Sequences of Two Related Multiple Antibiotic Resistance Virulence Plasmids Sharing a Unique IS26-Related Molecular Signature Isolated from Different Escherichia coli Pathotypes from Different Hosts

Enterohemorrhagic Escherichia coli (EHEC) and atypical enteropathogenic E. coli (aEPEC) are important zoonotic pathogens that increasingly are becoming resistant to multiple antibiotics. Here we describe two plasmids, pO26-CRL125 (125 kb) from a human O26:H- EHEC, and pO111-CRL115 (115kb) from a bovine O111 aEPEC, that impart resistance to ampicillin, kanamycin, neomycin, streptomycin, sulfathiazole, trimethoprim and tetracycline and both contain atypical class 1 integrons with an identical IS26-mediated deletion in their 3´-conserved segment. Complete sequence analysis showed that pO26-CRL125 and pO111-CRL115 are essentially identical except for a 9.7 kb fragment, present in the backbone of pO26-CRL125 but absent in pO111-CRL115, and several indels. The 9.7 kb fragment encodes IncI-associated genes involved in plasmid stability during conjugation, a putative transposase gene and three imperfect repeats. Contiguous sequence identical to regions within these pO26-CRL125 imperfect repeats was identified in pO111-CRL115 precisely where the 9.7 kb fragment is missing, suggesting it may be mobile. Sequences shared between the plasmids include a complete IncZ replicon, a unique toxin/antitoxin system, IncI stability and maintenance genes, a novel putative serine protease autotransporter, and an IncI1 transfer system including a unique shufflon. Both plasmids carry a derivate Tn21 transposon with an atypical class 1 integron comprising a dfrA5 gene cassette encoding resistance to trimethoprim, and 24 bp of the 3´-conserved segment followed by Tn6026, which encodes resistance to ampicillin, kanymycin, neomycin, streptomycin and sulfathiazole. The Tn21-derivative transposon is linked to a truncated Tn1721, encoding resistance to tetracycline, via a region containing the IncP-1α oriV. Absence of the 5 bp direct repeats flanking Tn3-family transposons, indicates that homologous recombination events played a key role in the formation of this complex antibiotic resistance gene locus. Comparative sequence analysis of these closely related plasmids reveals aspects of plasmid evolution in pathogenic E. coli from different hosts.


Enterohemorrhagic
Escherichia coli (EHEC) and enteropathogenic E. coli (EPEC) are important diarrheagenic pathotypes responsible for substantial human morbidity and mortality [1]. Both carry a chromosomally-located island known as the locus of enterocyte effacement (LEE) that produces essential effector molecules required for the formation of characteristic attaching and effacing lesions on gastrointestinal epithelial cells [2]. EHEC are a subset of Shiga toxin-producing E. coli (STEC) expressing phage-derived Shiga toxins and accessory virulence factors, including intimin (Eae) and the plasmid-encoded enterohemolysin EhxA, responsible for the development of serious post-infection sequelae, such as haemorrhagic colitis and haemolytic uremic syndrome (HUS) [3]. Although most EHEC infections cause self-limiting bloody diarrhoea, in 5 to 7% of cases patients develop HUS, the leading cause of acute renal failure in children [4].
Ruminants are a key reservoir for both EHEC and atypical EPEC (aEPEC), an emerging cause of diarrhoea in both humans and animals globally [5,6]. More than 400 STEC serotypes have been described many of which are recoverable from faeces [7,8,9]. EHEC serotype O157:H7 is responsible for most cases of HUS particularly in the United States, Japan, Scotland, Canada and England. In the USA, O157:H7 EHEC infection causes approximately 73,000 illnesses resulting in several thousand hospitalizations and over 60 deaths per annum [10]. However, other EHEC serotypes including O26:H11/H-and O111:H8/H2/H-are also responsible for both large and sporadic outbreaks of serious disease worldwide [9,11,12]. EHEC O26:H -/H11 is a leading cause of HUS in many European countries [13,14] and has recently been associated with severe paediatric cases [15]. The validity of antibiotic therapy in the treatment of EHEC infection is controversial [16,17] with reports of antibiotics both inducing the SOS response and influencing the stability and subsequent release of Shiga toxin phage [18,19]. Despite these concerns, the German Society for Infection recommended the use of antibiotics for the treatment of patients infected with O104:H4, responsible for the world's largest HUS outbreak [20].
Multiple antibiotic resistance in EHEC, particularly O157:H7, O26:H-/H11 and O111:H8, is a serious concern [21,22,23]. Genetic elements encoding multiple drug resistance (MDR) are often associated with complex antibiotic resistance gene loci (CRL) comprising mobile genetic elements, often located on transmissible plasmids of the IncI and IncF groups [24,25]. IS26 in association with Tn3-type transposons plays a particularly important role in the evolution of MDR plasmids and chromosomal islands in the Enterobacteriaceae [24,26,27,28]. Homologous and site-specific recombination events involving these mobilizable CRL are shaping the rapid evolution of MDR in the gut microflora resulting in the more frequent isolation of complex mosaic plasmid backbones carrying multiple replicons, and antimicrobial drug resistance and virulence genes [24,29].
Here, we report the complete sequence of a 125 kb MDR plasmid, identified as pO26-CRL 125 , isolated from human O26:H-strain O6877. Like co-resident plasmid pO26-CRL (renamed here as pO26-CRL 111 ), pO26-CRL 125 confers resistance to Ap, Km, Sm, Su and Tm but it also encodes resistance to Tc. We also fully sequenced a 115 kb plasmid (pO111-CRL 115 ) from O111 aEPEC strain D275, isolated from a bovine with gastrointestinal disease. pO111-CRL 115 also shares the unique molecular signature created by IS26-mediated deletion affecting the structural integrity of the 3ʹ-CS. Our analyses show that plasmids pO26-CRL 125 and pO111-CRL 115 are essentially identical. These two plasmids were recovered from serologically different E. coli, belonging to different pathotypes, isolated from different animal hosts, yet remarkably share identical genetic architecture with substantial sequence similarity. Analyses of their minor sequence differences revealed several important aspects of in vivo plasmid evolution in pathogenic E. coli from different hosts.

