Shared antibiotic resistance and virulence genes in Staphylococcus aureus from diverse animal hosts

The emergence of methicillin-resistant Staphylococcus aureus (MRSA) poses an important threat in human and animal health. In this study, we ask whether resistance and virulence genes in S. aureus are homogeneously distributed or constrained by different animal hosts. We carried out whole genome sequencing of 114 S. aureus isolates from ten species of animals sampled from four New England states (USA) in 2017–2019. The majority of the isolates came from cats, cows and dogs. The maximum likelihood phylogenetic tree based on the alignment of 89,143 single nucleotide polymorphisms of 1173 core genes reveal 31 sequence types (STs). The most common STs were ST5, ST8, ST30, ST133 and ST2187. Every genome carried at least eight acquired resistance genes. Genes related to resistance found in all genomes included norA (fluoroquinolone), arlRS (fluoroquinolone), lmrS (multidrug), tet(38) (tetracycline) and mepAR (multidrug and tigecycline resistance). The most common superantigen genes were tsst-1, sea and sec. Acquired antibiotic resistance (n = 10) and superantigen (n = 9) genes of S. aureus were widely shared between S. aureus lineages and between strains from different animal hosts. These analyses provide insights for considering bacterial gene sharing when developing strategies to combat the emergence of high-risk clones in animals.


Results
Phylogenetic diversity of animal-associated S. aureus in New England. We obtained a total of 114 high quality draft genomes of S. aureus isolates obtained through routine diagnostic tests of clinical specimens from diseased animals. These were submitted to the New Hampshire Veterinary Diagnostic Laboratory (NHVDL) from October 2017 to October 2019 ( Fig. 1A and Supplementary Table S1). We obtained the isolates from four states in the United States: New Hampshire (n = 74), Maine (n = 13), Massachusetts (n = 12) and Vermont (n = 15) (Fig. 1B). The majority of isolates came from cows (n = 30), dogs (n = 28) and cats (n = 25) (Fig. 1C). Other domestic animals from which we obtained isolates included horses, goats and rabbits, while wild animals included birds, rodents, deer, rabbits and a tortoise.
De novo assembly of the 114 genomes generated sequences of sizes ranging from 2.68 to 2.90 Mb (mean = 2.77 Mb). The number of predicted genes ranged from 2429 to 2739 per genome (mean = 2570) (Supplementary Table S1). The pan-genome of the New England S. aureus population consisted of 7500 orthologous gene families. The genes in the pan-genome were categorized into core genes (present in 99% of genomes), soft-core genes (present in 95% to < 99% of genomes), shell genes (present in 15% to < 95% of genomes), and cloud genes (present in < 15% of genomes). We identified 1773 core genes (present in 113-114 genomes), 116 soft core genes (present in 109-112 genomes), 1158 shell genes (present in 18-108 genomes) and 4453 cloud genes (present in 1-17 genomes) (Supplementary Table S2). The combined core and soft-core genes comprised 25.19% of the pan-genome, while the combined shell and cloud genes (which together make up the accessory genome) comprised 74.81% of the pan-genome. We identified 1716 genes, representing 22.88% of the species pan-genome, that are unique to a single strain.
The maximum likelihood phylogenetic tree based on the alignment of 89,143 SNPs of the core genes revealed many deep branching lineages consisting of 31 previously identified STs (Fig. 1A). The most common STs were ST5 (n = 15 genomes), ST8 (n = 7 genomes), ST30 (n = 8 genomes), ST133 (n = 7 genomes) and ST2187 (n = 16 genomes). Despite the unequal number of isolates from the four states (Fig. 1B), the phylogeny showed a lack of structure relative to the state from which the isolate originated.
Distribution of acquired genes related to antibiotic resistance. We determined the presence of acquired genes associated with resistance to different antibiotics. These genes represent a variety of resistance mechanisms (drug inactivation, target alteration, efflux, target replacement, target protection) based on definitions in the CARD database 17 . We identified a total of 30 resistance genes across the entire dataset ( Fig. 2A and Supplementary Table S3).
