Specificity of the High-Mannose Recognition Site between Enterobacter cloacae Pili Adhesin and HT-29 Cell Membranes

Transcription

1 INFECTION AND IMMUNITY, Oct. 1997, p Vol. 65, No /97/$ Copyright 1997, American Society for Microbiology Specificity of the High-Mannose Recognition Site between Enterobacter cloacae Pili Adhesin and HT-29 Cell Membranes Y. T. PAN, 1 BIN XU, 1 KEVIN RICE, 2 SAM SMITH, 3 RICHARD JACKSON, 3 AND ALAN D. ELBEIN 1 * Department of Biochemistry and Molecular Biology 1 and Department of Surgery, 3 University of Arkansas for Medical Sciences, Little Rock, Arkansas 72205, and Department of Medicinal Chemistry and Pharmaceutics, College of Pharmacy, The University of Michigan, Ann Arbor, Michigan Received 10 April 1997/Returned for modification 23 May 1997/Accepted 7 July 1997 Enterobacter cloacae has been implicated as one of the causative agents in neonatal infection and causes a septicemia thought to be initiated via the gastrointestinal tract. The adhesion of radiolabeled E. cloacae to HT-29 cells was concentration and temperature dependent and was effectively blocked by unlabeled bacteria or by millimolar concentrations of -mannosides and micromolar concentrations of high-mannose oligosaccharides. A variety of well-characterized mannose oligosaccharides were tested as inhibitors of adhesion. The best inhibitor was the Man 9 (GlcNAc) 2 -tyrosinamide, which was considerably better than other tyrosinamide-linked oligosaccharides such as Man 7 (GlcNAc) 2, Man 6 (GlcNAc) 2 or Man 5 (GlcNAc) 2. Further evidence that the bacteria preferred Man 9 (GlcNAc) 2 structures was obtained by growing HT-29 cells in the presence of glycoprotein processing inhibitors that block mannosidase I and increase the amount of protein-bound Man 9 (GlcNAc) 2 at the cell surface. Such cells bound 1.5- to 2-fold more bacteria than did control cells. The adhesin involved in binding to high-mannose structures was purified from isolated pili. On sodium dodecyl sulfate-gels, a 35-kDa protein was identified by its specific binding to a mannose-containing biotinylated albumin. The amino acid sequences of several peptides from the 35-kDa subunit showed over 85% identity to FimH, the mannose-specific adhesin of Salmonella typhimurium. Pili were labeled with 125 I and examined for the ability to bind to HT-29 cells. Binding showed saturation kinetics and was inhibited by the addition of Man 9 (GlcNAc) 2 -tyrosinamide but not by oligosaccharides with fewer mannose residues. Polyclonal antibody against this 35-kDa protein also effectively blocked adhesion of pili or E. cloacae, but no effect was observed with nonspecific antibody. These studies demonstrate that the 35-kDa pilus subunit is a lectin whose specificity is directed toward Man 9 (GlcNAc) 2 oligosaccharides. Downloaded from Enterobacter species are a major problem to clinicians because of their resistance to multiple drugs, including the new cephalosporins and penicillins (19), and since 1993, the incidence of these organisms in intensive care units has increased. Enterobacter cloacae is the species most frequently isolated from clinical specimens, followed by Enterobacter aerogenes and Enterobacter agglomerans (1, 12, 14, 21). This alarming increase in multiple-antibiotic-resistant organisms suggests that new chemotherapeutic strategies, such as inhibition of bacterial adhesion, may be a feasible approach to control colonization or infection in infants and patients at high risk. Bacterial attachment to mucosal surfaces is the first step in colonization and pathogenesis (2). Thus, an understanding of the mechanism involved in recognition of the host cell by the bacteria, and the chemical nature of the structures involved in attachment, can lead to new methods for preventing bacterial infections. If the attachment of the bacteria to the host cell surface involves a receptor-mediated process, it should be inhibited by substances that block this interaction. Such substances should include compounds that mimic the receptor binding site, as well as antibodies against the receptor binding site (5, 26). In addition, compounds that mimic the structure of the ligand being recognized by the receptor should also prevent this interaction. A number of bacteria bind to host cell surfaces via a proteincarbohydrate interaction (36). Usually in these cases, the carbohydrate molecule exists as a glycolipid or glycoprotein in the * Corresponding author. Phone: (501) Fax: (501) plasma membrane of the host cell, and this oligosaccharide is recognized by proteins, called lectins, on the bacterial surface. These lectins are frequently part of the pili or fimbriae that are produced by many bacteria, especially gram-negative organisms (34). One of the best-studied systems in this regard is the mannose-specific lectin that is part of the type I fimbriae of various members of the family Enterobacteriaceae, including Escherichia coli (40), Klebsiella pneumoniae (9), Shigella flexneri (4), Salmonella typhimurium (24), Serratia marcescens (23), and E. cloacae (18). Although many of these studies have demonstrated inhibition of bacterial adhesion by mannose oligosaccharides, none have compared a variety of homogeneous and well-characterized high-mannose oligosaccharides to detail the best carbohydrate recognition molecule. In this study, we examined the adhesion of E. cloacae to intestinal cells to determine the molecular details of the interaction between bacterial pili and cell surface N-linked glycoproteins. We also developed assay conditions to study the effects of a variety of well-characterized high-mannose oligosaccharides in inhibiting the binding of E. cloacae, or its isolated pili, to HT-29 cell monolayers. These experiments show that pilus binding is most effectively and specifically blocked by the addition of Man 9 GlcNAc 2 glycopeptides, rather than by oligosaccharides containing fewer mannose residues. In addition, binding is enhanced when HT-29 cells are grown in the presence of inhibitors that increase the number (and thus amount) of protein-bound Man 9 (GlcNAc) 2 structures at the cell surface. Studies using purified pili show that these structures bind in a dose-dependent manner and that this binding is effectively blocked by Man 9 GlcNAc 2 -tyrosinamide, as well as by antibodies directed against the 35-kDa pilus subunit. on September 25, 2018 by guest 4199

