Molecular Characterization of Equine Staphylococcus aureus Isolates Exhibiting Reduced Oxacillin Susceptibility

The detection of borderline oxacillin-resistant Staphylococcus aureus (BORSA) represents a challenge to both, veterinary and human laboratories. Between 2015 and 2017, 19 equine S. aureus with elevated minimal inhibitory concentrations for oxacillin were detected in routine diagnostics. The aim of this study was to characterize these isolates to identify factors possibly associated with the BORSA phenotype. All S. aureus were subjected to antimicrobial susceptibility testing and whole genome sequencing (WGS). A quantifiable β-lactamase activity assay was performed for a representative subset of 13 isolates. The WGS data analysis of the 19 BORSA isolates identified two different genomic lineages, sequence type (ST) 1 and ST1660. The core genome multilocus sequence typing (cgMLST) revealed a close relatedness of all isolates belonging to either ST1 or ST1660. The WGS analysis identified the resistance genes aadD, dfrG, tet(L), and/or blaZ and aacA-aphD. Phenotypic resistance to penicillins, aminoglycosides, tetracyclines, fluoroquinolones and sulfamethoxazole/trimethoprim was observed in the respective isolates. For the penicillin-binding proteins 1–4, amino acid substitutions were predicted using WGS data. Since neither transglycosylase nor transpeptidase domains were affected, these alterations might not explain the BORSA phenotype. Moreover, β-lactamase activity was found to be associated with an inducible blaZ gene. Lineage-specific differences regarding the expression profiles were noted.


Introduction
Staphylococcus aureus can be a harmless commensal residing on the skin and mucosal surfaces of healthy animals and humans [1], but it can also cause a broad spectrum of diseases, ranging from relatively mild skin infections to life-threatening pneumonia, sepsis and endocarditis [2,3]. The infections caused by antimicrobial-resistant staphylococci are a major threat in human and veterinary medicine, since treatment options are often limited [3][4][5]. A broad range of different staphylococcal species has been detected in horses, including S. aureus and Staphylococcus delphini as coagulase-positive species, but also a variety of coagulase-negative staphylococci [6][7][8][9][10]. Several studies on equine S. aureus and, in particular, equine methicillin-resistant S. aureus (MRSA) have been performed in recent years [11][12][13][14][15].
Besides MRSA, also borderline oxacillin-resistant S. aureus (BORSA) gained particular attention [16]. Commonly, BORSA were initially defined by their reduced susceptibility to oxacillin with minimal inhibitory concentrations (MICs) of 1-2 mg/L, but some BORSA strains can also exhibit oxacillin MICs of 4-16 mg/L [16]. These isolates lack an additional penicillin-binding protein (PBP), encoded by either mecA or mecC, which are located on mobile genetic elements, the Staphylococcal Cassette Chromosome mec (SCCmec) in MRSA [1,16] or mecB carried on a plasmid, as recently described [17].
Previous studies showed an increase of both, MRSA and BORSA in equine clinics [20,21]. Moreover, there has been evidence that, once resistant staphylococcal isolates enter a human or veterinary clinic [11,34,35], the number of nosocomial infections-in particular wound infections [34,35]-increases. Nosocomial settings provide a selective pressure to bacterial pathogens since antimicrobial agents and also biocides are widely used [36,37]. While horse clinics have been identified as "hot spots" for pathogens accumulating antibiotic resistance in the recent past, little is known about frequency and dimension of biocide resistance. Consequently, more research is needed in this particular field and biocide susceptibility of field isolates should be monitored. However, studies determining MICs for biocides often lack an approved method and are therefore difficult to compare [37].
The aim of this study was to thoroughly characterize S. aureus of equine origin with elevated oxacillin MICs ranging from 0.5 mg/L to ≥4 mg/L via VITEK2 while lacking known methicillin resistance genes, such as mecA and mecC.

Phylogenetic Analysis
The core genome multilocus sequence typing (cgMLST) analysis revealed 14 different allelic profiles and two different clusters for the 19 BORSA isolates (Figure 2). Cluster 1 represented the ST1 sequences and consisted of two closely related allelic profiles differing only in one target gene. Cluster 2 included the ST1660 isolates assigned to 12 distinct, but also closely related allelic profiles differing in 0 to 7 target genes only. The phylogenetic relationship of the two clusters was considerably distant, mirrored by the differences in 1697 of 1744 alleles included in the comparison.

