E-cadherin endocytosis regulates the activity of Rap1: a traffic light GTPase at the crossroads between cadherin and integrin function

Cadherins and integrins are the principal effectors of cell-cell and cell-extracellular matrix adhesion, respectively, and are critical determinants of tissue architecture and function both in developing and adult organisms (Wheelock and Johnson, 2003; Hynes, 2002).

Cadherins are single-pass transmembrane glycoproteins characterized by a variable number of extracellular `cadherin domains' that coordinate Ca²⁺ ions and mediate homophilic cell-cell adhesion in all solid tissues. In epithelia, E-cadherin is the essential component of adherens junctions (AJs), specialized calcium-dependent adhesive structures required for formation and maintenance of stable cell-cell adhesion (Wheelock and Johnson, 2003). Among the other core components of AJs, the roles of β-catenin, which interacts directly with the cadherin cytoplasmic tail, and α-catenin, which binds to both β-catenin and actin, and thereby connects the cadherin/catenin complex to the actin cytoskeleton, have been extensively characterized. In addition, recent growing evidence suggests an essential role for p120^ctn, which binds to the cadherin juxtamembrane domain (Peifer and Yap, 2003). However, while many potential regulatory proteins have been found to be associated with E-cadherin-catenin complexes, only recently has attention been focused on signaling events linked to the adhesive and tumor suppressor functions of cadherins. In particular, there is evidence that cadherins can modulate the activation and signaling of diverse Rho family GTPases and receptor tyrosine kinases (Wheelock and Johnson, 2003; Yap and Kovacs, 2003).

Integrins are heterodimeric transmembrane glycoproteins composed of noncovalently linked α and β subunits, which are endowed with both structural and regulatory functions, linking extracellular matrix to the actin cytoskeleton at focal adhesion sites, and providing bi-directional transmission of signals across the plasma membrane (Hynes, 2002). Through their outside-in and inside-out signaling, integrins regulate a number of critical cellular processes, including proliferation, differentiation, survival, migration and gene expression (Hynes, 2002).

It is generally postulated that a fine-tuned molecular crosstalk must be coordinated, both temporally and spatially, between cadherins and integrins, for the proper development and maintenance of tissue architecture. This idea is supported by the finding of the coordinate disruption of cadherin-dependent intercellular junctions and induction of integrin-dependent cell motility during the epithelial-mesenchymal transition (EMT) of most malignant tumors (Frame, 2002). However, although there is indeed evidence of crosstalk between members of these two adhesive receptor families (Gimond et al., 1999; von Schlippe et al., 2000), the molecules and molecular mechanisms involved remain ill defined. Therefore, clarification of how this crosstalk is regulated remains a fundamental challenge in understanding important physiological and pathological processes such as morphogenesis, wound healing, tumor invasion and metastasis.

The major candidate cross-regulatory molecules are small GTPases of the Ras and the Rho families. These small GTPases act as molecular switches by cycling between an inactive GDP-bound form and an active GTP-bound form, which allows interactions with effectors, thus controlling a wide range of essential biochemical pathways in all eukaryotic cells (Bar-Sagi and Hall, 2000). They are tightly regulated by guanine nucleotide exchange factors (GEFs), which stimulate GTP loading, GTPase activating proteins (GAPs), which catalyze GTP hydrolysis, and guanine nucleotide dissociation inhibitors (GDIs), which antagonize both GEFs and GAPs (Bar-Sagi and Hall, 2000).

Indeed, distinct members of the Rho GTPase family, including RhoA, Rac1 and Cdc42, have been showed to play a role in the regulation of both integrin and cadherin adhesive functions in important physiological and pathological processes (Arthur et al., 2002). In particular, Rac1 has been implicated in the regulation of cell-cell and cell-extracellular matrix adhesion during tumorigenesis (Lozano et al., 2003).

However, lessons from lower eukaryotes and Drosophila, as well as recently growing evidence in mice and in mammalian cells point to Rap1, a member of the Ras family of GTPases, as a crucial regulator of fundamental cell adhesion-dependent biological events such as morphogenesis, immune response, haemostasis and tumor invasion (Bos et al., 2001; Hattori and Minato, 2003). Indeed, a number of recent studies have firmly established that Rap1 regulates the inside-out activation of most integrins by mediating their polarized spatial redistribution and stabilization in an active conformation (Bos et al., 2003; Caron, 2003; Katagiri et al., 2003; Bivona et al., 2004; Dustin et al., 2004), suggesting that Rap1 activation is required for the induction and maintenance of integrin-mediated cell adhesion. At the same time, a study in Drosophila indicates that Rap1 might also play an important role in regulating morphogenetic processes through the control of AJ positioning during cell division and cell motility (Knox and Brown, 2002), whereas defective Rap1 activation has been recently associated with loss of AJs and cell scattering, suggesting that Rap1 may play a role in AJ formation and maintenance (Yajnik et al., 2003; Price et al., 2004; Hogan et al., 2004). In addition, recent works have indicated that either defective or excess Rap1 activation can contribute to malignancy by affecting some integrin and cadherin functions (Yajnik et al., 2003; Hattori and Minato, 2003). Nevertheless, our current understanding of Rap1 function in cell-matrix and cell-cell adhesion is far from complete: in particular, major ambiguities still persist regarding the effects of cell-matrix adhesion and integrin outside-in signaling on Rap1 activity (Bos et al., 2001; Bos et al., 2003), with reports showing either an increase or a decrease of GTP-loaded Rap1 in adherent versus suspended cells (Posern et al., 1998; Buensuceso and O'Toole, 2000), and studies that have detected either enhanced or unchanged levels of active Rap1 in response to the direct activation of integrins by activating antibodies (Franke et al., 2000; de Bruyn et al., 2002). Moreover, how Rap1 is related to cadherin functions remains elusive.

In the present study we provide insights into how the breakdown of E-cadherin-dependent cell-cell junctions (AJs) leads to the formation of integrin-dependent focal adhesions (FAs) via the regulation of Rap1 activity.

The mouse hybridoma producing the anti-human β₁ monoclonal antibody (mAb) TS2/16 was obtained from the American Type Culture Collection (ATCC). This hybridoma was injected into mice and mAb TS2/16 was affinity purified from ascitic fluid. The mouse mAb ECAD-FR1 against E-cadherin was derived from the B11/5D1 hybridoma produced in our laboratory by fusion of mouse myeloma cells and splenocytes from BALB/c mice immunized against a glutathione S-transferase (GST)-E-cadherin fusion protein containing the cytoplasmic domain of human E-cadherin. Immunofluorescence, immunohistochemistry, immunoprecipitation and western blotting determined its reactivity and specificity for either recombinant or endogenous E-cadherin proteins. Cross-reactivity was observed with human, rat and mouse E-cadherin (M. Cutufia, M. Enrietto and S.F.R., unpublished). Other primary antibodies used were commercial, and included mouse mAbs against E-cadherin (clone 34), β-catenin (clone 14), p120^ctn (clone 98) (Transduction Laboratories), a rabbit polyclonal antibody (pAb) to α-catenin (C-2081; Sigma Chemical Co.), a mouse mAb to c-Src (H-12) and a rabbit pAb to Rap1 (121) (Santa Cruz Biotechnology), a mouse mAb to Rac1 (clone 23A8; Upstate Biotechnology). The mAb DECMA-1 (Sigma; 1:100 dilution) was used for E-cadherin function-blocking experiments. Primary antibodies were detected using affinity purified HRP-, FITC- or RITC-conjugated secondary antibodies (Jackson Laboratories).

Bovine serum albumin (BSA), poly-L-lysine (PL), poly(2-hydroxyethyl methacrylate) (poly-HEMA), phenylarsine oxide (PAO), cytochalasin D (CytD), bafilomycin A₁ (BAF), N-ethylmaleimide (NEM), chlorpromazine (CPZ) and filipin were from Sigma. The inhibitors of Src kinase 4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrazolo[3,4-d]pyrimidine (PP2) and 4-amino-1-tert-butyl-3-(1′-naphthyl)pyrazolo[3,4-d]pyrimidine (PP1) were from Calbiochem. Fibronectin (FN) was purified from human plasma as described previously (Retta et al., 2001). Glutathione-Sepharose 4B, protein A-Sepharose and protein G-Sepharose were from Amersham Biosciences.

