Introduction

Vancomycin-resistant Enterococcus faecium (VREF) is a major pathogen causing healthcare-associated infections in many countries and has been particularly associated with outbreaks in hospitals. Multilocus sequence typing (MLST) has been successfully used to study the molecular epidemiology and population structure of E. faecium [1]. However, the genomic plasticity and propensity for recombination seen in the enterococci may be limiting factors for epidemiological investigations by MLST [1]. In particular, four MLST loci (atpA, gyd, pstS, and ddl) are located in recombination hotspots [2]. Recently, VREF strains lacking pstS have been reported in Australia [3], the UK [2], Scotland [4], and Denmark [5].

We have identified the nationwide emergence of VREF isolates lacking the pstS gene in Korea and hereby report the molecular characteristics of these isolates.

Materials and methods

Bacterial isolates and microbiological characteristics

This study was conducted at Samsung Medical Center (Seoul, Korea), a large tertiary referral hospital where more than 70% of hospitalized patients came from other parts of the country other than Seoul. A total of 300 VREF isolates were collected using rectal swab surveillance cultures from newly admitted patients between April 2014 and March 2015 according to the infection prevention policy of the hospital. In particular, screening tests were performed on patients transferred from other hospitals. In 46 (15.3%) of these isolates, typing by MLST was not possible because the pstS gene was not detected. MLST was conducted using PCR and sequencing of seven housekeeping genes (adk, atpA, ddl, gdh, gyd, purK, and pstS). Allelic profiles and sequence types (STs) were assigned according to the database available on the MLST website (https://pubmlst.org/efaecium/). Species identification and antimicrobial susceptibility testing were performed using the VITEK 2 system (bioMeriéux, Marcy-l’Etoile, France). The presence of genes encoding glycopeptide resistance (vanA, vanB, vanC, vanD, vanE, and vanG) was determined by PCR using primers previously described [6]. Virulence genes including enterococcal surface protein (esp) and hyaluronidase (hyl) were detected by PCR and confirmed by sequencing [7]. To identify the pstS2 gene used for MLST in the place of pstS [3], we performed PCR using the following specific primers; pstS2-Forward (5′-CCT TGC CAA TCA CGT TCG AC-3′) and pstS2-Reverse (5′-ATC CAT TAC GGC TGT TGG CT-3′) that amplified a 755 bp internal fragment of the gene. For pulsed-field gel electrophoresis (PFGE), bacterial DNA was digested with the SmaI restriction enzyme (TaKaRa Bio Inc., Shiga, Japan) and separated by electrophoresis using the CHEF DR II system (Bio-Rad Laboratories, Hercules, CA, USA). Potential clonal relatedness was determined at the ≥ 80% level of similarity.

Whole-genome sequencing

Four VREF strains (2014-VREF-041, 2014-VREF-063, 2014-VREF-114, and 2014-VREF-268) that could not be typed by MLST were randomly selected for whole-genome sequencing (WGS) based on the allelic profiles of six housekeeping genes and the presence of esp and hyl virulence genes. High-quality, high-molecular-weight genomic DNA (gDNA) was extracted using the G-spin™ Genomic DNA Extraction Kit (iNtRON, Korea) according to the manufacturer’s instructions. The gDNA was used to prepare approximately 20 kb size-selected SMRTbell templates. SMRT libraries were prepared by ligating hairpin adaptors to both ends using the PacBio DNA Template Prep Kit 1.0 (for 3–10 kb) for SMRT sequencing with C4 chemistry. Data (240 min movies) were collected using the PacBio RS II instrument (Pacific Biosciences, Menlo Park, CA, USA). Sequencing raw data was analyzed using the SMRT Analysis software (v2.3). The SMRT Portal allowed the execution of all HGAP steps in the web-based GUI.

