Involvement of hpap2 and dgkA Genes in Colistin Resistance Mediated by mcr Determinants

Plasmid-mediated colistin resistance (mcr) determinants are challenging the efficacy of polymyxins against Gram-negative pathogens. Among 10 mcr genes described so far, the major determinants mcr-1 and mcr-3 are found closely linked to hpap2 or dgkA genes, encoding a hypothetical phosphatidic acid phosphatase of type 2 (PAP2) and a diacylglycerol kinase, respectively, whose functions are still unknown. In this study, mcr-1, mcr-1–hpap2, mcr-3, and mcr-3–dgkA were expressed in Escherichia coli, and recombinant strains were analyzed to detect antimicrobial susceptibility and changes in the expression of genes involved in phospholipid metabolism. The mcr-1 or mcr-3 single genes were enough to drive growth on colistin selective media, although co-expression of linked genes conferred maximal antibiotic resistance. Expression of mcr determinants downregulated endogenous genes involved in lipopolysaccharide (LPS) modification or phospholipid recycling, although to different extents of repression: strong for arnB, ybjG, and pmrR; medium for eptA, lpxT, and dgkA; small for bacA and pgpB. Four of these genes (bacA, lpxT, pgpB, and ybjG) encode undecaprenyl pyrophosphate (UPP) phosphatases. In these conditions, cells presented resistance against bacitracin, an antibiotic that sequesters UPP from PAP2 enzymes. The hpap2 and dgkA genes might play a role in colistin resistance by compensating for phospholipid metabolism functions altered during LPS modification by colistin resistance determinants.

Interestingly, a recent study evidenced that, among 64628 genomes deposited in NCBI databases and screened for mcr sequences, three of the less commonly found genes lack hpap2or dgkA-linked sequences, like mcr-4 and mcr-5, or are clustered to a truncated hpap2 coding sequence in the even more rarely found mcr-2 gene [13]. Furthermore, mcr-9 and mcr-10, the two last and most closely related mcr genes identified so far, lacked both hpap2 or dgkA linked sequences and, although globally spread, had a weak role in colistin resistance [11,13,14]. Thus, mcr-1 and mcr-3 are the colistin resistance determinants more frequently mobilized, with plasmids carrying mcr-1-hpap2 or mcr-3-dgkA widely distributed among different species. The biological success of these mcr gene clusters suggests that expressed enzymes from hpap2 and dgkA genes could play a role connecting phospholipid metabolism and antimicrobial susceptibility to polymyxins.

Functionality of Plasmids Carrying mcr Genetic Elements
The mcr-1, mcr-1-hpap2, mcr-3, or mcr-3-dgkA sequence elements were amplified and cloned in pBAD24 (see details in Section 4; Table 1 and Figure 2A), a vector used previously to express, under control of the PBAD promoter, the transcriptional regulator AraC and l-arabinose induction, and the coding sequences of mcr-1 and mcr-3, among others [1]. Replication of plasmids carrying the two-gene constructs was as efficient as their relatives with single sequences (Figure 3), whereas E. coli strains carrying the different recombinant plasmids and induced by l-arabinose transcribed continuous messenger RNAs (mRNAs) from single-or two-gene constructs, indicating the co-transcription along mcr-1-hpap2 or mcr-3-dgkA genetic elements ( Figure 2B). 1 Underlined sequences indicate restrictions sites used for cloning in pBAD24 (EcoRI, SalI). 2 Amplicon lengths (bp). 3 Efficiency of the qPCR. 4 Correlation (R 2 ) of the qPCR calibration. 5 Calibration of the qPCR: slope. (B) Agarose gel electrophoresis of rtPCR, where reverse transcription was performed on RNA purified from the four recombinant strains after arabinose induction, corresponding to one of the biological replicates from every strain utilized in experiment shown in Figure 5; PCR amplification utilized the same primers used for cloning the four mcr gene constructs (Table 1).

