Background Pseudomonas aeruginosa is an opportunistic pathogen and a major cause of infections. Widespread resistance in human infections are increasing the use of last resort antimicrobials such as polymyxins. However, these have been used for decades in veterinary medicine. Companion animals are an understudied source of antimicrobial resistant P. aeruginosa isolates. This study evaluated the susceptibility of P. aeruginosa veterinary isolates to polymyxins to determine whether the veterinary niche represents a potential reservoir of resistance genes for pathogenic bacteria in both animals and humans.
Methods and results Clinical P. aeruginosa isolates (n=24) from UK companion animals were compared for antimicrobial susceptibility to a panel of human-associated isolates (n=37). Minimum inhibitory concentration (MIC) values for polymyxin B and colistin in the companion animals was significantly higher than in human isolates (P=0.033 and P=0.013, respectively). Genotyping revealed that the veterinary isolates were spread throughout the P. aeruginosa population, with shared array types from human infections such as keratitis and respiratory infections, suggesting the potential for zoonotic transmission. Whole genome sequencing revealed mutations in genes associated with polymyxin resistance and other antimicrobial resistance-related genes.
Conclusion The high levels of resistance to polymyxin shown here, along with genetic similarities between some human and animal isolates, together suggest a need for sustained surveillance of this veterinary niche as a potential reservoir for resistant, clinically relevant bacteria in both animals and humans.
- bacterial pathogenesis
Statistics from Altmetric.com
Antimicrobial resistance (AMR) is of great concern in both human and veterinary healthcare. This issue is confounded by the lack of novel antibacterials being developed in the past 30 years, in particular those with activity against serious Gram-negative infections.1–3 Therefore older, previously less favoured drugs, such as polymyxins, are increasingly being used.4–8 Polymyxin B and especially polymyxin E (colistin) are used in human and veterinary clinical practice, both systemically and topically, due to their bactericidal activity against a wide range of Gram-negative bacilli including major agents of nosocomial infections such as Acinetobacter baumannii, 9 carbapenem-resistant Enterobacteriaceae10 and Pseudomonas aeruginosa. 11 Polymyxins selectively bind to the lipopolysaccharide (LPS) of the outer cell membrane, disturbing its permeability and causing destabilisation and membrane lysis.12 Polymyxin use has been at relatively low levels in human medicine due to safety concerns, in particular nephrotoxicity and neurotoxicity.13 However, colistin has been re-introduced in clinical practice for treating pathogens resistant to all available antimicrobials and is therefore often regarded as a ‘last resort’ treatment in patients with cystic fibrosis (CF) and in nosocomial infections involving P. aeruginosa.14
P. aeruginosa is an opportunistic pathogen of both humans and companion animals. In humans, it can cause severe hospital-acquired infections such as pneumonia, keratitis, burn and wound infections, urinary tract infections, endocarditis and meningitis.15 16 P. aeruginosa can affect the lower respiratory system in humans and is an important pathogen in patients with CF and also other chronic lung diseases such as non-CF bronchiectasis.17 In companion animals, P. aeruginosa can cause pyoderma, chronic otitis externa, ulcerative keratitis, wound infections, respiratory tract infections and urinary tract infections in a number of species.18–22 A number of other diseases involving P. aeruginosa in animals are known, including equine endometritis23 and in chronic equine wounds where the ability of P aeruginosa to form and survive within protective biofilms aids persistence.24 25
Polymyxin B is a common first-line topical therapy to treat otitis externa in dogs and cats, and has also been used for systemic treatment of endotoxaemia-associated with severe colic and other gastrointestinal diseases in horses.26 27 High levels of P. aeruginosa have been identified in pet and laboratory chinchillas.28 A European Medicines Agency report on colistin usage in animals within the EU stated that across 30 EU/EEA European countries for which sales data are available for 2015, polymyxins were the fifth most sold group of antimicrobials (6.8 per cent), after tetracyclines (32.8 per cent), penicillins (25.0 per cent), sulphonamides (11.8 per cent), and macrolides (7.2 per cent).29 The use of polymyxins specifically in companion animals is not reported in the literature. However, there are several studies looking at antimicrobial prescribing in small animal practice.30–32
