ABSTRACT
Tebipenem pivoxil HBr (TBPM-PI-HBr) is a novel orally bioavailable carbapenem. The active moiety is tebipenem. Tebipenem pivoxil is licensed for use in Japan in children with ear, nose, and throat infections and respiratory infections. The HBr salt was designed to improve drug substance and drug product properties, including stability. TBPM-PI-HBr is now being developed as an agent for the treatment of complicated urinary tract infections (cUTI) in adults. The pharmacokinetics-pharmacodynamics of tebipenem were studied in a well-characterized neutropenic murine thigh infection model. Plasma drug concentrations were measured using liquid chromatography-tandem mass spectrometry. Dose fractionation experiments were performed after establishing dose-response relationships. The magnitude of drug exposure required for stasis was established using 11 strains of Enterobacteriaceae (Escherichia coli, n = 6; Klebsiella pneumoniae, n = 5) with a variety of resistance mechanisms. The relationship between drug exposure and the emergence of resistance was established in a hollow-fiber infection model (HFIM). Tebipenem exhibited time-dependent pharmacodynamics that were best described by the free drug area under the concentration-time curve (fAUC0-24)/MIC corrected for the length of the dosing interval (fAUC0–24/MIC · 1/tau). The pharmacodynamics of tebipenem versus E. coli and K. pneumoniae were comparable, as was the response of strains possessing extended-spectrum β-lactamases versus the wild type. The median fAUC0-24/MIC · 1/tau value for the achievement of stasis in the 11 strains was 23. Progressively more fractionated regimens in the HFIM resulted in the suppression of resistance. An fAUC0-24/MIC · 1/tau value of 34.58 to 51.87 resulted in logarithmic killing and the suppression of resistance. These data and analyses will be used to define the regimen for a phase III study of adult patients with cUTI.
TEXT
The emergence of multidrug-resistant (MDR) Gram-negative bacteria is a growing threat to public health. Antimicrobial agents to treat infections caused by extended-spectrum-β-lactamase (ESBL)-producing and quinolone-resistant Gram-negative pathogens represent an urgent unmet need. The incidence of ESBLs in Escherichia coli and Klebsiella pneumoniae is 14% to 23% in a wide range of health care settings (1, 2). Infections caused by ESBL-producing pathogens are generally treated with an agent from the carbapenem, quinolone, or aminoglycoside class of antibiotics (3). Unfortunately, however, 55% to 100% of Gram-negative pathogens harboring ESBLs have been reported to also be resistant to quinolones (4, 5), and coresistance to trimethoprim-sulfamethoxazole (TMP-SMX) is also common (6). Hence, there are few options for oral therapy, which then requires the placement of an intravenous (i.v.) line and admininstration of parenteral antibiotics. New orally biovailable agents with activity against ESBLs and quinolone-resistant Gram-negative bacteria would minimize the use of i.v. lines and facilitate the management of complex infections in ambulatory care settings.
Tebipenem pivoxil (TBPM-PI; Orapenem) is a broad-spectrum, orally bioavailable carbapenem that is currently licensed for use in Japan. Orapenem is used for ear, nose, and throat and respiratory infections in pediatric patients (7, 8). Tebipenem has broad-spectrum in vitro and in vivo activity against a variety of medically important Gram-negative and Gram-positive pathogens, including ESBL-producing Enterobacteriaceae (9). TBPM-PI is a prodrug, with the active moiety being tebipenem (TBPM; SPR859). Oral administration of the pivoxil prodrug is well tolerated in children and results in effective plasma concentrations in that population (7). Spero Therapeutics has modified the prodrug to form an HBr salt (TBPM-PI-HBr) to improve drug substance and drug product properties, such as stability. A phase III program in adult patients with complicated urinary tract infections (cUTI) has been accepted under an investigational new drug aplication by the U.S. Food and Drug Administration (FDA) and will commence enrollment in the near future, with ertapenem being used as the comparator.
Herein, we describe the preclinical pharmacokinetics (PK) and pharmacodynamics (PD) of tebipenem as a first critical step for the clinical development of this compound in adults. The principal goals of this program of work were the identification of the regimen for use in patients with cUTI, demonstration of in vivo activity against ESBL-producing E. coli and K. pneumoniae, and an understanding of the relationship between drug exposure and the emergence of antimicrobial resistance. To help place the experimental findings in a clinical context, the pharmacodynamics of tebipenem were benchmarked against those of ertapenem. Collectively, these data and analyses accelerate and derisk the clinical development of tebipenem.
RESULTS
In vitro susceptibility and genotype.The MICs against the strains used in this study are summarized in Table 1. MICs were determined using Clinical and Laboratory Standards Institute (CLSI) methodology (10). MICs were estimated using at least 10 independently conducted experiments, and the mode was used for the pharmacodynamic analyses. The MICs of the other carbapenems in clinical use were also estimated, using the same methodology and approach. The genotypes of the challenge strains are shown in Table 1.
