ABSTRACT
Previous studies have shown that topoisomerase IV and DNA gyrase interact with quinolones and coumarins in different ways. The MICs of coumarins (novobiocin and coumermycin) for MT5, a Staphylococcus aureus nov mutant, are higher than those for wild-type strains. Sequencing the gyrB gene encoding one subunit of the DNA gyrase revealed the presence of a double mutation likely to be responsible for this resistance: at codon 102 (Ile to Ser) and at codon 144 (Arg to Ile). For single-step flqA mutant MT5224c9, previously selected on ciprofloxacin, the fluoroquinolone MIC was higher and the coumarin MIC was lower than those for its parent, MT5. Sequencing the grlB andgrlA genes of topoisomerase IV of MT5224c9 showed a single Asn-470-to-Asp mutation in GrlB. Genetic outcrosses by transformation with chromosomal DNA and introduction of plasmids carrying either the wild-type or the mutated grlB gene indicated that this mutation causes both increased MICs of fluoroquinolones and decreased MICs of coumarins and that the mutant grlBallele is codominant for both phenotypes with multicopy alleles. Integration of these plasmids into the chromosome confirmed the codominance of fluoroquinolone resistance, butgrlB + appeared dominant over grlB(Asp-470) for coumarin resistance. Finally, the gyrA(Leu-84) mutation previously described as silent for fluoroquinolone resistance increased the MIC of nalidixic acid, a nonfluorinated quinolone. Combining the grlA (Phe-80) and grlB(Asp-470) mutations with this gyrA mutation also had differing effects. The findings indicate that alterations in topoisomerases may have pleiotropic effects on different classes of inhibitors as well as on inhibitors within the same class. A full understanding of drug action and resistance at the molecular level must take into account both inhibitor structure-activity relationships and the effects of different classes of topoisomerase mutants.
Staphylococcus aureus is a major pathogen, many strains of which are now susceptible to only a few antibiotics. Resistance has begun to compromise the utility of fluoroquinolones as antistaphylococcal agents.
Four genes related to fluoroquinolone resistance in S. aureus have been reported so far: norA (24),gyrA (15, 32), gyrB (15), and grlA (8, 25, 41). Resistance involving thenorA gene is due to increased expression of the NorA efflux pump. The other resistance genes encode the structural proteins of DNA topoisomerases: gyrA and gyrB encode the structural proteins of DNA gyrase, and grlA (referred to asparC in other species) encodes one of the two subunits of DNA topoisomerase IV. The grlB gene encoding the second subunit of topoisomerase IV has not been previously implicated in fluoroquinolone resistance in S. aureus, but the mutation of its homolog in Escherichia coli and Streptococcus pneumoniae, parE, has been shown to cause resistance (4, 31).
The fluoroquinolones act on type 2 topoisomerases by trapping or stabilizing an enzyme reaction intermediate in which both DNA strands are cleaved and covalently linked to the breakage-reunion subunits (GyrA in the case of gyrase and GrlA in the case of topoisomerase IV) (14). The stabilization of this cleavage complex initiates a series of events that result in cell death. In contrast to the findings for E. coli, in staphylococci and other gram-positive bacteria that have been studied, it appears that DNA gyrase is only a secondary target of many fluoroquinolones and that topoisomerase IV is the primary target (8, 25, 27). Mutations in gyrAare found only in more highly resistant clinical strains of S. aureus that also have mutations in grlA, andgrlA mutations precede gyrA mutations in single-step resistance (9). In addition, gyrAmutations producing fluoroquinolone resistance in S. aureus are silent in the absence of grlA mutations (25). Mutations in three codons were described forgrlA: codon 80 (Ser→Phe or Tyr), 84 (Glu→Lys), and 116 (Ala→Glu or Pro) (9, 25, 41).
