Polymyxin resistance in Klebsiella pneumoniae: multifaceted mechanisms utilized in the presence and absence of the plasmid-encoded phosphoethanolamine transferase gene mcr-1

Abstract

Objectives

Until plasmid-mediated mcr-1 was discovered, it was believed that polymyxin resistance in Gram-negative bacteria was mainly mediated by the chromosomally-encoded EptA and ArnT, which modify lipid A with phosphoethanolamine (pEtN) and 4-amino-4-deoxy-l-arabinose (l-Ara4N), respectively. This study aimed to construct a markerless mcr-1 deletion mutant in Klebsiella pneumoniae, validate a reliable reference gene for reverse transcription quantitative PCR (RT–qPCR) and investigate the interactions among mcr-1, arnT and eptA, in response to polymyxin treatments using pharmacokinetics/pharmacodynamics (PK/PD).

Methods

An isogenic markerless mcr-1 deletion mutant (II-503Δmcr-1) was generated from a clinical K. pneumoniae II-503 isolate. The efficacy of different polymyxin B dosage regimens was examined using an in vitro one-compartment PK/PD model and polymyxin resistance was assessed using population analysis profiles. The expression of mcr-1, eptA and arnT was examined using RT–qPCR with a reference gene pepQ, and lipid A was profiled using LC-MS. In vivo polymyxin B efficacy was investigated in a mouse thigh infection model.

Results

In K. pneumoniae II-503, mcr-1 was constitutively expressed, irrespective of polymyxin exposure. Against II-503Δmcr-1, an initial bactericidal effect was observed within 4 h with polymyxin B at average steady-state concentrations of 1 and 3 mg/L, mimicking patient PK. However, substantial regrowth and concomitantly increased expression of eptA and arnT were detected. Predominant l-Ara4N-modified lipid A species were detected in II-503Δmcr-1 following polymyxin B treatment.

Conclusions

This is the first study demonstrating a unique markerless deletion of mcr-1 in a clinical polymyxin-resistant K. pneumoniae. The current polymyxin B dosage regimens are suboptimal against K. pneumoniae, regardless of mcr, and can lead to the emergence of resistance.

Introduction

MDR Gram-negative bacteria present a serious threat to global health.¹ The lack of effective antimicrobial agents to combat MDR pathogens has led to the revival of ‘old’ antibiotics such as the polymyxins (i.e. polymyxin B and colistin).² Inevitably, the reintroduction of polymyxins has been followed by increasing reports of resistance in Gram-negative ‘superbugs’.³ In November 2015, polymyxin resistance mediated by a plasmid-borne phosphoethanolamine (pEtN) transferase gene, mcr-1, was reported and generated significant concerns of potential rapid dissemination of polymyxin resistance via horizontal transfer.⁴ Since then, eight mcr genes (mcr-1 to -8) have been discovered in several Gram-negative species, including Klebsiella pneumoniae.^4–12K. pneumoniae is a serious nosocomial pathogen and can rapidly evolve to gain MDR status, limiting the treatment options to the last-line polymyxins.¹³ Alarmingly, outbreaks of hospital- and community-acquired infections caused by polymyxin-resistant MDR K. pneumoniae have been increasingly reported globally over the last few years.^14–16

LPS is the predominant component in the outer leaflet of the Gram-negative outer membrane (OM) and the initial target of polymyxins.¹⁷ Polymyxin resistance in Gram-negative bacteria primarily occurs via the addition of positively charged moieties [e.g. pEtN and 4-amino-4-deoxy-l-arabinose (l-Ara4N)] to the lipid A component of LPS.¹⁸ Such lipid A modifications reduce the net negative charge and thereby diminish the electrostatic interaction with polymyxins.¹⁷ Lipid A modifications with pEtN can be mediated by either a chromosomally encoded pEtN transferase gene, eptA (pmrC) or a plasmid-encoded mcr gene. In contrast, l-Ara4N modification of lipid A is mediated by a transferase encoded by arnT (pmrK) that is located exclusively on the chromosome.^18,19 Recent preclinical and clinical studies indicated that the current recommended dosage regimens of intravenous polymyxin B and colistin are suboptimal.^20,21 In the present study, a markerless mcr-1 deletion mutant was constructed from a clinical K. pneumoniae isolate carrying mcr-1 on a large native plasmid. Our study provides the first set of pharmacokinetics/pharmacodynamics (PK/PD) data for polymyxins against mcr-carrying K. pneumoniae and highlight that the current polymyxin B dosage regimens are suboptimal against K. pneumoniae, irrespective of mcr, and potentially cause the rapid emergence of resistance.

