A Two-Component Regulatory System Interconnects Resistance to Polymyxins, Aminoglycosides, Fluoroquinolones, and β-Lactams in Pseudomonas aeruginosa

Abstract

Constitutive overexpression of the active efflux system MexXY/OprM is a major cause of resistance to aminoglycosides, fluoroquinolones, and cefepime in clinical strains of Pseudomonas aeruginosa. Upregulation of this pump often results from mutations occurring in mexZ, the local repressor gene of the mexXY operon. In this study, analysis of MexXY-overproducing mutants selected in vitro from reference strain PAO1Bes on amikacin (at a concentration 1.5-fold higher than the MIC) led to identification of a new class of mutants harboring an intact mexZ gene and exhibiting increased resistance to colistin and imipenem in addition to aminoglycosides, fluoroquinolones, and cefepime. Reverse transcription-quantitative PCR (RT-qPCR) experiments on a selected clone named PAOW2 demonstrated that mexXY overexpression was independent of mexZ and the PA5471 gene, which is required for drug-dependent induction of mexXY. Furthermore, the transcript levels of the oprD gene, which encodes the carbapenem-selective porin OprD, were found to be reduced drastically in PAOW2. Whole-genome sequencing revealed a single mutation resulting in an M59I substitution in the ParR protein, the response regulator of the ParRS two-component regulatory system (with ParS being the sensor kinase), which is required for adaptive resistance of P. aeruginosa to polycationic peptides such as colistin. The multidrug resistance phenotype was suppressed in PAOW2 by deletion of the parS and parRS genes and conferred to PAO1Bes by chromosomal insertion of the mutated parRS locus from PAOW2. As shown by transcriptomic analysis, only a very small number of genes were expressed differentially between PAOW2 and PAO1Bes, including the lipopolysaccharide (LPS) modification operon arnBCADTEF-ugd, responsible for resistance to polycationic agents. Exposure of wild-type PAO1Bes to different polycationic peptides, including colistin, was shown to result in increased mexY and repressed oprD expression via ParRS, independent of PA5471. In agreement with these results, colistin antagonized activity of the MexXY/OprM substrates in PAO1Bes but not in a ΔparRS derivative. Finally, screening of clinical strains exhibiting the PAOW2 resistance phenotype allowed the identification of additional alterations in ParRS. Collectively, our data indicate that ParRS may promote either induced or constitutive multidrug resistance to four different classes of antibiotics through the activation of three distinct mechanisms (efflux, porin loss, and LPS modification).

Aminoglycosides are invaluable drugs in the management of patients with acute or chronic infections caused by Pseudomonas aeruginosa. Previous studies have shown that an active efflux mechanism implying a tripartite pump called MexXY/OprM modulates the activity of aminoglycosides toward this major nosocomial pathogen (1, 35, 51). The RND (resistance-nodulation-cell division family) transporter MexY interacts with the outer membrane channel OprM and the periplasmic adaptor protein MexX to actively extrude aminoglycosides and various unrelated antibiotics (fluoroquinolones, macrolides, tetracyclines, and zwitterionic β-lactams) from the intracellular compartment to the external medium (31). At the gene level, the MexX and MexY proteins are encoded by a single transcriptional unit, mexXY, whereas OprM is encoded by the third gene of a constitutively expressed operon, mexAB-oprM, coding for another efflux system (MexAB-OprM) (1, 26, 35). Although produced at very low levels in wild-type bacteria, the MexXY proteins contribute through their interaction with OprM to the intrinsic resistance of P. aeruginosa when they are upregulated as part of the stress response to antibiotics (30). Expression of the mexXY operon has been found to be induced strongly when the ribosomal activity is impaired either by mutations (in ribosomal proteins L1 and L25 or in methionyl-tRNA^fmet formyltransferase) or by antibiotics (aminoglycosides, macrolides, tetracyclines, and chloramphenicol) (4, 8, 19, 51). The functional link between the ribosome and MexXY/OprM was elucidated partially with the demonstration that a gene (PA5471) coding for a protein of unknown function was required for drug-dependent induction of mexXY expression (36). Interestingly, PA5471 is cotranscribed with a gene (PA5470) encoding an alternative peptide releasing factor (RF) presumed to rescue stalled ribosomes.

MexXY-overproducing mutants with increased resistance (from 2- to 16-fold) to the pump substrates are quite common in cystic fibrosis (CF) and non-CF patients (15, 18, 28, 45, 50). Most of these resistant bacteria exhibit mutations in mexZ, a gene flanking the mexXY operon and coding for a TetR-like protein that, when intact, strongly represses mexXY expression (32). Consequently, inactivation of mexZ and/or MexZ activity leads to upregulation of MexXY and increased resistance. However, MexXY overproducers with intact mexZ genes have repeatedly been reported among clinical strains, suggesting the presence of an additional regulatory gene(s) for mexXY (18, 28, 45). To our knowledge, none of these strains was demonstrated to upregulate PA5471. These mutants were named agrW mutants to make a distinction from those exhibiting alterations in mexZ (called agrZ mutants, for aminoglycoside-resistant mutants dependent on mexZ) (28).

