Integrin-mediated Migration of Murine B82L Fibroblasts Is Dependent on the Expression of an Intact Epidermal Growth Factor Receptor*

To evaluate the mechanisms by which epidermal growth factor (EGF) regulates actin-based cellular processes such as cell migration, we first examined the effects of EGF on cell adhesion, which is essential for cell migration. In mouse B82L fibroblasts transfected with the full-length EGF receptor, EGF promotes cell rounding and attenuates cell spreading on fibronectin, laminin, and vitronectin, and thus appears to reduce the strength of cell adhesion. Moreover, EGF synergizes with multiple extracellular matrix (ECM) components in the promotion of integrin-mediated cell migration of several different cell types, including fibroblasts and various carcinoma and osteosarcoma cell lines. Interestingly, co-presentation (co-positioning) of EGF with laminin or fibronectin is essential for EGF-stimulated migration. When EGF is mixed with the cells instead of the ECM components, it has little effect on cell migration. These results suggest that co-presentation of EGF with ECM components can enhance the polarization events required for directional cell movement. To identify the EGF receptor elements critical for the EGF stimulation of cell migration, B82L fibroblasts were transfected with either mutated or wild-type EGF receptors. Surprisingly, we found that B82L-Parental cells that lack the EGF receptor are not able to migrate to fibronectin, even though they can adhere to fibronectin. However, the introduction of wild-type EGF receptors into these fibroblasts enables them to migrate toward fibronectin even in the absence of EGF. The requirement of the EGF receptor for cell migration does not appear to result from the secretion of EGF or TGF-α by the cells transfected with the EGF receptor. Furthermore, cells expressing EGF receptors that are kinase-inactive, or C-terminally truncated, exhibit little migration toward fibronectin, indicating that an intact EGF receptor kinase is required for fibronectin-induced cell migration. In addition, neutralizing anti-EGF receptor antibodies attenuate cell migration in the presence of EGF, and inhibit migration to fibronectin or laminin alone. These results further suggest that the EGF receptor is downstream of integrin activation in the signal transduction pathways leading to fibroblast migration.

EGF 1 triggers many biological responses, including cell proliferation and differentiation (1). In addition, EGF has been shown to induce the reorganization of the actin cytoskeleton, and the EGF receptor has been found to be associated with actin filaments (2)(3)(4)(5)(6)(7). In this regard, EGF has been reported to stimulate rapid cell rounding, extensive membrane ruffling, extension of filopodia, retraction of cells from the substratum (8,9), extensive cortical actin polymerization, and depolymerization of actin stress fibers (10,11). Moreover, numerous studies have shown that activation of the EGF receptor leads to increased cell motility (12)(13)(14)(15)(16)(17)(18)(19) and production of ECM degrading proteases (20 -23), thereby supporting a role for the EGF receptor in normal development and pathophysiological events such as tumor cell invasion and metastasis.
Cell migration plays a central role in a variety of biological processes including embryonic development, angiogenesis, wound healing, and tumor cell metastasis (24). Although the exact mechanisms of cell migration are not well established, it is generally believed that several coordinated events are involved, including morphological polarization, membrane extension, formation of cell-substratum attachments, contractile force and traction, and release of attachments (24). The adhesive interactions between cells and various ECM substrates are likely to be critical in determining cell migration capacity (24,25). Many investigations have shown that an intermediate adhesive strength generates maximal cell migration (26 -33). The adhesion between cells and substrate are largely mediated by the integrins, which are a family of cell surface heterodimeric receptors that bind to ECM proteins such as laminin, fibronectin, and vitronectin (34). Integrin expression, the affinity and specificity for their ligands, and the integrin-cytoskeleton linkages are regulated by various signals including those initiated by growth factors. In fact, integrins and growth factor receptors share many common signaling events, such as increased tyrosine phosphorylation, activation of mitogen-activated protein (MAP) kinases, protein kinase C isoforms, and small molecular weight GTP-binding proteins, as well as enhanced Ca 2ϩ fluxes (35)(36)(37)(38). Given their common signaling pathways, and the association of both EGF receptor and integrin systems with the actin cytoskeleton, it is conceivable that the EGF receptor can be involved in the modulation of integrinmediated cellular functions such as cell migration.
Previous analyses of the EGF receptor have revealed that the receptor tyrosine kinase activity and the C-terminal autophosphorylation sites are critical for EGF-stimulated signal transduction (1). The C-terminal domain of the receptor exerts a competitive autoinhibitory restraint on the receptor tyrosine kinase activity that can be removed by autophosphorylation of the tyrosine residues located within this domain (39,40). Furthermore, the autophosphorylated tyrosine residues at the EGF receptor C terminus provide anchoring sites for downstream Src homology 2 (SH2) or phosphotyrosine binding (PTB) domain-containing effectors that are involved in the transduction of EGF-stimulated intracellular signals (1). Chen et al. (17,18) have reported that the receptor kinase activity and at least one C-terminal tyrosine autophosphorylation site are required for cell movement. In addition, EGF stimulation of phospholipase C-␥ and protein kinase C has been linked to the EGF induction of cell motility (14,18).
In the present study, we evaluated the effects of EGF on cell migration and assessed the capacity of various EGF receptor constructs to modulate the motility of mouse B82L-Parental fibroblasts that possess no detectable endogenous EGF receptors. We found that EGF synergistically stimulates integrinmediated chemotaxis, and that the co-positioning of EGF and the chemoattractant is critical for EGF-stimulated migration. In particular, it is noteworthy that, although B82L-Parental cells do not exhibit fibronectin-induced chemotaxis, they do adhere to fibronectin, but it is the introduction of an intact EGF receptor into these cells that allows for fibronectin-induced motility. In addition, neutralizing anti-EGF receptor antibodies inhibit cell migration toward fibronectin or laminin alone. These findings demonstrate that the EGF receptor is important for B82L fibroblast motility, suggesting that the EGF receptor may act downstream of integrin activation and may directly engage in the signal transduction events leading to cell migration.