Figure 3.
Functionality DNA constructs for mcr expression. Concentration (copy number) and messenger RNA (mRNA) expression from the different mcr gene constructs were determined by qPCR with primers mcr1q, mcr3q, and recAq (Table 1) on DNA or complementary DNA (cDNA) (see above) from the four recombinant clones. Firstly, mcr DNA concentration (black bars) and mcr gene expression (white bars) from all strains were relativized to recA DNA or recA mRNA, respectively; secondly, every two-gene construct was normalized with respect to its single mcr-gene reference.
Furthermore, plasmids carrying two-gene constructs performed similarly for controlling expression of recombinant DNA since mRNA abundance from mcr genes did not change following the co-transcription with downstream sequences (Figure 3). Thus, hpap2 and dgkA coding sequences are, like upstream mcr genes, efficiently transcribed from the PBAD promoter. (B) Agarose gel electrophoresis of rtPCR, where reverse transcription was performed on RNA purified from the four recombinant strains after arabinose induction, corresponding to one of the biological replicates from every strain utilized in experiment shown in Figure 5; PCR amplification utilized the same primers used for cloning the four mcr gene constructs (Table 1). 1 Underlined sequences indicate restrictions sites used for cloning in pBAD24 (EcoRI, SalI). 2 Amplicon lengths (bp). 3 Efficiency of the qPCR. 4 Correlation (R 2 ) of the qPCR calibration. 5 Calibration of the qPCR: slope. (B) Agarose gel electrophoresis of rtPCR, where reverse transcription was performed on RNA purified from the four recombinant strains after arabinose induction, corresponding to one of the biological replicates from every strain utilized in experiment shown in Figure 5; PCR amplification utilized the same primers used for cloning the four mcr gene constructs (Table 1).

Figure 3.
Functionality DNA constructs for mcr expression. Concentration (copy number) and messenger RNA (mRNA) expression from the different mcr gene constructs were determined by qPCR with primers mcr1q, mcr3q, and recAq (Table 1) on DNA or complementary DNA (cDNA) (see above) from the four recombinant clones. Firstly, mcr DNA concentration (black bars) and mcr gene expression (white bars) from all strains were relativized to recA DNA or recA mRNA, respectively; secondly, every two-gene construct was normalized with respect to its single mcr-gene reference.
Furthermore, plasmids carrying two-gene constructs performed similarly for controlling expression of recombinant DNA since mRNA abundance from mcr genes did not change following the co-transcription with downstream sequences (Figure 3). Thus, hpap2 and dgkA coding sequences are, like upstream mcr genes, efficiently transcribed from the PBAD promoter.  (Table 1) on DNA or complementary DNA (cDNA) (see above) from the four recombinant clones. Firstly, mcr DNA concentration (black bars) and mcr gene expression (white bars) from all strains were relativized to recA DNA or recA mRNA, respectively; secondly, every two-gene construct was normalized with respect to its single mcr-gene reference.
Furthermore, plasmids carrying two-gene constructs performed similarly for controlling expression of recombinant DNA since mRNA abundance from mcr genes did not change following the Antibiotics 2020, 9, 531 6 of 15 co-transcription with downstream sequences ( Figure 3). Thus, hpap2 and dgkA coding sequences are, like upstream mcr genes, efficiently transcribed from the PBAD promoter.

Expression of hpap2 or dgkA Is Not Essential, Although It Slightly Increases Resistance to Colistin in E. coli
After induction with arabinose, E. coli strains that carry pBAD24 vectors expressing similar amounts of one or two gene constructs were inoculated in colistin media to detect antimicrobial resistance phenotypes conferred by functional expression of mcr and closely linked genes (Figure 4). Since a negative effect on cell viability was found after induction of mcr gene expression from recombinant plasmids, l-arabinose concentration in colistin-containing media was set at 0.02%, a concentration found to yield the best balance between transcription and toxicity [15]. In a gradient of colistin concentration between 0 and 3 mg/L and intervals every 0.25 mg/L, cells carrying mcr-3-dgkA or mcr-1-hpap2 reached maximal colistin resistance (2.75 and 2.5 mg/L, respectively), slightly higher than that of their relative single-gene constructs mcr-3 or mcr-1 (2.0 and 2.25, respectively).

