Deciphering MCR-2 Colistin Resistance

ABSTRACT Antibiotic resistance is a prevalent problem in public health worldwide. In general, the carbapenem β-lactam antibiotics are considered a final resort against lethal infections by multidrug-resistant bacteria. Colistin is a cationic polypeptide antibiotic and acts as the last line of defense for treatment of carbapenem-resistant bacteria. Very recently, a new plasmid-borne colistin resistance gene, mcr-2, was revealed soon after the discovery of the paradigm gene mcr-1, which has disseminated globally. However, the molecular mechanisms for MCR-2 colistin resistance are poorly understood. Here we show a unique transposon unit that facilitates the acquisition and transfer of mcr-2. Evolutionary analyses suggested that both MCR-2 and MCR-1 might be traced to their cousin phosphoethanolamine (PEA) lipid A transferase from a known polymyxin producer, Paenibacillus. Transcriptional analyses showed that the level of mcr-2 transcripts is relatively higher than that of mcr-1. Genetic deletions revealed that the transmembrane regions (TM1 and TM2) of both MCR-1 and MCR-2 are critical for their location and function in bacterial periplasm, and domain swapping indicated that the TM2 is more efficient than TM1. Matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) confirmed that all four MCR proteins (MCR-1, MCR-2, and two chimeric versions [TM1-MCR-2 and TM2-MCR-1]) can catalyze chemical modification of lipid A moiety anchored on lipopolysaccharide (LPS) with the addition of phosphoethanolamine to the phosphate group at the 4′ position of the sugar. Structure-guided site-directed mutagenesis defined an essential 6-residue-requiring zinc-binding/catalytic motif for MCR-2 colistin resistance. The results further our mechanistic understanding of transferable colistin resistance, providing clues to improve clinical therapeutics targeting severe infections by MCR-2-containing pathogens.

As a type of ␤-lactam antibiotic, carbapenems are generally used as the last resort against lethal infections by multidrug-resistant Gram-negative pathogens (3). Unfortunately the emergence of the notorious New Delhi ␤-lactamases (NDM-1 [3] and its variants NDM-5 [4], NDM-7 [5], and NDM-9 [6]) in diversified species of Enterobacteriaceae (e.g., Escherichia coli and Klebsiella pneumoniae) has almost pushed us to the cusp of the postantibiotic era. To conquer the carbapenemase (and/or extended-spectrum ␤-lactamase [ESBL])-producing pathogens, the cationic polypeptide antibiotic colistin (also referred to as polymyxin E; see Fig. S1 in the supplemental material) has to be reintroduced clinically as a choice of priority (7,8), even though it has an appreciable level of nephrotoxicity and neurotoxicity (9). Given the fact that colistin is used as a veterinary medicine (the old generation of antibiotics) and has been extensively applied in pig and poultry production for over 50 years, it is understood that in sporadic cases, chromosome-encoded colistin resistance is detected, which is frequently associated with point mutations in certain genes: two sets of two-component systems (pmrAB in Salmonella [10] and phoPQ in K. pneumoniae [11]) and a regulator gene, mgrB, in K. pneumoniae (11).
Earlier in 2016, Liu and colleagues (12) reported a new type of transferable colistin resistance, which is not spread by colonial expansion but by plasmid-mediated transmission in Enterobacteriaceae. The expression of the mobilized colistin resistance gene mcr-1 alone conferred the phenotype of robust resistance to colistin (13); the MCR-1 protein product is annotated as a putative member of the phosphoethanolamine (PEA) transferase family. Since its first discovery in southern China in late 2015, the mcr-1 gene has been found to have disseminated into over 30 countries (with even the mostly developed country, the United States, not exempt [14,15]) across 5 of 7 continents worldwide (16,17). This kind of colistin resistance was found to lie in the chemical addition of phosphoethanolamine to the 4= position of the lipid A moiety of lipopolysaccharide (LPS) on the outer leaflet of the bacterial outer membrane, which consequently gives a significant reduction in the affinity of binding to colistin (11,17,18).
