Searching for the Optimal Predictor of Ciprofloxacin Resistance in Klebsiella pneumoniae by Using In Vitro Dynamic Models

Abstract

There is growing evidence of applicability of the hypothesis of the mutant selection window (MSW), i.e., the range between the MIC and the mutant prevention concentration (MPC), within which the enrichment of resistant mutants is most probable. However, it is not clear if MPC-based pharmacokinetic variables are preferable to the respective MIC-based variables as interstrain predictors of resistance. To examine the predictive power of the ratios of the area under the curve (AUC₂₄) to the MPC and to the MIC, the selection of ciprofloxacin-resistant mutants of three Klebsiella pneumoniae strains with different MPC/MIC ratios was studied. Each organism was exposed to twice-daily ciprofloxacin for 3 days at AUC₂₄/MIC ratios that provide peak antibiotic concentrations close to the MIC, between the MIC and the MPC, and above the MPC. Resistant K. pneumoniae mutants were intensively enriched at an AUC₂₄/MIC ratio of 60 to 360 h (AUC₂₄/MPC ratio from 2.5 to 15 h) but not at the lower or higher AUC₂₄/MIC and AUC₂₄/MPC ratios, in accordance with the MSW hypothesis. AUC₂₄/MPC and AUC₂₄/MIC relationships with areas under the time courses of ciprofloxacin-resistant K. pneumoniae (AUBC_M) were bell shaped. These relationships predict highly variable “antimutant” AUC₂₄/MPC ratios (20 to 290 h) compared to AUC₂₄/MIC ratios (1,310 to 2,610 h). These findings suggest that the potential of the AUC₂₄/MPC ratio as an interstrain predictor of K. pneumoniae resistance is lower than that of the AUC₂₄/MIC ratio.

INTRODUCTION

An increasing number of reports on the isolation of resistant pathogens (1, 2) combined with a weak antibiotic pipeline suggests that optimization of antibiotic therapy should be aimed at the suppression of resistance (3, 4). In this regard, dynamic models that mimic antimicrobial pharmacokinetics in vitro have been proven to be a useful tool in predicting the amplification of resistant mutants at clinically achievable antibiotic concentrations (5). Using these models, bell-shaped relationships have been established between the emergence of resistance to fluoroquinolones and the ratios of 24-hour area under the concentration-time curve (AUC₂₄) to the MIC (6–17). Such relationships have been reported with moxifloxacin, gatifloxacin, levofloxacin, ciprofloxacin, the investigational fluoroquinolone ABÒ492, pazufloxacin, tosufloxacin, and garenoxacin against Staphylococcus aureus (6–8, 14–16); moxifloxacin against Streptococcus pneumoniae (12, 13); garenoxacin against Klebsiella pneumoniae (15); ciprofloxacin and moxifloxacin against Pseudomonas aeruginosa (11, 12); and ciprofloxacin, marbofloxacin, and enrofloxacin against Escherichia coli (9, 10, 17). In some of these studies (6–15, 17, 18), changes in susceptibility of antibiotic-exposed bacteria and/or their enrichment with resistant mutants was observed at fluoroquinolone concentrations above the MIC but below the mutant prevention concentration (MPC), i.e., inside the mutant selection window (MSW) but not outside the MSW, in accordance with the MSW hypothesis (19, 20). However, the potential of AUC₂₄/MIC and AUC₂₄/MPC ratios as interstrain predictors of bacterial resistance has been examined only infrequently (8, 10, 16–18, 21).

To compare the abilities of AUC₂₄/MPC and AUC₂₄/MIC ratios to predict interstrain thresholds that restrict mutant enrichment, ciprofloxacin-resistant mutants of three K. pneumoniae strains with various MPC/MIC ratios were studied over a wide range of AUC₂₄/MIC ratios simulated in an in vitro dynamic model.

MATERIALS AND METHODS

Antimicrobial agent, bacterial strains, and susceptibility testing.

Ciprofloxacin powder was purchased from AppliChem BioChemica Chemical Synthesis Services (Darmstadt, Germany).