Expression of hpap2 or dgkA Is Not Essential, Although It Slightly Increases Resistance to Colistin in E. coli
After induction with arabinose, E. coli strains that carry pBAD24 vectors expressing similar amounts of one or two gene constructs were inoculated in colistin media to detect antimicrobial resistance phenotypes conferred by functional expression of mcr and closely linked genes (Figure 4). Since a negative effect on cell viability was found after induction of mcr gene expression from recombinant plasmids, L-arabinose concentration in colistin-containing media was set at 0.02%, a concentration found to yield the best balance between transcription and toxicity [15]. In a gradient of colistin concentration between 0 and 3 mg/L and intervals every 0.25 mg/L, cells carrying mcr-3-dgkA or mcr-1-hpap2 reached maximal colistin resistance (2.75 and 2.5 mg/L, respectively), slightly higher than that of their relative single-gene constructs mcr-3 or mcr-1 (2.0 and 2.25, respectively). Moreover, pre-induction with L-arabinose was required for strains carrying mcr-3 elements to reach resistance to colistin, whereas cultures expressing mcr-1 were only faintly more resistant to colistin than control cells (0.5 and 0.25 mg/L, respectively), although resistance was increased up to 1.5 mg/L for the mcr-1-hpap2 gene construct ( Figure 4). On the other hand, a pBAD24 construct carrying the single coding sequence of hpap2 did not increase colistin resistance in the same conditions assayed for mcr-carrying strains (not shown).

Expression of mcr Downregulates Genes for Phospholipid Metabolism Enzymes
Genes controlled by PmrAB, the TCSS that signals Fe 3+ binding to cell surface [3], were found negatively regulated by the expression of mcr genes ( Figure 5), including those for reducing the negative charge of lipid A (eptA, arnB, and pmrR), for the recycling of DAG (dgkA), and for increasing the negative charge of lipid A (lpxT) [16]. Since this is related to hPAP2, whose coding sequence co-expressed with mcr-1 improved colistin resistance (see above), the expression analysis was extended to the four genes from the E. coli genome that are involved in UPP dephosphorylation [7], which were also found downregulated, although to a different extent. The strongest effect was detected for pmrR, YbjG, and arnB genes, reaching 80% reduction (fivefold) of transcript accumulation in at least one analyzed condition ( Figure 5A). Moreover, pre-induction with l-arabinose was required for strains carrying mcr-3 elements to reach resistance to colistin, whereas cultures expressing mcr-1 were only faintly more resistant to colistin than control cells (0.5 and 0.25 mg/L, respectively), although resistance was increased up to 1.5 mg/L for the mcr-1-hpap2 gene construct (Figure 4). On the other hand, a pBAD24 construct carrying the single coding sequence of hpap2 did not increase colistin resistance in the same conditions assayed for mcr-carrying strains (not shown).

Expression of mcr Downregulates Genes for Phospholipid Metabolism Enzymes
Genes controlled by PmrAB, the TCSS that signals Fe 3+ binding to cell surface [3], were found negatively regulated by the expression of mcr genes ( Figure 5), including those for reducing the negative charge of lipid A (eptA, arnB, and pmrR), for the recycling of DAG (dgkA), and for increasing the negative charge of lipid A (lpxT) [16]. Since this is related to hPAP2, whose coding sequence co-expressed with mcr-1 improved colistin resistance (see above), the expression analysis was extended to the four genes from the E. coli genome that are involved in UPP dephosphorylation [7], which were also found downregulated, although to a different extent. The strongest effect was detected for pmrR, YbjG, and arnB genes, reaching 80% reduction (fivefold) of transcript accumulation in at least one analyzed condition ( Figure 5A). Antibiotics 2020, 9, x 7 of 15 Figure 5. mRNA accumulation for phospholipid metabolism enzymes after expression of mcr-1-hpap2 and mcr-3-dgkA gene constructs. Gene expression was quantified by qPCR with primers and conditions mentioned in Table 1, whereas cell growth and induction, RNA extraction, cDNA synthesis, amplification, and quantification are explained in detail in Section 4. The significance of differences between every condition and control cells carrying empty pBAD24 vector is indicated by *p < 0.1, **p < 0.05, and ***p < 0.01 (Student's t-test). Error bars are standard deviations based on three fully independent biological replicates. Expression of genes is separately shown depending on maximal reduction achieved: (A) 5.0-fold; (B) 1.7-fold; (C) 1.25-fold.
A weaker effect was observed for dgkA, eptA, and lpxT gene expression, whose reduction was limited to 40% (1.7-fold), whereas bacA and pgpB genes shared a minimal although still significant downregulation of 20% (1.25-fold). The composition of mcr gene constructs conditioned this signaling; mcr-3 expression decreased more strongly than mcr-1 the accumulation of mRNA from ybjG, arnB, eptA, and bacA genes, whereas the mcr-1 effect was higher on pmrR and pgpB gene expression. In general, little or no differences were found upon comparing the co-expression of mcr-1/-3 genes alone or linked to hpap2 or dgkA, respectively. Thus, with subtle differences between mcr-1 and mcr-3 genes, their expression had a profound impact on the regulation of enzymes for phospholipid metabolism, decreasing the transcription of enzymes for decoration of lipid A with L-Ara4N and PEtN and for UPP dephosphorylation.  Table 1, whereas cell growth and induction, RNA extraction, cDNA synthesis, amplification, and quantification are explained in detail in Section 4. The significance of differences between every condition and control cells carrying empty pBAD24 vector is indicated by * p < 0.1, ** p < 0.05, and *** p < 0.01 (Student's t-test). Error bars are standard deviations based on three fully independent biological replicates. Expression of genes is separately shown depending on maximal reduction achieved: (A) 5.0-fold; (B) 1.7-fold; (C) 1.25-fold.

