In Vitro Selection and Characterization of Ceftobiprole-Resistant Methicillin-Resistant Staphylococcus aureus

Abstract

Methicillin-resistant Staphylococcus aureus (MRSA) is resistant to β-lactam antibiotics because it expresses penicillin-binding protein 2a (PBP2a), a low-affinity penicillin-binding protein. An investigational broad-spectrum cephalosporin, ceftobiprole (BPR), binds PBP2a with high affinity and is active against MRSA. We hypothesized that BPR resistance could be mediated by mutations in mecA, the gene encoding PBP2a. We selected BPR-resistant mutants by passage in high-volume broth cultures containing subinhibitory concentrations of BPR. We used strain COLnex (which lacks chromosomal mecA) transformed with pAW8 (a plasmid vector only), pYK20 (a plasmid carrying wild-type mecA), or pYK21 (a plasmid carrying a mutant mecA gene corresponding to five PBP2a mutations). All strains became resistant to BPR by day 9 of passaging, but MICs continued to increase until day 21. MICs increased 256-fold (from 1 to 256 μg/ml) for pAW8, 32-fold (from 4 to 128 μg/ml) for pYK20, and 8-fold (from 16 to 128 μg/ml) for pYK21. Strains carrying wild-type or mutant mecA developed six (pYK20 transformants) or four (pYK21 transformants) new mutations in mecA. The transformation of COLnex with a mecA mutant plasmid conferred BPR resistance, and the loss of mecA converted resistant strains into susceptible ones. Modeling studies predicted that several of the mecA mutations altered BPR binding; other mutations may have mediated resistance by influencing interactions with other proteins. Multiple mecA mutations were associated with BPR resistance in MRSA. BPR resistance also developed in the strain lacking mecA, suggesting a role for chromosomal genes.

Methicillin-resistant Staphylococcus aureus (MRSA) is a major cause of nosocomial and community-associated illness throughout the world. Infection with MRSA results in diverse clinical manifestations ranging from minor skin infections to life-threatening bacteremia and pneumonia. The recent, alarming emergence of vancomycin-intermediate and -resistant MRSA strains highlights the need for new antibiotics to treat infections with this highly drug resistant pathogen.

MRSA is resistant to most β-lactam antibiotics because it has acquired the gene mecA, most likely by horizontal transfer from coagulase-negative staphylococcal species (29). mecA encodes the low-affinity penicillin-binding protein 2a (PBP2a) which, unlike other PBPs, remains active and allows for cell wall biosynthesis at otherwise lethal β-lactam concentrations. Ceftobiprole (BPR) is an investigational, broad-spectrum cephalosporin that binds to PBPs, including PBP2a, with high affinity. It is active against MRSA, as well as other gram-positive and gram-negative pathogens, and is stable against the staphylococcal β-lactamase. For most clinical isolates of MRSA and methicillin-sensitive S. aureus (MSSA), BPR MICs are <2 μg/ml, below the preliminary breakpoint for BPR of 4 μg/ml (21). In vitro, the MIC₉₀ of BPR for most MRSA and MSSA strains has been reported to be <4 μg/ml (1). BPR is currently in phase III clinical studies for the treatment of complicated skin and soft tissue infections, as well as pneumonia. It has favorable pharmacokinetics in vivo and is currently under review by the FDA (18).

Although staphylococci have a proven ability to develop resistance to most antibiotics in clinical use, the potential for MRSA to become resistant to BPR is thought to be low, based on results from in vitro studies. In an attempt to generate BPR-resistant mutants in vitro, Heller et al. passaged three different MRSA strains on BPR-containing agar. After 50 passages, the BPR MIC for the strains was only 32 μg/ml (9). Similarly, other investigators performed 50 passages of MRSA and MSSA strains in 1-ml broth cultures containing subinhibitory concentrations of BPR and found that the maximal BPR MIC achieved was 8 μg/ml, only a fourfold increase over that for the parental strain (2).

Investigators in our laboratory have previously generated an MRSA strain resistant to L-695,256, another investigational β-lactam that binds tightly to PBP2a (10). This strain, COL52, contains five amino acid changes in PBP2a, and the MIC of nafcillin for the strain is >2,000 μg/ml, more than eightfold greater than that for strains with wild-type PBP2a. We hypothesized that mutations in mecA, the gene encoding PBP2a, are necessary and sufficient to confer resistance to BPR. Unlike the investigators in the studies described above, we passaged large-volume broth cultures in BPR to increase the inocula and the potential for the identification of a resistant mutant.

