Antimicrobial Resistance and Genetic Diversity of Pseudomonas aeruginosa Strains Isolated from Equine and Other Veterinary Samples

Pseudomonas aeruginosa is one of the leading causes of healthcare-associated infections in humans. This bacterium is less represented in veterinary medicine, despite causing difficult-to-treat infections due to its capacity to acquire antimicrobial resistance, produce biofilms, and persist in the environment, along with its limited number of veterinary antibiotic therapies. Here, we explored susceptibility profiles to antibiotics and to didecyldimethylammonium chloride (DDAC), a quaternary ammonium widely used as a disinfectant, in 168 P. aeruginosa strains isolated from animals, mainly Equidae. A genomic study was performed on 41 of these strains to determine their serotype, sequence type (ST), relatedness, and resistome. Overall, 7.7% of animal strains were resistant to carbapenems, 10.1% presented a multidrug-resistant (MDR) profile, and 11.3% showed decreased susceptibility (DS) to DDAC. Genomic analyses revealed that the study population was diverse, and 4.9% were ST235, which is considered the most relevant human high-risk clone worldwide. This study found P. aeruginosa populations with carbapenem resistance, multidrug resistance, and DS to DDAC in equine and canine isolates. These strains, which are not susceptible to antibiotics used in veterinary and human medicine, warrant close the setting up of a clone monitoring, based on that already in place in human medicine, in a one-health approach.


Introduction
Pseudomonas aeruginosa (P. aeruginosa) is a Gram-negative bacterium commonly found in water and soil and is considered an opportunistic pathogen for humans, animals, and plants [1,2]. It can simply be carried in humans or can lead to various types of acute or chronic infection, typically nosocomial infections such as hospital-acquired pneumonia, bloodstream infections, and urinary tract infections, especially in urinary catheterization in humans and in immunocompromised individuals [3].
In veterinary medicine, even though P. aeruginosa naturally colonizes animal surface tissues [4], infection is relatively uncommon. However, it does cover a diverse spectrum, from otitis, ulcerative keratitis, urinary tract infections, and pyoderma in dogs and cats [2,[5][6][7], to mastitis in dairy cows, sheep and goats [2,8], hemorrhagic pneumonia in mink, otitis in chinchillas, and necrotic upper and lower respiratory tract lesions in snakes [9]. P. aeruginosa infections in animals, and particularly dog otitis cases, occur following improperly use, from food and agriculture to leisure and medical equipment [41]. In veterinary settings, DDAC is considered one of the most commonly used biocides in Europe [42], where it is used for its detergent and disinfectant actions on floors, walls, accessories, examination tables, medicated equipment, and noninvasive medical devices [43].
Here, we studied animal P. aeruginosa strains isolated from 1996 to 2020 to: (1) determine their antimicrobial resistance profile and susceptibility to DDAC detergent-disinfectant; and (2) describe the genetic diversity of circulating populations and their resistome.

P. aeruginosa Bacterial Strains
We used 4 reference strains with the available genomes: ATCC15442 and ATCC27853 were obtained from the American Type Culture Collection (ATCC), and PAO1 and PA14 from the Institut Pasteur collection (Paris, France). ATCC15442 is recommended for disinfectant susceptibility testing [44], ATCC27853 is the reference for Pseudomonas spp. antibiotic susceptibility testing [45], PAO1 is the reference genome for the P. aeruginosa species [23], and PA14 is a highly virulent strain that represents the most common P. aeruginosa clonal group worldwide [46].
A further 168 P. aeruginosa strains isolated from animals (135 from equid origin, 30 from canine, 2 from feline, and 1 from bovine) were selected retrospectively ( Figure 1 and Supplementary Table S1 Part 1): 24 were collected at the Anses, Normandy Laboratory for Animal Health (Goustranville, France), and the other 144 were isolated at the LABÉO diagnostic laboratory (Saint-Contest, France), from samples received for diagnostic (ante or postmortem) or screening analysis. Selected strains covered a diverse set in terms of the year of isolation and origin (animal species and type of sample). The Anses strains represented all isolated P. aeruginosa strains during the 1996-2017 period and included 22 from necropsies and two from other/unspecified sampling types. The LABÉO strains represented all veterinary antibiotic-resistant strains isolated and conserved during the 2017-2018 period (n = 27), and then all isolated P. aeruginosa strains during the 2019-2020 period (n = 117). The LABÉO's strains were isolated from genital samples (n = 89), auricular samples (n = 27), respiratory samples (n = 12), cutaneous/wound samples (n = 8), ocular samples (n = 5), other/unspecified samples (n = 2), and digestive samples (n = 1). All strains were stored at −65 to −80 • C. Species identification was confirmed using matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-TOF) (Microflex; Bruker Daltonik, Bremen, Germany).