A quinolone resistance locus referred to as flqA was localized to the A fragment of chromosomal DNA of first-step ciprofloxacin- and ofloxacin-selected resistant mutants digested with SmaI between the thr and trp loci (37). This region has been shown to contain thegrlB and grlA genes (25). In a previous work (25), some flqA mutants were shown to have grlA mutations but one single-step flqAmutant selected on ciprofloxacin (MT5224c9) (37) showed no nucleotide change in the quinolone resistance-determining region ofgrlA, suggesting that additional resistance mutations may occur either in grlA outside the region sequenced, or in the adjacent grlB gene, or in another as yet undefined but linked locus. In addition to quinolone resistance, mutant MT5224c9, which was selected from novobiocin-resistant parent strain MT5 (nov-142), showed a substantial decrease in novobiocin resistance (37).
Coumarins such as novobiocin and coumermycin act differently from quinolones: they act by competitive inhibition of ATP hydrolysis by the B subunit of DNA gyrase (10, 12). The gyrase B protein consists of two domains: a 43-kDa N-terminal domain containing the ATPase site as well as the coumarin-binding site and a 47-kDa C-terminal domain responsible for interaction with the A protein and DNA (21). The binding of coumarin and ATP is competitive because of overlap in their binding sites (20). All reported mutations that result in coumarin resistance lie at the periphery of the ATP-binding site of GyrB. In contrast, quinolone resistance mutations in GyrB have been localized to a different subunit domain in a position midway between the amino and carboxy termini.
The aim of this study was to determine the origin of the modification of the quinolone and coumarin resistance in the mutant MT5224c9 and to determine if the nov-142 locus was a mutant allele ofgyrB. We have found that a novel mutation in thegrlB gene resulting in the replacement of an asparagine by an aspartate at position 470 is responsible for the phenotype of MT5224c9. In addition, we found that the nov-142 locus is a novel doubly mutant allele of gyrB.
MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions.The strains and plasmids used in this study are listed in Table1. E. coli strains were grown in Luria-Bertani medium, and S. aureus strains were grown in Trypticase soy (TS) medium. All strains were grown at 37°C except theS. aureus strains carrying the thermosensitive plasmids derived from pSK950, which were grown at 30°C.
Bacterial strains and plasmids used in this study
Determination of susceptibility patterns.Antibiotics were purchased from Sigma Chemical Co. (St. Louis, Mo.), except sparfloxacin and DU-6859a, which were kindly provided by Rhone-Poulenc Rorer and Daiichi Pharmaceuticals, respectively. MICs were determined with Mueller-Hinton agar supplemented with serial twofold-increasing concentrations of antibiotics. The cells were grown at 30°C when thermosensitive plasmids were used (free or integrated). For other determinations of MICs, cells were grown at 37°C.
Gene amplification and cloning.The grlB genes from strains ISP794 and MT5224c9 were amplified by PCR with two primers, both of which contain an engineered BamHI site (5′-AAC GTA CGG ATC CAG GAG GCG AAA T-3′; 5′ nucleotide at position 352 in the oligonucleotide coordinates used by Yamagishi et al. [41]; and 5′-GGC AAT GGA TCC TCT TGA ATA-3′; position 2470). PCR included annealing at 53°C with Vent DNA polymerase (New England Biolabs). PCR products were digested with BamHI and ligated into the BamHI site of pGEM3-zf(+). Recombinant plasmids were introduced into E. coli DH5α. In a similar manner, the grlA genes of ISP794 and MT5224c9 were amplified by PCR with two primers containing BamHI sites (5′-AGT GGA TCC TGA AGT GCG TT-3′; position 2196; and 5′-TAA GGA TCC GTA CTT GGT CA-3′; position 4881) and were cloned into pGEM3-zf(+).
The cloning into pSK950 was performed by using a PCR product amplified with the following primers: 5′-GAT TAA TAT ATG GGA TCC AGC TAT GAA AGT (position 259) and 5′-ACT GGA TCC TCC AAA GCG AT-3′ (position 2441). The PCR products were digested with BamHI, ligated into theBamHI site of pSK950, and introduced into E. coliDH5α
The 5′ termini of the gyrB genes of ISP794, MT5, and MT5224c9 were amplified by PCR with primers 5′-AAA GCG ATG GTG ACT GCA TTG-3′ (position 275) and 5′-GTG GCA TAT CCT GAG TTA TAT-3′ (position 1102 in the sequence published by Brockbank and Barth [5]), and PCR products were sequenced directly.