MATERIALS AND METHODS
Reagents and Antibodies-Mouse laminin was obtained from Collaborative Research (Bedford, MA). Human fibronectin and vitronectin were generous gifts from Dr. Deane F. Mosher (University of Wisconsin, Madison, WI). Recombinant human EGF was purchased from Upstate Biotechnology Inc. (UBI, Lake Placid, NY). Nucleopore membranes and the 48-well chemotaxis chamber were purchased from Nucleopore Corp. (Pleasanton, CA). Anti-EGF receptor monoclonal antibodies were obtained from the following sources: antibody 528 was affinity-purified from hybridoma cell line HB 8509 (ATCC, Rockville, MD), antibody C225 was obtained from Imclone Systems, Inc. (New York, NY), and antibody LA22 was purchased from UBI. Neutralizing anti-mouse EGF and anti-human TGF-␣ antibodies were purchased from UBI and R&D Systems (Minneapolis, MN), respectively.
Cell Culture Conditions-Murine B82L-Parental fibroblasts and those expressing the wild-type or the K721M or c'1022 EGF receptor constructs were provided by Dr. Gordon Gill (University of California, San Diego, CA). Mouse B82L-Parental fibroblasts were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% cosmic calf serum (HyClone, Logan, UT). Cells transfected with wild-type or mutated EGF receptors were cultured in the same medium that contained 10 M methotrexate because a mutant dihydrofolate reductase gene was used as a selectable marker (41). The Clone B3 cells were derived from B82L-wt cells based on their ability to bind laminin (42), and were cultured in the same medium as the B82L-wt cells. The GD25 cells are differentiated and immortalized cells derived from the embryonic stem cell clone G201, which are deficient in the integrin subunit ␤ 1 as the result of the introduction of a null mutation in the ␤ 1 integrin gene via homologous recombination (43). The stably transfected cell line GD25-␤ 1A was obtained by electroporating wild-type integrin ␤ 1A cDNA into GD25 cells. In the present study, the GD25 cells were cultured in nonselection medium consisting of DMEM plus either 10% fetal calf serum (HyClone) or 10% cosmic calf serum, whereas the GD25-␤ 1A cells were continuously cultured in the same medium plus puromycin (10 g/ml) (43). Both the GD25 and the GD25-␤ 1A cells were provided by Dr. Deane F. Mosher. Other cell lines used in this study, including human mammary carcinoma cell line MDA468, human osteosarcoma cell line MG63 and human epidermal carcinoma cell line A431, were cultured in DMEM plus 10% fetal calf serum. 3T3-F442A fibroblasts were cultured in DMEM plus 10% bovine calf serum (HyClone).
Cell Spreading Analysis-Cell culture plates (24-well) were coated with purified matrix or control proteins (laminin, 20 g/ml; fibronectin, 10 g/ml; vitronectin, 10 g/ml; polylysine, 10 g/ml) diluted in phosphate-buffered saline using either an overnight incubation at 4°C or a 2-h incubation at 37°C. Nonspecific binding sites were blocked using a 2% (w/v) bovine serum albumin (BSA) solution (in phosphate-buffered saline) and a 1-h incubation at 37°C. The Clone B3 cells were detached by digestion with 0.1% trypsin, and the cells were then resuspended in DMEM containing 0.1% BSA. The cells were mixed with EGF (at the indicated final concentrations), seeded at a density of 2 ϫ 10 5 cells/ml, and incubated at 37°C in a humidified incubator containing 5% CO 2 . Samples were viewed using a light microscope equipped with phase contrast (Nikon Inc., Instrument Group, Melville, NY), and photographs were taken at different times (magnification, ϫ40).
Cell Motility Assays-The experiments assessing chemotactic motility were performed using a 48-well migration chamber as described previously (42). In these studies, cells were grown as described under "Cell Culture Conditions" for 3-4 days. After being detached from the plastic dishes using a 0.1% trypsin solution, the cells were stabilized for 1 h at 37°C in DMEM containing 0.1% BSA. Cells were then counted and resuspended in DMEM plus 0.1% BSA at a final concentration of 1 ϫ 10 5 cells/50 l. The lower compartment of the migration chamber was filled with the indicated proteins dissolved in DMEM ϩ 0.1% BSA (29 l/well), and the cells were added to the upper compartment of the migration chamber (50 l/well). To ascertain the effects of EGF, the growth factor was added either to the upper or to the lower compartments at indicated concentrations. To evaluate the effects of anti-EGF receptor, anti-EGF, or anti-TGF-␣ antibodies, the antibodies were incubated with the cells for 30 min before the cells were placed into the upper wells. In all experiments, the two compartments of the migration chamber were separated by a polycarbonate filter (5-m pore size, Nucleopore Corp.). The cells were allowed to migrate for 4 h at 37°C in a humidified atmosphere containing 5% CO 2 . The cells that did not migrate through the membrane remained on the upper surface of the filter and were removed mechanically by scraping; the migrant cells on the lower surface were fixed in methanol/acetone (1:1) for 2 min, and then stained with 1% crystal violet. The filters were densitometrically analyzed using the OFOTO program (Light Source Computer Images, Inc.) and quantified by Scanner Analysis (Biosoft, Ferguson, MO).