EGFP-tagged canine Rab11 was kindly provided by C. Bucci (University of Lecce, Italy). mRFP-Rap1A was produced by subcloning a HindIII-BamHI fragment containing the coding sequence of human Rap1A from pcDNA3-Rap1A (provided by S. Paganini, University of Pavia, Italy) into the pmRFP-C1 vector encoding a monomeric red fluorescent protein tag (provided by K. Rottner, Gesellschaft fur Biotechnologische Forschung, Braunschweig, Germany). The EGFP-zyxin and EGFP-paxillin constructs were described previously (Rottner et al., 2001).

As in vitro model systems, we used the Fisher rat thyroid (FRT), the mouse GE11 and GE11-β₁A, and the human MCF10A epithelial cell lines, the monkey COS7, the mouse endothelial cell line sEnd1, and the mouse NIH/3T3, GD25, GD25-β₁A and GD25-β₁TR fibroblastic cell lines. GE11 and GD25 cell lines do not express integrins of the β₁ family and have been a valuable model for the functional characterization of distinct β₁ isoforms as well as for the analysis of crosstalk mechanisms between either distinct integrins or integrins and cadherins (Retta et al., 2001; Gimond et al., 1999). GD25 cells expressing the human β₁A integrin isoform or β₁TR mutant, lacking the entire cytoplasmic domain, were previously described (Retta et al., 2001). All cell lines were cultured at 37°C and 5% CO₂ atmosphere in Dulbecco's modified Eagle's medium (DMEM; Invitrogen), supplemented with 10% fetal calf serum (FCS; Invitrogen), 2 mM glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin, except for FRT and MCF10A cells, which were cultured in Coon's F-12 medium (Biochrom AG) and mammary epithelial cell medium (MEGM; Cambrex Corporation), respectively.

Confluent cells were serum-starved overnight, washed three times with PBS, and either lysed directly onto tissue culture dishes (Ad, adherent cells), or harvested and dissociated into single cells by treatment with 5 mM EDTA in PBS for 30 minutes at 37°C and pipetting. Harvested cells were then either washed once with PBS and lysed (Su, suspension cells), or resuspended in serum-free medium and allowed to adhere for the indicated time to tissue culture dishes that had been coated by overnight incubation with 10 μg/ml substratum proteins before lysis. In some experiments, cells were plated in the presence of 1 μM cytochalasin D (CytD) to perturb actin cytoskeleton dynamics.

Cell monolayers were harvested by trypsinization and pipetting to generate a maximal dispersed cell suspension, replated at various densities in serum-free medium, and allowed to recover overnight at 37°C before measurement of Rap1 activity.

Ca²⁺ switch experiments were performed using a well established procedure (Nakagawa et al., 2001). Briefly, epithelial cell monolayers were incubated in serum-free DMEM supplemented with 4 mM EGTA for 30 minutes at 37°C to disrupt E-cadherin-mediated cell-cell contacts. Treated cells were then either lysed or allowed to reform adherens junctions by further incubation with fresh medium containing 1.8 mM Ca²⁺ for 30-60 minutes at 37°C.

As an alternative method to EGTA treatment, in some experiments cadherin-mediated intercellular adhesion was disrupted by treating confluent epithelial cells for various times with serum-free medium containing 5 μM PAO, a selective tyrosine phosphatase inhibitor (Retta et al., 1996).

For experiments requiring inhibition of Src family kinase (SFK) activity, cells were preincubated for 1 hour at 37°C in serum-free medium supplemented with the specific SFK inhibitor PP2 (10 μM).

Cells were harvested from tissue culture dishes, completely dissociated into single cells by 5 mM EDTA treatment and pipetting, resuspended in Hank's balanced salt solution supplemented with either 4 mM EGTA or 1.8 mM Ca²⁺, seeded on suspension dishes (Petri dishes coated with poly-HEMA to keep cell in suspension), and incubated for 1 hour at 37°C on a gyratory shaker at low speed.

To block E-cadherin endocytosis promoted by the disassembly of AJs, epithelial cell monolayers were preincubated with CytD (1 μM) in serum-free medium for 30 minutes at 37°C to disrupt the actin cytoskeleton, prior to disruption of cell-cell adhesion by EGTA (4 mM) treatment for 30 minutes at 37°C. For experiments requiring blocking of E-cadherin endocytic trafficking, cells were incubated at 37°C in serum-free medium containing either 1 μM bafilomycin A₁ or 1 mM N-ethylmaleimide for 1 hour and 30 minutes, respectively.

Cell lysates were produced from both adherent and suspended cells by incubation in ice-cold cell lysis buffer for 15 minutes on ice. Lysates were centrifuged at 12,000 g for 10 minutes at 4°C and total protein concentration in the supernatants was determined using the Bio-Rad Protein Assay. Supernatant aliquots containing equal amounts of total proteins (∼2 mg) were precleared with a mixture of protein A- and protein G-Sepharose and used for pull-down assays and immunoprecipitations.

Rap1 activation was examined using an established pull-down method based on the specific binding of a GST fusion protein containing the Rap-binding domain of RalGDS (RalGDS-RBD/GST) to the active, GTP-bound form of Rap1 (Franke et al., 2000). Briefly, TOPF10 Escherichia coli were transformed with the expression vector pGEX-RalGDS-RBD, and RalGDS-RBD/GST fusion proteins were induced with 0.1 mM isopropyl-1-thio-β-D-galactopyranoside (IPTG). Bacteria were then resuspended in lysis buffer (see Solvent concentrations) and sonicated. RalGDS-RBD/GST fusion proteins were purified from sonicate supernatant by incubation with glutathione-coupled Sepharose 4B beads (Amersham) for 1 hour at 4°C. The beads were washed three times in lysis buffer and the amount of bound fusion proteins was estimated by SDS-PAGE and Coomassie Blue staining. Aliquots of glutathione-Sepharose beads containing about 50 μg of RalGDS-RBD/GST proteins were then used to precipitate GTP-bound Rap1 from cell lysate supernatants by incubation for 1 hour at 4°C with gentle rotation. The beads were then washed three times with an excess of lysis buffer, and bound proteins were eluted in Laemmli sample buffer.

For detection of active Rac1 by the pull-down assay, a GST fusion protein containing the Rac1-binding domain of PAK (GST-PAK-CRIB) was used (Nakagawa et al., 2001).

Affinity precipitated GTP-bound Rap1 and Rac1 were identified by SDS-PAGE (12%) and western blotting using specific antibodies. In parallel, western blotting of whole lysate supernatants ensured that relative equal amounts of total proteins were used for the pull-down assays.

Immunoprecipitation and western blotting analysis were performed as previously described (Retta et al., 2001). Briefly, cell lysates containing equal amounts of total proteins (∼2 mg) were incubated overnight at 4°C with the appropriate dilutions of specific antibodies and a mixture of protein A- and protein G-Sepharose beads. Thereafter, beads were washed four times with lysis buffer, and immunoprecipitated proteins were eluted with Laemmli buffer and subjected to SDS-PAGE followed by western blotting. In parallel, beads alone or equal amount of nonspecific IgG were used as negative control, whereas western blotting of whole lysates ensured that relative equal amounts of total proteins were used for immunoprecipitations.

For western blotting analysis, proteins separated by SDS-PAGE were electroblotted onto Hybond-C transfer membrane (Amersham). The blots were blocked with 5% BSA in Tris-buffered saline (TBS) containing 0.1% Tween 20 for 1 hour at 42°C, incubated with appropriate dilutions of primary antibodies overnight at 4°C and subsequently with HRP-conjugated secondary antibodies for 2 hours at room temperature (RT). Proteins were then visualized by an enhanced chemiluminescence (ECL) detection system (Amersham).

