were similar to those of MDCK^5aa cells, we determined whether the HGF-induced cell scattering could also be modulated by the inhibition of metalloproteinase activity. Experiments were performed as for cell scattering except that cells were exposed to Batimastat (5, 10 and 20 μM) for 3-5 hours before adding HGF/SF (20 ng/ml). MDCK^con cells scattered in the presence of all the doses of Batimastat (data not shown), but the HGF-induced scattering was markedly inhibited in MDCK^pro cells+Batimastat (20 μM) compared with MDCK^pro cells that received only HGF (Fig. 8A).

Because proHB-EGF is known to complex with CD9 and heparin-sulfate proteoglycans (Lagaudriere-Gesbert et al., 1997; Nakamura et al., 1995), we tested whether abrogating this complex formation could induce cell scattering in MDCK^5aa cells in response to HGF. We used a monoclonal CD9 blocking antibody and heparinase I for this purpose. Cells were pretreated with CD9 blocking antibody (1 μg/ml) or heparinase I (1 U/ml) for 3-5 hours before exposing cells to HGF. As a control for the CD9 antibody, a group of cells was exposed to mouse IgG (1 μg/ml). Although neither CD9 blocking antibody nor heparinase I affected the HGF-induced scattering in MDCK^con cells, they both induced scattering in MDCK^5aa cells in response to HGF compared with the MDCK^5aa cells receiving only HGF (Fig. 8B)

MDCK cells form cysts when cultured in three-dimensional collagen gels, and addition of exogenous HGF/SF induces these cysts to develop into branched tubules (Balkovetz, 1998; Stuart et al., 1995). However, as in the case of cell scattering, EGF administration does not have any effect on tubulogenesis in MDCK II cells (Barros et al., 1995). Because expression of membrane-anchored HB-EGF inhibited HGF/SF-induced cell scattering, we investigated whether it could also modulate HGF/SF-induced tubulogenesis. Experiments were performed as described in Materials and Methods, and cells were cultured for 7-9 days before photographs were taken. Interestingly, in nonHGF/SF-treated cells, all HB-EGF-transfected cells showed significant increases in the size of cysts compared with MDCK^con cells, indicating a possible proliferative role for membrane-anchored as well as soluble HB-EGF in collagen. However, tubular structures induced by HGF/SF treatment were very different among all transfected cells as well as compared with control cells. MDCK^5aa cells formed small irregular masses with many small-spike or needle-shaped outgrowths without any tubular extensions, whereas MDCK^sol cells developed into long tubules with reduced branching compared with MDCK^con. MDCK^pro cells once again showed an intermediate phenotype, with tubules exhibiting wider lumens compared with MDCK^con (Fig. 9).

During cell scattering, the adherent and tight junctions are disrupted/dissociated, and the adherent junction protein E-cadherin and the tight junction protein ZO-1 are both translocated to the cytoplasm (Grisendi et al., 1998; Kamei et al., 1999). During HGF/SF-induced tubulogenesis, epithelial cell polarity is transiently lost and components of cell-cell junctional complexes, including E-cadherin and ZO-1, undergo profound rearrangements (Pollack et al., 1998). Given that MDCK^5aa cells neither scattered nor formed well-defined tubular structures in response to HGF/SF, we determined the changes in E-cadherin and ZO-1 localization in control and MDCK^5aa cells in response to HGF/SF at various time-points. Cells were plated on glass coverslips and 48 hours post-seeding, cells were subjected to HGF/SF treatment as in the scattering studies. Cells were fixed at 4, 8 and 12 hours after the addition of HGF/SF. One group of cells did not receive any HGF/SF and served as a control. Immunofluorescent staining using anti-E-cadherin and anti-ZO-1 antibody showed that HGF/SF treatment induced disruption of adherent junctions by 4 hours post-treatment; at 12 hours after HGF/SF treatment almost all the adherent junctions were lost and E-cadherin was translocated to the cytoplasm in the MDCK^con cells. Disruption of tight junctions, as evidenced by ZO-1 localization, was delayed compared with adherent junctions disruption, but some degree of cytoplasmic translocation of ZO-1 was evident by 4 hours of HGF/SF treatment. At 12 hours post-treatment most of the tight junctions were disrupted, but the departing cells were still connected by individual strands of tight junction. By contrast, although there was a minor amount of cytoplasmic translocation of E-cadherin and ZO-1 in the MDCK^5aa cells, disruption of both adherent and tight junctions was largely prevented, even after 12 hours of HGF/SF treatment (Fig. 10a,b).

