Genomic and Transcriptomic Analyses of Colistin-Resistant Clinical Isolates of Klebsiella pneumoniae Reveal Multiple Pathways of Resistance

Abstract

The emergence of multidrug-resistant (MDR) Klebsiella pneumoniae has resulted in a more frequent reliance on treatment using colistin. However, resistance to colistin (Col^r) is increasingly reported from clinical settings. The genetic mechanisms that lead to Col^r in K. pneumoniae are not fully characterized. Using a combination of genome sequencing and transcriptional profiling by RNA sequencing (RNA-Seq) analysis, distinct genetic mechanisms were found among nine Col^r clinical isolates. Col^r was related to mutations in three different genes in K. pneumoniae strains, with distinct impacts on gene expression. Upregulation of the pmrH operon encoding 4-amino-4-deoxy-l-arabinose (Ara4N) modification of lipid A was found in all Col^r strains. Alteration of the mgrB gene was observed in six strains. One strain had a mutation in phoQ. Common among these seven strains was elevated expression of phoPQ and unaltered expression of pmrCAB, which is involved in phosphoethanolamine addition to lipopolysaccharide (LPS). In two strains, separate mutations were found in a previously uncharacterized histidine kinase gene that is part of a two-component regulatory system (TCRS) now designated crrAB. In these strains, expression of pmrCAB, crrAB, and an adjacent glycosyltransferase gene, but not that of phoPQ, was elevated. Complementation with the wild-type allele restored colistin susceptibility in both strains. The crrAB genes are present in most K. pneumoniae genomes, but not in Escherichia coli. Additional upregulated genes in all strains include those involved in cation transport and maintenance of membrane integrity. Because the crrAB genes are present in only some strains, Col^r mechanisms may be dependent on the genetic background.

INTRODUCTION

The increasing occurrence of multidrug-resistant (MDR) Klebsiella pneumoniae has expanded reliance on last-line therapies like colistin, a cationic antimicrobial peptide, for effective treatment (1). Of concern is the growing recovery of colistin-resistant (Col^r) strains from clinical settings (2–4). Colistin disrupts membrane integrity through displacement of cations like Mg²⁺ and Ca²⁺ in the outer membrane, leading to cell lysis (5). Resistance mechanisms described to date involve lipopolysaccharide (LPS) modification, particularly through derivatization of lipid A phosphate moieties with a sugar or ethanolamine. These modifications reduce the electrostatic affinity between the cationic colistin and anionic LPS. Mutations in the transcriptional regulatory systems controlling these LPS modifications are a common genetic mechanism leading to colistin resistance. For example, the PhoPQ and PmrAB two-component regulatory systems (TCRS) regulate expression of the gene (pmrC) that codes for the addition of phosphoethanolamine (pETN) and genes encoding biosynthesis and lipid A transfer of 4-amino-4-deoxy-l-arabinose (Ara4N) (pmrHFIJKLM) (Fig. 1A). Other regulatory components in this pathway include PmrD and MgrB, two connector proteins that convey feedback between the PmrAB and PhoPQ TCRS (6–9) (Fig. 1A). Mutations in pmrAB, phoPQ, and mgrB have been identified as mechanisms conferring Col^r in several Gram-negative pathogens, including K. pneumoniae (10–15).

FIG 1.

Model of Col^r mechanisms. (A) Diagram of genes involved in colistin resistance based on published summaries from Gram-negative bacteria. (B) Model of newly identified proteins and potential interactions with previously described pathways of Col^r. (C) Genome segment comprising differentially expressed genes, including the mutated histidine kinase gene in UHKPC26 and UHKPC28.

How colistin resistance mechanisms alter the global transcription profile and how transcriptome profiles vary in the genetic mechanism(s) that confers colistin resistance remain to be examined. By coupling complete genome sequence data and whole-genome transcriptional characterization of colistin-susceptible (Col^s) and Col^r K. pneumoniae, we were able to relate genetic mechanisms of colistin resistance to gene expression changes.

MATERIALS AND METHODS

Strains and genomic analyses.

