Emergence of High-Level Colistin Resistance in an Acinetobacter baumannii Clinical Isolate Mediated by Inactivation of the Global Regulator H-NS

ABSTRACT

Colistin is a crucial last-line drug used for the treatment of life-threatening infections caused by multidrug-resistant strains of the Gram-negative bacterium Acinetobacter baumannii. However, colistin-resistant A. baumannii isolates can still be isolated following failed colistin therapy. Resistance is most often mediated by the addition of phosphoethanolamine (pEtN) to lipid A by PmrC, following missense mutations in the pmrCAB operon encoding PmrC and the two-component signal transduction system PmrA/PmrB. We recovered a pair of A. baumannii isolates from a single patient before (6009-1) and after (6009-2) failed colistin treatment. These strains displayed low and very high levels of colistin resistance (MICs, 8 to 16 μg/ml and 128 μg/ml), respectively. To understand how increased colistin resistance arose, we sequenced the genome of each isolate, which revealed that 6009-2 had an extra copy of the insertion sequence element ISAba125 within a gene encoding an H-NS family transcriptional regulator. To confirm the role of H-NS in colistin resistance, we generated an hns deletion mutant in 6009-1 and showed that colistin resistance increased upon the deletion of hns. We also provided 6009-2 with an intact copy of hns and showed that the strain was no longer resistant to high concentrations of colistin. Transcriptomic analysis of the clinical isolates identified more than 150 genes as being differentially expressed in the colistin-resistant hns mutant 6009-2. Importantly, the expression of eptA, encoding a second lipid A-specific pEtN transferase but not pmrC, was increased in the hns mutant. This is the first time an H-NS family transcriptional regulator has been associated with a pEtN transferase and colistin resistance.

INTRODUCTION

Acinetobacter baumannii is a Gram-negative bacterial pathogen of significant medical importance (1). It causes a range of infections primarily in high-dependency patients in intensive care units, including ventilator-associated pneumonia, bloodstream infections, and urinary tract infections (2). While A. baumannii infections have historically been treated with carbapenem antibiotics, many isolates are now highly resistant to these and other commonly used antibiotics (1). As such, A. baumannii has been categorized as one of the top six priority dangerous organisms by the Infectious Diseases Society of America (3) and as a priority one organism for research and development of novel antibiotics by the World Health Organization (4).

To combat multidrug-resistant A. baumannii strains, colistin has been reintroduced as a therapeutic for the treatment of infections. In some cases, colistin is the only effective remaining treatment option, and it is therefore of major concern that resistance to this antibiotic has been observed in A. baumannii. The study of colistin pharmacology and colistin resistance mechanisms in A. baumannii is now imperative to address the potential public health threat posed by this organism.

Colistin targets the Gram-negative bacterial outer membrane via an initial charge-based interaction between the positively charged colistin and the negatively charged phosphate groups on the lipid A component of lipopolysaccharide (LPS). Three mechanisms of colistin resistance (MIC, >2 μg/ml) have been observed in A. baumannii, and all involve structural changes to lipid A, the membrane-anchoring component of the LPS molecule. First, very high-level colistin resistance (MIC, >128 μg/ml) can result from a complete loss of lipid A, resulting in an LPS-deficient outer membrane with a reduced negative charge (5). This mechanism of colistin resistance results from spontaneous mutations within one of the lipid A biosynthesis genes lpxA, lpxC, or lpxD; these mutations include point mutations, small deletions or insertions, or gene inactivation via transposon or insertion sequence element (IS) movement (5, 6). Second, high levels of colistin resistance (MIC, >32 μg/ml) can result from the addition of phosphoethanolamine (pEtN) to one or two positions on the lipid A phosphates, reducing the negative charge of the LPS and therefore the interaction with colistin. pEtN-mediated colistin resistance usually results from mutations in the pmrCAB operon that lead to constitutive activation of the PmrAB two-component signal transduction system, which in turn leads to the increased expression of the lipid A-specific pEtN transferase PmrC (7–9). pEtN-mediated resistance is the most common resistance mechanism observed in A. baumannii clinical isolates. Third, resistance can also occur via the addition of galactosamine to the lipid A phosphates, leading to a reduced negative surface charge and therefore reduced interaction with the positively charged colistin molecule. While the PmrB-regulated NaxD deacetylase has been identified as being essential for this galactosamine addition to lipid A (10), no galactosamine transferase has yet been identified in A. baumannii.

In this study, we report a novel colistin resistance mechanism of direct relevance to colistin treatment failure. We analyzed a pair of A. baumannii isolates obtained from a patient prior to colistin treatment (strain 6009-1; colistin broth MIC, 8 to 16 μg/ml) and from the same patient after 11 days of colistin therapy (strain 6009-2; broth MIC, 128 μg/ml). Comparative sequence analysis of the two isolates identified that the highly colistin-resistant strain had an extra copy of ISAba125 within a gene encoding an H-NS family transcriptional regulator. Furthermore, inactivation of this H-NS was shown to alter the expression of more than 150 genes, including a pEtN transferase gene. This is the first time that an H-NS family transcriptional regulator has been associated with the regulation of a lipid A-specific pEtN transferase and the first demonstration that colistin-resistant hns mutants can arise in the clinical setting.

RESULTS

Isolation and genome sequencing of a pair of A. baumannii MDR clinical isolates from the same patient.

To identify colistin resistance mechanisms directly relevant to colistin treatment failure, we analyzed a pair of multidrug-resistant (MDR) A. baumannii isolates recovered from a single patient before and after failed colistin treatment. The patient, a 98-year-old man, presented with severe respiratory distress and sepsis and was placed on mechanical ventilation and an intravenous (i.v.) catheter. Screening of bronchial secretions failed to identify any infection, and the patient was initially treated with amoxicillin and clavulanic acid (3,600 mg) and clarithromycin (1,000 mg) for 2 days. Subsequent testing of the bronchial secretions resulted in the isolation of A. baumannii (6009-1), which, according to Vitek-2 and Etest (confirmatory) methods was sensitive to tobramycin, amikacin, colistin, minocycline, and isepamicin, had intermediate resistance to gentamicin and rifampin, and was resistant to all other classes of antibiotics except tigecycline, which was only tested by Etest. The patient was treated over a period of 33 days (including the initial antibiotic treatment prior to the isolation of 6009-1) with meropenem (2,000 mg) for 18 days, erythromycin (500 mg) for 3 days during that same period, vancomycin (500 mg) for 25 days, colistin (2.5 million international units [MIU]) for 18 days, and erythromycin (375 mg) again for 4 days. Strain 6009-2 was isolated from bronchial secretions following 11 days of colistin treatment and according to Etest was susceptible to tobramycin, minocycline, and isepamicin, had intermediate resistance to gentamicin, rifampin, and tigecycline, and was resistant to all others, including colistin. Subsequent testing showed that the initial 6009-1 isolate displayed a low/intermediate level of resistance to colistin by both Etest (4 μg/ml) and broth microdilution (BMD) (8 to 16 μg/ml) (Table 1), and strain 6009-2 displayed significantly increased colistin MICs of 64 μg/ml (Etest) and 128 μg/ml (BMD) (Table 1). The colistin Etest is often unreliable for determining colistin resistance in Gram-negative bacilli and is known to produce false-negative colistin MICs in some colistin-resistant strains (11), which may explain the variation in results between Etest and BMD MICs and why strain 6009-1 was initially considered susceptible to colistin. All antibiotic treatments failed to clear the infection, and the patient eventually died.

