Increasing New Delhi metallo-β-lactamase-positive Escherichia coli among carbapenem non-susceptible Enterobacteriaceae in Taiwan during 2016 to 2018

New Delhi metallo-β-lactamase (NDM) had been reported to be the predominant carbapenemase among Escherichia coli in Taiwan. However, studies focusing on the clonal background and epidemiology of plasmids carrying NDM genes were limited. Between 2016 and 2018, all clinical E. coli and Klebsiella pneumoniae isolates that were non-susceptible to ertapenem, meropenem, and imipenem were tested for carbapenemase-encoding genes (CEGs) and antimicrobial susceptibilities. Molecular typing was performed on all carbapenemase-producing isolates. Whole genome sequencing (WGS) was performed on all NDM-positive E. coli isolates. Twenty-three (29.5%) of 78 carbapenem non-susceptible E. coli and 108 (35.3%) of 306 carbapenem non-susceptible K. pneumoniae isolates carried CEGs. The most prevalent CEGs in carbapenemase-producing E. coli (CPEc) were blaNDM (39.1%) and blaIMP-8 (30.4%), while that in carbapenemase-producing K. pneumoniae was Klebsiella pneumoniae carbapenemase (KPC) (72.2%). Fifteen sequence types were identified among 23 CPEc, and 55.6% of NDM-positive E. coli isolates belonged to ST410. WGS showed ST410 isolates were highly clonal and similar to those from other countries. All NDM-5-positive E. coli isolates carried identical IncX3 plasmid harboring blaNDM-5 but no other antimicrobial resistance (AMR) genes. In each of the four NDM-1-positive E. coli isolates, the blaNDM-1 was present in a ∼ 300 kb IncHI2/IncHI2A plasmid which carried an array of AMR genes. NDMs are the most prevalent carbapenemase among CPEc in Taiwan. Awareness should be raised as the prevalence of NDM-positive E. coli might increase rapidly with IncX3 plasmid and globally distributed strain ST410 being the potential vectors for wide dissemination.

Antimicrobial susceptibility testing of carbapenem-non-susceptible isolates. The susceptibility of antimicrobial agents and the distribution of MICs are presented in Table 2 and Supplementary Table 1. The carbapenem-non-susceptible isolates were highly resistant to extended spectrum β-lactams. For the non-βlactams, the susceptibility rate of CnSEc was the highest among tigecycline (97.4%), followed by colistin (94.9%) and amikacin (92.3%). Non-CP-CnSEc isolates were more susceptible to ciprofloxacin and levofloxacin than CPEc. CPEc isolates were 100% susceptible to tigecycline and colistin ( Table 2). NDM-positive CPEc were 100% non-susceptible to all β-lactams, and had lower susceptibility to aminoglycosides than IMP-, KPC-, or VIMpositive CPEc (Supplementary Table 2).
The susceptibility rate of CnSKP was the highest among amikacin (86.6%), followed by colistin (84.0%) and tigecycline (67.9%). The susceptibility rate between CPKP and non-CP-CnSKP isolates were also analyzed and compared. The most significant difference between these two groups were the higher rates of susceptibility to trimethoprim/sulfamethoxazole among CPKP compared to non-CP-CnSKP (34.3% and 16.2%, p < 0.01) ( Table 2). Table 1. The distribution of carbapenemases and sequence types of carbapenemase-producing E. coli and K. pneumoniae, 2016-2018. *Two carbapenemase-producing K. pneumoniae isolates were positive for dual carbapenemase: one positive for IMP-8 and VIM-1; the other positive for KPC-2 and NDM-4.    Fig. 2A,B. The sequence type of CPEc and CPKP isolates are listed in Table 1. CPEc isolates were highly diverse in terms of sequence type (ST), with a total of 15 STs identified. However, ST410 was the most predominant one among NDM-positive E. coli isolates (5/9, 55.6%) and all of them shared > 80% similarity in PFGE banding patterns to each other ( Fig. 2A). The remaining NDM-positive E. coli belonged to ST38, ST155, ST167, and ST359. The clonal relationship of the CPEc was further evaluated using the WGS data. Core genome MLST (cgMLST) analysis based on the WGS data on the chromosome also showed that our ST410 isolates are nearly identical to each other and highly similar to the ST410 isolates from China, South Africa and Ghana (Fig. 3). Among CPKP, more than 19 different STs were identified. Most of the KPC-positive K. pneumoniae were ST11 (59/78, 75.6%), followed by ST15 (17/78, 21.8%). All CPKP belonging to ST11 were positive for KPC. ST11 accounted for 75.7% (56/74) KPC-2, 100% (1/1) KPC-3 and 66.7% (2/3) KPC-17. The KPC-positive ST11 K. pneumoniae could be grouped into 4 main pulsotypes. The KPC-positive ST15 isolates belonging to another distinct pulsotype. The VIM-1 positive K. pneumoniae isolates were also distinct from the KPC-positive isolates by PFGE. (Fig. 2B).