Very recently, a novel colistin resistance gene, mcr-2, was elucidated by Xavier et al. (19), which exhibits about 76.7% nucleotide (81% amino acid) identity to the paradigm gene, mcr-1. In light of the higher prevalence of mcr-2 in Belgium relative to mcr-1, introduction of surveillance of mcr-2 worldwide is immediately required. In principle, mcr-2 represents one more threat to public health which parallels that of the mcr-1 gene. However, we are not aware of the transfer, origin, and mechanism for MCR-2 colistin resistance thus far. Here we report that this is the case through integrative approaches ranging from bioinformatics, comparative genomics, bacterial genetics, structural biology, and molecular microbiology to phylogeny. Certainly, our findings might add new knowledge and close the missing gap in the field of colistin resistance and also be helpful for better monitoring (and control) of the epidemiological dissemination of MCR-2-mediated colistin resistance.

RESULTS
Contrasting transmission patterns of MCR-1 and MCR-2. Unlike the IncI2-type plasmid pHNSHP45, which is the first identified plasmid carrying the mcr-1 colistin resistance gene (GenBank accession no. KP347127), the IncX4 family plasmid might represent the most prevalent version of the mcr-1-bearing plasmid reservoirs (20,21). As a novel variant of the mcr-1 gene, the mcr-2 gene was first reported in Belgium by Xavier and colleagues in July 2016 (19). Notably, it seems likely that the plasmid pKP37-BE (GenBank accession no. LT598652.1) from porcine and bovine E. coli is the only known mcr-2-producing plasmid thus far and also can be assigned to the IncX4 replicon type (19). The different adjacent sequences of the mcr-2 gene from those of mcr-1 suggested the possibility of varied patterns/modes for the integration, acquisition, or dissemination of the two colistin resistance genes (Fig. 1). To address this hypothesis, using blastn, we performed comparative genomics of the mcr-2-positive plasmid pKP37-BE following a search for all of the IncX4 plasmids deposited in the GenBank database (Fig. 1A) (22), is supposed to be an ancestor of the two mcr-1/mcr-2-bearing plasmids described above (Fig. 1B). There is awareness that the transposition of antibiotic resistance genes is a widespread genetic event catalyzed by specific mobile elements (either an insertion sequence [IS] or integron). As expected, insertion sequence ISApl1, originally found in Actinobacillus pleuropneumoniae, is located upstream of the mcr-1 gene in the IncI2type mcr-1-harbouring plasmid pHNSHP45 (12,13). However, the ISApl1 element is consistently absent in front of the mcr-1 gene on the most IncX4 plasmids (23,24), implying diversified conjugation-aided mechanisms for the transfer of the plasmidborne mcr-1 gene. In agreement with most of the mcr-1-harbouring IncX4 plasmids, the plasmid pESTMCR is also featured, with 2,600 bp of mcr-1 cassette whose 3= end consistently is a hypothetical phosphoesterase-encoding gene (hp) (25) (Fig. 1B). The relic of a transposition event in which a proposed inverted repeat right sequence (IRR2 [TTTTTAAGAAGGGTGAACAAGTTTAAT]) of ISApl1 remains at the 3= end of the hp gene described above (Fig. 1C) allowed us to believe that ISApl1 might be involved in the transposition of mcr-1 and then lost. In contrast, two predicted IS1595-like insertion sequences ( Fig. 1A and B) surround the mcr-2 gene in the IncX4 plasmid pKP37-BE, despite the lack of the insertion sequence ISApl1 (19). In brief, the IS1595-like element (714 bp) carries a transposase gene (654 bp) flanked by two inverted repeats (18 bp each) (Fig. 1B). Also, the transposase-encoding gene is similar to a fragment (bp 1531602 to 1532255) from Moraxella bovoculi strain 58069 (GenBank accession no. CP011374) with 75% identity and 100% query coverage, posing the possibility that it might be the origin of mcr-2 (19). In addition to the duplicated target sites (8 bp each), 176 bp of spacer was also proposed to be present between the two probable kinds of insertion events (Fig. 1B), implying a possible hot integration site/region for the IncX4 plasmids (26). This speculation will require further experimental evidence in the near future.
