Molecular characteristics of mcr-1-carrying plasmids and new mcr-1 variant recovered from polyclonal clinical Escherichia coli from Argentina and Canada

We have characterized nine mcr-1-harboring plasmids from clinical Escherichia coli isolates previously described in Argentina and Canada. Three of these plasmids carried a mcr-1-variant called here mcr-1.5. All these E. coli isolates were not clonally related and were recovered in different years and locations. However, their mcr-1-harboring plasmids showed high identity among them and to others characterized in other countries, which strongly suggests that this plasmid-type is playing an important role in spreading this mechanism of resistance to polymyxins.


Introduction
Since the first description of MCR-1, a plasmid-encoded phosphoethanolamine transferase, in November, 2015, in China [1], this mechanism of polymyxin resistance was detected around the world in enterobacterial isolates recovered from animals, environment, food samples and humans [2]. MCR-1-producing Escherichia coli was the most common species described in the literature, mainly of polyclonal origin. Similar plasmids harboring mcr-1 gene (pMCRs) were found elsewhere, belonging to replicon types IncI2, IncHI2 and IncX4, supporting the notion that horizontal transfer constitutes the major dissemination route of mcr genes [3,4].
The first gram-negative mcr-1-positive bacteria described in the Americas were E. coli isolates recovered in Argentina [9] and Canada [10]. Nine multidrug resistant E. coli were PLOS ONE | https://doi.org/10.1371/journal.pone.0180347 July 5, 2017 1 / 13 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 recovered from clinical human specimens in six Argentinian hospitals from three cities between July 2012 and January 2016. These isolates were part of a larger sample of 87 colistin resistant Gram-negative bacilli collected between 2008 and 2016, which were screened for the presence of mcr-1 by PCR [9]. The Canadian isolate was recovered from a gastrostomy tube site and rectum of a patient hospitalized in Ottawa, Ontario, Canada, in 2011, who previously received health care in Egypt [10,11]. This isolate was an OXA-48-and CTX-M-15-producing multidrug-resistant E. coli. As part of a database of~1600 Canadian bacterial whole-genome sequences, this was the only clinical isolate of three mcr-1-positive E. coli initially found in Canada after screening of this database [10]. Here, we describe the molecular characteristics of IncI2 pMCRs recovered from these clinical E. coli isolates from Argentina and Canada, as well as a new variant of the mcr-1 gene. We also compare them with other IncI2 mcr-harboring plasmids described elsewhere, strengthening the notion of this replicon-type as one of the main mcr-disseminator.

Materials and methods
The first ten mcr-1 isolates described in the Americas were included in this study, nine from Argentina (E. coli isolates M15049, M15224, M17056, M17059, M19241, M19242, M19441, M19736 and M19855) and one from Canada (E. coli GN775) [9][10][11]. Susceptibility profiles were obtained by Etest (bioMérieux) with the exception of colistin (broth dilution) and the results interpreted by the 2016 Clinical and Laboratory Standards Institute guidelines [12] except for colistin and tigecycline, interpreted according to the European Committee on Antimicrobial Susceptibility Testing guidelines [13]. E. coli clinical isolates were genotyped by MLST [14]. Plasmid profiles of these clinical isolates were obtained by pulsed-field gel electrophoresis analysis of S1 nuclease-digested DNA (S1-PFGE) [15]. The ones carrying the mcr-1 gene were identified and their sizes estimated by S1-PFGE followed by Southern blot analysis using a specific mcr-1 probe [15]. Plasmid content of each isolate was extracted with the QIAprep Spin miniprep kit (Qiagen) and used for transformation assays in chemically competent E. coli TOP10 (Life Technologies; colistin MIC of 0.016 μg/ml). mcr-1-transformant strains were selected using Luria-Bertani agar plates supplemented with colistin (1 μg/ml), and confirmed by PCR. pMCRs were then extracted from transformant E. coli strains using the Qiagen Large-Construct kit (Qiagen) and sequenced using Illumina's MiSeq system. The obtained contigs were assembled using CLC Genomics Workbench software (CLC bio, Qiagen). Gaps were filled by PCR amplification and Sanger sequencing. Open reading frames (ORFs) were annotated using the RAST server (rast.nmpdr.org) followed by manual comparative curation and determination of sequence similarity using the BLAST web server. Alignments with other IncI2 pMCRs were performed by using the BRIG tool [16].  