Challenge strains used in this study
Murine dose-response relationships.The dose-response relationship for tebipenem was defined over the course of 3 independently conducted experiments, and a compilation of these results is shown in Fig. 1. All doses were administered as prodrug. The experimental conditions used to establish this relationship were derived from preliminary dose-finding studies (data not shown) using E. coli ATCC 25922. The fit of an inhibitory sigmoid maximum-effect (Emax) model to the data was acceptable (Fig. 1). A half-maximal effect was observed with a total daily dose of 11.13 mg/kg of body weight (i.e., 3.71 mg/kg administered every 8 h [q8h]). The total daily doses that induced 20, 40, 60, and 80% of the maximal effect, as defined from the inhibitory sigmoid Emax model, were 5, 9, 14, and 25 mg/kg/day, respectively (i.e., 1.67, 3, 4.67, and 8.33 mg/kg q8h, respectively).
Pharmacodynamics of tebipenem against E. coli ATCC 25922 in a neutropenic thigh infection model. The data are the mean ± standard deviation for 3 mice collected over the course of three independently conducted experiments. The solid squares are the bacterial density at the commencement of therapy, which was at 2 h postinoculation. The mean of these points defines the stasis line, depicted by the broken horizontal line. The open circles are the data points from mice sacrificed after receiving 24 h of antibacterial therapy. The solid line is the fit of an inhibitory sigmoid Emax model.
Murine pharmacokinetics.The pharmacokinetics of tebipenem were linear over the regimens used in this study, which were primarily chosen to encompass the relevant pharmacodynamic relationships that had been initially determined from the dose-response relationships. The concentration-time profiles for total tebipenem from each dose are shown in Fig. 2, and the parameter values from a population PK model fitted to the total data are summarized in Table 2. Free drug concentrations were calculated by multiplying by the free fraction of 1.3%.
Pharmacokinetics of tebipenem. The data are the mean ± standard deviation for 3 mice. A destructive design was used. Mice sampled in the period from 16 to 24 h had received drug at 0, 8, and 16 h. The mice were infected with E. coli ATCC 25922.
Parameter estimates from the population PK analysis fitted to total drug concentrations
Persistent antibacterial effect.In order to provide an estimate of the magnitude of any persistent effect of tebipenem, the time course of bacterial density in the thigh was determined using a serial sacrifice design. E. coli ATCC 25922 was used as the challenge strain, and the neutropenic thigh infection model described in Materials and Methods was used for these experiments. A mathematical model was used to determine the difference between the time that the free plasma concentrations dipped beneath the MIC of 0.016 mg/liter (Table 1) and the time to regrowth. Each cohort of mice was treated as an individual. The population PK were initially solved, and the parameters from that analysis (Table 2) were fixed to fit a pharmacodynamic model to the pharmacodynamic data. The parameters for this analysis are shown in Table 3.
Parameter estimates from the population PD analysis of the single-dose experiment in mice
Experiments and PK-PD modeling showed that for each cohort, bacterial growth increased as soon as the free plasma concentrations were less than the MIC (Fig. 3). Hence, tebipenem exhibited negligible persistent effects, which is in keeping with the pharmacodynamics of the carbapenem class.
Pharmacodynamics of tebipenem following a single dose of prodrug, administered at 2 h postinoculation. Data are mean ± standard deviation for 3 mice. The solid lines are the fits of the PK-PD mathematical model to each cohort of mice (i.e., the Bayesian posterior estimates were used, with an individual being represented by the cohort of mice). The large arrows show the time of drug administration, and the small arrows show the predicted time that free plasma drug concentrations fell beneath the MIC of 0.015 mg/liter. The data point at the time of treatment initiation is common to each of the cohorts. (A) Controls; (B) 6.67 mg/kg once; (C) 16.67 mg/kg once; (D) 33.33 mg/kg once. There was a negligible persistent effect since bacterial growth commenced immediately after the free plasma drug concentrations fell beneath the MIC.
Murine dose fractionation studies.Dose fractionation studies were performed to define the pharmacodynamic index that best linked the administration of prodrug with the observed antibacterial effect. A single challenge strain was used, which was E. coli ATCC 25922. Dose fractionation studies were designed using oral regimens of prodrug that induced 20, 40, 60, and 80% of the maximal effect. The total daily dose at each value was administered as 4 quarter dosages every 6 h (q6h), 2 half dosages every 12 h (q12h), and a single daily dose every 24 h (q24h). The bacterial burden in each mouse (n = 4 for each condition) from each of these regimens was compared using analysis of variance (ANOVA), with a Bonferroni correction for multiple comparisons being used within each dose level where necessary. More fractionated schedules of administration generally resulted in more antibacterial activity, which was especially apparent for the q24h schedules. There were no differences for the 5-mg/kg/day group (P = 0.231) or the 25-mg/kg/day group (P = 0.113). However, there were statistically significant differences for the 9-mg/kg/day and 14-mg/kg/day groups, with P values of <0.001 and 0.001, respectively. For the 9-mg/kg/day group, the P values for the differences between the q6h and q12h, q6h and q24h, and q12h and q24h schedules were 0.066, <0.001, and 0.002, respectively. Similarly, for the 14-mg/kg/day group, the P values for the differences between the q6h and q12h, q6h and q24h, and q12h and q24h schedules were 0.604, 0.001, and 0.004, respectively.