It seems likely that unlike the ISApl1-mediated transposition of mcr-1, a unique IS1595-containing composite transposon enables the acquisition/transfer of the mcr-2 gene (Fig. 1B). Moreover, the mcr-2 colistin resistance gene with two identical IS elements might be enabled to form a circularized intermediate by homologous recombination of the IS elements (27). To test this hypothesis, we synthesized a cassette in which the mcr-2 gene is flanked by IS1595 on both sides ( Fig. 2A) and cloned into the vector pUC57, giving the plasmid pUC57-mcr-2 ( Fig. 2B). Following introduction into the recipient E. coli, the mcr-2-containing intermediate might be generated (Fig. 2C). As anticipated, two types of specific amplicons were produced in the reverse PCR assays ( Fig. 2D and E), which were verified by direct DNA sequencing. The reverse PCR results demonstrated the presence of mcr-2-containing circularized intermediates, implying the IS1595-like element was involved in the mobilization of mcr-2.
Therefore, we anticipated that the appearance of a circularized intermediate might accelerate the fast spread/dissemination of MCR-2 colistin resistance among diversified bacterial hosts. Together, we proposed that the transposal mechanism for the acquisition/transfer of the mcr-2-containing antibiotic cassette is far different from that of mcr-1 in the IncX4 plasmid (Fig. 1B).
Phylogeny of MCR-2. A blastp search (with a cutoff of Ͼ30% identity) with MCR-2 and/or MCR-1 as the primer sequence returned a collection of related homologues with the annotation of either PEA transferases or sulfatases (28). As a result, we retrieved 34 protein sequence candidates. The phylogeny of the MCR-2 protein constructed by MEGA7 (29) presented clearly two distinct groups: one group denotes a family of PEA transferases containing MCR-2, MCR-1, and Neisseria LptA (Fig. 3), whereas the other group comprises a series of putative sulfatases (Fig. 3). Of particular note, it seems likely that two subclades (I and II) can be assigned to the family of PEA transferases. Subclade I was featured with the plasmid-borne MCR-2 (and/or MCR-1) and the chromosomally encoded Paenibacillus PEA transferase (30-32) (Fig. 3). In contrast, subclade II was associated with the chromosomally encoded Neisseria LptA (Fig. 3). Obviously, this might suggest divergence in the evolutional patterns of the LptA/MCR-2 integral membrane proteins. In particular, MCR-2 and LptA can be functionally equivalent partially (if not all), despite the fact they are grouped into two distinct subclades (Fig. 3). However, we cannot exactly access the ancestor for the MCR-2 protein right now because that experimental evidence is not sufficient.
Domain analyses of MCR-2, a 538-amino-acid (aa) polypeptide, defines it as consisting of a transmembrane region connected to a sulfatase domain (see Fig. S2A in the supplemental material). As a core domain having the ability to hydrolyze a sulfate group, sulfatase is distributed extensively in three domains of life (33,34). Because deletion of the transmembrane domain (ΔTM2) impairs fully the phenotype in the colistin resistance conferred by MCR-2 (see Fig. 5), we thus hypothesize that the acquisition of a transmembrane domain enables the PEA transferase to localize correctly in the periplasm (Fig. 4) and then modify possible substrates like lipid A. Finally, the situation of MCR-2 in the phylogenetic tree allowed us to believe that a potentially parallel evolutionary path can be assigned to the two cousins (MCR-2 and LptA) under some environmental selection pressures (especially the extensive use of colistin as a veterinary/clinical medicine) (Fig. S2).
