|Home | About | Journals | Submit | Contact Us | Français|
The whole-genome sequence of an epidemic, multidrug-resistant Acinetobacter baumannii strain (strain ACICU) belonging to the European clone II group and carrying the plasmid-mediated blaOXA-58 carbapenem resistance gene was determined. The A. baumannii ACICU genome was compared with the genomes of A. baumannii ATCC 17978 and Acinetobacter baylyi ADP1, with the aim of identifying novel genes related to virulence and drug resistance. A. baumannii ACICU has a single chromosome of 3,904,116 bp (which is predicted to contain 3,758 genes) and two plasmids, pACICU1 and pACICU2, of 28,279 and 64,366 bp, respectively. Genome comparison showed 86.4% synteny with A. baumannii ATCC 17978 and 14.8% synteny with A. baylyi ADP1. A conspicuous number of transporters belonging to different superfamilies was predicted for A. baumannii ACICU. The relative number of transporters was much higher in ACICU than in ATCC 17978 and ADP1 (76.2, 57.2, and 62.5 transporters per Mb of genome, respectively). An antibiotic resistance island, AbaR2, was identified in ACICU and had plausibly evolved by reductive evolution from the AbaR1 island previously described in multiresistant strain A. baumannii AYE. Moreover, 36 putative alien islands (pAs) were detected in the ACICU genome; 24 of these had previously been described in the ATCC 17978 genome, 4 are proposed here for the first time and are present in both ATCC 17978 and ACICU, and 8 are unique to the ACICU genome. Fifteen of the pAs in the ACICU genome encode genes related to drug resistance, including membrane transporters and ex novo acquired resistance genes. These findings provide novel insight into the genetic basis of A. baumannii resistance.
Acinetobacter baumannii has emerged as an important opportunistic pathogen worldwide (15). It is responsible for large outbreaks of nosocomial infection which account for 2 to 10% of all infections caused by gram-negative bacteria, with the highest incidence occurring in intensive care units (ICUs) (17, 18). A. baumannii infections include nosocomial pneumonia, secondary meningitis, skin, soft tissue, and urinary tract infections, and bacteremia and result in high rates of morbidity and mortality (9, 13). A. baumannii has simple growth requirements and exploits a variety of nutritional sources, and it is adaptable to a range of temperatures, pHs, salinities, and levels of humidity (4). This contributes to the fitness of A. baumannii in the hospital, which represents the main reservoir of the bacterium. The rates of recovery of A. baumannii from natural environments and its incidence in the community are low, while its rate of carriage by hospitalized patients is high and its occurrence in the hospital setting is frequent (15).
Resistance to antimicrobial agents is the main advantage of A. baumannii in the nosocomial environment (15). Multidrug-resistant clones of A. baumannii are emerging and spreading throughout many geographic areas (10, 36, 40). The emergence of two pan-European epidemic clones, referred to as European clones I and II, was reported in northwestern Europe in the 1980s (15, 28), and since then these clones have been documented in many regions of Europe (8, 39). These epidemic lineages are endowed with resistance to a broad range of antibiotics, although individual strains can display different types and combinations of resistance determinants. The treatment options for multidrug-resistant A. baumannii infections are generally limited to carbapenems, but the emergence of imipenem and meropenem resistance is a growing source of concern (32). Recently, the OXA-58 carbapenemase has been shown to be involved in carbapenem resistance in members of the European clone II group (29).
The pathogenicity of A. baumannii is poorly understood, as this organism is commonly regarded as a low-virulence bacterium, despite the severity of the infections that it causes (15). Recently, novel insights into A. baumannii virulence have been gained from whole-genome analysis of strain ATCC 17978. About 17% of its ca. 4-Mb genome consists of 28 putative alien islands (pAs) acquired by horizontal gene transfer (HGT); and 20 of them harbor potential virulence genes involved in protein secretion, drug resistance, pilus formation, and cell wall biogenesis (35). These pAs were described as large chromosome portions that had little homology with the Acinetobacter baylyi ADP1 genome (1) or that showed anomalies in sequence composition and codon usage. Some of them were confirmed to be important for the fitness of ATCC 17978 by insertional mutagenesis and virulence assays with the Caenorhabditis elegans and Dictyostelium discoideum infection models (35). Moreover, comparative genomics revealed the presence in the European clone I AYE strain of an 86-kb resistance island (AbaR1) carrying 45 resistance genes acquired by HGT from other gram-negative species (30). AbaR1 is inserted into an ATPase gene that is in an “empty” state in antibiotic-susceptible strain A. baumannii SDF. This highlights the exceptional ability of A. baumannii to gather and assemble foreign genes.
