Background

The Pseudomonas genus includes important plant pathogens [1], agents of food spoilage [2] and opportunistic pathogens to animals and humans [3]. Previous metagenomic analyses have shown that Pseudomonas is the predominant bacterial genus found on retail food, including seafood and foods of plant and animal origin [4]. Bacteria on food may also carry antimicrobial resistance genes (ARGs), which may be transferred to other bacteria on food [5]. To date, few studies have described the genetic diversity of Pseudomonas on food and how this contributes to the clinical burden of disease either as a pathogen or as a potential reservoir of ARGs [6].

A total of 313 Pseudomonas species have been described and published [7], but the most clinically significant species is Pseudomonas aeruginosa, an opportunistic pathogen of humans, commonly associated with nosocomial infections [8], burn wound infections [9] and pneumonia in cystic fibrosis patients [10]. P. aeruginosa infections are difficult to treat as the species contains multiple intrinsic antimicrobial resistance (AMR) mechanisms, and readily adapts to antimicrobial pressure by accumulating mutations or acquiring ARGs via horizontal gene transfer [11]. Sources of P. aeruginosa infections are hard to identify due to the ubiquitous nature of the bacterium [12], which is found in soil and water environments, particularly those associated with human activity [13].

P. aeruginosa has also been isolated from diverse food types [14, 15] and P. aeruginosa from foods in hospitals has been identified as a potential source of nosocomial infections [16]. P. aeruginosa has been shown to translocate from the gastrointestinal tract to the lung [17], which could precede these infections.

Whole genome sequencing (WGS) is the most discriminatory method available to distinguish bacteria, and can be used to identify and characterize the genetic content of isolates, and assess putative sources of infections [18]. The aim of this study was to understand the potential role of Pseudomonas on food as a reservoir of ARGs for foodborne pathogens and to determine the prevalence of potentially clinically significant P. aeruginosa strains on food.

Results

Pseudomonas spp. analysis

We initially attempted to isolate P. aeruginosa from food by non-specifically culturing for Pseudomonas spp. Pseudomonas spp. presence was tested in 32 food samples and was isolated from all leafy green (8/8) and prawn samples (4/4), 7/8 of chicken, 7/8 pork samples, and 2/4 salmon samples. A total of 93 Pseudomonas isolates were cultured and sequenced from the 28 positive samples (Supplementary Table 1).

The Pseudomonas species cultured from food belonged to the Pseudomonas fluorescens, P. fragi, P. koreensis, P. putida, P. trivialis and P. veronii species and their presence on different food commodities (Table 1) were similar to that observed in previous studies (Supplementary Table 2). However, P. aeruginosa was not isolated from any of these food samples using this methodology, which utilized the ISO 13720 standard. There was species diversity amongst isolates collected from the same sample, with two different Pseudomonas species isolated from 15 samples, three different Pseudomonas species isolated from two samples, and five different Pseudomonas species isolated from one sample (Fig. 1). There was also diversity within species, with 15 samples containing isolates that belonged to different phylogenetic clades within the same species (Supplementary Figs. 1–4).

Table 1 Number of food samples from each commodity that cultured for different Pseudomonas species
Full size table
Fig. 1
figure 1

Pseudomonas species isolated from different food commodities. Number of Pseudomonas isolates analyzed from each food commodity, colored by species

Full size image

Of the 93 Pseudomonas spp. genomes analyzed, none contained known ARGs or plasmid replicons. Two P. fragi genomes contained five virulence genes (mrkA, mrkB, mrkC, mrkD and mrkF); virulence genes were not identified in the other Pseudomonas spp. genomes. The PATRIC database yielded 33 P. fluorescens, 10 P. koreensis and 17 P. putida genomes; similar results were obtained with these publicly available genomes, with only a single P. fluorescens genome containing a plasmid replicon and four P. putida isolates containing ARGs.

