The Cellular Form of Human Fibronectin as an Adhesion Target for the S Fimbriae of Meningitis-Associated Escherichia coli

Abstract

The adhesion of the S fimbriae of meningitis-associated Escherichia coli O18ac:K1:H7 to the cellular and the plasma forms of human fibronectin was studied. E. coli HB101(pAZZ50) expressing the complete S-fimbria II gene cluster of E. coli O18 adhered to cellular fibronectin (cFn) on glass but not to plasma fibronectin (pFn). Adhesion to cFn was specifically inhibited by neuraminidase treatment of cFn as well as by incubation of the bacteria with sialyl-α2-3-lactose, a receptor analog of the S fimbriae. No significant adhesion to cFn or pFn was detected with E. coli HB101(pAZZ50-67) expressing S fimbriae lacking the SfaS lectin subunit. Strain HB101(pAZZ50) also adhered to a human fibroblast cell culture known to be rich in cFn, and the adhesion was specifically inhibited in the presence of polyclonal antibodies to cFn. The results show that the SfaS lectin of the S fimbriae mediates the adherence of meningitis-associated E. coli to sialyl oligosaccharide chains of cFn.

The S fimbriae of Escherichia coli recognize terminal sialyl-α2-3-galactoside moieties of glycoproteins (11) and are associated with serogroup O18ac:K1:H7 E. coli as well as newborn meningitis (12). Genes encoding S-fimbrial adhesin (Sfa) complexes have been cloned and characterized from uropathogenic O6:K15:H31 E. coli 536 (the sfaI gene cluster) (4) and meningitis-associated O18:K1:H7 E. coli IHE3034 (sfaII) (5). Nine chromosomal genes in the sfa gene cluster are involved in the biogenesis and structure of the S fimbriae. The major subunit, SfaA, forms the bulk of the fimbrial filament, with which three minor subunits (SfaG, SfaH, and SfaS) are associated. SfaS is the sialic acid-binding adhesin and is responsible for S-fimbrial binding to various human epithelial and subepithelial tissue domains (33). In animal models of newborn meningitis, it has been found that the S fimbriae are expressed in vivo in blood and cerebrospinal fluid (21, 32) and that they bind to epithelial cells lining the choroid plexuses and brain ventricles and to subarachnoid endothelium (24). These findings are in agreement with the observation that the choroid plexus is the portal of entry into the cerebrospinal fluid by meningitis-associated bacteria (16) and have suggested that S fimbriae are involved in targeting E. coli O18:K1:H7 to the endothelium and epithelium in the mammalian choroid plexus (13).

Fibronectin is a high-molecular-mass (450- to 500-kDa) mammalian glycoprotein found in a soluble form in plasma and other body fluids as well as in an insoluble form in the interstitial connective tissues and extracellular matrices (1, 19, 28, 31, 45). Fibronectin is a dimer of two identical or very similar polypeptide chains that are assembled into a series of continuous structural and functional domains (28). Fibronectin is involved in many important functions of cells and tissues, such as cell adhesion, spreading, and migration, tissue development and differentiation, blood clot stabilization, and wound healing (1, 31, 45). The functions of fibronectin involve binding to a number of biological structures, e.g., eukaryotic cell surface receptors (integrins), extracellular matrix components (collagen and heparin), cytoskeletal components (actin), and plasma proteins (fibrin) (31, 45).

Fibronectins can be divided into two major forms: plasma fibronectin (pFn) and cellular fibronectin (cFn). pFn is produced by hepatocytes in the liver and secreted in a soluble form into plasma (36). cFn is produced locally in tissues by different cell types, such as fibroblasts (6, 19) and endothelial cells (26). cFn is mostly bound to the cell surface or deposited as an insoluble multimer in the extracellular matrix (7, 19). Structural differences between pFn and cFn in part result from different splicing patterns of the fibronectin gene. cFn contains three extra domain sequences: EDIIIA and EDIIIB, as well as a type III connecting strand lacking pFn (28, 35). Fibronectins contain 4 to 9% carbohydrate, mostly in N glycosidically linked complex oligosaccharide chains (1, 28). cFn contains sialic acid linked to galactose via an α2-3 linkage, whereas the α2-6 linkage is found in pFn (1, 20, 28).

Various bacteria pathogenic for humans recognize pFn (reviewed in references 25 and 43). Mutant strains of Staphylococcus aureus and Streptococcus sanguis with reduced fibronectin-binding activity have a decreased ability to cause infective endocarditis in rats (15, 17), suggesting that fibronectin binding is important for bacterial adhesion and colonization at damaged heart valves. Selective binding of the virulence-associated surface protein YadA of Yersinia enterocolitica to cFn has been reported (34). YadA exhibits multiple adhesive functions and promotes the invasiveness of yersiniae for orally infected mice (30, 37), but the role of cFn binding in bacterial spread has not been analyzed.

