Enterophilin-1 Interacts with Focal Adhesion Kinase and Decreases β1 Integrins in Intestinal Caco-2 Cells*

Intestinal cell growth and differentiation are tightly regulated by growth factors and extracellular matrix components along the crypt-villus axis. We previously described enterophilin-1 (Ent-1) as a new intestinal protein associated with growth arrest and enterocyte differentiation. Ent-1 interacted with sorting nexin 1 and decreased cell surface epidermal growth factor receptor. Because β1 integrins are mostly found in vivo in the proliferative crypt cells, we investigated the role of Ent-1 in the fate of β1 integrin subunits. In undifferentiated intestinal Caco-2 cells, overexpression of Ent-1 induces a marked decrease of α5β1 integrin pools, whereas α2β1 integrin is weakly affected. Conversely, overexpression of sorting nexin 1 has no effect on integrin levels despite its ability to interact with Ent-1. Interestingly, we identified focal adhesion kinase as a new Ent-1 partner using yeast two-hybrid screening and co-precipitation experiments. Furthermore by confocal microscopy, we observed that Ent-1 and β1 integrins partly co-localize on vesicular structures, suggesting a role for Ent-1 in integrin trafficking. Because focal adhesion kinase is able to bind both Ent-1 and β1 integrins, the kinase might act as a molecular bridge between the two proteins. Altogether, these results support a role of Ent-1 in regulating β1 integrin expression that could favor intestinal differentiation.

The intestinal epithelium undergoes continuous and rapid renewal, with proliferation of the multipotent stem cells limited to the crypts of Lieberkü hn. At the top of the crypt, the cells lose their proliferative ability and complete differentiation during a highly organized migration on the basement membrane, a specialized extracellular matrix consisting of a complex association of various molecules, including collagen IV, laminin, and fibronectin (1). At the tip of the villus, terminally differentiated cells are extruded to the lumen (2)(3)(4). Intestinal epithelial cell differentiation requires two major events: the transition from stem cells to committed proliferative cells and the irreversible loss of proliferative potential as the committed cells start to differentiate. However, the cellular and molecular mechanisms responsible for the fine coordination between proliferation, migration, and differentiation along the crypt-villus axis are still largely unknown.
The regulation of intestinal cell growth and differentiation is a multifactorial process, susceptible to various influences along the crypt-villus axis such as growth factors and basement membrane components. Indeed, in vivo and in vitro studies indicate that the expression of cell surface epidermal growth factor receptor (EGFR) 1 dramatically decreases during differentiation (5,6). Additionally, several studies have provided evidence for an important role for integrins during this process. Integrins are a family of heterodimeric cell surface adhesion receptors consisting of ␣ and ␤ subunits and mediating cellextracellular matrix interactions (for a review see Ref. 7). In epithelia, such interactions are mainly achieved by integrins belonging to the ␤ 1 and ␤ 4 classes. Interestingly, basement membrane components and ␤ integrins display a differential distribution along the crypt-villus axis. Briefly, fibronectin is abundant in the proliferative crypt compartment, whereas laminins are enriched at the villus tip (8,9). Similarly, ␤ 1 and ␤ 4 integrins show differential patterns of expression in concert to the distribution of their respective ligands. ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , and ␣ 5 ␤ 1 integrins are expressed in proliferative crypt cells and decrease in differentiated villus cells. In fact, ␤ 1 integrin classes are largely involved in the control of cell proliferation and particularly the specific fibronectin receptor, ␣ 5 ␤ 1 integrin, because of its link to the Ras-mitogen-activated protein kinase signaling pathway (10). Conversely, ␣ 3 ␤ 1 , ␣ 7B ␤ 1 , and ␣ 6 ␤ 4 integrins appear to be related to the process of intestinal cell differentiation (11,12). Nevertheless, the underlying mechanisms of such basement membrane and integrin compositional changes remain to be precisely defined.
We have previously cloned and characterized a new family of intestinal proteins named enterophilins (13). We have shown that enterophilin-1 (Ent-1) expression was positively correlated with growth arrest and terminal differentiation program in human colon adenocarcinoma cells. Furthermore, overexpression of Ent-1 inhibited cell growth and promoted intestinal differentiation (13). Using a yeast two-hybrid screen, we recently identified the sorting nexin 1 (SNX1) as an Ent-1-binding partner (14). The SNXs are an emerging family of proteins (for a review see Ref. 15), characterized by the presence of a phox homology domain (16). SNX1 has been particularly involved in cell surface EGFR removal as the result of enhancing trafficking in the endosome-to-lysosome pathway (17). Furthermore, we have reported a cooperative role of Ent-1 and SNX1 on the decrease of cell surface EGFR (14). All of these data led us to postulate a model involving Ent-1 in the down-regulation of mitogenic signals allowing cell growth arrest and enterocyte differentiation.
