Insights on the Horizontal Gene Transfer of Carbapenemase Determinants in the Opportunistic Pathogen Acinetobacter baumannii

Horizontal gene transfer (HGT) is a driving force to the evolution of bacteria. The fast emergence of antimicrobial resistance reflects the ability of genetic adaptation of pathogens. Acinetobacter baumannii has emerged in the last few decades as an important opportunistic nosocomial pathogen, in part due to its high capacity of acquiring resistance to diverse antibiotic families, including to the so-called last line drugs such as carbapenems. The rampant selective pressure and genetic exchange of resistance genes hinder the effective treatment of resistant infections. A. baumannii uses all the resistance mechanisms to survive against carbapenems but production of carbapenemases are the major mechanism, which may act in synergy with others. A. baumannii appears to use all the mechanisms of gene dissemination. Beyond conjugation, the mostly reported recent studies point to natural transformation, transduction and outer membrane vesicles-mediated transfer as mechanisms that may play a role in carbapenemase determinants spread. Understanding the genetic mobilization of carbapenemase genes is paramount in preventing their dissemination. Here we review the carbapenemases found in A. baumannii and present an overview of the current knowledge of contributions of the various HGT mechanisms to the molecular epidemiology of carbapenem resistance in this relevant opportunistic pathogen.


Introduction
Antimicrobial resistance has become a public health problem at a global scale; it limits therapeutic options and thereby increases morbidity, mortality and treatment costs. The emergence of resistance in pathogens and its dynamics in the last decades is the result of an evolutionary process, in part fueled by anthropogenic activities. Resistance can arise by mutation, transmitted vertically through populations by cell division, or through the acquisition of resistance gene(s) from other bacteria belonging to the same generation in a process called horizontal gene transfer (HGT). HGT is recognized as a major evolutionary force that is constantly reshaping genomes. Three principal mechanisms of intercellular transfer are described: natural transformation (transfer of naked DNA to a recipient cell), transduction (phage assisted transfer) and conjugation (direct cell to cell transfer through conjugative plasmids); the latter most often associated with the dissemination of resistance [1]. Additionally, resistance genes can be mobilized by transposable elements within the chromosome and to plasmids (intracellular movement) and vice-versa.
Acinetobacter baumannii emerged as an important opportunistic pathogen with a high ability to acquire antimicrobial resistance and nowadays, many strains are only susceptible to carbapenems and

Carbapenem Resistance in Acinetobacter baumannii
Carbapenems (imipenem, meropenem, biapenem, ertapenem, and doripenem) are the β-lactam antibiotics with the broadest spectrum of activity, showing great activity against Gram-negative and Gram-positive bacteria and are considered "last line agents". Also, they are relatively stable to the majority of β-lactamases, the more frequent reported mechanism of resistance to β-lactam antibiotics. Nevertheless, the resistance to carbapenems has been dramatically increasing over recent years among the Gram-negatives [22,23].
Diverse other mechanisms may be responsible for carbapenem resistance in A. baumannii. The loss or alterations of specific outer-membrane proteins have been reported to play a role in carbapenem resistance (CarO) [24]. Occasional reports associated the resistance to carbapenems to modification of penicillin binding proteins (PBPs) [25]. The AdeABC efflux pump, belonging to the resistance nodulation division family of efflux systems (RND) and associated with the efflux of antibiotics with diverse structures [26], was shown to be present in A. baumannii [27,28] and eventually to play some role in carbapenem resistance. However, the latter resistance mechanism appeared to give clinical resistance to carbapenems only when associated with others, namely the production of carbapenem-hydrolysing oxacillinases [29,30].

Intrinsic β-Lactamases
A. baumannii produces intrinsic β-lactamases such as an AmpC-type cephalosporinase expressed at low levels. However, when the insertion sequence ISAba1 is located upstream the bla AmpC gene, its expression is enhanced and provides resistance to 3rd generation cephalosporins, but not to carbapenems.
Additionally, it seems that the majority of A. baumannii strains produce oxacillinases represented by OXA-51/69 variants [31,32]. These enzymes show weak hydrolytic activity towards carbapenems. However, isolates bearing the ISAba1-bla OXA-51 -like gene, show higher rates of resistance to imipenem and meropenem [33,34]. The bla OXA-51 -type are usually located in the chromosome. It has been suggested its use as a marker for identification of the species [35] but it has also been shown that other species may eventually carry this oxacillinase in plasmids [36,37]. Detection of target site duplications surrounding transposon Tn6080, which carry the bla OXA-51 -like gene in A. baumannii plasmids, suggest that this genetic element has moved by transposition from the chromosome of this species [34]. Transfer of plasmid-carrying bla OXA-51 -like by HGT, probably conjugation, to non-baumannii species has probably occurred. Conjugation of plasmids carrying the bla OXA-51 -like gene from Acinetobacter nosocomialis to A. baumannii was observed [37].

Acquired β-Lactamases
The most striking activity against to carbapenems in A. baumannii has been described by the acquisition of diverse carbapenemase encoding-genes. Carbapenemases are a group of enzymes that are able to hydrolyse carbapenems even at low level. The emergence of infections caused by extended spectrum β-lactamases (ESBLs) Gram-negative producers in the early 1990s, concomitantly resistant to other classes of antibiotics such as quinolones and aminoglycosides, lead to the more frequent use of carbapenems in complicated infections, which may be associated with the emergence and spread of carbapenemases through all continents [38][39][40].
Carbapenemases belong to three classes of the Ambler classification [41]: the class B, comprising the metallo-β-lactamases (MBLs) that require a bivalent metal ion, usually Zn 2+ , for activity; and the class D, the typical class of oxacillinases, and to class A. The former are potent carbapenemases with the ability to hydrolyse all β-lactams, except aztreonam. To date, three families of MBLs have been identified in A. baumannii: the IMP-, the VIM-and NDM-types. Although the bla SIM-1 gene was described for the first time as part of the class 1 integron present in seven clinical isolates of  (Table 1). To our knowledge, only one study reports the finding of the bla OXA-48 gene in A. baumannii. However, identification of the location and genetic environment of the gene was not the focus of this report [47].
Both MBLs and CHDLs are resistant to the β-lactam inhibitors clavulanic acid and tazobactam [44].

Mechanisms of Acquisition of Carbapenemase Determinants in Acinetobacter baumannii
Despite the numerous reports on the mechanisms of the acquisition of exogenous DNA by bacteria, our knowledge is far from complete in explaining the rapid genetic evolution of bacteria. Some species seem more prone to acquiring exogenous DNA than others. A. baumannii has undoubtedly an enormous genetic plasticity in order to survive.
Members of the Acinetobacter genus can exploit four different HGT mechanisms to exchange genetic material. Conjugation is reported to have the greatest impact in the dissemination of antimicrobial resistance [48]. Transduction and transformation are apparently less important, yet recent studies suggest that their role may be more expressive than initially thought.
All of the HGT mechanisms have been specifically observed in A. baumannii. Many of the bla OXA genes and some MBLs determinants, such as bla NDM genes, are found to be plasmidic, suggesting that conjugation has an important role in the dissemination of these carbapenemases determinants. Nonetheless, a few studies failed to demonstrate conjugation of plasmids in this species [49-52] and quite often their dissemination is inferred from their plasmid location, especially in epidemiological studies where conjugation assays are not demonstrated. Moreover, the isolation and study of Acinetobacter plasmids has been shown to be a difficult task, which may be a limiting factor in experimental assays [52,53]. A few studies have experimentally shown transduction in A. baumannii [54,55]. Natural transformation is known to occur in the Acinetobacter genus [56]. However, this mechanism has not been demonstrated to occur in clinical A. baumannii isolates until recently. Indeed, two studies and our own current research identified naturally competent strains of A. baumannii; this species can under certain conditions take up both chromosomal and plasmid DNA [57,58], leading to the hypothesis that natural transformation may have some influence in the dynamics of resistance genes in A. baumannii. Another mechanism was also demonstrated in A. baumannii, where plasmid DNA transfer can be mediated by OMVs [59], a mechanism not frequently reported in explaining the dissemination of resistance genes and most probably underestimated.
The next chapters will review the mechanisms of the horizontal transfer of carbapenemase genes in A. baumannii. Unlike the majority of oxacillinases in other genera, none of the A. baumannii bla OXA genes associated with carbapenem resistance have been found in integrons [2,60]. On the other hand, some MBLs are embedded as gene cassettes in integrons, which are not self-mobile but are often inserted into mobile elements [2,61]. The bla OXA-23 (previously known as bla ARI-1 ) gene, identified for the first time in a clinical A. baumannii isolate collected in 1985 [62,63], can be inserted both in the chromosome and in plasmids and be surrounded by different genetic contexts, such as genomic islands (AbaR4, AbaR25 and AbaR26), transposons (Tn2006, Tn2007, Tn2008, Tn2009 and Tn6206) and insertion sequences (ISs) [64][65][66][67].
The bla OXA-23 gene, which is nowadays widely distributed in A. baumannii, was originally naturally present in Acinetobacter radioresistens strains, though not expressed and a sequence of different transposition and horizontal transfer events have probably resulted in the transfer of this gene to A. baumannii [66,121]. This gene has been identified in other Acinetobacter species, namely Acinetobacter baylyi [122], Acinetobacter johnsonii [123], Acinetobacter gandensis [124], Acinetobacter genomic species 14TU/13BJ [125], Acinetobacter genomic species 15TU [53], A. nosocomialis [125] and Acinetobacter pittii [123]. The same genetic context of the bla OXA-23 gene in A. baumannii is found in some of these species, suggesting occurrence of HGT [53, 67,122,125,126]. There are also three reports on the detection of the bla OXA-23 gene in two species of the Enterobacteriaceae family, namely Escherichia coli [127] and Proteus mirabilis [128,129]. While the bla OXA-23 gene in P. mirabilis is chromosomally-located [128,129], the gene found in E. coli is flanked by two IS1 elements and it is located on a non-self-conjugative plasmid, which failed conjugation to E. coli and A. baumannii recipient strains [127].
Expression of this gene in A. baumannii is associated with the presence of ISAba insertion sequences, especially ISAba1, preceding it [33,66,68,69]; the association of ISAba1 with bla OXA-23 has also been implicated in the movement of this gene by transposition, as experimentally shown [70] or as reflected by the identification of target site duplications surrounding the composite-transposons [66,67,69].
Horizontal transfer of this gene due to plasmid conjugation between A. baumannii strains has been demonstrated [64,71,72]; conjugation of a plasmid-carrying bla OXA-23 from A. baumannii to A. baylyi [130] has also been shown. Intracellular movement of transposons-carrying the bla OXA-23 gene by transposition can be suggested due to the identification of target site duplications surrounding the transposons, as determined by nucleotide sequencing of the genetic context of the gene in wildtype A. baumannii strains [64,71,73]; the same is valid for genomic islands [72].
Evidence of HGT has also been inferred from the different chromosomal insertion sites of the bla OXA-23 gene in isolates belonging to the same clonal complex [130]. Other evidence demonstrating the horizontal movement of the bla OXA-23 gene associated with the ISAba1 mobile element has come from the fact that the same genetic context is detected in several genetically unrelated isolates [68].

OXA-40/24 (and OXA-40-Like)
The bla OXA-40 (firstly identified as bla OXA-24 ) gene was identified for the first time in a clinical A. baumannii strain isolated in 1997 [74,131]. The bla OXA-40 and bla OXA-40 -like genes reported in A. baumannii are located both in the chromosome [74][75][76] and in plasmids [45, 76,77]. However, experimental demonstration of conjugation of A. baumannii plasmid-encoding bla OXA-40 -like genes into A. baumannii and E. coli have failed [76,78]. Nonetheless, the presence of this gene in plasmids belonging to the replicon group GR6, which are self-transmissible and in plasmids containing mob genes, which can be mobilized by self-transmissible plasmids, indicates a potential for horizontal dissemination by conjugation [52,78]. The lower GC content of the bla OXA-40 gene, as compared with the remaining genome of A. baumannii, suggests a different species as the source of this gene [3].
The bla OXA-40 gene found in the A. baumannii chromosome or in different plasmids is not embedded into the typical structures involved in DNA mobilization such as ISs but is instead flanked by conserved inverted repeats homologous to XerC/XerD binding sites [77,79,80]. These sites are the targets of the XerC and XerD recombinases, which are involved in site-specific recombination mechanisms. It is suggested that the bla OXA-40 gene can be mobilized by this mechanism within the A. baumannii isolates and then further spread by HGT. This hypothesis is supported by the fact that the same structure is located in different plasmids and in the chromosome [77,78] and that the same binding sites flank different DNA modules in other locations [79]. Identification of genes involved in the toxin/antitoxin system in plasmids harboring the bla OXA-40 gene might explain the wide dissemination and stability of the plasmids containing this gene [132].
This gene has been occasionally reported in other Acinetobacter species such as A. baylyi [133], Acinetobacter calcoaceticus [77], Acinetobacter haemolyticus [134] and A. pittii [125]. The same plasmid containing the bla OXA-40 gene flanked by the XerC/XerD binding sites has been found in A. baumannii, A. calcoaceticus [77], A. baumannii and A. haemolyticus [78], which suggests transfer of this genetic element from A. baumannii to the other species, by a HGT mechanism that has not yet been determined. The A. baumannii bla OXA-40 gene has also been detected in P. aeruginosa, in a plasmid similar to one found in A. baumannii [135].
OMVs of Gram-negative bacteria serve a wide number of biological functions and have been extensively studied in the delivery of proteins and toxins in order to target cells during infection, and therefore, are mostly associated with virulence factors [136]. Nevertheless, Rumbo and colleagues [59] have experimentally shown the release of OMVs containing the plasmid-encoded bla OXA-24 from carbapenem-resistant A. baumannii clinical strains that were able to transform a carbapenem-susceptible A. baumannii strain. The transformants maintained the capacity to release OMVs containing the plasmid-encoded bla OXA-24 . The OMVs-mediated transfer remains largely unexplored in the spread of antimicrobial resistance and this study has been the first experimental evidence that clinical isolates of A. baumannii may release OMVs carrying carbapenemase determinants as a mechanism of HGT.

OXA-58 (and OXA-58-Like)
The bla OXA-58 gene was identified for the first time in 2003 in a clinical A. baumannii isolate [120]. This gene is usually plasmid-encoded and it is found in different plasmids [67,81,82]. Despite the plasmid location, several attempts have failed to demonstrate conjugation of this genetic mobile element between A. baumannii or from Acinetobacter spp. to A. baumannii strains [50,51,120]. Therefore, the involvement of natural transformation, transduction and/or OMVs-mediated transfer could eventually explain the wide spread of the plasmid-encoded bla OXA-58 gene.
In A. baumannii the bla OXA-58 gene is surrounded by ISAba3 [3] however the upstream ISAba3-like element is often disrupted by different ISs such as ISAba1, ISAba2 [84], ISOur1, IS1008, IS15 [50] and ISAba825 [85], which enhance the expression of the bla OXA-58 gene. Despite the fact that the bla OXA-58 gene is surrounded by these mobile elements and that similar bla OXA-58 genetic contexts are found in different Acinetobacter species and strains [50, 84,141], there is little evidence that this structure has moved by transposition [3]. Rather, it is proposed that the acquisition of the bla OXA-58 gene and respective genetic context resulted from homologous recombination events, evidenced by the fact that different genetic structures containing the gene are flanked by two repeated sequences called Re27 [84]. The same recombination sites and the bla OXA-58 gene and respective gene surrounding has been identified in A. nosocomialis and A. pittii [50]. There is only one study that suggests that the bla OXA-58 gene flanked by an ISAba3 and an intact ISAba3-like elements as transposed into a plasmid in A. pittii [144]. The duplication of an ISAba2/ISAba3-bla OXA-58 -ISAba3 unit in clinical A. baumannii isolates has been linked to the action of two IS26 elements surrounding the previous unit; multiple copy numbers of the bla OXA-58 gene was correlated with increased Minimum Inhibitory Concentrations of carbapenems [86].
It has been suggested that the bla OXA-58 gene widely disseminated among A. baumannii strains has its source in a different species, reflected by the different GC content of the gene, when compared with its core genome [3]. However, the source has not been identified so far.
Recently, the mechanism of dissemination of OXA-58 CHDL was elucidated. Not being exactly a mechanism of gene transfer, the formation of OMVs by A. baumannii also contributed to the release of extracellular OXA-58 CHDL. OXA-58 is selectively released via OMVs after using the Sec-dependent periplasmic translocation transport system and its release was increased upon a carbapenem challenge [145]. These OMVs have a sheltering effect on carbapenem-susceptible bacteria allowing the unexpected survival of the cells in, for example, polimicrobial infections [145,146]. Yet not sharing genetic material, the overexpression of OXA-58 by A. baumannii protects other bacteria. However, as in the case of the transfer of bla OXA-40 OMVs, the genetic exchange cannot be discarded. One bla OXA-143 -like gene, bla OXA-253 , was also detected as part of a plasmid that resembles one plasmid carrying the bla OXA-40/24 gene; acquisition of the bla OXA-253 gene is suggested to have occurred by site-specific recombination in the XerC/XerD site [87].
None of the published studies have performed HGT assays that could be involved in the dissemination of these CHDLs. However, horizontal transfer might be inferred from the presence of the same gene in diverse genotypes [147].

Mechanisms Involved in the Movement and Dissemination of MBL Genes
Some MBLs are inserted in integrons that may be integrated in mobile genetic elements such as transposons; some are chromosomal, others are located in plasmids, which may more easily explain their dissemination. Integrons are genetic elements that can capture gene cassettes by site-specific recombination. Any gene cassette embedded in an integron that enters a cell by HGT mechanisms can potentially be recruited by an already existing integron [150] but this has not been experimentally demonstrated with MBL-encoding genes. The MBL-encoding genes present in A. baumannii can be found in a variety of different cassette arrays (http://integrall.bio.ua.pt/).

IMP-Type
From the 52bla IMP gene variants that have been described [151], at least 12 have been detected in A. baumannii [2,88] (GenBank accession number KT935306).
The A. baumannii class 1 integrons carrying bla IMP genes can be chromosomally- [89][90][91] or plasmid-located [88,92,93]. Class 1 integrons carrying bla IMP genes are also often found in Enterobacteriaceae, usually in transferable plasmids [61,152], which does not seem to be the case in A. baumannii [92,152]. Conjugation of plasmids-carrying the bla IMP-5 gene from A. bereziniae to A. baumannii has also been unsuccessful [153]. Conjugation of a bla IMP gene between A. baumannii strains has been reported by Takahashi and colleagues [154]. However, the authors did not detect plasmids in their donor isolates, often difficult to isolate in this species, but it opens up the possibility of the transfer of the bla IMP gene by another HGT mechanism. Nonetheless, bla IMP genes are not always embedded in integrons and conjugation of a plasmid-carrying the bla IMP-1 gene not associated with class 1 integron from A. baumannii to E. coli was reported [94].
The presence of the same class 1 integron in different bacterial species and genus is seen as an evidence of HGT [91,150,152].
Evidence of the intracellular movement by transposition of a class 1 integron-carrying the bla IMP-5 gene flanked by miniature inverted repeat transposable elements (MITEs) in A. baumannii can be inferred from the determination of target site duplications on both sides of the MITEs [95]; identical MITEs were found flanking the integron-carrying bla IMP-1 gene in A. baumannii isolates [96].
Intercellular transfer of chromosomally-located genes cannot be explained by conjugation. So far, there are no reports on the dissemination of carbapenem-resistance genes by natural transformation between A. baumannii strains, which could be explained by the fact that only very recently has this species been shown to be competent in specific conditions [57,58]. Nonetheless, the hypothesis that A. baumannii could be a source for carbapenem-resistance genes dissemination due to natural transformation events cannot be neglected. It was experimentally demonstrated that the chromosomal class 1 integron-carrying bla IMP-5 of the clinical A. baumannii 65FFC strain [89] was transferred by natural transformation to the integron-carrying A. baylyi SD2 (derivative of strain BD413) and incorporated into the recipient's chromosome by homologous recombination of the conserved regions of this class of integron [97]. From this study we might infer that carbapenem-resistance genes incorporated into class 1 integrons can be potentially acquired by all natural competent species that do not restrict the DNA uptake to the species level.

VIM-Type
Forty-eight bla VIM gene variants have been described so far [151]; among these, only six have been reported in A. baumannii [2]. The bla VIM genes embedded into class 1 integrons have been identified in a limited variety of gene cassette arrays in A. baumannii and are not frequently reported in this species; reports on these genes are more abundant and diverse in P. aeruginosa [155,156]. The bla VIM-2 gene has also been identified in a class 2 integron in A. baumannii (GenBank accession number LC107606).
Most of the studies do not report on the genetic environment of the class 1 integrons carrying the bla VIM genes [98][99][100], which hinders the unravelling of the horizontal mechanisms involved in their dissemination; these have been described in the chromosome of A. baumannii but without further analysis of the genetic context [101,102]. Up to now, there have been no studies showing the horizontal transfer of VIM determinants in A. baumannii. Horizontal transfer of VIM-encoding genes can be inferred from the presence of the same integron in genetically non-related A. baumannii isolates [98,157], in different species and genus [158]. Conjugation of bla VIM-1 genes from A. baumannii to E. coli has failed. However, it is not clarified if this failure is due to chromosomal location of the gene or due to inability of the plasmid to conjugate [98].

NDM-Type
Since the identification of the New Delhi metallo-β-lactamase (NDM) in 2008 in India, this MBL has been reported worldwide. The bla NDM genes have been recognized in a variety of species and genus and its location in diverse replicon type plasmids, such as IncA/C2 and IncFIIY plasmids, is the most likely mechanism in explaining this expanded dissemination [103]. Some examples are further mentioned.
Interestingly, a clinical case in China reports the finding of bla NDM-1 in E. coli, Citrobacter freundii and A. baumannii collected from the same patient but while bla NDM-1 was inserted in an identical plasmid in the Enterobacteria, in A. baumannii it was found to be located in a very large plasmid (>400-Kb) [104]. The genetic environment was not analysed. The bla NDM-1 determinant was also located in a plasmid of 100-Kb in an isolate from Brazil [105]. In Switzerland, the bla NDM-1 gene has been found in Enterobacteria located on conjugative IncA/C-or IncF-type plasmids.In all strains, part of ISAba125, previously identified in bla NDM-1 -negative A. baumannii, was found upstream of the bla NDM-1 gene, suggesting that the original dissemination occurred from A. baumannii [106]. This carbapenemase can also be found in animals. An isolate of A. baumannii positive for bla NDM-1 was collected from a lung sample of a pig with pneumonia and sepsis. The bla NDM-1 gene was located in a~47-Kb plasmid. This plasmid could be transferred by conjugation at the high frequency of 1.15 × 10 −2 per donor cell into E. coli J53 [107]. Therefore, it seems that in the case of bla NDM -type, conjugation plays an important role in the spread of carbapenem resistance and diverse conjugative plasmids are involved.
Nonetheless, the bla NDM-1 can also be chromosomally located. In fact, bla NDM-1 seems to be usually flanked by two ISAba125 insertion sequences, a structure corresponding to the composite transposon Tn125 [108,109,159]. This transposon was also associated with bla NDM-2 , suggesting that Tn125 is the major vehicle for the dissemination of bla NDM genes in A. baumannii [110].
A very recent study showed experimental evidence of the horizontal transfer of the Tn125 transposon carrying the bla NDM-1 gene by transduction from a carbapenem-resistant A. baumannii clinical strain (R2090) to a carbapenem-susceptible A. baumannii strain. The authors have excluded the involvement of the other three HGT mechanisms. The bla NMD-1 gene was chromosomally located and flanked by two different insertion elements (ISAb125 and ISCR21) that might explain the translocation of the gene into the strain. By full sequencing, it was possible to observe that this strain carried three intact prophages in the chromosome and it was suggested that activation of one of these prophages contributed to the transduction; nonetheless, it still remains to be tested if the phages detected are active [111]. So far, this is the first study demonstrating the possible involvement of these mobile elements in the dissemination of carbapenem resistance genes. There are about 20 A. baumannii phages described [161] and thus, bacteriophages may indeed contribute to the resistance evolution of A. baumannii under antibiotic selective pressure.

KPC-Type
KPC-type carbapenemases are rarely found in Acinetobacter spp. These beta-lactamases were identified for the first time in 3.4% (10/274) of clinical multidrug-resistant A. calcoaceticus-baumannii complex collected in 2009 in Puerto Rico. Sequencing of the bla KPC gene revealed the variants bla KPC-3 in seven isolates, bla KPC-4 , bla KPC-2 and a new variant bla KPC-10 in one isolate each [162]. A subsequent study identified 41 A. baumannii isolates with the bla KPC gene (the variant was not identified) [163]. Identification of the genetic background of bla KPC-3 in A. baumannii demonstrated that it was embedded into Tn4401b located in the chromosome within a 26.5 kb fragment, which included a KQ-like element similar to one described in a K. pneumoniae plasmid. This data suggests the acquisition of bla KPC-3 due to an ISEcp1 transposition event [112]. Very recently, whole genome sequencing revealed that the bla KPC-2 gene of A. baumannii (Puerto Rico), assigned to the international ST2 clone, was identified on a new truncated version of transposon Tn4401e (tentatively named Tn4401h), located in the chromosome within an IncA/C plasmid fragment derived from an Enterobacteriaceae, probably due to the insertion sequence IS26 [113]. This finding in the widely disseminated ST2 clone suggests further dissemination of bla KPC-2 in this species.
A study conducted in an Iranian Burn Care Center also identified the bla KPC gene in A. baumannii isolates but gene transfer was not demonstrated [164].

GES-Type
The GES enzymes are classified as extended-spectrum β-lactamases (ESBLs). However, some variants can act as carbapenemases [165]. Different levels of resistance or reduced susceptibility to carbapenems are observed with different variants [114,115,166]. Thirty-one bla GES gene variants have been described so far [151]; only six have been reported in A. baumannii [2,167] (GenBank accession number NG_049127).
The first bla GES gene reported in A. baumannii was the bla GES-11 , which was present as a gene cassette in a class 1 integron; this mobile element was embedded into a plasmid and transfer by conjugation between two A. baumannii isolates was observed [114]. The bla GES gene are embedded into class 1 integrons [116][117][118] and the majority are plasmid-located [115,116,119]. Nonetheless, chromosomal location has also been reported [116]. Transfer of the bla GES -encoding plasmid by conjugation between A. baumannii has been shown [115,117,119] but transfer to E. coli has failed [115].
The high GC content of the bla GES-11 gene as compared with the A. baumannii genome [114] can be seen as evidence of a different original source for this gene. The ocurrence of HGT has also been suggested due to the presence of the same bla GES gene in A. baumannii isolates belonging to different genotypes [167,168].

Conclusions
The fast increase of antimicrobial resistance in the last few decades has shown the remarkable ability of the genomic plasticity of Gram-negative bacteria, leading to the ineffective control of many infectious diseases and to a global public health problem of unexpected magnitude. The emergence and dissemination of antimicrobial resistance is an ample evidence of the biological response to the enhanced exposure to toxic molecules and dynamic genetic flux among bacteria. Mutations and lateral gene transfer mechanisms are the key for bacteria genome evolution and adaptation to multiple environments. Members of the Acinetobacter genus are considered environmental bacteria. For unclear reasons, A. baumannii emerged as an important nosocomial opportunistic pathogen. Not being a normal colonizer of non-hospitalized individuals, its reservoir remains unknown. Still considered a low virulence pathogen, infecting mostly immune-debilitated hosts, it acquired enormous clinical relevance when linked to the fast acquisition of resistance to the majority of antibiotic classes, including the last therapeutic resort antimicrobials, such as carbapenems. Since the 1980s, the carbapenems were seen as a new treatment option for serious resistant infections. However, its high activity has been hindered over time by the increase of carbapenem resistance. Many isolates of A. baumannii are often resistant to carbapenems and, as in other Gram-negative bacteria, have at their disposal a plethora of resistance mechanisms that may act in synergy. Production of carbapenemases is undoubtedly the most common mechanism of resistance to carbapenems. Indeed, this species produce the intrinsic carbapenemase OXA-51, that despite its low efficient hydrolysis, its expression and activity can be exacerbated by the upstream ISAba1 or/and work in conjunction with other mechanisms, as efflux pumps, to give carbapenem resistance [169]. Nevertheless, HGT plays an important role, not yet completely understood, in the dissemination of carbapenemase determinants.
A. baumannii seems to have a notorious ability to survive and it can use all the mechanisms of HGT and mobile genetic elements to spread resistance genes (Table 1). Understanding the extent of carbapenem resistance and how its mobilization takes place is essential to control the dissemination of these genes.
As in other Gram-negative bacteria, conjugation appears to have the greatest influence in the spread of carbapenem resistance determinants. Some of the CHDLs are disseminated by conjugative plasmids and the MBL bla NDM gene seems to be often located in different plasmids of diverse sizes. However, most reports associated bla NDM -type with Tn125, suggesting the intracellular "jumping" of the genetic elementintochromosome and to other plasmids of the new host bacteria. Indeed, bla NDM -type inserted in Tn125 has been found integrated in the chromosome [109], supporting the idea of the high dynamic of this gene, which might explain its global widespread in diverse genetic platforms.
For the first time, transduction has been recently suggested to occur as a mechanism of the spread of bla NDM genes in A. baumannii [111]. Not very widely reported, it is obvious that transduction is an important mechanism of lateral gene transfer when analyzing the genomes of diverse bacteria [170]. Moreover, many bacteria strains gained their virulence (production of some toxins) by the integration of stable prophage in their chromosomal DNA [170,171]. Therefore, it is evident that phage-mediated transduction can contribute to the transfer of resistance genes in A. baumannii. Nevertheless, phages are generally specific to the bacteria species [172], which might be a limitation for the spread to other species; however, interspecies transfer of genetic material by transduction was observed [173].
Natural transformation is usually an underestimated mechanism of lateral gene transfer. A barrier is that bacteria need to be naturally competent (and probably why it is thought that it occurs at a low frequency), i.e., to have protein-encoded genes necessary for the uptake of naked DNA, and then, stable integration in the chromosome (or reconstitution of plasmids in the bacterial cytoplasm) [174]. Some members of Acinetobacter spp. are naturally competent. It was demonstrated that integrons carrying the MBL bla IMP-5 gene can be inserted in the chromosome of A. baylyi, a naturally competent species, with a negligible biological cost to the recipient cell [97]. Some MBLs, especially from the IMP-and VIM-type families, are part of integrons and many of these isolates do not carry plasmids, suggesting that conjugation is not the involved mechanism (unless they lost the plasmid after transposition to the chromosome). IMP and VIM were the first metallo carbapenemases to be described and despite the increasing number of variants since then, they are not widely spread in A. baumannii. Natural transformation may contribute to their dissemination, even if at an initially low rate, which might be followed by high frequency dissemination [97]. This could explain the movement of integrons in the environment among diverse species [175,176] but not in clinical settings. However, recently clinical A. baumannii [57,58] and A. nosocomialis [177,178] strains were reported as naturally competent in very specific conditions, raising the hypothesis of a role of natural transformation in the resistance dissemination among clinical bacteria. Interestingly, this feature does not appear to be present in all members of the species or be characteristic of a clonal lineage of A. baumannii. Moreover, experimental conditions have a critical impact on the acquisition of foreign DNA [58,179]. Nevertheless, this is a mechanism that deserves to be explored with clinical strains and specifically what environmental conditions might influence the acquisition of naked exogenous DNA.
Another unusual reported HGT mechanism is related with the production of OMVs. The mechanisms of vesicle-mediated DNA delivery are not completely understood. A possibility is that competence proteins play a role in the uptake of DNA delivered by OMVs, since vesiculants (transformed cells obtained by vesicle-mediated gene transfer) were not formed with comdeficient mutants in A. baylyi. Interestingly, sub-inhibitory concentrations of gentamicin increased 10 times the transfer of plasmid DNA by OMVs, which might be explained by the altered surface potential by gentamicin-treated population [180].
Recent studies confirmed that A. baumannii are able to produce OMVs. It was demonstrated that OMVs released by A. baumannii are able to transfer bla OXA-24 [59], which reinforces the notion that the mediated-OMV mechanism must be taken into account for the dissemination of carbapenemase genes in this pathogenic species. Another study reported the extracellular delivery of OXA-58 in OMVs that shelter cells (even of other species) to the noxious action of carbapenems [145,146]. Not being a way of gene transfer, it is a way of protecting a diverse population of the toxic effects of antibiotics. The mechanisms of delivery from OMVs are still not very clear but the hypothesis suggested by Fulsundar et al. [180] and the recent findings of clinical competent strains, opens new fields of research.
Overall, the dissemination of carbapenemases in A. baumannii illustrates the complexity of the micro-dynamic flow of genes and their associated mobile genetic elements. Understanding these genetic mechanisms will allow gains and insights into the epidemiology of resistance genes and to predicting its spread, in order to act in the prevention of dissemination of antimicrobial resistance genes.