Bacterial strains and plasmids
EHEC O26:H-strain O6877 was originally isolated in 1998 [31] and carries two MDR plasmids, pO26-CRL 111 described earlier [26] and pO26-CRL 125 (described here). Strain O6877 and O111 aEPEC strain D275 (isolated between 1999 and 2002) were part of a larger collection of 512 serologically diverse MDR E. coli including aEPEC, STEC and EHEC of human and bovine origin that were screened for the presence of class 1 integrons [30]. Plasmids pO26-CRL 125 and pO111-CRL 115 were isolated from O6877 and D275 respectively and sequenced.

Plasmid isolation
Plasmids from E. coli strains O6877 and D275 were conjugated with E. coli DH5α as previously described [32]. Gel electrophoresis of plasmid preparations showed that the wildtype strains carried several plasmids of different molecular size. As this potentially complicates sequencing studies, purified plasmid preparations from each strain were used in transformation using E. coli TOP10 as recipient following standard protocols (Invitrogen, Mulgrave, Vic, Australia). Transconjugants and transformants were tested for resistance to appropriate panels of antibiotics as described previously [26] and examined by gel electrophoresis to ensure each carried a single plasmid species.

Sequencing
Plasmids were isolated from E. coli DH5α or TOP10 hosts using the plasmid Midi Prep extraction kit (Qiagen, Doncaster, Vic, Australia). Plasmid sequencing was performed using Roche 454 GS FLX technology at the Ramaciotti Center for Gene Function Analysis. MDR plasmids were multiplexed with equal total DNA concentrations providing approximately 25 to 700x sequence coverage. Plasmids pO26-CRL 125 and pO111-CRL 115 were sequenced to 49 and 25x coverage, respectively. Plasmid sequences were assembled de novo using the Newbler v2.3 software package (454 Life Sciences, a Roche company, Branford, CT, USA). Contigs were broken by long repeat sequences, typically IS26 and were assembled into a single sequence by PCR between adjacent contigs. Contaminant chromosomal sequences were present at significantly reduced coverage, typically less than 3x. Plasmid contigs were initially annotated automatically using the RAST server [33]. Subsequently, the annotation was manually curated.

Stability assays
Stability experiments were performed essentially as previously described [38]. Briefly, overnight cultures in LB Tc (20 μg/ml) of transformed E. coli TOP10 strains harboring pO26-CRL 125 or pO111-CRL 115 (37 °C with shaking) were washed to remove the antibiotics and resuspended in saline. Each day 4.88 μl from fresh cultures were transferred to 5 ml LB in order to obtain about 10 generations per 24 h growth cycle (37 °C with shaking at constant speed). At selected time points, the growing cultures were plated both onto LB agar and LB agar supplemented with Tc (20 μg/ml) and incubated overnight at 37 °C. Determining the fraction of plasmid free cells in the population was done by calculating the ratio between the total number of cells (LB only) and the number of plasmid-containing cells (LB Tc). For each strain, stability experiments were performed in triplicate over 80 generations. The data were plotted using the Graph Pad Prism software (Graph Pad Software, Inc., San Diego, CA, US).