After stripping with 2% SDS in 62.5 mM Tris-HCl (pH 6.8) at 42°C for 1 hour, to remove bound antibodies, some blots were reprobed sequentially with distinct primary antibodies to assess the presence and levels of specific proteins.

The kinase activity of endogenous Src was detected by an in vitro Src autophosphorylation assay. Briefly, cells were lysed by incubation with 1 ml of RIPA buffer for 5 minutes on ice. The lysates were centrifuged at 12,000 g for 10 minutes at 4°C, and total protein concentration in the supernatants was determined. Aliquots of supernatants containing equal amounts of total proteins (∼2 mg) were precleared with protein G-Sepharose and incubated with 5 μg of anti-Src mAb for 2 hours at 4°C with gentle rotation. Src immunocomplexes were subsequently collected by further incubation for 1 hour at 4°C with protein G-Sepharose, and washed four times with RIPA buffer and twice with kinase buffer. Immunoprecipitates were then resuspended in 100 μl of kinase buffer and split into two equal aliquots. Half of the immunoprecipitates was incubated with 1 nM ATP and 5 μCi [γ-³²P]ATP for 10 minutes at 30°C to perform the Src kinase assay. The reaction was terminated with the addition of an equal volume of 2× Laemmli buffer. Samples were subjected to SDS-PAGE (8%), and the gel was analyzed by autoradiography. The other half of the immunoprecipitates was subjected to SDS-PAGE and western blotting with anti-Src antibodies to confirm equal loading of Src proteins.

FRT cells were transiently transfected with 1.5 μg of cDNA constructs using the FuGENE 6 Transfection Reagent according to the manufacturer's instructions (Roche Diagnostics). At 12-24 hours after transfection, cells were replated in serum-free medium onto glass coverslips coated with 25 μg/ml FN.

Cells expressing EGFP-zyxin were microinjected with either GST-Rap1GAPv, a GST fusion protein containing a catalytically active fragment (residues 75-415) of the Rap1-specific Rap1GAP [sufficient for Rap1 inactivation (Brinkmann et al., 2002)], or GST alone as control. Protein injections were performed with sterile Femtotips 1 (Eppendorf) using a Leitz micromanipulator and an Eppendorf 5242 pressure supply, as described previously (Kaverina et al., 1999). Cells were injected using the backpressure mode (20-80 hPa) to give a continuous outflow from the needle. Proteins were microinjected at concentrations of 0.5-1 mg/ml.

Cells plated on glass coverslips were fixed in 3% paraformaldehyde for 10 minutes at room temperature (RT) and permeabilized with 0.5% Triton X-100 in TBS for 1 minute, incubated with 1% BSA in PBS for 30 minutes, and stained with appropriate dilutions of primary antibodies and either RITC- or FITC-conjugated secondary antibodies. Filamentous actin was revealed by staining with FITC-conjugated phalloidin (Sigma). Pictures were taken on an Olympus BX41 microscope equipped with a SPOT RT Slider CCD camera (Diagnostic Instruments). Colocalization studies were performed using a Leica TCS_SP2 confocal laser-scanning unit (Leica Microsystems).

Cells were observed and microinjected in an open heating chamber (Warner Instruments) maintained at 37°C on an inverted microscope (Axiovert 100M; Zeiss) equipped for multi-channel epifluorescence and phase contrast video microscopy as described previously (Kaverina et al., 1999). Injections were performed at an objective magnification of 40×, and video microscopy at either 63× or 100×. Phase contrast and fluorescence images were automatically acquired in rapid succession at 20-second intervals for up to 2 hours with a high-resolution cooled CCD camera (MicroMAX:512BFT, Princeton Research Instruments). The hardware was driven by IPLab software (Scanalytics). QuickTime movies were created from the original time-lapse images using the MetaMorph software (Universal Imaging).

Bacteria lysis buffer [50 mM Tris-HCl (pH 7.4), 0.1% Triton X-100, 100 mM NaCl, 5% glycerol, 1 mM EDTA, 1 mM DTT, 10 μg/ml leupeptin, 4 μg/ml pepstatin, 0.1 TIU/ml aprotinin, 1 mM PMSF]; Hank's balanced salt solution (HBSS) [1.3 mM CaCl₂, 0.4 mM MgSO₄, 5 mM KCl, 138 mM NaCl, 5.6 mM D-glucose, 25 mM Hepes (pH 7.4)]; cell lysis buffer [50 mM Tris-HCl (pH 7.4), 1% Igepal CA-630, 200 mM NaCl, 2.5 mM MgCl₂, 5% glycerol, 10 μg/ml leupeptin, 4 μg/ml pepstatin, 0.1 TIU/ml aprotinin]; Laemmli sample buffer [62.5 mM Tris-HCl (pH 6.8), 10% glycerol, 1% SDS, 5% β-mercaptoethanol]; RIPA buffer [1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, 150 mM NaCl, 50 mM Tris-HCl (pH 7.2), 0.4 mM Na₃VO₄, 10 μg/ml leupeptin, 4 μg/ml pepstatin, 0.1 TIU/ml aprotinin]; kinase buffer [50 mM Hepes (pH 7.4), 5 mM MgCl₂, 3 mM MnCl₂].

To examine the putative involvement of Rap1 in outside-in integrin signaling, and to circumvent possible cell context specificities, we analyzed cell adhesion effects on endogenous Rap1 activity in several epithelial, endothelial and fibroblastic cell lines. First we addressed the open question of whether the amount of active Rap1 is different in adherent and suspended cells. The results showed that the amount of active Rap1 was strongly increased upon cell detachment from the substratum in all cell lines tested (Fig. 1A, and not shown), thus supporting and extending a previous observation in CHO cells (Buensuceso and O'Toole, 2000), and implying that cell adhesion exerts an inhibitory effect on Rap1 activity.

To explore potential connections with integrin function, we performed cell adhesion experiments using both β₁ integrin-deficient and β₁A-transfected GD25 cells (Retta et al., 2001), which specifically attach and spread on fibronectin through α_Vβ₃ and α₅β₁ integrins, respectively, as well as GD25 cells expressing a β₁ integrin mutant lacking the entire cytoplasmic domain (β₁TR) and incapable of mediating β₁-specific signal transduction events (Retta et al., 2001). This model was well suited to examine the effects of the expression and function of specific integrins on Rap1 activity. The results showed a consistent slow but progressive down-regulation of Rap1 activity during adhesion to fibronectin (FN) of either GD25 (Fig. 1B) or GD25-β₁A (not shown) cells, as well as upon GD25-β₁A cell adhesion to the β₁ integrin-activating monoclonal antibody TS2/16 (Fig. 1D, left panel). However, the same effect was observed on the non-specific substratum poly-L-lysine (PL) (Fig. 1C), as well as when cell adhesion was mediated by the signaling-defective β₁TR mutant (Fig. 1D, right panel). Similar results were obtained using the epithelial GE11 and GE11-β₁A cell lines (not shown). Intriguingly, the down-regulation of Rap1 activity during cell adhesion was characterized by a constant time-course with a return to basal levels after about 1 hour of cell adhesion (Fig. 1B-D); at this time cells were completely spread on the substratum and intercellular adhesions were observed. Taken together, these results show that a down-regulation of Rap1 activity occurs during cell adhesion, but independently of integrin activation and function, suggesting that, at least in our experimental systems, Rap1 is not involved in outside-in integrin signaling.

To assess whether the down-regulation of Rap1 activity induced by cell adhesion was due to mechanosensory processes coupled to the remodeling of actin cytoskeleton, we examined the role of actin cytoskeleton dynamics in Rap1 activity modulation. As shown in Fig. 2A, cell pretreatment with cytochalasin D (CytD), although effective in disrupting the actin cytoskeleton (Fig. 2B), resulted in no significant effect upon the down-regulation of Rap1 activity during cell adhesion, indicating that an intact actin cytoskeleton was not required for this event.