2 4200 PAN ET AL. INFECT. IMMUN. MATERIALS AND METHODS Materials. HT-29 cells (ATCC HTB38) were obtained from the American Type Culture Collection. McCoy s 5A medium (modified) and fetal calf serum were purchased from Gibco BRL. E. cloacae was isolated from a patient at Arkansas Children s Hospital. Castanospermine was isolated from the seeds of the tree Castanospermum australe as previously described (17), and swainsonine was isolated from the leaves of the plant Astragalus lentiginosus as previously reported (32). Kifunensine was kindly supplied by M. Yamashita and M. Iwani, Fujisawa Pharmaceutical Co., Ibanaki, Japan. Deoxymannojirimycin was purchased from Boehringer Mannheim. [U- 3 H]glucose (2 to 10 Ci/mmol) was obtained from American Radiolabeled Chemicals, and concanavalin A, yeast mannan, ovalbumin, and jack bean -mannosidase were from Sigma Chemical Co. All other chemicals were purchased from reliable chemical sources and were of the highest grade available. Growth of mammalian cells and bacteria. HT-29 cells were maintained and grown in 75- or 150-cm 3 flasks in McCoy s 5A modified medium supplemented with 10% inactivated (56 C, 30 min) fetal calf serum. Growth was at 37 C in a 10% CO 2 atmosphere. HT-29 cells were released from the flasks with trypsin- EDTA, plated into 24-well plastic plates in McCoy s medium, and incubated as described above. The same number of cells was added to each well, and this number resulted in a confluent layer within 24 h. These cultures were used to measure bacterial binding. E. cloacae was grown in Trypticase soy broth for 24 to 48 h at 37 C. Cells were harvested by centrifugation, washed with phosphatebuffered saline (PBS), and then resuspended in PBS at a known concentration. Radiolabeling of E. cloacae. E. cloacae was grown in 250-ml Erlenmeyer flasks containing 50 ml of Trypticase soy broth without dextrose, and 100 Ci of [ 3 H]glucose was added to the medium. Bacteria were allowed to grow for 48 h and were harvested by centrifugation. Bacterial pellets were washed repeatedly with PBS until the supernatant liquid was free of radioactivity, and radiolabeled bacteria were suspended in PBS at a concentration of cells/ml (containing 860,000 cpm of radioactivity/ml). This suspension was aliquoted into a number of small tubes and stored in the freezer until used. This bacterial suspension was used throughout the study. There was no detectable loss in viability of the bacteria during this storage procedure. Adherence assay. Bacterial adherence was measured as follows. Confluent HT-29 cell monolayers, contained in 24-well plates, were washed two times with PBS. Various amounts of the labeled bacteria, all in the same final volume of buffer, were added to the monolayers and allowed to incubate at 37 C for 1 h with intermittent agitation. At least three wells received each concentration of bacteria. At the end of the incubation, unbound bacteria were removed by aspiration, and the monolayers were washed at least five times with PBS. After this thorough washing, the HT-29 cells were released from each well with a trypsin-edta mixture, and the cell suspension was placed in a scintillation vial. Each well was then washed with another 0.5 ml of water along with scraping with a plastic paddle to remove any firmly attached cells. This second wash was then added to the same vial that contained the first trypsin digest; 10 ml of scintillation fluid was added, and radioactivity was determined as a measure of the number of bacteria bound. No binding of bacteria was found to occur to plastic wells that did not contain HT-29 cell monolayers. Determination of sugar specificity for bacterial adherence. A variety of simple sugars (D-mannose, D-galactose, D-fructose, L-fucose, D-GlcNAc, and sialic acid) were tested at various concentrations to determine their ability to block the adhesion of E. cloacae to the HT-29 cells. In addition, p-nitrophenyl- -D-mannoside and methylumbelliferyl- -D-mannoside and glycopeptides with various high-mannose or complex types of oligosaccharides were tested as inhibitors. The structures of the most active of these N-linked oligosaccharides are shown in Fig. 3, and the details of their preparation are presented below. Oligosaccharides were also isolated from ovalbumin, and they and several partially characterized mannose polymers were examined as inhibitors of bacterial adhesion. In these experiments, the carbohydrate to be tested was added to the HT-29 cell monolayer just prior to the addition of the radiolabeled bacterial suspension, in order to compare its ability to compete with the HT-29 cell glycoproteins. The plates were then incubated and harvested as described above for the adherence assay, and the amount of radioactivity bound to the cell monolayer was determined. Preparation of various N-linked high-mannose oligosaccharide structures as potential inhibitors. High-mannose oligosaccharides containing five, six, and seven mannose residues were prepared as tyrosinamide oligosaccharides from ovalbumin as previously described (42). A Man 9 (GlcNAc) 2 -tyrosinamide oligosaccharide was prepared from soybean agglutinin as described below. Briefly, soybean agglutinin (3 g) was purified from untoasted soy flour (1 kg) as described by Lis et al. (28, 29). The N-linked oligosaccharides were released from the reduced and alkylated glycoprotein by using N-glycosidase F, and following the formation of a glycosylamine, the oligosaccharides were derivatized with Boctyrosine-N-hydroxysuccinimide ester as reported for ovalbumin oligosaccharides (42). The tyrosinamide-oligosaccharides were purified by gel filtration chromatography, and the major oligosaccharide (90% of the total) was resolved to homogeneity on an analytical reverse-phase high-pressure liquid chromatography column eluted with 0.1% acetic acid and 9% acetonitrile. The oligosaccharides could