Laminin-5-integrin interaction signals through PI 3-kinase and Rac1b to promote assembly of adherens junctions in HT-29 cells

Human intestinal cell differentiation is mediated by signaling pathways that remain largely undefined. We and others have shown that cell migration and differentiation along the crypt-villus axis is associated with temporal and spatial modulations of the repertoire, as well as with the function of integrins and E-cadherins and their substrates. Cross-talk between integrin and cadherin signaling was previously described and seems to coordinate this differentiation process. Here, we report that engagement of α6 and, to a lesser extent, α3 integrin subunits after HT-29 cell adhesion on laminin 5 increases the expression of E-cadherin, which then organizes into nascent adherens junctions. We further identify that phosphoinositide 3-kinase (PI 3-kinase) activation plays a key role in this cross-talk. Indeed, integrin-dependent adhesion on laminin 5 stimulates PI 3-kinase activity. Immunofluorescence and immunoprecipitation experiments revealed that activated PI 3-kinase is recruited at cell-cell contacts. Using LY294002, an inhibitor of PI 3-kinase activity, we found that this activation is essential for E-cadherin connection with the cytoskeleton and for biogenesis of adherens junctions. Finally, we demonstrated that PI 3-kinase could signal through Rac1b activation to control adherens junction assembly. Our results provide a mechanistic insight into integrin-cadherin cross-talk and identify a novel role for PI 3-kinase in the establishment of adherens junctions.


Introduction
described that cadherin engagement can also activate Rac in Madine-Darby Canine Kidney (MDCK) and Chinese Hamster Ovary (CHO) cells transfected with various cadherin isoforms (Nakagawa et al., 2001;Noren et al., 2001). Two reports have indicated that phosphoinositide 3-kinase (PI 3-kinase) was required for full Rac activation in this latter pathway. Signaling by PI 3-kinase is activated by a variety of extracellular stimuli, including growth factors and hemopoetic cytokines (Athie et al., 2000;Wymann and Pirola, 1998). Betson et al. demonstrated that epidermal growth factor (EGF) receptor signaling participates in the stimulation of Rac activity in keratinocytes upon junction formation (Betson et al., 2002). The cross-talk between cadherins and growth factor receptors might be analogous to the cross-talk between integrins and growth factor receptors that regulates cell survival and proliferation ( Giancotti and Ruoslahti, 1999) and might provide a link between integrin and cadherin signaling.
Signaling cross-talk between cadherins and integrins has been extensively investigated during some pathological processes including tumor progression (reviewed by Christofori, 2003). A correlation has been well established between breakage of intercellular junctions and loss of Ecadherin at the cell surface on the one hand and enhanced cell motility, invasiveness in vitro (Behrens et al., 1993;Vermeulen et al., 1995) or tumor progression in vivo (Perl et al., 1998;Toyoyama et al., 1999) on the other hand. Conversely, little is known about the mechanisms involved in the regulation of cadherin-mediated cellular interactions during embryonic development or cell differentiation (reviewed by Gumbiner, 1996). However, it has been described that, in migrating neural crest cells, ␤ 1 and ␤ 3 integrins are at the origin of a cascade of signaling events that ultimately control the surface distribution and activity of N-cadherin (Monier- Gavelle and Duband, 1997). Finally, in Caco-2 cells, integrins mediate functional cell polarization through complexes of E-cadherin and actin (Schreider et al., 2002); the signaling events that lead to this cross-talk have yet to be unraveled.
Since cell migration and differentiation along the cryptvillus axis are associated with dynamic modulations of the repertoire and function of adhesion receptors as well as their substrates (Beaulieu, 1997), intestinal epithelial cells provide a powerful paradigm for exploring integrin-cadherin cross-talk during renewal of the adult small intestine. In previous reports, we demonstrated that early enterocytic differentiation of the human adenocarcinoma cell line HT-29 is characterized by changes in the cell-adhesive properties and by laminin 5 secretion (Gout et al., 2004;Gout et al., 2001). Laminins represent the most abundant glycoproteins of the basement membrane and display the highest variability in their spatial and temporal expression either during intestinal development or in the adult (Simon-Assmann et al., 1994;Simon-Assmann et al., 1998;Teller and Beaulieu, 2001). Beside their fundamental role in organizing the basement membrane network, laminins promote several cellular processes such as adhesion, growth, polarization, differentiation and gene expression (Baker et al., 1996; Colognato and Yurchenco, 2000;De Arcangelis et al., 1996;Lampe et al., 1998; Vachon and Beaulieu, 1995). The distinct biological activities of laminins depend on both the isoform type and the repertoire of laminin receptors expressed.
The observation that laminin 5 is expressed both in vivo and in vitro during intestinal cell differentiation suggests that this matrix protein could participate in the differentiation process. Accordingly, in this study, we addressed the potential contribution of laminin 5-integrin interactions in the initiation of HT-29 cell differentiation and the resulting signaling events involved in this process. Here, we report that engagement of ␣ 6 and, to a lesser extent, ␣ 3 integrin subunits after HT-29 cell adhesion on laminin 5 increases the expression of E-cadherin, which organizes into nascent adherens junctions. Moreover, we showed that laminin 5 binding to integrins mediated a PI 3kinase-dependent activation of Rac1b that is involved in the assembly of adherens junctions, an initial step in enterocyte differentiation.

Results
Laminin 5 induces E-cadherin-mediated intercellular adhesion of HT-29 cells It has previously been reported that human colon carcinoma cells synthesize laminin 5 (Gout et al., 2004;Orian-Rousseau et al., 1998). Indeed, we have previously shown that HT-29 cells cultured for ten days in the differentiating medium produced a fivefold higher amount of laminin 5 than cells cultured in standard medium (Gout et al., 2001). Therefore, we hypothesized that laminin 5 could mimic the effect of the differentiating medium, and initiate HT-29 cell differentiation. Morphological changes associated with modification of intercellular junctions rapidly occur during HT-29 cell differentiation (Gout et al., 2004). Observed under phase contrast, HT-29 cells grow as homogenous cell layers with well-defined intercellular junctions. By contrast, HT-29 cells cultured in the differentiating medium for 10 days (HT-29 Gal) adopt a multicellular spheroid configuration leaving large cell-free spaces (Fig. 1A) owing to an increase in cell-cell adhesiveness. HT-29 cells cultured for 24 hours on laminin 5 in standard medium (HT-29 LN5) already adopt a compact organization similar to those observed for the differentiating cells, suggesting that laminin 5 might actually participate in this process. Therefore, using an in vitro cell aggregation assay, we compared the ability of HT-29 cells and of HT-29 cells cultured for 48 hours on laminin 5 to establish cell-cell contacts (Fig. 1B). We found that, in the presence of Ca 2+ , HT-29 cells spread on culture plastic aggregated poorly compared with HT-29 cells cultured on laminin 5 (Fig. 1B, upper panel). Laminin 5induced aggregation of HT-29 cells was inhibited by depletion of extracellular cation (EDTA) as well as by Ecadherin-blocking antibodies (HECD1; Fig. 1B, lower panel). These observations indicated that culture of HT-29 cells on laminin 5 favored intercellular adhesion mediated by E-cadherin as observed during enterocyte differentiation.