RESULTS
To explore the effects of EGF on cell migration, mouse B82L fibroblasts that lack measurable endogenous EGF receptors (B82L-Parental) were transfected with wild-type EGF receptors (these include two separate transfectants designated as B82L-wt and B82L-wt2) or mutant human EGF receptors. A cell clone possessing laminin binding activity was also isolated from B82L-wt cells (labeled as Clone B3) (42). The two EGF receptor mutants used in these studies include a construct encoding a kinase-inactive receptor that contains a lysine to methionine substitution at residue 721 (B82L-K721M), which is involved in ATP binding, and a construct encoding a kinaseactive receptor that has been truncated at residue 1022 (B82Lc'1022), which lacks four major receptor autophosphorylation sites. These cell lines were used to explore the involvement of EGF receptor activation in cell migration toward various ECM components.
Effect of EGF on B82L Fibroblast Spreading on ECM Substrates-Cell migration is a process that requires temporally and spatially coordinated cell attachment and detachment (24,25). Given the importance of adhesive strength between the cell and the substrate in determining the migration speed and persistence, we tested whether EGF could influence the adhesiveness of Clone B3 cells on several ECM substrates. As shown in Fig. 1A, when Clone B3 cells were plated onto fibronectincoated wells in the presence of different levels of EGF, the ability of these cells to spread on the fibronectin surface was attenuated in a time-and EGF dose-dependent manner. In the absence of EGF, nearly all the clone B3 cells were attached to the fibronectin surface in the first 30 min. Within 1 h, the cells became flattened and exhibited the characteristic protrusive structures of spreading cells, a process that was completed by 2.5 h after plating. However, upon the addition of EGF, cell spreading was impaired, i.e. the cells remained in a rounded shape for a longer period of time and appeared to form fewer focal contacts as assessed by immunofluorescent staining of vinculin (data not shown). As also illustrated in Fig. 1A, the addition of 10 nM EGF delayed cell spreading until 1.5 h after plating, whereas the presence of 50 nM EGF suppressed the initiation of cell spreading to approximately 2.5 h after plating.
Because Clone B3 cells express receptors for other ECM proteins such as laminin and vitronectin, we examined the effects of EGF on Clone B3 spreading onto laminin or vitronectin-coated wells. Similar to the results shown in Fig. 1A, the addition of EGF was found to dose-dependently reduce the capacity of the cells to spread on all three substrates tested (Fig. 1B). It was also noted that the spreading of these mouse fibroblasts onto laminin-or vitronectin-coated wells was less extensive than that observed on fibronectin-coated surfaces. Polylysine was used as a negative control for cell spreading onto ECM components because cells attach to polylysine without the activation of integrin-associated processes (45). Interestingly, EGF treatment had no effect on the attachment of Clone B3 cells onto fibronectin and laminin, as measured by the number of cells adhered to these ECM substrates (data not shown). These results suggest that EGF does not substantially regulate the initial interaction between the cell surface integrins, but that the later events (e.g. focal contacts maturation and stress fiber formation) are more likely to be EGF-sensitive.
Effect of EGF on Laminin-induced Chemotaxis of Clone B3 Fibroblasts-Based on the ability of EGF to modulate the adhesiveness of Clone B3 cells on various substrates, we investigated whether EGF could regulate the chemotactic capacity of Clone B3 cells using a 48-well chemotaxis chamber. As shown in Fig. 2A, EGF synergized with laminin in enhancing the migratory activity of these cells. This synergistic effect of EGF on cell chemotaxis was about 2-3-fold above that observed with laminin alone, whereas EGF alone had little or no effect on cell migration. Interestingly, the effect of EGF on laminin-induced chemotaxis could only be achieved when EGF was co-present with laminin in the lower wells. When EGF was mixed with the cells located in the upper wells, EGF was unable to augment laminin-induced chemotaxis. Because the development of cell polarization, such as the formation of distinct cellular leading and trailing edges, is essential for directional cell movement (24), our findings suggest that EGF, when co-present with laminin, may help to enhance the cell polarization required for directional cell movement by forming an EGF concentration gradient parallel to the laminin gradient. Conversely, when EGF is present with the cells in the upper wells, the cells were exposed to EGF from all directions, and thus may not be able to enhance a polarization process. Although several studies have shown that EGF can stimulate cell movement (14, 16 -18, 46), this appears to be the first report concerning the importance of EGF positioning in promoting directional cell movement.
To further test the hypothesis that EGF-induced cell polarization is important in facilitating cell migration toward laminin, EGF was added to both the lower and upper wells, based on the rationale that the uniform exposure of cells to EGF would disrupt the cell polarization induced by EGF present with laminin in the lower wells. As shown in Fig. 2B, the maximal cell migration induced by the co-presentation of laminin and EGF (1 nM) in the lower wells was reduced in a dose-dependent fashion (33-85%) when EGF (1-10 nM) was also mixed with cells in the upper wells. These studies further support the concept that the stimulation of cells with a cogradient of laminin and EGF facilitates cell polarization and efficient directional cell locomotion.
