INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Bacteroides fragilis is the gram-negative strictly anaerobic bacterium most frequently isolated from clinical infection. The major source of these infections is the normal resident colonic microbiota, where Bacteroides spp. outnumber facultatively anaerobic bacteria such as Escherichia coli by a factor of between 10² and 10³ (4). In the fecal microbiota, the predominant Bacteroides spp. is B. vulgatus, with B. fragilis a relatively minor component. Namavar and colleagues (6) report a relatively higher proportion of B. fragilis in the adherent mucosal microbiota; however, this was not confirmed by Poxton and colleagues (19). It therefore appears that the frequency with which B. fragilis is isolated from infection compared to other Bacteroides spp. of the resident microbiota cannot be explained simply by weight of numbers.

The potential virulence determinants of B. fragilis have been the subject of many investigations (9). It is clear that a number of factors may contribute to the virulence of B. fragilis, including surface structures, release of extracellular enzymes, iron-scavenging mechanisms, and enterotoxin production; however, extracellular polysaccharides have been considered to play a key role in B. fragilis virulence. Encapsulating structures have been implicated in resistance to complement-mediated killing, phagocytic uptake, and killing (21) and abscess formation in an animal model (27). Many studies have failed to take into account not only within-strain variation in capsule production but also between- and within-strain antigenic variation of different types of capsules. By electron microscopy, it is possible to identify within an individual strain of B. fragilis bacteria with either large or small capsules which are fibrous in appearance but are antigenically different, as well as bacteria with an encapsulating electron-dense layer (EDL) adjacent to the outer membrane (15, 16). The EDL bacteria are noncapsulate by light microscopy, whereas the small and large capsules are clearly visible with negative staining. Expression of the different capsular types is inheritable as populations can be enriched by subculture from different interfaces of Percoll step density gradients. Microscopical observation of the populations enriched for the three capsular types with monoclonal antibodies (MAbs) specific for surface polysaccharides shows that noncapsulate bacteria are antigenically different from bacteria with the small capsules but have shared epitopes with large-capsule bacteria. In addition, immunofluorescent and immunogold labelling for fluorescence and electron microscopy, respectively, reveals antigenic variation in populations which appear to be structurally homogeneous (5, 13, 22, 23). This phenomenon has been observed in recent clinical isolates from a variety of anatomical sites, in isolates from different geographical locations, and in culture collection type cultures (17). By polyacrylamide gel electrophoresis (PAGE) and immunoblotting with MAbs specific for surface polysaccharides, distinctive patterns are observed within the noncapsulate population of an individual strain. These results indicate that an individual B. fragilis strain may produce a number of antigenically different surface polysaccharides (5, 9, 10).

The aim of the present study was to investigate intrastrain variation in B. fragilis populations which were homogeneous with respect to encapsulation. The stability of expression of individual polysaccharide epitopes within B. fragilis populations which are noncapsulate by light microscopy (EDL enriched) was therefore examined.

We now report that populations already enriched for capsule type can also be enriched for expression of individual surface polysaccharide epitopes in both broth and plate cultures.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and culture methods. The strains used were B. fragilis NCTC 9343 (National Collection of Type Cultures, London, United Kingdom), LS66 and LS54 (clinical isolates from an abdominal abscess and a perianal abscess respectively; Craigavon Area Hospital, Northern Ireland, United Kingdom) (17), and JC17 (clinical isolate from an abscess; Belfast City Hospital, Northern Ireland, United Kingdom). All strains were enriched on Percoll density gradients for populations which were noncapsulate by light microscopy. Bacteria were grown in defined minimal medium (DM) broth or on DM plates (28) in an anaerobic cabinet (Mk. III anaerobic cabinet; 80% N₂, 10% CO₂, and 10% H₂; Don Whitley Scientific, Shipley, United Kingdom). Identification was confirmed with the API20A (Biomérieux, Marcy L'Etoile, France) system.

Production and characterization of polyclonal antisera and MAbs. Polyclonal antiserum specific for B. fragilis NCTC 9343 common antigen was produced as previously described (17). Polyclonal antiserum specific for B. fragilis NCTC 9344 common antigen was the kind gift of I. Poxton, University of Edinburgh. MAb production and initial characterization of some of the MAbs are detailed in reference 5. MAb QUBF 12 does not cross-react with E. coli but does cross-react with Bacteroides thetaiotaomicron, B. vulgatus, and B. ovatus. The other MAbs do not cross-react with E. coli, B. vulgatus, B. distasonis, B. ovatus, B. thetaiotaomicron, or Porphyromonas gingivalis.