Upregulation of E-cadherin expression upon laminin 5integrin interactions
To address further the role of laminin 5 in the regulation of E-cadherin-mediated cell-cell adhesion, we studied the expression of E-cadherin in HT-29 cells grown on this matrix protein. We observed that the total amount of E-cadherin protein gradually increased to reach a plateau after 24 hours of culture on laminin 5 ( Fig. 2A). Thereby, we subsequently used this time of culture in the following experiments. Using purified laminin 5, we showed that the effect of laminin 5 on E-cadherin expression is dose dependent, reaching a maximum around 5 g/ml and then gradually declining (Fig. 2B).
To determine whether upregulation of E-cadherin was specifically induced by laminin 5, we grew HT-29 cells for 24 hours on different ECM proteins, and the amount of E-cadherin in the cell lysates was estimated. Fig. 2C shows that laminin 5 was able to induce a significant and rapid increase in Ecadherin expression over a period of 24 hours. Culture on fibronectin allowed a limited and slower increase in E-cadherin levels whereas the kinetics of E-cadherin expression induced by other ECM proteins was similar to the one observed on plastic.
Adhesion of epithelial cells to laminin 5 mainly involves ␣ 3 ␤ 1 and ␣ 6 ␤ 4 integrins. To establish further the involvement of laminin 5 in the control of E-cadherin upregulation, we carried out similar experiments in the presence of functional blocking antibodies directed against laminin 5 (␣ 3 chain) or the ␣ 2 , ␣ 3 , ␣ 6 , ␤ 1 and ␤ 4 integrin subunits (Fig. 2D). The laminin 5-mediated increase in E-cadherin expression was completely inhibited by antibodies blocking ␣ 6 and ␤ 4 integrins. Whereas antibodies blocking ␣ 3 and ␤ 1 integrins inhibited the Ecadherin upregulation by 56% and 44%, respectively, no significant effect of the ␣ 2 -subunit-blocking antibody was observed. The laminin 5 ␣ 3 -subunit-blocking antibody also inhibited the expression of E-cadherin by 55%. Altogether, these results demonstrate that the interaction of laminin 5 with the integrin receptors ␣ 6 ␤ 4 and, to a lesser extent, ␣ 3 ␤ 1 increases the expression of E-cadherin.
Laminin 5-mediated cell-cell interaction is correlated with the association of E-cadherin with the actin cytoskeleton together with ␤-catenin but not p120 ctn Contrary to what was observed for E-cadherin, we found that the total amount of ␤and p120 catenins expressed in HT-29 cells did not vary during the culture on laminin 5 (Fig. 3). In intact cells, stable adherens junctions require the recruitment of both ␤-catenin and p120 ctn on distinct sites of the cytoplasmic domain of E-cadherin, and a linkage of the cadherin-catenin complexes to the actin cytoskeleton. This linkage can be assayed by the decreased solubility of the cadherin-catenin complex in Triton X-100 (Hinck et al., 1994). To examine the presence of actin-associated E-cadherin in HT-29 cells grown on laminin 5, we looked at the distribution of E cadherin, ␤and p120 catenins into Triton-insoluble (TI) and Triton-soluble (TS) fractions. As shown in Fig. 4A, E-cadherin was more abundant in the TS fraction in HT-29 cells. Scanning of autoradiograms showed that the proportion of E-cadherin in the TI fraction was only about 10% (HT-29). After 24 hours of culture on laminin 5 (HT-29 LN5), E-cadherin levels increased in the cytoskeleton-associated fraction and reached 33%. Similar results were obtained with ␤-catenin. Quantitative analysis performed on five independent experiments indicated a 3.2-and 3-fold increase in E-cadherin and ␤-catenin association with the actin cytoskeleton (TI fraction), respectively. These results were fairly consistent with the wellknown 1:1 stoichiometric association of these proteins in adherens junctions. By contrast, the fraction of p120 ctn associated with the actin cytoskeleton did not vary with the conditions of culture (Fig. 4A). The subcellular distribution of E-cadherin was further characterized by confocal analysis. Confocal Z axis imaging showed that E-cadherin was not concentrated into apicolateral cell junctions in control HT-29 cells, conversely to what was observed in HT-29 cells cultured on laminin 5 (Fig. 4B, upper panel). With those cells, the junctions were also enriched in F-actin that was fairly colocalized with E-cadherin (Fig. 4B, medium and lower panels), as revealed by the merge yellow signal. These experiments indicate that the gain in cell adhesion observed in HT-29 cells cultured on laminin 5 is accompanied by the engagement of E-cadherin and its redistribution together with actin at the cell-cell contacts.  , in differentiating medium for 10 days (HT-29 Gal), or on laminin 5 matrix in standard medium for 24 hours (HT-29 LN 5). Bars, 100 m. (B) Cell aggregation assays. Upper panel: HT-29 cells grown on plastic (PL) or laminin 5 matrix (LN 5) for 48 hours were resuspended in PBS containing 10 mM Hepes + 2 mM CaCl 2 . Suspensions were incubated for 60 minutes at 37°C in a gyratory shaker at 75 rpm. Cell aggregation was observed under phase contrast with a Zeiss Axiovert 135 inverted microscope. Lower panel: HT-29 cells grown for 48 hours on laminin 5 matrix were resuspended in PBS containing 10 mM Hepes + 2 mM CaCl 2 in the presence of 4 mM EDTA or 10 g/ml blocking anti-human Ecadherin (HECD1) antibody and then assayed for cell aggregation. Aggregation of HT-29 cells cultured on laminin 5 matrix was mediated by E-cadherin. Bars, 100 m.
Integrin-dependent cell adhesion to laminin 5 activates PI 3-kinase Recent data suggest a potential role for PI 3-kinase in the regulation of actin cytoskeleton dynamics and E-cadherin-␤catenin complexes (reviewed by Goodwin et al., 2003). Additionally, this kinase has been described as an intermediate of some integrin-mediated signaling pathways (Braga et al., 1997;Khwaja et al., 1997). Consequently, PI 3-kinase could be a mediator of the integrin-cadherin cross-talk. To test this hypothesis in intestinal epithelial cells, we used a lipid kinase assay to estimate PI 3-kinase activation directly by different ECM components (Fig. 5). Briefly, endogenous PI 3-kinase was immunoprecipited with an anti-p85 regulatory subunit antibody and its activity was visualized by the synthesis of phosphatidylinositol (3,4,5)-trisphosphate [PtdIns(3,4,5)P 3 ] (Vanhaesebroeck et al., 2001). When the cells were cultured for 24 hours on laminin 5, the pulled-down proteins showed a significant increase in PI 3-kinase enzymatic activity compared with cells cultured on collagen type IV, fibronectin or plastic (Fig. 5A). The PI 3-kinase activity was inhibited by the presence of 20 M LY294002 during the enzymatic assay (data not shown). To demonstrate further that spreading of HT-29 cells on laminin 5 induced the activation of PI 3-kinase, we tested the effect of the anti-laminin 5 ␣3 subunit blocking mAb (BM165) on PI 3-kinase activation. The laminin 5-dependent increase in PtdIns(3,4,5)P 3 synthesis was completely abolished when cells were cultured on laminin 5 in the presence of BM165 (Fig. 5A). ECM proteins differentially increase expression of E-cadherin. HT-29 cells were grown for various times (0-24 hours) on plastic (PL) or different ECM proteins used at 10 g/ml (FN, fibronectin; CO IV, collagen type IV; LN 5, laminin 5). Cell lysates were assayed for Ecadherin expression, using actin as an internal control. The density ratio of E-cadherin:actin is expressed as a relative factor of the value obtained from HT-29 cells in suspension (mean ± s.e.m. of three independent experiments). (D) Increase in E-cadherin expression is mediated by laminin 5-integrin interactions. Western blot analysis of the total amount of E-cadherin in HT-29 cells cultured for 24 hours on plastic (PL) or 5 g/ml purified laminin 5 (LN5) in the presence of mAbs directed against integrin subunits at a concentration of 10 g/ ml: P4C10 (anti-␤ 1 ), ASC8 (anti-␤ 4 ), BHA2.1 (anti-␣ 2 ), P1B5 (anti-␣ 3 ) and GoH3 (anti-␣ 6 ) or laminin 5 (␣ 3 chain), with a 1/10 diluted BM165 mAb (from cell culture supernatant). Expression of actin was used as a control. Band densitometry was normalized with actin to allow quantification and the % of control (i.e. laminin 5-induced E-cadherin expression without mAbs) was reported on the histogram. Data represent mean ± s.e.m. of three independent experiments. *P<0.05; **P<0.01, as determined by Student's t-test. Laminin 5-mediated increase in E-cadherin expression was dependent on ␣ 6 ␤ 4 and ␣ 3 ␤ 1 integrins.

