Molecular Analysis of VanA Outbreak of Enterococcus faecium in Two Warsaw Hospitals: The Importance of Mobile Genetic Elements

Vancomycin-resistant Enterococcus faecium represents a growing threat in hospital-acquired infections. Two outbreaks of this pathogen from neighboring Warsaw hospitals have been analyzed in this study. Pulsed-field gel electrophoresis (PFGE) of SmaI-digested DNA, multilocus VNTR analysis (MLVA), and multilocus sequence typing (MLST) revealed a clonal variability of isolates which belonged to three main lineages (17, 18, and 78) of nosocomial E. faecium. All isolates were multidrug resistant and carried several resistance, virulence, and plasmid-specific genes. Almost all isolates shared the same variant of Tn1546 transposon, characterized by the presence of insertion sequence ISEf1 and a point mutation in the vanA gene. In the majority of cases, this transposon was located on 50 kb or 100 kb pRUM-related plasmids, which lacked, however, the axe-txe toxin-antitoxin genes. 100 kb plasmid was easily transferred by conjugation and was found in various clonal backgrounds in both institutions, while 50 kb plasmid was not transferable and occurred solely in MT159/ST78 strains that disseminated clonally in one institution. Although molecular data indicated the spread of VRE between two institutions or a potential common source of this alert pathogen, epidemiological investigations did not reveal the possible route by which outbreak strains disseminated.


Introduction
Since the first isolation of vancomycin-resistant enterococci (VRE) in 1986 [1,2], this phenotype has spread rapidly and now is present in hospitals worldwide [3]. In Poland, the first VanA outbreak took place in the adult hematology ward of Gdansk Medical University in December 1996, followed by outbreaks in other centers [4]. The predominant species among VRE is Enterococcus faecium (VREfm). The majority of worldwide VREfm belongs to the meroclone CC17 (ciprofloxacin-and ampicillin-resistant and enriched in putative virulence traits), recently split into three distinct lineages, 17, 18, and 78, that evolved in hospital environment through horizontal gene transfer (HGT) and recombination processes [5]. These hospital-adapted lineages play a crucial role in the emergence and spread of VREfm.
The vanA gene cluster is a widely studied vancomycin/ teicoplanin resistance determinant, described as part of Tn1546-type transposons, generally carried on plasmids and thus effectively disseminated by HGT [6]. An acquisition of vanA plasmid by a strain of E. faecium representing hospitaladapted lineage may result in a spread of VREfm, first colonizing patients and then causing symptomatic infections. Therefore, both characterization of the Tn1546 structure and its linkage to particular plasmid groups is crucial for understanding of VRE dissemination in hospital environments. Several studies have shown the presence of various Tn1546 2 BioMed Research International types on Inc18, pRUM-like, pMG1-like, and pLG1 plasmids [7][8][9][10][11][12][13]; however, our knowledge of vanA plasmids and their epidemiology is still far from being satisfactory and the common presence of plasmids with Tn1546, belonging to unknown replicon types, has been shown [10,14].
The aim of this study was to characterize E. faecium VanA isolates from the outbreaks that concomitantly took place in hospital wards of two neighboring medical centers, The Institute of Oncology (IO) and The Institute of Hematology and Transfusion Medicine in Warsaw (IH). The investigation focused on the clonal relationships among isolates as well as analysis of the Tn1546 transposon structure and colocalization of vanA with other plasmid genes in order to elucidate the role of particular MGE during a VREfm outbreak in hospital settings.

Outbreak Description, Bacterial Isolates, and Susceptibility Testing.
Forty-four vancomycin-resistant E. faecium outbreak isolates were collected between February and June 2009 in two neighboring hospitals in Warsaw, The Institute of Oncology (IO) and The Institute of Hematology and Transfusion Medicine (IH), 776-and 198-bed hospitals, respectively. First VREfm was isolated from stool of 46-year-old patient on 4th February at the Gastroenterology Clinic of IO. Until the end of February, eight more cases were reported, in majority from the Clinic of Lymphatic System Cancers of IO. From the 31st March till the 18th of April, 18 VREfm were isolated, mainly from patients of this clinic (16 cases) and from two patients of the Gastroenterology Clinic. Simultaneously, VREfm cases were reported in IH wards, with the first two isolations on the 5th February from rectum and stool of the Hematology Ward patient and a patient from the ICU, respectively. One more isolate was obtained 10 days later in the Surgery Ward and till the end of June, 14 other VREfm cases were reported in the Hematology Ward of IH. Altogether, the outbreaks affected 42 patients, including 27 patients of IO (27 stool isolates) and 15 patients of IH (13 stool, 1 urine, 3 blood isolates). Antimicrobial susceptibility of collected isolates was determined using the Etest method (bioMérieux, Marcy l'Etoile, France) for glycopeptide susceptibility testing and broth microdilution method for other antimicrobials. The results were interpreted following the breakpoints of the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [15]; for chloramphenicol, erythromycin, ciprofloxacin, and tetracycline the Clinical and Laboratory Standards Institute (CLSI) [16] breakpoints were applied, and in the case of kanamycin and clindamycin, the breakpoints proposed by the Société Française de Microbiologie (SFM) [17] were used. The Enterococcus faecalis strain ATCC29212 was used for quality control purposes during testing. E. faecium BM4147 was used as a control VanA strain in this study.

