Exploiting lung adaptation and phage steering to clear pan-resistant Pseudomonas aeruginosa infections in vivo

Antibiotic resistance in Pseudomonas aeruginosa is a critical global health concern, with the WHO designating carbapenem-resistant P. aeruginosa as a priority 1 pathogen. P. aeruginosa causes severe nosocomial infections including hospital-acquired pneumonia (HAP) and bacteraemia, both associated with high morbidity and mortality. Conventional antibiotics often fail due to intrinsic and acquired resistance. Bacteriophages offer a potential alternative because of their specificity, in situ replication, low cost, and minimal toxicity. While case reports and murine models have shown phage efficacy against P. aeruginosa, the evolutionary dynamics of phage–bacteria interactions in vivo, particularly within different host niches (e.g., the lung microenvironment with low oxygen and abundant mucins/polysaccharides), remain poorly understood. This study aimed to: (1) develop a systemic in vivo model using a pan-resistant P. aeruginosa strain carrying a resistance megaplasmid; (2) evaluate efficacy and pharmacodynamics of a four-phage cocktail administered early or after systemic spread, via intranasal or intravenous routes; (3) determine whether P. aeruginosa develops phage resistance in vivo and identify underlying mechanisms; and (4) test whether phage steering can re-sensitise pan-resistant infections to antibiotics in vivo to enable bacterial clearance.

Prior clinical case studies and animal models have reported successful phage therapy for extensive pulmonary P. aeruginosa infections. In vitro work has shown that phage resistance can trade off with antibiotic resistance, re-sensitizing bacteria (phage steering). However, demonstration of phage steering leading to in vivo infection clearance had not been reported. Environmental factors relevant in vivo (hypoxia, mucins, polyamines) are known to alter phage susceptibility and resistance evolution. Previous murine models demonstrated single phage efficacy in chronic lung infection, but comprehensive evaluation of resistance dynamics across organs and re-sensitisation to multiple antibiotics in vivo remained a gap this study addresses.

- Bacterial strain and model: Four MDR clinical P. aeruginosa strains from Thailand (B3/T2101, B5/T2548, B7/T3582, B9/T2436) were screened in a murine invasive respiratory model (female BALB/c, 6–8 weeks). Intranasal infection with 1×10^6 CFU in 50 µL PBS; systemic dissemination to lungs, liver, blood by 6–24 h; kidneys and spleen by 24 h. B9 (T2436), harboring a resistance megaplasmid, was selected for subsequent experiments due to robust multi-organ burden. - Phage cocktail: Four phages (PELP20, PMN, PT6, 14/1) assembled based on MOI checkerboard assays against B9. Phages propagated via double-layer agar and titered; cocktail prepared with equal volumes targeting an overall starting dose of approximately 1×10^12 PFU administered to mice. PELP20 also tested as a single-phage control. - Host range screening: Direct spot tests screened 551 clinical isolates (primarily keratitis, plus UTI and wound isolates from UK, Kuwait, Thailand) for susceptibility to each phage and the cocktail; lysis recorded as complete, partial, or none. - Treatment regimens and routes: Two timing regimens—early (immediately after infection) and delayed (5–6 h post-infection after systemic spread). Two routes—intranasal and intravenous. Groups received PBS, single phage (PELP20), or the four-phage cocktail. CFU in lungs, liver, blood, kidneys, and spleen measured at 24 and 48 h post-infection. Phage titers (PFU) measured in tissues at 48 h to assess dissemination/amplification. - Phage resistance assays: Efficiency of plating (EOP) of isolates recovered from each organ (mock-treated, PELP20-treated, cocktail-treated) against each cocktail phage; adsorption assays with the cocktail at MOI≈1, sampling 5–60 min and titering filtrates. - Lung-environment adaptation experiments: In vitro evolution of B9 for 48 h in Healthy Lung Media (HLM) and LB under aerobic, microaerophilic, and anaerobic conditions; LB supplemented with polyamines (spermidine, spermine, putrescine) or mucin at HLM-equivalent concentrations. Clonal isolates tested by EOP for phage susceptibility. - Antibiotic susceptibility testing: Disk diffusion (EUCAST breakpoints) for a panel of antibiotics; MICs by E-test for meropenem and tobramycin on isolates from each organ and treatment condition. - Whole-genome sequencing: Illumina MiSeq of isolates from various organs/treatments; reads aligned to P. aeruginosa reference; variants called with GATK HaplotypeCaller, annotated with SnpEff; large SVs inspected visually. Assessed plasmid retention and AMR gene status. - Membrane assays: Outer membrane permeability (NPN uptake, propidium iodide fluorescence) and cytoplasmic membrane depolarization (DiSC3(5)) comparing input B9, non-phage-treated lung isolates, and delayed phage-treated lung isolates. - Phage–antibiotic sequencing therapy (phage steering): Mice infected with B9; treated with phage cocktail at 5 h post-infection, then antibiotics at 48 h (meropenem 1 mg/kg or tobramycin 5 mg/kg). CFUs in organs assessed; comparisons to monotherapy controls. - Statistics: GraphPad Prism; two-way ANOVA with Bonferroni correction for multi-group comparisons; organ- and treatment-specific p-values reported.

