Introduction

Vancomycin-resistant Enterococcus spp. (VRE) was first identified about three decades ago and has now become a major nosocomial pathogen. It typically infects immunocompromised patients and can cause endocarditis, bloodstream, urinary tract, and skin and skin structure infections [1]. VRE infections are generally more serious than those caused by vancomycin-susceptible enterococci, and are associated with higher mortality rates [2]. Among all VRE species, vancomycin-resistant Enterococcus faecium (VREfm) is responsible for the majority of hospital infections. VREfm has been recently listed as a high priority pathogen for research and development of new antibiotics by the World Health Organisation [3]. Various measures have been implemented to monitor the spread of VREfm infections, including multi-locus sequence typing (MLST); this relies on characterising the allelic profile of seven “house-keeping” genes, located in the E. faecium chromosome [4]. Although useful, this MLST scheme is of limited resolution to accurately capture the clonal type of isolated E. faecium strains [5–7]. Furthermore, VREfm isolates lacking the pstS MLST-gene locus have recently emerged in both England [7] and Australia [8].

Here, we have used whole genome sequencing to identify four VREfm isolates from two Scottish hospitals, which lack the pstS locus, and another Scottish VREfm isolate with a MLST profile that, to the best of our knowledge, has not been previously reported in the UK. Additionally, we provide information regarding their resistance profiles and epidemiology.

Materials and methods

VRE strains were isolated from five patients in two Scottish hospitals, between January to October 2017. These patients developed complications following either biliary or colonic surgery, and had been treated with various combinations of penicillin, amoxicillin, flucloxacillin, Tazocin, gentamicin, metronidazole and vancomycin during their hospital stay. The VRE strains were cultured on horse blood agar with single colonies transferred to Mueller Hinton broth (Oxoid) liquid. These were subsequently placed on Mueller Hinton agar. Single colony cultures of these were subsequently used for DNA isolation. Genomic DNA was extracted using an Isolate II genomic DNA kit (Bioline), using the manufacturer’s instructions for difficult to lyse Gram-positive bacteria. DNA libraries were then prepared using the NEBNext Fast DNA Fragmentation and Library Prep Set for Ion Torrent (New England Biolabs): briefly, 1 μg of genomic DNA was fragmented, and Ion Xpess barcode adapters (Life Technologies) were ligated to the DNA fragments; after clean-up using Agencourt AMPure XP beads (Beckman Coulter), 400 bp target fragments were isolated following 18 min electrophoresis on E-gel SizeSelect agarose gels (Life Technologies); these were subsequently amplified by PCR and, following another clean-up with Agencourt AMPure XP beads, the quality of the resulting DNA libraries was assessed on a 2100 Bioanalyzer (Agilent Technologies), using high sensitivity DNA chips (Agilent Technologies).Template positive Ion Sphere particles (ISPs) for semiconductor sequencing were prepared using the Ion Touch 2 System (Life Technologies). Enriched ISPs were loaded into ion v2 BC 316 chips (2 genomes per chip) and sequenced on an Ion PGM system (Life Technologies). Low quality reads (quality score threshold: 0.05) were trimmed using CLC genomics Workbench (Qiagen, version 9.5.2), and resulting reads were assembled using SPAdes (St. Petersburg genome assembler, version 3.9). Contigs having less than 1000 bp sequences were discarded. The remaining contigs were reordered on Mauve (version 20150226) using the complete E. faecium Aus0004 genome [9] as reference, and resulting genome sequences were submitted to GenBank under the accessions PJZU00000000 (VREF001), PJZT00000000 (VREF002), PJZS00000000 (VREF003), PJZR00000000 (VREF004) and PJZQ00000000 (VREF005). Antibiotic resistant genes in VREF001-5 genomes were predicted using the Resistance Gene Identifier tool within the comprehensive antibiotic resistance database (CARD) (https://card.mcmaster.ca/analyze/rgi). This tool utilizes algorithms for identification of perfect, strict or loose matches against resistant genes of the comprehensive antibiotic resistance database, based on homology and SNP models [10]. Sequence reads of pstS-null genomes from England and Australia were obtained from European Nucleotide Archive, and reads were assembled using CLC genomics Workbench; contigs having less than 200 bp were discarded. In silico MLST analysis was performed using the PubMLST website (http://pubmlst.org/) [11]. Alignments of genomes were done using the REALPHY (Reference sequence Alignment based Phylogeny builder) online tool (version 1.12) [12], with E. faecium Aus0004 [9], Aus0085 [13] as references, using default options (read length: 20, seed length: 22, polymorphism threshold: 0.95) and merging of reference alignments. All input genome sequences were assembled contigs in FASTA format and the two different reference genome alignments were combined to produce the resulting phylogenetic tree. Minimum inhibitory concentrations (MICs; μg/ml) for vancomycin, streptomycin, spectinomycin, tetracycline, oxytetracycline, doxycycline, minocycline and rifampicin were calculated using the microdilution method in cation adjusted Mueller-Hinton broth media (BD Biosciences), whereas MICs (μg/ml) for all other antibiotics were obtained using VITEK 2 (bioMerieux).