TABLE 1.

Colistin MICs of the A. baumannii strains used in this study, as determined using the BMD method

A. baumannii strain	MIC (μg/ml)
6009-1	8–16
6009-2	128
6009-2/pWH1266::hns	16–32
6009-2/pWH1266	64
6009-1/pWH1266::eptA	64
6009-1/pWH1266	8
6009-1 Δhns (directed hns deletion mutant)	128

The colistin resistance of strain 6009-2 was stable even in the absence of colistin selection (data not shown), suggesting that a specific mutation had occurred. Carbohydrate silver-stained PAGE gels containing whole-cell lysates of 6009-1 and 6009-2 revealed that the two strains produced identical LPS profiles (data not shown), indicating that the colistin-resistant phenotype displayed by 6009-2 was not due to the total loss of LPS.

To identify the genetic differences between 6009-1 and 6009-2, we determined draft genome sequences for both strains. As expected, the two strains had nearly identical genomes (see below) and were therefore a true isogenic pair. Phylogenetic analysis of the two 6009 genomes (6009-1 and 6009-2), together with 69 other A. baumannii genomes, indicated that 6009-1 and 6009-2 belonged to the global clone 2 (GC2) lineage (Fig. 1) and were closely related to the multidrug-resistant A. baumannii strain ACICU, which was isolated from a patient in Italy (12, 13), as well as two U.S. isolates, OIFC338 and UH41_448. Most commonly studied laboratory strains fall into GC1 (AB0057 and AYE) or GC3 (ATCC 19606 and ATCC 17978), and as such, 6009-1 and 6009-2 are quite divergent from these strains. Our comparison of the genomes identified a small number of regions that were unique to the 6009 genome, as well as a small number of regions that were present in the ACICU genome but absent (or with very low shared identity with) from the 6009 genome (see Fig. S1 in the supplemental material).

FIG 1.

Phylogenetic relationship of the 6009 clinical isolate to other A. baumannii strains based on conserved genomic region SNPs. Phylogeny was determined using ParSNP version 1.2, visualized and analyzed in Gingr version 1.1.1, and the final tree was rendered and manipulated in FigTree version 1.4.2. The colored outer arcs designate strains that are closely related to each other, with the GC1, GC2, and GC3 lineages designated in orange, pink, and blue, respectively.

The highly colistin-resistant strain 6009-2 has a novel ISAba125 insertion in the gene encoding the global transcriptional regulator H-NS.

To identify the genetic mechanism for the increased colistin resistance displayed by 6009-2, the genome sequences of 6009-1 and 6009-2 were compared. Only one synonymous point mutation was identified within a plasmid-carried gene that is homologous to ACICU_p0039 (as the 6009 clinical isolates are closely related to strain ACICU, the ACICU gene names are used here for simplicity). As the encoded amino acid sequence remained unchanged, we concluded that this change was unlikely to lead to increased colistin resistance. Further analyses of read depth differences between the two genomes indicated that there was a difference in the copy number of the insertion sequence ISAba125. This was confirmed by Southern hybridization using an ISAba125-specific probe, which revealed that strain 6009-2 had one extra hybridizing band in the DNA profile compared to 6009-1, corresponding to an additional insertion of the ISAba125 (data not shown). To identify all the positions of the ISAba125 element in each genome, we analyzed all sequence reads that contained either the left or right junction of ISAba125 and identified the adjoining A. baumannii-specific DNA sequence. Using this strategy, we identified that the additional copy of the ISAba125 element in the colistin-resistant strain 6009-2 was located at the 3′ end of a gene homologous to ACICU_00289 and whose product shared significant identity (99 to 100%) to a predicted protein containing an H-NS/histone family domain (Pfam PF00816, E value = 9 × 10⁻¹⁹) present in all sequenced A. baumannii genomes, including that of A. baumannii strain 17978 (14) (Fig. 2A). This unique site of ISAba125 insertion into the 6009-2 genome and the lack of an IS element in the same position in 6009-1 were confirmed by PCR (data not shown) and by Sanger sequencing using the 6009-2 genomic DNA as the template together with an outward-firing ISAba125-specific primer. Sequencing revealed that the DNA inserted into hns was composed of a partial (45-bp) ISAba125, followed by an intact copy ISAba125. The ISAba125 insertion, located 8 bp from the 3′ end of ACICU_00289, was predicted to result in a C-terminal fusion of the ACICU_00289 protein with an open reading frame present at the start of the IS element, resulting in a predicted change to the last two amino acids of H-NS and an extension of H-NS for a further three amino acids (Fig. 2B). Alignment of ACICU_00289 with known H-NS proteins from a wide range of bacteria showed that it contained a number of highly conserved H-NS residues (Fig. S2). H-NS proteins are global transcriptional regulators that have an N-terminal protein-protein interaction domain and C-terminal DNA binding domain and predominantly bind to genomic regions with high A+T content (15). They are often involved in the silencing of gene expression of horizontally acquired DNA (15). Interestingly, both the 6009 strains and the ACICU strain also have a second plasmid-encoded H-NS homolog that shares 71% identity (81% positive) with the genome-encoded H-NS ACICU_00289 (Fig. S2).

FIG 2.

The novel ISAba125 insertion in hns in strain 6009-2 and its effect on the amino acid sequence of the H-NS protein. (A) Schematic showing hns (ACICU_00289) and the surrounding region in the pretreatment clinical isolate 6009-1; below, the position of the ISAba125 element (apex of triangle) at the 3′ end of the hns gene contained in the highly colistin-resistant strain 6009-2 that was isolated from the same patient posttreatment. The direction of transcription (arrows) and the predicted function of each gene product (above gene) are shown. The equivalent ACICU locus tags are shown within each arrow. The ISAba125 element is composed of an intact ISAba125 element (open arrow, direction indicates orientation), as well as a 45-bp truncated region of ISAba125 (black filled box). (B) Amino acid sequence alignment of the predicted products from the hns gene in strain ACICU (ACICU_00289) and from strains 6009-1 and 6009-2. Altered residues predicted to be present at the C-terminal end of the mutated H-NS protein in 6009-2 are shown in bold and underlined.

