PmrAB, a Two-Component Regulatory System of Pseudomonas aeruginosa That Modulates Resistance to Cationic Antimicrobial Peptides and Addition of Aminoarabinose to Lipid A

Abstract

Spontaneous polymyxin-resistant mutants of Pseudomonas aeruginosa were isolated. The mutations responsible for this phenotype were mapped to a two-component signal transduction system similar to PmrAB of Salmonella enterica serovar Typhimurium. Lipid A of these mutants contained aminoarabinose, an inducible modification that is associated with polymyxin resistance. Thus, P. aeruginosa possesses a mechanism that induces resistance to cationic antimicrobial peptides in response to environmental conditions.

Cationic antimicrobial peptides (CAPs) are a widely conserved host defense mechanism of plants and animals. Their antimicrobial effects can be attributed to their amphipathic, detergent-like nature, which enables individual CAP molecules to interact with both anionic and hydrophobic components of the bacterial envelope. CAPs bind to lipopolysaccharide (LPS), a major component of the gram-negative cell surface, through interactions with phosphates and fatty acids of LPS core and lipid A moieties (31). These molecules cross the outer membrane and periplasm, disrupt the membrane potential of the inner membrane, and thereby cause cell death (25). In vertebrates, the CAPs that pathogens encounter at epithelial surfaces are a major component of innate immunity, an ancient system of host defense that is stimulated via receptors that recognize pathogen-associated molecular patterns (29).

Pseudomonas aeruginosa is an opportunistic pathogen of humans that causes infections in those with host defense defects such as epidermal injury, immunodeficiency, and impaired epithelial clearance mechanisms. In the human host, P. aeruginosa is exposed to endogenous CAPs such as β-defensins (37) and cathelicidins (5) at epithelial surfaces. It may also encounter exogenous CAPs in this setting, when agents such as the polymyxins, acylated cyclic CAPs synthesized by the gram-positive soil bacterium Bacillus polymyxa, are used as antibiotics. Since the discovery and initial clinical use of the polymyxins more than 50 years ago, both clinical (12, 23, 26) and experimental (7, 16, 32) P. aeruginosa polymyxin resistance has been reported. P. aeruginosa possesses proteases that can degrade some CAPs (35); in addition, physiological (or “adaptive”) polymyxin resistance may occur in response to membrane stresses such as divalent cation limitation (7, 13, 27, 30) and polymyxin exposure (9, 16, 36), the latter being associated with the modulation of lipid A fatty acid composition (9). The P. aeruginosa PhoPQ two-component system contributes to the induction of these resistance phenotypes; however, its role appears to be complex (13, 27), and the potential roles of other regulatory systems related to PmrAB, a response regulator-sensor kinase pair that regulates polymyxin resistance in Salmonella enterica serovar Typhimurium (18, 34), have not been defined.

Isolation of polymyxin-resistant mutants of P. aeruginosa.

Conditions that physiologically induce polymyxin resistance have not been fully defined for P. aeruginosa and could involve multiple regulatory systems. In order to identify regulators important for this resistance, spontaneous mutants of wild-type P. aeruginosa (strain PAK; obtained from S. Lory) were isolated from late-exponential-phase cultures by selection on Luria-Bertani (LB) plates containing 20 to 50 μg of polymyxin B (USB/Amersham) per ml. After incubation for 72 to 96 h at ambient temperature (approximately 25°C), 12 initial isolates were colony purified, of which 6 displayed a stable resistance pattern, as indicated by growth on LB plates containing polymyxin B (20 μg/ml) after two passages through LB broth lacking CAPs. Among these six isolates, two distinct phenotypes were observed, differing primarily in terms of growth rate and degree of polymyxin resistance. For each phenotype, representative mutant strains, designated PAKpmrB6 and PAKpmrB12, were selected for characterization. The polymyxin-resistant strains grew more slowly on solid media than the wild-type strain but had growth rates in liquid media that were similar to that of the wild-type strain (generation time in LB broth at 37°C at mid-log phase, ≃50 min). The polymyxin resistance of PAKpmrB6 and PAKpmrB12 strains was confirmed by a quantitative bactericidal assay performed as described previously (39), with incubation in the presence of polymyxin B for 30 min, dilution, and plating for the enumeration of surviving CFU. For all bactericidal assays, each strain was tested in triplicate at each peptide concentration. Relative to that for the wild-type strain, the polymyxin concentrations resulting in a 50% reduction in the number of CFU (50% lethal dose [LD₅₀]) for PAKpmrB6 and for PAKpmrB12 were about 6 and 16 times as high, respectively (Table 1).

