Enterocyte differentiation is a dynamic process during which reinforcement of cell-cell adhesion favours migration along the crypt-to-villus axis. Functional polarization of Caco-2 cells, the most commonly used model to study intestinal differentiation, is assessed by dome formation and tightness of the monolayer and is under the control of the extracellular matrix (ECM). Furthermore, our biochemical and confocal microscopy data demonstrate that the ECM dramatically reinforces E-cadherin targeting to the upper lateral membrane, formation of the apical actin cytoskeleton and its colocalization with E-cadherin in functional complexes. In our model, these effects were produced by native laminin-5-enriched ECM as well as by type IV collagen or laminin 2, which suggests a common pathway of induction through integrin receptors. Indeed, these effects were antagonized by blocking anti-β1-and anti-α6-integrin antibodies and directly induced by a stimulating anti-β1-integrin antibody. These results demonstrate that integrin-dependent cell to ECM adhesion reinforces E-cadherin-dependent cell-cell adhesion in Caco-2 cells and further support the notion that enterocyte differentiation is supported by a molecular crosstalk between the two adhesion systems of the cell.
The epithelium forms a barrier made of polarized cells joined by a complex set of cell-cell junctions. The assembly of adherens junctions through the interaction of E-cadherin of adjacent cells initiates this process(Gumbiner, 1996;Kemler, 1992). The importance of cell-cell adhesion in differentiation and in the maintenance of the differentiated phenotype is well established in epithelial cells(Braga et al., 1999). In addition, epithelial cells are separated from the underlying connective tissue by a basement membrane that is composed of a variety of extracellular matrix(ECM) molecules that control cell differentiation in many tissues through interactions with their cellular receptors, for example, with integrins(Boudreau and Bissell,1998).
The basement membrane is mostly composed of type IV collagen, different types of laminins, entactin and heparan sulfate proteoglycan(Beaulieu, 1997). ECM molecules, originating from both epithelial and underlying mesenchymal cells,create a framework that is essential for maintaining tissue integrity(Simon-Assmann and Kedinger,1993). Besides this structural role, ECM proteins are involved in the control of adhesion, migration, proliferation, differentiation and gene expression of adjacent cells, which emphasizes the dynamic reciprocity between epithelial and mesenchymal cells (Bissell et al., 1982). Additionally, ECM is able to control the effects of trophic factors by sequestration outside of the cell(Simon-Assmann et al., 1998)and by crosstalk between their signaling pathways(Yamada and Geiger, 1997). It is admitted that cell adhesion to the ECM contributes to the apical-to-basal axis of polarity, in vivo as well as in vitro. Appearance of polarized cells coincides with the expression of laminin 1 (LN1) in the developing kidney(Klein et al., 1990). Similarly, the addition of laminin boosts the formation of polarized alveoles in various types of epithelial cells, including mouse mammary(Li et al., 1987), human salivary (Hoffman et al.,1996) and rat lung (Matter and Laurie, 1994) cells in culture. ECM-integrin interactions have either been demonstrated to be directly involved in ECM control of cell functions or found to be aberrant in embryos or animals carrying mutations in integrin genes (Wang et al.,1999).
Both cell-ECM and cell-cell adhesion systems are connected to the cytoskeleton, which controls cell polarization. Numerous studies have established that the interaction between ECM and integrin results in cytoskeletal rearrangements (Larjava et al., 1990; Wang et al.,1999). Integrins are heterodimeric transmembrane receptors composed of α and β subunits associated in a noncovalent manner(Hynes, 1987;Yamada and Miyamoto, 1995). Integrin initiates, through its β1 cytoplasmic domain, the assembly of specialized cytoskeletal and signaling protein complexes at the contacting membrane (Gimond et al.,1999). In the same way, epithelial cells forming strong cell-cell junctions assemble a subcortical actin skeleton instead of focal adhesion and actin stress fibers (Larjava et al.,1990). Cadherins are also dependent on cytoskeletal organization(Tsukita et al., 1992);correct function of the E-cadherin—catenin complex requires association with the cytoskeleton (Skoudy et al.,1996). In epithelial cells, about one half of plasma membrane E-cadherin is connected to the actin cytokeleton: the rest is free within the membrane (Sako et al., 1998). The linkage between E-cadherin and the F-actin cytoskeleton is mediated through direct binding of the cytoplasmic domain of E-cadherin toβ-catenin, which binds to α-catenin(Aberle et al., 1994;Jou et al., 1995) in a 1:1:1 stochiometry. Crosstalk between the two adhesion systems has also been demonstrated in mammary epithelial cells through the integrin signaling pathway. In these cells, integrins promote the formation of morphologically differentiated acini-like structures, which involves the assembly of adherens junctions through the relocalization of E-cadherin at the lateral side of the cells (Weaver et al.,1997).
