Sequence variation in the active site of MCR proteins is evolutionarily accommodated through inter-domain interactions

Abstract

Sequence variation among homologous proteins can shed light on their function and ancestry. In this study, we analyse variation at catalytic residues among MCR (mobile colistin resistance) proteins, which confer resistance to the last resort antibiotic, colistin, in gram-negative bacteria. We show that not all naturally occurring variants at a lipid A-binding residue, Ser₂₈₄, are tolerated in MCR-1. In particular, the substitution of Ser₂₈₄ with Asp, found naturally in MCR-5, resulted in diminished colistin resistance. Using phylogenetic analyses and structure predictions we trace back variation at this site among MCRs to their ancestors, i.e. EptA phosphoethanolamine transferases that are encoded by diverse bacterial genomes. Mutational studies and Alphafold-based structural modelling revealed that the functional importance of position 284 varies between two phylogenetically distant MCRs, i.e. MCR-1 and MCR-5. Despite a high degree of similarity among their catalytic domains, inter-domain interactions were not conserved between MCR-1 and MCR-5 due to their different ancestries, providing a mechanistic basis behind the different phenotypes of similar mutations at position 284. Our study thus uncovers subtle differences in the organisation of domains among MCR proteins that can lead to substantial differences in their catalytic properties and mutational tolerance.

Introduction

Antimicrobial resistance (AMR) is expected to cause close to 10 million deaths annually by the year 2050 [1]. Failure of routinely used antibiotics to treat life-threatening infections requires the maintenance of certain antibiotics as “reserve” or “last-resort” drugs. Colistin, also known as polymyxin E, is one such antimicrobial that is used for the treatment of multidrug-resistant (MDR) bacterial infections [2, 3]. However, the surge of colistin resistance among gram-negative pathogens [4] threatens to weaken our defences against them. Colistin resistance is commonly mediated by plasmid-encoded mcr (Mobilized Colistin Resistance) genes, which encode phosphoethanolamine (PEA) transferases. MCR genes are found in diverse plasmid incompatibility groups, host bacterial strains and environments [3], and can co-exist with several other antibiotic resistance genes [5]. Their wide-spread prevalence and association with MDR bacteria has led to a rise in mortality, particularly in the context of nosocomial and ICU-acquired infections [5].

Colistin is a cationic polypeptide antibiotic with a lipophilic fatty acyl side chain [6]. The polycationic ring of colistin binds to the anionic lipopolysaccharide (LPS) layer of the gram-negative outer membrane. This in turn results in competitive displacement of membrane-stabilizing Ca²⁺ and Mg²⁺ ions and insertion of colistin into the outer-membrane via its fatty acyl side chain. Consequent disruption of the outer membrane and loss of cellular contents results in bacterial cell death [2, 7]. MCR proteins localise to the inner membrane and mediate colistin resistance by modifying the phosphate groups of lipid A in LPS with PEA, thus preventing colistin binding [3]. MCR proteins are thought to have evolved from chromosomally encoded PEA transferases belonging to the EptA family [8]. To date, ten mcr family genes (mcr-1to mcr-10) have been discovered [3], though the MCR-1 protein remains the most extensively characterised [6, 9–12]. The C-terminal catalytic domain of MCRs is stabilised by binding to Zn²⁺ and is connected to the N-terminal alpha-helical transmembrane domain through a flexible linker [13, 14]. Full-length structures are unavailable for MCR enzymes, however the catalytic domains of MCR-1 and 2 have been crystallised [10, 15]. Based on these structures and mutational analyses, distinct binding sites involving two different sets of active site amino acid residues for the two substrates, i.e. PEA and lipid A have been identified in MCR-1 and a ping-pong reaction mechanism has been proposed [2–4, 6, 10–12, 14, 16–22].

Though MCRs 1-10 show substantial conservation in their sequences, catalytic properties and domain organisation, a few functional differences among these proteins exist. For instance, the level of colistin resistance conferred by these proteins can vary [13, 14, 23] and inhibitors of catalysis show different efficacies against members of this family [24]. Additionally, the impact of MCR expression on bacterial fitness also varies across these enzymes [23]. These observations suggest subtle differences between MCRs 1-10. In this study, we evaluate the phenotypic consequences of sequence variation at catalytically relevant sites among MCR family members. Natural sequence variation among homologous proteins can result in significant changes in phenotypic and catalytic properties [25–28]. This is already known to be true for certain within-family sequence variants of MCR-1 (Liang et al., 2023), though a rigorous inter-family study across MCRs 1-10 has not yet been performed. From our analyses, we identify a naturally varying site in the lipid A binding pocket of MCRs, that displayed differences in mutational phenotypes and tolerances across two MCR members, i.e. MCR-1 and 5. Using this catalytic site as a case study, our study shines light on the evolutionary relationships of MCR family enzymes and identifies subtle differences among these proteins that may be traced back to their ancestries.