DNA Isolation and Genotyping of Isolates.
Total DNA of isolates was extracted using Genomic DNA Prep Plus kit (A&A Biotechnology, Gdansk, Poland), following the manufacturer's instructions. Additionally, as the above method may result in a low yield of small plasmids, plasmid DNA was isolated using the alkaline lysis method [18]. Pulsedfield gel electrophoresis (PFGE) was performed according to de Lancastre et al. [19] for agarose plugs preparation, followed by the procedure of Clark et al. [20] for total genomic DNA purification. Purified DNA in plugs was digested with the SmaI restriction enzyme (Fermentas, Vilnius, Lithuania). Electrophoresis was performed at 14 ∘ C for 22 h with a pulse time of 1-30 s at 6 V/cm 2 in 0.5x TBE buffer and the results were interpreted according to criteria proposed by Tenover et al. [21]. The Bionumeric software (Applied Maths, Kortrijk, Belgium) was used to analyze the similarity of PFGE-banding patterns, with an unweighted pair group method with arithmetic average (UPGMA) algorithm and Dice coefficient. Multilocus variable-number tandem repeat (VNTR) analysis (MLVA) was performed as described by Top et al. [22] with modifications given on the website (http://www.mlva.umcutrecht.nl). Multilocus sequence typing (MLST) was performed as described previously [23]. Allele numbers and sequence types (STs) were assigned using E. faecium MLST database (http://efaecium.mlst.net/; 16th December 2013, date last accessed). PCR detection of IS16 was performed as described by Werner et al. [24]. The Simpson index and Wallace index were calculated using the online tool available at http://darwin.phyloviz.net/ComparingPartitions/ (14th January 2014, date last accessed).

Plasmid Profiling, Hybridization Analyses, and Tn1546
Typing. DNA in agarose plugs obtained as described above was treated with 14U of S1 nuclease (Takara Bio, Japan) for 15 minutes at 37 ∘ C and separated by PFGE at 14 ∘ C for 22 h with pulse time 5-35 s at 6 V/cm 2 in 0.5x TBE buffer [39]. This method allows visualization and determination of the number and size of plasmids larger than approximately 30 kb. After electrophoresis, DNA was blotted onto the Hybond-N+ membrane (GE Healthcare, Buckinghamshire, UK) by capillary transfer. Probe labeling and signal detection for PFGE-S1 membranes were carried out using the Amersham ECL Random-Prime Labeling and Detection System (GE Healthcare), according to the manufacturer's protocol.
The 4.4 kb fragment of the vanRSHAX operon was amplified using Expand Long Template System (Roche Diagnostics GmbH, Mannheim, Germany) according to Palepou et al. [40] with the following amplification conditions: 94 ∘ C for 2 min; 10 cycles of 94 ∘ C for 10 s, 56 ∘ C for 30 s, and 68 ∘ C for 4 min; 20 cycles of 94 ∘ C for 10 s, 56 ∘ C for 30 s, and 68 ∘ C for 4 min (with the elongation time increased by 20 s/cycle); and 68 ∘ C for 7 min. L-PCR amplicons were analyzed by restriction fragment length polymorphism (RFLP) with DdeI (New England Biolabs, UK). The whole Tn1546 transposon was investigated by PCR mapping and sequencing (Table 1 and references therein).

Conjugation Experiments.
Conjugation transfer of vancomycin resistance was examined by cross-streak mating procedure with E. faecium strain 64/3 resistant to rifampin and fusidic acid as recipient. Fresh colonies of donors were crossstreaked with recipient on BHI-Agar plates and incubated overnight at 37 ∘ C. Bacterial cells from the streak crossing area were then incubated overnight in 37 ∘ C on selective media. Transconjugants were then confirmed by MLVA. For isolates negative for conjugation in this assay, a technique specific for bacteria with low frequency of transfer was used [44].