be detected by their absorption at 280 nm. The major oligosaccharide peak was pooled, dried, and characterized by proton nuclear magnetic resonance spectroscopy, which verified the presence of the -tyrosinamide linkage to a Man 9 (GlcNAc) 2 oligosaccharide. Effect of glycoprotein processing inhibitors on bacterial adhesion. Another approach to determine the N-linked oligosaccharide structure that binds the pili is to grow the HT-29 cells in the presence of inhibitors that block glycoprotein processing at various steps (6). Such inhibition results in the accumulation of immature N-linked oligosaccharides having high-mannose or partial hybrid structures. HT-29 cells were transferred at low density to multiwell plates and grown in McCoy s medium containing different concentrations of various processing inhibitors. Thus, one 24-well plate was divided into four sections, each with six wells having the following components: castanospermine at 0, 5, 10, 25, 50, and 100 g/ml; deoxymannojirimycin at 0, 5, 25, 50, 100, and 200 g/ml; kifunensine at 0, 2, 5, 10, 25, and 50 g/ml; and swainsonine at 0, 1, 5, 10, 25, and 50 g/ml. HT-29 cells were allowed to grow to confluency over a period of 4 to 5 days. Once the cells reached confluency, the medium was removed by aspiration and the monolayers were washed with PBS. Each well was then challenged with the same number of radioactive bacteria to determine the extent of bacterial adhesion as described above for the adhesion assay. Isolation of pili. Pili were isolated from E. cloacae as previously described (18). Cells were grown in Trypticase soy broth, and 6 liters of cells was harvested by centrifugation and resuspended in 250 ml of 5 mm Tris-HCl buffer (ph 7.4) containing 1 M NaCl. The cooled (4 C) cell suspension was homogenized for 4 min in a Waring blender, cooled, and homogenized again. This homogenization procedure was repeated four times. The bacteria were removed by centrifugation, resuspended in buffer, and subjected to another homogenization procedure. The supernatant liquids from the two homogenizations, containing the released pili, were combined, and solid ammonium sulfate was added slowly with stirring in an ice bath to reach a concentration of 10% (wt/vol). The mixture was allowed to stand for 60 min at 4 C, and the precipitate was removed by centrifugation and discarded. Solid ammonium sulfate was slowly added with stirring to reach a concentration of 30% (wt/vol), and the mixture was allowed to stand for 4 h at 4 C. The precipitate resulting from this treatment contained the pili, and this precipitate was collected by centrifugation and resuspended in sterile distilled water. This pilus solution was subjected to CsCl density gradient centrifugation (42%, wt/vol) for 8hat100,000 g, and the dense white band of pili was collected, diluted 1:5 in sterile distilled water, and pelleted in a fixed-angle rotor at 100,000 g for 2 h. The gelatinous pilus pellet was resuspended in sterile distilled water and stored at 4 C. Further purification of pili. The pilus preparations were frequently contaminated with outer membrane proteins such as the 35-kDa protein of OmpA. These proteins could be detected by Western blot analysis using antibody against OmpA, kindly supplied by H. Nakaido. The contaminated pili were resuspended in 4% sodium dodecyl sulfate (SDS) in 10 mm Tris buffer (ph 8) containing 15 mm dithiothreitol and stirred for 1 h at room temperature. Insoluble pili were sedimented by centrifugation at 25,000 g for1hat10 C. This pellet was resuspended in SDS-dithiothreitol buffer as described above, stirred, and subjected to centrifugation to reisolate the pili. This procedure was repeated several more times and effectively removed most of the outer membrane proteins (16). The purified pili were labeled with 125 I, using N-chloro-benzenesulfonamidederivatized polysterene beads (Iodo-beads) (30). Thus, 100 g of pili was labeled with 125 I, and then unlabeled pili were added to give a final protein concentration of 2 mg/ml. Identification and isolation of the adhesin subunit of pili. Pili were isolated and purified as described previously (27). SDS-gel electrophoresis of the pili under reducing conditions showed a major subunit of 20 kda and minor subunits of 35, 17, and 15 kda. Since the binding of E. cloacae to HT-29 cells is blocked by high-mannose structures, it was likely that the pili were type 1 pili which should bind to mannose-containing structures. Thus, purified pili were subjected to SDS 12% polyacrylamide gel electrophoresis (PAGE), and the resulting protein bands were transferred to nitrocellulose membranes. After being washed to remove SDS and incubated with PBS containing 5% bovine serum albumin, the membranes were incubated for 2 h at room temperature with a mannosylated albumin that contained a biotin tag. Some incubations also contained 50 mm D-mannose or 50 mm D-glucose. Only the 35-kDa band bound the mannosylated albumin-biotin complex, and this binding was blocked by the addition of high concentrations of D-mannose, indicating that it was specific for mannose. The 35-kDa band was obtained in sufficient amounts to obtain amino acid sequences of several different peptides. Isolation of peptides and amino acid sequencing were done at the Harvard Microchemistry Facility, Cambridge, Mass. Preparation of antibody against the 35-kDa pilus subunit. Antibody against the purified adhesin subunit (i.e., 35-kDa band) was raised in a rabbit, using standard methods of preparation. Briefly, 100 g of protein was mixed with Freund s complete adjuvant and injected into a rabbit. Four weeks later, another injection of 100 g of the pilus adhesin, mixed with Freund s incomplete adjuvant, was given to the animal. A blood sample taken after 2 weeks showed a good antibody titer, and the animal was bled for antibody isolation after 4 weeks. The immunoglobulin G (IgG) fraction was purified from the immune serum by ammonium sulfate fractionation (0 to 33% saturation), followed by gradient elution from DEAE-Sephadex. The IgG fraction was concentrated, dialyzed, aliquoted into small tubes, and stored at 80 C.

