Modifications in the pmrB gene are the primary mechanism for the development of chromosomally encoded resistance to polymyxins in uropathogenic Escherichia coli
Methods : Two complementary approaches , saturated transposon mutagenesis and spontaneous mutation induction with high concentrations of colistin and polymyxin B , were employed to select for mutations associated with resistance to polymyxins . 
Mutants were identified using transposon-directed insertion-site sequencing or Illumina WGS . 
A resistance phenotype was confirmed by MIC and further investigated using RT -- PCR . 
Competitive growth assays were used to measure fitness cost . 
Results : A transposon insertion at nucleotide 41 of the pmrB gene ( EC958pmrB41-Tn5 ) enhanced its transcript level , resulting in a 64 - and 32-fold increased MIC of colistin and polymyxin B , respectively . 
Three spontaneous mutations , also located within the pmrB gene , conferred resistance to both colistin and polymyxin B with a corresponding increase in transcription of the pmrCAB genes . 
All three mutations incurred a fitness cost in the absence of colistin and polymyxin B. 
Conclusions : This study identified the pmrB gene as the main chromosomal target for induction of colistin and polymyxin B resistance in E. coli . 
Introduction
Polymyxins are polypeptide antibiotics originally isolated in 1947 from Paenibacillus polymyxa subsp . 
colistinus .1,2 Both polymyxin B and E ( colistin ) are highly effective against many Gram-negative bacteria , including most members of the Enterobacteriaceae ; however , their use in clinical medicine has been limited due to their neurotoxic and nephrotoxic side effects .3 -- 5 Despite these limitations , the increasing incidence of infections caused by MDR Gram-negative pathogens , in particular carbapenem-resistant Enterobacteriaceae , has led to renewed interest in the revival of polymyxin B and colistin as last-line treatments .6 -- 8 The bactericidal activity of polymyxins against Gram-negative bacteria initiates through their interaction with negatively charged 
LPS in the outer membrane , subsequently permeabilizing the outer membrane and disrupting the inner membrane leading to lytic 9 cell death . 
Therefore , the main mechanism of polymyxin resistance occurs via modifications of LPS to decrease its net negative charge , weakening its interaction with cationic poly-10 myxins . 
These LPS modifications are directly controlled by the PmrAB two-component regulatory system ( TCS ) .11 In Escherichia coli , Salmonella enterica , Klebsiella pneumoniae , Pseudomonas aeruginosa and Acinetobacter baumannii , the most common poly-myxin resistance mechanism involves modification of the PmrAB 12 -- 16 TCS . 
The pmrCAB operon encodes the phosphoethanolamine transferase ( PmrC ; also called EptA ) , the response regulator PmrA and the sensor kinase PmrB . 
Mutations in the pmrA or pmrB genes can lead to activation of PmrA , which in turn upregulates PmrC and enzymes of the arnBCADTEF-pmrE operon that are responsible for the biosynthesis and transfer of phosphoetha-nolamine and 4-deoxyaminoarabinose to lipid A , respectively .8,11 Other polymyxin resistance mechanisms include mutations in the PhoPQ TCS that indirectly activate PmrAB via PmrD ,16 adsorption ofpolymyxins by surface polysaccharides ( K. pneumoniae and P. aeruginosa ) ,17,18 modification of the Kdo with phosphoethanol-amine ( E. coli ) 19,20 and transport via efflux pumps ( K. pneumoniae ) .21 Alarmingly , a plasmid-mediated polymyxin resistance gene ( mcr-1 ) was recently discovered22 and there are already .100 reports describing the identification of this gene in multiple Gram-negative species across the globe .23,24 The mcr-1 gene encodes a phosphoethanolamine transferase originally found on a transferable IncI2 plasmid ; 22 mcr-1 is mobilized by an ISApl1 composite transposon and has been subsequently shown to reside in 25 multiple plasmid and chromosomal locations . 
One of the most important groups of pathogens for which poly-myxin B and colistin are considered as last-line treatments is uro-pathogenic E. coli ( UPEC ) that cause urinary tract infection ( UTI ) . 
The most clinically important subgroup of UPEC belongs to the glo-bally disseminated MDR ST131 clone . 
