Comparative Efficacies of Candidate Antibiotics against Yersinia pestis in an In Vitro Pharmacodynamic Model

ABSTRACT Yersinia pestis, the bacterium that causes plague, is a potential agent of bioterrorism. Streptomycin is the “gold standard” for the treatment of plague infections in humans, but the drug is not available in many countries, and resistance to this antibiotic occurs naturally and has been generated in the laboratory. Other antibiotics have been shown to be active against Y. pestis in vitro and in vivo. However, the relative efficacies of clinically prescribed regimens of these antibiotics with streptomycin and with each other for the killing of Yersinia pestis are unknown. The efficacies of simulated pharmacokinetic profiles for human 10-day clinical regimens of ampicillin, meropenem, moxifloxacin, ciprofloxacin, and gentamicin were compared with the gold standard, streptomycin, for killing of Yersinia pestis in an in vitro pharmacodynamic model. Resistance amplification with therapy was also assessed. Streptomycin killed the microbe in one trial but failed due to resistance amplification in the second trial. In two trials, the other antibiotics consistently reduced the bacterial densities within the pharmacodynamic systems from 108 CFU/ml to undetectable levels (<102 CFU/ml) between 1 and 3 days of treatment. None of the comparator agents selected for resistance. The comparator antibiotics were superior to streptomycin against Y. pestis and deserve further evaluation.

centrations of the antibiotics that simulated the mean serum pharmacokinetic profiles of the drugs that were reported in humans. The simulated mean serum concentration-time profiles were for the clinical regimens of gentamicin (5 mg/kg of body weight intravenously [i.v.] every 24 h (Q24h), streptomycin (1 g i.v. Q12h), ciprofloxacin (500 mg orally [p.o.] Q12h), moxifloxacin (400 mg p.o. q24h), ampicillin (2 g i.v. Q6h), meropenem (1 g i.v. q8h), and doxycycline (100 mg p.o. q12h) (following a 200-mg doxycycline loading dose) ( Table 1). The targeted pharmacokinetic parameters were for the free (non-protein-bound) fraction of these regimens (24). Another HFPM arm served as a no-treatment control. Antibiotics were given for 10 days. Throughout an experiment, bacterial samples were removed from each HFPM arm and were replaced with the same volume of fresh medium. Washed bacterial samples were quantitatively cultured on antibiotic-free agar and agar supplemented with 2ϫ to 3ϫ MIC of the corresponding treatment drug to assess the effect of that antibiotic on the killing of the total bacterial population and for resistance amplification. The lower limit of quantification for the culture assay was 50 CFU/ml. Serial medium samples were collected from the HFPM arms for measurement of drug content by liquid chromatography/tandem mass spectrometry (LC/MS/MS) to validate that the targeted pharmacokinetic profiles were simulated. At the end of the 10-day study, the total volume of bacterial suspensions in the HFPM arms that were culture negative on day 8 were quantitatively cultured on drug-free agar to determine whether the culture-negative arms were sterile. The increases in antibiotic MICs between the colonies of bacteria that grew on antibiotic-supplemented agar and the parent isolate were determined. The comparative HFPM studies were conducted twice.
Time-kill study of Y. pestis and doxycycline. Forty-eight-hour time-kill studies were conducted with Y. pestis to assess the effect of 0.25ϫ to 2ϫ MIC concentrations of doxycycline on the killing of this microbe. The starting inoculum of approximately 10 8 CFU/ml (15 ml) of bacterium replicated the starting concentration of Y. pestis used in the HFPM. Bacteria and different doxycycline concentrations were added to flasks, which were incubated at 35°C in a water-shaker bath. Medium and drug were changed every 24 h. Quantitative cultures were conducted on washed bacterial samples at the 0-, 24-, and 48-h time points.

RESULTS
Susceptibility and mutation frequency studies. The MICs, MBCs, and mutation frequencies of 2ϫ to 3ϫ the baseline MICs of the antibiotics are shown in Table 2. The MIC values for the parent Y. pestis isolate were the same as the MBCs for ciprofloxacin, moxifloxacin, ampicillin, and meropenem. For gentamicin and streptomycin, the MBCs were 1 dilution higher than the respective MICs. Doxycycline was bacteriostatic since the MBC of Ͼ64 mg/liter was at least Ͼ32ϫ the MIC value of 2 mg/liter. The mutation frequencies were slightly higher at 2ϫ MIC than at 3ϫ MIC for each antibiotic.