Table 1 shows the susceptibility profiles of E. coli clinical isolates. Colistin MICs ranged from 4 to 16 μg/ml. All isolates were susceptible to amikacin and tigecycline. With the exception of the Canadian isolate (OXA-48 and CTX-M producer) [11] and 4 Argentinian E. coli (CTX-M producers) [9], the isolates were generally susceptible to ß-lactams. Conversely, most of the isolates were resistant to quinolones and tetracycline. By MLST all the clinical isolates belonged to different sequence types (ST) ( Table 1). Compared to the E. coli MLST database (http://mlst.ucc.ie/mlst/dbs/Ecoli; 7,412 STs; last accessed June 5, 2017), none of the STs assigned to the Argentinian isolates included in this work were found among the eight entries previously reported from Argentina, while ST624, which was assigned to the Canadian isolate GN775, was found in two entries reported from Canada. ST410, which was assigned to isolate M19441, was defined as a hyperepidemic clone and founder of the widely disseminated clonal complex 23 (CC23) [17]. This clone was also previously found to carry mcr-1 as well as bla CTX-M genes in isolates recovered from a turkey hen meat sample in Germany [18] and from a human blood culture in Brazil [19]. In these two last cases, mcr-1 gene was located on the bacterial chromosome [18] and on an IncX4 plasmid [19]. There is evidence that E. coli ST410 has been successful for interspecies transmission between food-producing animals, wildlife, humans, companion animals and the environment, increasing the risk of becoming a successful pandemic clone [20]. S1-PFGE showed a varied plasmid content in all the clinical isolates but Southern blot with mcr-1 probe showed that the pMCRs were very similar in size (~60 kb) (Fig 1).

Results and discussion
Nine mcr-1-transformant strains were obtained (8 from Argentina and 1 from Canada; no mcr-1 transformant strain was achieved for E. coli M17056). Colistin was the only antimicrobial drug for which all transformant isolates showed reduced susceptibility or resistance (MICs of 2 to 4 μg/ml). Comparison of the transferred mcr genes with the original mcr-1 [1] showed that 3 of the Argentinian isolates (M15049, M17059 and M19241, recovered from two hospitals) had the same missense point mutation in position 1,354 resulting in a H452Y change of the MCR protein. Other three new MCR-1 variants were recently published, MCR-1.2, MCR-1.3, and MCR-1.6 [6][7][8]. Other unpublished MCR-1 variants were also found in the Genbank database (Table 2). We noticed that one of them, called MCR-1.5 found in an E. coli isolated from human urinary tract sample in Argentina (GenBank accession number KY283125), had the same H452Y amino acid change described in our study. To avoid future confusions with the nomenclature, we called the new variant described here as MCR-1.5.
All plasmids included the conserved 2,607 bp DNA segment containing mcr (mcr-1 or mcr-1.5) and pap2 genes [21] with slightly different genetic environments (Fig 2). Two isolates presented the change TAAAAT instead of TATAAT at -10 box promoter region described previously [22]. The three isolates carrying the new variant mcr-1.5/pap2 (M15049, M17059 and M19241) had this DNA segment flanked by 2 copies of ISApI1 in the same orientation (Fig 2) but the characteristic 2 bp target site duplications (TSD) were not found [21].
This composite transposon was not described very often (usually only one copy, if any, of ISApI1 is present beside mcr-1) [4,20,23] and would be responsible for the chromosomal integration of mcr-1 in E. coli [18,[23][24][25]. ISApI1 was not present in pMCRs from the other 5 Argentinian and the Canadian isolates (Fig 2). The corresponding inverted repeat right and left (IRR and IRL) were also absent in those plasmids. However, the mcr-pap2 element was found in a similar location, between top (encoding a DNA topoisomerase III) and nikB (relaxase) genes (Fig 2), like in pHNSHP45 and other IncI2 plasmids, suggesting a hot spot for the insertion of this transposon in this plasmid-type. In these 6 isolates where ISApI1 was absent, the mcr-pap2 element was flanked by a conserved trinucleotide (5'-ATA-3'), a sequence found immediately downstream of the IRR of ISApl1 in both composite transposons and single-copy ISApl1 [21]. The absence of ISApI1 in some pMCRs described here and elsewhere could be explained by its mobilization, transposing the mcr-pap2 structure to conjugative plasmids with a subsequent loss of ISApI1 copies following integration, as postulated before [21,26,27].
