Evaluation of Imipenem for Prophylaxis and Therapy of Yersinia pestis Delivered by Aerosol in a Mouse Model of Pneumonic Plague

Abstract

It has been previously shown that mice subjected to an aerosol exposure to Yersinia pestis and treated with β-lactam antibiotics after a delay of 42 h died at an accelerated rate compared to controls. It was hypothesized that endotoxin release in antibiotic-treated mice accounted for the accelerated death rate in the mice exposed to aerosol Y. pestis. Imipenem, a β-lactam antibiotic, binds to penicillin binding protein 2 with the highest affinity and produces rounded cells. The binding of imipenem causes cells to lyse quickly and thereby to release less free endotoxin. Two imipenem regimens producing fractions of time that the concentration of free, unbound drug was above the MIC (fT>MIC) of approximately 25% (6/24 h) and 40% (9.5/24 h) were evaluated. In the postexposure prophylaxis study, the 40% and 25% regimens produced 90% and 40% survivorship, respectively. In the 42-h treatment study, both regimens demonstrated a 40 to 50% survivorship at therapy cessation and some deaths thereafter, resulting in a 30% survivorship. As this was an improvement over the results with other β-lactams, a comparison of both endotoxin and cytokine levels in mice treated with imipenem and ceftazidime (a β-lactam previously demonstrated to accelerate death in mice during treatment) was performed and supported the original hypotheses; however, the levels observed in animals treated with ciprofloxacin (included as an unrelated antibiotic that is also bactericidal but should cause little lysis due to a different mode of action) were elevated and significantly (7-fold) higher than those with ceftazidime.

INTRODUCTION

Yersinia pestis, the agent of plague, is susceptible to a number of different antimicrobials. Aminoglycosides, such as streptomycin or gentamicin, and tetracyclines, such as doxycycline, have been recommended as treatments of choice for this pathogen (1). Recently, we published an evaluation of gentamicin, doxycycline, and levofloxacin in an inhalational animal model of Yersinia pestis pneumonia (2). In this evaluation, we demonstrated that gentamicin was superior to doxycycline when the animals were rendered neutropenic by cyclophosphamide. We also demonstrated the superiority of the fluoroquinolone levofloxacin, in both the normal and neutropenic settings.

Resistance to a number of classes of antibacterial agents has been demonstrated in Y. pestis (3, 4). In these cases the appearance of antibiotic resistance appears to be by natural acquisition of transmissible plasmids. In a biowarfare/bioterrorism event, such resistances could be intentionally introduced; consequently, having different therapeutic options for this pathogen is a matter of great importance.

β-Lactam antibiotics have been a mainstay of therapy for decades, especially for seriously ill patients in an intensive care setting. It would be of considerable use to have this class of agents available and in the physician's therapeutic armamentarium for seriously infected patients, particularly those with pneumonic plague, where mortality rates are exceptionally high and are expected to be 80 to 100% when untreated (5).

Byrne and colleagues (6) extensively evaluated a large number of therapeutic agents in this inhalational mouse model of Y. pestis pneumonia. Surprisingly, β-lactam therapy resulted in uniformly poor outcomes when therapy was delayed for 42 h, despite the fact that the mice received appropriate doses frequently (on an every-6-h schedule). For all β-lactam antibiotics, death was faster than in the no-treatment control animals. This counterintuitive finding led us to hypothesize that β-lactam therapy caused a massive release of endotoxin, resulting in early death of the animals, even though cell killing was progressing adequately. Therefore, we decided to evaluate imipenem, an antimicrobial not tested in the previous study. Imipenem binds with greatest affinity to penicillin binding protein (PBP) 2 in the Enterobacteriaceae, which results in the formation of rounded forms after treatment, and the release of endotoxin is much less after cell lysis, at least in vitro (7).

We further hypothesized that the bacterial burden would have an impact on the outcome with these agents, as higher burdens would be likely to result in more endotoxin release during the initial stages of therapy. To examine this, we used different durations of therapeutic delay after the mice were exposed to a Y. pestis aerosol. A delay of 24 h before treatment represented a postexposure prophylaxis (PEP) scenario, where the bacterial burden would be smaller, and a 42-h delay in treatment represented a therapeutic scenario, when mice (and patients) would be expected to be symptomatic and where the bacterial burden would be considerably larger. Here, we wished to examine survivorship with three quite different antibiotics (imipenem, ceftazidime, and ciprofloxacin) using two different delay periods before the onset of therapy (24 and 42 h). In addition we wished to examine the substrate for any outcome differences noted by examining proinflammatory and anti-inflammatory cytokines and directly measuring circulating endotoxin.