The mammalian intestinal epithelium is peculiar in that it is a constantly renewing monocellular epithelium, which migrates `en cohorte' along the basement membrane from the proliferative undifferentiated compartment in the crypts to the tips of the villi. Enterocytes can probably glide over the basement membrane through loose adhesion, through them being tied to each other by strong cell-cell junctions. Whereas type IV collagen is constantly present in the basement membrane, LN2 is preferentially found in the proliferative compartment, LN5 in the villus and LN1 at the junction of the two compartments (Vachon et al.,1993; Lorentz et al.,1997). Similarly, villus and crypt epithelial cells display a different pattern of integrins, β1-containing integrins being more abundant in the villi than in the crypts. Furthermore, β1 is mainly associated with α2 in the crypt and with α3 integrins in the villus (Beaulieu, 1992). Whereas α2β1 integrin preferentially binds to collagen IV but also to LN1 and LN2, α3β1 integrin binds to both collagen IV and LN5(Beaulieu, 1999;Rousselle and Garrone, 1998). Integrin α6β4 binds to both LN1 and LN5(Fleischmajer et al., 1998). This differential pattern of expression of ECM proteins and their receptors along the crypt-to-villus axis parallels the differentiation process of epithelial cells. One can wonder whether changes in ECM-integrin interactions at the crypt to villus junction are accompanied by changes in cell-cell adhesion, which allow cell migration to the tip of the villus.
The colon cancer Caco-2 cell line in culture mimics enterocyte differentiation. We previously showed that ECM was required for the expression of the apoA-IV gene, an intestinal differentiation marker(Le Beyec et al., 1997). Here,we observed that the functional polarization of Caco-2 cells, assessed by dome formation and permeability of the monolayer, is under the control of integrin-mediated adhesion to ECM. Furthermore, we demonstrate that integrin activation by ECM reinforces cell-cell adhesion by targeting E-cadherin at the lateral membrane in functional complexes with actin cytoskeleton.
Materials and Methods
Antibodies and products
Antibodies used included monoclonal anti-β1 (6S6) and anti-α6(NKI-GoH3) human integrins with a blocking activity and anti-β1-integrin(B3B11) with a stimulating activity and purified control mouse anti-IgG(Chemicon); polyclonal anti-human E-cadherin (HECD-1) (Zymed); monoclonal anti-β-catenin (clone 14) (Transduction Laboratories); FITC-labeled antibodies (Sigma) and RITC-labeled antibodies (Boehringer). TRITC- or FITC-labeled phalloidin (Sigma) was used to visualize the actin cytoskeleton. We also used human merosin LN2 (Gibco) and mouse collagen type IV and synthetic poly-D-lysine (Becton-Dickinson).
Caco-2 cells (43rd to 50th passage) and HT29 cells adapted to 10-5 M of methotrexate(Lesuffleur et al., 1990) and cultured without the drug and named HT29-MTX (9th passage) were cultured at 37°C with 10% CO2 in Dulbecco's minimal essential medium (DMEM), 25 mM glucose (Gibco), pen/strept (50 μg/ml) and non-essential amino acid (1%) (Gibco) supplemented with 5% foetal calf serum(Boehringer). Mesenchymal intestinal cells C9, C11, C20 obtained from M. Kedinger (Fritsch et al.,1999) (28th, 29th and 14thpassages, respectively) were cultured at 37°C with 7.5% CO2 in RPMI 1640 medium, pen/strept (50 μg/ml) (Gibco), supplemented with 10%foetal calf serum (Boehringer). Muscle 129CB3 cells were cultured as described(Pinçon-Raymond et al., 1991) to form contracting myotubes and secrete a large amount of ECM.
