Optimization of a Meropenem-Tobramycin Combination Dosage Regimen against Hypermutable and Nonhypermutable Pseudomonas aeruginosa via Mechanism-Based Modeling and the Hollow-Fiber Infection Model

ABSTRACT

Hypermutable Pseudomonas aeruginosa strains are prevalent in patients with cystic fibrosis and rapidly become resistant to antibiotic monotherapies. Combination dosage regimens have not been optimized against such strains using mechanism-based modeling (MBM) and the hollow-fiber infection model (HFIM). The PAO1 wild-type strain and its isogenic hypermutable PAOΔmutS strain (MIC_meropenem of 1.0 mg/liter and MIC_tobramycin of 0.5 mg/liter for both) were assessed using 96-h static-concentration time-kill studies (SCTK) and 10-day HFIM studies (inoculum, ∼10^8.4 CFU/ml). MBM of SCTK data were performed to predict expected HFIM outcomes. Regimens studied in the HFIM were meropenem at 1 g every 8 h (0.5-h infusion), meropenem at 3 g/day with continuous infusion, tobramycin at 10 mg/kg of body weight every 24 h (1-h infusion), and both combinations. Meropenem regimens delivered the same total daily dose. Time courses of total and less susceptible populations and MICs were determined. For the PAOΔmutS strain in the HFIM, all monotherapies resulted in rapid regrowth to >10^8.7 CFU/ml with near-complete replacement by less susceptible bacteria by day 3. Meropenem every 8 h with tobramycin caused >7-log₁₀ bacterial killing followed by regrowth to >6 log₁₀ CFU/ml by day 5 and high-level resistance (MIC_meropenem, 32 mg/liter; MIC_tobramycin, 8 mg/liter). Continuous infusion of meropenem with tobramycin achieved >8-log₁₀ bacterial killing without regrowth. For PAO1, meropenem monotherapies suppressed bacterial growth to <4 log₁₀ over 7 to 9 days, with both combination regimens achieving near eradication. An MBM-optimized meropenem plus tobramycin regimen achieved synergistic killing and resistance suppression against a difficult-to-treat hypermutable P. aeruginosa strain. For the combination to be maximally effective, it was critical to achieve the optimal shape of the concentration-time profile for meropenem.

INTRODUCTION

Pseudomonas aeruginosa has an extraordinary capacity to develop resistance to antipseudomonal agents (1). Treatment failure due to emergence and amplification of antibiotic-resistant mutants is a frequent outcome of P. aeruginosa infections (2–4) and is particularly important in the management of chronic infections requiring prolonged treatment (5). As antimicrobial resistance has increased worldwide, hypermutation has become increasingly recognized as a major problem for antimicrobial therapy (6, 7). Hypermutable strains (i.e., strains with up to 1,000-fold increased mutation rates caused by defects in DNA repair or error avoidance systems, commonly due to mutations in the mutS gene) of P. aeruginosa develop resistance even more rapidly than nonhypermutable strains (8–12). Hypermutable P. aeruginosa strains are prevalent in patients with cystic fibrosis (CF) and have been linked to increased multidrug resistance (MDR) and reduced lung function (8, 11–16).

Hypermutable P. aeruginosa strains can become resistant to all available antibiotics administered as monotherapy (15, 17). Given the serious clinical consequences arising from infections caused by hypermutable strains and their frequent development of MDR, there is an urgent need to optimize antibiotic therapy to improve efficacy, including suppression of the emergence of resistance (18, 19). However, exacerbations of chronic P. aeruginosa infections are typically treated with either nonoptimized monotherapy or empirically chosen nonoptimized combinations (20), risking the emergence of MDR hypermutable strains. We therefore aimed to (i) characterize the effect of different concentrations of meropenem and tobramycin, in monotherapy and combination, on bacterial killing and resistance emergence in hypermutable and nonhypermutable P. aeruginosa, (ii) develop mechanism-based models (MBM) that support the selection of an optimized combination dosage regimen that maximizes bacterial killing and minimizes the emergence of resistance, and (iii) evaluate optimized and standard combination regimens in the hollow-fiber infection model (HFIM).

(Part of this work has been presented as a poster at the 55th Interscience Conference on Antimicrobial Agents and Chemotherapy, 18 to 21 September 2015, San Diego, CA. This work was also subject to oral presentations at the 27th European Congress of Clinical Microbiology and Infectious Diseases, 22 to 25 April 2017, Vienna, Austria, and the Australian Society of Clinical and Experimental Pharmacologists and Toxicologists Annual Meeting, 29 November to 2 December 2015, Hobart, Australia.)

