Lytic activity, stability, biofilm disruption capabilities, and genomic characterization of two bacteriophages active against respiratory MRSA

Abstract

Aims

This study aimed to characterize bacteriophages for potential therapeutic use against Staphylococcus aureus, focusing on clinical respiratory isolates of methicillin-sensitive (MSSA) and methicillin-resistant (MRSA) strains. Specifically, it sought to evaluate phage lytic activity, host range, stability, biofilm disruption capabilities, and overall safety for therapeutic use.

Methods and results

Novel phages, Koomba kaat 1 and Biyabeda mokiny 1, were identified and characterized using microbiological assays and bioinformatics. They exhibited lytic activity against clinical MSSA and MRSA isolates, disrupted biofilms from airway isolates, remained stable for at least one year in storage, and could be aerosolized without significant reductions in activity. Bioinformatic tools were used to assess safety, lifecycle, virulence, and prophage contamination when grown using their original isolation host. Receptor binding proteins within their genomes were also predicted, providing insight into their mechanisms of action. Both phages demonstrated strong efficacy against the clinical isolates tested and demonstrated robust stability under storage and delivery conditions.

Conclusions

Koomba kaat 1 and Biyabeda mokiny 1 are promising candidates for phage therapy. Their efficacy against clinical S. aureus isolates, ability to break down biofilm, and stability for airway implementation, positions them as valuable tools for addressing persistent airway infections caused by S. aureus.

Introduction

The use of bacteriophage (phage) is a possible solution to the issue of antibiotic resistance (Gordillo Altamirano and Barr 2019). Fortunately, phages are found wherever their bacterial host species grow (Clokie et al. 2011), and their isolation from various environmental sources has been well documented (Han et al. 2013, Khan Mirzaei and Nilsson 2015, Mattila et al. 2015, Rasool et al. 2016, Wintachai et al. 2022). Most phages have been derived from environments with high bacterial populations, particularly wastewater, owing to its abundance and ease of accessibility (Otawa et al. 2007, Batinovic et al. 2019). With the widespread adoption of whole genome analyses, the true diversity of phages has become apparent, resulting in a rapidly increasing number of diverse phages that are highly effective against antibiotic-resistant bacteria (Cook et al. 2021). Subsequently, the potential for phage as an alternative therapeutic option is quickly becoming realized, although achieving ‘on demand’ selection for individual clinical use cases still requires further infrastructure development (Mattila et al. 2015).

Recently, the drug resistant Staphylococcus aureus (S. aureus) phenotype (MRSA) has been reported as an especially lethal pathogen, causing over 100 000 deaths attributable to antimicrobial resistance (AMR) in 2019 (Murray et al. 2022). Currently, S. aureus is a leading cause of a broad range of infections from skin and soft tissues to bone and joints (Turner et al. 2019). This is due, in part, to S. aureus being a typical commensal pathogen (von Eiff et al. 2001), as well as a broad range of virulence factors that enable persistence within various bodily sites (Kelly et al. 2012, Hommes and Surewaard 2022, Kebriaei et al. 2023). While clinical phage use against S. aureus has been reported previously in a few clinical trials (Rhoads et al. 2009, McCallin et al. 2018, Petrovic Fabijan et al. 2020), phage use against MRSA infections have not become widespread as a standard clinical practice. Furthermore, despite the number of phages isolated against S. aureus, the measurements conducted for many isolated phages do not include application-specific analyses. Therapeutic applications of phage must undergo a lengthy process of characterization to ensure their suitability for human administration (Philipson et al. 2018).

There are numerous characteristics that are considered desirable for phage therapy, such as a broad host range and stable physiochemical characteristics. Host range reports the breadth of activity of a phage against planktonic-state bacteria. Another consideration for bacteria like S. aureus is their ability to produce biofilm. Biofilm production has been associated with various S. aureus disease conditions and increased resistance to antimicrobials and persistence at the site of infection (Schilcher and Horswill 2020, Kebriaei et al. 2023). Currently the effects of antimicrobial activity on these biofilms require further attention, with Staphylococcus-infecting phages demonstrating mixed activity against biofilms (Song et al. 2021, Joo et al. 2023, Verheul et al. 2024). Generally, phages that are suitable for phage therapy must be both obligately lytic so they cannot integrate into the bacterial genome and must also not contain any host-derived virulence factors (Dion et al. 2020, Strancar et al. 2023). Fortunately, the widespread implementation of high-throughput sequencing technologies (Rhoads and Au 2015, Rihtman et al. 2016, Lin et al. 2021) has led to the creation of numerous open-access tools that can be used to assemble genomes ‘de novo’ (Bankevich et al. 2012, Afiahayati et al. 2015, Wick et al. 2017). Interactions between the phages and their host are also complex, unique, and may involve prophages encoded within the bacterial genome that will parasitize the infecting phage to package its own genome in a process called the molecular piracy (Christie and Dokland 2012). As this is a known phenomenon within S. aureus phage-host interactions (Hawkins et al. 2021), prophages are an important parameter to investigate using whole genome sequencing (WGS) analysis of phage. Currently, information regarding the host organism and levels of prophage induction are often neglected, save from a select few characterization manuscripts (Strancar et al. 2023).

