Escherichia coli of sequence type 3835 carrying blaNDM-1, blaCTX-M-15, blaCMY-42 and blaSHV-12

New Delhi metallo-β-lactamase (NDM) represents a serious challenge for treatment and public health. A carbapenem-resistant Escherichia coli clinical strain WCHEC13-8 was subjected to antimicrobial susceptibility tests, whole genome sequencing and conjugation experiments. It was resistant to imipenem (MIC, >256 μg/ml) and meropenem (MIC, 128 μg/ml) and belonged to ST3835. blaNDM-1 was the only carbapenemase gene detected. Strain WCHEC13-8 also had a plasmid-borne AmpC gene (blaCMY-42) and two extended-spectrum β-lactamase genes (blaCTX-M-15 and blaSHV-12). blaNDM-1 and blaSHV-12 were carried by a 54-kb IncX3 self-transmissible plasmid, which is identical to plasmid pNDM-HF727 from Enterobacter cloacae. blaCMY-42 was carried by a 64-kb IncI1 plasmid and blaCTX-M-15 was located on a 141-kb plasmid with multiple F replicons (replicon type: F36:A4:B1). blaCMY-42 was in a complicated context and the mobilisation of blaCMY-42 was due to the transposition of ISEcp1 by misidentifying its right-end boundary. Genetic context of blaNDM-1 in strain WCHEC13-8 was closely related to those on IncX3 plasmids in various Enterobacteriaceae species in China. In conclusion, a multidrug-resistant ST3835 E. coli clinical strain carrying blaNDM-1, blaCTX-M-15, blaCMY-42 and blaSHV-12 was identified. IncX3 plasmids may be making a significant contribution to the dissemination of blaNDM among Enterobacteriaceae in China.

Strain WCHEC13-8 was subjected to whole genome sequencing on the Illumina HiSeq platform. A total of 7,149,960 reads and 643,496,400 bases were obtained from the genome sequencing with a 49.8% GC content. Reads were assembled to 268 contigs that were ≥ 500 bp in length (N50 metric, 118,912 bp) and contained 5,145,308 bp nucleotides. To identify whether strain WCHEC13-8 had any new genes or not, its genome sequence was compared to other E. coli genome sequences including 83 complete and 2,159 draft genomes available in GenBank. Despite comparing with such a large number of E. coli genomes, the strain still had three unique genes that were located on contigs belonging to the chromosome. Two of the three genes were clustered together and resided in the centre of a 149-kb large contig, which was part of the chromosome as it shows significant identity (74% coverage and 97% identity) with the chromosome of E. coli strain E2348/69 (GenBank accession number FM18568) identified using the BLAST program. The two novel genes encoded hypothetical proteins with 420 and 346 amino acids, respectively. Unfortunately, the function of the two hypothetical proteins were not able to be predicted using both Protein BLAST and InterProScan programs. The remaining unique gene was 2,976 bp on a small 3,322 bp contig and contained both an endonuclease/exonuclease/phosphatase and a reverse transcriptase domain identified using the InterProScan program.