P. aeruginosa analysis

As we were unable to isolate P. aeruginosa using non-specific Pseudomonas spp. culturing, we specifically cultured for this bacterial species by increasing the incubation temperature and using a different medium. P. aeruginosa presence was evaluated in 128 food samples and was isolated from 0/21 beef, 7/50 chicken, 1/18 lamb, 5/19 leafy greens, 0/10 pork and 1/10 salmon samples (Fig. 2, Supplementary Table 3), but these proportions were not significantly different (Fisher’s exact test: p = 0.101). Of the 19 leafy green samples tested for P. aeruginosa, 13 were labelled as “washed”, three were labelled were labelled as “wash before use”, and for three it was unknown. P. aeruginosa was only cultured from washed leafy green samples.

Fig. 2
figure 2

Prevalence of Pseudomonas aeruginosa found on different food commodities. Number of food samples belonging to each commodity that were cultured for and tested positive for P. aeruginosa

Full size image

The 56 P. aeruginosa isolates recovered belonged to 16 unique STs, but two of the STs had not been described before so were given the temporary names Novel-A and Novel-M (Supplementary Table 4). For one lamb and one leafy green sample, two STs were identified, whilst for the remaining twelve positive samples only a single ST was identified (Fig. 3).

Fig. 3
figure 3

Phylogenetic relationship between P. aeruginosa isolates collected from retail food samples. Maximum likelihood tree of the 56 P. aeruginosa isolates collected from food samples, colored by sample, food type (commodity) and sequence type (ST), along with a presence-absence matrix of the antimicrobial resistance genes (ARGs) they contained. The phylogenetic branch lengths are given in nucleotide substitutions per site, therefore a branch of length 0.0007 (as represented by the scale bar) equates to 3,356 substitutions, given that the core gene alignment consisted of 4,794,505 bp

Full size image

The 56 food-derived P. aeruginosa were compared to 606 publicly available genomes of the same species for context (Fig. 4). These isolates were mostly collected from human, industrial and environmental sources, predominantly from Europe and North America. This analysis showed food isolates were similar to isolates across the phylogeny of the context panel. All food isolates of P. aeruginosa in this study contained aph(3)_IIb, blaOXA, blaPDC, and fosA genes at similar frequencies to the context isolates: 99.7%, 99.5%, 99.0% and 84.2%, respectively. All food isolates contained the catB7 gene, except for those belonging to ST-253 and novel ST-M; 84.7% of the context P. aeruginosa genomes investigated contained this gene. Food isolates belonging to ST-17, ST-236, ST-253, ST-3700 and one of the ST-319 isolates contained the crpP genes, as did 61% of the context P. aeruginosa investigated. P. aeruginosa from one food sample belonging to ST-1212 contained the aac(3)_IVa and aph(4)_Ia genes, whilst none of the context P. aeruginosa contained these ARGs.

Fig. 4
figure 4

Phylogenetic relationship between Pseudomonas aeruginosa isolates collected from retail food samples and publicly available genomes. Maximum likelihood tree of 662 P. aeruginosa isolates, colored by source and continent of origin. The phylogenetic branch lengths are given in nucleotide substitutions per site, therefore a branch of length 0.009 (as represented by the scale bar) equates to 26,141 substitutions, given that the core gene alignment consisted of 2,904,504 bp

Full size image

None of the P. aeruginosa from food contained any known plasmid replicons, whilst only 2.81% of the 606 context P. aeruginosa genomes did so.

Amongst the 56 food and 606 context P. aeruginosa genomes, 239 virulence genes were identified. Individual isolates from food contained 192–231 virulence genes, similar to the context P. aeruginosa investigated: 152–236. There was a large amount of variation in the presence of virulence genes found amongst P. aeruginosa from each of the sources. The proportion of human and food isolates that contained each virulence gene were compared (Supplementary Fig. 5); the pchE, pchG, pchH, ppkA, pscH and tagF/pppB genes were associated with P. aeruginosa isolates from humans, whilst mucA, pvdM and pvdN genes were associated with P. aeruginosa from food (Supplementary Table 5).