MATERIALS AND METHODS

All work with Y. pestis, infected animals, and tissues was conducted in a fully accredited biosafety level 3 (BSL3) facility at USAMRIID, Ft. Detrick, MD. All methods were generally as published previously (2).

Mice.

Female BALB/c mice (20 g) were obtained from Charles River-National Cancer Institute and were used for all experiments. The mice had free access to food and water throughout the course of the study. During the study, if mice became moribund, they were humanely euthanized. The time of death was recorded as the time of euthanasia, and these animals were included in all analyses.

The research was conducted under an Institute Animal Care and Use Committee-approved protocol in compliance with the Animal Welfare Act and other federal statutes and regulations relating to animals and experiments involving animals and adhered to principles stated in the Guide for the Care and Use of Laboratory Animals (National Research Council, 1996). The facility where this research was conducted (USAMRIID) is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International.

Preparation of the Y. pestis challenge strain for aerosolization.

Y. pestis CO92 (provided by T. Quan, Centers for Disease Control and Prevention, Fort Collins, CO) was originally isolated in 1992 from a fatal human case of pneumonic plague (8). The 50% lethal dose (LD₅₀) of this strain in mice is 6.8 × 10⁴ CFU inhaled when administered as an aerosol (whole body) (2).

The inoculum for aerosol challenge was prepared as previously described (6), and the suspension of Y. pestis was diluted to the appropriate aerosol challenge dose. Colony counts were determined by serial dilution and plating on sheep blood agar plates (SBAP). These plates were incubated for 2 days at 28°C, and colonies were counted.

Aerosol infection.

Targeted inhaled doses of 20 LD₅₀s (1 LD₅₀ = 6.8 × 10⁴ CFU) of Y. pestis were administered to mice by whole-body aerosol. Aerosol was generated by a three-jet Collison nebulizer. All aerosol procedures were controlled and monitored using the Automated Bioaerosol Exposure System (Biaera Technologies, Hagerstown, MD) (9) operating with a whole-body rodent exposure chamber. Integrated air samples (6 liters/min) were obtained from the chamber during each exposure using an all-glass impinger (AGI). AGI samples were serially diluted and plated on SBAP, as described above. The inhaled dose (CFU/mouse) of Y. pestis was calculated from these counts over the time of exposure (10 min), and the mouse respiratory rate (minute volume) estimated as described by Guyton (10).

MIC determinations.

MIC values were determined using standard techniques as recommended by the Clinical and Laboratory Standards Institute (11) in cation-adjusted Mueller-Hinton broth and were carried out at 35°C. Due to the slow growth of Y. pestis, endpoints were determined at both 24 and 48 h. The 30 Y. pestis strains used in this study were from the USAMRIID collection, selected to represent the biovars, genotypes, and isotypes (12, 13) (see Table S1 in the supplemental material). Six strains were Y. pestis biovar Antiqua, six were Y. pestis biovar Medievalis, 16 were Y. pestis biovar Orientalis, and two atypical isolates could not be assigned to any biovar.

Antibiotics.

Imipenem-cilastatin was obtained from Merck Research Laboratories. Ciprofloxacin and ceftazidime were obtained from Bayer Healthcare Pharmaceuticals and GlaxoSmithKline Pharmaceuticals, respectively. Imipenem, ceftazidime, ciprofloxacin, or 0.85% saline (placebo) was administered by intraperitoneal (i.p.) injection in a volume of 0.2 ml for 5 days. Initiation of therapy occurred at 24 h after Y. pestis exposure for the prophylaxis (PEP) experiment and at 42 h after exposure for the treatment experiment. The antibiotic dose and schedule for imipenem-cilastatin were 25 mg/kg every 4 or 6 h to yield fractions of time that the concentration of free, unbound drug was above the MIC (fT>MIC) of 25% and 40% based on the mouse pharmacokinetic analysis. The imipenem half-life of 13 min in the mouse made it experimentally unfeasible to achieve a longer fT>MIC beyond 40%. As comparators, ceftazidime at 300 mg/kg every 6 h or ciprofloxacin at 30 mg/kg every 12 h was given to groups of 10 mice each.

Assessment of efficacy.

Aerosol exposures were conducted in 5 runs, with a mean of 10.4 LD₅₀s (range, 8.3 to 15.7) achieved. Two animals from each aerosol run were placed in each experimental group to balance the exposure population. The cohort size for statistical evaluation was 10 mice. Mortality was assessed and recorded every 4 to 6 h during antibiotic administration (5 days) and twice daily thereafter to 21 days after therapy initiation. Surviving animals were euthanized at the end of the study. Spleens from three animals in each group were removed, weighed, and homogenized in 1 ml of physiological saline, and 100 μl of undiluted sample and of a 10-fold dilution were plated on SBAP and incubated as described above to determine any residual infection.

Murine pharmacokinetics.