RESULTS

Static-concentration time-kill studies (SCTK).

Total viable counts are presented in Fig. 1, and mutation frequencies (see Table S1 the supplemental material), log₁₀ changes in viable counts (Table S2), and less-susceptible subpopulations (Fig. S1 and S2) are also presented.

FIG 1.

Bacterial counts with various concentrations of tobramycin and meropenem alone and in combination from the static time-kill model over 96 h. MEM, meropenem; TOB, tobramycin. For each strain, MIC_MEM is 1 mg/liter and MIC_TOB is 0.5 mg/liter.

PAO1 strain.

All tobramycin concentrations produced rapid (within 7 h) initial killing of strain PAO1 of ∼2 to 4 log₁₀ CFU/ml, followed by regrowth close to growth control values, with the emergence of a high proportion of resistant bacteria. Meropenem at 2 mg/liter resulted in ∼4-log₁₀ killing over the first 29 h, followed by regrowth due almost entirely to less susceptible bacteria; at 8 and 16 mg/liter, bacterial killing continued, with no colonies detected at 96 and 72 h, respectively. The combination of the two lowest concentrations of each antibiotic produced ∼3-log₁₀ CFU/ml initial killing and a slow decline in bacterial numbers thereafter, with no viable bacteria detected at 96 h. All other combinations produced rapid initial killing of ∼4-log₁₀ CFU/ml with synergy occurring at various times thereafter and no viable colonies from 24 to 48 h onwards.

PAOΔmutS strain.

Tobramycin at 1 mg/liter produced little PAOΔmutS strain bacterial killing without increases in resistant subpopulations. A 4 and 8 mg/liter initial ∼3- to 4-log₁₀ killing was followed by rapid regrowth with emergence of a high proportion of resistant bacteria. Meropenem at 2 mg/liter produced ∼2-log₁₀ killing at 6 h, whereas at 8 mg/liter ∼5-log₁₀ killing occurred over 29 h. In both cases rapid regrowth followed, with virtually the entire population growing on the 2.5-mg/liter meropenem-containing agar plates. Meropenem at 16 mg/liter produced sustained killing over 96 h but not eradication, with less susceptible subpopulations emerging. Combinations containing tobramycin (all concentrations) plus meropenem at 2 mg/liter produced rapid ∼3 to 4 log₁₀ CFU/ml killing over 6 h, followed by gradual regrowth (tobramycin at 1 mg/liter plus meropenem at 2 mg/liter) or suppression of regrowth (tobramycin at 4 or 8 mg/liter plus meropenem at 2 mg/liter) with synergy. In all combinations with 2 mg/liter meropenem, tobramycin-resistant subpopulations emerged by 48 h and subsequently increased, whereas subpopulations less susceptible to meropenem were detected after 24 h only with the two lowest tobramycin concentrations in combination. All other combinations, except 1 mg/liter tobramycin plus 16 mg/liter meropenem, produced synergistic killing from ∼24 to 48 h onwards and eventual eradication (no viable bacteria detected).

Mechanism-based mathematical modeling.

The SCTK data for both strains were comodeled and well described by the developed MBM, including subpopulation and mechanistic synergy (Fig. 2 and 3). A model with a single maximum killing rate constant (K_max; i.e., using the same K_max for each subpopulation and strain) per antibiotic was clearly inferior based on the significantly poorer (38 points; P < 0.0001) objective function (−1× log likelihood in S-ADAPT) and population fit plots. Thus, different estimates for K_max were required. The model yielded unbiased and precise curve fits for the total bacterial counts of both strains. The coefficient of correlation for the observed versus individual fitted viable counts was 0.989. The parameter estimates are reported in Table S3. The extended MBM adequately captured the growth and killing of less susceptible bacteria (Fig. S3).

FIG 2.

(A) Fit plots for PAO1. (B) Fit plots for PAOΔmutS strain.

FIG 3.

Observed versus individual and population fitted viable counts for meropenem and tobramycin alone and in combinations against PAO1 and PAOΔmutS strains.

The MBM predicted failure of all monotherapies against both strains in the HFIM, except for suppression of regrowth of PAO1 over 10 days by meropenem at 3 g/day as a continuous infusion (Table 1). Against PAO1, both combination regimens were predicted to suppress regrowth over 10 days. In contrast, the MBM predicted suppression of regrowth of the PAOΔmutS strain would be achieved only by the optimized combination regimen with continuous infusion of meropenem (Table 1).