Several meningitis-associated bacteria have recently been found to bind pFn; these bacteria include Haemophilus influenzae (40), Neisseria meningitidis (2), and Streptococcus pneumoniae (38). The present study was undertaken to determine whether meningitis-associated S-fimbriated E. coli also expresses fibronectin binding.

Adhesion of S-fimbriated E. coli strains to immobilized fibronectins.

We initially tested the adhesion of S-fimbriated E. coli strains to human pFn and cFn. pFn was from normal human plasma (Collaborative Biomedical Products, Bedford, Mass.), and cFn was from human foreskin fibroblasts (Fibrinogenex, Chicago, Ill.). Laminin, a major glycoprotein of basement membranes (18) from Engelbreth-Holm-Swarm mouse tumors (Upstate Biotechnology Inc., Lake Placid, N.Y.) has been shown to be recognized by S fimbriae (41) and was used as a positive control protein. Type IV collagen from human placenta (Sigma Chemical Co., St. Louis, Mo.) and bovine serum albumin (BSA; Sigma) were used as negative control proteins (41). Matrix proteins were immobilized on glass slides to obtain 2.5 pmol per well as described earlier (44). Recombinant E. coli strains HB101(pAZZ50), expressing the complete sfaII gene cluster from E. coli O18:K1:H7, and HB101(pAZZ50-67), expressing an sfaS-deficient gene cluster (5), as well as the nonfimbriated strain HB101(pBR322), with the vector plasmid alone, were grown overnight at 37°C on Luria agar plates supplemented with ampicillin. Before adhesion tests, the correct expression of the fimbriae by the bacteria was confirmed by hemagglutination of human blood group O erythrocytes, mannoside-sensitive yeast cell agglutination (8), and bacterial agglutination in S-fimbria-specific antiserum (41). Adhesion assays were performed as detailed recently (40). The number of adherent bacteria in 20 randomly chosen microscopic fields of 1.6 × 10⁴ μm² was determined by density slicing.

In accordance with our earlier findings (41), E. coli HB101(pAZZ50), with intact S-fimbrial filaments, strongly adhered to laminin. The strain also adhered efficiently to cFn (Fig. 1), whereas adherence to pFn was poor and equal to that seen with BSA. The difference in adhesiveness to cFn and pFn was statistically significant (P, <0.001). No adhesion to type IV collagen was detected. The sfaS mutant strain HB101(pAZZ50-67) and the nonfimbriated strain HB101(pBR322) exhibited only poor adhesiveness to the target proteins. These results suggested that the sialic acid-binding protein SfaS of the fimbrial filament is involved in adhesion.

FIG. 1.

Adhesiveness of E. coli HB101(pAZZ50), expressing the complete S-fimbria gene cluster, of HB101(pAZZ50-67), with SfaS-defective fimbriae, and of HB101(pBR322), with the vector plasmid alone, to target proteins immobilized on glass. The target proteins were laminin, cFn, pFn, type IV collagen, and BSA. Means ± standard deviations of adherent bacteria in 20 microscopic fields of 1.6 × 10⁴ μm² are shown.

Role of cFn carbohydrate in bacterial adhesion.

S fimbriae recognize terminal sialyl-α2-3-galactosides (11) present in numerous mammalian glycoproteins (14). Human pFn and cFn mainly carry biantennary oligosaccharide chains with terminal sialic acids. cFn contains sialic acid linked to galactose via an α2-3 linkage, in contrast to the α2-6 linkage found in pFn (1). To confirm the presence of terminal sialyl structures in the commercial target proteins used in this study, we used a dot blot assay to assess binding by digoxigenin-labeled Sambucus nigra agglutinin (SNA) and Maackia amurensis agglutinin (MAA), included in a glycan differentiation kit (Boehringer GmbH, Mannheim, Germany). SNA recognizes terminal sialyl-α2-6-galactosides, and MAA recognizes sialyl-α2-3-galactosides. Laminin, pFn, cFn, and type IV collagen were immobilized on nitrocellulose membranes (Bio-Rad Laboratories, Richmond, Calif.) at a concentration of 2 μg per dot. Lectin staining was performed as described by Boehringer. SNA lectin reacted strongly with pFn (Fig. 2A), whereas MAA lectin bound to laminin and cFn (Fig. 2B), indicating the presence of terminal sialyl-α2-3-galactosides in these proteins.

FIG. 2.

