Pharmacodynamics of Voriconazole against Wild-Type and Azole-Resistant Aspergillus flavus Isolates in a Nonneutropenic Murine Model of Disseminated Aspergillosis

ABSTRACT

Invasive aspergillosis (IA) due to Aspergillus flavus is associated with high mortality. Although voriconazole (VRC) is widely recommended as the first-line treatment for IA, emergence of azole resistance in Aspergillus spp. is translating to treatment failure. We evaluated the efficacy of voriconazole in a nonneutropenic murine model of disseminated A. flavus infection using two voriconazole-resistant isolates (one harboring the Y319H substitution in the cyp51C gene) and two wild-type isolates without mutations. All isolates exhibited a dose-response relationship, and voriconazole treatment improved mouse survival in a dose-dependent manner. At 40 mg/kg of body weight, 100% efficacy was observed for 1 susceptible isolate and 1 resistant isolate (with mutation), whereas for another susceptible isolate and resistant isolate (without mutation), survival rates were 81% and 72%, respectively. The Hill equation with a variable slope fitted the relationship between the area under the concentration-time curve (AUC)/MIC ratio and 14-day survival well for each strain. An F test showed the 50% effective doses to be significantly different from each other (P = 0.0023). However, contrary to expectation, there was a significant difference in exposure-response relationships between strains, and it appeared that the susceptible strains required a relatively higher exposure than the resistant ones to result in the same treatment effect, the 50% effective pharmacokinetic/pharmacodynamic (PK/PD) index (EI₅₀) required being negatively and log-linearly related to the MIC (P = 0.04). We conclude that the efficacy of voriconazole depended on drug exposure and the voriconazole MIC of the isolates, but lower exposures are required for strains with higher MICs. These findings may have profound significance in clinical practice with respect to dosing and drug choice.

INTRODUCTION

Invasive aspergillosis is an important opportunistic fungal infection, particularly in immunocompromised patients (1–3). After Aspergillus fumigatus, Aspergillus flavus is the second leading cause of invasive and noninvasive infections caused by Aspergillus spp. However, in certain tropical and subtropical countries, A. flavus has been reported as the predominant etiological agent causing mainly cerebral or sino-orbital aspergillosis or ocular infections (4–6). On the other hand, experimental data suggest that A. flavus is more virulent than A. fumigatus (7), except for pulmonary aspergillosis.

Of note, irrespective of the Aspergillus species involved, invasive aspergillosis causes significant mortality if not treated adequately with antifungals (8). The triazole voriconazole (VRC) is currently recommended as the first choice of treatment for disseminated infections caused by Aspergillus species (8, 9). However, emergence of acquired resistance to triazoles in A. fumigatus isolates is becoming a major clinical concern and has been associated with poor prognosis and treatment failures (10). Surveillance studies indicate that azole resistance is increasing in multiple European countries and in the Middle East, Asia, and Africa (11–15).

Although still uncommon, azole resistance has also been reported in A. flavus. Mutations in the cyp51A(T788G) and cyp51C(Y319H) (16) genes encoding the azole target enzyme have been associated with increased MICs of voriconazole. Given the prominent role of azoles in the management of aspergillosis due to A. flavus, it is important to determine if a mutation in the cyp51 gene in A. flavus corresponds with reduced in vivo efficacy. We therefore investigated the pharmacodynamics (PD) and dose-response and exposure-response relationships of voriconazole against wild-type (WT) and azole-resistant A. flavus isolates with and without mutation in the cyp51C gene in an immunocompetent murine model of disseminated aspergillosis.

(Parts of these results were presented at the 19th ISHAM [International Society for Human and Animal Mycology] Congress, 4 to 8 May 2015, Melbourne, Australia [poster no. 521].)

RESULTS

In vitro susceptibility.

The results of the in vitro susceptibility testing of the A. flavus isolates are shown in Table 1. Two isolates (NCCPF 761157 and 760815) showing voriconazole MICs of >1 mg/liter by the CLSI broth dilution technique were considered resistant or non-wild type based on the epidemiological cutoff values (ECVs) earlier described (17). The EUCAST MICs (MICs_EUCAST) of these isolates were ≥2 mg/liter. One isolate (NCCPF 761157) showed cross-resistance to itraconazole by both EUCAST (4 mg/liter) and CLSI (≥16 mg/liter), in contrast to the other voriconazole-resistant isolate (NCCPF 760815), which did not show any cross-resistance.

TABLE 1.