TABLE 1.

Meropenem and tobramycin dosage regimens, model predictions, and HFIM outcomes against PAO1 and PAOΔmutS strains^a

Treatmentb	MEM fT>MIC (%) for PAO1 and PAOΔmutS (MIC, 1 mg/liter)		TOB fAUC/MIC for PAO1 and PAOΔmutS (MIC, 0.5 mg/liter)	Predictions and outcomes for:
				PAO1		PAOΔmutS
	fT>1× MIC	fT>5× MIC		Predicted from MBM	Observed in the HFIM	Predicted from MBM	Observed in the HFIM
MEM, 1 g, every 8 h	61	36		∼5.2 log10 killing at 71 h, slow regrowth from 95 h, counts within ∼1 log10 of control values from 239 h	∼5.2 log10 killing at 71 h, regrowth from 168 h, counts within ∼1 log10 of control values from 239 h	∼3 log10 killing at 23 h (nadir), regrowth from 26 h, counts within ∼1 log10 of control values from 47 h	∼1 log10 killing at 1.5 to 23 h, regrowth from 26 h, counts within ∼1 log10 of control values from 47 h
MEM, 3 g/day, CI	100	100		∼6.5 log10 killing at 71 h, suppressed regrowth over 239 h	∼5.5 log10 killing at 71 h, limited regrowth at 239 h only	∼3.7 log10 killing at 26 h (nadir), regrowth from 47 h, counts within ∼1 log10 of control from 71 h	∼2 log10 killing at 23 h (nadir), regrowth from 26 h, counts within ∼1 log10 of control values from 71 h
TOB, 10 mg/kg, every 24 h			217	∼3.5 log10 killing, regrowth from 23 h, counts at control values from 47 h	∼7.2 log10 killing, regrowth from 47 h, counts within ∼1.5 log10 of control from 71 h	∼3.0 log10 killing at 5 h (nadir), regrowth from 23 h, counts at control values from 47 h	∼4.5 log10 killing at 5 h (nadir), regrowth from 23 h, counts at control from 71 h
MEM, 1 g, every 8 h, plus TOB	61	36	217	∼6.7 log10 killing (faster than MEM alone) at 47 h, counts of <1 log10 from 53 h until 239 h	∼6.9 log10 killing (faster than MEM alone) at 47 h, counts of <1 log10 from 53 h until 239 h	∼5.8 log10 killing at 53 h (nadir), ∼5 log10 killing at 71 h, regrowth above 4 log10 from 95 h, ∼7 log10 reached at 143 h	∼5.7 log10 killing at 53 h, ∼7.5 log10 killing at 71 h (nadir), regrowth above 4 log10 from 95 h, ∼7 log10 reached at 119 h
MEM, 3 g/day, CI, plus TOB	100	100	217	∼8.5 log10 killing (faster than MEM alone) from 47 h, counts of ∼1.0 log10 or below until 239 h	∼8.5 log10 killing (faster than MEM alone) from 26 h, counts of ∼1.0 log10 or below until 239 h	∼8.0 log10 killing at 71 h, counts of <1.0 log10 after that until 239 h	∼8.0 log10 killing at 71 h, counts of ∼1.0 log10 or below from 119 h until 239 h

^aMEM, meropenem; TOB, tobramycin; MBM, mechanism-based model; HFIM, hollow-fiber infection model; CI, continuous infusion; fT_{>1× MIC}, the cumulative percentage of a 24-h period that unbound concentrations exceed 1× MIC; fT_{>5× MIC}, the cumulative percentage of a 24-h period that unbound concentrations exceed 5× MIC; fAUC/MIC, ratio of the area under the unbound concentration-time curve to MIC. For less susceptible populations of PAO1, the extended MBM predicted the following: for 1 g MEM every 8 h, almost complete replacement of the total by populations less susceptible to MEM on day 10; for 3 g/day MEM CI, no quantifiable populations less susceptible to MEM from 23 h onwards; for 1 g MEM every 8 h plus TOB and 3 g/day MEM CI plus TOB, no quantifiable populations less susceptible to MEM and TOB from 23 h onwards; for 10 mg/kg TOB every 24 h, populations less susceptible to TOB within ∼1 log₁₀ CFU/ml of the total population from 47 h onwards. For less susceptible populations of the PAOΔmutS strain, the extended MBM predicted the following: for 1 g MEM every 8 h and 3 g/day MEM CI, almost complete replacement of the total by populations less susceptible to MEM from 47 h onwards; for 10 mg/kg TOB every 24 h, populations less susceptible to TOB within ∼1 log₁₀ CFU/ml of the total population from 47 h onwards; for 1 g MEM every 8 h plus TOB, populations less susceptible to MEM and TOB within ∼1 log₁₀ CFU/ml of the total population from 71 h onwards; for 3 g/day MEM CI plus TOB, no quantifiable populations less susceptible to MEM and TOB from 23 h onwards.