MATERIALS AND METHODS

Reagents.

BPR solution was prepared fresh daily at a concentration of 2 mg/ml and was provided by Johnson and Johnson Pharmaceutical Research and Development. Tetracycline, cefazolin, vancomycin, and nafcillin were obtained from Sigma Chemical Co., St. Louis, MO. Tetracycline was used at 10 μg/ml.

Bacterial strains.

All strains were grown in Trypticase soy agar (TSA), Trypticase soy broth (TSB), or blood agar (Remel) at 37°C with aeration. Strains used in this study are listed in Tables 1 and 2. COLnex is the parental strain in which chromosomal staphylococcal cassette chromosome mec has been precisely excised by the introduction of plasmid pSR, which carries two site-specific-recombinase genes, ccrA and ccrB (11). Plasmid pAW8 contains a tetracycline resistance marker and has been described previously (11). Plasmid pYK20 contains wild-type mecA cloned into the BamHI site of pAW8; pYK21 contains mutant mecA derived from strain COL52 cloned into pAW8 (10).

TABLE 1.

Parental strains and phenotypes relevant to this study^a

Strain	Description	Phenotype	Reference
COL	Homogeneously methicillin-resistant strain	Mcr Tcr	26
COLn	Tcs derivative of COL	Mcr Tcs	11
COL52	Antibiotic-selected COLn variant with mutant mecA	Highly methicillin resistant; Tcs	10
COLnex	SCCmec excision strain derived from COLn	Mcs Tcs	11

^aSCCmec, staphylococcal cassette chromosome mec; Mc^r, methicillin resistant; Tc^r, tetracycline resistant; Tc^s, tetracycline sensitive; Mc^s, methicillin sensitive.

TABLE 2.

Derivatives of COLnex with plasmid-carried mecA used in this study^a

Strain	Description	Phenotype	Reference
COLnex(pAW8)	mecA-negative strain	Mcs Tcr	11, 26
COLnex(pYK20)	Strain expressing plasmid-carried wild-type mecA	Mcr Tcr	11
COLnex(pYK21)	Strain expressing plasmid-carried mutant mecA from COL52	Highly methicillin resistant; Tcr	11
COLnex(pAW8)D28	Post-BPR passage strain	BPR resistant; Tcr	This study
COLnex(pYK20)D28	Post-BPR passage strain	BPR resistant; Mcr Tcr	This study
COLnex(pYK21)D28	Post-BPR passage strain	BPR resistant; Mcr Tcr	This study
COLnex(pAW8)c	Post-BPR passage strain cured of plasmid	BPR resistant; Tcs	This study
COLnex(pYK20)c	Post-BPR passage strain cured of plasmid	BPR sensitive; Mcs Tcs	This study
COLnex(pYK21)c	Post-BPR passage strain cured of plasmid	BPR sensitive; Mcs Tcs	This study
COLnex(pAW8)T	COLnex transformed with post-BPR passage plasmid from COLnex(pAW8)D28	BPR sensitive; Tcr	This study
COLnex(pYK20)T	COLnex transformed with post-BPR passage plasmid from COLnex(pYK20)D28	BPR resistant; Mcr Tcr	This study
COLnex(pYK21)T	COLnex transformed with post-BPR passage plasmid from COLnex(pYK21)D28	BPR resistant; Mcr Tcr	This study

^aMc^r, methicillin resistant; Tc^r, tetracycline resistant; Tc^s, tetracycline sensitive; Mc^s, methicillin sensitive.

Multipassage selection in BPR.

COLnex(pAW8), COLnex(pYK20), and COLnex(pYK21) were serially passaged for 28 days in subinhibitory concentrations of BPR. For each strain, 300-ml preparations of TSB containing 10 μg of tetracycline/ml and various concentrations of BPR were inoculated at a 1:100 dilution with overnight cultures containing 10⁹ CFU/ml. For each strain, approximately 10¹² to 10¹³ cells are estimated to have been exposed over the entire 28 days. The BPR concentration was doubled with each passage, as tolerated.

Plasmid curing.

