Mutant Prevention Concentration-Based Pharmacokinetic/Pharmacodynamic Indices as Dosing Targets for Suppressing the Enrichment of Levofloxacin-Resistant Subpopulations of Staphylococcus aureus

Abstract

MIC- and mutant prevention concentration (MPC)-based pharmacokinetic/pharmacodynamic (PK/PD) indices were compared for suitability as attainment targets for restricting amplification of levofloxacin-resistant mutant subpopulations. When three Staphylococcus aureus strains were examined with a hollow-fiber PK/PD model, area under the concentration-time curve over 24 h (AUC₂₄)/MPC values of >25 and maximum concentration of drug in serum (C_max)/MPC values of >2.2 predicted resistance outcome among different isolates with an interisolate kappa coefficient of 1. MIC-based mutant-restrictive PK/PD values varied >8-fold and exhibited only a moderate interisolate agreement (kappa coefficient of 0.5). Thus, MPC-based PK/PD indices are more suitable than MIC-based indices for predicting mutant-restricting fluoroquinolone doses when multiple bacterial isolates are considered.

INTRODUCTION

Optimization of dosing strategies has been advocated as a way to limit antimicrobial resistance. Early efforts focused on killing the major, susceptible pathogen population (1, 19). However, such a strategy is inadequate when infections contain bacterial subpopulations resistant to therapeutic drug concentrations, because mutants are recovered even when susceptible populations are eradicated (12). Descriptions of resistant subpopulations (2, 3, 7) led to the mutant selection window hypothesis (23, 24), which maintains that mutant subpopulations selectively amplify within a specific drug concentration range. The lower boundary of the range, approximated by the MIC, serves as a threshold for restricting the growth of the major, susceptible portion of the pathogen population (6); the mutant prevention concentration, or MPC (7), defines the upper boundary and serves to restrict growth of resistant mutant subpopulations (6, 25). Since MIC-based pharmacokinetic/pharmacodynamic (PK/PD) indices have successfully been used to predict microbiological outcome for the bulk of the bacterial population (5, 15, 17, 18), attempts have also been made to use MIC-based indices to establish PK/PD targets for the design of dosing regimens that limit selection of resistance (9). Such targets must be constant, or nearly constant, among different isolates of a given species with a given antimicrobial to be used readily to predict doses that restrict mutant outgrowth. Unfortunately, values of area under the concentration-time curve over 24 h (AUC₂₄)/MIC that restrict growth of mutant subpopulations vary substantially from one isolate to another with Staphylococcus aureus when treated with garenoxacin (20). We have proposed that MPC-based targets are more appropriate (25), because MPC is a direct measure of mutant subpopulation susceptibility while MIC is not and because MIC correlates poorly with MPC (8, 11, 13, 16). To date, no experiment has been performed to determine whether MPC-based, mutant-restrictive PK/PD indices for a given drug are similar among multiple isolates.

Three S. aureus strains having identical MICs but very different MPCs were chosen for analysis. If mutant-restricting values of AUC₂₄/MPC or maximum concentration of drug in serum (C_max)/MPC are superior dosing targets, they should be very similar for all three isolates despite significant differences in MPC. Moreover, mutant-restrictive AUC₂₄/MIC or C_max/MIC values should be very different among the isolates even though the MIC is the same for all three isolates. RN450 is a common laboratory strain (10), RN450-A1 is an isogenic garenoxacin-resistant gyrA mutant of RN450 (26), and SA99 is a sputum isolate obtained from PLA General Hospital. Bacteria were grown in Mueller-Hinton broth (MHB) or agar (MHA) at 37°C; MIC was determined using broth microdilution according to CLSI standards (4), and MPC was measured as described previously (26). Levofloxacin was from the National Institute for the Control of Pharmaceutical and Biological Products, Beijing, China.