Strains originated from a collection of KPC-producing K. pneumoniae isolates obtained from a consortium of tertiary-care hospitals previously described (16, 17). Colistin MICs were determined with a Sensititre system (17) and confirmed by Etest strips (bioMérieux). Genome sequencing was previously described (17) and consisted of Illumina HiSeq reads assembled with Newbler and annotated using the Comprehensive Microbial Resource annotation pipeline (18). Genome sequences were compared using Mauve (19), and gene content analysis used PanOCT (20). Single-nucleotide variants (SNVs) were determined using BWA for sequence read mapping (21) and Mpileup (22) for SNV detection in paired and closely related strains. A SNV-based phylogeny was constructed from kSNP software (23) output, as described in reference 17, with the inclusion of additional non-ST258 reference genomes. Two strains had paired isogenic Col^s isolates (UHKPC57 with UHKPC179, and UHKPC27 with UHKPC52). SNVs were associated with Col^r based on their presence in one of the Col^r strains (Table 1) and the absence of that allele in all 48 Col^s strains from reference 17 and additional genomes of the same MLST type in GenBank.

TABLE 1.

Klebsiella pneumoniae strains and predicted Col^r-associated mutations

Strain	GenBank accession no.a	SRA accession no.b	MLSTc	Locationd	Colistin MIC (μg/ml)e	KPC typec	Cols parent	Mutated geneg	Protein accession no.	Mutation
UHKPC45	ARVO00000000	SRR900179	258b	Cleveland OH, USA	4	3		mgrB		Deletion of 863 bp starting at mgrB38
UHKPC52	ARVN00000000	SRR921382	258a	Cleveland OH, USA	32	Nonef	UHKPC27	mgrB		IS1 family IS in mgrB with adjacent deletion, missing first 14 bp of mgrB
UHKPC81	APVQ00000000	SRR900190	234	Cleveland OH, USA	4	2		mgrB		IS4 family IS inserted at mgrB23
VAKPC280	APVZ00000000	SRR900205	258a	Cleveland OH, USA	16	2		mgrB		IS5 family IS inserted at mgrB41
280_1220	ARSQ00000000	SRR900126	258a	Northeast Ohio, USA	16	2		mgrB		IS5 family IS inserted with adjacent deletion; full mgrB deletion
UHKPC179	ARSM01000000	SRR900157	15	Cleveland OH, USA	16	Nonef	UHKPC57	mgrB	EPO88337	C28Y substitution
DMC1316	ARSB00000000	SRR900140	258b	Detroit, MI, USA	1	3		phoQ	EPO27247	D434N substitution
UHKPC26	APVT00000000	SRR900163	258a	Cleveland OH, USA	16	2		crrB	EOY97329	L94M substitution
UHKPC28	ARRU00000000	SRR900165	258a	Cleveland OH, USA	16	2		crrB	EPN94815	Q10L substitution
UHKPC27	APVR00000000	SRR900164	258a	Cleveland OH, USA	0.25	2		NA		Isogenic Cols strain
UHKPC57	ARPR00000000	SRR900182	15	Cleveland OH, USA	0.5	2		NA		Isogenic Cols strain

^aAccession number for the genomic sequence.

^bAccession number for the RNA-seq data set.

^cMultilocus sequence type (MLST) and bla_KPC type determined as described in reference 15.

^dGeographic location of source hospital.

^eDetermined by Etest.

^fPlasmid present in the paired Col^s strain but not in the Col^r strain.

^gNA, not applicable.

RNA-Seq experiments.

The 57 strains characterized in reference 17 were grown to mid-log phase at 37°C in LB broth. Cells were harvested and preserved with RNAprotect (Qiagen) until extraction with UltraClean RNA isolation kits (MoBio). cDNA libraries were constructed with ScriptSeq Complete Gold kits (Epicentre Biosciences) and were sequenced on an Illumina HiSeq instrument. Reads from each strain were mapped to the corresponding genome assembly and RPKM (number of mapped reads per kilobase of gene length per million total mapped reads) values were calculated in CLC (version 7.0.4). Genes with significantly different RPKM values were identified using the Significant Analysis for Microarray (SAM) (24) statistical analysis component of Multiexperiment Viewer (MeV version 4.9 [www.tm4.org]), where Col^r and Col^s strains represented the two unpaired classes.

Complementation assay.