Conjugal transfer of pO111-CRL 115 and pO26-CRL 125 into JM109 Rif r Nal r E. coli
Conjugal capacity of pO26-CRL 125 and pO111-CRL 115 was tested by mating the E. coli TOP10 transformants with JM109 Rif r Nal r , a recipient rifampicin resistant E. coli JM109 made resistant to nalidixic acid (Nal) by multiple subculture on selective LB agar with increasing Nal concentrations (2 to 30 μg/mL). To confirm conjugation frequencies, a second assay was carried out following the same protocol with the inclusion in the mating mixture of E. coli HB101 containing the chloramphenicol resistant conjugative helper plasmid pRK600 [39] (not shown). Transconjugants were plated onto LB agar co-selecting for both the MDR plasmid (Ap, 100 μg/ml) and the recipient JM109 Rif r Nal r (Nal 30 μg/ml). Each mating was also plated onto LB agar selecting for the recipient only (Nal 30 μg/ ml). Donor only and recipient only negative controls were also plated onto the same media. The conjugal transfer frequency was calculated as the number of transconjugants per number of recipient cells. Transconjugants were tested for resistance on media supplemented with the following antibiotics: Rif (100 μg/ml), Ap (32 μg/ml), Km (10 μg/ml), Tc (20 μg/ml), and Su (550 μg/ml).

Nucleotide sequences accession number
The complete nucleotide sequences for pO26-CRL 125 and pO111-CRL 115 have been submitted to the GenBank database under accession numbers KC340959 for pO111-CRL 115 and KC340960 for pO26-CRL 125 .

Results
Sequence analysis of pO26-CRL 125 and pO111-CRL 115 pO26-CRL 125 and pO111-CRL 115 comprise 124,908 bp and 115,452 bp respectively. Both display an average G + C content of about 53%, around 2.5-3% higher than the G + C content of sequenced E. coli O26 and O111 chromosomes [40]. The general structure of pO26-CRL 125 and pO111-CRL 115 is shown in Figure 1. The two plasmids are almost identical except for: i) a 9,726 bp fragment found only in pO26-CRL 125 (leading region), ii) differences in the number of repeats in two unrelated repeat regions (RD1 and RD2), iii) a 135 bp deletion in the traH open reading frame (orf) in pO26-CRL 125 (RD3), iv) one small indel and two point mutations ( Figure 1; Table 1). pO26-CRL 125 and pO111-CRL 115 contain 147 and 136 predicted orfs respectively (Table S1). The plasmid backbones encode genes for replication, stability and maintenance, and conjugal transfer ( Figure 2; Table S1), and display a mosaic structure where modules characteristic of plasmids belonging to different incompatibility types are assembled in a novel arrangement. Large portions of the backbones share high sequence identity (>95%) with IncI EHEC plasmids, such as IncI1 pO113 (GenBank AY258503) ( Figure 2), with some modules showing homology to specific elements of plasmids belonging to incompatibility groups IncZ (pIE545; GenBank M93064.1), IncB (p3521; GenBank GU256641) and IncP-1α (pBS228; GenBank NC_008357) ( Figure 1). In this mosaic backbone structure, divergent G + C content for separate regions suggests assembly by multiple horizontal gene transfer events ( Figure  1). The accessory gene load in both plasmids consists of a CRL containing derivate Tn21 and Tn1721 transposons, and a virulence module, encoding putative virulence factors, including a novel serine protease autotransporter of Enterobacteriaceae (SPATE) (Figure 2).

IncZ and IncQ replicons
pO26-CRL 125 and pO111-CRL 115 contain two separate replication regions, a complete IncZ replicon identical to that of pIE545 (GenBank M93064.1) from Klebsiella pneumoniae and a partial IncQ replicon within the Tn21 derivative transposon in the antibiotic resistance module (Figure 1; Figure 3). The IncZ replicon contains genes coding for RepZ and RepY proteins Overall mean G + C content is indicated in percentages (GC%). Highest sequence homology to specific Inc groups is indicated by labeled horizontal brackets. RD1, RD2 and RD3 indicate the position of the regions of sequence where pO26-CRL 125 and pO111-CRL 115 differ substantially. Potential recombination signatures were identified by BLAST analysis. Regions of interest in plasmid architecture are labeled with red boxes (bold italics). b Depiction of the unique shufflon found in pO111-CRL 115 and pO26-CRL 125 . This module presents a single invertible fragment selecting for two possible PilV C-termini. The sequence of the characteristic 19 bp repeats, thought to be involved in the site-specific recombination events leading to pilV inversions, is shown. Drawings not to scale. and an RNAI encoding sequence with homology to that of other plasmids of the IncI complex (Table S1; Figure S1a). Inc RNAI encodes antisense RNA for repYZ mRNA and is one of the elements responsible for plasmid incompatibility [41,42]. IncZ plasmids are compatible with IncI1 and IncK plasmids but incompatible with IncB plasmids [41]. Consistent with these reports, IncZ plasmid pO26-CRL 125 (this study) and IncI1 plasmid pO26-CRL 111 [26] are co-resident in O26:H-EHEC strain O6877 and appear to be stably maintained. In both pO26-CRL 125 and pO111-CRL 115 , the defective IncQ replicon, comprising the repC gene and a partial repA sequence, is identical to the IncQ replicon in the Tn21 derivative transposon of plasmid pO26-CRL 111 (Figure 3).
The sequence separating the IncZ oriV from the downstream CRL shows homology to ColIb plasmids sequence (99%), except for a 3 kb fragment with low G + C content (41%) containing five orfs with little or no nucleotide homology to known sequences ( Figure 1; Table S1). A BLASTn search against sequences deposited in the NCBI database revealed two gaps in alignment bordering the 5´ and 3´ flanks of this intervening sequence likely representing DNA recombination signatures. Two of these orfs encode a putative novel StbD/E type II antitoxin/toxin system with homology to the RelB/E toxin-antitoxin stability systems of E. coli and other Gramnegative species [43,44]. The StbD antitoxin partner belongs to the PHD antitoxin family but showed no specific amino acid homology to the RelB component, while the predicted StbE protein showed 80 to 95% identity with representatives of the characterized cytotoxic translational repressor RelE of E. coli [43].