In this paper, we investigated the role of Ent-1 and SNX1 on the fate of ␤ 1 integrin subunits that are involved in the control of enterocyte differentiation. We herein report a role for Ent-1 in ␤ 1 integrin variations occurring during the early stages of intestinal differentiation that could be linked to the ability of Ent-1 to interact with focal adhesion kinase p125FAK, a nonreceptor protein-tyrosine kinase co-localized to sites of integrin clustering and involved in integrin-initiated signaling events (for a review see Ref. 18).
Yeast Two-hybrid Screening-First, full-length Ent-1 cDNA containing vector pAS2-1 (Clontech Laboratories) was used as a bait to screen a HeLa cell cDNA library constructed in pGAD GH vector at EcoRI/ XhoI restriction sites (Clontech Laboratories) as already described (14). To pinpoint the interacting domain of Ent-1, the leucine zipper region or the B30.2 domain were used in a one to one interaction test with positive library plasmids in the yeast strain SFY526 and tested as previously described (14).
Quantification of ␤ 1 Integrins by Real Time Reverse Transcription-PCR-Nonconfluent Caco-2 cells were transfected with pEGFP-C2-Ent-1, pEGFP-C1-SNX1, or empty vector for 72 h. Then total RNA was extracted using a Trizol reagent (Invitrogen), following the manufacturer's instructions. To remove traces of genomic DNA, 5 g of total RNA were treated with RNase-free DNase I (Invitrogen), and subsequently reverse transcription reaction was performed in a reaction buffer containing dNTP, "random" primers (Roche Applied Science) and Superscript II (Invitrogen). Real time quantitative reverse transcription-PCR was performed in an ABI Prism 7000 Sequence Detector (Applied Biosystems) using SYBR Green JumpStart TM detection system (Sigma). All of the primers were designed using Primer Express TM software (PE Biosystems). Briefly, 125 ng of the reversed transcribed cDNA were used for each PCR with 300 nM of the human ␤ 1 integrin specific primers (forward, 5Ј-TGCCGGGTTTCACTTTGC-3Ј, bases 940 -957; reverse, 5Ј-GTGACATTGTCCATCATTTGGTAAA-3Ј, bases 985-1009 (19)). The amplicon was 70 bp and was in the 5Ј coding region of the human ␤1A integrin cDNA (20). To normalize the difference in loading amounts, hepatic nuclear factor 4␣ (HNF4␣), which is expressed along the entire length of the crypt-villus axis (21), was used at 100 nM (forward primer, 5Ј-GAGGAACCAGTGCCGCTACT-3Ј; reverse primer, 5Ј-CATTCTGGACGGCTTCCTTCT-3Ј). The PCR program with specific primers was as follows: 50°C for 2 min, then 95°C for 10 min followed by 40 cycles of 95°C for 15 s, and 60°C for 1 min. The threshold cycle (C T ) values were obtained for the reactions reflecting quantity of the template in the samples. ␤ 1 integrin ⌬C T was calculated by subtracting the HNF4␣ C T value from the ␤ 1 integrin C T value and thus represented the relative quantity of the target molecule after normalizing with the internal standard HNF4␣. The ␤ 1 integrin ⌬C T values of Caco-2 cells transfected with pEGFP-C2-Ent-1 or pEGFP-C1-SNX1 were expressed as percentages of ␤ 1 integrin ⌬C T values of GFP control cells.
Confluence-induced Differentiation of Caco-2 Cells-Caco-2 cells were seeded at 18 000 cells/cm 2 and grown in 60-mm dishes in the complete medium. The cells were washed twice in phosphate-buffered saline (PBS) and scraped at 4°C in PBS containing protease and phosphatase inhibitors (1 mM phenylmethanesulfonyl fluoride, 10 g/ml leupeptin, 1 g/ml aprotinin, 20 mM NaF, 2 mM Na 3 VO 4 ) at various times up to 23 days after plating. The samples were sonicated, and the protein concentration was determined according to the method of Bradford (22).
