|Home | About | Journals | Submit | Contact Us | Français|
ADP-ribosylation factor (Arf) 6 activity is crucially involved in the regulation of E-cadherin–based cell–cell adhesions. Erythropoietin-producing hepatocellular carcinoma (Eph)-family receptors recognize ligands, namely, ephrins, anchored to the membrane of apposing cells, and they mediate cell–cell contact-dependent events. Here, we found that Arf6 activity is down-regulated in Madin-Darby canine kidney cells, which is dependent on cell density and calcium ion concentration, and we provide evidence of a novel signaling pathway by which ligand-activated EphA2 suppresses Arf6 activity. This EphA2-mediated suppression of Arf6 activity was linked to the induction of cell compaction and polarization, but it was independent of the down-regulation of extracellular signal-regulated kinase 1/2 kinase activity. We show that G protein-coupled receptor kinase-interacting protein (Git) 1 and noncatalytic region of tyrosine kinase (Nck) 1 are involved in this pathway, in which ligand-activated EphA2, via its phosphorylated Tyr594, binds to the Src homology 2 domain of Nck1, and then via its Src homology 3 domain binds to the synaptic localizing domain of Git1 to suppress Arf6 activity. We propose a positive feedback loop in which E-cadherin–based cell–cell contacts enhance EphA-ephrinA signaling, which in turn down-regulates Arf6 activity to enhance E-cadherin–based cell–cell contacts as well as the apical-basal polarization of epithelial cells.
E-cadherin–mediated cell–cell adhesion is essential for the integrity of epithelial cell layers, as well as their normal functions (Takeichi, 1991 ; Gumbiner, 2005 ). ADP-ribosylation factor (Arf) 6 primarily regulates recycling of plasma membrane components, as well as remodeling of the membrane and actin cytoskeleton at the cell peripheries via its GTPase cycle (Donaldson, 2003 ). It has been shown that expression of an inactive form of Arf6, Arf6T27N, blocks hepatocyte growth factor (HGF)-induced internalization of E-cadherin–based junctional components in Madin-Darby canine kidney (MDCK) epithelial cells, whereas expression of a constitutively active form of Arf6, Arf6Q67L, causes disassembly of adherens junctions (Palacios et al., 2001 , 2002 ). The results of small interfering RNA (siRNA)-mediated knockdown of GEP100, a guanine nucleotide exchanging factor for Arf6, also support a similar notion that inactivation of Arf6 renders resistance to HGF-induced disruption of adherens junctions of human epidermoid carcinoma CaSki cells (Hiroi et al., 2006 ). Experiments using fetal hepatocyte cells prepared from Arf6−/− mice also support this notion (Suzuki et al., 2006 ). Moreover, we have shown that activation of Arf6 by GEP100 plays a pivotal role in the invasive phenotypes of different breast cancer cells, which is accompanied by the disruption of E-cadherin–mediated cell–cell adhesion in breast cancer MCF7 cells (Morishige et al., 2008 ). Therefore, down-regulation of Arf6 activity seems to be pivotal for the formation and maintenance of E-cadherin–mediated cell–cell adhesions. However, the mechanism by which Arf6 activity is suppressed in cell–cell contacts has not yet been identified.
Eph receptor tyrosine kinases are classified into either EphA or EphB subfamilies based on the identity of their ligands, ephrinA- and ephrinB-subfamily members, respectively, and both are anchored to the cell membrane by different mechanisms (Pasquale, 2005 ). Eph receptors and ephrins each have overlapping specificity: several receptors can bind to one ligand, and, in turn, several ligands can bind to one receptor (Pasquale, 2005 ). The physiological roles of Eph–ephrin interactions have been well characterized in the nervous system, such as during axon guidance and synapse formation, and also in somite and vascular development (Pasquale, 2005 ). Eph receptors, when ligand-activated, become tyrosine phosphorylated and evoke a variety of different intracellular signaling cascades, which mostly exert negative regulatory effects, such as on migration and proliferative signaling (Miao et al., 2000 , 2001 , 2003 ; Noren et al., 2006 ).