The PK model was used to transform drug exposure from the dose to the free drug area under the concentration-time curve (fAUC0-24)/MIC, the free drug maximum concentration in plasma (fCmax)/MIC, and the fraction of the dosing interval that free tebipenem concentrations were greater than the MIC (fT>MIC), which was estimated to be 0.015 mg/liter. The murine plasma protein binding was estimated to be 98.7% (i.e., 1.3% free tebipenem), and free drug concentrations were estimated using this value (i.e., the total drug concentration was multiplied by 0.013). None of the standard pharmacodynamic indices described the data well (Fig. 4). The coefficient of determination (r2) for the maximum concentration in plasma (Cmax)/MIC and the area under the concentration-time curve (AUC)/MIC were 0.33 and 0.73, respectively. The plot of fT>MIC versus effect showed clumping of the data at or near 100%, with considerable differences in the pharmacodynamics occurring at this value. Hence, it was apparent that the fT>MIC metric (which is bounded below and above by 0 and 100%, respectively) could not be readily used to describe the pharmacodynamics of tebipenem in murine models of infection. Since time-dependent pharmacodynamics were apparent from inspection of the raw data as well as from ANOVA, two additional metrics that enable time-dependent pharmacodynamics to be described were explored: the minimum concentration in plasma (Cmin)/MIC and the AUC/MIC per length of dosing interval (AUC/MIC · 1/tau). The regressions obtained using these two indices are shown in Fig. 4. Both described the data well, with r2 values being 0.90 and 0.96 for Cmin/MIC and AUC/MIC · 1/tau, respectively. The AUC/MIC · 1/tau index performed slightly better and was used in the subsequent analyses.
Dose fractionation study. Regression of the various pharmacodynamic indices versus the observed antibacterial effect when various total daily dosages were administered q8h, q12h, and q24h. Each data point is the mean ± standard deviation for 4 mice. There was not a strong relationship between fCmax/MIC and effect (panel A). Tebipenem appeared to display time-dependent pharmacodynamics, but the fT>MIC plot (B) showed clumping of the data at 100% fT>MIC. There was a relationship between fAUC/MIC and effect, although there were some outlying points (panel C). The relationships with fCmin/MIC versus effect (D) and fAUC0-24/MIC · 1/tau (E) are two alternative time-dependent indices.
Pharmacodynamics of tebipenem against Enterobacteriaceae.The pharmacodynamics of tebipenem against E. coli (n = 6) and K. pneumoniae (n = 5) were determined to identify the magnitude of the AUC/MIC · 1/tau index that was required to achieve stasis in the murine thigh infection model. The MIC and the genotypes of these strains are summarized in Table 1. Tebipenem induced orders of logarithmic killing in all strains. There was no difference in the pharmacodynamics between E. coli and Klebsiella pneumoniae or between ESBL-producing and non-ESBL-producing strains (Fig. 5). A histogram of the magnitude of the AUC/MIC · 1/tau values that induced stasis is shown in Fig. 6. There was a wide range of pharmacodynamic targets, and the distribution was log normal. The median fAUC0-24/MIC · 1/tau value was 23 (Fig. 6).
Pharmacodynamics of tebipenem against E. coli and K. pneumoniae. Data are the mean ± standard deviation for 3 mice. (A) Pharmacodynamics of E. coli versus K. pneumoniae; (B) pharmacodynamics of ESBL-producing Enterobacteriaceae versus the wild type. Data from the strains were comodeled using fAUC0-24/MIC · 1/tau.
Histogram of the magnitude of the fAUC0-24/MIC · 1/tau value required to achieve stasis for tebipenem against strains of E. coli (n = 6) and K. pneumoniae (n = 5). The median fAUC0-24/MIC · 1/tau value was 23.
Pharmacodynamics of ertapenem.The pharmacodynamics of ertapenem were investigated to benchmark the behavior of tebipenem and were estimated using 3 strains of E. coli with different MICs. An inhibitory sigmoid Emax model fitted the data well and is shown in Fig. 7. Drug exposure was quantified in terms of the fT>MIC and fAUC/MIC · 1/tau, and stasis was achieved with values of 0.6 and 48, respectively.