ST131 was originally identi-fied in 2008 as a major clone associated with dissemination of the CTX-M-15-type ESBL gene .26 -- 29 Multiple epidemiological studies have since highlighted the importance of ST131 and its significance as the predominant fluoroquinolone-resistant UPEC clone worldwide .30 -- 34 Of major concern , resistance to last-line carbape-nems22 ,35 -- 37 and colistin via acquisition of the mcr-1 gene38 has been reported in ST131 . 
In light of this , a detailed analysis of the mechanisms of induced resistance to polymyxin B and colistin in ST131 would enhance our understanding of resistance development against this important last-line treatment . 
In this study , we used the complementary laboratory approaches of saturated transposon mutagenesis and spontaneous induction in the presence of high concentrations of polymyxin B and colistin to identify mutations leading to polymyxin resistance in the ST131 reference strain EC958 . 
Overall , we showed that the primary mechanism of induced polymyxin resistance involves modification of the pmrB gene . 
Materials and methods
Bacterial strains and growth conditions
The E. coli strain EC958 was isolated from the urine of a patient with community-acquired UTI in the UK .39 EC958 belongs to ST131 subclade C2 or H30Rx32 ,40 and is resistant to fluoroquinolones and third-generation cephalosporins .41 Bacterial strains were cultured on LB broth or agar with appropriate selection ( 30 mg/L chloramphenicol ) as required . 
Antimicrobial susceptibility testing
MICs of colistin and polymyxin B were determined using the broth dilution method .42 Briefly , 5 % 105 cfu/mL of bacteria were inoculated into a series of Mueller -- Hinton broths containing 2-fold dilutions of each antibiotic in 0.01 % acetic acid and 0.4 % BSA . 
The MIC was defined as the lowest concentration of antibiotic that completely inhibited bacterial growth after incubation at 37 C for 18 -- 24 h. 
Selection of polymyxin-resistant mutants
The EC958 mini-Tn5 saturated mutant library has been described previously .43 Approximately 2 % 108 cells from the mutant library were screened on LB agar supplemented with either colistin ( 5 mg/L ) or poly-myxin B ( 14 mg/L ) at 37 C for 18 h to select for resistant mutants . 
All trans-poson-resistant mutants were pooled for subsequent analysis . 
To identify 8 spontaneous resistant mutants , 2 % 10 WT EC958 cells were grown under the same conditions ; resistant colonies were subcultured and stored individually in 10 % glycerol in LB at # 80 C. 
Transposon-directed sequencing, WGS and sequencing data analysis
A total of 282 colistin-resistant transposon mutants and 243 polymyxin B-resistant transposon mutants were pooled and DNA samples were extracted for transposon-directed insertion-site sequencing ( TraDIS ) analysis as described previously .43,44 In brief , the Illumina library preparation was performed using the Nextera DNA Sample Prep Kit ( Illumina ) following the manufacturer 's instructions with modifications for TraDIS , and sequenced with 100 bp single-end reads on the Illumina MiSeq platform . 
Spontaneous resistant mutants were sequenced using Illumina technology . 
The sequencing libraries were prepared using the Nextera DNA Library Prep Kit . 
The libraries were sequenced with 2 % 100 bp paired-end reads using the Illumina HiSeq 2500 . 
Illumina reads were analysed using an in-house pipeline for quality control . 
High quality Illumina reads were mapped to the EC958 genome ( HG941718 ) using SHRIMP 2.045 and nucleo-tide variations were identified using Nesoni ( www.vicbioinformatics.com / software.nesoni.shtml ) . 
Quantitative RT–PCR (qRT–PCR) and 50-RACE
RNA samples were extracted from cells grown to mid-log ( OD600 0.6 ) in LB broth using the RNAeasy Mini Kit ( QIAGEN ) and then converted into cDNA using SuperScript III Reverse Transcriptase ( Invitrogen , Life Technologies ) . 
qRT -- PCR analysis of the pmrCAB genes was performed in triplicate using the ABI SYBR Green PCR Master Mix on the ViiA 7 Real-Time PCR System ( Life Technologies ) . 