The periplasmic location of MCR-2 is a prerequisite for its activity. The synthesis of LPS-lipid A begins with UDP-GlcNAc as the primer substrate in the cytoplasm of E. coli (Fig. 4A). Subsequently, type II fatty acid synthesis (FAS II) provides the intermediates of fatty acid (C 12 to C 14 ) that enter the lipid A synthesis pathway involving nine enzymes (namely, LpxA, LpxC, LpxD, LpxH, LpxB, LpxK, LpxL, LpxM, and KdtA in Fig. 4A). Once being flipped from the cytoplasm by the ABC transporter MsbA into the periplasm (35), the nascent lipid A can be modified by the membrane-anchoring enzyme MCR-2, giving the final product PEA-4=-lipid A ( Fig. 4A and B). The replacement of lipid A with the PEA-4=-lipid A exported onto the outer leaflet of bacterial outer Mechanism of MCR-2 Colistin Resistance ® membrane significantly reduces the negative membrane charge (36), which consequently results in the decreased affinity of the surface in binding to colistin, the cationic polypeptide antibiotic (i.e., colistin resistance) (Fig. 4A). Therefore, it is reasonable that the subcellular location of the MCR-2 (and/or MCR-1) protein in bacterial periplasm is a prerequisite for its catalytic activity and colistin resistance. However, this requires further experimental evidence. Also, it is of much interest to compare the transcription profiles between mcr-2 and mcr-1. Thus, we applied LacZ-based transcriptional analyses to the mcr-2 gene. Unexpectedly, we noted that the mcr-2 promoter is appreciably stronger than that of the mcr-1 gene (Fig. 5A).
Bioinformatic analyses indicated that MCR-2 is an integral membrane protein with N-terminal 5 transmembrane helices (Fig. S2A). Using the arabinose-inducible E. coli expression system pBAD24/MG1655, we probed the in vivo role of the transmembrane  (53). The bootstrap consensus tree inferred from 1,000 replicates (54) is taken to represent the evolutionary history of the taxa analyzed (54). Branches corresponding to partitions reproduced in less than 50% of bootstrap replicates are collapsed. The percentages of replicate trees in which the associated taxa are clustered in the bootstrap test (1,000 replicates) are shown next to the branches (54). Initial trees for the heuristic search were obtained automatically by applying neighbor-joining (NJ) and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model and then selecting the topology with superior log likelihood value. A discrete gamma distribution was used to model evolutionary rate differences among sites: 5 categories (ϩG), parameter ϭ 1.5280. The rate variation model allowed for some sites to be evolutionarily invariable (ϩI; 10.3586% of sites). The analysis involved 34 amino acid sequences. All positions with less than 95% site coverage were eliminated. That is, fewer than 5% alignment gaps, missing data, and ambiguous bases were allowed at any position. There were a total of 511 positions in the final data set. Evolutionary analyses were conducted in MEGA7 (29). It seems likely that the phylogeny carries two groups: sulfatases (in gray) and phosphoethanolamine transferases (PEA transferases). The latter consisted of two subclades: subclade I (in yellow), including MCR-2 (in red) and the PEA transferase of Paenibacillus sophorae (in blue), and subclade II (in orange) with the Neisseria gonorrhoeae LptA (in purple) included.  (Fig. 5B). Second, the method of domain swapping was applied to generate two chimeric versions of the colistin resistance gene (namely, TM1-mcr-2 and TM2-mcr-1). "TM1-mcr-2" denotes the altered mcr-2 gene where the transmembrane region TM2 (bp 1 to 534) is replaced with the counterpart TM1 (bp 1 to 540) of the mcr-1 gene. In contrast, "TM2-mcr-1" represents the chimeric form of mcr-1 whose extracellular transferase domain is fused to the 3= end of the transmembrane region TM2 (bp 1 to 534) of the mcr-2 gene (Fig. 6A).