EGF Effects on Cell Migration with Multiple Cell Lines and ECM Substrates-Clone B3 cells were isolated by their ability to bind laminin (42). To assess whether the effects of EGF on cell migration were restricted to laminin, we performed similar migration assays using fibronectin and vitronectin as the chemoattractants. We found that EGF addition to the upper wells had no measurable effects on fibronectin or vitronectin-induced FIG. 1. EGF attenuates the spreading of Clone B3 fibroblasts, which contain the full-length active EGF receptor, on various ECM substrates. A, the time-and EGF dose-dependent spreading of Clone B3 fibroblasts on fibronectin. Cells were collected by trypsinization and allowed to attach and spread on wells coated with fibronectin (10 g/ml) in the absence or presence of the specified concentrations of EGF as described under "Materials and Methods." The cells were photographed at the indicated times using 40X magnification. B, EGF dose-dependent spreading of Clone B3 fibroblasts on various ECM components. Cells were collected by trypsinization and allowed to attach and spread on wells coated with either fibronectin (10 g/ml), laminin (20 g/ml), vitronectin (10 g/ml), or the non-integrin-activating protein polylysine (10 g/ml) in the absence or presence of the specified concentrations of EGF. At the end of a 2-h incubation, the cells were photographed (original magnification, ϫ40). For both A and B, analogous results were observed in three other experiments. chemotaxis (data not shown); however, when EGF was mixed with fibronectin or vitronectin in the lower wells, EGF substantially elevated Clone B3 cell migration (Fig. 3, A and B). Moreover, because Clone B3 cells were isolated from B82L fibroblasts that were transfected with the human EGF receptor (B82L-wt) (42), we compared the chemotactic capacity of Clone B3 and B82L-wt cells in order to ascertain whether EGF stimulation of cell migration reflected clonal variation. We observed that both cell types exhibited comparable responses to EGF, fibronectin, and vitronectin (Fig. 3C).
The arginine-glycine-aspartate (RGD) sequence on many matrix proteins, including fibronectin and vitronectin, has been shown to interact directly with cell surface receptor integrins (34). Synthetic RGDS peptides can block such matrixintegrin interactions. To identify whether ECM-induced chemotaxis was mediated by integrins, Clone B3 cells were mixed with RGDS peptides before they were placed into the chemotaxis chamber. As shown in Fig. 4, the migration of Clone B3 cells toward fibronectin or vitronectin in the absence or presence of EGF were almost completely blocked by RGDS peptides but not by the control peptide SDGRG, whereas laminin-induced migration was not affected by RGDS peptides at the same concentrations (data not shown). This observation is consistent with the previous finding that, although the RGD sequence is present in the A chain of murine laminin, it is cryptic, and therefore not accessible in native laminin (47,48). In addition, the laminin receptor on Clone B3 cells mediating motility has been identified to be ␣ 6 -containing integrins (42), whose recognition sites on laminin are different from the RGDcontaining region (49). These results support the concept that fibronectin-or vitronectin-induced chemotaxis of Clone B3 cells is mediated by integrins.
To verify that the EGF stimulation of chemotaxis is not limited to B82L cells, we examined the chemotactic behavior of another mouse fibroblastic cell line in response to EGF treatment, namely GD25 cells. These cells were derived from the mouse embryonic stem cell clone G201, which are deficient in the integrin subunit ␤ 1 due to the introduction of a null mutation in the ␤ 1 integrin gene. In addition, we evaluated a related cell line, i.e. GD25-␤ 1A cells, which were derived from GD25 cells that were transfected with the integrin ␤ 1A gene and have laminin binding ability. Both cell lines attach to fibronectin and form focal contacts that contain ␣ v ␤ 3 integrins (43). As shown in Fig. 5, we observed that the migration of both GD25 and GD25-␤ 1A cells toward fibronectin (Fig. 5A) and vitronectin (Fig. 5B), and of GD25-␤ 1A cells toward laminin (Fig. 5A), was stimulated by the addition of EGF (1 nM) to the lower wells. There was no detectable difference between the two cell lines in terms of their migration toward ECM components and their FIG. 2. The synergistic action of EGF in promoting laminininduced migration of Clone B3 fibroblasts. Cell migration toward laminin was determined using a 48-well microchemotaxis chamber as discussed under "Materials and Methods." In the experiments using laminin as the chemoattractant, 200 g/ml laminin was placed in the lower wells. A, EGF is required to be co-positioned with laminin to stimulate chemotaxis. EGF (0 -10 nM) was mixed either with the laminin in the lower wells or with B82L-Clone B3 fibroblasts in the upper wells. Cell motility induced by EGF alone (placed in lower wells) was also measured. B, the direct addition of EGF to the cells reduces the maximal migration conferred by the EGF co-positioned with laminin. In the indicated cases (filled circles), 1 nM EGF was added to the lower wells, whereas the EGF concentration in the upper wells was varied from 0 to 10 nM. The cell migration to laminin alone (square) or in the absence of both EGF and laminin (triangle) was also measured. In all cases, the data points represent the mean Ϯ S.D. of triplicate determinations. Similar results were obtained in at least five different experiments. EGF responses. Moreover, in all other cell lines tested, as long as the cells were able to migrate toward a given ECM component, EGF could further enhance their migration when it was co-positioned with the ECM component, and had no effect when mixed with the cells. Specifically, these cell lines include the 3T3-F442A murine fibroblasts, the human osteosarcoma cell line MG63, the human mammary carcinoma cell line MDA468, and the human epidermal carcinoma cell line A431 (data not shown). The broad spectrum of cell lines that are able to respond to EGF in terms of their migration toward ECM components suggests that the synergistic effect of EGF on chemotaxis is a general phenomenon.
EGF Receptor Expression Is Critical for Fibronectin-induced Cell Migration-To identify EGF receptor elements that are necessary for mediating the stimulatory effects of EGF on integrin-mediated chemotaxis, B82L-Parental cells, which contain no detectable endogenous EGF receptors, were used as a negative control for cells transfected with either wild-type or mutated EGF receptors. B82L-Parental cells have been shown to adhere and spread on fibronectin-coated surfaces (42), indicating that they express functional fibronectin receptors. Although it was predicted that the B82L-Parental cells would migrate toward fibronectin but not show EGF responses due to the lack of EGF receptor expression, we instead unexpectedly observed that these cells exhibited little detectable migration toward fibronectin (Ϯ EGF) (Fig. 6, A and B). In addition, cell migration was not observed, even when the concentrations of fibronectin and EGF were varied from 50 to 200 g/ml and from 0.1 nM to 10 nM, respectively (data not shown). However, the introduction of functional EGF receptors into these cells clearly enabled them to migrate toward fibronectin alone (B82L-wt or Clone B3) ( Fig. 6A; also see Figs. 2 and 3). To evaluate whether these responses were unique to the B82L-wt transfectant and Clone B3 cells, we generated a separate population of transfected B82L cells expressing wild-type EGF receptors, which we termed B82L-wt2. As shown in Fig. 6B, B82L-wt2 cells were also capable of migrating toward fibronectin, albeit at a lower level than that observed with B82L-wt cells. These observations indicate that the introduction of the wild-type EGF receptor into B82L fibroblasts enable them to switch from a non-migratory cell type into a migratory one.