With comparatively few phages identified to be active against S. aureus compared to other high priority bacteria (Asokan et al. 2019), this study is aimed to characterize two lytic bacteriophages we recently isolated: Biyabeda mokiny 1 and Koomba kaat 1 (Iszatt et al. 2022, Iszatt et al. 2022), both of which we found are active against a wide range of MRSA and MSSA clinical isolates.

Materials and methods

Phage acquisition

We recently reported the genomes of two novel phages Koomba kaat 1 and Biyabeda mokiny 1 (Iszatt et al. 2022, Iszatt et al. 2022), which were obtained, propagated, and prepared as a crude filtered lysate.

Phage genomics analysis

Phage DNA was extracted and sequenced for genome assembly in our prior publications (Iszatt et al. 2022, Iszatt et al. 2022). A multi-factorial genomic analysis was further conducted in this study and raw reads were re-assembled using the Phanta (v0.4) short read assembly pipeline (Iszatt 2024).

Genome annotation and reordering

As per previously published guidelines (Shen and Millard 2021, Turner et al. 2021), genomes that had passed the assembly curation process were annotated and reordered based on the small or large terminase subunits. Genome annotations were performed using Prokka (Seemann 2014) in conjunction with the PHROGs database (accessed May 2023) (Terzian et al. 2021) to provide annotations. Coding capacity was calculated as the sum of the length of all coding features divided by the total genome length.

Lifecycle analysis

To determine lifecycle, lysogeny-associated genes were searched for within the annotation data. Lysogeny-associated proteins include integrases, excisionases, recombinases, transposases, repressors, ParA, and ParB (Beamud et al. 2023, Strancar et al. 2023). The presence of any of these within the genomes of phages was flagged as lysogens. Next, each coding sequence, regardless of annotation, was subject to BLASTp (v2.5.0) analysis against all proteins extracted from the Inphared database (Cook et al. 2021) (accessed 1st May 2023) containing 1 799 354 proteins in total. Parameters included 10 maximum target sequences and a coverage of at least 50%. If any of the resulting hits, per coding sequence (CDS), were lysogeny-associated proteins, the flagged protein was investigated further as a potential lysogen.

Screening for virulence factors

Identification of antibiotic resistance genes was carried out using internal databases available through ABRicate v1.0.1 (Seemann Torsten 2023). In addition, the NCBI Antimicrobial Resistance Gene Finder Plus (Feldgarden et al. 2019) (AMRFinderPlus) version 3.11.20 was used to download the latest published databases from NCBI (Database version: 2023–09-26.1) and screen both genomes for the presence of AMR-related genes.

Phylogenetic UPGMA tree

Phylogenetic alignment of all Staphylococcus phage sequences (n = 574, accessed May 2023) listed within the Inphared database (Cook et al. 2021) was performed using MAFFT (v7.520) (Katoh and Standley 2013), and a UPGMA tree drawn using Biopython modules (Cock et al. 2009). To compare the genomic synteny and arrangement between phage genomes, the closest phages used for comparisons and phylogenetic inference were identified using a BLAST (v2.5.0) (Camacho et al. 2009) search against these phage genomes.

Receptor binding proteins

Phage receptor binding proteins (RBPs) were putatively identified within both phage genomes using a combination of methods. The first was prediction using a previously published machine learning pipeline: protein CDS were taken directly from Prokka (v1.14.6) (Seemann 2014) outputs produced during the annotation process. Briefly, protein sequences were transformed into numerical vectors using the ProtTransBertBFDEmbedder module in Python, which uses an embedding tool designed to capture sequence features using the ProtTrans-BFD language model (Elnaggar et al. 2022). A pre-trained XGBoost classifier model (Boeckaerts et al. 2022) using known phage RBPs was then used to generate predictions for each phage protein analysed. Once these proteins were identified, they were manually curated using the HHpred webserver for homologue detection (Soding et al. 2005) and compared with closely related genomes that have empirically determined RBPs via collinear alignment.

For comparisons, Staphylococcus-infecting phages with previously identified and characterized RBPs were included within the analysis, these phages were Staphylococcus phages: SA012 (Accession: AB903967), MR003 (Accession: AP019522), and SLT (Accession: AB045978) (Kaneko et al. 2009, Takeuchi et al. 2016, Peng et al. 2019). All genomes were reordered based on the small terminase subunit for collinear comparisons. Similarities were calculated using Mmseqs2 (v12-113e3 + ds-3 + b1) (Steinegger and Soding 2017) and visualized using pyGenomeViz (v0.4.3) (Yuki Shimoyama 2022).

