Whole-genome-sequence-based characterization of an NDM-5-producing uropathogenic Escherichia coli EC1390

Background Urinary tract infection (UTI) is one of the most common outpatient bacterial infections. In this study, we isolated and characterized an extensively-drug resistant (XDR) NDM-5-producing Escherichia coli EC1390 from a UTI patient by using whole-genome sequencing (WGS) in combination with phenotypic assays. Methods Antimicrobial susceptibility to 23 drugs was determined by disk diffusion method. The genome sequence of EC1390 was determined by Nanopore MinION MK1C platform. Conjugation assays were performed to test the transferability of EC1390 plasmids to E. coli recipient C600. Phenotypic assays, including growth curve, biofilm formation, iron acquisition ability, and cell adhesion, were performed to characterize the function of EC1390 plasmids. Results Our results showed that EC1390 was only susceptible to tigecycline and colistin, and thus was classified as XDR E. coli. A de novo genome assembly was generated using Nanopore 73,050 reads with an N50 value of 20,936 bp and an N90 value of 7,624 bp. WGS analysis showed that EC1390 belonged to the O101-H10 serotype and phylogenetic group A E. coli. Moreover, EC1390 contained 2 conjugative plasmids with a replicon IncFIA (pEC1390-1 with 156,286 bp) and IncFII (pEC1390-2 with 71,840 bp), respectively. No significant difference was observed in the bacterial growth rate in LB broth and iron acquisition ability between C600, C600 containing pEC1390-1, C600 containing pEC1390-2, and C600 containing pEC1390-1 and pEC1390-2. However, the bacterial growth rate in nutrition-limited M9 broth was increased in C600 containing pEC1390-2, and the cell adhesion ability was increased in C600 containing both pEC1390-1 and pEC1390-2. Moreover, these plasmids modulated the biofilm formation under different conditions. Conclusions In summary, we characterized the genome of XDR-E. coli EC1390 and identified two plasmids contributing to the antimicrobial resistance, growth of bacteria in a nutrition-limited medium, biofilm formation, and cell adhesion. Supplementary Information The online version contains supplementary material available at 10.1186/s12866-022-02562-6.

Cephalosporins, fosfomycin, trimethoprim, fluoroquinolones, and amoxicillin in combination with β-lactamase inhibitors, are considered effective antibiotics to reduce the duration of E. coli-causing UTI symptoms in the last decade [6]. However, the extensive use of antibiotics has led to the emergence of antibiotic-resistant UTI pathogens [7]. E. coli isolated from patients with UTIs showed relatively low resistance to carbapenems (imipenem, ertapenem, meropenem, and doripenem), therefore, carbapenems were considered the drug of last resort for the treatment of UTIs caused by extended-spectrum β-lactamases-producing E. coli [8]. The mechanisms of carbapenem resistance in Enterobacteriaceae are strongly associated with carbapenemase production, overexpression of efflux pumps, and porin loss [9]. Moreover, the emergence of conjugative plasmids carrying carbapenemase has been increasingly reported among Enterobacteriaceae and is a matter of major clinical concern [10]. Currently, the application of next-generation sequencing (NGS) extends from microbial identification to epidemiological insight, microbial community investigation, and antimicrobial resistance prediction [11]. Therefore, in this study, we performed whole-genome sequencing (WGS) in combination with phenotypic assays to characterize an NDM-5-producing E. coli strain EC1390 isolated from a UTI patient.

Sampling and isolation of E. coli
E. coli isolates were recovered from patients with UTIs at National Cheng Kung University Hospital (NCKUH) with the approval of the NCKUH Research Ethics Committee (Approval No: B-ER-110-144). Informed consent was waived because of the retrospective nature of the study, the isolates were remnants from patient samples, and the analysis used anonymous clinical data. A total of 844 non-duplicate E. coli isolates were identified in the clinical laboratory by colony morphology, Gram stain, biochemical tests, and the Vitek system (bioMérieux, Marcy l'Etoile, France) according to the manufacturer's recommendations [12]. E. coli isolates were stored at − 80 °C in lysogeny broth (LB) containing 20% glycerol (v/v) until tested.

