Hospital transmission of borderline oxacillin-resistant Staphylococcus aureus evaluated by whole-genome sequencing

Introduction Staphylococcus aureus is a major cause of hospital infections worldwide. Awareness towards methicillin-resistant S. aureus (MRSA) infections is high but attention towards borderline oxacillin-resistant S. aureus (BORSA) is limited, possibly due to an underestimated clinical relevance, presumption of low incidence and diagnostic limitations. Gap statement BORSA surveillance has not been routinely implemented, and thus consensus with regard to a definition and infection control measures is lacking. Aim Our goals were to investigate the occurrence, molecular characteristics and clinical manifestations of BORSA infections in the hospital setting. Methodology Following an increased incidence in 2016, BORSA cases in 2014/2016 (in our institution) were more specifically evaluated. Medical records were reviewed to investigate epidemiological links, clinical characteristics and outcomes. Resistance and virulence markers were assessed by whole genome sequencing (WGS). Conventional methods: amplified fragment length polymorphism (AFLP) ; multilocus sequence typing (MLST) and multiple locus variable-number tandem repeat analysis (MLVA) were compared with core genome MLST (cgMLST) and whole-genome single nucleotide polymorphism (wgSNP) analysis to confirm genetic clusters. Results From 2009 to 2013, BORSA comprised 0.1 % of all clinical S. aureus strains. In 2016, the incidence was six-fold higher in comparison to the baseline. Whole-genome SNP and cgMLST confirmed two BORSA clusters among patients with dermatological conditions. Patients with BORSA presented with skin infections, and one case developed a severe invasive infection with a fatal outcome. Infection control measures successfully prevented further transmission in both clusters. WGS findings showed that BORSA strains carried multiple resistance and virulence genes with increased pathogenic potential. Conclusion WGS and cgMLST effectively characterized and confirmed BORSA clusters among at-risk patients with clinical manifestations ranging from mild skin infections to life-threatening bacteraemia. Clinical awareness and active monitoring are therefore warranted for the timely implementation of infection control measures to prevent BORSA transmission in high-risk patients.


INTRODUCTION
Staphylococcus aureus is a common bacterial pathogen that causes a wide range of community-acquired and healthcareassociated infections. β-Lactam antibiotics are the treatment of choice as most S. aureus are susceptible to penicillinaseresistant penicillins, including oxacillin and methicillin [1]. The increasing occurrence and transmission of methicillin-resistant S. aureus (MRSA) in hospital settings are of concern due their association with more difficult-to-treat infections and increased mortality [2]. Effective infection prevention and control measures are important to prevent ongoing transmission to vulnerable patients and healthcare workers (HCW) [3,4].
MRSA carry modified penicillin-binding protein (PBP) 2a or 2c with a low affinity to β-lactam antibiotics encoded by mecA or mecC genes [5]. Thus, antibiotic options are severely limited with vancomycin being the first-choice regimen for severe infections requiring parental treatment [6,7]. Unfortunately, vancomycin has a slower bactericidal rate against staphylococci and is inferior when compared to β-lactam treatment of severe S. aureus infections such as bacteraemia and endocarditis [8][9][10][11][12].
Another type of oxacillin-resistant S. aureus is the borderline oxacillin-resistant S. aureus (BORSA) phenotype that contains alternative resistance mechanisms. These strains include β-lactamasehyperproducing S. aureus and S. aureus with point mutations in other PBP genes [13][14][15][16][17][18]. There are no strict laboratory diagnostic criteria to define these resistance phenotypes. Other diagnostic limitations include oxacillin breakpoint guidelines differences (EUCAST, >2 mg l −1 ; CLSI, ≥4 mg l −1 ) and the use of cefoxitin, which fails to detect oxacillin resistance due to production of β-lactamase. The clinical relevance of these resistance mechanisms is of debate, and with a presumption of limited clinical relevance. A lack of a definition and urgency have led to the fact that BORSA infection control measures and surveillance have not been routinely implemented [19]. However, β-lactam antibiotics are not proven to be effective for the treatment of severe BORSA infections, and reports of treatment failures exist [20,21].
In this outbreak analysis, we evaluated the occurrence and clinical manifestations of BORSA infections in the hospital setting, after an increase in incidence in 2016. We performed whole-genome sequencing (WGS) of all isolates for an assessment of the genetic relatedness of the presumed outbreak isolates using two different methods, core genome multilocus sequence typing (cgMLST) and a whole genome single nucleotide polymorphism (wgSNP) analysis.
The primary objective of this study was to elucidate the role and clinical relevance of nosocomial BORSA transmission using molecular typing tools.