Microbiological assays

Plaque characteristics

Purified phages were morphologically characterized based on size, shape, presence of a halo, and the plaque turbidity on a black background. To characterize each plaque, a titration was performed followed by a whole plate overlay to obtain countable numbers (30–300) of plaques on a single plate. Plates with plaques were imaged using the ChemiDoc (BioRad, Hercules, CA, USA). Twenty plaques from each plate were chosen and measured using ImageJ (v1.54d) (Schneider Rasband et al. 2012) software to generate averages. The diameters of the plaques were averaged, and plaque size categorized as small (<1 mm), medium (1-2 mm), or large (>2 mm) based on the mean value. Colonies were also visually inspected for the presence of a halo (Negative | Positive), lysis type (Clear | Turbid), and shape (Round | Irregular).

Transmission electron microscopy

Phage samples that were verified as ‘complete’ or ‘finished’, free of contamination, lysogenic proteins, and absent of any virulence-associated factors, were sent for imaging via transmission electron microscopy (TEM). Five microlitres of high titre (1 × 10⁹ PFU/mL) phage samples were aliquoted onto formvar-coated grids for 2 minutes, and then subsequently stained then with 2% (v/v) aqueous uranyl acetate, before washing with 5 μL of ddH₂O. Images were taken using a Tecnai G2 Spirit 120 kV electron microscope. The acceleration voltage was set at 100 kV to enhance image contrast. Images were recorded using an AMT Nanosprint 15 camera equipped with accompanying software V7.0.1. Phage dimensions such as head length, head diameter, tail length, and tail width were measured using ImageJ (v1.54d). From these measurements, the phages were classified into the following morphotypes: Myovirus, Podovirus or Siphovirus (Turner et al. 2023).

Long-term storage assessment

To determine stability in storage, phages were stored for up to one year at room temperature (RT), 4, −20, and −80°C. Phages in storage were enumerated via serial dilution, spot plating, and plaque counting on their host bacterial strain at the storage timepoints of 1 week, 1 month, 3 months, 6 months, and 1 year, for comparison to their starting concentrations. Three replicate tubes were used for each phage at each timepoint, and all tubes were placed at RT 1 hour prior to performing titrations to enable aliquots to thaw (if frozen).

Aerosol stability

To determine the aerosol stability of phage via nebulization; we used the Aerogen Pro vibrating mesh nebulizer (Aerogen Ltd, Dangen, Galway, Ireland) with the Aeroneb Lab control module (Kent Scientific, Torrington, CT 06790, United States). Briefly, 2 mL of phage lysate were purified to remove endotoxin to recreate a ‘therapeutic’ sample of phage using high-pressure liquid chromatography (Supplementary methods 4) and diluted to a final concentration of 1 × 10⁹ PFU/mL. Samples were aerosolized into 25 mL collection tubes and allowed to settle for 10 minutes. The remaining suspensions were then collected to determine volume, and the phage concentration was determined via titration.

Host range and specificity

Host range assays were performed by spot plating 10 μL of 1 × 10^8–9 phage lysate onto host-overlay agar plates inoculated with 100 μL of bacterial overnight culture. In total, 138 strains of clinical S. aureus were assayed from various sites, including respiratory (n = 24), blood (n = 90), tissue non-respiratory (n = 12), and a combination of wound sites (n = 12). Additional species of bacteria were used to determine if either phage could infect multiple species; these included Pseudomonas aeruginosa (P. aeruginosa) (n = 30) clinical isolates, Group A Streptococci (GAS) (n = 8), and Group B Streptococci (GBS) (n = 13) clinical isolates. For the P. aeruginosa isolates, the culturing medium used was Luria Bertani (LB) broth, and agar was made to manufacturers specifications (BD DifcoTM). Heart-infusion broth and agar (BD DifcoTM) were used for GAS and GBS, these were prepared according to the manufacturer’s instructions. Each of these experiments were repeated three times independently to ensure reproducibility.

Biofilm disruption

To assess biofilm disruption capabilities, five MRSA isolates from CF airways (SA04, SA05, SA07, SA08, SA09) and a laboratory strain of S. aureus (ATCC 6538) were chosen for biofilm production capacity and subsequent susceptibility to phage. Briefly, wells of a flat bottomed 96-well plate were inoculated with 200 µL of bacterial overnight culture diluted in TSB to an OD600nm of 0.05. Four replicate wells were conducted per bacterial isolate. Plates were transferred to an incubator for 24-hour static incubation at 37°C to enable biofilm growth. After incubation, the inoculation medium was removed, and excess bacterial planktonic cells were removed from the biofilm by gently washing wells three times with 200 µL of phosphate buffered saline (PBS). Remaining biofilm bacteria that had adhered to the wells were stained using 125 µL of 0.1% crystal violet v/v (CV) solution for 10 minutes at RT and subsequently dried for 1 hour. Washing was then repeated with 200 µL of PBS, performed three times to remove unbound CV. Stained biofilm was then solubilized in 125 µL of 30% (v/v) acetic acid and read at OD550nm using the BioTekTM Synergy TM Mx Multi Detection Top Monochromator Based Microplate Reader with Gen5 Software (Thermo Fisher Scientific, Waltham, MA, USA). Biomass remaining was calculated by comparing the OD550nm readings of phage-treated wells to untreated control wells.

