Intimin Types α, β, and γ Bind to Nucleolin with Equivalent Affinity but Lower Avidity than to the Translocated Intimin Receptor

The outer membrane adhesins of enteropathogenic Escherichia coli, Citrobacter rodentium, and enterohemorrhagic E. coli (EHEC) O157:H7 that mediate attach and efface intestinal lesions are classified as intimin α, β, and γ, respectively. Each of these intimin types binds to its cognate, bacterially encoded receptor (called Tir for translocated intimin receptor) to promote tight adherence of the organism to the host-cell plasma membrane. We previously reported that γ intimin of EHEC O157:H7 also bound to a eucaryotic receptor that we determined was nucleolin. The objective of this study was to investigate in vitro and in vivo the interactions of intimins α, β, and γ with nucleolin in the presence of Tir from EHEC O157:H7. Protein binding experiments demonstrated that intimin of types α, β, and γ bound nucleolin with similar affinity. Moreover, all three intimin types co-localized with regions of nucleolin expressed on the surface of HEp-2 cells. When intimin α, β, or γ bound to Tir in vitro, the intimin interaction with nucleolin was blocked. Both Tir and nucleolin accumulated beneath intimin-presenting bacteria that had attached to the surface of HEp-2 cells. Taken together, these findings suggest that nucleolin is involved in bacterial adherence promoted by all intimin types and that Tir and nucleolin compete for intimin during adherence.

quence varies among members of the intimin family, a group that includes five to nine members, depending on the particular typing scheme employed (9,10). All such schema place E. coli O157:H7 intimin as type gamma (␥), most human EPEC intimin as type alpha (␣), and C. rodentium intimin as type beta (␤). These sequence differences among the extracellular receptor-binding domains of the intimin types may contribute to both immune system avoidance and tissue tropism of bacterial adherence (11)(12)(13)(14).
All intimin types studied to date bind to an associated translocated intimin receptor, Tir (15)(16)(17)(18), that is produced by the bacteria and inserted into the host-cell membrane through a type III secretion system (19,20). The interaction between intimin and Tir triggers actin condensation beneath the bacterium and permits attachment to the host-cell cytoskeleton (21,22). The genes for intimin (eae) and Tir (tir) are contained in a pathogenicity island called the locus of enterocyte effacement (LEE) (23)(24)(25). All bacteria that display the A/E phenotype possess the LEE genes.
Various intimin types bind to protein receptors expressed by eucaryotic cells (26 -28), and the extracellular domains of different intimin types can bind to tissue culture cells in the absence of Tir (29). Frankel et al. (30) demonstrated that intimin-␣ binds to ␤ 1 integrin, as does invasin, a closely related adhesin produced by some Yersinia species (31). We recently reported that intimin-␥ from E. coli O157:H7 binds to nucleolin that is localized to the plasma membrane of HEp-2 cells (32). Nucleolin is a ubiquitous, conserved protein that is involved in cell proliferation and is produced by all vertebrate species (33). The role in bacterial adherence of the interaction of intimin with these eucaryotic receptors in animals remains unclear. However, we have demonstrated that antiserum raised against nucleolin can inhibit the adherence of enterohemorrhagic E. coli O157:H7 to HEp-2 cells. This result suggests that bacterial interaction with nucleolin may contribute to the pathogenic process initiated by EHEC O157.
The goal of this research was to examine the association of different intimin types with both Tir and nucleolin. Molecular structures obtained by nuclear magnetic resonance and x-ray crystallography have been used to map the interaction of intimin-␣ with Tir (34,35). Tir binds to a C-type lectin motif of intimin that is located at the carboxyl terminus of the extracellular domain. The intimin binding site for nucleolin is unknown. Reece et al. (13) reported that mutations in intimin-␣ engineered by site-directed mutagenesis of the eae locus in the region encoding the C-type lectin domain influenced EPEC colonization but had no affect on Tir binding. This finding suggests that the eucaryotic receptor-binding site also resides in the lectin domain (36). To better understand the role of each type of receptor in bacterial adherence, we used in vitro methods as well as a tissue culture model to investigate the inter-actions of three intimin types (␣, ␤, and ␥) with human nucleolin and Tir from EHEC O157:H7.