In this study, the genome of epidemic multidrug-resistant strain A. baumannii ACICU, representative of a clone that caused an outbreak in Rome in 2005 (21, 22), has been entirely sequenced. This clone belongs to the European clone II group and carries the plasmid-mediated blaOXA-58 gene (37). Since A. baumannii ATCC 17978 was isolated in the early 1950s, prior to the development of macrolides, glycopeptides, cephalosporins, and the latest beta-lactams (2), we compared the ACICU genome with the genomes of A. baumannii ATCC 17978 and A. baylyi ADP1 with the aim of identifying novel traits related to the drug resistance, virulence, and pathogenesis of A. baumannii European clone II.
A. baumannii ACICU (referred to as isolate H34) was described elsewhere (21, 22). A. baumannii ACICU is characterized by resistance to ampicillin, piperacillin, piperacillin-tazobactam, ticarcillin, ticarcillin-clavulanic acid, aztreonam, cefepime, ceftazidime, ciprofloxacin, trimethoprim-sulfamethoxazole, chloramphenicol, amikacin, gentamicin, tobramycin, imipenem, and meropenem (21). This strain was assigned to molecular group 1 by the multiplex PCR-based method described by Turton et al. (37). Escherichia coli DH5α [80dlacZΔM15 Δ(lacZYA-argF)U169 recA1 endA1 hsdR17 supE44 thi-1 gyrA96 relA1 Nalr] was used as the host strain for the DNA cloning experiments. E. coli strain ATCC 25922 was used as the control for the double-disk synergy test (DDST). DDST was performed to assess the contribution of AmpC to ceftazidime resistance. Disks containing 30 μg of ceftazidime or 30 μg of cefotaxime were placed on a Mueller-Hinton agar plate on which the bacterial suspension had been spread, with a center-to-center distance of 18 mm to a 500-μg cloxacillin-containing disk or a 250-μg boronic acid-containing disk (Rosco Diagnostica A/S, Taastrup, Denmark), respectively. The plates were incubated at 37°C overnight (33, 41).
The complete sequencing work flow of the Genome Sequencer FLX system is described at www.454.com. In brief, preparation of the genomic DNA of strain ACICU was performed by using the Wizard Genomic DNA purification kit (Promega Corporation, Madison, WI), according to the manufacturer's procedure. The genomic DNA preparation (3 to 5 μg) was fractionated into fragments ranging from 400 to 800 bp by nebulization and was ligated to adaptor A or B (4-base nonpalindromic sequences) by priming amplification and sequencing reactions. The fragments were also immobilized onto streptavidin beads by the biotin tag on adaptor B. Part of the DNA library was analyzed by the GS20 system, which produced sequences with an average length of 100 bp, while a second fraction was analyzed with the Genome Sequencer FLX system, which produced sequences of 200 to 250 bp. A total of 900,363 high-quality sequences and 96,454,548 bp were obtained with a 23-fold coverage of the genome. The 454 de Novo Assembler software was used to assemble the sequences that were read. This first automatic step produced 1,036 contigs with a maximum length of 200,179 bp.
Contigs were aligned by paired-end analysis (http://www.454.com), with 47 scaffolds obtained by semiautomatic assembly analysis with DNAStar Lasergene software (http://www.dnastar.com/products/lasergene.php). This approach permitted a manual check of every added sequence read to confirm the correct assembly.
To solve problems due to the assembly of repetitive DNA regions, short sequences that mapped into contigs carrying repeated regions were reassembled by manual rereading of the DNA sequences. If a single nucleotide variation was identified, the degenerated character defined in IUB code was used, while for larger variations (2 or more nucleotides), different contigs were created and their localization was validated by PCR. Since homopolymeric stretches into the genome can determine a high probability of frameshift error during the assembly of the sequence, potential errors were checked by visual inspection of the sequences read.