From these 168 strains, we selected 41 strains to constitute the panel for genomic characterization, including 31 equine strains, 7 canine strains, 2 feline strains, and 1 bovineassociated strain ( Figure 1 and Table S1 Part 2). Selection was based on antibiotic susceptibility tests (AST), disinfectant minimum inhibitory concentration (MIC), and diversity in year of isolation and origin (animal species and type of sample) of the strains. Appendix A gives a summary of the temporal distribution and origin of the strains.

Quaternary Ammonium Compound Susceptibility Testing
For the 168-strain study panel, the MIC of the detergent-disinfectant DDAC was assessed in triplicate per strain by the reference broth microdilution method [49] using cation-adjusted Mueller-Hinton Broth w/TES (Thermo Fisher Scientific, Waltham, MA, USA) at concentrations ranging from 0.5 to 1024 mg/L. Here, the threshold of decreased susceptibility (DS) was set at MIC > 62.9 mg/L, corresponding to the concentration of DDAC when combined with an alkylamine in a routine disinfectant diluted according to the manufacturer's instructions for use.

Whole Genome Sequencing and Bioinformatics Analysis
Genomes of the reference strains ATCC27853 and PA14 were obtained from the European Nucleotide Archive (ENA) database with accession numbers CP015117 and ASWV01000001, respectively. The ATCC15442 and PAO1 sequences were obtained from GenBank with accession numbers GCF_000504485.1 and GCA_000006765.1, respectively.

Statistical Analysis
All statistical tests were performed with GraphPad Prism version 9.0.0 for macOS (GraphPad Software, San Diego, CA, USA).
Populations were tested for independence using Fisher's exact test to determine whether DS to DDAC was linked to the nonsusceptibility of strains to antibiotics (MDR or XDR profiles). Then, to assess whether the distribution of the DS-to-DDAC phenotype was significant, Fisher's tests were performed by comparing the numbers of each animal species to the numbers of the remaining population. The same method was applied to compare the distribution of the DS-to-DDAC phenotype by type of sample. For a given category (animal species and then type of sample), if the total number was less than five, it was not given a p-value as it was considered non-representative.
Among the sequenced strains (n = 41), we found numerous genes conferring antibiotic resistance: some were constitutive for P. aeruginosa, and others were acquired. At least one resistance gene was associated with each main antibiotic class ( Figure 3). A total of 11.9% (n = 20) of all veterinary strains were resistant to anti-Pseudomonas penicillins (at least one resistance among PIL, PTZ, TIC, TCC), 7.7% (n = 13) were resistant to carbapenems (IPM and/or MEM), 1.2% (n = 2) were resistant to monobactams (ATM), and 20.2% (n = 34) were resistant to cephalosporins (FEP, CZD, CLT, CEQ) ( Figure 2A). For equine strains, 11.1% (n=15) were resistant to penicillins, 7.4% (n = 10) to carbapenems, and only one strain was resistant to monobactams ( Figure 2B). For both classes of antibiotics, canine strains were the most resistant, with 16.7% (n = 5), 10.0% (n = 3) and 3.3% (n = 1) of resistance, respectively ( Figure 2B). These resistances were associated with the presence of resistance genes for β-lactams, including those coding CARB-2, the PSE family carbenicillin-hydrolyzing class A beta-lactamase (bla CARB-2 ), PDC, the cephalosporin-hydrolyzing class-C β-lactamase (bla PDC variants), and the OXA-family oxacillin-hydrolyzing class-D β-lactamases (bla OXA ). The genes indicated as bla OXA -like and bla PDC -like were variants of bla OXA and bla PDC , respectively, that had less than 100% identity but were not known from the database. Note that no carbapenemase-coding genes were identified in the sequenced strains ( Figure 3). For fosfomycin, only 0.6% (n = 1) of the veterinary population showed resistance, whereas the fosfomycin resistance glutathione transferase (fosA) coding gene was found in all sequenced strains. It was observed that 18.5% (n = 31) of the veterinary strains tested showed at least one resistance to aminoglycosides (AKN, GMN, NTM, TMN) ( Figure 2A). Overall, 14.8% (n = 20) of equine strains were resistant to at least one aminoglycoside and 36.7% (n = 11) of canine strains ( Figure 2B). Various antimicrobial resistance genes targeting this class of antibiotics were identified in the tested strains, including genes encoding aminoglycoside Ophosphotransferases (aph variants), N-acetyltransferases (aac variants), and aminoglycoside nucleotidyltransferases (aad variants) ( Figure 3). For fluoroquinolones (CIP, LVX), at least one resistance was observed in 19.0% (n = 32) of veterinary strains. This rate was close (22.0%) when also considering veterinary antibiotics (ENR, MAR) ( Figure 2A). The equine strains were resistant in 8.9% of cases (n = 12), although canine strains were resistant in 66.8% of cases (n = 20) ( Figure 2B). For fluoroquinolone resistance, only gene-encoding protein CrpP was retrieved. For quinolone resistance, we identified genes encoding for the pentapeptide repeat protein QnrVC1 and modification of the amino acid sequence in the quinolone-resistance-determining region (QRDR) due to sequence alterations in position 473 of the parE gene. For sulfonamide, we found only the resistance gene encoding dihydropteroate synthase (sul1). For phenicols, we found type B chloramphenicol Oacetyltransferase (catB variants) and chloramphenicol efflux major facilitator superfamily (MFS) transporters (cmlA/floR variants and cmx). For tetracycline, genes coding the efflux MFS transporter Tet(G) and resistance-nodulation-division (RND) transporter efflux pump MexCD-OprJ were found. Interestingly, the qacEdelta1, qacG2, and qacL genes encoding small multidrug resistance (SMR) protein transporters were also found ( Figure 3).  . Antimicrobial-resistance-associated genes of strains from the panel for genomic characterization and for reference strains (n = 41 + 4). Strains were organized according to the cgMLST minimum distance tree. The figure lists serotype, sequence type, presence of a carbapenem resistance (imipenem and/or meropenem) phenotype, whether the strain was multidrug-resistant or not, and presence of decreased susceptibility to DDAC phenotype. The sub-variants of the aph(3 ) gene were grouped, as well as the sub-variants of aac (6 ). Co-occurrences of these sub-variants are indicated and separated by a backslash. The "-like genes" were variants not referenced on AMRFinder that had less than 100% shared identity. C: canine, B: bovine, DDAC: didecyldimethylammonium chloride, E: equine, F: feline, MDR: multidrug-resistant, O: serotype, ST: sequence type.
Regarding DDAC, the MICs ranged from 8 to 128 mg/L, and decreased susceptibility (DS, MIC > 62.9 mg/L) was observed in 11.3% of tested strains (n = 19/168) ( Figure 2C) and 11.9% of equine strains (n = 16/135) ( Figure 2D). The DS to DDAC phenotypes was associated with patterns of antimicrobial drug resistance to up to two antibiotic classes: 4.8% showed no antibiotic resistance (3.7% for equine strains), 4.2% were resistant to one antibiotic class (5.2% for equine strains), and 2.4% were resistant to two antibiotic classes (3.0% for equine strains) ( Figure 2C,D). The DS to DDAC phenotypes were observed since at least 2017 (from two in 2017 to seven in 2020; Table 1), largely in equine strains (n = 16) but also in feline (n = 2) and canine strains (n = 1) ( Table 1). Before 2017, its presence cannot be evaluated, due to a lack of representativeness of the population. The DS to DDAC phenotype was significantly found in respiratory samples (n = 8, Fisher's exact test, p < 0.0001), and it was also found in genital samples (n = 7); both were specifically associated with equines. The DS to DDAC phenotypes was also observed from ocular, cutaneous/wound, auricular, and digestive strains, but each only once (Table 1).

