The University of Maryland Medical Center (UMMC) is a 656-bed tertiary-care hospital located in Baltimore, Maryland, with a 19 bed surgical intensive care unit (SICU) and a 29 bed medical intensive care unit (MICU) that provides care to adult patients who have acute or potentially life-threatening medical conditions. This study utilized an existing and ongoing prospective cohort of adult patients admitted to the SICU and MICU at the UMMC from January 2008 to June 2008 who had routine (admission, weekly and discharge) peri-anal surveillance cultures obtained 27 . Bloodstream isolates were obtained from cohort participants with an imipenem-resistant A. baumannii blood stream infection (BSI). Peri-anal samples for culture were obtained using Staplex II cotton swabs (Staplex, Etobiocoke, Ontario, Canada), which are used to swab the peri-anal area in a circular motion moving from the rectum out 27 . Surveillance swab cultures were stored by placing swabs in tryptic soy broth containing 30% glycerol and stored at -80°C to allow for future recovery of isolates 28 .

Clinical isolates (blood and wound) were isolated from UMMC's microbiology laboratory using standard procedures. Peri-anal swabs were cultured for A. baumannii in the research laboratory by streaking swabs onto MacConkey agar (Remel, Lenexa, KS) supplemented with 6 µg/ml of imipenem and incubated at 37°C for 24 to 48 hours. Recovery of organisms from archived frozen peri-anal specimens has been shown to be 98% effective 28 . Non-lactose fermenting organisms (white colonies on MacConkey agar) were further identified as A. baumannii using API NE 20 Identification Strips (BioMérieux Inc.; Hazelwood, MO, USA) or Vitek II (BioMerieux, Durham, NC).

We selected one isolate from blood, peri-anal swab and a wound for further characterization. These three isolates selected were all isolated during a time period in 2008 where a localized outbreak of A. baumannii was ongoing at the UMMC. UMB001 was a virulent outbreak strain isolated from the blood of a patient in the SICU; UMB002 was a strain isolated from a peri-anal sample that was not part of the outbreak but was isolated at the same time in the MICU; UMB003 was also present at the time of the outbreak, but was isolated from a wound sample and was not related to the outbreak strain. In addition to the isolates selected for genome sequencing, a collection of 74 A. baumannii isolates was used for the polymerase chain reaction (PCR) screening of genomic markers specific to either invasive or colonization isolates (see below). This collection included 20 randomly sampled peri-anal surveillance isolates collected as part of an ongoing cohort of intensive care patients; 18 isolates were collected from February to May 2008 and 2 isolates were historical isolates (2006 and 2007) from patients later known to have A. baumannii bloodstream infections in 2008. The collection also included 54 invasive isolates from blood infections at UMMC collected from 9/2008-6/2010.

Antimicrobial susceptibility testing was performed on all isolates by the Kirby-Bauer disk diffusion method. Susceptibility testing was performed and interpreted in accordance with Clinical and Laboratory Standard and Institute (CLSI) guidelines 29 .

PFGE was performed on the three selected isolates. PFGE was performed following the protocol described at http://www.cdc.gov/pulsenet/protocols.htm with slight modifications described below 30 . Agarose plugs were prepared as previously described by Maslow et al 31 and the DNA was digested with ApaI 32 . DNA was subsequently separated in 1% agarose on a contour-clamped homogeneous-field machine (CHEF-DR II, Bio-Rad, Richmond, CA). Electrophoresis was performed at 6 V/cm for 18.5 hours, with pulse times ranging from 7-20 seconds. Following electrophoresis, the gels were stained with ethidium bromide and photographed under ultraviolet illumination. Interpretation of PFGE was preformed using the criteria outlined by Tenover et al. 33 , where band patterns are visually compared and classified as indistinguishable (clonal), closely related (clonal variants, three or less band differences), possibly related (four to six band differences) and unrelated (seven or more band differences).

Genomic DNA was extracted from the three cultures grown overnight at 37ºC using standard methods 34 . Genomes were sequenced with the 454 Titanium platform and raw reads were assembled with Newbler (454 Life Sciences). The resulting assembly was annotated with an automated annotation pipeline at the Institute for Genome Sciences (http://www.igs.umaryland.edu/). Assembly details are included in Additional File 2, Table S2.