Providing the hns mutant 6009-2 with an intact copy of hns decreased colistin resistance.

To confirm that the increased colistin resistance observed in A. baumannii strain 6009-2 was due to the insertion of ISAba125 into hns, the colistin-resistant clinical isolate 6009-2 was complemented with an intact version of hns on the replicating plasmid pWH1266, generating the strain 6009-2/pWH1266::hns. As a control, 6009-2 was separately transformed with the empty vector plasmid, generating the strain 6009-2/pWH1266. MIC assays revealed that the 6009-2 strain provided with an intact copy of hns showed significantly reduced colistin resistance (MIC, 16 to 32 μg/ml, Table 1) compared to 6009-2 alone and 6009-2 with empty vector (colistin MICs, 128 μg/ml and 64 μg/ml, respectively; Table 1). To verify the role of hns inactivation in colistin resistance, we deleted hns in strain 6009-1 using site-directed allelic exchange mutagenesis, generating strain 6009-1 Δhns. This hns deletion mutant had an MIC for colistin that was identical to the MIC of the colistin-resistant clinical isolate 6009-2 (MIC, 128 μg/ml; Table 1). Together, these data confirm that inactivation of hns in strain 6009-1, either by insertion of the IS element ISAba125 or by deletion of the hns gene via directed allelic replacement, results in a 4- to 8-fold increase in colistin resistance in A. baumannii strain 6009. Interestingly, the fitness of the 6009-1 Δhns mutant in normal growth media in vitro was greatly reduced in comparison to the fitness of 6009-1 (competitive growth index at 24 h, 0.003) and, to a lesser extent, the fitness of 6009-2. The 6009-1 Δhns mutant had a significantly longer lag phase during growth in the first 5 h, and the culture had a lower optical density (OD₆₀₀) at the end of the experiment at 24 h (Fig. S3). Supporting this finding, the 6009-1 Δhns mutant generated smaller colonies when grown on MH agar without supplementation than the parent strain 6009-1 grown for the same period of time (data not shown).

Transcriptional analysis of strains 6009-1 and 6009-2.

As H-NS proteins generally act as global transcriptional regulators, we predicted that the increased colistin resistance displayed by strain 6009-2 was likely due to the altered expression of one or more genes whose expression is controlled by H-NS. Therefore, we sought to identify the genes regulated by H-NS by comparing the transcriptomes of strains 6009-1 and 6009-2. Both strains were grown in the absence of colistin to early log phase (OD₆₀₀, 0.5), and then total RNA was purified and cDNA samples were generated and sequenced. A total of 164 genes were identified as being differentially expressed in the hns mutant strain 6009-2, as defined by a change in expression of >2-fold (log₂ ≥1.0), with a false-discovery rate (FDR) of <0.01. In total, 136 genes showed increased expression (Table S2) when hns was inactivated, but only 28 genes showed reduced expression (Table S3), supporting what has been observed in other bacteria that H-NS acts primarily as a negative regulator of gene expression (15).

The expression of a pmrC homologue was increased following hns inactivation, but that of the pmrCAB locus was not.

Analysis of the 6009 genomes indicated that they each contained two genes encoding predicted lipid A-specific pEtN transferases. The first gene, pmrC (ACICU_03004), is located within the three-gene pmrCAB locus in A. baumannii (Fig. 3). Missense mutations in the adjacent genes encoding the PmrAB two-component signal transduction system have been shown in A. baumannii to result in constitutive expression of pmrC and the addition of pEtN to lipid A with a concomitant increase in colistin resistance (7, 9). However, in strains 6009-1 and 6009-2, the pmrCAB loci were identical at the nucleotide level and contained none of the nucleotide changes known to be associated with pmrAB-mediated colistin resistance. Moreover, analysis of the RNA sequencing (RNA-seq) data indicated that while pmrC was transcribed, its expression was not significantly increased in 6009-2 (log₂ fold change, 0.49; FDR, 0.01). This low-level expression may partially explain the colistin resistance observed in the initial 6009-1 isolate. However, a second gene, eptA (ACICU_01072, 1,647 nucleotides in length), that encoded a PmrC homolog, which we designated EptA, was identified elsewhere on the genome (Fig. 3). Importantly, eptA was expressed at significantly higher levels in the 6009-2 hns mutant strain (log₂ fold change, 2.3; FDR, 4 × 10⁻¹¹) than in 6009-1. This pEtN gene was present in both the 6009-1 and 6009-2 strains and was not adjacent to any genes encoding two-component signal transduction system components (Fig. 3). A multiple-sequence alignment and the associated phylogenetic tree of known lipid A-specific pEtN transferases revealed that the two 6009 pEtN transferases, PmrC and EptA, were highly similar to each other and shared a high level of identity with the characterized lipid A pEtN transferases (Fig. 4). The 6009 PmrC protein shared 93% identity with the 6009 EptA protein and greater than 99% identity with PmrC proteins from A. baumannii ATCC 19606 and ATCC 17978, both of which are also encoded within the pmrCAB operon and known to be regulated by the PmrAB two-component system (8). In contrast, the 6009 EptA protein shared lower levels of identity with PmrC proteins from A. baumannii ATCC 19606 and ATCC 17978 (93% and 94% identity, respectively) but was highly similar to two pEtN transferases, EptA-1 and EptA-2 (97% and 99% amino acid identity, respectively), present in strain MSRN3004 and other extremely drug-resistant strains of A. baumannii (16). BLAST analyses of the publicly available A. baumannii ACICU genome revealed that it also contained a second lipid A-specific pEtN transferase gene (ACICU_01072) that shared 99.6% identity with eptA from strain 6009 (Fig. 4). However, it is not known if eptA is functional in ACICU, as the available annotation indicates that, in comparison to 6009-1, the ACICU gene is missing nucleotides 110(T) and 111(T), which means that the annotated start site as given for 6009-1 is out of frame in ACICU; therefore, a later ATG in ACICU is given as the start codon (nucleotide 298 in 6009-1).

FIG 3.

