Analysis of Virulence and Antimicrobial Resistance Gene Carriage in Staphylococcus aureus Infections in Equids Using Whole-Genome Sequencing

ABSTRACT While Staphylococcus aureus is associated with significant morbidity and mortality in equids (horses, donkeys, and mules), few studies have performed whole-genome sequencing to fully categorize large collections of equine isolates. Such sequencing allows for a comprehensive analysis of the genetic lineage and relationships of isolates, as well as the virulence genes present in each, which can be important for understanding the epidemiology of strains and their range of infections. Seventy-two clinical Staphylococcus aureus isolates from equids were collected at the Texas A&M University Veterinary Medical Teaching Hospital between 2007 and 2017. Whole-genome sequencing was performed to characterize the isolates according to sequence typing, biofilm association, antimicrobial resistance, and toxin gene carriage. Of the 72 isolates, 19% were methicillin resistant, of which the majority belonged to clonal complex 8. Eighteen distinct sequence types (STs) were represented, with the most common being ST1, ST133, ST8, and ST97. Most isolates had weak or negative overall biofilm production. Toxin and antimicrobial resistance gene carriage was varied; of note, this study revealed that a large proportion of North American equine isolates carry the leucocidin PQ toxin (66% of isolates). One isolate (17-021) carried genes imparting lincosamide and high-level mupirocin resistance, a combination not previously reported in equine-derived S. aureus isolates. IMPORTANCE This is one of the first studies to perform whole-genome sequencing (WGS) of a large collection of Staphylococcus aureus isolates, both methicillin resistant and susceptible, collected from horses. A large proportion of the isolates carry leucocidin PQ (LukPQ), making this one of the first reports of such carriage in the United States. The presence of lincosamide and high-level mupirocin resistance in a methicillin-susceptible S. aureus (MSSA) isolate highlights the importance of MSSA as a reservoir of important antimicrobial resistance genes. As microbial resistance genes on mobile genetic elements can pass between S. aureus strains and livestock-associated strains can be transferred to humans, these findings have important public health implications.

on the skin, respiratory tract, and genital tract (7). S. aureus can be grouped based on multilocus sequence type (MLST), spa type, and ribosomal MLST (rMLST) with broad groupings organized into clonal complexes (CCs) (15)(16)(17). MRSA can also be further divided into dru and SCCmec types based on differences in the methicillin resistance cassette (18,19). MRSA has important public health implications because it can be transferred between humans and horses, as well as among groups of horses (20,21).
Antimicrobial resistance, often to multiple classes of drugs, is a significant problem in S. aureus, and resistance genes are frequently carried on mobile genetic elements (MGEs) and can be passed among isolates. Trimethoprim-sulfamethoxazole (TMS) drugs and aminoglycosides are commonly used to treat infections in horses as they provide broad spectrum antimicrobial coverage (22,23). The genes conferring highlevel aminoglycoside resistance in S. aureus are aac (69)-Ie/aph (29') and aph (39)-IIIa, and kanamycin-neomycin resistance genes ant (49)-Ia or aadD are carried on plasmids (24). The major chloramphenicol resistance gene cat is carried on the plasmids pC221 and pC223 (25), while TMS drug resistance is mediated by the dfr genes (26). In horses, long-term treatment of staphylococcal eye infections with topical fluoroquinolones is a risk factor for developing resistance to fluroquinolones and other drug classes (27). The tet genes encode efflux pumps mediating tetracycline resistance, while the related multidrug efflux pump encoded by norA provides fluoroquinolone resistance (26). In France, the proportion of equine S. aureus isolates resistant to aminoglycosides, tetracyclines, or sulfonamides has increased from 2016 to 2019, with multidrug-resistant (MDR) strains rising from 26% to 52% (28).
