Genome sequences of two clinical Escherichia coli isolates harboring the novel colistin-resistance gene variants mcr-1.26 and mcr-1.27

In 2018, the two colistin-resistant E. coli isolates 803-18 and 844-18 were sent from two hospitals in the federal state of Hesse, Germany, to the Robert Koch Institute for confirmation of colistin resistance and identification of the genetic resistance determinant. The E. coli isolate no. 803-18 was isolated from blood culture of a 79 years old male patient presenting fever. The second E. coli (no. 844-18) was isolated from an intraoperative swab of a 48 years old female patient.

In the Robert Koch Institute species identification and antimicrobial susceptibility testing was performed by broth microdilution according to EUCAST (clinical breakpoints (v 10.0) or epidemiological cut-off values (ECOFFs), (http://​www.​eucast.​org)). The following antibiotic substances and substance combinations were tested: ampicillin, cefotaxime, ceftazidime, cefoxitin, meropenem, gentamicin, amikacin, streptomycin, nalidixic acid, ciprofloxacin, chloramphenicol, tetracycline, sulfamethoxazole-trimethoprim and colistin.

PCR screening for the presence of colistin resistance gene mcr-1 and in E. coli frequently occurring β-lactamase genes (bla_TEM, bla_SHV, bla_{CTX-M-groups-1-2-9}) was performed as previously described [4, 7]. Furthermore, a PCR-based method to determine phylogenetic groups of E. coli was applied [8].

The transferability of mcr-1 genes of isolates 803-18 and 844-18 was investigated by broth mate conjugation experiments; the sodium azide-resistant strain E. coli J53 Azi^r served as the recipient. Transconjugants were selected on Luria–Bertani agar plates containing sodium azide (200 mg/L) and a colistin disk (10 µg). Antimicrobial susceptibilities and presence of mcr-1 and β-lactamase genes were tested for selected transconjugants. To further verify the transfer of plasmids, general plasmid content and plasmid size were determined by S1-nuclease restriction and pulsed-field gel electrophoresis (PFGE) as described before [9].

DNA extraction was performed using the DNeasy Blood & Tissue kit (Qiagen) and extracted DNA was quantified using the Qubit dsDNA HS Assay Kit (Invitrogen), both according to the manufacturer’s protocols. Genomic libraries were generated with the NexteraXT kit (Illumina). Whole-genome sequencing (WGS) was carried out using the Illumina HiSeq 1500 (2 × 250 bp; HiSeq Rapid SBS Kit v2) benchtop device in ‘Rapid Run Mode’.

Raw reads were processed using the pipeline QCumber (v 2.1.1), where the FastQC (v 0.11.5), Trimmomatic (v 0.36; options ‘sliding window 4:20’, ‘MINLEN: 50 bp’) and Kraken (v 1.0.0) algorithms were included (https://​gitlab.​com/​RKIBioinformatic​sPipelines/​QCumber/​). The draft de novo reconstruction was done using the SPAdes algorithm (v 3.12.0) with default parameters. In a subsequent filtering step, all contigs < 200 bp were excluded. Using the QUAST algorithm without a reference sequence, the quality of draft genome sequences was investigated [10].

The de novo reconstructed sequences were used to extract multilocus sequence types (MLST; Achtmann scheme) and complex types (CT), based on core genome multilocus sequence typing (cgMLST; 2513 allele targets) by utilizing the SeqSphere⁺ software (v 6.0.0, Ridom GmbH) as described before [11, 12]. Gene annotation was determined by the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) [13]. To predict plasmid content in silico, the PlasmidFinder web tool (v 2.1) was used [14]. The NCBI blastn database was used to search for known replicon types, in case of contigs carrying a predicted replicon. Further, the SerotypeFinder (v 2.0) and the VirulenceFinder (v 2.0) web tools were used to characterize the isolates [15, 16].

Using raw reads, the tool ResFinder (v 3.1.0) was used to identify mcr genes [17]. Identified mcr-1-like genes were extracted from the contigs and aligned to a mcr-1.1 reference sequence (gene accession no: NG_050417.1) to calculate a gene-based phylogeny using PhyML (Jukes-Cantor; 500 bootstraps) [4, 18]. Sequences were translated and checked for synonymous and non-synonymous mutations using the Geneious Prime software (v 2020.0.5). To verify identified non-synonymous mutations, primers were designed (Mcr-1a FWD 5′-CAGTATGGGATTGCGCAATGA-3′, Mcr-1a REV 5′-GGGCATTTTGGAGCATGGTC-3′; product size 482 bp, Tm = 59 °C) to perform Sanger sequencing after PCR amplification. The resulting mcr-1-like gene sequences were submitted to NCBI (National Center for Biotechnology Information)/NLM (National Library of Medicine) to determine novel allele numbers (https://​www.​ncbi.​nlm.​nih.​gov/​pathogens/​submit-beta-lactamase/​) as it has been proposed [19]. Contigs, on which the mcr-1-like genes were located, were investigated by BLAST for known plasmid origins of replication (as of December 2019; https://​blast.​ncbi.​nlm.​nih.​gov/​Blast.​cgi).