In addition to bla NDM-1 , strain WCHEC13-8 had other resistance genes including bla CTX-M-15 (an extended-spectrum β -lactamase [ESBL] which is globally distributed), bla SHV-12 (an ESBL gene), bla CMY-42 (a plasmid-borne AmpC cephalosporinase gene), bla OXA-1 (a non-ESBL oxacillinase gene), bla ampC (a chromosome-based AmpC gene), aac(6′)-Ib-cr (encoding an aminoglycoside acetyltransferase with low-level activity against fluoroquinolones), aac(3)-II (encoding an aminoglycoside acetyltransferase), mph (encoding macrolide 2′ -phosphotransferase I) and tetB (conferring resistance to tetracycline). Of note, CMY-42, the AmpC enzyme encoded by bla CMY-42 , differs from CMY-2 by a single amino acid substitution, i.e. Ser for Val at Ambler's position 211 10 . The co-existence of bla NDM-1 , bla CMY-42 , bla CTX-M-15 and bla SHV-12 , all of which encode a β -lactamase with the ability to hydrolyse broad-spectrum cephalosporins, in a single isolate has not been reported before. The co-presence of bla NDM-1 , bla CMY-42 and bla CTX-M-15 has been found in an E. coli of ST101 from a two-year-old child who had recently travelled to India 11 , while two E. coli isolates of ST167 in China have been found carrying bla NDM-1 , bla CMY-42 and bla CTX-M-14 6 . In addition, bla CMY-42 has also been found in two E. coli isolates from a river in India (GenBank accession numbers KJ661335 and KJ661336) and one from a surgical wound at a university hospital in Germany 10 , but STs of the three E. coli isolates have not been reported. Of note, a report from Egypt also claimed that bla CMY-42 has been detected in 8 E. coli isolates 12 , but the gene detected was 99% similar to bla CMY-42 and the amino acid sequence of CMY was not available for analysis. Therefore, It remains unclear whether these Egyptian isolates carried bla CMY-42 or another variant. Strain WCHEC13-8 had three plasmids, which were approximately 50, 60 and 140 kb in size, as revealed by S1 nuclease pulse-field gel electrophoresis (S1-PFGE). The three plasmids were completely circularised using PCR and Sanger sequencing. bla NDM-1 and bla SHV-12 were carried by a 54-kb plasmid, assigned pNDM1_EC8 here, of the IncX3 group. bla CMY-42 was located on a 64-kb IncI1 plasmid, assigned pCMY42_EC8 here, while bla CTX-M-15 was carried by a 141-kb IncF plasmid, assigned pCTXM15_EC8.
pCTXM15_EC8 contained multiple F replicons including one FIA, one FIB and two FII replicons. Based on the replicon sequence typing (RST) for IncF plasmids 13 , the replicon types of pCTXM15_EC8 were F36:A4:B1 (FII 36, FIA 4 and FIB 1) with both FII replicons belonging to the same type. The F36:A4:B1 type has also been seen in plasmid p6409-151.583 kb (GenBank accession number CP010372), which was recovered from an E. coli strain in Colombia. However, p6409-151.583 kb has an additional F31 type FII replicon and does not carry bla CTX-M . Furthermore, the link between pCTXM15_EC8 and p6409-151.583 kb could not be established and the origin of pCTXM15_EC8 remains unknown.
The genetic context of bla CMY-42 has not been described before. On pCMY42_EC8, bla CMY-42 is adjacent to the insertion sequence ISEcp1, which was truncated by the insertion of IS1, at upstream and a gene encoding outer membrane lipoprotein at downstream (Panel A, Fig. 1). The remaining part of ISEcp1 was found 24.7-kb further upstream of bla CMY-42 and was also truncated by IS1. After carefully examining the flanking sequences of ISEcp1 and IS1, it became evident that the complex genetic context of bla CMY-42 has been formed by the transposition of ISEcp1 and the insertion of two copies of IS1 followed by homologous recombination between the two IS1 (Panel B, Fig. 1) as explained as below. First, both of the two IS1 were flanked by a remnant of ISEcp1 and a part of the traB gene, which is involved in plasmid conjugation. The two remnants of ISEcp1 form a complete ISEcp1 plus a 9-bp repeat, which abuts the two IS1 and is characteristic of the insertion of IS1. Similarly, the two parts of traB make the complete traB plus the characteristic 9-bp repeat. It is therefore clear that both ISEcp1 and traB had been interrupted by the insertion of IS1 and the subsequent homologous recombination between the two copies of IS1 could result in the inversion of the intervening region (Panel B, Fig. 1). Second, it has been proven that a single copy of ISEcp1 is able to mobilise its downstream genetic components by using alternative sequences as the right-hand inverted repeat (IRR) 15 . By comparing genetic contexts of bla CMY-2 and its variants available in GenBank using BLAST, the same 4 kb element comprising ISEcp1, bla CMY , blc (encoding an outermembrane lipoprotein) and sugE (encoding a quaternary ammonium compound-resistance protein) has been found on different plasmids but with varied abutting sequences. This element is bounded by a 14 bp sequence (AACCAGAAAGTCGA) at one end, which shows some similarity with the left-hand inverted repeat (IRL) of ISEcp1 (Panel B, Fig. 1) and might have served as the alternative IRR (IRR2) for ISEcp1 to mobilise bla CMY to different locations. Indeed, when we put the two remnants of ISEcp1 back together, we found that the finQ gene (the transcriptional inhibitor of plasmid transfer) has been interrupted by the ISEcp1-bla CMY-42 -IRR2 region with the presence of 5-bp direct target repeats (DR), which characterises the insertion of ISEcp1 (Panel B, Fig. 1). This confirms that the mobilisation of bla  has been mediated by ISEcp1 by misidentifying its IRR.