Materials and methods

Antibiotic solutions

Sterile solutions of polymyxin B (sulphate; Beta Pharma, Shanghai, China), colistin (sulphate; Beta Pharma) and kanamycin (sulphate; Astral Scientific, NSW, Australia) were prepared in Milli-Q^™ water (Millipore, USA) and filtered through 0.22 μm syringe filters (Sartorius, Germany).

Bacterial strains

K. pneumoniae II-503 was isolated from a urine sample of a hospitalized patient (Zhejiang, China). A markerless mcr-1 deletion mutant (II-503Δmcr-1) was constructed from K. pneumoniae II-503 in the present study (below).

Construction of a markerless mcr-1 deletion mutant and complementation

Information on all strains and plasmids used in this study is provided in Table S1 (available as Supplementary data at JAC Online) and primers in Table S2. The native plasmid, pII-503, was first isolated from K. pneumoniae II-503, then transformed into NEB^® 10-beta electrocompetent Escherichia coli. The mutagenesis cassette was constructed using a splice overlap extension (SOE) PCR method.²² The kanamycin resistance gene, aph, flanked by the flippase recognition target was PCR-amplified from the template plasmid pKD4.²³ The left and right regions of the SOE PCR product (∼1000 bp), representing the left and right regions flanking the mcr-1, were generated using pII-503 as the template. Double crossover mutagenesis involving lambda red recombinase (pKD46) was performed and mutants were selected on Luria-Bertani agar supplemented with 50 mg/L kanamycin.²³ The pII-503Δmcr-1:Kan^R was confirmed with PCR and sequencing. Plasmid incompatibility was exploited to introduce pII-503Δmcr-1:Kan^R into K. pneumoniae II-503 and ‘cure’ K. pneumoniae II-503 of pII-503 as follows. The pII-503Δmcr-1:Kan^R was introduced into K. pneumoniae II-503 via electroporation and transformants were selected on Luria-Bertani agar containing 50 mg/L kanamycin to preferentially select for pII-503Δmcr-1:Kan^R. Loss of pII-503 and retention of pII-503Δmcr-1:Kan^R were confirmed by PCR. To make a markerless mcr-1 deletion mutant, the kanamycin resistance cassette was removed from pII-503Δmcr-1:Kan^R in K. pneumoniae using the flippase (FLP) recombinase-recombination target (FRT) site-specific recombination system.²⁴ PCR analysis confirmed the loss of the kanamycin resistance cassette in kanamycin-susceptible K. pneumoniae II-503Δmcr-1 transformants. Complementation was performed by introducing a functional copy of mcr-1 on the recombinant plasmid, pBBR1MCS-2_mcr-1 into K. pneumoniae II-503Δmcr-1.²⁵

In vitro plasmid stability assay

The stability of pII-503 and pII-503Δmcr-1 in K. pneumoniae was determined by daily passaging (1:1000 dilution) into antibiotic-free CAMHB over 4 days. Each day, the culture was diluted and plated onto Mueller–Hinton (MH) agar to obtain isolated colonies. Fifty colonies were randomly picked and checked for the presence of the correct plasmid using colony PCR.

Polymyxin susceptibility testing

MICs of polymyxin B and colistin were determined in CAMHB for K. pneumoniae II-503, II-503Δmcr-1 and the complemented strain using broth microdilution.²⁶

Lipid A extraction and structural analysis

Lipid A from K. pneumoniae II-503 and II-503Δmcr-1 following polymyxin B treatment (8 mg/L) for 24 h was extracted using a mild acid hydrolysis method and analysed using LC-MS;^27,28 a group without polymyxin B treatment was included as the control.