In silico detection of the mecA gene from the genome sequences revealed seven MRSA isolates and 107 methicillin-susceptible S. aureus (MSSA) isolates. The mecA gene encodes an extra penicillin-binding protein (PBP2a) that has low affinity to virtually all beta-lactam antibiotics 18,19 . The mecA gene is carried by a mobile chromosomal cassette SCCmec and is classified into types based on the combination of the ccr and mec complexes they carry 19 . To date, there are 14 structurally distinct SCCmec types (I-XIV) that have been described [19][20][21] . SCCmec typing of the New England S. aureus population revealed the presence of types II (n = 4 genomes), III (n = 1 genomes) and IV (n = 2 genomes) ( Fig. 2A and Supplementary Table S1). SCCmec type II was detected in ST 5, type III in ST1 and type IV in STs 8 and 72. We did not detect the gene mecC, which is a divergent form of mecA and also mediates beta-lactam resistance 22 . We found a slight discrepancy between the in silico detection of the mecA gene and in vitro phenotypic testing for methicillin resistance. There were two isolates (3659B and 4052C) whose genomes did not contain the mecA gene but were phenotypically tested as MRSA. These alternative mechanisms warrant further study. www.nature.com/scientificreports/ Resistance-related genes associated with antibiotic efflux systems were ubiquitous in our dataset. We identified 11 such acquired genes, of which seven were found in all 114 genomes. These seven genes included norA (fluoroquinolone resistance), tet (38) (tetracycline resistance), arlRS (fluoroquinolone resistance), lmrS (multidrug resistance) and mepAR (multidrug and tigecycline resistance). The expression of norA is affected by the two-component regulatory system ArlRS 23 . lmrS confers resistance to aminoglycosides, macrolides, phenicols, diaminopyrimidine and oxazolidinone 24 . mepA is an efflux protein regulated by mepR and part of the mepRAB cluster, which makes up the multidrug and toxin extrusion (MATE) mechanism in S. aureus 25 . Also widely detected was mgrA (also known as norR), which is a regulator for norA and tet (38) 26 . The mgrA gene was detected in all genomes except two.
Other resistance genes were less frequently detected. The gene fosB (fosfomycin resistance) was detected in 60 genomes and was present in some genomes from STs 5, 6, 8, 30, 72 and 133. We detected the gene blaZ (betalactam resistance) in 43 genomes and was present in some genomes from STs 5, 6, 8, 30, 72 and 398. In all, every genome carried at least eight resistance-related genes, of which seven genomes have at least 13 resistance related genes (Fig. 2B). Among these less common genes, their distribution across disparate parts of the phylogeny indicates the potential for multiple independent HGT events across different genetic backgrounds, rather than the spread of resistance through clonal expansion.
We next compared the frequency of resistance-related genes among isolates from different animal species. To maintain consistency in our comparison, we focused only on those animals with the highest number of isolates (cats, cows, dogs). We detected a median of ten resistance-related genes (range = 8-15) in S. aureus from cats, ten resistance-related genes (range = 8-18) in S. aureus from dogs and eight resistance-related genes (range = 8-11) in   (Fig. 2C). The aad(6) gene (aminoglycoside resistance) was found only in one isolate from a cat, whereas the resistance genes dfrC (diaminopyrimidine resistance), lnuG (lincosamide resistance), mecI and mecR1 (beta-lactam resistance), and tetM (tetracycline resistance) appeared only in isolates from dogs. Several other resistance genes including mecA, mphC (macrolide resistance), msrA (erythromycin and streptogramin resistance), tetK (tetracycline resistance), APH(3')-IIIa (aminoglycoside resistance), ermA (resistance to streptogramin, macrolide and lincosamide) and SAT-4 (nucleoside resistance) were present in S. aureus from dogs and cats, but not from cows. Three STs (STs 151, 352, 2187) which were isolated exclusively from cows exhibited a lack of accessory antimicrobial resistance genes, except for ermC found in a single isolate in those three STs. All other resistance genes found in these three STs were core genes found across all other genomes in the study. The total number of resistance genes varied between isolates from cats and cows, as well as between isolates from dogs and cows (p = 2.74e-05 and 4.03e-05, respectively; Welch's t-test), but not between isolates from dogs and cats (p = 0.527; Welch's t-test) (Fig. 2C).

Distribution of staphylococcal virulence genes.
