The purpose of this study was to examine the relationship between pharmacokinetics and the evolution of resistance in bacterial populations. We chose to do this with S. aureus
and ciprofloxacin in an in vitro system. The system allowed us to expose the bacteria to several different pharmacokinetic profiles and to monitor changes in bacterial subpopulations, MICs, efflux, and mutations in the topoisomerase genes. Other investigators have examined the pharmacodynamics of fluoroquinolones (47
) and elucidated various mechanisms of resistance in S. aureus
), but to our knowledge, we are the first to examine the interplay of pharmacokinetics, resistance mechanisms, and the subpopulation dynamics underlying the evolution of fluoroquinolone resistance in these bacteria.
The starting cultures used in our in vitro system experiments were mainly comprised of ciprofloxacin-susceptible bacteria possessing silent mutations or wild-type sequences in the QRDRs of grlA/B
, as expected (8
). However, minor subpopulations with low-level ciprofloxacin resistance were recovered on agar containing 0.5 to 2.0 μg/ml from the starting cultures at frequencies previously reported for first-step S. aureus
). Genotyping of our variants revealed heterogeneity in their fluoroquinolone resistance determinants, consistent with the classic belief that mutations occur randomly during nonselective bacterial growth. Some variants possessed the point mutations (corresponding to S80F, S80Y, or A116P) in the QRDR of grlA
that are known to cause low-level fluoroquinolone resistance in genetic studies and associated with resistance in clinical isolates (15
). Other resistant subpopulations did not have mutations in the QRDR of either grlA/B
. Resistance among these variants may have been due to mutations outside the recognized topoisomerase gene QRDRs or to efflux (26
). The two- to fourfold lowering of the ciprofloxacin MICs in the presence of reserpine suggests that a reserpine-sensitive efflux pump, possibly NorA, was present in some bacteria in the starting populations. This finding is in concordance with a previous report describing reserpine-sensitive efflux systems in a majority of S. aureus
clinical strains (22
The pattern of bacterial killing and regrowth in the in vitro system varied with the pharmacokinetic profile simulated and the bacterial strain tested. As Cavg ss increased, the initial bacterial killing rate approached a maximum, and the rate of regrowth decreased. The emergence of ciprofloxacin resistance also varied in different pharmacokinetic environments. The smallest increases in MIC were noted when static ciprofloxacin concentrations were maintained just above the MIC. In these experiments, ciprofloxacin concentrations were maintained within the MSW the entire time. The low concentrations appeared to enrich variants with low-level resistance but apparently did not provide sufficient selective pressure for evolution toward higher levels of resistance. The greatest MIC increases were observed with the simulated clinical dosing regimens (400 mg every 8 and 12 h) providing Cavg ss values which fell between the MIC and the MPC for both strains. Within 24 h, low-level resistant variants present in the starting cultures became the predominant population; their numbers increased, and grlA/gyrA double mutants appeared in many cases by 96 h. A less pronounced increase in ciprofloxacin MIC was noted with the simulated pharmacokinetic profile of 750 mg every 12 h, which produced a value for Cavg ss near the MPC. In this case, selective enrichment of resistant subpopulations occurred, but no mutations in the gyrA QRDR were detected. Resistance to ciprofloxacin did not develop in pharmacokinetic environments in which concentrations exceeded the MPC over the entire dosing interval. Instead, viable counts declined to the limit of detection, and grlA mutants that were present in the starting populations appeared to be eliminated. These findings suggest that fluoroquinolone dosage regimens designed to eradicate topoisomerase mutants and other low-level resistant variants typically present in S. aureus cultures may prevent the emergence of resistance.
Our work supports the concept that resistant subpopulations of S. aureus
are selectively enriched when ciprofloxacin concentrations fall within the MSW, but unlike other investigators we did not find a clear relationship between the time that antimicrobial levels remained within the MSW and the degree of resistance that occurred (17
). For instance, high-level resistant variants appeared in a fluctuating environment (400 mg every 12 h) with a percentage of time in the MSW during the first 24-h period (T0-24 MSW
) of ≥96.9%, but did not appear in a continuous-infusion regimen with a T0-24 MSW
of 100%. These findings suggest that, in addition to TMSW
, other pharmacokinetic parameters, possibly including the Cavg ss
within the MSW, are also important in the emergence of resistance. Although others have proposed that selection is more intense in fluctuating than static antimicrobial environments (4
), we cannot draw conclusions about the influence of exposure pattern on the evolution of resistance with our data. This would require a different study design in which the effects of static and fluctuating pharmacokinetic profiles producing identical values for Cavg ss
are simulated and compared.
Our data showed that resistance did not occur with the Cmax
/MIC ratio of 47.2 and the AUC24
/MIC ratio of 584 h afforded by the dosing simulation of 2,800 mg every 12 h but did occur with the Cmax
/MIC ratios up to 13.4 and AUC24
/MIC ratios up to 159 h provided by other regimens. The latter findings do not support the suggestion made by others that Cmax
/MIC ratios of ≥8 to 12 and AUC24
/MIC ratios of ≥125 h are required for the eradication of bacteria and the prevention of resistance (6
). This observation, however, is based on only five simulated pharmacokinetic profiles. Additional experiments using a wider range of concentrations and incremental changes in Cmax
/MIC or AUC24
/MIC ratios would be needed to fully evaluate these pharmacodynamic concepts and determine optimal values for our bacteria in the in vitro system.
