Sophia Zborowsky (1)*, Jérémy Seurat (2,3)*, Quentin Balacheff (1,4), Chau Nguyen Ngoc Minh (1,5), Marie Titécat (6), Emma Evrard (1), Rogelio A. Rodriguez-Gonzalez (2,7), Jacopo Marchi (2,8), Joshua S. Weitz (2,3,8)# , Laurent Debarbieux (1)#
(1) Institut Pasteur, Université Paris Cité, Bacteriophage Bacterium Host, Paris 75015, France (2) School of Biological Sciences, Georgia Institute of Technology, Atlanta GA 30332, USA (3) Institut de Biologie, Ecole Normale Supérieure, Paris 75005, France (4) CHU Felix Guyon, Service des maladies respiratoires, La Réunion, France (5) Sorbonne Université, Collège Doctoral, Paris, France (6) Université de Lille, INSERM, CHU Lille, U1286-INFINITE-Institute for Translational Research in Inflammation, Lille 59000, France (7) Interdisciplinary Graduate Program in Quantitative Biosciences, Georgia Institute of Technology, Atlanta GA 30332, USA (8) Department of Biology, University of Maryland, College Park MD 20742, USA *: equal contribution #: corresponding authors
Objectives:
The rise of antimicrobial resistance has led to renewed interest in evaluating the efficacy of phage therapy [1]. In murine models highly effective treatment of acute pneumonia caused by Pseudomonas aeruginosa relies on the synergistic antibacterial activity of bacteriophages with neutrophils [2]. However, it remains unclear the importance of other components of the innate immune system. Here, we sought to assess the role and interactions of alveolar macrophages (AM) in the clearance of P. aeruginosa during acute pneumonia and its treatment by phage. We combined an in vivo and mathematical modeling approach to conceptualize and experimentally test the interactions between bacteria, phage, AM and neutrophils.
Methods:
In addition to control groups, mice were initially AM-depleted by clodronate, then infected by bioluminescent P. aeruginosa strain PAK [3] and treated with phage PAK_P1 at multiplicity of infection (MOI) 10, i.e. 10 times more phages than bacterial inoculum. Bioluminescent bacterial strains allowed longitudinal measurement of light converted in equivalent bacterial load in lungs. Moreover, blood neutrophils, blood monocytes, as well as lung neutrophils, AM, bacterial load and phage concentrations have been collected. The model was built following a sequential approach to estimate the key parameters of phage-immune cells-bacteria interactions. Inference procedures were performed in a non-linear mixed effect model framework using the SAEM algorithm implemented in Monolix software version 2021R2 [4]. The model selection was based on corrected Bayesian information criterion [5]. The final model was evaluated from goodness of fits as well as NPDE [6]. Simulations from the model have been performed with R using the RsSimulx package (version 2.0.2).
Results:
Here, we show that without bacteriophage, depletion of alveolar macrophages (AM) did not result in boosting the P. aeruginosa load in the lungs. Unexpectedly, upon phage treatment, although infection control was better than in untreated mice, pulmonary levels of P. aeruginosa were significantly lower in AM-depleted than in immunocompetent mice. To explore potential mechanisms underlying the benefit of AM-depletion in treated mice, we developed a mathematical model. Integration of model simulations suggests that AM reduce bacteriophage density in the lungs. From uninfected mice, we experimentally confirmed that the in vivo decay of phage is faster in immunocompetent compared to AM-depleted animals: 0.069 h-1 vs. 0.023 h‑1, respectively.
The final mathematical model consisted in an ordinary differential equation system describing the evolution of phage, bacteria (sensitive or resistant), neutrophils, AM and clodronate in the lungs, as well as blood neutrophils and monocytes. It included key interactions such as the recruitment of immune cells after infection, the neutrophils capacity to kill bacteria, the phage-bacteria lytic cycle and the phage phagocytosis by AM. In addition to this phagocytosis phenomenon, the differences observed between immunocompetent and AM-depleted for both bacterial load and phage density have enabled our model to identify the presence of another macrophage-phage interaction. We estimated the maximal adsorption rate of the phage to the bacteria at 1.2 10-7 mL/PFU/h and assumed that a high AM density could inhibit the adsorption by up to 30%. This final model fits adequately the data from the different mice groups. From our model-based simulations, we predict a bacterial rebound of over 7 log for at least 25% of individuals 30 h post-infection in immunocompetent, whereas in the AM-depleted group, over 75% of individuals would show a bacterial load control (below 4 log). In median immunocompetent mice, our model also predicts a rebound in bacterial load at MOI 0.1, i.e., with a dose 100 times lower than that used in our study, which is not the case for AM-depleted mice, even at MOI 0.01.
Conclusion:
This work demonstrate the involvement of an ecological feedback between bacteriophage, bacteria, and the immune system in shaping the outcomes of phage therapy, which is of particular relevance in clinical settings with patients displaying various immune status.
Pre-print of this work; https://www.biorxiv.org/content/10.1101/2024.01.16.575879
References:
[1] Gordillo Altamirano, F.L., and Barr, J.J. (2019). Phage therapy in the postantibiotic Era. Clin. Microbiol. Rev. 32, e00066-18. 10.1128/CMR.00066-18.
[2] Roach, D.R., Leung, C.Y., Henry, M., Morello, E., Singh, D., Di Santo, J.P., Weitz, J.S., and Debarbieux, L. (2017). Synergy between the host immune system and bacteriophage is essential for successful phage therapy against an acute respiratory pathogen. Cell Host Microbe 22, 38-47.e4. 10.1016/j.chom.2017.06.018.
[3] Debarbieux, L., Leduc, D., Maura, D., Morello, E., Criscuolo, A., Grossi, O., Balloy, V., and Touqui, L. (2010). Bacteriophages can treat and prevent Pseudomonas aeruginosa lung infections. J. Infect. Dis. 201, 1096–1104. 10.1086/651135.
[4] Kuhn, E., and Lavielle, M. (2005). Maximum likelihood estimation in nonlinear mixed effects models. Comput. Stat. Data Anal. 49, 1020–1038. 10.1016/j.csda.2004.07.002.
[5] Delattre, M., Lavielle, M., and Poursat, M.-A. (2014). A note on BIC in mixed-effects models. Electron. J. Stat. 8, 456–475. 10.1214/14-EJS890.
[6] Brendel, K., Comets, E., Laffont, C., and Mentré, F. (2010). Evaluation of different tests based on observations for external model evaluation of population analyses. J. Pharmacokinet. Pharmacodyn. 37, 49–65. 10.1007/s10928-009-9143-7.
Reference: PAGE 32 (2024) Abstr 11074 [www.page-meeting.org/?abstract=11074]
Poster: Drug/Disease Modelling - Infection