Macrolides such as erythromycin are not commonly used to treat staphylococcal infections, but their widespread use in treating other equine bacterial infections, such as pneumonia attributable to Rhodococcus spp., has provided secondary exposure to S. aureus, resulting in resistance genes (erm, msrA, and mph) becoming more common (23,26). Treatment with lincosamides, such as clindamycin, is typically contraindicated in horses due to the potentially fatal enterocolitis complications that arise from their use, but resistance (due to the lincosamide nucleotidyltransferase lnu genes) is occasionally reported in animal-derived staphylococci and lincosamides are used to treat soft tissue infections in people (29). Mutations of the gene encoding the b subunit of the bacterial RNA polymerase rpoB in MRSA strains can confer rifampin resistance (26,30). Fosfomycin is used to treat MDR strains, but resistance mediated by fosB is an emerging problem and is often observed in equine-derived isolates belonging to CC8 (31,32). Carriage of the purported quaternary ammonium compound resistance genes qacA/B, qacC, or qacJ is also relatively common in S. aureus isolates, particularly of the ST1 lineage (33,34), while high-level mupirocin resistance is mediated by transfer of the plasmid carrying mupA (26). Mupirocin is sometimes used to decolonize the nasal and pharyngeal passages of MRSA-positive people, but its use in equids is limited to topical treatment of pastern folliculitis that is resistant to TMS (26,35).
Toxin genes in staphylococci are often carried on large MGEs known as pathogenicity islands that can be horizontally transferred. In S. aureus, pore-forming toxins include alpha-hemolysin (Hla), leucocidins, and phenol-soluble modulins (36). Important S. aureus leucocidins include Panton-Valentine leucocidin (lukSF-PVL) and leucocidin PQ (lukPQ). The purported equine host-adapted leucocidin is lukPQ and causes death of neutrophils, while PVL is a significant virulence factor in human-associated strains (37). Staphylococcal superantigens cause indiscriminate stimulation of T lymphocytes resulting in a mass release of cytokines that can cause shock and organ failure in the host. Most S. aureus strains carry 5 to 6 superantigen genes, including toxic shock syndrome toxin (TSST-1; or SEF); enterotoxins A to E, G to J, and R to T (SEA, etc.); and the enterotoxin-like (SEL) agents K to Q, U, V, and X (SEL-K, etc.) (36). Staphylococcal superantigen carriage versus development of clinical disease has not been extensively studied in equids; however, tsst-1-positive isolates have been associated with toxic shock syndrome in horses (38). In addition to toxin genes, the bacteriophage immune evasion cluster (IEC) genes (sak, scn, and chp) are often carried with SEA and SEL-P and inhibit host neutrophil chemotaxis, cleave host defense proteins, and inhibit host opsonization (39). Toxin and IEC gene carriage differences have not been well categorized in horse S. aureus isolates.
Biofilms are another important virulence factor in staphylococcal species. They complicate the treatment of infections and can delay healing of chronic wounds in equines (40,41). The intercellular adhesion (ica) locus icaADBC mediates the production of polysaccharide intercellular adhesion (PIA) in staphylococcal species (42). A surface adhesion protein, BAP, has also been associated with the increased ability to form biofilms in bovine S. aureus isolates (43). Genes encoding fibronectin binding (fnbB) and bone sialoprotein-binding (bbp) proteins are additional virulence factors that influence biofilm formation. Few studies have characterized the biofilm-forming capabilities of clinical equine S. aureus isolates.
While whole-genome sequencing (WGS) has been used to retrospectively investigate and track emerging outbreaks of MRSA in human hospitals and communities (44), this technique has not been widely utilized with a large collection of MSSA and MRSA collected from horses. Few studies have examined the complete toxin gene carriage profiles of equine MRSA and MSSA strains, instead focusing on carriage of only a few toxin genes. In this study, WGS was performed on a convenience sample of 72 S. aureus isolates collected from equines that presented to the Texas A&M University Veterinary Medical Teaching Hospital over a 10-year period in order to investigate the relationship between strain type, toxin gene carriage, and antibiotic resistance.

RESULTS
Clinical and postmortem findings. The most common clinical findings in the equids were subcutaneous abscesses/cellulitis or dermatitis (n = 17), osteomyelitis and/or joint infections (n = 15), pleuropneumonia (n = 6), and corneal ulcers (n = 5). The majority of the dermatitis cases were lesions located on the limbs (n = 10). In the 22 cases where cytologic examination of fluid or tissue aspirates was performed, neutrophilic to suppurative inflammation was observed in 19 cases, and of those, 10 cases involved coccoid bacteria. In 65% (15/23) of the cases, the death or reason for euthanasia could be attributed to bacterial infections in which S. aureus was cultured. The most common gross necropsy findings included bronchopneumonia (n = 6), chronic joint infections (n = 5), wounds or skin abscesses (n = 5), corneal ulcers (n = 2), a nasal mass (n = 1), and abdominal abscesses (n = 1). One case involved death from a severe, secondary bacterial infection at the sites of previous erythema multiforme. Histologic findings included necrosuppurative bronchopneumonia, synovitis, and skin or abdominal abscesses.