Three clinical isolates of K. pneumoniae, 1885, 1145, and 185, with different ciprofloxacin MICs (1, 0.5, and 0.125 μg/ml, respectively) and MPCs (8, 12, and 8 μg/ml, respectively) were selected for the study. The MICs were determined prior to, during, and after 3-day simulated treatments with ciprofloxacin. Susceptibility testing was performed at least in duplicate by broth microdilution techniques (22) at 24 h postexposure with organisms grown in Ca²⁺- and Mg²⁺-supplemented Mueller-Hinton broth (MHB) at an inoculum size of 10⁶ CFU/ml.

The MPCs were determined as described elsewhere (8). Briefly, the tested microorganisms were cultured in MHB and incubated for 24 h. Then, the suspension was centrifuged (4,000 × g for 10 min) and resuspended in MHB to yield a concentration of ∼10¹⁰ CFU/ml. A series of agar plates containing known ciprofloxacin concentrations was then inoculated with ∼10¹⁰ CFU of K. pneumoniae. The inoculated plates were incubated for 48 h at 37°C and screened visually for growth. To estimate the MPC, logarithms of bacterial numbers were plotted against antibiotic concentrations. MPC was taken as the point where the plot intersected the lower limit of detection (log CFU/milliliter = 1).

In vitro dynamic model and simulated pharmacokinetic profiles.

A previously described dynamic model (23) was used in the study. The concordance between measured and designed concentrations of ciprofloxacin has been reported elsewhere (9). Briefly, the model consisted of two connected flasks, one containing fresh MHB and the other with a magnetic stirrer, the central unit, with the same broth containing a bacterial culture plus antibiotic. Peristaltic pumps circulated fresh nutrient medium to the flasks and from the central 100-ml unit at a flow rate of 17.3 ml/h. The system was filled with sterile MHB and placed in an incubator at 37°C. The central unit was inoculated with an 18-hour culture of K. pneumoniae. After a 2-hour incubation, the resulting exponentially growing cultures reached approximately 10⁸ CFU/ml, and ciprofloxacin solutions were injected into the central unit at 12-h intervals. The duration of the experiments was at least 72 h.

A series of monoexponential profiles that mimic twice-daily dosing of ciprofloxacin with a half-life (t_1/2) of 4 h was simulated for 3 consecutive days. The simulated t_1/2 represented weighted means of values reported for humans, 3.2 to 5.0 h (24). The simulated pharmacokinetic profiles were designed to provide 24-hour ratios of area under the curve (AUC₂₄) to the MIC from 15 to 2,880 h (corresponding to an AUC₂₄ range from 15 to 2,880 μg · h/ml [0.7- to 131-μg · h/ml range of ciprofloxacin clinical doses against K. pneumoniae 1885]) and from 7.5 to 2,880 h with K. pneumoniae 1145 and 185 (corresponding to an AUC₂₄ range from 3.75 to 1,440 and from 0.9 to 360 μg · h/ml, respectively [0.2- to 66- and 0.04- to 16.4-μg · h/ml range of the clinical dose, respectively]).

Quantitation of antibiotic effect on susceptible and resistant subpopulations.

In each experiment, bacterium-containing medium from the central unit of the model was sampled every 24 h to determine bacterial concentrations throughout the observation period. One-hundred-microliter samples were serially diluted if necessary, in order to account for antibiotic carryover prior to plating, thereby reducing the antibiotic concentration below the MIC of the drug. Samples were spirally plated onto Mueller-Hinton agar (MHA) plates without antibiotic (“0× MIC”) and those containing 4×, 8×, and 16× MIC of ciprofloxacin using the Easy Spiral Pro system (Interscience, St. Nom, France). Colonies were counted by an automated colony counter (Interscience Scan 1200; Interscience, St. Nom, France).

The lower limit of counting susceptible and resistant cells was 200 CFU/ml (equivalent to 20 colonies per plate); the lower limit of quantification of resistant mutants was 10 CFU/ml (equivalent to at least one colony per plate). To reveal changes in susceptibility of ciprofloxacin-exposed bacterial cultures, MICs were reassessed every 24 h during the experiment and after the simulated treatment. To confirm the enhanced ciprofloxacin resistance in isolates with established MIC elevations, repetitive MIC testing was performed after several passages on antibiotic-free medium.