Type of carbapenemase
Location and surrounding structure of the NDM-positive E. coli. In all five NDM-5-positive isolates (069, 278, 321, 467, 568), the NDM-5 gene was located in a ~ 46 kb IncX3 plasmid that are nearly identical to each other. These plasmids are similar to pP785-NDM5 from China (GenBank accession No. MG547511) 11 . The inverted repeat (IR) sequence of the ISAba125 was found at the immediate upstream of the NDM-5 gene (Fig. 4). Same as pP785-NDM5, these plasmids did not carry other antimicrobial resistance (AMR) genes.
Carbapenem consumption. Annual carbapenem consumption showed increasing trends during the analysis period, from 34.7 DDD per 1000 patient-day in 2010 to 73.2 DDD per 1000 patient-days in 2018 (p = 0.006, Supplementary Fig. 2). There was significant correlation between carbapenem consumption and the rate of CPKP among CnSKP (r = 1.00, p < 0.01) while the rate of CPEc among CnSEc (r = 0.60, p = 0.28) was insignificant.   www.nature.com/scientificreports/ Non-CPE isolates had accounted for the majority of carbapenem-non-susceptible Enterobacteriaceae (CnSE) in Taiwan 2 . Combination of Amp-C β-lactamse and loss of outer membrane porin used to be the main mechanism of carbapenem non-susceptibility among Enterobacteriaceae isolates in Taiwan 9,10,12 . However, an up-surging CPE trend had been observed, with the prevalence of carbapenemase production increased from 6% in 2010 to 46.5% in 2017 among CnSKP isolates in Taiwan [13][14][15][16] .
Carbapenemase-producing isolates were less common in E. coli than in K. pneumoniae. Among E. coli isolates that were non-susceptible to either ertapenem, imipenem, or meropenem, the rate of CPE was 1.4% during 2010-2012 17 , but has increased significantly to 7.6% during 2012-2015 10 . In the present study, the CPE accounted for 29.5% of E. coli isolates that were non-susceptible to all of the three carbapenem agents. It should be noted that the difference in the definition of carbapenem non-susceptibility might affect the prevalence of CPE reported in different studies. Nevertheless, increasing carbapenemase-positive E. coli isolates is still a call to intensify infection control intervention because the common commensal E. coli from the digestive tract might act as a vector to accelerate epidemic dissemination 18 .
Enterobacteriaceae carrying bla NDM was first identified in Taiwan in 2010 from K. pneumoniae in a patient returned from New Delhi where he had been hospitalized for a gunshot injury 8 . Later, the first NDM-1-positive E. coli (ST345) was discovered in 2012 9 . NDM has rapidly distributed and become the predominant carbapenemase among E. coli during 2012-2015 10 . The present study echoes that NDM are still the predominant carbapenemase in CPEc after year 2015 in Taiwan.
The increase of the NDM-positive E. coli isolates in our hospital likely resulted from horizontal gene transfer of the NDM gene carried on mobile genetic elements. Plasmids with similar backbones to the IncX3 bla NDM-5 and the IncHI2/IncHI2A bla NDM-1 -carrying plasmids found in the present study have been detected in multiple species of the Enterobacteriaceae 11,19 . These plasmids are considered important vectors involved in the dissemination   www.nature.com/scientificreports/ of NDM variants in different geographic regions and widespread in different hosts 11,19 . IncX3 plasmids are the most common type of plasmid carrying bla NDM , mainly in China and its neighboring countries 19 . E. coli with bla NDM -carrying IncX3 plasmids have been found in the livestock animals and poultry in China 7,11,20 . It is of concern that people in community may be exposed to these pathogens through the food chain. NDM-positive Enterobacteriaceae in both human and animal guts may serve as the environmental reservoir and source for bla NDM dissemination 7,21 .