Reactivity of the sialyl-α2-6-galactoside-binding SNA lectin (A) and of the sialyl-α2-3-galactoside-binding MAA lectin (B) with laminin (Lam), pFn, cFn, and type IV collagen (CIV). The target proteins were immobilized on a nitrocellulose membrane, and the binding was visualized with digoxigenin-labeled lectins.

To analyze the role of terminal sialyl-α2-3-galactosides in the observed bacterial adhesion to cFn, we tested the effect of neuraminidase treatment of cFn on bacterial adhesiveness. Immobilized cFn and type IV collagen were incubated with neuraminidase (100 mU/ml in Dulbecco’s phosphate-buffered saline [PBS]; Boehringer) at 37°C for 3 h before bacteria (10⁹ cells/ml) were added; control wells were treated with buffer alone. Neuraminidase treatment significantly (P, <0.001) decreased bacterial adhesion to cFn (Fig. 3A) but did not affect low-level adhesiveness to type IV collagen. We also tested the effect of sialyl-α2-3-lactose and lactose on the adhesiveness of E. coli HB101(pAZZ50). Bacteria were incubated with 28.5 mM sialyl-α2-3-lactose (10, 23) or lactose (Merck) for 30 min over crushed ice and then pipetted onto cFn- and type IV collagen-coated glass slides. Lactose had no significant effect on adhesion to cFn, whereas sialyl-α2-3-lactose inhibited adhesion significantly (P, <0.001) to close to the level seen with type IV collagen (Fig. 3B).

FIG. 3.

Inhibition of the adherence of E. coli HB101(pAZZ50) to cFn and type IV collagen (CIV) immobilized on glass. (A) Bacterial adherence to neuraminidase-treated cFn and CIV; the control bars show adherence to target proteins incubated in buffer alone. (B) Effect of lactose and sialyl-α2-3-lactose on adherence. The carbohydrates were tested at a concentration of 28.5 mM. Means ± standard deviations of adherent bacteria in 20 microscopic fields of 1.6 × 10⁴ μm² are shown.

Adhesion to human fibroblasts.

cFn is very efficiently produced by human fibroblast cell lines in patches or long fibers along cell surfaces and in intercellular spaces (7). We next assessed whether S-fimbriated bacteria also recognize cFn present on cell surfaces; this assessment was performed by double staining (10) of the fibroblast culture with anti-cFn antibody and bacterial cells. Human embryonic skin fibroblasts (6) were grown to subconfluence on glass slides in MEM medium (Life Technologies Gibco BRL, Paisley, Scotland) supplemented with 10% (wt/vol) fetal calf serum (PAA Laboratories GmbH, Linz, Austria) and 2 mM l-glutamine (Life Technologies Gibco BRL). E. coli clones HB101(pAZZ50) and HB101(pAZZ50-67) were labeled with fluorescein isothiocyanate (FITC; Sigma) as described earlier (10). FITC-labeled bacteria (10⁹ cells/ml in PBS) were pipetted onto glass slides and incubated at 4°C for 2 h. After being washed, the cells were fixed with cold 3.5% paraformaldehyde in PBS for 10 min and then were washed again with PBS. For localization of cellular fibronectin, cells were stained with monoclonal mouse anti-cFn antibodies specific for EDIIIA (diluted 1:100; Biohit, Helsinki, Finland) and tetramethyl rhodamine isothiocyanate R (TRITC)-labeled secondary anti-mouse antibodies (diluted 1:50; Dako A/S, Glostrup, Denmark). The cells were mounted with Nicethamid and examined in an Olympus standard fluorescence microscope (Olympus Optical Co., Hamburg, Germany).

FITC-labeled S-fimbriated HB101(pAZZ50) bacteria strongly adhered to fibroblast cells (Fig. 4A to C), whereas FITC-labeled HB101(pAZZ50-67) bacteria, with sfaS-deficient type S fimbriae, failed to adhere (Fig. 4D and E). HB101(pAZZ50) cells were detected at sites recognized by the anti-cFn monoclonal antibody (Fig. 4A to C, arrows). However, bacterial cells were also detected at locations that were not stained by the anti-cFn antibody (Fig. 4A to C, arrowheads). We estimated that 75% of the adherent HB101(pAZZ50) cells colocalized with the anti-cFn antibody-binding sites on fibroblasts. We also assessed the binding of purified S fimbriae to fibroblasts. Fimbriae from strain HB101(pAZZ50) were purified by use of deoxycholate and concentrated urea (9). Fibroblasts on glass slides were incubated with purified S fimbriae (950 μg/ml in 28.5 mM lactose or in 28.5 mM sialyl-α2-3-lactose) at 4°C for 2 h. After being washed, the cells were fixed with 3.5% paraformaldehyde and then were washed again. To detect S-fimbrial binding, cells were incubated first with anti-S-fimbria antibodies (diluted 1:50) and then with FITC-labeled secondary antibodies (Dako A/S) (diluted 1:50). S fimbriae bound to fibroblast cells, and the binding was specifically inhibited by sialyl-α2-3-lactose (Fig. 4F to I).