The gene sequences for seven housekeeping genes (gltA, pyrG, rplB, recA, cpn60, fusA, rpoB) were informatically extracted from genomic contigs with Mugsy 35 and were compared to allele profiles in the A. baumannii PF8MLST database (http://www.pasteur.fr/recherche/genopole/PF8/mlst/Abaumannii.html); sequence types were then assigned for matching profiles.

A BSR analysis 36 was performed on the predicted proteome of the three genomes included in this study and 6 finished A. baumannii genomes (Table 1). Peptides were classified as unique for a BSR > 0.80, divergent for a BSR < 0.80 and > 0.40, and unique for a BSR < 0.40 as described in previous multi-genome analyses 37 38 39 .

The AbaR1 gene sequence was downloaded from Genbank (Locus ID CT025832). Each genome was aligned against this sequence with NUCmer 40 . The resulting percent identity plot was generated showing conservation over the entire length of the gene, as has been done recently 41 .

A whole genome alignment was performed with Mugsy 35 on a total of 23 available A. baumannii genomes (Additional File 2, Table S2) using a maximum gap length between orthologous blocks of 100 base pairs. Columns with gaps were removed from the alignment with mothur 42 to generate an ungapped multiple sequence alignment. For phylogenetic studies, extracting polymorphic positions reduced the size of the alignment. A phylogenetic tree was then inferred on this reduced alignment with FastTree 43 , with 1000 boostrap replicates and using Acinetobacter calcoaceticus as the outgroup.

The collection of 74 A. baumannii isolates was screened by PCR for the genomic regions identified from the Mugsy alignment as being associated with the colonization or invasion phenotypes. Single colonies were grown overnight in Luria broth at 37ºC. Aliquots of the overnight cultures were boiled at 94ºC for 10 minutes, and then the cellular debris was removed by centrifugation for 5 min at 4000 × RPM; this crude cell lysate was then used as the PCR template. PCR primers were designed with Primer3 44 . A list of primers used in this study is shown in Table 2. Reaction conditions were as follows: 95ºC for 5 minutes followed by 30 cycles of 94ºC for 30 s, 58ºC for 30 s, and 72ºC for 45 s.

Genome sequence data are available in GenBank under accession numbers AEPK00000000 (UMB001), AEPL00000000 (UMB002), and AEPM00000000 (UMB003).

Three isolates were chosen for sequencing: a colonizing isolate from the peri-anal site (UMB002), an outbreak isolate from the bloodstream (UMB001), and a non-outbreak wound isolate (UMB003). Each of the isolates has a different PFGE pattern (Additional File 3, Figure S1), providing evidence that, while these isolates were obtained from the same hospital setting in a limited timeframe, they do not represent highly similar clones that are part of a single outbreak.

The allelic profiles were obtained from the genome sequencing projects of the three new genomes analyzed in this study. The PF8MLST (http://www.pasteur.fr/recherche/genopole/PF8/mlst/Abaumannii.html) system assigned the genomes into sequence types (STs) 2 (UMB001), 16 (UMB002), and 25 (UMB003) (Table 1). The comparison of these sequence types to the previously sequenced genomes identified that there is not a consistent MLST pattern associated with the isolates from invasive or colonization sources (i.e. there is not a predominant sequence type for invasion isolates). Additionally, the MLST sequence type did not extend to other phenotypes, for example, strain AB0057 and AB307-0294 are both ST1; AB0057 shows a MDR phenotype, while AB307-0294 shows a susceptible phenotype.

In an effort to assess the phylogeny of these genomes in a gene/annotation independent manner we used the whole genome comparative analysis described previously 34 . The whole genome alignment method identified 2.25 Mb of homologous sequence data shared between all available A. baumannii isolates (Additional File 1, Table S1). From this multiple sequence alignment, we identified 330,631 polymorphic positions (14.7% of the total alignment). As has been demonstrated previously 19 , multi-drug resistant (MDR) strains do not share a monophyletic origin (Figure 1). Limited genetic variation was observed between all A. baumannii genomes; an uncorrected distance matrix for this alignment showed < 2% nucleotide variation for all 17 genomes.