Comparison of the genomic regions containing the lipid A-specific phosphoethanolamine (pEtN) transferases in the 6009 strains and strain ACICU. (A) The fully characterized pEtN transferase-encoding gene, pmrC, in strains 6009-1, 6009-2, and ACICU is located in a three-gene operon upstream of the two-component response regulator pmrA and the sensor histidine kinase pmrB, which together control pmrC expression. (B) The second pEtN gene, eptA (ACICU_01072), is not associated with pmrAB and is located downstream of a gene encoding a TonB-dependent receptor protein (ACICU_01071), a number of hypothetical genes, and a gene encoding P4 phage site-specific integrase (ACICU_01075).

FIG 4.

Phylogenetic analysis of known and predicted lipid A-specific phosphoethanolamine (pEtN) transferases. The PmrC and EptA homologues from a number of A. baumannii strains and other bacterial species were compared. The Lpt_3 transferase, specific for the addition of pEtN to the 3 position of the second inner core heptose in Neisseria meningitidis, has been included in the analysis as an outlier. The name given at the right of each branch indicates the species, strain (plasmid), and gene/locus tag as listed in order from left to right and separated by hyphens. Bacterial species are as follows: Ab, Acinetobacter baumannii; Ec, Escherichia coli; Nm, N. meningitidis; and ST, Salmonella enterica subsp. enterica serovar Typhimurium.

To confirm the expression data obtained using RNA-seq, real-time PCR (qPCR) was performed using primers specific for eptA, pmrA, and pmrC. The qPCR results confirmed that eptA was significantly increased in expression in the hns mutant 6009-2 (log₂ fold change, 1.2; P < 0.01) compared to expression in 6009-1 (Fig. 5), while the expression of pmrC and pmrA was unchanged (log₂ fold change, 0.4 and P = 0.16; and a log₂ fold change, −0.2 and P = 0.88, respectively). These data confirm that the expression of eptA is regulated by H-NS and that the expression of the pmrCAB operon is independent of H-NS activity.

FIG 5.

Real-time qRT-PCR on selected genes identified as differentially expressed in 6009-2. qRT-PCR was used to determine the expression (log₂ ratio) of pmrC, pmrA, and eptA in 6009-1 and 6009-2 grown in cation-adjusted Mueller-Hinton broth (no antibiotics). The data generated for gene expression in 6009-2 were normalized against the expression of the housekeeping gene gyrB and then normalized to the expression of the corresponding gene in strain 6009-1. Each data point shows the expression value for one qRT-PCR measurement, with the horizontal lines showing mean ± standard error.

To determine if colistin resistance in strain 6009-2 was due solely to the increased expression of eptA, we cloned this gene into the expression vector pWH1266. The expression plasmid (pWH1266::eptA) and the empty vector control (pWH1266) were then separately introduced into the parent strain 6009-1. Colistin MICs (Table 1) determined for this set of strains revealed that the overexpression of eptA in 6009-1 resulted in significantly increased colistin resistance (MIC, 64 μg/ml) compared to the level of resistance in the parent strain 6009-1 (MIC, 8 to 16 μg/ml) and 6009-1 containing empty vector (MIC, 8 μg/ml) (Table 1). To confirm this observation, we attempted to inactivate eptA in strains 6009-1 and 6009-2 using allelic exchange mutagenesis, but despite numerous attempts, we were unable to generate a directed eptA mutant in either strain.

Strain 6009-2 shows increased pEtN modification of lipid A.

Colistin resistance in A. baumannii is mediated by modification of the LPS structure via the addition of pEtN and/or galactosamine (GalN) to lipid A (9, 10), or by the generation of lipid A mutants leading to the total loss of LPS (5). To assess the LPS produced by the 6009 strains, we used high-resolution electrospray ionization mass spectrometry (HR-ESI-MS) in the negative-ion mode to analyze the lipid A isolated from the isogenic clinical isolates 6009-1 and 6009-2 (hns mutant), the 6009-1 directed hns mutant (6009-1 Δhns), and 6009-1 provided with intact copy of eptA on pWH1266 (Fig. 6). Also included in the study was strain 6009-2 provided with an intact copy of hns on pWH1266 and 6009-2 provided with pWH1266 vector only (Fig. 6). All strains produced lipid A that was modified to some extent with pEtN and GalN, but the highly colistin-resistant clinical isolate 6009-2 (6009-2 with or without pWH1266) and the hns deletion mutant (6009-1 Δhns) showed an increase in the amount of pEtN-modified lipid A species (Fig. 6). HR-ESI-MS spectra of the lipid A derivatives isolated from 6009-1 (colistin MIC, 8 to 16 μg/ml) and the H-NS-complemented strain 6009-2/pWH1266::hns (colistin MIC, 16 to 32 μg/ml) (Fig. 6A) showed strong peaks at m/z 1,729.11, 1,883.26, and 1,911.28 that corresponded to the three different forms of unmodified lipid A (species 1, 2, and 3, respectively; Fig. 6B). These strains also showed peaks at m/z 1,852.12, 2,006.26, and 2,034.29, which correspond to each of the lipid A species with a single pEtN addition. In addition, two other minor peaks were identified, one at m/z 2,157.30 that corresponded to lipid A species 3 with two pEtN additions, and another at m/z 2,095.79 that corresponded to GalN modification of lipid A species 3 with an Na adduct. This level of modification of lipid A is unsurprising, as A. baumannii strains with colistin resistance as low as 1 μg/ml can exhibit pEtN or GalN modifications to lipid A (17). In contrast, the dominant peaks in the HR-ESI-MS spectra belonging to the highly colistin-resistant strains 6009-2 (with, or without, pWH1266) and 6009-1 Δhns were at m/z 1,772.16, 1,852.12, 1,954.32, and 2,034.29, which all correlate to lipid A species that are modified with a single pEtN. There was also a very significant peak at m/z 2,157.30, which corresponds to lipid A (species 3) modified at two positions with pEtN (Fig. 6A). These data strongly support the proposition that the mutation of hns results in increased levels of pEtN on lipid A. Furthermore, overexpression of eptA in 6009-1 (strain 6009-1/pWH1266::eptA) resulted in dominant peaks at m/z 1,852.12 and 2,034.29, which correspond to lipid A with a single pEtN addition. However, only the peak corresponding to pEtN addition to species 1 (m/z 1,852.12) was significantly higher (m/z 1,729.11 representing unmodified species 1) following overexpression of eptA (6009-1 compared to 6009-1 with pWH1266::eptA; Fig. 6). Together, these data suggest that EptA is able to transfer pEtN to lipid A, but it may not be the only transferase responsible for lipid A pEtN modification in the highly resistant 6009-2 and 6009-1 Δhns strains.