Journal of Cell
Time-course experiments performed on HT-29 cells cultured on laminin 5 indicated that the PI 3-kinase activity gradually increased to reach a maximum at 24 hours of adhesion. For longer incubation times, this enzymatic activity fell rapidly (Fig. 5B). Similar assays were performed on collagen IV and fibronectin. Whatever the time of adhesion tested on these ECM proteins, there was no variation in the PtdIns(3,4,5)P 3 production (data not shown). Altogether, these data suggested that laminin 5 specifically induces the activation of PI 3-kinase in HT-29 cells.
Both integrin and cadherin engagements have been described to signal through PI 3-kinase. Furthermore, culture on laminin 5 results in both integrin and cadherin engagement.
To determine the respective contribution of integrin and Ecadherin receptors in the induction of PI 3-kinase activity, PI 3-kinase assays were performed after 24 hours of cell adhesion onto laminin 5 in the presence of blocking antibodies raised against specific integrin subunits or E-cadherin (Fig. 5C).
Blocking E-cadherin ligation by either antibody or Ca 2+ chelation with EGTA did not abolish laminin 5-dependent activation of PI 3-kinase activity in our biological system. Conversely, the mAbs blocking ␣ 6 and ␤ 1 inhibited the laminin 5-induced PI 3-kinase activity by approximately 96% and 73%, respectively. By contrast, the ␣ 3 mAb had no significant effect. These observations suggested a major involvement of integrin receptors in laminin 5-dependent PI 3-kinase activation. Moreover, activation of PI 3-kinase is probably mediated by the binding of laminin 5 to ␣ 6 ␤ 4 integrin since it has been reported previously that ␣ 6 associates preferentially with ␤ 4 compared with ␤ 1 when both subunits are expressed in the same cell (Lee et al., 1992;Lotz et al., 1990). Therefore, the inhibition of PI 3-kinase activity by the anti-␤ 1 mAb is probably associated with a global inhibition of cell adhesion under these conditions, although we cannot exclude the participation of ␣ 6 ␤ 1 integrin. We also observed a decrease in PI 3-kinase activity when cells were maintained in suspension (data not shown). Taken together, these data indicate that, in HT-29 cells, laminin 5 binding to ␣ 6 -containing integrin results in the activation of PI 3-kinase.

PI 3-kinase controls E-cadherin-dependent cell-cell adhesion
The combined observations that laminin 5 upregulates Ecadherin expression, favors its linkage to the cytoskeleton and activates PI 3-kinase activity suggested that PI 3-kinase signaling upon integrin ligation might affect E-cadherin availability and recruitment at cell-cell contacts. Consistent with this idea, we found that the inhibition of PI 3-kinase by LY294002 (50 M) had no effect on E-cadherin expression in HT-29 cells cultured on plastic (not shown) but abolished laminin 5-induced upregulation of E-cadherin expression (Fig.  6A). The effect of LY294002 was almost maximal at 50 M (and did not significantly increase at higher concentrations such as 100 M), a dose previously described to provide a 100% blockade of PI 3-kinase in all cells. The role played by PI 3-kinase signaling in laminin 5-dependent E-cadherin upregulation was confirmed by expression of a dominantnegative construct of the p85 subunit, lacking the inter-SH2 domain (⌬p85; Fig hours of culture on laminin 5; by contrast, in HT-29 cells expressing ⌬p85, cell interactions were weaker and Ecadherin/actin colocalisation was decreased. Taken together, these results indicate that PI 3-kinase participates in the regulation of E-cadherin expression and favors its association with the cytoskeleton.

