Analysis of Mupirocin Resistance and Fitness in Staphylococcus aureus by Molecular Genetic and Structural Modeling Techniques

The mutants described in the present study are derivatives of S. aureus 8325-4 (22). Strains were cultured aerobically at 37°C by using Iso-Sensitest broth (ISB) or Iso-Sensitest agar (ISA) (Oxoid, Basingstoke, United Kingdom).

The lithium salt of mupirocin was a gift from Glaxo SmithKline Pharmaceuticals (Harlow, United Kingdom). MICs were determined by agar dilution on ISA with an inoculum in ISB of 10⁶ CFU/spot (31). The MIC was defined as the lowest concentration of mupirocin that inhibited visible growth after 18 h of incubation at 37°C.

Mutation frequencies were determined by previously described methods (23). The mutant prevention concentration (MPC) is defined as the lowest concentration of an antibiotic that prevents bacterial colony formation from a culture containing 10¹⁰ CFU (33). Accordingly, late-logarithmic-phase cultures of the S. aureus 8325-4 growing in ISB were harvested by centrifugation and resuspended in fresh ISB at a final concentration of 10¹¹ CFU/ml. Aliquots (200 μl) were incorporated into ISA plates containing serial twofold mupirocin dilutions in the range 0.06 to 1.0 μg/ml. The MPC was determined after incubation at 37°C for 48 h.

Maximum doubling time and relative competitive fitness determinations were performed exactly as described previously (15).

PCR amplification and sequencing of ileS genes from genomic DNA was performed as described previously (15).

As a measure of in vivo fitness, the survival of S. aureus strains was determined in a murine wound abscess infection model (14). Mice were subcutaneously injected with 10⁸ bacteria suspended in phosphate-buffered saline. After 7 days, the mice were killed, and the abscesses were recovered and homogenized in ISB. Bacteria in the homogenized lesions were quantified by plating on ISA, and the survival of mupirocin-resistant mutants was compared statistically with strain 8325-4 by using one-way analysis of variance and evaluation of minimum significance difference (P < 0.05) between means by the T method (27). The stability of mupirocin resistance genotypes during infection was confirmed by comparing the growth rates and susceptibility to mupirocin of in vivo isolates with their starting strains.

Fitness-restored mutants were selected from unfit parent strains after serial transfer in nonselective ISB over a period of 7 to 9 days. Cultures of mupirocin-resistant mutants were grown in ISB at 37°C for 24 h, followed by transfer of aliquots to fresh ISB to initiate cultures containing ca. 10⁶ CFU/ml. Serial transfer was continued until an increase in growth rate was detected, which is indicative of increased fitness in the population. Cultures were then plated onto nonselective ISA, and large, fast-growing colonies were selected for further study.

Molecular models of the mutated enzymes were constructed by using the Swiss-PDB viewer v3.7 software (11) in conjunction with the structure files 1FFY.pdb (26) and 1JZQ.pdb (21). The lowest-energy conformations of the resulting mutated enzymes were then subjected to full energy minimization by using the MM2 force field within MacroModel (18) on Silicon Graphics O2 or Indy workstations. In order to simplify the computations, all calculations were performed on a 25-Å spherical portion of the protein centered on the ligand. For the energy minimizations, a 12-Å spherical region within this substructure centered on the ligand was defined in which all atoms, including those from the ligand, were allowed to freely move and atoms outside this region were frozen.

Selection of antibiotic-resistant mutants through chromosomal mutations occurs in a mutant selection window in which the drug concentration lies between the MIC and the MPC (33). A mutant selection index of 31 (i.e., MPC divided by MIC) was obtained for S. aureus 8325-4 on the basis of an MPC of 0.5 μg/ml and MIC of 0.016 μg/ml. This indicates an intermediate vulnerability to the development of mupirocin resistance in S. aureus, which is less than that exhibited by rifampin but greater than that displayed by the majority of fluoroquinolones (33). We isolated and characterized mutants of strain 8325-4 that arose spontaneously after selection with 0.064 μg of mupirocin/ml (four times the MIC). The mutants displayed low-level resistance to mupirocin with MICs in the range 1 to 16 μg/ml (Table 1). The mutants arose with an overall frequency of (7.2 ± 0.9) × 10⁻⁸, which is consistent with single mutational events. Several mutants selected for sequencing of ileS revealed the V₅₈₈F and V₆₃₁F IRS amino acid substitutions also found in clinical isolates (Table 1). However, a third mutant class was identified involving a G₅₉₃V mutation (Table 1). This genotype has not previously been identified in mupirocin-resistant S. aureus, but mutation at the equivalent locus in Methanobacterium thermoautotrophicum Marburg (G₅₈₉D) confers mupirocin resistance (17) (Fig. 1).

