INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Intra-abdominal infections present serious clinical problems that are difficult to diagnose and treat, frequently require surgical intervention, and often result in complications, including death. Bacteroides fragilis accounts for only 0.5% of the normal human colonic flora; however, it is the anaerobic species most frequently isolated from clinical infections, particularly within the abdominal cavity (16, 37). The capsular polysaccharides of B. fragilis are the most important virulence factors in the formation of intra-abdominal abscesses by this organism, and purified capsule is able to induce abscesses in experimental animal models (27, 31).

The B. fragilis capsule is composed of two distinct high-molecular-weight polysaccharides, termed polysaccharide A (PS A) and polysaccharide B (PS B) (34), that are coexpressed (56). The chemical composition and structure of each polysaccharide have been determined for one B. fragilis strain, NCTC 9343 (56). PS A is composed of the following repeating unit: [3)-D-AATp(14)[-D-Galf(13)]-D-GalpNAc(13) -D-Galp(1], where AAT is 2-acetamido-4-amino-2,4,6-trideoxygalactose. A pyruvate substituent having the R configuration spans O-4 and O-6 of the -D-galactopyranosyl residue. PS B is composed of the following repeating unit: [4)-L-QuipNAc(13)-D-QuipNAc(14)[-L-Fucp(12) -D-GalpA(13)-D-GlcpNAc(13)]-D-Galp(1], with a 2-aminoethylphosphonate substituent on O-4 of the N-acetyl--D-glucopyranosyl residue. The most striking feature of PS A and PS B is the presence of both positively and negatively charged groups on each repeating unit. Of the many bacterial capsular polysaccharides whose structures have been determined, relatively few contain both positively and negatively charged substituent groups. The charge motif of these polysaccharides is essential for abscess formation (54).

The capsule of B. fragilis initiates a unique immune response in the host: the formation of abscesses. This process is dependent upon T cells (46, 53) and therefore is distinct from responses to most other polysaccharide antigens, which are considered to be T cell independent. Studies in our laboratory have shown that the B. fragilis capsule acts as an immunomodulator regulating the formation of intra-abdominal abscesses (53).

Despite the importance of B. fragilis capsular polysaccharides as virulence factors, as unique immunomodulators, and as interesting bacterial molecules with a rare charge motif, nothing is known about the genetics of their biosynthesis or regulation. This study describes the characterization of the first such locus of B. fragilis.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacteria, plasmids, and media. Bacterial strains and plasmids are described in Table 1. Escherichia coli strains were grown in Luria broth or on Luria agar plates. B. fragilis strains were grown anaerobically in supplemented brain heart infusion (BHI) broth (BHIS; Randolph Biomedical, West Warwick, R.I.) or on BHI agar plates supplemented with hemin (50 µg/ml) and menadione (0.5 µg/ml). The following concentrations of antibiotic supplements were added when appropriate: ampicillin, 100 µg/ml; kanamycin, 20 µg/ml; erythromycin, 7.5 µg/ml; trimethoprim, 100 µg/ml; gentamicin, 200 µg/ml.

Transposon mutagenesis and screening. Plasmid R7514 has been used previously to create transposon insertions in the chromosomes of Bacteroides species; the result has often been the insertion of the entire plasmid (47). Because R7514 is a large plasmid, its integration was not advantageous for these studies. Therefore, the Tn4351 tandem transposon was excised from R7514 with SalI and ligated into the unique SalI site of the Bacteroides suicide vector pNJR6, resulting in pNJR64. This plasmid is smaller and, when integrated with the transposon, permits cloning of the Bacteroides DNA at the junction of the transposon with various restriction enzymes by using the E. coli origin of replication and the Km^r gene of pNJR6. E. coli DH5 containing pNJR64 and the conjugal helper plasmid R751 was mated with B. fragilis 638R by a modification of a previously described protocol (49). B. fragilis was grown to an optical density at 550 nm of 0.2, centrifuged, and mixed with DH5 containing pNJR64 and R751. These strains were mated aerobically overnight on BHIS plates and then plated onto BHIS plates containing erythromycin and gentamicin and incubated anaerobically. The resulting colonies were screened by immunoblotting with monoclonal antibody (MAb) 4D5 (638R capsule specific).

