A collection (n = 253) of geographically diverse Pseudomonas aeruginosa isolates from dogs with otitis externa were collected from the UK Royal Veterinary College (RVC; n = 48), CAPL Nationwide Laboratories, UK (n = 99), and an EU collection from the University of Copenhagen (n = 106). Confirmation of P. aeruginosa identity was performed as described by the UK Health Security Agency standards for microbiology investigations (UKHSA, 2024) using cetrimide agar and the oxidase test, followed by 16S PCR. Pseudomonas aeruginosa PAO1 was used as a positive control for all assays in this study.

Prior to each experiment, agar plates were prepared [Lysogeny Broth (LB), Miller, 25 g/L, Sigma-Aldrich; supplemented with agar, 15 g/L, Fisher Scientific], and were inoculated with the required P. aeruginosa strain and streaked to obtain single colonies. Plates were inverted and incubated statically at 37° C for 16 h. Following incubation, 10 mL of LB was inoculated with a single colony using a sterile inoculating loop and incubated at 37° C with shaking at 200 rpm for 16 h.

Antimicrobial sensitivity testing (AST) was performed as described by the Clinical & Laboratory Standards Institute (CLSI) using the disk diffusion method (CLSI, 2012). A suspension of bacterial cells was prepared by inoculating 8 mL of Mueller-Hinton broth (MH; Sigma-Aldrich) with bacterial colonies selected from an LB plate until a turbidity equivalent to a 0.5 McFarland standard had been reached, as determined by visual comparison using a Wickerham card. A swab was then used to inoculate the surface of a MH agar 2 plate (Sigma-Aldrich) ensuring the entire plate is covered evenly by rotating the plate 90° between swabbing the first time with a final diagonal swab. The plates were then incubated at 37°C for 16 h, and the results recorded by measuring the inhibition zone diameter using a digital calliper.

AST was performed using an antimicrobial panel comprising six antimicrobial classes (Supplementary Table 1). All antimicrobials were tested against all strains except for Ticarcillin + Clavulanic acid which was discontinued part way through this study, and were acquired from two suppliers: ProLabs and Oxoid. E. coli ATCC 25922 and P. aeruginosa ATCC 27853 were the control strains used in this assay. Antimicrobial sensitivity was determined using breakpoints outlined by the CLSI, where available canine veterinary breakpoints were used (CLSI, 2018b) otherwise human breakpoints were used (CLSI, 2018a).

The ability of clinical P. aeruginosa isolates to form biofilm was tested in vitro as described previously with some modifications (Coffey and Anderson, 2014). Briefly, overnight cultures were diluted 1 in 100 into LB and 100 μL aliquots were separately transferred to wells of the 96-well microtiter plates in triplicate in the same microtiter plate; two microtiter plates were used for each technical repeat. A negative control, comprising three wells containing sterile LB, was included on each plate. The cultures were then grown statically at 37° C for 24 h. Following incubation, planktonic cells were removed, and the wells were washed using 125 μL of Ca/HEPES buffer; the wells were then stained using 0.1% (v/v) crystal violet solution. Excess crystal violet was removed, and the wells were then de-stained using ethanol (100%), and absorbance measured at 595 nm using a Tecan GENios Pro. The assay was repeated to achieve three biological repeats. Biofilm production was scored as either as either strong, moderate, weak, or non-biofilm producing as previously described (Stepanović et al., 2000).

Swarming motility assays were modified according to (Ugurlu et al., 2016). Briefly, 1 μL of bacterial culture was spotted into the centre of swarming motility agar (8 g/L nutrient broth, Oxoid; 5 g/L agar select, Sigma-Aldrich; 0.5% v/v glucose, Fisher Scientific), and then incubated at 37° C for 16 h. The distance travelled from the point of inoculation was measured using a digital calliper, and swarming motility assessed by phenotype on the plates in combination with the distance travelled from the point of inoculation.