In strain WCHEC13-8, bla NDM-1 was able to be transferred to E. coli strain J53, suggesting that pNDM1_ EC8 was a self-transmissible plasmid. None of the other resistance genes listed above except bla SHV-12 were co-transferred with bla NDM-1 . PCR-based replicon typing (PBRT) failed to assign pNDM1_EC8 to a replicon type as all replicon typing PCR reactions were negative. pNDM1_EC8 was an IncX3 plasmid and IncX3 has not been integrated into the PBRT scheme. pNDM1_EC8 is identical to pNDM-HF727 (GenBank accession number KF976405), an IncX3 plasmid carrying bla NDM-1 from an Enterobacter cloacae strain in China 9 .
IncX3 plasmids are narrow-host-range plasmids of Enterobacteriaceae 16 . A few IncX3 plasmids carrying ESBL genes (bla SHV-11 or bla SHV-12 ) and/or carbapenemase genes (bla KPC-2 , bla NDM-1 or bla NDM-5 ) from various species of Enterobacteriaceae in several countries have been completely sequenced 9,17-21 . In particular, IncX3 plasmids carrying bla NDM-1 and bla SHV-12 have been identified in various species of the Enterobacteriaceae at multiple locations in China. The completely-sequenced IncX3 plasmids except pKP13d are almost identical in their backbone sequences with less than 9 single nucleotide variations between any two plasmids (Fig. 2). The remarkable similarity between IncX3 plasmids suggests that a common IncX3 plasmid has spread to different continents, acquired the carbapenemase gene that is prevalent locally, e.g. bla KPC-2 in USA and bla NDM-1 in China, and then were transformed as a vehicle mediating the interspecies spread of carbapenemase genes in local settings.
IncX3 plasmids carrying bla NDM-1 and bla SHV-12 have also been found in strains causing an outbreak in China 22 . IncX3 plasmids carrying bla NDM-1 and bla SHV-12 have also been found among the Enterobacteriaceae in the United Arab Emirates (UAE). Of note, one of the UAE strain was recovered two years earlier than the first Chinese strain that harboured an IncX3 plasmid carrying bla NDM-1 and bla SHV-12 was isolated. It remains unclear where the exact origin of emergence of IncX3 plasmids carrying bla NDM-1 and bla SHV-12 is. More studies are warranted to characterise the prevalence of IncX3 plasmids in the Enterobacteriaceae and to understand the emergence and spread of IncX3 plasmids 23 .
The genetic context of bla NDM-1 on pNDM1_EC8 is identical to those on pNDM-HF727 and several other IncX3 plasmids from China. In such a context, bla NDM-1 is adjacent to an IS5-interrupted IS3000-truncated ISAba125 at upstream and is linked with bla SHV-12 at downstream (Fig. 3). bla SHV-12 is flanked by IS26, which could form a composite transposon with the potential to mobilise bla SHV-12 into different locations.