Antibiotic Resistance Phenotypes, Antimicrobial Resistance
Determinants, and Virulence Genes. All analyzed isolates were resistant to vancomycin and teicoplanin and exhibited the presence of vanA determinant (Table 2). Additionally, all of them were penicillin-, ampicillin-, ciprofloxacin-, clindamycin-, and erythromycin-resistant. The vast majority of isolates from both IO and IH showed resistance to rifampin. High-level resistance to gentamicin (HLGR), kanamycin (HLKR), and streptomycin (HLSR) was more prevalent among IH isolates, which were particularly enriched in aminoglycoside resistance genes aac(6 )-Ie-aph(2 )-Ia, aph(3 )-IIIa, and aad6 ( Figure 1). The aph(2 )-Ib gene occurred in nine isolates and three other tested genes, coding for aminoglycoside resistance; that is, aph(2 )-Ic, aph(2 )-Id, and ant(4 )-Ia were not detected. Isolates from both groups commonly carried erm(B) and tet(M) genes. Resistance and intermediate susceptibility to tetracycline was typical for 61% and 18% of isolates, respectively. Intermediate susceptibility to chloramphenicol and quinupristin-dalfopristin was shown for 51% and 29% of isolates, respectively. All isolates were susceptible to linezolid and tigecycline.
Among virulence determinants studied, the ℎ Efm gene was prevalent in both outbreaks, while the Efm gene was present mainly in IH (Table 2 and Figure 1). All Efmpositive isolates harbored the intA integrase gene. PGC genes fms21 (PGC-1), fms5 (PGC-3), and fms19 (PGC-4) commonly occurred in the whole studied collection, while the fms17  (PGC-2) was more prevalent among IH than IO isolates. Genes gel, asa, and cyl were not detected.

Tn1546 Structures and Transferability of Vancomycin
Resistance. All isolates exhibited the presence of 4.4 kb L-PCR product containing the vanRSHAX operon and showed that the DdeI restriction pattern is identical to the E. faecium BM4147 control VanA strain. Further PCR mapping and sequencing showed the presence of ISEf1, inserted at the position 9147 nt of Tn1546 (numbering according to the GenBank sequence M97297), that is, within the vanX-vanY intergenic region. The 5 GACTGAAA duplication was observed at the insertion site. ISEf1 was present in all but two isolates with the prototype Tn1546. One of the isolates was derived from IO and the other from IH, and each of them showed a unique PFGE type, PT3 and PT10, respectively (Table 2). Similarly, all isolates, except for the two mentioned above, exhibited the G7747C point mutation of Tn1546 located within the vanA gene, resulting in the amino acid substitution V257F. Both isolates with the prototype Tn1546 showed higher teicoplanin MIC values compared to the isolates with Tn1546: ISEf1.
Conjugation experiments were performed for all 44 isolates and 34 of them were able to transfer vancomycin resistance to the E. faecium 64/3 recipient. All donors produced transconjugants with cross-streak mating except for a single isolate, which required use of the method designed for strains with low-level conjugation frequencies [44]. Susceptibility testing of transconjugants (a single transconjugant for each donor) showed a concomitant transfer of erythromycin resistance in 32 cases. One of these transconjugants showed also HLGR phenotype and one was additionally resistant to tetracycline. , and pMG1 ( = 18) were also detected. First three of them were characteristic mainly for the IH outbreak, while pMG1 occurred mainly in isolates from IO ( Figure 1). The number of rep genes per isolate varied from three to seven, and the average number of plasmid rep genes per isolate was 4.50; however, this value was lower for IO (4.07) compared to IH (5.18). Analysis of distribution of relaxase genes revealed the common presence of two relaxases, pCIZ2 and pEF1 , while pHT was predominantly detected in IO and the distribution of this gene was completely concordant with the presence of pMG1 . Additionally, one IH isolate had the pAD1 gene. Screening for plasmid toxin-antitoxin systems (TA) resulted in the detection of axe-txe and --, while other TA genes, including ccd, higBA, mazEF, par, parDE, phd-doc, relBE, and vagCD were absent in the studied group. All but one ---positive isolates were also 2 pRE25 -positive and only one 2 pRE25 -positive lacked the --gene.