Immunoprecipitation Experiments-Caco-2 cells were transfected with pcDNA3/Myc-His-Ent-1 for 48 h. All of the procedures were done at 4°C. The cells were washed twice with PBS, harvested, and then lysed for 30 min in lysis buffer (20 mM Tris-HCl, pH 7.8, 150 mM NaCl, 1% Nonidet P-40, 10% glycerol) containing protease and phosphatase inhibitors as described above. After sonication, insoluble debris was removed by centrifugation at 14,000 ϫ g for 15 min, and the supernatant was subjected to a preclearing with 30 l of protein G-Sepharose (Amersham Biosciences) for 30 min. Ent-1 was immunoprecipitated with 5 g of monoclonal anti-Myc antibody, followed by 50 l of protein G-Sepharose for 1 h. After two washes in lysis buffer containing 0.1% Nonidet P-40, protease, and phosphatase inhibitors, followed by one wash in lysis buffer without detergent, the bound proteins were eluted by boiling the beads in Laemmli buffer (23). The precipitated proteins were then submitted to SDS-polyacrylamide gel electrophoresis and detected by immunoblotting.
Immunoblotting-The proteins were separated by SDS-polyacrylamide gel electrophoresis under reducing conditions, transferred onto nitrocellulose membrane, and analyzed by immunoblotting according to standard protocols (23,24). Rabbit polyclonal anti-Ent-1 peptide antibody has been previously described (13). Rabbit polyclonal anti-␤ 1 integrin antibody (anti-cyto-␤ 1 ) raised against a synthetic peptide corresponding to the cytoplasmic domain of the ␤ 1 integrin (25) was a generous gift from Dr. Albigès-Rizo (La Tronche, France). Monoclonal antibodies against ␣ 5 and ␣ 2 integrin subunits were purchased from BD Transduction Laboratories. Monoclonal antibody against ␤-actin was purchased from Sigma, and monoclonal antibody against GFP was from Roche Applied Science. Monoclonal antibodies against c-Myc epitope (9E10) and against the C terminus of FAK (H-1) were from Santa Cruz Biotechnology. Revelation was done with horseradish peroxidase-conjugated anti-mouse IgG or with anti-rabbit IgG antibody (Promega) and ECL (Amersham Biosciences) detection.
Confocal Microscopy-Caco-2 were grown on fibronectin-coated sterile glass coverslips for 38 h and then transfected with pEGFP-C2-Ent-1. After 48 h, the cells were washed twice with ice-cold PBS, fixed for 15 min with 3% formaldehyde, permeabilized with 0.2% Triton X-100, and saturated for 30 min with PBS containing 1% bovine serum albumin. The cells were then incubated 60 min with the primary antibody (monoclonal anti-FAK (clone 77) antibody from BD Transduction Laboratories or rabbit polyclonal anti-␤ 1 integrin (anti-cyto-␤ 1 ) antibody). Immunostaining was performed respectively with tetramethylrhodamine isothiocyanate-conjugated anti-mouse IgG (DAKO) or with Cy TM 5-conjugated anti-rabbit IgG (Jackson ImmunoResearch) secondary antibodies. The coverslips were mounted in Vectashield (Vector laboratories) and examined with a confocal laser scanning microscope (Zeiss; LSM 510).
Analysis of Cell Surface Integrin Density by Flow Cytometry-Caco-2 cells were transfected with pEGFP-C2-Ent-1, pEGFP-C1-SNX1, or empty vector and grown for 72 h. Then they were harvested in PBS containing 2 mM EDTA and fixed in 1% formaldehyde for 30 min, and nonspecific sites were saturated with 1% bovine serum albumin for 30 min. The cell surface ␤ 1 integrins were detected by an incubation with primary monoclonal antibodies against ␤ 1 integrin subunits (clone K20; Immunotech), ␣ 5 ␤ 1 integrin (clone P1D6; DAKO), or ␣ 2 ␤ 1 integrin (clone P1E6; DAKO) for 60 min, followed by incubation with a Rphycoerythrin (RPE)-conjugated goat anti-mouse IgG secondary antibody (DAKO) for 60 min. The cells were then analyzed by flow cytometry (Coulter XL 4C). Expression of the GFP-fused proteins was monitored by fluorescence measurement, and the surface binding of the anti-␤1, anti-␣ 5 ␤ 1 , or anti-␣ 2 ␤ 1 integrin antibody was measured specifically on the transfected cell population. The quantity of cell surface ␤ 1 integrin subunits and ␣ 5 ␤ 1 or ␣ 2 ␤ 1 integrins was measured by RPE fluorescence mean and was expressed as percentages of ␤ 1 subunit or ␣ 5 ␤ 1 integrin or ␣ 2 ␤ 1 integrin values in GFP-transfected control cells. The results were analyzed by paired t test and were considered significantly different from GFP-transfected control cells (*, p Ͻ 0.05; **, p Ͻ 0.01; *** p Ͻ 0.00005).
