Genetic characterization of blaNDM-harboring plasmids in carbapenem-resistant Escherichia coli from Myanmar

The bacterial enzyme New Delhi metallo-β-lactamase hydrolyzes almost all β-lactam antibiotics, including carbapenems, which are drugs of last resort for severe bacterial infections. The spread of carbapenem-resistant Enterobacteriaceae that carry the New Delhi metallo-β-lactamase gene, blaNDM, poses a serious threat to public health. In this study, we genetically characterized eight carbapenem-resistant Escherichia coli isolates from a tertiary care hospital in Yangon, Myanmar. The eight isolates belonged to five multilocus-sequence types and harbored multiple antimicrobial-resistance genes, resulting in resistance against nearly all of the antimicrobial agents tested, except colistin and fosfomycin. Nine plasmids harboring blaNDM genes were identified from these isolates. Multiple blaNDM genes were found in the distinct Inc-replicon types of the following plasmids: an IncA/C2 plasmid harboring blaNDM-1 (n = 1), IncX3 plasmids harboring blaNDM-4 (n = 2) or blaNDM-7 (n = 1), IncFII plasmids harboring blaNDM-4 (n = 1) or blaNDM-5 (n = 3), and a multireplicon F plasmid harboring blaNDM-5 (n = 1). Comparative analysis highlighted the diversity of the blaNDM-harboring plasmids and their distinct characteristics, which depended on plasmid replicon types. The results indicate circulation of phylogenetically distinct strains of carbapenem-resistant E. coli with various plasmids harboring blaNDM genes in the hospital.

Routine worldwide surveillance is essential to understand and prevent CRE transmission; however, in many countries, including Myanmar, surveillance is limited or non-existent. Recently, Myat et al. reported the isolation of three Escherichia coli and three Klebsiella pneumoniae strains harboring bla NDM genes in a screening of 592 blood cultures in three hospitals in Yangon, Myanmar [6]. No other carbapenemase-encoding genes were identified in their study, suggesting that NDM rather than other carbapenemases is prevalent in Myanmar. The bla NDM genes are usually located adjacent to or in between mobile genetic elements, including transposons and insertion sequences, which facilitate transposition between replicons [7], [8]. Accordingly, bla NDM genes are currently found in an array of plasmid replicon types, such as IncF, IncX3, IncL/M, and IncH, as well as in plasmids with a broad-host-range, including IncA/C 2 . Consequently, the bla NDM gene has spread from the putative original reservoir, Acinetobacter, to enteric bacteria [9]. Additionally, accumulation of nucleotide substitutions in bla NDM has produced several variants from the originally identified enzyme NDM-1 [10]. Organisms harboring bla NDM genes have been detected not only in hospitals, but also in the environment in some Asian countries [11], [12]. Therefore, it is necessary to characterize the bacterial host, the bla NDM -harboring plasmid, and the genetic environment associated with bla NDM in order to understand the gene acquisition mechanism, track its spread, and investigate possible preventive measures [9].
In this study, we genetically characterized eight carbapenem-resistant E. coli isolates from a tertiary care hospital in Yangon, Myanmar, using whole-genome sequencing (WGS). Comparative analysis highlighted the diversity of nine plasmids carrying bla NDM , providing insight into an evolutionary relationship with the already identified bla NDM plasmids.

Materials and methods
Bacterial isolates E. coli isolates were obtained from clinical blood or urine specimens collected from patients at Yangon General Hospital, Yangon, Myanmar, from April to August 2015. Ethical approval for the collection of patient specimens was obtained from the Ethics Committee of Osaka University Graduate School of Medicine and the Department of Medical Research, Myanmar, with a waiver of informed consent. All samples were anonymized before analysis. Specimens were cultured on blood agar plates at 37˚C overnight, and each single colony formed was subjected to species identification and antimicrobial-susceptibility testing using a VITEK2 automated system (Sysmex bioMérieux, Marcy l'Etoile, France). Enterobacteriaceae strains with minimum inhibitory concentration (MIC) > 2 μg/mL for meropenem were collected and stored at −80˚C in phosphate-buffered saline supplemented with 25% glycerol. Isolates were also grown on M-ECC [13], CHROMagar ECC (CHROMagar Microbiology, Paris, France) supplemented with 0.25 μg/mL meropenem and 70 μg/mL ZnSO 4 , and analyzed using an API 20E system (Sysmex bioMérieux) and an EIKEN dry plate (Eiken Chemical, Tokyo, Japan) to reconfirm species identity and antimicrobial susceptibility.

