INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Bacteria belonging to the Bacteroides fragilis group are the clinically most important of the anaerobic pathogens and are commonly isolated from intraabdominal infections and other infections below the diaphragm (9, 25). Today, carbapenems, nitroimidazoles, chloramphenicol, and beta-lactam agents in combination with beta-lactamase inhibitors are used for treatment of infections when the B. fragilis group is involved. Newly developed quinolones (clinafloxacin, moxifloxacin, levofloxacin, and trovafloxacin) with an extended spectrum of antimicrobial activity compared to earlier quinolones, such as ciprofloxacin, may also be a choice of treatment. However, rates of resistance to these antibiotics are increasing. In an ongoing surveillance study, 16.4% of 1,220 clinical B. fragilis group strains from 19 European countries were considered resistant to moxifloxacin, with MICs of 4 µg/ml (M Hedberg, Karolinska Institute, Sweden, personal communication).

The quinolones act by interfering with type II topoisomerases, DNA gyrase and topoisomerase IV, which are responsible for cleavage, passage, and rejoining of double-stranded DNA in an ATP-dependent reaction (5). The major difference between the two enzymes is that DNA gyrase has a supercoiling activity, while topoisomerase IV has the ability to decatenate replicated daughter chromosomes. DNA gyrase consists of two subunits, GyrA and GyrB, encoded by the gyrA and gyrB genes, respectively. Quinolone resistance attributable to gyrA mutations was first reported in Escherichia coli (12). The gyrA quinolone resistance-determining region (QRDR), a highly conserved motif in E. coli, is located from amino acid residues Ala-67 through Gln-106. Mutations in this region correlated to quinolone resistance have by now been reported in a number of aerobic bacteria (2, 4, 7, 13, 15, 24, 30). A high level of resistance is commonly associated with substitutions of amino acid residues equivalent to Ser-83 and Asp-87 in E. coli. In gram-negative species, quinolone resistance often arises initially from mutations in gyrA. If additional mutations in parC occur, these usually lead to higher resistance levels (6). Recent data suggest that the gyrA gene also plays a role in quinolone resistance in the anaerobe B. fragilis. In a study by Onodera et al. (26), the genes encoding the DNA gyrase A and B subunits of B. fragilis ATCC 25285 were sequenced. In laboratory mutants of a reference strain resistant to levofloxacin, Ser82 (equivalent to E. coli Ser83) of gyrA was replaced with Phe, but no mutations were found in gyrB (26). Bachoual et al. (1) detected Ser82Phe changes in gyrA in 3 of 12 trovafloxacin-resistant (MIC = 4 µg/ml) clinical isolates. This mutation was also found in two of three second-step mutants selected with ciprofloxacin (1). Alterations in GyrA might not be the only explanation for quinolone resistance. Other mechanisms, such as efflux and changes in the outer membrane proteins or alterations in ParC, may also confer quinolone resistance.

Drug efflux has been shown to be an important component in the development of high levels of resistance to this class of agents (8, 10, 17, 27, 31). The presence of an efflux pump in B. fragilis, actively pumping out norfloxacin, has been reported by Miyamae et al. (18).