EphA2 is normally expressed at high levels in adult epithelial cells, as well as in restricted regions of the embryo during early development (Ruiz and Robertson, 1994 ; Surawska et al., 2004 ). EphA2 is frequently overexpressed in different types of human carcinomas, including those of the breast, lung, prostate, esophagus, and kidney (Surawska et al., 2004 ). The EphA2 gene is a direct transcription target of the Ras/Raf/mitogen-activated protein kinase kinase (Mek)/extracellular signal-regulated kinase (Erk)1/2 pathway (Macrae et al., 2005 ). Overexpression of EphA2 in some carcinomas may hence merely be a result of the activation of this pathway in carcinomas rather than an etiologic event (Macrae et al., 2005 ). In contrast, it has been demonstrated that stimulation of overexpressed EphA2 in some tumor cells by high concentrations of exogenous ligands can negatively regulate the growth, survival, migration, and invasion of these cells (Zelinski et al., 2001 ; Noblitt et al., 2004 ). Moreover, disruption of the EphA2 gene in mice leads to increased susceptibility to skin carcinogenesis (Guo et al., 2006 ), suggesting a tumor suppressing role of EphA2.
Here, we found that EphA2, when ligand activated, suppresses Arf6 activity. We show that EphA2 uses G protein-coupled receptor kinase-interacting protein (Git) 1 to suppress Arf6 activities. We describe the precise mechanism by which EphA2 is linked to Git1, and we demonstrate that this pathway acts to enhance E-cadherin–based cell–cell adhesions and the apical-basal polarization of epithelial cells.
MDCK cells, obtained from Sh. Tsukita (Kyoto University), and human embryonic kidney (HEK) 293T cells were cultured as described previously (Mazaki et al., 2001 ). Fetal calf serum was purchased from HyClone Laboratories (Logan, UT).
For the culture of MDCK cells under the “sparse density” and the “dense density,” 1 × 106 and 5 × 106 cells, respectively, were seeded onto a Φ 90-mm plastic dish and cultured for a further 24 h before subjecting to analysis.
For ephrinA1 stimulation, cells were treated with nonclustered ephrinA1-Fc (R&D Systems, Minneapolis, MN) at 125 ng/ml or control Fc (R&D Systems) at 62.5 ng/ml for 30 min, unless otherwise indicated.
The rabbit polyclonal antibody against Git2 was generated as described previously (Mazaki et al., 2001 ). Other antibodies were purchased from commercial sources: rat monoclonal antibodies against E-cadherin (clone ECCD-2, Takara, Kyoto, Japan; and clone DECMA-1, Sigma-Aldrich, St. Louis, MO); rabbit polyclonal antibody against Git1 (H-170; Santa Cruz Biotechnology, Santa Cruz, CA); EphA2 (C-20; Santa Cruz Biotechnology); zona occludens (ZO)-1 (Zymed Laboratories, South San Francisco, CA); Erk1/2 (Cell Signaling Technology, Danvers, MA); FLAG (Sigma-Aldrich); mouse monoclonal antibody (mAb) against Arf6 (3A-1; Santa Cruz Biotechnology); noncatalytic region of tyrosine kinase (Nck) (BD Biosciences, San Jose, CA); Ezrin (EZ-1; Biodesign International, Kennebunk, ME); phospho-Erk1/2 (Cell Signaling Technology), hemagglutinin (HA) (16B12; BAbCo, Richmond, CA); glutathione transferase (GST) (Millipore, Billerica, MA); FLAG (M2; Sigma-Aldrich); nonspecific rabbit and mouse immunoglobulin G (IgG) (Sigma-Aldrich); horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse IgG; F(ab′)2 fragments of biotin-conjugated goat anti-rabbit or anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA); and Alexa-labeled goat anti-rabbit, anti-mouse, or anti-rat IgG (Invitrogen, Carlsbad, CA). All other chemical reagents were purchased from Sigma-Aldrich and Wako Pure Chemicals (Kyoto, Japan), unless otherwise stated.