P. aeruginosa within-sample diversity

All P. aeruginosa isolates collected from the same food sample belonged to one or two STs; isolates belonging to the same ST and sample were not identical, differing by up to 19–67 core non-recombinant SNPs and 5–29 virulence genes (Table 2). 71 virulence genes varied amongst isolates belonging to the same ST and sample.

Table 2 Diversity of P. aeruginosa isolates collected from the same food sample
Full size table

P. aeruginosa clinical isolates

Of the 16 STs within the 56 food P. aeruginosa genomes, publicly available genomes from human samples were available for four: ST-17, ST-253, ST-319 and ST-699. The minimum number of core non-recombinant SNPs between the isolates from human and food samples for these STs was 41–312 (Supplementary Table 6). The most closely related human isolates differed by 0–1 ARGs, no plasmid replicons and 3–6 virulence genes, and were collected between 1991 and 1997 from France and Turkey, or this information was not available.

Virulence of P. aeruginosa using the Galleria mellonella larvae infection model

Results from the Galleria mellonella larvae infection model varied between replicates (Supplementary Fig. 6; Supplementary Table 7). Of the 16 P. aeruginosa STs collected from food 15, were more virulent than the low virulence control (PAO1) (Supplementary Fig. 6); these differences were not statistically significant (Supplementary Table 8). None of the strains isolated were as virulent as the high virulence control (PA14).

Discussion

The Pseudomonas genus has been described as one of the most ubiquitous bacterial genera found in environmental, human and animal sources globally [19]. This was evident when we cultured 88% of retail food samples for this bacterial genus, but did not identify any isolates belonging to the P. aeruginosa species, suggesting that although Pseudomonas make up a large portion of the food metagenome [4], P. aeruginosa is not a large contributor. The six other Pseudomonas species isolated from food varied in the types of food from which they were isolated (Supplementary Table 2). There were some differences to previous studies; for examples, we isolated P. koreensis from prawns, and, to the best of our knowledge, this is the first time this species has been isolated from seafood; we also failed to isolate P. putida from any seafood samples or P. trivialis from any leafy greens, even though they have previously been reported from similar sources [20, 21]. However, we only examined 32 food samples with the ISO 13720 standard method, and may find particular Pseudomonas species on other food types with further sampling.

The risk of human infection varies by Pseudomonas species. P. fluorescens has been associated with infections, but usually these are in immunocompromised patients associated with contaminated pharmaceuticals [22]. P. koreensis was previously isolated from a keratitis case but was part of a mixed infection with Aspergillus fumigatus, likely the result of a contaminated contact lens [23]. P. putida has been isolated from bloodstream infections, but these were assumed to be nosocomial infections [24]. To the best of our knowledge, P. fragi, P. trivalis and P. veronii have not been isolated from any clinical samples. Therefore, the presence of these non-aeruginosa Pseudomonas species on food is unlikely to be a direct health concern. Furthermore, they did not contain any known ARGs. This suggests that although the non-aeruginosa Pseudomonas comprise a large proportion of the bacteria found on food, they are unlikely to act as a major reservoir of ARGs for pathogens.

The potential impacts of non-aeruginosa Pseudomonas spp. on food goes beyond clinical concerns; the genus is recognized as a major cause of food spoilage [25]. Of the Pseudomonas species identified in this study, P. fluorescens is the best described cause of food spoilage, having been associated with spoilage in a wide range of food types from fruits and vegetables, dairy products, meat and seafood [26]. In addition, P. fragi, P. koreensis and P. putida have also been found to cause the spoilage of specific meats and seafoods [27, 28]. A mrk gene cluster was identified in two P. fragi genomes cultured from food; this gene cluster originates from Klebsiella and encodes type III fimbriae that are used for cell adhesion and biofilm formation [29]. In Enterobacterales, the mrk gene cluster is associated with increased biofilm formation and isolates from catheter-associated urinary tract infections [30]. P. fragi has not been isolated from clinical samples, but improved biofilm formation could facilitate persistence during food processing. The high prevalence and diverse populations of these bacteria on the food types described in this study highlights the need to identify how food becomes contaminated with Pseudomonas, what food processing techniques facilitate or prevent their contamination, and how the relative proportions of the Pseudomonas species change over the shelf-life of food.