Genomic Diversity and Resistome Analysis
The 41 strains of the panel for genomic characterization were distributed into seven serotypes ( Table 2). The three main serotypes were the same for all veterinary strains, and for the Equidae population in particular: serotypes O6 (39.0%), O11 (26.8%), and O5 (14.6%). Only O6 and O11, like serotypes O1 and O10, were found in more than one animal species. The 41 sequenced strains were distributed among 28 sequence types (ST) ( Figure 4A): primarily ST395 (14.6%), ST309 (9.8%), ST3709 (7.3%), ST235 (4.9%), ST253 (4.9%), and ST655 (4.9%). The other 22 STs were represented once. A closer focus on the equine strains ( Figure 4B) showed that the main STs were ST395 (19.4%), ST3709 (9.7%), and ST235-ST309-ST655 (6.5% for each), with the other 16 ST represented by one strain. In the other animal species, only ST309 was represented twice, and ST162, ST252, ST253, ST260, ST261, ST1007, ST2683, and ST3314 were represented once ( Figure 4C). Only ST309 and ST253 were found in more than one animal species.  This apparent diversity in serotypes and sequence types was confirmed by the cgMLST results charted in Figure 4. Of a total of 3867 loci searched, 3314 loci were present in the genome of all strains. Overall, strains diverged in distance from 14 to 3256 loci (average distance was 2893) and were well distributed according to origin (equine, canine, feline, bovine, human or environmental), year of collection, serotype, sequence type, carbapenem-resistance phenotype (major antibiotics reserved for human medicine), DS to DDAC phenotype, and whether or not the strain was multidrug-resistant ( Figure 5). None of the 41 sequenced strains showed a high level of antibiotic resistance (carbapenem resistance and MDR status) associated with DS to DDAC. However, three equine strains (E-18-20621-1-1, E-18-40793-1-1, and E-18-42174-1-1) showed a DS to DDAC associated with resistance to carbapenems. These strains were serotype O6: two were ST395, and one was ST233. The other equine strains presented either a DS-to-DDAC profile (n = 13), an MDR profile associated or not with resistance to carbapenems (n = 3 and n = 5, respectively), carbapenem resistance only (n = 4), or none of these phenotypes (n = 107). MDR status was assigned to 5 out of 7 of tested canine strains: three had resistance to carbapenems, and only one had DS to DDAC (C-19-49802-1-1). These strains were not associated with a single serotype or ST. The two feline strains, F-20-32054-2-1 (O11 and ST309) and F-20-12619-1-1 (O10 and ST253) did not show resistance to carbapenems or an MDR profile but had DS to DDAC. The bovine strain B-20-37098 (O6 and ST2683) was susceptible to antibiotics and DDAC.
This study set out to review the level of resistance of veterinary strains to the main human and veterinary antibiotics and to a common disinfectant, and to highlight the lack of anti-Pseudomonas therapies available in the veterinary field. It also enabled us to investigate and report the genomic diversity of these populations and the different antimicrobial resistance genes represented in them, and argue for the need to jointly study the Human-Animal-Environmental reservoirs. We selected 111/263 (42.0%) of the strains isolated from 2017 to 2020 at LABÉO to ensure a diversity of sample years and origins. This panel was completed by 24 equine strains provided by Anses and isolated during necropsies performed in the 1996-2017 period. Several animal species were included in this study; however, the small number of feline and bovine strains ruled out including these strains in a cross-species comparison. However, these feline and bovine strains did make it possible to determine whether some phenotypes are specific to one of the species.
For ASTs, the only antibiotics cited for testing by the veterinary CASFM 2021 [78] are gentamicin, amikacin, and ciprofloxacin. The observed resistance rates in our population were 18.5% for gentamicin and less than 10% for amikacin and ciprofloxacin. However, among the list of anti-pseudomonas antibiotics tested in this study, only ceftiofur, cefquinome, gentamicin, marbofloxacin, and enrofloxacin are currently marketed for veterinary use according to the Index of Veterinary Medicines authorized in France by the Anses [79]. Polymyxins can also be used but were not tested here. In this population, resistance rates for these antibiotics are above 20% for cephalosporins (up to 98.8% for ceftiofur) but less than 20% for fluoroquinolones.