IncI modules form the majority of the plasmid backbone
The backbone of pO111-CRL 115 and pO26-CRL 125 contains plasmid maintenance, stability and transfer modules characteristic of IncI plasmids. The maintenance and stability modules of the two plasmids are identical (Figure 2) except for a difference (RD2) in the number of contiguous repeats (8 bp; aacaagat) found in a set between the col operon and the parA gene ( Figure 1; Table 1). Sequence comparison with other IncI plasmids in the NCBI database identified alignment gaps in the RD2 region, indicating a potential recombination hotspot. The transfer modules in pO26-CRL 125 and pO111-CRL 115 are also virtually identical and contain type IV conjugative transfer operons sharing extensive nucleotide sequence identity (95-100%) with the transfer regions of pO113 (GenBank AY258503) ( Figure 2). These modules comprise the traABC regulatory genes (~4 kb), the pil operon for thin pilus biogenesis (pilI to pilV; ~13 kb), and trb/tra genes for conjugal transfer (~29 kb). The oriT region, including the nikA and nikB genes and the 85 bp oriT sequence identical to the oriT of IncB plasmid p3521 (GenBank GU256641; Figure 1), was located in close proximity to the trb operon as in other IncI plasmids (R64 GenBank AP005147; pO113 GenBank AY258503; ColIb-P9 GenBank NC_002122.1). The oriT specific sequence with its two sets of imperfect repeats was identified by comparison with the well characterized oriT of IncI1 plasmid R64 [45] ( Figure  S1b). The tra operon presents the same gene arrangement observed in pO113 except for 1 kb of sequence between traU and traT displaying no nucleotide identity with other IncI plasmids. The traH sequence of pO111-CRL 115 is identical to that in pO113 and unique to these plasmids. In pO26-CRL 125 , a 135 bp deletion (45 amino acids) was identified in this orf (RD3; Table 1). This deletion does not disrupt the reading frame in the putative traH gene, therefore a functional conserved lipoprotein product can still be expressed in both plasmids.
The nucleotide sequence of the shufflon recombinase that follows the tra operon presents no homology to pO113 but is almost identical (92%) to a gene found in a phage sequence from Salmonella enterica serovar Hadar ICESe4 (GenBank FR686852). The unique shufflon of pO26-CRL 125 and pO111-CRL 115 consists of a single invertible segment (Figure 1b) that represents the 3´ variable portion of the PilV adhesin. Shufflons previously described in IncI plasmids contain four (A, B, C, D) or three 3´-terminal segments (A, B, and C or D) of the pilV orf [42,46], while the pO111-CRL 115 and pO26-CRL 125 shufflon contains only a B homologous segment and a second region with unique nucleotide sequence encoding an homolog of shufflon protein C (Figure 1b; Table S1). The only other shufflons presenting a single invertible portion have been described in Salmonella enterica serovar Typhi [46] and in S. enterica serovar Hadar ICESe4 (GenBank FR686852).

