Genome-wide recombination drives diversification of epidemic strains of Acinetobacter baumannii

Abstract

Acinetobacter baumannii is an emerging human pathogen and a significant cause of nosocomial infections among hospital patients worldwide. The enormous increase in multidrug resistance among hospital isolates and the recent emergence of pan-drug-resistant strains underscores the urgency to understand how A. baumannii evolves in hospital environments. To this end, we undertook a genomic study of a polyclonal outbreak of multidrug-resistant A. baumannii at the research-based National Institutes of Health Clinical Center. Comparing the complete genome sequences of the three dominant outbreak strain types enabled us to conclude that, despite all belonging to the same epidemic lineage, the three strains diverged before their arrival at the National Institutes of Health. The simultaneous presence of three divergent strains from this lineage supports its increasing prevalence in international hospitals and suggests an ongoing adaptation to the hospital environment. Further genomic comparisons uncovered that much of the diversification that occurred since the divergence of the three outbreak strains was mediated by homologous recombination across 20% of their genomes. Inspection of recombinant regions revealed that several regions were associated with either the loss or swapping out of genes encoding proteins that are exposed to the cell surface or that synthesize cell-surface molecules. Extending our analysis to a larger set of international clinical isolates revealed a previously unappreciated ability of A. baumannii to vary surface molecules through horizontal gene transfer, with subsequent intraspecies dissemination by homologous recombination. These findings have immediate implications in surveillance, prevention, and treatment of A. baumannii infections.

Fig. 1.

Occurrence of different A. baumannii strain types during the NIHCC polyclonal outbreak. Isolates from patients were classified with pulsed-field gel electrophoresis. Three MDR strain types were observed (A, blue; B, red; and C, green) as well as non-MDR strains (other, gray).

Fig. 2.

Distribution of nucleotide differences among three outbreak strains with the genome of ACICU as reference. SNPs in the genome alignments of A, B, and C relative to ACICU were determined by using Mauve and displayed with Circos (39). Positions differing from ACICU (outer black circle) are indicated by colored marks in the three inner circles for A, B, and C. Different colors represent shared alleles among different sets of genomes. Blue marks represent a position in which A differs from ACICU, with a blue mark also occurring in the circles for B and/or C if either of them also share this variant. Red marks represent a position in which B differs from both A and ACICU, with a red mark also occurring in the circle for C if it shares this variant. Finally, green marks represent a position in which C differs from ACICU, A, and B. In the outer circle, gray regions represent sequence present in ACICU but not in all three of the other strains, which were therefore not considered for the SNP analysis. Such variable sequences include both truly variable regions and repetitive sequences that may be collapsed in the assemblies of A, B, and/or C.

Fig. 3.

Phylogeny of EC II strains based on SNPs in nonrecombinant regions. A maximum likelihood tree including the EC II strains, and rooted with EC I strain AB0057 (not shown for clarity), was constructed by using only polymorphisms that did not occur in the clusters observed in Fig. 2. Numbers on internal nodes are based on 100 bootstraps.

Fig. 4.

Variable gene content in the O-antigen biosynthetic clusters of the three dominant outbreak strains. The ∼30-kb regions surrounding the O-antigen biosynthetic gene clusters of strains A, B, and C are each represented by semicircles as marked. The colored bands on each semicircle indicate the Clusters of Orthologous Groups of Proteins (COG) functions of genes at the given position, with the COG/color relationship indicated in the key on the right. The gray ribbons represent sequences that can be aligned between a pair of genomes. The sequences flanking the O-antigen biosynthetic cluster align in all three genomes, but the biosynthetic clusters themselves contain unique genes.

Fig. 5.

Alignment of regions of variable gene content in EC II recombinant regions to other sequenced Acinetobacter genomes. Each column in the table represents a different Acinetobacter strain that has been sequenced either here (A, B, or C) or previously. Phylogeny among the strains (shown above the table) was determined based on SNPs in regions common to all genomes by using the neighbor-joining algorithm. Each row of the table corresponds to variable genes found in the recombinant regions (I, II, III, and IV in Fig. 2). A gray/white box at a given position indicates that the variable region, corresponding to the row, is present/absent in the genome, corresponding to the column. A region is considered present in a given genome if more than 80% of the region aligns to the genome, as determined by BLAST. The alignments are shown in SI Appendix, Fig. S3.