3 VOL. 65, 1997 ADHESION OF E. CLOACAE TO HT-29 CELLS 4201 FIG. 1. Kinetics of adhesion of 3 H-labeled E. cloacae to HT-29 cell monolayers. HT-29 cells were grown as described in the text. Various amounts of the labeled bacterial suspension were added, and the number of bacteria bound was determined by scintillation counting as described in Materials and Methods. RESULTS Adherence of E. cloacae to HT-29 cells. E. cloacae bound to the HT-29 cell monolayers as demonstrated in Fig. 1. This binding showed saturation kinetics, and the number of bacteria that were bound increased with increasing numbers of bacteria added to the monolayers. Thus, at saturation, bacteria bound to HT-29 cells (average of about 55 bacteria per animal cell). The number of bacteria bound was calculated based on the specific activity of the bacterial suspension (counts/min/bacterium) and was quite reproducible, with a variation of about 10% from experiment to experiment. The binding of bacteria to the monolayers was temperature sensitive and was considerably higher at 37 C than at 5 C (data not shown). The binding was also proportional to time of incubation and increased with increasing time up to about 30 min. These data are indicative of a receptor-mediated process. There was essentially no binding of the bacteria to the plastic dishes in the absence of the HT-29 cells. Binding of the labeled bacteria to the cell monolayer was inhibited by adding unlabeled E. cloacae as well as by adding various simple sugars. Thus, 50% inhibition of binding was observed in the presence of 500 M D-mannose, 150 M -D-methylmannoside, 20 M p-nitrophenyl- -D-mannoside, or 10 M 4-methylumbilliferyl- -D-mannoside (data not shown). Other simple sugars, such as D-glucose, D-galactose, L-fucose, N-acetylglucosamine, N-acetylgalactosamine, and sialic acid, did not inhibit binding even at 100 mm. Binding was reversed by adding 5 mm -methylmannoside or 0.5 mm p-nitrophenyl- -D-mannoside to the monolayers after the bacteria had bound. Since binding could be inhibited by 80 to 90% at 100 M methylumbelliferyl- mannoside, it seems likely that the E. cloacae binding to HT-29 cells is mostly mediated by type 1 pili. Adherence involves recognition of high-mannose oligosaccharides. The binding of E. cloacae to the HT-29 cell monolayers was blocked by a mixture of high-mannose oligosaccharides that were isolated by pronase digestion of ovalbumin. The inhibition curve shown in Fig. 2 demonstrates that this mixture caused a 50% inhibition of binding at about 0.1- mol equivalents of mannose (as measured by the anthrone method [41]). Figure 2 also shows that complex or hybrid types of oligosaccharides were not effective inhibitors even at high micromolar concentrations. This finding suggests that the receptor molecule on the bacteria recognizes high-mannose oligosaccharides on the HT-29 cell surface. To determine the specificity for high-mannose structures, a series of oligosaccharides that contained three to nine mannose residues, as shown in Table 1, were tested at a concentration of 25 M for the ability to inhibit the binding of E. cloacae to the HT-29 cell monolayers. The data clearly show that the Man 9 (GlcNAc) 2 -T was the best inhibitor, while the Man 7 (GlcNAc) 2 -T and Man 6 (GlcNAc) 2 -T could also block adhesion but less effectively than the Man 9 (GlcNAc) 2 structure. Table 2 further substantiates that Man 9 (GlcNAc) 2 was the best blocking agent, with 50% inhibition occurring at 5 M. However, those oligosaccharides having fewer mannoses required concentrations of 37.5 M or higher to cause 50% inhibition. Interestingly, Man 5 (GlcNAc) 2 or simpler structures were not effective at inhibiting binding, even when tested at substantially higher concentrations. Effect of inhibition of N-linked oligosaccharide processing on bacterial adhesion. As indicated above, N-linked oligosaccharides can be altered by growing eucaryotic cells in the presence of inhibitors of glycoprotein processing (25). A number of compounds are known to inhibit the glycosidases involved in the initial stages of processing (6). Two such compounds, deoxymannojirimycin and kifunensine, inhibit the -mannosidases that trim the first four mannoses from the Man 9 (GlcNAc) 2 oligosaccharide (8), whereas castanospermine prevents the removal of the glucose residues (6, 38). As a result, mannosidase I inhibitors increase the number (i.e., the amount) of Man 9 (GlcNAc) 2 structures present at the cell surface by effectively preventing their conversion to complex (and hybrid) structures. These inhibitors thus provide valuable tools to determine the in vivo role of high-mannose structures on the adhesion of E. cloacae. HT-29 cells were grown in the presence of various amounts of each inhibitor as described in Materials and Methods, and the binding of radioactive bacteria to the cells was measured. The data in Table 3 show that with kifunensine or deoxymannojirimycin, there was a 50 to 75% increase in the number of radioactive bacteria bound to the monolayers. On the other hand, castanospermine, which prevents the removal of glucose and therefore almost completely blocks any trimming (38), did not cause any significant change in the amount of radioactivity bound. Swainsonine, an inhibitor that acts on mannosidase II FIG. 2. Effects of various oligosaccharide structures on the inhibition of binding of E. cloacae to HT-29 cell monolayers. HT-29 cells were incubated with radiolabeled bacteria in the presence of various amounts of oligosaccharide prepared as indicated in Materials and Methods. The number of radioactive bacteria bound was measured as described for Fig. 1.