In the present investigation, six American Type Culture Collection (ATCC) strains and 44 clinafloxacin-resistant and -susceptible fecal isolates of the B. fragilis group were analyzed as follows: (i) antimicrobial susceptibilities to ciprofloxacin, levofloxacin, moxifloxacin, and clinafloxacin were determined, and (ii) the QRDR of the gyrA gene was sequenced.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains. A total of 44 clinafloxacin-resistant and -susceptible strains of the B. fragilis group (eight B. fragilis, three Bacteroides ovatus, five Bacteroides thetaiotaomicron, six Bacteroides uniformis, and 22 Bacteroides vulgatus) were investigated. All resistant strains derived from a previous study on the impact of clinafloxacin on the intestinal microflora (23). These had been isolated from stool specimens of 12 healthy volunteers treated with clinafloxacin capsules (Parke-Davis Pharmaceutical Research, Berlin, Germany), 200 mg twice a day for 7 days, and had been selected during the administration period or within 2 weeks after withdrawal of the drug using screening plates containing clinafloxacin at 4 µg/ml. The susceptible isolates were derived from stool samples collected prior to clinafloxacin administration or clinical strains from an ongoing European surveillance study (M. Hedberg, Karolinska Institute, Sweden, personal communication). The isolates were identified to genus level by morphology, biochemical tests, and gas-liquid chromatography analysis of metabolic end products (20, 29). Identification to species level was performed phenotypically by fermentation of rhamnose, trehalose, saccharose, and arabinose and biochemical reaction with esculine, indole, and catalase (19, 21). The reference strains B. fragilis ATCC 25285, B. distasonis ATCC 8503, B. ovatus ATCC 8483, B. thetaiotaomicron ATCC 29741, B. uniformis ATCC 8492, and B. vulgatus ATCC 8482 were also included in the study.

Antibiotics. The antimicrobial agents tested were ciprofloxacin and moxifloxacin (Bayer, Leverkusen, Germany), levofloxacin (Roussel Uclaf, Paris, France), and clinafloxacin (Parke Davis Pharmaceutical Research, Ann Arbor, Mich.).

MIC determination. The MICs of ciprofloxacin, levofloxacin, moxifloxacin, and clinafloxacin against the B. fragilis group strains were determined by the agar dilution method using brucella agar supplemented with 5% lysed sheep blood according to the published standard (21). The quinolone agents were suspended and diluted according to the manufacturers' instructions. An inoculum of 10⁵ CFU per spot was delivered with a Steers replicator. The plates were incubated anaerobically for 48 h at 37°C. The MIC was defined as the lowest concentration of the drug that caused a marked change in the appearance of growth compared to the control plate.

DNA isolation. Chromosomal DNA was extracted from the isolates by suspending fresh colonies in 100 µl of sterile water and boiling for 5 min. After centrifugation for 5 min at 10,000 × g, supernatants were collected, and 1:10 dilutions in sterile water were used for PCR.

Amplification of QRDR of gyrA gene. PCR amplification of the gyrA QRDR of the strains was carried out with primers Pr-BFGBA03 (5'-ATGCTTGAACAAGACAGAATTATAAAG-3') and Pr-BFGA02 (5'-GACTGTCGCCGTCTACAGAACCG-3') published by Onodera et al. (26). The primers were purchased from Life Technologies AB, Täby, Sweden. PCRs were carried out in a final volume of 50 µl containing each primer at a concentration of 0.2 pM, 1× PCR buffer, 1.0 mM deoxyribonucleoside triphosphates, 2.0 to 3.0 µM MgCl₂, 2.5 U of Taq polymerase (Sigma, St. Louis, Mo.), and 10 ng of template. PCR conditions were based on these described previously: 25 cycles of 94°C for 0.5 min (denaturation), 60 to 62°C for 5 min (annealing), and 72°C for 1 min (extension). Optimal conditions varied slightly for different strains and species. PCR products were analyzed by electrophoresis on a 1% agarose gel in TBE buffer (Tris base, boric acid, and EDTA [pH 8.0]). The gel was stained in a 0.5-µg/ml ethidium bromide bath, visualized by UV translumination, and photographed using Polaroid films. DNA fragments 282 to 296 bp long of the QRDR of gyrA were obtained.

DNA sequencing and sequence analysis. Free primers and nucleotides from the PCR products were removed by the QIAquick-spin PCR purification kit (Qiagen Inc., Chatsworth, Calif.). Sequencing was carried out using ABI Prism dye terminator cycle sequencing kits with AmpliTaq DNA polymerase, dye terminator chemistry, and an ABI 310 genetic analyzer (Perkin Elmer). Nucleotide and deduced amino acid sequences between the primers were analyzed using the Macintosh DNA program ABI automated DNA sequencer viewer EditView (Perkin Elmer), ClustalW interactive multiple sequence alignment at European Bioinformatics Institute (http://www2.ebi.ac.uk), and the ExPASY molecular biology server at the Swiss Institute of Bioinformatics (Geneva, Switzerland; http://expasy.hcuge.ch). All PCR amplification and sequencing procedures were performed at least twice for each strain. Amino acid sequences of the isolates between the primers were compared between resistant and susceptible isolates and with the corresponding ATCC type strain.