Based on the predicted structure of MCR-1 [1,28] and the crystal structure of the catalytic domain [29], the amino acid change H452Y found in MCR-1.5 was located between helix α6 and sheet β10 (Fig 3).  Like MCR-1, a H452 was found in MCR-2, which showed some amino acid changes surrounding this position. Homology comparison with the phosphoethanolamine transferases (PAE) LptA from Neisseria meningitidis and EptC from Campylobacter jejuni showed an arginine (position 440) and glutamine (position 415), respectively, corresponding to the Y452 in MCR-1.5. A tyrosine residue was also observed at this position in the PAE of Enhydrobacter aerosaccus (WP_007116571.1), Paenibacillus sophorae (WP_036596266.1) and Dichelobacter nodosus (WP_041729850.1) [1]. A recent evolutionary analysis grouped MCR-1 with those of E. aerosaccus, P. sophorae and D. nodosus, and very close to the PAE of Moraxella catarrhalis (subclade I), while LptA was grouped in a second subclade of PAEs [30]. The lack of differences in colistin resistance levels observed between MCR-1-and MCR-1.5-transformant strains would suggest that the amino acid change found in MCR-1.5 is not affecting its PAE activity.
Full sequences of these nine pMCR showed a GC content ranging between 42.3% and 43%. As showed by S1-PFGE and Southern blot analysis, their sizes were similar and all belonged to the IncI2 incompatibility group (Fig 4).
pMCRs of Argentinian origin were highly similar, with a genetic identity ranging between 87.2% and 99.8%. Six of them (from clinical isolates M15049, M15224, M17059, M19241,  M19242, and M19441) showed maximum genetic similarities to plasmids pABC149-mcr-1 (GenBank accession no. KX013538), isolated from a pathogenic human E. coli in the United Arab Emirates [31], and pHNSHP45 (KP347127) the first reported pMCR [1]. The other 2 pMCRs (from M19736 and M19855) were more similar to plasmid pECJS-61-63 (KX084393) from an E. coli recovered in China (95.1% identity; 84.6% identity with pHNSHP45). The Canadian pMCR (from GN775) was the most divergent, with an overall 83% identity with the Argentinian plasmids, 80.3% identity with pHNSHP45, and a maximum genetic identity with plasmid pECJS-61-63 (91%). All backbone and transfer genes in the pMCRs characterized in this study were highly conserved. Differences were mainly due to the presence of insertion sequences (e.g. ISApI1 in three Argentinian plasmids flanking the mcr-pap2 structure, or some transposases downstream of repA in the Canadian pMCR), absence of open reading frames encoding for hypothetical proteins and reorganization of the pilV shufflon (Fig 4). Same kind of similarities and variations were observed in the alignment of different IncI2 plasmids, harboring or not mcr-1 gene, with the ones described in our study (Fig 5).
In Argentina, mcr spread would be due to the increased use of polymyxins against emergent, highly resistant clinical pathogens such as KPC-producing Enterobacteriaceae or extremely drug-resistant Acinetobacter baumannii [9], which could create selective pressure in  [32]. However, a recent report describes a broad dissemination of mcr-1 gene in E. coli recovered from healthy poultry in Argentina, suggesting an  Table 3.
https://doi.org/10.1371/journal.pone.0180347.g005 mcr-1-carrying plasmids from polyclonal clinical Escherichia coli from Argentina and Canada important source of this colistin resistant microorganism since at least 2013 [33]. In contrast, in countries like Canada where colistin is not commonly used in clinical therapy schemes, low detection of this mechanism in clinical isolates [34] was the result of 'imported' pathogens by patients hospitalized in countries where the mcr gene is more common to be found (e.g. China) [35], or where it was detected but its prevalence is unknown (e.g. Egypt) [11,36].
The worldwide emergence of colistin resistant bacteria without any prior clinical colistin exposure could be considered a consequence of polymyxins use in veterinary medicine [37][38][39] and as promoter factors in the agriculture and food production sectors [1,33,[37][38][39][40]. Abundant bibliography supports this presumption: mcr-carrying Enterobacteriaceae are frequently isolated from different animal species such as cattle, chicken, and pigs [4,41]. Moraxella spp., which includes different species that are mainly animal pathogens, were identified very recently as potential natural reservoir of chromosomal mcr-like genes, mobilizing them to Enterobacteriaceae [42]. Detection of mcr-1 also in river water, vegetables and wild birds [43][44][45] should impulse more collaborative studies with a One Health perspective to bring light to this complex matter. In addition, strong restrictions to the use of antimicrobials in the agriculture sector have to be implemented, particularly for the ones that are still useful drugs for human disease treatments (such as polymyxins).
In summary, we have characterized pMCRs from the first mcr-1 positive Enterobacteriaceae described in the Americas, detected in Argentina and Canada. All these clinical E. coli isolates were not clonally related and were recovered in different years and locations. However, their pMCRs had high identity among them and other IncI2 pMCRs characterized in other countries, which strongly suggest that this plasmid-type is playing an important role in spreading this mechanism of resistance to polymyxins.