Relevant primers are listed in Table S1 ( available as Supplementary data at JAC Online ) . 
Relative expression levels were calculated by the 2 #DDCT method46 using gapA as an endogenous control and the WT pmrCAB genes as a reference . 
The transcription start site of pmrB in the EC958pmrB41-Tn5 mutant was identified using the 50-RACE system ( QIAGEN ) according to the manufacturer 's instructions . 
Synthesis of cDNA specific for pmrB was performed using reverse transcriptase with the pmrBspecific primers pmrB_GSP1 and pmrB_GSP2 ( Table S1 ) . 
These PCR amplicons were sequenced using the BigDye Terminator v3 .1 Cycle Sequencing Kit ( Life Technology ) . 
Competitive growth assays
The growth of each spontaneous polymyxin-resistant mutant was examined in a mixed competitive growth experiment against EC958lacZ , which has an identical growth rate to WT EC958 .43 Overnight cultures of the spontaneous polymyxin-resistant mutants and EC958lacZ were standardized to OD600 '' 0.05 in LB before mixing at a 1:1 ratio and incubating at 37 C with shaking at 250 rpm for 18 h. cfu counts at time '' 0 and time '' 18 h were performed on MacConkey agar to allow differentiation of EC958Dlac ( white colo-nies ) and resistant mutants ( red colonies ) . 
The relative competitive fitness index ( W ) of each polymyxin-resistant mutant compared with WT EC958 was calculated using the formula W '' ln ( cfumutant at t18/cfumutant at t0 ) / ln ( cfu 47 EC958lacZ at t18/cfuEC958lacZ at t0 ) . 
Competition assays were performed as 10 independent biological replicate experiments . 
Results
100
Upregulation of pmrB leads to colistin and polymyxin B resistance
A mini-Tn5 mutant library comprising 1 million independent EC958 mutants was screened to identify mutants resistant to poly-myxins following growth in the presence of polymyxin B and colistin , respectively . 
The mini-Tn5 transposon employed to mutagenize EC958 was designed such that there was no transcriptional terminator downstream of the chloramphenicol resistance ( cat ) gene ; thus , the promoter of the cat gene can drive the transcription of a downstream gene when the insertion position is favourable . 
Consequently , within the transposon mutant library there were two types of polymyxin B/colistin-resistant mutants that could be selected ; mutants in which the transcription of a gene is induced by the upstream insertion of the transposon and mutants in which a gene is insertionally inactivated . 
From a total of 2 % 108 EC958 mutants plated on solid LB me-dium supplemented with 5 mg/L colistin or 14 mg/L polymyxin B , 282 colistin-resistant and 243 polymyxin B-resistant colonies , respectively , were obtained . 
These mutants were pooled , genomic DNA was extracted and the mini-Tn5 insertion sites were identified using TraDIS . 
The overwhelming majority of transposon-directed reads mapped to the pmrB gene ( EC958_4592 ) , accounting for 95.54 % and 95.78 % of total reads for colistin-resistant and poly-myxin B-resistant mutants , respectively ( Figure 1 and Table S2 ) . 
Interestingly , the insertion sites of mini-Tn5 were largely specific to one location in pmrB ; corresponding to nucleotide 41 ( 94.81 % and 95.75 % of total reads in colistin - and polymyxin B-resistant mutants , respectively ) ( Table S2 ) . 
All mini-Tn5 inserted at this site possessed the cat promoter in the same direction as the pmrB CDS ( Figure 2 ) . 
Forty-seven mini-Tn5 mutant colonies resistant to colistin , and 42 colonies resistant to polymyxin B were subcultured and screened by PCR for insertion of the mini-Tn5 at nucleotide 41 of pmrB ( pmrB41 : : Tn5 ) . 
The results revealed 100 % and 88 % of colo-nies , respectively , contained the insertion at this specific location . 
The MICs for two independently isolated pmrB41 : : Tn5 mutants ( EC958pmrB41-Tn5 ) of colistin and polymyxin B were increased 64 - and 32-fold compared with WT EC958 , respectively ( Table 1 ) . 