In these studies, it is conceivable that the necessity of the EGF receptor for B82L cells to migrate toward fibronectin alone may reflect an artifact of EGF receptor transfection, e.g.
EGF receptor overexpression may lead to ligand-independent activation of the EGF receptor. However, in our system, we found that the EGF receptor is expressed at ϳ100,000 molecules/cell (42), which is within the physiological range, suggesting that the EGF receptor in B82L cells is not likely to be activated due to the overexpression of the receptor. A second possibility, which may be of physiological significance, is that the cells may secrete an EGF receptor ligand such as EGF or TGF-␣, which in turn stimulates fibronectin-induced migration in an autocrine manner. Although this feature would still strongly support the notion that the EGF receptor is critical for cell migration, it was important to assess if this mechanism was occurring in our system. To examine this possibility, antibodies that can neutralize the most common EGF receptor ligands, EGF and TGF-␣, were used to assess whether they could reduce the migration of Clone B3 cells toward fibronectin. As shown in Fig. 6C, neither anti-EGF nor anti-TGF-␣ antibodies, alone or together, at the concentrations tested, exhibited any effects on the fibronectin-induced migration of Clone B3 cells. These results suggest that the capacity of B82L cells expressing the EGF receptor to migrate toward fibronectin alone does not appear to result from an EGF or TGF-␣ autocrine activation of the EGF receptor.

The EGF Receptor Kinase Activity and C Terminus Are Important for Conferring Fibronectin-induced Chemotaxis-Upon
EGF binding, EGF receptor autophosphorylation is linked to a cascade of signaling events leading to cellular responses such as cell proliferation and cell motility (50). Both the receptor kinase activity and the C terminus, which provides docking sites for SH2 and PTB domain-containing cellular proteins, are critical for EGF-induced signal transduction (1). Chen et al. (17,18) have shown that EGF-elicited random cell movement requires receptor tyrosine kinase activity and autophosphorylation. In order to map EGF receptor regions that are involved in the synergistic stimulation of EGF on integrin-mediated motility, B82L-Parental cells were transfected with EGF receptors that are kinase-inactive (B82L-K721M) or C-terminally truncated (B82L-c'1022). As shown in Fig. 7A, B82L-K721M cells did not migrate toward fibronectin alone, nor did they respond to EGF stimulation. Similar results were obtained with B82L cells expressing EGF receptors that retain only one of the five C-terminal autophosphorylation sites (c'1022) (Fig.  7B), although these cells exhibit increased kinase activity due to the removal of the C-terminal autoinhibitory restraint (51, 52). The observation that the migratory behavior of K721M or c'1022 cells are indistinguishable from the Parental cells suggest that both the receptor tyrosine kinase activity and the C-terminal region are important for the EGF-induced enhancement of cell motility and for the chemotaxis induced by fibronectin alone. In addition, Tyr-992, which has been shown to be sufficient to confer EGF-elicited random cell movement (17), does not appear to be sufficient to support EGF-stimulated chemotaxis, suggesting that EGF regulation of chemokinesis and chemotaxis may involve different signaling mechanisms.
Effects of Neutralizing Anti-EGF Receptor Antibodies on Cell Migration-Our studies have shown that the expression of functional EGF receptors is critical for the induction of motility in B82L fibroblasts. These results suggest that the EGF receptor may be used as a downstream mediator by integrins in transducing signals leading to cell migration. Such ligandindependent activation of the EGF receptor by other receptor systems has been reported for G-protein-coupled-receptor in Rat-1 cells (53). Conversely, the EGF receptor could act upstream of integrins, for example, by regulating integrin expression, activation or coupling to key components involved in cell motility. To further explore this system, we examined the effect of neutralizing anti-EGF receptor antibodies on Clone B3 cell migration. Monoclonal anti-EGF receptor antibodies (LA22, 528, and C225) were used that recognize the EGF binding site, compete for EGF binding, and block EGF-induced receptor autophosphorylation. As shown in Fig. 8 (A and B), neutralizing anti-EGF receptor antibodies inhibited not only EGF stimulation of fibronectin-induced cell migration, but also cell migration induced by fibronectin alone. This effect appeared to be overcome as the EGF concentration was increased. These results support the hypothesis that the EGF receptor may participate, at least in part, as a downstream mediator in the integrin-stimulated signaling pathways leading to cell motility. This concept is consistent with the studies of Klemke et al. (14), where the EGF-selective inhibitor tyrphostin 25 blocks human carcinoma cell migration on vitronectin.