cDNA encoding mouse Git1 was amplified by polymerase chain reaction (PCR) from mouse brain first-strand cDNA. Other cDNAs were provided by the following researchers: human EphA2 was from N. Mochizuki (National Cardiovascular Center, Osaka, Japan), mouse E-cadherin was from M. Takeichi (RIKEN CDB, Kobe, Japan); Venus was from A. Miyawaki (RIKEN BSI, Wako, Japan); mouse Arf1 and Arf6 were from K. Nakayama (Kyoto University), and mouse Nck1 was from T. Shishido (NAIST, Nara, Japan). pcDNA3 FLAG N dest, pcDNA3 HA N dest., and pEBG dest vectors were generated using the Gateway vector conversion kit (Invitrogen). EphA2, Git1, and Nck1 cDNAs were inserted into the pENTR/D topo vector and transferred into their destination vectors by using LR clonaseII (Invitrogen). pBabe puro Arf6-HA, Arf6T27N-HA, Arf6Q67L-HA, and Arf1Q71L-HA were generated by cloning HindIII-XbaI fragments from pcDNA3 Arf6-HA, Arf6T27N-HA, Arf6Q67L-HA, and Arf1Q71L-HA into the SnaBI site of the pBabe puro vector after filling of the ends. pVenus N1 E-cadherin was described previously (Bauer et al., 2008 ).
cDNA transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.
To establish cells stably expressing specific cDNAs, transfection-positive cells were selected by culturing in the presence of the appropriate drugs (4 μg/ml puromycin for the pBabe puro vector or 1 mg/ml G418 for the pcDNA3 and pVenus N1 vectors). Cells were then cloned by limited dilution, and the results were confirmed by at least two independent cell clones for each plasmid.
MDCK cells were transfected with oligonucleotide duplexes by using a reverse transfection method, according to the manufacturer's instructions (Invitrogen). Briefly, cells were trypsinized, washed, and suspended in DMEM with 10% fetal calf serum. Then, they were plated onto a plastic dish in the presence of 10 nM oligonucleotide duplexes and Lipofectamine 2000 in Opti-MEM (Invitrogen), Finally, they were incubated for 48 h before being subjected to analysis. Nucleotide sequences used were as follows: for Git1, sense 5′-GAGGUGGAUAGAAGAGAAAAU-3′; and for Nck1, sense 5′-UCCUGGUGGCGAGUUCGAA-3′. An siRNA duplex with an irrelevant sequence (5′-GCGCGCUUUGUAGGAUUCG-3′) was purchased from Dharmacon RNA Technologies (Lafayette, CO).
For the rescue experiment of the Git1 siRNA treatment, mouse Git1 cDNA, tagged with FLAG, in which the siRNA target sequence was mutated into 5′-GAAGTGGATCGGCGGGAGAAC-3′ was used.
Calcium switch was performed according to the method described previously (Zantek et al., 1999 ). Briefly, 8 mM EGTA was added to MDCK cells cultured at the dense density, and then cells were incubated for a further 30 min at 37°C. The medium was then exchanged with DMEM supplemented with 10% fetal calf serum and 4 mM CaCl2, and cells were further incubated for the times indicated.
Arf6 activities were measured using GST-GGA (Santy and Casanova, 2001 ). Cells were lysed in a lysis buffer (1% Triton X-100, 0.05% sodium cholate, 0.005% SDS, 50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 10 mM MgCl2, 10% glycerol, 2 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 5 μg/ml aprotinin, 2 μg/ml leupeptin, and 3 μg/ml pepstatin A). After clarifying by centrifugation at 1500 × g for 5 min, supernatants were incubated with GST-GGA3 conjugated to glutathione-Sepharose for 40 min at 4°C.
Immunoprecipitation assays were performed using antibodies coupled with protein G-Sepharose, as described previously (Morishige et al., 2008 ), in which cells were lysed in NP-40 buffer (1% Nonidet P-40, 20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1 mM Na3VO4, 1 mM phenylmethylsulfonyl fluoride, 5 μg/ml aprotinin, 2 μg/ml leupeptin, and 3 μg/ml pepstatin A). When cells were pretreated with ephrinA1-Fc or control Fc, biotin-conjugated F(ab′)2 fragments of goat anti-rabbit or anti-mouse IgG, coupled to streptavidin-Sepharose beads, were used. Three hundred micrograms of cell lysates was used for each GGA pulldown and immunoprecipitation.
Immunoblotting was performed as described previously (Morishige et al., 2008 ). To enhance the signal from the endogenous Arf6 protein, Can Get Signal (Toyobo Engineering, Osaka, Japan) was used to dilute the anti-Arf6 antibody.