Separation and enrichment of bacterial populations. Bacterial populations which were noncapsulate by light microscopy were enriched by subculture from the 60 to 80% interface layer of Percoll (Pharmacia, Uppsala, Sweden) discontinuous density gradients after centrifugation as previously described (11). Encapsulation was monitored by eosin-carbol fuchsin negative staining for light microscopy (1). When required, these populations were further enriched by using immunomagnetic beads as detailed below. Dynal M-280 magnetic beads precoated with sheep anti-mouse immunoglobulin [Dynal (UK) Ltd., Merseyside, United Kingdom] were used as described by Patrick and Larkin (11). In brief, the beads were washed by placing the tube containing the beads in a magnetic particle concentrator and removing the PBS by pipette. Washed beads were suspended in hybridoma culture supernatant containing the relevant antibody, incubated at room temperature with gentle rocking for 24 h, and again washed in PBS. The following steps were all carried out inside an anaerobic cabinet to ensure bacterial viability. The beads were incubated with bacterial suspension (10⁸ CFU/ml) in DM plus 0.02% (vol/vol) Tween 20 (Sigma Chemical Co. Ltd., Poole, United Kingdom) with gentle rocking for 2 min and washed three times in DM. The beads were then inoculated into DM and incubated at 37°C in the anaerobic cabinet until bacterial growth was visible.

Immunofluorescence microscopy. Bacterial suspensions in PBS were applied to multiwell microscope slides, dried at 37°C, and fixed either in methanol at 20°C for 10 min or in paraformaldehyde (4% [wt/vol] in 3× PBS [390 mM NaCl, 30 mM Na₂HPO₄, 30 mM NaH₂PO₄ · 2H₂O {pH 7.2}) at 4°C for 30 min (20). For single labelling, the bacteria were then reacted with a suitable dilution of concentrated mouse MAb hybridoma supernatant followed by goat anti-mouse immunoglobulin G (heavy and light chain) conjugated to fluorescein isothiocyanate (FITC; Sigma) as previously described (11). For dual labelling, after incubation of the slides with MAb, they were washed, incubated with anti-B. fragilis common antigen polyclonal rabbit antiserum (17), washed, and incubated with sheep anti-rabbit FITC and goat anti-mouse tetramethyl rhodamine isothiocyanate (TRITC; Sigma) before a final wash (11). Slides were examined with a Leitz fluorescence microscope. An estimate of the proportion of bacteria fluorescing was obtained either by comparing FITC-labelled bacteria with bright-field phase-contrast view or by comparing populations dual labelled with anti-rabbit polyclonal antiserum and an anti-rabbit FITC conjugate and with mouse MAb and an anti-mouse TRITC conjugate. All estimations of percentage labelling involved counting a minimum of 100 bacterial cells per well.

Colony lifts and immunoreaction. For colony lifts, total viable counts were carried out after serial 10-fold dilution in Ringer's solution (25% [wt/vol]) with cysteine (0.05%) and spread plating onto DM agar. Unless stated otherwise, the plates were incubated for 48 h. Plates with approximately 150 or fewer colonies were chosen and processed as follows. Discs of nitrocellulose (Millipore UK Ltd., Watford, United Kingdom) were gently applied to the agar plate and lifted once the whole nitrocellulose disc appeared to be wet. A maximum of three lifts were carried out on each plate. For sterile lifts, the nitrocellulose discs were autoclaved prior to use. The nitrocellulose discs were then air dried at room temperature (approximately 22°C) under aerobic conditions for a minimum of 1 h, blocked with dried milk (5% [wt/vol]; Marvel, Chivers Ireland Ltd., Dublin, Ireland) in PBS with Tween 20 (0.05%) for 1 h at 37°C, washed five times rapidly in PBS-Tween followed by five 5-min washes, and either allowed to dry and stored at room temperature in sealed polythene bags for later immunoreaction or used immediately. For immunoreaction, the discs were incubated with the appropriate mouse MAb diluted in PBS for 1 h at 37°C with gentle rocking and washed rapidly five times in PBS-Tween followed by 5-min washes with gentle rocking. The discs were then incubated for 1 h as described above with goat anti-mouse alkaline phosphatase conjugate (Bio-Rad Laboratories Ltd., Hemel Hempsted, United Kingdom). The discs were washed as before except that the final wash was carried out in Tris buffer (50 mM Tris HCl [pH 9.4]) prior to incubation in the substrate (Bio-Rad alkaline phosphatase substrate kit; p-nitroblue tetrazolium chloride and 5-bromo-4-chloro-3-indolylphosphate) according to the manufacturer's instructions. Once the substrate color had reached sufficient intensity, the reaction was stopped by immersion in distilled water and the nitrocellulose was blotted dry with filter paper. Control reactions in which the MAb was replaced with PBS were also processed. Colonies from the control reaction appeared light grey in color, which may have been due to background bacterial alkaline phosphatase activity; however, positive colonies were obviously bright purple-blue. Positive colonies were counted and compared with the initial colony count of the agar plate.