Mutants of S. aureus exhibiting elevated levels of resistance to mupirocin (MIC ≥ 16 μg/ml) have been isolated in vitro after stepwise exposure to increasing concentrations of the antibiotic (28). Although these resistant derivatives probably arise by accumulation of separate mutations within ileS, there are no DNA sequence data to support this hypothesis. Strains JGH-1, JGH-4, JGH-5, JGH-6, and JGH-8, which contain single mutations in ileS (Table 1), were therefore used as starting strains for the selection of higher-level resistant second-step mutants.

Second-step mutants were selected by plating onto ISA containing mupirocin at four times the respective MICs for the starting strain. This resulted in the recovery of mutants for which mupirocin MICs were in the range 16 to 128 μg/ml (Table 1). Short regions of the ileS were resequenced and resistance mutations detected. However, where no mutations were identified (JGH-13, JGH-15), the entire ileS gene was sequenced to reveal second-step mutations. In many cases the second mutation occurred at one of the sites (V₅₈₈F or V₆₃₁F) already identified in low-level mutants, resulting in the generation of double V₅₈₈F/V₆₃₁F or G₅₉₃V/V₆₃₁F mutants (Table 1). Antonio et al. (2) have suggested that V₆₃₁F and V₅₈₈F mutations may be mutually exclusive since considerable distortion of the hydrophobic domain of IRS would occur. Nevertheless, we show here that these mutations can coexist in IRS. Similarly, G₅₉₃V and V₆₃₁F mutations can be simultaneously tolerated in IRS. Three second-step mutations occurred at loci not previously implicated in mupirocin resistance (Table 1) involving the following changes: R₈₁₆C (strain JGH-13), F₅₆₃L (strain JGH-14) and H₆₇Q (strain JGH-15).

Second step mutants arose with an overall frequency in the range of 4 × 10⁻¹⁰ to 8 × 10⁻¹⁰, which is less than the frequency for selection of first-step mutants. The reason that second-step mutations occur less frequently is unclear. However, this does not appear to be related to mutational target size since the numbers of sites at which new mutations arose during first- and second-step selections were similar (Fig. 1).

Resistance to antibiotics arising from mutations in target genes can result in fitness costs that affect bacterial growth and survival (1, 20). Fitness costs associated with development of mutational resistance to mupirocin could be a factor that influences the prevalence of certain staphylococcal resistance genotypes in the clinic. In preliminary studies, we recently reported that the V₅₈₈F mutation in clinical isolates of GISA is associated with low fitness costs (15). The fitness of other genotypes has not been examined, but it is possible that fitness costs may account for the low frequency of detection of the V₆₃₁F resistance allele and, similarly, the apparent absence of G₅₉₃V in clinical specimens. We have now expanded our preliminary analysis by determining the in vitro fitness of a number of strains derived from 8325-4 that contain V₅₈₈F, V₆₃₁F, or G₅₉₃V mutations in IRS. Fitness was assessed by determining the doubling times of mutants and also by comparing their performance with the parental strain 8325-4 in mixed culture competition assays (Table 1 and Fig. 2).

The doubling times of six mutants containing the V₅₈₈F alteration, two mutants containing the V₆₃₁F alteration and two mutants containing the G₅₉₃V alteration were not considerably different from the parental starting strain 8325-4 (Table 1). In mixed culture competition assays five of six strains containing the V₅₈₈F mutation displayed a small loss of fitness, but the relative competitive fitness (W) never fell to <0.9. The G₅₉₃V mutation was also associated with a small loss (6 to 8%) of fitness detectable by the mixed culture competition assay.

Strains carrying the V₆₃₁F mutation (JGH-3 and JGH-4) were as fit as the parental strain (Table 1). However, although there were no significant differences in doubling times, there was an apparent reduction in W of ∼0.2 for three other strains containing V₆₃₁F mutations (data not shown). In other systems, it is well known that compensatory mutations can arise in unfit antibiotic-resistant mutants that restore fitness and yet retain resistance (1, 20). Therefore, the V₆₃₁F mutation might be initially associated with a small fitness cost that is ameliorated during subculture through intragenic or extragenic suppression. Nevertheless, no evidence for intragenic compensatory adaptation was found in fit strains possessing the V₆₃₁F mutation.