Static time–kill assay

Static time–kill assays with polymyxin B at 2, 4 and 8 mg/L were performed against K. pneumoniae II-503 and II-503Δmcr-1 with a starting inoculum of ∼6 log₁₀ cfu/mL. Bacterial cultures were collected at 0, 1, 4 and 24 h to determine the bacterial killing kinetics.²⁹

In vitro one-compartment PK/PD model (IVM)

PK/PD of polymyxin B against K. pneumoniae II-503 and II-503Δmcr-1 was performed using an IVM to simulate the PK of polymyxin B in patients with average steady-state concentrations (C_{ss, avg}) of 1 and 3 mg/L.^30–32 The reservoirs contained 80 mL of CAMHB and were maintained at 37°C. The flow rate of CAMHB was set at 4.7 mL/h to mimic the t_½ of polymyxin B (i.e. 11.9 h) in patients. Polymyxin B doses were administered every 12 h as 1 h infusions. Samples (1.5 mL) were collected at 0, 1, 4, 8, 24, 25, 28, 32 and 48 h for viable counting and expression analysis of mcr-1, eptA and arnT. Population analysis profiles (PAPs) were conducted at 0, 24 and 48 h using polymyxin B-containing plates (0.5, 1, 4 and 8 mg/L). Three biological replicates were performed for each dosage regimen.

RNA extraction and reverse transcription quantitative PCR (RT–qPCR)

Trizol^® (Invitrogen, USA) was employed to isolate RNA from bacterial cultures harvested from the IVM. DNase treatment was performed with an RNase-free DNase set (QIAGEN, Germany) and RNA was purified using the phenol:chloroform extraction method.³³ Synthesis of cDNA was done with an AffinityScript qPCR cDNA synthesis kit (Agilent Technologies, USA) and RT–qPCR was conducted using Brilliant III ultra-fast qPCR master mix (Agilent Technologies) with an Agilent AriaMx real-time PCR system. Four candidate reference genes (uvrD, arcA, mfd and pepQ) were selected based on previous transcriptomics data generated in our laboratory (data not shown). Two algorithms, geNorm and NormFinder, were used to determine the most stably expressed gene.^34,35 Using GeNorm, a gene with an M-value below the threshold of 1.5 is considered to be stably expressed and a lower M-value indicates greater stability.³⁴ NormFinder calculates the expression stability value (SV) and greater stability is indicated by a lower SV.³⁵ Statistical analyses were performed using non-parametric Mann–Whitney U test or one-way ANOVA followed by Dunn’s multiple comparison test.

In vivo infection model

Ethics approval was obtained from the Monash University Animal Ethics Committee and the animal infection experiment was conducted in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. Female Swiss mice (8 weeks old) were housed with food and water available ad libitum. The efficacy of polymyxin B treatment against K. pneumoniae II-503 and II-503Δmcr-1 was investigated in a neutropenic mouse thigh infection model with two mice (i.e. four thighs) per group.³⁶ Polymyxin B was administered subcutaneously at 60 mg/kg/day in three divided doses over 24 h. Mice were euthanized at 24 h and the infected thigh was removed, homogenized and filtered through a sterile filter bag (pore size 280 μm), followed by plating on antibiotic-free MH agar and MH agar containing 4 mg/L polymyxin B. Statistical analysis was performed using non-parametric Mann–Whitney U test. To determine the stability of pII-503 and pII-503Δmcr-1 in this K. pneumoniae isolate following infection in mice, colony PCR was performed with 20 randomly chosen colonies recovered from each mouse to detect the presence of plasmid. The fAUC/MIC values were calculated based on previously generated PK data.³⁷