HFPM comparative antibiotic efficacy studies. In the first trial, the HFPM arm treated with the simulated concentrationtime profile for streptomycin (1 g i.v. q12h) showed an initial decrease in bacterial density from 10 8 to approximately 5 ϫ 10 4 CFU/ml on day 1, followed by regrowth such that the bacterial densities matched those of the control arm by day 3 (Fig. 1A). In the second trial, streptomycin sterilized the HFPM arm (Fig. 1B). Subpopulations of Y. pestis in the control arm increased in proportion to the total population, showing that microbes with decreased susceptibilities to the treatment antibiotics were present in the bacterial inoculum at the start of therapy ( Fig. 2A).
Regrowth (and treatment failure) of streptomycin therapy in the first trial was due to rapid amplification of the drug-resistant subpopulation (Fig. 2B) that was present in the bacterial suspension prior to initiation of antibiotic therapy (Table 3). These mutants had MICs to streptomycin of Ͼ256 mg/liter, compared to an MIC of 2 mg/liter for the parent strain. The streptomycin-resistant mutants did not show cross-resistance to gentamicin, since the isolates with very high MICs to streptomycin and the parent Y. pestis strain all had MICs to gentamicin of 0.5 mg/liter. In contrast, in trial 2 (in which streptomycin therapy was successful), colonies that grew on agar that was supplemented with this antibiotic had MICs to streptomycin that were 4 to 8 mg/liter. These isolates did demonstrate decreased susceptibilities to gentamicin (gentamicin MICs of 1 to 2 mg/liter) ( Table 3).
In two trials, gentamicin, ampicillin, meropenem, ciprofloxacin, and moxifloxacin consistently and rapidly killed the drugsusceptible Y. pestis population and did not amplify the subpopulations with decreased susceptibilities to these antibiotics ( Fig. 1 and 2C through 2H). Doxycycline was ineffective in our HFPM at the dose examined, despite its proven effectiveness in in vivo experimental and clinical studies (5,23). Approximately 10 2 CFU/ml of mutants resistant to 2ϫ MIC of doxycycline were present in the HFPM systems at initiation of therapy (Fig.  2H). These mutants were amplified with doxycycline therapy. Time-kill studies showed that the lack of response of Y. pestis was not due to the HFPM systems, since doxycycline exposures higher than those used in the HFPM had no effect or a marginal effect against Y. pestis in the time-kill studies (Fig. 3).

DISCUSSION
Because naturally occurring human disease due to Y. pestis is uncommon and it is unethical to intentionally infect people with a bacterium that can cause severe morbidity and mortality, it is impossible to conduct controlled, randomized, double-blinded comparative clinical trials to define the relative efficacies of different antibiotics for the treatment of human plague. Furthermore, the half-lives of drugs in small animals are frequently shorter than those in humans. Thus, animal models may underestimate the true antimicrobial efficacy of some drugs, and this makes it difficult to determine the relative efficacies of different antibiotic classes (11). The serum concentration-time profiles of antibiotics can be accurately simulated within HFPMs. This enabled our group to conduct the only study that has compared the relative antimicrobial efficacies of clinically prescribed regimens with the "gold standard," streptomycin, and other commercially available antibiotics on the killing of a pan-antibiotic-susceptible isolate and on resistance amplification. In our in vitro HFPM studies, the efficacies of ampicillin, meropenem, ciprofloxacin, moxifloxacin, and gentamicin were superior to that of the "gold standard," streptomycin. These antibiotics consistently and rapidly sterilized the HFPM arms, while streptomycin failed in one of two trials due to amplification of resistance. The failure of streptomycin in the in vitro HFPM was concordant with the results reported by McCrumb et al. (22), who reported that 2 of 7 nonhuman primates with experimental plague pneumonia failed streptomycin therapy due to emergence of resistance. Emergence of resistance has not been reported in murine models of plague pneumonia (2), perhaps because the Y. pestis inoculum of 2.6 to 4.6 log CFU/ mouse used in murine models of septicemic and inhalationinduced plague pneumonia were less than the streptomycin mutation frequency of resistance of Ϫ6.55 to Ϫ7.78 log CFU for this microbe. Furthermore, the inoculum examined in mice was substantially lower than the 1.5 ϫ 10 9 CFU of bacteria that were examined in our HFPM arms. The inoculum used in the HFPM experiments approximated the 2 ϫ 10 11 CFU total burden of Y. pestis that has been documented in the bloodstream of some people with severe plague (based on a report that up to 4 ϫ 10 7 CFU/ml of Y. pestis can be cultured from the blood of humans with plague septicemia [4] and the estimate that the mean blood volume in a human male is approximately 5 liters [15]). Also, the starting inoculum used in the HFPM studies approximated the bacterial challenge used by McCrumb and colleagues in their nonhuman primate models of severe bubonic-septicemic and pneumonic plagues (22).