Whole genome sequencing of 35 isolates was performed using both short (Illumina MiSeq) and long (Oxford Nanopore) read technologies by MicrobesNG (Birmingham, United Kingdom). Isolates were selected for sequencing to ensure isolates from the CALP nationwide laboratories (n = 13), Denmark (n = 7) and the RVC (n = 9) were represented (PRJNA1180571) and included some sequences from previous work (n = 6; Gigante et al., 2024; PRJNA107813).

Illumina paired-end reads were quality-assessed and trimmed using FastQC v0.11.8 (Andrews, 2010) and FastP v0.12.4 (Chen et al., 2018) respectively. Subsequently, contigs were assembled de novo using Flye v2.9.2-b1786 (Kolmogorov et al., 2019), followed by post-processing with Circlator v.1.5.5 (Hunt et al., 2015) and Bandage v0.8.1 (Wick et al., 2015). This was followed by one round of long read polishing and two rounds of short read polishing using Medaka v1.11.1 (ONT, 2023), Polypolish v0.5.0 (Wick and Holt, 2022) and POLCA from the MaSuRCA toolkit v4.0.9 (Zimin and Salzberg, 2020) respectively. Finally, genomes were reoriented to begin with dnaA using dnaapler v0.4.0 (Bouras et al., 2024) and annotated using Bakta v1.8.2 (Schwengers et al., 2021).

To aid in the comparison of isolates from the present study to a wider population of P. aeruginosa isolates, PanACoTA v1.4.0 (Perrin and Rocha, 2021) was used to download and filter P. aeruginosa genomes from the RefSeq database as previously described (Botelho et al., 2023). Briefly, P. aeruginosa sequences were selected using taxid (−T 287) and filtered so that genomes with more than 100 contigs were removed (--nbcont 100). Subsequently, redundant and misclassified genomes were removed using Mash distances (--min_dist 0.0001; --max_dist 0.05). Metadata for the RefSeq genomes was downloaded using NCBI Datasets v16.40.1 (O’Leary et al., 2024).

Whole genome average nucleotide identity (ANI) was calculated using FastANI v1.34 (Jain et al., 2018). This was used in combination with hierarchical clustering using Euclidean distance and complete linkage to assign isolates to a phylogroup (Botelho et al., 2023; Supplementary Figure 1). A neighbour joining phylogenetic tree was constructed using Mashtree v1.4.6 (Katz et al., 2019) using bootstrapping with 100 replicates and mindepth 0. Trees were then visualised using iTOL (Letunic and Bork, 2024).

Multilocus sequence typing (MLST) was performed in silico using the PubMLST database (Jolley et al., 2018). Where an unknown ST was identified, genome sequences were uploaded to the PubMLST database, and a new ST was assigned.

Screening of virulence genes was performed using ABRicate (Seemann, 2015) using the VFDB (Liu et al., 2021). Acquired antimicrobial resistance genes were identified using AMRFinderPlus v4.0.3 using assembled nucleotide sequences (−n), this included searching for point mutations (−O; Feldgarden et al., 2021).

Biofilm associated genes were identified using a custom database, created as described by (Seemann, 2015). Genes located within the psl, pel and alginate (alg) operons (Franklin et al., 2011) in addition to other associated biofilm genes (Liu et al., 2021; Kiel et al., 2022) were downloaded from NCBI using P. aeruginosa PAO1 (NC_002516.2). A complete list of the genes is shown in Supplementary Table 2.

A total of 10,300 genomes were downloaded from the RefSeq database and filtered for quality and genetic distance. The resulting set of 3,721 genomes was then randomly subsampled to yield a subset of 1,000 genomes. This included P. aeruginosa PAO1 (GCF_000006765.1) and PA14 (GCF_000014625.1) but PA7 (GCF_000017205.1) was removed based on Mash genetic distance which is in line with previous reports proposing that this strain be reclassified to a novel species Pseudomonas paraeruginosa (Rudra et al., 2022). These genomes in addition to 35 from the present study were compared using Mashtree. The distances calculated were then used to construct a phylogenetic tree (Figure 1). The strains were ascribed to three phylogroups; A (n = 765), containing P. aeruginosa PAO1; B (n = 244), containing P. aeruginosa PA14, and a third minor group, C (n = 26). The classification of P. aeruginosa into three phylogroups has been described previously (Stewart et al., 2014; Ozer et al., 2019); with one study suggesting that phylogroup C should be split into three groups (Freschi et al., 2019). However, with the exclusion of PA7, the results in the present study are in line with a recent study of P. aeruginosa phylogeny (Botelho et al., 2023).