Bioinformatic analyses

An overall comparison of the architecture of the isolates was performed with the Mauve genome alignment viewer (Biomatters Ltd., New Zealand) [8] and the BLAST Ring Image Generator (BRIG) [9], as previously described. Specific plasmid genes containing Tn1546 were further examined with BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi). Bacterial Pan-Genome Analysis (BPGA) was performed for the core genomes and dispensable/accessory genomes among the strains [10]. Each core genome was extracted by BPGA using the USEARCH clustering algorithm, and multiple sequences were aligned using MUSCLE. All alignments were concatenated, and a neighbor-joining phylogenetic tree was constructed [10]. The phylogenetic tree was generated using FigTree v1.4.3 (http://tree.bio.ed.ac.uk /software/fig tree/). A bootstrap consensus tree with 1000 replicates was constructed by the maximum likelihood method based on the JTT matrix-based model [11].

Supplementary experiments using clinical isolates from nationwide surveillance collection

We determined the prevalence of VREF isolates lacking the pstS gene among the collection of VREF clinical isolates. These isolates came from the Asian Bacterial Bank (the Asia Pacific Foundation for Infectious Diseases, Seoul, Korea) and had been collected through Korean national surveillance studies of bacteremia and urinary tract infection during 2006–2007 (9 university hospitals participating) and 2012–2013 (16 university hospitals) funded by the Korea Centers for Disease Control and Prevention [12, 13].

Nucleotide sequence accession numbers

Genome read sequences for the four VREF ST1421 strains and annotation data of the isolates have been deposited in National Center for Biotechnology Information (NCBI) under project accession number (PRJNA358851) as shown in Table 1.

Table 1 Summary report of the de novo-assembled vancomycin-resistant E. faecium ST1421 strains from this study
Full size table

Results

Among the 46 isolates that lacked the pstS gene and were non-typable at the beginning of this study, 45 could be typed as ST1421 and one as ST1424 using a new PubMLST designation scheme. In this new scheme, the missing housekeeping gene allele, pstS, was set to 0 and ST17-related strains and ST80-related strains were assigned as ST1421 and ST1424, respectively. All these strains were isolated at the time of admission from patients who were referred from eight different provinces in Korea. The ST1421 strains were further subdivided into three groups according to the presence of the esp and hyl virulence genes. Thirty-nine ST1421 strains (84.7%) containing both esp and hyl were classified into ST1421-V1 (Fig. 1). ST1421 strains carrying only either esp or hyl were grouped into ST1421-V2 (8.7%) and ST1421-V3 (4.4%), respectively. All isolates contained the vanA gene, and analysis of the PFGE patterns revealed 32 pulsotypes (Fig. 1). All strains belonging to the predominant pulsotype P5 were ST1421-V1, and province of residence of the patients with these isolates was diverse (6 provinces).

Fig. 1
figure 1

Dendrogram of the PFGE patterns of 46 vancomycin-resistant Enterococcus faecium (VREF) strains with a pstS gene deletion. The boxes in dashed lines indicate the strains on which whole-genome sequencing was performed

Full size image

The genome sizes of the four ST1421 strains ranged from 3,255,197 to 3,401,730 bp. Transposon Tn1546 was found within the plasmids of each strain, but the structure of Tn1546 varied among strains. The structures of the plasmids were analyzed. All isolates harbored Tn1546 components (vanR, vanS, vanHAX, vanY, and vanZ) with various transposase and insertion sequences (Fig. 2). Only the original structure (vanRS, vanHAX, and vanYZ) was found in 2014-VREF-041 and 2014-VREF-114. However, additional cassettes with a vanYZ deletion in the original structure were detected in 2014-VREF-063 and 2014-VREF-268. We confirmed that the additional cassette was due to two copies in 2014-VREF-063. 2014-VREF-041 and 2014-VREF-268 carried the aph(3′)-IIIa gene mediating aminoglycoside resistance and the erm(B) gene causing macrolide resistance. 2014-VREF-063 had one resistance gene, aph(2″)-Ia, while 2014-VREF-114 had no resistance genes.