Protocol for determination of antigen expression in broth and plate cultures. For estimates of the proportion of bacteria expressing an epitope in broth culture, four replicate 100-ml volumes were inoculated and incubated at 37°C in the anaerobic cabinet until they had reached an optical density at 600 nm of 0.3 (equivalent to approximately 3 × 10⁸ CFU/ml). The proportion of bacteria within each broth which labelled with a given MAb was estimated by immunofluorescence microscopic examination of 20 separate-microscope slide wells for each broth. A minimum of 100 bacteria were counted in each of the 20 wells.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Epitope expression in broth culture. Immunomagnetic bead separation and broth enrichment with MAbs QUBF 6 and 7, specific for high-molecular-mass polysaccharide with an associated fine ladder pattern, and MAb QUBF 12, which recognizes an antigen similar in molecular mass to the B. fragilis common antigen (Fig. 1), was successful as assessed by immunofluorescence microscopy. The proportions of bacteria within a population which expressed the various epitopes could be increased from 15 to 76% (QUBF 6), from 21 to 74%, (QUBF 7), and from 22 to 56% (QUBF 12) with two subsequent immunomagnetic bead separations and broth enrichments. The degree of enrichment obtained on different occasions was not, however, always consistent. On a separate occasion, the QUBF 6 epitope was increased only to 32% after two enrichment steps.

View larger version (104K): [in this window] [in a new window]   FIG. 1.   Immunoblots of hot phenol-water extracts from B. fragilis NCTC 9343 after PAGE reacted with rabbit anti-B. fragilis NCTC 9343 common antigen antiserum (track 1), rabbit anti-B. fragilis NCTC 9344 common antigen antiserum (track 2), and MAbs QUBF 12 (track 3), QUBF 5 (track 4), QUBF 6 (track 5), QUBF 7 (track 6), and QUBF 8 (track 7).

Stability of QUBF 6 and 7 epitope expression in B. fragilis noncapsulate populations. The reactivity of MAbs QUBF 6 and 7 which are specific for high-molecular-mass polysaccharide with an associated fine ladder pattern was examined on pre- and post-immunomagnetic bead-enriched cultures of B. fragilis NCTC 9343. The proportion of bacteria labelling within replicate broth cultures was determined by immunofluorescence microscopy, and the proportion of colonies which labelled on spread plates prepared from the broth cultures was determined by colony blotting. The proportion of colonies which were positive was similar to that of the proportion of individual bacterial cells positive by immunofluorescence microscopy in the original broth culture (Table 1). Similar results were obtained for three other strains of B. fragilis (Table 2), although the proportion of bacteria that labelled was not the same for different strains labelled with the same MAb. The failure to detect any positive colonies by immunofluorescence microscopy after random selection of 20 colonies per plate, where it was estimated by colony lifts that 14% of the colonies were positive, is probably due to the small sample size.

View larger version (102K): [in this window] [in a new window]   FIG. 2.   Light micrographs of B. fragilis NCTC 9343, prepared from single colonies, immunolabelled with both mouse MAb plus anti-mouse TRITC conjugate and rabbit anti-B. fragilis polyclonal antiserum plus anti-rabbit FITC conjugate viewed (100× objective) with fluorescein filters (a) and the same field viewed with rhodamine filters (b). (i) Colony labelled with MAb QUBF 6 in which 95% or more of the bacteria are labelled; (ii) another colony labelled with MAb QUBF 6 in which only a small proportion of the total bacterial population is labelled.