The goal of the present study was to identify new regulators of mexXY expression through a whole-genome sequencing strategy applied to several one-step agrW mutants selected on aminoglycosides.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The bacterial strains and plasmids used or constructed for this study are listed in Table 1. Cultures were grown in Mueller-Hinton broth (MHB) with adjusted concentrations of the divalent cations Ca²⁺ and Mg²⁺ (Becton Dickinson, Microbiology Systems, Cockeysville, MD), on Mueller-Hinton agar (MHA) (Bio-Rad, Marnes-la-Coquette, France), or on BM2 glucose minimal medium containing a low (20 μM) or high (2 mM) MgSO₄ concentration (11). Isolation of one-step mutants with increased MexXY/OprM efflux activity was performed by plating 100-μl aliquots of log-phase P. aeruginosa cultures (A₆₀₀ equal to 1) on MHA supplemented with 6 μg ml⁻¹ of amikacin (at a concentration 1.5-fold higher than the MIC for strain PAO1Bes). Recombinant plasmids were introduced into P. aeruginosa strains by triparental matings, using the mobilization properties of the broad-host-range helper plasmid pRK2013 (6). Transconjugants were selected on Pseudomonas isolation agar (PIA; Becton Dickinson) supplemented with appropriate antibiotic concentrations, as follows: for Escherichia coli, ampicillin at 100 μg/ml, tetracycline at 15 μg/ml, and streptomycin at 50 μg/ml; and for P. aeruginosa, ticarcillin at 150 μg/ml, tetracycline at 200 μg/ml, and streptomycin at 2,000 μg/ml.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Phenotype or genotypea	Source or reference
Strains
P. aeruginosa strains
PAO1Bes	Wild-type reference strain	K. Stover
CMZ091	PAO1Bes ΔmexZ	This study
CM090	PAO1Bes ΔPA5470	This study
CM093	PAO1Bes ΔPA5471	This study
CM095	PAO1Bes ΔparS	This study
CM096	PAO1Bes ΔparRS	This study
CM099	PAO1Bes ΔpmrAB	This study
CM106	CM096 cis complemented with parRS genes from PAOW2, inserted at the attB site; Tcr	This study
PAOW1	PAO1Bes spontaneous mutant; Amkr	This study
PAOW2	PAO1Bes spontaneous mutant with Met59-to-Ile change in ParR; Amkr	This study
CMZ089	PAOW2 ΔmexZ	This study
CM092	PAOW2 ΔPA5470	This study
CM094	PAOW2 ΔPA5471	This study
CM097	PAOW2 ΔparS	This study
CM098	PAOW2 ΔparRS	This study
CM100	PAOW2 ΔpmrAB	This study
CMC	Wild-type susceptible clinical strain	This study
CMCW2	CMC spontaneous mutant with Leu14-to-Gln change in ParS; Amkr	This study
CM107	CM096 cis complemented with parRS genes from CMCW2, inserted at the attB site; Tcr	This study
E. coli strains
S17.1	recA thi pro hsdR Tra+ Smr	44
DH5α	F−supE44 endA1 hsdR17(rK− mK−) thi-1 recA1 Δ (argF-lacZYA)U169 φ80dlacZΔM15 phoA gyrA96 relA1 deoR λ−	Invitrogen
CC118	A(ara-leu) araD ΔlacX74 galE galK phoA20 thi-1 rpsE rpoB argE(Am) recAl	29
CC118 λpir	CC118 lysogenized with λpir phage	12
HB101	supE44 hsdS20(rB− mB−) recA13 ara-14 proA2 lacY1 galK2 rpsL20 xyl-5 mtl-1 leuB6 thi-1	25
Plasmids
pCR2.1	Cloning vector for PCR products; lacZΔColE1 f1 ori Apr Kmr	Invitrogen
pKNG101	Suicide vector in P. aeruginosa; sacB Smr	21
pRK2013	Helper plasmid; ColE1 ori Tra+ Mob+ Kmr	6
Mini-CTX1	Self-proficient integration vector with tet, V-FRT-attPMCS, ori, int, and oriT; Tcr	14
pKNGΔmexZ	BamHI/ApaI 1.212-kb fragment composed of sequences flanking 5′ and 3′ ends of mexZ, cloned into pKNG101; Smr	This study
pKNGΔ5470	BamHI/ApaI 1.194-kb fragment composed of sequences flanking 5′ and 3′ ends of PA5470, cloned into pKNG101; Smr	This study
pKNGΔ5471	BamHI/ApaI 1.241-kb fragment composed of sequences flanking 5′ and 3′ ends of PA5471, cloned into pKNG101; Smr	This study
pKNGΔparS	ApaI/ApaI 1.079-kb fragment composed of sequences flanking 5′ and 3′ ends of parS, cloned into pKNG101; Smr	This study
pKNGΔparRS	ApaI/ApaI 1.045-kb fragment composed of sequences flanking 5′ and 3′ ends of parRS, cloned into pKNG101; Smr	This study
pKNGΔpmrAB	BamHI/ApaI 1.241-kb fragment composed of sequences flanking 5′ and 3′ ends of pmrAB, cloned into pKNG101; Smr	This study
CTX-W2	parRS from PAOW2 cloned into mini-CTX1 at SpeI/EcoRV sites; Tcr	This study
CTX-CW2	parRS from CMCW2 cloned into mini-CTX1 at SpeI/EcoRV sites; Tcr	This study

^aAntibiotic resistance phenotypes; Tc^r, tetracycline resistance; Amk^r, amikacin resistance; Sm^r, streptomycin resistance; Ap^r, ampicillin resistance; Km^r, kanamycin resistance. MCS, multiple cloning sites.

Antibiotic susceptibility testing.

The MICs of selected antibiotics in MHA were determined by recommended agar dilution procedures (5). Bacterial susceptibility in BM2 glucose minimal medium was determined by the macrodilution broth method at two MgSO₄ concentrations (20 μM and 2 mM), as described previously (52). Growth was assessed after 24 h of incubation at 37°C. ParRS-dependent drug induction of the mexY gene was assayed by the disk diffusion method. MHA plates were inoculated with calibrated suspensions of strains PAO1Bes and CM096 as recommended by the CLSI. Disks containing meropenem (MEM; 10 μg), cefepime (FEP; 30 μg), gentamicin (GEN; 15 μg), and ciprofloxacin (CIP; 10 μg) were deposited on the agar surface 3 h after deposition of the colistin (CST; 50 μg) disk (Bio-Rad, Marnes-la-Coquette, France). The checkerboard technique for investigating antibiotic interactions has been described elsewhere (5).

Drug killing experiments.

Overnight cultures of strain PAO1Bes and its mutant CM096 were diluted 1:50 in fresh MHB containing indolicidin or colistin (at a concentration equivalent to 0.5× MIC) and grown with constant shaking (250 rpm) at 37°C to an A₆₀₀ value of 0.5 ± 0.05. The bacteria from 1 ml of culture were then collected by centrifugation (5,000 × g) and resuspended in 1 ml of drug-free MHB. This cell suspension was diluted 1:10 in prewarmed MHB prior to the addition of 16 μg/ml gentamicin (a concentration 8-fold higher than the MIC) and then was reincubated at 37°C. The survivors of gentamicin action were counted on MHA plates inoculated with serial 10-fold dilutions of culture aliquots taken at designated time points.

Construction of deletion mutants in strain PAO1 and mutant PAOW2.

