Characterization of Epidemiologically Unrelated Acinetobacter baumannii Isolates from Four Continents by Use of Multilocus Sequence Typing, Pulsed-Field Gel Electrophoresis, and Sequence-Based Typing of bla_OXA-51-like Genes

Bacterial isolates.Forty-four isolates of A. baumannii from 22 countries, including three standard representatives of the endemic pan-European clonal lineages I, II, and III, were included in this study. All isolates were identified as A. baumannii as described previously (8) and by the presence of a bla _OXA-51-like gene (45). All the isolates were recovered from cases of invasive disease in various hospitals between 1982 and 2006.

PFGE.PFGE was performed essentially as described previously (27), but with minor modifications. The agarose-embedded bacterial genomic DNA was digested with ApaI (Promega, Southampton, United Kingdom) at 37°C overnight. Electrophoresis was performed on agarose (1%, wt/vol) gels with 0.5× Tris-borate-EDTA buffer. The following PFGE parameters were applied: voltage of 6 V/cm, initial switch time of 5 s, final switch time of 35 s, and run time of 24 h. Electrophoresis was performed in a CHEF DRII apparatus (Bio-Rad, Hemel Hempstead, United Kingdom). The stained gels were scanned using the Diversity Database software image-capturing system (Bio-Rad). The Dice coefficient was used to calculate similarities, and the unweighted-pair group method using average linkages (UPGMA) was used for cluster analysis with BioNumerics software, version 4.0 (Applied Maths, St-Martens-Latem, Belgium).

MLST analysis.DNA was purified using a Puregene DNA purification system genomic DNA purification kit (Gentra Systems, Minneapolis, MN). PCRs for the seven housekeeping genes, gltA, gyrB, gdhB, recA, cpn60, gpi, and rpoD, were performed in 50-μl volumes containing 10 μl 5× Green GoTaq Flexi buffer, 1.5 mM MgCl₂, 800 μM PCR nucleotide mix, and 1.25 U GoTaq DNA polymerase (Promega). The primers and PCR conditions were those described by Bartual et al. (1), except that the annealing temperatures were modified to 45°C for gyrB and rpoD, to 56°C for recA, and to 58°C for cpn60 and gpi.

A Px2 thermal cycler (Thermo Fisher Scientific, Waltham, MA) was used for all PCRs. Analysis of the PCR products was on agarose (1.5%, wt/vol) gels, followed by staining with ethidium bromide and visualization using the Diversity Database software image-capturing system (Bio-Rad). Following purification using a QIAquick PCR purification kit (Qiagen, Crawley, United Kingdom), the products were sequenced in both directions on a 3730 DNA analyzer (Applied Biosystems, Warrington, United Kingdom). The resulting sequences were analyzed using the online BLAST (http://www.ncbi.nlm.nih.gov/BLAST/ ) and MultAlin (http://bioinfo.genopole-toulouse.prd.fr/multalin/ ) software. The isolates were then assigned to sequence types (STs) using the tools on the A. baumannii MLST webpage (http://pubmlst.org/abaumannii/ ).

(i) Tree congruence.Neighbor-joining trees for all identified alleles of the seven genes were constructed in Geneious (version 2.0.10) software using the Jukes-Cantor genetic distance model, with statistical support for the nodes being assessed via the bootstrap resampling method with 1,000 resamplings. Comparisons of the tree topologies were conducted visually, and differences were confirmed using the quartet measure of tree-to-tree distances implemented in COMPONENT (version 2.0) software (http://taxonomy.zoology.gla.ac.uk/rod/cpw/index.html ).

(ii) PHI test.The pairwise homoplasmy index (PHI) test was used to investigate the possibility of recombination among the alleles within the seven housekeeping genes (2) and was implemented in SplitsTree4 software (http://www.splitstree.org/ ).

(iii) Phylogenetic reconstruction.The phylogeny for the 44 isolates was estimated via the neighbor-joining method using Geneious (version 2.0.10) software and the Jukes-Cantor genetic distance model, with statistical support for the nodes being assessed via the bootstrap resampling method with 1,000 resamplings. Concatenated sequences comprising the seven loci (gltA, gdhB, rpoD, cpn60, recA, gyrB, and gpi) and five loci (gltA, gdhB, rpoD, cpn60, and recA) were utilized (see Results).

(iv) eBURST analysis.eBURST analysis was used to investigate the evolutionary relationships and clonal complexes (CCs) within the isolates, using the software on the eBURST website (http://eburst.mlst.net/v3/enter_data/single/ ), with statistical support for the complexes being assessed via the bootstrap resampling method with 1,000 resamplings (12). Analysis was performed by the use of both stringent (a minimum of six shared alleles) and relaxed (a minimum of five shared alleles) grouping parameters.

(v) I _A^S.The standardized index of association (I _A ^S) for the seven loci was calculated using START2 software (http://pubmlst.org/software/analysis/start2/ ) and 1,000 iterations.