Strains were cured of plasmids by daily passaging in 10 ml of TSB lacking antibiotics for 7 days. Cultures were then streaked onto blood agar, and individual colonies were replica plated onto TSA plates with or without tetracycline. PCR amplification of the mecA gene was performed for those colonies that did not grow in tetracycline to confirm the loss of plasmid. Plasmids containing mecA were included in each PCR experiment as positive controls.

Spectrophotometric β-lactamase assay.

Broth cultures were grown overnight with tetracycline in the absence or presence of the β-lactamase-inducing antibiotic cefoxitin at a concentration of 0.5 μg/ml. Cells were pelleted by centrifugation, resuspended in 2 volumes of phosphate-buffered saline, and aliquoted into 0.9-ml fractions. After the addition of 0.1 ml of 1 mM cephaloridine, each tube was incubated at 37°C for 0, 30, or 60 min. Cells were then pelleted by centrifugation at 13,000 × g for 3 min, and the absorbance of the supernatants at 254 nm was read.

Population analyses.

Population analyses were done by the agar plate method, in which a sample of approximately 10⁹ CFU was serially diluted and quantitatively inoculated onto a series of TSA plates containing increasing concentrations of antibiotic. Drug plates were used within 1 week.

Growth curves.

Volumes of 100 ml of TSB with 10 μg of tetracycline/ml were inoculated at a 1:100 dilution with overnight broth cultures. Cultures were grown at 37°C with aeration, and the optical density at 578 nm was measured every hour for 6 consecutive hours by using a spectrophotometer. All growth experiments were done in duplicate.

DNA manipulations.

Primers used in this study were as follows: for mecA, K23 (5′-TCGTGTCAGATACATTTCGATTG-3′) and K27 (5′-GTTGTAGCAGGAACACAAATGAATAAC-3′); for the PBP1 gene, PBP1F (5′-AGCAACAACCACAAACTAAGC-3′) and PBP1R (5′-CCTCGTCTACCTTAAAATTCTC-3′); for the PBP2 gene, PBP2F (5′-TGCATATCAACAAAAAGGTATTG-3′) and PBP2R (5′-CTATTTAGATGTTTCAAAATGTATG-3′); for the PBP3 gene, PFP3F (5′-GTTTGTTTTCACGTGAACAGAA-3′) and PBP3R (5′-ATTTTGGAATGTAGTTAACTGGG-3′); and for the PBP4 gene, PBP4F (5′-GACATGACTGGGAAGGTGAATT-3′) and PBP4R (5′-TAACACCTTTAGCTACACACGT-3′). Genomic and plasmid DNA was prepared with the Qiagen DNeasy kit and the Qiagen Qiaprep spin miniprep kit, respectively, by using an initial cell lysis step with lysostaphin (Sigma). PCR was performed using 1.25 U of Taq DNA polymerase (AllStar Scientific) per 20 μl of reaction mixture and a PTC200 thermal cycler (MJ Research) under the following conditions: 95°C for 10 min followed by 30 cycles of 95°C for 30 s, 45°C for 30 s, and 70°C for 3 min. PCR products were analyzed by gel electrophoresis and purified using the Roche High Pure PCR product purification kit according to the manufacturer's instructions. Purified PCR products were sequenced by the University of California, Berkeley, sequencing facility.

Transformations were performed using an Eppendorf 2510 electroporator.

Molecular modeling.

Mutant proteins were produced in silico by using Protein Data Bank structures 1VQQ (S. aureus apo [unliganded] PBP2a residues 27 to 668) and 1MWS (S. aureus PBP2a residues 27 to 668 covalently bound to the cephalosporin nitrocefin) with UCSF Chimera (22). Mutant structures were subjected to energy minimization with Chimera by employing the AMBER1994 force field (4) with ANTECHAMBER charge determination (28). Modeling was confirmed independently with the Swiss-Pdb Viewer (8) energy minimization routine, which uses the GROMOS96 force field (27), with 400 steps of steepest descent. Electrostatic surface values were calculated (for 1VQQ structures only) using PyMOL (6) and its corresponding adaptive Poisson-Boltzmann solver (APBS) plug-in (15). All figures were prepared using UCSF Chimera.

RESULTS

Generation of BPR-resistant mutants.