HGF/SF treatment of MDCK cells has been shown to result in activation of ERK 1/2 and phosphoinositide 3-kinase/Akt pathways, and specific pharmacological inhibitors of these pathways inhibit cell scattering induced by HGF/SF (Khwaja et al., 1998; Sipeki et al., 1999; Tanimura et al., 1998). Also, in a recent report, decreased JNK/SAPK phosphorylation was implicated in HGF/SF-induced cell scattering in MDCK cells (Paumelle et al., 2000). Therefore, we determined whether differences in the activation of one or more of these signaling pathways could be responsible for the lack of scattering observed in MDCK^5aa cells. The change in the phosphorylation status of Akt has been used as an index of HGF/SF-induced phosphoinositide 3-kinase activation, and inhibition of cell scattering with a specific pharmacological inhibitor (LY294002) of phosphoinositide 3-kinase also inhibits HGF/SF-induced Akt phosphorylation (Khwaja et al., 1998). HGF/SF treatment resulted in the slow but sustained activation of ERK 1/2 and Akt in the quiescent control cells and no significant differences were observed in the transfected cells compared with control cells. The basal activity as well the inducible activity of ERK and Akt were relatively lower in MDCK^sol cells, although the relative degree of overall activation of these proteins in response to HGF in these cells was not different than other cells under the study (Fig. 10). However, notably, the antibody against P-Akt recognized an additional band of relatively higher molecular weight in addition to the ∼65 kDa band for P-Akt in the MDCK^5aa cells, and the phosphorylation patterns of these two bands were different. While the higher molecular band showed transient phosphorylation with the peak at 15 minutes followed by a decrease in the phosphorylation, the ∼65 kDa band showed a slow but sustained phosphorylation similar to control cells. A similar higher molecular weight band but of lesser intensity was also recognized in the MDCK^pro cells (Fig. 11). Pretreatment of the cells with LY294002, a specific inhibitor of the phosphoinositide 3-kinase, resulted in complete inhibition of both HGF-induced bands recognized by the anti-phospho-Akt antibody in the MDCK^5aa cells (data not shown). The identity of this higher molecular weight band remains to be determined. No differences in the phosphorylation of JNK/SAPK were observed at any time-point after the treatment in the control cells or among HB-EGF transfected cells (data not shown).

Among membrane-anchored growth factors, HB-EGF is of special interest because it forms complexes with different proteins, functions as a receptor for Diphtheria toxin, serves as a ligand for two different receptors (erbB1 and erbB4) and is a precursor for a soluble growth factor (Iwamoto and Mekada, 2000; Raab and Klagsbrun, 1997). HB-EGF has been shown to be involved in various physiological, as well as pathological, processes; however, to date no study of the biological effects attributed to HB-EGF has unequivocally differentiated between mediation by the membrane-associated versus the soluble form of the growth factor, in which the processing of the transmembrane form may alter juxtacrine to autocrine/paracrine signaling. In the present study, we have defined distinct roles for membrane-associated HB-EGF and soluble HB-EGF in epithelial cell morphogenesis and behavior, using cultured polarized epithelial cells stably expressing either soluble secreted HB-EGF or a noncleavable membrane-anchored form of HB-EGF.

We used these stably transfected cell lines to show that in MDCK cells: first, deletion of the putative cleavage site of the proHB-EGF inhibited its metalloproteinase-dependent cleavage and did not affect the membrane localization or the ability of the molecule to act in a juxtacrine manner; second, the expression of membrane-anchored HB-EGF enhanced collagen-dependent cell-adhesion and induced tighter cell-cell contact while inhibiting cell migration; third, soluble HB-EGF decreased collagen-dependent cell-adhesion and cell-cell contact and significantly enhanced cell migration; and fourth, the expression of membrane-anchored HB-EGF inhibited HGF/SF-induced cell-cell dissociation and the formation of branched tubular structures.

The observation that expression of a noncleavable form of membrane-associated HB-EGF resulted in a significant increase in collagen I-dependent cell-adhesion supports our earlier observation that overexpression of proHB-EGF enhanced cell attachment in tubular epithelial cells (NRK52-E cells) (Takemura et al., 1997), and indicates that this effect is not cell specific. Differences in cell adhesion were not due to differential expression of β₁-integrins. Our observation that membrane-anchored HB-EGF enhanced cell adhesion is also consistent with the biological effects attributed to membrane-anchored HB-EGF that is induced in the mouse luminal epithelium solely at the site of blastocyst apposition before the attachment reaction (Paria et al., 1999).