Laminin 5 promotes direct interaction between PI 3kinase and E-cadherin
Agonist activation of PI 3-kinase frequently involves the translocation of this enzyme to the plasma membrane where it can gain access to its lipid substrates (reviewed by Wymann and Pirola, 1998). Thus, it was tempting to speculate that a similar mechanism might underlie PI 3-kinase activation upon adhesion to laminin 5. Analysis of PI 3-kinase distribution after HT-29 cell fractionation indicated that the p85 regulatory subunit of PI 3-kinase was distributed in both the cytosol and at the membrane in cells cultured on plastic. Culture on laminin 5 induced a significant redistribution of PI 3-kinase at the cell membrane ( Fig. 7, upper panel). The membrane shift in PI 3kinase localization was further confirmed by confocal immunofluorescence performed on HT-29 cells labeled with antibodies raised against Ecadherin (green) and the p85 regulatory subunit of PI 3-kinase (red). In subconfluent HT-29 cells cultured on plastic, E-cadherin was detected at the sites of cell-cell contacts, whereas p85 staining was mostly cytoplasmic (Fig. 7, middle panel) without any colocalization with E-cadherin. When the cells were cultured on laminin 5, p85 was still partially localized in the cytoplasm, but was also clearly located at the sites of cell-cell contacts with a significant colocalization with E-cadherin. This pattern was only observed in confluent cells in which E-cadherins are engaged. The colocalization of E-cadherin and PI 3-kinase suggested a possible interaction between these proteins. This association was assayed after 24 hours of culture on plastic or on laminin 5, by coimmunoprecipitation performed with whole-cell lysates of HT-29 cells (Fig. 7, lower panel). Ecadherin-p85 association in HT-29 cells was enhanced when cells were cultured on laminin 5. These findings suggest that the PtdIns(3,4,5)P 3 signal is being generated in a spatially confined region of the E-cadherin contact zones, namely in the nascent adherens junctions before the association of the cadherin-catenin complexes with the cytoskeleton.
PI 3-kinase signaling controls the biogenesis of adherens junctions Using the specific PI 3-kinase inhibitor LY294002 (50 M), we further assayed the role of PI 3kinase in the biogenesis of adherens junctions in HT-29 cells. We performed E-cadherin and F-actin staining in a two-step experiment in which adherens junctions of HT-29 cells cultured for 24 hours on laminin 5 were disrupted by chelating extracellular Ca 2+ with EGTA and were subsequently allowed to re-assemble upon restoration of extracellular Ca 2+ (Pece et al., 1999), in the absence or presence of LY294002 (Fig. 8). In control cells, E-cadherin colocalized with F-actin (panel A). After EGTA treatment for 30 minutes, cells adopted a more rounded shape Journal of Cell Science 119 (1) with some spaces between cells; E-cadherin staining formed a diffuse ring at the cell periphery and cortical actin was mostly depolymerized and, therefore, poorly stained with TRITC-conjugated phalloidin (panel B).
Following an incubation period of 60 minutes with Ca 2+ , E-cadherin redistributed at the sites of cell-cell contacts and was colocalized with F-actin while the cells reacquired their epithelial shape (panel C), suggesting that adherens junctions were being reformed. In LY294002-and EGTA-treated cells, we noticed a drastic effect on E-cadherin staining that was weaker and remained partially diffuse in the cytoplasm (panel D compared with panel B). Cortical actin was partly depolymerized and Ca 2+ restoration neither stimulated actin polymerization nor favored cell epithelial shape: E-cadherin staining was still weaker and diffuse (panel E compared with panel C). By contrast, the addition of the PI 3-kinase inhibitor on pre-existing cell-cell contacts (i.e. not disrupted by EGTA), neither affected E-cadherin nor F-actin expression and localization (data not shown). We concluded that PI 3-kinase activity was probably required for the biogenesis of E-cadherin-mediated cell-cell contacts rather than for the maintenance of pre-existing adherens junctions.
PI 3-kinase acts upstream of Rac1 in the laminin 5-integrin signaling pathway Recent studies support the view that Ecadherin engagement coincides with an increase in Rac1 activity that depends on PI 3kinase (Nakagawa et al., 2001;Pece et al., 1999). To test whether this pathway could be generalized to HT-29 cells, we first determined the impact of PI 3-kinase activity on Rac1 activation (Rac-GTP) by pull-down assays. A GST fusion protein containing the CRIB domain of PAK was used to isolate active GTP-bound Rac in lysates from cells cultured on laminin 5 or plastic, respectively (Fig. 9). Using anti-Rac1 antibodies, we observed a slight but reproducible increase in the GTPbound Rac1 fraction during the first 24 hours of culture on laminin 5 before reaching a plateau. Conversely, after 72 hours of culture on plastic, we were not able do detect any significant increase in the amount of activated Rac1 (Fig. 9A, upper panel). The anti-Rac1 mAb recognizes both Rac1 and Rac1b, a spliced variant that was predominantly identified in skin and epithelial tissues from the intestinal tract (Jordan et al., 1999). Hence, we performed the experiment with an antibody directed against the specific 19 amino acid stretch of Rac1b. We observed an important rise in Rac1b-GTP after 24 hours of culture on To test the specificity of the interaction with laminin 5, HT-29 cells were preincubated with the anti-␣3 chain of laminin 5 BM165 at a concentration of 20 g/ml and then allowed to adhere to laminin 5. (B) Time course of PI 3-kinase activation by laminin 5 matrix. HT-29 cells were allowed to adhere for 1, 4, 24 or 48 hours to dishes coated with laminin 5 matrix. In (A) and (B), total cell lysates were prepared and equivalent amounts of proteins were used in immunoprecipitation and kinase assays. Western blots performed on immunoprecipitated p85 subunit were carried out as controls. Phosphorylated lipids were resolved by thin layer chromatography. The amount of radiolabeled PtdIns(3,4,5)P 3 (PIP3) was determined for each condition by densitometry with ImageQuant software. Data represent the average of percentage of control activity (mean ± s.e.m. of three to four independent experiments) and one representative chromatogram is shown. *P<0.05; **P<0.01, as determined by Student's t-test. (C) HT-29 cells were incubated with 2 mM EGTA or 10 g/ml of integrin or E-cadherin-specific antibodies and allowed to adhere to plates coated with laminin 5 matrix for 24 hours. mAbs specific for ␣ 3 (P1B5), ␣ 6 (GoH3) or ␤ 1 (P4C10) integrin subunits and E-cadherin (HECD1) were used at 10 g/ml. Total cell lysates were prepared and equivalent amounts of proteins were subjected to immunoprecipitation and kinase assays as described previously. **P<0.01, as determined by Student's t-test.
laminin 5 followed by a plateau, similarly to what was observed with the anti-Rac1 antibody (Fig. 9A, upper middle panel). This indicated that laminin 5 predominantly activates the Rac1b isoform in HT-29 cells. Western blots performed in parallel showed no differences in the total Rac content in the cell lysates; however, Rac1 expression in HT-29 cells was higher than Rac1b (Fig. 9A, lower middle panel). Finally, to evaluate whether PI 3-kinase controls Rac1b activation upon laminin 5-cell adhesion, we carried out Rac1b-GTP pull-down assays under conditions of PI 3-kinase inhibition (Fig. 9B). Treatment with LY294002 at concentrations that efficiently inhibited E-cadherin expression and biogenesis of adherens junctions also inhibited laminin 5-mediated increase in Rac1b-GTP. This indicated that PI 3-kinase acts upstream of Rac1b in the laminin 5-integrin signaling pathway.
It is noteworthy that the extra 19 amino acid long domain of Rac1b appears to confer selectivity towards downstream Rac signaling targets. In particular, it was found that Rac1b could neither stimulate the formation of lamellipodia nor the activation of the protein kinase JNK (Matos et al., 2003). Consequently, we looked at the distribution of Rac1b in HT-29 cells. We found that Rac1b was mainly distributed as distinct spots at the cell surface and that adhesion to laminin 5 concentrated these structures at cell-cell contacts (Fig. 9C).
Using the specific PI 3-kinase inhibitor LY294002 (50 M), we further assayed the role of PI 3-kinase-mediated activation of Rac1b in the biogenesis of adherens junctions in HT-29 cells (Fig. 10). In control cells, we found that Rac1 was mainly distributed as distinct spots at cell-cell contacts (panel A). After an incubation of 30 minutes in EGTA, the cells rounded and Rac1 was no longer distributed at cell-cell junctions and some punctuated staining was observed in the cytosol (panel B). Following a period of 60 minutes in Ca 2+ to restore E-cadherin activity, Rac1 redistributed at the sites of cell-cell contacts Journal of Cell Science 119 (1) were grown on plastic or laminin 5 matrix for 24 hours. E-cadherin and F-actin were detected by immunofluorescence using HECD1 antibody and TRITC-conjugated phalloidin, respectively. Optical sections were performed using a confocal microscope. Bars, 50 m.
where adherens junctions were being reformed (panel C). However, in LY294002-and EGTA-treated cells, Ca 2+ could not restore the original distribution of Rac1. The staining was weaker and remained diffuse all over the cell, similar to what was observed in the absence of Ca 2+ (panel E).
These findings demonstrate that cell adhesion on laminin 5 is sufficient to activate Rac1b signaling through PI 3-kinase stimulation. Rac1b accumulates at the cell-cell contacts where it may participate in the laminin 5-mediated biogenesis of adherens junctions in HT-29 cells.