Bacterial Strains and Plasmids
Bacterial strains and protein constructs used in this study are listed in Table I. EPEC O127:H6 strain E2348/69 that expresses intimin-␣ and plasmid pCVD438 that contains the eae gene from E2348/69 were provided by Dr. Ann Jerse. Construction of pCVD438 has been previously described (1). Citrobacter rodentium that expresses intimin-␤ was supplied by Dr. James Kaper. EHEC O157:H7 strain 86-24, which expresses intimin-␥, was contributed by Dr. Phillip Tarr. The plasmid pEB310 that contains the eae gene from EHEC O157:H7 was previously described by McKee et al. (8). E. coli strains BL21(DE3) (Novagen) and XL1-Blue (Stratagene) were used for the expression of recombinant proteins.
The eae genes from EPEC, C. rodentium, and EHEC O157 were each cloned into high copy plasmid vectors with eae under control of the T7 promoter. Plasmid pCDV438 (intimin-␣) was digested with the restriction enzymes NarI and HindIII to yield a DNA fragment that contained the entire eae gene with some additional flanking sequence. This fragment was ligated into pBluescriptKSϪ (Stratagene) cut with the restriction enzymes AccI and HindIII to produce plasmid pInta939. The intimin-␤ eae gene was amplified directly from C. rodentium genomic DNA by PCR with primers derived from the published gene sequence. These primers, ATTCCCGACAGATAATCCTAAC and TGCCGGGTT-TATATGATTAGAC, annealed 338 bases upstream of the eae sequence and to the 3Ј end of the gene, respectively. The PCR product was ligated into the vector pCR2.1 (Invitrogen TA cloning kit) to give plasmid pIntb936. Plasmid pEB310 (intimin-␥) was digested with the restriction enzymes BamHI and HindIII, and the eae gene was ligated into vector pBluescriptKSϪ cut with the same enzymes to yield plasmid pIntg934.
For each intimin type, the portion of the eae gene that encodes the intimin extracellular domain was cloned into the expression vector pQE32 (Qiagen) so as to place the truncated eae gene immediately downstream from a sequence that encodes a 6-histidine tag and under the control of the lac promoter. Plasmid pCVD438 was digested with SalI to give a DNA fragment that contained the 3Ј 1157 nucleotides of the intimin-␣ eae gene. This fragment was ligated into SalI-digested pQE32 to give plasmid p6his-Inta385. Plasmid pIntb936 was cut with the restriction enzymes SalI and HindIII to produce a DNA fragment that contained the 3Ј 1157 nucleotides of the intimin-␤ eae gene. This fragment was ligated into vector pQE32 cut with the same enzymes to produce plasmid p6his-Intb385. Plasmid pEB310 was digested with SalI and HindIII to produce a DNA fragment with the 3Ј 1143 nucleotide segment of the intimin-␥ eae gene. This fragment was ligated into pQE32 that was then cut with the same enzymes to produce plasmid p6his-Intg380.
The construction of pTir, a plasmid that contains the EHEC O157:H7 tir gene under control of the T7 promoter, was previously described (32). A plasmid that expresses recombinant nucleolin was constructed as follows. A cDNA clone (IMAGE 591039) of the ncl gene from a human pancreas adenocarcinoma was obtained from the American Type Culture Collection (ATCC catalog number 749447). The entire ncl gene with some flanking sequence was amplified by PCR with the DNA primer pair TCGGATCCACTTGTCCGCTTCACACTC and AATCTG-CAGTCAAAACCAACACG that incorporated restriction sites for BamHI and PstI, respectively. After cutting the amplified product with these enzymes, the resultant DNA fragment was ligated into the 6-histidine tag expression vector pQE32 (Qiagen) in an attempt to produce a recombinant histidine-tagged version of the nucleolin protein. This construct was transformed into E. coli strain BL21 (Novagen), but no expression of the recombinant human protein was observed. Therefore, the ncl gene, along with the sequence that encoded the amino-terminal 6-histidine tag, was cut from the pQE32 vector with the restriction enzymes XhoI and NheI. This linear fragment was cloned into the yeast epitope tagging vector pESC-Trp (Stratagene) that had been restricted with the enzymes SalI and NheI. This construct, designated p6his-ncl, contained the tagged nucleolin gene behind a galactose-inducible promoter as well as the trp1 gene for tryptophan biosynthesis. The plasmid was electroporated into Saccharomyces cerevisiae strain YPH500 (Stratagene) that is auxotropic for several amino acids, including tryptophan. Yeast were selected and maintained on synthetic dextrose minimal medium without tryptophan.

