Sulfatide Recognition by Colonization Factor Antigen CS6 from Enterotoxigenic Escherichia coli

The first step in the pathogenesis of enterotoxigenic Escherichia coli (ETEC) infections is adhesion of the bacterium to the small intestinal epithelium. Adhesion of ETEC is mediated by a number of antigenically distinct colonization factors, and among these, one of the most commonly detected is the non-fimbrial adhesin coli surface antigen 6 (CS6). The potential carbohydrate recognition by CS6 was investigated by binding of recombinant CS6-expressing E. coli and purified CS6 protein to a large number of variant glycosphingolipids separated on thin-layer chromatograms. Thereby, a highly specific binding of the CS6-expressing E. coli, and the purified CS6 protein, to sulfatide (SO3-3Galβ1Cer) was obtained. The binding of the CS6 protein and CS6-expressing bacteria to sulfatide was inhibited by dextran sulfate, but not by dextran, heparin, galactose 4-sulfate or galactose 6-sulfate. When using recombinantly expressed and purified CssA and CssB subunits of the CS6 complex, sulfatide binding was obtained with the CssB subunit, demonstrating that the glycosphingolipid binding capacity of CS6 resides within this subunit. CS6-binding sulfatide was present in the small intestine of species susceptible to CS6-mediated infection, e.g. humans and rabbits, but lacking in species not affected by CS6 ETEC, e.g. mice. The ability of CS6-expressing ETEC to adhere to sulfatide in target small intestinal epithelium may thus contribute to virulence.


Introduction
Attachment of microbes to cell surface receptors on the target tissue is considered an essential step in the initiation, establishment and maintenance of infection. As a result great interest has been shown in the elucidation and identification of potential microbial host receptors, the majority of which appear to be glycoconjugates [1][2][3]. Glycoconjugates exhibit a characteristic and specific pattern of expression, which is dependent on the animal species, individual and cell type [4], thus explaining the phenomenon of tropism of infection.
Adherence of enterotoxigenic Escherichia coli (ETEC) is mediated by colonization factors (CFs), which usually are fimbrial structures present on the bacterial cell surface. Infecting ETEC adhere to and colonize the intestinal epithelium, and cause diarrhea primarily by the production of heat-labile and/or heat-stable enterotoxin (LT and ST respectively). Around 25 different CFs have been identified [5], and one of the most commonly detected is coli surface antigen 6 (CS6) [5][6][7][8]. CS6 is non-fimbrial, but the overall structure of this adhesin has not yet been defined.
The CS6 operon required for assembly of the CS6 adhesin contains four open reading frames [9]. Two heterologous major structural subunits, CssA and CssB, are encoded by the cssA gene and the cssB gene, respectively. The cssC gene encodes a chaperone that was assumed to assist in the folding of CssA and CssB, and cssD encodes a tentative usher involved in the transport of CssA and CssB across the outer membrane. In a recent study a series of deletions were made in each of the genes of the CS6 operon, and the effects on the expression of CssA and CssB were examined [10]. Deletion of the chaperone CssC gave reduced levels of the CssA subunit, while the expression of the CssB subunit was not affected. Reduced levels of the CssA subunit was also obtained when the cssB gene was deleted. Thus, the CssA subunit requires CssC for folding, and is stabilized by interaction with CssB. Surprisingly, the expression of the CssA and CssB subunits was not affected by deletion of the usher CssD, leading to the suggestion that the usher is not involved in the assembly or surface expression of CS6.
Using the rabbit non-ligated intestinal model (RITARD) it has been shown that CS6 mediates binding of ETEC in rabbit intestine, where colonization was obtained by a CS6-positive strain, but not with the isogenic CS6-deficient strain [11]. A previous study has also shown that the binding of CS6 to rabbit enterocytes, and to mucus from rabbit and human intestines, was abolished by treatment with meta-periodate, indicating the involvement of carbohydrates in the binding process [12].
To further examine the potential role of carbohydrates as adhesion receptors for CS6, the binding of purified CS6 protein and recombinant CS6-expressing E. coli to glycosphingolipids was investigated in the present study. Thereby, a specific interaction between CS6-expressing bacteria, and purified CS6 protein, and sulfatide (SO 3 -3Galb1Cer) was detected. Binding assays with purified CssA and CssB subunits demonstrated that the CssB subunit carries the sulfatide binding capacity. Furthermore, a correlation between the expression of sulfatide in target cells and susceptibility to CS6-mediated ETEC infection was found.