Based on time-kill data, the area between the cutoff level at 10⁸ CFU/ml and the time-kill curve (ABBC [25]) was calculated. The ABBC-log AUC₂₄/MIC curve was fitted by the sigmoid function:

$$Y=Y_0+a/\{1+\exp [-(x-x_0)/b]\}$$ (1)

where Y is ABBC; x is log(AUC₂₄/MIC); Y₀ and a are the minimal and maximal values of the antimicrobial effect, respectively; x₀ is x corresponding to a/2; and b is a parameter reflecting sigmoidicity.

To delineate AUC₂₄/MIC and AUC₂₄/MPC relationships with resistance, areas under the bacterial mutant concentration-time curves (AUBC_Ms [8]) were determined for subpopulations resistant to 4×, 8×, and 16× MIC of antibiotic from the beginning of treatment to 72 h, corrected for the areas under the lower limit of quantification over the same time interval. To reveal changes in susceptibility of ciprofloxacin-exposed bacterial cultures, MICs were reassessed every 24 h during the experiment and after antibiotic exposures (MIC_final). Each sample from the central compartment was plated on Mueller-Hinton agar (MHA). After 18 to 24 h of incubation, a suspension of isolated colonies was made in MHB to achieve an 0.5 McFarland turbidity standard. Susceptibility was evaluated by the microdilution technique (22).To account for the different susceptibilities of K. pneumoniae strains, the MIC_finals were related to the MICs determined with the starting inoculum (MIC_initials). A modified Gaussian-type function was used to fit the AUBC_M or MIC_final/MIC_initial versus AUC₂₄/MIC or AUC₂₄/MPC data sets:

$$Y=Y_0+a\hspace{0.17em}\exp \{[0.5{(x-x_0)}^c]/b\}$$ (2)

where Y is AUBC_M or MIC_final/MIC_initial, x is log(AUC₂₄/MIC) or log(AUC₂₄/MPC), Y₀ is the minimal value of Y, x₀ is log(AUC₂₄/MIC) or log(AUC₂₄/MPC) that corresponds to the maximal value of Y, and a, b, and с are parameters.

Determination of “antimutant” AUC₂₄/MIC and AUC₂₄/MPC ratios.

To determine the AUC₂₄/MIC and AUC₂₄/MPC ratios that protect against the enrichment of mutants (antimutant ratios), points belonging to the descending portion of the AUBC_M-AUC₂₄/MIC and AUBC_M-AUC₂₄/MPC curves were used. The AUBC_Ms were plotted against the respective AUC₂₄/MIC ratios and AUC₂₄/MPC ratios. The antimutant ratio was taken as the point where the regression line reaches the level of AUBC_M of 30 log(CFU/ml) · h.

$$Y=Y_0+ax$$ (3)

where Y is AUBC_M, x is AUC₂₄/MIC or AUC₂₄/MPC ratio, Y₀ is the minimal value of Y, and a is a parameter.

Mechanisms of resistance.

Nucleotide sequences of the quinolone-resistance-determining regions (QRDRs) of gyrA and parC genes were determined for both parental strains and mutants with enhanced ciprofloxacin resistance as previously described (26, 27) with some modifications. The PCR and sequencing primers were redesigned based on alignment of all K. pneumoniae gyrA and parC gene sequences available from GenBank. The original forward primer, gyrA6 (5′-CGACCTTGCGAGAGAAAT-3′), described by Weigel et al. (26), was used in combination with the modified primer, gyrAR2 (5′-CCTTCAATGCTGATGTCTTCA-3′); the parC forward and reverse primers (27) were extended by 2 bases at the 3′ end (parC-F2, 5′-TACGTGATCATGGACAGGGC-3′) and shortened by one base at the 5′-end (parC-R2, 5′-CCACTTCCCGCAGGTTG-3′), respectively. Bacterial genomic DNA was isolated using the InstaGene matrix kit (Bio-Rad Laboratories Inc., Hercules, CA). Amplified fragments were purified by exonuclease I and shrimp alkaline phosphatase treatment (Thermo Scientific Fermentas, Vilnius, Lithuania) and were sequenced on both strands using the same primers as those for PCR, the BigDye Terminator v3.1 cycle sequencing kit, and the ABI Prism 310 genetic analyzer (Applied Biosystems, Foster City, CA). Sequences obtained were compared with those previously reported for gyrA and parC genes of K. pneumoniae ATCC 43816 (GenBank nucleotide sequence accession no. APWN00000000.1).