Acquisition of confocal images of cells was performed using confocal laser scanning microscopes (LSM510, Carl Zeiss, Jean, Germany and FV1000, Olympus, Tokyo, Japan), as described previously (Mazaki et al., 2001 ). Z-sections were obtained at 0.5-μm step size. Focuses adjusted across the center of the majority of cell bodies were used to show the localization of E-cadherin, EphA2, Git1-FLAG, Arf-HA, and F-actin. To show the localization of ZO-1, projection images made by summing up all confocal sections into one image were used. Each figure of microscopic analysis shows representative results observed in at least three independent experiments.
We first found that the cellular activities of Arf6, measured by the GST-GGA pulldown method (Santy and Casanova, 2001 ), are suppressed in MDCK cells cultured under a dense density as compared with a sparse density, whereas Arf6 protein levels do not notably differ between these two cell densities (Figure 1A). Suppression of Arf6 activity has been implicated in inhibition of the internalization of E-cadherin from cell–cell contact areas, as well as stabilization of the cell–cell contacts (Palacios et al., 2001 , 2002 ). Consistently, E-cadherin seemed to be more densely accumulated at cell–cell contact areas in cells cultured under the dense density than cultured under the sparse density, and this accumulation was accompanied by substantial reduction of the amount of E-cadherin molecules in the cytoplasm and in the apical areas of the cell surface (Figure 1B). Moreover, observation from the z-axis revealed that E-cadherin is more clearly segregated from a tight junction protein, ZO-1, as well as from an apical marker protein, Ezrin, in the dense cells than in the sparse cells (Figure 1B and Supplemental Figure S1), indicating enhanced apical-to-basal polarization of the dense cells compared with the sparse cells.
To obtain clues as to how E-cadherin–based cell–cell adhesions are involved in the suppression of Arf6 activity, we used the calcium switch assay (Zantek et al., 1999 ), in which calcium ions are first deprived from the culture medium by the addition of EGTA and then added again. We found that Arf6 is swiftly activated upon EGTA treatment and then gradually down-regulated after readdition of calcium (Figure 1C). Activities of Erk1/2 have been shown to be increased by perturbation of E-cadherin–mediated cell–cell contacts in intestinal epithelial cells (Laprise et al., 2004 ). We confirmed that activities of Erk1/2, as assessed by their phosphorylation, are also increased upon EGTA treatment and then gradually decreased after readdition of calcium in these MDCK cells (Figure 1C; also see below). Likewise, activities of Erk1/2 were suppressed in MDCK cells cultured under a dense density compared with a sparse density (Figure 1A).
Among the EphA receptors, MDCK cells were found to predominantly express EphA2 (Supplemental Figure S2). We found that tyrosine phosphorylation of EphA2 is dramatically increased in MDCK cells cultured under the dense density compared with the sparse density (Figure 2A). In human mammary epithelial MCF-10A cells, tyrosine phosphorylation of EphA2 has been shown to be dependent on calcium ions in the culture medium (Zantek et al., 1999 ). Consistently, tyrosine phosphorylation of EphA2 in MDCK cells, which was induced by culturing cells under the dense density, was substantially diminished upon EGTA treatment and was gradually recovered after readdition of calcium (Figure 2B). MDCK cells also express ephrinA1, and, to a lesser extent, ephrinA4 (Supplemental Figure S2), and the above-described enhanced activation of EphA2 under the dense cell density is likely to be due to paracrine stimulation of EphA2 by its ligands upon cell–cell contacting, which might occur more efficiently at the dense cell density than at the sparse cell density. In contrast, ephrins themselves are also known to transduce intracellular signals, upon binding to Eph receptors (Pasquale, 2005 ). We found that stimulation of MDCK cells, cultured under the sparse density, with ephrinA1 fused to the Fc portion of human immunoglobulin G (ephrinA1-Fc) down-regulates Arf6 activity, whereas a control Fc fragment or EphA2-Fc did not (Figure 2C). Dense tyrosine phosphorylation of EphA2 upon ephrinA1-Fc stimulation was confirmed under the same condition (data not shown). This ephrinA1-Fc stimulation of MDCK cells also induced cell compaction, which was accompanied by the enhanced accumulation of E-cadherin to cell–cell contact areas (Figure 2, D and E, Supplemental Figure S3 and Supplemental Video S1). Observation from the z-axis revealed that E-cadherin is more clearly segregated from ZO-1 and Ezrin in ephrinA1-Fc–treated cells than in control Fc-treated cells, indicating that these ephrinA1–Fc-treated cells were more polarized in the apical-to-basal direction, as seen with cells cultured under the dense density (Figure 2D and Supplemental Figure S1). Therefore, ephrinA1-stimulated MDCK cells resemble those cells cultured under the dense density, in their EphA2 phosphorylation, Arf6 activities, morphology and apical-to-basal polarization. We also found that treatment of MDCK cells with an anti-E-cadherin antibody, DECMA-1, which blocks E-cadherin function (Vestweber and Kemler, 1985 ), inhibits ephrinA1–Fc-mediated cell compaction (Figure 2F), suggesting the involvement of homophilic E-cadherin adhesion in this event.