Single- or multiple-knockout mutants in the mexZ, PA5470, PA5471, parS, parRS, and pmrAB genes were constructed by using overlapping PCR and recombination events according to the method reported by Kaniga et al. (21). First, the 5′ and 3′ regions flanking each of the genes were amplified by PCR (ca. 500 bp in length) (a list of the primers used is provided in Table 2) under the following conditions: 5 min at 95°C followed by 30 cycles of 30 s at 95°C, 30 s at 60°C, and 30 s at 72°C, with a final extension step of 5 min at 72°C. The resultant amplicons were then used as templates for overlapping PCR with external pairs of primers (Table 2) to generate the mutagenic DNA fragments. The reaction mixtures contained a 200 μM concentration of each deoxynucleoside triphosphate (dNTP), 6% dimethyl sulfoxide (DMSO), 1× polymerase buffer, a 0.5 μM concentration of each primer, and 0.5 U of BioTaq Red DNA polymerase (Bioline, London, United Kingdom). The amplified products were cloned into plasmid pCR2.1 according to the manufacturer's instructions (Invitrogen, Carlsbad, CA) and next subcloned into the suicide vector pKNG101 and E. coli CC118 λpir, as BamHI/ApaI or ApaI/ApaI fragments (21). The recombinant plasmids were transferred into P. aeruginosa by conjugation, and the deletion mutants were selected on PIA plates containing 5% sucrose and 2,000 μg ml⁻¹ streptomycin. The allelic exchanges were confirmed by PCR. Nucleotide sequencing experiments confirmed deletion of 627 bp, 522 bp, 957 bp, 1,037 bp, 1,815 bp, and 1,941 bp in mexZ, PA5470, PA5471, parS, parRS, and pmrAB, respectively.

TABLE 2.

Oligonucleotides used for gene inactivation, gene cloning, and RT-qPCR

Oligonucleotide	Sequence (5′→3′)
Oligonucleotides for gene inactivation
mexZ
PCRimexZC1	CGTAGCGCTTGAGCTTGTC
PCRimexZC2	GCCCTCACACTGAACGTCCTCACAAGG
PCRimexZC3	TTCAGTGTGAGGGCGTCAATCGTCC
PCRimexZC4	GGTTCGTAGATCGGCATCTC
PA5471
PCRi5471C1	CTCCTGCGCGAGGTCTAC
PCRi5471C2	GGCGGTCGCGGACAAAGGCTTGATGTA
PCRi5471C3	TGTCCGCGACCGCCAGCTGATCTACG
PCRi5471C4	GATCGCCAGGCGTTTGTT
PA5470
PCRi5470C1	CTGGAACTGCTCGACATCAA
PCRi5470C2	GAAGCTGTCCTCGATTACCTCGACCTC
PCRi5470C3	TCGAGGACAGCTTCACTGGCTCTTCG
PCRi5470C4	CCGACCTCTTCCACCTGTT
parS
PCRiparSC1	AGGAGGACGATCTCGATCAC
PCRiparSC2	AGATCTTGGTCGACCACGAAGATCG
PCRiparSC3	GTCGACCAAGATCTTCGAACCCTTTTCG
PCRiparSC4	GCGTTGATCCTGTGCTGTC
parRS
PCRiParRSC1	GGATTCAGCTTCTGCTCGAC
PCRiParRSC2	AGATCTTCATAGGGTTTCATCGGTCG
PCRiParRSC3	CCCTATGAAGATCTTCGAACCCTTTTCG
PCRiParRSC4	GCGTTGATCCTGTGCTGTC
pmrAB
PCRiPmrABC1	TGTCGAACTGACCCAGCTAC
PCRiPmrABC2	CTTCCAGGTCACCCATTCCACGGTATC
PCRiPmrABC3	GGGTGACCTGGAAGTGCAGGTGTTCCT
PCRiPmrABC4	GTGCTGAGCTCCTCGATCTT
Oligonucleotides for complementation of parRS
CloparRSC1	GAGGGAAAAGCAGAAGTCACC
CloparRSC2	CGAGGTGTCCCATGCTAGG
Oligonucleotides for RT-qPCR
uvrD1	GCAGCCTCGCCCTACAGCA
uvrD2	GGATCTGGAAGTTCTGCTCAGC
mexY1A	GCAGCCTCGCCCTACAGCA
mexY1B	GGATCTGGAAGTTCTGCTCAGC
mexZRTA1	TTACCTCCTCCAGCGGC
mexZRTA2	GTGAGGCGGGCGTTGTG
PA5470C1	CCAAAGAGGAATCCCAGAAA
PA5470C2	CAGGCAGACCTCGATCTTGT
PA5471C1	TCGAGGTAATCGAGGAGGTG
PA5471C2	AGGGTCTGCAAACGGATCTC
oprDRTC1	CATCTACCGCACAAACGATG
oprDRTC2	ACAGAGTTGGCGAGGAAAATC
parRSTC1	CGAACTGGAGGAAATGGTCT
parRSTC2	ATGCGGATCTGTTCGACCT
pmrARTC1	GATACCGTGGAATGGGTGAC
pmrARTC2	GGCTTGGTCAGGTAGTCGTC
pmrBRTC1	ACGAACTCAACCTCCTGCTG
pmrBRTC2	ATCTGCTCCATCAAGGTGCT
PA1559RTC1	GCAGCAACTGGTGGACTACA
PA1559RTC2	CATGCGGAAGACCAGAAGAT
PA1797RTC1	GGACCCTTTGCAGATGACTC
PA1797RTC2	CGGAGTGTTTCCTGAGAAGC
PA2358RTC1	GTACTGTTCGCCGGAACAAC
PA2358RTC2	CCTGGAGCAGGAATATCTGG
PA2655RTC1	GTGCTGGTGTTCCTGTTGG
PA2655RTC2	CGTAGGTCCCCCAGATCG
PA3554RTC1	GTGGCTCGAATACCATGTGA
PA3554RTC2	TGCCGTATTTCACGCAGTAG
PA4773RTC1	CAGTGGATCGAGGAAAGCAT
PA4773RTC2	GTACTCCGGCCAGGTATGG
PA4774RTC1	CGCCGAACCACTTCTATTTC
PA4774RTC2	TCGTGGTACAGCGACTCATC

Chromosomal complementation with full-length parRS.

The parRS loci of mutants PAOW2 and CMCW2 were PCR amplified from genomic DNA by using the oligonucleotide pair CloparRSC1/CloparRSC2 (Table 2). The resulting 2,314-bp DNA fragments were cloned into pCR2.1 and next subcloned into the HindIII/SacI restriction sites of plasmid mini-CTX1 (14). The recombinant plasmids were then transferred from E. coli CC118 to P. aeruginosa strain CM096 (PAO1 ΔparRS) by conjugation, using PIA plates supplemented with tetracycline. Flippase-promoted excision of the chromosomally integrated FRT cassettes (tetracycline resistance and integrase genes) was achieved by conjugational transfer of plasmid pFLP2 from donor E. coli S17.1 to strain CM096 and subsequent selection of the transconjugants on ticarcillin-MHA plates (13). Plasmid pFLP2 was subsequently cured by streaking selected recipient clones on MHA medium supplemented with 5% sucrose.

Transcriptional profiling.