Seven contigs representing the whole genome were finally obtained and were interrupted at the 4.4-kb rRNA gene clusters. To complete the assembly of the contigs, a PCR strategy with 14 primer pairs designed on the basis of univocal sequences flanking each rRNA cluster was used. The use of all possible combinations of these primers and the seven contigs as DNA templates was attempted with the Elongase amplification system (Invitrogen, Milan, Italy). Seven amplicons were obtained and were sequenced with an ABI 3730 DNA sequencing instrument (Applied Biosystems, Monza, Italy). Adjacent contigs were then recognized and joined to obtain a full-length genome sequence.
To avoid hybrid contigs, plasmid DNA was analyzed independently from genomic DNA by using the same assembly protocols.
Coding genes were identified by crossing the predictions from the FGENESB package (24) (http://www.softberry.com/), the GeneMark program (7), and the GLIMMER program (14). We considered an open reading frame (ORF) prediction to be good when it was identified by each of the three prediction tools. Discrepant ORFs were manually verified by the Artemis viewer (5) and by identification of putative ribosomal binding sites.
Each gene was functionally classified by assigning a cluster of orthologous group (COG) number (39) or a Kyoto encyclopedia of genes and genomes (KEGG) number (23), and each predicted protein was compared against every protein in the nonredundant (nr) protein databases (http://ncbi.nlm.nih.gov) (1). In order to associate a function with a predicted gene, we used a minimum cutoff of 30% identity and 80% coverage of the gene length, checking at least two best hits among the COG, KEGG, and nonredundant protein databases.
The syntenies of the A. baumannii ACICU, A. baumannii ATCC 17978 (EMBL accession no. CP000521), and A. baylyi ADP1 (EMBL accession no. CR543861) genomes were investigated by the use of Mauve software (12).
The rRNA genes were identified by the FGENESB tool on the basis of sequence conservation, while tRNA genes were detected with the tRNAscan-SE program (23).
The strain ACICU and ATCC 17978 genomes were submitted to the http://www.membranetransport.org/ website to determine the presence of efflux systems and for the identification of transporter families, subfamilies, and substrates by the TransAAP (Transporter Automatic Annotation Pipeline) tool. The membrane transport systems in ACICU and ATCC 17978 were also compared with the systems previously identified by the TransAAP tool in different bacterial genomes (31). The entire list of transporters in ACICU and ATCC 17978 is available at the www.itb.cnr.it/genome-project website.
Insertion sequences (ISs) were identified by submission of the whole genome to the IS Finder website (http://www-is.biotoul.fr).
The BLASTp algorithm was used to search for protein similarities. The criterion used to evaluate the deduced amino acid sequence synteny was >50% similarity at the amino acid level and >50% coverage of protein length. Coding DNA sequences (CDSs) not matching these criteria were considered absent in the comparison of the A. baumannii ATCC 17978 and ACICU genomes.
Small-scale and large-scale genomic variations were detected in the sequence of strain ACICU compared with the sequence of strain ATCC 17978. Small-scale variations are defined as genetic changes affecting individual genes or small clusters (less than five CDSs) and will not be discussed in detail. Large-scale variations were interpreted when changes involved clusters of at least five adjacent CDSs.
The strain ACICU genome was analyzed with the extended PyPhy system (the SPyPhy system) (36; T. Sicheritz-Ponten, unpublished data) to make an automated wide primary screen of the genome for HGT, as described previously (36). Briefly, each protein sequence was compared to the sequences in the UniProt database by use of the BLASTp algorithm. All compatible sequences with >30% amino acid identity and >50% coverage of the protein length were aligned by using the ClustalW program. After filtering and alignment of the sequences, the GBLOCKS program (9) was used to remove poorly aligned and diverged positions (allowed gap positions, half; minimum length of a block, 2; maximum number of contiguous nonconserved positions, 20). Neighbor-joining analysis with 100 bootstrap replicates was performed with the PAUP* program (38) for those alignments for which there was a sufficient number of homologs for tree construction. Only BLAST matches involving at least five consecutive CDSs were interpreted and considered potential candidates for HGT.