Sequence-based typing of bla _OXA-51-like genes (SBT-bla _OXA-51-like genes).Genomic DNA was extracted by boiling two to three colonies for 10 min in 50 μl sterile distilled water. The PCR mixtures contained 20 mM Tris-HCl (pH 8.8), 10 mM KCl, 10 mM (NH₄)₂SO₄, 2 mM MgSO₄, nuclease-free bovine serum albumin (0.1 mg/ml), 0.1% Triton X-100, 1.5 mM MgCl₂, 800 μM PCR nucleotide mix, and 1.25 U of Pfu DNA polymerase (Promega) in a total volume of 50 μl. Primers OXA-69A and OXA-69B (16), external to the bla _OXA-51-like gene, were used to amplify the entire sequence under the following conditions: 95°C for 2 min, followed by 30 cycles of 95°C for 1 min, 48°C for 40 s, and 72°C for 3 min and then 72°C for 6 min. Primers were used at a final concentration of 0.25 μM, and reactions were performed with 0.5 μl crude DNA template. PCR product analysis and sequencing were performed as described above for the MLST products.

PFGE.Six small clusters of isolates, each with a similarity of >80%, were identified (Fig. 1). The isolates forming each cluster were considered to belong to the same epidemic lineage. The remaining isolates had PFGE profiles with similarities of <80% and were considered to be unrelated.

• Open in new tab
• Download powerpoint

FIG. 1.

Dendrogram constructed following determination of PFGE profiles using UPGMA. The OXA-51 type, the ST, and the MLST allelic profiles of seven housekeeping genes are shown for each isolate. ID, isolate identifier; ^, novel allelic profile; *, novel sequence.

MLST analysis.Seven housekeeping genes, gltA, gdhB, recA, cpn60, rpoD, gyrB, and gpi, were amplified and sequenced for each isolate. Twenty-four different STs were identified, of which 19 (designated ST33 to ST51) were novel, and five novel alleles, officially named gdhB21, gdhB22, recA16, rpoD22, and gpi26, were identified (Fig. 1).

(i) Tree congruence.Comparison of the tree topologies for all identified alleles of the seven genes revealed inconsistencies with the trees for gyrB and gpi in relation to one another and to the other five trees. The topologies of the trees for the gltA, cpn60, gdhB, recA, and rpoD sequences were broadly consistent, with the same isolates being grouped together in the same major clades with few exceptions. However, the gpi tree splits the 12 isolates usually grouped with isolate A297 (the representative of European clone I) across two major clades, while the gyrB tree splits the 17 isolates usually grouped with isolate A320 (the representative of European clone II) across two major clades. This observation was confirmed via the quartet measure of tree-to-tree distances (7). This method measures the similarity between two trees by breaking them down into quartets, subtrees containing just four isolates. By analyzing whether all possible quartets across both trees are resolved identically to or differently from one another, it can be determined whether the topologies of the two trees are similar or not. For the gyrB and gpi trees, the number of differently resolved trees was greater than the number of similarly resolved trees, in contrast to the results for the other five genes (Table 1), confirming the visual observation that the topologies of these two trees are inconsistent with those of the other five genes.

(ii) PHI test.The PHI test utilizes DNA sequence data and infers whether patterns of nucleotide polymorphisms are consistent with a model of vertical transmission (clonal population structure) or not. Instances where they are not may indicate recombination events (2). Analysis of alleles for all seven genes with the PHI test detected statistically significant (P = 0.0036) evidence for recombination within the alleles for gyrB but not in gpi.

(iii) eBURST analysis.eBURST analysis of the allelic profiles for all seven loci by the use of both stringent and relaxed grouping parameters produced the same result (Fig. 2 A), revealing four CCs and nine singletons. The largest clonal complex, CC1, contained ST34, which was identified as a potential founder, with ST46, ST36, ST40, ST4, and ST22 radiating from it. A second major clonal complex, CC2, contained ST49, ST51, ST47, and ST25 surrounding ST33, which was identified as a potential founder for the complex. The other two minor clonal complexes identified contained ST35 and ST38 (CC3) and ST50 and ST48 (CC4), respectively. Analysis using only five loci (Fig. 2B) reduced the number of different allelic profiles such that all of the STs grouped in CC1 in Fig. 2A were identical to ST34; all of the STs, with the exception of ST25, grouped in CC2 were identical to ST33; and the two STs in each of CC3 and CC4 were also identical.

• Open in new tab
• Download powerpoint

FIG. 2.

Results of eBURST analysis used to assign CCs within the 44 isolates of A. baumannii utilizing seven loci (A) and five loci (B). (A) The CCs are circled, and the predicted clonal ancestors for CC1 and CC2 are shown by the central points. (B) ST33 encompasses ST33, ST37, ST49, ST51, and ST47; ST34 encompasses ST34, ST4, ST36, ST40, ST46, and ST22; and ST35 includes ST35 and ST38, while ST48 covers ST48 and ST50. The sizes of the points are proportional to the number of isolates assigned to each ST.

