Structural modification of LPS in colistin-resistant, KPC-producing Klebsiella pneumoniae

Abstract

Background

Colistin resistance in Klebsiella pneumoniae typically involves inactivation or mutations of chromosomal genes mgrB, pmrAB or phoPQ, but data regarding consequent modifications of LPS are limited.

Objectives

To examine the sequences of chromosomal loci implicated in colistin resistance and the respective LPS-derived lipid A profiles using 11 pairs of colistin-susceptible and -resistant KPC-producing K. pneumoniae clinical strains.

Methods

The strains were subjected to high-throughput sequencing with Illumina HiSeq. The mgrB gene was amplified by PCR and sequenced. Lipid profiles were determined using MALDI-TOF MS.

Results

All patients were treated with colistimethate prior to the isolation of colistin-resistant strains (MIC >2 mg/L). Seven of 11 colistin-resistant strains had deletion or insertional inactivation of mgrB. Three strains, including one with an mgrB deletion, had non-synonymous pmrB mutations associated with colistin resistance. When analysed by MALDI-TOF MS, all colistin-resistant strains generated mass spectra containing ions at m/z 1955 and 1971, consistent with addition of 4-amino-4-deoxy-l-arabinose (Ara4N) to lipid A, whereas only one of the susceptible strains displayed this lipid A phenotype.

Conclusions

The pathway to colistin resistance in K. pneumoniae primarily involves lipid A modification with Ara4N in clinical settings.

Introduction

Klebsiella pneumoniae producing KPC-type carbapenemase is the most common and clinically problematic carbapenem-resistant Enterobacteriaceae (CRE) in healthcare institutions in the USA.^1,2 Infections due to KPC-producing K. pneumoniae are difficult to manage and often require the use of colistin, which is a decades-old cyclic polypeptide that is active against most K. pneumoniae strains.³ Colistin exerts its activity by binding the bacterial membrane through electrostatic interactions with the lipid A moiety of LPS, causing disruption of the outer membrane and cell death.⁴ Bacteria can develop resistance to colistin by modifying the structures of the lipid A moiety, hindering colistin binding.⁵ In K. pneumoniae, the LPS structure has been characterized for the spontaneous colistin-resistant strain OM-5.⁶ In strain OM-5, lipid A was mostly modified with 4-amino-4-deoxy-l-arabinopyranose (Ara4N), whereas little or no modification was observed in its colistin-susceptible parental strain, suggesting this as a primary mechanism of resistance in this organism. Other studies have also implicated palmitoylation⁷ and/or hydroxylation⁸ as conferring resistance in K. pneumoniae to antimicrobial peptides, including colistin.

The increasing clinical use of colistin has led to development of colistin resistance in K. pneumoniae, and outbreaks due to colistin-resistant KPC-producing K. pneumoniae have been reported.^9,10 Clinically, the most commonly reported genetic pathway for colistin resistance has involved loss-of-function mutations of mgrB, encoding a negative regulator of PhoPQ, which in turn modulates PmrAB, leading to activation of the pmrHFIJKLM operon.^10–12 Using a pair of colistin-susceptible and -resistant clinical strains identified from the same patient, a non-synonymous mutation in the gene encoding PmrB, the sensor kinase component of the PmrAB two-component regulatory system, has been associated with up-regulation of pmrA and pmrHFIJKLM, the latter encoding an LPS modification system that is presumed to facilitate biosynthesis and lipid A transfer of Ara4N.^13,14 The potential role of another two-component regulatory system, CrrAB, has also been suggested.¹⁴ However, little information is available regarding the lipid A modifications that occur in response to mutations or modulation of these genes in KPC-producing K. pneumoniae. The aim of the present study was to investigate mutations of mgrB, pmrAB, phoPQ, crrAB and pmrHFIJKLM with Ara4N diesterization of the lipid A phosphate, as determined by MS analysis, that characterize colistin resistance using multiple pairs of colistin-susceptible and -resistant KPC-producing K. pneumoniae strains identified within patients prior to and following treatment with colistimethate.

Methods

Bacterial strains

Twenty-two carbapenem-resistant K. pneumoniae clinical strains were collected at the University of Pittsburgh Medical Center between 2008 and 2012 under waiver of informed consent (University of Pittsburgh IRB # 0510165 and PRO12060302) and tested for colistin susceptibility using the standard agar dilution method. Strains with MICs ≥4 mg/L were considered resistant to colistin.¹⁵ Strains from the same patients that demonstrated discordant colistin susceptibility (susceptible and resistant) were included in the study.