Fig. 2
figure 2

Genetic environment of Tn1546 transposon components in four VREF ST1421 isolates. Red and blue arrows indicate antibiotic resistance genes and transposon Tn1546 components, respectively

Full size image

The architecture of the chromosomes of the four ST1421 strains relative to a reference strain (E. faecium Aus0085) was studied using the Mauve tool (Fig. 3a). Mauve alignment of the four ST1421 chromosomes identified 25 locally collinear blocks (LCBs) with a variety of inversions. A loss of one or more LCBs compared with Aus0085 was identified in all strains. To identify a genetic deletion in the pstS gene, we compared the positions containing the pstS gene. All ST1421 isolates showed deletion of the pstS gene and hypothetical proteins (Fig. 3b). All ST1421 isolates except one (2014-VREF-114) showed an insertion of prophage ps2 probable integrase (DNA recombination and DNA integration function genes) in front of the 50S ribosomal protein L19 (rplS gene), the first gene of LCB containing pstS gene (Fig. 3b). One strain (2014-VREF-114) was found to contain a putative transposase (InsK) for the insertion sequence element IS150 inserted in the pstS gene position.

Fig. 3
figure 3

Comparative analysis of the chromosome of four VREF ST1421 strains relative to the reference strain E. faecium Aus0085. a Sequences were aligned using the Mauve genome alignment viewer. Each genome panel is composed of different color blocks that represent locally collinear blocks (LCBs) that are aligned to the parts of other genomes. White areas between the LCBs contain sequence elements that are specific to the indicated genome and that are not aligned. LCBs containing the pstS gene are indicated with a dark triangle. b Genetic organization of the LCBs containing the pstS gene is shown diagrammatically for the strains used in this study and reference strain. Genes are color-coded according to predicted function of the corresponding protein product, with associated Cluster of Orthologous Groups (COGs) domains indicated. The inserted genes compared with the reference strain are shown with a red arrow, and an asterisk (*) represents the position of the pstS gene that was deleted in our four VREF ST1421 strains. The scale represents DNA length in 0.5-kb pairs

Full size image

Chromosome structures of the four ST1421 isolates were visually compared with that of Aus0085 using BRIG (Fig.4). The locations of the seven housekeeping genes and the pstS2 gene, which had been reported in Australia [14], are shown. Compared with the reference strain, four strains in this study were missing several gene segments in common. The 41 VREF ST1421 isolates and one VREF ST1424 isolate that were not sequenced were also determined to contain a chromosomal pstS2 gene.

Fig. 4
figure 4

Comparison of chromosome analysis using the BRIG bacterial pan-genome analysis pipeline. Each ring corresponds to one of the four VREF ST1421 chromosomes (indicated at the right of the figure together with the color code). The E. faecium Aus0085 chromosome (inner black ring) was used as a reference. Genes highlighted in red indicate the seven housekeeping genes in E. faecium

Full size image

We selected 25 E. faecium strains with a query coverage percentage of over 80% and an identity greater than 99% compared with the genomes of our four ST1421 strains using BLAST. By BPGA, we selected 1506 conserved core gene clusters from the 25 E. faecium reference strains, the four ST1421 strains of this study, and an Australian strain (DMG1500501) showing a pstS deletion. The phylogenetic tree generated using FigTree showed that only one ST1421 strain (2014-VREF-114) grouped with the reference strains, whereas the other three ST1421 strains and DMG1500501 (highlighted in blue) were not related to the reference strains (Fig. 5).

Fig. 5
figure 5

Maximum likelihood, circular phylogenetic tree of 30 Enterococcus faecium strains based on the JTT matrix-based model. Initial trees for the heuristic search were obtained automatically by applying the Neighbor-Join and BIONJ algorithms to a matrix of pairwise distances estimated using a JTT model and then selecting the topology with superior log likelihood value. The scale bar (2.0) indicates the number of nucleotide substitutions per site. The four VREF ST1421 isolates in this study are highlighted in red, whereas the Australian strain, DMG1500501, that showed deletion of pstS gene is highlighted in blue

Full size image

A supplementary investigation with VREF clinical isolates collected through Korean national surveillance studies of bacteremia and urinary infection revealed that VREF strains with a deleted pstS gene accounted for 12.1% of 33 isolates in 2006–2007 and accounted for 11.8% of 59 isolates in 2012–2013. Among these eleven isolates, eight were ST1421 with esp and hyl genes belonging to group ST1421-V1, although the hospitals where the strains were collected varied. Of those, seven strains were isolated during 2012–2013 (Table 2).

