ABSTRACT
The resistome of the multidrug-resistant Escherichia coli strain 271 carrying the plasmid-mediated blaNDM-1 carbapenemase gene was analyzed by high-throughput genome sequencing. The p271A plasmid carrying the blaNDM-1 gene was 35.9 kb in size and possessed an IncN-type backbone that harbored a novel replicase gene. Acquisition of the blaNDM-1 gene on plasmid p271A had been likely the result of a cointegration event involving the transposase of Tn5403. The expression of blaNDM-1 was associated with the insertion sequence ISAba125 likely originating from Acinetobacter baumannii. E. coli 271 accumulated multiple resistance determinants, including five β-lactamase genes (comprising the extended-spectrum β-lactamase CTX-M-15), two 16S RNA methylase ArmA- and RmtB-encoding genes, and the qepA gene encoding an efflux pump involved in resistance to fluoroquinolones. These resistance genes were located on three additional plasmids, of 160 kb (IncA/C), 130 kb (IncF), and 110 kb (IncI1). In addition, several chromosomally encoded resistance determinants were identified, such as topoisomerase mutations, porin modifications and truncations, and the intrinsic ampC gene of E. coli that was weakly expressed. The multidrug resistance pattern observed for E. coli 271 was therefore the result of combined chromosome- and plasmid-encoded mechanisms.
INTRODUCTION
Recent reports show unambiguously that the blaNDM-1 gene encoding the metallo-β-lactamase (MBL) NDM-1 is spreading worldwide (31, 35, 36). The blaNDM-1 gene was initially identified in Klebsiella pneumoniae and Escherichia coli isolates but has been recently reported to occur in Citrobacter freundii, Morganella morganii, Providencia spp., and Enterobacter cloacae isolates (24, 32, 40). It is noteworthy that the blaNDM-1 gene was most often reported to be located on plasmid supports. Additionally, the blaNDM-1 and blaNDM-2 genes have been identified recently on the chromosome of Acinetobacter baumannii and, very recently, in the genome of Pseudomonas aeruginosa isolates (14, 18, 19). Spread of blaNDM-1 is considered a serious threat, because the enzyme it encodes possesses a broad spectrum of activity.
Our study was initiated by isolation of a multidrug-resistant E. coli strain (strain 271) in Australia from a patient transferred from Bangladesh (38). The E. coli 271 isolate was resistant to all β-lactams, including carbapenems, all aminoglycosides, fluoroquinolones, nitrofurantoin, and sulfonamides, remaining susceptible only to tetracycline, tigecycline, fosfomycin, and colistin (38).
We previously showed that E. coli 271 harbored the blaNDM-1 gene on a plasmid (p271A) that does not carry other resistance determinants and that was not typeable by using the PCR-based replicon typing (PBRT) technique (4).
In order to identify which were the resistance mechanisms involved in the multidrug resistance pattern of E. coli 271, and to identify the plasmid backbone that carried the blaNDM-1 gene, the whole genome of this isolate was determined.
MATERIALS AND METHODS
High-density pyrosequencing and sequence assembly.The complete sequencing work flow of a Illumina genome analyzer IIx system (Illumina, Inc., San Diego, CA) was performed by the DNAVision company (Gosselles, Belgium). In brief, preparation of the genomic DNA of strain 271 was performed by using a Qiagen DNA purification kit (Qiagen, Courtaboeuf, France). The genomic DNA preparation (3 to 5 μg) was fractionated by sonication, and a library was prepared with 12 cycles of PCR enrichment using an Illumina paired-end DNA sample prep kit (Illumina). The amplified libraries were sequenced with an Illumina GAIIx instrument using six Illumina 36-cycle sequencing kits (v4) and a PhiX control kit v2, on a 2- by 100-bp paired-end run. A total of 25,455,000 reads were obtained, corresponding to 3,202 contigs (>50 bp) with a 234-fold median coverage. Velvet2 assembler software was used to assemble the sequences that were read (47).
Genome assembly and annotation.Reads from each sample were trimmed to remove poor-quality sequence by using the following procedure: keeping only reads for which the quality of all the 50 first bases was greater or equal to 10 in both reads; for each read, trimming the tail of the sequence from the first position for which the quality was smaller than 10; trimming poly(A) sequence artifacts; and trimming adapter sequences.
