An 11-amino acid sequence from c-met initiates epithelial chemotaxis via phosphatidylinositol 3-kinase and phospholipase C.

Interaction of hepatocyte growth factor with its high affinity receptor c-met initiates a cascade of intracellular events leading to epithelial motility. An 11-amino acid sequence from the c-met receptor has been found to cause cell transformation in transfected fibroblasts (Ponzetto, C., Bardelli, A., Zhen, Z., Maina, F., Dalla, Z. P., Giordano, S., Graziani, A., Panayotou, G., and Comoglio, P. M.(1994) Cell 77, 261-271). We inserted this sequence into a mutant platelet-derived growth factor receptor (F5) to determine if this region of c-met can initiate cell motility and which signaling pathways it activates. The platelet-derived growth factor (PDGF) receptor/c-met hybrid (F5 met) initiated PDGF-dependent chemotaxis in renal epithelial cells (8.0 +/- 2.3 versus 70.5 +/- 4.8 cells/mm2), while the parental construct, F5, did not. Addition of PDGF to cells expressing F5 met caused activation of the phosphatidylinositol (PI) 3-kinase (control 2.0 +/- 0.8, +PDGF 17.1 +/- 5.1, n = 3, p < 0.05) and phospholipase C (control 478.5 +/- 67 dpm/well, +PDGF 1049.3 +/- 93, n = 4, p = 0.003), while neither pathway was activated in cells expressing F5. The chemotactic response of F5 met was inhibited by both the PI 3-kinase inhibitor wortmannin and the phospholipase C inhibitor U-71322. Selective activation of the PI 3-kinase utilizing a PDGF receptor mutant (F3) containing the native high affinity PI 3-kinase binding site also resulted in PDGF stimulated chemotaxis, although less than that generated by the c-met sequence. These findings demonstrate that the 11-amino acid sequence from c-met initiates epithelial motility via coincident activation of the PI 3-kinase and phospholipase C and that selective activation of the PI 3-kinase can initiate a partial chemotactic response.

Hepatocyte growth factor (HGF) 1 has been independently characterized by its ability both to induce mitogenesis (1) and to cause "scattering" of epithelial cells in culture (2). Further study has implicated HGF and its high affinity receptor, c-met, in diverse biologic events that involve the motile response: chemotaxis and branching tubulogenesis in renal epithelial cells (3,4) and wound healing and angiogenesis in endothelial cells (3,5). The motile response induced by c-met is a complex one, which involves dissociation of cell contacts, ruffling of the leading edge, extension of lamellipodia, and finally migration (6). The role of c-met in this series of events has not yet been well characterized, but it is likely to involve the integration of multiple signaling pathways.
Examination of the COOH-terminal region of c-met reveals a single 11-amino acid sequence, Y 1349 VHVNATY 1356 VNV, which is predicted to be a low affinity site for the two SH2 domains of phospholipase C␥ (PLC␥) and the phosphatidylinositol 3-kinase (PI 3-K), and a high affinity site for the ras activating complex GRB2-hSos1 (7). No other domain outside of the tyrosine kinase region of c-met is predicted to bind any known signaling proteins, making it possible that this 11amino acid domain is a site for competitive interaction of many, if not all, of the receptor's signaling proteins. Indeed, Ponzetto et al. (8) were able to demonstrate co-immunoprecipitation of overexpressed c-met with PLC␥, GRB2-hSos1, and PI 3-K as well as pp60 src (8). Altering both tyrosines in this 11-amino acid sequence eliminated association with all four proteins, leading these researchers to term this region a "multifunctional docking site." Thus, this 11-amino acid sequence might be sufficient to activate all of the signaling activity induced by HGF and mediate the motility response.
Work in our laboratory has demonstrated the importance of the PI 3-K in HGF-mediated motility (4). Using the PI 3-K inhibitors wortmannin and LY-294002, we found striking inhibition of both motogenesis and morphogenesis. These experiments demonstrate that c-met-mediated activation of the PI 3-K is required for the full chemotactic response, but they do not address whether the PI 3-K alone is sufficient to initiate chemotaxis.