Assembly statistics. The average coverage achieved over the combined Illumina MiSeq runs for all isolates was 64-fold (range, 41-to 90-fold), with removal of 1 outlier of 174-fold coverage. Nine isolates had coverage estimates lower than the benchmark goal of $50-fold. While the overall genome coverage was high, coverage of some areas containing highly repetitive regions, such as the spa gene, the region flanking egc and seh, and enterotoxin gene clusters, was low in 10 isolates (09-047, 26-006, 28-062, 33-029, 36-009, 38-086, 49-063, 54-075, 58-006, and 63-019), resulting in fragmentation of genes. In these instances, either Sanger sequencing (Eton Biosciences, San Diego, CA) or conversion and BLAST query of the reads for each isolate was performed to accurately assess type and gene carriage.
Typing characteristics. The multilocus sequence type (MLST) and spa type combinations observed are illustrated in Fig. 1. The WGS spa types of 10 isolates did not match the Sanger sequencing, which is not uncommon in the tandem repeat regions; consequently, the Sanger spa types were used. Eighteen distinct ST were represented, with ST1 (n = 18), ST133 (n = 11), ST8 (n = 9), and ST97 (n = 9) most frequently encountered. Three new MLSTs were identified due to single, unique point mutations in the aroE (4214, 4215) and arcC (4277) genes. Of the 14 mecA-positive isolates, 64% were ST8, 14% ST612, 14% ST398, and 7% ST5. There were 2 distinct groups of ST8 MRSA, namely, 5 isolates characterized by SCCmec type IVa, spa type t008, and dru type dt9g; and 4 isolates having SCCmec IVd, spa t064, and dru dt10a. The 2 horses with travel history outside the United States cultured ST1-t127 MSSA, a major ST circulating within human communities in Canada and the northern United States and one of the predominate equine MSSA strains found in Denmark (11). New spa types included t17107, t17108, t17109, t17110, t17118, and t17146.
Biofilm production. Eight isolates were positive for biofilm production via the crystal violet microtiter plate assay. Of the 8 positives, only 1 was MRSA and the rest were MSSA.
Biofilm-associated gene carriage. The intercellular adhesion genes icaA, icaB, icaC, and icaD were present in all isolates. None of the isolates carried the bap gene. All of the ST8, ST612, ST816, and ST1181 isolates were positive for fnbB (18/72, 25%). Ten isolates were positive for bbp. Carriage of biofilm-associated genes did not correlate with strain types or the ability to produce biofilms in vitro on crystal violet assay (CVA).
Toxin and virulence gene carriage. Toxin and virulence gene carriage was diverse and was correlated with typing characteristics (Fig. 3). All isolates carried aureolysin and the gamma-hemolysin A, B, and C components. Alpha-hemolysin was carried by a majority (92%, 66/72) of the isolates, as well as b-hemolysin (97%; 70/72). None of the isolates carried eta, etb, etd, or lukMF'. Panton-Valentine leucocidin was found in only five ST8-IVa-t008-dt9g isolates, which also exclusively carried the arginine catabolic mobile element (ACME). Most of the isolates carried lukPQ (66%; 48/72), and carriage was associated with the presence of the strain 3711 prophage (99% identity to reference LT671578.1). All of the lukPQ-carrying isolates were MSSA. Carriage of lukAB was also common (97%, 70/72), and all of these isolates also carried lukPQ. None of the isolates carried phenol-soluble modulin genes.