Results

The Scottish VRE strains sequenced (VREF001-5) had 2.9–3.0 Mb genomes with 37.6–37.7% GC content (Table 1). Analysis of average nucleotide identity and Tetra Correlation Search (TCS) against database genome sequences, using JSpeciesWS (version 3.0.12) [14], showed that all five VRE isolates were E. faecium. VREF001, VREF002, VREF004 and VREF005 were highly related with over 99.9% identity among aligned (>97%) sequences, whereas VREF003 displayed around 16% unique sequences compared with the other four genomes. In silico MLST analysis of VREF001-5 genomes identified VREF003 as sequence type (ST) 262, which has not been previously reported in the UK. The other four genomes had exact matches for six MLST alleles (atpA-9, ddl-1, gdh-1, purK-1, gyd-12, adk-1), but no match for pstS. This MLST profile has now been assigned as ST1424 in the E. faecium MLST database (https://pubmlst.org/efaecium/). However, despite the missing pstS in these four genomes, all VREfm strains sequenced in this study did have a pstS homologue within a pst operon (also referred to as pstS2), which is thought to be the actual pstS housekeeping gene in E. faecium [7,8].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Features of assembled sequences of the five isolates.

https://doi.org/10.1371/journal.pone.0218185.t001

We then searched for antibiotic resistance genes with perfect or strict matches to reference genes of the comprehensive antibiotic resistance database (CARD), but also considering loose matches with exceptional low e-value (<10⁻¹⁰⁰) and/or very high identity (>66%) to the reference gene. Analysis of the gene profiles of these five isolates identified the multi-aminoglycoside resistance gene aac(6')-Ii, in all five strains, and aac(6')-aph(2''), in all but VREF004. The isolates had at least one macrolide-lincosamide-streptogramin resistance gene (ermB), a pbp5 variant (designated pbp5-R) conferring resistance to β-lactams [15], for which no reference gene exists in CARD, and the vancomycin-teicoplanin resistance gene, vanA. With the exception of VREF003, all strains had the spectinomycin resistance gene ant(9)-Ia. Lastly, VREF003 was the only strain with a tetracycline transporter gene (tetL) (Table 2). It is notable that all isolates had two loose matches for the tetracycline resistance gene tetM. By further examining the region where these tetM-like sequences are located, we found that the five genomes had a Tn5801-like transposon (full-length or fragment) containing a sequence similar to Staphylococcus aureus Mu50 tetM gene [16], but with a 3,229 bp insertion. The trimethoprim-resistance gene dfrG and two more genes with unknown function were located within this insertion. A similar insertion has been reported previously in some ST17 and ST18 E. faecium strains [16] which is thought to result in defective protection against tetracyclines [17]. Lastly, all strains had a homologue of the E. faecalis IsaA gene which confers resistance to quinupristin-dalfopristin. However, the product of this gene (known as EatA), unlike the >99% identical EatA_v product found in some E. faecium strains, has a threonine at position 450, which is associated with increased susceptibility towards quinupristin-dalfopristin [18,19].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. MIC values (μg/ml) of selected antibiotics against the five Scottish VREfm isolates.