Canine strains from the present study were distributed across all three phylogroups indicating that they were not closely related and did not cluster according to host source or geographical origin. This was supported by MLST analysis which identified 30 different STs among the clinical isolates. One isolate, 488613, was assigned a novel ST (5180). The most common ST was 788 which was identified in 9% (3/35) of clinical isolates.

Antimicrobial resistance (AMR) in P. aeruginosa from cases of canine otitis externa was determined using the disk diffusion method with 14 antimicrobials from 6 antimicrobial classes (Figure 2). Isolate 29130 failed to grow under the tested conditions and was excluded from this assay. The highest level of resistance was recorded for enrofloxacin (25.4%; 64/252) while tobramycin showed the highest sensitivity (99.6%; 251/252). Intermediate resistance was mainly seen against ticarcillin + clavulanic acid (60.7%; 127/252). Finally, 8.3% (21/252) and 0.79% (2/252) of isolates were resistant to imipenem and meropenem, respectively.

Of the 252 isolates tested, 65% (164/252) were susceptible to all of the tested antimicrobials. The remaining 35% (88/252) were resistant to at least one antimicrobial. Moreover, 1.6% (4/252) were classified as multidrug resistant—defined as resistant to three or more different classes of antimicrobials (Sweeney et al., 2018; Figure 3).

The 35 genome sequences of clinical P. aeruginosa isolates from canine OE were screened for the presence of ARG (Figure 4). Despite the variation in phenotypic susceptibility to antimicrobials between the strains, their complement of antimicrobial resistance genes remained remarkably consistent. Five genes, aph(3′)-IIb, fosA, catB7, blaOXA and blaPDC were detected in almost all of the isolates. Some notable exceptions are the presence of sul1, aadA7 and qacEΔ1 in 25181 and the absence of a blaOXA gene in 84269. The genes aadA7 and sul1 are linked to integrons and encode for a aminoglycoside nucleotidyltransferase that confers resistance to streptomycin and spectinomycin (Ahmed et al., 2004) and a sulfonamide resistant dihydropteroate synthase, respectively, (Sköld, 2000).

Development of resistance through the acquisition of mutations is one of the main drivers of resistance in P. aeruginosa (López-Causapé et al., 2018). Twelve different mutations were identified in the present study (Supplementary Table 3); the most common occurring in oprD. Other common mutations included those in the quinolone resistance-determining region (QRDR), which are the primary way in which P. aeruginosa can become resistant to fluoroquinolones (Piddock, 1999). In E. coli these mutations occur between amino acid residues 67–107 for GyrA, and 63–102 for ParC (Egorov et al., 2018). Studies with P. aeruginosa have reported the QRDR regions in GyrB as 429–585 and ParE as 357–503 (Bruchmann et al., 2013).

QRDR mutations were identified in almost all of the isolates that were resistant to at least one fluoroquinolone antimicrobial—with the exception of 84269 that was resistant to enrofloxacin but did not contain a QRDR mutation. When examining ARG individually, gyrA mutations were the most common, occurring in 20% (7/35) of isolates. The mutation T83I in GyrA was the most common among strains resistant to at least one fluoroquinolone. S466F, S87L/S331T and D533E were the most common mutations in GyrB, ParC and ParE respectively. These are within the QRDR regions, and have been previously associated with resistance (Akasaka et al., 2001; Bruchmann et al., 2013). Furthermore, some of the isolates contained more than one QRDR mutation which is known to further increase the MIC of ciprofloxacin (Bruchmann et al., 2013).