In this report, we used an altered platelet-derived growth factor receptor (PDGFR), which is unable to bind to PI 3-K, RasGAP, Syp/PTP, or PLC␥ (9) as a vehicle to introduce and study the role of the 11-amino acid c-met sequence in the motility response of epithelial cells. Use of a PDGFR mutant in our inner medullary collecting duct cells (which do not express endogenous PDGFR) makes it possible to study selective tyrosine phosphorylation of this met sequence in the setting of normal epithelial signaling machinery. In addition, a second PDGFR mutant with the native high affinity PI 3-K binding site (Y 740 MDM . . . Y 751 VPM) intact but with Tyr 3 Phe substitutions of the tyrosines critical for association with RasGAP, Syp/PTP and PLC␥ (F3) were utilized to examine more selective activation of the PI 3-K in epithelial chemotaxis.

MATERIALS AND METHODS
Cell Culture and Reagents-mIMCD-3 cells are a murine renal tubular epithelial cell line that expresses the c-met receptor and exhibits striking chemotaxis to a gradient of HGF (4,10,11). Screening of these cells by Northern analysis, polymerase chain reaction, and Western blots revealed no evidence of the PDGFR. In addition, mIMCD-3 cells did not demonstrate a mitogenic (data not shown) nor chemotactic response to PDGF (see Fig. 2). NIH 3T3 cells were used as a control for expression of the native PDGFR. All cells were cultured in Dulbecco's modified Eagle's medium with 5% fetal calf serum using standard * This work was supported in part by National Institutes of Health Grants DK 44503 and DK 18078. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
techniques. PDGF (Upstate Biotechnology, Inc.) and HGF (a gift from Dr. George Vande Woude) were used in concentrations from 1 to 100 ng/ml for dose-response curves (not shown) with maximal effects on chemotaxis detected at 10 ng/ml in NIH 3T3 cells for PDGF and 40 ng/ml in mIMCD-3 cells for HGF. In addition, recombinant HGF (Institute of Immunology Co., Ltd., Tokyo, Japan) was tested with equivalent results. Wortmannin was obtained from Sigma; U-71322 was from Biomol (Plymouth Meeting, PA); and diacylglycerol was from Boehringer Manheim.
Creation of PDGF Receptor Clones-Two PDGFR mutants (F5, in which the tyrosine residues at 740, 751, 771, 1009, and 1021 have been replaced by phenylalanines, and F3, in which the residues at 740 and 751 remain tyrosines) were provided to us as a gift by Dr. Andrius Kazlauskas (9). The kinase insert region of the F5 PDGFR was then modified to encode novel SalI and AatII restriction sites and allow insertion of the 33-nucleotide cDNA sequence encoding the 11-amino acid motif YVHVNATYVNV from the c-met receptor (designated F5 met) (Fig. 1). The identity of each construct was confirmed by sequencing.
Each of the constructs was then cloned into pCMV expression vectors and transfected into mIMCD-3 cells using the lipofectin reagent (Life Technologies, Inc.), and independent clones were selected on G418 (400 mcg/ml).
Protein Electrophoresis-Confluent plates of cells were serum starved for 24 h, stimulated with either vehicle control, PDGF, or HGF for 5 min followed by phosphate-buffered saline wash, lysis in ice-cold lysis buffer (137 mM NaCl, 20 mM Tris, 1 mM MgCl 2 , 1 mM CaCl 2 , 10% glycerol, 1% Nonidet P-40, 2 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, pH 7.5), and centrifuged for 10 min at 12,000 ϫ g. The supernatant was resolved by SDS-polyacrylamide gel electrophoresis and transferred to Immobilon (Millipore). Expression of full-length PDGFR was detected by blotting with an anti-PDGFR antibody known to detect the F5 mutant receptor (a generous gift of Dr. Andrius Kazlauskas) or anti-phosphotyrosine (ICN Biomedicals Inc, Costa Mesa, CA). Appropriate Western blots were performed using anti-PDGFR or anti-phosphotyrosine immunoprecipitates of equal sample size as determined by protein assay (Bio-Rad).