Animal-associated S. aureus carry numerous virulence-related genes. We detected 80 virulence genes in all 114 genomes and of which ten are known as superantigens ( Fig. 3A and Supplementary Table S4). Superantigens constitute a family of secreted toxins that trigger excessive non-specific T-cell activation and proliferation, resulting in the overproduction of cytokines 27 . These potent toxins cause a variety of human diseases from transient food poisoning to lethal toxic shock 27 . To date, there are at least 24 known superantigens in S. aureus 28 . Of the 114 genomes in our dataset, 39 genomes harbor 1-6 distinct superantigens that were distributed in divergent parts of the phylogeny. Similar to the phylogenetic distribution of the antibiotic resistance genes described above, the distribution of the superantigens across  . STs 151, 352, and 2187 which were isolated exclusively from cows contained no superantigen genes, while ST 2187 exhibited fewer virulence genes compared to all other genomes. ST 398, collected from diverse animal hosts, also showed fewer virulence genes and was additionally void of superantigen genes. A notable virulence factor in S. aureus is the Panton-Valentine Leukocidin (PVL) pore-forming cytotoxin assembled by the genes lukF-PV and lukS-PV 29 . PVL toxin-producing strains causes leukocytolysis and tissue necrosis and are often associated with community acquired MRSA infections in humans 30 . In our dataset, we detected 96 and 3 genomes that carry lukF-PV and lukS-PV, respectively. We also compared the number of virulence genes among isolates from the three most common animal hosts. Staphylococcal virulence genes were unevenly distributed among the three animal hosts. Isolates from cats harbored a median of 60 virulence genes (range = 55-67), 62 (range = 55-67) in dogs and 54 (range = 52-65) in cows (Fig. 3B). The total number of virulence genes varied between strains from cats and cows, as well as between isolates dogs and cows (p = 6.19e-06 and 9.86e-06 respectively; Welch's t-test), but not between isolates from dogs and cats (p = 0.331; Welch's t-test) (Fig. 3B). When superantigens were considered, the number of superantigen genes carried by a genome ranged from 0-2, 0-3 and 0-6 in isolates from cats, cows and dogs, respectively (Fig. 3C). There was also no significant difference in the number of superantigen genes between strains from dogs and cats (p = 0.366; Welch's t-test). Three superantigen genes (sec, sell, tsst-1) were found in isolates from dogs, cats and cows. Six superantigen genes (eta, sea, sed, seh, selk, selq) were found only in dogs and cats, but not cows. Lastly, seb was found only in a single dog.
Widespread gene sharing between animal-associated S. aureus. Evidence for HGT in S. aureus and its contributions to adaptation in animal hosts has been demonstrated previously 16,31,32 . However, little is known about the spread of potentially transferrable resistance and superantigen genes among the animal hosts of S. aureus. We mapped the presence or absence of these genes shared between isolates from the three most www.nature.com/scientificreports/ common animal hosts (cats, cows, dogs). For the 15 resistance genes found in more than one isolate, the genes blaZ and fosB were present in isolates from all three animal hosts (Fig. 4A). We detected blaZ in 2, 17 and 17 isolates from cows, cats and dogs, respectively. We detected fosB in 2, 19 and 23 isolates from cows, cats and dogs, respectively. The genes which were detected in isolates from both cats and dogs, but not cows, were APH(3')-IIIa, ermA, SAT-4, mecA, mphC, msrA, and tet(K). The gene ermC was present in isolates from cats and cows, but not dogs. Lastly, there were no genes that were shared between dogs and cows, but not cats. Staphylococcal superantigens genes are often located in mobile genetic elements, such as prophages, transposons, plasmids and pathogenicity islands 33,34 , thus facilitating their mobilization. We found that superantigen genes were widely distributed among isolates from different animal hosts (Fig. 4B). We detected sec in 1, 4, and 2 isolates from cows, cats and dogs, respectively. We detected tsst-1 in 1, 2, and 7 isolates from cows, cats and dogs, respectively. The genes sea, sed, seh, selk, selg, and eta were present in isolates from both dogs and cats, but not cows. No superantigens were shared between isolates from dogs and cows, but not cats, Lastly, no superantigen genes were shared between isolates from cats and cows, but not dogs. Overall, these results showed widespread gene sharing between S. aureus gene pools from different host species.

Discussion
While humans are considered its primary reservoir, S. aureus can readily cross species barriers and infect new hosts 16 . Host-jumping events of bacterial pathogens are likely amplified by agricultural intensification, habitat encroachment and animal domestication 35,36 . In the case of S. aureus, the range of eukaryotic species that it can colonize as a commensal or opportunistic pathogen remains unclear. Regardless, the remarkable capacity of S. aureus to adapt to new or multiple hosts thus makes it a formidable bacterium that can threaten animal health, agriculture and the economy. Host-jumping events are often associated with acquisition of genetic elements from host-specific gene pools that confer traits required for survival and adaptation in the new host niche 16 . These traits include a variety of virulence factors such as superantigens that can be used to manipulate innate and adaptive immune responses 28 . For example, HGT in S. aureus mediated by mobile genetic elements occurs rapidly after a host-jumping event, potentially affecting the innate immune response of the new host 16 . During animal colonization, frequent HGT is facilitated by few genetic barriers to HGT in vivo (e.g., restriction-modification systems) and the successful replication and integration of different mobile elements 37,38 .
