A Trilocus Sequence Typing Scheme for Hospital Epidemiology and Subspecies Differentiation of an Important Nosocomial Pathogen, Enterococcus faecalis^{▿ †}

PCR, sequencing, analysis, and allele and ST assignment.

Genomic DNA was isolated from a single-colony culture after overnight growth in 5 ml of brain heart infusion broth (Becton Dickinson, Sparks, MD), using a commercial kit (DNeasy tissue kit; Qiagen, Valencia, CA). The intragenic sequences of the ace, salA, and lsa genes were amplified and sequenced using the primers listed in Table S2 in the supplemental material. Due to observed sequence heterogeneity in the forward primer region of salAF, salAF1 was used for isolates TX2148, TX2137, TX2142, TX2134, and TX4245 (see Table S2 in the supplemental material). PCR mixtures contained the following in a 25-μl volume: 0.5 μl of template DNA, 0.5 μl of forward primer (25 pmol/μl), 0.5 μl of reverse primer (25 pmol/μl), 2.5 μl of deoxynucleoside triphosphates, 5 μl of optimized buffer B (1× buffer is 60 mM Tris-HCl [pH 8.5], 15 mM ammonium sulfate, 2 mM MgCl₂) obtained from Invitrogen (Carlsbad, CA), 0.1 μl of Invitrogen Taq DNA polymerase (5 units/μl), and 15.9 μl of distilled H₂O. PCR was performed with an initial denaturation at 94°C for 2 min, followed by 32 cycles of 94°C for 15 s, 55°C for 15 s, and 72°C for 1 min and a final extension of 72°C for 7 min.

The PCR amplicons were purified using a QIAquick PCR purification kit (Qiagen, Valencia, CA). Sequencing of the amplicons was performed using an Applied Biosystems BigDye Terminator V3.1 cycle sequencing kit (Foster City, CA). Each allele was sequenced from two independent amplicons, using the primers listed in Table S2 in the supplemental material. The sequences were resolved using an ABI Prism 3730 and/or Prism 3130 genetic analyzer. Sequences were assembled using DNASTAR software's SeqMan program (Lasergene, Madison, WI). Sequence types were determined by allelic variations in the three loci (no weight was given to the degree of sequence divergence between alleles). The presence of one or more nucleotide changes was defined as indicating a distinct allele type, which was assigned a number arbitrarily. An ST^T was defined as the combination of the types for the three alleles. In all cases except one (TX4241), the ST^T number assigned was chosen to match the previously assigned/published ST^M sequence type. TX4241 had an identical ST^T to another isolate, TX4251, although the ST^M differed, and therefore was assigned the ST^T of isolate TX4251. To determine the overall divergence of the sequenced gene fragments of the three loci, these sequences were then spliced together to obtain a composite, concatenated sequence for each isolate. Sequence alignments were performed by the Jotun Hein method, using the MegAlign program of DNASTAR software (Madison, WI).

PFGE and MLST.

A previously described method for PFGE was used, with some modifications (12, 29). Agarose plugs containing genomic DNA were digested with SmaI (NEB, Ipswich, MA), and electrophoresis was carried out using a clamped homogenous electrical field (CHEF-DRIII device; Bio-Rad Laboratories, Richmond, CA), with ramped pulse times beginning at 2 s and ending with 50 s, at 200 V for 23 h. The PFGE patterns were interpreted using the criteria suggested by Tenover et al. (48). MLST based on seven housekeeping genes was performed using the criteria and primers for E. faecalis designated by Ruiz-Garbajosa et al. (41). Sequence types were assigned in accordance to the database available at http://efaecalis.mlst.net/.

Establishment of TLST.