Chemotaxis Assay-Chemotaxis was evaluated using a modified Boyden chamber assay with 24 individual blind wells as described previously (Neuro Probe Inc., Cabin John, MD) (4,12). Serum-free medium containing either vehicle control, PDGF, or HGF was added to the bottom chamber and then covered with a polycarbonate filter (Nucleopore Corp., Pleasanton, CA) coated with rat tail collagen type I (Collaborative Biomedical, Bedford, MA). 2.5 ϫ 10 4 cells were added to the upper compartment. After 4 h of incubation at 37°C, filters were removed, the cells were fixed and stained with Diff-Quik (Baxter Healthcare Corp., Miami, FL), and the upper surface was wiped with a cotton applicator to remove nonchemotaxing cells. Cells that had passed through the pores were then counted in five nonoverlapping 10ϫ fields, and the mean value of cells/mm 2 was calculated. Although total numbers of migrating cells varied moderately between experiments, the ratio of control to stimulated chemotaxis remained consistent. In some experiments, the PI 3-K inhibitor wortmannin was added to both upper and lower chambers. In other experiments, the PLC inhibitor U-71322 was similarly employed.
In Vitro Phosphoinositide 3-Kinase Assay-Immunoprecipitation of the PI 3-kinase was performed as described previously (11). Briefly, following 24 h of serum deprivation, confluent cells were stimulated with ligand and lysed as above. Aliquots of supernatant containing equal amounts of protein were immunoprecipitated for 3 h at 4°C with monoclonal antibody directed against phosphotyrosine using protein A-Sepharose beads (Sigma). Beads were washed and incubated for 10 min at room temperature in kinase buffer (62.5 mM ATP, 6.25 mM MgCl 2 , 1.25 mM Hepes, 0.25 mg/ml phosphatidylinositol (Avanti Polar Lipids), 50 Ci of [␥-32 P]ATP, pH 7.0). Lipids were then extracted and separated on oxalate-coated thin-layer chromatography plates (EM Separations), and autoradiography was performed. The labeled lipids migrating as phosphatidylinositol 3-phosphate (PI3P) were then cut out and counted in a ␤ scintillation counter for quantification of amount of PI3P generated.
In Vitro Phospholipase C Assay-24-well plates of either F5 or F5 met cells were incubated with inositol-free medium and 5 Ci/ml [ 3 H] inositol for 48 h. Cells were then washed twice in PBS, followed by a 20-min incubation with inositol-free media. PDGF or vehicle control was added to appropriate wells for 2 min. Cells were lysed with ice-cold trichloroacetic acid, and samples were extracted with water-saturated ether, dried in a vacuum concentrator, and separated by HPLC by the method of Auger et al. (13).
Statistical Analysis-Results were averaged, and statistical relevance was determined by the Student's t test. Data are presented as mean Ϯ S.E.

Cells Expressing F5 met Chemotax in Response to PDGF-
G418-resistant clones from transfections with the F5 and F5 met constructs were simultaneously screened for expression of the PDGFR and chemotaxis in response to a PDGF gradient.
None of the F5 clones tested exhibited PDGF-dependent chemotaxis (Fig. 2), regardless of the level of expression of the full-length F5 protein. One clone, F5 16 , which had the highest level of PDGFR expression, was then utilized as a control for further experiments. Evaluation of the F5 16 mIMCD-3 cell clone revealed no chemotaxis in response to PDGF (control, 6.3 Ϯ 4.0 cells/mm 2 , n ϭ 4; PDGF, 2.7 Ϯ 1.2 cells/mm 2 , n ϭ 4; p ϭ not significant), despite a full chemotactic response to HGF (154 Ϯ 17.3, p Ͻ 0.001).
In contrast, we identified eight F5 met clones that demonstrated chemotaxis in response to PDGF and found that all eight clones expressed the PDGFR. Additionally, no clone exhibited PDGF-dependent chemotaxis, which did not express the receptor. Among these clones, the amount of chemotaxis correlated well with the level of expression of the PDGFR (Fig.   FIG. 1. Schematic depiction of the PDGF receptor mutants. The wild-type PDGFR is shown highlighting the 5 tyrosines critical for association with the PI 3-kinase (740/751), RasGAP (771), SHPTP2 (1009), and PLC␥ (1021). The F5 clone has phenylalanine substitutions for all 5 residues. F5 met has the sequence from 750 -760 replaced with the 11-amino acid c-met sequence YVHVNATYVNV. F3 clone retains the 740/751 tyrosines for PI 3-kinase binding.