Genome comparison.The BLASTp algorithm was used to search for protein similarities by using as a reference the E. coli IAI1 chromosome sequence (GenBank accession no. NC_011741). The criterion used to evaluate the deduced amino acid sequence homology was >50% similarity at the amino acid level and >50% coverage of protein length.
Nucleotide sequence accession number.The p271A plasmid sequence was submitted to the GenBank database and can be found under accession number JF785549.
RESULTS AND DISCUSSION
General characteristics of E. coli 271.Sequence analysis of the seven housekeeping genes selected for multilocus sequence typing (MLST) (45) showed that E. coli 271 belongs to the MLST ST101 and is therefore classified in the phylogenetic subgroup B1 that corresponds to a commensal strain. It is noteworthy that two other ST101-type E. coli isolates producing NDM-1 have been recently identified in Germany and Canada (32, 33), with a link with the Indian subcontinent in both cases. In contrast, the community-acquired E. coli GUE that had been recovered from a French woman who was living in India and that coexpressed NDM-1 and CTX-M-15 belonged to the ST131 type (6, 37).
We previously identified a series of antibiotic resistance genes in E. coli 271, including the extended-spectrum β-lactamase (ESBL) blaCTX-M-15 gene and the broad-spectrum β-lactamase blaTEM-1 gene that were located on a conjugative plasmid that was distinct from that carrying the blaNDM-1 gene (38). Also, we showed that E. coli 271 harbored a total of four plasmids whose sizes had been determined by agarose gel electrophoresis analysis after plasmid DNA extraction using the Kieser technique (21). Mating-out experiments were performed as described previously (38) by using as selective agents either ampicillin (50 μg/ml), amikacin (30 μg/ml), kanamycin (30 μg/ml), or chloramphenicol (30 μg/ml) that allowed us to obtain the four respective E. coli transconjugants harboring plasmids p271A, p271B, p271C, and p271D. Analysis of the antibiotic resistance patterns of these different E. coli transconjugants, together with PCR assays with primers specific for the different resistance genes identified (sequences available upon request) and with analysis of data obtained from genome sequencing, allowed us to more precisely determine the sizes of the four plasmids, namely, p271A to p271D, which were, respectively, ca. 35, 110, 130, and 160 kb, and to assign all the resistance genes to their respective plasmid supports. Incompatibility groups of plasmids p271B, p271C, and p271D corresponded to IncI1, IncF, and IncA/C, respectively. The blaCTX-M-15 and blaTEM-1 genes were located on the IncF plasmid (see Table 1).
Features of plasmid p271A carrying the blaNDM-1 gene.Plasmid p271A carrying the blaNDM-1 gene was 35,947 bp in size, with an average G+C content of 50.2%, which was very different from that of the blaNDM-1 gene (61.5%). It possessed 34 open reading frames, including those corresponding to the replication, conjugation, and antibiotic resistance modules. Its scaffold corresponded to that of IncN-type plasmids, commonly identified among enterobacterial isolates, such as the well-characterized R46 plasmid (7). In particular, it showed similarities with plasmid pJIE137, which was recently identified at the origin of the blaCTX-M-62 gene acquisition in a K. pneumoniae isolate in Australia (48). Also, the overall structure of plasmid p271A shared significant homologies but to a lesser extent than with other IncN-type plasmids, including the two plasmids P9 and P12 isolated from United States isolates and, respectively, harboring the blaKPC-2 and blaKPC-3 genes that have been fully sequenced (3, 13). Consequently, plasmids p271A and pJIE237 should be considered IncN2-type plasmids, a denomination that means their scaffolds are similar to those of IncN plasmids, but both of them constitute a subgroup of IncN plasmids, with ca. 70% similarity over the entire backbone. As shown in Fig. 1, the tra locus coding for proteins involved in conjugation was complete. As observed for other IncN plasmids, the traK-traJ-traI locus was separated and located on the right extremity of the ori sequence. The stbC, stbB, and stbA genes encoding plasmid stability proteins were identified, as observed for other IncN plasmids (3). As observed only with a few IncN plasmids recently described that harbored unrelated antibiotic resistance genes, the integrative element that carried the resistance module in p271A was inserted into the fipA gene, causing the same deletion of the 3′ extremity of that gene (Fig. 1). That observation is noteworthy considering that the corresponding FipA protein has been demonstrated to inhibit the conjugative transfer or mobilization of some other plasmids (43). More importantly, FipA was shown to delay the growth of E. coli (43). The lack of production of FipA could therefore result in an increase in the capacity of the bacterial host to accumulate plasmids and could also result in a growth advantage.