Alkaline Phosphatase Activity-Alkaline phosphatase activity was determined in the Caco-2 cell extracts as previously described (13).

Enterophilin-1 Expression Is Correlated to the Decrease of ␤ 1 Integrins during Confluence-induced Differentiation of Caco-2
Cells-Because ␤ 1 integrin expression was described to decrease in vivo during intestinal differentiation along the cryptvillus axis, we first investigated the ␤ 1 integrin variations in human colon carcinoma Caco-2 cells. These cells have proven to be the most useful in vitro model because of their unique ability to spontaneously differentiate at confluence, displaying morphological and functional characteristics of enterocytes (26). We thus analyzed the ␤ 1 integrin expression pattern during the confluence-induced differentiation of Caco-2 cells. After 7 days, the cells reached confluence and began to express the intestinal differentiation marker, alkaline phosphatase (Fig. 1A). As expected, Western blot analysis showed a marked decrease of ␤ 1 integrin subunits beginning at day 14 (Fig. 1B), which correlated with the appearance of the 65-kDa Ent-1 protein. As previously reported (13,14), the human ortholog of Ent-1 was weakly expressed at day 9, with a high level of expression from days 12 to 23 of culture. Interestingly, we had previously observed a similar correlation between Ent-1 expression and EGFR decrease during Caco-2 cell differentiation (14).
Enterophilin-1 Overexpression Induces ␤ 1 Integrin Decrease in Intestinal Caco-2 Cells-Undifferentiated Caco-2 cells were transfected with pEGFP-C2-Ent-1, pEGFP-C1-SNX1, or empty vector. Then cell surface ␤ 1 integrin subunits were immunolabeled and specifically measured in the transfected cell population by flow cytometry. Interestingly, after 72 h of transfection, overexpression of Ent-1 induced a significant decrease of ϳ25% of plasma membrane ␤ 1 integrin subunits, compared with the GFP-transfected cell population used as control ( Fig. 2A). Conversely, overexpressed SNX1 was unable to decrease plasma membrane ␤ 1 integrin subunit pool ( Fig. 2A). Thus, our results indicate that Ent-1 could down-regulate not only the cell sur-face EGFR as we previously demonstrated (14) but also cell surface ␤ 1 integrin subunits.
To further document the ␤ 1 integrin diminution, transfected Caco-2 cells were also analyzed by immunoblotting after 72 h of transfection (Fig. 2B). Consistent with the flow cytometry data, we observed that Ent-1 induced a marked decrease of ␤ 1 integrin subunit expression compared with the GFP-transfected control cells, whereas SNX1 did not. Altogether, these results suggest that Ent-1 contributed to the down-regulation of ␤ 1 integrin subunits in intestinal cells.
Enterophilin-1 Induces ␣ 5 ␤ 1 and ␣ 2 ␤ 1 Integrin Variations in Caco-2 Cells-␣ 5 ␤ 1 integrin is largely involved in the proliferative status of intestinal crypt cells. However, other changes in ␤ 1 integrins such as ␣ 2 ␤ 1 integrin were also monitored in vivo during epithelial intestinal differentiation. We thus checked for the effect of Ent-1 on both ␣ 5 ␤ 1 and ␣ 2 ␤ 1 integrins in Caco-2 cells. We first verified that in the Caco-2 cell model, both integrins displayed the same variations of expression as reported during in vivo intestinal differentiation. Equal amounts of proteins from undifferentiated (day 4) and fully differentiated (day 23) Caco-2 cells were analyzed by immunoblotting with an anti-␣ 5 or anti-␣ 2 subunit antibody (Fig. 3A). We thus confirmed that ␣ 5 and ␣ 2 subunits, like the ␤ 1 subunits (Fig.  1B), were highly expressed in undifferentiated proliferative Caco-2 cells, but they were strongly decreased in differentiated cells.