Single doses of a range of three dose sizes (50, 75, and 100 mg/kg) of imipenem-cilastatin were administered intraperitoneally to naive BALB/c mice. Mice were humanely euthanized in groups of three at seven time points over 8 h. Blood was obtained by cardiac puncture, separated into plasma, and assayed for drug concentration by high-performance liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) as previously published (14). While determination of levels in infected animals would be desirable, the added process time for decontamination and transfer of plasma samples from biocontainment would result in degradation of the imipenem.

All samples were analyzed simultaneously using a nonparametric determination approach employing the ADAPT II program of D'Argenio and Schumitzky (15). Multiple models were evaluated and discriminated by the likelihood ratio test and the Akaike information criterion (16). Weighting was as the inverse of the observation variance. Point estimates of the mean parameter values were used to calculate the fT>MIC for a dosing interval at steady state after correcting for protein binding.

Serum cytokine values.

For the cytokine/endotoxin studies, aerosol exposures were conducted in 5 runs with a mean of 10.0 LD₅₀s (range, 7.1 to 13.4). Six animals from each aerosol run were placed in each experimental group to balance the exposure population of the four experimental groups. Groups of 30 mice received either saline, imipenem at 25 mg/kg every 6 h, ceftazidime at 300 mg/kg every 6 h, or ciprofloxacin at 30 mg/kg every 12 h beginning 42 h after challenge. Serum samples were collected by terminal cardiac bleeds at 0, 24, 36, and 42 h for the control group (saline only) and at 45, 48, 51, 54, 57, and 60 h for the control group and all treatment groups. The amounts of specific cytokines in the sera of mice (3 mice per time point) during treatment with or without antibiotics were determined using the BDFACSArray Bioanalyzer system with the Cytometric Bead Array assay (BD Biosciences, San Diego, CA). Mouse cytokine tumor necrosis factor alpha (TNF-α,) gamma interferon (IFN-γ), interleukin-12 p70 (IL-12p70), IL-4, and IL-10 flex sets were used following the manufacturer's directions with detection limits of 2.8 pg/ml, 0.5 pg/ml, 1.9 pg/ml, 0.3 pg/ml, and 9.6 pg/ml, respectively. The results were reported as the geometric mean of the cytokine level for each time point. In some cases we were unable to recover enough serum to measure the cytokine level at a specific time period from a set of mice. The missing data are noted in the figure legends.

Endotoxin assays.

Serum endotoxin levels were measured by a standardized, quantitative turbidimetric Limulus amoebocyte lysate assay (Associates of Cape Cod, Woods Hole, MA). Samples were handled using endotoxin-free glassware, reagents, and pipettes. Serum was heat treated and diluted 1:10 to remove serum inhibitors of the reaction as recommended by the manufacturer. The limit of detection for endotoxin in serum by the assay was 20 pg/ml.

Statistical analysis.

Groups were compared by hypothesis. The first hypothesis was that imipenem-cilastatin would work pharmacodynamically in much the same way that it does in other Enterobacteriaceae (17). The second hypothesis was that increasing the bacterial load would blunt differences in outcomes between dosing groups because of the septic changes induced by the longer interval before therapy initiation. All analyses were performed employing a stratified Kaplan-Meyer analysis with a log rank test and a Cox proportional hazards model as implemented in SYSTAT for Windows v 11.0. Treatment groups of 10 animals allowed adequate (>80%) power. This sample size would allow the experiment to detect a minimum efficacy rate of 90% (9/10 surviving) in the treatment group compared to ≤30% (3/10 surviving) in the control group with a 95% confidence level.

RESULTS

MIC values.

The MIC distributions for 30 genetically and geographically diverse strains of Y. pestis are shown in Fig. 1. The MICs for the Y. pestis strains with imipenem ranged from a low of 0.12 mg/liter to a high of 1.0 mg/liter (Fig. 1A). The MICs for the Y. pestis strains against ceftazidime ranged from 0.03 mg/liter to a high of 0.5 mg/liter (Fig. 1B). The MICs for ciprofloxacin against the Y. pestis strains gave the narrowest range: from a low of 0.008 mg/liter to a high of 0.03 mg/liter (Fig. 1C). The MICs for Y. pestis CO92, used in the present study, for imipenem, ceftazidime, and ciprofloxacin were 0.5 mg/liter, 0.25 mg/liter, and 0.03 mg/liter, respectively.

FIG 1.

Susceptibility (MIC) distribution for genetically and geographically diverse strains of Y. pestis (n = 30). (A) Susceptibility to imipenem (mg/liter). (B) Susceptibility to ceftazidime (mg/liter). (C) Susceptibility to ciprofloxacin (mg/liter). Susceptibilities of Y. pestis strains were determined by the broth microdilution method according to CLSI M45-A2 (11).