Genome sequencing, assembly, annotation, and analysis
The genomic DNA (gDNA) of EC1390 was extracted using a Presto ™ gDNA Bacteria Advanced Kit (Geneaid Biotech, Ltd., Taiwan) from a 5 mL broth culture grown in LB, followed by the protocol for Gram-negative bacteria. One μg gDNA was used to construct the sequencing library by using a Ligation Sequencing Kit (SQK-LSK109, Oxford Nanopore techonologies, Oxford, UK). KAPA HyperPure beads purchased from Oxford Nanopore techonologies were used to purify the gDNA fragments. The Oxford Nanopore MinION MK1C was used to determine the whole genome sequence of EC1390. A total of 300 ng gDNA was loaded onto R9.4.1 flow cell. The quality of reads generated was assessed using Fast QC v0.11.5 (https:// www. bioin forma tics. babra ham. ac. uk/ proje cts/ fastqc/). Raw signals were translated into a DNA sequence by using ONT Gussy basecalling program (version 4.2.3). EC1390 genome was constructed with Flye de novo assembler (version 2.9), and the options for backward compatibility (plasmids) and uneven coverage mode (meta) were used in this analysis [14].

Phylogenetic analysis of pEC1390-1 and pEC1390-2 closely related plasmids
The sequences of assembled plasmids, pEC1390-1 and pEC1390-2, were submitted to BLAST and compared to previously sequenced plasmids in NCBI GenBank database. Top twenty-five plasmids with the highest degree of similarity to the plasmid sequences of pEC1390-1 and pEC1390-2 (also with greater than 70% query coverage), respectively, were selected to perform CulstalW alignment (Gap opening penalty, 15.00; Gap extension penalty, 6.66) to construct Maximum Likelihood trees inferred with the Jukes-Cantor model by using Molecular Evolutionary Genetics Analysis (MEGA) software version 11.0. Bootstrap confidence values were obtained applying 300 replications.

Phenotypic detection of carbapenemase
mCIM and eCIM were performed on EC1390 and TCGs to detect the presence of carbapenemase [19]. A 1-μL loopful of bacteria was resuspended in a 2-mL tube of tryptone soy broth (TSB) (BD Difco, Detroit, MI, USA). Another 1-μL loopful of bacteria was resuspended in a 2-mL tube of TSB supplemented with EDTA (Thermo Fisher Scientific, Carlsbad, CA, USA) at a final concentration of 5 mM. A meropenem disk was placed in each tube, and the tubes were incubated at 35 °C for 4 h ± 15 min. Subsequently, the disks were removed and applied to Mueller-HInton agar plates (BD Difco) freshly plated with a 0.5 McFarland suspension of a carbapenem-susceptible E. coli ATCC 25922 strain. The plates were incubated at 35 °C for 16 to 20 h and the mCIM and eCIM results were interpreted as previously described [19]. K. pneumoniae ATCC BAA-1706 (carbapenemase-negative), K. pneumoniae ATCC BAA-1705 (bla KPC -positive), and K. pneumoniae ATCC BAA-2146 (bla NDM -positive) were used as internal controls for mCIM and eCIM tests.

Growth of bacteria in M9 and LB medium
The rate of growth was determined over 24 h at 37 °C using a BioTek Synergy HTX multimode reader [20]. 200 μL of sterile M9 (BD Difco) or LB (Neogen, Lansing, Michigan, USA) broth was dispensed into clear sterile 96-well microplates and inoculated with overnight culture of each strain to give a final inoculum of 1%. Readings were taken every fifteen minutes of absorbance of each well (scanned at OD 600 nm) in the microplates over the 24 h period. Each strain was analyzed in triplicate wells on at least three separate occasions to give nine data sets for analysis.

Biofilm formation assay
Biofilm formation by E. coli was assayed using a 96-well flat-bottom polystyrene microtiter plate, as described previously [21], with some modifications. Biofilms were stained with 0.1% crystal violet for 30 min. Crystal violet dye associated with biofilms was eluted with 75% ethanol for 30 min, and was quantified by absorbance at 590 nm.

CAS agar diffusion assay
Chrome azurol S (CAS) agar plates were prepared as described previously [22]. The modified CAS agar plate was punched with 6.5-mm diameter holes by using a tip (Corning ® Pasteur pipettes, no. L, 5 3/4 inch (146 mm)). Each hole was filled with 50 μL of the LB broth containing equal CFU of E. coli, incubated at 37˚C for 16 h, and monitored for the formation of an orange halo. CAS remains blue when complexed with iron, but it turns orange when iron is chelated by other iron chelators. Siderophore activity was expressed as the halo diameter.