The relatively modest fitness costs associated with the V₅₈₈F and V₆₃₁F mutations are in agreement with our earlier studies on GISA and consistent with the occurrence of these resistance genotypes in clinical isolates (2, 8, 15). Since the G₅₉₃V resistance mutation does not appear to incur significant fitness costs, at least in vitro, it might arise in clinical isolates. Currently, this mutation has not been identified in clinical strains (2, 8, 15). However, more extensive surveys may reveal this genotype in the future.

In addition to the very low mutation frequencies for the generation of double ileS mutants, another reason for the absence of such bacteria in clinical isolates might be poor fitness of strains with two mutations in IRS. Indeed, second-step mupirocin-resistant mutants displayed reduced fitness both in vitro and in vivo (Table 1 and Fig. 3).

The doubling time of mutants containing two ileS mutations was longer than that for first-step mutants or for the original mupirocin-sensitive parental strain 8325-4. For example, JGH-15 (carrying mutations V₅₈₈F and H₆₇Q) grew approximately three times more slowly than first-step mupirocin-resistant mutants or 8325-4. The second-step mutants also displayed loss of fitness in the mixed-culture competitive fitness assay with W values from 0.24 to 0.63 (Table 1).

As previously mentioned, the exponential growth of resistant bacteria was affected by mupirocin resistance, and this was most pronounced in second-step mutants. Furthermore, a positive (R² = 0.89), correlation was observed between the doubling times and W of mupirocin-resistant staphylococci (Fig. 2). Consequently, at least in the case of mupirocin resistance in S. aureus, growth rate determinations provide a simple representative measure of fitness trends without the need for laborious mixed-culture competition assays.

Significant costs of resistance observed in vitro usually indicate situations where resistance would impose a fitness burden on the organism in natural populations (1). Nevertheless, the survival of S. aureus 8325-4 and selected first- and second-step mupirocin-resistant mutants, which carry mupirocin resistance genotypes previously identified in clinical strains, was also examined in a murine wound abscess infection model (Fig. 3). The number of viable organisms in the infectious dose was established at the start of the experiment, and the number of survivors after 7 days of infection was determined by plating samples of homogenized skin lesions onto ISA. The surviving fraction of strain JGH-12 (double IRS mutant, V₆₃₁F/V₅₈₈F) was significantly lower than the wild-type 8325-4 (Fig. 3). In contrast, survival of the first-step mupirocin-resistant mutants JGH-4 (V₆₃₁F) and JGH-6 (V₅₈₈F) was not significantly different from 8325-4. The mupirocin MICs and doubling times of independent colonies recovered from the mice were similar to the values for the strains used to initiate the infections (data not shown). This implies genetic stability of mupirocin resistance within ileS during the murine infections.

Although the experiments reported here did not reveal a statistically significant difference in the in vivo fitness of strains 8325-4 and JGH-6, there is, nevertheless, an indication that the V₅₈₈F mutation in strain JGH-6 might be associated with increased fitness compared to the parental starting strain 8325-4 (Fig. 3). However, further experiments involving infection of five mice with mixed cultures of strains 8325-4 and JGH-6 provided no evidence that JGH-6 was fitter than 8325-4 (data not shown).

Overall, the murine data provide a correlation between in vitro and in vivo fitness and further indicate that the clinically occurring V₅₈₈F and V₆₃₁F mutations are not associated with significant fitness costs, whereas second-step mutants, such as JGH-12 (V₆₃₁F/V₅₈₈F), which have not been reported clinically, possess significant fitness burdens.

In the absence of continued antibiotic selection pressure, the fitness of unfit antibiotic-resistant mutants can be restored by several mechanisms, one of which involves reversion of resistant alleles by back mutation (1). Consequently, if mupirocin-resistant mutants with more than one mutation in ileS can arise in the clinic, the failure to detect them in clinical isolates could reflect reversion to produce fitter strains containing only one resistance allele. We therefore examined whether resistance was lost and fitness restored when unfit second-step mupirocin-resistant mutants (JGH-11 to JGH-15) were serially subcultured in the absence of mupirocin for up to 9 days by daily transfer into fresh medium.

