Antimicrobial resistance in clinical bacterial isolates from horses in the UK

Background: Surveillance of antimicrobial resistance (AMR) in horses is important to aid empirical treatment decisions and highlight emerging AMR threats. Objective: To describe the AMR patterns of common groups


| INTRODUC TI ON
Antimicrobial resistance (AMR) is a global problem with implications for both human and equine health. 1 AMR in horses poses a threat not only to the individual horse but also to the owners and caregivers as well as to the environment from faecal and urine excretion of antimicrobials and their metabolites. 2 This problem is more concerning since transmission of multidrug-resistant (MDR, isolates with acquired non-susceptibility to at least one antimicrobial in three or more antimicrobial classes) pathogenic strains from animals to humans has also been reported. 3 There are also few antimicrobials available for use in UK horses due to a limited number of drugs being authorised for use in this species, cost implications and safety concerns due to hindgut fermentation. Certain antimicrobial classes, such as macrolides, which are commonly used in humans and other veterinary species, are rarely used in horses over 12 months of age (although macrolides are used in foals in the treatment of Rhodococcus pneumonia). Similarly, lincosamides are never used in horses due to risk of severe and potentially fatal colitis. 4,5 Some antimicrobials such as doxycycline and enrofloxacin, which are considered safe for use in horses but are not authorised for equine use in the UK, are frequently prescribed under the cascade for treating equine infections. 6 Other antimicrobials authorised for use in other veterinary species are rarely used in adult horses due to cost (eg amoxicillin), even though they are considered safe to use in adult horses. 6 Surveillance of AMR in clinical isolates is important in order to monitor and detect emerging resistance patterns, which may be a threat to horse or human health. In addition, surveillance data can be used to guide policies on antimicrobial use and local geographical empirical therapy. Antimicrobial stewardship and appropriate antimicrobial prescribing practices are also important to ensure antimicrobials remain effective, especially with limited treatment options in the horse. Intrinsic resistance (IR), the innate ability of wild-type bacterial species to resist activity of a particular antimicrobial, 7 is particularly high in some bacterial species that are commonly isolated in horses, eg Enterococcus spp. and Pseudomonas spp., 8 which further limits treatment options and may be compounded by acquired resistance also present in such bacteria.
Previous reports in horses have mostly focused on susceptibility patterns of particular bacteria 9 or from a particular sample site 10 or age group, 11 or used results from a single hospital or laboratory. 9 Recent publications from France have reported on susceptibility patterns from a variety of bacteria from clinical submissions from 2012 to 2016 and identified increasing resistance to trimethoprim-sulfamethoxazole in Streptococcus spp. and E. coli. 12 Another report from France identified a decrease in MDR in E. coli and Staphylococcus aureus clinical isolates from 2006 to 2016, however, prevalence of MDR still remained above 18% and 22.5% for S. aureus and E. coli respectively. 13 The Defra AHT BEVA Equine Quarterly Disease Surveillance Report 14  In the UK, a variety of different types of independent diagnostic laboratories operate; these include those based within large private equine hospitals, university-based laboratories, large commercial veterinary laboratories that predominantly process small animal submissions with fewer equine submissions as well as small in-house laboratories with mainly internal submissions. Currently, there are no standardised veterinary laboratory methods in the UK, although most laboratories use Clinical and Laboratory Standards Institute (CLSI) standards for performing antimicrobial susceptibility testing (AST) and for interpretation of clinical breakpoints. 15 Culture and susceptibility data are crucial for informing treatment decisions and determining emerging AMR threats. Therefore, the aim of the study was to describe the prevalence of bacteria most commonly isolated from clinical specimens and patterns of AMR among bacterial isolates from equine clinical samples submitted to diagnostic laboratories in the UK over a 12-month period in 2018. We hypothesised that there would be increased MDR from submissions from referral practices compared with first opinion practices, as referral caseloads are more likely to have already been administered firstline antimicrobial treatment, with subsequent referral only following treatment failure.
2018, from six equine diagnostic laboratories across England, including commercial, practice-based and University-based laboratories.
Microorganisms isolated from positive cultures were identified using commercial biochemical tests including API kits (Biomerieux) and GNID and GPID Sensititre Identification plates (TREK Diagnostic Systems) at four of the laboratories, while two used the Matrix-assisted laser desorption ionisation time-of-flight mass spectrometry (MALDI-TOF MS) platform for bacterial species identification (Bruker Daltonics).
AST was performed using minimum inhibitory concentration (MIC) at two laboratories while the remaining four used Kirby-Bauer disc diffusion testing. All laboratories used CLSI methods and used CLSI breakpoints where available for horses. When no breakpoints were available for horses, other veterinary breakpoints were used, followed by human breakpoints (CLSI or EUCAST) if no other veterinary breakpoints were available. 15 The individual breakpoints used by each laboratory for the most common bacteria in this study are listed in Table   S1 and although many breakpoints were identical (eg benzyl-penicillin for β-haemolytic Streptococcus spp., Pasteurella spp. and Actinobacillus spp., oxytetracycline and doxycycline for Enterobacteriaceae and folate pathway inhibitors for Acinetobacter spp.), some differed between laboratories (eg oxytetracycline and doxycycline for bacteria other than Enterobacteriaceae). No laboratories used urine-specific breakpoints.
Breakpoints were displayed to reflect whether MIC or Kirby-Bauer disc diffusion testing was performed. Laboratories also used different antimicrobial susceptibility panels (Table S2). From all laboratories, a range of information was provided including a unique submission identification code: first part of the postcode of the submitting veterinary practice address; date the results report was produced; the type or anatomical location of the submitted clinical specimen; the bacterial culture and AST results for each bacterial species isolated from clinical specimens.
Due to laboratories using different antimicrobial panels, antimicrobials were grouped by class. IR was not included when determining the susceptibility of isolates. Bacteria giving intermediate results, ie not fully susceptible, were categorised as resistant. 8 Table 1   and the National Statistics Postcode Look-up (NSPL) were used to determine if the submitting practice was a practice that accepted referral cases. Submissions from practices that accepted referral cases and with an ambulatory branch also were categorised as referral.
Orthopaedic infections included positive synovial cultures, septic tendinitis and osteomyelitis submissions and were grouped with SSI and CRI as these infections are difficult to treat and often require surgery and hospitalisation. Unknown submissions included those where no site was reported (n = 520 isolates from 447 submissions), while 'others' were those present in less than 100 isolates and included the following sites; faecal (n = 25), peritoneal fluid (n = 33), liver (n = 11), dental (n = 4), gastric (n = 7) and rectal (n = 5)