Plasmid DNA pyrosequencing was performed by using template purified with a PureLink Hipure plasmid midiprep kit (Invitrogen). Contigs containing repetitive sequences were analyzed for colinearity by restriction fragment length polymorphism analsis. SacI- and SalI-restricted plasmids were separated on a 1% agarose gel, transferred onto positively charged nylon membranes (Roche Diagnostics, Monza, Italy), and hybridized with PCR-generated specific probes with a DIG DNA labeling and detection kit (Roche Diagnostics). To complete the sequence, a SacI library was obtained in the pZErO-2 vector (Invitrogen) and transformed in competent MAX Efficiency E. coli DH5α cells (Invitrogen). The transformants were selected on Luria-Bertani agar plates containing 1 mM isopropyl-β-d-thiogalactopyranoside and kanamycin (40 μg/ml) alone or in combination with ampicillin (20 μg/ml). The nucleotide sequences of the SacI-generated inserts were confirmed with universal and walking primers (see details and primer sequences at the www.itb.cnr.it/genome-project website) with the ABI 3730 DNA sequencing instrument (Applied Biosystems, Monza, Italy).
The A. baumannii ACICU chromosome and plasmid sequences (pACICU1 and pACICU2) were submitted to the GenBank database and can be found under accession numbers CP000863, CP000864, and CP000865, respectively. The general properties of this genome are summarized in the supporting information available at the www.itb.cnr.it/genome-project website.
A. baumannii ACICU is an invasive multidrug-resistant strain isolated from the cerebrospinal fluid of a patient cared for at the ICU of Hospital S. Giovanni-Addolorata in Rome, Italy (21, 22). A. baumannii ACICU was assigned to the European clone II group; produces amplicons of 355, 702, and 599 bp; and is referred to as molecular group 1 (37). This result was confirmed by arbitrarily primed PCR and pulsotyping, and identity or a high degree of similarity to prototypic European clone II isolate RUH134 was shown (data not shown).
A. baumannii ACICU contains a single circular chromosome of 3,904,116 bp and two plasmids (pACICU1 and pACICU2) of 28,279 and 64,366 bp, respectively; 3,758 genes were annotated in the ACICU chromosome, including 3,670 predicted protein-encoding CDSs, 64 tRNA genes, and 8 rRNA operons (Table (Table1).1). Nearly 70% of the CDSs (n = 2,670) were assigned to a COG functional category; several genes belonged to more than one COG class (see Table S1 in the supplemental material). For many COG classes, the number of CDSs identified in ACICU largely exceeds the number identified in ATCC 17978, since in the latter strain only 60.1% of the genes were assigned to a COG class. The discrepancy observed is justified by the fact that almost the 7% of the ACICU CDSs do not show a significant match (>50% similarity, >50% coverage) with any CDS of ATCC 17978, explaining why many COG classes are more abundant in ACICU than in ATCC 17978.
ACICU also contains 14 ISs in the chromosome, including 7 ISAba125 elements, 4 ISAba2 elements, 2 IS26 elements, and 1 ISPu12 element, and 11 on plasmids, including 3 ISAba3 elements, 3 IS26 elements, 4 ISAba2 elements, and 1 ISAba125 element. ISAba1 elements were not identified in this strain. The chromosome is composed of 0.38% short repetitive mini- and microsatellite DNA sequences (www.itb.cnr.it/genome-project).
The A. baumannii ACICU genome was compared with that of its closest sequenced relatives, namely, A. baumannii ATCC 17978 (35) and A. baylyi ADP1 (1), with Mauve software (12). That comparison enabled the alignment of orthologous and xenologous regions. Comparison of the genomes showed that the A. baumannii ACICU genome had 86.4% synteny with the A. baumannii ATCC 17978 genome and 14.8% synteny with the A. baylyi ADP1 genome. An interactive graphical output made it possible to locate potential regions unique to the A. baumannii ACICU genome (Fig. (Fig.11).
Thirty-six putative pAs were identified in the strain ACICU genome (designated pAICU and numbered from 1 to 36; Table Table2),2), for a total of 1,030 CDSs, by genome comparison, use of the SPyPhy program, and G+C content analyses. Of these, 24 showed partial or total correspondence with the pAs previously identified in A. baumannii ATCC 17978 (35), while 4 of the 28 A. baumannii ATCC 17978 pAs (pA3, pA4, pA20, and pA25) were absent from the ACICU genome. Four additional regions of the ACICU genome (pAICU4, pAICU9, pAICU26, and pAICU34) were identified as potential candidates of HGT, showed significantly different G+C contents compared with that for the whole ACICU genome, and are predicted to encode proteins involved in metabolism. These four regions are present in the ATCC 17978 genome, but they were not designated pAs in the previous annotation (Table (Table2).2). Finally, eight regions (pAICU3, pAICU6, pAICU8, pAICU20, pAICU24, pAICU29, pAICU30, and pAICU32) were exclusive to the ACICU genome and showed no match with any region of the ATCC 17978 genome.