Cloning the DNA at the transposon junction of mutant 2-42. The chromosome of mutant 2-42 (MAb 4D5 negative) yielded a 15-kb HindIII fragment that contained the origin of replication and the Km^r gene of pNJR6 as well as ~900 bp of B. fragilis DNA. This fragment was cloned by ligation of HindIII-digested 2-42 chromosomal DNA (favoring intramolecular ligations), transformation of DH5, and selection with kanamycin. The resulting plasmid (pLEC6) has a very low copy number; therefore, the 900-bp B. fragilis DNA of pLEC6 was subcloned into pBluescript II SK (Stratagene, La Jolla, Calif.) by using the EcoRI site at the end of the transposon and the HindIII site at the end of the cloned B. fragilis DNA; the result was plasmid pLEC6.1. The insert of pLEC6.1 was sequenced with primers complementary to vector sequences.

Construction of plasmids pMJC2 to pMJC2.6. Plasmid pMJC2 was selected from a B. fragilis 9343 gene bank with the insert of pLEC6.1 as a probe. The gene bank was constructed with Sau3AI partial digests of chromosomal DNA cloned into the BamHI site of pHC79. For construction of pMJC2 subclones (pMJC2.1 to pMJC2.6), restriction fragments of pMJC2 were separated on agarose gels and eluted with a Qiagen gel extraction kit (<10 kb) or electroeluted (>10 kb). All subclones except pMJC2.5 were cloned into pBluescript II SK. The restriction sites used and the sizes and locations of the subclones constructed are illustrated in Fig. 1 and Table 1. Plasmid pMJC2.5 is an EcoRI fragment of pMJC2 which was circularized and utilizes the -lactamase gene and origin of replication of pHC79.

View larger version (23K): [in this window] [in a new window]   FIG. 1.   ORF map of a 24,454-bp region of the 9343 chromosome contained on cosmid pMJC2. Restriction sites within this region are shown for the following enzymes: PvuII (P), EcoRI (E), Asp718 (A), StuI (St), SmaI (Sm), SalI (Sl), MluI (M), and Sau3AI/BamHI junction with the pHC79 vector (Sa/B). wcfD-L and wcfF represent the regions that were deleted in the chromosomes of mutants 9343wcfD-L and 9343wcfF. The arrows indicate direction of transcription of each ORF. The ORFs that do not demonstrate homology with genes involved in polysaccharide biosynthesis (outside the wcf locus) are shaded. The lower diagram demonstrates the average G+C content for each ORF of the region. The dashed-line boxes represent a clustering of adjacent ORFs with similar G+C contents. A transcriptional terminator-like stem-loop region downstream of wcfL is indicated.

Creation of deletion mutants 9343wcfD-L and 9343wcfF. Mutant 9343wcfD-L is devoid of 10 of the 16 genes in the wcf region, beginning at the StuI site in wcfD and continuing to the SmaI site just downstream of the wcf locus in orf5 (Fig. 1). To create this mutant, pMJC2 was digested with PvuII and StuI, and the 9.2-kb fragment representing bp 3 through bp 9277 was recovered. pMJC2 was separately digested with SmaI and PvuII, and the 7.3-kb fragment representing bp 19,823 through bp 27,125 was eluted. The PvuII site of the 7.3-kb SmaI-PvuII fragment lay within the pHC79 cloning vector, and this 7.3-kb fragment contained the ori and the Ap^r gene of pHC79. The 9.2-kb fragment was ligated with this 7.3-kb fragment, and the ligation was transformed into DH5 and plated with ampicillin selection. The resulting colonies were tested by PCR for the correct ligation of the two PvuII ends and the StuI end of the 9.2-kb fragment with the SmaI end of the 7.3-kb fragment. The correct clone, designated pMJC2, was linearized with PvuII and cloned into the unique StuI site of the mobilizable Bacteroides suicide vector pNJR6, creating pMJC2.1. Plasmid pMJC2.1 was transferred into the chromosome of wild-type 9343 with the conjugal helper plasmid R751, and cointegrates were selected by Em^r encoded by the suicide vector. The cointegrate was passaged three times in BHIS to allow for resolution of the cointegrate and plated onto BHIS. The resulting colonies were replica plated onto BHIS containing erythromycin, and the Em^s colonies were tested for the mutant genotype by PCR. The loss of the 10.5-kb region of wcfD to -L in the 9343 chromosome was confirmed by Southern blotting. This mutant was designated 9343wcfD-L.