Consequently, we investigated the effect of Ent-1 overexpression on the cell surface integrin pool by flow cytometry using an anti-␣ 5 ␤ 1 or anti-␣ 2 ␤ 1 integrin antibody. Undifferentiated Caco-2 cells were transfected with pEGFP-C2-Ent-1, pEGFP-C1-SNX1, or empty vector for 72 h, and cell surface integrins were immunolabeled and analyzed by flow cytometry as described above. The results showed a significant 30% decrease of the plasma membrane pool of ␣ 5 ␤ 1 integrins in Ent-1-trans- FIG. 1. Ent-1 expression correlates with the decrease of ␤ 1 integrin expression during Caco-2 cell spontaneous differentiation. Caco-2 cells were grown in 60-mm dishes and harvested by scraping at various times up to 23 days after plating. A, alkaline phosphatase activity was determined at the indicated times as described under "Experimental Procedures." The data are expressed as IU (moles of substrate hydrolyzed per min)/gram of proteins. B, equal amounts of proteins were subjected to SDS-polyacrylamide gel electrophoresis under reducing conditions. Ent-1 was detected with rabbit polyclonal anti-Ent-1 peptide antibody and ␤ 1 integrin with rabbit polyclonal anti-␤ 1 integrin antibody (anti-cyto-␤ 1 ) as described under "Experimental Procedures." fected cells, compared with the GFP-transfected control cells (Fig. 3B, left panel), whereas a weaker diminution (ϳ15%) was observed for ␣2␤ 1 integrin (Fig. 3B, right panel). Conversely, as shown in Fig. 2A, overexpressed SNX1 induced no change on cell surface ␣ 5 ␤ 1 or ␣2␤ 1 integrins (Fig. 3B).
Additionally, ␣ 5 or ␣ 2 integrin expression patterns were evaluated by immunoblotting in transfected Caco-2 cells (Fig. 3C). Concordant with flow cytometry analysis, Ent-1-transfected cell extracts displayed a lower level of ␣ 5 integrin subunit compared with the GFP-transfected cells (Fig. 3C), whereas the decrease of ␣ 2 ␤ 1 integrin was smaller. SNX1 had no effect on ␣ 5 as well as on ␣ 2 integrin subunit content. Altogether, these results support a role of Ent-1 in regulating the ␤ 1 integrin changes that occur during the early stages of intestinal differentiation.
Enterophilin-1 Down-regulates ␤ 1 Integrin mRNA Expression-Because Ent-1 induced a marked decrease of ␤ 1 integrin pools, we investigated whether this was related to changes in ␤ 1 integrin gene expression. We thus analyzed ␤ 1 integrin mRNA amounts by real time quantitative reverse transcription-PCR. Undifferentiated Caco-2 cells were transfected with pEGFP-C2-Ent-1, pEGFP-C1-SNX1, or empty vector, and ␤ 1 integrin mRNA were quantified. Consistent with the decrease of ␤ 1 integrin protein levels ( Fig. 2A), we observed a significant reduction (ϳ50%) in the amount of ␤ 1 integrin mRNA in Ent-1-transfected Caco-2 cells compared with the control cells after 72 h of transfection (Fig. 4). By contrast, SNX1 overexpression did not diminish ␤ 1 integrin mRNA, even inducing a statistically nonsignificant increase (Fig. 4). This is in good agreement with the lack of effect on ␤ 1 integrin protein levels (Fig. 2). Altogether, these results suggest that Ent-1 overexpression modifies the expression of ␤ 1 integrin both at the protein and mRNA levels.
FAK Is a New Partner of Enterophilin-1-Using full-length Ent-1 cDNA as a bait in a HeLa cell cDNA library yeast two-hybrid screening, we identified FAK as an Ent-1 partner. Interestingly, the tyrosine kinase FAK is largely involved in integrin signaling pathways, and its N-terminal region is known to bind ␤-integrin cytoplasmic domains (for reviews see Refs. 7 and 18). Two positive clones encoding FAK were isolated in a first screening (Fig. 5A). We then performed the two-hybrid one to one interaction assay with the leucine zipper part or the B30.2 domain as bait to pinpoint the domain of Ent-1 interacting with FAK. The results showed that the B30.2 domain of Ent-1 was sufficient for interaction with FAK. Furthermore, analysis of FAK clones led us to conclude that the C-terminal focal adhesion targeting signal (FAT) domain is the minimal region of FAK interacting with Ent-1. Interestingly, the FAT domain is necessary and sufficient for recruiting FAK to focal adhesion contact sites and also contains binding sites for a number of signaling molecules and cytoskeletal proteins such as talin and paxillin (27,28) that can associate with ␤-integrin cytoplasmic domains. Thus, in addition to SNX1 involved in the plasma membrane receptor trafficking (14), FAK, the major integrin signaling transducing element, is a new partner of Ent-1.
