The cell-binding domain of intimin from enteropathogenic Escherichia coli binds to beta1 integrins.

Bacteria interact with mammalian cells surface molecules, such as integrins, to colonize tissues and evade immunological detection. Herein, the ability of intimin, an outer membrane protein from enteropathogenic Escherichia coli, to bind β1 integrins was investigated. Solid-phase binding assays revealed binding of the carboxyl-terminal 280 amino acids of intimin (Int280) to α4β1 and α5β1 integrins. The binding required divalent ions (in particular, it was enhanced by Mn2+) and was inhibited by an RGD-containing peptide. Nonderivatized Int280, but not Int280CS (like Int280 but with Cys-937 replaced by Ser) blocked the binding of biotinylated Int280 to integrins. Int280 did not efficiently inhibit β1 integrin binding of invasin from Yersinia pseudotuberculosis. Both intimin and invasin, immobilized on plastic surfaces, mediated adherence of resting or phorbol 12-myristate 13-acetate-activated human CD4+ T cells, whereas fibronectin mediated the adherence of only activated T cells. T cell binding to intimin and invasin was integrin mediated because it was specifically blocked by an RGD-containing peptide and by antibodies directed against the integrin subunits β1, α4, and α5. These results demonstrate a specific integrin binding activity for intimin that is related to, but distinct from, that of invasin.


Int280 to integrins. Int280 did not efficiently inhibit ␤ 1 integrin binding of invasin from Yersinia pseudotuberculosis. Both intimin and invasin, immobilized on plastic surfaces, mediated adherence of resting or phorbol 12-myristate 13-acetate-activated human CD4 ؉ T cells, whereas fibronectin mediated the adherence of only activated T cells. T cell binding to intimin and invasin was integrin mediated because it was specifically blocked by
an RGD-containing peptide and by antibodies directed against the integrin subunits ␤ 1 , ␣ 4 , and ␣ 5 . These results demonstrate a specific integrin binding activity for intimin that is related to, but distinct from, that of invasin.
Enteropathogenic Escherichia coli (EPEC) 1 is a major cause of diarrhea in early childhood (1). Characteristic "attachment and effacing" lesions can be visualized in electron micrographs of intestinal epithelial cells from patients with EPEC infections (2). During the formation of attachment and effacing lesions, major changes are triggered in the eukaryotic cell, including actin polymerization (3) and phosphorylation of several host proteins (4). Multiple bacterial proteins are required for the formation of attachment and effacing lesions (5). These include EaeB (6, 7), a secreted protein involved in activation of host cell signal transduction pathways, and intimin, a 94-kDa outer membrane protein that mediates the intimate attachment of EPEC to epithelial cells in vitro and in vivo via its carboxylterminal domain (8 -10). As a route toward understanding the pathogenesis of EPEC disease, we have recently expressed the 280-amino acid carboxyl-terminal domain of intimin as a maltose-binding protein (MBP) fusion (MBP-Int280) and characterized the cell binding activities of the fusion protein (10,11).
EPEC shows a limited ability to invade intestinal epithelial cells in vitro (12) and has been observed inside enterocytes in biopsies from infected patients (13). Intimin, which is a homologue of the invasin polypeptide of Yersinia (14), is necessary but not sufficient for invasion of EPEC into epithelial cells (6). The cell surface receptors for invasin are the ␣ 3-6 ␤ 1 integrins (15), proteins that mediate communication between the cytoskeleton and extracellular matrix in eukaryotes (16,17). A 76-amino acid loop formed by disulfide linkage between two Cys residues (at positions 907 and 982) of Yersinia invasin is required for invasin-mediated cell binding and invasion (18). The amino acid sequences of intimin and invasin are highly conserved along their amino-terminal regions (19), but the carboxyl-terminal cell-binding domains show a more limited similarity. Nevertheless, each domain contains two Cys residues in similar locations (10,18,19). Recently, we have shown that substitution of Cys-937 with Ser (Int280CS) abrogated the cell binding capacity of Int280 (11). In addition, MBP-Int280, but not MBP, MBP-Int280CS, or MBP-Inv280 (an MBP fusion protein containing the carboxyl-terminal 280-amino acid cellbinding domain of invasin from Yersinia pseudotuberculosis) inhibited EPEC entry into HEp-2 cells (11), whereas preincubation with invasin, but not with intimin, inhibited cell spreading on invasin-coated surfaces (10). These results suggest that intimin and invasin either bind different receptors or bind the same receptor with different mechanisms.