FIG 6.

Lipid A analysis of A. baumannii strains 6009-1 and 6009-2 and derivatives. (A) Mass spectra for the lipid A isolated from the following strains: strain 6009-1, isolated before colistin therapy; 6009-2, the spontaneous hns mutant isolated from the same patient post-colistin treatment; 6009-2 provided with an intact copy of hns (6009-2/pWH1266::hns) or with vector only (6009-2/pWH1266); and the 6009-1 hns deletion mutant and 6009-1 with eptA overexpressed (6009-1/pWH1266::eptA). The highlighted peaks (m/z) represent various forms of unmodified or modified lipid A (see panel B for details). Peaks highlighted in the same color share the same unmodified lipid A species (with or without OH or H₂PO₄; see panel B). Peaks connected by arrows are derivatives of the same lipid A species that have one or two phosphoethanolamine (pEtN) residues added. (B) Structures related to each of the three lipid A species are color-coded (exact mass and mass without one OH group or one H₂PO₄ shown below the images). Each species is shown together with the mass of the modified derivatives with one or two pEtN residues or one galactosamine (GalN). Lipid A samples were prepared and analyzed in two separate experiments, hence the variance in background peaks and x axis scales.

Other genes expressed at increased levels in the A. baumannii hns mutant.

As noted above, a large number of genes displayed significantly increased expression in the highly colistin-resistant hns mutant strain 6009-2 (Table S2), including many involved in the biosynthesis of pili, other adhesive surface structures, and numerous protein secretion system genes, as well as genes involved in poly-β-1,6-N-acetylglucosamine (PNAG) biosynthesis. Genes encoding the CsuA/BABCDE-dependent pili (ACICU_02418 to ACICU_02414) were expressed at highly elevated levels (∼32-fold), as were multiple genes predicted to encode proteins required for type 1 pili/fimbriae (ACICU_01810 to ACICU_01813) and P pilus (ACICU_1548, ACICU_1550, and ACICU_1551); similar increased expression of pilus-associated genes has been observed in an hns mutant of A. baumannii strain ATCC 17978 (14).

Genes expressed at decreased levels in the 6009-2 strain.

In the 6009-2 hns mutant, 28 genes were expressed at reduced levels compared to expression in 6009-1 (Table S3). Genes showing significantly reduced expression in 6009-2 included those encoding proteins required for cardiolipin biosynthesis, metal iron transport (hemF and hemN) and storage (ACICU_00759 and ACICU_03377), and choline transport and biosynthesis (betIAB; ACICU_00888, ACICU_00889, and ACICU_00890).

DISCUSSION

Polymyxins, such as colistin, are positively charged lipopeptides that have high binding affinity for the negatively charged lipid A component of LPS. This polymyxin-lipid A interaction leads to destabilization of the LPS layer and reduces bacterial outer membrane integrity (18), although the exact mode of killing by colistin is currently unknown. Polymyxin resistance via pEtN addition to lipid A by pEtN transferases has been characterized in several organisms, including E. coli and species within the genera Neisseria and Campylobacter (19–21). This modification leads to polymyxin resistance, as it results in a change to the overall charge of the lipid A (22); colistin-resistant strains of A. baumannii have a reduced negative membrane charge compared to colistin-sensitive strains (23). This reduced negative charge abrogates the ability of colistin to interact with the LPS and destabilize the outer membrane of Gram-negative bacteria.

IS elements are highly prevalent in pathogenic A. baumannii genomes; currently, 35 elements have been identified within the species (24). Several of these have been implicated in antibiotic resistance, most commonly by inserting upstream of carbapenemase and cephalosporinase genes and providing a strong promoter that increases the expression of these genes, leading to higher levels of antibiotic resistance (25–27). ISAba125 is a member of the IS30 family of ISs, which are typically between 1 kb and 1.6 kb in length and consist of a single open reading frame encoding a DDE family transposase. Multiple copies of the element are present in the same position in the 6009-1 and 6009-2 genomes, all of which are intact with no mutations observed in a conserved region for the transposase enzyme catalytic function and therefore highly likely to be active and capable of further transposition. The highly colistin-resistant strain 6009-2 contains an additional copy of ISAba125 located at the 3′ end of the regulatory gene hns. The movement of an IS element into a global regulatory gene in 6009-2 represents a novel genetic mechanism of increased colistin resistance in A. baumannii. This insertion is predicted to only alter the extreme C-terminal end of the H-NS protein, and it is possible that the insertion has a transcriptional effect on adjacent genes. However, deletion of the entire hns gene via allelic replacement resulted in increased resistance to colistin, and this phenomenon is reversed by providing the cells with an intact copy of hns on a replicating plasmid. Therefore, together, these data indicate that a functional H-NS results in the suppression of colistin resistance. While there is a second hns gene encoded on the endogenous plasmid present in this strain, our data suggest that it is not compensating for the chromosomally carried hns in relation to colistin resistance.

H-NS proteins in other bacterial species have been shown to have several roles in the cell, most notably in gene regulation and the silencing of horizontally acquired foreign DNA that often encodes virulence factors and antibiotic resistance determinants (28). Foreign DNA sequences can often be distinguished from the host genome by a higher A+T nucleotide content, although the cause of this bias is unknown. This feature is exploited by H-NS proteins, which do not recognize a specific consensus sequence but instead bind to regions that are A+T rich relative to the rest of the host genome (29). In Salmonella enterica, H-NS can bind to more than 10% of the genome, including all of its pathogenicity islands and many other virulence loci (15, 30). The suppression of gene expression provided by H-NS ensures that bacteria can tolerate and selectively regulate the genes encoded on horizontally acquired DNA, many of which may be detrimental to the cell (31).

The A. baumannii H-NS protein shares significant identity with orthologous proteins from other species, but there are important differences. Most notably, the A. baumannii H-NS protein is shorter, and the C-terminal conserved motifs known to be involved in dimerization and oligomerization of H-NS in other bacteria do not seem to be present in the A. baumannii protein (32, 33). However, the A. baumannii H-NS is predicted to be fully functional; when H-NS from Acinetobacter sp. strain 20 (which shares 76.8% amino acid identity with the 6009-1 H-NS and contains a short C-terminal domain) was used to complement an E. coli H-NS mutant, several wild-type phenotypes were restored (28).