We further investigated the interaction between Ent-1 and FAK by co-immunoprecipitation experiments in intestinal mammalian cells. Caco-2 cells were transfected with pcDNA3/Myc-His-Ent-1 vector, and Ent-1 was immunoprecipitated with an anti-Myc antibody after 48 h of transfection. The results revealed that FAK was co-precipitated with Ent-1, confirming the interaction of these two proteins in intestinal cells (Fig. 5B). Therefore FAK, which is able to bind both Ent-1 and ␤ 1 integrins, could act as a molecular bridge between the two proteins.
Enterophilin-1, ␤ 1 Integrins, and FAK Partly Co-localize in Caco-2 Cells-We previously demonstrated that Ent-1 co-localizes with SNX1 on vesicular structures and suggested the involvement of these proteins in the intracellular trafficking of EGFR (14). Because Ent-1 was also able to decrease the cell surface ␤ 1 integrin pool and to interact with FAK, we further analyzed by confocal microscopy the cellular distribution of the three proteins in intestinal Caco-2 cells. Each horizontal column of images represents a 0.2-m optical Z section of the same cells (Fig. 6). As previously observed after 48 h of transfection (13,14), Ent-1-transfected Caco-2 cells displayed an intracellular punctate distribution (Fig. 6, column A, green). ␤ 1 integrin pools showed membrane staining as well as a strong intracellular patched staining (Fig. 6, column B, blue) as already shown in Caco-2 and breast carcinoma cells (29,30). Interestingly, some of Ent-1-containing vesicles merged with part of the intracellular ␤ 1 integrin signal (Fig. 6, column D, white arrowheads), suggesting that Ent-1 could play a role in ␤ 1 integrin trafficking in intestinal cells. Furthermore, FAK exhibited typical staining at the cell membrane periphery in addition to adhesion complexes throughout the cell (Fig. 6,   FIG. 2. Ent-1

enhances the decrease of ␤ 1 integrin subunits in Caco-2 cells.
A, undifferentiated Caco-2 cells were transfected with pEGFP-C2-Ent-1 (stippled bar), pEGFP-C1-SNX1 (hatched bar), or empty vector (control, black bar) and grown for 72 h. The cells were then labeled with monoclonal anti-␤ 1 integrin (K20) primary antibody and RPE-conjugated secondary antibody. The relative density of cell surface ␤ 1 integrin subunit was specifically evaluated in the transfected cell population by flow cytometry. The percentage of cell surface ␤ 1 integrin was measured by RPE fluorescence mean and compared with the GFPtransfected cell population used as a control. The values represented the mean and standard error of at least three independent experiments. The results were analyzed by a paired t test and were considered significantly different from control cells (***, p Ͻ 0.00005) or nonsignificantly different (ns). B, equal amounts of proteins from transfected cell extracts were separated and analyzed by immunoblotting using rabbit polyclonal anti-␤ 1 integrin antibody (anti-cyto-␤ 1 ) and monoclonal anti-␤ actin antibody for loading control. Lane 1, empty vector; lane 2, GFP-Ent-1; lane 3, GFP-SNX1. column C, red). We noticed that part of the three protein staining co-located at the cell membrane periphery as well as on intracellular patched structures (Fig. 6, column D, white  arrowheads). In fact, Ent-1 and FAK could be both found in the same location, confirming their possible interaction in intestinal cells. DISCUSSION Little is known regarding the mechanisms involved in the regulation of cell growth and differentiation in the human intestinal epithelium. A key question in intestinal development is what triggers the differentiation process. Determination of integrin expression patterns along the crypt-villus axis has provided valuable information relative to the involvement of each of these extracellular matrix receptors in proliferative or differentiated cell status. Indeed, cell surface ␤ 1 integrin pools are mainly found in proliferative crypt cells and markedly decrease during enterocyte differentiation (9,12,31). Among ␤ 1 integrin classes, ␣ 5 ␤ 1 integrin, the specific fibronectin receptor, is particularly involved in cell growth by activating mitogenactivated protein kinases (10). Interestingly, we showed that Ent-1 overexpression induced a significant decrease of cell surface ␤ 1 integrin pools in undifferentiated Caco-2 cells. This was FIG. 3. Ent-1 induces a significant decrease of ␣ 5 ␤ 1 integrin and a slighter diminution of ␣ 2 ␤ 1 integrin in Caco-2 