The results of this study, which uses whole-genome comparative methods, demonstrate that the A. baumannii genome has the potential to accommodate a range of pathogenic phenotypes. This is the first study to use whole genome alignment and phylogenetic methods to classify A. baumannii. While methods such as MLST and PFGE can identify differences between related isolates, whole genome sequencing allows us unprecedented insight into the evolution and relatedness of cultured isolates. The variable nature of the A. baumannii pangenome [7, 19, 20, 58] suggests that additional whole genome sequence analysis will help to identify additional genes associated with antibiotic resistance and pathogenesis.
Our results demonstrated a wide variety of sequence types circulating within an individual hospital. With the PF8MLST typing system, UMB001 was assigned to ST2, which also contains the outbreak, multi-drug resistant isolate ACICU, and is one of the most commonly encountered STs. UMB003 was assigned to ST25 that contains outbreak strains isolated in Greece and Turkey . The UMB002 sequence type, ST16, is a much rarer sequence type; for example, out of a study of ~130 A. baumanniii isolates, only one, an outbreak isolate from the Netherlands, has the same allele profile . Each of these A. baumannii isolates was circulating within UMMC during the same limited time window, providing the potential for interaction and additional isolate-to-isolate horizontal gene transfer.
From a clinical perspective, the utility of typing a new isolate is uncertain and provides little, if any data that is utilized in patient treatment; however, genomics can provide insights that may impact clinical decisions. For example, a recent study examined at micro-evolution between three (AB056, AB058, AB059) phylogenetically similar clinical isolates , but found significant divergent susceptibilities to antimicrobials. This is highlighted in the current study in that included on the same phylogenetic branch and showing very little evolutionary distance are isolates expressing both MDR and susceptible phenotypes (Figure 1). This supports the mounting evidence that whole genome sequencing is necessary as a diagnostic tool, because it provides more detailed information than MLST or other typing methods, and it provides additional information about the genetic machinery necessary for antimicrobial resistance and pathogenesis. A recent study by Lewis et al.  utilized whole genome sequencing of A. baumannii for the investigation of a hospital outbreak in UK. This is one of the first uses of whole genome sequencing in an outbreak investigation or epidemiological setting. The fine scale detail analysis allowed by whole genome sequencing extended the findings of traditional epidemiological studies that were underway. The combination of traditional methodologies with next generation sequencing technologies will provide detailed analyses of these pathogens.
A. baumannii isolates have been identified from a wide range of human clinical sources including urine, blood, cerebrospinal fluid, and wounds [19, 20, 58]. One of the goals of this work was to identify if there were unique genetic features that could be attributed to a specific pathogenic phenotype. Bioinformatic analysis of available genomes tested the hypothesis that the gene repertoire of an invasion isolate (blood, urine) would be different than a colonization isolate (peri-anal, fecal matter, wound). In this study, peri-anal isolates were considered to be a colonization phenotype; surveillance from the peri-anal site has been shown in prior studies to represent the gastrointestinal flora (90% sensitivity, 100% specificity compared to stool samples) . Furthermore, studies have shown that bacteria found on peri-anal surveillance culturing often persist for prolonged periods [63, 64] and are associated with subsequent invasive disease .
Genetic features were identified that were exclusively present in genomes from each phenotype, which suggested that these genes encode proteins that aid in either A. baumannii invasion causing clinical infection or colonization. For example, a putative hemolysin gene was present in all colonization phenotype isolates (Figure 4), but was absent in all invasion phenotype isolates. A phylogenetic analysis demonstrated that the gene is present in ancestral Acinetobacter lineages and does not appear to be a recent introduction via horizontal transfer (Figure 4). However, when genes unique to each isolation phenotype were screened against a broader collection of isolates including colonization and invasion isolates, no correlation between phenotype and the pattern of gene presence/absence was observed. The design of the PCR-based assays was based on a limited number of isolates from each phenotype, and it's not surprising that the presence/absence pattern fails in a broader analysis. These findings demonstrate that results based solely on comparative genomic analysis of a limited number of genomes may not be broadly applicable to diverse culture collections, especially an organism with the genomic mosaicism of A. baumannii. The study highlights the difficulty in correlating clinical phenotypes with genotypes in some pathogens.
While the comparative analysis did not identify a consistent genotype associated with each of the isolation source groups, this study did identify a number of very interesting traits that highlight the genetic variability of A. baumannii and could potentially expand the survival of this pathogen in the hospital setting. For example, genes associated with phenol metabolism were exclusively found in isolate UMB002; the absence of these genes in other A. baumannii isolates suggests that these genes were acquired through horizontal gene transfer as they are similar to genes found in another Acinetobacter species isolated from oil refinery wastewater . However, it is possible that this isolate was selected for over time, as the primary surface cleanser at UMB contains nonoxynol, which is phenolic based, and an organism that would potentially be resistant to these types of cleaners would have an increased rate of transmission; this is supported by the fact that this complex is more frequently found in UMMC isolates, compared to the examined genomic space of the A. baumannii species. Additional experimental work is needed to verify that UMB002 can metabolize phenolic compounds, but the presence of these genes suggests a mechanism by which this isolate may persist and thrive on hospital surfaces.
In contrast to the variable genomic features highlighted here, we also identified a gene that encodes a large, conserved peptide that was found in all sequenced A. baumannii genomes; no homologous peptide was found in any other sequenced Acinetobacter species. A PCR screen demonstrated that this gene is present in A. baumannii isolates (~99% of isolates screened), suggesting that it can be used as a biomarker for the positive identification of clinical A. baumannii isolates. Additional population screening will help determine how well conserved this peptide is among a more diverse collection of clinical isolates.