Structure of the complex antibiotic resistance module
Plasmids pO26-CRL 125 and pO111-CRL 115 were isolated from pathogenic E. coli of human and bovine origin by PCR  [26]. In pO111-CRL 115 , a region encoding an A/T rich segment and nine tandem repeats, identical to the vegetative origin of replication (oriV) of the IncP-1α plasmid pBS228 of Pseudomonas aeruginosa (GenBank NC_008357) [47], separates the mer module of the Tn21 derivative transposon from ΔTn1721 (Figure 3). The oriV-∆Tn1721 sequence displays a G + C content of approximately 63% suggesting it was acquired by lateral transfer. IncP plasmids possess a very specific and well characterized oriV (Figure 3) [48]. A functional IncP replicon depends on the interaction between the oriV locus and the trfA gene products [48], but trfA was not present in either pO111-CRL 115 or pO26-CRL 125 . In pO26-CRL 125 , the IncP-oriV sequence is identical to that of pO111-CRL 115 except for 160 bp of missing sequence, containing two DnaA binding sites, one IHF binding moiety and iterons five and four (RD1) (Figure 3). In both pO26-CRL 125 and pO111-CRL 115 , the junction between oriV and ΔTn1721 is identical to that found in pBS228 (GenBank NC_008357; Figure 3). Also, the tetR and tetA(A) genes for tetracycline resistance and the pecM orf within ΔTn1721 are identical to both the Tn1721 prototype sequence (GenBank X61367) [49] and the same genes found in pBS228. Unlike pBS228, however, in both plasmids a truncated version of the Tn1721 tnpA gene (∆tnpA) is present. Database searches indicated that sequence identical to the entire ΔTn1721 module is specifically found in IncP-1β plasmid pB10 (GenBank NC_004840) and in IncN plasmid pRSB201 (GenBank JN102341.1), both recovered from wastewater [50,51].
In pO26-CRL 125 and pO111-CRL 115 , the 5´ and 3´ sequences flanking the CRL were virtually identical to orfs associated with The CRL comprises a Tn21 derivate, encoding resistance to six antibiotics -aphA1 (kanamycin and neomycin), strA-strB (streptomycin), sul2 (sulfathiazole), bla TEM-1 (ampicillin), and dfrA5 (trimethoprim) -and to mercuric chloride (mer operon), and a truncated Tn1721 (ΔTn1721) transposon encoding tetA(A) (resistance to tetracycline). The Tn21 derivative contains a modified class 1 integron, carrying the dfrA5 cassette, but missing the 3´-CS due to IS26-mediated insertion of transposon Tn6026. Tn6026 harbors a copy of Tn6029, comprising a truncated Tn2 with bla TEM-1 and an RSF1010 segment with strA-strB and sul2, and a copy of Tn4352 carrying aphA1. In pO111-CRL 115 , Tn21 and ΔTn1721 are separated by a region identical to the oriV of the IncP-1α pBS228 (GenBank NC_008357). IncP plasmids have a very specific and well characterized oriV, comprising a G/C rich region, an A/T rich region, several tandem direct repeats (iterons) and host factors (DnaA, IHF) binding motifs . In pO111-CRL 115 , as in pBS228, one IHF and three DnaA (DnaA boxes) binding sites plus nine iterons are present. In pO26-CRL 125 , part of this oriV sequence, with iterons 5 and 4, two DnaA boxes and the IHF binding motif, is deleted (red frame). The boundaries of the CRL components are labeled with brackets. The insertion point of the multiple resistance locus within the colIb operon is shown in the figure. Transcriptional direction is indicated by arrows. Δ denotes a partially deleted or truncated gene. IR denotes inverted repeats: IR IS26 , inverted repeats of IS26; IR t-In2 and IR i-In2 inverted repeats flanking the class 1 integron found within transposon Tn21; IR mer21 , 38 bp inverted repeat of transposon Tn21 adjacent to the merR gene; IR tnp21 , 38 bp inverted repeat of transposon Tn21 adjacent to the tnpA gene; IR tet1721 , 38 bp inverted repeat of transposon Tn21 adjacent to the ΔtnpA gene. Drawing not to scale. doi: 10.1371/journal.pone.0078862.g003 the colicin operon in IncI plasmids (98 to 100% nucleotide identity to ColIb-P9 and pO113) (Figure 1; Figure 3). Comparative analysis with the homologous region of ColIb plasmids showed deletion of the 3´ end of the conserved yagA orf of unknown function, and of the 5´ end of the adjacent colIb gene, responsible for the production of colicin Ib (Figure 3). Insertion of Tn21-RD1-∆Tn1721 in the col operon is unique to pO26-CRL 125 and pO111-CRL 115 . However, in other IncI plasmids, IS and other mobile genetic elements are known to insert in this same location [42,52]. No identical matches to either the CRL in pO111-CRL 115 or in pO26-CRL 125 , with the iteron deletions, were found in public databases. However, the arrangement of these CRL is consistent with observations that the oriV-trfA junction acts as a hot spot for the insertion of antibiotic resistance transposons and IS in the IncP backbone [53,54]. In pO26-CRL 125 and pO111-CRL 115, target repeats characteristic of Tn21 or Tn1721 transposition are missing at the CRL insertion site. This indicates that homologous recombination events may have played here a role in the formation of the Tn21-RD1-∆Tn1721 CRL.
Virulence module: novel serine protease autotransporter sequence pO26-CRL 125 and pO111-CRL 115 contain several genes encoding both putative and established E. coli virulence factors ( Figure 2; Table S1). An almost complete col operon, encoding bacteriocidal properties as well as host-specific colicin immunity [55], is found immediately adjacent to ∆Tn1721 (Figure 1; Figure 3). In both plasmids, five genes, with putative roles in virulence, were identified clustered together in the region between the trb and tra operons as in EHEC plasmid pO113 [56] (Figure 1; Figure 2). Among these is a 4,089 bp orf that encodes a novel SPATE. Sequence analysis shows that the putative protein product displays all the features characteristic of SPATEs [57], including a signal peptide sequence, a peptidase S6 domain, a functional passenger domain, and a β-barrel autotransporter domain ( Figure S2). At the nucleotide level the gene shows only partial identity with known sequences and presents a unique nucleotide sequence for the peptidase S6 and passenger domains. At the amino acid level, the autotransporter domain is identical to that of EspP and EspC from EHEC, while the passenger domain shares 33% sequence identity with EspP of E. coli O157:H7 strain Sakai (GenBank NP_052685.1) ( Figure S2).