- Efficacy of phage therapy: - Early intranasal treatment immediately post-infection significantly reduced CFUs at 24–48 h versus PBS. With the phage cocktail, CFUs were significantly reduced in lungs (p<0.0001), liver (p=0.003), and blood (p=0.0021) at 24 h, and in kidneys (p=0.002) and spleen (p=0.0001) at 48 h. The cocktail often cleared detectable bacteria in multiple organs by 48 h; single phage PELP20 reduced but did not clear all organs. - Delayed treatment after systemic spread: Intravenous cocktail outperformed intranasal delivery. At 24 h (post-treatment), significant CFU reductions in lungs (p=0.0021) and blood (p=0.014); at 48 h, liver (p=0.0007), kidneys (p=0.002), spleen (p=0.0031). Blood, kidneys, and spleen were cleared at 48 h in cocktail-treated mice; lungs and liver had low residual CFUs (<10^1 CFU/ml). PELP20 alone cleared blood only at 48 h. - Phage dissemination/amplification: Following intranasal or intravenous cocktail administration, phages were detected across organs; significant amplification in lungs of infected mice compared to uninfected controls (intranasal p=0.005; intravenous p<0.001). - In vivo phage resistance and niche effects: - Even without phage treatment, lung isolates at 48 h evolved near-complete resistance to all cocktail phages by EOP, whereas kidney isolates remained susceptible; liver, blood, spleen showed variable resistance. Adsorption assays indicated non-phage treated lung isolates had little to no adsorption, explaining resistance. - Phage-treated isolates from multiple sites were resistant to input phages due to reduced adsorption, indicating receptor modification or altered outer-membrane properties. - Lung-like conditions drive resistance in absence of phage: - Growth in Healthy Lung Media and under microaerophilic/anaerobic conditions increased resistance to all four phages. Oxygen limitation was a key driver; mucin and polyamines also contributed to resistance phenotypes. - Antibiotic re-sensitisation after phage exposure in vivo: - Non-phage-treated isolates showed minimal changes versus the ancestral strain (mostly ≤5 mm zone changes; small MIC shifts). Lung mock isolates had a +5 mm meropenem zone but remained clinically resistant. - Delayed phage-treated isolates displayed large increases in susceptibility across multiple classes (typically +15–25 mm in disk diffusion). For tobramycin, >2-fold MIC reduction in 14/15 isolates; multiple isolates shifted from meropenem resistant to susceptible by EUCAST breakpoints. Aztreonam susceptibility largely unchanged. - Mechanisms: - WGS of non-phage-treated lung isolates revealed a frameshift in FC629.24630 (glycosyltransferase family 2; homologous to PAO1 rhamnosyltransferase involved in LPS) and an ~810 kb duplication; early treated liver isolates also showed large SVs (including a ~241 kb deletion). Kidney isolates featured SNPs in LPS-associated genes and a ~64 kb deletion. These changes likely alter adsorption receptors (e.g., LPS) and contribute to resistance. - In delayed phage-treated isolates with broad re-sensitisation, there was no loss/mutation of AMR plasmid genes; data suggest post-transcriptional changes increasing outer membrane permeability. Membrane assays showed significantly greater outer membrane permeability in delayed phage-treated lung isolates versus input and non-phage-treated isolates, without significant changes in cytoplasmic depolarization. - Phage steering (phage then antibiotic) in vivo: - Pre-exposure to the phage cocktail followed by meropenem (1 mg/kg) or tobramycin (5 mg/kg) at 48 h enabled marked CFU reductions and, in many cases, clearance across organs. For cocktail+tobramycin, all organs cleared except occasional residual in liver (2/10 mice). Significant improvements versus antibiotic alone were observed in lungs and other organs in specific comparisons (e.g., lung CFU reduction p=0.0233 for tobramycin vs PBS; significant differences for meropenem in liver p=0.0133 and kidney p=0.0046).