Extracellular matrix preparation and coating
Native ECM was prepared from 129CB3 myotubes, mesenchymal C9, C11, C20 cells (at confluence), HT29-MTX cells (3 days postconfluence) or Caco-2 cells(12d post-confluence) as described previously(Le Beyec et al., 1997). Coating of plastic petri dishes was performed by overnight incubation with poly-D-lysine, 5 μg/cm2, collagen type IV, 10μg/cm2 and merosin LN2, 8.4 μg/cm2 at 4°C.
Caco-2 cells were seeded at 125,000 cells/cm2 (pre-confluence)in 24-well plates coated or not with native ECM or ECM components. At the time of plating, cells were mixed with control mouse IgG or anti-β1-integrin monoclonal blocking antibody (6S6) or anti-α6-integrin used to blockβ4 integrin (CD49F) or anti-E-cadherin monoclonal blocking antibody(HECD-1) at the indicated dilutions. Under these conditions, control cells were confluent within 24 hours. For each kinetics experiment, triplicate wells were observed using a phase contrast microscope. Confluence was evaluated, and counting triplicate wells on a phase contrast microscope numerated the domes.
Ribonuclease protection assay
A specific 400 bp cDNA encoding the human apoA-IV gene was obtained by RT-PCR using the coding oligonucleotide HindIII-AIV(5′-CTGGAGAAGCTT+149ACACTTACGCAGGTGACCTG-CAG+171-3′)and the noncoding oligonucleotide Xba-AIV(5′-CT-GCAGTCTAGA+550AGGGCGTAAGGCGTCCCTTGA+530-3′). The PCR product was digested using XbaI and HindIII, and ligated into the XbaI/HindIII-digested PSK vector to obtain the pAIV-RPA plasmid. For E-cadherin mRNA analysis, a specific 407 bp cDNA encoding the human E-cadherin gene was obtained using the coding oligonucleotide(5′-+2660GACCAGGACTATGACTACTTG-AACG+2684-3′)and the noncoding oligonucleotide(5′-+3067ATC-TGCAAGGTGCTGGGTGAACCTT+3043-3′)inserted into PCR 2.1 vector. An antisense AIV RNA probe (445 bp) was generated by in vitro transcription of the HindIII-digested pAIV-RPA plasmid using [α-32P]UTP and T3 RNA polymerase (Promega). An antisense E-cadherin RNA probe (523 bp) was generated by in vitro transcription of the kpn1-digested E-cadherin-PCR2.1 plasmid using[α-32P]UTP and T7 RNA polymerase (Promega). An antisenseβ-actin RNA probe (Human Internal Standards kit, Ambion Inc.) was synthesized with T3 as an internal control. Total RNA was extracted from cells using an RNAzol kit (Bioprobe Systems). Equal amounts (6 μg) of total RNA samples were subjected to the RNase protection assay using the RPAII kit(Ambion Inc) following the manufacturer's recommendations. The protected A-IV RNA (400 bp), E-cadherin RNA (407 bp) and β-actin RNA (245 bp) probes were separated on a 5% denaturing polyacrylamide-urea gel in Tris borate-EDTA buffer. The gel was dried and exposed to X-ray film at -80°C.
Cell surface biotinylation
Caco-2 cells were seeded on plastic coated or uncoated dishes with native ECM or ECM components and grown for 6 days. All manipulations were performed at 4°C according to Sander et al.(Sander et al., 1998). Briefly, cells were incubated for 15 minutes in phosphate-buffered saline(supplemented with 1 mM MgCl2 and 0.5 mM CaCl2)containing 500 μg/ml sulfo-NHS-biotin (Pierce Chemical Co.), washed three times in phosphate-buffered saline containing 50 mM glycine, pH 7.4, lysed in RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, 0.1% SDS, protease inhibitors, 10% glycerol, 1 mM EDTA, 3 mM MgCl2, 1 mM dithiothreitol) and centrifuged for 15 minutes at 13,000 g. The supernatant was incubated with avidin-coated agarose beads (Sigma Chemical Co.) for 1 hour. Immunoprecipitates of biotinylated surface proteins bound to avidin-agarose were washed five times in RIPA buffer and analysed for E-cadherin (HECD-1) by western blotting.