Doxycycline, the only other drug (besides streptomycin) that is approved by the U.S. FDA for the treatment of plague infections, could not be assessed in the in vitro HFPM since Y. pestis exposed to a simulated clinical regimen for this drug showed the same growth profile as the no-treatment controls. This finding is consistent with our previous findings in immunenormal and neutropenic murine models of plague pneumonia using the fully virulent Y. pestis strain CO92. That study dem-onstrated that this bacteriostatic drug required the presence of neutrophils in order to optimize treatment benefit. In contrast, the bactericidal drug gentamicin provided the same microbiological effect in immune-normal and neutropenic mice (16).
The gentamicin and streptomycin regimens simulated in our studies had similar free area under the concentration-time curve (AUC)/MIC, maximum concentration of drug in serum (C max )/MIC, and time-above-MIC (TϾMIC) values (Table 1) and similar mutation frequency values ( Table 2). Yet gentamicin was successful in two trials, while streptomycin failed in one of two trials. The parent Y. pestis strain had an MIC to gentamicin of 0.5 mg/liter. MIC determinations conducted with the isolates that grew on antibiotic-supplemented agar plates in the nontreatment control arms of both hollow-fiber experiments showed that the bacterial suspensions that were inoculated into the hollow fiber systems did contain a subpopulation of bacteria that had gentamicin MICs of 1 and 2 mg/ liter. CLSI guidelines categorize Y. pestis isolates with MICs to gentamicin of Յ4 mg/liter as susceptible to this aminoglycoside antibiotic (7). Thus, the strains with decreased susceptibilities to gentamicin were predicted to be killed by the clinically prescribed regimens of gentamicin that were simulated in our experiments.
For the hollow-fiber study in which streptomycin therapy was successful, the colonies that grew on streptomycin-supplemented agars in the nontreatment control arm and in the streptomycin therapy arm prior to initiation of therapy had MICs to this antibiotic of 4 to 8 mg/liter (compared to 2 mg/liter for the wild-type strain). CLSI guidelines suggest that Y. pestis isolates with streptomycin MICs of Յ4 mg/liter are susceptible to this antibiotic and isolates with an MIC of 8 mg/liter fall in the "intermediate" category, in that treatment success may occur if high enough drug exposure is achieved in the site of infection (7). In the second trial, streptomycin therapy rapidly killed the wild-type and less-susceptible bacterial populations, suggesting that the clinical regimen of streptomycin of 1 g given every 12 h would be successful against Y. pestis infections in humans in which a small proportion of the bacterial population has an MIC to streptomycin as high as 8 mg/liter.
In contrast, in the hollow-fiber experiment in which streptomycin therapy had failed, the streptomycin MICs for isolates in a Streptomycin therapy failed in trial 1 but was successful in trial 2. The wild-type Y. pestis isolate had a streptomycin MIC of 2 mg/liter, and the antibiotic-supplemented agar contained 4 mg/liter of streptomycin.
b These isolates had gentamicin MICs of 0.5 mg/liter, identical to the gentamicin MIC for the wild-type Y. pestis strain.
c These isolates had gentamicin MICs of 1 to 2 mg/liter compared to a gentamicin MIC for the wild-type strain of 0.5 mg/liter. d NG, no growth on agar supplemented with streptomycin at 2ϫ the MIC of the wild-type strain.