At the end of each subculture, samples were taken for determination of mupirocin MICs. Serial subculture of mupirocin-resistant strains JGH-11, JGH-13, JGH-14, and JGH-15 led to progressive reduction of organisms expressing higher levels of resistance to mupirocin within the cultures and replacement by derivatives for which mupirocin MICs were lower (Fig. 4). For instance, the MIC of mupirocin against JGH-15 was initially 128 μg/ml. However, after serial subculture in the absence of mupirocin for 4 days, the culture contained daughter cells with only low-level resistance (mupirocin MIC = 8 μg/ml), which remained stable for the remainder of the serial subculture. With the exception of JGH-12 (see below), similar results were obtained with the other higher level mupirocin-resistant mutants, which all gave rise to stable, lower-level mupirocin-resistant populations within 4 to 7 days of serial subculture (Fig. 4).

In contrast to the other higher level mupirocin-resistant strains, resistance was stable in strain JGH-12 during the course of 9 days serial subculture since starting and final MICs were 32 μg/ml. However, organisms recovered at the end of this experiment were fitness compensated (see below).

During serial subculture the emergence of lower-level mupirocin-resistant subpopulations in cultures of JGH-11, JGH-13, JGH-14, and JGH-15 coincided with progressive increases in growth rates of the mixed populations (Fig. 4). This is consistent with reversion of one of the mutated alleles in the ileS gene of the parent strains to produce derivatives with restored fitness and reduced resistance or accumulation of further mutations that act to suppress both the resistance phenotype and fitness cost of the existing allele (pseudoreversion). Elevated growth rates were also recorded in cultures of JGH-12 after 3 days of subculture even though resistance of JGH-12 to mupirocin was unaltered through 9 days of serial subculture. This probably reflects the appearance of an intragenic or extragenic compensatory mutation that restores fitness but allows resistance to be retained (1).

The more rapidly growing colonies obtained from the mupirocin-resistant strains were therefore presumptive fitness-restored mutants. A number of these colonies were retained for further analysis. They were designated strains JGH-16 to JGH-27 (Table 1). Strains JGH-16 to JGH-27 exhibited doubling times (range, 38 to 52 min) and W values (range, 0.79 to 1) that indeed indicated partial or complete restoration of fitness compared to their original unfit parental starting strains (Table 1).

To determine whether compensatory or back mutations were responsible for fitness restoration, the entire ileS gene of the fitness-restored mutants was sequenced and mutations were identified. Back mutation of the second-step mutation (Q₆₇H) in strain JGH-15 to create JGH-25 restored fitness by reinstating the HMGH sequence (Fig. 1 and Table 1). However, back mutation of resistance alleles was not detected in other fitness-restored strains. Intragenic and extragenic compensatory mutations that suppressed the effects of the resistance genotypes were inferred from sequencing data. For example, in strains JGH-22 and JGH-23 (both derived from JGH-13) either E₁₉₅K (JGH-22) or A₁₉₆V (JGH-23) mutations were found, in addition to the previously identified resistance mutations V₅₈₈F and R₈₁₆C. One strain, JGH-21 (derived from strain JGH-13) seemed to have acquired an extragenic mutation in addition to the compensatory mutation E₁₉₅K. Acquisition of this uncharacterized extragenic mutation may explain why strain JGH-21 showed greater susceptibility to mupirocin (MIC = 0.5 μg/ml) compared to JGH-22 (MIC = 4 μg/ml). Compensatory mutations (E₁₉₀K or A₁₉₆V) were detected in strains JGH-16 to JGH-19 (each derived from JGH-11) in regions adjacent to compensatory mutations found in JGH-22 and JGH-23 (each derived from JGH-13). Intragenic mutations were not found in JGH-20 (derived from JGH-12) or JGH-24 (derived from JGH-14), a finding consistent with a role for extragenic compensatory mutations in restoring the fitness of these strains.

Restoration of fitness by intragenic suppression usually involves compensation by second site mutations that restore fitness and retain resistance (1, 20). Therefore, the situation described here is somewhat unusual since intragenic (and indeed some extragenic) mutations produced derivatives that exhibited lower-level resistance to mupirocin, despite retention of both original mutations that conferred high-level resistance. Fitness costs may be alleviated in these pseudorevertants by introduction of compensatory mutations that alter IRS conformation, restoring catalytic activity and susceptibility to mupirocin, despite the presence of mutations previously conferring high-level mupirocin resistance whose expression is now partially suppressed.