In our assays, the negative-control strain (MG1655 with/without the pBAD24 vector) can only grow on the Luria-Bertani agar (LBA) plate with less than 1 to 2 mg/liter of colistin, whereas the positive-control strain MG1655 with the pBAD24-borne expression of the mcr-2 in the wild type, exhibits significant growth on LBA medium with 16 mg/liter of colistin (Fig. 5B). Despite the fact that the growth of E. coli expressing both mcr-1 and mcr-2 appears on the LBA plates with colistin at 16 mg/liter (Fig. 5B), MCR-2 can confer to the recipient E. coli strain growth phenotype that is quite a bit better (see Fig. S3 in the supplemental material). Unlike the scenario seen with the   (Table S1) were sampled. Data are collected from three independent experiments and expressed as average Ϯ standard deviation (SD). The data suggested that the transcriptional level of the mcr-2 promoter is appreciably higher than that of the mcr-1. (B) Removal of the transmembrane regions (TM1 of MCR-1 and TM2 of MCR-2) impairs its role in colistin resistance by MCR-1 and/or MCR-2. The two wild-type (WT) strains denote the E. coli MG1655 strain carrying either pBAD24::mcr-1 or pBAD24::mcr-2. The ΔTM mutant includes ΔTM1 (the mcr-1 mutant with deletion of N-terminal transmembrane region from bp 1 to 540) and ΔTM2 (the mcr-2 mutant with deletion of the N-terminal transmembrane region from bp 1 to 534), respectively. The method culture dilution on LBA plates was used, and a representative result from three independent experiments is given. An asterisk indicates the growth of the recipient E. coli strain that express the mcr-2 gene is appreciably better than that of the counterpart expressing the mcr-1 gene under the condition with 16 mg/liter of colistin.

Sun et al.
® positive control carrying the wild-type mcr-2 gene, we noticed that the E. coli strain with the expression of the mcr-2 mutant lacking its transmembrane region (ΔTM2) consistently fail to grow on LBA plates with over 1 to 2 mg/liter of colistin (Fig. 5B). As anticipated, the transmembrane domain-swapping experiments (TM1 and TM2) also proved the observations presented above. First, Western blotting illustrated that four mcr genes (mcr-1, mcr-2, and two chimeric genes [TM1-mcr-2 and TM2-mcr-1]) can be expressed at an appreciable level in E. coli (Fig. 6B). More importantly, functional assays of domain swapping between mcr-1 and mcr-2 showed that the presence of the resultant two chimeric genes (TM1-mcr-2 and TM2-mcr-1) can confer significant resistance of the colistin-susceptible strain MG1655 to up to 16 mg/liter of colistin (Fig. 6C).
To gain further metabolic evidence, we isolated LPS-lipid A for extensive analyses by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) (Fig. 7A). The MS results demonstrated that unlike the negative-control MG1655 with or without empty vector pBAD24 (Fig. 7B), the lipid A modification occurs consistently in the strains carrying any one of the following four MCR versions: MCR-1 (Fig. 7C), MCR-2 (Fig. 7D), TM1-MCR-2 (Fig. 7E), or TM2-MCR-1 (Fig. 7F). Also, the engineered strains with the transmembrane deletion mutations of either MCR-1 (ΔTM1) or MCR-2 (ΔTM2) consistently lack the unique alteration of lipid A ( Fig. 7G and H). Obviously, the above data represent solid evidence that the transmembrane regions TM2 and TM1 play a critical role in MCR-2-mediated colistin resistance. Therefore, we are allowed to believe that the catalytic activity of PEA transferase depends on the correct location of MCR-2 in bacterial periplasm (Fig. 4A).

Structure-guided functional mapping of MCR-2.
To the best of our knowledge, MCR-1 (12, 13) and its newly identified variant, MCR-2 (19), are the only two known plasmid-borne genes encoding the PEA transferases so far. However, limited information is available regarding to the genetic determinant and functional aspects of MCR-2 (and/or MCR-1). We aimed to address these unanswered questions. As expected, the two MCR homologues (MCR-1 and MCR-2) are pretty conserved in the extracellular domain of PEA transferase in comparison with its cousin, LptA (EptA) of Neisseria. Multiple sequence alignment suggests that a putative zinc-binding/catalytic motif comprising six conservative residues (E244, T283, H393, D463, H464, and H476) is present in MCR-2 (Fig. S2B). Of note, the locations of the critical residues in MCR-1 denote E246, T285, H395, D465, H466, and H478, respectively (Fig. S2B).