Sequencing.The sequences were determined with the ABI Fluorescent System and Taq Dye terminators (Qiagen). In order to verify that sequenced mutations were not introduced by Vent polymerase, each mutation was verified by direct sequencing on PCR products.
DNA transformation.Transformation of plasmids in S. aureus was performed by electroporation as previously described (4).
Chromosomal DNA was obtained by the procedure of Stahl and Pattee (33). Transformation with high-molecular-weight chromosomal DNA was performed as previously described (33).
Integration of plasmids derived from pSK950 into the chromosome.Plasmids derived from pSK950 were first introduced into S. aureus strains and grown at 30°C. pSK950 is derived from plasmid pCL84, which contains the attP site of staphylococcal phage L54a (18) and which is capable of integrating specifically into the chromosomal attB site located just 3′ of the geh gene, which encodes staphylococcal lipase. Integration is facilitated by the presence of plasmid pYL112Δ19, which carries the L54a int gene encoding integrase. pYL112Δ19 was introduced into the strains carrying the derivatives of pSK950 and selected at 30°C for 48 h on tetracycline (5 μg/ml) and chloramphenicol (10 μg/ml). A first shift to 42°C was done in TS broth with chloramphenicol (10 μg/ml) for 24 h, and another incubation was performed at 42°C on TS agar for 24 h. One colony was isolated on TS agar containing 2 μg of tetracycline per ml at 37°C.
The site-specific integration of the plasmid into the S. aureus chromosome was confirmed by scoring for loss of lipase activity resulting from the disruption of the lipase gene. The lipase assay was performed as previously described by plating overnight cultures on J1 medium containing egg yolk (19). In order to determine that plasmids derived from pSK950 did not remain in the cells, the absence of these plasmids was verified by minipreps (Promega).
RESULTS
Sequencing of grlB and grlA. The entiregrlA genes from ISP794 and MT5224c9 were amplified by PCR, generating a 2.7-kb fragment that was cloned into the BamHI site of pGEM3-zf(+). The entire grlA sequence of MT5224c9 was identical to that of ISP794 (data not shown).
The 2.2-kb PCR products containing the entire grlB gene were cloned into pGEM3-zf(+). The sequence of grlB from strain MT5224c9 revealed a single mutation in comparison to that ofgrlB from ISP794, an A-to-G change resulting in the replacement of an asparagine (codon AAT) by an aspartate (codon GAT) at position 470 (Table 2).
Sites associated with quinolone resistance in the quinolone-resistance-determining regions of GrlB and GyrB
Sequencing of gyrB. The 5′ regions of gyrBgenes (830 bp) from strains ISP794, MT5, and MT5224c9 were amplified by PCR, and the region of each gene from nucleotide 342 (codon 21) to 983 (codon 243) in the sequence published by Brockbank and Barth (5) was directly sequenced after PCR. Two mutations producing the substitution of two amino acids were observed in MT5 and MT5224c9 in comparison to the sequence of gyrB from strain ISP794: a T→G substitution at nucleotide 586 modifying codon 102 (ATT [Ile] to AGT [Ser]) and a G→T substitution at position 702 modifying codon 144 (AGA [Arg] to ATA [Ile]) (Fig.1). These two mutations characterize thegyrB142 allele. Two other silent mutations were observed in the DNA from MT5 and MT5224c9 in comparison to that from strain ISP794: a C-to-T change at nucleotide 617 and an A-to-G change at position 872. The nov-142 marker originated in strain 655 (30) and was introduced into strain 8325 by several DNA outcrosses to produce MT5 (30, 37). As ISP794 was derived from strain 8325, the two silent mutations are likely due to strain-to-strain variabilities in the sequence of gyrB.
Alignment of the amino acid sequences of GyrB involved in coumarin resistance (positions 71 to 190) from S. aureus(5), S. pneumoniae (22), and E. coli (1) and identification of positions of the mutations responsible for coumarin resistance. The asterisks and the circles indicate the sites involved in ATP (39) and novobiocin binding (20), respectively. The mutants ofS. aureus (except MT5), S. pneumoniae, andE. coli GyrB were previously described (references34, 22, and 6, respectively). Numbers at the right refer to the positions of the last amino acids.
