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

Abstract

Pseudomonas aeruginosa is a major nosocomial pathogen that causes severe disease including sepsis. Carbapenem-resistant P. aeruginosa is recognised by the World Health Organisation as a priority 1 pathogen, with urgent need for new therapeutics. As such, there is renewed interest in using bacteriophages as a therapeutic. However, the dynamics of treating pan-resistant P. aeruginosa with phage in vivo are poorly understood. Using a pan-resistant P. aeruginosa in vivo infection model, phage therapy displays strong therapeutic potential, clearing infection from the blood, kidneys, and spleen. Remaining bacteria in the lungs and liver displays phage resistance due to limiting phage adsorption. Yet, resistance to phage results in re-sensitisation to a wide range of antibiotics. In this work, we use phage steering in vivo, pre-exposing a pan resistant P. aeruginosa infection with a phage cocktail to re-sensitise bacteria to antibiotics, clearing the infection from all organs.

Personalized aerosolised bacteriophage treatment of a chronic lung infection due to multidrug-resistant Pseudomonas aeruginosa

Article Open access 27 June 2023

Thilo Köhler, Alexandre Luscher, … Christian van Delden

Optimized preparation pipeline for emergency phage therapy against Pseudomonas aeruginosa at Yale University

Article Open access 01 February 2024

Silvia Würstle, Alina Lee, … Benjamin K. Chan

Bacteriophage-antibiotic combination therapy against extensively drug-resistant Pseudomonas aeruginosa infection to allow liver transplantation in a toddler

Article Open access 29 September 2022

Brieuc Van Nieuwenhuyse, Dimitri Van der Linden, … Jean-Paul Pirnay

Introduction

Antibiotic resistance is a global health issue and the need for new therapeutics was highlighted by the World Health Organisation (WHO), with Pseudomonas aeruginosa deemed a priority 1 pathogen in urgent need of new therapeutic strategies¹. P. aeruginosa occurs naturally in various environments and can cause nosocomial, life-threatening diseases including burn wound and pulmonary infections. It also causes acute issues in hospitals due to its ability to cause sepsis and its intrinsic resistance to many classes of antibiotics.

Hospital-acquired pneumonia (HAP) is the leading cause of death from nosocomial infections in critically ill patients. In 2017, the European Centre for Disease Prevention and Control (ECDC) reported that 8.3% of patients in ICU for >2 days presented with HAP and the most frequently isolated organism was P. aeruginosa, with carbapenem resistance reported in 26% of isolates^2,3. Furthermore, P. aeruginosa is the third most prevalent Gram-negative species isolated from clinical bacteraemia^4,5, and P. aeruginosa associated bacteraemia is linked with increased mortality compared with bacteraemia caused by other Gram-negative pathogens⁶. This highlights the critical importance of developing effective new therapeutics against pan-resistant pathogens.

Since the advent of increasing antibiotic resistance, there is renewed interest in using bacteriophages to treat bacterial infections. Phage therapy has many advantages compared to antibiotics, such as increased specificity, replication at the site of infection, low manufacturing cost and little reported toxicity⁷. Phage therapy has also shown efficacy in a variety of in vivo murine models^8,9,10,11. In a P. aeruginosa chronic lung infection murine model, phage PELP20 was highly effective against an established 6-day lung infection, completely clearing bacteria from the lungs in 70% of mice¹². However, complex evolutionary phage-bacteria dynamics in vivo remain unclear as environmental factors present in vivo, such as limited oxygen availability, and the presence of mucin and polyamines, have been shown to contribute to phage resistance^13,14,15. Thus far, bacterial adaptation to in vivo factors and the effect on phage resistance has not been explored.

Previous case studies show that phage therapy is effective clinically in the management of extensive pulmonary P. aeruginosa infections^16,17. Additionally, phage therapy can "re-sensitise" previously antibiotic resistant bacteria to antibiotics in vitro, a concept known as phage steering^18,19,20. However, to date, phage steering has not been demonstrated to clear an infection in vivo. Overall, bacteriophage therapy is an exciting potential alternative for treating multi-drug resistant (MDR) P. aeruginosa infections, needing further exploration.