To investigate the ability of intimin to bind integrins in vitro, solid-phase binding assays using the intimin cell-binding domain and purified integrins and T cell adherence assays to intimin-coated surfaces were performed. MBP-Inv280 and MBP-Inv280CS (as Inv280 with Cys-to-Ser substitution at position 982) (11) were used as a reference and controls. We found that biotinylated MBP-Int280 (b-MBP-Int280) bound to purified integrins and that the binding could be inhibited by preincubating the wells with unlabeled MBP-Int280 but not with MBP-Int280CS. MBP-Int280 was unable to block binding of MBP-Inv280 to the purified ␤ 1 integrins; however, binding of both intimin and invasin was inhibited by an Arg-Gly-Asp (RGD)-containing peptide. In addition, the cell-binding domains of intimin and invasin were able to mediate adherence of T cells to coated plastic wells, in both resting and PMA-activated states. The adherence of T cells to the coated surfaces was specifically blocked by RGD-containing peptide and by some anti-integrin antibodies but not by anti-CD4 and anti-Lselectin antibodies. These results indicate that intimin exhibits a distinct ␤ 1 integrin binding activity in vitro.
Solid-phase Binding of MBP-Int280 to Purified Integrins-Solidphase ligand receptor binding was performed by a modification of the method of Charo et al. (23). Ligands were mixed with equal amounts of ImmunoPure sulfo-N-hydroxysuccinimide biotin (Pierce) and rotary mixed for 30 -40 min at room temperature. The mixture was then dialyzed against several changes of 150 mM NaCl, 25 mM Tris-HCl (pH 7.4) to remove free biotin. Purified integrins (at a concentration of 0.1-0.5 mg/ml) were diluted 1:100 with PBS containing divalent cations, and 100-l aliquots were added to the wells of Immulon 96-well (Dynatech Laboratories, Inc.) plates. Plates were incubated overnight at 4°C and were then blocked with 200 l of 5% BSA, 150 mM NaCl, 0.05% NaN 3 , 10 mM Tris-HCl (pH 7.4) overnight. Wells were then washed twice with 200 l of 150 mM NaCl, 5 mM MnCl 2 , 25 mM Tris-HCl (pH 7.4), 1 mg/ml BSA (buffer A), and 100-l aliquots of 0.1-0.2 g/ml biotinylated ligands in buffer A were added, with or without competitors. In experiments examining the cation dependence of integrin-ligand interactions, ligand binding was also measured in the presence of 5 mM MgCl 2 , CaCl 2 , or EDTA in place of MnCl 2 . The plates were then incubated at 30°C for 3 h. Biotinylated ligands were aspirated, and the wells were washed three times with buffer A. Bound ligand was quantified by addition of 100 l of 1:200 Avidin-peroxidase conjugate (Sigma) in buffer A, and color was developed using ABTS (Sigma). The level of nonspecific binding was measured in every experiment by determining the level of binding to wells coated with BSA alone. These values were subtracted from the corresponding values for receptor-coated wells.
T Cell Adherence Assays-Before their addition to the coated microtiter wells, T cells were radiolabeled with 51 Cr as described previously (25), washed, and resuspended in adhesion medium (RPMI 1640 supplemented with 1% BSA, 1 mM Ca 2ϩ , 1 mM Mg 2ϩ , 1 mM Mn 2ϩ , 1% sodium pyruvate, 1% HEPES). 10 5 labeled T cells in 100 l adhesion medium with or without 25 ng PMA (Sigma) were added to the coated wells. Where indicated, the CD4 ϩ T cells were pretreated with blocking monoclonal antibodies (diluted 1:200 -450) specific for the integrin subunits ␤ 1 (code MCA667, Serotec Ltd., Oxford, United Kingdom), ␣ 1 (clone 1B3.1 kindly provided by Dr. Ilan Bank, Sheba Medical Center, Tel Hashomer, Israel) (26,27), ␣ 4 and ␣ 5 (clones P4C2 and CLB705, respectively, both obtained from Chemicon International, Inc.), and L-selectin and CD4 markers. To determine the ability of RGD-containing peptide to inhibit adherence of T cells to the coated wells, the assay was performed in the presence of either GRGDSPK or GRGESP (250 g/ml) peptides. The microtiter plates were then incubated (30 min, 37°C, 10% CO 2 humidified atmosphere) and washed three times with PBS to remove nonadherent cells. The adherent T cells were lysed (25), and the resulting supernatants were removed for counting. For each experimental group, the results were expressed as the mean percent of adherent T cells in quadruplicate wells.