Table 2 Characteristics of 11 VREF clinical isolates lacking the pstS gene
Full size table

Discussion

In our study, 46 (15.3%) of 300 VREF rectal isolates collected from 2014 to 2015 were ST1421 or ST1424 strains lacking the pstS gene. The failure in MLST genotyping due to the pstS gene deletion in a substantial proportion of VREF isolates coincides with a previous report from Australia [3]. WGS in the four ST1421 strains in this study determined that DNA recombination and transposition genes were inserted, resulting in considerable inversion and loss of the pstS gene. The same insertion sequence was identified in front of the pstS gene region of three VREF ST1421 strains (2014-VREF-041, -063 and -268), whereas a putative transposase was inserted into the pstS gene position in one strain (2014-VREF-114) with a deletion of the pstS gene. Thus, significant recombination in the chromosome and considerable inversions identified in our strains might have affected the housekeeping gene region. E. faecium is known to have a high rate of recombination [14]. Our results support the suggestion that WGS should be used in tracing the epidemiology of E. faecium nosocomial infections [15]. The majority VREF strain lacking the pstS gene in our study was ST1421, which has an identical allelic profile to ST17 for six housekeeping genes, all except pstS. Considering that ST17 is the predominant VREF genotype isolated from healthcare-associated infections in Korea [16], these strains can be assumed to be derived from ST17. There are 16 single-locus variants (SLVs) that have the same allelic profile as ST17 except for the pstS gene according to the MLST web site (ST180, ST181, ST182, ST202, ST233, ST386, ST472, ST485, ST538, ST723, ST803, ST1021, ST1061, ST1335, ST1394, and ST1471). Therefore, another possibility is that the ST1421 strains originated from the SLVs of ST17. In fact, analysis of the PFGE patterns revealed considerable heterogeneity among the 45 VREF ST1421 strains. The finding that very diverse ST1421 strain pulsotypes were found in different geographical regions in the country suggests that deletion of the pstS gene occurred sporadically in diverse clones.

Phylogenetic analysis showed that pstS gene-deleted strains emerged into new clones with distinct phylogenetic patterns and suggested that VREF ST1421 strains emerged and spread in a non-clonal manner. The mutation in which the pstS gene is deleted from VREF appears to have occurred nationwide in Korea. The results of our supplementary experiments using VREF isolates collected nationwide for the surveillance of bacteremia and urinary tract infections support this. Notably, these results indicate that VREF with the pstS deletion had already emerged nationwide in Korea in 2006, much earlier than when similar strains were found in Australia [3].

Multiple copies of the vanA gene were identified in our isolates through the analysis of Tn1546. The copy number of some resistance genes has been reported to affect the minimum inhibitory concentrations (MICs) of corresponding antibiotics [17]. In our study, however, no correlation between vanA gene copy number and vancomycin MICs was observed, and all isolates represented high-level resistance to vancomycin (MIC > 32 mg/l). It is suggested that our ST1421 strains acquired a plasmid-mediated vanA gene with multiple copies in clones in which the pstS gene is missing.

There are some limitations in our study. First, WGS was performed in only four ST1421 strains. In the other 42 isolates (ST1421 and ST1424) that failed to amplify the pstS gene, details about the genome structures near the pstS region were not determined. Second, rectal bacterial isolates were collected from a single center; however, the strains were isolated from patients who were referred from various provinces in Korea. Additionally, clinical isolates from bacteremia and urinary tract infections were collected from multiple centers through a national surveillance. Therefore, our results could be interpreted as representing the country.

In conclusion, VREF ST1421 strains lacking the pstS gene have emerged in Korea and account for a substantial portion among rectal and clinical isolates. Although isolates with deletion of the pstS gene have been assigned as new sequence types, the emergence and spread of pstS-deleted VREF strains pose a serious challenge for epidemiological investigation. Alternative molecular typing methods to MLST will be increasingly necessary.