Susceptibility to a given antibiotic is indicated in bold. ND: MIC value not determined. Resistant genes identified which correspond to strict or perfect matches against antibiotic resistance genes of the comprehensive antibiotic resistance database (CARD) are shown. Additionally, some loose matches (underlined) to CARD reference genes, which either had > 99% identity with reference CARD gene (tetL), or had zero e-value (rpoB2), or were the sole genes that could explain differences in antimicrobial susceptibility among the five isolates (ant(9)-Ia, marA and mtrR), are shown. The variant of a full-length penicillin-binding protein 5 gene conferring resistance to penicillins (pbp5-R), for which no reference gene in CARD exists, was found in all genomes. Superscripted numbers at top right of resistant genes (far-right column) correspond to the sequenced strains (last digit of isolate ID) carrying these genes.

https://doi.org/10.1371/journal.pone.0218185.t002

All strains exhibited high level resistance to vancomycin; 4/5 had high level resistance to teicoplanin, associated with the presence of vanSA (absent from VREF005, which displayed low level resistance) and 4/5 had high level resistance to spectinomycin (apart from VREF003 which displayed low level resistance). The strains had low level resistance to streptomycin, as well as resistance to a variety of other drugs. The low level resistance to streptomycin and spectinomycin, in the absence of any gene coding for resistance to those antibiotics, is due to decreased uptake of these drugs by E. faecium [20]. All isolates were susceptible to linezolid, chloramphenicol, tigecycline and Synercid (quinupristin-dalfopristin). VREF003 was the only isolate found to be resistant to tetracycline, doxycycline, oxytetracycline and minocycline, but unlike the other four isolates, it was susceptible to rifampicin (Table 2). Since all five isolates had an identical rpoB2 variant (assumed to exert some level of rifampicin-resistance), the decreased resistance of VREF003 towards rifampicin, compared to the other four strains, may be due to the lack of any mtrR and marA homologues in the VREF003 genome. Conversely, loose matches of these were present in the other four isolates, and such genes have been shown to confer resistance to multiple drugs in other bacteria, including rifampicin [21,22]. Alternatively, VREF001-2 and VREF004-5 may contain an mtrR-independent mechanism of reduced permeability for rifampicin compared with VREF003 [23]. Antibiotic susceptibility tests indicated that all five isolates are classified as multi-drug resistant enterococci, according to standardised international terminology [24].

We then performed a phylogenetic comparison of VREF001-5 genomes with other known VREfm genomes reported to have a missing pstS locus; these included: 14 (out of 66 reported) Australian strains isolated from 9 different hospitals between 2014 and 2015 [8], 5 Australian strains (out of 202 reported within 6 local health districts) isolated from 2 different local health districts (LHD-1 and LHD-2) in New South Wales during 2016 [25], and all five English pstS-null E. faecium strains isolated from a single hospital (Kathy Raven, personal communication) in 2005 [7]. In our phylogenetic analysis we also included: five hospital-associated isolates of different MLST sequence type each, three isolates (BM4538, UAA1025 and E2883) previously reported to have an insertion in the tetM-locus [16] and the first complete E. faecium genome, Aus0004 [9]. The 14 Australian pstS-null genomes were carefully selected to represent pstS-null isolates from the 66 totally reported across Australia by Carter and co-workers [8], covering all hospitals, health jurisdictions and importantly, phylogenetically distinct clades.