In this study, we developed an in vivo model for P. aeruginosa systemic infection using a megaplasmid carrying, pan-resistant P. aeruginosa strain²¹. Our phage cocktail had strong therapeutic potential within this in vivo model. We found that phage resistance developed in vivo even the absence of phage treatment, with bacteria reisolated from the lungs of untreated mice displaying increased phage resistance. When investigated further, we found bacterial adaptation to in vivo factors such as oxygen availability, and the presence of mucin and polyamines contributes to the development of phage resistance. We also observed that bacterial isolates from phage treated mice gained phage resistance by limiting phage adsorption but were more susceptible to a wide range of antibiotics, resulting in a shift from carbapenem resistant to susceptible. The therapeutic potential for phage steering was demonstrated as pre-exposure to the phage cocktail re-sensitised the infection to antibiotics, permitting bacterial clearance in our pan-resistant P. aeruginosa in vivo infection model.

Results

Phage infectivity of clinical P. aeruginosa isolates

The susceptibility of 551 clinical bacterial isolates from a range of sources, including the UK and Thailand, to the four phages (PELP20, PNM, PT6 and 14/1) used in our phage cocktail was determined (Supplementary Data 1, Supplementary Table 1). PELP20 has previously been shown to be effective in vivo against P. aeruginosa Liverpool Epidemic strains¹², whilst phages PNM and 14/1 have previously been included in a well-defined cocktail available for use in clinical human trials²². Here, PELP20 displayed the broadest infection range (Fig. 1a; >76% infectivity of 551 clinical isolates) and 14/1 the narrowest (Fig. 1d; >41% infectivity). Overall, 454 (>80%) clinical isolates were susceptible to ≥1 phage and 190 (~35%) isolates demonstrated at least intermediate susceptibility to all 4 phages (Fig. 1f). All four phages were effective against the pan-resistant B9 (T2436) strain used in our mouse model (Fig. 1).

**Fig. 1: Phage cocktail screening against 551 clinical isolates.**

Invasive respiratory model for pan-resistant P. aeruginosa

An invasive respiratory infection model of pan-resistant P. aeruginosa was developed to test efficacy of phage therapy in vivo. Four different strains of MDR P. aeruginosa isolated from clinical sputum samples from patients in Thailand were tested; these included strains B3 (T2101), B8 (T2584), B9 (T2436), and C7 (T3582) (Supplementary Fig. 1). Mice were intra-nasally infected with each strain at 10⁶ cfu/ml and 10⁷ cfu/ml challenge doses, and the bacterial load in the lungs, liver, blood, spleen and kidneys was determined over a 48 h period post-infection. Secondary to lung infection, systemic spread was observed with all four strains at both challenge doses. Strain B9 (T2436) was chosen to continue forward to phage testing, as it had consistently high CFU loads in all organs and was both pan-resistant and harboured an unusual megaplasmid carrying antimicrobial resistance (AMR) genes²¹.

Following respiratory infection with B9, bacteria were found in the lungs (>10⁴ cfu/ml) and liver (>10³ cfu/ml) within 6 h post-infection (Supplementary Fig. 2a, b) while bacteraemia developed rapidly by 4 h in blood (>10⁴ cfu/ml) (Supplementary Fig. 2c). By 24 h post-infection, bacteria could also be isolated from the kidneys and spleen (Supplementary Fig. 2d, e). The bacterial load in each of these infection sites were maintained for the 48 h infection period, with CFU load increasing in liver, kidney, and spleen.

Early phage treatment significantly reduces P. aeruginosa in vivo

To investigate the timing and route of administration, treatment was trialled both early (simultaneously with bacteria) and late (5 h following bacteraemia) with administration via the intranasal and intravenous routes.