Disease classification, underlying azole resistance mechanisms, and in vitro antifungal susceptibilities of A. flavus isolates

Strain no.	Diagnosis	cyp51C substitution	MIC/MEC (mg/liter) ofa:
			Voriconazole		Itraconazole		Posaconazole		Caspofungin		Micafungin		Anidulafungin		Amphotericin B
			EUCAST	CLSI	EUCAST	CLSI	EUCAST	CLSI	EUCAST	CLSI	EUCAST	CLSI	EUCAST	CLSI	EUCAST	CLSI
NCCPF 761157	Chronic obstructive pulmonary disease	Y319H	4.0	4.0	4	≥16	0.12	0.25	1	0.03	1	0.015	1	0.007	8	2
NCCPF 760815	Granulomatous fungal rhinosinusitis	No	2.0	2.0	0.5	1	0.25	0.5	2	4	2	0.12	1	0.25	4	4
NCCPF 761100	Allergic fungal rhinosinusitis	Wild type	0.5	0.5	0.12	0.25	0.03	0.06	0.12	0.03	0.12	0.015	0.03	0.25	16	1
NCCPF 760690	Invasive fungal rhinosinusitis	Wild type	0.25	0.125	0.03	0.06	0.03	0.03	0.03	0.03	0.03	0.015	0.25	0.06	4	4

^aEUCAST, European Committee for Antimicrobial Susceptibility Testing; CLSI, Clinical and Laboratory Standards Institute; MIC, minimum inhibitory concentration; MEC, minimum effective concentration. The MIC/MEC values represent the mode value of the triplicate experiments.

Survival curves.

Figure 1 shows the survival curves of all four different isolates and different doses of voriconazole. Control groups of each of the four isolates injected intraperitoneally with normal saline showed 90% mortality and a median survival time of 7 to 9 days. Survival at day 14 was significantly better for the doses of 10 mg/kg and above in each of the four isolates tested compared to that of the controls. At the dose of 40 mg/kg, 100% efficacy was observed for the isolates with the VRC MIC of 0.25 mg/liter, with 72% efficacy for the voriconazole-resistant isolate (VRC MIC_EUCAST of 2 mg/liter) without harboring the Y319H substitution in the cyp51C gene.

FIG 1.

Efficacy of voriconazole against voriconazole-resistant (MICs of 2 and 4 mg/liter) and voriconazole-susceptible (MICs of 0.25 and 0.5 mg/liter) A. flavus isolates. Control group received sterile saline. For all groups, n = 11.

Pharmacokinetics of voriconazole.

The plasma voriconazole concentration versus time profile is shown in Fig. 2, and the pharmacokinetic (PK) parameters are provided in Table 2. The dose-normalized area under the concentration-time curve (AUC) increased (AUC_{∞_D_pred}), and clearance (CL) decreased with increasing dosages, confirming the nonlinear PK of voriconazole (18). There was a difference by a factor of more than 5 per dose unit between the 5- and 40-mg/kg doses.

FIG 2.

Plasma concentration of voriconazole following intraperitoneal administration of 5, 10, 20, 40 mg/kg to immunocompetent mice infected with A. flavus. Each symbol corresponds to the geometric mean and standard error of the mean plasma levels of three mice.

TABLE 2.

Pharmacokinetic parameters of voriconazole following intraperitoneal administration of dosages ranging from 5 to 40 mg/kg

Dose group (mg/kg)	Cmax (mg/liter)	Clast (mg/liter)	t1/2 (h)	AUCa
				AUC∞_pred (h · mg/liter)	AUC∞_D_pred (h · mg)/(liters · kg)
5	2.50	1.05	0.85	2.44	0.49
10	4.65	1.25	1.33	7.56	0.76
20	14.20	3.15	3.68	36.63	1.83
40	22.55	3.70	5.24	103.30	2.58

^aAUC_∞__pred, AUC extrapolated to infinity using the last predicted concentration; AUC_{∞_D_pred}, dose-normalized AUC_{∞_pred}.

Dose- and exposure-response analysis.

