Medical records and the patient information were retrospectively reviewed and collected. A 19-year-old female was admitted into a hospital in Guangzhou in April 2015. The patient had a history of pelvic fracture, urethral disruption and pelvic organ injury treated with colostomy and cystostomy. Surgery was performed during her hospitalization. She subsequently presented with fever, fistula of colon and ureter, complex intra-abdominal infections, systemic inflammatory response syndrome and urinary tract infections. From May to July, three carbapenem-resistant E. coli (GZ1, GZ2, and GZ3) were successively isolated from abdominal drainage and urine obtained from cystostomy. While she received various antimicrobials, including cefuroxime, gentamycin, metronidazole, ceftazidime, amikacin, meropenem, imipenem, levofloxacin, cefixime, vancomycin and piperacillin–tazobactam before GZ2 was isolated, due to clinical improvement, the treatment was changed to levofloxacin, cefixime and metronidazole after the isolation. The study was approved by the research ethics board at Peking University People's Hospital. Informed consent was not needed as this study was retrospective and participants were anonymized.

The minimum inhibitory concentrations (MICs) of the following antimicrobials, including ceftriaxone (Roche, Shanghai, China), ceftazidime, cefoxitin, cefepime, amikacin (National Institutes for Food and Drug Control, Beijing, China), piperacillin–tazobactam, tigecycline (Pfizer, NY, USA), imipenem (MSD, Hangzhou, China), meropenem (Sumitomo Pharmaceuticals, Suzhou, China), ciprofloxacin (Bayer, Leverkusen, Germany) and colistin (Amresco, Solon, USA) were determined by agar dilution method, according to Clinical and Laboratory Standards Institute (CLSI) guidelines (Clinical and Laboratory Standards Institute, 2015). The results were interpreted according to CLSI breakpoints (Clinical and Laboratory Standards Institute, 2016). Polymerase chain reaction (PCR) was used to detect the presence of mcr-1, carbapenemase genes and other resistance genes as previously described (Wang et al., 2014; Liu, Y. Y. et al., 2016). Multilocus sequence typing (MLST) was performed as described on the E. coli MLST website (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli).

Fitness was investigated by using growth curve assay and in vitro competition experiments (Liu, D. et al., 2016). Three isolates (GZ1, GZ2, and GZ3) cultured overnight in LB broth were diluted to an OD₆₀₀ of 0.01 and grew at 37°C with vigorous aeration (200 rpm). The culture cell density was determined every 0.5 h by measuring the OD₆₀₀.

In the in vitro competition assays, GZ2 and GZ3 were separately cultured overnight in LB broth at 37°C. The bacteria were diluted and equivalent numbers of GZ2 and GZ3 were pooled and cultured together at 37°C. At 0, 4, and 24 h, aliquots of the mixed bacteria were diluted with 0.9% saline solution and plated on Mueller–Hinton agar (BD, Sparks MD, USA) plates with or without colistin (4 mg/L). Colony-forming unit (CFU) were counted after 0, 4, and 24 h of incubation at 37°C. Each isolate was tested three times. The competitive index (CI) was determined as follows: CI = (GZ3/GZ2)/(Inoculated GZ3/Inoculated GZ2) as previously described (Liu, D. et al., 2016).

Serum killing assay was conducted to determine the virulence in vitro as previously described (Abate et al., 2012). An inoculum of 25 μL prepared from the mid-log phase was diluted by 0.9% saline solution and was added to 75 μL of pooled human sera contained in a 10 × 75 mm Falcon polypropylene tube (BD Biosciences, Franklin Lakes, NJ, USA). Viable counts were checked at 0, 1, 2, and 3 h of incubation at 37°C. The mean results were expressed as percentage of inoculation and a strain was classified as serum sensitive, intermediately sensitive and resistant.

