Non-canonical phage-bacterium interactions during topical phage therapy in the PhagoBurn clinical trial: a...

The PhagoBurn trial compared the efficacy and tolerability of a cocktail of lytic anti-Pseudomonas aeruginosa bacteriophages with standard of care for patients with burns but did not meet its primary endpoint. In this follow-up study using samples from the PhagoBurn trial, we aimed to investigate non-canonical phage-bacterium interactions and longitudinal changes in the phenotypic and genotypic characteristics of P aeruginosa during phage therapy.

In this retrospective, observational, reverse translational study, we collected viable P aeruginosa isolates from the wound swabs of individuals who completed at least 7 days of treatment with the phage cocktail product (PP1131) in the multicentre PhagoBurn trial conducted in Switzerland, France, and Belgium. Isolates were subjected to longitudinal phenotypic screening for phage susceptibility using standard drop assays and reverse double-layer assays. Isolates with unique phage susceptibility profiles were selected for genomic characterisation. After the unexpected detection of PP1131-derived phage DNA in many bacterial isolates, non-canonical phage-bacterium interactions were assessed using double-layer agar lysis assays and ultrastructural assessments using electron microscopy.

Of the 27 individuals enrolled in the PhagoBurn trial, 13 received phage therapy; among them, ten individuals completed the therapy, and longitudinal P aeruginosa isolates from seven individuals were available for analysis. 184 (mean 26 [SD 21·9] isolates per individual) bacterial isolates were phenotypically assessed. Phage susceptibility varied substantially between and within individuals over time, with rapid emergence of resistance in some cases. 98 (53%) of 184 isolates (mean 14 [11·8] isolates per individual) had unique phage susceptibility profiles and were selected for genome sequencing. In these 98 isolates, mutations were detected in genes involved in virulence regulation, secretion and efflux, chemotaxis, and biosynthesis. Complete or partial phage genomes derived from PP1131 were detected in 41 (42%) of the 98 sequenced bacterial isolates; 16 (39%) of 41 phage-positive isolates continuously secreted infectious phage particles, consistent with a productive chronic infection phenotype. In 23 (56%) of 41 isolates, phage genomic DNA persisted without detectable phage release, suggesting non-productive chronic infection (including putative pseudolysogeny). These alternative infection states remained stable and detectable after serial passages in vitro and were observed in isolates from individuals with both favourable and unfavourable clinical outcomes.

Bacterial adaptation, phage adaptation, and non-canonical phage-bacterium interactions have been largely overlooked in clinical trial design and might contribute to variability in therapeutic outcomes and warrant further investigation. The non-canonical phage interactions (eg, chronic infection and pseudolysogeny) observed in this study highlight the need for protocols that account for the emergence of persistent non-lytic phage-bacterium interactions.

Stiftung für die Forschung in Anästhesiologie und Intensivmedizin, Swiss National Science Foundation, and Placide Nicod Foundation.