When the effects of anti-EGF receptor antibodies on laminininduced cell migration were examined, we observed that the addition of monoclonal anti-EGF receptor antibodies LA22 and 528 to the cells not only reversed EGF stimulation of laminininduced chemotaxis, but also blocked cell migration toward laminin alone (Fig. 8, C and D). As also shown in Fig. 8D, a relatively weaker inhibition of cell migration was observed FIG. 5. EGF synergistically stimulates the migration of GD25 and GD25-␤ 1A murine fibroblasts toward multiple ECM substrates. Cell migration was determined using a 48-well microchemotaxis chamber as discussed under "Materials and Methods." The migration of GD25 and GD25-␤ 1A fibroblasts toward medium alone, EGF (1 nM) alone, or laminin (100 g/ml) (A), fibronectin (100 g/ml) (B), and vitronectin (100 g/ml) (C) in the absence or presence of EGF (1 nM), in the lower wells (co-positioned with the chemoattractant) was measured. Each bar represents the mean (Ϯ S.D.) of triplicate determinations. Similar results were obtained in three different experiments.
when the anti-EGF receptor antibodies were put into the lower wells with laminin. This may result from the lower effective antibody concentration available to the cells due to diffusion and dilution of the antibodies upon entry into the upper wells.
These results appear to be specific because only anti-EGF receptor antibodies could block cell motility induced by laminin alone whereas nonspecific IgG had no effect (Fig. 8D). Trypan blue exclusion assays showed that all the cells treated with or FIG. 6. Expression of wild-type EGF receptors is required for B82L fibroblasts to migrate toward fibronectin. In panels A and B, B82L-Parental cells expressing no endogenous EGF receptor were transfected on two separate occasions with the wild-type (full-length) EGF receptor designated wt (A) and wt2 (B). Clone B3 cells were used as positive control for fibronectin-induced chemotaxis in response to EGF. The migration of B82L cells toward medium alone, EGF alone (1 nM), fibronectin (100 g/ml) alone, fibronectin (100 g/ml) and EGF (1 nM) co-positioned in the lower wells, or fibronectin in the lower well and EGF in the upper well (co-positioned with the cells), was measured. In panel C, the effects of neutralizing anti-EGF or anti-TGF-␣ antibodies on the fibronectin-induced migration of B82L-Clone B3 cells was determined. Clone B3 cells were incubated with anti-EGF and/or anti-TGF-␣ monoclonal antibodies at the indicated concentrations (legend at right) (for 0.5 h before they were loaded to the migration chamber. Each bar represents the mean Ϯ S.D. of triplicate determinations. Similar results were obtained in two different experiments. without the antibodies exhibited similar viability (data not shown), indicating that the inhibition of laminin-induced cell motility by anti-EGF receptor antibodies does not appear to be due to cell death. DISCUSSION In the present study, we demonstrate that EGF can potently stimulate the motility of mouse B82L fibroblasts that express a full-length EGF receptor. Furthermore, these investigations reveal that the co-positioning of EGF and the chemoattractant is critical for EGF stimulation of cell migration, i.e. it is essential that the cell is exposed to an EGF and substrate gradient oriented in the same direction. This EGF-dependent stimulation of cell motility may in part be due to the cell rounding and the reduction of cell spreading and focal adhesion formation caused by EGF treatment. Another major observation made in these studies is that a fully functional EGF receptor appears critical for integrin-mediated migration, in particular, the EGF receptor C-terminal domain and an intact kinase domain are critical for fibronectin and fibronectin plus EGF-stimulated cell migration. In this regard, it is noteworthy that B82L-Parental cells that do not express the EGF receptor exhibit little fibronectin-induced chemotaxis, whereas the introduction of the wild-type EGF receptor enables these cells to now undergo fibronectin-induced migration. Moreover, the inhibition of fibronectin or laminin-induced chemotaxis by neutralizing anti-EGF receptor antibodies supports the concept that the EGF receptor may act downstream of integrin activation.
Many studies have shown that cell migration is a process that requires dynamic interactions between the cell, the substrate, and the cytoskeleton-associated motile apparatus (24,25). Reduced cell motility can result if cell-substratum interactions conferred by growth factor/cytokine (32,33) or anti-integrin antibody stimulation (28,29), as well as by modulation of the levels of substrates (26,27), integrins (30,31), or cytosketal proteins present at focal contacts are too strong or too weak. These observations are consistent with earlier findings that EGF can induce rapid cell rounding (9), membrane ruffling and retraction (8), and the promotion of extensive cortical actin polymerization and depolymerization of actin stress fibers (10,11). The present studies suggest that an EGF-induced reduction of adhesiveness may augment the motility of B82L fibroblasts expressing functional EGF receptors as well as the mobility of other EGF-responsive cell types.