Protein Expression and Purification
Tir-All salts and chemicals used for protein purification were of reagent grade and were purchased from Sigma-Aldrich unless otherwise noted. Plasmid pTir was transformed into the E. coli strain BL21, and expression of the recombinant Tir protein was induced from mid-log cultures with 2 mM isopropyl-␤-D-thiogalactopyranoside. Induced cells were lysed by sonication, and the cell lysate was subjected to a 20% ammonium sulfate precipitation. The lysate supernatant was extensively dialyzed against phosphate-buffered saline (PBS). The dialyzed protein was centrifuged at 10,000 rpm to remove insoluble material. Tir protein in the supernatant was purified from contaminating proteins by passage over a column of DEAE-Sepharose (Amersham Biosciences) and eluted with a linear sodium chloride gradient. Column fractions were assessed by Western blot with rabbit anti-Tir serum (32), and fractions with the greatest reactivity were pooled. We refer to this recombinant Tir protein as Tir O157 in the text.
Nucleolin-Production of the recombinant nucleolin protein was induced by growth of the S. cerevisiae strain YPH500 transformed with p6his-ncl at 30°C for 36 h in yeast synthetic minimal medium with 2% galactose and without tryptophan (Bio 101, Inc., Vista, CA). Western blot analysis of yeast whole cell lysate showed that the induced culture produced a protein with an apparent size of 114 kDa that was recognized by rabbit anti-nucleolin antiserum (Santa Cruz Biotechnology) as well as mouse anti-his tag antibody (Qiagen). We refer to this recombinant 6-histidine-tagged human nucleolin protein as 6H-nucleolin in the text. Total cellular protein was extracted from induced cells with a pH 8.0 denaturation buffer (100 mM sodium phosphate, 10 mM Tris, 6 M guanidinium hydrochloride, and 0.2% Tween 20). The 6H-nucleolin was purified from the cell lysate by nickel-affinity column chromatography (Qiagen). Guanidinium-extracted proteins were loaded onto the nickel affinity column, and contaminants were washed from the resin with 20 column volumes of pH 7 buffer (100 mM sodium phosphate, 10 mM Tris, 0.2% Tween 20). Bound protein was eluted from the column with pH 4.5 buffer (10 mM sodium phosphate, 100 mM sodium chloride, 0.2% Tween 20).

Histidine-tagged Intimin Extracellular Domains
E. coli XL1-Blue-competent cells were transformed with plasmids p6his-Inta385, p6his-Intb385, or p6his-Intg380. Transformed cells were selected for ampicillin resistance. Mid-log phase LB cultures of the transformed cells grown at 37°C were induced with 2 mM isopropyl-␤-D-thiogalactopyranoside. These cells produced truncated intimin with an amino-terminal 6-histidine tag and the carboxyl-terminal 385 amino acids of intimin-␣ or intimin-␤, or 380 amino acids of intimin-␥. We refer to these molecules in the text as intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 , respectively. Purification of the His-tagged intimins by nickelaffinity chromatography was accomplished as described above for 6H-nucleolin.

Protein Binding Experiments
Purified intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 (1 mg/ml) were biotin-labeled by reaction with 2 g of biotinamidocaproate N-hydroxysuccinimide ester (Sigma) in 0.1 M sodium borate buffer, pH 9.0, for 4 h at room temperature. Unreacted ester was quenched by the addition of 5 mM ammonium chloride, and the samples were dialyzed against PBS. Polyvinylchloride microtiter plates (96-well format) were coated for 3 h at room temperature with 50 l of a 20 g/ml solution of either Tir O157 or 6H-nucleolin. Nonspecific binding sites in the wells were blocked with 3% bovine serum albumin (Sigma) in PBS incubated in the wells overnight at 4°C. A 2-fold dilution series was made for each biotinylated intimin sample. These dilution series were incubated with the immobilized receptor proteins in a binding buffer that contained PBS with 3% bovine serum albumin and 0.2% Tween 20. After 1-h incubation at room temperature, intimin that had not bound to the immobilized receptors was rinsed from the wells with three washes of PBS. The amount of intimin bound nonspecifically to the wells was estimated from 2-fold dilutions added to wells that contained only bovine serum albumin. Bound intimin was incubated with streptavidin-conjugated horseradish peroxidase (Amersham Biosciences) diluted 1:1000 in binding buffer. The intimin-peroxidase complexes were detected by the enzyme-catalyzed color change of 3,3Ј,5,5Ј-tetramethylbenzidine (TMB) using the TMB Peroxidase EIA Substrate Kit (Bio-Rad). The bound intimin concentration was estimated from the change in absorbance at 450 nm of the solution in each well, measured with an Elx800 Universal Microplate reader (Bio-Tek Instruments). These values were corrected for nonspecific binding by subtraction of the absorbance reading in wells that contained bovine serum albumin only. In a separate experiment designed to test the affect of Tir on the intimin interaction with nucleolin, various concentrations of Tir O157 were mixed with an intimin-␥ 380 dilution series in binding buffer. The intimin-Tir mixtures were incubated with nucleolin that had been immobilized on a microtiter plate. After incubation for 1 h at room temperature, intimin-␥ 380 that had bound to 6H-nucleolin, was detected as described.