the nontreatment control arm that grew on agar supplemented with this antibiotic had streptomycin MICs of Ͼ256 mg/liter, compared with an MIC of 2 mg/liter for the parent strain. Additional studies using agar plates that contained 128 mg/liter of streptomycin determined that the mutation frequency was Ϫ8.46 to Ϫ9.03 log CFU for isolates that had MICs to streptomycin of Ͼ256 mg/liter. Since the hollow-fiber cartridges were inoculated with 10 9 CFU of bacteria (10 ml of 10 8 CFU of Y. pestis), it was a "chance event" whether or not a hollowfiber system was inoculated with Y. pestis isolates with highlevel streptomycin resistance. Genetic studies to define the mechanism of streptomycin resistance that resulted in treatment failure with this antibiotic in one but not the other trial were not conducted in this project. Streptomycin treatment failure was due to Y. pestis isolates that had very high MICs to this antibiotic. However, these isolates did not show cross-resistance with gentamicin. Only bacteria that have a mutation in the 30S ribosomal binding site for streptomycin manifest this aminoglycoside-susceptible profile (10,14,26). The isolates with MICs for streptomycin of 4 to 8 mg/liter that were detected in the trial in which streptomycin therapy was successful likely expressed aminoglycoside-modifying enzymes with or without efflux pump overexpression (10,26), since these isolates did have increased MICs to gentamicin.
Gentamicin and ciprofloxacin have been used successfully for the treatment of plague pneumonia in animal models (5) and consistently and rapidly killed Y. pestis in the current investigation. Case reports and noncomparative studies suggest that these antibiotics are efficacious for the treatment of plague in humans (18,23). Our in vitro results suggest that these drugs would be equally effective in the treatment of Y. pestis infections in humans.
Ampicillin and meropenem have in vitro efficacy (5, 12, 27; this study). Little data are available regarding the efficacy of these two antibiotics for the treatment of plague in humans. However, in one murine study, the administration of amoxicillin to mice 24 and 48 h after infection with Y. pestis was as efficacious as that of gentamicin (5). Mice with experimental plague pneumonia that were treated with ampicillin beginning 24 h after infection had an 85% survival rate. The regimen given to those mice provided a time-above-MIC of 33% (5), which was much less than the TϾMIC value of 100% that was simulated in our HFPM. In a separate project, dose fractionation studies conducted in our HFPM showed that a timeabove-MIC value of Յ79% of the dosing interval was associated with treatment failure and a time-above-MIC value of Ն92% was predictive of treatment success for ampicillin (21). Since the human clinical regimen of ampicillin of 2 g i.v. given every 6 h that was simulated in the current project generated a time-above-MIC value of 100%, it is likely that the efficacy of this drug in humans would be better than what is predicted from murine infection models.
However, it is important to note that when initiation of ampicillin therapy was delayed to 42 h after infection, the treated mice died more rapidly than the nontreated controls (5). Although not measured, this suggests that the more rapid death was due to endotoxin release by this bacterium as it is killed by beta-lactam antibiotics. Thus, studies need to be conducted to determine if the earlier deaths were indeed due to endotoxin release. Also, studies in other animal species are needed to define whether the severe sepsis induced by betalactam therapy for Y. pestis in mice occurs in other animal infection models.
In summary, the experiments using the HFPM found that simulated clinical regimens for ciprofloxacin, moxifloxacin, gentamicin, ampicillin, and meropenem were superior to the gold standard, streptomycin, in eradicating a high bacterial burden of Y. pestis that could be found in plague septicemia and pneumonic plague. While streptomycin therapy failed due to amplification of resistant mutants in one of two trials, the comparator antibiotics rapidly eradicated the bacteria in both trials. Since streptomycin is currently not available in many countries, including the United States, and since persons with severe plague infections may harbor high bacterial burdens that may include subpopulations with high-level streptomycin resistance, our data suggest that the other antibiotics that were evaluated in the current study should be preferred over streptomycin for the treatment of plague infections.
Importantly, the doses of the drugs examined in this project simulated the mean concentration-time profiles for clinically prescribed regimens. This suggests that at least 50% of the people who would receive the dosages of the antibiotics that were simulated in the HFPM would be cured of plague. Doserange studies and mathematical modeling (i.e., Monte Carlo simulations) are needed in order to provide a more accurate prediction of the overall efficacy of the evaluated drugs for the treatment of Y. pestis infections in humans.