Pharmacokinetic-Pharmacodynamic Characterization of Omadacycline against Haemophilus influenzae Using a One-Compartment In Vitro Infection Model

Omadacycline is a novel aminomethylcycline with activity against Gram-positive and -negative organisms, including Haemophilus influenzae, which is one of the leading causes of community-acquired bacterial pneumonia (CABP). The evaluation of antimicrobial agents against H. influenzae using standard murine infection models is challenging due to the low pathogenicity of this species in mice. Therefore, 24-h dose-ranging studies using a one-compartment in vitro infection model were undertaken with the goal of characterizing the magnitude of the ratio of the area under the concentration-time curve (AUC) to the MIC (AUC/MIC ratio) associated with efficacy for a panel of five H. influenzae isolates.

with CABP, including H. influenzae (8). Intravenous (i.v.) and oral formulations of omadacycline were recently approved by the U.S. Food and Drug Administration for the treatment of patients with CABP (12). As part of the omadacycline drug development program, in vivo studies were carried out to characterize the pharmacokineticspharmacodynamics (PK-PD) of omadacycline against Staphylococcus aureus and S. pneumoniae (13,14). However, PK-PD evaluations for H. influenzae using in vivo infection models are challenging. This is due primarily to the fact that H. influenzae does not grow well and auto-clears in murine infection models (15,16), even the neutropenic thigh infection model, the workhorse infection model that has been used to characterize the PK-PD of essentially all classes of antimicrobial agents to date (17). In such cases, in vitro infection models have allowed for the characterization of the PK-PD of a given antimicrobial agent against H. influenzae (18)(19)(20)(21).
Herein, we describe the results of a series of in vitro studies evaluating the activity of omadacycline against H. influenzae. The objectives of these studies were 2-fold. The first objective was to evaluate the frequency of mutation of clinical H. influenzae isolates to omadacycline. As the ratio of the area under the concentration-time curve from 0 to 24 h to the MIC (AUC/MIC ratio) is generally considered to be the PK-PD index most closely associated with efficacy for tetracyclines (22)(23)(24)(25), the second objective was to determine the magnitude of the omadacycline AUC/MIC ratio associated with efficacy for a panel of H. influenzae isolates evaluated using a one-compartment in vitro infection model. Given the importance of considering effect site drug exposures (26)(27)(28), which for patients with CABP is epithelial lining fluid (ELF), the studies described herein were carried out in the context of omadacycline total-drug ELF concentration-time profiles to characterize AUC/MIC ratio targets associated with efficacy.

RESULTS
In vitro susceptibility studies. The omadacycline MIC values for the panel of five H. influenzae isolates are presented in Table 1. Omadacycline MIC values ranged from 1 to 2 mg/liter and from 0.5 to 1 mg/liter for the microbroth and agar dilution methods, respectively.
Mutation frequency studies. The inocula evaluated in the frequency of resistance studies failed to produce or identify a drug-resistant subpopulation for all isolates evaluated, as presented in Table 2. The lack of colonies found to grow on the antibiotic-supplemented agar plate suggests that the frequency of mutations resulting in omadacycline resistance was less than that of the inocula (2.9 ϫ 10 9 CFU/ml) examined.
One-compartment in vitro infection model. The targeted concentration-time profiles representing the beta and terminal half-lives of omadacycline in human epithelial lining fluid (ELF) were well simulated in the one-compartment in vitro infection model, as evidenced by the agreement between observed and targeted concentrations shown for all data evaluated in Fig. 1A and by the 100-mg every 12 h (q12h) example dosing regimen in Fig. 1B. The assessments of the mass spectrometer assay performance demonstrated that the interassay percent coefficient of variation (%CV) for the quality control samples at concentrations of 0.300, 1.05, and 8.40 g/ml were 28.3, 9.41, and 9.09%, respectively.
Pharmacokinetic (PK) models were fit to the samples collected for the evaluation of the omadacycline concentration-time profiles. The final PK model, which was a onecompartment model utilizing two different clearance terms over the 24-h study duration, fit the data with excellent precision, as evidenced by a coefficient of determination (r 2 ) value of 0.964 and a slope of 0.970 (which is close to 1) for the observed versus model-predicted concentrations (see Fig. S1 in the supplemental material). No bias was evident when goodness-of-fit plots were examined (data not shown). Using the final PK model, omadacycline total-drug ELF AUC values were computed through numerical integration for each dosing regimen studied in the one-compartment in vitro infection models, which are described in Table 3.
