E. coli C159/11 was isolated from calves on a dairy farm in the United Kingdom in 2004 (15,16). Investigation and manipulation of the C159/11 plasmid (pCT) was conducted in an E. coli DH5α transformant created for the current study. We also investigated 15 CTX-M-14–producing E. coli isolates collected during 2006–2009 from cattle feces on farms in different geographic locations in the United Kingdom and 15 E. coli clinical isolates from England (P. Hawley, unpub. data), Germany (17), Spain (11,18), the People’s Republic of China (19), and Australia (20) (Table 1).

Plasmid pCT was extracted from E. coli DH5α transconjugants by using an alkaline sodium dodecyl sulfate Maxi preparation (21) and cesium chloride density gradient centrifugation (22). Conjugation was by solid mating on a filter (Whatman, Maidstone, UK), by using rifampin-resistant E. coli (DH5α) as a recipient and selection of transconjugants on Luria-Bertani agar containing 50 µg/mL rifampin and 8 µg/mL cefotaxime.

Susceptibility of C159/11 and pCT transconjugants to a panel of antimicrobial drugs (ampicillin, cefotaxime, cefoxitin, chloramphenicol, ciprofloxacin, naladixic acid, streptomycin, and tetracycline) was determined by using a microtiter broth double dilution method (www.bsac.org.uk/_db/_documents/Chapter_2_Determination_of_MICs_2006.pdf). Susceptibility of E. coli DH5α to the antimicrobial drugs tested was also determined.

The plasmid DNA sequence was determined by using a 454/Roche GS FLX analyzer (Roche, Basel, Switzerland). The de novo assembly generated 93631 bases in 7 contigs by using the 454/Roche Newbler assembly program with an N50 of 52,495 bp. The sequence represents improved high-quality draft sequence (23) with no discernable misassembles having undergone multiple rounds of computational gap closure. Annotation was completed by using Artemis (Sanger, Cambridge, UK). Further comparative analysis of the DNA sequence used Double ACT and Artemis Comparison Tool (Sanger, Cambridge, UK), DNAstar (Lasergene; Madison, WI, USA) and BLAST (www.ncbi.nlm.nih.gov/guide).

Boiled bacterial cell lysates provided template DNA, 1 µL of which was added to PCR ReddyMix Master mixture (Abgene, Epson, UK). Typically, PCR conditions were 30 cycles of 95°C for 30 sec, 51°C for 30 sec, and 72°C for 30 sec. PCR was used to detect bla_CTX-M-14 and insertion sequences ISEcp1 and IS903 as described (24–26). All primers used are shown in Table 2. To determine whether the pCT bla_CTX-M-14 shares a common insertion site with bla_CTX-M-14 on other plasmids, PCRs were designed to amplify the sequence from bla_CTX-M-14 into both pCT flanking genes.

Using the pCT sequence, we designed primer pairs to amplify novel regions of pCT for rapid identification of potential pCT-like plasmids in CTX-M-14–producing bacteria (Table 2). PCRs to amplify DNA encoding the putative sigma factor, pilN gene, and shufflon recombinase were developed into a multiplex PCR. An additional primer pair was designed to a unique region of pCT and compared with other known sequences for amplification across coding sequences (CDSs) pCT008–pCT009 for further discrimination of pCT-like plasmids. Sequencing of the relaxase gene has been reported to further categorize plasmids within the IncI complex (11,28). A modified primer pair was designed for this region (nikB) by using sequence data from pCT and other related sequenced plasmids. Resulting amplicons were sequenced by using BigDye Terminator version 3.1 cycle sequencing (Applied Biosystems, Foster City, CA, USA) at the functional genomics laboratory of the University of Birmingham (Birmingham, UK) sequences were aligned by using MEGA 4.0 (29) for phylogenetic analysis (30). Primers amplifying group 9 bla_CTX-M genes were used as a positive control in each instance. The complete DNA sequence of plasmid pCT was assigned GenBank accession nos. F868832 and NC_014477.1.

The bla_CTX-M-14–carrying plasmid isolated from E. coli C159/11 was demonstrated to be conjugative by successful transfer to E. coli DH5α by using filter mating and previously determined to be of the incompatibility group IncK (16). Analysis of transconjugants showed resistance to β-lactam antimicrobial drugs as the only transferrable resistance phenotype. Whole-plasmid sequencing showed that pCT was 93,629 bp (Figure 1) with an average G+C content of 52.67%. Annotation of the plasmid showed 115 potential protein CDSs, 89 of which were homologous to proteins of known function (Technical Appendix ). No genes known to play a role in determining virulence were identified, and sequencing confirmed bla_CTX-M-14 as the only antimicrobial drug resistance gene on pCT. The pCT bla_CTX-M-14 gene is found between insertion sequences ISEcp1 and IS903 as described (31).