To generate BPR-resistant MRSA, 300-ml broth cultures of three plasmid transformants of strain COLnex were serially passaged in subinhibitory and increasing BPR concentrations for a total of 28 days. COLnex(pYK21) was BPR resistant initially, with the MIC for the strain being 16 μg/ml, above the BPR breakpoint of 4 μg/ml. Within 9 days of passaging, strains COLnex(pAW8) and COLnex(pYK20) became resistant to BPR, with MICs for the strains reaching 8 μg/ml (Fig. 1, inset). For all strains, BPR MICs continued to increase in a stepwise fashion until approximately 3 weeks (Fig. 1). For the pAW8 (mecA-negative) transformant, MICs plateaued for several days at 64, 128, and then 256 μg of BPR/ml. For the pYK20 and pYK21 transformants, MICs plateaued at 32, 64, and 128 μg/ml. The pAW8 transformant achieved the highest level of resistance. Population analyses demonstrated homogeneous resistance in all strains and confirmed the MICs obtained using broth cultures (Fig. 2). MICs increased as follows: 256-fold (from 1 to 256 μg/ml) for COLnex(pAW8), 32-fold (from 4 to 128 μg/ml) for COLnex(pYK20), and 8-fold (from 16 to 128 μg/ml) for COLnex(pYK21). All the BPR-resistant, passaged mutants had significantly reduced doubling times compared to those of their parent strains. This difference was most pronounced for the mecA-negative strains COLnex(pAW8) and COLnex(pAW8)D28, the pAW8 transformant obtained after 28 days of passaging (Fig. 3).

FIG. 1.

BPR resistance developed in multiple steps during the serial passage of strains COLnex(pAW8) (filled squares), COLnex(pYK20) (filled triangles), and COLnex(pYK21) (open circles). The highest BPR concentration in which strains grew each day is shown on the y axis. The inset is a close-up view of results from the first 2 weeks of passaging. The dashed line represents the preliminary BPR breakpoint of 4 μg/ml.

FIG. 2.

Population analyses showing BPR susceptibilities of prepassage strains [filled squares, COLnex(pAW8); filled triangles, COLnex(pYK20); and filled circles, COLnex(pYK21)] and strains passaged in BPR for 28 days [open squares, COLnex(pAW8)D28; open triangles, COLnex(pYK20)D28; and open circles, COLnex(pYK21)D28]. The y axis indicates the number of cells (expressed as the log₁₀ number of CFU per milliliter) growing on BPR-containing agar.

FIG. 3.

Growth curves of strains before and after BPR passage. Filled squares, COLnex(pAW8); open squares, COLnex(pAW8)D28; filled triangles, COLnex(pYK20); open triangles, COLnex(pYK20)D28; filled circles, COLnex(pYK21); and open circles COLnex(pYK21)D28. OD₅₇₈, optical density at 578 nm.

BPR-resistant strains developed PBP2a mutations.

The sequences of mecA genes in post-BPR passage strains COLnex(pYK20)D28 and COLnex(pYK21)D28 were determined. In COLnex(pYK20)D28, six new mutations in mecA had arisen; five resulted in amino acid changes within the transpeptidase domain of PBP2a, and one change occurred in the non-penicillin-binding domain. mecA in COLnex(pYK21)D28 had four new mutations, three corresponding to the transpeptidase domain and one corresponding to the non-penicillin-binding domain (Fig. 4). Two changes, E447K and S649A, occurred in passaged COLnex(pYK20)D28 and COLnex(pYK21)D28. Intriguingly, the highest level of BPR resistance was observed in COLnex(pAW8)D28, a strain lacking mecA entirely. The absence of mecA was confirmed by PCR given this surprising result.

FIG. 4.

Schematic of PBP2a and amino acid substitutions in BPR-passaged strains containing plasmid-carried mecA, COLnex(pYK20) and COLnex(pYK21). The day of serial passage and the corresponding amino acid changes identified are shown. In the schematic, vertical black lines indicate three penicillin-binding motifs, the arrowhead denotes a transmembrane anchor, the speckled region denotes the non-penicillin-binding domain (nonPBD), and the diagonally striped region denotes the transpeptidase domain. Underlined amino acid substitutions arose independently in derivatives of both COLnex(pYK20) and COLnex(pYK21). D0, D13, D15, and D28, days 0, 13, 15, and 28.