Resistance mechanisms in P. aeruginosa to polymyxins include adaptive and intrinsic resistance, often through genetic mutations,33 and are generally characterised by alterations in the outer membrane.34 These mechanisms have recently been reviewed by Jeannot and others.35 A study by Fernández and others reported a relatively small polymyxin B resistome involving 17 susceptibility/intrinsic resistance determinants, including evidence of cross-resistance between colistin and polymyxin B.36 In addition, previous studies have shown that selective pressure generated by the increased use of colistin can lead to emergence of colistin resistance.37 38 This raises concern regarding use of polymyxins within the veterinary community, whereby their use may help drive development of polymyxin B/colistin-resistant P. aeruginosa strains.39 Such strains could then act as a source of resistant P. aeruginosa populations. The occurrence and risk of P. aeruginosa transmission between animals and humans is unknown. While it is possible that animals may act as sources of bacteria for humans, transmission is likely to occur both ways. Transmission of P. aeruginosa between a patient with CF and its pet cat has previously been reported.40
In this study, the authors investigated resistance prevalence to polymyxins in veterinary-associated companion animal (including dog, cat and horse) P. aeruginosa isolates, and compared their phenotype and genotype with those from a reference panel of human isolates. This provides an important insight into P. aeruginosa AMR in this niche, and suggests that potential resistance in animals is a critically important area for future consideration.41 42
Materials and methods
Clinical P. aeruginosa (n=24) isolates from 22 companion animals attending the Small Animal or the Philip Leverhulme Equine Hospital (both referral hospitals; University of Liverpool, UK) in 2012 were collected for this study as part of routine microbiological diagnostic workup (table 1). Human-associated P. aeruginosa isolates (n=37 (25 CF and 12 non-CF clinical isolates)) from a previously published panel were used for comparison.43 44 All isolates were stored in Luria-Bertani Broth (LB) (Oxoid) supplemented with glycerol and stored at −80°C. When needed, bacteria were grown on Columbia agar (Oxoid) aerobically for 24 hours at 37°C.
Minimum inhibitory concentration assays (polymyxin B and colistin)
For all isolates, minimum inhibitory concentration (MIC) for polymyxin B and colistin (Sigma, UK) were performed from overnight broths using EUCAST standardised methods.45 In brief, broth dilutions were performed in 96-well plates with antibiotic concentrations from 128 to 0.25 mg/l. Positive and negative controls were used throughout and five replicate assays were performed. Controls strains P. aeruginosa ATCC 27853, the laboratory reference strain PAO1 and Escherichia coli NCTC 13846 were used in every experiment. The plates were incubated aerobically at 37°C and assessed after 24 and 48 hours of growth in room air.
For colistin, standard breakpoints were used (>2 mg/l), consistent with the EUCAST guidelines used by human clinical microbiology diagnostic laboratories in the UK.45 MIC values for polymyxin B were interpreted at breakpoints of sensitive ≤2 mg/l, intermediate=4 mg/l, resistant ≥8 mg/l (M100-ED29 Clinical and Laboratory Standards Institute).46
Further antimicrobial susceptibility testing
To test susceptibility to a panel of veterinary-relevant antimicrobials, subcultures were made of the human and companion animal samples by passage on blood agar or Columbia agar. After aerobic incubation at 37°C overnight, colonies were suspended in 5 ml sterile diluted water (SDW) to produce a uniform turbidity of 0.5 McFarland units. A 10 µl aliquot of this solution was transferred to 11 ml cation-adjusted Müeller-Hinton broth (Sigma, UK) that was then transferred to the Sensititre COMPAN1F 96-well microtitre plate via the Thermoscientific Sensititre AIM (ThermoScientific) automated inoculation delivery system. This system determines resistance to ampicillin, amoxicillin/clavulanic acid, ticarcillin, trimethoprim/sulfamethoxazole, gentamicin, penicillin, ceftiofur, enrofloxacin, cefovecin, amikacin, cefpodoxime, imipenem, erythromycin, marbofloxacin, oxacillin, cefoxitin, ticarcillin/clavulanic acid, clindamycin, doxycycline, chloramphenicol, cefazolin and rifampin. The plates were incubated aerobically at 37°C for 24 hours and the plates read using the Sensititre OptiRead Automated Fluorometric Plate Reading SystemTrek Sensititre (ThermoFisher Scientific) to measure the MIC for inhibition of growth (breakpoints provided by Thermo Scientific SENSITITRE COMPAN1F protocol).