As a consequence of the integration of the resistance module, and unlike other IncN plasmids, p271A lacked the ccgC and ccgD genes encoding products involved in the protection of plasmid DNA from type I restriction enzymes, which could make this plasmid more susceptible to those endonucleases (Fig. 1) (8).
Plasmid p271A harbors a new replicase gene.Although the overall scaffold of p271A matched with that of IncN-type plasmids, the corresponding replicase gene was not identified. This observation agrees with the negative results previously obtained using the PBRT technique aimed to type a plasmid scaffold by targeting its replicase gene. A new replicase gene was identified, encoding a 295-amino-acid-long protein named Rep271. In silico analysis showed that the gene encoding Rep271 was identical to that recently identified from plasmid pJIE137 carrying the blaCTX-M-62 gene (48). In silico analysis showed that the most closely related replicase proteins were those of plasmids identified from Salmonella enterica serovar Typhimurium (68% amino acid identity) (1, 10) and also from Pseudomonas aeruginosa (68%) (23) and Xanthomonas axonopodis (68%) (22). While this work was in progress, the complete sequence of a blaNDM-1-bearing plasmid from an E. coli isolate in Hong Kong was published (15). That plasmid possessed an IncL/M-type scaffold that shares no genetic link with the plasmid p271A described here. This indicates that completely different genetic platforms are currently vehiculating the same blaNDM-1 gene.
Since Rep271 shared features of replicase proteins identified from nonenterobacterial rods, and since the blaNDM-1 gene has been identified also in A. baumannii, it prompted us to evaluate whether plasmid p271A could replicate in other Gram-negative species, such as A. baumannii and P. aeruginosa. Electrotransformation assays performed as described previously (39, 41) using A. baumannii CIP70.10 and P. aeruginosa PU21 recipient strains and a ticarcillin (100 μg/ml)-based selection revealed that p271A replicated in A. baumannii and may confer to that species a high level of resistance to carbapenems (data not shown). In contrast, no transformant was obtained using P. aeruginosa as the recipient. This constitutes the very first in vitro demonstration of transfer of the blaNDM-1 gene between unrelated Gram-negative species.
Detailed analysis of the blaNDM-1 genetic environment.Preliminary findings showed that the blaNDM-1 gene was bracketed by two novel insertion sequences, namely, ISEc33 and ISSen4, belonging to the IS630 and IS3 families, respectively (38). Also, we previously showed that the promoter sequences involved in blaNDM-1 expression were not part of ISEc33 (38). A detailed analysis of the 194 bp separating the blaNDM-1 start codon from ISEc33 identified a DNA fragment corresponding to insertion sequence ISAba125 (Fig. 2). ISAba125 is an insertion sequence involved in the acquisition of the blaOXA-58 carbapenemase gene and has been identified for Acinetobacter genomospecies 3 (11) and for A. baumannii, with seven copies distributed in the genome of strain ACICU (16). Actually, a 101-bp fragment corresponding to the left extremity of ISAba125 was present upstream of blaNDM-1 in isolate 271 (Fig. 2). This revealed that the −35 box of the blaNDM-1 promoter was located inside the inverted repeat left (IRL) of ISAba125. Expression of the blaNDM-1 gene is therefore under the control of a hybrid promoter with the −35 sequences being provided by ISAba125 and the −10 sequences likely originating together with blaNDM-1 from its natural progenitor. Further analysis of the sequences located at the right-end extremity of ISEc33 identified the following ISAba125 sequences. The 2-bp target site duplication identified on both extremities of ISEc33 confirmed that its insertion into ISAba125 had occurred through an independent transposition event. However, the right-end extremity of ISAba125 was not identified, this IS being truncated. In addition to demonstrating that ISAba125 was involved in the blaNDM-1 expression, this finding suggests that this IS element could originally have played a role in the mobilization of the blaNDM-1 gene. Downstream of the blaNDM-1 gene, the insertion sequence ISSen4 was identified (Fig. 1).