HEp-2 Cell Binding Assay
HEp-2 (ATCC CCL23) human laryngeal epithelial cell cultures were maintained in EMEM (BioWhittaker) supplemented with 10% fetal calf serum, 20 mM L-glutamine, 100 g/ml gentamicin, 10 units/ml penicillin G, and 10 g/ml streptomycin. Confluent cells grown at 37°C in an atmosphere of 5% CO 2 were released from the culture flasks by treatment with trypsin, and then 0.3 ml of cell suspension was seeded into 8-well chamber slides (Lab-Tek) at a cell density of 6 ϫ 10 4 cells/ml. The chamber slides were then incubated for 24 h at 37°C in an atmosphere of 5% CO 2 . Protein binding buffer consisted of RPMI 1640 (BioWhittaker) medium with 200 mM HEPES buffer and 0.5% bovine serum albumin. Purified samples of histidine-tagged intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 were diluted into 200 l of protein binding buffer to a final concentration of 10 g/ml. HEp-2 cell monolayers were incubated with the protein samples for 1 h at 37°C, and then unbound intimin was rinsed from the cells by three washes with PBS. Bound intimin was immunolabeled with Penta-His Alexa Fluor 555 Conjugate (Qiagen) monoclonal antibody directed against the 6-histidine tag. The HEp-2 cell monolayer was fixed with PBS-buffered 3% formaldehyde. Nucleolin on the surface of the fixed cells was detected with primary rabbit anti-nucleolin serum (Santa Cruz Biotechnology) followed with secondary goat anti-rabbit-IgG serum conjugated to Alexa Fluor 488 (Molecular Probes), both of which were diluted in PBS that contained 3% bovine serum albumin. In a separate experiment, cell monolayers were infected with EHEC strain 86 -24 eae⌬10 to insert the bacterially encoded Tir receptor into the HEp-2 cell plasma membrane. Samples of the purified histidine-tagged intimin (0.5 g/ml) in protein binding buffer were incubated with these infected cells. Cell surface-bound intimin was immunolabeled as described. Nucleolin was stained by the addition of primary rabbit antiserum labeled with Zenon Alexa Fluor 488 (Molecular Probes). Primary rabbit antiserum raised against Tir was covalently labeled with Alexa Fluor 350 using a Molecular Probes protein labeling kit following the manufacturer's directions. Infected and uninfected cells were stained for Tir by the addition of the labeled anti-Tir serum. Both labeled anti-nucleolin and anti-Tir sera were diluted in PBS that contained 3% bovine serum albumin and were precleared of antibodies against bacterial proteins by reaction with formaldehyde-fixed preparations of E. coli strain BL21.

Bacterial Adherence Assay
HEp-2 cell monolayers were grown in 8-well chamber slides as described for the cell binding assay. EHEC O157:H7 strain 86 -24, strain 86 -24eae⌬10, and EPEC strain E2348/69 were grown overnight at 37°C in static cultures of EMEM. Cultures of E. coli strain BL21(DE3) that expressed intimin-␣ 939 , intimin-␤ 936 , or intimin-␥ 934 were grown overnight with agitation in LB supplemented with 0.1 mg/ml ampicillin to maintain selection for the intimin plasmids. HEp-2 cells were infected with 8 l of overnight cultures inoculated into 300 l of EMEM and incubated at 37°C. After a 4-h infection period, the liquid on the cells was gently aspirated with a pipette to minimize the disturbance of bacteria that had adhered to the cell surface. The infected cells were fixed for 20 min in 2% formaldehyde, washed with PBS, permeabilized with 0.1% Tween 20 for 4 min, and then washed again with PBS. Nucleolin and Tir were stained by primary antiserum labeled with Alexa Fluor 488 and Alexa Fluor 350, respectively, as described. Polymerized actin was visualized with Alexa Fluor 594 phalloidin (Molecular Probes). All immunostaining procedures were carried out in PBS that contained 3% bovine serum albumin.

Deconvolution Microscopy
All images were obtained with an Olympus BX60 system microscope with a BX-FLA reflected light fluorescence attachment. Images were captured with a SPOT RT charge-coupled device digital camera (Diagnostic Instrument, Inc.) in 8-bit gray-scale format. For each microscopic field shown, four images were obtained with red, green, and blue filters through sequential focal planes separated by 1-m increments. These images were processed into stacks by the public domain program Im-ageJ (developed at the National Institutes of Health and available on the Internet at the web address rsb.info.nih.gov/ij/). Out of focus light was removed from the stacks through the process of deconvolution. The point-spread function for deconvolution was modeled by the radial symmetric equation, A*exp(Ϫr 2 /2*k 1 )*cos(2r/) 2 ϩ B*exp(Ϫr 2 / 2*k 2 )*sin(2r/) 2 , where the variables A and B represent the contributions of the cos and sin functions, k 1 and k 2 represent the decay of the functions, and and represent the period of the functions at distance r from the center. The ImageJ module DeconvolutionJ, written by N. Linnenbrü gger, accomplished deconvolution of image stacks. Deconvo-lutionJ uses a regularized Wiener filter to deconvolve in-focus light from low frequency out-of-focus light. One or two image stack sections were selected for presentation, with the maximal intensity from each slice added to the final z-section. For each image, a background value was subtracted, and the signal intensity was normalized between the values of 0 and 255. Images were colored red, green, or blue, and overlaid and annotated in the final figures with the program Adobe Photoshop.