Biofilm infection

To assess biofilm penetration and infection, 1.5 mL tubes were inoculated with 1 mL of bacterial overnight culture diluted in TSB to an OD600nm of 0.05. Six replicate tubes were used for each bacterial isolate. Once loaded, tubes were transferred to an incubator for 24-hour static incubation at 37°C to enable biofilm growth. After incubation, the inoculation medium was removed, and excess bacterial planktonic cells gently removed from the biofilm via washing the wells three times with 1.2 mL of phosphate PBS. Phages suspended in 1 mL of PBS at a concentration of 1 × 10⁹ PFU/mL were then added to the tubes and left in static incubation at 37°C for 6 hours. Post-exposure, free phage were removed by washing three times with 1.2 mL of phosphate buffered saline, as described above. Following this, 1 mL of sterile PBS was added to each tube and sonicated at 20–40 Hz for 15 minutes to dislodge biofilms. Finally, suspended biofilms were serially diluted from 10^–1 to 10^–9 and then plated onto whole plate overlay cultures to determine CFU counts. Remaining bacteria were calculated by comparing the colony counts of phage-treated tubes to untreated control tubes.

Statistics

Statistical analysis

Statistical analyses were conducted using GraphPad Prism v8.4.3 (GraphPad Software, La Jolla, CA, USA). Comparisons containing mixed model analyses were performed using 2-way ANOVA. Plates that did not yield any viable bacterial counts were reported as zero. P-values <0.05 were considered significant for all tests used unless alpha has been specified otherwise. For dosage curve analysis, area under the curve (AUC) analyses were performed.

Results

Phages Koomba kaat 1 and Biyabeda mokiny 1 have desirable genomic characteristics.

Both genomes were successfully reassembled, annotated, and reordered based on the small terminase subunits. Data shown are from the reordered genomes and values were matched to the raw assembly values to check no errors were introduced in the reordering process.

The genome annotation of Biyabeda mokiny 1 reveals a total of 224 coding sequences (CDS), of which 140 are identified as hypothetical proteins. The genome displays a coding capacity of 90.17% and includes tRNAs for the amino acids tryptophan (Trp), phenylalanine (Phe), and aspartic acid (Asp). For Koomba kaat 1, the annotation shows 187 CDS, with 107 hypothetical proteins. This phage has a coding capacity of 88.67% and contains no identified tRNAs. Neither phage were found to contain any lysogenic markers. The annotation process revealed no known integrases or lysogeny-associated markers (Fig. 1).

Screening for resistance genes using AMRFinderPlus (v3.11.20) and ABRicate (v1.0.1) revealed no antimicrobial resistance genes within the genomes of any contigs produced by Biyabeda mokiny 1 and Koomba kaat 1. This included all known Staphylococcal antimicrobial resistance genes. Furthermore, both phages were predicted to have a lytic life cycle based on their genome using Bacphlip v0.9.6 bacphlip (Hockenberry and Wilke 2021) (lysogeny score < 0.2 for both phages).

Identification of receptor binding proteins

Koomba kaat 1 and Biyabeda mokiny 1 were determined to be polyvalent phages with 2 predicted RBPs. Using comparisons to known Staphylococcus phages as a reference, these RBPs matched with known genomes of Staphylococcus phages belonging to the same genus (Takeuchi et al. 2016, Peng et al. 2019). Overall, these phages share highly similar annotation profiles (Fig. 2).

Both Koomba kaat 1 and Biyabeda mokiny 1 had a single CDS score above the default 0.5 required to classify as an RBP detection. The CDS were different between the two phages. The prediction for Koomba kaat 1 (score = 0.968) was a hypothetical protein (ORF 158) product, 3219 bp in length, belonging to a category of unknown function within the PHROGs database. For Biyabeda mokiny 1, the prediction (score = 0.998) was for a tail fibre protein (ORF 42), 1377 bp in length.

Both phages had similar genomic structures to SA012 and MR003 (Fig. 2). The primary receptor binding proteins of SA012 and MR003 were found to be in the same regions as orthologous proteins identified within Koomba kaat 1 (ORF 46 and 48) and Biyabeda mokiny 1 (ORF 40 and 42) (Fig. 2). Orthologous pairs of CDS between Biyabeda mokiny 1 and SA012 were high (phrog_2665 similarity = 87.3%, phrog_2691 similarity = 91.3%), whereas comparison to MR003 was both low (<60%). Orthologous pairs of CDS between Koomba kaat 1 and MR003 were high (phrog_2665 similarity = 91%, phrog_2691 similarity = 99.3%), whereas comparison to SA012 was low (<60%).