For the establishment of TLST, 50 isolates of 42 diverse ST^M collected in nine different countries and representing human and animal sources were chosen based on their seven-housekeeping-gene ST^M. For the majority of the isolates, the lengths of the loci analyzed for allele assignment for ace, salA, and lsa were 828 bp (from bp 197 to 1024 of the complete open reading frame), 863 bp or 866 bp (from bp 89 to 954 or 957, due to a 3-bp in-frame deletion [positions 940 to 942] present in 52% of the allele types), and 868 bp (from bp 270 to 1137), respectively. The exceptions were salA of TX2136 (due to a 15-bp in-frame insertion between positions 688 and 689) and lsa of TX4254 (due to a 1-bp insertion between positions 210 and 211). Note that the frameshift mutation in lsa of TX4254 resulted in susceptibility to quinupristin-dalfopristin (Etest MIC = 1 to 1.5 μg/ml); this finding is consistent with published reports on other E. faecalis strains (7, 45, 47). The total number of different allele types identified for 50 isolates was 27, 27, and 33 for ace, salA, and lsa, respectively. As shown in Tables S3 to S5 in the supplemental material, the number of variable sites in each locus ranged from 65 (salA) to 71 (lsa).

Comparison of TLST to MLST.

The discriminatory abilities of MLST and TLST were first compared by the numbers of unique and related ST^M and ST^T determined by each method. The 50 isolates used for the development of the TLST scheme, representing 42 distinct ST^M, were differentiated into 41 different sequence types by TLST (Table ​(Table1).1). For 36 isolates representing singletons (a single isolate representing one ST) by MLST, 35 ST^T were found. Thirty-four of the 36 were distinguished as singletons by TLST, and the remaining two MLST singletons (ST^M-58 and ST^M-99) were represented by a single ST^T (ST^T-58). Closer examination identified that ST^M-58 is a single-locus variant (SLV) of ST^M-99 differing in the yqiL locus by one nucleotide; thus, the grouping of these two isolates into one ST^T resulted because they were indeed closely related. The remaining 14 isolates represented small groups of 2 to 4 isolates belonging to six ST^M. All of the isolates in each of the six groups were differentiated identically by both TLST and MLST. In conclusion, 49 of the 50 isolates were typed similarly between MLST and TLST. The remaining isolate, TX4241 (which had an identical ST^T to that of TX4251), was only 1 bp different from TX4251 by MLST and thus was also closely related by MLST. Furthermore, there were no differences between Simpson's indexes of diversity for TLST (D = 99.0%; confidence interval, 0.981 to 1.0) and MLST (D = 99.1%; confidence interval, 0.981 to 1.0), suggesting that the two methods have very similar discriminatory abilities.

To compare the congruence between type assignments by MLST and TLST, the adjusted Rand (4, 16) and Wallace (4, 55) coefficients were also calculated. This showed that the probability that a pair of strains which were classified as the same type by MLST were also classified as the same type by TLST was 100% (based on the Wallace coefficient) and the probability of two strains having the same ST^T also sharing the same ST^M was 91.7% (as stated above, a single ST^T corresponded to two ST^M that were SLVs that differed by a single nucleotide). The adjusted Rand index (compares the clustering/grouping of isolates sharing similar characteristics according to a given method) for MLST and TLST in this study was 0.956, meaning that the concordance between the two methods was extremely good.

Next, since a previous study had suggested that virulence genes may evolve too rapidly to be useful for assessing the similarity of strains due to evolution from a common ancestor (41), we looked further at evolutionary relationships. It is important that accurate interpretations of clonal relationships between SLVs and double-locus variants (DLVs) by TLST cannot be assessed because of the limited number of loci in the scheme; therefore, we used the percent identities calculated from the concatenated sequences of all three genes (total sequence divergence as opposed to allelic differences) for this analysis. As shown in Fig. ​Fig.11 and Fig. S1 in the supplemental material, the evolutionary relationships (clustering and branching between isolates/ST) defined between two different ST were in agreement at times and differed substantially at other times between the two techniques. For example, ST^M-21/ST^T-21 and ST^M-5/ST^T-5 were separated by a single branch in the tree created via the unweighted-pair group method using average linkages (UPGMA) by both techniques (Fig. ​(Fig.1;1; see Fig. S1 in the supplemental material). ST^M-21 and ST^M-5 are SLVs by MLST (99.9% identical by concatenated sequence), and ST^T-21 and ST^T-5 are also highly identical by TLST (99.8%), differing by only one allele in both cases. On the other hand, ST^M-40 was several branches away from ST^M-27 in the MLST UPGMA tree (see Fig. S1 in the supplemental material) due to differences in five of the seven allele types (99.5% identical); however, ST^T-40 and ST^T-27 were separated by only a single branch (differing by one allele) in the TLST UPGMA tree (Fig. ​(Fig.1)1) and were 99.8% identical by the concatenated TLST genes. This may be indicative of homoplasy (similarity due to convergent evolution) of the genes in the TLST scheme, which implies that this scheme is more suitable for the study of hospital epidemiology than long-term epidemiology.