However, during the course of our experiments, we noticed an inverse relationship between the level of GTP-loaded Rap1 and cell density. To assess this relationship, Rap1 activity was measured for different periods of time in cells plated at various densities. Fig. 2C shows typical results, which demonstrated that the GTP-loading of Rap1 was inversely proportional to cell density and was minimal when cells reached confluence, suggesting a major role for cell-cell adhesion in the regulation of Rap1 activity. Interestingly, basal levels of active Rap1 were reached within 1 hour after plating (see Fig. 1), and persisted even when cells became fully confluent, indicating a constitutive basal activity of Rap1. Conversely, the activation of Rap1 induced by cell resuspension became less evident with the decrease in cell density (not shown).

It is noteworthy that the time required for maximal down-regulation of Rap1 activity during cell adhesion was very similar to that required for strengthening of cadherin-dependent cell-cell adhesion (45 minutes to 1 hour) (Niessen and Gumbiner, 2002), suggesting that cadherin adhesive function might affect Rap1 activity.

Cadherin functions are commonly studied by comparing the behavior of cultured cells as they grow to confluence, or where cell-cell contacts are abruptly broken and allowed to reform through manipulation of extracellular calcium or cellular tyrosine phosphatase activity, and by using blocking antibodies to inhibit cadherin adhesive function. To assess the potential role of cadherin-mediated cell-cell junctions in the modulation of Rap1 activity, we combined all these approaches with the pull-down assay for GTP-loaded Rap1. In particular, we focused our attention on the role of E-cadherin, using epithelial cells as in vitro model systems. E-cadherin-dependent cell-cell junctions were modulated using a well established Ca²⁺ switch procedure (Nakagawa et al., 2001), which involves the disruption of AJs by removal of extracellular Ca²⁺ with the specific chelator EGTA, and the re-establishment of cadherin-mediated cell-cell contacts by the subsequent restoration of Ca²⁺ ions. In addition, to circumvent possible artifacts attributable to Ca²⁺ chelation by EGTA, we also disrupted E-cadherin-dependent cell adhesion by treating cells with the selective tyrosine phosphatase inhibitor phenylarsine oxide (PAO) (Retta et al., 1996), which has been previously shown to decrease cadherin-dependent cell adhesion by increasing tyrosine phosphorylation of AJ components (Staddon et al., 1995). The outcomes of these experiments clearly showed that the disruption of AJs by both EGTA and PAO treatments led to a strong activation of Rap1 with a time course of about 30 minutes (Fig. 3A,C), and a subsequent return to basal levels of Rap1 activity within 45 minutes as E-cadherin-mediated cell-cell adhesion was restored (Fig. 3A, and not shown), suggesting that AJ dynamics plays a major role in the regulation of Rap1 activity. Immunofluorescent staining confirmed the effectiveness of both the Ca²⁺ switch and tyrosine phosphatase inhibition procedures (Fig. 3B,D), and showed that cells acquired a fibroblast-like morphology upon both EGTA and PAO treatment, suggesting that cell-matrix adhesion was at least preserved during disruption of intercellular junctions. Accordingly, we have previously shown that inhibition of PAO-sensitive tyrosine phosphatases is a sufficient stimulus for triggering integrin-mediated focal adhesion formation (Retta et al., 1996).

To consolidate these findings, we coupled the analysis of GTP-bound Rap1 with that of GTP-bound Rac1 within the same experimental procedure. In fact, we reasoned that the well established effect of the modulation of E-cadherin-mediated adhesion upon Rac1 activity (Nakagawa et al., 2001) could represent a useful control of the reliability of the methods used. Strikingly, we observed that the changes in Rap1 activity that accompanied the assembly/disassembly of AJs were closely paralleled by inverse changes in Rac1 activity (Fig. 3E), indicating that the modulation of E-cadherin-mediated adhesion exerts opposite effects upon the activity of Rap1 and Rac1.

To investigate the relationship between Rap1 activity and E-cadherin adhesive function, we next examined the effect of DECMA-1, a neutralizing antibody against E-cadherin, during the modulation of cell-cell adhesion by the Ca²⁺ switch procedure. The results clearly showed that the decrease in Rap1 activity induced by restoration of cell-cell adhesion was inhibited by DECMA-1 (Fig. 4A), demonstrating that E-cadherin engagement was required for the down-regulation of Rap1 activity.

This critical role of E-cadherin in the regulation of Rap1 activity was further confirmed by a cell-cell aggregation assay in which cells were allowed to reaggregate from single cell suspensions by restoring Ca²⁺ to the medium either in the absence or in the presence of DECMA-1 to prevent E-cadherin-mediated cell-cell adhesion (Fig. 4B,C). This result indicates that E-cadherin-adhesion-dependent modulation of Rap1 activity can occur even in suspended cells.

Taken together these results show that the disruption of AJs by distinct mechanisms leads to a strong activation of Rap1, whereas their restoration induces a down-regulation of Rap1 activity, which is independent of cell-matrix adhesion but requires E-cadherin function. In addition, they allow a fine reinterpretation of the results reported in Fig. 1, demonstrating that the activation of Rap1 occurs even before cell resuspension, and that it is indeed the formation/disruption of cadherin-mediated cell-cell contacts that plays a major role in the modulation of Rap1 activity during cell adhesion/de-adhesion processes.

In an attempt to understand the mechanism by which cadherin-mediated cell-cell contacts regulate Rap1 activity, we examined the role of the non-receptor tyrosine kinase Src. In epithelial cells, Src is localized both at AJs and focal adhesions, and its activity is required for the turnover of these adhesive structures (Frame, 2002).

As a first step we tested the kinase activity of endogenous Src under the Ca²⁺ switch procedure in order to determine whether changes in Src activity accompanied the observed changes in Rap1 GTP-loading. Using an in vitro Src autophosphorylation assay, we detected a significant increase in Src kinase activity upon disruption of cell-cell adhesion by EGTA treatment, whereas the activity of Src returned to the basal level observed in untreated cells following Ca²⁺-induced restoration of cell-cell adhesion (Fig. 5A,B). There was thus a parallel correspondence between changes in Src and Rap1 activities during the modulation of E-cadherin-mediated cell-cell adhesion. Accordingly, it has been reported that Src activity is increased following disruption of cell-substratum adhesion by EDTA treatment in confluent but not in subconfluent cells (Maher, 2000), whereby the catalytic activity of Src family kinases (SFK) is needed for destabilization and turnover of cadherin-dependent cell-cell adhesions (Frame, 2002).

To assess whether Src activity was required for Rap1 activation following disruption of E-cadherin-mediated cell-cell adhesion, confluent cells were pretreated with PP2, a selective inhibitor of SFK, before EGTA treatment. The results showed that the presence of PP2 prevented the activation of Rap1 induced by EGTA treatment (Fig. 5C). Similar results were obtained using PP1, another inhibitor of SFK (see Fig. S1 in supplementary material), thus establishing that the catalytic activity of SFK was required for Rap1 activation that accompanied AJs disassembly.

Taken together, these results suggest that the activation of Src by disruption of E-cadherin-mediated cell adhesion represents a plausible mechanism for initiating a signaling pathway that involves Rap1 as a downstream effector.

In order to further investigate the mechanisms underlying the cadherin function-dependent modulation of Rap1 activity and to gain insight into the physiological significance of this event, we examined the possibility that Rap1 and E-cadherin might associate in epithelial cells. Immunofluorescence experiments revealed a clearly distinct localization of Rap1 and E-cadherin in confluent epithelial cells, with Rap1 being diffuse in the cytoplasm and E-cadherin concentrated at the plasma membrane; by contrast, E-cadherin label localized to the cytoplasm upon disruption of cell-cell adhesion by EGTA treatment (see Fig. 7E). Therefore, we performed coimmunoprecipitation experiments from lysates of both untreated and EGTA-treated epithelial cells. As shown in Fig. 6A (left panels, upper blot), immunoprecipitation of Rap1 (IP Rap1) from untreated confluent epithelial cells resulted in the coprecipitation of some of the E-cadherin, whereas treatment with EGTA greatly increased the amount of E-cadherin recovered, suggesting that a protein complex containing both Rap1 and E-cadherin may form in vivo and that the formation of this complex is up-regulated in response to the disruption of cell-cell adhesion.