To determine in vivo virulence, six pathogen-free, 6–8-week-old, male BALB/c mice were used as a sample population for each bacterial concentration. Ten-fold serial dilution of CFU of E. coli was made from a starting concentration of 10⁹ CFU/mL to 10⁵ CFU/mL, and BALB/c mice were infected intraperitoneally with 0.1 mL of each concentration. Symptoms and mortality rates were observed for 14 days. Inoculation dose was confirmed on LB agar and survival curves were assessed by Kaplan–Meier analysis. The study was approved by the research ethics board at Peking University People's Hospital.

To exclude the impact on fitness caused by different sequence type, we constructed the same strains carrying different plasmids. Conjugation experiment was performed using GZ2 and GZ3 as the donors and E. coli J53 as the recipient according to previous studies (Wang et al., 2014). The donor and the recipient were mixed at a ratio of 1:1 for 24 h. Transconjugants were selected on Chinese Blue lactose agars (OXOID, Basingstoke Hampshire, UK), supplemented with sodium azide (100 mg/L) and imipenem (1 mg/L), sodium azide (100 mg/L) and colistin (4 mg/L) or sodium azide (100 mg/L), colistin (4 mg/L), and imipenem (1 mg/L), respectively. PCR was used to screen for mcr-1 and bla_NDM-5 as previously described (Wang et al., 2014; Liu, Y. Y. et al., 2016).

The plasmid profiles of the transconjugants were identified by PCR based on replicons of the major plasmid incompatibility groups as previously described (Carattoli et al., 2005). To verify the association between the resistance plasmids and the fitness, growth kinetics of the transconjugants and the recipient E. coli J53 were investigated (Liu, D. et al., 2016). The strains cultured overnight in LB broth were diluted to an OD₆₀₀ of 0.01 and grew at 37°C with vigorous aeration (200 rpm). The culture cell density was determined every 0.5 h by measuring the OD₆₀₀ for 24 h.

Whole genome sequencing (WGS) of GZ2 and GZ3 was conducted using a PacBio RS II system (Pacific Biosciences, Menlo Park, USA) with a 10 kb size library and P6/C4 chemistry. De novo assembly was performed with SMRTanalysis version 2.3 (Pacific Biosciences). The plasmid sequence was annotated using the online Rapid Annotation Subsequencing Technology (RAST) (http://rast.nmpdr.org/) automatically and BLASTn analysis. Antibiotic resistance genes and plasmid incompatibility groups were analyzed through the website of the Center for Genomic Epidemiology (http://www.genomicepidemiology.org/). Virulence genes were identified using the virulence factor database (http://www.mgc.ac.cn/VFs/) and PathogenFinder (https://cge.cbs.dtu.dk/services/PathogenFinder/). The nucleotide sequences of the genome and plasmids of GZ2 and GZ3 have been submitted to GenBank with accession no. MCRE00000000, CP017980, and CP017981.

Statistical analysis was performed with the software GraphPad Prism version 5 using one-way analysis of variance (ANOVA) followed by Tukey–Kramer tests. P < 0.05 was considered to be statistically significant.

Results pertaining to susceptibility testing showed that all isolates were resistant to most of the tested antimicrobials including carbapenems. While GZ1 and GZ2 exhibited resistance to colistin (Table 1) and co-harbored bla_NDM-5 and mcr-1 belonging to ST10, GZ3 (ST4981) carried only bla_NDM-5.

The growth of GZ3 was significantly higher than the other two isolates (P < 0.001) (Figure 1) and in vitro competition experiment revealed that GZ2 was overgrown by GZ3 with a competitive index (CI) of 1.0157 (4 h) and 2.5207 (24 h). GZ3 demonstrated intermediate to serum killing (grade 3), whereas GZ1 and GZ2 displayed sensitive (grade 1 or 2), thus suggesting that in human sera, GZ3 exhibited higher defense against bactericidal activity than GZ1 and GZ2. A significant difference in mice lethality was identified between GZ3 and the other isolates. All mice injected with GZ3 died within the first day at a concentration of 10⁹ CFU/mL, while mice infected with GZ1, GZ2 and the control survived over 14 days.