(iv) I _A ^S. I _A ^S is a measure of how clonal a population is and can be calculated using the allelic profiles generated by MLST. In a completely clonal population, alleles will be in linkage disequilibrium and the I _A ^S approaches 1. Under panmixis, linkage equilibrium will be observed and I _A ^S approaches 0. Values for I _A ^S that differ significantly from 0 are considered to indicate a clonal population structure (26). Analysis of the entire data set of 44 isolates yielded an I _A ^S value of 0.4907 (P < 0.001). This was decreased to 0.3569 (P < 0.001) when only one representative of each sequence type was included. Both values are significantly greater than 0 and indicate that the population is clonal.

(v) Phylogenetic reconstruction.Neighbor-joining phylogenies for the 44 isolates were estimated using concatenated sequences for all seven loci (Fig. 3 A) and for five loci (gyrB and gpi were excluded) (Fig. 3B). The seven-locus phylogeny was not supported by very good percentage support values for the nodes, and isolates assigned to the same CCs by eBURST analysis were quite divergent and, in the case of CC2, occupied completely separate halves of the tree. As described above, there was evidence for horizontal gene transfer within the gyrB and gpi alleles and noncongruence between the individual locus trees. In order to assess the effect that the inclusion of the gyrB and gpi sequences was having on the tree topology, a neighbor-joining phylogeny for the 44 isolates was estimated using concatenated sequences for the other five genes only (Fig. 3B). The phylogeny was well supported with high percent support values for the nodes. The majority of isolates fell into one of three distinct and highly related clades: 14 isolates were grouped with the European clone II strain, 12 isolates were grouped with the European clone I strain, and 4 isolates were grouped with the European clone III strain. There was complete agreement between the CC groupings by eBURST analysis and the clades in the five-locus tree.

• Open in new tab
• Download powerpoint

FIG. 3.

Neighbor-joining phylogenetic tree based on seven (A) and five (B) housekeeping genes. The corresponding OXA-51-like enzymes that the isolates encode and their sequence types are shown beside each isolate number. The CC that isolates were assigned to by eBURST analysis is indicated. Branches are labeled with percent support. OXA-69*, the OXA-69 enzyme encoded by a bla _OXA-69 gene containing 5 silent nucleotide substitutions.

SBT-bla _OXA-51-like genes.Genes for 12 different OXA-51-like enzymes were identified, of which OXA-66 (n = 11), OXA-69 (n = 6), and OXA-71 (n = 5) constituted three major groups (Fig. 1). OXA-51 and OXA-107 were each found in three isolates, while OXA-83 was found in two isolates. OXA-68, OXA-82, OXA-92, OXA-109, and OXA-112 were found in a single isolate each (Fig. 1). As reported previously (8), representatives of the three major European clonal lineages were found to encode enzymes representative of the major groups, with clone I encoding OXA-69, clone II encoding OXA-66, and clone III encoding OXA-71.

Evaluation of SBT-bla _OXA-51-like gene data and its correlation with MLST and PFGE data.The SBT-bla _OXA-51-like gene data correlated well with the MLST data, with respect to the identification of the major European lineages. The 14 isolates that grouped with the European clone II isolate each encoded closely related OXA-51-like enzymes of the OXA-66 group (Fig. 3A). Similarly, the 12 isolates that grouped with the European clone I isolate each encoded closely related members of the OXA-69 group of OXA-51-like enzymes, while the 4 isolates that grouped with the European clone III isolate each encoded OXA-71. A few exceptions were noticed in which isolate STs and OXA gene sequences were not consistent. Isolates A187 and A483, which had the same ST, ST20, encoded different OXAs, OXA-68 and OXA-51 (8 amino acid differences); and isolate A92, which had an OXA-69 enzyme, was not grouped with the other OXA-69-encoding isolates. However, the bla _OXA-69 gene in A92 contained 5 silent nucleotide substitutions (G426 → A, C474 → A, C511 → T, G540 → A, and T801 → C), suggesting that this isolate may be quite different from the other OXA-69-encoding isolates. For the three major epidemic lineages, the SBT-bla _OXA-51-like gene data were in concordance with the MLST-derived phylogeny for 93% of the isolates related to European clone I, 94% of the isolates related to European clone II, and 100% of the isolates related to European clone III. In contrast, typing by PFGE was not always in concordance with the SBT-bla _OXA-51-like gene data or the MLST data. There were isolates with the same PFGE type (types A4, A6, and A24) that shared the same OXA-51-like gene, whereas others had the same ST but a different PFGE type (e.g., isolates A25 and A4 and isolates A6 and A24) (Fig. 1).