Phylogenetic tree (UPGMA)

Koomba-kaat 1 and Biyabeda-mokiny 1 (positions indicated with red and blue stars, respectively; Fig. 1) both cluster alongside Silviavirus and Kayvirus genera when aligned using the UPGMA method. Phages belonging to Silviavirus and Kayvirus to date have all been absent of integrase genes.

Both phages display typical morphological characteristics expected of their genus

Koomba kaat 1 and Biyabeda mokiny 1 plaques did not maintain consistent morphology during the consecutive whole plate overlays required for plaque purification. Despite undergoing three rounds of plaque isolation, whole plate overlays of the final prospective phage samples against their host bacterial strain of S. aureus revealed the presence of multiple plaques with different sizes. Biyabeda mokiny 1 grown using its host MSSA strain SA01 showed turbid, medium-sized plaques (Fig. 3a) with a mean diameter of 1.65 mm (± 0.59). Koomba kaat 1 grown using its host MRSA strain SA09 showed clear, larger plaques (Fig. 3b) with a mean diameter of 2.97 mm (± 0.73).

Under TEM analysis, both phages were found to belong to the myovirus morphotype. Each phage had a typically sized icosahedral head and a long contractile tail. Koomba kaat 1 has an icosahedral head with a length of 93.8 nm (± 1.7) and a diameter of 91.4 nm (± 3.6). Its tail length is 164.1 nm (± 5.7), and it has a tail diameter of 26.1 nm (± 0.9) (Fig. 3c). Biyabeda mokiny 1 has an icosahedral head with a head length of 94.6 nm (± 1.4) and diameter of 96.8 nm (± 0.9). Its tail measures 202.1 nm (± 4.3) in length and has a diameter of 20.2 nm (± 1.3) (Fig. 3d).

Both phages can be recovered from long-term storage and aerosolized without significant losses in activity

Phages Koomba kaat 1 and Biyabeda mokiny 1 were assessed for optimal storage temperature over the course of 1 year. Optimal preservation temperature for both phages was 4°C, followed by −80 and −20°C. After 1 year at room temperature, neither phage was recoverable and so room temperature storage concentrations were removed from further analysis.

Phages Koomba kaat 1 and Biyabeda mokiny 1 were also assessed for their ability to be aerosolized without damage. Both phages were recoverable and maintained their activity at both a high (1 × 10⁹) and low (1 × 10⁶) concentration post-nebulization (Table 1). There were no significant differences between Koomba kaat 1 and Biyabeda mokiny 1 in either their aerosol stability (P = 0.138) or the percentage loss between high or low starting concentrations (P = 0.407).

Both phages effectively kill planktonic S . aureus in vitro

In the host range screen against S. aureus isolates, Koomba kaat 1 successfully infected all four laboratory strains, whereas Biyabeda mokiny 1 only infected one. In the larger clinical strain comparison, Koomba kaat 1 was highly effective at infecting 83% (n = 104/126) of clinical strains. A lower infection rate of 37% was observed for Biyabeda mokiny 1 (n = 47/126). This difference was statistically significant (P = 0.028). When the bacterial isolates were subset based upon their anatomical site of isolation, Koomba kaat 1 had greater activity from each site apart from the non-respiratory tissue S. aureus strains (Fig. 4).

Both phages are able to disrupt and infect biofilms produced by clinical MRSA isolates

The isolates MRSA-1, MRSA-2, MRSA-3, MRSA-4, and MRSA-5 were chosen due to their isolation from the airways of people with cystic fibrosis.

Biofilm disruption results (Fig. 5a) indicated that overall, Koomba kaat 1 was significantly better at disrupting the biofilms produced by clinical MRSA strains (n = 5) when compared to Biyabeda mokiny 1 (P < 0.0001, mean difference = 25.64%). The anti-biofilm activity of Koomba kaat 1 was not 100% effective across the selected MRSA isolates; when assessed individually Koomba kaat 1 was able to significantly reduce biofilms produced from ATCC-6538 (P < 0.001), MRSA-1 (P < 0.001), MRSA-2 (P = 0.004), MRSA-3 (P = 0.004), and MRSA-5 (P = 0.001), but not MRSA-4 (P = 0.111). Biofilms infected by Biyabeda mokiny 1 were observed to be visibly reduced; however only MRSA-1 biofilms were significantly disrupted (P = 0.003). Also noted was a significant increase in the production of ATCC-6538 biofilm when infected with Biyabeda mokiny 1 (P = 0.002).