Most of the identified putative coding open reading frames are typical for an IncI group plasmid backbone (32). Two conjugal transfer systems were identified. The first is the more commonly described tra operon, which encodes the primary pilus for conjugation transfer. The pCT tra locus is analogous to the tra operon of R64/ColIb-P9 and is found in a similar conformation, although minor differences existed between the 3 plasmids throughout. The second is the pil locus encoding a thin pilus, which is believed to increase conjugation rates in liquid media. The tip of this thin pilus is variable, and the exact nature of the expressed epitope is determined by the orientation and order of pilV shufflon components (pCT_094-pCT_098), which can be inverted by the action of the recombinase protein (pCT_093) encoded downstream. Analysis of the pCT plasmid assembly showed that this region was present in multiple forms (data not shown), which is consistent with site-specific recombination mediated by a shufflon recombinase. The pCT shufflon potentially differs from that of the other closely related plasmids (pO113, pO26_vir and pSERB1) because each of these apparently has an inactive shufflon, which can be attributed to the absence of the recombinase or an insertion element within this CDS. The antirestriction and segregation genes on pCT are typical of this type of plasmid.

When we compared complete sequences deposited in GenBank of plasmids carrying bla_CTX-M genes with pCT, we found that outside the bla_CTX-M gene those plasmids with different replication mechanisms, such as those within the IncFII group or of the IncN group (pKP96) (33), have almost no DNA homology to pCT. The only other bla_CTX-M carrying IncI complex plasmid to be sequenced and deposited in GenBank thus far is a bla_CTX-M-3 carrying IncI plasmid pEK204 (EU935740) (32). pEK204 shares sequence conservation over ≈60% of the pCT genome, including most of the core IncI complex–related genes for replication or transfer. Further similarities were found in the minimal carriage of resistance genes (Figure 2, panel D).

The pCT genome was also compared with other IncI group sequenced plasmids to identify regions considered core or backbone and to determine novel encoded genes. IncI complex reference plasmids R64 (AP005147) and ColIb-P9 (AB021078) shared 99% identity with 64% of the pCT sequence primarily within genes involved in replication and conjugation (Figure 2, panel C). Other plasmids compared with pCT included R387, the Sanger IncK reference plasmid, which shared a high percentage identity to pCT across IncK-specific core genes (Figure 2, panel B).

We investigated novel genes by identifying regions of the pCT sequence absent from those plasmid sequences with most homology to pCT: pO26_vir (FJ38659), pO113 (AY258503), and the partially sequenced pSERB1 (AY686591), none of which carry bla_CTX-M genes. Both pO26_vir (168,100 bp) and pO113 (165,548 bp) are large plasmids that carry several virulence genes and share 99% homology with 85% and 83% of the pCT genome, respectively (Technical Appendix Table; Figure 2, panel A). pSERB1 also has high DNA sequence identity (91% of the pSERB1 deposited genome sequence has 99% identity to two thirds of the pCT sequence), however, because this plasmid is deposited in GenBank only as a partial sequence, the total identity cannot be assessed.

Fifteen CTX-M-14–producing E. coli isolates collected from different cattle farms around the United Kingdom were examined for pCT-like plasmids with a series of PCRs that amplified characteristic regions of pCT. Veterinary isolates I779 and I780 were obtained 2 years apart and from different geographic areas in the United Kingdom. These isolates carried plasmids encoding all 6 pCT regions that were investigated (bla_CTX-M-14, nikB, putative sigma factor, rci, pilN, and pCT008–009) and when compared with pCT had identical flanking regions adjacent to bla_CTX-M-14. Therefore, these plasmids were deemed pCT-like. Fifteen CTX-M-14–producing E. coli human clinical isolates from England, Germany (17), Spain (11,18), Australia (20), and China (19) also were examined for pCT (Table 1). No pCT-like plasmids were detected in the isolates from the United Kingdom and Germany. However, pCT-like plasmids were identified in 4 of 5 clinical isolates from Spain (C559; C567; C574; FEC383), 3 of 6 isolates from Australia (JIE 052, JIE 182, JIE 201), and the isolate from China (E. coli 8, CH13) because all sequences specific to pCT could be amplified by PCR. pCT-like plasmids have also been shown to be present in other clinical E. coli isolates from the United Kingdom by using the same multiplex assay (M. Stokes et al., pers. comm.). Sequencing of amplicons generated during PCR amplification of nikB showed pCT-like plasmids had nikB sequences with >98% DNA identity to the pCT nikB sequence and clustered when these sequences were used to construct a phylogenetic tree (Figure 3). nikB sequences from non–pCT-like plasmids clustered further from pCT within the phylogenetic tree (Figure 3). This analysis further supports the hypothesis that pCT has disseminated broadly between bacteria in animal and human ecosystems.