FIG. 2. Directional chemotaxis of mIMCD-3 cell clones toward a PDGF gradient. Independent clones for each mutant PDGFR were screened for the ability to migrate in response to PDGF. F5 PDGFR clones did not exhibit chemotaxis, while F5 met clones did chemotax toward PDGF.

An 11-Amino Acid Sequence of c-met Mediates Chemotaxis
3). Interestingly, clones F5 met 29 and F5 met 41 subsequently failed to chemotax to PDGF and were found to no longer express the PDGFR by Western analysis. One clone, F5 met 23 , which stably expressed the mutant PDGFR, was used for further experiments and displayed consistent ability to chemotax toward PDGF (control, 4.6 Ϯ 3.5 cells/mm 2 , n ϭ 9; PDGF, 78.6 Ϯ 12.7 cells/mm 2 , n ϭ 11; p Ͻ 0.001).
F5 met Chemotaxis Correlates with Ligand-dependent Association with the PI 3-K-Following demonstration that the F5 16 and F5 met 23 clones exhibited similar levels of receptor protein expression and PDGF-dependent tyrosine phosphorylation (Fig. 4), we examined the ability of these clones to associate with the active form of the PI 3-K (Fig. 5). Antiphosphotyrosine immunoprecipitates of F5 met 23 cells revealed a marked increase in receptor association with the active form of the PI 3-K following stimulation with PDGF, whereas parental mIMCD-3 and F5 cells revealed essentially no change in PI 3-K activity in the stimulated immunoprecipitates.
The importance of the association of the mutant PDGFR with the PI 3-K in the initiation of chemotaxis was confirmed by incubation of the cells with the PI 3-K inhibitor wortmannin. In these experiments, 10 nM wortmannin was found to markedly inhibit PDGF-dependent chemotaxis in F5 met cells (ϩPDGF, 179.1 Ϯ 42.1 versus ϩPDGF ϩ 10 nM wortmannin, 70 Ϯ 10.3, n ϭ 9, p ϭ 0.023).
Selective Activation of the PI 3-K Initiates a Chemotactic Response-To examine the effect of selective activation of the PI 3-K on cell motility, we also transfected mIMCD-3 cells with a mutant PDGF receptor which has Tyr 3 Phe substitutions at binding sites for PLC, Syp(PTP), and RasGAP but with the high affinity binding site for PI3-K intact (F3). Four G418resistant F3 clones were identified that expressed the receptor and demonstrated PDGF-dependent chemotaxis (Table I). Again, all clones that expressed the receptor at high levels were found to chemotax to PDGF, and no clones demonstrated chemotaxis in the absence of receptor expression. As expected, the high affinity native PDGFR PI 3-K binding site in the F3 1 cells mediated a marked increase in PI 3-K association following stimulation with PDGF (control 2.12 Ϯ 0.34 dpm PI3P/g, ϩ10 ng/ml PDGF 76.69 Ϯ 17.55 dpm PI3P/g (n ϭ 6, p ϭ 0.002) Fig.  5). Despite this 4-fold greater association of the F3 receptor with the PI 3-K than observed with F5 met, chemotactic rates to a gradient of PDGF were less for the F3 1 cells as compared with the F5 met 23 cells in all experiments (Table I, Fig. 2). Thus selective activation of the PI 3-K can initiate a partial chemotactic response. DISCUSSION The potent chemotactic effect of HGF on epithelial cells makes the study of its receptor, c-met, a particularly relevant system for determining the signaling cascade involved in the epithelial motility response. The use of a hybrid F5-c-met receptor (F5 met) allowed us to selectively examine the role of an 11-amino acid region of c-met in epithelial cell motility and signal transduction.