Cell line and adhesion assay
The bladder epithelial cell line 5637 was a kind gift of Prof. Ching-Hao Teng (National Cheng-Kung University, Taiwan). 5637 cells were seeded into 12-well plates in RPMI-1640 media (Thermo Fisher Scientific) with 10% fetal bovine serum (FBS) (HyClone Laboratories, Logan, UT, USA) and 1% penicillin-streptomycin (HyClone) and grown to a monolayer. Before 2 h of infection, the culture medium was replaced with fresh medium without 1% penicillin-streptomycin. The cells were infected with a MOI of 10 of the E. coli strains, centrifuged at 600 g for 5 min to synchronize bacteria-host cell contact, and incubated at 37 °C for 2 h. The cells were washed three times with phosphate buffered saline (PBS). The cells were lysed by incubation with saponin at 37 °C for 15 min and plated on LB plates. The resulting colonies were counted to determine the adherent bacteria [23].

Statistical analysis
All analyses were performed in GraphPad Prism 8.0 software (GraphPad Software, Inc., San Diego, CA). Student's t test was performed to evaluate the statistical differences in phenotypic assay experiments. A value of p < 0.05 was considered statistically significant.
EC1390 was isolated from a 50-year-old woman admitting to NCKUH on September 4, 2020, for acute hypoxemic respiratory failure with acute respiratory distress syndrome. The patient also had chronic lymphocytic leukemia. The patient was hospitalized at NCKUH from September 4 to December 3. The timeline of the first 4 weeks of history of drug use and pathogen isolation of this patient was described in Fig. 1. EC1390 was isolated from the urine sample on day 14 of patient's hospitalization. Cefepime (day 1-8), doxycycline (day 1-7), and sulfamethoxazole/trimethoprim (day 4), were used for the treatment of bacterial infections before EC1390 was isolated ( Fig. 1).

Complete genome sequence of EC1390 and its characteristics
Nanopore WGS data to construct the EC1390 genome were obtained using the MinION flow cell according to the manufacturer's protocol. A de novo assembly was generated using Nanopore 73,050 reads with an N 50 value of 20,936 bp and an N 90 value of 7,624 bp (mean coverage was 316). Reads were assembled and returned three contigs with the head segment was almost identical to the tail segment, indicating the circular nature of the contigs. The closed EC1390 genome comprises a 4,779,543-bp chromosome, a 156,286-bp plasmid named pEC1390-1, and a 71,840-bp plasmid named pEC1390-2 ( Fig. 2 and Table 1). The number of plasmid in EC1390 was verified by Kado-Liu's plasmid extraction method (Fig. S1), however, 3 putative plasmids with an expected size lower than 10 kbp were not identified by Nanopore WGS. GC content of the EC1390 chromosome was 50.8% and that of plasmid pEC1390-1 and pEC1390-2 was 52.3% and 51.9%, respectively (Table 1).
Serotype was determined by the genotype of fliC, wzx, and wzy, and the results showed that EC1390 belonged to the O101-H10 serotype. Phylogenetic group analysis showed that EC1390 was group A E. coli. Acquired antimicrobial resistance genes were not identified on EC1390 chromosome sequence. However, an amino acid substitution E84K of ParC associated with nalidixic acid and quinolone resistance was identified on EC1390 chromosome sequence. The MLST scheme of EC1390 which is based on the sequences of seven housekeeping genes (gyrB999, recA2, mdh8, purA13, icd996, adk922, fumC11) confirmed that EC1390 belonged to a novel sequence type, and the results were further verified by sanger's sequencing.
To characterize CRISPR-Cas systems in EC1390, the presence of true CRISPRs in the EC1390 genome was assessed by the CRISPRCasFinder tool. To discriminate spurious CRISPR-like elements from the true CRISPRs, only CRISPRs classified with evidence levels 4 were considered for further analyses. Based on the selection criteria, one CRISPR type IE loci/arrays containing 10 spacers were detected in EC1390 (Table S1). To look for potential targets of EC1390 CRISPR spacers, CRISPRTarget was used to find the spacer against the RefSeq-Plasmid and PHAST databases. Three spacer sequences (TGT GTT TGC GGC ATT AAC GCT CAC CAG TAT TTC, CGA CGT GGT CAT GGG TGC TGC TGT TGC AGA GCCA, and CGA ATC GCG CAT ACC CTG CGC GTC GCC GCC TGC) were shown to target plasmids and/or phages (Table S1). However, the function/activity of CRISPR type IE in EC1390 was not further characterized.
The sequence of pEC1390-2 was highly similar to plasmids pNCYU-29-69-2 (accession number CP042601.   (1) shows the GC skew (G-C/G + C). The value is plotted as the deviation from the average GC skew of the entire sequence. Circle (2) shows the GC content, plotted using a sliding window. Circle (3,4) illustrate the coding sequences, 3 is backward strand, 4 is forward strand. The region containing multiple drug resistance genes on pEC1390-1 was indicated was identified in a mcr-3 positive E. coli from diseased swine. The phylogenetic tree showed that pEC1390-2 showed the highest similarity to plasmid pCRKP-1215_2 (coverage 92%, identity 99.02%) (Fig. 3B), which was identified in Enterobacteriaceae in South Korea.