Results

Polymyxin susceptibility and static time–kill kinetics

After successful construction of the mcr-1 deletion mutant, the in vitro stability assay conducted over 4 days confirmed that both plasmids, pII-503 and pII-503Δmcr-1, were stably maintained in the clinical K. pneumoniae isolate. MICs of polymyxin B and colistin were 8 mg/L for K. pneumoniae II-503. In contrast, K. pneumoniae II-503Δmcr-1 displayed significantly reduced polymyxin resistance with an MIC of 0.5 mg/L. Importantly, complementation of the mcr-1 deletion mutant with a functional copy of mcr-1 restored the polymyxin resistance to an MIC of 4 mg/L, proving that mcr-1 was responsible for the level of resistance observed in this isolate. Static time–kill studies with K. pneumoniae II-503 treated with polymyxin B at 2, 4 and 8 mg/L over 24 h revealed that the growth of K. pneumoniae II-503 was reduced by 1.4 and 3.3 log₁₀ cfu/mL following polymyxin B treatment at 4 and 8 mg/L, respectively, compared with the control group at 4 h (Figure 1). In contrast, a bactericidal effect (≥3 log₁₀ cfu/mL reduction compared with the initial inoculum) was observed after 1 h treatment with all three concentrations against K. pneumoniae II-503Δmcr-1 (Figure 1). Notwithstanding, extensive regrowth at 24 h was observed for both isolates with all treatment groups (Figure 1).

Figure 1.

Static time–kill kinetics of K. pneumoniae II-503 and II-503Δmcr-1 treated with polymyxin B (PMB) at 2, 4 and 8 mg/L over 24 h.

Lipid A profiling

Four predominant peaks at m/z 1745, 1761, 1825 and 1841 were identified in the lipid A profiles generated from K. pneumoniae II-503 and II-503Δmcr-1, grown in the presence or absence of polymyxin B (8 mg/L) (Figure 2a–d). Based on the literature,³⁸ the peak at m/z 1825 corresponds to hexa-acylated species with two glucosamines, two phosphates, four 3-OH-C₁₄ and two C₁₄, whereas m/z 1841 represents 2-hydroxylated hexa-acylated lipid A species. The peaks at m/z 1745 and 1761 represent the dephosphorylated forms of lipid A species at m/z 1825 and 1841, respectively (Figure 2e and f). The MS analyses of lipid A from K. pneumoniae II-503, in the absence and presence of polymyxin B, revealed four major peaks corresponding to pEtN-modified hexa-acylated lipid A species at m/z 1868, 1884, 1948 and 1964 (Figure 2a, b and e). These predominant pEtN-modified peaks were not observed in the lipid A profiles of the mcr-1 deletion mutant, irrespective of the treatment (Figure 2c and d). In K. pneumoniae II-503Δmcr-1, two predominant peaks at m/z 1892 and 1972 were observed after exposure to polymyxin B, corresponding to the l-Ara4N-modified lipid A (Figure 2d and f).

Figure 2.

Lipid A profiles of K. pneumoniae II-503 (a) without treatment, and (b) treated with 8 mg/L polymyxin B; and K. pneumoniae II-503Δmcr-1 (c) without treatment, and (d) treated with 8 mg/L polymyxin B. The pEtN-modified lipid A species are indicated by grey vertical arrows and l-Ara4N-modified lipid A species are indicated by black vertical arrows. Structures of unmodified lipid A species are shown on the left of the arrow with the corresponding lipid A modified with (e) pEtN (grey solid line box) and (f) l-Ara4N (black dashed line box) shown on the right.

IVM

For K. pneumoniae II-503, the polymyxin B regimen with C_{ss, avg} of 1 mg/L led to a slight reduction of bacterial viability (−1.07±1.24 log₁₀ cfu/mL) at 4 h, compared with the untreated group (Figure 3a). In contrast, exposure to a higher polymyxin B dosage regimen (C_{ss, avg} of 3 mg/L) resulted in a greater reduction of bacterial viability at 4 h (−2.74±0.65 log₁₀ cfu/mL) and 8 h (−1.70±0.60 log₁₀ cfu/mL) (Figure 3a). Rapid initial bacterial killing was observed for K. pneumoniae II-503Δmcr-1 at both polymyxin B dosage regimens, with bacterial reduction of −3.72±1.09 (C_{ss, avg} of 1 mg/L) and −4.88±1.42 log₁₀ cfu/mL (C_{ss, avg} of 3 mg/L) at 1 h; and no viable bacteria were recovered at 4 h with polymyxin B at C_{ss, avg} of 3 mg/L (Figure 3b). Despite this significant initial bacterial killing effect, regrowth of K. pneumoniae II-503Δmcr-1 rapidly occurred. The PAPs at 48 h revealed the emergence of resistance in K. pneumoniae II-503Δmcr-1 after exposure to polymyxin B (Figure 3d) and similar PAPs were observed at 24 h (data not shown).