4 4202 PAN ET AL. INFECT. IMMUN. TABLE 1. Comparison of inhibition of binding by various well-characterized high-mannose oligosaccharides a Downloaded from a The oligosaccharides were prepared as described in Materials and Methods, and each oligosaccharide was tested at 25 M to determine its ability to function as an inhibitor. Binding was determined as for Fig. 1. and causes an increase in the formation of hybrid types of structures (7), also did not cause any change in the binding of bacteria to the HT-29 cells. These data support the hypothesis that the pili recognize Man 9 (GlcNAc) 2 and that this oligosaccharide is the cause of the initial binding of bacteria to the cells. Isolation and characterization of the adhesin subunit of E. cloacae pili. E. cloacae pili were isolated by centrifugation (Fig. 3, lane B) and purified by ammonium sulfate precipitation (lane C) and then density gradient centrifugation (lane D) as described in Materials and Methods. The protein profile of each fraction upon SDS-PAGE is shown in Fig. 3. In most cases, four protein bands with molecular masses of 35, 20, 17, and 15 kda were observed, with the 20-kDa band being the major component. However, the 17-kDa band was frequently difficult to see on the gels since it was masked by the large amount of the 20-kDa protein. The binding subunit was identified by its reaction with an albumin analog that contained covalently bound mannose as well as a biotin tag. Figure 4, lane C, shows that only the 35-kDa band became labeled with this tag, and this labeling was inhibited by adding 50 mm D-mannose (lane E), but not 50 mm D-glucose (lane D), to these incubations. Lane B in Fig. 5 shows the nitrocellulose membrane of the transferred pilus proteins stained with amido black. The data in this figure suggest that the 35-kDa protein is the adhesive subunit that recognizes Man 9 (GlcNAc) 2 structures. The 35-kDa pili subunit was transferred to polyvinylidene difluoride membranes and subjected to endoproteinase Lys-C digestion, isolation of the peptides by high-pressure liquid chromatography and amino acid sequencing of several of these peptides. Figure 5 shows the sequences of several of these on September 25, 2018 by guest

5 VOL. 65, 1997 ADHESION OF E. CLOACAE TO HT-29 CELLS 4203 TABLE 2. Effects of concentrations of various oligosaccharides on bacterial adhesion Oligosaccharide added Concn ( M) cpm Adherence a % of control None 1, Man 9 GlcNAc 2 -T Man 7 GlcNAc 2 -T Man 6 GlcNAc 2 -T Man 5 GlcNAc 2 -T 50 1, a Glycopeptides were added to assay mixtures as described in the text. The amount of radiolabeled bacteria adhering to HT-29 monolayers was measured by scintillation counting. TABLE 3. Effects of inhibitors of N-linked glycoprotein processing on the adhesion of E. cloacae to HT-29 cells a Inhibitor Concn ( g/ml) Adhesion b (cpm) Increase in adhesion (% of control) Castanospermine 0 13, , , , , ,943 8 Swainsonine 0 13, , , , , ,965 8 Deoxymannojirimycin 0 13, , , , , , Kifunensine 0 13, , , , , , a HT-29 cell were grown in the culture medium in the presence of various concentrations of processing inhibitors. When the cell layer had reached confluency (about 7 days after the last passage), the adhesion assay was performed. b Average of three set of experiments. FIG. 3. SDS-gel electrophoresis of pili at various stages of purification. Lanes: A, molecular weight standards; B, supernatant fraction after homogenization and low-speed centrifugation; C, precipitate obtained by ammonium sulfate fractionation of the supernatant fraction from lane B; D, pilus fraction isolated by CsCl density gradient fractionation; E, highly purified pili after removal of membrane proteins with Tris-HCl buffer. Proteins were detected by Coomassie blue staining. peptides compared to sequences of the FimH protein of S. typhimurium. One peptide of 15 amino acids showed 85% identity to FimH, while a 26-amino-acid sequence showed 69% identity, and another 16-amino-acid peptide showed 68% identity. This very strong similarity between these proteins is additional evidence that the 35-kDa protein from E. cloacae is the mannose-binding subunit of this pilus. Binding of the pili to HT-29 cell monolayers. The purified pili were labeled with 125 I and tested for the ability to bind to HT-29 cells. Figure 6A shows that the pili bound in a concentration-dependent manner and showed saturation kinetics. In addition, this binding was blocked in a dose-dependent manner by the addition of increasing amounts of unlabeled pili. The binding of radioactive pili was also inhibited by the addition of increasing amounts of Man 9 (GlcNAc) 2 -T as shown in Fig. 6B, with 50% inhibition occurring at 5 nm. However, Man 5 (GlcNAc) 2 did not inhibit, nor did other smaller mannosecontaining oligosaccharides. These results are similar to those FIG. 4. Detection of the pilus adhesive subunit by binding of -mannosylated-biotinylated albumin. Lanes: A, molecular weight standards as in Fig. 3; B, isolated pili after transfer to nitrocellulose membranes and detection with amido black; C, modified Western blot generated by using -mannosylated-biotinylated albumin, followed by an avidin-horseradish peroxidase conjugate and then visualization with 4-chloro-naphthol according to the manufacturer s directions; D, reaction system as for lane C, but incubated in the presence of 50 mm -Dglucose; E, reaction system as for lane C but incubated in the presence of 50 mm -D-mannose.