Bacterial conjugation and transformation
All bla NDM -harboring plasmids identified in this study were transferred to laboratory strains by conjugation or transformation and the presence of bla NDM and the plasmid-replicon type was confirmed by PCR [23][24][25].
Bacterial conjugation was carried out as described previously [26], with some modifications. Luria-Bertani (LB) broth cultures of E. coli isolates in the early exponential phase were mixed with the recipient strain, E. coli ML4909 [27], at a 1:10 ratio. The bacterial mixture was pelleted by centrifugation, transferred onto nitrocellulose membranes on an LB agar plate, and incubated at 37˚C for 6 h. Transconjugants were selected on the LB plate supplemented with 2 μg/mL meropenem and 100 μg/mL rifampicin.
For transformation, plasmids were extracted from overnight cultures of the isolates using the Plasmid Midi kit (Qiagen, Hilden, Germany). HST08 (Takara Bio, Shiga, Japan), a derivative of E. coli K12, was electroporated with the extracted plasmids using a Gene Pulser Xcell System (Bio-Rad, Hercules, CA, USA), and transformants were selected on brain-heart-infusion agar (BD Bacto) supplemented with 0.25 μg/mL meropenem.

Characterization of carbapenem-resistant E. coli
Carbapenem-resistant E. coli strains (n = 8) were isolated at Yangon General Hospital, Yangon, Myanmar, from April to August 2015. Patients that tested positive for carbapenem-resistant E. coli were admitted to the following wards: hematology (n = 5), surgery (n = 2), and physical medicine (n = 1). Six isolates were obtained from blood specimens derived from patients in the hematology (M105, M109, M110, M214, and M217) and physical medicine (M107) wards, and two were isolated from urine derived from patients in the surgery ward (M213 and M216). All eight isolates were non-susceptible to not only β-lactams, including carbapenems, but also most other antibiotics tested, including aminoglycoside, quinolone, and chloramphenicol (Table 1). Interestingly, isolates from urine were highly resistant to chloramphenicol (MIC > 128), unlike isolates from blood (MIC: . Colistin appeared to be the only drug effective against all of the isolates examined, with fosfomycin representing a viable alternative for all strains, except for M110 (MIC: 16). By database searches using ResFinder and Plasmid-Finder, all isolates were found to carry different types of plasmids encoding several antimicrobial-resistance determinants, including β-lactamases, supporting their multidrug-resistance phenotype ( Fig 1B and S1 Table). Phylogenetic analysis based on genome-wide SNPs showed that the isolates belong to different previously defined phylogroups [28], with some belonging to pathogenic phylogroups D and F and others belonging to the commensal and less virulent phylogroups A and B1 (Fig 1A). The isolates were further classified into five multilocussequence types; no predominant sequence types were found ( Fig 1B). Notably, the five isolates from the hematology ward are not closely related genetically, except for M105 and M109, indicating that different resistant strains disseminated in the ward. M105 and M109 differ in the types of NDM and bla NDM -harboring plasmid (Fig 1B), suggesting that they had acquired the plasmids independently. This is in contrast to clonal expansion of other drug-resistant