The synergistic action of EGF and matrix proteins to enhance fibroblast motility suggests that an interaction exists between the signaling pathways triggered by the activation of these two receptor systems. One possible site of convergence could be at the level of the tyrosine kinase Src, which has been reported to modulate EGF-induced mitogenesis (54), actin-cytoskeleton reorganization (via the small molecular weight Gprotein Rho) (55), and integrin-initiated signal transduction. Of note, the transformation of chicken embryo fibroblasts by the Rous sarcoma virus is associated with a general loss of substratum adhesion, changes in cell shape, and the disorganization of the cytoskeleton (56), all of which are similar to those observed in EGF-treated cells. Moreover, the integrin ␤ 1 subunit is tyrosine-phosphorylated in v-Src-transformed cells, and the highly homologous integrin ␤ 3 subunit has been reported to be tyrosine-phosphorylated (57) and to bind the Shc PTB do- FIG. 7. The EGF receptor tyrosine kinase activity and the Cterminal region are important for conferring fibronectin-induced chemotaxis in B82L fibroblasts. B82L-Parental cells expressing no endogenous EGF receptor were transfected with an EGF receptor that carries a lysine to methionine substitution at position 721, which abolishes the receptor kinase activity (K721M) (A), or a kinase active EGF receptor that has been C-terminally truncated from residue 1022 through residue 1186 (c'1022) (B). This mutant lacks four of the major EGF receptor tyrosine autophosphorylation sites. The migration of B82L cells toward medium alone, EGF alone (1 nM), fibronectin (100 g/ml) alone, fibronectin (100 g/ml) and EGF (1 nM) co-positioned in the lower wells, or toward fibronectin placed in the lower well and EGF added to the upper well (co-positioned with the cells), was measured. Each bar represents the mean (Ϯ S.D.) of triplicate determinations. Similar results were obtained in three different experiments. main upon platelet aggregation (58). Tyrosine-phosphorylated ␤ 1 integrin subunit exhibits a decreased ability to bind to both fibronectin and talin (59), implying that the phosphorylated ␤ 1 subunit may be able to escape from focal contacts. The release of ␤ 1 integrins from focal adhesions may result in the disruption of focal contacts, and a reduction of adhesion strength that allows the cell to enter a motile state. In fact, certain Tyr to Phe mutations in the ␤ 1 cytoplasmic domain results in the loss of cell motility (60). Although the ␤ 1 subunit could be a direct substrate of Src, it is also likely that the phosphorylation of ␤ 1 is mediated by other kinases such as focal adhesion kinase (FAK), which has been shown to bind a peptide mimicking the ␤ 1 cytoplasmic domain (61). Additionally, Src and the Src family kinase Fyn are known to form stable complexes with FAK through phosphotyrosine-SH2 domain interactions, thereby bringing these kinases into focal contacts (62)(63)(64). Src can also phosphorylate FAK at additional sites to fully activate its tyrosine kinase activity (65,66). Therefore, as a downstream mediator of both EGF receptor and integrin activation, Src or Src family kinases may integrate signals from both receptor systems, thus leading to events such as the release of phospho-rylated integrins from focal contacts, thereby reducing cell adhesiveness and enhancing cell motility.
The observation that a non-polarized addition of EGF to the cells does not affect cell migration (Fig. 2) suggests that a generalized EGF-induced attenuation of cell adhesion strength is not sufficient to fully explain EGF action. EGF increases cell migration toward matrix proteins only when EGF and the chemoattractant are co-presented to the cells at the same time and in the same direction. The exposure of the cell to EGF and chemoattractant gradients originating from the same direction is likely to augment the cell polarization known to be necessary for directional cell movement (24). When EGF is added directly to the cells, cell polarization can be generated only by the chemoattractant and not by an EGF gradient, and thus EGF does not increase cell migration.
EGF-enhanced cell polarization may occur in several nonexclusive ways. For example, a high local EGF concentration at the leading edge of a migrating cell may facilitate the induction of various active membrane processes. EGF is known to promote F-actin redistribution (10,11) and the formation of lamellipodia and filopodia (8,9), which are processes occurring at the FIG. 8. Neutralizing anti-EGF receptor antibodies inhibit migration of B82L-Clone B3 cells toward fibronectin or laminin. A, cell migration toward fibronectin (50 g/ml) alone, or fibronectin (50 g/ml) together with different doses of EGF (0 -1 nM) was measured. Cells were incubated with anti-EGF receptor monoclonal antibodies LA22 (10 or 100 nM) for 0.5 h before they were loaded to the migration chamber. B, cell migration toward fibronectin alone (0 -200 g/ml) was measured. Cells were incubated with anti-EGF receptor monoclonal antibodies LA22 (30 nM), 528 (30 nM), or C225 (30 nM) for 0.5 h before they were loaded to the migration chamber. C, cell migration toward laminin (200 g/ml) alone, or laminin (200 g/ml) together with different doses of EGF (0 -1 nM) was measured. Cells were incubated with anti-EGF receptor monoclonal antibodies LA22 (10 or 100 nM) for 0.5 h before they were loaded to the migration chamber. D, cell migration toward laminin (200 g/ml) alone, or laminin (200 g/ml) and EGF (1 nM) together was measured. Anti-EGF receptor monoclonal antibodies 528, LA22, or nonspecific IgG at 10 nM were added either to the cells in the upper wells or to the laminin in the lower wells. LA22, 528, and C225 can compete with EGF for the EGF receptor binding and block EGF-induced autophosphorylation of the EGF receptor. Each point represents the mean Ϯ S.D. of triplicate determinations. Similar results were obtained in two different experiments. leading edge of migrating cells (24). In addition, EGF may induce integrin association/interaction with the EGF receptor at the cell leading edge, and thus generate an asymmetric distribution of integrins when the cells are co-exposed to an EGF gradient. Growth factors such as PDGF (58) and insulin (67) have been reported to induce the association of integrins (such as ␣ v ␤ 3 ) with these growth factor receptors. Furthermore, as integrins engaged in cell migration become clustered at the leading edge in the form of macroaggregates (68,69), EGF receptors may co-cluster with the integrins at the cell leading edge when the cells are exposed to a chemoattractant gradient. The idea that the clustering of EGF receptors by activated integrins could be involved is supported by the observation that an accumulation of EGF receptors occurs on fibronectin-coated beads where integrin interaction with fibronectin takes place (45). Moreover, a recent report by Moro et al. (70) indicates that ␤ 1 integrin stimulation can enhance EGF receptor tyrosine phosphorylation. The consequences of this EGF receptor distribution/activation on the leading edge of the cell surface may then be further amplified by the higher concentration of EGF at the cell front when cells are exposed to an EGF gradient in addition to a chemoattractant gradient. The enriched distribution of activated EGF receptors at the leading edge may trigger asymmetric phosphorylation of cellular proteins resulting in an augmentation of the cell polarization induced by chemoattractant alone. This hypothesis is consistent with the observation that tyrosine-phosphorylated proteins are found at the tips of growth cone filopodia and that this process is correlated with the length of the filopodia (71).