A 9.7 kb element distinguishes pO26-CRL 125 from pO111-CRL 115
The 9,726 bp fragment present in pO26-CRL 125 contains an IncI-associated genetic module implicated in the stable establishment in recipient cells following conjugation. This fragment is found in most IncI conjugative plasmids and contains an ssb gene coding for a single-stranded DNA (ssDNA) binding protein, a parB homolog for plasmid partitioning, psiB and psiA genes encoding SOS-response inhibition functions, and the conserved ardA and ccgAII orfs both with antirestriction function [42]. ccgAII is adjacent to a transposase encoding gene and four orfs with unknown function (Figure 4a; Table S1). Sequence analysis of the 9.7 kb region of pO26-CRL 125 revealed the presence of three imperfect repeats (R1, R2, R3) with homology to the ssDNA promoter Frpo sequence (Figure 4b) [58,59]. The 9.7 kb module was entirely missing in the pO111-CRL 115 backbone, but continuous sequence (R1 pO111 ) almost identical to the pO26-CRL 125 repeats was found at the point of insertion of the missing fragment (Figure 4a). The R1 pO111 sequence in fact is identical to parts of both R1 (bases 1 to 239) and R3 (last 222 bases). These observations are consistent with the 9.7 kb fragment inserting via a double reciprocal crossover event.
Frpo-containing sequences in sequenced IncI1 family plasmids are always associated with leading region genes exactly as seen in pO26-CRL 125 (Figure 4a) or with minor variations, and are mostly located in proximity to DNA modifying genes and the oriT and nikAB orfs as seen in both pO26-CRL 125 and pO111-CRL 115 (Figure 4a). Variants may contain transposase or IS elements between repeats (pETEC_73 GenBank NC_009788.1; R621a GenBank NC_015965.1; p746 GenBank NC_014234.1) or the hok-mok post-segregational killing system replacing the ccgAII gene (p1658/97 GenBank NC_004998.1; p53638_75 GenBank NC_010720.1). Uniquely in pO111-CRL 115 , R1 pO111 was not associated with any of the leading region components.
In order to determine whether the differences in the sequence of pO111-CRL 115 and pO26-CRL 125 compromise the ability of each plasmid to mobilize and establish in recipient cells, plasmid conjugation and stability experiments were performed in parallel. In mating assays using E. coli JM109 Rif r Nal r as recipient strain, both plasmids were able to selftransfer and showed comparable conjugation frequency ( Table  2). In transformed E. coli TOP10 strains, no significant difference in stability between pO111-CRL 115 and pO26-CRL 125 was observed after 80 generations ( Figure 5).