Pharmacodynamics of ertapenem against E. coli. Data are the mean ± standard deviation for 3 mice. (A) Pharmacodynamics determined using fT>MIC, which is the traditional measure of drug exposure for the carbapenems. Stasis was achieved with a fT>MIC of 0.60. (B) The same data from the assay whose results are presented in panel A, but using fAUC0-24/MIC · 1/tau as the pharmacodynamic index. Stasis was achieved with a value of 46.
Emergence of resistance.The pharmacodynamics of the emergence of resistance to tebipenem were studied using a hollow-fiber infection model (HFIM) with ESBL-producing E. coli strain SPT719 as the challenge organism (Fig. 8). A preliminary dose-ranging experiment was performed to identify the regimens that resulted in information-rich portions of the relationships for drug exposure versus cell kill and drug exposure versus the emergence of resistance. A fractionation experiment was then performed where the total daily dose was administered once, two half dosages were administered q12h, three one-third dosages were administered q8h, and four quarter dosages were administered q6h. The total bacterial burden and a resistant subpopulation able to grow on agar containing 0.125 mg/liter SPR859 were quantified. Progressively more fractionated regimens resulted in progressively more cell kill and suppression of resistance. The parameter estimates from the mathematical model fitted to the entire dataset are shown in Table 4. An fAUC0-24/MIC · 1/tau value of between 34.58 and 51.87 resulted in logarithmic killing, with a value of 69.15 causing suppression of resistance (Fig. 8).
Hollow-fiber infection model. The challenge strain (SPT719) used in this experiment is an ESBL-producing E. coli strain (tebipenem MIC, 0.03 mg/liter). Each panel represents the data from an individual fiber. The data represent the total bacterial population (open red triangles), a resistant subpopulation (open blue circles), and the pharmacokinetic profile of tebipenem (open black squares). The outputs from the mathematical model are shown using the same colors. The Bayesian posterior estimates for each fiber are shown. The fAUC0-24/MIC · 1/tau index values associated with the q24h, q12h, q8h, and q6h schedules are 17.3, 34.58, 51.87, and 69.15, respectively. Logarithmic killing was observed with at fAUC0-24/MIC · 1/tau values of 34.58 to 51.87, and suppression of resistance was observed at a value of 69.15. Panel A shows the control. In panels B to E, the same total amount of SPR859 has been administered in different schedules.
Parameter estimates from the mathematical model fitted to the hollow fiber dataset
DISCUSSION
The spectrum of antimicrobial activity of tebipenem is comparable to that of currently available carbapenems, particularly ertapenem. Tebipenem has potent in vitro activity against Enterobacteriaceae and is unaffected by commonly encountered β-lactamases, such as extended-spectrum β-lactamases (ESBLs) and AmpC. The MICs of Pseudomonas aeruginosa typically range from 8 to 32 mg/liter, and Acinetobacter species MICs showed a broad range of values of between 0.06 and 8 mg/liter, meaning that overall there is little clinical utility against these pathogens (11). The rise of quinolone resistance (approximately 25% to 35% in recent studies) and coresistance to TMP-SMX means that patients frequently have no options for oral agents for either induction or maintenance therapy (6, 11, 12). This mandates the use of an injectable antibiotic, with the attendant cost, inconvenience, and infection risk of an i.v. line (13). The orally bioavailable formulation of tebipenem, TBPM-PI-HBr, provides an alternative oral treatment option for otherwise resistant Gram-negative pathogens and the ability to manage MDR infections in both inpatient and ambulatory care settings.
There have been significant advances in the use of PK-PD to develop new antibacterial and antifungal drugs. The European Medicines Agency (EMA) and FDA each have recommendations for preclinical data that should be considered to progress to human clinical trials, and this topic has also been considered in a recent publication sponsored by NIH/NIAID (14). A deep understanding of dose-exposure-response relationships substantially derisks development programs (15). PK-PD can be used to identify regimens for early- and later-phase clinical trials that are likely to be safe and effective, provide evidence for the biological information transmitted by the MIC, and provide evidence for activity against resistance mechanisms that are likely to be encountered in clinical studies (16). More recently, PK-PD have also been used to design regimens that minimize the emergence of resistance (17, 18). All of these facets were used in this study: tebipenem displayed time-dependent pharmacodynamics, pharmacodynamic targets that can be used for PK-PD bridging studies were defined (free drug area under the concentration-time curve from 0 to 24 h [fAUC0–24 ]/MIC · 1/tau = 23; see below), the activity of tebipenem against ESBL-producing E. coli and K. pneumoniae was established (exposure-response relationships from these strains were the same as those from the wild type), and incorporation of MIC into exposure-response relationships collapses the dose-exposure-response relationships, suggesting that it accounts for a significant portion of the observed variance.
