More than mcr: canonical plasmid- and transposon-encoded mobilized colistin resistance genes represent a subset of phosphoethanolamine transferases

Abstract

Mobilized colistin resistance genes (mcr) may confer resistance to the last-resort antimicrobial colistin and can often be transmitted horizontally. mcr encode phosphoethanolamine transferases (PET), which are closely related to chromosomally encoded, intrinsic lipid modification PET (i-PET; e.g., EptA, EptB, CptA). To gain insight into the evolution of mcr within the context of i-PET, we identified 69,814 MCR-like proteins present across 256 bacterial genera (obtained by querying known MCR family representatives against the National Center for Biotechnology Information [NCBI] non-redundant protein database via protein BLAST). We subsequently identified 125 putative novel mcr-like genes, which were located on the same contig as (i) ≥1 plasmid replicon and (ii) ≥1 additional antimicrobial resistance gene (obtained by querying the PlasmidFinder database and NCBI's National Database of Antibiotic Resistant Organisms, respectively, via nucleotide BLAST). At 80% amino acid identity, these putative novel MCR-like proteins formed 13 clusters, five of which represented putative novel MCR families. Sequence similarity and a maximum likelihood phylogeny of mcr, putative novel mcr-like, and ipet genes indicated that sequence similarity was insufficient to discriminate mcr from ipet genes. A mixed-effect model of evolution (MEME) indicated that site- and branch-specific positive selection played a role in the evolution of alleles within the mcr-2 and mcr-9 families. MEME suggested that positive selection played a role in the diversification of several residues in structurally important regions, including (i) a bridging region that connects the membrane-bound and catalytic periplasmic domains, and (ii) a periplasmic loop juxtaposing the substrate entry tunnel. Moreover, eptA and mcr were localized within different genomic contexts. Canonical eptA genes were typically chromosomally encoded in an operon with a two-component regulatory system or adjacent to a TetR-type regulator. Conversely, mcr were represented by single-gene operons or adjacent to pap2 and dgkA, which encode a PAP2 family lipid A phosphatase and diacylglycerol kinase, respectively. Our data suggest that eptA can give rise to "colistin resistance genes" through various mechanisms, including mobilization, selection, and diversification of genomic context and regulatory pathways. These mechanisms likely altered gene expression levels and enzyme activity, allowing bona fide eptA to evolve to function in colistin resistance.

Keywords: MCR; antimicrobial resistance; colistin; horizontal gene transfer; mobile genetic element; phosphoethanolamine transferase; plasmid.

Figure 1

(A) Schematic diagram of colistin’s mode of action and resistance mechanism. Colistin, a cationic antimicrobial peptide, binds to the negatively charged lipid A, displacing membrane-bound cations and disrupting membrane integrity by inserting the hydrophobic tail into the membrane’s lipid (panel A, left). Lipid A modification neutralizes the negative membrane charge, thus reducing colistin binding and cell susceptibility to colistin (panel A, right). IM, inner membrane; LPS, lipopolysaccharide; MCR, mobilized colistin resistance protein; OM, outer membrane; PE, phosphoethanolamine. (B) Schematic outline of the main methodologies used to acquire and characterize MCR and MCR-like proteins from publicly available whole-genome sequencing (WGS) data (see the Materials and Methods section for details). The number of MCR, putative novel MCR-like, and/or intrinsic lipid modification phosphoethanolamine transferase (i-PET) proteins produced at relevant steps are shown in green text; the number of such genes/proteins used in final analyses are shown in blue text. AMR, antimicrobial resistance; BLASTP, protein basic local alignment search tool; CDSs, coding sequences; DB, database; IPG, Identical Protein Group; LPS, lipopolysaccharide; MCR, mobilized colistin resistance amino acid sequences; ML, maximum likelihood; mOTUs, marker gene-based operational taxonomic units; NCBI, National Center for Biotechnology Information; NDARO, National Database of Antibiotic Resistant Organisms; nr, non-redundant protein sequence; (i) -PET, (intrinsic) phosphoethanolamine transferases; tBLASTN, translated nucleotide basic local alignment search tool.