Colocalization Correlation Determination
To evaluate the colocalization between intimin and nucleolin on the surface of HEp-2 cells a cross-correlation analysis was performed. Intimin and nucleolin were immunostained with the fluorescent dyes Alexa Fluor-555 and Alexa Fluor-480, respectively. Images obtained through the red and green filter sets were deconvolved and normalized as described above. Numerical values of the pixel intensities for the separate images were converted to text strings by the program ImageJ. The intensity of staining was compared at each X-Y coordinate pair of the separate images. For the intimin-stained images, pixels with intensity values greater than 50 were considered valid for inclusion in the correlation analysis. The Pearson correlation coefficient, r ϭ n(⌺XY) Ϫ (⌺X)(⌺Y)/((n⌺X 2 Ϫ (⌺X) 2 )(n⌺Y 2 Ϫ (⌺Y) 2 ) 1 ⁄2, was evaluated by the program Microsoft Excel from valid X-Y intensity pairs. The significance of the correlation value was determined from a two-tailed paired t test.

Intimin Types ␣, ␤, and ␥ Bound to Nucleolin with Similar
Affinity-Previously we demonstrated that intimin-␥ of EHEC O157:H7 binds to nucleolin isolated from HEp-2 cells (32). One purpose of the present study was to determine if other intimin types interact with nucleolin. To answer this question, recombinant versions of nucleolin and various intimin extracellular domains were purified and used in an in vitro binding assay. Recombinant 6H-nucleolin was absorbed onto the wells of a microtiter plate. To measure the affinity of intimin for nucleolin, a series of 2-fold dilutions of biotinylated intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 were incubated with the immo-

Various Intimins Bind to Nucleolin and Tir
bilized 6H-nucleolin. The amount of intimin that bound in each well was estimated by a change in solution color catalyzed by streptavidin-peroxidase used to detect the biotin label. The results of this experiment are shown in Fig. 1. The intimin types tested (␣, ␤, and ␥) all bound to human nucleolin. A Scatchard analysis of each titration was done to estimate the equilibrium dissociation constant (K d ) for the interaction of intimin with nucleolin. The average K d values from three separate determinations are presented in Table II. These results show that the affinity for nucleolin of all three intimin types was on the order of 100 nm. Marginal amounts of recombinant intimin bound nonspecifically to wells coated only with bovine serum albumin (Fig. 1). The binding data presented in Fig. 1 have been corrected for the nonspecific background values. Binding of biotinylated intimin to immobilized 6H-nucleolin was almost completely blocked by the addition of 10-fold excess of unlabeled recombinant intimin (data not shown), a finding that demonstrates the specificity of the binding reaction.
Intimin Types ␣, ␤, and ␥ Bound to Tir O157 with Similar Affinity-Several reports have shown that different intimin types are able to bind the same Tir molecule, i.e. that the intimin-binding domain of Tir is functionally interchangeable between organisms (17,37,38). Because the interaction between intimin and Tir has been well documented, we used the association between the different intimin types and Tir O157 as a positive control for the protein binding titration. Purified Tir O157 was absorbed onto the wells of a polyvinylchloride microtiter plate. Serial 2-fold dilutions of intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 were incubated with immobilized Tir O157 . Biotin-labeled intimin that had bound in the wells was detected as described. The results of this experiment, corrected for nonspecific background binding, are presented in Fig. 1. Scatchard analysis of the titrations gave equilibrium dissociations from the binding data. The average K d value from three separate determinations is shown in Table II. This experiment confirms that the three different intimin types had comparable equilibrium binding constants for Tir O157 .