When assessing the infection of the bacteria within the biofilms (Fig. 5b), Koomba kaat 1 could infect and significantly reduce the viable bacterial load of biofilms produced by ATCC-6538, MRSA-1 (mean reduction: 1.5 × 10⁷ CFU/mL), MRSA-2 (mean reduction: 1.6 × 10⁷ CFU/mL), MRSA-3 (mean reduction: 6.4 × 10⁶ CFU/mL), and MRSA-5 (mean reduction: 6.9 × 10⁶ CFU/mL) (all P values < 0.001), but again did not impact MRSA-4 (mean reduction: 4.3 × 10⁵ CFU/mL) biofilms (P = 0.933). For Biyabeda mokiny 1, viable bacteria were also significantly reduced in biofilms produced by MRSA-1 (mean reduction: 1.5 × 10⁷ CFU/mL, P < 0.001), MRSA-2 (mean reduction: 2.5 × 10⁶ CFU/mL, P = 0.008, Fig. 5b), and MRSA-5 (mean reduction: 2.3 × 10⁶ CFU/mL, P = 0.014). There was a significant increase in the viable bacterial growth of ATCC-6538 when infected with Biyabeda mokiny 1 (mean increase: 5.8 × 10⁶ CFU/mL, P < 0.001).

Discussion

The antibiotic-resistant pathogen MRSA remains a critical global health challenge across both hospital and community settings. Currently, there is an urgent need for alternative antimicrobial strategies to combat MRSA infections, as traditional antibiotics are becoming ineffective against the broad ranges of infections S. aureus can cause. Phages may offer a promising alternative to antibiotics for S. aureus isolated from various clinical sites (Abatangelo et al. 2017, Ajuebor et al. 2018) and the development of phage products against S. aureus infections could offer a much-needed solution to the limitations of current antibiotic pipelines (Morris et al. 2011). For successful therapeutic intervention, phages must have highly effective lytic life cycles, able to reach the site of infection without significant loss of activity, and they must be able to overcome bacterial defences at the site of infection like biofilm (Cairns et al. 2009, Hyman 2019, Glonti and Pirnay 2022). In this work, two phages were thoroughly characterized for these properties. Both Koomba kaat 1 and Biyabeda mokiny 1 were resilient during steps required for therapeutic preparation and delivery. Both phages could be reliably recovered from storage, propagated, and aerosolized without significant reductions in activity. When phages were assayed against MRSA and MSSA from various clinical sources, including respiratory, non-respiratory tissue, non-respiratory fluid, and blood cultures, wide ranges of activity were seen which further corroborate previous data generated for Kayviruses and Silviaviruses targeting Staphylococcus species (Hsieh et al. 2011, Lubowska et al. 2019, Peng et al. 2019, Kitamura et al. 2020).

General morphological characteristics were as expected for lytic Staphylococcus phages assigned to the Herelleviridae family (Fig. 3a and b) (Peng et al. 2019, Kitamura et al. 2020). This, in addition to the variable nature of plaque morphology across different culture conditions (Ramesh et al. 2019), suggests plaque characterization may be necessary. Whole genome sequencing is typically able to determine whether or not a phage culture is contaminated with prophage to a more sensitive degree and for considerably lower costs in recent years (Philipson et al. 2018). An in-depth look at the genomes of Koomba kaat 1 and Biyabeda mokiny 1 revealed clustering alongside multiple phages previously characterized for therapeutic applications (Peng et al. 2019). Leveraging this information, multiple receptor binding proteins belonging to both phages were identified and thus these phages were determined to be polyvalent. Koomba kaat 1 and Biyabeda mokiny 1 are likely to share similar receptor binding proteins (Fig. 2). Both phages are predicted to have two RBPs, and likely share the same receptors of phage SA012 (Fig. 2, phrogs: 2665, 2691) (Terzian et al. 2021), receptors shown to bind to the Wall Teichoic Acid (WTA) backbone or N-acetylglucosamine (ɑ-GlcNAc) residues on the WTA. These receptors were previously described for phage SA012 (Takeuchi et al. 2016) and a similar process of genome comparison was produced for phage MR003 (Peng et al. 2019), whose genome also contained an N-acetylglucoamidase motif (Fig. 2, phrog_10 431) within a similar region that the authors speculate to be responsible for this Silviaviruses broad range of activity. In our analysis these factors were identified independently (RBPs, endolysin motifs) and labelled according to their functional assignments from the PHROGs database (Terzian et al. 2021) (Fig. 2). While phage activity does depend on the ability for phages to recognize and bind to their target bacterium, the infection process can be resisted by bacteria across all steps of the phage life cycle (Labrie et al. 2010).