IncA/C 2 plasmids
IncA/C 2 plasmids are broad-host-range plasmids that are predominantly found among NDM-1 producers [29,30]. Indeed, a plasmid we designated as pM214_A/C 2 represents an IncA/C 2 plasmid harboring bla NDM-1 (Fig 2A). A BLAST search indicated that pM214_A/C 2 shares the highest degree of similarity with pNDM-1_Dok01 (GenBank accession no. AP012208.1), a bla NDM-1 plasmid isolated from a clinical E. coli isolate in Japan, but of Indian origin, with 99% identity and 93% query coverage [31]. Both plasmids encode bla NDM-1 in the ARI-A region, a resistance island designated for IncA/C 2 [32]; however, the gene clusters containing bla NDM-1 are inverted between the two plasmids. In pM214_A/C 2 , the resistance island consists of two parts: an ISAba125-mediated composite transposon (Tn125) and a Tn1548-like transposon. The former region, which harbors bla NDM-1 , completely matches that in pNDM-BJ01 (Gen-Bank accession no. JQ001791.1) (Fig 2B), which was isolated from a clinical isolate of  Acinetobacter lwofii [33]. Notably, Tn125 is often detected in Acinetobacter; however, the intact form of this transposon has never been found in Enterobacteriaceae [34]. This gene cluster is bracketed by IS26-insertion sequences in pM214_ A/C 2 , suggesting that the plasmid acquired bla NDM-1 from pNDM-BJ01 or a closely related plasmid via a single IS26-mediated gene transfer. Additionally, the Tn1548-like transposon is located adjacent to Tn125 (Fig 2B). This 15,525-bp fragment contains a class 1 integron comprising several antimicrobial-resistance genes and is identical to an unrelated plasmid (pCTX-M-3, GenBank accession no. AF550415.2) [35], except for one nucleotide. This cluster is partially conserved in other IncA/C 2 plasmids, such as pNDM-1_Dok01 and pM216_A/C 2 , isolated from strain M216. Therefore, a Tn1548-like transposon might have contributed to the ARI-A island in the putative ancestor of these plasmids, with prototypic features being well conserved in pM214_A/C 2 . However, pM216_A/C 2 encodes neither bla NDM-1 nor bla CMY-4 , both of which are present in pM214_A/C 2 , but not in the putative precursor plasmid pR148 (GenBank accession no. JX141473.1) [36]. Moreover, the gene array in the ARI-A resistance island significantly differs from that in pM214_A/C 2 (Fig 2A), suggesting that these two plasmids were derived from different precursors.
Concerning the genetic context of bla NDM , pM109_FII is distinct from other IncFII Myanmar plasmids, but similar to the bla NDM-1 plasmid pGUE-NDM (Fig 4B). Both plasmids share a 12-kbp resistance-gene region surrounding bla NDM . In pM109_FII, an additional gene cassette bracketed by two IS26 sequences containing the class A β-lactamase gene bla-TEM-1 and the aminoglycoside resistance-coding gene rmtB is located downstream of this region. Interestingly, this gene cassette was also found in the bla NDM region of other plasmids, including pMC-NDM (Fig 4B). pM109_FII differs from other IncFII Myanmar plasmids in that it lacks the macrolide-resistance gene ermB and the toxin-antitoxin module pemI-pemK (Fig 4A and 4B).  Fig 2, except that genes involved in conjugal transfer are depicted in green. A 92-bp deletion that is found in pM110_X3, pM213_X3, pM216_X3, pKpN01-NDM7, and pOM26-1 is indicated by the arrow. (B) Evolutionary relationship of fully sequenced IncX3 plasmids harboring bla NDM as inferred from genetic variances. Encoded NDM variants are shown in parentheses along with geographical origin. The plasmids identified in this study are boxed. pJEG027, pNDM-MGR194, and pKpN01-NDM7 were isolated from K. pneumoniae isolates and are shaded gray. pOM26-1 was isolated from an E. coli isolate. SNS, single nucleotide substitution.
https://doi.org/10.1371/journal.pone.0184720.g003 bla NDM -harboring plasmids from Myanmar pM107_FII and pM214_FII clustered with pCC1409-1 and pCC1410-1 in the phylogenetic tree, whereas pM217_FII was assigned to a different branch (Fig 4C). The transfer operon in the former four plasmids differs slightly from that in other IncFII plasmids, with pM214_FII and pM217_FII sharing 81% nucleotide identity in this region. Interestingly, the transfer operon in pM214_FII is completely identical to that in known plasmids, such as pC15-1a [46]. Therefore, the transfer region in pM214_FII and related plasmids might be unrelated to other plasmids, possibly as a result of recombination.
Additionally, we found two IncFII plasmids lacking bla NDM (pM110_FII and pM105_FII). pM110_FII encodes resistance genes against quinolone and tetracycline and is phylogenetically distinct from other IncFII plasmids harboring β-lactamases and macrolide-resistance genes (Figs 1B and 4C). pM105_FII, a plasmid identified in strain M105, clustered with the bla NDM-5 -harboring plasmids pM107_FII and pM214_FII. Indeed, pM105_FII is very similar to pM214_FII, except for the region surrounding bla NDM-5 (Fig 4A, arrow), with 99.3% nucleotide identity outside of this region. Therefore, it is likely that pM105_FII had been encoded previously, but recently lost bla NDM-5 along with other neighboring genes. This region is bracketed by two IS26 sequences, implying that IS26 was involved in the loss of this fragment [47]. Additionally, the M105 isolate encodes bla NDM-5 on a multireplicon IncF plasmid designated as pM105_mF, and, notably, the genetic context of bla NDM-5 in this plasmid is identical to that in pM214_FII and related plasmids (Fig 5). Furthermore, the resistance island in pM107_FII and pM214_FII bracketed by two IS26 sequences is conserved in pM105_mF ( Fig  5, yellow). This implies that pM105_mF likely acquired bla NDM-5 from the pM105_FII precursor or related IncFII plasmids via IS26.