The A. baumannii ACICU genome shows 367 genes involved in membrane transport, equivalent to 76.6 genes per Mb of genome (Fig. (Fig.2A)2A) (http://www.itb.cnr.it/genome-project/index.php?mod=6). This value is much higher than the values for the other Acinetobacter genomes (275 transporters in ATCC 17978, equivalent to 57.2 genes per Mb, and 225 transporters in ADP1, equivalent to 62.5 genes per Mb), while it is comparable to the values for the E. coli, Staphylococcus aureus, and Pseudomonas sp. genomes (Fig. (Fig.2A).2A). A feature common to all Acinetobacter strains is the paucity of sugar-specific phosphotransferase systems compared with the number in fermenting species like E. coli and S. aureus, which have about 20. None of the Acinetobacter phosphotransferase systems shows a predicted specificity for glucose, while only one fructose-specific type IIC enzyme was found in all three Acinetobacter genomes. Moreover, Acinetobacter spp. lack predicted monosaccharide-specific transporters of the major facilitator superfamily (MFS). These observations, combined with the lack of exokinase and glucokinase enzymes involved in the first glycolytic step (35; this work; http://www.itb.cnr.it/genome-project/index.php?mod=7), account for the noteworthy inability of members of the genus Acinetobacter to catabolize glucose. On the other hand, A. baumannii ACICU shows specificity for the uptake of a variety of alternative carbon sources, including benzoate, citrate, galactonate, gluconate, glycerol, lactate and 4-hydroxyphenylacetate (data not shown), which is consistent with the peculiar catabolic profile of the species (3, 34). Interestingly, 11 chromosomal ORFs were predicted to encode porins which could have a role in diffusion of substrates through the outer membrane.
The 36 pAICUs represent the most interesting part of the A. baumannii ACICU genome. The relevant features of the pAICUs in comparison with the features of the A. baumannii ATCC 17978 pAs are outlined below and are summarized in Table Table22.
pAICU1 is remarkably different from ATCC 17978 pA1. The genes involved in cell wall biogenesis (wecC, mviM, wbbJ, wecE, and the O-antigen translocase-encoding gene) detected in ATCC 17978 pA1 do not match any of the genes identified in pAICU1. However, newly acquired functions related to cell wall biogenesis were identified in pAICU1, including the UDP-N-acetyl-d-mannosaminuronate dehydrogenase (78% amino acid identity with the Pasteurella multocida homologue), the spore coat polysaccharide biosynthesis protein (44% amino acid identity with the Alkalilimnicola erlichei homologue), the sialic acid synthase (75% amino acid identity with the Pseudomonas stutzeri homologue), and the membrane protein involved in the export of O antigen and teichoic acid (35% amino acid identity with the Pseudomonas aeruginosa homologue). All these functions plausibly originated by HGT, since the overall G+C content of pAICU1 (34.0%) differs significantly from that of the ACICU genome (39.1%). Thus, pAICU1 is predicted to retain a general function in cell envelope biogenesis, but with a low level of gene conservation with respect to that of its counterpart in the ATCC 17978 genome.
A homologue of ATCC 17978 pA3 was not detected in the ACICU genome. At the same genomic location, the pAICU3 drug resistance island, which contains transposases, integrases, insertion sequences, antibiotic resistance genes, and heavy metal detoxification proteins, was identified (this is further explained below).
pAICU6, pAICU9, pAICU24, and pAICU30 are exclusive to the ACICU genome and mostly contain genes encoding hypothetical proteins, in addition to phage-related proteins. It was not possible to predict any specific function conferred by the acquisition of these phages (designated ACICU phages 1, 2, 3, and 4, respectively), since except for pAICU24, the encoded proteins do not show significant similarity with any known protein. pAICU24 contains a toxin-antitoxin stability system whose importance for the maintenance of this region deserves further investigation.