To examine whether Rap1 immunocomplexes contained other AJ proteins besides E-cadherin, we reprobed Rap1 immunoprecipitations with antibodies against distinct catenins, including β-catenin, α-catenin and p120^ctn, as well as against well established cadherin regulatory partners, including EGFR and Src. These analyses revealed that β-catenin and some α-catenin, but not p120^ctn, were coimmunoprecipitated with Rap1 and E-cadherin, particularly upon AJ disassembly by EGTA treatment (Fig. 6A, left panels). Rap1 can thus associate with E-cadherin-catenin complexes that do not contain p120^ctn, and this association increases following the disassembly of AJs. However, neither E-cadherin nor catenins were detected in Rap1 immunoprecipitations from suspended cells (Fig. 6A), nor E-cadherin cleavage products containing the cytoplasmic domain coprecipitated with Rap1 in any condition (Fig. 6A, compare left and right panels; see also Fig. 6B), suggesting that a functional full-length E-cadherin is required for the formation of Rap1/E-cadherin-catenin complexes, and that these complexes could form early during cell resuspension and decrease later on, when, in the absence of cell-cell contact restoration, the increased pool of cell surface E-cadherin undergoing endocytosis is targeted for degradation. Consistently, immunoprecipitation experiments performed at different time points during cell resuspension and in suspended cells have confirmed that the presence of Rap1/E-cadherin-catenin complexes decreases over time during cell resuspension (data not shown).

It is noteworthy that the enhancement in the level of E-cadherin-catenin complexes associated with Rap1 observed upon EGTA treatment was suppressed when cells were pretreated with the Src family kinase inhibitor PP2 (Fig. 6B, left panels), indicating that SFK activity was required. By contrast, the inhibition of SFK did not influence the lack of p120^ctn association with Rap1 immunocomplexes containing E-cadherin (Fig. 6B, left panels), although PP2 treatment clearly enhanced p120^ctn association with E-cadherin, as detected by parallel immunoprecipitation and western blotting analyses (Fig. 6B, right panels). In addition, we constantly observed that the EGTA treatment that led to an increased formation of Rap1/E-cadherin/catenin complexes (Fig. 6B, left panels) resulted in a reduction in the amount of p120^ctn associated with E-cadherin (Fig. 6B, right panels), strongly suggesting that a dynamic and mutually exclusive association of Rap1 and p120^ctn with E-cadherin-catenin complexes may occur in certain physiological conditions.

Taken together, these data indicate that a SFK-dependent increased association of Rap1 with E-cadherin-catenin complexes lacking p120^ctn occurs as consequence of the disassembly of AJs, suggesting that this association must be a tightly regulated event that responds to proper cadherin function-dependent and SFK-mediated spatial and temporal signals.

There is increasing evidence suggesting a major role for Src in regulating cadherin endocytosis (Fujita et al., 2002; Palovuori et al., 2003), which in turn has recently emerged as an important determinant of E-cadherin function (Le et al., 1999; Palacios et al., 2001; Pece and Gutkind, 2002). Therefore, to provide additional information on the mechanisms by which E-cadherin adhesive function regulates Rap1 activity, we assessed whether the endocytic trafficking of E-cadherin could contribute to the SFK-dependent modulation of Rap1 activity coupled to the remodeling of AJs. We initially took advantage of the established requirement of a functional actin cytoskeleton for receptor endocytosis (Akhtar and Hotchin, 2001; Engquist-Goldstein and Drubin, 2003; Ascough, 2004). In particular, to block E-cadherin internalization promoted by the disassembly of AJs, we disrupted the actin cytoskeleton by pretreating epithelial cell monolayers with CytD before disruption of intercellular junctions by EGTA treatment. Using immunofluorescence we observed that disruption of epithelial cell-cell contacts by EGTA treatment resulted in a general marked reduction in levels of E-cadherin at the plasma membrane and its increased localization to the perinuclear cytoplasm, with several cells showing internalization of E-cadherin into diffuse small vesicles and redistribution of F-actin to the perimeter of these vesicles (Fig. 7Ad,h). By contrast, in cells pretreated with CytD prior to exposure to EGTA, a majority of E-cadherin remained at the cell surface despite the loss of cell-cell adhesion, accumulating in large peripheral spot-like structures containing residual F-actin (Fig. 7Ac,g), showing that a dynamic actin cytoskeleton is required for endocytosis of E-cadherin induced by disruption of cell-cell adhesion. However, the disruption of the actin cytoskeleton per se by treatment with CytD alone did not affect significantly the surface level of E-cadherin nor the integrity of intercellular junctions (Fig. 7Ab,f).

With the assurance that the disruption of the actin cytoskeleton by CytD treatment was effective in preventing E-cadherin endocytosis but not loss of cell-cell junctions promoted by exposure to EGTA, we next sought to address whether E-cadherin endocytosis was required for the modulation of Rap1 activity. To this end, immunofluorescence experiments were paralleled with pull-down assays to measure Rap1 activity. As shown in Fig. 7B, CytD pretreatment clearly prevented the activation of Rap1 induced by cell exposure to EGTA, suggesting that the disruption of cell-cell junctions is not sufficient for stimulating Rap1 activation, but requires the internalization of E-cadherin as an additional crucial step.

To further support this conclusion, we performed additional experiments using specific pharmacological inhibitors of endocytosis as alternative approaches to block E-cadherin endocytosis. In particular, we used chlorpromazine (CPZ) and filipin, which have been shown to preferentially interfere with clathrin-dependent and clathrin-independent endocytosis, respectively (Wang et al., 1993; Sigismund et al., 2005). Using these tools in our experimental system, we demonstrated by immunofluorescence experiments that in FRT cells the EGTA-induced internalization of E-cadherin was inhibited by filipin but not chlorpromazine (see Fig. S2A in supplementary material), suggesting that in FRT cells E-cadherin is mainly internalized through a clathrin-independent pathway. Accordingly, clathrin-independent pathways for E-cadherin endocytosis have been previously shown (Akhtar and Hotchin, 2001; Lu et al., 2003; Paterson et al., 2003). Moreover, by parallel pull-down assays we demonstrated that filipin pre-treatment effectively inhibited also the EGTA-induced activation of Rap1, whereas CPZ pre-treatment was ineffective (see Fig. S2B in supplementary material), providing further support to our conclusion that E-cadherin endocytosis is required for Rap1 activation induced by adherens junction disassembly.

Rap1 activation stimulated by E-cadherin internalization could occur in different steps of E-cadherin endocytic trafficking, including the inductive process of internalization, when cadherin-containing vesicles bud into the cell, as well as subsequent vesicular fusion and budding events. Therefore, to further elucidate the mechanism of Rap1 activation upon cell-cell adhesion disruption we examined the effects of drugs acting at defined points along the endocytic/recycling pathways. In particular, to demonstrate a requirement for E-cadherin trafficking through the early endosome compartment, we used bafilomycin A₁ (BAF), a specific inhibitor of vacuolar type H⁺-ATPase proton pumps that blocks budding of endosomal carrier vesicle (ECV) by preventing endosome acidification (Bayer et al., 1998). This drug allows the early steps of endocytosis, including the internalization process and the formation of early endosome, but prevents the subsequent vesicular traffic, leading to accumulation of cargo in early endosomes (Bayer et al., 1998). Indeed, BAF has been previously shown to block the recycling of EGTA-induced endocytosed E-cadherin upon Ca²⁺ restoration, but not the disruption of cell-cell contacts and the internalization of E-cadherin induced by Ca²⁺ chelation (Le et al., 1999). Using BAF in our experimental system, we confirmed these observations (Fig. 7Cf, and Fig. S3Ac in supplementary material). At the same time, we showed that BAF pretreatment clearly prevented the activation of Rap1 induced by cell exposure to EGTA (Fig. 7D; BAF+EGTA), confirming that the disruption of cell-cell junctions and the internalization of E-cadherin, although necessary, are not sufficient for stimulating Rap1 activation, but require the subsequent vesicular traffic. Importantly, by comparing the activation state of Rap1 in cell monolayers pretreated with EGTA to induce E-cadherin endocytosis, and then treated with Ca²⁺ in the absence or presence of BAF to induce or prevent recycling of endocytosed E-cadherin, respectively, we showed that the effect of Ca²⁺ restoration in inducing a downregulation of Rap1 activity was prevented by the presence of BAF (Fig. S3A,B in supplementary material), suggesting that the recycling of endocytosed E-cadherin to the cell surface is required for the downregulation of Rap1 that occurs upon reformation of cell-cell contacts.