Discussion
Laminin 5 and upregulation of E-cadherin Laminin 5 has been described to promote gap junction assembly and intercellular communication in three different cultured cell populations through ␣ 3 ␤ 1 engagement (Lampe et al., 1998). In many cell types, the assembly of gap junctions, tight junctions or desmosomes depends on cell-cell adhesions mediated by cadherins (Cereijido et al., 2000;Fujimoto et al., 1997;Jongen et al., 1991;Musil and Goodenough, 1990). Therefore, it is conceivable that interactions of integrin with laminin 5 could trigger the assembly of adherens junctions.
Culture of HT-29 cells on laminin 5 leads to a threefold increase in the total amount of E-cadherin and thereby favors E-cadherin-mediated cell aggregation. Indeed, it has been demonstrated that a twofold increase in the neural cell adhesion molecule N-CAM (which is also involved in homophilic binding), resulted in a >30-fold increase in the rate of aggregation (Hoffman and Edelman, 1983). The mechanism responsible for E-cadherin upregulation is under investigation. However, our data indicate that it partly depends on PI 3-kinase activation. We showed that activated PI 3-kinase is recruited to the plasma membrane at sites of cadherin-mediated cell adhesion. At this stage, PI 3-kinase is associated with Ecadherin, possibly indirectly through ␤-catenin (Goodwin et al., 2003;Woodfield et al., 2001). This interaction might stabilize ␤-catenin binding to E-cadherin, which is usually disrupted by presenilin. Presenilin plays a role in the Ecadherin degradation process and binds the E-cadherin cytoplasmic domain (Baki et al., 2001).
Using antibodies that block integrin subunits, we found that laminin 5 upregulation of E-cadherin expression was predominantly mediated by ␣ 6 ␤ 4 and to a lesser extent by ␣ 3 ␤ 1 , the two major receptors for laminin 5. The lower participation of ␣ 3 ␤ 1 to the process might be explained by the fact that, in HT-29 cells, activation of ␣ 3 ␤ 1 is not constitutive but is activated by the secreted laminin 5 (Gout et al., 2001). Furthermore, in keratinocytes, ␣ 6 ␤ 4 activation antagonized ␣ 3 ␤ 1 integrin-mediated adhesion and relocated this integrin from sites of basal clustering where it displayed increased conformational activation to cell-cell contacts (Russell et al., 2003). This process might also occur in HT-29 cells. The more effective contribution of ␣ 6 ␤ 4 to the regulation of E-cadherin expression and function when HT-29 cells adhere onto laminin 5 is in agreement with the work of Hintermann and colleagues: in their system, the integrin ␣ 6 ␤ 4 can stimulate cell-cell aggregation, increase colony size in plated HaCaT keratinocytes, and upregulate E-cadherin localization at cellcell contacts (Hintermann et al., 2005). This effect was mediated by an increase in erbB-2 and PI 3-kinase activity, which in turn inhibits ␣ 3 ␤ 1 -controlled keratinocyte hypotaxis on laminin 5 (Hintermann et al., 2001). Thus, binding to laminin 5 through ␣ 3 ␤ 1 in focal contacts or ␣ 6 ␤ 4 in hemidesmosomes transmits distinct molecular signals, which support cell migration and static adhesion respectively (Goldfinger et al., 1999;Nguyen et al., 2000b). By favoring both adherens junction-controlled cell-cell adhesion and hemidesmosome-mediated stable cell-ECM interactions, the integrin ␣ 6 ␤ 4 appeared to be a key integrin in the control of epithelial cell motility. This property might be required to maintain cellular polarity and cohesion during epithelial cell migration along the crypt-villus axis. In this context, disruption of the Lama3 gene, which encodes the ␣ 3 subunit of laminin 5 in mice, reveals abnormalities in survival and late-stage differentiation of epithelial cells (Ryan et al., 1999). It has been suggested that the resulting hemidesmosome alterations (consecutive to the discontinuity in protein localization of ␣ 6 ␤ 4 and BP230) affect the stability of cell-cell junctions. The phenotype obtained in these mice was similar to a lethal variant of the human disease epidermolysis bullosa, JEB-G, whose clinical features include mechanical fragility of the skin, oral erosions, gastrointestinal and genitourinary tract involvement, hypoplastic hemidesmosomes and high morbidity (Fine et al., 1991). JEB-G can be caused by additive mutations in genes encoding laminin chains (LAMA3, LAMB3, LAMC2) or collagen XVII, thus suggesting that combined alterations of different isoforms of matrix proteins might suppress compensation of each other for the cell organization and function, thus leading to this pathology (Castiglia et al., 2001;Floeth and Bruckner-Tuderman, 1999). However, no direct reports on defects in cadherin-mediated adherens junctions in mice or humans with null defects in laminin 5 have been published, probably because other ECM proteins and receptor signaling pathways might control in vivo expression and localization of E-cadherin at adherens junctions. By contrast, the half-life of junctional proteins in cultured epithelial cells is about 5 hours ( Shore and Nelson, 1991). The assembly of newly synthesized proteins into the junction must balance this rapid turnover. By efficiently regulating the expression of E-cadherin, laminin 5 might favor assembly of cell-cell adhesion complexes.
Laminin 5 and adherens junction assembly Initial cell-cell contacts are formed by the engagement of two opposing E-cadherin-␤-catenin complexes at the tips of filipodia and/or lamellipodia projections. Then anchoring of cadherin-catenin complexes to the cortical actin cytoskeleton promotes clustering and stabilization of the junction proteins to form adjacent punctua organized as a zipper-like structure, which later 'zips' to seal the membranes into mature epithelial-sheet adherens junctions (Vaezi et al., 2002;Vasioukhin et al., 2000). The cytosolic protein p120 ctn interacts with the membrane-proximal domain of classical cadherins and has been suggested to participate in the zipper stage by regulating cadherin clustering (Yap et al., 1998).
Using biochemical analysis and confocal microscopy, we have shown that, after 24 hours of culture on laminin 5, E-cadherin colocalizes with actin at the apicolateral side of the cells where adherens junctions take place. This observation suggests that laminin 5 is not only involved in the control of E-cadherin expression, but also mediates signaling that controls post-translational events. Indeed, involvement of ECM proteins in the reinforcement of E-cadherin-actin complexes has been described in Caco-2 cells (Schreider et al., 2002). In HT-29 cells, our results suggest that laminin 5 favors the formation of stable punctum corresponding to the immature E-cadherin-␤catenin complexes anchored to the actin cytoskeleton but not mature junctions. The parallel increase in both ␤-catenin and E-cadherin associated with the actin cytoskeleton is consistent with the well-known stoichiometric interaction between these proteins and correlates with the fact that ␤catenin to E-cadherin association is required for the transport of the newly synthesized E-cadherin to the plasma membrane (Chen et al., 1999). By contrast, association of p120 ctn with the cytoskeleton did not vary with the conditions of culture, suggesting that this catenin is not recruited to adherens junctions at this stage. This hypothesis is further confirmed by coimmunoprecipitation experiments showing that only 8% of total p120 ctn is complexed to E-cadherin (even upon adhesion to laminin 5), and correlates with the inability of these junctions to seal the membranes into epithelial sheets, thus leading to biotin diffusion (data not shown).
How does PI 3-kinase control assembly of adherens junctions? PI 3-kinases are known to play a central role in several cellular processes, including mitogenic signaling, cell survival, cytoskeletal remodeling, as well as metabolic control and Journal of Cell Science 119 (1) Fig. 8. PI 3-kinase signaling is necessary for adherens junction biogenesis. HT-29 cells were cultured for 24 hours on laminin 5 matrix and subsequently treated with 4 mM EGTA for 30 minutes in the absence or presence of 50 M LY294002. The EGTA-containing medium was then replaced with Ca 2+ -containing medium for 60 minutes. Cells were fixed and stained with anti-E-cadherin (green) and TRITCconjugated phalloidin (red) before confocal microscopy analysis on optical sections. Bars, 25 m. vesicular trafficking (reviewed by Wymann and Pirola, 1998). Adhesion to laminin 5 through ␣ 3 ␤ 1 activates PI 3-kinase and regulates cell spreading and migration (Choma et al., 2004;Enserink et al., 2004). This effect could be modulated by ␣ 6 ␤ 4 adhesion (Nguyen et al., 2000a;Russell et al., 2003). Our data clearly indicate that laminin 5-and ␣ 6 ␤ 4 -mediated activation of PI 3-kinase is further required for adherens junction biogenesis in HT-29 cells. This fits with previous studies indicating that the ␣ 6 integrin subunit associated with ␤ 4 rather than ␤ 1 when these subunits are expressed in the same cell (Lee et al., 1992;Lotz et al., 1990) and that the PI 3-kinase pathway is better activated by ␣ 6 ␤ 4 than ␣ 6 ␤ 1 or other ␤ 1 integrins (Shaw et al., 1997). Our findings that PI 3-kinase activity regulates the recruitment of F-actin at cell-cell contacts and the connection of E-cadherin with the cytoskeleton are consistent with other data (Laprise et al., 2002;Somasiri et al., 2000). However, this process is usually attributed to cadherin ligation (Kovacs et al., 2002;Laprise et al., 2002;Nakagawa et al., 2001). Recent studies suggest that cadherin ligation alone is not sufficient and that additional cooperating signals from tyrosine kinase receptors are required (Betson et al., 2002;Pang et al., 2005).