Conclusions
An ST3835 E. coli clinical isolate carrying multiple genes, e.g. bla NDM-1 , bla CTX-M-15 , bla CMY-42 and bla SHV-12 , which are able to confer resistance to broad-spectrum cephalosporins, was identified in China. Carbapenemase gene bla NDM-1 was carried by a self-transmissible IncX3 plasmid in this strain. IncX3 plasmids may have served as a common vehicle mediating the dissemination of bla NDM among the Enterobacteriaceae. The recently-identified plasmid-born AmpC gene bla CMY-42 was in a complex genetic Δ and shapes with a forked tail represent truncated genes or insertion sequences. The broken line represents a 27.4 kb region between traC and finQ, which is not scaled. The putative alternative IRR (IRR2) of ISEcp1 is depicted as a red pole and the alignment of the IRL of ISEcp1 and the IRR2 is shown. ISEcp1 and traB are both interrupted into two parts. The 9-bp nucleotide repeats belonging to ISEcp1 are depicted in blue and those of traB are in red, while the 5-bp nucleotide repeats belonging to finQ is shown in bold. Panel B, The proposed scheme for the formation of the context of bla CMY-42 . The ISEcp1-bla CMY-42 -IRR2 region (structure 1) is inserted into the finQ gene of the traA-trbA region (structure 2) on an IncI1 plasmid and then generates the structure 3 with the 5-bp characteristic DR (GATAA). Two copies of IS1 (structure 4) are then inserted into traB and ISEcp1, respectively, generating the structure 5 with the 9-bp DR (CCGTCAATA for traB and CGTTTGAAG for ISEcp1), which is the characteristic of IS1 insertion. The subsequent homologous recombination between the two copies of IS1 mediates the inversion (shown by two broken arrows) of the intervening region (grey part) and then forms the context of bla CMY-42 on pCMY42_EC8 (structure 6). context, which was formed by the transposition of ISEcp1 and the insertion of two copies of IS1 followed by homologous recombination between the two IS1.

Strain and in vitro susceptibility. Strain WCHEC13-8 was a clinical isolate recovered in West China
Hospital of Sichuan University, Chengdu, China. Species identification and in vitro susceptibility test were performed using the Vitek II automated system (bioMerieux, Lyon, France). MICs of imipenem, meropenem, amikacin, ceftazidime and ciprofloxacin were also determined using the microdilution broth method followed recommendations of the Clinical Laboratory Standards Institute 24 .
Detection of carbapenemase-encoding genes. The strain was screened for acquired carbapenemase-encoding genes bla GES , bla KPC , bla IMP , bla IMI , bla NDM , bla OXA-48 and bla VIM using PCR as described previously [25][26][27][28] . The complete coding sequence of bla NDM was amplified with an additional pair of primers (NDM-up/NDM-dw) 25   The trees are generated using MEGA based on the complete sequence of IncX3 plasmids. Panel A, a tree of IncX3 plasmids including pKP13d, which carries bla KPC-2 in a K. penumoniae strain from Brazil (GenBank accession number CP003997). pKP13d is distinct from other IncX3 plasmids in sequence. Panel B, a higher resolution tree of IncX3 plasmids excluding pKP13d. The carbapenemase and ESBL genes, the host species, the locations of recovery and the GenBank accession number of each plasmid are shown. For host species, CF refers to Citrobacter freundii, EC to E. coli, ECL to E. cloacae, KP to K. pneumoniae and RP to Raoultella planticola. The plasmid carrying bla NDM-1 identified in this study is identical to pNDM-HF727, which is underlined. The GenBank accession number of pNDM5_0215 17 is not available. Genome sequencing and analysis. Genomic DNA of strain WCHEC13-8 was prepared using QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) and was subjected to whole genome sequencing with a ca. 140 × coverage using the Hiseq 2500 Sequencer (Illumina, San Diego, CA, USA) following the manufacturer's protocol at the Beijing Genomics Institute. Reads were assembled to contigs using the Spades program 29 . The Prokka program 30 was employed for annotating the genomic sequence. To identify whether this strain had unique genes, its assembled sequence was compared to the complete chromosome sequences of all of the 83 E. coli strains with their complete genome available in GenBank using the Gegenees program 31 . Potential unique genes were called using a stringent threshold of a 0.9 or higher score generated by Gegenees and then were confirmed or rejected using the BLAST program (http:// blast.ncbi.nlm.nih.gov) against the 2,159 draft genomes of E. coli (accessed on December 18, 2014) to test whether they were absent from the E. coli draft genome database or not. Function of the products encoded by unique genes was predicted using both the Protein BLAST and InterProScan (http://www. ebi.ac.uk/interpro/) programs. The plasmids carrying bla NDM-1 , bla CMY-42 or bla CTX-M-15 were completely circularised in sequence with intervals between contigs being filled by PCR with primers designed based on available sequences and Sanger sequencing.