Using membrane protein expression technology, we attempted to prepare the four MCR proteins (MCR-1, MCR-2, TM1-MCR-2, and TM2-MCR-1 in Fig. S4A to D in the supplemental material) and purified them to homogeneity (Fig. S4A to D). Size exclusion analyses suggested that they both can form in a monomer in vitro (Fig. S4A to D). Besides the confirmation by Western blotting using an anti-6ϫHis primary antibody (Fig. 6B), MS-based determination validated the molecular mass and its identity for the recombinant MCR-2 transmembrane protein (see Fig. S5A and B in the supplemental material). To gain further structural insight into the catalytic mechanism of MCR-2 (i.e., the addition of PEA to 4=-phosphate of lipid A from phosphatidylethanolamine), structural modeling by the Swiss-Model program was performed using Neisseria LptA (PDB accession no. 4KAV) (18) as a structural template (Fig. 8). The ribbon structures of MCR-2 (Fig. 8B) together with MCR-1 (Fig. 8A) were generated with the PyMol software. Intriguingly, structural comparison of the two MCR proteins illustrated that the putative 6 conserved residues (Fig. S2B) do constitute almost identical motifs ( Fig. 8A and B). Thus, we are extremely interested in examining their physiological role in the MCR-2 catalytic activity-dependent colistin resistance phenotype.
Driven by the above structural and bioinformatics speculation, we applied site-directed PCR mutagenesis to generate the following 6 point mutations of mcr-2 (E244A, T283A, H393A, D463A, H464A, and H476A). Similarly, six point mutations of mcr-1 were also used here. In the functional assays, the expression of all the versions of MCR-2 (and/or MCR-1) was triggered by the arabinose-inducible promoter of the pBAD24 expression vector in E. coli MG1655. As expected, strain MG1655 with or without vector alone (negative control) cannot show any growth when colistin is supplemented at over 2 mg/liter, whereas the positive-control strain MG1655 with expression of wild-type MCR-2 (or MCR-1) can grow well under conditions with up to 16 mg/liter colistin (Fig. 8C). In contrast to the positive control, none of the six mcr-2 (or mcr-1) point mutations can confer robust growth to the recipient strain MG1655 under conditions with over 2 mg/liter of colistin, which is similar to the finding with the negative control (Fig. 8C). The data constitute in vivo evidence that the six-residuecontaining motif plays critical roles in the maintenance of the biochemical mechanism of MCR-2 and its phenotype of colistin resistance.

DISCUSSION
In the Gram-negative bacteria, modifications of the LPS-lipid A moiety might reduce the net negative charge of the bacterial outer membrane, which consequently impairs the attachment of cationic antimicrobial peptide (CAMP), like polymyxins ( Fig. 1B;  Fig. S1A), to the bacterial surface (37,38). A total of three types of chemical modifica-  (Fig. 4). The bacterial LPSs were isolated as described above.
tions of the lipid A have been identified to account for bacterial resistance to colistin (37,38). They included (i) addition of phosphoethanolamine to 4=-phosphate position of sugar originally found in K. pneumoniae (11), (ii) modification of sugar with aminoarabinose at the 4=-phosphate position in Cupriavidus metallidurans (38), and (iii) glycine (and/or diglycine) modification at 3=-linked secondary acyl chain of lipid A in Vibrio cholerae El Tor (37). A single enzyme, ArnT, modifies the lipid A with aminoarabinose in the periplasm of C. metallidurans (38), whereas, a three-protein system is required for the modification of the lipid A with glycine (and/or diglycine) in the cytoplasm of V. cholerae El Tor, comprising (i) glycine carrier protein AlmF (Vc1578), (ii) amino acid ligase AlmE (Vc1579), and (iii) glycine transferase AlmG (Vc1577) (37). Given the fact that the three genes (almG-amlF-almE) are acquired by gene horizontal transfer in V. cholerae El Tor, comprising the operon Vc1577-Vc1578-Vc1579 (37), it is reasonable to ask whether it can be transferred into closely related enterobacteria like E. coli and Salmonella and in turn promote extensive resistance to polymyxin.