Phylogenetic Analysis
The core genome multilocus sequence typing (cgMLST) analysis revealed 14 different allelic profiles and two different clusters for the 19 BORSA isolates (Figure 2). Cluster 1 represented the ST1 sequences and consisted of two closely related allelic profiles differing only in one target gene. Cluster 2 included the ST1660 isolates assigned to 12 distinct, but also closely related allelic profiles differing in 0 to 7 target genes only. The phylogenetic relationship of the two clusters was considerably distant, mirrored by the differences in 1697 of 1744 alleles included in the comparison.

Phylogenetic Analysis
The core genome multilocus sequence typing (cgMLST) analysis revealed 14 different allelic profiles and two different clusters for the 19 BORSA isolates (Figure 2). Cluster 1 represented the ST1 sequences and consisted of two closely related allelic profiles differing only in one target gene. Cluster 2 included the ST1660 isolates assigned to 12 distinct, but also closely related allelic profiles differing in 0 to 7 target genes only. The phylogenetic relationship of the two clusters was considerably distant, mirrored by the differences in 1697 of 1744 alleles included in the comparison.  isolates based on a core genome multilocus sequence typing (cgMLST) analysis including 1744 genes using the SeqSphere+ software. Each circle represents an allelic profile and the connecting lines display the number of different alleles between the distinct profiles. The individual isolate IDs are shown within the circles, the collection date is given in brackets, while the sequence types (ST) types are indicated by color: ST1 in light blue, ST1660 in red.

IMT39129 -
All isolates harbored a similar set of genes associated with adherence, including ebpS (encoding the elastin binding protein) and efb (encoding the extracellular fibrinogen binding protein). All isolates, except isolates IMT37341, IMT39841 and IMT41899, were positive for the fibronectin binding protein gene fnbA. Isolate IMT41899 was also negative for the gene fnbB. All isolates harbored the intercellular adhesion gene cluster (ica), which is associated with biofilm formation.
The ST1660 isolates carried genes for staphylococcal enterotoxins (SE) and SE-like toxins associated with the egc enterotoxin gene cluster, namely sei, selm, seln, and selo. Moreover, pseudogene φent2 was present in all but one [IMT39841] and selq was present among all ST1660 isolates, while the ST1 isolates harbored the enterotoxin gene seh. All isolates harbored the leukotoxin-encoding genes lukD and lukE. All isolates were positive for the bacteriophage Saeq1 (acc. no. LT671578) harboring genes (lukP and lukQ) which code for a further leukocidin (Table 1). For the staphylococcal complement inhibitor (Scin)-also encoded on this phage-96% amino acid (aa) identity compared to the reference, bacteriophage Saeq1, was observed. All isolates were negative for the genes encoding the toxic shock syndrome toxin 1 (tst) and the Panton-Valentin leucocidin (PVL) genes lukF-PV and lukS-PV.

