Globally, drug resistant infections are projected to cause 10 million human deaths at a cost of 100 trillion USD annually by 2050 if current trends continue (O’Neill, 2014). Selection pressure caused by the widespread use of antimicrobials in human medicine, veterinary medicine and agriculture has increased the dissemination and prevalence of AMR in bacteria (Laxminarayan and Heymann, 2012). Improving our understanding of the emergence and spread of AMR bacteria within and between host species populations is essential to inform effective control policies to prevent or reduce dissemination.

In the EU, over 100,000 cases of enterocolitis, costing an estimated €3 billion annually, are attributed to non-typhoidal Salmonella infections, of which Salmonella enterica subsp. enterica serovar Typhimurium is the second most common serovar (European Food Safety Authority and European Centre for Disease Prevention and Control, 2017). Salmonella Typhimurium is a zoonotic pathogen and the primary reservoir is thought to be food-production animals, with the majority of human cases deriving through the food chain (Majowicz et al., 2010), although more recent studies have suggested a more nuanced situation (Mather et al., 2013). Compared to antimicrobial sensitive strains, those resistant to therapeutically relevant antimicrobials pose a greater threat to public health because they are associated with higher morbidity and mortality rates (Helms et al., 2002; Cosgrove, 2006; Depuydt et al., 2008). Multi-drug resistant (MDR) strains, most notably definitive type (DT)104, have been documented to disseminate globally, causing infections in multiple host species including humans and food-production animals (Leekitcharoenphon et al., 2016).

National surveillance of S. Typhimurium in United Kingdom animal populations is conducted primarily for outbreak identification control purposes to limit the public health risk. Mandatory active surveillance is conducted for poultry, whilst passive surveillance is conducted for other food producing animals (DEFRA, 2007, 2008, 2009). Passive surveillance relies upon submission of samples by veterinarians or farmers for clinical diagnostic purposes. Identification of Salmonella from an animal source is reportable to the APHA where confirmation and phenotypic antimicrobial susceptibility testing and phage typing are conducted (Zoonoses Order, 1989; Animal and Plant Health Agency, 2017). Phenotypic AMR profile may not consistently correlate with bacterial genetic lineage, due to the importance of horizontal gene transfer for the transmission of AMR, and phage types can be polyphyletic (Chen et al., 2005; Petrovska et al., 2016). The resolution afforded through analysis of phage type and AMR phenotypic profiles is therefore restricted, but the combination of phage type and phenotypic AMR profile have proved useful for distinguishing between isolates for outbreak detection purposes.

Variation in the prevalence of resistance of bacterial pathogens to individual antimicrobials between food-production animal groups has been attributed to differences in selective pressures between production systems and dissemination of AMR clones, e.g., DT104. However, our understanding of the evolution and dissemination of resistant bacteria within and between food-production animal populations is limited (Young et al., 2009). AMR profiles unique to individual host species would not be expected if there was transfer of strains between host species populations. The rate of transfer of strains between host species could be influenced by ecological or pathogen related barriers which could result in differences in the observed AMR profiles between host species (Rabsch et al., 2002). The presence of shared AMR profiles between multiple host species can arise through a common source of infection, transmission of bacteria or mobile genetic elements (MGEs) carrying AMR genes between the host species populations or through independent emergence or acquisition of resistance determinants.

Opportunities for transfer of non-host-restricted strains between host species are numerous and include mixed-species farms, markets, common grazing, contaminated feed products and movements of fomites, personnel or wildlife (Davies and Wray, 1997; Robinson and Christley, 2007; Skov et al., 2008). The risk of transfer between host species is likely to vary between food-production animal species due to the differences in industry structure and standard biosecurity protocols (Lindström et al., 2012). Enhanced surveillance and control measures for United Kingdom poultry through the National Control Program (NCP) were introduced to reduce the prevalence of Salmonella Typhimurium and Salmonella Enteritidis on farms (DEFRA, 2007, 2008, 2009). Combined with biosecurity practices and a pyramidal industry structure, the United Kingdom S. Typhimurium population in chickens could be predicted to be relatively isolated from S. Typhimurium circulating in other host species. However, free-range poultry could be exposed to Salmonella transferred by wild birds and flies (Wales et al., 2010; Andrés et al., 2013).