PCR and sequencing of the mgrB gene

The mgrB gene was amplified and sequenced by the Sanger method for all study strains using the primer set reported previously¹⁶ to identify clinical strains with full or partial deletion of the gene, the inactivation of which reportedly accounts for the majority of colistin-resistant K. pneumoniae strains worldwide.^12,16

High-throughput sequencing

The genomes of all study strains were sequenced on a HiSeq 2500 (Illumina, San Diego, CA, USA). In addition, strain C2 was sequenced using a PacBio RS II (Pacific Biosciences, Menlo Park, CA, USA) to provide the internal reference genome for the analysis, which yielded a total of 150292 reads with an average read length of 9408 bp. De novo assembly of the C2 PacBio reads using the hierarchical genome assembly process (HGAP) available in the SMRT Analysis v2.1 software generated six contigs with ×84 coverage. The assembly was annotated using Prokka. The reads were aligned to reference genome C2 using BWA-MEM (http://bio-bwa.sourceforge.net/). Duplicate reads were marked using Picard MarkDuplicates. SNPs were identified using GATK HaplotypeCaller with ploidy of 1.¹⁷ SNPs with low mapping quality (RMSMappingQuality<20), evidence of strand bias (FisherStrand >60.0), low variant confidence (QualByDepth <2), only seen near the ends of the reads (ReadPosRankSum <−8.0), or low depth (Depth <5) were filtered and removed using GATK VariantFiltration. The ST was determined by uploading contigs assembled de novo by CLC Genomics Workbench version 7.5.1 using the default settings to the K. pneumoniae MLST website (http://bigsdb.web.pasteur.fr/klebsiella/klebsiella.html). Assembled genomes were submitted to GenBank under accession numbers SAMN06445922 to 06445943.

Lipid A isolation from whole cells

Membrane lipids were extracted and LPS was converted to lipid A by an optimized small-scale hot ammonium isobutyrate-based protocol.¹⁸ Bacteria were streaked onto lysogeny broth agar plates to obtain pure colonies. Overnight cultures (1–5 mL at 37°C) in lysogeny broth with or without 2 mg/L colistin sulphate were harvested at 4000 g for 10 min. Bacterial pellets were treated with a 5:3 mixture of 70% (v/v) isobutyric acid/1 M ammonium hydroxide (250 μL/150 μL) and incubated at 100°C for 30–45 min. Reactions were spun down at 2000 g for 15 min to remove cell debris, and supernatants were transferred to clean tubes, combined in a 1:1 ratio of distilled water (400 μL), frozen on dry ice, and lyophilized overnight. The resulting dry pellets contain whole-cell extracts of membrane lipids.

Lipid A characterization by mass spectrometry

Dry lipid extracts were washed twice with 1 mL of methanol and then resuspended in 200 μL of a 2:1:0.25 chloroform/methanol/water solvent mixture. Aliquots of 1 μL each of norharmane matrix (10 mg/mL in 2:1 v/v chloroform/methanol) then analyte were spotted directly onto stainless steel target plates. Mass spectra were recorded in negative ion mode using a Bruker microflex LRF MALDI-TOF mass spectrometer (Bruker Daltonics Inc., Billerica, MA, USA) operated in reflectron mode. The instrument was equipped with a 337 nm nitrogen laser, and analyses were performed at 39.5% global intensity. Typically, 900 laser shots were summed to acquire each spectrum. Data were acquired and processed using flexControl and flexAnalysis version 3.4 (Bruker Daltonics Inc.).

Structural analysis was carried out on a Waters Synapt G2 Q-TOF mass spectrometer (Waters Corporation, Milford, MA, USA) operated in sensitivity mode. Lipid extract resuspensions were diluted 2:1 in a solvent mixture of chloroform/methanol (v/v). The solution was infused at 3 μL/min flow rate. The source block temperature was set to 150°C. Tandem MS was carried out using trap CID. For mass selection in the quadrupole, we used the instrument standard values (low mass resolution 4.7 and high mass resolution 15.0). To mitigate the effect of instantaneous signal fluctuations, the data were averaged for 3 min.

Results

Characteristics of clinical cases associated with colistin-resistant K. pneumoniae

Both colistin-susceptible and -resistant K. pneumoniae strains were identified during the colistin treatment span in 11 patients between 2008 and 2012. All except one of them were solid organ transplant recipients, and nine of them were in an ICU at the time of the culture (Table 1). While the time between isolation of the susceptible and resistant strains varied (range, 6–101 days), all had received treatment with intravenous colistin methanesulfonate (the prodrug of colistin) and three had received prolonged inhaled colistin methanesulfonate prior to the isolation of colistin-resistant K. pneumoniae.