Triplicate cultures of mutant strain PAOW2 and its parent, PAO1Bes, were grown at 37°C with shaking (1:100 dilution of an overnight culture in 25 ml MHB) to an A₆₀₀ value of 1 ± 0.05. The bacteria were collected by centrifugation at 15,000 × g for 1 min. Total RNA was extracted from the pellet by using an RNeasy Plus Mini kit (Qiagen SA, Courtaboeuf, France), treated with DNase (RQ1 RNase-free DNase; Promega, Charbonnières les Bains, France) for 1 h at 37°C, and purified with an RNeasy Mini Elute cleanup kit (Qiagen SA). Ten micrograms of RNA was next retrotranscribed into cDNA, fragmented, labeled, and hybridized on an Affymetrix P. aeruginosa GeneChip by DNA Vision (Charleroi, Belgium). Normalization and signal value extraction were achieved with the Robust Multiarray Average (RMA) software package (16, 17). Differential gene expression between PAO1Bes and PAOW2 was analyzed with Student's t test, using a nominal significance level (P) of ≤0.05 for each univariate test. Changes in gene expression of ≥2-fold were considered significant.

RT-qPCR.

Specific gene expression was measured by real-time reverse transcription-quantitative PCR (RT-qPCR) as described previously (7). Briefly, 2 μg of total RNA was reverse transcribed with ImpromII reverse transcriptase as specified by the manufacturer (Promega). The amounts of specific cDNA were assessed in a Rotor Gene RG6000 real-time PCR instrument (Qiagen) by using a Fast SybrGreen kit (Qiagen) and primers designed from the sequence of Pseudomonas (Pseudomonas genome database v2 [http://v2.pseudomonas.com]) (Table 2), with uvrD transcripts as an internal control (19). The transcript levels of a given gene in a given strain were normalized with those of uvrD (20) and expressed as a ratio (fold change) to that for wild-type strain PAO1Bes, used as a reference. Gene expression values were calculated from three independent bacterial cultures, each of which was tested in duplicate.

SNP identification.

Single-nucleotide polymorphisms (SNPs) between the PAO1 strain used in our laboratory, called PAO1Bes (for Besançon), its derived mutant PAOW2, and the published PAO1 reference sequence (GenBank accession no. NC-002516) (46) were established with Illumina's ELAND aligner as described previously (22). Briefly, the genomic DNAs from PAO1Bes and PAOW2 were extracted and purified with a QIAamp DNA Mini kit (Qiagen). Illumina libraries were prepared from 10 μg of each DNA preparation by GATC Biotechn GA (Konstanz, Germany), using a Genome Analyzer I apparatus. Mapping of the sequence reads (76 bp) of PAOW2 (2,297,644 reads) to the published PAO1 reference led to the identification of potential sequence variations (SNPs). A SNP was considered reliable if the coverage was ≥5-fold and its percentage was ≥75%. The sequence variations predicted for PAOW2 relative to PAO1 and PAO1Bes (2,289,792 reads) were verified on both DNA strands in an Applied Biosystems 3130 automatic sequencer (Applied Biosystems, Courtaboeuf, France) after PCR amplification with proper primers.

RESULTS AND DISCUSSION

Overexpression of mexY independent of MexZ and PA5471.

A functional link has been established between the efflux system MexXY/OprM and the ribosomal machinery (8, 19). In order to identify novel physiological functions associated with this transporter, we selected spontaneously MexXY-overproducing mutants of wild-type strain PAO1Bes (the PAO1 strain studied in our laboratory) by culture on MHA supplemented with amikacin at a concentration 1.5-fold higher than the MIC. Mutants developing on this medium were easily obtained, at rates ranging from 7.7 × 10⁻⁷ to 7.7 × 10⁻⁸. Drug susceptibility tests followed by sequencing experiments on 15 randomly selected colonies led to the identification of MexXY overproducers exhibiting either an intact or mutated mexZ gene (13 agrW and 2 agrZ mutants, respectively). Consistent with MexXY being upregulated in the agrW mutants, all of the colonies tested were 2- to 8-fold more resistant than PAO1Bes to the MexXY/OprM substrates, including aminoglycosides (gentamicin, amikacin, tobramycin, and apramycin), cefepime, and ciprofloxacin (28, 45, 50) (Table 3). RT-qPCR data confirmed overexpression of the mexY gene in these bacteria (data not shown). More interestingly, the agrW mutants could be divided into two phenotypic groups, named PAOW1 and PAOW2, with respect to their susceptibility to carbapenems (imipenem and meropenem) and colistin (Table 3). While the PAOW1 type (n = 6) exhibited the same susceptibility as PAO1Bes to these drugs, the PAOW2 type (n = 7) was 2- to 8-fold more resistant. Since MexXY overexpression had thus far never been associated with a decreased susceptibility to carbapenems and polymyxins, we focused our attention on a typical PAOW2 mutant (results on PAOW1 will be reported in a separate paper).

TABLE 3.

Drug susceptibilities of various mutants derived from strains PAO1Bes and CMC

Strain	Genotype	MIC (μg/ml)a
		GEN	AMK	TOB	APR	FEP	CIP	IMP	MEM	CST
PAO1Bes	Wild type	2	4	1	16	4	0.25	2	1	1
CMC	Wild type	2	4	1	16	4	0.25	2	1	1
PAOW1	PAO1Bes agrW1 mutant	16	32	4	64	16	1	2	1	1
PAOW2	PAO1Bes agrW2 mutant	8	32	2	64	8	1	16	4	2
CMCW2	CMC agrW2 mutant	4	8	2	32	8	0.5	16	4	2
CMZ091	PΑΟ1 ΔmexZ	8	16	2	32	8	1	2	1	1
CM090	PΑΟ1 ΔPA5470	2	4	1	16	4	0.25	2	1	1
CM093	PΑΟ1 ΔPA5471	≤0.12	2	0.5	4	4	0.12	2	1	1
CM095	PΑΟ1 ΔparS	2	4	1	16	4	0.25	2	1	0.5
CM096	PAO1 ΔparRS	2	4	1	16	4	0.25	2	1	0.5
CM099	PΑΟ1 ΔpmrAB	2	4	1	16	4	0.25	2	1	0.5
CMZ089	PΑΟW2 ΔmexZ	8	32	2	64	8	1	16	4	2
CM092	PΑΟW2 ΔPA5470	8	32	2	64	8	1	16	4	2
CM094	PΑΟW2 ΔPA5471	8	32	2	64	8	1	16	4	2
CM097	PΑΟW2 ΔparS	2	4	1	16	4	0.25	2	1	0.5
CM098	PΑΟW2 ΔparRS	2	4	1	16	4	0.25	2	1	0.5
CM100	PΑΟW2 ΔpmrAB	8	32	2	64	8	1	16	4	1
CM106	CM096 with parRS from PAOW2	4	8	1	32	8	0.5	16	4	1
CM107	CM096 with parRS from CMCW2	4	8	1	32	8	0.5	16	2	1

^aMICs in bold (or underlined) indicate that the bacteria were at least 2-fold more (or less) resistant than the reference strain PAO1Bes. The data are representative of four distinct experiments. Abbreviations: GEN, gentamicin; AMK, amikacin; TOB, tobramycin; APR, apramycin; FEP, cefepime; CIP, ciprofloxacin; IMP, imipenem; MEM, meropenem; and CST, colistin.