The biological function of plasmids pEC1390-1 and pEC1390-2
Compare to phylogenetic group B2, group A strains usually have fewer virulence factors and are less virulent [24].
To characterize the biological function of two plasmids in EC1390 in addition to antibiotic resistance, we compared the bacterial growth rate, biofilm formation, iron acquisition and bladder epithelial cell adhesion, between EC1390, C600, and 3 TCGs (Fig. 4). Growth curves of E. coli strains were measured in triplicate for each of M9 broth (nutrition-limited) and LB broth (Fig. 4A). There was no significant difference in the growth rates between 5 strains in LB. Compared to C600 and 3 TCGs, UPEC 1390 showed a higher growth rate in M9 broth. Interestingly, Table 2 Antimicrobial resistance genes are present in the pEC1390-1 and pEC1390-2 a Antimicrobial resistance genes were identified only by using NCBI prokaryotic annotation pipeline but not ResFinder database  EC1390-TCG(S) showed a significant increase in growth rate in M9 broth (p < 0.05), compared to C600, EC1390-TCG(S), and EC1390-TCG(S + L) (Fig. 4A). Our results indicated that 5 examined strains showed a similar ability of iron acquisition (Fig. 4B). The results of biofilm formation in LB showed that EC1390-TCG(S) had an increase in biofilm formation after 1-day incubation (Fig. 4C). All 3 TCGs showed lower biofilm formation in M9 broth after 1-day incubation, compared to the C600, but EC1390-TCG(S) and EC1390-TCG(S + L) had increasing biofilm formation in M9 after 2-day and 3-day incubation, respectively ( Fig. 4C). Moreover, EC1390-TCG(S + L) had a higher cell adhesion ability compared to the C600 (Fig. 4D).