For antimicrobials commonly associated with human use, disk diffusion assays were performed to determine antimicrobial susceptibilities for ceftazidime (30 µg), ciprofloxacin (5 µg), meropenem (10 µg), tazobactam/piperacillin (85 µg) and tobramycin (10 µg) (all from Oxoid, Basingstoke, UK) using current EUCAST guidelines.45
Alere array tube genotyping for P. aeruginosa
Alere array tubes (Alere Technologies) were used to genotype the isolates as described previously.47 In brief, a sweep of each isolate was emulsified in 1 ml of SDW and centrifuged at 10,000 g for two minutes. The supernatant was removed and resuspended in 200 µl of SDW, boiled for five minutes and centrifuged at 10,000 g for two minutes; 5 µl of this supernatant was added to 5 µl of a labelling master mix and amplified by PCR (five minutes at 96 for 1 cycle, 50 cycles of 20 seconds at 62°C, 40 seconds at 72°C and 60 seconds at 96°C). Following hybridisation and washing, reagent C3 (containing Horse Radish Peroxidase conjugate) was used to label the chip. Detection was performed using the Iconoscan, Iconoclust Software (Alere Technologies). Tube array images were transformed into array types as previously described.47 In order to study the veterinary isolates in the context of the wider population, an array type database of more than 900 recorded P. aeruginosa strains was used.48 For displaying the wider P. aeruginosa population, the eBURST algorithm was applied.49 The number of isolates=981, number of sequence types=256, number of loci per isolate=16 and number of resampling for bootstrapping=1000 (for statistical confidences).
SigmaPlot (V.13.0) was used to compare the median MIC values (of five replicates) for each isolate. The MIC results for each of the polymyxins (ie, colistin vs polymyxin B) and for each set of isolates (ie, companion animal vs human) were compared. To assess resistance to polymyxin between the human and veterinary isolates Fisher’s exact test (with P<0.05 considered significant) was used.
Whole genome sequencing of bacterial isolates
Genomic DNA from the seven companion animal-associated polymyxin-resistant isolates was extracted from overnight cultures using the DNeasy Blood and Tissue Kit (QIAGEN). Genomic DNA (500 ng) was mechanically fragmented for 40 seconds using a Covaris M220 (Covaris, Woburn, Massachusetts, USA) with default settings. Fragmented DNA was transferred to PCR tubes and library synthesis was performed with the Kapa Hyperprep kit (Kapa Biosystems, Wilmington, Massachusetts, USA) according to manufacturer’s instructions. TruSeq HT adapters (Illumina, SanDiego, California, USA) were used to barcode the samples and libraries were sequenced along with 41 other bacterial genomes in an Illumina MiSeq 300 bp paired-end run at the Plateforme d’Analyses Génomiques of the Institut de Biologie Intégrative et des Systèmes (Laval University, Quebec, Canada).50 Resistant isolate genomes were assembled using the A5 assembler version A5-miseq 2014052151 and annotated using prokka V.1.5.52 Accession numbers are available in online supplementary table S1.