To assess whether ephrinA1–Fc-induced cell compaction is related to Arf6 activities, we next established MDCK cell lines stably expressing HA-tagged Arf6 and its mutants (Figure 3A). Expression of a guanosine triphosphate (GTP) hydrolysis-defective form of Arf6-HA, Arf6Q67L-HA, but not wild-type Arf6-HA, caused a spread out cell morphology, which was resistant to ephrinA1–Fc-mediated cell compaction (Figure 3, B and C). Such resistance to ephrinA1-Fc was not observed with a similar type of Arf1 mutant, Arf1Q71L-HA (Figure 3, B and C). In contrast, expression of a GTP binding-defective form of Arf6-HA, Arf6T27N-HA, induced cell compaction (Figure 3, A–C). However, expression of Arf6T27N-HA was very low compared with wild-type Arf6-HA and Arf6Q67L-HA (Figure 3A), and these cells expressing Arf6T27N-HA could respond to ephrinA1-Fc to become more compacted (Figure 3, B and C). These results are consistent with the notion that down-regulation of Arf6 activity plays an integral role in the pathway by which ligand-activated EphA2 induces cell compaction.
We then sought to clarify the mechanism by which ligand-activated EphA2 down-regulates Arf6 activity. EphA2 signaling pathways have already been extensively characterized (Pasquale, 2005 ). However, none of the known signaling pathways can adequately explain the suppression of Arf6 by EphA2. We examined whether some GTPase activation proteins (GAPs) for Arf GTPases are linked to EphA2. Git1 exhibits efficient GAP activity against different Arf isoforms, including Arf6 (Premont et al., 1998 ; Meyer et al., 2006 ). We found that a significant amount of Git1 is coimmunoprecipitated with EphA2, when MDCK cells are stimulated by ephrinA1-Fc (Figure 4A). Conversely, an anti-Git1 antibody also coimmunoprecipitated significant amounts of EphA2, which was dependent on ephrinA1-Fc stimulation (Figure 4B). Git1 has a structurally conserved isoform, namely, Git2 (Hoefen and Berk, 2006 ). Git2 did not coprecipitate with EphA2 (Figure 4A). Consistent with these results, siRNA-mediated knockdown of Git1 substantially abolished ephrinA1–Fc-mediated suppression of Arf6 activity, as well as ephrinA1–Fc-mediated cell compaction and apical-to-basal polarization (Figure 4, C–F and Supplemental Figure S4). To confirm the specificity of the Git1 siRNA, we then established MDCK cells expressing a rescue construct of Git1-FLAG, in which the Git1 siRNA-target nucleotides were mutated (Figure 4C), and we confirmed that ephrinA1–Fc-mediated suppression of Arf6 activity, as well as ephrinA1–Fc-mediated cell compaction were still observed in these “rescued” cells, even when the cells were treated with the Git1 siRNA (Figure 4, D–F). These results suggest that EphA2, when ligand activated, uses Git1 to suppress Arf6 activity and that this engagement of Git1 is important for the EphA2-mediated modulation of cell–cell contacts.