Variation in AMR profile diversity within or between host species could indicate a greater diversity of AMR bacteria sources entering a host species population or indicate continuous evolution in the host population in response to greater selective pressures. Additionally, distinct differences in AMR profiles between host species could indicate epidemiological barriers to transmission. Here, the available national surveillance data for S. Typhimurium isolates, collected over an 11-year period, have been used to determine whether or not there are detectable differences in circulating AMR profiles within and between S. Typhimurium isolates from cattle, chicken and pig populations.

Between 2003 and 2014 a total of 3537 isolates (1115 from cattle, 248 from chickens and 2174 from pigs) were submitted to the APHA. The majority (95.8%) of submissions were obtained though passive surveillance with the remainder obtained through the NCP active surveillance system for poultry. Phenotypic antimicrobial test data were available for all except one isolate which was excluded from analysis. The data for these 3537 samples were used for the general analysis. A summary table of the antimicrobial resistance profiles observed in each host species is available in Supplementary Table 1. Production level data were available for 920 (82%) of isolates from cattle, 233 (94%) of isolates from chickens and 1180 (54.3%) of isolates from pigs.

Some isolates were collected on the same day from the same farm, as indicated by a shared submission reference. The proportions of isolates with shared submission reference varied between the host species 24.6% (274/1115) for cattle, 26.6% (66/248) for chickens and 48.9% (1062/2174) for pigs. The submission references indicate whether multiple isolates have been collected from the same farm on the same day, however, whether the isolates are from the same animal group or different animal groups on the farm is unknown. Multiple AMR profiles of isolates with a shared submission reference have been observed and these samples may have been taken from different animal groups (Table 2), however, this metadata was not available for analysis in this study. Isolates with the same submission reference which share an AMR profile could be the same strain, however, this cannot be determined given the resolution of the data.

Salmonella Typhimurium is a common cause of foodborne illness globally, with MDR strains of particular public health concern due to association with poorer patient outcomes. Reduction of prevalence in the food-production animal reservoir is important for both animal and human health and improving understanding of S. Typhimurium epidemiology including AMR dynamics in the food-production animal populations and has the potential to inform interventions. Using national surveillance data collected over an 11-year time period, AMR profiles of S. Typhimurium were compared within and between food-production animal host species populations, as variation in observed AMR profiles and AMR profile diversity could indicate variation in AMR dynamics between the host species.

Active surveillance samples from chickens from 2007 onward were included, however, the majority of samples were obtained through passive surveillance and the data are therefore biased toward strains causing clinical disease in food-production animals or identified through diagnostic testing for other diseases. Asymptomatic/subclinical infection is common in some animal species and the full diversity of circulating S. Typhimurium at the population level is therefore unlikely to be captured (Rostagno et al., 2012; Troxell et al., 2015). Submissions are also influenced by variation in farmer and veterinarian behavior (Schwaber et al., 2004; Laupland et al., 2007; Rempel and Laupland, 2009). Passive surveillance systems have been found to capture greater diversity of rare AMR profiles of Salmonella in food-production animals in Canada (Mather et al., 2016), and clinical isolates from food-production animals have been observed to have higher diversity of AMR phenotypes compared to asymptomatic food-production animals (Perron et al., 2007; Afema et al., 2015). Utilizing the national surveillance data should enable meaningful comparison of AMR profile diversity between and within food-production animal populations, although variations in surveillance between host populations could also contribute to any observed differences.

Ecological diversity analyses revealed variation in observed AMR profile diversity both between host species and between production types for chickens and pigs. The AMR profile diversity of isolates from pigs was greater than for isolates from chickens and cattle across SE, SD and BP diversity measures. The percentage of isolates resistant to one or more antimicrobials is almost two-fold higher for pigs (98.1%) compared to chickens (52.4%). This may in part account for the higher ecological diversity measures of isolates from pigs compared to chickens. The observation of 15.9% of isolates from pigs having an AMR profile unique to pigs suggests that some strains causing clinical disease in pigs are either not present, do not persist or are not causing clinical disease in cattle or chicken populations. The distinct separation of the distribution of the observed number of AMR profiles unique to pigs from that of chickens, controlling for sample size, indicates this finding is unlikely due to differential sampling intensity. Inference is limited due to the nature of passive surveillance; however, these observations suggest alternate sources of AMR bacteria and/or variation in selective evolutionary pressures between host species populations.