TABLE 1.

Resistance of P. aeruginosa PAKpmrB strains to CAPs

CAP	PAKpmrB+ LD50 (μg/ml)	PAKpmrB6		PAKpmrB12
		LD50 (μg/ml)	Relative resistance	LD50 (μg/ml)	Relative resistance
Polymyxin B	0.5	3	6	8	16
β-Defensin-2	0.5	1.5	3	6	12
Protegrin-1	1	6	6	60	60
C18G	1	16	16	>200	>200

Cross-resistance of polymyxin-resistant strains to additional CAP classes.

In other gram-negative bacteria, resistance to polymyxin may confer cross-resistance to other structural classes of CAPs due to structural modification of a common drug target (e.g., LPS), regulatory mutation, or both. Therefore, polymyxin-resistant strains of P. aeruginosa were tested for cross-resistance to additional CAPs. Quantitative bactericidal assays (39) were performed with the following modifications: cells were diluted to 2 × 10⁴ CFU per ml prior to assay, and assays of human β-defensin activity were performed in 1.4% tryptic soy broth with 10 mM Na phosphate (pH 7.4) rather than Mueller-Hinton broth. The PAKpmrB6 and PAKpmrB12 strains displayed cross-resistance to defensins, protegrin, and α-helical peptides, as reflected by relative LD₅₀s (compared to those of the PAK parental strain) for human β-defensin-2 (4), protegrin-1 (39), and C18G, an α-helical peptide derived from the carboxy terminus of platelet factor IV (11), that ranged from 3 to >200 (Table 1). In addition, these strains were also resistant to human β-defensin-1, rabbit α-defensin NP1, and the α-helical cathelicidins CAP18, SMAP29, and LL37 (data not shown). These results indicate that the P. aeruginosa PmrAB system regulates resistance to a variety of structural classes of CAP.

Identification of the P. aeruginosa PmrAB homologue.

Potential homologues of the S. enterica serovar Typhimurium pmrAB locus were identified by BLAST homology comparisons (1) of the P. aeruginosa Genome Project database (www.pseudomonas.com). This analysis revealed strong matches for pmrA in the P. aeruginosa open reading frames designated PA2479 (probability score of 6e⁻⁴⁶) and PA4776 (4e⁻⁴⁸) and corresponding matches for pmrB in PA2480 (4e⁻²⁶) and PA4777 (2e⁻²⁸). Therefore, insertion mutations targeting these loci were constructed in the polymyxin-resistant strains. A loss of resistance was associated only with the disruption of the locus corresponding to PA4776 (pmrA) and PA4777 (pmrB) in P. aeruginosa, which encode a response regulator and a sensor histidine kinase displaying, respectively, 44 and 32% identity and 59 and 48% similarity to their S. enterica serovar Typhimurium homologues. A PAKpmrB6-derived strain with a gentamicin resistance cassette inserted within pmrA was as susceptible to killing by polymyxin as the parental PAK strain (Fig. 1A). The disruption of pmrA in PAKpmrB12 gave similar results (data not shown). The polymyxin resistance phenotypes of these strains were completely dependent on an intact pmrAB locus. In contrast, the disruption of phoP in strain PAKpmrB6 did not diminish the resistance phenotype (Fig. 1A).

FIG. 1.

Role of the pmrAB locus in P. aeruginosa polymyxin resistance. (A) Resistance of strains with an aacC1 (Gm^r) cassette insertion. Triangles, PAKpmrB6 pmrA::aacC1; circles, PAKpmrB6; squares, PAKpmrB6 phoP::aacC1. (B) Resistance of strains with an episomal copy of the indicated pmrAB allele. Diamonds, PAKpmrA::aacC1(pMMB67HE::pmrAB6); circles, PAKpmrB6; inverted triangles, PAKpmrA::aacC1(pMMB67HE::pmrAB⁺).

Constitutive expression of the polymyxin resistance phenotype due to mutations in pmrB.