Ligand-activated EphA2 is known to attenuate the activities of mitogen-activated protein kinases (Miao et al., 2001 ). Moreover, the Arf6 GTPase cycle has been implicated in the activation of Erk kinases (Tague et al., 2004 ). We however found that knockdown of Git1 does not affect ephrinA1–Fc-induced suppression of Erk1/2 activities (Figure 4D). Therefore, the EphA2–Git1 pathway down-regulating Arf6 activity is substantially independent of the EphA2 pathway down-regulating Erk1/2 activity in MDCK cells.
Git1 does not seem to have protein interaction modules that can directly bind to ligand activated EphA2. In contrast, Git1 has been shown to bind to several adaptor proteins, such as Nck (Frese et al., 2006 ). Among the Nck isoforms, MDCK cells predominantly express Nck1 (Supplemental Figure S1). We found that Nck1 is readily coprecipitated with Git1 in MDCK cells even without ephrinA1-Fc stimulation (Figure 5A). In contrast, Nck1 coprecipitated with EphA2 only after ephrinA1-Fc stimulation of these cells (Figure 5A). Coprecipitation of Git1 with EphA2 was abolished upon siRNA-mediated knockdown of Nck1 (Figure 5, B and C). Similar to the Git1 knockdown, knockdown of Nck1 also abolished ephrinA1-Fc-mediated suppression of Arf6 activity (Figure 5D), as well as ephrinA1–Fc-mediated cell compaction and apical-to-basal polarization (Figure 5, E and F, and Supplemental Figure S4). Therefore, Nck1 seems to link Git1 with ligand-activated EphA2.
The above-mentioned results indicate that the association of Nck1 and EphA2 occurs only after ephrinA1 stimulation, whereas that of Nck1 and Git1 occurs constitutively. We then investigated the precise mechanisms involved in the protein interactions of this EphA2–Nck1–Git1 complex. Because EphA2 makes a complex with Git1 upon ephrinA1-Fc stimulation, we first examined the possible involvement of ligand-induced tyrosine phosphorylation of EphA2 in this complex formation. Phosphorylation of highly conserved tyrosine residues on the juxtamembrane region of Eph receptors, which correspond to Tyr588 and Tyr594 of EphA2, has been shown to be the major Nck binding sites (Kullander and Klein, 2002 ). We expressed Git1-FLAG in HEK293T cells together with HA-tagged EphA2 (EphA2-HA) or its tyrosine nonphosphorylation mutants Y588F and Y594F, in which Tyr588 and Tyr594 are mutated into phenylalanine, respectively (Figure 6A). HEK293T cells express Nck endogenously (Figure 6B). Moreover, EphA2 overexpressed in HEK293T cells was notably tyrosine-phosphorylated even without exogenous stimulation by ephrinA1-Fc (data not shown), similar to those observed with other Eph receptors overexpressed in HEK293 cells (Becker et al., 2000 ). We found that mutation of Tyr594, but not Tyr588, substantially abolishes coprecipitation of EphA2-HA with Git1-FLAG (Figure 6B). We then examined whether phosphorylation of Tyr594 is necessary for the association of EphA2 with Nck1. Nck1 has one Src homology 2 (SH2) domain, a domain known to bind to phosphotyrosines. We found that a GST fusion form of the SH2 domain of Nck1 (GST-Nck1 SH2) binds to wild-type EphA2-HA and its Y588F mutant (EphA2 Y588F-HA) but not the Y594F mutant (EphA2 Y594F-HA) (Figure 6C).
We next investigated which portion(s) of Git1 is necessary for its complex formation with EphA2. It has been reported that Tyr392 of Git1 can be phosphorylated and that this phosphorylation enables its binding to the SH2 domain of Nck (Frese et al., 2006 ). We however found that mutation of Tyr392 into phenylalanine (Git1Y392F-FLAG) does not affect its complex formation with EphA2-HA (Figure 6D). In contrast, we showed above that EphA2 does not associate with Git2 (Figure 6D). The synaptic localizing domain, present in Git1, is not well conserved in Git2 (Supplemental Figure S5). We found that the synaptic localizing domain-deletion mutant of Git1 (Git1ΔSLD-FLAG) no longer associates with EphA2-HA (Figure 6D). Consistently, the synaptic localizing domain of Git1 alone (Git1 SLD-FLAG) was able to associate with EphA2-HA (Figure 6D). The synaptic localizing domain of Git1 contains several repeats of proline-rich sequences at its C terminus, which apparently conform to Src homology 3 (SH3) binding motifs. Nck1 contains three tandem repeats of SH3 domains. Deletion of these proline-rich regions from the synaptic localizing domain of Git1 (Git1 SLDΔC-FLAG) abolished its complex formation with EphA2-HA (Figure 6D). We also confirmed that the SH3 domains of Nck1, fused to GST (GST-Nck1 3xSH3), bind to Git1 SLD-FLAG but not Git1 SLDΔC-FLAG (Figure 6E). Together, it is most likely that in the EphA2–Nck1–Git1 complex, the phosphorylated Tyr594 of EphA2 binds to the SH2 domain of Nck1, and the SH3 domains of Nck1 bind to the synaptic localizing domain of Git1. Our results also indicate that Tyr392 of Git1 is dispensable for its association with EphA2. Tyr392 is conserved in Git2, whereas the synaptic localizing domain is not. Lack of the synaptic localizing domain in Git2 may explain why Git2 is not directly engaged by EphA2.