Expression of mcr Genes, But Not hpap2, Confers Bacitracin Resistance
A weaker effect was observed for dgkA, eptA, and lpxT gene expression, whose reduction was limited to 40% (1.7-fold), whereas bacA and pgpB genes shared a minimal although still significant downregulation of 20% (1.25-fold). The composition of mcr gene constructs conditioned this signaling; mcr-3 expression decreased more strongly than mcr-1 the accumulation of mRNA from ybjG, arnB, eptA, and bacA genes, whereas the mcr-1 effect was higher on pmrR and pgpB gene expression. In general, little or no differences were found upon comparing the co-expression of mcr-1/-3 genes alone or linked to hpap2 or dgkA, respectively. Thus, with subtle differences between mcr-1 and mcr-3 genes, their expression had a profound impact on the regulation of enzymes for phospholipid metabolism, decreasing the transcription of enzymes for decoration of lipid A with L-Ara4N and PEtN and for UPP dephosphorylation.

Expression of mcr Genes, But Not hpap2, Confers Bacitracin Resistance
Resistance to bacitracin, an antimicrobial peptide that binds to UPP and sequesters it from PAP2 enzymes [7], was previously shown to occur, although to a limited extent, in E. coli strains expressing mcr-1 whose resistance increased from 1 to 2 mg/mL [17]. Accordingly, expression of mcr-1 or mcr-3 from pBAD24 vector conferred a modest increase in bacitracin resistance, from 2 up to 4 mg/mL, independently of the linked gene hpap2, which cloning in the same vector designed for optimal expression and translation of its coding sequence did not show any increase in bacitracin resistance ( Figure 6). Furthermore, co-expression of mcr-3 and dgkA further increased bacitracin resistance to 6 mg/mL. Antibiotics 2020, 9, x 8 of 15 Resistance to bacitracin, an antimicrobial peptide that binds to UPP and sequesters it from PAP2 enzymes [7], was previously shown to occur, although to a limited extent, in E. coli strains expressing mcr-1 whose resistance increased from 1 to 2 mg/mL [17]. Accordingly, expression of mcr-1 or mcr-3 from pBAD24 vector conferred a modest increase in bacitracin resistance, from 2 up to 4 mg/mL, independently of the linked gene hpap2, which cloning in the same vector designed for optimal expression and translation of its coding sequence did not show any increase in bacitracin resistance ( Figure 6). Furthermore, co-expression of mcr-3 and dgkA further increased bacitracin resistance to 6 mg/mL. Figure 6. Level of bacitracin resistance from mcr-1-hpap2 and mcr-3-dgkA gene constructs. XL1-Blue MRF' cells carrying the different recombinant plasmids were grown, induced with 0.2% arabinose, and spotted on Mueller-Hinton agar plates or liquid medium containing 0.02% arabinose supplemented with bacitracin at various levels, following the protocol described in Section 4. pBAD corresponds to XL1-Blue MRF' cells carrying intact pBAD vector.

Discussion
The mcr-1 and mcr-3 genes can be traced back, with their linked sequences hpap2 and dgkA, to far-related organisms like Moraxella and Aeromonas, respectively [1]. The conservation of these gene clusters during evolution (Figure 1), assuming that natural selection tends to eliminate superfluous and costly characters, suggests the existence of still unknown functional connections, the discovery of which could be the key to designing new strategies against antimicrobial resistance.
The strong effect of mcr gene expression on the regulation of phospholipid metabolism is, to our knowledge, shown in this work for the first time ( Figure 5). Under this condition, the genes downregulated are arnB, dgkA, eptA, pmrR, ybjG, bacA, lpxT, and pgpB, with all but the last three being positively controlled by PmrAB (Figure 7) [2,3,6,7]. Since PEtN transferase activity would decrease LPS negative charge, mcr genes expressed in E. coli may interfere with iron signaling by reducing the negative charge of cellular surface, decreasing Fe 3+ availability and the activation status of the PmrB sensor [3]. Thus, cells downregulated their endogenous systems for LPS modification with PEtN and L-Ara4N, tagged here by arnB and eptA, respectively (Figure 5 A,B).