Statistics-Data from representative experiments are presented as mean Ϯ S.E. Means were compared by using Student's t test, and P values of 0.05 or less were considered statistically significant.

Binding of b-MBP-Int280 to Purified
Integrins-The amino acid sequence similarity and the apparent comparable structural requirements (11,18) suggest that, like invasin, intimin may have the ability to bind integrins. To investigate this, the biotinylated cell-binding domain of intimin, purified as MBP-Int280 (b-MBP-Int280), was incubated under different conditions with purified integrins in solid-phase binding assays. b-MBP-Inv280 was used as a positive control. Preliminary assays using b-MBP-Int280 and a mixed ␤ 1 integrin preparation revealed that this protein showed affinity for integrins and that the binding was more efficient in the presence of Mn 2ϩ ions compared with the binding in the presence of Ca 2ϩ or Mg 2ϩ ions (Fig. 1). b-MBP-Int280 and b-MBP-Inv280 bound to both purified ␣ 4 ␤ 1 and ␣ 5 ␤ 1 integrins, although b-MBP-Inv280 showed better binding to the latter ( Fig. 2A). Significantly, reduced binding was seen when the reaction was performed in the presence of 5 mM EDTA (Figs. 1 and 2A). The reduction in Int or Inv binding in the presence of EDTA was the same as that caused when a large excess of unlabeled Int or Inv was added to inhibit the binding of biotinylated proteins (data not shown). The component of the binding that cannot be inhibited by EDTA is therefore probably nonspecific.
Binding of b-MBP-Int280 to Purified ␣ 5 ␤ 1 Integrins in the Presence of an RGD-containing Peptide-The RGD motif is present in the active site of several integrin ligands, and Tranvan-Nhieu and Isberg (28) have shown that 75% of integrinbound MBP-Inv479 can be displaced by an RGD-containing peptide. To determine the ability of such a peptide to compete with the binding of b-MBP-Int280, binding experiments to purified ␣ 5 ␤ 1 integrins in the presence of GRGDS and GRDGS peptides (100 g/ml) were performed. These binding assays revealed that the GRGDS peptide inhibited b-MBP-Int280 binding by 64%, to the same level as EDTA (i.e. to the level of nonspecific binding), whereas no effect on the binding was seen with the control GRDGS peptide (Fig. 2B). Similar results were obtained with b-MBP-Inv280 (not shown).
Solid-phase Cross-inhibition Binding Assays-To further determine the specificity of the binding exhibited by b-MBP-Int280 to the purified integrins, cross-inhibition binding assays were performed using unlabeled MBP-Int40, an MBP fusion protein that contains 40 amino acids of intimin upstream from Cys-860 (11), and unlabeled MBP-Int280 and MBP-Int280CS as competitors. In the presence of unlabeled MBP-Int280, the binding of b-MBP-Int280 was inhibited in a dose-dependent manner (Fig. 3A). Binding of b-MBP-Int280 to ␤ 1 integrins in the presence of MBP-Int40 and MBP-Int280CS was not significantly affected (Fig. 3A). In parallel experiments, it was shown that although unlabeled MBP-Inv280 inhibited binding of b-MBP-Inv280, binding was reduced to a much lesser extent by coincubation with MBP-Int280 (Fig. 3B).
Adherence of T Cells to MBP-Int-coated Surfaces-To test the integrin binding activity exhibited by the intimin cell-binding domain using a physiological assay, the ability of MBP-Int280, immobilized to the surface of 96-well plates, to mediate adherence of human CD4 ϩ T cells was investigated. Previous studies have shown that nonactivated human T cells bind invasin (29), unlike binding to other T cell-specific integrin ligands (e.g. fibronectin and laminin) (30). Accordingly, in this study the binding of both resting and PMA-activated T cells was investigated. Fibronectin, MBP-Int280CS, MBP-Inv280, MBP-Inv280CS, and BSA were used as controls. T cells adhered in a similar manner to both MBP-Int280-and MBP-Inv280-coated wells, whether in activated or resting states (Fig. 4, A-D). In contrast, and in agreement with published results (31), adherence to fibronectin-coated surfaces was observed only by T cells undergoing PMA-mediated activation (Fig. 4E). No adherence above that observed for BSA (Fig. 4E) was observed when the T cells were added to microtiter wells coated with either MBP-Int280CS or MBP-Inv280CS (Fig. 4, A-D). Thus, immobilized intimin appears to mediate the binding of PMA-activated and resting CD4 ϩ T lymphocytes in a ligand-specific manner.