In order to probe the molecular effects of mutations upon mupirocin binding, we constructed molecular models of both the first- and second-step IRS mutants and used these to both quantify the effect of the mutations upon the binding energy for mupirocin and to try to establish the precise interactions responsible for any changes in binding energy. The calculated binding energies of mupirocin within these complexes are summarized below (Table 2). In general, there appears to be a broad correlation between the calculated mupirocin-IRS binding energy and the measured mupirocin MIC.

As can be seen, all of the mutant enzymes contained lower mupirocin-protein binding energies than that calculated for the wild type (Table 2). In addition, the most dramatic decreases were found within the second-site mutants, in keeping with elevated levels of mupirocin resistance. Detailed examination of the molecular models appeared to reveal some of the structural factors affecting the mupirocin binding of these mutants.

Glycine 593 is located in the loop that makes direct contact with the tRNA via a hydrogen bond between the backbone carbonyl and the 2′-OH of G-68 (Fig. 5). Substitution for valine locates the new hydrophobic side chain close to this hydroxyl and appears to result in a net repulsion. This induces a slight movement of the loop toward the side chain of the neighboring methionine 589, which in turn appears to move the backbone carbonyl unit of valine 588 nearer to the hydrophobic region of mupirocin bordering C-8′ and C-9′. The resulting repulsion also appears to lengthen the H-bond between the valine 588 NH and the O = C1 carbonyl of mupirocin by 0.2 Å (Fig. 5). This apparent relaying of the steric interaction experienced within the newly mutated residue also appears to move some of the carbon atoms within the fatty acid side chain of mupirocin. Since this resides in a fairly narrow hydrophobic channel, this movement may also contribute to weaker binding of antibiotic to the enzyme.

In the wild-type enzyme, valine 588 lines the hydrophobic pocket that interacts with the fatty acid side chain of mupirocin. Upon mutation to phenylalanine, the increased bulk of the benzyl side chain results in steric repulsion at C-1 and C-2 of mupirocin and lengthens the valine 588 NH/O = C1 H bond by 0.67 Å (Fig. 5).

As in the case of G₅₉₃V, V₆₃₁F appears to affect mupirocin binding via a “steric relay” mechanism. Valine 631 is stacked above valine 588 and, upon mutation to phenylalanine, the increased bulk of the benzyl side chain repels that of valine 588. The resulting torsional movement pushes the carbonyl oxygen of valine 588 closer to mupirocin and again, distorts the fatty acid side chain of the antibiotic, and weakens the valine 588 derived H bond, which becomes lengthened by 0.22 Å (Fig. 5).

Increasing the size of the side chain at position 631 from isopropyl to benzyl in combination with the effect of the G₅₉₃F mutation, as described previously, increases further the steric repulsion between the side chains at positions 631 and 588 and, again, leads to further stretching of the V₅₈₈-mupirocin H bond by 0.13 Å and, perhaps more importantly, further increases the distortion along the fatty acid side chain of the antibiotic (Fig. 5).

Since these two amino acids are essentially stacked one above the other, the steric repulsion between the two benzyl side chains in this mutant appears to be considerable such that the movement of F₅₈₈ away from F₆₃₁ appears to abolish the F₅₈₈-mupirocin H bond with considerable distortion of the mupirocin fatty acid moiety by up to 1.18 Å from its position in the wild-type enzyme.

In addition to the effect of the V₅₈₈F mutation via its direct interaction with mupirocin as noted above, the F₅₆₃L mutation appears to be a further example of a “relayed” steric clash with mupirocin. The calculated structure indicates that the leucyl side chain at position 563, which is located within a looped region, will undergo steric repulsion with L₅₈₂, which is directly opposite on the adjacent peptide chain. This appears to result in torsional strain within the loop and in particular moves the side chain of tryptophan 562 close to the C-14 methyl group of mupirocin.

As for the previously described double mutant, in addition to the direct effect of the V₅₈₈F mutation on mupirocin binding, the H₆₇Q mutation also appears to directly affect the interaction with mupirocin and locates the carbonyl group from the amide side chain of glutamine in close proximity to the hydrophobic portion of the sugar-like ring of mupirocin, centered around C-16 and C-8.