All seven polymyxin-resistant genomes were aligned to reference genome PAO1 and separately, PA14 using bwa mem V.0.7.5a.53 Resulting .sam alignment files were sorted, converted to .bam format and duplicates marked and removed using picardtools V.1.85. The Genome Analysis Toolkit (GATK) V.3.353 was used to create indel targets, realign them and call variants using the Unified Genotyper (UG) module. All variants were filtered using vcffilter version54 ‘DP more than 9’ and ‘QUAL more than 10’ to produce the final .vcf files. For all isolates, the estimated genome coverage was over x15.
Reference genomes PAO1, PA14, PA7 and LESB58 and 338, which were deemed to represent the wider population,55 were downloaded from the Pseudomonas Genome Database.56 The wider population was a random subset of isolates from a previous published study that included isolates from North America, South America, Europe and Asia.55 The core genome was defined and extracted by Panseq57 as 500 bp fragments matching in all 342 genomes with greater than and equal to 85 per cent similarity. MEGA658 was used for all phylogenetic analyses. Phylogeny was estimated using the maximum likelihood method (ML) and Tamura-Nei substitution model from concatenated polymorphic sites within the defined core genome. Inner node support was based on 100 bootstrap replicates. The tree was drawn using Figtree.59
Ortholog sequences for 31 polymyxin resistance-associated genes were downloaded from the Pseudomonas Database.56 A custom blast database was curated consisting of the seven polymyxin-resistant isolate genomes using the BLAST+60 makeblastdb software. Each polymyxin resistance-associated gene was aligned against the custom database using BLAST+ blastall. A python script was used to extract matching regions and convert to amino acid sequences from the seven polymyxin-resistant isolate genomes.
Resistance to polymyxin antimicrobials
For the veterinary isolates, 92 per cent displayed resistance to polymyxin B (table 2). The highest MIC value was observed in an isolate (984) detected from a canine ear, which had an MIC of greater than 128 mg/l (figure 1). In comparison, 68 per cent of the human-associated P aeruginosa isolates were resistant to polymyxin B (P=0.033). The proportion resistant to colistin among the veterinary isolates was also significantly higher compared with the human-associated panel (54 per cent compared with only 22 per cent, P=0.013) (table 2). The number of colistin-resistant isolates was surprisingly high; however, comparison with previous data43 is challenging as the MIC cut-off value has recently changed from greater than 4 mg/l to greater than 2 mg/l61. Many of the human isolates displayed MICs of around 4 and 2 mg/l and therefore caution must be taken when interpreting these results. The isolates with the highest MIC value to colistin were isolates 856 and 903 (figure 1). These isolates were both of equine origin, from an abdominal incision and a urine sample, respectively. MIC 50 values for polymyxin B in the companion animal and human isolates were 16 and 8 mg/l, respectively, while the MIC50 values for colistin were 4 and 2 mg/l, respectively (table 2).
Resistance of veterinary-associated P aeruginosa to other antimicrobials
Testing of a panel of antimicrobials included in the Sensititre COMPAN1F microtitre plates (table 3) showed that there was no resistance to amikacin among the companion animal isolates and only intermediate resistance to imipenem. Four per cent of isolates were resistant to gentamicin and ticarcillin/clavulanic acid. Higher levels of resistance were detected to ticarcillin (21 per cent), marbofloxacin (21 per cent) and enrofloxacin (33 per cent). As expected, high resistance to ceftiofur was detected with 92 per cent of isolates classed as resistant (table 3).
The human isolates were also tested using the Sensititre COMPAN1F microtitre susceptibility assay. A similar trend in resistance was observed with the highest number of isolates showed resistance to ceftiofur and the lowest to amikacin (table 3).
Based on disc diffusion susceptibility assays, no resistance was detected to ceftazidime, meropenem or tobramycin (data not shown). Two isolates were resistant to ciprofloxacin and a further three displayed intermediate resistance. The two resistant isolates (1055 and 823) were both of canine origin (online supplementary figure S1). The isolates with intermediate resistance were from a canine buccal swab (1095) and the left and right ear (467L and R) of the same dog.