Blocking of T Cell Adherence by Monoclonal Antibodies and RGD-containing Peptide-To determine whether the binding of T cells to the coated wells was indeed mediated by ␤ 1 integrins, cell adherence assays were performed in the presence of different monoclonal antibodies directed against the ␣ or ␤ chains of the ␤ 1 integrins. Monoclonal antibodies to L-selectin and CD4 were used as negative controls. Anti-␤ 1 , ␣ 4 , and ␣ 5 integrin monoclonal antibodies blocked the adherence of resting and PMA-activated human CD4 ϩ T cells to intimin-and invasincoated surfaces and, as expected, of the PMA-activated T cells to fibronectin (Fig. 5, A and B). In contrast, anti-L-selectin (Fig.  5, A-D), anti-␣ 1 integrin (Fig. 5A), and anti-CD4 (Fig. 5, C and D) monoclonal antibodies did not have a significant effect on the adherence to the coated surfaces under identical test conditions. In addition, the ability of an RGD-containing peptide to inhibit T cell adherence to the coated wells was tested. As shown in Fig. 6, the RGD-containing peptide, but not the control RGE-containing peptide, inhibited adherence of PMA-acti- T cells adhered to plastic surfaces coated with MBP-Int280. T cell adherence mediated by MBP-Int280 was specifically blocked by an RGD-containing peptide and by anti-␤ 1 , ␣ 4 , and ␣ 5 integrin monoclonal antibodies but not by anti-␣ 1 integrin, anti-L-selectin, and anti-CD4.
Intimin is a homologue of the invasin polypeptide of Yersinia. A unique feature of invasin is its ability to confer a high level invasion phenotype to noninvasive E. coli HB101 (14). An extensive investigation of invasin led to the identification of integrins as its cellular receptors (15). Integrins are ␣/␤ heterodimeric glycoproteins composed of non-covalently associated subunits, which have the ability to transmit signals across the plasma membrane after interaction with a variety of extracellular ligands (16,17). A structural and functional feature that is common to many integrin ligands is a critical Asp residue (32,33). Alteration of the Asp in the binding motif abrogates adhesive activity of many integrin binding ligands. Consistent with this, Asp-911 was recently found to be essential for invasin-integrin interaction (34). A homologous residue within the amino acid sequence of the cell-binding domain of intimin has not yet been identified.
Arg-Gly-Asp (RGD) and Leu-Asp-Val (LDV) are two of the best characterized motifs of extracellular matrix-derived glycoproteins present in the active site of several integrin ligands, such as fibronectin, laminin, and collagen (30,33). However, although invasin and intimin contain neither RGD nor LDV, the binding of both polypeptides to purified ␣ 5 ␤ 1 integrins and the adherence of CD4 ϩ T cell to intimin-and invasin-coated wells can be specifically inhibited by an RGD-containing peptide. It is important to note, however, that despite this, intimin and invasin appear to bind to different sites on the integrin.
It seems that the high affinity of the invasin-integrin inter-action is responsible for internalization of invasin-coated particles, in the form of either live bacteria or inert particles. Other integrin ligands that interact with the receptor with lower affinity, such as fibronectin, do not trigger uptake of the coated particles but rather mediate cell adherence (35). One reason for this might be the fact that unlike fibronectin, invasin-integrin interaction does not rely on terminal oligosaccharide processing of the receptor (36). The outcome of binding of monoclonal antibodies to ␤ 1 integrins is dependent on the affinity of the binding and not on the site of recognition, as interaction with high affinity leads to uptake of coated particles, whereas interaction with low affinity leads to cell adherence (35). Our finding that MBP-Int280 does not effectively compete with MBP-Inv280 using a solid-phase integrin binding assay is in agreement with our previous reports showing no cross-inhibition between invasin and intimin (10,11). The reduced capability of intimin to mediate cell invasion, compared with that of invasin, may be attributed to receptor binding with lower affinity, but this remains to be determined experimentally. EPEC and invasin-mediated phagocytosis are inhibited by cytochalasin D, a drug that causes actin depolymerization within the host cells (12,37). Integrins have been co-localized immunocytochemically with actin-containing microfilaments, and biochemical evidence has been obtained to indicate a direct binding of integrin cytoplasmic domain to certain cytoskeletal proteins, including talin ␣-actinin and paxillin, which are localized to the sites of contact between cells and their substrata (38 -40). Similar cytoskeletal proteins colocalize with EPEC at the bacterial-host cell interface (41). A consistent feature of cell contact sites is their high content of protein tyrosine phosphate, which suggests a prominent role for tyrosine kinases in cytoskeleton assembly (42). Consistent with this, internalization of EPEC and invasin-encoded E. coli is inhibited by several protein kinase inhibitors (4,43).