The dose-response curves for the dosing regimens and control groups of voriconazole are shown in Fig. 3. Treatment improved the survival of mice in a dose-dependent manner. Higher doses were required for mice infected with resistant strains to result in in the same effect compared to those infected with the susceptible isolates, as can be observed from a right shift of the curve fits for the more resistant strains. The 50% effective doses (ED₅₀s) based on survival were 13.21 mg/kg voriconazole (95% confidence interval [CI], 10.55 to 16.53 mg/kg) for the WT strain with a MIC of 0.25 mg/liter, 14.57 (95% CI, 9.70 to 21.86 mg/kg) for the WT strain with a MIC of 0.5 mg/liter, 25.82 (95% CI, 17.76 to 37.52 mg/kg) for the non-WT strain with a MIC of 4 mg/liter, and 23.54 (95% CI, 18.36 to 32.20 mg/kg) for the non-WT strain with a MIC of 2 mg/liter. An F test showed the ED₅₀s to be significantly different from each other (P = 0.0023).

FIG 3.

Voriconazole dose-survival relationship for voriconazole-resistant and voriconazole-susceptible A. flavus isolates. The curve indicates a fit with the Hill equation for each isolate.

The relationship between AUC/MIC ratio and survival was determined for each isolate by using the AUCs from Table 2. The Hill equation with a variable slope showed that the relationship between the AUC/MIC ratio and 14-day survival fitted well for each strain (Fig. 4). However, contrary to expectation, there was a significant difference in exposure-response relationships, and it appeared that the more susceptible strains required a lower exposure than the resistant ones to result in the same treatment effect. A further analysis indicated that there was a log-linear relationship between MIC and exposure, as shown in Fig. 5, showing the 50% effective PK/PD index (EI₅₀) of the exposure-response curves as a function of the MIC. Mice infected with strains with a higher MIC require less exposure than those infected with strains with a lower MIC to result in the same effect.

FIG 4.

Percentage of survival as a function of the voriconazole AUC/MIC ratio against A. flavus voriconazole-resistant and -susceptible isolates. The curve is the model fit with the Hill equation for each strain. MIC values are in milligrams per liter determined as per the EUCAST methodology.

FIG 5.

EI₅₀ of the AUC/MIC ratio of voriconazole in relation to the MICs of different voriconazole-resistant and -susceptible strains.

DISCUSSION

The present study describes the efficacy of voriconazole against voriconazole-resistant and wild-type A. flavus isolates in an immunocompetent murine model of disseminated aspergillosis. The results show that treatment with voriconazole is effective against the voriconazole-resistant as well as voriconazole-susceptible isolates. Increasing doses increased the survival of the mice in a dose-dependent manner; a maximum response was achieved at the higher doses irrespective of the susceptibility of the isolates. However, overall the AUC and AUC/MIC ratio showed a better exposure-survival relationship. This improved relationship with AUC/MIC can be readily explained by the nonlinear clearance of the voriconazole (18, 19), as was observed for A. fumigatus (18, 20). The pharmacokinetics was not determined after one dose only. In the present study, the PK at day 2 was considered similar to that at day 7 on the basis of our earlier study (20). In our earlier study, the survival-AUC/MIC relationship and EI₅₀ were similar up to day 7, indicating that induced autometabolism, which is a known issue with mice, is likely to not significantly affect the PK at day 2 or 7 (20).

A surprising finding was that 100% survival could be achieved with the highest dose (40 mg/kg) for both a susceptible isolate as well as a resistant isolate—one wild-type isolate (NCCPF 760690) and another non-wild-type isolate (NCCPF 761157) carrying the Y319H substitution in the cyp51C gene—whereas in the non-wild-type isolate not carrying the Y319H substitution in cyp51C, the survival rate at this dose was only 72%. This result is contrary to those seen in A. fumigatus in which a higher dose was required for the mice infected with the TR₃₄/L98H mutation compared to the wild-type isolate (18, 20) and the AUC/MIC response relationship could be fully explained by exposure and the MIC independent of the mutation.

In the present study, the interdependence between MIC, mutation, and overall effect is further demonstrated by the significant relationship between the EI₅₀s of the AUC/MIC and MIC, in that lower exposures were required for strains with higher MICs to result in the same effect. This indicates that although MIC values represent the activity of the drug in vitro, the underlying mutations in the target gene may significantly impact the dose or exposure required for the treatment of infection. It appears that there is a cost to the microorganism to become less susceptible. Similar observations were also made for some antibacterials (21). For instance, in a study looking at the efficacy of ceftazidime against a number of resistant strains, it was observed that several resistant strains required far less a percentage of time (24 h) that the percentage of free, unbound fraction was greater than the MIC (%fT_MIC) than expected (21). It is difficult to explain this phenomenon in this study. The fitness of the strains was determined by in vitro growth kinetics to identify the differences in growth rates, as one of the indicators of viability. There was no significant difference in the growth kinetics of the wild type and the mutant (data not shown). Also, we recently investigated the impact of the Y319H substitution in cyp51A (which was observed in one non-wild-type isolate, NCCPF 761157) by performing homology modeling and molecular dynamic simulation studies. This model indicated that the Y319H substitution was located far from the iron porphyrin complex and hence did not have a direct effect on the docking of the azoles at the binding site. Instead this mutation exhibited increased conformational flexibility, probably due to polar nature of histidine which causes interatomic repulsion leading to reduced drug binding affinity (16). A third explanation could be a difference in levels of virulence between susceptible and resistant strains. However, we did not find any evidence for that as there was no major difference in the 90% lethal doses (LD₉₀s) for the strains tested (data not shown). In any case, the results found need to be confirmed by more isolates.