It has been reported (3) that an ethyl ester variant of mupirocin (therefore containing a much-shortened alkyl chain) shows a 10-fold reduction in activity against S. aureus IRS compared to mupirocin (k_i = 0.19 versus 0.017 nM, respectively). Using the equation ΔG = −RT × lnk_i (wherein ΔG is the change in free energy, R is the gas constant, T is the temperature, and k_i is the dissociation constant of the enzyme-inhibitor complex), free-energy changes upon dissociation (ΔG) of 13.25 and 14.68 kcal/mol, respectively, are obtained. Clearly, the magnitudes of these measured free energies reflect the total change in the energetics of the system during binding (including changes in solvation, entropy, bond enthalpies, etc.). Therefore, they cannot be directly compared to the calculated gas-phase enthalpies of binding shown in Table 2, which simply reflect the relative changes in bond enthalpy between substrate and enzyme resulting from substrate binding. Nevertheless, the reduced affinity of the shortened alkyl chain variant for IRS compared to that of mupirocin is consistent with our present findings concerning the effect of mutations on the region of the protein bordering the binding pocket of the alkyl chain of mupirocin.

The variation in fitness costs associated with the various mutants could result from changes in the binding of the enzyme to the reactive isoleucyl-AMP intermediate generated during acylation of the tRNA. We have therefore probed the structural details within a series of models that examine the binding energies of the reactive isoleucyl-AMP intermediate (Fig. 6, structure 2) generated during acylation of the tRNA. The models of this reactive complex were constructed first by docking a model of structure 2 into the enzyme active site by using the automated molecular docking program AutoDock (19). The lowest-energy complex was then used to create the various mutant structures, which were then subjected to energy minimization as described above. To date, there is no available X-ray structural data for the S. aureus IRS containing the reactive intermediate, with which direct comparison of IRS/inhibitor models could be made.

However, the X-ray crystal structure of a sulfonamide-based inhibitor (Fig. 6, structure 3), designed to mimic a reactive intermediate (Fig. 6, structure 2) with an IRS, has been reported (21). Comparison of the models with this structure reveals remarkable similarity with essentially all of the observed inhibitor/enzyme contacts present within the active complex/enzyme models (Fig. 7). In particular, hydrogen bonds are predicted between the 2′-OH and backbone N of Q₅₅₄; the 3′-OH and side chain of Q₅₅₄; a phosphate oxygen and backbone NH of Y₅₈, and between the leucyl NH₃ and the carbonyl of P₅₆. Similarly, the purine ring takes part in H bonding involving N5 and the backbone NH of V₅₈₈, and between N7 and H₅₈₅ (Fig. 7). In addition, the reactive intermediate and mupirocin bind in very similar regions within the enzyme (Fig. 8), with the epoxide ring of mupirocin occupying the same space as the phosphate unit in the intermediate. The central sugar core of both systems is also seen to overlap. The close similarity in binding regions for mupirocin and the reactive intermediate may further explain why major conformational changes that result in higher-level resistance also incur significant fitness reductions. However, an important difference occurs in the region of the extended hydrophobic side chain of mupirocin, which is essentially unoccupied in the reactive intermediate-containing models.

Table 2 lists the calculated binding energies of the model of the reactive intermediate within the mutant enzymes. Interestingly, the binding energies within all of the first-step mutants are essentially the same as that calculated for the wild-type system. In particular, as outlined above, the G₅₉₃V mutation occurs in a region several angstroms away from the phosphorylated intermediate binding cavity. Similar observations apply for the V₆₃₁F and V₅₈₈F first-step mutations (Fig. 9). These mutations appear to have little effect upon the binding of the acylated intermediate due to the absence of any part of the ligand in these regions.

For the second-step mutants the calculated binding affinities show a marked decrease relative to the first-step mutants, with that from V₅₈₈F/F₅₆₃L exhibiting the lowest value (Table 2). Inspection of the models reveals the origins of the decreased binding energies. For the G₅₉₃V/V₆₃₁F mutant, the presence of the V₆₃₁F mutation in particular appears to induce a small movement in the adjacent peptide strand at V₅₈₈ which in turn induces a further movement (by ∼0.55 Å) in the neighboring peptide loop at position M₆₅. This widens the hydrophobic pocket holding the purine ring of the substrate, and results in movement of the substrate to better fit the wider pocket. This in turn alters the hydrogen bond distances between the substrate and enzyme in the region of the isoleucine fragment, most noticeably involving Y₅₈ NH and one of the phosphate oxygens, which is lengthened by ∼0.1 Å and also that between the NH₃ of the substrate and the side chain of D₉₅, which increases in length by 0.2 Å. This widening of the pocket binding the purine ring is also seen in the V₆₃₁F/V₅₈₈F mutant and analogous side chain interactions appear to induce the same type of distortion in substrate binding and H-bond lengths.