Drug pharmacokinetics.

The pharmacokinetic parameter values for imipenem (in the presence of cilastatin) in BALB/c mice are displayed in Table 1. The clearance of imipenem in the mouse is roughly 2 times the human rate (18). Using these mouse imipenem-pharmacokinetic values and protein binding of 4% (18) to calculate fT>MIC values of 25% and 40%, respectively, doses of 25 mg/kg every 6 h and every 4 h would be required.

TABLE 1.

Pharmacokinetic parameters at steady state for imipenem in BALB/c mice

Parameter (unit)a	Value
Vc (liters)	0.0233
CLd (liters/h)	0.361
Vp (liters)	0.0011
Ka (h−1)	11.65
SCL (liters/h)	0.095
t1/2 (h)	0.21

^aV_c, volume of the central compartment; CL_d, distributional clearance; V_p, volume of the peripheral compartment; K_a, absorption rate constant; SCL, serum clearance; t_1/2, half-life.

Regimen efficacy.

Survivorship is displayed in a Kaplan-Meyer plot for imipenem for the two doses studied in Fig. 2A for the PEP experiment. For the treatment experiment, these data are displayed in Fig. 2B. As the experiments were conducted simultaneously, there was a single control (no-treatment) group for both experiments. Survivorships for ciprofloxacin and ceftazidime therapy, determined as a separate experiment, are displayed in Fig. 3A and B.

FIG 2.

Effects of two imipenem treatment regimens on survival of BALB/c mice exposed to aerosolized Y. pestis CO92. Imipenem at 25 mg/kg was given every 4 h (fT>MIC, 40%) (triangles) or given every 6 h (fT>MIC, 25%) (circles) after an inhalational exposure to Y. pestis CO92. Because the studies were done simultaneously, the same no-treatment saline control group at an fT>MIC of 0 (squares) was used for both panels. (A) PEP treatment with imipenem was started at 24 h postinfection (overall Mantel test, P < 0.0000009. Directly contrasting 25% and 40% fT>MIC is also significant (P = 0.018), even after α-decay using a Bonferroni adjustment. (B) Imipenem treatment was the same as for panel A except that treatment was started at 42 h postinfection (delayed treatment). The difference noted between treatment groups and the no-treatment control was not statistically significant (n = 10 mice for all groups).

FIG 3.

Effects of two different treatment regimens of ciprofloxacin or ceftazidime therapy on survival of BALB/c mice infected with aerosolized Y. pestis CO92. Ceftazidime was given at 300 mg/kg every 6 h (circles), and ciprofloxacin was given at 30 mg/kg every 12h (triangles). The same no-treatment control group (squares) is shown in panels A and B, as the studies were performed simultaneously. (A) PEP treatment of mice with ceftazidime or ciprofloxacin was started at 24 h postinfection. The difference in the number of survivors between the ceftazidime or ciprofloxacin PEP treatment group and the control group of mice was found to be significant (P < 0.0001). (B) A delayed treatment of mice with ceftazidime or ciprofloxacin was started at 42 postinfection. The number of survivors in the ciprofloxacin delayed-treatment group was found to be significant compared with that of the control group (P = 0.0061), but the difference between the number of survivors in the ceftazidime delayed-treatment group compared to that in the control group was not significant (n = 10 mice for all groups).

For imipenem-cilastatin, the contrast of the control group with the two groups treated after a 24-h delay in the PEP experiment was significant (Mantel test P value, ≪0.001). The first treatment group was 25 mg/kg administered i.p. every 6 h, which covered 25% of the dosing interval. The second group was treated with 25 mg/kg of imipenem-cilastatin every 4 h, which covered 40% of the dosing interval with free drug in excess of the MIC. The two treatment regimens were significantly different when compared directly (P = 0.018, α-decay by Bonferroni adjustment). When fT>MIC was examined as a covariate in a Cox model, the results were statistically significant (parameter estimate = −0.1861; standard error [SE] = 0.05916; P = 0.001656; 95% confidence interval for the estimate = −0.302 to −0.070).

When treatment was delayed for 42 h (treatment experiment), the contrast was no longer statistically significant, although the imipenem group had 30% survivors remaining, while the control group had 100% mortality (Fig. 2B). Neither treatment regimen alone was significantly different from controls in the treatment experiment. The contrast of treatment and PEP was significant (P = 0.013 by the Mantel test).

When ciprofloxacin and ceftazidime were tested, both performed well in the PEP experiment (100% and 90% survivorship, respectively) (Fig. 3A). However, in the 42-h delay treatment experiment, ciprofloxacin therapy resulted in 50% survivorship, while treatment with ceftazidime resulted in only 10% survivorship (Fig. 3B).