Setting and data collection
Leiden University Medical Centre (LUMC) is a tertiary medical centre with an average of 25 000 admissions per year.
The Infection Control department initiated an outbreak investigation in May 2016 following consecutive BORSA infection cases at the Dermatology outpatient clinic. This outbreak analysis includes eight clinical BORSA isolates obtained from patients presenting at LUMC from January 2014 to December 2016 with a BORSA infection. Medical records were reviewed to obtain relevant clinical data concerning the BORSA infection, the infection date, site, treatment and outcome, and to investigate common exposures and an epidemiological link between the cases. All Dermatology HCW and other contacts of BORSA-positive patients were screened for BORSA carriage to investigate transmission.
The study describes routine outbreak investigation procedures initiated by the Infection Control department, and permission from the Ethics board was obtained to process and present this information.

BORSA definition
BORSA isolates were defined as S. aureus with an oxacillin MIC ≥2 mg l −1 as measured with an E-test (bioMérieux) and without the presence of mecA and mecC genes.

Routine diagnostics and antimicrobial susceptibility testing
Clinical cultures were ordered by the attending physician, and processed in the laboratory as per routine. In the case of a clinically relevant culture with S. aureus, susceptibility testing was performed using a Vitek2 Gram-Positive Susceptibility card (AST-P633; bioMérieux), which included susceptibility results for oxacillin, clindamycin, vancomycin, rifampicin and fusidic acid, and a cefoxitin screening.
If the Vitek2 oxacillin MIC value was ≥4 mg l −1 , an E-test was performed using an inoculum of a McFarland 0.5-1.0 suspension in saline and incubated for 24 h on Mueller Hinton agar with 2 % NaCl, in ambient air at 35 °C. To assess the presence of MecA/C genes, a PCR (BDmax StaphSR PCR Assay; BD Diagnostics) was performed on all strains with an oxacillin E-test value of ≥2 mg l −1 .

Contact screening cultures
All dermatology HCW were consequently screened for BORSA carriage when an outbreak was suspected. HCW provided a swab from the throat and nose, and, if applicable, skin defects. Patient contacts were also screened for BORSA carriage and provided a throat, nose and perineum swab [22]. Swabs were incubated in a Brain Heart Infusion broth with 2.5 % NaCl and colistin. The broth was plated on ChromID agar plates (bioMérieux), and on selected colonies, a MecA/C PCR (BDmax StaphSR PCR Assay; BD Diagnostics) was performed after 18-24 h of incubation.
Susceptibility testing was performed as described previously.

Conventional molecular typing methods
Molecular typing was used in the outbreak investigation to assess the relatedness of the BORSA strains.
AFLP (amplification fragment length polymorphism) analysis was performed in our institution using a previously described protocol [23]. AFLP patterns were analysed using BioNumerics software, version 7.1 (Applied Maths NV), and the similarity between normalized AFLP patterns (range 60-600 bp) was calculated with the Pearson productmoment correlation coefficient and UPGMA (unweighted pair group method with arithmetic mean) algorithms. Strains with more than 90 % similarity were considered to be closely related.
Both multiple loci variable-number tandem repeat analysis (MLVA) and MLST was performed in the national reference centre for MRSA, The National Institute for Public Health and the environment [24,25]. MLST is performed using DNA nucleotide sequences of seven housekeeping genes (arcC, aroE, glpF, gmk, pta, tpi, yqi) and analysed using http:// saureus. mlst. net/