Tir Blocks the Association between Intimin and Nucleolin-
One objective of this research was to observe the simultaneous interactions between intimin, Tir, and nucleolin. If intimin has separate, distinct binding sites for Tir and nucleolin, than both molecules should bind concurrently to intimin. On the other hand, if the binding sites for the receptors are overlapping or in close proximity, then steric hindrance would prevent both molecules from binding to intimin simultaneously. To test these possibilities we examined the competition between Tir O157 and 6H-nucleolin for a limited number of intimin binding sites. The recombinant protein 6H-nucleolin was absorbed onto the wells of a microtiter plate. Intimin-␥ 380 was incubated with immobilized nucleolin in a solution containing various concentrations of Tir O157 . After equilibrium had been established, intimin not bound to nucleolin was washed from the wells, and bound intimin was detected as described. The results of this experiment are shown in Fig. 2A. With increasing concentrations of Tir O157 , the amount of intimin-␥ 380 bound to 6H-nucleolin was diminished. There was no association between Tir O157 and immobilized 6H-nucleolin as measured by anti-Tir enzymelinked immunoassay (data not shown). This observation ruled out the possibility that intimin adherence was blocked by Tir O157 that had bound nucleolin. The fraction of intimin-␥ 380 that was bound to 6H-nucleolin as a function of Tir O157 concentration is presented in Fig. 2B. These data were consistent with a simple competition between the receptors in which one molecule of Tir was able to block one molecule of intimin from binding to nucleolin. Tir O157 was also able to block the interaction between 6H-nucleolin and intimin-␣ 385 and intimin-␤ 385 (data not shown).
Intimin of Types ␣, ␤, and ␥ Associated with Endogenous Nucleolin and Tir on the Surface of HEp-2 Cells-In an earlier report, we showed by immunofluorescence staining that recombinant intimin-␥ of EHEC O157 binds to the surface of HEp-2 cells in regions that contain nucleolin (32). In the present study, we investigated the association of the three different intimin types with native nucleolin in the presence or absence of translocated Tir. Intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 were added to the supernatant of HEp-2 cell monolayers. Intimin bound to the cell surface was detected with a labeled antibody directed against the amino-terminal histidine tag. Native nucleolin expressed by the cells and localized to the plasma membrane was detected with the appropriate primary and secondary anti-sera. Fig. 3A shows that the staining pattern for each intimin type (stained red) demonstrated overlap with the staining pattern for cell surface-localized nucleolin (stained green) as revealed by the orange/yellow color in the merged images. A typical cross-correlation plot of the staining intensity for nucleolin compared with that of intimin is presented in Fig. 4A. Values of the Pearson product moment correlation coefficient were determined to be 0.38, 0.42, and 0.40, all with p Ͻ 0.0005, for intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 , respectively. These results demonstrate that there was a statistically significant co-localization between intimin and The solid line through each data set is a theoretical binding curve calculated from the equilibrium constants listed in Table II. Each data point represents the average of three repeats. The error bars that bracket each symbol approximate the standard error of the mean of three replicate determinations. Each data set was corrected for nonspecific background intimin binding. Representative background measurements are shown for intimin bound in wells coated with bovine serum albumin (x and asterisk).

Various Intimins Bind to Nucleolin and Tir
nucleolin. The intimin deletion mutant EHEC O157:H7 86 -24eae⌬10 was used to insert Tir into the HEp-2 cell plasma membrane. Because this mutant bacterium did not express a functional intimin, it was unable to adhere to the infected cells by means of Tir. HEp-2 cells with both nucleolin and Tir on the surface showed a distinctly different pattern for binding of recombinant intimin. Fig. 3B illustrates that in cells with Tir in the plasma membrane, there was an increase in the amount of intimin-␣ 385 bound to the cell surface, and there was not extensive overlap between the intimin-and nucleolin-staining patterns. Tir (stained blue) localization did overlap with regions of increased intimin binding as denoted by the purple color in the merged image. When Tir was present, the association between intimin and nucleolin decreased, and there were clearly regions of intimin staining without nucleolin. Fig. 4B shows a typical cross-correlation plot for the intimin and nucleolin staining intensity on cells with Tir inserted into the plasma membrane. There was no significant correlation between intimin and nucleolin staining intensity on these cells, with the Pearson correlation coefficients calculated to be 0.06, 0.06, and 0.04 (all with p Ͼ 0.25) for intimin-␣ 385 , intimin-␤ 385 , and intimin-␥ 380 , respectively.
Nucleolin and Polymerized Actin Segregate beneath Intiminexpressing Bacteria-HEp-2 cell monolayers were co-infected with EHEC O157:H7 strain 86 -24eae⌬10 (to insert Tir) and E. coli K12 strains that produced intimin-␣ 939 , intimin-␤ 936 , or intimin-␥ 934 . These K12 transformants were used to analyze bacterial adherence promoted by the various intimin types without the confounding influences of other adherence factors produced by the different wild-type pathogenic strains. Following a 4-h infection period with the intimin-expressing E. coli K12 strains, the HEp2 cells were fluorescently stained for Tir, nucleolin, and actin. Photomicrographs of these stained cells are presented in Fig. 5. Although most adherent bacteria stained for Tir (blue) and actin (red), a subset of bacteria also displayed intense nucleolin immunostaining (green) at the site of adherence. Little to no nucleolin stain was observed in re-