To ensure pure cultures and to phenotypically verify the species, single colonies were repeatedly cultivated on different media (Müller-Hinton agar with sheep blood and Bile-Chrysoidin-Glycerol agar). Further, automated species identification (VITEK 2 GN) was performed. For DNA extraction, single colonies were used. After sequencing, the Kraken algorithm results, also implemented in the QCumber pipeline, were inspected for potential contaminations [20]. De novo assembled genome sequences were quality checked using QUAST.

Both E. coli isolates 803-18 and 844-18 were resistant to colistin (MIC = 4 mg/L), ampicillin, sulfamethoxazole/trimethoprim, nalidixic acid, ciprofloxacin and tetracycline (Table 1). Isolate 803-18 was additionally resistant to streptomycin, and isolate 844-18 was additionally resistant to chloramphenicol. Both isolates remained fully susceptible to cephalosporins and carbapenems (Table 2).

The PCR-confirmed mcr-1-like genes in both isolates could be transferred in conjugation experiments. Transconjugant 803-18 Tc1 harbored two plasmids (ca. 33 kb and 90 kb) and showed a colistin MIC of 2 mg/L. Additional resistance to streptomycin and ampicillin was detected; presence of mcr-1-like and β-lactamase gene bla_TEM was confirmed by PCR. Transconjugant 844-18 Tc1 was positive for the mcr-1-like gene, showed a colistin MIC of 2 mg/L and harbored one plasmid of ca. 33 kb size. Additional resistance to chloramphenicol, tetracycline and sulfamethoxazole-trimethoprim was detected (Table 2). These resistances might be encoded on smaller plasmids but plasmids smaller than 20 kb were not detectable by S1-PFGE.

A total of 1,349,261 raw reads were obtained for E. coli no. 803-18 and 1,700,507 for E. coli no. 844-18. After de novo reconstruction of isolate 803-18, 154 scaffolds (155 contigs) were assembled, with N50: 119,280 bp and L50: 13. On average, the assembled draft genome was covered 84x. The draft genome size was determined as 4.92 Mb, with 50.6% GC content; and 4854 genes, encoding 4516 proteins, were predicted. The draft assembly of isolate 844-18 resulted in 192 scaffolds (193 contigs), with N50: 145,421 bp and L50: 12; with 45 × genome coverage. The determined draft genome size was 5.31 Mb, with 50.6% GC content; and 5228 genes, encoding 4923 proteins, were predicted.

ResFinder detected the presence of several resistance genes in isolates 803-18 and 844-18, respectively (Table 1) contributing to resistance to colistin (mcr-1-like), penicillins (bla_TEM-1B), sulfonamides (sul1, sul2), trimethoprim (dfrA1, dfrA14-like), aminoglycosides (str-A-like, str-B-like, aadA1), tetracyclines (tetA) and phenicols (catA1-like) (Table 1). These results corresponded to the phenotype of the isolates, which highlights the general applicability of WGS-based data also for antibiotic resistance predictions, as it was discussed before [21].

VirulenceFinder detected genes in both isolates that were associated with fitness or virulence traits (colonization and fitness factors) in E. coli, named iroN, gad, lpfA and iss encoding enterobactin siderophore receptor protein, glutamate decarboxylase, long polar fimbriae and increased serum survival, respectively (Table 1). For isolate 844-18 three further genes were detected: cma, encoding the bacteriocin colicin M, air encoding the adhesin enteroaggregative immunoglobulin repeat protein and its regulator eilA (hilA homolog in Salmonella) [22, 23]. However, virulence genes (e.g. eae and stx) that are associated with a specific pathotype (e.g. EAEC and EHEC) were not detected in the two isolates.