There are some limitations of this study. As to the study design, one might consider the infection was induced via the intravascular route and that a nonneutropenic model was used. However, similar studies with posaconazole and voriconazole in both neutropenic and nonneutropenic models of A. fumigatus have shown that the exposure–response relationships of azoles are of the same order of magnitude: in fact, slightly lower exposures were required in the neutropenic model (10). One of the major reasons here is that the inocula in each model are chosen to allow no or very low survival rates if treatment is not given and near 100% survival for the maximal effect. Furthermore, infections with A. flavus often occur in immunocompetent patients.

In the absence of a clinical MIC value to categorize an A. flavus isolate as resistant or susceptible, wild-type MIC distribution and epidemiological cutoff (ECOFF) values are usually used to separate the non-wild-type population. The CLSI and EUCAST ECOFF values defined for voriconazole are 1 mg/liter and 2 mg/liter, respectively (22, 23). In the present study, we therefore considered these values to define the isolate as wild type or non-wild type. However, even this does not fully explain the obtained results, as we observed that the exposure-response curves of the two WT strains were significantly different.

According to the Infectious Diseases Society of America (IDSA) guidelines, voriconazole is the recommended primary therapy for invasive aspergillosis (8). In the present study, we observed that the AUC level achieved for the highest dose (40 mg/kg) was 103.30 h · mg/liter. Given the voriconazole protein-free fraction of 29.87% in mice (20), therefore, the free drug AUC value of 28.91 resulted in therapeutic success against both susceptible and resistant A. flavus isolates with the MICs_EUCAST ranging from 0.25 to 4 mg/liter. Considering the kinetics of voriconazole in humans, the standard dose on the basis of 200 mg oral voriconazole or an intravenous dose of 4 mg/kg every 12 h would produce a free drug AUC value of approximately 25 mg · h/liter (24). Therefore, one could consider that standard dosing of voriconazole could be used for the successful treatment of infections caused by A. flavus for which MICs are as high as 2 to 4 mg/liter.

Alarmingly, azole resistance is now reported in cases of aspergillosis due to A. flavus (16). In countries like India, where A. flavus is the primary fungus causing cerebral or sino-orbital aspergillosis, the situation is far more crucial (4, 6). The first case of invasive aspergillosis refractory to voriconazole treatment was reported from China, from a patient with acute myeloid leukemia who had undergone an allogeneic stem cell transplant (25). The A. flavus isolate had a T788G missense mutation in the cyp51C gene, which was an undocumented mutation (25). Previously, a study spanning 9 years from 2001 until 2009 used epidemiological cutoff values (ECVs) to explore the susceptibility of Aspergillus to azoles. They reported 4 resistant A. flavus isolates from a total of 235 strains tested (1.7% of strains had MICs greater than the ECV) (17), whereas in a Spanish cohort, 58 strains of A. flavus were tested, with 15 strains and 4 strains exhibiting intermediate susceptibility and resistance to voriconazole, respectively (26). Unfortunately, patient outcome data are not available for these studies, and the impact of voriconazole resistance could not be assessed. Earlier from India, we reported around 5% of A. flavus isolates had high voriconazole MICs (27). Another recent study from India reported two voriconazole-resistant isolates out of a total of 8 A. flavus isolates tested (28). Recently a single-center study of 120 A. flavus isolates from India reported that MICs of three isolates were above the ECOFF of voriconazole. Interestingly, they also identified novel substitutions in cyp51C that could possibly play a role in the azole resistance (29). All of these data suggest that voriconazole resistance in A. flavus is emerging in clinical settings.

The model indicates that the efficacy of voriconazole against A. flavus depends on the drug exposure and the MIC of the isolate. Contrary to the general belief, this study showed that there was a significant difference in the exposure-response relationship. The more susceptible isolates required more drug exposure than the resistant isolates to have the same treatment effect.