Cytokines in serum after challenge and antibiotic treatment.

Because it was hypothesized previously that the accelerated death of mice exposed to aerosol Y. pestis was possibly the result of released endotoxin after treatment of infected mice with β-lactam antibiotics, we monitored the serum for the appearance of proinflammatory (tumor necrosis factor alpha [TNF-α], gamma interferon [IFN-γ], and interleukin-12 p70 [IL-12p70]) and anti-inflammatory (IL-4 and IL-10) cytokines starting 3 h after beginning treatment at 42 h with imipenem, ceftazidime, or ciprofloxacin as a possible indication of endotoxin release. For the control group (saline only), we measured cytokines from the time of infection. In the control group (saline only), we first detected the appearance of a large amount of TNF-α (473 pg/ml) at 45 h postinfection (p.i.) in the untreated mice, which quickly decreased between 48 and 51 h p.i. and rose slightly to 169 pg/ml at 54 h before decreasing slightly (103 pg/ml) up to 60 h p.i. (Fig. 4A). In mice that received imipenem, TNF-α levels remained relatively low throughout the course of the study (80 pg/ml at 45 h). In mice that received ceftazidime, there was a moderate increase in TNF-α levels at 51 h to 54 h (168 to 173 pg/ml, respectively), after which the levels dropped to barely detectable (5 pg/ml) at 57 h p.i. In contrast to the case for other antibiotic-treated mice, TNF-α levels reached a maximum at 48 h (491 pg/ml) in the ciprofloxacin-treated mice, before decreasing quickly at 51 h p.i. (30 pg/ml), only to increase slightly again at 54 and 57 h (72 and 106 pg/ml, respectively).

FIG 4.

Postinfection levels of proinflammatory and anti-inflammatory cytokines in the sera of control (saline) and antibiotic-treated mice. Mice were challenged with aerosolized Y. pestis CO92 and treated with imipenem, ceftazidime, ciprofloxacin, or saline (control) at 42 h postinfection. The cytokine levels of TNF-α (A), IFN-γ (B), IL-12p70 (C), IL-10 (D), and IL-4 (E) were measured at 3- to 6-h intervals up to 60 h in the control mice, and the cytokine levels were measured every 3 h in the antibiotic-treated mice beginning 3 h after (45 h postinfection) the start of the delayed antibiotic treatment (42 h). Each bar represents the mean from three animals. Treatments were saline (no-treatment control) (black), imipenem (blue), ceftazidime (green), or ciprofloxacin (red). Cytokine levels were not obtained for the following samples and time points: (C) IL-12p70, control at 51 h; (D) IL-10, control at 51 h; (E) IL-4, control at 51 h and imipenem 54 h and 60 h.

Unlike TNF-α levels in the sera of untreated mice, we saw a moderate amount of IFN-γ (719 pg/ml) at 45 h, which then dropped quickly to baseline (26 pg/ml) at 51 h. It then rose to 4,404 pg/ml 54 h p.i. but quickly fell at 57 h (1,690 pg/ml) and 60 h p.i. (99 pg/ml) (Fig. 4B). In mice that received imipenem, we saw only a slight increase in IFN-γ levels after 48 h and 51 h p.i. (866 and 681 pg/ml, respectively) before they decreased to baseline levels (30 pg/ml) at 60 h p.i. In ceftazidime-treated mice, there was a slow rise in IFN-γ levels from 45 h to 51 h p.i. (155 to 1,054 pg/ml, respectively) before there was a large increase (8,669 pg/ml) at 54 h p.i., which occurred at the same time as seen in the control mice, before it dropped down to 36 pg/ml at 57 h p.i. and 135 pg/ml at 60 h p.i. On the other hand, in ciprofloxacin-treated mice, IFN-γ first appeared at 45 h p.i. (2,234 pg/ml) and reached a maximal level at 48 h p.i. (8,177 pg/ml), before it rapidly dropped at 51 h p.i. (270 pg/ml) and further decreased to 94 pg/ml at 60 h p.i. The appearance and changes in relative levels of IL-12p70 in the serum were similar to those seen with IFN-γ with all antibiotic treatments (Fig. 4C).