Across the omadacycline dosing regimens evaluated in the one-compartment in vitro infection model dose-ranging studies for the five H. influenzae isolates, an exposure-response relationship was observed for each isolate, as demonstrated by the  time course data shown in Fig. 2. As evidenced by the data for each of the five isolates, the growth control grew well, reaching bacterial densities greater than 1.0 ϫ 10 8 log 10 CFU/ml by the 24-h time point, representing a 2-log 10 increase in bacterial burden. A range of responses was achieved over the total-drug ELF AUC range for each isolate, with low exposures resulting in treatment failure (i.e., as evidenced by the growth matching that of the control by 24 h), intermediate exposures reaching only net bacterial stasis at 24 h, and higher exposures achieving reductions in the bacterial burden over the 24-h period.
The relationships between the change in the log 10 CFU/ml from baseline at 24 h and omadacycline total-drug ELF AUC/MIC ratio were evaluated based on the data for each individual H. influenzae isolate as well as a pooled collection. As evidenced by a high r 2 of 0.88 and the dispersion of data about the fitted function, the relationship between the change in the log 10 CFU/ml from baseline at 24 h and total-drug ELF AUC/MIC ratio FIG 2 Results of dose-ranging studies carried out using the one-compartment in vitro infection model for the panel of five H. influenzae isolates exposed to omadacycline total-drug ELF AUC values ranging from 2.38 to 82.7 mg · h/liter, which were representative of dosing regimens of 12.5 to 400 mg q12h. shown in Fig. 3 described the activity of omadacycline well across the panel of five H. influenzae isolates with MIC values ranging from 1 to 2 mg/liter. As shown in Table 4, the parameter estimates (standard errors) for the Hill-type model describing this relationship were as follows: the change in the log 10 CFU/ml from baseline at 24 h in the absence of drug (E 0 ) was 2.80 (0.34), the maximum change in the log 10 CFU/ml from E 0 (E max ) was 8.81 (0.75), the Hill coefficient was 1.66 (0.29), and the omadacycline total-drug ELF AUC/MIC ratio associated with half-maximal effect (50% effective concentration [EC 50 ]) was 9.29 (1.13). The magnitude of the total-drug ELF AUC/MIC ratio associated with net bacterial stasis and 1-and 2-log 10 CFU/ml reductions from baseline at 24 h based on the pooled data were 5.87, 7.87, and 10.4, respectively.
Hill-type models describing the relationship between the change in the log 10 CFU/ml from baseline at 24 h and the total-drug ELF AUC/MIC ratio for each H. influenzae isolate also described the data well (r 2 values of 0.89 to 0.94). Parameter estimates (standard errors) for these Hill-type models are shown in Table 4. The median

DISCUSSION
The studies described herein were carried out to accomplish two objectives. The first was to evaluate the frequency of mutation of clinical H. influenzae isolates to omadacycline. The second objective was to determine the magnitude of the omadacycline AUC/MIC ratio associated with efficacy for the above-described panel of H. influenzae isolates evaluated using a one-compartment in vitro infection model.
The frequency of mutation to omadacycline for the H. influenzae isolates evaluated could not be identified for bacterial populations less than or equal to 2.9 ϫ 10 9 CFU/ml at omadacycline concentrations representing 3 or 5 times the baseline MIC. These values are similar to those identified by Clark et al. (29) and Min et al. (30) using a variety of agents against H. influenzae. Mutation frequency values like those described herein are favorable compared to those of drug-organism combinations such as fosfomycin and Escherichia coli, the mutation frequency for which is approximately 1 ϫ 10 5 to 1 ϫ 10 6 CFU/ml (31). For drug-organism combinations with mutation frequencies in this range, the probability of on-therapy emergence of resistance is expected to be higher than those with lower mutation frequencies. While data from these 24-h studies assessing mutation frequency are useful, studies using a hollow-fiber in vitro infection model, a system that allows for the evaluation of the resistance suppression over a longer study period, are beneficial for characterizing the potential for development of on-therapy resistance and the identification of exposure that prevents resistance amplification.