In addition, we used N-ethylmaleimide (NEM), which has been shown to block at least one membrane fusion step in the different endocytic processes (macropinocytosis, surface membrane internalization and phagocytosis) by inhibiting the function of the NEM-sensitive factor (NSF) (Robinson et al., 1997; Thompson and Bretscher, 2002). NSF is a cytosolic hexameric ATPase that uses energy from ATP hydrolysis to dissociate SNARE complexes after membrane fusion, allowing the individual SNARE proteins to be recycled for subsequent rounds of fusion (May et al., 2001; Bonifacino and Glick, 2004). In our experimental system, cell pretreatment with NEM resulted in a complete block of EGTA-induced E-cadherin endocytosis (Fig. 7Ce), and this event was again associated with the inhibition of Rap1 activation (Fig. 7D, NEM+EGTA), lending further support to the conclusion that E-cadherin endocytosis and Rap1 activation are correlated processes. In addition, in the attempt to discriminate between the role of E-cadherin endocytic and exocytic pathways in the stimulation of Rap1 activity, we examined the effects of NEM in different steps of the calcium switch procedure. In particular, by parallel immunofluorescence and biochemical analyses, we compared the subcellular localization of E-cadherin and the activation state of Rap1 in cell monolayers pretreated with or without NEM before exposure to EGTA, or pretreated with EGTA, to induce E-cadherin endocytosis, and then treated with or without NEM before Ca²⁺ restoration. Using these approaches, we found that NEM, while preventing the EGTA-induced E-cadherin internalization (Fig. S4Ad in supplementary material), did not block the recycling of endocytosed E-cadherin to the cell surface and the reformation of cell-cell contacts upon Ca²⁺ restoration (Fig. S4Af in supplementary material). Furthermore, we observed that a depletion of E-cadherin stores in the perinuclear recycling compartment occurred in steady-state cell monolayers as a consequence of NEM treatment (Fig. S4A in supplementary material), suggesting that, in contrast to the NSF function-requirement for E-cadherin endocytosis, a NSF function-independent mechanism may operate for both the constitutive and the Ca²⁺ restoration-induced recycling of endocytosed E-cadherin from intracellular storage compartments to the plasma membrane. Accordingly, it has been suggested that NSF is not needed for some exocytic processes based on a single membrane fusion event, such as the recycling from some endosome compartments, but is instead required for multi-step trafficking processes (Jahn and Sudhof, 1999; May et al., 2001; Bonifacino and Glick, 2004), and there is indeed evidence for NSF function-independent exocytosis (Sandvig et al., 2000; Bittner and Holz, 2005). At the same time, the presence of NEM, while preventing the activation of Rap1 induced by EGTA (Fig. S4B, NEM/EGTA in supplementary material), did not affect the downregulation of Rap1 activity that occurred in EGTA-treated cells upon Ca²⁺ restoration (Fig. S4B, EGTA/NEM/Ca²⁺ in supplementary material). Taken together, these results indicate that Rap1 activity slows down as NSF-dependent trafficking processes cease, suggesting that the activation of Rap1 does not occur during the exocytic route of E-cadherin, but instead during the NEM-sensitive endocytic pathway, and that, conversely, the downregulation of Rap1 activity may occur along the BAF-sensitive recycling pathway, probably during the last steps of Rap1-dependent exocytic routes. Consistent with this finding, previous reports showed the presence of an intracellular gradient of Rap1 activity in response to growth factor stimulation, with a predominant Rap1 activation at the perinuclear membrane compartment and inhibition at the plasma membrane (Mochizuki et al., 2001; Ohba et al., 2003), and, intriguingly, it has also been suggested that cell surface RapGAPs may serve as signal terminators for the vectorial transport of vesicles or Rap1-associated molecules to the plasma membrane (Polakis et al., 1991).

Interestingly, by confocal immunofluorescence analysis, we observed a distinct subcellular localization of E-cadherin and Rap1 in untreated confluent cells (Fig. 7Ea-c), as well as in cells pretreated with CytD prior to exposure to EGTA (Fig. 7Ed-f), while upon EGTA treatment alone both E-cadherin and Rap1 were redistributed to and partially colocalized at a perinuclear region (Fig. 7Eg-i). The perinuclear colocalization of E-cadherin and Rap1 was also prevented by pretreatment with NEM prior to exposure to EGTA (see Fig. 7Ce), suggesting that a transit along the endocytic pathway was required and that the perinuclear area of E-cadherin and Rap1 accumulation was an endosome compartment. Thus, we next sought to identify this compartment.

The staining pattern of this compartment was reminiscent of the pericentriolar recycling endosome (Ullrich et al., 1996), and also the time needed for E-cadherin and Rap1 perinuclear colocalization (20-30 minutes) was consistent with an accumulation at this compartment (Ullrich et al., 1996). We therefore transiently cotransfected FRT cells with EGFP-tagged Rab11, a known marker of the perinuclear recycling endosome (Ullrich et al., 1996; Sonnichsen et al., 2000), and RFP-tagged Rap1, and used confocal microscopy to assess whether Rap1 colocalized with this marker upon the EGTA-induced cell-cell junction disruption. There was a significant colocalization of RFP-Rap1 and EGFP-Rab11 in untreated cells (Fig. 7Fa-c). However, this colocalization was enhanced and prominently concentrated in a perinuclear region upon EGTA treatment (Fig. 7Fd-f). Similar results were obtained using antibodies to either endogenous Rap1 or E-cadherin (not shown), suggesting that the formation of Rap1/E-cadherin complexes probably occurs within the perinuclear Rab11-positive endosome compartment.

Taken together, our results clearly indicate that E-cadherin endocytosis and Rap1 activation are connected processes, and strongly suggest that the activation of Rap1 triggered by E-cadherin endocytosis occurs upon E-cadherin transit from early endosomes to Rab11-positive recycling endosomes.

Because of the well established Rap1 activity requirement for integrin adhesive function (Bos et al., 2001; Bos et al., 2003; Caron, 2003), we hypothesized that the activation of Rap1 triggered by E-cadherin endocytosis could promote the formation of integrin-dependent adhesive structures. To test this hypothesis, we performed time-lapse video microscopy analyses of the dynamics of EGFP-zyxin in transiently transfected FRT cells during cell-cell junction disassembly. EGFP-zyxin, in fact, is a well established marker for the detection of focal adhesion assembly, as it associates with large and stable focal adhesions and not with small and transient focal complexes (Rottner et al., 2001; Zaidel-Bar et al., 2004), and indeed it has been previously used as a molecular marker for the transformation of integrin-mediated adhesions from dot-shaped focal complexes (which contain paxillin and vinculin but not zyxin) into definitive focal adhesions, or as an exclusive marker for mature focal adhesions (Rottner et al., 2001; Zaidel-Bar et al., 2003). In addition, it is well known that the transition of focal complexes into focal adhesions is not manifested just by the incorporation of zyxin and a growth in size of the adhesion site, but also by and the concomitant assembly of an actin bundle and increase in adhesion strength (Zaidel-Bar et al., 2004). Conversely, zyxin has been shown to dissociate from disassembling focal adhesions earlier than vinculin and paxillin (Rottner et al., 2001), again pointing to zyxin as a useful marker for focal adhesion dynamics.