Furthermore , 50-RACE analysis demonstrated that transcription of the pmrB gene in EC958pmrB41-Tn5 was driven by the cat promoter , confirming that insertion of the mini-Tn5 cassette in this mutant was responsible for the transcription of the downstream pmrB gene . 
TraDIS also identified several genes with a high number of insertion sites but a low number of mapped reads ( ,0.4 % of the total reads ) : envZ , ompR and ompC , indicating that insertional inactivation of these genes might result in tolerance but not growth upon exposure to colistin/polymyxin B ( Figure S1 and Table S3 ) . 
Defined mutants inactivated for these genes were generated and shown to exhibit the same MIC as WT EC958 ( data not shown ) . 
Thus , these mutations might confer a subtle advantage in survival that we could not detect based on MIC analysis , an interpretation consistent with a recent study that showed antibiotic tolerance facilitates the rapid subsequent evolution of resistance .48 
Spontaneous mutations in the pmrB gene also promote resistance to colistin and polymyxin B
In parallel to the selection of mini-Tn5 mutants resistant to colistin and polymyxin B , the WT EC958 strain was also cultured under the same conditions ( solid medium supplemented with 5 mg/L co-listin or 14 mg/L polymyxin B ) to select for spontaneously arising resistant mutants . 
In this analysis , nine mutants were identified ( six resistant to colistin and three resistant to polymyxin B ) . 
WGS of these resistant mutants revealed that they all contained mutations within the pmrB gene ( Table S4 ) . 
The most common pmrB mutation was a G to A non-synonymous substitution at nucleotide 251 ( Cys-84 ! 
Tyr ; EC958pmrB-C84Y ) , which was found in seven of the nine mutants . 
The mutations in the other two mutants were a G445T substitution ( Asp-149 ! 
Tyr ; EC958pmrBD149Y ) and a 12 bp deletion from nucleotide 258 to nucleotide 269 ( GlnAlaValArgArg-86 -- 90 ! 
His ; EC958pmrB-D12bp ) , respect-ively . 
The MICs for these mutants of colistin and polymyxin B increased by 32 -- 64-fold , regardless of their mode of selection ( Table 1 ) . 
Of note , EC958pmrB-D149Y and EC958pmrB-D12bp possessed the highest level of resistance against colistin and poly-myxin B ( MIC '' 8 mg/L ; 64-fold increase compared with the WT ) . 
Colistin and polymyxin B resistance is associated with increased transcription of the pmrCAB genes
To understand better the mechanism by which mutations in the pmrB gene lead to colistin and polymyxin B resistance in EC958 , transcription of the pmrCAB genes was assessed in each mutant by qRT -- PCR . 
In EC958pmrB41-Tn5 , the pmrB gene was 3.1-fold upregulated compared with WT EC958 . 
Similarly , the transcription of pmrC and pmrA ( which lie upstream of pmrB ) was also upregulated 6.9 - and 6.6-fold , respectively ( Figure 3 ) , suggesting a feedback loop linking pmrB with their transcription . 
In congruence with this result , the spontaneously induced resistant mutants also possessed elevated pmrCAB transcript levels compared with WT 
EC958 ( Figure 3 ) . 
However , the magnitude of pmrCAB upregulation varied ; the largest increase in pmrCAB transcription was observed in EC958pmrB-D149Y ( 74.8 - , 39.9 - and 14.3-fold increase in pmrC , pmrA and pmrB transcript level , respectively ) . 
The pmrCAB transcript level in EC958pmrB-C84Y and EC958pmrBD12bp was similar ; 10 - , 6 - and 3-fold increase compared with WT EC958 ( Figure 3 ) . 
Spontaneous mutants resistant to polymyxins incur a fitness cost
We hypothesized that the spontaneous mutations in pmrB that led to increased colistin and polymyxin B resistance may incur a fitness cost . 
Thus , the growth of each of the polymyxin-resistant mutants was examined in LB broth using a mixed competitive assay against EC958Dlac , a derivative of EC958 that has an identical growth rate as the WT strain ( Figure 4 ) but can be phenotypically differentiated on MacConkey agar . 