To test their ability to confer constitutive polymyxin resistance on recipient P. aeruginosa strains, pmrAB alleles from strains PAKpmrB6 and PAKpmrB12 were amplified by PCR from chromosomal DNA and cloned into the IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible broad-host-range expression plasmid pMMB67HE (15). Strains carrying a plasmid with these alleles were then tested for resistance to polymyxin. Bacteria were grown in the presence of IPTG under Mg²⁺-replete conditions that do not induce physiological polymyxin resistance. The pmrAB6 allele (but not the wild-type pmrAB allele or pmrA6 alone) conferred constitutive resistance on the pmrAB-null strain PAKpmrA::aacC1(Fig. 1B). These results suggested the presence of a mutation in the pmrB6 allele. Similar results were obtained for the pmrAB12 allele, indicating the presence of a mutation in pmrB12. Sequencing of these pmrB alleles revealed distinct missense mutations, L243Q in PAKpmrB6 (single nucleotide substitutions T5365486A) and A248V in PAKpmrB12 (C5365501T), in the histidine box motif of the sensor kinase, adjacent to the putative active-site histidine, H249. The H-box motif mediates phosphotransfer to the response regulator (PmrA) following sensor kinase activation and dimerization. Although the effect of any given H-box mutation cannot be readily predicted, the P. aeruginosa pmrB mutations are quite similar to activating mutations previously identified in ntrB (glnL), the nitrogen regulator II sensor kinase of Escherichia coli (2). Specifically, both the pmrB6 allele and the glnL1012 allele carry mutations at a position corresponding to a conserved leucine 6 residues towards the amino terminus from the active-site histidine, and the pmrB12 allele and glnL1004 allele carry identical mutations at a position corresponding to a conserved alanine immediately amino terminal to this histidine. Because the PmrAB system is known to stimulate aminoarabinose synthesis in S. enterica serovar Typhimurium, it is plausible that the pmrB H-box mutations selectively impair PmrB phosphatase activity, leading to constitutive activation of the PmrA regulon.

The pmrAB locus modulates the addition of aminoarabinose to lipid A in P. aeruginosa.

Loci within the genome of P. aeruginosa strain PAO1, designated PA3552 to PA3559 (pmrHFIJKLME), PA3540 (algD), and PA2022 (ugd), are homologues of PmrA-regulated S. enterica serovar Typhimurium genes that encode aminoarabinose synthetic enzymes essential for polymyxin resistance (17, 19). Because the polymyxin-resistant strains were thus expected to have aminoarabinose-modified LPS, lipid A was purified from them and analyzed. P. aeruginosa strains PAK, PAKpmrB6, PAKpmrB12, PAKpmrB6ΔpmrPAB::aacC1, and PAKpmrB12ΔpmrPAB::aacC1 were grown to stationary phase under conditions that do not induce physiologic aminoarabinose addition (LB broth with 1 mM MgCl₂ but without polymyxin B). Lipid A was isolated and analyzed as described previously (13) by negative-ion-matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (MS) by using a BIFLEX-III mass spectrometer (Bruker Daltonics Inc., Billerica, Mass.). PAK lipid A had mass peaks corresponding to the previously determined structures of P. aeruginosa lipid A (6, 22) containing five (m/z 1,447) or six (m/z 1,617) fatty acid substitutions (Fig. 2A). In contrast, PAKpmrB6 lipid A had additional mass peaks at m/z 1,748 and m/z 1,879 (Fig. 2B) corresponding to the addition of one or two 4-aminoarabinose moieties (change in m/z [Δm/z], 131) to the wild-type lipid A structure (Fig. 2C). The mass spectrum for PAKpmrB12 lipid A was indistinguishable from that of PAKpmrB6 (data not shown). Analyses of mass spectra for lipid A isolated from the PAKpmrB6ΔpmrPAB::aacC1 and PAKpmrB12ΔpmrPAB::aacC1 strains, in which the pmrPAB genes have been replaced by a gentamicin cassette, gave results similar to those for the wild type, indicating the PmrAB dependence of this modification in these strains. Because MALDI-TOF analysis is not quantitative, differences in polymyxin resistance observed among mutant strains may be due to differences in aminoarabinose content. Alternatively, differences may be due to changes in labile modifications of lipid A (e.g., phosphoethanolamine) lost during sample preparation (41) or to changes in nonlipid A surface structures, such as proximal LPS core sugar phosphates (21) or LPS-associated lipoproteins (20). Despite these possibilities, these results indicate that the P. aeruginosa PmrAB system mediates the addition of aminoarabinose to lipid A and provides additional support for the importance of this outer membrane modification in the polymyxin resistance of gram-negative bacteria.

FIG. 2.

