Combination of in vivo phage therapy data with in silico model highlights key parameters for treatment efficacy
Raphaëlle Delattre (1,2,3)*, Jérémy Seurat (2)*, Feyrouz Haddad (1,2), Thu-Thuy Nguyen (2), Baptiste Gaborieau (1,2,3), Rokhaya Kane (1), Nicolas Dufour (4), Jean-Damien Ricard (2), Jérémie Guedj (2) and Laurent Debarbieux (1)
(1) Institut Pasteur, Bacteriophage, Bacterium, Host Laboratory, Paris F-75015 France (2) Université de Paris, INSERM, IAME, Paris F-75018, France (3) Université de Paris, Cellule Pasteur, Paris F-75013, France (4) Centre Hospitalier René Dubos, Cergy Pontoise F-95503, France
Objectives: Bacteriophage (phage) therapy provides an attractive option to the increasing failure of antibiotics treatments. Phages are viruses that specifically target, infect, replicate and destroy bacteria [1]. A major milestone for their widespread use in human is the determination of the optimal administration scheme, which is particularly complex for phages, as standard PK/PD models are not adapted to phages. From a series of specific experiments in a murine model of acute pneumonia [2], we built a semi-mechanistic model that can be used to understand the tripartite phage-bacteria-host interactions in vivo.
Methods: The model was built sequentially. First, some parameters of the lytic cycle of phage 536_P1 infecting the Escherichia coli strain 536 have been determined in vitro. Next, in vivo studies involved infected and uninfected mice (N=241 in total) to collect data characterizing phage-bacteria-host interactions, animals being treated with different treatment doses and administration routes. Uninfected animals received 108 phages intravenously (IV, N=37) or intratracheally (IT, N=16) and were sacrificed at different times to measure phage concentration in lungs to characterize phage PK in lungs. Then, infected animals (N=188) by an inoculum of 4.107 CFU of E. coli 536 were treated by phages IV or IT at various doses of phages (given in PFU): 0 (control animals, to identify the bacterial growth), 4.105 (Lo), 4.107 (Med) or 4.108 PFU (Hi). Importantly, bacteria strains were bioluminescent, allowing longitudinal measurement of light converted in equivalent bacterial load (CFUeq/g of lung) [3]. From infected mice, a new model was built, combining a model of viral dynamics [4] with a model of bacterial kinetics [5]. This model consisted in a predator-prey ODE system. It was composed of bacteria susceptible to phage, which become infected by phages and latent, then produce new phages at the time of their lysis. The model also includes the effect of the immunity on the bacteria, as well as refractory bacteria to phage (either phage-resistant or phage-inaccessible). Finally, we used a joint model to link the bacterial load and the survival probability of mice, which accounted for the potential bias in longitudinal parameters [6]. Parameters of the phage/bacteria model were estimated using SAEM in Monolix 2018R2.
Results: Our new model described the interactions between phage 536_P1 and strain E. coli 536, including the lytic cycle parameters of the phage and its kinetics in the lung compartment. In uninfected animals, the maximal concentration of 536_P1 in lungs after an IV administration was reached in 4 h. The relative bioavailability (IV/IT) was 0.02%. In infected animals, the bacterial growth in lungs in absence of phages was characterized by a doubling time estimated at 3 h until reaching a plateau at 10.4 log10 CFUeq/g. An impact of the effective inoculum (i.e. the initial bacterial load reaching the lung compartment) on the bacterial dynamics was observed. Indeed, in absence of phages, when the inoculum was greater than 6.8 log10 CFUeq/g, the bacterial load increased continuously, while, below 6.8 log10 CFUeq/g, the bacteria were cleared by the immune system with an elimination half-life of 2 h.
In infected and treated animals, the model showed that, in synergy with the immune system, the benefit of phage therapy was visible mostly for animals having an intermediate effective inoculum size (between 6 and 8 log10 CFUeq/g). Although phages could induce a bacterial decay at the individual level when bacterial load reached 8 log10 CFUeq/g, they were not sufficient to cure the animals in median. With an intermediate bacterial inoculum (7 log10 CFUeq/g), the IV route had a lower efficiency to stop bacterial growth, with a time to peak bacterial load that could be as large as 4.0 h (IV Lo) compared to less than 1.1 h with IT route (IT Lo), and as low as 0.3 h in case of IT Hi.
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. The model developed during this work could be used to predict the efficacy of virtually any phage for which a minimum set of in vitro and in vivo data must be obtained, which will considerable lower the number of experiments needed to validate such a phage during pre-clinical development.
*: the first two authors equally contributed
References:
[1] Merril CR, Scholl D, Adhya SL. The prospect for bacteriophage therapy in Western medicine. Nat Rev Drug Discov. 2003;2:489–97.
[2] 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.
[3] 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.
[4] Perelson AS, Guedj J. Modelling hepatitis C therapy—predicting effects of treatment. Nat Rev Gastroenterol Hepatol. 2015;12:437–45.
[5] Friberg LE, Guedj J. Acute bacterial or viral infection-What’s the difference? A perspective from PKPD modellers. Clin Microbiol Infect Off Publ Eur Soc Clin Microbiol Infect Dis. 2020;26:1133–6.
[6] Desmée S, Mentré F, Veyrat-Follet C, Guedj J. Nonlinear mixed-effect models for prostate-specific antigen kinetics and link with survival in the context of metastatic prostate cancer: a comparison by simulation of two-stage and joint approaches. AAPS J. 2015;17:691–9.