submissions.
Bacterial species were separated based on their IR and ge- were those which were susceptible to all classes of antimicrobials tested (IR excluded) and described in Table 1; 'Resistant to 1 or 2 classes' were those resistant to one or two antimicrobials from different classes; MDR was defined as isolates with acquired non-susceptibility to at least one antimicrobial in three or more antimicrobial classes. 'XDR isolates' were those which were resistant to all classes of antimicrobials considered. 8 Isolates with 'no readily available treatment for adult horses in the UK' included those that were not susceptible to any of the following antimicrobials; penicillin (penicillin G), 3rd-generation cephalosporins (3GC; ceftiofur), aminoglycosides (gentamicin/amikacin), tetracyclines (oxytetracycline/doxycycline), folate pathways inhibitors (trimethoprim-sulfamethoxazole), fluoroquinolones (enrofloxacin) or phenicols (chloramphenicol). Polymyxin B, although tested for using Kirby-Bauer disc diffusion methods by two laboratories was not included due to inaccuracy of this method; testing using MIC by microbroth dilution is advocated. 22 TA B L E 1 List of antimicrobial classes and agent used to define multidrug resistance (MDR) for common bacterial isolates in horses (modified from resources in literature such as Magiorakos et al. (2012) isolates respectively (breakdown shown in Table S4).
The submissions came from 208 veterinary practices distributed across the UK (shown in Figure 1). The most common

| Antimicrobial resistance
The proportion of resistance of bacterial isolates is shown in Tables 2 and 3

| Submitting practice demographics
Gram-negative and 3,142 Gram-positive isolates with 2,820 isolates from referral and 2,744 isolates from first opinion practices. The proportions of MDR in bacterial isolates were significantly different in referral hospitals compared with first opinion practices (Table 5). MDR was significantly higher in submissions from referral hospitals in E. coli   Note: Multidrug-resistant (MDR) isolates were those with acquired non-susceptibility to at least one antimicrobial in three or more different antimicrobial classes. Extensively drug-resistant (XDR) isolates were those, which were resistant to all classes of antimicrobials tested. 'No readily available treatment for adult horses in the UK' included those isolates, which were resistant to commonly used (authorised or non-authorised) antimicrobials available for adult horses in the UK. All calculations are based on antimicrobials considered in Table 1 and excludes intrinsic resistance. a For α-haemolytic Streptococcus spp. and Enterococcus spp. only three classes on antimicrobials were considered, hence multidrug resistance is the same as resistance to all classes of antimicrobials tested (XDR).