Pseudomonas spp. was isolated from 88% of food samples using the ISO 13720 standard method, but P. aeruginosa isolation required different culture conditions, including raising the incubation temperature from 25 ̊C to 37 ̊C, a 24-hour enrichment step and a more selective medium (Pseudomonas centrimide (CN) instead of Pseudomonas cephalothin-sodium fusidate-cetrimide (CFC) agar). Despite these adjustments, P. aeruginosa was still only isolated from 11% of food samples. P. aeruginosa seems to make up a small proportion of the Pseudomonas spp. found on food, if it is present at all. It is possible that P. aeruginosa could have been recovered using the ISO 13720 standard method if we had analyzed more isolates from each sample. However, the number required to reach the sensitivity of the CN agar-37 ̊C method remains to be determined.

P. aeruginosa has been isolated from a wide range of human clinical samples [8,9,10], but source of infections are hard to identify due to the ubiquitous nature of the bacterium [12]. Wheatley et al. [17] demonstrated that P. aeruginosa from the gastrointestinal tract could travel to colonize the lungs of a patient, showing that carrying a reservoir of P. aeruginosa in the gut can be a risk factor for serious disease. Although P. aeruginosa has been found in public water supplies, representing one route of transmission [31], our data suggest gastrointestinal colonization could also be due to contaminated food as we isolated P. aeruginosa from 11% of food samples of different types. Four out of 16 P. aeruginosa STs isolated were within 41–312 SNPs to an isolate from a human clinical sample [32,33,34,35]. This indicates that humans are regularly exposed to P. aeruginosa isolated from food and this exposure has the potential to lead to opportunistic infections in vulnerable people.

The G. mellonella larvae model is a simple model for investigating P. aeruginosa virulence [36]. P. aeruginosa kills G. mellonella larvae very effectively, particularly in comparison to other pathogens [37]. For this reason, approximately three P. aeruginosa cells were inoculated into each larva. There were no significant differences in larvae survivability between the P. aeruginosa strains isolated from food and the low virulence control strain, but there was variation in survivability between replicates. More consistent methods are required to compare the virulence of these P. aeruginosa strains.

P. aeruginosa was isolated from chicken, lamb, leafy greens and salmon samples. No beef or pork samples were positive, but further sampling may identify P. aeruginosa from these commodities. Wong et al. [38] cultured eight chicken and eight pork samples for meropenem-resistant Pseudomonas, but only isolated P. aeruginosa from one of the pork samples. Most of the food samples investigated in this study consisted of meat and seafood samples that are likely to be cooked prior to consumption. However, leafy greens were the food commodity with the highest proportion of P. aeruginosa isolates, and they are unlikely to be further processed prior to consumption, especially as the five leafy green samples from which P. aeruginosa was cultured were already washed and one of these samples contained an ST previously identified in human clinical samples. Further studies are required to investigate P. aeruginosa from a wider range of food commodities to obtain better insight into which food types are most frequently contaminated with the bacterium, if the P. aeruginosa isolated are associated with clinical infections and what factors influence its prevalence on food, such as geographical origin, cut of meat and storage temperature.