H-NS inactivation in strains of E. coli and S. enterica confers multidrug resistance by increasing the expression of acrEF or mdtEF, which encode proteins involved in drug efflux and export (35, 36). These studies also showed that H-NS inactivation increased resistance to various antimicrobials and toxic compounds. Although colistin was not included in these investigations in E. coli and S. enterica, efflux pumps in Neisseria meningitidis are involved in mediating resistance to polymyxin B (37). All A. baumannii strains examined to date have cation and multidrug efflux pump systems encoded in their genomes, but none were identified as being differentially expressed in the hns mutant 6009-2, indicating that these drug efflux mechanisms are not controlled by H-NS. Though highly resistant to colistin (MIC, 128 μg/ml), the directed hns deletion mutant exhibited decreased overall fitness and had a smaller colony morphology than the 6009-1 parent strain. In contrast, there was no difference in colony morphology between 6009-1 and the colistin-resistant clinical isolate and hns mutant 6009-2, and 6009-2 displayed only a slight lag during growth. These observations suggest that the complete abrogation of H-NS function leads to a more significant dysregulation of global functions and that the C-terminal changes in the H-NS produced by 6009-2 might only partially attenuate function.

The transcriptomic analysis of the two clinical strains isolated before and after colistin treatment identified many genes with increased expression in the hns mutant 6009-2, including one we have called eptA, encoding a predicted pEtN transferase. This gene shared very high identity with pmrC but was located elsewhere on the genome and, unlike pmrC, was not colocalized with pmrAB or any other system that encoded a two-component regulator system. Other A. baumannii strains also contain multiple paralogs of pEtN transferase genes. A longitudinal study of 28 isolates taken from seven wounded war veterans identified strains containing up to three “alternative” pmrC or eptA-like genes, which had >95% identity to the canonical pmrC but <90% identity to a variant pmrC1 gene (16). Many of the genes identified around the eptA genes were located near prophages or integrase genes, indicating prior horizontal acquisition or duplication (16). Similarly, eptA in the 6009 strains is located downstream of the phage integrase gene ACICU_0175. Importantly, the regulation of any eptA gene via H-NS has not been previously described. Our study also showed that overexpression of the “orphan” eptA in the pre-colistin-treatment strain 6009-1 led to increased colistin resistance, thus proving that eptA encodes a functional pEtN transferase that can mediate colistin resistance.

Although increased eptA expression is likely to contribute significantly to the increased colistin resistance observed in the 6009 strains, our lipid A analysis indicated that this gene may not be solely responsible and that other H-NS-controlled genes may be involved. Genes identified as being differentially expressed in the 6009-2 hns mutant included those within the csu type I pilus operon (ACICU_02418 to ACICU_02414). Interestingly, this operon is found beside a putative transposase gene, suggesting that these genes may have been horizontally acquired in the past and therefore are likely repressed by H-NS. A. baumannii Csu proteins have also been identified as being differentially produced in a laboratory-generated colistin-resistant derivative of ATCC 19606; however, the genetic mechanism behind the resistance was not characterized in this study (38). In our study, increased csu gene expression correlated with increased expression of the orphan eptA and increased colistin resistance. In contrast, another study found that the expression of csu genes was decreased in ColR of a colistin-resistant clinical isolate, Ab347. However, resistance in this strain was shown to be mediated by pmrB mutations, leading to uncontrolled expression of pmrCAB encoding the primary lipid A pEtN transferase (39), rather than via increased expression of a secondary pEtN transferase. Further research is required to elucidate the full mechanisms of csuA and BABCDE pilus regulation. Interestingly, a recent analysis suggests that the csu cluster may be nonfunctional in both of the A. baumannii strains ACICU and ATCC 17978 (40). In our study, numerous protein secretion system genes were also expressed at increased levels in the 6009-2 hns mutant, including type VI secretion system (T6SS) genes vgrG1 (ACICU_01117) and vgrG3 (ACICU_3563), hcp (ACICU_01119), and other T6SS-associated genes (ACICU_01307 to ACICU_01312), the type V secretion system-autotransporter gene ata (ACICU_00993), the adjacent gene encoding an OmpA-like protein (ACICU_00994), and three contiguous genes (ACICU_01910 to ACICU_01912) predicted to encode a hypothetical protein, a hemagglutinin and its secretion/activation protein ShlB (also called FhaC). Notably, type VI secretion system genes are involved in the regulation of type I pili in some Gram-negative pathogens (41). Therefore, it is possible that the increased expression of the type I pilus genes in A. baumannii strain 6009-2 is primarily due to the dysregulation of the T6SS as a result of H-NS inactivation. Genes involved in poly-β-1,6-N-acetylglucosamine (PNAG) biosynthesis (ACICU_02362 to ACICU_02364) were also expressed at significantly increased levels in 6009-2. PNAG gene expression in A. baumannii has been shown previously to increase in response to colistin treatment, and PNAG production by E. coli during growth in biofilms helps resist the action of cationic antimicrobials (42).

Mutations in hns have previously been reported in several bacterial species, including A. baumannii. A hypermotile A. baumannii ATCC 17978 variant strain, 17978hm, was found to have an IS element (A1S_0268) inserted within the hns gene, thereby inactivating its function (14). Microarray data analysis of 17978hm revealed that 78 genes showed increased expression and 60 genes showed decreased expression (>4-fold change in expression) compared to expression in the parent strain ATCC 17978 (14). Importantly, the ATCC 17978 genome contains only one pEtN transferase gene, pmrC (A1S_2752), located within the pmrCAB operon, and this gene was not differentially expressed in the ATCC 17978 hns mutant (14). Moreover, the ATCC 17978 hns mutant (17978hm) is as sensitive to colistin as the parent strain (M. Brown, personal communication). When we compared the differential expression gene list generated for the hns mutant 6009-2 with that generated for the ATCC 17978 hns mutant (17978hm), only 13 genes were identified in both, eight genes with increased expression (Table S2) and five genes with reduced expression (Table S3). These included a type I pilus locus (A1S_1507 to A1S_1510), A1S_1033 (ACICU_00994) encoding an ompA homologue, and A1S_1032 (ACICU_00993) that encodes the trimeric autotransporter Ata. In ATCC 17978, the Ata protein has been shown to be involved in biofilm formation and adhesion to host matrix proteins (43). Thirteen of the ATCC 17978 genes identified with increased expression in strain 17978hm had no clear homologues in the 6009 genome. Conversely, 56 of the genes that showed increased expression in the hns mutant 6009-2 had no known homologues in the ATCC 17978 genome, including those within a single locus encoding a hypothetical protein, filamentous hemagglutinin, and a hemolysin secretion/activation protein ShlB (ACICU_01910, ACICU_01911, and ACICU_01912, respectively). Our data show that the clinical isolates 6009-1 and 6009-2 used in this study are very closely related to strain ACICU, which belongs to the global clone 2 lineage. In contrast, strain ATCC 17978 and its variant 17978hm belong to the global clone 3 lineage. We therefore predict that H-NS regulates the subset of genes common to strains 6009-1 and ATCC 17978, as well as a larger unique set of genes present in each strain. These differences in the genes regulated by H-NS are likely due to strain divergence and the type and amount of foreign DNA acquired and integrated into the genome. Future work should include an analysis of the consequences of expression of the 6009 eptA in a fully colistin-susceptible background. Furthermore, analysis of the role of H-NS in other A. baumannii strains encoding secondary phosphoethanolamine homologs would be of interest to gain a clearer understanding of the importance of H-NS inactivation as a general mechanism of colistin resistance in A. baumannii.

