Model-based rationale for the in vitro characterisation of antimicrobial combinations in Buruli Ulcer
Umberto Villani (1,2), Salvatore D’Agate (1,2), Emma Sáez López (3), Santiago Ramon Garcia(3,4), Oscar Della Pasqua(1,2)
(1) Clinical Pharmacology & Therapeutics Group, University College London, London, United Kingdom; (2) Istituto per le Applicazioni del Calcolo, National Research Council, Rome, Italy; (3) Department of Microbiology, Faculty of Medicine, University of Zaragoza, Spain; (4) Research and Development Agency of Aragón (ARAID) Foundation, Zaragoza, Spain
Objectives: Buruli Ulcer (BU) is a chronic skin and soft tissue disease that causes significant disability, particularly among young people in tropical areas. Today, the standard-of-care (SoC) treatment for BU is a 8-week oral antimicrobial regimen of rifampicin (RIF) and clarithromycin (CLA) [1]. Although well-tolerated and effective, this therapy can still cause relevant side effects. Recently, there has been a strong interest in adding amoxicillin-clavulanate (AMX-CLV) to the SoC to reduce treatment duration, driven by the promising in vitro efficacy results of this four-drug combination [2]. However, establishing the adequate treatment duration for achieving clinical cure through empirical protocols can be very inefficient and costly approach. Thus, here we describe the development of a drug-disease model to quantify the contribution of each antibiotic in the overall regimen, expanding the modelling framework recently proposed for TB [3]. This approach aims to rank antimicrobial regimens in vitro and assess the probability of success of selected regimens in the clinic.
Methods: In vitro time-kill assays of several clinical isolates of M. ulcerans were available for this analysis. [4] The time course of bacterial growth and killing was studied over a period of 35 days in three different conditions: No treatment, drug monotherapy at various nominal concentrations, and 2-,3-,4- drug combination therapy. The selection of concentration ranges to test for all compounds was based on a previous static microbiological study [2]. Due to the long incubation times required to obtain observable colonies in vitro, light production by luciferase enzymes was used as a real-time marker for bacterial viability [5].
The pharmacodynamic (PD) data analysis comprised three stages: (1) development of a model of natural bacterial growth, (2) parametrisation of the effect of single drugs on M. ulcerans and (3) estimation of the effect of drugs in combination. Regarding the natural bacterial growth, different structural alternatives were tested (e.g. Gompertz, Velhurst and the delay logistic growth model). The effect of single antimicrobials was investigated as a bactericidal and/or bacteriostatic effect with linear or sigmoidal EMAX models. Drug concentrations within the culture medium were parametrised with an exponential kinetic function. The overall antimicrobial activity of a combination was characterised by a shift in the potency and/or maximum killing of the backbone drug (i.e. RIF), similarly to a standard covariate analysis.
Model selection was based on goodness-of-fit plots, changes in the objective function value (OFV), the biological plausibility of estimates and their precision. Parameter estimation was conducted using NONMEM 7.5 and PsN 5.2.6. Graphical analyses were performed in R version 4.1.2.
Results: The delay logistic growth was found to be the most suitable model to parametrize the natural growth of M. ulcerans, with distinct estimates for each clinical isolate (median:0.141, range: 0.116–0.185 [day-1]). The effect of single drugs on bacterial growth was best described by a joint inhibitory effect and a time-delayed killing effect, both implemented as a sigmoidal EMAX function. The effect of companion drugs as shifts in potency of the backbone drug provided an adequate description of the time kill curves. In general, the four-drug combination (RIF, CLA, and AMX-CLV) was found to consistently feature the lowest apparent potency (aEC50) across the tested combinations (geometric mean=0.95, range=0.017–12 [μg/L]). Significant inter-strain variability was also identified in the aEC50 of RIF-CLA (gmean=12, range=0.5-169 [μg/L]) and RIF-AMX-CLV (gmean=1, range=0.017–20 [μg/L]) combinations.
Conclusions: The proposed framework allows to systematically rank antimicrobial combinations in vitro for the treatment of BU in a pragmatic, yet semi-mechanistic fashion. Notably, the delay logistic growth parametrisation allowed the characterisation of complex bacterial growth profiles. We were able to disentangle the contribution of AMX and CLA to the overall antimicrobial activity of the combination and show that the apparent synergistic effect depends on the clinical isolate. These results will provide the basis for further evaluation of suitable dosing regimens and minimum treatment time for bacterial eradication in patients.
References:
[1] WHO, “Buruli Ulcer.” http://www.who.int/mediacentre/factsheets/fs199/en/ (accessed Mar. 06, 2023).
[2] M. P. Arenaz-Callao, R. González del Río, A. Lucía Quintana, C. J. Thompson, A. Mendoza-Losana, and S. Ramón-García, “Triple oral beta-lactam containing therapy for Buruli ulcer treatment shortening,” PLoS Negl Trop Dis, vol. 13, no. 1, 2019, doi: 10.1371/journal.pntd.0007126.
[3] M. Muliaditan and O. della Pasqua, “Evaluation of pharmacokinetic-pharmacodynamic relationships and selection of drug combinations for tuberculosis,” Br J Clin Pharmacol, vol. 87, no. 1, 2021, doi: 10.1111/bcp.14371.
[4] E. Sáez López, A. Cris Millán Placer, A. Lucía Quintana, and S. Ramón García, “In vitro time-kill assays of amoxicillin/clavulanate in combination with rifampicin/clarithromycin against mycobacterium ulcerans”, Abstract no. 1767, Abstract book of the 2022 annual meeting of the American Society of Tropical Medicine & Hygiene.
[5] T. Zhang, S. Y. Li, P. J. Converse, J. H. Grosset, and E. L. Nuermberger, “Rapid, Serial, Non-invasive Assessment of Drug Efficacy in Mice with Autoluminescent Mycobacterium ulcerans Infection,” PLoS Negl Trop Dis, vol. 7, no. 12, 2013, doi: 10.1371/journal.pntd.0002598.