Promoter sequences of the blaNDM-1 gene in E. coli 271. The −35 and −10 boxes are double underlined. The inverted repeat left (IRL) of insertion sequences ISEc33 and ISAba125 are highlighted in grey. The start codon of the blaNDM-1 gene is in boldface.
The entire so-called resistance module (defined by the fragment containing the blaNDM-1 resistance gene inserted onto the IncN scaffold) was then analyzed. It started on the left extremity by the inverted repeat right (IRR) sequence of Tn5403 (Fig. 1). Transposon Tn5403 is a Tn501-related transposon that has been often identified among Enterobacteriaceae, corresponding to a Tn3-like class II transposon usually carrying mercury resistance-encoding genes (2). The respective transposase and resolvase genes of Tn5403 were fully identified, as were its inverted repeats. Since no duplication of the target site of Tn5403 was identified at its extremities, its acquisition through a transposition event was not evidenced. However, a duplication of 8 bp (CAGGGGTC) was identified on both ends of the resistance module (Fig. 1). On the left extremity, that sequence was made of 2 bp (CA) belonging to the plasmid scaffold, and the following sequence (GGGGTC) actually corresponded to the first bp of the IRL of Tn5403. Since Tn501-like transposons have been previously shown to transpose via a cointegrate intermediate (2), it is possible that the blaNDM-1-bearing resistance module had been acquired by a similar cointegration process.
Other plasmid-mediated resistance determinants.In addition to plasmid p271A carrying the blaNDM-1 carbapenemase gene, other plasmids carried additional antibiotic resistance genes. As mentioned above, the blaCTX-M-15 ESBL gene (associated with insertion sequence ISEcp1) and the blaTEM-1 narrow-spectrum β-lactamase gene were identified on the IncF plasmid p271B, together with the blaOXA-9 gene (Table 1). Aminoglycoside resistance genes were identified, including in particular the two 16S RNA methylase ArmA (located on plasmid p271B)- and RmtB (located on plasmid p271C together with blaTEM-1)-encoding genes (Table 1). A series of genes encoding resistance to macrolides (ermB, mph2, mel), to trimethoprim (dfrA1, dfrA12), to rifampin (arr2), to chloramphenicol (cmlA5, catB4), to fluoroquinolones (qepA), and to sulfonamides (sul1) were also plasmid borne (Table 1). The arr2-cml5A-blaOXA-10-aadA1 gene cassette array was identified in a class 1 integron, although the armA-mel-mph array was associated with ISCR1 as previously reported (46) and the blaTEM-1-rmtB with ISCR3 as also previously reported (9).
Resistance determinants identified for E. coli 271
Chromosomally located antibiotic resistance determinants in E. coli 271.To gain further insight into the genetic basis of multidrug resistance of that isolate, the E. coli 271 chromosome was also analyzed for antibiotic resistance determinants. Resistance to fluoroquinolones (ciprofloxacin MIC, >32 μg/ml) was likely explained by the substitutions observed for the GyrA amino acid sequence—a Ser-to-Leu substitution at position 83 and an Asp-to-Asn substitution at position 87—and in the ParC sequence, which had a Ser-to-Ile substitution at position 80. The fact that E. coli 271 expressed a plasmid-encoded QepA efflux pump suggests that this mechanism acted synergistically with these chromosomal substitutions to give rise to a high level of resistance to fluoroquinolones as shown previously (5). The naturally occurring ampC gene that is intrinsic to the E. coli species was identified, being 100% identical to that of E. coli K-12. This sequence corresponded to that of a wild-type AmpC and not to an extended-spectrum AmpC that may compromise, although only slightly, the efficacy of carbapenems (26, 27). Analysis of its promoter sequences revealed that they were of wild-type form and therefore that the ampC gene was not expressed at a high level in E. coli 271.