Results obtained from the dose-ranging studies demonstrated that the data for the five isolates comodeled well. As evidenced by a high r 2 of 0.88 and the dispersion of data about the fitted function, the relationship between the change in the log 10 CFU/ml from baseline at 24 h and total-drug ELF AUC/MIC ratio described the efficacy of omadacycline against the panel of H. influenzae isolates well. Hill models describing the relationship for individual isolates also described these data well (r 2 values of 0.89 to 0.94). The median omadacycline total-drug ELF AUC/MIC ratio associated with net bacterial stasis and the 1-and 2-log 10 CFU/ml reductions from baseline at 24 h for the H. influenzae isolates studied were 6.91, 8.91, and 11.1, respectively. As described above, PK-PD targets for H. influenzae were characterized using an in vitro infection model, given the difficulty of achieving sufficient growth and potential for auto-clearance of the pathogen in in vivo infection models. Thus, while information derived from in vitro infection models represents an optimal approach to characterize the PK-PD targets for H. influenzae efficacy, such targets may be overestimated relative to what is expected using an in vivo infection model as a result of the enhanced effect of growth media on bacterial growth and the lack of an immune system.
The evaluation of effect site exposures, which for CABP is total-drug ELF, and thus the characterization of total-drug ELF AUC/MIC ratio targets represented an important strength of the studies described herein. Mimicking the shape of total-drug ELF, which involved applying two different half-live estimates observed in healthy volunteers (32), was easier to accomplish in an in vitro model than would have been possible using a standard murine infection model.
In the current era of drug development for antimicrobial agents, PK-PD assessments are carried out during all stages of drug development (27,(33)(34)(35)(36)(37). Evaluation of PK-PD targets for efficacy from nonclinical infection models, together with phase 1 PK data and Monte Carlo simulation, prior to the execution of clinical trials with patients ensures that efficacious dosing regimens are selected for studying infected patients and serves to make the drug development program less prone to risk. In late-stage development, the use of such data provides the basis for refining dosing regimens, especially in special populations, and supports decisions for interpretive criteria for in vitro susceptibility testing for the antimicrobial agent against a given pathogen (27).
Given the empirical nature of the treatment of patients with CABP, it is important that the approved omadacycline dosing regimen for CABP be optimized from a PK-PD perspective for key pathogens, such as H. influenzae. The data described herein, which allowed for the characterization of the PK-PD of omadacycline against H. influenzae, were used to support both omadacycline dose selection for patients with CABP and evaluations for interpretive criteria for in vitro susceptibility testing for omadacycline against H. influenzae (38).

MATERIALS AND METHODS
Bacteria and study drug. A panel of five H. influenzae isolates was utilized in the studies described herein and was obtained from the American Type Culture Collection (ATCC) and JMI Laboratories (North Liberty, IA). The challenge panel included four clinical isolates and one ATCC reference strain (H. influenzae ATCC 49247). Omadacycline was provided by Paratek Pharmaceuticals, Inc. (King of Prussia, PA).
Media and in vitro susceptibility studies. Omadacycline MIC values were determined for the challenge isolate panel according to Clinical and Laboratory Standards Institute guidelines (39) using agar and broth microdilution methodologies. These different assays were used as described below. All MIC values were determined using freshly prepared Haemophilus test medium (HTM), in triplicate, over a 2-day period and are reported as the modal values.
Frequency of resistance studies. The frequency of omadacycline resistance was estimated for each challenge isolate. In brief, 2 ml of log-phase growth suspension, consisting of approximately 1.0 ϫ 10 9 CFU/ml, was plated using 200-l volumes on 10-by 100-mm HTM agar plates supplemented with three times or five times the omadacycline baseline agar dilution MIC. The bacterial concentration in the suspension was determined by quantitative culture using drug-free agar. The inoculated agar plates were incubated at 35°C and then observed for growth, with individual colonies counted 48 h postincubation. The frequency of resistance was determined as the ratio of the number of colonies found growing on the drug-supplemented agar plates to the total number of colonies plated. If no isolates were observed after 48 h of incubation, then the resultant mutation frequency was determined to be less than that of the inoculum evaluated.