In the mecA-containing strains, the level of BPR resistance increased with the number of mutations in PBP2a. Comparisons of PBP2a sequences and BPR MICs for strains passaged for roughly 2 and 4 weeks revealed that the acquisition of S649A was associated with the doubling of the BPR MIC from 64 to 128 μg/ml for COLnex(pYK21)D28. Similarly, the acquisition of S649A, F467Y, and R589K was associated with a fourfold increase in the BPR MIC (from 32 to 128 μg/ml) for COLnex(pYK20)D28 (Fig. 4).

To determine if mutations in other PBPs in BPR-resistant strains also occurred, PBP1 to PBP4 in all pre- and postpassage strains were sequenced. No changes in PBP1 in any strain were identified. There was a conservative amino acid substitution in PBP2 in COLnex(pYK20)D28 (A172V) and in PBP3 in COLnex(pYK21)D28 (D562E). There were two amino acid substitutions within PBP4 in COLnex(pAW8)D28 (E183A and F241R).

PBP2a mutations confer BPR resistance on mecA-containing strains.

To determine if the mutations in PBP2a were sufficient to confer BPR resistance, we exploited the fact that our strains contained plasmid-carried mecA. When COLnex was transformed with plasmids containing mutant mecA (generated by passaging in BPR), this BPR-sensitive strain was converted into a BPR-resistant strain (Fig. 5A). The BPR MICs for transformed strains were virtually the same as those for serially passaged strains, suggesting that the presence of mutant mecA alone was sufficient to confer full BPR resistance. There was no difference between resistance due to mecA from pYK20 in COLnex(pYK20)D28 and that from pYK21 in COLnex(pYK21)D28. As expected, transformation with the post-BPR passage plasmid pAW8D28 did not confer BPR resistance, as this plasmid does not contain the mecA insert (Fig. 5A).

FIG. 5.

(A) Population analyses of COLnex transformed with plasmids derived from BPR-passaged strains: COLnex(pAW8)T (filled squares), COLnex(pYK20)T (filled triangles), and COLnex(pYK21)T (open circles). The y axis indicates the number of cells (expressed as the log₁₀ number of CFU per milliliter) growing on BPR-containing agar. (B) Population analyses of BPR-resistant strains cured of plasmid by passage in the absence of tetracycline: COLnex(pAW8)c (filled squares), COLnex(pYK20)c (filled triangles), and COLnex(pYK21)c (open circles). The y axis indicates the number of cells (expressed as the log₁₀ number of CFU per milliliter) growing on BPR-containing agar.

Conversely, the loss of PBP2a converted BPR-resistant strains into BPR-sensitive strains. Postpassage BPR-resistant strains were cured of their plasmids by serial passage in the absence of tetracycline. Their susceptibilities to BPR were assessed by population analyses. Cured of the mecA-containing plasmid, COLnex(pYK20)c and COLnex(pYK21)c were highly susceptible to BPR (Fig. 5B). Only COLnex(pAW8)c remained resistant to BPR after the loss of the plasmid.

Molecular modeling of PBP2a mutant forms.

Molecular models of PBP2a mutant forms were generated based on the crystal structure of wild-type PBP2a from S. aureus (17). Amino acid substitutions found in COLnex(pYK20)D28 and COLnex(pYK21)D28 fall into three main groups. Group 1 mutations introduce charged or polar groups into the interior of the protein (not directly at the active site but nearby, within the transpeptidase domain) and make contact with a helix (α2) that plays a key role in the acylation of PBP2a by β-lactam substrates and inhibitors. An example of a group 1 mutation is V470E, which introduces a negatively charged or polar glutamate into the otherwise hydrophobic protein interior, adjacent to helix α2; likewise, the group 1 mutations F467Y and I563T both introduce a polar hydroxyl (-OH) group into the hydrophobic interior of the protein, near this helix (Fig. 6A). The perturbation of this helix may have implications for the rate of inhibition of PBP2a (see Discussion).

FIG. 6.