^bSimulated half-lives: meropenem, 0.8 h; tobramycin, 2.5 h. No loading dose was administered for intermittent dosing. The meropenem CI (3 g/day) was started at the average steady-state concentration of ∼8 mg/liter.

Hollow-fiber in vitro infection model.

For each of the simulated profiles, the observed meropenem and tobramycin concentrations were, on average, within 20% of the targeted concentrations. Changes in viable counts, total and less susceptible populations, and MICs are shown in Fig. 4 and 5 and Table S6. Changes in mutation frequencies are shown in Tables S4 and S5.

FIG 4.

Bacterial counts following meropenem (3 g/day, continuous infusion, or 1 g via 30-min intravenous infusion every 8 h) and tobramycin (10 mg/kg every 24 h as a 1-h infusion) in monotherapies and combination studied in the HFIM. MEM, meropenem; TOB, tobramycin. For each strain, MIC_MEM is 1 mg/liter and MIC_TOB is 0.5 mg/liter.

FIG 5.

Bacterial counts on agar plates containing 2.5 mg/liter tobramycin, 2.5 mg/liter meropenem, and 5 mg/liter meropenem from the HFIM over 10 days. MEM, meropenem; TOB, tobramycin. For each strain, MIC_MEM is 1 mg/liter and MIC_TOB is 0.5 mg/liter.

PAO1.

The growth control grew rapidly to ∼10.5 log₁₀ CFU/ml, with less susceptible populations for both meropenem (2.5 and 5 mg/liter on agar) and tobramycin (2.5 mg/liter on agar), plateauing at ∼3 to 4 log₁₀ CFU/ml. Tobramycin monotherapy produced killing of ∼7 log₁₀ over the first day, followed by rapid regrowth to control values. Regrowth was accompanied by an ∼5-log₁₀ increase in less susceptible bacteria compared to the control. Both the intermittent and continuous meropenem regimens produced nearly identical killing of ∼5-log₁₀ over the first 7 days, with few or no less susceptible bacteria detected. After that, rapid regrowth to control levels occurred with intermittent meropenem due almost entirely to an increase in less susceptible bacteria. Growth with the continuous meropenem regimen remained low (∼2.0 to 2.5 log₁₀ CFU/ml) until day 9 before increasing to ∼5 log₁₀ at day 10; no less susceptible bacteria were detected at any time. Both combination regimens suppressed regrowth over 10 days, with few if any colonies detected after 24 h.

PAOΔmutS strain.

For the PAOΔmutS strain, the growth control grew rapidly to ∼10.5 log₁₀ CFU/ml, with less susceptible subpopulations plateauing at ∼7 to 8 log₁₀ CFU/ml and ∼4 to 5 log₁₀ CFU/ml on agar containing 2.5 mg/liter and 5 mg/liter meropenem, respectively. Less susceptible populations plateaued at ∼4 log₁₀ CFU/ml on the tobramycin-containing agar. With monotherapy, initial killing of ∼4 log₁₀ with tobramycin and <∼1.5 log₁₀ with both meropenem regimens was followed in all cases by rapid regrowth due almost entirely to less susceptible bacteria. Both combinations produced nearly identical killing of ∼7.5 log₁₀ over the first 3 days. With the standard combination (containing 1 g meropenem given every 8 h), rapid regrowth of ∼6 log₁₀ due almost entirely to less susceptible subpopulations subsequently occurred, with growth plateauing at ∼7 log₁₀. With the optimized combination, bacterial killing continued such that no viable bacteria were detected at 7 and 10 days.