Other virulence genes included the agr, ACME, and IEC. The most common agr type was agr type I (57%, 41/72), followed by agr type III (26%, 19/72), type II (15%, 11/72), and type IV (1%; 1/72). This is similar to the prevalence seen in a study of S. aureus isolates collected from donkeys in Tunisia, where the most common agr types were also types I and III (45). ACME was seen in all of the ST8 isolates. The IEC genes were carried in 11 isolates, with the following ST pairings observed: A in ST30 (n = 1); B in ST8 (n = 5), ST5 (n = 2), ST398 (n = 2), ST188 (n = 1), and ST45 (n = 1); C in ST582 (n = 1); D in ST8 (n = 1); and G in ST1181 (n = 1). Types B and D are the most common IEC types seen in isolates of agr type I, while type A is commonly seen in isolates of agr type III (21).
Antimicrobial resistance gene carriage. Most of the isolates were MSSA (81%), while only 14 were MRSA. Methicillin-resistant strains were collected from two of the horses that died, but only one was infected with a strain that directly contributed to death. Of the 9 surgical site infections, 3 were attributed to MRSA strains, namely, ST8, ST398, and ST612. MSSA associated with surgical site infections included ST1 (n = 2), ST6, ST97, ST398, and ST816. ST97, ST398, and ST816 are associated with livestock S. aureus infections. A full minimum spanning tree based on MLST was drawn for the mecA-positive and -negative isolates (Fig. 4), and a summary of 14 mecA-positive isolates is given in Table 1. Thirteen of the methicillin-resistant isolates were phenotypically resistant to oxacillin and carried mecA, while one isolate (60-070) was phenotypically oxacillin susceptible but was mecA positive. Isolate 60-070 has a T-to-C missense mutation in the ribose-phosphate pyrophosphokinase (prs) gene that results in a serine-to-proline substitution at residue 292. None of the other mecA-positive isolates had prs mutations. Mutations in prs have been associated with phenotypic susceptibility to  b-lactam drugs in mecA-positive, oxacillin-susceptible S. aureus (48), and 60-070 did not have mutations in other correlated loci.
Antimicrobial resistance gene carriage of the isolates sequenced is summarized in Fig. 5. The b-lactamase-encoding blaZ gene was present in 44% (32/72) of the isolates. Of the 27 penicillin-resistant isolates identified by MIC assay and nitrocefin test, 24 carried blaZ. One isolate carried blaZ but was susceptible to penicillin. All isolates had vancomycin MICs of #1 mg/ml by the broth microtiter dilution method. One isolate (17-021) was resistant to mupirocin due to carriage of a mupA plasmid. Aminoglycoside resistance genes were carried by 40% (29/72) of the isolates. All of the isolates that showed gentamicin resistance with MIC testing carried aminoglycoside resistance genes. Seven isolates carried the chloramphenicol acetyltransferase gene encoded on the pC221 plasmid, and all of the chloramphenicol-resistant (n = 5) isolates determined by MIC carried this plasmid.

DISCUSSION
The overall prevalence of MRSA in this collection was 19%, and most of the equine MRSA cases were attributed to the USA 300 clone (ST8-IV-t008; PVL positive) (49), a clone implicated in the majority of outbreaks of community-associated MRSA in the United States, as well as the Canadian USA500 clone (ST8-IV-t064; PVL-negative). This high percentage of ST8-associated MRSA is similar to veterinary hospital cases from the northeastern United States, Ohio, and Canada attributed to USA500 as well as infections observed at French stud farms (8,(50)(51)(52). However, while previous studies with European horses found the porcine-associated ST398 to be a predominant clone (53)(54)(55), only two of the MRSA isolates in this study were ST398. The most common clonal complex in our study CC8 is one of the most frequently identified and associated with MRSA in horses (32). CC133 and CC97 groups are associated with livestock (45,46), while CC15 has occasionally been cultured from donkeys in Tunisia (45). Strains belonging to CC5 (USA100), CC30 (USA200), and CC45 (USA600) are often associated with human infections and are the most common CC associated with blood infections and endocarditis in people (45). Additionally, CC30 is the predominant cause of mucosal infections in people, and CC45 is a common colonizer of human skin in the United States and northern Europe (45). Only 11% of isolates were found to produce biofilm via the crystal violet microtiter plate assay. This result is different than what is generally seen in S. aureus isolates, although no studies have addressed biofilm production of horse clinical isolates; in human clinical isolates, overall biofilm production under in vitro conditions has been found to vary from 50% to 70% (56,57). A 2009 study of 228 clinical S. aureus isolates found all strains were positive for crystal violet biofilm production, with MRSA producing stronger biofilms (58). The presence of icaADBC genes is observed in the majority of isolates presented here, despite the low biofilm production. Due to the complex nature and multitude of factors mediating biofilm formation (e.g., glucose, salt, osmotic pressure, pretreatment of culture plate wells, aerobic conditions, environmental DNA (eDNA), proteins, and expression of modulating genes such as sarA and rbf), ica-containing species may fail to produce biofilms under testing conditions-a challenge highlighting the importance of considering both genotypic and phenotypic testing methods for biofilm formation (42,59,60). These discrepancies highlight the complex nature of biofilm testing and biofilm regulation (61). While carriage of biofilm genes such as icaA and icaD is high across S. aureus, clinical isolates have demonstrated variable biofilm-producing capabilities under in vitro testing conditions (19,56,58). It is possible that the in vitro conditions of these clinical isolates did not capture their full biofilm-forming capabilities. Additionally, biofilm production has been found to be stronger in MRSA strains than that in MSSA strains (62), and the majority of isolates in this collection were MSSA.