Regimens of levofloxacin that restricted the amplification of resistant mutant subpopulations were identified by escalating doses in a CellMax hollow-fiber bioreactor system (Spectrum Laboratories Inc., Rancho Dominguez, CA) as described previously (21). S. aureus (1.5 × 10⁶ CFU [∼10⁵ CFU/ml]) was placed in the extracapillary space of the hollow-fiber cartridge, and cultures were allowed to grow overnight to a cell density of 10⁹ to 10¹⁰ CFU/ml. Fresh MHB was passed through the system continuously from a reservoir at a rate of 0.5 ml/min, which allowed simulation of a human plasma half-life for levofloxacin (∼8 h); bolus levofloxacin doses were administered once daily into the central compartment. Three incremental doses of levofloxacin were used for each strain to achieve AUC₂₄/MPC values that spanned from 22 to 25, a range that was reported previously as being restrictive for resistant-subpopulation amplification (6, 14). Levofloxacin concentration from samples collected at various times from the central reservoir was measured by high-performance liquid chromatography (HPLC) (6). The interday and intraday coefficients of variation for the HPLC measurement were 1 to 5.3% and 1.8 to 11%, respectively. The coefficient of correlation between simulated levofloxacin concentrations in the extracapillary compartment and experimentally determined concentrations was >98%. Pharmacokinetic parameters were calculated according to a two-compartmental model using DAS software (version 2.0; DAS, Wuhu, China). Values of AUC₂₄ that blocked mutant outgrowth and MIC increase were confirmed for each isolate with replicate experiments.

To monitor levofloxacin effects, bacterial samples (300 μl) were removed from the extracapillary space at various times after treatment. Half of each sample was inoculated into drug-free MHB for overnight growth to expand bacterial cultures and to reverse induced phenotypic resistance (if any). Then, the MIC was measured for each postexposure sample to determine the ratio of the MIC at each time point (MIC_time) relative to the MIC of the starting culture (MIC₀) as a surrogate for emergence of resistance. The other half of each sample was washed and diluted serially with sterile saline (0.9% NaCl) before being applied to MHA either lacking levofloxacin or containing levofloxacin at 2× or 4× MIC of the starting culture to determine the number of total or resistant cells, respectively (2× and 4× MIC were chosen because they are near the middle of the mutant selection window). Several resistant colonies recovered from 72-h samples were subjected to nucleotide sequence determination of the gyrA and parC (grlA) quinolone resistance-determining regions (QRDR), as described previously (26).

Initial MIC for the three strains of S. aureus was 0.125 μg/ml; MPCs were 1, 3, and 8 μg/ml for SA99, RN450, and RN450-A1, respectively. With each strain, three levofloxacin doses were applied (Fig. 1A). An initial reduction of total bacterial count was observed in each case, but persistent killing (i.e., without a nadir or regrowth) was evident only with the highest dose (Fig. 1B, right). Mutant subpopulation growth/regrowth during treatment was observed at both low and intermediate drug exposures (Fig. 1C, left and center). At the highest dose, mutant subpopulation growth was absent, and mutant killing was observed (Fig. 1C, right). An increase in total population MIC was observed with each strain at low and intermediate drug doses that maintained drug concentrations mostly between the MIC and the MPC (Fig. 1D, left and center). The highest drug dose, which kept drug concentrations above the MPC for more than 50% of the time, prevented mutant subpopulations from growing (Fig. 1D, right).

Fig. 1.

Levofloxacin treatment of S. aureus strain SA99 in a hollow-fiber dynamic in vitro model. (A) Simulated pharmacokinetics for levofloxacin to generate successively increasing values of AUC₂₄, as indicated at the top of each column (panels A to D have the same AUC₂₄ for a given column). Levofloxacin concentration was measured by HPLC, with samples taken at 0, 0.5, 1, 2, 5, 7, 8, 24, 48, and 72 h after initiation of treatment from the central compartment. The levofloxacin concentration in the extracapillary space was then simulated and displayed. MPC and MIC determined with the starting culture are indicated. (B) Survival measured with total bacterial population, determined by plating and incubation on drug-free agar. (C) Change in size of resistant mutant subpopulation, determined by plating and incubation at a levofloxacin concentration twice the MIC of the starting culture. The dashed line and the dotted symbol indicate that the viable bacterial number was below the detection limit (e.g., 10 CFU/ml) 72 h after initiation of levofloxacin treatment. (D) Ratio of MIC of samples treated with levofloxacin for the indicated times (MIC_time) relative to MIC of starting sample (MIC₀). Similar data were obtained with strains RN450 and RN450-A1.