This study demonstrates that bacteriophage therapy can substantially reduce and often clear systemic infections caused by a pan-resistant P. aeruginosa strain in vivo. Crucially, the work elucidates niche-specific evolutionary dynamics: the lung environment promotes phage resistance even in the absence of phage exposure, likely via altered adsorption (e.g., LPS modification) driven by hypoxia and lung-associated components (mucins, polyamines). Despite this, phage exposure can steer survivors toward antibiotic susceptibility by increasing outer membrane permeability without loss of AMR determinants, enabling effective antibiotic action. Thus, phage steering transforms an untreatable, pan-resistant infection into one amenable to standard-of-care antibiotics, addressing the central question of whether in vivo phage steering can be harnessed therapeutically. Route and timing affect efficacy; intravenous delivery after systemic spread achieved broader organ-level reductions than intranasal dosing, aligning with phage pharmacokinetics and RES-mediated distribution. These findings underscore the importance of combining phage with antibiotics in a sequenced regimen and of tailoring administration strategies to infection stage and site.

Using a clinically relevant systemic murine model with a pan-resistant P. aeruginosa strain, a four-phage cocktail reduced or cleared infection across multiple organs. The lung environment independently drove phage resistance via reduced adsorption, yet phage exposure re-sensitized bacteria to diverse antibiotics through increased outer membrane permeability, enabling successful phage steering in vivo. This provides proof-of-concept that phage pre-treatment followed by antibiotics can clear otherwise intractable infections. Future work should optimize dosing, timing, and route; purify phage preparations for clinical safety; dissect individual phage dynamics within cocktails; expand host-range validation to diverse clinical isolates beyond keratitis; investigate mechanistic bases of permeability changes and regulatory adaptations; and test phage–antibiotic ordering across additional drug classes and infection models.

- Biological niche variability: Efficacy and resistance evolved differently across organs; liver frequently retained residual bacteria and showed distinct resistance dynamics. - Phage preparation: Cocktail not purified for endotoxins; clinical translation requires purification which can lower titer and potentially affect efficacy. - Dose comparisons: Cocktail contained higher total phage than single-phage PELP20; efficacy differences may partly reflect total dose rather than synergy. Higher-dose single-phage controls were not fully explored. - Host-range assessment methodology: Direct spot tests can underestimate true infectivity due to lysis-from-without; the screening set was dominated by keratitis isolates, with fewer hospital-acquired and UTI isolates. - Timing/route constraints: Delayed intranasal dosing was less effective systemically, likely due to limited translocation to blood; comprehensive PK/PD for each phage in vivo was not performed. - Genomic interpretations: Some variant calls were low frequency in the ancestral stock; causality between specific mutations/structural variants and resistance phenotypes was inferred but not functionally validated.