Bacteria with high-level resistance and grlA/gyrA
double mutations appeared at the end of some of our simulated regimens. The origin of these bacteria is unclear. It is unlikely that the appearance of these mutants at 96 h was simply due to selection, because we were unable to detect them in any of the 24 starting cultures. The probability that double mutants would arise from wild-type bacteria should be very low (10−15
) since the frequency of our grlA
mutations was 10−6
, and the frequency of mutations in gyrA
may be even lower (~10−9
). However, mutation frequencies may change in the presence of ciprofloxacin, since the fluoroquinolone is known to be mutagenic (13
High-level resistant variants with gyrA
mutations may have originated from grlA
mutants in the starting populations. In our experiments, grlA
mutants recovered at 96 h were always preceded by low-level resistant variants in the starting cultures that had the same grlA
mutation. This is consistent with previous reports showing that sequential mutations in grlA
of S. aureus
result in incremental increases in MIC (15
Although stepwise mutations in fluoroquinolone targets may explain the evolution of resistance during some dosage regimen simulations, some high-level resistant bacteria had no mutations in the QRDRs of grlA/B
. Others have also reported resistant S. aureus
variants lacking the expected mutations in the QRDRs of target genes (12
). In one of our experiments we found a mutation (corresponding to A176G) outside the traditional QRDR that has recently been reported to be associated with fluoroquinolone resistance (25
). However, other high-level resistant bacteria recovered during some of our experiments did not have mutations outside the conventional QRDR of grlA
, suggesting that mutations in an unrecognized determinant(s) may have contributed to the high-level resistance. In addition, the reserpine screening studies suggest that efflux may be responsible, in part, for the emergence of some resistant bacteria. The relative contributions of target mutations and efflux in the evolution of resistance cannot be elucidated without additional ciprofloxacin uptake and genetic studies.
Although we cannot explain the exact pathways taken by the bacteria in their evolution from susceptibility or low-level resistance to high-level resistance, bacteria recovered at the end of the 96-h experiments with the same MIC often had different mutations in the QRDRs of the topoisomerase genes and different levels of efflux. These differences occurred in several cases even though the bacteria had been exposed to the same simulated ciprofloxacin regimen. This suggests that the bacteria follow heterogeneous pathways of evolution and that the sequence of resistance mechanisms employed is not predetermined.
Of note, ciprofloxacin-resistant variants recovered from the in vitro system experiments maintained their resistant phenotype after seven passages on ciprofloxacin-free agar. Furthermore, the MICs for highly resistant bacteria that emerged during a 96-h experiment did not change even when the bacteria were maintained in the in vitro system for an additional 4 days without exposure to ciprofloxacin. These findings are consistent with the observation that susceptible bacteria readily evolve to resistance but rarely revert to a susceptible phenotype (33
). For this reason, it is essential to prevent resistance from occurring in the first place.
The in vitro system has several characteristics that may have influenced our results. Resistant variants may have been more likely to appear because of the poorly described effects of cellular crowding, relative anoxia (37
), and the absence of an immune system. Conversely, resistant bacteria may have been less likely to emerge because drug distribution in the peripheral compartment of the in vitro system was uniform and did not mimic the gradients often present within tissues (2
Despite potential limitations of the in vitro system, it is interesting to speculate on the clinical relevance of our in vitro findings. Of all the ciprofloxacin environments simulated in our system, the clinical pharmacokinetic profiles (400 mg every 8 or 12) were most likely to result in the emergence of S. aureus
variants with high-level resistance and mutations in grlA
. This may explain, in part, the high prevalence of grlA/gyrA
double mutant genotypes reported in collections of fluoroquinolone-resistant S. aureus
clinical isolates (50
). These findings suggest that novel approaches to fluoroquinolone dosing are needed.
Our results differ from those of others who examined the killing and regrowth of S. aureus
exposed to ciprofloxacin in in vitro systems (6
). The discrepancies between our results in which resistance (often high level) emerged and those of others in which resistance did not emerge may be due to factors such as the type of in vitro system used, phenotypic and genotypic heterogeneity of the bacteria tested, methods of drug administration, and the duration of the experiments.
Our work suggests that careful attention should be paid to population size, genotypes and phenotypes of resistant subpopulations, experiment duration, and profiles of drug exposure when examining the evolution of fluoroquinolone resistance in an in vitro system. Bacterial numbers must be sufficiently large (>1/resistance frequency) to ensure that resistant variants are present. Resistant subpopulation and mechanistic studies should be considered because of the genotypic and phenotypic heterogeneity in the resistance determinants observed in some bacterial populations. Experiments must be of sufficient duration to allow for the emergence of resistant variants since their appearance may be delayed if there is a fitness cost associated with resistance (1
). The concentration-time profiles simulated with the in vitro system should mimic as closely as possible the patterns observed in vivo since small concentration differences may exert strong selection pressure. In the case of antimicrobial agents administered by intermittent infusion, the drug should be infused and not administered as a bolus into the central compartment to allow concentrations to accumulate until a steady state is reached. We propose that a consensus method for conducting studies of antimicrobial resistance in in vitro systems would be beneficial. If standards are adopted, data from different studies might be pooled, enhancing their utility. Standardization of pharmacodynamic terminology has already been proposed (38
We evaluated the emergence of resistance when test strains were exposed to five different ciprofloxacin concentration-time profiles in the in vitro system. To fully characterize the relationship between pharmacokinetics and the emergence of resistance, additional concentration-time profiles should be simulated using dose escalation and fractionation approaches. Unfortunately, in vitro system studies of resistance are lengthy and resource and labor intensive. This limits the number of concentration-time profiles that can reasonably be simulated. The usefulness of the in vitro system might be enhanced if it were paired with a pharmacodynamic model integrating pharmacokinetic, subpopulation, and resistance mechanism data. The pharmacodynamic model could then be used to predict the fate of bacterial populations as a function of antimicrobial concentration. We demonstrate the potential utility of pharmacodynamic modeling in predicting the emergence of ciprofloxacin resistance in S. aureus
populations in a companion paper (9