DNA sequencing and analysis. DNA sequencing was performed with an automated DNA sequencer (Prism model 377; Applied Biosystems, Inc.) with AmpliTaq FS DyeDeoxy terminator cycle sequencing kits (Applied Biosystems) at the Automated Sequencing and Genotyping Center at Brigham and Women's Hospital.

SDS-polyacrylamide gel electrophoresis and Western blotting (immunoblotting). Bacterial cell lysates were run on discontinuous sodium dodecyl sulfate (SDS)-polyacrylamide gels (10% acrylamide; ESA, Inc. Chelmsford, Mass.) and transferred to nitrocellulose by standard techniques. The blots were blocked in Tris-buffered saline (TBS) containing 5% nonfat dry milk (TBS-milk) and then incubated for 1 h in TBS-milk containing primary antibody. The blots were washed with TBS and incubated with alkaline phosphatase-conjugated goat anti-rabbit or anti-mouse immunoglobulin G (Sigma Chemical Co.) at 1:1,000 in TBS for 1 h. The blots were washed with TBS and developed with a colorimetric detection solution. Ascitic fluid containing MAb CE3 (9343 PS A specific) was used at 1:1,000. Culture supernatant containing MAb 4D5 (638R capsule specific) was used at 1:10.

Capsular polysaccharide extraction, purification, and analysis. A 16-liter culture of B. fragilis 9343wcfD-L was grown in a fermentor, and the capsular polysaccharide was extracted with phenol and water as previously described (34). The resulting material was resuspended in 274 ml of deoxycholate buffer (3% sodium deoxycholate, 50 mM glycine, 10 mM EDTA [pH 9.8]). One-quarter of the sample was applied to a Sepharose S400 column equilibrated with deoxycholate buffer. This procedure completely separated lipopolysaccharide from capsular polysaccharide as determined by silver-stained SDS-polyacrylamide gels. Fractions containing capsular material were pooled, concentrated, precipitated, resuspended in 20 ml of distilled H₂O (dH₂O) (pH 9.0), and dialyzed against pH 9.0 dH₂O for 48 h to remove the detergent. The material was then dialyzed against pH 7.0 dH₂O and lyophilized, and the total weight of the capsular polysaccharide was determined. The capsular material was resuspended in 50 mM Tris (pH 7.4) and applied to an anion-exchange column (Q-Sepharose; Pharmacia). The column was washed with 2 bed volumes of buffer and eluted with a linear gradient to 2 M NaCl. Column fractions were analyzed by immunoelectrophoresis as described previously (34). Fractions appearing identical by immunoelectrophoresis were pooled, concentrated, dialyzed, lyophilized, and analyzed by nuclear magnetic resonance to confirm the identity of the polysaccharide.