Discussion
In this study, we combined Nanopore WGS with phenotypic assays to characterize an NDM-5-producing E. coli isolated from UTI patient. EC1390 was isolated from the urine sample on day 14 of patient's hospitalization. Cefepime, doxycycline, and sulfamethoxazole/trimethoprim, were used for the treatment of bacterial infections before EC1390 was isolated (Fig. 1). The history of antibiotic therapies has been reported to be associated with the development of antibiotic resistance [25], however, the limit of our study is that we did not isolate EC1390 before antibiotic therapies to evaluate the association between the therapy and resistance development. On day 18 of the hospitalization, a carbapenem-resistant E. coli (CREC) was isolated from the patient's rectal swab, however, we did not further determine the clonality of CREC isolated from urine and rectal swab. The inconsistency of annotation results from NCBI PGAP and RAST server suggests the assemblies generated only from MinION data contained many indels and substitutions that affected the following analyses (Table 1 and Fig. S2). In addition, in the presence of a frameshift, the NCBI PGAP annotation pipeline generates a 'corrected' sequence of the genome, adding or subtracting bases to repair the translation frame [26]. Therefore, the number of coding sequences predicted by NCBI PGAP annotation was lower than the RAST server. Moreover, our Kado-Liu results showed the other 3 plasmids with sizes below 10 kbp, however, Nanopore WGS results showed only two large plasmids (Fig. 2 and Fig. S1). These results suggest the loss of small plasmids by using long-read Nanopore WGS platform. Many studies showed that the Nanopore/ Illumina hybrid sequencing platform enhances the accuracy of genome assemblies [27]. However, this study aimed to combine WGS and phenotypic assays to characterize EC1390, therefore, we did not use a hybrid sequencing platform to further verify the EC1390 genome. Instead, we increased the depth of coverage to enhance the accuracy of EC1390 genome.
Our results showed that EC1390 belonged to the O101-H10 serotype, phylogenetic group A, and untypable sequence type E. coli. The phylogenetic groups A, B1, and D had fewer virulence determinants than B2 strains, and O101-H10 serotype phylogenetic group A E. coli was relatively rare reported [24,28]. Mamani et al. reported that of the 2,427 E. coli bloodstream isolates in a Spanish hospital, 96 (4.0%) were extended-spectrum β-lactamase producers. Moreover, only 2 out of 96 ESBL-producer belonged to the O101-H10 serotype (1 ST32 strain contained bla CTX-M-32 , and 1 ST617 strain contained bla CTX-M-14 ) [28]. Therefore, the characteristics of O101-H10 serotype pathogenic E. coli were rarely determined. Currently, XDR-E. coli ST7624 belonged to O − :H34 serotype harboring bla CTX−M−14b was identified from milk samples [29]. The characteristics and circulation of these XDR pathogenic E. coli that can cause human diseases should be monitored.
Currently, international circulation of NDM-5-producing E. coli was reported but only a few UPEC isolates harboring bla NDM-5 were identified [30][31][32]. To our knowledge, there are only 2 studies that reported the E. coli carrying bla NDM-5 in Taiwan [33,34]. Chang et al. collected 237 carbapenem-nonsusceptible E. coli from 2012 to 2015, and only one ST117 E. coli contained bla NDM-5 [34]. From 2016 to 2018, Huang et al. showed 78 (0.5%) of the 17,018 E. coli clinical isolates collected at National Taiwan University Hospital were carbapenem-non-susceptible. Moreover, among the 78 carbapenem-nonsusceptible E. coli isolates, 23 (29.5%) were carbapenemase-producing E. coli, and 5 NDM-5-producers were identified (ST38, 167, 359, 410 (n = 2)) [33]. However, our results showed the distribution of pEC1390-1 similar bla NDM-5 -positive plasmids worldwide. Therefore, it is worth monitoring the spread of bla NDM-5 -positive plasmids in Taiwan. WGS is becoming widely used in pathogen identification, genome characterization, and phenotypic prediction. However, the virulence of plasmids carrying bla NDM-5 is less investigated. Although our results showed that the plasmid pEC1390-2 promoted the growth of bacteria in nutrition-limited M9 broth (Fig. 4A), however, we could not find out the potential annotated genes involved in this phenotypic change. There are 17 hypothetical proteins identified on pEC1390-2. Therefore, the roles of these genes in bacterial growth remain to be identified. Although our results indicated that pEC1390-1 and pEC1390-2 were not involved in iron acquisition (Fig. 4B), we still can not rule out the possibility that these two plasmids may play critical roles in biofilm formation and iron acquisition in vivo. Moreover, our recipient C600 showed high biofilm formation and iron acquisition, even compared to EC1390, a UPEC strain. These results showed that C600 may not a suitable for these two phenotypic assays.
Our results showed that the biofilm formation of TCGs compared to C600 was significantly different in different broth and incubation periods (Fig. 4C). Moreover, EC1390-TCG(S + L) had a higher cell adhesion ability compared to the C600 (Fig. 4D). These results suggest the cross-talk between pEC1390-1, pEC1390-2, and chromosomal genes, in different culture conditions and growth stages. Therefore, a transcriptome analysis to determine the cross-talk between genes located on chromosomal and plasmids would be useful to further elucidate the function(s) of two plasmids in UPEC.

Conclusions
In summary, we characterized an NDM-5-producing E. coli isolated from UTI patients by using a combination of WGS and phenotypic assays. We also demonstrated that these two conjugative plasmids were involved in antibiotic resistance, bacterial growth, biofilm formation, and bacterial adhesion to host cells.