We expected to find this high rate (up to 98.8%) of resistance to ceftiofur [80][81][82], but as this antibiotic is taken into consideration for the choice of therapy by some diagnostic laboratories, we wanted to provide additional evidence of the low activity of ceftiofur on P. aeruginosa. For gentamicin and fluoroquinolones, the rates obtained were lower than those found by van Spijk et al. [83] between 2012 and 2015 on an equine hospital population.
Note that for some of these antibiotics, there is no marketed medicine suitable for use in every animal species. Consequently, there is a significant lack of antibiotic-based solutions against P. aeruginosa in veterinary medicine. For all classes of antibiotics tested, canine strains had higher resistance rates than equine strains. This was particularly marked for fluoroquinolones (57.9% difference) and aminoglycosides (21.9% difference). The resistance values obtained here can be compared with a previous study performed at the nearby Caen University Hospital on strains isolated from patients between 2011 and 2020 [84]. The resistance rates found in P. aeruginosa strains were much lower than those obtained in human medicine for phosphonic acid (−32.7% for the veterinary population) and penicillins (−15.2%), and measurably lower for monobactams (−5.2%), cephalosporins (−4.8%) and carbapenems (−1.8%), but higher for fluoroquinolones (+3.4%) and aminoglycosides (+7.2%). It is surprising to note such a small difference in resistance rates in respect of carbapenems between veterinary and human populations, especially as carbapenems are not authorized and used in veterinary medicine in Europe, except in exceptional cases in university clinics. As found in other studies, the presence of a carbapenem resistance phenotype in animals is rarely associated with the presence of a carbapenemase. The presumed mechanisms would be the decreased permeability by deficiency of the outer membrane protein OprD2 [85], hyperproduction of the chromosomal cephalosporinase AmpC [86][87][88], or preferential overexpression of efflux pumps [89]. However, some rare cases of carbapenemase expression, especially VIM-2, have been reported in different countries in P. aeruginosa strains isolated from dogs, cattle, and fowl [90,91], but to our knowledge not yet in horses. Some studies even suggest the existence of zoonotic transmission from animals to humans [92] and from humans to animals [93,94], but in an anecdotal manner, including the case of a transmission of a VIM-2 strain from humans to animals in Brazil [95].
No pandrug-resistant or extensively drug-resistant strains were found, but 10.1% of the strains studied were categorized as MDR. Note that MDR strains were found in the more recent strains isolated at LABÉO, suggesting that equine P. aeruginosa strains have adapted to 3-4 classes of antibiotics. This finding corroborates similar previous studies [89] showing the existence of MDR strains in the veterinary population and extends this problem to Equidae. In comparison, in human medicine, 12.6% of P. aeruginosa strains analyzed have either MDR (11.9%) or XDR (0.7%) profiles [84].
From a genomic point of view, the 41 veterinary strains were quite diverse, which is consistent with the strain sampling system. These strains came from different types of environments depending on animal species and geographical location. In total, 28 different STs were identified. Note that among these STs, ST395 (n = 6), ST235 (n = 2), ST253 (n = 2), ST233 (n = 1), and ST27 (n = 1) were also found in the hospital isolates in a previous human study (patient only) [84]. Thus, 29.3% of the animal strains sequenced in this study shared an ST identified in patients at the University Hospital of Caen. ST235, representing 4.9% of the sequenced strains, was even considered in 2020 as among the top 10 P. aeruginosa high-risk clones worldwide based on prevalence, global spread, and association with MDR/XDR profiles and the extended-spectrum β-lactamases and carbapenemases [96]. However, in our study, ST235 strains were neither XDR, MDR, nor carbapenem-resistant, but only associated with a DS-to-DDAC phenotype. ST235 had already been found in Japan in dogs and cats at a rate of up to 21.1% [94]. Four ST309 strains were also identified in this study and were found in the human hospital environment [84]. It would be informative to study the genetic distance between veterinary and hospital strains in more detail, in particular using a cgMLST approach.