Here, we sought to elucidate the population genomic structure of 114 S. aureus isolates sampled from diseased animals in New England, USA from 2017-2019. We found that many multidrug resistant STs were detected in multiple wild and domestic animals. Notable were STs 5, 8 and 30, which are major S. aureus clones that are implicated in nosocomial and community-associated infections in humans 3,39 . ST133 isolates has been identified in healthy donkeys destined for food consumption in Tunisia 40 , caprine and ovine animals in Australia 41 , and the  www.nature.com/scientificreports/ gut of healthy humans in Spain 42 . ST2187 has a long history of association with cows, and more specifically their milk 43,44 . Although some S. aureus lineages are specifically adapted to a narrow host range on a short evolutionary time scale 45 , as in the case of STs 151, 352 and 2187 in our study, this may be due to uneven and scarce sampling done in animals, particularly of wildlife species. Also notable is that 43 out of 114 genomes from six STs presented blaZ. For comparison, high penicillin susceptibility rate has been reported in MSSA from bloodstream infections in humans, primarily from CCs 5 and 398 46 . Penicillin may be considered a therapeutic option in the treatment of animal infections, but ST identity should be carefully considered as blaZ appears to be more phylogenetically widespread in animal-associated S. aureus. As for the two isolates whose genomes did not contain the mecA gene but were phenotypically tested as MRSA, a similar finding has been reported in four isolates from the Scottish MRSA Reference Laboratory and which has been posited to indicate the existence of alternative mechanisms of beta-lactam resistance 47 . Whole genome sequencing also revealed that those isolates phenotypically tested as MSSA often harbor numerous antibiotic resistance determinants. This result highlights the need to further investigate resistance characteristics beyond methicillin resistance in animal-associated S. aureus, which are often overlooked in many surveillance studies. In addition, such efforts will be instrumental in advancing the One Health concept, focused on the interconnectedness of animal, human and environmental well-being 10 . Genes that encode for antibiotic resistance and superantigens were shared not only between divergent genetic backgrounds (or STs) of S. aureus but also between the animal hosts in which they reside. Hence, the ability of S. aureus to colonize multiple animal hosts means that mobilizable DNA may be disseminated more widely, creating a shared pool of resistance and virulence that is not limited by the animal hosts that harbor them. Our study also showed that cats, cows and dogs were frequent carriers of numerous S. aureus clones, each with distinct repertoire of resistance and superantigen genes. Hence, these animals are a major reservoir of clinically relevant genes and high-risk clones that can be transmitted to humans through frequent close contact. Overlapping ecological niches and/or physical proximity between animal hosts (e.g., pets in the same household, livestock animals in the same or nearby farms, interactions at the interface between wild and domestic animals) can certainly promote multihost colonization and frequent gene sharing between S. aureus isolates. Domestic animals can therefore act as melting pots whereby genetic elements from various S. aureus lineages are combined in new genetic backgrounds. A variety of genetic assortments can promote the rapid emergence of high-risk clones with novel phenotypic characteristics. Large-scale genomic changes derived through HGT can generate "hopeful monsters" that may potentially cause public and animal health threats in ways that are hard to predict 48,49 . Because S. aureus can also acquire DNA from other Staphylococcus species with which it shares its niche with 31,50 , its gene pool may be further augmented with DNA that can confer additional adaptive or pre-adaptive features.