Table 1.

Characteristics and outcomes of patients with carbapenem- and colistin-resistant K. pneumoniae

Patient	Colistin- resistant strain	Date (dd/mm/yy)	Days from the susceptible strain	Age (years)	Sex	Underlying disease	Culture site	Type of infection	ICU	Prior intravenous CMS (days)	Prior inhaled CMS (days)	Treatment of colistin-resistant infection	30 day mortalitya
1	A5	25/02/08	31	49	male	multivisceral transplant	BAL	VAP, BSI	yes	23	34	AMK, TGC	no
2	B6	10/03/10	47	63	male	liver transplant	urine	asymptomatic bacteriuria	no	2	–	none	no
3	B9	13/04/10	24	25	female	multivisceral transplant	abdominal clot	intra-abdominal abscess	yes	18	–	CMS, TZP, TGC	no
4	C3	17/04/10	14	42	male	liver transplant	blood	intra-abdominal abscess	yes	12	–	CMS, DOR, TGC, MTZ	no
5	C5	01/06/10	49	67	female	lung transplant	bronchial washing	acute cellular rejection	no	5	24	CMS, DOR, then DOX, MEM (inh)	no
6	D7	24/06/11	49	47	male	heart/lung transplant	bronchial washing	VAP	yes	2	43	CMS, DOR, ETP, TOB (inh)	no
7	E6	12/01/12	21	60	male	liver transplant	BAL	tracheal colonization	yes	19	–	none	no
8	F3	13/03/12	11	60	male	necrotizing pancreatitis	blood	intra-abdominal abscess	yes	8	–	CMS, DOR, GEN	yes
9	F9	30/03/12	6	63	female	liver transplant	blood	intra-abdominal abscess, BSI	yes	8	–	CMS, DOR, ETP	no
10	H5	12/07/12	101	37	female	liver transplant	liver tissue	intra-abdominal abscess	yes	5	–	CMS, DOR, FEP	no
11	I2	19/11/12	40	58	female	liver transplant	BAL	intra-abdominal abscess	yes	38	–	CMS, DOR, TGC	no

BAL, bronchoalveolar lavage, VAP, ventilator-associated pneumonia; BSI, bloodstream infection; CMS, colistin methanesulfonate; AMK, amikacin; TGC, tigecycline; TZP, piperacillin/tazobactam; MTZ, metronidazole; DOR, doripenem; DOX, doxycycline; MEM, meropenem; inh, inhaled; ETP, ertapenem; TOB, tobramycin; GEN, gentamicin; FEP, cefepime.

^aDeath within 30 days from the day the culture growing the colistin-resistant strain was collected.

Overview of the colistin-susceptible and -resistant K. pneumoniae strains

For 9 of the 11 patients, both the colistin-susceptible and -resistant strains belonged to the epidemic ST258 (Table 2).¹⁹ A phylogenetic tree based on 2718 SNPs identified by the GATK method was generated using RAxML v8.2.9 by running 100 bootstrap replicates under the generalized time-reversible model (GTRCAT) and Lewis correction for ascertainment bias, which supported within-patient development of colistin resistance upon exposure to colistin methansulfonate for all patients except one (Figure S1, available as Supplementary data at JAC Online). Both the susceptible and resistant strains from the first patient identified belonged to ST17 and were also closely related on the expanded phylogenetic tree (data not shown). A bla_KPC-2-carrying K. pneumoniae ST17 strain has been reported from Greece previously.²⁰ The likely exception to within-patient evolution of colistin resistance was Patient 7, who initially had a colistin-susceptible ST258 strain but then a colistin-resistant ST37 strain 21 days later, which suggests either reinfection with the latter strain or an initial coinfection with ST258 and ST37 strains that was undetected at the time.

Table 2.