Results

Sequence variation in the active site influences colistin resistance of MCR-1

To identify sequence variation within the catalytic domain of MCR family enzymes, we aligned the consensus amino acid sequences of MCRs 1-10 (Figure S1) and analyzed variation at substrate-binding and catalytic residues (Figure 1A). Most catalytic residues were conserved across MCRs, though sequence variation was observed at positions corresponding to Ser₂₈₄, Tyr₂₈₇, Asn₃₂₉, Ser₃₃₀, and Asn₄₈₂ in MCR-1 (Figure 1A). Of these, Ser₂₈₄, Tyr₂₈₇, and Asn₄₈₂ were particularly interesting, as they are present in a highly conserved region of the active site directly involved in lipid A binding [12] (Figure 1A, B). To test whether sequence variation at these three sites was functionally equivalent, we replaced them in MCR-1 with amino acids present in other MCR family enzymes. In addition to single mutations, we also generated the Tyr₂₈₇Thr/Asn₄₈₂Tyr double mutation as these variants co-occurred in certain MCR family sequences (Figure 1A). The catalytically inactive Thr₂₈₅Ala mutant [6] served as a negative control. Wild-type and mutant MCR-1 proteins were heterologously expressed in E. coli K-12 BW25113 and the Minimum Inhibitory Concentrations (MICs) of colistin were measured. As expected, wild-type MCR-1 increased the MIC of colistin by ~12-fold, while the Thr₂₈₅Ala mutant did not alter the MIC. Mutations at Tyr₂₈₇ and Asn₄₈₂ as well as their combination did not alter the colistin resistance conferred by MCR-1 to E. coli (Figure 1C). The Ser₂₈₄Ala mutation also did not alter colistin MIC. However, substitution of Ser₂₈₄ with Asp resulted in a ~4-fold lower MIC than wild-type MCR-1 (Figure 1C). Thus, variation at this position likely had functional consequences for the activity of MCR proteins.

Figure 1. Natural sequence variation at a catalytic site impacts colistin resistance of MCR-1.

A. Multiple sequence alignment of consensus sequences of all MCR family members. Sequence blocks containing catalytically important amino acids in the catalytic domain are shown and numbered according to MCR-1. Catalytically important residues are highlighted based on their known/predicted function. Positions that vary across MCRs are mentioned below the alignment. B. The catalytic domain of MCR-1 with substrate analogues, ethanolamine (PBD ID: 5yle), or Glucose (PDB ID: 5ylf) is shown as a cartoon. Regions in the protein are coloured according to secondary structure elements: beta-sheet - red, loop - purple, helix - blue. Residues Ser₂₈₄, Thr₂₈₅, Tyr₂₈₇, and Asn₄₈₂ that are relevant to this study are represented as sticks and coloured by element (C-green, H-white, O-red, N-blue). Ethanolamine (EtA) and Glucose (Glc) are shown as yellow sticks. The zinc atom bound to the active site is shown as a grey sphere. C. Colistin MIC of E. coli heterologously expressing wild-type MCR-1 or indicated mutants. Mean ± S.D. from three independent measurements is plotted. Statistical significance was tested by comparing with wild-type MCR-1. Asterisk (*) indicates P-value <0.000001.

Small uncharged residues are preferred at position 284 by MCR-1

Next, we generated a series of substitutions at Ser₂₈₄ of MCR-1 to change side-chain polarity, charge, and volume at this site. Mutation of Ser₂₈₄ to charged residues compromised colistin MIC regardless of whether the introduced charge was negative or positive (Figure 2A, B). Further, MIC of Lys and Glu substitutions approached that of the inactive Thr₂₈₅Ala mutant, indicating that a large side chain volume at position 284 was not tolerated by MCR-1 (Figure 2A, B). In agreement with this finding, colistin MIC of uncharged mutants at this site showed an inverse trend with side-chain volume (Figure 2B) [29]. Introduction of a large amino acid such as Trp completely inactivated the MCR-1 enzyme, whereas intermediate MIC values were observed for Ser₂₈₄Thr, Ser₂₈₄Leu, and S₂₈₄Asn mutations (Figure 2A, B). Thus, we concluded that a small and uncharged residue was preferred at position 284 of MCR-1. Curiously, Gly at this site inactivated MCR-1, indicating that the complete absence of a space-filling group was also detrimental (Figure 2A, B).

Figure 2. Residue 284 of MCR-1 modulates lipid A binding and catalysis.

A. Colistin MICs conferred by indicated mutants of MCR-1 to E. coli. Wild-type MCR-1 was used as a reference for statistical comparisons. Mean ± SD from three independent measurements is plotted. Asterisk (*) indicates P-value <0.000001. B. The influence of charge and side chain volume on colistin MIC. Individual points represent the indicated amino acid substitutions at position 284 of MCR-1. MIC values are represented on the colour scale shown next to the plot. Side-chain charge is plotted on the X-axis and side-chain volumes are plotted on the Y-axis. C. Contacts made by the side-chain of Ser₂₈₄ of MCR-1 with glucose (Glc) as visualized on the crystal structure of MCR-1 catalytic domain (PDB ID: 5YLF). Polar contacts are shown by dotted yellow lines and distances are shown in Angstrom units. Amino acid sidechains are coloured by element (red: oxygen, blue: nitrogen, green: carbon). Glucose is shown as sticks and coloured by element (red: oxygen, purple: carbon). D. SDS-EDTA sensitivity of E. coli expressing MCR-1 or indicated mutants. Representative data of three independent replicates are shown.