The collection of multiple P. aeruginosa isolates from each food samples helped determine within-sample bacterial diversity. Multiple STs were collected from two samples, and isolates belonging to the same ST and sample differed by up to 19–67 core non-recombinant SNPs, along with variable presence of ARGs and virulence genes. Most P. aeruginosa strains have a substitution rate of 4.3 × 10− 6 – 1.0 × 10− 5 substitutions site− 1 year− 1, equating to 27–63 SNPs per year, although strains with faster substitution rates have been described [39]. Therefore, it would take 3 to 30 months for isolates collected from the same food sample to accumulate the number of SNPs identified. This is possible with salmon that are usually harvested after 24 months [40] and lamb that are usually slaughtered at 2–6 months [41], but not leafy greens such as lettuce that are usually harvested after 2–3 months [42] or chicken that is usually slaughtered at 1–2 months [43]. Therefore, for leafy greens and chicken samples, and samples with multiple STs identified, they were likely contaminated with a heterogenous population of P. aeruginosa or were contaminated with P. aeruginosa at multiple time points. This complicates P. aeruginosa source attribution studies as multiple isolates will need to be collected from food to capture ST and SNP diversity.

P. aeruginosa contain multiple virulence genes that allow them to cause opportunistic infections. These include genes involved in biofilm formation, quorum sensing, intracellular survival, and acquiring nutrients [44]. We identified three virulence genes associated with a larger proportion of food isolates compared to human isolates: mucA, pvdM and pvdN. mucA encodes a negative regulator of alginate production, and mucA mutations are associated with a mucoid P. aeruginosa phenotype [45]. In cystic fibrosis patients, mucoidal P. aeruginosa are often selected for via inactivation of the mucA gene [46], which could explain why a smaller proportion of human isolates contained an intact version of this gene compared to food isolates, as it is likely that many human clinical isolates were derived from cystic fibrosis patients. pvdM encodes an enzyme essential for the production of the siderophore pyoverdine [47], whilst pvdN encodes an enzyme that modifies pyoverdine, but the effect of this modification is not clear [48]. O’Brien et al. [49] investigated P. aeruginosa from long-term cystic fibrosis patients and found many have subpopulations of this bacterium with reduced pyoverdine. Therefore, the lack of the pvdM and pvdN genes in many human isolates may have been selected for in chronic infections. In addition, six virulence genes were identified that were found in a larger proportion of human isolates than food: pchE, pchG, pchH, ppkA, pscH and tagF/pppB. pchE [50] and pchG [51] encode enzymes essential for the production of pyochelin, a siderophore that contributes to pathogenicity, whilst pchH is found in the same operon as these genes but its exact function is unknown. ppkA encodes a serine/threonine protein kinase and knockout studies found it is associated with biofilm formation, pyocyanin production, tolerance to oxidative and osmotic stresses, and host cell invasion [52]. pscH encodes a protein involved in the type III secretory system of P. aeruginosa involved in virulence by injecting effector proteins into host cells [53]. tagF/pppB regulates the type VI secretion system that is used to deliver effector proteins into eukaryotic host cells or bacterial competitors [54]. P. aeruginosa are involved in a wide range of opportunistic infections, and the lack of these virulence genes in many food isolates may prevent them forming certain opportunistic infections if given the chance.

P. aeruginosa infections are difficult to treat due to their large number of inherent and acquired AMR mechanisms [11]. These mechanisms were evident with those collected from food, where all isolates contained aph(3)_IIb, blaOXA_50, blaPDC, and fosA genes, whilst they varied in the presence of the aac(3)_IVa, aph(4)_Ia, catB7 and crpP genes, similarly to the other P. aeruginosa genomes investigated. However, as no P. aeruginosa were isolated when non-specifically culturing for Pseudomonas and the Pseudomonas species that were contained no ARGs, the Pseudomonas genus does not likely contribute much to the ARG reservoir of food.

Conclusions

P. aeruginosa isolates likely to be able to cause human infections are present on a relatively small but still substantial proportions of retail foods examined. Whilst the P. aeruginosa isolates contained a multidrug resistant genotype, the most frequent non-P. aeruginosa recovered carried no ARGs and are more likely to play a role in food spoilage rather than infection. Given the frequency of exposure, the potential contribution of food to the burden of P. aeruginosa infections in humans should be evaluated more closely.