Expression of soluble HB-EGF resulted in significantly decreased cell adhesion and increased migration compared with MDCK^5aa cells. Ectodomain shedding of EGF receptor ligands, particularly HB-EGF, is required for keratinocyte migration in cutaneous wound healing (Piepkorn et al., 1998; Tokumaru et al., 2000b). In addition, N-arginine dibasic convertase (NRDc), a metalloendopeptidase of the M16 family, which has been recently identified as a receptor for HB-EGF, induces HB-EGF-dependent cell migration (Nishi et al., 2001). Thus, our data are in agreement with the previous reports suggesting a role for sHB-EGF in wound healing and enhancing cell migration.

HGF/SF-induced cell scattering/motility in MDCK cells involves changes in cell-matrix interactions and disassembly of cell-cell junctions, with adherent junctions disassembled before the disruption of tight junctions (Grisendi et al., 1998; Kamei et al., 1999). Immunofluorescent localization of E-cadherin and ZO-1 indicated persistent E-cadherin at adherent junctions and ZO-1 at the tight junctions in MDCK^5aa cells compared with control cells after exposure to HGF/SF. Compared with MDCK^con or MDCK^sol cells, MDCK^5aa cells also developed significantly higher resistance, suggesting tighter cell-cell contact. Cell-cell junctions and polarity undergo profound although transient changes during the process of tubulogenesis (Pollack et al., 1998), and persistence of tight and adherens junction integrity may underlie the failure of MDCK^5aa cells to develop normal branched structures.

By contrast, MDCK^sol cells developed only minimal TER when grown on permeable membrane supports. In this regard, decreased cell-adhesion, increased motility and lack of resistance due to compromised cell junctions are characteristics of transformed cells (Schoenenberger et al., 1991), and the members of EGF family of growth factors are known to induce epithelial-mesenchymal transformation (Cattan et al., 2001).

Increases in TER were delayed in MDCK^pro cells compared with MDCK^5aa cells, and HGF/SF-induced scattering was not prevented in MDCK^pro cells. We postulate that the initial low resistance in MDCK^pro cells was due to sHB-EGF released by cleavage of proHB-EGF at the time of plating and cell attachment, and the high TER at later points represented a shift to the predominant expression of membrane-associated HB-EGF. This hypothesis was supported by the fact that pretreatment of the MDCK^pro cells with a metalloproteinase inhibitor Batimastat (BB-94) not only selectively eliminated the delayed increase in TER in MDCK^pro cells, such that the timing of increase in TER was similar to MDCK^5aa cells, but also markedly inhibited the HGF-induced scattering in these cells.

In the present study, HGF/SF treatment induced comparable levels of both c-Met and EGFR tyrosine phosphorylation in MDCK^con and MDCK^5aa cells. Furthermore, identical patterns and degrees of activation of both ERK 1/2 and Akt were detected. HGF/SF treatment of MDCK cells is known to induce activation of ERK 1/2 and phosphoinositide 3-kinase, as well as phosphoinositide 3-kinase-dependent Akt activation (Khwaja et al., 1998; Tanimura et al., 1998). Interestingly, lower levels of baseline phosphorylation of Akt and ERK 1/2 were observed in MDCK^sol cells compared with control cells; however, the trend of HGF/SF-induced activation of Akt and ERK 1/2 in MDCK^sol cells was not different to control cells. We postulate that the lower basal and inducible activation of the ERK 1/2 and Akt proteins in these cells could be a result of constitutive activation of EGF receptors.

Interestingly, the antibody against P-Akt recognized an additional band of relatively higher molecular weight in addition to the ∼65 kDa band for P-Akt, and the signal intensity and timing of phosphorylation for the higher molecular weight band (∼67-70 kDa) were different between MDCK^con and MDCK^5aa cells. Pretreatment of the cells with LY294002 inhibited not only the activation of the ∼65 kDa (Akt) band but also the activation of the higher molecular weight band (data not shown). Recently, three Akt isoforms, Akt1/PKBα, Akt2/PKBβ and Akt3/PKBγ have been reported (Datta et al., 1999). Although the literature regarding the role of different Akt-isoforms is still in its nascent phase, these isoforms have been shown to be differentially regulated by EGF stimulation (Okano et al., 2000). It is tempting to speculate that the observed unidentified band could be an Akt isoform and its activation is induced by the expression of membrane-anchored HB-EGF. However, the identity and significance of this band remains to be determined because of the current lack of specific reagents for canine cells.