The crystal structure of the catalytic domain of MCR-1 showed that Ser₂₈₄ is located in the lipid A binding pocket and is within interaction distance (i.e. < 4 Å) of substrate analogs glucose and xylose (Figure 2C) [12, 16]. Therefore, we wondered whether the reduced colistin MIC of Ser₂₈₄Asp MCR-1 may be due to lower lipid A binding and/or catalysis. To test this hypothesis, we exploited MCR-1-induced hypersensitivity of E. coli to membrane destabilization by SDS and EDTA. Earlier work has established that this hypersensitivity is due to disruption of LPS homeostasis and depends on the binding of MCR-1 to lipid A, but not catalysis [30]. We used the Ala₂₈₆Gly mutant of MCR-1, which has impaired lipid A binding capacity [30], as a control for these experiments. As expected, wild-type MCR-1 but not the Ala₂₈₆Gly mutant, rendered E. coli hypersensitive to SDS/EDTA (Figure 2D). MCR-1 Ala₂₈₆Gly also did not confer colistin resistance (Figure 2A). The catalytically inactive PEA binding deficient Thr₂₈₅Ala mutant, on the other hand, induced SDS/EDTA hypersensitivity (Figure 2D) indicating that it continued to bind lipid A. Interestingly, both Ser₂₈₄Ala and Ser₂₈₄Asp mutants showed similar sensitivity to SDS/EDTA as wild-type MCR-1 (Figure 2D), demonstrating that neither of these mutations appreciably reduced lipid A binding. On the other hand, the inactive Ser₂₈₄Gly mutant did not confer SDS/EDTA hypersensitivity (Figure 2D). Based on these data, we concluded that while residue 284 of MCR-1 influenced binding to lipid A, the presence of Asp at this site did not compromise lipid A binding. Instead, Asp at position 284 may reduce colistin MIC by directly impeding catalysis.

Natural sequence variation in the catalytic pocket of MCR members reflects individual evolutionary ancestry

The MCR gene family is thought to have originated due to the mobilization of chromosomal eptA genes that encode Phosphoethanolamine Transferases [8]. To ascertain whether sequence variation at the catalytic site of MCR-s arose before or after their diversification from the EptA family, we aligned 1270 sequences belonging to the EptA_B_N domain obtained from the Pfam database [31] with 109 MCR protein sequences acquired from Genbank. The catalytic residue Thr₂₈₅ of MCR-1 was conserved in 96.5% of all EptA sequences reflecting a conserved reaction mechanism across this enzyme family. Interestingly, the position equivalent to Ser₂₈₄ varied greatly among EptA proteins (Figure 3A). Ala was the most common residue at this site and was seen in 37.6% of all EptA sequences (Figure 3A), followed by Ser in 26.7% of sequences (Figure 3A). On the other hand, Asp, though found naturally in MCR-5, was infrequent (1.96%) at this position (Figure 3A). Other charged, bulkier residues, such as Glu (0.05%) or Asn (0.07%) were also rare, while Gly (0.0007%) and Trp (0.0002%) were almost completely absent at this site (Figure 3A). These trends resonated with our observation that Ser and Ala at position 284 of MCR-1 were functionally equivalent and that small uncharged amino acids were most suitable for catalysis.

Figure 3. Ancestral origin of variation at position 284 in MCRs.

A. Amino acid frequencies at position 284 among 1270 EptA and 109 MCR sequences. B. Maximum Likelihood cladogram built using EptA and MCR protein sequences. The labels of EptA family sequences and the labels and branches corresponding to MCR family sequences are coloured according to amino acid at position 284 as shown in the key. Uncoloured labels indicate sequences in which a gap was observed at position 284 in the alignment. The positions of MCRs 1-10 are shown in the tree.

Next, we used the above sequence alignment to build a cladogram of EptA and MCR family proteins to determine their evolutionary relationships (Figure 3B). Sequences of different MCR family members did not cluster together in the EptA-MCR tree. Instead, they were distributed across 5 branches (Figure 3B), reinforcing that MCRs may have evolved from different EptA ancestors [13, 14, 18–22]. This branching pattern resonated strongly with earlier work that has identified non-mobilised ancestors of different MCR-s in bacteria such as Moraxella (for MCR-1) [19] and Shewanella (for MCR-4) species [22]. In the case of MCR-8, our analyses identified EptA from Lampropedia as the closest homolog, while earlier studies have proposed EptA from Kosakonia to be the ancestor [18]. However, both Kosakonia and Lampropedia EptA show similar sequence identity with MCR-8 (68.91% and 68.19%, respectively) indicating that either of the two enzymes could be considered the closest non-mobilised bacterial homolog based on protein sequence alone. Importantly, position 284 strongly reflected the evolutionary relationships between MCR-s and EptA-s. MCR proteins with Ser or Ala residues at position 284 branched with EptA enzymes containing the same residue at the equivalent site (Figure 3B). MCR-5, the only family of MCRs that harboured an Asp at position 284 was separated substantially from the other MCR families in the EptA tree. Further, MCR-5 branched with EptA proteins that also had an Asp at the equivalent position (Figure 3B). Similar trends were observed for positions 287 and 482 as well (Figure S2). Based on these results we concluded that sequence variation in the active site among MCR family members reflected their different ancestries rather than divergence after mobilization.