MCR-1 represents the unique plasmid-borne machinery that also allows the modification of the lipid A with phosphoethanolamine at the 4=-phosphate position, thereafter conferring efficient resistance to colistin (12,13). Unlike the chromosome-encoded  (18) as a structural template and were generated via the PyMol software. In addition to the five essential residues (E246, T285, H395, D465, and H466) we recently determined, one more key residue, H478, has been assigned to the MCR-1 function. In the MCR-2 protein, the critical motif included the following six residues: E244, T283, H393, D463, H464, and H476. (C) Site-directed mutagenesis-based assay for the functional motif of MCR-2. The mcr-2 point mutants were assayed using LBA plates with colistin in a series of dilutions. A representative result from our trials is shown. The negative control denotes MG1655 alone (or with empty vector). Vec, vector pBAD24.

Sun et al.
® enzyme ArnT which features 11 N-terminal transmembrane regions (38), the MCR-1 protein is an integral membrane protein with five periplasmic transmembrane helices (12,13). In light of the lack of significant homology (27.1% similarity and 12.7% identity) between the two enzymes (ArnT and MCR-1), we are very interested in examining in the near future whether or not they are functionally replaced in E. coli. In light of the transferability of colistin resistance by the MCR-1, we are not surprised by its fast sweeping across nearly the entire world in less than 1 year since its first discovery late in 2015 (17). The results from our group together with those from other groups worldwide suggest that MCR-1 is a very conservative protein that has undergone little selective pressure.
Very recently, Xavier and colleagues reported a novel plasmid-borne gene, mcr-2, that also confers resistance to colistin (19), although it seems unusual that the mcr-2 gene is detected only in Belgium and nowhere else worldwide. In fact, we and our collaborators in China attempted to perform a PCR screen for mcr-2 using almost 10,000 animal/clinical samples. It seems likely that the absence of mcr-2 in China might verify its low prevalence (39). This posed a hypothesis that might be due to a mechanism for mcr-2 dissemination different from that of the paradigm mcr-1 gene. In this report, we illustrated a working model for the transfer of the mcr-2-containing cassette by a unique transposon-like event (Fig. 1). The phylogeny suggested that MCR-2 and MCR-1 might share similar evolutionary history for functional acquisition of colistin resistance (Fig. 3). Consistent with that of MCR-1, the correct location of MCR-2 in periplasm ensures its enzymatic activity (Fig. 4). In addition to the five essential residues (E246, T285, H395, D465, and H466) for MCR-1 function we determined very recently (40), we demonstrated the requirement for one more residue, H478. Equivalently, we here have defined the genetic requirement of a substrate-binding/catalytic domain containing six residues (E244, T283, H393, D463, H464, and H476) for MCR-2 colistin resistance function (Fig. 6). In particular, during the review of our submitted manuscript, three reports added the X-ray structure of the catalytic domain of the MCR-1 and also added a similar structural clue to a full set of functional motifs, but it lacked full experimental verification for either MCR-1 or MCR-2 (41)(42)(43).
In summary, our findings establish mechanisms for the transfer, origin, and function of MCR-2 in transferable colistin resistance with relevance to lipid A modification and extend our understanding of a membrane charge-based remodeling strategy used for acquisition of bacterial antibiotic resistance, which might provide clues to the design of compounds targeted at reversing MCR-2/MCR-1 colistin resistance.

MATERIALS AND METHODS
Strains and growth conditions. All of the bacterial strains used here are E. coli K-12 derivatives (see Table S1 in the supplemental material). Luria-Bertani (LB) liquid medium and LB agar (LBA) plates were used for the maintenance of E. coli strains. DH5␣ was a recipient strain for cloning of the mcr-2 gene and its promoter, whereas strain BL21(DE3) was a protein expression host (13). DH5␣(-pir) was used to propagate pAH125 and its derivatives, and MC4100 (and/or MC1061), a lacZ mutant strain, was subjected to producing the mcr-2 promoter-driven lacZ transcriptional fusion (pAH125-Pmcr-2) integrated on the chromosome. The colistin-susceptible strain MG1655 was utilized to assay the phenotype of MCR-2 (and/or its mutants) in resistance to colistin. All of the strains were maintained at either 37 or 30°C.