Open in a separate window

FIG. 1.

UPGMA tree generated from the TLST allelic profiles of 50 isolates. The phylogenetic tree was based on the matrix of pairwise differences in TLST allelic sequences, as determined by UPGMA, and was drawn using the sequence type analysis and recombinational tests, version 2 (START2), program (http://pubmlst.org/software/analysis/start2/).

Correlation between PFGE and TLST.

To compare the discriminatory abilities of TLST and PFGE, we studied 10 isolates belonging to four pulsotype groups (which differed by fewer than six bands in each group), as identified by published PFGE results (Table ​(Table2).2). Four different ST^T, corresponding to the different PFGE types, were identified, and the isolates within each PFGE group had identical ST^T (i.e., 100% identical concatenated sequences), thus suggesting that TLST's usefulness in the nosocomial setting is similar to that of PFGE (Table ​(Table2).2). In contrast, the sequence divergence between the four different PFGE/ST^T groups for the concatenated sequences ranged from 1.1 to 2.4% (27 to 60 bp different). Also, as previously noted, PFGE is often not suitable for determining long-term epidemiological relationships, which can be identified by sequence typing techniques. For example, isolates which were SLVs or DLVs by MLST were generally considered to be in the same clonal complex. However, as shown in Fig. ​Fig.2,2, isolates belonging to different ST^M but considered clonally linked by MLST had differences of more than six bands by PFGE, and consequently the clonal relationships were not widely discernible by this method. Thus, TLST appears to be a fitting alternative for use in hospital epidemiology.

Open in a separate window

FIG. 2.

PFGE of isolates which were SLVs or DLVs by MLST. Lanes: a, TX2783; b, TX2148; c, TX0645; d, TX4243; e, TX2621-2; f, V583; g, TX4239; h, TX2141; i, TX4249; j, TX4259. Black brackets indicate SLVs by MLST, and gray brackets indicate DLVs by MLST.

Application of TLST to a hospital collection.

We next applied the TLST scheme to representative isolates of a recently published hospital collection of 48 vancomycin-resistant E. faecalis strains previously typed by PFGE and MLST (Table ​(Table3)3) (1). By PFGE, these isolates were all considered to be a related clone (seven PFGE subtypes with ≤4 band differences) which had disseminated in six hospitals in Bogota, Colombia. The TLST scheme was applied to a representative isolate of each PFGE subtype as well as to additional isolates to represent each isolation year (Table ​(Table3).3). By TLST, all isolates typed were ST^T-2, which matched the previously characterized isolate TX2486 (representing the index isolate for the HV1 clone), an outbreak VanB-type E. faecalis strain isolated in Houston, TX, in 1994. These results are in agreement with the published PFGE and MLST results, which suggested a link between the U.S. strain and this Colombian outbreak clone (1).

We thank K. V. Singh for his advice and for help with MIC determination. We thank L. S. Jensen (Denmark), T. Coque (Spain), R. del Campo (Spain), and E. Sadowy (Poland) for sending isolates.

This work was supported in part by an NIH grant (R37 AI 47923 to B.E.M.) from the Division of Microbiology and Infectious Diseases, National Institute of Allergy and Infectious Diseases (NIAID). C.A.A. is supported by a K99/R00 Pathway to Independence award (1K99-AI72961) from NIAID. R.J.L.W. is supported by the European Union Sixth Framework Program under contract LSHE-CT-2007-037410. This work was carried out in a facility supported by NIH (CTSA) grant UL1 RR024148.