The anti-inflammatory cytokines IL-10 and IL-4 were detected in the control mice at the time of challenge (Fig. 4D and E, respectively). We detected a low level of IL-10 from 0 to 42 h p.i. (6 to 7 pg/ml) in the animals treated with saline only. A peak of IL-10 expression occurred at 45 h p.i. (18 pg/ml) in the control mice, and then levels declined before slightly rising (14 to 15 pg/ml) from 54 to 60 h p.i. For the imipenem-treated mice, the levels of IL-10 remained within a narrow range of 8 to 13 μg/ml throughout the study (45 to 60 h p.i.). With the ceftazidime-treated mice, there was a slight increase in IL-10, which peaked at 54 h p.i. (19 pg/ml) before decreasing to 12 pg/ml at 60 h p.i. In the ciprofloxacin-treated mice we saw a slow rise in IL-10 levels at 45 to 54 h p.i. before they peaked at 57 h p.i. (22 pg/ml) and dropped down to 14 pg/ml at 60 h p.i. The amount of IL-4, on the other hand, remained relatively constant (5.0 pg/ml) from 0 to 60 h p.i. in the control mice, and similarly, the levels of IL-4 varied little in all the antibiotic-treated mice from 45 h to 60 h p.i. (5 to 6 pg/ml) when it could be detected.

Endotoxin concentrations.

Endotoxin concentrations were measured for no-treatment control mice as well as for each group of antibiotic-treated mice. As expected, the no-treatment control mice had, by far, the highest endotoxin concentrations (Fig. 5A; Table 2). Very little endotoxin was seen through 45 h p.i. At 48 h p.i., we saw a plasma concentration of 321 ng/ml, and the concentration stayed elevated through 60 h p.i. It should be noted that in the control groups in the two efficacy experiments, 100% of animals survived through 60 h p.i., but 70 to 80% of mice succumbed by 72 h p.i., and 100% of the mice succumbed by 96 to 120 h p.i.

FIG 5.

Postinfection serum levels of endotoxin measured at 3- to 6-h intervals up to 60 h in antibiotic-treated mice. Mice were aerosolized with Y. pestis CO92 and treated with saline (no-treatment control) (A) or with imipenem (blue), ceftazidime (green), or ciprofloxacin (red) (B). Note that the scale in panel B is 25% of that in panel A. Each bar represents the mean from three animals except for the following samples and time points: control at 36 and 48 h, n = 1, and at 42 h, n = 2; imipenem at 45 and 60 h, n = 2, and at 54 h, n = 0; ceftazidime at 45 h, n = 1, and at 48 h and 51 h, n = 2; and ciprofloxacin at 54 h, n = 1.

TABLE 2.

Total serum endotoxin from treatment groups between 42 h and 60 h postchallenge

Treatment	Area under the curve, μg · h/ml (% of control value)
Control	2,688 (100)
Imipenem	34 (1)
Ceftazidime	76 (3)
Ciprofloxacin	379 (14)

For the antibiotic-treated animals, different endotoxin profiles and lower levels were generally seen (Fig. 5B; Table 2). For the imipenem-treated mice, no endotoxin was seen until 45 h p.i., when only 8.1 ng/ml was detected. It then rapidly declined through 60 h p.i., when only 0.05 ng/ml was measured. For the ceftazidime-treated mice, as seen with the no-treatment control mice and imipenem-treated mice, no endotoxin was seen until 45 h p.i., when 1.4 ng/ml was detected. We then saw a small peak at 54 h p.i., when 13.4 ng/ml was measured. At 57 h p.i., little endotoxin was detected, but at 60 h p.i., 3.6 ng/ml was detected. This profile is distinct from that for imipenem-treated mice, where lower concentrations were seen and for a prolonged time frame. Somewhat surprising were the endotoxin results seen with the ciprofloxacin-treated mice, where the first detection of endotoxin was seen at 45 h p.i. (8.4 ng/ml), as seen in the other antibiotic-treated mice. The endotoxin level peaked at 48 h p.i. at 69.5 ng/ml before it declined to less than 1 ng/ml at 51 h p.i., and then it rebounded to a second peak at 57 h p.i. to 48.8 ng/ml and then decreased to ∼4 ng/ml at 60 h p.i.

The time frames of detection and linkage between TNF-α and endotoxin increases with each therapy are also of interest. For controls, there is a large spike of endotoxin after hour 45, which remains elevated. TNF-α also spiked at 45 h and remained quite elevated through 60 h (Fig. 6). For imipenem, the endotoxin spike is detected at hour 45 and is immediately followed by a TNF spike at 48 and 51 h. This TNF spike is smaller than that seen with the other active treatments (Fig. 6). For ceftazidime, we see some endotoxin at 45 h, which then climbs and peaks at hour 54 but remains measurable through the 60-h time point. TNF-α climbs from 45 through 51 h, where it peaks at a substantial concentration of 342.3 pg/ml (Fig. 6). For ciprofloxacin, we see some endotoxin detectable at 45 h, which peaks at 48 h, and a profile of TNF-α where quite high concentrations are seen early (189.4 pg/ml at 45 h and 685.2 pg/ml at 48 h) and elevated levels are seen through 57 h (Fig. 6). All this is consistent with TNF-α release being stimulated by the profile of endotoxin release.

FIG 6.