Using a pangenome analysis pipeline maintained at the Institute for Genome Sciences 45 46 47 , we estimated the size of the A. baumannii pangenome based on 6 complete, publicly available A. baumannii genomes (Table 1). We estimate the pangenome size to be approximately 6500 genes (Figure 2A) and approximately 66 new genes are expected with the sequencing of each additional genome (Figure 2B). These values are similar to other A. baumannii pan-genome calculations 19 .

The BLAST score ratio analysis will identify the level of conservation of genes that have been predicted in any set of genomes 36 . Using a conservative BSR threshold of 0.4 (similar to <30% amino acid identity over <30% of the peptide) we identified the unique genes in each of the new genomes sequenced in this study when compared to the complete genomes in GenBank (Additional File 2, Table S2). The BSR analysis for the genomes sequenced in this study identified 180 unique genes in UMB001, 290 in UMB002 and 496 in UMB003. These genes and functional annotations for UMB001-UMB003 are included in Additional Files 4, 5, 6, 7, Tables S3-S6), respectively. For each of these genomes there is no functional annotation for ~ 60% of the unique gene set (UMB001, 106 genes = 58.9%; UMB002, 182 genes = 62.8% UMB003, 288 genes = 58.1%). The majority of genes with functional annotation are similar to genes that are involved in horizontal gene transfer; these genes include transposases or genes involved in Type IV secretion systems that constitute plasmid mobility and conjugation systems. These plasmid transfer genes further suggest the presence of plasmids in these isolates.

The largest unique gene set was observed in UMB003. Interestingly these genes with functional annotation have similarities that suggest recent lateral gene transfer between not just other Acinetobacter species, but other unrelated species. A chloramphenicol acetyltransferase gene was identified that showed high homology (99%) to a Klebsiella pneumoniae homolog 48 . While this gene has been identified in another A. baumannii isolate by single gene sequencing 49 , the previous version demonstrates a much lower level of homology (81%); this gene is not present in any publically available A. baumannii complete or draft genome sequence. Interestingly, in UMB003 a phenol hydroxylase protein complex was identified. This complex is responsible for the breakdown of phenol, the base for many disinfectants; a homologous gene cluster (99% identity) was originally identified in an Acinetobacter calcoaceticus genome isolated from oil refinery wastewater 50 . A primer set was designed for a conserved protein in the phenol hydroxylase complex. A screen of the 74 A. bumannii isolates showed that 25 of them were positive for presence of the amplified gene fragment. A BLAST search of this conserved fragment against all 23 sequenced A. baumannii genomes only showed hits against UMB003 and the draft genome sequence AB_4190 51 .

Clinical laboratory testing identified resistance in all three isolates to several antibiotics including amikacin, ceftazimide, and gentamicin (Table 3). A genomic comparison identified that each isolate contained a gentamicin 3'-acetyltransferase and an RND type efflux pump (AdeT) previously demonstrated to be involved in aminoglycoside resistance 52 . All Isolates contained an AmpC cephalosporinase, which would help explain the clinical observation of ceftazidime resistance 53 ; the insertion sequence ISAbar1, which has been associated with over-production of AmpC 54 , was also present in UMB001 and UMB002. The oxacillinase like-gene bla _OXA-69, which has the ability to hydrolyze carbapenem 55 , was present in all genomes.