pAICU13 has a predicted function in phenyl acetic acid degradation and is homologous to part of pA10, previously identified in ATCC 17978. However, strain ACICU shows 35 novel CDSs within this pA, including four MFS transporters, likely involved in drug resistance on the basis of predictions obtained with the TransAAP tool. Only one MFS transporter was identified in the corresponding pA of ATCC 17978, pA10 (outlined below and in Table S2 in the supplemental material). Furthermore, one additional transporter of the nucleobase:cation symporter 1 (NCS1) was identified in ACICU but was absent from ATCC 17978, together with three ATP binding cassette (ABC) transporters, two of which showed substrate specificity for the ferric ion, and one MSF not involved in drug resistance. These findings strongly suggest that pAICU13 is a candidate relevant to ACICU virulence and drug resistance.
pAICU20 and pAICU32 were identified by screening with the SPyPhy system, and they are exclusive to the ACICU genome, being absent from strain ATCC 17978. A complete ABC transport system for molybdate, a cluster of genes involved in the coenzyme pyrroloquinolinequinone (PQQ) biosynthesis (73% amino acid identity with the Pseudomonas aeruginosa homologue), and an additional cluster for pilus biosynthesis were identified in pAICU20. Hypothetical proteins and a putative hemagglutinin/hemolysin-related protein (48% amino acid identity with the Marinobacter algicola homologue) were identified in pAICU32. These traits plausibly contribute to the virulence potential of strain ACICU. pAICU32 has likely been acquired by HGT, as it showed a significantly higher G+C content (50.4%) than the core ACICU genome.
pAICU21 and pAICU22 correspond to pA17 and pA18 of ATCC 17978, respectively, and are implicated in drug resistance, as they encode specific membrane transport systems. In particular, pAICU21 carries the cluster encoding the AdeABC system, which has been demonstrated to be involved in antimicrobial resistance in A. baumannii (29). Interestingly, pAICU22 contains one MFS transporter and one drug/metabolite transporter (DMT) likely involved in multidrug resistance, a surface adhesion protein fused to an RTX calcium-binding cytotoxin (33% identity with the Pseudomonas putida homologue), a hemagglutinin FhaB-like protein, and a hemolysin activator FhaC-like protein previously identified in ADP1 but not in ATCC 17978. The surface adhesion protein is predicted to belong to the family of the RTX cytotoxins, which are involved in invasion and hemolysis. Notably, the Vibrio vulnificus RtxA1 toxin causes hemolysis through pore formation and helps with invasion of the bloodstream from ligated ileal loops of CD-1 mice (20). Other genes are present in this region, including a second MFS permease, esterase/lipase, and catalase enzymes, whose roles in ACICU pathogenesis are unclear at present.
pAICU29 is exclusive to the ACICU genome and contains several genes encoding membrane transport systems related to drug resistance. In particular, a permease of the MFS group and two DMTs were identified in this region.
To gain further insight into the genetic basis of resistance, the A. baumannii ACICU and ATCC 17978 genomes were inspected for drug resistance determinants, and the drug resistance regions were then compared with those identified in the multiresistant strain A. baumannii AYE (16). An 86-kb resistance island (AbaR1) carrying 82 CDSs was recently described in A. baumannii AYE (16). A different resistance island, designated AbaR2, was identified in the genome of ACICU as pAICU3 (Fig. (Fig.3A).3A). AbaR2 was shorter than AbaR1, showing only 17 CDSs, including seven resistance genes. Within this region, a class 1 integron carrying the aminoglycoside acetyltransferase gene aacA4, a putative CDS encoding an unknown function, and the blaOXA-20 beta-lactamase gene were identified. This gene cassette array is different from that described for AbaR1, even if genes flanking the 5′ and 3′ conserved segments of the integron perfectly match in AbaR1 and AbaR2. In both islands the integrons are flanked by the aphA1-IS26 element at the 5′ conserved segment and the ISPpu12-like transposon at the 3′ conserved segment. However, 116 nucleotides at the 3′ end of the aphA1 gene are missing in AbaR2 as well as the remaining portion of AbaR1. The 3′ end of an ATPase gene, located at one boundary and considered the integration site of AbaR1 within the AYE genome, is missing in ACICU. Both strains show the 5′ end of the ATPase gene, located at the other boundary of the island. This configuration is compatible with a deletion event in AbaR2 involving a large portion of AbaR2 from the 3′ end of the flanking ATPase gene to a site internal to the aphA1 gene. As a consequence of this deletion, AbaR2 lacks 69 CDSs, including the arsenic, mercury, and tetracycline resistance operons, and a large integron carrying the blaVEB-1 and blaOXA-10 genes.
A. baumannii AYE and ACICU belong to different clonal groups (European clones I and II, respectively) (37). However, the presence of related resistance islands in both lineages suggests that AbaR1 and AbaR2 derived from an island acquired by a common A. baumannii ancestor before their divergence into two different clonal lineages.
Besides the blaOXA-20 gene identified within the integron, strain ACICU contains three additional beta-lactamases, including a putative class A beta-lactamase previously reported in strain AYE (18), the previously described AmpC beta-lactamase, and the naturally occurring OXA-66 oxacillinase (Table (Table3).3). OXA-66 is predicted to be not expressed since it is not flanked by any IS previously described to be necessary to drive expression of the carbapenem resistance phenotype (38). Interestingly, the ampC gene is preceded by an ISAba125 element that could influence its expression. Notably, ISAba1-directed overexpression of the naturally occurring ampC gene has been identified as a mechanism conferring ceftazidime resistance in A. baumannii clinical strains (19). However, DDST performed with class C β-lactamase inhibitors, such as cloxacillin and boronic acid, did not restore ceftazidime susceptibility in ACICU, suggesting that AmpC is not the only mechanism accounting for ceftazidime resistance and that multidrug efflux pumps could contribute to this phenotype.
Resistance to fluoroquinolones in ACICU (ciprofloxacin MIC > 16 μg/ml ) is likely ascribed to the previously characterized mutation occurring in the gyrA gene, with a substitution of Ser for Leu at position 83 of its amino acid sequence. In comparison with the sequences of strains AYE and ATCC 17978, several amino acid substitutions that were not previously described to be associated with fluoroquinolone resistance in any Acinetobacter strain, however, were observed in the parC and parE genes, while no mutations were observed in the gyrB gene of ACICU.
Transporters plausibly involved in drug efflux likely make an important contribution to the overall multidrug-resistant phenotype of ACICU (see Table S2 in the supplemental material), and these transporters are more numerous in ACICU than in other Acinetobacter strains (Fig. (Fig.2B).2B). Forty MFS drug transporters were identified in ACICU, whereas 33 were identified in ATCC 17978 and only 11 were identified in ADP1. Twenty-six DMTs were also detected, with 17 detected in ATCC 17978 and 12 detected in ADP1. Furthermore, eight drug transporters of the ABC superfamily, seven of the resistance-nodulation-cell division (RND) family, and three of the multidrug/oligosaccharidyl-lipid/polysaccharide family were also identified in ACICU (the ATCC 17978 genome contained six, six, and four of these transporters, respectively) (Fig. (Fig.2B;2B; see also Table S2 in the supplemental material). Among the RND transporters, both AdeABC (ACICU_01823, ACICU_01824, ACICU_01825) and AdeIJK (ACICU_02987, ACICU_02988, ACICU_02989) were identified in ACICU. These RND systems were demonstrated to act as pumps for the efflux of multiple antibiotics in A. baumannii (11, 25). Resistance to arsenic compounds was due to one transporter of the arsenical resistance-3 family, which was also detected in ATCC 17978 but not in ADP1 (Fig. (Fig.2B;2B; see Table S2 the supplemental material). These transporters may account for the antimicrobial resistance phenotype of strain ACICU, including the resistance to those drugs for which specific genes have not yet been identified (i.e., tetracycline, trimethoprim, and chloramphenicol).
The carbapenem resistance in A. baumannii ACICU is conferred by the carbapenem-hydrolyzing OXA-58 oxacillinase, whose gene was detected in plasmid pACICU1. Notably, the chromosomal gene encoding outer membrane protein CarO (ACICU_02813), which is responsible for the influx of carbapenem antibiotics, should not contribute to carbapenem resistance in this strain, since it is not interrupted by any IS, as is the case in other A. baumannii strains (27).
Plasmid pACICU1 carries two copies of the blaOXA-58 gene, and these likely originated from direct repetition of the ISAba2-blaOXA-58-ISAba3 mobile element (Fig. (Fig.3B).3B). The duplication of the blaOXA-58 gene has been already described in other unrelated A. baumannii strains and is generally associated with increased resistance to imipenem and meropenem (6). The duplication of this genetic determinant has likely been facilitated by the presence of multiple IS26 elements dispersed within this plasmid region.
Plasmid pACICU1 also contains two replication systems. It carries two replicase genes (repAci1 and repAciX), both of which are flanked by cis regulatory elements (iterons). repAciX is close to genes encoding an antitoxin-toxin system, which ensures plasmid maintenance, and to the par gene, which confers stability during plasmid partitioning. These systems are very well known in plasmids of the family Enterobacteriaceae (26), but to the best of our knowledge, this is the first description of their presence in an Acinetobacter plasmid. repAci1 is associated with a retro-type reverse transcriptase-like gene and a gene encoding a phage protein. This part of the plasmid could therefore derive from a bacteriophage and could be implicated in horizontal mobilization, since a conjugation system(s) was not identified in this plasmid. A complete tra locus, which encodes the conjugative apparatus and the type IV secretion system, was identified in plasmid pACICU2, which coresides with pACICU1 in A. baumannii ACICU. It is therefore plausible that pACICU2 could provide in trans the mobilization functions for pACICU1 conjugation. It is intriguing that the composite scaffold of pACICU1, which carries two replicons and phage-related proteins, is the vehicle of the blaOXA-58 gene. It is worth emphasizing that the blaOXA-58 gene has been identified worldwide as one of the most common mechanisms for carbapenem resistance in clinical A. baumannii isolates. Thus, elucidation of the mechanism(s) implicated in the mobilization and diffusion of a blaOXA-58-carrying plasmid, as well as determination of the plasmid scaffold sequence, would help to set a genetic screen for the rapid monitoring for and detection of this carbapenem resistance plasmid in other strains unrelated to A. baumannii.
A. baumannii has gained a leading position among opportunistic pathogens, and epidemic clones are rapidly evolving toward antimicrobial panresistance. This feature greatly facilitates their spread in health care facilities. In this study, we demonstrated the usefulness of high-density pyrophosphate DNA sequencing as a tool to achieve rapid insight into the physiology, pathogenicity, and molecular basis of antimicrobial resistance of A. baumannii ACICU, a pathogenic multidrug-resistant strain typed as a member of the European clone II group. Whole-genome comparison of A. baumannii ACICU and the A. baumannii ATCC 17978 reference strain revealed a number of distinctive traits. The key feature of the ACICU genome is a conspicuous number of additional functions that were not identified in ATCC 17978. In several cases these regions are clusters of genes homologous to those identified in pathogenic or commensal proteobacteria, such as Neisseria, Burkholderia, and Pseudomonas. These clusters were presumably acquired by HGT from these genera or, alternatively, originated in Acinetobacter and were then transferred to other organisms. Given the propensity of A. baumannii to acquire exogenous genetic material, it is tempting to speculate that HGT events may occur with phylogenetically related species sharing a common ecological niche, such as the hospital environment or the skin and mucosal surfaces of the colonized human host. Among the novel functions gained by A. baumannii ACICU, there are (i) improved membrane transport potential (76.2 transporters per Mb of genome compared with 57.2 transporters per Mb in ATCC 17978 and 62.5 transporters per Mb in ADP1); (ii) expanded drug resistance (which accounts for the panresistant phenotype); and (iii) alteration of the cell envelope (novel genes for the synthesis of the cell wall), adherence properties (surface adhesion proteins fused to putative cytotoxin binding domains), potential changes in the pathogenic potential (hemagglutinin/hemolysin-related proteins), and a high number of hypothetical and/or functionally uncharacterized proteins, often associated with regions of prophage origin. Altogether, these findings provide a solid genetic background for future insights into the functional basis of the biology, pathogenicity, and multidrug resistance of A. baumannii.
This work was supported by contract DRESP2 (6th PCRD, LSHM-CT-2005-018705) from the European Commission and a contract for the FIRB Project Costruzione di un Laboratorio Nazionale per lo Studio delle Resistenze Batteriche agli Antibiotici from the Italian Ministry of Research and University.
We thank Qinghu Ren for his high-throughput transporter analysis pipeline (TranSAPP http://www.membranetransport.org/) and assistance.
Published ahead of print on 14 April 2008.
†Supplemental material for this article may be found at http://aac.asm.org/.