The current MLVA typing scheme for Enterococcus faecium is less discriminatory than MLST and PFGE for epidemic-virulent, hospital-adapted clonal types

Background MLVA (multiple-locus variable-number tandem repeat analysis) is a reliable typing technique introduced recently to differentiate also isolates of Enterococcus faecium. We used the established VNTR (variable number of tandem repeats) scheme to test its suitability to differentiate 58 E. faecium isolates representing mainly outbreaks and clusters of infections and colonizations among patients from 31 German hospitals. All isolates were vancomycin-resistant (vanA type). Typing results for MLVA are compared with results of macrorestriction analysis in PFGE (pulsed-field gel electrophoresis) and MLST (multi-locus sequence typing). Results All 51 but one hospital isolates from 1996–2006 were assigned to the clonal complex (CC) of epidemic-virulent, hospital-adapted lineages (MLST CC-17; MLVA CC-1) and differed from isolates of sporadic infections and colonizations (n = 7; 1991–1995) and other non-hospital origins (n = 27). Typing of all 58 hospital VRE revealed MLVA as the least discriminatory method (Simpson's diversity index 0.847) when compared to MLST (0.911) and PFGE (0.976). The two most common MLVA types MT-1 (n = 16) and MT-159 (n = 14) combined isolates of several MLST types including also major epidemic, hospital-adapted, clonal types (MT-1: ST-17, ST-18, ST-280, ST-282; MT-159: ST-78, ST-192, ST-203). These data clearly indicate that non-related E. faecium could possess an identical MLVA type being especially critical when MLVA is used to elucidate supposed outbreaks with E. faecium within a single or among different hospitals. Stability of a given MLVA profile MT-12 (ST-117) during an outbreak over a period of five years was also shown. Conclusion MLVA is a suitable method to assign isolates of E. faecium into distinct clonal complexes. To investigate outbreaks the current MLVA typing scheme for E. faecium does not discriminate enough and cannot be recommended as a standard superior to PFGE.


Background
Effective typing of microorganisms is a prerequisite for establishing epidemiological or phylogenetic links between corresponding isolates. A plethora of different methods has been successfully applied to type and differ-entiate bacterial strains and clonal groups from each other [1]. A critical point to all of these methods is their applicability to answer distinct questions ranging from investigation of outbreaks to establishing rather broad phylogenetic trees of relatedness and arrangement of strains within major clonal complexes. Each method has its respective weaknesses and strengths according to the question(s) addressed and the methodology behind [2][3][4][5][6][7].
Recently, a new method was introduced using small repetitive elements appearing in a variable number and distributed among the genome of a given species. Accordingly this technique based on a variable number of tandem repeats (VNTR) was named multiple-locus variablenumber tandem repeat analysis (MLVA; [8]). Initially MLVA was established to differentiate high-risk pathogens such as Bacillus anthracis and Francisella tularensis [9][10][11] but has been extended to a numerous number of other bacterial species and scientific questions [8,12] including outbreak investigations for pathogenic bacteria [4,13].
MLVA was applied also recently to type isolates of Enterococcus spp. Its discriminatory power was compared to MLST for a collection of 392 E. faecium [14] and to macrorestriction analysis in PFGE for 83 E. faecalis [15]. In both cases it was described that MLVA showed similar and rather concordant discrimination when compared to the respective reference method. Although the selected VNTRs were different between the two species, the overall conclusion would suggest MLVA as a typing method on one hand to discriminate highly enough between strains and on the other hand indicate the possibility to establish rather broad phylogenetic relatednesses. To support this hypotheses and to test the applicability of the established MLVA scheme for E. faecium to indicate and differentiate hospital-adapted clonal types appearing in increasing numbers among hospital patients worldwide [16], we investigated hospital isolates representing outbreaks and clusters of infections and colonizations from German hospital patients from the last 15 years using MLVA, SmaImacrorestriction analysis in PFGE, and MLST. A collection of 27 E. faecium from commensal, animal, and environmental origins was included for reasons of comparison and also typed by MLVA.