In a first attempt to find out the genetic mechanisms leading to MexXY upregulation in PAOW2, we sequenced (PA5470, PA5471, rplA, rplY, and nuoG) and quantified (mexZ, PA5470, and PA5471) the transcripts of a number of genes known to influence mexXY operon expression (8, 32, 36, 51). Since all of these experiments yielded negative results, we looked at whether some of these genes would be required for mexXY overexpression in PAOW2. We thus constructed mexZ, PA5470, and PA5471 deletion mutants of PAO1Bes and PAOW2. As expected, inactivation of the repressor gene mexZ in PAO1Bes (mutant strain CMZ091) resulted in a strong increase in mexY expression and in more resistance to all of the MexXY/OprM substrates tested (Fig. 1 and Table 3). In PAOW2, however, deletion of mexZ (mutant strain CMZ089) upregulated mexY expression only 2-fold, without notable effects on the mutant's susceptibility (Fig. 1 and Table 3). Concordant with data from the literature (36), inactivation of PA5471 (mutant strain CM093), but not that of the adjacent gene PA5470 (mutant strain CM090), significantly reduced mexY expression in wild-type PAO1Bes, leading to supersusceptibility to most of the antibiotics exported by MexXY/OprM (compare strains CM090, CM093, and PAO1Bes in Fig. 1 and Table 3). In contrast, suppression of PA5471 had no influence on the mexY level and on drug susceptibility in PAOW2 (compare strains CM094 and PAOW2). Altogether, these results suggested the existence of a new regulatory pathway for MexXY/OprM that is able to override the control exerted by MexZ and PA5471.

FIG. 1.

Mutant strain PAOW2 upregulates the mexXY operon independently of the mexZ, PA5470, and PA5471 genes. mexY gene levels in strains PAO1Bes and PAOW2 and in their respective ΔmexZ (CMZ091 and CMZ089), ΔPA5470 (CM090 and CM092), and ΔPA5471 (CM093 and CM094) mutants were quantified by RT-qPCR. The bars represent means of three independent determinations ± standard deviations (SD).

It is well established that the OprD porin is the specific uptake pathway for basic amino acids and carbapenems across the outer membrane of P. aeruginosa (47). Since qualitative and/or quantitative alterations of this channel are a major cause of resistance to carbapenems in this organism (24, 27, 38, 42), we sequenced and measured the expression levels of the oprD gene in PAOW2 and PAO1Bes. PAOW2 turned out to harbor an intact oprD gene whose expression was 7.1-fold ± 0.1-fold less than that in PAO1Bes (Fig. 2). Reminiscent of the case for nfxC mutants, which overproduce the efflux system MexEF-OprN while downregulating OprD (23, 39), our results provided good evidence for coordinated regulation between two complementary resistance mechanisms aimed at limiting intracellular drug accumulation (i.e., active efflux and outer membrane impermeability) in the agrW2 mutants.

FIG. 2.

Impact of mutations in parR and parS on expression levels of the mexY and oprD genes. The mRNA levels of the mexY (black bars) and oprD (white bars) genes were determined by RT-qPCR. The bars represent means of three independent determinations ± SD.

Assuming that the elevated rates at which the agrW2 mutants were recovered on selective medium were the result of single mutations, we carried out whole-genome sequencing experiments to identify the genetic alteration of PAOW2.

Mutational alteration of ParRS two-component regulatory system.

Because of the genomic polymorphism that may exist between PAO1 laboratory strains (22), we aligned the sequence reads from PAO1Bes (2,289,792) and its mutant PAOW2 (2,297,644) with the PAO1 genome sequenced by Stover et al. (46). An initial analysis of the sequencing data pointed to the potential existence of 13 SNPs in PAOW2. Only one of these was confirmed by PCR and resequencing: a G-to-A change at nucleotide position 177 of the PA1799 gene was predicted to generate an M59I substitution in the response regulator (ParR) of a recently described two-component regulatory system, ParRS (for peptide-adaptive resistance regulator and sensor) (9). The mutation was located in the receiver domain of this regulator, near the conserved phosphorylation site D57. According to the Pseudomonas genome database (http://v2.pseudomonas.com), the parS gene (PA1798), which codes for the sensor kinase ParS, is assumed to be cotranscribed with parR (PA1799). The ParRS system has been reported to be required for the activation of the lipopolysaccharide (LPS) modification operon arnBCADTEF-ugd in the presence of subinhibitory concentrations of various bacterial (polymyxin B and colistin) and eukaryotic (indolicidin) polycationic peptides and, consequently, to be responsible for adaptive resistance to these peptides (9).

To ascertain the role of mutated ParRS in the multidrug resistance phenotype of PAOW2, we constructed parS and parRS deletion mutants of PAOW2 (CM097 and CM098, respectively) and PAO1Bes (CM095 and CM096, respectively). The MICs of MexXY/OprM substrates (aminoglycosides, ciprofloxacin, and cefepime) and of carbapenems were restored to wild-type levels for PAOW2 upon inactivation of parS or parRS, while they remained unchanged for PAO1Bes (Table 3). Notably, these deletions decreased the colistin MIC 2-fold for PAO1Bes and 4-fold for PAOW2, thus suppressing the difference in resistance between the two strains. To confirm these data, we carried out complementation experiments by inserting a single copy of the mutated parRS operon from PAOW2 into the chromosome of PAO1Bes ΔparRS (CM096). The observation that the complemented mutant (CM106) displayed a resistance pattern similar to that of PAOW2 unambiguously established that multidrug resistance may arise in P. aeruginosa as a result of mutations in parR. Upon complementation, a slight but reproducible difference between the colistin resistances of CM096 and CM106 could be observed by the disk method (inhibition zones of 22 and 20 mm, respectively, for a disk load of 50 μg) but failed to produce a significant (2-fold) difference in the MIC by the microdilution method.

Since the genetic background of strains may potentially influence the phenotypic effects of mutations (22), we selected new agrW2 mutants on amikacin, using a wild-type clinical strain (CMC) isolated in our hospital. Sequencing of the parRS operon from imipenem-resistant clones allowed the identification of a single-step mutant, dubbed CMCW2, harboring a single T-to-A substitution at position 41 of the parS gene (L14Q; located in the first transmembrane domain). As for PAOW2, transfer of mutated parRS from CMCW2 into the chromosome of CM096 (resulting in mutant CM107) generated a multidrug resistance phenotype due to up- and downregulation of the mexY and oprD genes, respectively (Table 3; Fig. 2). Taken together, these data indicated that amino acid changes in ParS could also constitutively coactivate several resistance mechanisms.