Genetic linkage of quinolone resistance and novobiocin hypersusceptibility phenotypes.To determine the relationship of the quinolone resistance and novobiocin susceptibility of MT5224c9, we performed both incross and outcross transformation experiments to define the linkage of these phenotypes to Ω(chr::Tn917lac)2, a Tn917transposon insertion encoding erythromycin resistance (Ermr) and previously shown to be linked to grlAand grlB (25). For the incross experiment, high-molecular-weight chromosomal DNA from strain ISP2133 Ω(chr::Tn917lac)2 grlA + grlB + was used to transform MT5224c9, selecting for Ermr. Forty-four percent of 246 Ermr transformants showed both loss of quinolone resistance and an increase in novobiocin resistance (Table3). The remaining Ermrtransformants had no change in these resistances from those of MT5224c9.
Outcross and incross of flqA(grlB543) locus by transformation with chromosomal DNA
For the outcross experiment, chromosomal DNA was prepared from strain EN3 grlB543Ω(chr::Tn917lac)2 and used to transform strain MT5 (nov-142), selecting for Ermr. Thirty-eight percent of 149 Ermrtransformants exhibited increases in quinolone resistance and decreases in novobiocin resistance, with the remaining transformants showing no change relative to the recipient (Table 3). Thus, both the quinolone resistance and novobiocin susceptibility phenotypes of thegrlB543 mutants are 38 to 44% linked to Ω(chr::Tn917lac)2, and no dissociation of phenotypes was seen. These data indicate that the quinolone resistance and novobiocin susceptibility ofgrlB543 are tightly linked and likely due to the singlegrlB543 mutation.
Dominance characteristics of grlB543 in multicopy and single copy.To determine the dominance characteristics ofgrlB543, the grlB genes of strains ISP794 (grlB +) and MT5224c9 (grlB543) were cloned into hybrid shuttle plasmid pSK950. This plasmid carries the thermosensitive replicon of staphylococcal plasmid pE194, which is stable at 30°C but unstable at 42°C, and the replicon of E. coli plasmid pSC101. grlB was cloned directly from PCR products into pSK950, which has a low copy number in E. coli (10 copies per cell). We were able to clonegrlB without its promoter in E. coliplasmid pGEM3-zf(+), a high-copy-number plasmid, but were unable to clone grlB with its promoter in this plasmid. The resulting plasmids, derived from pSK950, pBFISB (grlB +), and pBFC9B (grlB543), were introduced into strains RN4220 and EN22, the latter being a strain derived from RN4220 in which the Ω(Tn917lac)2 marker is linked to thegrlB543 chromosomal mutation. The effects of the introduction of these plasmids on drug susceptibility are shown in Table 4. RN4220 (grlB +) carrying pBFC9B (grlB543) increased its resistance to norfloxacin (eightfold) relative to RN4220 containing pSK950. The effect of the mutated gene on resistance to quinolones ciprofloxacin, ofloxacin, and sparfloxacin was less (twofold). A surprising result was the absence of an effect of pBFC9B on the MICs of nalidixic acid and oxolinic acid. In addition to the effects of grlB543 on quinolone resistance, novobiocin and coumermycin MICs were decreased fourfold. These data definitively demonstrated that the grlB543 allele itself causes the quinolone resistance and novobiocin hypersusceptibility phenotypes. For the merodiploid of EN22 (grlB543) containing pBFISB (grlB +) there were decreases in the MICs of the tested fluoroquinolones relative to those of the merodiploid containing plasmid-encoded grlB543 (in strain RN4220); the coumarin MICs were similar. Thus,grlB543 is codominant to grlB +, with the phenotype determined by the allele present on the multicopy plasmid.