For toxin carriage, of interest were the lukPQ-positive isolates, as them confirm lukPQ carriage in S. aureus isolates from the United States. All isolates carrying the lukPQ leucocidin genes were methicillin susceptible, while the CC traditionally associated with human infections (such as CC8, CC15, and CC30 [47]) were lukPQ negative. None of the PVL-positive isolates also carried lukPQ. In the first paper describing lukPQ (37), the genes were found in 15% of the 87 isolates tested from the Netherlands, Italy, and Portugal; the genes were determined to be carried in ST1, ST133, ST398, and ST1660 isolates from horses. The authors did not mention the methicillin resistance status of the isolates they examined in the study; however, lukPQ was not seen in the subset of isolates sourced from a study by The Ohio State University (51) that consisted of 26 MRSA isolates. Consequently, most of the isolates used in the initial lukPQ prevalence study were likely also MRSA. A subsequent study including MSSA from healthy horses at slaughter in Spain and a WGS study of borderline oxacillin-resistant S. aureus in Germany also demonstrated a high prevalence of lukPQ carriage in ST1 and ST1660 isolates (63,64). As we did not observe any lukPQ-positive MRSA isolates in our collection, the higher prevalence of lukPQ in our study of U.S. isolates is likely due to including MSSA or due to potential differences between U.S. and European MRSA isolates. Enterotoxin gene carriage was common in this collection, although none of the ST398 isolates carried toxin genes, similar to other prevalence reports in Europe that evaluated enterotoxin gene carriage (65). Carriage of sel-x was most common, followed by seh1sel-x in the ST1 isolates. All of the ST816 isolates carried a prophage that encoded sea, sec2, and sel-l and a novel form of tsst-1. ST816-t124 has been isolated from the nasal passages of healthy horses at slaughter in Spain (63).
Most of the MSSA isolates in this collection were resistant to ampicillin and penicillin. Resistance to ceftiofur (45%, 21/47), tetracycline (34%, 16/47), and gentamicin (31%, 15/49) was common. Antimicrobial resistance genes norA, fosB, blaZ, aac6-aph2, and aadD were common in the collection. The high level of phenotypic resistance to penicillin and tetracycline in the isolates is similar to previous reports in S. aureus from horses in Canada and Europe (8,66). However, the percentage of gentamicin-, TMS-, and rifampin-resistant isolates in the collection is higher than the zero resistance prevalence observed in one report from Canada but lower than reports from French stud farms and a Hungarian clinic (8,53,66). Macrolide resistance was seen in 10 isolates. Macrolides, while contraindicated in adult animals, are primarily used for treating rhodococcal infections in foals. None of the horses that cultured macrolide-resistant isolates were being treated with a macrolide antimicrobial drug or mupirocin while in the hospital. As most isolates in this study came from adults with an unknown prior treatment history for rhodococcal infections, it is not surprising that there is some degree of macrolide resistance in this population.