Competitive ELISA inhibition. A competitive enzyme-linked immunosorbent assay (ELISA) was performed as previously described (35). Immulon-2 96-well plates (Dynatech) were coated with purified PS B and frozen at 20°C until use. Antigens to be tested were dissolved at a concentration of 100 µg/ml in phosphate buffer, diluted in triplicate, and added to PS B-coated plates. Polyclonal rabbit antiserum enriched for antibodies to PS B was added to all wells at a 1:4,000 dilution. The plates were thoroughly washed, and a goat anti-rabbit horseradish peroxidase conjugate was added for 1 h. After thorough washing, substrate was added for 1 h. Absorbance was read at 405 nm. Percent inhibition calculations for each antigen were based on comparisons with uninhibited control wells.

Rat abscess model. An animal model for intra-abdominal sepsis was utilized for these studies (30). Male Wistar rats (180 to 200 g; Charles River Laboratories, Wilmington, Mass.) were anesthetized with a single intraperitoneal injection of 0.15 ml of pentobarbitalsodium (Nembutal) (50 mg/ml; Abbott Laboratories, North Chicago, Ill.). The inocula, containing serial 10-fold dilutions of B. fragilis 9343 or 9343wcfF, were mixed with sterile rat cecal contents and 10% BaSO₄ (wt/vol) as previously described (32, 53). An anterior midline incision (0.5 cm) was made through the abdominal wall and peritoneum, and 0.5 ml of inoculum was inserted into the pelvis with a pipette. The incisions were closed with 3.0 silk sutures. The animals that did not survive the surgical procedure were removed from the experiment. Six days postchallenge, the surviving animals were necropsied in an observer-blinded fashion and examined for the presence of intra-abdominal abscesses. The presence of one or more abscesses in an animal was scored as a positive result. The results are reported as the number of animals with one or more abscesses per total number of animals. The number of organisms necessary to cause abscesses in 50% of the animals (AD₅₀) was calculated by the method of Reed and Muench (39).

Nucleotide sequence accession numbers. The sequence discussed in this paper has been assigned GenBank accession no. AF048749.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Cloning of a polysaccharide biosynthesis locus of 9343. The prototypical B. fragilis strain used to study abscess formation, NCTC 9343, is somewhat resistant to the introduction of foreign DNA; thus, transposon mutagenesis of this strain is difficult. To locate genes involved in capsular polysaccharide biosynthesis in strain 9343, transposon mutants were generated with B. fragilis 638R. MAb 4D5 (638R capsule specific [33]) was used to screen the mutant bank. The DNA at the junction of the transposon insertion of mutant 2-42 (4D5 negative) was homologous to rmlA, a gene known to be present in various bacterial polysaccharide biosynthesis loci. This rmlA probe hybridized with 9343 chromosomal DNA; therefore, this region was further studied in the 9343 chromosome. The 638R rmlA probe (pLEC6.1 insert) was used to select the 9343 gene bank clone pMJC2 (Fig. 1). This cosmid clone was mapped with restriction enzymes and subcloned as diagrammed in Fig. 1. The subclones were used as sequencing templates. In total, 24,454 bp were sequenced. Analysis of this sequence revealed 23 complete ORFs and two partial frames at the ends. The central 15,379 bp of this sequence was found to contain 16 ORFs, all of which encode putative products with varying degrees of similarity to proteins implicated in polysaccharide biosynthesis in other organisms (see Table 3). In accordance with changes in the nomenclature of polysaccharide biosynthesis genes (40), the ORFs were named wcfA through wcfL. Four genes were given previously designated names due to a high degree of homology or conserved sequence characteristics: rmlA, rmlC, wzx, and wzy.

Phenotype conferred by the wcf locus. Because the transposon insertion of mutant 2-42 was in strain 638R rather than 9343, and because MAb 4D5 does not react with 9343, no phenotype was ascribed to the wcf locus. Therefore, to determine which polysaccharide is synthesized by this region, 10.5 kb of wcf were deleted from the 9343 chromosome, removing the last 10 genes of the region (Fig. 1). The resulting strain, 9343wcfD-L, was then tested for phenotype. MAbs that react with 9343 PS A are available; however, no MAbs that are specific to 9343 PS B are available. The MAb G9, which was previously shown to react with PS B (33), also cross-reacts slightly with purified PS A. The PS A-specific MAb CE3 (34) reacted with 9343wcfD-L (data not shown). Given this result, as well as the many studies showing that 9343 does not express an O antigen and is therefore a rough strain (24, 57), it seemed most likely that this region synthesizes PS B.