Using this strategy, we observed that in cells forming mature cell-cell contacts, EGFP-zyxin was mostly diffuse in the cytoplasm, and only few, small EGFP-zyxin-containing focal adhesions were detected, mainly restricted to membrane regions where cell-cell contacts were weak or absent. By contrast, upon cell-cell junction disruption by EGTA treatment, EGFP-zyxin was progressively and strongly accumulated in numerous newly formed focal adhesions, which gradually enlarged and eventually became stable within 30 minutes (Fig. 8A and Movie 1A,B in supplementary material). This process was entirely reversed upon the induced reformation of cell-cell junctions by Ca²⁺ restoration (Movie 2A,B in supplementary material), suggesting the existence of an inverse relationship between the assembly/disassembly of cadherin- and integrin-dependent adhesive structures. Accordingly, there is consistent evidence for a `tug-of-war' between cell-cell and cell-substratum adhesions (Lu et al., 1998; Gimond et al., 1999; Ryan et al., 2001; Avizienyte et al., 2002). In addition, similar results were obtained using EGFP-paxillin as an additional marker of the formation of integrin-mediated adhesive structures (Fig. S5 and Movie 3A,B in supplementary material).

To test whether the formation of focal adhesions that accompanied the loss of E-cadherin-dependent cell-cell adhesions required the activation of Rap1 triggered by E-cadherin endocytosis, we microinjected EGFP-zyxin-expressing cells with either GST-Rap1GAPv, a GST-fusion protein containing a catalytically active fragment of the Rap1-specific Rap1GAP (sufficient for Rap1 inactivation) (Brinkmann et al., 2002), or GST alone as control, and followed the dynamics of EGFP-zyxin during the EGTA-induced disassembly of cell-cell junctions. As shown in Fig. 8B and Movie 4A,B in supplementary material, the formation of EGFP-zyxin-containing focal adhesions was specifically and completely inhibited in cells injected with Rap1GAPv, suggesting that Rap1 activity plays a pivotal role in the functional relationship between E-cadherin and integrin-mediated adhesive structures that occurs during the modulation of cell-cell adhesion.

It is well established that, besides their structural roles, cadherins and integrins can provide bi-directional transmission of signals across topographically discrete regions of the plasma membrane (Hynes, 2002; Wheelock and Johnson, 2003). Moreover, it is generally accepted that the activity of these major cell-cell and cell-matrix adhesion receptors must be finely coordinated during important physiological and pathological processes, including morphogenesis and cancer (Frame, 2002). Nevertheless, the functional crosstalk between cadherins and integrins still remains largely unexplained at the molecular level.

Here we show that the small GTPase Rap1 plays a pivotal role at the crossroads between cadherin function and integrin-based adhesive structures. In particular, we demonstrate that E-cadherin internalization and endocytic trafficking triggers Rap1 activation, providing the first evidence of an E-cadherin-modulated endosomal signaling pathway involving Rap1 as a downstream target. Conversely, the formation of E-cadherin-mediated cell-cell junctions induces a progressive down-regulation of Rap1 activity, which is dependent on E-cadherin homophilic adhesion. The E-cadherin endocytosis-dependent activation of Rap1 is paralleled by the colocalization of Rap1 and E-cadherin at the perinuclear Rab11-positive recycling endosome compartment, and by the formation of Rap1/E-cadherin-catenin complexes that do not contain p120^ctn. Furthermore, an enhanced Src activity accompanies and is required for both the activation of Rap1 and the formation of Rap1/E-cadherin complexes, thus suggesting a potential molecular and signaling link between the remodeling of E-cadherin-mediated cell-cell junctions and the modulation of Rap1 activity (Fig. 9). In addition, we show that the E-cadherin endocytosis-dependent activation of Rap1 is associated with and is required for the assembly of integrin-dependent adhesive structures, indicating that Rap1 may play a pivotal role in transmitting information from E-cadherin-based to integrin-based protein complexes by inside-out signaling. However, we also demonstrate that the modulation of Rap1 activity occurs independently of integrin-mediated outside-in signaling, suggesting that the link between Rap1 activity and integrin function is unidirectional and uniquely related to the inside-out signaling to integrins (Fig. 9).

In the attempt to address the role of integrin adhesive function in the regulation of Rap1 activity, we found that Rap1 was strongly activated upon cell detachment from the substratum, whereas Rap1 activity was progressively down-regulated during cell-matrix adhesion. However, the effect of cell adhesion on Rap1 activity was clearly independent of integrin adhesive function, but entirely dependent on the formation of cell-cell contacts. Indeed, using E-cadherin-blocking antibodies and two distinct and popular methods to manipulate E-cadherin-dependent intercellular adhesion, namely the calcium switch and the tyrosine phosphatase inhibition procedures, we demonstrated that E-cadherin function was required for the cell-cell contact-dependent down-regulation of Rap1 activity in epithelial cells, and that this down-regulation could occur even in suspended cells if E-cadherin-mediated intercellular contacts were allowed to form. Conversely, the up-regulation of Rap1 activity occurred even before cell detachment from the substratum, as it was caused just by the breakdown of E-cadherin-mediated cell-cell junctions. These findings shed light on the controversial issue of the effect of cell-matrix adhesion on Rap1 activity, and suggest that Rap1 is a target for E-cadherin-modulated signaling events that take place during the establishment and reorganization of cell-cell junctions.

Although the down-regulation of Rap1 activity was paralleled by the well established E-cadherin homophilic ligation-dependent activation of Rac1 (Nakagawa et al., 2001; Noren et al., 2001), our finding that E-cadherin-mediated adhesion leads to the down-regulation of Rap1 activity seems to contradict a recent study that shows activation of Rap1 upon epithelial cell adhesion to beads coupled with an Fc-tagged extracellular domain of E-cadherin (Hogan et al., 2004). One possible explanation is that different cell lines have been used in each study, as we have examined Rap1 activity in normal epithelial cells, and the previous study was based on the transformed MCF7 cell line. However, most probably the differences are attributable to the distinct methods applied, as we analyzed the effects of physiological E-cadherin-based cell-cell contacts, and in the other study only the effect of cell adhesion to a non-physiological rigid substratum (coated beads) was considered. Indeed, it has been previously reported that the incubation of cell monolayers with beads coated with cadherin ectodomain-recombinant proteins [which is the experimental approach used by Hogan et al. (Hogan et al., 2004)] results in the alteration of the even distribution of cadherins within the cells interacting with the beads, and can even trigger endocytic processes (Levenberg et al., 1998; Yam and Theriot, 2004), suggesting that the cell-bead interaction may induce a redistribution of E-cadherin similar to that occurring during the remodeling of cell-cell junctions triggered by external stimulation, thus resulting in the transient activation of Rap1. Accordingly, in their report Hogan and colleagues suggest that Rap1 activity is required for the recruitment of E-cadherin into nascent cell-cell contact sites, while it is less important for the maintenance of E-cadherin at mature cell-cell adhesions.

Interestingly, our finding that the regulation of Rap1 activity by AJ assembly/disassembly is paralleled by an inverse modulation of Rac1 activity raises the possibility that Rap1 crosstalks with Rac1 during the establishment and reorganization of AJs, thus opening a new interesting area for future studies. Consistently, it has been reported in budding yeast that the Rap1 orthologue Bud1 acts upstream of a Rho GTPase signaling pathway (Gulli and Peter, 2001), whereas, more recently it has been reported that constitutively active Rap1 antagonizes the activation of Rac1 in migrating epithelial cells (Valles et al., 2004).

In many cell types, the activation status of Rap1 has been shown to be spatio-temporally modulated by a large variety of extracellular stimuli, including growth factors, neurotransmitters and cytokines. Indeed, several, markedly distinct Rap1GEFs and Rap1GAPs with unique expression profiles and subcellular localization in different cells have been identified (Bos et al., 2001; Caron, 2003; Ohba et al., 2003). Furthermore, there is growing evidence that endocytic transport is important in regulating signal transduction and in mediating the formation of specialized signaling complexes (York et al., 2000; Sorkin and von Zastrow, 2002). In this context, our finding that E-cadherin endocytosis leads to the activation of Rap1 implicates E-cadherin internalization as a novel starting or convergence point for upstream signals that regulate Rap1 functions.