DISCUSSION

The PAO1 wild-type reference strain and its isogenic hypermutable PAOΔmutS strain were used in this study. The PAOΔmutS strain differs from PAO1 only by the absence of mutS, which encodes a component of the mismatch repair system that detects and repairs replication errors. The strains shared the same MICs for each antibiotic. Deletion of mutS is one of the most frequent mutations in clinical hypermutable P. aeruginosa isolates, and it represents a nearly worst-case scenario, as it has a large impact on increasing mutation rate (12, 21, 22). The increased resistance of the PAOΔmutS strain is due to the ascent to dominance of resistant mutant subpopulations, but increases in minimum bactericidal concentrations observed with single agents (meropenem, imipenem, and ceftazidime) can be minimized using combinations of two bactericidal antipseudomonal agents (10). Current guidelines endorse the use of combination antipseudomonal therapy for the treatment of acute exacerbations in patients with CF (23), with antipseudomonal β-lactams and aminoglycosides most commonly used (24–26). However, combination dosage regimens have never been optimized. Antipseudomonal β-lactams are commonly administered via intermittent infusion (every 6 to 8 h over 5 to 30 min) (25).

Based on synergistic bacterial killing of the PAOΔmutS strain by clinically relevant meropenem and tobramycin concentrations in SCTK, we evaluated this combination in standard and optimized dosage regimens (the latter selected using MBM, which incorporated the SCTK data) in the HFIM. Short-term infusions remain the standard method of administration for β-lactams, including meropenem, in a large part of the world, as demonstrated by two recent studies that included up to 53 countries (27, 28). However, prolonged infusions are used in other countries. We performed simulations with our MBM for 0.5-h, 3-h, and continuous-infusion meropenem dosing in monotherapy and combination with tobramycin. Our MBM predicted that only the optimized combination regimen (containing meropenem at 3 g/day with continuous infusion) would suppress bacterial regrowth and minimize resistance emergence over 10 days for both strains. Regrowth and resistance emergence of the PAOΔmutS strain was predicted to occur with both the standard combination (containing 1 g meropenem every 8 h as 0.5-h infusions) and the combination including meropenem as prolonged infusions (1 g every 8 h as 3-h infusions; data not shown). Therefore, the combinations containing standard (short-term) and continuous-infusion meropenem were evaluated in the HFIM. Both of these combination regimens produced nearly identical killing of PAOΔmutS organisms across the first 3 days in the HFIM; extensive regrowth of less susceptible bacteria ensued with the standard combination, whereas killing continued across 10 days with the optimized combination regimen, similar to the MBM predictions.

Our results are instructive in a number of ways. First, they highlight the utility of MBM to select optimized regimens that maximize bacterial killing and minimize resistance emergence against hypermutable P. aeruginosa, an especially important finding given these strains can rapidly develop MDR. Second, they show the potentially different pharmacodynamic responses to therapy of hypermutable and nonhypermutable strains, with bacterial killing and emergence of resistance (for both mono- and combination therapy) differing substantially between otherwise identical strains. Third, as will be discussed below, they show the importance of shape in relation to concentration-time profiles.

PK/PD approaches to optimize the administration of β-lactams have typically sought to maximize the fT_>MIC, i.e., the fraction of the dosing interval for which the unbound concentration remains above 1× MIC of the infecting pathogen (29, 30). For carbapenems such as meropenem, in vitro, ex vivo, and animal in vivo studies, usually of relatively short duration (≤24 h), have shown fT_>MIC values of 20% and 40% to be necessary for bacteriostasis and near-maximal bacterial killing, respectively (31–33). In the present HFIM studies, meropenem monotherapy with both intermittent (fT_>MIC of 61%) and continuous (fT_>MIC of 100%) infusions did suppress regrowth over at least 7 days against the wild-type strain. However, even the continuous infusion, where concentrations remained at ∼8× MIC at all times, was completely ineffective against the hypermutable strain. This comprehensive failure of both intermittent and continuous infusion as monotherapy against the PAOΔmutS strain, due to the rapid ascent to dominance of mutants less susceptible to meropenem (Fig. 5), strongly argues against the use of monotherapy, especially against hypermutable strains.