Other plasmids
Multiple replicons might broaden the plasmid host range and enable the maintenance of incompatible replicons [24]. To the best of our knowledge, the pM105_mF plasmid backbone comprising IncFIA-, FIB-, and Q1-replication genes (Fig 1B, designated FIA-FIB-Q1) has not been previously reported. A BLAST search revealed a partial match to pV228-a (Fig 5, dark  green), with 99% identity, but query coverage of only 54%. pV228-a, containing IncFIA-, FIB-, and FII-replication genes, has been identified in an E. coli strain isolated from a sewage-treatment plant in India [48], and similar multireplicon F plasmids have been found in three other Myanmar isolates (Fig 1B), with one containing an IncFII-replication gene (FIA-FIB-FII-Q1). These plasmids commonly encode several antimicrobial-resistance determinants, such as bla-TEM-1 , and resistance genes against sulphonamide, trimethoprim, and tetracycline. Collectively, these data indicate that multireplicon plasmids with several antimicrobial-resistance genes might be widespread in Myanmar.
Although it is unclear whether the bla NDM-5 gene cluster was transferred directly from a precursor of pM105_FII onto pM105_mF, our findings demonstrated that mobilization of bla NDM between plasmids resulted in a novel resistance plasmid. Additionally, the number of antimicrobial-resistance genes in pM105_mF was higher than that in IncFII plasmids modules are illustrated in green and black, respectively. (B) Schematic representation of the genetic context of bla NDM . Identical regions are shaded gray, whereas arrows denote gene-cluster insertions. (C) Maximum-likelihood tree of IncFII plasmids harboring bla NDM . Encoded NDM variants are shown in parentheses along with geographical origin. IncFII plasmids lacking bla NDM and found in Myanmar isolates were also included and are indicated in italics. Bootstrap support values based on 1000 replications are indicated at the branching points. All the plasmids, except for pCC1409-1 and pCC1410-1, which were identified in K. pneumoniae isolates, were isolated from E. coli.
https://doi.org/10.1371/journal.pone.0184720.g004 bla NDM -harboring plasmids from Myanmar harboring bla NDM-5 (Fig 1B and S1 Table), suggesting that recombination increased the number of resistance genes encoded on the same plasmid as bla NDM . As shown in Fig 1B, several multidrug-resistant plasmids, such as multireplicon F plasmids, co-exist with bla NDM plasmids in all isolates. Therefore, clinical settings in Myanmar appear to provide a breeding ground for novel resistance plasmids, and thus require more vigilance from national and local public health authorities.

Conclusions
We genetically characterized clinical carbapenem-resistant E. coli isolates from Myanmar and discovered multiple-resistance plasmids and several NDM variants in the diverse genetic backgrounds of the bacteria, even within the eight isolates examined. These findings suggest dissemination of NDM via multiple introductions and/or prolonged presence in the hospital allowing for recombination and dissemination via horizontal gene transfer into various genetic backgrounds. The situation seems to be similar to that reported in India, where NDM is already endemic and is found not only in healthcare settings, but also in the community [11], [49]. Therefore, further surveillance, including both the hospital and the community, is warranted to understand the dissemination of CRE in Myanmar.