Discussion
The mcr-1 and mcr-3 genes can be traced back, with their linked sequences hpap2 and dgkA, to far-related organisms like Moraxella and Aeromonas, respectively [1]. The conservation of these gene clusters during evolution (Figure 1), assuming that natural selection tends to eliminate superfluous and costly characters, suggests the existence of still unknown functional connections, the discovery of which could be the key to designing new strategies against antimicrobial resistance.
The strong effect of mcr gene expression on the regulation of phospholipid metabolism is, to our knowledge, shown in this work for the first time ( Figure 5). Under this condition, the genes downregulated are arnB, dgkA, eptA, pmrR, ybjG, bacA, lpxT, and pgpB, with all but the last three being positively controlled by PmrAB (Figure 7) [2,3,6,7]. Since PEtN transferase activity would decrease LPS negative charge, mcr genes expressed in E. coli may interfere with iron signaling by reducing the negative charge of cellular surface, decreasing Fe 3+ availability and the activation status of the PmrB sensor [3]. Thus, cells downregulated their endogenous systems for LPS modification with PEtN and L-Ara4N, tagged here by arnB and eptA, respectively ( Figure 5A,B). Antibiotics 2020, 9, x 9 of 15 Between the two members of the pap2 family also showing their mRNA accumulation to be decreased, ybjG and lpxT, the latter was unexpected since it was not previously found to be controlled by PmrAB [6]. However, the stronger repression of pmrR suggests that a significant amount of the LpxT enzyme could remain active to balance the lower expression of ybjG and, thus, the regulation mediated by PmrAB would adjust UPP recycling during LPS modification (Figure 7). Nevertheless, this condition might correspond to a disturbed metabolic status, since LpxT and MCR Between the two members of the pap2 family also showing their mRNA accumulation to be decreased, ybjG and lpxT, the latter was unexpected since it was not previously found to be controlled by PmrAB [6]. However, the stronger repression of pmrR suggests that a significant amount of the LpxT enzyme could remain active to balance the lower expression of ybjG and, thus, the regulation mediated by PmrAB would adjust UPP recycling during LPS modification (Figure 7). Nevertheless, this condition might correspond to a disturbed metabolic status, since LpxT and MCR enzymes would compete for the same substrate, the 1 -phosphate lipid A, and that downregulation of the endogenous dgkA gene may limit DAG recycling.
Members of the PAP2 protein family, also known as lipid phosphate phosphatases (LPPs), are Mg 2+ -independent phosphatases that, in eukaryotes, hydrolyze phosphatidic acid to yield inorganic phosphate and diacylglycerol (DAG) for biosynthesis of triacylglycerol and phospholipids [18]. In prokaryotes, PAP2 proteins are functionally diverse and, among other functions, three enzymes from E. coli are involved in C 55 -P synthesis from UPP, a closely regulated committed step (Figure 7) [7]. C 55 -P is required as the biosynthetic platform for peptidoglycan synthesis, as well as for LPS modification by L-Ara4N, but Gram-negative bacteria can only synthetize it from UPP via its dephosphorylation. hPAP2 proteins encoded closely linked to mcr-1-like genes are far-related to UPP phosphatases from E. coli ( Figure 8A). One of these, PgpB, is a phosphatidylglycerol phosphatase integrally located in the plasma membrane by six transmembrane helices, and its topological and 3D studies revealed a periplasm-oriented active site built by six residues from three conserved motifs C1, C2, and C3 [19][20][21]. These conform the phosphate binding and the catalytic triad signature, His C2 -Asp C3 -His C3 , which is strictly conserved among the three PAP2 enzymes from E. coli, but not in hPAP2 proteins that mismatch the central residue of the motif ( Figure 8A), suggesting different substrate specificities. Accordingly, the fact that hpap2 expression in E. coli could not increase bacitracin resistance ( Figure 6) may indicate that hPAP2 activity does not have UPP phosphatase activity unlike BacA, PgpB, or YbjG [21].   . Key residues for protein functioning that were previously determined are indicated by blue (role in activity) or red (role in folding) triangles [5].
The moderated increase in bacitracin resistance conferred by the expression of mcr-1 or mcr-3 genes could be mediated by the strong decrease in pmrR, which may result in a higher level of LpxT activity, and/or by the reduction of the negative charge around bacterial cells that blocks interactions Key residues for protein functioning that were previously determined are indicated by blue (role in activity) or red (role in folding) triangles [5].
The moderated increase in bacitracin resistance conferred by the expression of mcr-1 or mcr-3 genes could be mediated by the strong decrease in pmrR, which may result in a higher level of LpxT activity, and/or by the reduction of the negative charge around bacterial cells that blocks interactions with positively charged antimicrobials. Thus, a hypothesis to contrast in the near future is that colistin and bacitracin co-resistance conferred by mcr genes may not be the result of co-selection but derived from the cross-talk between LPS decoration and UPP recycling.
Co-expression of hpap2 and mcr-1 genes is required for maximal expression of colistin resistance, although the increase is too modest to be detected in a classical, two-dilution fold, minimal inhibitory concentration (MIC) assay (Figure 4), which may explain why previous studies described contradictory results like a null effect or an essential role of hpap2 expression in antimicrobial susceptibility [9,10]. hPAP2 belongs to the PAP2 family, which predicts membrane location, with active residues oriented toward the periplasm and a putative phosphatase activity on unknown substrate(s), which might not be UPP (see above) [7]. Clues about hPAP2 function might be derived from the homology to far-related sequences from Francisella tularensis and Helicobacter pylori [22][23][24]. Two PAP2 proteins, LpxE and LpxF, which are required for the natural resistance to colistin of these human pathogens, are phosphatases for 1 -phosphate or 4 -phosphate lipid A, respectively. Interestingly, LpxE from H. pylori is encoded closely linked and downstream the eptA ortholog from this organism (not shown), a genetic organization that invokes that of mcr-1-hpap2. Moreover, expression of LpxE or LpxF from Francisella in E. coli increases polymyxin resistance [22,23], although a similar role for hPAP2 could be ruled out since it does not produce a similar phenotype when mcr-1 is not co-expressed (not shown). Since the strongest repression of pmrR was found after expressing mcr-1, and a relatively lower repression of lpxT was detected in cells carrying the mcr-1-hpap2 construct, suggesting a higher availability of LpxT activity in this condition, the most consistent hypothesis for a role of hPAP2 in colistin resistance would be to convert 1 -bis-phosphate to 1 -phosphate lipid A, withdrawing the action of LpxT and increasing accessibility for PEtN addition by MCR-1-like enzymes (Figure 7).
In contrast to hpap2 genes, the dgkA sequence closely linked to mcr-3 determinant that was also shown to slightly improve colistin resistance (Figure 4) is closely related to its ortholog found in the genome of E. coli ( Figure 8B). The hypothetical role of dgkA in colistin resistance can also be postulated on the basis of the strong conservation of their structural and functional determinants. Indeed, DAGK is the enzyme involved in recycling the membrane-disrupting DAG produced from PA by PEtN transferases like EtpA or MCR enzymes (Figure 7) [5]. E. coli DAGK is a protein integral in the plasma membrane with the active site oriented to the cytosol. All functional and structural determinants placed in the cytosol-contacting surface of its three helical segments are strongly conserved among the endogenous DAGK from E. coli and those encoded in the vicinities of mcr-3-like genes ( Figure 7B), suggesting conservation of functionality. However, the dgkA gene linked to mcr-8, the less closely related genetic element, presents four polymorphisms in key residues for enzyme activity and/or folding of the protein [5].
This work presents evidence for the involvement of genes hpap2 and dgkA in the function of mcr-1 and mcr-3, two major determinants for colistin resistance. Their possible role may be to compensate for alterations of phospholipid metabolism during colistin resistance, which can provides clues to understand bacterial physiology and might contribute to a better control of antimicrobial resistance determinants.