RESULTS

Time courses of simulated ciprofloxacin concentrations, related to MIC, and ciprofloxacin-exposed K. pneumoniae grown without antibiotic (0× MIC) and on medium containing, for example, 4× MIC of ciprofloxacin at three characteristic AUC₂₄/MIC ratios are shown in Fig. 1. As seen in the left column, small and transient, if any, reductions in the number of susceptible K. pneumoniae cells were observed at the lowest AUC₂₄/MIC ratio (15 h). Mutants resistant to 4× MIC of ciprofloxacin were enriched starting from the second day of treatment with each of the studied organisms. At the AUC₂₄/MIC ratio of 60 h (middle column in Fig. 1), both killing of susceptible cells and enrichment of resistant K. pneumoniae were more pronounced than at the lower AUC₂₄/MIC ratio. At least with K. pneumoniae 1145, ciprofloxacin-susceptible cells were completely replaced by resistant mutants by the end of treatment. With an increase in the AUC₂₄/MIC ratio to 720 h (right column in Fig. 1), the numbers of susceptible cells decreased significantly except for K. pneumoniae 1145 (maximal reduction of 3 log CFU/ml). Although resistant mutants were enriched beginning from the 3rd to 4th day of treatment, their final numbers were 1.1 to 2.4 log CFU/ml smaller than at the intermediate simulated AUC₂₄/MIC ratio.

FIG 1.

Simulated pharmacokinetic profiles of ciprofloxacin and time courses of susceptible (0× MIC) and resistant (4× MIC) subpopulations of K. pneumoniae exposed to ciprofloxacin. The simulated AUC₂₄/MIC ratios (in hours) are indicated as boxed numbers.

Mutants resistant to 8× and 16× MIC were isolated with each ciprofloxacin-exposed organism at each AUC₂₄/MIC ratio except for the lack of growth of K. pneumoniae 185 mutants resistant to 16× MIC at an AUC₂₄/MIC of 15 h. Time courses of the more resistant mutants were similar to those observed with the mutant resistant to 4× MIC of ciprofloxacin (data not shown). As a whole, the higher the resistance level the less pronounced was the respective mutant's growth. The enrichment of resistant mutants at AUC₂₄/MIC ratios of 15 and 60 h was accompanied by concomitant MIC elevations. The enhanced ciprofloxacin resistance in these isolates was confirmed by repetitive MIC testing after several passages on antibiotic-free medium. With K. pneumoniae 1885 and 1145, the maximal ratio of MIC_final to MIC_initial (14 and 8, respectively) was observed at an AUC₂₄/MIC ratio of 60 h, and that with K. pneumoniae 185 was observed at an AUC₂₄/MIC ratio of 720 h (MIC_final/MIC_initial = 14). More detailed MIC_final/MIC_initial data obtained over a wider range of simulated AUC₂₄/MIC ratios are discussed below.

To determine the mechanisms of decreased ciprofloxacin susceptibility in mutants, partial nucleotide sequences of gyrA (nucleotides 21 to 605 with respect to the translation start site) and parC (nucleotides 82 to 585 with respect to the translation start site) genes spanning the entire quinolone-resistance-determining regions (QRDRs) were analyzed. The sequences of the parental strains and mutant strains were identical and contained no missense mutations compared to the wild-type K. pneumoniae ATCC 43816 (GenBank nucleotide sequence accession no. APWN00000000.1).