In mice treated with phage immediately after bacterial infection via the intranasal route (Fig. 2), phage treatment was able to target the site of infection (the lungs) before bacteria had entered the bloodstream. A single phage treatment of either the 4-phage cocktail or PELP20 alone were trialled. Compared to mock treated mice, phage cocktail treated and PELP20 treated mice had significantly reduced bacterial load in the lung (p < 0.0001), liver (p = 0.0013), and blood (p = 0.0021) 24 h post-infection, and significantly reduced bacterial load in the kidneys (p = 0.002) and spleen (p = 0.0001) 48 h post-infection (Fig. 2).The phage cocktail was more effective than PELP20 alone: phage cocktail treated mice completely cleared the infection from all organs tested, except from the liver which had residual CFU left ( ≤ 10² cfu/ml). In comparison, bacteria were still present in the lungs (<10⁴ cfu/ml), liver (<10² cfu/ml), kidney (<10² cfu/ml) and spleen (<10¹ cfu/ml) of PELP20 treated mice, 48 h post-infection (Fig. 2).

**Fig. 2: Early intranasal phage treatment in vivo.**

To determine whether intranasal administration of the phage cocktail resulted in phage dissemination into the bloodstream and amplification in vivo (i.e., auto-dosing), phage density (plaque forming units; PFU) was determined at 48 h for both uninfected and B9 infected mice treated with the phage cocktail. Intranasal administration of phage cocktail reached all organs tested, and in most animals, persisted for 48 h even in the absence of bacterial infection (Fig. 2f). Significant phage amplification was seen in the lungs of B9 infected mice, as the PFU was significantly increased (p = 0.0051) compared with uninfected, phage cocktail treated mice. In summary, early respiratory phage treatment (prior to systemic spread of P. aeruginosa), significantly reduced or totally cleared bacterial infection from all tissue sites.

Delayed phage treatment significantly reduces P. aeruginosa in vivo

We next determined whether phage treatment administered after systemic spread of P. aeruginosa, could reduce bacterial load. The efficacy of both intranasal and intravenous delivery of phage was tested. To mimic treating a MDR systemic infection, phage treatment was administered 5 h post-infection, after infection had become systemic. Intranasal administration of PELP20 5 h post infection was not effective at reducing CFU loads, however, treatment with phage cocktail significantly reduced (or cleared) CFU loads in lung, liver and blood by 24–48 h post-infection, compared to mock treated mice (Supplementary Fig. 3).

Intravenous treatment with phage was significantly more efficacious than intranasal administration, with significantly reduced bacterial loads in the lung (p = 0.0021), and blood (p = 0.014) 24 h post infection; and significantly reduced bacterial loads in the liver (p = 0.0007), kidneys (p = 0.002) and spleen (p = 0.0031) 48 h post-infection compared to mock-treated mice. Although bacteria were still detectable at low levels in the lungs (≤10² cfu/ml) and liver (≤10² cfu/ml) in phage cocktail treated mice, no bacteria were detected in the blood, kidneys and spleen at 48 h post infection (Fig. 3). In contrast, in single phage PELP20 treated mice, at 48 h the infection was only cleared from blood. Similar to intranasal delivery, when PFU was measured 48 h after intravenous administration of the phage cocktail, phages could be detected in all tissues tested, including significant phage amplification in the lungs of B9 infected mice (p < 0.0001) (Fig. 3f). The phage cocktail was therefore highly effective at reducing (and in some cases even clearing) CFU from infected tissues and blood in this challenging systemic infection model.

**Fig. 3: Delayed intravenous phage treatment in vivo.**

P. aeruginosa develops resistance to phage in vivo

To determine whether resistance to phage developed in vivo, P. aeruginosa present in tissues at 48 h post-infection from mock-treated, phage cocktail treated and PELP20 treated mice were harvested. Efficiency of plating (EOP) of the phages present in the cocktail (PELP20, PNM, 14/1 and PT6) was conducted on all the in vivo recovered isolates to determine whether phage resistance developed in vivo.