To determine whether the mutation in crrB was necessary to confer colistin resistance, a pUC-19-derived plasmid containing the wild-type crrB gene as well as the upstream and downstream flanking sequences was introduced into UHKPC26 and UHKPC28. To generate a plasmid vector with a zeocin marker that could be used for selection of transformants in the MDR background of these strains, the bla gene in the pUC19 vector was replaced with the open reading frame of the resistance gene for zeocin. Specifically, by using the primers Zeo_ORF_F and Zeo_ORF_R (primer sequences are shown in Table S1 in the supplemental material), a 412-bp PCR fragment representing the open reading frame (ORF) of the zeocin resistance gene was amplified from a pAF6-derived plasmid (25). A vector fragment ending with the bla promoter and terminator, but not extending into the bla ORF, was amplified as a 1,857-bp fragment from pUC19 using the primers Zeo-pUC19_F and Zeo-pUC19_R. These fragments were purified using a PCR clean-up kit (ZymoResearch) and combined using Gibson assembly (26). The assembly product was used for transformation of NEB5α Escherichia coli cells (New England BioLabs). Plasmids were prepared from the transformants using a miniprep kit (Qiagen) and confirmed using restriction analysis with SmaI (New England BioLabs).

To clone a fragment containing the crrB gene into the zeocin plasmid, a 2,125-bp fragment was amplified from the genomic DNA sample of the Col^s strain UHKPC27. The crrB locus is wild type in this strain. This amplification was first performed using Q5 polymerase and the primers MW_F_5Phos and MW_R. The amplified fragment was purified using a PCR purification kit (ZymoResearch) and further amplified using PrimeSTAR MAX polymerase (Clontech) and the primers MW_F and MW_R. The amplified fragment included the crrAB ORFs and the 5′ and 3′ flanking sequences. The zeocin vector was amplified using PrimeSTAR Max polymerase and the primers pUC19_MW_GA_F and pUC19_MW_GA_R. The amplified vector and insert fragments were purified using a PCR purification kit and used in Gibson assembly. The assembly mixture was used to transform NEB5α cells. Plasmids were purified from the transformants and confirmed using restriction analysis with HinfI and Sanger sequencing.

Electroporation of UHKPC26 and UHKPC28 was conducted as described previously (27) except that >1 μg of DNA was introduced into the transformation mixture and cells were recovered in 500 μl low-salt LB broth. The transformed cells were selected on an agar plate containing low-salt LB broth and 1,250 μg/ml zeocin. The presence of this plasmid in the colonies was confirmed using PCR with the primers Jct_Up_F and Jct_Up_R, where the amplification of the product in this PCR required the junction between the crrA upstream sequence and the vector to be present.

A broth microdilution assay was used to determine the MIC of UHKPC26 and UHKPC28 strains carrying the empty vector or the vector with the crrAB genes. Overnight cultures of UHKPC26 and UHKPC28 grown in cation-adjusted Mueller-Hinton broth (MHB) plus 1,250 μg/ml zeocin were washed, resuspended in MHB broth, and inoculated into MHB with a series of colistin concentrations. The Col^s strain UHKPC27 was used as a control.

Nucleotide sequence accession numbers.

Sequence reads are available from the Short Read Archive at NCBI under accession numbers SRR896011, SRR900124 to SRR900127, SRR900132 to SRR900135, SRR900138 to SRR900144, SRR900151 to SRR900158, SRR900160 to SRR900169, SRR900178 to SRR900184, SRR900187 to SRR900191, SRR900193, SRR900198, SRR900199, SRR900202 to SRR900208, and SRR921382.

RESULTS

Genomic characterization.

Nine independent clinical strains representing three MLST sequence types (15, 234, and 258) were previously sequenced as part of a survey of K. pneumoniae strains in Midwestern U.S. hospitals (17) (Table 1). The most common mechanism of Col^r involved mgrB alteration, a previously identified Col^r mechanism in K. pneumoniae (6, 28) (Table 1). MgrB is a negative regulator of PhoQ, and inactivation leads to overexpression of phoPQ (9). Six strains had independent mutations of mgrB. Four strains experienced mgrB disruption mediated by different classes of insertion sequence (IS) families inserted at different positions in the mgrB gene. The MgrB amino acid substitution in UHKPC179 (C28Y) occurred at a cysteine residue previously identified as involved in a key disulfide bond in MgrB (29); thus, the substitution of a tyrosine here likely interferes with its ability to repress PhoQ. Genome analysis indicated that an 863-bp deletion resulted in loss of the entire mgrB gene and an adjacent ORF from UHKPC45. Two of these strains (UHKPC52 and UHKPC179) had a matched Col^s isolate (UHKPC27 and UHKPC57, respectively) from the same patient obtained before the initiation of colistin treatment. In each case, the mgrB gene was intact in the parental isolates and disrupted in the Col^r isolate.