Antimicrobial Resistance Properties
Antimicrobial susceptibility testing was performed via broth micro-and macrodilution according to the Clinical and Laboratory Standards Institute (CLSI) recommendations [38,39]. The MIC value distribution and their classification into the categories susceptible, intermediate (if available) or resistant is displayed in Table 2. Penicillin MIC values from 2 to ≥64 mg/L were determined for the 19 isolates reported. Hence, all isolates were classified as resistant [38,39] and harbored the β-lactamase gene blaZ.
Grey shading indicates concentrations not included in the test panel. Isolates with growth throughout the panel have MIC values equal to or larger than the highest concentration tested and are, therefore, displayed in the next higher concentration with grey shading; classification as susceptible, intermediate or resistant, according to CLSI [38,39] is indicated by black vertical bars; species-specific clinical breakpoints for horses were applied for penicillin, enrofloxacin and doxycycline, the remaining breakpoints were adopted from human medicine [38,39]. a amoxicillin and trimethoprim MIC values were used for the combinations amoxicillin/clavulanic acid (2:1) and sulfamethoxazole/trimethoprim (19:1), respectively.
Broth microdilution revealed comparatively low oxacillin MICs of 0.25 or 0.5 mg/L for the ST1 isolates, and, as expected, the WGSs lacked the methicillin resistance genes mecA, mecB or mecC. Resistance to gentamicin and kanamycin, tetracyclines, and the combination sulfamethoxazole/trimethoprim was observed. The respective aminoglycoside resistance genes aacA-aphD and aadD, the tetracycline resistance gene tet(L) as well as the trimethoprim resistance gene dfrG were detected in the WGSs. The enrofloxacin MICs of 0.5 or 1 mg/L classified the respective isolates as resistant and the analysis of the quinolone resistance determining regions (QRDR) of GyrA/GyrB and GrlA/GrlB [40,41] revealed that the isolates displayed the same aa alteration Ser80Tyr in GrlA compared to a wildtype sequence (acc. no. D67075.1).
The ST1660 isolates had comparably higher MICs for oxacillin of 1 or 2 mg/L. Resistance to gentamicin and kanamycin was found in all isolates, which corresponded well to the presence of the aacA-aphD gene. The neomycin MIC values of 0.5 or 1 mg/L were lower than those obtained for the ST1 isolates with MICs of 8 or 16 mg/L, which might be caused by the absence of the aadD gene, which confers kanamycin/neomycin resistance. A single isolate was classified as intermediate to erythromycin. In comparison to the ST1 isolates, the ST1660 isolates displayed lower enrofloxacin MICs of 0.03 to 0.25 mg/L, resulting in the classification of one isolate as intermediate based on its MIC of 0.25 mg/L. However, a single aa exchange Glu434Asp in GrlB was observed in the QRDRs. Six isolates with doxycycline MICs of 0.25 mg/L were classified as intermediate.
The results of the quality control strain S. aureus ATCC ® 29213 were always within the respective QC ranges [38,39].

Investigation of the Reduced Susceptibility to Oxacillin
Agar disk diffusion for penicillin resulted in zone diameters of 8-18 mm, which classified the isolates as resistant [38,39]. The ST1 isolates displayed remarkably larger zone diameters (16-18 mm) than the ST1660 isolates (8-12 mm) ( Table 3). A similar situation was observed for oxacillin, ampicillin, ampicillin-sulbactam and amoxicillin-clavulanic acid, whereas the amoxicillin zone diameters were comparable for the isolates of both STs and measured up to 8 mm (Table 3). Comparing the results for ampicillin, and ampicillin-sulbactam as well as for amoxicillin and amoxicillin-clavulanic acid, increased zone diameters (of ≥5 mm) were detected for the β-lactamase inhibitor containing compounds. The results obtained with the quality control strains S. aureus ATCC ® 25923 and Escherichia coli ATCC ® 35218 were always within the respective QC ranges [38,39]. Regarding the reduced susceptibility to oxacillin, the Fem (factors essential for methicillin resistance) proteins [42] were further investigated, including the aa sequences of FemA, FemB, FemX, FemC and FemD. Compared to the β-lactam susceptible S. aureus reference strain ATCC ® 25923 (acc. no. CP009361.1), the following situation was observed. Within the FemA protein sequence, all isolates displayed the aa alteration Glu234Asp, while the ST1 isolates had an additional difference, Tyr195Phe. Moreover, the ST1 isolates displayed the aa alteration Arg199Ser in the FemB sequence. Within the FemX protein sequence, all isolates had an Asn18His alteration. While all ST1 isolates exhibited only one additional aa exchange (Ile51Val), the ST1660 isolates showed two additional exchanges (Asn155Thr and Thr262Lys) in the FemX protein sequence. There were no differences regarding the aa sequences of FemC and FemD.
All isolates had aa differences in the catabolite control protein A (CcpA), a global transcriptional regulator, shown to be associated with oxacillin resistance in staphylococci [43,44]. The following differences were observed: Lys171Glu and Ser207Gly, and the ST1 isolates showed an additional Ala197Glu exchange.
All ST1660 isolates had two aa differences (Ile456Val and Asp561Glu) in the phosphodiesterase GdpP, for which is known, that mutations can lead to elevated oxacillin MICs [45,46]. The position Ile456Val is located in the so-called desert hedgehog (DHH) domain [47,48].
Being the primary targets of β-lactam antibiotics, the penicillin binding proteins (PBPs) [42] were also comparatively investigated. The PBP aa sequences of the analyzed isolates were compared to the respective ones of the reference strain S. aureus ATCC ® 25923 (acc. no. CP009361.1). Differences could be detected in all investigated PBP aa sequences. All ST1 isolates showed two aa exchanges in PBP1 (Asp118Asn, Val617Met), three in PBP3 (Gly167Arg, Lys504Arg, Asp563Glu) and one in PBP4 (Thr189Ser). The ST1660 isolates displayed a single variation in PBP2 (Val102Met), four differences in PBP3 (Gly167Glu, Ala330Ser, Lys504Arg, Asp563Glu) and one in PBP4 (Glu398Ala). Moreover, all isolates had an elongation of eleven aa at position 717 in the PBP2 aa sequence.
Furthermore, the aa sequences of the additional proteins that have been shown to be involved in methicillin resistance, like the efflux pump regulator MgrA [49] and the multiple peptide resistance factor MprF [50] involved in cell wall synthesis, showed no differences to the respective aa sequences of S. aureus ATCC ® 25923.