This analysis indicated that Scottish pstS-null isolates are most closely related to Australian pstS-null isolates (Fig 1A; subclade 3–1), and particularly to certain ST1424 clones isolated from New South Wales (hospital 3 and Local Health District 1). Due to the existence of long branches within subclade 3–1 (Fig 1A), the phylogeny of that subclade was further refined by analysis of genomes belonging only to this subclade [12] (Fig 1B). Within this subclade, all ST1421, ST1422 and ST1424 clones were found to be closely related, whereas the sole ST1423 clone (vanB type DMG1500761) was found to diverge the most among all Australian and Scottish pstS-null strains (Fig 1A–1C). English pstS-null isolates, which are all ST1477, all cluster together (Fig 1A; subclade 2–2) and are distinct form Australian and Scottish pstS-null strains. Nevertheless, all pstS-null clones appear to be related to ST17 clones (Fig 1A; subclade 3–2 and Fig 1C), which suggests that the former may have derived from ST17 strains. Most of the Australian pstS-null genomes we analysed were ST1421 (Fig 1C), in agreement with recent data showing that ST1421 is the most common and widespread pstS-null VRE strain in Australia, accounting for more than 70% of cases [25,26]. Since both Australian [8] and Scottish (this study) VREfm pstS-null strains were found in different hospitals (Fig 1C), it is likely that such clones have spread in the community. Our phylogenetic analysis further supports an intercontinental ST1424 clone spread between hospitals of New South Wales and Scotland (Fig 1). On the other hand, all English VREfm strains with missing pstS locus [7] cluster together and are likely to have spread within the hospital from which they were isolated. VREF003 is more phylogenetically related to ST18 E. faecium isolates (Fig 1A), in agreement with a previous study showing a phylogenetic relationship between ST262 and ST18 E. faecium strains [27]. Unlike the sole pstS-null vanB type VREfm which had an intact tetM gene sequence in its genome, all the pstS-null vanA type VREfm strains included in our phylogenetic analysis (4/4 Scottish, 18/18 Australian and 5/5 English) (Fig 1C) had a Tn5801-like transposon with an insertion in their tetM locus. To identify how common this insertion is, we additionally tested the remaining pstS-null genomes (50 vanA-type and one van-negative) from the ones reported by Carter et al. All these, without exception, were identified to be ST1421 and to have the same tetM insertion within a Tn5801-like transposon. Such insertion, which has been previously documented in the MLST-typeable Aus0004, BM4538, E2883 and UAA1025 [16] (also included in our phylogenetic analysis), was further observed in VREF003, ERR374934 and ERR374968. Collectively, it appears that the tetM insertion in VREfm is strongly associated with vanA type strains missing pstS (STs: 1421, 1422, 1424 and 1477), but may also be found in some VREfms with STs: 17, 18, 80 and 262 (Fig 1C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Scottish pstS-null VREfm isolates are related to vanA-type Australian pstS-null VREfm isolates.

(A) Phylogenetic tree of selected E. faecium genomes, constructed using the REALPHY (Reference sequence Alignment based Phylogeny builder) online tool, with Aus0004 and Aus0085 used as reference genomes. pstS-null genomes are present in subclades 2–2 (English isolates) and 3–1 (Australian and Scottish isolates), (B) The phylogeny of strains clustering in subclade 3–1 (A) was further refined using REALPHY with Aus0085 and ERR374968 as references. (C) The MLST sequence type (ST) and alleles, country of origin, van genotype, hospital and year of isolation are given for each sequenced genome. * Only the Local Health District (LHD) is known for Australian strains isolated in New South Wales hospitals during 2016. Hospitals 3 and 9 are also in New South Wales jurisdiction, but their LHD is not known. Information is given on whether an insertion encompassing dfrG and two uncharacterised genes was found within tetM in these genomes. N/A: not applicable (no tetM sequences found). ** Although no such insertion was identified in Aus0085, there was a frameshift mutation within its tetM gene. pstS-null genomes of same ST are highlighted with the same colour.

https://doi.org/10.1371/journal.pone.0218185.g001

Discussion

MLST analysis is inferior to whole genome sequence analysis for outbreak investigations of VREfm [7], due to the high rate of recombination events occurring within the E. faecium chromosome that cannot be captured adequately by MLST analysis [27,28]. Furthermore, the emergence of non-typeable VREfm strains poses an extra obstacle in implementing the MLST scheme for VREfm phylogenetic analysis. Non-typeable VREfm strains have recently emerged and have been shown to be very rare in the UK. In a study encompassing whole genome sequencing of nearly 500 E. faecium healthcare-associated isolates from 2001–2011 in the UK and Ireland, there were only five cases of pstS-null VRE strains, all of which were isolated from a single English hospital, between 2004 and 2005 [7].The emergence of such strains was even more recent in Australia: the first two strains were isolated in 2013, but numbers increased rapidly to a total of 89 cases, by the end of 2015 [8,29,30] and to about 300 cases, by the end of 2016 [25,30,31].