For carbapenems, the traditional PK-PD index that best links drug exposure to antibiotic activity is the fraction of the dosing interval that the free drug concentrations are above the MIC (fT>MIC) (19). Currently licensed carbapenems generally achieve their maximal effect when concentrations remain above the MIC 40% of the time (20), although in the current study ertapenem achieved stasis with an fT>MIC of 60%. Tebipenem displays unequivocal time-dependent pharmacodynamics. More fractionated regimens in the mouse and HFIM resulted in statistically greater antibacterial activity in murine models and HFIM. However, in the murine studies, this time dependency could not be quantified using fT>MIC because of clumping of the data at 100% fT>MIC. Both submaximal and nearly maximal antibacterial activities were observed at this value (Fig. 4B). This is primarily a result of the inherent limitation of fT>MIC, which is bounded from below and above by 0 and 100%, respectively. We have encountered this issue for other drug-pathogen combinations and have previously used Cmin/MIC as the index to describe the time-dependent pharmacodynamics (21). In this study, we used fAUC0-24/MIC · 1/tau, which is a recently described PD index that enables time-dependent PD to be described (22). This index has the advantage of being unbounded from above and therefore able to adequately describe pharmacodynamic data at the extremes of dosing.
The dose fractionation study showed layering of the data with each schedule (Fig. 4). Each schedule of drug administration is associated with its own unique exposure-response relationship that can each be quantified in terms of AUC/MIC. Dividing by the length of the dosing interval (tau) collapses these relationships so that they can be described using a single metric, which is fAUC0–24/MIC · 1/tau. This index can be used to describe the behavior of other drugs that display time-dependent pharmacodynamics and yields results for dose prediction similar to (but not identical to) those yielded by other indices, such as fT>MIC and the free drug minimum concentration in plasma (fCmin)/MIC. Hence, the pharmacodynamics of other carbapenems can be described using fAUC0–24/MIC · 1/tau, as demonstrated in Fig. 7. A logical conclusion of being able to describe the pharmacodynamic data using fAUC0–24/MIC · 1/tau is that the magnitude of the target is proportional to the length of the dosing interval. If the length of the interval triples (e.g., q8h is increased to q24h), the fAUC0–24/MIC target value must also triple. Hence, the fAUC0–24/MIC · 1/tau target of 23 in this study is contingent on the use of a q8h schedule. The PD index can also be written as fAUC0–8/MIC = 23. The corresponding targets for q12h and q24h schedules are 34.5 and 69, respectively (i.e., fAUC0–12/MIC = 34.5 and fAUC0–24/MIC = 69, respectively). The schedule-dependent changes in the magnitude of the PD index are at odds with more traditional pharmacodynamic metrices, where the target is fixed and is independent of the schedule (e.g., the AUC/MIC target of 125 for the quinolones is independent of the schedule [23]).
A variety of endpoints from preclinical pharmacodynamic models are used to define the magnitude of the relevant pharmacodynamic index for preclinical-to-clinical bridging studies (20, 24–27). For murine studies, these include stasis and various orders of logarithmic killing. Stasis has generally been considered suitable for dose identification in patients with cUTI, whereas orders of logarithmic killing are used for more serious high-burden infections, such as hospital-acquired pneumonia and ventilator-associated pneumonia (14). Importantly, however, there are some additional subtleties. A determination of the magnitude of the pharmacodynamic index requires definition of both the study endpoint (e.g., stasis, 1-log kill, 2-log kill) and a collective measure of the response for the challenge strains used to determine the relevant magnitude. This may include the mean or median value or some other statistical measure that captures the pharmacodynamics of the collection of strains. In the current study, the median was used, given the nature of the distribution (Fig. 5). Stasis has been used as an endpoint for dose justification for cUTI for recently approved compounds, such as plazomicin, meropenem-vaborbactam, ceftazidime-avibactam, and ceftolozane-tazobactam (28–31). The relevance of the use of stasis as an endpoint was further supported by the benchmarking studies with ertapenem (Fig. 8). The drug exposure in humans (quantified in terms of fT>MIC and fAUC0-24/MIC · 1/tau) generated by 1 g i.v. q24h also produces stasis in the murine thigh infection model.
Tebipenem has a relatively long half-life in mice (circa 2.72 h) and high murine protein binding. However, the pharmacodynamic data from the murine models and HFIM (using a PK profile based on the findings of studies with healthy volunteers) were remarkably consistent. Confidence in the findings was further increased by the same conclusions from dose fractionation studies performed in two experimental model systems (mice and HFIM), the pharmacodynamic assessment of 11 different strains of Enterobacteriaceae with a variety of resistance mechanisms, and an understanding of the liability for the emergence of resistance with the candidate regimen for patients with cUTI. Together with human pharmacokinetic data, these pharmacodynamic models provide a rationale for the selection of a TBPM-PI-HBr regimen for a phase III trial comparing oral tebipenem to intravenous ertapenem in patients with cUTI.
MATERIALS AND METHODS
Drug.Tebipenem (pure active compound) was supplied by Spero Therapeutics as SPR859 (pure active compound) for liquid chromatography-tandem mass spectrometry (LC-MS/MS), hollow-fiber studies, MIC determinations, and incorporation into agar for evaluation of the emergence of resistance. The orally bioavailable prodrug was used for all murine studies. Dose formulations were stored on the bench at ambient temperature for the length of the study, and stability under these conditions was established (no longer than 24 h).