Susceptibility patterns of S. aureus RN4220 strains diploid for grlB + and mutant grlB
To determine further the dominance characteristics ofgrlB543 in the presence of the novobiocin resistance mutations carried by gyrB142, plasmids were extracted from RN4220 and introduced by electroporation into derivatives of strain ISP794 (gyrB + grlB +), MT5 (gyrB142 grlB +), or MT5224c9 (gyrB142 grlB543) (Table 5). A pattern similar to that for RN4220 and EN22 was seen, confirming the codominance of fluoroquinolone resistance in grlBmerodiploids. The modifications of the MICs of ciprofloxacin, norfloxacin, and ofloxacin in the presence of the different plasmids were, however, somewhat greater in these strains than in RN4220: 8-, 8- to 32-, and 4-fold, respectively. In the MT5 (gyrB142) background, the grlB543 mutation decreased by four- to eightfold the MICs of coumarins. Conversely,grlB + increased by four- to eightfold the MICs of coumarins.
Susceptibility patterns of S. aureus MT5gyrB142 strains diploid for grlB + and mutant grlBa
To determine the dominance characteristics of single copies ofgrlB543 and grlB +, pBFISB and pBFC9B were integrated into the chromosomal lipase gene. MICs for the strains carrying both grlB + and grlB543 on the chromosome were identical for both merodiploids and corresponded to an average of the MICs of quinolones for the wild-type and resistant strains, indicating codominance for quinolone resistance (Table 5). In contrast, in MT5 (gyrB142) with chromosomally integrated merodiploids, grlB + appeared dominant overgrlB543 for coumarin resistance (Table 5).
Effects of mutations in grlA, grlB, andgyrA alone and in combination on susceptibility to quinolones and novobiocin.In a previous work, we showed that a mutation in gyrA encoding GyrA (Ser-84 to Leu [Ser84Leu]) caused no fluoroquinolone resistance in the absence of agrlA mutation (25). To expand these observations, we determined the MICs of additional quinolones for strains with single, dual, and triple mutations (Table6). In the absence of mutations in thegrlB and grlA genes, the MICs of the fluoroquinolones tested were unchanged for the strain carrying thegyrA mutation. However, the MIC of nalidixic acid was increased fourfold. When the grlA mutation was combined with the gyrA mutation, a major increase in the MIC of sparfloxacin in comparison to that produced by the grlAmutation alone was observed: 133-fold. Increases of 16-, 8-, and 4-fold were observed for the MICs of ciprofloxacin, DU6859a, and norfloxacin, respectively. In contrast, for the combination of the grlBand gyrA mutations, MICs for most of the quinolones were unchanged or increased only twofold in comparison with MICs associated with grlB alone. The only substantial effect of this combination was observed with sparfloxacin: a 16-fold increase. For nalidixic acid, neither the grlA nor grlBmutation conferred resistance when present alone, nor did they increase resistance in combination with the gyrA mutation.
Effects of grlA, grlB, andgyrA mutations on the activities of selected quinolones
DISCUSSION
Spontaneous mutant MT5224c9 was previously obtained in one step by plating MT5 gyrB142 on ciprofloxacin (37). This mutant showed an increase in the MIC of ciprofloxacin (about fourfold) and a decrease in the MIC of novobiocin (about eightfold) in comparison to its parent, MT5. In a previous study, the mutation in MT5224c9 was localized to SmaI chromosomal fragment A by its linkage to a Tn551 insertion in this fragment (37), which also carries the grlB and grlA genes (25). This previous observation suggested that the resistance to quinolones was due to a mutation in topoisomerase IV. However, no mutation was found in the region of grlA in which several mutations responsible for quinolone resistance in other mutants had been found (8, 25, 41). These mutations in grlA did not affect novobiocin resistance (25). We have shown here that MT5224c9 contains a novel grlB543 mutation encoding Asp-470 of GrlB, which is responsible for both quinolone resistance and coumarin hypersusceptibility.