The presence of mobile genetic element ISAba125 immediate upstream of the NDM-5 gene likely played a role in the mobilization of the NDM-5 gene or its expression of carbapenem resistance 19 . The presence of multiple copies of the NDM-1 gene cassettes in tandem repeat, seen in one of our bla NDM-1 -carrying plasmid, has been reported in a Klebsiella pneumoniae isolate KPX 22 . Such phenomenon indicated its active mobilization capability likely due to selective pressure brought about by antibiotic use.
The NDM-positive E. coli isolates are mostly genetically diverse with respect to MLST and PFGE in this present study. This is quite different from the spread of KPC-positive K. pneumoniae and CTX-M-15-positive E. coli, which both disseminates through the successful pandemic clones [23][24][25][26] . A total of 15 sequence types (ST) were detected among the 23 CPEc isolates. Among these, ST410 accounted for a majority of NDM-positive E. coli (5/9, 55.6%) and all of them showed > 80% similarity by PFGE. Transmission of the ST410 NDM-positive isolates among patients cannot be ruled out since isolates having indistinguishable PFGE patterns were also detected. ST410 E. coli has been considered as a reservoir of resistance genes in both human and animals, particularly of the bla CTX-M-15 gene 27,28 . In China, the carriage of CPEc is also closely associated with specific clones, with ST410 being the second most prevalent among bla NDM carrying isolates. Although less common than ST167 and ST617, ST 410 strains with bla NDM are widely distributed in several countries and are regarded as a potential emergent high-risk clone for the global dissemination of NDM in E. coli 19,29 . Our cgMLST analysis show that the bla NDMpositive ST410 isolates are highly clonal and similar to those from other countries, which attest to the global dissemination of this carbapenem-resistant E. coli lineage 30 .
Colistin remained one of the limited treatment options against CnSE. In our study, the resistant rate to colistin among CnSKP and CnSEc isolates was 16.0% and 5.1%, respectively. CnSE become resistant to colistin via different mechanism, including chromosomal gene mutations and plasmid-borne mcr-1 genes 31 . The plasmidmediated mcr-1 gene raised the concerns that colistin resistance could become more widespread. Tigecycline is another treatment option but decreasing susceptibility rate had been observed in our CPKP (78.6% in 2016 to 58.6% in 2018, data not shown). Prolonged-infusion of high-dose meropenem in combination with other active antimicrobial agent has been proposed for treating CnSE isolates with a meropenem MIC ≤ 8 mg/L 32 . However, in the present study, only 4.6% of CPKP and 8.7% of CPEc isolates display a meropenem MIC ≤ 8 mg/L, which indicates that high-dose prolonged infusion meropenem is not the drug of choice for carbapenemase-producing E. coli or K. pneumoniae. Several new antimicrobial agents have been developed for carbapenem-non-susceptible bacteria, such as ceftazidime/avibactam, meropenem/vaborbactam and imipenem/relebactam, but none of them have the ability to inhibit metallo-β-lactamases (MBLs), which accounted for 78.2% of CPEc and 28.7% of CPKP in this study. While the prevalence of NDM has increased and become the major carbapenemase in E. coli, development of new therapeutic agent is in urgent need because treatment options of MBLs are limited.
There are several limitations to our study. First, this was a single center study with only the two most common CnSE species included. Nevertheless, our study revealed the dissemination of carbapenemase-producing gene among diverse clones in a single center, which should raise the concerns about the emergent resistance from the expansion of multiple bacterial clones and further inter-clonal spread of the resistant determining genes. Second, isolates susceptible to any of the three tested carbapenem would be excluded so we might fail to identify all carbapenemase-positive isolates. Third, harboring carbapenemase-producing genes is not equal to gene expression with carbapenemase production. We regarded them the same because gene expression may be induced under antibiotic pressure even though carbapenemase was not detected.