Characterization of the CS6 protein and the CssA and CssB subunits
The three proteins were purified by the chromatgraphic steps described in the Materials and methods section. The purified proteins migrated as single bands, and the apparent molecular weights were in agreement with the predicted molecular masses for the constructed proteins 18483.61 for CssA (with the polyhistidine tag) and 43,997.93 for CssB (fused to glutathione-S-transferase carrying a His tag (Fig. 1).
Specific recognition of sulfatide by CS6-expressing E. coli and the CS6 protein In order to expose the CS6 protein to a large number of potentially binding-active carbohydrate structures, mixtures of glycosphingolipids from various sources separated on thin-layer plates were used in the initial screening for carbohydrate recognition by the CS6 adhesin. When using mixtures of acid glycosphingolipids, i.e. sulfate-containing glycosphingolipids and sialic acid-containing glycosphingolipids (gangliosides), a distinct binding of the protein to a fast-migrating compound in the acid fraction of human small intestine (Fig. 2, lane 2), human meconium (lane 5), and human colon cancer (lane 6) was observed. No binding of the CS6 protein to the more slowmigrating compounds (mainly gangliosides) in the acid fractions, or to any non-acid glycosphingolipids was obtained.
The binding of the CS6 protein to a fast-migrating compound in the acid fraction of human small intestine and human meconium directed our interest towards sulfate-containing glycosphingolipids, since the major glycosphingolipid of these fractions is sulfatide [13,14].
The TOP10 background strain binds to glycosphingolipids of the ganglio-series, as e.g. fucosyl-gangliotetraosylceramide (Fuca2-Galb3GalNAcb4Galb4Glcb1Cer; No. 8 in Table 1; Fig. 3C, lane 1, lower band) [15], and this binding was also obtained with the recombinant TOP10-CS6 strain (Fig. 3D). However, while both the CS6 protein and the CS6-positive E. coli bound to sulfatide, no binding of the TOP10 background strain to this compound was obtained, demonstrating that the recognition of sulfatide by the TOP10-CS6 strain was indeed dependent on the CS6 protein.

Sulfatide binding is mediated by the CssB subunit
In order to evaluate the roles of the CssA and CssB structural subunits of CS6 in the sulfatide binding process, these two proteins were expressed separately. Chromatogram binding experiments using 125 I-labeled subunits gave a specific binding of the CssB subunit to sulfatide, the major acid glycosphingolipid of piglet intestine [16] (Fig. 6D, lane 8), in the same manner as the CS6 protein (Fig. 6B, lane 8). The CssA subunit, on the other hand, bound to several major compounds both in the non-acid fractions (Fig. 6C, lanes 1-5) and in the acid fractions (Fig. 6C, lanes 6-8), in an unspecific manner demonstrating a general stickiness of this

Presence of sulfatide in target small intestinal target cells
To assess the potential role of the sulfatide recognition by CS6 in target tissue adherence, the binding of recombinant CS6expressing E. coli to total acid glycosphingolipid fractions from human, mouse and rabbit small intestines was examined. The CS6-positive bacteria bound to a compound migrating as sulfatide in the acid glycosphingolipid fractions from human small intestine (Fig. 7, lanes 1-3), and rabbit small intestine (Fig. 7, lane 5), while no binding to the acid glycosphingolipids from mouse small intestine (Fig. 7, lane 4) occurred. Sulfatide isolated from human small intestine was also recognized by the CS6-expressing bacteria (Fig. 7D, lanes 3 and 4).