While our data suggest that combination therapy is required for treatment of P. aeruginosa infections that are chronic or known to involve hypermutable strains, the design of such regimens must consider ways to minimize the emergence of resistance (10). With maximal bactericidal activity of the β-lactams occurring at 4× to 5× MIC (34, 35) and regrowth often observed as soon as concentrations fall below the MIC (36, 37), pharmacodynamic targets such as 100% fT_>4–5×MIC and unbound minimum concentration (fC_min)/MIC have been investigated in relation to suppression of emergence of resistance (38–42). We observed that the standard meropenem regimen, in combination with tobramycin, eliminated the wild-type strain but not the hypermutable strain, which demonstrated significant regrowth comprising large numbers of bacteria less susceptible to both antibiotics. However, greatly enhanced bacterial killing and resistance suppression were achieved with the optimized combination. For the standard intermittent meropenem regimen in the HFIM, concentrations dropped below the MIC ∼4.2 h after administration, with essentially no drug remaining at the end of each dosing interval (C_min of 0.07 mg/liter for meropenem every 8 h). However, with the optimized combination regimen, meropenem concentrations greatly exceeded the MIC (∼8× MIC) at all times, even when tobramycin concentrations had declined. For meropenem, an fC_min/MIC of ≥2 to 6 and 100% fT_>5×MIC have been associated with suppression of resistance emergence (38–41), with both targets substantially exceeded with the continuous infusion regimen. Thus, despite each meropenem regimen having the same daily dose, the shape of the optimized meropenem concentration-time profile maximized fT_>5×MIC and fC_min/MIC compared to the standard regimen, such that in combination with tobramycin, emergence of resistance was minimized. Achieving an optimal shape of the concentration-time profile for meropenem was therefore critical for the combination to be maximally effective.

Previous work has shown that combining antibiotics that require separate and independent mutations for resistance development may help minimize selection of mutants resistant to multiple drugs (10). Resistance to meropenem occurs primarily via reduced expression of the gene for the outer membrane porin OprD (which allows carbapenems to enter the cell) and enzymatic inactivation via carbapenemases (43). The most important resistance mechanisms of P. aeruginosa against aminoglycosides are increased expression of MexXY-OprM (particularly common in strains from patients with CF) (44), modification of the binding site, and enzymatic inactivation (43). The different resistance mechanisms of each antibiotic in the combination, together with the maintenance of higher meropenem concentrations, may have contributed to the substantially enhanced effectiveness of the optimized combination regimen against the hypermutable strain.

We have examined standard and optimized regimens against a hypermutable and a nonhypermutable P. aeruginosa strain. The focus on P. aeruginosa may be regarded as a limitation of the study, as the airways of patients with CF are often colonized with multiple bacterial species simultaneously (45). Nevertheless, P. aeruginosa is the predominating species in most adults with CF and remains the most common pathogen associated with morbidity and mortality among patients with CF (46–48). Future studies may be directed at evaluating the optimized combination regimen against other P. aeruginosa isolates or other species, alone or in combination. The HFIM utilized lacks an immune system, in particular granulocytes, which work in combination with antibiotic treatment to control bacterial infections. Therefore, future animal studies may be warranted to assess immune system effects on residual populations following initial bacterial killing by the antibiotics. However, given the ethical limitations on study duration inherent in animal studies, suppression of the emergence of resistance, a key component of the present study, is best examined in the HFIM where longer study durations can be employed.

In summary, standard regimens of meropenem and tobramycin, both as monotherapy and in combination, produced substantially less bacterial killing and suppression of emergence of resistance against a hypermutable P. aeruginosa strain than its isogenic wild-type reference strain. MBM was applied effectively to SCTK data to select an optimized dosage regimen. Against the hypermutable strain the optimized combination regimen, when subjected to human pharmacokinetics in the HFIM, prevented the regrowth and emergence of resistance observed with the standard combination regimen. Achieving an optimal shape of the concentration-time profile for meropenem was critical for the combination to be maximally effective against both strains. The optimized combination dosage regimen is expected to be highly promising for evaluation in future clinical studies.

MATERIALS AND METHODS

Antibiotics, bacterial strains, media, and susceptibility testing.

Stock solutions of meropenem (Hospira, Melbourne, Australia) and tobramycin (AK Scientific, Union City, MD, USA) were prepared as previously described (49). Viable counting was performed on cation-adjusted Mueller-Hinton agar (CAMHA; containing 25 mg/liter Ca²⁺ and 12.5 mg/liter Mg²⁺; Media Preparation Unit, University of Melbourne, Melbourne, Australia). Drug-containing agar plates were prepared by adding appropriate volumes of stock solution to CAMHA (BD, Sparks, MD, USA). All studies used CAMHB (BD, Sparks, MD, USA) containing 25 mg/liter Ca²⁺ and 12.5 mg/liter Mg²⁺.

The P. aeruginosa wild-type reference strain PAO1 and its PAOΔmutS isogenic hypermutable strain (constructed from PAO1 by Mena et al. [50] via mutS deletion) were used. The meropenem and tobramycin MICs prior to drug exposure were determined in duplicate on separate days (51). For both strains the meropenem and tobramycin MICs were 1 mg/liter and 0.5 mg/liter, respectively. Susceptibility and resistance were defined as MICs of ≤2 and >8 mg/liter for meropenem and ≤4 and >4 mg/liter for tobramycin according to EUCAST guidelines (52).