Nucleotide sequence accession numbers. The partial DNA sequences corresponding to the gyrA gene of the five ATCC strains of the B. fragilis group have been assigned the following GenBank accession numbers: AJ279040 (B. distasonis ATCC 8503), AJ279043 (B. ovatus ATCC 8488), AJ279039 (B. thetaiotaomicron ATCC 2974), AJ 279042 (B. uniformis ATCC 8492), and AJ279041 (B. vulgatus ATCC 8482).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The MIC values of the various quinolones are listed in Table 1. Clinafloxacin was the most active of all the quinolones, followed by moxifloxacin, levofloxacin, and ciprofloxacin. Thirty-one isolates were classified as resistant to all four quinolones. In the absence of breakpoint criteria for the quinolones tested, the recommended breakpoint concentration for trovafloxacin, 4 µg/ml, was used (22). In general, strains resistant to moxifloxacin and clinafloxacin usually had higher MIC values (32 µg/ml) to ciprofloxacin and levofloxacin than the moxifloxacin- and clinafloxacin-susceptible strains.

The sequenced fragment of the gyrA QRDR in B. fragilis ATCC 2528 was identical in nucleotide and amino acid sequences to that previously assigned in GenBank (accession no. AB017712) and had 72% identity with E. coli. The pairwise scores among the study ATCC strains regarding amino acid sequences matched with B. fragilis ATCC 25285 and E. coli as follows: B. distasonis ATCC 8503, 95 and 73%; B. ovatus ATCC 8483, 100 and 72%; B. thetaiotaomicron ATCC 2974, 98 and 72%; B. uniformis ATCC 8492, 89 and 72%; and B. vulgatus ATCC 8482, 92 and 72%, respectively. It was also observed among the clinical isolates that the partial gyrA sequences of almost all B. vulgatus isolates had higher similarity with B. uniformis ATCC 8492 than with B. vulgatus ATCC 8482. Mutations in the gyrA QRDR are listed in Table 2. At resistance hotspots 82 (equivalent to Ser-83 in E. coli) and 86 (equivalent to Asp-87 in E. coli), amino acid exchanges were found in 15 and 8 strains, respectively. Thirteen isolates (four B. fragilis, five B. uniformis, and four B. vulgatus) had Ser82Leu changes, while in two B. ovatus strains Ser82 was replaced with Phe or Cys. The strains with Ser82Leu mutations were resistant to all four quinolones tested, while those with Phe or Cys substitutions of Ser82 were susceptible to at least two of the quinolones. Eight isolates (five B. fragilis, one B. ovatus, and two B. thetaiotaomicron) carried amino acid exchanges at position 86. Two B. thetaiotaomicron isolates with a single mutation of Tyr86 to Phe were resistant to all four quinolones, while one B. ovatus isolates with a single mutation of Phe86 to Tyr showed a susceptibility pattern similar to that of the type strain. Phe86Gly substitutions were present in five strains, of which four with additional Ser82Leu mutations were resistant to all quinolones tested. Eight isolates, three B. fragilis, three B. thetaiotaomicron, one B. uniformis, and one B. vulgatus, lacked mutations and were also susceptible to at least two of the quinolones tested. The remaining 36 fecal isolates, of which 31 were resistant to all quinolones tested, had one to nine mutations in their QRDRs. Other more unusual amino acid substitutions occurred most frequently at positions 30, 49, 53, 61, 68, 94, and 100 and could not be clearly related to resistance.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The similarities of the partial QRDR sequences of the ATCC strains with B. fragilis ATCC 25285 varied between 89 and 100%. Ser82 was conserved in all studied B. fragilis group reference strains, but at position 86 Phe, Gly, or Tyr was present. These substitutions for the Asp (in E. coli) may help to explain why B. fragilis group strains are intrinsically resistant to many of the quinolones, since these mutations are common in quinolone-resistant E. coli strains. The partial gyrA sequences of almost all clinical B. vulgatus isolates had higher similarity with B. uniformis ATCC 8492 than with B. vulgatus ATCC 8482. ATCC strains for specific species might not necessarily represent a "wild-type" sequence for that species. Some variability found in clinical isolates may be due to incomplete genetic information about the range of mutations in clinical strains. A mosaic of amino acid changes might be found among a large database for each Bacteroides spp.