Macrorestriction analysis in PFGE
Altogether 58 E. faecium were investigated by SmaI-macrorestriction analysis originating from 31 German hospitals. Thirty-eight different PFGE types were assigned based on a 90 % similarity cut-off and recommendations described recently [17,18]. However, larger clusters of strains possessing at least related patterns (> 80 % identity) could be identified ( [18]; see below and Fig. 1).

MLVA
Altogether 14 MLVA types were assigned. All 51 group II and III isolates possessed MLVA types associated with clonal complex C1 of hospital-adapted types (Table 1). Cluster assignments of new MLVA types were done by the curator of the MLVA internet pages (J. Top, University Utrecht, Netherlands) based on pairwise similarities, UPMGA clustering and similarities within a minimum spanning tree of relatedness. Four group I VRE belonged to MLVA clonal complex C1, the other three each one to A, B or C (Table 1, see also [14]). The 27 isolates of different non-hospital origins showed ten MLVA types. All but two belonged to MLVA clonal complexes A or B (J. Top, pers. communication). Main type was MT-89 (n = 9; CC-A). Six isolates were none-typeable due to an incapability of an amplification of one of the six VNTR fragments.

Concordance of PFGE, MLVA and MLST results
Altogether 58 hospital VRE (groups I -III) were typed by all three molecular typing methods. SmaI-macrorestriction in PFGE revealed 38, MLST 19, and MLVA 14 different types. PFGE was the most discriminatory method with a discriminatory index of 0.972, MLST the median (0.91) and MLVA the least discriminatory (0.842) ( Table 2). Differences in discriminatory power between those different typing methods are even more obvious and pronounced when results involving only outbreak isolates of groups II and III are compared (PFGE: 0.965; MLST: 0.891; MLVA: 0.814). MLVA appears in all cases as the least discriminatory method which derives from the fact that (i) it revealed the least number of different types and in addition, (ii) more than half of all the investigated VRE belong to only two types MT-1 and MT-159.
Some isolates with identical MLVA types showed different MLST types (Table 1; Fig. 2). But also the opposite was found; isolates with identical MLST types could possess different MLVA types (Table 1 Results of MLST partly confirm cluster assignments based on PFGE data, whereas our data did not show a visible concordance between MLVA and PFGE (Fig. 1). Many MLVA types were distributed among different main PFGE clusters. Especially the most common types such as MT-1 (n = 16) and MT-159 (n = 14) were evenly distributed among different major PFGE branches. This fact is also reflected in the much lower concordance value for results of MLVA/PFGE (0.863) compared to MLST/PFGE (0.935; Table 2). Eight of nine ST-203 isolates cluster in a group of related PFGE patterns (> 83 % identity) together with two ST-282 VRE (ST-282 is a double-locus variant [DLV] of ST-203). A single ST-203 isolate clusters in a side branch but at least within a cluster of > 75 % identity. These eleven isolates possess three different MLVA types. All eleven ST-18 E. faecium cluster in a group with > 70 % identical patterns. However, also two ST-78 and a single ST-17 isolates belong to this cluster but are located at both edges of this branch. These altogether 14 VRE possess four    Comparison of phylogenetic relatedness between identified MLST (a) and MLVA types (b) for all 58 hospital VRE