Coplots of TNF-α and endotoxin levels over the duration of sampling for each therapy. Black, TNF-α; red, endotoxin. Note that the endotoxin scales for antibiotic treatments are 25% of the control scale. Each point represents the mean from three animals except for the following endotoxin samples and time points: control at 36 and 48 h, n = 1, and at 42 h, n = 2; imipenem at 45 and 60 h, n = 2, and at 54 h, n = 0; ceftazidime at 45 h, n = 1, and at 48 h and 51 h, n = 2; and ciprofloxacin at 54 h, n = 1.

DISCUSSION

The use of injectable β-lactam antibiotics for treating pneumonic plague is highly controversial. Previous mouse studies by Byrne and colleagues (6) evaluated multiple β-lactam antibiotics and also evaluated early and late intervention, as was the case here. In the study by Byrne et al., five cephalosporins were examined, as well as aztreonam and ampicillin, but a carbapenem was not evaluated. In early intervention (PEP) studies, good activity was seen, with drugs with longer half-lives doing better. The exception was ampicillin, which had the overall shortest half-life but still performed well. When late intervention was examined, all the β-lactam antibiotics had times to death as short as or shorter than that with the normal saline control. This indicates an active process causing the earlier deaths relative to those of the control animals, and the authors hypothesized that endotoxin release was one possible explanation.

In our present experiments, imipenem produced good survivorship when the delay between exposure and intervention was short (PEP scenario) and when the bacterial burden would be smaller. However, when the delay in treatment was increased to 42 h (treatment scenario), when the bacterial burden would be larger (6, 19), there was a decrease in survivorship. However, in the two different imipenem regimens, five or six deaths occurred at the same time as in the untreated controls, with the two treatment groups having one or two extra deaths after therapy ended at day 5, leaving 30% survivorship overall for the two imipenem treatment regimens. This was quite different from the previously reported study by Byrne et al. (6), where all β-lactam regimens killed as fast or faster than the no-treatment controls in the treatment (nonprophylaxis, 42-h delay) scenario and all regimens had 10% or fewer survivors at day 5 (end of therapy). The difference in the two studies, besides the use of imipenem in the present study, may have been because the amount of challenge dose and method of challenge were different. In the study by Byrne et al., 1 LD₅₀ for the inhaled dose was reported as 2.3 × 10⁴ CFU and 100 LD₅₀s were given (6) (7.3 ×10⁶ CFU/mouse), while in the present study 1 LD₅₀ was 6.8 × 10⁴ CFU by whole-body aerosol exposure, which was determined in an earlier study (2), and 20 LD₅₀s were given (1.36 ×10⁶ CFU/mouse). The other difference between the study by Byrne et al. and the present study was that in the former the challenge dose was given by a nose-only exposure (6), while in the present study a whole-body aerosol exposure was used during the challenge. It should be noted that the mean times to death were similar in both of these studies. In addition, a natural history study using the whole-body exposure indicates that the differences in the two exposures may not be that great, as the initial lung load was observed to be between 2 and 3 logs (19).

In the PEP scenario, imipenem-cilastatin performed well, and there were differences noted in the number of survivors between the two regimens (25% and 40% fT>MIC). Both Kaplan-Meyer analysis and Cox proportional hazards modeling demonstrated this finding. In other Enterobacteriaceae, near-maximal cell death was demonstrated in a mouse thigh infection model when the fT>MIC met or exceeded 40% of the dosing interval (18). In our study, the regimen producing 40% fT>MIC resulted in 90% survival, whereas the 25% fT>MIC regimen (near stasis) resulted in only a 40% survivorship. In contrast to the treatment scenario, all deaths in the PEP scenario occurred after therapy ceased at day 5. This suggests that at 40% fT>MIC, imipenem was able to clear the vast majority of Y. pestis CO92 by day 5. On the other hand, for the less intense 25% fT>MIC regimen, the majority of the animals still had a significant bacillary burden at day 5 postinfection. The difference in survivorship between regimens was significant when compared directly (P = 0.018), in the absence of the no-treatment control group and decaying α for multiple comparisons.

Note that the very short half-life of imipenem, even in the presence of cilastatin (approximately 13 min), in BALB/c mice means that significantly prolonging the fT>MIC can be done most efficiently only by shortening the dosing interval and not by simply increasing the dose, even at relatively short dosing intervals such as 4 h. It should be noted that for imipenem, each dose doubling provided only 13 min more fT>MIC coverage. To go from 25% fT>MIC (25 mg/kg every 6 h) to 40% fT>MIC and keeping the same 6-h interval, one would need to administer a dose of approximately 450 mg/kg. Additionally, application of even shorter dosing intervals below 4 h becomes increasingly more difficult under biocontainment conditions and because of the large number of mice needed for the study.

