Impact of Two-Component Regulatory Systems PhoP-PhoQ and PmrA-PmrB on Colistin Pharmacodynamics in Pseudomonas aeruginosa

Abstract

The in vitro pharmacodynamics of colistin against Pseudomonas aeruginosa PAO1 wild-type and isogenic knockout strains of phoP and pmrA were evaluated. Colistin killing at subinhibitory concentrations was greater against the phoP and pmrA mutants than the wild type within the first 8 h: the concentration that results in 50% of maximal effect (EC₅₀) of the pmrA mutant (0.413 mg/liter) was less than that of the wild type (0.718 mg/liter) (P < 0.05). An in vitro pharmacodynamic model simulating human colistin regimens displayed initial killing followed by regrowth in the phoP mutant and gradual regrowth in the pmrA mutant and wild type.

TEXT

Pseudomonas aeruginosa is a Gram-negative opportunistic pathogen which has developed resistance to nearly all currently available antibiotics. However, many of the multidrug-resistant (MDR) P. aeruginosa strains remain susceptible to colistin (or polymyxin E) (11, 16). Although colistin was first introduced in 1959 for the treatment of severe Gram-negative bacterial infections, there is little information on how to optimally dose this antimicrobial agent. This has resulted in suboptimal dosing whereby the increased usage of colistin monotherapy has resulted in increasing colistin heteroresistance (2, 4, 14, 17, 19).

Although the complete mechanism of colistin heteroresistance has not been fully elucidated, it has been suggested that alterations in lipopolysaccharide (LPS) structure, loss of LPS, and global regulatory systems may play a role (7, 8, 10, 18, 20). Colistin exerts its bactericidal antimicrobial activity by binding to the LPS, a major component of the Gram-negative cell surface, through interactions with phosphates and fatty acids of the LPS core and lipid A moieties in the outer membrane (24). Earlier studies have suggested that colistin resistance is adaptive and regulated by two separate two-component regulatory genes, phoP-phoQ (phoPQ) and pmrA-pmrB (pmrAB) (7, 8, 18, 20). The PhoPQ system is a global regulatory system that autoregulates the oprH-phoP-phoQ operon under divalent cation-limiting conditions and in the presence of polyamines, which induce resistance to cationic peptides such as colistin (8, 13, 26). pmrAB directly controls the pmrHIJKLM operon, whose gene products are responsible for the synthesis of N₄-aminoarabinose, which binds to lipid A moieties and neutralizes the negatively charged phospholipids, causing a change in the negatively charged cell membrane which leads to colistin resistance (23, 28, 30). There exists a paucity of studies that define the impact of these two-component regulatory systems on colistin pharmacodynamics at clinically relevant concentrations. Here, we compared the pharmacodynamics of P. aeruginosa PAO1 with those of isogenic strains with mutations in phoP and pmrA to determine the contributions of each of these regulatory systems to the rate and extent of killing by colistin at clinically relevant concentrations using time-kill experiments and an in vitro pharmacodynamic model.

The genetically characterized P. aeruginosa strain PAO1 (strain H103) and its respective isogenic knockout isolates in phoP (strain H851), phoP::xylE-aacC1;Gm^r, and pmrA (strain H988), pmrA::xylE-aacC1;Gm^r, were utilized in this study. The construction of these mutants and their characteristics were previously described (27). All bacterial strains were obtained as a gift from R. E. W. Hancock at the University of British Columbia, Vancouver, Canada. Colistin sulfate was obtained from Sigma-Aldrich (St. Louis, MO). Fresh stock solutions were prepared before each experiment. Luria-Bertani (LB) broth was used as the growth medium for all experiments and was supplemented with 25 mg/liter of calcium and 12.5 mg/liter of magnesium. Susceptibility testing was completed based on the modified Clinical and Laboratory Standards Institute (CLSI) guidelines (3) utilizing cation-adjusted LB broth. Time-kill experiments were completed as previously described (29). In brief, fresh bacterial colonies were grown overnight prior to each experiment. The target initial inocula were 10⁶ and 10⁸ CFU/ml. The following concentrations were evaluated for colistin: 0, 0.5, 1, 2, 4, 8, 16, 32, 64, 128, and 256 mg/liter. Serial samples were collected at 0, 0.25, 0.5, 1, 2, 4, 8, 12, and 24 h.