Structural perspectives of BPR-resistant PBP2a mutant forms. (A and B) Molecular modeling of mutant forms of PBP2a bound to the cephalosporin nitrocefin showing nitrocefin in blue (cephalosporin core structure), orange (variable group R1), and green (group R2); the native protein backbone and residues in gray; residues subject to mutation in cyan (native residue) and purple (corresponding mutant residue); and noncarbon atoms in CPK. (A) Group 1 mutations (F467Y, V470E, and I563T) likely affect BPR binding by perturbing helix α2 (shown as a yellow ribbon) by introducing adjacent polar or charged groups into the hydrophobic interior of PBP2a. (B) Group 2 substitutions probably lower BPR-binding affinity by directly affecting the PBP2a active site. These substitutions include Y446L, predicted to disrupt van der Waals contacts and pi bond stacking (π-π) interactions between PBP2a and conjugated double-bond systems of cephalosporin R2 groups; E447K, which may interact electrostatically with E460 to reposition Y446; S649A, predicted to destabilize the alpha helix bearing M641 that interacts with aromatic or hydrophobic R2 groups; and S643N, which may alter the polarity at the active-site entrance. (C) APBS modeling of PBP2a surface charge potential shows native residues E150 and E239 located within an extended swath of negative charge in the putative dimerization domain (the APBS surface is colored blue, white, and red, corresponding to values of +15, 0, and −15 kT/e, respectively, with shading by linear interpolation). Each of these residues is mutated, respectively, in pYK20 and pYK21, possibly contributing to BPR resistance by influencing protein-protein interactions.

Group 2 substitutions occur within or proximal to the β-lactam-binding site and so most likely affect BPR binding directly. Group 2 substitutions include S649A, which, by deleting the (-OH) group of serine 649, abolishes a hydrogen bond that ordinarily stabilizes the alpha helix terminating in M641, a residue shown previously to interact with the R2 group of the cephalosporin nitrocefin (17) (see Fig. S1 in the supplemental material; Fig. 6B); Y446L, which removes the aromatic ring of tyrosine 446 that interacts with the nitrocefin R2 group (17) (Fig. 6B); and E447K, which causes a charge reversal (negative to positive) adjacent to the active-site groove. Subsequent electrostatic attraction between mutant residue lysine 447 and native residue glutamate 460 (on the protein surface, ∼4 Å away) may further distort the active site, perhaps also altering the position of tyrosine 446 (Fig. 6B). Finally, S643N may also affect substrate binding via a polarity change at the active-site entrance. Of note, both S649A and E447K arose independently in COLnex(pYK20)D28 and COLnex(pYK21)D28.

Group 3 substitutions occur far from the active site in a domain not involved in penicillin binding and are thus likely to mediate resistance indirectly via interactions with other proteins. Group 3 substitutions include E150K and E239K, found in pYK20 and pYK21 mutants, respectively. These mutations introduce a positive charge within the same patch of negatively charged protein surface located in the solvent-accessible, N-terminal region of PBP2a (Fig. 6C). For the group 3 mutations, therefore, increased resistance may result from modified protein-protein interactions that rely on the electronegativity of this region of the enzyme. The remaining mutations did not fit neatly into any group, had no explicable effect, and may merely be incidental: I397T and R589K are located on the protein surface, do not involve formal charge differences, and furthermore did not arise independently of other mutations.

BPR resistance in COLnex(pAW8) (mecA-negative strain) appears to be mediated by chromosomal genes.

The mecA-negative strain COLnex(pAW8)D28 displayed the highest level of resistance to BPR among all strains used. The resistance mechanism was not plasmid mediated, as the pAW8 plasmid contained no insert and the removal or addition of the plasmid had no effect on the level of BPR resistance (Fig. 5). Resistance was not due to β-lactamase production, as the COLnex strain and its transformants are all β-lactamase negative (data not shown). BPR resistance in COLnex(pAW8)D28 was also stable over at least 7 days of passaging in the absence of BPR.

DISCUSSION

This is the first study to demonstrate that a strain of MRSA can develop high-level BPR resistance in vitro and that BPR resistance can be mediated by mutations in PBP2a. Unlike other groups, we were able to generate strains that were highly resistant to BPR, most likely because we used larger inocula during the serial passage of cultures. Maximal resistance to BPR in mecA-containing strains required multiple mutations in PBP2a that occurred in at least three steps and developed after prolonged passage in subinhibitory BPR concentrations.

The clinical implications of BPR resistance generated in vitro are unclear. Although the inocula used in our study were similar to concentrations seen in endocarditis or soft tissue abscesses (10⁸ to 10⁹ CFU per ml or gram of tissue) (13, 14), cultures were passaged for 9 days before resistant mutants were detected (Fig. 1). Several days of infection with such high organism burden is unlikely to occur in an individual patient. However, a prolonged period at high concentration may occur with a strain of MRSA that is transmitted over time among several patients.