The nov-142 locus of S. aureus has been used for genetic mapping and has been presumed to be gyrB, as are novobiocin resistance loci in other species (6, 13, 21, 22,34-36), based on its close linkage to gyrA, which is adjacent to gyrB in S. aureus (25). Sequencing the gyrB gene of MT5, a nov-142 mutant (37), revealed two different mutations, Ile102Ser and Arg144Ile. These two mutations were previously described individually, but not together, as being in the gyrB genes of coumarin-resistant strains of S. aureus and had not been linked specifically to the nov genetic locus (34) (Fig. 1). The second mutation at position 144 has also been described for E. coli gyrB (7). As the novobiocin and coumermycin MICs for MT5 are particularly high, it is not surprising that two mutations may be involved in this level of coumarin resistance. MICs of coumarins for strains carrying the individual Ile102Ser and the Arg144Ile mutations are 8 and 16 μg of novobiocin/ml and 0.03 and 1 μg of coumermycin/ml, respectively (34). The MICs for MT5 are 40 μg of novobiocin/ml and 5 μg of coumermycin/ml, values compatible with the incremental effects of the two mutations.
grlB has similarities to gyrB, including conservation of the domains involved in ATPase activity, resistance to novobiocin, and resistance to quinolones (Fig. 1; Table 2) (8). A single GrlB mutation (Asn470Asp) was found in MT5224c9, in comparison to strain ISP794. Since this mutation was associated with two phenotypes, increased quinolone resistance and decreased resistance to coumarins, it seemed possible that an additional undetected mutation might be responsible for one of the phenotypes and that the grlB mutation might be responsible for the other. Several findings indicated, however, that thegrlB mutation alone is responsible for both effects on quinolone and coumarin susceptibility. First, the mutant was obtained in a single step with a frequency of 5 × 10−9(37), a frequency consistent with a single mutational event. Second, no other mutation was found in the entire extent of thegrlB and grlA genes. Third, the genetic incross and outcross experiments showed no segregation of the quinolone and novobiocin phenotypes. Finally, the introduction of the plasmid carrying the mutant grlB gene into a quinolone-susceptible strain increased quinolone resistance and decreased novobiocin resistance.
In E. coli a region (PLKGKILN; amino acids 451 to 458, numbered as in S. aureus GrlB) is highly conserved between ParE (homologous to GrlB) and GyrB (4). This region is also conserved between GrlB (ParE) subunits of S. aureus(8) and S. pneumoniae (28). The Asn470Asp mutation encoded by the grlB543 allele is located slightly outside this conserved region. A quinolone resistance mutation also in this region in Salmonella typhimurium gyrB (position 463) has been described (11) (Table 2). By alignment with the sequence of the yeast topoisomerase II, position Asn-470 was found to be homologous to Asn-493 (data not shown). The crystal structure of yeast (Saccharomyces cerevisiae) topoisomerase II indicated that Asn-493 is on the B′α3 helix, which is distant from the active-site tyrosine directly involved in DNA strand breakage (2). The mechanism by which a mutation distant from the active site may affect quinolone activity is unknown. GrlA and GyrA mutations involved in quinolone resistance are, in contrast, close to the active-site tyrosine. It has been postulated that the active-site region of DNA gyrase, and by homology of topoisomerase IV, may constitute the region of quinolone binding, since GyrA mutations result in decreased drug binding (40). No structural data on topoisomerase-DNA-quinolone complexes have, however, been reported. Therefore, the exact mode of binding of quinolones to the enzyme-DNA complex remains unknown. The pattern of resistance due to mutations inE. coli gyrB had led to the hypothesis that direct electrostatic interactions of quinolones with GyrB amino acid residues might be involved in quinolone binding (42). This possibility seems unlikely for the GrlB (Asn470Asp) mutation, however, because the negative charge of the mutant amino acid is predicted by this hypothesis to increase the affinity of GrlB for quinolones with a positively charged piperazinyl group (ciprofloxacin or norfloxacin), thereby increasing susceptibility, rather than causing resistance, to these drugs (42). Because the phenotype of GrlB (Asp-470) is resistance to amphoteric quinolones, our findings are not consistent with such a model.