All test articles were prepared based on a dosing volume of 10 ml/kg of body weight. Prodrug was stored at room temperature. Solutions of prodrug at the appropriate concentration were prepared in 2.5% ethanol–2.5% Tween 80–95% citric acid-sodium citrate buffer (pH 4.0) prior to administration. Prodrug powder was initially suspended by vortex mixing for 30 s in 200 μl 50% ethanol–50% Tween 80, after which the mixture was allowed to stand for 2 min. Subsequently, the mixture was subjected to sonication for 10 min in a water bath at ambient temperature before being diluted with 3.8 ml filter-sterilized (pore size, 0.22 μm) citric acid-sodium citrate buffer (pH 4.0), and the pH was adjusted to between 4.5 and 5.0 with 6 N HCl. Further sonication followed by alternating cycles of vortex mixing and sonication (if required) were used to completely solubilize the prodrug. The formulation solutions were stable for up to 24 h at ambient temperature.
Isolates.Challenge organisms with a range of resistance mechanisms and tebipenem MICs were chosen (Table 1). A variety of wild-type and non-wild-type isolates of the Enterobacteriacae were investigated. The isolates were stored long term at −80°C. Colonies were confirmed to possess the correct characteristics and purity at the time of experimentation.
Molecular characterization of β-lactam resistance mechanisms.Molecular characterization of β-lactam resistance mechanisms was performed as JMI study 18-SPT-06.
(i) DNA extraction. Total genomic DNA of selected isolates was extracted using fully automated Thermo Scientific KingFisher Flex magnetic particle processors (Cleveland, OH, USA).
(ii) Whole-genome sequencing. Total genomic DNA was used as input material for library construction. DNA libraries were prepared using the Nextera XT library construction protocol and index kit (Illumina, San Diego, CA, USA) and sequenced on a MiSeq sequencer (Illumina) using a MiSeq reagent kit (v2, 500 cycles; v3, 600 cycles) with a minimum of 20× coverage.
(iii) DNA assembly and data analysis. FASTQ format sequencing files for each sample set were quality ensured, error corrected, and assembled independently using the de novo assembler SPAdes (v3.9.0). In-house-designed software was applied to the assembled sequences to align against the sequences of known β-lactam resistance genes. E. coli sequences of OmpC/OmpF and the respective homologues from K. pneumoniae, OmpK36/OmpK35 and OmpK37, were retrieved from FASTQ format sequencing files for each sample set and compared to those of reference, wild-type, and surveillance control isolates.
Measurement of tebipenem.Total drug concentrations of tebipenem in murine plasma were measured using an Agilent 6420 LC-MS/MS. A working solution of the internal standard, cefotaxime (Sigma-Aldrich, Dorset, UK), was prepared in methanol and acetonitrile (0.05 mg/liter in 50:50 methanol-acetonitrile) before being added to a 96-well Waters Sirocco protein precipitation plate (300 μl). Mouse plasma samples along with blanks, calibrators, and quality control samples were then aliquoted (30 μl) into the plate and mixed with the cefotaxime working solution on an orbital shaker for 2 min. Using a positive-pressure manifold, the liquid was drawn through the Sirocco plate into a collection plate. Supernatant (200 μl) from each well in the collection plate was transferred into a 96-well plate, which was then sealed and was ready for analysis by LC-MS/MS. The lower limit of quantification (LLOQ) was 0.05 mg/liter. The coefficient of variation (CV) was 0.29 to 11.55% over the concentration range of 0.05 to 10 mg/liter. The intra- and interday variation was −12.21 to 10.25%.
Murine thigh model of infection.All murine experiments were conducted under UK Home Office project license PAC022930 and approved by the Animal Welfare Ethics Review Board at the University of Liverpool. Male CD1 mice were provided by Charles River and weighed approximately 25 to 30 g at the time of experimentation. Food and water were provided ad libitum. Mice were housed in individually ventilated cages.
The challenge organism was recovered from beads that had been stored at −80°C, cultured onto Mueller-Hinton (MH) agar, and grown overnight at 37°C. A sweep of colonies was then placed in 2 × 30 ml of Mueller-Hinton broth (Sigma-Aldrich, Dorset, UK) and placed on an orbital shaker at 37°C for 4 or 24 h, depending on the species and strain. The final inoculum was prepared by centrifuging 30 ml at 1,438 × g for 5 min. The supernatant was removed and discarded. The final inoculum was dependent on the species, strain, and route of infection and was adjusted to the desired optical density with progressive dilution in sterile phosphate-buffered saline (PBS) using a spectrophotometer. The inoculum was checked with quantitative cultures.