The possibility of misidentification in the present investigation is low, since phenotyping results were clear and identical on repeated testing. The routine phenotypic identification has also been shown to agree with the more accurate genotyping by restriction fragment length polymorphism analysis of amplified 16S rDNA (28).

Strains with Ser82Leu mutations were resistant to all quinolones tested, while strains with other Ser82 substituents than Leu were susceptible to two or more of the quinolones tested. In two previous studies on quinolone-resistant B. fragilis strains, Ser82 was replaced with Phe (1, 23). The MIC values of the in vivo-selected isolates in our study were considerably higher than those reported by Bachoual et al. (1) and Onodera et al. (26). Mutations at hotspot position 86 were found in eight isolates (five B. fragilis, one B. ovatus, and two B. thetaiotaomicron). Five of these showed resistance to all quinolones tested, of which two carried single point mutations of Tyr86 to Phe. One strain with a single mutation of Phe86 to Tyr was susceptible to the two most active quinolones. Four of the five isolates in which Phe86 was replaced with Gly were resistant to all quinolones tested, although these had additional Ser82Leu mutations. In contrast, one isolate with the Phe86Gly substitution but no Ser82Leu replacement showed quinolone susceptibility similar to that of the type strain. These results suggest that replacement at codon 86 with Phe might contribute to high quinolone resistance, while substitutions with Tyr or Gly do not. In a similar way, isolates with Leu replacing Ser at position 82 were resistant to moxifloxacin and clinafloxacin, while those with Phe or Cys substitutions at the same position were not. It has previously been shown in E. coli that different amino acid substitutions in the same position in GyrA can produce different levels of resistance (11). Among some strains with identical GyrA sequences, a wide range of MIC values could be noted, indicating involvement of other resistance mechanisms. In the present study, resistant isolates had been selected using only clinafloxacin, and the correlation between mutations and MIC values for other agents might be inconsequent, since each agent may select for different mutations. However, all clinafloxacin-resistant isolates also showed resistance to moxifloxacin, levofloxacin, and ciprofloxacin. The resistance pattern among the isolates was consistent; no remarkable differences were shown in the order of activity between the different quinolones.

Newer quinolones like moxifloxacin and clinafloxacin could be potential agents against infections caused by the B. fragilis group since they have excellent in vitro activity against anaerobes. Comparative studies have shown that clinafloxacin has the highest activity among the quinolones (3, 14, 16). A possible disadvantage is that B. fragilis group strains resistant to quinolones can readily be selected in the intestinal microflora. In a previous study by our group, it has been shown that a shift from susceptible to resistant intestinal strains occurred in 9 of 12 subjects during clinafloxacin treatment (23). Thus, it is important to restrict the quinolone class of antibiotics to appropriate indications, due to the risk of emergence of resistance.

In conclusion, strains of the B. fragilis group with reduced susceptibility to quinolones harbored a number of amino acid substitutions, including mutations at hotspot positions in their QRDRs of GyrA. Fifteen of 31 strains resistant to all quinolones tested had mutations at hotspot positions, mainly Ser82Leu. Sixteen highly resistant strains lacked hotspot mutations, indicating the involvement of other mechanisms of resistance. Further studies are needed to increase the understanding on quinolone resistance in Bacteroides species.