Figure 3.

Bacterial counts (mean ± SD, n = 3) of (a) K. pneumoniae II-503 and (b) K. pneumoniae II-503Δmcr-1 following polymyxin B (PMB) treatment in an IVM over 48 h. PAPs (mean ± SD, n = 3) of (c) K. pneumoniae II-503 and (d) K. pneumoniae II-503Δmcr-1 at 48 h.

Gene expression analysis of mcr-1, eptA and arnT

All the candidate reference genes demonstrated an M-value below the threshold, with the lowest M-value shown by pepQ. Similarly, the lowest SV was demonstrated by pepQ (0.157), compared with uvrD (0.465), arcA (0.531) and mfd (0.496). These results showed that pepQ was the most stably expressed gene under our experimental conditions. RT–qPCR analysis revealed that over 48 h the expression of mcr-1 in K. pneumoniae II-503 was ∼100-fold higher than the expression levels of eptA and arnT, irrespective of polymyxin B exposure (Figure 4a–c). Polymyxin B at C_{ss, avg} of 3 mg/L did not result in any significant change in the expression of any of the three genes in K. pneumoniae II-503, compared with the untreated controls (Figure 4a–c). Due to the significant killing of K. pneumoniae II-503Δmcr-1 at the early timepoints, 24 h was the earliest timepoint to obtain sufficient cells for RNA extraction; hence, only samples at 24 and 48 h were subjected to RT–qPCR analysis. Overall, an increasing trend in the expression of eptA and/or arnT was observed for K. pneumoniae II-503Δmcr-1 after treatment with polymyxin B, relative to the untreated controls (Figure 4d). The expression of eptA and arnT significantly increased in K. pneumoniae II-503Δmcr-1, following the treatment with polymyxin B at C_{ss, avg} of 3 mg/L at 24 h (Figure 4d). However, at 48 h the levels of eptA and arnT expression were comparable to those of the control except for one biological replicate, which showed relatively greater expression of eptA following polymyxin B at C_{ss, avg} of 1 mg/L (Figure 4e). Collectively, these results show that in the absence of mcr-1, increased expression of eptA and/or arnT could be essential for bacteria to develop resistance to polymyxin B.

Figure 4.

The normalized relative expression (mean ± SD, n = 3) of mcr-1, eptA and arnT in the WT strain K. pneumoniae II-503 at (a) 1 h, (b) 24 h and (c) 48 h. The normalized relative expression (mean ± SD; n = 3) of eptA and arnT in the mcr-1 mutant, K. pneumoniae II-503Δmcr-1 at (d) 24 h and (e) 48 h. The asterisk denotes a statistically significant difference with P < 0.05; ns, not significant.

In vivo infection study

The PCR analysis conducted on colonies recovered from the thighs of infected mice demonstrated that both pII-503 and pII-503Δmcr-1 were stably maintained in vivo. The initial bacterial burdens were 6.44±0.12 and 6.27±0.14 log₁₀ cfu/thigh for K. pneumoniae II-503 and II-503Δmcr-1, respectively. Polymyxin B at 60 mg/kg/day was ineffective against K. pneumoniae II-503, with comparable bacterial load between the untreated control group (9.0±0.26 log₁₀ cfu/thigh) and the treated group (8.6±0.18 log₁₀ cfu/thigh) at 24 h (Figure 5a). In contrast, 60 mg/kg/day polymyxin B led to significant killing of K. pneumoniae II-503Δmcr-1, with 3.0±0.18 log₁₀ cfu/thigh reduction in the bacterial load compared with the untreated group (Figure 5b). No emergence of polymyxin resistance was observed at 24 h with K. pneumoniae II-503Δmcr-1 following polymyxin B treatment (0.0087% ± 0.017%), as there was no significant increase in the polymyxin-resistant subpopulation compared with the untreated mice (0.0016% ± 0.00067%, P = 0.314) (Figure 5c).

Figure 5.