SCTK.

To characterize the effect of different meropenem and tobramycin concentrations in monotherapy and in combination on PAO1 and PAOΔmutS strains, static-concentration time-kill experiments (SCTK) were performed (53, 54). Bacteria were grown on CAMHA at 36°C for ∼20 h, followed by overnight incubation in a shaking water bath at 36°C in sterile CAMHB. The optical density of the bacterial suspension was measured using a spectrophotometer, and the suspension was appropriately diluted to achieve the targeted initial inoculum of ∼10^7.8 CFU/ml in 15 ml. Antibiotic stock solutions were added at the initiation of the experiments (0 h) to achieve concentrations of 2, 8, and 16 mg/liter for meropenem and 1, 4, and 8 mg/liter for tobramycin, i.e., concentrations within the range of those observed in patients (55, 56). Bacterial suspensions were centrifuged (10 min, 3,220 × g, 36°C) and resuspended in prewarmed, antibiotic-containing sterile CAMHB (100% of the initial antibiotic concentration) every 24 h (53). Meropenem was additionally supplemented at 30% of the initial concentration at 6, 30, 54, and 78 h to compensate for its known thermal degradation (57).

Samples for total viable counting were collected at 0, 1, 3, 6, 24, 29, 48, 72, and 96 h and for quantification of resistant bacteria at 0, 24, 48, and 96 h. To minimize antibiotic carryover, samples were twice centrifuged at 4,000 × g for 5 min and resuspended in sterile saline (36°C) (42). For total viable counting, 100 μl of undiluted or appropriately diluted sample was manually plated onto CAMHA. Given the expected low number of resistant bacteria in samples from some treatments, 200 μl of sample was plated onto CAMHA containing 2.5 mg/liter meropenem or 5 mg/liter tobramycin to increase sensitivity. Antibiotic-free and meropenem-containing plates were incubated at 36°C for 48 h and tobramycin-containing plates for 72 h, followed by manual counting of colonies. The limit of counting was 1.0 log₁₀ CFU/ml for antibiotic-free (i.e., one colony per plate) and 0.7 log₁₀ CFU/ml for antibiotic-containing plates. The log₁₀(mutation frequency), i.e., log₁₀(MF), was determined by the following equation: log₁₀(MF) = log₁₀(CFU/ml on antibiotic-containing agar) − log₁₀(CFU/ml on antibiotic-free agar).

Microbiological responses were quantified using the log₁₀ change method, which compares the change in log₁₀(CFU/ml) from 0 h (CFU₀) to time t (1, 3, 6, 24, 29, 48, 72 and 96 h; CFU_t): log₁₀ change = log₁₀(CFU_t) − log₁₀(CFU₀). Synergy was defined as a reduction of ≥2 log₁₀ CFU/ml for the combination compared to the most active monotherapy component at the specified time (58).

MBM.

Simultaneous mechanism-based modeling (MBM) of SCTK data from both strains was performed using importance sampling (pmethod of 4) in S-ADAPT (v1.57) with S-ADAPT-TRAN (59–61). The S-ADAPT objective function value (−1× log likelihood), standard diagnostic plots, coefficients of correlation, biological plausibility of parameter estimates, and visual predictive checks were used for model evaluation. A life cycle growth model described bacterial growth and replication (62, 63). The final models included three subpopulations with different susceptibilities to meropenem and tobramycin and direct bacterial killing by both antibiotics. Subpopulation synergy and mechanistic synergy were incorporated to describe the effect of the combinations (54). Mechanistic synergy was described as tobramycin enhancing the target site penetration of meropenem via disrupting the bacterial outer membrane (53, 64, 65). The differential equations are available in the supplemental material. The models were then used in in silico simulations to predict the expected bacterial outcomes for clinically relevant monotherapy and combination regimens and the pharmacokinetics as observed in CF patients. Berkeley Madonna (v8.3.18) was used for all in silico simulations. We extended the MBM by simultaneously fitting the time courses of the total bacterial population and the populations less susceptible to meropenem or tobramycin (quantified on plates containing 2.5 mg/liter meropenem and 5 mg/liter tobramycin) for both strains, as previously described (49, 64).

Hollow-fiber in vitro infection model.