We finally examined the possible colocalization of Git1, and also its mutants, with EphA2. Because immunostaining of endogenous Git1 in MDCK cells was very weak, we stably expressed Git1-FLAG and its mutants in MDCK cells and visualized them by use of an anti-FLAG antibody (Figure 7A). EphA2 is known to localize to cell–cell contact areas (Zantek et al., 1999 ). We found that although only a small amount of Git1-FLAG is localized to cell–cell contact areas when cells are cultured under the sparse density, a significant fraction is swiftly recruited to cell–cell contact areas upon ephrinA1-Fc stimulation and colocalizes well with EphA2 (Figure 7B). The synaptic localizing domain of Git1 was previously identified as being responsible for its localization to neuronal synapses (Zhang et al., 2003 ). In MDCK cells, however, we found that Git1ΔSLD-FLAG, which lacks the synaptic localizing domain, is still able to localize to cell–cell contacts, although with a lower efficiency than that of intact Git1-FLAG (Figure 7, B and C). In contrast, expression of the synaptic localizing domain alone (Git1 SLD-FLAG) blocked ephrinA1–Fc-mediated cell compaction, whereas this domain alone seemed to be very inefficient in localizing to the cell–cell contacts (Figure 7, B and C). These results again suggest the importance of the synaptic localizing domain of Git1 in Eph signaling, although this domain per se is not essential for its localization to cell–cell contacts.
EphA receptors have been implicated in the inhibition of cell migration and growth. To accomplish such roles, several EphA receptors have been shown to employ signaling molecules that down-regulate the activities of Ras and Rac small GTPases, such as p120RasGAP and α-chimaerin (Pasquale, 2005 , 2008 ). In this paper, we show that EphA2 has a novel signaling pathway that downregulates the activity of another small GTPase, Arf6, by recruiting Git1. Through this pathway, we show that the ligand-activation of EphA2, which may occur during cell-cell contacting of epithelial cells, contributes to the cell compaction as well as the maturation of E-cadherin-based cell-cell adhesion and apical-basal polarization. Our results also show that E-cadherin function is necessary for the EphA2-mediated down-regulation of the Arf6 activity during cell-cell contacting. Therefore, it is plausible to assume that a positive feedback loop exists between EphA2 and E-cadherin in some normal epithelial cells, in which E-cadherin-based cell-cell contacts enhance binding of EphA2 receptors with their ligands, which in turn activates the signaling pathway at the cell-cell contacting areas to down-regulate Arf6 activity, and this down-regulation then acts to enhance E-cadherin-based cell-cell adhesions as well as maturation of apical-basal polarization.
Git1 has been shown to localize and function at cell–cell junctions formed in neuronal synapses (Zhang et al., 2003 ) and immune synapses (Phee et al., 2005 ). However, these studies have mostly dealt with Git1 as a scaffold protein, which links cell surface receptors to the Pix–Pak pathway to remodel actin cytoskeletal architecture and stabilize these cell–cell junctions. In contrast, our present study clarifies the role of Git1 as a GTPase-activating protein for Arf6 in the stabilization of cell–cell junctions.
