A novel method of consensus pan-chromosome assembly and large-scale comparative analysis reveal the highly flexible pan-genome of Acinetobacter baumannii

Abstract

Background: Infections by pan-drug resistant Acinetobacter baumannii plague military and civilian healthcare systems. Previous A. baumannii pan-genomic studies used modest sample sizes of low diversity and comparisons to a single reference genome, limiting our understanding of gene order and content. A consensus representation of multiple genomes will provide a better framework for comparison. A large-scale comparative study will identify genomic determinants associated with their diversity and adaptation as a successful pathogen.

Results: We determine draft-level genomic sequence of 50 diverse military isolates and conduct the largest bacterial pan-genome analysis of 249 genomes. The pan-genome of A. baumannii is open when the input genomes are normalized for diversity with 1867 core proteins and a paralog-collapsed pan-genome size of 11,694 proteins. We developed a novel graph-based algorithm and use it to assemble the first consensus pan-chromosome, identifying both the order and orientation of core genes and flexible genomic regions. Comparative genome analyses demonstrate the existence of novel resistance islands and isolates with increased numbers of resistance island insertions over time, from single insertions in the 1950s to triple insertions in 2011. Gene clusters responsible for carbon utilization, siderophore production, and pilus assembly demonstrate frequent gain or loss among isolates.

Conclusions: The highly variable and dynamic nature of the A. baumannii genome may be the result of its success in rapidly adapting to both abiotic and biotic environments through the gain and loss of gene clusters controlling fitness. Importantly, some archaic adaptation mechanisms appear to have reemerged among recent isolates.

Fig. 1

Analysis of the A. baumannii pan-genome. The distribution of protein cluster sizes generated from the comparison of 249 A. baumannii genomes using PanOCT [35] indicates the number of singleton and core genes (a). The pan-genome size (left panel) and the number of novel genes discovered with the addition of each new genome (right panel) were estimated for all 249 genomes (b) and a set of 100 representative genomes identified by hierarchical clustering of all 249 genomes (c) using a pan-genome model based on the original Tettelin et al. model [42]. Purple circles are the median of each distribution (gray circles). Power law (red lines) and exponential (blue lines) regressions were plotted to determine α (open/closed status) and tg(θ), the average extrapolated number of strain-specific/novel genes, respectively [41]

Fig. 2

Pan-chromosome and fGIs. The core gene cluster (cGC), flexible gene cluster (fGC) and flexible genomic region (fGR) locations of the A. baumannii pan-genome (a) were computed from PanOCT output and are illustrated as a circle where each concentric circle is numbered from the outermost to the innermost circle (b). fGR locations are depicted in circles 1 (>20,000 bp), 2 (10,001–20,000 bp), and 3 (1000–10,000 bp) on a core backbone of genes in circles 4 (positive strand) and 5 (negative strand). Refer to the key for details on color representations, circle number and bar height. Key fGRs are noted by black numbering and letters K (K-antigen), N (Novel), and O (O-antigen) as in Table 2. Gray numbers indicate positions on the pan-chromosome in megabase pairs. Predicted functions are noted in gray words and the following abbreviations: RI resistance island, GI genomic island, ɸ phage. Genes associated with RI insertions are labeled in gray and indicated by a black line. The frequency of individual fGIs within six size class bins was also determined (c)

Fig. 3

Clone-type specific RI signatures and RI insertion frequencies over time. a–c Genomic locations of RI signatures specific to clonal complexes (CCs) 1–3 and ST 25. In CC1 isolates, RI insertions are detected in the comM and/or pho loci (a). In CC2 isolates, RI insertions are found in the comM locus, and in addition the astA, or acetylT locus. A novel RI insertion (composite IS26) was identified in CC3 isolates at the acyl-CoA synthase (acylCS) locus (b). Gene annotation of the composite IS26 RI is shown to the right. Drug resistance genes (green); immediate flanking genes acylCS and a transporter protein (dark orange). A novel GI was identified in ST 25 isolates at the acetylT locus (c). Gene annotation of the 7.8 kb GI is shown to the right. Salicylate monooxygenase (blue); immediate flanking genes acetylT and mdtL (dark orange). A cumulative frequency graph showing the number of RI-positive isolates carrying single, double, or triple RI insertions collected between 1950 and 2011 (d). Isolates that represent the first occurrence of a given RI signature are labeled on the graph (numbered 1–9)

Fig. 4

Virulence and fitness factors displaying variable gene content based on centroid-to-ortholog BSR. Specific genomic regions involved in the assembly of type I pili, siderophore production and efflux were highly variable and showed specific gain or loss of entire gene clusters in isolates analyzed. In the BSR-based heat map, the presence, absence, and low similarity of a protein ortholog compared with its centroid is shown in yellow (BSR = 1, presence), blue (BSR = 0, absence), and gray (low similarity or truncated), respectively. The year of collection is shown above strain names. Isolates collected prior to year 2000 are indicated with an asterisk. In summary, gene gain/loss events involved in virulence and survival were detected in decades-old isolates and appeared to have reemerged among recent isolates. The list of virulence genes analyzed and the full size version of the BSR-derived heat map are provided in Additional files 17 and 19, respectively

Fig. 5

Loss of pili cluster 1 (csu gene cluster) and gain of siderophore cluster 2 in specific isolates. a Two types of deletion were observed which led to a complete loss of the type I pilus csuAB-E gene cluster. A novel 17 kb deletion was detected in NIPH 335 and OIFC098, whereas a previously reported 42 kb deletion was found in MDR-ZJ06 and nine UH clade B isolates (e.g., UH6207). b Siderophore cluster 2 was detected only in a small subset of isolates across all 249 analyzed. Two apparently independent molecular events were observed among the siderophore cluster 2-positive isolates. In decades-old isolate ATCC 17978, insertion of the gene cluster was detected at a genomic position corresponding to 3.0 Mbp of the ACICU reference genome. c In the remaining siderophore cluster 2-positive modern isolates (e.g., MRSN 3405), insertion was detected at a different location, which corresponds to 3.8 Mbp of the ACICU reference genome. Since ATCC 17978 was isolated in 1951 while other isolates were isolated more recently between 2007 and 2011, the acquiring of siderophore cluster 2 among modern isolates could be an example of the reemergence of a survival mechanism of A. baumannii. The functional significance of siderophore cluster 2 is yet to be determined. Key: pairwise nucleotide identity shown in red to blue (100 % identity) color scale; contig breaks (pink vertical bars); open reading-frames (thick arrows); type I pilus cluster 1 and siderophore cluster 2 genes (green); deleted genes (gray scale); genes bordering insertions/deletions (dark orange and brown); other flanking genes (orange); other genes (light brown)