In terms of cgMLST, note that strains isolated from different animal species could have greater genomic proximity than strains isolated from the same animal species, which would be quite diversified. However, the data collected on our samples do not allow us to distinguish between cases of infection or colonization in the animal. This factor would determine whether or not strains from colonization and/or infection would be grouped on the phylogenetic tree. It would be also interesting to determine how close the human and animal strains were in order to get a picture of their potential capacity for human-animal transmission.
Concerning the resistome of the strains, some genes were only found to be associated with one animal species or within one ST. Except for aph(3 )-IIb, which is systematically present in P. aeruginosa, all the aminoglycoside resistance genes were specific to equine strains, as was also the case for some oxacillinases and the bla CARB-2 gene and for the genes for resistance to sulfonamides, phenicols (except catB7), tetracycline, the qnrVC1 gene targeting quinolones, and resistance genes to quaternary ammonium. The resistome associated with the other species was thus less diverse. On the whole, for the same ST, the strains had similar, but not always identical, resistomes. In ST252, ST655, and ST3709, only one resistome was found by ST, whereas for ST155, ST235, ST253, ST309, and ST395, various antimicrobial resistance gene profiles were found in each, although similar. Moreover, the same resistome could lead to various resistance profiles. For example, the canine-associated MDR strain C-19-50802-1-1 presented the same resistome as the non-MDR reference strain ATCC15442, implying potentially mutational mechanisms conferring higher resistance to the canine-associated strain.
This study also determined the MIC of P. aeruginosa strains to a quaternary ammonium compound, DDAC, that is widely used in veterinary hospital disinfectants and as a biocide in various applications. Considering a MIC > 62.9 mg/L (corresponding to the concentration of DDAC in the commercial disinfectant solution), there was DS to DDAC in 11.0% of our strains and this DS was not associated with MDR P. aeruginosa profiles. MICs of DDAC greater than or equal to our DS threshold have also been shown for veterinary strains isolated between 1994 and 2003 in the USA, but on a smaller scale (2.9% of 175 P. aeruginosa strains) [81]. In contrast, at Caen University Hospital, this profile was found for 38.9% of strains, spanning both human strains and hospital-environment strains, that were significantly more associated with MDR and XDR P. aeruginosa profiles and more prevalent in the hospital environment (62.5% of them) than in human strains (28.2%) [84]. In our population, the fact that we included strains potentially from individuals and non-hospital veterinary environments likely leads to a lower calculated rate of resistance. These strains must have been less frequently exposed to DDAC than strains from the hospital environment, which is obviously regularly disinfected. The underlying molecular mechanisms of this phenotype are not yet fully elucidated, but the initial evidence points to MexAB-OprM pump efflux overexpression [84].

Conclusions
To the best of our knowledge, this is the biggest and most representative phenotypic and genomic study on P. aeruginosa strains isolated from Equidae. With the implementation of whole-genome sequencing and genomic approaches, we were able to assess the diversity and the resistome of the different strains. Such a strategy is destined to become an indispensable tool for monitoring infections and the dissemination of resistance. In contrast to hospital-acquired human infections, our results point to a high diversity of P. aeruginosa populations as causative agents in equine infections. Because more than 10% of animal strains showed an MDR phenotype, horses may be considered reservoirs of antimicrobial resistance in P. aeruginosa. Moreover, a significant number of isolates were resistant to carbapenems (7.7%), which are antimicrobials non-authorized in veterinary medicine. This highlights the need for a global approach in epidemiological studies. Our data also pointed out that attention should be paid to the use of disinfectants such as DDAC, constituting a selective pressure for the persistence of less susceptible strains. Our robust data are the foundation for further monitoring P. aeruginosa resistant strains and optimizing antimicrobial therapies in veterinary medicine via a one-health approach.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/pathogens12010064/s1, Table S1: Characteristics of the strains used in the study (Part 1) and for genomic characterization (Part 2). Table S2: Excel file of antimicrobial resistance-associated genes of strains from the panel for genomic characterization and for reference strains (n = 41 + 4).

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.
Appendix A Table A1. Distribution of the decreased susceptibility-to-DDAC phenotype in the panel for genomic characterization (n = 41), according to the year of strain isolation, animal species and sampling.  Appendix C Figure A1. Bioinformatics pipeline for whole-genome sequencing and sequence analysis. . cg: core genome; MLST: multilocus sequence typing; WGS: whole-genome sequencing.