Aspartate has been evolutionarily accommodated at position 284 by inter-domain interactions in MCR-5

Given their different evolutionary origins, we asked whether the phenotypic consequences of variation at position 284 would differ across MCRs. To test this idea, we chose MCR-5, which naturally harbours an Asp at position 284 and branched differently from MCR-1 in the EptA tree. Wild-type MCR-5 conferred a similar level of colistin resistance to E. coli as MCR-1. Like in MCR-1, introducing a Gly or Trp residue at position 284 significantly compromised the MIC for colistin (Figure 4A). However, contrary to expectation, the substitution of Asp₂₈₄ with Ser or Ala in MCR-5 did not enhance colistin resistance (Figure 4A). Further, SDS/EDTA hypersensitivity was seen upon expression of MCR-5 as well but was unaffected by the substitution of Asp₂₈₄ with Ala, Ser, or Gly (Figure 4B). These observations showed that while position 284 was important for catalysis across MCRs, its role in lipid A binding was less significant in MCR-5 than in MCR-1, and Asp had been accommodated at this position in the former such that it allowed for maximum functionality.

Figure 4. Mutations at residue 284 have context-dependent phenotypes in MCRs.

A. Colistin MICs of E. coli heterologously expressing MCR-5 or its indicated mutants and MCR-1 or indicated single/double mutants. For statistical significance, MIC values of MCR-5 mutants were compared with wild-type MCR-5. Mean ± SD from five independent measurements is plotted. Asterisk (*) indicates P-value < 0.000001. B. SDS-EDTA sensitivity of E. coli expressing wild type or indicated mutants of MCR-5. E. coli expressing MCR-1 is shown for comparison. LA plates were used as growth controls. Data shown are representative of three replicates. C. Structural alignment of models of full-length MCR-1 with MCR-5 and EptA from Pigmentiphaga sp, i.e. the closest EptA homolog of MCR-5. Both reference and matched structures are coloured according to C-alpha RMSD values as shown. Structural elements absent in the reference structure but present in the matched structure are coloured in light blue, while those present in the reference but absent in the matched structure are coloured in light brown. D. Inter-domain contacts in MCR-1 and MCR-5 determined from the model of full proteins generated by Alphafold. Interface between the catalytic domain (magenta) and linker-transmembrane domain (cyan) are shown for both proteins. All residues at the interface are shown as lines. Residues within interaction distance (< 4 Å) are shown as sticks, connected by yellow dashes, and listed below the structures. Both proteins are represented as cartoons. Side-chain oxygens are coloured red and side-chain nitrogens are coloured blue.

Since Asp₂₈₄ co-occurred with Thr₂₈₇ and Tyr₄₈₂ in MCR-5 (Figure 1A), we argued that differences at these neighbouring sites within the catalytic pocket itself might be responsible for the different phenotypes of analogous mutations in MCR-1 and MCR-5. However, neither the Ser₂₈₄Asp/Tyr₂₈₇Thr nor the Ser₂₈₄Asp/Asn₄₈₂Tyr double mutation in MCR-1 restored colistin resistance (Figure 4A). These results hinted at a larger structural basis for the accommodation of Asp at this position. Lipid A binding by MCRs has recently been demonstrated to involve the linker and transmembrane domains in addition to the catalytic domain [32]. We, therefore, wondered if inter-domain interactions may play a role in the accommodation of Asp at position 284. Since crystal structures of full-length MCR-1 and MCR-5 are currently unavailable, we generated AI-based structural models for both proteins using Alphafold 3 [33] and compared them. Models of MCR-1 and MCR-5 showed high structural conservation in their catalytic domains (Figure 4C). However, their linker and transmembrane domains aligned less well (> 5 Å C_a RMSD) (Figure 4C). A lack of structural concordance was also seen between the linker and transmembrane domains of MCR-1 and the closest EptA homolog of MCR-5, i.e. Pigmetiphaga sp. EptA, though MCR-5 and Pigmetiphaga EptA aligned well over the entire protein (Figure 4C). Indeed, we observed a similar trend between MCR-1 and MCRs 3, 4, and 8, which have different ancestries (Figure S3). Thus, while catalytic domains were highly conserved across MCRs, transmembrane and linker regions varied and these differences could be traced to different ancestries.

Interestingly, the structural models revealed that Ser₂₈₄ may itself participate in inter-domain interactions by forming polar contacts with Asn₁₀₈ from the transmembrane domain in MCR-1. In MCR-5, Asp₂₈₄ interacted with the structurally analogous residue Asn₁₁₂ (Figure 4D). However, we found several other pairs of amino acids that could potentially maintain inter-domain interactions in MCR-5 that were missing in MCR-1. These included at least 3 possible salt bridges indicating strong electrostatic forces holding the catalytic and transmembrane domains together in MCR-5 (Figure 4D). On the other hand, in MCR-1 the only other residue pair that could potentially form inter-domain contacts was Gln₁₀₇-Asp₃₀₄ (Figure 4D). Since these conclusions were based on Alphafold models that can make erroneous predictions for side-chain orientations, we performed a similar analysis on models of EptA from Moraxella (homolog of MCR-1) and Pigmentiphaga (homology of MCR-5). Both the number and identity of interacting residues at the domain-interface were completely conserved in these two proteins (Figure S4), strongly supporting our conclusions for MCR-1 and MCR-5. This difference suggested that the overall structure of MCR-5 would be more tolerant to mutations at position 284 than MCR-1. In agreement with MIC and SDS/EDTA data (Figure 2), the predicted structures of Ser₂₈₄Ala and wild-type MCR-1 were identical (Figure 5). However, models of Ser₂₈₄Asp and Ser₂₈₄Gly showed a dramatic reorganization of the linker and transmembrane domains of the protein relative to the catalytic domain (Figure 5). As a result of this structural reorganization, the lipid A binding helix contributed by the linker (Pro₁₈₈ to Pro₁₉₅) [32] was displaced, explaining the phenotype of mutations at Ser₂₈₄ in MCR-1 (Figure 5). On the other hand, none of the mutations at position 284 of MCR-5 were predicted to induce these large structural changes (Figure 5), corroborating that MCR-5 showed greater mutational tolerance at this site than MCR-1. Thus, we concluded that ancestrally inherited structural accommodation of Asp at position 284 due to inter-domain interactions altered the consequences of mutations at this site among MCR enzymes.