Discussion
MLVA was introduced to molecular bacterial strain typing as a promising alternative to established typing standard methods such as AFLP, MLST and partly PFGE or as an alternative for microorganisms where other common typing techniques could not be applied due to several reasons [19]. Therefore it was not possible in all cases to compare its discriminatory power to established standards which is expected to vary also from genus to genus. For nosocomial pathogens, such as Staphylococcus aureus, MLVA was introduced as an alternative typing method showing matching results with macrorestriction typing in PFGE and clusters generated by spa-typing and MLST [4]. For E. faecium discriminatory power of MLVA was compared to MLST and AFLP (Amplified fragment length polymorphism) and found to be reliably sufficient as a suitable typing alternative [14]. But already when comparing data from its original description in detail, some uncertainties emerged challenging the overall applicability of this method for typing in general including especially investigation of outbreaks or following routes of disseminated outbreak strains. For instance, MT-1 combined 58 isolates represented by eleven STs [14]. Since MT-1 is the primary founder and one of the major MLVA types representing the cluster of hospital-adapted, clonal types (MLVA C1) its applicability to investigate single outbreak situations could be questioned. Indeed, this was one of the results of our study: Isolates representing several MTs such as the most common MT-1 and MT-159 neither cluster based on their SmaI-macrorestriction patterns in PFGE nor possess all identical or related MLST types ( Fig. 1 and 2).
As shown by our results, discrimination of E. faecium based on results of MLVA, MLST and PFGE does not always give concordant results as it was shown for rather clonal populations such as S. aureus [7] or for an analysis of E. faecium within a single hospital over a period of five years [20]. Concerning the latter a more detailed comparison of the results of Abele-Horn et al., and ours may be required. These, at the first moment rather contradictory results reflect major differences in the used strain collection (strains from a single hospital vs. 31 hospitals) and correspondingly resulting MLVA and MLST type diversity.
A collection of isolates from a single hospital may not be diverse enough to elucidate the non-congruencies we found here between some MLVA and MLST types (e.g., MT-159 absent in [20]). The discriminatory power of MLVA in strains recovered from patients at the University Hospital in general, i.e. not only associated with the described outbreaks, revealed a Simpson's diversity index of 0.846 [20] and was almost identical to what we calculated for our study population (0.842).
Clonal types might be affected by genomic rearrangements to a different extent. Whereas ST-18 or ST-203 isolates showed rather similar SmaI-macrorestriction patterns over time and geographical distribution, isolates of ST-17 were rather diverse. The ST-17 isolates were distributed among several subclusters or represent single PFGE patterns (Fig. 1). It could be speculated if isolates of ST-17 represent a rather ancient clonal type prone to recombination and thus divergent PFGE patterns for a longer time than ST-203 and ST-18 which could constitute rather recent hospital-adapted clones. Investigation of ST-17 from the early 1990ies as shown in our small collection of group I isolates and in a collection of bacteremia isolates from the mid to the late 1990ies (Werner, unpublished results) supports this hypothesis. The role of mobile and conjugative elements in rearranging bacterial genomes is well-known [21]. The genome sequence of E. faecalis V583 revealed one of the highest numbers of mobile, integrative and conjugative elements known so far from the bacterial world [22]. Similar numbers are suggested for E. faecium, however whole genome data for E. faecium are still incomplete. Recent comparative genomic hybridizations for a set of 97 E. faecium from different origins confirm the role of mobile and conjugative elements for shaping the E. faecium genome [23]. IS element driven recombination appeared higher in a subset of evolutionary linked epidemic E. faecium than in non-epidemic E. faecium (based on a comparison of typing results for MLST and PFGE). In this respect PFGE may be "overdiscriminatory" to a certain extent when applied to investigate socalled hospital clade or epidemic E. faecium [23]. Other factors and impacts that affect distinct clonal types in a different manner could also be discussed [24,25].
Clearly, MLVA has its advantages when applied to type E. faecium. It is a suitable method to identify strains belonging to the complex of hospital-adapted, epidemic clones (MLVA CC-1). The method is quick (intra-day results), easy to perform, comparably cheap, with excellent reproducibility (intra-and inter-laboratory) and allows data transfer and comparison. The comparably long repeat lengths of the VNTR loci used to type E. faecium make the assay suitable for agarose gel separation and therefore capable for many standard laboratories. Introduction of fluorescence-labelled primers in future may lead to a com-bined approach of six single reactions into one or two multiplex PCRs [26].
However, the advantage of comparably long repeat lengths may lead to an unintentional side-effect. MLVA was primarily established as a typing method to differentiate between strains. The VNTR repeats are, in general, very short and in a range of 6 to 30 bp [11,12,27]. The idea behind MLVA typing is mainly based on errors caused by the DNA polymerase resulting in slippage and/or recombination in a somehow expectable manner within these sequences. It could be speculated if repeat lengths of the sizes used to type isolates of E. faecium (120 -279 bp) may be too long to let these events appear in an essential frequency. So differences between strains would be less detected and the identified changes would only be suitable to establish phylogenetic links between strain types similar to or even less discriminatory than MLST. Indeed, some VNTR fragments showed length of incomplete repeats suggesting recombinational events within the distinct single repeat units. Assignment of VNTR repeat numbers might be rather erroneous in those cases. Analysis of tandem-repeat fragment length was recently reported to provide misleading results about the phylogeny and subspecies affiliation of Francisella tularensis isolates. This was found to be due to non-homologous, sequence-different VNTR fragments of identical lengths in unrelated isolates [28]. It should be emphasized again that the models predicting stability and flexibility of the distinct genomic repetitive structures used for VNTR analysis in E. faecium and other bacteria are rather theoretical and speculative.
Nothing is really known about the conditions affecting stability of these structures. Due to this uncertainty, establishing phylogenetic relatedness in appropriate trees by a link via single-locus or double-locus variants of MLVA profiles which might suggest an evolutionary link should be interpreted with caution.