View larger version (38K): [in this window] [in a new window]   FIG. 2.   Western immunoblot of cell lysates of B. fragilis 9343 and mutants probed with 9343wcfD-L-adsorbed antiserum. The location of the 182-kDa marker is shown.

Abscess induction by PS B mutant. To test the effect of loss of PS B on the ability of 9343 to cause abscesses in the animal model, a second mutant was created (9343wcfF) that is mutated in only one gene, wcfF. This mutant was created to ensure that only PS B was lost from the mutant strain. The 9343wcfD-L mutant was lacking wcfH, a gene whose product may be involved in synthesis of the free amino group of PS A. In addition, orf5 was mutated in the large deletion strain, and the product of this gene is undefined. WcfF is very similar to dehydrogenases, which would be necessary for the synthesis of PS B but not PS A. The 9343wcfD-L-adsorbed antiserum is unable to react with 9343wcfF (Fig. 2), demonstrating that the 9343wcfF mutant contains no immunoreactive molecule compared to 9343wcfD-L, which has been shown to be devoid of PS B. When 9343wcfF and the wild type were compared in the animal model, the ability of the mutant to cause abscesses was not attenuated (Table 2). The AD₅₀s were calculated to be 3.9 × 10⁵ for the wild type and 3.5 × 10⁵ for the mutant.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The importance of the capsular polysaccharides of B. fragilis in the formation of abscesses has been recognized for decades (31). The two capsular polysaccharides of the prototype strain NCTC 9343 have been purified, the structures have been determined (56), and the charge motif necessary for abscess formation has been well studied (54, 55). The host response to these polysaccharides leading to the formation of abscesses is beginning to be understood at the molecular level (14). Despite these advances, this is the first description of the cloning and sequencing of a B. fragilis capsular polysaccharide biosynthesis locus.

We have shown that the wcf region is necessary for the synthesis of PS B. To determine the contribution of PS B to abscess formation in the natural context of the bacterial cell surface, a defined mutation that abolished the synthesis of PS B was made. Animal studies showed that this mutant was able to cause abscesses in the animal model at the same level as the wild type. This result is not surprising, considering that each of the B. fragilis 9343 capsular polysaccharides has been demonstrated to be a potent abscess-inducing polysaccharide. In the rat intra-abdominal abscess model, the AD₅₀ of purified PS A is 0.67 µg, in contrast to that of purified PS B (AD₅₀ = 25 µg) or capsular polysaccharide complex (AD₅₀ = 22 µg). Although PS B is not necessary for abscess formation when PS A is present, it is likely that synthesis of either PS A or PS B alone is sufficient to cause abscesses, although a PS A mutant may be somewhat attenuated. It is possible that abrogation of both polysaccharides will be necessary to completely attenuate abscess formation by B. fragilis.

The wcf locus is typical of other bacterial polysaccharide biosynthesis loci in regard to gene composition, G+C content, and genetic organization. Putative functions for some of the gene products encoded by the region are assigned based on similarity to other known products.