Transcriptomic data.

In order to delineate the set of genes directly or indirectly responding to mutated ParRS, we compared the whole-genome expression profiles of PAOW2 and PAO1Bes, as determined with Affymetrix microarrays. To our surprise, only a very small number of open reading frames (ORFs) (8 in total) were differentially expressed (≥2-fold) in the two strains, in addition to mexY, arnBCADTEF-ugd, and oprD (Table 4). Some of the ParRS-regulated genes belong to operons (PA1559-PA1560, PA4773-PA4775-pmrAB, and arnBCADTEF-ugd) that are known to be regulated by the two-component signal transduction system PmrAB (33). With the exception of the arnBCADTEF-ugd locus, no overlap was found between the genes responding to mutated ParRS and those controlled by PhoP-PhoQ, another regulatory system which, like PmrAB, is involved in the adaptive response of P. aeruginosa to Mg²⁺ starvation (33). To validate our transcriptomic data, we measured the mRNA levels of selected genes by RT-qPCR. As indicated in Table 5, gene expression changes were concordant, though sometimes more pronounced (up to 15.9 times), for RT-qPCR compared to Affymetrix chips. For example, the pmrB gene (PA4777), which together with pmrA (PA4776) is cotranscribed with PA4773 to PA4775, was found to be upregulated in PAOW2 by RT-qPCR but not with the DNA chips. Thus, it is possible that some genes of the ParRS regulon may have been missed by the transcriptomic approach, despite an excellent (>98%) interassay reproducibility.

TABLE 4.

Microarray analysis of differentially expressed genes (at least 2-fold) in the PAOW2 mutant compared with PAO1Bes

Gene ID	Name (alias)	Fold change	Description
PA0958	oprD	−7.5	OprD porin
PA1559		28.9	Hypothetical protein
PA1560		14.3	Hypothetical protein
PA1797		45.7	Hypothetical protein
PA2018	mexY	4.6	RND transporter MexY
PA2358		25.5	Hypothetical protein
PA2655		33.6	Hypothetical protein
PA3552	arnB (pmrH)	11.0	Hypothetical protein
PA3553	arnC (pmrF)	11.6	Probable glycosyl transferase
PA3554	arnA (pmrI)	15.3	Hypothetical protein
PA3556	arnT (pmrK)	8.5	4-Amino-4-deoxy-l-arabinose transferase
PA3557	arnE (pmrL)	3.1	Hypothetical protein
PA3558	arnF (pmrM)	9.2	Hypothetical protein
PA3559	ugd (pmrE)	16.1	Probable UDP-glucose dehydrogenase
PA4773		36.8	Hypothetical protein
PA4774		12.5	Hypothetical protein
PA4775		7.1	Hypothetical protein

TABLE 5.

Influence of ParRS on expression of selected genes in strain PAO1Bes and mutant strain PAOW2

Gene	Fold change in gene expressiona
	PAOW2/PAO1Bes	CM096/PAO1Bes
PA0958 (oprD)	−7.1 ± 0.1	1.1 ± 0.3
PA1559	50.1 ± 14	1.6 ± 0.3
PA1797	315.5 ± 83	1.1 ± 0.5
PA1799 (parR)	−1.2 ± 0.4
PA2018 (mexY)	13.1 ± 2.8	−1.1 ± 0.1
PA2358	27.8 ± 5.6	1.9 ± 0.1
PA2655	63.7 ± 14.1	1.2 ± 0.3
PA3554 (arnA)	50.6 ± 8.2	2.3 ± 0.4
PA4773	90.6 ± 15.7	1.6 ± 0.2
PA4774	131.3 ± 25.5	1.5 ± 0.5
PA4777 (pmrB)	15.9 ± 2.6	1.4 ± 0.2

^aValues are means of three independent determinations ± SD.

PmrAB-independent activation of LPS modification.

The signal transduction system PmrAB is able to promote bacterial resistance to polymyxins and cationic peptides in response to Mg²⁺ starvation through modification of LPS molecules. These adaptive changes, which include neutralization of negatively charged phosphate residues of lipid A by addition of 4-aminoarabinose, limit the penetration of polycations across the bacterial outer membrane (33, 34, 37). Because of this, we wondered whether PmrAB, whose expression is increased in PAOW2, would account for the lower susceptibility (2-fold) to colistin of the mutant than of PAO1Bes. Actually, elimination of the pmrAB genes had similar effects in PAOW2 (mutant strain CM100) and PAO1Bes (mutant strain CM099), leading to a modest 2-fold decrease in the colistin MIC (Table 3). At the transcriptional level, suppression of the pmrAB locus did not significantly influence expression of the arnA gene from the LPS modification operon arnBCADTEF-ugd for both strains (50.6-fold ± 8.2-fold for PAOW2 versus 40.5-fold ± 7.3-fold for CM100 and 1-fold for PAO1Bes versus 1.9-fold ± 0.25-fold for CM099). Contrasting with these results, deletion of the parRS locus drastically reduced arnA levels in PAOW2 (−1.1-fold ± 0.15-fold for mutant strain CM098) but not in PAO1Bes (2.3-fold ± 0.4-fold for mutant strain CM096). Therefore, in addition to the recent demonstration of the role of ParRS in adaptive resistance to bacterial and eukaryotic polycationic peptides (9), we found that the ParRS system may provide P. aeruginosa with constitutive resistance to these agents, through upregulation of the arnBCADTEF-ugd operon, when activated by mutations affecting either the response regulator ParR or the sensor kinase ParS.

ParRS mediates multidrug resistance in P. aeruginosa.

