Salma Bahnasawy1, A. Leoni Swart2, Rusudan Okujava3, Caterina Bissantz4, Neil Parrott4, Lena Friberg1, Urs Jenal2, Elisabet Nielsen1
1Department of Pharmacy, Uppsala University, 2Biozentrum, University of Basel, 3Roche Pharma Research and Early Development, Infectious Diseases Therapeutic Area, Roche Innovation Center Basel, F. Hoffmann-La Roche Ltd, 4Roche Pharma Research and Early Development, Pharmaceutical Sciences, Roche Innovation Center Basel, F. Hoffmann-La Roche Ltd
Background: Traditional in vitro time-kill curve (TKC) experiments, combined with semi-mechanistic PKPD modelling, provide valuable insights into antibiotic efficacy and in vivo predictions. [1,2] However, these setups do not fully replicate the complexity of in vivo environments, particularly in lung infections where factors like tissue penetration, mucus thickness, pH, and nutrient availability influence drug distribution and bacterial dynamics. [3,4] Additionally, bacterial-induced tissue disruption may further impact drug penetration. Recently, Swart et al. developed an in vitro human lower airway transwell tissue infection model, offering a physiologically relevant setup for TKC experiments. [5] The present study aimed to evaluate the lung-mimicking transwell model for antibiotic PKPD assessment, comparing it to conventional broth-based TKC experiments. Methods: TKC experiments were conducted in broth and lung tissue models to assess the antibacterial effects of levofloxacin and meropenem against Pseudomonas aeruginosa (PAO1, MIC = 0.5 µg/mL). In the broth model, bacterial kill kinetics were assessed over 24 hours across antibiotic concentrations (0.25- to 64-fold MIC) at two starting inocula (~5 × 10³ and ~5 × 106 CFU/mL). In the lung tissue model, tissues were infected apically (~5 × 104 CFU/tissue insert), and antibiotics (0.5- to 32-fold MIC) were added basally two hours post-infection. Bacterial counts were quantified from tissue homogenates and the basal compartment over 24 hours. Permeability assays characterised drug penetration across infected tissue at two basal concentrations per drug (levofloxacin: 1- and 32-fold MIC; meropenem: 2- and 32-fold MIC), with drug concentrations measured in apical mucus and the basal compartment. Bacterial and drug concentration data were used to develop PKPD models in NONMEM 7.5 via Finch Studio using PsN. Simulations of dose fractionation studies were conducted to evaluate differences in PKPD indices and efficacy between models, incorporating published human epithelial lining fluid PK models. Results: A total of 326 levofloxacin and 341 meropenem bacterial count observations were collected in broth TKC experiments, while 240 levofloxacin and 233 meropenem counts were obtained in tissue experiments. Levofloxacin exhibited a faster bacterial killing rate than meropenem in both models. Bacterial regrowth was observed for both drugs at 24 h in broth and tissue. At high basal drug concentration, meropenem exhibited a low tissue penetration ratio (basal-to-apical concentration) of approximately 0.26. However, this ratio increased to 1 at low basal drug concentration, where cell breaching was more likely. In contrast, levofloxacin maintained a consistently high tissue penetration ratio of 3.5 across all tested concentrations. Bacteria were modelled as either a growing drug-susceptible or a resting insusceptible population, with regrowth driven by a pre-existing resistant subpopulation. Bacterial growth was slower in tissue than broth (kg = 0.794 vs. 1.39 h?¹). The lung tissue PKPD model accounted for bacterial invasion from the apical to basal compartment, using a cumulative hazard function driven by the area under the bacterial growth curve on the apical side. Additionally, incorporating the additive effect of the cumulative hazard of tissue invasion on increased drug transfer (basal to apical) significantly improved the model fit for meropenem (?OFV = -89.9), but not for levofloxacin (?OFV = 0). A sigmoidal Emax model described drug effects, with lower Emax in tissue than broth (levofloxacin: 4.28 vs. 100 h?¹; meropenem: 1.79 vs. 3.27 h?¹). Pre-existing resistant bacteria had the same growth rate as susceptible ones but lower drug-killing capacity (lower Emax). Simulations revealed that the PKPD index best correlating with efficacy remained consistent between models: fAUC/MIC for levofloxacin and %fT>MIC for meropenem. However, for clinically recommended pneumonia dosing, the tissue model predicted 1.5 log kill/tissue insert, whereas the broth model predicted 3.5 log kill/mL. Unlike broth-based predictions, bacterial regrowth was observed in all dosing scenarios in the tissue model. Conclusion: This study highlights significant PKPD differences between lung tissue and broth-based TKC models for levofloxacin and meropenem. Further research should explore additional pathogen-drug combinations and assess how well the tissue model reflects in vivo bacterial killing to improve PKPD characterisation and optimise dose selection.
[1] Khan DD, Friberg LE, Nielsen EI. A pharmacokinetic–pharmacodynamic (PKPD) model based on in vitro time–kill data predicts the in vivo PK/PD index of colistin. J Antimicrob Chemother 2016;71:1881–4. https://doi.org/10.1093/jac/dkw057. [2] Sy SKB, Zhuang L, Xia H, Schuck VJ, Nichols WW, Derendorf H. A model-based analysis of pharmacokinetic–pharmacodynamic (PK/PD) indices of avibactam against Pseudomonas aeruginosa. Clin Microbiol Infect 2019;25:904.e9-904.e16. https://doi.org/10.1016/j.cmi.2018.10.014. [3] Satta G, Cornaglia G, Foddis G, Pompei R. Evaluation of ceftriaxone and other antibiotics against Escherichia coli, Pseudomonas aeruginosa, and Streptococcus pneumoniae under in vitro conditions simulating those of serious infections. Antimicrob Agents Chemother 1988;32:552–60. [4] van Os W, Zeitlinger M. Predicting Antimicrobial Activity at the Target Site: Pharmacokinetic/Pharmacodynamic Indices versus Time-Kill Approaches. Antibiotics 2021;10:1485. https://doi.org/10.3390/antibiotics10121485. [5] Leoni Swart A, Laventie B-J, Sütterlin R, Junne T, Lauer L, Manfredi P, et al. Pseudomonas aeruginosa breaches respiratory epithelia through goblet cell invasion in a microtissue model. Nat Microbiol 2024:1–13. https://doi.org/10.1038/s41564-024-01718-6.
Reference: PAGE 33 (2025) Abstr 11330 [www.page-meeting.org/?abstract=11330]
Poster: Drug/Disease Modelling - Other Topics