Quantifying adaptive resistance in bacteria using well-designed dynamic time-kill curve experiments
Niklas Hartung (1), Christoph Hethey (1), Eva B. Goebgen (2), Charlotte Kloft (2), Wilhelm Huisinga (1)
(1) University of Potsdam, Potsdam, Germany; (2) Dept. Clinical Pharmacy & Biochemistry, Institute of Pharmacy, Freie Universitaet Berlin, Germany
Objectives: In vitro time-kill curve (TKC) experiments allow to study the dose-effect relationship of antimicrobial treatments, either in a static (constant drug concentrations) or a dynamic setting (time-varying drug concentrations, usually to mimic PK in vivo). Mathematical models contribute to this understanding by offering the possibility to simulate microbial growth under clinically relevant conditions and under combination therapy. However, some dynamic processes such as adaptive resistance and persister formation (dormant state), both stress-triggered phenotypic changes, are difficult to quantify reliably from these experiments [1]. Here, we explore the use of a mathematical model for bacterial growth parametrized from static TKC data to design dynamic TKC experiments aiming not at mimicking typical PK, but at differentiating adaptive resistance and persister formation processes.
Methods: We used a previously developed cell-level bacterial population growth model under meropenem treatment [2]. Reflecting cell-level processes, adaptive resistance and persister formation are incorporated into this model. While persister formation could be parametrized from static TKCs on methicillin susceptible Staphylococcus aureus with a meropenem minimal inhibitory concentration (MIC) of 0.13 mg/L [3], assumptions had to be made on adaptive resistance parameters. We explored intermittent exposure designs (static exposure, then fast decrease to low concentrations, then again static exposure at a different concentration) to determine the unknown maximum adaptive resistance (MaxAR) and loss rate of adaptive resistance (LossAR). Within this class of designs, DS-optimal designs were determined [4]. All simulations were carried out in R software.
Results: Initial exposure at 10*MIC for 3 h, followed by a 7 h rest and subsequent low exposure close to MIC leads to good identifiability of LossAR for a range of assumed true parameters. In contrast, MaxAR was difficult to quantify reliably with any design even if bacterial load was very sensitive to the parameter, which is probably due to correlations with other model parameters such as persister formation.
Conclusions: Mathematical models parametrized from static TKC data can be leveraged for the design of dynamic TKC experiments. Model-based experimental design allow for a better characterization of bacterial resistance. As the next step, further scenarios for MaxAR quantification will be considered.
References:
[1] Jacobs M, Grégoire N, Couet W, Bulitta JB. Distinguishing Antimicrobial Models with Different Resistance Mechanisms via Population Pharmacodynamic Modeling. PLoS Comput Biol (2016) 10.1371/journal.pcbi.1004782.
[2] Hethey C, Wicha SG, Kloft C, Huisinga W. Impact of the intracellular ribosomal concentration on in vitro bacterial growth kinetics and the antibacterial effect of linezolid on S. aureus in time-kill assays. PAGE 25 (2016) Abstr
[3] Wicha SG, Kees MG, Kuss J, Kloft C. Pharmacodynamic and response surface analysis of linezolid or vancomycin combined with meropenem against Staphylococcus aureus. Pharmaceutical Research (2014), 32(7):2410-8.
[4] Studden WJ. Ds-Optimal Designs for Polynomial Regression Using Continued Fractions. The Annals of Statistics (1980) 8(5):1132-41.