MATERIALS AND METHODS

Fungal isolates.

Four clinical A. flavus isolates obtained from patients with different manifestations of aspergillosis with EUCAST MICs (MICs_EUCAST) ranging from 0.25 to 4 mg/liter, were used in the experiments (Table 1). Two wild-type isolates without mutations in the cyp51C gene (NCCPF 760690, VRC MIC_EUCAST of 0.25 mg/liter; NCCPF 761100, VRC MIC_EUCAST of 0.5 mg/liter) and two voriconazole-resistant isolates (NCCPF 761157, VRC MIC_EUCAST of 4 mg/liter, harboring a Y319H substitution in the cyp51C gene; NCPPF 760815, VRC MIC_EUCAST of 2 mg/liter, without mutations in the cyp51C gene).

Strain identifications and the cyp51C gene substitutions were confirmed by sequence-based analysis as described previously (11). The isolates were stored in 10% glycerol broth at −80°C and were cultured on Sabouraud's dextrose agar (SDA) supplemented with 0.02% chloramphenicol for 48 to 72 days at 30°C. All isolates were cultured again on SDA for 5 to 7 days at 35 to 37°C before preparation of the inoculum.

In vitro antifungal susceptibility testing.

In vitro antifungal susceptibility testing (MICs [13] and minimum effective concentrations [MECs]) was performed by the EUCAST (The European Committee on Antimicrobial Susceptibility Testing) method as per protocol EUCAST E.DEF 9.2 for voriconazole (Pfizer Central Research, Groton, CT), posaconazole (Merck Sharp and Dohme Pharmaceuticals, India), itraconazole (Sigma-Aldrich, Bengaluru, India), caspofungin (Merck Sharp and Dohme Pharmaceuticals, India), anidulafungin (Pfizer Central Research), micafungin (Astellas, Pharma US, Inc., Deerfield, IL), and amphotericin B (Sigma-Aldrich, Bengaluru, India). A. flavus ATCC 204304 was used as a quality control strain for antifungal susceptibility testing. Antifungal susceptibility testing was also performed by the CLSI (Clinical and Laboratory Standards Institute) broth microdilution technique based on the M38A2 document (27, 30, 31). Candida krusei ATCC 6258 and A. flavus ATCC 204304 were used as quality control strains.

Infection model.

The efficacy of voriconazole monotherapy was determined in an immunocompetent mouse model of disseminated aspergillosis following intravenous inoculation, as described previously (18, 32–34). Animals were infected via injection into the lateral tail of the mouse of 0.1 ml of the conidial suspension corresponding to the LD₉₀ for each isolate. A total of 220 outbred Swiss albino female mice (Advanced Small Animal Research Facility, PGIMER, Chandigarh, India), 4 to 5 weeks old and weighing 20 to 25 g, were randomized into groups of 11 mice as control or treatment groups.

Before the experiment was performed, the isolates were cultured once on Sabouraud's dextrose agar (SDA) for 5 days at 35 to 37°C. The conidia were then harvested in 20 ml of sterile phosphate-buffered saline (PBS) plus 0.1% Tween 80 (Sigma-Aldrich, Bengaluru, India). The conidial suspension was filtered through a sterile gauze folded four times to remove any hyphae, and the number of conidia was counted in a hemocytometer. After the inoculum was adjusted to the required concentration, the conidial suspension was stored overnight at 4°C.

The 90% lethal dose (LD₉₀) was determined for each isolate separately. The LD₉₀s were 8 × 10⁵ (wild-type control, NCCPF 760690), 1 × 10⁶ (wild-type control, NCCPF 761100), 6 × 10⁵ (resistant isolate, NCCPF 760815), and 8 × 10⁵ (resistant isolate 761157, harboring the Y319H mutation) conidia, respectively. Postinfection viability counts of the injected inoculum were determined to ensure that the correct inoculum had been injected.

Mice were infected with the A. flavus isolate through the lateral tail vein, and after 24 h, voriconazole (Pfizer, Mumbai, India) was administered in doses of 5, 10, 20, and 40 mg/kg intraperitoneally once daily for 7 consecutive days. The control group received single doses of saline. The animals were housed under standard conditions, with drink and feed supplied ad libitum. All animal procedures were approved by the Animal Ethics Committee of PGIMER, Chandigarh (IAEC 324).