To delineate concentration-resistance relationships, the AUBC_Ms were plotted against AUC₂₄/MIC and AUC₂₄/MPC ratios. Figure 2 shows AUBC_M-AUC₂₄/MIC and AUBC_M-AUC₂₄/MPC relationships observed with K. pneumoniae mutants resistant to 4×, 8×, and 16× MIC of ciprofloxacin. With each organism, AUBC_M versus AUC₂₄/MIC and AUC₂₄/MPC curves were bell shaped. The AUBC_M increased with an increase in the AUC₂₄/MIC or AUC₂₄/MPC ratio, reaching a maximum followed by a subsequent decrease in numbers of resistant mutants. The maximal amplification of mutants resistant to 4×, 8×, and 16× MIC of ciprofloxacin was bacterial strain specific, much more pronounced for K. pneumoniae 185 and 1145 than for K. pneumoniae 1885. At the lower levels of resistance (4× and 8× MIC), maximal AUBC_Ms for resistant mutants of the studied organisms differ by 1.3 to 1.8 orders of magnitude, whereas with the highest level of resistance (16× MIC) the strain-specific difference in the maximal AUBC_Ms was 2.5-fold. Regardless of bacterial strain, AUBC_Ms that reflect amplification of mutants resistant to 4× MIC of ciprofloxacin were more pronounced than those with mutants resistant to 8× and especially 16× MIC.

FIG 2.

AUC₂₄/MIC and AUC₂₄/MPC relationships with AUBC_M (for mutants resistant to 4×, 8×, and 16× MIC of ciprofloxacin) observed with individual K. pneumoniae isolates.

Unlike maximal AUBC_Ms, AUC₂₄/MIC and AUC₂₄/MPC ratios that correspond to maximums on the bell-shaped curves did not depend on the level of resistance, but they were strain specific. Maximal amplification of resistant K. pneumoniae 1145 occurred at a lower AUC₂₄/MIC ratio (60 h) than that for K. pneumoniae 1885 (60 to 120 h) and especially K. pneumoniae 185 (180 to 360 h). The respective AUC₂₄/MPC ratios were 2.5, 7.6 to 15, and 2.8 to 5.6 h. Also, bacterial strain-specific differences were inherent in the antimutant AUC₂₄/MIC and AUC₂₄/MPC ratios determined for individual organisms (Table 1). As seen in Table 1, with mutants of the same level of resistance, estimated thresholds of AUC₂₄/MIC ratio were less variable than those of AUC₂₄/MPC ratio among the studied strains. For example, with mutants resistant to 4× MIC of ciprofloxacin, individual antimutant AUC₂₄/MIC ratios varied over a 2-fold range (from 1,312 to 2,612 h), whereas the respective AUC₂₄/MPC ratios varied over an almost 10-fold range (from 20 to 290 h). With K. pneumoniae 1885 and 1145 but not with strain 185, strain-associated differences in the antimutant AUC₂₄/MIC or AUC₂₄/MIC ratio were less pronounced with mutants resistant to 8× and 16× MIC (1.2- to 1.7- versus 4.5- to 7.5-fold ranges, respectively). Overall, both AUC₂₄/MIC and AUC₂₄/MPC ratio thresholds were lower with more resistant than with less resistant mutants.

TABLE 1.

Predicted antimutant (AUC₂₄/MIC)/(AUC₂₄/MPC) ratios

K. pneumoniae strain	Predicted antimutant (AUC24/MIC)/(AUC24/MPC) ratio (h) for mutant resistant to ciprofloxacin at:
	4× MIC	8× MIC	16× MIC
1885	2,320/290	1,200/150	720/90
1145	2,610/110	1,440/60	1,000/40
185	1,310/20	1,270/20	1,250/20

Despite individual differences in the AUBC_M versus AUC₂₄/MIC or AUC₂₄/MPC curves, a Gaussian-type function fits the AUBC_M-AUC₂₄/MIC or AUBC_M-AUC₂₄/MPC combined data on the three K. pneumoniae mutants resistant to 4× MIC of ciprofloxacin (Fig. 3, upper plot) with relatively high r²s (0.76 and 0.72, respectively). Less clear correlations were found for mutants resistant to 8× MIC (0.63 and 0.60, respectively) and 16× MIC (0.61 and 0.55, respectively) (data not shown). As seen in Fig. 3, the greatest enrichment of mutants resistant to 4× MIC of ciprofloxacin was seen at 90 h (AUC₂₄/MIC ratio) and 5 h (AUC₂₄/MPC ratio), i.e., within the ranges established for individual organisms (60 to 360 h and 2.5 to 15 h, respectively).

FIG 3.