Discussion
Susceptibility to CS6-mediated infections varies among species, and while e.g. humans and rabbits are highly susceptible, mice are not colonized [11,17]. Here we report that the CssB subunit of the CS6 protein confers a specific binding of this colonization factor to a glycosphingolipid present in human and rabbit small intestine. This CS6-binding compound was identified as sulfatide (SO 3 -3Galb1Cer), which is the major acid glycosphingolipid of the human small intestinal epithelial cells [13]. Binding of CS6expressing E. coli to sulfatides isolated from human small intestinal epithelium was also demonstrated.
The binding studies with the recombinant CssA and CssB subunits were done with proteins having remnant affinity tags. We have now developed an affinity system based on a sulfate-carrying polymer (Lebens et al., to be published). The CS6 protein and the native CssB subunit, but not the native CssA subunit, are retained on this affinity column, supporting the suggestion that the sulfatide recognition of CS6 is mediated by the CssB subunit. Furthermore, in a recent study the binding to CaCo-2 cells by recombinant E. coli with deletion of the cssA gene was reported [10]. In contrast,  Radiolabeled CS6 protein and recombinant CS6-expressing bacteria were incubated with dextran or dextran sulfate (0.01-1 mg/ml) for 1 h at room temperature. Thereafter the suspensions were utilized in the chromatogram binding assay or the microtiter well assay as described under ''Materials and Methods''. Autoradiograms obtained by binding of 35 S-labeled E. coli TOP10-CS6 strain incubated with dextran (A), and with dextran sulfate (B). Autoradiography was for 12 h. The lanes were: Lane 1, sulfatide (SO 3 -3Galb1Cer), 4 mg; Lane 2, sulfatide, 2 mg; Lane 3, sulfatide, 1 mg. Binding of 125 I-labeled CS6 protein, and 125 I-labeled CS6 protein incubated with dextran sulfate, to pure glycosphingolipids in microtiter wells (C). doi:10.1371/journal.pone.0004487.g005 upon deletion of the cssB gene, the cell binding by the bacteria was not statistically different from that obtained with the background strain.
Although sulfatide is a common constituent among small intestinal glycosphingolipids of most species, its presence is not obligatory, since it is not found in the small intestine of e.g. mouse, rat, cod [18] or calves [19]. In mouse small intestinal epithelium, the major sulfo-glycosphingolipid is sulf-gangliotetraosylceramide (SO 3 -3Galb3GalNAcb4Galb4Glcb1Cer) [20], which, as we show here, is not recognized by the CS6 protein or CS6-positive E. coli.
A dose-dependent reduction of the CS6-mediated sulfatide binding was obtained with the highly sulfated polymer dextran sulfate, but not with dextran, demonstrating the importance of the sulfate group for binding to occur. Preincubation with heparin, galactose 4-sulfate, galactose 6-sulfate had no inhibitory effect, suggesting a specific recognition of the sulfate group in 3-position of galactose as in sulfatide, or glucose as in dextran sulfate. The recognition of sulfolactosylceramide (SO 3 -3Galb4Glcb1Cer) by CS6 protein and the CS6-expressing bacteria is in agreement with this suggestion.
However, sulfo-gangliotetraosylceramide (SO 3 -3Galb3GalNAcb4-Galb4Glcb1Cer) was not recognized by the CS6 protein or CS6positive bacteria, although it has a terminal SO 3 -3Gal. Thus, a SO 3 -3Galb3GalNAc sequence is not tolerated, suggesting a sterical hindrance caused by the N-acetylgalactosamine.
Phenotypic expression of CS6 is associated with expression of the heat-stable enterotoxin (ST-only and ST/LT strains), while it is rarely observed in ETEC strains expressing only the heat-labile enterotoxin [5,21]. There are two subtypes of the heat-stable toxin, designated STa and STb. The receptor for STa, which is associated with diarrhea in humans and animals, piglets, is the extracellular domain of a trans-membrane guanylate cyclase type C protein [22]. Interestingly, the STb subtype, which is primarily associated with diarrhea in piglets, binds to sulfatide [23]. The interaction of STb with sulfatide is not affected by dextran sulfate, in contrast to the inhibitory effect observed for the sulfatide binding of the CS6 protein.
The sulfatide binding of CS6-expressing ETEC differs distinctly from the glycosphingolipid specificity of CFA/I-fimbriated ETEC [15]. The CFA/I minor tip subunit CfaE mediates hemagglutination and binding to CaCo-2 cells by binding to an as yet unidentified receptor [32]. The major CfaB subunit of CFA/I binds to a number of non-acid glycosphingolipids, including glucosylceramide, lactosylceramide with phytosphingosine and/or hydroxy fatty acids, neolactotetraosylceramide, gangliotriaosylceramide, gangliotetraosylceramide, the H5 type 2 pentaglycosylceramide, the Le a -5 glycosphingolipid, the Le x -5 glycosphingolipid and the Le y -6 glycosphingolipid. These non-acid glycosphingolipids were also recognized by fimbriae with high sequence similarity to CFA/I, i.e. CS1 and CS4 [12], but not by the CS6 adhesin (present study). However, there is a large repertoire of CFs expressed by ETEC, and for the major part of these CFs the attachment sites are not yet defined. The mechanisms of adherence mediated by ETEC CFs thus merits further studies.