Circulating strain types of P. aeruginosa
The Alere array tube has been used to determine the population structure of P. aeruginosa 47 and a database of 955 isolates can be used to infer the wider population.48 This typing technique was used here for the first time to classify the companion animal isolates. From the 24 isolates, 20 different array types were identified (table 1). Eight were novel array types not previously identified in the database.48 One array type had previously been associated only with the environment (isolate 2C12 from water). However, the remaining array types had been previously associated with human infections such as keratitis, catheter-associated and respiratory tract infection in CF and pneumonia (table 1). Using the database of array types to generate a P. aeruginosa population structure, the isolates were mostly distributed among the main P.aeruginosa clades. Five isolates were located as outliers (isolates 811, 1107, 856) and the two isolates from the same dog (467L and 467R) (figure 2). The latter two isolates share an array type with the widely used laboratory strain of P aeruginosa, PAO1.
Genome sequencing and analysis of the polymyxin resistome
Seven veterinary-associated isolates that displayed resistance to colistin and polymyxin B were selected for whole genome sequencing. A phylogenetic tree of the veterinary isolates and 342 of the available P. aeruginosa available genomes (figure 3) shows that four of the isolates were located in clade I and another two in clade II of the main P. aeruginosa population. Although none of the isolates was found to be PA7-like, isolate 856 (equine origin) was also diverse from the main population, located on a new arm of the phylogenetic tree (figure 3). This genome had 125303 single-nucleotide polymorphisms (SNPs) and 1884 indels compared with PAO1 and 130312 SNPs and 1985 indels compared with PA14 (online supplementary table S2). This was double the number of SNPs and indels observed for other sequenced isolates.
Thirty-one genes associated with resistance to polymyxins were analysed (online supplementary table S3). Multiple stop codons were identified in isolate 856. A stop codon in pyrC was identified in isolate 1098. Additional amino acid modifications resulting in a change in hydrophobicity were identified in all seven genomes (table 4).
In addition, the genome sequences were analysed using the Comprehensive Antibiotic Resistance Database (CARD) database62 (online supplementary figure S2). This showed the presence of resistance determinants including beta-lactamase (PDC-1–7, amr), efflux (smeB, mexBDFIY), fluoroquinolone and elfamycin resistance. Despite being the most divergent isolate by sequencing, isolate 856 had the fewest additional resistance genes.
In this study, we describe a higher prevalence of resistance to the polymyxin antimicrobials polymyxin B and colistin in a panel of P aeruginosa isolates from companion animals from the UK compared with human-associated isolates.43 To tour knowledge, this is the first description of increased resistance to colistin among companion animal clinical isolates. Using molecular typing methods, we identified that some of the strain types from animals have previously been identified in human infections. Genome sequencing revealed mutations leading to changes in the amino acid sequences that could contribute to resistance and revealed the highly diverse nature of some of the isolates. No resistance to some commonly used human antimicrobials (including meropenem, ceftazidime and tobramycin) was found; however, five isolates showed either resistance or intermediate resistance to a high priority critically important antimicrobials fluoroquinolone (ciprofloxacin). For the majority of the human clinically relevant antimicrobials there was little evidence for cross-resistance with veterinary antimicrobials; however, there did appear to be potential cross-resistance between the fluoroquinolones enrofloxacin and ciprofloxacin.
Several isolates within the companion animal sample set were from dog skin (including wounds) or ears. In canids, P. aeruginosa is a frequently isolated pathogen in chronic otitis externa and otitis media.22 The recommended management of canine otitis externa consists of identifying and treating the predisposing factors and primary disease, ear cleaning/flushing, appropriate topical therapy and if indicated, systemic antimicrobial medications.63 First-line topical therapies for canine otitis externa commonly contain polymyxin B as one of the active ingredients. It is authorised for use in several ear conditions and skin infections. A recent report examining antimicrobial susceptibility profiles of bacterial isolates in the canine ear in Australia identified P. aeruginosa as one of the five most commonly isolated bacterial pathogens.64 Of 3541 canine ear swabs, 35.5 per cent isolated P. aeruginosa. However, although they raised concerns of resistance levels in the other Gram-negative bacterial isolates (E. coli and Proteus species) to polymyxin B, the resistance in P. aeruginosa isolates was comparatively low to polymyxin B (7 per cent) and gentamicin (5 per cent).64 This is in contrast to the findings presented in the study, whereby polymyxin resistance levels of veterinary isolates were high. The precise reason for such differences is uncertain but may represent local differences in antimicrobial usage patterns and that the centres involved in this study were specialist referral centres and therefore potentially biased towards infections that display greater antimicrobial resistance.