To further evaluate the findings from the time-kill experiments, an in vitro one-compartment model (IVPM) study was utilized to determine the killing activity of human dosing regimens of colistin against P. aeruginosa as previously described (9). The initial inoculum was ∼10⁸ CFU/ml, and bacteria were allowed to equilibrate for approximately 30 min before dosing. Serial samples were taken at 0 (predose), 0.5, 1, 2, 4, 8, 12, 13, 14, 24, 26, 28, 32, 36, 37, 38, and 48 h to assess viable counts. The simulated regimens used free (ƒ) maximum concentrations of the drug in serum (C_maxs) of 1.2 mg/liter every 12 h, 2 mg/liter every 12 h, 4.5 mg/liter every 12 h, and 14.6 mg/liter every 12 h (concentrations refer to the colistin base). A half-life of 4.6 h was utilized for all in vitro model experiments based on the mean colistin half-life in cystic fibrosis patients (15). The log ratio area method was utilized as the pharmacokinetic/pharmacodynamic measurement (29) and fit by a Hill pharmacodynamic model using the linear trapezoidal rule and nonlinear regression using WinNonlin (version 5.3). Model parameters were estimated to describe the concentration-effect relationship over the 10 studied colistin concentrations for each strain. The model fitting yielded the standard errors and confidence intervals for each parameter and strain. These parameter estimates were then compared between the wild type and the mutants using a Student t test (2-tailed) with an alpha of 0.05.

The colistin MICs to all strains were 2.0 mg/liter. The bacterial killing activities of colistin against all strains at two different initial inocula are presented in Fig. 1. Colistin displayed concentration-dependent pharmacodynamics with concentrations of >8 mg/liter, resulting in sustained bactericidal activity until 24 h (Fig. 1A to C). Against a low inoculum of phoP and pmrA mutant strains, the overall rate and extent of killing were similar to those against the wild type; however, there were differences in the early bactericidal activity profiles for colistin at sub-MICs of 0.5 mg/liter when comparing isolates. For example, at 1 h, 0.5 mg/liter of colistin resulted in stasis (+0.02 log₁₀ CFU/ml) against the wild type while rapidly reducing bacterial counts by 2.20 and 2.96 log₁₀ CFU/ml against pmrA and phoP mutant strains, respectively (Fig. 1A to C). Against a high initial inoculum of 10⁸ CFU/ml of the wild type, colistin activity was attenuated at earlier time points, with concentrations of ≤4 mg/liter resulting in diminished activity compared to that with the low inoculum. There were no significant differences at the high inoculum when comparing the killing profiles of colistin against phoP and pmrA mutants to that against the wild type (Fig. 1D to F).

Fig 1.

Time-kill experiments of colistin versus P. aeruginosa PAO1 isogenic strains—the wild type (A and D), a pmrA mutant (B and E), and a phoP mutant (C and F)—at two different initial inocula of 10⁶ CFU/ml (top) and 10⁸ CFU/ml (bottom).

The pharmacodynamic relationship characterizing the concentration-response relationship using the log area ratio is shown in Fig. 2. Overall, there were significant differences for colistin killing activity against the mutant strains within the first 8 h at the low inoculum (Fig. 2A). With this inoculum level, pmrA and phoP mutant strains were more sensitive to colistin than was the wild type, as supported by the model-fitted parameter estimates. The concentration that results in 50% of maximal effect (EC₅₀) of the pmrA mutant (0.413 mg/liter; relative standard error [SE], 5.15%) was significantly (P < 0.05) lower than that in the wild type (0.718 mg/liter; SE of 10.7%), while there were no statistically significant differences in maximal effect (E_max). There was a similar trend for colistin against the phoP mutant, as the EC₅₀ of the phoP mutant (0.432 mg/liter; SE of 19.2%) was lower than that of the wild type (P > 0.05). However, for the log area ratios calculated over 24 h, there were no significant differences among the three strains, as supported by the parameter estimates for all three strains and the overlapping of the concentration-effect curves (Fig. 2B). With the high inoculum level, there were no significant differences between the three isogenic strains (Fig. 2C).

Fig 2.

Dose-response relationships for colistin profiling the initial bacterial response over the first 8 h at 10⁶ CFU/ml (A), the bacterial response over 24 h at 10⁶ CFU/ml (B), and the bacterial response over 24 h at 10⁸ CFU/ml (C).