The characteristics of the transconjugants (designated as Tc_pmcr-1GZ2, Tc_pNDM-5GZ2, Tc_pNDM-5GZ3) were summarized in Table 1. In the conjugation experiments, bla_NDM-5 and mcr-1 could be transferred separately from GZ2 and GZ3 to the recipient E. coli J53. No transconjugants co-harboring mcr-1 and bla_NDM-5 were detected. In addition to plasmids containing mcr-1 or bla_NDM-5, the plasmid profile of the transconjugants was the same as their corresponding parent strains. According to the plasmid replicon typing results, there were three plasmids in Tc_pmcr-1GZ2 (IncI2, IncHI2, IncFII), three plasmids in Tc_pNDM-5GZ2 (IncX3, IncHI2, IncFII) and one plasmid in Tc_pNDM-5GZ3 (IncFII).

No significant difference in growth was observed in Tc_pmcr-1GZ2, Tc_pNDM-5GZ2, Tc_pNDM-5GZ3, and E. coli J53 (P = 0.6238) (Figure 2). The results indicated that no considerable fitness costs were identified in carriage of plasmid containing mcr-1 or bla_NDM-5 compared to the recipient strain, suggesting that resistance was not associated with the decreased fitness of GZ2. Chromosomal characteristics, such as different sequence type (ST10 and ST4981) were considered to be the potential explanations.

WGS and virulence factor database analysis revealed that GZ2 and GZ3 had 120 and 124 virulence-associated genes, respectively. Compared with GZ2, GZ3 had a higher number of hemolysin genes and 28 specific virulence genes in the genome. These genes were associated with chorismate binding-like protein, iron acquisition, CFA/I fimbria and type III secretion protein (such as eiv, epr and epa), potentially contributing to the increased virulence in GZ3.

In this study, the plasmids harboring mcr-1 and bla_NDM-5 were designated as pGZ2-mcr, pGZ2-NDM, and pGZ3-NDM. The plasmid carrying mcr-1 from GZ2 (named pGZ2-mcr) was 59,476 bp in length, belonging to IncI2 incompatibility group. The results of BLASTn analysis revealed that plasmid pGZ2-mcr with 100% query coverage displayed 99% identity to plasmid pEC13-1 (GenBank accession CP016186.1) and plasmid pBA77-MCR-1 (GenBank accession KX013539.1) which have been isolated from E. coli in Malaysia and United Arab Emirates. No other resistance genes were found in pGZ2-mcr.

Sequence analysis showed that the two plasmids carrying bla_NDM-5 were not identical and belonged to different incompatibility group (Figure 3A). While plasmid pGZ2-NDM was 102,512 bp in length and possessed an IncX3-type backbone, pGZ3-NDM (91,451 bp) belonged to IncFII group. Both the plasmids encoded replication, stability, horizontal transfer, antimicrobial resistance, and maintenance functions (Figure 3A). pGZ2-NDM and pGZ3-NDM were found to harbor additional resistance genes, including aadA2, mph (A), sul1, aac (3)-IId, dfrA12, bla_TEM−1B, and erm (B) (Table 1). Sequence alignments with BLAST revealed that pGZ2-NDM shared 53% query coverage and 99% identity with plasmid pNDM-ECN49 (GenBank accession KP765744.1), isolated from an Enterobacter cloacae strain. The genetic environment of bla_NDM-5 differed between these two plasmids. pGZ3-NDM exhibited more than 70% query coverage and 99% identity with plasmid pHNFP460-1 (GenBank accession KJ020575.1), whereas no bla_NDM-5 gene was found in pHNFP460-1.

Comparative analysis of the genetic contexts of bla_NDM-5 (~4,900 bp) in pGZ2-NDM and pGZ3-NDM were almost identical and were also similar as that previously reported in pNDM-QD28 (GenBank accession KU167608.1) and pCQ02-121 (GenBank accession KU647721.2) with 100% query coverage and 99% identity (Figure 3B). The regions containing bla_NDM-5 were inserted into different genes. Furthermore, these plasmids belonged to different incompatibility groups. Therefore, the results indicated that similar genetic context containing bla_NDM-5 have been inserted into different plasmid backbones, thus forming a “new” plasmid.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.