Surprisingly, bacteria isolated from the lungs of mock-treated mice displayed increased phage resistance compared to the input bacterial strain (Fig. 4a, b). These non-phage treated isolates from the lungs were found to be nearly completely resistant to all the phages present in the cocktail (Fig. 4a, c). This has important implications, showing that resistance developed in the lungs independently of phage treatment, potentially limiting the efficacy of phage therapy in this niche. However, the development of resistance in non-phage treated isolates was not universal across all organs. Isolates from the kidneys were still susceptible to the input phages, while phage resistance in isolates recovered from the liver, blood and spleen was more variable (Fig. 4a). This indicates that some biological niches may be more difficult to treat using phage therapy due to resistance developing independently of phage treatment.

**Fig. 4: Phage resistance develops in vivo due to reduced phage adsorption.**

Organ-related differences in resistance was observed when comparing the non-phage treated isolates and phage treated isolates (Fig. 4b & Supplementary Fig. 4): whilst phage treated isolates from the lungs were slightly less resistant to the input phages than mock-treated isolates, phage treated isolates from the liver and kidneys developed more resistance compared to the mock-treated isolates. This suggests that bacterial adaption to environmental factors could impact the development of phage resistance (Fig. 4b). We then explored whether phage resistance was due to modification of phage surface receptors via phage adsorption assays. Non-phage treated bacterial isolates had variable phage adsorption rates, with only isolates from the lungs showing no adsorption, explaining the phage resistance observed for non-phage treated lung isolates. All phage treated isolates were resistant to the input phages due to reduced adsorption (Fig. 4c, d). This may indicate that phage treated bacteria either modify phage receptors or alter their outer membrane properties resulting in phage resistance. Overall, while our phage cocktail showed efficacy, phage resistance still occurs in vivo, which has important implications for future use of phage therapy.

P. aeruginosa adaptation to the lung results in phage resistance

The lung environment is unique²³, with many environmental stressors to which P. aeruginosa adapts, which could result in phage resistance as a secondary effect. These factors include variable oxygen availability¹³, the presence of polyamines¹⁴ and mucin¹⁵, all of which have been linked with phage resistance. To explore this, an experimental evolution approach was conducted, where B9 was grown in nutrient rich broth (LB Broth) or Healthy Lung Media (HLM)²³ for 48 h, then isolates were recovered and tested for phage resistance (via EOP). The effect of oxygen availability was determined by repeating the experiment under microaerophilic and anaerobic conditions. Finally, the effect of mucin and polyamines was investigated by growing B9 in LB broth supplemented with either polyamines (spermidine-200 ng/ml, spermine-32.5ug/l, and putrescine-616 ug/l) or mucin (1.2 mg/ml), at the same concentrations used in HLM and testing recovered isolates for resistance via EOP compared to the input strain and controls (i.e., LB in aerobic conditions).

We found P. aeruginosa grown in media mimicking the lung environment acquired resistance to all four phages present in the cocktail (Fig. 5a). Furthermore, we show that even in nutrient rich broth, oxygen availability was a key driver of phage resistance (Fig. 5a). While bacteria grown in the presence of polyamines only developed resistance to the phage PNM, P. aeruginosa isolates grown in the presence of mucin demonstrated resistance to all four phages (Fig. 5b).

**Fig. 5: Adaptation to the lung environment results in the development of phage resistance in the absence of phage treatment.**

We have shown that bacterial adaption to the lung, which includes limited oxygen availability, mucins, and polyamines, can result in phage resistance as a secondary effect. Bacterial adaption to different biological niches may reduce the efficacy of phage therapy and should be taken into consideration.

Phage exposure alters antimicrobial susceptibility of P. aeruginosa

Previous in vitro studies showed that phage resistant bacteria can become re-sensitised to antibiotics^18,19,20. However, there is little evidence for re-sensitisation within in vivo models. Therefore, altered antimicrobial sensitivity in vivo was explored via disk diffusion assays for a range of antibiotics and E-Test to determine the minimum inhibitory concentration (MIC) of two clinically relevant antibiotics, tobramycin and meropenem.