Discussion
The O26:H-EHEC strain O6877 is resistant to Ap, Km, Nm, Sm, Su, Tm and Tc and has been implicated as the causative agent of haemorrhagic colitis in an elderly patient [31]. We showed that two MDR plasmids, one a 124,908 bp IncZ plasmid pO26-CRL 125 reported here, and a second 111,481 bp IncI1 plasmid pO26-CRL, renamed pO26-CRL 111 , reported earlier [26], co-exist in strain O6877. These plasmids carry different virulence gene combinations and plasmid incompatibility markers, but all genes encoding antibiotic resistance except for Tc are localized within the IR boundaries of identical derivate Tn21 transposons. Five base pair direct repeats characteristic of Tn21 transposition sites were found flanking the derivative transposon in the traC gene in pO26-CRL 111 [26], but not in pO26-CRL 125 or related plasmid pO111-CRL 115 . The transposition (tnpA) and resolvase (tnpR) genes and the IR tnp21 and IR mer21 inverted repeats are intact in pO26-CRL 125 . In EHEC strain O6877, it is likely that the derivate Tn21 CRL originated in pO26-CRL 125 and transposed into pO26-CRL 111 . STEC are readily isolated from the faeces of all ruminant species and many carry the enterohaemolysin gene ehxA on large plasmids that also carry other virulence genes [6,7,8,60]. The spread of the derivate Tn21 transposon . Schematic representation of the IncI leading region of pO26-CRL 125 . a. A 9.7 kb fragment, carrying genes involved in plasmid establishment in recipient cells during the early phases of conjugation, is the main distinguishing feature between pO26-CRL 125 and pO111-CRL 115 . The insertion point in the backbone shared by pO26-CRL 125 and pO111-CRL 115 is indicated. The genes in the leading region are transcribed in the same direction (black arrows) from the single-stranded DNA promoter Frpo located in imperfect direct repeat sequences (R). The pO111-CRL 115 sequence contains a R sequence (R pO111 ) identical to portions of both R1 and R3, but no leading region genes. Drawing not to scale. b. Alignment of R sequences found in pO26-CRL 125 and pO111-CRL 115 with Frpo containing sequences from IncI plasmids ColIb-P9 (GenBank NC_002122) and R64 (GenBank AP005147). The base differences between repeat sequences are highlighted in yellow. Alignment was obtained using the EMBO ClustalO online tool [36] (http://www.ebi.ac. uk/clustalO). doi: 10.1371/journal.pone.0078862.g004 described in this study represents a mechanism by which virulence plasmids widely found in ruminant STEC populations can rapidly acquire resistance to multiple antibiotics in a single transposition event and is cause for serious concern. Little is known about the lateral transfer of EHEC plasmids and derivate mercury resistant transposons harboring CRL, between different pathotypes of E. coli. Recent studies provide evidence that food-producing animals represent one of the major reservoirs of E. coli causing urinary tract infections (UTI) [61]. Although antibiotics have limited use in the treatment of diarrheagenic E. coli infections, they are critical for management of extraintestinal infections. In this study, we also reported the complete sequence of a large MDR plasmid, named pO111-CRL 115 , isolated from bovine O111 aEPEC strain D275. Bovine strain D275 is resistant to the same panel of antibiotics as human O26:H-EHEC strain O6877 and was positive for a diagnostic PCR we developed to detect atypical class 1 integrons [30]. The PCR, which has one primer in intI1 and another in IS26, produced an 848 bp amplicon indicating that D275 carried a similar derivative Tn21 transposon to that found in O26:H-EHEC strain O6877 [26,30]. The same 848 bp amplicon was previously identified in a group of serologically diverse multiply antibiotic resistant E. coli from cattle on farms located far apart on the east coast of Australia and in two isolates recovered from human UTI patients [30]. More recently, we also identified the 848 bp amplicon using the intI1-IS26 PCR in several clinical isolates [62]. Composite transposons such as Tn6029/Tn6026, that carry bla TEM , strAB, and sul2 genes flanked by IS26, are globally disseminated and contribute to the evolution of CRL [26,28,63,64].
In this study, we showed that pO111-CRL 115 carries an identical copy of the derivate Tn21 transposon found in pO26-CRL 125 and pO26-CRL 111 . The CRL in pO26-CRL 125 and pO111-CRL 115 starts at the IR tnp21 and ends with IR tet1721 . Tn21 and Tn1721 both belong to the Tn3 family of transposons. Consequently, transposase-IR regions and their resolvases are functionally interchangeable and both generate 5 bp direct repeats at the site of transposition [65,66]. The structure of the CRL described here and the lack of 5 bp direct repeats suggest a complex evolutionary history involving laterally acquired DNA derived from disparate plasmids with different incompatibility, assembling within the IR boundaries via homologous recombination. These data reveal the critical role of derivate Tn21 transposons in spreading CRL among diverse plasmids carrying different virulence gene combinations in E. coli belonging to different pathotypes. PCR targeting key loci that arise during the molecular evolution of CRL can be exploited to track mercury resistance transposons containing CRL and the plasmids that carry them. We provide direct evidence of the latter here.
Despite their different host origins, pO26-CRL 125 and pO111-CRL 115 are essentially identical plasmids. The main difference between them is a 9,726 bp fragment found only in pO26-CRL 125 . The 9,726 bp fragment, known as the IncI1 leading region, is characteristic of large IncI conjugative plasmids and contains a number of genes, including ardA, psiAB, and ssb, that function, often in a host-specific manner, during conjugation to protect single stranded DNA as it first enters the recipient cell [67,68,69]. The SOS response is known to influence the frequency of genetic rearrangement, particularly within integrons, as well as protect bacteria from external stresses in promiscuous environments [68,70]. Trimethoprim and β-lactam antibiotics are known to induce the SOS response [18]. The leading region genes are transiently expressed during conjugation and their expression is regulated by the specific Frpo promoter situated within repeat regions (R1, R2 and R3 in pO26-CRL 125 ) [59]. To our knowledge, pO111-CRL 115 is the first plasmid shown to completely lack the 9.7 kb fragment, but retain a Frpo sequence located exactly where the pO26-CRL 125 leading region inserts. This suggests that the 9.7 kb fragment may actually be a novel mobile element. The presence of a putative transposase gene in the leading region sequence adds some weight to this speculation. We were unable to detect any difference in the conjugative efficiencies of pO111-CRL 115 and pO26-CRL 125 in standard laboratory conditions (Table 2). Further studies are needed to characterize the function of the 9,726 bp leading region and determine whether it can transpose independently.
Further comparative analysis of pO26-CRL 125 and pO111-CRL 115 sequences showed that regions of difference were invariably associated with features that either pertain to plasmid stability and may influence host range (leading region; traH) or confer flexibility to the plasmid backbone (repeats -RD1, RD2). The RD1 region, which separates the derivate Tn21 transposon from ∆Tn1721, is essentially identical to the oriV of the IncP-1α plasmid pBS228, originally described in Pseudomonas [47]. While the oriV-∆Tn1721 module is identical to one described in IncP-1β plasmid pB10 isolated from wastewater [51], the Tn21-RD1-ΔTn1721 arrangement in pO26-CRL 125 and pO111-CRL 115 is unique. In both plasmids the oriV sequence found in Tn21-RD1-ΔTn1721 is not expected to be functional as a replication feature because the trfA gene is missing [48]. Nonetheless, plasmids that carry oriV-type sequences are likely to gain an evolutionary advantage by virtue of the propensity of this region to serve as a hotspot for the acquisition of exogenous DNA [51,71].
pO26-CRL 125 and pO111-CRL 115 are conjugative, chimeric plasmids that contain genetic signatures common to both narrow host range IncI1 family plasmids and broad host range IncP family plasmids. The distinctive features in pO26-CRL 125 and pO111-CRL 115 , including the unique shufflon, the IncZ replicon, unique toxin/antitoxin system and unique SPATE, and the regions of difference distinguishing the two plasmids, may be representative of adaptive responses to a lifestyle where the bacteria that house these plasmids move between bovine and human gastrointestinal tracts. This existence provides an opportunity for genes to be acquired laterally from microbial populations found in the soil and in wastewater ponds generated by food animal production. Members of the well characterized IncP-1 family are frequently isolated in the environment and carry readily mobilizable antibiotic resistance modules that play an important role in the lateral transfer of accessory genes between unrelated bacteria [50,72,73]. Conjugative IncI plasmids display a narrow host range and function as vehicles for the dissemination of antimicrobial resistance determinants in pathogenic Enterobacteriaceae [25]. The sequence differences observed in pO26-CRL 125 and pO111-CRL 115 suggest a likely role in bacterial adaptation to rapidly changing environments or host-specific recognition [74].