Combined data on all three K. pneumoniae isolates (for mutants resistant to 4× MIC of ciprofloxacin). (Top) AUC₂₄/MIC and AUC₂₄/MPC relationships with AUBC_M. Estimated coefficients of equation 2: Y₀ = 1, x₀ = 1.926, a = 350.8, b = 0.6633, c = 2.007 (AUBC_M versus AUC₂₄/MIC ratio); Y₀ = 1, x₀ = 0.4464, a = 349, b = 0.6887, c = 2.01 (AUBC_M versus AUC₂₄/MPC ratio). (Bottom) Antimutant AUC₂₄/MIC and AUC₂₄/MPC ratio determination using the respective descending portions of the AUBC_M versus AUC₂₄/MIC and AUC₂₄/MPC curves. Estimated coefficients of equation 3: Y₀ = 953.0, a = −283.8 (AUBC_M versus AUC₂₄/MIC ratio); Y₀ = 442.5, a = −194.8 (AUBC_M versus AUC₂₄/MPC ratio).

To determine the antimutant AUC₂₄/MIC and AUC₂₄/MPC ratios more precisely, only points on the descending branch of the bell were considered (Fig. 3, lower plot). Although AUC₂₄/MPC and AUC₂₄/MIC relationships with the AUBC_M were similar, a linear function (equation 3) fits AUC₂₄/MIC data combined for the three K. pneumoniae strains better than the AUC₂₄/MPC ratio: r² of 0.74 versus 0.59. The respective antimutant AUC₂₄/MIC ratio of 1,800 h and AUC₂₄/MPC ratio of 130 h were within the ranges established for individual organisms (1,310 to 2,610 h and 20 to 290 h, respectively), too.

The same equation 2 also fits MIC_final/MIC_initial versus AUC₂₄/MIC ratio and MIC_final/MIC_initial versus AUC₂₄/MPC data (Fig. 4). These relationships were bell shaped, similarly to the respective AUC₂₄/MIC or AUC₂₄/MPC relationships with the AUBC_M described above. As seen in the left plot in Fig. 4, the maximal MIC_final elevations were observed at the intermediate AUC₂₄/MIC ratios (180 to 360 h), whereas at the lower and higher AUC/MIC ratios the MIC_finals did not increase. MIC_final/MIC_initial ratio versus AUC₂₄/MIC ratio showed good correlation (r² of 0.75), comparable to AUBC_M-AUC₂₄/MIC or AUBC_M-AUC₂₄/MPC relationships. However, the MIC_final/MIC_initial-AUC₂₄/MPC relationship, shown in the right plot in Fig. 4, was described with significantly lower r² (0.43) because of pronounced scattering of the points. As seen in Fig. 4, the greatest losses in susceptibility occurred at the AUC₂₄/MIC ratio of 300 h and AUC₂₄/MPC ratio of 8.5 h. The respective antimutant AUC₂₄/MIC and AUC₂₄/MPC ratios, at which the MIC_final/MIC_initial ratio was 1, were 2,560 and 300 h.

FIG 4.

AUC₂₄/MIC and AUC₂₄/MPC relationships with MIC_final/MIC_initial combined data on all three K. pneumoniae isolates. Estimated coefficients of equation 2: Y₀ = 1, x₀ = 2.309, a = 20.74, b = 0.5127, c = 2.175 (MIC_final/MIC_initial ratio versus AUC₂₄/MIC ratio); Y₀ = 1, x₀ = 0.8559, a = 15.63, b = 0.7454, c = 2.207 (MIC_final/MIC_initial ratio versus AUC₂₄/MPC ratio).

Pronounced individual variability was inherent also in killing of susceptible subpopulations of K. pneumoniae exposed to ciprofloxacin (Fig. 5). Using combined data on the three organisms, a sigmoid function (equation 1) fits the ABBC versus AUC₂₄/MIC ratio with an r² of 0.78.

FIG 5.

AUC₂₄/MIC ratio-dependent effects of ciprofloxacin on susceptible subpopulation of K. pneumoniae fitted by equation 1: Y₀ = 0, x₀ = 2,389, a = 208.2, b = 0.4083.