As well as this use for canine otitis externa, polymyxin B, in the form of polymyxin B sulphate administered intravenously, is reported for use in equids as a treatment of endotoxaemia.65 Colistin is not used therapeutically in companion animals in the UK, although it is available as an authorised product for administration in various food production animals for the indication of enteric infections. However, there is little evidence in this study as to what the drivers of resistance may be within the companion animal niche of P. aeruginosa isolates.
Resistance to polymyxins has been attributed to modifications to the outer membrane including modifications to lipid A and LPS as well as two-component regulators. A study by Fernandez et al reported a relatively small polymyxin B resistome involving 17 susceptibility/intrinsic resistance determinants.36 They also demonstrated considerable cross-resistance in susceptibility to polymyxin B and to colistin,36 suggesting resistance determinants confer resistance to both antimicrobials. However, other genes have also been implicated in resistance.66 Seven of the veterinary isolates, chosen for their extreme resistance, were whole genome sequenced and mutations, surprisingly this included stop codons (a severe mutation resulting in early termination of transcription and therefore no or highly altered protein production), that were detected in genes previously implicated in polymyxin resistance such as pyrC, wapR, mpl and ampR. Many of the sequences studied showed modifications in genes leading to an altered amino acid, several of which had altered hydrophobic/hydrophilic properties which can result in conformational or activity changes in the protein. It is possible that a combination of these mutations may lead to a change in phenotype through membrane remodelling,66 although further studies would be needed to confirm this.
Using the array tube typing method, the P. aeruginosa veterinary isolates were found to be distributed throughout the P. aeruginosa population. These findings highlight the potential of transmission of these bacteria between humans and animals. Using whole genome sequencing on a limited subset of isolates, the veterinary isolates were generally clustered within the main P. aeruginosa population; however, one isolate, 856, did not cluster with any other previously sequenced isolate. Through mapping to the genomes of both PAO1 and PA14, the authors found that this isolate had double the number of SNPs and indels shown by any of the other isolates sequenced. The isolate was of equine origin and highlights the importance of studying alternative and under-represented niches in order to fully characterise the P. aeruginosa pan genome.
In conclusion, these findings raise concerns regarding the use of polymyxins within the veterinary community. Such isolates (from both dogs and horses) could then potentially act as a source of resistant isolates for the human-associated P. aeruginosa population. The findings suggest that future surveillance of polymyxin resistance in animal-associated P. aeruginosa strains is warranted. In this study, there was a small sample size from only one geographical location and no detailed history of prior antibiotic usage. These limitations warrant further, much larger studies. The veterinary setting is an often ignored niche that provides close proximity between humans and companion animals and therefore cross-infection of resistant organisms would be possible. The preservation of colistin as a ‘last resort’ effective antipseudomonal drug is of great concern.
The sequencing was performed as part of the International Pseudomonas Genomics Consortium.
AS and SP contributed equally.
Funding The work was supported by a University of Liverpool/Wellcome Trust Research taster fellowship and internal funding. AS is supported by a University of Liverpool/Wellcome Trust Research taster fellowship and JLF is supported by a Leverhulme Trust Early Career Fellowship. JJ is supported by a Cystic Fibrosis Canada postdoctoral fellowship. RCL is funded by Cystic Fibrosis Canada and by the Canadian Institute for Health Research (CIHR).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Data are available in a public, open access repository.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.