Data were also obtained to support the safety of using the initial isolation hosts (clinical isolates) as the host propagating strains for these particular phages (Supplementary Table S1). Using these data, in addition to previous warnings flagged throughout the genome assembly pipeline, Biyabeda mokiny 2 [data not shown, but previously published alongside Koomba kaat 1 (Iszatt et al. 2022)] was removed from downstream analysis due to transduction of prophages within phage propagations. Despite the genome itself being of appropriate quality and meeting our coverage checkpoint of >90% phage-mapped reads to the assembly, this was a borderline value, and a more stringent cutoff point of 95% is suggested by others (Philipson et al. 2018). In addition, the unmapped reads that were assembled formed multiple large contigs that were likely from host contamination. A limitation of this procedure is the need to re-purify and characterize phages within different isolation hosts, or use a prophage-removed strain of S. aureus such as RN4420; however the efficiency of phage replication within this strain was noticeably lower (data not shown). It has been demonstrated previously that the use of RN4420 can be used as a high-efficiency propagation strain using an adapted growth protocol to improve lytic efficiency and provide homogenous populations of phages (Nirmal Kumar et al. 2012). While this might be useful in producing prophage contaminant-free samples, the resulting phage progeny could harbour differences from the originating phage and require recharacterization. This is because the host bacteria may impart a selection pressure onto the phage or, more deliberately, impart epigenetic changes to the phage genome that will affect its ability to infect other S. aureus bacteria effectively (Moller et al. 2019, 2021).

Koomba kaat 1 demonstrated a much broader range of activity against S. aureus than Biyabeda mokiny 1 (Fig. 4), yet unlike Biyabeda mokiny 1, Koomba kaat 1 was not able to infect any other Staphylococcal species tested. Unfortunately, the low number of non-aureus Staphylococci within this bacterial repository (n = 5) prevented further work to truly understand the range diversity. However, it could be possible for the ability of Biyabeda mokiny 1 to infect S. xylosus to be leveraged as a non-pathogenic propagation host in manufacturing. This approach has been reported using a food-grade S. xylosus strain previously, a key benefit of which is avoiding the concerns around large-scale propagation (required for clinical translation of any phage) using large amounts of a human pathogenic bacterium (Gonzalez-Menendez et al. 2018). Outside of the Staphylococcus genus, neither phage was able to infect any of the other bacterial species assayed. Collectively these data are supported by what is commonly seen in the literature for Staphylococcus infecting Kayvirus and Silviavirus phages (Leskinen et al. 2017, Lubowska et al. 2019, Peng et al. 2019).

The antibiofilm activity of these phages was also demonstrated using an adapted form of a previously published method of in vitro biofilm production (Neopane et al. 2018). Koomba kaat 1 and Biyabeda mokiny 1 were able to differentially disrupt and infect the biofilms produced by clinical MRSA isolates when grown in culture (Fig. 5). We observed that in our set of MRSA clinical isolates, Koomba kaat 1 had a broader range of activity against their biofilms. Previous literature reporting the ability for Silviaviruses to prevent regrowth from biofilm corroborates this; however, comparisons are difficult as the authors did not quantify the exact biofilm CFU within the study (Kebriaei et al. 2023). Biyabeda mokiny 1 was also able to infect biofilms and effectively reduce biofilm CFU counts but had less of an ability to disrupt the biofilms biomass (Fig. 5, panels a and b). Despite this, activity has already been reported for Kayvirus phages, including the well-studied Phage K representative of this genus (Kelly et al. 2012). The differences in antibiofilm activity between these phages may be attributed to the presence of multiple tail-associated lysins (GenBank protein ID: UXE02954.1) identified within the Koomba kaat 1 genome that are absent from Biyabeda mokiny 1. These enzymes are known to play a role in biofilm degradation against some bacterial species (Chang et al. 2022). Tail-associated lysins may enhance Koomba kaat 1’s ability to disrupt biofilm structure and also contribute to its ability to infect the bacterial cells within. Our data suggests both phages would be able to reduce the bacterial viability of biofilm-embedded communities in addition to reducing the overall physical biomass of biofilms produced by these bacterial strains. However, our data is limited by assessing biofilm produced in vitro in the absence of tissue, which may not fully reflect biofilm encountered in vivo. Future work should attempt to model this with cell culture-based model and include the use of confocal microscopy (Grossman et al. 2021, Turner et al. 2024) to visualize biofilm composition or structure, which could help to clarify why biofilm activity is specific to individual bacterial isolates biofilm.

The stability of both phages when aerosolized was high, averaging a 2.5% reduction in viable titre for Koomba kaat 1 and 12.5% average reduction in viable titre for Biyabeda mokiny 1 across both concentrations tested (Table 1). When compared to previous literature regarding mesh-type nebulization; the effects of nebulization on a P. aeruginosa infecting phage (PEV44) reported a 50%–60% increase in ‘broken’ phage particles (Astudillo et al. 2018). While there are no studies looking at the effects of nebulizing Staphylococcus phages specifically in vitro, phages have been nebulized previously with successful treatment of MRSA infections in rats (Prazak et al. 2022). In more in vivo studies, it has been shown within macaque monkeys (Le Guellec et al. 2023) that nebulizer type (mesh, jet, ultrasonic) had no effect on phage viability and that large amounts of phage can be delivered using nebulization (Guillon et al. 2021). Overall, there were no significant drops in concentration during nebulization for either phage within this study and based on previous use cases of Staphylococcus phages in vivo, there is a good chance aerosol viability will remain high in vivo (Prazak et al. 2022).