A second resistance mechanism was identified as a mutation in the phoQ gene in DMC1316. The PhoQ substitution (D434N) in the cytoplasmic GHKL domain (ATP-lid loop) in DMC1316 is likely critical for phosphate transfer to PhoP (30).

Strains UHKPC26 and UHKPC28 did not have any mutations in known colistin resistance pathways, but each strain possessed an independent mutation in the histidine kinase component of a previously uncharacterized two-component regulatory system, which we designate CrrAB to signify colistin resistance regulation. In each case, the “wild-type” allele is observed in Col^s strains of ST258 (e.g., locus H239_3061 from strain UHKPC45) (Table 1). This histidine kinase gene (crrB) and an adjacent response regulator (crrA) are present in all K. pneumoniae ST258 strains with the exception of UHKPC06, where an insertion sequence (IS) event resulted in the deletion of the region, and in many other K. pneumoniae strain types (Fig. 2A). It appears that all K. pneumoniae strains have either the crr genes or an IS-mediated deletion or substitution of the region. The crrAB genes and the two flanking genes (Fig. 1C) do not have clear orthologs in E. coli or Salmonella, but orthologs of these genes are present several Enterobacter genomes but in a different genomic context than in K. pneumoniae.

FIG 2.

Comparative analyses of the crrAB region. (A) Presence of crrAB plus upstream (conserved hypothetical membrane protein ORF) and downstream (glycosyltransferase-like ORF) region among K. pneumoniae strains shown with a genome-wide SNV-based phylogeny constructed using kSNP output. Shared circle color indicates shared genome content and organization; gray squares indicate the presence of a shared ISKpn26 insertion upstream of the conserved hypothetical membrane protein H239_3059. (B) Maximum-likelihood tree of protein sequence alignment of CrrB (353 amino acids) in publicly available K. pneumoniae genomes.

Because no matched Col^s parental strains were available for UHKPC26 or UHKPC28, we considered whether other variants in these strains might be involved in the Col^r phenotype by identifying SNVs present in these strains relative to closely related Col^s strains. The only shared nonsynonymous mutation in the two strains was in MenB (A244T), a naphthoate synthase protein that is part of the menaquinone biosynthesis pathway. Strain UHKPC28 had an additional mutation in RstA (F166Y), the response regulator component of RstAB, and a mutation in a transmembrane domain of WecA (G134A), which is involved in cell envelope synthesis. The complete list of sequence differences between UHKPC24, UHKPC26, and UHKPC28 genomes and the reference genome HS11286 are given in Table S2 in the supplemental material. None of these candidates were deemed to be stronger candidates as the Col^r mutation than the crrB mutations.

Transcriptome characterization.

Two primary patterns of changes in gene expression were observed in Col^r strains relative to Col^s strains (Table 2). Common to both patterns was increased expression of the pmrHFIJKLM operon relative to Col^s strains. This operon encodes genes responsible for the biosynthesis of Ara4N and modification of the lipid A component of LPS with this moiety. The six strains with mgrB modifications and the one phoQ mutant strain shared a pattern characterized by elevated phoPQ expression relative to Col^s strains with no difference in pmrCAB transcript levels. The mutational spectrum in mgrB suggests loss-of-function mutations in this negative regulator of phoPQ, and each strain showed higher phoPQ expression levels than Col^s strains from the collection. Of these strains, all but UHKPC52 also exhibited significantly elevated transcript levels for pmrD.

TABLE 2.

RPKM values for differentially expressed genes or those that have been previously associated with PhoPQ or PmrAB regulons in other organisms

^aA representative locus ID for each gene from the K. pneumoniae UHKPC45 genome sequence (ARVO01000000) is given. Locus IDs in orange are predicted to be part of the PhoPQ regulon.

^bMeans and standard deviations are given for all Col^s strains (n = 45). RPKM values in green are significantly different (P < 0.05) and >2-fold higher in Col^r strains than the mean value in Col^s strains. Yellow indicates significantly different (P < 0.05) but <2-fold-different RPKM values. NA, gene not present in the genome.

^cSD, standard deviation.