Overall , all of the spontaneously derived polymyxin-resistant mutants displayed a significantly reduced fitness compared with the EC958Dlac strain ( P , 0.0001 , one-way ANOVA ; Figure 4 ) . 
The fitness index varied among the pmrB mutants , with EC958pmrB-D149Y , which possessed the greatest resistance to colistin and polymyxin B , conversely exhibiting the greatest growth deficiency compared with EC958Dlac ( W '' 0.90 ) . 
Both EC958pmrB-C84Y and EC958pmrB-D12bp grew at a reduced rate relative to EC958Dlac ( W '' 0.95 and 0.97 , respectively ) ( Figure 4 ) . 
Finally , to investigate the possibility of cross-resistance to other antimicrobial cationic peptides , we measured the MICs for all of the mutants generated in this study of LL-37 , a soluble cationic peptide that contributes to host innate immunity ,49 and a key mediator of mucosal immunity in the urinary tract .50 Our results showed that all mutants exhibited a 2-fold increase in MIC of LL-37 ( from 16 to 32 mg/L ) compared with WT EC958 ( Table 1 ) . 
Discussion
The 50-year-old cationic peptide antibiotics polymyxin B and colis-tin have recently undergone a revival for clinical use in response to the increasing incidence of infections caused by MDR Gramnegative pathogens .6,51 They are among the few effective antibiotics reserved for ESBL-producing bacteria that are also resistant to gentamicin and carbapenems .6 Despite this , resistance to these antibiotics has been reported in a number of Gramnegative bacteria , including E. coli ,52,53 Salmonella typhimurium ,54 K. pneumoniae ,55 P. aeruginosa56 and A. baumannii .14 Here , we employed two parallel approaches , transposon mutagenesis and spontaneous induction , to understand the evolution of polymyxin resistance in the reference E. coli ST131 strain EC958 . 
We used a highly saturated mini-Tn5 mutant library in EC95843 to select for gain-of-function mutants resistant to colis-tin and polymyxin B. Based on our approach , we hypothesized that gain-of-function mutants could be achieved by two mechanisms , upregulation of a downstream gene via the promoter on the mini-Tn5 cassette or insertional inactivation . 
Remarkably , one transposon insertion site ( at nucleotide 41 of the pmrB gene ) accounted for .95 % of sequencing reads from TraDIS data acquired from 525 resistant colonies . 
This particular site was enriched from among 311 possible insertion sites within the pmrCAB locus from the input mutant library ( Figure 1 ) . 
Investigation using 50-RACE revealed that the transcription of pmrB in this mutant ( EC958pmrB41-Tn5 ) was driven by the cat promoter on the mini-Tn5 . 
The consequence of this insertion was increased transcription of the pmrCAB genes , leading to a 64 - and 32-fold increase in the MIC of colistin and polymyxin B , respectively ( Table 1 ) . 
Notably , the insertion also truncated the pmrB gene , and thus it is possible that the resulting change in the N-terminus of the modified PmrB changed its activity . 
Transposon mutagenesis was used previously in P. aeruginosa to identify genes involved in polymyxin B susceptibility .57 However , this previous study only identified one mutant with an insertion in the phoQ gene that showed increased resist-ance to polymyxin B . 
The second approach employed to select for resistance to colis-tin and polymyxin B involved the isolation of spontaneous mutants following challenge in the presence of high concentrations of the two antibiotics . 
In total , nine mutants representing three different types of sequence changes were identified . 
The Cys-84 ! 
Tyr substitution was the most common type ( found in seven mutants ) , but conferred the lowest MICs of colistin and polymyxin B ( Figure 2 and Table 1 ) . 
In contrast , the 12 bp deletion and Asp-149 ! 
Tyr mutation were found once , respectively , and conferred a higher level of resistance to both antibiotics ( Table 1 ) . 
The Asp-149 ! 
Tyr substitution is located near the autophosphorylation site of PmrB at histidine residue 152 in the histidine kinase domain ( Figure 2 ) . 
It is possible that this change may increase PmrB activity , resulting in more phosphorylated PmrA regulator , which increased the transcription of the whole pmrCAB operon as observed by our qRT -- PCR analysis ( Figure 3 ) . 