Figure 5. Differential mutational tolerance at position 284 in MCR-1 and MCR-5.

Aligned structural models of MCR-1 with its Ser₂₈₄Ala, Ser₂₈₄Asp, and Ser₂₈₄Gly mutants generated by Alphafold are shown in the top three panels. Aligned structural models of MCR-5 with its Asp₂₈₄Ala, Asp₂₈₄Ser, and Asp₂₈₄Gly mutants are shown in the lower three panels. Structures of wild-type MCR-1 and MCR-5 are represented as grey cartoons. The structures of mutants are shown as orange ribbons. Lipid A binding residues in the catalytic site, i.e. residues 284, 287, and 482 in MCR-1 or their equivalent in MCR-5 are shown as sticks and coloured by element (carbon: magenta, oxygen: red, nitrogen: blue). The lipid A binding helix from the linker is highlighted in red in the wild-type and mutant structures. Displaced helix position is indicated (*). The lipid A binding pocket, constituted by catalytic residues as well as residues from the linker-transmembrane regions, is indicated with an arrow. Predicted Template Modelling (pTM) score for all models used in these analyses are provided. A score of >0.5 indicates high confidence in the model.

Discussion

Sequence variation among homologous proteins can have profound effects on their function. In this study, we explored sequence divergence at catalytic residues among different MCR enzymes to identify an active site variant that has context-specific effects on colistin resistance. The residue 284 in MCR-1, which was the focus of our study, is adjacent to a conserved Thr that binds PEA and is indispensable for catalysis. Despite being in the active site of the enzyme, we found substantial sequence variation at this site among MCRs, as well as in the larger EptA family. Our experiments showed that Ser₂₈₄ of MCR-1 contributed to both, catalysis and lipid A binding, and these two properties can be decoupled by mutations. MCR-1 Ser₂₈₄Asp, for instance, was able to bind to lipid A, but had a significantly lower colistin MIC. Further, mutation of this position to Gly or Trp in MCR-1 and MCR-5 almost completely inactivated both proteins, indicating that residue 284 is necessary for catalysis in both enzymes, however had different consequences for the capacity to bind lipid A. We explain the role of residue 284 in catalysis using the “loading-transferring” and the “ping-pong trade-off” models that have been proposed for MCR-1 [21, 32]. According to the ping-pong trade-off model, MCR-1 is initially linked to a PEA moiety which then is transferred to an incoming lipid A molecule. Similarly, the loading-transferring model proposes that the catalytic cycle of MCR-1 requires a binding step, in which lipid A is loaded onto a molecule of MCR-1 via the latter’s linker domain and then transferred to the catalytic site to be modified by an already recruited PEA molecule. Ser₂₈₄ is located between the binding sites for PEA and glucose in MCR-1 [12] and hence, we propose that a bulky or charged residue at position 284 may impair the transfer step of the catalytic cycle. Curiously, Gly was an exception to this trend, and inactivated both MCR-1 and MCR-5. We hypothesize that Gly at position 284 may induce large conformational changes or result in structural destabilisation that inactivates MCR-1 and MCR-5. Indeed, the near absence of Gly residues at the equivalent position in EptA proteins suggests a strong purifying selection against it, and the established role of Gly as an inducer of conformational dynamics may hinder the inter-domain contacts that we have found to be essential for MCR functioning.

A role for Ser₂₈₄ in directly interacting with lipid A has been proposed based on the co-crystal structure of the catalytic domain of MCR-1 and glucose/xylose [11, 12]. However, the neutral phenotype of an Ala substitution at this site in MCR-1 and MCR-5 indicates that direct interaction with lipid A may not be a major role of residue 284 in MCRs. Instead, structural modelling of the full-length MCR-1 from the present study showed that residue 284 may influence lipid A binding indirectly by participating in inter-domain interactions. Though residues for enzymatic activity are provided by the catalytic domain, a role for the linker and transmembrane domains of MCRs in substrate binding and catalysis is increasingly becoming evident. For example, deletion mutants of the linker and/or transmembrane domain inactivate MCRs-1, 2, and 3 [17, 32, 34]. Similarly, the inactivity of some natural MCR variants has been traced to mutations in linker and transmembrane regions [13, 14, 34], and a crucial helix in the linker that is necessary for lipid A binding has also been identified [32]. Taken together, the lipid A binding site in MCRs appears to be constituted by all three domains. A key finding from the present study was that residue 284 formed interactions with amino acids from the transmembrane domain and that inter-domain contacts varied between MCRs. Notably, MCR-5 had a greater number of interacting residues between domains compared to MCR-1, reducing its dependence on residue 284 to hold the protein structure together. This difference may explain several experimental observations: Firstly, the tolerance of MCR-5 for mutations at residue 284 was greater than MCR-1. As a result, Asp at this position allowed for maximum activity in MCR-5, but reduced activity in MCR-1. Additionally, Gly at position 284 disrupted the structure and abolished lipid A binding only in MCR-1. Secondly, attempts to generate chimaeras between catalytic and transmembrane domains across MCRs have been met with limited success [13, 14, 34]. Our attempts to generate functional chimaeras between MCR-1 and MCR-5 similarly failed (not shown). We discovered that the precise inter-domain amino acid contacts in MCR-1 and its closest relative Moraxella EptA (A0A1T0CMP3_9GAMM) were identical (Figure 4D, S4). Similarly, inter-domain contacts in MCR-5 and Pigmentiphaga EptA (A0A208Y8E5_9BURK) were also conserved (Figure 4D, S4). Thus, domain organisation and inter-domain contacts appears to be ancestrally inherited in MCRs and result in differences in how the three domains are coordinated in these enzymes. In further support of this idea, functional chimaeras have been successfully made between MCR-1 and MCR-2, which have a common ancestry based on our phylogenetic analyses [17].