Genetic characteristics of E. coli 271.The colIA gene encoding production of colicin was identified. No additional specific toxin was identified, whereas the previously identified toxin-antitoxin systems encoded, respectively, by the yofQ-dimJ, chapA-chapR, and mazF-E tandems were found. The gene locus encoding flagellum biosynthesis was identified, being made of previously identified genes, including the fla ATP synthase gene (29). Fimbria biosynthesis was associated with a series of genes, including the fimA gene encoding the major type 1 fimbria subunit, the fimZ gee encoding a transcription factor activating fimA transcription, the fimI gene encoding type 1 fimbria biosynthesis, the fimD gene encoding the type 1 fimbria outer membrane protein, and the sfmH, -F, and -A genes encoding a putative fimbria-like adhesin (12). Also, the fimH gene together with the hra gene encoding adhesins considered virulence factors in E. coli were identified (44).
Analysis of the porin-encoding genes, and in particular those previously associated with β-lactam resistance, showed that the OmpA was 100% identical to the wild-type OmpA identified for E. coli K-12 (42). However, the OmpC protein showed only 89% amino acid identity with that of E. coli K-12, with 32 amino acid changes, including a 3-amino-acid deletion shortening the protein size. Interestingly, this OmpC protein sequence was 100% identical to that of a carbapenem-nonsusceptible but carbapenemase-negative E. coli isolate previously described (25), which would suggest the involvement of that OmpC variant in the resistance to carbapenems in E. coli 271. Finally, a mutation that generated a stop codon was identified in the ompF gene, thus leading to a short (34-amino-acid long) protein and nonfunctional OmpF porin. Those porin modifications may contribute significantly to a higher level of resistance to several β-lactams (17, 28).
Concluding remarks.The current emergence of NDM-1 producers worldwide represents a significant threat, for several reasons. (i) NDM-1 producers are most often resistant to many if not all antibiotics. (ii) The blaNDM-1 gene is plasmid located and associated with mobile elements, and its spread appears to be efficient at least among Enterobacteriaceae. (iii) The blaNDM-1 gene has been identified not only in K. pneumoniae, a hospital-based enterobacterial species, but also in E. coli, which is the most important community-acquired Gram-negative pathogen (34). For those reasons, we have sequenced the whole genome of an NDM-1-positive E. coli isolate with the aim of providing evidence for special features associated either with the blaNDM-1-carrying plasmid or with the genetic background of the strain.
NDM-1 producers have accumulated as exemplified here an incredibly high number of unrelated resistance determinants. This observation suggests that they are the products of multiple rounds of in vivo antibiotic selection. Since the blaNDM-1 gene was located here on a specific plasmid without any other resistance markers, it can be speculated that its selection resulted from usage of β-lactams only, and in particular, carbapenems.
The plasmid carrying the blaNDM-1 gene was peculiar since it possessed the scaffold of an IncN-type plasmid, now named IncN2, but contained a different replicase gene. Further studies will evaluate to what extent the replicase properties impact the plasmid replication, possibly modifying the plasmid copy number and incompatibility properties. Interestingly, we demonstrated its broad host range with a replication ability observed at least for A. baumannii. Since a very similar plasmid scaffold has been identified once from Australia in K. pneumoniae, it could be speculated that plasmid p271A may correspond to a plasmid carrying unrelated resistance genes that are widespread in Australia, resulting from transfers of strains from Bangladesh.
Since a partial sequence of an IS element specific for Acinetobacter species was identified upstream of blaNDM-1, this may suggest that the spread of the blaNDM-1-containing genetic structures occurred initially in A. baumannii, its acquisition in Enterobacteriaceae corresponding to a further event. Of note, in China, the blaNDM-1 gene has been seen only for A. baumannii so far, whereas it has been reported for both A. baumannii and Enterobacteriaceae in India (20, 24).
Finally, our study provides the first genome sequence of an NDM-1-producing isolate, revealing interesting features related to the plasmid scaffold and accumulation of resistance mechanisms. The genetic background of the bacterial host corresponded to that of a commensal pathogen that may easily act as a reservoir of resistance genes in the human digestive tract with a subsequent uncontrolled spread into the environment.
ACKNOWLEDGMENTS
This work was funded entirely by a grant from INSERM (U914).
We are grateful to A. Carattoli for her precious advice on the plasmid annotation.
FOOTNOTES
- Received 8 February 2011.
- Returned for modification 1 April 2011.
- Accepted 25 June 2011.
- Accepted manuscript posted online 11 July 2011.
- Copyright © 2011, American Society for Microbiology. All Rights Reserved.