To further investigate the impact of pmrA and phoP mutation on colistin pharmacodynamics, we conducted IVPM experiments with 4 regimens of human dose-simulated pharmacokinetics. For the wild type, the maximum bacterial reductions for regimens with free (ƒ) C_maxs of 1.2 mg/liter, 2 mg/liter, 4.5 mg/liter, and 14.6 mg/liter every 12 h were 2.24, 2.36, 3.62, and 4.70 log₁₀ CFU/ml, respectively. Interestingly, there was greater maximal killing in mutant strains than in the wild type (Fig. 3): the pmrA mutant displayed 1.97, 3.04, 4.38, and 5.51 log₁₀ CFU/ml maximal decreases for the four regimens, respectively, while the phoP mutant displayed 2.22, 3.57, 5.14, and 6.30 log₁₀ CFU/ml maximal decreases, respectively. Compared to the pmrA mutant, the phoP mutant was more sensitive to subsequent dosing every 12 h, as presented by a “seesaw” pattern after each dose. For example, after the first redose at 12 h, there was a rapid bacterial reduction of 2.02 log₁₀ CFU/ml in the phoP mutant, while the pmrA mutant displayed a 0.312 log₁₀ CFU/ml reduction after exposure to a colistin ƒC_max of 14.6 mg/liter every 12 h. This was a trend that continued every 12 h throughout the course of dosing for the study duration.

Fig 3.

In vitro one-compartment pharmacodynamic model simulating humanized colistin regimens versus P. aeruginosa PAO1 isogenic strains—the wild type (A), a pmrA mutant (B), and a phoP mutant (C)—at 10⁸ CFU/ml.

Although the activation of the two-component regulatory systems PhoPQ and PmrAB in P. aeruginosa has been associated with resistance to cationic antimicrobial peptides, including colistin, there is a paucity of data regarding the effect that these mutations have on bacterial dynamics during the natural course of colistin dosing. We determined that colistin against pmrA and phoP mutants displayed an increased killing activity compared to that against the wild type at subinhibitory colistin concentrations. This translated to a statistically significant difference in EC₅₀ between the wild type and the pmrA mutant. This observation is in line with that seen in studies by other investigators. McPhee et al. determined that exposure to subinhibitory concentrations of polymyxin B caused the induction of pmrB genes, which lead to alteration in LPS structure, resulting in decreased susceptibility to colistin or cationic antibiotics (20). Adams et al. observed a similar finding in Acinetobacter baumannii, whereby pmrA expression is constitutively high in colistin-resistant mutants (1). Additionally, the development of colistin resistance through the acquisition of mutations has been reported to occur in clinical isolates. Recently, Miller et al. (21) and Moskowitz et al. (22) observed gain-of-function pmrB and loss-of-function phoQ genes in colistin-resistant P. aeruginosa strains isolated from cystic fibrosis patients. Our results support the theory that the induction of the two-component regulatory systems PhoPQ and PmrAB is due to exposure to subinhibitory colistin concentrations.

Interestingly, the differential effects of killing against the pmrA or phoP mutant were not evident under conditions in which colistin concentrations were above the MIC or at high bacterial densities in time-kill experiments. However, at concentrations below the MIC, near the EC₅₀, colistin achieved greater killing against the pmrA and phoP mutant strains than against the wild type. This finding shows that intrinsic genetic differences in these strains may play a greater role in the alteration of response than in the adaptation of P. aeruginosa in the development of resistance to colistin. Additionally, this may suggest that these regulatory systems confer low-level tolerance and that tolerance to higher colistin concentrations is not mediated by a single signature mutation but is rather influenced by a number of regulator systems, environmental factors, and genes (12, 25). For example, Fernandez et al. suggested that another two-component regulatory system, ParR-ParS, is also responsible for cationic peptide adaptive resistance (5). Furthermore, as recently suggested by Henry et al. (10), other mechanisms involving mutational inactivation of genes essential for lipid A biosynthesis that result in subsequent total loss of LPS may play a greater role in the development of colistin tolerance and resistance. These data may also support the notion that both pmrAB and phoPQ systems played a role in colistin pharmacodynamics. However, such differences are likely to depend on the microenvironments in which bacteria are exposed to antimicrobials (31). It is important to note a potential limitation of this study. For the in vitro one-compartment model, the simulated pharmacokinetic profiles were based on the pharmacokinetics of colistin in cystic fibrosis patients (23) who showed a fast conversion from colistin methanesulfonate (CMS) to colistin and a shorter colistin half-life than the apparent colistin half-life seen in critically ill patients (6). Therefore, care should be taken when extrapolating our in vitro results to the critically ill patient population. Additional studies in animal models are warranted to translate the impact of the two-component regulatory systems PhoPQ and PmrAB on the pharmacodynamics of polymyxins in humans.