Of concern is the finding of an ST5-t002 MSSA isolate with dual carriage of lnuA and mupA in an animal that had been treated only with penicillin and gentamicin. This combination of resistance genes is rare in S. aureus and to the authors' knowledge has not been reported in an equine isolate. The profile is unusual because mupirocin and lincosamides are not routinely used in horses and suggests that there was transfer from a person to the horse. As previously mentioned, only 19% of the collection were MRSA isolates; the high association of MSSA with the majority of these equine infections, with most cultured in significant amounts and confirmed with cytologic or histopathologic findings, indicates that MSSA can be pathogenic in horses. Additionally, the diverse array of antimicrobial resistance and toxin genes in these equine MSSA isolates could contribute to patient morbidity as well as serve as reservoirs for the transfer of virulence genes between staphylococci of similar lineage.

MATERIALS AND METHODS
Bacterial isolates and demographics. The collection consisted of a convenience sample of 72 S. aureus isolates collected between 2007 and 2017 from 65 horses, 6 donkeys, and 1 mule that presented to the Texas A&M University Veterinary Medical Teaching Hospital. One environmental S. aureus isolate was also included, which was collected in 2008 during a hospital procedure in a large animal ward. All patient isolates were residual diagnostic specimens with written owner consent to be retained for research purposes. Fourteen breeds of horse were represented as isolate hosts, with American Quarter horse (n = 29), Thoroughbred (n = 8), and American Paint (n = 7) horses being the most common. Three of the donkeys were of unspecified breed, 2 were miniature, and 1 was an American mammoth donkey. The average age of the equids was 8.4 years 6 7.6 years (range, 1 day to 31 years). Most isolates were cultured from mares (n = 36) and geldings (n = 25), with 10 collected from stallions and 1 recovered from the lung of an aborted male, full-term Clydesdale fetus. The most common culture site was the skin (n = 28), respiratory secretions or tissues (n = 15), bones or joint tissues (n = 15), blood (n = 6), and the cornea (n = 5). One isolate was collected from urine and 1 from lymph nodes. Two horses had documented recent travel history from Canada and France. Two horses were cultured as part of a breeding soundness exam. Nine of the cases involved infection of surgical sites, with two involving infection of joint implants and seven involving skin infections related to incisional dehiscence. Cytologic examination of fluid or tissue aspirates was performed by a clinical pathologist in 22 of the cases involving joint (n = 8), eye (n = 5), respiratory (n = 5), abdominal (n = 2), skin (n = 1), and urinary (n = 1) infections. Of the 72 equids, 23 died or were euthanized within 3 weeks of discharge from the hospital, and 21 of those received a postmortem examination.
Phenotypical characterization of biofilm production on crystal violet microtiter plate assay. Overall biofilm production was assessed by the ability of S. aureus isolates to adhere to a 96-well microtiter plate as previously described (67). The crystal violet assay (CVA) was performed based on the referenced protocol, with the following changes: cultures were diluted 1:200 in tryptic soy broth (TSB) with 1% glucose in flat-bottomed, tissue culture-treated (TC) polystyrene microtiter plates (Falcon 96-well TC plates; Corning, USA). The wells were then carefully rinsed and left to air dry, and the dried biofilms were stained with 0.1% crystal violet dye. The optical density at 570 nm (OD 570 ) values of each plate were adjusted by subtracting the average of the blank control wells that contained only TSB broth. Biofilm production cutoff values were established by the average negative-control value and two standard deviations of the negative control [Neg 1 2 * (std dev)] per plate. Values below the cutoff were considered negative for biofilm production and those higher were considered positive. Microtiter plate assays were performed in technical and biological duplicate using a symmetrically inverted loading pattern to control for edge effects (68).
DNA extraction and library preparation. DNA was extracted from individual bacterial pellets with the DNeasy blood and tissue kit (Qiagen, Germantown, MD, USA) according to the manufacturer's recommendations for Gram-positive bacteria, except 1 ml of a solution of 5 mg/ml lysostaphin (L7386; Sigma-Aldrich) was added to the lysis buffer per isolate. DNA was quantified via Life Technologies Qubit high sensitivity double-stranded DNA (dsDNA) assay, and all samples were normalized to 1 ng total DNA for library preparation. Sequencing libraries were prepared using the Nextera XT library prep kit (Illumina, San Diego, CA) per their standard protocol. The Qubit high-sensitivity (HS) dsDNA assay was used to determine the concentration and the Agilent TapeStation D1000 HS system was used to determine the average fragment size of the prepared libraries. All 72 samples were normalized to 4 nM, pooled for two independent runs, and sequenced on the Illumina MiSeq system with the 300 by 300 cycle v3 sequencing kit. All run data and FASTQ files were uploaded to BaseSpace (Illumina) for downstream analysis.