Here, we report four additional non-typeable VREfm strains isolated in 2017, which are the first identified in Scotland. Their striking genomic similarity, combined with the fact that one was isolated from a different hospital, indicates that, although deriving from a single strain type, they are not necessarily hospital-acquired, but may be health-care associated, or spread within the community. Although transmission may have occurred through the frequent transfer of equipment, specimens, staff and patients between these two hospitals (located in the same health board just 14 miles apart), evidence suggests that VRE strains are also transmitted outside hospitals, such as in long term care facilities [6]. Additionally, VRE clones can disperse outside health-care facilities, since strains can be carried in patients for long periods, ranging from a few months to a few years following discharge from hospital [32,33]. For the above reasons VRE infections may spread over long periods and across long distances. Indeed, our phylogenetic analysis is indicative of an inter-continental spread between Australia and Scotland: the closest phylogenetic relationships were identified among Australian and Scottish ST1424 clones. Such six-locus sequence type has been identified in 28.7% (58/202) of the recently sequenced pstS-null VREfm strains from New South Wales, Australia [25]. In addition, the spread of such strains in New South Wales is also suggestive of inter-hospital transmission for this VREfm clone [25]. Even within the same hospital, VRE transmission routes can be prolonged and complex with closely related VRE strains re-appearing in the same ward after several months, or transmitted to patients located in different wards [34]. We also observed re-appearance of a pstS-null strain (isolate VREF005) in the same hospital, seven months after discharge of the last carrier patient.

With the exception of one vancomycin-susceptible and one vanB type strain, all other (69) previously sequenced pstS-null E. faecium strains have been found to be vanA type [7,8]. Consistent with these studies, the newly sequenced Scottish VREfm strains with the missing pstS locus were also vanA type. In fact, the vast majority (84.2%) of all sequenced vanA E. faecium 2016 isolates from New South Wales, have been identified as pstS-null strains [25]. Hence, pstS-null VREfm appears to be a very successful vanA clone. We further identified that loss of pstS in vanA type VREfm is strongly associated with an insertion in a tetM gene (74/74 cases tested). Such insertion was present in all ST4121, ST1422 and ST1424 strains tested, isolated from around a dozen hospitals across Australia (New South Wales, Victoria and Australian Capital Territory) and the UK (England and Scotland), and was only absent in the sole vanB type pstS-null E. faecium strain identified. Although we cannot exclude the possibility that vanA-type pstS-null members with an intact or no tetM sequence exist (the vast majority of the recently isolated ST1421 and ST1424 stains from New South Wales were excluded in our analysis), our data strongly suggest that the Tn5801-like transposon with the tetM insertion is an extremely common feature of vanA-type pstS-null strains. How strongly associated these two features are (missing pstS and tetM insertion), why they were co-selected in numerous vanA type VREs and what is the significance of this association, are issues that need to be investigated further. The insertion within the tetM gene apparently leads to a mutant TetM protein which is unable to confer resistance towards tetracyclines, as it can be inferred from relevant susceptibility tests against tetracycline, doxycycline and minocycline, using Aus0004 [17], VREF001-2 and VREF004-5 (this study). All sequenced isolates were resistant to trimethoprim, and the cause of such resistance is most likely the dfrG gene inserted into the tetM locus. None of the isolates sequenced contained a gene for tigecycline, Synercid, or linezolid resistance, and as a consequence they were susceptible to these three drugs. Since these antibiotics are frequently used for management of multi-drug-resistant enterococcal infections [1,35], they may still be the favourable option for combating VREfm strains with a missing pstS locus.

The emergence of pstS-null VREfm strains in Scotland raises concerns for a potential outbreak. These strains were found to be highly similar to Australian pstS-null strains, which have very rapidly expanded to become the leading strain in Australia (2016). They are causing infections in numerous hospitals and in different health jurisdictions [8,25,26]. Furthermore, the missing pstS locus of such strains renders the—limited in resolution—MLST analysis even less suitable for VREfm outbreak investigations. Hence, our study, like previous ones [7,8,25], highlights the need for whole genome sequencing approaches in monitoring VREfm epidemiology.

Acknowledgments

We are very thankful to Gavin Paterson and Alison MacFadyen for reading the manuscript and for their useful comments.