A well-characterized neutropenic murine thigh infection model was used throughout (20, 32). Neutropenia was induced with 150- and 100-mg/kg cyclophosphamide (Baxter Healthcare Ltd., UK) administered intraperitoneally on study days −4 and −1, respectively, relative to the time of infection. Mice were inoculated while they were under temporary general anesthesia, induced with 2% isoflurane. A total of 0.05 ml per thigh of the diluted inoculum suspension was injected into each lateral thigh muscle of each mouse. Mice were anesthetized for approximately 5 min.
At 2 h postinfection, three animals were sacrificed using pentobarbitone overdose to provide a pretreatment control group. Thigh muscles (from immediately above the knee joint to the hip joint) were removed, weighed, and homogenized using a stainless steel sawtooth dispersing element (VWR, UK) in sterile PBS (2 ml). Thigh homogenates were serially diluted 10-fold in sterile PBS and plated onto Mueller-Hinton (MH) agar. Total bacterial counts were determined following overnight incubation at 37°C. The remaining mice were treated orally every 8 h (q8h) with a range of tebipenem doses. Ertapenem was administered on the same schedule. Groups of 3 mice were used for each regimen. At 26 h postinfection, the remaining mice were euthanized using a pentobarbitone overdose, and the thighs were processed in the same manner described above and used as the 2-h controls. The endpoint of the studies was the average bacterial density (number of CFU per gram of tissue) of both thighs.
Murine dose fractionation study.Male CD-1 mice were housed and rendered neutropenic as described above in the thigh model of infection. Mice were infected intramuscularly with E. coli ATCC 25922. Treatment was initiated at 2 h postinfection with total daily doses of prodrug at 5, 9, 14, and 25 mg/kg. These dosages of prodrug were determined from the dose-ranging studies (Fig. 1) and produced 20% of the maximal effect (EC20), EC40, EC60, and EC80 calculated from dose-ranging studies (Fig. 1). The doses were fractionated to be given 4 times, twice, or once a day over a period of 24 h (i.e., prodrug at 1.25 mg/kg q6h, 2.5 mg/kg q12h, and 5 mg/kg q24h, respectively, for the EC20; 2.25 mg/kg q6h, 4.5 mg/kg q12h, and 9 mg/kg q24h, respectively, for the EC40; 3.5 mg/kg q6h, 7 mg/kg q12h, and 14 mg/kg q24h, respectively, for the EC60; and 6.25 mg/kg q6h, 12.5 mg/kg q12h, and 25 mg/kg q24h, respectively, for the EC80). Groups of 4 mice were used per regimen. All mice were sacrificed at a single time point at 26 h postinfection. The study endpoint was the average bacterial density (CFU per gram of tissue) of both thighs, processed as described above.
Persistent antibacterial effect.The neutropenic murine thigh model of infection described above was used with E. coli ATCC 25922 to determine the persistent effect of tebipenem over 26 h. Treatment was intitiated at 2 h postinfection with doses of prodrug at 6.67, 16.67, and 33.33 mg/kg q24h, and ertapenem at 200 mg/kg q24h. An untreated cohort was also included. Groups of 3 mice per dosage-time point were euthanized via pentobarbitone injection at 2, 6, 10, 14, 18, 22, and 26 h postinfection. At each time point postinfection, the thighs were dissected and processed as previously described. The study endpoint was the average bacterial density (number of CFU per gram of tissue) of both thighs for each mouse.
Pharmacokinetic study.The murine pharmacokinetics were performed in immunosuppressed infected mice, using the murine thigh infection model described above with E. coli ATCC 25922. The regimens chosen for study were selected after preliminary pharmacodynamic experiments had been completed. Mice received prodrug at 3.33, 8.33, 16.67, or 33.33 mg/kg q8h over 24 h. Plasma samples for PK bioanalysis of tebipenem were taken in the 1st and 3rd dosing intervals. Mice from each cohort were euthanized in a serial sacrifice design at 0.083, 0.25, 0.5, 1, 2, 4, and 8 h postdose. Groups of 3 mice were used for each dose-time point combination. A combination of 5% isofluorane and 95% oxygen was used to anesthetize the mice for at least 5 min, until a deep plane of anesthesia was achieved. Whole blood was taken from the mice via terminal cardiac puncture using heparinized syringes at the required time postdose. Whole blood was centrifuged, and the plasma supernatant was stored at −80°C for bioanalysis.
Hollow-fiber infection model.E. coli SPT 719 (supplied by Spero Therapeutics) was used as the challenge strain against tebipenem in a hollow-fiber infection model (HFIM). The MIC was 0.03 mg/liter (Table 1). Cation-adjusted MH broth (Ca-MHB) was pumped from the central compartment through a hollow-fiber cartridge (FiberCell Systems, Frederick, MD, USA) before being returned to the central compartment. A peristaltic pump (model 205 U; Watson-Marlow, United Kingdom) was used. SPR859 (tebipenem) was administered to the central compartment via a programmable syringe driver (CME Medical, Blackpool, UK) over a 1-h period to achieve the required Cmaxs. Fresh Ca-MHB was pumped into the central compartment from a peripheral reservoir. Concomitantly, drug-containing broth in the central compartment was removed (at the same rate) into a waste reservoir. The simulated half-life was 35 min.