WGS methods
We performed WGS of all BORSA strains as part of the outbreak investigation. The genetic relatedness of these strains was analysed using two different methods, cgMLST and a whole genome SNP (wgSNP) analysis. A BORSA For cgMLST analysis, raw sequence data were transferred to a FASTA file and analysed using SeqSphere+ software version 6.0.2 (Ridom) [26]. The number of core-genome targets for S. aureus is 1861 with a Cluster Alert of 24 differences or more [27].
For SNP analysis, we used the Basty pipeline (version 0.9.0) from BIOPET [28] to produce multisample vcf files. Using this multisample vcf file, the SNP distances between strains were calculated using VCFtools (version 0.1.16; with  parameter '--relatedness2') [29]. The distances between strains were visualized with R software version 3.6.1 [30] using the pheatmap package [31] based on the manhattan clustering distance. More details regarding this pipeline can be found in the supplementary material.
To detect mutations in the gdpP, PBP4 and BlaZ genes, we aligned the genes to each strain with the software program Geneious (version 10.2.6) using sequences from blast.
VirulenceFinder and ResFinder web-based tools [32,33] were used to find the virulence-and antimicrobial resistance genes (see Table 1). The Resistance Gene Identifier (RGI), which relies on the Comprehensive Antibiotic Resistance Database (CARD), was used to find antimicrobial resistance genes for all samples [34].

Epidemiology
In 2016, an increased number of BORSA isolates were reported over a period of several months, leading to suspicion that there may have been nosocomial transmission occurring. In this period, five consecutive BORSA patients presented among 750 S. aureus-positive patients, an incidence of 0.66. In 2009-2015 preceding the start of the outbreak investigation, the incidence of BORSA was only 0.1 % with eight BORSA patients among 8345 S. aureuspositive patients identified. This increase in BORSA incidence was statistically significant, with a P-value of <0.001 as calculated with the Pearson chi-squared test.

Outbreak investigation
The Infection Control Unit initiated an outbreak investigation in 2016 following the epidemiological increase of BORSA in early 2016. During this investigation, patients with a BORSApositive culture in 2016 were retrospectively reviewed. This investigation revealed that the outbreak may have started as early as 2014, as epidemiological links with the Dermatology department among BORSA cases from 2014 to 2016 were identified. A total of eight BORSA patients from this period were identified of whom seven had contact with the Dermatology department (Table 2). Six out of eight patients visited the Dermatology Outpatient Clinic between January and May 2016, and one patient visited the Dermatology Outpatient Clinic only once in 2015. Two out of eight patients received light therapy involving twice weekly visits to the clinic, and several times their visits were scheduled on coinciding days. Other epidemiological links included two patients who were admitted to the hospital and were seen by the same consulting Dermatologist, on the same dates but on different wards. Two patients were admitted to the same Internal medicine ward on overlapping days, but did not share a room.
Patient 8 seemed epidemiologically unrelated, had no contact with the dermatology department and carried a strain with a distinct susceptibility pattern; for these reasons, this patient was excluded from further outbreak analysis. Individual phenotypic susceptibility results of the outbreak cluster strains are given in Table 1.
The strains were isolated from a wide range of specimens, and a clinical overview of patients is depicted in Table 2. All patients manifested (recurrent) skin infections, and one case (patient 4) developed a severe invasive infection with a fatal outcome. Several patients received β-lactam antibiotics in the year prior to the first positive BORSA culture.
A common source was suspected, and several infection control measures were initiated at the Dermatology Outpatient Clinic. Contact isolation precautions for positive patients were applied. Hand hygiene compliance and the importance of individualized use of creams and lotions was emphasized. A total of 47 dermatology HCW were screened, but their cultures revealed no BORSA carriage. No new cases were identified during a 6-month follow-up period after the outbreak investigation.