FIG. 2. Binding competition for the interaction of intimin-␥ 380
with Tir O157 and 6H-nucleolin. A, 6H-nucleolin was immobilized in the wells of a polyvinylchloride microtiter plate. Intimin-␥ 380 titrations (x-axis) were performed in the presence of Tir O157 at concentrations of 1 M (ϩ), 500 nM (triangle), 250 nM (Ϫ), 125 nM (diamond), 62 nM (square), 16 nM (X), and without Tir O157 (circle). The fraction of immobilized 6H-nucleolin with intimin bound (y-axis) was measured as described under "Experimental Procedures." Tir O157 in solution with the recombinant intimin blocked binding to 6H-nucleolin in a concentrationdependant manner. The data in A were corrected for nonspecific binding of intimin-␥ 380 to control wells coated with bovine serum albumin. The solid line through each data set is calculated from a simple twostate binding equilibrium. B, the fraction of immobilized nucleolin with intimin-␥ 380 bound (y-axis) is shown for one intimin-␥ 380 concentration of 250 nM and various Tir O157 concentrations (x-axis). The solid line is a theoretical fit to the data for a competitive inhibition models for two ligands with equilibrium dissociation constants of 20 and 100 nM. The error bars approximate the average standard error of the mean between replicate determinations.

FIG. 3. HEp-2 cell binding assay for various types or recombinant intimin.
A, solutions of intimin-␣ 385 (top), intimin-␤ 385 (middle), or intimin-␥ 380 (bottom) were incubated with sub-confluent HEp-2 cell monolayers. Intimin (red) and nucleolin (green) were immunostained with the appropriate primary and secondary antiserum as described under "Experimental Procedures." Merged fluorescence micrographs (right column) of the two stains demonstrate a strong correlation between the location of bound intimin and nucleolin for each intimin type. B, subconfluent HEP-2 cell monolayers were infected with intimin deletion strain EHEC 86 -24eae⌬10 to insert Tir into the HEp-2 plasma membrane. Intimin-␣ 385 (red), nucleolin (green), and Tir (blue) were stained with labeled primary immunoserum as described. A merged fluorescence micrograph of the three stains demonstrates that in preinfected cells the recombinant intimin associates with both Tir and nucleolin while in uninfected cells intimin only associates with regions of nucleolin expression. All micrographs were obtained at 40ϫ original magnification.
gions with actin polymerized beneath the adherent bacteria. These observations suggest that nucleolin was excluded from the regions of Tir-mediated intimate adherence promoted by all intimin types. To observe the association of nucleolin with wild-type bacteria that express intimin, HEp-2 cells were infected with EPEC O126:H6 or EHEC O157:H7. The cells were stained for Tir, nucleolin, and actin following bacterial infection (Fig. 5). Actin and Tir were evident beneath the wild-type bacteria, but nucleolin appeared to be excluded from those regions directly under the bacteria where polymerization of actin was evident. We surmise that such actin engagement indicates tight bacterial adherence. That nucleolin clustered around the sites of bacterial association with the HEp-2 cells suggests that the intimin-expressing bacteria bound nucleolin at the initial contact of the organism with the host cell but that these intimin-nucleolin interactions were displaced by the Tirdriven formation of the actin-rich pedestal. DISCUSSION In this investigation, we attempted to analyze the interactions between intimin and two receptors, bacterially encoded Tir and host cell-expressed nucleolin. We found that intimin of types ␣, ␤, and ␥ each bound to nucleolin with similar dissociation constants that were on the order of 100 nM. We previously reported (32) that purified intimin-␥ binds to the surface of HEp-2 cells with a dissociation constant of ϳ90 nM, a value HEp-2 cell monolayers were co-infected with the intimin deletion strain EHEC 86 -24eae⌬10 and with E. coli K12 strains that expressed full-length intimin-␣ 939 , intimin-␤ 936 , or intimin-␥ 934 . After a 4-h infection the bacteria and cells were fixed with buffered 2% formaldehyde and then permeabilized with 0.1% Tween 20. Tir and nucleolin were stained with labeled primary immunoserum as described under "Experimental Procedures." Polymerized actin was stained with labeled phalloidin. In addition HEp-2 monolayers were infected with wild-type EPEC O126:H6 or EHEC O157:H7 for 4 h, and the cells were stained in the same manner as the co-infection. Fluorescence photomicrographs (original magnification of ϫ100) of the staining patterns for Tir (blue), nucleolin (green), and actin (red) are presented for the bacteria shown in the phase contrast images in the column to the right. Images in the top three rows are bacteria that expressed intimin types ␣, ␤, and ␥. Micrographs of cells infected with the pathogenic bacteria are presented in the bottom two rows. Merged pictures of the three-color staining revealed that regions of nucleolin accumulation were distinct from regions of Tirinduced actin polymerization beneath the adherent bacteria.
close to that which we obtained for the interaction with purified nucleolin. Frankel et al. (29) first demonstrated that the extracellular domain of these different intimin types bound in a punctate distribution to HEp-2 cells. Our data extend this observation to show that the different intimin types each bind to cells in regions that contained nucleolin. The binding data and immunofluorescence experiments support the hypothesis that nucleolin is the primary intimin receptor on uninfected tissue culture cells. Differences in sequence between the various intimin types were previously shown to influence the site of bacterial adherence in the host intestine (13,14,38,39). Our finding that the three types of intimin had similar affinity for nucleolin suggests that this interaction is not involved in intimin-mediated tissue tropism. The association between intimin and nucleolin may represent a common host-cell adherence mechanism exhibited by all A/E lesion-forming pathogens.