Both the Cys-84 ! 
Tyr substitution and the 12 bp deletion are located within the HAMP domain of PmrB , which is proposed to be important for signal transduction from the peri-plasmic input to the kinase domain .58 Mutations within the HAMP domain in other histidine kinases of E. coli such as EnvZ and CpxA result in constitutive activation of their respective receptor histidine kinase .59 -- 61 Thus , the mutations in the HAMP domain of PmrB might lead to constitutive activation of PmrA . 
Notably , all three mutants were recovered in reduced numbers compared with WT EC958 in mixed competitive growth experiments , indicating that these mechanisms of resistance to colistin and polymyxin B occurred at a fitness cost to the cell ( Figure 4 ) . 
The precise mechanism by which this occurs and the capacity for compensating mutations to alleviate this effect remains an area of future investigation . 
Mutations in the PmrAB two-component system have been associated with resistance to colistin and polymyxin B in clinical E. coli isolates ( Table S5 ) . 
One study reported a Val-161 ! 
Gly mutation in PmrB and a double mutation of Ser-39 ! 
Ile and Arg-81 ! 
Ser in PmrA in two colistin-resistant E. coli strains isolated from swine faeces ; both mutations resulted in an MIC of colistin of 4 mg/L .62 Another report identified a range of mutations in PmrB associated with colistin resistance , including Pro-7-Gln-12del ( deletion of six amino acids ) , Ala-159 ! 
Val , Thr-156 ! 
Lys and Ile-91-Thr-92-ins-Ile ( insertion of isoleucine at position 92 ) .63 The three PmrB mutations reported here are novel in E. coli , and in add-ition to resistance against colistin and polymyxin B these mutations also resulted in a 2-fold increase in the MIC of the human cathelicidin peptide LL-37 . 
LL-37 is a cationic peptide produced by urothelial cells that contributes to innate protection against UTI .50 Importantly , these observations are consistent with the concerning trend of cross-resistance to LL-37 through the clinical use of colistin .64 -- 66 Overall , the impact of acquired resistance to colistin and polymyxin B on sensitivity to soluble cationic peptides present in human urine such as LL-37 remains to be assessed . 
Our analysis did not identify mutations in other regulatory pathways that led to colistin or polymyxin B resistance . 
This includes the PhoPQ TCS and mgrB , both of which play a major role in colistin and polymyxin B resistance in other Gram-negatives .11,16,19,20,67 As our mutant library contained multiple insertions within phoPQ 43 and mgrB , this suggests that mutation of these genes does not lead to significant colistin and polymyxin B resistance in EC958 under selection conditions employed in our experiments . 
Since its discovery in late 2015 , the plasmid-mediated polymyxin resistance gene mcr-1 has made the threat of pan-22 ,38,68 resistance imminent . 
However , despite increased awareness of polymyxin resistance and the global prevalence of mcr-1 ,23 there are very few reports that describe the impact of chromosomal mutations that lead to colistin and polymyxin B resistance in E. coli . 
While the work described in this study demonstrates that resistance to colistin and polymyxin B in E. coli occurs through modification of the pmrB gene , the fact that mcr-1 has now been 38 described in ST131 highlights the alarming scenario that its spread may be dramatically enhanced in a clone of E. coli that has already demonstrated its capacity to disseminate rapidly across the globe . 
The combined effect of chromosomal modifications in pmrB and acquisition of mcr-1 on polymyxin resistance remains to be evaluated . 
Funding
This work was supported by a grant from the National Health and Medical Research Council ( NHMRC ) of Australia ( GNT1067455 ) and High Impact Research ( HIR ) grants from the University of Malaya ( UM-MOHE HIR Grant UM C/625/1 / HIR/MOHE/CHAN / 14/1 , no . 
H-50001-A000027 ; UM-MOHE HIR Grant UM C/625/1 / HIR/MOHE/CHAN / 01 , no . 
A000001 -- 50001 ) . 
N. T. K. N. is supported by an Australian Government Research Training Program Scholarship . 
M. A. S. , S. A. B. , M. J. W. and D. L. P. are supported by NHMRC Fellowships . 
Supplementary data