Concluding remarks.

In this study, we have shown that both the modification and inactivation of H-NS in A. baumannii strain 6009 can increase colistin resistance. Providing the posttreatment colistin-resistant hns mutant 6009-2 with an intact copy of hns in trans reduced colistin resistance back to levels similar to those observed in pretreatment clinical isolate 6009-1, proving that the global transcriptional regulator H-NS negatively regulates genes that can confer colistin resistance. Transcriptomic analyses identified more than 150 genes expressed at altered levels in the colistin-resistant H-NS mutant, including increased expression of the predicted pEtN transferase gene eptA, a close homologue of pmrC located elsewhere on the genome. Overexpression of wild-type eptA in the colistin-intermediate strain 6009-1 parent strain increased colistin resistance and the amount of pEtN decoration on some lipid A species but not others, suggesting that other factors may also be involved in the lipid A changes observed in 6009-2. This is the first study to show that H-NS plays a role in regulating genes involved in colistin resistance and that inactivation of hns can occur in clinically relevant strains during failed colistin treatment regimens.

MATERIALS AND METHODS

Bacterial strains, plasmids, and culture conditions.

The strains used in this study are outlined in Table 2. All A. baumannii strains were maintained on Mueller-Hinton (MH) agar or cultured in cation-adjusted MH broth (CAMHB; Oxoid) at 37°C, supplemented with 10 μg/ml of colistin sulfate and/or 12.5 μg/ml tetracycline where appropriate. Plasmid pWH1266 was obtained from the American Type Culture Collection. E. coli was maintained on lysogeny broth (LB) agar or LB broth (Oxoid) supplemented with the appropriate antibiotics, as required.

TABLE 2.

Strains and plasmids used in this study

Strain or plasmid (laboratory reference no.)	Descriptiona	Reference or source
Strains
A. baumannii
6009-1	A. baumannii clinical strain isolated before colistin treatment from a patient at Egas Moniz Hospital, Portugal; displays intermediate levels of colistin resistance (MIC, 4–16 μg/ml)	Miguel Viveiros, Institute of Hygiene and Tropical Medicine, Universidade Nova de Lisboa
6009-2	A. baumannii clinical strain isolated from the same patient as described above following 11 days of colistin treatment; high/very high colistin resistance (MIC, 64–128 μg/ml)	Miguel Viveiros, Institute of Hygiene and Tropical Medicine, Universidade Nova de Lisboa
6009-2/pWH1266::hns (AL2430)	Strain 6009-2 containing a functional copy of hns (amplified from 6009-1) cloned into pWH1266; Tetr	This study
6009-2/pWH1266 (AL2434)	Strain 6009-2 containing the empty vector pWH1266; Ampr Tetr	This study
6009-1/pWH1266::eptA (AL2702)	Strain 6009-1 containing a functional copy of eptA cloned into pWH1266; Tetr	This study
6009-1/pWH1266 (AL2706)	Strain 6009-1 containing empty vector pWH1266; Ampr Tetr	This study
6009-1 Δhns (AL2835)	6009-1 hns deletion mutant, generated by double-crossover allelic exchange mutagenesis; Tetr	This study
E. coli
DH5α	E. coli, F− Φ80lacZΔM15 Δ(lacZYA-argF) U169 recA1 endA1 hsdR17 (rK− mK+) phoA supE44 λ− thi-1 gyrA96 relA1	Bethesda Research Laboratories
Plasmids
pWH1266	A. baumannii/E. coli shuttle vector, 8.9 kb, Ampr Tetr	61
pWH1266::hns	pWH1266 containing the wild-type hns gene PCR amplified from A. baumannii strain 6009-1	This study
pWH1266::eptA	pWH1266 containing the eptA gene PCR amplified from A. baumannii strain 6009-2	This study

^aTet^r, tetracycline resistant; Amp^r, ampicillin resistant.

Isolation and manipulation of DNA.

A. baumannii genomic DNA extraction, purification, PCR, and cloning were performed as described previously (5). The plasmids and oligonucleotides used in this study are outlined in Tables 2 and S1, respectively. For initial identification of ISAba125 insertion into ACICU_00289, primers BAP7277 and BAP7278 were used for PCR amplification and genomic sequencing. For recombinant expression of A. baumannii 6009 genes, oligonucleotides were designed to amplify the full-length genes of interest from isolated genomic DNA. Forward oligonucleotides carried a PvuI restriction site, the reverse oligonucleotides carried a PstI restriction site for cloning into pWH1266, hns was amplified from 6009-1 using BAP 7353/7354, and eptA was amplified from 6009-2 using BAP 7584/7609 (Table S1). The purified PCR products were digested with PvuI and PstI and ligated into similarly digested pWH1266. Ligation reactions were used to transform competent E. coli DH5α cells and transformants selected on LB agar with tetracycline (12.5 μg/ml). Transformants containing the correct recombinant plasmid were identified using colony PCR, followed by DNA sequencing of the appropriate region in the plasmid. Recombinant plasmids were then separately introduced into A. baumannii by electroporation, as previously described (5).

Site-directed double-crossover mutagenesis of A. baumannii.

The hns gene was inactivated by allelic exchange mutagenesis, as described previously (44), with the following changes. Using A. baumannii 6009-1 genomic DNA as the template, the regions representing 1.5 kb upstream of hns and 1.5 kb downstream of hns were amplified with primers BAP 7806/BAP 7807 and BAP 7808/BAP 7809 (Table S1), respectively, using KOD Hot Start DNA polymerase (Merck Millipore). The tetracycline gene was separately amplified using purified pWH1266 plasmid DNA with primers BAP 7791/BAP 7792 (Table S1). A splice overlap extension PCR (SOE PCR) was then performed using all three PCR products together with the appropriate forward (BAP 7806) and reverse (BAP 7809) primers to generate a final PCR product (4 kb) representing the region upstream and downstream of the target gene separated by the tetracycline cassette. The SOE PCR product was then introduced into the A. baumannii strain 6009-1 (Table 2) by electroporation.