The different typing approaches showed that the two isolates were genotypically dissimilar; at core-genome level (cgMLST-analysis) the isolates showed a distance of 2362 alleles to each other. Isolate 803-18 was assigned to phylogenetic group B1, serotype H45, sequence type ST155 and cgMLST-based complex type CT7500; isolate 844-18 was identified as phylogenetic group D, serotype O15:H18, ST69 and CT7508 (Table 1). Phylogenetic group B1 is known to mainly comprise environmental and animal isolates, whereas phylogenetic group D is known to include more (urogenital-) pathogenic E. coli [24]. This result seems to be concordant with MLST, since E. coli-ST155 has been described as sequence type with zoonotic potential and plasmid-mediated spread of antibiotic resistance, whereas E. coli-ST69 was described as a pandemic and pathogenic lineage [25, 26]. The latter is supported by the presence of additional virulence genes in E. coli-ST69 isolate 844-18 that are involved in adherence to epithelial cells and biofilm formation (adhesion AIR and regulator protein EilA), and the fitness factor colicin M, a bacteriocin that kills other sensitive E. coli strains [22, 23]. The typing results of the two isolates, led to the assumption that there is no certain Mcr-1-like producing strain in the hospitals, instead this might be a hint for a potentially community-based influx of mcr-1-like mediated colistin-resistance via different strains into hospitals, as discussed in other studies [27, 28].

Based on the PlasmidFinder results, for both isolates several plasmids could be predicted; and S1-PFGE analysis confirmed the presence of at least three plasmids in each of the two isolates (Table 1). PlasmidFinder was able to predict more plasmids, e.g. several Col-like plasmids (Table 1). These plasmids of small size could not be seen in S1-PFGE analysis and were not further analyzed in the present study.

Blast analyses of the scaffolds carrying the mcr-1-like genes predicted their location on IncX4 plasmids, with high similarity (> 99%) to mcr-1 of an IncX4 plasmid of 33 kb size (GenBank accession: CP042970.1) of an E. coli isolate from raw milk cheese in Egypt. For our isolates, the IncX4 plasmid could be reconstructed with two scaffolds: for isolate 803-18 a 32,744 bp mcr-1-like-positive scaffold and an 820 bp scaffold; for isolate 844-18 a 32,738 bp mcr-1-like-positive scaffold and an 821 bp scaffold. However, in CP042970.1 and in both reconstructed IncX4 plasmids the mcr-1-like gene was not part of the ISApl1 transposon that is known to be associated with mcr-1 dissemination as described previously [29, 30]. Instead both plasmids included the IS6-like element (encoding an IS26 family transposase). We reconstructed a 33 kb ring structure, with 100% coverage and 99.6% pairwise identity compared to plasmid CP042970.1. Furthermore, high identity with further mcr-1 carrying plasmids in the NCBI database was detected (GenBank accession: MF449287.1; MK172815.1). These were from fresh water from Italy (MF449287.1) and human origin from Russia (MK172815.1), also showing > 99% coverage and > 99% pairwise identity with the reconstructed 33 kb plasmid of 803-18 and 844-18. This indicates a worldwide spread of this type of plasmid with colistin resistance gene mcr-1 in E. coli.

Alignment of the extracted mcr-1-like genes of isolates 803-18 and 844-18 and known mcr-1 variants (as of December 2019) to the reference sequence of mcr-1.1 (NG_050417.1) revealed putative point mutations (Fig. 1). These point mutations were confirmed by PCR amplification and Sanger sequencing. Subsequent translation revealed these point mutations were non-synonymous mutations, resulting in amino acid substitutions Met1Thr (isolate 803-18) and Tyr9Cys (isolate 844-18) (Fig. 1B). The substitution Met1Thr in isolate 803-18 was due to the ACG (Thr) codon that has been reported by Hecht et al. for its potential role in non-canonical initiation in E. coli [31]. It is important to note that in mcr-1.26 an ATG (Met) is present immediately after ACG (Thr) and therefore we are uncertain of the actual effect of Met1Thr on the translation initiation of mcr-1.26 in isolate 803-18. This warrants further investigation. However, the conjugation experiment confirmed an increase in colistin MIC of the transconjugant (2 mg/L, Table 2).

Both mcr-1-like sequences were submitted to NCBI/NLM and assigned with two novel mcr-1 allele numbers: mcr-1.26 (isolate 803-18; NCBI Reference Sequence: NG_068217.1; RefSeq CDS region in nucleotide: JAAGSA010000042.1 3574-5196(+); protein accession: WP_034169413.1) and mcr-1.27 (isolate 844-18; NCBI Reference Sequence: NG_068218.1; RefSeq CDS region in nucleotide: JAAGSB010000042.1 27547-29172(−); protein accession: WP_163397051.1). The identification of two novel mcr-variants in hospitals in the same region and within 1 year shows that the spread of plasmid-mediated colistin-resistance seems to rapidly progress and new variants are constantly emerging [28].