Figure 2

Maximum likelihood (ML) phylogeny inferred from a nucleotide back-translation-based multiple sequence alignment (NT_btn-MSA) of (i) 98 mcr alleles and (ii) 125 unique sequences of mcr-like genes located on the same contig as ≥1 plasmid replicon and ≥1 other antimicrobial resistance (AMR) gene (i.e., “putative novel mcr-like genes”). Sequences were aligned using MUSCLE. The ML phylogeny was constructed with RAxML, using the GTRGAMMA nucleotide substitution model and 100 bootstrap replicates. The tree was edited using the iTOL web server (https://itol.embl.de/) and rooted at the midpoint, with branch lengths reported in substitutions per site. Branches with bootstrap values ≥70% are denoted by blue stars. Clades exclusively composed of putative novel mcr-like genes were collapsed for clarity (green branches; see Supplementary Figure S1 for a fully expanded tree). Tip label shading corresponds to known mcr alleles (light blue) and putative novel mcr-like genes identified in this study (light green). The colors of the inner ring represent different mcr families (mcr-1 to -10). The outer graph represents the maximum reported colistin minimum inhibitory concentration (MIC) values of native strains harboring different mcr alleles. The colistin breakpoint established by the Clinical and Laboratory Standards Institute (CLSI) is 2 mg/L (red line). Isolates with colistin MIC ≥2 mg/L are considered colistin-resistant. Colistin MIC values were compiled from different mcr-harboring Gram-negative species as listed in Supplementary Table S8 ; metadata associated with each gene can be found in Supplementary Table S6 .

Figure 3

Maximum likelihood (ML) phylogeny inferred from a nucleotide back-translation-based multiple sequence alignment (NT_btn-MSA) of (i) 98 mcr alleles, (ii) 125 unique sequences of mcr-like genes located on the same contig as ≥1 plasmid replicon and ≥1 other antimicrobial resistance (AMR) gene (i.e., “putative novel mcr-like genes”), and (iii) 237 chromosomal phosphoethanolamine transferase (ipet) genes. Sequences were aligned using MUSCLE. The ML phylogeny was constructed with RAxML, using the GTRGAMMA nucleotide substitution model and 100 bootstrap replicates. The tree was edited using the iTOL web server (https://itol.embl.de/) and rooted at the midpoint, with branch lengths reported in substitutions per site. Branches with bootstrap values ≥70% are denoted by blue stars. Linages and clades described in the main text are numbered in blue, while subclades and clusters are numbered in magenta. Clades exclusively composed of genes from the same family were collapsed and color-coded, as shown in the left color legend key (see Supplementary Figure S2 for a fully expanded tree). Five PET families clustered based on an 80% AA identity threshold representing 26 putative novel mcr genes are identified by black boxes around the taxa names with capital letters A to E and asterisks in color strip (A). Color-coded regions in strip (A) denote mcr families, putative novel mcr-like genes, and eptA, eptB, and cptA homologues. Color-coded regions in strip (B) denote gene localization as chromosomally encoded (cyan), plasmid-encoded (green), or unplaced (white); localization was assigned based on the first gene reported in the literature (note that for some gene families, e.g., mcr-3, chromosomally encoded genes have also been reported). Color-coded regions in strip (C) represent the regulatory system juxtaposing phosphoethanolamine transferase (PET)-encoding genes: magenta regions represent genes contained within the same operon as a two-component response sensor-regulatory system; cyan regions represent genes located divergent of a two-component response sensor-regulatory system; orange regions represent genes located adjacent to a TetR-type regulator; pink regions represent genes located adjacent to an AraC-type regulator; yellow region represents gene located adjacent to a YebC-type regulator; white regions represent genes represented by single-gene operons with no upstream or downstream regulatory protein. Color-coded regions in strip (D) represent PET-encoding genes localized adjacent to accessory enzyme-encoding genes, including pap2 encoding lipid A phosphatase (light green), dgkA encoding diacylglycerol kinase (red), or both pap2 and dgkA genes (dark green); white regions represent the absence of accessory enzyme-encoding genes upstream or downstream of PET-encoding genes.