Strain typing. Phylogenetic group (A, B1, B2 and D) typing was performed as described previously 32 .
Strain WCHEC13-8 was assigned to a ST using the assembled genomic sequence to query the seven alleles of the multi-locus sequence typing scheme for E. coli (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli) 33 . Clonal complexes were assigned according to the founder ST identified using eBURST.
Mating experiments and plasmid typing. Conjugation experiment was carried out in broth using azide-resistant E. coli strain J53 as the recipient. Transconjugants were selected on plates containing 4 μ g/ml meropenem plus 150 μ g/ml sodium azide. The presence of bla NDM in transconjugants was confirmed using PCR and ERIC-PCR was used for further distinguishing transconjugants from the donor strain. Plasmid DNA that was prepared from the transconjugant using alkaline lysis was subjected to PCR-based replicon typing (PBRT) 34 . Sequences of IncX3 plasmids were retrieved from GenBank and aligned using MEGA (version 6). Maximum likelihood phylogenetic tree was constructed using MEGA using backbone sequences. IncI1 plasmid was assigned to ST and the replicon types of the IncF plasmids were assigned after querying their sequence against the plasmid MLST databases (http://pubmlst.org/ plasmid/). Plasmids of the same type as pCTXM15_EC8 were identified using concatenated sequence of the IncF replicon alleles of pCTXM15_EC8 for BLAST. S1-PFGE. S1-PFGE was performed to determine the number and size of plasmids carried by strain WCHEC13-8 as described previously 35 . Briefly, agarose plugs containing whole-cell DNA of strain WCHEC13-8 were treated with 8 U of S1 nuclease (Fermentas, Thermo Scientific; Waltham, MA, US) and the reaction was stopped by adding 0.5 M EDTA (pH 8). PFGE was conducted with a 1% SeaKem Gold agarose gel (Lonza, Basal, Switzerland) using a CHEF DRII system (Bio-Rad, Hercules, CA, US) at 14°C, with a 6-V/cm current and run times of 12 h at switch time of 5 to 40 s followed by 8 h at switch time of 3 to 8 s. MidRange I PFG Marker (NEB, Ipswich, MA, US) was used for size estimation.
Nucleotide sequence accession number. Reads and the Whole Genome Shotgun project of WCHEC13-8 genomic sequence have been deposited into DDBJ/EMBL/GenBank under accession SRR2012643 and LCWG00000000, respectively. The sequence of pCMY42_EC8 and pCTXM15_EC8 has been deposited into DDBJ/EMBL/GenBank under accession numbers KP789019 and KP789020, respectively. . Genetic context of bla NDM-1 in E. coli strain WCHEC13-8 . ISAba125 was interrupted by the insertion of IS5. Genes shown from the left in order are mpr (encoding a zinc metalloproteinase), bla NDM-1 , ble (mediating bleomycin resistance), trpF (encoding a phosphoribosylanthranilate isomerase), dsbC (encoding an oxidoreductase), ctuA1 (encoding an ion tolerant protein), groL (encoding a chaperonin subunit), ygbI (a putative dehydrogenase gene), a truncated ygbJ (a putative DEOR-type transcriptional regulator gene), bla SHV-12 and a truncated umuD (encoding a mutagenesis protein).