The extracapillary space of each HFIM was inoculated with ∼40 ml of bacterial suspension, and the desired inoculum was confirmed by quantitative cultures. The HFIM system was incubated at 37°C in ambient air. The PK were estimated by sampling from the central reservoir using the same bioanalytical assay described above. Samples were taken throughout the dosing interval (the specific times depended on the schedule). Bacterial densities were determined by removing 1 ml from the extracapillary space via a sampling port. Serial dilutions were then plated onto both drug-free and drug-containing (4× tebipenem MIC at 0.125 mg/liter) Ca-MH agar to enumerate the total and resistant subpopulations.
Pharmacokinetic-pharmacodynamic modeling.The drug exposure-response relationships were modeled using an inhibitory sigmoid Emax model that took the following form:
The PK of tebipenem were estimated using a population methodology and the Pmetrics program. The following structural model was used:
Equations 1 to 3 represent the rate of change of the mass of drug in the gut, bloodstream, and peripheral compartment, respectively. Ka is the first-order rate constant describing the absorption of drug from the gut into the central compartment, Kcp and Kpc are the respective first-order intercompartmental rate constants, SCL is the first-order clearance of drug from the central compartment, and V is the volume of the central compartment. XP(1), XP(2), and XP(3) are the rates of change of mass in compartments 1, 2, and 3, respectively. Y(1) is the output of the model.
The PD data from the experiments to describe the persistent effect were modeled using the following differential equation, which was combined with PK equations 1 to 3 above. The value of the PK parameters estimated from fitting of the PK model were fixed.
Equation 4 describes that rate of change (dN/dt) of the number of organisms (N) in the thigh at time t, which is a balance between bacterial growth and drug-induced bacterial killing. The maximum rate of growth is given by Kgmax (number of log10 CFU per gram of thigh tissue per hour), which is modulated by the drug concentrations, given by X(2)/V from equation 2 above. popmax is the maximum theoretical bacterial density. C50g is the concentration of drug at which there is half-maximal suppression of growth, and Hg is the associated slope function. The maximum rate of drug-induced killing is given by Kkmax (number of log10 CFU per gram of thigh tissue per hour), which is similarly affected by the drug concentrations. C50k and Hk are the concentration at which the rate of killing is half maximal and the slope function, respectively.
The hollow-fiber data were modeled using the following set of inhomogeneous differential equations that enabled the PK in each cartridge to affect bacterial killing of a susceptible and resistant bacterial subpopulation.
Equation 6 represents a 1-compartment model that describes the PK of tebipenem in the hollow-fiber circuit with input of drug represented by R(1). Equations 7 and 8 represent the pharmacodynamics of susceptible and resistant subpopulations, respectively. In each case, the rate of change of the organism is modeled as growth minus drug-induced killing. The pharmacodynamic equations take the same structural form with parameters that describe the maximum rate of growth of susceptible and resistant subpopulations (Kgmax_s and Kgmax_r, respectively), the maximum rate of drug-induced killing of susceptible and resistant subpopulations (Kkmax_s and Kkmax_r, respectively), the concentrations of tebipenem where the rate of killing is half maximal (C50k_s and C50k_r, respectively), and the respective slope functions for the susceptible and resistant subpopulations (Hks and Hkr, respectively). The total bacterial population consists of the sum of the susceptible and resistant subpopulations and is given in equation 10, while the resistant subpopulation is given by equation 11.
ACKNOWLEDGMENTS
We thank JMI for the challenge strains used in this study.
This study was supported by Spero Therapeutics.
W.H. holds or has recently held research grants with F2G, AiCuris, Astellas Pharma, Spero Therapeutics, Matinas Biosciences, Antabio, Amplyx, Allecra, Bugworks, NAEJA-RGM, AMR Centre, and Pfizer. He holds awards from the National Institutes of Health, Medical Research Council, National Institute of Health Research, FDA, and the European Commission (FP7 and IMI). W.H. has received personal fees in his capacity as a consultant for F2G, Amplyx, Ausperix, Spero Therapeutics, and BLC/TAZ. W.H. is an Ordinary Council Member for the British Society of Antimicrobial Chemotherapy. A.J., N.C., I.C., D.M., and T.P. are employees of Spero Therapeutics. P.G.A. holds research grants and is also a consultant for Spero Therapeutics.
FOOTNOTES
- Received 22 March 2019.
- Returned for modification 23 April 2019.
- Accepted 14 May 2019.
- Accepted manuscript posted online 20 May 2019.
- Copyright © 2019 American Society for Microbiology.