There are limitations in our study that need to be acknowledged. We recognize the sampling bias in our dataset that heavily favored isolates from cats, cows and dogs. Moreover, most isolates in our dataset were collected in New Hampshire where NHVDL is located, resulting in comparatively low sample sizes from the surrounding regions. Such bias did not allow us to carry out a more systematic analyses of the host distribution of STs and patterns of gene sharing between animal hosts that included other eukaryotes. The limited number of eukaryotic hosts means that those STs identified as host-restricted may in fact be found in multiple animal species. This also means that certain STs were overrepresented and rare ones were overlooked. Wildlife-associated S. aureus may likely harbor novel genetic variants or mobile genetic elements that pose an unknown level of risk to humans, companion animals and livestock. The structure and gene content of mobile genetic elements, such as pathogenicity islands and phages, and how they shape patterns of gene sharing in animal-associated S. aureus should also be investigated. Future work should therefore include a broader surveillance of S. aureus in other less commonly studied domestic animals and wildlife species, especially those species that often interface with livestock and/or exist at the junction of urban and natural landscapes. It is likely that the known range of species that S. aureus can colonize will expand as studies continue to examine S. aureus in wild animals. Our dataset also included only isolates from disease cases; hence, we do not have data to describe the extent of S. aureus carriage in animals and its contributions to the overall population genetic structure. Comparison of S. aureus in carriage and infections is critical to understanding the genetic basis of pathogenicity and hence, reduce the threats to animal health. Lastly, our study encompassed only three years of bacterial sampling, which was not sufficient to elucidate the long-term evolution of highly virulent and/or multidrug resistant lineages. Future investigations therefore necessitate close monitoring of high-risk clones and the underlying reasons that contribute to their persistence or expansion in the population.
In summary, our findings highlight the role of animals in disseminating resistance and virulence determinants of S. aureus. The remarkable ability of S. aureus as a versatile, multi-host pathogen lies partly on its ability to acquire and disseminate genetic material between lineages and between animal hosts within a short period of time. This study reveals widespread gene sharing between bacterial strains colonizing different animal hosts and highlights the need for routine surveillance to capture the dynamic genetic context of S. aureus.

Methods
Sample collection in New England. The New England S. aureus collection consisted of 133 isolates that were retrospectively sampled from September 2017 through March 2020. Isolates were obtained as culture swabs from routine clinical specimen submissions to the New Hampshire Veterinary Diagnostic Laboratory (NHVDL), New Hampshire, USA. The clinical specimens were received from multiple veterinary practices from the states of Connecticut, New Hampshire, Maine, Massachusetts and Vermont. These states are located in the northeastern part of the country. All isolates were from animals with confirmed clinical infections. No live vertebrates were used in this study; hence, the NHVDL was exempt from the IACUC approval process. Pure isolates were cultured in commercially prepared tryptic soy agar with 10% sheep red blood cells and brain heart infusion broth. Initial species identification was carried out using matrix-assisted laser desorption/ionization time-of- Pan-genome analysis and phylogenetic reconstruction. We used Roary v3.13.0 57 to characterize core genes and accessory genes that make up the pan-genome 57 . To balance the tradeoff between inferring robust phylogenetic relationships versus accounting for assembly errors, we included core genes if they were present in ≥ 99% of the genomes. Nucleotide sequences of each orthologous gene family were aligned using MAFFT v7.475 58 . Aligned core genes were concatenated to generate a core genome alignment. Phylogenetically informative single nucleotide polymorphisms (SNPs) in the core genome alignment were extracted using SNP-sites 59 . We used the core SNP alignment to construct a maximum likelihood phylogenetic tree using RAxML v8.2.12 60 employing a general time-reversible nucleotide substitution model 61 and four gamma categories for rate heterogeneity. Phylogenetic trees were visualized using the online platform Interactive Tree of Life (IToL) 62 .
In silico sequence typing, detection of resistance genes, virulence genes and SCCmec. Using the contig files, we determined the multilocus sequence types (ST) for all genomes used in this study using the program mlst v2. 19.0 (https:// github. com/ tseem ann/ mlst). STs pertain to allelic profiles that characterize nucleotide differences in partial sequences of seven housekeeping genes 63 . In S. aureus, these seven genes consist of arcC, aroE, glpF, gmk, pta, tpi and yqiL 64 . Allelic profiles of the genomes used in this study were compared to those in the S. aureus MLST database (https:// pubml st. org/) 65 . We screened for the presence of horizontally acquired antibiotic resistance genes and virulence factors using ABRicate v1.0.1 (https:// github. com/ tseem ann/ abric ate) utilizing the Comprehensive Antibiotic Resistance Database (CARD) 17 and the Virulence Factor Database (VFDB) 66 . Finally, we used SCCmecFinder 67 implemented in SCCion v0.1 (https:// github. com/ estei nig/ sccion) to determine the presence and type of the mobile genetic element SCCmec. We used the minimum thresholds of > 60% for sequence coverage and > 90% sequence identity to identify the SCCmec. Visualization of the distribution of resistance and virulence genes was carried out using Circos 68 . We used the default parameters for each program unless indicated otherwise.
Statistical tests. We used Welch's t-test to compare the number of resistance and virulence genes of isolates from different animal hosts. Results were considered significant when p < 0.05.