The protein concentration of Caco-2 lysates, biotinylated or not, was assessed by the Biorad `Dc' protein assay. A 20 μg aliquot of each sample mixed with Laëmmli buffer was boiled and submitted to 7% SDS polyacrylamide gel electrophoresis. Samples were then transferred onto nitrocellulose and blocked in 1% non-fat milk overnight at 4°C. After a 2 hour incubation with the primary antibody in the blocking solution at room temperature, blots were washed in PBS 1× pH 7.4,incubated with appropriate HRP-conjugated secondary antibody and washed again. The blots were visualized by chemiluminescence (Amersham ECL system). Signals were scanned (Umax vistaScan S6E) from chemiluminescence into Adobe Photoshop.
Caco-2 cells were grown on Lab-Tek chambered borosilicate coverglasses(Nunc), coated or not with native ECM or ECM components. At the indicated time, cells were fixed in 4% paraformaldehyde in phosphate-buffer saline, then permeabilised in 0.1% Triton X-100 during all incubations. Non-specific antigens were blocked for 30 minutes in 3% bovine serum albumin. Double labeling was performed sequentially to avoid crossreactions. Anti-β1-integrin (6S6) primary antibodies diluted in the blocking solution were incubated for 1 hour 30 minutes, followed by a 1 hour 30 minute incubation with RITC-labeled secondary antibodies, followed by overnight incubation at 4°C with an anti-E-cadherin antibody (HECD-1). These antibodies were visualized with FITC-labeled secondary antibodies after a 1 hour 30 minute incubation. Images were acquired with a Zeiss LSM-510 laser-scanning confocal microscope (Carl Zeiss, Oberkochen, Germany) equiped with Zeiss Axiovert 100M (plan Apochromat 63×1.40 NA oil immersion objective). The contrast and brightness settings were constant during the course of image acquisition. The E-cadherin/actin colocalization visualized by confocal analysis was quantified using a program from Zeiss LSM 510 confocal. The data were recorded from cells in the upper half of the cell in six random fields from three independent experiments.
Native ECM induces the expression of a differentiation marker gene and cell polarity
Of the criteria for epithelial cell differentiation, modifications in cell shape (polarization) and gene expression are those most often reported. We have previously reported that the expression of a differentiation marker gene of enterocytes, apolipoprotein A-IV (apoA-IV), was induced in human colon carcinoma Caco-2 cells when they were grown on filters coated with a native extracellular matrix (ECM)(Le Beyec et al., 1997). To determine whether this effect was due either to ECM-cell adhesion or to cell polarization, which would be induced independently by culturing cells on a filter, we grew Caco-2 cells on native ECM deposited on a plastic support.Fig. 1 clearly shows that ECM is able to induce apoA-IV expression independently of the polarization effect of the filter. Furthermore, the level of induction varies according to the origin of the different ECM tested.Table 1 summarizes available data on the production of ECM components by the various mesenchymal and epithelial cell types used to deposit native ECM on the plastic support. Most of these data were obtained by measuring mRNA levels by RT-PCR, which is insufficient to predict the amount of laminin synthesized and deposed by the cells. Indeed discrepancies between mRNA and protein measurements were reported for α1 laminin in HT29-MTX and Caco-2 cells. Furthermore, C9,C11 and C20 intestinal mesenchymal cells have been reported to express laminin chain mRNA in the same range. Nevertheless, the native ECM deposited by C20 cells was much more efficient in inducing apoA-IV gene expression than that from the other clones. However,Fig. 1 indicates that the most effective ECM to induce apoA-IV expression is the native LN5-rich ECM from HT29 cells adapted to 10-5 M methotrexate.