FIG. 4.

Binding of S-fimbriated E. coli and of S fimbriae to human fibroblasts. (A) Adhesiveness of FITC-labeled E. coli HB101(pAZZ50). (B) Same microscopic field as in panel A double stained with the anti-cFn monoclonal antibody and TRITC-labeled secondary antibodies. (C) Same field as in panel A visualized by light microscopy. Arrows indicate bacterial adhesion to cFn-containing sites, and arrowheads indicate adhesion to other cell surface structures lacking cFn. (D) Adhesiveness of FITC-labeled HB101(pAZZ50-67), devoid of the SfaS lectin. (E) Same field as in panel D visualized by light microscopy. (F and H) Binding of purified S fimbriae in the presence of 28.5 mM lactose (F) and in the presence of 28.5 mM sialyl-α2-3-lactose (H). (G and I) Same fields as in panels F and H, respectively, visualized by light microscopy. Bar, 10 μm.

Terminal sialyl-α2-3-galactosides occur commonly in mammalian glycoproteins (14), and our observations above suggested that the fibroblast surface may express other receptor-active glycoproteins in addition to cFn. We analyzed the effect of antibodies specific for cFn on the adhesiveness of E. coli HB101(pAZZ50). Bacteria (10⁹ cells/ml) and antibodies, polyclonal anti-collagen type IV immunoglobulin G (IgG) (2 mg/ml) or anti-cFn IgG, were pipetted onto fibroblast cells and incubated as described above. In the presence of polyclonal anti-collagen type IV IgG, bacterial adhesion was not significantly decreased (Fig. 5, bars 1 and 2). In contrast, polyclonal anti-cFn IgG (Fig. 5, bars 1 and 3) decreased bacterial adhesion significantly (P, <0.001) but did not completely abolish it.

FIG. 5.

Adherence of E. coli HB101(pAZZ50) to human fibroblasts. Bar 1 shows bacterial adhesiveness (10⁹ cells/ml) in PBS, bar 2 shows adhesiveness in the presence of polyclonal IgG (2 mg/ml) against type IV collagen, and bar 3 shows adhesiveness in the presence of polyclonal IgG (2 mg/ml) against cFn. Bar 4 shows adherence of SfaS mutant HB101(pAZZ50-67). The adhesiveness shown in bars 3 and 4 differed statistically significantly (P, <0.001) from that shown in bar 1, whereas the adhesiveness shown in bar 2 did not differ significantly from that shown in bar 1. Error bars indicate standard deviations.

Our results demonstrate that sialyl-α2-3-oligosaccharide chains of cFn serve as adhesion targets for S-fimbriated E. coli. The adhesion of S-fimbriated bacteria to cFn was inhibited by sialyl-lactose as well as by the removal of terminal sialic acid from the cFn molecule. Furthermore, SfaS, the lectin protein of the S-fimbrial filament, was needed for cFn binding. On cultured human fibroblasts, which express cFn efficiently (6), partial overlap of the S-fimbria- and anti-cFn antibody-binding sites was seen. Furthermore, the anti-cFn antibody caused specific but partial inhibition of bacterial adhesion. The latter results are compatible with the common occurrence of sialyl-α2-3-oligosaccharides in mammalian glycoproteins and the hypothesis that the fibroblast surface expresses various glycoproteins recognized by S fimbriae.

The common expression of fibronectin adhesiveness by meningitis-associated or invasive bacterial species suggests that it provides a pathogenic function for the bacteria. At present, we can only speculate on the role of cFn recognition in bacterial meningitis. In embryonic tissues, cFn is found in developing basement membranes, but in adult tissues, cFn is found mostly in vascular endothelial cells (39). Both tissue types are strongly recognized by S fimbriae (10, 13, 42). During tissue injury and repair, the distribution of fibronectin in the body changes. During vascular injury, cFn is released and becomes widely distributed at inflamed tissue sites. It accumulates at sites of injury and inflammation, where it appears to provide a provisional matrix for repair processes (27). cFn has been detected in tissues with local trauma, such as bacterial infection (22, 29), tumor metastasis (39), or wound healing and regeneration (3). Local production of cFn may increase the colonization potential of S-fimbriated E. coli at sites of tissue trauma or inflammation.