Another critical observation of this study is that B82L fibroblasts appear to require an intact EGF receptor to efficiently migrate toward fibronectin alone. B82L-Parental cells contain functional fibronectin receptors in that they are able to adhere to a fibronectin-coated surface (42) , yet they are essentially non-migratory. Interestingly, the introduction of the wild-type EGF receptor into these cells converts them from a non-migratory cell type into a migratory one. Although we cannot rule out the possibility that the transfection of intact EGF receptor may alter the expression of certain proteins that are critical for cell motility, the present investigations support the concept that the EGF receptor can act downstream of integrin activation and serves as a key component of the motility machinery of B82L fibroblasts. In this regard, if the EGF receptor is not directly involved in cell migration, the blockade of EGF receptor function should not abolish cell migration ability, yet anti-EGF receptor antibodies can block cell migration toward fibronectin alone.
Altogether, these observations suggest that the EGF receptor may be activated in an EGF-independent manner upon fibronectin engagement and that the EGF receptor plays a direct role in the migration of these cells. This situation may be analogous to that seen in other systems, i.e. the EGF receptor has been reported to be activated by G-protein-coupled receptors (53) and by UV light (72). Moreover, Miyamoto et al. (45) and Moro et al. (70) have shown an interaction between integrins and growth factors for triggering the tyrosine phosphorylation of the EGF receptor under various conditions. Although this kind of cross-talk between the EGF receptor and integrins may be accomplished by intervening proteins such as Src or FAK (discussed above), it may also occur directly between the two receptors. The two NPXY motifs in the cytoplasmic domain of the integrin ␤ 1 subunit are homologous to the tyrosine autophosphorylation sites on the EGF receptor (1, 34), indicating that the ␤ 1 integrin and the EGF receptor may interact directly or may be modulated by a common system. Similarly, a physical association between integrin ␣ v ␤ 3 and growth factor receptors such as the PDGF receptor and the insulin receptor substrate (IRS-1) upon PDGF or insulin treatment (58,67,73) has been reported. The present study provides evidence for an EGF receptor-integrin system interaction, which is consistent with earlier observations that the EGF receptor interacts directly with ␤-catenin, a vinculin homologous protein involved in cell-cell adhesion (74), and that EGF receptors are accumulated and potentially activated by integrins (45,70).
Inactivation of the EGF receptor kinase activity by replacing Lys-721 with Met reduces EGF-induced tyrosine phosphorylation of cellular proteins, including phospholipase C-␥, and inhibits EGF-induced receptor internalization. Although cells expressing this tyrosine kinase negative EGF receptor still bind EGF with the same affinity, and exhibit EGF-induced MAP kinase activation (75), EGF-stimulated cell proliferation is blocked (76). Our study indicates that an intact kinase domain appears essential for cell migration as well. The inability of this kinase-inactive mutant to mediate cell motility may reside in the transient nature of EGF-induced MAP kinase activation (75), which has been shown to be important for focal adhesion disassembly (77), and/or in the blockage of activation of phospholipase C-␥, which has been shown to be important for mediating random cell movement (17,18) and haptotaxis (14). Unlike the kinase-negative mutant, the c'1022 mutant EGF receptor mediates enhanced tyrosine phosphorylation of multiple cellular substrates (39), it undergoes normal receptor internalization and Ca 2ϩ mobilization (78), and it can be tyrosine-phosphorylated (39) upon EGF treatment. The observation that c'1022-expressing cells cannot migrate suggest that an SH2-or PTB domain-containing protein that binds to the C-terminal autophosphorylation sites is involved in mediating signals in the motility pathway, or that the altered substrate specificity of this mutant may modify the activity of key effectors that regulate the motility pathway.
One additional issue that is related to the present work is that the inhibition of fibronectin-and laminin-induced chemotaxis by function-blocking anti-EGF receptor antibodies could arise from EGF contamination of the ECM preparations, i.e. EGF accounts for part of the ECM-induced chemotaxis. Although this scenario would still strongly support a key role for the EGF receptor system in B82L cell migration, the argument of EGF contamination is not supported by the observations that EGF alone, at all the concentrations tested, induced little chemotaxis, that EGF was not detected in the laminin preparations as assessed by immunoblotting using anti-EGF and TGF-␣ antibodies, and that anti-EGF or TGF-␣ antibodies do not alter fibronectin-induced chemotaxis. A related issue is that because laminin contains EGF-like motifs and it may bind EGF receptor to induce cell migration, and that the anti-EGF receptor antibodies may block laminin-induced chemotaxis by abolishing laminin binding to the EGF receptor. However, this is not likely because intact laminin (79) and a peptide encompassing the EGF-like fragment of laminin (80) cannot compete with EGF receptor binding. Thus, the data are most indicative of a system wherein anti-EGF receptor antibodies eliminate laminin-induced chemotaxis by binding to the EGF receptor and antagonizing the receptor dimerization/activation, thereby suggesting that EGF receptor function is directly important for ECM-induced chemotaxis.
In summary, EGF can exert a synergistic effect on integrinmediated chemotaxis, and this process requires the co-exposure of the cell to both an EGF and an ECM substrate gradient. We also provide evidence to support the concept that the EGF receptor can act downstream of integrin activation and may play a direct role in cell migration toward matrix proteins.
These events are not only likely to be critical for both normal growth and development (e.g. embryogenesis, organ development, and wound healing), but also may be critical for some of the pathobiological actions of EGF, especially given that the aberrant expression of the EGF receptor, and the other members of the erbB receptor family, have been implicated in cancer progression. In the latter case, whereas tumor progression may result from the strong mitogenic effects ascribed to EGF, our study suggests that the chemotactic effects of EGF could contribute to tumor cell metastasis. Given that integrins appear to transmodulate the EGF receptor, abnormal EGF receptor expression in various tumor cells may contribute to their metastatic potential.