Observation of Caco-2 cells during these experiments revealed that cells grown on native LN5-rich ECM formed domes 2 days earlier than cells grown on the plastic support, and the ECM-grown domes were larger(Fig. 2A). It is known that, at confluence, epithelial cells grown on a non-porous support such as plastic are elevated by the fluid accumulated under the monolayer and form domes(Pinto et al., 1983). Comparison, every 2 days for 14 days, of Caco-2 cells grown on native ECM or on plastic shows that this dramatic increase in domes formed by Caco-2 cells on native LN5-rich ECM (Fig. 2C) does not rely on the confluence rate of the cells(Fig. 2B), which is the same under both conditions. The permeability of the monolayer was further assessed by the use of FITC-biotin, an outside marker to which cells are impermeable.Fig. 2D confirms an overall inductive effect of ECM on the tightness of cell-cell junctions and functionality of tight junctions by displaying the ability of FITC-labelled biotin to penetrate between adjacent cells within the monolayer. Clearly, this molecule remained apical on the monolayer grown on ECM substrate(Fig. 2Db) whereas it penetrated much deeper between cells grown on plastic without ECM (a) or on polylysine (not shown), an artificial substrate which does not binds to integrins (Machesky and Hall,1997). In contrast to the purpose of the experiment, which was to differentiate between the effects of ECM and filter-induced cell polarization,it suggests that functional polarization of Caco-2 cells, as assessed by dome formation and permeability of the monolayer, is under the control of ECM.
Native ECM triggers E-cadherin accumulation at the lateral membrane and colocalization with actin cytoskeleton
The aggregation of E-cadherin molecules at the adherens junctions is the primary event, which organizes the formation of the other cell-cell junctions,that is gap, desmosome, and tight junctions, which ensure the formation of an impermeable polarized epithelium(Cereijido et al., 2000;Fujimoto et al., 1997;Jongen et al., 1991;Lampe et al., 1998). We therefore studied the expression of E-cadherin in our system. We saw that native LN5-rich ECM induced a threefold increase in apoA-IV gene expression. At the same time, the total amount of E-cadherin protein(Fig. 3B) and mRNA(Fig. 3A) remained in the same range, as did that of β-catenin protein, a partner of E-cadherin required for an efficient exit from endoplasmic reticulum in MDCK cells(Chen et al., 1999). The amount of E-cadherin associated with the membrane was obtained after surface biotinylation in the presence of 0.5 mM Ca2+, a concentration resulting in a slight loosening of tight junctions but still too high for inducing the disruption of adherens junctions, which occurs under 0.1 mM Ca2+ (Cereijido et al. 2001;Braga et al., 1997)(Fig. 3C). Similar to the observation by Sander et al. in MDCK cells expressing Tiam1/Rac(Sander et al. 1998),Figure 3C shows that the association of E-cadherin with the membrane was increased fourfold in cells grown on ECM compared with those grown on plastic without ECM, although ECM did not influence the total amount of E-cadherin—β-catenin.
The targeting of E-cadherin to the membrane induced by ECM was further characterized by confocal analysis. The signal detected by indirect immunofluorescence was stronger and cell-cell junctions were better delineated in cells grown on ECM (Fig. 4Ab) compared with cells grown on an inert support(Fig. 4Aa). In addition,confocal 3D analysis shows that E-cadherin was clearly visible at the base of cells, which form a flat monolayer on an inert support(Fig. 4Ac), whereas the signal almost disappeared from the base of cells forming domes on ECM and concentrated in focal spots in the upper third of the lateral membrane(Fig. 4Ad).Fig. 4B (c,d,e) shows that purified ECM components such as type IV collagen and laminin 2 were as efficient as native ECM in inducing E-cadherin targeting to the lateral membrane of cells that do not form domes. Since E-cadherin localized to adherens junctions is intimately associated with actin cytoskeleton in polarized epithelial cells, we also investigated by confocal analysis the actin cytoskeleton and its association with E-cadherin. In addition, culturing cells on ECM components reinforced the formation of the apical actin cytoskeleton and its colocalization with E-cadherin at the upper part of the lateral membrane (Fig. 4Bc', d',e'), as revealed by the merge yellow signal, as compared to an inert support (Fig. 4Ba'). Similar observations were made using a stimulating anti-β1-integrin antibody in Caco-2 cells grown on an inert support, resulting in the formation of a cortical network of actin at the apical side of the cell(Fig. 4C). Altogether, these results favour a role of native ECM or of its components in the accumulation of E-cadherin at the lateral membrane in functional complexes anchored to the apical actin cytoskeleton.