Overexpression and Induction of blaZ
The representative isolates (all three ST1 and ten ST1660) were subjected to β-lactamase activity testing using a nitrocefin assay. The induction with ampicillin revealed a 5.05-19.10-fold increase of β-lactamase production for the ST1660 isolates, while the ST1 isolates showed only a moderate increase (3.17-3.40-fold) (Figure 3a). The Mann-Whitney-U-Test revealed a significant difference (U = 0.0; p = 0.007) between the ST1 and the ST1660 isolates. Susceptibility testing after induction resulted, for all but one isolate, in oxacillin MICs that were one to two dilution steps higher than the corresponding values of the uninduced isolates (Figure 3b). The isolate IMT37083 (ST1660) showed no change in its oxacillin MIC of 2 mg/L. For the remaining nine ST1660 isolates tested, oxacillin MICs increased to 4 mg/L, thus they were classified as resistant. Even though the ST1 isolates showed increased oxacillin MICs of 1 mg/L, they were still classified as susceptible according to the CLSI standards [38,39]. The investigation of the β-lactamase operon containing genes encoding the proteins BlaZ, BlaR1 and BlaI revealed the differences between the sequence types. The aa sequences were compared to the originally described sequence of S. aureus transposon Tn552 (acc. no. X52734.1) [51]. For the repressor BlaI, only one aa alteration, Gly21Asp, was observed which was present in all ST1 isolates and the ST1660 isolate IMT37083. The BlaZ proteins of the ST1 isolates had seven aa exchanges (Ser22Pro, Val86Ile, Glu145Gly, Tyr220Cys, Val9Ala, Ala112Glu and Pro217Ser). The aforementioned ST1660 isolate IMT37083 shared the latter three aa exchanges (Val9Ala, Ala112Glu and Pro217Ser) with the three ST1 isolates and harbored the additional aa exchange Lys169Arg. In the BlaR1 sensor protein, all ST1 isolates showed 32 aa exchanges, with 19 of them also occurring in the aforementioned ST1660 isolate IMT37083 (Figure 4). The remaining ST1660 isolates had four aa exchanges in BlaR1 (Ala91Thr, Ser106Cys, Asp447Asn, and Phe491Leu).

Susceptibility to Biocides
Comparative biocide susceptibility testing via broth macrodilution revealed MICs of 0.00006-0.0005% for benzalkonium chloride (BAC), 0.125-0.5% for glutardialdehyde (GLU) and 0.00006-0.00025% for chlorhexidine (CHX). The broth microdilution results were 0.000125-0.0005% (BAC), 0.25-0.5% (GLU) and 0.000125-0.00025% (CHX) ( Table 4). Overall, the results vary between three to four dilution steps in broth macrodilution and between two to three dilution steps in broth microdilution. The biocide MIC values of the ST1 and ST1660 isolates did not show major differences. Only a slight difference could be seen regarding the CHX MIC values. The S. aureus reference strain ATCC ® 6538 was tested for comparative reasons and showed comparable results to previous studies [37]. Table 4. MIC distribution obtained with broth micro-and macrodilution.