Molecular typing andwhole-genome sequence analysis
Conventional typing methods identified two different clusters by using AFLP, MLVA, and MLST ( Further analysis based on whole-genome sequence result showed successful extraction >97% of cgMLST targets in all strains.
Cluster 2 cgMLST results confirmed that isolates from patients 1 and 3 were related with a genetic distance of only 5 alleles and formed one genetic cluster that belonged to CT3242. SNP analysis showed a slightly larger genetic distance of 113 SNP. The exact moment of transmission between these patients is unknown, however both patients visited the Dermatology outpatient clinic on consecutive days for a routine check up.
The cgMLST results for Cluster 1 isolates from patient 2, 4, 5, 6 and 7 could not confirm one genetic cluster. Isolates 4, 5, 6 and 7 belonged to CT3243 with a genetic distance ranging from 0 to 2 allelic differences and patient 2 belonged to another complex type, CT3248. (Fig. 1). The genetic distance between the complexes was 163 alleles, and the presence of additional resistance genes in CT3243 (ermC) and absence of virulence genes sak, scn and resistance gene BlaZ suggested a high genetic heterogeneity.
Whole-genome SNP analysis showed that the differences ranged from 5 to15 SNP between the strains from patients 4, 5, 6 and 7 as can be seen in Fig. 2.
Patient 8 belonged to CT6016 and was highly unrelated with a genetic distance of 1664 alleles or more.
Whole-genome sequence analysis using CARD and Viru-lenceFinder showed that BORSA isolates from 2014 to 2016 carried multiple virulence factors and resistance genes, as summarized in Table 2. Cluster strains harbored β-lactamase, quinolone, and macrolide resistance genes. The wide range of virulence genes encoded for aureolysins, serine proteases, staphylokinases, leukocidins, b-and g-hemolysins, superantigens, complement inhibitors, and enterotoxins.
Analysis of mutations in the gdpP, PBP4 and blaZ genes reveiled that these genes were identical within each cluster and carried several mutations. Cluster 1 gdpP gene had an insertion at N263I. The cluster 1 PBP4 gene had 6 mutations: F12C, A25T, R101T, S189T, A398E and A409T. The Cluster 1 blaZ gene had a E112A mutation, and belonged to type A.