E-cadherin targeting to the lateral membrane involves β1 integrin
ECM components interact at the cell surface with their receptor integrins,which are mainly α3β1 and α6β4 for LN5, in differentiated intestinal epithelial cells. In order to see whether ECM induced modification in integrin distribution in Caco-2 cells, we performed confocal analysis after double labeling against β1 or β4 integrin and E-cadherin. As expected, we observed that β1 integrin colocalized with E-cadherin at the lateral membrane of cells forming domes, mostly when cells were grown on ECM, a condition in which domes are much more numerous than in cells grown on an inert support (data not shown). We also verified that β4 integrin was only found at the basal membrane of cells forming domes on ECM but not on an inert support (data not shown).
To further establish the role of ECM on Caco-2 cell polarization and E-cadherin targeting to the membrane, we performed perturbation experiments using functional blocking antibodies against β1 integrin, the βchain of the major integrin receptor for laminins and type IV collagen. Indeed, dome formation in cells grown on ECM was drastically impaired by the anti-β1-integrin blocking antibody, in a dose dependent manner, and it was reduced to the range observed in cells grown on plastic support(Fig. 5A). Similarly, upon treatment with antibodies, cells grown on ECM reached confluence 1 day later than those not treated, at a time similar to that observed with cells grown on plastic. The effect of anti-β1 antibody on dome formation was observed when cells were at confluence whereas non-specific mouse IgG displayed no effect (Fig. 5B).
Under these conditions, we investigated the effects of anti-β1 or anti-α6 blocking antibodies on E-cadherin accumulation at the lateral membrane and anchoring to the cortical actin cytoskeleton. Colocalization of E-cadherin and actin at the apical-lateral side of cells was estimated by computer analysis of confocal stack series. Pixel count and pixel intensity measurements gave the same results. Fig. 6A shows that Col IV and LN2 increased the amount of colocalization of E-cadherin and actin signals up to 30% as compared to cells grown on plastic. Thus, either ECM component could be used in the experiments. Confocal 3D reconstruction of cells grown on Col IV reveals double, cortical and basal rows of actin cytoskeleton with an important level of E-cadherin and actin colocalization (Fig. 6Ba). The addition of anti-β1-integrin blocking antibody resulted in dramatic disorganization of the cortical row of the actin cytoskeleton and, in parallel, a reduction in E-cadherin and actin colocalization (Fig. 6Bb). The blockade of either of the β1 integrins in cells grown on LN2(Fig. 6C) or α6β4 integrin in cells grown on native ECM (Fig. 6D) resulted in a significant reduction in the colocalization of E-cadherin and actin at the apical lateral side of Caco-2 cells. Altogether,these results demonstrate that ECM, by interacting with its receptor integrins, influences the association of E-cadherin and actin at the level of adherens junction as functional complexes responsible for Caco-2 cell polarization.
Enterocyte differentiation is a dynamic process that takes place within a polarized epithelium migrating `en cohorte' from the proliferative compartment located in the crypt. Proliferative cells arise by successive asymmetric divisions from stem cells, which themselves are part of the polarized intestinal epithelium, the integrity of which is essential for its barrier function. The localization of stem cells at a fixed position within the crypt,the gradual loss of stem cell properties in the upwardly migrating cells and changes in cell adhesion to the ECM during cell migration toward the villus indicate that the cell environment may control differentiation through different attachment properties (Booth and Potten, 2000). The composition of ECM varies, as does the integrin repertoire expressed by enterocytes(Potten et al., 1997). At the same time, E-cadherin is mostly localized at the apical junctional complexes in the villus, contrary to the crypt(Hermiston et al., 1996). Altogether, these changes might allow enterocytes to glide over the basement membrane through a looser type of adhesion and might reinforce cell-cell junctions to perform the driving force.