Bacterial strains, culture and labeling
Recombinant CS6 adhesin was expressed in the E. coli strain TOP10 as described by Tobias et al. [10], and the resulting recombinant strain was designated TOP10-CS6. The TOP10-CS6 strain was grown in CFA broth containing Casamino acids 10 g, yeast extract 1.5 g, MgSO 4 .7H 2 O 102 mg, and MnCl 2 .4H 2 O 8 mg per liter (pH 7.4), supplemented with ampicillin (100 mg/ml) at 37uC overnight with shaking. For metabolic labeling, 100 ml of this bacterial suspension was transferred to 10 ml of the same medium supplemented with 10 ml 35 S-methionine (400 mCi; Amersham Pharmacia Biotech), and grown for 2 h at 37uC with shaking. Thereafter, isopropyl-b-D-thiogalactopyranoside (IPTG; Saveen Werner AB, Malmö, Sweden) to the final concentration of 1 mM was added, and the bacteria were then incubated overnight. Bacteria were harvested, washed three times in phosphate-buffered saline (PBS, pH 7.3), then resuspended in PBS) containing 2% (w/v) bovine serum albumin, 0.1% (w/v) NaN 3 and 0.1% (w/v) Tween 20 (BSA/ PBS/TWEEN) to a bacterial density of 1610 8 colony forming units/ml. The specific activity of bacterial suspensions was approximately 1 cpm per 100 bacteria.
The same conditions (with omission of ampicillin and IPTG) were used for culture and labeling of the background E. coli strain TOP10.

Isolation of CS6 protein
The CS6 protein was isolated from the TOP10-CS6 recombinant strain. An overnight culture of TOP10-CS6, grown in CFA broth supplemented with 100 mg/ml of ampicillin, was diluted 1/ 100 in the same medium and incubated for 2 h at 37uC and 150 rev/min. IPTG was then added into the medium to the final concentration of 1 mM, and the bacteria were grown at 37uC and 150 rpm for 17 h. The bacterial culture was centrifuged twice at 100006g for 15 min. The supernatant was then filtered (0.2 mm Amicon; Millipore, Bedford, MA) to remove bacterial cells. The supernatant containing the CS6 was subjected to overnight ammonium sulfate precipitation (50% saturation) at 4uC, and the protein was recovered by centrifugation at 130006g for 30 min. The protein was dialyzed extensively against PBS and stored at 270uC.