We show that EphA2 signaling causes enhanced polarization in the apical-to-basal direction, which is accompanied by clear segregation of ZO-1 and Ezrin from E-cadherin. Interestingly, suppression of Git1 and Nck1 expression also affects the subcellular localization of ZO-1. Therefore, activation of the EphA2–Nck1–Git1 signaling pathway seems to lead to maturation of tight junction structures, together with the maturation of adherens junctions. In contrast, it has been shown that an active form of the Arf6 mutant does not perturb localization of ZO-1 at cell–cell contacts in MDCK cells under a condition in which this mutant disrupts E-cadherin–based adherens junctions (Palacios et al., 2001 ). It will be interesting to investigate whether the EphA2–Nck1–Git1 pathway, as well as the GAP activity of Git1, is directly involved also in the maturation of tight junctions. Moreover, Eph receptor signaling also regulates activities of other small GTPases, including Rac1, RhoA, and Rap1 (Pasquale, 2005 ), which may also be involved in modulating the architecture of cell–cell adhesions. In particular, it has been shown that Rac1 activities can be tightly coupled with activities of Arf6 (Radhakrishna et al., 1999 ). It will be interesting to examine whether EphA2 signaling, as well as cell densities affect activities of other small GTPases.
Penela et al. (2008) have reported that Git1 plays a role in enhancing cell migration and proliferation of HeLa cells and COS7 cells, by interacting with G protein-coupled receptor kinase 2. These cells are completely transformed; moreover, HeLa cells have lost the expression of E-cadherin. Their report, together with our results suggest that Git1 may be pleiotropic, having different roles in different signaling pathways or cellular contexts. It will be interesting to examine whether Git1 also acts as a GAP for Arf6 in G protein-coupled receptor signaling or just acts as a scaffold protein in this signaling pathway.
The synaptic localizing domain of Git1 contains both typical and atypical PXXP motifs at its C terminus and has been predicted to bind to some yet unidentified SH3 domains (Bagrodia et al., 1999 ). We show that the Nck1 SH3 domain binds to this synaptic localizing domain, an interaction which requires the C-terminal proline-rich region of this synaptic localizing domain. Recently, it was reported that Git1 forms a closed conformation by intramolecular interaction between its N- and C-terminal regions. When the synaptic localizing domain or the Ankyrin repeat is deleted, Git1 seems to change to an open conformation and shows increased affinities to paxillin and liprin α (Totaro et al., 2007 ). We observed that the synaptic localizing domain of Git1 alone shows higher affinity to the Nck SH3 domain than full-length Git1 (data not shown). Therefore, it will be interesting to examine whether Eph signals have a role in changing Git1 to an open conformation and hence up-regulating its binding to other proteins. It will also be interesting to examine whether Eph signals up-regulate the GAP activity of Git1.
Disruption of E-cadherin–mediated cell–cell adhesions is the major cause for the acquisition of invasive and metastatic properties in most types of carcinomas (Takeichi, 1991 ). The EphA2 gene has been implicated in tumor suppression, as mentioned. It will thus be interesting to clarify the precise molecular mechanism by which Arf6 activity, which is under the regulation of the EphA2–Nck–Git1 signaling pathway, participates in the processes maintaining E-cadherin–mediated cell–cell adhesions, especially those regulating the cellular dynamics and fates of E-cadherin molecules. Moreover, loss of EphB receptors have also been highly implicated in the tumorigenesis of different types of cancers (Pasquale, 2008 ), and EphB signaling has also been shown to enhance cell–cell adhesion by recruiting E-cadherin to the plasma membrane (Cortina et al., 2007 ). Furthermore, EphB receptors possess Nck SH2 binding sites (Kullander and Klein, 2002 ). It will thus be interesting to examine whether EphB receptors also use a similar pathway involving the suppression of Arf6 activity.
We are grateful to T. Yoneda, Y. Shibata, A. Arakawa, M. Minamimoto, and M. Hiraishi for technical assistance and Y. Okuda for secretarial work. We also thank M. Takeichi, A. Miyawaki, K. Nakayama, and T. Shishido for cDNAs; Sh. Tsukita for cells; Y. Mazaki and A. Suzuki for technical suggestions; and H. A. Popiel for critical reading of the manuscript. This work was supported in part by grants-in-aid from the Ministry of Education, Science, Sports and Culture of Japan and by the Takeda Science Foundation.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E08-06-0549) on February 4, 2009.