Resistance islands are common in A. baumannii as genomic locations for the acquisition and accumulation of genes required for antibiotic resistance 20 22 . The most well studied of these RIs is AbaR1 20 . We utilize the AbaR1 structure as a reference as it appears to be the largest of the RI and all others appear smaller, however it must be noted that the complete AbaR1 structure has not been observed in any other isolate to date. Comparison of all A. baumannii genomes in this genomic region identifies a wide spectrum of genomic conservation (Figure 3). Limited, but highly conserved regions (BSR > 0.80) of the AbaR1 RI are present in UMB001 and UMB003, but absent in UMB002 (Figure 3). Bioinformatic analysis identified that UMB001 and UMB003 contain an aminoglycoside resistance protein and the transposase TN21 that are present in the prototype AbaR1. UMB002 contains only 5 genes from AbaR1, and none of them associated with antibiotic resistance. An analysis of genomic contigs that contain comM from all three UMB strains failed to identify additional genes from known RIs; this suggests that homologous genes to those in known RIs are either in previously unknown RIs, or are not present in a genomic island.

Previous studies have demonstrated the importance of efflux pumps at removing antimicrobials from the cell and conferring the MDR phenotype. These pump systems are present in the three genomes sequenced in this study, but are also present in genomes of drug susceptible isolates. Three systems have been demonstrated to contribute to the MDR phenotype, AdeABC, AdeFGH and AdeIJK 24 25 26 . The three genomes sequenced in this study contain the adeFGH operon, as well as the gene encoding the regulator, AdeL. A multiple sequence alignment of the predicted adeL coding sequence demonstrated no single nucleotide polymorphisms (SNPs) that were unique to either the MDR or drug susceptible phenotypes (data not shown). Additionally, each of the three new genomes contained highly conserved adeA and adeB genes from the AdeABC efflux pump complex; however, UMB002 and UMB003 lack the gene encoding the outer membrane protein AdeC 56 . A multiple alignment of the AdeABC regulator, AdeR 57 , also showed no MDR-specific SNPs.

Detailed comparative genomic analysis identified one gene encoding a hypothetical protein (accession YP_001713249) that is conserved in all 23 A. baumannii genomes and has no identified homologue in 12 other Acinetobacter species genomes. A pairwise alignment demonstrates that there is a high level of conservation for this peptide (Additional File 8 Figure S2). PCR primers were designed from this coding region (Table 2) and a PCR based screen of 74 putative A. baumannii isolates identified 73 positive amplicons (98.6% sensitivity). A BLAST homology search identified the closest homolog as a protein of unknown function with 47% peptide identity in the actinomycete Catenulispora acidiphila.

The majority of A. baumannii genomes sequenced to date demonstrate an invasion phenotype, as they are isolated from fluids within the human body (Table 1 and Additional File 1, Table S1). In fact, only one previously available draft genome, AB900, was isolated from what the present study defines as a colonization isolate. In the case of AB090, the colonization site was the perinea 19 . The current study adds two draft genomes (UMB002, UMB003) that were isolated from colonization of the peri-anal region and a wound, respectively.

A comparative whole genome approach was employed to determine if the organisms with either a colonization (AB900, UMB002, UMB003) or invasion phenotype (AYE, AB0057, AB307-0294, ACICU, ATCC 17978, UMB001, AB056, AB058, AB059) contain unique genomic content. Three additional A. baumannii genomes (AB6013113, AB6013150, AB6014059) were not included in this analysis due to their uncertain isolation location. The results of the whole genome alignment approach demonstrate that unique genomic regions in the invasion isolates consisted mainly of hypothetical proteins (Table 4). There were also identified regions within the colonization isolates that were found to be absent in the invasion isolates. One region in the colonization genotyped isolates contained a large (>5000 aa) putative hemaglutinin/hemolysin-related protein related to a gene that was also present in the body louse strain A. baumanii SDF 20 . Phylogenetic analysis of homologous genes in other pathogens showed that this gene has not been horizontally transferred recently into the Acinetobacter species group (Figure 4). We developed a PCR-based assay to determine if these genetic features were consistently present in a larger group of colonization or invasive isolates. A screen of 74 A. baumannii isolates with primers specific to genes present in either colonization or invasion isolates (Table 2) demonstrated that the genes common to the isolates from each phenotype in the genomic analysis were not restricted to that phenotype in a broader population based analysis (Table 4). Only the two sequenced colonization isolates amplified with the "colonization specific" primer sets. Furthermore, the primers designed from "invasion specific" genomic regions showed no specificity to A. baumannii obtained from blood compared to peri-anal isolates.