Colistin MICs, lipid A modification and mgrB, pmrAB and phoPQ loci

Patient	Strain	ST	KPC type	Colistin MIC (mg/L)	Extra m/z consistent with Ara4N	mgrB	crrB	pmrA	pmrB	pmrF	pmrI	pmrJ	pmrK	phoP	phoQ
1	A2	ST17	KPC-3	0.5		WT
	A5	ST17	KPC-3	>256	1955/1971a	Gln30Arg	Leu94Met		His340Arg
2	B5	ST258	KPC-2	0.25		WT
	B6	ST258	KPC-2	>256	1955/1971	nt75insISKpn26-like	Pro151Gln			Phe280Leu
3	B8	ST258	KPC-2	0.25		WT
	B9	ST258	KPC-2	>256	1955/1971	nt19frameshift
4	C2	ST258	KPC-2	0.25	1955/1971	WT
	C3	ST258	KPC-2	128	1955/1971	WT			Ser85Arg
5	C4	ST258	KPC-2	0.5		WT
	C5	ST258	KPC-2	128	1955/1971	Gln30stop
6	D4	ST258	KPC-2	0.5		WT
	D7	ST258	KPC-2	256	1955/1971a	WT
7	E5	ST258	KPC-2	0.25		WT
	E6	ST37	KPC-2	128	1955/1971a	nt70insIS903B-like	Leu296Gln			Lys322Gln		Glu25Ala	Ile117Val
												Arg29Lys	His156Gln
												Ile53Val	Asp441Glu
												Leu94Ile
8	F2	ST258	KPC-2	0.25		WT
	F3	ST258	KPC-2	128	1955/1971	not detected by PCR or WGS			Thr157Pro
9	F8	ST258	KPC-2	0.5		WT
	F9	ST258	KPC-2	128	1955/1971a	nt75insISKpn26-like
10	H4	ST258	KPC-2	0.5		WT
	H5	ST258	KPC-2	64	1955/1971	WT
11	I1	ST258	KPC-2	0.25		WT
	I2	ST258	KPC-2	>256	1955/1971	not detected by PCR or WGS	Leu133Arg

The sequences of the colistin-resistant strains were compared with those of the colistin-susceptible strains from the same patients.

^aMS profiles detected in antibiotic medium.

Genetic changes consistent with colistin resistance

The mgrB gene encodes a regulatory transmembrane protein that negatively regulates the PhoPQ signalling system, which in turn activates the PmrAB signalling system leading to LPS modification and colistin resistance.^12,16 Major loss-of-function mutations in mgrB (premature stop codon, insertional inactivation, or deletion) were detected in 7 of the 11 colistin-resistant strains (Table 2). In one of these strains (F3), a non-synonymous mutation in pmrB that has been associated with colistin resistance (T157P) was also detected.²¹ Of the four colistin-resistant strains with intact mgrB, two strains (A5 and C3) contained non-synonymous mutations in pmrB (S85R and H340R). The S85R substitution has been previously associated with colistin resistance in K. pneumoniae,¹¹ whereas the H340R substitution has been associated with resistance in Pseudomonas aeruginosa.²² Two colistin-resistant strains (D7 from Patient 6 and H5 from Patient 10) did not contain any non-synonymous mutations in any of the genetic loci that have been implicated in colistin resistance in K. pneumoniae (mgrB, crrAB, pmrAB, phoPQ, pmrHFIJKLM). The pairwise comparison results are shown in Table S1.

Identification of Ara4N addition to lipid A isolated from K. pneumoniae LPS

All strains were subjected to MS analysis of extracted membrane lipids to characterize lipid A modifications associated with the observed genetic changes. The strains showed ions at m/z 1824 and 1840 (Figure 1) that represent a bisphosphorylated, hexa-acylated lipid A molecule either with or without a hydroxylation, respectively, on the C′-2 fatty acyl chain. Most strains, excluding B9, D4, D7 and E6, also showed ions at m/z 2063 and 2079, which result from a palmitoylation at the C-1 acyl-oxo-acyl position of the structures at m/z 1824 and 1840. These findings are consistent with previous studies,⁶ and lipid A structures and the corresponding m/z values of these signature ions can be found in Figure 2. Mass spectra of colistin-resistant K. pneumoniae strains are characterized by the presence of ions at m/z 1955 and 1971, a mass shift of m/z 131 caused by addition of Ara4N to the hexa-acylated lipid A structures at m/z 1824 and 1840 (Figure 1). The addition of Ara4N was confirmed using tandem MS. Collision-induced dissociation of m/z 1955 and 1971 resulted in loss of the neutral mass of Ara4N (131 mass units), and m/z 1824 and 1840 were the main product ions, respectively (Figures S2–S4). In general, these mass spectra correlated well with observed disruptions to mgrB, except for colistin-susceptible strain C2 and resistant strains C3, D7 and H5, which possess the WT form of the gene. With the exception of strain C2, these also correlated with resistance to colistin (Table 2).

Figure 1.