Since the oprD and mexY genes are constitutively down- and upregulated in the PAOW2 mutant, respectively, we considered the possibility that polycationic peptides might induce increased resistance of wild-type P. aeruginosa strains to carbapenems, aminoglycosides, cefepime, and fluoroquinolones through their activation of ParRS. This hypothesis was tested first by measuring the transcriptional levels of the mexY and oprD genes in bacteria cultured for 4 h in the presence of a concentration equivalent to a 1:2 ratio of the MIC of colistin (0.5 μg/ml), indolicidin (4 μg/ml), polymyxin B (0.5 μg/ml), or polymyxin B nonapeptide (PMBN) (64 μg/ml) (Fig. 3A and B). The latter compound, which lacks the fatty acid tail of polymyxin B, has lost most of the antibacterial properties of its parent molecule on P. aeruginosa but retains a strong outer membrane-permeabilizing activity (40, 48, 49). Because the PA5471 gene is required for drug-induced expression of mexXY in wild-type bacteria (36), the induction experiments were carried out with mutants lacking either PA5471 (CM093) or parRS (CM096), in addition to PAO1Bes. Exposure of PAO1Bes to colistin, the most therapeutically relevant cationic peptide from this series, was associated with substantial changes in mexY and oprD expression compared with that in untreated cells (4.6-fold ± 0.6-fold and −2.5-fold ± 0.35-fold, respectively). Suppression of the PA5471 gene (mutant strain CM093) had no influence on these transcript levels (4.2-fold ± 0.7-fold and −2.6-fold ± 0.2-fold, respectively), indicating a PA5471-independent regulation pathway for mexXY by colistin (Fig. 3A and B). On the other hand, the colistin-induced responses of mexY and oprD were completely abolished in the parRS null mutant CM096 (−1.1-fold ± 0.15-fold and 1.29-fold ± 0.1-fold changes in expression, respectively) (Fig. 3A and B and 4). Similar results were obtained with indolicidin and polymyxin B. Notably, despite the use of a fairly high concentration (64 μg/ml), PMBN increased mexY activity only marginally (Fig. 3 and 4).

FIG. 3.

Influence of antibiotic stress on mexY and oprD expression. Changes in expression levels of the mexY (A) and oprD (B) genes were assessed by RT-qPCR for wild-type strain PAO1Bes (black bars), ΔPA5471 mutant strain CM093 (white bars), and ΔparRS mutant strain CM096 (gray bars), which were preincubated for 4 h in the presence of indolicidin, colistin, polymyxin B, PMBN, or gentamicin. An uninduced culture was performed in parallel as a negative control. Values are means of three independent determinations ± SD.

FIG. 4.

Interactions of polymyxin B and PMBN with gentamicin. Antibiograms were created for strain PAO1Bes (A) and ΔparRS mutant strain CM096 (B) on MHA according to CLSI recommendations. PB, polymyxin B; PMBN, polymyxin B nonapeptide. As shown in the upper part of panel A, polymyxin B antagonizes the bacteriostatic activity of gentamicin (GM) on PAO1Bes.

The possibility that gentamicin might trigger mexXY expression via ParRS was also examined by culturing PAO1Bes, CM093, and CM096 in the presence of a concentration of the antibiotic equivalent to 0.5× MIC (1 μg/ml, 0.06 μg/ml, and 1 μg/ml, respectively). Confirming previous results from our laboratory (19), sub-MIC gentamicin markedly induced mexY gene expression (18.5-fold ± 1.7-fold) in wild-type PAO1Bes compared with untreated cells. However, this adaptive response was independent of parRS, as the mexY levels of CM096 (13.2-fold ± 2-fold) were not really different from those of PAO1Bes. In agreement with this, using a pPA3552::lux fusion, Fernandez et al. reported no effect of aminoglycosides on ParRS-dependent activation of the LPS modification operon arnBCADTEF-ugd (9). Thus, ParRS is unlikely to respond to the presence of aminoglycosides in the external medium. Similarly, imipenem (2 μg/ml), meropenem (0.5 μg/ml), and cefepime (2 μg/ml) did not show any ParRS-dependent effects on mexY and oprD expression (data not presented).

Role of ParRS in antagonistic drug interactions.

Collectively, the results mentioned above suggested that subinhibitory concentrations of polycationic peptides might promote antagonistic interactions between antipseudomonal antibiotics through distinct mechanisms, such as outer membrane impermeability and active drug efflux (Fig. 4). To demonstrate this, checkerboard assays were performed with wild-type PAO1Bes and its ΔparRS mutant CM096 grown in MHB. Consistent with our gene expression experiments, colistin at concentrations near the MIC strongly antagonized the bacteriostatic activities of gentamicin, meropenem, cefepime, and ciprofloxacin in PAO1Bes (fractional inhibitory concentration [FIC] indexes of 2.5, 4.5, 4.5, and 4.5, respectively) but not in CM096 (FIC indexes of 1, 1, 1.5, and 1, respectively). The FIC index reflects synergism (value of ≤0.5), additivity (value of 0.51 to 0.99), indifference (value of 1 to 2), or antagonism (value of >2) between the bacteriostatic effects of two drugs.

Such an antagonistic effect of colistin toward aminoglycosides is of considerable clinical importance for cystic fibrosis. Patients suffering from this inherited genetic disease are indeed treated over long periods with aerosols of colistin and aminoglycosides, often given sequentially, when they are colonized by P. aeruginosa in the lungs (10). These patients are also used to receiving repeated courses of intravenous aminoglycosides during infectious exacerbations. To better evaluate the negative impact of colistin on aminoglycoside efficacy, we carried out time-kill experiments with gentamicin at 16 μg/ml (a concentration 8-fold higher than the MIC), using PAO1Bes and CM096 preincubated for 4 h with or without 0.5 μg/ml of colistin (equivalent to a 1:2 ratio of the MIC). Under noninducing conditions, PAO1Bes and its mutant CM096 exhibited similar killing profiles, with 3 orders of reduction in survivors after 30 min of gentamicin exposure (Fig. 5). As anticipated from our checkerboard assays, preincubation with colistin drastically reduced killing of PAO1Bes (−1.15-fold ± 0.2 log₁₀-fold), confirming the development of a refractory state in the bacteria. Of greater importance, such a colistin-induced adaptation was not observed in mutant strain CM096, which exhibited a decline in living cells averaging −3.6 ± 0.2 log₁₀ after 30 min of gentamicin treatment. Similar results were obtained with amikacin and tobramycin at concentrations 8-fold higher than the MIC. If confirmed in vivo by clinical trials, these data would preclude the use of combined therapy with colistin and aminoglycosides in the management of CF lung infection.

FIG. 5.

ParRS is required for colistin-induced resistance to the killing activity of gentamicin. Survivors of wild-type strain PAO1Bes (triangles) and its ΔparRS derivative CM096 (circles) in the presence of 16 μg/ml gentamicin were determined in triplicate after 4 h of preincubation in drug-free MHB (open symbols) or MHB supplemented with a concentration equivalent to a 1:2 ratio of the MIC of colistin.

Clinical strains with alterations in ParRS.