Another hypothesis is that the grlB mutation attenuates the intrinsic catalytic efficiency of the enzyme. Two mutants of yeast topoisomerase II carrying mutations in the PLRGKMLN region (Leu480Pro and Arg476Gly, each in combination with Leu475Ala, corresponding to positions 457, 453, and 452 in GrlB of S. aureus, respectively) were selected for resistance to amsacrine, a DNA topoisomerase-targeting drug that acts analogously to quinolones in forming drug-enzyme-DNA complexes. The mutant enzymes also exhibited decreased relaxation of supercoiled DNA with or without amsacrine (38). Thus, in such a model, if quinolone binding to the enzyme-DNA complex occurs only at certain steps in the catalytic cycle, then enzyme conformational changes that slow the catalytic cycle might reduce the steady-state level of enzyme-DNA complexes in the proper conformation to bind quinolones. This reduction is postulated to be sufficient to reduce the number of drug-enzyme-DNA complexes that initiate events leading to cell death and thereby cause drug resistance but to be insufficient to impair cell growth in the absence of a drug. Such conformational changes might act by reducing DNA binding, by altering subunit association, by altering enzyme affinity for ATP, or by impairing other steps in the catalytic cycle. We currently have no information about the catalytic efficiency of purified topoisomerase IV containing GrlB (Asp-470). Although MT5224c9 grows identically to its parent strain, MT5 (data not shown), this does not exclude a possible catalytic defect in topoisomerase IV containing GrlB543, since the gyrB (Arg-136) coumarin resistance mutation has been shown to impair enzyme catalytic function but not cell growth (6).
The codominance of fluoroquinolone resistance seen with multicopy and single-copy mutant alleles of grlB is similar to that seen with multicopy alleles of either parE + or the resistance mutant parE gene (4) andparC + or the resistance mutant parCgene (equivalent to grlA) in E. coli(16). In contrast, the dominance ofgrlA + by gene dosage in S. aureus was reported to require the additional presence of plasmid-encodedgrlB (41). This discrepancy might be explained ifgrlA alone was not expressed from the complementary plasmid. Our preliminary results indicated that the grlA gene cloned alone with its putative promoter is not expressed (or is only poorly expressed) and may need the grlB promoter to be expressed (data not shown).
The grlB543 mutation also increases coumarin susceptibility in both nov and nov + cells (Table 4). Coumarins act by competitively inhibiting ATP binding to the B subunit of DNA gyrase (10, 12). All of the mutations that result in coumarin resistance lie at the periphery of the ATP-binding site of the B subunit of gyrase (6, 7, 13, 22, 34, 36) (Fig. 1). Recently, the crystal structure of a complex between the 24-kDa N-terminal fragment of the E. coli DNA gyrase B protein and novobiocin was described (20). Novobiocin and ATP, which are dissimilar in chemical structure, bind to the protein in different ways, but their binding is competitive because of the overlap of their binding sites (Fig. 1) (20). Hypersusceptibility to coumarins might be envisioned to occur if topoisomerase IV containing GrlB543 had reduced affinity for ATP. In such a circumstance coumarins become more effective as competitors of ATP binding. Unfortunately, no crystal structure that includes both the region of amino acid 470 of GrlB (or its equivalent) and the coumarin-binding site has yet been reported.
The 50% inhibitory concentration (IC50) of novobiocin for inhibition of decatenation of wild-type S. aureustopoisomerase IV (30 μg/ml) (3) is 54- to 167-fold higher than the IC50s of novobiocin for inhibition of DNA supercoiling by wild-type S. aureus DNA gyrase (0.18 to 0.56 μg/ml) (23, 26). Thus, in S. aureus, topoisomerase IV is considerably less sensitive to novobiocin than DNA gyrase. In the gyrB142 background, topoisomerase IV may, however, become the actual target of novobiocin when gyrase is first protected from inhibition by mutation, since grlB543 is epistatic to gyrB142 for coumarin susceptibility.