The free (non-protein-bound) meropenem and tobramycin concentration-time profiles were simulated based on pharmacokinetics reported from patients with CF (55, 56). The standard regimens were 1 g meropenem every 8 h as a 30-min infusion and 10 mg/kg tobramycin every 24 h as a 1-h infusion (Table 1); these doses correspond to those used clinically in CF patients (66). An optimized meropenem regimen of 3 g/day as continuous infusion, chosen based on MBM predictions, was also evaluated in monotherapy and in combination with tobramycin (Table 1). Based on in silico simulations, the meropenem and tobramycin concentration-time profiles, in monotherapy and combination therapy, were reproduced experimentally in the HFIM for each bacterial strain (simulated clearances and half-lives were the following: meropenem, 15.9 liters/h and 0.8 h; tobramycin, 4.9 liters/h and 2.5 h). The continuous infusion of meropenem was started at the steady-state concentration of ∼8 mg/liter. A growth control was included for each strain.

HFIM studies were conducted over 10 days as previously described (42, 67, 68), utilizing cellulosic cartridges (C3008-1; FiberCell Systems Inc., Frederick, MD, USA) in a humidified incubator at 36°C. For each strain, one colony was grown overnight in CAMHB at 36°C to prepare the bacterial stock solution. The optical density of the overnight culture was measured spectrophotometrically. A bacterial suspension was prepared to achieve the targeted initial inoculum of ∼10^8.4 CFU/ml, and 17 ml was injected into each HFIM cartridge. Samples for viable counting (1.0 to 1.5 ml) were collected at 0, 1.5, 5, 23, 26, 29, 47, 53, 71, 95, 119, 143, 167, 191, 215, and 239 h. Less susceptible bacteria were quantified on CAMHA containing meropenem at 2.5 and 5 mg/liter (2.5× and 5× MIC) and tobramycin at 2.5 mg/liter (5× MIC). The total and less susceptible bacteria and log₁₀(MF) were quantified per the SCTK methods described above. MICs were determined at 0, 95, 215, and/or 239 h by taking a subset of at least three colonies from antibiotic-free and antibiotic-containing plates.

Meropenem and tobramycin assays for pharmacokinetics.

Samples (1.0 ml) were collected in duplicate from the outflow of the HFIM central reservoir and immediately stored at −80°C until assayed. Meropenem and tobramycin in CAMHB were measured using validated liquid chromatography-tandem mass spectrometry (LC-MS/MS) assays. An Agilent 1200 high-performance liquid chromatograph (HPLC) coupled with an Agilent 6430 triple-quadrupole mass spectrometer equipped with a turbo ion electrospray ionization (ESI) source (Agilent Technologies, Santa Clara, CA) was used.

For measurement of tobramycin and meropenem, 50 μl of internal standard solution (10 μg/ml metformin in acetonitrile) was added to 50 μl of sample and vortex mixed for 30 s. Fifty microliters of acetonitrile and 100 μl of 0.05% trichloroacetic acid then were added, followed by vortex mixing for 1 min before centrifugation for 10 min at 15,000 × g. One hundred microliters of the supernatant was transferred to a polypropylene HPLC vial, diluted with 100 μl of distilled water, and vortex mixed for 30 s, and then 5 μl was injected onto a Synergi Polar-RP column (150 by 2.0 mm, 4.0 μm; Phenomenex, Torrance, CA, USA) using a binary gradient mobile phase composed of 0.25% formic acid (A) and acetonitrile (B), programmed as A:B at 80:20 (0 to 1 and 3.6 to 10 min) and A:B at 10:90 (1.1 to 3.5 min). The flow rate of the mobile phase was 0.2 ml/min (0 to 2.5 and 4.9 to 10.0 min) and 0.5 ml/min (2.6 to 4.8 min). The column oven temperature was 30°C, and total run time was 10 min. The ESI source was operated in positive mode, and the ionization gas temperature was set at 350°C. Mass transitions of the precursor/product ion pairs were monitored at m/z of 468.3→163.1, 384.2→141.0, and 130.1→71.0 for tobramycin, meropenem, and metformin, respectively. Mass spectrometric data were processed by the MassHunter Quantitative Analysis software (Agilent Technologies, Santa Clara, CA, USA). For both antibiotics the lower limit of quantification was 50 ng/ml and the correlation coefficient was >0.99. The interday precision was 2.3 to 6.9% for meropenem and 2.5 to 7.7% for tobramycin; interday accuracy was 95.9 to 100% for meropenem and 94.3 to 103.5% for tobramycin.