Transformation of E. coli and A. baumannii.

Plasmids were introduced into chemically competent E. coli cells (45), as previously described. A. baumannii strains 6009-1 and 6009-2 were transformed by electroporation, as previously described (46). Transformants were selected on LB agar plates containing 12.5 μg/ml tetracycline.

Genome sequence and bioinformatic analyses.

General nucleotide sequence analyses were performed using Vector NTI (Invitrogen) or NEBCutter 2 (47). Multiple-sequence alignments were performed using Clustal Omega (48). The draft genome sequences of A. baumannii strains 6009-1 and 6009-2 were determined using an Illumina genome analyzer IIx (Illumina, USA) at the Micromon Sequencing Facility (Monash University, Australia). Raw sequence reads from both strains were aligned independently to the A. baumannii ACICU genome using SHRiMP version 2.2.3 (49) and differentiating single-nucleotide polymorphisms (SNPs) identified as previously described (34). Raw read data were also assembled de novo using Velvet 1.2.10 (50) and annotated using Prokka 1.9 (51). The eptA gene from the 6009-1 and 6009-2 strains shared a high level of identity with the pmrC gene (97% nucleotide identity, 85% coverage) also present in both genomes. Therefore, the sequence reads belonging to the two regions were manually curated, and PCR amplicons representing the two pEtN-encoding regions were sequenced again using Sanger sequencing technology to confirm the nucleotide sequence of each gene.

Whole-genome transcriptomic studies.

For RNA isolation, all A. baumannii cultures were grown in the absence of colistin. First, a fresh overnight culture was subcultured 1:100 into 5 ml MH broth and incubated at 37°C with 200 rpm agitation until the mid-log-growth phase was reached, i.e., an optical density at 600 nm (OD₆₀₀) of approximately 0.5. Cultures were then subcultured again (1:100) into 30 ml prewarmed MH broth in 250-ml conical flask and incubated at 37°C with 200 rpm agitation until an OD₆₀₀ of 0.5 ± 0.05 was reached. Total RNA was purified and quantified as described previously (52), rRNA was depleted, and cDNA libraries were prepared as per Henry et al. (53). The cDNA libraries were then sequenced on an Illumina HiSeq 2000 platform (Macrogen, South Korea), and the RNA-seq reads aligned to the 6009-2 draft genome using SHRiMP (53). Data were analyzed using the Voom and Limma methods (54). Differentially expressed genes were defined as those showing a >2-fold (log₂ = 1.0) change in expression with a false-discovery rate (FDR) of <0.01.

Phylogenetic analyses and whole-genome comparisons.

The relationship of the 6009-1 and 6009-2 strains to other A. baumannii strains was determined using the Harvest suite of programs (55). Specifically, ParSNP version 1.2 (55) was used for core genome alignments (default parameters except for “–C 5000”) and Gingr version 1.1.1 for initial tree visualization and analysis. The phylogenetic tree was then manipulated further in FigTree version 1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/). The BLAST Ring Image Generator (Brig) (56) was used to identify differences between the 6009 and ACICU genomes.

Quantitative reverse transcription-PCR.

The RNA preparations used for the RNA-seq reactions were also used for quantitative reverse transcription-PCR (qRT-PCR). Oligonucleotides were designed using Primer 3 (57). Reverse transcription was performed using the AffinityScript qPCR cDNA synthesis kit (Agilent), and qRT-PCRs (biological triplicates) were performed using Brilliant III ultrafast qPCR master mix (Agilent) using a Mastercycler Ep Realplex (Eppendorf) with the following cycle conditions: 95°C for 3 min, 40 cycles of 95°C for 10 s, and 60°C for 20 s, followed by melt curve analysis. Gene-specific hydrolysis probes were used due to the high nucleotide identity between the pmrC and eptA genes with the appropriate primers, as well as primers and probes specific to pmrA and gyrB (Table S1). Gene expression levels were normalized relative to the housekeeping gene gyrB, and the relative standard curve method was used to determine gene expression levels, as previously described (52, 53). Standard curves were generated for each gene using known concentrations of genomic DNA extracted from A. baumannii strain 6009-1. The significance of each gene expression difference was determined using an unpaired Student's t test.

Colistin sulfate MIC BMD assays.

Each A. baumannii strain was subcultured 1:100 from an overnight culture, grown to an OD₆₀₀ of 0.4, and then diluted 1:100 in CAMHB to obtain a cell density of approximately 1 × 10⁶ CFU/ml. Colistin sulfate at various concentrations (1 to 512 μg/ml) was prepared in CAMHB, and 50 μl of each colistin concentration and 50 μl of each bacterial culture to be tested were added to the wells of polystyrene 96-well microtiter plates (Corning) and mixed thoroughly by pipetting. The final range of colistin concentrations tested was 0.5 μg/ml to 256 μg/ml. A positive control for growth consisting of 50 μl of CAMHB and 50 μl of bacterial culture and a sterility control consisting of 100 μl CAMHB were also included. The microtiter plates were incubated overnight at 37°C and each well assessed for visible growth of the bacteria, as indicated by turbidity. The MIC for each tested strain was recorded as the lowest concentration of colistin that completely inhibited the growth of the bacteria.

Lipid A purification and analysis by liquid chromatography-mass spectroscopy.

Crude lipid A was extracted as per Que et al. (58), with modifications as described previously (59). Lipid A analysis was performed using liquid chromatography-mass spectroscopy (LC-MS), as previously described (59).

Growth curve.

Overnight cultures were subcultured 1/50 and incubated at 37°C, with shaking at 200 rpm until an OD₆₀₀ of 0.2 ± 0.05 was reached. Once all of the strains had reached the desired OD₆₀₀, each strain was used for inoculation (1/100) into conical flasks of prewarmed CAMHB. Flasks were incubated at 37°C, with shaking at 200 rpm. The OD₆₀₀ of each culture was measured hourly for the first 7 h and then at 20, 24, and 28 h.

Competitive growth assay.

Competitive growth assays were performed as previously described (60), pooled samples were serially diluted and plated at 0 and 24 h, and samples were plated on MH agar and MH agar supplemented with colistin sulfate (10 μg/ml) and tetracycline (12.5 μg/ml) to select for 6009-1 Δhns.

Accession number(s).

The RNA-seq data have been deposited in the NCBI Gene Expression Omnibus with accession number GSE107964. Final draft annotated genomes have been given the GenBank accession number PUEI00000000.