Colocalization of vanA Determinant and Other Plasmid
Genes. Twenty-seven selected isolates (11 from IO and 16 from IH) of various clonal types, as defined by MLVA, MLST, and PFGE, were subjected to PFGE-S1 analysis, followed by Southern blot hybridization (Figure 3 and Table 3) with probes specific for genes detected earlier by PCR, such as vanA, seven rep genes ( 2 pRE25 , 11 pEF1071 , 14 pRI1 , 17 pRUM , 18 pEF418 , pLG1 , and pMG1 ), genes of two plasmid TA systems ( --and axe-txe), and three other plasmid-associated genes (pilA, ℎ Efm , and aac(6 )-Ieaph(2 )-Ia). Altogether, 122 plasmid bands were visualized in PFGE-S1, with 56 megaplasmids bands greater than 100 kb. The average number of plasmid bands in PFGE-S1 analysis was 4.35, with very similar values for both IO and IH outbreaks. Thirty plasmid bands hybridized with the vanA probe; that is, three of the analyzed isolates carried two vanA plasmids. The majority of vanA plasmids were <30-100 kb in size; additionally, four megaplasmids (170, 200, 240, and 315 kb) were associated with the vanA determinant. Among vanA plasmids, 24 were 17 pRUM replicons, mostly of 50 kb and 100 kb. The 100 kb plasmid was present in 15 isolates of various clonal backgrounds in both IH and IO and most of these isolates easily produced transconjugants. Moreover, the first observed VREfm isolates in both IH and IO carried such plasmids but in different clonal backgrounds. The 50 kb plasmid was associated exclusively with MT159 isolates, which differed, however, in their PFGE patterns. Six such isolates occurred exclusively in IH and all of them were deficient in conjugation. Five of vanA-17 pRUM plasmids of various sizes hybridized also with other rep genes, such as pLG1 (two 100 kb plasmids and one 315 kb plasmid), 18 pEF418 (one 100 kb plasmid), and both 2 pRE25 and 18 pEF418 (a 40 kb plasmid). Two isolates of MT1/PT4/ST18 from IO and IH both had 100 kb 17 pRUM plasmids but they differed in the content of other plasmids (Table 3, isolates labeled C and X); moreover, 100 kb plasmids from IO isolate additionally carried 18 pEF418 . Among the remaining plasmids, other than 17 pRUM replicons, vanA plasmids, a single 70 kb plasmid had 2 pRE25 and a 240 kb megaplasmid hybridized with pLG1 and 18 pEF418 . Considering other tested genes, the pilA gene was associated with 18 vanA plasmids, including all the 100 kb plasmids with 17 pRUM ; genes of the --TA system were present on a single 40 kb plasmid with 2 pRE25 , 17 pRUM , and 18 pEF418 and on a 70 kb plasmid carrying 2 pRE25 . The 240 kb vanA megaplasmid carried also the aac(6 )-Ie-aph(2 )-Ia resistance gene. Each of the two isolates with the prototype Tn1546 carried two vanA plasmids (<30 kb in both isolates and megaplasmids of 170 and 200 kb) that did not hybridize with any of probes used in this study. In summary, almost all IO isolates showed the presence of 100 kb vanA plasmids with 17 pRUM and pilA genes, while in IH the diversity of vanA plasmids was higher, with pRUM-like replicons of both 50 kb and 100 kb Information concerning other than vanA plasmids of E. faecium that was obtained during the study is summarized in Table 3.