We initially determined that all three of the transfected receptor constructs were successfully expressed in mIMCD-3 cells and underwent PDGF-dependent tyrosine phosphorylation. The F5 mutant PDGFR, which does not bind PI 3-K, PLC␥, PTP, or RasGAP but is still capable of associating with src (via Y 579 IYV) and possibly GRB-2 (via Y 716 SNA) (15) did not initiate epithelial chemotaxis in response to PDGF. However, expression of the F5 met hybrid receptor did initiate chemotaxis in a ligand-dependent manner, demonstrating that phosphorylation of the inserted c-met sequence Y 1349 VHVNATYVNV activated those signaling pathways necessary for epithelial cell motility.
Based on our observation that the PI 3-K inhibitors wortmannin and LY294002 could partially block HGF/c-met-mediated epithelial chemotaxis (4), as well as data from Kundra and co-workers (16) that demonstrated that PDGF constructs excluding PLC activation exhibited diminished chemotactic responses (16), we examined the ability of the 11-amino acid c-met sequence to activate these two candidate signaling pathways in epithelial cells. Indeed, both the PI 3-K and PLC were activated in a ligand-dependent manner by the F5 met hybrid receptor, but not by the parental F5 construct. The importance of these two pathways in chemotaxis was then examined using inhibitors of PI 3-K and PLC as well as a receptor construct, which selectively activates PI 3-K. Both wortmannin, an inhibitor of PI 3-K, and U-73122, an inhibitor of PLC, caused marked reductions of F5 met-stimulated chemotaxis. Thus, the PI 3-K and PLC are activated in vivo by the 11-amino acid sequence from c-met and contribute to the chemotactic response initiated by this receptor.
In addition, the F3 PDGFR in which the native high affinity PI 3-K binding site (Y 740 MDM . . . Y 751 VPM) has been selectively restored (but which still lacks PLC, RasGAP, and PTP binding sites) also caused epithelial chemotaxis in response to PDGF. It should be noted that one study has described the small adaptor molecule Nck as competing for binding to the Y 751 VPM sequence of F3 as well (17). The above data demonstrate a critical role for both the PI 3-K and PLC in epithelial cell migration and are the first to show that selective activation of PI 3-K is sufficient to initiate chemotaxis.
Our findings are complemented by data from several laboratories. Wennstrom et al. (18) examined the ruffling response in porcine aortic endothelial cells transfected with a mutant PDGFR, which selectively excluded binding of the PI 3-K (Tyr 3 Phe substitutions at 740 and 751) (18). This construct, the opposite of our F3 PDGFR, failed to mediate ruffling in response to PDGF, implicating a requirement for PI 3-K in membrane ruffling. Likewise, Kundra et al. (16) examined the same Y740F/Y751F mutation in chemotactic assays of a canine kidney epithelial cell line (16). Again, the chemotactic response was obliterated by an inability to activate the PI 3-K, as well as  a mutant excluding activation of PLC. Here, we demonstrate that selective activation of PI 3-K initiates a chemotactic response, while coactivation of PI 3-K and PLC produces a marked increase in this response.
Other signaling pathways important in cell movement involve the small GTP binding proteins Ras, Rac and Rho. In a recent study by Ridley et al. (19), microinjection of Madin-Darby canine kidney cells with constituitively active Ras reproduced HGF-mediated ruffling but did not initiate cell scattering (19). Thus, it may be that isolated activation of ras is sufficient to generate actin filament rearrangements but not actual cell movement. In agreement with this, our F5 PDGFR clones did not mediate PDGF-dependent chemotaxis, even though ras activation may occur via GRB2/hSos1 nucleotide exchange factor association at Y 716 SNA in the F5 receptor (20). However, since both the F5 met and F3 receptors are capable of Grb2/hSos1/Ras activation, we cannot rule out a contributory or co-stimulatory effect of Ras activation on chemotaxis in these experiments.
The c-met receptor mediates diverse phenotypic changes in several cell types including cell motility and tubule formation. An 11-amino acid sequence from this receptor, when phosphorylated in a ligand dependent manner, activates both PLC and the PI 3-K, triggering the intracellular events necessary for epithelial chemotaxis.