DISCUSSION

Aimed at examination of the AUC₂₄/MPC and AUC₂₄/MIC ratios as interstrain predictors of bacterial resistance, this study demonstrates applicability of the MSW hypothesis to fluoroquinolone-resistant K. pneumoniae. With each of three clinical isolates, the maximal amplification of mutants resistant to 4×, 8×, and 16× MIC of ciprofloxacin was observed when antibiotic concentrations fell into the MSW for most of the dosing interval, and both AUC₂₄/MIC and AUC₂₄/MPC relationships with resistance expressed by the population analysis data (AUBC_M) or susceptibility testing (ratio of the postexposure MIC to preexposure MIC) were bell shaped. These observations are consistent with earlier findings reported in studies with fluoroquinolones (6–17).

Findings obtained in the present study predict the antimutant AUC₂₄/MIC ratios for K. pneumoniae strains 1885, 1145, and 185 (2,320, 2,610, and 1,310 h, respectively) that are much higher than the respective clinically attainable AUC₂₄/MIC ratios (22, 44, and 176 h, respectively). The same trend also was seen in our recent study with ciprofloxacin against four P. aeruginosa strains (11): antimutant AUC₂₄/MIC ratios (220 to 1,100 h) were higher than the respective clinically attainable AUC₂₄/MIC ratios (44 and 176 h). In contrast, in our studies with E. coli (9), despite relatively high antimutant AUC₂₄/MIC ratios (720 to 1,440 h), these ratios can be achieved at clinically attainable exposures (AUC/MIC ratios of 2,750 and 1,375 h).

Despite pronounced stratification of AUBC_M versus AUC₂₄/MIC or AUC₂₄/MPC curves seen with individual organisms, it appeared possible to establish bacterial strain-independent AUC₂₄/MIC or AUC₂₄/MPC relationships with resistance using combined data with all three K. pneumoniae strains. Similar relationships were established between the MIC_final/MIC_initial ratio and AUC₂₄/MIC and AUC₂₄/MPC ratios. However, AUC₂₄/MIC and AUC₂₄/MPC ratios that correspond to the greatest loss in susceptibility were shifted toward higher values compared to those that correspond to the greatest AUBC_M. A similar shift was reported with two of four strains in our study with ciprofloxacin-exposed P. aeruginosa (11).

As in our recent study with ciprofloxacin against E. coli (10), to test significant levels of K. pneumoniae resistance, we chose mutants resistant to 4×, 8×, and 16× MIC, ignoring more abundant nontarget (e.g., efflux) mutants resistant to 2× MIC of ciprofloxacin because exposure of bacteria to twice their MICs can almost be regarded as a positive control (28). In the present study, we did not identify any changes in nucleotide sequences of gyrA and parC QRDRs of ciprofloxacin-resistant K. pneumoniae mutants compared to those of parental or reference strains, suggesting that the enhancement of resistance was likely due to altered drug accumulation, although mutations causing resistance to fluoroquinolones in Enterobacteriaceae most often occur in the QRDR of the primary target gene, gyrA (9, 29–31). Additional mutations that further increase resistance to fluoroquinolones may occur in the secondary target gene, parC (29–31), and, in K. pneumoniae, in various regulatory genes that affect the expression of efflux pumps, e.g., AcrAB (32–34), and major porins, OmpK35 and OmpK36 (35, 36), leading to decreased accumulation of the drug in bacterial cells.

In conclusion, findings obtained in the present study are in support of the MSW hypothesis that explains the bell shape of the AUBC_M versus AUC₂₄/MIC and AUC₂₄/MPC curves. However, highly variable AUC₂₄/MPC thresholds call into question the reliability of their interstrain predictions of the enrichment of K. pneumoniae mutants. From this point of view, the AUC₂₄/MIC ratio should be preferred to the AUC₂₄/MPC ratio for predictions of the emergence of K. pneumoniae resistance. Similar conclusions were drawn in studies with ciprofloxacin-exposed E. coli (9, 10) and P. aeruginosa (11). In fact, a priori expectations of better predictive force of the AUC₂₄/MPC ratio were confirmed only with fluoroquinolone-exposed S. aureus (5, 8).

Further studies with other bacterial species are needed to provide a more complete understanding of AUC₂₄/MIC and AUC₂₄/MPC ratios as potential predictors of the enrichment of resistant mutants.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Grant 14-15-00970 supported the study performed at the Department of Pharmacokinetics & Pharmacodynamics of Gause Institute of New Antibiotics.