MALDI-TOF MS of K. pneumoniae with differential colistin susceptibility. Within-patient strains B5 and B6 were grown overnight in liquid culture at 37 °C, and lipids were extracted and analysed by MALDI-TOF MS. (a) Colistin-susceptible B5 shows a mass ion at m/z 1824 corresponding to a base lipid A structure that also exists with an additional hydroxylation (m/z 1840), palmitoylation (m/z 2063), or hydroxylation and palmitoylation (m/z 2079). (b) Colistin-resistant B6 shows m/z 1840 as the base peak with additional ions at m/z 1955 and 1971, indicating an Ara4N addition to the base structures at m/z 1824 and 1840. (c and d) Molecular structures of the lipid A molecules found in mass spectra from resistant isolates.

Figure 2.

Lipid A structures with corresponding m/z values found in clinical isolates. Lipid A m/z values and molecular structures found in mass spectra of the K. pneumoniae clinical isolates are shown, with descriptions of the modifications responsible for the observed mass shifts. Asterisks denote ions associated with colistin resistance.

Discussion

The rapid dissemination of CRE has resulted in hard-to-treat infections and high mortality rates. Facing a shortage of effective antimicrobials, clinicians have turned to colistin, a last-line antibiotic that has now become a valuable clinical treatment option.³ Increased colistin consumption has led to the emergence of colistin-resistant Gram-negative Enterobacteriaceae and Acinetobacter baumannii. In K. pneumoniae, lipid A is modified with the addition of Ara4N by the pmrHFIJKLM operon and under the control of phoPQ and pmrAB.¹⁴ Colistin resistance is associated with mutations to phoPQ and pmrAB or the negative regulator mgrB, resulting in the up-regulation of this LPS modification system.^13,16

Here we present a study of colistin-susceptible and -resistant pairs of K. pneumoniae clinical strains collected from 11 patients. We examined only strains obtained from patients with independently acquired infections in which evolution of resistance occurred as a response to colistin methanesulfonate treatment. We found that 7 of 11 resistant strains possessed loss-of-function mgrB mutations, consistent with previous findings.^{10–12,16,23} While there is strong association between mutations in specific genetic loci regulating lipid A synthesis and modifications to lipid A in susceptible versus resistant strains, different locations or types of loss-of-function mutations did not produce significantly different MICs or MS profiles of lipid A (Table 2). Interestingly, resistant strains A5, D7, E6 and F9 only displayed the Ara4N peaks following culture in a sub-lethal colistin concentration, indicating that there is regulation of LPS modification. The findings in this study are congruent with previous work demonstrating high variability in resistance profiles and in which strains incur some fitness cost of resistance yet resistance persists in the absence of selection.^7,24

This work demonstrated that genetic alterations previously implicated in colistin resistance of K. pneumoniae displayed a corresponding LPS modification as demonstrated by an altered MS profile (Figures 1 and 2). Notably, strains C2, D7 and H5 shared this same MS profile, but contained no non-synonymous mutations of the aforementioned genes. Park et al.²⁵ similarly identified a colistin-resistant A. baumannii with WT pmrAB. Furthermore, strain C2 was found to be susceptible to colistin upon repeated susceptibility testing. Clearly there may be additional genes involved in regulation or modification of LPS, as was previously posited.^5,11 Other factors may contribute to this phenomenon, including heterogeneity or hetero-resistance between cells within a strain (MIC measurements and MS phenotyping were conducted on the same cultures from single colonies), which has been observed in A. baumannii²⁶ and P. aeruginosa,²⁷ or emergence of cross-resistance due to contact with host antimicrobial peptides.

Finally, the discovery of a colistin resistance mechanism mediated by a plasmid-encoded mobile colistin resistance (MCR) phosphoethanolamine transferase should be addressed: effective colistin resistance monitoring must include this novel mechanism.²⁸ The current study links alterations in the bacterial genome with a physiological response in these resistant organisms and lends further understanding to the role these mutations play in resistance acquisition, and this could certainly be applied to the study of the impact of MCR-conferred, colistin-resistant organisms. We have also recently shown that mcr-1-containing plasmids expressed in colistin-resistant K. pneumoniae strains gave an altered resistance ion profile in which phosphoethanolamine is added to lipid A, overriding Ara4N modification.²⁹ Furthermore, our determination of structural differences between resistant and susceptible strains elucidated by MS in both studies suggests the potential of this platform as a diagnostic to improve antibiotic stewardship of colistin-resistant organisms.

Nucleotide accession numbers

The sequence reads were deposited in the sequence read archive under BioProject PRJNA375812. De novo contigs were assigned GenBank accession numbers SAMN06445922 to 06445943.