When PK parameters were normalized to MIC or MPC to correlate PK/PD indices with selection of resistance among different isolates, only the highest drug exposures prevented mutant subpopulation enrichment (MIC₇₂/MIC₀ = 1) (Table 1). Lower exposures allowed enrichment of single-step parC mutants containing QRDR alleles commonly observed in clinical resistance (Table 1). When PK/PD indices (Table 1) were plotted as point scatter for each experiment (Fig. 2), with MPC-based indices, a line separating all cases that blocked mutant subpopulation enrichment from those that did not could readily be drawn (Fig. 2, squares). This line served as a sharp threshold for predicting resistance outcome among different isolates. Analysis (22) of these data showed a kappa coefficient of 1, indicating a perfect interisolate agreement. In contrast, with either AUC₂₄/MIC or C_max/MIC, a threshold separating doses that blocked mutant subpopulation enrichment from those that did not could not be drawn (Fig. 2, circles). The MIC-based, resistant-mutant-restrictive indices showed a high interisolate fluctuation (kappa coefficient of 0.5). The high interisolate precision among MPC-based, mutant-restrictive PK/PD values makes these values better than MIC-based indices as dosing targets for limiting the emergence of resistance.

Table 1.

Selective enrichment of levofloxacin-resistant mutant subpopulations at various PK/PD values

S. aureus strain	Expt no.	Cmax/MICa	Cmax/MPCa	AUC24/MICa	AUC24/MPCa	MIC72/MIC0b	ParC amino acid substitutionc	GyrA amino acid substitutionc
SA99	1	11.2	1.40	135	16.9	4	Glu84→Lys	None
	2	12.7	1.58	154	19.2	4	Ser80→Phe	None
	3	18.5	2.31	199	24.8	1	NAe	NA
	4	24.0	3.00	244	30.6	1	NA	NA
RN450	5	14.9	0.623	187	7.8	4	Ser80→Phe	None
	6	32.0	1.33	334	13.9	8	Ser80→Phe	None
	7	52.9	2.20	612	25.5	1	NA	NA
	8	55.1	2.30	626	26.1	1	NA	NA
RN450-A1	9	40.4	0.631	420	6.60	32	Glu84→Lys	Ser84→Leud
	10	74.5	1.16	871	13.6	32	Ser80→Phe	Ser84→Leud
	11	145	2.27	1,590	24.8	1	NA	NA
	12	159	2.48	1,740	27.1	1	NA	NA

^aPK/PD indices were calculated with the C_max and AUC from the first 24-h period.

^bMIC₇₂/MIC₀ represents the ratio of MIC for colonies recovered at 72 h during simulated levofloxacin treatment to initial (time zero) MIC and serves as a surrogate for emergence of resistance.

^cDNA sequence analysis was performed with colonies recovered from samples exposed to a 72-h levofloxacin treatment and then selected on 4× MIC agar.

^dStrain RN450-A1 had this GyrA resistance substitution prior to the present work. Note that this substitution did not affect the levofloxacin MIC because the primary target for levofloxacin is topoisomerase IV, while that for garenoxacin is gyrase.

^eNA, not available (no resistant mutant was recovered).

Fig. 2.

Correlation of MPC- and MIC-based PK/PD indices with resistance outcome among different bacterial isolates. PK/PD indices from all 12 bacterium-dose combinations listed in Table 1 were treated as independent entries and plotted as a function of experiment number to evaluate point scatter. AUC₂₄- and C_max-based indices are presented. MIC-based indices are shown as circles, and MPC-based indices are shown as squares. Open symbols indicate no amplification of resistant mutant subpopulation, while filled symbols represent selective amplification of resistant mutants. Dotted lines indicate a threshold value for either AUC₂₄/MPC or C_max/MPC separating PK/PD indices that resulted in enrichment of resistant mutant subpopulation from those that did not. The data showed perfect interisolate agreement (kappa coefficient of 1 [22]). No single line could be drawn for either AUC₂₄/MIC or C_max/MIC when multiple isolates were considered. Only a moderate interisolate agreement (kappa coefficient of 0.5) was observed when MIC-based PK/PD indices were used to predict resistance outcome.

Two additional comments are relevant. First, AUC₂₄/MPC may be preferable to C_max/MPC as a PK/PD target, because the former takes into account dosing frequency while the latter does not. Second, due to the use of only 3 bacterial strains and one drug in an artificial (hollow-fiber) system, the present work has limitations and thus should be treated only as a proof-of-principle study. Evaluation of a more extensive data set with multiple drug-bacterium combinations is now required to test whether MPC-based PK/PD indices are of general use for combating antimicrobial resistance.