Small Rho-GTPases have distinct functions during assembly of adherens junctions
Integrin-mediated adhesion to laminin 5 induces signaling pathways that regulate the activities of Rho GTPases, which in turn control actin dynamics. Depending on the type of integrins and ECM proteins, various G proteins can be activated (reviewed by Etienne- Manneville and Hall, 2002). For example, laminins 10 and 11 are more active than fibronectin in promoting cell migration by regulating ␣ 3 ␤ 1 -dependent Rac activation by the p130cas-crkII-DOCK180 pathway (Gu et al., 2001). By contrast, cells adhering to fibronectin develop stress fibers and focal contacts by regulating integrin-dependent Rho activation. On laminin 5, ␣ 3 ␤ 1 -dependent activation of Rac1 contributes to the formation of stable lamellipodia necessary for cell migration (Choma et al., 2004). In keratinocytes, deposit of laminin 5 induces a change in signaling from a Rhoto a PI 3-kinase-dependent pathway (Nguyen et al., 2000a).
This switch has already been described to participate in organization of adherens junctions (Kovacs et al., 2002;Pece et al., 1999). Here, we have shown that Rac1b activation is triggered by integrin ligation rather than by engagement of cadherins. Rac1b was discovered in human tumors as an alternative splice variant of Rac1 (Jordan et al., 1999). Since Rac1b is unable to interact with Rho-GDI, it is constitutively associated with membranes (Matos et al., 2003). This localization leaves the Rac1b variant in a favorable spatial position to become activated by exchange factors, such as Tiam1. We demonstrated that, upon adhesion to laminin 5, Rac1b is redistributed at cell-cell contacts where adherens junctions organize. Rac1b is not involved in lamellipodia formation but is activated in the presence of low stimuli, which fail to activate Rac1 (Matos et al., 2003). In this way, our data suggest that, whereas a more persistent activation of Rac1 by PI 3-kinase can be involved in lamellipodia formation, a transient activity of this kinase upon integrin engagement might specifically control adherens junction assembly through the activation of Rac1b.

Role of PI 3-kinase in intestinal cell differentiation
Data presented by Laprise et al. (Laprise et al., 2002) suggest that PI 3-kinase plays a crucial role in the control of differentiation since the inhibition of PI 3-kinase in Caco-2/15 cells represses sucrase-isomaltase and villin protein expression, both of which are markers of enterocytic differentiation. However, in HT-29 cells, laminin 5, which activates PI 3-kinase, is unable to induce full cell differentiation (data not shown). This discrepancy might be explained by the fact that contrary to HT-29 cells, Caco-2 cells at confluence produce laminin 1, which has been previously described to induce both enterocytic differentiation (Basson et al., 1996;De Arcangelis et al., 1996;Vachon and Beaulieu, 1995) and PI 3-kinase activation (Shaw et al., 1997). The HT-29 cell model indicates that, although necessary, PI 3kinase signaling is not sufficient for full differentiation, which requires additional factors. The view that PI 3kinase only initiates the early stages of cell differentiation is consistent with the transient activation of this lipid kinase by laminin 5. Wang et al. showed that PI 3-kinase inhibition through overexpression of the antagonist phosphatase PTEN or through wortmannin treatment resulted in enterocytic differentiation (Wang et al., 2001). Indeed, during differentiation of HT-29 cells induced by glucose starvation, we also observed a rapid decrease in PI 3kinase activity yet these cells produced laminin 5 (data not shown). Therefore, a precise temporal regulation of PI 3-kinase seems to be required during the whole enterocytic differentiation process. Sustained activation of PI 3-kinase might contribute to pathogenicity (e.g. cancer) by favoring cell survival, a function also associated with PI 3-kinase through the serine/threonine kinase Akt [or protein kinase B (PKB) (Kulik et al., 1997;Marte and Downward, 1997;Murga et al., 1998)]. This finely regulated equilibrium in PI 3-kinase activation during epithelial cell differentiation might be assumed by the PTEN-MAGI-1b signalosome (Kotelevets et al., 2005). Indeed, targeting of PTEN to adherens junctions by MAGI-1b and the resulting local downregulation in the PtdIns(3,4,5)P 3 pool play a crucial role in preventing both disruption of junctional complexes and induction of tumor cell invasion.
In conclusion, our results demonstrate that, whereas laminin 5 does not appear to play a major role in enterocyte differentiation, its interaction with integrins controls the biogenesis of adherens junctions. For the first time, we have shown that this process is specifically regulated by PI 3-kinase and involves the activation of Rac1b, a splice variant of the small GTPase Rac1. Thus, PI 3-kinase seems to be a major link in the integrin-cadherin cross-talk in HT-29 cells.

