Detection of a novel mcr-5.4 gene variant in hospital tap water by shotgun metagenomic sequencing

Sir,

Colistin is considered a last-resort antibiotic for treating serious infections caused by MDR Gram-negative bacteria. The efficacy of this antibiotic is challenged by the emergence and global spread of mobile colistin resistance (mcr) determinants, which threaten human, animal and environmental health. The first mobile colistin resistance gene (mcr-1) was reported in 2015 and since then up to eight different variants have been described.¹ In 2017, Borowiak et al.² described a new transposon-associated phosphoethanolamine transferase mediating colistin resistance, named mcr-5, in d-tartrate-fermenting Salmonella enterica subsp. enterica serovar Paratyphi B isolated from poultry. The mcr-5.3 variant has been recently reported in Stenotrophomonas spp. from sewage water.³ Here we report for the first time (to the best of our knowledge) the detection of an mcr-5 gene in a hospital water environment using short-read metagenomic sequencing (SRMseq) and subsequent characterization using long-read metagenomic sequencing (LRMseq) to reveal its genetic environment.

In June 2017, eight tap-water samples (900 mL) were collected at the University Medical Center Groningen. Water samples were filtered (0.2 μm) and after DNA extraction (PowerWater DNA Extraction Kit, QIAGEN), SRMseq was performed on a MiSeq instrument (500 cycles) (Illumina). Antibiotic resistance genes were identified in the metagenome assemblies (CLC Genomics Workbench v10.1.1, QIAGEN) using ABRicate-0.7 (https://github.com/tseemann/abricate) and applying the following thresholds: >70% identity and >80% coverage. One sample contained an mcr-type gene (5× sequencing depth), with the nucleotide change 313C>T (amino acid change F105L) with respect to the original mcr-5.1 gene, which was designated mcr-5.4 by NCBI (accession no. MK965519). This sample was selected for LRMseq; the DNA libraries were prepared using the Rapid PCR Barcoding Kit (SQK-RPB004) from Oxford Nanopore Technologies (ONT) and loaded into a FLO-MIN106 R9.4 flow cell. The run was performed on a MinION device (ONT) and it proceeded for 24 h. The data were basecalled using Albacore (https://github.com/rrwick/Basecalling-comparison) and further processed with Poretools⁴ and Porechop (https://github.com/rrwick/Porechop). Trimmed reads from SRMseq and LRMseq were used for hybrid-assembly analysis by metaSPAdes-3.13.0.⁵ After a BLAST search using the hybrid contig containing the mcr-5.4 gene, the plasmid pSE13-SA01718 (accession no. KY807921.1) was listed as one of the hits with the highest identity and we used it as a reference for genome comparison with the Artemis Comparison Tool (ACT) v1.0.⁶ The mcr-5.4-carrying contig from the hybrid assembly was annotated using PATRIC v3.5.27.⁷ Trimmed reads from SRMseq were used to investigate the bacterial composition by OneCodex.⁸ Finally, in order to predict the bacterial host of the mcr-5.4 gene, a contig-binning analysis of the hybrid-assembled metagenome was performed using MaxBin2 v2.2.4 (https://sourceforge.net/projects/maxbin2/), probability threshold 0.9 and minimum contig length 1000 bp. The resulting bin containing the mcr-5.4 gene was selected for taxonomy classification using Kraken2 (https://github.com/DerrickWood/kraken2) (minikraken2 DB v1).

SRMseq showed the mcr-5.4 gene detected in a contig of 2113 bp flanked by two truncated protein-coding sequences (CDSs), encoding the ChrB domain protein (involved in chromate resistance) and the Major Facilitator Superfamily (MFS) transporter. The hybrid-assembly analysis resulted in a contig of 8456 bp consisting of nine CDSs and four truncated CDSs (Figure 1). Comparative analysis of the genetic environment of the mcr-5 gene, between the annotated hybrid metagenome contig and the reference plasmid pSE13-SA01718, showed a region of 4670 bp with 98% identity, corresponding to the backbone of the Tn6452 transposon (Figure 1). We observed three truncated CDSs for the MFS-type transporter in our contig instead of two as previously described in the reference sequence pSE13-SA01718. These differences did not appear to be due to sequencing errors when we checked the sequence MK965519, (i) using pilon (https://github.com/broadinstitute/pilon) to correct for errors in short-read sequencing data and (ii) using CLC Genomic Workbench to update the hybrid contig by mapping both long and short reads against the hybrid contig. We also observed a region of 3786 bp, with no identity either with the reference plasmid pSE13-SA01718 (Figure 1) or with any other sequence in the GenBank database.

Figure 1.

Comparative analysis of the genetic environment of mcr-5 between the reference plasmid pSE13-SA01718 (accession no. KY807921.1) and the annotated hybrid metagenome contig (accession no. MK965519). The contig carrying the mcr-5.4 gene consists of the following putative gene products: 7-carboxy-7-deazaguanine synthase (queE), 7-cyano-7-deazaguanine synthase (queC), glycine cleavage system transcriptional antiactivator GcvR (gcvR), thiol peroxidase (tpx), sulphurtransferase TusA family protein (sirA), hypothetical protein (hp), truncated MFS-type transporter (Δmsf), lipid A phosphoethanolamine transferase (mcr-5.4), ChrB domain protein (chrB), transposon resolvase (tnpR) and truncated transposon transposase (ΔtnpA). Areas with 98% identity between sequences are represented in light grey. Arrows indicate the position and direction of the genes. The transposon Tn6452 sequence in the reference plasmid pSE13-SA01718 is bounded by inverted repeats: IRL and IRR.

Species previously described to harbour an mcr-5 gene are Escherichia coli, Pseudomonas aeruginosa, Salmonella enterica, Aeromonas hydrophila and Cupriavidus gilardii. The bacterial composition analysis of the water sample using SRMseq showed the presence of Pseudomonas spp. (relative abundance: 0.004%), Cupriavidus spp. (relative abundance: 0.001%) and Aeromonas spp. (relative abundance: 0.0003%). The binning analysis produced a bin positive for the mcr-5.4 gene consisting of 1336 contigs (genome size: 5 175 285 bp; genome completeness: 68.2%). This bin was taxonomically classified as bacteria (70.73%) and proteobacteria (64.90%), and from this the most abundant class was Gammaproteobacteria (37.20%) (order Pseudomonadales, 15.57%), followed by Betaproteobacteria (14.90%) (order Burkholderiales, 10.63%).

Colistin resistance determinants (mcr) have been rarely reported in water environments; mcr-1 has been detected in both hospital sewage and in environmental water streams and mcr-3 in environmental water.^9,10 To the best of our knowledge, this is the first-time description of an mcr-5 gene in an indoor and healthcare water environment. Despite the fact that the comparative analysis showed the hybrid contig covering a large region of Tn6452, neither the left inverted repeat (IRL) nor the right inverted repeat (IRR) have been found. In addition, the lack of the right transposon region does not allow us to search for other possible inverted repeats. Thus, it is not possible to conclude whether the described mcr-5.4 gene is transferable or not. Taxonomic analysis suggested the order of Pseudomonadales as the most probable host of the mcr-5.4 gene in the water sample. Further studies are needed to determine the frequency of this gene in hospital water and other water environments and to evaluate the potential risks for patients and healthcare workers.