Discussion
Between 2015 and 2017, BORSA isolates caused infections in horses of a German equine clinic. These isolates were initially noticed as they showed elevated MICs for oxacillin using VITEK2. However, these MICs were classified as susceptible.
An analysis of whole genome sequencing (WGS) data revealed two lineages being associated with the conspicuous oxacillin phenotypes, ST1-t127 and ST1660-t3043, -t2484 and -t549. ST1 is attributed to CC1, which is known to be a livestock-associated putative pathogen, causing zoonotic infections [18]. The isolates with similar characteristics to our isolates (ST1-t127, ST1660-t549 and ST1660-t3043) have been previously described as MRSA and MSSA in equine samples [6,11,21,25,26] and have also been detected in samples obtained from humans in different European and non-European countries [28][29][30][31][32][33]. In addition, MSSA ST1660 with other spa types, t2484 for S. aureus from a horse in Germany [52] and t3043 for an isolate obtained from a donkey in Tunisia [53] were described. The isolates attributed to ST1-t127 belong to the three most prevalent lineages of MRSA in the Italian pig industry [54,55] and were also detected in cattle in Italy and China, [56,57] and wild boars in Germany [58].
The two clusters obtained by cgMLST were in accordance with the multi locus sequence typing (MLST) and could further differentiate the isolates. Interestingly, the four ST1660-t3043 isolates obtained in 2015 cluster very closely together, showing 0-2 allelic differences only, while the isolates obtained in other years were more distantly related ( Figure 2). The correlation of the cgMLST and the spa types revealed that the 14 isolates with spa type t3043 could be assigned into 11 allelic profiles by cgMLST. However, it should be noted that one allelic profile was shared by two isolates with different spa types, namely t549 and t3043. A former study on the relatedness of S. aureus outbreak isolates using the SeqSphere+ cgMLST approach revealed that genomes with 0 to 8 allelic differences should be considered as related, while those with 9 to 29 allelic differences seemed to be possibly related, and those with 30 or more differences were rated as unrelated [59]. However, a lack of epidemiological metadata concerning possible relatedness of individual cases forbids further speculation here. Moreover, a recent study on equine MRSA obtained from horses directly at hospital admission revealed a very limited number of genomic differences for unrelated equine ST398 MRSA as well [13].
All isolates were negative for PVL and the toxic shock syndrome toxin 1, a T-cell activating superantigen (SAg). The ST1660 isolates carried further genes encoding SAgs, including enterotoxins and enterotoxin-like proteins, which are beyond others associated with egc. However, only non-egc encoded SAgs have been implicated in toxin-mediated diseases [60]. Thus, the role of egc-encoded SAgs in equine S. aureus requires further investigation. At present, there is an ongoing discussion about the impact of enterotoxins on colonization abilities of S. aureus. The ST1 isolates harbored the enterotoxin gene seh, which is usually attributed to CC1 and the corresponding protein is known for its binding affinity to human major histocompatibility complex class II [61]. The ST1 and ST1660 isolates were positive for the leukocidin genes lukP/Q located on a bacteriophage most similar to Saeq1 (acc. no. LT671578) [62]. Very recently, this phage was reported for MRSA-ST398 isolated from horses in the same area [13]. LukPQ preferentially kills equine neutrophils, but it also showed activity towards human neutrophils at high concentrations [62]. Moreover, an important immune-modulating factor, a variant of Scin, is also located on that phage. Previous research indicated a C3-inhibiting activity of eqSCIN in plasma of a much broader range of hosts, including horses, humans, and pigs [63]. According to Monecke et al., most isolates of CC1 harbor a β-hemolysin converting phage, which is supported by our findings [64].
Regarding the antimicrobial resistance properties, the ST1 isolates were resistant to penicillins, aminoglycosides, enrofloxacin, sulfamethoxazole/trimethoprim and tetracyclines and accordingly classified as multiresistant, based on their resistance to three or more classes of antimicrobial agents [65]. Similar resistance profiles were also detected among BORSA isolates of ST1 and ST1660 and MRSA CC398 from horses in equine clinics [13,21]. The ST1660 isolates were resistant to penicillins, and aminoglycosides. It should be mentioned, that except for penicillin, enrofloxacin and doxycycline, no equine-specific clinical breakpoints for the tested antimicrobial agents were available [38].
Regarding the reduced oxacillin susceptibility, the VITEK2 results were compared with the broth microdilution results. The broth microdilution results were generally lower than the results obtained by VITEK2. Moreover, higher MICs were determined for the ST1660 isolates compared to the ST1 isolates.
All isolates lacked the known mec genes, encoding an additional PBP and causing resistance to virtually all β-lactams, except specific anti-MRSA compounds [38]. Moreover, no gene with considerable homology to mec genes could be detected in the wholegenome sequences so far. Therefore, other potential causes were analyzed. An analysis of the PBPs, compared to the susceptible S. aureus ATCC ® 25923 (acc. no. CP009361.1) revealed aa alterations in PBP1, PBP2, PBP3 and PBP4, even though none of these differences were within the functional transglycosylase or transpeptidase domains [1,2,[66][67][68]. Since Morroni et al. compared the PBPs of ceftobiprole-resistant MRSA isolates with the vancomycin-resistant mecA-carrying MRSA Mu50 [69], a comparison of our isolates with Mu50 (acc. no. NC_002758.2) was performed. Some of the mutations present in Mu50 compared to S. aureus ATCC ® 25923, were also present in our collection, including the eleven aa terminal elongation in PBP2, but also the aa differences Gly167Arg, Lys504Arg and Asp563Glu in PBP3, as well as Thr189Ser (ST1) in PBP4. Since Mu50 is an MRSA isolate with an alternative PBP2a, it cannot be stated whether these aa changes are involved in the generation of elevated oxacillin MICs.