The interaction of intimin-␥ with Tir produced by EHEC O157:H7 was reported by DeVinney et al. (17) to have a dissociation constant of ϳ10 nM. We measured intimin-Tir dissociation constants of ϳ20 nM. These values are relatively close, and the discrepancy between them may be attributed to different experimental methods. From these data, we can conclude that the affinity of intimin for Tir is 5 to 10 times stronger than the affinity for nucleolin. When Tir and nucleolin were presented simultaneously, we found that Tir blocks intimin from adhering to nucleolin. This observation indicates that either nucleolin and Tir have overlapping binding sites on the C-type lectin domain of intimin, or that Tir binding induces a conformational change in intimin that precludes nucleolin association. There is no apparent sequence homology between Tir and human nucleolin, so we feel it is unlikely that nucleolin acts as a Tir mimic. Although we do not understand the exact mechanism by which Tir blocks nucleolin binding, the affinity results would suggest that Tir is the preferred receptor for intimin. Nevertheless, when endogenous nucleolin and Tir were presented simultaneously in the milieu of the plasma membrane, we found that recombinant intimin bound to both receptors.
For A/E lesion-forming bacteria, intimin-mediated adherence occurs in conjunction with many other adherence factors (40 -42). To focus on potential differences in adherence caused by the different intimin types, the eae genes were cloned from the wild-type bacteria and expressed in an E. coli K12 background. Previous work showed that K12 strains that express intimin, or latex beads coated with intimin can induce pedestal formation when Tir is inserted into the plasma membrane by an intimin-deletion strain (15,37). Immunofluorescence staining showed nucleolin accumulation beneath adherent bacteria only in regions where there was no Tir-induced actin polymerization. This observation supports the idea that intimin on the bacterial surface can bind to the bacterial receptor or the eucaryotic receptor but can not adhere simultaneously through both. When HEp-2 cells were infected by EPEC or EHEC, only Tir and polymerized actin staining were observed directly beneath the adherent bacteria. Nucleolin staining was observed in proximity to the site of attachment, but significant accumulation did not appear directly beneath the bacteria. The difference in nucleolin staining observed between the E. coli K12 strains that express intimin and the wild-type A/E pathogens may be due to the fact that the K12 strains express significantly more intimin than the pathogens (verified by Western blot analysis).
The data that we have presented suggest that intimin-mediated bacterial adherence involves competition between Tir and nucleolin. How does this competition relate to the process of infection by these pathogenic microorganisms? Our current understanding of the infection process is that the host ingests the bacteria, which then pass through the stomach and into the large intestine. Once in the intestine these bacteria move through the mucus layer and come into contact with the luminal mucosal epithelium were Tir is inserted into host-cell membranes. The role of Tir in linking the bacterium to the host-cell cytoskeleton is well established. The observation that Tir interaction with intimin appeared to dominate the later stages of adherence to tissue culture cells implies that any interactions between intimin and nucleolin must occur prior to the insertion of Tir and the formation of actin-rich pedestals. The interaction between intimin and nucleolin is not sufficient to promote bacterial attachment, and for that reason it is unlikely that nucleolin is acting solely as a passive receptor for intimin. One characteristic of C. rodentium infection in mice is the induction of colonic hyperplasia. Higgins et al. (43) have shown that intimin alone is able to produce a similar hyperplasia in mice and have proposed that intimin interactions with some hostcell receptor produce this response. The interaction between intimin and nucleolin may play role in the pro-inflammatory response induced in host cells during infection by the A/E lesion-forming pathogens. Adherence of these bacteria has been reported to induce the mitogen-activated protein (MAP) kinase cascade, leading to the production of pro-inflammatory cytokines (44 -47). Activation of the MAP-kinases has been reported to be both intimin-dependent and -independent (44,48). Nucleolin is one target of several MAP-kinases, including p38 (49), extracellular-regulated kinase (50), and casein kinase 2 (51,52). Of particular note is the recent work of Savkovic et al. (53) who showed that protein kinase C-is involved in EPEC-induced inflammatory responses of infected cells. Nucleolin is a protein kinase C-substrate and has been shown to be involved in cell-surface signaling (54). One possible result of the competition between nucleolin and Tir is that the initial contacts between intimin and the host cell may trigger the extracellular signaling cascade that promotes bacterial dissemination while the later contacts with Tir promote tight adherence. Although we do not yet fully understand the role of nucleolin in intimin-mediated bacterial adherence, these data in aggregate support a model in which intimin interactions with nucleolin are distinct from the interactions with Tir that occur during adherence.