Discussion
This study provides the molecular characteristics of VREfm outbreak isolates with the special focus on the role of  Letter code of each isolate corresponds to the designation used in Figure 2; nd: not determined; two isolates with the prototype Tn1546 marked with an asterisk; x,y isolates from the same patients "X" and "Y"; # isolates positive in conjugation.  Figure 3: PFGE of S1-digested total DNA of selected 27 VREfm isolates, visualized by ethidium bromide staining (a) and subjected to Southern hybridization with the following probes: vanA (b), 17 pRUM (c), axe-txe (d), pLG1 (e), pilA (f), aac(6 )-Ie-aph(2 )-Ia (g), 18 pEF418 (h), 2 pRE25 (i), and --(j). Lanes A-a, isolates designation as described in Table 3.
MGE, such as Tn1546-type transposons and vanA plasmids, acting as mediators of vancomycin resistance transfer. The investigated group of isolates originated from two hospitals, The Institute of Oncology and The Institute of Hematology and Transfusion Medicine in Warsaw, where two VREfm outbreaks occurred concomitantly. Immunocompromised patients of oncological and hematological wards are known to be of special risk for VRE colonization and infection [45]. Such susceptibility was especially evident during the outbreak in the IH, where three bloodstream infections caused by VRE were reported. The proximity of the two hospitals in the city and the simultaneous emergence of both outbreaks, which lasted for a few months, suggested the possibility of VRE transmission between the two institutions, although the investigation of available medical documentation, done independently in IO and IH, revealed no obvious routes, such as patient transfer between the two hospitals or from the same third hospital or utilization of common diagnostic equipment just before or during the outbreak period. The involvement of hospital personnel in VREfm transfer was also excluded.
Molecular typing methods, such as PFGE analysis and MLVA, have shown a divergent clonal structure of isolates. Such a situation is typical for VanA hospital outbreaks [46][47][48][49]. MLST performed for representative isolates included all of them into the hospital-associated lineages 17, 18, and 78, formerly described as CC17 complex [3]. Isolates belonging to this meroclone display common features such as ampicillin and ciprofloxacin resistance, the prevalence of IS16, Efm , and ℎ Efm , and enrichment in microbial surface components recognizing adhesive matrix molecules (MSCRAMMs), including pili genes [26,50]. However, isolates from each institution showed some specific features, such as predominance of certain PTs/MTs and differences in distribution of virulence, resistance, and plasmid-specific genes. This observation would suggest the existence of two separate endemic subpopulations without much exchange of strains prior to the introduction of vanA-carrying MGE.
VanA phenotype in both outbreaks was associated, in the vast majority of cases, with an acquisition of the same specific variant of Tn1546 transposon with ISEf1 and a mutation in the vanA gene. Tn1546-type transposons show significant variability due to point mutations, deletions, and presence of various ISs [41,42,47,51], and thus analysis of transposon structure provides valuable epidemiological information for investigation of VRE outbreaks. The ISEf1 insertion in the vanX-vanY intergenic region was described previously for hospital VREfm in Portugal and Germany [47,51], and the mutation in the vanA gene was not reported before. The presence of the same type of Tn1546 in both outbreaks provides a strong indication of either a common source or transmission of VREfm between the two hospitals.
Hybridization studies on representative isolates revealed the presence of a 100 kb plasmid carrying the vanA determinant as well as rep17, typical for pRUM plasmid and the pilA gene in several isolates from both institutions. Most of these isolates readily produced transconjugants, suggesting that this plasmid might play the principal role in the outbreak. The observed concomitant transfer of erythromycin resistance is in agreement with the colocalization of the ermB gene on pRUM [52]. Other, frequently encountered vanA plasmid was 50 kb in size and also represented the 17 pRUM replicon; however, it lacked pilA and all isolates with this plasmid were negative in conjugation. The 50 kb plasmid was exclusively associated with isolates of MT159/ST78 and observed solely in IH. The recently emerged lineage 78 of the hospital-adapted E. faecium shows increased epidemic properties and plays an important role in HAIs [53]. Thus, strains of this lineage, harboring the 50 kb nonconjugative vanA plasmid, were likely to be spreading by efficient clonal dissemination during the outbreak in IH. Association of van determinants with pRUM-type plasmids was described also by others [7,10,54]. Both 100 kb and 50 kb plasmids lacked the axe-txe TA system genes, typical for pRUM [10,52], suggesting the possible common origin of these two plasmids. Further studies, based on whole plasmid sequencing, are indispensable to elucidate the possible evolution of these plasmids during the outbreak. Although 100 kb plasmids were found in two isolates of the same clonal characteristics from IO and IH (Table 3, isolates C and X), differences in the plasmid content do not allow us to indicate these isolates as a direct epidemiological link between two hospitals. In a few cases, the vanA gene was associated with other replicons, typical for pLG1, pRE25, and pEF418. Such vanA plasmids were observed also in other studies [7,10]. The prevalent distribution of pLG1 as well as its predominant presence on plasmids over 200 kb in size is also in agreement with earlier studies which showed that all VREfm megaplasmids with the defined replicon type were always pLG1-like [7,8]. The association of VanA determinants with various plasmids during one outbreak may be caused by the Tn1546 transposition among plasmids and/or plasmid recombination. The latter process may yield plasmids with more than a single rep gene, which was also observed in the current study, both for vanA-and other plasmids. The role of plasmid mosaics in the dissemination of Tn1546 among VRE was emphasized recently by Freitas et al. [7]. Finally, for some vanA plasmids and other plasmids the replicon types could not be established (13% and 34% of observed plasmids, resp.), indicating that the pool of E. faecium plasmids remains only partly explored [10] and that there is the need for further studies of these epidemiologically important elements.

Conclusions
Molecular analysis of VanA VREfm outbreaks revealed that Tn1546::ISEf1 elements associated with pRUM-like plasmids were the key mediators of vancomycin-resistant E. faecium dissemination among the investigated group of oncological/hematological patients. Horizontal gene transfer of the whole vanA plasmids and/or Tn1546 transposons in endemic populations of nosocomial E. faecium is suggested as the potential way of VanA phenotype spread in the analyzed outbreaks. The enrichment in different plasmid-associated genes, antimicrobial resistance, and potential virulence determinants in the investigated population emphasizes the impact of mobile genetic elements on the epidemiology and evolution of VREfm.