The mutations in the gdpP gene, which encodes a phosphodiesterase that regulates gene expression, have also been described in association with borderline oxacillin resistance [45]. The GdpP protein has two functional domains, GGDEF and DHH. The GGDEF domain contains a diguanylate cyclase, conferring the capacity to synthesize the second nucleotide messenger cyclic di-GMP, and the DHH domain contains the phosphodiesterase characteristic catalytic DHH motif, mediating hydrolysis of cyclic di-AMP, which is, besides others, involved in cell wall homeostasis [47,48]. Only one aa difference of the ST1660 isolates (Ile456Val) is located in the DHH domain. The fact that one aa alteration is located in a functional domain might indicate a possible contribution to the reduced oxacillin susceptibility of the ST1660 isolates. These alterations were only present among the ST1660 isolates and, therefore, did not explain the BORSA phenotype of the ST1 isolates. However, the comparatively lower oxacillin MICs confirmed for the ST1 isolates versus the ST1660 isolates could possibly be in accordance with this finding. A study by Griffiths and O'Neil revealed that neither a mutation Asp418Ala within the DHH domain nor a truncation of GdpP to 370 aa, causing a deletion of this domain, contributed to oxacillin resistance [45]. Therefore, the aa changes in the GdpP protein reported here for ST1660 isolates might have little or no effect on the phenomenon of elevated oxacillin MICs in this study.
In the case of MRSA, some genes, e.g., those of the fem family, are important for methicillin resistance [70]. However, the respective Fem proteins were found in the WGSs of susceptible and resistant S. aureus isolates. In MRSA, the inactivation of these factors results in a Fem-specific reduced resistance to oxacillin in the corresponding strains, ranging from slightly decreased MICs to complete hypersusceptibility to β-lactam antibiotics [42,71]. Thus far, studies only showed a reduction of oxacillin resistance in usually resistant MRSA isolates, when these factors were altered or deleted [42,70,72]. Consequently, the involvement of the Fem alterations in the increase of oxacillin MICs is not likely.
In line with this, the deletion or inactivation of the carbon catabolite protein CcpA was only shown to increase the β-lactam susceptibility in S. aureus, including MRSA, Staphylococcus epidermidis as well as different streptococci [43,44,73,74]. Therefore, the detected aa differences did not seem to play a role in the reduced oxacillin susceptibility.
Among staphylococci, the resistance to penicillins is commonly mediated by the blaZ gene. This is in accordance with this study, since all isolates were classified as penicillin-resistant by broth microdilution and agar disk diffusion and carried the blaZ gene. As expected, the β-lactam compounds containing a β-lactamase inhibitor revealed larger zone diameters than the respective β-lactam compound alone. All isolates showed zone diameter differences of ≥5 mm for the combinations with a β-lactamase inhibitor, indicating the presence of an active β-lactamase. This is in accordance with the results from previous studies. Maalej and colleagues revealed the differences of at least 5 mm for oxacillin zone diameters with and without clavulanic acid [75].
Borderline oxacillin resistance can be caused by the overexpression of the β-lactamase gene blaZ [16]. β-Lactamase hyperproduction was evaluated as an underlying mechanism, since oxacillin is usually no target for β-lactamases. Using a nitrocefin assay for the selected isolates, inducible β-lactamase production was detected in all 13 isolates tested. Interestingly, the ST1660 isolates showed a higher degree of induction, which might explain the higher oxacillin MICs (Figure 3). Subsequent susceptibility testing after induction with ampicillin, revealed that all isolates but one showed higher oxacillin MICs than without induction. While all ST1 isolates were still classified as susceptible, nine of ten tested ST1660 isolates reached oxacillin MICs of 4 mg/L and were classified as resistant. While all other ST1660 isolates displayed almost no differences to the originally described BlaZ-BlaR1-BlaI proteins of transposon Tn552 of S. aureus strain NCTC 9789 (acc. no. X52734.1), ST1660 isolate IMT37083 shared 19 of the 32 aa differences within the respective proteins with the ST1 isolates. Since BlaR1 is the sensor protein for extracellular β-lactam antibiotics [51,[76][77][78], it is possible, that these aa changes result in a comparatively lower expression of blaZ in ST1 isolates and IMT37083 in the presence of β-lactam antibiotics. The observed increase of the oxacillin MICs after induction of all but one isolate (IMT37083) tested points towards borderline oxacillin resistance due to β-lactamase hyperproduction [16,75].
Biocide susceptibility testing was performed for BAC, GLU, and CHX using broth macro-and microdilution as an additional characterization of the isolates [37,79]. In the equine clinic, only BAC is used, as a compound of a floor disinfectant (7.6 g BAC per 100 g disinfectant). The MICs for GLU and CHX of the isolates in this study were below the standard concentrations used and did not differ remarkably from the MICs of the S. aureus reference strain ATCC ® 6538. BAC is often used as an additive and not as a single antibacterial agent. Regarding BAC, the highest MIC of the isolates corresponds to the lowest used concentration of this agent.
In conclusion, in a two-year period, two closely related lineages of S. aureus, causing infections in an equine clinic, were identified. These isolates were attributed to sequence types/clonal complexes (including MRSA) that are commonly isolated from equine samples, but have also a zoonotic potential [21][22][23][24][25][26]. Here, borderline oxacillin resistance seems to be associated with the hyperproduction of the β-lactamase BlaZ. The detection and correct classification of isolates expressing the BORSA phenotype is of major importance since the effectiveness of β-lactams is limited and therapy failure might occur when these isolates cause infections.

