Jérémy Seurat (1) & Raphaëlle Delattre (1,2,3) – Jean Damien Ricard (1,4) – Laurent Debarbieux (2) – Jérémie Guedj (1)
(1) Université de Paris, IAME, INSERM, F-75018 Paris, France (2) Department of Microbiology, Institut Pasteur, Paris, France (3) AP-HP, Hôpital Beaujon, Service d'Anesthésie-Réanimation, Clichy, France (4) AP-HP, Hôpital Louis Mourier, Service de Réanimation Médico-Chirurgicale, Colombes, France
the first two authors equally contributed
Objectives:
In 2015, the WHO initiated a global action plan to combat the emergence of antimicrobial resistance, which could cause nearly 10 million deaths yearly by 2050 [1,2]. Bacteriophage (phage) therapy provides an attractive option to the increasing failure of antibiotics. Lytic phages are viruses that specifically target, infect, replicate and destroy bacteria [3]. Using a murine model of acute pneumonia we previously demonstrated that a dose of phages 536_P1 could prevent the death of animals infected by a lethal dose of the Escherichia coli strain 536, displaying an efficacy comparable to an antibiotic treatment [4]. To accelerate the transition towards human treatment, we aim to characterize quantitatively the dynamics of phage 536_P1 and their impact on E. coli 536 dynamics. The model of those two processes and their interaction is particularly complex. It should describe how fast phages may replicate once encountering bacteria and how efficient the immune response is in clearing bacteria, taking into account dosing regimen for phage administration. Mechanistic modeling has long been limited by the absence of longitudinal data due to destructive sampling and large within and inter-subject data variability associated with the immune response [5].
In this project, three in vitro and two in vivo studies, in mice, were conducted on the couple E. coli 536 / 536_P1. The aim is to develop a mechanistic model of the interaction phage-bacteria.
Methods:
In vitro studies were notably made to characterize the lytic cycle of phage 536_P1. Mainly, in the ‘one-step growth’ study, the phage concentration (in presence of E. coli 536) were measured longitudinally during one lytic cycle. It therefore allowed to estimate the burst size (i.e. the number of phages released after the infection by phage and the lysis of the bacteria) and the eclipse period (time for phages to produce new virions before lysing bacteria). In vivo studies involved infected and uninfected mice (N=215 in total). Some animals (N=32) received 108 phages intravenously (IV) or intratracheally (IT). Those uninfected animals were sacrificed at different times to measure phage concentration in various organs (including lungs), in order to characterize phage distribution and kinetics in lungs. Other animals (N=183) were infected by receiving an inoculum of 4.107 of E. coli 536. Two hours (h) after infection, those infected animals were treated by phages intravenously (IV) or intratracheally (IT). Various doses of phages were given, expressed as Multiplicity of infection (MOI, i.e. the dose of phage divided by the bacterial inoculum), varying between 0 (control animals) and 10. Importantly, bacteria strains were bioluminescent, allowing longitudinal measurement up to 2 days post infection expressed in equivalent bacterial load (CFUeq/g) [6]. Some animals were sacrificed at days 0.25, 0.5, 1 and 2 to quantify phage concentrations in lungs.
The mathematical model was built sequentially. First, in vitro data were fitted using transit compartments. Second, the kinetics of 536_P1 in lungs was described in uninfected mice. Then a model in infected mice was built which combined standard model of viral dynamics [7] with model of bacterial kinetics [5], using some parameter values of steps 1 and 2. The model consisted in a predator-prey system, composed of uninfected bacteria, which become infected by phages, then lysed, releasing new phages. It also includes the effect of the immune system on the bacteria. Parameter (fixed effects, standard error of random effects and residual errors) estimation was performed using SAEM in Monolix 2018R2. Model selection was based on physiology and BIC.
Simulations (N=1000) were performed with different bacterial inoculums and phage administration schedules to compare the dynamics of phages and bacteria. These simulations enabled comparisons of bacterial load at 24 or 48 h, notably between the two administration routes IT and IV and between different doses of phages (MOI from 0 to 1000).