Interestingly, though mcr family genes have arisen from EptA-like ancestors, the role of chromosomally encoded EptA enzymes in intrinsic resistance to colistin remains poorly understood. Unequivocal evidence is available for only a few members of this family, e.g. EptA from Neisseria meningitidis which can lead to colistin resistance when overexpressed [20]. Indeed, some EptA enzymes have other substrates altogether. For instance, an EptA homolog from Campylobacter jejuni modifies the flagellar rod protein FlgG with PEA and is essential for flagellar assembly and motility [35, 36]. Thus, the ability to confer high levels of colistin resistance may have evolved independently starting from different EptA ancestors and is reflected in the diversity of existing alleles of MCRs [3]. While the precise evolutionary trajectory remains to be elucidated, the discovery of non-mobile colistin resistance determinants such as the NMCR-3 gene in Aeromonas suggests that chromosomally encoded EptA enzymes may serve as a reservoir for the evolution of antibiotic resistance genes in Gram-negative bacteria [37]. Importantly, different ancestries result in subtle differences in catalytic properties and active site architecture, which may pose a challenge in the design of small molecule inhibitors of MCRs. For example, Hanpaibool et al. discovered that Pyrazolone, an MCR-1 enzyme inhibitor, is inactive against MCR-3 [24]. Similarly, Xie et al. found that Pogostone, though active against both MCR-1 and MCR-3, bound the active sites of the two enzymes differently, suggesting that differences in the active site architecture of MCRs need to be taken into account during inhibitor design [38]. It is also likely that the propensity to evolve resistance to MCR inhibitors may be determined by ancestrally inherited mutational tolerances warranting further investigations into the differences between MCR family proteins. Finally, our mutational and modelling data suggest that inhibitors that disrupt the inter-domain interface are likely to be effective against a broader suite of MCR proteins. Small molecules that bind to and prevent protein-protein interactions, for instance rapamycin, are known [39] and a similar approach may be worth exploring for disrupting the interaction between protein domains in MCRs. This strategy may also have the additional benefit of reducing the mutation tolerance of MCRs and decreasing the chance for resistance evolution.

Materials and Methods

Antibiotic solutions

Sterile solutions of colistin (Polymyxin E, HiMedia, India) and kanamycin (SRL, India) were prepared in Milli-Q filtered de-ionised water (Millipore, India). Stock concentrations of 10-30 mg/mL were prepared as needed and stored at -20 °C until required.

Bacterial strains, plasmids, and culture conditions

E. coli DH10B was used for DNA manipulation and cloning. E. coli K-12 BW25113 was used for expression and characterisation of MCR-1, MCR-5, and their mutants. Bacterial strains were cultured at 37 °C in Luria-Bertani Broth (LB) in a shaker incubator (150-200 rpm) or on Luria-Bertani Agar (LA) plates. Media was supplemented with kanamycin at a concentration of 30 μg/mL to select for plasmid-transformed E. coli as required. A list of all plasmids used in this study is provided in Table 1. The pGDP2 MCR-1 plasmid (Addgene plasmid #118404), containing the colistin-resistance encoding mcr-1 gene in a pGDP2 vector, was a gift from Prof. Gerard Wright [40]. The mcr-5 gene was chemically synthesized and cloned into a pEX-A258 vector clone (Eurofins Genomics, India). This plasmid was used to generate the pGDP2 MCR-5 plasmid using restriction-free cloning, as previously described [41].

Table 1. List of plasmids used in this study.