Background Information and Bacterial Isolates
During the years 2015-2017, routine diagnostics identified 19 S. aureus isolates with elevated MICs for oxacillin via VITEK2. These isolates originated from 17 equine patients of a veterinary clinic (Table 5). One sample was from an injured horse and one from respiratory disease. All other samples were from surgical site infections. Most of the cases (eleven samples from nine patients) were from orthopedic procedures. Four samples were from colic patients and two from surgeries of the genito-urinary tract. All surgical patients were treated with a combination of amoxicillin and gentamicin, either as single shot therapy before surgery or up to six days after surgery. The injured patient and four with surgical site infections were additionally treated with the combination of sulfamethoxazole/trimethoprim from three up to 13 days. One of these patients had to undergo a second surgery where a single shot dose of amikacin was injected.

Characterization of the Isolates
The DNA extraction for WGS was performed using the QIAamp ® DNA Mini Kit (QIAGEN, Hilden, Germany) with some adaptations for staphylococci. Before starting the official protocol, the cells were mixed with 25 µL lysostaphin solution (0.1 mg/mL) and incubated for 25 min at 37 • C. Then, 75 µL TE buffer and 25 µL proteinase K (0.1 mg/L) were added and the cells were incubated for 25 min at 37 • C. Then, 75 µL PBS and 2 µL RNAse A (2 µg/µL) were added and slightly mixed. After this, the protocol for the kit was followed starting with the addition of AL buffer. The libraries for WGS were prepared using the Nextera XT library preparation kit (Illumina Inc., San Diego, CA, USA) according to the manufacturer's instructions. The 2 × 300 bp paired-end sequencing in 30-fold multiplexes was performed on the Illumina MiSeq platform. The genome sequences were de novo assembled using Newbler (Roche, Basel, Switzerland) and SPAdes v3.12.0 [80]. The nucleotide sequences were analyzed with Geneious v11.1.4 (Biomatters Ltd., Auckland, NewZealand) and annotated with the subsystem technology server (RAST) [81] and Prokka [82] which were compared with BLAST (National Center for Biotechnology Information, Rockville Pike, USA) [83] results. Further investigations were performed by using ResFinder [84] of the Center for Genomic Epidemiology (http://www.genomicepidemiology.org/). The virulence factors were identified by using VFanalyzer of VFDB (http://www.mgc.ac.cn/cgi-bin/VFs/v5/main.cgi?func=VFanalyzer) and checked using Geneious v11.1.4. The associated mobile genetic elements including pathogenicity islands and phages were determined using Geneious v11.1.4 software. The STs were derived from the pubmlst database (https: //pubmlst.org) and spa types were deduced using the Ridom Spa Server (http://www.spa.ridom.de).