Figure 4

Heatmap showcasing colistin minimum inhibitory concentration (MIC) values reported for MCR alleles in their native strains and in heterologous expression systems. MIC values were obtained from a review of the literature (see the Materials and Methods section for details). Heatmap cell colors correspond to base-2 logarithm-transformed MIC values, where white cells denote a value of 1 (i.e., the log2-transformed colistin resistance breakpoint of 2 mg/L, established by the Clinical and Laboratory Standards Institute [CLSI]); blue shading denotes MIC values below this breakpoint, and red shading denotes MIC values above. MCR alleles with no data reported for a given species (due to the allele not being present in the species, or due to an MIC not being reported) are denoted in the heatmap by black shading. For individual species with multiple MIC values reported for a single MCR allele, the median MIC value is reported ( Supplementary Table S8 ). For mcr-2.1, Salmonella enterica was selected to represent an unspecified “Salmonella spp.” reported in the literature for the native strain ( Supplementary Table S8 ). The maximum likelihood (ML) phylogeny displayed to the left of the heatmap denotes type strains and/or NCBI representative genomes of reported species for the native strains ( Supplementary Table S8 ). Phylogeny tip label colors correspond to taxonomic families assigned via the Genome Taxonomy Database Toolkit (GTDB-Tk). The phylogeny was constructed with IQ-TREE, using an alignment of amino acid sequences produced via GTDB-Tk as input. The phylogeny is rooted along the midpoint, with branch lengths reported in substitutions per site. Branch labels denote branch support percentages obtained using the ultrafast bootstrap approximation. The phylogeny was edited and displayed using iTOL.

Figure 5

Identification of amino acid (AA) residues under branch-specific positive selection. A phylogeny of 98 known mcr alleles produced and partitioned by GARD (https://www.datamonkey.org/gard) shows partition 1 (A) and partition 2 (B) branch-specific AA sites evolving under positive selection. Sequences were aligned using MUSCLE, and the resulting multiple sequence alignment (MSA) was supplied as input to GARD. The partitioned dataset produced by GARD was supplied as input to MEME (mixed-effect model of evolution-based selection analysis; https://www.datamonkey.org/meme), which was used to identify AA residues under branch-specific positive selection. The tree (default output/rooting produced by GARD/MEME, with branch lengths reported in substitutions per site) was edited using the iTOL web server (https://itol.embl.de/). Color-coded regions and triangles represent AA residues under allele-specific and branch-specific positive selection, respectively.

Figure 6

Structural localization and conservation of amino acid (AA) residues under branch-specific positive selection. (A) Structural model of the MCR-1 protein, constructed based on Neisseria meningitidis phosphoethanolamine transferase EptA (Anandan et al., 2017). The MCR structural model was constructed using the Phyre2 server, and the structure was viewed and edited using UCSF ChimeraX. The structural model shows the transmembrane-anchored domain (yellow) and the soluble periplasmic catalytic domain (light green) connected by a bridging helix and extended loop (light blue). AA residues involved in zinc (yellow circle) and phosphoethanolamine (pEtN) binding are colored dark blue and green, respectively. Branch-specific AA residues evolving under positive selection are shown in magenta (partition 1) and red (partition 2). (B) Close-up of the MCR active site showing the localization of branch-specific AA residues under positive selection in relation to the active site and the substrate entry tunnel. The structure shows Gln¹⁰⁷, which is located at a periplasmic loop connecting the third and fourth transmembrane segments juxtaposing the substrate entry tunnel; Ser¹⁹¹, which is located at a bridging region that connects the N-terminal membrane domain and the C-terminal catalytic domain; and Pro³⁹⁷, which is located adjacent to the active site. (C) Web-logo of nucleotide and AA sequences showing the conservation of 19 AA residues identified by MEME (https://www.datamonkey.org/meme) to have evolved under positive selection among the 98 known MCR alleles.