Plasmid Name	Description	Source
pGDP2 MCR-1	mcr-1 gene cloned into the low copy number	Addgene plasmid # 118404
	vector pGDP2	[40]
pGDP2 MCR-5	mcr-5.2 gene cloned into the low copy number vector pGDP2	This study
pEX-A258- mcr5.2	mcr-5.2 gene cloned into the cloning vector pEX-A258-mcr5.2	This study
pGDP2 MCR-1 T285A	pGDP2 MCR-1 containing a Thr285Ala point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284A	pGDP2 MCR-1 containing a Ser284Ala point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284D	pGDP2 MCR-1 containing a Ser284Asp point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 Y287V	pGDP2 MCR-1 containing a Tyr287Val point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 Y287T	pGDP2 MCR-1 containing a Tyr287Thr point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 N482Y	pGDP2 MCR-1 containing an Asn482Tyr point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 Y287TN482Y	pGDP2 MCR-1 containing Tyr287Thr and Asn482Tyr mutations in the mcr-1 gene	This study
pGDP2 MCR-1 Y287VN482Y	pGDP2 MCR-1 containing Tyr287Val and Asn482Tyr mutations in the mcr-1 gene	This study
pGDP2 MCR-1 S284E	pGDP2 MCR-1 containing a Ser284Glu point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284N	pGDP2 MCR-1 containing a Ser284Asn point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284K	pGDP2 MCR-1 containing a Ser284Lys point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284L	pGDP2 MCR-1 containing a Ser284Leu point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284T	pGDP2 MCR-1 containing a Ser284Thr point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284G	pGDP2 MCR-1 containing a Ser284Gly point mutation in the mcr-1 gene	This study
pGDP2 MCR-1 S284W	pGDP2 MCR-1 containing a Ser284Trp point mutation in the mcr-1 gene	This study
pGDP2 MCR-1	pGDP2 MCR-1 containing Ser284Asp and	This study
S284DY287T	Tyr287Thr mutations in the mcr-1 gene
pGDP2 MCR-1 S284DN482Y	pGDP2 MCR-1 containing Ser284Asp and Asn482Tyr mutations in the mcr-1 gene	This study
pGDP2 MCR-1 A286G	pGDP2 MCR-1 containing an Ala286Gly point mutation in the mcr-1 gene	This study
pGDP2 MCR-5 D284A	pGDP2 MCR-5 containing a Asp284Ala point mutation in the mcr-5 gene	This study
pGDP2 MCR-5 D284S	pGDP2 MCR-5 containing a Asp284Ser point mutation in the mcr-5 gene	This study
pGDP2 MCR-5 D284W	pGDP2 MCR-5 containing a Asp284Trp point mutation in the mcr-5 gene	This study
pGDP2 MCR-5 D284G	pGDP2 MCR-5 containing a Asp284Gly point mutation in the mcr-5 gene	This study

MCR consensus sequences and alignment

Reference amino acid sequences for each MCR family member (MCR families 1-10) were obtained from the NCBI National Database of Antibiotic Resistant Organisms (NDARO) (https://www.ncbi.nlm.nih.gov/pathogens/antimicrobial-resistance/). To avoid sampling bias, all full-length available amino acid reference sequences within each MCR family member were aligned by multiple sequence alignment using Clustalw with default parameters in MEGA (version 10.2.6) [42], and a consensus reference sequence for each MCR family member was then generated using EMBOSS Cons [43]. Full-length consensus sequences were re-aligned using MEGA (10.2.6) [42] and this alignment was used to identify variant residues at catalytic sites across MCR family members.

Alignment and cladogram for EptA proteins

All available full-length Pfam EptA/B domain containing sequences [31] and all available MCR family reference sequences obtained from the NCBI National Database of Antibiotic Resistant Organisms (NDARO) (https://www.ncbi.nlm.nih.gov/pathogens/antimicrobial-resistance/) were aligned in MEGA (10.2.6) [42]. Supplementary File 1 contains the accession numbers of all sequences used for this analysis. Linux command line was employed to enumerate motifs of interest from the generated EptA-MCR alignment. This alignment was used to construct a phylogenetic tree using the ‘Maximum Likelihood’ method on MEGA (10.2.6) [42]. The iTOL (EMBL) phylogenetic tree viewer [44] was used to visualize the resulting cladogram.

Structural Prediction and Analyses

All protein structures were visualised on UCSF Chimera (version 1.17.3) [45] and Pymol (version 2.5.2). Structural models of full-length MCR-1, MCR-5 or their mutants were generated on the AlphaFold 3 server [33] using default parameters. Structural models of the full-length EptA and other MCR family of proteins were obtained from the AlphaFold Structure Database [33]. Structural alignments were performed on UCSF Chimera (version 1.17.3) or Pymol (version 2.5.2) using default parameters. Structural alignments were coloured according to average Cα-RMSD values on Chimera using the following colour scale thresholds: white ≥12 Å, red 5-12 Å, dark blue 0-5 Å.

Site-directed mutagenesis

All mutations in the mcr-1 and mcr-5 genes were made by PCR based site-directed mutagenesis using the protocol described by Shenoy and Visweswariah [46], with the slight modification of setting up 20 μL PCR reactions with both forward and reverse mutagenic primers. All mutagenic primers used are listed in Table 2. Mutants were confirmed by restriction digestion and Sanger sequencing (Barcode BioSciences, India).

Table 2. List of PCR primers.