6 4204 PAN ET AL. INFECT. IMMUN. FIG. 5. Sequence homologies between the E. cloacae adhesive subunit (i.e., Query) and the FimH protein of S. typhimurium. Sequences were compared by a BLAST search. with the intact bacteria and strongly suggest that the 35-kDa protein is the pilus adhesin. Effect of antipilus antibody on adhesion. The IgG fraction was isolated from immune serum as described in Materials and Methods, and this antibody fraction was tested for its ability to block the adhesion of the pilus subunit, or the intact bacteria, to HT-29 cell monolayers. Figure 7 shows that this antibody was quite effective in inhibiting pilus binding, whereas nonspecific serum was ineffective. Similar results were obtained when binding of intact bacteria was examined in the presence of antipilus IgG (data not shown). Binding of the 35-kDa protein was also blocked by antipilus IgG. It should also be mentioned FIG. 6. (A) Effect of pilus concentration on binding to HT-29 cell monolayers and inhibition by unlabeled pili. Various amounts of radiolabeled pili were added to monolayers, and the extent of binding was determined. At the arrow, increasing amounts of unlabeled pili were mixed with labeled pili and added to the monolayers. (B) Pili were mixed with increasing amounts of the Man 9 (GlcNAc) 2 -T (E) orman 5 (GlcNAc) 2 -T (F), and the mixtures were added to HT-29 cells. The amount of pili bound was determined by scintillation counting. FIG. 7. Inhibition of pilus binding to HT-29 cells by antipilus antibody. Radiolabeled pili were incubated with partially purified IgG prepared against the isolated 35-kDa pilus subunit (F) or with nonspecific rabbit antibody ( ). After an incubation of 30 min at 37 C, the mixtures were added to HT-29 cells and the number of pili bound was determined as indicated for Fig. 6. that in preliminary studies with punches of fetal rabbit small intestine, radiolabeled E. cloacae or the isolated 125 I-pili showed specific and saturable binding to these tissue slices, suggesting that this interaction is a naturally and physiologically significant process. DISCUSSION The results reported in this report demonstrate that the binding of E. cloacae to HT-29 cells, a cell line derived from an intestinal tumor, involves a mannose-specific adherence event. A number of previous studies have investigated the interaction of enteric bacteria with various animal cells and shown that in many cases this binding involves recognition of mannose-containing structures on the animal cell surface by type 1 pili on the bacteria (10, 11, 33 36). However, the specific carbohydrate structure that is recognized by type 1 pili either has not been well defined or varies depending on the source of the type 1 pili. Thus, based on the ability of various mannosides to inhibit the agglutination of yeast by enteric bacteria, Firon et al. (11) concluded that the fimbrial lectins of various genera exhibit differences in sugar specificity, but all strains within a genera exhibit the same sugar specificity. These workers postulated that the combining site of the E. coli type 1 lectin corresponded to a trisaccharide of mannose that fit into a pocket (or depression) on the surface of the lectin (36). Similar patterns of specificity were observed by Nesser et al. (33), who used guinea pig erythrocytes as a model system and examined the effects of various mannose oligosaccharides on the agglutination of erythrocytes by enteropathogenic E. coli. In our studies, we have used an intestinal tumor cell line, HT-29, as a model system since these cells bind E. cloacae very well and are relatively easy to grow in cell culture. There is some question as to whether HT-29 cells are a better model of the intestinal basement membrane than Caco-2 cells, since they are apparently not as well differentiated. However, for our purposes, i.e., to study the molecular details of this interaction and to isolate sufficient amounts of the cell surface glycoprotein involved, these cells are quite appropriate, representing a useful and readily usable model. Using some well-characterized and natural N-linked oligosaccharides, having from three to nine mannose residues, as inhibitors of bacterial or pilus adhesion to HT-29 cells, we were able to establish that the

7 VOL. 65, 1997 ADHESION OF E. CLOACAE TO HT-29 CELLS 4205 E. cloacae adhesin preferred the Man 9 (GlcNAc) 2 structure over Man 7 (GlcNAc) 2 or smaller oligomers. In addition, blocking the natural processing of N-linked oligosaccharides in the HT-29 cells with specific inhibitors caused an increase in the number of Man 9 (GlcNAc) 2 structures on the HT-29 cell surface and also resulted in a considerable increase in the number of bacteria that bound to these cells. These studies suggest that once the technology for preparing large amounts of this material becomes available, the Man 9 (GlcNAc) 2 structure may be a useful agent to test in animals as an antiadhesin. In addition to the recognition of the oligosaccharide by type 1 pili, these lectins may also have a hydrophobic region that interacts with a hydrophobic site(s) on the membrane glycoprotein of the animal cells. Thus, studies by Firon et al. (10) indicated that aromatic -glycosides of mannose were more effective as inhibitors of E. coli binding to epithelial cells than was -methylmannoside. In fact, these workers demonstrated that p-nitrophenyl- -D-mannoside was considerably better as an inhibitor than was -methylmannoside, and methylumbelliferyl- -mannoside was even better than the p-nitrophenylderivative. Thus, the more hydrophobic the structure to which mannose is attached, the better the inhibitory capacity. Our studies with the E. cloacae