Codominance for coumarin resistance was observed when thegrlB + or grlB543 alleles were on multicopy plasmids. In contrast, when plasmids were integrated into the chromosome, the wild-type gene appeared dominant over the mutant gene (Table 5). The codominance of fluoroquinolone resistance in these chromosomal merodiploids makes it unlikely that the dominance ofgrlB + for coumarin susceptibility is due to a modification of the expression of the grlB gene integrated in the lipase gene. One possible explanation for this finding is that GrlB543 subunits form complexes with GrlA subunits poorly. WhengrlB543 is hyperexpressed on plasmids, the GrlB543 subunits are in excess and form mutant enzymes in sufficient numbers to affect the cell phenotype, but when GrlB543 and GrlB+ subunits are present in similar numbers, holoenzyme complexes form dominantly with GrlB+ subunits.
The gyrA (Leu-84) mutation is silent for fluoroquinolone resistance in the absence of grlA and grlBmutations, but unexpectedly GyrA (Leu-84) causes resistance to nalidixic acid, a nonfluorinated member of the quinolone group. The opposite effect is observed with grlA and grlBmutations: increased MICs of fluoroquinolones (ciprofloxacin and norfloxacin) but no modification of the MIC of nalidixic acid (Table6). Although chromosomal grlA + orgrlB + is epistatic to chromosomal mutantgyrA in determining fluoroquinolone susceptibility (25), this is not the case for nalidixic acid susceptibility. This pattern suggests that the fluorine substituent at position 6 on the quinolone nucleus may specifically enhance activity against topoisomerase IV.
When the gyrA mutation was combined with the grlAmutation, MICs of all tested fluoroquinolones were substantially increased in comparison to those associated with the individual mutations. In contrast, the combination of a gyrA mutation with a grlB mutation only increased the MIC of sparfloxacin. The consequences of a grlA mutation in the presence of a gyrA mutation differ from those of a grlBmutation, indicating that the combined effects of these two mutations are different. Also noteworthy is the observation that for sparfloxacin and DU6859a either a grlA or a gyrA mutation alone had little or no effect on activity. The combination of these same two mutations, however, produced substantial increments in resistance (16-fold for DU6859a and 128-fold for sparfloxacin). This pattern is consistent with the hypothesis that these drugs have similar activities in vivo against both DNA gyrase and topoisomerase IV (4). The findings with nalidixic acid, which appears to target gyrase, and those with other fluoroquinolones (ciprofloxacin and norfloxacin), which appear to target topoisomerase IV, amplify previous observations for S. pneumoniae (29), indicating that either or both enzymes may be targets of the quinolone drugs depending on drug structure. A study of a larger series of compounds should allow better definition of structural features that determine the selectivity of drug targeting. Our findings also indicate that alterations in topoisomerases may have pleiotropic effects not only on different classes of inhibitors such as fluoroquinolones and coumarins but also on inhibitors within the same class such as ciprofloxacin and sparfloxacin for grlB and grlA mutations or nalidixic acid and oxolinic acid for the gyrA mutation. Full understanding of drug action and resistance at the molecular level must take into account both inhibitor structure-activity relationships and the effects of different classes of topoisomerase mutants.
In summary, the nov-142 locus in MT5 has a double mutation in the GyrB protein: Ile102Ser and Arg144Ile, which likely explains the high level of resistance of the strain to coumarins. ThegrlB mutation encoding the Asn470Asp substitution in the GrlB protein of MT5224c9 causes an increase in quinolone resistance and a decrease in coumarin resistance. The combination of grlA(Phe-80) and grlB (Asp-470) mutations with thegyrA (Leu-84) mutation had differing effects on drug susceptibility. For sparfloxacin and DU6859a the pattern of minimal to no resistance caused by single grlA, grlB, andgyrA mutations and substantial resistance caused by the same mutations together suggests these drugs have dual primary targets.
ACKNOWLEDGMENTS
We thank G. L. Archer for providing plasmid pSK950 and C. Y. Lee for plasmid pYL112Δ19. We also thank J. C. Wang for helpful discussions.
This work was supported in part by U.S. Public Health Service grant AI23988 (to D.C.H.) from the National Institutes of Health and by a grant from Daiichi Pharmaceuticals.
FOOTNOTES
- Received 28 July 1997.
- Returned for modification 20 September 1997.
- Accepted 30 October 1997.
- Copyright © 1998 American Society for Microbiology