Furthermore, the ability to develop and tolerate mutations in mecA or other genes mediating BPR resistance may depend on the genetic background, an issue that our experiments did not address. COL, the only parent strain used in our experiments, is an early MRSA strain (24) that is homogeneously resistant, not particularly virulent, and not representative of MRSA strains existing outside the laboratory. Moreover, the BPR-resistant strains exhibited significantly impaired growth compared to the parent strain, indicating that they had a fitness cost that may impair survival outside the laboratory (Fig. 3).

We have shown that in mecA-containing strains, BPR resistance results from changes in PBP2a alone. Because we used strains in which mecA had been excised from the chromosome and expressed from a plasmid, we were able to easily assess the effect of the addition or loss of mutant PBP2a on BPR susceptibility. The introduction of mutant PBP2a into a BPR-sensitive strain conferred full resistance to BPR. Conversely, the loss of mutant PBP2a converted a BPR-resistant strain into a fully sensitive one. The contribution of chromosomal loci in these strains appeared to be minor, indicating that in mecA-containing strains, PBP2a is the overwhelming target for BPR.

Our molecular modeling data suggest that PBP2a mutations can lead to BPR resistance through three different mechanisms: the inhibition of acylation, the inhibition of substrate binding, and interference with protein-protein interactions. All three mechanisms arose independently in both mecA-containing strains.

PBP2a resists β-lactam inactivation primarily via a highly inefficient acylation step that results from the high energetic cost needed to rearrange strand β3 and helix α2 (which bears the PBP2a serine 403 nucleophile) (17). The polar group 1 mutations (V470E, F467Y, and I563T) (Fig. 6A) likely further inhibit the acylation step by perturbing helix α2 and increasing the energetic penalty associated with the acylation of S403. Cephalosporins with activity against PBP2a, including BPR (5) and nitrocefin (19), have improved initial binding to the protein and overall increased acylation efficiency compared to those of other β-lactam antibiotics. The group 2 mutations identified here (S649A, Y446L, E447K, and S643N) appear to interfere with substrate binding by altering the functional groups that bind to BPR. Key substitutions appear to be S649A and E447K, as they both arose independently in COLnex(pYK20)D28 and COLnex(pYK21)D28.

PBPs, including PBP2a, are thought to be part of a large, multimembered cell wall-synthesizing holoenzyme complex (3). The N-terminal domain region of PBP2a, where group 3 substitutions occurred, is thought to be involved in PBP dimerization (InterPro domain IPR005311; http://www.ebi.ac.uk/interpro) (20). In addition, the farthest N-terminal domain may have a role in cell wall turnover, as it bears homology to YoeB, a protein that enhances cell survival in the presence of cell wall-targeting antibiotics by mediating autolysin activity via protein-protein interactions (25). As the group 3 substitutions E150K and E239K (Fig. 6C) introduced positive charges within an area of otherwise negatively charged surface, they may affect cooperative interactions with other S. aureus PBPs in the process of β-lactam resistance (16). Future studies that will be useful for the analysis of active-site architecture and the design of β-lactam antibiotics will include enzyme kinetics, molecular dynamics, and drug resistance studies of PBP2a mutant forms containing one or more of the amino acid substitutions described here.

The most surprising result of our study is that the highest level of BPR resistance developed in the mecA-negative strain COLnex(pAW8)D28. BPR resistance in this strain is unlikely to be explained by mutations in the high-affinity PBPs; no changes in PBP1, PBP2, or PBP3 were detected, and two amino acid changes in PBP4 were found. However, PBP4 is not essential and its inactivation has minimal effects on methicillin resistance (12).

Unidentified chromosomal loci have been implicated previously in oxacillin resistance in clinical MRSA isolates lacking mecA (26). A number of chromosomal genes may affect β-lactam resistance in mecA-negative strains. Transposon mutagenesis experiments have previously identified a number of auxiliary genes that influence methicillin resistance and have proven or proposed roles in cell wall synthesis (murE, femA, and femV), the regulation of the global stress response (the σ^B gene), or metabolism (protein kinase and ABC transporter genes) (7). In addition, proteins involved in the recruitment of PBPs to the septum (the site of new cell wall synthesis) or autolysins that control cell wall turnover may affect susceptibility to drugs active against the cell wall (23). Future experiments will focus on identifying the novel resistance determinant in COLnex(pAW8)D28.