A large body of evidence now indicates a close link between integrin activation and receptor association with divalent cations (44). The affinity and specificity of integrin-ligand binding is dependent on the type of the divalent cations. In this study, we have found that, like most integrin interactions, binding of both intimin and invasin was optimal in the presence of Mn 2ϩ . The ability of integrins to undergo conformational changes is a key mechanism underlining bidirectional regulation of adhesion and transduction of certain signals by regulating cytoskeletal assembly. Integrin-mediated adhesion of T cells is dependent upon cell activation; resting T cells show only limited adhesive activity (31). Here we have shown that like invasin (29), intimin can mediate equal binding of both activated and resting CD4 ϩ T cells. Site-directed mutagenesis of intimin-and invasin-distal Cys residues (Cys-937 and Cys-982, respectively) abolished cell binding activity. In agreement with this, MBP-Int280CS was unable to inhibit binding of MBP-Int280 in the solid-phase assay. Our results show that adherence of human CD4 ϩ T cells to invasin-and intimin-coated surfaces can be blocked by both anti-␣ 4 and anti-␣ 5 interin antibodies to levels of ϳ85 and ϳ70%, respectively, compared with untreated cells. The reason for the high degree of inhibition of T cell adherence to intimin-and invasin-coated (and FN-coated) wells by both antibodies might reflect the complexity of the preparation obtained from the peripheral blood of several donors. In this study we found that intimin and invasin can bind to immobilized ␣ 4 ␤ 1 and ␣ 5 ␤ 1 integrins. In an independent report, Ennis et al. (29) implicated ␣ 4 ␤ 1 and not ␣ 5 ␤ 1 integrins as T cell receptors for invasin. The reasons for these differences are unclear, although monoclonal antibody specificities and experimental design may be partially responsible. The reason for the inability of MBP-Int280 to induce epithelial cell spreading (10), while the fusion protein is capable of inducing T cell adhesion, is also not yet clear.
What is the apparent benefit for microorganisms interacting with adhesion receptors expressed on immune and nonimmune host cells? The answer to this question is obscure at present. However, the growing number of recent studies demonstrating recognition, binding, and interaction between infectious microorganisms and integrins, probably through mimicry of eukaryotic adhesion epitopes, suggests that such a mechanism could be advantageous for infectious agents in vivo. One may postulate that these modes of interaction may (i) facilitate the motility, colonization, and homing of microorganisms within the host body's compartments; (ii) inhibit proper functioning of immune cells; (iii) facilitate intracellular infection via internalization of the receptor-pathogen complex; and (iv) sequester the invading microorganism from the immune system's surveillance, providing a mechanism of escape and bypass of killing threats inflicted by phagocytic cells (45)(46)(47)(48).
The mechanism by which EPEC causes diarrhea remains equivocal. Normally, EPEC adheres to the mucosal surface of both the small and large intestine and does not reach the lamina propria. However, EPEC shows a limited but reproducible ability to invade intestinal epithelial cells in vitro (12) and has been visualized inside enterocytes in biopsies from infected patients (13). The role of enterocyte invasion in the pathogenesis of EPEC infections remains controversial. RDEC-1, a rabbit E. coli pathogen expressing intimin, attaches specifically to the apical surface of M cells. The preferred cell type infected by EPEC in vivo is still not well documented. It is possible that EPEC intimately attaches to M cells, at least during early stages of infection, and that intimin is essential in this stage of infection. An example of such an interaction can be found in the enteroinvasive bacterium Y. enterocolitica, which specifically attaches to and enters through M cells of the Peyer's patches of the intestine. Invasin plays a central role in this process (49). Nonetheless, the precise relevance of the intimin-integrin interaction described here for gut infections involving intiminexpressing Gram-negative microbial pathogens requires further investigation. In particular, the availability of integrin on the apical surface of enterocytes and the mechanism that might control their distribution needs to be studied. Two studies have investigated the effect of EPEC infection on the integrity of tight junctions in tissue cultured cells and have produced conflicting results (50,51). It remains to be seen whether natural EPEC-like infection alters the integrity of the tight junctions and whether integrins can be colocalized to the sites of intimate bacterial cell association, or if intimins can target integrins on other host cell types in vivo.