To gain insight into the clinical relevance of parRS mutants, we screened our laboratory collection in search of CF and non-CF isolates showing a typical AgrW2 resistance phenotype. One CF (3020R) and two non-CF (4922 and 5024) isolates over 6 strains were found to fit this phenotypic profile. Subsequent determination of the colistin MIC demonstrated a modest but reproducible resistance to this agent in the three isolates (MIC of 2 μg/ml). Compared to reference strain PAO1Bes, all of the bacteria appeared to overexpress mexY (10.6-fold ± 1.6-fold, 4.7-fold ± 0.5-fold, and 7-fold ± 0.7-fold, respectively) and to downregulate oprD (−5.2-fold ± 0.5-fold, −6.4-fold ± 0.8-fold, and −5.7-fold ± 0.6-fold, respectively), as assayed by RT-qPCR. Interestingly, sequence analysis of the parRS loci from these strains showed nonsynonymous nucleotide substitutions when the sequences were aligned with the PAO1 reference genome (http://v2.pseudomonas.com). These changes were predicted to lead to single amino acid substitutions in the sensor kinase ParS (V101M in isolate 4922 and L137P in isolate 5024) or the response regulator ParR (E156K in isolate 3020R). In support of a role of these alterations in the constitutive activation of ParRS, we noted that the gene flanking the parRS operon, PA1797, was overexpressed in the selected strains (108.2-fold ± 23.3-fold in 3020R, 33.6-fold ± 7.5-fold in 4922, and 4.6-fold ± 0.3-fold in 5024), as observed in mutant strain PAOW2.

Conclusions.

Among the numerous two-component regulatory systems (64 response regulators, 63 classical histidine kinases, and 16 atypical kinases) possessed by P. aeruginosa (41), only a few have been reported to provide significant antibiotic resistance in clinical strains (e.g., PmrB and PhoQ) (3, 43). To our knowledge, this study is the first showing the role played by a two-component signal transduction system (ParRS) in bacterial adaptation to no fewer than four different classes of antibiotics (polymyxins, aminoglycosides, fluoroquinolones, and β-lactams) through three distinct resistance mechanisms (i.e., LPS modification, increased drug efflux, and a reduced porin pathway). Such a multiresistance phenotype may be induced by bacterial exposure to polycationic peptides or may be inherited stably as a result of mutations affecting the sensor kinase ParS or the response regulator ParR. Adding to the recent observation that ParRS mediates the adaptive response of P. aeruginosa to polycations such as polymyxin B, colistin, and indolicidin at Mg²⁺ concentrations similar to those found in the human body (1 to 2 mM) (9), this work demonstrates that when it is activated, ParRS also leads to up- and downregulation of the efflux system MexXY/OprM and the porin OprD, respectively. Some of our data strongly suggest that the transcriptional regulator ParR is activated rather than inhibited by ParS once a specific signal is detected in the bacterial environment. Indeed, deletion of parS or parRS did not affect antibiotic resistance in wild-type bacteria, except for inducers such as colistin. Overexpression of plasmid-borne wild-type parRS genes did not result in increased resistance either (data not shown). The nature of the signal sensed by the regulatory system remains to be identified. However, this signal is unlikely to be related to disorganization of the outer membrane architecture, since (i) some polycationic peptides (e.g., CP-28 and LL-37) were unable to activate ParRS-dependent expression of the arnBCADTEF-ugd operon (9) and (ii) PMBN proved to be a much weaker inducer of mexY expression than polymyxin B, despite their having similar outer membrane-permeabilizing capabilities (Fig. 5). Undoubtedly, identification of the molecular signal perceived by ParS is a prerequisite for the development of innovative molecules able to block the adaptive resistance of P. aeruginosa to polycations.

The molecular basis of ParRS-dependent regulation of MexXY/OprM and OprD is another crucial issue that should be addressed in order to understand and try to prevent the adaptation of P. aeruginosa to antibiotic stress. A search for motif elicitation in the promoter regions upstream of the genes responding to ParRS activation in PAOW2 by use of MEME software (2) identified a predicted region of 40 nucleotides (P < 10⁻¹⁰) upstream of mexY, oprD, PA1797, PA2358, and PA2655. Thus, it is tempting to assume that MexXY/OprM and OprD act in concert to limit (i.e., by active efflux and by outer membrane impermeability) the intracellular accumulation of a toxic substrate that directly or indirectly results from the antibacterial action of polycationic peptides. Since genes coding for putative homologs of polyamine biosynthetic/degradative enzymes (e.g., PA4773 and PA4774) are overexpressed in PAOW2 or under polycation exposure, we looked at whether their inactivation would normalize the expression of mexY and oprD to baseline levels. The deletion of the PA4773 gene, which is predicted to encode an adenosylmethionine decarboxylase, failed to produce the expected results (data not shown). In the same line, plasmid-mediated overexpression or deletion of the PA2358 and PA2655 genes, whose functions are virtually unknown, had no influence on mexY or oprD expression in PAO1Bes (not presented). A search for a toxic metabolite recognized by both MexXY/OprM and OprD is currently under way in our laboratory.

Finally, this work demonstrates the existence of at least two novel types of agrW mutants in addition to those exhibiting alterations in the ribosomal machinery (8, 19). A preliminary study conducted on 94 non-CF, MexXY-overproducing clinical strains showed that 21 (22%) were agrW mutants and 73 (78%) were of the agrZ type. The characterization of these mutants is in progress. The involvement of some of the agrW isolates in severe infections (e.g., bacteremia) suggests that they are fully virulent. From a physiological perspective, the active efflux system MexXY/OprM now appears to be a key element of bacterial adaptation to antibiotics targeting the ribosome (aminoglycosides, tetracyclines, macrolides, and chloramphenicol) or the cellular envelope (colistin and polymyxin B). Figure 6 summarizes our current knowledge of the complex regulation of MexXY.

FIG. 6.

Schematic representation of regulation of the mexXY operon in P. aeruginosa. (A) Wild-type strain grown in drug-free medium. During bacterial growth, a translation attenuation mechanism involving a small peptide, PA5471.1, prevents transcription of PA5471, a downstream gene of unknown function whose product directly or indirectly modulates the activity of MexZ, the local repressor of the mexXY operon. (B) Growth of wild-type strain in the presence of MexXY-inducing antibiotics. Following bacterial exposure to ribosome-targeting antimicrobials, MexXY is produced at high levels conferring intrinsic resistance to the inducer, provided that it is a good substrate for the efflux pump (as is the case for aminoglycosides but not for chloramphenicol). This MexZ-dependent induction process is linked in part to increased levels of the PA5471 protein modulating MexZ repressor activity. A second pathway leading to drug-dependent overexpression of mexXY is independent of both the MexZ and PA5471 proteins. Our results demonstrate that polycationic antibiotics (colistin and polymyxin B) are able to promote expression of the pmrAB, arnBCDTEF-ugd, and mexXY operons and to coordinately downregulate the oprD gene through the activation of ParRS. This regulation results in multidrug resistance.