Conclusion
Altogether

Bacterial strains
Majority of isolates (n = 43) originated from recent outbreaks and clusters of infections and colonizations (CCI) in German hospital patients between 2004 and 2006 (group III). Each isolate represents a single case/patient. We stick to the term CCI since outbreak means cases of infections. Our sample collection also includes cases of colonizations (stool samples) and cases of infections where the role of E. faecium is not clear (pneumonia, UTI). According to common outbreak definitions, a cluster is represented by n ≥ 2 cases which were epidemiologically linked. Isolates UW5248-UW5258 (n = 6) originated from a large private laboratory service provider in Southwestern Germany with a representative coverage of hospitals in five federal states (Labor Limbach, Heidelberg, Germany). Six representative isolates were chosen from a larger set of isolates representing clusters in five hospitals [29]. The other 37 group III E. faecium originated from 14 other German hospitals. Additional eight isolates representing each one an older outbreak sampled all over Germany between 1996-1999 were also included (group II; [30]). All seven but one (UW786) older hospital isolates represent sporadic infections or colonizations in hospital patients between 1991-1995 (group I; [31]). All hospital E. faecium were vancomycin-resistant and possessed the vanA gene cluster. For reasons of comparison 27 E. faecium isolates from non-hospital origins as well as the vanA reference strain BM4147 were also included [31,32].

DNA isolation
DNA was isolated by a column-based technique as recommended by the manufacturer (DNeasy Tissue Kit, Qiagen, Hilden, Germany). Standard protocol was slightly modified starting with an initial cell wall lysis step of 1 ml overnight culture centrifuged and resuspended in 400 μl TES buffer (10 mM Tris, 1 mM EDTA, 10 % saccharose) plus 10 mg/ml lysozyme and incubated at 37°C for 30 min.
Macrorestriction analysis in PFGE. Genomic DNA was isolated, digested with restriction endonuclease SmaI, and treated as described recently [29]. The agarose gel concentration was 1 %, the CHEF-DR III apparatus (BIO-RAD laboratories, Hercules, CA, USA) was used for PFGE. The ramped pulsed times were as follows: 1-11 sec for 15 h and 11-30 sec for 14 h at 14°C. Relatedness between banding patterns was calculated using a band based similarity coefficient (Dice) and UPMGA clustering (BioNumerics, Applied Maths, Sint-Martens-Latem, Belgium). A composite tree of all 58 hospital VRE of groups I -III resolved in 18 independent PFGE gels was generated the