Non-phage treated bacteria displayed only small differences in antibiotic susceptibility compared to the input isolate (Fig. 6). For the majority of antibiotics, there was either no change or changes <5 mm in zone diameter and MIC for tobramycin and meropenem (Fig. 6a). For non-phage treated isolates from the lungs, the zone of inhibition to meropenem increased by 5 mm, however isolates were still classed as resistant overall. For bacteria isolated from other organs of non-phage treated mice, there was an increase in the zone of inhibition to cefepime of between 5 and 9 mm (Fig. 6b–f) however, this did not result in a change in resistance based on EUCAST clinical breakpoints. Small increases in resistance to ticarcillin-clavulanic acid was observed and, in some cases, resulted in a change in resistance class (from intermediate to resistant) (Fig. 6f). While isolates derived from different in vivo niches displayed small differences in antibiotic susceptibility, the isolates from mock-treated mice were largely unchanged based on clinical breakpoints which are designed to indicate clinical efficacy in systemic infection.

**Fig. 6: Antibiotic susceptibility of non-phage treated isolates recovered from systemic infection in vivo model.**

However, isolates from delayed phage treatment group did demonstrate altered antibiotic susceptibility based on clinical breakpoints. When phage treatment of infected mice was delayed by 5 h, an increase in antibiotic susceptibility in phage resistant isolates was seen compared to the input strain. Large changes in antibiotic susceptibility were observed across multiple classes of antibiotics, typically a 15–25 mm increase in the zone of inhibition (Fig. 7a–e). Additionally, >2-fold reduction in MIC was seen in 14 out of 15 isolates for meropenem and >4-fold reduction was in 11 out of 15 isolates for tobramycin (Fig. 7f). Across all isolates, only resistance to aztreonam remained unaltered when compared to the input isolate.

**Fig. 7: Increased antibiotic susceptibility of delayed phage treated isolates recovered from systemic infection in vivo model.**

Of the 15 isolates from delayed phage treated mice, 13 isolates showed multiple shifts in susceptibility classification (Fig. 7). Isolates from delayed phage treated lungs and liver swung from meropenem-resistant to sensitive and for one isolate treated with the cocktail, there were 12 shifts in susceptibility classification. These results highlight major and clinically relevant, changes in antibiotic susceptibility following phage treatment.

Timing of phage administration appears to be a key factor in antibiotic sensitisation as only small changes in antibiotic susceptibility were seen in early phage treated isolates (Supplementary Fig. 5)

Genetic alterations in phage resistant isolates

To further investigate the mechanism of phage resistance and antibiotic re-sensitisation, we conducted whole genome sequencing to compare acquisition of mutations between non phage treated in vivo adapted isolates and the phage treated in vivo adapted isolates. Across all isolates sequenced, there was no evidence of loss of the megaplasmid, and no mutations on the plasmid to indicate altered expression of plasmid-encoded AMR genes.

In isolates recovered from the lungs of non-phage treated mice, we found evidence of a frameshift variant in gene FC629_24630, a glycosyltransferase family 2 protein which has homology with PA01 migA, an LPS associated alpha-1,5-rhamnosyltransferase²⁴. Alterations in LPS biosynthesis could explain the development of resistance in the absence of phage treatment via modification of the phage adsorption receptor for LPS-targeting phages (14/1 and PELP20). There was also evidence of an 810 kb duplication in isolates recovered from the lungs, which could have an impact of phage resistance by altering expression of many genes within this region (Fig. 8, Supplementary data 2). Furthermore, a number of SNPs present at low frequency in the B9 infection stock have become fixed in nearly all the in vivo adapted isolates, suggesting they are advantageous for P. aeruginosa survival in vivo (Supplementary Fig. 6, Supplementary data 2). Overall, there is evidence of P. aeruginosa adapting to the in vivo lung environment.

**Fig. 8: Genetic variation and outer membrane permeability of non-phage and phage treated in vivo adapted isolates.**

Naegleria Fowleri Symptoms