The V₅₈₈F/F₅₆₃L mutation also appears to disrupt the H bonding and perhaps also the hydrophobic interactions at the isoleucyl end of the substrate. This seems to result from loss of the edge-to-face interaction between the original aromatic ring present at F₅₆₃ and the phenolic ring of Y₅₅₉. This appears to widen the hydrophobic cavity carrying the isoleucyl chain of the substrate and also lengthens the H bonds within this region, particularly that between the NH₃ of the substrate and the side chain of D₉₅ (by 0.23 Å).

This same distortion of the isoleucyl binding pocket also appears to be responsible for the decreased affinity observed for the substrate in the V₅₈₈F/H₆₇Q mutant. In this case, however, it originates from loss of the original face-to-face interaction involving the imidazole rings of H₆₇ and H₆₄. In the mutant, the amide side chain at Q₆₇ is forced to lie nearer the substrate resulting in a steric repulsion, which again lengthens the H bond at the NH3 and phosphate positions of the substrate by 0.1 and 0.2 Å, respectively.

The general trend within the binding energies broadly correlates with the fitness cost measurements and appears to show that the fitness within the various mutant IRS enzymes is a direct result of changes in the efficiency of binding of the key isoleucyl phosphate ester intermediate.

Mutations conferring antibiotic resistance with only low fitness costs may survive for very long periods in natural populations since there will be no counterselection favoring loss of resistance. The fitness studies reported here suggest that first-step V₅₈₈F and V₆₃₁F ileS mutations conferring low-level mupirocin resistance are low-cost mutations both in vitro and in vivo, a finding consistent with the occurrence of these mutations in clinical isolates (2, 8).

We observed in the laboratory that higher levels of mupirocin resistance could be generated by selection of second-step mutants from first-step hosts. Such second-step mutants exhibited fitness costs associated with the accumulation of a second mutation in ileS. Studies on fitness costs imposed by other antibiotic resistance mutations (1, 20) do not necessarily preclude the clinical appearance of mupirocin-resistant mutants with more than one ileS mutation because compensatory mutations to alleviate fitness might occur. Indeed, the studies presented here show that fitness burdens imposed by two mutations in ileS are more frequently relieved by independent compensatory mutations rather than back mutations in the specific allele conferring resistance, at least in vitro. Nevertheless, there is currently no evidence that mupirocin-resistant mutants with more than one mutation in ileS arise in the clinic (2, 8, 15). Furthermore, although several mutations of uncertain significance have been located in the ileS of clinical isolates (2, 8), none of them corresponds to the intragenic compensatory mutations we report here in laboratory strains. We therefore conclude that clinical selection of mupirocin-resistant strains possessing two mutations in ileS, if it occurs, must be a rare event. There are a number of possible explanations for this situation. In the first instance the probability of a strain simultaneously acquiring two mutations in ileS appears to be very low since combination of the individual mutation frequencies for generation of first- and second-step mutants provides a theoretical simultaneous two-step mutation frequency of ∼10⁻¹⁸. Nevertheless, stepwise selection of two mutations conferring resistance could occur as a result of poor compliance or suboptimal application of mupirocin. Consequently, lack of fitness is the most likely explanation for the absence of resistant double mutants in clinical isolates.

In addition to the clinical implications of the work reported here, we provide a structural understanding of the mutations conferring mupirocin resistance and their effects upon bacterial fitness. For example, there was a correlation between binding energies for mupirocin-IRS interactions and degree of resistance to the antibiotic. Furthermore, the lack of fitness in some mupirocin-resistant mutants could be explained by decreased binding of the reactive intermediate Ile-AMP in the IRS enzyme. Unfortunately, it was not possible to extend the structural studies to the third-site fitness-compensated mutants since these were generally located at positions outside the 12-Å sphere centered on mupirocin within which the energy calculations could be performed.