DISCUSSION
In this study, WGS analysis confirmed hospital transmission events during an increase from the baseline incidence, and our findings highlight that clinically relevant outbreaks can occur among high-risk patients. Skin conditions are a known risk factor for the acquisition and transmission of S. aureus, including MRSA. This is especially the case for atopic dermatitis [35][36][37][38].
Previous BORSA outbreaks have been suggested by others, using conventional molecular techniques, including PFGE, but these methods have a relatively low discriminatory power and cluster confirmation is therefore limited. Our results provided evidence that a strain that seemed related to a cluster by AFLP and MLVA was unrelated using WGS data with 162 allelic cgMLST and 435 wgSNP differences. cgMLST established itself as a reliable and user-friendly tool that can be used for prospective surveillance and comparison between laboratories and differentiate with a higher resolution between relatedness due to transmission events compared to conventional molecular techniques [27,[39][40][41]. In our study, cgMLST confirmed four different genetic lineages and two clusters. wgSNP analysis confirmed the relatedness of the strains and may have the potential to provide even more in-depth data as it takes into account all genetic differences between strains. A cgMLST scheme counts only one allelic change when multiple nucleotide changes within the same gene are observed. Thus, SNP analysis may yield a higher number of differences between strains. This will probably explain why the number of SNP differences between the strains is slightly higher than the number of allelic changes. This is shown in patients 4 and 6, who are typed as identical by cgMLST but have 14 SNP differences. As we do not known the exact sequalea of transmission it is unknown whether the SNP differences are related to the number of events that were needed for the strains to spread from the source. The genetic distance within our clusters ranged from one to five allelic differences, which is within the eight allelic differences criterion defining recent S. aureus transmission as proposed by others [40,42].
BORSA outbreaks have previously been reported among dermatology patients [43,44]. All BORSA patients in our outbreak analysis had a history of dermatological conditions. Infection control investigations at the Dermatology department and outpatient clinic indicated possible epidemiological links. Culturing of HCW revealed no source of transmission and environmental samples could not be obtained due to the time that elapsed. It remains unclear what caused the increase in cases and how transmission could have occurred. Potentially shared creams or lotions could be a point source. However, a diversity of potential epidemiological links was identified, as some patients were admitted to the same ward, were seen by the same physician on different wards or had light therapy on the same day. It is plausible that transmission routes varied among patients, and multiple modes of transmission are of importance when it comes to preventative measures for this specific patient group. Fortunately, early reinforcement of infection control measures halted ongoing nosocomial spread.
Recent surveillance studies suggest a BORSA incidence in hospital settings of approximately 1-5 % [13,[45][46][47]. We describe a lower incidence, but this may be due to underreporting in our study. Diagnostic limitations may include the routine use of automated susceptibility testing methods with no active BORSA screening thresholds, the lack of addition of NaCl in routine cultures to improve the synthesis of β-lactamases and application of cefoxitin susceptibility testing for oxacillin resistance screening.
No exact criteria exist to define a BORSA phenotype, and oxacillin MIC thresholds ranging from 1 to 8 mg l −1 have been suggested. However, most authors classify S. aureus with an MIC of at least 2-4 mg l −1 as borderline resistant for oxacillin [13]. EUCAST clinical breakpoints suggest breakpoints above 2 mg l −1 for oxacillin as suspect for either BORSA or MRSA. This is in line with our findings, where most isolates had an oxacillin MIC of 2-4 mg l −1 . Future studies may focus on optimizing BORSA detection using automated susceptibility testing by defining thresholds for oxacillin that could trigger more accurate MIC testing in selected isolates. It is interesting to note that in one of the patients, one strain belonging to the same genotypic cluster exhibited MICs ranging from 2 to 6 mg l −1 and showed documented treatment failure. These findings suggest a threshold of maximum 4 mg l −1 in automated testing and adopting an MIC of 2 mg l −1 or higher for oxacillin as measured by an E-test to define BORSA.
In the literature, a β-lactam at high dose is suggested as a therapeutic option for the treatment of uncomplicated BORSA infections [20,[48][49][50][51][52]. Our findings indicate a risk of therapeutic failure using high-dose β-lactam for the treatment of severe BORSA infections, including those causing endocarditis. The initial treatment of a case of BORSA endocarditis consisted of 12 g/24 h i.v. flucloxacillin. It was switched to vancomycin after 5 days when susceptibility results showed an oxacillin MIC of 4 mg l −1 , a susceptible cefoxitin screen and absence of MecA/C which excluded MRSA. High-dose β-lactam treatment failure was reported in a previous case of BORSA endocarditis [21]. Both cases demonstrated in vivo development of oxacillin resistance after initial high-dose β-lactam treatment and subsequent recurrence. Caution should be exercised when prescribing β-lactam treatment for BORSA infections.
Molecular characterization showed that all BORSA isolates carried a wide range of resistance and virulence markers. Interestingly, the cluster strains had several mutations in the PBP4 and gdpP genes. Some of those mutations (Y208F, V381F and R430I) have been associated with decreased oxacillin susceptibility by Argudin et al. [53]. Cluster 1 strains carried a mutation in blaZ E112A that very recently has been associated with increased oxacillin resistance [54]. BORSA isolates may be characterized into β-lactamasehyperproducing S. aureus (BORSA), modified S. aureus with point mutations in PBP genes expressing a reduced affinity for β-lactams (MODSA) and other mechanisms [13,45,55]. Evaluation of more genetic markers is an object for future studies [53]. Several questions remain unanswered concerning the emergence and genetic diversity of BORSA in nosocomial and community settings. Further understanding of the molecular epidemiology and resistance mechanisms of BORSA is of clinical relevance [56] and can be accomplished by implementing systematic surveillance programmes and by establishing a local reference database. Further research should focus on identifying the underlying molecular mechanisms for oxacillin resistance in BORSA.
In conclusion, our findings highlight that clinically relevant BORSA outbreaks can occur in a low-incidence hospital setting and that WGS-based analysis can be used for accurate outbreak confirmation. This should be confirmed in studies with higher numbers, as our outbreak included only a low study number to investigate.
Implementing cgMLST and reference databases in infection control practice will aid in defining transmission pathways and improving prevention and control measures to avoid transmission to HCW and vulnerable patients.

Funding information
The authors received no specific grant from any funding agency.