Two of the sequenced strains were resistant to gentamycin. Resistance to aminoglycosides can be due to inactivation by enzymes but also though reduced permeability or increased efflux (Poole, 2005). More recently, mutations to fusA1 in P. aeruginosa have been shown to confer resistance to aminoglycosides (Bolard et al., 2018). One chromosomal aminoglycoside resistance gene, 3′-phosphotransferase (aph(3′)-IIb), was identified which conferred resistance to kanamycin A and B, neomycin B and C, butirosin and seldomycin F5 in all the sequenced isolates (Zeng and Jin, 2003). However, aph(3′)-IIb does not provide resistance to the clinically relevant aminoglycosides tested in the present study (Atassi et al., 2023).

Only a small number of sequenced isolates (2/35) harboured a mutation in fusA1. Strain 488958, which was resistant to all of the tested aminoglycosides, contained T671A. Mutations in domains II, IV and V have been reported to increase the MIC of aminoglycosides up to 16-fold. T671A has been shown to significantly decrease the susceptibility of P. aeruginosa to aminoglycosides including gentamicin, amikacin and tobramycin (Bolard et al., 2018). Conversely, strain 87895 contained a Q678L mutation in the fusA1 gene which is known to result in a four-fold increase in the minimum inhibitory concentration (MIC) of tobramycin, however this strain was sensitive to tobramycin, and other tested aminoglycosides, in the present study (Scribner et al., 2020).

Other mutations identified in the present study include the introduction of a premature stop codon in mexZ in 8.6% (3/35) of the isolates. Inactivation of mexZ results in the overproduction of the MexXY-OprM efflux pump which can result in decreased susceptibility to fluoroquinolones, aminoglycosides and cefepime (López-Causapé et al., 2018). Finally, 14% (5/35) of the isolates harboured a V15I mutation in the pmrB gene, previously identified in clinical strains with an increased MIC of colistin (Lee and Ko, 2014) although this antibiotic was not included in the present study.

Clinical isolates of P. aeruginosa were tested for their ability to form a biofilm in vitro (Figure 5; Supplementary Figure 2). In total, 82% (207/253) of strains produced strong levels of biofilm, 9% (23/253) produced moderate biofilm, 5% (13/253) produced weak biofilm and 4% (10/253) made no quantifiable biofilm under the tested conditions.

To further understand the differences between levels of biofilm formation, the genomes of 35 P. aeruginosa isolates were screened for the presence of 53 genes known to be associated with biofilm formation (Supplementary Figure 3). All of the genes were highly conserved when compared to PAO1, with the exception of algP—a histone-like regulatory protein whose function in alginate synthesis is disputed and may be strain-dependent (Cross et al., 2020). All but one of the isolates contained all 53 genes, with the exception of 464429 which did not contain pslABCD. These genes encode the psl polysaccharide, which contributes to submerged biofilm (Friedman and Kolter, 2004), which may explain the poor biofilm forming ability of this strain.

Most strains (71%, 179/253) tested positive for swarming motility (Figure 6). The positive control strain (PAO1) consistently migrated to the perimeter of the agar plate within 16 h of incubation, and 16% (41/253) of the clinical isolates showed a similar swarming ability. Of the 205 isolates that produced strong levels of biofilm, 72% (147) also tested positive for swarming. Interestingly, strain 29051—the strongest biofilm producer—tested negative for swarming. When the relationship between biofilm formation and swarming motility was investigated no correlation was identified (r² = 0.002; Supplementary Figure 4).

The presence and absence of virulence genes was investigated using ABRicate with VFDB. Comparison of the canine OE isolates from this study to other P. aeruginosa isolates from human, animal and environmental sources using hierarchical clustering highlighted that there is no specific virulence profile associated with canine OE infection (Figure 7). Additionally, this highlighted that the strains did not cluster within phylogroups based on virulence factors. However, one notable exception is a cluster of 12 strains– including some from canines—which lacked numerous virulence factors, namely a series of genes involved in type III secretion systems. The majority (92%; 11/12), of these isolates belonged to phylogroup C, specifically, the section of the tree identified as group 5 by Freschi et al., 2019. These results are in line with the aforementioned study as they also identified that isolates in this group were missing 36 genes that encode for a Type III secretion system. The remaining isolate belonged to phylogroup A and was missing 32/36 of the type three secretion genes.