Cross-reactivity of epitope-enriched populations with other B. fragilis-specific MAbs. Populations enriched using immunomagnetic beads coated with either QUBF 6 (specific for high-molecular-mass polysaccharide) or QUBF 12 (specific for a band similar in molecular mass to the common antigen) and subcultured in DM broth were examined for reactivity with other MAbs by immunofluorescence microscopy (Table 4), as were populations enriched by subculture of colonies immunoblot positive for QUBF 6 and 7 (specific for high-molecular-mass polysaccharide) into DM broth (Table 5). Reactivity of the enriched populations with MAbs showed that with none of the MAbs tested was the proportion of bacteria labelled as great as that labelled with the MAb for which the population had been enriched.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The results clearly illustrate intrastrain antigenic variation of the surface polysaccharides of B. fragilis. The epitopes recognized by the MAbs are present on distinct bacterial cells and are detectable on highly variable proportions of bacterial cells within individual natural populations as determined by immunofluorescence and immunoelectron microscopy (9).

The results show conclusively that the proportion of bacterial cells within a population which express these variable polysaccharides can be enriched by using immunomagnetic beads coated with the relevant MAb, followed by broth culture. The lack of reproducibility in the level of enrichment obtained by using immunomagnetic beads on different occasions was probably due to varying numbers of outer membrane vesicles (14) attaching to the magnetic bead bound antibodies and thus preventing bacterial attachment.

The proportion of bacteria which were labelled with MAbs QUBF 6 and 7, which both label polysaccharide in the high-molecular-mass region and an associated fine ladder pattern, was different within the same population of a given strain (Tables 1 and 2). For example, in two strains 50% or more of the bacteria were labelled with MAb QUBF 7, but very few or none of the cells labelled with MAb QUBF 6. This finding suggests that although these polysaccharides appear to be immunochemically similar after PAGE and immunoblotting, they are antigenically different and that the number of bacteria within a population which express these epitopes varies from strain to strain. The level of expression of MAbs QUBF 6 and 7 within individual colonies was, however, strikingly reproducible. Examination of individual colonies, picked from an agar plate, showed either a high proportion (>94%) or low proportion (<7%) of bacteria labelled for all the strains examined. These proportions were maintained on subculture of individual colonies into broth. This is similar to the phenomenon of enrichment for expression of different capsule types after Percoll density gradient centrifugation (9), which is also maintained on subculture into broth. The relationship between the number of colonies positive for a given epitope and the proportion of bacteria that were labelled in the broth culture from which the colonies were derived was also consistent. It is likely that colonies with a low proportion of bacteria expressing the epitope were derived from a cell which did not initially carry the epitope and that some type of switching mechanism has generated the variants. The rate of switching appeared to occur at a constant rate for the epitopes examined, as the estimated level of expression remained constant for different colonies and between experiments carried out at different times. The proportion of bacteria expressing an epitope was maintained when colonies were subcultured into broth continuously for up to 5 days. This finding suggests that either the switching mechanism observed within the colonies does not function in the relatively homogeneous environment of the broth culture or it occurs at a much slower rate. Given the apparent constant rate of change in the colonies, it is likely that there is an underlying genetic basis to the mechanism that generates the variation. This is a well-documented phenomenon in other pathogenic bacteria (24). It remains to be determined if there is also an environmental influence on this type of variation. Possibilities include external influences such as nutrient availability or changes in the redox potential in the microenvironment. Another possibility is that bacterial signalling molecules such as for example the N-acyl homoserine lactones (2) are involved. These have been well characterized in other gram-negative bacteria, although to date no evidence has been presented for the production of these or similar molecules in Bacteroides spp. An environmental influence could perhaps explain the differences observed between the colonies and the broth culture.

The lack of cross-reactivity of populations enriched from colonies in which 98% or more labelled with a given MAb with other MAbs (Table 5) suggests that enrichment for one polysaccharide epitope does not result in coenrichment for a second epitope. Whether this represents exclusive production of one antigenic type or the masking of one by another remains to be determined. The patterns generated by PAGE and immunoblotting, illustrated in Fig. 1, suggest that B. fragilis may express three distinct components extractable by the hot phenol-water method. It is likely that these three components are the smooth lipopolysaccharide (QUBF 5), high-molecular-mass polysaccharide (QUBF 6 to 8), and common antigen (QUBF 12) described by Poxton and Brown (18). Furthermore, the lack of cross-reactivity of the populations enriched from QUBF 6- or 7-positive colonies with the other MAbs examined (Table 5) indicates that there are at least three antigenic types of high-molecular-mass polysaccharide with an associated ladder pattern. A further two MAbs, which also did not cross-react with populations enriched from QUBF 6- or 7-positive colonies, have a PAGE pattern indicative of high-molecular-mass polysaccharide but lacking the ladder pattern (unpublished data).