Production and purification of CssA and CssB
DNA fragments encoding the mature CssA and CssB proteins were each amplified from the recombinant plasmid pJT-CS6 [10]. Each fragment was amplified so that it could be inserted directly into an appropriate expression vector. The primers for the amplification were: The amplified cssA fragment contained terminal BsaI and KpnI sites that allowed directional cloning into a modified pBAD/His plasmid [10] such that the resulting protein carried an amino terminus polyhistidine tag that could be removed by enterokinase digestion resulting in the native CssA protein. Protein expression was induced by growth of the recombinant TOP10 strain in the presence of 0.1% arabinose. This resulted in the accumulation of the recombinant protein as cytoplasmic inclusion bodies, since the signal peptide required for transport through the inner membrane to the periplasm had been removed. The cells were harvested by centrifugation and disrupted by lysozyme and DNase I treatment followed by sonication. The inclusion bodies were separated from the other cell debris by low speed centrifugation. Following extensive washing of the inclusion bodies they were finally dissolved in 6.5 M urea. The urea was removed by dialysis against 20 mM phosphate buffer, pH 7.4. When the urea had been removed the his tagged protein was purified by affinity chromatography using a nickel loaded His trap column (Amersham Pharmacia Biotech, Uppsala, Sweden) attached to an Ä KTA FPLC apparatus (Pharmacia). The protein was loaded onto the column in 20 mM phosphate buffer, pH 7.4, made 10 mM with respect to imidazole, essentially according to the instructions of the manufacturers. The protein was eluted using an imidazole gradient from 10 mM up to 500 mM. The peak containing the recombinant protein was identified by SDS-PAGE electrophoresis. Fractions containing the protein were pooled and dialyzed against 20 mM phosphate buffer, pH 7.4, in order to remove the imidazole. Purity of the resulting CssA product was estimated from SDS-PAGE to be greater than 95%. Protein concentration was determined by a modified Biuret assay.
Following cloning, CssB was not successfully expressed in the same system. However, the mature CssB protein was also fused to the 26 kDa glutathione-S-transferase (sj26GST) from Schistosoma japonicum using an expression vector constructed in this laboratory. The fusion protein was constructed such that cssB protein could be cleaved from sj26GST carrier in two ways. Using thrombin for cleavage, the CssB carries an amino terminus extension containing a polyhistidine tag. Cleavage with enterokinase, however, leads to removal of the native CssB protein from the sj26GST, carrier which retains the polyhistidine tag. This allows the carrier protein to be removed from the CssB using a chelating column charged with nickel ions, even if the glutathione-S-transferase (GST) is not active and cannot act as an affinity ligand. A strain carrying the recombinant plasmid was grown up under inducing conditions in which IPTG was added to the growth medium to a final concentration of 1 mM. As with CssA, recombinant protein accumulated in the cytoplasm of the bacteria as insoluble inclusion bodies. The inclusion bodies were isolated and dissolved as described above, and treated in exactly the same way as CssA, taking advantage of the polyhistidine tag and avoiding difficulties related to the reconstitution of active GST.

Electrophoresis
The protein preparations were separated on 10% or 12% NuPAGE BisTris gels (Life Technologies Inc., Carlsbad, CA) according to the instructions of the manufacturer, and stained by Coomassie Brilliant Blue R-250.

I-labeling
Aliquots of 100 mg of protein were labeled with 125 I, using Na 125 I (100 mCi/ml; Amersham Pharmacia Biotech, Little Chalfont, U.K.), according to the IODO-GEN protocol of the manufacturer (Pierce, Rockford, IL), giving approximately 5610 3 cpm/mg protein.

Reference glycosphingolipids
Total acid and non-acid glycosphingolipid fractions were prepared as described earlier [33]. Individual glycosphingolipids were isolated by repeated chromatography on silicic acid columns and by HPLC, and identified by mass spectrometry [34], and 1 H-NMR spectroscopy [35].

Thin-layer chromatography
Aluminum-or glass-backed silica gel 60 high performance thinlayer chromatography plates (Merck, Darmstadt, Germany) were used for thin-layer chromatography, and eluted with chloroform/ methanol/water (60:35:8, by volume) as solvent system. The different glycosphingolipids were applied to the plates in quantities of 1-4 mg of pure glycosphingolipids and 40 mg of glycosphingolipid mixtures. Chemical detection was done with anisaldehyde [36].