Following infection, the monitoring of survival was performed by experienced individuals blind to the animal treatment. The infected mice were examined at least three times daily. Clinical inspections focused on dehydration, torticollis, staggering, high weight loss (a decrease of 15% within 48 h or 20% within 24 h), or drop in body temperature to below 33°C. Mice demonstrating these clinical signs were euthanized. On day 15 postinfection, all surviving mice were euthanized under isoflurane anesthesia, and blood and internal organs were collected.

Pharmacokinetic analysis of voriconazole in mice.

A total of 90 outbred Swiss albino (Advanced Small Animal Research Facility, PGIMER, Chandigarh, India) female mice, 4 to 5 weeks old and weighing 20 to 25 g, were used for the PK experiments. On day zero, mice were infected with the wild-type and non-wild-type A. flavus isolates through the lateral tail vein, and after 24 h, voriconazole monotherapy was initiated at dosages of 5, 10, 20, and 40 mg/kg of body weight intraperitoneally. At day 2 of treatment (day 3 after infection), blood samples were drawn from 3 mice for each predefined sampling time point (6 time points in total); immediately before administration of drugs and subsequently at 0.5, 1, 2, 4, and 8 h postdose. The blood samples were drawn through the orbital vein or heart puncture into lithium-heparin-containing tubes and were cooled and then centrifuged for approximately 10 min at 1,000 × g within 30 min of collection. Plasma was aspirated, transferred in two 2-ml plastic tubes, and stored at −80°C.

Analytical assay of voriconazole.

Voriconazole concentrations in plasma samples were measured by high-performance liquid chromatography (HPLC) with UV detection (Hitachi D-2000 system; Hitachi High-Technologies, Tokyo, Japan) using a C₁₈ column (Hitachi analytical steel column; 5 μm, 4.6 by 250 mm); the mobile phase was achieved with absolute acetonitrile (HPLC grade; Sigma, Bengaluru, India) and (NH₄)H₂PO₄ (0.04 M [pH 7.4]) in the ratio 50:50. The assay parameters were set as an injection volume of 100 μl, flow rate of 0.8 ml/min, run time of 10 min, and temperature of 25°C. A different concentration of voriconazole was used as the internal standard, and the UV detector was set at 255 nm. The lower limit of detection was 0.05 mg/liter, and accuracy ranged between 97% and 103%. Free fractions of voriconazole in plasma samples were determined by pooling three samples at identical time points for all four dosing regimens to obtain at least around 400 μl of voriconazole-containing plasma. The plasma was transferred to 1.5-ml microcentrifuge tubes (Axygen, Inc., Corning Brand, Union City, CA). The samples were prepared adding 250 μl of plasma to 500 μl of acetonitrile and incubated for 15 min at 4°C followed by centrifugation at 12,000 rpm for 10 min at 4°C. The filtrate was injected directly onto the HPLC system.

Pharmacokinetic analysis.

Geometric mean concentrations of voriconazole were calculated for each time point (n = 3 mice). Pharmacokinetic parameters (area under the concentration-time curve from 0 to 8 h [AUC_0–8], maximum concentration of drug in serum (C_max), concentration at 24 h (C₂₄), half-life (t_1/2), volume of distribution (V), clearance (CL), and terminal elimination rate constant [k_el]) were calculated using noncompartmental analysis (Phoenix version 6.3). The AUC from 0 to infinity (AUC_0–∞) was calculated using the linear up-log down trapezoidal rule. In addition, C_max and C₂₄ were directly observed from the data. Half-life was calculated by ln 2/k_el, in which k_el was determined by linear regression of the terminal points of the log-linear plasma concentration time curve. V was calculated using the formula V = dose/AUC × k_el, and CL was calculated as dose/AUC_0–8.

Statistical analysis.

All data analyses were performed using GraphPad Prism version 6.05 for Windows (GraphPad Software, San Diego, CA). Mortality data were analyzed by the log rank test. The survival data were plotted against dose/MIC and AUC/MIC. The Hill equation with a variable slope was fitted to the data. The fits were performed for survival data of each strain. The goodness of fit was checked by the R2 and visual inspection. Dose/MIC and AUC/MIC ratios were calculated by dividing the dose (in milligrams per kilogram) or AUC by the MIC. Dose/MIC and AUC/MIC ratio data were log₁₀ transformed to approximate a normal distribution prior to statistical analysis. The 50% effective PK/PD index (EI₅₀) of the PK/PD index best correlating with efficacy was determined. For comparison between strains, the F test was performed to define whether EI₅₀ differed between the four groups. Statistical significance was defined as a P value of <0.05 (two-tailed).