Cell culture
The human colon adenocarcinoma HT-29 cell line (kindly provided by J. Marvaldi, Université Aix-Marseille, Marseille, France) was routinely cultured at 37°C in a 5% CO 2 atmosphere in Dulbecco's Modified Eagle Medium (DMEM) containing 25 mM glucose (Invitrogen) and supplemented with 10% fetal calf serum, penicillin and streptomycin (standard medium). The medium was changed every day to avoid glucose exhaustion, which leads to differentiation. The differentiation of HT-29 cells was initiated by replacing the standard medium for glucose-free DMEM (Invitrogen) supplemented with 10% dialyzed fetal calf serum, 5 mM galactose, 15 mM HEPES, selenous acid (10 -2 g/ml), penicillin and streptomycin. This medium (Gal-medium or differentiating medium) was changed every day. The cells were Journal of Cell Science 119 (1) Fig. 10. Rac1 localization at nascent cell-cell contacts is dependent on PI 3kinase signaling. HT-29 cells were cultured for 24 hours on laminin 5 matrix and then treated with 4 mM EGTA for 30 minutes in the absence or presence of 50 M LY294002. The EGTA-containing medium was then replaced by Ca 2+ -containing medium for 60 minutes. Cells were fixed and stained for Rac1 before confocal microscopy analysis on optical sections. Bars, 25 m. harvested in phosphate-buffered saline (PBS) supplemented with 1 mM EDTA and 0.05% trypsin (w/v).

ECM preparation and coating
Glass coverslips and tissue culture dishes were coated with laminin 5 using either one of the following methods, which generated equivalent results. (1) A431 epidermoid cells were cultured to confluence on various surfaces at 37°C to allow for the deposit of laminin 5, then cells were removed as previously described (Wayner et al., 1993;Weitzman et al., 1993). Briefly, confluent monolayers were sequentially extracted with 1% (v/v) Triton X-100 in PBS, followed by 2 M urea in 1 M NaCl. All extraction buffers contained protease inhibitors (1 mM phenylmethylsulfonyl fluoride and 2 mM N-ethylmaleimide). Plates were washed in PBS, incubated with 1% bovine serum albumin (BSA), and stored at -20°C. (2) Laminin 5 purified from the culture medium of human SCC25 cells was kindly provided by P. Rousselle (Institut de Biologie et Chimie des Proteines, Lyon, France) (Kantengwa et al., 1997;Rousselle and Aumailley, 1994;Rousselle et al., 1991). Except otherwise stated, the first method was used preferentially to prepare large numbers of surfaces. Human collagen type IV from placenta was obtained from Life Technologies. Human plasma fibronectin was purified according to a previously described method (Engvall and Ruoslahti, 1977). Coating of plastic Petri dishes was performed by overnight incubation with ECM proteins (10 g/cm 2 ) at 4°C.

Retrovirus production and infection
The pSG5⌬p85 construct encoding the dominant-negative PI 3-kinase regulatory subunit was kindly provided by B. Vanhaesebroeck (Ludwig Institute for Cancer Research, London, UK). The ⌬p85 cDNA was subcloned into the pBabe retroviral expression vector. Amphotropic retrovirus stocks were obtained by transient transfection [using the transfection reagent Exgen (Euromedex)] of the retroviral vector DNAs into Phoenix packaging cells (gift from C. Bagnis, EFS, Marseille, France). Target HT-29 cells were infected in the presence of 4 g/ml polybrene (Sigma Aldrich) then cultured in medium containing 2 g/ml puromycin (BD bioscience) to select for virus-infected cells. Following selection, cells were pooled, expanded and tested for the expression of ⌬p85 by western blotting, using antibodies to the p85 regulatory subunit of PI 3-kinase.

Cell aggregation assay
Cell-cell adhesion was qualitatively evaluated in an aggregation assay based on Bracke et al. (Bracke et al., 1993). Briefly, HT-29 cells grown on plastic or laminin 5 for 48 hours were detached by incubation in 10 mM Hepes-buffered saline solution (HBSS) containing 0.01% trypsin and 2 mM CaCl 2 for 10 minutes at 37°C [conditions previously described to maintain E-cadherin integrity (Takeichi, 1977)]. After trypsinization, single cell suspensions were made by trituration with a needle. Cell viability assessed by Trypan Blue dye exclusion was greater than 95%. Cells were washed twice in HBSS + 2 mM CaCl 2 and resuspended at 5ϫ10 5 cells per well in 500 l of HBSS + 2 mM CaCl 2 in the presence or absence of 10 g/ml anti-E-cadherin antibody (HECD1) or 4 mM EDTA as indicated. Aggregation assays were performed at 37°C in a gyratory shaker at 75 rpm for 60 minutes in 24-well non-tissue-culture-treated plates that had been blocked with PBS, 2% BSA for 30 minutes at 37°C to prevent attachment of cells to the plastic. Cell aggregation was evaluated by observation of cell clusters under phase contrast microscopy.

In vitro PI-3 kinase assay and lipid analysis
Immunoprecipitation of p85 was carried our with 10 7 cells lyzed for 15 minutes on ice with 1 ml of RIPA buffer (50 mM Hepes, pH 7.4, 10 mM EDTA, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) containing phosphatase and protease inhibitor cocktails. Lysates were clarified by centrifugation for 10 minutes at 8000 g at 4°C. Lysates (500 g protein) were precleared with protein A-Sepharose (Amersham-Pharmacia Biotech) and then incubated overnight with anti-p85 antibodies (0.5 g/ml). After capture by protein A-Sepharose beads and three washes with PBS, half of the immunocomplexes was tested immediately for in vitro PI 3-kinase activity, and the other half was analyzed by SDS-PAGE and immunoblotting with relevant antibodies.
Immunoprecipitates were suspended in 30 l of PI 3-kinase activity buffer (0.5 mM EDTA, 100 mM NaCl and 50 mM Tris-HCl, pH 7.4, plus 50 M ATP and 10 mM MgCl 2 ) and incubated with a mixture of phosphatidylinositol/ phosphatidylserine vesicles (30 g/60 g) and 15 Ci of [␥ 32 P]ATP (3000Ci/mmole from Amersham Pharmacia Biotech) for 30 minutes at 37°C with shaking. The reaction was stopped by the addition of 80 l of 1.2 N HCl and 400 l of a 1:1 mixture of chloroform and methanol. Lipids extracts were separated by thin layer chromatography on Silica Gel G plates (Merck) previously coated with 1% (w/v) potassium oxalate, 2 mM EDTA in a mixture of water and methanol (90:60 v/v). Chromatograms were developed for 2 hours in chloroform, methanol (90:70 v/v) and 3 M ammoniac. The radioactive spots were visualized by a PhosphorImager 445 SI (Molecular Dynamics) after 1 hour exposure and quantified by ImageQuant software.