The precise nature of the biochemical differences which generate these different patterns of labelling are unknown. It is possible for polysaccharides to be biochemically similar in terms of components (e.g., sugar moieties) but for a wide variety of antigenic variation to be generated by alteration of either the linkage of the substituent moieties, their chemical substitution, or both. Antigenic variation generated by these means is well documented in the polysaccharides of other pathogenic bacteria such as Haemophilus influenzae, E. coli, and Neisseria meningitidis (12). As yet, no chemical analyses have been carried out on the antigens described in this report. Their relationship with the chemically characterized polysaccharides A and B described by Pantosti and coauthors (7, 8) is therefore unknown. Polysaccharides A and B were obtained from the fraction referred to as capsular polysaccharide (CP) by boiling in 5% acetic acid for 1 h. MAbs specific for either A or B revealed a broad high molecular mass band after PAGE and immunoblotting of the CP fraction. After acetic acid hydrolysis, a narrower band in the highest-molecular-mass region of the broad band was visible for both fractions A and B and their respective MAbs. It therefore appears that polysaccharides A and B also give a similar pattern after PAGE and immunoblotting. A ladder pattern associated with the broad band was detected in the CP fraction by using polyclonal antiserum. These authors suggested that this ladder pattern represented the ladder pattern described by ourselves in the small capsule subpopulation enriched from strain NCTC 9343 by Percoll density gradient centrifugation (23) and also the ladder pattern interpreted by Poxton and Brown as a possible O antigen (18). The ladder pattern illustrated by Poxton and Brown is similar to that detected by our MAb QUBF 5 (Fig. 1) and has wider-spaced bands than that illustrated by Pantosti and coauthors (8). Their ladder pattern is, however, similar in appearance to that observed in the small-capsule population (5, 23). As strain NCTC 9343 normally contains a mixture of large-capsule, small-capsule, and noncapsulate bacteria, it is likely that polysaccharides A and B were extracted from a population with a mixture of types of capsule as well as antigenic types.

The biological activity of our phenol-water extracts was not investigated. Delahooke and coauthors (3), however, compared the immunomodulatory activity of polysaccharide material extracted from B. fragilis by different methods and reported a high level of in vitro biological activity in phenol-water extracts. Material extracted from B. fragilis NCTC 9343 grown in the same defined medium as we used (28) was 10 times more active in the Limulus amoebocyte lysate assay than material obtained from E. coli O18:K. These authors acknowledge that their B. fragilis extracts were probably heterogeneous both in M_r and molecular composition. It would be interesting to relate this biological activity to the potential variety of components identified in the present study.

The capacity for antigenic variation could clearly be advantageous to the survival of B. fragilis in both its pathogenic mode of existence and in its role as a member of the normal intestinal microbiota. These results also have implications for any studies of the virulence of B. fragilis and the chemical nature of these polysaccharides. Inter- and intrastain variation can be clearly observed not only in recent clinical isolates but also directly in pus samples with MAb labelling and immunofluorescence microscopy (17, 25, 26). Studies of the immunological diversity of B. fragilis which rely only on whole-cell dot immunobinding or enzyme-linked immunosorbent assay (7) will not detect this within-strain diversity. The intrastrain variation will result in titers in such assays which reflect the proportion of the bacteria within the population which are expressing the epitope. This intrinsic variability of the surface polysaccharides will also need to be taken into account in the production of potential vaccines and immunodiagnostic tests based on polysaccharides.

In conclusion, surface polysaccharides of noncapsulate B. fragilis are antigenically highly variable within individual strains. Different antigenic types can be detected on different bacterial cells within an individual population of a given strain. The proportion of bacterial cells carrying any given epitope will be variable depending on the strain and how it has been cultured in the laboratory. Production of the same polysaccharide by all the bacteria within a strain of B. fragilis is unlikely and cannot be assumed.