TA B L E 3 (Continued)
sepsis (orthopaedic infection). 33 However, depending on the sever-   and other veterinary studies 39 where information was commonly missing from diagnostic submission forms. Isolates from unknown sample sites often also had high prevalence of MDR, however, this is of limited value without knowing the source of the samples. We submissions are likely to originate from some large equine practices that have both hospital and ambulatory branches and may undertake more poor performance/subclinical respiratory disease screening in sport and racehorses rather than sampling horses with overt clinical disease which may have biased these results. In these horses, S.
zooepidemicus, for example, is viewed as performance limiting, which may well be associated with tracheal mucus and inflammatory airway disease and is likely to be treated with antimicrobials. 48 As it was not possible to distinguish between upper and lower respiratory tract submissions, it is not possible to explore this further.
It is important to highlight that despite there being higher prevalence of MDR in β-haemolytic Streptococcus spp. in first opinion submissions than referral submissions, overall MDR in β-haemolytic Streptococcus spp. was only 8.3% and importantly 97.5% of isolates were susceptible to penicillin, which is the current first-line treatment for equine respiratory infections listed in the BEVA Protect ME toolkit. 49  should not be overlooked due to their low MDR as they pose a therapeutic challenge when involved in infection. 55 These bacteria, as well as posing a risk for the individual horse, are also of zoonotic concern as they have also been reported in humans. A genotypically identical strain of Pseudomonas spp. from a water source has been reported as a cause of an outbreak of equine endometritis in Australia, 56 from a variety of equine samples in Ireland, 57 from companion animals 58 and from human cystic fibrosis patients. 59 Enterococcus spp.
are common pathogens in hospital-acquired infections in humans, 60 equine synovial infections 61 and companion animals 62 and have been associated with increased mortality in foals. 63 However, they are often present in human and animal gut flora, 64 on skin 64 and urogenital mucosa 65 and therefore are often present in clinical specimens as contaminants. 66 penicillins and some cephalosporins due to better accuracy. 85,86 Larger and more modern laboratories are commonly using automated microbroth dilution methods due to its versatility and ability to determine the MIC likely to achieve effective antimicrobial plasma concentration. This means that if the MIC indicates that an isolate is susceptible but at the higher end of the range, near the epidemiological cut-off value (ECOFF), it may require a higher dose to achieve therapeutic concentrations. 87 Although there are also inaccuracies in MIC, such that the accepted MIC ranges of quality control strains, often span over two to three dilutions and even four dilutions in some cases. 88 Smaller laboratories often use Kirby-Bauer disc diffusion methods due to lower costs and no requirement for extensive equipment. Furthermore, another limitation was that a pooled approach to reporting was used by combining some bacterial species based on their similarities in intrinsic resistance patterns. This is similar to human studies 8 and was done in order to avoid having several smaller groups making conclusions and presentation of results difficult, but the authors acknowledged that this does somewhat limit the application of these pooled results. may be crucial in this harmonisation process. However, veterinary laboratories must adopt the same laboratory standards in order to achieve this. 99 There are many barriers to implementation of harmonised methods including cost and availability of equipment, skills and training of the laboratory staff, and the time-consuming nature of updating the latest breakpoints while running a commercial service. As there is no governing body which veterinary laboratories have to subscribe to that regulates or audits methods and results, laboratories are able to use their own in-house methods. Despite these limitations, the results from this study provide relevant and updated information on the current AMR situation in clinical bacterial isolates from horses in the UK.
Apart from establishing if practices were referral or first opinion, it was not possible to determine further practice characteristics such as case load. Descriptive spatial analysis suggested there may be geographical differences in levels of resistance prevalence, as has been observed in humans. 100,101 However, in the current study, data were based on the submitting practice postcode, rather than horse or owner location, and it is therefore not accurate to compare geographical differences. Furthermore, the submissions were from a limited number of diagnostic laborato- are not responding to first-line treatment.

| CON CLUS ION
This study provides important information about patterns of AMR in major equine pathogens in the UK. Our results are useful for veterinarians to guide their initial empirical treatment. Our results also emphasise the importance of antimicrobial stewardship and judicious use of antimicrobials especially in horses undergoing surgery as SSI/CRI and orthopaedic infections had increased levels of MDR.
It also highlights the need for concerted efforts for harmonisation and standardisation of culture and susceptibility methods at least at national level to support AMR surveillance. Furthermore, resistance patterns were different in referral and first opinion submission, which is vital information for risk assessment and implementation of biosecurity measures. This study only provides information on equine isolates submitted during 2018 and ongoing surveillance is recommended to determine differences in seasonality and to detect emerging trends in AMR.

ACK N OWLED G EM ENTS
The authors would like to thank all the contributing diagnostic laboratories for their time and effort in collating these data includ- Laboratory. We would also like to thank Ruth Ryvar for laboratory assistance and David Singleton for help in generating the choropleths.

CO N FLI C T O F I NTE R E S T S
No competing interests have been declared.

AUTH O R CO NTR I B UTI O N S
This project was executed by C. Isgren with assistance from N.

E TH I C A L A N I M A L R E S E A RCH
Ethical approval for the study was granted by the University of Liverpool Veterinary Research Ethics Committee (VREC544). Data were collected confidentially, and all laboratories provided written consent to participate in the study.

I N FO R M E D CO N S E NT
Explicit owner informed consent for inclusion of samples from animals in this study was not sought but owners were given the option to opt out of research. Data from laboratory submissions were excluded where the option to exclude data from future research had been selected.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/evj.13437.

DATA ACCE SS I B I LIT Y S TATE M E NT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.