Western blot analysis
Cells were grown to confluence and lyzed with a buffer made of 1% Triton X-100, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and supplemented with 1 mM PMSF, 2 g/ml aprotinin, 10 g/ml leupeptin, 10 M pepstatin, 2 mM CaCl 2 and 2 mM MgCl 2 for 15 minutes on ice. Protein concentrations in lysates were determined using the copper reduction/bicinchonic acid (BCA) assay (Pierce Chemical Co) according to the manufacturer's instructions. Proteins (50 g in SDS-␤-mercaptoethanol sample buffer) were resolved on 10% polyacrylamide gels, transferred onto PVDF membranes (Hybond-C super; Amersham), and blocked in 5% fat-free dry milk in 0.1% Tween 20 in PBS for 1 hour at room temperature. After overnight incubation at 4°C with primary antibodies diluted in the blocking solution, blots were washed in PBS, 0.1% Tween 20 and then incubated with appropriate horseradish peroxidase-conjugated secondary antibodies (dilution of 1:3000) for 1 hour at room temperature before extensive washes. The blots were visualized by chemiluminescence (Amersham ECL reagents) using the Image Master VDS-CL device (Amersham-Pharmacia Biotech) and quantified with Image J software from NIH. Primary antibodies were used at the following dilutions: anti-p85 (1:2000), anti-Rac1 (1:1000), anti-Rac1b (1:1000), anti-human E-cadherin (1:1000) and anti-actin (1:100).

Subcellular fractionations
10 7 cells were lyzed in 1 ml of hypotonic buffer (0.34 M saccharose, 1 mM EDTA pH 7.4) by 40 passages through a dounce homogenizer and subsequent centrifugation for 15 minutes at 500 g at 4°C. The pellet corresponded to the nuclear fraction. The supernatant was withdrawn and further centrifugated for 40 minutes at 60,000 g at 4°C. The supernatant was assimilated to the cytosol and the pellet containing membranes was resuspended in 50 mM Tris-HCl pH 7.2, 500 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 10 mM MgCl 2 , and containing protease inhibitor cocktail.
Soluble and cytoskeletal fractions were prepared essentially as described by Nelson and coworkers (Hinck et al., 1994). The cells were rinsed in PBS supplemented with 1 mM CaCl 2 and homogenized in CSK buffer (50 mM NaCl, 10 mM PIPES, pH 6.8, 3 mM MgCl 2 , 0.5% Triton X-100, 300 mM sucrose) supplemented with 1 mM PMSF, 10 g/ml leupeptin, 0.5 mM sodium vanadate and 20 M phenylarsine oxide for 10 minutes at 4°C with gentle rocking. After centrifugation for 10 minutes at 15,000 g at 4°C, the supernatant constituted the Triton-soluble (TS) fraction. The pellet was triturated in the same volume of SDS buffer (20 mM Tris, pH 7.5, 5 mM EDTA, 2.5 mM EGTA, 1% SDS) and boiled at 100°C for 10 minutes. After centrifugation for 10 minutes at 15,000 g at 4°C, the cleared supernatant constituted the Triton-insoluble (TI) fraction. This fraction usually contained 5-8-fold less protein than the TS fraction, as determined by BCA assay. Protein amounts corresponding to the same number of cells were routinely analyzed.

Immunofluorescence microscopy
Cells grown on glass coverslips were fixed with 3% paraformaldehyde, 2% sucrose for 10 minutes at 37°C and further permeabilized with 0.2% Triton X-100 in PBS for 15 minutes at room temperature. Cells were washed twice with PBS containing 3% fat-free dried milk and 0.1% Tween 20, and blocked in 10% goat serum in PBS for 1 hour at room temperature to reduce background before staining. Cells were then stained for 1 hour at 37°C with the primary antibody diluted in the blocking solution, washed three times with PBS containing 0.1% Tween 20 and incubated with a secondary antibody coupled to AlexaFluor488 or AlexaFluor546 (used at the dilution of 1:500) for 45 minutes at 37°C. Coverslips were permanently mounted with Mowiol (Calbiochem). Fluorescence photomicrographs were taken using a confocal laser-scanning microscope (Zeiss LSM 510). Primary antibodies were used at the following dilutions: anti-human E-cadherin (1:500), anti-p85 (1:100), anti-Rac1 (1:200) and anti-Rac1b (1:100).

Co-immunoprecipitation experiments
The cells were lyzed in RIPA buffer (50 mM Hepes, pH 7.4, 10 mM EDTA, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) containing protease inhibitor cocktails for 15 minutes on ice. The cell lysates (500 g of protein) were incubated 2 hours at 4°C with 2 g of IgG2a isotype coupled to protein A-Sepharose (Amersham-Pharmacia Biotech). After a 10 minute centrifugation at 15,000 g, the supernatants were incubated overnight at 4°C with 2 g of antibodies raised against p85 and immobilized on protein A-Sepharose. Beads were washed four times with the lysis buffer, and bound proteins were eluted from the beads by boiling in 20 l of SDS-PAGE Laemmli's sample buffer for 5 minutes. The samples were analyzed by western blot.
Inhibition of E-cadherin-mediated cell-cell contacts Cells (80% confluence) were cultured for 24 hours on plastic or laminin 5. Adherens junctions were then disrupted by treatment with 4 mM EGTA for 30 minutes at 37°C. Intercellular contacts were subsequently allowed to re-establish by restoration of the extracellular Ca 2+ concentration by replacing the EGTA-containing medium with fresh medium [1.8 mM CaCl 2 (Pece et al., 1999;Volberg et al., 1986)]. In some experiments, LY294002 (50 M) was added in the EGTA-containing medium. After the selected time of Ca 2+ restoration, the cells were fixed for subsequent immunofluorescence studies.

Rac1-GTP pull-down assays
Pull-down assays, using the glutathione S-transferase (GST)-PAK-Rac-binding domain (CRIB) fusion protein (GST-PAK-CRIB) (kindly provided by C. Gauthier-Rouviere, CRBM, CNRS/INSERM, Montpellier, France) were performed essentially as described (Sander et al., 1998). One 10 cm dish of confluent HT-29 cells was used per data point. Cells were rapidly washed in ice-cold PBS and lyzed in a buffer containing 1% Triton, 50 mM Tris, pH 7.2, 5 mM EGTA, 5 mM EDTA and a protein inhibitor mixture. Lysates were centrifuged for 10 minutes at 17,000 g at 4°C, and samples were taken from the supernatant to estimate the total protein concentration. GST-PAK-CRIB fusion bound to Sepharose beads (50 l) was added to 2 mg of cell lysate proteins and incubated for 1 hour at 4°C. Beads were washed four times in lysis buffer, and bound proteins were eluted in Laemmli's sample buffer at 95°C for 5 minutes. GTP-bound Rac1 was analyzed by SDS-PAGE using a 12% polyacrylamide gel, transferred onto a PVDF membrane, and probed with anti-Rac1 mAb and anti-Rac1b polyclonal antibodies. Whole-cell lysates were run in parallel.

Statistical analysis
All the experiments were performed at least three times. For statistical analysis of data, Student's t-test was used. Values are expressed as mean ± s.e.m. Data were considered statistically significant at a P value of <0.01(**) or <0.05 (*).