Bacterial counts (mean ± SD, n = 4) recovered on MH agar and 4 mg/L polymyxin B (PMB)-containing MH agar for (a) K. pneumoniae II-503 and (b) K. pneumoniae II-503Δmcr-1 at 24 h in the neutropenic mouse thigh infection model. (c) Percentage of K. pneumoniae II-503Δmcr-1 colonies recovered on 4 mg/L PMB-containing MH agar over antibiotic-free MH agar at 24 h in the neutropenic mouse thigh infection model. The asterisk denotes a statistically significant difference with P < 0.05; ns, not significant.

Discussion

Plasmid-mediated polymyxin resistance may be a potential contributor to the emergence of pandrug-resistant Enterobacteriaceae.³⁹E. coli has the highest prevalence of mcr-harbouring isolates, while the occurrence of mcr in K. pneumoniae remains at a lower prevalence, with mcr-1 being the most common type among the eight mcr genes reported to date.¹² However, mcr-harbouring K. pneumoniae in clinical settings is of concern due to limited therapeutic options for this problematic pathogen.⁴⁰ To the best of our knowledge, the present study is the first to: (i) use markerless mutagenesis to remove mcr-1 from a clinical K. pneumoniae isolate while still retaining the native plasmid; this unique approach allows other native plasmid functions to be maintained; (ii) identify and validate a reliable reference gene, pepQ, for the RT–qPCR experiment in the absence and presence of polymyxins; and (iii) integrate gene expression, lipid A profiling and in vitro and in vivo PK/PD to investigate the interplay among mcr-1, eptA and arnT.

Similar to other mcr-carrying Gram-negative bacteria, a moderate level of polymyxin resistance (MIC = 8 mg/L) was observed with the clinical isolate K. pneumoniae II-503. Notably, the markerless mcr-1 deletion plasmid was stable in K. pneumoniae host cells for at least 4 days. Removal of mcr-1 restored the polymyxin susceptibility and the polymyxin B MIC of K. pneumoniae II-503Δmcr-1 (0.5 mg/L) was below its epidemiological cut-off value (ECOFF, 2 mg/L).⁴¹ The complementation of the mcr-1 deletion mutant with functional mcr-1 restored the MIC to 4 mg/L, which was lower than the MIC observed in the WT. This was very likely due to the difference of the mcr-1 expression driven by the native promoter and the lac promoter on the complementing plasmid. In addition, the instability of the complementing plasmid in the absence of kanamycin during the MIC measurement could also lead to the loss of the complementing plasmid.

It should be emphasized that validation of the reference gene is essential prior to RT–qPCR analysis. The expression of commonly used reference genes gyrB and rpoB, in which rpoB has also been utilized to study the expression of mcr-1 in E. coli, was examined in our transcriptomics data.⁴² Surprisingly, both gyrB and rpoB were differentially expressed following polymyxin treatment in both K. pneumoniae and an mcr-1-carrying E. coli strain⁴³ (and S. C. Nang, M. Li, Y. Zhu, T. Velkov and J. Li, 2018, unpublished data). In our present study, pepQ was validated as a stably expressed gene under polymyxin treatment in our clinical K. pneumoniae isolate.