The separation of expected and observed distributions of AMR profiles shared by all food-production animal populations indicates that interchange of S. Typhimurium strains between host species is not complete, with differences in the circulating S. Typhimurium population between host species groups. Variation in AMR profile diversity between host species could be attributable to differences in observed lineages between host species. Bacterial typing information was limited to phage type, which affords limited resolution and does not necessarily correlate with lineage; multiple lineages can have the same phage type and AMR profile combination, and phage switching can occur (Baggesen et al., 2010; Pang et al., 2012). The consistency in AMR profile AmCSSuT and phage type DT104 combination could indicate the presence of a common single lineage between multiple host species; this particular AMR profile is also known to be chromosomally encoded in DT104 (Boyd et al., 2001). However, the most common phage type was inconsistent across host species for other AMR profiles including T and AmCSSuSxtT, suggesting different genetic lineages with the same phenotypic AMR profile may be circulating in different host species.

The selective pressures driving the maintenance of high proportions of MDR S. Typhimurium are not well understood and may not be due to ongoing selective pressure of antimicrobial use in food-production animal populations alone (Davis et al., 2002). Several factors, in addition to antimicrobial use, may influence variation in S. Typhimurium between host species including host immunity, vaccination status, biosecurity and industry structure. Vaccination can result in reduction in S. Typhimurium prevalence, reducing the risk of selection of AMR strains (Dagan and Klugman, 2008). Lower AMR profile diversity of isolates from chickens compared to from cattle and pigs could be expected due to widespread vaccination of laying poultry (>85%) and broiler-breeder flocks against S. Typhimurium, maintenance of Salmonella free grandparent and parent flocks and a stringent active NCP with surveillance and control measures to contain outbreaks (DEFRA, 2007, 2008, 2009; British Egg Industry Council, 2013).

Investigation of association between AMR profile diversity and risk factors within host species populations were restricted by the metadata available for retrospective analysis and therefore limited to high level observations. Higher AMR diversity was observed in isolates from breeder compared to rearer and finisher pigs. No isolates from piglets or at farrowing stage were available. This is consistent with previous findings of acquisition of AMR with successive production stages and higher pulse-field gel electrophoresis cluster diversity of monophasic S. Typhimurium in sows compared to piglets (Fernandes et al., 2016). Compared to layer chicken flocks, broilers were observed to have higher AMR diversity. The relatively high percentage of isolates from layers sensitive to all tested antimicrobials compared to isolates from broilers contributes to the lower diversity of AMR profiles in isolates from layers. Stringent restrictions on antimicrobial use in layer poultry (NOAH, 2016), may place greater emphasis on the need to prevent incursion of infection and reduce selective pressure for AMR to a wider range of antimicrobials. In addition, the high vaccination rates of laying poultry (British Egg Industry Council, 2013) and host genetic differences (Lumpkins et al., 2010) may contribute to variation in AMR profile diversity between the poultry sectors.

Animals are a potential reservoir for AMR bacteria and AMR genes of public health concern, therefore it is important to understand the trends of AMR in animal populations. Despite the limited discriminatory ability of phenotypic AMR profiles and phage type, variation in AMR dynamics of S. Typhimurium were observed within and between host species populations. The transition to whole genome sequencing technology by the APHA provides an enhanced utility for surveillance over time and space, and importantly an opportunity for improved understanding of AMR dynamics between food-production animal populations and humans at the national level. If combined with enhanced metadata, this could ultimately result in higher food security by identifying where intervention strategies may be most effectively applied and provide the opportunity for delivery of an enhanced One Health strategy.

Data was collected through routine national surveillance and therefore did not require study permissions.

KM and AM contributed to the design of the study. KM performed the analyses. KH and LP were responsible for provision of data. KM wrote the first draft of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.