The online sequencing pipeline offered by the Pathosystems Resource Integration Center (PATRIC version 3.4.6; https://www.patricbrc.org) was used to assemble and annotate the S. aureus genomes (69) for analysis. Illumina MiSeq paired read libraries for each isolate were uploaded to PATRIC and de novo assembled using the recommended MiSeq assembly strategy parameter (69). Genome coverage of the combined MiSeq runs was estimated by comparing the total nucleotide bases per isolate to the 2.82-Mb S. aureus subsp. aureus NCTC 8325 reference genome in GenBank (NC_007795.1). The annotation was performed using the Rapid Annotation using Subsystem Technology tool kit (RASTtk) on the PATRIC pipeline (70).
Strain typing and eBURST analyses. In order to account for the propensity for alignment errors during WGS in the tandem repeat regions of the S. aureus spa gene, spa sequences were amplified via PCR as previously described (15), and the forward and reverse strands were Sanger sequenced (Eton Biosciences, San Diego, CA) to generate a consensus sequence. The tool spaTyper (http://spatyper .fortinbras.us) was used to determine the spa type for each isolate and compared with the sequences in the WGS assemblies; new spa types were submitted to the Ridom SpaServer (http://www.spaserver .ridom.de) for inclusion in the database.
The MRSA isolates were screened via PCR for amplification of the dru segment (18), and dru type was assigned by querying the mecA cassette sequences for each isolate WGS against the online dru repeat and typing database (http://dru-typing.org/site/; curator, Richard V. Goering). The SCCmec type was assigned via BLAST query for cassettes I to VI (19). Minimum spanning trees were generated using PHYLOViZ (73), and the full goeBURST plugin was used to determine clonal complexes in the MLST database (as of March 2020).
Toxin, antimicrobial resistance, and virulence gene queries. The assembly files and annotated genomes were queried using the standalone BLAST1 suite version 2.6.0 and BLAST nucleotide and protein interface in the PATRIC database, respectively. The bacterial antimicrobial resistance gene database Antibiotic Resistance Gene-ANNOTation (ARG-ANNOT) and the staphylococcal VirulenceFinder 1.5 database files from the Center for Genomic Epidemiology (March 2017 versions) were used to make the database for the BLAST1 queries (74,75). Nucleotide sequences in GenBank for lukPQ (LT671578.1), qacA/B (GU565967.1), qacC (M37889.1), phenol soluble modulins 1 to 4 (BK006301.1), and chp (AF285146.1) were also added to the database. A positive hit for a gene was defined as a query having at least 95% identity with and covering at least 90% of the length of the database gene reference. The agr and IEC types were assigned based on a BLAST query using primer sequences from the previously described PCR typing methods (39, 76). The UniProt protein sequences for the ica genes (Q9RQP9, Q9RQP7, Q9RQP6, and Q9RQP8), fnbB (A0A0H2XKG3), bap (Q79LN3), and bbp (Q14U76) were queried using the PATRIC BLAST feature. Due to nucleotide variation, a positive hit for a protein was defined as a query having at least 95% identity with and covering at least 90% of the length of the UniProt reference. To ensure complete coverage of genes in repetitive areas, the raw MiSeq FASTQ files were converted to FASTA files with fastq2fasta (TM Software, Inc., Arcadia, CA) and requeried as above; a cutoff of .90% identity to a known toxin or virulence proteins cataloged in Uniprot and GenBank was used to identify the genes.
Data availability. All assemblies are freely available in the public workspace of PATRIC under the name "Equine S. aureus isolates" and in the NCBI GenBank database under BioProject PRJNA604020 and accessions JAAFKO000000000 to JAAFNH000000000. Illumina reads were deposited under accessions SRR14923786 to SRR14923857.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. DATA SET S1, XLSX file, 0.1 MB.