adhesin have shown the same phenomenon, with methylumbelliferyl- -mannoside being the most effective of the simple mannosides as an inhibitor. Once we have identified the specific glycoprotein involved in binding, it will be of considerable interest and importance to examine the protein structure around the mannose oligosaccharide, to look for hydrophobic pockets that increase the binding capacity. It may be that hydrophobic peptides that mimic such sites will be valuable as antiadhesins, and a possible combination of the Man 9 (GlcNAc) 2 structure with a hydrophobic site may produce an especially potent antiadhesive agent. We have not yet isolated a specific glycoprotein from the HT-29 surface membranes that binds to the bacterial adhesin, and in vitro studies indicate that several glycoproteins can bind. Thus, in preliminary studies, SDS-treated membranes were subjected to SDS-PAGE and then overlaid with 125 I- adhesin. Several protein bands became labeled by this treatment. However, these studies may be subject to artifacts that arise once the membranes are disrupted. That is, in the intact membrane or cell, only certain glycoproteins or high-mannose structures may have the appropriate orientation to interact with the pili. Alternatively, the hydrophobic region of specific glycoproteins may be necessary for a strong interaction between pili and glycoprotein. We will explore this possibility further by doing studies to cross-link the pilus adhesin to intact HT-29 cells and attempt to isolate a radiolabeled cross-linked complex. In this regard, it should be mentioned that Rodriguez-Ortega et al. did studies to isolate the mannose-containing receptors from polymorphonuclear leukocytes and found three bands of 70 to 80, 100, and 150 kda, with the 150-kDa band being the major receptor band (39). However, they were not able to conclusively establish whether all three bands or only the 150-kDa glycoprotein was the natural receptor. The amino acid sequence of the high-mannose adhesin of E. cloacae shows very high homology to FimH of S. typhimurium, the mannose-binding protein in this organism s fimbriae (20, 27), further indicating that this protein is one of the class of mannose recognition proteins of bacterial pili. Now that we have sufficient amino acid sequence available, we will clone the gene for this adhesin to determine the binding site for Man 9 (GlcNAc) 2 as well as for the hydrophobic regions of the glycoprotein. This will also allow us to do site-directed mutagenesis studies to alter these sites. There is a considerable amount of information available in the literature concerning the structure of type 1 pili (20), and the type 1 fimbrial gene clusters of E. coli (22, 37) and K. pneumoniae (3) have been cloned. Some interesting studies showed that mutant bacteria lacking the receptor binding function were as ineffective at colonizing the bladder, as were mutants lacking the entire type 1 pili (22). These latter studies lend support to the hypothesis that interfering with the interaction of bacterial type 1 adhesin and high-mannose glycoprotein, either by occupying the highmannose site or by altering the pili in some way, can provide a useful chemotherapy for these bacterial infections. The results described here, as well as those reported in the literature, suggest an alternative to antibiotic therapy for neonatal infections that involve the gastrointestinal tract. Such an approach would utilize agents that block the attachment of bacteria to intestinal epithelial brush border cells. Thus, if a high-mannose structure that had greater affinity for the pili than those currently available, and perhaps also had a hydrophobic component, could be designed, it could be infused into the intestinal tract of neonates to prevent infection. The problem with the Man 9 (GlcNAc) 2 structure used in this study is that it is a monovalent ligand and probably has a lower K i of binding. Thus, it seems likely that in interactions of the type described here, polyvalent ligands are involved in the binding. If a polyvalent ligand having multiple oligosaccharides could be developed, such a compound would probably be more effective. Studies are under way to develop such compounds and to test their efficacy as inhibitors. ACKNOWLEDGMENTS The first two authors contributed equally to this work. This study was supported in part by a grant from the National Institutes of Health (DK-21800) to A.D.E. and research development funds from the Division of Pediatric Surgery, University of Arkansas for Medical Sciences, to S.S. REFERENCES 1. Acolet, D., Z. Ahmet, and E. Houang Enterobacter cloacae in a neonatal intensive care unit: account of an outbreak and its relationship to use of third generation cephalosporins. J. Hosp. Infect. 28: Beachy, E. H Bacterial adherence: adherence-receptor interactions mediating the attachment of bacteria to mucosal surface. J. Infect. Dis. 143: Cleeg, S., J. Pruckler, and B. K. Purcell Complementation analyses of recombinant plasmids encoding type 1 fimbriae of members of the family Enterobacteriaceae. Infect. Immun. 50: Duguid, J. P., and D. C. Old Adhesive properties of Enterobacteriaceae, p In E. H. Beachey (ed.), Bacterial adherence, series B, vol. 6. Chapman and Hall, London, England. 5. Eden, C. S., B. Andersson, G. Aniansson, R. Lindstedt, P. deman, A. Nielsen, H. Leffler, and A. Wald Inhibition of bacterial attachment: examples from the urinary and respiratory tracts. Curr. Top. Microbiol. Immunol. 151: Elbein, A. D Glycosidase inhibitors: inhibition of N-linked oligosaccharide processing. FASEB J. 5: Elbein, A. D., R. Solf, P. R. Dorling, and K. Vosbeck Swainsonine, an inhibitor of glycoprotein processing. Proc. Natl. Acad. Sci. USA 78: Elbein, A. D., J. E. Tropea, M. Mitchell, and G. P. Kaushal Kifunensine, a potent inhibitor of the glycoprotein processing mannosidase I. J. Biol. Chem. 265: Fader, R. C., K. L. Duffy, C. P. Davis, and A. Kurosky Purification and chemical characterization of type I pili isolated from Klebsiella pneumoniae. J. Biol. Chem. 257: Firon, N., S. Ashkenazi, D. Mirelman, I. Ofek, and N. Sharon Aromatic -glycosides of mannose are powerful inhibitors of the adherence of type I fimbriated Escherichia coli to yeast and intestinal cells. Infect. Immun. 55: Firon, N., I. Ofek, and N. Sharon Carbohydrate-binding sites of the mannose-specific fimbrial lectins of enterobacteria. Infect. Immun. 43: Gallagher, P. G Enterobacter bacteremia in pediatric patients. Rev. Infect. Dis. 12:

8 4206 PAN ET AL. INFECT. IMMUN. 13. Giampapa, C. S., S. N. Abraham, T. M. Chiang, and E. H. Beachey Isolation and characterization of a receptor for type 1 fimbriae of Escherichia coli from guinea pig erythrocytes. J. Biol. Chem. 263: Gladstone, I. M., R. A. Ehrenkranz, and S. C. Edberg A 10 year review of neonatal sepsis in comparison with the previous 50 year experience. Pediatr. Infect. Dis. J. 9: Hanson, M. S., and C. C. Brinton, Jr Identification and characterization of E. coli type 1 pilus tip adhesion protein. Nature (London) 332: Hanson, M. S., J. Hemper, and C. C. Brinton Purification of the Escherichia coli typi 1 pilin and minor pilus proteins and partial characterization of the adhesin protein. J. Bacteriol. 170: Hohenschutz, L. D., E. A. Bell, P. J. Jewess, P. P. Leworthy, R. J. Pryce, E. Arnold, and J. Clardy Castanospermine, a 1,6,7,8-tetrahydroxy-octahydroindolizidine alkaloid from the seeds of Castanospermum australe. Phytochemistry 20: Hornick, D. B., B. L. Allen, M. A. Horn, and S. Clegg Fimbrial types among respiratory isolates belonging to the family of Enterobacteriaceae. J. Clin. Microbiol. 29: Husser, M. F., J. E. Patterson, and A. P. Kuritza Emergence of resistance to multiple beta lactams in Enterobacter cloacae during treatment for neonatal meningitis with cefotaxime. Pediatr. Infect. Dis. J. 9: Jones, C. H., J. S. Pinkner, R. Roth, J. Heuser, A. V. Nicholes, S. N. Abraham, and S. J. Hultgren FimH adhesin of type 1 pili is assembled into a fibrillar tip structure in the Enterobacteriaceae. Proc. Natl. Acad. Sci. USA 92: Karpuch, J., M. Goldberg, and D. Kohelet Neonatal bacteremia: a 4 year prespective study. Isr. J. Med. Sci. 19: Keith, B. R., L. Maurer, P. A. Spears, and P. E. Orndorff Receptorbinding function of type 1 pili effects bladder colonization by a clinical isolate of Escherichia coli. Infect. Immun. 53: Kohno, K., T. Yamamoto, A. Kuroiwa, and K. Amako Purification and characterization of Serratia marcescens US5 pili. Infect. Immun. 46: Korhonen, T. K., K. Lounatmaa, H. Ranta, and N. Kuusi Characterization of type I pili of Salmonella typhimurium LT2. J. Bacteriol. 144: Kornfeld, R., and S. Kornfeld Assembly of asparagine-linked oligosaccharides. Annu. Rev. Biochem. 54: Krogfelt, K. A Bacterial adhesion: genetics, biogenesis, and role in pathogenesis of fimbrial adhesins of Escherichia coli. Rev. Infect. Dis. 13: Krogfelt, K. A., H. Bergmans, and P. Klemm Direct evidence that the Editor: P. E. Orndorff FimH protein is the mannose-specific adhesin of Escherichia coli type I fimbriae. Infect. Immun. 58: Lis, H., and N. Sharon Soybean agglutinin a plant glycoprotein: structure of the carbohydrate unit. J. Biol. Chem. 253: Lis, H., N. Sharon, and E. Katchalski Soybean hemagglutinin, a plant glycoprotein: isolation of glycopeptides. J. Biol. Chem. 241: Markwell, M. A A new solid-state reagent to iodinate proteins. Anal. Biochem. 125: Minion, F. C., S. N. Abraham, E. H. Beachey, and J. D. Goguen The genetic determinant of adhesive function in type 1 fimbriae of Escherichia coli is distinct from the gene encoding the fimbrial subunit. J. Bacteriol. 165: Molyneux, R. J., and J. P. James Loco intoxication: indolizidine alkaloids of spotted locoweed (Astragalus lentiginosus). Science 216: Nesser, J.-R., B. Koellreutter, and P. Wuersch Oligomannoside-type glycopeptides inhibiting adhesion of Escherichia coli strains mediated by type 1 pili: preparation of potent inhibitors from plant glycoproteins. Infect. Immun. 52: Ofek, I., and R. J. Doyle Bacterial adhesion to cells and tissues. Chapman and Hall, New York, N.Y. 35. Ofek, I., A. Mosek, and N. Sharon Mannose-specific adherence of Escherichia coli freshly excreted in the urine of patients with urinary tract infections, and of isolates subcultured from the infected urine. Infect. Immun. 34: Ofek, I., and N. Sharon Adhesins as lectins: specificity and role in infection. Curr. Top. Microbiol. Immunol. 151: Orndorff, P. E., and S. Falkow Organization and expression of genes responsible for type 1 piliation in Escherichia coli. J. Bacteriol. 159: Pan, Y. T., H. Hori, R. Saul, B. A. Sanford, R. J. Molyneux, and A. D. Elbein Castanospermine inhibits processing of the oligosaccharide portion of the influenza viral hemagglutinin. Biochemistry 22: Rodriguez-Ortega, M., I. Ofek, and N. Sharon Membrane glycoproteins of human polymorphonuclear leukocytes that act as receptors for mannose-specific Escherichia coli. Infect. Immun. 55: Salit, I. E., and E. C. Gotschlick Hemagglutination by purified Escherichia coli pili. J. Exp. Med. 146: Seifter, S., S. Dayton, B. Novic, and E. Muntwyler The estimation of glycogen with the anthrone reagent. Arch. Biochem. Biophys. 25: Silva, M., H. J. Stubbs, T. Tamura, and K. C. Rice H-NMR characterization of a hen ovalbumin tyrosinamide N-linked oligosaccharide library. Arch. Biochem. Biophys. 318: