Klebsiella pneumoniae Multiresistance Plasmid pMET1: Similarity with the Yersinia pestis Plasmid pCRY and Integrative Conjugative Elements

Background Dissemination of antimicrobial resistance genes has become an important public health and biodefense threat. Plasmids are important contributors to the rapid acquisition of antibiotic resistance by pathogenic bacteria. Principal Findings The nucleotide sequence of the Klebsiella pneumoniae multiresistance plasmid pMET1 comprises 41,723 bp and includes Tn1331.2, a transposon that carries the bla TEM-1 gene and a perfect duplication of a 3-kbp region including the aac(6′)-Ib, aadA1, and bla OXA-9 genes. The replication region of pMET1 has been identified. Replication is independent of DNA polymerase I, and the replication region is highly related to that of the cryptic Yersinia pestis 91001 plasmid pCRY. The potential partition region has the general organization known as the parFG locus. The self-transmissible pMET1 plasmid includes a type IV secretion system consisting of proteins that make up the mating pair formation complex (Mpf) and the DNA transfer (Dtr) system. The Mpf is highly related to those in the plasmid pCRY, the mobilizable high-pathogenicity island from E. coli ECOR31 (HPIECOR31), which has been proposed to be an integrative conjugative element (ICE) progenitor of high-pathogenicity islands in other Enterobacteriaceae including Yersinia species, and ICEKp1, an ICE found in a K. pneumoniae strain causing primary liver abscess. The Dtr MobB and MobC proteins are highly related to those of pCRY, but the endonuclease is related to that of plasmid pK245 and has no significant homology with the protein of similar function in pCRY. The region upstream of mobB includes the putative oriT and shares 90% identity with the same region in the HPIECOR31. Conclusions The comparative analyses of pMET1 with pCRY, HPIECOR31, and ICEKp1 show a very active rate of genetic exchanges between Enterobacteriaceae including Yersinia species, which represents a high public health and biodefense threat due to transfer of multiple resistance genes to pathogenic Yersinia strains.


Introduction
Plasmids are important contributors to the rapid acquisition of antibiotic resistance by pathogenic bacteria through their ability to acquire resistance genetic determinants and to transfer them among bacteria belonging to the same or different genera and species [1,2]. These resistance genes are usually located inside transposable elements, integrons, or ISCRs (insertion sequence common regions), which facilitate their mobility at the molecular level and at least in the case of transposons could help expanding plasmid's host range [1,3,4,5]. In the past few years, considerable efforts have been made to completely sequence resistance plasmids and understand their biological properties [1].
K. pneumoniae, an important opportunistic pathogen, is the causative agent of community-acquired infections and more frequently of infections in the urinary tract and soft tissue, pneumonia, septicemia, and meningitis in hospitalized patients [6]. Hospital outbreaks caused by multiresistant K. pneumoniae strains have been described throughout the world [7,8]. It has been proposed that Crohn's disease as well as ankylosing spondylitis are linked to Klebsiella infection possibly through the mechanism of molecular mimicry [9]. K. pneumoniae has also been identified as the causative agent of liver abscess [10]. However, K. pneumoniae infections are especially dangerous in neonatal wards [6]. With the rise in the number of strains harboring resistance genes, usually plasmid-borne, the mortality of K. pneumoniae infections has increased and treatment has become more complicated [11,12,13]. A multiresistant K. pneumoniae isolate that caused high mortality in a neonatal ward harbors the selftransmissible plasmid pMET1, which carries the genetic determi- nants for resistance to several aminoglycosides and b-lactams [12].
In this paper we report the complete sequence and analysis of pMET1, and a comparative study with the Yersinia pestis cryptic plasmid pCRY, the mobilizable high-pathogenicity island found in Escherichia coli strain ECOR31 (HPI ECOR31 ), an integrative conjugative element (ICE) that is a possible progenitor of highpathogenicity islands in other Enterobacteriaceae including Yersinia species, and an ICE (ICE Kp1 ) present in a clinical K. pneumoniae isolate. Our results indicate that pMET1 and pCRY are closely related and most possibly can be mobilized between Yersinia and other Enterobateriaceae, and be stably maintained in these bacteria. Furthermore the presence of elements of these plasmids in HPI ECOR31 and at least another ICE is an indication of a very active genetic exchange between these bacteria, which represents a high public health and biodefense threat due to transfer of multiple resistance genes to pathogenic Yersinia strains.

Results and Discussion
The 41,723 bp multiresistance plasmid pMET1 (GenBank accession number EU383016) includes the replication region, a type 4 secretion system (T4SS), a partition system, and a transposable element that includes two genes coding for aminoglycoside modifying enzymes and two b-lactamases. Table 1 and Fig. 1 summarize the characteristics of pMET1 including the predicted functions of the ORFs on the basis of amino acid sequence comparison.
The transposable element Tn1331.2 The plasmid pMET1 includes an 11,042-bp transposon, named Tn1331.2 [12] that is highly related to Tn1331 (see Fig. 2) [14,15]. Tn1331 can be considered an evolutionary product of Tn3 by the addition of a 3,047-bp DNA region, which has the structure of the variable portion of the integrons, flanked by 520-bp direct repeats (Fig. 2). This region includes three resistance genes with a structure consisting of aac(69)-Ib-attC-aadA1-bla OXA-9 -attC, potentially defining two gene cassettes including aac(69)-Ib-attC and aadA1-bla OXA-9 -attC [14,16,17]. As a result of the presence of these three genes and bla TEM , cells harboring Tn1331 are resistant to several aminoglycosides and b-lactams. Possible mechanisms for the generation of Tn1331 have been discussed previously [16,18]. Tn1331.2 has a perfect duplication of the 3,047-bp DNA region and as a consequence a 520-bp including a C-terminal fragment of tnpR is found as three direct repeats flanking the potential gene cassettes (Fig. 2). As it is the case for Tn1331 [14], the Tn1331.2 tnpA differs from the Tn3 version of the gene in the 9-nucleotide repeats that in Tn3 code for the amino acid sequence GFHGFH. Tn1331.2 includes only one copy of this 9-bp fragment coding for the amino acid sequence GFH.
The duplication of the 3,047-bp DNA fragment including the resistance genes could be a strategy to increase the gene dosage when the transposon is present in a low copy number plasmid. Mechanisms of gene duplication when they are flanked by direct repeats have been described before [19,20,21]. Although we have observed a transition from Tn1331.2 to Tn1331 by the loss of one of the 3,047-bp repeats in the laboratory, we do not know if this is a process that happens in nature (dashed arrow, Fig. 2). While for many years no evolutionary derivatives of Tn1331 other than the duplication to render Tn1331.2 had been isolated, there was a recent report of such a phenomenon. The transposon Tn1332, isolated from Pseudomonas putida, can be considered as Tn1331 with the insertion of three DNA fragments that include the metallo-ßlactamase gene bla VIM-2 , and the insertion sequences ISPpu17 and ISPpu18, respectively ( Fig. 2) [22].

Replication and partition
Blastn analysis revealed a pMET1 region of high homology to the Y. pestis 91001 plasmid pCRY putative replication region [23] (schematically shown in Fig. 1A). The characteristics of this region include two AT-rich DNA segments flanking ORF1, named RepA, whose amino acid sequence showed 89% identity and 95% similarity to the pCRY RepA putative protein ( Fig. 3A and B). Lower but still significant similarities were found to other replication proteins from plasmids isolated from Buchnera aphidicola [24,25] and Y. enterocolitica plasmid pYVe227 [26]. To functionally identify the replication region of pMET1, three fragments 12, 34, and 45 (coordinates 41298-1655, 522-2039, and 1-2039 respectively) (Fig. 3A) were ligated to the kanamycin resistance fragment and used to transform E. coli TOP10. Kanamycin resistant colonies were obtained only when the recombinant clones included the fragments 12 or 45 (Fig. 3A). These results were confirmed by cloning all three fragments using pCR2.1 as vector, and attempting to transform the DNA polymerase I deficient E.  (Fig. 3A). Therefore the results of both experiments defined a replication region encompassing nucleotides 1 to 1655, which includes two essential features: one complete ORF, named repA, and the AT-rich sequence between nucleotides 1 and 522 (Fig. 3A). This latter region likely includes the ori locus. These two features are conserved in the pCRY replication region: the AT rich sequence is conserved as the G+C content profile is almost identical to pMET1, and repA shows 95% similarity and 89% identity to pMET1 RepA when translated (Fig. 3B). On the other hand, the overall identity at the nucleotide level between pMET1 and pCRY for this locus is 89%. This high relatedness between the pMET1 and pCRY replication regions strongly suggests that pMET1 can be replicated in Yersinia species. Plasmid pMET1 includes a putative partition region that has the general organization of that of plasmid pTP228, also known as the parFG locus [27]. The pMET1 partition region encompasses nucleotides 22,100 to 23,011 and includes two ORFs (ORF25 and ORF26, Table 1) encoding two putative proteins with strong homologies to ParF and ParG. The ParF homolog is a 205-amino acids protein that has shown significant homology with numerous proteins that belong to the type Ib subgroup of the ParA superfamily that is related to the MinD subgroup of cell division proteins [28,29,30]. These proteins are Walker-type ATPases that include a deviant motif A, which contains a conserved 'signature' lysine (XKGGXXKT) [29,31,32]. This motif as well as motif A' green: replication and partition; red: antibiotic resistance; purple: virB/pilX-like; blue: transposition; grey: unknown. Since pCRY is smaller than pMET1, to represent it in circular form a dotted line was added to fill the gap. Solid line represents non-homologous DNA. B. Comparison of a pMET1 region with chromosomal HPIs or ICEs is shown using a linearized version of the plasmid. The HPIs shown are those from E. coli ECOR31 (HPI ECOR31 ) [43], K. pneumoniae NTUH-K2044 (ICE Kp1 ) [44], and Y. pestis KIM (HPI Yp ) [43]. The diagram shows the HP core regions, which are not at scale and are represented as in [43], and the RB-HPIs. The sequence described in this manuscript has been deposited in GenBank, accession number is EU383016. doi:10.1371/journal.pone.0001800.g001 [29] are shown in Fig. 4A. The maximum pMET1 ParF identities and similarities were found with those encoded by the cryptic Y. enterocolitica plasmid p29930 [33] (90% similarity and 77% identity), the multidrug resistance plasmid pTP228 from Salmonella enterica subsp. Enterica serovar Newport [27] (87% similarity and 72% identity), and the Erwinia amylovora Ea88 plasmid pEA29 [34] (85% similarity and 73% identity).
Unlike with ParF, the pMET1 ParG has significant homologues only in p29930 (70% similarity and 53% identity), pTP228 (71% similarity and 58% identity), and pEA29 (68% similarity and 51% identity), generally known as ParB-like proteins. Studies on the pTP228 ParG protein showed that it acts as a dimer, may interact in a sequence-specific manner with a direct repeats-containing DNA region located upstream of parF known as parH, stimulates ATP hydrolysis by ParF, and modulates polymerization of ParF [35]. In addition, the ParG dimer binds an operator and acts as an autoregulator of the parFG operon [31]. Structures of parB-like proteins alone or in complex with DNA have been determined [36,37,38]. Other plasmids that have ParF homologues include proteins that are functional analogs but heterologs to ParG. The nucleoprotein complex has been recently called ''segrosome'' and is essential to ensure proper partition of the plasmid molecules [31]. Analysis of the pMET1 region upstream of the parF and parG genes showed an organization similar to that found in pTP228, which includes the parH locus and a putative operator (Fig. 4A) [31,39]. Interestingly, although the ParG proteins from pTP228 and pMET1 share homology, the repeats found upstream of parF in pMET1 are unrelated to those in pTP228. This region includes eight imperfect direct repeats that stretch over the promoter region. A graphical representation (sequence logo) of the multiple alignment of these repeats is shown in Fig. 4B. A region that includes some or all of these direct repeats could play the role of parH in pMET1 partition. Fig. 4A also shows the putative promoter, the 235 sequence is located within two partially overlapping direct repeats and upstream of this structure the repeat number 6 is part of an inverted repeat. One or more of these structures could function as operators, binding of ParG to one or both of these two structures may result in inhibition of transcription of parF and parG. Within the putative parH region there is also a perfect inverted repeat (Fig. 4A) unrelated to the putative ParG binding sites. The function of this structure is unknown.

Conjugation
Type 4 secretion systems (T4SSs) encoded by plasmids are involved in bacterial conjugation and transport of effectors to eukaryotic cells during infection [40]. Bacterial conjugation plays an important role in dissemination of antibiotic resistance genes at the cellular level [41,42]. Previous studies indicated that pMET1 is self-transmissible [12], analysis of its nucleotide sequence shows that the transfer region of pMET1 lies within a ca. 16-kbp DNA segment, the ATG of the first gene, virB1, occurs at coordinate 3534 and the ATG of the last gene, ardC (which is transcribed opposite to all the rest) at coordinate 19,631. A block of predicted coding products (named VirB1-VirB11) are highly homologous to proteins that make up the mating pair formation complex (Mpf) in a large number of plasmids and other DNA elements with the highest identity with those in the Y. pestis 91001 plasmid pCRY [23], the high pathogenicity island HPI ECOR31 [43], and ICE Kp1 , an ICE found in the primary liver abscess causing K. pneumoniae NTUH-K2044 [44] (see Fig 1B, Table 1, and Table 2). The putative functions of these proteins are discussed in several reviews [40,42,45]. In all four cases the virB3 and virB4 genes are fused. An homolog of published VirB7 proteins is absent in all four strucutures. However, a 45-amino acids hypothetical protein (ORF9) is present between virB6 and virB8 in all four elements. This high conservation and the synteny suggest that this ORF might play a role equivalent to that of VirB7. The pMET1 vir-like genes are followed by ORFs of unknown function and the predicted DNA transfer (Dtr) group of genes that potentially code for MobB, MobC, and an endonuclease. While the pMET1 MobB and MobC showed 99% identity to those in the Y. pestis plasmid pCRY, the endonuclease showed highest identity with that of the K. pneumoniae plasmid pK245, which carries genes coding for resistance to quinolones and extended-spectrum b-lactamases [46], and the cryptic Y. enterocolitica plasmid p29930 [33]. Interestingly, no significant homology was found with the nuclease encoded by pCRY. Further analysis of the pMET1 mobilization region showed a structure oriT-mobB-mobC, similar to those in other elements such as the mobilizable HPI ECOR31 [43] or the plasmid CloDF13 [47]. The pMET1 region upstream of mobB has 90% identity with that of the HPI ECOR31 . As it is the case in this element, there are two nic (or oriT) sites in pMET1 due to their location inside inverted repeat 2 (IR2). It is then possible that either strand of pMET1 can be transferred to the recipient cell, an unusual phenomenon [48]. Furthermore, studies on the plasmid R1162, which also possesses two oriTs, suggested that transfer can be initiated at one of them and terminated at the other with the inverted repeat playing a role in termination [49].
Downstream of the endonuclease gene, there is an ORF that is transcribed in the opposite direction to all the genes described above with homology to the p29930 ardC [33], which codes for a putative antirestriction protein. The pSA ArdC protein has been studied in some detail. It has been shown to bind ssDNA and to protect single-stranded but not double-stranded plasmid DNA in vitro against the activity of HhaI, an enzyme that cleaves both single-and double-stranded DNA. It was proposed that ArdC is transported to the recipient cell with the transferred singlestranded DNA and protects the incoming single-stranded DNA from the host endonucleases [50].
Comparative analysis of pMET1 and the Y. pestis pCRY plasmid pCRY is a 21,742-bp cryptic plasmid recently isolated from Y. pestis strain 91001 [23]. Comparative analysis of pMET1 and pCRY DNA sequences revealed highly similar backbones (see Fig. 1A). These analyses showed that the ca. 18-kb fragment including the replication region and the T4SS involved in conjugation of pMET1 shares high sequence homology as well as gene synteny with that of pCRY. These results suggest that they have a common ancestor and it would not be surprising if other plasmids with the same backbone but carrying other mobile genetic elements including resistance genes were found in other Enterobacteriaceae isolates. In the particular case of Y. pestis, our findings, taken together with the recent description of other The ability to be maintained in E. coli C2110 (a polA mutant) of the recombinant plasmids made using pCR2.1 as vector is indicated to the right by a + or 2 sign. The ability to generate kanamycin resistant colonies in E. coli TOP10 of the indicated fragments when ligated to the aph cassette from pUC4K is also represented by a + or 2 sign. B. BLASTP comparison of the amino acid sequences of the putative RepA proteins from pMET1 and pCRY. doi:10.1371/journal.pone.0001800.g003 multiresistance plasmids of Yersinia that share their backbone with plasmids of Salmonella [51], and the finding of a common ancestry between the Y. pestis pMT1/pFra plasmid and a cryptic plasmid from a clinical isolate of S. enterica serovar Typhi [52] indicate that there must be an active transfer of plasmids between Yersinia and other Enterobacteriaceae by direct contact or through other intermediate bacteria resulting in a high risk of dissemination of antibiotic resistance genes to pathogenic Y. pestis strains.
Comparative analysis of pMET1 and the integrative and conjugative element of E. coli ECOR31 The pMET1 T4SS is related to those found in the ICE Kp1 and HPI ECOR31 (Fig. 1B). A study on HPI ECOR31 showed that it shares common structures, such as the genes for synthesis of the siderophore yersiniabactin, with other high-pathogenicity islands found in Enterobacteriaceae including Yersinia species; but differs in the presence of a DNA region within an end of the element, known as RB-HPI ECOR31 (right border-high pathogenicity island ECOR31) (see Fig. 1B, ICEs from E. coli ECOR31 and Y. pestis KIM) [43]. This region includes a pMET1-related T4SS, which made HPI ECOR31 the first pathogenicity island to be found harboring a complete set of conjugative plasmid genes [43]. Furthermore, the complete HPI ECOR31 can be excised from the chromosome, circularized, and transferred indicating that it is an integrative conjugative   [43,53]. These properties taken together with the structure of high-pathogenicity islands found in Yersinia species and other Enterobacteriaceae led to the suggestion that HPI ECOR31 is their progenitor [43]. It is possible that this is also the origin of the T4SS found in pCRY and pMET1, or conversely it has been captured from a plasmid related to pCRY or pMET1. These findings demonstrate the high rate of genetic exchanges that occurs between Yersinia species and other Enterobacteriaceae. These considerations also speak in favor of a high risk of acquisition of antibiotic resistance or virulence traits by pathogenic Yersinia species.