Phylogenetic Analysis
The molecular epidemiology was investigated using the previously generated WGSs as FASTA files for the S. aureus core genome multilocus sequence typing (cgMLST) approach. For this, the software SeqSphere+ version 6.0.2 (Ridom GmbH, Münster, Germany) was used [85,86]. To illustrate the clonal relationship between the different isolates, a minimum-spanning tree was built based on a distance matrix of the core genome allelic profiles including 1744 of 1861 possible target genes, using the "pairwise ignoring missing values" option of the software.

Antimicrobial Susceptibility Testing
The routine diagnostics performed antimicrobial susceptibility testing via VITEK2 according to the manufacturer's instructions. Additional susceptibility testing to 31 antimicrobial agents was performed by broth microdilution according to the CLSI recommendations [38,39] using sensititre TM microtiter plates. S. aureus ATCC ® 29213 was used for quality control. The antimicrobial susceptibility testing by broth microdilution was repeated for the β-lactam antibiotics using selected isolates (all three ST1 and ten ST1660 isolates), after the induction with ampicillin [0.25 µg/mL (ST1) or 32 µg/mL (ST1660) ampicillin (Roth ® , Karlsruhe, Germany)], to investigate the effects of induced β-lactamase production. The inoculum for the induction testing was prepared using the growth method, where the isolates were incubated with the respective amount of ampicillin for 4 h in cation-adjusted Mueller-Hinton broth (CAMHB) and then, the bacterial suspension was adjusted to McFarland 0.5. Additionally, the susceptibility to kanamycin (Roth ® , Karlsruhe, Germany) was tested via broth macrodilution [38,39].
Using a nitrocefin-based β-Lactamase Activity Assay Kit (Sigma-Aldrich ® , Munich, Germany), the β-lactamase activity of all ST1 isolates (n = 3) and selected ST1660 isolates (n = 10) were quantitatively investigated. Therefore, the isolates were cultured overnight in brain-heart-infusion (BHI, Oxoid, Wesel, Germany) at 37 • C. The next day, 5 mL BHI with and without 0.25 µg/mL (ST1) or 32 µg/mL (ST1660) ampicillin (Roth ® , Karlsruhe, Germany) were inoculated with 200 µL of the overnight cultures and incubated for 4 h at 37 • C. Following the manufacturer's instructions, the reaction was prepared in duplicate with a sample volume of 20 µL per isolate and the absorbance was measured at a wavelength of 490 nm, every 60 s for one hour. The standard curves were evaluated for every microtiter plate and β-lactamase activity was calculated, according to the manufacturer's instructions.

Biocide Susceptibility Testing
Comparative biocide susceptibility testing was performed using a broth macrodilution method [37] and a broth microdilution method, which has been developed in this research group [79]. S. aureus ATCC ® 6538 was tested for comparative reasons and the results were compared to those of a previous interlaboratory trial [37]. The optical densities were adjusted according to German Veterinary Medical Society (DVG) standards for biocide efficacy testing [87]. For broth microdilution, twofold dilution series were prepared in 100 µL per well on a 96 well plate. The biocide solutions were prepared in deionized water and the inoculum was prepared in tryptic soy broth (TSB). The inoculum was added to a final testing volume of 200 µL and the results were read after incubation for 24 h at 37 • C. The MIC was defined as the first well concentration without visible growth. The tested biocides were BAC (Roth ® , Karlsruhe, Germany), GLU (Roth ® , Karlsruhe, Germany) and CHX (Sigma ® , Munich, Germany). The test ranges were prepared in twofold dilution series; 0.000008-0.004% for BAC, 0.008-4% for GLU and 0.00001-0.0005% for CHX.

Statistical Analysis
The statistical analysis was performed using IBM ® SPSS ® Statistics Version 25. To compare the β-lactamase inducibility of the isolates, non-parametric Mann-Whitney-U-test was performed.