Association of pmrB mutations with constitutive addition of aminoarabinose to P. aeruginosa lipid A. (A) MALDI-TOF negative-ion mode analysis of lipid A purified from P. aeruginosa strain PAK. In the negative-ion mode, observed molecular species lack at least one proton: [M—H]⁻. The difference between [M—H]⁻ at m/z 1,447 and m/z 1,617 (Δm/z = 170) in the mass spectrum indicates the loss of 3-hydroxydecanoate from position 3 of the reducing diglucosamine (right-hand ring in panel C). (B) Mass spectrum for lipid A from strain PAKpmrB6. The difference between [M—H]⁻ at m/z 1,617, 1,748, and 1,879 (Δm/z = 131) indicates the addition of aminoarabinose to the 1 and 4′ phosphates of lipid A. (C) Structure of P. aeruginosa lipid A without and with aminoarabinose. The X⁻ symbol at the right side of the lower structure represents either H (corresponding to m/z 1,748) or aminoarabinose (corresponding to m/z 1,879). The fatty acids depicted are 3-hydroxydecanoate (3-OH C_10:0), laurate (C_12:0), 2-hydroxylaurate (2-OH C_12:0), and 3-hydroxylaurate (3-OH C_12:0).

The isolation of spontaneous polymyxin-resistant mutants of P. aeruginosa, described here for the first time, was a prerequisite to defining the PmrAB two-component system as an important regulator of P. aeruginosa resistance to polymyxin and other CAPs. The polymyxin resistance phenotypes of P. aeruginosa strains were associated with mutations in the H-box motif of the PmrB sensor kinase. Both the polymyxin resistance of these mutants and the addition of aminoarabinose to their lipid A under Mg²⁺-replete conditions were dependent on the mutated pmrAB locus. These regulatory mutants were highly resistant to a variety of CAPs, indicating that the P. aeruginosa PmrAB system can induce CAP resistance.

The physiologic conditions that induce P. aeruginosa CAP resistance through PmrAB two-component signaling are not known. In S. enterica serovar Typhimurium, PmrB may function to sense the ionization state of iron (40); however, its periplasmic domain, the presumed site of this sensing capability, lacks homology to that of P. aeruginosa. Unlike S. enterica serovar Typhimurium, the viability of P. aeruginosa at high ferrous iron concentrations is not diminished by a disruption of pmrAB, and extracellular iron does not induce polymyxin resistance in a pmrAB-dependent fashion (S. M. Moskowitz and S. I. Miller, unpublished results). Moreover, the P. aeruginosa PAO1 genome lacks a homologue of the pmrD gene (33), an important regulator of PmrAB activation in S. enterica serovar Typhimurium (24). Thus, the activation of PmrAB in P. aeruginosa may differ significantly from that in S. enterica serovar Typhimurium.

The levels of transcriptional regulation of the pmrAB locus in P. aeruginosa strains also appear to differ. Polymyxin B and other CAPs induce transcription of the P. aeruginosa pmrAB locus in a PhoPQ-independent fashion (28). In contrast, in S. enterica serovar Typhimurium, subinhibitory concentrations of CAPs induce CAP resistance in a PhoPQ-dependent fashion (3). Nonetheless, both divalent cation deficiency (7, 13, 27) and acidity (Moskowitz and Miller, unpublished results) induce mild polymyxin resistance in wild-type P. aeruginosa, similar to effects in S. enterica serovar Typhimurium that are mediated by the PhoPQ and PmrAB systems, respectively (38). Thus, in both organisms, the induction of CAP resistance, triggered by various physiologic conditions acting through these two-component systems, may represent an important step in adaptation to host environments.

P. aeruginosa strains isolated from infants with cystic fibrosis (8) have aminoarabinose addition to lipid A (13) as an early adaptation, consistent with the notion that CAPs impose selective pressure on P. aeruginosa in the airways of individuals with cystic fibrosis (37). Moreover, inhaled polymyxin E (colistin) is commonly used to treat cystic fibrosis airway infection in Europe and Australia (12, 14, 26), and its intravenous use has also been advocated for the treatment of multidrug-resistant P. aeruginosa (10). Colistin-resistant strains of P. aeruginosa isolated from patients receiving inhaled colistin as routine maintenance therapy (12, 14) have alterations in the lipid A structure, including the addition of aminoarabinose (S. M. Moskowitz, R. K. Ernst, and S. I. Miller, unpublished results). Such lipid A modifications indicate potential targets for the development of novel antipseudomonal agents that could act synergistically with polymyxin.

Nucleotide sequence accession numbers.

Sequences of PAK pmrB wild-type, pmrB6, and pmrB12 alleles have been registered with the GenBank database under accession numbers AY493419 to AY493421.