Materials and methods

Pseudomonas spp. culturing

Retail food samples used for Pseudomonas spp. culturing were collected as part of a previously reported repeated cross-sectional study of retail food in Norfolk, UK [55]. Chicken (n = 8), leafy greens (n = 8), pork (n = 8), prawn (n = 4) and salmon (n = 4) products were collected on 25/11/2019. Pseudomonas spp. were cultured from these samples using a method adapted from ISO 13720 [56]. For each food sample, 100 g was aseptically transferred into a sterile filtered stomacher bag (Corning, New York, USA). All sample types were homogenized in 225 ml of buffered peptone water (BPW) at 100 rpm for 30 s (Seward stomacher 400 C laboratory blender, Worthing, UK). For samples containing bones or shells, homogenization was performed manually for two minutes. A 10 µl loopful of stomached food was inoculated onto CFC agar (Oxoid, Basingstoke, UK) and streaked for single colonies. Inoculated CFC agar plates were incubated at 25 ̊C for 48 h. Eight colonies were subcultured onto separate tryptic soy agar (TSA) plates (Trafalgar Scientific Ltd., Leicester, UK), streaked for single colonies and incubated at 25 ̊C for 48 h. An oxidase test (Oxoid) was performed on all colonies on TSA and up to five of those that tested positive underwent WGS to confirm they were Pseudomonas spp.

P. aeruginosa culturing

Retail food samples used for specific P. aeruginosa culturing were collected in Norwich, Norfolk, UK. Beef (n = 21), chicken (n = 50), lamb (n = 18), leafy greens (n = 19), pork (n = 10) and salmon (n = 10) were collected between 14/03/2021-30/09/2022. 100 g of each food sample was aseptically transferred into a sterile filtered stomacher bag. All sample types were homogenized in 225 mL of BPW at 100 rpm for 30 s. Stomached food samples were pre-enriched by incubating them at 37 ̊C for 24 h in BPW. A 10 µL loopful of pre-enriched food was inoculated onto Pseudomonas CN agar (Oxoid) and streaked for single colonies. Inoculated Pseudomonas CN agar plates were incubated at 37 ̊C for 48 h. Four colonies morphologically consistent with being P. aeruginosa (producing green/blue pigment) were subcultured onto TSA plates (Trafalgar Scientific Ltd.) and streaked for single colonies, before being incubated at 37 ̊C for 24 h. An oxidase test (Oxoid) was performed on all colonies cultured on TSA and isolates that tested positive were assumed to be P. aeruginosa. WGS was used to confirm P. aeruginosa identity.

Whole genome sequencing

DNA was extracted using the Maxwell® RSC Cultured Cells DNA Kit (Promega, Madison, Wisconsin, USA) using the manufacturer’s instructions. Libraries were created using the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, California, USA) and sequenced on a NextSeq 550 System (Illumina) as 2 × 150 bp paired-end reads.

Genomic analysis

Genomic analyses were performed using the Cloud Infrastructure for Microbial Bioinformatics (CLIMB) [57]. Raw paired-end reads were trimmed using Trimmomatic v0.36 [58] (Supplementary material) and assembled using Spades v3.11.1 [59] in “careful” mode. Centrifuge v1.0.3 [60] was used to predict Pseudomonas species. The quality of genome assemblies was assessed using QUAST v4.6.3 [61], CheckM v1.1.2 [62] and by aligning reads to the assemblies using the Burrows-Wheeler aligner (BWA) v0.7.17 [63]. Assemblies were accepted if they consisted of less than 500 contigs that were over 500 bp, less than 50 duplicate genes and had a mean read depth of the four largest contigs above 30. ARGs, virulence genes and plasmid replicons were identified using ARIBA v2.14.4 [64] and the NCBI AMR [65], virulence finder database (VFDB) [66] and PlasmidFinder [67] databases, respectively.