In UHKPC26 and UHKPC28, the two strains with mutations in crrB, a distinct transcriptional pattern was observed, characterized by elevated transcription of pmrCAB, with no difference in phoPQ RPKM values compared to Col^s strains. The sequences of the pmrCAB locus and flanking regions in these strains were identical to those in other ST258 strains. Expression levels were higher in UHKPC26 and UHKPC28 for crrB, a gene encoding an adjacent conserved hypothetical membrane protein, and a gene encoding an adjacent putative glycosyltransferase than in other Col^r and Col^s strains (Table 2 and Fig. 1C). In UHKPC26, the mutation is in the HAMP domain of the histidine kinase. In contrast, the histidine kinase mutation in UHKPC28 affects the predicted signal peptide domain, and transcript levels of the response regulator crrA were unaffected. Neither rstA nor wecA, two other genes with sequence variants in UHKPC28, had altered expression in UHKPC28 relative to Col^s strains.

Other components of the PhoPQ regulon, such as slyB (encoding an outer membrane lipoprotein) and mgtA (encoding a magnesium-importing ATPase), were upregulated in all Col^r strains (31–33), but pagP (encoding an outer membrane protein that adds a palmitate chain to lipid A), lpxO (encoding a dioxygenase that modifies lipid A) (34, 35), and ybjG (encoding a putative undecaprenyl pyrophosphate phosphatase) (36) were upregulated only in the mgrB and phoQ mutant strains. Other genes with altered transcription include those associated with cation transport, membrane integrity, and the macAB efflux transporters (Table 2). Expression of pmrD was elevated in some strains yet was not correlated with expression of pmrCAB under these experimental conditions for strains with mgrB or phoQ alterations. RPKM values for mgrB for strains with full-length sequences did not vary between Col^r and Col^s strains, indicating that the phoQ mutation in DMC1316 and related increased expression did not activate the feedback inhibition mediated by MgrB. No correlation was observed between the colistin MIC and the genetic mechanism of resistance: the MIC of strains with mgrB mutations ranged from 4 to 32 μg/ml. This could indicate that variation in additional genes, including genes in the phoPQ regulon, may modulate the resistance phenotype.

Confirmation of the role of crrB in colistin resistance.

To confirm that the crrB mutations in UHKPC26 and UHKP28 were necessary to confer colistin resistance, we cloned the wild-type crrAB genes from UHKPC27 into a pUC19-derived vector carrying a zeocin resistance marker and introduced this plasmid into the Col^r strains. The UHKPC27 sequence is identical to that in UHKCP26 and UHKPC28 except for the hypothesized Col^r-associated mutations. Colistin susceptibility was restored to UHKPC26 and UHKPC28 cells carrying the crrAB plasmid but not to cells with the vector alone (Table 3).

TABLE 3.

Complementation of crrB mutations

Strain	MIC (μg/ml)a
UHKPC27	<0.5
UHKPC26	>16
UHKPC26 + vector	>16
UHKPC26 crrAB	<0.5
UHKPC28	>16
UHKPC28 + vector	>16
UHKPC28 crrAB	<0.5

^aDetermined by broth microdilution in cation-adjusted Mueller-Hinton broth.

DISCUSSION

Two distinct transcription profiles were observed; common to both was increased expression of the Ara4N pathway. Previous studies have shown linkage between Col^r and pmrCAB expression (6, 12) and cross talk between pmrAB and phoPQ (37). The strains reported here demonstrate upregulation of either the pmrAB or phoPQ genes, but not both in the same Col^r strain. The identification of the crr genes as additional regulators of colistin resistance expands the number of known genes that modulate this phenotype and highlights multifaceted ways that cells respond to antimicrobial peptide challenge.

The CrrAB proteins are not orthologs of other TCRS linked to PhoPQ or cell envelope stress response, such as RcsAB (38), RstAB (32), EvgAS (36), CpxAR (39), and BaeSR (40). Comparative analyses indicate that the crrAB region is variably present in K. pneumoniae, although it is present in nearly all ST258 strains (green circles in Fig. 2) and is found in other Klebsiella and Enterobacter species, albeit in different genomic contexts. In Enterobacter sp. strain SST3, a plant endophyte, the region is adjacent to the sap (sensitivity to antimicrobial peptides) operon (41). In K. pneumoniae genomes without these genes, either a lipoprotein of unknown function is present at this genomic location, or a phage-related integrase and transposase are present. The GC content of this region is approximately 40%, much lower than the average GC content of >50% for the K. pneumoniae chromosome, suggesting that it was laterally acquired. Genomic analysis also suggests that IS events are reshaping the region. ST258a strains described in reference 17 all have the same ISKpn26 insertion upstream of the conserved hypothetical protein (H239_3062), which also resulted in the deletion of 468 bp and truncation of an adjacent ABC transporter (H239_3063). Interestingly, eight ST258b strains described in reference 42 carried the same IS, ISKpn26, inserted at the same chromosomal position, suggesting a recombination event at this location. A phylogeny of CrrB is largely congruent with the SNV-based whole-genome phylogeny (Fig. 2B).