Conclusions
This study and others [20] are indicative of how the problem of antibiotic resistance in humans may be linked to how antibiotics are used in food-animal production, aquaculture and horticulture. It is not possible to compartmentalize the problem because lateral gene transfer drives the movement of antibiotic resistance genes through these reservoirs and profoundly influences the delicate interplay between pathogenic and commensal bacterial populations [75]. Plasmids play a key role in the evolution of MDR Enterobacteriaceae, a key group in the struggle to curtail antibiotic resistance in the clinical environment [25]. Comprehensive analysis of complete sequences of plasmids carrying multiple antibiotic resistance genes is necessary to fully understand how CRL evolve and move through microbial populations in diverse settings. Figure S1. Features of the oriT region of plasmids pO26-CRL 125 and pO111-CRL 115 . a Inc RNAI sequence of pO26-CRL 125 and pO111-CRL 115 compared to that of other IncI family plasmids: EHEC plasmids pO26-CRL 111 [26] and pO113 [56], and the prototype IncI1 plasmid R64 [S1]. Inc RNAI is a small antisense RNA essential for control of IncI plasmids replication. Due to the trans-acting nature of this type of replication control the Inc RNAI determines also the incompatibility of IncI family members. About 70 bases in length, it is encoded downstream of the repYZ genes and regulates copy number by binding to a complementary mRNA sequence in the 5´ end of repZ and silencing repZ [S2,S3]. The four Inc RNAI sequences shown here are not identical but present conserved features (underlined) conferring the specific secondary stem-loop structure involved in target binding. IncZ plasmids are compatible with IncI1 plasmids [S3]. b minimum oriT sequence of pO26-CRL 125 and pO111-CRL 115 compared to that of R64. The oriT minimal region is located immediately upstream of nikA in IncI1 plasmids such as R64. It can be identified by the presence of two sets of inverted repeats (17 bp in blue, and 8 bp in red) involved in protein binding [45]. In pO26-CRL 125 and pO111-CRL 115 , the oriT sequence was immediately adjacent to the starting codon of nikA and contained both sets of repeats. The 8 bp repeats are identical to those of R64 while the 17 bp differ as it may be expected since the 17 bp inverted repeats constitute part of the recognised binding site for NikA and the NikA proteins of R64 and pO26-CRL 125 and pO111-CRL 115 share homology but are not identical. (TIFF) Figure S2.

Supporting Information
Comparison of novel SPATE sequence identified in pO26-CRL 125 and pO111-CRL 115 with characterized EHEC SPATEs. The novel SPATE sequence presents all the features characteristic of SPATEs: a conserved unusually long signal sequence (in bold magenta); a functional domain (underlined), containing a peptidase S6 domain (blue highlight) with a conserved serine protease motif GDSGS (bold, underlined), where the first S is the catalytic serine (red bold); and a very well conserved β-barrel autotransporter domain (green highlight). The functional domains are specific and show low homology to the characterized EHEC EspP from O157:H7 str Sakai E. coli (NP_052685.1), while the autotransporter domain is identical to that of EspP. Other conserved residues involved in protease activity are shown in red bold font. (TIFF)