Primer name	Primer sequence (5’ to 3’)
mcr1_T285A_fwd	CAATGTCACCAGCTGCGGCACATCGGCGGCGTATTCTGT
mcr1_T285A_rev	AATACGCCGCCGATGTGCCGCAGCTGGTGACATTGCTA
mcr1_S284A_fwd	CAATGTCACCAGCTGCGGCACAGCGACGGCGTATTCT
mcr1_S284A_rev	TACGCCGTCGCTGTGCCGCAGCTGGTGACATTGCTAA
mcr1_S284D_fwd	CAATGTCACCAGCTGCGGCACAGATACGGCGTATTCTG
mcr1_S284D_rev	TACGCCGTATCTGTGCCGCAGCTGGTGACATTGCTA
mcr1_Y287V_fwd	CGTGCGGCACGTCGACGGCGGTGTCTGTGCCGTGTATGTTC
mcr1_Y287V_rev	ACGGCACAGACACCGCCGTCGACGTGCCGCACGATGTGAC
mcr1_Y287T_fwd	CGTGCGGCACGTCGACGGCGACCTCTGTGCCGTGTATGTTC
mcr1_Y287T_rev	ACGGCACAGAGGTCGCCGTCGACGTGCCGCACGATGTGAC
mcr1_N482Y_fwd	TGGTATGCCATATGCCTTTGCACCAAAAGAACAG
mcr1_N482Y_rev	CAAAGGCATATGGCATACCATGTAGATAGACACCG
mcr1_S284E_fwd	GCAATGTCACCAGCTGCGGCACAGAGACGGCGTATTCTGTGCC
mcr1_S284E_rev	ACGCCGTCTCTGTGCCGCAGCTGGTGACATTGCTAAAATTG
mcr1_S284N_fwd	GCAATGTCACCAGCTGCGGCACAAATACGGCGTATTCTGTGC
mcr1_S284N_rev	TACGCCGTATTTGTGCCGCAGCTGGTGACATTGCTAAAATTG
mcr1_S284K_fwd	CAATGTCACCAGCTGCGGCACAAAGACGGCGTATTCTGTGC
mcr1_S284K_rev	ACGCCGTCTTTGTGCCGCAGCTGGTGACATTGCTAAAATTG
mcr1_S284L_fwd	GCAATGTCACCAGCTGCGGCACATTGACGGCGTATTCTGTGC
mcr1_S284L_rev	ACGCCGTCAATGTGCCGCAGCTGGTGACATTGCTAAAATTG
mcr1_S284T_fwd	GCAATGTCACCAGCTGCGGCACAACGACGGCGTATTCTGTG
mcr1_S284T_rev	CGCCGTCGTTGTGCCGCAGCTGGTGACATTGCTAAAATTG
mcr1_S284G_fwd	GCAATGTCACCAGCTGCGGCACAGGGACGGCGTATTCTG
mcr1_S284G_rev	GAATACGCCGTCCCTGTGCCGCAGCTGGTGACATTGCTAAAATTG
mcr1_S284W_fwd	GCAATGTCACCAGCTGCGGCACATGGACGGCGTATTCTG
mcr1_S284W_rev	GAATACGCCGTCCATGTGCCGCAGCTGGTGACATTGCTAAAATTG
mcr1_A286G_fwd	CGTGCGGCACGTCGACGGGGTATTCTGTGCCG
mcr1_A286G_rev	CACAGAATACCCCGTCGACGTGCCGCACGATGTGAC
mcr5_D284A_fwd	GCGGGACGGCTACGGCTACAAGCTTACCCTGCATGTTTTCC
mcr5_D284A_rev	CATGCAGGGTAAGCTTGTAGCCGTAGCCGTCCCGCAACTGGTG
mcr5_D284S_fwd	TTGCGGGACGTCGACGGCTACATCCC
mcr5_D284S_rev	GTAGCCGTCGACGTCCCGCAACTGGTGAC
mcr5_D284W_fwd	GCGGGACGTGGACGGCTACAAGCTTACCCTGCATGTTTTCCC
mcr5_D284W_rev	CATGCAGGGTAAGCTTGTAGCCGTCCACGTCCCGCAACTGGTG
mcr5_D284G_fwd	GCGGGACGGGTACGGCTACAAGCTTACCCTGCATGTTTTCCC
mcr5_D284G_rev	CATGCAGGGTAAGCTTGTAGCCGTACCCGTCCCGCAACTGGTG
pGDP2_mcr5_RF_F	GTTTAACTTTAAGAAGGAGATATACCATGCGGTTGTCTGCATTTA TCACTTTC
pGDP2_mcr5_RF_R	GGTGGTGGTGGTGGTGCTCGAGTCATTGTGGTTGTCCTTTTCTGC ATG

Colistin susceptibility measurement

Colistin resistance of all bacterial strains was measured using a broth-microdilution assay. The frozen stock of appropriate strain (5 μL) was inoculated in 1 mL LB broth and grown for 5.5 hours at 37°C with shaking (180 rpm). The resulting culture was diluted 1:10 in fresh LB. Diluted culture (15 μL) was inoculated into the wells of a sterile 96-well microtiter plate, with each well containing a two-fold serial dilution of colistin (highest concentration 100 μg/mL) in 150 μL of LB. Peripheral wells of the microtiter plate were filled with autoclaved Milli-Q filtered water. The microtiter plate was incubated at 37°C for 20 hours, after which optical density was measured in a plate reader (PerkinElmer, India). Absorbance readings were normalized to the no-antibiotic growth control. MIC was defined as the lowest concentration of colistin at which no bacterial growth was visible and O.D. values ceased to change significantly upon a further increase in colistin concentration.

SDS-EDTA sensitivity assays

Sensitivity to SDS-EDTA was measured using spot assays on LA plates, as previously described by Feng et al., 2022 [30]. Briefly, 5 μL of frozen stock of required strains was revived in 1 mL LB broth for 7 hours. The resulting cultures were serially diluted (10^-1 to 10^-6) in LB media. Diluted cultures (5 μL) were spotted onto LA plates supplemented with 0.25% SDS and 1 mM EDTA. LA plates were used as a growth control. The plates were incubated at 37°C for 20 hours and then imaged.