Context collection of Pseudomonas spp.

The PATRIC database [68] was searched for genomes belonging to the Pseudomonas species identified in this study. These genomes were downloaded and the quality checked. Those that passed QC had ARGs, plasmid replicons and virulence genes identified. To compare Pseudomonas spp. genomes belonging to the same species, the genomes were annotated using Prokka v1.13 [69], Roary v3.11.2 [70] was used to cluster the genes of each genome using a 95% identity cut-off and classifying genes that were found in 95% of isolates as core. RAxML v8.2.4 [71] was used to form a maximum likelihood tree from the core gene alignments of each species using a generalized-time reversible (GTR) substitution model [72]. TreeCluster v1.0.3 [73] was used to predict clades from the maximum likelihood trees for these species using a 0.02 length threshold.

P. aeruginosa sequence type analysis

The sequence types (STs) of P. aeruginosa genomes were predicted in silico using MLST v2.16.1 (https://github.com/tseemann/mlst). Single nucleotide polymorphisms (SNPs) were identified using an alignment-based approach using NC_002516 as the reference genome. Phaster [74] was used to identify prophage regions in the reference genome and phage regions were blocked out. Snippy v3.1 (https://github.com/tseemann/snippy) was used to align P. aeruginosa trimmed reads from each ST to the prophage-free reference. Gubbins v2.3.1 [75] was used to remove SNPs putatively associated with recombination and RAxML was used to generate a maximum likelihood tree based on non-recombinant SNPs using a GTR substitution model.

Virulence of P. aeruginosa using the Galleria mellonella larvae infection model

The Galleria mellonella larvae infection model has been used previously to investigate the virulence of P. aeruginosa [76]. Here, it was used to compare 16 selected food isolates to known virulent control strains PA14 and PAO1 [77]. Preliminary investigations determined the LD50 (lethal dose required to kill 50% of larvae per replicate after 48 h) of the less virulent control strain PAO1 was approximately three colony-forming units (CFU) per inoculum. Fresh bacterial cultures were normalized to an optical density of 0.1 at 600 nm and diluted in sterile phosphate-buffered saline (PBS) (Sigma-Aldrich, St Louis, MO, USA) to achieve the LD50 inoculum concentration. The inoculum concentrations were confirmed by plating 50 µL of inoculum on Luria-Bertani (LB) agar (ThermoFisher Scientific, Waltham, MA, USA), incubating the plates at 37 °C and counting the colonies the following day. The wax moth larvae (Livefood UK, Rooks Bridge, UK) chosen for infection were similarly sized with no signs of pupation or melanization. Each larva was injected with 10 µL of inoculum using a 10 µL syringe (Hamilton, Reno, NV, USA) into the left penultimate proleg. Syringes were sterilized with 70% ethanol (VWR International, Radnor, PA, USA) and washed with sterile PBS between each infection. Each strain was injected into 10 larvae and the percentage of surviving larvae was determined every 3–17 h over 48 h. Controls included ten larvae injected with sterile PBS only and ten uninjected larvae. The infection model was repeated on three independent occasions.

Statistics and reproducibility

Fisher’s exact test was used to determine if there were any associations between the proportion of samples that cultured positive for P. aeruginosa and food commodity.

For each virulence gene, the proportion of P. aeruginosa isolates from human and food sources that contained them were calculated. Two-population t-tests were used to determine if the proportions were significantly different. Bonferroni’s correction for multiple hypothesis testing was used to take into consideration the total number of virulence genes investigated (p = 0.0002).

The survivorship of G. mellonella larvae was modelled using a mixed-model logistic regression model using the package glmmTMB v1.1.8 [78] in R v4.1.2 [79], with strain and replicate as fixed effects and the interaction between run and strain as a random effect.