Internalization and recycling of cadherins has recently emerged as a major route for controlling AJ remodeling and maintenance (Le et al., 1999; Palacios et al., 2001; Fujita et al., 2002; Palovuori et al., 2003). There is evidence, in fact, that a small pool of cell surface E-cadherin is constitutively and constantly trafficked through endocytosis and recycling, and that this pool increases markedly in preconfluent cells and when cell-cell contacts are weakened or disrupted by distinct mechanisms, including manipulation of extracellular calcium and Src tyrosine kinase activation (Le et al., 1999; Palacios et al., 2001; Fujita et al., 2002; Palovuori et al., 2003; Pece and Gutkind, 2002). As we show, the activation of Rap1 triggered by E-cadherin endocytosis occurs upon E-cadherin endocytic transit to Rab11-positive recycling endosomes, raising the possibility that this event is implicated in the formation of specialized signaling complexes that modulate specific recycling processes. In this regard, it is remarkable that integrins also undergo endocytic recycling (Ng et al., 1999; Pierini et al., 2000; Powelka et al., 2004), and, in particular, that in epithelial cells internalized integrins have been recently shown to accumulate at the Rab11-positive recycling endosome, which mediates their rapid and targeted delivery to cell-matrix adhesion sites upon external stimulation (Powelka et al., 2004). Intriguingly, it has been consistently reported that Rap1 is activated mainly on endosomal compartments (York et al., 2000; Mochizuki et al., 2001; Ohba et al., 2003), whereas active Rap1 has been consistently implicated in mediating both endocytic and exocytic pathways for membrane proteins, including integrins and integrin function regulators (Zhu et al., 2002; Tohyama et al., 2003; Katagiri et al., 2003; Bivona et al., 2004; Dustin et al., 2004). In this respect, our finding that the E-cadherin endocytosis-dependent activation of Rap1 is associated with and required for the enhanced assembly of focal adhesions, suggests that Rap1 might mediate the crosstalk between cadherins and integrins by linking E-cadherin endocytosis to the control of integrin distribution and/or function at the plasma membrane. In support of this, it has been reported recently that the Rap1 regulation of integrin-mediated cell adhesion is sensitive to agents that block endosome recycling (Bivona et al., 2004). Further studies are underway to test this stimulating hypothesis.

Indeed, Rap1 might also regulate the endocytic recycling of cadherins to the plasma membrane. This is suggested by our finding that the E-cadherin endocytosis-dependent activation of Rap1 is coupled to the enhanced formation of Rap1/E-cadherin-catenin complexes that do not contain p120^ctn, a key regulator of E-cadherin stability at the cell surface (Peifer and Yap, 2003), combined with recent evidence suggesting that Rap1 activity is required for the recruitment of E-cadherin into nascent cell-cell contacts but not for the maintenance of mature cell-cell junctions (Price et al., 2004; Hogan et al., 2004). In this context, our findings would argue for a complementary, but not synergistic, role for Rap1 and p120^ctn in regulating E-cadherin turnover. In particular, whereas Rap1 might regulate the balance between endocytosis and exocytosis of E-cadherin, this balance might be shifted to accumulate either more or less E-cadherin on the plasma membrane depending on the functional state of stabilizing factors, such as p120^ctn. Alternatively, the putative role of Rap1 in the formation of E-cadherin-based cell-cell junctions might be an indirect consequence of the modulation of integrin adhesive function, as integrin-mediated adhesion is involved in the formation of new membrane protrusions required for the establishment of cell-cell contacts.

The catalytic activity of Src has been clearly implicated in the destabilization and turnover of epithelial cell-cell adhesions through the tyrosine phosphorylation of AJ components, including E-cadherin and p120^ctn, and/or the activation of the endocytic machinery underlying E-cadherin endocytosis (Frame, 2002; Fujita et al., 2002; Palovuori et al., 2003), whereas it has also been reported that Src activity is required for Rap1 function in promoting integrin-mediated adhesion (Li et al., 2002). Consistent with these opposing effects on E-cadherin- and integrin-mediated cell adhesion, it has been suggested that one major role of Src activity in normal epithelial cells is to cooperate with activated growth factor receptors to couple the disruption of cell-cell adhesions with the formation of prominent integrin-mediated cell-matrix adhesions (Avizienyte et al., 2002; Frame, 2002). In this context, our data suggest that Src activity may coordinate E-cadherin-mediated cell-cell and integrin-mediated cell-matrix adhesion by coupling the regulation of AJ remodeling with the modulation of Rap1 function at the crossroads between E-cadherin outside-in and integrin inside-out signaling. In particular, our findings that E-cadherin adhesive function modulates Src activity, and that this activity is in turn required for the activation of Rap1 in response to the disassembly of AJs suggest the existence of a cadherin function-dependent and Src-mediated signaling pathway that probably involves either GEFs or GAPs as important mediators for Rap1 activity regulation (Fig. 9). A likely GEF candidate is C3G, the only Rap1GEF that has definitively been linked to the tyrosine kinase signaling pathway (Ohba et al., 2001). Intriguingly, C3G has very recently been identified as a new binding protein for the cytoplasmic domain of E-cadherin, and shown to interact with E-cadherin when cell-cell contacts are weak or lost (Hogan et al., 2004). However, our initial attempt failed to detect endogenous C3G in Rap1/E-cadherin-catenin complexes in FRT cells during the calcium switch procedure, raising the possibility that additional GEFs may be involved. It is noteworthy that the identification of a β-catenin-interacting protein with a GEF activity for Rap1 has also been reported (Kawajiri et al., 2000). However, the reversibility of Rap1 activation that accompanies the reformation of AJs might implicate the recruitment and/or activation of GAPs for Rap1. Intriguingly, recent kinetic estimates show that Rap1GAP activity is high at the plasma membrane and low at endomembranes (Ohba et al., 2003). Thus, in the future it would be interesting to examine whether the modulation of E-cadherin adhesive function regulates the activity of specific GEFs and GAPs for Rap1, as well as whether Rap1GAPs located at plasma membrane may serve as signal terminators for the vectorial transport of Rap1-associated molecules from the Rab11-positive recycling endosome to specific plasma membrane sites, including sites of cell-cell and cell-matrix adhesion.

In summary, our results provide new important insights into Rap1 functions and indicate that E-cadherin may have a novel modulatory role in the regulation of integrin-based adhesive structures by fine-tuning Rap1 activation, with important implications for fundamental physiological and pathological processes, including morphogenesis and cancer. In fact, we can postulate that AJ disassembly needs to be associated with increased integrin-mediated cell-matrix adhesion in order to prevent uncontrolled cell dissemination, and it is tempting to speculate that Rap1 is a key regulator of this process.

We are grateful to R. Fässler (Max Planck Institute for Biochemistry, Martinsried, Germany), A. Sonnenberg and E. Danen (The Netherlands Cancer Institute, Amsterdam, The Netherlands) for providing the GE11 and GE11-β₁ cell lines, A. Wittinghofer (Max Planck Institute for Molecular Physiology, Dortmund, Germany) for providing the Rap1GAPv protein and cDNA, and C. Bucci (University of Lecce, Italy) for providing the pEGFP-Rab11 construct. We also want to thank M. Cutufia and M. Enrietto for technical support on the production of the mAb ECAD-FR1 (clone B11/5D1) against E-cadherin, S. Auinger, S. Graziano and F. Fraioli for helping in some experiments, and G. Tarone, V. Poli, S. Dewilde and S. Barbaro for critical reading of the manuscript and helpful discussion. This work was supported by grants from the University of Torino (60%–2003, 2004) and the MIUR (COFIN 2004) to S.F.R., the Austrian-Italian agreement for scientific collaboration (2004-2005) to S.F.R. and J.V.S. and the MIUR (COFIN 2002) to L.S.