De novo synthesis of the mcr-2 gene and its promoter. As we described earlier (44), with little change, overlapping PCR was adopted to synthesize the mcr-2 gene at full length. Briefly, 35 pieces of DNA primers (see Table S2 in the supplemental material) were designed to amplify five overlapped DNA fragments that fully cover the mcr-2 gene. Subsequently, the mixture of the above five PCR products diluted appropriately was subjected as the templates to the second round of overlapping PCR with a pair of specific primers (MCR-2-F and MCR-2-R [ Table S2]). As a result, the mcr-2 PCR product of the expected size was directionally cloned into the arabinose-inducible pBAD24 via the two restriction enzymes EcoRI and SalI, giving the recombinant plasmid pBAD24::mcr-2 (Table S1). Similarly, the mcr-2 promoter region (~300 bp) was also synthesized in vitro by overlapping PCR (see Tables S3 and S4 in the supplemental material) and inserted directionally into the promoter-less pAH125 with SalI and EcoRI, giving the plasmid pAH125-Pmcr-2. All of the acquired plasmids were verified by PCR detection (Table S3) and Sanger sequencing, and the point mutations detected in the mcr-2 gene were recovered through site-directed mutagenesis.
Assay for colistin resistance. As recommended by the Clinical and Laboratory Standards Institute guidelines (CLSI M100-S25), the agar dilution method was applied in colistin susceptibility experiments (47), which involves the breakpoint of EUCAST (European Committee on Antimicrobial Susceptibility Testing, 2015) for colistin resistance. The strain expressing MCR-1 refers to the positive control, whereas the MG1655 strain with/without plasmid pBAD24 alone denotes the negative control. The log-phase cultures of all the other strains carrying MCR-2 or its derivatives (Table S1) were diluted appropriately and spotted on LBA plates with colistin at various levels (0, 0.5, 1.0, 2.0, 4.0, 8.0, 16.0, and 32.0 mg/liter) (46). Of note, 0.2% arabinose was supplemented to trigger expression of MCR-2 (and/or MCR-1) (13).
Lipid A isolation and MALDI-TOF MS. The LPS-lipid A was routinely isolated by the method of hot phenol-water with minor change (18,48). The E. coli samples were collected through stripping of colonies grown on the LBA plates with/without the addition of colistin (8 mg/liter) overnight. The resultant crude LPS was resuspended in a solution of 30 mM Tris-HCl (pH 8.0) and 0.2% SDS and subjected to (i) digestion by DNase I (25 g/ml) and RNase A (100 g/ml) at 37°C for 2 h followed by (ii) 1 h of incubation with proteinase K at 56°C for 1 h. After other routine purification process, the acquired LPS was loaded for 12% PAGE coupled with silver staining to judge its purity and sent out for MALDI-TOF MS-based identification as described by Hankins et al. (49).
␤-Galactosidase assays. The two E. coli strains that carry the transcriptional fusion of either Pmcr-2-lacZ or Pmcr-1-lacZ (Table S1) were prepared for the measurement of LacZ activity. Mid-log-phase cultures grown in LB medium were directly tested after bacterial lysis with SDS-chloroform (50,51). The data were recorded in triplicate from three independent experiments.
Bioinformatics. To probe its topological structure, the MCR-2 protein sequence was subjected to the TMHMM server v2.0 (http://www.cbs.dtu.dk/services/TMHMM/). The protein sequences of MCR-2 and MCR-1 were aligned using the program Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/), and the final output was given after proceeding with the program ESPript 2.2 (http://espript.ibcp.fr/ESPript/ cgi-bin/ESPript.cgi) (52). A blastp search with MCR-2 (and/or MCR-1) as the primer sequences was carried out to collect the related homologues with the cutoff of above 30% identity. All returned hits were annotated with either PEA transferases or sulfatases. The phylogeny of the MCR-2 protein was constructed by MEGA7 (29). Structural modeling of MCR-2 (and MCR-1) was conducted with the Swiss-Model program, in which the Neisseria lipooligosaccharide phosphoethanolamine transferase A (LptA) acted as a structural template (PDB accession no. 4KAV) (18), and the ribbon structure was presented with the PyMol software.