We examined both proinflammatory cytokines and anti-inflammatory cytokines and also measured endotoxin. Some clear lessons were learned. As one may expect, endotoxin was measured at much higher levels (∼4 to 5-fold) in the no-treatment control animals than in the antibiotic-treated animals. Among the antibiotic-treated animals, imipenem treatment gave the lowest measurable endotoxin concentrations and suppressed the release of TNF-α, which was lower than seen with the other two antibiotics. A dichotomy appeared, however, in the murine responses to ceftazidime and ciprofloxacin, where treatment with ciprofloxacin resulted in a larger endotoxin and TNF-α release than seen with ceftazidime or imipenem. The discordance was in the number of the mice that survived in the treatment scenario (42-h delay). Ciprofloxacin, which gave the highest endotoxin and TNF-α levels, had the best survivorship (50%), while treatment with imipenem resulted in 30% survivors. However, it should be noted that two of the imipenem deaths occurred substantially after the end of therapy at day 5. Immediately after therapy was initiated, both ciprofloxacin and imipenem had 50% survivorships, while ceftazidime had a 10% survivorship. It is likely that there will be a counterbalance between the ability to kill more bacterial cells with higher doses of imipenem versus an increase in endotoxin released with subsequent production of other immune mediators at these higher doses. Again, the results with imipenem seen here in the treatment scenario differed from those seen in the earlier study reported by Byrne et al. (6) for other β-lactam antibiotics and provide hope that a different dose and schedule may result in improved survivorship. Part of the answer to the difference in the numbers of surviving mice after treatment with the different antibiotics may lie in the level of IFN-γ (and IL-12) expressed and the time of appearance of the inflammatory cytokine. Both TNF-α and IFN-γ were shown to be important for survival of mice challenged with Y. pestis (20, 21), although in the case of imipenem treatment where endotoxin release was suppressed, very little TNF-α and IFN-γ expression was seen. With ciprofloxacin treatment, a larger amount of endotoxin was seen (compared to that in mice treated with imipenem and ceftazidime), but we saw a large amount of the proinflammatory cytokine TNF-α after 48 h and at the same time a large amount of IFN-γ and a moderate amount of IL-12. On the other hand, in mice treated with ceftazidime, we saw a moderate amount of endotoxin, a moderate amount of TNF-α (compared to that in the control mice), and the appearance of a relatively large amount of IFN-γ similar to that seen with ciprofloxacin but later after infection. As discussed below, the levels of proinflammatory/inflammatory cytokines are important to control infection by the host of Y. pestis (22). The studies by Smiley's laboratory suggest that the presence of both TNF-α and IFN-γ is important for protection against Y. pestis (20, 21) .

We also saw the presence of IL-10 and IL-4, both anti-inflammatory cytokines (23), immediately after infection and even before antibiotic treatment was started. The presence of IL-10 may be important to attenuate the inflammatory response in the normal uninfected host (24). The level of IL-10 more than doubled after placebo was started in the control animals. The increase in the anti-inflammatory IL-10 may be responsible for the suppression of IFN-γ expression seen in the control animals (25). The consequences of this suppression of IFN-γ expression could be the inhibition of the establishment of a Th1 immune response needed to develop an acquired immune response against the pathogen (25). There is a balance between the levels of IL-10 and IFN-γ, where IFN-γ can also modulate the expression of the anti-inflammatory IL-10 (26, 27). The expression of IL-10 in the control animals could be triggered by activation of Toll-like receptor 2 (28) by the presence of modified lipopolysaccharide (LPS) residues on lipid A of the pathogen (22, 29, 30, 31). Changes in lipid A modification by Y. pestis are temperature dependent and appear to be a mechanism adapted by the pathogen to establish an infection in the host. IL-4, on the other hand, promotes the development of a Th2-type immune response and inhibits Th1-type development in the host (23). Our results suggest that these observations may also indicate that reduced viability of the infection and an alternate innate immune response to Y. pestis infection following treatment with imipenem may contribute to increased survival. The reduction in inflammatory cytokine production as a positive outcome may seem paradoxical; however, the altered LPS of Y. pestis has been shown to be associated with increased survival of the bacterium in vivo (22). Inflammation does take place during Y. pestis infection; however, it may be the timing and cell types involved that drive what is essentially a nonspecific anti-host-directed inflammation versus a more productive antibacterial response following treatment with imipenem. Studies to determine if this response includes other classes of bacterial compounds and other innate immune receptors should be conducted.

Imipenem-cilastatin may be an antibiotic that can be added to the physician's therapeutic armamentarium for prophylaxis of Y. pestis aerosol exposure. Further work needs to be done to determine whether higher fractions of the dosing interval covered with free drug imipenem will result in better survivorship in the treatment scenario.