Results:
The model described the interactions between phage 536_P1 and E. Coli 536, including the lytic cycle parameters of the phage and its kinetics in the lung compartment. The burst size of 536_P1 was estimated at 570 virus per bacteria and the median eclipse time was 34 minutes. In healthy animals, the maximal concentration of 536_P1 in lungs after an IV administration was reached in 4 h. The relative bioavailability (IV/IT) was 0.1% and phage elimination was monophasic, with a half-life of 12 h. In infected animals, the bacterial growth in lungs in absence of phages was characterized by a doubling time estimated at 3.4 h until reaching a plateau at 10.4 log10 CFUeq/g. An impact of the effective inoculum (i.e. the actual number of bacteria reaching the lung compartment) on the bacterial dynamics was observed in all infected mice. This was captured by accounting for the role of the innate immune system in controlling the infection: at low bacterial inoculum (e.g less than 106 CFUeq), the immune system (such as macrophages and neutrophils) could control the bacterial growth and clear bacteria with a half-life of 2.3 h. However, at high inoculum, bacteria increased exponentially until death of the mice within 2 days. The model showed that, after reaching a sufficient amount and in synergy with the immune system, phages induced a bacterial decay, even when bacterial load reached 8 log10 CFUeq/g. According to the model, median start of the bacterial decay ranged from 4 h (i.e. 2 h after treatment) for the best treatment group (MOI 10/IT) to 24 h for the less effective one (MOI 0.01/IV).
Simulations with a low inoculum of 4.106 CFUeq showed that phages increased bacterial decay: at 24 h, median bacterial load was 4.5 log10 CFUeq/g in 10/IT vs. 6.5 in control. When the bacterial inoculum was 4.107 CFUeq, the immunity was not sufficient in control mice: bacterial load at 48h was at the plateau for at least 95% of simulated individuals. In phage-treated groups (excepted 0.01/IV), the median bacterial load at 48h was lower than baseline. An IT administration at MOI 1 led to a lower bacterial load at 24 and 48h than an IV administration at MOI 100, and may therefore reduce the mortality.
Conclusion:
This work proposes the first detailed pharmacometric model of phage therapy in vivo. The data were generated and analyzed by a bacterial/viral dynamic model showing the synergism between the immune system and phage therapy in clearing bacteria. At low bacterial load, the immune system could clear bacteria with or without phage treatment. The model showed that at higher bacterial load, phages were instrumental to increase the clearance rate of bacteria and to bring the bacterial load to levels low enough such that it could be controlled by the immune system. The results of this work will help design future clinical studies and discuss relevant dosing regimens in humans.
References:
[1] The review on antimicrobial resistance, chaired by Jim O’Neill. Antimicrobial resistance: tackling a crisis for the health and wealth of nations. [Internet]. 2014 Dec. Available from: http://www.jpiamr.eu/wp-content/uploads/2014/12/AMR-Review-Paper-Tackling-a-crisis-for-the-health-and-wealth-of-nations_1-2.pdf
[2] Global action plan on antimicrobial resistance. Geneva: World Health Organization; 2015.
[3] Merril CR, Scholl D, Adhya SL. The prospect for bacteriophage therapy in Western medicine. Nat Rev Drug Discov. 2003;2:489–97.
[4] Dufour N, Debarbieux L, Fromentin M, Ricard J-D. Treatment of highly virulent extraintestinal pathogenic Escherichia coli pneumonia with bacteriophages. Crit Care Med. 2015;43:e190-198.
[5] Friberg LE, Guedj J. Acute bacterial or viral infection—What’s the difference? A perspective from PKPD modellers. Clin Microbiol Infect [Internet]. 2019 [cited 2020 Feb 25]; Available from: https://linkinghub.elsevier.com/retrieve/pii/S1198743X19306639
[6] Debarbieux L, Leduc D, Maura D, Morello E, Criscuolo A, Grossi O, et al. Bacteriophages can treat and prevent Pseudomonas aeruginosa lung infections. J Infect Dis. 2010;201:1096–104.
[7] Perelson AS, Guedj J. Modelling hepatitis C therapy—predicting effects of treatment. Nat Rev Gastroenterol Hepatol. 2015;12:437–45.
Reference: PAGE () Abstr 9537 [www.page-meeting.org/?abstract=9537]
Poster: Oral: Drug/Disease Modelling