Plasmid Constructs, Strains, and Growth Conditions
Specific primers (Table 1) were designed to amplify the full coding sequence of mcr-1 (mcr1c), mcr-1-hpap2 (mcr-pap2c), mcr-3 (mcr3c), mcr-3-dgkA (mcr3dgkAc), and hpap2 (pap2c). PCR was performed with High-Fidelity PCR Enzyme Mix (New England Biolabs, Ipswich MA, USA) according to the manufacturer's protocol. PCR conditions for amplification included a primary denaturation step of 5 min at 98 • C, followed by 30 cycles of 30 s at 98 • C, 30 s at 59.5 • C for annealing, and 2 min at 72 • C for elongation, and a final elongation step of 10 min at 72 • C. DNA samples for amplification of PCR products were obtained by boiling overnight cultures. The mcr-1 or mcr-3 determinants were obtained from HSP38 or Eco46 strains, whose genomes were fully sequenced after their isolation from a human infection or from the feces of a healthy bovine, respectively [25,26]. PCR products were purified (MEGAquick-spin plus fragment DNA purification kit, iNtRON Biotechnology, Seongnam-Si Korea, Republic of (South)), digested with enzymes EcoRI and SalI (New England Biolabs), and ligated (T4 DNA ligase, Thermofisher, Waltham MA, USA) to the arabinose-inducible pBAD24 vector (Life Science Market, Nova lifetech Limited, Hongkong). XL1 blue MRF', a K12 derivative that was obtained from Statagene (Agilent, Santa Clara, CA, USA), was electroporated for selection of strains carrying recombinant plasmids according to classical methods [27]. XL1-Blue MRF' cells carrying intact pBAD vector were used as a control strain for gene expression and antimicrobial resistance studies.