In the present paper, we show in vitro that cell-ECM adhesion improves cell-cell adhesion through the reinforcement of E-cadherin—actin complexes at the level of adherens junctions in Caco-2 cells. This effect is specific for cell-ECM adhesion as it is antagonized by function-blocking anti-integrin antibodies.
In vitro, epithelial cells form polarized monolayers at confluence, even though full differentiation is not reached. Studying the influence of native ECM on the expression of the apoA-IV gene, an enterocytic marker, in Caco-2 cells we observed that the ECM boosted the formation of domes by the monolayer. Formation of domes by confluent epithelial cells cultured on a non-porous support signals the formation of an impermeable monolayer, which rises owing to the fluid accumulated underneath. This requires the setting of intercellular tight junctions, the activation of pumps for electrolytes and water and a decrease in adherence to the substrate. The formation of domes,while occurring spontaneously on plastic support(Pinto et al., 1983), has been shown to be enhanced by differentiation inducers such as dimethyl sulfoxide(DMSO) or 8-Br-cAMPC in LA7 epithelial cells(Zucchi et al., 1998). Here,time of dome formation, their number and size and monolayer tightness specifically depend on cell-ECM interactions, as FITC-labelled biotin penetrated much deeper between Caco-2 cells grown on plastic or polylysine as compared to cells grown on native ECM. It should be emphasized that the time for the delayed formation of domes by Caco-2 cells grown on plastic was compatible with the time necessary for the deposition of ECM material produced by Caco-2 cells themselves (Vachon and Beaulieu, 1995).
Assembly of tight junctions, as well as gap and desmosomal junctions,depends on E-cadherin recruitment at adherens junctions(Cereijido et al., 2000;Jongen et al., 1991;Matsuzaki et al., 1990;Mege et al., 1988;Fujimoto et al., 1997;Green et al., 1987;Gumbiner et al., 1988;van Hengel et al., 1997). Therefore, we investigated E-cadherin status in our cells. Native ECM did not affect the total amount of E-cadherin and of β-catenin protein, as shown by biochemical analysis and confocal microscopy, but we demonstrated that ECM dramatically increases E-cadherin localization to the plasma membrane. Confocal microscopy revealed that, in cells grown on native LN5-rich ECM, the E-cadherin signal focused at the cell-cell junction domain in the upper third of the lateral membrane, where adherens junctions are known to be localized. Furthermore, the observation that ECM induced an increase in E-cadherin—actin colocalization suggests a reinforcement of E-cadherin anchoring to the actin cytoskeleton and a better organization of the subcortical network of actin by ECM. It is well established that tethering of E-cadherin to the actin cytoskeleton underlies strong cell-cell adhesion and is loosened in weak adhesion (Adams and Nelson, 1998; Kaibuchi et al.,1999b). The lateral membrane targeting of E-cadherin is produced not only by a native LN5-rich ECM but also by Col IV alone, which is a common component of all native ECMs (Rousselle and Garrone, 1998).
In our model, the coordinated reorganization of cell-cell adhesion and the F-actin cytoskeleton was produced by native laminin-5-enriched ECM as well as by type IV collagen or laminin 2, suggesting a common pathway of induction. We therefore questioned the role of ECM receptors expressed in Caco-2 cells (i.e.α3β1 and α6β4 integrins). Blocking experiments with anti-β1-integrin or anti-α6-integrin antibodies in Caco-2 cells grown on ECM substrates resulted in a phenotype similar to that obtained on an inert support: a random distribution of E-cadherin along the basolateral membrane, a looser organisation of the F-actin network and a reduction in the merge signal from E-cadherin and F-actin cytoskeleton. These results demonstrate that recruitment of integrin receptors by their external ligands results in the reinforcement of E-cadherin—actin functional complexes. But, upon ligand binding, integrin linkage to the F-actin cytoskeleton is known to be reinforced (Calderwood et al.,2000). This apparent contradiction might be explained by the existence of distinct pools of F-actin forming functional complexes with E-cadherin and integrin. Alternatively, translocation of regulatory proteins from E-cadherin to integrin complexes has been proposed to mediate crosstalk between N-cadherin and β1 integrin in neural retina explants(Arregui et al., 2000). Both hypotheses are challenged by the colocalization of E-cadherin and β1 orβ4 integrin that we observed by confocal microscopy in the lateral membrane of Caco-2 cells grown on ECM, whereas no colocalization was found on an inert support. Such a colocalization of β1 integrin and E-cadherin has already been reported at cell-cell junctions in keratinocytes(Braga et al., 1997), although keratinocytes form a different system in which integrin loses contact with ECM while migrating towards the superficial layers of this stratified epithelium,where E-cadherin finally downregulates integrin expression(Hodivala and Watt, 1994).