One of the most striking findings of the present study was the persistence of the cell-junction integrity and significantly higher TER in MDCK^5aa cells. Furthermore, administration of exogenous EGF/soluble HB-EGF or inhibition of EGFR tyrosine phosphorylation did not alter HGF/SF-induced cell scattering in the control cells. Also, no major differences were observed in EGF receptor tyrosine phosphorylation under basal conditions or in response to HGF administration among the cells under study, whereas a blocking antibody against the extracellular domain of CD9, as well as pretreatment of cells with heparinase I, both induced cell scattering in MDCK^5aa cells in response to HGF. These findings suggest that juxtacrine EGFR activation per se may not be necessary and/or sufficient for the observed phenotypic changes in MDCK^5aa cells and that membrane-anchored HB-EGF may be functioning as a scaffolding protein, as proHB-EGF is known to complex with CD9, a member of the tetraspanin superfamily (Lagaudriere-Gesbert et al., 1997). CD9 associates only with membrane-associated proHB-EGF but not soluble HB-EGF by interacting with the extracellular heparin-binding domain (Nakamura et al., 2000). CD9 interacts with selected integrins, especially β1 integrins, which are important in cell-cell and cell-extracellular matrix adhesive properties (Nakamura et al., 1995). In Vero cells expressing HB-EGF and CD9, HB-EGF-CD9 complexes localize to the cell-cell contact interface in association with α-catenin and vinculin, and CD9 interacts specifically with α3β1 integrins at the adherens junctions (Nakamura et al., 1995). Membrane-associated HB-EGF also interacts with cell-surface heparin-sulfate proteoglycans (Paria et al., 1999). Therefore, our results suggest that the phenotypic changes observed in MDCK^5aa cells may result from membrane-bound HB-EGF interacting with proteins such as the tetraspanin, integrins or heparin-sulfate proteoglycans, rather than by juxtacrine receptor activation.

Disruption of cell-cell and cell-matrix interaction is of central importance in epithelial cell injury, including renal injuries such as ischemia-reperfusion (Racusen, 1997; Sutton and Molitoris, 1998). As discussed above, owing to their basolateral localization in the epithelial cells, EGF receptor, proHB-EGF, CD9 and α3β1 integrins are capable of making a large protein complex involving heparin-sulfate proteoglycans at the cell-cell contact site. We postulate that proHB-EGF helps to maintain the integrity of epithelial sheet through these interactions, providing stronger cell-cell and cell-matrix interaction. This hypothesis is supported by significantly higher cell-matrix and cell-cell adhesion in our cells expressing the noncleavable mutant of proHB-EGF. When the integrity of a tissue is disrupted, surrounding cells such as platelets, monocytes/macrophages and other cells in the damaged tissue release several factors including lysophosphatidic acid (LPA). Activation of G protein-coupled receptors induces ectodomain shedding of proHB-EGF (Prenzel et al., 1999). LPA is one of the ligands of GPCR that stimulates ectodomain shedding of proHB-EGF (Hirata et al., 2001). LPA and other factors possibly induce ectodomain shedding of proHB-EGF to yield sHB-EGF. sHB-EGF may then help in the regeneration of the damaged tissue by stimulating proliferation as well as migration of cells. This proliferation and migration of the cells would reinstate the cell-cell contact, and interactions of proHB-EGF with EGFR and other proteins would be reconstituted and cell-cell integrity re-established.

In summary, we have developed novel engineered MDCK cell lines that allow the study of membrane-associated and soluble HB-EGF. Using these cell lines we have shown that expression of membrane-anchored HB-EGF inhibited HGF/SF-induced epithelial cell scattering and tubulogenesis. We have also shown that the expression of membrane-anchored HB-EGF resulted in enhanced cell adhesion as well as tighter cell-cell contact and inhibited cell migration and HGF/SF-induced cell scattering and tubulogenesis. By contrast, expression of soluble HB-EGF resulted in reduced cell-extracellular matrix and cell-cell interactions and enhanced cell migration. These studies describe the distinctive roles of membrane-associated and soluble HB-EGF in epithelial cell morphogenesis and indicate that an EGFR ligand can modulate HGF/SF-induced responses in MDCK cells.

This work was supported by National Institutes of Health grant DK51265 (R.C.H.), funds from the Department of Veterans Affairs (R.C.H.) and American Heart Association Southeast Affiliate grant 4043756112-2003 (A.B.S.). Dr Roy Zent is a clinician scientist of National Kidney Foundation (NKF) and is funded by the Advanced Career Development award from the Department of Veterans Affairs. Experiments were performed in part through the use of the VUMC Cell Imaging Core Resource (supported by NIH grants CA68485 and DK20593).