Concern around bacterial resistance to phages is a widely discussed topic, with numerous manuscripts reporting host-pathogen interactions between bacteria and their phages (Labrie et al. 2010, Coulter et al. 2014, Drilling et al. 2017, Moulton-Brown and Friman 2018, Oechslin 2018, Bernheim and Sorek 2020, Bodner et al. 2020, Broniewski et al. 2020). The consensus seems to be that while phage usage does lead to resistance in vitro, these effects have fitness costs associated with them such as increased susceptibility to antibiotics and/or host immune factors (Oechslin 2018), which may prevent their translation into in vivo or clinical studies. General understanding of these interactions is developing at a rapid rate, and recent data have shown a hierarchical nature of resistance in certain bacterial species (Moller et al. 2019), or modular cross-resistance reported in others (Wright et al. 2018). While more data are required to determine the complexity and reach of these effects, and whether they are relevant in vivo, it is clear that these interactions are not only phage-host specific, but also specific to their host microenvironment where the immune system may play a role for or against the introduction of phage (Sunagar et al. 2010, Shivshetty et al. 2014, Oechslin 2018). With this in mind, future work that may make use of the data within this study may include the acquisition of bacterial genome sequencing data to accompany the lytic profiles generated for Koomba kaat 1 and Biyabeda mokiny 1. An advantage of sequencing the bacterial isolates is the standardization of information since, to the authors knowledge, there is no standardized host range panel for S. aureus bacteria. Attempts have been made for other bacterial genera (De Soyza et al. 2013) but this was not widely adopted. Furthermore, the use of laboratory strains may not accurately represent the bacteria causing infections today. By treating WGS of bacterial isolates as a crucial aspect of host range and lytic efficiency data, analyses that compare ranges of activity across the wider literature may be a possible future.

Overall, this work has demonstrated that these phages are able to be recovered from storage, grown, purified, and withstand the effects of mesh nebulization with minimal loss in viable activity. The data provided here provides strong evidence that these phages may be used to great effect in targeting S. aureus from a range of clinical sources, including S. aureus isolates that generate biofilm. Future work should validate Koomba kaat 1 and Biyabeda mokiny 1 are safe for human therapeutic use.

Acknowledgements

Member institutions of the Western Australian Epithelial Research Program (WAERP) include the Wal-Yan Respiratory Research Centre, Kids Research Institute Australia, The University of Western Australia, Perth, Western Australia, Australia, and St John of God Hospital, Subiaco, Western Australia, Australia. We acknowledge that this project was conducted on the traditional homelands of the Noongar people, with phages isolated from waters across Noongar Wadjak. The respiratory Staphylococcus aureus clinical isolates for host range assays were generously provided by Scott Bell Queensland Institute of Medical Research (QIMR Berghofer, Queensland, Australia). All non-respiratory Staphylococcus aureus clinical isolates were generously provided by Professor Geoff Coombs from Murdoch University. The transmission electron microscopy was performed by Dr. Christopher Leigh at the University of Adelaide.

Author contributions

Joshua J. Iszatt (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing), Alexander N. Larcombe (Data curation, Formal analysis, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing), Luke W. Garratt (Data curation, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing), and Stephen M. Stick (Funding acquisition, Project administration, Supervision), Anthony Kicic (Funding acquisition, Project administration, Supervision, Writing – review & editing)

Conflict of interest

None declared.

Funding

This study was supported by a Telethon—Perth Children’s Hospital Research Fund, a Perpetual IMPACT Philanthropy Grant (#IPAP2018/0704), a Wesfarmers Centre of Vaccines and Infectious Diseases Seed Funding Grant, and an NHMRC 2021 MRFF Chronic Respiratory Conditions Grant (#2023559). JJI was supported by a Curtin University PhD scholarship. ANL was partially supported by Curtin University. LWG was supported by a NHMRC Early Career Fellowship (#141479) and a Department of Health Research Excellence Award. AK is a Rothwell Family Fellow at the Kids Research Institute Australia. The contents of the published material are solely the responsibility of The Kids Research Institute Australia or the individual authors, they do not reflect the views of the Commonwealth.

Data availability

The empirical data underlying this article will be shared on reasonable request to the corresponding author. The genomics data underlying this article are available in GenBank and can be accessed with accession numbers: OP263967 and OP263969 under the BioProject accession: PRJNA862682.

References