Our results favour cooperation between ligand-bound integrin and E-cadherin in the organization of the subcortical F-actin cytoskeleton. In accordance with our results, it has been shown in kidney epithelial cells that E-cadherin—catenin complexes at cell-cell junctions were not sufficient to maintain the subcortical F-actin cytoskeleton in the absence ofα3β1 integrin (Wang et al.,1999). Similarly, laminin-5-activated α3β1 integrin has been demonstrated to promote gap junctional communication in keratinocytes(Lampe et al., 1998). In contrast, expression of β1 integrin splice variants in β1-deficient epithelium-like cells resulted in downregulation of cadherin function,disruption of cell-cell adhesion and induction of cell scattering(Gimond et al., 1999), all of which underlie the cell-type specificity of cadherin localization(Braga et al., 1999).
The extrinsic spatial cues mediated by cell-cell and cell-substratum adhesions and trophic factor signaling need to be coordinated to ensure a differentiated phenotype. The Rho family GTPases (Rho, Rac and Cdc42) are good candidates for a central role in coordinating adhesion systems. Rho GTPases have been demonstrated to intervene in the inside-outside control of cell-substrate adhesion (Calderwood et al., 2000). Reciprocally, Rho, Rac1 and Cdc42 play roles in parallel and convergent signaling pathways triggered by cell adhesion to an ECM substrate (Clark et al.,1998). The control of E-cadherin-mediated cell-cell adhesion by the Rho family GTPases and their modulators has been recently characterized in the context of epithelial-mesenchymal transition, where the loss of cell-cell junctions promotes cell migration (Braga et al., 1997; Hordijk et al.,1997; Takaishi et al.,1997; Kuroda et al.,1998; Braga et al.,1999; Fukata et al.,1999). At the same time, it was established that Rho GTPases play a key role in the control of actin polymerization, cell shape and motility(Kaibuchi et al., 1999a).
Our findings lend support to the notion that enterocyte differentiation is an active process supported by molecular crosstalk involving cell-ECM and cell-cell adhesions (Hermiston and Gordon,1995). We report for the first time that integrin-dependent cell-ECM adhesion reinforces E-cadherin-dependent cell-cell adhesion in epithelial cells (see Note in Proof). This reinforcement most probably allows cell migration along the crypt to the villus of the intestinal epithelium(Hermiston et al., 1996). By contrast, it is well documented that cell migration is promoted by the loss of cell-cell junctions during the epithelial-mesenchymal transition of epithelial cells. Our results supports the hypothesis that crosstalk between integrin and cadherin, as well as regulation of E-cadherin localization and function,depends on the cell fate (Braga et al.,1999). The identification of the Rho GTPase and its partners that are involved in the network will contribute to the understanding of the mechanisms set up at the crypt-to-villus transition checkpoint. Coordinated changes in ECM components, integrin repertoire and E-cadherin localization might also result in migration of differentiating enterocytes along the crypt-to-villus axis.
Note in Proof
A similar conclusion was drawn from experiments performed in fibroblasts and recently published in this journal(Whittard and Akiyama, 2001a;Whittard and Akiyama,2001b).
We would like to acknowledge financial support from Université Paris VI, INSERM and CNRS. This work was performed using IFR 58 facilities. Cyrille Schreider is recipient of a fellowship from MRT then FRM (Fondation pour la Recherche Médicale). We would like to thank Christophe Klein (IFR 58) for excellent technical assistance.