Conclusions
In conclusion, the NDM genes appear to disseminate and become the most common carbapenemase among CPEc in Taiwan. The globally distributed E. coli strain ST410 accounted for a high proportion of NDM-positive E. coli isolates, and the majority of bla NDM was found on IncX3 plasmids. Awareness should be raised as the prevalence of NDM-positive E. coli might increase rapidly with IncX3 and ST410 being the potential vectors for wide dissemination.

Methods
Bacterial isolates and antibiotic susceptibility testing. The study was conducted at National Taiwan University Hospital (NTUH), a major teaching hospital located in northern Taiwan with a total capacity of 2600 beds. Between 2016 and 2018, all clinical K. pneumoniae and E. coli isolates with non-susceptibility to all of ertapenem, imipenem and meropenem (minimum inhibitory concentrations [MICs] of imipenem and meropenem > 1 μg/ml and MIC of ertapenem > 0.5 μg/ml) were enrolled for subsequent microbiological studies. We used the VITEK 2 Automated System (BioMe ŕieux, Marcy l'Etoile, France) for bacterial identification and for determining the susceptibility to piperacillin/tazobactam, cefmetazole, cefotaxime, ceftazidime, cefepime, ertapenem, meropenem, imipenem, ciprofloxacin, levofloxacin, gentamicin, amikacin, trimethoprime-sulfamethoxazole and tigecycline. Susceptibility to colistin was determined by broth microdilution method as the Clinical & Laboratory Standards Institute (CLSI) suggested 33 . Susceptibilities to colistin and tigecycline were interpreted based on the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines 34 and the US Food and Drug Administration breakpoint 35 for Enterobacteriaceae, respectively. Susceptibilities to other agents were based on CLSI guidelines 36  Detection of carbapenemase-encoding genes (CEGs). DNA was extracted using Viogene DNA extraction kit (Blood Genomic DNA Extraction Midiprep System; Viogene, Taipei, Taiwan) following instruction of the manufacturer. All of the isolates fulfilled the predefined definition of carbapenam nonsusceptibiltiy were subjected to polymerase chain reaction (PCR) detection of genes encoding carbapenemases, including class A (bla KPC ), class B (bla IMP , bla VIM , bla NDM ), and class D (bla OXA-48 ), using the primers described previously 37,38 . All carbapenemase gene PCR positive amplicons were sequenced to check for amplicon specificity. In this present study, isolates positive for the CEGs were considered as CPE, and those that were negative were considered as non-carbapenemase-producing carbapenem-non-susceptible Enterobacteriaceae (non-CP-CnSE).

Molecular typing.
To analyze the clonal relatedness, the CPE isolates were subject to pulsed-field gel electrophoresis (PFGE) and multi-locus sequence typing (MLST) following previously published protocols and information from the MLST website (http://bigsd b.paste ur.fr/ecoli /ecoli .html and https ://bigsd b.paste ur.fr/ klebs iella /prime rs_used.html) 39  was performed on all NDM-positive E. coli isolates (5 NDM-5 and 4 NDM-1-positive). De novo sequencing of each of the E. coli strain was conducted by a shotgun approach using Oxford Nanopore MinION rapid 1D sequencing and illumina 2 × 150-bp pair-end shotgun reads. The long reads and short reads were assembled using Unicycler following the program operating instructions 40 . Annotation of the genomes was performed using RAST 41 . Subsequent analyses of the assembled genomes were performed using CLC genomic workbench (Qiagen, Düsseldorf, Germany) and MAUVE 42 .

Consumption of carbapenems.
The data of carbapenem consumption at NTUH were obtained for the years 2010-2018 in order to observe the annual fluctuation. The consumption of each carbapenem was provided in the form of defined daily dose (DDD) per 1000 patient-days. DDD is the assumed average treatment dose per day for a drug prescribed for its main indication in adults, as defined by the World Health Organization (WHO) 43 .
Statistical analysis. Statistical analysis was performed using STATA software (version 11, StataCorp, College Station, TX). The χ2 test was used for categorical variables. Carbapenem consumption and the prevalence of CPE during every half year period were assessed for trends by utilizing the chi-square test for trend. Correlations between CPE incidence and carbapenem consumption were assessed with Spearman's correlations, using total carbapenem consumption one year prior to the predefined time period. A value of p < 0.05 was considered to be statistically significant.

Ethics approval and consent to participate. The study was approved by the NTUH Research Ethics
Committee (registration number 201908060RIN) and written or oral informed consent was waived.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.