Polymyxin resistance via the addition of pEtN to lipid A can be mediated by pEtN transferases, EptA or Mcr, whereas the modification with l-Ara4N is mediated only by the l-Ara4N transferase ArnT.^18,19 Our data demonstrated that, in the presence of mcr-1, the constitutive addition of pEtN to lipid A played a key role in polymyxin resistance in K. pneumoniae II-503 and this is supported by the lipid A profiles, which revealed predominant peaks representing pEtN-modified lipid A species (Figure 2a and b). Our RT–qPCR data also showed that mcr-1 was constitutively expressed at similar levels in the presence or absence of polymyxin B (Figure 4a–c). Considering the structure–activity relationship of polymyxins,¹⁷ the lipid A phosphate groups are crucial for the antibacterial activity of polymyxins and our results showed that the level of mcr-1 expression was sufficient for the modification of lipid A with pEtN moieties to confer resistance to both clinically relevant dosage regimens. Overexpression of mcr-1 on a recombinant plasmid exerted an adverse effect on bacterial viability.⁴² It is very likely that the level of mcr-1 expression in the clinical K. pneumoniae II-503 was optimally maintained to provide a balance between polymyxin resistance and bacterial growth. Our results showed that in the mcr-1 deletion mutant the expression of eptA and arnT was both concentration- and time-dependent (Figure 4d and e). Interestingly, despite the increase in expression of eptA and arnT, the lipid A of K. pneumoniae II-503Δmcr-1 was mainly decorated with l-Ara4N following polymyxin B treatment, indicating that l-Ara4N modification was preferred over pEtN in the absence of mcr-1 (Figure 2d). Indeed, lipid A molecules of most polymyxin-resistant K. pneumoniae isolates analysed to date are modified with l-Ara4N.^44,45 We postulate that the expression of eptA and arnT, which are typically regulated by PhoPQ and/or PmrAB,^18,19,46 is dependent on a number of regulatory factors whereas the expression of mcr-1 on the plasmid is constitutive. The activity and t_½ of each of the transferases that act on the same site of lipid A could play a part in the dominance of certain lipid A modifications. Understanding the complex interplay between the plasmid-mediated mcr-1, chromosomal-mediated eptA and arnT in K. pneumoniae is crucial in optimizing polymyxin use in patients, and is currently conducted in our laboratory.

Both polymyxin B dosage regimens examined (C_{ss, avg} of 1 and 3 mg/L) are ineffective against mcr-1-carrying K. pneumoniae. Our data highlight the rapid emergence of resistance to polymyxins in K. pneumoniae in vitro in the absence of mcr-1 (Figure 3b). Interestingly, our in vivo data at 24 h revealed that polymyxin resistance did not emerge following polymyxin B treatment at 60 mg/kg/day for 24 h in mice infected with the mcr-1 deletion mutant (Figure 5c). The difference in the emergence of polymyxin resistance between our studies in vitro and in infected mice might be due to the presence of cationic antimicrobial peptides in vivo,^47,48 which might play a key role in limiting the emergence of polymyxin resistance. In addition, it is also possible that the 24 h experimental period in mice was not sufficient for the emergence of detectable polymyxin-resistant bacterial cells. In the clinical setting, polymyxin-resistant strains might be isolated from patients after many days of treatment.^49,50

The most predictive PK/PD index for the antibacterial effect of polymyxins is fAUC/MIC.^36,37 Among the limited PK/PD information on polymyxin B obtained in a neutropenic mouse thigh infection model, the fAUC/MIC values for stasis and 1 log₁₀ reduction against K. pneumoniae were 1.22–13.5 and 3.72–28, respectively.³⁷ Using the same infection model, 60 mg/kg/day polymyxin B was employed in our present study to mimic a dosage regimen of ∼6 mg/kg/day in patients, doubling the upper limit of the currently recommended dosage regimens (i.e. 2.5–3 mg/kg).³⁰ The corresponding fAUC/MIC against mcr-1-carrying K. pneumoniae II-503 was 1.5, but stasis was not achieved. This is not surprising, as an fAUC/MIC of 1.5 is within the lower range of the target reported.³⁷ With an fAUC/MIC of 24.2 against K. pneumoniae II-503Δmcr-1, 0.63±0.09 log₁₀ cfu/thigh (P < 0.05) reduction in bacterial viability was observed, compared with the initial bacterial inoculum. Our result with K. pneumoniae II-503Δmcr-1 (MIC = 0.5 mg/L) is reasonably consistent with the 1 log₁₀ reduction fAUC/MIC target of polymyxin B against K. pneumoniae.³⁷

Overall, this study is the first to construct a markerless mcr deletion mutant from a clinical isolate for the investigation of the interactions among mcr-1, arnT and eptA in response to polymyxin treatment. Our results demonstrated that, regardless of the mcr gene, resistance to polymyxins can occur in K. pneumoniae with polymyxin monotherapy, which highlights precautions with intravenous polymyxins as monotherapy. Therefore, there is an urgent need for novel effective interventions (e.g. synergistic combination therapies) to prevent the emergence of polymyxin resistance.