cells. A, equal amounts of proteins from undifferentiated (lanes 1 and 3, day 4 after plating) and differentiated (lanes 2 and 4, day 23 after plating) Caco-2 cells were subjected to SDS-polyacrylamide gel electrophoresis under reducing conditions. ␣ 5 or ␣ 2 integrin was detected with monoclonal anti-␣ 5 or anti-␣ 2 integrin antibody. B, nonconfluent Caco-2 cells were transfected with pEGFP-C2-Ent-1 (stippled bar), pEGFP-C1-SNX1 (hatched bar), or empty vector (control, black bar) and grown for 72 h. The cells were then labeled with monoclonal anti-␣ 5 ␤ 1 integrin (P1D6) or anti-␣ 2 ␤ 1 integrin (P1E6) primary antibody and RPE-conjugated secondary antibody and then analyzed by flow cytometry. The percentage of cell surface ␣5␤1 or ␣ 2 ␤ 1 integrin in the transfected cell population was measured by RPE fluorescence mean and compared with the GFP-transfected cell population used as control, as explained in Fig.  2. The values represented the means and standard errors of three independent experiments. The results were analyzed by paired t test and were considered significantly different from control cells (*, p Ͻ 0.05; **, p Ͻ 0.01) or nonsignificantly different (ns). C, equal amounts of proteins from transfected cell extracts were subjected to SDS-polyacrylamide gel electrophoresis. ␣ 5 or ␣ 2 integrin subunits were detected with monoclonal anti-␣ 5 or anti-␣ 2 integrin antibody. Lanes 1, 3, and 5, empty vector; lanes 2 and 6, GFP-Ent-1; lanes 4 and 7, GFP-SNX1. consistent with the observed correlation between the increased Ent-1 expression and the diminution of ␤ 1 integrins during spontaneous Caco-2 cell differentiation. In addition, we observed that the Ent-1-induced cell surface ␤ 1 integrin decrease led to a subsequent diminution of ␤ 1 integrin mRNA. Thus, Ent-1 will participate in the integrin changes occurring during the intestinal differentiation process by triggering cell surface ␤ 1 integrin removal, and also, this will induce the down-regulation of ␤ 1 integrin gene expression, thus sustaining the differentiation process as already noticed in Caco-2 cells (29,32). Interestingly in human epidermal keratinocytes, commitment to terminal differentiation first results in modulation of preexisting ␤ 1 integrin subunits at the cell surface, and later the receptor is lost from the cell surface and the level of the subunits mRNA declines (33).
Furthermore, investigating the fate of ␣ integrin subunits enriched in proliferative crypt cells, we reported a dramatic decrease of ␣ 5 ␤ 1 integrin in Ent-1-transfected undifferentiated cells. Additionally, we observed that overexpressed Ent-1 caused a slight decrease of ␣2␤ 1 integrin. These data can be correlated with our previously published results demonstrating a role of Ent-1 on cell surface EGFR removal (14).
Growth factor receptors and integrins display not only physical association (34) but also functional cooperation (35). However, the interconnection between growth factor receptor and integrin signaling pathways in the control of cell functions is a complex phenomenon. Although both EGFR and ␣ 5 ␤ 1 integrin can independently stimulate Ras/mitogen-activated protein kinase signaling pathway, an emerging concept suggests a minimum activation threshold to drive cell proliferation that would require input from both EGFR and integrins (36 -38). In Caco-2 cells, Kuwada and Li (39) described that ␣ 5 ␤ 1 integrin potentiates the EGFR-dependent cell proliferation. Additionally, ␣2␤ 1 integrin, which is proposed to be a collagen receptor, has also been reported to function in EGFR activation (35,40). The extensive cross-talk occurring between integrins and growth factor receptors is particularly relevant in intestinal dynamic processes. In this context, the Ent-1-induced decrease of EGFR as well as ␣ 5 ␤ 1 and ␣2␤ 1 integrins could be functionally important in the early stages of growth arrest leading to enterocyte differentiation. This is particularly interesting because in vivo Ent-1 expression at the mid-villus could be related to the reported decrease of EGFR and ␤ 1 integrins (5,9). Furthermore, we previously showed a synergetic effect of Ent-1 and SNX1 on EGFR cell surface removal. Conversely, we found that SNX1 had effect on neither ␣ 5 ␤ 1 nor ␣ 2 ␤ 1 integrin expression. We hypothesized that Ent-1 could interact with another SNX protein or another yet unidentified Ent-1 partner to mediate ␤ 1 integrin variations.