Genes upstream and downstream of wcf. It is unclear whether either of the two small ORFs upstream of rmlA (orf3 and orf4) is involved in the synthesis of PS B. These ORFs exhibit no significant homology to other sequences, and they are relatively small. The DNA upstream of orf3 is intergenic, with no ORF of significant size in either direction for 892 bp. Upstream of this gap is a 636-bp ORF that is similar over its central portion to S-adenosyl-sterol-C-methyltransferases from plant species (6) and to methyltransferases involved in menaquinone biosynthesis in bacteria (25).

rml genes. The first two genes of the wcf locus, rmlA and rmlC, are easily recognized by the high degree of similarity of their predicted products to other glucose-1-phosphate thymidyltransferase (RmlA) and dTDP-4-keto-6-deoxy-D-glucose 3,5 epimerase (RmlC) protein sequences (Table 3). The products encoded by rmlA and rmlC may not be involved in the synthesis of PS B. These highly conserved proteins are usually described as rhamnose biosynthesis enzymes, but neither of the structurally characterized B. fragilis 9343 capsular polysaccharides contains rhamnose as a substituent component. Homologues of some or all of the four rml genes (rmlA, -B, -C, and -D) are frequently found at the margins of polysaccharide biosynthesis loci.

Gene products involved in the formation of nucleotide sugar precursors (WcfF and WcfK). wcfF encodes a 425-amino-acid (aa) product that is similar to 6-dehydrogenase enzymes of other bacterial polysaccharide loci (Table 3). The amino-terminal portion of WcfF contains the 28-aa consensus NAD-binding domain found in enzymes of this class (9, 59). The gene product of wcfF may be responsible for the synthesis of UDP-galacturonic acid (the only uronic acid of PS B) from UDP-galactose.

Sugar transferase genes (wcfB, wcfC, wcfE, wcfG, wcfI, wcfJ, and wcfL). The predicted product of wcfL is similar to several Rfe proteins, which catalyze the transfer of an N-acetylglucosamine residue to undecaprenylphosphate as the initial step in the synthesis of enterobacterial common antigen and of some E. coli lipopolysaccharide O antigens (41). WcfL displays a hydrophobicity profile indicative of an integral membrane protein. It is likely that WcfL transfers the first monosaccharide to a lipid carrier in the synthesis of B. fragilis PS B.

Polymerization and export genes (wzy and wzx). The characteristics of low G+C content, multiple membrane-spanning domains, and sequence similarity were used to putatively identify the genes encoding Wxy and Wzx within the wcf locus. The eighth gene of the operon, wzy, has a G+C content of 26.9% (the lowest of the region) and encodes a protein of 365 aa. This putative protein is 44% similar to the O-antigen polymerase of Salmonella enterica C1 (26) and is extremely hydrophobic, with nine potential membrane-spanning domains.

Acetyltransferase genes (wcfA and wcfD). The gene products of wcfA and wcfD are both similar to galactoside-O-acetyltransferases of the CysE-LacA-NodL family, which acetylate a variety of substrates. A motif has been described for this family of acetyltransferases that has 21 conserved amino acids over a 31-aa stretch (11). WcfA aligns with this consensus sequence with just two mismatches, and WcfD aligns with five mismatches. Although the published structure of PS B does not contain O-acetylated sugars, some lots of purified PS B have been shown to be O acetylated to varying degrees (56a), which may explain the presence of these genes in the wcf locus.

WcfH, a putative deacetylase. The presence of both positively and negatively charged groups on the repeating-unit structures of PS A and PS B has been shown to be critical to the virulence of the B. fragilis capsular polysaccharides (54). The positively charged free amino group of PS A may be formed by removal of an acetyl group from an N-acetylated precursor sugar by a deacetylase. WcfH is 40% similar over half of the protein to NodB of Rhizobium meliloti. NodB proteins are deacetylases involved in the hydrolysis of the N-acetamido group of N-acetyl-D-glucosamine during the formation of the lipo-chitin-oligosaccharides (21). There is ample precedent for a gene of one polysaccharide biosynthesis locus encoding a product that is involved in the synthesis of a different polysaccharide (3, 17, 23, 58). Due to the rare occurrence of genes encoding deacetylase-like proteins in polysaccharide biosynthesis loci and the importance of the free amino group to the abscess-inducing capabilities of both PS A and PS B, future studies will include analysis of the enzymatic activity of WcfH.