One-compartment in vitro infection model. A one-compartment in vitro infection model, as previously described (40,41), was used for the studies described herein. In brief, the in vitro model consisted of a central infection compartment, which contained growth medium (HTM), the challenge organism, and magnetized stir bars to ensure that the test compound, media, and organism remained in a homogenous state. The central infection compartment was placed on a magnetic stir plate, which was then housed in a temperature-controlled incubator set to 35°C. The bacterial suspension in the central compartment was then exposed to dynamic concentration-time profiles of omadacycline designed to simulate human ELF concentration-time profiles in healthy volunteers following i.v. administration (32). A series of computer-controlled peristaltic pumps infused drug-free media into the central compartment, while simultaneously removing media into a waste container. The challenge isolate was then inoculated directly into the central infection compartment, and the test compound was infused by a computer-controlled syringe pump in order to simulate selected beta-and terminal half-lives, dosing frequencies, and durations of infusion, as described below. The diffusion rates simulated by the peristaltic pumps were such that the desired concentration-time profile of omadacycline imitated human totaldrug ELF concentration-time profiles. Samples were aseptically collected from the central compartment for determination of bacterial density and drug concentration at the predetermined time points described below.
The bacterial suspension of each challenge isolate was prepared from a culture grown overnight on chocolate II agar plates, purchased from BD Laboratories (Franklin Lakes, NJ). Isolates were collected from the overnight cultures, suspended in HTM broth, and grown to mid-logarithmic growth phase in an Erlenmeyer flask immersed in a shaking water bath set to 35°C and 125 rotations per minute. The optical density of the bacterial suspension growing in the flask was compared to those of previously completed growth curves using a spectrophotometer set to a wavelength of 630 nm. Isolates were then inoculated in the system to achieve an initial bacterial burden of 1.0 ϫ 10 6 CFU/ml. The bacterial suspension in the central compartment was then exposed to changing omadacycline concentrations, simulating human beta and terminal half-lives of 1 and 14.9 h in total-drug ELF (32). Omadacycline dosing regimens of 12.5, 25, 50, 100, 150, 200, 300, and 400 mg administered every 12 h were linearly scaled based on ELF concentrations observed following a 100-mg i.v. dose. All omadacycline treatment regimens were compared to a no-treatment control.
In order to evaluate the effect that omadacycline had on the bacterial suspension in the central compartment, a series of 1-ml samples were collected for the determination of bacterial density at 0, 2, 4, 8, 12, and 24 h. In order to eliminate the potential for drug carryover, each bacterial sample was centrifuged, aspirated, and resuspended to the initial sample volume with sterile normal saline in duplicate to eliminate drug carryover. The washed sample was diluted serially by 10-fold volumes in sterile normal saline, cultured on chocolate agar, and placed in a humidified incubator for 24 h. In order to measure drug concentrations in the in vitro system, a second 1-ml sample aliquot was collected at 1, 2, 3, 5, 7, 11, 13, and 23 h postinoculation. All samples for evaluation of omadacycline concentration were immediately frozen at -80°C until they were assayed via liquid chromatography-tandem mass spectrometry (LC/MS/MS).

Drug assay.
Calibration standard and quality control samples were prepared in HTM and processed concurrently with collected samples. All samples were assayed via LC/MS/MS on a Sciex 5500 mass spectrometer using a Thermo Hypercil Gold C8 column (3 by 100 mm) with a mobile phase of 10 mM ammonium formate (pH 3.0) and acetonitrile. The standard curve for omadacycline was linear, ranging from 0.10 to 10.0 mg/liter, with a correlation coefficient (r 2 ) value ranging from 0.9926 to 0.9992. The lower limit of quantification was 0.100 mg/liter. PK-PD analyses. PK models were fit to the samples collected for the evaluation of the total-drug ELF concentration-time profiles. Omadacycline total-drug ELF AUC values were calculated for each dosing regimen simulated in the one-compartment in vitro infection model. Given the assumption that protein binding in ELF is negligible, total-drug ELF AUC values were evaluated and not further adjusted to account for protein binding. Data from the omadacycline dose-ranging studies were evaluated using Hill-type models and nonlinear least-squares regression. All data were weighted using the inverse of the estimated measurement variance. The relationship between the change in log 10 CFU/ml from baseline at 24 h and total-drug ELF AUC/MIC ratio was evaluated by utilizing the broth dilution MIC values. The magnitude of the total-drug AUC/MIC ratio associated with net bacterial stasis and 1-and 2-log 10 CFU/ml reductions from baseline at 24 h was determined for each H. influenzae isolate individually and based on the pooled data for the isolates studied.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.2 MB.