Moreover, seeking for Ent-1 cellular partners to precisely determine the molecular mechanisms involved in Ent-1-induced cell surface ␤ 1 integrin removal, we demonstrated that Ent-1 interacted with FAK. This interaction required the Cterminal B30. will open perspectives to precisely define the B30.2 domain function and then the role of proteins belonging to the B30.2 domain family of proteins.
It is well documented that FAK plays important functions in integrin-initiated signaling events (for reviews see Refs. 7 and 18). The FAT domain of FAK is necessary and sufficient for recruiting the kinase to focal adhesion contact sites. It contains binding sites for cytoskeletal proteins such as talin and paxillin (27,28) that can associate with ␤-integrin cytoplasmic domains (45,46). In addition, the group of Schlaepfer (47) has recently demonstrated that the N-terminal domain of FAK can associate, through indirect interaction, with an activated EGFR complex, emphasizing the critical position of FAK as a connecting component to both integrin and growth factor receptors. Interestingly, the authors suggested that such interaction may be mediated by one or more intermediary bridging proteins. Therefore FAK, which is able to bind both Ent-1 and ␤ 1 integrins, could act as a common and important connection between the two proteins. Furthermore, identification of both SNX1 and FAK as Ent-1 partners indicated that Ent-1 could be part of a multiprotein complex involved in the cross-talk between integrins and growth factor signaling pathways.
In addition, by confocal microscopy, we observed a partial co-localization of Ent-1 and ␤ 1 integrin subunits on vesicular structures, suggesting that Ent-1 could regulate the integrin trafficking. Indeed, we previously reported that Ent-1 and SNX1 co-localized on vesicular and tubulovesicular structures, which were different from early endosome antigen 1-containing endosomes (14). However, more studies are needed to determine the identity of the transport vesicles carrying Ent-1 and ␤ 1 integrins as well as its association with the other endocytic markers.
The exact mechanisms regulating the intracellular integrin trafficking are still poorly understood. It has been postulated that cell migration is favored by internalizing integrins at the rear of the cells and transporting them forward within vesicles at the leading edge to form new contacts with the extracellular matrix (48). However, those studies were mainly done with migrating fibroblasts that behave differently than polarized epithelial cells. In fact during their differentiation process and their cell life, the intestinal cells are continuously moving in coherent sheets along the villus, each cell tightly linked to its neighbors via lateral adhesive complexes. The Ent-1-induced ␤ 1 integrin removal from the cell surface of intestinal Caco-2 cells will contribute to the rapid switch in the integrin repertoire that accompanies intestinal differentiation and allows cell migration along the crypt-villus axis (9,49). Thus, our study is providing new insight into the ␤ 1 integrin trafficking in polarized intestinal cells.
To summarize, the Ent-1-induced-␤ 1 integrin decrease in intestinal cells adds to our understanding of the role of Ent-1 in the regulation of growth arrest in intestinal epithelium that we previously reported (13,14). Moreover, early changes in ␣ 5 ␤ 1 and ␣ 2 ␤ 1 integrin variations have a decisive impact on the behavior of committed crypt cells toward terminal differentiation. Alterations in controlled integrin expression could be a crucial step in colorectal tumorigenesis, as previously reported in epidermal squamous cell carcinoma (50) and in breast cancer (51). Because we previously reported that Ent-1 in cooperation with its binding partner, SNX1, was able to regulate EGFR vesicular trafficking (14), we can hypothesize that Ent-1 may similarly link integrin-containing vesicles to molecular motors involved in vesicle transport along the cytoskeleton network. The next challenge will be to explore the molecular mechanisms involving Ent-1 in both EGFR and integrin variations during intestinal cell differentiation.