Both UHKPC26 and UHKPC28 had significantly elevated expression of an adjacent conserved hypothetical protein (H239_3059), a putative glycosyltransferase with a TupA-like ATP grasp domain (PF14305), which is predicted to be involved in surface polysaccharide biosynthesis, and an adjacent uncharacterized conserved membrane protein (H239_3062). These strains also had elevated expression of both the pmrH and pmrCAB operons, which encode the Ara4N and pETN pathways of LPS modification. Our working hypothesis is that CrrAB induces expression of the glycosyltransferase-like protein which transfers an as-yet-unidentified sugar to lipid A phosphate in a manner analogous to PmrC (Fig. 1B). The basis for the upregulation of pmrCAB has not been established. It is possible that CrrB, the novel TCRS histidine kinase, directly phosphorylates PmrA or that a yet-to-be-identified connector protein, perhaps the conserved hypothetical membrane protein (H239_3062), facilitates feedback between the two systems.

UHKPC26 and UHKPC28 do carry other single-nucleotide variants (SNVs) compared with Col^s strains (see Table S1 in the supplemental material). RstA has been shown to be induced by PhoPQ in other Gram-negative pathogens (32, 43); however, the variant seen in UHKPC28 is unlikely to have contributed to the Col^r phenotype as neither rstAB nor phoPQ expression was altered in UHKPC28. Furthermore, the UHKPC26 sequence was identical to Col^s strains across this region, and the expression profiles are similar between the two strains. An additional variant was observed in a transmembrane region of WecA, which is involved in O-antigen biosynthesis, in strain UHKPC28 but is not predicted to be a key residue (44–46). However, the UHKPC26 wecA RPKM value was significantly lower than that in Col^s strains even though there was no mutation in this gene. An examination of a potential role for WecA in LPS modification and a better characterization of what regulates its expression is needed to better understand any potential role in Col^r.

Other transcription changes common to Col^r strains included increased expression of cation transporters and other efflux pumps, which could be a response to altered membrane charge and difficulty in transporting cations through a LPS layer with a more neutral charge. Other genomic changes include the loss of bla_KPC plasmids in UHKPC52 and UHKPC179 which were present in a closely related strain in the case of UHKPC52, or isogenic paired strain UHKPC57 in the case of UHKPC179. The loss of these plasmids suggests that there may be a fitness cost associated with acquiring colistin resistance that is partially mitigated by the plasmid loss, and this should be more rigorously explored with formalized fitness comparisons.

This study identified transcriptional changes of Col^r strains during growth without colistin. Further experiments to examine how colistin exposure alters gene expression may provide additional insight into the resistance mechanism(s). Analysis of LPS and lipid A structures may identify novel modifications in UHKPC26 and UHKPC28. In addition, formal analysis of the fitness cost of resistance along with follow-up experiments to examine how the observed LPS modifications alter K. pneumoniae host interaction is important to understand the impact of colistin resistance on virulence.

Conclusions.

The combination of genomic and transcriptomic analysis revealed three genetic mechanisms conferring colistin resistance with distinct global gene expression profiles depending on the genetic mechanism. Interestingly, we found that each strain possessed unique mutations predicted to confer colistin resistance. Although the sample collection was not designed to be comprehensive or represent a formal epidemiological study, the absence of repeat observations of specific mutations suggests that independently derived Col^r strains of K. pneumoniae may be more common than patient-to-patient spread of resistant strains. Moreover, the constellation of changes that can lead to a “final common phenotype” will challenge the future development of cationic antimicrobial peptides and molecular diagnostics, as targets may vary across strains. The fact that mutations were found in genes without E. coli orthologs highlights the importance of studying colistin resistance across the range of clinically significant pathogens.