Quantitative PCR
Determination of the relative concentration of plasmids within cells was performed by real-time PCR using mcr1q or mcr3q primer pairs and recAq as a calibration reference (Table 1), whereas DNA samples were obtained using the non-selective boiling method (see above).
Gene expression analyses were performed with cells growing in liquid cultures (Mueller-Hinton broth supplemented with 100 mg/L ampicillin). After overnight growth, cultures were renewed by diluting 1/10 with fresh media supplemented with 0.2% arabinose and incubated at 37 • C with strong shaking (200 rpm). When cell cultures reached 0.3-0.5 optical density (OD) at 600 nm, they were quickly cooled on ice, centrifuged, and processed for RNA extraction (Aurum Total RNA Minikit, Bio-Rad), and then reverse-transcribed (PrimeScript™ RT reagent Kit, Takara); next, genomic DNA was removed (TURBO DNA-free kit, Ambion), according to the manufacturer's protocols. SYBRgreen real-time quantitative assays were carried out using the SYBR ® Premix Ex Taq™ II (Tli RNase H Plus; Takara Bion Inc.) and an Applied Biosystems ® Step One PCR System. Oligo Primer Analysis Software v. 7 was utilized to design primer sequences with optimal amplification efficiencies ( Table 1). The normalized relative quantities (NRQ) of transcripts were obtained using the 2 −∆∆Ct calculation method with the expression of recA gene used as a calibration reference, with every experimental condition including two technical replicates (duplicate reactions in the same qPCR) and three biological replicates from fully independent experiments. Ratios between the mean NRQ for every treatment (the different mcr-1 ± hpap2 and mcr-3 ± dgkA constructs cloned in pBAD24 and transformed in XL1-Blue MRF' cells) and control condition (XL1-Blue MRF' cells carrying intact pBAD24 vector) and the standard error of the ratios were calculated according to previously reported methods [28].

Determination of Antibiotic Resistance
Colistin resistance was determined using cation-adjusted Mueller-Hinton agar (Sigma Aldrich). Liquid cultures (Mueller-Hinton broth) grown overnight with 100 mg/L ampicillin were renewed by diluting 1/10 with fresh media supplemented with 0.2% arabinose and incubated at 37 • C with strong shaking (200 rpm). When cell cultures reached OD 600 nm = 1.0, bacterial growth was diluted to 0.5 of the McFarland standard, corresponding approximately to OD 600 nm = 0.1 or 1.5 × 10 8 CFU/mL (colony-forming units per mL), and 10-µL aliquots were spotted on Mueller-Hinton agar plates containing 0.02% arabinose and colistin or bacitracin at indicated concentrations, incubated at 37 • C overnight. The empty vector pBAD24 cloned in the same genetic background was used as a negative control. Colistin susceptibility was defined as the highest concentration of antibiotic that yielded visible growth of bacteria, whose values were confirmed using three independent experiments.

Conclusions
hpap2 and dgkA genes closely linked to mcr-1 and mcr-3 might play a role on colistin resistance by compensating phospholipid metabolism functions altered during LPS modification by colistin resistance determinants.